|
1. |
Self-assembled Monolayers for Biosensors |
|
Analyst,
Volume 122,
Issue 4,
1997,
Page 43-50
Th. Wink,
Preview
|
|
摘要:
Tutorial Review Self-assembled Monolayers for Biosensors Th. Wink*, S. J. van Zuilen, A. Bult and W. P. van Bennekom Department of Pharmaceutical Analysis, Faculty of Pharmacy, Utrecht University, P.O. Box 80082, 3508 TB Utrecht, The Netherlands The use of self-assembled monolayers (SAMs) in various fields of research is rapidly growing. In particular, many biomedical fields apply SAMs as an interface-layer between a metal surface and a solution or vapour. This review summarises methods for the formation of SAMs upon the most commonly used materials and techniques used for monolayer characterisation.Emphasis will lie on uniform, mixed and functionalised monolayers applied for immobilisation of biological components including (oligo-)nucleotides, proteins, antibodies and receptors as well as polymers. The application of SAMs in today’s research, together with some applications will be discussed. Keywords: Self-assembled monolayer; immunoassay; electrochemistry; surface plasmon resonance; (electrochemical) quartz crystal microbalance; biosensor High selectivity provided by biomolecules (antibodies, enzymes, nucleic acids) or biological systems (receptors, whole cells) is exploited in biosensors, in which a biological sensing element is integrated with an electrochemical, optical or piezoelectric transducer.1,2 Most commonly, the biological component (capture molecule) is immobilised on, or in close proximity to, the surface of the transducer.As a consequence, immobilisation strategies for biomolecules are of paramount importance in order to preserve their biological activity. At biological recognition, i.e., when the immobilised molecule and its ligand (the analyte) have interaction, a signal is generated which is (proportionally) related to the concentration or amount of the analyte in the sample. Direct immobilisation of biomolecules involves physisorption and is applied, e.g., in microtiterplate immunoassays, in which usually no control over the orientation of the molecule is achieved.3 Interaction sensors based upon surface plasmon resonance (SPR) or monomode dielectric waveguides,4 as well as microtiterplate immunoassays,3 require more sophisticated approaches to avoid random orientation.Protein A or G5 as well as biotin-avidin chemistry6–11 can improve control over the immobilisation process. Other strategies exploit hydrogels12 or poly-l-lysine5 with specific linkage reagents (e.g., carbodiimide, 13,14 or periodate15).Photochemical16 activation and, most of all, functionalised self-assembled monolayers (SAMs)17–20 offer promising possibilities. Almost any surface can be equipped with functionalised monolayers which possesses a required specific electrical, optical or chemical property. An ideal monolayer is depicted as perfectly aligned, closely packed alkane chains, attached to a smooth surface (Fig. 1). There are two methods to deposit molecular layers on solid substrates (e.g., glass or metal): Langmuir–Blodgett transfer21 and self assembly.The Langmuir –Blodgett technology21 will not be discussed in this review. The formation of stable monolayers on glass or aluminium oxide by alkanesilanes22 is also beyond the scope of this review. Emphasis, however, lies on another class of monolayers, based on the strong adsorption of disulfides (R–S– S–R), sulfides (R–S–R) and thiols (R–SH) on a metal (particularly gold) surface.Nuzzo and Allara, pioneers in the assembly of sulfur-containing molecules,23 noticed that dialkane sulfides form highly ordered monolayers on metal surfaces. Sulfur donor atoms coordinate strongly on a gold substrate. Van der Waals forces between methylene groups orient and stabilise the monolayer. Porter et al.24 showed that long-chain (number of methylene groups n > 10) alkanethiols assemble in a crystalline-like way. A reduction in chain length leads to less ordered structures.Adsorption studies of unsymmetrical dialkane sulfides25 revealed that these monolayers are significantly less densely Thijs Wink has been a PhD researcher at Utrecht University, Faculty of Pharmacy, Department of Pharmaceutical Analysis, since 1993, after graduating in Pharmacy. Dr. Wout van Bennekom and Professor Dr. Auke Bult are, respectively, his co-supervisor and supervisor. Steven van Zuilen is a Pharmacy student. Thijs Wink is working on the development of a method to analyse interferons and interleukins using surface plasmon resonance at the low picomolar level.His interests include most optical analytical systems, as well as electrochemistry. Fig. 1 Schematic drawing of a pure and a mixed monolayer. Sulfur (black) is attached to a gold surface. Functional groups (shaded) are elevated into the solution or gaseous phase. Analyst, April 1997, Vol. 122 (43R–50R) 43Rpacked and less ordered, compared to alkanethiol monolayers.A series of w-substituted long-chain alkanethiols26 also form both highly oriented and ordered monolayers. Almost simultaneously a mixed monolayer, composed of different thiols, was introduced which offers a very promising basis for immobilising biomolecules via a functional group.27 The formation and characterisation of organic SAMs and their applications in sensor technology will be discussed in this review. Formation of Monolayers Disulfides, sulfides or thiols coordinate very strongly onto a variety of metals, e.g., gold, silver, platinum or copper.The structure of a self-assembled monolayer depends on the morphology of the metal. Au(111) is mostly applied for the formation of monolayers. Gold films adopt this crystallographic orientation predominantly when deposited upon polished glass, silicon or freshly cleaved mica. Gold is favoured, because it is reasonably inert. Thermal evaporation and deposition upon silicon wafers is the most convenient way to prepare gold substrates.To render a hydrophilic wafer surface, it is pretreated with ‘piranha’ solution27 (hot H2O2–H2SO4) and successively rinsed with deionised water and absolute alcohol. Another method is treatment with hot basic and acidic oxidising mixtures.28,29 An adlayer of chromium (50–150 Å)30 or titanium ( ~ 40 Å)31 promotes the adhesion of gold. Deposition with a rate of 2 Å s21 at a pressure of ~1028 Torr (1 Torr = 133.322 Pa) produces a polycrystalline surface, but extended Au(111) terraces are present.30 The thiols form the highest ordered and oriented monolayers upon these terraces.An uncontaminated gold surface is important but not essential; the high affinity of the thiol moiety for gold even displaces contaminants.25,32 Most commonly, as a precaution, possible contaminants are removed by ‘piranha’ solution and a hydrophilic gold surface will be obtained. Residual contaminants on bare gold can be removed by scanning the potential between +0.5 and +1.4 V versus SCE in dilute (0.5 m) sulfuric acid.Clean gold shows a characteristic anodic peak current near +1.1 V and a single cathodic peak near +0.9 V.33 Thiols, sulfides or disulfides can be directly adsorbed from appropriate high purity solvents (commonly ethanol for nonpolar or water for polar w-substituted alkanethiols). Alkanethiols from dilute solution form a densely packed monolayer in less than 1 h. The adsorption time seems to be independent of the chain length, but high concentrations lead to shorter adsorption times.A general protocol for self-assembly of monolayers is hard to give, because the preparation depends on the desired properties. The mole ratio of a mixture of thiols in solution results in the same ratio in the mixed SAM. The two components do not phase segregate into islands.34 This interesting feature can be exploited to immobilise biomolecules in such a manner that steric hindrance between these molecules and their binding partners is avoided.Although dense monolayers assemble quickly, well ordered monolayers can take days to form.35 For alkanethiols (n > 5) adsorption stops at the monolayer level, a stable multilayer is not formed. However, multilayer formation of an alkanethiol is observed after 6 d of immersion.36 The assembling kinetics of a monolayer is biphasic: the diffusion-controlled adsorption is followed by a slow (re-)crystallisation process.Characterisation of Monolayers Monolayers can be characterised by a wide variety of methods, of which the non-electrochemical methods are not extensively reviewed. Several chapters35,37 and a book38 discuss this topic. Of interest are the (a) pinholes or defect structures, (b) gold– sulfur bonding, (c) molecular orientation of the polymethylene chains and (d) order and orientation of the tail groups. (a) Pinholes or Defect Structures Monolayers may completely cover the metal surface.Microscopy techniques (scanning tunnelling39–41 and atomic force42) have revealed that normally ‘a very low concentration’ of pinholes is present. Electrochemistry can easily detect pinholes in a monolayer covering gold.33 A bare gold electrode in an acidic aqueous solution yields a well defined set of peaks, due to surface oxidation. Suppression of these faradaic currents implies that even water is excluded from a SAM-covered electrode. Residual oxidation can be caused by pinholes in the SAM or defect structures in the gold surface, so care should be taken when interpreting experimental results.Porter et al.24 have estimated the maximum pinhole radius as 8 mm and a fractional uncovered area of one per cent. or less.37 An extreme decrease in the faradaic current of a redox couple [Ru(NH3)6 3+ or Fe(CN)6 32] is indicative of a (hydrophobic) SAM. A small residual current is a strong indication of the presence of pinholes.A pinhole-free monolayer is impermeable to aqueous ions, and will act as an ideal capacitor.43 A SAMcovered electrode, compared to a bare one, demonstrates a strong decrease in capacity. Charging currents will be relatively larger if the monolayer is not impermeable. Differential capacitance measurements are less sensitive than voltammetry, but they additionally provide a manner in which to measure the film thickness and the permeability for simple ions. (b) Gold-Sulfur Bonding Voltammetric studies indicate that thiol groups are deprotonated upon adsorption.44 The assumed formation of a gold– thiolate bond is: RSH + Au " RS–Au + e2 + H+.A new route for fast electrodeposition of monolayers has been proposed by Weisshaar et al.45 In this study it is proven that a monolayer can be reversibly adsorbed and desorbed by electrochemical means. This process can serve as a basis for the determination, by capacitance measurements, of the monolayer surface coverage G,46 by measuring the charge, Q, needed to desorb an w-mercaptoalkane ferrocenecarboxylate monolayer. From G = Q/nFA, where n is the number of electrons involved in the electron-transfer process, F the Faraday constant and A the geometric electrode surface area, the value of G can be calculated.The values of Q are determined by integration of the area under the i–E curves (obtained in 1.0 m HClO4) after compensating for charging current and compared to those obtained from infrared reflection absorption spectroscopy.46 Monolayers are stable in the potential range from 2400 to +1400 mV versus SCE in dilute sulfuric acid solutions.33 Surface methods based on electron or photon irradiation should not destroy the gold–sulfur bond.47 Techniques used for the determination of the chemical composition of the monolayers, including infrared,23,48 X-ray photoelectron spectroscopy, 49 and near edge X-ray absorption fine structure measurements50 are reviewed by Finklea.37 (c) Molecular Orientation of the Polymethylene Chains Infrared data show that monolayers of w-substituted alkanethiols (n > 10) are densely packed crystalline-like structures, exhibiting a typical tilt angle in the range of 28–40° from the surface normal and a twist of chain axes of approximately 55°.31 The methylene groups exhibit strong Van der Waals inter- 44R Analyst, April 1997, Vol. 122actions, stabilising the monolayer. Helium diffraction and transmission electron microscopy studies51 reveal that methylene groups are ordered (crystalline like) at low temperatures and less ordered (semi-crystalline) at room temperature.(d) Order and Orientation of the Tail Groups These groups are important for the interaction of a monolayer with a biomolecule. Grazing incidence infrared spectroscopy31 shows that wsubstituted alkanethiols also are densely packed, highly oriented and ordered. As long as the end group (NH2, OH) is relatively small ( < 5Å),43 the orientation of the monolayer is not influenced.More bulky groups (COOH, ferrocene) decrease the density of packing and ordering.43,46 A mixed monolayer, consisting of a w-substituted and a (shorter) alkanethiol is adsorbed in the same mole fraction as in the solution. A phase separation is not observed, indicating random ordering of the two constituents.34 This feature offers the possibility to ‘dilute’ w-substituted alkanethiols with shorter non-substituted thiols in order to have anchor groups available for immobilisation procedures in which steric hindrance is possibly reduced (Fig. 2). Wetting studies by contact angle measurements between the monolayer surface and a liquid provide structural information.27 Comparing contact angles of polar and non-polar liquids can provide insight into the three dimensional structure of the monolayer. Contact angle measurements probe the outermost few angstroms of the surface.26 High contact angles of water and hexadecane on methyl-terminated thiol monolayers and the low contact angles on carboxylic acid- and alcohol-terminated monolayers indicate that the surface of the monolayers consist of densely packed arrays of the tail groups of the thiols.Applications SAM technology is most advantageous for electrochemical, SPR and (electrochemical) quartz crystal microbalance [(E)QCM] sensors. Modifying a sensor surface with SAMs generates model systems with a specific property or function.Some considerations to use SAMs in analytical applications include: (a) the (alkane)thiols form easy-to-manufacture, pinhole- free, stable monolayers from dilute solutions, ensuring a uniform immobilisation surface; (b) SAMs shield biological substances from the sensor surface, preventing possible denaturation; 52–56 (c) contamination of metal surfaces (non-specific adsorption altering the hydrophobic c.q. hydrophilic properties) impairs analysis and has to be avoided; and (d) the monolayer can be tailored with functional terminal groups for immobilisation purposes.This last feature offers numerous challenges, e.g., improvement of detection limits, the ability to regenerate the sensor, prevention of aspecific adsorption, as well as development of generic assays. In the following sections a number of applications of SAMs in electrochemical, optical (SPR) and (E)QCM sensors is presented. A representative selection of articles has been made in this very rapidly expanding research field.Electrochemical Sensors Within the potential limits as stated before, SAMs based on thiols or related compounds are not desorbed from electrode surfaces in aqueous solutions. SAMs can either be used for studying (non-electroactive) blocking properties, or for obtaining a selective electroactive surface. The blocking properties of a SAM have already been presented in the section dealing with the detection of pinholes.SAMs with long alkanethiols (n > 10) cause a dramatic decrease in electrode/electrolyte capacitance compared with bare gold.33 This involves a reduction in noise and a relative enhancement of the faradaic current (provided that this current is not significantly attenuated by the monolayer). On the other hand SAMs can decrease electron-transfer rates, which can be used for kinetic studies. Detailed theoretical aspects are presented by Finklea.37 An ‘electroactive electrode’ consists of a SAM with both blocking behaviour and selective electron tunnelling or ‘gates’ for the analyte.In the following section, several approaches for using SAMs in electrochemical sensors are presented. In an amperometric sensor a carboxylic acid-terminated SAM [HS(CH2)nCOOH with n = 15 or 11] is used to immobilise cytochrome c (a mediator in cell redox reactions) via carbodiimide activation.57 In contrast to electrostatically adsorbed cytochrome c, carbodiimide-mediated attached molecules could not be desorbed by a solution of saturated potassium nitrate. Also, carboxylic acid-terminated SAMs (n = 2, 5 or 10) are applied in detecting the neurotransmitter dopamine in the presence of ascorbic acid.58 At neutral pH, the negatively charged SAM repels ascorbic acid, while positively charged dopamine can be detected at sub-millimolar levels.An optimum in electrochemical discrimination between ascorbic acid and dopamine is achieved using the medium length thiol (n = 5).Electrochemically polymerised pyrrole (polypyrrole), a conducting polymer, is widely used for incorporating biomolecules. 59,60 Polypyrrole initially deposits as oligomers onto the electrode surface. Growth upon this nucleation sites continues in a fairly uncontrolled manner. Scanning tunnelling micrographs reveal a ‘cauliflower’-like structure. A novel approach is controlled growth on a SAM from a pyrrole-derivatised thiol.61,62 Pyrrole is electrochemically polymerised to form an adlayer.Control over the final deposition of the polymer film is provided by the number and location of the ordered nucleation sites. Adhesion of multilayers of polypyrrole on this adlayer is enhanced. Growth of the film proceeds layer by layer, as Fig. 2 Schematic representation of a SAM of thiols and the binding of streptavidin to them. (Top) A pure monolayer. Binding of streptavidin is severely sterically hindered. (Center) A monolayer of a mixture of thiols with the same length of the alkane moiety.There still is steric hindrance. (Bottom) Addition of a spacer allows binding without steric hindrance.83 (With kind permission from Elsevier Science Ltd.) Analyst, April 1997, Vol. 122 45RN N N Co N O O O O S S HS S gold film indicated by a less porous surface. The resulting film, in contrast to deposition of polypyrrole on a bare surface, is not easily removed. The thickness-corrected conductivity is enhanced by a factor of 3.Redox enzymes linked to SAMs have been studied by Creager and Olsen.63 A glucose sensor is prepared by crosslinking glucose oxidase (GOx) to a SAM from w-hydroxy alkanethiol by glutaric dialdehyde. Electrical communication between electrode and GOx is achieved via freely diffusing hydroxymethylferrocene. Normally, redox currents are diminished by a SAM, but the response of this redox mediator was less affected. Background currents of interferents (uric acid, ascorbic acid and 4-acetamidophenol) were dramatically reduced relative to untreated electrodes.Rikling and Willner64 coupled GOx enzyme, derivatised with ferrocene, to a monolayer of cystamine: [NH2(CH2)2S]2. Since one adlayer of GOx did not yield a detectable signal, seven consecutive adlayers were covalently attached, enhancing the signal substantially. A highly selective and thermodynamically favourable electron transfer pathway consists of a SAM from 3,3A-dithiobissulfosuccinimidylpropionate on which the enzyme GOx is immobilised.The active site of the enzyme is assumed to be in direct contact with the electrode.65 Another approach for a glucose sensor is presented by Rubin et al.66 They used a mixed monolayer of ferrocenylhexadecanethiol, as mediator for electron transfer, and aminoethanethiol, for immobilising GOx. A monolayer from 2-mercaptobenzothiazole can electrically be opened and closed for underpotential deposition of copper,67 without affecting the electron-transfer rate of an Fe2+/Fe3+ redox reaction.Cysteamine [NH2(CH2)2SH] was assembled and a quinone mediator (pyrroloquinolinquinone: a catalyst for electro-oxidation of NADPH and NADH) covalently linked.68 This yielded a pH-sensitive detector for nicotinamide cofactors. When to this quinone (NADP+-dependent) malic enzyme is covalently bound, maleic acid can be measured amperometrically in the range from 1027 to 1023 m. Porphyrins (catalysts for oxygen reduction) linked with thiols on the ‘edges’ of the molecule are assembled on an optical transparent electrode.69 The effect of the electrode surface on the mechanism of oxygen reduction by metalloporphyrin films could be studied (Fig. 3). Using a transparent electrode, UV/VIS absorption spectra can be obtained simultaneously for the adsorption process. Surprisingly, long adsorption times led to the formation of multilayers, but these could be removed by extensive rinsing. Thioctic acid, a cyclic disulfide containing a carboxylic acid, is used as a pH-dependent SAM70,71 for the redox response of Ru(NH3)6 3+ and Fe(CN)6 32.At increasingly higher pH values, the SAM becomes more negatively charged and the faradaic current for Fe(CN)6 32 is reduced. In this way it acts as a cationic sensor. At low pH values when carboxylic acid is neutral, both Fe(CN)6 32 and Ru(NH3)6 3+ show a fast response. Thioctic acid serves as a basis for a ‘novel separation-free sandwich-type enzyme immunoassay’ for proteins.72 On a goldcoated microporous nylon membrane a capture antibody against human chorionic gonadotrophin (hCG) is covalently immobilised.Simultaneously, hCG and an alkaline phosphatase labelled detecting antibody are incubated. Aminophenyl phosphate is enzymatically converted into aminophenol, which is detected electrochemically.73 The detection limit for hCG was determined to be 0.8 U l21 ( Å 160 ng l21). There is a growing demand for detection systems of specific DNA sequences.74 An electrochemical method for the determination of the amount of DNA adsorbed onto an electrode surface is presented by Pang et al.75 Single-stranded DNA is covalently attached to a monolayer of thioctic acid, activated by carbodiimide and N-hydroxysuccinimide.The amount of immobilised DNA is determined indirectly through the redox reaction of tris(2,2A-bipyridyl)cobalt(iii), which complexes strongly with DNA.76 Surface Plasmon Resonance When p-polarised laser light reflects internally at the interface of an optically dense and less dense medium, an evanescent field is generated.This field will, at a specific angle, be enhanced by collective motions (resonance) of conducting electrons (plasmons) in a thin ( Å 50 nm) metal layer.77 Silver and gold films are commonly used, because of a low imaginary dielectric constant. Changes in refractive index at the gold/air or gold/solution interface shifts the resonance angle.Interactions can be followed in real time, without any labelling of the reaction partners. Again, SAM technology has already proven its merits in this sensor type. A hydrogel78 can be linked with a SAM onto a gold surface. Electrostatic interactions preconcentrate positively charged proteins (at low pH, below their isoelectric point) at residual negative charges in the hydrogel. Activation with carbodiimide and N-hydroxysuccinimide covalently binds the proteins to the matrix.12 Other methods for (covalent) coupling chemistries for this carboxymethylated dextran hydrogel matrix are presented by L�ofås et al.79 A versatile and useful biotin-functionalised SAM has been presented by Knoll and coworkers.80,81 Biotin (vitamin H) has an extremely high binding constant (Kdiss = 10215 m) for tetravalent (strept)avidin.When (strept)avidin is bound to this monolayer, a biotinylated antibody (anti-hCG) is readily immobilised. The accessibility for (strept)avidin can be optimised by preparing a monolayer of biotinylated long-chain alkanethiols, diluted with shorter hydroxythiols.82 The longchain alkanethiol acts as a spacer molecule, elevating the biotin group from the monolayer surface.Cyclodextrins are cyclic oligosaccharides that chelate with small molecules. Derivatised with multiple long-chain thiol spacers, these cyclodextrin molecules assemble parallel to the surface. Oriented in this way they are freely accessible for dye molecules.83 Specific recognition with high sensitivity for cholera toxin is achieved by creating an artificial membrane with entrapped ganglioside.84 A lipid layer (membrane) can be immobilised through hydrophobic interaction upon a SAM of an alkanethioTo form a lipid membrane upon this SAM, various deposition methods were compared. A simple and fast way to deposit an artificial membrane has been achieved by a lipid/detergent dilution technique that finally yields a micellar solution.These Fig. 3 Schematic illustration of the most probable average surface structure of tetra-thiolated cobalt porphyrin.69 (With kind permission from the American Chemical Society.) 46R Analyst, April 1997, Vol. 122micelles spontaneously assemble onto the hydrophobic SAM as an artificial membrane. The detection limit for cholera toxin (M Å 50 kDa for the B5 subunit) can be estimated as about 1029 m with SPR. Thiol-derivatised synthetic peptides (Mr Å 2 kDa) form a sensing layer for specific protein molecules.85,86 An optimum response for an antibody (Mr Å 150 kDa) is found when the SAM consisted of only 3 mol % thiolated peptide.To study the effects of surface properties on protein adsorption,87 SAMs are excellent model systems. Hydrophilic hexa(ethylene glycol)-terminated SAMs effectively reduce non-specific adsorption for a wide range of proteins found in biological fluids. A similar type of SAM is used by Duschl et al.88 11-Mercaptoundecanol-linked antigenic peptide (mimicking the surface of a malaria parasite) is diluted with underivatised 11-mercaptoundecanol. Non-specific antibodies did not adsorb, but the specific antibody (1029 m) gave a small, but noticeable shift when incubated.A series of six histidine moieties (a His tag) is incorporated in the primary sequence of recombinant proteins, in order to simplify purification. Thiols, terminated with a nitrilotriacetic acid (NTA) group and a tri(ethylene glycol) group (mentioned before) in a molar ratio 1 : 10 form a mixed SAM.89 NTA coordinates with Ni2+, leaving two vacant binding sites, which selectively chelates with the His tag.Detection of DNA hybridisation with SPR is demonstrated by Piscevic et al.90 A ten-mer oligonucleotide, with the 5A-phosphate group attached to mercaptopropyl has been immobilised on gold. Surprisingly, formation of an ‘incomplete’ SAM is preferred over dilution.The self-assembly process was stopped in an early stage of the formation. In real-time, the hybridisation process is followed, Fig. 4 Schematic representation of the molecular structure and interaction of a self-assembled monolayer of resorcin[4]arenes on an Au(111) surface with perchloroethylene molecules from the gas phase.96 (With kind permission from the American Association for the Advancement of Science). Analyst, April 1997, Vol. 122 47Rproviding a promising basis for detection of DNA hybridisation reactions with SPR.Quartz Crystal Microbalance A QCM senses mass changes that result in a shift in resonance frequency. A piezoelectric crystal oscillates at a very sharp frequency. Small mass changes on the crystal surface shifts this frequency. In liquids, the performance of a QCM is in many cases inferior to that in air.91 Sensitivity and selectivity are comparable with SPR.92 If both sides of the crystal are coated with gold, these layers can be used as electrodes.Simultaneous measurements of electrochemical parameters and mass changes are an interesting feature of the EQCM.93 With a mass-sensitive instrument, organophosphonates (nerve agents) can be selectively and reversibly detected, as shown by Kepley et al.94 Copper(ii)-ions, bound to a SAM of carboxylate-terminated alkanethiols act as the sensing layer. EQCM has been used to study the kinetics of the formation, and the oxidative or reductive desorption of alkanethiol SAMs.94 It was concluded that better SAMs are formed from poorer solvents for the thiol.Choosing the appropriate conditions (solvent, formation time, thiol concentration) is of importance when well defined monolayers are required. Fundamental knowledge about the adsorption kinetics of alkanethiols can be obtained both from SPR and QCM measurements, because the process is monitored in realtime. 95 Very small mass changes of small gaseous molecules were measured with QCM using a resorcin[4]arene monolayer (Fig. 4) for detection.96 Perchloroethylene molecules are, according to the authors, recognised by these hydrophobic cavitants and could be measured in the nanogram range. However, other authors doubt their conclusions.97 A synthetic antigenic peptide [of the foot-and-mouth disease virus (FMDV)] has been derivatised with w-hydroxyundecanethiol (HUT) and was adsorbed onto gold.91 Various procedures to prepare sensing layers from (mixed) SAMs are compared and (contrarily) the best layer is the undiluted modified-peptide SAM.The common strategy to avoid steric hindrance, diluting the HUT-modified peptide in a mixed monolayer with pure HUT, yielded a lower response to anti- FMDV-antibody. Uniform Immobilisation Strategies In developing sensitive sensors, the self-assembly technique offers interesting perspectives. By generating functionalised surfaces through modification of thiols, it is possible to selectively attach the biomolecules of interest. Another option is to derivatise a biomolecule with a thiol functionality.For repetitive measurements (cost savings), the sensor surface ought to be regenerable. A sensor surface for phosphate biomolecules can be regenerated using a pH-dependent, electrostatically attached pentamidine layer upon a monolayer of mercaptoalkanoic acid.98 An N-hydroxysuccinimide ester functionalised SAM can be applied for a uniform covalent immobilisation of amino groupcontaining biomolecules.99 Photo-immobilisation for proteins via benzophenon derivatization can be another strategy (Fig. 5). Upon UV irradiation, a 10,10A-dithiobis(decanoic acid N-hydroxysuccinimide ester) derivatised with benzophenon crosslinks with an antibody. A homogeneous single layer of antibodies results, with retention (at least in part) of the activity.100 Polymer layers are widely applied in sensors and reviewed by Hars�anyi.101 The films can be deposited upon a substrate by (a) spinning or casting, (b) electrochemical polymerisation or (c) vacuum deposition.Polymerisation to a thin film, after being adsorbed as thiol monomers, was first described by Ford et al.102 An SAM of 4-(mercaptomethyl)styrene was polymerised in an aqueous solution of azo-initiator by irradiation with a laser, yielding a hydrophobic surface. General Conclusions Monolayers from thiol-containing molecules are easy to prepare, quickly assembled and well ordered.A general protocol for self-assembly conditions is hard to give; the preparation route depends on the desired properties. Electrochemistry, especially cyclic voltammetry, can detect pinholes or defect structures in the monolayers. SAMs are stable in a potential range from 2400 to +1400 mV versus SCE; beyond these potentials the layer desorbs, yielding a clean gold surface. Small w-functional groups exhibit no influence on the formation of the monolayer. Preparing a mixed SAM from a long-chain w-functionalised thiol ‘diluted’ with a shorter-chain alkanethiol offers great challenges for analytical purposes.In this way steric hindrance of the functionality can largely be reduced. The SAM technology is widely applied in electrochemical, SPR-based and (E)QCM-based sensors. Fig. 5 The immobilisation process used for patterning of proteins. a, Avidin with photobiotin immobilised onto the surface. b, Exposure of selected areas to light through a mask results in activation of the photobiotin molecule, specifically immobilising the antibody in the solution.c, Unbound material is removed by washing. d, The entire surface is exposed to light, and a blocking molecule bound to all unreacted photobiotin groups. e, Following washing the surface is exposed to fluorescently labeled antigen, which is bound by the patterned antibody.100 (With kind permission from Elsevier Science Ltd.) 48R Analyst, April 1997, Vol. 122References 1 Turner, A.F. P., in Biosensors: Fundamentals and Applications, ed. Turner, A. F. P., Karube, I., and Wilson, G. S., Oxford University Press, 1987, pp. V–VIII. 2 Byfield, M. P., and Abuknesha, R. A., Biosens. Bioelectron., 1994, 9, 373. 3 Lu, B., Smyth, M. R., a OAKennedy, R., Analyst, 1996, 121, 29R. 4 Goddard, N. J., Pollard-Knight, D., and Maule, C. H., Analyst, 1994, 119, 583. 5 Schramm, W., Paek, S.-H., and Voss, G., Immunomethods, 1993, 3, 93. 6 Ebersole, R.C., Miller, J. A., Moran, J. R., and Ward, M. D., J. Am. Chem. Soc., 1990, 112, 3239. 7 Morgan, H., Taylor, D. M., DASilva, C., and Fukushima, H., Thin Solid Films, 1992, 210/211, 773. 8 Achtnich, U. R., Tiefenauer, L. X., and Andres, R. Y., Biosens. Bioelectron., 1992, 7, 279. 9 Schmidt, A., Spinke, J., Bayerl, T., Sackmann, E., and Knoll, W., Biophys. J., 1992, 63, 1185. 10 Zhao, S., and Reichert, W. M., Langmuir, 1992, 8, 2785. 11 Hoshi, T., Anzai, J., and Osa, T., Anal. Chim.Acta, 1994, 289, 321. 12 F�agerstam, L. G., Frostell-Karlsson, A., Karlsson, R., Persson, B., and R�onnberg, I., J. Chromatogr., 1992, 597, 397. 13 Scouten, W. H., Methods Enzymol., 1987, 135, 30. 14 Carraway, K. L., and Koshland, Jr., D. E., Methods Enzymol., 1979, 25, 616. 15 Wolfe, C. A., and Hage, D. S., Anal. Biochem., 1995, 231, 123. 16 Matsuda, T., and Sugawara, T., Langmuir, 1995, 11, 2267. 17 Mandler, D., and Turyan, I., Electroanalysis, 1995, 8, 207. 18 Allara, D. L., Biosens.Bioelectron., 1995, 10, 771. 19 G�opel, W., and Heiduschka, P., Biosens. Bioelectron., 1995, 10, 853. 20 Xu, J., and Li, H.-L., J. Colloid Interface Sci., 1995, 176, 138. 21 Roberts, G. G., and Pitt, C. W., Langmuir-Blodgett Films, 1982, Elsevier, Amsterdam, 1983. 22 Sagiv, J., J. Am. Chem. Soc., 1980, 102, 2. 23 Nuzzo, R. G., and Allara, D. L., J. Am. Chem. Soc., 1983, 105, 4481. 24 Porter, M. D., Bright, T. B., Allara, D. L., and Chidsey, C. E. D., J. Am. Chem.Soc., 1987, 109, 3559. 25 Troughton, E. B., Bain, C. D., Whitesides, G. M., Nuzzo, R. G., Allara, D. L., and Porter, M. D., Langmuir, 1988, 4, 365. 26 Bain, C. D., Troughton, E. B., Tao, Y.-T., Evall, J., Whitesides, G. M., and Nuzzo, R. G., J. Am. Chem. Soc., 1989, 111, 321. 27 Bain, C. D., Evall, J., and Whitesides, G. M., J. Am. Chem. Soc., 1989, 111, 7155. 28 Bertilsson, L., and Liedberg, B., Langmuir, 1993, 9, 141. 29 J�onsson, U., Malmquist, M., Olofsson, G., and R�onnberg, I., Methods Enzymol., 1988, 137, 381. 30 Nuzzo, R. G., Fusco, F. A., and Allara, D. L., J. Am. Chem. Soc., 1987, 109, 2358. 31 Nuzzo, R. G., Dubois, L. H., and Allara, D. L., J. Am. Chem. Soc., 1990, 112, 558. 32 Keller, H., Schrepp, W., and Fuchs, H., Thin Solid Films, 1992, 210/211, 799. 33 Finklea, H. O., Avery, S., Lynch, M., and Furtsch, T., Langmuir, 1987, 3, 409. 34 Bain, C. D., and Whitesides, G. M., J. Am. Chem. Soc., 1989, 111, 7164. 35 Dubois, L. H., and Nuzzo, R.G., Annu. Rev. Phys. Chem., 1992, 43, 437. 36 Kim, Y.-T., McCarley, R. L., and Bard, A. J., Langmuir, 1993, 9, 1941. 37 Finklea, H. O., in Electroanalytical Chemistry: A Series of Advances, ed. Bard, A. J., and Rubinstein, I., Marcel Dekker. New York, 1996, vol. 19, pp. 110–317. 38 Ulman, A., Characterization of Organic Thin Films, Butterworth– Heinemann, Boston, 1995. 39 Edinger, K., G�olzhaeuser, A., Demota, K., W�oll, C., and Grunze, M., Langmuir, 1993, 9, 4. 40 Sun, L., and Crooks, R.M., Langmuir, 1993, 9, 1951. 41 Delamarche, E., Michel, B., Kang, H., and Gerber, C., Langmuir, 1994, 10, 4103. 42 Pan, J., Tao, N., and Lindsay, S. M., Langmuir, 1993, 9, 1556. 43 Chidsey, C. E. D., and Loiacono, D. N., Langmuir, 1990, 6, 682. 44 Widrich, C. A., Chung, C., and Porter, M. D., J. Electroanal. Chem., 1991, 310, 335. 45 Weisshaar, D. E., Lamp, B. D., and Porter, M. D., J. Am. Chem. Soc., 1992, 114, 5860. 46 Walczak, I. M., Popenoe, D. D., Deinhammer, R.S., Lamp, B. D., Chung, C., and Porter, M. D., Langmuir, 1991, 7, 2687. 47 De Weldige, K., Rohwerder, M., Vago, H., Viefhaus, H., and Stratman, M., Fresenius’ J. Anal. Chem., 1995, 353, 329. 48 Truong, K. D., and Rowntree, P. A., Prog. Surf. Sci., 1995, 50, 207. 49 Nuzzo, R. G., Zegarski, B. R., and Dubois, L. H., J. Am. Chem. Soc., 1987, 109, 733. 50 Grunze, M., Phys. Scr., 1993, 49, 711. 51 Strong, L., and Whitesides, G. M., Langmuir, 1988, 4, 546. 52 Ivarsson, B. A., Hegg, P.-O., Lundstr�om, I., and J�onsson, U., Colloids.Surface., 1985, 13, 169. 53 Liedberg, B., Ivarsson, B., Lundstr�om, I., and Salaneck, W. R., Progr. Colloid Polym., 1985, 70, 67. 54 Lundstr�om, I., Progr. Colloid Polym. Sci., 1985, 70, 76. 55 Lundstr�om, I., and Salaneck, W. R., J. Colloid Interf. Sci., 1985, 108, 288. 56 Sundgren, J.-E., Bod�o, P., Ivarsson, B., and Lundstr�om, I., Progr. Colloid Polym. Sci., 1986, 113, 530. 57 Collison, M., Bowden, E. F., and Tarlov, M.J., Langmuir, 1992, 8, 1247. 58 Malem, F., and Mandler, D., Anal. Chem., 1993, 65, 37. 59 Van Os, P. J. H. J., Bult, A., and Van Bennekom, W. P., Anal. Chim. Acta, 1995, 305, 18. 60 Sadik, O. A., and Wallace, G. G., Anal. Chim. Acta, 1993, 279, 209. 61 Sayre, C. N., and Collard, D. M., Langmuir, 1995, 11, 302. 62 Willicut, R. J., and McCarley, R. L., Adv. Mater., 1995, 7, 759. 63 Creager, S. E., and Olsen, K. G., Anal. Chim. Acta, 1995, 307, 277. 64 Riklin, A., and Willner, I., Anal.Chem., 1995, 67, 4118. 65 Jiang, L., McNeil, C. J., and Cooper, J. M., J. Chem. Soc., Chem. Commun., 1995, 1293. 66 Rubin, S. R., Chow, J. T., Ferraris, J. P., and Zawodzinski, Jr., T. A., Langmuir, 1996, 12, 363. 67 Bharathi, S., Yegnaraman, V., and Rao, G. P., Langmuir, 1993, 9, 1614. 68 Willner, I., and Riklin, A., Anal. Chem., 1994, 66, 1535. 69 Postlethwaite, T. A., Hutchison, J. E., Hathcock, K. W., and Murray, R. W., Langmuir, 1995, 11, 4109. 70 Cheng, Q., and Brajter-Toth, A., Anal.Chem., 1992, 64, 1998. 71 Cheng, Q., and Brajter-Toth, A., Anal. Chem., 1995, 67, 2767. 72 Duan, C. M., and Meyerhoff, M. E., Anal. Chem., 1994, 66, 1369. 73 Heineman, W. R., and Halsall, H. B., Anal. Chem., 1985, 57, 1321A. 74 Downs, M. E. A., Kobayashi, S., and Karube, I., Anal. Lett., 1987, 20, 1897. 75 Pang, D.-W., Zhang, Z.-L., Qi, Y.-P., Cheng, J.-K., and Liu, Z.-Y., J. Electroanal. Chem., 1996, 403, 183. 76 Millan, K. M., and Mikkelsen, S. R., Anal.Chem., 1993, 65, 2317. 77 Kretschmann, E., and Raether, H., Z. Naturforsch., 1968, 23a, 2135. 78 L�ofås, S., and Johnsson, B., J. Chem. Soc., Chem. Commun., 1990, 1526. 79 L�ofås, S., Johnsson, B., Edstr�om, Å., Hansson, A., M�uller Hillgren, R.-M., and Stigh, L., Biosens. Bioelectron., 1995, 10, 813. 80 H�aussling, L., Ringsdorf, H., Schmitt, F.-J., and Knoll, W., Langmuir, 1991, 7, 1837. 81 Schmitt, F.-J., H�aussling, L., Ringsdorf, H., and Knoll, W., Thin Solid Films, 1992, 210/211, 815. 82 Spinke, J., Liley, M., Schmitt, F. J., Guder, H. J., Angermaier, L., and Knoll, W., J. Chem. Phys., 1993, 99, 7012. 83 Mittler-Neher, S., Spinke, J., Liley, M., Nelles, G., Weisser, M., Back, R., Wenz, G., and Knoll, W., Biosens. Bioelectron., 1995, 10, 903. 84 Terrettaz, S., Stora, T., Duschl, C., and Vogel, H., Langmuir, 1993, 9, 1361. 85 Van den Heuvel, D. J., Kooyman, R. P. H., Drijfhout, J. W., and Welling, G. W., Anal. Biochem., 1993, 215, 223. 86 Kooyman, R.P. H., Van den Heuvel, D. J., Drijfhout, J. W., and Welling, G. W., Thin Solid Films, 1994, 244, 913. Analyst, April 1997, Vol. 122 49R87 Mrksich, M., Sigal, G. B., and Whitesides, G. M., Langmuir, 1995, 11, 4383. 88 Duschl, C., S�evin-Landais, A.-F., and Vogel, H., Biophys. J., 1996, 70, 1985. 89 Sigal, G. B., Bamdad, C., Barberis, A., Strominger, J., and Whitesides, G. M., Anal. Chem., 1996, 68, 490. 90 Piscevic, D., Lawall, R., Veith, M., Liley, M., Okahata, Y., and Knoll, W., Appl.Surf. Sci., 1995, 90, 425. 91 Rickert, J., Weiss, T., Kraas, W., Jung, G., and G�opel, W., Biosens. Bioelectron., 1996, 11, 591. 92 Minunni, M., Mascini, M., Guilbault, G. G., and Hock, B., Anal. Lett., 1995, 28, 749. 93 Schneider, T. W., and Buttry, D. A., J. Am. Chem. Soc., 1993, 115, 12 391. 94 Kepley, L. J., Crooks, R. M., and Ricco, A. J., Anal. Chem., 1992, 64, 3191. 95 Karpovich, D. S., and Blanchard, G. J., Langmuir, 1994, 10, 3315. 96 Schierbaum, K.D., Weiss, T., Thoden van Velzen, E. U., Engbersen, J. F. J., Reinhoudt, D. N., and G�opel, W., Science, 1994, 265, 1413. 97 Grate, J. W., Patrash, S. J., Abraham, M. H., and Du, C. M., Anal. Chem., 1996, 68, 913.., and Unger, K., Anal. Chem., 1996, 68, 402. 99 Wagner, P., Hegner, M., Kernen, P., Zaugg, F., and Semenza, G., Biophys. J., 1996, 70, 2052. 100 Morgan, H., Pritchard, D. J., and Cooper, J. M., Biosens. Bioelectron., 1995, 10, 841. 101 Hars�anyi, G., Mater. Chem. Phys., 1996, 43, 199. 102 Ford, J. F., Vickers, T. J., Mann, C. K., and Schlenoff, J. B., Langmuir, 1996, 12, 1944. Paper 6/06964I Received October 11, 1996 Accepted December 26, 1996 50R Analyst, April 1997, Vol. 122 Tutorial Review Self-assembled Monolayers for Biosensors Th. Wink*, S. J. van Zuilen, A. Bult and W. P. van Bennekom Department of Pharmaceutical Analysis, Faculty of Pharmacy, Utrecht University, P.O. Box 80082, 3508 TB Utrecht, The Netherlands The use of self-assembled monolayers (SAMs) in various fields of research is rapidly growing.In particular, many biomedical fields apply SAMs as an interface-layer between a metal surface and a solution or vapour. This review summarises methods for the formation of SAMs upon the most commonly used materials and techniques used for monolayer characterisation. Emphasis will lie on uniform, mixed and functionalised monolayers applied for immobilisation of biological components including (oligo-)nucleotides, proteins, antibodies and receptors as well as polymers.The application of SAMs in today’s research, together with some applications will be discussed. Keywords: Self-assembled monolayer; immunoassay; electrochemistry; surface plasmon resonance; (electrochemical) quartz crystal microbalance; biosensor High selectivity provided by biomolecules (antibodies, enzymes, nucleic acids) or biological systems (receptors, whole cells) is exploited in biosensors, in which a biological sensing element is integrated with an electrochemical, optical or piezoelectric transducer.1,2 Most commonly, the biological component (capture molecule) is immobilised on, or in close proximity to, the surface of the transducer.As a consequence, immobilisation strategies for biomolecules are of paramount importance in order to preserve their biological activity. At biological recognition, i.e., when the immobilised molecule and its ligand (the analyte) have interaction, a signal is generated which is (proportionally) related to the concentration or amount of the analyte in the sample.Direct immobilisation of biomolecules involves physisorption and is applied, e.g., in microtiterplate immunoassays, in which usually no control over the orientation of the molecule is achieved.3 Interaction sensors based upon surface plasmon resonance (SPR) or monomode dielectric waveguides,4 as well as microtiterplate immunoassays,3 require more sophisticated approaches to avoid random orientation.Protein A or G5 as well as biotin-avidin chemistry6–11 can improve control over the immobilisation process. Other strategies exploit hydrogels12 or poly-l-lysine5 with specific linkage reagents (e.g., carbodiimide, 13,14 or periodate15). Photochemical16 activation and, most of all, functionalised self-assembled monolayers (SAMs)17–20 offer promising possibilities. Almost any surface can be equipped with functionalised monolayers which possesses a required specific electrical, optical or chemical property.An ideal monolayer is depicted as perfectly aligned, closely packed alkane chains, attached to a smooth surface (Fig. 1). There are two methods to deposit molecular layers on solid substrates (e.g., glass or metal): Langmuir–Blodgett transfer21 and self assembly. The Langmuir –Blodgett technology21 will not be discussed in this review. The formation of stable monolayers on glass or aluminium oxide by alkanesilanes22 is also beyond the scope of this review.Emphasis, however, lies on another class of monolayers, based on the strong adsorption of disulfides (R–S– S–R), sulfides (R–S–R) and thiols (R–SH) on a metal (particularly gold) surface. Nuzzo and Allara, pioneers in the assembly of sulfur-containing molecules,23 noticed that dialkane sulfides form highly ordered monolayers on metal surfaces. Sulfur donor atoms coordinate strongly on a gold substrate.Van der Waals forces between methylene groups orient and stabilise the monolayer. Porter et al.24 showed that long-chain (number of methylene groups n > 10) alkanethiols assemble in a crystalline-like way. A reduction in chain length leads to less ordered structures. Adsorption studies of unsymmetrical dialkane sulfides25 revealed that these monolayers are significantly less densely Thijs Wink has been a PhD researcher at Utrecht University, Faculty of Pharmacy, Department of Pharmaceutical Analysis, since 1993, after graduating in Pharmacy.Dr. Wout van Bennekom and Professor Dr. Auke Bult are, respectively, his co-supervisor and supervisor. Steven van Zuilen is a Pharmacy student. Thijs Wink is working on the development of a method to analyse interferons and interleukins using surface plasmon resonance at the low picomolar level. His interests include most optical analytical systems, as well as electrochemistry.Fig. 1 Schematic drawing of a pure and a mixed monolayer. Sulfur (black) is attached to a gold surface. Functional groups (shaded) are elevated into the solution or gaseous phase. Analyst, April 1997, Vol. 122 (43R–50R) 43Rpacked and less ordered, compared to alkanethiol monolayers. A series of w-substituted long-chain alkanethiols26 also form both highly oriented and ordered monolayers. Almost simultaneously a mixed monolayer, composed of different thiols, was introduced which offers a very promising basis for immobilising biomolecules via a functional group.27 The formation and characterisation of organic SAMs and their applications in sensor technology will be discussed in this review.Formation of Monolayers Disulfides, sulfides or thiols coordinate very strongly onto a variety of metals, e.g., gold, silver, platinum or copper. The structure of a self-assembled monolayer depends on the morphology of the metal. Au(111) is mostly applied for the formation of monolayers. Gold films adopt this crystallographic orientation predominantly when deposited upon polished glass, silicon or freshly cleaved mica.Gold is favoured, because it is reasonably inert. Thermal evaporation and deposition upon silicon wafers is the most convenient way to prepare gold substrates. To render a hydrophilic wafer surface, it is pretreated with ‘piranha’ solution27 (hot H2O2–H2SO4) and successively rinsed with deionised water and absolute alcohol. Another method is treatment with hot basic and acidic oxidising mixtures.28,29 An adlayer of chromium (50–150 Å)30 or titanium ( ~ 40 Å)31 promotes the adhesion of gold.Deposition with a rate of 2 Å s21 at a pressure of ~1028 Torr (1 Torr = 133.322 Pa) produces a polycrystalline surface, but extended Au(111) terraces are present.30 The thiols form the highest ordered and oriented monolayers upon these terraces. An uncontaminated gold surface is important but not essential; the high affinity of the thiol moiety for gold even displaces contaminants.25,32 Most commonly, as a precaution, possible contaminants are removed by ‘piranha’ solution and a hydrophilic gold surface will be obtained.Residual contaminants on bare gold can be removed by scanning the potential between +0.5 and +1.4 V versus SCE in dilute (0.5 m) sulfuric acid. Clean gold shows a characteristic anodic peak current near +1.1 V and a single cathodic peak near +0.9 V.33 Thiols, sulfides or disulfides can be directly adsorbed from appropriate high purity solvents (commonly ethanol for nonpolar or water for polar w-substituted alkanethiols).Alkanethiols from dilute solution form a densely packed monolayer in less than 1 h. The adsorption time seems to be independent of the chain length, but high concentrations lead to shorter adsorption times. A general protocol for self-assembly of monolayers is hard to give, because the preparation depends on the desired properties.The mole ratio of a mixture of thiols in solution results in the same ratio in the mixed SAM. The two components do not phase segregate into islands.34his interesting feature can be exploited to immobilise biomolecules in such a manner that steric hindrance between these molecules and their binding partners is avoided. Although dense monolayers assemble quickly, well ordered monolayers can take days to form.35 For alkanethiols (n > 5) adsorption stops at the monolayer level, a stable multilayer is not formed.However, multilayer formation of an alkanethiol is observed after 6 d of immersion.36 The assembling kinetics of a monolayer is biphasic: the diffusion-controlled adsorption is followed by a slow (re-)crystallisation process. Characterisation of Monolayers Monolayers can be characterised by a wide variety of methods, of which the non-electrochemical methods are not extensively reviewed. Several chapters35,37 and a book38 discuss this topic. Of interest are the (a) pinholes or defect structures, (b) gold– sulfur bonding, (c) molecular orientation of the polymethylene chains and (d) order and orientation of the tail groups.(a) Pinholes or Defect Structures Monolayers may completely cover the metal surface. Microscopy techniques (scanning tunnelling39–41 and atomic force42) have revealed that normally ‘a very low concentration’ of pinholes is present. Electrochemistry can easily detect pinholes in a monolayer covering gold.33 A bare gold electrode in an acidic aqueous solution yields a well defined set of peaks, due to surface oxidation.Suppression of these faradaic currents implies that even water is excluded from a SAM-covered electrode. Residual oxidation can be caused by pinholes in the SAM or defect structures in the gold surface, so care should be taken when interpreting experimental results. Porter et al.24 have estimated the maximum pinhole radius as 8 mm and a fractional uncovered area of one per cent.or less.37 An extreme decrease in the faradaic current of a redox couple [Ru(NH3)6 3+ or Fe(CN)6 32] is indicative of a (hydrophobic) SAM. A small residual current is a strong indication of the presence of pinholes. A pinhole-free monolayer is impermeable to aqueous ions, and will act as an ideal capacitor.43 A SAMcovered electrode, compared to a bare one, demonstrates a strong decrease in capacity. Charging currents will be relatively larger if the monolayer is not impermeable.Differential capacitance measurements are less sensitive than voltammetry, but they additionally provide a manner in which to measure the film thickness and the permeability for simple ions. (b) Gold-Sulfur Bonding Voltammetric studies indicate that thiol groups are deprotonated upon adsorption.44 The assumed formation of a gold– thiolate bond is: RSH + Au " RS–Au + e2 + H+. A new route for fast electrodeposition of monolayers has been proposed by Weisshaar et al.45 In this study it is proven that a monolayer can be reversibly adsorbed and desorbed by electrochemical means.This process can serve as a basis for the determination, by capacitance measurements, of the monolayer surface coverage G,46 by measuring the charge, Q, needed to desorb an w-mercaptoalkane ferrocenecarboxylate monolayer. From G = Q/nFA, where n is the number of electrons involved in the electron-transfer process, F the Faraday constant and A the geometric electrode surface area, the value of G can be calculated.The values of Q are determined by integration of the area under the i–E curves (obtained in 1.0 m HClO4) after compensating for charging current and compared to those obtained from infrared reflection absorption spectroscopy.46 Monolayers are stable in the potential range from 2400 to +1400 mV versus SCE in dilute sulfuric acid solutions.33 Surface methods based on electron or photon irradiation should not destroy the gold–sulfur bond.47 Techniques used for the determination of the chemical composition of the monolayers, including infrared,23,48 X-ray photoelectron spectroscopy, 49 and near edge X-ray absorption fine structure measurements50 are reviewed by Finklea.37 (c) Molecular Orientation of the Polymethylene Chains Infrared data show that monolayers of w-substituted alkanethiols (n > 10) are densely packed crystalline-like structures, exhibiting a typical tilt angle in the range of 28–40° from the surface normal and a twist of chain axes of approximately 55°.31 The methylene groups exhibit strong Van der Waals inter- 44R Analyst, April 1997, Vol. 122actions, stabilising the monolayer. Helium diffraction and transmission electron microscopy studies51 reveal that methylene groups are ordered (crystalline like) at low temperatures and less ordered (semi-crystalline) at room temperature. (d) Order and Orientation of the Tail Groups These groups are important for the interaction of a monolayer with a biomolecule.Grazing incidence infrared spectroscopy31 shows that wsubstituted alkanethiols also are densely packed, highly oriented and ordered. As long as the end group (NH2, OH) is relatively small ( < 5Å),43 the orientation of the monolayer is not influenced. More bulky groups (COOH, ferrocene) decrease the density of packing and ordering.43,46 A mixed monolayer, consisting of a w-substituted and a (shorter) alkanethiol is adsorbed in the same mole fraction as in the solution.A phase separation is not observed, indicating random ordering of the two constituents.34 This feature offers the possibility to ‘dilute’ w-substituted alkanethiols with shorter non-substituted thiols in order to have anchor groups available for immobilisation procedures in which steric hindrance is possibly reduced (Fig. 2). Wetting studies by contact angle measurements between the monolayer surface and a liquid provide structural information.27 Comparing contact angles of polar and non-polar liquids can provide insight into the three dimensional structure of the monolayer.Contact angle measurements probe the outermost few angstroms of the surface.26 High contact angles of water and hexadecane on methyl-terminated thiol monolayers and the low contact angles on carboxylic acid- and alcohol-terminated monolayers indicate that the surface of the monolayers consist of densely packed arrays of the tail groups of the thiols.Applications SAM technology is most advantageous for electrochemical, SPR and (electrochemical) quartz crystal microbalance [(E)QCM] sensors. Modifying a sensor surface with SAMs generates model systems with a specific property or function. Some considerations to use SAMs in analytical applications include: (a) the (alkane)thiols form easy-to-manufacture, pinhole- free, stable monolayers from dilute solutions, ensuring a uniform immobilisation surface; (b) SAMs shield biological substances from the sensor surface, preventing possible denaturation; 52–56 (c) contamination of metal surfaces (non-specific adsorption altering the hydrophobic c.q.hydrophilic properties) impairs analysis and has to be avoided; and (d) the monolayer can be tailored with functional terminal groups for immobilisation purposes. This last feature offers numerous challenges, e.g., improvement of detection limits, the ability to regenerate the sensor, prevention of aspecific adsorption, as well as development of generic assays.In the following sections a number of applications of SAMs in electrochemical, optical (SPR) and (E)QCM sensors is presented. A representative selection of articles has been made in this very rapidly expanding research field. Electrochemical Sensors Within the potential limits as stated before, SAMs based on thiols or related compounds are not desorbed from electrode surfaces in aqueous solutions.SAMs can either be used for studying (non-electroactive) blocking properties, or for obtaining a selective electroactive surface. The blocking properties of a SAM have already been presented in the section dealing with the detection of pinholes. SAMs with long alkanethiols (n > 10) cause a dramatic decrease in electrode/electrolyte capacitance compared with bare gold.33 This involves a reduction in noise and a relative enhancement of the faradaic current (provided that this current is not significantly attenuated by the monolayer).On the other hand SAMs can decrease electron-transfer rates, which can be used for kinetic studies. Detailed theoretical aspects are presented by Finklea.37 An ‘electroactive electrode’ consists of a SAM with both blocking behaviour and selective electron tunnelling or ‘gates’ for the analyte. In the following section, several approaches for using SAMs in electrochemical sensors are presented.In an amperometric sensor a carboxylic acid-terminated SAM [HS(CH2)nCOOH with n = 15 or 11] is used to immobilise cytochrome c (a mediator in cell redox reactions) via carbodiimide activation.57 In contrast to electrostatically adsorbed cytochrome c, carbodiimide-mediated attached molecules could not be desorbed by a solution of saturated potassium nitrate. Also, carboxylic acid-terminated SAMs (n = 2, 5 or 10) are applied in detecting the neurotransmitter dopamine in the presence of ascorbic acid.58 At neutral pH, the negatively charged SAM repels ascorbic acid, while positively charged dopamine can be detected at sub-millimolar levels.An optimum in electrochemical discrimination between ascorbic acid and dopamine is achieved using the medium length thiol (n = 5). Electrochemically polymerised pyrrole (polypyrrole), a conducting polymer, is widely used for incorporating biomolecules. 59,60 Polypyrrole initially deposits as oligomers onto the electrode surface. Growth upon this nucleation sites continues in a fairly uncontrolled manner. Scanning tunnelling micrographs reveal a ‘cauliflower’-like structure.A novel approach is controlled growth on a SAM from a pyrrole-derivatised thiol.61,62 Pyrrole is electrochemically polymerised to form an adlayer. Control over the final deposition of the polymer film is provided by the number and location of the ordered nucleation sites. Adhesion of multilayers of polypyrrole on this adlayer is enhanced.Growth of the film proceeds layer by layer, as Fig. 2 Schematic representation of a SAM of thiols and the binding of streptavidin to them. (Top) A pure monolayer. Binding of streptavidin is severely sterically hindered. (Center) A monolayer of a mixture of thiols with the same length of the alkane moiety. There still is steric hindrance. (Bottom) Addition of a spacer allows binding without steric hindrance.83 (With kind permission from Elsevier Science Ltd.) Analyst, April 1997, Vol. 122 45RN N N Co N O O O O S S HS S gold film indicated by a less porous surface. The resulting film, in contrast to deposition of polypyrrole on a bare surface, is not easily removed. The thickness-corrected conductivity is enhanced by a factor of 3. Redox enzymes linked to SAMs have been studied by Creager and Olsen.63 A glucose sensor is prepared by crosslinking glucose oxidase (GOx) to a SAM from w-hydroxy alkanethiol by glutaric dialdehyde.Electrical communication between electrode and GOx is achieved via freely diffusing hydroxymethylferrocene. Normally, redox currents are diminished by a SAM, but the response of this redox mediator was less affected. Background currents of interferents (uric acid, ascorbic acid and 4-acetamidophenol) were dramatically reduced relative to untreated electrodes. Rikling and Willner64 coupled GOx enzyme, derivatised with ferrocene, to a monolayer of cystamine: [NH2(CH2)2S]2.Since one adlayer of GOx did not yield a detectable signal, seven consecutive adlayers were covalently attached, enhancing the signal substantially. A highly selective and thermodynamically favourable electron transfer pathway consists of a SAM from 3,3A-dithiobissulfosuccinimidylpropionate on which the enzyme GOx is immobilised. The active site of the enzyme is assumed to be in direct contact with the electrode.65 Another approach for a glucose sensor is presented by Rubin et al.66 They used a mixed monolayer of ferrocenylhexadecanethiol, as mediator for electron transfer, and aminoethanethiol, for immobilising GOx.A monolayer from 2-mercaptobenzothiazole can electrically be opened and closed for underpotential deposition of copper,67 without affecting the electron-transfer rate of an Fe2+/Fe3+ redox reaction. Cysteamine [NH2(CH2)2SH] was assembled and a quinone mediator (pyrroloquinolinquinone: a catalyst for electro-oxidation of NADPH and NADH) covalently linked.68 This yielded a pH-sensitive detector for nicotinamide cofactors. When to this quinone (NADP+-dependent) malic enzyme is covalently bound, maleic acid can be measured amperometrically in the range from 1027 to 1023 m.Porphyrins (catalysts for oxygen reduction) linked with thiols on the ‘edges’ of the molecule are assembled on an optical transparent electrode.69 The effect of the electrode surface on the mechanism of oxygen reduction by metalloporphyrin films could be studied (Fig. 3). Using a transparent electrode, UV/VIS absorption spectra can be obtained simultaneously for the adsorption process. Surprisingly, long adsorption times led to the formation of multilayers, but these could be removed by extensive rinsing. Thioctic acid, a cyclic disulfide containing a carboxylic acid, is used as a pH-dependent SAM70,71 for the redox response of Ru(NH3)6 3+ and Fe(CN)6 32. At increasingly higher pH values, the SAM becomes more negatively charged and the faradaic current for Fe(CN)6 32 is reduced. In this way it acts as a cationic sensor.At low pH values when carboxylic acid is neutral, both Fe(CN)6 32 and Ru(NH3)6 3+ show a fast response. Thioctic acid serves as a basis for a ‘novel separation-free sandwich-type enzyme immunoassay’ for proteins.72 On a goldcoated microporous nylon membrane a capture antibody against human chorionic gonadotrophin (hCG) is covalently immobilised. Simultaneously, hCG and an alkaline phosphatase labelled detecting antibody are incubated.Aminophenyl phosphate is enzymatically converted into aminophenol, which is detected electrochemically.73 The detection limit for hCG was determined to be 0.8 U l21 ( Å 160 ng l21). There is a growing demand for detection systems of specific DNA sequences.74 An electrochemical method for the determination of the amount of DNA adsorbed onto an electrode surface is presented by Pang et al.75 Single-stranded DNA is covalently attached to a monolayer of thioctic acid, activated by carbodiimide and N-hydroxysuccinimide.The amount of immobilised DNA is determined indirectly through the redox reaction of tris(2,2A-bipyridyl)cobalt(iii), which complexes strongly with DNA.76 Surface Plasmon Resonance When p-polarised laser light reflects internally at the interface of an optically dense and less dense medium, an evanescent field is generated. This field will, at a specific angle, be enhanced by collective motions (resonance) of conducting electrons (plasmons) in a thin ( Å 50 nm) metal layer.77 Silver and gold films are commonly used, because of a low imaginary dielectric constant.Changes in refractive index at the gold/air or gold/solution interface shifts the resonance angle. Interactions can be followed in real time, without any labelling of the reaction partners. Again, SAM technology has already proven its merits in this sensor type. A hydrogel78 can be linked with a SAM onto a gold surface.Electrostatic interactions preconcentrate positively charged proteins (at low pH, below their isoelectric point) at residual negative charges in the hydrogel. Activation with carbodiimide and N-hydroxysuccinimide covalently binds the proteins to the matrix.12 Other methods for (covalent) coupling chemistries for this carboxymethylated dextran hydrogel matrix are presented by L�ofås et al.79 A versatile and useful biotin-functionalised SAM has been presented by Knoll and coworkers.80,81 Biotin (vitamin H) has an extremely high binding constant (Kdiss = 10215 m) for tetravalent (strept)avidin. When (strept)avidin is bound to this monolayer, a biotinylated antibody (anti-hCG) is readily immobilised.The accessibility for (strept)avidin can be optimised by preparing a monolayer of biotinylated long-chain alkanethiols, diluted with shorter hydroxythiols.82 The longchain alkanethiol acts as a spacer molecule, elevating the biotin group from the monolayer surface.Cyclodextrins are cycc oligosaccharides that chelate with small molecules. Derivatised with multiple long-chain thiol spacers, these cyclodextrin molecules assemble parallel to the surface. Oriented in this way they are freely accessible for dye molecules.83 Specific recognition with high sensitivity for cholera toxin is achieved by creating an artificial membrane with entrapped ganglioside.84 A lipid layer (membrane) can be immobilised through hydrophobic interaction upon a SAM of an alkanethiol.To form a lipid membrane upon this SAM, various deposition methods were compared. A simple and fast way to deposit an artificial membrane has been achieved by a lipid/detergent dilution technique that finally yields a micellar solution. These Fig. 3 Schematic illustration of the most probable average surface structure of tetra-thiolated cobalt porphyrin.69 (With kind permission from the American Chemical Society.) 46R Analyst, April 1997, Vol. 122micelles spontaneously assemble onto the hydrophobic SAM as an artificial membrane.The detection limit for cholera toxin (M Å 50 kDa for the B5 subunit) can be estimated as about 1029 m with SPR. Thiol-derivatised synthetic peptides (Mr Å 2 kDa) form a sensing layer for specific protein molecules.85,86 An optimum response for an antibody (Mr Å 150 kDa) is found when the SAM consisted of only 3 mol % thiolated peptide. To study the effects of surface properties on protein adsorption,87 SAMs are excellent model systems.Hydrophilic hexa(ethylene glycol)-terminated SAMs effectively reduce non-specific adsorption for a wide range of proteins found in biological fluids. A similar type of SAM is used by Duschl et al.88 11-Mercaptoundecanol-linked antigenic peptide (mimicking the surface of a malaria parasite) is diluted with underivatised 11-mercaptoundecanol. Non-specific antibodies did not adsorb, but the specific antibody (1029 m) gave a small, but noticeable shift when incubated.A series of six histidine moieties (a His tag) is incorporated in the primary sequence of recombinant proteins, in order to simplify purification. Thiols, terminated with a nitrilotriacetic acid (NTA) group and a tri(ethylene glycol) group (mentioned before) in a molar ratio 1 : 10 form a mixed SAM.89 NTA coordinates with Ni2+, leaving two vacant binding sites, which selectively chelates with the His tag.Detection of DNA hybridisation with SPR is demonstrated by Piscevic et al.90 A ten-mer oligonucleotide, with the 5A-phosphate group attached to mercaptopropyl has been immobilised on gold. Surprisingly, formation of an ‘incomplete’ SAM is preferred over dilution. The self-assembly process was stopped in an early stage of the formation. In real-time, the hybridisation process is followed, Fig. 4 Schematic representation of the molecular structure and interaction of a self-assembled monolayer of resorcin[4]arenes on an Au(111) surface with perchloroethylene molecules from the gas phase.96 (With kind permission from the American Association for the Advancement of Science).Analyst, April 1997, Vol. 122 47Rproviding a promising basis for detection of DNA hybridisation reactions with SPR. Quartz Crystal Microbalance A QCM senses mass changes that result in a shift in resonance frequency. A piezoelectric crystal oscillates at a very sharp frequency.Small mass changes on the crystal surface shifts this frequency. In liquids, the performance of a QCM is in many cases inferior to that in air.91 Sensitivity and selectivity are comparable with SPR.92 If both sides of the crystal are coated with gold, these layers can be used as electrodes. Simultaneous measurements of electrochemical parameters and mass changes are an interesting feature of the EQCM.93 With a mass-sensitive instrument, organophosphonates (nerve agents) can be selectively and reversibly detected, as shown by Kepley et al.94 Copper(ii)-ions, bound to a SAM of carboxylate-terminated alkanethiols act as the sensing layer.EQCM has been used to study the kinetics of the formation, and the oxidative or reductive desorption of alkanethiol SAMs.94 It was concluded that better SAMs are formed from poorer solvents for the thiol. Choosing the appropriate conditions (solvent, formation time, thiol concentration) is of importance when well defined monolayers are required.Fundamental knowledge about the adsorption kinetics of alkanethiols can be obtained both from SPR and QCM measurements, because the process is monitored in realtime. 95 Very small mass changes of small gaseous molecules were measured with QCM using a resorcin[4]arene monolayer (Fig. 4) for detection.96 Perchloroethylene molecules are, according to the authors, recognised by these hydrophobic cavitants and could be measured in the nanogram range. However, other authors doubt their conclusions.97 A synthetic antigenic peptide [of the foot-and-mouth disease virus (FMDV)] has been derivatised with w-hydroxyundecanethiol (HUT) and was adsorbed onto gold.91 Various procedures to prepare sensing layers from (mixed) SAMs are compared and (contrarily) the best layer is the undiluted modified-peptide SAM.The common strategy to avoid steric hindrance, diluting the HUT-modified peptide in a mixed monolayer with pure HUT, yielded a lower response to anti- FMDV-antibody.Uniform Immobilisation Strategies In developing sensitive sensors, the self-assembly technique offers interesting perspectives. By generating functionalised surfaces through modification of thiols, it is possible to selectively attach the biomolecules of interest. Another option is to derivatise a biomolecule with a thiol functionality. For repetitive measurements (cost savings), the sensor surface ought to be regenerable.A sensor surface for phosphate biomolecules can be regenerated using a pH-dependent, electrostatically attached pentamidine layer upon a monolayer of mercaptoalkanoic acid.98 An N-hydroxysuccinimide ester functionalised SAM can be applied for a uniform covalent immobilisation of amino groupcontaining biomolecules.99 Photo-immobilisation for proteins via benzophenon derivatization can be another strategy (Fig. 5). Upon UV irradiation, a 10,10A-dithiobis(decanoic acid N-hydroxysuccinimide ester) derivatised with benzophenon crosslinks with an antibody.A homogeneous single layer of antibodies results, with retention (at least in part) of the activity.100 Polymer layers are widely applied in sensors and reviewed by Hars�anyi.101 The films can be deposited upon a substrate by (a) spinning or casting, (b) electrochemical polymerisation or (c) vacuum deposition. Polymerisation to a thin film, after being adsorbed as thiol monomers, was first described by Ford et al.102 An SAM of 4-(mercaptomethyl)styrene was polymerised in an aqueous solution of azo-initiator by irradiation with a laser, yielding a hydrophobic surface.General Conclusions Monolayers from thiol-containing molecules are easy to prepare, quickly assembled and well ordered. A general protocol for self-assembly conditions is hard to give; the preparation route depends on the desired properties. Electrochemistry, especially cyclic voltammetry, can detect pinholes or defect structures in the monolayers.SAMs are stable in a potential range from 2400 to +1400 mV versus SCE; beyond these potentials the layer desorbs, yielding a clean gold surface. Small w-functional groups exhibit no influence on the formation of the monolayer. Preparing a mixed SAM from a long-chain w-functionalised thiol ‘diluted’ with a shorter-chain alkanethiol offers great challenges for analytical purposes. In this way steric hindrance of the functionality can largely be reduced. The SAM technology is widely applied in electrochemical, SPR-based and (E)QCM-based sensors.Fig. 5 The immobilisation process used for patterning of proteins. a, Avidin with photobiotin immobilised onto the surface. b, Exposure of selected areas to light through a mask results in activation of the photobiotin molecule, specifically immobilising the antibody in the solution. c, Unbound material is removed by washing. d, The entire surface is exposed to light, and a blocking molecule bound to all unreacted photobiotin groups.e, Folling washing the surface is exposed to fluorescently labeled antigen, which is bound by the patterned antibody.100 (With kind permission from Elsevier Science Ltd.) 48R Analyst, April 1997, Vol. 122References 1 Turner, A. F. P., in Biosensors: Fundamentals and Applications, ed. Turner, A. F. P., Karube, I., and Wilson, G. S., Oxford University Press, 1987, pp. V–VIII. 2 Byfield, M. P., and Abuknesha, R.A., Biosens. Bioelectron., 1994, 9, 373. 3 Lu, B., Smyth, M. R., and OAKennedy, R., Analyst, 1996, 121, 29R. 4 Goddard, N. J., Pollard-Knight, D., and Maule, C. H., Analyst, 1994, 119, 583. 5 Schramm, W., Paek, S.-H., and Voss, G., Immunomethods, 1993, 3, 93. 6 Ebersole, R. C., Miller, J. A., Moran, J. R., and Ward, M. D., J. Am. Chem. Soc., 1990, 112, 3239. 7 Morgan, H., Taylor, D. M., DASilva, C., and Fukushima, H., Thin Solid Films, 1992, 210/211, 773. 8 Achtnich, U.R., Tiefenauer, L. X., and Andres, R. Y., Biosens. Bioelectron., 1992, 7, 279. 9 Schmidt, A., Spinke, J., Bayerl, T., Sackmann, E., and Knoll, W., Biophys. J., 1992, 63, 1185. 10 Zhao, S., and Reichert, W. M., Langmuir, 1992, 8, 2785. 11 Hoshi, T., Anzai, J., and Osa, T., Anal. Chim. Acta, 1994, 289, 321. 12 F�agerstam, L. G., Frostell-Karlsson, A., Karlsson, R., Persson, B., and R�onnberg, I., J. Chromatogr., 1992, 597, 397. 13 Scouten, W. H., Methods Enzymol., 1987, 135, 30. 14 Carraway, K. L., and Koshland, Jr., D. E., Methods Enzymol., 1979, 25, 616. 15 Wolfe, C. A., and Hage, D. S., Anal. Biochem., 1995, 231, 123. 16 Matsuda, T., and Sugawara, T., Langmuir, 1995, 11, 2267. 17 Mandler, D., and Turyan, I., Electroanalysis, 1995, 8, 207. 18 Allara, D. L., Biosens. Bioelectron., 1995, 10, 771. 19 G�opel, W., and Heiduschka, P., Biosens. Bioelectron., 1995, 10, 853. 20 Xu, J., and Li, H.-L., J. Colloid Interface Sci., 1995, 176, 138. 21 Roberts, G.G., and Pitt, C. W., Langmuir-Blodgett Films, 1982, Elsevier, Amsterdam, 1983. 22 Sagiv, J., J. Am. Chem. Soc., 1980, 102, 2. 23 Nuzzo, R. G., and Allara, D. L., J. Am. Chem. Soc., 1983, 105, 4481. 24 Porter, M. D., Bright, T. B., Allara, D. L., and Chidsey, C. E. D., J. Am. Chem. Soc., 1987, 109, 3559. 25 Troughton, E. B., Bain, C. D., Whitesides, G. M., Nuzzo, R. G., Allara, D. L., and Porter, M. D., Langmuir, 1988, 4, 365. 26 Bain, C. D., Troughton, E. B., Tao, Y.-T., Evall, J., Whitesides, G.M., and Nuzzo, R. G., J. Am. Chem. Soc., 1989, 111, 321. 27 Bain, C. D., Evall, J., and Whitesides, G. M., J. Am. Chem. Soc., 1989, 111, 7155. 28 Bertilsson, L., and Liedberg, B., Langmuir, 1993, 9, 141. 29 J�onsson, U., Malmquist, M., Olofsson, G., and R�onnberg, I., Methods Enzymol., 1988, 137, 381. 30 Nuzzo, R. G., Fusco, F. A., and Allara, D. L., J. Am. Chem. Soc., 1987, 109, 2358. 31 Nuzzo, R. G., Dubois, L. H., and Allara, D. L., J. Am. Chem. Soc., 1990, 112, 558. 32 Keller, H., Schrepp, W., and Fuchs, H., Thin Solid Films, 1992, 210/211, 799. 33 Finklea, H. O., Avery, S., Lynch, M., and Furtsch, T., Langmuir, 1987, 3, 409. 34 Bain, C. D., and Whitesides, G. M., J. Am. Chem. Soc., 1989, 111, 7164. 35 Dubois, L. H., and Nuzzo, R. G., Annu. Rev. Phys. Chem., 1992, 43, 437. 36 Kim, Y.-T., McCarley, R. L., and Bard, A. J., Langmuir, 1993, 9, 1941. 37 Finklea, H. O., in Electroanalytical Chemistry: A Series of Advances, ed. Bard, A.J., and Rubinstein, I., Marcel Dekker. New York, 1996, vol. 19, pp. 110–317. 38 Ulman, A., Characterization of Organic Thin Films, Butterworth– Heinemann, Boston, 1995. 39 Edinger, K., G�olzhaeuser, A., Demota, K., W�oll, C., and Grunze, M., Langmuir, 1993, 9, 4. 40 Sun, L., and Crooks, R. M., Langmuir, 1993, 9, 1951. 41 Delamarche, E., Michel, B., Kang, H., and Gerber, C., Langmuir, 1994, 10, 4103. 42 Pan, J., Tao, N., and Lindsay, S. M., Langmuir, 1993, 9, 1556. 43 Chidsey, C.E. D., and Loiacono, D. N., Langmuir, 1990, 6, 682. 44 Widrich, C. A., Chung, C., and Porter, M. D., J. Electroanal. Chem., 1991, 310, 335. 45 Weisshaar, D. E., Lamp, B. D., and Porter, M. D., J. Am. Chem. Soc., 1992, 114, 5860. 46 Walczak, I. M., Popenoe, D. D., Deinhammer, R. S., Lamp, B. D., Chung, C., and Porter, M. D., Langmuir, 1991, 7, 2687. 47 De Weldige, K., Rohwerder, M., Vago, H., Viefhaus, H., and Stratman, M., Fresenius’ J. Anal. Chem., 1995, 353, 329. 48 Truong, K.D., and Rowntree, P. A., Prog. Surf. Sci., 1995, 50, 207. 49 Nuzzo, R. G., Zegarski, B. R., and Dubois, L. H., J. Am. Chem. Soc., 1987, 109, 733. 50 Grunze, M., Phys. Scr., 1993, 49, 711. 51 Strong, L., and Whitesides, G. M., Langmuir, 1988, 4, 546. 52 Ivarsson, B. A., Hegg, P.-O., Lundstr�om, I., and J�onsson, U., Colloids. Surface., 1985, 13, 169. 53 Liedberg, B., Ivarsson, B., Lundstr�om, I., and Salaneck, W. R., Progr. Colloid Polym., 1985, 70, 67. 54 Lundstr�om, I., Progr.Colloid Polym. Sci., 1985, 70, 76. 55 Lundstr�om, I., and Salaneck, W. R., J. Colloid Interf. Sci., 1985, 108, 288. 56 Sundgren, J.-E., Bod�o, P., Ivarsson, B., and Lundstr�om, I., Progr. Colloid Polym. Sci., 1986, 113, 530. 57 Collison, M., Bowden, E. F., and Tarlov, M. J., Langmuir, 1992, 8, 1247. 58 Malem, F., and Mandler, D., Anal. Chem., 1993, 65, 37. 59 Van Os, P. J. H. J., Bult, A., and Van Bennekom, W. P., Anal. Chim. Acta, 1995, 305, 18. 60 Sadik, O. A., and Wallace, G.G., Anal. Chim. Acta, 1993, 279, 209. 61 Sayre, C. N., and Collard, D. M., Langmuir, 1995, 11, 302. 62 Willicut, R. J., and McCarley, R. L., Adv. Mater., 1995, 7, 759. 63 Creager, S. E., and Olsen, K. G., Anal. Chim. Acta, 1995, 307, 277. 64 Riklin, A., and Willner, I., Anal. Chem., 1995, 67, 4118. 65 Jiang, L., McNeil, C. J., and Cooper, J. M., J. Chem. Soc., Chem. Commun., 1995, 1293. 66 Rubin, S. R., Chow, J. T., Ferraris, J. P., and Zawodzinski, Jr., T. A., Langmuir, 1996, 12, 363. 67 Bharathi, S., Yegnaraman, V., and Rao, G. P., Langmuir, 1993, 9, 1614. 68 Willner, I., and Riklin, A., Anal. Chem., 1994, 66, 1535. 69 Postlethwaite, T. A., Hutchison, J. E., Hathcock, K. W., and Murray, R. W., Langmuir, 1995, 11, 4109. 70 Cheng, Q., and Brajter-Toth, A., Anal. Chem., 1992, 64, 1998. 71 Cheng, Q., and Brajter-Toth, A., Anal. Chem., 1995, 67, 2767. 72 Duan, C. M., and Meyerhoff, M. E., Anal. Chem., 1994, 66, 1369. 73 Heineman, W. R., and Halsall, H. B., Anal. Chem., 1985, 57, 1321A. 74 Downs, M. E. A., Kobayashi, S., and Karube, I., Anal. Lett., 1987, 20, 1897. 75 Pang, D.-W., Zhang, Z.-L., Qi, Y.-P., Cheng, J.-K., and Liu, Z.-Y., J. Electroanal. Chem., 1996, 403, 183. 76 Millan, K. M., and Mikkelsen, S. R., Anal. Chem., 1993, 65, 2317. 77 Kretschmann, E., and Raether, H., Z. Naturforsch., 1968, 23a, 2135. 78 L�ofås, S., and Johnsson, B., J. Chem. Soc., Chem. Commun., 1990, 1526. 79 L�ofås, S., Johnsson, B., Edstr�om, Å., Hansson, A., M�uller Hillgren, R.-M., and Stigh, L., Biosens. Bioelectron., 1995, 10, 813. 80 H�aussling, L., Ringsdorf, H., Schmitt, F.-J., and Knoll, W., Langmuir, 1991, 7, 1837. 81 Schmitt, F.-J., H�aussling, L., Ringsdorf, H., and Knoll, W., Thin Solid Films, 1992, 210/211, 815. 82 Spinke, J., Liley, M., Schmitt, F. J., Guder, H. J., Angermaier, L., and Knoll, W., J. Chem. Phys., 1993, 99, 7012. 83 Mittler-Neher, S., Spinke, J., Liley, M., Nelles, G., Weisser, M., Back, R., Wenz, G., and Knoll, W., Biosens. Bioelectron., 1995, 10, 903. 84 Terrettaz, S., Stora, T., Duschl, C., and Vogel, H., Langmuir, 1993, 9, 1361. 85 Van den Heuvel, D. J., Kooyman, R. P. H., Drijfhout, J. W., and Welling, G. W., Anal. Biochem., 1993, 215, 223. 86 Kooyman, R. P. H., Van den Heuvel, D. J., Drijfhout, J. W., and Welling, G. W., Thin Solid Films, 1994, 244, 913. Analyst, April 1997, Vol. 122 49R87 Mrksich, M., Sigal, G. B., and Whitesides, G. M., Langmuir, 1995, 11, 4383. 88 Duschl, C., S�evin-Landais, A.-F., and Vogel, H., Biophys. J., 1996, 70, 1985. 89 Sigal, G. B., Bamdad, C., Barberis, A., Strominger, J., and Whitesides, G. M., Anal. Chem., 1996, 68, 490. 90 Piscevic, D., Lawall, R., Veith, M., Liley, M., Okahata, Y., and Knoll, W., Appl. Surf. Sci., 1995, 90, 425. 91 Rickert, J., Weiss, T., Kraas, W., Jung, G., and G�opel, W., Biosens. Bioelectron., 1996, 11, 591. 92 Minunni, M., Mascini, M., Guilbault, G. G., and Hock, B., An., and Buttry, D. A., J. Am. Chem. Soc., 1993, 115, 12 391. 94 Kepley, L. J., Crooks, R. M., and Ricco, A. J., Anal. Chem., 1992, 64, 3191. 95 Karpovich, D. S., and Blanchard, G. J., Langmuir, 1994, 10, 3315. 96 Schierbaum, K. D., Weiss, T., Thoden van Velzen, E. U., Engbersen, J. F. J., Reinhoudt, D. N., and G�opel, W., Science, 1994, 265, 1413. 97 Grate, J. W., Patrash, S. J., Abraham, M. H., and Du, C. M., Anal. Chem., 1996, 68, 913. 98 Sellergren, B., Swietlow, A., Arnebrant, T., and Unger, K., Anal. Chem., 1996, 68, 402. 99 Wagner, P., Hegner, M., Kernen, P., Zaugg, F., and Semenza, G., Biophys. J., 1996, 70, 2052. 100 Morgan, H., Pritchard, D. J., and Cooper, J. M., Biosens. Bioelectron., 1995, 10, 841. 101 Hars�anyi, G., Mater. Chem. Phys., 1996, 43, 199. 102 Ford, J. F., Vickers, T. J., Mann, C. K., and Schlenoff, J. B., Langmuir, 1996, 12, 1944. Paper 6/06964I Received October 11, 1996 Accepted December 26, 1996 50R Analyst, April 1997, Vol.
ISSN:0003-2654
DOI:10.1039/a606964i
出版商:RSC
年代:1997
数据来源: RSC
|
2. |
Critical Review. Approaches to Predicting Stability ConstantsACritical Review |
|
Analyst,
Volume 122,
Issue 4,
1997,
Page 51-58
Robert D. Hancock,
Preview
|
|
摘要:
Critical Review Approaches to Predicting Stability Constants A Critical Review Robert D. Hancock Department of Chemistry, University of the Witwatersrand, Johannesburg, South Africa Summary of Contents Introduction Monodentate Ligands Complexes of Polyamine Ligands Linear Free Energy Relationships of logK1 for N-donor Ligands and O-donor Ligands and logK1 (NH3) Accuracy of the ‘E, C and D’ Equation Multidentate Ligands References Keywords: Stability constants; formation constants; prediction methods; review Introduction In a recent excellent and useful review in this journal, Dimmock et al.1 covered all the methods that are currently available to researchers for estimating unknown formation constants. They pointed out the importance of a knowledge of formation constants in analytical chemistry, ranging from the need to understand speciation in developing many analytical procedures to the prediction of speciation of metal ion complexes in the environment.Clearly, it is not possible to determine the formation constants of complexes of all possible ligands and metal ions, particularly the ternary complexes that arise in complex mixtures, so that methods for the estimation of unknown formation constants will be an important aspect of modelling such systems.One aspect that was not touched on in their review, however, was an evaluation of the reliability of the different methods presented that was independent of the claims of the proponents of the methods.The reader was left with the impression that all the methods are reliable and work equally well. Three important points in particular were not touched on. (1) Most empirical methods can be made to reproduce reasonably well the data used in the fitting of some equation or correlation that is expected to predict formation constants. The test of a method of prediction, however, is the performance of the method with regard to formation constants that were not used in fitting the equation or correlation.This extends also to formation constants that are not currently known, or are perhaps even unmeasurable for reasons such as hydrolysis or insolubility. One has to ask the question, ‘Is this prediction consistent with the known chemistry of this metal ion?’ The latter point is well illustrated2,3 by the complexes of metal ions with ammonia and polyamine ligands. For example, BiIII had until recently no known solution chemistry with the latter ligands.4–7 Do methods for estimating formation constants predict logK1 with ammonia and BiIII consistently with the absence of chemistry of the metal ion with such ligands? For an ammonia complex to be resistant to hydrolysis, it should not be decomposed by water according to the equation M(NH3)n++H2O?M(OH)(n21)++NH4 + (1) In eqn.(1), with pKw = 14.0 and pKa for NH3 = 9.22,8 logK1 (NH3) for any metal ion must not be significantly less than logK1(OH2) by more than 14 2 9.22 = 4.78 log units for there to be detectable amounts of the ammonia complex present in solution.Any method of prediction that suggests that logK1(NH3) for a metal ion is large enough for it to survive hydrolysis according to eqn. (1), which metal ion is known to have no aqueous chemistry with sp3 hybridized nitrogen donor ligands, must therefore come under suspicion. Most of the methods of prediction described in the review by Dimmock et al.1 fail this type of test, as discussed below, in that they predict that many metal ions that form no complexes with ammonia should have a rich chemistry with such nitrogen donor ligands.2 For example, the Edwards equation9 predicts that logK1(NH3) for PbII is 6.5.If this were correct, PbII [logK1(OH2) = 6.38] would form very stable complexes with ammonia in aqueous solution. A subsequent study10 has shown that, in fact, logK1(NH3) for PbII is 1.6, which complex can only barely be stabilized against hydrolysis in 5 m NH4 + ion.The intention here is not to state that all these methods of prediction are without value, but simply that the user should be aware that they may be deficient if extended to ligand types that are different from those used in fitting the equation. (2) Have the proponents of methods of prediction of unknown logK1 values actually made any effort to validate their methods of prediction by checking predictions that their Rob Hancock obtained his BSc at Rhodes University, and PhD at the University of Cape Town.He carried out research in ligand design at the National Institute for Metallurgy in Johannesburg, followed by a period at the University of the Witwatersrand as Professor of Inorganic Chemistry. He is currently carrying out research at IBC Advanced Technologies in the USA. He is author of 185 research papers, and a book ‘Metal Complexes in Aqueous Solutions’, co-authored with A. E. Martell. Current interests include ligand design for analysis of metal ions by fluorescence spectroscopy, and complexes for positron emission tomography.Analyst, April 1997, Vol. 122 (51R–58R) 51Rmethods make, particularly those that are not of a more limited nature? Although these are obviously useful, one would regard as limited the use of LFER (linear free energy relationships) between two very similar metal ions, e.g., CoII and NiII, with a set of very similar ligands, e.g., bidentate ligands containing two negative oxygen donors, such as oxalate, malonate and catecholate (see Fig. 1 for ligand abbreviations).Much more significant is the prediction of logK1 for a complex where no currently known chemistry would allow for a guess at the logK1 value. Validation of such predictions renders the method of prediction much more trustworthy. One should point out here for readers who are not familiar with the determination of formation constants, and who may feel that checking constants is an inherently inaccurate process, that with use of techniques such as glass electrode potentiometry and modern computer programs, it is usual for experienced researchers to achieve accuracy of formation constant determination of better than 0.1 log unit.Disagreements in the literature between reported values for the same constant can be traced8 to such factors as the use of impure ligands and application of incorrect models in analyzing the potentiometric data.One can usually discern whether a constant is likely to be accurate8 by reading the paper reporting it, and checking off 10 criteria that indicate the quality of the experimental work. (3) An important aspect of the prediction of formation constants is the extent to which methods are purely empirical. The LFER is an important method of estimation of unknown logK values. However, if used in a purely empirical way, without knowledge of accompanying theories on how structural and bonding factors control complex stability, misleading predictions can be made.This is shown in Fig. 2, where logK1 for PbII complexes with polyaminocarboxylates is plotted against logK1 for the corresponding complexes with ZnII. Fig. 2 contains all the data points that are available for these two metal ions with ligands of this type in the compilation of Martell and Smith.8 If we suppose that logK1 for TMDTA with PbII had not been known, and we were to use the LFER in Fig. 2 to estimate this logK1 value, the remainder of the points present an impressive correlation with a correlation coefficient of 0.98, and we would be confident that our prediction of logK1 for PbII with TMDTA of 16.5 was correct within likely experimental error. However, the experimental value8 is 13.7. This is probably not an acceptable level of accuracy of prediction for most purposes. How can this single point be such an outlier? The answer is that it involves an important structural feature not present in any of the other ligands, i.e., that TMDTA contains a six-membered chelate ring, where the other ligands all contain five-membered chelate rings. One has to be aware of theories regarding the role of chelate ring size11–13 and metal ion size in complex stability before one might expect such an effect. Purely empirical approaches without accompanying insight into theories of complex formation run the risk of giving incorrect predictions.Fig. 1 Ligands discussed in this paper. 52R Analyst, April 1997, Vol. 122The more a method of prediction is grounded in theory, the less likely it is that it will make highly erroneous predictions. These three aspects of prediction of stability constants are dealt with below in relation to unidentate ligands first of all, followed by a discussion of the role of architectural features in multidentate ligands and how these relate to complex stability and its prediction. How these factors might all be built into a computer program which is an expert system, i.e., able to function as an aid to ligand design, is discussed.Monodentate Ligands The obvious patterns present in the formation constants of complexes of monodentate ligands have led14 several workers to propose classification schemes15–17 and equations2,3,9,17–23 for the prediction of formation constants. All these schemes can be summarized in the HSAB (Hard and Soft Acids and Bases) classification of Pearson.17 The HSAB classification is based on the preferences of metal ions for ligands with different donor atom types, as indicated by their formation constants.Thus, for example, complexes of soft metal ions have stability orders with halide ligands F2 << Cl2 < Br2 < I2, while complexes of hard metal ions have stability orders F2 >> Cl2 > Br2 > I2. To model these differing stability sequences, Edwards proposed a dual parameter equation:9 log(K1/K0) = aEn+bH (2) This equation is typical of most of the equations that have been used to predict formation constants in an attempt to mimic HSAB behaviour.A pair of parameters is used, one of each of which relates to the metal ion, in this case a and b, and the other pair to the ligand, in this case En and H. One pair of parameters represents the tendency to covalence or softness (a and Ea) in the M–L bond, while the other represents the tendency to hardness or ionicity (b and H) in the M–L bond.The limitations and inaccuracy of some of the predictions made by this type of equation have been discussed elsewhere.2,3 We shall confine ourselves here to discussing the predictions made by the ‘E and C’ equation of Drago, adapted and extended2,3 to apply to aqueous solutions, and the equations of Brown and coworkers21 –23 (the BSE equations), which take a completely different type of approach to that adopted by other, more empirical, approaches. In their review, Dimmock et al.1 gave a very comprehensive account of the BSE equations, so that only a brief description will be given here.Brown and co-workers developed the BSE equations which predicted logK1(OH2) for a large number of metal ions using physical properties of the ions such as charge and ionic radius and factors such as the protonation constant and a quantity defined as the electronicity of the ligand, and so on. This worthwhile approach predicts logK1(OH2) with considerable accuracy, and has the potential to predict formation constants for metal ions such as Lr3+, about which very little is known.These authors also developed parameters for ammonia complexes, and here it appears that they did not thoroughly examine the implications of their predictions. The BSE equations can be used to predict logK1(NH3) for all metal ions for which parameters are available. Many of these predictions fail the test of eqn. (1). Thus, for example, La3+ is predicted to form a complex with ammonia with logK1(NH3) = 2.55, which means that La3+ [logK1(OH2) = 5.38] should have a rich chemistry of ammonia complexes, and also with polyamines, similar to that of ZnII [logK1(NH3) = 2.1, logK1(OH2) = 5.08], which is not the case.One can see from Fig. 3 that the parameters in the BSE equation as currently developed basically model logK1(NH3) as 0.6 logK1(OH2). In Fig. 3, the values of logK1(NH3) predicted by the BSE equation for metal ions are plotted against their experimental8 values of logK1(OH2), and this plot has a slope of approximately 0.6.Clearly, the parameters for NH3 are similar to those for OH2, which later discussion shows should not be the case. Fig. 2 LFER of logK1 for PbII complexes plotted against logK1 for the corresponding complexes of ZnII. Ligands have been limited to aminocarboxylates, i.e., those containing sp3 hybridized nitrogens and N-acetate groups, or other similar donor groups such as phenolates.Ligands: 1, glycine; 2, N,N-bis(2-hydroxyethyl)alanine; 3, IDA, iminodiacetate; 4, MIDA, N-methyliminodiacetate; 5, EDMA, ethylenediaminemonoacetate; 6, N-2-methoxyethyliminodiacetate; 7, HIDA, 2-hydroxyethyliminodiacetate; 8, N-tetrahydropyranylmethyliminodiacetate; 9, ethylenediaminedi- 2-propionic acid; 10, ethylenediaminediserine; 11, 1,6-dicarboxypiperidine- N-acetate; 12, NTA, nitrilotriacetate; 13, EDDA, ethylenediamine-N,NA-diacetate; 14, N-2-aminoethyliminodiacetate; 15, EGTA, 4-oxa-1,7-diazaheptane-N,N,NA,NA-tetraacetate; 16, HEDTA, 2-hydroxyethylethylenediaminetriacetate; 17, EEDTA, 4,7-dioxa- 1,10-diazadecane-N,N,NA,NA-tetraacetate; 18, EDTA; 19, HBED, N,NAbis( 2-hydroxybenzyl)ethylenediaminediacetate; 20, DTPA, diethylenetriaminepentaacetate; and 21, CDTA, trans-1,2-diaminocyclohexanetetraacetate. The point for TMDTA (trimethylenediaminetetraacetate) should be noted, as discussed in the text, in that it falls well off the correlation. Formation constants from ref. 4. Fig. 3 Values of logK1(NH3) predicted (5) by the BSE equation of Brown and co-workers,21–23 plotted against experimental8 values of logK1(OH2). The dotted lines to the open circles show the positions of the experimentally known values of logK1(NH3) or those which have been estimated as discussed later, shown in Fig. 5. The point that the figure makes is that the parameters currently employed in the BSE equation to predict logK1(NH3) are too like those for predicting logK1(OH2), and do not distinguish these ligands in terms of HSAB trends.In other words, metal ions that have high logK1(OH2) should also have high logK1(NH3), which is not found to be so, as discussed in the text. Analyst, April 1997, Vol. 122 53RMany predictions of logK1(NH3) by the BSE equation fall into a gray area, as far as validation by eqn. (1) is concerned. The prediction for BiIII of logK1(NH3) is 7.86, which with logK1(OH2) of 12.98 means that ammonia complexes will tend to be hydrolyzed, in agreement with observation.This is substantially different from the value of logK1(NH3) of 5.0 predicted by the equation of Hancock and Marsicano.2,3 Complexes of BiIII with ammonia do not exist in appreciable quantities in aqueous solution, so how can one decide which is the more reliable estimate? Two approaches exist that allow us to make a judgement. Complexes of Polyamine Ligands The affinity of metal ions for ammonia parallels fairly closely their affinity for polyamine ligands.A major difference, however, is that for polyamine ligands the complexes may be sufficiently stabilized by the chelate effect24 to be stable in aqueous solution, where the ammonia complexes are not. Thus, complexes of PbII with ammonia are easily hydrolyzed,10 whereas those with polyamines such as dien or trien are not. One may develop theoretical equations25 that relate the stability of complexes of n-dentate polyamines to those of their ammonia complexes: logK1(polyamine) = 1.152.logbn(NH3) + (n21)log55.5 (3) The factor of 1.152 models the greater inductive effects of the primary to tertiary amines present in polyamines (pKa Å 10.6) compared with ammonia (pKa 9.22) and the (n21)log55.5 term is the entropic contribution25 to the chelate effect.Eqn. (3) can be further extended to predict logK1 for aminocarboxylate ligands such as EDTA from the known values of logK1 for ammonia and acetate. The logK1 value for BiIII with EDTA8 suggests that logK1(NH3) with BiIII is 5.0.This prediction suggested that BiIII should have an extensive chemistry with polyamine ligands, both open-chain ligands such as dien and trien and azamacrocycles such as cyclen and cyclam. We have made an extensive effort to check the predictions of logK1 for BiIII with polyamine ligands. This has resulted in polarographic studies of these complexes, with the results appearing to be more consistent with logK1(NH3) for BiIII = 5.0 than the 7.86 predicted by the BSE equations.In fact, eqn. (3), modified25 to include estimates of the stepwise decrease in logKm(NH3) values as m increases 1, 2, . . ., n, predicts logK1 for the dien and trien complexes of BiIII to be 17.7 and 22.3, respectively, in reasonable agreement with the experimental values in Table 1. The predicted value of logK1 for BiIII with NH3 suggested a previously unsuspected chemistry of BiIII with azamacrocycles.This led to the synthesis7 of the complex of BiIII with cyclen, the structure of which is shown in Fig. 4, and of a complex of BiIII with THP–cyclen.6 We have validated predictions of logK1(NH3) made2,3 by eqn. (4) below for complexes of several metal ions such as LaIII,26 [VO]2+ complexes,27 complexes of GaIII and InIII,28 [UO2]2+ complexes29 and PuIV complexes30 by studying complexes of polyamines and more hydrolysisresistant complexes of amine ligands such as THEEN, BIPY and AMPY.LFER of LogK1 for N-donor Ligands and O-donor Ligands and LogK1(NH2) It has been pointed out10 that the process where a ligand such as IDA replaces a ligand such as ODA resembles the formation of an ammonia complex in that an sp3 hybridized O-donor is being replaced by an sp3 hybridized N-donor in both cases (see Scheme 1). Thus, for several ligand pairs where there is a neutral oxygen donor in one member of the pair in place of a neutral nitrogen donor in the other member of the pair, a plot such as that in Fig. 5 may be drawn. Fig. 5 shows an LFER of logK1(IDA) 2 logK1(ODA) versus logK1(NH3) for a variety of metal ions. The correlation includes experimental points for logK1(NH3) and also values estimated by the combined use of eqns. (3) and (4) together with other correlations similar to that shown in Fig. 5. The degree of agreement between logK1(NH3) values predicted by eqn. (3) and by correlations such as that shown in Fig. 5, and also eqn. (4), are encouraging, and suggest that the estimates are substantially correct. Accuracy of the ‘E, C and D’ Equation The following multiparameter equation, which is of the type reported by Drago et al.,31 is typical of all such multiparameter equations, and is here modified2,3 to predict formation constants in aqueous solution rather than heats of complex formation in solvents of low dielectric constant: logK1 = EAEB + CACB (4) where the E parameters represent the ionic contribution to the M–L bond and the C parameters represent the covalent contribution for the Lewis acid A and base B.This equation was used successfully to reproduce all known formation constants of complexes of F2, OH2 and NH3. The fitting of the equation2,3 included experimentally inaccessible logK1(NH3) values estimated by independent methods,10,25 as described above. The parameters in an equation such as eqn. (4) can be set31 arbitrarily, or they can be set with some physical model in mind.Since logK1(F2) and logK1(OH2) are known for virtually all metal ions,8 and considerable evidence suggests that M–F bonds are essentially ionic, EB for F2 was set as 1 and CB as zero, with effectively no covalence in the M–F bond, and EA for each metal ion = logK1(F2). Since M–OH bonds are certainly more covalent than M–F bonds, CA for each metal ion was set as logK1(OH2)/14. A best fit gives EB = 21.08 and cB = 12.34 for ammonia. Rearrangement of eqn.(4) gives Table 1 LogK1 values for various complexes. Formation constants from refs. 5–8 Ligand Parameter NH3 ampy dien trien cyclen LogK1 (BiIII) (5.0)* 9.4 17.5 21.9 23.5 LogK1 (CuII) 4.1 9.5 16.0 20.1 23.3 LogK1 (HgII) 8.8 (12.0)† 21.8 24.8 25.5 LogK1 (PbII) 1.6 3.95 7.5 10.4 15.9 * Estimated,2,3 as discussed above. † Estimated14 as 1 2 [logK1(BIPY) + logK1(en)]. Fig. 4 Structure of the complex of BiIII with cyclen [Bi(cyclen)- (ClO4)3H2O], as determined by X-ray crystallography.7 That BiIII would have an extensive chemistry with N-donor ligands was based on the prediction2,3 of logK1(NH3) by eqn.(4). Three of the four coordinated oxygens are derived from perchlorate anions and the fourth is from a coordinated water molecule. 54R Analyst, April 1997, Vol. 122logK1(NH3) = 0.881[logK1(OH2)]21.08[logK1(F2)] (5) Eqn. (5) indicates that for a constant value of logK1[OH2], as logK1(F2) increases for a metal ion so its value of logK1(NH3) must decrease.This is the very essence of the HSAB idea. There is no one order of acid strength for metal ions. Rather, a soft metal ion such as HgII will have a high logK1(NH3) and low logK1(F2), whereas a hard metal ion such as ThIV, which has a similar affinity to HgII for the hydroxide ion, will have the reverse situation (Table 2). Eqn. (5) can be tested graphically, as shown in Fig. 6, where logK1(NH3) values, both experimental and predicted by other means,10,25 are plotted against 0.881[logK1(OH2)] 2 1.08[logK1(F2)].Fig. 6 is a graphical representation of HSAB ideas. Eqn. (4) can be successfully fitted to logK1 values only for F2, NH3 and pyridine and unidentate ligands containing a negative oxygen donor such as OH2, phenols or carboxylic acids. It cannot deal with ligands with donor atoms from the third or lower periods in the Periodic Table, such as Cl, Br, I, S, Se, P or As. To do so, a third pair of parameters, the D parameters, has to be included, as in the equation logK1 = EAEB + CACB2DADB (6) The DA parameters for the metal ions decrease as the metal ion increases in size, whereas the DB parameters for the ligand increase as the ligand donor atom increases in size.The D parameters have been interpreted2,3 as reflecting the steric strain that arises when ligands with large donor atoms replace the small water molecule from the coordination sphere of a metal ion, and also may reflect the loss of ability to stabilize the complex by hydrogen bonding between ligands coordinated to the metal ions and the solvent water.Thus, for large metal ions such as Ag+, Hg2+ and Pb2+, which have ionic radii in excess of 1.0 Å, DA is zero, and is at a maximum for the small proton. The accuracy of prediction obtained with use of eqn. (6) has already been discussed.2,3 In the original papers, predictions of logK1 were included for those ligands and metal ions examined whose logK1 values were unknown at the time of publication.Since then, logK1 values have been published8 for the azide complexes of some of these metal ions, and the predicted2,3 and subsequently observed logK1 values can be compared (Table 3). The most recent set of E, C and D parameters for use in eqn. (6) appear in ref. 14. Eqn. (6) is philosophically not very different from those multiparameter equations that have preceded it.9,17–20 The superiority of eqn. (6) at predicting logK1 for unidentate ligands derives from the care with which we have developed the parameters in it and the extensive studies which have been aimed at validating its predictions.No doubt eqn. (6) can be improved further, and possibly the parameters in it can be derived theoretically rather than empirically. The BSE equation21 –23 represents a good start at developing a more theorybased approach to the prediction of logK1 values, and possibly Fig. 5 Estimation10 of logK1(NH3) for metal ions where ammonia complexes do not exist in aqueous solution because of hydrolysis.The solid circles are for experimental logK1 values and the open circles are for metal ions for which the ammonia complexes do not exist in aqueous solution, and for which logK1(NH3) has been estimated by a number of correlations of the type shown here. The correlation is for logK1 for the IDA complex minus logK1 for the ODA complex versus logK1(NH3), as discussed in the text. Formation constants from ref. 8. Table 2 LogK1 values for HgII and ThIV LogK1(F2) LogK1(OH2) LogK1(NH3) HgII (soft) 1.5 10.6 8.8 ThIV (hard) 8.44 10.8 0.4* * Estimated from eqn. (5) and Fig. 5; other logK1 values from ref. 8. Scheme 1 Analyst, April 1997, Vol. 122 55Rcould be refined to predict more accurately the formation constants of ligands such as ammonia, and also ligands with heavy donor atoms, taking into account steric considerations that appear to be important to the success of eqn. (6). Multidentate Ligands Equations such as eqn.(3), and its extension to polyaminocarboxylates, have been shown25 to predict the formation constants of polyamines such as dien and trien and aminocarboxylates such as NTA and EDTA very well, provided that the ligands form only five-membered chelate rings on complex formation. The question that arises is how to extend equations such as eqn. (3) to ligands that contain other architectural features, such as six-membered chelate rings, or neutral oxygen donors, C-alkyl substituents or macrocyclic rings. The way in which this can be done is simple.It has been shown14 that many such structural changes produce changes in complex stability that are related to metal ion radius.32 Thus, replacement of a five-membered chelate ring with a six-membered chelate ring will increase the selectivity for smaller relative to larger metal ions,11,33 as will addition of C-alkyl groups34 (by selectivity in this review is meant the difference in logK1).Addition of groups containing neutral oxygen donors11 will shift the selectivity in favour of large metal ions. In Fig. 7–9 are shown examples of these effects, where the change in logK1, DlogK, caused by the structural changes indicated, is plotted as a function of ionic radius.32 In Fig. 7 is shown the effect on metal ion size-based selectivity of an increase in chelate ring size from all fivemembered in the NTA complex to all six-membered in the NTP complex, or one five- and two six-membered chelate rings in NADP.The small BeII ion actually shows a small increase in logK1 on increase of chelate ring size, but as the metal ion size increases further there is a steady decrease in logK1 for the NTP or NADP complex relative to the NTA complex. The type of relationship in Fig. 7 is general for a wide range of metal ions, and is best illustrated by the relationship involving EDTA and TMDTA.11 in Fig. 8 is shown the relationship between metal ion size and the effect on complex stability of adding C-methyl substituents to EDTA. It is seen that the addition of two methyl groups to the ethylene bridge between the two nitrogens of EDTA causes a metal ion size related change in complex stability such that the complexes of larger metal ions are stabilized less than those of smaller metal ions.A similar effect is produced by adding isopropyl groups to two acetates of EDTA to give DP-EDTA in Fig. 8.The difference is that the small methyl groups produce general stabilization, whereas the bulky isopropyl groups produce destabilization of the complexes. Fig. 6 Plot demonstrating the relationship between log K1 for the F2, OH2 and NH3 complexes of metal ions. The relationship is a graphical validation of equations (4) and (5). Experimental formation constants (5) from ref. 8, or log K1(NH3) estimated (2) as described in the text, with log K1 for OH2 and F2 experimental. Table 3 Comparison of predicted and observed logK1 values Metal ion LogK1(azide) GaIII InIII UO2 2+ LaIII Predicted2,3 4.28* 4.31 2.10 1.33 Observed8† 4.40 4.09 2.14 1.52 * The E and C parameters originally reported2,3 for GaIII have since been revised,14 owing to a better value8 of logK1(OH2) for GaIII.The original E, C and D parameters for GaIII gave logK1(azide) = 3.84, still a reasonable level of agreement. † Experimental logK1 values from ref. 8, corrected to ionic strength zero.Fig. 7 Effect of chelate ring size on complex stability. As shown in structure (a), a five-membered chelate ring of the ethylenediamine type forms11–13 with least steric strain with a large metal ion of M–N bond length 2.5 Å and N–M–N bond angle 69°. In structure (b) it is seen that lowest steric strain occurs for a six-membered chelate ring of the 1,3-diaminopropane type when the coordinated metal ion is small with an M–N length of 1.6 Å and an N–M–N angle of 109.5°.A consequence of the metal ion size preference of chelate rings of different sizes is shown in (c), where logK1(NTP) 2 logK1(NTA) (5) is plotted as a function of metal ion radius.32 NTP = nitrilotripropionate and NTA = nitrilotriacetate. Also shown is the plot of logK1(NADP) 2 logK(NTA) (2) as a function of metal ion radius. NADP = nitriloacetatedipropionate. The correlations in (c) show how increasing metal ion size causes a decrease in logK1 as the chelate ring size increases from all five-membered in NTA to all six-membered in NTP, or one five-membered and two six-membered in NADP.Formation constants from ref. 8. 56R Analyst, April 1997, Vol. 122The addition of groups bearing neutral oxygen donors can also be viewed as an architectural change in the ligand. In the precursor complex, e.g., the en complex shown in Fig. 9, the remaining coordination sites are filled with water molecules. Thus, adding connecting bridges to the en complex to give either THEEN or 18-ane-N2O4 as shown can be viewed as merely adding connecting bridges between the N-donors and Odonors on the complex.Fig. 9 shows the metal ion size-related change in complex stability that occurs on adding neutral oxygen donor groups to en complexes to give complexes of THEEN, of 18-aneN2O4 and cryptand-222. It may turn out that all purely architectural changes in complex structure produce changes in complex stability that are directly related to metal ion radius, which seems to be the case from the available8 formation constant data.One can thus further modify eqn. (3) to predict unknown formation constants simply by adding a further term that models the slopes and intercepts of diagrams such as Figs. 7–9. We have devised a computer program, LOGKEST, that runs on IBM PC compatible machines, based on the ‘E, C and D’ equation [eqn. (6)], which predicts formation constants for unidentate ligands. These predictions are then used in equations such as eqn.(3) to predict the formation constants of ligands such as trien or NTA. For known ligands the effects on complex stability of architectural features such as larger chelate rings, addition of neutral oxygen donors, the presence of macrocyclic or cryptate structures or C-alkyl substituents can be corrected for from a simple term involving the ionic radius of the metal ion, which models the slopes and intercepts of diagrams such as Figs. 7–9. As input for a metal ion, the data required to be known are ideally logK1 for the F2 and OH2 complexes, and the ionic radius, while knowing logK1 for one polydentate ligand such as EDTA is useful.From this simple input the program can make reasonable predictions of logK1 for a great number of ligands for the metal ion, ranging from unidentate ligands such as thiosulfate or cyanide, through ligands of intermediate complexity such as trien or EDTA, to ligands of considerable complexity such as cyclam or cryptand-222.The three steps that LOGKEST uses to predict a formation constant for a complex are as follows: (1) The ‘E, C and D’ equation [eqn. (6)] is used to predict logK1 for complexes of unidentate ligands. (2) For non-macrocyclic multidentate ligands that form only five-membered chelate rings on complex formation, the chelate effect equations such as eqn. (3) are used together with the predicted values for the unidentate groups from step 1 to predict logK1 for the complex of the multidentate ligand.(3) For ligands that have other architectural features present, such as chelate rings of sizes other than five-membered, C-alkyl substituents, groups bearing neutral oxygen donors and macrocyclic rings, further corrections are made to the predictions in step 2. This involves two parameters that represent the slope and intercept of relationships such as those shown in Figs. 7–9. These are used to derive formation constants from the formation constants calculated for the analogous ligands containing fivemembered chelate rings only. Fig. 8 The effect of C-alkyl substituents on complex stability. At (a) above is shown diagrammatically the way in which a small metal ion causes a large curvature in a coordinated ligand, so that C-alkyl substituents are well separated, while at (b) is shown a ligand coordinated to a large metal ion where the small curvature of the ligand pushes the C-alkyl substituents closer together. The result of this effect is shown at (c), where it is seen that addition of C-alkyl substituents to EDTA to give DMEDTA (2) or DPEDTA (5) produces a change in complex stability that is strongly related to metal ion size.The DlogK values at logK1(EDTA) 2 logK1(DMEDTA), or logK1(EDTA 2 logK1 DP-EDTA). Formation constants from ref. 8. Fig. 9 Effect of substituents bearing neutral oxygen donors on complex stability. In (a) it is shown how the changes in going from an en complex to a THEEN or 18-ane-N2O4 complex can be regarded as a structural change brought about by inserting ethylene bridges between nitrogens and coordinated water molecules.In (b) it is shown how the addition of neutral oxygen donors produces a change in complex stability such that the complexes of larger metal ions are stabilized relative to those of smaller metal ions. The effect on complex stability is similar whether an open-chain ligand or a macrocycle is formed by the structural change in the ligand.Ionic radii from ref. 32 and stability constants from ref. 8. Analyst, April 1997, Vol. 122 57RA possible limitation of LOGKEST at present in some cases may lie in predicting the response that will be obtained to changes in ligand architecture in step 3 above. If it were possible to predict the metal ion size-related changes that would occur for unknown ligands, it would then be possible to make LOGKEST an expert ligand design system, where reasonable predictions of logK1 could be made for ligands that were designed on the computer screen.Since these metal ion sizerelated changes in complex stability are steric effects, it would seem that the best approach to predicting them without recourse to synthesis of the ligand and measurement of some stability constants would be molecular mechanics (MM) calculations.12 This aspect of ligand design is currently being explored. A further important aspect that has not been touched on here is that of higher complexes, MLn where n > 1, and ternary complexes, MLaLb, where La and Lb are different ligands, which should be the subject of future investigations. Clearly, this is a subject of greater complexity than that of a single ligand attaching to a metal ion, as the two separate ligands may coordinate in a variety of ways with many different steric possibilities, and again here MM analysis should be a useful tool.The author thanks the University of the Witwatersrand and the Foundation for Research Development for generous financial support for this work.References 1 Dimmock, P. W., Warwick, P., and Robbins, R. A., Analyst, 1995, 120, 2159. 2 Hancock, R. D., and Marsicano, F., Inorg. Chem., 1978, 17, 560. 3 Hancock, R. D., and Marsicano, F., Inorg. Chem., 1980, 19, 2709. 4 Hancock, R. D., Cukrowski, I., and Mashishi, J., J. Chem. Soc., Dalton Trans., 1993, 2895. 5 Hancock, R. D., Cukrowski, I., Cukrowska, E., Antunes, I., Mashishi, J., and Brown, K., Polyhedron, 1995, 14, 1699. 6 Luckay, R., Reibenspies, J. H., and Hancock, R. D., J. Chem. Soc., Chem. Commun., 1995, 2365. 7 Luckay, R., Cukrowski, I., Mashishi, J., Reibenspies, J. H., Bond, A. H., Rogers, R. D., and Hancock, R. D., J. Chem. Soc., Dalton Trans., 1997, 901. 8 Martell, A. E., and Smith, R. M., Critical Stability Constants, Plenum Press, New York, 1974–89, vols. 1–6. 9 Edwards, J. O., J. Am. Chem. Soc., 1954, 76, 1540. 10 Mulla, F., Marsicano, F., Nakani, B.S., and Hancock, R. D., Inorg. Chem., 1985, 24, 3076. 11 Hancock, R. D., Pure Appl. Chem., 1986, 58, 1445. 12 Hancock, R. D., Prog. Inorg. Chem., 1989, 37, 187. 13 Hancock, R. D., Acc. Chem. Res., 1990, 23, 253. 14 Martell, A. E., and Hancock, R. D., Metal Complexes in Aqueous Solutions, Plenum Press, New York, 1996. 15 Schwarzenbach, G., Adv. Inorg. Radiochem., 1961, 3, 257. 16 Ahrland, S., Chatt, J., and Davies, N. R., Q. Rev. Chem. Soc., 1958, 12, 265. 17 Pearson, R. G., Chem. Br., 1967, 3, 103. 18 Yingst, A., and McDaniel, D. H., Inorg. Chem., 1967, 6, 1067. 19 Yamada, S., and Tanaka, M., J. Inorg. Nucl. Chem., 1975, 37, 587. 20 Misono, M., and Saito, Y., Bull. Chem. Soc. Jpn., 1970, 43, 3680. 21 Brown, P. L., Sylva, R. N., and Ellis, J., J. Chem. Soc., Dalton Trans., 1985, 723. 22 Brown, P. L., and Sylva, R. N., J. Chem. Res, 1987, (S) 4; (M) 0110. 23 Brown, P. L., Talanta, 1989, 3, 351. 24 Schwarzenbach, G., Helv.Chim. Acta, 1952, 35, 2344. 25 Hancock, R. D., and Marsicano, F., J. Chem. Soc., Dalton Trans., 1976, 1096. 26 Hancock, R. D., Jackson, G. J., and Evers, A., J. Chem. Soc., Dalton Trans., 1979, 1384. 27 Duma, T. W., and Hancock, R. D., J. Coord. Chem., 1994, 31, 135. 28 Duma, T. W., Marsicano, F., and Hancock, R. D., J. Coord. Chem., 1991, 23, 221. 29 Jarvis, N. V., de Sousa, A. S., and Hancock, R. D., Radiochim. Acta, 1992, 57, 33. 30 Jarvis, N. V., de Sousa, A. S., and Hancock, R.D., Radiochim. Acta, 1994, 64, 15. 31 Drago, R. S., Vogel, G. C., and Needham, T. E., J. Am. Chem. Soc., 1971, 93, 6014. 32 Shannon, R. D., Acta Crystallogr., Sect. A, 1976, 32, 751. 33 Hancock, R. D., and Martell, A. E., Chem. Soc. Rev., 1989, 89, 1875. 34 Hancock, R. D., Maumela, H., and de Sousa, A. S., Coord. Chem. Rev., 1996, 148, 315. Paper 6/07993H Received November 26, 1996 Accepted January 16, 1997 58R Analyst, April 1997, Vol. 122 Critical Review Approaches to Predicting Stability Constants A Critical Review Robert D.Hancock Department of Chemistry, University of the Witwatersrand, Johannesburg, South Africa Summary of Contents Introduction Monodentate Ligands Complexes of Polyamine Ligands Linear Free Energy Relationships of logK1 for N-donor Ligands and O-donor Ligands and logK1 (NH3) Accuracy of the ‘E, C and D’ Equation Multidentate Ligands References Keywords: Stability constants; formation constants; prediction methods; review Introduction In a recent excellent and useful review in this journal, Dimmock et al.1 covered all the methods that are currently available to researchers for estimating unknown formation constants.They pointed out the importance of a knowledge of formation constants in analytical chemistry, ranging from the need to understand speciation in developing many analytical procedures to the prediction of speciation of metal ion complexes in the environment. Clearly, it is not possible to determine the formation constants of complexes of all possible ligands and metal ions, particularly the ternary complexes that arise in complex mixtures, so that methods for the estimation of unknown formation constants will be an important aspect of modelling such systems.One aspect that was not touched on in their review, however, was an evaluation of the reliability of the different methods presented that was independent of the claims of the proponents of the methods.The reader was left with the impression that all the methods are reliable and work equally well. Three important points in particular were not touched on. (1) Most empirical methods can be made to reproduce reasonably well the data used in the fitting of some equation or correlation that is expected to predict formation constants. The test of a method of prediction, however, is the performance of the method with regard to formation constants that were not used in fitting the equation or correlation. This extends also to formation constants that are not currently known, or are perhaps even unmeasurable for reasons such as hydrolysis or insolubility.One has to ask the question, ‘Is this prediction consistent with the known chemistry of this metal ion?’ The latter point is well illustrated2,3 by the complexes of metal ions with ammonia and polyamine ligands. For example, BiIII had until recently no known solution chemistry with the latter ligands.4–7 Do methods for estimating formation constants predict logK1 with ammonia and BiIII consistently with the absence of chemistry of the metal ion with such ligands? For an ammonia complex to be resistant to hydrolysis, it should not be decomposed by water according to the equation M(NH3)n++H2O?M(OH)(n21)++NH4 + (1) In eqn. (1), with pKw = 14.0 and pKa for NH3 = 9.22,8 logK1 (NH3) for any metal ion must not be significantly less than logK1(OH2) by more than 14 2 9.22 = 4.78 log units for there to be detectable amounts of the ammonia complex present in solution. Any method of prediction that suggests that logK1(NH3) for a metal ion is large enough for it to survive hydrolysis according to eqn.(1), which metal ion is known to have no aqueous chemistry with sp3 hybridized nitrogen donor ligands, must therefore come under suspicion. Most of the methods of prediction described in the review by Dimmock et al.1 fail this type of test, as discussed below, in that they predict that many metal ions that form no complexes with ammonia should have a rich chemistry with such nitrogen donor ligands.2 For example, the Edwards equation9 predicts that logK1(NH3) for PbII is 6.5.If this were correct, PbII [logK1(OH2) = 6.38] would form very stable complexes with ammonia in aqueous solution. A subsequent study10 has shown that, in fact, logK1(NH3) for PbII is 1.6, which complex can only barely be stabilized against hydrolysis in 5 m NH4 + ion.The intention here is not to state that all these methods of prediction are without value, but simply that the user should be aware that they may be deficient if extended to ligand types that are different from those used in fitting the equation. (2) Have the proponents of methods of prediction of unknown logK1 values actually made any effort to validate their methods of prediction by checking predictions that their Rob Hancock obtained his BSc at Rhodes University, and PhD at the University of Cape Town.He carried out research in ligand design at the National Institute for Metallurgy in Johannesburg, followed by a period at the University of the Witwatersrand as Professor of Inorganic Chemistry. He is currently carrying out research at IBC Advanced Technologies in the USA. He is author of 185 research papers, and a book ‘Metal Complexes in Aqueous Solutions’, co-authored with A. E. Martell. Current interests include ligand design for analysis of metal ions by fluorescence spectroscopy, and complexes for positron emission tomography.Analyst, April 1997, Vol. 122 (51R–58R) 51Rmethods make, particularly those that are not of a more limited nature? Although these are obviously useful, one would regard as limited the use of LFER (linear free energy relationships) between two very similar metal ions, e.g., CoII and NiII, with a set of very similar ligands, e.g., bidentate ligands containing two negative oxygen donors, such as oxalate, malonate and catecholate (see Fig. 1 for ligand abbreviations). Much more significant is the prediction of logK1 for a complex where no currently known chemistry would allow for a guess at the logK1 value. Validation of such predictions renders the method of prediction much more trustworthy. One should point out here for readers who are not familiar with the determination of formation constants, and who may feel that checking constants is an inherently inaccurate process, that with use of techniques such as glass electrode potentiometry and modern computer programs, it is usual for experienced researchers to achieve accuracy of formation constant determination of better than 0.1 log unit.Disagreements in the literature between reported values for the same constant can be traced8 to such factors as the use of impure ligands and application of incorrect models in analyzing the potentiometric data. One can usually discern whether a constant is likely to be accurate8 by reading the paper reporting it, and checking off 10 criteria that indicate the quality of the experimental work.(3) An important aspect of the prediction of formation constants is the extent to which methods are purely empirical. The LFER is an important method of estimation of unknown logK values. However, if used in a purely empirical way, without knowledge of accompanying theories on how structural and bonding factors control complex stability, misleading predictions can be made. This is shown in Fig. 2, where logK1 for PbII complexes with polyaminocarboxylates is plotted against logK1 for the corresponding complexes with ZnII. Fig. 2 contains all the data points that are available for these two metal ions with ligands of this type in the compilation of Martell and Smith.8 If we suppose that logK1 for TMDTA with PbII had not been known, and we were to use the LFER in Fig. 2 to estimate this logK1 value, the remainder of the points present an impressive correlation with a correlation coefficient of 0.98, and we would be confident that our prediction of logK1 for PbII with TMDTA of 16.5 was correct within likely experimental error.However, the experimental value8 is 13.7. This is probably not an acceptable level of accuracy of prediction for most purposes. How can this single point be such an outlier? The answer is that it involves an important structural feature not present in any of the other ligands, i.e., that TMDTA contains a six-membered chelate ring, where the other ligands all contain five-membered chelate rings.One has to be aware of theories regarding the role of chelate ring size11–13 and metal ion size in complex stability before one might expect such an effect. Purely empirical approaches without accompanying insight into theories of complex formation run the risk of giving incorrect predictions. Fig. 1 Ligands discussed in this paper. 52R Analyst, April 1997, Vol. 122The more a method of prediction is grounded in theory, the less likely it is that it will make highly erroneous predictions. These three aspects of prediction of stability constants are dealt with below in relation to unidentate ligands first of all, followed by a discussion of the role of architectural features in multidentate ligands and how these relate to complex stability and its prediction. How these factors might all be built into a computer program which is an expert system, i.e., able to function as an aid to ligand design, is discussed.Monodentate Ligands The obvious patterns present in the formation constants of complexes of monodentate ligands have led14 several workers to propose classification schemes15–17 and equations2,3,9,17–23 for the prediction of formation constants. All these schemes can be summarized in the HSAB (Hard and Soft Acids and Bases) classification of Pearson.17 The HSAB classification is based on the preferences of metal ions for ligands with different donor atom types, as indicated by their formation constants. Thus, for example, complexes of soft metal ions have stability orders with halide ligands F2 << Cl2 < Br2 < I2, while complexes of hard metal ions have stability orders F2 >> Cl2 > Br2 > I2.To model these differing stability sequences, Edwards proposed a dual parameter equation:9 log(K1/K0) = aEn+bH (2) This equation is typical of most of the equations that have been used to predict formation constants in an attempt to mimic HSAB behaviour. A pair of parameters is used, one of each of which relates to the metal ion, in this case a and b, and the other pair to the ligand, in this case En and H.One pair of parameters represents the tendency to covalence or softness (a and Ea) in the M–L bond, while the other represents the tendency to hardness or ionicity (b and H) in the M–L bond.The limitations and inaccuracy of some of the predictions made by this type of equation have been discussed elsewhere.2,3 We shall confine ourselves here to discussing the predictions made by the ‘E and C’ equation of Drago, adapted and extended2,3 to apply to aqueous solutions, and the equations of Brown and coworkers21 –23 (the BSE equations), which take a completely different type of approach to that adopted by other, more empirical, approaches. In their review, Dimmock et al.1 gave a very comprehensive account of the BSE equations, so that only a brief description will be given here.Brown and co-workers developed the BSE equations which predicted logK1(OH2) for a large number of metal ions using physical properties of the ions such as charge and ionic radius and factors such as the protonation constant and a quantity defined as the electronicity of the ligand, and so on. This worthwhile approach predicts logK1(OH2) with considerable accuracy, and has the potential to predict formation constants for metal ions such as Lr3+, about which very little is known.These authors also developed parameters for ammonia complexes, and here it appears that they did not thoroughly examine the implications of their predictions. The BSE equations can be used to predict logK1(NH3) for all metal ions for which parameters are available. Many of these predictions fail the test of eqn. (1). Thus, for example, La3+ is predicted to form a complex with ammonia with logK1(NH3) = 2.55, which means that La3+ [logK1(OH2) = 5.38] should have a rich chemistry of ammonia complexes, and also with polyamines, similar to that of ZnII [logK1(NH3) = 2.1, logK1(OH2) = 5.08], which is not the case.One can see from Fig. 3 that the parameters in the BSE equation as currently developed basically model logK1(NH3) as 0.6 logK1(OH2). In Fig. 3, the values of logK1(NH3) predicted by the BSE equation for metal ions are plotted against their experimental8 values of logK1(OH2), and this plot has a slope of approximately 0.6.Clearly, the parameters for NH3 are similar to those for OH2, which later discussion shows should not be the case. Fig. 2 LFER of logK1 for PbII complexes plotted against logK1 for the corresponding complexes of ZnII. Ligands have been limited to aminocarboxylates, i.e., those containing sp3 hybridized nitrogens and N-acetate groups, or other similar donor groups such as phenolates.Ligands: 1, glycine; 2, N,N-bis(2-hydroxyethyl)alanine; 3, IDA, iminodiacetate; 4, MIDA, N-methyliminodiacetate; 5, EDMA, ethylenediaminemonoacetate; 6, N-2-methoxyethyliminodiacetate; 7, HIDA, 2-hydroxyethyliminodiacetate; 8, N-tetrahydropyranylmethyliminodiacetate; 9, ethylenediaminedi- 2-propionic acid; 10, ethylenediaminediserine; 11, 1,6-dicarboxypiperidine- N-acetate; 12, NTA, nitrilotriacetate; 13, EDDA, ethylenediamine-N,NA-diacetate; 14, N-2-aminoethyliminodiacetate; 15, EGTA, 4-oxa-1,7-diazaheptane-N,N,NA,NA-tetraacetate; 16, HEDTA, 2-hydroxyethylethylenediaminetriacetate; 17, EEDTA, 4,7-dioxa- 1,10-diazadecane-N,N,NA,NA-tetraacetate; 18, EDTA; 19, HBED, N,NAbis( 2-hydroxybenzyl)ethylenediaminediacetate; 20, DTPA, diethylenetriaminepentaacetate; and 21, CDTA, trans-1,2-diaminocyclohexanetetraacetate.The point for TMDTA (trimethylenediaminetetraacetate) should be noted, as discussed in the text, in that it falls well off the correlation.Formation constants from ref. 4. Fig. 3 Values of logK1(NH3) predicted (5) by the BSE equation of Brown and co-workers,21–23 plotted against experimental8 values of logK1(OH2). The dotted lines to the open circles show the positions of the experimentally known values of logK1(NH3) or those which have been estimated as discussed later, shown in Fig. 5. The point that the figure makes is that the parameters currently employed in the BSE equation to predict logK1(NH3) are too like those for predicting logK1(OH2), and do not distinguish these ligands in terms of HSAB trends. In other words, metal ions that have high logK1(OH2) should also have high logK1(NH3), which is not found to be so, as discussed in the text.Analyst, April 1997, Vol. 122 53RMany predictions of logK1(NH3) by the BSE equation fall into a gray area, as far as validation by eqn. (1) is concerned. The prediction for BiIII of logK1(NH3) is 7.86, which with logK1(OH2) of 12.98 means that ammonia complexes will tend to be hydrolyzed, in agreement with observation. This is substantially different from the value of logK1(NH3) of 5.0 predicted by the equation of Hancock and Marsicano.2,3 Complexes of BiIII with ammonia do not exist in appreciable quantities in aqueous solution, so how can one decide which is the more reliable estimate? Two approaches exist that allow us to make a judgement.Complexes of Polyamine Ligands The affinity of metal ions for ammonia parallels fairly closely their affinity for polyamine ligands. A major difference, however, is that for polyamine ligands the complexes may be sufficiently stabilized by the chelate effect24 to be stable in aqueous solution, where the ammonia complexes are not.Thus, complexes of PbII with ammonia are easily hydrolyzed,10 whereas those with polyamines such as dien or trien are not. One may develop theoretical equations25 that relate the stability of complexes of n-dentate polyamines to those of their ammonia complexes: logK1(polyamine) = 1.152.logbn(NH3) + (n21)log55.5 (3) The factor of 1.152 models the greater inductive effects of the primary to tertiary amines present in polyamines (pKa Å 10.6) compared with ammonia (pKa 9.22) and the (n21)log55.5 term is the entropic contribution25 to the chelate effect.Eqn. (3) can be further extended to predict logK1 for aminocarboxylate ligands such as EDTA from the known values of logK1 for ammonia and acetate.The logK1 value for BiIII with EDTA8 suggests that logK1(NH3) with BiIII is 5.0. This prediction suggested that BiIII should have an extensive chemistry with polyamine ligands, both open-chain ligands such as dien and trien and azamacrocycles such as cyclen and cyclam. We have made an extensive effort to check the predictions of logK1 for BiIII with polyamine ligands. This has resulted in polarographic studies of these complexes, with the results appearing to be more consistent with logK1(NH3) for BiIII = 5.0 than the 7.86 predicted by the BSE equations.In fact, eqn. (3), modified25 to include estimates of the stepwise decrease in logKm(NH3) values as m increases 1, 2, . . ., n, predicts logK1 for the dien and trien complexes of BiIII to be 17.7 and 22.3, respectively, in reasonable agreement with the experimental values in Table 1. The predicted value of logK1 for BiIII with NH3 suggested a previously unsuspected chemistry of BiIII with azamacrocycles. This led to the synthesis7 of the complex of BiIII with cyclen, the structure of which is shown in Fig. 4, and of a complex of BiIII with THP–cyclen.6 We have validated predictions of logK1(NH3) made2,3 by eqn. (4) below for complexes of several metal ions such as LaIII,26 [VO]2+ complexes,27 complexes of GaIII and InIII,28 [UO2]2+ complexes29 and PuIV complexes30 by studying complexes of polyamines and more hydrolysisresistant complexes of amine ligands such as THEEN, BIPY and AMPY.LFER of LogK1 for N-donor Ligands and O-donor Ligands and LogK1(NH2) It has been pointed out10 that the process where a ligand such as IDA replaces a ligand such as ODA resembles the formation of an ammonia complex in that an sp3 hybridized O-donor is being replaced by an sp3 hybridized N-donor in both cases (see Scheme 1). Thus, for several ligand pairs where there is a neutral oxygen donor in one member of the pair in place of a neutral nitrogen donor in the other member of the pair, a plot such as that in Fig. 5 may be drawn. Fig. 5 shows an LFER of logK1(IDA) 2 logK1(ODA) versus logK1(NH3) for a variety of metal ions. The correlation includes experimental points for logK1(NH3) and also values estimated by the combined use of eqns. (3) and (4) together with other correlations similar to that shown in Fig. 5. The degree of agreement between logK1(NH3) values predicted by eqn. (3) and by correlations such as that shown in Fig. 5, and also eqn.(4), are encouraging, and suggest that the estimates are substantially correct. Accuracy of the ‘E, C and D’ Equation The following multiparameter equation, which is of the type reported by Drago et al.,31 is typical of all such multiparameter equations, and is here modified2,3 to predict formation constants in aqueous solution rather than heats of complex formation in solvents of low dielectric constant: logK1 = EAEB + CACB (4) where the E parameters represent the ionic contribution to the M–L bond and the C parameters represent the covalent contribution for the Lewis acid A and base B.This equation was used successfully to reproduce all known formation constants of complexes of F2, OH2 and NH3. The fitting of the equation2,3 included experimentally inaccessible logK1(NH3) values estimated by independent methods,10,25 as described above. The parameters in an equation such as eqn. (4) can be set31 arbitrarily, or they can be set with some physical model in mind.Since logK1(F2) and logK1(OH2) are known for virtually all metal ions,8 and considerable evidence suggests that M–F bonds are essentially ionic, EB for F2 was set as 1 and CB as zero, with effectively no covalence in the M–F bond, and EA for each metal ion = logK1(F2). Since M–OH bonds are certainly more covalent than M–F bonds, CA for each metal ion was set as logK1(OH2)/14. A best fit gives EB = 21.08 and cB = 12.34 for ammonia. Rearrangement of eqn. (4) gives Table 1 LogK1 values for various complexes.Formation constants from refs. 5–8 Ligand Parameter NH3 ampy dien trien cyclen LogK1 (BiIII) (5.0)* 9.4 17.5 21.9 23.5 LogK1 (CuII) 4.1 9.5 16.0 20.1 23.3 LogK1 (HgII) 8.8 (12.0)† 21.8 24.8 25.5 LogK1 (PbII) 1.6 3.95 7.5 10.4 15.9 * Estimated,2,3 as discussed above. † Estimated14 as 1 2 [logK1(BIPY) + logK1(en)]. Fig. 4 Structure of the complex of BiIII with cyclen [Bi(cyclen)- (ClO4)3H2O], as determined by X-ray crystallography.7 That BiIII would have an extensive chemistry with N-donor ligands was based on the prediction2,3 of logK1(NH3) by eqn.(4). Three of the four coordinated oxygens are derived from perchlorate anions and the fourth is from a coordinated water molecule. 54R Analyst, April 1997, Vol. 122logK1(NH3) = 0.881[logK1(OH2)]21.08[logK1(F2)] (5) Eqn. (5) indicates that for a constant value of logK1[OH2], as logK1(F2) increases for a metal ion so its value of logK1(NH3) must decrease.This is the very essence of the HSAB idea. There is no one order of acid strength for metal ions. Rather, a soft metal ion such as HgII will have a high logK1(NH3) and low logK1(F2), whereas a hard metal ion such as ThIV, which has a similar affinity to HgII for the hydroxide ion, will have the reverse situation (Table 2). Eqn. (5) can be tested graphically, as shown in Fig. 6, where logK1(NH3) values, both experimental and predicted by other means,10,25 are plotted against 0.881[logK1(OH2)] 2 1.08[logK1(F2)]. Fig. 6 is a graphical representation of HSAB ideas. Eqn. (4) can be successfully fitted to logK1 values only for F2, NH3 and pyridine and unidentate ligands containing a negative oxygen donor such as OH2, phenols or carboxylic acids. It cannot deal with ligands with donor atoms from the third or lower periods in the Periodic Table, such as Cl, Br, I, S, Se, P or As. To do so, a third pair of parameters, the D parameters, has to be included, as in the equation logK1 = EAEB + CACB2DADB (6) The DA parameters for the metal ions decrease as the metal ion increases in size, whereas the DB parameters for the ligand increase as the ligand donor atom increases in size.The D parameters have been interpreted2,3 as reflecting the steric strain that arises when ligands with large donor atoms replace the small water molecule from the coordination sphere of a metal ion, and also may reflect the loss of ability to stabilize the complex by hydrogen bonding between ligands coordinated to the metal ions and the solvent water.Thus, for large metal ions such as Ag+, Hg2+ and Pb2+, which have ionic radii in excess of 1.0 Å, DA is zero, and is at a maximum for the small proton. The accuracy of prediction obtained with use of eqn. (6) has already been discussed.2,3 In the original papers, predictions of logK1 were included for those ligands and metal ions examined whose logK1 values were unknown at the time of publication.Since then, logK1 values have been published8 for the azide complexes of some of these metal ions, and the predicted2,3 and subsequently observed logK1 values can be compared (Table 3). The most recent set of E, C and D parameters for use in eqn. (6) appear in ref. 14. Eqn. (6) is philosophically not very different from those multiparameter equations that have preceded it.9,17–20 The superiority of eqn. (6) at predicting logK1 for unidentate ligands derives from the care with which we have developed the parameters in it and the extensive studies which have been aimed at validating its predictions. No doubt eqn.(6) can be improved further, and possibly the parameters in it can be derived theoretically rather than empirically. The BSE equation21 –23 represents a good start at developing a more theorybased approach to the prediction of logK1 values, and possibly Fig. 5 Estimation10 of logK1(NH3) for metal ions where ammonia complexes do not exist in aqueous solution because of hydrolysis.The solid circles are for experimental logK1 values and the open circles are for metal ions for which the ammonia complexes do not exist in aqueous solution, and for which logK1(NH3) has been estimated by a number of correlations of the type shown here. The correlation is for logK1 for the IDA complex minus logK1 for the ODA complex versus logK1(NH3), as discussed in the text. Formation constants from ref. 8. Table 2 LogK1 values for HgII and ThIV LogK1(F2) LogK1(OH2) LogK1(NH3) HgII (soft) 1.5 10.6 8.8 ThIV (hard) 8.44 10.8 0.4* * Estimated from eqn. (5) and Fig. 5; other logK1 values from ref. 8. Scheme 1 Analyst, April 1997, Vol. 122 55Rcould be refined to predict more accurately the formation constants of ligands such as ammonia, and also ligands with heavy donor atoms, taking into account steric considerations that appear to be important to the success of eqn. (6). Multidentate Ligands Equations such as eqn.(3), and its extension to polyaminocarboxylates, have been shown25 to predict the formation constants of polyamines such as dien and trien and aminocarboxylates such as NTA and EDTA very well, provided that the ligands form only five-membered chelate rings on complex formation. The question that arises is how to extend equations such as eqn. (3) to ligands that contain other architectural features, such as six-membered chelate rings, or neutral oxygen donors, C-alkyl substituents or macrocyclic rings.The way in which this can be done is simple. It has been shown14 that many such structural changes produce changes in complex stability that are related to metal ion radius.32 Thus, replacement of a five-membered chelate ring with a six-membered chelate ring will increase the selectivity for smaller relative to larger metal ions,11,33 as will addition of C-alkyl groups34 (by selectivity in this review is meant the difference in logK1).Addition of groups containing neutral oxygen donors11 will shift the selectivity in favour of large metal ions. In Fig. 7–9 are shown examples of these effects, where the change in logK1, DlogK, caused by the structural changes indicated, is plotted as a function of ionic radius.32 In Fig. 7 is shown the effect on metal ion size-based selectivity of an increase in chelate ring size from all fivemembered in the NTA complex to all six-membered in the NTP complex, or one five- and two six-membered chelate rings in NADP. The small BeII ion actually shows a small increase in logK1 on increase of chelate ring size, but as the metal ion size increases further there is a steady decrease in logK1 for the NTP or NADP complex relative to the NTA complex.The type of relationship in Fig. 7 is general for a wide range of metal ions, and is best illustrated by the relationship involving EDTA and TMDTA.11 in Fig. 8 is shown the relationship between metal ion size and the effect on complex stability of adding C-methyl substituents to EDTA.It is seen that the addition of two methyl groups to the ethylene bridge between the two nitrogens of EDTA causes a metal ion size related change in complex stability such that the complexes of larger metal ions are stabilized less than those of smaller metal ions. A similar effect is produced by adding isopropyl groups to two acetates of EDTA to give DP-EDTA in Fig. 8. The difference is that the small methyl groups produce general stabilization, whereas the bulky isopropyl groups produce destabilization of the complexes.Fig. 6 Plot demonstrating the relationship between log K1 for the F2, OH2 and NH3 complexes of metal ions. The relationship is a graphical validation of equations (4) and (5). Experimental formation constants (5) from ref. 8, or log K1(NH3) estimated (2) as described in the text, with log K1 for OH2 and F2 experimental. Table 3 Comparison of predicted and observed logK1 values Metal ion LogK1(azide) GaIII InIII UO2 2+ LaIII Predicted2,3 4.28* 4.31 2.10 1.33 Observed8† 4.40 4.09 2.14 1.52 * The E and C parameters originally reported2,3 for GaIII have since been revised,14 owing to a better value8 of logK1(OH2) for GaIII.The original E, C and D parameters for GaIII gave logK1(azide) = 3.84, still a reasonable level of agreement. † Experimental logK1 values from ref. 8, corrected to ionic strength zero. Fig. 7 Effect of chelate ring size on complex stability. As shown in structure (a), a five-membered chelate ring of the ethylenediamine type forms11–13 with least steric strain with a large metal ion of M–N bond length 2.5 Å and N–M–N bond angle 69°. In structure (b) it is seen that lowest steric strain occurs for a six-membered chelate ring of the 1,3-diaminopropane type when the coordinated metal ion is small with an M–N length of 1.6 Å and an N–M–N angle of 109.5°. A consequence of the metal ion size preference of chelate rings of different sizes is shown in (c), where logK1(NTP) 2 logK1(NTA) (5) is plotted as a function of metal ion radius.32 NTP = nitrilotripropionate and NTA = nitrilotriacetate.Also shown is the plot of logK1(NADP) 2 logK(NTA) (2) as a function of metal ion radius. NADP = nitriloacetatedipropionate. The correlations in (c) show how increasing metal ion size causes a decrease in logK1 as the chelate ring size increases from all five-membered in NTA to all six-membered in NTP, or one five-membered and two six-membered in NADP.Formation constants from ref. 8. 56R Analyst, April 1997, Vol. 122The addition of groups bearing neutral oxygen donors can also be viewed as an architectural change in the ligand. In the precursor complex, e.g., the en complex shown in Fig. 9, the remaining coordination sites are filled with water molecules. Thus, adding connecting bridges to the en complex to give either THEEN or 18-ane-N2O4 as shown can be viewed as merely adding connecting bridges between the N-donors and Odonors on the complex. Fig. 9 shows the metal ion size-related change in complex stability that occurs on adding neutral oxygen donor groups to en complexes to give complexes of THEEN, of 18-aneN2O4 and cryptand-222. It may turn out that all purely architectural changes in complex structure produce changes in complex stability that are directly related to metal ion radius, which seems to be the case from the available8 formation constant data.One can thus further modify eqn. (3) to predict unknown formation constants simply by adding a further term that models the slopes and intercepts of diagrams such as Figs. 7–9. We have devised a computer program, LOGKEST, that runs on IBM PC compatible machines, based on the ‘E, C and D’ equation [eqn. (6)], which predicts formation constants for unidentate ligands. These predictions are then used in equations such as eqn.(3) to predict the formation constants of ligands such as trien or NTA. For known ligands the effects on complex stability of architectural features such as larger chelate rings, addition of neutral oxygen donors, the presence of macrocyclic or cryptate structures or C-alkyl substituents can be corrected for from a simple term involving the ionic radius of the metal ion, which models the slopes and intercepts of diagrams such as Figs. 7–9. As input for a metal ion, the data required to be known are ideally logK1 for the F2 and OH2 complexes, and the ionic radius, while knowing logK1 for one polydentate ligand such as EDTA is useful.From this simple input the program can make reasonable predictions of logK1 for a great number of ligands for the metal ion, ranging from unidentate ligands such as thiosulfate or cyanide, through ligands of intermediate complexity such as trien or EDTA, to ligands of considerable complexity such as cyclam or cryptand-222.The three steps that LOGKEST uses to predict a formation constant for a complex are as follows: (1) The ‘E, C and D’ equation [eqn. (6)] is used to predict logK1 for complexes of unidentate ligands. (2) For non-macrocyclic multidentate ligands that form only five-membered chelate rings on complex formation, the chelate effect equations such as eqn. (3) are used together with the predicted values for the unidentate groups from step 1 to predict logK1 for the complex of the multidentate ligand.(3) For ligands that have other architectural features present, such as chelate rings of sizes other than five-membered, C-alkyl substituents, groups bearing neutral oxygen donors and macrocyclic rings, further corrections are made to the predictions in step 2. This involves two parameters that represent the slope and intercept of relationships such as those shown in Figs. 7–9. These are used to derive formation constants from the formation constants calculated for the analogous ligands containing fivemembered chelate rings only.Fig. 8 The effect of C-alkyl substituents on complex stability. At (a) above is shown diagrammatically the way in which a small metal ion causes a large curvature in a coordinated ligand, so that C-alkyl substituents are well separated, while at (b) is shown a ligand coordinated to a large metal ion where the small curvature of the ligand pushes the C-alkyl substituents closer together. The result of this effect is shown at (c), where it is seen that addition of C-alkyl substituents to EDTA to give DMEDTA (2) or DPEDTA (5) produces a change in complex stability that is strongly related to metal ion size.The DlogK values at logK1(EDTA) 2 logK1(DMEDTA), or logK1(EDTA 2 logK1 DP-EDTA). Formation constants from ref. 8. Fig. 9 Effect of substituents bearing neutral oxygen donors on complex stability. In (a) it is shown how the changes in going from an en complex to a THEEN or 18-ane-N2O4 complex can be regarded as a structural change brought about by inserting ethylene bridges between nitrogens and coordinated water molecules.In (b) it is shown how the addition of neutral oxygen donors produces a change in complex stability such that the complexes of larger metal ions are stabilized relative to those of smaller metal ions. The effect on complex stability is similar whether an open-chain ligand or a macrocycle is formed by the structural change in the ligand.Ionic radii from ref. 32 and stability constants from ref. 8. Analyst, April 1997, Vol. 122 57RA possible limitation of LOGKEST at present in some cases may lie in predicting the response that will be obtained to changes in ligand architecture in step 3 above. If it were possible to predict the metal ion size-related changes that would occur for unknown ligands, it would then be possible to make LOGKEST an expert ligand design system, where reasonable predictions of logK1 could be made for ligands that were designed on the computer screen.Since these metal ion sizerelated changes in complex stability are steric effects, it would seem that the best approach to predicting them without recourse to synthesis of the ligand and measurement of some stability constants would be molecular mechanics (MM) calculations.12 This aspect of ligand design is currently being explored. A further important aspect that has not been touched on here is that of higher complexes, MLn where n > 1, and ternary complexes, MLaLb, where La and Lb are different ligands, which should be the subject of future investigations. Clearly, this is a subject of greater complexity than that of a single ligand attaching to a metal ion, as the two separate ligands may coordinate in a variety of ways with many different steric possibilities, and again here MM analysis should be a useful tool. The author thanks the University of the Witwatersrand and the Foundation for Research Development for generous financial support for this work. References 1 Dimmock, P. W., Warwick, P., and Robbins, R. A., Analyst, 1995, 120, 2159. 2 Hancock, R. D., and Marsicano, F., Inorg. Chem., 1978, 17, 560. 3 Hancock, R. D., and Marsicano, F., Inorg. Chem., 1980, 19, 2709. 4 Hancock, R. D., Cukrowski, I., and Mashishi, J., J. Chem. Soc., Dalton Trans., 1993, 2895. 5 Hancock, R. D., Cukrowski, I., Cukrowska, E., Antunes, I., Mashishi, J., and Brown, K., Polyhedron, 1995, 14, 1699. 6 Luckay, R., Reibenspies, J. H., and Hancock, R. D., J. Chem. Soc., Chem. Commun., 1995, 2365. 7 Luckay, R., Cukrowski, I., Mashishi, J., Reibenspies, J. H., Bond, A. H., Rogers, R. D., and Hancock, R. D., J. Chem. Soc., Dalton Trans., 1997, 901. 8 Martell, A. E., and Smith, R. M., Critical Stability Constants, Plenum Press, New York, 1974–89, vols. 1–6. 9 Edwards, J. O., J. Am. Chem. Soc., 1954, 76, 1540. 10 Mulla, F., Marsicano, F., Nakani, B. S., and Hancock, R. D., Inorg. Chem., 1985, 24, 3076. 11 Hancock, R. D., Pure Appl. Chem., 1986, 58, 1445. 12 Hancock, R. D., Prog. Inorg. Chem., 1989, 37, 187. 13 Hancock, R. D., Acc. Chem. Res., 1990, 23, 253. 14 Martell, A. E., and Hancock, R. D., Metal Complexes in Aqueous Solutions, Plenum Press, New York, 1996. 15 Schwarzenbach, G., Adv. Inorg. Radiochem., 1961, 3, 257. 16 Ahrland, S., Chatt, J., and Davies, N. R., Q. Rev. Chem. Soc., 1958, 12, 265. 17 Pearson, R. G., Chem. Br., 1967, 3, 103. 18 Yingst, A., and McDaniel, D. H., Inorg. Chem., 1967, 6, 1067. 19 Yamada, S., and Tanaka, M., J. Inorg. Nucl. Chem., 1975, 37, 587. 20 Misono, M., and Saito, Y., Bull. Chem. Soc. Jpn., 1970, 43, 3680. 21 Brown, P. L., Sylva, R. N., and Ellis, J., J. Chem. Soc., Dalton Trans., 1985, 723. 22 Brown, P. L., and Sylva, R. N., J. Chem. Res, 1987, (S) 4; (M) 0110. 23 Brown, P. L., Talanta, 1989, 3, 351. 24 Schwarzenbach, G., Helv. Chim. Acta, 1952, 35, 2344. 25 Hancock, R. D., and Marsicano, F., J. Chem. Soc., Dalton Trans., 1976, 1096. 26 Hancock, R. D., Jackson, G. J., and Evers, A., J. Chem. Soc., Dalton Trans., 1979, 1384. 27 Duma, T. W., and Hancock, R. D., J. Coord. Chem., 1994, 31, 135. 28 Duma, T. W., Marsicano, F., and Hancock, R. D., J. Coord. Chem., 1991, 23, 221. 29 Jarvis, N. V., de Sousa, A. S., and Hancock, R. D., Radiochim. Acta, 1992, 57, 33. 30 Jarvis, N. V., de Sousa, A. S., and Hancock, R. D., Radiochim. Acta, 1994, 64, 15. 31 Drago, R. S., Vogel, G. C., and Needham, T. E., J. Am. Chem. Soc., 1971, 93, 6014. 32 Shannon, R. D., Acta Crystallogr., Sect. A, 1976, 32, 751. 33 Hancock, R. D., and Martell, A. E., Chem. Soc. Rev., 1989, 89, 1875. 34 Hancock, R. D., Maumela, H., and de Sousa, A. S., Coord. Chem. Rev., 1996, 148, 315. Paper 6/07993H Received November 26, 1996 Accepted January 16, 1997 58R Analyst, April 1997, Vol. 122
ISSN:0003-2654
DOI:10.1039/a607993h
出版商:RSC
年代:1997
数据来源: RSC
|
3. |
Five-way ANOVA Interaction Analysis of the Selective Extraction ofCarbaryl, Pirimicarb and Aldicarb From Soils by Supercritical FluidExtraction |
|
Analyst,
Volume 122,
Issue 4,
1997,
Page 303-308
Iain A. Stuart,
Preview
|
|
摘要:
Five-way ANOVA Interaction Analysis of the Selective Extraction of Carbaryl, Pirimicarb and Aldicarb From Soils by Supercritical Fluid Extraction† Iain A. Stuart*a, Ray O. Ansella, John Maclachlana, Peter A. Bathera and William P. Gardinerb a Department of Physical Sciences, Glasgow Caledonian University, Cowcaddens Road, Glasgow, UK G4 0BA b Department of Mathematics, Glasgow Caledonian University, Cowcaddens Road, Glasgow, UK G4 0BA Due to the biopersistence of organophosphate and organochloride compounds, the hydrolytic degradation of the carbamate insecticides has proved attractive in the reduction of persistent insecticides in the biosphere and food chain.Their susceptibility to hydrolysis, however, can complicate their analysis and care is required in the selection of the extraction conditions and the analytical technique employed. The work described here is an investigation into method optimisation in the extraction and final analysis of selected carbamate insecticides from selected soils. A method using the inert extraction medium of supercritical carbon dioxide has been developed for determination of three carbamates relevant to the soft fruit growing industry (carbaryl, aldicarb and pirimicarb).Determinations were completed using HPLC-postcolumn reaction-fluorescence with orthophthalaldehyde-mercaptoethanol derivatisation. The resultant methylisoindole fluorophore was detected at lex : 330 nm and lem : 450 nm with pirimicarb detection at lex : 315 nm and lem : 380 nm.A five variable ANOVA analysis was carried out to determine both the most significant independent factors in the extractions and their most significant statistical interaction(s). Retrospectively optimised extraction conditions were obtained from the ANOVA where many of the carbamates were successfully extracted from each soil using CO2 at 300 atm modified with 10% dimethyl sulfoxide. The mean recoveries obtained were 91.5–107.8% for all carbamates from many soils.Keywords: Pesticides; supercritical fluid extraction; ANOVA; matrix effects In the UK, the Food and Environmental Protection Act 1985 Part III and the Control of Pesticides Act 1986 were designed to protect the health of consumers, the biosphere of the surrounding area and to secure safe, efficient and humane methods of pest control. The enforcement of this legislation has led to an increased need from the analyst to provide reliable, effective methods of qualitative and quantitative pesticide residue analysis from environmental and food matrices.Methods of efficiently monitoring concentrations of carbamate insecticides became an issue with the detection of the aldicarb in groundwater which led to the US Environmental Protection Agency (US EPA) proposing a maximum admissible total carbamate concentration of 0.2 mg l21. UK legislation on carbamates is guided by the EC directive on Drinking Water Quality 80/778/EEC which states that the maximum for any single carbamate is 0.1 mg l21.Local enforcement is placed under The Control of Pesticide Regulations 1986 (MAFF 01690). Due to the thermal lability of this class of compounds and consequent incompatibility with GC methods, previous work involving the analysis of carbamates has focused on the use of reverse phase C18 liquid chromatography with both UV and fluorescence methods of detection. The derivatisation of the carbamates using 2,4-dinitrophenyl phenyl ether1 or trichloroacetate2 has also been used in this context with Dorough and Thorstenson3 providing a thorough review on determining carbamates and their metabolites by GC.Previous work on alternative methods has included mass spectrometric detection (HPLC–thermospray-MS)4 and ‘spot-test’ colorimetric reactions in off-line determinations.5–13 As many of the carbamates used in crop protection possess natural fluorescence, they can be detected successfully without further derivatisation.Increased detection limits for the carbamates have been achieved by using postcolumn reaction fluorescence (PRF) through the formation of a highly fluorescent N-methylisoindole derivative (solely for N-methylcarbamates) and has been extensively studied since its introduction in the 1970s.14–16 Extensive reviews of HPLC reaction detectors and HPLC analysis of carbamates by Frei17 and Sharp et al.,18 respectively, can be used for further reference.The quantification of pesticides such as the carbamates in solid matrices has provided further problems in the selection of an appropriate method for extraction and sample preparation that will not contribute to further analyte breakdown. It is here that one must question the use of basic solvents, e.g., triethylamine in high temperature extractions for compounds that are susceptible to hydrolysis. The introduction of analytical- scale supercritical fluid extraction in the mid to late 1980s has provided an efficient, flexible extraction method capable of solvating pesticides of medium to high hydrophobicity directly and those of low hydrophobicity using reaction–extraction protocols in an inert solvent.19 Further to the solubility of the analyte in the supercritical flow, the second main component that affects extraction recovery is the rate of mass transfer from the matrix into the supercritical flow.20 This parameter acts as a function of the strength of the analyte-to-matrix interactions present in the sample and is of greatest interest in many of the studies carried out in extracting pesticides from environmental matrices by SFE.Due to the complexity of interaction between a supercritical fluid and the host matrix, many workers have attempted to use statistical techniques in order to quantify the significance of a particular factor or interaction. Lopez-Avila and Beckert21 investigated several instrumental variables (cell geometry, fluid flow rate and selection of appropriate collection media) for their respective and cumulative effect on analyte recovery of 42 † Presented, in part, at the International Symposium on Supercritical Fluid Chromatography and Extraction, Indianapolis, Indiana, USA, March 31–April 4, 1996.Analyst, April 1997, Vol. 122 (303–308) 303organochlorine pesticides from spiked sand and soil samples (this work was further extended to cover selected OPPs22).Method optimisation was carried out on seven variables each set at low and high to define relative changes in recovery for eight extractions with a second test used to investigate the presence/ absence of a modifer, glass bead sorbent and static extraction. Conclusions made on recovery evidence and %RSDs for the 42 compounds proved the existence of both independent and intervariable dependences. The use of statistical techniques in this way have provided the analyst with a quick, relatively accurate optimisation technique that would otherwise have been costly in both time and materials. In performing a full factorial design for a seven variable system, 27 experiments would have had to have been performed to obtain a set of comparable data trends. Additional optimisation procedures such as multilinear regression (MLR) in experimental design has been used by Kane et al.23 and by van der Velde et al.24 for the extraction of triazines from soil with the objective being again to reduce the number of runs required in method optimisation.Other techniques such as multivariate optimisation25 and two and three level factorial designs26 have also been used with this aim in mind. This work concentrates on developing an interaction model for the extraction of selected carbamates from soil matrices of differing chemical and physical characteristics. The use of ANOVA techniques in this context when used on an exhaustive data set can produce information on the most important fluid characteristics with respect to analyte solvation in the extracting fluid and matrix effects.In this study, five soil samples with four soil matrix parameters, namely: moisture content; pH; particle size; and organic content were measured against variation in pressure, analyte, co-solvent type and % co-solvent added to the fluid. The analytes selected for this study were pirimicarb, aldicarb and carbaryl as they are used exclusively for both crop protection (Fig. 1) and as pest deterrents in both domestic use and the soft fruit growing industry in Scotland.27 Experimental Instrumentation All extractions were completed on an SFE-723M Supercritical Fluid Extraction system (Dionex, Camberley, UK) using two 16 ml capacity extraction cells (Keystone Scientific, Bellefonte, USA) in parallel (in positions 1 and 8 on the manifold fitted with 1200 ml linear restrictors) for each extraction run. 99.99% SFEgrade CO2 was used as the primary solvent supplied with a 110 bar He overpressure (BOC Speciality Gases, Guildford, UK).Organic and moisture content were determined by thermogravimetry (model TG-750/70, Stanton-Redcroft, London, UK) using software developed in-house. Particle sizing was completed on a low angle laser scattering particle size analyser (Malvern, Malvern, UK) with sample introduction by gravity feeding. Extract Assay All chromatographic and modifier solvents used for the determination of the Certified Pesticide Standards (Promochem Ltd.Hertfordshire, UK) were of HPLC-grade (Sigma-Aldrich, Poole, Dorset, UK). The HPLC pump used was a LC 9012 Solvent Delivery System (Varian, Walton-on-Thames, UK) with a 10 ml injection loop and detection by postcolumn reaction fluorescence on a scanning wavelength detector (model 9070, Varian) at lex : 330 nm and lem : 450 nm for carbaryl, carbofuran (internal standard) and aldicarb. Pirimicarb was detected at lex : 315 nm and lem : 380 nm without the presence of reagents.An isocratic mobile phase of 55% water–45% methanol with a flow rate of 1 ml min21 was used in all assays. All determinations were completed on a 150 mm 3 4.6 mm C18 carbamate column at 42 °C (Pickering Laboratories, Mountain View, CA, USA) contained in the postcolumn reaction module (PCX 5100, Pickering Laboratories). Fluorescence reagents, orthophthalaldehyde (OPA), sodium hydroxide hydrolyser (0.4% at 100 °C), OPA diluent (0.3% boric acid) and Thiofluor (N,N-dimethyl-2-mercaptoethylamine hydrochloride) were all of chromatographic-grade (Pickering Laboratories).The monitored extraction parameters are given in Table 1. Soil Sample Preparation Soil samples were taken and left exposed overnight to come into equilibrium with the laboratory atmosphere moisture content. Representative samples of the soils (2 g) were taken using a ‘coning and quartering’ technique and each extraction cell was packed at each end with methanol-cleaned glass wool to prevent end-cell frit blockage.Soil spiking was completed by evaporating 1 ml of a methanolic carbamate mixture (containing 50 mg of each insecticide) directly onto each sample of soil. Evaporation was allowed to occur in the open laboratory atmosphere. Post-extraction, each extract was spiked with the internal standard (carbofuran, 50 mg ml21) and then reduced to 1 ml prior to chromatographic determination. Kratochvil and Peak28 have provided a useful review on the subject of pesticide sampling/handling from a number of sources.Fig. 1 Carbamates used in soft fruit protection Table 1 Monitored extraction parameters Extraction temperature 70 °C Extraction pressures 10 minute cell ‘conditioning’ at 150 atm followed by individual extractions at 200, 300 and 450 atm Total extraction time 40 min Restrictor temperature 70 °C Restrictor volume 1200 ml Flow rate (gas state)* 1650 ml min21 Modifiers used CH2Cl2 (5, 10 and 20%), DMSO (5, 10, 20%), CH3OH (5, 10, 20%) Solvent collection Liquid collection in vial (15 ml) Solvent collection temperature 21 to 1 °C Extraction cell geometry 14 mm3100 mm (16 ml capacity) Cell packing Methanol-cleaned glass wool with celite wet support * Mean flow rate 304 Analyst, April 1997, Vol. 122Results and Discussion In the investigation of extraction conditions, 30 supercritical fluids were used to investigate optimum extraction conditions for the carbamates selected, i.e., three modifiers at three different concentrations and three pressures including pure CO2.Initially, the relative solubility of each carbamate was plotted (Fig. 2) to determine if fluid modification was required and to ascertain at which fluid pressure consistently presented the highest recoveries. This was completed by evaporating 1 ml of the methanolic carbamate mixture onto 2 g of celite prior to extraction. It was found that analyte-to-matrix interactions between the hydrogen bonding of the carbamates and the celite support was not as significant an interaction as in previous studies in polar pesticide extraction. This would then indicate that the pH of the matrix may not be a significant factor in the later studies on soil extractions.Evidence to support this conclusion is shown in the extraction profiles for all three compounds at 300 and 450 atm (200 atm omitted for clarity), Fig. 2, whereby extraction of the compounds with pure CO2 implied very low matrix interaction.The extraction profile also illustrates that using a fluid pressure of 300 atm recovers the carbamates and in particular, pirimicarb, exceptionally well without the presence of fluid polarity modification. Percentage recoveries from the inert substrate in pure CO2 for all three pesticides are consistent with their relative polarity and is a satisfactory initial test of the suitability of the extraction conditions. Further to this, consequent extract reverse phase chromatograms indicate that the order of increasing hydrophobicity is aldicarb, carbaryl, carbofuran (internal standard), pirimicarb.It was also noted that in increasing the fluid density above 300 atm, comparison of individual pesticide extraction profiles indicated a reduction in solubility. As a result, this study provided basic analyte solubility information in the non-polar solvent prior to solvent modification. In selected appropriate co-solvents for the study, dichloromethane (DCM, 3.1), methanol (5.1) and dimethyl sulfoxide (DMSO, 7.2) were selected from their relative solvent strengths, i.e., low to high polarity index.Using these solvents, it is then possible to obtain relevant information on the significance of co-solvent polarity on extraction recovery. Additional issues pertaining to their potential for successful extraction and the benefits of minimal co-extractant throughput to the collection solvent were also considered.Parametric Significance of Soil Types The soils used were chosen from areas characteristically classed as fruit growing land. Soils C1 (forestry) and C2 (forestry drainage) were selected for comparison against soil B (arable land) for organic content significance with soil D (top soil) compared against all four (soil A, coastal soil) remaining soils for moisture content significance. The properties of the soils used are given in Table 2 indicating significant organic and moisture content in all test soils used for SFE extractions.The significance of these factors has been previously shown29,30 as the corresponding organic mass loss can provide a guide to the organic activity (aorg) of the host matrix and so provide a relative strength index of the analyte-to-matrix interaction. pH is used as an indication of humic/organic acid concentration although cation exchange capacity (CEC) has also been used to assess the cation activity in soil matrices.29 The relative percentage moisture content can also be a significant factor in the competition between inherent moisture content and cosolvent strength in the removal of analytes from the matrix.In addition to this issue, particle size has also been measured although found to be comparable in all cases. Experimental Data The experiment was designed to obtain exhaustive statistical information on all possible effects and interactions between the selected variable set (Table 3).As a result, all recovery data effects with respect to fluid pressure, co-solvent and co-solvent type were acquired from using low, medium and high settings. These results were set against each soil and analyte type. Statistical Analysis The use of ANOVA models provides the user with a statistically based technique capable of producing meaningful models on the importance of the studied factors in the experiment.In addition to this, it is also possible to observe possible factor interactions within the data set that may have a bearing in the development of optimised extraction conditions. The use of modelling in this context with SFE has proved highly relevant due mainly to the ability to accurately vary the extracting fluids physical properties. In addition to this, the importance of other parameters such as analyte polarity and matrix properties on extraction recovery are also demonstrated in ANOVA results.A five factor ANOVA analysis was selected to investigate factor effects and interactions with respect to recorded recovery rates and to use a retrospective optimisation technique to obtain the most appropriate extraction conditions for each soil and analyte type. Due to the complexity of the analysis, it was not possible to obtain recordings for certain treatment combinations. These missing entries complicated the use of standard ANOVA procedures with respect to the full data set.To overcome this, the mean values for each treatment combination were treated as the response with missing entries estimated using a weighted average of means for the factor levels corresponding to the missing treatment combinations: n1x – 1 + n2x – 2 + ··· + n5x – 5 Xmean = n1 + n2 + ··· + n5 Fig. 2 Extraction profiles for supercritical fluid solubility optimisation of carbaryl, pirimicarb and aldicarb at 70 °C and at 300 and 450 atm, respectively (rfluid = 0.79 g ml21 at 300 atm).A, 300 atm aldicarb; B, 450 atm pirimicarb; C, 300 atm carbaryl; D, 450 atm carbaryl; E, 300 atm pirimicarb; F, 450 atm aldicarb. Table 2 Properties of investigated soils Mean particle Soil Moisture (%) Organic (%) pH size/mm A 32.06 7.28 4.64 549.1 B 26.44 13.85 4.26 452.5 C1 23.89 4.09 4.80 429.9 C2 24.70 3.23 3.89 432.2 D 14.31 6.75 4.47 558.4 Analyst, April 1997, Vol. 122 305ANOVA assessments of the standard deviations of the treatment combinations was also complicated by missing entries.This was rectified by using the mean of the two known standard deviations for the first four factor combinations as an estimate of the missing value. Use of the full model with all possible effects results in no degrees of freedom for error estimation. To overcome this, a model of the form: SoilýPressureýAnalyteýCo-solventý%Co-solvent-Soil* Pressure*Analyte*Co-solvent*%Co-solvent was fitted to both mean and standard deviation data.The first part of this model refers to the full model with all possible effects (factors and interactions) with the subtraction of the five factor interaction necessary to provide an error term for effect estimation and test statistic construction. Table 3 provides a summary of the ANOVA results, modified to account for missing values, in respect of average recovery. The results show that most effects are significant (P < 0.05) with all main effects and several two and three factor interactions highly significant.For the ANOVA assessment of the standard deviations, only soil type was shown to have any significant effect (P = 0.0287) on the variability in recovery rate (P < 0.05), as might be expected. To maintain succinctness and to avoid repetition, three two factor interaction plots and five three factor interaction plots are included (Fig. 3) which make up the majority of the most statistically important and relevant interactions presented by the data.Four factor interactions are not represented due to graphical complexity. Two Factor Interactions Fig. 3A illustrates the interaction between soil type and each analyte. It has been found that throughout the entire data set, aldicarb is recovered to a greater extent than carbaryl. This contradicts with the relative hydrophobicities of the selected analytes and the initial solubility extractions using unmodified CO2, although pirimicarb, having the greatest log P, is recovered to the highest level.The interaction between each analyte and each fluid pressure (Fig. 3B) has indicated that the optimum extraction pressure/density of the extracting fluid is 300 atm in all cases of modified fluid. This has been shown to be a confirmation of the initial studies for analyte solubility in unmodified CO2 (Fig. 2). Using the co-solvent DMSO (Fig. 3C) has provided the greatest recovery of the analytes at 300 atm pressure, although DCM, which has a lower polarity index value than methanol, recovers the analytes to a higher level than using methanol as the co-solvent. Three Factor Interactions Conclusions taken from two factor interactions can be used as an indicator of relevance of the ANOVA model in the three factor analysis.Fig. 3D exhibits this incremental aspect in the comparison of interaction effects between soil type, pressure and analyte where recovery in all cases is as follows, pirimicarb > aldicarb > carbaryl, as shown in Fig. 3A.The coastal soil A demonstrates the poorest recovery but also possesses the highest moisture content and the second highest organic content. Conformation of the significance of organic and moisture content is shown for the arable land soil B which possesses the second lowest response of the data set. Soil B characteristics include having the highest organic content and the second highest moisture content. In Fig. 3E, the selection of an appropriate co-solvent is made using Fig. 3C as a basis. Recovery from soil B is improved by the use of DMSO. Reproducibility is high for soils B–D although recovery from soil A is reduced. Implying that water has a polarity index value of 10.2, the action of DCM (3.1) will not be curtailed by the presence of water, although co-solvents that are more miscible in water will be subject to the competitive action of water to retain the analyte to the surface of the matrix. This incompatibility is shown in the fact that DCM possesses the highest recovery from most soils with methanol (5.1) next and finally DMSO (7.2).Fig. 3F depicts the optimised concentration of co-solvent in the modified extraction fluid. For the majority of soil types, extractions using co-solvent at a concentration of 10% appear best. In observing soil, analyte and % co-solvent interactions (Fig. 3G), previous problems with the extraction from matrices of higher moisture content is further illustrated here.In this interaction, it is difficult to divorce organic content from moisture content effects in soil C1 and C2 as the former has lower recovery yet lower moisture content than the latter although the former’s organic content is significantly larger than the latter. Fig. 3H of the interactions between fluid parameters (pressure, co-solvent and % cosolvent) demonstrates the optimised extraction conditions for the experiment for the majority of the soil types investigated.Using a CO2 pressure of 300 atm modified with 10% DMSO, the statistical conclusion over the entire data set is that these conditions provide the most satisfactory reproducible extraction recoveries for the carbamates. Due to the co-solvent incompatibility with soils of higher moisture content, the selection of a co-solvent of lower polarity index will result in comparable recoveries for all analytes (Fig. 3E). As expected, P-values for average response suggested that each individual factor (Fig. 3A–E) provided a significant effect to the recoveries obtained. As a result, a main effects plot (Fig. 4) is capable of providing a graphical interpretation of the Table 3 Experiment design Variable Level* Analyte Aldicarb, carbaryl and pirimicarb Soil A, B, C1, C2, D Pressure/atm 200, 300, 450 Co-solvent type Methanol, dichloromethane, dimethyl sulfoxide Co-solvent concentration (%) 5, 10, 20 Effect† Significance level P < 0.001 0.001 < P < 0.05 Main effects‡ A, B, C, D, E — Two factor interactions A3C, A3D, A3E, B3D A3B, B3C Three factor interactions A3B3C, A3C3D, A3D3E A3B3D, A3B3E, A3C3E, C3D3E Four factor interactions — A3B3C3D, A3B3D3E, A3C3D3E * All experiments were carried out at n = 3 replicates.† In order of relevance/importance. ‡ Codes used: A, soil; B, pressure; C, analyte; D, cosolvent; E, % co-solvent. 306 Analyst, April 1997, Vol. 122Fig. 3 Two factor and three factor interaction plots.A, soil type versus analyte; B, analyte versus fluid pressure; C, co-solvent versus pressure; D, soil versus pressure versus analyte at 300 atm; E, soil versus pressure versus co-solvent at 300 atm; F, soil versus pressure versus co-solvent at 300 atm; G, soil versus analyte versus co-solvent (% co-solv 10); and H, pressure versus co-solv versus % co-solvent (% co-solv 10). Analyst, April 1997, Vol. 122 307relative significance of each of the factors investigated.Further interpolation can also indicate at which state each factor returns a maximum effect. This technique can be useful as a backup check when using the model as a retrospective optimisation method for extraction conditions. Conclusions It can been seen by undertaking an exhaustive survey of a selective band of variables, it is possible to determine the significance of each factor to both the entire experiment and to its significance over others. Previously, many studies have relied on experience in SFE to obtain optimised extraction parameters although it can also be seen that the technique in this work has shown other, less obvious, yet important interactions.It was shown that optimised fluid density for carbamate solvation using pure CO2 does not alter on addition of a modifier to the extracting fluid. Successful recoveries using DMSO have illustrated the competitive nature between moisture and the co-solvent in analyte solvation from a matrix where the greatest recovery has occurred using a co-solvent that possesses a polarity index value closest to water.As a result, retrospectively optimised fluid conditions for the extraction of the carbamates from the majority of the soil types were found at 300 atm using 10% DMSO modification. With respect to matrix properties, the effect of moisture has been shown to have a greater effect on extraction recoveries than organic content. This is indicated in comparing % moisture (Table 2) against mean soil recovery (Fig. 4) where soils of comparable moisture content are close together in recovery. It has also been possible to explain many of these interactions although the reduction in expected recovery for carbaryl compared to aldicarb and its maintenance throughout the data warrants further investigation and could be due to the significant DPvap that exists between them. The authors gratefully acknowledge the assistance of A. Evans, Centre for Industrial Bulk Solids Handling, Glasgow Caledonian University, in the particle sizing work presented here. References 1 Caro, J.H., Freeman, H. P., and Turner, B. C., J. Agric. Food Chem., 1974, 22, 860. 2 Butler, L. I., and McDonough, L. M., J. Ass. Off. Anal. Chem., 1965, 53, 495. 3 Dorough, H. W., and Thorstenson, J. H., J. Chrom. Sci., 1975, 13, 212. 4 Rule, G. S., Mordehal, A. U., and Henion, J., Anal. Chem., 1994, 66, 230. 5 Quintero, M. C., Silva, M., and Perez-Bendito, D., Talanta, 1988, 35(12), 943. 6 Krause, R. T., J. Chrom., 1983, 255, 497. 7 Krause, R. T., and August, E. M., J. Ass. Off. Anal. Chem., 1983, 66(2), 234. 8 Hsu, J. P., Scattenberg III, H. J., and Garza, M. M., J. Ass. Off. Anal. Chem., 1991, 74(5), 886. 9 Krause, R. T., J. Ass. Off. Anal. Chem., 1985, 68(4), 726 10 Blaicher, G., Pfannhauser, W., and Woidich, H., Chromatographia, 1980, 13(7), 438. 11 Jansen, H., Brinkman, U. A. Th., and Frei, R. W., Chromatographia, 1985, 20, 453. 12 Frister, H., Meisel, H., and Schlimme, E., Fresenius’ Z. Anal. Chem., 1988, 330, 631. 13 Nondek, L., Frei, R. W., and Brinkman, U. A. Th., J. Chrom., 1983, 282, 141. 14 She, L. W., Brinkman, U. A. Th., and Frei, R. W., Anal. Lett., 1984, 17(A10), 915. 15 Holstege, D. M., Scharberg, D. L., Tor, E. R., Hart, L. C., and Galey, F. D., JAOAC, 1994, 77(5), 1263. 16 Bowman, M. C., and Beroza, M., Residue Rev., 1967, 17, 23. 17 Frei, R. W., Chromatographia, 1982, 15(3), 161. 18 Sharp, G.J., Brayan, J. G., Dilli, S., Haddad, P. R., and Desmarchelier, J. M., Analyst, 1988, 113, 1493. 19 Stuart, I. A., Maclachlan, J., and McNaughtan, A., Analyst, 1996, 121, 11R. 20 Bartle, K. D., Clifford, A. A., Hawthorne, S. B., Langenfeld, J. J., Miller, D. J., and Robinson, R., J. Supercrit. Fluids, 1990, 3, 143. 21 Lopez-Avila, V., and Beckert, W. F., Supercritical Fluid Technology, ACS Symposium Series No. 488, American Chemical Society, Washington DC, ch. 14. 22 Lopez-Avila, V., Dodhiwala, N.S., and Beckert, W. F., J. Chromatogr. Sci., 1990, 28, 468. 23 Kane, M., Dean, J. R., Hitchen, S. M., Dowle, C. J., and Tranter, R. L., Anal. Chim. Acta., 1993, 271, 83. 24 van der Velde, E. G., Ramlal, M. R., van Beuzekom, A. C., and Hoogerbrugge, R., J. Chromatogr. A., 1994, 683, 125. 25 Zhou, M. M., Trubey, R. K., Keil, Z. O., and Sparks, D. L., Abstr. Pap. Am. Chem. Soc., 1996, 212, Pt 1, 3-AGRO. 26 Llompart, M. P., Lorenzo, R. A., and Cola, R., J.Microcol. Sep., 1996, 8(3), 163. 27 Snowden, J. P., Bowen, H. M., and Dickson, J. M., Arable Crops 1990 Survey Report 87, Pesticide Usage in Scotland, Scottish Office Agriculture and Fisheries Department. 28 Kratochvil, B., and Peak, J., Anal. Methods Pestic. Plant Growth Regul., 1989, 17, 1. 29 Dean, J. R, Barnabas, I. J., and Owen, S. P., Analyst, 1996, 121, 465. 30 Izquierdo, A., Tena, M. T., Luque de Castro, M. D., and Valcarcel, M., Chromatographia, 1996, 42(3–4), 206.Paper 6/08060J Received November 28, 1996 Accepted January 17, 1997 Fig. 4 Main effects plot for response means. 308 Analyst, April 1997, Vol. 122 Five-way ANOVA Interaction Analysis of the Selective Extraction of Carbaryl, Pirimicarb and Aldicarb From Soils by Supercritical Fluid Extraction† Iain A. Stuart*a, Ray O. Ansella, John Maclachlana, Peter A. Bathera and William P. Gardinerb a Department of Physical Sciences, Glasgow Caledonian University, Cowcaddens Road, Glasgow, UK G4 0BA b Department of Mathematics, Glasgow Caledonian University, Cowcaddens Road, Glasgow, UK G4 0BA Due to the biopersistence of organophosphate and organochloride compounds, the hydrolytic degradation of the carbamate insecticides has proved attractive in the reduction of persistent insecticides in the biosphere and food chain.Their susceptibility to hydrolysis, however, can complicate their analysis and care is required in the selection of the extraction conditions and the analytical technique employed.The work described here is an investigation into method optimisation in the extraction and final analysis of selected carbamate insecticides from selected soils. A method using the inert extraction medium of supercritical carbon dioxide has been developed for determination of three carbamates relevant to the soft fruit growing industry (carbaryl, aldicarb and pirimicarb). Determinations were completed using HPLC-postcolumn reaction-fluorescence with orthophthalaldehyde-mercaptoethanol derivatisation. The resultant methylisoindole fluorophore was detected at lex : 330 nm and lem : 450 nm with pirimicarb detection at lex : 315 nm and lem : 380 nm.A five variable ANOVA analysis was carried out to determine both the most significant independent factors in the extractions and their most significant statistical interaction(s). Retrospectively optimised extraction conditions were obtained from the ANOVA where many of the carbamates were successfully extracted from each soil using CO2 at 300 atm modified with 10% dimethyl sulfoxide. The mean recoveries obtained were 91.5–107.8% for all carbamates from many soils.Keywords: Pesticides; supercritical fluid extraction; ANOVA; matrix effects In the UK, the Food and Environmental Protection Act 1985 Part III and the Control of Pesticides Act 1986 were designed to protect the health of consumers, the biosphere of the surrounding area and to secure safe, efficient and humane methods of pest control.The enforcement of this legislation has led to an increased need from the analyst to provide reliable, effective methods of qualitative and quantitative pesticide residue analysis from environmental and food matrices. Methods of efficiently monitoring concentrations of carbamate insecticides became an issue with the detection of the aldicarb in groundwater which led to the US Environmental Protection Agency (US EPA) proposing a maximum admissible total carbamate concentration of 0.2 mg l21.UK legislation on carbamates is guided by the EC directive on Drinking Water Quality 80/778/EEC which states that the maximum for any single carbamate is 0.1 mg l21. Local enforcement is placed under The Control of Pesticide Regulations 1986 (MAFF 01690). Due to the thermal lability of this class of compounds and consequent incompatibility with GC methods, previous work involving the analysis of carbamates has focused on the use of reverse phase C18 liquid chromatography with both UV and fluorescence methods of detection.The derivatisation of the carbamates using 2,4-dinitrophenyl phenyl ether1 or trichloroacetate2 has also been used in this context with Dorough and Thorstenson3 providing a thorough review on determining carbamates and their metabolites by GC. Previous work on alternative methods has included mass spectrometric detection (HPLC–thermospray-MS)4 and ‘spot-test’ colorimetric reactions in off-line determinations.5–13 As many of the carbamates used in crop protection possess natural fluorescence, they can be detected successfully without further derivatisation.Increased detection limits for the carbamates have been achieved by using postcolumn reaction fluorescence (PRF) through the formation of a highly fluorescent N-methylisoindole derivative (solely for N-methylcarbamates) and has been extensively studied since its introduction in the 1970s.14–16 Extensive reviews of HPLC reaction detectors and HPLC analysis of carbamates by Frei17 and Sharp et al.,18 respectively, can be used for further reference.The quantification of pesticides such as the carbamates in solid matrices has provided further problems in the selection of an appropriate method for extraction and sample preparation that will not contribute to further analyte breakdown. It is here that one must question the use of basic solvents, e.g., triethylamine in high temperature extractions for compounds that are susceptible to hydrolysis.The introduction of analytical- scale supercritical fluid extraction in the mid to late 1980s has provided an efficient, flexible extraction method capable of solvating pesticides of medium to high hydrophobicity directly and those of low hydrophobicity using reaction–extraction protocols in an inert solvent.19 Further to the solubility of the analyte in the supercritical flow, the second main component that affects extraction recovery is the rate of mass transfer from the matrix into the supercritical flow.20 This parameter acts as a function of the strength of the analyte-to-matrix interactions present in the sample and is of greatest interest in many of the studies carried out in extracting pesticides from environmental matrices by SFE.Due to the complexity of interaction between a supercritical fluid and the host matrix, many workers have attempted to use statistical techniques in order to quantify the significance of a particular factor or interaction. Lopez-Avila and Beckert21 investigated several instrumental variables (cell geometry, fluid flow rate and selection of appropriate collection media) for their respective and cumulative effect on analyte recovery of 42 † Presented, in part, at the International Symposium on Supercritical Fluid Chromatography and Extraction, Indianapolis, Indiana, USA, March 31–April 4, 1996.Analyst, April 1997, Vol. 122 (303–308) 303organochlorine pesticides from spiked sand and soil samples (this work was further extended to cover selected OPPs22). Method optimisation was carried out on seven variables each set at low and high to define relative changes in recovery for eight extractions with a second test used to investigate the presence/ absence of a modifer, glass bead sorbent and static extraction. Conclusions made on recovery evidence and %RSDs for the 42 compounds proved the existence of both independent and intervariable dependences.The use of statistical techniques in this way have provided the analyst with a quick, relatively accurate optimisation technique that would otherwise have been costly in both time and materials. In performing a full factorial design for a seven variable system, 27 experiments would have had to have been performed to obtain a set of comparable data trends. Additional optimisation procedures such as multilinear regression (MLR) in experimental design has been used by Kane et al.23 and by van der Velde et al.24 for the extraction of triazines from soil with the objective being again to reduce the number of runs required in method optimisation.Other techniques such as multivariate optimisation25 and two and three level factorial designs26 have also been used with this aim in mind. This work concentrates on developing an interaction model for the extraction of selected carbamates from soil matrices of differing chemical and physical characteristics.The use of ANOVA techniques in this context when used on an exhaustive data set can produce information on the most important fluid characteristics with respect to analyte solvation in the extracting fluid and matrix effects. In this study, five soil samples with four soil matrix parameters, namely: moisture content; pH; particle size; and organic content were measured against variation in pressure, analyte, co-solvent type and % co-solvent added to the fluid.The analytes selected for this study were pirimicarb, aldicarb and carbaryl as they are used exclusively for both crop protection (Fig. 1) and as pest deterrents in both domestic use and the soft fruit growing industry in Scotland.27 Experimental Instrumentation All extractions were completed on an SFE-723M Supercritical Fluid Extraction system (Dionex, Camberley, UK) using two 16 ml capacity extraction cells (Keystone Scientific, Bellefonte, USA) in parallel (in positions 1 and 8 on the manifold fitted with 1200 ml linear restrictors) for each extraction run. 99.99% SFEgrade CO2 was used as the primary solvent supplied with a 110 bar He overpressure (BOC Speciality Gases, Guildford, UK). Organic and moisture content were determined by thermogravimetry (model TG-750/70, Stanton-Redcroft, London, UK) using software developed in-house. Particle sizing was completed on a low angle laser scattering particle size analyser (Malvern, Malvern, UK) with sample introduction by gravity feeding.Extract Assay All chromatographic and modifier solvents used for the determination of the Certified Pesticide Standards (Promochem Ltd. Hertfordshire, UK) were of HPLC-grade (Sigma-Aldrich, Poole, Dorset, UK). The HPLC pump used was a LC 9012 Solvent Delivery System (Varian, Walton-on-Thames, UK) with a 10 ml injection loop and detection by postcolumn reaction fluorescence on a scanning wavelength detector (model 9070, Varian) at lex : 330 nm and lem : 450 nm for carbaryl, carbofuran (internal standard) and aldicarb. Pirimicarb was detected at lex : 315 nm and lem : 380 nm without the presence of reagents.An isocratic mobile phase of 55% water–45% methanol with a flow rate of 1 ml min21 was used in all assays. All determinations were completed on a 150 mm 3 4.6 mm C18 carbamate column at 42 °C (Pickering Laboratories, Mountain View, CA, USA) contained in the postcolumn reaction module (PCX 5100, Pickering Laboratories).Fluorescence reagents, orthophthalaldehyde (OPA), sodium hydroxide hydrolyser (0.4% at 100 °C), OPA diluent (0.3% boric acid) and Thiofluor (N,N-dimethyl-2-mercaptoethylamine hydrochloride) were all of chromatographic-grade (Pickering Laboratories). The monitored extraction parameters are given in Table 1. Soil Sample Preparation Soil samples were taken and left exposed overnight to come into equilibrium with the laboratory atmosphere moisture content.Representative samples of the soils (2 g) were taken using a ‘coning and quartering’ technique and each extraction cell was packed at each end with methanol-cleaned glass wool to prevent end-cell frit blockage. Soil spiking was completed by evaporating 1 ml of a methanolic carbamate mixture (containing 50 mg of each insecticide) directly onto each sample of soil. Evaporation was allowed to occur in the open laboratory atmosphere.Post-extraction, each extract was spiked with the internal standard (carbofuran, 50 mg ml21) and then reduced to 1 ml prior to chromatographic determination. Kratochvil and Peak28 have provided a useful review on the subject of pesticide sampling/handling from a number of sources. Fig. 1 Carbamates used in soft fruit protection Table 1 Monitored extraction parameters Extraction temperature 70 °C Extraction pressures 10 minute cell ‘conditioning’ at 150 atm followed by individual extractions at 200, 300 and 450 atm Total extraction time 40 min Restrictor temperature 70 °C Restrictor volume 1200 ml Flow rate (gas state)* 1650 ml min21 Modifiers used CH2Cl2 (5, 10 and 20%), DMSO (5, 10, 20%), CH3OH (5, 10, 20%) Solvent collection Liquid collection in vial (15 ml) Solvent collection temperature 21 to 1 °C Extraction cell geometry 14 mm3100 mm (16 ml capacity) Cell packing Methanol-cleaned glass wool with celite wet support * Mean flow rate 304 Analyst, April 1997, Vol. 122Results and Discussion In the investigation of extraction conditions, 30 supercritical fluids were used to investigate optimum extraction conditions for the carbamates selected, i.e., three modifiers at three different concentrations and three pressures including pure CO2. Initially, the relative solubility of each carbamate was plotted (Fig. 2) to determine if fluid modification was required and to ascertain at which fluid pressure consistently presented the highest recoveries.This was completed by evaporating 1 ml of the methanolic carbamate mixture onto 2 g of celite prior to extraction. It was found that analyte-to-matrix interactions between the hydrogen bonding of the carbamates and the celite support was not as significant an interaction as in previous studies in polar pesticide extraction. This would then indicate that the pH of the matrix may not be a significant factor in the later studies on soil extractions. Evidence to support this conclusion is shown in the extraction profiles for all three compounds at 300 and 450 atm (200 atm omitted for clarity), Fig. 2, whereby extraction of the compounds with pure CO2 implied very low matrix interaction. The extraction profile also illustrates that using a fluid pressure of 300 atm recovers the carbamates and in particular, pirimicarb, exceptionally well without the presence of fluid polarity modification. Percentage recoveries from the inert substrate in pure CO2 for all three pesticides are consistent with their relative polarity and is a satisfactory initial test of the suitability of the extraction conditions.Further to this, consequent extract reverse phase chromatograms indicate that the order of increasing hydrophobicity is aldicarb, carbaryl, carbofuran (internal standard), pirimicarb. It was also noted that in increasing the fluid density above 300 atm, comparison of individual pesticide extraction profiles indicated a reduction in solubility.As a result, this study provided basic analyte solubility information in the non-polar solvent prior to solvent modification. In selected appropriate co-solvents for the study, dichloromethane (DCM, 3.1), methanol (5.1) and dimethyl sulfoxide (DMSO, 7.2) were selected from their relative solvent strengths, i.e., low to high polarity index. Using these solvents, it is then possible to obtain relevant information on the significance of co-solvent polarity on extraction recovery.Additional issues pertaining to their potential for successful extraction and the benefits of minimal co-extractant throughput to the collection solvent were also considered. Parametric Significance of Soil Types The soils used were chosen from areas characteristically classed as fruit growing land. Soils C1 (forestry) and C2 (forestry drainage) were selected for comparison against soil B (arable land) for organic content significance with soil D (top soil) compared against all four (soil A, coastal soil) remaining soils for moisture content significance.The properties of the soils used are given in Table 2 indicating significant organic and moisture content in all test soils used for SFE extractions. The significance of these factors has been previously shown29,30 as the corresponding organic mass loss can provide a guide to the organic activity (aorg) of the host matrix and so provide a relative strength index of the analyte-to-matrix interaction.pH is used as an indication of humic/organic acid concentration although cation exchange capacity (CEC) has also been used to assess the cation activity in soil matrices.29 The relative percentage moisture content can also be a significant factor in the competition between inherent moisture content and cosolvent strength in the removal of analytes from the matrix. In addition to this issue, particle size has also been measured although found to be comparable in all cases.Experimental Data The experiment was designed to obtain exhaustive statistical information on all possible effects and interactions between the selected variable set (Table 3). As a result, all recovery data effects with respect to fluid pressure, co-solvent and co-solvent type were acquired from using low, medium and high settings. These results were set against each soil and analyte type. Statistical Analysis The use of ANOVA models provides the user with a statistically based technique capable of producing meaningful models on the importance of the studied factors in the experiment.In addition to this, it is also possible to observe possible factor interactions within the data set that may have a bearing in the development of optimised extraction conditions. The use of modelling in this context with SFE has proved highly relevant due mainly to the ability to accurately vary the extracting fluids physical properties.In addition to this, the importance of other parameters such as analyte polarity and matrix properties on extraction recovery are also demonstrated in ANOVA results. A five factor ANOVA analysis was selected to investigate factor effects and interactions with respect to recorded recovery rates and to use a retrospective optimisation technique to obtain the most appropriate extraction conditions for each soil and analyte type. Due to the complexity of the analysis, it was not possible to obtain recordings for certain treatment combinations.These missing entries complicated the use of standard ANOVA procedures with respect to the full data set. To overcome this, the mean values for each treatment combination were treated as the response with missing entries estimated using a weighted average of means for the factor levels corresponding to the missing treatment combinations: n1x – 1 + n2x – 2 + ··· + n5x – 5 Xmean = n1 + n2 + ··· + n5 Fig. 2 Extraction profiles for supercritical fluid solubility optimisation of carbaryl, pirimicarb and aldicarb at 70 °C and at 300 and 450 atm, respectively (rfluid = 0.79 g ml21 at 300 atm). A, 300 atm aldicarb; B, 450 atm pirimicarb; C, 300 atm carbaryl; D, 450 atm carbaryl; E, 300 atm pirimicarb; F, 450 atm aldicarb. Table 2 Properties of investigated soils Mean particle Soil Moisture (%) Organic (%) pH size/mm A 32.06 7.28 4.64 549.1 B 26.44 13.85 4.26 452.5 C1 23.89 4.09 4.80 429.9 C2 24.70 3.23 3.89 432.2 D 14.31 6.75 4.47 558.4 Analyst, April 1997, Vol. 122 305ANOVA assessments of the standard deviations of the treatment combinations was also complicated by missing entries. This was rectified by using the mean of the two known standard deviations for the first four factor combinations as an estimate of the missing value. Use of the full model with all possible effects results in no degrees of freedom for error estimation. To overcome this, a model of the form: SoilýPressureýAnalyteýCo-solventý%Co-solvent-Soil* Pressure*Analyte*Co-solvent*%Co-solvent was fitted to both mean and standard deviation data.The first part of this model refers to the full model with all possible effects (factors and interactions) with the subtraction of the five factor interaction necessary to provide an error term for effect estimation and test statistic construction. Table 3 provides a summary of the ANOVA results, modified to account for missing values, in respect of average recovery.The results show that most effects are significant (P < 0.05) with all main effects and several two and three factor interactions highly significant. For the ANOVA assessment of the standard deviations, only soil type was shown to have any significant effect (P = 0.0287) on the variability in recovery rate (P < 0.05), as might be expected. To maintain succinctness and to avoid repetition, three two factor interaction plots and five three factor interaction plots are included (Fig. 3) which make up the majority of the most statistically important and relevant interactions presented by the data. Four factor interactions are not represented due to graphical complexity. Two Factor Interactions Fig. 3A illustrates the interaction between soil type and each analyte. It has been found that throughout the entire data set, aldicarb is recovered to a greater extent than carbaryl. This contradicts with the relative hydrophobicities of the selected analytes and the initial solubility extractions using unmodified CO2, although pirimicarb, having the greatest log P, is recovered to the highest level.The interaction between each analyte and each fluid pressure (Fig. 3B) has indicated that the optimum extraction pressure/density of the extracting fluid is 300 atm in all cases of modified fluid. This has been shown to be a confirmation of the initial studies for analyte solubility in unmodified CO2 (Fig. 2). Using the co-solvent DMSO (Fig. 3C) has provided the greatest recovery of the analytes at 300 atm pressure, although DCM, which has a lower polarity index value than methanol, recovers the analytes to a higher level than using methanol as the co-solvent. Three Factor Interactions Conclusions taken from two factor interactions can be used as an indicator of relevance of the ANOVA model in the three factor analysis. Fig. 3D exhibits this incremental aspect in the comparison of interaction effects between soil type, pressure and analyte where recovery in all cases is as follows, pirimicarb > aldicarb > carbaryl, as shown in Fig. 3A. The coastal soil A demonstrates the poorest recovery but also possesses the highest moisture content and the second highest organic content. Conformation of the significance of organic and moisture content is shown for the arable land soil B which possesses the second lowest response of the data set. Soil B characteristics include having the highest organic content and the second highest moisture content. In Fig. 3E, the selection of an appropriate co-solvent is made using Fig. 3C as a basis. Recovery from soil B is improved by the use of DMSO. Reproducibility is high for soils B–D although recovery from soil A is reduced. Implying that water has a polarity index value of 10.2, the action of DCM (3.1) will not be curtailed by the presence of water, although co-solvents that are more miscible in water will be subject to the competitive action of water to retain the analyte to the surface of the matrix.This incompatibility is shown in the fact that DCM possesses the highest recovery from most soils with methanol (5.1) next and finally DMSO (7.2). Fig. 3F depicts the optimised concentration of co-solvent in the modified extraction fluid. For the majority of soil types, extractions using co-solvent at a concentration of 10% appear best. In observing soil, analyte and % co-solvent interactions (Fig. 3G), previous problems with the extraction from matrices of higher moisture content is further illustrated here. In this interaction, it is difficult to divorce organic content from moisture content effects in soil C1 and C2 as the former has lower recovery yet lower moisture content than the latter although the former’s organic content is significantly larger than the latter. Fig. 3H of the interactions between fluid parameters (pressure, co-solvent and % cosolvent) demonstrates the optimised extraction conditions for the experiment for the majority of the soil types investigated.Using a CO2 pressure of 300 atm modified with 10% DMSO, the statistical conclusion over the entire data set is that these conditions provide the most satisfactory reproducible extraction recoveries for the carbamates. Due to the co-solvent incompatibility with soils of higher moisture content, the selection of a co-solvent of lower polarity index will result in comparable recoveries for all analytes (Fig. 3E). As expected, P-values for average response suggested that each individual factor (Fig. 3A–E) provided a significant effect to the recoveries obtained. As a result, a main effects plot (Fig. 4) is capable of providing a graphical interpretation of the Table 3 Experiment design Variable Level* Analyte Aldicarb, carbaryl and pirimicarb Soil A, B, C1, C2, D Pressure/atm 200, 300, 450 Co-solvent type Methanol, dichloromethane, dimethyl sulfoxide Co-solvent concentration (%) 5, 10, 20 Effect† Significance level P < 0.001 0.001 < P < 0.05 Main effects‡ A, B, C, D, E — Two factor interactions A3C, A3D, A3E, B3D A3B, B3C Three factor interactions A3B3C, A3C3D, A3D3E A3B3D, A3B3E, A3C3E, C3D3E Four factor interactions — A3B3C3D, A3B3D3E, A3C3D3E * All experiments were carried out at n = 3 replicates.† In order of relevance/importance. ‡ Codes used: A, soil; B, pressure; C, analyte; D, cosolvent; E, % co-solvent. 306 Analyst, April 1997, Vol. 122Fig. 3 Two factor and three factor interaction plots. A, soil type versus analyte; B, analyte versus fluid pressure; C, co-solvent versus pressure; D, soil versus pressure versus analyte at 300 atm; E, soil versus pressure versus co-solvent at 300 atm; F, soil versus pressure versus co-solvent at 300 atm; G, soil versus analyte versus co-solvent (% co-solv 10); and H, pressure versus co-solv versus % co-solvent (% co-solv 10).Analyst, April 1997, Vol. 122 307relative significance of each of the factors investigated. Further interpolation can also indicate at which state each factor returns a maximum effect. This technique can be useful as a backup check when using the model as a retrospective optimisation method for extraction conditions. Conclusions It can been seen by undertaking an exhaustive survey of a selective band of variables, it is possible to determine the significance of each factor to both the entire experiment and to its significance over others.Previously, many studies have relied on experience in SFE to obtain optimised extraction parameters although it can also be seen that the technique in this work has shown other, less obvious, yet important interactions. It was shown that optimised fluid density for carbamate solvation using pure CO2 does not alter on addition of a modifier to the extracting fluid. Successful recoveries using DMSO have illustrated the competitive nature between moisture and the co-solvent in analyte solvation from a matrix where the greatest recovery has occurred using a co-solvent that possesses a polarity index value closest to water.As a result, retrospectively optimised fluid conditions for the extraction of the carbamates from the majority of the soil types were found at 300 atm using 10% DMSO modification. With respect to matrix properties, the effect of moisture has been shown to have a greater effect on extraction recoveries than organic content.This is indicated in comparing % moisture (Table 2) against mean soil recovery (Fig. 4) where soils of comparable moisture content are close together in recovery. It has also been possible to explain many of these interactions although the reduction in expected recovery for carbaryl compared to aldicarb and its maintenance throughout the data warrants further investigation and could be due to the significant DPvap that exists between them.The authors gratefully acknowledge the assistance of A. Evans, Centre for Industrial Bulk Solids Handling, Glasgow Caledonian University, in the particle sizing work presented here. References 1 Caro, J. H., Freeman, H. P., and Turner, B. C., J. Agric. Food Chem., 1974, 22, 860. 2 Butler, L. I., and McDonough, L. M., J. Ass. Off. Anal. Chem., 1965, 53, 495. 3 Dorough, H. W., and Thorstenson, J. H., J. Chrom. Sci., 1975, 13, 212. 4 Rule, G. S., Mordehal, A. U., and Henion, J., Anal. Chem., 1994, 66, 230. 5 Quintero, M. C., Silva, M., and Perez-Bendito, D., Talanta, 1988, 35(12), 943. 6 Krause, R. T., J. Chrom., 1983, 255, 497. 7 Krause, R. T., and August, E. M., J. Ass. Off. Anal. Chem., 1983, 66(2), 234. 8 Hsu, J. P., Scattenberg III, H. J., and Garza, M. M., J. Ass. Off. Anal. Chem., 1991, 74(5), 886. 9 Krause, R. T., J. Ass. Off. Anal. Chem., 1985, 68(4), 726 10 Blaicher, G., Pfannhauser, W., and Woidich, H., Chromatographia, 1980, 13(7), 438. 11 Jansen, H., Brinkman, U. A. Th., and Frei, R. W., Chromatographia, 1985, 20, 453. 12 Frister, H., Meisel, H., and Schlimme, E., Fresenius’ Z. Anal. Chem., 1988, 330, 631. 13 Nondek, L., Frei, R. W., and Brinkman, U. A. Th., J. Chrom., 1983, 282, 141. 14 She, L. W., Brinkman, U. A. Th., and Frei, R. W., Anal. Lett., 1984, 17(A10), 915. 15 Holstege, D. M., Scharberg, D. L., Tor, E. R., Hart, L. C., and Galey, F. D., JAOAC, 1994, 77(5), 1263. 16 Bowman, M. C., and Beroza, M., Residue Rev., 1967, 17, 23. 17 Frei, R. W., Chromatographia, 1982, 15(3), 161. 18 Sharp, G. J., Brayan, J. G., Dilli, S., Haddad, P. R., and Desmarchelier, J. M., Analyst, 1988, 113, 1493. 19 Stuart, I. A., Maclachlan, J., and McNaughtan, A., Analyst, 1996, 121, 11R. 20 Bartle, K. D., Clifford, A. A., Hawthorne, S. B., Langenfeld, J. J., Miller, D. J., and Robinson, R., J. Supercrit. Fluids, 1990, 3, 143. 21 Lopez-Avila, V., and Beckert, W. F., Supercritical Fluid Technology, ACS Symposium Series No. 488, American Chemical Society, Washington DC, ch. 14. 22 Lopez-Avila, V., Dodhiwala, N. S., and Beckert, W. F., J. Chromatogr. Sci., 1990, 28, 468. 23 Kane, M., Dean, J. R., Hitchen, S. M., Dowle, C. J., and Tranter, R. L., Anal. Chim. Acta., 1993, 271, 83. 24 van der Velde, E. G., Ramlal, M. R., van Beuzekom, A. C., and Hoogerbrugge, R., J. Chromatogr. A., 1994, 683, 125. 25 Zhou, M. M., Trubey, R. K., Keil, Z. O., and Sparks, D. L., Abstr. Pap. Am. Chem. Soc., 1996, 212, Pt 1, 3-AGRO. 26 Llompart, M. P., Lorenzo, R. A., and Cola, R., J. Microcol. Sep., 1996, 8(3), 163. 27 Snowden, J. P., Bowen, H. M., and Dickson, J. M., Arable Crops 1990 Survey Report 87, Pesticide Usage in Scotland, Scottish Office Agriculture and Fisheries Department. 28 Kratochvil, B., and Peak, J., Anal. Methods Pestic. Plant Growth Regul., 1989, 17, 1. 29 Dean, J. R, Barnabas, I. J., and Owen, S. P., Analyst, 1996, 121, 465. 30 Izquierdo, A., Tena, M. T., Luque de Castro, M. D., and Valcarcel, M., Chromatographia, 1996, 42(3–4), 206. Paper 6/08060J Received November 28, 1996 Accepted January 17, 1997 Fig. 4 Main effects plot for response means. 308 Analyst, April 1997, Vol. 122
ISSN:0003-2654
DOI:10.1039/a608060j
出版商:RSC
年代:1997
数据来源: RSC
|
4. |
Integrated Automatic Determination of Nitrate, Ammonium and OrganicCarbon in Soil Samples |
|
Analyst,
Volume 122,
Issue 4,
1997,
Page 309-313
Evaristo Ballesteros,
Preview
|
|
摘要:
Integrated Automatic Determination of Nitrate, Ammonium and Organic Carbon in Soil Samples Evaristo Ballesteros, Angel R�ýos and Miguel Valc�arcel* Department of Analytical Chemistry, Faculty of Sciences, University of C�ordoba, E-14004 C�ordoba, Spain An automated photometric method for the sequential determination of nitrate, ammonium and organic carbon in soils by use of a single-line flow injection assembly is reported, but differently programmed for the assay of the individual analytes.These analytes, which are determined by using suitable photometric reagents, are directly related to the nutritional properties of soil. A filtration probe was used for the continuous filtration of soil extract. The precision, expressed as RSD, was 2.0, 2.6 and 2.1% for nitrate, ammonium and organic carbon, respectively. The proposed method features a high throughput and low reagent consumption, and requires minimal sample handling. Its performance was tested in routine analyses of soil samples, all with satisfactory results.Keywords: Flow system; automatic photometric determination; nitrate; ammonium; organic carbon; soil samples The bound inorganic nitrogen in soils is predominantly in the form of ammonium and nitrate; nitrite is seldom present in detectable amounts, so its determination is usually unwarranted. Several other inorganic forms of nitrogen, including hydroxylamine, hyponitrous acid, and imidonitric acid (nitramide), have been postulated as intermediates in microbial transformations of nitrogen; however, none of these compounds has been detected in soils.Organic carbon is contained in the soil organic fraction, which consists of cells of microorganisms, plant and animal residues at various stages of decay, stable humus synthesized from residues and highly carbonized compounds such as charcoal, graphite and coal.1 Flow injection (FI) methodology is a powerful tool for performing a wide variety of on-line treatments. Over 99% of reported FI applications are based on the introduction of pretreated liquid samples.2,3 The preliminary operations of the analytical process (sampling and sample conditioning and treatment) are rarely considered systematically in FI.These steps are crucial because they are tedious, time consuming and the source of major errors affecting the final results. Direct analyses of heterogeneous or solid samples by use of automatic systems pose special difficulties.4 Automatic leaching can be accomplished in a continuous-flow configuration by using a dissolving stream together with an external source of energy (e.g.a high temperature, an electrical discharge or ultrasound). Ultrasonic irradiation has been applied to the direct determination of iron in plant materials and available boron in soil.5 A continuous system involving microwave treatment of shellfish samples was used for aluminium determination by ETAAS.6 McLeod7 developed a continuous method for the determination of nitrogen in soils and plants from a Kjeldahl digest, by using a flow system with a microdistillation unit.Zhi et al.8 reported a flow system integrating a soil sample pre-treatment unit and a flow manifold for the sequential determination of nitrate and nitrite. Few methods for convenient, automatic multi-determinations are available.9 In previous work, we developed a method for the determination of pH, nitrite, copper, iron and aluminium in waste water by use of a continuous configuration including both potentiometric and photometric detectors,10 in addition to a determination for pH, conductivity, residual chlorine, ammonium and nitrate ion in water with an unsegmented flow configuration.11 More recently, continuous flow systems have been integrated with chromatographic equipment in order to automate sample preparation prior to introduction into a chromatograph, thus facilitating multi-determinations.12–14 The aim of this work was to develop a methodology for the integrated determination of ammonium, nitrate and organic carbon in soils by using a flow module including a diode-array detector. Previously, nitrate and ammonium have been widely determined by using flow injection methods; in this work, however, the determination of these two analytes was integrated with that of organic carbon.Soil sample extracts were cleaned up in the manifold by means of a filtration probe accommodating a hydrophobic PTFE membrane.Experimental Apparatus A Hewlett-Packard (Avondale, PA, USA) Model 8451A diodearray spectrophotometer furnished with an HP Think-Jet printer was used. The proposed continuous system consisted of two Gilson (Worthington, OH, USA) Minipuls-3 peristaltic pumps fitted with poly(vinyl chloride) tubes; five Rheodyne (Cotati, CA, USA) Model 5041 valves; PTFE tubing (0.5 mm id) for coils; a Hellma (Jamaica, NY, USA) Model 178.12 QS flow cell (inner volume 18 ml; pathlength 10 mm); and a reduction column (50 3 3 mm id) packed with 0.8–1.0 mm grain-size cadmium granules.The filtration probe was developed in our laboratory (Fig. 1), from two glass tubes [1.5 cm and 1.0 cm id, for the internal (B) and external tube (C), respectively] and a PTFE tube (A, 0.5 mm id) connected with the pump tube. The PTFE tube was held on a PTFE support (D). Glass tubes B and C were sealed with a hydrophobic membrane PTFE from Millipore (Bedford, MA, USA) (Mitex, 1.0 mm pore size) through an O-ring (see details in Fig. 1). The filtration probe was included in the tube for aspiration of the sample in the flow system. When the probe was used to filter soil extracts (in 2 m KCl) for the determination of nitrate and ammonium, tube B was sealed with filter-paper (0.45 mm pore size); however, a paper filter cannot be used with the sulfuric acid–dichromate extract handled in the determination of organic carbon.At the end of each working day, the filtration probe was cleaned with distilled water. With this maintenance, the probe can be used for about 500 analyses. Reagents Phenol–nitroprusside reagent. An amount of 7 g of phenol (0.74 m) (Merck, Darmstadt, Germany) and 34 mg of sodium pentacyanonitrosylferrate(iii) (sodium nitroprusside) (1.14 mm) Analyst, April 1997, Vol. 122 (309–313) 309(Sigma, St. Louis, MO, USA) were dissolved in 80 ml of distilled water and then diluted to 100 ml.The solution was stored refrigerated in a dark-coloured bottle. Buffered hypochlorite reagent. A 1.480 g amount of sodium hydroxide was dissolved in 70 ml of distilled water (0.37 m) and supplied with 4.98 g of sodium monohydrogenphosphate (0.35 m) (Merck) and 20 ml of sodium hypochlorite solution (5–5.25% v/v). The solution was adjusted to pH 11.4–12.3 with sodium hydroxide and diluted to 100 ml. Ethylenediaminetetracetic acid (EDTA) reagent. A 6.0 g amount of EDTA disodium salt (0.16 m) (Merck) was dissolved in 80 ml of distilled water, adjusted to pH 7 and diluted to a final volume of 100 ml.Sulfanilamide solution. A 2.5 g amount of sulfanilamide (Merck) was dissolved in 13 ml of concentrated hydrochloric acid and diluted to 250 ml with distilled water (0.06 m). N-(1-naphthyl)ethylenediamine solution. A 0.25 g amount of the dihydrochloride (Merck) was dissolved in 250 ml of distilled water (3.9 mm). Conditioning reagent for the cadmium reduction column.A solution containing 0.1% (w/v) copper(ii) sulfate in 0.1 m EDTA was used for this purpose. Ammonium standard stock solution, 1 g l21. Prepared from ammonium chloride ( > 99.8%, Merck). Nitrate standard stock solution, 100 g l21. Prepared from sodium nitrate ( > 99.5%, Merck). Glucose standard solution, 100 g ml21 ( > 99.5%, Sigma). Sample Treatment Determination of nitrate and ammonium A battery of 50 samples was prepared as follows: 10 g of each soil sample was placed in a PTFE bottle and 100 ml of 2 m potassium chloride were added. The bottle was stoppered and the solution stirred mechanically for 1 h.The samples were then sequentially filtered through the filtration/dialysis probe and successively introduced into the manifold. Determination of organic carbon A battery of 50 samples was prepared as follows: 1 g of each soil sample was mixed with 8% potassium dichromate and 15 ml otrated sulfuric acid.The solution was stirred mechanically for 15 min, diluted to 100 ml and finally continuously filtered through the filtration/dialysis probe for introduction into the manifold. Procedure The manifold used for the sequential determination of nitrate, ammonium and organic carbon in soils is depicted in Fig. 2(a). Each analyte was quantified with a different reagent (A, B, or C, Table 1). The same tubes were used in all the determinations, the flow rate being adjusted as required by changing the speed of pumps.The positions (I or II) of the switching valves (SV1– Fig. 1 Scheme of the filtration probe used for the continuous introduction of soil extract. A, 0.5 mm id PTFE tubing; B and C, glass tubing; D, PTFE support; E, O-ring; F, PTFE hydrophobic membrane; G, filter paper (or hydrophobic PTFE membrane). Fig. 2 Overall flow manifold (a) and details of the continuous determination of nitrate (b), ammonium (c) and organic carbon (d). A, B and C, reagents (see Table 1); RC, reduction column; SV, switching valve; IV, injection valve; R, reactor; D, diode-array spectrophotometer. 310 Analyst, April 1997, Vol. 122SV4) used with each analyte are given in Table 1. For the determination of nitrate [Fig. 2(b)], the cadmiun reduction column was conditioned daily by passing a solution containing 0.1% copper(ii) sulfate in 0.1 m EDTA for 3 min. For the determination of ammonium [Fig. 2(c)], reactor R (300 cm) was heated to 50 °C by immersing the coil in a thermostated waterbath, 2 m potassium chloride was used as carrier in both determinations and 15% sulfuric acid in that for organic carbon [Fig. 2(d)]. Soil samples were passed through the filtration probe in all instances (Fig. 1). Results and Discussion The photometric technique has been widely used for determining nitrate, ammonium and organic carbon. The manifold depicted in Fig. 2(a) was optimized for the determination of these three parameters in soil samples by different photometric reactions.Determination of Nitrate Nitrate is usually determined by direct photometric methods or measurement after derivatization (usually reduction to nitrite); these methods utilize a homogeneous reductant or a heterogeneous reductant containing amalgamated zinc,15 amalgamated cadmium,16 copperized cadmium7,8,17 or copperized cadmium– silver.18 The nitrite thus produced is determined by diazotization with sulfanilamide and coupling with N-(1-naphthyl)ethylenediamine to form a highly coloured azo dye that is measured photometrically.19 The optimum flow conditions for the determination of nitrate [Fig. 2(a)] are given in Table 1.The effect of the concentrations of the reagents [sulfanilamide, N-(1-naphthyl)ethylenediamine and hydrochloric acid] were individually studied. Solutions of 1% sulfanilamide, 0.1% N-(1-naphthyl)ethylenediamine and 5% v/v hydrochloric acid were examined. Also, a solution consisting of a mixture of the three reagents at these concentrations was tested for reaction with nitrite (after reduction of nitrate on the cadmium column).Although the reaction took place, the sensitivity was about 40% lower. The best results were obtained when the reagents were placed in two solutions consisting of 0.1% N-(1-naphthyl)ethylenediamine on the one hand and of 1% sulfanilamide and 5% v/v HCl on the other. In a previous study,20 we found the most efficient reduction of nitrate to nitrite was achieved with a column packed with copperized cadmium in medium to large grain sizes.A column packed with 0.8–1.0 mm grain-sized cadmium granules and copperized with a solution containing 0.1% v/v copper(ii) sulfate in 0.1 m EDTA was finally used. The analytical features for the determination of nitrate are summarized in Table 2. The sampling frequency was relatively low owing to the time needed for reproducible signals to be obtained (about 3 min). Determination of Ammonium There are many methods for the determination of ammonium, some of which have been adapted to the flow injection technique (e.g., several methods based on the Nessler21,22 and Berthelot23,24 reactions).We tested two configurations based on Table 1 FI conditions for the determination of nitrate, ammonium and organic carbon Organic Nitrate Ammonium carbon A N-(1-Naphthyl)ethylenediamine Buffered hypochlorite — q/ml min21 0.45 0.35 B Sulfanilamide Phenol–nitroprusside reagent — q/ml min21 0.45 0.35 Conditioning reagent Copper(ii) sulfate in EDTA — — q/ml min21 0.45 Carrier 2 m KCl 2 m KCl 15% H2SO4 q/ml min21 2.90 1.00 2.90 Sample q/ml min21 0.75 0.55 0.75 C — EDTA — q/ml min21 0.33 SV1 Ia I II SV2 I II I SV3 I II II SV4 I–II I I IV loop 200 80 200 R, temperature/°C — 50 — Pump 1 On On On Pump 2 On On Off l/nm 540 636 600 a Switching valves (SV) were in position I or II.Table 2 Figures of merit of the determination of nitrate, ammonium and organic carbon Regression Detection Linear RSD‡ Sampling Analyte Equation coefficient limit range (%) frequency/h21 Nitrate A = 0.090 [NO32] 2 0.002* 0.996 0.1* 0.25–4* 2.0 15 Ammonium A = 0.039 [NH4 +] + 0.008* 0.998 0.1* 0.4–8* 2.6 30 Organic carbon A = 0.170 [C] + 0.009† 0.998 0.05† 0.1–2† 2.1 60 * Concentration in mg ml21. † Concentration in % w/v carbon as glucose.‡ Concentration for RSD (n = 11): 2 mg ml21 nitrate, 3 mg ml21 ammonium and 0.8% w/v carbon. Analyst, April 1997, Vol. 122 311these two reactions for the determination of ammonium. The method based on the Nessler reaction posed many problems with the soil samples, possibly because of precipitated calcium hydroxide and other interferents present in soil, the alkaline pH needed for the reaction and the decreased signal resulting from the high concentration of potassium chloride needed to extract the ammonium from the soil. The system depicted in Fig. 2 (based on the Berthelot reaction) was freed from the above problems posed by calcium and other interferents present in the soil by using a stream of EDTA that was merged with the sample.In this case, a high potassium chloride concentration had no effect on the sensitivity. Table 1 gives the optimum sample, reagent and carrier flow rates used for the determination of ammonium. The phenol–nitroprusside solution contained 70 g l21 phenol and 0.34 g l21 sodium nitroprusside. The optimum pH range for the hypochlorite buffer [containing 14.8 g l21 sodium hydroxide, 49.8 g l21 sodium monohydrogenphosphate and 20% of sodium hypochlorite solution (5–5.25% v/v)] was 11.4–12.3.Table 2 shows the analytical figures of merit of the method. The precision, assessed on 11 samples containing 4 mg ml21 ammonium, was 2.6% (RSD). Determination of Organic Carbon The proposed continuous method for the determination of organic carbon in soil (Fig. 2) was based on the manual method of Sims and Haby,25 in which the organic carbon is oxidized by using a potassium dichromate solution in sulfuric acid and photometric measurement at 600 nm of the CrIII produced in the reduction of dichromate.The flow conditions used in the determination of organic carbon are summarized in Table 1. Table 2 gives the figures of merit of the method; organic carbon contents are expressed as the percentage (m/v) of carbon in samples containing various amounts of glucose (the method uses glucose as a standard).Alternatively, the organic matter content of a soil may be estimated by multiplying the organic carbon concentration by a constant factor based on the proportion of carbon in the organic matter; however, this factor varied between 1.7 and 2 depending on the soil sample and must be determined experimentally for each soil. Therefore, ‘organic carbon concentration’ is preferable to the term ‘soil organic matter content’ because the latter cannot be determined directly.1 Analysis of Soil Samples The proposed method was applied to the determination of nitrate, ammonium and organic carbon in 20 soil samples from different countryside areas of C�ordoba (southern Spain).Samples were collected by means of a sampling probe and dried. Subsequently, they were e-treated as described under Experimental and analysed in triplicate. Three standard additions of nitrate, ammonium and glucose were performed on five soils. Table 3 gives the recoveries of nitrate, ammonium and organic carbon (expressed as carbon produced from the glucose added).The average recoveries (n = 3) of these analytes ranged from 95.2 to 100.6, 94.5 to 101.5 and 86.9 to 95.3% for nitrate, ammonium and organic carbon, respectively. Any systematic error due to nitrogen labile species was observed in the determination of ammonium and nitrate. The organic carbon recoveries were smaller than those for the other analytes, consistent with reported data1 (maximum recoveries around 85–90%).The results obtained for 20 samples of five different soils withdrawn from different depths are listed in Table 4. The contents in the three analytes depend on the type of soil and depth; the concentration of organic carbon decreased with increasing depth. On the other hand, the maximum concentra- Table 4 Determination of nitrate, ammonium and organic carbon in soil samples (n = 3) from different depths by using the proposed and conventional methods1 Nitrate/mg kg21 Ammonium/mg kg21 Organic carbon/g kg21 Proposed Conventional Proposed Conventional Proposed Conventional Depth/cm Sample method method method method method method 0–15 1 8.1 ± 0.2 8.0 ± 0.2 5.1 ± 0.1 5.9 ± 0.2 4.4 ± 0.1 4.5 ± 0.1 2 11.5 ± 0.2 11.6 ± 0.3 8.9 ± 0.2 9.5 ± 0.3 5.5 ± 0.1 4.8 ± 0.2 3 3.3 ± 0.1 4.0 ± 0.1 6.3 ± 0.2 5.8 ± 0.1 6.3 ± 0.2 6.4 ± 0.2 4 2.2 ± 0.1 1.9 ± 0.1 3.5 ± 0.1 3.4 ± 0.1 3.2 ± 0.1 3.5 ± 0.1 5 9.9 ± 0.2 10.0 ± 0.2 7.4 ± 0.2 7.4 ± 0.2 4.4 ± 0.1 4.2 ± 0.2 15–30 1 22.5 ± 0.5 21.9 ± 0.6 16.3 ± 0.3 16.9 ± 0.4 2.5 ± 0.1 2.5 ± 0.1 2 9.3 ± 0.1 10.0 ± 0.2 7.2 ± 0.2 7.1 ± 0.2 3.7 ± 0.1 3.3 ± 0.1 3 3.8 ± 0.1 4.0 ± 0.1 6.0 ± 0.2 6.8 ± 0.2 3.4 ± 0.1 3.8 ± 0.1 4 20.9 ± 0.4 21.6 ± 0.6 13.8 ± 0.4 14.8 ± 0.3 3.5 ± 0.1 3.4 ± 0.1 5 9.3 ± 0.2 8.7 ± 0.2 6.3 ± 0.1 6.9 ± 0.1 3.7 ± 0.1 3.2 ± 0.1 30–60 1 13.4 ± 0.3 13.3 ± 0.4 8.2 ± 0.2 8.0 ± 0.3 1.7 ± 0.1 1.8 ± 0.1 2 2.7 ± 0.1 3.1 ± 0.1 4.7 ± 0.1 4.9 ± 0.1 2.4 ± 0.1 2.3 ± 0.1 3 3.3 ± 0.1 3.3 ± 0.1 4.5 ± 0.1 5.0 ± 0.2 1.7 ± 0.1 1.6 ± 0.1 4 16.9 ± 0.4 18.0 ± 0.4 9.8 ± 0.2 10.0 ± 0.3 2.8 ± 0.1 2.9 ± 0.1 5 7.8 ± 0.1 7.6 ± 0.2 5.8 ± 0.2 5.7 ± 0.2 3.2 ± 0.1 3.1 ± 0.1 60–90 1 2.9 ± 0.1 2.4 ± 0.1 4.7 ± 0.1 4.5 ± 0.1 1.6 ± 0.1 1.7 ± 0.1 2 2.5 ± 0.1 2.8 ± 0.1 4.1 ± 0.1 4.4 ± 0.2 2.1 ± 0.1 2.0 ± 0.1 3 2.8 ± 0.1 3.0 ± 0.1 4.2 ± 0.1 4.6 ± 0.1 1.1 ± 0.1 1.2 ± 0.1 4 10.8 ± 0.2 10.3 ± 0.3 7.8 ± 0.2 7.9 ± 0.2 2.5 ± 0.1 2.3 ± 0.1 5 5.4 ± 0.1 6.1 ± 0.2 4.6 ± 0.1 4.3 ± 0.1 2.4 ± 0.1 2.6 ± 0.1 Table 3 Recoveries (n = 3) obtained from different types of soils spiked with nitrate (8 mg kg21), ammonium (1 mg kg21) and carbon (8 g kg21 as glucose) Sample No.Nitrate Ammonium Organic carbon* 1 100.2 ± 2.1 94.5 ± 2.7 93.1 ± 2.0 2 95.2 ± 1.8 101.1 ± 2.8 95.3 ± 2.3 3 100.6 ± 2.0 96.8 ± 2.4 94.6 ± 2.1 4 98.3 ± 1.9 101.5 ± 2.9 86.9 ± 1.9 5 99.1 ± 2.2 100.6 ± 3.0 89.8 ± 1.9 * Expressed as glucose. 312 Analyst, April 1997, Vol. 122tions of nitrate and ammonium were significantly increased in samples 1 and 4, obtained at depths between 15 and 30 cm, possibly as a result of migration of these analytes from ground soil (0–15 cm) to deeper layers from the effect of rain water.The three methods were compared with the existing batch alternatives for the determination of nitrate,17 organic carbon26 and ammonium23 based on the same photometric reactions. As can be seen in Table 4, the concentrations provided by the conventional and proposed methods for the three analytes were fairly consistent, with Student’s t-values of 0.2, 0.4 and 0.5 for nitrate, ammonium and organic carbon, respectively (the critical value of t for n = 20 and P = 0.05 is 2.1).Conclusions We have demonstrated the suitability of the continuous configuration used for the integrated sequential determination of nitrate, ammonium and organic carbon in soil samples simply by switching among photometric reagents via the manifold valves.The use of a continuous filtration probe prior to introducing the soil extract into the flow system avoids clogging of the manifold. The proposed method surpasses conventional procedures for the determination of these three parameters in soils; it involves little sample manipulation, which results in increased precision and throughput, and in high automatability. The authors are grateful to the Spanish Direcci�on General de Investigaci�on Cient�ýfica y T�ecnica (Project No.PB95-0977) for financial support. They also acknowledge Dr. Lop�ez Bellido for kindly supplying the soil samples used. References 1 Page, A. L., Miller, R. H., and Keeney, D. R., Methods of Soil Analysis. Part 2. Chemical and Microbiological Properties, American Society of Agronomy and Soil Science Society of America, Madison, WI, 2nd edn., 1982. 2 Valc�arcel, M., and Luque de Castro, M. D., Flow Injection Analysis: Principles and Applications, Ellis Horwood, Chichester, 1987. 3 R°u�zi�cka, J., and Hansen, E. H., Flow Injection Analysis, Wiley, New York, 2nd edn., 1988. 4 Valc�arcel, M., and Luque de Castro, M. D., Fresenius’ J. Anal. Chem., 1990, 337, 662. 5 Lazaro, F., Luque de Castro, M. D., and Valc�arcel, M., Anal. Chim. Acta, 1991, 242, 283. 6 Arruda, M. A. Z., Gallego, M., and Valc�arcel, M., J. Anal. At. Spectrom., 1995, 10, 501. 7 McLeod, S., Anal. Chim. Acta, 1992, 266, 107. 8 Zhi, Z., R�ýos, A., and Valc�arcel, M., J. Environ. Anal.Chem., 1994, 57, 279. 9 Luque de Castro, M. D., and Valc�arcel, M., Analyst, 1984 , 109, 413. 10 R�ýos, A., Luque de Castro, M. D., and Valc�arcel, M., Analyst, 1984, 109, 1487. 11 Ca�nete, F., R�ýos, A., Luque de Castro, M. D., and Valc�arcel, M., Analyst, 1988, 113, 739. 12 Farran, A., Cortina, J. L., de Pablo, J., and Barcel�o, D., Anal. Chim. Acta, 1990, 234, 119. 13 Ballesteros, E., Gallego, M., and Valc�arcel, M., Anal. Chem., 1990, 62, 1587. 14 Ballesteros, E., Gallego, M., and Valc�arcel, M., Anal.Chem., 1993, 65, 1773. 15 Bajic, S. J., and Jaselskis, B., Talanta, 1985, 32, 115. 16 Marczenko, Z., Spectrophotometric Determination of Elements, Ellis Horwood, Chichester, 1976. 17 Kojlo, A., and Gorodkiewicz, E., Anal. Chim. Acta, 1995, 302, 283. 18 Windholz, M., The Merck Index. An Encyclopedia of Chemicals, Drugs and Biologicals, Merck, Rahway, NJ, 10th edn., 1983. 19 Standard Methods for the Examination of Water and Wastewater, ed. Greenberg, A.E., Connors, J. J., and Jenkins, D., American Public Health Association, Washington, DC, 15th edn., 1980, pp. 370– 373. 20 Berm�udez, B., R�ýos, A., Luque de Castro, M. D., and Valc�arcel, M., Talanta, 1988, 35, 810. 21 Krug, F. J., R°u�zi�cka, J., and Hansen, E. H., Analyst, 1979, 104, 47. 22 Zhi, Z., R�ýos, A., and Valc�arcel, M., Anal. Chim. Acta, 1994, 293, 163. 23 Schulze, G., Liu, C. Y., Brodowski, M., Elsholz, O., Frenzel, W., and M�oller, J., Anal. Chim. Acta, 1988, 214, 121. 24 Cerd`a, A., Oms, M. T., Forteza, R., and Cerd`a, V., Anal. Chin. Acta, 1995, 311, 165. 25 Sims, J. R., and Haby, V. A., Soil Sci., 1971, 112, 137. 26 Kempers, A.er 19, 1996 Accepted January 13, 1997 Analyst, April 1997, Vol. 122 313 Integrated Automatic Determination of Nitrate, Ammonium and Organic Carbon in Soil Samples Evaristo Ballesteros, Angel R�ýos and Miguel Valc�arcel* Department of Analytical Chemistry, Faculty of Sciences, University of C�ordoba, E-14004 C�ordoba, Spain An automated photometric method for the sequential determination of nitrate, ammonium and organic carbon in soils by use of a single-line flow injection assembly is reported, but differently programmed for the assay of the individual analytes.These analytes, which are determined by using suitable photometric reagents, are directly related to the nutritional properties of soil. A filtration probe was used for the continuous filtration of soil extract.The precision, expressed as RSD, was 2.0, 2.6 and 2.1% for nitrate, ammonium and organic carbon, respectively. The proposed method features a high throughput and low reagent consumption, and requires minimal sample handling. Its performance was tested in routine analyses of soil samples, all with satisfactory results. Keywords: Flow system; automatic photometric determination; nitrate; ammonium; organic carbon; soil samples The bound inorganic nitrogen in soils is predominantly in the form of ammonium and nitrate; nitrite is seldom present in detectable amounts, so its determination is usually unwarranted.Several other inorganic forms of nitrogen, including hydroxylamine, hyponitrous acid, and imidonitric acid (nitramide), have been postulated as intermediates in microbial transformations of nitrogen; however, none of these compounds has been detected in soils. Organic carbon is contained in the soil organic fraction, which consists of cells of microorganisms, plant and animal residues at various stages of decay, stable humus synthesized from residues and highly carbonized compounds such as charcoal, graphite and coal.1 Flow injection (FI) methodology is a powerful tool for performing a wide variety of on-line treatments.Over 99% of reported FI applications are based on the introduction of pretreated liquid samples.2,3 The preliminary operations of the analytical process (sampling and sample conditioning and treatment) are rarely considered systematically in FI.These steps are crucial because they are tedious, time consuming and the source of major errors affecting the final results. Direct analyses of heterogeneous or solid samples by use of automatic systems pose special difficulties.4 Automatic leaching can be accomplished in a continuous-flow configuration by using a dissolving stream together with an external source of energy (e.g.a high temperature, an electrical discharge or ultrasound). Ultrasonic irradiation has been applied to the direct determination of iron in plant materials and available boron in soil.5 A continuous system involving microwave treatment of shellfish samples was used for aluminium determination by ETAAS.6 McLeod7 developed a continuous method for the determination of nitrogen in soils and plants from a Kjeldahl digest, by using a flow system with a microdistillation unit.Zhi et al.8 reported a flow system integrating a soil sample pre-treatment unit and a flow manifold for the sequential determination of nitrate and nitrite. Few methods for convenient, automatic multi-determinations are available.9 In previous work, we developed a method for the determination of pH, nitrite, copper, iron and aluminium in waste water by use of a continuous configuration including both potentiometric and photometric detectors,10 in addition to a determination for pH, conductivity, residual chlorine, ammonium and nitrate ion in water with an unsegmented flow configuration.11 More recently, continuous flow systems have been integrated with chromatographic equipment in order to automate sample preparation prior to introduction into a chromatograph, thus facilitating multi-determinations.12–14 The aim of this work was to develop a methodology for the integrated determination of ammonium, nitrate and organic carbon in soils by using a flow module including a diode-array detector.Previously, nitrate and ammonium have been widely determined by using flow injection methods; in this work, however, the determination of these two analytes was integrated with that of organic carbon. Soil sample extracts were cleaned up in the manifold by means of a filtration probe accommodating a hydrophobic PTFE membrane. Experimental Apparatus A Hewlett-Packard (Avondale, PA, USA) Model 8451A diodearray spectrophotometer furnished with an HP Think-Jet printer was used.The proposed continuous system consisted of two Gilson (Worthington, OH, USA) Minipuls-3 peristaltic pumps fitted with poly(vinyl chloride) tubes; five Rheodyne (Cotati, CA, USA) Model 5041 valves; PTFE tubing (0.5 mm id) for coils; a Hellma (Jamaica, NY, USA) Model 178.12 QS flow cell (inner volume 18 ml; pathlength 10 mm); and a reduction column (50 3 3 mm id) packed with 0.8–1.0 mm grain-size cadmium granules. The filtration probe was developed in our laboratory (Fig. 1), from two glass tubes [1.5 cm and 1.0 cm id, for the internal (B) and external tube (C), respectively] and a PTFE tube (A, 0.5 mm id) connected with the pump tube. The PTFE tube was held on a PTFE support (D). Glass tubes B and C were sealed with a hydrophobic membrane PTFE from Millipore (Bedford, MA, USA) (Mitex, 1.0 mm pore size) through an O-ring (see details in Fig. 1). The filtration probe was included in the tube for aspiration of the sample in the flow system.When the probe was used to filter soil extracts (in 2 m KCl) for the determination of nitrate and ammonium, tube B was sealed with filter-paper (0.45 mm pore size); however, a paper filter cannot be used with the sulfuric acid–dichromate extract handled in the determination of organic carbon. At the end of each working day, the filtration probe was cleaned with distilled water. With this maintenance, the probe can be used for about 500 analyses. Reagents Phenol–nitroprusside reagent. An amount of 7 g of phenol (0.74 m) (Merck, Darmstadt, Germany) and 34 mg of sodium pentacyanonitrosylferrate(iii) (sodium nitroprusside) (1.14 mm) Analyst, April 1997, Vol. 122 (309–313) 309(Sigma, St. Louis, MO, USA) were dissolved in 80 ml of distilled water and then diluted to 100 ml. The solution was stored refrigerated in a dark-coloured bottle. Buffered hypochlorite reagent. A 1.480 g amount of sodium hydroxide was dissolved in 70 ml of distilled water (0.37 m) and supplied with 4.98 g of sodium monohydrogenphosphate (0.35 m) (Merck) and 20 ml of sodium hypochlorite solution (5–5.25% v/v).The solution was adjusted to pH 11.4–12.3 with sodium hydroxide and diluted to 100 ml. Ethylenediaminetetracetic acid (EDTA) reagent. A 6.0 g amount of EDTA disodium salt (0.16 m) (Merck) was dissolved in 80 ml of distilled water, adjusted to pH 7 and diluted to a final volume of 100 ml. Sulfanilamide solution. A 2.5 g amount of sulfanilamide (Merck) was dissolved in 13 ml of concentrated hydrochloric acid and diluted to 250 ml with distilled water (0.06 m).N-(1-naphthyl)ethylenediamine solution. A 0.25 g amount of the dihydrochloride (Merck) was dissolved in 250 ml of distilled water (3.9 mm). Conditioning reagent for the cadmium reduction column. A solution containing 0.1% (w/v) copper(ii) sulfate in 0.1 m EDTA was used for this purpose. Ammonium standard stock solution, 1 g l21. Prepared from ammonium chloride ( > 99.8%, Merck).Nitrate standard stock solution, 100 g l21. Prepared from sodium nitrate ( > 99.5%, Merck). Glucose standard solution, 100 g ml21 ( > 99.5%, Sigma). Sample Treatment Determination of nitrate and ammonium A battery of 50 samples was prepared as follows: 10 g of each soil sample was placed in a PTFE bottle and 100 ml of 2 m potassium chloride were added. The bottle was stoppered and the solution stirred mechanically for 1 h. The samples were then sequentially filtered through the filtration/dialysis probe and successively introduced into the manifold.Determination of organic carbon A battery of 50 samples was prepared as follows: 1 h soil sample was mixed with 8% potassium dichromate and 15 ml of concentrated sulfuric acid. The solution was stirred mechanically for 15 min, diluted to 100 ml and finally continuously filtered through the filtration/dialysis probe for introduction into the manifold.Procedure The manifold used for the sequential determination of nitrate, ammonium and organic carbon in soils is depicted in Fig. 2(a). Each analyte was quantified with a different reagent (A, B, or C, Table 1). The same tubes were used in all the determinations, the flow rate being adjusted as required by changing the speed of pumps. The positions (I or II) of the switching valves (SV1– Fig. 1 Scheme of the filtration probe used for the continuous introduction of soil extract.A, 0.5 mm id PTFE tubing; B and C, glass tubing; D, PTFE support; E, O-ring; F, PTFE hydrophobic membrane; G, filter paper (or hydrophobic PTFE membrane). Fig. 2 Overall flow manifold (a) and details of the continuous determination of nitrate (b), ammonium (c) and organic carbon (d). A, B and C, reagents (see Table 1); RC, reduction column; SV, switching valve; IV, injection valve; R, reactor; D, diode-array spectrophotometer. 310 Analyst, April 1997, Vol. 122SV4) used with each analyte are given in Table 1.For the determination of nitrate [Fig. 2(b)], the cadmiun reduction column was conditioned daily by passing a solution containing 0.1% copper(ii) sulfate in 0.1 m EDTA for 3 min. For the determination of ammonium [Fig. 2(c)], reactor R (300 cm) was heated to 50 °C by immersing the coil in a thermostated waterbath, 2 m potassium chloride was used as carrier in both determinations and 15% sulfuric acid in that for organic carbon [Fig. 2(d)]. Soil samples were passed through the filtration probe in all instances (Fig. 1). Results and Discussion The photometric technique has been widely used for determining nitrate, ammonium and organic carbon. The manifold depicted in Fig. 2(a) was optimized for the determination of these three parameters in soil samples by different photometric reactions. Determination of Nitrate Nitrate is usually determined by direct photometric methods or measurement after derivatization (usually reduction to nitrite); these methods utilize a homogeneous reductant or a heterogeneous reductant containing amalgamated zinc,15 amalgamated cadmium,16 copperized cadmium7,8,17 or copperized cadmium– silver.18 The nitrite thus produced is determined by diazotization with sulfanilamide and coupling with N-(1-naphthyl)ethylenediamine to form a highly coloured azo dye that is measured photometrically.19 The optimum flow conditions for the determination of nitrate [Fig. 2(a)] are given in Table 1.The effect of the concentrations of the reagents [sulfanilamide, N-(1-naphthyl)ethylenediamine and hydrochloric acid] were individually studied. Solutions of 1% sulfanilamide, 0.1% N-(1-naphthyl)ethylenediamine and 5% v/v hydrochloric acid were examined. Also, a solution consisting of a mixture of the three reagents at these concentrations was tested for reaction with nitrite (after reduction of nitrate on the cadmium column). Although the reaction took place, the sensitivity was about 40% lower.The best results were obtained when the reagents were placed in two solutions consisting of 0.1% N-(1-naphthyl)ethylenediamine on the one hand and of 1% sulfanilamide and 5% v/v HCl on the other. In a previous study,20 we found the most efficient reduction of nitrate to nitrite was achieved with a column packed with copperized cadmium in medium to large grain sizes. A column packed with 0.8–1.0 mm grain-sized cadmium granules and copperized with a solution containing 0.1% v/v copper(ii) sulfate in 0.1 m EDTA was finally used. The analytical features for the determination of nitrate are summarized in Table 2.The sampling frequency was relatively low owing to the time needed for reproducible signals to be obtained (about 3 min). Determination of Ammonium There are many methods for the determination of ammonium, some of which have been adapted to the flow injection technique (e.g., several methods based on the Nessler21,22 and Berthelot23,24 reactions).We tested two configurations based on Table 1 FI conditions for the determination of nitrate, ammonium and organic carbon Organic Nitrate Ammonium carbon A N-(1-Naphthyl)ethylenediamine Buffered hypochlorite — q/ml min21 0.45 0.35 B Sulfanilamide Phenol–nitroprusside reagent — q/ml min21 0.45 0.35 Conditioning reagent Copper(ii) sulfate in EDTA — — q/ml min21 0.45 Carrier 2 m KCl 2 m KCl 15% H2SO4 q/ml min21 2.90 1.00 2.90 Sample q/ml min21 0.75 0.55 0.75 C — EDTA — q/ml min21 0.33 SV1 Ia I II SV2 I II I SV3 I II II SV4 I–II I I IV loop 200 80 200 R, temperature/°C — 50 — Pump 1 On On On Pump 2 On On Off l/nm 540 636 600 a Switching valves (SV) were in position I or II.Table 2 Figures of merit of the determination of nitrate, ammonium and organic carbon Regression Detection Linear RSD‡ Sampling Analyte Equation coefficient limit range (%) frequency/h21 Nitrate A = 0.090 [NO32] 2 0.002* 0.996 0.1* 0.25–4* 2.0 15 Ammonium A = 0.039 [NH4 +] + 0.008* 0.998 0.1* 0.4–8* 2.6 30 Organic carbon A = 0.170 [C] + 0.009† 0.998 0.05† 0.1–2† 2.1 60 * Concentration in mg ml21.† Concentration in % w/v carbon as glucose. ‡ Concentration for RSD (n = 11): 2 mg ml21 nitrate, 3 mg ml21 ammonium and 0.8% w/v carbon. Analyst, April 1997, Vol. 122 311these two reactions for the determination of ammonium. The method based on the Nessler reaction posed many problems with the soil samples, possibly because of precipitated calcium hydroxide and other interferents present in soil, the alkaline pH needed for the reaction and the decreased signal resulting from the high concentration of potassium chloride needed to extract the ammonium from the soil.The system depicted in Fig. 2 (based on the Berthelot reaction) was freed from the above problems posed by calcium and other interferents present in the soil by using a stream of EDTA that was merged with the sample. In this case, a high potassium chloride concentration had no effect on the sensitivity.Table 1 gives the optimum sample, reagent and carrier flow rates used for the determination of ammonium. The phenol–nitroprusside solution contained 70 g l21 phenol and 0.34 g l21 sodium nitroprusside. The optimum pH range for the hypochlorite buffer [containing 14.8 g l21 sodium hydroxide, 49.8 g l21 sodium monohydrogenphosphate and 20% of sodium hypochlorite solution (5–5.25% v/v)] was 11.4–12.3.Table 2 shows the analytical figures of merit of the method. The precision, assessed on 11 samples containing 4 mg ml21 ammonium, was 2.6% (RSD). Determination of Organic Carbon The proposed continuous method for the determination of organic carbon in soil (Fig. 2) was based on the manual method of Sims and Haby,25 in which the organic carbon is oxidized by using a potassium dichromate solution in sulfuric acid and photometric measurement at 600 nm of the CrIII produced in the reduction of dichromate.The flow conditions used in the determination of organic carbon are summarized in Table 1. Table 2 gives the figures of merit of the method; organic carbon contents are expressed as the percentage (m/v) of carbon in samples containing various amounts of glucose (the method uses glucose as a standard). Alternatively, the organic matter content of a soil may be estimated by multiplying the organic carbon concentration by a constant factor based on the proportion of carbon in the organic matter; however, this factor varied between 1.7 and 2 depending on the soil sample and must be determined experimentally for each soil.Therefore, ‘organic carbon concentration’ is preferable to the term ‘soil organic matter content’ because the latter cannot be determined directly.1 Analysis of Soil Samples The proposed method was applied to the determination of nitrate, ammonium and organic carbon in 20 soil samples from different countryside areas of C�ordoba (southern Spain).Samples were colcted by means of a sampling probe and dried. Subsequently, they were pre-treated as described under Experimental and analysed in triplicate. Three standard additions of nitrate, ammonium and glucose were performed on five soils. Table 3 gives the recoveries of nitrate, ammonium and organic carbon (expressed as carbon produced from the glucose added). The average recoveries (n = 3) of these analytes ranged from 95.2 to 100.6, 94.5 to 101.5 and 86.9 to 95.3% for nitrate, ammonium and organic carbon, respectively.Any systematic error due to nitrogen labile species was observed in the determination of ammonium and nitrate. The organic carbon recoveries were smaller than those for the other analytes, consistent with reported data1 (maximum recoveries around 85–90%). The results obtained for 20 samples of five different soils withdrawn from different depths are listed in Table 4. The contents in the three analytes depend on the type of soil and depth; the concentration of organic carbon decreased with increasing depth.On the other hand, the maximum concentra- Table 4 Determination of nitrate, ammonium and organic carbon in soil samples (n = 3) from different depths by using the proposed and conventional methods1 Nitrate/mg kg21 Ammonium/mg kg21 Organic carbon/g kg21 Proposed Conventional Proposed Conventional Proposed Conventional Depth/cm Sample method method method method method method 0–15 1 8.1 ± 0.2 8.0 ± 0.2 5.1 ± 0.1 5.9 ± 0.2 4.4 ± 0.1 4.5 ± 0.1 2 11.5 ± 0.2 11.6 ± 0.3 8.9 ± 0.2 9.5 ± 0.3 5.5 ± 0.1 4.8 ± 0.2 3 3.3 ± 0.1 4.0 ± 0.1 6.3 ± 0.2 5.8 ± 0.1 6.3 ± 0.2 6.4 ± 0.2 4 2.2 ± 0.1 1.9 ± 0.1 3.5 ± 0.1 3.4 ± 0.1 3.2 ± 0.1 3.5 ± 0.1 5 9.9 ± 0.2 10.0 ± 0.2 7.4 ± 0.2 7.4 ± 0.2 4.4 ± 0.1 4.2 ± 0.2 15–30 1 22.5 ± 0.5 21.9 ± 0.6 16.3 ± 0.3 16.9 ± 0.4 2.5 ± 0.1 2.5 ± 0.1 2 9.3 ± 0.1 10.0 ± 0.2 7.2 ± 0.2 7.1 ± 0.2 3.7 ± 0.1 3.3 ± 0.1 3 3.8 ± 0.1 4.0 ± 0.1 6.0 ± 0.2 6.8 ± 0.2 3.4 ± 0.1 3.8 ± 0.1 4 20.9 ± 0.4 21.6 ± 0.6 13.8 ± 0.4 14.8 ± 0.3 3.5 ± 0.1 3.4 ± 0.1 5 9.3 ± 0.2 8.7 ± 0.2 6.3 ± 0.1 6.9 ± 0.1 3.7 ± 0.1 3.2 ± 0.1 30–60 1 13.4 ± 0.3 13.3 ± 0.4 8.2 ± 0.2 8.0 ± 0.3 1.7 ± 0.1 1.8 ± 0.1 2 2.7 ± 0.1 3.1 ± 0.1 4.7 ± 0.1 4.9 ± 0.1 2.4 ± 0.1 2.3 ± 0.1 3 3.3 ± 0.1 3.3 ± 0.1 4.5 ± 0.1 5.0 ± 0.2 1.7 ± 0.1 1.6 ± 0.1 4 16.9 ± 0.4 18.0 ± 0.4 9.8 ± 0.2 10.0 ± 0.3 2.8 ± 0.1 2.9 ± 0.1 5 7.8 ± 0.1 7.6 ± 0.2 5.8 ± 0.2 5.7 ± 0.2 3.2 ± 0.1 3.1 ± 0.1 60–90 1 2.9 ± 0.1 2.4 ± 0.1 4.7 ± 0.1 4.5 ± 0.1 1.6 ± 0.1 1.7 ± 0.1 2 2.5 ± 0.1 2.8 ± 0.1 4.1 ± 0.1 4.4 ± 0.2 2.1 ± 0.1 2.0 ± 0.1 3 2.8 ± 0.1 3.0 ± 0.1 4.2 ± 0.1 4.6 ± 0.1 1.1 ± 0.1 1.2 ± 0.1 4 10.8 ± 0.2 10.3 ± 0.3 7.8 ± 0.2 7.9 ± 0.2 2.5 ± 0.1 2.3 ± 0.1 5 5.4 ± 0.1 6.1 ± 0.2 4.6 ± 0.1 4.3 ± 0.1 2.4 ± 0.1 2.6 ± 0.1 Table 3 Recoveries (n = 3) obtained from different types of soils spiked with nitrate (8 mg kg21), ammonium (1 mg kg21) and carbon (8 g kg21 as glucose) Sample No.Nitrate Ammonium Organic carbon* 1 100.2 ± 2.1 94.5 ± 2.7 93.1 ± 2.0 2 95.2 ± 1.8 101.1 ± 2.8 95.3 ± 2.3 3 100.6 ± 2.0 96.8 ± 2.4 94.6 ± 2.1 4 98.3 ± 1.9 101.5 ± 2.9 86.9 ± 1.9 5 99.1 ± 2.2 100.6 ± 3.0 89.8 ± 1.9 * Expressed as glucose. 312 Analyst, April 1997, Vol. 122tions of nitrate and ammonium were significantly increased in samples 1 and 4, obtained at depths between 15 and 30 cm, possibly as a result of migration of these analytes from ground soil (0–15 cm) to deeper layers from the effect of rain water.The three methods were compared with the existing batch alternatives for the determination of nitrate,17 organic carbon26 and ammonium23 based on the same photometric reactions. As can be seen in Table 4, the concentrations provided by the conventional and proposed methods for the three analytes were fairly consistent, with Student’s t-values of 0.2, 0.4 and 0.5 for nitrate, ammonium and organic carbon, respectively (the critical value of t for n = 20 and P = 0.05 is 2.1).Conclusions We have demonstrated the suitability of the continuous configuration used for the integrated sequential determination of nitrate, ammonium and organic carbon in soil samples simply by switching among photometric reagents via the manifold valves. The use of a continuous filtration probe prior to introducing the soil extract into the flow system avoids clogging of the manifold. The proposed method surpasses conventional procedures for the determination of these three parameters in soils; it involves little sample manipulation, which results in increased precision and throughput, and in high automatability.The authors are grateful to the Spanish Direcci�on General de Investigaci�on Cient�ýfica y T�ecnica (Project No. PB95-0977) for financial support. They also acknowledge Dr. Lop�ez Bellido for kindly supplying the soil samples used. References 1 Page, A. L., Miller, R.H., and Keeney, D. R., Methods of Soil Analysis. Part 2. Chemical and Microbiological Properties, American Society of Agronomy and Soil Science Society of America, Madison, WI, 2nd edn., 1982. 2 Valc�arcel, M., and Luque de Castro, M. D., Flow Injection Analysis: Principles and Applications, Ellis Horwood, Chichester, 1987. 3 R°u�zi�cka, J., and Hansen, E. H., Flow Injection Analysis, Wiley, New York, 2nd edn., 1988. 4 Valc�arcel, M., and Luque de Castro, M. D., Fresenius’ J. Anal. Chem., 1990, 337, 662. 5 Lazaro, F., Luque de Castro, M. D., and Valc�arcel, M., Anal. Chim. Acta, 1991, 242, 283. 6 Arruda, M. A. Z., Gallego, M., and Valc�arcel, M., J. Anal. At. Spectrom., 1995, 10, 501. 7 McLeod, S., Anal. Chim. Acta, 1992, 266, 107. 8 Zhi, Z., R�ýos, A., and Valc�arcel, M., J. Environ. Anal. Chem., 1994, 57, 279. 9 Luque de Castro, M. D., and Valc�arcel, M., Analyst, 1984 , 109, 413. 10 R�ýos, A., Luque de Castro, M. D., and Valc�arcel, M., Analyst, 1984, 109, 1487. 11 Ca�nete, F., R�ýos, A., Luque de Castro, M. D., and Valc�arcel, M., Analyst, 1988, 113, 739. 12 Farran, A., Cortina, J. L., de Pablo, J., and Barcel�o, D., Anal. Chim. Acta, 1990, 234, 119. 13 Ballesteros, E., Gallego, M., and Valc�arcel, M., Anal. Chem., 1990, 62, 1587. 14 Ballesteros, E., Gallego, M., and Valc�arcel, M., Anal. Chem., 1993, 65, 1773. 15 Bajic, S. J., and Jaselskis, B., Talanta, 1985, 32, 115. 16 Marczenko, Z., Spectrophotometric Determination of Elements, Ellis Horwood, Chichester, 1976. 17 Kojlo, A., and Gorodkiewicz, E., Anal. Chim. Acta, 1995, 302, 283. 18 Windholz, M., The Merck Index. An Encyclopedia of Chemicals, Drugs and Biologicals, Merck, Rahway, NJ, 10th edn., 1983. 19 Standard Methods for the Examination of Water and Wastewater, ed. Greenberg, A. E., Connors, J. J., and Jenkins, D., American Public Health Association, Washington, DC, 15th edn., 1980, pp. 370– 373. 20 Berm�udez, B., R�ýos, A., Luque de Castro, M. D., and Valc�arcel, M., Talanta, 1988, 35, 810. 21 Krug, F. J., R°u�zi�cka, J., and Hansen, E. H., Analyst, 1979, 104, 47. 22 Zhi, Z., R�ýos, A., and Valc�arcel, M., Anal. Chim. Acta, 1994, 293, 163. 23 Schulze, G., Liu, C. Y., Brodowski, M., Elsholz, O., Frenzel, W., and M�oller, J., Anal. Chim. Acta, 1988, 214, 121. 24 Cerd`a, A., Oms, M. T., Forteza, R., and Cerd`a, V., Anal. Chin. Acta, 1995, 311, 165. 26 Kempers, A. J., Geoderma, 1974, 12, 201. Paper 6/07849D Received November 19, 1996 Accepted January 13, 1997 Analyst, April 1997, Vol. 122 313
ISSN:0003-2654
DOI:10.1039/a607849d
出版商:RSC
年代:1997
数据来源: RSC
|
5. |
Determination of Primary Explosive Azides in Environmental Samplesby Sequential Injection Amperometry |
|
Analyst,
Volume 122,
Issue 4,
1997,
Page 315-319
Roger T. Echols,
Preview
|
|
摘要:
Determination of Primary Explosive Azides in Environmental Samples by Sequential Injection Amperometry Roger T. Echols*a, Ryan R. Jamesa and Joseph H. Aldstadtb a Division of Science and Mathematics, University of Minnesota Morris, Morris, MN 56267, USA b Environmental Research Division, Argonne National Laboratory, 9700 S. Cass Avenue ER/203, Argonne, IL 60439, USA The application of flow injection methodology to the determination of trace concentrations of primary explosives is presented.The approach is demonstrated with a sequential injection amperometric method for the determination of the azide ion (N32). The proposed method can be applied to the determination of sodium azide or lead azide, a primary explosive, without regard to other sources of lead in environmental samples. The sequential injection system used for the analysis forms the basis for a proposed field-portable instrument for the analysis of primary explosives. A microporous gas permeable membrane in a gas diffusion unit (GDU) is used to separate the analyte from other anions that can also be oxidized at the amperometric cell.The behaviour of the GDU was optimized with respect to the pH of the donor stream and the timing of the preconcentration step. A study of anions that are commonly found in environmental samples showed that the species that will interfere with the analytical signal can be removed by the GDU. Results from three water samples that were spiked with 0.40 ppm of azide are presented.RSDs in the range 3–5% were typically obtained using the method. The useful working range of the method was linear up to 0.5 ppm and non-linear up to 20 ppm (second-order model). The limit of detection was 24.6 ppb. Keywords: Flow injection; sequential injection; explosives; lead azide; sodium azide; amperometry; gas diffusion The contamination that has resulted from the production, storage, testing and disposal of explosives and the concomitant health risks that affect military and civilian personnel is a growing environmental problem.Most of the contamination is a consequence of the presence of secondary explosives, which are the principle species used in most shells and munitions. Examples of secondary explosives include well known nitroaromatics, nitric esters and nitramines: 2,4,6-tritinitrotoluene (TNT), 2,6- or 2,4-dinitrotoluene (DNT), nitroglycerin, nitrocellulose, 2,4,6-N-tetranitro-N-methylaniline (tetryl), 1,3,5-trinitro-1,3,5-triazacyclohexane (RDX) and 1,3,5,7-tetranitro- 1,3,5,7-tetrazacyclooctane (HMX).Contaminated water that contains TNT and DNT is commonly termed ‘red’ or ‘pink water’. Another category of explosives that are of environmental interest are primary explosives, which are readily ignited species used to detonate secondary explosives. Various inorganic salts comprise the list of primary explosives and include lead styphnate, mercury fulminate and metal azides.Lead azide is a common primary explosive that has been used extensively in blasting caps and military ammunition as a consequence of its high sensitivity to shock or ignition.1 Although the majority of the research in the area of determinations of trace concentrations of explosives has focused on secondary explosives, there have been a few reports of methods for the determination of azide (from lead or sodium azide). Instrumental methods of analysis (spectrophotometric, chromatographic and electrochemical methods) have supplanted cerium2 and other oxidation–reduction titrimetric methods.Spectrophotometric methods for the determination of the concentration of azide ion are based on the ability of azide (N32) to form charge-transfer complexes. These methods include those based upon coloured metal–azide complexes that form between azide and iron (or an iron complex)3–6 or between azide and copper.7 Other spectrophotometric methods include a method based upon a simple derivitization reaction of azide with carbon disulfide8 and a method based upon the formation of a complex between azide and Arsenazo III.9 Spectrophotometric detection has also been used in chromatographic methods.Swarin and Waldo10 proposed a chromatographic method with post-column derivitization for the determination of azide and reported a detection limit of 10 ppb with 3% reproducibility. Of the few papers on the determination of lead azide by electroanalytical methods, only one can be considered a recent application of modern electrochemistry.Previous electrochemical methods include an amperometric titration11 and a polarographic determination.12 The recent work on the determination of azide by an electrochemical method involved potentiometric flow injection:13 an azide ion-selective electrode consisting of an iron(ii) or nickel(ii) bathophenanthroline–azide complex embedded in PVC was constructed and tested on a variety of solutions. The electrode was used to determine azide with a limit of detection of 0.8 ppm and to determine the solubility product constants of insoluble metal azides.A table in this paper13 summarized the limit of detection, relative error and interfering species for most of the methods for the determination of azide. This research was undertaken to develop a method for the determination of the azide ion and to design a sequential injection system that will form the basis for a small, fieldportable or process laboratory instrument for the determination of this analyte in waters and soil extracts.The method is applicable to sodium azide, which has been used as a pesticide and herbicide, and metal azide primary explosives. The analytical method is based upon amperometric detection of the azide ion at a potential at which the azide ion is oxidized14 under basic conditions. N32?3–2 N2 + e2 The on-line isolation of the azide analyte from other electroactive ions that will be present in environmental samples is an important feature of the method and is made possible by use of a gas permeable membrane and flow injection (FI) methodology.In particular, sequential injection (SI) techniques are used in the system; SI15,16 is a subset of FI methods that was introduced in the early 1990s and has been applied recently to solve a number of analytical problems. Gas diffusion methods Analyst, April 1997, Vol. 122 (315–319) 315have been used extensively in FI systems as a means by which analytes can be isolated from difficult matrices or preconcentrated. 17,18 The principle involves passing the species of interest (as a gas) across a membrane from a donor stream into an acceptor stream within the FI system.The ultimate goal of this work is to design a field or process monitor that can be used on-site for measuring the concentrations of primary explosives. A monitor the size of a small suitcase that can be interfaced to a portable computer and power supply is envisioned.It is hoped that the method and instrument will aid military and civilian officials in the disposal of explosives and in risk assessment of contaminated areas by providing the advantages of on-site determinations. Moreover, the ability of this method to distinguish between a particular lead azide and other compounds containing lead (e.g., tetraethyllead from gasoline spills and lead azide from explosives) is a useful feature for issues of hazardous waste disposal and accountability; methods of analysis that only determine the concentration of lead would be inadequate for assessing the contamination due to explosives.Experimental Apparatus An SI system was used for all of the work with the exception of the preliminary work, which was performed with a single-line FI system. The SI system is shown in Fig. 1 and was composed of the following: standard flow tubing (0.76 mm id) and connectors (Global FIA, Gig Harbor, WA, USA) syringe pump (Cavro, Sunnyvale, CA, USA) multiport valve (Valco, Houston, TX, USA), peristaltic pump (Alitea, Medina, WA, USA), gas diffusion unit (GDU) with a PTFE membrane (Global FIA) and UniJet amperometric cell (3 mm glassy carbon disc electrode) with Petit Ampere potentiostat (BAS, W.Lafayette, IN, USA). The SI system was controlled using FlowTEK software distributed by Global FIA and required the use of a computer (80286, DOS) and an A/D converter.The amperometric cell was operated at a potential of 1.0 V (versus Ag/AgCl) and both the peristaltic pump and syringe pump were operated at a volumetric flow rate of 1.0 ml min21. The GDU has a channel that is 16 cm in length, 1 mm in width and 0.8 mm in depth; the PTFE membrane is standard plumber’s tape. Reagents All chemicals were of analytical-reagent grade; ultra-pure water was provided from a high purity (18 MW) filtration system.Azide solutions were prepared from a sodium azide stock solution. Phosphate buffer solutions that were 0.01 m in KCl were prepared from monobasic and dibasic phosphate salts. Buffer solutions with a range of pH from 1.1 to 5.2 were used as the carrier stream in the SI system. The anions that were used in the study of interfering species were prepared from common salts. Natural water samples (tap, ground and Lake Michigan) were spiked with sodium azide. Procedures Most experiments were conducted using the SI procedure detailed in Table 1.The sample is loaded into the holding coil (see Fig. 1) and then propelled towards the GDU with an appropriate change in position on the multiport valve. The sample is then manipulated on the donor side of the GDU using forward and reverse flows of the carrier stream. The acceptor stream is stopped and then resumed to send the analyte to the amperometric cell; the procedure results in the preconcentration of the analyte across the membrane.Peak height and peak area data were obtained using the FlowTEK software. Optimization of the passage of azide through the GDU was conducted by variation of the two factors that were the most important for the transfer of the analyte across the membrane: the pH of the donor stream and the number of flow reversals used in the method. Peak height was used as the figure of merit. The response of 1 ppm solutions of common anions at 1.0 V was obtained by removing the GDU from the system and operating the SI system in a manner similar to a single-line FI manifold. The response to 5 ppm anion solutions with the complete SI system was measured using the standard procedure (Table 1).Natural water samples were spiked with 0.4 ppm of azide and analyzed using the standard procedure. Preliminary work with the SI amperometric system was conducted with bleach in phosphate buffer;19 this system is a Fig. 1 Schematic diagram of the sequential injection system.Flow tubing (0.76 mm id) is shown as solid lines and electrical connections as dashed lines. The gas diffusion unit has a channel that is 16 cm in length, 1 mm in width and 0.8 mm in depth. Table 1 Timed events of the sequential injection system Multiport Syringe Peristaltic Time/s valve pump pump Comment 0.00 Reset Stop Off Prepares MPV for next command 0.50 Step Turn MPV to port 2— connected to sample 0.75 Reset 1.75 Reverse Load sample into HC 2.50 Forward Begin flow of acceptor stream 21.75 Stop 22.75 Step Turn MPV to port 3— connected to GDU 23.00 Reset 24.00 Forward Propel sample to GDU (donor stream) 44.00 Off Stop flow of acceptor stream 50.00 Stop Flow agitations: 52.00 Reverse Reverse flow 62.00 Stop 64.00 Forward Forward flow 71.00 Stop 73.00 Reverse Reverse flow 81.00 Stop 83.00 Forward Forward flow 100.00 Forward Begin flow of acceptor stream 197.00 Stop 307.00 Off Stop flow of acceptor stream 308.00 Home Returns MPV to port 1 316 Analyst, April 1997, Vol. 122good ‘test’ reaction for setting-up and troubleshooting the SI system without generating hazardous waste. Preliminary work with azide was conducted with a simple single-line FI system. It was demonstrated that azide could be detected (in a single-line FI system) over a linear range from 0.10 to 2.5 ppm. The initial reaction conditions (phosphate buffer of pH 6.6 and potential) were based on work by Ward and Wright:14 a peak potential of 1.08 V was reported for the oxidation of azide at platinum electrodes in phosphate buffer solutions for a range of pH from 6.0 to 10.5.A potential of 1.0 V was used because of equipment limitations: the amperometric system did not have a current range large enough to accommodate background signals of the carrier solution at a peak potential of 1.08 V. Hence the analytical signals would be slightly larger with a different potentiostat. Results and Discussion Optimization of the Sequential Injection System The principal component of the SI system used for the determination of azide is the GDU that is used to isolate the azide from other anions.Anions that are present in environmental samples (e.g., chloride) will oxidize under these electrochemical conditions and interfere with the analytical method. The procedure for transporting azide across the microporous PTFE membrane in the GDU prevents the passage of most anions because they do not form neutral molecules at the pH used in this experiment.When the azide ion is protonated (pKa = 4.72), the HN3 passes through the membrane in the gaseous state. The relatively high volatility of HN3 (boilingpoint 37 °C) allows the separation to occur. Because the transport of HN3 across the gas permeable membrane is not 100% efficient, the pH of the donor stream and the timing of the SI system were chosen as the primary variables of the SI system for optimization. Sample volume was not varied as a consequence of choosing the time for an individual experiment as a design criterion for the work; the sample load time of 20.0 s (approximately 333 ml) was chosen on this basis to provide an overall experiment time of less than 5 min.Other parameters of the SI system (e.g., flow rate and tubing lengths) were not optimized with respect to the performance of the GDU because the effect of these variables is secondary to that of the donor stream pH and timing.In addition to microporous PTFE, a silicone-rubber membrane was used in the GDU. The results indicated that the transport of HN3 across the silicone-rubber membrane was significantly less than that with the PTFE membrane. The pH of the donor stream was optimized by varying the pH and by measuring the current at peak height for a 1 ppm solution. Results from this experiment are shown in Fig. 2. The mechanism for transport of azide across the gas permeable membrane involves the deprotonation of this species in the more basic acceptor stream (pH 6.6) and the formation of HN3(g) in the donor stream; it was expected that a donor stream pH less than the pKa of hydrazoic acid (4.72) would provide the most efficient transport across the membrane.The constant response for experiments for which the donor stream pH was less than the pKa confirmed the expectation and a pH of 3.78 was chosen for subsequent work. Moreover, calculation of the fraction of dissociation for hydrazoic acid confirmed that only 10% of azide is in the basic form at pH 3.78.The center of the sample (greatest concentration of analyte) was determined to be in the GDU at 40 s. The acceptor stream was halted for a time before and after this time in order to allow the diffusion of HN3 across the membrane and into the pH 6.6 acceptor stream. The method was optimized by performing a number of flow reversals using the syringe pump (0, 2, 4 and 8 flow reversals).The procedure has been termed ‘agitating the sample’ because of the similarity to what is done when solvent extractions are conducted in separating funnels. The results from an experiment investigating the technique are illustrated in Fig. 3. The greatest amount of azide was transferred across the membrane when four ‘agitations’ (four steps of reverse flow and forward flow) were performed within the established time interval that the acceptor stream was stopped. An approximately 25% signal enhancement was realized using this protocol (see Fig. 3). The ability to reverse the flow in the experiment is one of the advantages that SI methods have over traditional FI methods in which the flow is not typically reversed. Study of the Interference of Common Anions Anions that contribute significantly to the amperometric signal for azide were identified by Ward and Wright14 in their study of the electrochemistry of azide. Their list formed the basis for the species that were chosen for the study of interfering anions.Results from the study are shown in Table 2. The response of the anions in the study without the GDU in the SI system (first two columns in Table 2) can be grouped into three categories: (1) no response (perchlorate, bromide and nitrate); (2) low to moderate response (sulfite, phosphate, hypochlorite and cyanide); and (3) strong response (iodide, chloride, thiocyanate, nitrite and sulfide). The signals decrease significantly when the GDU is incorporated in the SI system for the step that separates the analyte from the matrix (second two columns in Table 2).The signals for sulfide, nitrite and hypochlorite are high enough at a concentration of 5 ppm to interfere with the azide signal. Other Fig. 2 Optimization of pH of the donor stream of the gas diffusion unit (1.0 ppm azide solutions). Error bars represent plus or minus one standard deviation (n = 3). Fig. 3 Optimization of number of flow reversals (‘agitations’) in the experimental method (1.0 ppm azide solutions).Error bars represent plus or minus one standard deviation (n = 3). Analyst, April 1997, Vol. 122 317anions (with the exception of iodide, which has a negative response) have corrected signals that are scattered around the blank signal and lie within the experimental uncertainty (95% confidence interval) of the SI experiment for the determination of azide (±0.019). The passage of hydrogen sulfide (from sulfide), nitrous acid (from nitrite) and hypochlorous acid (from hypochlorite) through the membrane was expected on the basis of the chemistry of those species: H2S has a pKa1 of 7.02 and a pKa2 of 13.9; HNO2 has a pKa of 3.15 and HOCl has a pKa of 7.53.20 The three anions are subjected to the same chemistry as the azide ion: conjugate acids are formed in the donor stream before passing across the membrane and before being oxidized at the working electrode.Nitrite is dissociated to a greater extent than the other two species, which might explain the larger signal for nitrite than sulfide and hypochlorite.On the basis of previous work,14 it was expected that hydrogen cyanide (from cyanide) would interfere to a greater extent than that observed in the study of interference. At the pH used in the experiment, the cyanide should cross the membrane as HCN (pKa = 9.21) and remain in the neutral form; hence the oxidation of cyanide ion would not be expected.It was difficult to determine how much of the azide in each sample was passing through the PTFE membrane (efficiency of transport) as a consequence of the agitations (reversals of flow) that were an integral part of the procedure. A comparison could be made on the basis of the signals from experiments including and excluding the GDU from the SI system. The donor stream was connected to the acceptor stream such that the GDU was by-passed and a 1.0 ppm solution of azide was run through the procedure (with agitations).The signal from this experiment was 0.98 nA, which was 40% larger than that obtained using the GDU and the normal method (signal of 0.58 nA). The simple comparison does not indicate that transport of the azide across the membrane is 60% efficient in an absolute sense because the azide is preconcentrated during the steps of forward and reverse flow while the acceptor stream is halted. Determination of Azide in Water Samples Results from the spiked samples of Lake Michigan, ground and tap waters are summarized in Table 3.The calculation of the concentrations was based on a regression line obtained using standard azide solutions (0–0.5 ppm): y = 0.0131 (nA) +1.21 (nA ppm21) x with r2 = 0.990. The blank signal of 0.0219 nA was subtracted from each standard and sample; the response for the blank is caused by the change in background current that results from the stopping of the donor stream during the preconcentration/separation step using the GDU.The limit of detection was determined to be 24.6 ppb of azide on the basis of three times the standard deviation of the blank signals (0.0429 nA). While the response of the standards was linear over this range (0–0.5 ppm), the response for a broader range of standards (0–20 ppm) was non-linear and was fitted by a second-order polynomial model: y = 0.0798 + 0.597x 2 0.138x2 (r2 = 0.988); the non-linearity of the upper range of the standards was thought to be a feature of the gas diffusion unit.The precision for repeated experiments was influenced by the experimental procedure in which the donor stream was halted and the baseline current allowed to decay before flow was resumed. Nevertheless, RSDs of 3–5% were typical for calibration standard solutions (clean samples). For the spiked water samples that contained interfering anions, the RSDs were 1.7, 3.0 and 7.7%. The accuracy of the results was assessed by examination of the percentage difference from the expected value of 0.40 ppm azide (amount spiked) and the population means (Table 3).Of the three samples, the expected value lies outside the population mean at 95% confidence for the ground water sample. The three water samples (not spiked with azide) were introduced directly into the amperometric cell to verify that there were other anions present and that the GDU was separating azide from the matrix. Normalized current–time profiles (‘peaks’) for tap water and ground water samples that were introduced directly into the amperometric cell (GDU removed from the SI system) are illustrated in Fig. 4 (A and B). The interfering signal from background anions that would result from the determination being performed without the separation of the analyte using the GDU is clearly illustrated. Normalized peaks for the spiked tap water sample (C) and the unspiked tap water sample (D), which were obtained with the GDU in the SI system, are also shown in Fig. 4. The height of the peak of the background anion signal when the GDU is not in use (A) is twice that of the spiked tap water sample when the GDU is used (C). Additionally, the elimination of the interfering signal can be seen by comparing peaks A and D in Fig. 4: the GDU completely eliminates the background anion signal. Future Work Future work will include an assessment of the field performance of the proposed SI instrument, which will necessitate the inclusion of an extraction procedure for lead azide.A study of the discrimination of lead species (e.g., determination of lead azide in the presence of other sources of lead) will also be included in future work. The principles applied in this work for the determination of the target analyte (azide) can be employed for the determination of other explosives. Other analytical methods that incorporate FI methodologies can be designed for the different chemistries needed to determine the concentrations of other explosives that are of environmental interest (e.g., TNT Table 2 Results from the study of the interference of common anions Without gas diffusion unit With gas diffusion unit Signal/nA Percentage Signal/nA Percentage (1 ppm of azide (5 ppm of azide Anion solutions) signal solutions)* signal Azide 2.731 — 1.365 — Bromide 0.004 0.1 20.015 21.1 Chloride 1.173 42.9 20.003 20.2 Cyanide 0.063 2.3 0.015 1.1 Hypochlorite 0.225 8.2 0.068 5.0 Iodide 3.429 122.1 20.021 21.5 Nitrate 0.006 0.2 0.009 0.7 Nitrite 2.030 74.3 0.211 15.5 Perchlorate 0.009 0.3 20.007 20.5 Phosphate 0.029 1.0 20.011 20.8 Sulfide 3.110 113.9 0.132 9.7 Sulfite 0.254 9.0 20.003 20.2 Thiocyanate 0.753 27.6 0.006 0.4 * Signal is corrected for response of blank (0.053 nA).Table 3 Results from water samples spiked with 0.40 ppm of sodium azide Difference Concentration, from 0.400 Sample ppm* RSD (%) ppm (%) Lake Michigan water 0.385 ± 0.047 7.7 23.8 Tap water 0.409 ± 0.011 1.7 2.3 Ground water 0.429 ± 0.020 3.0 7.3 * 95% confidence limits of the mean; n = 4. 318 Analyst, April 1997, Vol. 122and DNT). An interesting possibility for the determination of TNT and DNT is the formation of coloured sulfite complexes that can be detected with visible spectrophotometry.21 The Division of Educational Programs of Argonne National Laboratory is gratefully acknowledged for supporting R.T.E. through the Faculty Research Program and R.R.J.through the Science and Engineering Research Semester Program. Argonne National Laboratory is operated by the University of Chicago under Contract No. W-31-109-ENG-38. This research was presented in part at the 1996 Pittsburgh Conference and at the 1996 FACSS Conference. References 1 Yinon, J., and Zitrin, S., Modern Methods and Applications in Analysis of Explosives, Wiley, New York, 1993. 2 Grove, E. L., Braman, R. S., Combs, H. F., and Nicholson, S.B., Anal. Chem., 1962, 34, 682. 3 Roberson, C. E., and Austin, C. M., Anal. Chem., 1957, 29, 855. 4 Anton, A., Dodd, J. G., and Harvey, A. E., Anal. Chem., 1960, 32, 1209. 5 Ilcheva, L., and Todorova, G., Acta Chim. Acad. Sci. Hung., 1979 (102) 113. 6 Mehra, M. C., and Garvie, R., Microchem. J., 1980, 25, 223. 7 Neves, E. A., de Oliveira, E., and Sant’agostino, L., Anal. Chim. Acta, 1976, 87, 243. 8 Franco, D. W., Neves, E. A., and Andrade, J. F., Anal. Lett., 1977, 10, 243. 9 Kubaszewski, E., and Kurzawa, Z., Chem. Anal. (Warsaw), 1985, 30, 609. 10 Swarin, S. J., and Waldo, R. A., J. Liq. Chromatogr., 1982, 5, 597. 11 Dziegiec, J., and Ignaczak, M., Acta Chim. Soc. Sci. Lodz, 1971, 16, 69. 12 Bryant, J. I., and Kemp, M. D., Anal. Chem., 1960, 32, 759. 13 Hassan, S. S., El Zawawy, F. M., Marzouk, S. A. M., and Elnemma, E. E., Analyst, 1992, 117, 1683. 14 Ward, G. A., and Wright, C. M., J. Electroanal. Chem., 1964, 8, 302. 15 Ruzicka, J., and Marshall, G.D., Anal. Chim. Acta, 1990, 237, 329. 16 Gubeli, T., Christian, G. D., and Ruzicka, J., Anal. Chem., 1991, 69, 2407. 17 Kuban, V., Crit. Rev. Anal. Chem., 1992, 23, 323, and references cited therein. 18 Fang, Z., Flow Injection Separation and Pre-concentration, VCH, New York, 1993. 19 Tsaouis, A. N., and Huber, C. O., Anal. Chim. Acta, 1985, 178, 319. 20 Martell, A. E., and Smith, R. M., Critical Stability Constants, Plenum Press, New York, 1974. 21 Jenkins, T. F., US Army Corps of Engineers Special Report 90-38, Cold Regions Research and Engineering Laboratory, Hanover, NH, USA, 1990.Paper 6/05662H Received August 13, 1996 Accepted January 7, 1997 Fig. 4 Normalized current–time profiles for samples analyzed with and without the GDU: unspiked tap water (A) and unspiked ground water (B) samples (without the GDU) and spiked (0.40 ppm azide) tap water (C) and unspiked tap water (D) samples (with the GDU). Note the magnitude of the signal for the unspiked tap water sample (A) and the elimination of the interfering signal for the unspiked tap water sample (D) when the GDU is used in the SI system.Analyst, April 1997, Vol. 122 319 Determination of Primary Explosive Azides in Environmental Samples by Sequential Injection Amperometry Roger T. Echols*a, Ryan R. Jamesa and Joseph H. Aldstadtb a Division of Science and Mathematics, University of Minnesota Morris, Morris, MN 56267, USA b Environmental Research Division, Argonne National Laboratory, 9700 S.Cass Avenue ER/203, Argonne, IL 60439, USA The application of flow injection methodology to the determination of trace concentrations of primary explosives is presented. The approach is demonstrated with a sequential injection amperometric method for the determination of the azide ion (N32). The proposed method can be applied to the determination of sodium azide or lead azide, a primary explosive, without regard to other sources of lead in environmental samples.The sequential injection system used for the analysis forms the basis for a proposed field-portable instrument for the analysis of primary explosives. A microporous gas permeable membrane in a gas diffusion unit (GDU) is used to separate the analyte from other anions that can also be oxidized at the amperometric cell. The behaviour of the GDU was optimized with respect to the pH of the donor stream and the timing of the preconcentration step.A study of anions that are commonly found in environmental samples showed that the species that will interfere with the analytical signal can be removed by the GDU. Results from three water samples that were spiked with 0.40 ppm of azide are presented. RSDs in the range 3–5% were typically obtained using the method. The useful working range of the method was linear up to 0.5 ppm and non-linear up to 20 ppm (second-order model). The limit of detection was 24.6 ppb. Keywords: Flow injection; sequential injection; explosives; lead azide; sodium azide; amperometry; gas diffusion The contamination that has resulted from the production, storage, testing and disposal of explosives and the concomitant health risks that affect military and civilian personnel is a growing environmental problem.Most of the contamination is a consequence of the presence of secondary explosives, which are the principle species used in most shells and munitions. Examples of secondary explosives include well known nitroaromatics, nitric esters and nitramines: 2,4,6-tritinitrotoluene (TNT), 2,6- or 2,4-dinitrotoluene (DNT), nitroglycerin, nitrocellulose, 2,4,6-N-tetranitro-N-methylaniline (tetryl), 1,3,5-trinitro-1,3,5-triazacyclohexane (RDX) and 1,3,5,7-tetranitro- 1,3,5,7-tetrazacyclooctane (HMX).Contaminated water that contains TNT and DNT is commonly termed ‘red’ or ‘pink water’. Another category of explosives that are of environmental interest are primary explosives, which are readily ignited species used to detonate secondary explosives. Various inorganic salts comprise the list of primary explosives and include lead styphnate, mercury fulminate and metal azides.Lead azide is a common primary explosive that has been used extensively in blasting caps and military ammunition as a consequence of its high sensitivity to shock or ignition.1 Although the majority of the research in the area of determinations of trace concentrations of explosives has focused on secondary explosives, there have been a few reports of methods for the determination of azide (from lead or sodium azide). Instrumental methods of analysis (spectrophotometric, chromatographic and electrochemical methods) have supplanted cerium2 and other oxidation–reduction titrimetric methods.Spectrophotometric methods for the determination of the concentration of azide ion are based on the ability of azide (N32) to form charge-transfer complexes.These methods include those based upon coloured metal–azide complexes that form between azide and iron (or an iron complex)3–6 or between azide and copper.7 Other spectrophotometric methods include a method based upon a simple derivitization reaction of azide with carbon disulfide8 and a method based upon the formation of a complex between azide and Arsenazo III.9 Spectrophotometric detection has also been used in chromatographic methods. Swarin and Waldo10 proposed a chromatographic method with post-column derivitization for the determination of azide and reported a detection limit of 10 ppb with 3% reproducibility.Of the few papers on the determination of lead azide by electroanalytical methods, only one can be considered a recent application of modern electrochemistry. Previous electrochemical methods include an amperometric titration11 and a polarographic determination.12 The recent work on the determination of azide by an electrochemical method involved potentiometric flow injection:13 an azide ion-selective electrode consisting of an iron(ii) or nickel(ii) bathophenanthroline–azide complex embedded in PVC was constructed and tested on a variety of solutions.The electrode was used to determine azide with a limit of detection of 0.8 ppm and to determine the solubility product constants of insoluble metal azides. A table in this paper13 summarized the limit of detection, relative error and interfering species for most of the methods for the determination of azide.This research was undertaken to develop a method for the determination of the azide ion and to design a sequential injection system that will form the basis for a small, fieldportable or process laboratory instrument for the determination of this analyte in waters and soil extracts. The method is applicable to sodium azide, which has been used as a pesticide and herbicide, and metal azide primary explosives. The analytical method is based upon amperometric detection of the azide ion at a potential at which the azide ion is oxidized14 under basic conditions.N32?3–2 N2 + e2 The on-line isolation of the azide analyte from other electroactive ions that will be present in environmental samples is an important feature of the method and is made possible by use of a gas permeable membrane and flow injection (FI) methodology. In particular, sequential injection (SI) techniques are used in the system; SI15,16 is a subset of FI methods that was introduced in the early 1990s and has been applied recently to solve a number of analytical problems.Gas diffusion methods Analyst, April 1997, Vol. 122 (315–319) 315have been used extensively in FI systems as a means by which analytes can be isolated from difficult matrices or preconcentrated. 17,18 The principle involves passing the species of interest (as a gas) across a membrane from a donor stream into an acceptor stream within the FI system.The ultimate goal of this work is to design a field or process monitor that can be used on-site for measuring the concentrations of primary explosives. A monitor the size of a small suitcase that can be interfaced to a portable computer and power supply is envisioned. It is hoped that the method and instrument will aid military and civilian officials in the disposal of explosives and in risk assessment of contaminated areas by providing the advantages of on-site determinations.Moreover, the ability of this method to distinguish between a particular lead azide and other compounds containing lead (e.g., tetraethyllead from gasoline spills and lead azide from explosives) is a useful feature for issues of hazardous waste disposal and accountability; methods of analysis that only determine the concentration of lead would be inadequate for assessing the contamination due to explosives. Experimental Apparatus An SI system was used for all of the work with the exception of the preliminary work, which was performed with a single-line FI system. The SI system is shown in Fig. 1 and was composed of the following: standard flow tubing (0.76 mm id) and connectors (Global FIA, Gig Harbor, WA, USA) syringe pump (Cavro, Sunnyvale, CA, USA) multiport valve (Valco, Houston, TX, USA), peristaltic pump (Alitea, Medina, WA, USA), gas diffusion unit (GDU) with a PTFE membrane (Global FIA) and UniJet amperometric cell (3 mm glassy carbon disc electrode) with Petit Ampere potentiostat (BAS, W.Lafayette, IN, USA). The SI system was controlled using FlowTEK software distributed by Global FIA and required the use of a computer (80286, DOS) and an A/D converter. The amperometric cell was operated at a potential of 1.0 V (versus Ag/AgCl) and both the peristaltic pump and syringe pump were operated at a volumetric flow rate of 1.0 ml min21. The GDU has a channel that is 16 cm in length, 1 mm in width and 0.8 mm in depth; the PTFE membrane is standard plumber’s tape.Reagents All chemicals were of analytical-reagent grade; ultra-pure water was provided from a high purity (18 MW) filtration system. Azide solutions were prepared from a sodium azide stock solution. Phosphate buffer solutions that were 0.01 m in KCl were prepared from monobasic and dibasic phosphate salts. Buffer solutions with a range of pH from 1.1 to 5.2 were used as the carrier stream in the SI system. The anions that were used in the study of interfering species were prepared from common salts.Natural water samples (tap, ground and Lake Michigan) were spiked with sodium azide. Procedures Most experiments were conducted using the SI procedure detailed in Table 1. The sample is loaded into the holding coil (see Fig. 1) and then propelled towards the GDU with an appropriate change in position on the multiport valve. The sample is then manipulated on the donor side of the GDU using forward and reverse flows of the carrier stream.The acceptor stream is stopped and then resumed to send the analyte to the amperometric cell; the procedure results in the preconcentration of the analyte across the membrane. Peak height and peak area data were obtained using the FlowTEK software. Optimization of the passage of azide through the GDU was conducted by variation of the two factors that were the most important for the transfer of the analyte across the membrane: the pH of the donor stream and the number of flow reversals used in the method.Peak height was used as the figure of merit. The response of 1 ppm solutions of common anions at 1.0 V was obtained by removing the GDU from the system and operating the SI system in a manner similar to a single-line FI manifold. The response to 5 ppm anion solutions with the complete SI system was measured using the standard procedure (Table 1). Natural water samples were spiked with 0.4 ppm of azide and analyzed using the standard procedure.Preliminary work with the SI amperometric system was conducted with bleach in phosphate buffer;19 this system is a Fig. 1 Schematic diagram of the sequential injection system. Flow tubing (0.76 mm id) is shown as solid lines and electrical connections as dashed lines. The gas diffusion unit has a channel that is 16 cm in length, 1 mm in width and 0.8 mm in depth. Table 1 Timed events of the sequential injection system Multiport Syringe Peristaltic Time/s valve pump pump Comment 0.00 Reset Stop Off Prepares MPV for next command 0.50 Step Turn MPV to port 2— connected to sample 0.75 Reset 1.75 Reverse Load sample into HC 2.50 Forward Begin flow of acceptor stream 21.75 Stop 22.75 Step Turn MPV to port 3— connected to GDU 23.00 Reset 24.00 Forward Propel sample to GDU (donor stream) 44.00 Off Stop flow of acceptor stream 50.00 Stop Flow agitations: 52.00 Reverse Reverse flow 62.00 Stop 64.00 Forward Forward flow 71.00 Stop 73.00 Reverse Reverse flow 81.00 Stop 83.00 Forward Forward flow 100.00 Forward Begin flow of acceptor stream 197.00 Stop 307.00 Off Stop flow of acceptor stream 308.00 Home Returns MPV to port 1 316 Analyst, April 1997, Vol. 122good ‘test’ reaction for setting-up and troubleshooting the SI system without generating hazardous waste. Preliminary work with azide was conducted with a simple single-line FI system. It was demonstrated that azide could be detected (in a single-line FI system) over a linear range from 0.10 to 2.5 ppm.The initial reaction conditions (phosphate buffer of pH 6.6 and potential) were based on work by Ward and Wright:14 a peak potential of 1.08 V was reported for the oxidation of azide at platinum electrodes in phosphate buffer solutions for a range of pH from 6.0 to 10.5. A potential of 1.0 V was used because of equipment limitations: the amperometric system did not have a current range large enough to accommodate background signals of the carrier solution at a peak potential of 1.08 V.Hence the analytical signals would be slightly larger with a different potentiostat. Results and Discussion Optimization of the Sequential Injection System The principal component of the SI system used for the determination of azide is the GDU that is used to isolate the azide from other anions. Anions that are present in environmental samples (e.g., chloride) will oxidize under these electrochemical conditions and interfere with the analytical method.The procedure for transporting azide across the microporous PTFE membrane in the GDU prevents the passage of most anions because they do not form neutral molecules at the pH used in this experiment. When the azide ion is protonated (pKa = 4.72), the HN3 passes through the membrane in the gaseous state. The relatively high volatility of HN3 (boilingpoint 37 °C) allows the separation to occur. Because the transport of HN3 across the gas permeable membrane is not 100% efficient, the pH of the donor stream and the timing of the SI system were chosen as the primary variables of the SI system for optimization. Sample volume was not varied as a consequence of choosing the time for an individual experiment as a design criterion for the work; the sample load time of 20.0 s (approximately 333 ml) was chosen on this basis to provide an overall experiment time of less than 5 min.Other parameters of the SI system (e.g., flow rate and tubing lengths) were not optimized with respect to the performance of the GDU because the effect of these variables is secondary to that of the donor stream pH and timing.In addition to microporous PTFE, a silicone-rubber membrane was used in the GDU. The results indicated that the transport of HN3 across the silicone-rubber membrane was significantly less than that with the PTFE membrane. The pH of the donor stream was optimized by varying the pH and by measuring the current at peak height for a 1 ppm solution.Results from this experiment are shown in Fig. 2. The mechanism for transport of azide across the gas permeable membrane involves the deprotonation of this species in the more basic acceptor stream (pH 6.6) and the formation of HN3(g) in the donor stream; it was expected that a donor stream pH less than the pKa of hydrazoic acid (4.72) would provide the most efficient transport across the membrane. The constant response for experiments for which the donor stream pH was less than the pKa confirmed the expectation and a pH of 3.78 was chosen for subsequent work.Moreover, calculation of the fraction of dissociation for hydrazoic acid confirmed that only 10% of azide is in the basic form at pH 3.78. The center of the sample (greatest concentration of analyte) was determined to be in the GDU at 40 s. The acceptor stream was halted for a time before and after this time in order to allow the diffusion of HN3 across the membrane and into the pH 6.6 acceptor stream.The method was optimized by performing a number of flow reversals using the syringe pump (0, 2, 4 and 8 flow reversals). The procedure has been termed ‘agitating the sample’ because of the similarity to what is done when solvent extractions are conducted in separating funnels. The results from an experiment investigating the technique are illustrated in Fig. 3. The greatest amount of azide was transferred across the membrane when four ‘agitations’ (four steps of reverse flow and forward flow) were performed within the established time interval that the acceptor stream was stopped.An approximately 25% signal enhancement was realized using this protocol (see Fig. 3). The ability to reverse the flow in the experiment is one of the advantages that SI methods have over traditional FI methods in which the flow is not typically reversed. Study of the Interference of Common Anions Anions that contribute significantly to the amperometric signal for azide were identified by Ward and Wright14 in their study of the electrochemistry of azide.Their list formed the basis for the species that were chosen for the study of interfering anions. Results from the study are shown in Table 2. The response of the anions in the study without the GDU in the SI system (first two columns in Table 2) can be grouped into three categories: (1) no response (perchlorate, bromide and nitrate); (2) low to moderate response (sulfite, phosphate, hypochlorite and cyanide); and (3) strong response (iodide, chloride, thiocyanate, nitrite and sulfide).The signals decrease significantly when the GDU is incorporated in the SI system for the step that separates the analyte from the matrix (second two columns in Table 2). The signals for sulfide, nitrite and hypochlorite are high enough at a concentration of 5 ppm to interfere with the azide signal.Other Fig. 2 Optimization of pH of the donor stream of the gas diffusion unit (1.0 ppm azide solutions). Error bars represent plus or minus one standard deviation (n = 3). Fig. 3 Optimization of number of flow reversals (‘agitations’) in the experimental method (1.0 ppm azide solutions). Error bars represent plus or minus one standard deviation (n = 3). Analyst, April 1997, Vol. 122 317anions (with the exception of iodide, which has a negative response) have corrected signals that are scattered around the blank signal and lie within the experimental uncertainty (95% confidence interval) of the SI experiment for the determination of azide (±0.019).The passage of hydrogen sulfide (from sulfide), nitrous acid (from nitrite) and hypochlorous acid (from hypochlorite) through the membrane was expected on the basis of the chemistry of those species: H2S has a pKa1 of 7.02 and a pKa2 of 13.9; HNO2 has a pKa of 3.15 and HOCl has a pKa of 7.53.20 The three anions are subjected to the same chemistry as the azide ion: conjugate acids are formed in the donor stream before passing across the membrane and before being oxidized at the working electrode.Nitrite is dissociated to a greater extent than the other two species, which might explain the larger signal for nitrite than sulfide and hypochlorite. On the basis of previous work,14 it was expected that hydrogen cyanide (from cyanide) would interfere to a greater extent than that observed in the study of interference. At the pH used in the experiment, the cyanide should cross the membrane as HCN (pKa = 9.21) and remain in the neutral form; hence the oxidation of cyanide ion would not be expected.It was difficult to determine how much of the azide in each sample was passing through the PTFE membrane (efficiency of transport) as a consequence of the agitations (reversals of flow) that were an integral part of the procedure. A comparison could be made on the basis of the signals from experiments including and excluding the GDU from the SI system.The donor stream was connected to the acceptor stream such that the GDU was by-passed and a 1.0 ppm solution of azide was run through the procedure (with agitations). The signal from this experiment was 0.98 nA, which was 40% larger than that obtained using the GDU and the normal method (signal of 0.58 nA). The simple comparison does not indicate that transport of the azide across the membrane is 60% efficient in an absolute sense because the azide is preconcentrated during the steps of forward and reverse flow while the acceptor stream is halted. Determination of Azide in Water Samples Results from the spiked samples of Lake Michigan, ground and tap waters are summarized in Table 3.The calculation of the concentrations was based on a regression line obtained using standard azide solutions (0–0.5 ppm): y = 0.0131 (nA) +1.21 (nA ppm21) x with r2 = 0.990. The blank signal of 0.0219 nA was subtracted from each standard and sample; the response for the blank is caused by the change in background current that results from the stopping of the donor stream during the preconcentration/separation step using the GDU.The limit of detection was determined to be 24.6 ppb of azide on the basis of three times the standard deviation of the blank signals (0.0429 nA). While the response of the standards was linear over this range (0–0.5 ppm), the response for a broader range of standards (0–20 ppm) was non-linear and was fitted by a second-order polynomial model: y = 0.0798 + 0.597x 2 0.138x2 (r2 = 0.988); the non-linearity of the upper range of the standards was thought to be a feature of the gas diffusion unit.The precision for repeated experiments was influenced by the experimental procedure in which the donor stream was halted and the baseline current allowed to decay before flow was resumed. Nevertheless, RSDs of 3–5% were typical for calibration standard solutions (clean samples).For the spiked water samples that contained interfering anions, the RSDs were 1.7, 3.0 and 7.7%. The accuracy of the results was assessed by examination of the percentage difference from the expected value of 0.40 ppm azide (amount spiked) and the population means (Table 3). Of the three samples, the expected value lies outside the population mean at 95% confidence for the ground water sample.The three water samples (not spiked with azide) were introduced directly into the amperometric cell to verify that there were other anions present and that the GDU was separating azide from the matrix. Normalized current–time profiles (‘peaks’) for tap water and ground water samples that were introduced directly into the amperometric cell (GDU removed from the SI system) are illustrated in Fig. 4 (A and B). The interfering signal from background anions that would result from the determination being performed without the separation of the analyte using the GDU is clearly illustrated.Normalized peaks for the spiked tap water sample (C) and the unspiked tap water sample (D), which were obtained with the GDU in the SI system, are also shown in Fig. 4. The height of the peak of the background anion signal when the GDU is not in use (A) is twice that of the spiked tap water sample when the GDU is used (C). Additionally, the elimination of the interfering signal can be seen by comparing peaks A and D in Fig. 4: the GDU completely eliminates the background anion signal. Future Work Future work will include an assessment of the field performance of the proposed SI instrument, which will necessitate the inclusion of an extraction procedure for lead azide. A study of the discrimination of lead species (e.g., determination of lead azide in the presence of other sources of lead) will also be included in future work. The principles applied in this work for the determination of the target analyte (azide) can be employed for the determination of other explosives.Other analytical methods that incorporate FI methodologies can be designed for the different chemistries needed to determine the concentrations of other explosives that are of environmental interest (e.g., TNT Table 2 Results from the study of the interference of common anions Without gas diffusion unit With gas diffusion unit Signal/nA Percentage Signal/nA Percentage (1 ppm of azide (5 ppm of azide Anion solutions) signal solutions)* signal Azide 2.731 — 1.365 — Bromide 0.004 0.1 20.015 21.1 Chloride 1.173 42.9 20.003 20.2 Cyanide 0.063 2.3 0.015 1.1 Hypochlorite 0.225 8.2 0.068 5.0 Iodide 3.429 122.1 20.021 21.5 Nitrate 0.006 0.2 0.009 0.7 Nitrite 2.030 74.3 0.211 15.5 Perchlorate 0.009 0.3 20.007 20.5 Phosphate 0.029 1.0 20.011 20.8 Sulfide 3.110 113.9 0.132 9.7 Sulfite 0.254 9.0 20.003 20.2 Thiocyanate 0.753 27.6 0.006 0.4 * Signal is corrected for response of blank (0.053 nA).Table 3 Results from water samples spiked with 0.40 ppm of sodium azide Difference Concentration, from 0.400 Sample ppm* RSD (%) ppm (%) Lake Michigan water 0.385 ± 0.047 7.7 23.8 Tap water 0.409 ± 0.011 1.7 2.3 Ground water 0.429 ± 0.020 3.0 7.3 * 95% confidence limits of the mean; n = 4. 318 Analyst, April 1997, Vol. 122and DNT). An interesting possibility for the determination of TNT and DNT is the formation of coloured sulfite complexes that can be detected with visible spectrophotometry.21 The Division of Educational Programs of Argonne National Laboratory is gratefully acknowledged for supporting R.T.E. through the Faculty Research Program and R.R.J.through the Science and Engineering Research Semester Program. Argonne National Laboratory is operated by the University of Chicago under Contract No. W-31-109-ENG-38. This research was presented in part at the 1996 Pittsburgh Conference and at the 1996 FACSS Conference. References 1 Yinon, J., and Zitrin, S., Modern Methods and Applications in Analysis of Explosives, Wiley, New York, 1993. 2 Grove, E. L., Braman, R. S., Combs, H. F., and Nicholson, S. B., Anal. Chem., 1962, 34, 682. 3 Roberson, C. E., and Austin, C. M., Anal. Chem., 1957, 29, 855. 4 Anton, A., Dodd, J. G., and Harvey, A. E., Anal. Chem., 1960, 32, 1209. 5 Ilcheva, L., and Todorova, G., Acta Chim. Acad. Sci. Hung., 1979 (102) 113. 6 Mehra, M. C., and Garvie, R., Microchem. J., 1980, 25, 223. 7 Neves, E. A., de Oliveira, E., and Sant’agostino, L., Anal. Chim. Acta, 1976, 87, 243. 8 Franco, D. W., Neves, E. A., and Andrade, J. F., Anal. Lett., 1977, 10, 243. 9 Kubaszewski, E., and Kurzawa, Z., Chem. Anal. (Warsaw), 1985, 30, 609. 10 Swarin, S. J., and Waldo, R. A., J. Liq. Chromatogr., 1982, 5, 597. 11 Dziegiec, J., and Ignaczak, M., Acta Chim. Soc. Sci. Lodz, 1971, 16, 69. 12 Bryant, J. I., and Kemp, M. D., Anal. Chem., 1960, 32, 759. 13 Hassan, S. S., El Zawawy, F. M., Marzouk, S. A. M., and Elnemma, E. E., Analyst, 1992, 117, 1683. 14 Ward, G. A., and Wright, C. M., J. Electroanal. Chem., 1964, 8, 302. 15 Ruzicka, J., and Marshall, G. D., Anal. Chim. Acta, 1990, 237, 329. 16 Gubeli, T., Christian, G. D., and Ruzicka, J., Anal. Chem., 1991, 69, 2407. 17 Kuban, V., Crit. Rev. Anal. Chem., 1992, 23, 323, and references cited therein. 18 Fang, Z., Flow Injection Separation and Pre-concentration, VCH, New York, 1993. 19 Tsaouis, A. N., and Huber, C. O., Anal. Chim. Acta, 1985, 178, 319. 20 Martell, A. E., and Smith, R. M., Critical Stability Constants, Plenum Press, New York, 1974. 21 Jenkins, T. F., US Army Corps of Engineers Special Report 90-38, Cold Regions Research and Engineering Laboratory, Hanover, NH, USA, 1990. Paper 6/05662H Received August 13, 1996 Accepted January 7, 1997 Fig. 4 Normalized current–time profiles for samples analyzed with and without the GDU: unspiked tap water (A) and unspiked ground water (B) samples (without the GDU) and spiked (0.40 ppm azide) tap water (C) and unspiked tap water (D) samples (with the GDU). Note the magnitude of the signal for the unspiked tap water sample (A) and the elimination of the interfering signal for the unspiked tap water sample (D) when the GDU is used in the SI system. Analyst, April 1997, Vol. 122 319
ISSN:0003-2654
DOI:10.1039/a605662h
出版商:RSC
年代:1997
数据来源: RSC
|
6. |
Application of Tryptamine as a Derivatizing Agent for theDetermination of Airborne Isocyanates.Part 7. Selectionof Impinger Solvents and the Evaluation Against Dimethyl Sulfoxide Used inUS NIOSH Regulatory Method 5522 |
|
Analyst,
Volume 122,
Issue 4,
1997,
Page 321-323
Weh S. Wu,
Preview
|
|
摘要:
Application of Tryptamine as a Derivatizing Agent for the Determination of Airborne Isocyanates. Part 7.† Selection of Impinger Solvents and the Evaluation Against Dimethyl Sulfoxide Used in US NIOSH Regulatory Method 5522 Weh S. Wu*, Roman S. Szklar and Roy Smith Occupational Health Laboratory, Ontario Ministry of Labour, 101 Resources Road, Weston, Ontario, Canada M9P 3T1 The application of tryptamine to derivatize airborne isocyanates has been evaluated and adopted by the National Institute for Occupational Safety and Health (NIOSH) as the latest isocyanates regulatory method (Method 5522) in the USA.Method 5522 uses dimethyl sulfoxide (DMSO) as the impinger solvent in which tryptamine is dissolved to sample isocyanates for analysis. Since DMSO is both extremely hygroscopic and corrosive, it is not a satisfactory solvent for impinger air sampling of isocyanates. The high boiling and freezing points also present some distinct drawbacks. In a search for a suitable impinger solvent, the efficiencies of various solvents, which all were potentially more suitable for sampling isocyanates viz., N,NA-dimethylformamide, butyl acetate, isobutyl acetate, sec-butyl acetate, tert-butyl acetate and octane, were investigated.Simulated air sampling was conducted on two commonly used isocyanates in industry, hexamethylene diisocyanate (HDI) and toluene diisocyanate (TDI), which were vaporized and sampled into impingers containing dissolved tryptamine.Butyl acetate and octane were found to be most suitable for using as impinger solvents. Recoveries of isocyanates at two concentration levels of HDI and TDI were 93.4–108% in comparison with those of using DMSO. Keywords: Air sampling; isocyanates; selection of impinger solvent; NIOSH Method 5522; tryptamine The potential use of tryptamine (TRYP) in the impinger solution to sample airborne isocyanates for regulatory monitoring in the workplace has proven to be more advantageous than other types of reagents.Various discussions regarding this topic can be found in previous publications of this series.1–4 The US National Institute for Occupational Safety and Health (NIOSH) has recently completed an evaluation and has proposed the adoption of the TRYP method as their latest methodology for monitoring airborne isocyanates (Method 5522).5 The impinger sampling medium used in Method 5522 is TRYP dissolved in dimethyl sulfoxide (DMSO). DMSO is not reactive to isocyanates and is one of the few solvents in which TRYP and its derivatized isocyanates are readily dissolved.Also, DMSO is not very volatile, which removes the requirement to replenish the impinger solvent during the sampling period. However, there are serious drawbacks associated with using DMSO in both sampling and analyzing airborne isocyanates. Since DMSO is extremely hygroscopic and the absorbed moisture is reactive to isocyanates, an excessive amount of TRYP needs to be used in the impinger solution for derivatizing isocyanates to compete with the potentially high content of moisture.Thus, Method 5522 employs 9 mg of TRYP in each impinger for sampling whereas the previous method, Method 5521,6,7 used only 0.65 mg of 1-(2-methoxyphenyl)piperazine (MPP) for the same purpose. In previous publications from this laboratory reflecting the initial development of the TRYP method for air sampling, a range of only 0.1–0.4 mg of TRYP in 15 cm23 of isooctane was used in each impinger.3 Taking into consideration the molecular masses of TRYP and MPP of 160 and 192, respectively, the amount of TRYP used in DMSO for sampling isocyanates is excessive. Since TRYP is eluted earlier than TRYP-derivatized isocyanates in HPLC, a large excess of TRYP would inevitably interfere with the elution of isocyanate derivatives having shorter retention times.This results in a noticeably higher limit of detection (LOD) for hexamethylene 1,6-diisocyanate (HDI), since HDI has the shortest retention time among the diisocyanates commonly used in industry.To minimize the overlap of derivative peaks from the huge peak of excess TRYP during elution, the polarity of the eluting solvent must be kept low for maximum separation. For this reason, Method 5522 uses an isocratic 40 + 60 (v/v) acetonitrile–0.6% aqueous sodium acetate mixture for the early-eluting HDI derivative whereas 60 + 40 (v/v) acetonitrile–0.6% aqueous sodium acetete is used for the later eluting derivatives.A further inconvenience of using DMSO is that it freezes at 18 °C so that acetonitrile must be added to DMSO at a level of 20% ratio when the ambient temperature is below 15.6 °C (60 °F) during sampling. DMSO is also known to be extremely corrosive to many organic materials. It has been found in our work to corrode the air sampling pump, in particular the filter housing of the pump, in a relatively short period of time. The corrosion of the air inlet of the pump and the filter housing is easily spotted visually without even having to dismantle the pump unit.Based on the above shortcomings of DMSO, we felt it was necessary to revisit the topic of the impinger solvent for the TRYP method with the aim of finding a more satisfactory solvent to overcome the noted problems. It should be mentioned that the solvent isooctane, used in our original TRYP method, is also not very satisfactory owing to its high volatility.Experimental Chemicals and Apparatus Tryptamine was purchased from Sigma (St. Louis, MO, USA) with a purity of 99+% and was recrystallized from acetonitrile before use. HDI and toluene 2,4-diisocyanate (TDI) were obtained from Aldrich (Milwaukee, WI, USA) and used as received. Methylene diisocyanate (MDI), which was found to be present in Mondur MR (commercial name of a polymeric † For Part 6 of this series, see reference 4. Analyst, April 1997, Vol. 122 (321–323) 321isocyanate product) was supplied by Mobay Chemical (Pittsburgh, PA, USA). DMSO of spectrophotometric grade (99.9% purity) was obtained from Aldrich. Butyl acetate, octane, N,NAdimethylformamide (DMF) and dichloromethane were all of glass-distilled quality and supplied by Caledon Laboratory (Georgetown, Ontario, Canada). The instrument used for HPLC analysis was described in previous parts of this series. The air sampling pump used in this work was a Model 224-PCXR3 from SKC (Eighty Four, PA, USA). Preparation of Impinger Solutions The impinger solution of TRYP in DMSO was prepared at a concentration of 450 mg cm23.The concentration of TRYP in other solvents was 25 mg cm23. a 20 cm3 impinger solution was placed in the impinger in all cases for sampling. Simulated Air Sampling for Isocyanate Recoveries The sampling strategy was carried out according to the procedure suggested by NIOSH. A sampling tube to deposit the desired amount of isocyanate solution for vaporization was prepared by cutting off part of the long stem of the drying tube (with a bulb at the end of the constricted stem type).The constricted stem was bent, under a strong flame, to a shape of about 90° elbow. The stem was then connected to the air inlet of the impinger so that the shortened drying tube was in a horizontal position. All connections were made by using solvent-resistant Tygon tubing. Approximate concentrations of 0.4 and 3 mg cm23 of HDI and TDI, to reflect two levels of air sampling, were prepared in dichloromethane for this recovery study.In each case, 10 mm3 of solution were deposited in the bulb with a syringe and the sampling pump was then immediately turned on. The sampling pump was operated at a flow rate of 1 dm3 min21 and the sampling time was 3 h. Experiments on the relative recoveries of the above isocyanates as TRYP derivatives in various solvents were conducted against DMSO.Details of the excellent recoveries using DMSO were cited in NIOSH Method 5522 for three concentration levels of each isocyanate as 91, 103 and 90% for 2,4-TDI, 2,6-TDI and HDI, respectively. HPLC Analysis of Simulated Air Samples Since DMSO and DMF are difficult to remove by evaporation owing to their low volatility, impinger solutions of these solvents were used directly for analysis. A 50 mm3 volume of acetic anhydride was added to the respective solution and set aside for 10 min.A 50 mm3 volume of water was then added and again set aside for 30 min before injection into the HPLC system for quantification. Analytical standards were prepared in the respective solvent for accurate calibration. Impinger solvents of higher volatility, such as butyl acetate or octane, were transferred, with subsequent rinsing with acetonitrile, into a 100 cm3 distillation flask. It should be noted that despite the fact that octane and acetonitrile are only slightly miscible, rinsings with Table 1 Simulated air sampling recoveries for HDI vapour in the impinger solvents butyl acetate, octane and DMSO Sample Impinger Amount of Impinger Amount of set solvent HDI found/mg solvent HDI found/mg 1 Butyl 2.9 DMSO 2.9 acetate 2.4 2.7 2.9 2.9 3.2 3.1 3.4 2.6 Av. 3.0 ± 0.4 Av. 2.9 ± 0.2 (93.8 ± 12%) (90.6 ± 6.3%) 2 Butyl 25.4 DMSO 23.3 acetate 23.7 22.1 22.0 24.9 25.6 23.8 24.9 22.4 Av. 24.3 ± 1.5 Av. 23.3 ± 1.2 (97.6 ± 6.0%) (93.6 ± 4.8%) 3 Octane 4.7 DMSO 4.4 4.3 4.3 4.2 4.5 4.6 3.8 4.0 4.4 Av. 4.4 ± 0.3 Av. 4.3 ± 0.3 (95.7 ± 6.5%) (93.5 ± 6.5%) 4 Octane 36.7 DMSO 31.5 37.6 28.1 35.5 33.6 30.6 34.9 31.5 32.3 Av. 34.5 ± 3.1 Av. 32.0 ± 2.6 (106 ± 9.5%) (98.2 ± 8.0%) Table 2 Simulated air sampling recoveries for TDI vapour in the impinger solvents butyl acetate, octane and DMSO Sample Impinger Amount of Impinger Amount of set solvent TDI found/mg solvent TDI found/mg 1 Butyl 3.0 DMSO 2.9 acetate 2.5 2.8 2.9 2.5 2.8 2.4 2.7 2.9 Av. 2.8 ± 0.2 Av. 2.7 ± 0.2 (96.6 ± 6.9%) (93.1 ± 6.9%) 2 Butyl 31.4 DMSO 24.2 acetate 30.2 30.4 28.2 26.8 27.0 26.8 24.6 26.4 Av. 28.3 ± 2.7 Av. 27.0 ± 2.2 (93.7 ± 8.9%) (89.4 ± 7.3%) 3 Octane 25.6 DMSO 28.8 28.7 27.1 23.9 29.1 27.6 26.2 27.7 31.5 Av. 26.7 ± 1.9 Av. 28.6 ± 2.1 (90.5 ± 6.4%) (96.9 ± 7.1%) Table 3 Relative recoveries for HDI and TDI vapor using butyl acetate and octane in comparison with that for DMSO Relative recovery Set Impinger Amount of isocyanate compared with No.* solvent found/mg using DMSO (%) 1 Butyl acetate 3.0 ± 0.4 (HDI) 103 ± 16 2 Butyl acetate 24.3 ± 1.5 (HDI) 104 ± 8 3 Octane 4.4 ± 0.3 (HDI) 102 ± 10 4 Octane 34.5 ± 3.1 (HDI) 108 ± 18 5 Butyl acetate 2.8 ± 0.2 (TDI) 104 ± 11 6 Butyl acetate 28.3 ± 2.7 (TDI) 105 ± 13 7 Octane 26.7 ± 1.9 (TDI) 93.4 ± 9.5 * References to sets 1–7 for recoveries using DMSO are 2.9 ± 0.2, 23.3 ± 1.2, 4.3 ± 0.3, 32.0 ± 2.6, 2.7 ± 0.2, 27.0 ± 2.2 and 28.6 ± 2.1%. 322 Analyst, April 1997, Vol. 122acetonitrile proved to be effective.The solvent was evaporated under a rotary evaporator between 40 and 45 °C. Acetonitrile was used to rinse the flask and the rinsings were collected in a 10 cm3 volumetric flask. Acetic anhydride and water were added to the solution as described previously except that the volume was reduced to 10 mm3 before diluting to volume for HPLC analysis. Results and Discussion Simulated air sampling recoveries for vaporized HDI and TDI using TRYP were conducted in the impinger solvents butyl acetate, octane and DMSO.Tables 1 and 2 summarize the results. The mean RSD of all the recoveries is 7.8 ± 1.9%. Using an anticipated pump error of not more than 5% as assumed by NIOSH, the total sampling and analytical error, RSDt, is calculated to be 9% based on RSDt 2 = 7.82 + 52. The expected accuracy of recoveries based on a maximum bias of 5% pump error is 5 + 2 RSDt (i.e., smaller than 25%), which meets the NIOSH criterion specified in the NIOSH Standards Completion Program.6,8 Also, considering the standard deviations, the relative recoveries using butyl acetate and octane for sampling compared with that of DMSO as the impinger solvent are excellent, as shown in Table 3.The results confirm that these solvents are comparable in recoveries to DMSO employed in NIOSH Method 5522. The lower detection limit of HDI and TDI using either butyl acetate or octane was approximately 0.01 mg cm23. These detection limits are substantially lower than those obtained using DMSO.It was also found that the recoveries using DMF relative to those of DMSO for vaporized HDI at levels of 4 and 40 mg were 116 ± 12 and 111 ± 14%, respectively. Although butyl acetate and octane were replaced with acetonitrile for HPLC analysis, DMF was used throughout the procedure owing to its difficulty of removal. Unlike DMSO, the detection limit of HDI in DMF was unaffected since an excessive amount of TRYP did not need to be added to the sampling solution.Since DMF has been documented as a suspected carcinogen and butyl acetate or octane was found to be a highly effective solvent, the application of DMF for airborne isocyanates sampling was abandoned. Other solvents such as isobutyl acetate, sec-butyl acetate and tert-butyl acetate were also briefly tested as possible impinger solvents but were dropped when unsuspected analytical interferences were encountered. Since MDI was highly unevaporative, sampling recoveries were carried out in a similar fashion to that of NIOSH by spiking MDI solution into the impinger solutions.2,4,7 The relative recoveries of MDI using butyl acetate and octane compared with DMSO were 104 and 105%, respectively.Conclusions DMSO is a solvent with extremely high polarity, relatively high freezing point and low volatility. To replace DMSO used in NIOSH Method 5522, this study showed that the solvents butyl acetate and octane were more advantageous and practical for sampling and determining airborne isocyanates.Butyl acetate and octane also proved to be just as effective as DMSO with regard to isocyanates recoveries but provided a much lower detection limit for HDI. It was observed that corrosion of the air sampling pump by DMSO was too significant to be ignored. The long term adverse effect of DMSO on an HPLC system is also likely unfavourable. References 1 Wu, W. S., Nazar, M. A., Gaind, V.S., and Calovini, L., Analyst, 1987, 112, 863. 2 Wu, W. S., Szklar, R. S., and Gaind, V. S., Analyst, 1988, 113, 1209. 3 Wu, W. S., Stoyanoff, R. E., Szklar, R. S., Gaind, V. S., and Rakanovic, M., Analyst, 1990, 115, 801. 4 Wu, W. S., and Gaind, V. S., Analyst, 1994, 119, 1043. 5 NIOSH Manual of Analytical Methods, National Institute for Occupational Safety and Health, Cincinnati, OH, 4th edn., 1993, Method 5522, in draft. 6 Bagon, D. A., Warwick, C. J., and Brown, R. H., Am.Ind. Hyg. Assoc. J., 1984, 45, 39. 7 NIOSH Manual of Analytical Methods, National Institute for Occupational Safety and Health, Cincinnati, OH, 1989, Method 5521. 8 Documentation of the NIOSH Validation Tests, NIOSH Publication No. 77–185, National Institute for Occupational Safety and Health, Cincinnati, OH, 1977. Paper 6/06136B Received September 6, 1996 Accepted November 25, 1996 Analyst, April 1997, Vol. 122 323 Application of Tryptamine as a Derivatizing Agent for the Determination of Airborne Isocyanates.Part 7.† Selection of Impinger Solvents and the Evaluation Against Dimethyl Sulfoxide Used in US NIOSH Regulatory Method 5522 Weh S. Wu*, Roman S. Szklar and Roy Smith Occupational Health Laboratory, Ontario Ministry of Labour, 101 Resources Road, Weston, Ontario, Canada M9P 3T1 The application of tryptamine to derivatize airborne isocyanates has been evaluated and adopted by the National Institute for Occupational Safety and Health (NIOSH) as the latest isocyanates regulatory method (Method 5522) in the USA.Method 5522 uses dimethyl sulfoxide (DMSO) as the impinger solvent in which tryptamine is dissolved to sample isocyanates for analysis. Since DMSO is both extremely hygroscopic and corrosive, it is not a satisfactory solvent for impinger air sampling of isocyanates. The high boiling and freezing points also present some distinct drawbacks. In a search for a suitable impinger solvent, the efficiencies of various solvents, which all were potentially more suitable for sampling isocyanates viz., N,NA-dimethylformamide, butyl acetate, isobutyl acetate, sec-butyl acetate, tert-butyl acetate and octane, were investigated.Simulated air sampling was conducted on two commonly used isocyanates in industry, hexamethylene diisocyanate (HDI) and toluene diisocyanate (TDI), which were vaporized and sampled into impingers containing dissolved tryptamine. Butyl acetate and octane were found to be most suitable for using as impinger solvents.Recoveries of isocyanates at two concentration levels of HDI and TDI were 93.4–108% in comparison with those of using DMSO. Keywords: Air sampling; isocyanates; selection of impinger solvent; NIOSH Method 5522; tryptamine The potential use of tryptamine (TRYP) in the impinger solution to sample airborne isocyanates for regulatory monitoring in the workplace has proven to be more advantageous than other types of reagents.Various discussions regarding this topic can be found in previous publications of this series.1–4 The US National Institute for Occupational Safety and Health (NIOSH) has recently completed an evaluation and has proposed the adoption of the TRYP method as their latest methodology for monitoring airborne isocyanates (Method 5522).5 The impinger sampling medium used in Method 5522 is TRYP dissolved in dimethyl sulfoxide (DMSO). DMSO is not reactive to isocyanates and is one of the few solvents in which TRYP and its derivatized isocyanates are readily dissolved.Also, DMSO is not very volatile, which removes the requirement to replenish the impinger solvent during the sampling period. However, there are serious drawbacks associated with using DMSO in both sampling and analyzing airborne isocyanates. Since DMSO is extremely hygroscopic and the absorbed moisture is reactive to isocyanates, an excessive amount of TRYP needs to be used in the impinger solution for derivatizing isocyanates to compete with the potentially high content of moisture.Thus, Method 5522 employs 9 mg of TRYP in each impinger for sampling whereas the previous method, Method 5521,6,7 used only 0.65 mg of 1-(2-methoxyphenyl)piperazine (MPP) for the same purpose. In previous publications from this laboratory reflecting the initial development of the TRYP method for air sampling, a range of only 0.1–0.4 mg of TRYP in 15 cm23 of isooctane was used in each impinger.3 Taking into consideration the molecular masses of TRYP and MPP of 160 and 192, respectively, the amount of TRYP used in DMSO for sampling isocyanates is excessive. Since TRYP is eluted earlier than TRYP-derivatized isocyanates in HPLC, a large excess of TRYP would inevitably interfere with the elution of isocyanate derivatives having shorter retention times.This results in a noticeably higher limit of detection (LOD) for hexamethylene 1,6-diisocyanate (HDI), since HDI has the shortest retention time among the diisocyanates commonly used in industry. To minimize the overlap of derivative peaks from the huge peak of excess TRYP during elution, the polarity of the eluting solvent must be kept low for maximum separation.For this reason, Method 5522 uses an isocratic 40 + 60 (v/v) acetonitrile–0.6% aqueous sodium acetate mixture for the early-eluting HDI derivative whereas 60 + 40 (v/v) acetonitrile–0.6% aqueous sodium acetete is used for the later eluting derivatives.A further inconvenience of using DMSO is that it freezes at 18 °C so that acetonitrile must be added to DMSO at a level of 20% ratio when the ambient temperature is below 15.6 °C (60 °F) during sampling. DMSO is also known to be extremely corrosive to many organic materials. It has been found in our work to corrode the air sampling pump, in particular the filter housing of the pump, in a relatively short period of time. The corrosion of the air inlet of the pump and the filter housing is easily spotted visually without even having to dismantle the pump unit.Based on the above shortcomings of DMSO, we felt it was necessary to revisit the topic of the impinger solvent for the TRYP method with the aim of finding a more satisfactory solvent to overcome the noted problems. It should be mentioned that the solvent isooctane, used in our original TRYP method, is also not very satisfactory owing to its high volatility.Experimental Chemicals and Apparatus Tryptamine was purchased from Sigma (St. Louis, MO, USA) with a purity of 99+% and was recrystallized from acetonitrile before use. HDI and toluene 2,4-diisocyanate (TDI) were obtained from Aldrich (Milwaukee, WI, USA) and used as received. Methylene diisocyanate (MDI), which was found to be present in Mondur MR (commercial name of a polymeric † For Part 6 of this series, see reference 4. Analyst, April 1997, Vol. 122 (321–323) 321isocyanate product) was supplied by Mobay Chemical (Pittsburgh, PA, USA).DMSO of spectrophotometric grade (99.9% purity) was obtained from Aldrich. Butyl acetate, octane, N,NAdimethylformamide (DMF) and dichloromethane were all of glass-distilled quality and supplied by Caledon Laboratory (Georgetown, Ontario, Canada). The instrument used for HPLC analysis was described in previous parts of this series. The air sampling pump used in this work was a Model 224-PCXR3 from SKC (Eighty Four, PA, USA).Preparation of Impinger Solutions The impinger solution of TRYP in DMSO was prepared at a concentration of 450 mg cm23. The concentration of TRYP in other solvents was 25 mg cm23. a 20 cm3 impinger solution was placed in the impinger in all cases for sampling. Simulated Air Sampling for Isocyanate Recoveries The sampling strategy was carried out according to the procedure suggested by NIOSH. A sampling tube to deposit the desired amount of isocyanate solution for vaporization was prepared by cutting off part of the long stem of the drying tube (with a bulb at the end of the constricted stem type).The constricted stem was bent, under a strong flame, to a shape of about 90° elbow. The stem was then connected to the air inlet of the impinger so that the shortened drying tube was in a horizontal position. All connections were made by using solvent-resistant Tygon tubing. Approximate concentrations of 0.4 and 3 mg cm23 of HDI and TDI, to reflect two levels of air sampling, were prepared in dichloromethane for this recovery study.In each case, 10 mm3 of solution were deposited in the bulb with a syringe and the sampling pump was then immediately turned on. The sampling pump was operated at a flow rate of 1 dm3 min21 and the sampling time was 3 h. Experiments on the relative recoveries of the above isocyanates as TRYP derivatives in various solvents were conducted against DMSO. Details of the excellent recoveries using DMSO were cited in NIOSH Method 5522 for three concentration levels of each isocyanate as 91, 103 and 90% for 2,4-TDI, 2,6-TDI and HDI, respectively.HPLC Analysis of Simulated Air Samples Since DMSO and DMF are difficult to remove by evaporation owing to their low volatility, impinger solutions of these solvents were used directly for analysis. A 50 mm3 volume of acetic anhydride was added to the respective solution and set aside for 10 min. A 50 mm3 volume of water was then added and again set aside for 30 min before injection into the HPLC system for quantification.Analytical standards were prepared in the respective solvent for accurate calibration. Impinger solvents of higher volatility, such as butyl acetate or octane, were transferred, with subsequent rinsing with acetonitrile, into a 100 cm3 distillation flask. It should be noted that despite the fact that octane and acetonitrile are only slightly miscible, rinsings with Table 1 Simulated air sampling recoveries for HDI vapour in the impinger solvents butyl acetate, octane and DMSO Sample Impinger Amount of Impinger Amount of set solvent HDI found/mg solvent HDI found/mg 1 Butyl 2.9 DMSO 2.9 acetate 2.4 2.7 2.9 2.9 3.2 3.1 3.4 2.6 Av. 3.0 ± 0.4 Av. 2.9 ± 0.2 (93.8 ± 12%) (90.6 ± 6.3%) 2 Butyl 25.4 DMSO 23.3 acetate 23.7 22.1 22.0 24.9 25.6 23.8 24.9 22.4 Av. 24.3 ± 1.5 Av. 23.3 ± 1.2 (97.6 ± 6.0%) (93.6 ± 4.8%) 3 Octane 4.7 DMSO 4.4 4.3 4.3 4.2 4.5 4.6 3.8 4.0 4.4 Av. 4.4 ± 0.3 Av. 4.3 ± 0.3 (95.7 ± 6.5%) (93.5 ± 6.5%) 4 Octane 36.7 DMSO 31.5 37.6 28.1 35.5 33.6 30.6 34.9 31.5 32.3 Av. 34.5 ± 3.1 Av. 32.0 ± 2.6 (106 ± 9.5%) (98.2 ± 8.0%) Table 2 Simulated air sampling recoveries for TDI vapour in the impinger solvents butyl acetate, octane and DMSO Sample Impinger Amount of Impinger Amount of set solvent TDI found/mg solvent TDI found/mg 1 Butyl 3.0 DMSO 2.9 acetate 2.5 2.8 2.9 2.5 2.8 2.4 2.7 2.9 Av. 2.8 ± 0.2 Av. 2.7 ± 0.2 (96.6 ± 6.9%) (93.1 ± 6.9%) 2 Butyl 31.4 DMSO 24.2 acetate 30.2 30.4 28.2 26.8 27.0 26.8 24.6 26.4 Av. 28.3 ± 2.7 Av. 27.0 ± 2.2 (93.7 ± 8.9%) (89.4 ± 7.3%) 3 Octane 25.6 DMSO 28.8 28.7 27.1 23.9 29.1 27.6 26.2 27.7 31.5 Av. 26.7 ± 1.9 Av. 28.6 ± 2.1 (90.5 ± 6.4%) (96.9 ± 7.1%) Table 3 Relative recoveries for HDI and TDI vapor using butyl acetate and octane in comparison with that for DMSO Relative recovery Set Impinger Amount of isocyanate compared with No.* solvent found/mg using DMSO (%) 1 Butyl acetate 3.0 ± 0.4 (HDI) 103 ± 16 2 Butyl acetate 24.3 ± 1.5 (HDI) 104 ± 8 3 Octane 4.4 ± 0.3 (HDI) 102 ± 10 4 Octane 34.5 ± 3.1 (HDI) 108 ± 18 5 Butyl acetate 2.8 ± 0.2 (TDI) 104 ± 11 6 Butyl acetate 28.3 ± 2.7 (TDI) 105 ± 13 7 Octane 26.7 ± 1.9 (TDI) 93.4 ± 9.5 * References to sets 1–7 for recoveries using DMSO are 2.9 ± 0.2, 23.3 ± 1.2, 4.3 ± 0.3, 32.0 ± 2.6, 2.7 ± 0.2, 27.0 ± 2.2 and 28.6 ± 2.1%. 322 Analyst, April 1997, Vol. 122acetonitrile proved to be effective.The solvent was evaporated under a rotary evaporator between 40 and 45 °C. Acetonitrile was used to rinse the flask and the rinsings were collected in a 10 cm3 volumetric flask. Acetic anhydride and water were added to the solution as described previously except that the volume was reduced to 10 mm3 before diluting to volume for HPLC analysis. Results and Discussion Simulated air sampling recoveries for vaporized HDI and TDI using TRYP were conducted in the impinger solvents butyl acetate, octane and DMSO.Tables 1 and 2 summarize the results. The mean RSD of all the recoveries is 7.8 ± 1.9%. Using an anticipated pump error of not more than 5% as assumed by NIOSH, the total sampling and analytical error, RSDt, is calculated to be 9% based on RSDt 2 = 7.82 + 52. The expected accuracy of recoveries based on a maximum bias of 5% pump error is 5 + 2 RSDt (i.e., smaller than 25%), which meets the NIOSH criterion specified in the NIOSH Standards Completion Program.6,8 Also, considering the standard deviations, the relative recoveries using butyl acetate and octane for sampling compared with that of DMSO as the impinger solvent are excellent, as shown in Table 3.The results confirm that these solvents are comparable in recoveries to DMSO employed in NIOSH Method 5522. The lower detection limit of HDI and TDI using either butyl acetate or octane was approximately 0.01 mg cm23. These detection limits are substantially lower than those obtained using DMSO.It was also found that the recoveries using DMF relative to those of DMSO for vaporized HDI at levels of 4 and 40 mg were 116 ± 12 and 111 ± 14%, respectively. Although butyl acetate and octane were replaced with acetonitrile for HPLC analysis, DMF was used throughout the procedure owing to its difficulty of removal. Unlike DMSO, the detection limit of HDI in DMF was unaffected since an excessive amount of TRYP did not need to be added to the sampling solution.Since DMF has been documented as a suspected carcinogen and butyl acetate or octane was found to be a highly effective solvent, the application of DMF for airborne isocyanates sampling was abandoned. Other solvents such as isobutyl acetate, sec-butyl acetate and tert-butyl acetate were also briefly tested as possible impinger solvents but were dropped when unsuspected analytical interferences were encountered. Since MDI was highly unevaporative, sampling recoveries were carried out in a similar fashion to that of NIOSH by spiking MDI solution into the impinger solutions.2,4,7 The relative recoveries of MDI using butyl acetate and octane compared with DMSO were 104 and 105%, respectively. Conclusions DMSO is a solvent with extremely high polarity, relatively high freezing point and low volatility. To replace DMSO used in NIOSH Method 5522, this study showed that the solvents butyl acetate and octane were more advantageous and practical for sampling and determining airborne isocyanates. Butyl acetate and octane also proved to be just as effective as DMSO with regard to isocyanates recoveries but provided a much lower detection limit for HDI. It was observed that corrosion of the air sampling pump by DMSO was too significant to be ignored. The long term adverse effect of DMSO on an HPLC system is also likely unfavourable. References 1 Wu, W. S., Nazar, M. A., Gaind, V. S., and Calovini, L., Analyst, 1987, 112, 863. 2 Wu, W. S., Szklar, R. S., and Gaind, V. S., Analyst, 1988, 113, 1209. 3 Wu, W. S., Stoyanoff, R. E., Szklar, R. S., Gaind, V. S., and Rakanovic, M., Analyst, 1990, 115, 801. 4 Wu, W. S., and Gaind, V. S., Analyst, 1994, 119, 1043. 5 NIOSH Manual of Analytical Methods, National Institute for Occupational Safety and Health, Cincinnati, OH, 4th edn., 1993, Method 5522, in draft. 6 Bagon, D. A., Warwick, C. J., and Brown, R. H., Am. Ind. Hyg. Assoc. J., 1984, 45, 39. 7 NIOSH Manual of Analytical Methods, National Institute for Occupational Safety and Health, Cincinnati, OH, 1989, Method 5521. 8 Documentation of the NIOSH Validation Tests, NIOSH Publication No. 77–185, National Institute for Occupational Safety and Health, Cincinnati, OH, 1977. Paper 6/06136B Received September 6, 1996 Accepted November 25, 1996 Analyst, April 1997, Vol. 122 323
ISSN:0003-2654
DOI:10.1039/a606136b
出版商:RSC
年代:1997
数据来源: RSC
|
7. |
Microwave-assisted Extraction of Monoterpenols in MustSamples |
|
Analyst,
Volume 122,
Issue 4,
1997,
Page 325-329
N. Carro,
Preview
|
|
摘要:
Microwave-assisted Extraction of Monoterpenols in Must Samples N. Carro, C. M. Garc�ýa and R. Cela* Departamento Qu�ýmica Anal�ýtica, Nutrici�on y Bromatolog�ýa, Universidad de Santiago de Compostela, Avda. de las Ciencias s/n, 15706 Santiago de Compostela, Spain. E-mail: QNRCTD@USC.ES The microwave-assisted extraction of five terpenic compounds associated with the varietal aroma of Vitis vinifera was developed and optimized by means of threeand two-level factorial designs. Four variables (extractant solvent volume, extraction temperature, amount of sample and extraction time) were considered as factors in the optimization process.The results suggest that the solvent volume and the amount of sample to be extracted are statistically significant for the overall recovery of the studied species, although compromise conditions have to be established in order to avoid losses of the extracted compounds in the concentration steps needed when the solvent volume increases.The optimum conditions established were applied to the extraction of real grape must samples and compared with the results given by conventional alternative procedures. Keywords: Microwave-assisted extraction; terpenic compounds; factorial designs The first applications of microwave heating in analytical sample preparation were devoted to sample dissolution and acidic digestion. Since then, the technique has become standard for the preparation of many kinds of biological and environmental samples.1–3 Many of these applications have been performed using domestic devices with open or closed vessels as sample containers.In the mid-1980s, the first experiments on sample extraction were reported. Several types of compounds from plant materials, food and soil were extracted.4–5 Later, the use of solvents transparent to microwaves added new perspectives to the technique and reopened the debate about the ‘microwave effect.’6–13 Logically, the ability to absorb microwave energy and to heat varies with the chemical nature of the species being subjected to microwave irradiation (matrix and solvent).In general, the higher the dielectric constant, the higher the level of absorption of microwave energy. When the matrix to be extracted is non-transparent to microwaves and is capable of inducing microwave heating, a transparent solvent (with a low dielectric constant) is necessary; in this way it is used as a natural coolant that solubilizes the extracted compounds of the matrix and it can induce a rapid temperature increase to well over the boiling-point or superheating, allowing very rapid extractions and the use of low volumes of extractant.Moreover, with a transparent solvent local sample heating allows mild extractions without noticeable degradation of thermally labile compounds. In contrast, if it is necessary to extract a transparent matrix, a non-transparent solvent (with a high dielectric constant) has to be used.Mixed (transparent and non-transparent) solvents can also be used, thus exploiting both the extraction mechanism and the associated features. Although microwave-assisted extraction (MAE) processes working with flammable solvents can be hazardous in the hands of inexperienced operators or when using devices not specifically designed for the purpose, the recent availability of commercial equipment furnished with all necessary security compliances allows safe work in closed-vessel systems, thus rendering the technique a practical and routinely used alternative to more conventional extraction procedures such as Soxhlet extraction, hydrodistillation and the shake-flask, method.Terpenic compounds are responsible for varietal aromas and require multistage time-consuming procedures for extraction from musts or wines. In this paper, a procedure for extracting terpenic compounds from grape musts by means of microwave heating is presented.For this purpose, dichloromethane, a microwave partially transparent (dielectric constant e = 9)8,13,14 and well known universal aroma components extractant, was used. Using this solvent, microwave heating affects mainly the sample matrix and not the species once extracted, thus minimizing artifacts. The procedure was optimized by resorting to two- and three-level factorial designs. Four experimental variables were considered as factors in the optimization process: the solvent volume, the extraction temperature and time and the amount of sample to be extracted.The optimization process was carried out on a synthetic mixture of the five main monoterpenols appearing in Vitis vinifera musts and then applied to real must samples, comparing the results obtained with those offered by more conventional alternatives such as countercurrent liquid–liquid extraction with Freon 1115 and solid–liquid extraction, retaining the terpenic compounds on Amberlite XAD-2 non-ionic resin.14 Experimental Material and Apparatus The terpenic compound standards (linalool, a-terpineol, citronellol, nerol and geraniol), ethanol (99%), diethyl ether and dichloromethane were supplied by Merck (Darmstadt, Germany), non-ionic Amberlite XAD-2 (20-60 mesh) by Sigma– Aldrich (Madrid, Spain), acetonitrile and methanol by Rhomil (Teknokroma, Spain) and Freon 11 by Carburos Met�alicos (Madrid, Spain) Optimization experiments were performed on ethanol–water working standard solutions that contained 0.1 mg l21 of each terpene, prepared by appropriate dilution of a stock standard solution (1000 mg l21) in ethanol.For GC determinations, calibration was carried out at four concentration levels for all species spanning the range 1–16 mg l21. Calibration standards were obtained by dilution of a stock standard solution (100 mg l21) in dichloromethane. All solutions were stored at 4 °C when not in use. The real samples to which the method was applied were white musts produced from cv Albari�no grapes produced in the Controlled Brand of Origin (CBO) R�ýas Baixas (Galicia, Spain) region.MAE experiments were performed on an MES 1000 system (CEM, Matthews, NC, USA) operating at 950 W. This extractor has provision for 12 simultaneous extractions in Teflon-lined closed vessels allowing a temperature of 200 °C, a pressure of Analyst, April 1997, Vol. 122 (325–329) 325200 lb in22 and a nominal volume of 100 ml.One of the vessels is used to control actual temperature and pressure values in the system. The control lined extraction vessel has a different cap and cover to permit the connection of a fiber-optic temperature probe and a pressure sensing tube for monitoring the internal temperature and pressure of the vessel. The fiber-optic temperature probe allows for temperature control of the extraction run (±2 °C). The fiber-optic probe is microwave transparent and is positoned in the control vessel using a glass thermal well.The extracts obtained were concentrated to 0.5 ml under a nitrogen stream in a TurboVap II prepstation (Zymark, Hopkinton, MA, USA). The concentrated extracts (1 ml injected) were analysed by GC using a Hewlett-Packard (Avondale, PA, USA) Model 5890 Series II gas chromatograph equipped with a flame ionization detector (FID) and a Hewlett- Packard Model 7673 A autosampler. A BP-20 (Scientific Glass Engineering, Ringwood, Australia), polyethylene glycol analytical column ( 25 m30.22 mm id, 0.25 mm phase thickness) was used.Chromatographic data were acquired and processed with a Hewlett-Packard Model 3365A data station. Table 1 summarizes the chromatographic conditions used. The identification of the extracted compounds in the real samples was performed on a Varian (Walnutt Creek, CA, USA) Saturn 4 gas chromatograph–ion trap detector mass spectrometer furnished with a Varian Model 1093 septum-equipped programmable injector (SPI) and a capillary column identical with that described above. A NIST 90 mass spectral library fit and retention indexes of peaks against pure standards were the procedures used to identify the extracted species positively or tentatively.Operating conditions are summarized in Table 2. Numerical analysis of the experimental designs by means of the Statgraphics Plus V.6.0 statistical package (Manugistics Rockville, MD, USA).16 Sample Preparation–Extraction Procedure Irrespective of the working conditions imposed by the particular experiment in the factorial design, all samples were prepared by following the same procedure.An amount (fixed or dictated by the factorial design, depending on the experiment) of sample (synthetic mixture of terpenes or real sample) was placed in the Teflon-lined extraction vessel and then the extractant solvent (dichloromethane, its volume also being fixed or dictated by the factorial).The extraction vessels were closed after ensuring that a new rupture membrane was used for each extraction. For this study, single extractions were performed using 50% power and programming extraction times and temperatures as a function of the values dictated by the factorial design. At the end of the extraction program, the sample carousel was removed from the microwave cavity and cooled in a water-bath. The control vessel was returned to the microwave system to check that the extract was at room temperature before opening.The organic phase was separated and dried over anhydrous sodium sulfate, then transferred into a TurboVap concentration tube and the solvent was evaporated under nitrogen (at a pressure of 12 lb in22 and thermostated at 20 °C) to a final volume of 0.5 ml. A 1 ml volume of this concentrated extract was injected into the chromatograph under the operating conditions given in Table 1. Study of the Losses at the Solvent Evaportion Stage Terpene losses at the solvent evaporation stage carried out in the TurboVap prepstation were checked by processing three replicates of 5, 10 and 15 ml of a 0.1 mg l21 solution of each terpene in dichloromethane.These solutions were evaporated to a final volume of 0.5 ml under nitrogen at a pressure of 12 lb in22 and thermostated at 20 °C. Concentrated solutions were then injected into the chromatograph under the conditions given in Table 1. Alternative Conventional Procedures for Terpene Extraction Countercurrent liquid–liquid extraction using Freon 1115 and solid–liquid extraction using the non-ionic resin Amberlite XAD-214 were used to obtain results for comparison with those given by the proposed MAE procedure. Results and Discussion Validation of the Analytical Procedure for Terpenic Compounds by Gas Chromatography As stated in the experimental section, calibration curves were obtained at four concentration levels using appropriately diluted standards.Each concentration level was injected in triplicate and peak areas were fitted by linear regression, the results of Table 1 GC operating conditions Detector temperature 200 °C Injector temperature 190 °C Carrier gas (N2) flow rate 0.9 ml min21 Carrier gas pressure at column head 13 lb in22 Split flow 5.5 ml min21 Injection mode Splitless Purge time 1 min Injected volume 1 ml Temperature program— Initial temperature 30 °C Initial time 0 min Ramp 6 °C min21 Final temperature 150 °C Final time 10 min Table 2 GC–MS conditions Gas chromatography SPI temperature program— Initial temperature 35 °C Rate 250 °C min21 Final temperature 200 °C Final time 10 min Carrier gas (helium) flow rate 1 ml min21 Injection mode On-column Injected volume 0.5–1 ml Oven temperature program— Initial temperature 40 °C Ramp 1 3 °C min21 Final temperature 125 °C Ramp 2 1 °C min21 Final temperature 140 °C Ramp 3 3 °C min21 Final temperature 190 °C Final time 26 min Mass Spectrometry Emission current 80 mA Manifold heater 170 °C Multiplier voltage 2000 V Maximum ionization time 25 000 ms AGC prescan ionization time 100 ms AGC prescan storage level m/z 20 Rf dump value m/z 20 Axial modulation 4 V Scan range m/z 35–300 326 Analyst, April 1997, Vol. 122which are summarized in Table 3.The linearity range was 1–16 mg l21. Preliminary Evaluation of Experimental Conditions of the Analytical System. First Factorial Design The number of variables potentially affecting the efficiency of the microwave extraction is not very large, so we decided to apply a first factorial design at three levels with the aim of evaluating from the first moment the curvature of the response surface.In this way, a preliminary evaluation of the experimental variables could be performed. The sample volume was fixed at 15 ml (the minimum sample size to allow quantitative results taking into account the limit of quantification of the overall procedure).On the other hand, microwave extractions are usually very fast and, according to recent investigations, the total extraction time factor affects the extraction efficiency only slightly,6,17 so, this factor was also fixed at 10 min, which was considered a sufficient time to extract all target compounds in model mixtures and real samples. Moreover, in the system used only the total time is a controllable factor, while the relevant factor is the time the extraction vessel takes to attain the programmed temperature.In all experiments, the applied power was set at 50%; values higher than 50% often lead to sudden increases in the pressure of the vessel with occasional breaking of the rupture membranes, thus stopping the system. In contrast, when values lower than 50% were applied, the extraction times were unnecessarily long. Once these factors had been fixed, only the solvent volume and the extraction temperature remained as parameters to be optimized.A three level orthogonal 3 3 2 full factorial design, which implies 11 randomized experiments, was used to evaluate the response surface defined by these two factors. The factor levels for variables were solvent volume 5–15 ml and extraction temperature 60–95 °C. Obviously, the upper level for the solvent volume factor is limited by the nominal capacity of the extraction vessels, but the extraction temperature (through the inner pressure in the vessel) also limits this level.Extraction yields of linalool, a-terpineol, citronellol, nerol and geraniol in each of 11 runs were used to fit the model response surface. The numerical analysis of the recovery results in this factorial design show that all the species considered exhibit exactly the same behavior. Solvent volume appears as the only statistically significant factor; extraction temperature and the quadratic terms do not. Although extraction temperature does not appear to be statistically significant, it is affected by a positive sign.Also, the interaction between both factors appears with a positive sign, meaning that an increase in both factors values will increase the extraction yield. The conclusion was that the maximum extraction yield should be obtained when both factors are at their maximum tested values, thus making it necessary to develop a new factorial design shifted in this direction compared with the first one.Evaluation of Analyte Losses During the Evaportion Stage During the evaluation of the preliminary factorial design described above, it was also observed that the residuals in the mathematical model increased on increasing the volume of the solvent used to extract the samples. Solvent losses in the extraction vessels during the microwave process were checked in several replicate runs and were found to be below 1%. Hence the solvent evaporation process of the extracts had to be responsible for the large residuals observed.However, nitrogen steam evaporation is usually considered a soft evaporation method and important evaporation losses have been reported18 depending of the nature of the species to be analysed and the nature of the solvent. To evaluate this effect, several evaporation processes were carried out directly on a standard mixture of the compounds but using different solvent volumes. The results of these experiments are given in Table 4, where it can be clearly seen that analyte losses are dependent on not only the nature of the compound but also the solvent volume subjected to evaporation, and can even be 40% of the initial amount.Because the evaporation device used (TurboVap) is fully automated, it is to be expected that the results shown could not be improved upon. Hence the only means of reducing these losses would be to decrease the solvent volume used in the extraction process. Optimization of Microwave-assisted Extraction Process Using a Fixed Solvent Volume.Second Factorial Design The solvent volume was decreased and fixed at 10 ml. A new factorial design was developed using an orthogonal Central Composite 2 3 3 + star design which involved 14 experiments with two center points to model the response surface. Here, because the solvent volume was fixed, the amount of sample to be processed was considered as a factor to be optimized. Table 5 shows the levels used in this design, and also the design matrix and the recoveries [corrected for evaporative losses (average values, as a function of the considered species)] obtained in each experiment for the five species considered.From the numerical analysis of these results, the combined standardized main effects Pareto chart shown in Fig. 1 was Table 3 Calibration data Parameter Linalool a-Terpineol Citronellol Nerol Geraniol Intercept 2237 2325 2258 2280 2205 SE of estimate 51 68 47 67 39 Slope 869 1033 886 931 779 SE of estimate 12 15 11 15 9 r2 1.000 0.999 1.000 1.000 0.999 Table 4 Analyte losses in the solvent evaporation process as a function of the solvent volume Loss (%) Solvent volume/ml Linalool a-Terpineol Citronellol Nerol Geraniol 5 7.9 7.2 8.9 11.1 11.8 10 18.4 17.5 16.9 20.6 22.4 15 29.9 28.3 26.7 32.7 30.6 20 37.4 37.7 30.5 40.1 38.4 Analyst, April 1997, Vol. 122 327Standardized main effects –15 –10 5 0 –5 BC AC CC AB BB AA C(Time) B(Temperature) A(Sample) Geraniol Nerol Citronellol Terpineol Linalool obtained. This chart is the result of combining the individual Pareto charts for each species.This process of mixing Pareto charts violates the condition of sorting the effects but allows direct comparison of the results for the five compounds considered. As can be seen, all the species considered exhibit the same behavior. As can be expected, the amount of sample (having a negative effect) is the most significant factor.However, the temperature (also having a negative sign) and some of the interaction and quadratic terms also appear to be significant for some species. Fig. 2 shows the estimated response surface for nerol obtained by plotting this model. The response surfaces for the other terpenes studied were similar to that depicted in Fig. 2. Better recoveries were obtained when sample amount was minimized and the optimum temperature was around 85–95 °C depending on the particular species.In summary, sample amounts of 5 ml, can be optimally extracted with 10 ml of dichloromethane at a temperature of 90 °C in 10 min, setting the microwave extractor at half power. Under these conditions, the last row in Table 5 summarizes the recoveries and the precision for each compound obtained through series of six independent extractions carried out on different days. Application of the Optimized Procedure to Real Samples and Comparison With Alternative Extraction Procedures To illustrate the applicability of the proposed procedure, must samples obtained from white Albari�no grapes produced in the Controlled Brand of Origin (CBO) R�ýas Baixas (Galicia, Spain) region were extracted and analysed.The GC–MS trace in Fig. 3 Fig. 1 Pareto charts of main effects for the second (central composite) factorial design, obtained by using the terpenic compound recoveries. The vertical lines indicate the statistical significance bound for the effects. Table 5 Factor levels in the second factorial design; design matrix and recoveries for the studied species (a = 1.215 to force orthogonality) Factor Key Low level (2) High level (+) Fixed Amount of sample/ml A 10 40 Extraction time/min C 5 15 Extraction temperature/°C B 80 120 Solvent volume/ml 10 Recovery (%) Expt.No. A B C Linalool a-Terpineol Citronellol Nerol Geraniol 1 0 0 0 71.1 76.4 78.6 71.6 77.1 2 2 2 2 64.2 67.8 76.7 70.4 69.8 3 + 2 2 42.6 44.1 49.2 44.1 42.6 4 2a 0 0 83.7 81.4 89.3 84.4 77.7 5 0 0 + 60.6 64.6 67.7 59.1 60.6 6 +a 0 0 41.4 42.9 45.4 43.1 43.3 7 0 +a 0 43.0 48.0 46.5 40.0 40.1 8 + 2 + 55.4 52.9 57.2 53.9 50.1 9 + + + 41.1 47.1 44.8 38.1 40.2 10 2 + 2 58.1 60.4 68.9 65.1 62.2 11 + + 2 40.3 42.7 44.2 40.6 41.9 12 0 0 2a 55.5 58.4 66.2 55.8 58.2 13 2 + + 76.9 84.8 91.8 70.6 70.7 14 2 2 + 70.3 72.6 84.0 68.3 78.9 15 0 2a 0 58.1 60.6 66.6 57.7 58.0 16 0 0 0 42.4 44.5 47.1 44.8 45.5 Recovery at the optimum 80 ± 5 87 ± 8 86 ± 2 83 ± 1 81 ± 3 Fig. 2 Nerol response surface estimated for the second design, obtained by plotting the two main influential factors. 328 Analyst, April 1997, Vol. 122show the results for a must sample extract obtained by means of the proposed microwave procedure. Table 6 summarizes the identification of the marked peaks in Fig. 3 and compares these results with those obtained using countercurrent liquid–liquid extraction with Freon 11 as extractant15 and solid–liquid extraction on Amberlite XAD-2 non-ionic resin.14 Several differences can be observed regarding the species extracted preferentially in each case and also the recoveries.Restricting the discussion to terpenes, the comparative results between two common extraction procedures and that proposed here are summarized in Table 6, together with some significant operating parameters characterizing each procedure. It can be seen that not only does the proposed procedure give better recoveries but also it is advantageous from the point of view of sample handling, amounts of sample and extractant, total run time and automation.Conclusions Microwave-assisted extraction is a practical alternative for extracting terpenic compounds from must samples. Terpenes and many other aroma components in samples can be extracted with good recoveries using a small volume of organic extractant (10 ml) in only 10 min. Reduced sample handling and the possibility of carrying out several simultaneous extractions by means of commercial non-focused microwave devices make the proposed procedure clearly advantageous from the point of view of sample throughput.Also, the use of transparent or partially transparent extractants together with low temperatures (90 °C) ensures mild extractions without noticeable degradation of the extracted species. This work was supported by the Xunta de Galicia, Direcci�on Xeral de Universidades e Investigaci�on, in the framework of the Project XUGA 20906B95.References 1 Introduction to Microwave Sample Preparation, ed. Kingston, H. M., and Jassie, L. B., American Chemical Society, Washington, DC, 1988. 2 Samra, A. A, Morris, J. S., and Koirtyoham, S. R., Anal. Chem., 1975, 47, 175. 3 Zlotorzynski, A., Crit. Rev. Anal. Chem., 1995, 25, 43. 4 Ganzler, K., Salg�o, A., and Valko, J., J. Chromatogr., 1986, 371, 299. 5 Ganzler, K., and Salg�o, A., Z. Lebensm.-Unters. Forsch., 1987, 184, 274. 6 L�opez-Avila, V., and Young, R., Anal.Chem., 1994, 66, 1097. 7 L�opez-Avila, V., Young, R., Benedicto, J., Ho, P., and Kim, R., Anal. Chem., 1995, 67, 2096. 8 Par�e, J. R. J., Sigouin, M., and Lapointe, J., US Pat., 5 002 784, 1991. 9 Par�e, J. R. J., Sigouin M., and Lapointe J., Eur. Pat., 0485668A1, 1992. 10 Par�e, J. R. J., Belanger, M. R., and Stafford, S. S., TrAC, Trends. Anal. Chem. (Pers. Ed), 1994, 13, 176. 11 Par�e, J. R. J., US Pat., 5 338 557, 1994; 5 377 426, 1995; 5 458 557, 1995. 12 Jean, F.I., Collin, G. J., and Lord, D., Perfum. Flavor., 1992, 17, 35. 13 Barnabas, I. J., Dean, J. R., Fowlis, I. A., and Owen, S. P., Analyst, 1995, 120, 1897. 14 G�unata, Z., Th`ese Docteur Ingenieur, Universit�e de Montpellier, 1984. 15 Garc�ýa-Jares, C. M., Carro-Mari�no, N., Garc�ýa-Mart�ýn, S., and Cela- Torrijos, R., in Fifteenth International Symposium on Capillary Chromatography, ed. Sandra, P., and Devos, G., H�uthig, Heidelberg, 1993, vol. II, pp. 1302–1306. 16 Statgraphics Plus, Version 6, Reference Manual, Manugistics, Rockville, MD, 1992, pp.R47–R130. 17 Vazquez, M. J., Carro A. M., Lorenzo R. A., a. 18 Rodriguez, I., Turnes, M. I., Mejuto, M. C. and Cela R., J. Chromatogr., 1996, 721, 297. Paper 6/05856F Received August 22, 1996 Accepted January 6, 1997 Fig. 3 Chromatogram of an Albari�no grape must extract obtained by the proposed MAE procedure. Peak numbers as in Table 6. Table 6 GC–MS identification and peak areas (normalized to sample volume processed in each technique) of some volatile compounds of Albari�no musts extracted by the three methods Peak area/sample volume Peak Freon XAD MAE No.Compound extraction extraction extraction 1 Isoamyl alcohol* 6 413 41 916 662 032 2 Hexanol* 110 433 15 223 91 865 3 trans-Hex-3-en-1-ol* — — — 4cis-Hex-3-en-1-ol* 722 478 8 527 5 trans-Hex-2-en-1-ol* 4 392 5 321 6 426 6 Ethyl octanoate* 100 40 1 352 7 cis-Furanlinalool oxide 36 15 10 525 8 Benzaldehyde* 91 311 5 153 9 Linalool* 493 187 13 688 10 Hotrienol 983 2 586 201 977 11 Benzeneacetaldehyde* 332 — 55 681 12 a-Terpineol* 657 87 25 786 13 trans-Pyranlinalool oxide 3 344 703 38 603 14 Methyl-2-hydroxybenzoate* 1 137 187 6 171 15 cis-Pyranlinalool oxide 801 118 14 539 16 Citronellol* 54 51 11 460 17 Nerol* 203 103 2 303 18 Geraniol* 160 52 1 333 19 Hexanoic acid* 14 570 7 714 456 307 20 Benzyl alcohol* 8 336 2 332 235 082 21 2-Phenylethanol* 3 057 2 735 670 603 22 Hydroxycitronellol 162 — 7 502 23 Dienediol 1 10 839 5 020 38 304 24 Diethyl malate* 656 889 41 948 25 Octanoic acid* 1 000 503 373 723 26 Dienediol 2 1 378 702 5 809 27 Unknown diol 1 107 230 10 995 28 Decanoic acid* — 217 113 047 29 Unknown terpene 1 375 — 22 114 30 a-Ionine 367 359 15 059 31 Lauric acid* 1 762 — 8 588 32 3-Hydroxy-b-damascone 8 749 4 455 33 187 33 Vanillin* 4 404 4 854 104 321 * Compounds positively identified by comparison with NIST 90 spectral library and retention indices versus pure standards.Analyst, April 1997, Vol. 122 329 Microwave-assisted Extraction of Monoterpenols in Must Samples N. Carro, C. M. Garc�ýa and R. Cela* Departamento Qu�ýmica Anal�ýtica, Nutrici�on y Bromatolog�ýa, Universidad de Santiago de Compostela, Avda. de las Ciencias s/n, 15706 Santiago de Compostela, Spain. E-mail: QNRCTD@USC.ES The microwave-assisted extraction of five terpenic compounds associated with the varietal aroma of Vitis vinifera was developed and optimized by means of threeand two-level factorial designs.Four variables (extractant solvent volume, extraction temperature, amount of sample and extraction time) were considered as factors in the optimization process. The results suggest that the solvent volume and the amount of sample to be extracted are statistically significant for the overall recovery of the studied species, although compromise conditions have to be established in order to avoid losses of the extracted compounds in the concentration steps needed when the solvent volume increases. The optimum conditions established were applied to the extraction of real grape must samples and compared with the results given by conventional alternative procedures.Keywords: Microwave-assisted extraction; terpenic compounds; factorial designs The first applications of microwave heating in analytical sample preparation were devoted to sample dissolution and acidic digestion.Since then, the technique has become standard for the preparation of many kinds of biological and environmental samples.1–3 Many of these applications have been performed using domestic devices with open or closed vessels as sample containers. In the mid-1980s, the first experiments on sample extraction were reported. Several types of compounds from plant materials, food and soil were extracted.4–5 Later, the use of solvents transparent to microwaves added new perspectives to the technique and reopened the debate about the ‘microwave effect.’6–13 Logically, the ability to absorb microwave energy and to heat varies with the chemical nature of the species being subjected to microwave irradiation (matrix and solvent).In general, the higher the dielectric constant, the higher the level of absorption of microwave energy. When the matrix to be extracted is non-transparent to microwaves and is capable of inducing microwave heating, a transparent solvent (with a low dielectric constant) is necessary; in this way it is used as a natural coolant that solubilizes the extracted compounds of the matrix and it can induce a rapid temperature increase to well over the boiling-point or superheating, allowing very rapid extractions and the use of low volumes of extractant.Moreover, with a transparent solvent local sample heating allows mild extractions without noticeable degradation of thermally labile compounds. In contrast, if it is necessary to extract a transparent matrix, a non-transparent solvent (with a high dielectric constant) has to be used.Mixed (transparent and non-transparent) solvents can also be used, thus exploiting both the extraction mechanism and the associated features. Although microwave-assisted extraction (MAE) processes working with flammable solvents can be hazardous in the hands of inexperienced operators or when using devices not specifically designed for the purpose, the recent availability of commercial equipment furnished with all necessary security compliances allows safe work in closed-vessel systems, thus rendering the technique a practical and routinely used alternative to more conventional extraction procedures such as Soxhlet extraction, hydrodistillation and the shake-flask, method.Terpenic compounds are responsible for varietal aromas and require multistage time-consuming procedures for extraction from musts or wines.In this paper, a procedure for extracting terpenic compounds from grape musts by means of microwave heating is presented. For this purpose, dichloromethane, a microwave partially transparent (dielectric constant e = 9)8,13,14 and well known universal aroma components extractant, was used. Using this solvent, microwave heating affects mainly the sample matrix and not the species once extracted, thus minimizing artifacts. The procedure was optimized by resorting to two- and three-level factorial designs.Four experimental variables were considered as factors in the optimization process: the solvent volume, the extraction temperature and time and the amount of sample to be extracted. The optimization process was carried out on a synthetic mixture of the five main monoterpenols appearing in Vitis vinifera musts and then applied to real must samples, comparing the results obtained with those offered by more conventional alternatives such as countercurrent liquid–liquid extraction with Freon 1115 and solid–liquid extraction, retaining the terpenic compounds on Amberlite XAD-2 non-ionic resin.14 Experimental Material and Apparatus The terpenic compound standards (linalool, a-terpineol, citronellol, nerol and geraniol), ethanol (99%), diethyl ether and dichloromethane were supplied by Merck (Darmstadt, Germany), non-ionic Amberlite XAD-2 (20-60 mesh) by Sigma– Aldrich (Madrid, Spain), acetonitrile and methanol by Rhomil (Teknokroma, Spain) and Freon 11 by Carburos Met�alicos (Madrid, Spain) Optimization experiments were performed on ethanol–water working standard solutions that contained 0.1 mg l21 of each terpene, prepared by appropriate dilution of a stock standard solution (1000 mg l21) in ethanol.For GC determinations, calibration was carried out at four concentration levels for all species spanning the range 1–16 mg l21. Calibration standards were obtained by dilution of a stock standard solution (100 mg l21) in dichloromethane.All solutions were stored at 4 °C when not in use. The real samples to which the method was applied were white musts produced from cv Albari�no grapes produced in the Controlled Brand of Origin (CBO) R�ýas Baixas (Galicia, Spain) region. MAE experiments were performed on an MES 1000 system (CEM, Matthews, NC, USA) operating at 950 W. This extractor has provision for 12 simultaneous extractions in Teflon-lined closed vessels allowing a temperature of 200 °C, a pressure of Anal. 122 (325–329) 325200 lb in22 and a nominal volume of 100 ml.One of the vessels is used to control actual temperature and pressure values in the system. The control lined extraction vessel has a different cap and cover to permit the connection of a fiber-optic temperature probe and a pressure sensing tube for monitoring the internal temperature and pressure of the vessel. The fiber-optic temperature probe allows for temperature control of the extraction run (±2 °C).The fiber-optic probe is microwave transparent and is positoned in the control vessel using a glass thermal well. The extracts obtained were concentrated to 0.5 ml under a nitrogen stream in a TurboVap II prepstation (Zymark, Hopkinton, MA, USA). The concentrated extracts (1 ml injected) were analysed by GC using a Hewlett-Packard (Avondale, PA, USA) Model 5890 Series II gas chromatograph equipped with a flame ionization detector (FID) and a Hewlett- Packard Model 7673 A autosampler.A BP-20 (Scientific Glass Engineering, Ringwood, Australia), polyethylene glycol analytical column ( 25 m30.22 mm id, 0.25 mm phase thickness) was used. Chromatographic data were acquired and processed with a Hewlett-Packard Model 3365A data station. Table 1 summarizes the chromatographic conditions used. The identification of the extracted compounds in the real samples was performed on a Varian (Walnutt Creek, CA, USA) Saturn 4 gas chromatograph–ion trap detector mass spectrometer furnished with a Varian Model 1093 septum-equipped programmable injector (SPI) and a capillary column identical with that described above.A NIST 90 mass spectral library fit and retention indexes of peaks against pure standards were the procedures used to identify the extracted species positively or tentatively. Operating conditions are summarized in Table 2. Numerical analysis of the experimental designs was carried out by means of the Statgraphics Plus V.6.0 statistical package (Manugistics Rockville, MD, USA).16 Sample Preparation–Extraction Procedure Irrespective of the working conditions imposed by the particular experiment in the factorial design, all samples were prepared by following the same procedure.An amount (fixed or dictated by the factorial design, depending on the experiment) of sample (synthetic mixture of terpenes or real sample) was placed in the Teflon-lined extraction vessel and then the extractant solvent (dichloromethane, its volume also being fixed or dictated by the factorial).The extraction vessels were closed after ensuring that a new rupture membrane was used for each extraction. For this study, single extractions were performed using 50% power and programming extraction times and temperatures as a function of the values dictated by the factorial design. At the end of the extraction program, the sample carousel was removed from the microwave cavity and cooled in a water-bath. The control vessel was returned to the microwave system to check that the extract was at room temperature before opening.The organic phase was separated and dried over anhydrous sodium sulfate, then transferred into a TurboVap concentration tube and the solvent was evaporated under nitrogen (at a pressure of 12 lb in22 and thermostated at 20 °C) to a final volume of 0.5 ml. A 1 ml volume of this concentrated extract was injected into the chromatograph under the operating conditions given in Table 1.Study of the Losses at the Solvent Evaportion Stage Terpene losses at the solvent evaporation stage carried out in the TurboVap prepstation were checked by processing three replicates of 5, 10 and 15 ml of a 0.1 mg l21 solution of each terpene in dichloromethane. These solutions were evaporated to a final volume of 0.5 ml under nitrogen at a pressure of 12 lb in22 and thermostated at 20 °C.Concentrated solutions were then injected into the chromatograph under the conditions given in Table 1. Alternative Conventional Procedures for Terpene Extraction Countercurrent liquid–liquid extraction using Freon 1115 and solid–liquid extraction using the non-ionic resin Amberlite XAD-214 were used to obtain results for comparison with those given by the proposed MAE procedure. Results and Discussion Validation of the Analytical Procedure for Terpenic Compounds by Gas Chromatography As stated in the experimental section, calibration curves were obtained at four concentration levels using appropriately diluted standards.Each concentration level was injected in triplicate and peak areas were fitted by linear regression, the results of Table 1 GC operating conditions Detector temperature 200 °C Injector temperature 190 °C Carrier gas (N2) flow rate 0.9 ml min21 Carrier gas pressure at column head 13 lb in22 Split flow 5.5 ml min21 Injection mode Splitless Purge time 1 min Injected volume 1 ml Temperature program— Initial temperature 30 °C Initial time 0 min Ramp 6 °C min21 Final temperature 150 °C Final time 10 min Table 2 GC–MS conditions Gas chromatography SPI temperature program— Initial temperature 35 °C Rate 250 °C min21 Final temperature 200 °C Final time 10 min Carrier gas (helium) flow rate 1 ml min21 Injection mode On-column Injected volume 0.5–1 ml Oven temperature program— Initial temperature 40 °C Ramp 1 3 °C min21 Final temperature 125 °C Ramp 2 1 °C min21 Final temperature 140 °C Ramp 3 3 °C min21 Final temperature 190 °C Final time 26 min Mass Spectrometry Emission current 80 mA Manifold heater 170 °C Multiplier voltage 2000 V Maximum ionization time 25 000 ms AGC prescan ionization time 100 ms AGC prescan storage level m/z 20 Rf dump value m/z 20 Axial modulation 4 V Scan range m/z 35–300 326 Analyst, April 1997, Vol. 122which are summarized in Table 3. The linearity range was 1–16 mg l21.Preliminary Evaluation of Experimental Conditions of the Analytical System. First Factorial Design The number of variables potentially affecting the efficiency of the microwave extraction is not very large, so we decided to apply a first factorial design at three levels with the aim of evaluating from the first moment the curvature of the response surface. In this way, a preliminary evaluation of the experimental variables could be performed. The sample volume was fixed at 15 ml (the minimum sample size to allow quantitative results taking into account the limit of quantification of the overall procedure).On the other hand, microwave extractions are usually very fast and, according to recent investigations, the total extraction time factor affects the extraction efficiency only slightly,6,17 so, this factor was also fixed at 10 min, which was considered a sufficient time to extract all target compounds in model mixtures and real samples.Moreover, in the system used only the total time is a controllable factor, while the relevant factor is the time the extraction vessel takes to attain the programmed temperature. In all experiments, the applied power was set at 50%; values higher than 50% often lead to sudden increases in the pressure of the vessel with occasional breaking of the rupture membranes, thus stopping the system. In contrast, when values lower than 50% were applied, the extraction times were unnecessarily long. Once these factors had been fixed, only the solvent volume and the extraction temperature remained as parameters to be optimized.A three level orthogonal 3 3 2 full factorial design, which implies 11 randomized experiments, was used to evaluate the response surface defined by these two factors. The factor levels for variables were solvent volume 5–15 ml and extraction temperature 60–95 °C. Obviously, the upper level for the solvent volume factor is limited by the nominal capacity of the extraction vessels, but the extraction temperature (through the inner pressure in the vessel) also limits this level.Extraction yields of linalool, a-terpineol, citronellol, nerol and geraniol in each of 11 runs were used to fit the model response surface. The numerical analysis of the recovery results in this factorial design show that all the species considered exhibit exactly the same behavior. Solvent volume appears as the only statistically significant factor; extraction temperature and the quadratic terms do not.Although extraction temperature does not appear to be statistically significant, it is affected by a positive sign. Also, the interaction between both factors appears with a positive sign, meaning that an increase in both factors values will increase the extraction yield. The conclusion was that the maximum extraction yield should be obtained when both factors are at their maximum tested values, thus making it necessary to develop a new factorial design shifted in this direction compared with the first one.Evaluation of Analyte Losses During the Evaportion Stage During the evaluation of the preliminary factorial design described above, it was also observed that the residuals in the mathematical model increased on increasing the volume of the solvent used to extract the samples. Solvent losses in the extraction vessels during the microwave process were checked in several replicate runs and were found to be below 1%. Hence the solvent evaporation process of the extracts had to be responsible for the large residuals observed.However, nitrogen steam evaporation is usually considered a soft evaporation method and important evaporation losses have been reported18 depending of the nature of the species to be analysed and the nature of the solvent. To evaluate this effect, several evaporation processes were carried out directly on a standard mixture of the compounds but using different solvent volumes.The results of these experiments are given in Table 4, where it can be clearly seen that analyte losses are dependent on not only the nature of the compound but also the solvent volume subjected to evaporation, and can even be 40% of the initial amount. Because the evaporation device used (TurboVap) is fully automated, it is to be expected that the results shown could not be improved upon. Hence the only means of reducing these losses would be to decrease the solvent volume used in the extraction process.Optimization of Microwave-assisted Extraction Process Using a Fixed Solvent Volume. Second Factorial Design The solvent volume was decreased and fixed at 10 ml. A new factorial design was developed using an orthogonal Central Composite 2 3 3 + star design which involved 14 experiments with two center points to model the response surface. Here, because the solvent volume was fixed, the amount of sample to be processed was considered as a factor to be optimized.Table 5 shows the levels used in this design, and also the design matrix and the recoveries [corrected for evaporative losses (average values, as a function of the considered species)] obtained in each experiment for the five species considered. From the numerical analysis of these results, the combined standardized main effects Pareto chart shown in Fig. 1 was Table 3 Calibration data Parameter Linalool a-Terpineol Citronellol Nerol Geraniol Intercept 2237 2325 2258 2280 2205 SE of estimate 51 68 47 67 39 Slope 869 1033 886 931 779 SE of estimate 12 15 11 15 9 r2 1.000 0.999 1.000 1.000 0.999 Table 4 Analyte losses in the solvent evaporation process as a function of the solvent volume Loss (%) Solvent volume/ml Linalool a-Terpineol Citronellol Nerol Geraniol 5 7.9 7.2 8.9 11.1 11.8 10 18.4 17.5 16.9 20.6 22.4 15 29.9 28.3 26.7 32.7 30.6 20 37.4 37.7 30.5 40.1 38.4 Analyst, April 1997, Vol. 122 327Standardized main effects –15 –10 5 0 –5 BC AC CC AB BB AA C(Time) B(Temperature) A(Sample) Geraniol Nerol Citronellol Terpineol Linalool obtained. This chart is the result of combining the individual Pareto charts for each species. This process of mixing Pareto charts violates the condition of sorting the effects but allows direct comparison of the results for the five compounds considered. As can be seen, all the species considered exhibit the same behavior. As can be expected, the amount of sample (having a negative effect) is the most significant factor.However, the temperature (also having a negative sign) and some of the interaction and quadratic terms also appear to be significant for some species. Fig. 2 shows the estimated response surface for nerol obtained by plotting this model. The response surfaces for the other terpenes studied were similar to that depicted in Fig. 2. Better recoveries were obtained when sample amount was minimized and the optimum temperature was around 85–95 °C depending on the particular species. In summary, sample amounts of 5 ml, can be optimally extracted with 10 ml of dichloromethane at a temperature of 90 °C in 10 min, setting the microwave extractor at half power.Under these conditions, the last row in Table 5 summarizes the recoveries and the precision for each compound obtained through series of six independent extractions carried out on different days. Application of the Optimized Procedure to Real Samples and Comparison With Alternative Extraction Procedures To illustrate the applicability of the proposed procedure, must samples obtained from white Albari�no grapes produced in the Controlled Brand of Origin (CBO) R�ýas Baixas (Galicia, Spain) region were extracted and analysed. The GC–MS trace in Fig. 3 Fig. 1 Pareto charts of main effects for the second (central composite) factorial design, obtained by using the terpenic compound recoveries. The vertical lines indicate the statistical significance bound for the effects.Table 5 Factor levels in the second factorial design; design matrix and recoveries for the studied species (a = 1.215 to force orthogonality) Factor Key Low level (2) High level (+) Fixed Amount of sample/ml A 10 40 Extraction time/min C 5 15 Extraction temperature/°C B 80 120 Solvent volume/ml 10 Recovery (%) Expt. No. A B C Linalool a-Terpineol Citronellol Nerol Geraniol 1 0 0 0 71.1 76.4 78.6 71.6 77.1 2 2 2 2 64.2 67.8 76.7 70.4 69.8 3 + 2 2 42.6 44.1 49.2 44.1 42.6 4 2a 0 0 83.7 81.4 89.3 84.4 77.7 5 0 0 + 60.6 64.6 67.7 59.1 60.6 6 +a 0 0 41.4 42.9 45.4 43.1 43.3 7 0 +a 0 43.0 48.0 46.5 40.0 40.1 8 + 2 + 55.4 52.9 57.2 53.9 50.1 9 + + + 41.1 47.1 44.8 38.1 40.2 10 2 + 2 58.1 60.4 68.9 65.1 62.2 11 + + 2 40.3 42.7 44.2 40.6 41.9 12 0 0 2a 55.5 58.4 66.2 55.8 58.2 13 2 + + 76.9 84.8 91.8 70.6 70.7 14 2 2 + 70.3 72.6 84.0 68.3 78.9 15 0 2a 0 58.1 60.6 66.6 57.7 58.0 16 0 0 0 42.4 44.5 47.1 44.8 45.5 Recovery at the optimum 80 ± 5 87 ± 8 86 ± 2 83 ± 1 81 ± 3 Fig. 2 Nerol response surface estimated for the second design, obtained by plotting the two main influential factors. 328 Analyst, April 1997, Vol. 122show the results for a must sample extract obtained by means of the proposed microwave procedure. Table 6 summarizes the identification of the marked peaks in Fig. 3 and compares these results with those obtained using countercurrent liquid–liquid extraction with Freon 11 as extractant15 and solid–liquid extraction on Amberlite XAD-2 non-ionic resin.14 Several differences can be observed regarding the species extracted preferentially in each case and also the recoveries.Restricting the discussion to terpenes, the comparative results between two common extraction procedures and that proposed here are summarized in Table 6, together with some significant operating parameters characterizing each procedure. It can be seen that not only does the proposed procedure give better recoveries but also it is advantageous from the point of view of sample handling, amounts of sample and extractant, total run time and automation.Conclusions Microwave-assisted extraction is a practical alternative for extracting terpenic compounds from must samples. Terpenes and many other aroma components in samples can be extracted with good recoveries using a small volume of organic extractant (10 ml) in only 10 min.Reduced sample handling and the possibility of carrying out several simultaneous extractions by means of commercial non-focused microwave devices make the proposed procedure clearly advantageous from the point of view of sample throughput. Also, the use of transparent or partially transparent extractants togetheth low temperatures (90 °C) ensures mild extractions without noticeable degradation of the extracted species. This work was supported by the Xunta de Galicia, Direcci�on Xeral de Universidades e Investigaci�on, in the framework of the Project XUGA 20906B95.References 1 Introduction to Microwave Sample Preparation, ed. Kingston, H. M., and Jassie, L. B., American Chemical Society, Washington, DC, 1988. 2 Samra, A. A, Morris, J. S., and Koirtyoham, S. R., Anal. Chem., 1975, 47, 175. 3 Zlotorzynski, A., Crit. Rev. Anal. Chem., 1995, 25, 43. 4 Ganzler, K., Salg�o, A., and Valko, J., J. Chromatogr., 1986, 371, 299. 5 Ganzler, K., and Salg�o, A., Z.Lebensm.-Unters. Forsch., 1987, 184, 274. 6 L�opez-Avila, V., and Young, R., Anal. Chem., 1994, 66, 1097. 7 L�opez-Avila, V., Young, R., Benedicto, J., Ho, P., and Kim, R., Anal. Chem., 1995, 67, 2096. 8 Par�e, J. R. J., Sigouin, M., and Lapointe, J., US Pat., 5 002 784, 1991. 9 Par�e, J. R. J., Sigouin M., and Lapointe J., Eur. Pat., 0485668A1, 1992. 10 Par�e, J. R. J., Belanger, M. R., and Stafford, S. S., TrAC, Trends. Anal. Chem. (Pers. Ed), 1994, 13, 176. 11 Par�e, J. R. J., US Pat., 5 338 557, 1994; 5 377 426, 1995; 5 458 557, 1995. 12 Jean, F. I., Collin, G. J., and Lord, D., Perfum. Flavor., 1992, 17, 35. 13 Barnabas, I. J., Dean, J. R., Fowlis, I. A., and Owen, S. P., Analyst, 1995, 120, 1897. 14 G�unata, Z., Th`ese Docteur Ingenieur, Universit�e de Montpellier, 1984. 15 Garc�ýa-Jares, C. M., Carro-Mari�no, N., Garc�ýa-Mart�ýn, S., and Cela- Torrijos, R., in Fifteenth International Symposium on Capillary Chromatography, ed. Sandra, P., and Devos, G., H�uthig, Heidelberg, 1993, vol. II, pp. 1302–1306. 16 Statgraphics Plus, Version 6, Reference Manual, Manugistics, Rockville, MD, 1992, pp. R47–R130. 17 Vazquez, M. J., Carro A. M., Lorenzo R. A., and Cela R., Anal. Chem., 1997, 69, 221. 18 Rodriguez, I., Turnes, M. I., Mejuto, M. C. and Cela R., J. Chromatogr., 1996, 721, 297. Paper 6/05856F Received August 22, 1996 Accepted January 6, 1997 Fig. 3 Chromatogram of an Albari�no grape must extract obtained by the proposed MAE procedure. Peak numbers as in Table 6. Table 6 GC–MS identification and peak areas (normalized to sample volume processed in each technique) of some volatile compounds of Albari�no musts extracted by the three methods Peak area/sample volume Peak Freon XAD MAE No. Compound extraction extraction extraction 1 Isoamyl alcohol* 6 413 41 916 662 032 2 Hexanol* 110 433 15 223 91 865 3 trans-Hex-3-en-1-ol* — — — 4cis-Hex-3-en-1-ol* 722 478 8 527 5 trans-Hex-2-en-1-ol* 4 392 5 321 6 426 6 Ethyl octanoate* 100 40 1 352 7 cis-Furanlinalool oxide 36 15 10 525 8 Benzaldehyde* 91 311 5 153 9 Linalool* 493 187 13 688 10 Hotrienol 983 2 586 201 977 11 Benzeneacetaldehyde* 332 — 55 681 12 a-Terpineol* 657 87 25 786 13 trans-Pyranlinalool oxide 3 344 703 38 603 14 Methyl-2-hydroxybenzoate* 1 137 187 6 171 15 cis-Pyranlinalool oxide 801 118 14 539 16 Citronellol* 54 51 11 460 17 Nerol* 203 103 2 303 18 Geraniol* 160 52 1 333 19 Hexanoic acid* 14 570 7 714 456 307 20 Benzyl alcohol* 8 336 2 332 235 082 21 2-Phenylethanol* 3 057 2 735 670 603 22 Hydroxycitronellol 162 — 7 502 23 Dienediol 1 10 839 5 020 38 304 24 Diethyl malate* 656 889 41 948 25 Octanoic acid* 1 000 503 373 723 26 Dienediol 2 1 378 702 5 809 27 Unknown diol 1 107 230 10 995 28 Decanoic acid* — 217 113 047 29 Unknown terpene 1 375 — 22 114 30 a-Ionine 367 359 15 059 31 Lauric acid* 1 762 — 8 588 32 3-Hydroxy-b-damascone 8 749 4 455 33 187 33 Vanillin* 4 404 4 854 104 321 * Compounds positively identified by comparison with NIST 90 spectral library and retention indices versus pure standa
ISSN:0003-2654
DOI:10.1039/a605856f
出版商:RSC
年代:1997
数据来源: RSC
|
8. |
Determination of Cadmium in Environmental Samples by HydrideGeneration with In Situ Concentration and Atomic AbsorptionDetection |
|
Analyst,
Volume 122,
Issue 4,
1997,
Page 331-336
Henryk Matusiewicz,
Preview
|
|
摘要:
Determination of Cadmium in Environmental Samples by Hydride Generation with In SituConcentration and Atomic Absorption Detection† Henryk Matusiewicz*a, Mariusz Koprasa and Ralph E. Sturgeonb a Politechnika Pozna�nska, Department of Analytical Chemistry, 60-965 Pozna�n, Poland b Institute for National Measurement Standards, National Research Council of Canada, Ottawa, Ontario, K1A 0R9, Canada A volatile Cd species (presumed to be the hydride) was generated from aqueous solutions by merging sample and tetrahydroborate reductant in a continuous-flow system.The gaseous analyte was transferred onto the inner wall of a graphite tube furnace for in situ preconcentration at 200 °C. Calibration was achieved via the method of standard additions. An absolute detection limit (3sblank) of 10 pg was obtained using KBH4 as reductant. The precision of the determination was 12% (RSD) for a Cd concentration of 0.2 ng ml21 using KBH4. The method was successfully applied to the determination of Cd in several certified environmental reference materials (soil, sediment, sea-water, biological samples) following pressure digestion of the samples by microwave heating in a mixture of acids.Keywords: In situ preconcentration; electrothermal atomic absorption spectrometry; cadmium generation; volatile species Cadmium is one of the most hazardous of elements to human health. For this reason, years of effort have been devoted to the development of more effective, fast, precise and accurate approaches to the determination of this element in biological and environmental materials using numerous classical and modern analytical methods.1 Hydride generation techniques in atomic spectrometry are very popular because of the enhanced concentration detection limits attained by comparison with conventional nebulization of sample into a given atom reservoir.Additionally, isolation and chemical separation of the analyte from the matrix can be achieved.2 However, the number of elements forming volatile hydrides is limited (e.g., As, Bi, Ge, Hg, In, Pb, Sb, Se, Sn, Te, Tl) and only a few can be conveniently transformed into volatile species at room temperature.The development of new vapor generation systems for the determination of additional elements has, therefore, become an important research area in atomic spectrometry. Fortunately, in recent years, determination of Cd by vapor generation has become feasible.3–13 Derivatization with sodium tetraethylborate to yield volatile metal alkyls has been shown to be a viable vapor generation technique for Cd.3–5 Batch methods, coupled with AAS,3 or interfaced with AFS,4 and continuous-flow generation coupled with ICP-AES5 have been studied.Using NaBH4 as the reducing agent, others have reported the generation of a volatile Cd species in organized media,6–8,13 and its determination by AAS,6 ICP-AES7 and cold vapor (CV)AAS.8 Recently, the reaction between Cd and potassium tetrahydroborate (KBH4) was used for the determination of ultratrace levels of Cd with detection by AFS9 and CVAAS.10 This was repeated and confirmed using NaBH4 with CVAAS.11 The sensitivity of vapor generation systems can be increased still further by in situ preconcentration (trapping) techniques,12 which couple hydride generation with the graphite furnace.The subsequent atomization/vaporization procedure is the same as in conventional electrothermal atomization systems.Recently, Infante et al.13 have utilized this method to advantage to enhance the detection limit (LOD) for Cd, achieving an LOD of 60 ng l21, and Bermejo-Barrera et al.14 applied a continuous flow injection cold vapor generation technique coupled with ETAAS for Cd determination. To date, however, no determination of Cd in environmental samples has been reported using such in situ trapping techniques, except for the determination of Cd in sea-water.14 This work describes the generation of a volatile Cd species (presumed to be cadmium hydride8,10,11,13,15,16) utilizing the reaction with tetrahydroborate (NaBH4 and KBH4) in aqueous solutions at room temperature and its application to in situ trapping within a graphite furnace for the determination of trace amounts of Cd in a variety of environmental reference materials.Experimental Instrumentation A Carl Zeiss Jena (Jena, Germany) Model AAS 3 atomic absorption spectrometer (components described in Table 1 in ref. 17) equipped with a Model EA 3 graphite furnace (components described in ref. 18) and a deuterium background corrector was used for some measurements by ETAAS. A Cd hollow cathode lamp (Veky, Beijing, China) was operated at a lamp current of 3 mA. The 228.8 nm resonance line of Cd was selected as the analysis line with an instrumental spectral bandpass of 0.2 nm. High-purity argon was employed as the internal gas. Measurements were performed with Carl Zeiss (Oberkochen, Germany) standard graphite tubes.Atomic absorption measurements were also made using a Perkin-Elmer (Norwalk, CT, USA) Model 5000 spectrometer fitted with a Model HGA-500 graphite furnace and Zeemaneffect background correction. In-house software was used to permit the visualization of the transient signals for each atomization event and the evaluation of peak height and integrated absorbance. A Cd hollow cathode lamp (Hamamatsu Photonics, Bridgewater, NJ, USA) was run at 4 mA with the spectral bandpass of the spectrometer set at 0.7 nm centered at the 228.8 nm resonance line.Although standard Perkin-Elmer pyrolytic graphite coated graphite tubes were used throughout, the coating of pyrolytic graphite was removed from the interior surface of the tube using sandpaper wrapped on a glass rod. It should be noted that uncoated tubes are currently available from † Presented, in part, at the 2nd European Furnace Symposium, St.Petersburg, Russia, May 26–30, 1996, and at the Fourth Rio Symposium on Atomic Spectrometry, Buenos Aires, Argentina, November 24–30, 1996. Analyst, April 1997, Vol. 122 (331–336) 331the manufacturer (Perkin-Elmer, Part No. B007 0699). Gilson (Worthington, OH, USA) Minipuls-2 peristaltic pumps were used to deliver samples and all reagent solutions to the vapor generation cell. For sample digestion, a Plazmatronika (Wroclaw, Poland) Model BM-2S microwave heated digestion system, equipped with exhaust module and closed TFM-PTFE (Hostaflon TFM is a chemically modified PTFE) vessel (30 ml internal volume), was used.A laboratory-built vapor generation system utilizing continuous introduction of the reaction solutions, shown in Fig. 1, was used for the generation of volatile Cd species based on NaBH4 reduction. This system consisted of a manually controlled peristaltic pump (Gilson, Minipuls-3), a three-way Y-shaped tube, a gas handling network, a reaction cell (volume 25 ml) and a silicone transfer line terminating in a quartz tip.One end of this quartz tip (40 mm length, 1.3 mm id, 1.7 mm od) was beveled at 45° to provide an unobstructed exit for gas when it was placed in the graphite furnace. The bevel was required to prevent obstruction of the outlet by the furnace wall which would result in poor precision. The other end was connected to the outlet of the reaction cell via silicone tubing of dimensions 1.8 mm id 3 2.5 mm od having a length of 20 cm.The distance between the Y-shaped tube and the inlet of the reaction cell was about 10 cm. This arrangement allows the sample and reaction solutions to be introduced simultaneously into the reaction cell, where the vapor generation reaction occurs. Presumably, the vapor generation process begins immediately when the two reacting solutions merge. The reaction cell also serves as the gas–liquid separator. A second, continuous vapor generation system, shown in Fig. 2, was used for evaluation of KBH4 as a reductant. This system is similar to that described previously18 except that the dead volume within the reactor was minimized and the lower portion was filled with glass beads to promote the smooth degassing of solution. Additionally, provision was made for rapidly quenching the continued generation of hydrogen and analyte species from the waste solution by ing a solution of 0.5% m/v NaOH at a flow rate of 3 ml min21 into the lower portion of the generator.This was undertaken in an effort to Fig. 1 Schematic diagram of the generation manifold coupled with in situ trapping of Cd (using NaBH4). Fig. 2 Schematic diagram of the generation manifold for use with KBH4. 332 Analyst, April 1997, Vol. 122separate the contribution to the signal arising from the initial ‘primary’ generation from that due to any continued slower evolution from the waste solution which can invariably arise in such systems.A quartz line was used between the generator and the furnace to transfer the generated Cd species. Mass flow controllers (DHN, Warsaw, Poland and MKS Instruments, Andover, MA, USA) were used to regulate the purge and transfer gas flow rates accurately and reproducibly. Reagents A stock standard solution of 1000 mg l21 CdII was prepared by dissolution of the nitrate salt (BDH, Poole, Dorset, UK). Working standard solutions were freshly prepared daily by diluting appropriate aliquots of the stock solution in high-purity water (quartz apparatus, Bi18, Heraeus, Hanau, Germany) containing 0.5 mol l21 HCl.A 2% m/v solution of sodium tetrahydroborate(iii) was prepared by dissolving 0.25 g pellets of NaBH4 (Alfa Inorganics, Ward Hill, MA, USA) in high-purity water, stabilized in 0.1% m/v NaOH solution, and was used without filtration. Solutions were prepared as required before use, although this concentration of base was usually sufficient to keep the reductant stable for 1 d.A 4% m/v solution of potassium tetrahydroborate(iii) was prepared by dissolving KBH4 (Aldrich, Milwaukee, WI, USA) in de-ionized water obtained from a NanoPure (Barnsted, Dubuque, IA, USA) water purification system, stabilized in 0.4% m/v KOH solution, and used without filtration. Solutions were prepared daily. Vesicles of didodecyldimethylammonium bromide (DDAB) were prepared by dissolving the surfactant powder (Aldrich) in water to yield a 0.01 mol l21 solution.All mineral acids used were of the highest quality (Suprapur, Merck, Darmstadt, Germany) or were prepared in-house by subboiling distillation of reagent-grade feedstocks. High-purity water was used throughout. All sample and standard solutions were kept in quartz containers which were washed prior to use with pure HNO3. Sample Preparation Several environmental certified reference materials were analysed for Cd for assessment of accuracy, namely: TORT-2 (Lobster Hepatopancreas), PACS-1 (Marine Sediment) and CASS-3 (Nearshore Seawater) supplied by the National Research Council of Canada (NRCC), CRM-01 (Pepperbush) from The National Institute for Environmental Studies; NIES (Japan), and SRM 2710 (Montana Soil II) supplied by NIST.Approximately 100 mg sub-sample of reference material, 2.5 ml of concentrated HNO3 and 1 ml of 30% H2O2 were placed in the TFM-PTFE vessels of the microwave digestion system. When working with inorganic materials, approximately 200 mg of sample, 2 ml of concentrated HNO3, 1 ml of 40% HF and 0.5 ml of 30% H2O2 were used.The microwave-assisted digestion technique used for samples has been described previously.19 The digested solutions were transferred into 10 ml calibrated flasks and diluted to volume with water. In all cases, a suite of reagent blanks was processed through identical steps. Analytical Procedures All sample and analytical manipulations were conducted in a routine laboratory environment.An aliquot of sample or blank containing Cd was placed in a 2 ml vessel. A 0.1 ml volume of 5% m/v thiourea solution (Fluka, Buchs, Switzerland) and 0.1 ml of Co solution were added (to yield a final concentration of 1 mg ml21 Co) and the mixture was diluted to 1 ml with highpurity water. The quartz transfer line from the reaction cell was placed in the sample insertion hole of the warm (200 °C) graphite tube. The sample and the reducing agent solution (NaBH4) were peristaltically pumped for 60 s at flow rates of 1 ml min21.Vapor generation was accomplished in this continuous- flow mode at the merge point of the Y-tube. The gas and liquid mixture passed through the reaction cell and the volatile Cd evolved was transferred into the furnace where it was sequestered onto the tube surface. A further 60 s purge of the reaction cell with Ar at a flow rate of 200 ml min21 completed the analyte transfer process, whereupon the quartz transfer line was removed from the furnace tube.The sample was atomized at 1800 °C for 3 s using maximum power heating and the tube was subsequently cleaned by heating to 2000 °C for 2 s. Continuous-flow generation measurements of volatile species were also studied using the system shown schematically in Fig. 2. An aliquot containing Cd dissolved in 0.3 mol l21 HCl was placed in a 1 ml vessel. A 0.1 ml volume of 5% m/v thiourea solution, 0.1 ml of Co solution and 0.1 ml of 0.01 mol l21 DDAB were added and the mixture was diluted to a final volume of 1 ml with water.The quartz transfer line from the reaction cell was placed in the sample insertion hole of the warm (200 °C) furnace. The Cd sample was continuously introduced at a rate of 1 ml min21 to merge with a 4% m/v solution of KBH4 (flow rate, 1 ml min21). The merging solution feeds the gas–liquid separator to the graphite tube furnace. A further 60 s Ar purge of the reaction cell completed the analyte transfer process whereupon the quartz transfer line was removed from the furnace tube and the sample atomized as described above.The liquid phase was continuously removed to waste after neutralization with a 0.5% NaOH solution. All samples and blanks were analysed in triplicate by the method of standard additions for each aliquot taken. In addition, calibration was achieved by generating the appropriate hydride from Cd spikes added to acidified high-purity water.The efficiency of the generation and collection process was assessed in each case by comparison of response from known amounts of analyte taken through the vapor deposition procedure with signals derived from atomization of aqueous spikes of analyte introduced directly into the furnace. The experimental conditions are given in Table 1. Results and Discussion The results of this study relate to two distinct aspects of the overall investigation: vapor generation with in situ preconcentration, and its application to practical analysis.Although these are related, there are developmental details for each which are independent. Hence, it is convenient to discuss the development of the vapor collection (concentration) and analysis technique separately. Preliminary experiments were performed by ETAAS using aqueous standard solutions of Cd. It was determined that the generation efficiency of the volatile Cd species greatly depends on the type of reactor used.This was first attempted using a batch-type reaction vessel described previously.20 Signals were erratic and their intensities decreased with time. The simple generator consisting of the reaction cell and Y-shaped mixing tube illustrated in Fig. 1 proved successful for the separation and trapping of a volatile Cd compound. This was probably due to the miniaturized flow system which significantly reduces dead volume and minimizes the decomposition of the volatile species and its deposition onto the cool parts of the reaction vessel and gas transfer line.20 Alternatively, if the volatile Cd species is unstable and prone to decomposition in the aqueous phase, rapid gas–liquid phase separation may prove essential, as with volatile Cu generation.21 Hence, batch generation techniques cannot be efficiently used.The length of tubing connecting the Analyst, April 1997, Vol. 122 333reaction cell to the graphite furnace was also important in determining the magnitude of the analytical signal.In practice, as short a connection as possible was required, the signal disappearing completely when a 40 cm transfer line was used. These phenomena are in agreement with the results reported by Sanz-Medel et al.8 Substantial optimization of the generation parameters was not undertaken, as this information was readily available from the literature pertaining to Cd detection by both atomic fluorescence9 and atomic absorption.10 A 2% (NaBH4) or 4% (KBH4) solution of reductant was used, to which a small concentration of NaOH or KOH stabilizer was added in an effort to maintain the reagent blank as low as possible.The generation efficiency greatly depends on whether sodium or potassium tetrahydroborate is used.9,10 For this reason, the potassium salt was ultimately preferred because a better generation efficiency and detection limit were obtained. Use of KBH4 enhanced sensitivity by 3–4-fold over NaBH4, although the additional influence of the presence of the DDAB vesicles in the former system may also play a role in this enhancement. In the light of the work of Guo and Guo,9 whose results showed that thiourea, particularly in the presence of Co, enhanced the generation efficiency of the volatile Cd species, these reagents were added to the sample solution (to a final concentration of 0.5% and 1 mg ml21, respectively).These reagents improved the final detection limit approximately 8–10-fold.Sanz-Medel and co-workers7,8,13,15,16 have shown that organized media, particularly DDAB vesicles, enhance the generation of volatile Cd species from aqueous solution at room temperature. Therefore, DDAB was also added to the solution (final concentration 0.01 mol l21) when KBH4 was studied. Additionally, it was confirmed that HCl is the most appropriate acid to use.9,10 The nature of the graphite tube was expected to affect the efficiency of trapping.Uncoated and pyrolytic graphite coated tubes were investigated and the efficiency was about 50% better using the former tube, probably due to its higher porosity and surface area. The significant advantage of working with uncoated tubes is that the surface characteristics do not change with use. Although the morphology may be altered with ageing, this has no effect on the collection efficiency. Although several workers have noted that an Ir- or Pd-coated graphite tube generally exhibits a more efficient adsorption for volatile analyte hydrides,12,22 use of reduced Pd or Ir as a scavenger in this work had no beneficial effects on the analytical performance of this system.Infante et al.13 and Bermejo-Barrera et al.,14 however, reported significant enhancement in the efficiency of trapping when a reduced Pd modifier was used to sequester the Cd; no such advantage was observed in these studies. This may be related to the use of uncoated tubes as opposed to the pyrolytic graphite coated tubes investigated by these workers.13,14 Variation in the atomization programme of time and temperature was shown to have little influence on the absorbance signal within the ranges 800–2000 °C and 1–5 s.As high atomization temperatures and prolonged times decrease the useful lifetime of the tubes, an atomization temperature of 1800 °C was chosen, while a time of 3 s was adopted as being optimum. Optimization of the trapping temperature was necessary as no information on this parameter was previously available.Preliminary tests showed that adsorption (trapping) of the analyte was better in a warm, rather than a cold, graphite tube. Thus, in order to choose the optimum furnace temperature, this was varied over the range 20–500 °C in experiments conducted with a 10 ng ml21 solution of Cd, while running a blank in parallel. The optimum trapping temperature was determined to be 200 °C. At lower temperatures, the analyte could not be efficiently decomposed and retained on the graphite surface; at higher temperatures, volatilization of the decomposed analyte species probably occurs.To estimate the collection time necessary for maximum response, different times for vapor generation and for sweeping the residual volatiles into the graphite tube were examined. Maximum adsorption efficiency was reached at 60 s. A time of 120 s was selected for further experiments in order to ensure, with greater certainty, that all of the volatile species had been stripped from solution and transferred into the furnace.The effect of the temperature of the reaction vessel and transport line to the graphite tube was also examined, the reaction vessel being thermostated at different temperatures ranging from 0 to 50 °C. It was found that temperature significantly affects the kinetics and efficiency of the generation reaction. For these experiments, the generation was carried out using a solution having a Cd concentration of 10 ng ml21.Sensitivity was observed to deteriorate as the temperature decreased and, with increasing reaction temperatures, formation of a Cd species and its transport to the graphite tube increased. An improvement of 3-fold could be obtained at 50 °C as opposed to 20 °C. Therefore, operating at temperatures higher than room temperature for the vapor generation can provide a further enhancement in the LOD for Cd.This is in agreement with the results of Vald�es-Hevia y Temprano et al.,5 who produced volatile Cd species by ethylation reactions in a batch generator, but in distinct contrast to the conclusions noted by Sanz-Medel et al.8 for generation using NaBH4. This discrepancy may be the result of differences in the nature of the gas– liquid separation process used in these two studies. The generation efficiency may be influenced by the temperature in such a manner that lower temperatures favor the chemical formation of the species in solution but elevated temperatures enhance its separation from the liquid phase.With very efficient phase separation (e.g., with a grid nebulizer), the former effect would predominate. The flow rate of the Ar carrier gas, used to strip and transport the volatile species to the graphite tube, was varied from 80 to 1600 ml min21. For generation with NaBH4, an increase in carrier gas flow rate from 80 to 350 ml min21 caused an increase in the analytical signal.This may be connected with the more efficient separation of the volatile species from the Table 1 Operating conditions for volatile Cd generation by ETAAS ETAAS experimental conditions NaBH4 KBH4 Wavelength/nm 228.8 228.8 Lamp current/mA 3 4 Band width/nm 0.2 0.7 Background currection Deuterium Zeeman-effect Integration time/s 3 3 Tube type Graphite Graphite Signal measurement Peak height Peak height Trapping temperature/°C 200 200 Atomization temperature/°C 1800 1800 Chemical parameters HCl/sample NaBH4–0.5 mol l21 (flow rate 1 ml min21) KBH4–0.3 mol l21 (flow rate 1 ml min21) NaBH4 2% m/v in 0.1% m/v NaOH (flow rate 1 ml min21) KBH4 4% m/v in 0.4% m/v KOH (flow rate 1 ml min21) Reagents Thiourea (0.1 ml, 5% m/v); Co (1 ppm) Surfactant (KBH4 system) DDAB (0.1 ml, 0.01 mol l21) 334 Analyst, April 1997, Vol. 122reaction solution. The signal reached a plateau over the range 350–1600 ml min21 with the result that a carrier gas flow rate of 350 ml min21 was selected for the preconcentration and determination of Cd using NaBH4.For generation using KBH4, the influence of carrier gas flow rate was also studied between 50 and 700 ml min21. The response reached a plateau over the range 500–700 ml min21, and hence 500 ml min21 was selected for the preconcentration and determination of Cd using KBH4. The effect of sample and reagent flow rates was investigated while keeping a constant 1 : 1 sample : reagent flow rate ratio and the generation and trapping time (60 + 120 s) constant.As expected, the signal increased linearly with the sample flow rate. The final value selected for subsequent experiments was 1 ml min21, which was convenient for use with flow-injection applications. Figures of Merit Absolute blanks were assessed for each of the sample matrices. These were available as decomposition/dissolution blanks for the environmental reference materials and, in this respect, these blanks were analysed as samples.For CASS-3, a simple highpurity water matrix was used as the blank. The procedural blank reflects the total Cd burden arising from several sources, including reagents (high-purity water, HCl, HNO3, HF, H2O2, thiourea, Co, DDAB, reductant), surfaces in contact with the gas and liquid phases, adventitious contamination arising fr manual manipulations, as well as the intrinsic instrument blank. Although these experiments were not conducted in a clean room atmosphere, but in an ordinary laboratory, the major source of the blank was determined to be the tetrahydroborate reagents.Attempts to purify the BH42 reagents and thiourea in an effort to reduce the blank (using Ar sparging of the solution or filtration before use) were unsuccessful. The amount of Cd present as an impurity in the NaBH4 was 3–4-fold greater than that in the KBH4. Using the continuous-flow system (Fig. 2) and KBH4 as reductant, an absolute blank of 3 pg was achieved.Procedural blanks derived from the NaBH4 system (Fig. 1) amounted to 100 pg absolute with digested samples. LOD and precision were evaluated under room temperature conditions (for ease of operation) using 1 ml sample volumes. The absolute LOD, based on a 3sblank criterion, was found to be 10 pg for the KBH4 system, giving a concentration LOD of 10 ng l21. This concentration LOD can be linearly enhanced by processing larger sample volumes.13 The precision of replicate determinations was calculated from the RSD of the mean of ten replicate measurements of a Cd standard using a mass 100-fold above the LOD.Precision was in the range from 10% (NaBH4) to 14% (KBH4) and reflects the cumulative imprecision of all of the sample handling, vapor generation, trapping, atomization and detection steps. Infante et al.13 and Bermejo-Barrera et al.14 achieved a precision of 1.7 and 0.2–1.2% RSD, respectively, with their systems.Such an improvement may be a consequence of the use of Pd as a scavenger and atomization from a platform and incorporating continuous-flow equipment. Whereas this material did not improve the sensitivity of determinations in this work, its effect on precision was not evaluated. Although sample aliquoting and handling was manual, the precision can be significantly improved if an automated system such as the FIAS-400 is used14 instead of the manual technique used here.The overall efficiency of the in situ trapping and generation processes was evaluated by a comparison of the integrated response from the direct injection of a 20 ml volume of aqueous Cd standard solution containing the same mass of Cd on the same tube surface and subjected to an identical thermal programme. Using NaBH4, the efficiency was estimated to be only 2%. Infante et al.13 reported a room temperature efficiency of 50% for Cd generation and trapping using NaBH4. This higher value probably reflects the advantages offered by the DDAB vesicles and, in particular, the rapid gas–liquid phase separation arising from incorporation of a grid nebulizer in the system and possible increased loss of Cd during transport through the silicone tubing as opposed to the glass line used by Infante et al.13 For KBH4, the efficiency was estimated to be 30%.It should be noted that the overall efficiency reflects those of both the vapor generation step and the trapping process, but the latter is likely not to be influenced by the nature of the reductant; hence, KBH4 is a much more efficient generator of volatile Cd, as reported previously by Guo and Guo.9,10 Application to Environmental Samples To establish the accuracy of the approach, several solutions of various certified reference materials were analysed.The results are summarized in Table 2 and have been corrected for dry mass where appropriate. Calibration was achieved using the method of standard additions.Comparison of standard additions with the slope of the calibration graph showed that there were no significant interferences from the sample matrices, and a Cd spike added before the digestion procedure was recovered quantitatively. Therefore, analyses can be performed using a direct calibration graph. The accuracy of these approaches is clear from the agreement of the results with the certified values. Although no interference study was undertaken, it is obvious from the data that there are no systematic errors due to the presence of any of the matrices.Interference from concomitant metals likely to be present in these samples has been shown previously to pose no problems to the generation of the hydrides.9,10 However, to eliminate possible matrix effects, only the method of standard additions was used to obtain accurate results. The precision of replicate determinations is typically better than 10% RSD. Conclusions The combination of a continuous flow system for vapor generation with in situ concentration of the Cd species (presumably CdH2) in a graphite furnace has been evaluated for analytical applications to real samples.This approach provides a viable alternative to the conventional ETAAS technique. The advantageous features offered by this technique for the analysis of environmental samples, which often have complicated matrices, are clear. Although the results presented here relate only to Cd, the system could potentially be extended to other hydride-forming elements such as Tl and probably Zn.Table 2 Analytical results Cd concentration/mg g21 CRM/SRM This work* Certified value† NRCC TORT-2 30.5 ± 1.5‡ 26.7 ± 0.6 NRCC PACS-1 2.8 ± 0.4‡ 2.38 ± 0.2 NIES CRM-01 5.6 ± 0.4‡ 6.7 ± 0.5 NIST SRM 2710 24 ± 2‡ 21.8 ± 0.6 NRCC CASS-3 0.034 ± 0.004§,¶ 0.030 ± 0.005§ * Mean ± standard deviation (n = 3). † Mean and 95% confidence limits. ‡ Analysis using NaBH4.§ Concentration, mg l21. ¶ Analysis using KBH4. Analyst, April 1997, Vol. 122 335Financial support by the Pozna�n University of Technology, Grant No. BW31-499/96, is gratefully acknowledged. H.M. thanks the NRCC for financial support while in Canada. References 1 Stoeppler, M., Int. J. Environ. Anal. Chem., 1986, 27, 231. 2 Nakahara, T., in Advances in Atomic Spectroscopy, ed. Sneddon, J., JAI Press, Greenwich, 1995, vol. 2, pp. 139–178. 3 D’Ulivo, A., and Chen, Y., J.Anal. At. Spectrom., 1989, 4, 319. 4 Ebdon, L., Goodall, P., Hill, S. J., Stockwell, P. B., and Thompson, K. C., J. Anal. At. Spectrom., 1993, 8, 723. 5 Vald�es-Hevia y Temprano, M. C., Fern�andez de la Campa, M. R., and Sanz-Medel, A., J. Anal. At. Spectrom., 1994, 9, 231. 6 Cacho, J., Beltr�an, I., and Nerin, C., J. Anal. At. Spectrom., 1989, 4, 661. 7 Vald�es-Hevia y Temprano, M. C., Fern�andez de la Campa, M. R., and Sanz-Medel, A., J. Anal. At. Spectrom., 1993, 8, 847. 8 Sanz-Medel, A., Vald�es-Hevia y Temprano, M.C., Bordel Garcia, N., and Fern�andez de la Campa, M. R., Anal. Chem., 1995, 67, 2216. 9 Guo, X., and Guo, X., Anal. Chim. Acta, 1995, 310, 377. 10 Guo, X.-W., and Guo, X.-M., J. Anal. At. Spectrom., 1995, 10, 987. 11 Matsumoto, K., Bunseki, 1996, 4, 293. 12 Matusiewicz, H., and Sturgeon, R. E., Spectrochim. Acta, Part B, 1996, 51, 377. 13 Infante, H. G., Fern�andez S�anchez, M. L., and Sanz-Medel, A., J. Anal. At. Spectrom., 1996, 11, 571. 14 Bermejo-Barrera, P., Moreda-Pi�neiro, J., Moreda-Pi�neiro, A., and Bermejo-Barrera, A., J.Anal. At. Spectrom., 1996, 11, 1081. 15 Fern�andez de la Campa, M. R., Segovia Garc�ýa, E., Vald�es Hevia y Temprano, M. C., Aizp�un Fern�andez, B., Marchante Gay�on, J. M., and Sanz-Medel, A., Spectrochim. Acta, Part B, 1995, 50, 377. 16 Sanz-Medel, A., Vald�es-Hevia y Temprano, M. C., Bordel Garc�ýa, N., and Fern�andez de la Campa, M. R., Anal. Proc., 1995, 32, 49. 17 Matusiewicz, H., J.Anal. At. Spectrom., 1989, 4, 265. 18 Matusiewicz, H., Chem. Anal. (Warsaw), 1995, 40, 667. 19 Matusiewicz, H., Anal. Chem., 1994, 66, 751. 20 Sturgeon, R. E., Willie, S. N., and Berman, S. S., Anal. Chem., 1985, 57, 2311. 21 Sturgeon, R. E., Liu, J., Boyko, V. J., and Luong, V. T., Anal. Chem., 1996, 68, 1883. 22 Shuttler, I. L., Feuerstein, M., and Schlemmer, G., J. Anal. At. Spectrom., 1992, 7, 1299. Paper 6/06747F Received October 2, 1996 Accepted January 6, 1997 336 Analyst, April 1997, Vol. 122 Determination of Cadmium in Environmental Samples by Hydride Generation with In SituConcentration and Atomic Absorption Detection† Henryk Matusrgeonb a Politechnika Pozna�nska, Department of Analytical Chemistry, 60-965 Pozna�n, Poland b Institute for National Measurement Standards, National Research Council of Canada, Ottawa, Ontario, K1A 0R9, Canada A volatile Cd species (presumed to be the hydride) was generated from aqueous solutions by merging sample and tetrahydroborate reductant in a continuous-flow system. The gaseous analyte was transferred onto the inner wall of a graphite tube furnace for in situ preconcentration at 200 °C.Calibration was achieved via the method of standard additions. An absolute detection limit (3sblank) of 10 pg was obtained using KBH4 as reductant. The precision of the determination was 12% (RSD) for a Cd concentration of 0.2 ng ml21 using KBH4.The method was successfully applied to the determination of Cd in several certified environmental reference materials (soil, sediment, sea-water, biological samples) following pressure digestion of the samples by microwave heating in a mixture of acids. Keywords: In situ preconcentration; electrothermal atomic absorption spectrometry; cadmium generation; volatile species Cadmium is one of the most hazardous of elements to human health. For this reason, years of effort have been devoted to the development of more effective, fast, precise and accurate approaches to the determination of this element in biological and environmental materials using numerous classical and modern analytical methods.1 Hydride generation techniques in atomic spectrometry are very popular because of the enhanced concentration detection limits attained by comparison with conventional nebulization of sample into a given atom reservoir.Additionally, isolation and chemical separation of the analyte from the matrix can be achieved.2 However, the number of elements forming volatile hydrides is limited (e.g., As, Bi, Ge, Hg, In, Pb, Sb, Se, Sn, Te, Tl) and only a few can be conveniently transformed into volatile species at room temperature.The development of new vapor generation systems for the determination of additional elements has, therefore, become an important research area in atomic spectrometry. Fortunately, in recent years, determination of Cd by vapor generation has become feasible.3–13 Derivatization with sodium tetraethylborate to yield volatile metal alkyls has been shown to be a viable vapor generation technique for Cd.3–5 Batch methods, coupled with AAS,3 or interfaced with AFS,4 and continuous-flow generation coupled with ICP-AES5 have been studied.Using NaBH4 as the reducing agent, others have reported the generation of a volatile Cd species in organized media,6–8,13 and its determination by AAS,6 ICP-AES7 and cold vapor (CV)AAS.8 Recently, the reaction between Cd and potassium tetrahydroborate (KBH4) was used for the determination of ultratrace levels of Cd with detection by AFS9 and CVAAS.10 This was repeated and confirmed using NaBH4 with CVAAS.11 The sensitivity of vapor generation systems can be increased still further by in situ preconcentration (trapping) techniques,12 which couple hydride generation with the graphite furnace.The subsequent atomization/vaporization procedure is the same as in conventional electrothermal atomization systems.Recently, Infante et al.13 have utilized this method to advantage to enhance the detection limit (LOD) for Cd, achieving an LOD of 60 ng l21, and Bermejo-Barrera et al.14 applied a continuous flow injection cold vapor generation technique coupled with ETAAS for Cd determination. To date, however, no determination of Cd in environmental samples has been reported using such in situ trapping techniques, except for the determination of Cd in sea-water.14 This work describes the generation of a volatile Cd species (presumed to be cadmium hydride8,10,11,13,15,16) utilizing the reaction with tetrahydroborate (NaBH4 and KBH4) in aqueous solutions at room temperature and its application to in situ trapping within a graphite furnace for the determination of trace amounts of Cd in a variety of environmental reference materials.Experimental Instrumentation A Carl Zeiss Jena (Jena, Germany) Model AAS 3 atomic absorption spectrometer (components described in Table 1 in ref. 17) equipped with a Model EA 3 graphite furnace (components described in ref. 18) and a deuterium background corrector was used for some measurements by ETAAS. A Cd hollow cathode lamp (Veky, Beijing, China) was operated at a lamp current of 3 mA. The 228.8 nm resonance line of Cd was selected as the analysis line with an instrumental spectral bandpass of 0.2 nm. High-purity argon was employed as the internal gas.Measurements were performed with Carl Zeiss (Oberkochen, Germany) standard graphite tubes. Atomic absorption measurements were also made using a Perkin-Elmer (Norwalk, CT, USA) Model 5000 spectrometer fitted with a Model HGA-500 graphite furnace and Zeemaneffect background correction. In-house software was used to permit the visualization of the transient signals for each atomization event and the evaluation of peak height and integrated absorbance.A Cd hollow cathode lamp (Hamamatsu Photonics, Bridgewater, NJ, USA) was run at 4 mA with the spectral bandpass of the spectrometer set at 0.7 nm centered at the 228.8 nm resonance line. Although standard Perkin-Elmer pyrolytic graphite coated graphite tubes were used throughout, the coating of pyrolytic graphite was removed from the interior surface of the tube using sandpaper wrapped on a glass rod. It should be noted that uncoated tubes are currently available from † Presented, in part, at the 2nd European Furnace Symposium, St.Petersburg, Russia, May 26–30, 1996, and at the Fourth Rio Symposium on Atomic Spectrometry, Buenos Aires, Argentina, November 24–30, 1996. Analyst, April 1997, Vol. 122 (331–336) 331the manufacturer (Perkin-Elmer, Part No. B007 0699). Gilson (Worthington, OH, USA) Minipuls-2 peristaltic pumps were used to deliver samples and all reagent solutions to the vapor generation cell. For sample digestion, a Plazmatronika (Wroclaw, Poland) Model BM-2S microwave heated digestion system, equipped with exhaust module and closed TFM-PTFE (Hostaflon TFM is a chemically modified PTFE) vessel (30 ml internal volume), was used.A laboratory-built vapor generation system utilizing continuous introduction of the reaction solutions, shown in Fig. 1, was used for the generation of volatile Cd species based on NaBH4 reduction. This system consisted of a manually controlled peristaltic pump (Gilson, Minipuls-3), a three-way Y-shaped tube, a gas handling network, a reaction cell (volume 25 ml) and a silicone transfer line terminating in a quartz tip.One end of this quartz tip (40 mm length, 1.3 mm id, 1.7 mm od) was beveled at 45° to provide an unobstructed exit for gas when it was placed in the graphite furnace. The bevel was required to prevent obstruction of the outlet by the furnace wall which would result in poor precision. The other end was connected to the outlet of the reaction cell via silicone tubing of dimensions 1.8 mm id 3 2.5 mm od having a length of 20 cm.The distance between the Y-shaped tube and the inlet of the reaction cell was about 10 cm. This arrangement allows the sample and reaction solutions to be introduced simultaneously into the reaction cell, where the vapor generation reaction occurs. Presumably, the vapor generation process begins immediately when the two reacting solutions merge. The reaction cell also serves as the gas–liquid separator.A second, continuous vapor generation system, shown in Fig. 2, was used for evaluation of KBH4 as a reductant. This system is similar to that described previously18 except that the dead volume within the reactor was minimized and the lower portion was filled with glass beads to promote the smooth degassing of solution. Additionally, provision was made for rapidly quenching the continued generation of hydrogen and analyte species from the waste solution by pumping a solution of 0.5% m/v NaOH at a flow rate of 3 ml min21 into the lower portion of the generator.This was undertaken in an effort to Fig. 1 Schematic diagram of the generation manifold led with in situ trapping of Cd (using NaBH4). Fig. 2 Schematic diagram of the generation manifold for use with KBH4. 332 Analyst, April 1997, Vol. 122separate the contribution to the signal arising from the initial ‘primary’ generation from that due to any continued slower evolution from the waste solution which can invariably arise in such systems.A quartz line was used between the generator and the furnace to transfer the generated Cd species. Mass flow controllers (DHN, Warsaw, Poland and MKS Instruments, Andover, MA, USA) were used to regulate the purge and transfer gas flow rates accurately and reproducibly. Reagents A stock standard solution of 1000 mg l21 CdII was prepared by dissolution of the nitrate salt (BDH, Poole, Dorset, UK).Working standard solutions were freshly prepared daily by diluting appropriate aliquots of the stock solution in high-purity water (quartz apparatus, Bi18, Heraeus, Hanau, Germany) containing 0.5 mol l21 HCl. A 2% m/v solution of sodium tetrahydroborate(iii) was prepared by dissolving 0.25 g pellets of NaBH4 (Alfa Inorganics, Ward Hill, MA, USA) in high-purity water, stabilized in 0.1% m/v NaOH solution, and was used without filtration. Solutions were prepared as required before use, although this concentration of base was usually sufficient to keep the reductant stable for 1 d.A 4% m/v solution of potassium tetrahydroborate(iii) was prepared by dissolving KBH4 (Aldrich, Milwaukee, WI, USA) in de-ionized water obtained from a NanoPure (Barnsted, Dubuque, IA, USA) water purification system, stabilized in 0.4% m/v KOH solution, and used without filtration. Solutions were prepared daily. Vesicles of didodecyldimethylammonium bromide (DDAB) were prepared by dissolving the surfactant powder (Aldrich) in water to yield a 0.01 mol l21 solution. All mineral acids used were of the highest quality (Suprapur, Merck, Darmstadt, Germany) or were prepared in-house by subboiling distillation of reagent-grade feedstocks.High-purity water was used throughout. All sample and standard solutions were kept in quartz containers which were washed prior to use with pure HNO3. Sample Preparation Several environmental certified reference materials were analysed for Cd for assessment of accuracy, namely: TORT-2 (Lobster Hepatopancreas), PACS-1 (Marine Sediment) and CASS-3 (Nearshore Seawater) supplied by the National Research Council of Canada (NRCC), CRM-01 (Pepperbush) from The National Institute for Environmental Studies; NIES (Japan), and SRM 2710 (Montana Soil II) supplied by NIST.Approximately 100 mg sub-sample of reference material, 2.5 ml of concentrated HNO3 and 1 ml of 30% H2O2 were placed in the TFM-PTFE vessels of the microwave digestion system.When working with inorganic materials, approximately 200 mg of sample, 2 ml of concentrated HNO3, 1 ml of 40% HF and 0.5 ml of 30% H2O2 were used. The microwave-assisted digestion technique used for samples has been described previously.19 The digested solutions were transferred into 10 ml calibrated flasks and diluted to volume with water. In all cases, a suite of reagent blanks was processed through identical steps. Analytical Procedures All sample and analytical manipulations were conducted in a routine laboratory environment.An aliquot of sample or blank containing Cd was placed in a 2 ml vessel. A 0.1 ml volume of 5% m/v thiourea solution (Fluka, Buchs, Switzerland) and 0.1 ml of Co solution were added (to yield a final concentration of 1 mg ml21 Co) and the mixture was diluted to 1 ml with highpurity water. The quartz transfer line from the reaction cell was placed in the sample insertion hole of the warm (200 °C) graphite tube.The sample and the reducing agent solution (NaBH4) were peristaltically pumped for 60 s at flow rates of 1 ml min21. Vapor generation was accomplished in this continuous- flow mode at the merge point of the Y-tube. The gas and liquid mixture passed through the reaction cell and the volatile Cd evolved was transferred into the furnace where it was sequestered onto the tube surface. A further 60 s purge of the reaction cell with Ar at a flow rate of 200 ml min21 completed the analyte transfer process, whereupon the quartz transfer line was removed from the furnace tube.The sample was atomized at 1800 °C for 3 s using maximum power heating and the tube was subsequently cleaned by heating to 2000 °C for 2 s. Continuous-flow generation measurements of volatile species were also studied using the system shown schematically in Fig. 2. An aliquot containing Cd dissolved in 0.3 mol l21 HCl was placed in a 1 ml vessel.A 0.1 ml volume of 5% m/v thiourea solution, 0.1 ml of Co solution and 0.1 ml of 0.01 mol l21 DDAB were added and the mixture was diluted to a final volume of 1 ml with water. The quartz transfer line from the reaction cell was placed in the sample insertion hole of the warm (200 °C) furnace. The Cd sample was continuously introduced at a rate of 1 ml min21 to merge with a 4% m/v solution of KBH4 (flow rate, 1 ml min21). The merging solution feeds the gas–liquid separator to the graphite tube furnace.A further 60 s Ar purge of the reaction cell completed the analyte transfer process whereupon the quartz transfer line was removed from the furnace tube and the sample atomized as described above. The liquid phase was continuously removed to waste after neutralization with a 0.5% NaOH solution. All samples and blanks were analysed in triplicate by the method of standard additions for each aliquot taken. In addition, calibration was achieved by generating the appropriate hydride from Cd spikes added to acidified high-purity water.The efficiency of the generation and collection process was assessed in each case by comparison of response from known amounts of analyte taken through the vapor deposition procedure with signals derived from atomization of aqueous spikes of analyte introduced directly into the furnace. The experimental conditions are given in Table 1. Results and Discussion The results of this study relate to two distinct aspects of the overall investigation: vapor generation with in situ preconcentration, and its application to practical analysis. Although these are related, there are developmental details for each which are independent.Hence, it is convenient to discuss the development of the vapor collection (concentration) and analysis technique separately. Preliminary experiments were performed by ETAAS using aqueous standard solutions of Cd. It was determined that the generation efficiency of the volatile Cd species greatly depends on the type of reactor used.This was first attempted using a batch-type reaction vessel described previously.20 Signals were erratic and their intensities decreased with time. The simple generator consisting of the reaction cell and Y-shaped mixing tube illustrated in Fig. 1 proved successful for the separation and trapping of a volatile Cd compound. This was probably due to the miniaturized flow system which significantly reduces dead volume and minimizes the decomposition of the volatile species and its deposition onto the cool parts of the reaction vessel and gas transfer line.20 Alternatively, if the volatile Cd species is unstable and prone to decomposition in the aqueous phase, rapid gas–liquid phase separation may prove essential, as with volatile Cu generation.21 Hence, batch generation techniques cannot be efficiently used. The length of tubing connecting the Analyst, April 1997, Vol. 122 333reaction cell to the graphite furnace was also important in determining the magnitude of the analytical signal.In practice, as short a connection as possible was required, the signal disappearing completely when a 40 cm transfer line was used. These phenomena are in agreement with the results reported by Sanz-Medel et al.8 Substantial optimization of the generation parameters was not undertaken, as this information was readily available from the literature pertaining to Cd detection by both atomic fluorescence9 and atomic absorption.10 A 2% (NaBH4) or 4% (KBH4) solution of reductant was used, to which a small concentration of NaOH or KOH stabilizer was added in an effort to maintain the reagent blank as low as possible.The generation efficiency greatly depends on whether sodium or potassium tetrahydroborate is used.9,10 For this reason, the potassium salt was ultimately preferred because a better generation efficiency and detection limit were obtained.Use of KBH4 enhanced sensitivity by 3–4-fold over NaBH4, although the additional influence of the presence of the DDAB vesicles in the former system may also play a role in this enhancement. In the light of the work of Guo and Guo,9 whose results showed that thiourea, particularly in the presence of Co, enhanced the generation efficiency of the volatile Cd species, these reagents were added to the sample solution (to a final concentration of 0.5% and 1 mg ml21, respectively).These reagents improved the final detection limit approximately 8–10-fold. Sanz-Medel and co-workers7,8,13,15,16 have shown that organized media, particularly DDAB vesicles, enhance the generation of volatile Cd species from aqueous solution at room temperature. Therefore, DDAB was also added to the solution (final concentration 0.01 mol l21) when KBH4 was studied. Additionally, it was confirmed that HCl is the most appropriate acid to use.9,10 The nature of the graphite tube was expected to affect the efficiency of trapping.Uncoated and pyrolytic graphite coated tubes were investigated and the efficiency was about 50% better using the former tube, probably due to its higher porosity and surface area. The significant advantage of working with uncoated tubes is that the surface characteristics do not change with use. Although the morphology may be altered with ageing, this has no effect on the collection efficiency. Although several workers have noted that an Ir- or Pd-coated graphite tube generally exhibits a more efficient adsorption for volatile analyte hydrides,12,22 use of reduced Pd or Ir as a scavenger in this work had no beneficial effects on the analytical performance of this system.Infante et al.13 and Bermejo-Barrera et al.,14 however, reported significant enhancement in the efficiency of trapping when a reduced Pd modifier was used to sequester the Cd; no such advantage was observed in these studies. This may be related to the use of uncoated tubes as opposed to the pyrolytic graphite coated tubes investigated by these workers.13,14 Variation in the atomization programme of time and temperature was shown to have little influence on the absorbance signal within the ranges 800–2000 °C and 1–5 s.As high atomization temperatures and prolonged times decrease the useful lifetime of the tubes, an atomization temperature of 1800 °C was chosen, while a time of 3 s was adopted as being optimum.Optimization of the trapping temperature was necessary as no information on this parameter was previously available. Preliminary tests showed that adsorption (trapping) of the analyte was better in a warm, rather than a cold, graphite tube. Thus, in order to choose the optimum furnace temperature, this was varied over the range 20–500 °C in experiments conducted with a 10 ng ml21 solution of Cd, while running a blank in parallel. The optimum trapping temperature was determined to be 200 °C.At lower temperatures, the analyte could not be efficiently decomposed and retained on the graphite surface; at higher temperatures, volatilization of the decomposed analyte species probably occurs. To estimate the collection time necessary for maximum response, different times for vapor generation and for sweeping the residual volatiles into the graphite tube were examined. Maximum adsorption efficiency was reached at 60 s. A time of 120 s was selected for further experiments in order to ensure, with greater certainty, that all of the volatile species had been stripped from solution and transferred into the furnace.The effect of the temperature of the reaction vessel and transport line to the graphite tube was also examined, the reaction vessel being thermostated at different temperatures ranging from 0 to 50 °C. It was found that temperature significantly affects the kinetics and efficiency of the generation reaction.For these experiments, the generation was carried out using a solution having a Cd concentration of 10 ng ml21. Sensitivity was observed to deteriorate as the temperature decreased and, with increasing reaction temperatures, formation of a Cd species and its transport to the graphite tube increased. An improvement of 3-fold could be obtained at 50 °C as opposed to 20 °C. Therefore, operating at temperatures higher than room temperature for the vapor generation can provide a further enhancement in the LOD for Cd.This is in agreement with the results of Vald�es-Hevia y Temprano et al.,5 who produced volatile Cd species by ethylation reactions in a batch generator, but in distinct contrast to the conclusions noted by Sanz-Medel et al.8 for generation using NaBH4. This discrepancy may be the result of differences in the nature of the gas– liquid separation process used in these two studies. The generation efficiency may be influenced by the temperature in such a manner that lower temperatures favor the chemical formation of the species in solution but elevated temperatures enhance its separation from the liquid phase.With very efficient phase separation (e.g., with a grid nebulizer), the former effect would predominate. The flow rate of the Ar carrier gas, used to strip and transport the volatile species to the graphite tube, was varied from 80 to 1600 ml min21. For generation with NaBH4, an increase in carrier gas flow rate from 80 to 350 ml min21 caused an increase in the analytical signal.This may be connected with the more efficient separation of the volatile species from the Table 1 Operating conditions for volatile Cd generation by ETAAS ETAAS experimental conditions NaBH4 KBH4 Wavelength/nm 228.8 228.8 Lamp current/mA 3 4 Band width/nm 0.2 0.7 Background currection Deuterium Zeeman-effect Integration time/s 3 3 Tube type Graphite Graphite Signal measurement Peak height Peak height Trapping temperature/°C 200 200 Atomization temperature/°C 1800 1800 Chemical parameters HCl/sample NaBH4–0.5 mol l21 (flow rate 1 ml min21) KBH4–0.3 mol l21 (flow rate 1 ml min21) NaBH4 2% m/v in 0.1% m/v NaOH (flow rate 1 ml min21) KBH4 4% m/v in 0.4% m/v KOH (flow rate 1 ml min21) Reagents Thiourea (0.1 ml, 5% m/v); Co (1 ppm) Surfactant (KBH4 system) DDAB (0.1 ml, 0.01 mol l21) 334 Analyst, April 1997, Vol. 122reaction solution. The signal reached a plateau over the range 350–1600 ml min21 with the result that a carrier gas flow rate of 350 ml min21 was selected for the preconcentration and determination of Cd using NaBH4.For generation using KBH4, the influence of carrier gas flow rate was also studied between 50 and 700 ml min21. The response reached a plateau over the range 500–700 ml min21, and hence 500 ml min21 was selected for the preconcentration and determination of Cd using KBH4. The effect of sample and reagent flow rates was investigated while keeping a constant 1 : 1 sample : reagent flow rate ratio and the generation and trapping time (60 + 120 s) constant.As expected, the signal increased linearly with the sample flow rate. The final value selected for subsequent experiments was 1 ml min21, which was convenient for use with flow-injection applications. Figures of Merit Absolute blanks were assessed for each of the sample matrices. These were available as decomposition/dissolution blanks for the environmental reference materials and, in this respect, these blanks were analysed as samples. For CASS-3, a simple highpurity water matrix was used as the blank.The procedural blank reflects the total Cd burden arising from several sources, including reagents (high-purity water, HCl, HNO3, HF, H2O2, thiourea, Co, DDAB, reductant), surfaces in contact with the gas and liquid phases, adventitious contamination arising from manual manipulations, as well as the intrinsic instrument blank.Although these experiments were not conducted in a clean room atmosphere, but in an ordinary laboratory, the major source of e blank was determined to be the tetrahydroborate reagents. Attempts to purify the BH42 reagents and thiourea in an effort to reduce the blank (using Ar sparging of the solution or filtration before use) were unsuccessful. The amount of Cd present as an impurity in the NaBH4 was 3–4-fold greater than that in the KBH4. Using the continuous-flow system (Fig. 2) and KBH4 as reductant, an absolute blank of 3 pg was achieved. Procedural blanks derived from the NaBH4 system (Fig. 1) amounted to 100 pg absolute with digested samples. LOD and precision were evaluated under room temperature conditions (for ease of operation) using 1 ml sample volumes. The absolute LOD, based on a 3sblank criterion, was found to be 10 pg for the KBH4 system, giving a concentration LOD of 10 ng l21. This concentration LOD can be linearly enhanced by processing larger sample volumes.13 The precision of replicate determinations was calculated from the RSD of the mean of ten replicate measurements of a Cd standard using a mass 100-fold above the LOD.Precision was in the range from 10% (NaBH4) to 14% (KBH4) and reflects the cumulative imprecision of all of the sample handling, vapor generation, trapping, atomization and detection steps. Infante et al.13 and Bermejo-Barrera et al.14 achieved a precision of 1.7 and 0.2–1.2% RSD, respectively, with their systems.Such an improvement may be a consequence of the use of Pd as a scavenger and atomization from a platform and incorporating continuous-flow equipment. Whereas this material did not improve the sensitivity of determinations in this work, its effect on precision was not evaluated. Although sample aliquoting and handling was manual, the precision can be significantly improved if an automated system such as the FIAS-400 is used14 instead of the manual technique used here.The overall efficiency of the in situ trapping and generation processes was evaluated by a comparison of the integrated response from the direct injection of a 20 ml volume of aqueous Cd standard solution containing the same mass of Cd on the same tube surface and subjected to an identical thermal programme. Using NaBH4, the efficiency was estimated to be only 2%. Infante et al.13 reported a room temperature efficiency of 50% for Cd generation and trapping using NaBH4.This higher value probably reflects the advantages offered by the DDAB vesicles and, in particular, the rapid gas–liquid phase separation arising from incorporation of a grid nebulizer in the system and possible increased loss of Cd during transport through the silicone tubing as opposed to the glass line used by Infante et al.13 For KBH4, the efficiency was estimated to be 30%. It should be noted that the overall efficiency reflects those of both the vapor generation step and the trapping process, but the latter is likely not to be influenced by the nature of the reductant; hence, KBH4 is a much more efficient generator of volatile Cd, as reported previously by Guo and Guo.9,10 Application to Environmental Samples To establish the accuracy of the approach, several solutions of various certified reference materials were analysed. The results are summarized in Table 2 and have been corrected for dry mass where appropriate.Calibration was achieved using the method of standard additions. Comparison of standard additions with the slope of the calibration graph showed that there were no significant interferences from the sample matrices, and a Cd spike added before the digestion procedure was recovered quantitatively. Therefore, analyses can be performed using a direct calibration graph. The accuracy of these approaches is clear from the agreement of the results with the certified values.Although no interference study was undertaken, it is obvious from the data that there are no systematic errors due to the presence of any of the matrices. Interference from concomitant metals likely to be present in these samples has been shown previously to pose no problems to the generation of the hydrides.9,10 However, to eliminate possible matrix effects, only the method of standard additions was used to obtain accurate results. The precision of replicate determinations is typically better than 10% RSD.Conclusions The combination of a continuous flow system for vapor generation with in situ concentration of the Cd species (presumably CdH2) in a graphite furnace has been evaluated for analytical applications to real samples. This approach provides a viable alternative to the conventional ETAAS technique. The advantageous features offered by this technique for the analysis of environmental samples, which often have complicated matrices, are clear. Although the results presented here relate only to Cd, the system could potentially be extended to other hydride-forming elements such as Tl and probably Zn. Table 2 Analytical results Cd concentration/mg g21 CRM/SRM This work* Certified value† NRCC TORT-2 30.5 ± 1.5‡ 26.7 ± 0.6 NRCC PACS-1 2.8 ± 0.4‡ 2.38 ± 0.2 NIES CRM-01 5.6 ± 0.4‡ 6.7 ± 0.5 NIST SRM 2710 24 ± 2‡ 21.8 ± 0.6 NRCC CASS-3 0.034 ± 0.004§,¶ 0.030 ± 0.005§ * Mean ± standard deviation (n = 3). † Mean and 95% confidence limits. ‡ Analysis using NaBH4. § Concentration, mg l21. ¶ Analysis using KBH4. Analyst, April 1997, Vol. 122 335Financial support by the Pozna�n University of Technology, Grant No. BW31-499/96, is gratefully acknowledged. H.M. thanks the NRCC for financial support while in Canada. References 1 Stoeppler, M., Int. J. Environ. Anal. Chem., 1986, 27, 231. 2 Nakahara, T., in Advances in Atomic Spectroscopy, ed. Sneddon, J., JAI Press, Greenwich, 1995, vol. 2, pp. 139–178. 3 D’Ulivo, A., and Chen, Y., J. Anal. At. Spectrom., 1989, 4, 319. 4 Ebdon, L., Goodall, P., Hill, S. J., Stockwell, P. B., and Thompson, K. C., J. Anal. At. Spectrom., 1993, 8, 723. 5 Vald�es-Hevia y Temprano, M. C., Fern�andez de la Campa, M. R., and Sanz-Medel, A., J. Anal. At. Spectrom., 1994, 9, 231. 6 Cacho, J., Beltr�an, I., and Nerin, C., J. Anal. At. Spectrom., 1989, 4, 661. 7 Vald�es-Hevia y Temprano, M. C., Fern�andez de la Campa, M. R., and Sanz-Medel, A., J. Anal. At. Spectrom., 1993, 8, 847. 8 Sanz-Medel, A., Vald�es-Hevia y Temprano, M. C., Bordel Garcia, N., and Fern�andez de la Campa, M. R., Anal. Chem., 1995, 67, 2216. 9 Guo, X., and Guo, X., Anal. Chim. Acta, 1995, 310, 377. 10 Guo, X.-W., and Guo, X.-M., J. Anal. At. Spectrom., 1995, 10, 987. 11 Matsumoto, K., Bunseki, 1996, 4, 293. 12 Matusiewicz, H., and Sturgeon, R. E., Spectrochim. Acta, Part B, 1996, 51, 377. 13 Infante, H. G., Fern�andez S�anchez, M. L., and Sanz-Medel, A., J. Anal. At. Spectrom., 1996, 11, 571. 14 Bermejo-Barrera, P., Moreda-Pi�neiro, J., Moreda-Pi�neiro, A., and Bermejo-Barrera, A., J. Anal. At. Spectrom., 1996, 11, 1081. 15 Fern�andez de la Campa, M. R., Segovia Garc�ýa, E., Vald�es Hevia y Temprano, M. C., Aizp�un Fern�andez, B., Marchante Gay�on, J. M., and Sanz-Medel, A., Spectrochim. Acta, Part B, 1995, 50, 377. 16 Sanz-Medel, A., Vald�es-Hevia y Temprano, M. C., Bordel Garc�ýa, N., and Fern�andez de la Campa, M. R., Anal. Proc., 1995, 32, 49. 17 Matusiewicz, H., J. Anal. At. Spectrom., 1989, 4, 265. 18 Matusiewicz, H., Chem. Anal. (Warsaw), 1995, 40, 667. 19 Matusiewicz, H., Anal. Chem., 1994, 66, 751. 20 Sturgeon, R. E., Willie, S. N., and Berman, S. S., Anal. Chem., 1985, 57, 2311. 21 Sturgeon, R. E., Liu, J., Boyko, V. J., and Luong, V. T., Anal. Chem., 1996, 68, 1883. 22 Shuttler, I. L., Feuerstein, M., and Schlemmer, G., J. Anal. At. Spectrom., 1992, 7, 1299. Paper 6/06747F Received October 2, 1996 Accepted January
ISSN:0003-2654
DOI:10.1039/a606747f
出版商:RSC
年代:1997
数据来源: RSC
|
9. |
Electrothermal Atomic Absorption Spectrometric Determination ofLead and Tin in Slurries. Optimization Study |
|
Analyst,
Volume 122,
Issue 4,
1997,
Page 337-343
V. I. Slaveykova,
Preview
|
|
摘要:
Electrothermal Atomic Absorption Spectrometric Determination of Lead and Tin in Slurries. Optimization Study V. I. Slaveykova*† and M. Hoenig Ministere des Classes Moyennes et de l’Agriculture, Institut de Recherches Chimiques, Leuvensesteenweg 17, B3080, Tervuren, Belgium A comparative study of the efficiencies of W, Mg, Mg + Pd, Ir + Mg and Ir + W + PO4 32 as chemical modifiers for the thermal stabilization of Pb and Sn in slurries was performed. The Ir + Mg modifier contributes more than the others to minimizing matrix effects and preventing double peak formation.The influences of slurry concentration, amount of modifier and pyrolysis step on the integrated absorbance signals for Pb and Sn in sediment slurry were studied. The potential of fractional factorial design was explored to evaluate the effect on the absorbance signals of different factors such as drying and pyrolysis times, pyrolysis and atomization temperatures, presence of modifier and incorporation of a ‘cool-down’ step in the electrothermal program.The Pb and Sn integrated absorbance signals are increased by lower and higher pyrolysis and atomization temperatures, respectively. The Ir + Mg modifier also increases the absorbance signal for Sn. The use of an Ir + Mg modifier and ultrasonic agitation in the direct introduction of slurries into the graphite atomizer provides a consistent performance with good sensitivity, precision and analytical recovery. Keywords: Electrothermal atomic absorption spectrometry; slurry; iridium + magnesium modifier; fractional factorial design; lead; tin The methodology of slurry sampling (SS) ETAAS is now well established and generally used in the determination of trace elements in environmental samples, particularly when there is a real need to analyse microsamples.1,2 The most attractive advantages of slurry analysis over acid digestion or fusion procedures and direct solid sampling can be briefly summarized as follows: (i) reduced sample pretreatment and increase in the speed of the whole analytical procedure; (ii) low contamination risk; (iii) fewer possibilities of analyte losses during the sample pre-treatment; (iv) use of conventional sample introduction systems without instrumental modification; (v) adequate calibration with aqueous standards; and (vi) good precision and accuracy.Some problems and potential drawbacks of SS-ETAAS worth addressing are the high background levels, lack of slurry homogeneity, irreproducible sample introduction and inhomogeneous distribution of the analytes in samples.3 To ensure good repeatability of the measurements, a representative number of particles must be analyzed. It has been shown that only a very fine granulometry of the slurry ensures correct results.Thus, the grinding efficiency is one of the critical factors in the analysis. For slurry analysis of soils and sediments, where trace elements are associated with the very fine granulometric fraction only, the grinding efficiency is less important than for the other type of samples, where the analytes are distributed more homogeneously (e.g., plants, glass materials).1–4 The errors associated with slurry analysis can be minimized by working with small particle sizes, larger total masses of material and narrower particle size distributions.4 Homogeneous slurries can be obtained by using ultrasonic and magnetic stirring, vortex or gas mixing of the slurries5–7 and the addition of various stabilizing agents such as glycerol,8 Viscalex9 or Triton X-100.10 Methods based on modern furnace technology, e.g., platforms, appropriate modifiers, fast absorbance recording and efficient background correction, were usually applied to slurry analysis, but matrix interferences were not avoided completely. 11 Two approaches to slurry analysis by ETAAS has been described.2 The first includes elimination of the pyrolysis step and injection of the sample into the pre-heated furnace in order to decrease the analytical time and increase the sample throughput.12–14 Hoenig and Cilissen5 demonstrated that the precisions reached with conventional and fast furnace programs were comparable, but also stressed the importance of peak area measurements to obtain correct results. The second approach is based on conventional programs and chemical modification techniques.Most often a Pd + Mg modifier has been used to determine, for example, Pb in mussels, fruit, coal fly ash and sediment slurries,10,15–17 and Sn in sediments and coal fly ash,17,18 while NH4H2PO4 has been used for Pb in paprika16 and food.19 The aim of this study was to improve the analytical performance of the ETAAS determination of Pb and Sn in slurries.The capabilities of fractional factorial design were explored to study the influence of various parameters, such as drying and pyrolysis times, pyrolysis and atomization temperatures, presence of modifier, incorporation of a ‘cool-down’ step in the electrothermal program and the volume of the samples introduced, on the absorbance signal.The efforts were aimed at using the advantages provided by both a chemical modifier and a fast temperature program. The stabilizing actions of different modifiers such as Mg, W, Mg + Pd, Mg + Ir and Ir + W + PO4 32 were compared for Pb and Sn in marine sediments and plants slurries. Special attention was paid to a detailed investigation not only of the pyrolysis stage but also of the peak shape and modifier potential for peak isoformation in slurries and aqueous solutions.A comparison between fast and conventional furnace programs in the absence and the presence of a modifier was also made. Experimental Apparatus Measurements were carried out using a Varian (Palo Alto, CA, USA) SpectrAA 400 Zeeman-effect atomic absorption spectrometer, equipped with a GTA 96 graphite atomizer and a † On leave from the Faculty of Chemistry, University of Sofia, 1 James Bourchier Blvd., Sofia 1126, Bulgaria.Analyst, April 1997, Vol. 122 (337–343) 337programmable sample dispenser. A Hamamatsu (Phonotonics K.K., Shimokanzo, Japan) hollow cathode lamp for Pb and a Juniper (Harlow, Essex, UK) hollow cathode lamp for Sn were used. The instrument parameters for the Varian SpectrAA 400 Zeeman-effect atomic absorption spectrometer were adjusted according to the manufacturer’s recommendations.Pyrolytic graphite coated tubes with specially design fork platforms, RW 0961/3 PyC N 299990/89, Series V/64/01701 (Ringsdorff, Bonn, Germany), were used. A Vibra Cell VC 50 high intensity ultrasonic processor (Sonics and Materials, Danbury, USA) was utilized in the preparation of the slurries. The optimized heating program for the GTA 96 graphite tube atomizer is given in Table 1. A ‘cool-down’ step at 100 °C was incorporated in the program in order to normalize the atomization conditions for experiments at different pre-treatment temperatures.After each heating cycle the tube was cleaned up using the maximum permissible temperature for the tube. Reagents Stock standard solutions of Pb (1 g l21 PbII in 2% HNO3) and Sn (1 g l21 SnII in 20% HCl) were obtained from Servi-Lab (Wakken, Belgium). Working standard solutions containing 200 mg l21 of the analytes were prepared daily from the stock standard solutions by dilution with de-mineralized water and acidified with HNO3 (0.1% final concentration).Chemical modifier solutions, each 1% m/V, were prepared using the following reagents: (NH4)2IrCl6, (NH4)2PdCl4 and WO3 (Specpur; Johnson-Matthey, Royston, Hertfordshire, UK) and Mg(NO3)2 (analytical-reagent grade; Merck, Darmstadt, Germany). The Pd, Ir and Mg modifier solutions were prepared as described by Hoenig20 and W modifier solution by dissolution of WO3 in NH3 according to Slaveyhova et al.21 The modifier solutions were mixed in different proportions and the composition of the modifiers used is pointed out in the text for each of the systems studied. Certified Reference Materials Plant and sediment materials with certified or reference contents were obtained from the National Research Council of Canada (Marine Sediments MESS-1, BCSS-1 and PACS-1), the Community Bureau of Reference (BCR, now SMT, Geel, Belgium) (RM 62, Olea europaea) and the National Institute for Standards and Technology (NIST) (Gaithersburg, MD, USA) (SRM 1572, Citrus Leaves).Sample Pre-treatment Procedure Plant and sediment materials were ground in a boron carbide mortar to reduce the particle size to < 50 mm. Amounts of 3–10 mg of finely powdered solid samples were weighed directly into Varian autosampler microvials and treated with 50 ml of concentrated HNO3 overnight, before addition of 950 ml of demineralized water. This ensured that the sample was partly digested and the analytes were mobilized into solution.A portion of the trace elements present remained in the solid phase, which was dispensed together with the enriched solution into the graphite tube during sampling. Just before each sampling, the slurry was homogenized with a high intensity ultrasonic probe. The autosampler was programmed to aspirate 10 ml of each slurried sample and 5 ml of the modifier solution successively, which were mixed in situ (sampler automixing mode) during the injection on to the platform. The hot injection mode of the autosampler at 120 °C was used.Thermal Pre-treatment Study The optimum pyrolysis temperatures were determined at fixed atomization temperatures by their systematic variation and measurements of the integrated absorbance. The possibility of a sensitivity drift effect was avoided by randomizing measurements and using the mean of two or three values at each temperature. Normalized absorbance measurements were used, assigning a value of 1 to the plateau absorbance for the species.When there was no pronounced plateau on the pyrolysis curves, the maximum integrated absorbance signal was used to normalize the signal. Results and Discussion Comparison of Various Chemical Modifiers Efforts were directed to using the advantages provided by W, Mg, Ir + Mg and Ir + W + PO4 32 modifiers and to compare their influence on the atomization signal and thermal pre-treatment losses of Pb and Sn in sediment and plant slurries.The Wcontaining modifiers are very effective thermal stabilizers for Pb and Sn in aqueous solutions and biological samples21,22 and Mg is both a well recognized modifier in ETAAS and an efficient ashing aid (for a review, see ref. 23). Iridium was chosen as a main component of the mixed modifiers because of its pronounced stabilizing efficiency in the determination of volatile analytes.20,24,25 It is expected that addition of Mg and W to an Ir solution will further stabilize Ir to higher pyrolysis temperatures.Tsalev et al.24 observed volatilization losses of Ir at temperatures > 2050–2100 °C in WC-coated platforms. The presence of Mg and W would also lead to a more homogeneous distribution of Ir on the platform surface.23,26 Thus, migration of metal droplets at high temperature and possible coalescence resulting in larger modifier droplets would be minimized. Therefore, certain effects of overstabilization24,25 and incomplete analyte vaporization from Ir during atomization could be avoided.The performance of each chemical modifier was established by considering maximum loss-free pyrolysis temperatures, characteristic masses and impact of the peak shape. The minimum possible amounts of modifiers providing stabilization to at least 1000 °C were used in order to avoid rapid deterioration of the furnace surface and drift in the sensitivity. The possible condensation of chemical modifiers on the cooler end of the tube and light scattering by microparticles and therefore a high background level could also be reduced by using the minimum possible amount of modifier.27,28 The maximum pyrolysis temperatures for Pb and Sn in marine sediments and plant slurries in the presence of mixed Ir + Mg and Ir + W + PO4 32 modifiers, in addition to the single components, are given in Table 2.A comparison with Mg + Pd modifier also is made. Further details and specific features of Table 1 Temperature program for GTA 96 graphite tube atomizer Ar flow rate/ Step Temperature/°C Time/s l min21 Read 1 300 30 3.0 No 2 Var.* 5 3.0 No 3 Var.* 10 3.0 No 4 100 5 3.0 No 5 100 1 0 No 6 2000† 0 0 Yes 7 2000† 5 0 Yes 8 120 10 3.0 No * Temperatures varied within the ranges given in Figs. 1 and 2. † 2500 °C for Sn. 338 Analyst, April 1997, Vol. 122the pyrolysis curves are presented below. It is evident that maximum loss-free pyrolysis temperatures in slurries are much lower than those in aqueous solutions for all of the modifiers studied.The modifiers applied have a substantial stabilizing effect for Pb in slurries, but for Sn the increase in maximum pyrolysis temperatures is only 100–200 °C. The maximum pyrolysis temperatures obtained in marine sediment are higher for Pb and lower for Sn than those reported in the presence of Mg + Pd.18 Typical examples of thermal pre-treatment curves for Pb and Sn in marine sediment and Olea europaea slurries are presented in Figs. 1 and 2, respectively. In the absence of modifier, the integrated absorbance signal for Pb in plant slurries gradually decreases with increasing pyrolysis temperature, whereas plant matrix components stabilize Sn, for which a maximum pyrolysis temperature of 800 °C was observed. The addition of a modifier strongly stabilizes Pb in plants and moderately stabilizes Sn. The stabilizing action of the modifiers decreases in the series Mg Å Mg + Pd < Ir + Mg < Ir + W + PO4 32 for Pb and Mg Å Mg + Pd < Ir + W + PO4 32 < Ir + Mg for Sn.Moreover, both Pb and Sn are stabilized to almost identical temperatures in the presence of the same modifier. The main components of marine sediments, Si, Al, Fe and K, could act as thermal stabilizers, thus the maximum loss-free temperature for Pb and Sn in slurries is 800 and 900 °C, respectively. At low pyrolysis temperatures, losses in the integrated absorbance signal were observed for both Pb and Sn in sediments.The diminution of the signal at low pre-treatment temperatures could be explained by redistribution of the sample in the atomizer and as a result the analyte is less effectively atomized in the cooler region towards the end of the tube. The vaporization of volatile Pb and Sn compounds at low temperature and their transformation into less volatile form during thermal pre-treatment could also be taken into account to explain the effect observed.The formation of Cl-, O- and Scontaining compounds, in competition with free atom formation, was observed when slurries were vaporized and Pb and Sn gas phase molecule species were studied.29 However, signal losses for Pb and Sn in sediment slurries were not observed at low temperatures in the presence of any of the modifiers studied. Moreover, the stabilizing efficiency increases in the order Mg Å W Å Mg + Pd < Ir + Mg < Ir + W + PO4 32 for Pb.The maximum pyrolysis temperatures for Sn are similar for all modifiers studied, which is consistent with the decisive role of the matrix component on the behaviour of Sn in the graphite atomizer. The peak area/peak height ratio was used as an indicator for peak shape in terms of whether sharp and narrow or broad and wide profiles were observed. A peak area/peak height ratio near 1 s is the optimum from the viewpoint of the determination of low concentrations near the LOD.The optimum peak should be sharp and narrow to be distinguishable from the noise and with a sufficiently high peak area (peak area measurements). In the absence of a modifier and in the presence of Mg and W, the absorbance profiles were relatively fast and narrow with peak area/peak height ratios << 1 s for both Pb and Sn in sediment and plant slurries. The addition of a platinum group metal (Ir or Pd) as a modifier component produced a shift in the peak maximum and signal broadening.The addition of Ir + Mg produced the most symmetrical peaks and the peak area/peak height ratios were near 1 s for Pb in all samples and for Sn in sediments. In plant slurry a peak area/peak height ratio close to 1 s was also observed for Sn in plants in the presence of Ir + W + PO4 32 and for Pb in the presence of Pd + Mg. The existence of double absorbance peaks and also the modifier capability to ‘level off’ an absorbance signal in samples with different matrix components were also studied.In contrast to Bermejo-Barrera et al.’s results,18 double peak formation was not observed for Pb under our conditions and the background signal was almost negligible. Modification with Ir + Mg resulted in the best signal isoformation and minimum background absorbance in both sediment and plant slurries for Sn. A certain overcompensation of the signal and/or the appearence of double peaks were observed for Sn in sediments modified with Pd + Mg or Ir + W + PO4 32 and, moreover, this effect was independent of pyrolysis and atomization temperature variations.A possible cause of double peak formation is the existence of two different forms of an analyte in sediment.25 Also, Sn leached into the nitric acid solution may behave differently from the Sn still in the sediment. Amount of Chemical Modifier The amounts of chemical modifiers were varied between 1 and 50 mg for W, Mg and Ir. Higher amounts led to an increase in the maximum pyrolysis temperatures, which approach 1100, 1100 and 1200 °C for Pb and 1400, 1400 and 1500 °C for Sn in sediments.With more than 20 mg of Mg or W, enhancements of the background absorbance were observed. The addition of more than 10 mg of Ir led to overstabilization and atomization temperatures above 2600–2700 °C were required for the effective release of analyte atoms. The peaks were very broad and real time integration was impossible. The effect of the ratio of the two modifier components Ir and Mg on the maximum pre-treatment temperatures and absorbance signals for Pb and Sn in sediments was studied.An increase of the Ir : Mg ratio led to higher stabilization temperatures for both elements but affected only slightly the absorbance peak characteristics. Influence of Slurry Concentration The influence of different slurry concentrations on the analytical signal, sensitivity and precision was investigated. The slurry concentration for Sn determination was varied between 1 Table 2 Pyrolysis temperatures for Pb and Sn in sediment and plant slurries in the presence of various modifiers Pb Sn Citrus Olea Citrus Olea Aqueous Sediment leaves europaea Aqueous Sediment leaves europaea Modifier solution slurry slurry slurry solution slurry slurry slurry None 600* 800† 300* 200* 400* 700† 600* 700 15 mg Mg 900 1100 1000 1000 1400 900 1000 900 15 mg W 1000 1000 1000 1100 1500 1000 1000 1000 5 mg Ir + 15 mg Mg 1200 1100 1000 1000 1500 1000 1000 1000 5 mg Ir + 10 mg W + 25 mg PO4 32 1200 1200 1200 1200 1450 1100 1100 1000 10 mg Pd + 15 mg Mg 1200 1100 1000 1000 1400 1000 1000 900 * The signal diminished gradually with increase in temperature.†A signal decrease at low temperature was observed. Analyst, April 1997, Vol. 122 339ratio is > 1 with Ir + Mg and Mg + Pd and < 1 with the other modifiers studied. Optimization by Fractional Factorial Design To study the influence of different factors on the atomization of Pb and Sn in marine sediment slurries, the potential of fractional factorial design was explored.The fractional factorial design at two levels, 27–3, was applied to determine the main effects bi and some of the two-factor interactions bij.30 The variables considered and levels studied are given in Table 3. For x5, minus and plus levels correspond to absence of a modifier and the presence of Ir + Mg modifier, respectively. The plus level for x6 refers to the incorporation of a ‘cool-down’ step at 100 °C in the program.The Ir + Mg modifier was chosen in the optimization study because of its pronounced stabilizing efficiency and the relative isoformation of the analytical signal in slurries and aqueous solutions. The volume of sample injected is also believed to influence the reproducibility of the analysis. Droplets of different size are spread over the inner surface of the platform and different sample distributions on the substrate could be expected.On the other hand, the amount of the analyte in the graphite furnace is the same for different volumes of the sample injected, because the slurry concentration is different. Hence the duration of the drying stage might play an important role in the slurry particle distribution on the platform and also on the rate of nucleation and crystal growth of the part of the sample leached in the HNO3 and thus influence the reproducibility.Table 4 shows the 27–3 design. Factors x5, x6 and x7 are obtained by multiplying columns x1x2x3, x2x3x4, and x1x3x4, respectively. A similar approach was used in the determination of Cd by ETAAS for different atomization systems.31 The statistical evaluation reveals that the main effects b3–b6 for Pb and b4–b6 for Sn are significant. They correspond to pyrolysis and atomization temperatures, the presence of a modifier and a ‘cool-down’ step. These statistical estimates suggest that an increase in the drying time leads to a moderately increased absorbance signal and better reproducibility for both Pb and Sn in a slurry.Lower atomization and pyrolysis temperatures for Pb also give higher atomization signals, whereas the Sn absorbance increases significantly with increase in the pyrolysis and atomization temperatures (see Table 5). The differences observed between Pb and Sn as regards the influence of pyrolysis and atomization temperatures on the atomization signal are probably due to different distributions of the analytes in the sediment.31 Moreover, the addition of HNO3 during sample preparation produces about 80% extraction of the lead from the particles into the solution and only 40% for tin.Thus a higher atomization temperature was required to insure release of Sn from slurry particles. The addition of a chemical modifier also is of prime importance for Sn but not for Pb determination. This could be related to differences in the efficiency of the sediment matrix as a thermal stabilizer by itself.The confounded terms formed by the two-factor interactions b12 + b35 + b67, b13 + b25 + b47, b14 + b37 + b56, b23 + b15 + b46 and b34 + b26 + b17 for Pb and Sn and b45 + b16 + b27 for Sn are significant at the 95% probability level. The interaction diagrams were used to resolve these interaction terms and examples are represented in Fig. 4. Each box shows the combined effect of the confounded terms.The value presented in each quadrant is the average signal obtained in each case. Among the confounded terms, only for Pb do the interaction terms b12 and b14 show a tendency to improve the signal slightly when shorter drying times and lower atomization temperatures are used. For Sn there is a pronounced tendency for signal enhancement when a modifier is added to the sample and the atomization temperature increases. Figures of Merit The calibration graph and standard additions methods were compared for both Pb and Sn in plant and different sediment slurries.The equations obtained in the presence of Ir + Mg modifier are presented in Table 6. To compare the slopes of the calibration and standard addition graphs for different slurries in the presence of 5 mg of Ir and 15 mg of Mg, the t-test was applied. For Pb, no significant difference was found between the two slopes in any of the systems studied, including the aqueous Table 3 Experimental variables Levels Pb Sn Variable Coded 2 + 2 + Drying time/s x1 5 40 5 40 Pyrolysis time/s x2 1 20 1 20 Pyrolysis temperature/°C x3 600 1000 600 1200 Atomization temperature/°C x4 1800 2200 2000 2400 Presence of modifier x5 No Yes No Yes Cool-down step x6 No Yes No Yes Volume injected/ml x7 2 10 2 10 Table 4 Fractional factorial design matrix 27–3 and experimental results Run x1 x2 x3 x4 x5 x6 x7 APb sPb ASn sSn 1 2 2 2 2 2 2 2 0.304 0.003 0.034 0.002 2 + 2 2 2 + 2 + 0.369 0.002 0.043 0.001 3 2 + 2 2 + + 2 0.383 0.005 0.016 0.003 4 + + 2 2 2 + + 0.281 0.002 0.076 0.006 5 2 2 + 2 + + + 0.373 0.003 0.052 0.003 6 + 2 + 2 2 + 2 0.295 0.005 0.064 0.004 7 2 + + 2 2 2 + 0.362 0.036 0.023 0.003 8 + + + 2 + 2 2 0.374 0.005 0.020 0.002 9 2 2 2 + 2 + + 0.174 0.001 0.131 0.006 10 + 2 2 + + + 2 0.307 0.002 0.443 0.008 11 2 + 2 + + 2 + 0.251 0.004 0.353 0.001 12 + + 2 + 2 2 2 0.223 0.002 0.144 0.002 13 2 2 + + + 2 2 0.162 0.001 0.500 0.003 14 + 2 + + 2 2 + 0.253 0.005 0.081 0.006 15 2 + + + 2 + 2 0.094 0.003 0.048 0.005 16 + + + + + + + 0.341 0.001 0.602 0.004 Analyst, April 1997, Vol. 122 341solution. However, for Sn in NRCC MESS-1 and BCSS-1 slurries, the standard additions and calibration graphs show significantly different slopes. For this reason the standard additions technique was used for determination of Sn. Table 7 presents compiled data for Pb and Sn determinations in slurried reference material. The recoveries are within the ranges 91.8–101% for Sn and 95.2–102% for Pb in marine sediment and plant slurries. The precisions are within the ranges 2–15% for Sn and 5–13% for Pb in the slurries studied and the t-test indicates the absence of systematic errors.The characteristic masses based on integrated absorbances are of the same order of magnitude (although several times higher) as the manufacturer’s data for simple aqueous solutions: for Sn, 15 pg in PACS-1 and plant slurries and 25 and 30 pg in MESS-1 and BCSS-1 slurries, and for Pb, 18 pg, versus 10 pg and 5.5 pg for Sn and Pb, respectively.The limits of detection (LOD) calculated as LOD = 3s/m, where m is the slope of the calibration graph, were 13 mg kg21 for Pb in PACS-1 and 7 mg kg21 in the other samples and 9 mg kg21 for Sn in each of the analysed standard materials. Conclusion A comparison of the thermal stabilizing efficiencies of W, Mg, Ir + Mg, Ir + Mg + PO4 32 with respect to Pb and Sn in sediment and plant slurries was made.Each of the modifiers studied provides good stabilization for both Pb and Sn. The maximum loss-free pyrolysis temperatures approach 1000–1100 °C and are of the same order for all modifiers. The Ir + Mg modifier is the best for analyte peak isoformation and produces the most symmetrical absorbance profiles. The addition of modifiers contributes to the minimization of matrix effects, which depend on slurry concentration and are more pronounced for sediments.The influences on the absorbance signal of the drying and pyrolysis times, pyrolysis and atomization temperatures, presence of a modifier, incorporation of a ‘cool-down’ step in the program and volume injected were investigated by a 27–3 fractional factorial design. Drying time, pyrolysis and atomization temperatures and presence of a modifier were the most critical parameters. For Pb, lower pyrolysis and atomization temperatures increase the integrated absorbance signals in sediment slurries, whereas for Sn pyrolysis and atomization temperatures and addition of Ir + Mg modifier increase the absorbance signal.The use of Ir + Mg and ultrasonic agitation in the direct introduction of slurries into the graphite atomizer provide a consistent performance, with good sensitivity, precision and analytical recovery. Table 5 Estimated effects Values Values Effect Pb Sn Effect Pb Sn b1 0.021 0.022 b12 + b35 + b67 20.081 0.297 b2 0.005 20.004 b13 + b25 + b47 0.039 20.009 b3 20.038 0.008 b14 + b37 + b56 0.102 0.039 b4 20.072 0.125 b23 + b15 + b46 0.021 0.009 b5 0.036 0.088 b24 + b36 + b57 20.009 20.042 b6 0.042 0.025 b34 + b26 + b17 20.033 0.036 b7 0.016 0.007 b45 + b16 + b27 0.009 0.291 1 0.027 0.031 Fig. 4 Examples of interaction diagrams of confounded terms for Sn: (a) b45 + b16 + b27; and (b) b12 + b35 + b67.Table 7 Analytical results for certified reference materials Found ± Certified*/ RSD Relative Sample Analyte s/mg kg21 mg kg21 (%) error (%) tcalc.(n) ttab. NRCC PACS-1 Sn 41.4 ± 3.6 41.1 ± 3.1 8.7 0.7 0.25 (9) 2.26 Pb 405 ± 24 404 ± 20 5.9 0.2 0.09 (5) 2.57 NRCC MESS-1 Sn 3.6 ± 0.3 3.98 ± 0.44 8.2 28.3 3.12 (2) 4.35 Pb 34.1 ± 2.5 34.0 ± 6.1 7.3 0.2 0.09 (6) 2.45 NRCC BCSS-1 Sn 1.6 ± 0.3 1.85 ± 0.2 10.7 211.4 2.04 (6) 2.45 Pb 23.3 ± 3.4 22.7 ± 3.4 14.6 2.6 0.43 (6) 2.45 NBS 1572 Sn 0.35 ± 0.05 < 0.25† 12.7 — — — Pb 12.4 ± 1.7 13.3 ± 2.4 13.7 26.8 1.06 (4) 2.87 CRM 062 Sn 0.2 ± 0.05 —† 2.4 — — — Pb 25.5 ± 3.5 25.0 ± 1.5 4.3 1.8 0.35 (6) 2.45 * Certified or reference values with confidence interval.† No certified values. Table 6 Calibration and standard additions graph equations Correlation Analyte Sample Standard additions* coefficient Pb† NRCC PACS-1 A = 0.214 + 0.220m 0.998 NRCC MESS-1 A = 0.229 + 0.230m 0.989 NRCC BCSS-1 A = 0.140 + 0.232m 0.991 CRM 062 A = 0.162 + 0.242m 0.969 Aqueous solution A = 20.013 + 0.238m 0.982 Sn NRCC PACS-1 A = 0.283 + 0.314m 0.991 NRCC MESS-1 A = 0.179 + 0.267m 0.986 NRCC BCSS-1 A = 0.112 + 0.175m 0.971 CRM 062 A = 0.164 + 0.321m 0.989 Aqueous solution A = 0.014 + 0.315m 0.995 * A in the integrated absorbance signal and m is the analyte mass in ng.† Ten times more diluted slurries than for Sn. 342 Analyst, April 1997, Vol. 122Financial support to V. I. Slaveykova from the Ministere des Classes Moyennes et de l’Agriculture, Institut de Recherches Chimiques, Tervuren, Belgium, is gratefully acknowledged.References 1 Hoenig, M., and De Kersabiec, A.-M., Spectrochim. Acta, Part B, 1996, 51, 1297. 2 Bendicho, C., and de Loos-Vollebregt, M. T. C., J. Anal. At. Spectrom., 1991, 6, 353. 3 Bendicho, C., and de Loos-Vollebregt, M. T. C., Spectrochim. Acta Part B, 1990, 45, 679. 4 Holcombe, J. A., and Majidi, V., J. Anal. At. Spectrom., 1989, 4, 423. 5 Hoenig, M., and Cilissen, A., Spectrochim. Acta, Part B, 1993, 48B, 1303. 6 Miller-Ihli, N. J., J. Anal. At. Spectrom., 1989, 4, 295. 7 Miller-Ihli, N. J., J. Anal. At. Spectrom., 1994, 9, 1129. 8 Hoenig, M., and Van Hoeyweghen, P., Anal. Chem., 1986, 58, 2614. 9 Littlejohn, D., Stephen, S. C., and Ottaway, J. M., presented at SAC’86/IIIrd BNASS, Bristol, UK, July 1986, Poster PF13. 10 Bermejo-Barrera, P., Aboal-Somoza, M., Soto-Ferreiro, R., and Dominguez-Gonzalez, R., Analyst, 1993, 118, 665. 11 Miller-Ihli, N. J., Spectrochim. Acta, Part B, 1995, 50, 477. 12 Garcia, I. L., Cortez, J. A., and Cordoba, M.H., Anal. Chim. Acta, 1993, 283, 167. 13 Vinas, P., Campillo, N., Garcia, I. L., and Hernandez-Cordoba, M., Talanta, 1995, 42, 527. 14 Halls, D. J., J. Anal. At. Spectrom., 1995, 10, 169. 15 Cabrera, C., Lorenzo, M. L., and Lopez, M. C., J. AOAC Int., 1995, 78, 1061. 16 Hernandez-Cordoba, M., and Lopez-Garcia, I., Talanta, 1991, 38, 1247. 17 Shan, X.-Q., and Wen, B., J. Anal. At. Spectrom., 1995, 10, 791. 18 Bermejo-Barrera, P., Barciela-Alonso, M. C., Moreda-Pineiro, J., Gonzalez-Sixto, C., and Bermejo-Barrera, A., Spectrochim.Acta, Part B, 1996, 51, 1235. 19 Vinas, P., Campillo, N., Garcia, I. L., and Hernandez-Cordoba, M., Fresenius’ J. Anal. Chem., 1994, 349, 306. 20 Hoenig, M., Analusis, 1991, 19, 41. 21 Slaveykova, V. I., and Tsalev, D. L., Anal. Lett., 1990, 23, 1921. 22 Tsalev, D. L., Slaveykova, V. I., and Georgieva, R. B., Anal. Lett., 1996, 29, 73. 23 Tsalev, D. L., Slaveykova, V. I., and Mandjukov, P. B., Spectrochim. Acta Rev., 1990, 13, 225. 24 Tsalev, D. L., D’Ulivo, A., Lampugnani, L., Di Marco, M., and Zamboni, R., J. Anal. At. Spectrom., 1995, 10, 1003. 25 Rademeyer, C. J., Radziuk, B., Romanova, N., Skaugset, N. P., Skogstad, A., and Thomassen, Y., J. Anal. At. Spectrom, 1995, 10, 739. 26 Qiao, H., and Jackson, K., Spectrochim. Acta, Part B, 1991, 46, 1841. 27 Hughes, D. M., Chakrabarti, C. L., Goltz, D. M., Sturgeon, R. E., and Gregoire, D. C., Spectrochim. Acta Part B, 1995, 50, 715. 28 Hughes, D. M., Chaktabarti, C.L., Lamoureux, M. M., Hutton, J. C., Goltz, D. M., Sturgeon, R. E., Gregoire, D. C., and Gilmutdinov, A. K., Spectrochim. Acta, Part B, 1996, 51, 973. 29 Anselmi, A., Tittarelli, P., Biffui, C., and Kmetov, V., Riv. Combust., 1994, 48, 53. 30 Deming, S., and Morgan, S., Experimental Design: a Chemometric Approach, Elsevier, Amsterdam, 1987. 31 Araujo, P. W., Gomez, C. V., Marcano, E., and Benzo, Z., Fresenius’ J. Anal. Chem., 1995, 351, 204. 32 Qiao, H., and Jackson, K., Spectrochim.Acta, Part B, 1992, 47, 1267. Paper 6/07920B Received November 22, 1996 Accepted January 27, 1997 Analyst, April 1997, Vol. 122 343 Electrothermal Atomic Absorption Spectrometric Determination of Lead and Tin in Slurries. Optimization Study V. I. Slaveykova*† and M. Hoenig Ministere des Classes Moyennes et de l’Agriculture, Institut de Recherches Chimiques, Leuvensesteenweg 17, B3080, Tervuren, Belgium A comparative study of the efficiencies of W, Mg, Mg + Pd, Ir + Mg and Ir + W + PO4 32 as chemical modifiers for the thermal stabilization of Pb and Sn in slurries was performed.The Ir + Mg modifier contributes more than the others to minimizing matrix effects and preventing double peak formation. The influences of slurry concentration, amount of modifier and pyrolysis step on the integrated absorbance signals for Pb and Sn in sediment slurry were studied. The potential of fractional factorial design was explored to evaluate the effect on the absorbance signals of different factors such as drying and pyrolysis times, pyrolysis and atomization temperatures, presence of modifier and incorporation of a ‘cool-down’ step in the electrothermal program.The Pb and Sn integrated absorbance signals are increased by lower and higher pyrolysis and atomization temperatures, respectively. The Ir + Mg modifier also increases the absorbance signal for Sn. The use of an Ir + Mg modifier and ultrasonic agitation in the direct introduction of slurries into the graphite atomizer provides a consistent performance with good sensitivity, precision and analytical recovery.Keywords: Electrothermal atomic absorption spectrometry; slurry; iridium + magnesium modifier; fractional factorial design; lead; tin The methodology of slurry sampling (SS) ETAAS is now well established and generally used in the determination of trace elements in environmental samples, particularly when there is a real need to analyse microsamples.1,2 The most attractive advantages of slurry analysis over acid digestion or fusion procedures and direct solid sampling can be briefly summarized as follows: (i) reduced sample pretreatment and increase in the speed of the whole analytical procedure; (ii) low contamination risk; (iii) fewer possibilities of analyte losses during the sample pre-treatment; (iv) use of conventional sample introduction systems without instrumental modification; (v) adequate calibration with aqueous standards; and (vi) good precision and accuracy.Some problems and potential drawbacks of SS-ETAAS worth addressing are the high background levels, lack of slurry homogeneity, irreproducible sample introduction and inhomogeneous distribution of the analytes in samples.3 To ensure good repeatability of the measurements, a representative number of particles must be analyzed. It has been shown that only a very fine granulometry of the slurry ensures correct results.Thus, the grinding efficiency is one of the critical factors in the analysis. For slurry analysis of soils and sediments, where trace elements are associated with the very fine granulometric fraction only, the grinding efficiency is less important than for the other type of samples, where the analytes are distributed more homogeneously (e.g., plants, glass materials).1–4 The errors associated with slurry analysis can be minimized by working with small particle sizes, larger total masses of material and narrower particle size distributions.4 Homogeneous slurries can be obtained by using ultrasonic and magnetic stirring, vortex or gas mixing of the slurries5–7 and the addition of various stabilizing agents such as glycerol,8 Viscalex9 or Triton X-100.10 Methods based on modern furnace technology, e.g., platforms, appropriate modifiers, fast absorbance recording and efficient background correction, were usually applied to slurry analysis, but matrix interferences were not avoided completely. 11 Two approaches to slurry analysis by ETAAS has been described.2 The first includes elimination of the pyrolysis step and injection of the sample into the pre-heated furnace in order to decrease the analytical time and increase the sample throughput.12–14 Hoenig and Cilissen5 demonstrated that the precisions reached with conventional and fast furnace programs were comparable, but also stressed the importance of peak area measurements to obtain correct results.The second approach is based on conventional programs and chemical modification techniques. Most often a Pd + Mg modifier has been used to determine, for example, Pb in mussels, fruit, coal fly ash and sediment slurries,10,15–17 and Sn in sediments and coal fly ash,17,18 while NH4H2PO4 has been used for Pb in paprika16 and food.19 The aim of this study was to improve the analytical performance of the ETAAS determination of Pb and Sn in slurries.The capabilities of fractional factorial design were explored to study the influence of various parameters, such as drying and pyrolysis times, pyrolysis and atomization temperatures, presence of modifier, incorporation of a ‘cool-down’ step in the electrothermal program and the volume of the samples introduced, on the absorbance signal. The efforts were aimed at using the advantages provided by both a chemical modifier and a fast temperature program. The stabilizing actions of different modifiers such as Mg, W, Mg + Pd, Mg + Ir and Ir + W + PO4 32 were compared for Pb and Sn in marine sediments and plants slurries.Special attention was paid to a detailed investigation not only of the pyrolysis stage but also of the peak shape and modifier potential for peak isoformation in slurries and aqueous solutions. A comparison between fast and conventional furnace programs in the absence and the presence of a modifier was also made. Experimental Apparatus Measurements were carried out using a Varian (Palo Alto, CA, USA) SpectrAA 400 Zeeman-effect atomic absorption spectrometer, equipped with a GTA 96 graphite atomizer and a † On leave from the Faculty of Chemistry, University of Sofia, 1 James Bourchier Blvd., Sofia 1126, Bulgaria.Analyst, April 1997, Vol. 122 (337–343) 337programmable sample dispenser. A Hamamatsu (Phonotonics K.K., Shimokanzo, Japan) hollow cathode lamp for Pb and a Juniper (Harlow, Essex, UK) hollow cathode lamp for Sn were used.The instrument parameters for the Varian SpectrAA 400 Zeeman-effect atomic absorption spectrometer were adjusted according to the manufacturer’s recommendations. Pyrolytic graphite coated tubes with specially design fork platforms, RW 0961/3 PyC N 299990/89, Series V/64/01701 (Ringsdorff, Bonn, Germany), were used. A Vibra Cell VC 50 high intensity ultrasonic processor (Sonics and Materials, Danbury, USA) was utilized in the preparation of the slurries.The optimized heating program for the GTA 96 graphite tube atomizer is given in Table 1. A ‘cool-down’ step at 100 °C was incorporated in the program in order to normalize the atomization conditions for experiments at different pre-treatment temperatures. After each heating cycle the tube was cleaned up using the maximum permissible temperature for the tube. Reagents Stock standard solutions of Pb (1 g l21 PbII in 2% HNO3) and Sn (1 g l21 SnII in 20% HCl) were obtained from Servi-Lab (Wakken, Belgium).Working standard solutions containing 200 mg l21 of the analytes were prepared daily from the stock standard solutions by dilution with de-mineralized water and acidified with HNO3 (0.1% final concentration). Chemical modifier solutions, each 1% m/V, were prepared using the following reagents: (NH4)2IrCl6, (NH4)2PdCl4 and WO3 (Specpur; Johnson-Matthey, Royston, Hertfordshire, UK) and Mg(NO3)2 (analytical-reagent grade; Merck, Darmstadt, Germany).The Pd, Ir and Mg modifier solutions were prepared as described by Hoenig20 and W modifier solution by dissolution of WO3 in NH3 according to Slaveyhova et al.21 The modifier solutions were mixed in different proportions and the composition of the modifiers used is pointed out in the text for each of the systems studied. Certified Reference Materials Plant and sediment materials with certified or reference contents were obtained from the National Research Council of Canada (Marine Sediments MESS-1, BCSS-1 and PACS-1), the Community Bureau of Reference (BCR, now SMT, Geel, Belgium) (RM 62, Olea europaea) and the National Institute for Standards and Technology (NIST) (Gaithersburg, MD, USA) (SRM 1572, Citrus Leaves).Sample Pre-treatment Procedure Plant and sediment materials were ground in a boron carbide mortar to reduce the particle size to < 50 mm. Amounts of 3–10 mg of finely powdered solid samples were weighed directly into Varian autosampler microvials and treated with 50 ml of concentrated HNO3 overnight, before addition of 950 ml of demineralized water.This ensured that the sample was partly digested and the analytes were mobilized into solution. A portion of the trace elements present remained in the solid phase, which was dispensed together with the enriched solution into the graphite tube during sampling. Just before each sampling, the slurry was homogenized with a high intensity ultrasonic probe. The autosampler was programmed to aspirate 10 ml of each slurried sample and 5 ml of the modifier solution successively, which were mixed in situ (sampler automixing mode) during the injection on to the platform.The hot injection mode of the autosampler at 120 °C was used. Thermal Pre-treatment Study The optimum pyrolysis temperatures were determined at fixed atomization temperatures by their systematic variation and measurements of the integrated absorbance. The possibility of a sensitivity drift effect was avoided by randomizing measurements and using the mean of two or three values at each temperature.Normalized absorbance measurements were used, assigning a value of 1 to the plateau absorbance for the species. When there was no pronounced plateau on the pyrolysis curves, the maximum integrated absorbance signal was used to normalize the signal. Results and Discussion Comparison of Various Chemical Modifiers Efforts were directed to using the advantages provided by W, Mg, Ir + Mg and Ir + W + PO4 32 modifiers and to compare their influence on the atomization signal and thermal pre-treatment losses of Pb and Sn in sediment and plant slurries.The Wcontaining modifiers are very effective thermal stabilizers for Pb and Sn in aqueous solutions and biological samples21,22 and Mg is both a well recognized modifier in ETAAS and an efficient ashing aid (for a review, see ref. 23). Iridium was chosen as a main component of the mixed modifiers because of its pronounced stabilizing efficiency in the determination of volatile analytes.20,24,25 It is expected that addition of Mg and W to an Ir solution will further stabilize Ir to higher pyrolysis temperatures.Tsalev et al.24 observed volatilization losses of Ir at temperatures > 2050–2100 °C in WC-coated platforms. The presence of Mg and W would also lead to a more homogeneous distribution of Ir on the platform surface.23,26 Thus, migration of metal droplets at high temperature and possible coalescence resulting in larger modifier droplets would be minimized.Therefore, certain effects of overstabilization24,25 and incomplete analyte vaporization from Ir during atomization could be avoided. The performance of each chemical modifier was established by considering maximum loss-free pyrolysis temperatures, characteristic masses and impact of the peak shape. The minimum possible amounts of modifiers providing stabilization to at least 1000 °C were used in order to avoid rapid deterioration of the furnace surface and drift in the sensitivity.The possible condensation of chemical modifiers on the cooler end of the tube and light scattering by microparticles and therefore a high background level could also be reduced by using the minimum possible amount of modifier.27,28 The maximum pyrolysis temperatures for Pb and Sn in marine sediments and plant slurries in the presence of mixed Ir + Mg and Ir + W + PO4 32 modifiers, in addition to the single components, are given in Table 2.A comparison with Mg + Pd modifier also is made. Further details and specific features of Table 1 Temperature program for GTA 96 graphite tube atomizer Ar flow rate/ Step Temperature/°C Time/s l min21 Read 1 300 30 3.0 No 2 Var.* 5 3.0 No 3 Var.* 10 3.0 No 4 100 5 3.0 No 5 100 1 0 No 6 2000† 0 0 Yes 7 2000† 5 0 Yes 8 120 10 3.0 No * Temperatures varied within the ranges given in Figs. 1 and 2. † 2500 °C for Sn. 338 Analyst, April 1997, Vol. 122the pyrolysis curves are presented below. It is evident that maximum loss-free pyrolysis temperatures in slurries are much lower than those in aqueous solutions for all of the modifiers studied. The modifiers applied have a substantial stabilizing effect for Pb in slurries, but for Sn the increase in maximum pyrolysis temperatures is only 100–200 °C. The maximum pyrolysis temperatures obtained in marine sediment are higher for Pb and lower for Sn than those reported in the presence of Mg + Pd.18 Typical examples of thermal pre-treatment curves for Pb and Sn in marine sediment and Olea europaea slurries are presented in Figs. 1 and 2, respectively. In the absence of modifier, the integrated absorbance signal for Pb in plant slurries gradually decreases with increasing pyrolysis temperature, whereas plant matrix components stabilize Sn, for which a maximum pyrolysis temperature of 800 °C was observed.The addition of a modifier strongly stabilizes Pb in plants and moderately stabilizes Sn. The stabilizing action of the modifiers decreases in the series Mg Å Mg + Pd < Ir + Mg < Ir + W + PO4 32 for Pb and Mg Å Mg + Pd < Ir + W + PO4 32 < Ir + Mg for Sn. Moreover, both Pb and Sn are stabilized to almost identical temperatures in the presence of the same modifier. The main components of marine sediments, Si, Al, Fe and K, could act as thermal stabilizers, thus the maximum loss-free temperature for Pb and Sn in slurries is 800 and 900 °C, respectively.At low pyrolysis temperatures, losses in the integrated absorbance signal were observed for both Pb and Sn in sediments. The diminution of the signal at low pre-treatment temperatures could be explained by redistribution of the sample in the atomizer and as a result the analyte is less effectively atomized in the cooler region towards the end of the tube. The vaporization of volatile Pb and Sn compounds at low temperature and their transformation into less volatile form during thermal pre-treatment could also be taken into account to explain the effect observed.The formation of Cl-, O- and Scontaining compounds, in competition with free atom formation, was observed when slurries were vaporized and Pb and Sn gas phase molecule species were studied.29 However, signal losses for Pb and Sn in sediment slurries were not observed at low temperatures in the presence of any of the modifiers studied.Moreover, the stabilizing efficiency increases in the order Mg Å W Å Mg + Pd < Ir + Mg < Ir + W + PO4 32 for Pb. The maximum pyrolysis temperatures for Sn are similar for all modifiers studied, which is consistent with the decisive role of the matrix component on the behaviour of Sn in the graphite atomizer. The peak area/peak height ratio was used as an indicator for peak shape in terms of whether sharp and narrow or broad and wide profiles were observed.A peak area/peak height ratio near 1 s is the optimum from the viewpoint of the determination of low concentrations near the LOD. The optimum peak should be sharp and narrow to be distinguishable from the noise and with a sufficiently high peak area (peak area measurements). In the absence of a modifier and in the presence of Mg and W, the absorbance profiles were relatively fast and narrow with peak area/peak height ratios << 1 s for both Pb and Sn in sediment and plant slurries.The addition of a platinum group metal (Ir or Pd) as a modifier component produced a shift in the peak maximum and signal broadening. The addition of Ir + Mg produced the most symmetrical peaks and the peak area/peak height ratios were near 1 s for Pb in all samples and for Sn in sediments. In plant slurry a peak area/peak height ratio close to 1 s was also observed for Sn in plants in the presence of Ir + W + PO4 32 and for Pb in the presence of Pd + Mg.The existence of double absorbance peaks and also the modifier capability to ‘level off’ an absorbance signal in samples with different matrix components were also studied. In contrast to Bermejo-Barrera et al.’s results,18 double peak formation was not observed for Pb under our conditions and the background signal was almost negligible. Modification with Ir + Mg resulted in the best signal isoformation and minimum background absorbance in both sediment and plant slurries for Sn. A certain overcompensation of the signal and/or the appearence of double peaks were observed for Sn in sediments modified with Pd + Mg or Ir + W + PO4 32 and, moreover, this effect was independent of pyrolysis and atomization temperature variations.A possible cause of double peak formation is the existence of two different forms of an analyte in sediment.25 Also, Sn leached into the nitric acid solution may behave differently from the Sn still in the sediment. Amount of Chemical Modifier The amounts of chemical modifiers were varied between 1 and 50 mg for W, Mg and Ir.Higher amounts led to an increase in the maximum pyrolysis temperatures, which approach 1100, 1100 and 1200 °C for Pb and 1400, 1400 and 1500 °C for Sn in sediments. With more than 20 mg of Mg or W, enhancements of the background absorbance were observed. The addition of more than 10 mg of Ir led to overstabilization and atomization temperatures above 2600–2700 °C were required for the effective release of analyte atoms.The peaks were very broad and real time integration was impossible. The effect of the ratio of the two modifier components Ir and Mg on the maximum pre-treatment temperatures and absorbance signals for Pb and Sn in sediments was studied. An increase of the Ir : Mg ratio led to higher stabilization temperatures for both elements but affected only slightly the absorbance peak characteristics. Influence of Slurry Concentration The influence of different slurry concentrations on the analytical signal, sensitivity and precision was investigated.The slurry concentration for Sn determination was varied between 1 Table 2 Pyrolysis temperatures for Pb and Sn in sediment and plant slurries in the presence of various modifiers Pb Sn Citrus Olea Citrus Olea Aqueous Sediment leaves europaea Aqueous Sediment leaves europaea Modifier solution slurry slurry slurry solution slurry slurry slurry None 600* 800† 300* 200* 400* 700† 600* 700 15 mg Mg 900 1100 1000 1000 1400 900 1000 900 15 mg W 1000 1000 1000 1100 1500 1000 1000 1000 5 mg Ir + 15 mg Mg 1200 1100 1000 1000 1500 1000 1000 1000 5 mg Ir + 10 mg W + 25 mg PO4 32 1200 1200 1200 1200 1450 1100 1100 1000 10 mg Pd + 15 mg Mg 1200 1100 1000 1000 1400 1000 1000 900 * The signal diminished gradually with increase in temperature.†A signal decrease at low temperature was observed. Analyst, April 1997, Vol. 122 339ratio is > 1 with Ir + Mg and Mg + Pd and < 1 with the other modifiers studied.Optimization by Fractional Factorial Design To study the influence of different factors on the atomization of Pb and Sn in marine sediment slurries, the potential of fractional factorial design was explored. The fractional factorial design at two levels, 27–3, was applied to determine the main effects bi and some of the two-factor interactions bij.30 The variables considered and levels studied are given in Table 3.For x5, minus and plus levels correspond to absence of a modifier and the presence of Ir + Mg modifier, respectively. The plus level for x6 refers to the incorporation of a ‘cool-down’ step at 100 °C in the program. The Ir + Mg modifier was chosen in the optimization study because of its pronounced stabilizing efficiency and the relative isoformation of the analytical signal in slurries and aqueous solutions. The volume of sample injected is also believed to influence the reproducibility of the analysis.Droplets of different size are spread over the inner surface of the platform and different sample distributions on the substrate could be expected. On the other hand, the amount of the analyte in the graphite furnace is the same for different volumes of the sample injected, because the slurry concentration is different. Hence the duration of the drying stage might play an important role in the slurry particle distribution on the platform and also on the rate of nucleation and crystal growth of the part of the sample leached in the HNO3 and thus influence the reproducibility. Table 4 shows the 27–3 design.Factors x5, x6 and x7 are obtained by multiplying columns x1x2x3, x2x3x4, and x1x3x4, respectively. A similar approach was used in the determination of Cd by ETAAS for different atomization systems.31 The statistical evaluation reveals that the main effects b3–b6 for Pb and b4–b6 for Sn are significant.They correspond to pyrolysis and atomization temperatures, the presence of a modifier and a ‘cool-down’ step. These statistical estimates suggest that an increase in the drying time leads to a moderately increased absorbance signal and better reproducibility for both Pb and Sn in a slurry. Lower atomization and pyrolysis temperatures for Pb also give higher atomization signals, whereas the Sn absorbance increases significantly with increase in the pyrolysis and atomization temperatures (see Table 5).The differences observed between Pb and Sn as regards the influence of pyrolysis and atomization temperatures on the atomization signal are probably due to different distributions of the analytes in the sediment.31 Moreover, the addition of HNO3 during sample preparation produces about 80% extraction of the lead from the particles into the solution and only 40% for tin. Thus a higher atomization temperature was required to insure release of Sn from slurry particles.The addition of a chemical modifier also is of prime importance for Sn but not for Pb determination. This could be related to differences in the efficiency of the sediment matrix as a thermal stabilizer by itself. The confounded terms formed by the two-factor interactions b12 + b35 + b67, b13 + b25 + b47, b14 + b37 + b56, b23 + b15 + b46 and b34 + b26 + b17 for Pb and Sn and b45 + b16 + b27 for Sn are significant at the 95% probability level.The interaction diagrams were used to resolve these interaction terms and examples are represented in Fig. 4. Each box shows the combined effect of the confounded terms. The value presented in each quadrant is the average signal obtained in each case. Among the confounded terms, only for Pb do the interaction terms b12 and b14 show a tendency to improve the signal slightly when shorter drying times and lower atomization temperatures are used. For Sn there is a pronounced tendency for signal enhancement when a modifier is added to the sample and the atomization temperature increases.Figures of Merit The calibration graph and standard additions methods were compared for both Pb and Sn in plant and different sediment slurries. The equations obtained in the presence of Ir + Mg modifier are presented in Table 6. To compare the slopes of the calibration and standard addition graphs for different slurries in the presence of 5 mg of Ir and 15 mg of Mg, the t-test was applied.For Pb, no significant difference was found between the two slopes in any of the systems studied, including the aqueous Table 3 Experimental variables Levels Pb Sn Variable Coded 2 + 2 + Drying time/s x1 5 40 5 40 Pyrolysis time/s x2 1 20 1 20 Pyrolysis temperature/°C x3 600 1000 600 1200 Atomization temperature/°C x4 1800 2200 2000 2400 Presence of modifier x5 No Yes No Yes Cool-down step x6 No Yes No Yes Volume injected/ml x7 2 10 2 10 Table 4 Fractional factorial design matrix 27–3 and experimental results Run x1 x2 x3 x4 x5 x6 x7 APb sPb ASn sSn 1 2 2 2 2 2 2 2 0.304 0.003 0.034 0.002 2 + 2 2 2 + 2 + 0.369 0.002 0.043 0.001 3 2 + 2 2 + + 2 0.383 0.005 0.016 0.003 4 + + 2 2 2 + + 0.281 0.002 0.076 0.006 5 2 2 + 2 + + + 0.373 0.003 0.052 0.003 6 + 2 + 2 2 + 2 0.295 0.005 0.064 0.004 7 2 + + 2 2 2 + 0.362 0.036 0.023 0.003 8 + + + 2 + 2 2 0.374 0.005 0.020 0.002 9 2 2 2 + 2 + + 0.174 0.001 0.131 0.006 10 + 2 2 + + + 2 0.307 0.002 0.443 0.008 11 2 + 2 + + 2 + 0.251 0.004 0.353 0.001 12 + + 2 + 2 2 2 0.223 0.002 0.144 0.002 13 2 2 + + + 2 2 0.162 0.001 0.500 0.003 14 + 2 + + 2 2 + 0.253 0.005 0.081 0.006 15 2 + + + 2 + 2 0.094 0.003 0.048 0.005 16 + + + + + + + 0.341 0.001 0.602 0.004 Analyst, April 1997, Vol. 122 341solution.However, for Sn in NRCC MESS-1 and BCSS-1 slurries, the standard additions and calibration graphs show significantly different slopes. For this reason the standard additions technique was used for determination of Sn.Table 7 presents compiled data for Pb and Sn determinations in slurried reference material. The recoveries are within the ranges 91.8–101% for Sn and 95.2–102% for Pb in marine sediment and plant slurries. The precisions are within the ranges 2–15% for Sn and 5–13% for Pb in the slurries studied and the t-test indicates the absence of systematic errors. The characteristic masses based on integrated absorbances are of the same order of magnitude (although several times higher) as the manufacturer’s data for simple aqueous solutions: for Sn, 15 pg in PACS-1 and plant slurries and 25 and 30 pg in MESS-1 and BCSS-1 slurries, and for Pb, 18 pg, versus 10 pg and 5.5 pg for Sn and Pb, respectively. The limits of detection (LOD) calculated as LOD = 3s/m, where m is the slope of the calibration graph, were 13 mg kg21 for Pb in PACS-1 and 7 mg kg21 in the other samples and 9 mg kg21 for Sn in each of the analysed standard materials. Conclusion A comparison of the thermal stabilizing efficiencies of W, Mg, Ir + Mg, Ir + Mg + PO4 32 with respect to Pb and Sn in sediment and plant slurries was made.Each of the modifiers studied provides good stabilization for both Pb and Sn. The maximum loss-free pyrolysis temperatures approach 1000–1100 °C and are of the same order for all modifiers. The Ir + Mg modifier is the best for analyte peak isoformation and produces the most symmetrical absorbance profiles.The addition of modifiers contributes to the minimization of matrix effects, which depend on slurry concentration and are more pronounced for sediments. The influences on the absorbance signal of the drying and pyrolysis times, pyrolysis and atomization temperatures, presence of a modifier, incorporation of a ‘cool-down’ step in the program and volume injected were investigated by a 27–3 fractional factorial design. Drying time, pyrolysis and atomization temperatures and presence of a modifier were the most critical parameters. For Pb, lower pyrolysis and atomization temperatures increase the integrated absorbance signals in sediment slurries, whereas for Sn pyrolysis and atomization temperatures and addition of Ir + Mg modifier increase the absorbance signal.The use of Ir + Mg and ultrasonic agitation in the direct introduction of slurries into the graphite atomizer provide a consistent performance, with good sensitivity, precision and analytical recovery.Table 5 Estimated effects Values Values Effect Pb Sn Effect Pb Sn b1 0.021 0.022 b12 + b35 + b67 20.081 0.297 b2 0.005 20.004 b13 + b25 + b47 0.039 20.009 b3 20.038 0.008 b14 + b37 + b56 0.102 0.039 b4 20.072 0.125 b23 + b15 + b46 0.021 0.009 b5 0.036 0.088 b24 + b36 + b57 20.009 20.042 b6 0.042 0.025 b34 + b26 + b17 20.033 0.036 b7 0.016 0.007 b45 + b16 + b27 0.009 0.291 1 0.027 0.031 Fig. 4 Examples of interaction diagrams of confounded terms for Sn: (a) b45 + b16 + b27; and (b) b12 + b35 + b67.Table 7 Analytical results for certified reference materials Found ± Certified*/ RSD Relative Sample Analyte s/mg kg21 mg kg21 (%) error (%) tcalc. (n) ttab. NRCC PACS-1 Sn 41.4 ± 3.6 41.1 ± 3.1 8.7 0.7 0.25 (9) 2.26 Pb 405 ± 24 404 ± 20 5.9 0.2 0.09 (5) 2.57 NRCC MESS-1 Sn 3.6 ± 0.3 3.98 ± 0.44 8.2 28.3 3.12 (2) 4.35 Pb 34.1 ± 2.5 34.0 ± 6.1 7.3 0.2 0.09 (6) 2.45 NRCC BCSS-1 Sn 1.6 ± 0.3 1.85 ± 0.2 10.7 211.4 2.04 (6) 2.45 Pb 23.3 ± 3.4 22.7 ± 3.4 14.6 2.6 0.43 (6) 2.45 NBS 1572 Sn 0.35 ± 0.05 < 0.25† 12.7 — — — Pb 12.4 ± 1.7 13.3 ± 2.4 13.7 26.8 1.06 (4) 2.87 CRM 062 Sn 0.2 ± 0.05 —† 2.4 — — — Pb 25.5 ± 3.5 25.0 ± 1.5 4.3 1.8 0.35 (6) 2.45 * Certified or reference values with confidence interval.† No certified values. Table 6 Calibration and standard additions graph equations Correlation Analyte Sample Standard additions* coefficient Pb† NRCC PACS-1 A = 0.214 + 0.220m 0.998 NRCC MESS-1 A = 0.229 + 0.230m 0.989 NRCC BCSS-1 A = 0.140 + 0.232m 0.991 CRM 062 A = 0.162 + 0.242m 0.969 Aqueous solution A = 20.013 + 0.238m 0.982 Sn NRCC PACS-1 A = 0.283 + 0.314m 0.991 NRCC MESS-1 A = 0.179 + 0.267m 0.986 NRCC BCSS-1 A = 0.112 + 0.175m 0.971 CRM 062 A = 0.164 + 0.321m 0.989 Aqueous solution A = 0.014 + 0.315m 0.995 * A in the integrated absorbance signal and m is the analyte mass in ng.† Ten times more diluted slurries than for Sn. 342 Analyst, April 1997, Vol. 122Financial support to V. I. Slaveykova from the Ministere des Classes Moyennes et de l’Agriculture, Institut de Recherches Chimiques, Tervuren, Belgium, is gratefully acknowledged. References 1 Hoenig, M., and De Kersabiec, A.-M., Spectrochim. Acta, Part B, 1996, 51, 1297. 2 Bendicho, C., and de Loos-Vollebregt, M. T. C., J. Anal. At. Spectrom., 1991, 6, 353. 3 Bendicho, C., and de Loos-Vollebregt, M. T. C., Spectrochim. Acta Part B, 1990, 45, 679. 4 Holcombe, J. A., and Majidi, V., J. Anal. At. Spectrom., 1989, 4, 423. 5 Hoenig, M., and Cilissen, A., Spectrochim. Acta, Part B, 1993, 48B, 1303. 6 Miller-Ihli, N. J., J. Anal. At. Spectrom., 1989, 4, 295. 7 Miller-Ihli, N. J., J. Anal. At. Spectrom., 1994, 9, 1129. 8 Hoenig, M., and Van Hoeyweghen, P., Anal. Chem., 1986, 58, 2614. 9 Littlejohn, D., Stephen, S. C., and Ottaway, J. M., presented at SAC’86/IIIrd BNASS, Bristol, UK, July 1986, Poster PF13. 10 Bermejo-Barrera, P., Aboal-Somoza, M., Soto-Ferreiro, R., and Dominguez-Gonzalez, R., Analyst, 1993, 118, 665. 11 Miller-Ihli, N. J., Spectrochim. Acta, Part B, 1995, 50, 477. 12 Garcia, I. L., Cortez, J. A., and Cordoba, M. H., Anal. Chim. Acta, 1993, 283, 167. 13 Vinas, P., Campillo, N., Garcia, I. L., and Hernandez-Cordoba, M., Talanta, 1995, 42, 527. 14 Halls, D. J., J. Anal. At. Spectrom., 1995, 10, 169. 15 Cabrera, C., Lorenzo, M. L., and Lopez, M. C., J. AOAC Int., 1995, 78, 1061. 16 Hernandez-Cordoba, M., and Lopez-Garcia, I., Talanta, 1991, 38, 1247. 17 Shan, X.-Q., and Wen, B., J. Anal. At. Spectrom., 1995, 10, 791. 18 Bermejo-Barrera, P., Barciela-Alonso, M. C., Moreda-Pineiro, J., Gonzalez-Sixto, C., and Bermejo-Barrera, A., Spectrochim. Acta, Part B, 1996, 51, 1235. 19 Vinas, P., Campillo, N., Garcia, I. L., and Hernandez-Cordoba, M., Fresenius’ J. Anal. Chem., 1994, 349, 306. 20 Hoenig, M., Analusis, 1991, 19, 41. 21 Slaveykova, V. I., and Tsalev, D. L., Anal. Lett., 1990, 23, 1921. 22 Tsalev, D. L., Slaveykova, V. I., and Georgieva, R. B., Anal. Lett., 1996, 29, 73. 23 Tsalev, D. L., Slaveykova, V. I., and Mandjukov, P. B., Spectrochim. Acta Rev., 1990, 13, 225. 24 Tsalev, D. L., D’Ulivo, A., Lampugnani, L., Di Marco, M., and Zamboni, R., J. Anal. At. Spectrom., 1995, 10, 1003. 25 Rademeyer, C. J., Radziuk, B., Romanova, N., Skaugset, N. P., Skogstad, A., and Thomassen, Y., J. Anal. At. Spectrom, 1995, 10, 739. 26 Qiao, H., and Jackson, K., Spectrochim. Acta, Part B, 1991, 46, 1841. 27 Hughes, D. M., Chakrabarti, C. L., Goltz, D. M., Sturgeon, R. E., and Gregoire, D. C., Spectrochim. Acta Part B, 1995, 50, 715. 28 Hughes, D. M., Chaktabarti, C. L., Lamoureux, M. M., Hutton, J. C., Goltz, D. M., Sturgeon, R. E., Gregoire, D. C., and Gilmutdinov, A. K., Spectrochim. Acta, Part B, 1996, 51, 973. 29 Anselmi, A., Tittarelli, P., Biffui, C., and Kmetov, V., Riv. Combust., 1994, 48, 53. 30 Deming, S., and Morgan, S., Experimental Design: a Chemometric Approach, Elsevier, Amsterdam, 1987. 31 Araujo, P. W., Gomez, C. V., Marcano, E., and Benzo, Z., Fresenius’ J. Anal. Chem., 1995, 351, 204. 32 Qiao, H., and Jackson, K., Spectrochim. Acta, Part B, 1992, 47, 1267. Paper 6/07920B Received November 22, 1996 Accepted January 27, 1997 Analyst, April 1997, Vol. 122 343
ISSN:0003-2654
DOI:10.1039/a607920b
出版商:RSC
年代:1997
数据来源: RSC
|
10. |
Flow Injection Spectrophotometric Determination of L-Dopa andCarbidopa in Pharmaceutical Formulations Using a Crude Extract of SweetPotato Root [Ipomoea batatas (L.) Lam.] as EnzymaticSource |
|
Analyst,
Volume 122,
Issue 4,
1997,
Page 345-350
Orlando Fatibello-Filho,
Preview
|
|
摘要:
Flow Injection Spectrophotometric Determination of L-Dopa and Carbidopa in Pharmaceutical Formulations Using a Crude Extract of Sweet Potato Root [ Ipomoea batatas (L.) Lam.] as Enzymatic Source Orlando Fatibello-Filho* and Iolanda da Cruz Vieira Departamento de Qu�ýmica, Grupo de Qu�ýmica Anal�ýtica, Centro de Ci�encias Exatas e de Tecnologia, Universidade Federal de S�ao Carlos, Caixa Postal 676, CEP 13.560-970, S�ao Carlos, SP, Brazil A flow injection (FI) spectrophotometric method is proposed for the determination of L-dopa and carbidopa in pharmaceutical formulations.After selection of the extraction medium (e.g., buffer-to-tissue ratio, pH, buffer concentration, protective agents and/or stabilizers) and storage conditions, crude extract of sweet potato root [Ipomoea batatas (L.) Lam.] was used as an enzymatic source of polyphenol oxidase (Tyrosinase; catechol oxidase; EC.1.14.18.1) directly in the carrier. This enzyme catalyses the oxidation of these catecholamines to the corresponding dopaquinone. Further, dopaquinone undergoes a rapid spontaneous auto-oxidation to leucodopachrome, which is in turn oxidized to dopachrome; this last compound has a strong absorption at 480 and 360 nm for L-dopa and carbidopa, respectively. For the optimum extraction conditions found the enzyme activity of the crude extract did not vary for at least 5 months when stored at 4 °C and decreased by only 4–5% during an 8 h working period at 25 °C.The results obtained for L-dopa and carbidopa by the proposed enzymatic FI method were in close agreement with the label values (r1 = 0.9699 and r2 = 0.9999) and also with those obtained using a pharmacopeial method (r3 = 0.9675). The throughput was 26 samples h21, and 2.30 ml of crude extract were consumed in each determination, corresponding to only 72 mg of the original sweet potato root. The detection limit (three times the signal blank/slope) was 1.5 31025 and 2.0 3 1025 mol l21 for L-dopa and carbidopa, respectively; the recovery of L-dopa and carbidopa from three samples ranged from 98.6 to 106.3% of the added amount.Keywords: L-Dopa; carbidopa; flow system; polyphenol oxidase; tyrosinase; catechol oxidase; sweet potato root [Ipomoea batatas (L.) Lam.]; pharmaceutical formulations l-Dopa [(2)-3-(3,4-dihydroxyphenyl)-l-alanine] and carbidopa [(2)-l-2-(3,4-dihydroxybenzyl)-2-hydrazinopropionic acid] are catecholamines with an alkylamine chain attached to a benzene ring bearing two hydroxyl groups.l-Dopa is an important neurotransmitter that is used for the treatment of neural disorders such as Parkinson’s disease. It is effective in relieving hypokinesia and can also decrease rigidity, oculogyric crises and tremor.1 After its oral administration, l-dopa is absorbed through the bowel at the level of the small intestine and metabolized by a decarboxylation process to dopamine and then to other metabolites. Both carbidopa and benserazide are used as inhibitors for the decarboxylase activity.2 Hence, the development of a method for the selective determination of ldopa and carbidopa is important, since they are frequently found together in some pharmaceutical formulations.Several methods have been proposed for the determination of these catecholamine drugs in biological specimens and/or pharmaceutical formulations. The determination of catecholamines in biological specimens normally requires the use of trace analysis techniques, mainly chromatography with fluorimetric or electrochemical detection.3 On the other hand, catecholamines are present in relatively large amounts in pharmaceutical formulations and much effort has been devoted to the development of simple, rapid, accurate and precise analytical methods.HPLC has been widely used for the determination of l-dopa in brain, plasma, urine, liver, serum, tissue and biological fluids.4–8 A photokinetic method based on the strong inhibitory effect of the catecholamines on the photochemical reaction between Rose Bengal and EDTA has been proposed.9 A fluorimetric method10 based on the inhibition by catechlolamines of the photoreduction of phloxin (tetrachlorotetrabromofluorescein) by EDTA has also been proposed.A flow injection (FI) determination of epinephrine, norepinephrine, dopamine and l-dopa by their chemiluminogenic oxidation with potassium permanganate in acidic medium in the presence of formaldehyde has been reported.11 Spectrophotometric methods have also been proposed for the determination of catecholamines in pharmaceutical formulations based on the direct measurement of l-dopa at 280 and 290 nm12 and by first-derivative spectrophotometry at 276 nm.13 With other UV spectrophotometric methods, this compound has been determined in the presence of germanium dioxide,14 boric acid,15 Na2B4O7 16 and Na2HPO4 17 at 292, 239.5, 287 and 292.5 nm, respectively.In addition, catecholamines have been determined in the visible region after reaction with metaperiodate,18 isonicotinic acid hydrazide in alkaline medium,19 FeIII and o-phenanthroline in a moderately acidic medium,20 PdCl2,21 molybdophosphoric acid,22 ninhydrin,23 ammonium metavanadate24and p-aminophenol in the presence of KIO4 after its oxidation in alkaline medium.25 Nevertheless, there are few enzymatic methods for determining catecholamines described in the literature26–30 and of these only one laborious manual spectrophotometric method30 has been proposed.l-Dopa was incubated with mammalian tyrosinase in the presence of an optimum concentration of ZnII ions for 50–60 min and the melanochrome formed in this reaction was monitored at 540 nm. Uchiyama and co-workers31,32 used a cucumber juice solution (crude extract) containing ascorbate oxidase as a carrier in an FI system for the determination of l-ascorbic acid31 and dehydroascorbic acid.32 In another paper,33 they proposed an FI procedure for the determination of polyphenols using banana pulp and spinach leaf solution as carriers. The oxygen consumption in the enzymatic reaction was monitored by an oxygen electrode and its concentration decrease was inversely proportional to the phenolic substrate concentration.However, the former enzymatic source is easily oxidized in air and its Analyst, April 1997, Vol. 122 (345–350) 345colour changes rapidly from white–yellow to brown after the peel has been removed. In addition, it is necessary to preoxidize the banana pulp completely in air and store it for 1 month before use.Also, it is difficult to use banana pulp solution as a carrier solution in an FI procedure owing to the instability of the baseline. On the other hand, the spinach leaf solution did not exhibit high enzyme activity immediately and had to be stored at 4 °C for 24 h in a refrigerator to release the enzyme polyphenol oxidase from the cells before a centrifugation step in the presence of 0.3% m/v sodium azide to prevent oxidation. Even in the presence of this antioxidant, the long-term stability of the spinach leaf solution at 4 °C is only 7 d.In this work, an FI spectrophotometric procedure is reported for determining l-dopa and carbidopa in pharmaceutical formulations. A crude extract of sweet potato root [Ipomoea batatas (L.) Lam.] was used as the enzymatic source of polyphenol oxidase (PPO; tyrosinase; catechol oxidase; EC.1.14.18.1 ) directly in the carrier.This enzyme catalyses the oxidation of these catecholamines to the corresponding dopaquinone. Further, dopaquinone is converted to leucodopachrome by a rapid spontaneous auto-oxidation which is in turn oxidized to dopachrome which presents a strong absorption at 480 and 360 nm for l-dopa and carbidopa, respectively. The use of an insoluble poly(vinylpyrrolidone) (PVP) such as Polyclar SB-100 to remove natural phenolic compounds (e.g., chlorogenic and isochlorogenic acids) from solution in the preparation of the crude extract of sweet potato root led to a considerable increase in the enzyme activity, storage time and stability of the baseline as well as to a decrease in the time required to obtain the enzymatic carrier, since no previous storage is needed to release the enzyme.The enzymatic FI methoere to determine either l-dopa or carbidopa selectively could be employed as an inexpensive alternative to those procedures that use pure enzymes and/or chromogenic reagents.Experimental Apparatus A DuPont Instruments (Newtown, CT, USA) Model RC-5B centrifuge, provided with a Model SS-34 rotor, was used. A Hewlett-Packard (Boise, ID, USA) Model 8452A UV– visible spectrophotometer was used in all spectrophotometric measurements. An eight-channel Ismatec (Zurich, Switzerland) Model 7618-40 peristaltic pump supplied with Tygon pump tubing was used for the propulsion of the fluids.The manifold was constructed with polyethylene tubing (0.8 mm id). Sample injection was performed using a laboratory-constructed threepiece manual commutator34 made of Perspex, with two fixed side bars and a sliding central bar, which is moved for sampling and injection. FI spectrophotometric measurements were carried out using a Femto (S�ao Paulo, Brazil) Model 435 spectrophotometer with a glass flow cell (optical path 1.0 cm) connected to a Cole Parmer (Niles, IL, USA) Model 12020000 two-channel strip-chart recorder.The effect of temperature on the enzymatic reaction was evaluated using a Tecnal (Piracicaba, Brazil) Model TE184 thermostatically controlled water-bath. Reagents and Solutions All reagents were of analytical-reagent grade and all solutions were prepared with water from a Millipore (Bedford, MA, USA) Milli-Q system (Model UV Plus Ultra-Low Organics Water). Sucrose, glucose, fructose, lactose, starch, poly- (ethylene glycol) 1500, sodium chloride, magnesium stearate and indigo carmine were purchased from Sigma (St.Louis, MO, USA). l-Dopa was purchased from BDH (Poole, Dorset, UK) and carbidopa was kindly provided by PRODOME Chemical and Pharmaceutical (Campinas, SP, Brazil); 1.0 31022 mol l21 stock solutions were prepared daily in 0.1 mol l21 phosphate buffer of pH 7.0 and standardized by a conventional method.29 Standard solutions from 4.0 31024 to 1.0 31022 mol l21 were prepared from the stock solutions in 0.1 mol l21 phosphate buffer of pH 7.0.Amberlite CG-400 ion-exchange resin from Aldrich (Milwaukee, WI, USA) was used as received. The PVPs Polyclar AT, Polyclar SB-100, Polyclar R and Polyclar K-30 were obtained from GAF (Wayne, NJ, USA) and were purified essentially as described by McFarlane and Vader,35 i.e., the PVPs were boiled for 10 min in 10% v/v HCl and washed with distilled water until free of chloride ion, then washed with acetone and dried.Healthy sweet potato roots [Ipomoea batatas (L.) Lam.] purchased from a local producer were selected, washed, handpeeled, chopped and frozen in liquid nitrogen or in a freezer. Methods Preparation of the crude extract Twenty-five grams of the frozen sweet potato root were homogenized in a liquefier with 100 ml of 0.1 mol l21 phosphate buffer of pH 7.0, containing 2.5 g of Polyclar SB-100 or other stabilizer/protective agent, for 2 min at 4–6 °C.The suspension was filtered through four layers of cheesecloth and centrifuged at 25 000 3g (18 000 rev min21) for 30 min at 4 °C; it was stored at this temperature in a refrigerator and utilized as the enzymatic source in the FI spectrophotometric procedure after the determination of the PPO activity and total protein. Measurement of PPO activity The activity of soluble PPO present in the crude extract was determined in triplicate by measurement of the absorbance at 410 nm of melanin-like pigments formed in the polymerization of quinone produced by the reaction between 0.2 ml of supernatant solution and 2.8 ml of 0.05 mol l21 catechol solution in 0.1 mol l21 phosphate buffer (pH 7.0) at 25 °C. The initial rate of the enzyme-catalysed reaction was a linear function of time for 1.5–2.0 min.One activity unit (U) is defined as the amount of enzyme that causes an increase of 0.001 A min21 under the conditions described above.36 Total protein determination The protein concentration was determined in triplicate by the method of Lowry et al.37 using bovine serum albumin as standard. PPO solution in phosphate buffer A 120 U PPO solution in 0.1 mol l21 phosphate buffer was prepared daily by dilution of 33 ml of a 908 U PPO solution in 0.1 mol l21 phosphate buffer in a 250 ml calibrated flask using the same buffer solution.FI procedure Fig. 1 shows a schematic diagram of the spectrophotometric flow system used. A 120 U PPO solution in 0.1 mol l21 phosphate buffer was used as the carrier (C) at a flow rate of 1.01 ml min21.A solution contained in the sample loop (L, 500 ml) was injected and transported by the carrier after the baseline had reached a steady-state value. A 400 cm tubular coiled reactor maintained in a water-bath (R) at 25 °C was placed in the analytical path in order to provide better reaction conditions. The coloured zone was measured in the flow-through spectrophotometric cell (SC) at 370 nm (carbidopa) or 500 nm (ldopa). 346 Analyst, April 1997, Vol. 122Determination of L-dopa and carbidopa in pharmaceutical formulations The contents of 20 tablets were well mixed; from the fine powder an accurately weighed portion was taken and dissolved in phosphate buffer (pH 7.0; 0.1 mol l21 at 25 °C). Using a mechanical shaker or an ultrasonic bath, the powder was completely disintegrated and the solution was clarified by passing it through No. 1 filter-paper, after which appropiate dilutions were made.Results and Discussion Preparation of the Crude Extract The activity and total protein extracted varied according to the extraction procedure and medium used. The buffer-to-tissue ratio was an important factor in the extraction of PPO from sweet potato root. In this study, the enzyme was extracted using ratios of 2–6 : 1 v/m and the highest specific activity was obtained at a ratio of 4 : 1 v/m. It was also found that PPO could be extracted with a phosphate buffer solution of low molarity such as 0.05–0.4 mol l21 with the maximum yield achieved at a concentration of 0.1 mol l21 (Table 1). The effect of buffer pH on the extraction of PPO was also investigated in the pH range 6.0–8.0.The highest enzymatic activity was reached at pH 7.0, as shown in Table 1. In addition, it is well known38 that the enzymatic browning tendency of sweet potato is related to natural phenolic compounds (natural substrates present in the root), particularly chlorogenic and isochlorogenic acids which comprise about 80% or more of the total phenolic compounds in the root. This process and the oxidation by atmospheric oxygen are responsible for the decrease in the PPO activity in the crude extract.In order to minimize this effect, several protective agents and/or stabilizers were investigated such as Polyclar K- 30, l-cysteine, l-cysteine + Amberlite CG- 400 ion-exchange resin, Amberlite CG-400 ion-exchange resin + EDTA + Polyclar AT, Amberlite CG-400 ion-exchange resin, Polyclar AT, Polyclar R and Polyclar SB-100.Table 2 presents the activity (U ml21), total protein (mg ml21) and specific activity (U mg21 of protein) of the crude extracts obtained in triplicate using these compounds. As can be seen, Polyclar SB-100 in the concentration ratio of 2.5 : 25.0 m/m was the best compound among those studied. Polyclar SB-100 is a PVP of high molecular mass which is supplied as a dry, finely divided white powder.Its remarkable ability to remove phenolic compounds from solutions39 and its low solubility made it particularly useful in the preparation of the crude extract, since it can be easily removed by filtration or by centrifugation. By using insoluble polymers, several workers29,39–42 have separated phenolic compounds that form strong H-bonded complexes (i.e., those with isolated hydroxyl groups) from many crude extracts, obtaining very active soluble enzymes.The inhibition of PPO by –SH compounds and other reducing agents is well documented.38 l-Cysteine used in this work probably inhibited the enzyme by interaction with the enzyme’s copper, leading to a decrease in the crude extract activity (see Table 2). The same effect was observed in our previous work29 and also in that of Lourenço et al.43 in the preparation of ude extracts of yam (Alocasia marcrohiza) and sweet potato root [Ipomoea batatas (L.) Lam.], respectively.Storage Time and Stability of Crude Extract For the optimum extraction conditions described above (Polyclar SB-100), the enzyme activity of the crude extract did not vary for at least 5 months when the extract was stored in a refrigerator at 4 °C and decreased by only 4–5% after an 8 h working period at 25 °C. Cysteine plus Amberlite CG-400 ionexchange resin also provided a good enzymatic stabilization, but not cysteine alone, as can be seen in Fig. 2. The long storage time achieved and the low background absorbances of the crude extract obtained with Polyclar SB-100 in comparison with other substances used here and/or sodium azide used by Uchiyama and Suzuki33 shows the advantage of the medium, preparation method and biological material used in this work.Reaction of L-Dopa and/or Carbidopa with PPO Fig. 3 shows the reaction steps of the catalytic oxidation of ldopa by PPO.27,30,44 This enzyme catalyses the oxidation of ldopa to dopaquinone.Further, dopaquinone is converted to leucodopachrome by a rapid spontaneous auto-oxidation which is in turn oxidized to dopachrome which has a strong absorption at 480 nm. Dopachrome can slowly be converted to melanin through a series of reactions catalysed by Zn2+ ions, as shown in this Fig. 3. An extensive spectrophotometric study of the reaction of 120 U of PPO from the crude extract of sweet potato with 5.0 3 1023 mol l21 l-dopa at 25 °C and pH 7.0 (0.1 mol l21 phosphate buffer) was carried out and the absorption spectrum of the dopachrome produced after 3 min is shown in Fig. 1 Schematic diagram of the FI system used for l-dopa and carbidopa determinations. The peristaltic pump is not shown and the broken line in the central bar of the manual injector shows the injection position after commutation. C, Carrier solution, 120 U of PPO in 0.1 mol l21 phosphate buffer (pH 7), flowing at 1.01 ml min21; L, sample loop (100 cm, 500 ml); S, sample or standard solution; R, tubular coiled reactor (400 cm) in a waterbath at 25 °C; SC, spectrophotometric cell [l = 500 nm (l-dopa) and 370 nm (carbidopa)]; W, waste.Table 1 Effect of phosphate buffer concentration and pH on the extraction of PPO from sweet potato root at 25 °C Phosphate buffer concentration/mol l21 0.05 0.1 0.2 0.3 0.4 Specific activity/ U mg21 protein 788 910 623 522 339 pH 6.0 6.5 7.0 7.5 8.0 Specific activity/ U mg21 protein 630 747 916 733 656 Table 2 Effect of protective agents and/or stabilizers on the extraction of PPO from sweet potato root at 25 °C Protective agent Activity/ Total protein/ Specific activity/ and/or stabilizer U ml21 mg ml21 U mg21 protein Polyclar K-30 1482 3.51 422 l-Cysteine 1605 3.68 436 l-Cysteine + Amberlite CG-400 1570 3.25 483 Amberlite 1930 3.80 508 CG-400 + EDTA + Polyclar A.T.Amberlite CG-400 2270 3.75 605 Polyclar A.T. 2440 3.80 642 Polyclar R 2795 3.85 726 Polyclar SB-100 2997 3.30 908 Analyst, April 1997, Vol. 122 347Fig. 4. The absorption of this chromophore did not increase significantly ( Å 15%) for the time range 2–10 min. In another study, the addition of 5.0 3 1023 mol l21 Zn2+ ions to the reaction medium led to an increase in the wavelength of absorption, which reached 540 nm after 1 h, indicating the formation of melanochrome.30,44 A similar spectrophotometric study was conducted for carbidopa and maximum absorption was obtained at 360 nm. As can be seen from these two spectra (Fig. 4), it is possible to determine each of these drugs selectively. Although dopachrome has a maximum absorption at 480 nm, all l-dopa determinations were performed at 500 nm in order to eliminate the interference from the carbidopa reaction product. Taking into account the enzymatic system characteristics, an FI system was developed, as shown in Fig. 1. FI Parameters and Reaction Conditions The effect of varying the sample loop length from 50 to 200 cm (250–1000 ml) on the absorbance signal was initially evaluated. The best sample loop length was found to be 100 cm (500 ml).With respect to sensitivity and analytical frequency, the best compromise was attained using a coiled reactor 400 cm long and a carrier flow rate of 1.01 ml min21. The effect of varying the enzyme concentration in the carrier solution from 30 to 185 U on the analytical signal (absorbance) for l-dopa at concentrations of 5.0 3 1024, 1.0 3 1023, 5.0 3 1023 and 1.0 31022 mol l21 was also investigated. In the range of l-dopa solution concentrations studied, the absorbance signal increased with an increase in the concentration of the enzyme solution used up to 100–120 U of PPO and then levelled off between 120 and 185 U.Consequently, a concentration of 120 U was used in this work. Table 3 shows the effect of pH in the range 5.0–8.0 on the absorbance of a 5.0 3 1023 mol l21 l-dopa solution under the conditions specified in the legend of Fig. 1. The optimum pH for PPO activity was 7.0. The same optimum pH was found by Lourenço et al.43 in pure and crude extracts of sweet potato. The tubular coiled reactor was placed in a water-bath and the effect of temperature was studied between 5 and 50 °C. The enzyme exhibited the highest activity in the temperature range 15–25 °C after which a gradual decline in its activity by heat inactivation was observed between 25 and 50 °C (see Table 3). In a bath study, heating at 90 °C for 40 min led to inactivation of 45% of the enzyme compared with its initial activity using l-dopa as substrate; thus, this shows the high stability of PPO in the medium used.Interference and Recovery Studies The effect of excipient substances frequently found with catecholamines in pharmaceutical formulations, such as sucrose, glucose, fructose, lactose, starch, poly(ethylene glycol), sodium chloride, magnesium stearate and indigo carmine, was evaluated using an FI system similar to that shown in Fig. 1. The ratios of the concentrations of l-dopa or carbidopa to those of the excipient substances were fixed at 0.1, 1.0 and 10.0. None of these substances interfered in the proposed FI method. Recoveries of 98.6 and 106.3% of l-dopa and carbidopa, respectively, from two pharmaceutical formulation samples (n = 6) were obtained using the FI spectrophotometric procedure. In this study, 4.9, 9.9 and 14.8 mg of l-dopa or carbidopa were added to each sample (Table 4).This is good evidence of the accuracy of the proposed method. In addition, the relative standard deviations were 1.28 and 1.33% for solutions containing 4.0 3 1023 mol l21 l-dopa or carbidopa (n = 12), respectively; furthermore, the baselines were stable. Analytical Curves and Applications The conditions determined above, i.e., a sample loop (L) of 100 cm (500 ml), a reactor length of 400 cm, a carrier (C) flow rate of 1.01 ml min21, an enzyme concentration of 120 U in 0.1 Fig. 2 Effect of the extraction medium (A, Polyclar SB-100; B, l-cysteine + Amberlite CG-400; and C, l-cysteine) on the storage time and/or stability of the crude extract of sweet potato root. Fig. 3 Reaction steps of the catalytic oxidation of l-dopa by PPO. 348 Analyst, April 1997, Vol. 122mol l21 phosphate buffer of pH 7.0 and a temperature of 25 °C, were used for the proposed method. Triplicate signals for nine l-dopa standard solutions and six consecutive signals for three pharmaceutical formulation samples and triplicate signals for ten carbidopa standard solutions and six consecutive signals for three samples are shown in Figs. 5(a) and (b), respectively. The calibration graphs obtained for l-dopa and carbidopa in the concentration range from 4.0 3 1024 to 1.0 3 1022 mol l21, using the flow system depicted in Fig. 1, were A = 0.0033 + 64.72C1, r = 0.9993, and A = 0.0087 + 130.28C2, r = 0.9990, where C1 and C2 are the concentrations of l-dopa and carbidopa in mol l21, respectively.After an 8 h working period, no baseline drift was observed and only a slight variation (4–5%) of the enzyme activity of the crude extract was observed. Table 5 presents the results obtained for these samples using a pharmacopeial method45 (a spectrophotometric method that detects both catecholamines at 280 nm) and the proposed enzymatic FI method and also those declared on the labels. The results obtained for l-dopa (1) and carbidopa (2) by the proposed enzymatic FI method are in close agreement with those reported (r1 = 0.9699 and r2 = 0.9999).Additionally, the total concentrations (l-dopa + carbidopa) obtained by the FI Fig. 4 Absorption spectra of the chromophores formed in the enzymatic reaction of PPO with 5.0 3 1023 mol l21 catecholamines: A, carbidopa and B, l-dopa at 25 °C and pH 7. Table 3 Effect of the pH of the 0.1 mol l21 phosphate buffer carrier solution and of the temperature of the tubular coiled reactor (R) bath on the absorbance value obtained for l-dopa at 500 nm.Experimental conditions as in Fig. 1 pH 5.0 5.5 6.0 6.5 7.0 7.5 8.0 Absorbance 0.201 0.218 0.263 0.343 0.375 0.351 0.339 Temperature/ °C 5.0 10.0 20.0 25.0 30.0 40.0 50.0 Absorbance 0.326 0.358 0.385 0.394 0.323 0.225 0.145 Table 4 Results of addition–recovery experiments using l-dopa and carbidopa with three different standard concentrations l-Dopa/mg* Carbidopa/mg* Recovery Recovery Sample Added Found (%) Added Found (%) 4.90 5.01 102.2 4.90 4.83 98.6 Sinemet 9.90 10.03 101.3 9.90 9.92 100.2 14.80 14.70 99.3 14.80 14.73 99.5 4.90 5.12 104.5 4.90 5.21 106.3 Cronomet 9.90 10.17 102.7 9.90 10.24 103.4 14.80 14.91 100.7 14.80 15.05 101.7 * n = 6.Table 5 Determination of l-dopa and carbidopa in pharmaceutical formulations using the pharmacopeial45 and enzymatic FI spectrophotometric procedures Label value/mg Pharmacopeia*/mg Enzymatic FI*/mg Relative error (RE) (%) Sample l-Dopa Carbidopa l-Dopa + Carbidopa l-Dopa Carbidopa RE1 † RE2 ‡ RE3 § Sinemet 250 25 274.5 ± 0.2 251.0 ± 0.1 24.5 ± 0.1 +0.4 22.0 +0.4 Cronomet 200 50 248.0 ± 0.3 211.2 ± 0.2 52.6 ± 0.1 +5.6 +2.5 +6.4 Prolopa 200 0 206.3 ± 0.1 197.7 ± 0.1 0.0 21.2 0.0 +4.2 * n = 6, confidence level 95%.† RE1 = Enzymatic FI versus l-dopa label value. ‡ RE2 = Enzymatic FI versus carbidopa label value. § RE3 = Enzymatic FI (l-dopa + carbidopa) versus pharmacopeial value. Fig. 5 Transient absorbance signals obtained in triplicate for standard catecholamines solutions: (a), 4.0; 6.0; 8.0; 10.0; 20.0; 40.0; 60.0; 80.0; and 100.0 3 1024 mol l21 l-dopa and three samples (A, sinemet; B, cronomet; and C, prolopa) and the standard solutions again and (b), 4.0; 6.0; 8.0; 10.0; 20.0; 40.0; 60.0; 70.0; 80.0; and 100.0 3 1024 mol l21 carbidopa, the same three samples and the standard solutions again.Analyst, April 1997, Vol. 122 349method agreed with those obtained using the pharmacopeial method (r3 = 0.9675) and within an acceptable range of error. The detection limits were 1.5 3 1025 and 2.0 3 1025 mol l21 (three times the signal blank/slope), for l-dopa and carbidopa, respectively, and the throughput was 26 samples h21 with a relative standard deviation of less than 1.0% (n = 6).Also, 2.30 ml of crude extract were consumed in each determination, corresponding to only 72 mg of the original sweet potato root. Financial support of FAPESP (Processes 91/2637-5, 92/2637-5), PADCT/CNPq (Process 62.0060/91-3), CNPq (Process 50.1638/91-1), and also the scholarship granted by CAPES to I.C.V. are gratefully acknowledged.References 1 Stryer, L., Biochemistry, Freeman, New York, 3rd edn., 1988. 2 Moffat, A. C., Clarke’s Isolation and Identification of Drugs, Pharmaceutical Press, London, 2nd edn., 1986. 3 Pesce, A. J., and Kaplan, L. A., in Methods in Clinical Chemistry, ed. Bircher, S., C. V. Mosby, St. Louis, MO, 1987, pp. 944–963. 4 Deleu, D., Sarre, S., Herregodts, P., Ebinger, G., and Michotte, Y., J. Pharm. Biomed. Anal., 1991, 9, 159. 5 Cedarbaum, M. J., Williamson, R., and Kutt, H., J. Chromatogr. Biomed. Appl., 1987, 415, 393. 6 Baruzzi, A., Contin, M., Albani, F., and Riva, R., J. Chromatogr. Biomed. Appl., 1986, 375, 165. 7 Cummings, J., Matheson, M. L., and Smyth, F. J., J. Chromatogr. Biomed. Appl., 1990, 528, 43. 8 Rihbany, A. L., and Delaney, F. M., J. Chromatogr., 1982, 248, 125. 9 Mart�ýnez-Lozano, C., P�erez-Ruiz, T., Tom�as, V., and Val, O., Analyst, 1991, 116, 857. 10 Perez-Ruiz, T., Mart�ýnez-Lozano, C., Tom�as, V., and Val, O., Talanta, 1993, 40, 1625. 11 Deftereos, N. T., Calokerinos, A. C., and Efstathiou, C. E., Analyst, 1993, 118, 627. 12 Mie , X., and Yang, G., Fenxi Zaahi, 1992, 12, 172. 13 Hassib, S. T., Anal. Lett., 1990, 23, 2195. 14 Davidson, G. A., J. Pharm. Sci., 1984, 73, 1582. 15 Yucesoy, C., Gazi. Univ. Ecza. Cilik. Fak. Derg., 1990, 7, 43. 16 Hassib, S. T., and El-Khateeb, S. Z., Anal. Lett., 1990, 23, 255. 17 Davidson, G. A., J. Pharm. Biomed. Anal., 1985, 3, 235. 18 El-Kommos, E. M., Mohamed, A. E., and Khedr, S. A., J. Assoc. Off. Anal. Chem., 1990, 73, 516. 19 Maggi, N., and Cometti, A., J. Pharm. Sci., 1972, 61, 924. 20 Issopoulos, P. B., Fresenius’ J. Anal. Chem., 1990, 336, 124. 21 Zivanovic, L., Vasiljevic, M., Radulovic, D., and Kustrin, S. A., J. Pharm. Biomed. Anal., 1991, 9, 1157. 22 Issopoulos, P. B., Pharm.Acta Helv., 1989, 64, 82. 23 Steup, A., Metzner, J., and Voll, A., Pharmazie, 1986, 41, 739. 24 Sane, R. T., Bhounsule, G. J., and Sawant, S. V., Indian Drugs, 1987, 24, 207. 25 Hasan, B. A., Khalaf, K. D., and de la Guardia, M., Talanta, 1995, 42, 627. 26 Campanella, L., Beone, T., Sammartino, M. P., and Tomasseti, M., J. Pharm. Biomed. Anal., 1993, 11, 1099. 27 Behbahani, I., Miller, S. A., and O’Keefe, D. H., Microchem. J., 1993, 47, 251. 28 Hasebe, Y., Tanaka, Y., and Uchiyama, S., Anal.Lett., 1994, 27, 41. 29 Signori, C. A., and Fatibello-Filho, O., Qu�ým. Nova, 1994, 17, 38. 30 Vachtenheim, J., Duchon, J., and Matous, B., Anal. Biochem., 1985, 146, 405. 31 Uchiyama, S., Tofuku, Y., and Suzuki, S., Anal. Chim. Acta, 1988, 208, 291. 32 Uchiyama, S., and Suzuki, S., Bunseki Kagaku, 1990, 39, 793. 33 Uchiyama, S., and Suzuki, S., Anal. Chim. Acta, 1992, 261, 361. 34 Fatibello-Filho, O., Capelato, M. D., and Calafatti, S. A., Analyst, 1995, 120, 2407. 35 McFarlane, W. D., and Vader, M. J., J. Inst. Brewing, 1962, 68, 254. 36 Lourenço, E. J., Le�ao, J. S., and Neves, V. A., J. Sci. Food Agric., 1990, 52, 249. 37 Lowry, O. H., Rosebrough, N. J., Farr, A. L., and Randall, R. J., J. Biol. Chem., 1951, 193, 265. 38 Mayer, A. M., and Harel, E., Phytochemistry, 1979, 18, 193. 39 Loomis, W. D., and Battaile, J., Phytochemistry, 1966, 5, 423. 40 Hsu, A. F., Thomas, C. E., and Brauer, D., J. Food Sci., 1988, 53, 1743. 41 Walter, W.M., and Purcell, A. E., J. Agric. Food Chem., 1979, 27, 943. 42 Andersen, R. A., and Sowers, J. A., Phytochemistry, 1968, 7, 293. 43 Lourenço, E. J., Neves, V. A., and Da-Silva, M. A., J. Agric. Food Chem., 1992, 40, 2369. 44 Mason, H. S., in Advances in Enzymology, ed. Nord, F. F., Interscience, New York, 1955, vol. XVI, p. 164. 45 United States Pharmacopeia National Formulary XXI, US Pharmacopeial Convention, Rockville, MD, 1985, p. 586. Paper 6/06852I Received October 10, 1996 Accepted January 6, 1997 350 Analyst, April 1997, Vol. 122 Flow Injection Spectrophotometric Determination of L-Dopa and Carbidopa in Pharmaceutical Formulations Using a Crude Extract of Sweet Potato Root [ Ipomoea batatas (L.) Lam.] as Enzymatic Source Orlando Fatibello-Filho* and Iolanda da Cruz Vieira Departamento de Qu�ýmica, Grupo de Qu�ýmica Anal�ýtica, Centro de Ci�encias Exatas e de Tecnologia, Universidade Federal de S�ao Carlos, Caixa Postal 676, CEP 13.560-970, S�ao Carlos, SP, Brazil A flow injection (FI) spectrophotometric method is proposed for the determination of L-dopa and carbidopa in pharmaceutical formulations. After selection of the extraction medium (e.g., buffer-to-tissue ratio, pH, buffer concentration, protective agents and/or stabilizers) and storage conditions, crude extract of sweet potato root [Ipomoea batatas (L.) Lam.] was used as an enzymatic source of polyphenol oxidase (Tyrosinase; catechol oxidase; EC.1.14.18.1) directly in the carrier.This enzyme catalyses the oxidation of these esponding dopaquinone. Further, dopaquinone undergoes a rapid spontaneous auto-oxidation to leucodopachrome, which is in turn oxidized to dopachrome; this last compound has a strong absorption at 480 and 360 nm for L-dopa and carbidopa, respectively. For the optimum extraction conditions found the enzyme activity of the crude extract did not vary for at least 5 months when stored at 4 °C and decreased by only 4–5% during an 8 h working period at 25 °C.The results obtained for L-dopa and carbidopa by the proposed enzymatic FI method were in close agreement with the label values (r1 = 0.9699 and r2 = 0.9999) and also with those obtained using a pharmacopeial method (r3 = 0.9675). The throughput was 26 samples h21, and 2.30 ml of crude extract were consumed in each determination, corresponding to only 72 mg of the original sweet potato root.The detection limit (three times the signal blank/slope) was 1.5 31025 and 2.0 3 1025 mol l21 for L-dopa and carbidopa, respectively; the recovery of L-dopa and carbidopa from three samples ranged from 98.6 to 106.3% of the added amount. Keywords: L-Dopa; carbidopa; flow system; polyphenol oxidase; tyrosinase; catechol oxidase; sweet potato root [Ipomoea batatas (L.) Lam.]; pharmaceutical formulations l-Dopa [(2)-3-(3,4-dihydroxyphenyl)-l-alanine] and carbidopa [(2)-l-2-(3,4-dihydroxybenzyl)-2-hydrazinopropionic acid] are catecholamines with an alkylamine chain attached to a benzene ring bearing two hydroxyl groups.l-Dopa is an important neurotransmitter that is used for the treatment of neural disorders such as Parkinson’s disease. It is effective in relieving hypokinesia and can also decrease rigidity, oculogyric crises and tremor.1 After its oral administration, l-dopa is absorbed through the bowel at the level of the small intestine and metabolized by a decarboxylation process to dopamine and then to other metabolites.Both carbidopa and benserazide are used as inhibitors for the decarboxylase activity.2 Hence, the development of a method for the selective determination of ldopa and carbidopa is important, since they are frequently found together in some pharmaceutical formulations. Several methods have been proposed for the determination of these catecholamine drugs in biological specimens and/or pharmaceutical formulations. The determination of catecholamines in biological specimens normally requires the use of trace analysis techniques, mainly chromatography with fluorimetric or electrochemical detection.3 On the other hand, catecholamines are present in relatively large amounts in pharmaceutical formulations and much effort has been devoted to the development of simple, rapid, accurate and precise analytical methods.HPLC has been widely used for the determination of l-dopa in brain, plasma, urine, liver, serum, tissue and biological fluids.4–8 A photokinetic method based on the strong inhibitory effect of the catecholamines on the photochemical reaction between Rose Bengal and EDTA has been proposed.9 A fluorimetric method10 based on the inhibition by catechlolamines of the photoreduction of phloxin (tetrachlorotetrabromofluorescein) by EDTA has also been proposed.A flow injection (FI) determination of epinephrine, norepinephrine, dopamine and l-dopa by their chemiluminogenic oxidation with potassium permanganate in acidic medium in the presence of formaldehyde has been reported.11 Spectrophotometric methods have also been proposed for the determination of catecholamines in pharmaceutical formulations based on the direct measurement of l-dopa at 280 and 290 nm12 and by first-derivative spectrophotometry at 276 nm.13 With other UV spectrophotometric methods, this compound has been determined in the presence of germanium dioxide,14 boric acid,15 Na2B4O7 16 and Na2HPO4 17 at 292, 239.5, 287 and 292.5 nm, respectively.In addition, catecholamines have been determined in the visible region after reaction with metaperiodate,18 isonicotinic acid hydrazide in alkaline medium,19 FeIII and o-phenanthroline in a moderately acidic medium,20 PdCl2,21 molybdophosphoric acid,22 ninhydrin,23 ammonium metavanadate24and p-aminophenol in the presence of KIO4 after its oxidation in alkaline medium.25 Nevertheless, there are few enzymatic methods for determining catecholamines described in the literature26–30 and of these only one laborious manual spectrophotometric method30 has been proposed.l-Dopa was incubated with mammalian tyrosinase in the presence of an optimum concentration of ZnII ions for 50–60 min and the melanochrome formed in this reaction was monitored at 540 nm. Uchiyama and co-workers31,32 used a cucumber juice solution (crude extract) containing ascorbate oxidase as a carrier in an FI system for the determination of l-ascorbic acid31 and dehydroascorbic acid.32 In another paper,33 they proposed an FI procedure for the determination of polyphenols using banana pulp and spinach leaf solution as carriers.The oxygen consumption in the enzymatic reaction was monitored by an oxygen electrode and its concentration decrease was inversely proportional to the phenolic substrate concentration. However, the former enzymatic source is easily oxidized in air and its Analyst, April 1997, Vol. 122 (345–350) 345colour changes rapidly from white–yellow to brown after the peel has been removed. In addition, it is necessary to preoxidize the banana pulp completely in air and store it for 1 month before use. Also, it is difficult to use banana pulp solution as a carrier solution in an FI procedure owing to the instability of the baseline. On the other hand, the spinach leaf solution did not exhibit high enzyme activity immediately and had to be stored at 4 °C for 24 h in a refrigerator to release the enzyme polyphenol oxidase from the cells before a centrifugation step in the presence of 0.3% m/v sodium azide to prevent oxidation.Even in the presence of this antioxidant, the long-term stability of the spinach leaf solution at 4 °C is only 7 d. In this work, an FI spectrophotometric procedure is reported for determining l-dopa and carbidopa in pharmaceutical formulations. A crude extract of sweet potato root [Ipomoea batatas (L.) Lam.] was used as the enzymatic source of polyphenol oxidase (PPO; tyrosinase; catechol oxidase; EC.1.14.18.1 ) directly in the carrier.This enzyme catalyses the oxidation of these catecholamines to the corresponding dopaquinone. Further, dopaquinone is converted to leucodopachrome by a rapid spontaneous auto-oxidation which is in turn oxidized to dopachrome which presents a strong absorption at 480 and 360 nm for l-dopa and carbidopa, respectively.The use of an insoluble poly(vinylpyrrolidone) (PVP) such as Polyclar SB-100 to remove natural phenolic compounds (e.g., chlorogenic and isochlorogenic acids) from solution in the preparation of the crude extract of sweet potato root led to a considerable increase in the enzyme activity, storage time and stability of the baseline as well as to a decrease in the time required to obtain the enzymatic carrier, since no previous storage is needed to release the enzyme. The enzymatic FI method proposed here to determine either l-dopa or carbidopa selectively could be employed as an inexpensive alternative to those procedures that use pure enzymes and/or chromogenic reagents.Experimental Apparatus A DuPont Instruments (Newtown, CT, USA) Model RC-5B centrifuge, provided with a Model SS-34 rotor, was used. A Hewlett-Packard (Boise, ID, USA) Model 8452A UV– visible spectrophotometer was used in all spectrophotometric measurements. An eight-channel Ismatec (Zurich, Switzerland) Model 7618-40 peristaltic pump supplied with Tygon pump tubing was used for the propulsion of the fluids.The manifold was constructed with polyethylene tubing (0.8 mm id). Sample injection was performed using a laboratory-constructed threepiece manual commutator34 made of Perspex, with two fixed side bars and a sliding central bar, which is moved for sampling and injection. FI spectrophotometric measurements were carried out using a Femto (S�ao Paulo, Brazil) Model 435 spectrophotometer with a glass flow cell (optical path 1 cm) connected to a Cole Parmer (Niles, IL, USA) Model 12020000 two-channel strip-chart recorder.The effect of temperature on the enzymatic reaction was evaluated using a Tecnal (Piracicaba, Brazil) Model TE184 thermostatically controlled water-bath. Reagents and Solutions All reagents were of analytical-reagent grade and all solutions were prepared with water from a Millipore (Bedford, MA, USA) Milli-Q system (Model UV Plus Ultra-Low Organics Water).Sucrose, glucose, fructose, lactose, starch, poly- (ethylene glycol) 1500, sodium chloride, magnesium stearate and indigo carmine were purchased from Sigma (St. Louis, MO, USA). l-Dopa was purchased from BDH (Poole, Dorset, UK) and carbidopa was kindly provided by PRODOME Chemical and Pharmaceutical (Campinas, SP, Brazil); 1.0 31022 mol l21 stock solutions were prepared daily in 0.1 mol l21 phosphate buffer of pH 7.0 and standardized by a conventional method.29 Standard solutions from 4.0 31024 to 1.0 31022 mol l21 were prepared from the stock solutions in 0.1 mol l21 phosphate buffer of pH 7.0.Amberlite CG-400 ion-exchange resin from Aldrich (Milwaukee, WI, USA) was used as received. The PVPs Polyclar AT, Polyclar SB-100, Polyclar R and Polyclar K-30 were obtained from GAF (Wayne, NJ, USA) and were purified essentially as described by McFarlane and Vader,35 i.e., the PVPs were boiled for 10 min in 10% v/v HCl and washed with distilled water until free of chloride ion, then washed with acetone and dried.Healthy sweet potato roots [Ipomoea batatas (L.) Lam.] purchased from a local producer were selected, washed, handpeeled, chopped and frozen in liquid nitrogen or in a freezer. Methods Preparation of the crude extract Twenty-five grams of the frozen sweet potato root were homogenized in a liquefier with 100 ml of 0.1 mol l21 phosphate buffer of pH 7.0, containing 2.5 g of Polyclar SB-100 or other stabilizer/protective agent, for 2 min at 4–6 °C.The suspension was filtered through four layers of cheesecloth and centrifuged at 25 000 3g (18 000 rev min21) for 30 min at 4 °C; it was stored at this temperature in a refrigerator and utilized as the enzymatic source in the FI spectrophotometric procedure after the determination of the PPO activity and total protein. Measurement of PPO activity The activity of soluble PPO present in the crude extract was determined in triplicate by measurement of the absorbance at 410 nm of melanin-like pigments formed in the polymerization of quinone produced by the reaction between 0.2 ml of supernatant solution and 2.8 ml of 0.05 mol l21 catechol solution in 0.1 mol l21 phosphate buffer (pH 7.0) at 25 °C.The initial rate of the enzyme-catalysed reaction was a linear function of time for 1.5–2.0 min. One activity unit (U) is defined as the amount of enzyme that causes an increase of 0.001 A min21 under the conditions described above.36 Total protein determination The protein concentration was determined in triplicate by the method of Lowry et al.37 using bovine serum albumin as standard.PPO solution in phosphate buffer A 120 U PPO solution in 0.1 mol l21 phosphate buffer was prepared daily by dilution of 33 ml of a 908 U PPO solution in 0.1 mol l21 phosphate buffer in a 250 ml calibrated flask using the same buffer solution. FI procedure Fig. 1 shows a schematic diagram of the spectrophotometric flow system used. A 120 U PPO solution in 0.1 mol l21 phosphate buffer was used as the carrier (C) at a flow rate of 1.01 ml min21.A solution contained in the sample loop (L, 500 ml) was injected and transported by the carrier after the baseline had reached a steady-state value. A 400 cm tubular coiled reactor maintained in a water-bath (R) at 25 °C was placed in the analytical path in order to provide better reaction conditions.The coloured zone was measured in the flow-through spectrophotometric cell (SC) at 370 nm (carbidopa) or 500 nm (ldopa). 346 Analyst, April 1997, Vol. 122Determination of L-dopa and carbidopa in pharmaceutical formulations The contents of 20 tablets were well mixed; from the fine powder an accurately weighed portion was taken and dissolved in phosphate buffer (pH 7.0; 0.1 mol l21 at 25 °C). Using a mechanical shaker or an ultrasonic bath, the powder was completely disintegrated and the solution was clarified by passing it through No. 1 filter-paper, after which appropiate dilutions were made. Results and Discussion Preparation of the Crude Extract The activity and total protein extracted varied according to the extraction procedure and medium used. The buffer-to-tissue ratio was an important factor in the extraction of PPO from sweet potato root. In this study, the enzyme was extracted using ratios of 2–6 : 1 v/m and the highest specific activity was obtained at a ratio of 4 : 1 v/m.It was also found that PPO could be extracted with a phosphate buffer solution of low molarity such as 0.05–0.4 mol l21 with the maximum yield achieved at a concentration of 0.1 mol l21 (Table 1). The effect of buffer pH on the extraction of PPO was also investigated in the pH range 6.0–8.0. The highest enzymatic activity was reached at pH 7.0, as shown in Table 1. In addition, it is well known38 that the enzymatic browning tendency of sweet potato is related to natural phenolic compounds (natural substrates present in the root), particularly chlorogenic and isochlorogenic acids which comprise about 80% or more of the total phenolic compounds in the root.This process and the oxidation by atmospheric oxygen are responsible for the decrease in the PPO activity in the crude extract. In order to minimize this effect, several protective agents and/or stabilizers were investigated such as Polyclar K- 30, l-cysteine, l-cysteine + Amberlite CG- 400 ion-exchange resin, Amberlite CG-400 ion-exchange resin + EDTA + Polyclar AT, Amberlite CG-400 ion-exchange resin, Polyclar AT, Polyclar R and Polyclar SB-100.Table 2 presents the activity (U ml21), total protein (mg ml21) and specific activity (U mg21 of protein) of the crude extracts obtained in triplicate using these compounds. As can be seen, Polyclar SB-100 in the concentration ratio of 2.5 : 25.0 m/m was the best compound among those studied.Polyclar SB-100 is a PVP of high molecular mass which is supplied as a dry, finely divided white powder. Its remarkable ability to remove phenolic compounds from solutions39 and its low solubility made it particularly useful in the preparation of the crude extract, since it can be easily removed by filtration or by centrifugation. By using insoluble polymers, several workers29,39–42 have separated phenolic compounds that form strong H-bonded complexes (i.e., those with isolated hydroxyl groups) from many crude extracts, obtaining very active soluble enzymes.The inhibition of PPO by –SH compounds and other reducing agents is well documented.38 l-Cysteine used in this work probably inhibited the enzyme by interaction with the enzyme’s copper, leading to a decrease in the crude extract activity (see Table 2). The same effect was observed in our previous work29 and also in that of Lourenço et al.43 in the preparation of crude extracts of yam (Alocasia marcrohiza) and sweet potato root [Ipomoea batatas (L.) Lam.], respectively.Storage Time and Stability of Crude Extract For the optimum extraction conditions described above (Polyclar SB-100), the enzyme activity of the crude extract did not vary for at least 5 months when the extract was stored in a refrigerator at 4 °C and decreased by only 4–5% after an 8 h working period at 25 °C. Cysteine plus Amberlite CG-400 ionexchange resin also provided a good enzymatic stabilization, but not cysteine alone, as can be seen in Fig. 2. The long storage time achieved and the low background absorbances of the crude extract obtained with Polyclar SB-100 in comparison with other substances used here and/or sodium azide used by Uchiyama and Suzuki33 shows the advantage of the medium, preparation method and biological material used in this work. Reaction of L-Dopa and/or Carbidopa with PPO Fig. 3 shows the reaction steps of the catalytic oxidation of ldopa by PPO.27,30,44 This enzyme catalyses the oxidation of ldopa to dopaquinone.Further, dopaquinone is converted to leucodopachrome by a rapid spontaneous auto-oxidation which is in turn oxidized to dopachrome which has a strong absorption at 480 nm. Dopachrome can slowly be converted to melanin through a series of reactions catalysed by Zn2+ ions, as shown in this Fig. 3. An extensive spectrophotometric study of the reaction of 120 U of PPO from the crude extract of sweet potato with 5.0 3 1023 mol l21 l-dopa at 25 °C and pH 7.0 (0.1 mol l21 phosphate buffer) was carried out and the absorption spectrum of the dopachrome produced after 3 min is shown in Fig. 1 Schematic diagram of the FI system used for l-dopa and carbidopa determinations. The peristaltic pump is not shown and the broken line in the central bar of the manual injector shows the injection position after commutation. C, Carrier solution, 120 U of PPO in 0.1 mol l21 phosphate buffer (pH 7), flowing at 1.01 ml min21; L, sample loop (100 cm, 500 ml); S, sample or standard solution; R, tubular coiled reactor (400 cm) in a waterbath at 25 °C; SC, spectrophotometric cell [l = 500 nm (l-dopa) and 370 nm (carbidopa)]; W, waste.Table 1 Effect of phosphate buffer concentration and pH on the extraction of PPO from sweet potato root at 25 °C Phosphate buffer concentration/mol l21 0.05 0.1 0.2 0.3 0.4 Specific activity/ U mg21 protein 788 910 623 522 339 pH 6.0 6.5 7.0 7.5 8.0 Specific activity/ U mg21 protein 630 747 916 733 656 Table 2 Effect of protective agents and/or stabilizers on the extraction of PPO from sweet potato root at 25 °C Protective agent Activity/ Total protein/ Specific activity/ and/or stabilizer U ml21 mg ml21 U mg21 protein Polyclar K-30 1482 3.51 422 l-Cysteine 1605 3.68 436 l-Cysteine + Amberlite CG-400 1570 3.25 483 Amberlite 1930 3.80 508 CG-400 + EDTA + Polyclar A.T.Amberlite CG-400 2270 3.75 605 Polyclar A.T. 2440 3.80 642 Polyclar R 2795 3.85 726 Polyclar SB-100 2997 3.30 908 Analyst, April 1997, Vol. 122 347Fig. 4. The absorption of this chromophore did not increase significantly ( Å 15%) for the time range 2–10 min. In another study, the addition of 5.0 3 1023 mol l21 Zn2+ ions to the reaction medium led to an increase in the wavelength of absorption, which reached 540 nm after 1 h, indicating the formation of melanochrome.30,44 A similar spectrophotometric study was conducted for carbidopa and maximum absorption was obtained at 360 nm.As can be seen from these two spectra (Fig. 4), it is possible to determine each of these drugs selectively. Although dopachrome has a maximum absorption at 480 nm, all l-dopa determinations were performed at 500 nm in order to eliminate the interference from the carbidopa reaction product. Taking into account the enzymatic system characteristics, an FI system was developed, as shown in Fig. 1. FI Parameters and Reaction Conditions The effect of varying the sample loop length from 50 to 200 cm (250–1000 ml) on the absorbance signal was initially evaluated.The best sample loop length was found to be 100 cm (500 ml). With respect to sensitivity and analytical frequency, the best compromise was attained using a coiled reactor 400 cm long and a carrier flow rate of 1.01 ml min21. The effect of varying the enzyme concentration in the carrier solution from 30 to 185 U on the analytical signal (absorbance) for l-dopa at concentrations of 5.0 3 1024, 1.0 3 1023, 5.0 3 1023 and 1.0 31022 mol l21 was also investigated.In the range of l-dopa solution concentrations studied, the absorbance signal increased with an increase in the concentration of the enzyme solution used up to 100–120 U of PPO and then levelled off between 120 and 185 U. Consequently, a concentration of 120 U was used in this work. Table 3 shows the effect of pH in the range 5.0–8.0 on the absorbance of a 5.0 3 1023 mol l21 l-dopa solution under the conditions specified in the legend of Fig. 1. The optimum pH for PPO activity was 7.0. The same optimum pH was found by Lourenço et al.43 in pure and crude extracts of sweet potato. The tubular coiled reactor was placed in a water-bath and the effect of temperature was studied between 5 and 50 °C. The enzyme exhibited the highest activity in the temperature range 15–25 °C after which a gradual decline in its activity by heat inactivation was observed between 25 and 50 °C (see Table 3).In a bath study, heating at 90 °C for 40 min led to inactivation of 45% of the enzyme compared with its initial activity using l-dopa as substrate; thus, this shows the high stability of PPO in the medium used. Interference and Recovery Studies The effect of excipient substances frequently found with catecholamines in pharmaceutical formulations, such as sucrose, glucose, fructose, lactose, starch, poly(ethylene glycol), sodium chloride, magnesium stearate and indigo carmine, was evaluated using an FI system similar to that shown in Fig. 1. The ratios of the concentrations of l-dopa or carbidopa to those of the excipient substances were fixed at 0.1, 1.0 and 10.0. None of these substances interfered in the proposed FI method. Recoveries of 98.6 and 106.3% of l-dopa and carbidopa, respectively, from two pharmaceutical formulation samples (n = 6) were obtained using the FI spectrophotometric procedure. In this study, 4.9, 9.9 and 14.8 mg of l-dopa or carbidopa were added to each sample (Table 4).This is good evidence of the accuracy of the proposed method. In addition, the relative standard deviations were 1.28 and 1.33% for solutions containing 4.0 3 1023 mol l21 l-dopa or carbidopa (n = 12), respectively; furthermore, the baselines were stable. Analytical Curves and Applications The conditions determined above, i.e., a sample loop (L) of 100 cm (500 ml), a reactor length of 400 cm, a carrier (C) flow rate of 1.01 ml min21, an enzyme concentration of 120 U in 0.1 Fig. 2 Effect of the extraction medium (A, Polyclar SB-100; B, l-cysteine + Amberlite CG-400; and C, l-cysteine) on the storage time and/or stability of the crude extract of sweet potato root. Fig. 3 Reaction steps of the catalytic oxidation of l-dopa by PPO. 348 Analyst, April 1997, Vol. 122mol l21 phosphate buffer of pH 7.0 and a temperature of 25 °C, were used for the proposed method. Triplicate signals for nine l-dopa standard solutions and six consecutive signals for three pharmaceutical formulation samples and triplicate signals for ten carbidopa standard solutions and six consecutive signals for three samples are shown in Figs. 5(a) and (b), respectively. The calibration graphs obtained for l-dopa and carbidopa in the concentration range from 4.0 3 1024 to 1.0 3 1022 mol l21, using the flow system depicted in Fig. 1, were A = 0.0033 + 64.72C1, r = 0.9993, and A = 0.0087 + 130.28C2, r = 0.9990, where C1 and C2 are the concentrations of l-dopa and carbidopa in mol l21, respectively.After an 8 h working period, no baseline drift was observed and only a slight variation (4–5%) of the enzyme activity of the crude extract was observed. Table 5 presents the results obtained for these samples using a pharmacopeial method45 (a spectrophotometric method that detects both catecholamines at 280 nm) and the proposed enzymatic FI method and also those declared on the labels.The results obtained for l-dopa (1) and carbidopa (2) by the proposed enzymatic FI method are in close agreement with those reported (r1 = 0.9699 and r2 = 0.9999). Additionally, the total concentrations (l-dopa + carbidopa) obtained by the FI Fig. 4 Absorption spectra of the chromophores formed in the enzymatic reaction of PPO with 5.0 3 1023 mol l21 catecholamines: A, carbidopa and B, l-dopa at 25 °C and pH 7. Table 3 Effect of the pH of the 0.1 mol l21 phosphate buffer carrier solution and of the temperature of the tubular coiled reactor (R) bath on the absorbance value obtained for l-dopa at 500 nm. Experimental conditions as in Fig. 1 pH 5.0 5.5 6.0 6.5 7.0 7.5 8.0 Absorbance 0.201 0.218 0.263 0.343 0.375 0.351 0.339 Temperature/ °C 5.0 10.0 20.0 25.0 30.0 40.0 50.0 Absorbance 0.326 0.358 0.385 0.394 0.323 0.225 0.145 Table 4 Results of addition–recovery experiments using l-dopa and carbidopa with three different standard concentrations l-Dopa/mg* Carbidopa/mg* Recovery Recovery Sample Added Found (%) Added Found (%) 4.90 5.01 102.2 4.90 4.83 98.6 Sinemet 9.90 10.03 101.3 9.90 9.92 100.2 14.80 14.70 99.3 14.80 14.73 99.5 4.90 5.12 104.5 4.90 5.21 106.3 Cronomet 9.90 10.17 102.7 9.90 10.24 103.4 14.80 14.91 100.7 14.80 15.05 101.7 * n = 6.Table 5 Determination of l-dopa and carbidopa in pharmaceutical formulations using the pharmacopeial45 and enzymatic FI spectrophotometric procedures Label value/mg Pharmacopeia*/mg Enzymatic FI*/mg Relative error (RE) (%) Sample l-Dopa Carbidopa l-Dopa + Carbidopa l-Dopa Carbidopa RE1 † RE2 ‡ RE3 § Sinemet 250 25 274.5 ± 0.2 251.0 ± 0.1 24.5 ± 0.1 +0.4 22.0 +0.4 Cronomet 200 50 248.0 ± 0.3 211.2 ± 0.2 52.6 ± 0.1 +5.6 +2.5 +6.4 Prolopa 200 0 206.3 ± 0.1 197.7 ± 0.1 0.0 21.2 0.0 +4.2 * n = 6, confidence level 95%. † RE1 = Enzymatic FI versus l-dopa label value.‡ RE2 = Enzymatic FI versus carbidopa label value. § RE3 = Enzymatic FI (l-dopa + carbidopa) versus pharmacopeial value.Fig. 5 Transient absorbance signals obtained in triplicate for standard catecholamines solutions: (a), 4.0; 6.0; 8.0; 10.0; 20.0; 40.0; 60.0; 80.0; and 100.0 3 1024 mol l21 l-dopa and three samples (A, sinemet; B, cronomet; and C, prolopa) and the standard solutions again and (b), 4.0; 6.0; 8.0; 10.0; 20.0; 40.0; 60.0; 70.0; 80.0; and 100.0 3 1024 mol l21 carbidopa, the same three samples and the standard solutions again. Analyst, April 1997, Vol. 122 349method agreed with those obtained using the pharmacopeial method (r3 = 0.9675) and within an acceptable range of error.The detection limits were 1.5 3 1025 and 2.0 3 1025 mol l21 (three times the signal blank/slope), for l-dopa and carbidopa, respectively, and the throughput was 26 samples h21 with a relative standard deviation of less than 1.0% (n = 6). Also, 2.30 ml of crude extract were consumed in each determination, corresponding to only 72 mg of the original sweet potato root.Financial support of FAPESP (Processes 91/2637-5, 92/2637-5), PADCT/CNPq (Process 62.0060/91-3), CNPq (Process 50.1638/91-1), and also the scholarship granted by CAPES to I.C.V. are gratefully acknowledged. References 1 Stryer, L., Biochemistry, Freeman, New York, 3rd edn., 1988. 2 Moffat, A. C., Clarke’s Isolation and Identification of Drugs, Pharmaceutical Press, London, 2nd edn., 1986. 3 Pesce, A. J., and Kaplan, L. A., in Methods in Clinical Chemistry, ed.Bircher, S., C. V. Mosby, St. Louis, MO, 1987, pp. 944–963. 4 Deleu, D., Sarre, S., Herregodts, P., Ebinger, G., and Michotte, Y., J. Pharm. Biomed. Anal., 1991, 9, 159. 5 Cedarbaum, M. J., Williamson, R., and Kutt, H., J. Chromatogr. Biomed. Appl., 1987, 415, 393. 6 Baruzzi, A., Contin, M., Albani, F., and Riva, R., J. Chromatogr. Biomed. Appl., 1986, 375, 165. 7 Cummings, J., Matheson, M. L., and Smyth, F. J., J. Chromatogr. Biomed. Appl., 1990, 528, 43. 8 Rihbany, A. L., and Delaney, F. M., J. Chromatogr., 1982, 248, 125. 9 Mart�ýnez-Lozano, C., P�erez-Ruiz, T., Tom�as, V., and Val, O., Analyst, 1991, 116, 857. 10 Perez-Ruiz, T., Mart�ýnez-Lozano, C., Tom�as, V., and Val, O., Talanta, 1993, 40, 1625. 11 Deftereos, N. T., Calokerinos, A. C., and Efstathiou, C. E., Analyst, 1993, 118, 627. 12 Mie , X., and Yang, G., Fenxi Zaahi, 1992, 12, 172. 13 Hassib, S. T., Anal. Lett., 1990, 23, 2195. 14 Davidson, G. A., J. Pharm. Sci., 1984, 73, 1582. 15 Yucesoy, C., Gazi. Univ. Ecza. Cilik. Fak. Derg., 1990, 7, 43. 16 Hassib, S. T., and El-Khateeb, S. Z., Anal. Lett., 1990, 23, 255. 17 Davidson, G. A., J. Pharm. Biomed. Anal., 1985, 3, 235. 18 El-Kommos, E. M., Mohamed, A. E., and Khedr, S. A., J. Assoc. Off. Anal. Chem., 1990, 73, 516. 19 Maggi, N., and Cometti, A., J. Pharm. Sci., 1972, 61, 924. 20 Issopoulos, P. B., Fresenius’ J. Anal. Chem., 1990, 336, 124. 21 Zivanovic, L., Vasiljevic, M., Radulovic, D., and Kustrin, S. A., J. Pharm. Biomed. Anal., 1991, 9, 1157. 22 Issopoulos, P. B., Pharm. Acta Helv., 1989, 64, 82. 23 Steup, A., Metzner, J., and Voll, A., Pharmazie, 1986, 41, 739. 24 Sane, R. T., Bhounsule, G. J., and Sawant, S. V., Indian Drugs, 1987, 24, 207. 25 Hasan, B. A., Khalaf, K. D., and de la Guardia, M., Talanta, 1995, 42, 627. 26 Campanella, L., Beone, T., Sammartino, M. P., and Tomasseti, M., J. Pharm. Biomed. Anal., 1993, 11, 1099. 27 Behbahani, I., Miller, S. A., and O’Keefe, D. H., Microchem. J., 1993, 47, 251. 28 Hasebe, Y., Tanaka, Y., and Uchiyama, S., Anal. Lett., 1994, 27, 41. 29 Signori, C. A., and Fatibello-Filho, O., Qu�ým. Nova, 1994, 17, 38. 30 Vachtenheim, J., Duchon, J., and Matous, B., Anal. Biochem., 1985, 146, 405. 31 Uchiyama, S., Tofuku, Y., and Suzuki, S., Anal. Chim. Acta, 1988, 208, 291. 32 Uchiyama, S., and Suzuki, S., Bunseki Kagaku, 1990, 39, 793. 33 Uchiyama, S., and Suzuki, S., Anal. Chim. Acta, 1992, 261, 361. 34 Fatibello-Filho, O., Capelato, M. D., and Calafatti, S. A., Analyst, 1995, 120, 2407. 35 McFarlane, W. D., and Vader, M. J., J. Inst. Brewing, 1962, 68, 254. 36 Lourenço, E. J., Le�ao, J. S., and Neves, V. A., J. Sci. Food Agric., 1990, 52, 249. 37 Lowry, O. H., Rosebrough, N. J., Farr, A. L., and Randall, R. J., J. Biol. Chem., 1951, 193, 265. 38 Mayer, A. M., and Harel, E., Phytochemistry, 1979, 18, 193. 39 Loomis, W. D., and Battaile, J., Phytochemistry, 1966, 5, 423. 40 Hsu, A. F., Thomas, C. E., and Brauer, D., J. Food Sci., 1988, 53, 1743. 41 Walter, W. M., and Purcell, A. E., J. Agric. Food Chem., 1979, 27, 943. 42 Andersen, R. A., and Sowers, J. A., Phytochemistry, 1968, 7, 293. 43 Lourenço, E. J., Neves, V. A., and Da-Silva, M. A., J. Agric. Food Chem., 1992, 40, 2369. 44 Mason, H. S., in Advances in Enzymology, ed. Nord, F. F., Interscience, New York, 1955, vol. XVI, p. 164. 45 United States Pharmacopeia National Formulary XXI, US Pharmacopeial Convention, Rockville, MD, 1985, p. 586. Paper 6/06852I Received October 10, 1996 Accepted January 6, 1997 350 Analyst, April
ISSN:0003-2654
DOI:10.1039/a606852i
出版商:RSC
年代:1997
数据来源: RSC
|
|