摘要:

Critical Review Hydrogen Measurement by Prompt Gamma-ray Activation Analysis A Review Rick L. Paul Analytical Chemistry Division, National Institute of Standards and Technology, Gaithersburg, MD 20899, USA Summary of Contents Introduction Facilities Selection and Preparation of Samples Preparation of Standards and Blanks Sample Irradiation and Gamma-ray Counting Evaluation and Minimization of Uncertainties Counting Uncertainties Neutron Flux Variation and Sample Positioning Gamma-ray Background Neutron Self-shielding and Scattering Gamma-ray Attenuation Sample Heterogeneity Spectral Interferences Other Sources of Uncertainty Expanded Standard Uncertainties of Measurements Limits of Detection of Hydrogen Applications Future Developments References Keywords: Hydrogen; prompt gamma rays; activation analysis; cold neutrons Introduction Hydrogen, the ninth most abundant element (by mass) in the Earth’s atmosphere and crust, is present in virtually all materials.It has long been known that chemical properties of many substances are directly related to hydrogen content.The hydrogen content of organic compounds determines the degree of saturation and hence reactivity of carbon–carbon bonds. A major concern for several US industries today is the effect of trace hydrogen on materials properties. Hydrogen is known to cause embrittlement in steels at levels of 10 mg g21 and to affect the cracking strength of titanium alloys at mass fractions below 50 mg g21.1 In semiconductors, hydrogen causes bulk electrical effects at concentrations of 10 mg cm23 or even lower.2 Prompt gamma-ray activation analysis (PGAA) is suitable for the detection of hydrogen, both at high levels (mass fractions > 1%) and at lower levels, in a wide variety of materials.3–5 The technique is described in detail elsewhere.6–11 The sample is irradiated by a beam of neutrons, inducing nuclei of many elements to undergo neutron capture.Upon de-excitation, these nuclei emit prompt gamma-rays, which are measured using a high resolution gamma-ray detector.Qualitative analysis is accomplished by identification of gamma-ray energies, while comparison of gamma-ray intensities with those emitted by standards yields quantitative analysis. Hydrogen absorbs neutrons via the 1H(n,g)2H reaction, resulting in the emission of a 2223.3 keV gamma-ray. The analysis is nondestructive and, since both the neutron and gamma radiation are penetrating, the entire volume of sample irradiated by the neutron beam is analyzed.Because the analytical signal results from a nuclear and not a chemical reaction, the results are independent of the chemical form of the element. Finally, since the analysis is performed in situ, extraction of hydrogen from the sample is not required. Two separate PGAA facilities are available at the National Institute of Standards and Technology Research Reactor. The University of Maryland–NIST thermal neutron† PGAA (TNPGAA) facility, in operation since 1978, has proven useful for the determination of > 1 mg of hydrogen (mass fraction > 0.1% for a 1 g sample) in materials.Because the presence of hydrogenous shielding materials gives rise to a hydrogen gamma-ray background of 1 mg, this instrument is not normally used to measure hydrogen at trace levels. A second PGAA spectrometer, constructed as part of the Cold Neutron Research Facility and in operation since late 1990, uses a beam of ‘cold’ neutrons (l > 4 nm, energies < 0.005 eV) to irradiate samples.Because the hydrogen background of the cold neutron (CN)PGAA spectrometer is low ( < 10 mg), due to its nearly hydrogen-free construction, this instrument has been useful for analysis of 10 mg to 1 mg of hydrogen in samples. In this paper † Uranium atoms undergo fission to produce fast neutrons with energies 100 keV to 10 MeV. Moderation produces thermal neutrons with an average energy of 0.025 eV.Moderation at low temperature produces cold neutrons with energies < 0.25 eV. Rick L. Paul earned a B.S. in chemistry from Western Illinois University and a Ph.D. from Purdue University, where his research interests included neutron activation analysis of meteorites. He was a postdoctoral research associate at San Jose State University before coming to the National Institute of Standards and Technology in 1991. His current areas of research include cold neutron prompt gamma-ray activation analysis as well as radiochemical neutron activation analysis.He is an author of over 30 papers on PGAA. Analyst, March 1997, Vol. 122(35R–41R) 35Rwe describe these two facilities and their capabilities for hydrogen measurement, outline a procedure for measuring hydrogen by PGAA, discuss sources of error in PGAA measurements and how to correct for these errors, and relate applications for analysis of hydrogen in materials.Facilities Both the thermal and cold neutron PGAA facilities at NIST are described in detail elsewhere.7–11 The thermal neutron PGAA facility is located above the reactor core. Neutrons pass through a series of collimators in a 6 m long vertical aluminum beam tube. The 2.5 cm diameter unfiltered beam exits the top of the reactor, intercepting the target at 1 m above the exit. Targets are irradiated at a neutron flux of 3.3 3 108 cm22 s21 (at 20 MW reactor power).Prompt gamma-rays are measured by a germanium detector (27% efficiency relative to a 7.6 cm 3 7.6 cm sodium iodide crystal), surrounded by a heavily-shielded, split annulus, thallium-doped sodium iodide detector, which gives the facility both anti-Compton and pair spectrometry capabilities. The analyzer consists of an Ethernet-based Canberra Nuclear Acquisition Interface Module‡ connected to a Digital VAXstation 2000, and routinely collects 16,384 channels of Compton-suppressed data over a gamma-ray energy range of 0 to 8 MeV.The CNPGAA instrument is located on a straight guide in the Cold Neutron Research Facility. Reactor neutrons are moderated by passage through D2O ice at 30 K and then pass down the guide and through a cold beryllium–bismuth filter on their way to the CNPGAA station. The neutron beam is free of reactor gamma-rays and fast neutrons. The neutron flux at the sample position is 1.5 3 108 cm22 s21 (thermal equivalent). Prompt gamma-rays are measured by a high purity germanium detector (26% relative efficiency, 1.9 keV resolution), which is surrounded by a bismuth germanate Compton shield (Fig. 1). Gamma-ray spectra up to 10 MeV are collected using Canberra Nuclear Data Software on a VAXstation 3100. The hydrogen sensitivity (counts s21 mg21) for this instrument is a factor of three to four better than that measured for the TNPGAA instrument, and the hydrogen background is a factor of more than 100 lower. Selection and Preparation of Samples Although it is not always possible to choose the size and shape of a sample, as when analyzing an intact turbine blade or silicon disk, certain guidelines exist when a choice is possible.If analysis of the entire sample is desired, the sample must be smaller than the collimated neutron beam. For the CNPGAA instrument, sample diameter is also limited by presence of an upper neutron guide, which is 30 mm above the center of the CNPGAA neutron beam.Because of this, samples larger than 60 mm in diameter may be analyzed only on the outer edges. In order to avoid significant neutron self-shielding and gamma-ray attenuation during irradiation, the sample thickness should preferably not exceed 10 mm, and should be considerably less if the sample contains large amounts of elements which are strong neutron absorbers (such as B, Cd, Sm, Gd). The mass of the sample should be large enough to produce an observable signal for the element(s) of interest, but not so large that the total sample count rate saturates the detector. For the majority of materials analyzed, a relatively thin target with a diameter of 10 mm to 20 mm and a mass of 0.1 to 1 g represents the ideal sample size and geometry.Although PGAA has occasionally been used for analysis of thin films, it should be remembered that the technique measures total hydrogen. Samples analyzed should be homogeneous whenever possible, in order to simplify corrections for neutron self-shielding, scattering and gamma-ray attenuation. If a thin layer on a substrate is analyzed for hydrogen, the substrate should preferably consist of a material which does not significantly absorb neutrons and does not contain hydrogen or any other element which emits a gamma-ray near 2223 keV.Sample preparation is minimal. In order to avoid irregular sample geometries, powdered samples are usually made into cylindrical pellets using a hydraulic press and a 12 mm diameter die.Powders containing large quantities of elements which strongly absorb neutrons may be diluted with graphite if desired to avoid high gamma-ray count rates and large self-shielding corrections. If the surface of the sample contains grease or other contamination, it should be cleaned with ethanol or another suitable solvent prior to analysis. Small samples (those weighing < 2 g) are usually sealed into bags of FEP Teflon in a clean hood.Since the hydrogen content of this Teflon is negligible, this material is ideal for packaging samples for hydrogen analysis. Preparation of Standards and Blanks Accurate measurement of element mass fractions by PGAA requires precise determination of element sensitivities. Since element sensitivities are dependent upon the matrix and geometry of the material being analyzed, it is imperative that matrix and geometry of standards and samples be as closely matched as possible. This is especially important when analyzing highly hydrogenous samples, as element sensitivities are greatly affected by neutron scattering by hydrogen.If samples of widely differing hydrogen mass fraction are being analyzed, it may be necessary to prepare multiple standards in order to calibrate the analytical signal as a function of hydrogen mass fraction (see discussion below). As an alternative to measuring the absolute mass fraction of hydrogen in a sample, it is sometimes sufficient to measure the ratio of the mass fraction of hydrogen to that of another (comparator) element in the sample.In this case it may be necessary to simply measure a standard containing both hydrogen and the comparator element, or even to perform the analysis without standards, using tabulated values of sensitivity ratios. The use of ratio measurements to eliminate measurement errors in PGAA is discussed below. When determining hydrogen at trace levels, it is necessary to measure a blank. The blank should closely match the sample matrix and geometry, but should be free of hydrogen.Measurement of a blank allows correction for background generated from neutron capture by environmental hydrogen ‡ The identification of certain commercial equipment, instruments, or materials does not imply recommendation or endorsement by the National Institute of Standards and Technology. These identifications are made only in order to specify the experimental procedures in adequate detail.Fig. 1 Side view of the CNPGAA spectrometer in the NIST Cold Neutron Research Facility. 36R Analyst, March 1997, Vol. 122(which often varies with sample matrix) and also for background due to spectral interferences. Sample Irradiation and Gamma-ray Counting Targets (i.e., samples and standards) for irradiation are usually suspended in the neutron beam by means of Teflon strings at a 45° angle to both the detector and the neutron beam. For large and irregularly shaped samples, special sample holders and mounting procedures are employed.Samples for determination of trace hydrogen by CNPGAA are often irradiated inside an evacuated magnesium sample box in order to eliminate hydrogen background from neutron capture by water vapor in the air. Irradiation times are dependent upon the quantity of hydrogen in the target, the desired precision of the analysis, and the available beam time. Typical analysis times range from a few minutes to > 48 h.Target count rates are corrected for variations in the neutron flux over the course of an analysis. These variations are monitored by irradiating a Ti foil at regular intervals. For the TNPGAA instrument, the measured flux variation is small (generally < 0.5% on a day-to-day basis). Larger variations have been measured for the CNPGAA instrument. Data reduction and spectral manipulation are accomplished using Nuclear Data spectroscopy applications software, which includes a peak search and the SUM program written at NIST.12 The SUM program allows peak and background channels to be entered manually and is useful for integration of small peaks on high backgrounds.Evaluation and Minimization of Uncertainties Experimental uncertainties are evaluated using guidelines set by the International Organization for Standardization (ISO).13 Uncertainties from individual sources are added in quadrature to obtain the combined standard uncertainty for each measurement. The combined standard uncertainty is then multiplied by a coverage factor (usually 2) to obtain an expanded standard uncertainty corresponding to a confidence interval of approximately 95%.A discussion of sources of uncertainty and their relative contribution to the combined standard uncertainty of PGAA measurements is given below. Methods for minimizing each uncertainty are also described. Counting Uncertainties Counting uncertainties arise from the fact that radioactive decay is a random process, hence any measurement based on observing radiation emitted in nuclear decay is subject to some degree of statistical fluctuation.Because the statistics are straight forward, this uncertainty in any measurement may be accurately determined. Both the PEAK and SUM programs calculate the uncertainties associated with the integrated peaks. Uncertainties arising from gamma-ray counting can be lowered by increasing the counting time, hence long counting times are usually required when peak-to-background ratio is low.When measuring milligram amounts of hydrogen by CNPGAA, relative standard uncertainties of < 1% are obtainable with reasonably short analysis times. For measurement of hydrogen at trace levels, the counting uncertainty often represents the major uncertainty in the analysis, even when long counting times are employed. For example, CNPGAA measurement of 60 mg kg21 hydrogen in a 500 mg sample of a titanium alloy counted for 10 h yielded a relative standard counting uncertainty of 11%.The evaluation of counting uncertainties is discussed in detail by Knoll.14 Neutron Flux Variation and Sample Positioning The standard uncertainty associated with variations in the flux of the neutron beam and positioning of the sample may be determined by repeated measurement of a titanium foil whose geometry is comparable to that of the sample. The relative standard uncertainty is approximately 1% for measurements by CNPGAA, and @0.5% for TNPGAA measurements.Gamma-ray Background Neutron capture by hydrogen in the shielding and in the atmosphere gives rise to hydrogen background gamma rays in both the TNPGAA and CNPGAA facilities. The magnitude of this background is dependent upon the matrix of the sample. If the sample contains nuclides which significantly scatter neutrons, additional neutrons may be scattered into the shielding and surrounding environment, thus enhancing the hydrogen gamma-ray background.Scattering-induced background enhancement has been observed with both the TNPGAA15 and CNPGAA instruments. For this reason, the hydrogen gammaray background is determined from irradiation of a blank whenever possible. Although hydrogen itself has by far the largest scattering cross section (bound ss = 80 b) of any element,§ significant enhancement of background H, Al, and Pb gamma-rays has been observed with neutron scattering by other elements. Irradiation of graphite and beryllium targets by CNPGAA yielded up to a factor of 10 enhancement in the hydrogen background.Measurement errors arising from background enhancement may be minimized by calibrating background count rates as a function of target total neutron scattering cross-section.15 In the future, background enhancement in CNPGAA may be minimized by lining the inside of the sample chamber with neutron absorbing 6Li-glass.Subtraction of a hydrogen background count rate from the sample count rate does not introduce significant additional measurement uncertainty unless the background count rate is a significant fraction of the total hydrogen count rate. For the CNPGAA instrument, in the absence of significant background enhancement, a 12 h background determination yields a relative standard uncertainty in background hydrogen count rate of ~ 10%. If the sample count rate is @5 times the background count rate, i.e., if the sample contains less than 50 mg of hydrogen, the relative standard uncertainty of the measurement arising from background subtraction is > 1%.Neutron Self-shielding and Scattering Perhaps the greatest potential source of error in PGAA measurements arises from the interaction of neutrons with the target matrix. If the target matrix contains large quantities of strongly absorbing nuclides, self-shielding occurs, resulting in a decrease in the average neutron flux within the sample, and a corresponding decrease in element sensitivities.If the target composition, density and appropriate absorption cross-sections are known, and if the sample size and shape are well defined (e.g., a sphere, a slab, or a cylinder), this self-shielding effect may be corrected for using well known general absorption laws.16 Corrections for neutron self-shielding are larger for CNPGAA than for TNPGAA because of larger neutron capture cross-sections.Shielding corrections may be significant even for very thin ( < 1 mm) targets if appreciable amounts of B, Cd, Gd or Sm are present. For a 0.8 mm thick geological sample containing mass fraction of Å 1% m/m of boron the calculated self-shielding correction factor is about 5% for TNPGAA and 13% for CNPGAA. Neutron scattering is a significant source of error in PGAA when hydrogenous samples are measured. The effects of neutron scattering by hydrogen on element sensitivities in TNPGAA have been studied extensively.17–20 Measurements of 2 mm thick, disk-shaped targets have shown that sensitivities § Average neutron scattering cross sections for most elements are < 10b.Analyst, March 1997, Vol. 122 37Rfor most elements increase by 1.8% per 1% increase in H mass fraction. For a series of hydrogenous disks approximately 12 mm in diameter, and ranging in thickness from 1 to 12 mm, sensitivities are enhanced relative to sensitivities for nonhydrogenous disks.These enhancements are largely the result of elastic scattering which changes the mean free path of the neutron within the target and therefore changes the probability for neutron absorption. Experiments have shown that the effects of elastic scattering are minimized when spherical targets are measured. The effects of neutron scattering on CNPGAA sensitivities are different and more pronounced than those observed for TNPGAA.21,22 Sensitivities for CNPGAA decrease markedly with both increasing H mass fraction and target thickness.Studies indicate that interaction of cold neutrons with a warm target results in an increase in the average energy of the neutron beam. The effect is observed for both cylindrical and spherical targets. In TNPGAA, uncertainties arising from neutron scattering by hydrogen are minimized either by careful matching of geometry and matrix of the standards and samples or by calibrating element sensitivities as a function of H mass fraction.The latter procedure was used in the analysis of hydrogen and boron in 40 food and mineral supplements with hydrogen mass fractions ranging from 0.83 to 9.82%.23 Mixtures of D2O–H2O–B(OH)3 and pellets prepared from homogeneous mixtures of graphite, mannitol, and boron carbide were used to calibrate H and B sensitivities. Both H and B sensitivities showed good linear correlation with hydrogen mass fraction over the range of 0 to 11%. Relative standard uncertainties associated with calibration of TNPGAA sensitivities have been evaluated at @1%, based on calibration curve data.Such procedures give less satisfactory results for CNPGAA. Because scattering effects are more pronounced for CNPGAA, even small differences in sample scattering power and geometry can result in significant changes in element sensitivities. For this reason, CNPGAA has not often been used to measure element mass fractions in highly hydrogenous samples.A more common approach in CNPGAA has been to use a method of internal standards by measuring ratios of element mass fractions. Recent studies indicate that measurement errors arising from the effects of neutron self-shielding and scattering are largely eliminated when ratios are measured.24,25 This method has been used in the determination of hydrogen in titanium alloys.5 The ratio of hydrogen mass fraction to titanium mass fraction is measured and, since the titanium content of the alloy is known, the hydrogen mass fraction is easily calculated.A future approach to eliminating errors due to neutron scattering in CNPGAA may be to cool the target in order to bring it into thermal equilibrium with the neutron beam.22 The effects of neutron scattering on CNPGAA sensitivities in the cold target would then be similar to those observed for targets in TNPGAA, and could be mitigated by the use of spherical targets. Gamma-ray Attenuation Attenuation of the gamma-ray signal prior to detection may result in significant measurement error.The degree of attenuation is proportional to the thickness, density, and mass number (Z) of the attenuating medium, and is inversely proportional to gamma-ray energy. If the signal is collimated, the collimator should be composed of a high Z, high density material (e.g., lead or tungsten) in order to minimize the number of attenuated gamma-rays reaching the detector. Gamma rays emitted from a target are also subjected to a certain amount of attenuation by the emitting material.If the target is very thick, is of high density, or contains large quantities of high Z elements, this selfattenuation may be significant, particularly for gamma-rays of low energy (7500 keV). Errors arising from gamma-ray attenuation are not eliminated when element ratios are measured, especially when count rates given by gamma-rays of vastly different energies are compared.25 As an example, consider the measurement of hydrogen in a 1 cm thick cylinder of titanium mounted 45° to the detector.Gamma-ray attenuation within the sample results not only in a 12% decrease in the 2223 keV hydrogen signal, but also in a 3% decrease in the ratio of the 1381 keV Ti to the 2223 keV hydrogen signal. Several approaches may be used to minimize measurement errors arising from gamma-ray attenuation. Errors due to gamma-ray self-attenuation may be minimized by matching matrix and geometry of samples and standards.Alternatively, if the geometry and composition of the sample are known, selfattenuation of gamma rays may be corrected using derived equations.26 If one of the elements present in the sample matrix emits a spectrum of gamma-ray lines covering a range of energies, a third option is to calibrate gamma attenuation as a function of energy by comparing the relative intensities of these lines with those measured in a thin foil of the same material (where self-attenuation is minimal).Sample Heterogeneity Sample heterogeneity can introduce errors into PGAA measurements, regardless of whether absolute mass fractions or ratios are measured.25 Element sensitivities may be dependent on the distribution of the elements in a target. Corrections for neutron self-shielding or gamma-ray attenuation are difficult, if not impossible, for heterogeneous targets, since equations derived for these corrections assume a homogeneous composition for the sample.One method of minimizing measurement errors due to sample heterogeneity is to analyze the sample in several different orientations, then report the average of the measurements. Spectral Interferences Because of the complexity of prompt gamma-ray spectra, spectral interferences often constitute a major source of error in PGAA measurements. Current tabulations of prompt gammarays list few interferences for the H 2223.3 keV gamma-ray. Lone et al.27 list gamma-rays by Ba (2220.0 keV), Os (2223.3 keV), and Xe (2225.2 keV) in the H energy region, while Tuli’s compilation28 includes Gd (2220.3 keV), Ge (2223.0 keV), Nd (2223.5 keV), Ru (2223.8 keV), Sn (2224.7 keV), and Ba (2224.8 keV).However, these lists are incomplete and subject to error. Spectra of some elements have many small peaks which are not listed in these compilations. When measuring hydrogen at very low levels by CNPGAA, even tiny peaks in the hydrogen region can become major sources of error.For this reason, accurate determination of hydrogen at low levels in many materials is possible only after rigorous examination of spectra of samples and carefully prepared blanks. Although materials are available that have been certified as hydrogen-free by another analytical technique, it is often necessary to prepare blanks by removal of hydrogen from these materials. Many metals (e.g., Ti, Zr) readily lose hydrogen when heated in a vacuum.More volatile materials or those that decompose upon heating may require freeze-drying to remove residual moisture. It is desirable to prepare more than one blank in order to check the material for complete loss of hydrogen. If analysis of a blank reveals a peak in the hydrogen region, the relative intensity of this line to that of a known peak for an element present in the material should be calculated. If this ratio is the same for each blank analyzed, the peak is likely to be an interference and not due to residual hydrogen in the blank.CNPGAA analyses of samples and blanks containing Co and Cu have revealed previously unlisted peaks for these elements 38R Analyst, March 1997, Vol. 122near 2223 keV. A small peak at 2219 keV interfered with measurement of hydrogen in titanium, however, this peak was reduced with the use of Compton suppression.5 The magnitude of error introduced by spectral interferences in hydrogen measurements depends upon the mass fraction of the interfering element, the intensity of the interfering gamma-ray peak, and the mass fraction of hydrogen in the sample.For example, failure to correct for the titanium interference peak in the analysis of a 0.2 g sample of titanium containing hydrogen at 215 mg kg21 yields a value for hydrogen that is 25 to 30% too high. Other Sources of Uncertainty Measurements by PGAA are affected by many of the same sources of uncertainty that affect measurements by delayed neutron activation analysis (NAA). These include uncertainties arising from pulse pileup correction and sample weighing, which are negligible for most PGAA measurements.For hygroscopic samples, absorption of water represents an additional source of error, which can generally be eliminated by storing samples in a desiccator prior to analysis. A discussion of sources of error in NAA is given elsewhere.29 Expanded Standard Uncertainties of Measurements For determination of hydrogen at percentage levels in biological and geological samples by TNPGAA, expanded standard uncertainties are often less than 5% with counting times of < 24 h.Determination of large amounts of hydrogen by CNPGAA usually yields higher expanded uncertainties than TNPGAA, except where it is possible to measure the ratio of hydrogen to another element. Higher expanded uncertainties are often obtained when neutron self-shielding, gamma-ray attenuation, or sample heterogeneity are significant, or when measuring hydrogen at very low levels.Measurement of 50 to 200 mg kg21 in titanium alloys resulted in expanded uncertainties of 10 to 30% with analysis times of 5 h or longer.5 Future improvements to the CNPGAA instrument are expected to result in smaller uncertainties for measurement of trace hydrogen. Limits of Detection of Hydrogen Limits of detection (LOD) for hydrogen in selected matrices by thermal and cold neutron PGAA are given in Table 1.The LOD for a 1 g sample counted for 24 h is calculated as described by Currie,30 using the equation LOD = 4.65 (Rb/t)1/2/S where Rb is the background counting rate (counts s21), t is the duration of the count (s), and S is the hydrogen sensitivity in counts s21 mg21. The hydrogen signals from coal fly ash, bovine liver, and quartz are presumed to be free of spectral interferences. Because of the presence of a Ti interference peak at 2219 keV, the CNPGAA LOD for hydrogen in Ti-6Al-4V is higher than in the other three matrices.Applications CNPGAA has been used to measure hydrogen in a wide variety of materials (see Tables 2 and 3). Considerable emphasis has been placed upon the ability of the technique to measure hydrogen in metals. Hydrogen mass fractions of 50 to 750 mg kg21 have been measured in titanium alloy jet engine compressor blades in order to determine whether hydrogen embrittlement may be responsible for blade failure.5,31 The technique has also been used to analyze a series of titanium alloy standards containing 100 to 20 000 mg kg21 hydrogen,5 and to measure < 100 mg kg21 hydrogen in samples of nanocrystalline Fe, Pd, and Cu.Attempts to measure hydrogen in semiconductor materials have met with limited success. Less than 10 mg kg21 hydrogen has been measured in hydrothermal quartz, while hydrogen mass fractions of 50 to 80 mg kg21 have been measured in crystals of semiconductor grade germanium. 4,32 Other industrially important materials analyzed for H by CNPGAA have included pure and substituted fullerenes (as little as 0.02% m/m H measured),33 solid proton conductors,34 hydrofluorocarbons and solid acid catalysts,35 zeolite catalysts, and samples of substituted lithium niobate. Hydrogen has also been measured in geological materials, including the Allende meteorite,10,36 United States Geological Survey (USGS) rock standards and terrestrial basalts. TNPGAA has proven valuable for the determination of hydrogen (at mass fractions of 0.83 to 12%) and up to 22 other elements in foods and mineral supplements.23,37–39 Analysis of 22 USGS reference standards yielded hydrogen mass fractions from 0.02 to 3.1%.40 TNPGAA has also measured 580 to 920 mg kg21 hydrogen in ground ash from Mt.St. Helens,41 3.7 to 4% m/m hydrogen in bituminous coal standard reference materials,42 and 0.09 to 0.61% m/m hydrogen in construction materials (sand, cement and concrete).43 Future Developments Future developments in the CNPGAA instrument, such as the addition of improved neutron and gamma-ray shielding to the system and continued fine tuning of Compton suppression electronics are expected to result in lower gamma-ray background, higher signal to noise ratio and better H LOD.Such improvements should also result in smaller uncertainties in hydrogen measurements and also shorter analysis times. The replacement of the D2O cold source with a liquid hydrogen source will result in a 5–10 fold increase in the neutron capture rate, reducing the sample size required for measurement.The installation of a neutron focusing lens will make it possible to analyze samples less than a millimeter in diameter without loss in sensitivity as well as to probe larger samples for compositional mapping. The lens, which has been tested at the CNPGAA station, focuses the neutron beam to 0.5 mm, resulting in an increase by a factor of 80 in neutron current density over the unfocused beam.48 Future plans also call for the development of a lens which will bend the neutron beam as well as focus it.The new lens will lower the sample position by more than 40 mm, removing samples from the immediate vicinity of the upper neutron guide. Much work still remains in order to ensure accurate measurement of hydrogen in materials by CNPGAA. Reliable blanks need to be prepared for many materials in order to minimize errors due to spectral interferences.Further studies are also needed with cryogenic samples in order to develop a method for minimizing the effects of neutron scattering on CNPGAA measurements of highly hydrogenous samples. The potential of PGAA for measurement of hydrogen in materials is great and has yet to be fully realized. Research areas which should be more fully exploited in the future include the use of PGAA to study hydrogen embrittlement in a variety of metals, to accurately measure the degree of deuteration of Table 1 LOD for hydrogen in selected matrices for thermal and cold neutron PGAA Limit of detection/mg kg21 Coal Fly Ash Bovine Liver (SRM 1633) (SRM 1577) Quartz Ti-6Al-4V* Thermal 27 46 nd† nd Cold 8.5 nd 5.0 15 * An alloy of titanium containing Å 6% aluminum and 4% vanadium by mass.† nd, not determined. Analyst, March 1997, Vol. 122 39RTable 2 Some materials analyzed for hydrogen by CNPGAA* Hydrogen content Material CNPGAA Other methods Titanium (alloyed and pure): SRM 354 (unalloyed) 224 ± 22 mg kg21a 215 ± 6 mg kg21b SRM 352c (unalloyed) 58 ± 20 mg kg21a 49 ± 0.9 mg kg21b Ti-6Al-4V standards 117 ± 39 mg kg21c 114 ± 60 mg kg21d 164 ± 138 mg kg21c 175 ± 115 mg kg21d Nanocrystalline metals: Fe 0.250 ± 0.078% m/ma 0.214 ± 0.040% m/ma Pd 0.0080 ± 0.0086% m/ma Cu 0.0031 ± 0.013% m/ma Quartz: 6 ± 12 mg kg21e 5.4 mg kg21f 10 ± 12 mg kg21e 12 mg kg21f @5 mg kg21e 4.9 mg kg21f @6 mg kg21e 4.6 mg kg21f Semiconductor-grade germanium: 80 ± 55 mg kg21e 75 ± 50 mg kg21e 50 ± 40 mg kg21e Pure and substituted fullerenes: C60 C:H (atom ratio) = 180 ± 30g K3C60 C:H (atom ratio) = 114 ± 11g Rb2.6K0.4C60 C:H (atom ratio) = 68 ± 6g Na2RbC60 C:H (atom ratio) = 180 ± 35g Solid proton conductors: Yb doped SrCeO3 3.1 ± 0.8 mol%h 2.5 mol%i Undoped SrCeO3 0.7 ± 0.1 mol%h 0.5 mol%i Allende meteorite: 160 ± 10 mg kg21j 190 mg kg21, 180 mg kg21k; 110 mg kg21, 150 mg kg21l * All stated uncertainties are expanded standard uncertainties as defined in the text.a Previously unpublished results. b NIST certified values determined by hot extraction of hydrogen at 1400 °C and vacuum fusion of the sample at 1950 °C. c Ref. 5. d Determined using a LECO inert gas fusion analyzer, ref. 5. e Ref. 4. f Determined by IR, ref. 4. g Ref. 33. h Ref. 34. i Determined by thermal gravimetry, ref. 34. j Ref. 36. k Ref. 44. l Ref. 45. Table 3 Some materials analyzed for hydrogen by TNPGAA* Hydrogen content Material TNPGAA Other methods Biological materials: SRM 1549 (Milk powder) 5.9 ± 0.8% m/ma SRM 1570 (Spinach) 5.45 ± 0.08% m/mb 5.57% m/md SRM 1571 (Orchard Leaves) 5.84 ± 0.08% m/mb 5.84 ± 0.26% m/md SRM 1577 (Bovine Liver) 6.83 ± 0.08% m/mc 6.97 ± 0.16% m/md Ground beef 9.89 ± 0.12% m/ma Whole milk (fluid) 10.77 ± 0.09% m/ma Potato chips 7.32 ± 0.11% m/ma Corn flakes 6.14 ± 0.07% m/ma USGS reference standards: GSP-1 785 ± 40 mg k21e 680 ± 170 mg kg21f STM-1 0.18 ± 0.02% m/me 0.19% m/mf AGV-1 0.18 ± 0.02% m/me 0.20 ± 0.04% m/mf PCC-1 0.58 ± 0.01% m/me 0.57 ± 0.03% m/mf Volcanic ash (Mt. St.Helens): 580 ± 90 mg kg21g 900 ± 100 mg kg21g 920 ± 70 mg kg21g 870 ± 40 mg kg21g Construction materials: Sand 0.28 ± 0.01% m/me Cement 0.088 ± 0.014% m/me Concrete 0.61 ± 0.01% m/me * All stated uncertainties are expanded standard uncertainties as defined in the text. a Ref. 37. b Ref. 38. c Ref. 23. d Ref. 46. e Ref. 40. f Ref. 47. g Ref. 41. h Ref. 43. 40R Analyst, March 1997, Vol. 122materials, and to measure absorption of hydrogen by materials to determine their potential for use in hydrogen fuel cells. The cooperation of the NIST reactor staff during the PGAA measurements is gratefully acknowledged. R. Lindstrom, E. Mackey and D. Anderson are thanked for their advice and interest in this work. References 1 Meyn, D. A., Metall. Trans., 1974, 5, 2405. 2 Chevallier, J., and Aucouturier, M., Ann.Rev. Mater. Sci., 1988, 18, 219. 3 Lindstrom, R. M., Paul, R. L., Vincent, D. H., and Greenberg, R. R., J. Radioanal. Nucl. Chem., 1994, 180, 271. 4 Paul, R. L., and Lindstrom, R. M., Diagnostic Techniques for Semiconductor Processing, MRS Symposium Proceedings Volume 324, ed. Glembocki, O. J., Pang, S. W., Pollak, F. H., Crean, G. M., and Larrabee, G., Materials Research Society, Pittsburgh, PA, 1994, pp. 403–408. 5 Paul, R. L., Privett III, H. M., Lindstrom, R. M., Richards, W.J., and Greenberg, R. R., Metall. Mater. Trans. A, 1996, 27A, 3682. 6 Lindstrom, R. M., J. Res. Nat. Inst. Stand., Techn 1993, 98, 127. 7 Failey, M. P., Anderson, D. L., Zoller, W. H., Gordon, G. E., and Lindstrom, R. M., Anal. Chem., 1979, 51, 2209. 8 Anderson, D. L., Failey, M. P., Zoller, W. H., Walters, W. B., Gordon, G. E., and Lindstrom, R. M., J. Radioanal. Chem., 1981, 63, 97. 9 Lindstrom, R. M., Zeisler, R., Vincent, D. H., Greenberg, R. R., Stone, C. A., Mackey, E.A. Anderson, D. L., and Clark, D. D., J. Radioanal. Nucl. Chem., 1993, 167, 121. 10 Paul, R. L., Lindstrom, R. M., and Vincent, D. H., J. Radioanal. Nucl. Chem., 1994, 180, 263. 11 Mackey, E. A., Anderson, D. L., Chen, H., Downing, R. G., Greenberg, R. R., Lamaze, G. P., Lindstrom, R. M., Mildner, D. F. R., and Paul, R. L., J. Radioanal. Nucl. Chem., 1996, 203(2), 411. 12 Lindstrom, R. M., Biol. Trace Elem. Res., 1994, 43–45, 597. 13 International Standards Organisation, Guide to the Expression of Uncertainty in Measurement, 1st edn.ISO, Geneva, 1993. 14 Knoll, G. F., Radiation Detection and Measurement, Wiley, New York, 1989, pp. 65–99. 15 Anderson, D. L., and Mackey, E. A., J. Radioanal. Nucl. Chem., 1993, 167, 145. 16 Fleming, R. F., Int. Appl. Radiat. Isot., 1982, 33, 1263. 17 Copley, J. R. D., and Stone, C. A., Nucl. Instrum. Meth., 1989, A281, 593. 18 Mackey, E. A., Gordon, G. E., Lindstrom, R. M., and Anderson, D. L., Anal. Chem., 1991, 63, 288. 19 Mackey, E. A., Gordon, G. E., Lindstrom, R. M., and Anderson, D. L., Anal. Chem., 1992, 64, 2366. 20 Mackey, E. A., and Copley, J. R. D., J. Radioanal. Nucl. Chem., 1993, 167, 127. 21 Paul, R. L., and Mackey, E. A., J. Radioanal. Nucl. Chem., 1994, 181, 321. 22 Mackey, E. A., Biol. Trace Element Res., 1994, 43–45, 103. 23 Anderson, D. L., Cunningham, W. C., and Mackey, E. A., Biological Trace Element Research, 1990, 27, 616. 24 Lindstrom, R. M., Fleming, R. F., Paul, R.L., and Mackey, E. A., Proceedings of the International k0 Users Workshop, Gent, 1992, pp. 121–124. 25 Paul, R. L., J. Radioanal. Nucl. Chem., 1995, 191, 245. 26 Debertin, K., and Helmer, R. G., Gamma- and X-Ray Spectrometry with Semiconductor Detectors, Elsevier, Amsterdam, 1988. 27 Lone, M. A., Leavitt, R. A., and Harrison, D. A., Atomic Data and Nuclear Data Tables, 1981, 26, 511. 28 Tuli, J. K., in Prompt Gamma Neutron Activation Analysis, ed. Alfassi, Z. B., and Chung, C., CRC Press, Boca Raton, FL, 1995, pp. 177–223. 29 Blaauw, M., Ammerlaan, M. J. J., and Bode, P., Appl. Radiat. Isotop., 1993, 44, 547. 30 Currie, L. A., Anal. Chem., 1968, 40, 586. 31 Paul, R. L., and Lindstrom, R. M., Review of Progress in Quantitative Nondestructive Evaluation, vol. 13, ed. Thompson, D. O., and Chimenti, D. E., Plenum Press, New York, 1994, pp. 1619–1624. 32 Paul, R. L., and Lindstrom, R. M., International Workshop on Semiconductor Characterization: Present Status and Future Needs, ed.Bullis, W. M., Seiler, D. G., and Diebold, A. C., AIP Press, Woodbury, NY, 1996, pp. 342–345. 33 Neumann, D. A., Copley, J. R. D., Reznik, D., Kamitakahara, W. A., Rush, J. J., Paul, R. L., and Lindstrom, R. M., J. Phys. Chem. Solids, 1993, 54, 1699. 34 Krug, F., Schober, T., Paul, R., and Springer, T., Solid State Ionics, 1995, 77, 185. 35 Crawford, M. K., Corbin, D. R., and Vernooy, P. D., Trans. Am. Nucl. Soc., 1994, 71, 168. 36 Paul, R. L., Lindstrom, R.M., and Vincent, D. H., J. Radioanal. Nucl. Chem., 1995, 190, 181. 37 Anderson, D. L., Cunningham, W. C., and Lindstrom, T. R., J. Food Comp. Anal., 1994, 7, 59. 38 Anderson, D. L., Cunningham, W. C., and Alvarez, G. H., J. Radioanal. Nucl. Chem., 1993, 167, 139. 39 Anderson, D. L., and Cunningham, W. C., Trans. Am. Nucl. Soc., 1994, 71, 21. 40 Anderson, D. L., Sun, Y., Failey, M. P., and Zoller, W. H., Geostand. News., 1985, 9, 219. 41 Anderson, D. L., Phelan, J. M., Vossler, T., and Zoller, W.H., J. Radioanal. Chem., 1982, 71, 47. 42 Germani, M. S., Gokmen, I., Sigleo, A. C., Kowalczyk, G. S., Ohmez, I., Small, A., Anderson, D. L., Failey, M. P., Gulovali, M. C., Choquette, C. E., Lepel, E. A., Gordon, G. E., and Zoller, W. H., Anal. Chem., 1980, 52, 240. 43 Anderson, D. L., Gordon, G. E., and Lepel, E. A., Can. J. Chem., 1983, 61, 724. 44 Scoon, J. H., in The Allende Meteorite Reference Sample, ed. Jarosewich, E., Clarke, Jr. R. S., and Barrows, J.N., Smithsonian Contrib. Earth Sci., 1987, 27, 39. 45 Willis, J. P., in: The Allende Meteorite Reference Sample, ed. Jarosewich, E., Clarke, Jr. R. S., and Barrows, J. N., Smithsonian Contrib. Earth Sci., 1987, 27, 39. 46 Gladney, E. S., O’Malley, B. T., Roelandts, I., and Gills, T. E., National Bureau of Standards Special Publication NBS SP 260-111, 1987. 47 Gladney, E. S., Burns, C. E., and Roelandts, I., Geostand. News., 1983, 7, 3. 48 Chen, H., Sharov, V. A., Mildner, D.F. R., Downing, R. G., Paul, R. L., Lindstrom, R. M., Zeissler, C. J., and Xiao, Q.-F., Nucl. Instrum. Meth. B, 1995, 95, 107. Paper 6/06419A Received September 17, 1996 Accepted December 4, 1996 Analyst, March 1997, Vol. 122 41R Critical Review Hydrogen Measurement by Prompt Gamma-ray Activation Analysis A Review Rick L. Paul Analytical Chemistry Division, National Institute of Standards and Technology, Gaithersburg, MD 20899, USA Summary of Contents Introduction Facilities Selection and Preparation of Samples Preparation of Standards and Blanks Sample Irradiation and Gamma-ray Counting Evaluation and Minimization of Uncertainties Counting Uncertainties Neutron Flux Variation and Sample Positioning Gamma-ray Background Neutron Self-shielding and Scattering Gamma-ray Attenuation Sample Heterogeneity Spectral Interferences Other Sources of Uncertainty Expanded Standard Uncertainties of Measurements Limits of Detection of Hydrogen Applications Future Developments References Keywords: Hydrogen; prompt gamma rays; activation analysis; cold neutrons Introduction Hydrogen, the ninth most abundant element (by mass) in the Earth’s atmosphere and crust, is present in virtually all materials.It has long been known that chemical properties of many substances are directly related to hydrogen content. The hydrogen content of organic compounds determines the degree of saturation and hence reactivity of carbon–carbon bonds.A major concern for several US industries today is the effect of trace hydrogen on materials properties. Hydrogen is known to cause embrittlement in steels at levels of 10 mg g21 and to affect the cracking strength of titanium alloys at mass fractions below 50 mg g21.1 In semiconductors, hydrogen causes bulk electrical effects at concentrations of 10 mg cm23 or even lower.2 Prompt gamma-ray activation analysis (PGAA) is suitable for the detection of hydrogen, both at high levels (mass fractions > 1%) and at lower levels, in a wide variety of materials.3–5 The technique is described in detail elsewhere.6–11 The sample is irradiated by a beam of neutrons, inducing nuclei of many elements to undergo neutron capture.Upon de-excitation, these nuclei emit prompt gamma-rays, which are measured using a high resolution gamma-ray detector. Qualitative analysis is accomplished by identification of gamma-ray energies, while comparison of gamma-ray intensities with those emitted by standards yields quantitative analysis.Hydrogen absorbs neutrons via the 1H(n,g)2H reaction, resulting in the emission of a 2223.3 keV gamma-ray. The analysis is nondestructive and, since both the neutron and gamma radiation are penetrating, the entire volume of sample irradiated by the neutron beam is analyzed. Because the analytical signal results from a nuclear and not a chemical reaction, the results are independent of the chemical form of the element.Finally, since the analysis is performed in situ, extraction of hydrogen from the sample is not required. Two separate PGAA facilities are available at the National Institute of Standards and Technology Research Reactor. The University of Maryland–NIST thermal neutron† PGAA (TNPGAA) facility, in operation since 1978, has proven useful for the determination of > 1 mg of hydrogen (mass fraction > 0.1% for a 1 g sample) in materials. Because the presence of hydrogenous shielding materials gives rise to a hydrogen gamma-ray background of 1 mg, this instrument is not normally used to measure hydrogen at trace levels.A second PGAA spectrometer, constructed as part of the Cold Neutron Research Facility and in operation since late 1990, uses a beam of ‘cold’ neutrons (l > 4 nm, energies < 0.005 eV) to irradiate samples. Because the hydrogen background of the cold neutron (CN)PGAA spectrometer is low ( < 10 mg), due to its nearly hydrogen-free construction, this instrument has been useful for analysis of 10 mg to 1 mg of hydrogen in samples.In this paper † Uranium atoms undergo fission to produce fast neutrons with energies 100 keV to 10 MeV. Moderation produces thermal neutrons with an average energy of 0.025 eV. Moderation at low temperature produces cold neutrons with energies < 0.25 eV. Rick L. Paul earned a B.S. in chemistry from Western Illinois University and a Ph.D. from Purdue University, where his research interests included neutron activation analysis of meteorites. He was a postdoctoral research associate at San Jose State University before coming to the National Institute of Standards and Technology in 1991. His current areas of research include cold neutron prompt gamma-ray activation analysis as well as radiochemical neutron activation analysis.He is an author of over 30 papers on PGAA. Analyst, March 1997, Vol. 122(35R–41R) 35Rwe describe these two facilities and their capabilities for hydrogen measurement, outline a procedure for measuring hydrogen by PGAA, discuss sources of error in PGAA measurements and how to correct for these errors, and relate applications for analysis of hydrogen in materials.Facilities Both the thermal and cold neutron PGAA facilities at NIST are described in detail elsewhere.7–11 The thermal neutron PGAA facility is located above the reactor core. Neutrons pass through a series of collimators in a 6 m long vertical aluminum beam tube. The 2.5 cm diameter unfiltered beam exits the top of the reactor, intercepting the target at 1 m above the exit.Targets are irradiated at a neutron flux of 3.3 3 108 cm22 s21 (at 20 MW reactor power). Prompt gamma-rays are measured by a germanium detector (27% efficiency relative to a 7.6 cm 3 7.6 cm sodium iodide crystal), surrounded by a heavily-shielded, split annulus, thallium-doped sodium iodide detector, which gives the facility both anti-Compton and pair spectrometry capabilities.The analyzer consists of an Ethernet-based Canberra Nuclear Acquisition Interface Module‡ connected to a Digital VAXstation 2000, and routinely collects 16,384 channels of Compton-suppressed data over a gamma-ray energy range of 0 to 8 MeV. The CNPGAA instrument is located on a straight guide in the Cold Neutron Research Facility. Reactor neutrons are moderated by passage through D2O ice at 30 K and then pass down the guide and through a cold beryllium–bismuth filter on their way to the CNPGAA station.The neutron beam is free of reactor gamma-rays and fast neutrons. The neutron flux at the sample position is 1.5 3 108 cm22 s21 (thermal equivalent). Prompt gamma-rays are measured by a high purity germanium detector (26% relative efficiency, 1.9 keV resolution), which is surrounded by a bismuth germanate Compton shield (Fig. 1). Gamma-ray spectra up to 10 MeV are collected using Canberra Nuclear Data Software on a VAXstation 3100.The hydrogen sensitivity (counts s21 mg21) for this instrument is a factor of three to four better than that measured for the TNPGAA instrument, and the hydrogen background is a factor of more than 100 lower. Selection and Preparation of Samples Although it is not always possible to choose the size and shape of a sample, as when analyzing an intact turbine blade or silicon disk, certain guidelines exist when a choice is possible. If analysis of the entire sample is desired, the sample must be smaller than the collimated neutron beam.For the CNPGAA instrument, sample diameter is also limited by presence of an upper neutron guide, which is 30 mm above the center of the CNPGAA neutron beam. Because of this, samples larger than 60 mm in diameter may be analyzed only on the outer edges. In order to avoid significant neutron self-shielding and gamma-ray attenuation during irradiation, the sample thickness should preferably not exceed 10 mm, and should be considerably less if the sample contains large amounts of elements which are strong neutron absorbers (such as B, Cd, Sm, Gd).The mass of the sample should be large enough to produce an observable signal for the element(s) of interest, but not so large that the total sample count rate saturates the detector. For the majority of materials analyzed, a relatively thin target with a diameter of 10 mm to 20 mm and a mass of 0.1 to 1 g represents the ideal sample size and geometry.Although PGAA has occasionally been used for analysis of thin films, it should be remembered that the technique measures total hydrogen. Samples analyzed should be homogeneous whenever possible, in order to simplify corrections for neutron self-shielding, scattering and gamma-ray attenuation. If a thin layer on a substrate is analyzed for hydrogen, the substrate should preferably consist of a material which does not significantly absorb neutrons and does not contain hydrogen or any other element which emits a gamma-ray near 2223 keV.Sample preparation is minimal. In order to avoid irregular sample geometries, powdered samples are usually made into cylindrical pellets using a hydraulic press and a 12 mm diameter die. Powders containing large quantities of elements which strongly absorb neutrons may be diluted with graphite if desired to avoid high gamma-ray count rates and large self-shielding corrections.If the surface of the sample contains grease or other contamination, it should be cleaned with ethanol or another suitable solvent prior to analysis. Small samples (those weighing < 2 g) are usually sealed into bags of FEP Teflon in a clean hood. Since the hydrogen content of this Teflon is negligible, this material is ideal for packaging samples for hydrogen analysis. Preparation of Standards and Blanks Accurate measurement of element mass fractions by PGAA requires precise determination of element sensitivities.Since element sensitivities are dependent upon the matrix and geometry of the material being analyzed, it is imperative that matrix and geometry of standards and samples be as closely matched as possible. This is especially important when analyzing highly hydrogenous samples, as element sensitivities are greatly affected by neutron scattering by hydrogen. If samples of widely differing hydrogen mass fraction are being analyzed, it may be necessary to prepare multiple standards in order to calibrate the analytical signal as a function of hydrogen mass fraction (see discussion below). As an alternative to measuring the absolute mass fraction of hydrogen in a sample, it is sometimes sufficient to measure the ratio of the mass fraction of hydrogen to that of another (comparator) element in the sample.In this case it may be necessary to simply measure a standard containing both hydrogen and the comparator element, or even to perform the analysis without standards, using tabulated values of sensitivity ratios.The use of ratio measurements to eliminate measurement errors in PGAA is discussed below. When determining hydrogen at trace levels, it is necessary to measure a blank. The blank should closely match the sample matrix and geometry, but should be free of hydrogen. Measurement of a blank allows correction for background generated from neutron capture by environmental hydrogen ‡ The identification of certain commercial equipment, instruments, or materials does not imply recommendation or endorsement by the National Institute of Standards and Technology.These identifications are made only in order to specify the experimental procedures in adequate detail. Fig. 1 Side view of the CNPGAA spectrometer in the NIST Cold Neutron Research Facility. 36R Analyst, March 1997, Vol. 122(which often varies with sample matrix) and also for background due to spectral interferences.Sample Irradiation and Gamma-ray Counting Targets (i.e., samples and standards) for irradiation are usually suspended in the neutron beam by means of Teflon strings at a 45° angle to both the detector and the neutron beam. For large and irregularly shaped samples, special sample holders and mounting procedures are employed. Samples for determination of trace hydrogen by CNPGAA are often irradiated inside an evacuated magnesium sample box in order to eliminate hydrogen background from neutron capture by water vapor in the air.Irradiation times are dependent upon the quantity of hydrogen in the target, the desired precision of the analysis, and the available beam time. Typical analysis times range from a few minutes to > 48 h. Target count rates are corrected for variations in the neutron flux over the course of an analysis. These variations are monitored by irradiating a Ti foil at regular intervals. For the TNPGAA instrument, the measured flux variation is small (generally < 0.5% on a day-to-day basis). Larger variations have been measured for the CNPGAA instrument.Data reduction and spectral manipulation are accomplished using Nuclear Data spectroscopy applications software, which includes a peak search and the SUM program written at NIST.12 The SUM program allows peak and background channels to be entered manually and is useful for integration of small peaks on high backgrounds.Evaluation and Minimization of Uncertainties Experimental uncertainties are evaluated using guidelines set by the International Organization for Standardization (ISO).13 Uncertainties from individual sources are added in quadrature to obtain the combined standard uncertainty for each measurement. The combined standard uncertainty is then multiplied by a coverage factor (usually 2) to obtain an expanded standard uncertainty corresponding to a confidence interval of approximately 95%.A discussion of sources of uncertainty and their relative contribution to the combined standard uncertainty of PGAA measurements is given below. Methods for minimizing each uncertainty are also described. Counting Uncertainties Counting uncertainties arise from the fact that radioactive decay is a random process, hence any measurement based on observing radiation emitted in nuclear decay is subject to some degree of statistical fluctuation. Because the statistics are straight forward, this uncertainty in any measurement may be accurately determined.Both the PEAK and SUM programs calculate the uncertainties associated with the integrated peaks. Uncertainties arising from gamma-ray counting can be lowered by increasing the counting time, hence long counting times are usually required when peak-to-background ratio is low. When measuring milligram amounts of hydrogen by CNPGAA, relative standard uncertainties of < 1% are obtainable with reasonably short analysis times.For measurement of hydrogen at trace levels, the counting uncertainty often represents the major uncertainty in the analysis, even when long counting times are employed. For example, CNPGAA measurement of 60 mg kg21 hydrogen in a 500 mg sample of a titanium alloy counted for 10 h yielded a relative standard counting uncertainty of 11%. The evaluation of counting uncertainties is discussed in detail by Knoll.14 Neutron Flux Variation and Sample Positioning The standard uncertainty associated with variations in the flux of the neutron beam and positioning of the sample may be determined by repeated measurement of a titanium foil whose geometry is comparable to that of the sample.The relative standard uncertainty is approximately 1% for measurements by CNPGAA, and @0.5% for TNPGAA measurements. Gamma-ray Background Neutron capture by hydrogen in the shielding and in the atmosphere gives rise to hydrogen background gamma rays in both the TNPGAA and CNPGAA facilities.The magnitude of this background is dependent upon the matrix of the sample. If the sample contains nuclides which significantly scatter neutrons, additional neutrons may be scattered into the shielding and surrounding environment, thus enhancing the hydrogen gamma-ray background. Scattering-induced background enhancement has been observed with both the TNPGAA15 and CNPGAA instruments. For this reason, the hydrogen gammaray background is determined from irradiation of a blank whenever possible.Although hydrogen itself has by far the largest scattering cross section (bound ss = 80 b) of any element,§ significant enhancement of background H, Al, and Pb gamma-rays has been observed with neutron scattering by other elements. Irradiation of graphite and beryllium targets by CNPGAA yielded up to a factor of 10 enhancement in the hydrogen background. Measurement errors arising from background enhancement may be minimized by calibrating background count rates as a function of target total neutron scattering cross-section.15 In the future, background enhancement in CNPGAA may be minimized by lining the inside of the sample chamber with neutron absorbing 6Li-glass.Subtraction of a hydrogen background count rate from the sample count rate does not introduce significant additional measurement uncertainty unless the background count rate is a significant fraction of the total hydrogen count rate.For the CNPGAA instrument, in the absence of significant background enhancement, a 12 h background determination yields a relative standard uncertainty in background hydrogen count rate of ~ 10%. If the sample count rate is @5 times the background count rate, i.e., if the sample contains less than 50 mg of hydrogen, the relative standard uncertainty of the measurement arising from background subtraction is > 1%. Neutron Self-shielding and Scattering Perhaps the greatest potential source of error in PGAA measurements arises from the interaction of neutrons with the target matrix. If the target matrix contains large quantities of strongly absorbing nuclides, self-shielding occurs, resulting in a decrease in the average neutron flux within the sample, and a corresponding decrease in element sensitivities.If the target composition, density and appropriate absorption cross-sections are known, and if the sample size and shape are well defined (e.g., a sphere, a slab, or a cylinder), this self-shielding effect may be corrected for using well known general absorption laws.16 Corrections for neutron self-shielding are larger for CNPGAA than for TNPGAA because of larger neutron capture cross-sections. Shielding corrections may be significant even for very thin ( < 1 mm) targets if appreciable amounts of B, Cd, Gd or Sm are present.For a 0.8 mm thick geological sample containing mass fraction of Å 1% m/m of boron the calculated self-shielding correction factor is about 5% for TNPGAA and 13% for CNPGAA. Neutron scattering is a significant source of error in PGAA when hydrogenous samples are measured.The effects of neutron scattering by hydrogen on element sensitivities in TNPGAA have been studied extensively.17–20 Measurements of 2 mm thick, disk-shaped targets have shown that sensitivities § Average neutron scattering cross sections for most elements are < 10b. Analyst, March 1997, Vol. 122 37Rfor most elements increase by 1.8% per 1% increase in H mass fraction. For a series of hydrogenous disks approximately 12 mm in diameter, and ranging in thickness from 1 to 12 mm, sensitivities are enhanced relative to sensitivities for nonhydrogenous disks. These enhancements are largely the result of elastic scattering which changes the mean free path of the neutron within the target and therefore changes the probability for neutron absorption. Experiments have shown that the effects of elastic scattering are minimized when spherical targets are measured.The effects of neutron scattering on CNPGAA sensitivities are different and more pronounced than those observed for TNPGAA.21,22 Sensitivities for CNPGAA decrease markedly with both increasing H mass fraction and target thickness. Studies indicate that interaction of cold neutrons with a warm target results in an increase in the average energy of the neutron beam. The effect is observed for both cylindrical and spherical targets. In TNPGAA, uncertainties arising from neutron scattering by hydrogen are minimized either by careful matching of geometry and matrix of the standards and samples or by calibrating element sensitivities as a function of H mass fraction. The latter procedure was used in the analysis of hydrogen and boron in 40 food and mineral supplements with hydrogen mass fractions ranging from 0.83 to 9.82%.23 Mixtures of D2O–H2O–B(OH)3 and pellets prepared from homogeneous mixtures of graphite, mannitol, and boron carbide were used to calibrate H and B sensitivities.Both H and B sensitivities showed good linear correlation with hydrogen mass fraction over the range of 0 to 11%. Relative standard uncertainties associated with calibration of TNPGAA sensitivities have been evaluated at @1%, based on calibration curve data. Such procedures give less satisfactory results for CNPGAA. Because scattering effects are more pronounced for CNPGAA, even small differences in sample scattering power and geometry can result in significant changes in element sensitivities.For this reason, CNPGAA has not often been used to measure element mass fractions in highly hydrogenous samples. A more common approach in CNPGAA has been to use a method of internal standards by measuring ratios of element mass fractions. Recent studies indicate that measurement errors arising from the effects of neutron self-shielding and scattering are largely eliminated when ratios are measured.24,25 This method has been used in the determination of hydrogen in titanium alloys.5 The ratio of hydrogen mass fraction to titanium mass fraction is measured and, since the titanium content of the alloy is known, the hydrogen mass fraction is easily calculated.A future approach to eliminating errors due to neutron scattering in CNPGAA may be to cool the target in order to bring it into thermal equilibrium with the neutron beam.22 The effects of neutron scattering on CNPGAA sensitivities in the cold target would then be similar to those observed for targets in TNPGAA, and could be mitigated by the use of spherical targets.Gamma-ray Attenuation Attenuation of the gamma-ray signal prior to detection may result in significant measurement error. The degree of attenuation is proportional to the thickness, density, and mass number (Z) of the attenuating medium, and is inversely proportional to gamma-ray energy.If the signal is collimated, the collimator should be composed of a high Z, high density material (e.g., lead or tungsten) in order to minimize the number of attenuated gamma-rays reaching the detector. Gamma rays emitted from a target are also subjected to a certain amount of attenuation by the emitting material. If the target is very thick, is of high density, or contains large quantities of high Z elements, this selfattenuation may be significant, particularly for gamma-rays of low energy (7500 keV).Errors arising from gamma-ray attenuation are not eliminated when element ratios are measured, especially when count rates given by gamma-rays of vastly different energies are compared.25 As an example, consider the measurement of hydrogen in a 1 cm thick cylinder of titanium mounted 45° to the detector. Gamma-ray attenuation within the sample results not only in a 12% decrease in the 2223 keV hydrogen signal, but also in a 3% decrease in the ratio of the 1381 keV Ti to the 2223 keV hydrogen signal.Several approaches may be used to minimize measurement errors arising from gamma-ray attenuation. Errors due to gamma-ray self-attenuation may be minimized by matching matrix and geometry of samples and standards. Alternatively, if the geometry and composition of the sample are known, selfattenuation of gamma rays may be corrected using derived equations.26 If one of the elements present in the sample matrix emits a spectrum of gamma-ray lines covering a range of energies, a third option is to calibrate gamma attenuation as a function of energy by comparing the relative intensities of these lines with those measured in a thin foil of the same material (where self-attenuation is minimal).Sample Heterogeneity Sample heterogeneity can introduce errors into PGAA measurements, regardless of whether absolute mass fractions or ratios are measured.25 Element sensitivities may be dependent on the distribution of the elements in a target.Corrections for neutron self-shielding or gamma-ray attenuation are difficult, if not impossible, for heterogeneous targets, since equations derived for these corrections assume a homogeneous composition for the sample. One method of minimizing measurement errors due to sample heterogeneity is to analyze the sample in several different orientations, then report the average of the measurements. Spectral Interferences Because of the complexity of prompt gamma-ray spectra, spectral interferences often constitute a major source of error in PGAA measurements.Current tabulations of prompt gammarays list few interferences for the H 2223.3 keV gamma-ray. Lone et al.27 list gamma-rays by Ba (2220.0 keV), Os (2223.3 keV), and Xe (2225.2 keV) in the H energy region, while Tuli’s compilation28 includes Gd (2220.3 keV), Ge (2223.0 keV), Nd (2223.5 keV), Ru (2223.8 keV), Sn (2224.7 keV), and Ba (2224.8 keV).However, these lists are incomplete and subject to error. Spectra of some elements have many small peaks which are not listed in these compilations. When measuring hydrogen at very low levels by CNPGAA, even tiny peaks in the hydrogen region can become major sources of error. For this reason, accurate determination of hydrogen at low levels in many materials is possible only after rigorous examination of spectra of samples and carefully prepared blanks.Although materials are available that have been certified as hydrogen-free by another analytical technique, it is often necessary to prepare blanks by removal of hydrogen from these materials. Many metals (e.g., Ti, Zr) readily lose hydrogen when heated in a vacuum. More volatile materials or those that decompose upon heating may require freeze-drying to remove residual moisture. It is desirable to prepare more than one blank in order to check the material for complete loss of hydrogen.If analysis of a blank reveals a peak in the hydrogen region, the relative intensity of this line to that of a known peak for an element present in the material should be calculated. If this ratio is the same for each blank analyzed, the peak is likely to be an interference and not due to residual hydrogen in the blank. CNPGAA analyses of samples and blanks containing Co and Cu have revealed previously unlisted peaks for these elements 38R Analyst, March 1997, Vol. 122near 2223 keV. A small peak at 2219 keV interfered with measurement of hydrogen in titanium, however, this peak was reduced with the use of Compton suppression.5 The magnitude of error introduced by spectral interferences in hydrogen measurements depends upon the mass fraction of the interfering element, the intensity of the interfering gamma-ray peak, and the mass fraction of hydrogen in the sample. For example, failure to correct for the titanium interference peak in the analysis of a 0.2 g sample of titanium containing hydrogen at 215 mg kg21 yields a value for hydrogen that is 25 to 30% too high.Other Sources of Uncertainty Measurements by PGAA are affected by many of the same sources of uncertainty that affect measurements by delayed neutron activation analysis (NAA). These include uncertainties arising from pulse pileup correction and sample weighing, which are negligible for most PGAA measurements. For hygroscopic samples, absorption of water represents an additional source of error, which can generally be eliminated by storing samples in a desiccator prior to analysis.A discussion of sources of error in NAA is given elsewhere.29 Expanded Standard Uncertainties of Measurements For determination of hydrogen at percentage levels in biological and geological samples by TNPGAA, expanded standard uncertainties are often less than 5% with counting times of < 24 h. Determination of large amounts of hydrogen by CNPGAA usually yields higher expanded uncertainties than TNPGAA, except where it is possible to measure the ratio of hydrogen to another element. Higher expanded uncertainties are often obtained when neutron self-shielding, gamma-ray attenuation, or sample heterogeneity are significant, or when measuring hydrogen at very low levels.Measurement of 50 to 200 mg kg21 in titanium alloys resulted in expanded uncertainties of 10 to 30% with analysis times of 5 h or longer.5 Future improvements to the CNPGAA instrument are expected to result in smaller uncertainties for measurement of trace hydrogen.Limits of Detection of Hydrogen Limits of detection (LOD) for hydrogen in selected matrices by thermal and cold neutron PGAA are given in Table 1. The LOD for a 1 g sample counted for 24 h is calculated as described by Currie,30 using the equation LOD = 4.65 (Rb/t)1/2/S where Rb is the background counting rate (counts s21), t is the duration of the count (s), and S is the hydrogen sensitivity in counts s21 mg21.The hydrogen signals from coal fly ash, bovine liver, and quartz are presumed to be free of spectral interferences. Because of the presence of a Ti interference peak at 2219 keV, the CNPGAA LOD for hydrogen in Ti-6Al-4V is higher than in the other three matrices. Applications CNPGAA has been used to measure hydrogen in a wide variety of materials (see Tables 2 and 3). Considerable emphasis has been placed upon the ability of the technique to measure hydrogen in metals.Hydrogen mass fractions of 50 to 750 mg kg21 have been measured in titanium alloy jet engine compressor blades in order to determine whether hydrogen embrittlement may be responsible for blade failure.5,31 The technique has also been used to analyze a series of titanium alloy standards containing 100 to 20 000 mg kg21 hydrogen,5 and to measure < 100 mg kg21 hydrogen in samples of nanocrystalline Fe, Pd, and Cu. Attempts to measure hydrogen in semiconductor materials have met with limited success.Less than 10 mg kg21 hydrogen has been measured in hydrothermal quartz, while hydrogen mass fractions of 50 to 80 mg kg21 have been measured in crystals of semiconductor grade germanium. 4,32 Other industrially important materials analyzed for H by CNPGAA have included pure and substituted fullerenes (as little as 0.02% m/m H measured),33 solid proton conductors,34 hydrofluorocarbons and solid acid catalysts,35 zeolite catalysts, and samples of substituted lithium niobate.Hydrogen has also been measured in geological materials, including the Allende meteorite,10,36 United States Geological Survey (USGS) rock standards and terrestrial basalts. TNPGAA has proven valuable for the determination of hydrogen (at mass fractions of 0.83 to 12%) and up to 22 other elements in foods and mineral supplements.23,37–39 Analysis of 22 USGS reference standards yielded hydrogen mass fractions from 0.02 to 3.1%.40 TNPGAA has also measured 580 to 920 mg kg21 hydrogen in ground ash from Mt.St. Helens,41 3.7 to 4% m/m hydrogen in bituminous coal standard reference materials,42 and 0.09 to 0.61% m/m hydrogen in construction materials (sand, cement and concrete).43 Future Developments Future developments in the CNPGAA instrument, such as the addition of improved neutron and gamma-ray shielding to the system and continued fine tuning of Compton suppression electronics are expected to result in lower gamma-ray background, higher signal to noise ratio and better H LOD.Such improvements should also result in smaller uncertainties in hydrogen measurements and also shorter analysis times. The replacement of the D2O cold source with a liquid hydrogen source will result in a 5–10 fold increase in the neutron capture rate, reducing the sample size required for measurement. The installation of a neutron focusing lens will make it possible to analyze samples less than a millimeter in diameter without loss in sensitivity as well as to probe larger samples for compositional mapping.The lens, which has been tested at the CNPGAA station, focuses the neutron beam to 0.5 mm, resulting in an increase by a factor of 80 in neutron current density over the unfocused beam.48 Future plans also call for the development of a lens which will bend the neutron beam as well as focus it. The new lens will lower the sample position by more than 40 mm, removing samples from the immediate vicinity of the upper neutron guide.Much work still remains in order to ensure accurate measurement of hydrogen in materials by CNPGAA. Reliable blanks need to be prepared for many materials in order to minimize errors due to spectral interferences. Further studies are also needed with cryogenic samples in order to develop a method for minimizing the effects of neutron scattering on CNPGAA measurements of highly hydrogenous samples.The potential of PGAA for measurement of hydrogen in materials is great and has yet to be fully realized. Research areas which should be more fully exploited in the future include the use of PGAA to study hydrogen embrittlement in a variety of metals, to accurately measure the degree of deuteration of Table 1 LOD for hydrogen in selected matrices for thermal and cold neutron PGAA Limit of detection/mg kg21 Coal Fly Ash Bovine Liver (SRM 1633) (SRM 1577) Quartz Ti-6Al-4V* Thermal 27 46 nd† nd Cold 8.5 nd 5.0 15 * An alloy of titanium containing Å 6% aluminum and 4% vanadium by mass.† nd, not determined. Analyst, March 1997, Vol. 122 39RTable 2 Some materials analyzed for hydrogen by CNPGAA* Hydrogen content Material CNPGAA Other methods Titanium (alloyed and pure): SRM 354 (unalloyed) 224 ± 22 mg kg21a 215 ± 6 mg kg21b SRM 352c (unalloyed) 58 ± 20 mg kg21a 49 ± 0.9 mg kg21b Ti-6Al-4V standards 117 ± 39 mg kg21c 114 ± 60 mg kg21d 164 ± 138 mg kg21c 175 ± 115 mg kg21d Nanocrystalline metals: Fe 0.250 ± 0.078% m/ma 0.214 ± 0.040% m/ma Pd 0.0080 ± 0.0086% m/ma Cu 0.0031 ± 0.013% m/ma Quartz: 6 ± 12 mg kg21e 5.4 mg kg21f 10 ± 12 mg kg21e 12 mg kg21f @5 mg kg21e 4.9 mg kg21f @6 mg kg21e 4.6 mg kg21f Semiconductor-grade germanium: 80 ± 55 mg kg21e 75 ± 50 mg kg21e 50 ± 40 mg kg21e Pure and substituted fullerenes: C60 C:H (atom ratio) = 180 ± 30g K3C60 C:H (atom ratio) = 114 ± 11g Rb2.6K0.4C60 C:H (atom ratio) = 68 ± 6g Na2RbC60 C:H (atom ratio) = 180 ± 35g Solid proton conductors: Yb doped SrCeO3 3.1 ± 0.8 mol%h 2.5 mol%i Undoped SrCeO3 0.7 ± 0.1 mol%h 0.5 mol%i Allende meteorite: 160 ± 10 mg kg21j 190 mg kg21, 180 mg kg21k; 110 mg kg21, 150 mg kg21l * All stated uncertainties are expanded standard uncertainties as defined in the text.a Previously unpublished results. b NIST certified values determined by hot extraction of hydrogen at 1400 °C and vacuum fusion of the sample at 1950 °C.c Ref. 5. d Determined using a LECO inert gas fusion analyzer, ref. 5. e Ref. 4. f Determined by IR, ref. 4. g Ref. 33. h Ref. 34. i Determined by thermal gravimetry, ref. 34. j Ref. 36. k Ref. 44. l Ref. 45. Table 3 Some materials analyzed for hydrogen by TNPGAA* Hydrogen content Material TNPGAA Other methods Biological materials: SRM 1549 (Milk powder) 5.9 ± 0.8% m/ma SRM 1570 (Spinach) 5.45 ± 0.08% m/mb 5.57% m/md SRM 1571 (Orchard Leaves) 5.84 ± 0.08% m/mb 5.84 ± 0.26% m/md SRM 1577 (Bovine Liver) 6.83 ± 0.08% m/mc 6.97 ± 0.16% m/md Ground beef 9.89 ± 0.12% m/ma Whole milk (fluid) 10.77 ± 0.09% m/ma Potato chips 7.32 ± 0.11% m/ma Corn flakes 6.14 ± 0.07% m/ma USGS reference standards: GSP-1 785 ± 40 mg k21e 680 ± 170 mg kg21f STM-1 0.18 ± 0.02% m/me 0.19% m/mf AGV-1 0.18 ± 0.02% m/me 0.20 ± 0.04% m/mf PCC-1 0.58 ± 0.01% m/me 0.57 ± 0.03% m/mf Volcanic ash (Mt.St. Helens): 580 ± 90 mg kg21g 900 ± 100 mg kg21g 920 ± 70 mg kg21g 870 ± 40 mg kg21g Construction materials: Sand 0.28 ± 0.01% m/me Cement 0.088 ± 0.014% m/me Concrete 0.61 ± 0.01% m/me * All stated uncertainties are expanded standard uncertainties as defined in the text.a Ref. 37. b Ref. 38. c Ref. 23. d Ref. 46. e Ref. 40. f Ref. 47. g Ref. 41. h Ref. 43. 40R Analyst, March 1997, Vol. 122materials, and to measure absorption of hydrogen by materials to determine their potential for use in hydrogen fuel cells. The cooperation of the NIST reactor staff during the PGAA measurements is gratefully acknowledged.R. Lindstrom, E. Mackey and D. Anderson are thanked for their advice and interest in this work. References 1 Meyn, D. A., Metall. Trans., 1974, 5, 2405. 2 Chevallier, J., and Aucouturier, M., Ann. Rev. Mater. Sci., 1988, 18, 219. 3 Lindstrom, R. M., Paul, R. L., Vincent, D. H., and Greenberg, R. R., J. Radioanal. Nucl. Chem., 1994, 180, 271. 4 Paul, R. L., and Lindstrom, R. M., Diagnostic Techniques for Semiconductor Processing, MRS Symposium Proceedings Volume 324, ed.Glembocki, O. J., Pang, S. W., Pollak, F. H., Crean, G. M., and Larrabee, G., Materials Research Society, Pittsburgh, PA, 1994, pp. 403–408. 5 Paul, R. L., Privett III, H. M., Lindstrom, R. M., Richards, W. J., and Greenberg, R. R., Metall. Mater. Trans. A, 1996, 27A, 3682. 6 Lindstrom, R. M., J. Res. Nat. Inst. Stand., Techn 1993, 98, 127. 7 Failey, M. P., Anderson, D. L., Zoller, W. H., Gordon, G. E., and Lindstrom, R.M., Anal. Chem., 1979, 51, 2209. 8 Anderson, D. L., Failey, M. P., Zoller, W. H., Walters, W. B., Gordon, G. E., and Lindstrom, R. M., J. Radioanal. Chem., 1981, 63, 97. 9 Lindstrom, R. M., Zeisler, R., Vincent, D. H., Greenberg, R. R., Stone, C. A., Mackey, E. A. Anderson, D. L., and Clark, D. D., J. Radioanal. Nucl. Chem., 1993, 167, 121. 10 Paul, R. L., Lindstrom, R. M., and Vincent, D. H., J. Radioanal. Nucl. Chem., 1994, 180, 263. 11 Mackey, E. A., Anderson, D. L., Chen, H., Downing, R.G., Greenberg, R. R., Lamaze, G. P., Lindstrom, R. M., Mildner, D. F. R., and Paul, R. L., J. Radioanal. Nucl. Chem., 1996, 203(2), 411. 12 Lindstrom, R. M., Biol. Trace Elem. Res., 1994, 43–45, 597. 13 International Standards Organisation, Guide to the Expression of Uncertainty in Measurement, 1st edn. ISO, Geneva, 1993. 14 Knoll, G. F., Radiation Detection and Measurement, Wiley, New York, 1989, pp. 65–99. 15 Anderson, D. L., and Mackey, E. A., J. Radioanal. Nucl. Chem., 1993, 167, 145. 16 Fleming, R. F., Int. Appl. Radiat. Isot., 1982, 33, 1263. 17 Copley, J. R. D., and Stone, C. A., Nucl. Instrum. Meth., 1989, A281, 593. 18 Mackey, E. A., Gordon, G. E., Lindstrom, R. M., and Anderson, D. L., Anal. Chem., 1991, 63, 288. 19 Mackey, E. A., Gordon, G. E., Lindstrom, R. M., and Anderson, D. L., Anal. Chem., 1992, 64, 2366. 20 Mackey, E. A., and Copley, J. R. D., J. Radioanal. Nucl. Chem., 1993, 167, 127. 21 Paul, R. L., and Mackey, E. A., J. Radioanal. Nucl. Chem., 1994, 181, 321. 22 Mackey, E. A., Biol. Trace Element Res., 1994, 43–45, 103. 23 Anderson, D. L., Cunningham, W. C., and Mackey, E. A., Biological Trace Element Research, 1990, 27, 616. 24 Lindstrom, R. M., Fleming, R. F., Paul, R. L., and Mackey, E. A., Proceedings of the International k0 Users Workshop, Gent, 1992, pp. 121–124. 25 Paul, R. L., J. Radioanal. Nucl. Chem., 1995, 191, 245. 26 Debertin, K., and Helmer, R. G., Gamma- and X-Ray Spectrometry with Semiconductor Detectors, Elsevier, Amsterdam, 1988. 27 Lone, M. A., Leavitt, R. A., and Harrison, D. A., Atomic Data and Nuclear Data Tables, 1981, 26, 511. 28 Tuli, J. K., in Prompt Gamma Neutron Activation Analysis, ed. Alfassi, Z. B., and Chung, C., CRC Press, Boca Raton, FL, 1995, pp. 177–223. 29 Blaauw, M., Ammerlaan, M. J. J., and Bode, P., Appl. Radiat. Isotop., 1993, 44, 547. 30 Currie, L. A., Anal. Chem., 1968, 40, 586. 31 Paul, R. L., and Lindstrom, R. M., Review of Progress in Quantitative Nondestructive Evaluation, vol. 13, ed. Thompson, D. O., and Chimenti, D. E., Plenum Press, New York, 1994, pp. 1619–1624. 32 Paul, R. L., and Lindstrom, R. M., International Workshop on Semiconductor Characterization: Present Status and Future Needs, ed. Bullis, W. M., Seiler, D. G., and Diebold, A. C., AIP Press, Woodbury, NY, 1996, pp. 342–345. 33 Neumann, D. A., Copley, J. R. D., Reznik, D., Kamitakahara, W. A., Rush, J. J., Paul, R. L., and Lindstrom, R. M., J. Phys. Chem. Solids, 1993, 54, 1699. 34 Krug, F., Schober, T., Paul, R., and Springer, T., Solid State Ionics, 1995, 77, 185. 35 Crawford, M. K., Corbin, D. R., and Vernooy, P. D., Trans. Am. Nucl. Soc., 1994, 71, 168. 36 Paul, R. L., Lindstrom, R. M., and Vincent, D. H., J. Radioanal. Nucl. Chem., 1995, 190, 181. 37 Anderson, D. L., Cunningham, W. C., and Lindstrom, T. R., J. Food Comp. Anal., 1994, 7, 59. 38 Anderson, D. L., Cunningham, W. C., and Alvarez, G. H., J. Radioanal. Nucl. Chem., 1993, 167, 139. 39 Anderson, D. L., and Cunningham, W. C., Trans. Am. Nucl. Soc., 1994, 71, 21. 40 Anderson, D. L., Sun, Y., Failey, M. P., and Zoller, W. H., Geostand. News., 1985, 9, 219. 41 Anderson, D. L., Phelan, J. M., Vossler, T., and Zoller, W. H., J. Radioanal. Chem., 1982, 71, 47. 42 Germani, M. S., Gokmen, I., Sigleo, A. C., Kowalczyk, G. S., Ohmez, I., Small, A., Anderson, D. L., Failey, M. P., Gulovali, M. C., Choquette, C. E., Lepel, E. A., Gordon, G. E., and Zoller, W. H., Anal. Chem., 1980, 52, 240. 43 Anderson, D. L., Gordon, G. E., and Lepel, E. A., Can. J. Chem., 1983, 61, 724. 44 Scoon, J. H., in The Allende Meteorite Reference Sample, ed. Jarosewich, E., Clarke, Jr. R. S., and Barrows, J. N., Smithsonian Contrib. Earth Sci., 1987, 27, 39. 45 Willis, J. P., in: The Allende Meteorite Reference Sample, ed. Jarosewich, E., Clarke, Jr. R. S., and Barrows, J. N., Smithsonian Contrib. Earth Sci., 1987, 27, 39. 46 Gladney, E. S., O’Malley, B. T., Roelandts, I., and Gills, T. E., National Bureau of Standards Special Publication NBS SP 260-111, 1987. 47 Gladney, E. S., Burns, C. E., and Roelandts, I., Geostand. News., 1983, 7, 3. 48 Chen, H., Sharov, V. A., Mildner, D. F. R., Downing, R. G., Paul, R. L., Lindstrom, R. M., Zeissler, C. J., and Xiao, Q.-F., Nucl. Instrum. Meth. B, 1995, 95, 107. Paper 6/06419A Received September 17, 1996 Accepted December 4, 1996 Analyst, March 1997, Vol. 122 41R

ISSN:0003-2654    

DOI:10.1039/a606419a

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Ozone Degradation of Residual Carbon in Biological Samples Using Microwave Irradiation Wenchun Jiang†, Stuart J. Chalk‡, H. M. ‘Skip’ Kingston* Department of Chemistry and Biochemistry, Duquesne University, Pittsburgh, PA 15282, USA In an attempt to produce complete oxidation of a biological matrix, bovine liver, ozone was investigated as an additional, potentially non-contaminating, oxidizing reagent after nitric acid digestion. Experiments were carried out to determine the decomposition efficiency of residual carbon species, primarily o-, m- and p-nitrobenzoic acids (NBAs), using ozone.The NBAs were degraded by purging sample digests with ozone, while heating the solutions with microwave energy at atmospheric pressure. The effects of the degradation temperature and solution pH on the ozonation of NBAs were determined. Solid phase extraction (C18) was used to extract NBAs from the acid digestate solutions prior to HPLC analysis. Reversed phase HPLC was used to determine NBA concentrations in digest solutions.After 2.5 h of purging ozone at 80 °C, 33.65 ± 3.80% o-NBA degradation, 19.39 ± 1.74% m-NBA degradation, and 26.47 ± 3.36% p-NBA degradation were obtained. Keywords: Ozone; biological samples; microwave digestion; residual carbon; nitrobenzoic acids Today, microwave energy has been widely applied to sample digestion for the analysis of biological, geological, and environmental materials. Microwave enhanced sample preparation techniques dramatically speed up sample dissolution and decomposition, but do not eliminate organic carbon residues due to the incomplete oxidation of organic compounds.Research on microwave sample decomposition has been based on comparison and optimization of conventional acid digestion procedures using microwave energy in open or closed vessels. The completeness of these techniques can be evaluated by considering the residual carbon content, recovery of analytes, and digestion time.Since organic carbon residues interfere with detection by some instrumental techniques, the completeness of decomposition can be a crucial factor for accurate analysis. Nitric acid, the most commonly used mineral acid, oxidizes the majority of organic matrix components with the primary exception of aromatic ring structures, particularly nitrobenzoic acids (NBAs).1–7 Pratt et al.1 identified o-, m- and p-NBAs as the major residual organic compounds after nitric acid digestion of bovine liver in sealed perfluoroalkoxy (PFA) vessels at a maximum temperature of 180 °C and a sustained maximum pressure of 8 atm (1 atm = 101.325 kPa).A number of researchers have identified that residues from incomplete nitric acid oxidation act as interferences with some analytical techniques such as polarography, voltammetry, and ETAAS.1,4,5,8,9 In spectroscopic techniques with high energy interfaces, such as ICP-OES, ICP-MS, and FAAS, small amounts of residual carbon is not a significant limitation.To take account of residual carbon Pratt et al. also successfully demonstrated that refluxing with perchloric acid (HClO4) can decompose the residual organic products remaining after nitric acid digestion.1 However, a drawback of using perchloric acid is the possible formation of explosive products if safety precautions are not taken.10 Thus, an alternative oxidizing reagent is desirable to completely and safely decompose organic carbon residues.The goal is the evaluation of ozone as an alternative decomposition reagent for analytical sample preparation, especially coupled to microwave decompositions. This work extends the studies of Pratt et al.1 and W�urfels et al.4,5 who identified that the major residual organic species in nitric acid digests of bovine liver are o-, m- and p-NBAs, arising from the nitration of aromatic amino acid constituents of original sample protein. This research focuses on the addition of ozone to aqueous acidic solutions to aid the microwave degradation of NBAs.The application of ozone to decompose o-, m- and p-NBAs also allows an investigation into the oxidation processes needed to efficiently destroy or minimize organic carbon residues. Experimental Reagents and Standards NIST SRM 1577a, bovine liver, was used as the biological matrix in this study. Samples were vacuum-dried at room temperature for 24 h in accordance with the instructions on the certificate of analysis.Compressed air (reagent grade) was used as the source of oxygen to produce ozone. Ozone concentrations were measured by the decolorization of potassium indigo trisulfonate (Aldrich, Milwaukee, WI, USA) solution. Doubly distilled 18 MW NanoPure water (Barnstead, Dubuque, IA, USA) was used to prepare all solutions. 2-Nitrobenzoic acid (o-NBA) (96%, Aldrich), 3-nitrobenzoic acid (m-NBA) (99%, Aldrich) and 4-nitrobenzoic acid (p-NBA) (99%, Eastman Kodak, Rochester, NY, USA) were used to prepare aqueous samples for the NBA degradation studies.All other reagents were certified ACS reagent grade. The HPLC eluent used was 5% methanol (HPLC grade)– 15% acetonitrile (HPLC grade)–80% H2O (18 MW) (by volume before mixing), to which 0.05% trifluoroacetic acid was added to acidify the solution to a pH of 2–2.3. Apparatus The ozonator used for ozone production in this work was a Model ss-150 (Pillar Technologies, Hartland, WI, USA).It was purchased from Radiation Disposal Systems (Charlotte, NC, USA) and was redesigned to a Model ss-300 (Select Industrial Systems, Waukesha, WI, USA) to double the ozone production rate. Fig. 1 shows a schematic representation of the ozonation apparatus set-up. The reaction vessel was a Teflon microwave vessel, 28 cm tall with an id of 3.8 cm (Prolabo, Paris, France). The total volume was approximately 320 ml. The gas stream was introduced into the reaction vessel via a 25 mm pore size † Present address: Caelum research Corporation, 7505 Standish Place, Rockville, MD 20855, USA.‡ Present address: Department of Natural Sciences, University of North Florida, Jacksonville, FL 32224, USA. Analyst, March 1997, Vol. 122 (211–215) 211Teflon gas sparger (Omnifit, Toms River, NJ, USA) located close to the bottom of the vessel. Ultraviolet and visible spectra were obtained using a Cary 1E UV/VIS spectrophotometer (Varian Australia, Springvale Victoria, Australia).The instrument was interfaced to a Grid 3861S-25 PC computer (Varian Australia) running Varian Cary 13E software Version 2. The pH of the solutions was calculated or measured using a Model 710A pH meter (Orion, Boston, MA, USA) equipped with a pH combination electrode (Orion, part No. 91–56). The atmospheric pressure microwave unit used in this research was a Maxidigest MX350 (Prolabo). A gas bulb thermometer (Prolabo) and a PC computer (Dell 450/MX, Austin, TX, USA) running MEGAL 500 software (Prolabo) were used to control microwave power based on the measurement of the temperature. The closed vessel microwave unit used for the dissolution of bovine liver was an MLS 1200 MEGA (Milestone, Sorisole, Italy).High pressure vessels were used for these digestions (Milestone). HPLC analysis was performed using a Model M-6000A (Waters Associates, Milford, MA) coupled to a Model 440 absorbance detector (Waters Associates) and an HP 3396A integrator (Hewlett-Packard, Avondale, PA).The reversed phase column used was a Vydac Equivalent C18 column (nonendcapped, 25 cm 3 4.6 mm, 5 mm particle size) (Whatman, Clifton, NJ, USA). Solid phase extraction was performed on a Mega Bond Elut octadecylsilyl (C18) cartridge (endcapped) (ChromTech, Apple Valley, MN, USA). A Bond Elut adapter (ChromTech) and poly(propylene)polyethylene syringe (Aldrich) were used to push solvents through the cartridge. Procedures The ozone transfer rate into water and the ozone decomposition rate in water were initially evaluated.Ozonated air was purged through 50 ml of doubly distilled water in the Teflon vessel for a certain period of time with different flow rates (110, 190 and 328 ml min21). The concentration of the ozone solutions were determined immediately using the Indigo Blue method.11 The ozone decomposition rate was determined measuring the decrease of the ozone absorbance at 260 nm over a period of 1 h. In order to study the degradation of o-, m- and p-NBAs at different pHs, solutions of the NBAs were made up in different concentration nitric acid solutions and water.Ozone was then purged through these solutions for 90 min at 80 or 115 °C, the solutions were cooled, made up to volume in 100 ml calibrated flasks and analyzed using solid phase extraction (SPE) and HPLC. Bovine liver samples were digested using a closed vessel high pressure microwave system. Each batch contained four samples of bovine liver ( ~ 0.25 g each) and two blanks in 10 ml of concentrated nitric acid using the high pressure vessels.Three of the samples were put in regular sample vessels and the remaining one was put in the monitor vessel to measure the pressure and temperature change during the digestion. The microwave energy was programmed to heat at 600 W until reaching 180 °C (in about 2 min), and the temperature was kept at 180 °C for about 8–9 min. The samples and rinses were transferred to calibrated flasks, diluted to 50 ml, and were directly used for the atmospheric microwave vessel ozonation study.o-, m- and p-NBA solutions were also prepared using a flow rate of 328 ml min21 to study the mechanism of the ozonation. After purging ozone (1–2 h) at a temperature of 80 °C or higher, the solutions were transferred and diluted to 100 ml calibrated flasks before SPE and HPLC analysis. The C18 cartridges used for SPE were conditioned by passing 5 ml of methanol through them, followed by at least 10 ml of doubly distilled water prior to extraction of an aqueous sample.An aliquot of 0.5 ml of the sample solution was passed through the cartridge, followed by a wash with 5 ml of pH 2 aqueous ophosphoric acid and finally eluted with 2 ml of methanol. Results and Discussion Ozone Transfer Rate and Decomposition Rate in Solution By varying the ozone gas flow rate, it was shown that higher gas flow rates produced greater ozone transfer rates into water.Therefore, a 328 ml min21 gas flow rate was chosen as the ozone purging flow rate for subsequent experiments. At this flow rate, the ozone generator produced 445 mg of ozone per hour. This ensured a fast ozone mass transfer and a high ozone concentration without violent bubbling of the samples, minimizing any loss of solution. The saturated ozone solution that was produced decomposed by 50% in 1 h once the gas flow was turned off.At the flow rate of 328 ml min21, pure water can be saturated with ozone in less than 10 min, which indicated that the dissolution rate of ozone was much faster than the decomposition rate of ozone under the experimental conditions; therefore, a saturated ozone solution could be maintained. SPE Sample Clean-up The efficiency of the extraction of NBAs by the C18 solid phase cartridges was investigated at the level of NBAs expected in the digests. Table 1 shows the SPE cartridge extraction efficiency for a 1 ml aliquot of 66.7 ppm of each of o-, m-, p-NBA solution (200 mg total).The efficiency of extraction for each NBA was excellent (99–103%). Replicate extractions on the same column and on different columns were statistically indistinguishable. Owing to the levels of NBAs being well below the capacity of the cartidges (50 mg), it was anticipated that this efficiency would also be acheivable on real sample digests even though other organic species may be present.HPLC of Nitrobenzoic Acid An interesting observation in the HPLC chromatograms was the changing retention time of the compounds. Continuous injections of bovine liver digestate solution onto the HPLC column caused the retention time of the compounds to increase significantly. This was attributed to the retention of some large organic species extracted by the C18 cartridge which changed the surface of the column packing and thus the chromatography. Fig. 1 Ozone apparatus set-up. 212 Analyst, March 1997, Vol. 122Pumping acetonitrile through the column for approximately 30 min washed out the retained organics and the retention times returned to their original values. The use of a guard column could alleviate this problem. o-, m- and p-NBA Degradation Table 2 shows the effects of pH and temperature on the degradation. Since this research focuses on the application of ozone with nitric acid to improve the oxidation of organic residues, in these experiments nitric acid was used to adjust the pH of the NBA solutions.The mechanism of ozonation of NBAs in acidic solution was expected to be an electrophilic reaction.12,13 However, the presence of nitrosyl (–NO2) and carboxylic acid (–COOH) groups (both electron-withdrawing) on the aromatic ring makes it electron deficient, so it is difficult for ozone to electrophilically attack the ring structure at low pH. The reaction rate observed should therefore be relatively slow in 25% nitric acid13 and concentrated nitric acid solutions as was observed.The pKa values of o-, m- and p-NBA in water at 25 °C are 2.17, 3.49 and 3.77 respectively.14 Increasing the solution pH increases the concentration of deprotonated forms of NBAs. When the NBAs are deprotonated (NBA2), carboxylate (–COO2), an electron-donating group, replaces –COOH and electrophilic attack by ozone is favored. Thus, the increase from pH 2 < 0.1 to 3.2 results in very different oxidation rates. Even at a pH of 0.8, most of the NBA molecules are protonated, but the existence of a small concentration of the deprotonated species provides partial oxidation. At a pH of 3.2, about 85% degradation of NBAs can be achieved in 90 min at 80 °C, suggesting that it would be better to neutralize the digestate solution, with a non-contaminating reagent, such as ammonia, prior to ozonation to enhance the destruction of NBAs.Both of these reagents could be introduced as high purity gases.Temperature also helps the oxidation process. The higher the temperature, the faster the degradation rate. As these experiments were designed for the purging of ozone through solution in an atmospheric pressure microwave system, the temperature of the reaction was limited by the boiling point of the reagent used (118 °C for nitric acid). At this temperature the experimental conditions are rather violent and thus, for safety reasons, a lower temperature (80–90 °C) was used and is recommended.Hydrogen peroxide has been shown to increase the rate of decomposition of nitrogenzoic acids in aqueous solution. The addition of hydrogen peroxide to the reaction mixture was also tested while bubbling ozone through the solution. No additional decomposition was seen when compared to ozone alone. Degradation of Bovine Liver Digestate The estimated total organics in the original SRM 1577a bovine liver sample was ~ 97.4% of the sample dry mass (calculated from ref. 15). After closed vessel nitric acid digestion at 180 °C, HPLC analysis of the digestate solution was performed and is shown in Fig. 2. The major organic residues are o-, m-, p-NBA and an unidentified peak at a retention time of ~ 10 min. The concentrations of the three NBA isomers were calculated and are summarized in Table 3. The average total NBA residue was 2.70% (m/m). Not taking into account the amount of the unidentified compound, the closed vessel nitric acid digestion decomposed ~ 94.7% (m/m) of the organics in the original Table 1 Extraction efficiency of o-, m- and p-nitrobenzoic acids on the same and different C18 columns Extraction* Cartridge I— A B C o-NBA 100 ± 0.9 100 ± 0.7 100 ± 0.6 m-NBA 99 ± 1.4 99 ± 1.5 99 ± 1.2 p-NBA 100 ± 0.7 100 ± 0.8 100 ± 0.8 Cartridge II— A B C o-NBA 102 ± 1.7 100 ± 1.6 100 ± 1.5 m-NBA 102 ± 1.7 100 ± 1.3 100 ± 1.5 p-NBA 103 ± 1.8 100 ± 1.1 101 ± 0.8 Cartridge III— A B C o-NBA 100 ± 1.3 100 ± 0.7 100 ± 1.1 m-NBA 100 ± 1.4 99 ± 1.2 99 ± 1.3 p-NBA 100 ± 0.8 100 ± 0.7 99 ± 0.7 * Mean ± % RSD of three replicate injections.Extractions A, B, and C were consecutive on each column. Table 2 Degradation of o-, m- and p-nitrobenzoic acids in acid solutions Temp- Degradation (% ± %RSD*) erature/ Time/ Solution pH °C min o-NBA m-NBA p-NBA 18 MW water ~ 3.2 80 90 84 ± 3.8 85 ± 3.2 88 ± 3.4 1% HNO3 0.8 80 90 58 ± 0.5 63 ± 3.2 69 ± 0.1 25% HNO3 < 0.1 80 90 23 ± 1.4 15 ± 2.4 19 ± 0.7 Conc.HNO3 < 0.1 80 90 27 ± 1.7 17 ± 1.7 21 ± 0.5 Conc. HNO3 < 0.1 115 90 42 ± 0.4 27 ± 1.6 28 ± 0.4 * %RSDs were calculated from three replicate digestions. Fig. 2 One hour degradation of bovine liver digestate solution with ozone purging at 80 °C using an atmospheric pressure microwave system, (a) before and (b) after. Analyst, March 1997, Vol. 122 213bovine liver sample. In comparison to the 3.01% (m/m) NBA residues found in the Pratt et al.1 work, the amount of total NBA residues in our experiments was 0.31% lower.However, in the work by Pratt et al., the major organic residues were o-, m- and p-NBA; and no unknown peak such as that appearing at ~ 10 min was reported. The decomposition of o-, m- and p-NBA in the digestate solutions was not as much as the decomposition of o-, m- and p-NBA in aqueous solution. The presence of the other residual carbon species that are more easily oxidized and react with ozone preferentially may be one reason for this, but is likely not to be the only contributing factor.Bovine liver biological tissue has significant protein content. Many biological tissues with proteins containing an aromatic side chain will also have significant quantities of o-, m- and p- NBA remaining after digestion. These digestion products are typical of protein containing tissues.1 Since dinitrobenzoic acid has been reportedly found in biological sample digestate solutions as one of the major organic residues,4 experiments were performed to identify whether or not the unknown peak was dinitrobenzoic acid.Fig. 3 shows the HPLC analysis of 2,4-dinitrobenzoic acid (top) and 3,5-dinitrobenzoic acid (bottom). Based on the retention times, neither seemed to be the unidentified peak. However, 2,4-dinitrobenzoic acid had a retention time of 6.4 min, which is close to a small peak (retention time 6.7 min) shown in Fig. 4. Spiking of the digestate solution by adding ~ 20 ppm of 2,4-dinitrobenzoic acid to the solution indicated that this peak is most likely to be 2,4-dinitrobenzoic acid [see Fig. 4 (b)]. The estimated concentration is ~ 0.01% (m/m) of the dry bovine liver sample. In the W�urfels et al. paper,4 5% 2,4-dinitrobenzoic acid was reported to be one of the organic residues after nitric acid digestion of bovine liver sample. However, in the Pratt et al. work, dinitrobenzoic acid was not reported.1 This possibly indicates that the difference of the temperature and pressure during digestion can lead to a shift in the distribution of digestion products.However, it should be noted that in the Pratt et al. experiments, isobutyl methyl ketone (IBMK) was used to extract the NBAs from the digestion solution before HPLC analysis. In our experiment, the NBAs were extracted from the digestate solution by using SPE. It is possible that the extraction method used in these experiments is more efficient, allowing the recovery of the additional unknown peak.The degradation could be improved by the addition of other reagents with the ozone such as peroxide for instance. However the goal of this study was to evaluate the degradation of ozone alone and not to add any other additional reagents to the ozone decomposition. Ozone is produced just prior to its addition in the decomposition system from filtered air or oxygen. It is not a classical reagent and may not contribute significantly to the analytical blank.After an hour of purging ozone at 80 °C through bovine liver digestate solution, 17.08 ± 2.69% o-NBA degradation, 4.78 ± 3.21% m-NBA degradation, and 6.88 ± 2.95% p-NBA degradation result (see Fig. 2). Most of the unknown peaks were degraded by ozone in the first hour of purging. After 2.5 h of purging ozone at 80 °C, 33.65 ± 3.80% o-NBA degradation, 19.39 ± 1.74% m-NBA degradation, and 26.47 ± 3.36% p-NBA degradation were obtained. Conclusions It has been shown in this research that ozone can oxidize o-, mand p-NBA, and has the potential to be used as an additional decomposition and/or finishing reagent.Deprotonated NBAs react faster with ozone than protonated NBAs, strongly suggesting an electrophilic attack of the benzene ring. Thus, increasing the pH of the solution to increase the concentration of the deprotonated NBA significantly improves the ozonation rate. When ozone was applied to closed vessel bovine liver digestate solution, the decomposition of residual o-, m- and p- NBA and other residual carbon was shown.A better understanding of the chemistry of the ozonation of biological sample digestate solutions is an important goal for further research. Since the pH of the solution is essential for the degradation of NBAs, the reaction efficiency could be increased if an appropriate reagent can be found to neutralize the digestate solution (i.e., ammonia) without introducing contamination.This research demonstrates the feasibility of ozone as an Table 3 Concentrations of o-, m- and p-nitrobenzoic acids after closed vessel digestion of bovine liver o-NBA m-NBA p-NBA Total NBA Sample Mass/g (%) (%) (%) (%) 1 0.2178 0.61 0.45 1.59 2.66 2 0.2181 0.59 0.42 1.58 2.59 3 0.2069 0.70 0.50 1.65 2.85 Mean 0.63 0.46 1.61 2.70 s 0.06 0.04 0.04 0.13 %RSD 9.52 8.70 2.48 4.81 Fig. 3 HPLC analysis of 2, 4- and 3, 5-dinitrobenzoic acid under the same conditions as nitrobenzoic acid analysis.Fig. 4 HPLC analysis of (a) bovine liver digestate and (b) a 2,4-dinitrobenzoic acid spike of the same solution. 214 Analyst, March 1997, Vol. 122alternative decomposition method. Additional studies to optimize its use seem appropriate. References 1 Pratt, K. W., Kingston, H. M., MacCrehan, W. A., and Koch, W. F., Anal. Chem., 1988, 60, 2024. 2 Kingston, H. M., and Jassie, L. B., J. Res. Nat. Bur. Stand., 1988, 93, 269. 3 Krushevska, A., Barnes, R. M., Amarasiriwaradena, C.J., Foner, H. A., and Martines, L., J. Anal. At. Spectrom., 1992, 7, 845. 4 W�urfels, M., Jackwerth, E., and St�oppler, M., Anal. Chim. Acta, 1989, 226, 17. 5 W�urfels, M., Jackwerth, E., and St�oppler, M., Anal. Chim. Acta, 1989, 226, 1. 6 Kingston, H. M., Walter, P. J., Chalk, S. J., Lorentzen, E., and Link, D. D., in Microwave Enhanced Chemistry, ed. Kingston, H. M., and Haswell, S., American Chemical Society, Washington, DC, 1997, ch. 3. 7 Walter, P. J., Chalk, S.J., Kingston, H. M., and Lorentzen, E. M. L., in Microwave Enhanced Chemistry, ed. Kingston, H. M., and Haswell, S., American Chemical Society, Washington, DC, 1997, ch. 2. 8 Jackwerth, E., and Gomiscek, S., Pure Appl. Chem., 1984, 56, 479. 9 Kotz, L., Henze, G., Kaiser, G., Pahlke, S., Veber, M., and Tolg, G., Talanta, 1979, 26, 681. 10 Hertz, J., and Pani, R., Fresenius’ Z. Anal. Chem., 1987, 328, 487. 11 Langlais, B., Reckhow, D. A., and Brink, D. R., Ozone in Water Treatment: Application and Engineering, Lewis Publishers, Chelsea, MI, 1991, p. 103. 12 Langlais, B., Reckhow, D. A., and Brink, D. R., Ozone in Water Treatment: Application and Engineering, Lewis Publishers, Chelsea, MI, 1991. 13 Jiang, W., Masters Thesis Duquesne University, 1996. 14 Lange’s Handbook of Chemistry, ed. Dean, J. A., McGraw Hill, New York, NY, 1979, 5. 15 Certificate of Analysis: Standard Reference Material 1577a—Bovine Liver, US Department of Commerce, The National Bureau of Standards, Gaithersburg, MD, 1982.Paper 6/05282G Received July 29, 1996 Accepted November 15, 1996 Analyst, March 1997, Vol. 122 215 Ozone Degradation of Residual Carbon in Biological Samples Using Microwave Irradiation Wenchun Jiang†, Stuart J. Chalk‡, H. M. ‘Skip’ Kingston* Department of Chemistry and Biochemistry, Duquesne University, Pittsburgh, PA 15282, USA In an attempt to produce complete oxidation ofcal matrix, bovine liver, ozone was investigated as an additional, potentially non-contaminating, oxidizing reagent after nitric acid digestion. Experiments were carried out to determine the decomposition efficiency of residual carbon species, primarily o-, m- and p-nitrobenzoic acids (NBAs), using ozone.The NBAs were degraded by purging sample digests with ozone, while heating the solutions with microwave energy at atmospheric pressure. The effects of the degradation temperature and solution pH on the ozonation of NBAs were determined. Solid phase extraction (C18) was used to extract NBAs from the acid digestate solutions prior to HPLC analysis.Reversed phase HPLC was used to determine NBA concentrations in digest solutions. After 2.5 h of purging ozone at 80 °C, 33.65 ± 3.80% o-NBA degradation, 19.39 ± 1.74% m-NBA degradation, and 26.47 ± 3.36% p-NBA degradation were obtained. Keywords: Ozone; biological samples; microwave digestion; residual carbon; nitrobenzoic acids Today, microwave energy has been widely applied to sample digestion for the analysis of biological, geological, and environmental materials.Microwave enhanced sample preparation techniques dramatically speed up sample dissolution and decomposition, but do not eliminate organic carbon residues due to the incomplete oxidation of organic compounds. Research on microwave sample decomposition has been based on comparison and optimization of conventional acid digestion procedures using microwave energy in open or closed vessels.The completeness of these techniques can be evaluated by considering the residual carbon content, recovery of analytes, and digestion time. Since organic carbon residues interfere with detection by some instrumental techniques, the completeness of decomposition can be a crucial factor for accurate analysis. Nitric acid, the most commonly used mineral acid, oxidizes the majority of organic matrix components with the primary exception of aromatic ring structures, particularly nitrobenzoic acids (NBAs).1–7 Pratt et al.1 identified o-, m- and p-NBAs as the major residual organic compounds after nitric acid digestion of bovine liver in sealed perfluoroalkoxy (PFA) vessels at a maximum temperature of 180 °C and a sustained maximum pressure of 8 atm (1 atm = 101.325 kPa).A number of researchers have identified that residues from incomplete nitric acid oxidation act as interferences with some analytical techniques such as polarography, voltammetry, and ETAAS.1,4,5,8,9 In spectroscopic techniques with high energy interfaces, such as ICP-OES, ICP-MS, and FAAS, small amounts of residual carbon is not a significant limitation.To take account of residual carbon Pratt et al. also successfully demonstrated that refluxing with perchloric acid (HClO4) can decompose the residual organic products remaining after nitric acid digestion.1 However, a drawback of using perchloric acid is the possible formation of explosive products if safety precautions are not taken.10 Thus, an alternative oxidizing reagent is desirable to completely and safely decompose organic carbon residues.The goal is the evaluation of ozone as an alternative decomposition reagent for analytical sample preparation, especially coupled to microwave decompositions. This work extends the studies of Pratt et al.1 and W�urfels et al.4,5 who identified that the major residual organic species in nitric acid digests of bovine liver are o-, m- and p-NBAs, arising from the nitration of aromatic amino acid constituents of original sample protein.This research focuses on the addition of ozone to aqueous acidic solutions to aid the microwave degradation of NBAs. The application of ozone to decompose o-, m- and p-NBAs also allows an investigation into the oxidation processes needed to efficiently destroy or minimize organic carbon residues. Experimental Reagents and Standards NIST SRM 1577a, bovine liver, was used as the biological matrix in this study.Samples were vacuum-dried at room temperature for 24 h in accordance with the instructions on the certificate of analysis. Compressed air (reagent grade) was used as the source of oxygen to produce ozone. Ozone concentrations were measured by the decolorization of potassium indigo trisulfonate (Aldrich, Milwaukee, WI, USA) solution. Doubly distilled 18 MW NanoPure water (Barnstead, Dubuque, IA, USA) was used to prepare all solutions. 2-Nitrobenzoic acid (o-NBA) (96%, Aldrich), 3-nitrobenzoic acid (m-NBA) (99%, Aldrich) and 4-nitrobenzoic acid (p-NBA) (99%, Eastman Kodak, Rochester, NY, USA) were used to prepare aqueous samples for the NBA degradation studies. All other reagents were certified ACS reagent grade. The HPLC eluent used was 5% methanol (HPLC grade)– 15% acetonitrile (HPLC grade)–80% H2O (18 MW) (by volume before mixing), to which 0.05% trifluoroacetic acid was added to acidify the solution to a pH of 2–2.3.Apparatus The ozonator used for ozone production in this work was a Model ss-150 (Pillar Technologies, Hartland, WI, USA). It was purchased from Radiation Disposal Systems (Charlotte, NC, USA) and was redesigned to a Model ss-300 (Select Industrial Systems, Waukesha, WI, USA) to double the ozone production rate. Fig. 1 shows a schematic representation of the ozonation apparatus set-up. The reaction vessel was a Teflon microwave vessel, 28 cm tall with an id of 3.8 cm (Prolabo, Paris, France).The total volume was approximately 320 ml. The gas stream was introduced into the reaction vessel via a 25 mm pore size † Present address: Caelum research Corporation, 7505 Standish Place, Rockville, MD 20855, USA. ‡ Present address: Department of Natural Sciences, University of North Florida, Jacksonville, FL 32224, USA. Analyst, March 1997, Vol. 122 (211–215) 211Teflon gas sparger (Omnifit, Toms River, NJ, USA) located close to the bottom of the vessel.Ultraviolet and visible spectra were obtained using a Cary 1E UV/VIS spectrophotometer (Varian Australia, Springvale Victoria, Australia). The instrument was interfaced to a Grid 3861S-25 PC computer (Varian Australia) running Varian Cary 13E software Version 2. The pH of the solutions was calculated or measured using a Model 710A pH meter (Orion, Boston, MA, USA) equipped with a pH combination electrode (Orion, part No. 91–56). The atmospheric pressure microwave unit used in this research was a Maxidigest MX350 (Prolabo).A gas bulb thermometer (Prolabo) and a PC computer (Dell 450/MX, Austin, TX, USA) running MEGAL 500 software (Prolabo) were used to control microwave power based on the measurement of the temperature. The closed vessel microwave unit used for the dissolution of bovine liver was an MLS 1200 MEGA (Milestone, Sorisole, Italy). High pressure vessels were used for these digestions (Milestone).HPLC analysis was performed using a Model M-6000A (Waters Associates, Milford, MA) coupled to a Model 440 absorbance detector (Waters Associates) and an HP 3396A integrator (Hewlett-Packard, Avondale, PA). The reversed phase column used was a Vydac Equivalent C18 column (nonendcapped, 25 cm 3 4.6 mm, 5 mm particle size) (Whatman, Clifton, NJ, USA). Solid phase extraction was performed on a Mega Bond Elut octadecylsilyl (C18) cartridge (endcapped) (ChromTech, Apple Valley, MN, USA).A Bond Elut adapter (ChromTech) and poly(propylene)polyethylene syringe (Aldrich) were used to push solvents through the cartridge. Procedures The ozone transfer rate into water and the ozone decomposition rate in water were initially evaluated. Ozonated air was purged through 50 ml of doubly distilled water in the Teflon vessel for a certain period of time with different flow rates (110, 190 and 328 ml min21). The concentration of the ozone solutions were determined immediately using the Indigo Blue method.11 The ozone decomposition rate was determined by measuring the decrease of the ozone absorbance at 260 nm over a period of 1 h.In order to study the degradation of o-, m- and p-NBAs at different pHs, solutions of the NBAs were made up in different concentration nitric acid solutions and water. Ozone was then purged through these solutions for 90 min at 80 or15 °C, the solutions were cooled, made up to volume in 100 ml calibrated flasks and analyzed using solid phase extraction (SPE) and HPLC.Bovine liver samples were digested using a closed vessel high pressure microwave system. Each batch contained four samples of bovine liver ( ~ 0.25 g each) and two blanks in 10 ml of concentrated nitric acid using the high pressure vessels. Three of the samples were put in regular sample vessels and the remaining one was put in the monitor vessel to measure the pressure and temperature change during the digestion.The microwave energy was programmed to heat at 600 W until reaching 180 °C (in about 2 min), and the temperature was kept at 180 °C for about 8–9 min. The samples and rinses were transferred to calibrated flasks, diluted to 50 ml, and were directly used for the atmospheric microwave vessel ozonation study. o-, m- and p-NBA solutions were also prepared using a flow rate of 328 ml min21 to study the mechanism of the ozonation. After purging ozone (1–2 h) at a temperature of 80 °C or higher, the solutions were transferred and diluted to 100 ml calibrated flasks before SPE and HPLC analysis.The C18 cartridges used for SPE were conditioned by passing 5 ml of methanol through them, followed by at least 10 ml of doubly distilled water prior to extraction of an aqueous sample. An aliquot of 0.5 ml of the sample solution was passed through the cartridge, followed by a wash with 5 ml of pH 2 aqueous ophosphoric acid and finally eluted with 2 ml of methanol.Results and Discussion Ozone Transfer Rate and Decomposition Rate in Solution By varying the ozone gas flow rate, it was shown that higher gas flow rates produced greater ozone transfer rates into water. Therefore, a 328 ml min21 gas flow rate was chosen as the ozone purging flow rate for subsequent experiments. At this flow rate, the ozone generator produced 445 mg of ozone per hour. This ensured a fast ozone mass transfer and a high ozone concentration without violent bubbling of the samples, minimizing any loss of solution.The saturated ozone solution that was produced decomposed by 50% in 1 h once the gas flow was turned off. At the flow rate of 328 ml min21, pure water can be saturated with ozone in less than 10 min, which indicated that the dissolution rate of ozone was much faster than the decomposition rate of ozone under the experimental conditions; therefore, a saturated ozone solution could be maintained. SPE Sample Clean-up The efficiency of the extraction of NBAs by the C18 solid phase cartridges was investigated at the level of NBAs expected in the digests.Table 1 shows the SPE cartridge extraction efficiency for a 1 ml aliquot of 66.7 ppm of each of o-, m-, p-NBA solution (200 mg total). The efficiency of extraction for each NBA was excellent (99–103%). Replicate extractions on the same column and on different columns were statistically indistinguishable. Owing to the levels of NBAs being well below the capacity of the cartidges (50 mg), it was anticipated that this efficiency would also be acheivable on real sample digests even though other organic species may be present.HPLC of Nitrobenzoic Acid An interesting observation in the HPLC chromatograms was the changing retention time of the compounds. Continuous injections of bovine liver digestate solution onto the HPLC column caused the retention time of the compounds to increase significantly. This was attributed to the retention of some large organic species extracted by the C18 cartridge which changed the surface of the column packing and thus the chromatography.Fig. 1 Ozone apparatus set-up. 212 Analyst, March 1997, Vol. 122Pumping acetonitrile through the column for approximately 30 min washed out the retained organics and the retention times returned to their original values. The use of a guard column could alleviate this problem. o-, m- and p-NBA Degradation Table 2 shows the effects of pH and temperature on the degradation.Since this research focuses on the application of ozone with nitric acid to improve the oxidation of organic residues, in these experiments nitric acid was used to adjust the pH of the NBA solutions. The mechanism of ozonation of NBAs in acidic solution was expected to be an electrophilic reaction.12,13 However, the presence of nitrosyl (–NO2) and carboxylic acid (–COOH) groups (both electron-withdrawing) on the aromatic ring makes it electron deficient, so it is difficult for ozone to electrophilically attack the ring structure at low pH.The reaction rate observed should therefore be relatively slow in 25% nitric acid13 and concentrated nitric acid solutions as was observed. The pKa values of o-, m- and p-NBA in water at 25 °C are 2.17, 3.49 and 3.77 respectively.14 Increasing the solution pH increases the concentration of deprotonated forms of NBAs. When the NBAs are deprotonated (NBA2), carboxylate (–COO2), an electron-donating group, replaces –COOH and electrophilic attack by ozone is favored.Thus, the increase from pH 2 < 0.1 to 3.2 results in very different oxidation rates. Even at a pH of 0.8, most of the NBA molecules are protonated, but the existence of a small concentration of the deprotonated species provides partial oxidation. At a pH of 3.2, about 85% degradation of NBAs can be achieved in 90 min at 80 °C, suggesting that it would be better to neutralize the digestate solution, with a non-contaminating reagent, such as ammonia, prior to ozonation to enhance the destruction of NBAs.Both of these reagents could be introduced as high purity gases. Temperature also helps the oxidation process. The higher the temperature, the faster the degradation rate. As these experiments were designed for the purging of ozone through solution in an atmospheric pressure microwave system, the temperature of the reaction was limited by the boiling point of the reagent used (118 °C for nitric acid). At this temperature the experimental conditions are rather violent and thus, for safety reasons, a lower temperature (80–90 °C) was used and is recommended. Hydrogen peroxide has been shown to increase the rate of decomposition of nitrogenzoic acids in aqueous solution.The addition of hydrogen peroxide to the reaction mixture was also tested while bubbling ozone through the solution. No additional decomposition was seen when compared to ozone alone.Degradation of Bovine Liver Digestate The estimated total organics in the original SRM 1577a bovine liver sample was ~ 97.4% of the sample dry mass (calculated from ref. 15). After closed vessel nitric acid digestion at 180 °C, HPLC analysis of the digestate solution was performed and is shown in Fig. 2. The major organic residues are o-, m-, p-NBA and an unidentified peak at a retention time of ~ 10 min. The concentrations of the three NBA isomers were calculated and are summarized in Table 3.The average total NBA residue was 2.70% (m/m). Not taking into account the amount of the unidentified compound, the closed vessel nitric acid digestion decomposed ~ 94.7% (m/m) of the organics in the original Table 1 Extraction efficiency of o-, m- and p-nitrobenzoic acids on the same and different C18 columns Extraction* Cartridge I— A B C o-NBA 100 ± 0.9 100 ± 0.7 100 ± 0.6 m-NBA 99 ± 1.4 99 ± 1.5 99 ± 1.2 p-NBA 100 ± 0.7 100 ± 0.8 100 ± 0.8 Cartridge II— A B C o-NBA 102 ± 1.7 100 ± 1.6 100 ± 1.5 m-NBA 102 ± 1.7 100 ± 1.3 100 ± 1.5 p-NBA 103 ± 1.8 100 ± 1.1 101 ± 0.8 Cartridge III— A B C o-NBA 100 ± 1.3 100 ± 0.7 100 ± 1.1 m-NBA 100 ± 1.4 99 ± 1.2 99 ± 1.3 p-NBA 100 ± 0.8 100 ± 0.7 99 ± 0.7 * Mean ± % RSD of three replicate injections.Extractions A, B, and C were consecutive on each column. Table 2 Degradation of o-, m- and p-nitrobenzoic acids in acid solutions Temp- Degradation (% ± %RSD*) erature/ Time/ Solution pH °C min o-NBA m-NBA p-NBA 18 MW water ~ 3.2 80 90 84 ± 3.8 85 ± 3.2 88 ± 3.4 1% HNO3 0.8 80 90 58 ± 0.5 63 ± 3.2 69 ± 0.1 25% HNO3 < 0.1 80 90 23 ± 1.4 15 ± 2.4 19 ± 0.7 Conc.HNO3 < 0.1 80 90 27 ± 1.7 17 ± 1.7 21 ± 0.5 Conc. HNO3 < 0.1 115 90 42 ± 0.4 27 ± 1.6 28 ± 0.4 * %RSDs were calculated from three replicate digestions. Fig. 2 One hour degradation of bovine liver digestate solution with ozone purging at 80 °C using an atmospheric pressure microwave system, (a) before and (b) after.Analyst, March 1997, Vol. 122 213bovine liver sample. In comparison to the 3.01% (m/m) NBA residues found in the Pratt et al.1 work, the amount of total NBA residues in our experiments was 0.31% lower. However, in the work by Pratt et al., the major organic residues were o-, m- and p-NBA; and no unknown peak such as that appearing at ~ 10 min was reported. The decomposition of o-, m- and p-NBA in the digestate solutions was not as much as the decomposition of o-, m- and p-NBA in aqueous solution.The presence of the other residual carbon species that are more easily oxidized and react with ozone preferentially may be one reason for this, but is likely not to be the only contributing factor. Bovine liver biological tissue has significant protein content. Many biological tissues with proteins containing an aromatic side chain will also have significant quantities of o-, m- and p- NBA remaining after digestion.These digestion products are typical of protein containing tissues.1 Since dinitrobenzoic acid has been reportedly found in biological sample digestate solutions as one of the major organic residues,4 experiments were performed to identify whether or not the unknown peak was dinitrobenzoic acid. Fig. 3 shows the HPLC analysis of 2,4-dinitrobenzoic acid (top) and 3,5-dinitrobenzoic acid (bottom). Based on the retention times, neither seemed to be the unidentified peak.However, 2,4-dinitrobenzoic acid had a retention time of 6.4 min, which is close to a small peak (retention time 6.7 min) shown in Fig. 4. Spiking of the digestate solution by adding ~ 20 ppm of 2,4-dinitrobenzoic acid to the solution indicated that this peak is most likely to be 2,4-dinitrobenzoic acid [see Fig. 4 (b)]. The estimated concentration is ~ 0.01% (m/m) of the dry bovine liver sample. In the W�urfels et al. paper,4 5% 2,4-dinitrobenzoic acid was reported to be one of the organic residues after nitric acid digestion of bovine liver sample.However, in the Pratt et al. work, dinitrobenzoic acid was not reported.1 This possibly indicates that the difference of the temperature and pressure during digestion can lead to a shift in the distribution of digestion products. However, it should be noted that in the Pratt et al. experiments, isobutyl methyl ketone (IBMK) was used to extract the NBAs from the digestion solution before HPLC analysis. In our experiment, the NBAs were extracted from the digestate solution by using SPE.It is possible that the extraction method used in these experiments is more efficient, allowing the recovery of the additional unknown peak. The degradation could be improved by the addition of other reagents with the ozone such as peroxide for instance. However the goal of this study was to evaluate the degradation of ozone alone and not to add any other additional reagents to the ozone decomposition.Ozone is produced just prior to its addition in the decomposition system from filtered air or oxygen. It is not a classical reagent and may not contribute significantly to the analytical blank. After an hour of purging ozone at 80 °C through bovine liver digestate solution, 17.08 ± 2.69% o-NBA degradation, 4.78 ± 3.21% m-NBA degradation, and 6.88 ± 2.95% p-NBA degradation result (see Fig. 2). Most of the unknown peaks were degraded by ozone in the first hour of purging.After 2.5 h of purging ozone at 80 °C, 33.65 ± 3.80% o-NBA degradation, 19.39 ± 1.74% m-NBA degradation, and 26.47 ± 3.36% p-NBA degradation were obtained. Conclusions It has been shown in this research that ozone can oxidize o-, mand p-NBA, and has the potential to be used as an additional decomposition and/or finishing reagent. Deprotonated NBAs react faster with ozone than protonated NBAs, strongly suggesting an electrophilic attack of the benzene ring. Thus, increasing the pH of the solution to increase the concentration of the deprotonated NBA significantly improves the ozonation rate. When ozone was applied to closed vessel bovine liver digestate solution, the decomposition of residual o-, m- and p- NBA and other residual carbon was shown.A better understanding of the chemistry of the ozonation of biological sample digestate solutions is an important goal for further research. Since the pH of the solution is essential for the degradation of NBAs, the reaction efficiency could be increased if an appropriate reagent can be found to neutralize the digestate solution (i.e., ammonia) without introducing contamination.This research demonstrates the feasibility of ozone as an Table 3 Concentrations of o-, m- and p-nitrobenzoic acids after closed vessel digestion of bovine liver o-NBA m-NBA p-NBA Total NBA Sample Mass/g (%) (%) (%) (%) 1 0.2178 0.61 0.45 1.59 2.66 2 0.2181 0.59 0.42 1.58 2.59 3 0.2069 0.70 0.50 1.65 2.85 Mean 0.63 0.46 1.61 2.70 s 0.06 0.04 0.04 0.13 %RSD 9.52 8.70 2.48 4.81 Fig. 3 HPLC analysis of 2, 4- and 3, 5-dinitrobenzoic acid under the same conditions as nitrobenzoic acid analysis. Fig. 4 HPLC analysis of (a) bovine liver digestate and (b) a 2,4-dinitrobenzoic acid spike of the same solution. 214 Analyst, March 1997, Vol. 122alternative decomposition method. Additional studies to optimize its use seem appropriate. References 1 Pratt, K. W., Kingston, H. M., MacCrehan, W. A., and Koch, W. F., Anal. Chem., 1988, 60, 2024. 2 Kingston, H. M., and Jassie, L. B., J. Res. Nat. Bur. Stand., 1988, 93, 269. 3 Krushevska, A., Barnes, R. M., Amarasiriwaradena, C. J., Foner, H. A., and Martines, L., J. Anal. At. Spectrom., 1992, 7, 845. 4 W�urfels, M., Jackwerth, E., and St�oppler, M., Anal. Chim. Acta, 1989, 226, 17. 5 W�urfels, M., Jackwerth, E., and St�oppler, M., Anal. Chim. Acta, 1989, 226, 1. 6 Kingston, H. M., Walter, P. J., Chalk, S. J., Lorentzen, E., and Link, D. D., in Microwave Enhanced Chemistry, ed. Kingston, H. M., and Haswell, S., American Chemical Society, Washington, DC, 1997, ch. 3. 7 Walter, P. J., Chalk, S. J., Kingston, H. M., and Lorentzen, E. M. L., in Microwave Enhanced Chemistry, ed. Kingston, H. M., and Haswell, S., American Chemical Society, Washington, DC, 1997, ch. 2. 8 Jackwerth, E., and Gomiscek, S., Pure Appl. Chem., 1984, 56, 479. 9 Kotz, L., Henze, G., Kaiser, G., Pahlke, S., Veber, M., and Tolg, G., Talanta, 1979, 26, 681. 10 Hertz, J., and Pani, R., Fresenius’ Z. Anal. Chem., 1987, 328, 487. 11 Langlais, B., Reckhow, D. A., and Brink, D. R., Ozone in Water Treatment: Application and Engineering, Lewis Publishers, Chelsea, MI, 1991, p. 103. 12 Langlais, B., Reckhow, D. A., and Brink, D. R., Ozone in Water Treatment: Application and Engineering, Lewis Publishers, Chelsea, MI, 1991. 13 Jiang, W., Masters Thesis Duquesne University, 1996. 14 Lange’s Handbook of Chemistry, ed. Dean, J. A., McGraw Hill, New York, NY, 1979, 5. 15 Certificate of Analysis: Standard Reference Material 1577a—Bovine Liver, US Department of Commerce, The National Bureau of Standards, Gaithersburg, MD, 1982. Paper 6/05282G Received July 29, 1996 Accepted November 15, 1996 Analyst, March 1997, Vo

ISSN:0003-2654    

DOI:10.1039/a605282g

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Determination of Organophosphorus Pesticides in Foods Using an Accelerated Solvent Extraction System Hirotaka Obana*a, Kosuke Kikuchib, Masahiro Okihashia and Shinjiro Horia a Osaka Prefectural Institute of Public Health, 1-3-69, Nakamichi, Higashinarik-ku, Osaka, 537, Japan. E-mail: obana@iph.pref.osaka.jp b Japanese Consumers’ Co-operative Union, 1-17-18, Nishiki-cho, Warabi, 335, Japan Residual organophosphorus pesticides in foods were determined by accelerated solvent extraction (ASE), gel permeation chromatography and GC–FPD.Pesticides were extracted at 100 °C under 1500 psi pressure in less than 20 min. Wet samples were extracted after mixing with Extrelut drying agent. Four foods were spiked with 19 pesticides at 0.1 ppm or less. The average recoveries of these pesticides were 80–90% and the precision was < 10%. Both methamidophos and acephate gave 37–50% recoveries and 6–40% precision. These results suggested that ASE can be used to extract residues of organophosphorus pesticides in foods. Keywords: Organophosphorus pesticides; foods; accelerated solvent extraction; gas chromatography–flame photometric detection Laboratories that monitor pesticides have developed various means of measuring residual levels in foods. Multiresidue methods are usually applied in pesticide monitoring because several compounds can be targeted simultaneously, which reduces labor and material costs.Extraction is one of the most important steps in the analysis.Conventional manual liquid– liquid partition consumes time and it is labor intensive. Moreover, it sometime results in an emulsion that can lower the reproducibility of the analysis. Japanese laboratories cannot freely drain waste containing dichloromethane owing to legal restrictions ( < 10 mg ml21).1 Dichloromethane is a frequently used solvent in multiresidue methods,2,3 so an alternative method is required that incorporates non-halogenated solvents.The latest automated extraction methods use less solvent and give improved analytical precision. Supercritical fluid extraction (SFE) is an example that has been adapted to residual pesticides analysis.4–8 Pesticides have also been analyzed after microwave-assisted extraction.9,10 On the other hand, Richter et al.11 reported that environmental contaminants such as PCBs and PAHs were rapidly and effectively extracted at temperatures above 100 °C with pressurized solvents.They showed that solvents solubilize pesticides and penetrate sample matrices better under these conditions. Both accelerated solvent extraction (ASE) and microwave extraction operate under the same principle of automated solvent extraction at high temperature. In this study, we extracted organophosphorus pesticides from foods using ASE. Experimental Pesticide Standards Organophosphorus pesticides were purchased from Wako (Osaka, Japan): acephate, chlorpyrifos, chlorpyrifos-methyl, diazinon, dichlorvos, dimethoate, dioxabenzofos, edifenphos, EPN, ethion, ethoprophos, fenthion, fenitrothion, malathion, methamidophos, methidathion, phosalone, pirimiphos-methyl and prothiofos.Each compound was dissolved in hexane to make a 1000 mg ml21 stock standard solution. The stock standard solution was diluted with acetone and used in the recovery tests at 1 mg ml21. Organophosphorus pesticides cause matrix effects that result in over-estimation of the recoveries in GC analysis.6,7 To prevent this, standard mixtures were diluted with the respective cleaned-up extract from pesticide-free samples. Reagents Acetone, hexane, ethyl acetate and anhydrous Na2SO4 were of pesticide analysis grade and cyclohexane was of HPLC grade.All reagents were purchased from Wako. Diatomaceous Earth Particles of diatomaceous earth (Extrelut for refilling; particle size 160–800 mm) (Merck, Darmstadt, Germany) were heated at 550 °C for 15 h to exclude interferences in the GC analysis.Food Preparation Samples were purchased at a local market in Osaka and before use we confirmed that organophosphorus pesticide residues were below detectable levels. Pesticide residues were analyzed using a liquid–liquid extraction described by Yoshida et al.12 About 500 g of broccoli or grapefruit were chopped in a conventional food processor (MK-K3, Matsushita, Osaka, Japan) for 5 min to obtain thoroughly mixed homogenates. The operating manual recommended that wet samples should be mixed with a drying agent to facilitate solvent penetration into sample matrices.13 An aliquot of 5 g of sample homogenate and 6 g of Extrelut particles was ground in a mortar (12 cm id) with a pestle until the mixture became homogeneous. In the fortification study, 0.5 ml of pesticide mixture at 1 mg ml21 was added to the sample homogenate to a final concentration of 0.1 ppm on a sample mass basis and the mixture was left for 30 min before adding Extrelut. Pesticides were added to flour at a final concentration of 0.05 ppm, then extracted without extra Extrelut.The procedure was performed twice for each sample because of GC separation preferences. One group consisted of methamidophos, acephate and dimethoate and the other of the remaining 16 pesticides. Mixtures of samples and Extrelut were placed in stainless-steel cells (33 ml; 11 cm 3 1.9 cm id). ASE Accelerated solvent extraction was performed with a Dionex AS 200 system (Dionex, Sunnyvale, CA, USA).The pre-set default conditions were as follows: extraction temperature, 100 °C; extraction pressure, 1500 psi (10.5 kPa); pre-heating period, 5 min; static extraction period, 5 min; solvent flash, 19.8 ml; nitrogen purge, 60 s; and collection, in 60 ml glass vials with Teflon coated rubber caps (I-CHEM, New Castle, DE, USA). The extraction cell was heated and pressurized immediately after introducing it into the heating port. The pressure valve was Analyst, March 1997, Vol. 122 (217–220) 217sometimes opened to maintained the pre-set pressure, and the extracted solvent was collected from pre-heating. After extraction, the cell was purged with fresh extraction solvent and additional nitrogen to collect the remaining extract in the cell. The extraction process was automated according to the assigned extraction method. The extracts were usually suspended and the water layer which lay separated in the bottom of vial because it was also co-extracted with solvent. The extract was transferred into a separating funnel.The collecting vial was washed with 10 ml of hexane twice, then hexane was added to the extract. As the suspension of the extract became heavier with hexane, the organic layer was left for about 2 h to form water droplets from the organic layer. After the water portion had been discarded, the extract was dried by adding anhydrous Na2SO4. The extract was carefully evaporated almost to dryness with warming at < 40 °C and the residue was dissolved in ethyl acetate– cyclohexane (1 + 1) for gel permeation chromatography (GPC) and diluted to 7 ml.GPC14 Extracts were injected into an AS-2000 system (abc Laboratories, Columbia, MO, USA) equipped with an Environsep- ABC column (300 3 21.2 mm id). The mobile phase was ethyl acetate–cyclohexane (1 + 1) at a flow rate of 5 ml min21. Food extract (5 ml) was loaded into the GPC system. The first 70 ml were discarded and the following 80 ml were collected as the pesticide fraction.A further 50 ml were eluted for the GPC wash. The pesticide fraction was concentrated to 3.5 ml with an evaporator for GC determination. GC–FPD Determination Pesticides were determined using a GC 17A instrument (Shimadzu, Kyoto, Japan) equipped with a flame photometric detector. A DB-210 column (5 m 3 0.25 mm id, 0.25 mm thickness) (J & W Scientific, Folsom, CA, USA) was used for methamidophos, acephate, and dimethoate determinations.The temperature program was initial temperature 80 °C (held for 1 min), increased at 10 °C min21 to 260 °C (held for 4 min). A DB-5 column (30 m 3 0.25 mm id, 0.25 mm thickness) (J & W) was used for the other pesticides. The temperature program was initial temperature 80 °C (held for 1 min), increased at 20 °C min21 to 120 °C (no hold), then at 4 °C min21 to 280 °C (held for 5 min). Carrier gas was helium. A 2ml volume was injected in the splitless mode at 250 °C.The detector temperature was set at 300 °C. The chromatogram was recorded using a CR7A integrator (Shimadzu). Results and Discussion Extraction From Flour First, a recovery test was performed with flour with a low water content, so that no drying agent needed to be added to the sample during extraction. Flour (10 g) spiked with 0.05 ppm of Table 1 ASE extraction of 17 organophosphorus pesticides in flour using three solvent mixtures Cyclohexane– Dichloromethane– Ethyl acetate– acetone (1 + 1) acetone (1 + 1) acetone (1 + 1) Mean Mean Mean recovery RSD recovery RSD recovery RSD Pesticide (%)* (%) (%)* (%) (%)* (%) Dichlorvos 36.6 12.1 42.9 37.0 46.6 30.3 Ethoprophos 100.7 2.5 95.1 3.8 141.4 24.1 Dioxabenzofos 100.8 3.6 83.2 18.1 95.5 27.7 Dimethoate 49.1 1.8 86.0 16.1 199.2 34.4 Diazinon 99.2 2.4 93.5 1.6 134.3 26.7 Chlorpyrifos-methyl 100.7 5.1 92.9 13.6 104.6 28.4 Fenitrothion 111.7 5.3 90.7 12.8 105.9 29.0 Pirimiphos-methyl 96.4 4.8 106.0 0.8 155.4 24.3 Malathion 106.3 4.6 93.4 5.4 137.8 19.1 Chlorpyrifos 94.9 3.4 93.8 7.0 125.3 25.2 Fenthion 90.1 4.8 88.3 0.1 146.1 24.4 Methidathion 113.6 14.1 88.5 13.9 106.9 32.1 Prothiofos 89.2 5.5 92.6 5.8 128.6 26.0 Ethion 85.5 7.3 96.3 5.2 142.3 25.6 Edifenphos 115.8 11.7 91.2 6.6 141.3 28.6 EPN 87.1 10.5 93.5 11.3 128.6 31.5 Phosalone 106.5 10.9 93.4 7.3 140.6 33.2 * Means of three experiments.Table 2 ASE extraction of 19 organophosphorus pesticides from grapefruit and orange juice by ASE using cyclohexane–acetone (1 + 1) Grapefruit Orange juice Mean Mean recovery RSD recovery RSD Pesticide (%)* (%) (%)* (%) Dichlorvos 82.8 15.3 87.1 10.5 Methamidophos 8.7 20.1 9.3 48.5 Ethoprophos 99.6 13.9 61.1 13.5 Dioxabenzofos 100.0 19.6 51.6 16.8 Acephate 10.2 19.4 10.6 45.3 Dimethoate 97.9 14.1 74.7 8.0 Diazinon 102.0 12.6 73.4 7.5 Chlorpyrifos-methyl 105.1 17.6 75.7 6.0 Fenitrothion 99.7 8.2 76.5 10.9 Pirimiphos-methyl 101.7 14.0 88.4 4.5 Malathion 98.2 9.1 76.3 6.3 Chlorpyrifos 105.1 10.4 79.2 7.5 Fenthion 105.6 10.2 80.9 5.9 Methidathion 100.5 4.4 73.7 9.0 Prothiofos 103.1 9.6 75.2 9.7 Ethion 97.5 21.8 110.5 19.1 Edifenphos 100.9 14.9 74.1 13.6 EPN 107.1 13.4 87.6 15.3 Phosalone 90.3 13.8 70.3 14.3 * Means of three experiments. 218 Analyst, March 1997, Vol. 122each pesticide was extracted with three solvent mixtures each containing 50% acetone. As shown in Table 1, extraction with cyclohexane–acetone (1 + 1) and dichloromethane–acetone (1 + 1) resulted in good recoveries. The recoveries of most of the pesticides was 83–115%.The RSDs were acceptable, usually < 10%, using cyclohexane–acetone (1 + 1). With ethyl acetate– acetone (1 + 1) the RSD varied from 19 to 34%. The recovery of dichlorvos was around 40% with all three mixtures and the precision was also poor. One explanation is that the compound was lost during sample preparation because dichlorvos is volatile with a vapor pressure of 290 mPa at 20 °C.15 Extraction From Wet Samples Since one aim of this study was to develop a dichloromethanefree extraction procedure, we extracted pesticides from grapefruit and orange juice with cyclohexane–acetone (1 + 1) (Table 2).In this study methamidophos and acephate were added in addition to the 17 pesticides in Table 1. The recoveries of the 17 pesticides except methamidophos and acephate varied from 82 to 105% in grapefruit and from 51 to 110% in orange juice.Little methamidophos or acephate was extracted under the conditions tested. The extraction precision of these pesticides was poor in orange juice. Extraction Conditions We studied the effects of static extraction time, temperature and pressure to improve the extraction rates of methamidophos and acephate in orange juice (Table 3). Other conditions were the same as those described above. Since two of the pesticides were water soluble,16 orange juice seemed to be one of the most difficult foods from which to obtain satisfactory recoveries.The appropriate conditions for orange juice should also be sufficient for drier samples. Time and temperature did not improve the recoveries under the conditions tested. Only a higher extraction pressure slightly improved the recovery. Further improvement by increasing the pressure was unlikely since the mechanical pressure limit of the ASE system is 3000 psi. Table 3 Effects of extraction time, temperature and pressure on recoveries of methamidophos and acephate* Recovery (%) Variable Methamidophos Acephate Time/min— 5 15.4 9.3 10 11.9 9.6 15 11.7 7.7 Temperature/°C— 50 5.8 4.5 100 8.1 8.0 150 6.4 4.2 Pressure/psi— 1500 9.5 16.0 2000 13.3 20.0 2500 15.0 22.2 * The extraction solvent was cyclohexane–acetone (1 + 1) and other conditions were the default values as described in the text.Values are means of two experiments. Table 5 ASE extraction of 19 organophosphorus pesticides in foods using ethyl acetate Orange juice Grapefruit Broccoli Flour (0.1 ppm) (0.1 ppm) (0.1 ppm) (0.05 ppm) Mean Mean Mean Mean recovery RSD recovery RSD recovery RSD recovery RSD Pesticide (%)* (%) (%)* (%) (%)* (%) (%)* (%) Dichlorvos 91.5 4.6 102.8 10.0 87.4 3.9 92.7 9.9 Methamidophos 56.6 7.9 60.3 17.0 53.7 6.2 37.3 41.5 Ethoprophos 68.9 6.5 91.4 4.7 114.0 4.8 90.1 9.8 Dioxabenzofos 67.0 4.9 88.6 6.8 97.0 4.8 60.3 7.9 Acephate 54.8 9.0 59.1 20.9 50.8 6.1 55.7 10.1 Dimethoate 91.8 5.0 109.3 9.9 103.2 3.0 83.5 9.2 Diazinon 81.9 4.3 86.6 5.2 88.0 4.6 90.9 2.2 Chlorpyrifos-methyl 76.9 2.9 87.0 5.6 87.8 4.6 72.3 3.1 Fenitrothion 97.6 7.3 89.0 6.0 102.8 5.5 91.4 4.3 Pirimiphos-methyl 85.3 2.2 85.5 7.0 88.0 5.4 105.7 1.7 Malathion 81.9 4.1 89.5 5.9 99.5 5.1 77.5 0.4 Chlorpyrifos 74.1 2.7 87.3 6.2 81.4 5.1 95.2 2.7 Fenthion 79.6 4.8 90.3 4.2 91.4 7.7 92.3 2.3 Methidathion 80.9 11.3 91.4 5.2 110.9 6.7 73.5 4.7 Prothiofos 68.9 4.4 90.1 3.7 88.5 6.2 96.1 2.3 Ethion 78.1 5.8 90.1 4.1 87.5 9.0 98.1 2.1 Edifenphos 90.9 7.4 90.9 5.4 101.5 7.2 67.9 5.1 EPN 83.1 9.1 85.7 3.5 91.9 5.7 98.4 2.0 Phosalone 97.5 9.5 84.0 6.2 93.4 6.4 96.8 1.5 Average 79.3 87.4 90.4 82.9 * Means of three experiments.Table 4 ASE extraction of methamidophos and acephate from orange juice using various solvents Recovery (%)* Solvent Methamidophos Acephate Cyclohexane–acetone (1 + 1) 9.3 10.6 Toluene–acetone (1 + 1) 46.6 45.3 Dichloromethane–acetone (1 + 1) 36.0 36.5 Ethyl acetate–acetone (1 + 1) 16.1 30.5 Acetonitrile 28.2 30.6 Ethyl acetate 56.1 47.8 * Means of two experiments. Analyst, March 1997, Vol. 122 219We studied the effects of extraction solvents on the recoveries of methamidophos and acephate in orange juice under the default extraction conditions (Table 4). Ethyl acetate recovered 56 and 47% of spiked methamidophos and acephate, respectively. Toluene–acetone (1 + 1) gave a secondary recovery of around 45%.The recovery with cyclohexane– acetone (1 + 1) was the lowest among the six solvent systems tested. Since ethyl acetate gave the best results, it was used to extract 19 pesticides from four spiked foods. As shown in Table 5, the average recoveries of the 19 pesticides in the four samples were considerably better: 79.3% in orange juice, 87.4% in grapefruit, 90.4% in broccoli and 82.9% in flour. The recovery of methamidophos and acephate ranged from 50 to 60% except for methamidophos in flour (37%). Most RSDs were < 10%.Although methamidophos and acephate tend to show poor precision, the poor reproducibility of these pesticide peaks on the chromatograms in GC–FPD might have caused the fluctuating recoveries. These results suggested that ethyl acetate is a useful solvent in ASE with which to determine residual organophosphorus pesticides. ASE extraction was compared with hexane extraction in pesticide-containing samples (Table 6). These samples were found in our routine surveillance and stored at 220 °C until analysis.Whereas the ASE extraction seemed to give slightly lower recoveries for the samples tested, the precision of ASE is better than that of hexane extraction.The colours of the hexane extracts were deeper than those in ASE extracts. The advantages of ASE over liquid–liquid extraction were as follows. The extraction procedure is simple, requiring only mixing of the samples with drying agent and transferring the mixture to an extraction cell.One sample can be extracted within 20 min, including washing the solvent line for the next extraction. The solvent volume with a 33 ml cell is 50–60 ml, which depends on the sample volume. Even though the volume is not large, the ratio of solvent (50 ml) to sample (5 g) is 10, whereas it is 2–5 in conventional procedures.2,14,17,18 Hence ASE should combine good recovery and precision with rapid extraction. Another advantage over the conventional method is continuous separation of the extract from sample residues during extraction.The disadvantage of ASE in this study is the 2 h waiting to remove water from the extract after adding hexane. Without hexane, small water droplets appeared after evaporation even when the extract was dried with anhydrous Na2SO4. Ethyl acetate at elevated temperature solubilized water to a greater extent during extraction than by conventional means at room temperature. Water remaining in the extract decreased the precision of the recovery tests (data not shown).Although thermal degradation of the target compounds is a potential concern because of elevated temperature during extraction, Richter et al.11 reported that the thermal degradation of DDT to DDE or DDD and of endrin to endrin aldehyde or endrin ketone did not occur during ASE at 150 °C. This study demonstrated that ASE automatically and rapidly extracted organophosphorus pesticides from foods with good accuracy and precision.ASE could be introduced as a means of determining residual levels of organophosphorus pesticides in foods. We are studying its applicability to multiresidue analyses of agrochemicals other than organophosphorus pesticides. References 1 Prime MinisterAs Office of Japan, Bulletin 54, December 27, 1993 (in Japanese). 2 Cairns, T., Luke, M. A., Chiu, K. S., Navarro, D., and Siegmund, E. G., Rapid Commun. Mass Spectrom., 1993, 7, 1070. 3 Luke, M.A., Froberg, J. E., Doose, G. M., and Masumoto, H. T., J. Assoc. Off. Anal. Chem., 1981, 64, 1187. 4 Skopec, Z. V., Clark, R., Harvey, P. M. A., and Wells, R. J., J. Chromatogr. Sci., 1993, 31, 445. 5 Hopper, M. L., and King, J. W., J. Assoc. Off. Anal. Chem., 1991, 74, 661. 6 Lehotay, S. J., Aharonson, N., Pfeil, E., and Ibrahim, M. A., J. AOAC Int., 1995, 78, 831. 7 Lehotay, S. J., and Eller, K. I., J. AOAC Int., 1995, 78, 821. 8 Stuart, I. A., MacLachlan, J., and McNaughtan, A., Analyst 1996, 121, 11R. 9 Lopezavila, V. L., Young, R., Benedicto, J., Ho, P., Kim, R., and Beckert, W. F., Anal. Chem., 1995, 67, 2096. 10 Lopezavila, V., Young, R., Kim, R., and Beckert, W. F., J. Chromatogr. Sci., 1995, 33, 481. 11 Richter, B. E., Jones, B. A., Ezzell, J. L., Porter, N. L., Avdalovic, N., and Pohl, C., Anal. Chem., 1996, 68, 1033. 12 Yoshida, S., Konaka, H., and Nishimune, T., J. Food Hyg. Soc. Japn., 1991, 32, 267. 13 ASE 200 Accelerated Solvent Extractor Operator’s Manual, Document No. 031149, Revision 01, Dionex, Sunnyvale, CA, 1995, Sect. 3–5. 14 Committee on Analytical Methods for Pesticide Residues, Food Sanit. Res., 1995, 45, 31 (in Japanese). 15 The Agrochemical Handbook, Royal Society of Chemistry, Cambridge, 3rd edn., 1991, p. A0141. 16 The Agrochemical Handbook, Royal Society of Chemistry, Cambridge, 3rd edn., 1991, pp. A0001 and A0265. 17 Pesticide Analytical Manual, ed. McMahon, B. M., and Hardin, N. F., Food and Drug Administration, Washington, DC, 3rd edn., 1994, vol. 1, pp. 302–305. 18 Fillion, J., Hindle, R., Lacroix, M., and Selwyn, J., J. AOAC Int., 1995, 78, 1252. Paper 6/06063C Received September 3, 1996 Accepted October 28, 1996 Table 6 Comparison between ASE and hexane extraction in pesticidecontaining samples ASE Hexane Mean RSD Mean RSD Food Pesticide (ppm)* (%) (ppm)* (%) Banana Chlorpyrifos 0.03 9.7 0.05 13.4 Okra Phosalone 0.05 4.0 0.08 27.5 Sweetie Chlorpyrifos 0.22 2.7 0.27 13.2 Malathion 0.03 20.9 0.03 35.9 * Means of three experiments. 220 Analyst, March 1997, Vol. 122 Determination of Organophosphorus Pesticides in Foods Using an Accelerated Solvent Extraction System Hirotaka Obana*a, Kosuke Kikuchib, Masahiro Okihashia and Shinjiro Horia a Osaka Prefectural Institute of Public Health, 1-3-69, Nakamichi, Higashinarik-ku, Osaka, 537, Japan. E-mail: obana@iph.pref.osaka.jp b Japanese Consumers’ Co-operative Union, 1-17-18, Nishiki-cho, Warabi, 335, Japan Residual organophosphorus pesticides in foods were determined by accelerated solvent extraction (ASE), gel permeation chromatography and GC–FPD.Pesticides were extracted at 100 °C under 1500 psi pressure in less than 20 min. Wet samples were extracted after mixing with Extrelut drying agent. Four foods were spiked with 19 pesticides at 0.1 ppm or less. The average recoveries of these pesticides were 80–90% and the precision was < 10%. Both methamidophos and acephate gave 37–50% recoveries and 6–40% precision.These results suggested that ASE can be used to extract residues of organophosphorus pesticides in foods. Keywords: Organophosphorus pesticides; foods; accelerated solvent extraction; gas chromatography–flame photometric detection Laboratories that monitor pesticides have developed various means of measuring residual levels in foods. Multiresidue methods are usually applied in pesticide monitoring because several compounds can be targeted simultaneously, which reduces labor and material costs.Extraction is one of the most important steps in the analysis. Conventional manual liquid– liquid partition consumes time and it is labor intensive. Moreover, it sometime results in an emulsion that can lower the reproducibility of the analysis. Japanese laboratories cannot freely drain waste containing dichloromethane owing to legal restrictions ( < 10 mg ml21).1 Dichloromethane is a frequently used solvent in multiresidue methods,2,3 so an alternative method is required that incorporates non-halogenated solvents.The latest automated extraction methods use less solvent and give improved analytical precision. Supercritical fluid extraction (SFE) is an example that has been adapted to residual pesticides analysis.4–8 Pesticides have also been analyzed after microwave-assisted extraction.9,10 On the other hand, Richter et al.11 reported that environmental contaminants such as PCBs and PAHs were rapidly and effectively extracted at temperatures above 100 °C with pressurized solvents.They showed that solvents solubilize pesticides and penetrate sample matrices better under these conditions. Both accelerated solvent extraction (ASE) and microwave extraction operate under the same principle of automated solvent extraction at high temperature. In this study, we extracted organophosphorus pesticides from foods using ASE. Experimental Pesticide Standards Organophosphorus pesticides were purchased from Wako (Osaka, Japan): acephate, chlorpyrifos, chlorpyrifos-methyl, diazinon, dichlorvos, dimethoate, dioxabenzofos, edifenphos, EPN, ethion, ethoprophos, fenthion, fenitrothion, malathion, methamidophos, methidathion, phosalone, pirimiphos-methyl and prothiofos.Each compound was dissolved in hexane to make a 1000 mg ml21 stock standard solution. The stock standard solution was diluted with acetone and used in the recovery tests at 1 mg ml21.Organophosphorus pesticides cause matrix effects that result in over-estimation of the recoveries in GC analysis.6,7 To prevent this, standard mixtures were diluted with the respective cleaned-up extract from pesticide-free samples. Reagents Acetone, hexane, ethyl acetate and anhydrous Na2SO4 were of pesticide analysis grade and cyclohexane was of HPLC grade. All reagents were purchased from Wako. Diatomaceous Earth Particles of diatomaceous earth (Extrelut for refilling; particle size 160–800 mm) (Merck, Darmstadt, Germany) were heated at 550 °C for 15 h to exclude interferences in the GC analysis.Food Preparation Samples were purchased at a local market in Osaka and before use we confirmed that organophosphorus pesticide residues were below detectable levels. Pesticide residues were analyzed using a liquid–liquid extraction described by Yoshida et al.12 About 500 g of broccoli or grapefruit were chopped in a conventional food processor (MK-K3, Matsushita, Osaka, Japan) for 5 min to obtain thoroughly mixed homogenates. The operating manual recommended that wet samples should be mixed with a drying agent to facilitate solvent penetration into sample matrices.13 An aliquot of 5 g of sample homogenate and 6 g of Extrelut particles was ground in a mortar (12 cm id) with a pestle until the mixture became homogeneous.In the fortification study, 0.5 ml of pesticide mixture at 1 mg ml21 was added to the sample homogenate to a final concentration of 0.1 ppm on a sample mass basis and the mixture was left for 30 min before adding Extrelut.Pesticides were added to flour at a final concentration of 0.05 ppm, then extracted without extra Extrelut. The procedure was performed twice for each sample because of GC separation preferences. One group consisted of methamidophos, acephate and dimethoate and the other of the remaining 16 pesticides. Mixtures of samples and Extrelut were placed in stainless-steel cells (33 ml; 11 cm 3 1.9 cm id).ASE Accelerated solvent extraction was performed with a Dionex AS 200 system (Dionex, Sunnyvale, CA, USA). The pre-set default conditions were as follows: extraction temperature, 100 °C; extraction pressure, 1500 psi (10.5 kPa); pre-heating period, 5 min; static extraction period, 5 min; solvent flash, 19.8 ml; nitrogen purge, 60 s; and collection, in 60 ml glass vials with Teflon coated rubber caps (I-CHEM, New Castle, DE, USA). The extraction cell was heated and pressurized immediately after introducing it into the heating port.The pressure valve was Analyst, March 1997, Vol. 122 (217–220) 217sometimes opened to maintained the pre-set pressure, and the extracted solvent was collected from pre-heating. After extraction, the cell was purged with fresh extraction solvent and additional nitrogen to collect the remaining extract in the cell. The extraction process was automated according to the assigned extraction method. The extracts were usually suspended and the water layer which lay separated in the bottom of vial because it was also co-extracted with solvent.The extract was transferred into a separating funnel. The collecting vial was washed with 10 ml of hexane twice, then hexane was added to the extract. As the suspension of the extract became heavier with hexane, the organic layer was left for about 2 h to form water droplets from the organic layer. After the water portion had been discarded, the extract was dried by adding anhydrous Na2SO4.The extract was carefully evaporated almost to dryness with warming at < 40 °C and the residue was dissolved in ethyl acetate– cyclohexane (1 + 1) for gel permeation chromatography (GPC) and diluted to 7 ml. GPC14 Extracts were injected into an AS-2000 system (abc Laboratories, Columbia, MO, USA) equipped with an Environsep- ABC column (300 3 21.2 mm id). The mobile phase was ethyl acetate–cyclohexane (1 + 1) at a flow rate of 5 ml min21.Food extract (5 ml) was loaded into the GPC system. The first 70 ml were discarded and the following 80 ml were collected as the pesticide fraction. A further 50 ml were eluted for the GPC wash. The pesticide fraction was concentrated to 3.5 ml with an evaporator for GC determination. GC–FPD Determination Pesticides were determined using a GC 17A instrument (Shimadzu, Kyoto, Japan) equipped with a flame photometric detector. A DB-210 column (5 m 3 0.25 mm id, 0.25 mm thickness) (J & W Scientific, Folsom, CA, USA) was used for methamidophos, acephate, and dimethoate determinations.The temperature program was initial temperature 80 °C (held for 1 min), increased at 10 °C min21 to 260 °C (held for 4 min). A DB-5 column (30 m 3 0.25 mm id, 0.25 mm thickness) (J & W) was used for the other pesticides. The temperature program was initial temperature 80 °C (held for 1 min), increased at 20 °C min21 to 120 °C (no hold), then at 4 °C min21 to 280 °C (held for 5 min).Carrier gas was helium. A 2ml volume was injected in the splitless mode at 250 °C. The detector temperature was set at 300 °C. The chromatogram was recorded using a CR7A integrator (Shimadzu). Results and Discussion Extraction From Flour First, a recovery test was performed with flour with a low water content, so that no drying agent needed to be added to the sample during extraction. Flour (10 g) spiked with 0.05 ppm of Table 1 ASE extraction of 17 organophosphorus pesticides in flour using three solvent mixtures Cyclohexane– Dichloromethane– Ethyl acetate– acetone (1 + 1) acetone (1 + 1) acetone (1 + 1) Mean Mean Mean recovery RSD recovery RSD recovery RSD Pesticide (%)* (%) (%)* (%) (%)* (%) Dichlorvos 36.6 12.1 42.9 37.0 46.6 30.3 Ethoprophos 100.7 2.5 95.1 3.8 141.4 24.1 Dioxabenzofos 100.8 3.6 83.2 18.1 95.5 27.7 Dimethoate 49.1 1.8 86.0 16.1 199.2 34.4 Diazinon 99.2 2.4 93.5 1.6 134.3 26.7 Chlorpyrifos-methyl 100.7 5.1 92.9 13.6 104.6 28.4 Fenitrothion 111.7 5.3 90.7 12.8 105.9 29.0 Pirimiphos-methyl 96.4 4.8 106.0 0.8 155.4 24.3 Malathion 106.3 4.6 93.4 5.4 137.8 19.1 Chlorpyrifos 94.9 3.4 93.8 7.0 125.3 25.2 Fenthion 90.1 4.8 88.3 0.1 146.1 24.4 Methidathion 113.6 14.1 88.5 13.9 106.9 32.1 Prothiofos 89.2 5.5 92.6 5.8 128.6 26.0 Ethion 85.5 7.3 96.3 5.2 142.3 25.6 Edifenphos 115.8 11.7 91.2 6.6 141.3 28.6 EPN 87.1 10.5 93.5 11.3 128.6 31.5 Phosalone 106.5 10.9 93.4 7.3 140.6 33.2 * Means of three experiments.Table 2 ASE extraction of 19 organophosphorus pesticides from grapefruit and orange juice by ASE using cyclohexane–acetone (1 + 1) Grapefruit Orange juice Mean Mean recovery RSD recovery RSD Pesticide (%)* (%) (%)* (%) Dichlorvos 82.8 15.3 87.1 10.5 Methamidophos 8.7 20.1 9.3 48.5 Ethoprophos 99.6 13.9 61.1 13.5 Dioxabenzofos 100.0 19.6 51.6 16.8 Acephate 10.2 19.4 10.6 45.3 Dimethoate 97.9 14.1 74.7 8.0 Diazinon 102.0 12.6 73.4 7.5 Chlorpyrifos-methyl 105.1 17.6 75.7 6.0 Fenitrothion 99.7 8.2 76.5 10.9 Pirimiphos-methyl 101.7 14.0 88.4 4.5 Malathion 98.2 9.1 76.3 6.3 Chlorpyrifos 105.1 10.4 79.2 7.5 Fenthion 105.6 10.2 80.9 5.9 Methidathion 100.5 4.4 73.7 9.0 Prothiofos 103.1 9.6 75.2 9.7 Ethion 97.5 21.8 110.5 19.1 Edifenphos 100.9 14.9 74.1 13.6 EPN 107.1 13.4 87.6 15.3 Phosalone 90.3 13.8 70.3 14.3 * Means of three experiments. 218 Analyst, March 1997, Vol. 122each pesticide was extracted with three solvent mixtures each containing 50% acetone.As shown in Table 1, extraction with cyclohexane–acetone (1 + 1) and dichloromethane–acetone (1 + 1) resulted in good recoveries. The recoveries of most of the pesticides was 83–115%. The RSDs were acceptable, usually < 10%, using cyclohexane–acetone (1 + 1). With ethyl acetate– acetone (1 + 1) the RSD varied from 19 to 34%. The recovery of dichlorvos was around 40% with all three mixtures and the precision was also poor.One explanation is that the compound was lost during sample preparation because dichlorvos is volatile with a vapor pressure of 290 mPa at 20 °C.15 Extraction From Wet Samples Since one aim of this study was to develop a dichloromethanefree extraction procedure, we extracted pesticides from grapefruit and orange juice with cyclohexane–acetone (1 + 1) (Table 2). In this study methamidophos and acephate were added in addition to the 17 pesticides in Table 1. The recoveries of the 17 pesticides except methamidophos and acephate varied from 82 to 105% in grapefruit and from 51 to 110% in orange juice.Little methamidophos or acephate was extracted under the conditions tested. The extraction precision of these pesticides was poor in orange juice. Extraction Conditions We studied the effects of static extraction time, temperature and pressure to improve the extraction rates of methamidophos and acephate in orange juice (Table 3). Other conditions were the same as those described above.Since two of the pesticides were water soluble,16 orange juice seemed to be one of the most difficult foods from which to obtain satisfactory recoveries. The appropriate conditions for orange juice should also be sufficient for drier samples. Time and temperature did not improve the recoveries under the conditions tested. Only a higher extraction pressure slightly improved the recovery. Further improvement by increasing the pressure was unlikely since the mechanical pressure limit of the ASE system is 3000 psi. Table 3 Effects of extraction time, temperature and pressure on recoveries of methamidophos and acephate* Recovery (%) Variable Methamidophos Acephate Time/min— 5 15.4 9.3 10 11.9 9.6 15 11.7 7.7 Temperature/°C— 50 5.8 4.5 100 8.1 8.0 150 6.4 4.2 Pressure/psi— 1500 9.5 16.0 2000 13.3 20.0 2500 15.0 22.2 * The extraction solvent was cyclohexane–acetone (1 + 1) and other conditions were the default values as described in the text.Values are means of two experiments. Table 5 ASE extraction of 19 organophosphorus pesticides in foods using ethyl acetate Orange juice Grapefruit Broccoli Flour (0.1 ppm) (0.1 ppm) (0.1 ppm) (0.05 ppm) Mean Mean Mean Mean recovery RSD recovery RSD recovery RSD recovery RSD Pesticide (%)* (%) (%)* (%) (%)* (%) (%)* (%) Dichlorvos 91.5 4.6 102.8 10.0 87.4 3.9 92.7 9.9 Methamidophos 56.6 7.9 60.3 17.0 53.7 6.2 37.3 41.5 Ethoprophos 68.9 6.5 91.4 4.7 114.0 4.8 90.1 9.8 Dioxabenzofos 67.0 4.9 88.6 6.8 97.0 4.8 60.3 7.9 Acephate 54.8 9.0 59.1 20.9 50.8 6.1 55.7 10.1 Dimethoate 91.8 5.0 109.3 9.9 103.2 3.0 83.5 9.2 Diazinon 81.9 4.3 86.6 5.2 88.0 4.6 90.9 2.2 Chlorpyrifos-methyl 76.9 2.9 87.0 5.6 87.8 4.6 72.3 3.1 Fenitrothion 97.6 7.3 89.0 6.0 102.8 5.5 91.4 4.3 Pirimiphos-methyl 85.3 2.2 85.5 7.0 88.0 5.4 105.7 1.7 Malathion 81.9 4.1 89.5 5.9 99.5 5.1 77.5 0.4 Chlorpyrifos 74.1 2.7 87.3 6.2 81.4 5.1 95.2 2.7 Fenthion 79.6 4.8 90.3 4.2 91.4 7.7 92.3 2.3 Methidathion 80.9 11.3 91.4 5.2 110.9 6.7 73.5 4.7 Prothiofos 68.9 4.4 90.1 3.7 88.5 6.2 96.1 2.3 Ethion 78.1 5.8 90.1 4.1 87.5 9.0 98.1 2.1 Edifenphos 90.9 7.4 90.9 5.4 101.5 7.2 67.9 5.1 EPN 83.1 9.1 85.7 3.5 91.9 5.7 98.4 2.0 Phosalone 97.5 9.5 84.0 6.2 93.4 6.4 96.8 1.5 Average 79.3 87.4 90.4 82.9 * Means of three experiments.Table 4 ASE extraction of methamidophos and acephate from orange juice using various solvents Recovery (%)* Solvent Methamidophos Acephate Cyclohexane–acetone (1 + 1) 9.3 10.6 Toluene–acetone (1 + 1) 46.6 45.3 Dichloromethane–acetone (1 + 1) 36.0 36.5 Ethyl acetate–acetone (1 + 1) 16.1 30.5 Acetonitrile 28.2 30.6 Ethyl acetate 56.1 47.8 * Means of two experiments.Analyst, March 1997, Vol. 122 219We studied the effects of extraction solvents on the recoveries of methamidophos and acephate in orange juice under the default extraction conditions (Table 4). Ethyl acetate recovered 56 and 47% of spiked methamidophos and acephate, respectively.Toluene–acetone (1 + 1) gave a secondary recovery of around 45%. The recovery with cyclohexane– acetone (1 + 1) was the lowest among the six solvent systems tested. Since ethyl acetate gave the best results, it was used to extract 19 pesticides from four spiked foods. As shown in Table 5, the average recoveries of the 19 pesticides in the four samples were considerably better: 79.3% in orange juice, 87.4% in grapefruit, 90.4% in broccoli and 82.9% in flour.The recovery of methamidophos and acephate ranged from 50 to 60% except for methamidophos in flour (37%). Most RSDs were < 10%. Although methamidophos and acephate tend to show poor precision, the poor reproducibility of these pesticide peaks on the chromatograms in GC–FPD might have caused the fluctuating recoveries. These results suggested that ethyl acetate is a useful solvent in ASE with which to determine residual organophosphorus pesticides.ASE extraction was compared with hexane extraction in pesticide-containing samples (Table 6). These samples were found in our routine surveillance and stored at 220 °C until analysis. Whereas the ASE extraction seemed to give slightly lower recoveries for the samples tested, the precision of ASE is better than that of hexane extraction.The colours of the hexane extracts were deeper than those in ASE extracts. The advantages of ASE over liquid–liquid extraction were as follows.The extraction procedure is simple, requiring only mixing of the samples with drying agent and transferring the mixture to an extraction cell. One sample can be extracted within 20 min, including washing the solvent line for the next extraction. The solvent volume with a 33 ml cell is 50–60 ml, which depends on the sample volume. Even though the volume is not large, the ratio of solvent (50 ml) to sample (5 g) is 10, whereas it is 2–5 in conventional procedures.2,14,17,18 Hence ASE should combine good recovery and precision with rapid extraction.Another advantage over the conventional method is continuous separation of the extract from sample residues during extraction. The disadvantage of ASE in this study is the 2 h waiting to remove water from the extract after adding hexane. Without hexane, small water droplets appeared after evaporation even when the extract was dried with anhydrous Na2SO4. Ethyl acetate at elevated temperature solubilized water to a greater extent during extraction than by conventional means at room temperature. Water remaining in the extract decreased the precision of the recovery tests (data not shown).Although thermal degradation of the target compounds is a potential concern because of elevated temperature during extraction, Richter et al.11 reported that the thermal degradation of DDT to DDE or DDD and of endrin to endrin aldehyde or endrin ketone did not occur during ASE at 150 °C.This study demonstrated that ASE automatically and rapidly extracted organophosphorus pesticides from foods with good accuracy and precision. ASE could be introduced as a means of determining residual levels of organophosphorus pesticides in foods. We are studying its applicability to multiresidue analyses of agrochemicals other than organophosphorus pesticides. References 1 Prime MinisterAs Office of Japan, Bulletin 54, December 27, 1993 (in Japanese). 2 Cairns, T., Luke, M. A., Chiu, K. S., Navarro, D., and Siegmund, E. G., Rapid Commun. Mass Spectrom., 1993, 7, 1070. 3 Luke, M. A., Froberg, J. E., Doose, G. M., and Masumoto, H. T., J. Assoc. Off. Anal. Chem., 1981, 64, 1187. 4 Skopec, Z. V., Clark, R., Harvey, P. M. A., and Wells, R. J., J. Chromatogr. Sci., 1993, 31, 445. 5 Hopper, M. L., and King, J. W., J. Assoc. Off. Anal. Chem., 1991, 74, 661. 6 Lehotay, S. J., Aharonson, N., Pfeil, E., and Ibrahim, M. A., J. AOAC Int., 1995, 78, 831. 7 Lehotay, S. J., and Eller, K. I., J. AOAC Int., 1995, 78, 821. 8 Stuart, I. A., MacLachlan, J., and McNaughtan, A., Analyst 1996, 121, 11R. 9 Lopezavila, V. L., Young, R., Benedicto, J., Ho, P., Kim, R., and Beckert, W. F., Anal. Chem., 1995, 67, 2096. 10 Lopezavila, V., Young, R., Kim, R., and Beckert, W. F., J. Chromatogr. Sci., 1995, 33, 481. 11 Richter, B. E., Jones, B. A., Ezzell, J. L., Porter, N. L., Avdalovic, N., and Pohl, C., Anal. Chem., 1996, 68, 1033. 12 Yoshida, S., Konaka, H., and Nishimune, T., J. Food Hyg. Soc. Japn., 1991, 32, 267. 13 ASE 200 Accelerated Solvent Extractor Operator’s Manual, Document No. 031149, Revision 01, Dionex, Sunnyvale, CA, 1995, Sect. 3–5. 14 Committee on Analytical Methods for Pesticide Residues, Food Sanit. Res., 1995, 45, 31 (in Japanese). 15 The Agrochemical Handbook, Royal Society of Chemistry, Cambridge, 3rd edn., 1991, p. A0141. 16 The Agrochemical Handbook, Royal Society of Chemistry, Cambridge, 3rd edn., 1991, pp. A0001 and A0265. 17 Pesticide Analytical Manual, ed. McMahon, B. M., and Hardin, N. F., Food and Drug Administration, Washington, DC, 3rd edn., 1994, vol. 1, pp. 302–305. 18 Fillion, J., Hindle, R., Lacroix, M., and Selwyn, J., J. AOAC Int., 1995, 78, 1252. Paper 6/06063C Received September 3, 1996 Accepted October 28, 1996 Table 6 Comparison between ASE and hexane extraction in pesticidecontaining samples ASE Hexane Mean RSD Mean RSD Food Pesticide (ppm)* (%) (ppm)* (%) Banana Chlorpyrifos 0.03 9.7 0.05 13.4 Okra Phosalone 0.05 4.0 0.08 27.5 Sweetie Chlorpyrifos 0.22 2.7 0.27 13.2 Malathion 0.03 20.9 0.03 35.9 * Means of three experiments. 220 Analyst, March 1997, Vol. 122

ISSN:0003-2654    

DOI:10.1039/a606063c

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Simultaneous Spectrofluorimetric Determination of Selenium(IV) and (VI) by Flow Injection Analysis M. J. Ahmed†, C. D. Stalikas, P. G. Veltsistas, S. M. Tzouwara-Karayanni and M. I. Karayannis* Laboratory of Analytical Chemistry, Department of Chemistry, University of Ioannina, Ioannina 451 10, Greece A simple, sensitive, highly selective, automatic spectrofluorimetric method for the simultaneous determination of selenium(IV) and (VI) as selenite–selenate by flow injection analysis (FIA) has been developed.The method is based on the selective oxidation of the non-fluorescent reagent 2-(a-pyridyl)thioquinaldinamide (PTQA) in acidic solution (1.5–3.0 m H2SO4) by SeIV to give an intensely fluorescent oxidation product (lex = 350 nm; lem = 500 nm). Selenium(VI) is reduced on-line to SeIV, in a reduction coil installed in a photo-reactor, which is then treated with PTQA and the fluorescence due to the sum of SeIV and SeVI is measured; SeVI is determined from the difference in fluorescence values.Various analytical parameters, such as effect of acidity, flow rate, sample size, dispersion coefficient, temperature, reagent concentration and interfering species were studied. The photo-reduction conditions were optimized, with an FIA procedure, for SeVI on the basis of its reduction efficiency. The calibration graphs were rectilinear for 0.1–2.4 mg ml21 of SeVI and 10 ng ml21–2.2 mg ml21 of SeIV, respectively. The method was applied to the determination of Se in several Standard Reference Materials (alloy, sediments and tea), as well as in some environmental waters (tap and surface water), food samples (flour and egg), a biological sample (human hair), soil sample and in synthetic mixtures.Up to 25 samples per hour can be analysed with an RSD Å 0.1–2%. Keywords: Flow injection; spectrofluorimetry; selenium speciation; 2-( a-pyridyl)thioquinaldinamide; on-line photo-reduction; environmental; biological; soil samples Recently, there has been increasing interest in trace determination of selenium because of its dual role, as an essential nutrient at low concentrations (10–40 mg ml21 in serum), or a highly toxic compound (selenosis) at an intake of 5 mg kg21 of Se (in the mammals of a seleniferous region).1 It is contained in the enzyme glutathione peroxidase (GSHPx), which affords cells protection against oxidative damage.2 A selenium deficiency in man may also result in cardiomyopathy.2 The narrow concentration range between the two opposite effects (0.1–4.0 mg kg21 in plants), requires accurate and precise knowledge of the selenium species present in the environment.3 In the environment, Se levels generally fall in the ranges 0.1–400 ng ml21 in natural waters,4 1 ng ml21 in the atmosphere3 and 0–80 mg g21 in soils.5 Selenium finds its way into the environment through its widespread use in the glass and electronics industries, as well as from the combustion of fossil fuels and uses in agriculture.Detailed information about the availability and mobility of Se in the environment and its biogeochemical cycle, however, requires the additional knowledge of the different chemical forms and oxidation states in which this element can exist. The inorganic Se species most frequently found in water and soils are selenite (SeO3 22) and selenate (SeO4 22).3 In this regard, a method for speciation is needed because its availability for plant uptake, mobility in soil, and toxicity in biota depends on the oxidation state of the Se.6 Therefore, its accurate determination at trace levels using simple and rapid methods is of paramount importance.Analytical techniques for Se speciation have recently been reviewed.7–9 Automatic flow techniques have hardly been applied to the determination of Se. Until now, only few methods have been described that use flow injection analysis (FIA) and these methods either use FIA combined with HG–AAS8,10 or FIA with spectrophotometric detection.11 But these methods suffer from several limitations: (i) indirect determination of SeVI because it does not give hydrides, and (ii) the matrix affects both hydride generation and reduction yields of SeVI to SeIV. The spectrophotometric methods suffer from sensitivity and selectivity due to various or many interferences.In this respect, interlaboratory comparisons of inorganic Se in biological and environmental samples showed unacceptable differences using these techniques.12 Recent developments in the automation of instrumentation leads to an improved precision.13 Recently, photochemical reactions have been applied to the on-line reduction and oxidation of inorganic and organic substances in flow injection methods.14,15 Measures and Burton16 studied the photochemical oxidation of an organometallic form of selenium( Se22) into inorganic SeIV with a high pressure mercury lamp.All of these systems have been used for on-line photochemical oxidation or reduction to enhance the detection of a variety of inorganic and organic compounds with AAS17 and spectrophotometric techniques.15 The aim of the present study was to develop a more simple FIA system for the simultaneous determination of SeIV and SeVI with 2-(a-pyridyl)thioquinaldinamide (PTQA) using a reduction coil installed in a photo-reactor in the reaction manifold.PTQA has been reported as a spectrofluorimetric reagent,18 but has not previously been used for the simultaneous determination of SeIV and SeVI in a flow injection system.This paper reports its use in a very sensitive, highly specific automatic spectrofluorimetric method for the simultaneous determination of SeIV and SeVI. The method is based on the selective oxidation of the nonfluorescent reagent, PTQA, in an acidic medium (1.5–3.0 m H2SO4) by SeIV to produce an intensely fluorescent product followed by the direct measurement of the fluorescence intensity in aqueous solution at room temperature.Oxidation is very rapid and no extraction is required. With suitable masking the reaction can be made highly selective. The reaction mechanism of the present method is as reported earlier.18 Experimental Apparatus The manifold for simultaneous determination of SeIV and SeVI was of Teflon tubing (0.8 mm id) and linear dual connectors were used (Fig. 1). It consisted of a four-way pneumatically † Present address: Department of Chemistry, University of Chittagong, Chittagong- 4331, Bangladesh.Analyst, March 1997, Vol. 122 (221–226) 221actuated injection valve (Rheodyne, Type 50 Teflon, Cotati, CA, USA), an eight-channel peristaltic pump (Ismatec, Glattburg- Zurich, Switzerland) and a spectrofluorimeter (RF-551, Shimadzu, Japan), equipped with a 12 ml flow-through cell for measurement. Data processing and collection was performed with an IBMcompatible Personal Computer (PC) by means of software written in Microsoft Q-Basic.The interface unit was an RTL 800/815 multifunction Input/Output board. A Varian AA-300 Atomic Absorption Spectrophotometer equipped with a hydride system at 196.1 nm using an air–acetylene flame was used for comparison of the results. A digital pH-meter (Model-PHM83 AUTOCAL, Radiometer, Copenhagen, Denmark) was used to measure the pH of the solutions. Photoreduction-reactor The photoreduction-reactor comprised a high pressure mercury light source (2 cm od, 25 cm long, 125 W, DESAGA) and a quartz coil (40 cm long 3 0.8 mm id).The source emits short wavelength light at lmax = 254 nm. The effective irradiation length was 6 cm. The unit is covered with aluminium foil or thick paper in order to increase the light intensity reaching the coil by reflectance and to prevent eye exposure to ultraviolet radiation. Reagents All chemicals used were of analytical-reagent grade or the highest purity available.Doubly distilled water and HPLCgrade propan-2-ol, which is non-fluorescent under ultraviolet radiation, were used throughout. SeIV standard solutions. A 100 ml amount of stock SeIV solution (1 mg ml21) was prepared by dissolving 333.1 mg of general-reagent grade sodium selenite (Merck, Darmastadt, Germany) in doubly distilled water. The solution was kept in a refrigerator in a polyethylene container for preservation. Working standard solutions were prepared daily by appropriate dilution in 2 m H2SO4.SeVI standard solutions. A 100 ml amount of stock SeVI solution (1 mg ml21) was prepared by dissolving 467.4 mg of ACS-grade sodium selenate (99%, Aldrich, Steinheim, Germany) in doubly distilled water. The solution was kept in a refrigerator in a polyethylene container. Working standard solutions were prepared daily by appropriate dilution in 2 m H2SO4. Carrier solution. 2.0 m H2SO4 (Merck) was used as the carrier solution. 2-( a-Pyridyl)thioquinaldinamide (PTQA), solution (1023 M).The reagent was synthesized according to the method of Porter.19 The contents, containing 2-aminopyridine (2 mol), quinaldine (1 ml) and sulfur powder (1.5 mol), were mixed and refluxed for 6 h in a 250 ml round bottomed flask fitted with bulb condenser under controlled temperature (140–150 °C) at 1 atm pressure over a sand bath. The reaction mixture was kept overnight. The thio-compound was filtered and crystallized using petroleum ether to give a bright yellow crystalline (needle) solid.The compound recrystallized from ethanol was dried under vacuum (0.1 mg of Hg) for 24 h. The melting point of this synthesized compound (PTQA) was 155 ± 2 °C and the elemental analysis data (C = 72.25, N = 13.35 and H = 4.25%) were very close to the literature values,19 e.g., melting point (155 ± 1 °C); C (N72.43); N (N13.55) and H (N4.55%). The reagent solution (1023 m) was prepared by dissolving the requisite amount (0.0266 g in 100 ml) of PTQA in HPLC-grade propan-2-ol.A freshly prepared reagent solution (1024 m) was used whenever required. Other solutions. Solutions of a large number of inorganic ions and complexing agents were prepared from their AnalaR grade or equivalent grade water soluble salts. In the case of insoluble substances, special dissolution methods were adopted.20 Stock solutions and environmental samples were kept in a refrigerator in poly(propylene) bottles.Preparation of the Samples Food samples (rice flour and egg) were purchased from a local supermarket. These samples were homogenized thoroughly. Soil samples was collected from local agricultural field sites and homogenized in a mortar. Human hair (3–5 cm long from male) was cut from the occipitonuchal region of the head. They were cleaned by stirring with acetone, rinsing with tap water, stirring in a detergent solution (which had no detectable Se), rinsing with tap water, doubly distilled water and finally with acetone.They were then dried at 45 °C and cut into small pieces for analyses (sampled person did not use Se-containing shampoo). Procedure The standards (0.01–2.2 mg ml21 SeIV or 0.1–2.4 mg ml21 SeVI) and samples were injected into a carrier stream by means of the peristaltic pump, P (Fig. 1). Then the sample was measured by different ways using a selector valve. The sample stream was firstly directed through path 1, treated with a 10–30-fold molar excess of the PTQA reagent solution and passed directly into the measuring cell of a spectrofluorimeter where the fluorescence intensity due to SeIV was measured at 500 nm with excitation at 350 nm.Then the sample stream was passed through path 2 to photo-reduction coil (R1) by using a second selector valve where SeVI was reduced to SeIV. The sample stream was then treated with the PTQA reagent at the end of the coil and the overall mixture was passed to the same cell of the spectrofluorimeter where the fluorescence intensity due to total Se was measured; SeVI was determined from the difference in fluorescence intensity values.The reaction is very rapid and the fluorescence intensity remains stable for 24 h. The PTQA reagent does not show any fluorescence in the absence of SeIV. The concentrations of SeIV and SeVI were evaluated from the peak heights of the signal by using the calibration curves prepared with standard solutions.Results and Discussion Optimization of the Flow Injection System Preliminary tests were carried out with the aid of different flow assemblies to select the optimal manifold configuration. The assembly in Fig. 1 was selected as the one producing the best compromise between peak height and the shape of the peak. Fig. 1 Schematic representation of FI manifold employed for the simultaneous determination of SeIV and SeVI. P, Pump; V, valve; S, selector valve; R1, photoreduction coil; R2, single bead string reactor (SBSR); D, detector; W, waste; and PC, personal computer. 222 Analyst, March 1997, Vol. 122In order to optimize the proposed flow injection manifold, the influence of the hydrodynamic and chemical parameters on the magnitude of the peak height, the shape of the peak and reproducibility of the results were studied. The univariate method was adopted for the optimization of the system. Table 1 shows results of optimization of working conditions for 0.5 mg ml21 of SeIV and 1.0 mg ml21 of SeVI.The optimum length of the photo-reduction coil (R1) was established by using a 1.0 mg ml21 SeVI solution, the single bead string reactor (SBSR) (R2), for better mixing and lower dispersion, being of length 100 cm and having an acidity of 1.5–3.0 m H2SO4. Photo-reduction coil (R1) lengths of 15, 30, 40, 60 and 80 cm were tested, keeping the power of the lamp and its distance from reaction coil constant. For any combination of the above parameters the efficiency of the reduction was determined by comparison of the plateau achieved with that corresponding to a 1.0 mg ml21 SeIV solution processed in the same way.A coil length of 40 cm was chosen, because the reduction was almost complete, reproducibility was good and back-pressure relatively low. Different wavelengths of the UV radiation were also tested. For each wavelength, the efficiency of the reduction was determined. The effective wavelength of photo-reduction of the SeVI was lmax = 254 nm.Different lamp powers were also tested but no significant effect on reduction was observed. Different distances of the reaction coil from the lamp were tested keeping the length of the coil and wavelength constant. A length of 6 cm was selected because maximum conversion efficiency was achieved. A length 100 cm for the SBSR reactor (R2), a sample size of 100 ml, an overall flow rate of 0.4 ml min21 and a reagent flow rate of 0.3 ml min21 were selected, these being a compromise between the sampling rate and the height of the peak.Of the various acids (sulfuric, hydrochloric, nitric and phosphoric) studied, sulfuric acid was found to be best acid for the system. Different concentrations of sulfuric acid were tested in the range shown in Table 1. The fluorescence intensity was at maximum and constant when the solution (1.0 mg ml21) contained 1.5–3.0 m H2SO4 (Fig. 2) at room temperature, which was the optimum acidity range.The photoreduction efficiency for 1.0 mg ml21 of SeVI in this acidity range was also tested. More than 97% of the SeVI can be reduced to SeIV in this acidity range. For all subsequent measurements 2.0 m H2SO4 was used as carrier for this manifold. The effect of propan-2-ol on the fluorescence was studied and no adverse effect was observed over a wide range of propan- 2-ol concentrations. A 1024 m solution of PTQA in propan-2-ol was sufficient to prevent any precipitation or turbidity or bubbling and to allow accurate measurements for this manifold.Other common organic solvents, e.g., chloroform, benzene, tetrachloromethane and ethanol, were also tried but no fluorescence was observed in the organic phase, with the exception of ethanol. The reaction is rapid. A constant maximum fluorescence intensity was obtained just after the dilution to volume and remained strictly unaltered for 24 h. Different concentrations of PTQA solution were tested in the ranges shown in Table 1.A reagent concentration of 2 3 1024 m was selected as optimum for this manifold. The fluorescence intensity of the SeIV–PTQA system within the prescribed acidity range was maximum and constant for SeIV to reagent molar ratios in the range 1 : 10–1 : 30 when the Se concentration was 1.0 mg ml21 (Fig. 3). At different SeIV concentrations (0.1 and 0.5 mg ml21), the effect of varying the reagent concentration was similar. Evaluation of the Method The reproducibility of the proposed procedure and sample throughput were determined by repeated injection of a sample containing 0.5 mg ml21 SeIV and 1.0 mg ml21 SeVI.The RSD (n = 5) was 0.1–2% for 0.01–2.2 mg ml21 SeIV and 0.1–2.4 mg ml21 SeVI indicating that this method is highly precise and reproducible. The calibration graphs obtained from the peak heights were rectilinear for 10 ng ml21 to 2.2 mg ml21 of SeIV and 0.1–2.4 mg ml21 of SeVI, respectively. The detection limits, defined as three times the baseline noise, were 1 ng ml21 for SeIV and 10 ng ml21 for SeVI.The sample throughput was 25 measurements per hour. The dispersion coefficients were estimated with a 0.5 mg ml21 SeIV and 1.0 mg ml21 SeVI standard solutions as described earlier.21 Important features of the proposed method for simultaneous determination of SeIV and SeVI are summarised in Table 2. The performance and reproducibility of the proposed method are also shown in Tables 3–6.The reliability of the proposed procedure was also assessed by analysing Certified Reference Materials. The results for total Se were in good agreement with certified values (Table 3). The method was also tested by analysing several synthetic mixtures containing standard SeIV and SeVI (Table 4). The reliability of the proposed procedure was also tested by performing recovery studies. The average Table 1 Selected chemical and FIA parameters obtained with the optimization experiments Parameter Studied range Selected value Size of sample loop/ml 30–180 100 Overall flow rate/ ml min21 0.20–1.0 0.40 Reagent flow rate/ ml min21 0.05–0.60 0.30 Length of the photoreaction coil, R1/cm 15–80 40 Length of the SBSR reactor, R2/cm 20–180 100 pH 0.1–1.3 0.40–0.75 (preferably 0.6) Concentration of reagent ( m)— H2SO4 0.50–4.5 1.50–3.0 (preferably 2.0) PTQA 331025–631024 231024 Fig. 2 Effect of acidity on the fluorescence intensity of the SeIV–PTQA system.Fig. 3 Effect of reagent on the fluorescence intensity of the Seiv–PTQA system. Analyst, March 1997, Vol. 122 223percentage recovery obtained for the addition of SeIV and SeVI spikes to some environmental water and flour samples was quantitative as shown in Table 5. The results of food, hair and soil analyses by the present method was in excellent agreement with those obtained by HG-AAS (Table 6). The precision and accuracy of the method are satisfactory. The interference of several ions which may occur in environmental samples was studied by using a solution containing a mixture of SeIV and SeVI at concentrations of 0.5 mg ml21 and 1.0 mg ml21, respectively, adding various concentrations of interfering ions up to the amounts where the relative error reached a value of about 5%.The errors were calculated by comparing the peak height to that obtained after the injection of an aqueous solution of SeIV and SeVI containing no interfering ions, as a reference.The results are summarized in Table 7. During the interference studies, if a precipitate was formed, it was removed by centrifugation. Positive interference from permanganate or hydrogen peroxide was eliminated by boiling the solution with sodium azide, a reducing agent which had no reducing effect on either SeIV or SeVI. Applications The proposed method was used to determine the total selenium content in a number of certified reference materials (sediments, tea and alloy) (Table 3).The method was also successfully applied to the simultaneous determination of the SeIV and SeVI content in a number of synthetic mixtures of SeIV and SeVI (Table 4). The method was also extended to the simultaneous determination of SeIV and SeVI in a number of environmental waters. The samples were spiked with one of the two concentrations of SeIV and SeVI and recoveries determined (Table 5). The results of the analyses of real samples (soil, human hair and food) by our procedure were in excellent agreement with those obtained by HG-AAS (Table 6).Determination of Total Selenium in Certified Reference Materials Sediment (1–2 g) or tea (2–5 g) or alloy (0.05–0.1 g) was placed in a 50 ml beaker and digested using a procedure described by Cutter.22 The beaker with the testing material and concentrated nitric acid was covered with a watch glass and heated gently for 3 h. Perchloric acid was added and the mixture was heated until Table 2 Analytical features of the proposed method Parameter SeIV SeVI Acidity/m 1.50–3.0 1.50–3.0 Fluorescence stability/h 24 24 Temperature/°C 25 25 Reagent (fold molar excess) 1 : 10–1 : 30 1 : 10–1 : 30 Linear range/mg ml21 0.01–2.2 0.1–2.4 Detection limit/ng ml21 1 10 Dispersion coefficient 1.65 1.70 Reproducibility (% RSD) 0.1–2 0.1–2 Sample throughput/samples h21 28 22 Table 3 Recoveries of total Se for Certified Reference Materials Se/mg g21 Certified Type value Found ± s* Marine Sediment (NRC-PACS 1) 1.09 ± 0.11 1.06 ± 0.08 Estuarine Sediment (CEC-CRM 277) 2.04 ± 0.18 1.95 ± 0.10 Tea (NRC-CRM C85-05) 0.041 ± 0.004 0.043 ± 0.004 Selenium eutectic alloy (%) 2.60 ± 0.10 2.58 ± 0.15 * n = 5.Table 4 Simultaneous determination of SeIV and SeVI in synthetic mixtures of standard SeIV and SeVI Added/mg ml21 Found*/mg ml21 Relative error (%) SeIV SeVI Total† SeIV Total† SeVI‡ SeIV Total† SeVI 0.20 0.20 0.40 0.20 0.40 0.20 +1.0 +1.3 +1.5 0.20 0.50 0.70 0.20 0.70 0.50 20.50 20.20 20.20 0.50 0.20 0.70 0.50 0.70 0.20 20.20 +0.30 +1.5 0.40 0.60 1.0 0.40 1.0 0.60 0.0 0.0 0.0 0.50 1.0 1.5 0.50 1.5 1.0 0.0 +1.0 +1.0 1.0 0.50 1.5 1.0 1.5 0.50 0.0 20.20 20.40 0.0 1.0 1.0 0.01 1.0 1.0 +1.0 0.0 21.0 1.0 0.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 * n = 5.† SeIV + SeVI. ‡ Calculated by subtraction of Seiv from total Se. Table 5 Analysis of spiked environmental water and flour samples SeIV added/ SeVI added/ Reduced through Total Se added*/ Total Se found*/ SeVI found†/ Recovery ± s Sample ng ml21 ng ml21 photo-reactor ng ml21 ng ml21 ng ml21 (%) Tap water 100 100 No 200 100 — 100 ± 0.3 100 100 Yes 200 199 99 99.5 ± 0.5 Lake water— Sample 1 200 80 No 280 201 — 100.5 ± 0.4 200 80 Yes 280 278 79 99.5 ± 0.6 Sample 2 500 50 No 550 499 — 99.8 ± 0.2 500 50 Yes 550 545 45 99.0 ± 0.8 Flour (rice) 300 60 No 360 301.5 — 100.5 ± 0.5 300 60 Yes 360 358.0 58 99.5 ± 0.6 * Determined as SeIV.† Calculated by subtracting the SeVI from the total Se.Table 6 Determination of total Se in real samples Total Se/mg g21 Sample AAS Proposed method* Relative error (%) Soil (surface) 0.375 0.380 +1.3 Hair (human) nd† nd — Egg (yolk) 0.170 0.167 21.7 Flour (rice) nd nd — * n = 5. † nd: not detected. 224 Analyst, March 1997, Vol. 122only a slight amount of moisture remained. A third 3 h nitric acid heating was performed and the sample was again carefully evaporated. Concentrated HCl was added under heating to ensure that all the Se was present as SeIV.The contents of the beaker were filtered through a Whatman No. 40 filter paper into a 25 ml calibrated flask. The solution was then diluted up to the mark with 2.0 m H2SO4. The total Se content was determined as described under Procedure using tartrate as masking agent. In the case of the tea sample, interference from permanganate was removed by adding sodium azide and boiling the solution before measurement. The results for total Se were in good agreement with certified values.The results of total Se by on-line photoreduction were also in excellent agreement with those obtained by HCl reduction. The results are shown in Table 3. Simultaneous Determination of SeIV and SeVI in Synthetic Mixtures Synthetic mixtures of standard SeIV and SeVI of different concentrations were prepared with 2.0 m H2SO4. The SeIV and SeVI contents were determined spectrofluorimetrically as described under Procedure. The precision for the determination of SeIV and SeVI was measured by analysing (n = 5) the samples listed in Table 4.The relative errors for all samples were < 2%. Analysis of Spiked Environmental Water and Flour Samples The proposed method was applied to the determination of SeIV and SeVI added to some environmental water and flour samples. A preliminary study showed SeIV and SeVI to be below the limits of detection in the samples. The samples were spiked with one of two concentrations of SeIV and SeVI with 2.0 m H2SO4 and the recoveries determined (the standard additions technique was used and in the case of the flour sample, the standard was added before digestion). The recoveries in all cases were high (between 99.0 and 100.5%) and are shown in Table 5.Determination of Total Selenium in Real Samples An air-dried homogenised soil sample (5–10 g) was weighed accurately and placed in a 100 ml beaker. The sample was digested and reduced following the method recommended by Cutter.22 The content of the beaker was filtered through a Whatman No. 40 filter paper into a 25 ml calibrated flask. It was then diluted up to the mark with 2.0 m H2SO4. Human hair (2–5 g) or egg yolk (5–10 g) or rice flour (5–10g) was placed in a 100 ml beaker. Following the procedure recommended by Bratakos et al.,23 the sample was digested with a mixture of nitric and perchloric acids and the SeVI species were reduced with HCl. The contents of the flask were filtered through a Whatman No. 40 filter paper into a 25 ml calibrated flask. The solution was then diluted up to the mark with 2.0 m H2SO4. Suitable aliquots of the above samples were transferred into a 10 ml calibrated flask and the total Se contents were determined as described under Procedure using tartrate as masking agent. The results of soil, hair, egg and flour analyses by the FIA method were found to be in good agreement with those obtained by HG-AAS. The results are shown in Table 6.The very low value of total selenium for the hair sample is probably due to the type, quality and quantity of the foods consumed by Greeks.24 The low value of total selenium for the rice flour sample is probably due to low selenium in soils. Bratakos et al.23 also reported such low selenium contents in foods produced in Greece. Occurrence of such low selenium contents have been reported in the soils of some countries.5,25 Conclusions The use of a photo-reduction-reactor with an FIA system has been shown to be effective in the speciation of selenium.Automation of the system has resulted in much shorter analysis times, with greater reduction efficiency than using conventional heated digestion methods. The proposed FIA method using PTQA is not only one of the most sensitive methods for the simultaneous determination of SeIV and SeVI but is also excellent in terms of selectivity and simplicity. It offers also a very efficient procedure for speciation analysis.Therefore, this method will be successfully applied to the monitoring trace amounts of selenium species in environmental, biological and soil samples. M.J.A. thanks the State Scholarship Foundation of Republic of Greece (I.K.Y.) for the award of a postdoctoral fellowship. References 1 Shamberger, R. J., Biochemistry of Selenium, Plenum, New York, 1983. 2 Levander, O. A., Trace Elements in Human and Animal Nutrition, Academic Press, New York, 1986, pp. 209–279. 3 Masscheleyen, P. H., and Patrick, W. H., Jr., Environ. Toxicol. Chem., 1993, 12, 2235. 4 Fishbein, L., Int. J. Environ. Anal. Chem., 1984, 17, 113. Table 7 Effect of interfering ions on the determination of 0.5 mg ml21 SeIV and 1.0 mg ml21 SeVI, respectively Maximum permissible concentration*/ mg ml21 Interfering ion SeIV SeVI Ammonium 100 100 Bromide 100 100 EDTA 100 100 Fluoride 100 100 Persulfate 100 100 Silicate 100 100 Tartrate 5000 5000 Aluminium 100 100 Antimony(v) 50 50 Arsenic(iii) 50 50 Beryllium 50 50 Calcium 50 50 Cerium(iii, iv) 50 50 Chromium(iii) 100 100 Cobalt(ii) 100 100 Copper(ii) 100 100 Iron(ii, iii) 50 50 Lanthanum 100 100 Magnesium 50 50 Manganese(ii) 100 100 Mercury(ii) 50 50 Molybdenium(vi) 100 100 Nickel(ii) 100 100 Potassium 2000 2000 Silver(i) 50 50 Tin(ii, iv) 50 50 Tungsten(vi) 50 50 Uranium(vi) 50 50 Vanadium(v) 50 50 Zinc+ 100 100 * For acetate, alkali metals, sodium azide, carbonate, citrate, chloride, dichromate, fluoride, iodide, nitrate, oxalate, perchlorate, phosphate and sulfate the maximum permissible concentration is 1000 mg ml21.A 5% error criterion is adopted for all the interferents. Analyst, March 1997, Vol. 122 2255 Iyengar, G. V., Kollmer, W. E., and Bowen, H. J. M., in The Elemental Composition of Human Tissues and Body Fluids, Verlag Chemie, Weinheim, 1978. 6 Masscheleyen, P. H., DeLaune, R. D., and Patrick, W. H., Jr., J. Environ. Health, 1991, 26, 555. 7 Robberecht, H., and Van Grieken, R., Talanta, 1982, 29, 823. 8 Olivas, R.M., Donard, O. F. X., Camara, C., and Quevauviller, Ph., Anal. Chim. Acta, 1994, 286, 357. 9 Sanz Alaejos, M., and Diaz Romero, C., Chem. Rev., 1995, 95, 227. 10 Welz, B., and Schubert-Jacobs, M., At. Spectrosc., 1991, 12, 91. 11 Linares, P., Luque de Castro, M. D., and Valcarcel, M., Analyst, 1986, 111, 1405. 12 Cai, Y., Cabanas, M., Fernandez-Turiel, J. L., Abalos, M., and Bayona, J. M., Anal. Chim. Acta, 1995, 314, 183. 13 Pitts, L., Worsfold, P. J., and Hill, S. J., Analyst, 1994, 119, 2785. 14 Liu, Ren-Min, Liu, Dao-Jie, and Sun, Ai-Ling, Analyst, 1992, 117, 1767. 15 Motomizu, S., and Sanada, M., Anal. Chim. Acta, 1995, 308, 406. 16 Measures, C. I., and Burton, J. D., Anal. Chim. Acta, 1980, 120, 177. 17 Atallah, R. H., and Kalman, D. A., Talanta, 1991, 38, 167. 18 Pal, B. K., Chakrabarti, A. K., and Ahmed, M. J., Anal. Chim. Acta, 1988, 206, 345. 19 Porter, H. D., J. Am. Chem. Soc., 1954, 76, 127. 20 Pal, B. K., and Chowdhury, B., Mikrochim. Acta, 1984, II, 121. 21 Ruzicka, J., and Hansen, E. H., Anal. Chim. Acta, 1978, 99, 37. 22 Cutter, G. A., Anal. Chem., 1985, 57, 2951. 23 Bratakos, M. S., Zafiropoulos, T. F., Siskos, P. A., and Ioannou, P. V., J. Food Sci., 1987, 52, 817. 24 Bratakos, M. S., Zafiropoulos, T. F., Siskos, P. A., and Ioannou, P. V., Int. J. Food Sci. Technol., 1988, 23, 585. 25 Forbes, S., Bound, G. P., and West, T. S., Talanta, 1979, 26, 473. Paper 6/06357H Received September 16, 1996 Accepted November 11, 1996 226 Analyst, March 1997, Vol. 122 Simultaneous Spectrofluorimetric Determination of Selenium(IV) and (VI) by Flow Injection Analysis M. J. Ahmed†, C. D. Stalikas, P. G. Veltsistas, S. M. Tzouwara-Karayanni and M. I. Karayannis* Laboratory of Analytical Chemistry, Department of Chemistry, University of Ioannina, Ioannina 451 10, Greece A simple, sensitive, highly selective, automatic spectrofluorimetric method for the simultaneous determination of selenium(IV) and (VI) as selenite–selenate by flow injection analysis (FIA) has been developed. The method is based on the selective oxidation of the non-fluorescent reagent 2-(a-pyridyl)thioquinaldinamide (PTQA) in acidic solution (1.5–3.0 m H2SO4) by SeIV to give an intensely fluorescent oxidation product (lex = 350 nm; lem = 500 nm).Selenium(VI) is reduced on-line to SeIV, in a reduction coil installed in a photo-reactor, which is then treated with PTQA and the fluorescence due to the sum of SeIV and SeVI is measured; SeVI is determined from the difference in fluorescence values.Various analytical parameters, such as effect of acidity, flow rate, sample size, dispersion coefficient, temperature, reagent concentration and interfering species were studied. The photo-reduction conditions were optimized, with an FIA procedure, for SeVI on the basis of its reduction efficiency. The calibration graphs were rectilinear for 0.1–2.4 mg ml21 of SeVI and 10 ng ml21–2.2 mg ml21 of SeIV, respectively.The method was applied to the determination of Se in several Standard Reference Materials (alloy, sediments and tea), as well as in some environmental waters (tap and surface water), food samples (flour and egg), a biological sample (human hair), soil sample and in synthetic mixtures. Up to 25 samples per hour can be analysed with an RSD Å 0.1–2%. Keywords: Flow injection; spectrofluorimetry; selenium speciation; 2-( a-pyridyl)thioquinaldinamide; on-line photo-reduction; environmental; biological; soil samples Recently, there has been increasing interest in trace determination of selenium because of its dual role, as an essential nutrient at low concentrations (10–40 mg ml21 in serum), or a highly toxic compound (selenosis) at an intake of 5 mg kg21 of Se (in the mammals of a seleniferous region).1 It is contained in the enzyme glutathione peroxidase (GSHPx), which affords cells protection against oxidative damage.2 A selenium deficiency in man may also result in cardiomyopathy.2 The narrow concentration range between the two opposite effects (0.1–4.0 mg kg21 in plants), requires accurate and precise knowledge of the selenium species present in the environment.3 In the environment, Se levels generally fall in the ranges 0.1–400 ng ml21 in natural waters,4 1 ng ml21 in the atmosphere3 and 0–80 mg g21 in soils.5 Selenium finds its way into the environment through its widespread use in the glass and electronics industries, as well as from the combustion of fossil fuels and uses in agriculture.Detailed information about the availability and mobility of Se in the environment and its biogeochemical cycle, however, requires the additional knowledge of the different chemical forms and oxidation states in which this element can exist. The inorganic Se species most frequently found in water and soils are selenite (SeO3 22) and selenate (SeO4 22).3 In this regard, a method for speciation is needed because its availability for plant uptake, mobility in soil, and toxicity in biota depends on the oxidation state of the Se.6 Therefore, its accurate determination at trace levels using simple and rapid methods is of paramount importance.Analytical techniques for Se speciation have recently been reviewed.7–9 Automatic flow techniques have hardly been applied to the determination of Se.Until now, only few methods have been described that use flow injection analysis (FIA) and these methods either use FIA combined with HG–AAS8,10 or FIA with spectrophotometric detection.11 But these methods suffer from several limitations: (i) indirect determination of SeVI because it does not give hydrides, and (ii) the matrix affects both hydride generation and reduction yields of SeVI to SeIV. The spectrophotometric methods suffer from sensitivity and selectivity due to various or many interferences.In this respect, interlaboratory comparisons of inorganic Se in biological and environmental samples showed unacceptable differences using these techniques.12 Recent developments in the automation of instrumentation leads to an improved precision.13 Recently, photochemical reactions have been applied to the on-line reduction and oxidation of inorganic and organic substances in flow injection methods.14,15 Measures and Burton16 studied the photochemical oxidation of an organometallic form of selenium( Se22) into inorganic SeIV with a high pressure mercury lamp.All of these systems have been used for on-line photochemical oxidation or reduction to enhance the detection of a variety of inorganic and organic compounds with AAS17 and spectrophotometric techniques.15 The aim of the present study was to develop a more simple FIA system for the simultaneous determination of SeIV and SeVI with 2-(a-pyridyl)thioquinaldinamide (PTQA) using a reduction coil installed in a photo-reactor in the reaction manifold.PTQA has been reported as a spectrofluorimetric reagent,18 but has not previously been used for the simultaneous determination of SeIV and SeVI in a flow injection system. This paper reports its use in a very sensitive, highly specific automatic spectrofluorimetric method for the simultaneous determination of SeIV and SeVI. The method is based on the selective oxidation of the nonfluorescent reagent, PTQA, in an acidic medium (1.5–3.0 m H2SO4) by SeIV to produce an intensely fluorescent product followed by the direct measurement of the fluorescence intensity in aqueous solution at room temperature.Oxidation is very rapid and no extraction is required. With suitable masking the reaction can be made highly selective. The reaction mechanism of the present method is as reported earlier.18 Experimental Apparatus The manifold for simultaneous determination of SeIV and SeVI was of Teflon tubing (0.8 mm id) and linear dual connectors were used (Fig. 1). It consisted of a four-way pneumatically † Present address: Department of Chemistry, University of Chittagong, Chittagong- 4331, Bangladesh. Analyst, March 1997, Vol. 122 (221–226) 221actuated injection valve (Rheodyne, Type 50 Teflon, Cotati, CA, USA), an eight-channel peristaltic pump (Ismatec, Glattburg- Zurich, Switzerland) and a spectrofluorimeter (RF-551, Shimadzu, Japan), equipped with a 12 ml flow-through cell for measurement.Data processing and collection was performed with an IBMcompatible Personal Computer (PC) by means of software written in Microsoft Q-Basic. The interface unit was an RTL 800/815 multifunction Input/Output board. A Varian AA-300 Atomic Absorption Spectrophotometer equipped with a hydride system at 196.1 nm using an air–acetylene flame was used for comparison of the results. A digital pH-meter (Model-PHM83 AUTOCAL, Radiometer, Copenhagen, Denmark) was used to measure the pH of the solutions.Photoreduction-reactor The photoreduction-reactor comprised a high pressure mercury light source (2 cm od, 25 cm long, 125 W, DESAGA) and a quartz coil (40 cm long 3 0.8 mm id). The source emits short wavelength light at lmax = 254 nm. The effective irradiation length was 6 cm. The unit is covered with aluminium foil or thick paper in order to increase the light intensity reaching the coil by reflectance and to prevent eye exposure to ultraviolet radiation.Reagents All chemicals used were of analytical-reagent grade or the highest purity available. Doubly distilled water and HPLCgrade propan-2-ol, which is non-fluorescent under ultraviolet radiation, were used throughout. SeIV standard solutions. A 100 ml amount of stock SeIV solution (1 mg ml21) was prepared by dissolving 333.1 mg of general-reagent grade sodium selenite (Merck, Darmastadt, Germany) in doubly distilled water. The solution was kept in a refrigerator in a polyethylene container for preservation.Working standard solutions were prepared daily by appropriate dilution in 2 m H2SO4. SeVI standard solutions. A 100 ml amount of stock SeVI solution (1 mg ml21) was prepared by dissolving 467.4 mg of ACS-grade sodium selenate (99%, Aldrich, Steinheim, Germany) in doubly distilled water. The solution was kept in a refrigerator in a polyethylene container. Working standard solutions were prepared daily by appropriate dilution in 2 m H2SO4.Carrier solution. 2.0 m H2SO4 (Merck) was used as the carrier solution. 2-( a-Pyridyl)thioquinaldinamide (PTQA), solution (1023 M). The reagent was synthesized according to the method of Porter.19 The contents, containing 2-aminopyridine (2 mol), quinaldine (1 ml) and sulfur powder (1.5 mol), were mixed and refluxed for 6 h in a 250 ml round bottomed flask fitted with bulb condenser under controlled temperature (140–150 °C) at 1 atm pressure over a sand bath.The reaction mixture was kept overnight. The thio-compound was filtered and crystallized using petroleum ether to give a bright yellow crystalline (needle) solid. The compound recrystallized from ethanol was dried under vacuum (0.1 mg of Hg) for 24 h. The melting point of this synthesized compound (PTQA) was 155 ± 2 °C and the elemental analysis data (C = 72.25, N = 13.35 and H = 4.25%) were very close to the literature values,19 e.g., melting point (155 ± 1 °C); C (N72.43); N (N13.55) and H (N4.55%).The reagent solution (1023 m) was prepared by dissolving the requisite amount (0.0266 g in 100 ml) of PTQA in HPLC-grade propan-2-ol. A freshly prepared reagent solution (1024 m) was used whenever required. Other solutions. Solutions of a large number of inorganic ions and complexing agents were prepared from their AnalaR grade or equivalent grade water soluble salts. In the case of insoluble substances, special dissolution methods were adopted.20 Stock solutions and environmental samples were kept in a refrigerator in poly(propylene) bottles.Preparation of the Samples Food samples (rice flour and egg) were purchased from a local supermarket. These samples were homogenized thoroughly. Soil samples was collected from local agricultural field sites and homogenized in a mortar. Human hair (3–5 cm long from male) was cut from the occipitonuchal region of the head. They were cleaned by stirring with acetone, rinsing with tap water, stirring in a detergent solution (which had no detectable Se), rinsing with tap water, doubly distilled water and finally with acetone.They were then dried at 45 °C and cut into small pieces for analyses (sampled person did not use Se-containing shampoo). Procedure The standards (0.01–2.2 mg ml21 SeIV or 0.1–2.4 mg ml21 SeVI) and samples were injected into a carrier stream by means of the peristaltic pump, P (Fig. 1). Then the sample was measured by different ways using a selector valve.The sample stream was firstly directed through path 1, treated with a 10–30-fold molar excess of the PTQA reagent solution and passed directly into the measuring cell of a spectrofluorimeter where the fluorescence intensity due to SeIV was measured at 500 nm with excitation at 350 nm. Then the sample stream was passed through path 2 to photo-reduction coil (R1) by using a second selector valve where SeVI was reduced to SeIV. The sample stream was then treated with the PTQA reagent at the end of the coil and the overall mixture was passed to the same cell of the spectrofluorimeter where the fluorescence intensity due to total Se was measured; SeVI was determined from the difference in fluorescence intensity values.The reaction is very rapid and the fluorescence intensity remains stable for 24 h. The PTQA reagent does not show any fluorescence in the absence of SeIV. The concentrations of SeIV and SeVI were evaluated from the peak heights of the signal by using the calibration curves prepared with standard solutions.Results and Discussion Optimization of the Flow Injection System Preliminary tests were carried out with the aid of different flow assemblies to select the optimal manifold configuration. The assembly in Fig. 1 was selected as the one producing the best compromise between peak height and the shape of the peak. Fig. 1 Schematic representation of FI manifold employed for the simultaneous determination of SeIV and SeVI.P, Pump; V, valve; S, selector valve; R1, photoreduction coil; R2, single bead string reactor (SBSR); D, detector; W, waste; and PC, personal computer. 222 Analyst, March 1997, Vol. 122In order to optimize the proposed flow injection manifold, the influence of the hydrodynamic and chemical parameters on the magnitude of the peak height, the shape of the peak and reproducibility of the results were studied. The univariate method was adopted for the optimization of the system.Table 1 shows results of optimization of working conditions for 0.5 mg ml21 of SeIV and 1.0 mg ml21 of SeVI. The optimum length of the photo-reduction coil (R1) was established by using a 1.0 mg ml21 SeVI solution, the single bead string reactor (SBSR) (R2), for better mixing and lower dispersion, being of length 100 cm and having an acidity of 1.5–3.0 m H2SO4. Photo-reduction coil (R1) lengths of 15, 30, 40, 60 and 80 cm were tested, keeping the power of the lamp and its distance from reaction coil constant.For any combination of the above parameters the efficiency of the reduction was determined by comparison of the plateau achieved with that corresponding to a 1.0 mg ml21 SeIV solution processed in the same way. A coil length of 40 cm was chosen, because the reduction was almost complete, reproducibility was good and back-pressure relatively low. Different wavelengths of the UV radiation were also tested. For each wavelength, the efficiency of the reduction was determined.The effective wavelength of photo-reduction of the SeVI was lmax = 254 nm. Different lamp powers were also tested but no significant effect on reduction was observed. Different distances of the reaction coil from the lamp were tested keeping the length of the coil and wavelength constant. A length of 6 cm was selected because maximum conversion efficiency was achieved. A length 100 cm for the SBSR reactor (R2), a sample size of 100 ml, an overall flow rate of 0.4 ml min21 and a reagent flow rate of 0.3 ml min21 were selected, these being a compromise between the sampling rate and the height of the peak.Of the various acids (sulfuric, hydrochloric, nitric and phosphoric) studied, sulfuric acid was found to be best acid for the system. Different concentrations of sulfuric acid were tested in the range shown in Table 1. The fluorescence intensity was at maximum and constant when the solution (1.0 mg ml21) contained 1.5–3.0 m H2SO4 (Fig. 2) at room temperature, which was the optimum acidity range. The photoreduction efficiency for 1.0 mg ml21 of SeVI in this acidity range was also tested. More than 97% of the SeVI can be reduced to SeIV in this acidity range. For all subsequent measurements 2.0 m H2SO4 was used as carrier for this manifold. The effect of propan-2-ol on the fluorescence was studied and no adverse effect was observed over a wide range of propan- 2-ol concentrations.A 1024 m solution of PTQA in propan-2-ol was sufficient to prevent any precipitation or turbidity or bubbling and to allow accurate measurements for this manifold. Other common organic solvents, e.g., chloroform, benzene, tetrachloromethane and ethanol, were also tried but no fluorescence was observed in the organic phase, with the exception of ethanol. The reaction is rapid. A constant maximum fluorescence intensity was obtained just after the dilution to volume and remained strictly unaltered for 24 h.Different concentrations of PTQA solution were tested in the ranges shown in Table 1. A reagent concentration of 2 3 1024 m was selected as optimum for this manifold. The fluorescence intensity of the SeIV–PTQA system within the prescribed acidity range was maximum and constant for SeIV to reagent molar ratios in the range 1 : 10–1 : 30 when the Se concentration was 1.0 mg ml21 (Fig. 3). At different SeIV concentrations (0.1 and 0.5 mg ml21), the effect of varying the reagent concentration was similar.Evaluation of the Method The reproducibility of the proposed procedure and sample throughput were determined by repeated injection of a sample containing 0.5 mg ml21 SeIV and 1.0 mg ml21 SeVI. The RSD (n = 5) was 0.1–2% for 0.01–2.2 mg ml21 SeIV and 0.1–2.4 mg ml21 SeVI indicating that this method is highly precise and reproducible. The calibration graphs obtained from the peak heights were rectilinear for 10 ng ml21 to 2.2 mg ml21 of SeIV and 0.1–2.4 mg ml21 of SeVI, respectively.The detection limits, defined as three times the baseline noise, were 1 ng ml21 for SeIV and 10 ng ml21 for SeVI. The sample throughput was 25 measurements per hour. The dispersion coefficients were estimated with a 0.5 mg ml21 SeIV and 1.0 mg ml21 SeVI standard solutions as described earlier.21 Important features of the proposed method for simultaneous determination of SeIV and SeVI are summarised in Table 2.The performance and reproducibility of the proposed method are also shown in Tables 3–6. The reliability of the proposed procedure was also assessed by analysing Certified Reference Materials. The results for total Se were in good agreement with certified values (Table 3). The method was also tested by analysing several synthetic mixtures containing standard SeIV and SeVI (Table 4). The reliability of the proposed procedure was also tested by performing recovery studies.The average Table 1 Selected chemical and FIA parameters obtained with the optimization experiments Parameter Studied range Selected value Size of sample loop/ml 30–180 100 Overall flow rate/ ml min21 0.20–1.0 0.40 Reagent flow rate/ ml min21 0.05–0.60 0.30 Length of the photoreaction coil, R1/cm 15–80 40 Length of the SBSR reactor, R2/cm 20–180 100 pH 0.1–1.3 0.40–0.75 (preferably 0.6) Concentration of reagent ( m)— H2SO4 0.50–4.5 1.50–3.0 (preferably 2.0) PTQA 331025–631024 231024 Fig. 2 Effect of acidity on the fluorescence intensity of the SeIV–PTQA system. Fig. 3 Effect of reagent on the fluorescence intensity of the Seiv–PTQA system. Analyst, March 1997, Vol. 122 223percentage recovery obtained for the addition of SeIV and SeVI spikes to some environmental water and flour samples was quantitative as shown in Table 5. The results of food, hair and soil analyses by the present method was in excellent agreement with those obtained by HG-AAS (Table 6).The precision and accuracy of the method are satisfactory. The interference of several ions which may occur in environmental samples was studied by using a solution containing a mixture of SeIV and SeVI at concentrations of 0.5 mg ml21 and 1.0 mg ml21, respectively, adding various concentrations of interfering ions up to the amounts where the relative error reached a value of about 5%. The errors were calculated by comparing the peak height to that obtained after the injection of an aqueous solution of SeIV and SeVI containing no interfering ions, as a reference.The results are summarized in Table 7. During the interference studies, if a precipitate was formed, it was removed by centrifugation. Positive interference from permanganate or hydrogen peroxide was eliminated by boiling the solution with sodium azide, a reducing agent which had no reducing effect on either SeIV or SeVI. Applications The proposed method was used to determine the total selenium content in a number of certified reference materials (sediments, tea and alloy) (Table 3).The method was also successfully applied to the simultaneous determination of the SeIV and SeVI content in a number of synthetic mixtures of SeIV and SeVI (Table 4). The method was also extended to the simultaneous determination of SeIV and SeVI in a number of environmental waters. The samples were spiked with one of the two concentrations of SeIV and SeVI and recoveries determined (Table 5).The results of the analyses of real samples (soil, human hair and food) by our procedure were in excellent agreement with those obtained by HG-AAS (Table 6). Determination of Total Selenium in Certified Reference Materials Sediment (1–2 g) or tea (2–5 g) or alloy (0.05–0.1 g) was placed in a 50 ml beaker and digested using a procedure described by Cutter.22 The beaker with the testing material and concentrated nitric acid was covered with a watch glass and heated gently for 3 h.Perchloric acid was added and the mixture was heated until Table 2 Analytical features of the proposed method Parameter SeIV SeVI Acidity/m 1.50–3.0 1.50–3.0 Fluorescence stability/h 24 24 Temperature/°C 25 25 Reagent (fold molar excess) 1 : 10–1 : 30 1 : 10–1 : 30 Linear range/mg ml21 0.01–2.2 0.1–2.4 Detection limit/ng ml21 1 10 Dispersion coefficient 1.65 1.70 Reproducibility (% RSD) 0.1–2 0.1–2 Sample throughput/samples h21 28 22 Table 3 Recoveries of total Se for Certified Reference Materials Se/mg g21 Certified Type value Found ± s* Marine Sediment (NRC-PACS 1) 1.09 ± 0.11 1.06 ± 0.08 Estuarine Sediment (CEC-CRM 277) 2.04 ± 0.18 1.95 ± 0.10 Tea (NRC-CRM C85-05) 0.041 ± 0.004 0.043 ± 0.004 Selenium eutectic alloy (%) 2.60 ± 0.10 2.58 ± 0.15 * n = 5.Table 4 Simultaneous determination of SeIV and SeVI in synthetic mixtures of standard SeIV and SeVI Added/mg ml21 Found*/mg ml21 Relative error (%) SeIV SeVI Total† SeIV Total† SeVI‡ SeIV Total† SeVI 0.20 0.20 0.40 0.20 0.40 0.20 +1.0 +1.3 +1.5 0.20 0.50 0.70 0.20 0.70 0.50 20.50 20.20 20.20 0.50 0.20 0.70 0.50 0.70 0.20 20.20 +0.30 +1.5 0.40 0.60 1.0 0.40 1.0 0.60 0.0 0.0 0.0 0.50 1.0 1.5 0.50 1.5 1.0 0.0 +1.0 +1.0 1.0 0.50 1.5 1.0 1.5 0.50 0.0 20.20 20.40 0.0 1.0 1.0 0.01 1.0 1.0 +1.0 0.0 21.0 1.0 0.0 1.0 1.0 1.0 0.0 0.0 0.0 0.0 * n = 5.† SeIV + SeVI. ‡ Calculated by subtraction of Seiv from total Se. Table 5 Analysis of spiked environmental water and flour samples SeIV added/ SeVI added/ Reduced through Total Se added*/ Total Se found*/ SeVI found†/ Recovery ± s Sample ng ml21 ng ml21 photo-reactor ng ml21 ng ml21 ng ml21 (%) Tap water 100 100 No 200 100 — 100 ± 0.3 100 100 Yes 200 199 99 99.5 ± 0.5 Lake water— Sample 1 200 80 No 280 201 — 100.5 ± 0.4 200 80 Yes 280 278 79 99.5 ± 0.6 Sample 2 500 50 No 550 499 — 99.8 ± 0.2 500 50 Yes 550 545 45 99.0 ± 0.8 Flour (rice) 300 60 No 360 301.5 — 100.5 ± 0.5 300 60 Yes 360 358.0 58 99.5 ± 0.6 * Determined as SeIV.† Calculated by subtracting the SeVI from the total Se. Table 6 Determination of total Se in real samples Total Se/mg g21 Sample AAS Proposed method* Relative error (%) Soil (surface) 0.375 0.380 +1.3 Hair (human) nd† nd — Egg (yolk) 0.170 0.167 21.7 Flour (rice) nd nd — * n = 5. † nd: not detected. 224 Analyst, March 1997, Vol. 122only a slight amount of moisture remained. A third 3 h nitric acid heating was performed and the sample was again carefully evaporated.Concentrated HCl was added under heating to ensure that all the Se was present as SeIV. The contents of the beaker were filtered through a Whatman No. 40 filter paper into a 25 ml calibrated flask. The solution was then diluted up to the mark with 2.0 m H2SO4. The total Se content was determined as described under Procedure using tartrate as masking agent. In the case of the tea sample, interference from permanganate was removed by adding sodium azide and boiling the solution before measurement.The results for total Se were in good agreement with certified values. The results of total Se by on-line photoreduction were also in excellent agreement with those obtained by HCl reduction. The results are shown in Table 3. Simultaneous Determination of SeIV and SeVI in Synthetic Mixtures Synthetic mixtures of standard SeIV and SeVI of different concentrations were prepared with 2.0 m H2SO4.The SeIV and SeVI contents were determined spectrofluorimetrically as described under Procedure. The precision for the determination of SeIV and SeVI was measured by analysing (n = 5) the samples listed in Table 4. The relative errors for all samples were < 2%. Analysis of Spiked Environmental Water and Flour Samples The proposed method was applied to the determination of SeIV and SeVI added to some environmental water and flour samples. A preliminary study showed SeIV and SeVI to be below the limits of detection in the samples.The samples were spiked with one of two concentrations of SeIV and SeVI with 2.0 m H2SO4 and the recoveries determined (the standard additions technique was used and in the case of the flour sample, the standard was added before digestion). The recoveries in all cases were high (between 99.0 and 100.5%) and are shown in Table 5. Determination of Total Selenium in Real Samples An air-dried homogenised soil sample (5–10 g) was weighed accurately and placed in a 100 ml beaker.The sample was digested and reduced following the method recommended by Cutter.22 The content of the beaker was filtered through a Whatman No. 40 filter paper into a 25 ml calibrated flask. It was then diluted up to the mark with 2.0 m H2SO4. Human hair (2–5 g) or egg yolk (5–10 g) or rice flour (5–10g) was placed in a 100 ml beaker. Following the procedure recommended by Bratakos et al.,23 the sample was digested with a mixture of nitric and perchloric acids and the SeVI species were reduced with HCl. The contents of the flask were filtered through a Whatman No. 40 filter paper into a 25 ml calibrated flask. The solution was then diluted up to the mark with 2.0 m H2SO4. Suitable aliquots of the above samples were transferred into a 10 ml calibrated flask and the total Se contents were determined as described under Procedure using tartrate as masking agent. The results of soil, hair, egg and flour analyses by the FIA method were found to be in good agreement with those obtained by HG-AAS.The results are shown in Table 6. The very low value of total selenium for the hair sample is probably due to the type, quality and quantity of the foods consumed by Greeks.24 The low value of total selenium for the rice flour sample is probably due to low selenium in soils. Bratakos et al.23 also reported such low selenium contents in foods produced in Greece. Occurrence of such low selenium contents have been reported in the soils of some countries.5,25 Conclusions The use of a photo-reduction-reactor with an FIA system has been shown to be effective in the speciation of selenium.Automation of the system has resulted in much shorter analysis times, with greater reduction efficiency than using conventional heated digestion methods. The proposed FIA method using PTQA is not only one of the most sensitive methods for the simultaneous determination of SeIV and SeVI but is also excellent in terms of selectivity and simplicity.It offers also a very efficient procedure for speciation analysis. Therefore, this method will be successfully applied to the monitoring trace amounts of selenium species in environmental, biological and soil samples. M.J.A. thanks the State Scholarship Foundation of Republic of Greece (I.K.Y.) for the award of a postdoctoral fellowship. References 1 Shamberger, R. J., Biochemistry of Selenium, Plenum, New York, 1983. 2 Levander, O. A., Trace Elements in Human and Animal Nutrition, Academic Press, New York, 1986, pp. 209–279. 3 Masscheleyen, P. H., and Patrick, W. H., Jr., Environ. Toxicol. Chem., 1993, 12, 2235. 4 Fishbein, L., Int. J. Environ. Anal. Chem., 1984, 17, 113. Table 7 Effect of interfering ions on the determination of 0.5 mg ml21 SeIV and 1.0 mg ml21 SeVI, respectively Maximum permissible concentration*/ mg ml21 Interfering ion SeIV SeVI Ammonium 100 100 Bromide 100 100 EDTA 100 100 Fluoride 100 100 Persulfate 100 100 Silicate 100 100 Tartrate 5000 5000 Aluminium 100 100 Antimony(v) 50 50 Arsenic(iii) 50 50 Beryllium 50 50 Calcium 50 50 Cerium(iii, iv) 50 50 Chromium(iii) 100 100 Cobalt(ii) 100 100 Copper(ii) 100 100 Iron(ii, iii) 50 50 Lanthanum 100 100 Magnesium 50 50 Manganese(ii) 100 100 Mercury(ii) 50 50 Molybdenium(vi) 100 100 Nickel(ii) 100 100 Potassium 2000 2000 Silver(i) 50 50 Tin(ii, iv) 50 50 Tungsten(vi) 50 50 Uranium(vi) 50 50 Vanadium(v) 50 50 Zinc+ 100 100 * For acetate, alkali metals, sodium azide, carbonate, citrate, chloride, dichromate, fluoride, iodide, nitrate, oxalate, perchlorate, phosphate and sulfate the maximum permissible concentration is 1000 mg ml21. A 5% error criterion is adopted for all the interferents. Analyst, March 1997, Vol. 122 2255 Iyengar, G. V., Kollmer, W. E., and Bowen, H. J. M., in The Elemental Composition of Human Tissues and Body Fluids, Verlag Chemie, Weinheim, 1978. 6 Masscheleyen, P. H., DeLaune, R. D., and Patrick, W. H., Jr., J. Environ. Health, 1991, 26, 555. 7 Robberecht, H., and Van Grieken, R., Talanta, 1982, 29, 823. 8 Olivas, R. M., Donard, O. F. X., Camara, C., and Quevauviller, Ph., Anal. Chim. Acta, 1994, 286, 357. 9 Sanz Alaejos, M., and Diaz Romero, C., Chem. Rev., 1995, 95, 227. 10 Welz, B., and Schubert-Jacobs, M., At. Spectrosc., 1991, 12, 91. 11 Linares, P., Luque de Castro, M. D., and Valcarcel, M., Analyst, 1986, 111, 1405. 12 Cai, Y., Cabanas, M., Fernandez-Turiel, J. L., Abalos, M., and Bayona, J. M., Anal. Chim. Acta, 1995, 314, 183. 13 Pitts, L., Worsfold, P. J., and Hill, S. J., Analyst, 1994, 119, 2785. 14 Liu, Ren-Min, Liu, Dao-Jie, and Sun, Ai-Ling, Analyst, 1992, 117, 1767. 15 Motomizu, S., and Sanada, M., Anal. Chim. Acta, 1995, 308, 406. 16 Measures, C. I., and Burton, J. D., Anal. Chim. Acta, 1980, 120, 177. 17 Atallah, R. H., and Kalman, D. A., Talanta, 1991, 38, 167. 18 Pal, B. K., Chakrabarti, A. K., and Ahmed, M. J., Anal. Chim. Acta, 1988, 206, 345. 19 Porter, H. D., J. Am. Chem. Soc., 1954, 76, 127. 20 Pal, B. K., and Chowdhury, B., Mikrochim. Acta, 1984, II, 121. 21 Ruzicka, J., and Hansen, E. H., Anal. Chim. Acta, 1978, 99, 37. 22 Cutter, G. A., Anal. Chem., 1985, 57, 2951. 23 Bratakos, M. S., Zafiropoulos, T. F., Siskos, P. A., and Ioannou, P. V., J. Food Sci., 1987, 52, 817. 24 Bratakos, M. S., Zafiropoulos, T. F., Siskos, P. A., and Ioannou, P. V., Int. J. Food Sci. Technol., 1988, 23, 585. 25 Forbes, S., Bound, G. P., and West, T. S., Talanta, 1979, 26, 473. Paper 6/06357H Received September 16, 1996 Accepted November 11, 1996 226 Analyst, March 1997, Vol. 122

ISSN:0003-2654    

DOI:10.1039/a606357h

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Concentration of Platinum(IV) From Acidic Chloride and Sulfate–Chloride Aqueous Media With Reversed Micellar Solutions of Oxyethylated Surfactant Alexander I. Bulavchenko*, Tat’yana Yu. Podlipskaya, Elena K. Batishcheva and Vladislav G. Torgov Institute of Inorganic Chemistry, Russian Academy of Sciences, Siberian Branch, 630090 Novosibirsk, Russia The possibility of employing reversed micelles of oxyethylated surfactants to concentrate platinum from acidic aqueous media was investigated.A procedure for reversed micellar concentration is described. Platinum concentration was effected by means of desolubilization in a non-traditional back-extraction stage by dilution of the reversed micellar solution with chloroform or a mixture of chloroform with hexane. In acidic sulfate–chloride media the distribution ratio of PtIV increased to about 102–5 3 103 in the presence of Br2 and I2 as complexing agents. In chloride media the distribution ratio could be increased to about 2 3 103 by using SnCl2.The maximum recovery of platinum from the extract did not exceed 85%. With I2 as a complexing agent, no back-extraction could be performed. With this back-extraction procedure, the PtIV concentration factor varied from about 102 to 103 depending on the aqueous feed composition. It was shown spectrophotometrically that PtIV complex species were the same in the feed, extract and desolubilized aqueous solution. The possibility of the spectrophotometric determination of platinum with SnCl2 directly in the reversed micellar solution is demonstrated.Keywords: Reversed micellar extraction; oxyethylated oil-soluble surfactant; platinum(IV); chloride and sulfate–chloride aqueous media Surfactants and reversed micellar solutions are widely used in many analytical procedures as auxiliary agents to increase analytical sensitivity and selectivity.1–3 In recent years there has been a trend to develop procedures with surfactants playing the main role.The use of reversed micellar surfactant solutions for the separation and concentration of metals has attracted considerable attention. Vijayalakshmi and co-workers4,5 studied the equilibrium extraction and concentration of polyvalent metal ions with water-in-oil (w/o) microemulsions and showed the possibility of separating trivalent from divalent metal ions. Apanasenko et al.6 performed the reversed micellar concentration of Al and Ga with a fivefold concentration factor.Earlier, a threefold concentration of NiII was achieved with the help of w/o microemulsions of ionic surfactants.7 The solubilization and complexation of metals in reversed micelles has been studied.8,9 Water-in-oil microemulsion formation and the extraction of metals with mixed micelles of surfactants and various types of extractants has been investigated.10 Neuman and co-workers11,12 considered the role of reversed micelles in extraction with the well known extractant di(2-ethylhexyl)phosphoric acid.Leodidis and Hatton13 studied the selective solubilization of alkali and alkaline earth metal ions with reversed micelles. The possibility of separating metals having similar properties with the help of reversed micelles in liquid membrane processes has been demonstrated for the pair NiII–CoII.14 This brief survey indicates that this area of research is still in its infancy. The unique properties of reversed micellar solutions are very under-utilized and at present there is no serious competition with traditional extraction techniques.We have developed a procedure for the reversed micellar concentration of metals15 differing from solvent extraction concentration16 in having a non-traditional back-extraction stage. In this stage the reversed micelles are broken by heating or by the the addition of chloroform17 and the solubilized metalenriched aqueous solution separates as an aqueous phase (Fig. 1). The non-use of an aqueous stripping agent solution and the low solubilizing capacity of reversed micelles have resulted in a considerably higher concentration of chloride metal complexes (FeIII, PtIV and PdII)18 compared with solvent extraction concentration. The aim of this work was to develop further the reversed micellar concentration procedure to determine low PtIV concentrations in aqueous acidic media. The effects of various complexing agents (Br2 and I2) and SnCl2 on the distribution ratio, recovery and concentration factor in the extraction and back-extraction stages were studied. The method was applied to the determination of low concentrations of PtIV in the presence of FeIII.Experimental Reagents Oxyethylated isononylphenol with an average degree of oxyethylation of 4, manufactured by Diagnosticum (Belgorod, Russia) as Neonol APh9-4 (an analogue of Triton N-42), was used as the oil-soluble micellation surfactant. The hydrophilic– lipophilic balance (HLB) was 9.1 ± 0.3 and the mass fraction of the main substance (the sum of the oxyethylated homologues) was 99.8%.The distribution functions of the homologues had a Poisson shape. The contribution of the homologue with a degree of oxyethylation of 4 was 30%. The surfactant was used as Fig. 1 Scheme for reversed micellar concentration of metals. Analyst, March 1997, Vol. 122 (227–232) 227received. The solvent (reversed micellar solution) contained 10% v/v Neonol APh9-4 in decane.Analytical-reagent grade organic diluents (decane, hexane and chloroform) were used without further purification. All other reagents were chemically pure. Distilled water was used to prepare aqueous solutions. Solutions of PtIV were prepared by diluting a solution of chloroplatinic acid. The acid was obtained by dissolving 99.9% pure metallic platinum in HNO3–HCl (1+3). The stock solution contained 0.2 mol l21 PtIV in 2.4 mol l21 HCl. A solution of FeIII (2.5 mol l21 + 1.5 mol l21 HCl) was obtained by dissolving FeCl3·6H2O in hydrochloric acid.Sulfate–chloride and chloride solutions had the compositions saturated Na2SO4 + 3 mol l21 H2SO4 + 0.05 mol l21 HCl, 4 mol l21 NaCl + 1.2 mol l21 HCl and 1 mol l21 HCl. The complexing agents (HBr and NaI) and SnCl2 were added to the feed as aliquot portions of the solutions: 7.1 mol l21 HBr, 0.5 mol l21 NaI and 0.5 mol l21 SnCl2 + 2 mol l21 HCl. Apparatus Spectrophotometric measurements were performed on a Specord M-40 UV/VIS spectrophotometer (Carl Zeiss, Jena, Germany).Colorimetric determinations of FeIII and PtIV were performed on a KFK-2MP photoelectric colorimeter (ZOMZ, Zagorsk, Russia). Determination of PtIV and FeIII in Aqueous Media Concentrations of PtIV and FeIII in stock and working standard solutions were determined colorimetrically using published procedures with SnCl2 19 and sulfosalicylic acid,20 respectively. Concentration With Reversed Micellar Solutions (Extraction and Back-extraction Stages) The procedure for the reversed micellar concentration of metals is shown schematically in Fig. 2. The solvent extraction stage was performed in a glass separating funnel with a volume of 100–1500 ml. A water-to-oil ratio of 5 : 1 was used in most instances. The feed served as the continuous phase. The phases were contacted for not more than 30 min under mild stirring with a propeller mixer. After extraction, the raffinate was analysed for PtIV and/or FeIII content.The extract was transferred into a conical-bottomed flask (Fig. 1) with polyethylene-coated walls and diluted with chloroform or chloroform–hexane [route (1), Fig. 2]. After the addition of chloroform, the solution became turbid and the solubilized aqueous solution of the metal coagulated into a separate transparent phase. This took 1–2 h for most of the systems. The separated aqueous solution was removed with a glass measuring pipette, measured and analysed for the metal content.Spectrophotometric Determination of PtIV Directly in Reversed Micellar Phase As an alternative procedure, the metal concentration can be determined directly in the reversed micellar phase without micelle breaking and separation of the solubilized aqueous solution [route (2), Fig. 2]. In this approach, the determination Fig. 2 Analytical procedure for reversed micellar concentration of platinum. 228 Analyst, March 1997, Vol. 122of PtIV is effected colorimetrically by a procedure developed for aqueous media. This ‘water’ procedure is applicable here since the interaction of analytical reagents with platinum takes place in the water cores of the micelles.To prepare a standard reversed micellar solution of PtIV, an aliquot of an aqueous solution of PtIV was solubilized by the injection method into the solvent. An aliquot (0.5–3.0 ml) of this standard solution of PtIV or of its extract was then transferred into a 25 ml calibrated flask and diluted with the solvent to about 20 ml [route (2) Fig. 2], then 0.21 ml of an aqueous solution of SnCl2 (2.66 mol l21 SnCl2 + 3 mol l21 HCl) was solubilized into the solution by the injection method. The resulting solution was diluted to volume with the solvent, allowed to stand for 40 min and then scanned photometrically against a reagent blank in a cell with a pathlength of 5 cm at a wavelength of 400 nm. A blank solution was prepared in the same way with the same composition except that it did not contain platinum.The calibration graph was linear for platinum concentrations in the range 1026–1025 mol l21. Determination of platinum directly in the reversed micellar solution was employed to calculate the extraction parameters in a number of systems. Results and Discussion Results for the reversed micellar concentration of PtIV in the extraction stage are given in Table 1. Extraction Stage Extraction of chloride complexes of PtIV The distribution isotherms of PtIV from acidic chloride and sulfate–chloride solutions are shown in Fig. 3. The dependences are linear for low metal concentrations and the distribution ratio of PtIV in the presence of Cl2 did not exceed 12.9 ± 0.7. To determine the PtIV complex species in the media under study, the corresponding absorption spectra were measured. The absorption maximum of PtIV in the visible region was at 453 nm, corresponding to the chloride complex (PtCl6 22).21 The electronic absorption spectra of PtIV measured in all stages of reversed micellar concentration (Fig. 4) indicated that the PtIV complex species was the same as that in the feed, extract and desolubilized aqueous solution. The bathochromic shift of the absorption maximum by 11 nm for the chloride complex of platinum in reversed micelles may be due to its interaction with protonated oxyethylate groups of the surfactant molecules. Effect of Br2 and I2 on extraction of PtIV from acidic sulfate–chloride media Spectrophotometric studies indicated that in acidic sulfate– chloride solutions of PtIV in the presence of Br2, the bromide complex of platinum (PtBr6 22) is formed.The equilibrium in the solutions at room temperature is established within 2 d. In the presence of Br2, the distribution ratio of PtIV increases by an Table 1 Results for reversed micellar concentration of PtIV in the extraction stage (95% confidence level). For abbreviations, see text [eqn.(1)] Feed Model Supporting system electrolyte and No. complexing agent Metal Cf/1024 mol l21 Vw/Vo D Re(%) Ke 1 Saturated Na2SO4 + 3 mol l21 H2SO4 + 0.05 mol l21 HCl Pt 1.72 ± 0.03 5 12.9 ± 0.7 72.1 ± 1.1 3.6 ± 0.2 2 Saturated Na2SO4 + 3 mol l21 H2SO4 + 0.001 mol l21 HCl + 0.05 mol l21 HBr Pt 1.37 ± 0.03 5 88.7 ± 3.5 94.7 ± 1.4 4.7 ± 0.2 3 Saturated Na2SO4 + 3 mol l21 H2SO4 + 0.002 mol l21 HCl + 0.002 mol l21 NaI Pt 1.82 ± 0.04 50 (4.74 ± 0.16) 3 103 99.0 ± 1.5 49.5 ± 2.1 4 1 mol l21 HCl + 0.02 mol l21 SnCl2 Pt 5.00 ± 0.09 5 (1.78 ± 0.07) 3 103 99.7 ± 1.5 5.0 ± 0.2 5 1 mol l21 HCl + 0.05 mol l21 SnCl2 Pt (5.0 ± 0.1) 3 1022 50 (2.00 ± 0.08) 3 103 97.6 ± 1.5 48.8 ± 2.0 6 Saturated Na2SO4 + 3 mol l21 H2SO4 + Pt 1.81 ± 0.03 5 12.9 ± 0.7 72.0 ± 1.1 3.6 ± 0.2 0.05 mol l21 HCl Fe (1.10 ± 0.02) 3 102 5 @0.05 1.00 ± 0.02 0.050 ± 0.002 7 4 mol l21 NaCl + Pt 3.92 ± 0.07 5 0.95 ± 0.05 16.1 ± 0.2 0.80 ± 0.03 1.2 mol l21 HCl Fe 4.34 ± 0.09 5 6.4 ± 0.4 56.2 ± 0.8 2.8 ± 0.1 Fig. 3 Distribution isotherms of PtIV and FeIII from acidic aqueous media: 1, saturated Na2SO4 + 3 mol l21 H2SO4 + 0.05 mol l21 HCl; 2–5, 4 mol l21 NaCl + 1.2 mol l21 HCl, with 1 and 4, PtIV, 2, FeIII 3, FeIII in the presence of PtIV [Cf(Fe) = Cf(Pt)] and 5, PtIV in the presence of FeIII [Cf(Pt) = Cf(Fe)]. Analyst, March 1997, Vol. 122 229order of magnitude and shows a maximum (Fig. 5, curve 1) at a 500-fold excess of HBr relative to PtIV. A greater increase in the distribution ratio (to about 5000) was obtained with I2 as the complexing agent.The dependence shows a broad maximum at a tenfold excess of I2 (Fig. 5, curve 2). Effect of SnCl2 on extraction of PtIV from dilute hydrochloric acid media The logarithmic dependence of the distribution ratio of PtIV when concentrated from 1 mol l21 HCl is linear over a wide range of SnCl2 concentrations (Fig. 5, curve 3). The addition of a 40-fold molar excess of SnCl2 relative to PtIV increases the distribution ratio by four orders of magnitude.Extraction of chloride complex of PtIV in the presence of FeIII The extraction of these two metals, which frequently accompany each other, is considered for the case of the collective extraction of PtIV and FeIII from acidic chloride and sulfate– chloride media in the absence of complexing agents. The distribution ratio of PtIV from chloride media at low concentrations does not change even at a 50-fold molar excess of FeIII.It was also found that in the extraction of metals with equal concentrations in the systems being studied, the distribution isotherms of FeIII and PtIV (Fig. 3) do not change in the region of constancy of the distribution ratios. Hence in the systems studied the metals are extracted into reversed micelles independently of each other. The results for the extraction of PtIV with oxyethylated surfactants in decane from acidic chloride and sulfate–chloride media follow the general regularities found for the extraction of platinum with oxygen-containing extractants (alcohols, ethers, etc.).22 The increase in recovery when using complexing agents results from the increased stability of the complexes formed.In this study we did not investigate in detail the chemical interaction between platinum and surfactants, but it seems most probable that there is an interaction between anionic platinum complexes and protonated oxyethylate groups of the surfactant molecules.From the point of view of methodology, the extraction with oxyethylated surfactants should be carried out with great care because of a high probability of the formation of w/o emulsions owing to the low interfacial tension. Back-extraction Stage Reversed micellar back-extraction (desolubilization) with polar organic diluents has rarely been studied in the past. Deso- Fig. 4 Absorption spectra of PtIV (against a reagent blank) in all stages of reversed micellar concentration. 1, Feed: saturated Na2SO4 + 3 mol l21 H2SO4 + 0.05 mol l21 HCl; Cf(Pt) = 7.5 31023 mol l21, using a 1.0 cm cell. 2, Extract: Co(Pt) = 7.5 3 1023 mol l21, using a 1.0 cm cell. 3, Desolubilized aqueous solution (back-extraction by chloroform addition up to 30% v/v); Cs(Pt) = 7.5 3 1022 mol l21, using a 0.1 cm cell. Table 2 Results for reversed micellar concentration of PtIV in the back-extraction stage and combined data (95% confidence level). For abbreviations, see text (eqn.(1)] Model system Cs/1022 No.* Metal mol l21 Vo/Vs Rbe (%) Kbe R (%) K 1 Pt 5.89 ± 0.15 110 86.4 ± 4.4 95.0 ± 3.7 62.3 ± 3.4 342 ± 11 2 Pt 5.52 ± 0.14 120 70.9 ± 3.6 89.0 ± 3.5 67.2 ± 3.7 403 ± 14 3 Pt† — — — — — — 4 Pt 6.08 ± 0.14 30 81.3 ± 4.1 24.4 ± 1.0 81.1 ± 4.4 122 ± 4 5 Pt 0.823 ± 0.019 42 80.4 ± 4.1 33.8 ± 0.3 78.4 ± 4.3 (1.64 ± 0.53) 3 103 6 Pt 6.54 ± 0.15 120 83.6 ± 4.2 100 ± 4 60.2 ± 3.3 361 ± 12 Fe 0.107 ± 0.002 120 1.62 ± 0.08 1.95 ± 0.07 1.62 ± 0.08 0.097 ± 0.003 7 Pt 1.00 ± 0.02 33 96.2 ± 4.9 31.8 ± 1.2 15.5 ± 0.8 25.5 ± 0.8 Fe 1.46 ± 0.03 33 36.3 ± 1.8 12.0 ± 0.5 20.4 ± 1.1 33.6 ± 1.0 * Numbering of model systems as in Table 1.† Back-extraction conditions were not found. Fig. 5 Distribution ratio of PtIV versus concentration of X (HBr, NaI or SnCl2). 1, Saturated Na2SO4 + 3 mol l21 H2SO4 + 0.001 mol l21 HCl; Cf(Pt) = 1.0 3 1024 mol l21, X = HBr; 2, saturated Na2SO4 + 3 mol l21 H2SO4 + 0.002 mol l21 HCl; Cf(Pt) = 2.0 3 1024 mol l21, X = NaI; and 3, 1 mol l21 HCl; Cf(Pt) = 5 3 1024 mol l21, X = SnCl2. 230 Analyst, March 1997, Vol. 122lubilization results for the systems studied are given in Table 2. In most cases, the desolubilization of the aqueous solution was carried out by the addition of one volume of chloroform to two volumes of reversed micellar solution. For systems with I2 as complexing agent, no desolubilization could be performed. As an example, consider in more detail finding the optimum backextraction conditions for a reversed micellar solution obtained by extraction from the feed phase of composition 5 3 1026 mol l21 PtIV + 0.05 mol l21 SnCl2 + 1 mol l21 HCl (system No. 5 in Tables 1 and 2). For this model system, the dilution of the reversed micellar solution to separate the solubilized aqueous solution was perfomed with chloroform or chloroform–hexane. Back-extraction data are shown as a schematic phase diagram in Fig. 6. Plotted on the abscissa is the concentration of chloroform (%v/v) in the decane–chloroform–hexane mixture formed after dilution of the reversed micellar solution, and on the ordinate the surfactant concentration (%v/v) in this mixture.The line connecting the axes of rectangular coordinates corresponds to the dilution of the reversed micellar solution with chloroform. The area inside the triangle corresponds to the dilution of the reversed micellar solution with the chloroform–hexane mixture. For example, point A in Fig. 6 corresponds to the dilution of one volume of reversed micellar solution with one volume of chloroform (after dilution the solution contains 5% v/v of the surfactant, 50% v/v of chloroform and 45% v/v of decane). Point B in Fig. 6 corresponds to the dilution of one volume of reversed micellar solution with four volumes of chloroform– hexane (5 + 3) (after dilution the solution contains 2% v/v of the surfactant, 50% v/v of chloroform, 30% v/v of hexane and 18% v/v of decane).It follows from the diagram that it is not always possible to separate the solubilized solution. Thus, no separation takes place at low chloroform concentrations and small dilutions of the reversed micellar solutions and also at high chloroform concentrations and large dilutions (single-phase regions in Fig. 6). Between these regions lie the region of backextraction and a three-phase region. In the region of backextraction the platinum recovery from the organic phase is not uniform.In the shaded portion it exceeds 50%. The best results given in Table 2 were obtained at the dilution of one volume of reversed micellar solution with 20 volumes of the chloroform– hexane (4 + 15). Results for Reversed Micellar Concentration The total reversed micellar concentration factor for platinum is determined as the overall concentration in the stages of extraction (Ke) and back-extraction (Kbe): K = KeKbe = Re ¥ Vw Vo ¥ Rbe ¥ Vo Vs (1) where K = Cs/Cf = total concentration factor, Ke = Co/ Cf = concentration factor in the extraction stage, Kbe = Cs/ Co = concentration factor in the back-extraction stage, C = metal concentration, Re = extraction (%), Rbe = backextraction (%), R = ReRbe/100 = total recovery (%), V = volume and subscripts f, o, w and s represent the feed, extract, raffinate and solubilized aqueous solution, respectively.It can be seen from eqn. (1) that the main difference between reversed micellar concentration and conventional concentration by solvent extraction is that in the back-extraction stage the volume ratio of phases is determined by the solubilizing capacity of the reversed micellar solution.The increase in the concentration factor is limited by the recovery in both the back-extraction and extraction stages. An increase in the recovery in the extraction stage can be achieved either by increasing the distribution ratio (D = Co/Cw) or by decreasing the w/o ratio: Re = 1+ Vw Vo ¥ 1 D Ê Ë Á � � � -1 (2) One of the ways of increasing the concentration in the backextraction stage is to decrease the solubilizing capacity of the reversed micellar solution.This reasoning is illustrated by the experimental results in Tables 1 and 2. The highest contribution to the concentration factor is made by the back-extraction stage. In sulfate–chloride media the back-extraction contribution was the determining factor (approximately 102). For chloride media this contribution was approximately 25, the difference being determined by the solubilizing capacity of the reversed micellar solution, which in turn depends on the type and concentration of the background electrolyte in the feed (a more detailed investigation of this problem will be performed in future studies). For the system 1 mol l21 HCl + 0.05 mol l21 SnCl2, the possibility of increasing the PtIV concentration in the extraction stage was demonstrated: a tenfold increase in the w/o ratio led to the same increase in the concentration factor (the total recovery was not decreased).Unfortunately, we failed to find a system for which high distribution ratios (approximately 103) would combine with a low solubilizing capacity of the reversed micellar solution (Vo/Vs Å 102). In the most promising system (containing I2), the solubilized aqueous solution could not be separated at all. Possibly here we have a situation often encountered in the concentration by solvent extraction when high recoveries values in the extraction stage lead to considerable worsening of the back-extraction. In Tables 1 and 2 are also presented data on the collective extraction of the PtIV–FeIII pair of metals, showing the possibility of a fairly good relative concentration of PtIV from acidic sulfate–chloride media in the presence of a large excess of FeIII.Conclusion The results of this study indicate that reversed micellar concentration with oxyethylated surfactants has some promise for the determination of platinum in dilute acidic chloride and sulfate–chloride aqueous media.The possibility of increasing the concentration factor in both the extraction and backextraction stages has been demonstrated. At the present stage of Fig. 6 Schematic phase diagram of the back-extraction of PtIV from the reversed micellar solution by addition of chloroform or chloroform–hexane. The region with back-extraction exceeding 50% is shaded.Analyst, March 1997, Vol. 122 231development, the problem remaining is the incomplete recovery of platinum in the back-extraction stage, which may result, in the best case, in an approximately 20% underestimation of the platinum content in the feed determined using reversed micellar concentration. Further research efforts should therefore be concentrated on the elucidation of the mechanism of the desolubilization process and on looking for new ways of separating the solubilized aqueous solutions (e.g., with the help of an electric field).From the point of view of methodology, ways of by-passing the problems are possible, such as repeated treatment of the raffinate with a reduced reversed micellar solution (i.e., after distilling off chloroform and hexane) and repeated desolubilization. However, this will complicate the analytical procedure considerably. The use of large volumes of organic diluents and working with small volumes of desolubilized aqueous solution are further drawbacks to desolubilization by dilution. The determination of platinum directly in the reversed micellar solutionferior to desolubilization with regard to absolute concentration, since the concentration of platinum is determined per unit volume of the whole reversed micellar solution and the reversed micellar concentration factor will be determined only by the concentration in the extraction stage.Nevertheless, attempts to determine the metal directly in the reversed micellar phase appear to be promising, the task consisting in finding a procedure which would determine the metal concentration only in the water cores of the micelles.The use of ion-selective electrodes with a hydrophilic membrane may turn out to be the simplest approach. References 1 Savvin, S. V., Chernova, R. K., and Shtykov, S. N., Surface-active Substances, Nauka, Moscow, 1991. 2 Issopoulos, P. B., and Economou, P. T., Fresenius’ J.Anal. Chem., 1992, 342, 439. 3 Pavon, J. L. P., and Cordero, B. M., Analyst, 1992, 117, 215. 4 Vijayalakshmi, Ch. S., Annapragala, L. V., and Gulary, E., Sep. Sci. Technol., 1990, 25, 711. 5 Vijayalakshmi, Ch. S., and Gulary, E., Sep. Sci. Technol., 1992, 27, 173. 6 Apanasenko, B. B., Reznic, A. I., and Sokolova, A. N., Dokl. Akad. Nauk SSSR, 1990, 315, 106. 7 Ovejero-Ecudero, F. J., Angelino, H., and Casamatta, G., J. Dispers. Sci. Technol., 1987, 8, 89. 8 Osseo-Asare, K., Sep.Sci. Technol., 1988, 23, 1269. 9 Robinson, B. H., Steytler, D. C., and Tack, R. D., J. Chem. Soc., Faraday Trans. 1, 1979, 75, 481. 10 Paatero, E., Sjoblom, J., and Datla, S. K., J. Colloid Interface Sci., 1990, 138, 88. 11 Neuman, R. D., Zhou, N.-F., Wu, J., Jones, M. A., Gaonkar, A. G., Park, S. J., and Agrawal, M. L., Sep. Sci. Technol., 1990, 25, 1655. 12 Neuman, R. D., Jones, M. A., and Zhou N.-F., Colloids Surf., 1990, 46, 45. 13 Leodidis, E. B., and Hatton, T.A., Langmuir, 1989, 5, 741. 14 Moutaoukel, J., and Christian, T., Langmuir, 1992, 8, 1039. 15 Bulavchenko, A. I., Batishcheva, E. K., and Torgov, V. G., Russ. Pat., 1 805 990, 1990. 16 Zolotov, Yu. A., and Kuzmin, N. M., Extractional Concentration, Khimiya, Moscow, 1971. 17 Bulavchenko, A. I., and Torgov, V. G., Abstracts of International Organic Substances Solvent Extraction Conference (ISECOS’92), Voronezh, Russia, 1992, Vol. 2, p. 201. 18 Bulavchenko, A. I., Batishcheva, E.K., and Torgov, V. G., Sep. Sci. Technol., 1995, 30, 239. 19 Milner, O. I., and Shipman, G. F., Anal. Chem., 1955, 27, 1476. 20 Cheng, K. L., Bray, R. H., and Kurtz, T., Anal. Chem., 1953, 25, 347. 21 Babaeva, A. B., Dokl. Akad. Nauk SSSR, 1938, 20, 365. 22 Zolotov, Yu. A., Iofa, B. Z., and Chuchalin, L. K., Extraction of Halide Complexes of Metals, Nauka, Moscow, 1973. Paper 6/05773J Received August 19, 1996 Accepted October 14, 1996 232 Analyst, March 1997, Vol. 122 Concentration of Platinum(IV) From Acidic Chloride and Sulfate–Chloride Aqueous Media With Reversed Micellar Solutions of Oxyethylated Surfactant Alexander I.Bulavchenko*, Tat’yana Yu. Podlipskaya, Elena K. Batishcheva and Vladislav G. Torgov Institute of Inorganic Chemistry, Russian Academy of Sciences, Siberian Branch, 630090 Novosibirsk, Russia The possibility of employing reversed micelles of oxyethylated surfactants to concentrate platinum from acidic aqueous media was investigated.A procedure for reversed micellar concentration is described. Platinum concentration was effected by means of desolubilization in a non-traditional back-extraction stage by dilution of the reversed micellar solution with chloroform or a mixture of chloroform with hexane. In acidic sulfate–chloride media the distribution ratio of PtIV increased to about 102–5 3 103 in the presence of Br2 and I2 as complexing agents. In chloride media the distribution ratio could be increased to about 2 3 103 by using SnCl2.The maximum recovery of platinum from the extract did not exceed 85%. With I2 as a complexing agent, no back-extraction could be performed. With this back-extraction procedure, the PtIV concentration factor varied from about 102 to 103 depending on the aqueous feed composition. It was shown spectrophotometrically that PtIV complex species were the same in the feed, extract and desolubilized aqueous solution. The possibility of the spectrophotometric determination of platinum with SnCl2 directly in the reversed micellar solution is demonstrated.Keywords: Reversed micellar extraction; oxyethylated oil-soluble surfactant; platinum(IV); chloride and sulfate–chloride aqueous media Surfactants and reversed micellar solutions are widely used in many analytical procedures as auxiliary agents to increase analytical sensitivity and selectivity.1–3 In recent years there has been a trend to develop procedures with surfactants playing the main role.The use of reversed micellar surfactant solutions for the separation and concentration of metals has attracted considerable attention. Vijayalakshmi and co-workers4,5 studied the equilibrium extraction and concentration of polyvalent metal ions with water-in-oil (w/o) microemulsions and showed the possibility of separating trivalent from divalent metal ions. Apanasenko et al.6 performed the reversed micellar concentration of Al and Ga with a fivefold concentration factor.Earlier, a threefold concentration of NiII was achieved with the help of w/o microemulsions of ionic surfactants.7 The solubilization and complexation of metals in reversed micelles has been studied.8,9 Water-in-oil microemulsion formation and the extraction of metals with mixed micelles of surfactants and various types of extractants has been investigated.10 Neuman and co-workers11,12 considered the role of reversed micelles in extraction with the well known extractant di(2-ethylhexyl)phosphoric acid.Leodidis and Hatton13 studied the selective solubilization of alkali and alkaline earth metal ions with reversed micelles. The possibility of separating metals having similar properties with the help of reversed micelles in liquid membrane processes has been demonstrated for the pair NiII–CoII.14 This brief survey indicates that this area of research is still in its infancy. The unique properties of reversed micellar solutions are very under-utilized and at present there is no serious competition with traditional extraction techniques.We have developed a procedure for the reversed micellar concentration of metals15 differing from solvent extraction concentration16 in having a non-traditional back-extraction stage. In this stage the reversed micelles are broken by heating or by the the addition of chloroform17 and the solubilized metalenriched aqueous solution separates as an aqueous phase (Fig. 1). The non-use of an aqueous stripping agent solution and the low solubilizing capacity of reversed micelles have resulted in a considerably higher concentration of chloride metal complexes (FeIII, PtIV and PdII)18 compared with solvent extraction concentration. The aim of this work was to develop further the reversed micellar concentration procedure to determine low PtIV concentrations in aqueous acidic media. The effects of various complexing agents (Br2 and I2) and SnCl2 on the distribution ratio, recovery and concentration factor in the extraction and back-extraction stages were studied. The method was applied to the determination of low concentrations of PtIV in the presence of FeIII.Experimental Reagents Oxyethylated isononylphenol with an average degree of oxyethylation of 4, manufactured by Diagnosticum (Belgorod, Russia) as Neonol APh9-4 (an analogue of Triton N-42), was used as the oil-soluble micellation surfactant. The hydrophilic– lipophilic balance (HLB) was 9.1 ± 0.3 and the mass fraction of the main substance (the sum of the oxyethylated homologues) was 99.8%.The distribution functions of the homologues had a Poisson shape. The contribution of the homologue with a degree of oxyethylation of 4 was 30%. The surfactant was used as Fig. 1 Scheme for reversed micellar concentration of metals. Analyst, March 1997, Vol. 122 (227–232) 227received. The solvent (reversed micellar solution) contained 10% v/v Neonol APh9-4 in decane.Analytical-reagent grade organic diluents (decane, hexane and chloroform) were used without further purification. All other reagents were chemically pure. Distilled water was used to prepare aqueous solutions. Solutions of PtIV were prepared by diluting a solution of chloroplatinic acid. The acid was obtained by dissolving 99.9% pure metallic platinum in HNO3–HCl (1+3). The stock solution contained 0.2 mol l21 PtIV in 2.4 mol l21 HCl. A solution of FeIII (2.5 mol l21 + 1.5 mol l21 HCl) was obtained by dissolving FeCl3·6H2O in hydrochloric acid.Sulfate–chloride and chloride solutions had the compositions saturated Na2SO4 + 3 mol l21 H2SO4 + 0.05 mol l21 HCl, 4 mol l21 NaCl + 1.2 mol l21 HCl and 1 mol l21 HCl. The complexing agents (HBr and NaI) and SnCl2 were added to the feed as aliquot portions of the solutions: 7.1 mol l21 HBr, 0.5 mol l21 NaI and 0.5 mol l21 SnCl2 + 2 mol l21 HCl. Apparatus Spectrophotometric measurements were performed on a Specord M-40 UV/VIS spectrophotometer (Carl Zeiss, Jena, Germany).Colorimetric determinations of FeIII and PtIV were performed on a KFK-2MP photoelectric colorimeter (ZOMZ, Zagorsk, Russia). Determination of PtIV and FeIII in Aqueous Media Concentrations of PtIV and FeIII in stock and working standard solutions were determined colorimetrically using published procedures with SnCl2 19 and sulfosalicylic acid,20 respectively. Concentration With Reversed Micellar Solutions (Extraction and Back-extraction Stages) The procedure for the reversed micellar concentration of metals is shown schematically in Fig. 2. The solvent extraction stage was performed in a glass separating funnel with a volume of 100–1500 ml. A water-to-oil ratio of 5 : 1 was used in most instances. The feed served as the continuous phase. The phases were contacted for not more than 30 min under mild stirring with a propeller mixer. After extraction, the raffinate was analysed for PtIV and/or FeIII content.The extract was transferred into a conical-bottomed flask (Fig. 1) with polyethylene-coated walls and diluted with chloroform or chloroform–hexane [route (1), Fig. 2]. After the addition of chloroform, the solution became turbid and the solubilized aqueous solution of the metal coagulated into a separate transparent phase. This took 1–2 h for most of the systems. The separated aqueous solution was removed with a glass measuring pipette, measured and analysed for the metal content.Spectrophotometric Determination of PtIV Directly in Reversed Micellar Phase As an alternative procedure, the metal concentration can be determined directly in the reversed micellar phase without micelle breaking and separation of the solubilized aqueous solution [route (2), Fig. 2]. In this approach, the determination Fig. 2 Analytical procedure for reversed micellar concentration of platinum. 228 Analyst, March 1997, Vol. 122of PtIV is effected colorimetrically by a procedure developed for aqueous media. This ‘water’ procedure is applicable here since the interaction of analytical reagents with platinum takes place in the water cores of the micelles.To prepare a standard reversed micellar solution of PtIV, an aliquot of an aqueous solution of PtIV was solubilized by the injection method into the solvent. An aliquot (0.5–3.0 ml) of this standard solution of PtIV or of its extract was then transferred into a 25 ml calibrated flask and diluted with the solvent to about 20 ml [route (2) Fig. 2], then 0.21 ml of an aqueous solution of SnCl2 (2.66 mol l21 SnCl2 + 3 mol l21 HCl) was solubilized into the solution by the injection method. The resulting solution was diluted to volume with the solvent, allowed to stand for 40 min and then scanned photometrically against a reagent blank in a cell with a pathlength of 5 cm at a wavelength of 400 nm. A blank solution was prepared in the same way with the same composition except that it did not contain platinum.The calibration graph was linear for platinum concentrations in the range 1026–1025 mol l21. Determination of platinum directly in the reversed micellar solution was employed to calculate the extraction parameters in a number of systems. Results and Discussion Results for the reversed micellar concentration of PtIV in the extraction stage are given in Table 1. Extraction Stage Extraction of chloride complexes of PtIV The distribution isotherms of PtIV from acidic chloride and sulfate–chloride solutions are shown in Fig. 3. The dependences are linear for low metal concentrations and the distribution ratio of PtIV in the presence of Cl2 did not exceed 12.9 ± 0.7. To determine the PtIV complex species in the media under study, the corresponding absorption spectra were measured. The absorption maximum of PtIV in the visible region was at 453 nm, corresponding to the chloride complex (PtCl6 22).21 The electronic absorption spectra of PtIV measured in all stages of reversed micellar concentration (Fig. 4) indicated that the PtIV complex species was the same as that in the feed, extract and desolubilized aqueous solution. The bathochromic shift of the absorption maximum by 11 nm for the chloride complex of platinum in reversed micelles may be due to its interaction with protonated oxyethylate groups of the surfactant molecules. Effect of Br2 and I2 on extraction of PtIV from acidic sulfate–chloride media Spectrophotometric studies indicated that in acidic sulfate– chloride solutions of PtIV in the presence of Br2, the bromide complex of platinum (PtBr6 22) is formed.The equilibrium in the solutions at room temperature is established within 2 d. In the presence of Br2, the distribution ratio of PtIV increases by an Table 1 Results for reversed micellar concentration of PtIV in the extraction stage (95% confidence level). For abbreviations, see text [eqn.(1)] Feed Model Supporting system electrolyte and No. complexing agent Metal Cf/1024 mol l21 Vw/Vo D Re(%) Ke 1 Saturated Na2SO4 + 3 mol l21 H2SO4 + 0.05 mol l21 HCl Pt 1.72 ± 0.03 5 12.9 ± 0.7 72.1 ± 1.1 3.6 ± 0.2 2 Saturated Na2SO4 + 3 mol l21 H2SO4 + 0.001 mol l21 HCl + 0.05 mol l21 HBr Pt 1.37 ± 0.03 5 88.7 ± 3.5 94.7 ± 1.4 4.7 ± 0.2 3 Saturated Na2SO4 + 3 mol l21 H2SO4 + 0.002 mol l21 HCl + 0.002 mol l21 NaI Pt 1.82 ± 0.04 50 (4.74 ± 0.16) 3 103 99.0 ± 1.5 49.5 ± 2.1 4 1 mol l21 HCl + 0.02 mol l21 SnCl2 Pt 5.00 ± 0.09 5 (1.78 ± 0.07) 3 103 99.7 ± 1.5 5.0 ± 0.2 5 1 mol l21 HCl + 0.05 mol l21 SnCl2 Pt (5.0 ± 0.1) 3 1022 50 (2.00 ± 0.08) 3 103 97.6 ± 1.5 48.8 ± 2.0 6 Saturated Na2SO4 + 3 mol l21 H2SO4 + Pt 1.81 ± 0.03 5 12.9 ± 0.7 72.0 ± 1.1 3.6 ± 0.2 0.05 mol l21 HCl Fe (1.10 ± 0.02) 3 102 5 @0.05 1.00 ± 0.02 0.050 ± 0.002 7 4 mol l21 NaCl + Pt 3.92 ± 0.07 5 0.95 ± 0.05 16.1 ± 0.2 0.80 ± 0.03 1.2 mol l21 HCl Fe 4.34 ± 0.09 5 6.4 ± 0.4 56.2 ± 0.8 2.8 ± 0.1 Fig. 3 Distribution isotherms of PtIV and FeIII from acidic aqueous media: 1, saturated Na2SO4 + 3 mol l21 H2SO4 + 0.05 mol l21 HCl; 2–5, 4 mol l21 NaCl + 1.2 mol l21 HCl, with 1 and 4, PtIV, 2, FeIII 3, FeIII in the presence of PtIV [Cf(Fe) = Cf(Pt)] and 5, PtIV in the presence of FeIII [Cf(Pt) = Cf(Fe)]. Analyst, March 1997, Vol. 122 229order of magnitude and shows a maximum (Fig. 5, curve 1) at a 500-fold excess of HBr relative to PtIV. A greater increase in the distribution ratio (to about 5000) was obtained with I2 as the complexing agent.The dependence shows a broad maximum at a tenfold excess of I2 (Fig. 5, curve 2). Effect of SnCl2 on extraction of PtIV from dilute hydrochloric acid media The logarithmic dependence of the distribution ratio of PtIV when concentrated from 1 mol l21 HCl is linear over a wide range of SnCl2 concentrations (Fig. 5, curve 3). The addition of a 40-fold molar excess of SnCl2 relative to PtIV increases the distribution ratio by four orders of magnitude.Extraction of chloride complex of PtIV in the presence of FeIII The extraction of these two metals, which frequently accompany each other, is considered for the case of the collective extraction of PtIV and FeIII from acidic chloride and sulfate– chloride media in the absence of complexing agents. The distribution ratio of PtIV from chloride media at low concentrations does not change even at a 50-fold molar excess of FeIII.It was also found that in the extraction of metals with equal concentrations in the systems being studied, the distribution isotherms of FeIII and PtIV (Fig. 3) do not change in the region of constancy of the distribution ratios. Hence in the systems studied the metals are extracted into reversed micelles independently of each other. The results for the extraction of PtIV with oxyethylated surfactants in decane from acidic chloride and sulfate–chloride media follow the general regularities found for the extraction of platinum with oxygen-containing extractants (alcohols, ethers, etc.).22 The increase in recovery when using complexing agents results from the increased stability of the complexes formed.In this study we did not investigate in detail the chemical interaction between platinum and surfactants, but it seems most probable that there is an interaction between anionic platinum complexes and protonated oxyethylate groups of the surfactant molecules.From the point of view of methodology, the extraction with oxyethylated surfactants should be carried out with great care because of a high probability of the formation of w/o emulsions owing to the low interfacial tension. Back-extraction Stage Reversed micellar back-extraction (desolubilization) with polar organic diluents has rarely been studied in the past. Deso- Fig. 4 Absorption spectra of PtIV (against a reagent blank) in all stages of reversed micellar concentration. 1, Feed: saturated Na2SO4 + 3 mol l21 H2SO4 + 0.05 mol l21 HCl; Cf(Pt) = 7.5 31023 mol l21, using a 1.0 cm cell. 2, Extract: Co(Pt) = 7.5 3 1023 mol l21, using a 1.0 cm cell. 3, Desolubilized aqueous solution (back-extraction by chloroform addition up to 30% v/v); Cs(Pt) = 7.5 3 1022 mol l21, using a 0.1 cm cell. Table 2 Results for reversed micellar concentration of PtIV in the back-extraction stage and combined data (95% confidence level). For abbreviations, see text (eqn.(1)] Model system Cs/1022 No.* Metal mol l21 Vo/Vs Rbe (%) Kbe R (%) K 1 Pt 5.89 ± 0.15 110 86.4 ± 4.4 95.0 ± 3.7 62.3 ± 3.4 342 ± 11 2 Pt 5.52 ± 0.14 120 70.9 ± 3.6 89.0 ± 3.5 67.2 ± 3.7 403 ± 14 3 Pt† — — — — — — 4 Pt 6.08 ± 0.14 30 81.3 ± 4.1 24.4 ± 1.0 81.1 ± 4.4 122 ± 4 5 Pt 0.823 ± 0.019 42 80.4 ± 4.1 33.8 ± 0.3 78.4 ± 4.3 (1.64 ± 0.53) 3 103 6 Pt 6.54 ± 0.15 120 83.6 ± 4.2 100 ± 4 60.2 ± 3.3 361 ± 12 Fe 0.107 ± 0.002 120 1.62 ± 0.08 1.95 ± 0.07 1.62 ± 0.08 0.097 ± 0.003 7 Pt 1.00 ± 0.02 33 96.2 ± 4.9 31.8 ± 1.2 15.5 ± 0.8 25.5 ± 0.8 Fe 1.46 ± 0.03 33 36.3 ± 1.8 12.0 ± 0.5 20.4 ± 1.1 33.6 ± 1.0 * Numbering of model systems as in Table 1.† Back-extraction conditions were not found. Fig. 5 Distribution ratio of PtIV versus concentration of X (HBr, NaI or SnCl2). 1, Saturated Na2SO4 + 3 mol l21 H2SO4 + 0.001 mol l21 HCl; Cf(Pt) = 1.0 3 1024 mol l21, X = HBr; 2, saturated Na2SO4 + 3 mol l21 H2SO4 + 0.002 mol l21 HCl; Cf(Pt) = 2.0 3 1024 mol l21, X = NaI; and 3, 1 mol l21 HCl; Cf(Pt) = 5 3 1024 mol l21, X = SnCl2. 230 Analyst, March 1997, Vol. 122lubilization results for the systems studied are given in Table 2. In most cases, the desolubilization of the aqueous solution was carried out by the addition of one volume of chloroform to two volumes of reversed micellar solution. For systems with I2 as complexing agent, no desolubilization could be performed. As an example, consider in more detail finding the optimum backextraction conditions for a reversed micellar solution obtained by extraction from the feed phase of composition 5 3 1026 mol l21 PtIV + 0.05 mol l21 SnCl2 + 1 mol l21 HCl (system No. 5 in Tables 1 and 2). For this model system, the dilution of the reversed micellar solution to separate the solubilized aqueous solution was perfomed with chloroform or chloroform–hexane. Back-extraction data are shown as a schematic phase diagram in Fig. 6. Plotted on the abscissa is the concentration of chloroform (%v/v) in the decane–chloroform–hexane mixture formed after dilution of the reversed micellar solution, and on the ordinate the surfactant concentration (%v/v) in this mixture.The line connecting the axes of rectangular coordinates corresponds to the dilution of the reversed micellar solution with chloroform. The area inside the triangle corresponds to the dilution of the reversed micellar solution with the chloroform–hexane mixture. For example, point A in Fig. 6 corresponds to the dilution of one volume of reversed micellar solution with one volume of chloroform (after dilution the solution contains 5% v/v of the surfactant, 50% v/v of chloroform and 45% v/v of decane). Point B in Fig. 6 corresponds to the dilution of one volume of reversed micellar solution with four volumes of chloroform– hexane (5 + 3) (after dilution the solution contains 2% v/v of the surfactant, 50% v/v of chloroform, 30% v/v of hexane and 18% v/v of decane).It follows from the diagram that it is not always possible to separate the solubilized solution. Thus, no separation takes place at low chloroform concentrations and small dilutions of the reversed micellar solutions and also at high chloroform concentrations and large dilutions (single-phase regions in Fig. 6). Between these regions lie the region of backextraction and a three-phase region. In the region of backextraction the platinum recovery from the organic phase is not uniform.In the shaded portion it exceeds 50%. The best results given in Table 2 were obtained at the dilution of one volume of reversed micellar solution with 20 volumes of the chloroform– hexane (4 + 15). Results for Reversed Micellar Concentration The total reversed micellar concentration factor for platinum is determined as the overall concentration in the stages of extraction (Ke) and back-extraction (Kbe): K = KeKbe = Re ¥ Vw Vo ¥ Rbe ¥ Vo Vs (1) where K = Cs/Cf = total concentration factor, Ke = Co/ Cf = concentration factor in the extraction stage, Kbe = Cs/ Co = concentration factor in the back-extraction stage, C = metal concentration, Re = extraction (%), Rbe = backextraction (%), R = ReRbe/100 = total recovery (%), V = volume and subscripts f, o, w and s represent the feed, extract, raffinate and solubilized aqueous solution, respectively.It can be seen from eqn. (1) that the main difference between reversed micellar concentration and conventional concentration by solvent extraction is that in the back-extraction stage the volume ratio of phases is determined by the solubilizing capacity of the reversed micellar solution.The increase in the concentration factor is limited by the recovery in both the back-extraction and extraction stages. An increase in the recovery in the extraction stage can be achieved either by increasing the distribution ratio (D = Co/Cw) or by decreasing the w/o ratio: Re = 1+ Vw Vo ¥ 1 D Ê Ë Á � � � -1 (2) One of the ways of increasing the concentration in the backextraction stage is to decrease the solubilizing capacity of the reversed micellar solution.This reasoning is illustrated by the experimental results in Tables 1 and 2. The highest contribution to the concentration factor is made by the back-extraction stage. In sulfate–chloride media the back-extraction contribution was the determining factor (approximately 102). For chloride media this contribution was approximately 25, the difference being determined by the solubilizing capacity of the reversed micellar solution, which in turn depends on the type and concentration of the background electrolyte in the feed (a more detailed investigation of this problem will be performed in future studies). For the system 1 mol l21 0.05 mol l21 SnCl2, the possibility of increasing the PtIV concentration in the extraction stage was demonstrated: a tenfold increase in the w/o ratio led to the same increase in the concentration factor (the total recovery was not decreased).Unfortunately, we failed to find a system for which high distribution ratios (approximately 103) would combine with a low solubilizing capacity of the reversed micellar solution (Vo/Vs Å 102). In the most promising system (containing I2), the solubilized aqueous solution could not be separated at all. Possibly here we have a situation often encountered in the concentration by solvent extraction when high recoveries values in the extraction stage lead to considerable worsening of the back-extraction. In Tables 1 and 2 are also presented data on the collective extraction of the PtIV–FeIII pair of metals, showing the possibility of a fairly good relative concentration of PtIV from acidic sulfate–chloride media in the presence of a large excess of FeIII.Conclusion The results of this study indicate that reversed micellar concentration with oxyethylated surfactants has some promise for the determination of platinum in dilute acidic chloride and sulfate–chloride aqueous media.The possibility of increasing the concentration factor in both the extraction and backextraction stages has been demonstrated. At the present stage of Fig. 6 Schematic phase diagram of the back-extraction of PtIV from the reversed micellar solution by addition of chloroform or chloroform–hexane. The region with back-extraction exceeding 50% is shaded.Analyst, March 1997, Vol. 122 231development, the problem remaining is the incomplete recovery of platinum in the back-extraction stage, which may result, in the best case, in an approximately 20% underestimation of the platinum content in the feed determined using reversed micellar concentration. Further research efforts should therefore be concentrated on the elucidation of the mechanism of the desolubilization process and on looking for new ways of separating the solubilized aqueous solutions (e.g., with the help of an electric field).From the point of view of methodology, ways of by-passing the problems are possible, such as repeated treatment of the raffinate with a reduced reversed micellar solution (i.e., after distilling off chloroform and hexane) and repeated desolubilization. However, this will complicate the analytical procedure considerably. The use of large volumes of organic diluents and working with small volumes of desolubilized aqueous solution are further drawbacks to desolubilization by dilution. The determination of platinum directly in the reversed micellar solution is inferior to desolubilization with regard to absolute concentration, since the concentration of platinum is determined per unit volume of the whole reversed micellar solution and the reversed micellar concentration factor will be determined only by the concentration in the extraction stage.Nevertheless, attempts to determine the metal directly in the reversed micellar phase appear to be promising, the task consisting in finding a procedure which would determine the metal concentration only in the water cores of the micelles. The use of ion-selective electrodes with a hydrophilic membrane may turn out to be the simplest approach. References 1 Savvin, S. V., Chernova, R. K., and Shtykov, S. N., Surface-active Substances, Nauka, Moscow, 1991. 2 Issopoulos, P. B., and Economou, P. T., Fresenius’ J. Anal. Chem., 1992, 342, 439. 3 Pavon, J. L. P., and Cordero, B. M., Analyst, 1992, 117, 215. 4 Vijayalakshmi, Ch. S., Annapragala, L. V., and Gulary, E., Sep. Sci. Technol., 1990, 25, 711. 5 Vijayalakshmi, Ch. S., and Gulary, E., Sep. Sci. Technol., 1992, 27, 173. 6 Apanasenko, B. B., Reznic, A. I., and Sokolova, A. N., Dokl. Akad. Nauk SSSR, 1990, 315, 106. 7 Ovejero-Ecudero, F. J., Angelino, H., and Casamatta, G., J. Dispers. Sci. Technol., 1987, 8, 89. 8 Osseo-Asare, K., Sep. Sci. Technol., 1988, 23, 1269. 9 Robinson, B. H., Steytler, D. C., and Tack, R. D., J. Chem. Soc., Faraday Trans. 1, 1979, 75, 481. 10 Paatero, E., Sjoblom, J., and Datla, S. K., J. Colloid Interface Sci., 1990, 138, 88. 11 Neuman, R. D., Zhou, N.-F., Wu, J., Jones, M. A., Gaonkar, A. G., Park, S. J., and Agrawal, M. L., Sep. Sci. Technol., 1990, 25, 1655. 12 Neuman, R. D., Jones, M. A., and Zhou N.-F., Colloids Surf., 1990, 46, 45. 13 Leodidis, E. B., and Hatton, T. A., Langmuir, 1989, 5, 741. 14 Moutaoukel, J., and Christian, T., Langmuir, 1992, 8, 1039. 15 Bulavchenko, A. I., Batishcheva, E. K., and Torgov, V. G., Russ. Pat., 1 805 990, 1990. 16 Zolotov, Yu. A., and Kuzmin, N. M., Extractional Concentration, Khimiya, Moscow, 1971. 17 Bulavchenko, A. I., and Torgov, V. G., Abstracts of International Organic Substances Solvent Extraction Conference (ISECOS’92), Voronezh, Russia, 1992, Vol. 2, p. 201. 18 Bulavchenko, A. I., Batishcheva, E. K., and Torgov, V. G., Sep. Sci. Technol., 1995, 30, 239. 19 Milner, O. I., and Shipman, G. F., Anal. Chem., 1955, 27, 1476. 20 Cheng, K. L., Bray, R. H., and Kurtz, T., Anal. Chem., 1953, 25, 347. 21 Babaeva, A. B., Dokl. Akad. Nauk SSSR, 1938, 20, 365. 22 Zolotov, Yu. A., Iofa, B. Z., and Chuchalin, L. K., Extraction of Halide Complexes of Metals, Nauka, Moscow, 1973. Paper 6/05773J Received August 19, 1996 Accepted October 14, 1996 232 Analyst, March 1997, Vol. 122

ISSN:0003-2654    

DOI:10.1039/a605773j

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Determination of Trace Amounts of Zinc in Water Samples by Flow Injection Isotope Dilution Inductively Coupled Plasma Mass Spectrometry Tarn-Jiun Hwang and Shiuh-Jen Jiang* Department of Chemistry, National Sun Yat-Sen University, Kaohsiung, 80424, Taiwan Flow injection isotope dilution ICP-MS was applied to the determination of zinc in several water samples. A matrix separation and preconcentration system was used for the separation of Mg, Ca, S and Cl matrix and preconcentration of trace amounts of zinc in high salt content samples.A complete preconcentration cycle was accomplished in 4 min. The isotope ratio for each injection was calculated from the peak areas of the flow injection peaks. The precision for the isotope ratio determination was better than 1.7%. A detection limit of 0.014 ng ml21 was obtained for zinc with this method. The method was successfully applied to the determination of trace levels of zinc in SLRS-2 riverine water, SLEW-2 estuarine water, NASS-3 open ocean sea-water and CASS-3 nearshore sea-water reference samples.Keywords: Flow injection; isotope dilution; inductively coupled plasma mass spectrometry; zinc; matrix separation; water samples ICP-MS is a powerful technique for trace multielement and isotopic analysis.1–3 This technique combines the characteristics of the ICP for atomizing and ionizing injected material with the sensitivity of MS. It has been satisfactorily applied to the analysis of diverse samples.However, highly saline solutions can cause both spectroscopic and non-spectroscopic interferences. 4–13 Spectral overlaps are seen from polyatomic ions derived from matrix elements such as Na, Ca and Cl. Changes in analyte count rates are observed with high levels of salts or heavy matrix ions. Orifice plugging is also a problem for samples of high solid content. Several calibration methods have been adapted to ICP-MS to help deal with matrix effects.10,14–19 The matrix interference problems can also be dealt with by separating the analytes from the matrix. Several matrix–analyte separation and preconcentration techniques have been adapted to ICP-MS analysis.20–30 We describe here the implementation of an on-line matrix–analyte separation and preconcentration technique with ICP-MS to determine Zn, using a miniature column packed with SO3– quinolin-8-ol carboxymethyl-cellulose.28–31 Isotope dilution (ID) techniques have been applied in several previous ICP-MS applications.14,17,27,29,32–34 Isotope dilution is well recognized as a definitive analytical technique for the determination of the trace elements.Since another isotope of the same element represents the ideal internal standard for that element, ID results are expected to be highly accurate even when the sample contains high concentrations of concomitant elements and/or there is sample loss during the preparation or pretreatment process. In this work, flow injection (FI)–ID–ICP-MS was used to determine the concentrations of zinc in water samples.Matrix separation and preconcentration were used for separation of interfering ions. The influence of instrumental operating conditions, preconcentration system conditions and non-spectroscopic and spectroscopic interferences due to the matrix on the precision and accuracy of isotope ratio determination was also investigated. The method was applied to the determination of zinc in SLRS-2 riverine water, SLEW-2 estuarine water, NASS-3 open ocean sea-water and CASS-3 nearshore seawater reference samples.Experimental ICP-MS Device and Conditions An ELAN 5000 ICP-MS instrument (Perkin-Elmer SCIEX, Thornhill, ON, Canada) was used. It was equipped with an ultrasonic nebulizer (U-5000 AT+ Cetac, Omaha, NE, USA). The gas flow rates were controlled by a four channel mass flow controller. The ICP and ultrasonic nebulizer operating conditions were selected to maximize the sensitivity of zinc using a simple flow injection system.A 100 ml volume of 10 ng ml21 Zn solution was injected into a 0.5 mol l21 HNO3 carrier solution and analyzed by ICP-MS. The operating conditions used throughout this work are summarized in Table 1. Data acquisition parameters used for isotope ratio measurement are listed in Table 1. The isotopes monitored were 66Zn and 67Zn. Version 2.2 of the ELAN 5000 software was used. Under the combinations of dwell time and sweeps/reading, a data point could be obtained in < 1 s for each isotope.The isotope ratio was calculated from the peak areas of the flow injection peaks. The mean isotope ratio and RSD were calculated from isotope ratios measured from several repeated injections. Table 1 Equipment and operating conditions* ICP-MS PE SCIEX ELAN 5000 Plasma conditions— Outer gas flow rate 16 l min21 Auxiliary gas flow rate 0.9 l min21 Nebulizer gas flow rate 1.1 l min21 Rf power 1.1 kW Mass spectrometer settings*— Photon stop (S2) 210.05 V Bessel box barrel +15.34 V Einzel lenses 1 and 3 (E1) 20.04 V Bessel box end lens (P) 279.10 V Resolution Normal Dwell time 40 ms Sweeps/reading 5 Readings/replicate 135 Number of replicates 1 Points/spectral peak 1 Isotopes monitored 66Zn and 67Zn Ultrasonic nebulizer— Desolvation tube temperature 135 °C Condenser temperature 0 °C * Definitions and descriptions of these terms are given in user manuals.Analyst, March 1997, Vol. 122 (233–237) 233Flow Injection In order to investigate the effect of the sample matrix on the precision and accuracy of isotope ratio determinations, a simple FI system was built. The system was composed of a Rheodyne (Cotati, CA, USA) Model 5041 rotary sample injection valve and a Gilson (Middleton, WI, USA) Minipuls-3 peristaltic pump. A 100 ml portion of the test mixture solution was injected into a 0.5 mol l21 nitric acid carrier solution and analyzed by ICP-MS.Preconcentration System A laboratory-made sample pretreatment system was used to separate trace amounts of zinc from matrix elements. A schematic diagram of the matrix separation and preconcentration system is shown in Fig. 1. This system was composed of three Gilson Minipuls-3 peristaltic pumps and two Rheodyne Model 5041 PTFE six-port four-way rotary valves whose switching was actuated by Rheodyne Model 5701 pneumatic actuators and Rheodyne Model 7163 solenoid valve kits.The automation of the flow injection system was controlled by a personal computer through a laboratory-made I/O card and switching circuit. The operating procedure for the preconcentration system is described below. The first rotary valve was used as a selection valve between buffer and sample. At the start of a preconcentration cycle, sample was pumped to wash the system while buffer solution was pumped to condition the column. After a suitable period, valve 1 was switched to load the sample and pump 2 was stopped. After 10.0 ml of sample had been loaded, pump 1 was stopped and valve 1 was rotated back to buffer solution to wash the column.Then valve 2 was switched to let 0.5 mol l21 HNO3 to pass through the column to elute the retained metal ion into the ultrasonic nebulizer. After elution, both valves were switched back and pump 2 was turned on to let the buffer solution condition the column for the next analysis. An Omnifit (Cambridge, UK) high performance column (50 mm 3 10 mm id) was used.It was equipped with a moveable end piece to allow for adjustment of the amount of ion-exchange material. During this study, the column was packed with about 150 mg of SO3–quinolin-8-ol carboxymethylcellulose (Knapp Logistic Automation, Graz, Austria). At a suitable pH, commonly encountered matrix components such as alkali and alkaline earth elements are not strongly retained on the ionexchange materials and are separated from the elements of interest.The operating parameters of the preconcentration system used in this experiment are shown in Table 2. The preconcentration system was connected to the nebulizer with PTFE tubing (20 cm 3 0.8 mm id). Reagents Mg(NO3)2, Ca(NO3)2, (NH4)2SO4 and NaCl were obtained from Fisher (Fair Lawn, NJ, USA). A zinc stock standard solution (1.0 3 106 ng ml21) was obtained from Fisher. Enriched isotope (67Zn) was purchased from Cambridge Isotope Laboratories (Andover, MA, USA).This stable isotope was received as an oxide. A stock standard solution of zinc of approximately 100 mg l21 was prepared by dissolution of an accurately weighed amount of the material in nitric acid and dilution to volume. The concentration of the spiked solution was verified by reverse spike isotope ID–ICP. All buffers and eluents were prepared using high purity deionized water (Milli-Q reagent water system, Millipore, Bedford, MA, USA). The ammonium acetate buffer solutions were adjusted with ammonia solution and acetic acid in the pH range 2–8.All the buffers were purified by triple filtration on an SO3–quinolin-8-ol carboxymethyl-cellulose column. After the uptake step, the absorbed analyte was eluted with 0.5 mol l21 HNO3 (Fisher, trace metal grade). Evaluation of Retention The effect of the pH of the loading solution on the chelation of zinc was studied with the preconcentration system in an off-line mode. Solutions of 100 ng ml21 zinc were buffered to suitable pH values, then 10 ml of these solutions were ‘loaded’ on to the SO3–quinolin-8-ol carboxymethylcellulose column at a flow rate of 2.3 ml min21.Finally, any retained metal was eluted with 5 ml of 1.0 mol l21 HNO3 into a 10 ml calibrated flask and the collected eluate was diluted to volume with distilled, deionized water. The amounts of metals were then determined by ICPMS. Recoveries were calculated against the theoretical concentrations.The tolerance of the matrix separation system for co-existing matrix ions was examined by adding relatively high concentrations of Mg2+, Ca2+, SO4 22 and Cl2 to 1 ng ml21 of zinc metal solutions (for simple FI analysis 5 ng ml21 Zn solution was used). Each solution was injected into FI–ICP-MS and matrix separation–ICP-MS systems successively and the isotope ratio of Zn was determined for comparison. Sample Preparation SLRS-2 riverine water reference material for trace metals, SLEW-2 estuarine water reference material for trace metals, Fig. 1 Schematic diagram of matrix–analyte separation and preconcentration system. This figure shows the flow path of the system in column wash step. Table 2 Preconcentration system and operating conditions Column SO3–quinolin-8-ol carboxymethylcellulose Buffer solution pH 4; 0.1 mol l21 ammonium acetate Eluent 0.5 mol l21 HNO3 Pump 1 (sample) flow rate 10.0 ml min21 Pump 2 (buffer) flow rate 4.5 ml min21 Pump 3 (eluent) flow rate 3.0 ml min21 Preconcentration method— Pump in Valve Step operation position Duration/s System wash 1, 2, 3 1 load 20 2 load Sample loading 1, 3 1 inject 60 2 load Column wash 2, 3 1 load 40 2 load Elution 3 1 load 60 2 inject Column wash 2, 3 1 load 60 2 load 234 Analyst, March 1997, Vol. 122NASS-3 open ocean sea-water reference material for trace metals and CASS-3 nearshore sea-water reference material for trace metals (National Research Council of Canada, Ottawa, Canada) were obtained to demonstrate the applicability of the method to real samples.A 25 ml aliquot of these acidified reference materials was adjusted to pH 4 with 5 ml of 1.0 mol l21 purified ammonium acetate buffer solution. After a suitable amount of enriched isotope has been added, this solution was diluted to 50 ml with distilled, deionized water. The analyte concentration in the sample was calculated using the equation given in a previous paper.27 Owing to a mass bias effect, the intensities obtained during isotope ratio determinations were used to calculate the isotopic abundance of zinc.Since the mass bias effect could be factored out during ID calculation, in this study the measured isotope ratio did not correct the mass bias effect. Results and Discussion Effect of pH on Retention The amount of zinc chelated with SO3–quinolin-8-ol carboxymethylcellulose is related to the pH of the solution, and this effect was examined for Zn in the pH range 2–8.The results are shown in Fig. 2. Zinc was chelated quantitatively in the pH range 4–8. For the analysis of the sea-water sample, a solution with pH between 3.5 to 4 should be chosen; at this pH, Zn was complexed by SO3–quinolin-8-ol carboxymethylcellulose and was retained on the column relatively completely. Meanwhile, as shown in Fig. 2, the interfering ions Mg2+ and Ca2+ chelated with the resin only when the pH value was > 4. Most of the Mg and Ca wash through the column at pH < 4, which can be used to reduce the interferences of Mg and Ca in the ICP-MS analysis.In subsequent experiments, all the loading sample solutions were adjusted to pH 4. Relationship Between Sample Loading Flow Rate and Recovery The sample uptake rate did not affect the recovery significantly in the range 3–13 ml min21. For the purposes of trace Zn concentration, a much faster sample flow rate should be used to save experimental time, and a 10 ml min21 sample loading flow rate was used in the following experiments.Selection of Eluent and Flow Rate Owing to the formation of extra molecular ion interferences when HCl and HClO4 were used as eluents, in order to avoid any possible interferences and reduce the background, these acids were not used in this study. The elution of Zn2+ from the column was examined with HNO3 at various concentrations. The recovery of Zn was complete if the concentration of HNO3 was !0.5 mol l21, since it was difficult to nebulize a concentrated acid solution owing to the viscosity and density and, when a high concentration of acid was nebulized, matrix effects were observed.To avoid matrix effects and signal suppression by high acid levels, the lowest possible acid concentration for elution which yielded 100% recovery was used. In subsequent experiments, 0.5 mol l21 HNO3 was used as the eluent to obtain best recovery and elution peaks for Zn. The recovery of Zn at pH 4 for replicate measurements (n = 7), using 10 ml of 1 ng ml21 standard solution, was 100.2 ± 0.4%.Fig. 3 shows the effect of the eluent flow rate on the ion signal. An increase in the eluent flow rate decreased the width of the elution peak and saved analysis time. However, an increased eluent flow rate also decreased the nebulization efficiency, causing the ion signal to decrease significantly. The effect of eluent flow rate on the peak shape of the elution curve was similar to that in previous work.28 Hence an eluent flow rate of 3.0 ml min21 was selected to increase the ion signals in the following experiments, although a slightly greater duration of analysis was required.Under the selected preconcentration conditions, a complete preconcentration cycle was accomplished in 4 min. A typical elution curve for a solution containing 1 ng ml21 of Zn is shown in Fig. 4. Under the operating conditions of the preconcentration system, the peak half-width for the flow injection peak was about 12 s, which is equivalent to 0.6 ml; this means that with an injection volume of 10 ml a preconcentration factor of about 17 could be obtained for Zn.Detection Limit and Reproducibility The calibration plot based on peak area for 66Zn was linear with a correlation coefficient better than 0.9995 at levels near the detection limit up to at least 50 ng ml21. The detection limit calculated from the calibration plot was 0.014 ng ml21. It was based on the conventional definition of the concentration of the analyte yielding a signal equivalent to three times the standard deviation of the blank.The detection limit obtained by this method is comparable or superior to previous results obtained with ICP-MS detection.35–38 The reagent blank was 0.13 ng ml21 for Zn. A possible explanation is that the system is contaminated with Zn. Zinc is a common contaminant in the Fig. 2 Dependence of recovery on pH of the loading solution.The loaded sample volume was 10 ml at a flow rate of 2.3 ml min21. The flow rate of the eluent (1.0 mol l21 HNO3) was 2.0 ml min21. Fig. 3 Dependence of ion signal on eluent flow rate. The eluent was 0.5 mol l21 HNO3 solution. The loaded sample volume was 10 ml at a flow rate of 10 ml min21. All data are relative to the first point. Analyst, March 1997, Vol. 122 235reagent and solutions; a better detection limit is to be expected with much more purified reagents. The repeatability was determined by seven injections of a 1 ng ml21 Zn solution.The repeatability of the peak areas of these seven injections was 1.4%. Non-spectroscopic and Spectroscopic Interferences As described earlier, the retentions of matrix elements on the ion-exchange material are usually different from that of zinc, hence zinc can be separated from the matrix and collected on the cellulose. The effectiveness of matrix elimination with this separation system is demonstrated by the following experiments.Solutions containing zinc and matrix elements were analyzed by matrix separation–ICP-MS and the results were compared with those obtained using a simple FI–ICP-MS system. Table 3 shows the results of adding various concentrations of S, Cl, Mg and Ca ions to a test solution containing 5 ng ml21 Zn. As shown, the 66Zn/67Zn ratio was increased when 500 mg ml21 S was added, which indicated an interference at m/z 66 from 34S16O16O+ and/or 32S34S+.39 Moreover, the 66Zn/67Zn ratio increased significantly when 1000 mg ml21 of Mg was added, which indicated an interference at m/z 66 from 40Ar26Mg+, although the 66Zn/67Zn ratio did not change significantly when 250 mg ml21 of Ca and 15000 mg ml21 of Cl were added.The precision of the isotope ratio determination was poor when simple FI sample introduction was used. This could be due to the suppression of the ion signal by the high concentration matrix elements. Finally, a solution containing 500 mg ml21 S, 250 mg ml21 Ca, 1000 mg ml21 Mg and 15 000 mg ml21 Cl was spiked to 5 ng ml21 Zn standard solution. The concentrations of the matrix ions were about half of the concentrations of these ions in the sea-water matrix.As shown in Table 3, the 66Zn/67Zn ratio increased to 16.34 when this matrix solution was added. Table 3 also shows that a stable 66Zn/67Zn ratio could be obtained when S, Cl, Mg and Ca were separated from the analyte with the matrix separation system.Hence, for the analysis of highly saline samples, the interfering ions must be separated before Zn is determined. In general, in order to achieve accurate, reliable and sensitive results, preconcentration and separation are needed when the concentrations of analyte elements in the original material or the prepared solution are too small to be determined directly by MS or when matrix elements interfere with the determination. The precision (RSD 0.7–1.7%), shown in Table 3, was slightly worse than that obtained for isotope ratio measurements during continuous introduction of aqueous solutions.Moreover, the precision for the measured isotope ratio was slightly worse than that calculated by the counting statistics (RSD 0.5–0.6%).40 A poorer precision in isotope ratio measurements is to be expected when transient sample introduction techniques are used because the ions of interest are observed for a shorter time and averaging of nebulizer fluctuations, etc., is less efficient.The slight but reproducible differences between the found and expected values are similar to those commonly seen in isotope ratio measurements with ICP-MS and were probably caused by some mass discrimination in ion extraction, focusing, mass analysis and detection.38 Determination of Zinc in Water Samples In order to demonstrate that the system is effective for practical analysis, four water reference samples (NRCC SLRS-2, SLEW- 2, NASS-3 and CASS-3) were analyzed.The amount of Zn in each sample was determined by ID–ICP-MS after on-line sample pretreatment using the matrix separation and preconcentration method. In order to decrease the retention of alkaline earth ions when analyzing sea-water samples, a low pH buffer solution (pH 4) and careful washing of the column with buffer solution were used. The results of the analysis of standard reference materials are given in Table 4. Good agreement Fig. 4 Typical elution curve for 1 ng ml21 Zn solution. The loaded sample volume was 10 ml at a flow rate of 10 ml min21. The eluent (0.5 mol l21 HNO3) flow rate was 3.0 ml min21. Time zero represents the start of elution. Table 3 Isotope ratios of Zn in various matrices with different methods of analysis* 66Zn/67Zn† Matrix Sample composition FI‡ separation§ Zn standard 6.41 ± 0.07 6.47 ± 0.05 Zn standard + S 500 mg ml21 7.59 ± 0.53 6.35 ± 0.11 Zn standard + Ca 250 mg ml21 6.60 ± 1.07 6.45 ± 0.08 Zn standard + Mg 1000 mg ml21 11.81 ± 0.71 6.59 ± 0.09 Zn standard + Cl 15 000 mg ml21 6.64 ± 1.56 6.53 ± 0.08 Zn standard + S 500 mg ml21 + Ca 250 mg ml21 + Mg 1000 mg ml21 + Cl 15 000 mg ml21 16.34 ± 8.55 6.49 ± 0.06 * The expected value of 66Zn/67Zn is 6.77.† Mean of three measurements ± standard deviation. ‡ FI carrier (0.5 mol l21 HNO3) flow rate = 3.0 ml min21; 100 ml sample loop. Concentration of Zn standard solution used in FI, 5 ng ml21. § The operating conditions for the matrix– analyte separation and preconcentration system are given in Table 2.Concentration of Zn standard solution used in matrix separation, 1 ng ml21. Table 4 Concentrations of Zn in reference materials as measured by ID–online preconcentration–ICP-MS (n = 3) Concentration*/ng ml21 Sample Found Certified SLRS-2 riverine water 3.35 ± 0.06 3.33 ± 0.15 SLEW-2 estuarine water 1.12 ± 0.05 1.10 ± 0.14 CASS-3 nearshore sea-water 1.30 ± 0.17 1.24 ± 0.25 NASS-3 open ocean sea-water 0.208 ± 0.035 0.178 ± 0.025 * Mean ± 95% confidence limit. 236 Analyst, March 1997, Vol. 122between the experimental results and certified values show the applicability of this system to real sample analysis. Conclusion The use of ICP-MS and an SO3–quinolin-8-ol carboxymethylcellulose miniature column flow system provide a simple, rapid and accurate technique to determine routinely trace amounts of Zn in water samples. This preconcentration method yielded an enrichment factor of 17 and sample consumption was thus reduced.This research was supported by a grant from the National Science Council of the Republic of China under contract NSC 85-2621-M- 110-009. References 1 Houk, R. S., Fassel, V. A., Flesch, G. D., Svec, H. J., Gray, A. L., and Taylor, C. E., Anal. Chem., 1980, 52, 2283. 2 Jarvis, K. E., Gray, A. L., and Houk, R. S., Handbook of Inductively Coupled Plasma Mass Spectrometry, Blackie, Glasgow, 1992. 3 Houk, R. S., Anal.Chem., 1986, 58, 97A. 4 Boomer, D. W., and Powell, M. J., Anal. Chem., 1987, 59, 2810. 5 Olivares, J. A., and Houk, R. S., Anal. Chem., 1986, 58, 20. 6 Vaughan, M. A., and Horlick, G., Appl. Spectrosc., 1986, 40, 434. 7 Tan, S. H., and Horlock, G., Appl. Spectrosc., 1986, 40, 445. 8 McLaren, J. W., Beauchemin, D., and Berman, S. S., Anal. Chem., 1987, 59, 610. 9 Vaughan, M. A., and Horlick, G., Appl. Spectrosc., 1987, 41, 523. 10 Thompson, J. J., and Houk, R. S., Appl. Spectrosc., 1987, 41, 801. 11 Gregoire, D. C., Appl. Spectrosc., 1987, 41, 897. 12 Gray, A. L., and Williams, J. G., J. Anal. At. Spectrom., 1987, 2, 599. 13 Tan, S. H., and Horlick, G., J. Anal. At. Spectrom., 1987, 2, 745. 14 Dean, J. D., Ebdon, L., and Massey, R., J. Anal. At. Spectrom., 1987, 2, 369. 15 Beauchemin, D., McLaren, J. W., Mykytiuk, A. P., and Berman, S. S., Anal. Chem., 1987, 59, 778. 16 Beauchemin, D., McLaren, J. W., and Berman, S. S., Spectrochim. Acta, Part B, 1987, 42, 467. 17 Beauchemin, D., McLaren, J. W., Mykytiuk, A. P., and Berman, S. S., J. Anal. At. Spectrom., 1988, 3, 305. 18 Date, A. R., Cheung, Y. Y., and Stuart, M. E., Spectrochim. Acta, Part B, 1987, 42, 3. 19 Lichte, F. A., Meier, A. L., and Crock, J. G., Anal. Chem., 1987, 59, 1150. 20 Heithmar, E. M., Hinners, T. A., Rowan, J. T., and Riviello, J. M., Anal. Chem., 1990, 62, 857. 21 Porta, V., Sarzanini, C., and Abolino, O., Anal. Chim. Acta, 1992, 258, 237. 22 Vaughan, M. A., and Templeton, D.M., Appl. Spectrosc., 1990, 44, 1685. 23 Beauchemin, D., and Berman, S. S., Anal. Chem., 1989, 61, 1857. 24 Yang, H.-J., Huang, K.-S., Jiang, S.-J., Wu, C.-C., and Chou, C.-H., Anal. Chim. Acta, 1993, 282, 437. 25 Orians, K. J., and Boyle, E. A., Anal. Chim. Acta, 1993, 282, 63. 26 Bloxham, M. J., Hill, S. J., and Worsfold, P. J., J. Anal. At. Spectrom., 1994, 9, 935. 27 Hwang, T.-J., and Jiang, S.-J., J. Anal. At. Spectrom., 1996, 11, 353. 28 Yang, K.-L., Jiang, S.-J., and Hwang, T.-J., J.Anal. At. Spectrom., 1996, 11, 139. 29 Lu, P.-L., Huang, K.-S., and Jiang, S.-J., Anal. Chim. Acta, 1993, 284, 181. 30 Huang, K.-S., and Jiang, S.-J., Fresenius’ J. Anal. Chem., 1993, 347, 238. 31 Schramel, P., Xu, L. Q., Knapp, G., and Michaelis, M., Mikrochim. Acta, 1992, 106, 191. 32 Gregoire, D. C., and Lee, J., J. Anal. At. Spectrom., 1994, 9, 393. 33 Bowins, R. J., and McNutt, R. H., J. Anal. At. Spectrom., 1994, 9, 1233. 34 Liaw, M.-J., and Jiang, S.-J., J.Anal. At. Spectrom., 1996, 11, 555. 35 Ebdon, L., Ford, M. J., Hutton, R. C., and Hill, S. J., Appl. Spectrosc., 1994, 48, 507. 36 Haraldsson, C., Lyven, B., Pollak, M., and Skoog, A., Anal. Chim. Acta, 1993, 284, 327. 37 Ebdon, L., Fisher, A. S., Worsfold, P. J., Crews, H., and Baxter, M., J. Anal. At. Spectrom., 1993, 8, 691. 38 Jiang, S.-J., Palmieri, M. D., Fritz, J. S., and Houk, R. S., Anal. Chim. Acta, 1987, 200, 559. 39 Reed, N. M., Cairns, R. O., Hutton, R.C., and Takaku, Y., J. Anal. At. Spectrom., 1994, 9, 881. 40 Warren, A. R., Allen, L. A., Pang, H.-M., Houk, R. S., and Janghorbani, M., Appl. Spectrosc., 1994, 48, 1360. Paper 6/06577E Received September 24, 1996 Accepted December 3, 1996 Analyst, March 1997, Vol. 122 237 Determination of Trace Amounts of Zinc in Water Samples by Flow Injection Isotope Dilution Inductively Coupled Plasma Mass Spectrometry Tarn-Jiun Hwang and Shiuh-Jen Jiang* Department of Chemistry, National Sun Yat-Sen University, Kaohsiung, 80424, Taiwan Flow injection isotope dilution ICP-MS was applied to the determination of zinc in several water samples.A matrix separation and preconcentration system was used for the separation of Mg, Ca, S and Cl matrix and preconcentration of trace amounts of zinc in high salt content samples. A complete preconcentration cycle was accomplished in 4 min. The isotope ratio for each injection was calculated from the peak areas of the flow injection peaks. The precision for the isotope ratio determination was better than 1.7%.A detection limit of 0.014 ng ml21 was obtained for zinc with this method. The method was successfully applied to the determination of trace levels of zinc in SLRS-2 riverine water, SLEW-2 estuarine water, NASS-3 open ocean sea-water and CASS-3 nearshore sea-water reference samples. Keywords: Flow injection; isotope dilution; inductively coupled plasma mass spectrometry; zinc; matrix separation; water samples ICP-MS is a powerful technique for trace multielement and isotopic analysis.1–3 This technique combines the characteristics of the ICP for atomizing and ionizing injected material with the sensitivity of MS.It has been satisfactorily applied to the analysis of diverse samples. However, highly saline solutions can cause both spectroscopic and non-spectroscopic interferences. 4–13 Spectral overlaps are seen from polyatomic ions derived from matrix elements such as Na, Ca and Cl.Changes in analyte count rates are observed with high levels of salts or heavy matrix ions. Orifice plugging is also a problem for samples of high solid content. Several calibration methods have been adapted to ICP-MS to help deal with matrix effects.10,14–19 The matrix interference problems can also be dealt with by separating the analytes from the matrix. Several matrix–analyte separation and preconcentration techniques have been adapted to ICP-MS analysis.20–30 We describe here the implementation of an on-line matrix–analyte separation and preconcentration technique with ICP-MS to determine Zn, using a miniature column packed with SO3– quinolin-8-ol carboxymethyl-cellulose.28–31 Isotope dilution (ID) techniques have been applied in several previous ICP-MS applications.14,17,27,29,32–34 Isotope dilution is well recognized as a definitive analytical technique for the determination of the trace elements.Since another isotope of the same element represents the ideal internal standard for that element, ID results are expected to be highly accurate even when the sample contains high concentrations of concomitant elements and/or there is sample loss during the preparation or pretreatment process.In this work, flow injection (FI)–ID–ICP-MS was used to determine the concentrations of zinc in water samples. Matrix separation and preconcentration were used for separation of interfering ions.The influence of instrumental operating conditions, preconcentration system conditions and non-spectroscopic and spectroscopic interferences due to the matrix on the precision and accuracy of isotope ratio determination was also investigated. The method was applied to the determination of zinc in SLRS-2 riverine water, SLEW-2 estuarine water, NASS-3 open ocean sea-water and CASS-3 nearshore seawater reference samples. Experimental ICP-MS Device and Conditions An ELAN 5000 ICP-MS instrument (Perkin-Elmer SCIEX, Thornhill, ON, Canada) was used.It was equipped with an ultrasonic nebulizer (U-5000 AT+ Cetac, Omaha, NE, USA). The gas flow rates were controlled by a four channel mass flow controller. The ICP and ultrasonic nebulizer operating conditions were selected to maximize the sensitivity of zinc using a simple flow injection system. A 100 ml volume of 10 ng ml21 Zn solution was injected into a 0.5 mol l21 HNO3 carrier solution and analyzed by ICP-MS.The operating conditions used throughout this work are summarized in Table 1. Data acquisition parameters used for isotope ratio measurement are listed in Table 1. The isotopes monitored were 66Zn and 67Zn. Version 2.2 of the ELAN 5000 software was used. Under the combinations of dwell time and sweeps/reading, a data point could be obtained in < 1 s for each isotope. The isotope ratio was calculated from the peak areas of the flow injection peaks. The mean isotope ratio and RSD were calculated from isotope ratios measured from several repeated injections. Table 1 Equipment and operating conditions* ICP-MS PE SCIEX ELAN 5000 Plasma conditions— Outer gas flow rate 16 l min21 Auxiliary gas flow rate 0.9 l min21 Nebulizer gas flow rate 1.1 l min21 Rf power 1.1 kW Mass spectrometer settings*— Photon stop (S2) 210.05 V Bessel box barrel +15.34 V Einzel lenses 1 and 3 (E1) 20.04 V Bessel box end lens (P) 279.10 V Resolution Normal Dwell time 40 ms Sweeps/reading 5 Readings/replicate 135 Number of replicates 1 Points/spectral peak 1 Isotopes monitored 66Zn and 67Zn Ultrasonic nebulizer— Desolvation tube temperature 135 °C Condenser temperature 0 °C * Definitions and descriptions of these terms are given in user manuals.Analyst, March 1997, Vol. 122 (233–237) 233Flow Injection In order to investigate the effect of the sample matrix on the precision and accuracy of isotope ratio determinations, a simple FI system was built.The system was composed of a Rheodyne (Cotati, CA, USA) Model 5041 rotary sample injection valve and a Gilson (Middleton, WI, USA) Minipuls-3 peristaltic pump. A 100 ml portion of the test mixture solution was injected into a 0.5 mol l21 nitric acid carrier solution and analyzed by ICP-MS. Preconcentration System A laboratory-made sample pretreatment system was used to separate trace amounts of zinc from matrix elements. A schematic diagram of the matrix separation and preconcentration system is shown in Fig. 1. This system was composed of three Gilson Minipuls-3 peristaltic pumps and two Rheodyne Model 5041 PTFE six-port four-way rotary valves whose switching was actuated by Rheodyne Model 5701 pneumatic actuators and Rheodyne Model 7163 solenoid valve kits. The automation of the flow injection system was controlled by a personal computer through a laboratory-made I/O card and switching circuit. The operating procedure for the preconcentration system is described below.The first rotary valve was used as a selection valve between buffer and sample. At the start of a preconcentration cycle, sample was pumped to wash the system while buffer solution was pumped to condition the column. After a suitable period, valve 1 was switched to load the sample and pump 2 was stopped. After 10.0 ml of sample had been loaded, pump 1 was stopped and valve 1 was rotated back to buffer solution to wash the column. Then valve 2 was switched to let 0.5 mol l21 HNO3 to pass through the column to elute the retained metal ion into the ultrasonic nebulizer.After elution, both valves were switched back and pump 2 was turned on to let the buffer solution condition the column for the next analysis. An Omnifit (Cambridge, UK) high performance column (50 mm 3 10 mm id) was used. It was equipped with a moveable end piece to allow for adjustment of the amount of ion-exchange material. During this study, the column was packed with about 150 mg of SO3–quinolin-8-ol carboxymethylcellulose (Knapp Logistic Automation, Graz, Austria).At a suitable pH, commonly encountered matrix components such as alkali and alkaline earth elements are not strongly retained on the ionexchange materials and are separated from the elements of interest. The operating parameters of the preconcentration system used in this experiment are shown in Table 2. The preconcentration system was connected to the nebulizer with PTFE tubing (20 cm 3 0.8 mm id).Reagents Mg(NO3)2, Ca(NO3)2, (NH4)2SO4 and NaCl were obtained from Fisher (Fair Lawn, NJ, USA). A zinc stock standard solution (1.0 3 106 ng ml21) was obtained from Fisher. Enriched isotope (67Zn) was purchased from Cambridge Isotope Laboratories (Andover, MA, USA). This stable isotope was received as an oxide. A stock standard solution of zinc of approximately 100 mg l21 was prepared by dissolution of an accurately weighed amount of the material in nitric acid and dilution to volume. The concentration of the spiked solution was verified by reverse spike isotope ID–ICP.All buffers and eluents were prepared using high purity deionized water (Milli-Q reagent water system, Millipore, Bedford, MA, USA). The ammonium acetate buffer solutions were adjusted with ammonia solution and acetic acid in the pH range 2–8. All the buffers were purified by triple filtration on an SO3–quinolin-8-ol carboxymethyl-cellulose column.After the uptake step, the absorbed analyte was eluted with 0.5 mol l21 HNO3 (Fisher, trace metal grade). Evaluation of Retention The effect of the pH of the loading solution on the chelation of zinc was studied with the preconcentration system in an off-line mode. Solutions of 100 ng ml21 zinc were buffered to suitable pH values, then 10 ml of these solutions were ‘loaded’ on to the SO3–quinolin-8-ol carboxymethylcellulose column at a flow rate of 2.3 ml min21. Finally, any retained metal was eluted with 5 ml of 1.0 mol l21 HNO3 into a 10 ml calibrated flask and the collected eluate was diluted to volume with distilled, deionized water.The amounts of metals were then determined by ICPMS. Recoveries were calculated against the theoretical concentrations. The tolerance of the matrix separation system for co-existing matrix ions was examined by adding relatively high concentrations of Mg2+, Ca2+, SO4 22 and Cl2 to 1 ng ml21 of zinc metal solutions (for simple FI analysis 5 ng ml21 Zn solution was used).Each solution was injected into FI–ICP-MS and matrix separation–ICP-MS systems successively and the isotope ratio of Zn was determined for comparison. Sample Preparation SLRS-2 riverine water reference material for trace metals, SLEW-2 estuarine water reference material for trace metals, Fig. 1 Schematic diagram of matrix–analyte separation and preconcentration system. This figure shows the flow path of the system in column wash step.Table 2 Preconcentration system and operating conditions Column SO3–quinolin-8-ol carboxymethylcellulose Buffer solution pH 4; 0.1 mol l21 ammonium acetate Eluent 0.5 mol l21 HNO3 Pump 1 (sample) flow rate 10.0 ml min21 Pump 2 (buffer) flow rate 4.5 ml min21 Pump 3 (eluent) flow rate 3.0 ml min21 Preconcentration method— Pump in Valve Step operation position Duration/s System wash 1, 2, 3 1 load 20 2 load Sample loading 1, 3 1 inject 60 2 load Column wash 2, 3 1 load 40 2 load Elution 3 1 load 60 2 inject Column wash 2, 3 1 load 60 2 load 234 Analyst, March 1997, Vol. 122NASS-3 open ocean sea-water reference material for trace metals and CASS-3 nearshore sea-water reference material for trace metals (National Research Council of Canada, Ottawa, Canada) were obtained to demonstrate the applicability of the method to real samples. A 25 ml aliquot of these acidified reference materials was adjusted to pH 4 with 5 ml of 1.0 mol l21 purified ammonium acetate buffer solution. After a suitable amount of enriched isotope has been added, this solution was diluted to 50 ml with distilled, deionized water.The analyte concentration in the sample was calculated using the equation given in a previous paper.27 Owing to a mass bias effect, the intensities obtained during isotope ratio determinations were used to calculate the isotopic abundance of zinc. Since the mass bias effect could be factored out during ID calculation, in this study the measured isotope ratio did not correct the mass bias effect.Results and Discussion Effect of pH on Retention The amount of zinc chelated with SO3–quinolin-8-ol carboxymethylcellulose is related to the pH of the solution, and this effect was examined for Zn in the pH range 2–8. The results are shown in Fig. 2. Zinc was chelated quantitatively in the pH range 4–8. For the analysis of the sea-water sample, a solution with pH between 3.5 to 4 should be chosen; at this pH, Zn was complexed by SO3–quinolin-8-ol carboxymethylcellulose and was retained on the column relatively completely.Meanwhile, as shown in Fig. 2, the interfering ions Mg2+ and Ca2+ chelated with the resin only when the pH value was > 4. Most of the Mg and Ca wash through the column at pH < 4, which can be used to reduce the interferences of Mg and Ca in the ICP-MS analysis. In subsequent experiments, all the loading sample solutions were adjusted to pH 4. Relationship Between Sample Loading Flow Rate and Recovery The sample uptake rate did not affect the recovery significantly in the range 3–13 ml min21.For the purposes of trace Zn concentration, a much faster sample flow rate should be used to save experimental time, and a 10 ml min21 sample loading flow rate was used in the following experiments. Selection of Eluent and Flow Rate Owing to the formation of extra molecular ion interferences when HCl and HClO4 were used as eluents, in order to avoid any possible interferences and reduce the background, these acids were not used in this study.The elution of Zn2+ from the column was examined with HNO3 at various concentrations. The recovery of Zn was complete if the concentration of HNO3 was !0.5 mol l21, since it was difficult to nebulize a concentrated acid solution owing to the viscosity and density and, when a high concentration of acid was nebulized, matrix effects were observed. To avoid matrix effects and signal suppression by high acid levels, the lowest possible acid concentration for elution which yielded 100% recovery was used.In subsequent experiments, 0.5 mol l21 HNO3 was used as the eluent to obtain best recovery and elution peaks for Zn. The recovery of Zn at pH 4 for replicate measurements (n = 7), using 10 ml of 1 ng ml21 standard solution, was 100.2 ± 0.4%. Fig. 3 shows the effect of the eluent flow rate on the ion signal. An increase in the eluent flow rate decreased the width of the elution peak and saved analysis time.However, an increased eluent flow rate also decreased the nebulization efficiency, causing the ion signal to decrease significantly. The effect of eluent flow rate on the peak shape of the elution curve was similar to that in previous work.28 Hence an eluent flow rate of 3.0 ml min21 was selected to increase the ion signals in the following experiments, although a slightly greater duration of analysis was required.Under the selected preconcentration conditions, a complete preconcentration cycle was accomplished in 4 min. A typical elution curve for a solution containing 1 ng ml21 of Zn is shown in Fig. 4. Under the operating conditions of the preconcentration system, the peak half-width for the flow injection peak was about 12 s, which is equivalent to 0.6 ml; this means that with an injection volume of 10 ml a preconcentration factor of about 17 could be obtained for Zn.Detection Limit and Reproducibility The calibration plot based on peak area for 66Zn was linear with a correlation coefficient better than 0.9995 at levels near the detection limit up to at least 50 ng ml21. The detection limit calculated from the calibration plot was 0.014 ng ml21. It was based on the conventional definition of the concentration of the analyte yielding a signal equivalent to three times the standard deviation of the blank. The detection limit obtained by this method is comparable or superior to previous results obtained with ICP-MS detection.35–38 The reagent blank was 0.13 ng ml21 for Zn.A possible explanation is that the system is contaminated with Zn. Zinc is a common contaminant in the Fig. 2 Dependence of recovery on pH of the loading solution. The loaded sample volume was 10 ml at a flow rate of 2.3 ml min21. The flow rate of the eluent (1.0 mol l21 HNO3) was 2.0 ml min21. Fig. 3 Dependence of ion signal on eluent flow rate.The eluent was 0.5 mol l21 HNO3 solution. The loaded sample volume was 10 ml at a flow rate of 10 ml min21. All data are relative to the first point. Analyst, March 1997, Vol. 122 235reagent and solutions; a better detection limit is to be expected with much more purified reagents. The repeatability was determined by seven injections of a 1 ng ml21 Zn solution. The repeatability of the peak areas of these seven injections was 1.4%. Non-spectroscopic and Spectroscopic Interferences As described earlier, the retentions of matrix elements on the ion-exchange material are usually different from that of zinc, hence zinc can be separated from the matrix and collected on the cellulose. The effectiveness of matrix elimination with this separation system is demonstrated by the following experiments.Solutions containing zinc and matrix elements were analyzed by matrix separation–ICP-MS and the results were compared with those obtained using a simple FI–ICP-MS system.Table 3 shows the results of adding various concentrations of S, Cl, Mg and Ca ions to a test solution containing 5 ng ml21 Zn. As shown, the 66Zn/67Zn ratio was increased when 500 mg ml21 S was added, which indicated an interference at m/z 66 from 34S16O16O+ and/or 32S34S+.39 Moreover, the 66Zn/67Zn ratio increased significantly when 1000 mg ml21 of Mg was added, which indicated an interference at m/z 66 from 40Ar26Mg+, although the 66Zn/67Zn ratio did not change significantly when 250 mg ml21 of Ca and 15000 mg ml21 of Cl were added.The precision of the isotope ratio determination was poor when simple FI sample introduction was used. This could be due to the suppression of the ion signal by the high concentration matrix elements. Finally, a solution containing 500 mg ml21 S, 250 mg ml21 Ca, 1000 mg ml21 Mg and 15 000 mg ml21 Cl was spiked to 5 ng ml21 Zn standard solution. The concentrations of the matrix ions were about half of the concentrations of these ions in the sea-water matrix.As shown in Table 3, the 66Zn/67Zn ratio increased to 16.34 when this matrix solution was added. Table 3 also shows that a stable 66Zn/67Zn ratio could be obtained when S, Cl, Mg and Ca were separated from the analyte with the matrix separation system. Hence, for the analysis of highly saline samples, the interfering ions must be separated before Zn is determined. In general, in order to achieve accurate, reliable and sensitive results, preconcentration and separation are needed when the concentrations of analyte elements in the original material or the prepared solution are too small to be determined directly by MS or when matrix elements interfere with the determination.The precision (RSD 0.7–1.7%), shown in Table 3, was slightly worse than that obtained for isotope ratio measurements during continuous introduction of aqueous solutions. Moreover, the precision for the measured isotope ratio was slightly worse than that calculated by the counting statistics (RSD 0.5–0.6%).40 A poorer precision in isotope ratio measurements is to be expected when transient sample introduction techniques are used because the ions of interest are observed for a shorter time and averaging of nebulizer fluctuations, etc., is less efficient.The slight but reproducible differences between the found and expected values are similar to those commonly seen in isotope ratio measurements with ICP-MS and were probably caused by some mass discrimination in ion extraction, focusing, mass analysis and detection.38 Determination of Zinc in Water Samples In order to demonstrate that the system is effective for practical analysis, four water reference samples (NRCC SLRS-2, SLEW- 2, NASS-3 and CASS-3) were analyzed.The amount of Zn in each sample was determined by ID–ICP-MS after on-line sample pretreatment using the matrix separation and preconcentration method.In order to decrease the retention of alkaline earth ions when analyzing sea-water samples, a low pH buffer solution (pH 4) and careful washing of the column with buffer solution were used. The results of the analysis of standard reference materials are given in Table 4. Good agreement Fig. 4 Typical elution curve for 1 ng ml21 Zn solution. The loaded sample volume was 10 ml at a flow rate of 10 ml min21. The eluent (0.5 mol l21 HNO3) flow rate was 3.0 ml min21. Time zero represents the start of elution.Table 3 Isotope ratios of Zn in various matrices with different methods of analysis* 66Zn/67Zn† Matrix Sample composition FI‡ separation§ Zn standard 6.41 ± 0.07 6.47 ± 0.05 Zn standard + S 500 mg ml21 7.59 ± 0.53 6.35 ± 0.11 Zn standard + Ca 250 mg ml21 6.60 ± 1.07 6.45 ± 0.08 Zn standard + Mg 1000 mg ml21 11.81 ± 0.71 6.59 ± 0.09 Zn standard + Cl 15 000 mg ml21 6.64 ± 1.56 6.53 ± 0.08 Zn standard + S 500 mg ml21 + Ca 250 mg ml21 + Mg 1000 mg ml21 + Cl 15 000 mg ml21 16.34 ± 8.55 6.49 ± 0.06 * The expected value of 66Zn/67Zn is 6.77.† Mean of three measurements ± standard deviation. ‡ FI carrier (0.5 mol l21 HNO3) flow rate = 3.0 ml min21; 100 ml sample loop. Concentration of Zn standard solution used in FI, 5 ng ml21. § The operating conditions for the matrix– analyte separation and preconcentration system are given in Table 2. Concentration of Zn standard solution used in matrix separation, 1 ng ml21. Table 4 Concentrations of Zn in reference materials as measured by ID–online preconcentration–ICP-MS (n = 3) Concentration*/ng ml21 Sample Found Certified SLRS-2 riverine water 3.35 ± 0.06 3.33 ± 0.15 SLEW-2 estuarine water 1.12 ± 0.05 1.10 ± 0.14 CASS-3 nearshore sea-water 1.30 ± 0.17 1.24 ± 0.25 NASS-3 open ocean sea-water 0.208 ± 0.035 0.178 ± 0.025 * Mean ± 95% confidence limit. 236 Analyst, March 1997, Vol. 122between the experimental results and certified values show the applicability of this system to real sample analysis.Conclusion The use of ICP-MS and an SO3–quinolin-8-ol carboxymethylcellulose miniature column flow system provide a simple, rapid and accurate technique to determine routinely trace amounts of Zn in water samples. This preconcentration method yielded an enrichment factor of 17 and sample consumption was thus reduced. This research was supported by a grant from the National Science Council of the Republic of China under contract NSC 85-2621-M- 110-009.References 1 Houk, R. S., Fassel, V. A., Flesch, G. D., Svec, H. J., Gray, A. L., and Taylor, C. E., Anal. Chem., 1980, 52, 2283. 2 Jarvis, K. E., Gray, A. L., and Houk, R. S., Handbook of Inductively Coupled Plasma Mass Spectrometry, Blackie, Glasgow, 1992. 3 Houk, R. S., Anal. Chem., 1986, 58, 97A. 4 Boomer, D. W., and Powell, M. J., Anal. Chem., 1987, 59, 2810. 5 Olivares, J. A., and Houk, R. S., Anal. Chem., 1986, 58, 20. 6 Vaughan, M. A., and Horlick, G., Appl.Spectrosc., 1986, 40, 434. 7 Tan, S. H., and Horlock, G., Appl. Spectrosc., 1986, 40, 445. 8 McLaren, J. W., Beauchemin, D., and Berman, S. S., Anal. Chem., 1987, 59, 610. 9 Vaughan, M. A., and Horlick, G., Appl. Spectrosc., 1987, 41, 523. 10 Thompson, J. J., and Houk, R. S., Appl. Spectrosc., 1987, 41, 801. 11 Gregoire, D. C., Appl. Spectrosc., 1987, 41, 897. 12 Gray, A. L., and Williams, J. G., J. Anal. At. Spectrom., 1987, 2, 599. 13 Tan, S. H., and Horlick, G., J. Anal. At. Spectrom., 1987, 2, 745. 14 Dean, J. D., Ebdon, L., and Massey, R., J. Anal. At. Spectrom., 1987, 2, 369. 15 Beauchemin, D., McLaren, J. W., Mykytiuk, A. P., and Berman, S. S., Anal. Chem., 1987, 59, 778. 16 Beauchemin, D., McLaren, J. W., and Berman, S. S., Spectrochim. Acta, Part B, 1987, 42, 467. 17 Beauchemin, D., McLaren, J. W., Mykytiuk, A. P., and Berman, S. S., J. Anal. At. Spectrom., 1988, 3, 305. 18 Date, A. R., Cheung, Y. Y., and Stuart, M. E., Spectrochim. Acta, Part B, 1987, 42, 3. 19 Lichte, F. A., Meier, A. L., and Crock, J. G., Anal. Chem., 1987, 59, 1150. 20 Heithmar, E. M., Hinners, T. A., Rowan, J. T., and Riviello, J. M., Anal. Chem., 1990, 62, 857. 21 Porta, V., Sarzanini, C., and Abolino, O., Anal. Chim. Acta, 1992, 258, 237. 22 Vaughan, M. A., and Templeton, D. M., Appl. Spectrosc., 1990, 44, 1685. 23 Beauchemin, D., and Berman, S. S., Anal. Chem., 1989, 61, 1857. 24 Yang, H.-J., Huang, K.-S., Jiang, S.-J., Wu, C.-C., and Chou, C.-H., Anal. Chim. Acta, 1993, 282, 437. 25 Orians, K. J., and Boyle, E. A., Anal. Chim. Acta, 1993, 282, 63. 26 Bloxham, M. J., Hill, S. J., and Worsfold, P. J., J. Anal. At. Spectrom., 1994, 9, 935. 27 Hwang, T.-J., and Jiang, S.-J., J. Anal. At. Spectrom., 1996, 11, 353. 28 Yang, K.-L., Jiang, S.-J., and Hwang, T.-J., J. Anal. At. Spectrom., 1996, 11, 139. 29 Lu, P.-L., Huang, K.-S., and Jiang, S.-J., Anal. Chim. Acta, 1993, 284, 181. 30 Huang, K.-S., and Jiang, S.-J., Fresenius’ J. Anal. Chem., 1993, 347, 238. 31 Schramel, P., Xu, L. Q., Knapp, G., and Michaelis, M., Mikrochim. Acta, 1992, 106, 191. 32 Gregoire, D. C., and Lee, J., J. Anal. At. Spectrom., 1994, 9, 393. 33 Bowins, R. J., and McNutt, R. H., J. Anal. At. Spectrom., 1994, 9, 1233. 34 Liaw, M.-J., and Jiang, S.-J., J. Anal. At. Spectrom., 1996, 11, 555. 35 Ebdon, L., Ford, M. J., Hutton, R. C., and Hill, S. J., Appl. Spectrosc., 1994, 48, 507. 36 Haraldsson, C., Lyven, B., Pollak, M., and Skoog, A., Anal. Chim. Acta, 1993, 284, 327. 37 Ebdon, L., Fisher, A. S., Worsfold, P. J., Crews, H., and Baxter, M., J. Anal. At. Spectrom., 1993, 8, 691. 38 Jiang, S.-J., Palmieri, M. D., Fritz, J. S., and Houk, R. S., Anal. Chim. Acta, 1987, 200, 559. 39 Reed, N. M., Cairns, R. O., Hutton, R. C., and Takaku, Y., J. Anal. At. Spectrom., 1994, 9, 881. 40 Warren, A. R., Allen, L. A., Pang, H.-M., Houk, R. S., and Janghorbani, M., Appl. Spectrosc., 1994, 48, 1360. Paper 6/06577E Received September 24, 1996 Accepted December 3, 1996 Analyst, March 1997, Vol. 122 237

ISSN:0003-2654    

DOI:10.1039/a606577e

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Determination of Gold in Geological Samples and Anode Slimes by Atomic Absorption Spectrometry After Preconcentration Wth Amberlite XAD-16 Resin A. Tunçeli and A. R. T�urker Gazi � Universitesi, Fen-Edebiyat Fak�ultesi, Teknikokullar, TR-06500 Ankara, Turkey A method was developed for the determination of trace levels of gold in geological materials and anode slimes involving the adsorption of gold on Amberlite XAD-16 resin followed by FAAS determination after elution. The adsorption behaviour of the chloro complex of gold from hydrochloric acid solution was also investigated and the recovery of gold was found to be 95.48 ± 0.04% at the 95% confidence level.The degree of preconcentration ranges from 10- to 75-fold for different sample volumes. The detection limit of gold was 0.046 mg l21. The adsorption isotherm and adsorption capacity of XAD-16 resin were investigated for gold. The adsorption isotherm was found to be of the Langmuir type and the adsorption capacity of the resin was 0.55 mmol g21 (108 mg g21).The influence of geological matrix elements on the adsorption of gold was also evaluated. The gold contents of anode slime and a reference standard ore were determined by applying the proposed method. Keywords: Resin preconcentration; separation; gold; geological samples; anodic slime; flame atomic absorption spectrometry Because of the low abundance of gold and high salt concentrations in geological samples and anode slimes, the direct determination of gold by FAAS in solutions obtained by wet decomposition of samples is often difficult or even impossible.For this reason, separation of gold from the sample matrix and increases in analyte concentrations are useful steps in the analysis. In trace element analysis, preconcentration and separation methods also enhance the sensitivity and precision of the determination.1 For sub-mg g21 levels in geological samples, ETAAS, ICP-AES and neutron activation analysis have been the dominant and preferred methods either with or without separation and preconcentration steps.2–6 However, FAAS has been widely used after a preconcentration step for trace analysis because it is a relatively simple method and requires less expensive instrumentation.7,8 The analysis of geological samples and anode slimes for gold is very difficult owing to the high concentrations of widely varying matrix elements such as Fe and Cu.Therefore, before the determination, the gold must be separated from the sample matrix and enriched. For this purpose, several separation and preconcentration techniques for gold, such as solvent extraction, 9–12 coprecipitation4 and ion-exchange or adsorption,13–15 have been developed.Among these techniques, sorption is one of the most widely used. For this purpose, XAD resins16–18 and various polymers19,20 have been used. Because of the purity and good adsorption properties, the commercially available Amberlite XAD copolymers (XAD-2, -4, -7 and -8) have been widely used as adsorbents.16–18 However, Amberlite XAD-16 has not been widely used for the preconcentration of trace elements.21 This paper describes the determination of gold in geological samples and anode slimes after preconcentration on a column packed with Amberlite XAD-16 resin, which has the largest surface area in the XAD series.The optimum conditions for adsorption and determination procedures are also discussed.The influence of interfering elements such as iron on the determination of gold was evaluated. Experimental Apparatus A Philips (Eindhoven, The Netherlands) PU 9285 flame atomic absorption spectrometer equipped with an air–acetylene burner and gold hollow cathode lamp operated at 11.2 mA was used for the determination of gold. The operating conditions were wavelength 242.8 nm, bandpass 0.5 nm and fuel flow rate 0.9 l min21. Deuterium lamp background correction was used.All pH measurements were made with a Consort (Turnhout, Belgium) digital pH meter and a combined glass electrode. Column Preparation The column diameter and amount of resin were chosen taking into account the information obtained from our previous studies.21 For subsequent experiments, a glass column (30 mm 3 10 mm id) with a glass-wool plug over its stopcock was used. A 0.2 g amount of resin was rested on the glass-wool and compressed in order to avoid channel formation.Another plug of glass-wool was placed on top, so that the resin was not disturbed during sample passage. Before use, 1 mol l21 hydrochloric acid was passed through the column for cleaning and preconditioning. Reagents and Standard Solutions Amberlite XAD-16 resin (Rohm and Haas, Philadelphia, PA, USA; surface area 800 m2 g21; wet mesh size 20–60) was used after washing with methanol, 1 mol l21 hydrochloric acid and triply distilled water and dried at 60 0C. Analytical-reagent grade chemicals were used without further purification.Triply distilled water was used in all experiments. A stock standard solution of gold (1000 mg l21) was prepared by dissolving the appropriate amount of tetrachloroauric( iii)acid trihydrate (HAuCl4·3H2O) in water. Working standard solutions were prepared by dilution. General Procedure for Optimization of the Preconcentration Step Volumes of 100 ml of spiked solutions containing 25 mg of gold (0.25 mg l21) were used.The pH of the solution was adjusted to the desired value with hydrochloric acid and potassium hydroxide solutions. The resulting solution was passed through the column at an optimum flow rate of 5 ml min21. After the sample solution had been passed through the column, the retained metal ions were eluted from the resin by using 5 ml of 0.3 mol l21 potassium iodide solution in methanol and 5 ml of pure methanol. Analyst, March 1997, Vol. 122 (239–242) 239The eluate in a 10 ml calibrated flask was aspirated into an air–acetylene flame and the gold was determined by FAAS.The resin was used repeatedly after washing with 1 mol l21 hydrochloric acid and water. Using the procedure described above, the recovery of gold was calculated from the ratio of the concentration found by FAAS to that calculated theoretically. Sample Preparation Samples were dissolved by a commonly used method.9,13,15 An anode slime sample and a geological standard reference material (Gold Ore MA-3 certified by the Canada Center For Mineral and Energy Technology) were dried at 110 0C for 2 h, then approximately 1 g of gold ore and 0.1 g of anode slime were weighed accurately into a beaker separately.In order to decompose the sample, about 20 ml of aqua regia were added to the beaker and the solution was evaporated to dryness. This decomposition procedure was repeated and 10 ml of water were added. The suspension formed was filtered through a fine filterpaper.The insoluble part retained on the filter-paper was washed with 1 mol l21 hydrochloric acid and water. The filtrate and washings were collected in a 100 ml calibrated flask. Potassium hydroxide was used to adjust the pH of the solution to 2. Gold was preconcentrated from this solution with the column procedure described above and determined by FAAS. Results and Discussion Effect of Hydronium Ion Concentration The retention of gold on the resin was studied as a function of pH and/or the concentration of hydrochloric acid.The results are shown in Fig.1. As can be seen, a quantitative recovery (!95%) of gold was obtained at pH 2, and this pH was chosen as the optimum for subsequent experiments. Effect of Elution Solutions Elution studies were performed with halide solutions (potassium iodide and potassium bromide) in water and methanol and acidic solutions (hydriodic and hydrochloric acid) at various concentrations. The results are shown in Table 1.The best elution ( > 95% recovery) was obtained by using 5 ml of 0.3 mol l21 potassium iodide solution in methanol. Gold was quantitatively eluted with this eluent. The use of potassium iodide in methanol as the eluent was very convenient because the stability constant for the iodide complex of gold was high. In addition, methanol improved the nebulization efficiency by decreasing the viscosity of the sampleion. A 5 ml volume of 0.3 mol l21 KI solution and 5 ml of pure methanol were used as eluents in all subsequent experiments.Effect of Sample Volume and Analyte Concentration In order to deal with real samples containing very low concentrations of gold, the maximum applicable sample volume must be determined. Therefore, the effect of changes in the volume of sample solution passed through the column on the sorption of Au was investigated. Volumes of 100, 250, 500, 750 and 1000 ml of sample solutions containing 0.25, 0.10, 0.05, 0.033 and 0.025 mg l21 of Au were passed through the column.The recovery of gold was approximately quantitative (about 95%) up to a sample volume of 750 ml. At higher sample volumes, the gold recovery decreased gradually with increasing volume of sample. In this work, the elution volume was 10 ml and the highest concentration factor was 75 for a 750 ml sample volume. These results also indicate the effect of analyte concentration on the retention of Au. Normally FAAS is unable to determine gold at levels below 0.1 mg l21,22 but the proposed technique allowed gold to be determined in the range 0.033–0.25 mg l21.Effect of Sample and Eluent Flow Rate on Sorption Since the retention of elements on an adsorbent depends on the flow rate of the metal solution, the effect of the flow rate was examined under the optimum conditions (pH, eluent type, etc.). The flow rate varied from 1 to 5 ml min21. It was found that the retention of gold was not affected by flow rate up to 5 ml min21.Since either suction or pressure is needed for flow rates higher than 5 ml min21, no test was carried out for such flow rates. The effect of eluent flow rate on the recovery was also examined. Three flow rates (1, 2.5 and 5 ml min21) were studied and recoveries of 95, 88 and 60%, respectively, at 5 ml min21 were obtained. Since the recovery increases with decreasing flow rate of the eluent, 1 ml min21 was adopted in all experiments.Precision of Method The precision of the method was investigated at the optimum pH of 2 for the spiked solution (0.25 mg l21 Au) described above. For this purpose, the general procedure was performed successively and gold was determined in the solution by FAAS. The mean recovery for eight determinations at the 95% confidence level was 95.48 ± 0.04%. As can be seen, the recovery is quantitative and it is sufficient for preconcentration purposes. The precision of the method is very good, the RSD being about 2.1%.Adsorption Isotherm and Adsorption Mechanisms For the determination of the adsorption behaviour of the resin, the amount of adsorbed gold was studied as a function of gold Fig. 1 Effect of hydronium ion concentration on the recovery of gold. Table 1 Effect of elution solutions on the recovery of gold Solution Recovery (%) 5 ml of 0.1 mol l21 KI in methanol and 5 ml of pure methanol 91.5 5 ml of 0.1 mol l21 KBr in methanol and 5 ml of pure methanol 9.0 5 ml of 0.3 mol l21 KI in methanol and 5 ml of pure methanol 95.0 5 ml of 1 mol l21 KI in methanol and 5 ml of pure methanol 95.0 5 ml of 1 mol l21 HCl 0.5 5 ml of 0.1 mol l21 HI in water 1.0 5 ml of 0.1 mol l21 HI in methanol and 5 ml of pure methanol 69.3 5 ml of 0.3 mol l21 KI in methanol 59.6 5 ml of pure methanol 2.6 240 Analyst, March 1997, Vol. 122H+AuCl4 – concentration (Fig. 2). The adsorption isotherm of gold conforms to the Langmuir equation.23 Graphical treatment of the data yielded Langmuir parameters of capacity of the resin for gold 0.55 mmol g21 (108 mg g21) and equilibrium binding constant, K, about 6 3 103 l mol21.The Langmuir treatment demonstrated that the surface is nearly fully covered with a monomolecular layer of gold at high concentrations. Therefore, at high concentrations adsorption does not increase linearly. In addition to the adsorption isotherm of the resin, another important parameter, the adsorption mechanism was studied.Amberlite XAD-16 is a styrene–divinylbenzene copolymer. The metal is probably adsorbed as a chloro complex such as HAuCl4. The adsorption might result in the formation of a complex compound such as that shown below:24 Effect of Matrix Elements The determination of gold in naturally occurring ores is difficult owing to the high concentrations of other (accompanying) elements such as Na, Fe, Ni and Cu. For this reason, a fixed amount of gold was taken with various combined amounts of matrix elements such as Fe, Cu, Ni, Na, K, Mg, Ca and the recommended procedure was applied.The results are summarized in Table 2. The recovery of gold decreased by in the presence of a large excess of matrix elements, because of the insufficient resin capacity. Especially iron, generally present in the ore matrix, decreased the gold adsorption.25 Gjerde and Fritz25 found that a 1000-fold excess of iron over gold decreased the gold recovery from 98% to 80%.For this reason, the effect of iron on the adsorbance and recovery of gold was specially investigated. The results are given in Table 3. As can be seen, by using the proposed method an approximately 5000-fold excess of iron over gold did not interfere; the recovery of gold was almost constant. In order to explain which element(s) cause the large decreases in recovery shown in Table 2, the effect of each individual matrix element on the recovery of gold was investigated.For this purpose, only the effect of 250 mg l21 of each matrix element on 0.25 mg l21 of gold was studied. The results are given in Table 4. The recovery of gold did not decrease when the matrix elements were studied individually, but when they were used all together the recovery of gold decreased by about 22.5% (Table 2). The reason for this result might be attributed to the insufficient column capacity. The retention of Cu and Ni on XAD-16 resin at pH 2 is less,21 but the retention of Fe is high (about 40%).Hence, it can be said that the capacity of resin has been fully utilized when the matrix elements were used all together at a high concentration (250 mg l21). Application The proposed method was applied to the determination of gold in an anode slime and a geological sample. As the geological Fig. 2 Adsorption isotherm of XAD-16 for gold. Table 5 Determination of gold in real samples Au concentration Recovery LOD‡ Sample* Certified Found† (%) (mg l21) Gold Ore MA-3 7.49 ± 0.19 mg g21 7.23 ± 0.12 mg g21 96 0.046 Anode slime 0.076% m/m§ 0.071 ± 0.003% m/m 93 0.046 * The composition of Gold Ore MA-3 (only gold was certified) was Si 25, Al 7.22, Fe 3.85, Ca 3.31, K 3.05, Na 2.42, C 0.76 and S 0.56% m/m, Ag 1.5 mg g21 and Au 7.49 mg g21.The composition of the anode slime was Cu 18.86, Pb 7.67, Sn 1.36, Ag 2.29, Au 0.076, Fe 0.10 and SiO2 1.69% m/m. † Average of five determinations at the 95% confidence level. ‡ Limit of detection based on 99.7% (3s) confidence level (n = 10). § Determined spectrophotometrically by Makina ve Kimya End�ustrisi Kurumu.Table 2 Effect of combined matrix elements on the recovery of gold Foreign metal (Fe, Cu, Ni, Au Na, K, Ca, Mg) concentration/ Added/ Found/ Recovery mg l21 mg l21 mg l21 (%) 0 0.25 0.24 95.0 25 0.25 0.22 86.4 125 0.25 0.12 46.2 250 0.25 0.06 22.5 Table 3 Effect of iron concentration on the recovery of gold Fe/ Au/ Recovery mg l21 mg l21 (%) 100 0.25 94.0 500 0.25 93.5 1000 0.25 94.6 1250 0.25 93.8 Table 4 Effect of individual matrix elements at a concentration of 250 mg l21 on the recovery of gold Recovery Element of gold (%) None 95.0 Cu 92.0 Ni 94.0 Na 99.5 K 97.6 Mg 96.0 Ca 104.0 Analyst, March 1997, Vol. 122 241sample a Certified Gold Reference Ore and as an anode slime the product from the brass plant of Makina ve Kimya End¡§ustrisi Kurumu were analysed. The results are shown in Table 5. The limit of detection was calculated on the basis of three times the standard deviation of the absorbance using a series of measurements of blank solutions prepared according to the procedure given under Sample Preparation without taking the sample.As can be seen, the preconcentration, separation and determination of trace amounts of gold by the proposed method in a geological material and ano slime in the presence of high concentrations of foreign cations and anions including iron (in MA-3 about a 5000-fold excess) could be performed successfully.The relative error was 3.5%, for gold ore and 6.6% for anode slime. Conclusion The proposed column preconcentration method provides a simple, sensitive and accurate method for the preconcentration of gold in geological materials and anode slimes and is also successful in the presence of iron. Gold can be retained on the column by adjusting the sample solution to pH 2 and then be eluted from the column using potassium iodide. Most of the metals including iron do not influence the gold determination. The precision and recovery were satisfactory.Further work is in progress to evaluate the use of XAD-16. References 1 Mizuike, A., Enrichment Techniques for Trace Analysis, Springer, Berlin, 1983. 2 McHugh, J. B., At. Spectrosc., 1983, 4(2), 66. 3 Kontas, E., At. Spectrosc., 1981, 2(2), 59. 4 Chung, Y. S., and Barnes, R. M., J. Anal. At. Spectrom., 1988, 3, 1079. 5 Elson, C. M., and Chatt, A., Anal.Chim. Acta, 1983, 155, 305. 6 Inductively Coupled Plasma Emission Spectroscopy, ed. Boumans, P. W. J. M., Wiley, New York, 1987, pt. 2, ch. 3, pp. 27¡©47. 7 Xu, S., Sun, L., and Fang, Z., Anal. Chim. Acta, 1991, 245, 7. 8 Elci, L., Anal. Lett., 1993, 26, 1025. 9 Brooks, R. R., Holzbecher, J., Ryan, D. E., and Zhang, H. F., At. Spectrosc., 1981, 2(5), 151. 10 Tsukahara, I., Talanta, 1977, 24, 633. 11 Marczenko, Z., and Kowalski, T., Anal. Chim. Acta., 1984, 156, 193. 12 McBryde, W.A. E., and Yoe, J. H., Anal. Chem., 1948, 20, 1094. 13 Elci, L., I`s©¥ldar, S., and Do¢�gan, M., Anal. Chim. Acta, 1994, 293, 319. 14 Koshima, H., and Onishi, H., Anal. Sci., 1988, 4, 611. 15 Rivoldini, A., and Haile, T., At. Spectrosc., 1989, 10(3), 89. 16 Yang, X. G., and Jackwerth, E., Fresenius¡� Z. Anal. Chem., 1989, 335, 712. 17 Yang, X. G., and Jackwerth, E., Fresenius¡� Z. Anal. Chem., 1989, 335, 483. 18 Yang, X. G., and Jackwerth, E., Fresenius¡� Z. Anal. Chem., 1990, 336, 588. 19 Hoshi, S., Tanaka, Y., Inoue, S., and Matsubara, M., Anal. Sci., 1989, 5, 471. 20 Ar©¥k, N., and T¡§urker, A. R., Fresenius¡� J. Anal. Chem., 1991, 339, 874. 21 T¡§urker, A. R., and Tunceli, A., Fresenius¡� J. Anal. Chem., 1993, 345, 755. 22 Taylor, M. J. C., Barnes, D. E., and Marshall, G. D., Anal. Chim. Acta, 1992, 265, 71. 23 Adamson, A. W., Physical Chemistry of Surfaces, Wiley-Interscience, New York, 4th edn., 1982, p. 521. 24 Koshima, H., Anal. Sci., 1986, 2, 255. 25 Gjerde, D. T., and Fritz, J. S., J. Chromatogr., 1980, 188,. 391. Paper 6/05008E Received July 17, 1996 Accepted November 27,1996 242 Analyst, March 1997, Vol. 122 Determination of Gold in Geological Samples and Anode Slimes by Atomic Absorption Spectrometry After Preconcentration Wth Amberlite XAD-16 Resin A. Tunçeli and A. R. T�urker Gazi � Universitesi, Fen-Edebiyat Fak�ultesi, Teknikokullar, TR-06500 Ankara, Turkey A method was developed for the determination of trace levels of gold in geological materials and anode slimes involving the adsorption of gold on Amberlite XAD-16 resin followed by FAAS determination after elution.The adsorption behaviour of the chloro complex of gold from hydrochloric acid solution was also investigated and the recovery of gold was found to be 95.48 ± 0.04% at the 95% confidence level. The degree of preconcentration ranges from 10- to 75-fold for different sample volumes. The detection limit of gold was 0.046 mg l21.The adsorption isotherm and adsorption capacity of XAD-16 resin were investigated for gold. The adsorption isotherm was found to be of the Langmuir type and the adsorption capacity of the resin was 0.55 mmol g21 (108 mg g21). The influence of geological matrix elements on the adsorption of gold was also evaluated. The gold contents of anode slime and a reference standard ore were determined by applying the proposed method. Keywords: Resin preconcentration; separation; gold; geological samples; anodic slime; flame atomic absorption spectrometry Because of the low abundance of gold and high salt concentrations in geological samples and anode slimes, the direct determination of gold by FAAS in solutions obtained by wet decomposition of samples is often difficult or even impossible.For this reason, separation of gold from the sample matrix and increases in analyte concentrations are useful steps in the analysis. In trace element analysis, preconcentration and separation methods also enhance the sensitivity and precision of the determination.1 For sub-mg g21 levels in geological samples, ETAAS, ICP-AES and neutron activation analysis have been the dominant and preferred methods either with or without separation and preconcentration steps.2–6 However, FAAS has been widely used after a preconcentration step for trace analysis because it is a relatively simple method and requires less expensive instrumentation.7,8 The analysis of geological samples and anode slimes for gold is very difficult owing to the high concentrations of widely varying matrix elements such as Fe and Cu.Therefore, before the determination, the gold must be separated from the sample matrix and enriched. For this purpose, several separation and preconcentration techniques for gold, such as solvent extraction, 9–12 coprecipitation4 and ion-exchange or adsorption,13–15 have been developed. Among these techniques, sorption is one of the most widely used.For this purpose, XAD resins16–18 and various polymers19,20 have been used. Because of the purity and good adsorption properties, the commercially available Amberlite XAD copolymers (XAD-2, -4, -7 and -8) have been widely used as adsorbents.16–18 However, Amberlite XAD-16 has not been widely used for the preconcentration of trace elements.21 This paper describes the determination of gold in geological samples and anode slimes after preconcentration on a column packed with Amberlite XAD-16 resin, which has the largest surface area in the XAD series.The optimum conditions for adsorption and determination procedures are also discussed. The influence of interfering elements such as iron on the determination of gold was evaluated. Experimental Apparatus A Philips (Eindhoven, The Netherlands) PU 9285 flame atomic absorption spectrometer equipped with an air–acetylene burner and gold hollow cathode lamp operated at 11.2 mA was used for the determination of gold.The operating conditions were wavelength 242.8 nm, bandpass 0.5 nm and fuel flow rate 0.9 l min21. Deuterium lamp background correction was used. All pH measurements were made with a Consort (Turnhout, Belgium) digital pH meter and a combined glass electrode. Column Preparation The column diameter and amount of resin were chosen taking into account the information obtained from our previous studies.21 For subsequent experiments, a glass column (30 mm 3 10 mm id) with a glass-wool plug over its stopcock was used.A 0.2 g amount of resin was rested on the glass-wool and compressed in order to avoid channel formation. Another plug of glass-wool was placed on top, so that the resin was not disturbed during sample passage. Before use, 1 mol l21 hydrochloric acid was passed through the column for cleaning and preconditioning. Reagents and Standard Solutions Amberlite XAD-16 resin (Rohm and Haas, Philadelphia, PA, USA; surface area 800 m2 g21; wet mesh size 20–60) was used after washing with methanol, 1 mol l21 hydrochloric acid and triply distilled water and dried at 60 0C.Analytical-reagent grade chemicals were used without further purification. Triply distilled water was used in all experiments. A stock standard solution of gold (1000 mg l21) was prepared by dissolving the appropriate amount of tetrachloroauric( iii)acid trihydrate (HAuCl4·3H2O) in water. Working standard solutions were prepared by dilution.General Procedure for Optimization of the Preconcentration Step Volumes of 100 ml of spiked solutions containing 25 mg of gold (0.25 mg l21) were used. The pH of the solution was adjusted to the desirochloric acid and potassium hydroxide solutions. The resulting solution was passed through the column at an optimum flow rate of 5 ml min21. After the sample solution had been passed through the column, the retained metal ions were eluted from the resin by using 5 ml of 0.3 mol l21 potassium iodide solution in methanol and 5 ml of pure methanol.Analyst, March 1997, Vol. 122 (239–242) 239The eluate in a 10 ml calibrated flask was aspirated into an air–acetylene flame and the gold was determined by FAAS. The resin was used repeatedly after washing with 1 mol l21 hydrochloric acid and water. Using the procedure described above, the recovery of gold was calculated from the ratio of the concentration found by FAAS to that calculated theoretically.Sample Preparation Samples were dissolved by a commonly used method.9,13,15 An anode slime sample and a geological standard reference material (Gold Ore MA-3 certified by the Canada Center For Mineral and Energy Technology) were dried at 110 0C for 2 h, then approximately 1 g of gold ore and 0.1 g of anode slime were weighed accurately into a beaker separately. In order to decompose the sample, about 20 ml of aqua regia were added to the beaker and the solution was evaporated to dryness.This decomposition procedure was repeated and 10 ml of water were added. The suspension formed was filtered through a fine filterpaper. The insoluble part retained on the filter-paper was washed with 1 mol l21 hydrochloric acid and water. The filtrate and washings were collected in a 100 ml calibrated flask. Potassium hydroxide was used to adjust the pH of the solution to 2. Gold was preconcentrated from this solution with the column procedure described above and determined by FAAS.Results and Discussion Effect of Hydronium Ion Concentration The retention of gold on the resin was studied as a function of pH and/or the concentration of hydrochloric acid. The results are shown in Fig.1. As can be seen, a quantitative recovery (!95%) of gold was obtained at pH 2, and this pH was chosen as the optimum for subsequent experiments. Effect of Elution Solutions Elution studies were performed with halide solutions (potassium iodide and potassium bromide) in water and methanol and acidic solutions (hydriodic and hydrochloric acid) at various concentrations.The results are shown in Table 1. The best elution ( > 95% recovery) was obtained by using 5 ml of 0.3 mol l21 potassium iodide solution in methanol. Gold was quantitatively eluted with this eluent. The use of potassium iodide in methanol as the eluent was very convenient because the stability constant for the iodide complex of gold was high.In addition, methanol improved the nebulization efficiency by decreasing the viscosity of the sample solution. A 5 ml volume of 0.3 mol l21 KI solution and 5 ml of pure methanol were used as eluents in all subsequent experiments. Effect of Sample Volume and Analyte Concentration In order to deal with real samples containing very low concentrations of gold, the maximum applicable sample volume must be determined. Therefore, the effect of changes in the volume of sample solution passed through the column on the sorption of Au was investigated. Volumes of 100, 250, 500, 750 and 1000 ml of sample solutions containing 0.25, 0.10, 0.05, 0.033 and 0.025 mg l21 of Au were passed through the column. The recovery of gold was approximately quantitative (about 95%) up to a sample volume of 750 ml.At higher sample volumes, the gold recovery decreased gradually with increasing volume of sample. In this work, the elution volume was 10 ml and the highest concentration factor was 75 for a 750 ml sample volume. These results also indicate the effect of analyte concentration on the retention of Au.Normally FAAS is unable to determine gold at levels below 0.1 mg l21,22 but the proposed technique allowed gold to be determined in the range 0.033–0.25 mg l21. Effect of Sample and Eluent Flow Rate on Sorption Since the retention of elements on an adsorbent depends on the flow rate of the metal solution, the effect of the flow rate was examined under the optimum conditions (pH, eluent type, etc.).The flow rate varied from 1 to 5 ml min21. It was found that the retention of gold was not affected by flow rate up to 5 ml min21. Since either suction or pressure is needed for flow rates higher than 5 ml min21, no test was carried out for such flow rates. The effect of eluent flow rate on the recovery was also examined. Three flow rates (1, 2.5 and 5 ml min21) were studied and recoveries of 95, 88 and 60%, respectively, at 5 ml min21 were obtained.Since the recovery increases with decreasing flow rate of the eluent, 1 ml min21 was adopted in all experiments. Precision of Method The precision of the method was investigated at the optimum pH of 2 for the spiked solution (0.25 mg l21 Au) described above. For this purpose, the general procedure was performed successively and gold was determined in the solution by FAAS. The mean recovery for eight determinations at the 95% confidence level was 95.48 ± 0.04%.As can be seen, the recovery is quantitative and it is sufficient for preconcentration purposes. The precision of the method is very good, the RSD being about 2.1%. Adsorption Isotherm and Adsorption Mechanisms For the determination of the adsorption behaviour of the resin, the amount of adsorbed gold was studied as a function of gold Fig. 1 Effect of hydronium ion concentration on the recovery of gold. Table 1 Effect of elution solutions on the recovery of gold Solution Recovery (%) 5 ml of 0.1 mol l21 KI in methanol and 5 ml of pure methanol 91.5 5 ml of 0.1 mol l21 KBr in methanol and 5 ml of pure methanol 9.0 5 ml of 0.3 mol l21 KI in methanol and 5 ml of pure methanol 95.0 5 ml of 1 mol l21 KI in methanol and 5 ml of pure methanol 95.0 5 ml of 1 mol l21 HCl 0.5 5 ml of 0.1 mol l21 HI in water 1.0 5 ml of 0.1 mol l21 HI in methanol and 5 ml of pure methanol 69.3 5 ml of 0.3 mol l21 KI in methanol 59.6 5 ml of pure methanol 2.6 240 Analyst, March 1997, Vol. 122H+AuCl4 – concentration (Fig. 2). The adsorption isotherm of gold conforms to the Langmuir equation.23 Graphical treatment of the data yielded Langmuir parameters of capacity of the resin for gold 0.55 mmol g21 (108 mg g21) and equilibrium binding constant, K, about 6 3 103 l mol21. The Langmuir treatment demonstrated that the surface is nearly fully covered with a monomolecular layer of gold at high concentrations. Therefore, at high concentrations adsorption does not increase linearly. In addition to the adsorption isotherm of the resin, another important parameter, the adsorption mechanism was studied.Amberlite XAD-16 is a styrene–divinylbenzene copolymer. The metal is probably adsorbed as a chloro complex such as HAuCl4. The adsorption might result in the formation of a complex compound such as that shown below:24 Effect of Matrix Elements The determination of gold in naturally occurring ores is difficult owing to the high concentrations of other (accompanying) elements such as Na, Fe, Ni and Cu.For this reason, a fixed amount of gold was taken with various combined amounts of matrix elements such as Fe, Cu, Ni, Na, K, Mg, Ca and the recommended procedure was applied. The results are summarized in Table 2. The recovery of gold decreased by in the presence of a large excess of matrix elements, because of the insufficient resin capacity. Especially iron, generally present in the ore matrix, decreased the gold adsorption.25 Gjerde and Fritz25 found that a 1000-fold excess of iron over gold decreased the gold recovery from 98% to 80%.For this reason, the effect of iron on the adsorbance and recovery of gold was specially investigated. The results are given in Table 3. As can be seen, by using the proposed method an approximately 5000-fold excess of iron over gold did not interfere; the recovery of gold was almost constant.In order to explain which element(s) cause the large decreases in recovery shown in Table 2, the effect of each individual matrix element on the recovery of gold was investigated. For this purpose, only the effect of 250 mg l21 of each matrix element on 0.25 mg l21 of gold was studied. The results are given in Table 4. The recovery of gold did not decrease when the matrix elements were studied individually, but when they were used all together the recovery of gold decreased by about 22.5% (Table 2). The reason for this result might be attributed to the insufficient column capacity. The retention of Cu and Ni on XAD-16 resin at pH 2 is less,21 but the retention of Fe is high (about 40%).Hence, it can be said that the capacity of resin has been fully utilized when the matrix elements were used all together at a high concentration (250 mg l21). Application The proposed method was applied to the determination of gold in an anode slime and a geological sample. As the geological Fig. 2 Adsorption isotherm of XAD-16 for gold. Table 5 Determination of gold in real samples Au concentration Recovery LOD‡ Sample* Certified Found† (%) (mg l21) Gold Ore MA-3 7.49 ± 0.19 mg g21 7.23 ± 0.12 mg g21 96 0.046 Anode slime 0.076% m/m§ 0.071 ± 0.003% m/m 93 0.046 * The composition of Gold Ore MA-3 (only gold was certified) was Si 25, Al 7.22, Fe 3.85, Ca 3.31, K 3.05, Na 2.42, C 0.76 and S 0.56% m/m, Ag 1.5 mg g21 and Au 7.49 mg g21. The composition of the anode slime was Cu 18.86, Pb 7.67, Sn 1.36, Ag 2.29, Au 0.076, Fe 0.10 and SiO2 1.69% m/m.† Average of five determinations at the 95% confidence level. ‡ Limit of detection based on 99.7% (3s) confidence level (n = 10). § Determined spectrophotometrically by Makina ve Kimya End�ustrisi Kurumu. Table 2 Effect of combined matrix elements on the recovery of gold Foreign metal (Fe, Cu, Ni, Au Na, K, Ca, Mg) concentration/ Added/ Found/ Recovery mg l21 mg l21 mg l21 (%) 0 0.25 0.24 95.0 25 0.25 0.22 86.4 125 0.25 0.12 46.2 250 0.25 0.06 22.5 Table 3 Effect of iron concentration on the recovery of gold Fe/ Au/ Recovery mg l21 mg l21 (%) 100 0.25 94.0 500 0.25 93.5 1000 0.25 94.6 1250 0.25 93.8 Table 4 Effect of individual matrix elements at a concentration of 250 mg l21 on the recovery of gold Recovery Element of gold (%) None 95.0 Cu 92.0 Ni 94.0 Na 99.5 K 97.6 Mg 96.0 Ca 104.0 Analyst, March 1997, Vol. 122 241sample a Certified Gold Reference Ore and as an anode slime the product from the brass plant of Makina ve Kimya End¡§ustrisi Kurumu were analysed. The results are shown in Table 5.The limit of detection was calculated on the basis of three times the standard deviation of the absorbance using a series of measurements of blank solutions prepared according to the procedure given under Sample Preparation without taking the sample. As can be seen, the preconcentration, separation and determination of trace amounts of gold by the proposed method in a geological material and anode slime in the presence of high concentrations of foreign cations and anions including iron (in MA-3 about a 5000-fold excess) could be performed successfully.The relative error was 3.5%, for gold ore and 6.6% for anode slime. Conclusion The proposed column preconcentration method provides a simple, sensitive and accurate method for the preconcentration of gold in geological materials and anode slimes and is also successful in the presence of iron.Gold can be retained on the column by adjusting the sample solution to pH 2 and then be eluted from the column using potassium iodide. Most of the metals including iron do not influence the gold determination. The precision and recovery were satisfactory. Further work is in progress to evaluate the use of XAD-16. References 1 Mizuike, A., Enrichment Techniques for Trace Analysis, Springer, Berlin, 1983. 2 McHugh, J. B., At. Spectrosc., 1983, 4(2), 66. 3 Kontas, E., At. Spectrosc., 1981, 2(2), 59. 4 Chung, Y. S., and Barnes, R. M., J. Anal. At. Spectrom., 1988, 3, 1079. 5 Elson, C. M., and Chatt, A., Anal. Chim. Acta, 1983, 155, 305. 6 Inductively Coupled Plasma Emission Spectroscopy, ed. Boumans, P. W. J. M., Wiley, New York, 1987, pt. 2, ch. 3, pp. 27¡©47. 7 Xu, S., Sun, L., and Fang, Z., Anal. Chim. Acta, 1991, 245, 7. 8 Elci, L., Anal. Lett., 1993, 26, 1025. 9 Brooks, R. R., Holzbecher, J., Ryan, D. E., and Zhang, H. F., At. Spectrosc., 1981, 2(5), 151. 10 Tsukahara, I., Talanta, 1977, 24, 633. 11 Marczenko, Z., and Kowalski, T., Anal. Chim. Acta., 1984, 156, 193. 12 McBryde, W. A. E., and Yoe, J. H., Anal. Chem., 1948, 20, 1094. 13 Elci, L., I`s©¥ldar, S., and Do¢�gan, M., Anal. Chim. Acta, 1994, 293, 319. 14 Koshima, H., and Onishi, H., Anal. Sci., 1988, 4, 611. 15 Rivoldini, A., and Haile, T., At. Spectrosc., 1989, 10(3), 89. 16 Yang, X. G., and Jackwerth, E., Fresenius¡� Z. Anal. Chem., 1989, 335, 712. 17 Yang, X. G., and Jackwerth, E., Fresenius¡� Z. Anal. Chem., 1989, 335, 483. 18 Yang, X. G., and Jackwerth, E., Fresenius¡� Z. Anal. Chem., 1990, 336, 588. 19 Hoshi, S., Tanaka, Y., Inoue, S., and Matsubara, M., Anal. Sci., 1989, 5, 471. 20 Ar©¥k, N., and T¡§urker, A. R., Fresenius¡� J. Anal. Chem., 1991, 339, 874. 21 T¡§urker, A. R., and Tunceli, A., Fresenius¡� J. Anal. Chem., 1993, 345, 755. 22 Taylor, M. J. C., Barnes, D. E., and Marshall, G. D., Anal. Chim. Acta, 1992, 265, 71. 23 Adamson, A. W., Physical Chemistry of Surfaces, Wiley-Interscience, New York, 4th edn., 1982, p. 521. 24 Koshima, H., Anal. Sci., 1986, 2, 255. 25 Gjerde, D. T., and Fritz, J. S., J. Chromatogr., 1980, 188,. 391. Paper 6/05008E Received July 17, 1996 Accepted November 27,1996 242 Analyst, March

ISSN:0003-2654    

DOI:10.1039/a605008e

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Sorption of Arsenic, Bismuth, Mercury, Antimony, Selenium and Tin on Dithiocarbamate Loaded Polyurethane Foam as a Preconcentration Method for Their Determination in Water Samples by Simultaneous Inductively Coupled Plasma Atomic Emission Spectrometry and Electrothermal Atomic Absorption Spectrometry Sonja Arpadjan, Lili Vuchkova and Elena Kostadinova University of Sofia, Faculty of Chemistry, 1126 Sofia, Bulgaria Column solid-phase extraction using dithiocarbamate loaded polyurethane foam was applied to the preconcentration of trace amounts of As, Bi, Hg, Sb, Se and Sn from water samples prior to their measurement by simultaneous ICP-AES and ETAAS.The sorption recoveries of all the analytes were higher than 97% for 150 ml water sample solutions passed through the column with pH Å 4.5 and sodium chloride concentration up to 3%. For As, Bi, Hg, Sb and Se, quantitative solid-phase extraction can be achieved over a wide pH range, from 0.5 to pH 5. The combination of the proposed preconcentration method with subsequent simultaneous analyte determination in the methanol eluates by ICP-AES permits the detection of 3 mg l21 As and Se, 8 mg l21 Bi, 0.12 mg l21 Hg, 2 mg l21 Sb and 6 mg l21 Sn in water samples.ETAAS measurement after dissolution of the sorbed analytes in isobutyl methyl ketone allows the detection of 0.06 mg l21 As and Sb, 0.1 mg l21 Bi and Sn, 0.08 mg l21 Se and 0.3 mg l21 Hg. Keywords: Column solid-phase extraction; trace element preconcentration water samples; inductively coupled plasma atomic emission spectrometry; electrothermal atomic absorption spectrometry In routine laboratory measurements, As, Bi, Sb, Se and Sn are determined in their covalent hydride form and Hg in its atomic form in the vapour phase using atomic spectroscopic methods.However, for extremely low analyte concentrations a preliminary preconcentration is required. In comparison with the most commonly used solvent extraction enrichment procedure, column solid-phase extraction allows preconcentration from a larger sample volume, establishing higher concentration factors, simple storage and transportation of the pre-treated samples.1–12 In a previous study a method was developed for the preconcentration of trace amounts of Cd, Co, Cu, Fe, Hg, Mn, Ni and Pb from analytical reagent-grade sodium salts and oxalic acid on polyurethane foam (PU) immobilized with ammonium hexamethylenedithiocarbamate (HMDC) or with a mixture of HMDC and methyltrioctylammonium chloride (MTOAC) prior to their measurement by AAS or ICP-AES.13,14 Elution with 4 mol l21 HNO3 for subsequent ICP-AES determinations was found to be inappropriate for Hg determination. Quantitative elution of all the sorbed complexes can be achieved only by their total dissolution in organic solvents.13–15 The aim of this work was (i) to optimize the conditions for the quantitative column solid-phase extraction preconcentration of AsIII, BiIII, HgII SbIII, SeIV and SnIV on PU–HMDC from various water samples, (ii) to investigate the analytical potential of the subsequent simultaneous analyte determination in methanol eluates by ICP-AES and (iii) to investigate the analytical potential of the subsequent analyte determination in isobutyl methyl ketone (IBMK) eluates by ETAAS.Experimental Apparatus ICP-AES measurements were performed with a simultaneous ICP spectrometer (Spectroflame ICP; SPECTRO, Kleve, Germany, holographic grating with 3600 lines mm21, plasma generator frequency 27.12 MHz).The plasma was run at 1750 W power with 21 l min21 plasma, 0.99 l min21 auxiliary and 0.66 l min21 nebulizer argon flow rates. The spectra were measured at an observation height of 14 mm. The integration time was 10 s. A Meinhard-type nebulizer was used throughout. The wavelengths used, viz., As 184.04, Bi 230.0, Hg 184.95, Se 196.09, Sb 206.8 and Sn 189.98 nm, offer the optimum signalto- background ratio for the analytes.ETAAS measurements were performed on a Perkin-Elmer (Norwalk, CT, USA) Zeeman 3030 atomic absorption spectrometer with an HGA-600 atomiser. The light sources used were electrodeless discharge lamps for As, Hg, Se, Sb and Sn and a hollow-cathode lamp for Bi. The spectral bandpass and the wavelength used were as recommended by Perkin-Elmer. Uncoated graphite tubes with pyrolytic graphite platforms were used as atomizers.The injection of the eluates (10 ml) into the tube and injection of the calibration standards (10 ml) and of the chemical modifier solution (10 ml) were carried out manually. The graphite furnace operating parameters are presented in Table 1. In all instances only peak areas (integrated absorbance) were used for quantification. Reagents All reagents were of analytical-reagent grade. Doubly distilled water was used throughout. Methanol and chloroform were used after additional purification by distillation. The organic solvent IBMK and the ligand HMDC (Fluka, Buchs, Switzerland) were used as received.The long-chain quaternary ammonium salt MTOAC was used as a 2% solution in IBMK. Preparation of Standard Solutions for ICP-AES A multi-element aqueous standard solution (100 mg ml21 As, Sb, Se and Sn, 200 mg ml21 Bi and 20 mg ml21 Hg) was prepared from atomic absorption standard solutions (BDH, Analyst, March 1997, Vol. 122 (243–246) 243Poole, Dorset, UK). Calibration standard solutions were prepared daily by appropriate dilution with 65% v/v aqueous methanol of HMDC (0.5% m/v HMDC).Preparation of Standard Solutions for ETAAS A multi-element aqueous standard solution containing 200 mg l21 AsIII, SbIII, SeIV and SnIV, 400 mg l21 SeIV and 800 mg l21 HgII was prepared from atomic absorption standard solutions (BDH). The organic calibration standards were prepared daily in the following way: appropriate volumes from the multielement aqueous standard solution were diluted to 10 ml with water in extraction tubes, then 2 ml of acetic acid buffer (pH 4.66) and 5.0 ml of a 1% solution of HMDC in IBMK were added.Extraction was performed for 1 min. Aliquots (10 ml) from the organic layer were used as calibration standards. Preparation of Palladium Chemical Modifier Soluble in IBMK A 1 ml volume of aqueous standard solution containing 30 mg l21 PdII was diluted with 10 ml of 6 mol l21 HCl in an extraction tube, then 3.0 ml of MTOAC solution in IBMK were added and extracted for 5 min.Using these conditions, Pd was extracted into IBMK as the ion association complex PdCl4(MTOA)2 . A definite volume (10 ml) from this organic layer was injected into the platform of the graphite tube for ETAAS determination of As, Se and Sn. Sorbent Preparation Cubical foam pieces 5 mm long were cut from commercially available polyurethane foam (polyether type). The cleaned and dried foam pieces were soaked in a chloroform solution of HMDC (1.5 g of HMDC for 1 g of PU) and then squeezed occasionally until all the solution was retained in the PU cubes.Finally, the organic solvent was evaporated by drying the foam in air. The dithiocarbamate-loaded PU foam pieces were stored in a dark borosilicate glass container. Column Preparation The HMDC-loaded PU foam pieces (approximately 0.3 g) were transferred into a disposable syringe after packing a small wad of cotton-wool in the end of the column and covering it with a disc of filter-paper.The sample solution was passed through the sorbent column using a peristaltic pump as described previously. 13 General Procedure for ICP-AES Measurements Aqueous test solutions containing 2 mg of As, 5 mg of Bi, 2 mg of Sb, 3 mg of Se and 2 mg of Sn were pumped through the sorbent column at a flow rate of 2 ml min21. Dilute NH3 solution (1 + 1) and HCl (1 + 1) were used for adjusting of the pH. The physically immobilized organic chelating ligand, HMDC, together with the analyte–dithiocarbamate complexes formed were completely dissolved in methanol.For this purpose, 3.0 ml of pure methanol were placed in a small, dry quartz beaker and was passed five consecutive times through the sorbent with the aid of the syringe plunger. The methanol solution was then diluted with 3.0 ml of doubly distilled water and nebulized into the argon plasma of the ICP-AES system. The emission intensity signals of the elements in this solution were measured against calibration standard solutions in order to evaluate the sorption efficiency.For each sample a new sorption column was applied. General Procedure for ETAAS Measurements Aqueous test solutions containing 50 ng of As, Bi, Sb and Sn, 100 ng of Se and 400 ng of Hg were pumped through the sorbent column at a flow rate of 2 ml min21 after adjusting of the pH with dilute NH3 and HCl. Elution was performed with 1.0 ml of IBMK as described for the ICP-AES technique.The atomic absorption signals of the elements in this IBMK solution were measured against organic calibration standard solutions. For each experiment a new column was prepared. Results and Discussion Influence of pH The results for the influence of the pH of the water sample solution on the solid-phase extraction preconcentration with PU–HMDC sorbent are presented in Table 2. AsIII, BiIII, HgII, SbIII and SeIV form stable dithiocarbamate complexes with HMDC in the solid state in the pH range 1–5 and SnIV in the pH range 4–5.However, AsV, SbV and SeVI do not form a complex Table 1 Temperature programme for the determination of analytes in sorbent extraction eluates (10 ml IBMK solution) using a pyrolytic graphite platform and uncoated graphite tube Step Element Parameter I II III IV Temperature/°C 120 Variable Variable Variable Ramp time/s 10 Variable Variable 2 Hold time/s 15 Variable Variable 3 Read — — On — Gas flow rate/ml min21 300 300 0 300 As* Temperature/°C 1200 2200 2500 Ramp/s 12 0 Hold/s 15 3 Bi Temperature/°C 700 1100 1500 Ramp/s 10 0 Hold/s 20 5 Hg Temperature/°C 250 900 1600 Ramp/s 5 1 Hold/s 10 14 Sb Temperature/°C 1100 1600 2200 Ramp/s 12 0 Hold/s 15 5 Se* Temperature/°C 1100 2100 2400 Ramp/s 12 0 Hold/s 15 3 Sn* Temperature/°C 1300 2300 2500 Ramp/s 13 0 Hold/s 10 4 * Addition of 10 ml of modifier solution [10 mg ml21 Pd as PdCl4(MTOA)2 in IBMK].Table 2 Effect of pH on the recovery of trace elements in the analysis of water samples after solid-phase extraction preconcentration on polyurethane –HMDC (n = 3 parallel determinations) Recovery ± s (%) pH AsIII BiIII HgII SbIII SeIV SnIV 1 97 ± 2 98 ± 1 99 ± 2 98 ± 2 98 ± 2 17 ± 3 2 94 ± 4 99 ± 2 99 ± 2 95 ± 3 94 ± 2 28 ± 6 3 95 ± 3 99 ± 2 99 ± 2 97 ± 2 96 ± 2 62 ± 5 4 99 ± 2 100 ± 2 100 ± 2 99 ± 2 99 ± 2 94 ± 3 5 95 ± 2 100 ± 2 100 ± 2 99 ± 2 97 ± 2 97 ± 3 6 62 ± 5 99 ± 2 100 ± 2 99 ± 2 70 ± 6 58 ± 4 7 < 5 99 ± 2 99 ± 2 99 ± 2 < 5 15 ± 5 244 Analyst, March 1997, Vol. 122with HMDC under any conditions. A reduction step must therefore precede the passage through the sorbent column. For AsIII, BiIII, HgII, SbIII and SeIV, sorption recoveries > 98% can also be achieved for water samples acidified with HCl (0.1–0.7 mol l21 HCl in the solution). Hence when the determination of tin is not required, the water sample can simply be acidified with HCl before preconcentration instead of adjusting the pH to 4.5 ± 0.5 with dilute ammonia solution (1 + 1) or HCl (1 + 1).Influence of Volume of Water Sample Solution Water sample solution volumes up to 150 ml permit the achievement of quantitative solid-phase extraction preconcentration of all the analytes studied on PU–HMDC sorbent. Some decrease in the recovery for As, Sb, Se and Sn was observed when the sample volume was increased from 150 to 200 ml. The sorption for Bi and Hg is quantitative for a 200 ml sample volume.Hence the enrichment factor achieved for the proposed sorbent preconcentration ETAAS method is 150 (elution of the sorbed analyte–dithiocarbamate complexes with 1.0 ml of IBMK). The necessity to perform the ICP-AES measurements in at least 5.0 ml of methanol eluates leads to an enrichment factor of only 30. Influence of Sodium Chloride Concentration The influence of the concentration of NaCl on the efficiency of sorption on PU–HMDC was investigated in order to elucidate the possibilities of determining the studied analytes in sea-water samples using the proposed preconcentration procedure.The results obtained are presented in Table 3. The retention of Bi and Hg on the PU–HMDC sorbent is quantitative for all water sample volumes and sodium chloride concentrations studied. For the simultaneous quantitative preconcentration of all the analytes studied from a 150 ml water sample, the concentration of NaCl in the solution pumped through the PU–HMDC sorbent column must not exceed 3.5%.Analysis of Water Samples For the determination of trace amounts of As, Bi, Hg, Sb, Se and Sn in water samples, 3 g of KI were dissolved in 150.0 ml of sample solution and the pH was adjusted to 4.5 ± 0.5 with dilute ammonia solution (1 + 1) or HCl (1 + 1). When the determination of Sn is not required, the sample (150.0 ml) after addition of 3 g of KI is acidified with 1.5 ml of concentrated HCl. The sample is heated gently at 60 °C for 1 h to ensure the reduction of arsenic, antimony and selenium to their lower oxidation state.After cooling to room temperature, the sample is pumped through the sorption column at a flow rate of 2 ml min21, then the column is washed with 5 ml of doubly distilled water. In this state, prior to elution, the sorbent columns are suitable for conservation and transportation. The elution and the measurements by ICP-AES and ETAAS were performed as described in the general procedure.Absolute blanks of the proposed procedure are presented in Table 4. The largest part of the blank signal, especially for Bi and Sn, is due to the KI, and also to the ligand HMDC. The limits of detection of each element, expressed as the blank value ± three times the standard deviation of the blank, are reported in Table 5. With the exception of Hg, the solid-phase extraction preconcentration ETAAS method ensures lower detection limits owing to the higher enrichment factor and the higher sensitivity of the graphite furnace measurement technique.The method using ICP does not offer any advantage over hydride generation ICP-AES. The 30-fold analyte preconcentration does not compensate for the poorer detection limit of the Table 3 Effect of the water sample volume and NaCl concentration (synthetic mixtures) on the solid-phase extraction recovery (pH 4.3) Recovery ± s (%) Volume*/ NaCl ml (%) AsIII BiIII HgII SbIII SeIV SnIV 100 (n = 5) 3.5 99 ± 2 100 ± 2 100 ± 2 99 ± 2 99 ± 2 95 ± 6 100 (n = 5) 5 97 ± 2 98 ± 1 99 ± 2 98 ± 2 98 ± 2 93 ± 3 150 (n = 4) 3.5 99 ± 2 99 ± 2 99 ± 2 99 ± 2 99 ± 2 95 ± 4 150 (n = 4) 5 87 ± 6 98 ± 2 99 ± 2 90 ± 5 91 ± 5 90 ± 7 * n = Number of parallel determinations.Table 6 Recovery of analytes added to sea-water (n = 4) Solid-phase extraction–ICP-AES Solid-phase extraction Added/ Found ±s/ Recovery ±s Added/ Found ±s/ Recovery ±s Element mg l21 mg l21 (%) mg l21 mg l21 (%) As 62.5 61.4 ± 2.8 98.2 ± 4.5 0.666 0.648 ± 0.042 97.3 ± 6.3 Bi 62.5 64.0 ± 3.5 102.4 ± 5.6 0.666 0.668 ± 0.026 100.3 ± 3.9 Hg 25.0 25.8 ± 1.6 103.2 ± 6.4 2.67 2.74 ± 0.07 102.6 ± 2.6 Sb 62.5 60.8 ± 3.2 97.3 ± 5.1 0.666 0.644 ± 0.028 96.7 ± 4.2 Se 125.0 122 ± 5 97.6 ± 4.0 1.33 1.26 ± 0.08 94 ± 6 Sn 125.0 127 ± 8 101.6 ± 6.4 0.666 0.687 ± 0.034 103.2 ± 5.1 Table 4 Reagent blanks for preconcentration procedure.Doubly distilled water (150 ml) + 3 g of dissolved KI was pumped though 0.30 g of sorbent containing 0.15 g of HMDC; n = 5 Element Blank ± s/ng As 4.5 ± 0.5 Bi 12.9 ± 0.6 Hg 6.8 ± 0.4 Sb 4.7 ± 0.4 Se < detection limit Sn 10.8 ± 1.2 Table 5 Detection limits (3s) of polyurethane-HMDC preconcentration procedure for trace determination in waters by ICP-AES and ETAAS Element ICP-AES*/mg l21 ETAAS*/mg l21 As 3.2 0.06 Bi 8 0.1 Hg 0.12 0.3 Sb 2 0.06 Se 3 0.08 Sn 6 0.1 * Elution of the retained on the sorbent analytes (150 ml water sample pumped though the sorbent column) in 2.5 ml of methanol + 2.5 ml doubly distilled water for ICP-AES and in 1 ml IBMK for ETAAS measurements.Analyst, March 1997, Vol. 122 245direct simultaneous ICP-AES measurement technique additionally owing to the lower analytical potential of the ICP method in methanol solutions than in aqueous solutions. However, the combination of the described preconcentration procedure with hydride generation in methanol solutions of the analyte– dithiocarbamate complexes and subsequent simultaneous ICPAES determination of As, Bi, Hg, Sb, Se and Sn leads to a substantial improvement in sensitivity.16 The accuracy of the procedure was investigated by determining the analyte content in spiked sea-water (Black Sea-water with analyte concentrations lower than the detection limit of the method).The results (Table 6) show sufficiently high recoveries and a precision (RSD) of 3–7%. The precision is very good despite working close to the detection limit.In addition, the accuracy and precision of the described methods were evaluated by comparing the results for waste waters with those obtained by direct ICP-AES measurements after hydride generation (Table 7). No significant differences in the results obtained were observed. Conclusion Solid-phase extraction on a PU–HMDC sorbent column is an effective preconcentration procedure for simple storage, transportation and measurement of trace levels of As, Bi, Hg, Sb, Se and Sn in various kinds of water samples using ICP-AES or ETAAS.The combination of the preconcentration procedure with the ETAAS measurement technique offers higher enrichment factors and higher sensitivity. The authors thank the Bulgarian Foundation of Sciences, Project X-519. References 1 Braun, T., Fresenius’ Z. Anal. Chem., 1983, 314, 652. 2 Burba, P., Rocha, J. C., and Schulte, A., Fresenius’ J. Anal. Chem., 1993, 346, 414. 3 Jambor, J., and Javorek, T., Collect.Czech. Chem. Commun., 1993, 58, 1821. 4 Prakash, N., Csanady, G. J., Michaelis, M., and Knapp, G., Mikrochim. Acta, 1989, 3, 257. 5 Maloney, M. P., Moody, G. J., and Thomas, J. D. R., Analyst, 1980, 105, 1087. 6 Horvath, Z., Lastztity, A., Szakacs, O., and Bozsai, G., Anal. Chim. Acta, 1985, 173, 273. 7 Terada, K., Matsumoto, K., and Nanao, Y., Anal. Sci., 1985, 1, 145. 8 Tong, A., and Akama, Y., Anal. Chim. Acta, 1990, 230, 175. 9 Tong, A., Akama, Y., and Tanaka, S., Analyst, 1990, 115, 947. 10 Yebra-Biurrun, M. C., Bermejo-Barrera, A., and Bermejo-Barrera, P., Mikrochim. Acta, 1992, 109, 243. 11 Schilling, T., Schramel, P., Michalke, B., and Knapp, G., Mikrochim. Acta, 1994, 116, 83. 12 Kantipuly, C., Katragadda, S., Chow, A., and Gesser, H., Talanta, 1990, 37, 491. 13 Alexandrova, A., and Arpadjan, S., Analyst, 1993, 118, 1309. 14 Alexandrova, A., and Arpadjan, S., Anal. Chim. Acta, 1995, 307, 71. 15 Sperling, M., Yin, X., and Welz, B., Analyst, 1992, 117, 629. 16 Vuchkova, L., and Arpadjan, S., Talanta, 1996, 43, 479.Paper 6/06917G Received October 9, 1996 Accepted December 3, 1996 Table 7 Results (mg l21 ± s) for the analysis of waste water (n = 4) PU–HMDC HG-ICP in aqueous Element ETAAS ICP-AES solution As 4.3 ± 0.2 4.5 ± 0.3 4.3 ± 0.2 Bi 6.7 ± 0.4 < 8 6.9 ± 0.4 Hg < 0.4 0.28 ± 0.03 0.28 ± 0.02 Se 5.8 ± 0.4 5.6 ± 0.4 5.9 ± 0.3 Sb 8.2 ± 0.4 8.3 ± 0.5 8.1 ± 0.4 Sn 9.4 ± 0.5 9.2 ± 0.6 9.5 ± 0.5 246 Analyst, March 1997, Vol. 122 Sorption of Arsenic, Bismuth, Mercury, Antimony, Selenium and Tin on Dithiocarbamate Loaded Polyurethane Foam as a Preconcentration Method for Their Determination in Water Samples by Simultaneous Inductively Coupled Plasma Atomic Emission Spectrometry and Electrothermal Atomic Absorption Spectrometry Sonja Arpadjan, Lili Vuchkova and Elena Kostadinova University of Sofia, Faculty of Chemistry, 1126 Sofia, Bulgaria Column solid-phase extraction using dithiocarbamate loaded polyurethane foam was applied to the preconcentration of trace amounts of As, Bi, Hg, Sb, Se and Sn from water samples prior to their measurement by simultaneous ICP-AES and ETAAS.The sorption recoveries of all the analytes were higher than 97% for 150 ml water sample solutions passed through the column with pH Å 4.5 and sodium chloride concentration up to 3%. For As, Bi, Hg, Sb and Se, quantitative solid-phase extraction can be achieved over a wide pH range, from 0.5 to pH 5.The combination of the proposed preconcentration method with subsequent simultaneous analyte determination in the methanol eluates by ICP-AES permits the detection of 3 mg l21 As and Se, 8 mg l21 Bi, 0.12 mg l21 Hg, 2 mg l21 Sb and 6 mg l21 Sn in water samples. ETAAS measurement after dissolution of the sorbed analytes in isobutyl methyl ketone allows the detection of 0.06 mg l21 As and Sb, 0.1 mg l21 Bi and Sn, 0.08 mg l21 Se and 0.3 mg l21 Hg. Keywords: Column solid-phase extraction; trace element preconcentration water samples; inductively coupled plasma atomic emission spectrometry; electrothermal atomic absorption spectrometry In routine laboratory measurements, As, Bi, Sb, Se and Sn are determined in their covalent hydride form and Hg in its atomic form in the vapour phase using atomic spectroscopic methods.However, for extremely low analyte concentrations a preliminary preconcentration is required. In comparison with the most commonly used solvent extraction enrichment procedure, column solid-phase extraction allows preconcentration from a larger sample volume, establishing higher concentration factors, simple storage and transportation of the pre-treated samples.1–12 In a previous study a method was developed for the preconcentration of trace amounts of Cd, Co, Cu, Fe, Hg, Mn, Ni and Pb from analytical reagent-grade sodium salts and oxalic acid on polyurethane foam (PU) immobilized with ammonium hexamethylenedithiocarbamate (HMDC) or with a mixture of HMDC and methyltrioctylammonium chloride (MTOAC) prior to their measurement by AAS or ICP-AES.13,14 Elution with 4 mol l21 HNO3 for subsequent ICP-AES determinations was found to be inappropriate for Hg determination.Quantitative elution of all the sorbed complexes can be achieved only by their total dissolution in organic solvents.13–15 The aim of this work was (i) to optimize the conditions for the quantitative column solid-phase extraction preconcentration of AsIII, BiIII, HgII SbIII, SeIV and SnIV on PU–HMDC from various water samples, (ii) to investigate the analytical potential of the subsequent simultaneous analyte determination in methanol eluates by ICP-AES and (iii) to investigate the analytical potential of the subsequent analyte determination in isobutyl methyl ketone (IBMK) eluates by ETAAS.Experimental Apparatus ICP-AES measurements were performed with a simultaneous ICP spectrometer (Spectroflame ICP; SPECTRO, Kleve, Germany, holographic grating with 3600 lines mm21, plasma generator frequency 27.12 MHz).The plasma was run at 1750 W power with 21 l min21 plasma, 0.99 l min21 auxiliary and 0.66 l min21 nebulizer argon flow rates. The spectra were measured at an observation height of 14 mm. The integration time was 10 s. A Meinhard-type nebulizer was used throughout. The wavelengths used, viz., As 184.04, Bi 230.0, Hg 184.95, Se 196.09, Sb 206.8 and Sn 189.98 nm, offer the optimum signalto- background ratio for the analytes.ETAAS measurements were performed on a Perkin-Elmer (Norwalk, CT, USA) Zeeman 3030 atomic absorption spectrometer with an HGA-600 atomiser. The light sources used were electrodeless discharge lamps for As, Hg, Se, Sb and Sn and a hollow-cathode lamp for Bi. The spectral bandpass and the wavelength used were as recommended by Perkin-Elmer. Uncoated graphite tubes with pyrolytic graphite platforms were used as atomizers.The injection of the eluates (10 ml) into the tube and injection of the calibration standards (10 ml) and of the chemical modifier solution (10 ml) were carried out manually. The graphite furnace operating parameters are presented in Table 1. In all instances only peak areas (integrated absorbance) were used for quantification. Reagents All reagents were of analytical-reagent grade. Doubly distilled water was used throughout. Methanol and chloroform were used after additional purification by distillation.The organic solvent IBMK and the ligand HMDC (Fluka, Buchs, Switzerland) were used as received. The long-chain quaternary ammonium salt MTOAC was used as a 2% solution in IBMK. Preparation of Standard Solutions for ICP-AES A multi-element aqueous standard solution (100 mg ml21 As, Sb, Se and Sn, 200 mg ml21 Bi and 20 mg ml21 Hg) was prepared from atomic absorption standard solutions (BDH, Analyst, March 1997, Vol. 122 (243–246) 243Poole, Dorset, UK). Calibration standard solutions were prepared daily by appropriate dilution with 65% v/v aqueous methanol of HMDC (0.5% m/v HMDC). Preparation of Standard Solutions for ETAAS A multi-element aqueous standard solution containing 200 mg l21 AsIII, SbIII, SeIV and SnIV, 400 mg l21 SeIV and 800 mg l21 HgII was prepared from atomic absorption standard solutions (BDH). The organic calibration standards were prepared daily in the following way: appropriate volumes from the multielement aqueous standard solution were diluted to 10 ml with water in extraction tubes, then 2 ml of acetic acid buffer (pH 4.66) and 5.0 ml of a 1% solution of HMDC in IBMK were added.Extraction was performed for 1 min. Aliquots (10 ml) from the organic layer were used as calibration standards. Preparation of Palladium Chemical Modifier Soluble in IBMK A 1 ml volume of aqueous standard solution containing 30 mg l21 PdII was diluted with 10 ml of 6 mol l21 HCl in an extraction tube, then 3.0 ml of MTOAC solution in IBMK were added and extracted for 5 min.Using these conditions, Pd was extracted into IBMK as the ion association complex PdCl4(MTOA)2 . A definite volume (10 ml) from this organic layer was injected into the platform of the graphite tube for ETAAS determination of As, Se and Sn. Sorbent Preparation Cubical foam pieces 5 mm long were cut from commercially available polyurethane foam (polyether type).The cleaned and dried foam pieces were soaked in a chloroform solution of HMDC (1.5 g of HMDC for 1 g of PU) and then squeezed occasionally until all the solution was retained in the PU cubes. Finally, the organic solvent was evaporated by drying the foam in air. The dithiocarbamate-loaded PU foam pieces were stored in a dark borosilicate glass container. Column Preparation The HMDC-loaded PU foam pieces (approximately 0.3 g) were transferred into a disposable syringe after packing a small wad of cotton-wool in the end of the column and covering it with a disc of filter-paper.The sample solution was passed through the sorbent column using a peristaltic pump as described previously. 13 General Procedure for ICP-AES Measurements Aqueous test solutions containing 2 mg of As, 5 mg of Bi, 2 mg of Sb, 3 mg of Se and 2 mg of Sn were pumped through the sorbent column at a flow rate of 2 ml min21. Dilute NH3 solution (1 + 1) and HCl (1 + 1) were used for adjusting of the pH.The physically immobilized organic chelating ligand, HMDC, together with the analyte–dithiocarbamate complexes formed were completely dissolved in methanol. For this purpose, 3.0 ml of pure methanol were placed in a small, dry quartz beaker and was passed five consecutive times through the sorbent with the aid of the syringe plunger. The methanol solution was then diluted with 3.0 ml of doubly distilled water and nebulized into the argon plasma of the ICP-AES system.The emission intensity signals of the elements in this solution were measured against calibration standard solutions in order to evaluate the sorption efficiency. For each sample a new sorption column was applied. General Procedure for ETAAS Measurements Aqueous test solutions containing 50 ng of As, Bi, Sb and Sn, 100 ng of Se and 400 ng of Hg were pumped through the sorbent column at a flow rate of 2 ml min21 after adjusting of the pH with dilute NH3 and HCl. Elution was performed with 1.0 ml of IBMK as described for the ICP-AES technique.The atomic absorption signals of the elements in this IBMK solution were measured against organic calibration standard solutions. For each experiment a new column was prepared. Results and Discussion Influence of pH The results for the influence of the pH of the water sample solution on the solid-phase extraction preconcentration with PU–HMDC sorbent are presented in Table 2. AsIII, BiIII, HgII, SbIII and SeIV form stable dithiocarbamate complexes with HMDC in the solid state in the pH range 1–5 and SnIV in the pH range 4–5.However, AsV, SbV and SeVI do not form a complex Table 1 Temperature programme for the determination of analytes in sorbent extraction eluates (10 ml IBMK solution) using a pyrolytic graphite platform and uncoated graphite tube Step Element Parameter I II III IV Temperature/°C 120 Variable Variable Variable Ramp time/s 10 Variable Variable 2 Hold time/s 15 Variable Variable 3 Read — — On — Gas flow rate/ml min21 300 300 0 300 As* Temperature/°C 1200 2200 2500 Ramp/s 12 0 Hold/s 15 3 Bi Temperature/°C 700 1100 1500 Ramp/s 10 0 Hold/s 20 5 Hg Temperature/°C 250 900 1600 Ramp/s 5 1 Hold/s 10 14 Sb Temperature/°C 1100 1600 2200 Ramp/s 12 0 Hold/s 15 5 Se* Temperature/°C 1100 2100 2400 Ramp/s 12 0 Hold/s 15 3 Sn* Temperature/°C 1300 2300 2500 Ramp/s 13 0 Hold/s 10 4 * Addition of 10 ml of modifier solution [10 mg ml21 Pd as PdCl4(MTOA)2 in IBMK].Table 2 Effect of pH on the recovery of trace elements in the analysis of water samples after solid-phase extraction preconcentration on polyurethane –HMDC (n = 3 parallel determinations) Recovery ± s (%) pH AsIII BiIII HgII SbIII SeIV SnIV 1 97 ± 2 98 ± 1 99 ± 2 98 ± 2 98 ± 2 17 ± 3 2 94 ± 4 99 ± 2 99 ± 2 95 ± 3 94 ± 2 28 ± 6 3 95 ± 3 99 ± 2 99 ± 2 97 ± 2 96 ± 2 62 ± 5 4 99 ± 2 100 ± 2 100 ± 2 99 ± 2 99 ± 2 94 ± 3 5 95 ± 2 100 ± 2 100 ± 2 99 ± 2 97 ± 2 97 ± 3 6 62 ± 5 99 ± 2 100 ± 2 99 ± 2 70 ± 6 58 ± 4 7 < 5 99 ± 2 99 ± 2 99 ± 2 < 5 15 ± 5 244 Analyst, March 1997, Vol. 122with HMDC under any conditions. A reduction step must therefore precede the passage through the sorbent column. For AsIII, BiIII, HgII, SbIII and SeIV, sorption recoveries > 98% can also be achieved for water samples acidified with HCl (0.1–0.7 mol l21 HCl in the solution). Hence when the determination of tin is not required, the water sample can simply be acidified with HCl before preconcentration instead of adjusting the pH to 4.5 ± 0.5 with dilute ammonia solution (1 + 1) or HCl (1 + 1).Influence of Volume of Water Sample Solution Water sample solution volumes up to 150 ml permit the achievement of quantitative solid-phase extraction preconcentration of all the analytes studied on PU–HMDC sorbent. Some decrease in the recovery for As, Sb, Se and Sn was observed when the sample volume was increased from 150 to 200 ml. The sorption for Bi and Hg is quantitative for a 200 ml sample volume.Hence the enrichment factor achieved for the proposed sorbent preconcentration ETAAS method is 150 (elution of the sorbed analyte–dithiocarbamate complexes with 1.0 ml of IBMK). The necessity to perform the ICP-AES measurements in at least 5.0 ml of methanol eluates leads to an enrichment factor of only 30. Influence of Sodium Chloride Concentration The influence of the concentration of NaCl on the efficiency of sorption on PU–HMDC was investigated in order to elucidate the possibilities of determining the studied analytes in sea-water samples using the proposed preconcentration procedure.The results obtained are presented in Table 3. The retention of Bi and Hg on the PU–HMDC sorbent is quantitative for all water sample volumes and sodium chloride concentrations studied. For the simultaneous quantitative preconcentration of all the analytes studied from a 150 ml water sample, the concentration of NaCl in the solution pumped through the PU–HMDC sorbent column must not exceed 3.5%.Analysis of Water Samples For the determination of trace amounts of As, Bi, Hg, Sb, Se and Sn in water samples, 3 g of KI were dissolved in 150.0 ml of sample solution and the pH was adjusted to 4.5 ± 0.5 with dilute ammonia solution (1 + 1) or HCl (1 + 1). When the determination of Sn is not required, the sample (150.0 ml) after addition of 3 g of KI is acidified with 1.5 ml of concentrated HCl.The sample is heated gently at 60 °C for 1 h to ensure the reduction of arsenic, antimony and selenium to their lower oxidation state. After cooling to room temperature, the sample is pumped through the sorption column at a flow rate of 2 ml min21, then the column is washed with 5 ml of doubly distilled water. In this state, prior to elution, the sorbent columns are suitable for conservation and transportation. The elution and the measurements by ICP-AES and ETAAS were performed as described in the general procedure.Absolute blanks of the proposed procedure are presented in Table 4. The largest part of the blank signal, especially for Bi and Sn, is due to the KI, and also to the ligand HMDC. The limits of detection of each element, expressed as the blank value ± three times the standard deviation of the blank, are reported in Table 5. With the exception of Hg, the solid-phase extraction preconcentration ETAAS method ensures lower detection limits owing to the higher enrichment factor and the higher sensitivity of the graphite furnace measurement technique.The method using ICP does not offer any advantage over hydride generation ICP-AES. The 30-fold analyte preconcentration does not compensate for the poorer detection limit of the Table 3 Effect of the water sample volume and NaCl concentration (synthetic mixtures) on the solid-phase extraction recovery (pH 4.3) Recovery ± s (%) Volume*/ NaCl ml (%) AsIII BiIII HgII SbIII SeIV SnIV 100 (n = 5) 3.5 99 ± 2 100 ± 2 100 ± 2 99 ± 2 99 ± 2 95 ± 6 100 (n = 5) 5 97 ± 2 98 ± 1 99 ± 2 98 ± 2 98 ± 2 93 ± 3 150 (n = 4) 3.5 99 ± 2 99 ± 2 99 ± 2 99 ± 2 99 ± 2 95 ± 4 150 (n = 4) 5 87 ± 6 98 ± 2 99 ± 2 90 ± 5 91 ± 5 90 ± 7 * n = Number of parallel determinations. Table 6 Recovery of analytes added to sea-water (n = 4) Solid-phase extraction–ICP-AES Solid-phase extraction Added/ Found ±s/ Recovery ±s Added/ Found ±s/ Recovery ±s Element mg l21 mg l21 (%) mg l21 mg l21 (%) As 62.5 61.4 ± 2.8 98.2 ± 4.5 0.666 0.648 ± 0.042 97.3 ± 6.3 Bi 62.5 64.0 ± 3.5 102.4 ± 5.6 0.666 0.668 ± 0.026 100.3 ± 3.9 Hg 25.0 25.8 ± 1.6 103.2 ± 6.4 2.67 2.74 ± 0.07 102.6 ± 2.6 Sb 62.5 60.8 ± 3.2 97.3 ± 5.1 0.666 0.644 ± 0.028 96.7 ± 4.2 Se 125.0 122 ± 5 97.6 ± 4.0 1.33 1.26 ± 0.08 94 ± 6 Sn 125.0 127 ± 8 101.6 ± 6.4 0.666 0.687 ± 0.034 103.2 ± 5.1 Table 4 Reagent blanks for preconcentration procedure. Doubly distilled water (150 ml) + 3 g of dissolved KI was pumped though 0.30 g of sorbent containing 0.15 g of HMDC; n = 5 Element Blank ± s/ng As 4.5 ± 0.5 Bi 12.9 ± 0.6 Hg 6.8 ± 0.4 Sb 4.7 ± 0.4 Se < detection limit Sn 10.8 ± 1.2 Table 5 Detection limits (3s) of polyurethane-HMDC preconcentration procedure for trace determination in waters by ICP-AES and ETAAS Element ICP-AES*/mg l21 ETAAS*/mg l21 As 3.2 0.06 Bi 8 0.1 Hg 0.12 0.3 Sb 2 0.06 Se 3 0.08 Sn 6 0.1 * Elution of the retained on the sorbent analytes (150 ml water sample pumped though the sorbent column) in 2.5 ml of methanol + 2.5 ml doubly distilled water for ICP-AES and in 1 ml IBMK for ETAAS measurements.Analyst, March 1997, Vol. 122 245direct simultaneous ICP-AES measurement technique additionally owing to the lower analytical potential of the ICP method in methanol solutions than in aqueous solutions. However, the combination of the described preconcentration procedure with hydride generation in methanol solutions of the analyte– dithiocarbamate complexes and subsequent simultaneous ICPAES determination of As, Bi, Hg, Sb, Se and Sn leads to a substantial improvement in sensitivity.16 The accuracy of the procedure was investigated by determining the analyte content in spiked sea-water (Black Sea-water with analyte concentrations lower than the detection limit of the method).The results (Table 6) show sufficiently high recoveries and a precision (RSD) of 3–7%. The precision is very good despite working close to the detection limit.In addition, the accuracy and precision of the described methods were evaluated by comparing the results for waste waters with those obtained by direct ICP-AES measurements after hydride generation (Table 7). No significant differences in the results obtained were observed. Conclusion Solid-phase extraction on a PU–HMDC sorbent column is an effective preconcentration procedure for simple storage, transportation and measurement of trace levels of As, Bi, Hg, Sb, Se and Sn in various kinds of water samples using ICP-AES or ETAAS. The combination of the preconcentration procedure with the ETAAS measurement technique offers higher enrichment factors and higher sensitivity. The authors thank the Bulgarian Foundation of Sciences, Project X-519. References 1 Braun, T., Fresenius’ Z. Anal. Chem., 1983, 314, 652. 2 Burba, P., Rocha, J. C., and Schulte, A., Fresenius’ J. Anal. Chem., 1993, 346, 414. 3 Jambor, J., and Javorek, T., Collect. Czech. Chem. Commun., 1993, 58, 1821. 4 Prakash, N., Csanady, G. J., Michaelis, M., and Knapp, G., Mikrochim. Acta, 1989, 3, 257. 5 Maloney, M. P., Moody, G. J., and Thomas, J. D. R., Analyst, 1980, 105, 1087. 6 Horvath, Z., Lastztity, A., Szakacs, O., and Bozsai, G., Anal. Chim. Acta, 1985, 173, 273. 7 Terada, K., Matsumoto, K., and Nanao, Y., Anal. Sci., 1985, 1, 145. 8 Tong, A., and Akama, Y., Anal. Chim. Acta, 1990, 230, 175. 9 Tong, A., Akama, Y., and Tanaka, S., Analyst, 1990, 115, 947. 10 Yebra-Biurrun, M. C., Bermejo-Barrera, A., and Bermejo-Barrera, P., Mikrochim. Acta, 1992, 109, 243. 11 Schilling, T., Schramel, P., Michalke, B., and Knapp, G., Mikrochim. Acta, 1994, 116, 83. 12 Kantipuly, C., Katragadda, S., Chow, A., and Gesser, H., Talanta, 1990, 37, 491. 13 Alexandrova, A., and Arpadjan, S., Analyst, 1993, 118, 1309. 14 Alexandrova, A., and Arpadjan, S., Anal. Chim. Acta, 1995, 307, 71. 15 Sperling, M., Yin, X., and Welz, B., Analyst, 1992, 117, 629. 16 Vuchkova, L., and Arpadjan, S., Talanta, 1996, 43, 479. Paper 6/06917G Received October 9, 1996 Accepted December 3, 1996 Table 7 Results (mg l21 ± s) for the analysis of waste water (n = 4) PU–HMDC HG-ICP in aqueous Element ETAAS ICP-AES solution As 4.3 ± 0.2 4.5 ± 0.3 4.3 ± 0.2 Bi 6.7 ± 0.4 < 8 6.9 ± 0.4 Hg < 0.4 0.28 ± 0.03 0.28 ± 0.02 Se 5.8 ± 0.4 5.6 ± 0.4 5.9 ± 0.3 Sb 8.2 ± 0.4 8.3 ± 0.5 8.1 ± 0.4 Sn 9.4 ± 0.5 9.2 ± 0.6 9.5 ± 0.5 246 Analyst, March 1997, Vol. 122

ISSN:0003-2654    

DOI:10.1039/a606917g

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Direct Determination of Amiloride in Urine Using Isopotential Fluorimetry Jos�e A. Murillo Pulgar�ýn*, Aurelia Ala �n�on Molina and Pablo Fern�andez L�opez Department of Analytical Chemistry and Foods Technology, University of Castilla La Mancha, 13071 Ciudad Real, Spain A method for the determination of amiloride at concentrations between 15 and 152 ng ml21 by means of matrix isopotential synchronous fluorescence spectrometry and derivative techniques is proposed. This method is useful for the determination of compounds in samples with unknown background fluorescence without the need for tedious pre-separation.As amiloride is widely used as a doping substance in sport, the method was successfully applied to the determination of amiloride in urine. To obtain maximum sensitivity and adequate selectivity, factors affecting fluorescence intensity were studied in the amiloride band centered at lex = 362 nm and lem = 415 nm. As a result, the determination was performed in an ethanol–water (1 + 1, v/v) medium at pH 6.3, adjusted by using sodium citrate–citric acid (0.1 M) as buffer solution.The concentration of amiloride in urine samples can be calculated by recording its total luminescence spectrum and applying the isopotential trajectory of the urine that cuts the selected band of amiloride. The unknown analytical signal of urine is eliminated in the MISF spectrum obtained, by means of its first derivative. A calibration graph was constructed by measuring first derivative values at lex = 357 nm and lem = 392 nm.Analytical parameters of the proposed method were calculated according to the error propagation theory. The sensitivity, repeatability, reproducibility and limit of determination achieved with the proposed method are adequate for the determination of amiloride in urine. Keywords: Amiloride; urine; spectrofluorimetry, diuretics Diuretics are defined as substances that produce an increase in the urinary elimination of sodium bound to an anion and water, leading to a reduction in extracellular liquids.Amiloride (Namidino- 3,5-diamino-6-chloropyrazine-2-carboxamide) is a weak diuretic, thrifty of potassium, since it inhibits the sodium– potassium interchange in kidneys by waterproofing the distal membrane to promote sodium excretion and potassium reabsorption. This natriuretic agent is widely used in the treatment of several diseases. It can also be applied as a doping substance. In sports, diuretics are abused mainly for two reasons.The first is to obtain a rapid diminution of corporal weight, which is important in sports that are divided into different weight categories. The second is to reduce the concentration of medical drugs in urine by diluting the latter by means of the rapid production of an elevated volume of urine, leading to a smaller possibility of detecting other doping substances. An advantage in the use of amiloride is that low doses lead to high-volume urine excretion, obstructing its determination and therefore highly sensitive methods are required.Further, amiloride is potassium-sparing that is essential for energy production. It is known that potassium is lost by the muscles during ischaemia. This loss reduces the rate of glucolysis owing to the dependence of quinases on potassium to reach maximum activity. Nevertheless, no medical reason can justify a rapid decrease of weight in any sport.Additionally, this abuse causes grave dangers to health because of possible serious secondary effects. Owing to the uncontrollable use of amiloride, the International Olympics Committee, since 1990 has included it in the list of forbidden substances.1 Consequently, there is a need for the development of effective methods for determining this doping substance. Here we propose a fast, simple, and sensitive fluorimetric method for determining amiloride in urine samples, reaching the same sensitivity as direct determination but without the need for prior separation and concentration or derivatization procedures.The therapeutic and doping dose of amiloride varies from 5 to 20 mg daily (one administration only). It is incompletely absorbed and it does not appear to be metabolized. The half-life in plasma varies from 6 to 10 h. About 50% of an oral dose is excreted in the unchanged form in urine and 40% is eliminated in the faeces within 72 h.2 Consequently, the determination of amiloride in urine demands highly sensitive methods.Amiloride is a pale yellow to greenish yellow powder that exhibits violet fluorescence in solution. There have been only a few reports on the determination of amiloride in tablets3–6 or in biological fluids.7,8 Normally the determination of amiloride at therapeutic levels by liquid chromatography requires various tedious preliminary procedures, such as extraction and preconcentration in an organic solvent.This causes many disadvantages (such as low recoveries), since all these procedures are based on equilibrium reactions. We have previously reported a fluorimetric synchronous technique called matrix isopotential synchronous fluorescence spectrometry (MISF),9 which is particularly useful for removing fluorescence matrix background effects and allows the determination of individual compounds in complex samples. It is essential that the matrix has an almost invariable composition. It is possible to maintain a constant background signal, even though its fluorescence intensity may vary, if a cut is made in the total fluorescence spectrum following one of the trajectories that joins points of equal intensity (isopotential trajectory) from an initial point to final excitation and emission wavelengths.This trajectory is obtained by means of a program developed in BASIC.10 It is always possible to find the matrix trajectory that passes through the maximum fluorescence excitation and emission wavelengths of the component to be determined.Therefore, the same sensitivity is achieved as in direct determinations in the absence of background fluorescence. This technique can be improved by applying derivative methods. Human urine is composed of numerous organic substances, most of which give a high absorbance in the ultraviolet region. A small number of these organic substances present fluorescence. 11 Owing to the high absorbance values in the ultraviolet region, the fluorescence intensity of the substance of interest (amiloride) does not vary linearly with concentration (A < 0.05).Consequently, it is recommended to measure fluorescence intensity at the maxima of the higher excitation wavelength and also to make an appropriate dilution of the urine. Additionally, as some of the urine components are Analyst, March 1997, Vol. 122 (247–252) 247fluorescent, they provide high background fluorescence and, as a result, interfere with the direct determination of amiloride.Recent work in our laboratory has concerned the application of this fluorimetric technique to the determination of different drugs in biological fluids with excellent results.9,12–17 Experimental Apparatus All fluorimetric measurements were performed on an Aminco Bowman Series 2 instrument equipped with a continuous 150 W xenon lamp, connected to software which runs on the OS2 operating system.Quartz glass cuvettes with a pathlength of 1.0 3 1.0 cm were used. Thermostatic equipment and a Crison Model 2001 pH meter with a glass–saturated calomel combination electrode and a Selecta Mixtaxel centrifuge were also used. Software A program was developed which enabled us to obtain the values of lex and lem for any constant value of the fluorescence intensity from a three-dimensional spectrum. As the values obtained for a particular curve were not equidistant, the Lagrange interpolation method was applied to all points, which were placed in order, using emission wavelengths at 0.4 nm intervals.Once the trajectory had been defined, the spectrum was obtained by means of the Ftotal program.10 The spectra obtained from it displayed the same format as those obtained directly with the Aminco Bowman Series 2 spectrofluorimeter. The Ftotal program not only generates information on a fluorescent compound through the isomentation of the three-dimensional spectrum as a level curve, but also processes the spectral data to obtain any type of spectrum of socalled new fluorimetric techniques.The statistical analysis is totally covered by means of a program developed by us. This program has a menu that includes procedures such as least median of squares regression (detection of outlier and leverage points), least squares regression with replicates or not, weighted least squares regression, tests of regression and correlation, detection and determination limits, ellipse graph for the 95% confidence region for the true slope and intercept on the ordinate estimated from the regression method, dispersion and confidence bands for the calibration graph and ANOVA test for linearity and for comparison of several regression lines.18–22 Reagents All experiments were performed with analytical reagent grade chemicals, pure solvent and Milli-Q-purified water.A stock standard solution of amiloride (Aldrich, Milwaukee, WI, USA) (250 mg dissolved in 1 l of water) was diluted to prepare suitable working standard solutions.The stock standard solution of amiloride was stored protected from light and maintained below 5 °C. Under these conditions, it was stable for 2 months. The working standard solutions of amiloride were stable for at least 2 h at room temperature. A 1.0 m buffer solution of pH 6.3 was prepared by mixing appropriate amounts of citric acid with sodium hydroxide.Urine samples were obtained from fasting, healthy people in the morning. Procedure Centrifuge urine for 15 min at 3800 rpm, transfer 10 ml of the clear supernatant solution into a 100 ml calibrated flask and dilute to volume with water. Store and maintain below 5 °C. For the preparation of the calibration graph, place an aliquot of amiloride equivalent to 500–5000 ng in a 25 ml calibrated flask, add 2.5 ml of buffer solution (pH 6.3), 12.5 ml of ethanol and 5 ml of a solution of urine free from amiloride and dilute to volume with water.Record 61 emission spectra of 192 nm width in steps of 0.4 nm, varying the excitation wavelength in 3.2 nm steps. Obtain the total luminescence spectra. Select a suitable trajectory which passes through the excitation and emission maxima of amiloride and obtain matrix isopotential synchronous spectra by means of Ftotal.10 Calculate the first derivative according to the Savitzky and Golay algorithm.23,24 Finally, determine the amiloride content by measuring the derivative signal at the emission wavelength of 392.0 nm and using the appropriate calibration graph.Results and Discussion Factors Affecting Fluorescence Intensity Chemical variables were studied to obtain the best measurement conditions and maximum fluorescence sensitivity. The influence of pH on the fluorescence intensity was studied by adding different amounts of HCl and NaOH to a urine solution and an amiloride solution.As can be seen in Fig. 1(a), the fluorescence intensity of amiloride is nearly constant at pH values between 2 and 7, whereas that due to urine decreases slowly from pH 4 to 8. A pH 6.3 was selected; this is approximately the pH value of urine of healthy people. The pH selected was adjusted by adding sodium citrate–citric acid buffer solution. The fluorescence intensity of amiloride and urine was not affected by the buffer and its concentration. A 0.1 m concentration of the buffer was therefore selected to obtain an adequate buffering capacity.Amiloride is highly soluble in water, so it was not necessary to use an organic solvent. Nevertheless, it was desirable to study the variation of amiloride and urine fluorescence with changes in the dielectric constant of the medium. Owing to the physical and chemical properties of ethanol, it is suitable for such a study. Hence the effect of ethanol content in the medium was investigated by preparing samples of amiloride and urine with ethanol concentrations between 0 and 100% v/v for the amiloride solutions and between 0 and 80% v/v for the urine solutions (when the ethanol content is more than 80% some of the urine components are precipitated).The fluorescence intensity due to amiloride increases when the ethanol content in the medium increases whereas that due to urine increases slowly [Fig. 1(b)]. For these reasons, it is preferred to use a high ethanol Fig. 1 (a) Influence of the pH on fluorescence intensity of urine (.) and amiloride (5), measuring emission at 414.8 nm after excitation at 362 nm. Urine diluted 1 : 50; concentration of amiloride, 380 ng ml21; detector voltage, 500 V.(b) Influence of ethanol concentration on fluorescence intensity of urine and amiloride, measuring emission at 414.8 nm after excitation at 362 nm. Urine diluted 1 : 50; concentration of amiloride, 380 ng ml21; detector voltage, 450 V. 248 Analyst, March 1997, Vol. 122percentage but with the precaution that precipitation does not occur. Therefore, 50% of ethanol was selected. Another factor that affects the fluorescence intensity is temperature. In both cases, fluorescence intensity decreased when the temperature was increased from 7 to 84 °C. Under the experimental conditions selected above, the temperature coefficients are about 0.99% °C21 for amiloride and about 0.59% °C21 for urine. This effect can be explained by a higher internal conversion as temperature increases, simplifying nonradiative deactivation of the excited singlet state.25 Therefore, the use of a thermostat is recommended and a measurement temperature of 20 °C was chosen.The influence of amiloride concentration on the fluorescence intensity was studied under the above conditions. The best range of pure amiloride concentration for the relationship between fluorescence intensity and concentration was found to be up to 152 ng ml21. Determination of Amiloride in Urine Previous experiments9,11–17 have shown that the qualitative composition of the fluorescent metabolites of urine from healthy people of both sexes and different diets, aged between 25 and 35 years, which is the usual age of people taking doping substances, is almost constant, a necessary condition for the application of the proposed technique.Different samples of urine display the same type of fluorescence, with hardly any variation in the form of the spectrum and the location of the fluorescence maxima, although it is possible to observe some variations in their intensity.All fluorimetric three-dimensional spectra were obtained by varying the emission wavelength from 310.0 to 502.0 nm and the excitation wavelength from 220.8 to 412.8 nm. Both excitation and emission bandpasses were 8 nm and the scan rate was 50 nm s21. Under these conditions, the time taken to determine contour spectra was 292 s. Matrix isopotential synchronous fluorescence was applied to the determination of amiloride in urine, a very fluorescent matrix, under the optimum conditions established above.As Fig. 2 shows, the fluorescence maxima characteristic of amiloride (solid line) are located in the UV region. As can readily be observed in this region, urine (broken line) shows two bands that prevent the determination of this compound without prior separation. Amiloride shows two broad peaks that are located at lex = 284.8 nm, lem = 412.8 nm and lex = 364.8 nm, lem = 412.8 nm.As at an excitation wavelength of 284.8 nm urine gives high absorbance values and there are more interferences, it is preferable to determine amiloride in urine samples with an excitation wavelength of 364.8 nm where urine needs only minor dilution and adequate selectivity is obtained. The isopotential trajectory for the urine spectrum also shown in Fig. 2 and, as can be observed, the trajectory A passes through the excitation and emission maxima of amiloride (lex = 364.8 nm, lem = 412.8 nm).Trajectory A was obtained from the spectrum corresponding to the arithmetic mean of 10 total fluorescence spectra of the different urine samples by means of the Ftotal program.10 Urine sample solutions containing amiloride gave signals smaller than those obtained with aqueous standard solutions, owing to some type of binding with other components of the urine. Total luminescence spectra of amiloride were obtained in different urine samples to construct calibration graphs and to carry out recovery experiments.Fig. 3 shows the effect of background fluorescence intensity (urine) on amiloride MISF and derivative spectra. It can be seen that the spectra are identical in form although their intensities are different by constant terms. They correspond to the values of the fluorescence intensity due to the three urines in the isopotential trajectory applied. It is easy to observe that derivatization totally removes the background effect.The first derivative of MISF spectra was applied to all samples. The number of points through which the derivative was obtained was optimized, with the conclusion that derivative spectra with a suitable signal-to-noise ratio were obtained with 25 points. In the same way we obtained the total luminescence spectra of amiloride in aqueous solution at the same concentrations. The MISF spectra were obtained by applying the isopotential trajectory (Fig. 2). We also obtained their first derivatives, as with the urine samples. As can be readily seen in Fig. 4(a), which shows the MISF spectra of amiloride obtained in urine, it is not possible to determine this drug by measuring the maximum fluorescence Fig. 2 Total fluorescence spectra of 152 ng ml21 of amiloride (solid line) and diluted urine (1 : 50) (broken line). The line labelled A is the isopotential trajectory for the urine. Analyst, March 1997, Vol. 122 249intensity (which is located at lem = 412.8 nm) with regard to the initial or final extremes (lem = 340.0 nm and lem = 480.0 nm) of the selected trajectory, since the fluorescence intensity does not reach a constant value. The first derivative technique was therefore applied. Calibration graphs were constructed by measuring the first derivative at lex = 357.0 nm lem = 392.0 nm, where maximum sensitivity is achieved. Fig. 4(b) shows the spectra derived from the calibration of amiloride in urine.In order to test the independence of the analytical signal of amiloride, i.e., to show that the signal measured is independent of the urine, three calibration graphs from the first derivative signals were constructed with different urine samples. The proposed method was evaluated by a statistical analysis of experimental data by fitting the least squares line according to y = a + bx, after discarding outliers with help of the least median of squares regression (LMS).21 As LMS is a robust regression method, it is able to detect outlier points.These outliers cause errors in the true line when experimental data are fitted according to the least squares regression. Although the LMS method considers outliers to the points that give an absolute value of standardized residual above 2.5, owing to the robustness of the LMS fit, we only reject points that give an absolute value of standardized residual higher than 10. No outliers were detected. Table 1 shows the results of the statistical analysis.To verify if the intercepts on the ordinate were negligible, significances were studied by applying Student’s t-test at the 95% confidence level and suitable degrees of freedom.19,20 If the intercepts on the ordinate for the lines calculated by the least Fig. 3 (a) Set of MISF spectra of 122 ng ml21 of amiloride in different urine samples. (b) Corresponding first derivative spectra. Fig. 4 (a) Set of MISF spectra of amiloride in urine.Amiloride concentrations: (1) 20; (2) 30; (3) 40; (4) 60; (5) 80; (6) 120; (7) 160; and (8) 200 ng ml21. (b) Corresponding first derivative spectra. Table 1 Statistical data for the determination of amiloride in urine by means of first derivative matrix isopotential synchronous fluorescence Slope Deter- without Fitting mination Intercept on Confidence interval Confidence interval Exp. SD of intercept, used* Sample coefficient ordinate, a Slope, b for intercept for slope t-test estimation bA 1 Urine I 0.9997 21.750 3 1024 4.056 3 1024 — — — 7.2 3 1024 4.040 3 1024 Urine II 0.9997 2.037 3 1023 3.733 3 1024 — — — 5.1 3 1024 4.190 3 1024 Urine III 0.9999 21.610 3 1023 3.981 3 1024 — — — 1.4 3 1024 3.631 3 1024 2 Urine I 0.9967 25.244 3 1024 4.070 3 1024 1.247 3 1023 to 4.286 3 1024– 0.725 1.2 3 1023 4.018 3 1024 22.295 3 1023 3.854 3 1024 Urine II 0.9974 1.233 3 1023 3.961 3 1024 2.764 3 1023 to 4.148 3 1024– 1.970 1.0 3 1023 4.085 3 1024 22.985 3 1024 3.775 3 1024 Urine III 0.9995 22.134 3 1023 4.172 3 1024 21.421 3 1023 to 4.259 3 1024– 7.323b 4.7 3 1024 3.958 3 1024 22.846 3 1023 4.086 3 1024 3 Global 0.9958 24.751 3 1024 4.068 3 1024 4.579 3 1024 to 4.182 3 1024– 1.056 1.2 3 1023 4.020 3 1024 21.408 3 1023 3.954 3 1022 * 1 = Least median of squares regression line according to y = a + bx and y = bAx. 2 = Least squares regression line according to y = a + bx (theoretical t = 2.447) and y = bAx. 3 = Overall least squares regression line according to y = a + bx (theoretical t = 2.074) and y = bAx.b Experimental t-value is greater than the theoretical t-value and therefore the slope calculated without an intercept is not reliable. 250 Analyst, March 1997, Vol. 122squares technique are negligible, it is necessary to perform the least squares regression again according to the function y = bAx and then the new value of the slopes of calibration graphs (bA) may be calculated.As Table 1 shows, the intercept on the ordinate is significantly different from zero for the true line of urine III and consequently the fit according to y = bAx is not recommended; nevertheless, the value of the slope for this regression is also given in Table 1. To obtain the most representative calibration graph, an overall least squares was developed. The regression line shows homoscedasticity19,20 and consequently the residuals manifest a uniform variance (the errors of measurements are independent of the amiloride concentration), it being unnecessary to weight the first derivative values according to the mean standard deviation.When the 95% confidence region for true slope and intercept19,20 estimated is represented, the zero intercept on the ordinate falls within the joint confidence region. This means that the intercept is not significantly different from zero. The confidence interval for the corresponding slope is 3.946 3 1024– 4.095 3 1024.Nevertheless, the intercept significance was tested by applying Student’s t-test. The result was not significantly different from zero, and therefore it is possible to establish a proportional relationship between the analytical signal and amiloride concentration. Therefore, the overall least squares fit according to y = bAx was performed. When the first derivative of matrix isopotential synchronous spectrofluorimetry was used and the overall least squares regression line according to y = a + bx was considered, a detection limit of 2.78 ng ml21 and a determination limit of 9.26 ng ml21 were obtained on applying the IUPAC21,22 definition, in which only the standard deviation of the blank is assumed.The propagation of errors approach will give values of detection and determination limits consistent with the reliability of the blank measurements and the signal measurements of the standards.21,22 In this case a detection limit of 4.45 ng ml21 and a determination limit of 14.82 ng ml21 were obtained.Precision, Accuracy and Interference Studies To study the precision of the method, a series of 10 solutions at three different concentrations were prepared containing 15, 61 and 152 ng ml21 of amiloride in urine and they were measured 15 times on the same day (repeatability study) and once a day for 10 d (reproducibility study). The SEs, assuming the error propagation theory, and RSD of repeatability at each of these levels were 2.2, 2.1 and 2.4 ng ml21 and 14.4, 3.5 and 1.6%, respectively.The SEs and RSDs of reproducibility at each of these levels were 2.0, 1.9, and 2.2 ng ml21 and 12.5, 3.1 and 1.4%, respectively. It can be seen that the SEs are approximately constant with the concentration, which is in agreement with the homoscedasticity that the calibration line presents. The accuracy of the method was established by testing the analytical signal corresponding to three replicates of three samples of 15, 61 and 152 ng ml21 of amiloride in different urine samples.The recoveries obtained were 107.0, 100.1 and 101.6%, respectively. A comparison study of six samples of amiloride in urine at therapeutic levels was performed by means of MISF and the currently accepted method for doping substances (HPLC screening and GC–MS confirmation). The concentrations calculated in reverse by both methods are given in Table 2. As both methods should give same concentrations for the same samples, if least squares pair regression is applied at these concentrations, zero intercept on the ordinate and a slope of unity must be obtained. In order to test whether these methods are significantly different, the confidence intervals for the parameters of the linear regression model must be studied together.If the intercept on the ordinate, a, is plotted against the slope, b, for repeated random samples the points will be found to fall elliptically about the true centre (a, b) and, conversely, any confidence interval for the true combination of a and b will take the form of an elliptical region about the best estimates (a, b) as the centre.19 Fig. 5 shows the 95% confidence region for the true slope and estimated intercept. As can be observed, the point corresponding to the zero intercept and unit slope falls within the joint confidence region. This means that the accuracies of the proposed and currently accepted method are not significantly different.The specificity was studied by adding some closely related drugs to urine and testing to see if they caused interference in the amiloride quantification. The amiloride concentration was 152 ng ml21 and the concentration of the drugs added (atenolol, furosemide, dipyridamole, aspirin, bumetanide, hydrochlorothiazide, hydroflumethazide, caffeine, metoprolol, quinidine and quinine) was slightly higher than would be expected at therapeutic doses. Except with quinidine and quinine, no significant variation of the analytical signal was observed from the value expected when amiloride was present alone.Nevertheless, quinidine and quinine do not present interferences if the calibration is performed at the minimum of the first derivative from the matrix isopotential synchronous spectrum that is located at lex = 363.2 nm and lem = 434.0 nm. Conclusions A method for the direct fluorimetric determination of amiloride in urine by matrix isopotential synchronous fluorescence, without the need for prior separation, has been described.The determination of this natriuretic drug in urine can be performed at lex = 357.0 nm and lem = 392.0 nm in the first derivative of the matrix isopotential synchronous scan. As can be readily observed from the detection limits, only a 63% of error is caused by the standard deviation of the blank. Table 2 Concentrations (ng ml21) of amiloride in urine samples calculated in reverse using the currently accepted method and the proposed method XAccepted YProposed 14 15 28 30 58 61 98 91 122 122 158 152 Fig. 5 The ellipse is the 95% confidence region for the true slope and intercept on the ordinate estimated from the overall least squares regression between the concentration calculated in reverse by the proposed MISF method and the currently accepted method for the determination of doping substances. The point (0,1) corresponds to zero intercept and unit slope. This point falls within the joint confidence region and consequently there are no significant differences in the accuracy of the two methods.Analyst, March 1997, Vol. 122 251Therefore, the IUPAC definition with a standard deviation proportionality factor of three is inadequate. An exhaustive statistical analysis was applied to all calibration graphs, including the least median squares robust regression and least squares. Owing to the homoscedasticity of the regression line, the weighted version is not recommended.The significance of the intercept on the ordinate was investigated through the ellipse method and Student’s t-test at the 95% confidence level and was found not to be significantly different from zero. To validate the method, precision, accuracy and interference studies were performed. As can be observed from the SEs and RSDs obtained from the repeatability and reproducibility experiments, excellent precision, better than in other spectrofluorimetric methods, is achieved without pre-separation procedures. 3–8 In a comparison of this method with the currently accepted method, their accuracies were not significantly different. The specificity of the proposed fluorimetric method is demonstrated by the few interferences from some closely related drugs to urine. The authors gratefully acknowledge financial support from the Direcci�on General de Investigaci�on Cient�ýfica y T�ecnica (Project No. PB 94-0743). The authors gratefully thank the Laboratorio de Control del Dopaje del Consejo Superior de Deportes del Ministerio de Educaci�on y Cultura for the determination of amiloride in urine samples by HPLC screening and GC–MS confirmation.References 1 Rodr�ýguez Bueno, C., Dopaje, Interamericana–McGraw-Hill, Madrid, 1992. 2 Clarke’s Isolation and Identification of Drugs, ed. Moffat, A. C., Pharmaceutical Press, London, 1986, p. 339. 3 Kurani, S. P., Desai, D. K., and Seshadrinathan, A. G., Indian Drugs, 1986, 23, 230. 4 Sastry, C. S. P., Suryanarayan, M. V., Tipirnani, A. S. R. P., and Satyanarayana, T., Indian Drugs, 1989, 26, 651. 5 Wahbi, A. A. M., Bedair, M. M., Galal, S. M., and Gazy, A. A., S. T. P. Pharm. Sci., 1933, 3(2), 182. 6 Garc�ýa Sanch�ez, F., Fern�andez Gutierrez, A., and Cruces Blanco, C., Anal. Chim. Acta, 1995, 306, 313. 7 Karola, R., Knauf, H., and Ernst, M., J. Chromatogr., 1982, 233, 432. 8 Reewijk, H. J. E. M., Tjaden, U. R., and Van der Greef, J., J.Chromatogr., 1992, 575, 269. 9 Murillo, J. A., and Ala�n�on, A., Anal. Chim. Acta, 1994, 296, 87. 10 Murillo, J. A., and Ala�n�on, A., Comput. Chem., 1993, 17, 34. 11 Leiner, M. J. P., Hubmann, M. R., and Wolfbeis, O. S., Anal. Chim. Acta, 1987, 198, 13. 12 Murillo Pulgar�ýn, J. A., and Ala�n�on, A., Analyst, 1994, 119, 1915. 13 Berzas Nevado, J. J., Murillo Pulgar�ýn, J. A., and G�omez Laguna, M. A., Analyst, 1995, 120, 171. 14 Murillo, J. A., Ala�n�on, A., and Fern�andez, P., Talanta, 1996, 43, 431. 15 Murillo, J. A., and Ala�n�on, A., Anal. Chim. Acta, 1996, 317, 359. 16 Murillo, J. A., Ala�n�on, A., and Fern�andez, P., Anal. Chim. Acta, 1996, 326, 117. 17 Murillo, J. A., and Ala�n�on, A., Microchem. J., 1995, 52, 341. 18 Rousseeuw, P. J., and Leroy, A. M., Robust Regression and Outlier Detection, Wiley, New York, 1987. 19 Lark, P. D., Craven, B. R., and Bosworth, R. C. L., The Handling of Chemical Data, Pergamon Press, Oxford, 1968, ch. 4. 20 Massart, D.L., Vandeginste, B. G. M., Deming, S. N., and Kaufman, L., Chemometrics: a Textbook, Elsevier, Oxford, 1988. 21 Miller, J. N., Analyst, 1991, 116, 3. 22 Gary, L., and Winefordner, J. D., Anal. Chem., 1983, 55, 712. 23 Savitzky, A., and Golay, M. J. E., Anal. Chem., 1964, 36, 1627. 24 Steinier, J., Termonia, Y., and Deltour, J., Anal. Chem., 1972, 44, 1906. 25 Seitz, W. R., in Treatise on Analytical Chemistry, ed. Elving, P. J., Meehan, E. J., and Kolthoff, I. M., Wiley, New York, 1981, pp. 194– 196.Paper 6/07219D Received October 23, 1996 Accepted December 10, 1996 252 Analyst, March 1997, Vol. 122 Direct Determination of Amiloride in Urine Using Isopotential Fluorimetry Jos�e A. Murillo Pulgar�ýn*, Aurelia Ala �n�on Molina and Pablo Fern�andez L�opez Department of Analytical Chemistry and Foods Technology, University of Castilla La Mancha, 13071 Ciudad Real, Spain A method for the determination of amiloride at concentrations between 15 and 152 ng ml21 by means of matrix isopotential synchronous fluorescence spectrometry and derivative techniques is proposed.This method is useful for the determination of compounds in samples with unknown background fluorescency used as a doping substance in sport, the method was successfully applied to the determination of amiloride in urine. To obtain maximum sensitivity and adequate selectivity, factors affecting fluorescence intensity were studied in the amiloride band centered at lex = 362 nm and lem = 415 nm.As a result, the determination was performed in an ethanol–water (1 + 1, v/v) medium at pH 6.3, adjusted by using sodium citrate–citric acid (0.1 M) as buffer solution. The concentration of amiloride in urine samples can be calculated by recording its total luminescence spectrum and applying the isopotential trajectory of the urine that cuts the selected band of amiloride.The unknown analytical signal of urine is eliminated in the MISF spectrum obtained, by means of its first derivative. A calibration graph was constructed by measuring first derivative values at lex = 357 nm and lem = 392 nm. Analytical parameters of the proposed method were calculated according to the error propagation theory. The sensitivity, repeatability, reproducibility and limit of determination achieved with the proposed method are adequate for the determination of amiloride in urine.Keywords: Amiloride; urine; spectrofluorimetry, diuretics Diuretics are defined as substances that produce an increase in the urinary elimination of sodium bound to an anion and water, leading to a reduction in extracellular liquids. Amiloride (Namidino- 3,5-diamino-6-chloropyrazine-2-carboxamide) is a weak diuretic, thrifty of potassium, since it inhibits the sodium– potassium interchange in kidneys by waterproofing the distal membrane to promote sodium excretion and potassium reabsorption. This natriuretic agent is widely used in the treatment of several diseases. It can also be applied as a doping substance.In sports, diuretics are abused mainly for two reasons. The first is to obtain a rapid diminution of corporal weight, which is important in sports that are divided into different weight categories. The second is to reduce the concentration of medical drugs in urine by diluting the latter by means of the rapid production of an elevated volume of urine, leading to a smaller possibility of detecting other doping substances. An advantage in the use of amiloride is that low doses lead to high-volume urine excretion, obstructing its determination and therefore highly sensitive methods are required.Further, amiloride is potassium-sparing that is essential for energy production. It is known that potassium is lost by the muscles during ischaemia. This loss reduces the rate of glucolysis owing to the dependence of quinases on potassium to reach maximum activity.Nevertheless, no medical reason can justify a rapid decrease of weight in any sport. Additionally, this abuse causes grave dangers to health because of possible serious secondary effects. Owing to the uncontrollable use of amiloride, the International Olympics Committee, since 1990 has included it in the list of forbidden substances.1 Consequently, there is a need for the development of effective methods for determining this doping substance.Here we propose a fast, simple, and sensitive fluorimetric method for determining amiloride in urine samples, reaching the same sensitivity as direct determination but without the need for prior separation and concentration or derivatization procedures. The therapeutic and doping dose of amiloride varies from 5 to 20 mg daily (one administration only). It is incompletely absorbed and it does not appear to be metabolized. The half-life in plasma varies from 6 to 10 h.About 50% of an oral dose is excreted in the unchanged form in urine and 40% is eliminated in the faeces within 72 h.2 Consequently, the determination of amiloride in urine demands highly sensitive methods. Amiloride is a pale yellow to greenish yellow powder that exhibits violet fluorescence in solution. There have been only a few reports on the determination of amiloride in tablets3–6 or in biological fluids.7,8 Normally the determination of amiloride at therapeutic levels by liquid chromatography requires various tedious preliminary procedures, such as extraction and preconcentration in an organic solvent.This causes many disadvantages (such as low recoveries), since all these procedures are based on equilibrium reactions. We have previously reported a fluorimetric synchronous technique called matrix isopotential synchronous fluorescence spectrometry (MISF),9 which is particularly useful for removing fluorescence matrix background effects and allows the determination of individual compounds in complex samples.It is essential that the matrix has an almost invariable composition. It is possible to maintain a constant background signal, even though its fluorescence intensity may vary, if a cut is made in the total fluorescence spectrum following one of the trajectories that joins points of equal intensity (isopotential trajectory) from an initial point to final excitation and emission wavelengths.This trajectory is obtained by means of a program developed in BASIC.10 It is always possible to find the matrix trajectory that passes through the maximum fluorescence excitation and emission wavelengths of the component to be determined. Therefore, the same sensitivity is achieved as in direct determinations in the absence of background fluorescence. This technique can be improved by applying derivative methods. Human urine is composed of numerous organic substances, most of which give a high absorbance in the ultraviolet region.A small number of these organic substances present fluorescence. 11 Owing to the high absorbance values in the ultraviolet region, the fluorescence intensity of the substance of interest (amiloride) does not vary linearly with concentration (A < 0.05). Consequently, it is recommended to measure fluorescence intensity at the maxima of the higher excitation wavelength and also to make an appropriate dilution of the urine.Additionally, as some of the urine components are Analyst, March 1997, Vol. 122 (247–252) 247fluorescent, they provide high background fluorescence and, as a result, interfere with the direct determination of amiloride. Recent work in our laboratory has concerned the application of this fluorimetric technique to the determination of different drugs in biological fluids with excellent results.9,12–17 Experimental Apparatus All fluorimetric measurements were performed on an Aminco Bowman Series 2 instrument equipped with a continuous 150 W xenon lamp, connected to software which runs on the OS2 operating system.Quartz glass cuvettes with a pathlength of 1.0 3 1.0 cm were used. Thermostatic equipment and a Crison Model 2001 pH meter with a glass–saturated calomel combination electrode and a Selecta Mixtaxel centrifuge were also used. Software A program was developed which enabled us to obtain the values of lex and lem for any constant value of the fluorescence intensity from a three-dimensional spectrum.As the values obtained for a particular curve were not equidistant, the Lagrange interpolation method was applied to all points, which were placed in order, using emission wavelengths at 0.4 nm intervals. Once the trajectory had been defined, the spectrum was obtained by means of the Ftotal program.10 The spectra obtained from it displayed the same format as those obtained directly with the Aminco Bowman Series 2 spectrofluorimeter.The Ftotal program not only generates information on a fluorescent compound through the isometric representation of the three-dimensional spectrum as a level curve, but also processes the spectral data to obtain any type of spectrum of socalled new fluorimetric techniques. The statistical analysis is totally covered by means of a program developed by us. This program has a menu that includes procedures such as least median of squares regression (detection of outlier and leverage points), least squares regression with replicates or not, weighted least squares regression, tests of regression and correlation, detection and determination limits, ellipse graph for the 95% confidence region for the true slope and intercept on the ordinate estimated from the regression method, dispersion and confidence bands for the calibration graph and ANOVA test for linearity and for comparison of several regression lines.18–22 Reagents All experiments were performed with analytical reagent grade chemicals, pure solvent and Milli-Q-purified water.A stock standard solution of amiloride (Aldrich, Milwaukee, WI, USA) (250 mg dissolved in 1 l of water) was diluted to prepare suitable working standard solutions. The stock standard solution of amiloride was stored protected from light and maintained below 5 °C. Under these conditions, it was stable for 2 months. The working standard solutions of amiloride were stable for at least 2 h at room temperature.A 1.0 m buffer solution of pH 6.3 was prepared by mixing appropriate amounts of citric acid with sodium hydroxide. Urine samples were obtained from fasting, healthy people in the morning. Procedure Centrifuge urine for 15 min at 3800 rpm, transfer 10 ml of the clear supernatant solution into a 100 ml calibrated flask and dilute to volume with water. Store and maintain below 5 °C. For the preparation of the calibration graph, place an aliquot of amiloride equivalent to 500–5000 ng in a 25 ml calibrated flask, add 2.5 ml of buffer solution (pH 6.3), 12.5 ml of ethanol and 5 ml of a solution of urine free from amiloride and dilute to volume with water.Record 61 emission spectra of 192 nm width in steps of 0.4 nm, varying the excitation wavelength in 3.2 nm steps. Obtain the total luminescence spectra. Select a suitable trajectory which passes through the excitation and emission maxima of amiloride and obtain matrix isopotential synchronous spectra by means of Ftotal.10 Calculate the first derivative according to the Savitzky and Golay algorithm.23,24 Finally, determine the amiloride content by measuring the derivative signal at the emission wavelength of 392.0 nm and using the appropriate calibration graph.Results and Discussion Factors Affecting Fluorescence Intensity Chemical variables were studied to obtain the best measurement conditions and maximum fluorescence sensitivity.The influence of pH on the fluorescence intensity was studied by adding different amounts of HCl and NaOH to a urine solution and an amiloride solution. As can be seen in Fig. 1(a), the fluorescence intensity of amiloride is nearly constant at pH values between 2 and 7, whereas that due to urine decreases slowly from pH 4 to 8. A pH 6.3 was selected; this is approximately the pH value of urine of healthy people. The pH selected was adjusted by adding sodium citrate–citric acid buffer solution.The fluorescence intensity of amiloride and urine was not affected by the buffer and its concentration. A 0.1 m concentration of the buffer was therefore selected to obtain an adequate buffering capacity. Amiloride is highly soluble in water, so it was not necessary to use an organic solvent. Nevertheless, it was desirable to study the variation of amiloride and urine fluorescence with changes in the dielectric constant of the medium. Owing to the physical and chemical properties of ethanol, it is suitable for such a study.Hence the effect of ethanol content in the medium was investigated by preparing samples of amiloride and urine with ethanol concentrations between 0 and 100% v/v for the amiloride solutions and between 0 and 80% v/v for the urine solutions (when the ethanol content is more than 80% some of the urine components are precipitated). The fluorescence intensity due to amiloride increases when the ethanol content in the medium increases whereas that due to urine increases slowly [Fig. 1(b)]. For these reasons, it is preferred to use a high ethanol Fig. 1 (a) Influence of the pH on fluorescence intensity of urine (.) and amiloride (5), measuring emission at 414.8 nm after excitation at 362 nm. Urine diluted 1 : 50; concentration of amiloride, 380 ng ml21; detector voltage, 500 V. (b) Influence of ethanol concentration on fluorescence intensity of urine and amiloride, measuring emission at 414.8 nm after excitation at 362 nm.Urine diluted 1 : 50; concentration of amiloride, 380 ng ml21; detector voltage, 450 V. 248 Analyst, March 1997, Vol. 122percentage but with the precaution that precipitation does not occur. Therefore, 50% of ethanol was selected. Another factor that affects the fluorescence intensity is temperature. In both cases, fluorescence intensity decreased when the temperature was increased from 7 to 84 °C. Under the experimental conditions selected above, the temperature coefficients are about 0.99% °C21 for amiloride and about 0.59% °C21 for urine.This effect can be explained by a higher internal conversion as temperature increases, simplifying nonradiative deactivation of the excited singlet state.25 Therefore, the use of a thermostat is recommended and a measurement temperature of 20 °C was chosen. The influence of amiloride concentration on the fluorescence intensity was studied under the above conditions. The best range of pure amiloride concentration for the relationship between fluorescence intensity and concentration was found to be up to 152 ng ml21.Determination of Amiloride in Urine Previous experiments9,11–17 have shown that the qualitative composition of the fluorescent metabolites of urine from healthy people of both sexes and different diets, aged between 25 and 35 years, which is the usual age of people taking doping substances, is almost constant, a necessary condition for the application of the proposed technique.Different samples of urine display the same type of fluorescence, with hardly any variation in the form of the spectrum and the location of the fluorescence maxima, although it is possible to observe some variations in their intensity. All fluorimetric three-dimensional spectra were obtained by varying the emission wavelength from 310.0 to 502.0 nm and the excitation wavelength from 220.8 to 412.8 nm. Both excitation and emission bandpasses were 8 nm and the scan rate was 50 nm s21.Under these conditions, the time taken to determine contour spectra was 292 s. Matrix isopotential synchronous fluorescence was applied to the determination of amiloride in urine, a very fluorescent matrix, under the optimum conditions established above. As Fig. 2 shows, the fluorescence maxima characteristic of amiloride (solid line) are located in the UV region. As can readily be observed in this region, urine (broken line) shows two bands that prevent the determination of this compound without prior separation.Amiloride shows two broad peaks that are located at lex = 284.8 nm, lem = 412.8 nm and lex = 364.8 nm, lem = 412.8 nm. As at an excitation wavelength of 284.8 nm urine gives high absorbance values and there are more interferences, it is preferable to determine amiloride in urine samples with an excitation wavelength of 364.8 nm where urine needs only minor dilution and adequate selectivity is obtained.The isopotential trajectory for the urine spectrum also shown in Fig. 2 and, as can be observed, the trajectory A passes through the excitation and emission maxima of amiloride (lex = 364.8 nm, lem = 412.8 nm). Trajectory A was obtained from the spectrum corresponding to the arithmetic mean of 10 total fluorescence spectra of the different urine samples by means of the Ftotal program.10 Urine sample solutions containing amiloride gave signals smaller than those obtained with aqueous standard solutions, owing to some type of binding with other components of the urine.Total luminescence spectra of amiloride were obtained in different urine samples to construct calibration graphs and to carry out recovery experiments. Fig. 3 shows the effect of background fluorescence intensity (urine) on amiloride MISF and derivative spectra. It can be seen that the spectra are identical in form although their intensities are different by constant terms.They correspond to the values of the fluorescence intensity due to the three urines in the isopotential trajectory applied. It is easy to observe that derivatization totally removes the background effect. The first derivative of MISF spectra was applied to all samples. The number of points through which the derivative was obtained was optimized, with the conclusion that derivative spectra with a suitable signal-to-noise ratio were obtained with 25 points. In the same way we obtained the total luminescence spectra of amiloride in aqueous solution at the same concentrations. The MISF spectra were obtained by applying the isopotential trajectory (Fig. 2).We also obtained their first derivatives, as with the urine samples. As can be readily seen in Fig. 4(a), which shows the MISF spectra of amiloride obtained in urine, it is not possible to determine this drug by measuring the maximum fluorescence Fig. 2 Total fluorescence spectra of 152 ng ml21 of amiloride (solid line) and diluted urine (1 : 50) (broken line).The line labelled A is the isopotential trajectory for the urine. Analyst, March 1997, Vol. 122 249intensity (which is located at lem = 412.8 nm) with regard to the initial or final extremes (lem = 340.0 nm and lem = 480.0 nm) of the selected trajectory, since the fluorescence intensity does not reach a constant value. The first derivative technique was therefore applied. Calibration graphs were constructed by measuring the first derivative at lex = 357.0 nm lem = 392.0 nm, where maximum sensitivity is achieved.Fig. 4(b) shows the spectra derived from the calibration of amiloride in urine. In order to test the independence of the analytical signal of amiloride, i.e., to show that the signal measured is independent of the urine, three calibration graphs from the first derivative signals were constructed with different urine samples. The proposed method was evaluated by a statistical analysis of experimental data by fitting the least squares line according to y = a + bx, after discarding outliers with help of the least median of squares regression (LMS).21 As LMS is a robust regression method, it is able to detect outlier points.These outliers cause errors in the true line when experimental data are fitted according to the least squares regression. Although the LMS method considers outliers to the points that give an absolute value of standardized residual above 2.5, owing to the robustness of the LMS fit, we only reject points that give an absolute value of standardized residual higher than 10.No outliers were detected. Table 1 shows the results of the statistical analysis. To verify if the intercepts on the ordinate were negligible, significances were studied by applying Student’s t-test at the 95% confidence level and suitable degrees of freedom.19,20 If the intercepts on the ordinate for the lines calculated by the least Fig. 3 (a) Set of MISF spectra of 122 ng ml21 of amiloride in different urine samples. (b) Corresponding first derivative spectra. Fig. 4 (a) Set of MISF spectra of amiloride in urine. Amiloride concentrations: (1) 20; (2) 30; (3) 40; (4) 60; (5) 80; (6) 120; (7) 160; and (8) 200 ng ml21. (b) Corresponding first derivative spectra. Table 1 Statistical data for the determination of amiloride in urine by means of first derivative matrix isopotential synchronous fluorescence Slope Deter- without Fitting mination Intercept on Confidence interval Confidence interval Exp.SD of intercept, used* Sample coefficient ordinate, a Slope, b for intercept for slope t-test estimation bA 1 Urine I 0.9997 21.750 3 1024 4.056 3 1024 — — — 7.2 3 1024 4.040 3 1024 Urine II 0.9997 2.037 3 1023 3.733 3 1024 — — — 5.1 3 1024 4.190 3 1024 Urine III 0.9999 21.610 3 1023 3.981 3 1024 — — — 1.4 3 1024 3.631 3 1024 2 Urine I 0.9967 25.244 3 1024 4.070 3 1024 1.247 3 1023 to 4.286 3 1024– 0.725 1.2 3 1023 4.018 3 1024 22.295 3 1023 3.854 3 1024 Urine II 0.9974 1.233 3 1023 3.961 3 1024 2.764 3 1023 to 4.148 3 1024– 1.970 1.0 3 1023 4.085 3 1024 22.985 3 1024 3.775 3 1024 Urine III 0.9995 22.134 3 1023 4.172 3 1024 21.421 3 1023 to 4.259 3 1024– 7.323b 4.7 3 1024 3.958 3 1024 22.846 3 1023 4.086 3 1024 3 Global 0.9958 24.751 3 1024 4.068 3 1024 4.579 3 1024 to 4.182 3 1024– 1.056 1.2 3 1023 4.020 3 1024 21.408 3 1023 3.954 3 1022 * 1 = Least median of squares regression line according to y = a + bx and y = bAx. 2 = Least squares regression line according to y = a + bx (theoretical t = 2.447) and y = bAx. 3 = Overall least squares regression line according to y = a + bx (theoretical t = 2.074) and y = bAx. b Experimental t-value is greater than the theoretical t-value and therefore the slope calculated without an intercept is not reliable. 250 Analyst, March 1997, Vol. 122squares technique are negligible, it is necessary to perform the least squares regression again according to the function y = bAx and then the new value of the slopes of calibration graphs (bA) may be calculated.As Table 1 shows, the intercept on the ordinate is significantly different from zero for the true line of urine III and consequently the fit according to y = bAx is not recommended; nevertheless, the value of the slope for this regression is also given in Table 1. To obtain the most representative calibration graph, an overall least squares was developed. The regression line shows homoscedasticity19,20 and consequently the residuals manifest a uniform variance (the errors of measurements are independent of the amiloride concentration), it being unnecessary to weight the first derivative values according to the mean standard deviation.When the 95% confidence region for true slope and intercept19,20 estimated is represented, the zero intercept on the ordinate falls within the joint confidence region.This means that the intercept is not significantly different from zero. The confidence interval for the corresponding slope is 3.946 3 1024– 4.095 3 1024. Nevertheless, the intercept significance was tested by applying Student’s t-test. The result was not significantly different from zero, and therefore it is possible to establish a proportional relationship between the analytical signal and amiloride concentration. Therefore, the overall least squares fit according to y = bAx was performed.When the first derivative of matrix isopotential synchronous spectrofluorimetry was used and the overall least squares regression line according to y = a + bx was considered, a detection limit of 2.78 ng ml21 and a determination limit of 9.26 ng ml21 were obtained on applying the IUPAC21,22 definition, in which only the standard deviation of the blank is assumed. The propagation of errors approach will give values of detection and determination limits consistent with the reliability of the blank measurements and the signal measurements of the standards.21,22 In this case a detection limit of 4.45 ng ml21 and a determination limit of 14.82 ng ml21 were obtained.Precision, Accuracy and Interference Studies To study the precision of the method, a series of 10 solutions at three different concentrations were prepared containing 15, 61 and 152 ng ml21 of amiloride in urine and they were measured 15 times on the same day (repeatability study) and once a day for 10 d (reproducibility study). The SEs, assuming the error propagation theory, and RSD of repeatability at each of these levels were 2.2, 2.1 and 2.4 ng ml21 and 14.4, 3.5 and 1.6%, respectively.The SEs and RSDs of reproducibility at each of these levels were 2.0, 1.9, and 2.2 ng ml21 and 12.5, 3.1 and 1.4%, respectively. It can be seen that the SEs are approximately constant with the concentration, which is in agreement with the homoscedasticity that the calibration line presents.The accuracy of the method was established by testing the analytical signal corresponding to three replicates of three samples of 15, 61 and 152 ng ml21 of amiloride in different urine samples. The recoveries obtained were 107.0, 100.1 and 101.6%, respectively. A comparison study of six samples of amiloride in urine at therapeutic levels was performed by means of MISF and the currently accepted method for doping substances (HPLC screening and GC–MS confirmation).The concentrations calculated in reverse by both methods are given in Table 2. As both methods should give same concentrations for the same samples, if least squares pair regression is applied at these concentrations, zero intercept on the ordinate and a slope of unity must be obtained. In order to test whether these methods are significantly different, the confidence intervals for the parameters of the linear regression model must be studied together.If the intercept on the ordinate, a, is plotted against the slope, b, for repeated random samples the points will be found to fall elliptically about the true centre (a, b) and, conversely, any confidence interval for the true combination of a and b will take the form of an elliptical region about the best estimates (a, b) as the centre.19 Fig. 5 shows the 95% confidence region for the true slope and estimated intercept. As can be observed, the point corresponding to the zero intercept and unit slope falls within the joint confidence region.This means that the accuracies of the proposed and currently accepted method are not significantly different. The specificity was studied by adding some closely related drugs to urine and testing to see if they caused interference in the amiloride quantification. The amiloride concentration was 152 ng ml21 and the concentration of the drugs added (atenolol, furosemide, dipyridamole, aspirin, bumetanide, hydrochlorothiazide, hydroflumethazide, caffeine, metoprolol, quinidine and quinine) was slightly higher than would be expected at therapeutic doses.Except with quinidine and quinine, no significant variation of the analytical signal was observed from the value expected when amiloride was present alone. Nevertheless, quinidine and quinine do not present interferences if the calibration is performed at the minimum of the first derivative from the matrix isopotential synchronous spectrum that is located at lex = 363.2 nm and lem = 434.0 nm. Conclusions A method for the direct fluorimetric determination of amiloride in urine by matrix isopotential synchronous fluorescence, without the need for prior separation, has been described.The determination of this natriuretic drug in urine can be performed at lex = 357.0 nm and lem = 392.0 nm in the first derivative of the matrix isopotential synchronous scan. As can be readily observed from the detection limits, only a 63% of error is caused by the standard deviation of the blank.Table 2 Concentrations (ng ml21) of amiloride in urine samples calculated in reverse using the currently accepted method and the proposed method XAccepted YProposed 14 15 28 30 58 61 98 91 122 122 158 152 Fig. 5 The ellipse is the 95% confidence region for the true slope and intercept on the ordinate estimated from the overall least squares regression between the concentration calculated in reverse by the proposed MISF method and the currently accepted method for the determination of doping substances.The point (0,1) corresponds to zero intercept and unit slope. This point falls within the joint confidence region and consequently there are no significant differences in the accuracy of the two methods. Analyst, March 1997, Vol. 122 251Therefore, the IUPAC definition with a standard deviation proportionality factor of three is inadequate. An exhaustive statistical analysis was applied to all calibration graphs, including the least median squares robust regression and least squares.Owing to the homoscedasticity of the regression line, the weighted version is not recommended. The significance of the intercept on the ordinate was investigated through the ellipse method and Student’s t-test at the 95% confidence level and was found not to be significantly different from zero. To validate the method, precision, accuracy and interference studies were performed. As can be observed from the SEs and RSDs obtained from the repeatability and reproducibility experiments, excellent precision, better than in other spectrofluorimetric methods, is achieved without pre-separation procedures. 3–8 In a comparison of this method with the currently accepted method, their accuracies were not significantly different. The specificity of the proposed fluorimetric method is demonstrated by the few interferences from some closely related drugs to urine. The authors gratefully acknowledge financial support from the Direcci�on General de Investigaci�on Cient�ýfica y T�ecnica (Project No. PB 94-0743). The authors gratefully thank the Laboratorio de Control del Dopaje del Consejo Superior de Deportes del Ministerio de Educaci�on y Cultura for the determination of amiloride in urine samples by HPLC screening and GC–MS confirmation. References 1 Rodr�ýguez Bueno, C., Dopaje, Interamericana–McGraw-Hill, Madrid, 1992. 2 Clarke’s Isolation and Identification of Drugs, ed. Moffat, A. C., Pharmaceutical Press, London, 1986, p. 339. 3 Kurani, S. P., Desai, D. K., and Seshadrinathan, A. G., Indian Drugs, 1986, 23, 230. 4 Sastry, C. S. P., Suryanarayan, M. V., Tipirnani, A. S. R. P., and Satyanarayana, T., Indian Drugs, 1989, 26, 651. 5 Wahbi, A. A. M., Bedair, M. M., Galal, S. M., and Gazy, A. A., S. T. P. Pharm. Sci., 1933, 3(2), 182. 6 Garc�ýa Sanch�ez, F., Fern�andez Gutierrez, A., and Cruces Blanco, C., Anal. Chim. Acta, 1995, 306, 313. 7 Karola, R., Knauf, H., and Ernst, M., J. Chromatogr., 1982, 233, 432. 8 Reewijk, H. J. E. M., Tjaden, U. R., and Van der Greef, J., J. Chromatogr., 1992, 575, 269. 9 Murillo, J. A., and Ala�n�on, A., Anal. Chim. Acta, 1994, 296, 87. 10 Murillo, J. A., and Ala�n�on, A., Comput. Chem., 1993, 17, 34. 11 Leiner, M. J. P., Hubmann, M. R., and Wolfbeis, O. S., Anal. Chim. Acta, 1987, 198, 13. 12 Murillo Pulgar�ýn, J. A., and Ala�n�on, A., Analyst, 1994, 119, 1915. 13 Berzas Nevado, J. J., Murillo Pulgar�ýn, J. A., and G�omez Laguna, M. A., Analyst, 1995, 120, 171. 14 Murillo, J. A., Ala�n�on, A., and Fern�andez, P., Talanta, 1996, 43, 431. 15 Murillo, J. A., and Ala�n�on, A., Anal. Chim. Acta, 1996, 317, 359. 16 Murillo, J. A., Ala�n�on, A., and Fern�andez, P., Anal. Chim. Acta, 1996, 326, 117. 17 Murillo, J. A., and Ala�n�on, A., Microchem. J., 1995, 52, 341. 18 Rousseeuw, P. J., and Leroy, A. M., Robust Regression and Outlier Detection, Wiley, New York, 1987. 19 Lark, P. D., Craven, B. R., and Bosworth, R. C. L., The Handling of Chemical Data, Pergamon Press, Oxford, 1968, ch. 4. 20 Massart, D. L., Vandeginste, B. G. M., Deming, S. N., and Kaufman, L., Chemometrics: a Textbook, Elsevier, Oxford, 1988. 21 Miller, J. N., Analyst, 1991, 116, 3. 22 Gary, L., and Winefordner, J. D., Anal. Chem., 1983, 55, 712. 23 Savitzky, A., and Golay, M. J. E., Anal. Chem., 1964, 36, 1627. 24 Steinier, J., Termonia, Y., and Deltour, J., Anal. Chem., 1972, 44, 1906. 25 Seitz, W. R., in Treatise on Analytical Chemistry, ed. Elving, P. J., Meehan, E. J., and Kolthoff, I. M., Wiley, New York, 1981, pp. 194– 196. Paper 6/07219D Received October 23, 1996 Acce

ISSN:0003-2654    

DOI:10.1039/a607219d

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

N N N N N N N N CH2CH2OH CH2CH2OH HOCH2CH2 HOCH2CH2 Phosphorimetric Determination of Dipyridamole in Pharmaceutical Preparations Jos�e A. Murillo Pulgar�ýn*, Aurelia Ala �n�on Molina and Pablo Fern�andez L�opez Department of Analytical Chemistry and Foods Technology, University of Castilla La Mancha, 13071 Ciudad Real, Spain Room temperature phosphorescence was applied to the determination of dipyridamole in pharmaceutical preparations. The response was linear in the concentration range 100–1600 ng ml21.The use of phosphorescence enhancers such as thallium(I) nitrate (external heavy atom), sodium dodecyl sulfate (microemulsion stabilizer) and sodium sulfite (deoxygenation agent) was studied and optimized to obtain maximum sensitivity and adequate selectivity. The determination was performed in 0.026 M sodium dodecyl sulfate, 0.0156 M thallium nitrate and 0.02 M sodium sulfite. The pH value was 11.5, adjusted by adding sodium hydroxide. The phosphorescence was totally developed in 15 min, after that the intensity was measured at lex = 303 nm and lem = 616 nm.The recovery of the method was tested on commercial formulations containing dipyridamole. The recoveries obtained were 94.67 ± 0.58% for Persantin and 96.75 ± 1.37% for Asasantin. The overall least squares regression method was applied to find the most exact straight line that fits the experimental data. The detection limit according to the error propagation theory was 16.4 ng ml21.The repeatability and relative standard deviation were also determined according to this theory. Keywords: Dipyridamole; pharmaceuticals; phosphorimetry; room temperature phosphorescence Phosphorimetry is a selective method for measuring different organic compounds, such as pesticides,1 PAHs2,3 and some drugs.4 Although room temperature fluorimetry is usually more sensitive than phosphorimetry, the broad band of spectra for different compounds limits the selectivity.Therefore, phosphorimetry is used because of its better selectivity over room temperature fluorescence and absorption spectrometry,5–9 since not all molecules that fluoresce will phosphoresce, often completely eliminating the spectral interference, and the phosphorescence is shifted to a less crowded spectral region. However, the difficulties associated with low temperature phosphorescence (LTP) make this technique unpopular. Owing to these difficulties, LTP has not been used extensively for the identification and determination of many compounds.Since its discovery in 1967 by Roth,10 room temperature phosphorimetry (RTP) has attracted great interest and has become a practical technique for the detection of many organic compounds.10–18 As the sample must be adsorbed on an inert substrate such as filter-paper,19,20 the technique has the disadvantages of cumbersome sample preparation, critical drying requirements and high phosphorescent background intensity from the filter-paper substrate.RTP can also be observed from many organic compounds in liquid solutions incorporating the phosphors into organized media such as cyclodextrins, micellar systems and microcrystalline media or by using sensitized RTP. In 1980, Cline Love and co-workers21,22 first investigated micelle-stabilized RTP and studied several PAHs. In a micellar solution the analytes included in the micellar assembly are apparently protected from the quenchers present in the solution.Observation of RTP in a micellar solution usually requires the presence of a heavy atom. It is placed as a counter ion outside the micelle, thus being in proximity to the hydrophobic molecules associated with the micelle. The high local concentration of the heavy atoms produces an efficient spin–orbit coupling that can diminish the fluorescence and increase the phosphorescence. Furthermore, the micelles can effectively screen molecules in the excited triplet state from the action of potential quenchers present in the bulk water phase.However, phosphorescence is not observed unless oxygen is removed, as it is a very efficient quencher that easily penetrates the micelles. D�ýaz Garc�ýa and Sanz-Medel23 have proposed the substitution of the troublesome deoxygenation with nitrogen by chemical deoxygenation with sodium sulfite, thus avoiding foam formation and ensuring more permanent protection of the solution against contamination with oxygen.In the presence of sulfite, however, the phosphorescence is not immediately observed. The oxygen in the bulk water phase is removed first, followed by that in the micelle pseudophases as it diffuses out. Equilibrium is achieved in a few minutes. In the spite of the interest in observing phosphorescence in micellar solutions, only a few analytical procedures have used this technique. In this work, the appropriate experimental conditions to obtain a reproducible and maximum phosphorescence signal, when sulfite is used to eliminate the oxygen from the micellar solutions, were studied.Dipyridamole (2,2A,2AA,2AAA-{(4,8-dipiperidinopyrimino[5,4- d]pyrimidine-2,6-diyl)dinitrilo}tetraethanol) is shown in Fig. 1. It is an intensely yellow crystalline powder. Its solution gives a yellowish blue fluorescence. It is almost insoluble in water. Dipyridamole is a vasodilator agent that is widely used in medicine. After oral administration, it stimulates a rise in the blood flow through the coronary circulation, providing more blood to the myocardium.This effect has been used in certain sports to increase energy production, such as via ATP molecules. Therefore, this vasodilator agent is classified, in doping terms, as a stimulant. Stimulants have been consumed in sports to increase efficiency and decrease tiredness. Stimulants include other substances that intensify the attention and could increase competitiveness and aggressiveness.Nevertheless, the uncontrolled use of such drugs could cause the loss of mental power and also have serious secondary effects that could cause grave danger to health. Fig. 1 Structure of dipyridamole. Analyst, March 1997, Vol. 122 (253–258) 253There have been various reports describing the spectrophotometric determination of dipyridamole in pharmaceutical tablets.24–34 In comparison with the method proposed here, these methods show low sensitivity and selectivity.The need for derivatization reactions31–34 to increase the selectivity makes these methods tedious, low accuracy is achieved and the detection limits are higher than when direct spectrophotometry is used. Only three methods have utilized the highly fluorescent nature of this drug.35–37 Although Shao et al.35 developed a direct spectrofluorimetric determination, the method has low sensitivity. The methods of Wilczynska-Wojtulewicz36 and Steyn37 require various preliminary procedures such as extraction and separation by means of TLC.We propose the application of RTP to determine dipyridamole in pharmaceutical tablets. The main advantages of using RTP over spectrophotometric and fluorimetric methods for this purpose are the high selectivity and sensitivity achieved, together with a detection limit of < 20 ng ml21 compared with approximately 2 mg ml21 and 100 ng ml21 for spectrophotometric and spectrofluorimetric methods, respectively.Experimental Apparatus All phosphorimetric measurements were performed on an Aminco Bowman Series 2 luminescence spectrometer, connected to software which runs on the OS2 operating system. The instrument utilizes a 7 W integral pulsed xenon lamp for phosphorescence measurements. The gated photomultiplier tube detection includes a unique masking method for detection in < 200 ms after the initiation of the flash lamp. Quartz glass cuvettes with pathlengths of 1.0 3 1.0 cm were used.Thermostatic equipment and a Crison Model 2001 pH meter with a glass–saturated calomel combination electrode were also used. Software The AB2 program allows the operation of the instrument to obtain excitation and emission spectra, total phosphorescence spectra and time traces, such as decay curves and time resolved curves. The Ftotal program38 was used to generate fluorescence and phosphoresectra. Statistical analysis was performed by means of a program developed by us, which has an option menu that includes all the procedures mentioned in this paper.Reagents All experiments were performed with analytical reagent grade chemicals, pure solvents and Milli-Q-purified water. Sodium dodecyl sulfate (SDS) was obtained from Sigma (St. Louis, MO, USA) and thallium(i) nitrate and sodium sulfite from Merck (Darmstadt, Germany). A stock standard solution of dipyridamole (Aldrich, Milwaukee, WI, USA) (100.0 mg disolved 100 ml of 0.1 m SDS) was diluted to prepare working standard solutions.The stock standard solution of dipyridamole was stored and protected from the light and maintained below 5 °C. Under these conditions, it was stable for at least 2 weeks. The working standard solutions of dipyridamole were stable for at least 2 d at room temperature. Stock standard solutions of 0.1 m SDS and 0.3 m thallium(i) nitrate were used. A 0.25 m sodium sulfite solution was prepared daily. Procedure For the preparation of the calibration graph, an aliquot of dipyridamole standard solution was pipetted into a 25 ml calibrated flask, then SDS as necessary to give a 0.026 m concentration, 2.5 ml of 0.025 m sodium hydroxide solution, 1.3 ml of 0.3 m thallium(i) nitrate solution and 2.0 ml of 0.25 m sodium sulfite were added.The solution was diluted to volume with water and shaken. After the flask had stood for 15 min in a thermostat at 20 °C, a portion of the solution was transferred into a phosphorescence cuvette and the RTP measured at an excitation wavelength of 303 nm and an emission wavelength of 616 nm.For the analysis of Persantin (Boehringer Ingelheim, Barcelona, Spain), 10 tablets were weighed and for the analysis of Asasantin (Boehringer Ingelheim) ten capsules (which contained inside dipyridamole in the form of minute tablets) were weighed. In all instances, the solid was powdered, homogenized and about 0.1 g was taken for analysis. Suitable dilutions were made with 0.1 m SDS. In all instances, the excipients were not soluble in SDS, so after shaking and maintaining in an ultrasonic bath for 5 min the solutions had to be centrifuged. Results and Discussion Spectral Characteristics Fig. 2 shows the total phosphorescence spectrum (solid line) and the total fluorescence spectrum (broken line) of dipyridamole. They are identical in form, varying only in the location of the emission wavelength of maximum intensity. In the Fig. 3(a) the phosphorescence (solid line) and fluorescence (broken line) excitation spectra at the emission wavelength of Fig. 2 Total phosphorescence spectrum (solid line) and total fluorescence spectrum (broken line) of dipyridamole. Fig. 3 Phosphorescence (solid line) and fluorescence (dashed line) spectra of dipyridamole. Photomultiplier voltage 800 V in phosphorescence and 410 V in fluorescence. (a) Excitation spectra; and (b) emission spectra. 254 Analyst, March 1997, Vol. 122maximum intensity are shown.Both excitation spectra are coincident, displaying two broad peaks at excitation wavelengths of 303 and 411 nm. The coincidence of the phosphorescence and fluorescence excitation spectra means an identical excitation process in both cases. This is justified in theory by the light absorption promoting an electron from the ground electronic state to the first and second excited singlets. The peak at 303 nm shows a higher luminescence intensity than that at 411 nm and the former excitation wavelength was therefore chosen for phosphorescence measurements.Fig. 3(b) shows the phosphorescence (solid line) and fluorescence (broken line) emission spectra. The fluorescence spectrum gives a band with a characteristic emission wavelength of 488 nm, corresponding to the transition from the first excited singlet to the ground state. The phosphorescence spectrum presents a band at 616 nm corresponding to the transition from the excited triplet to the singlet ground state.The fluorescence and phosphorescence spectra show different characteristic emission wavelengths. This difference of 128 nm gives an idea of the non-radiant energy that is lost in the intersystem crossing and the subsequent vibrational relaxation to the lowest vibrational level of the excited triplet. The phosphorescence lifetime of dipyridamole is approximately 1324 ms. This is the time required for the population of the excited triplet state to decrease to 1/e of its original value after the excitation source has been turned off.The lifetime is a means of considering the luminescence process in terms of rates. Fluorescence lifetimes are typically of the order of 1–20 ns. As phosphorescence is a spin forbidden process, phosphorescence lifetimes are considerably longer, generally ranging from milliseconds to seconds. Consequently, the phosphorescence lifetime is a measure of the forbiddenness of singlet–triplet transitions in a given molecule.Factors Affecting Phosphorescence Chemical variables were optimized to obtain maximum phosphorescence sensitivity and adequate selectivity. Dipyridamole is almost insoluble in water, so it was necessary to use an organic solvent. Owing to the micellar properties of SDS it was used to dissolve dipyridamole; further, the semirigid structure of the solution favours the development of phosphorescence. The extremely high sensitivity of the triplet state to quenching by oxygen requires deoxygenation of the sample.Owing to the high efficiency of this quenching process, it is necessary to eliminate totally the oxygen in the micellar solution. The method proposed for sample deoxygenation is based on the redox reaction of sulfite with molecular oxygen to product sulfate.23 It was observed that a concentration of Na2SO3 of 0.02 m was required to eliminate oxygen completely from the solutions, which was evident from the intensity of the phosphorescence signal.This decreases at higher concentrations of Na2SO3. The decrease in the signal for higher concentrations of Na2SO3 has been interpreted as the displacement of thallium(i) from the micelle because of the high concentration of sodium in the solution.23 An Na2SO3 concentration of 0.02 m was chosen. The influence of pH on phosphorescence intensity was studied by adding different amounts of H2SO4 and NaOH to a dipyridamole solution.The phosphorescence is not significant at pH values up to 9.5. In Fig. 4 phosphorescence intensity is plotted versus the pH of the solution and, as can be readily observed, dipyridamole has maximum intensity at pH 11.5. Phosphate buffer was first used to adjust this pH value, but the phosphorescence intensity diminished. This is a quenching situation due to the formation of exciplexes. The explanation for this behaviour is that an excited dipyridamole molecule associates with the phosphate in the ground state.Consequently, the phosphorescence intensity of excited dipyridamole diminishes. This pH was subsequently adjusted by adding 2.5 ml of 0.025 m NaOH. It was also necessary to establish how the phosphorescence intensity varies with changes in thallium concentration. Fig. 5(a) shows this relationship for thallium concentrations between 0.0024 and 0.0168 m. As can be readily observed, the phosphorescence intensity increases with increase in thallium concentration.For concentrations of thallium above 0.0168 m precipitation is observed. For lower concentrations of thallium, the heavy atom effect is very sensitive. When the thallium salt is not present the phosphorescence intensity disappears. A concentration of thallium of 0.0144 m gives high intensity without problems of precipitation. The effect of SDS concentration was investigated by preparing samples with SDS concentrations between 0.014 and 0.074 m. As Fig. 5(b) shows, phosphorescence intensity of dipyridamole diminishes as the SDS concentration increases.For SDS concentrations lower than 0.014 m precipitation occurred. A 0.022 m concentration of SDS was therefore selected, giving good sensitivity while being sufficient to dissolve the dipyridamole. From the above two experiments, the phosphorescence intensity is maximum with SDS and thallium concentrations in Fig. 4 Variation of phosphorescence intensity of dipyridamole with pH. The pH was varied from 9.5 to 12.5 by adding sodium hydroxide.Fig. 5 (a) Effect of thallium(i) concentration on phosphorescence intensity at an SDS concentration of 0.022 m; (b) effect of SDS concentration on phosphorescence intensity at a thallium(i) concentration of 0.0144 m; and (c) effect of SDS and thallium(i) on phosphorescence intensity at an SDS to thallium(i) ratio of 1.667. Analyst, March 1997, Vol. 122 255the proportion of 1.667, and Fig. 5(c) shows the phosphorescence intensity versus SDS concentration with this proportion maintained.As can be observed, the phosphorescence intensity increases as the concentration of SDS increases, reaching a constant value at SDS concentrations above 0.025 m. Concentrations of SDS of 0.026 m and of thallium(i) 0.0156 m were selected as adequate. At concentrations of SDS above 0.032 m precipitation occurred. Another factor that affects the phosphorescence intensity is temperature, the phosphorescence intensity decreasing when the temperature increased from 15 to 79 °C. This decrease is measured by plotting the relative signal increment versus temperature, i.e., the intensity at each temperature minus the intensity at the lowest temperature, divided by the intensity of the higher temperature and multiplied by 100.This relationship shows a linear behavior, the slope (which is the temperature coefficient of dipyridamole) being 1.36% °C21, the intercept on the ordinate 104.4% and the coefficient of the determination r2 = 0.995. The influence of dipyridamole concentration on the phosphorescence intensity was studied under the above conditions.The phosphorescence intensity reaches a constant value at dipyridamole concentrations above 20 mg ml21. The dipyridamole concentration range for a linear relationship between phosphorescence intensity and concentration was found to be up to 1600 ng ml21. The inner filtering effect is significant for concentrations of dipyridamole above 1600 ng ml21, and increases as the dipyridamole concentration increases.Consequently, calibration was performed for dipiridamole concentrations up to 1600 ng ml21 with three replicates per point. Determination of Dipyridamole Under the operating conditions outlined above, we propose a method to determine dipyridamole by direct measurement of phosphorescence intensity with an emission wavelength of 616 nm and an excitation wavelength of 303 nm in the concentration range 100–1600 ng ml21.The delay time required was 120 ms with the photomultiplier tube masked. The gate time appropriate for this delay time was 800 ms and the detector voltage was 740 V. Both excitation and emission bandpasses were 16 nm and the scan rate was 5 nm s21. A calibration graph was constructed with three replicates per point. Fig. 6 shows the average emission spectra of the calibration concentrations. The proposed method was evaluated by a statistical analysis of the experimental data by fitting the overall least squares line according to y = a + bx.39,40 Table 1 gives the results from the statistical analysis.The calibration line presents homoscedasticity (standard residuals have a uniform variance) and therefore it is not necessary to weight the phosphorescence intensity values according to the mean standard deviation. In order to test the linearity of the overall least squares regression, the ANOVA test was performed.39 The results are given in Table 2.When the 95% confidence region for true slope and intercept40 estimated is represented, the zero intercept on the ordinate falls within the joint confidence region. This means that the intercept on the ordinate is not significantly different from zero. Therefore, the relationship between phosphorescence intensity and dipyridamole concentration is proportional. The confidence interval for the corresponding slope is 1.257 3 1023–1.238 3 1023.Nevertheless, the significance of the intercept was tested by applying Student’s t-test (Student’s t-test is more restrictive than the confidence region of the true slope and intercept). The experimental t-value obtained was 0.324, which is less than theoretical value of 2.12. If the theory of error propagation is considered, the values of the detection and determination limits are consistent with the reliability of the blank measurements and the signal measurements of the standards.41,42 In this case a detection limit of 16.4 ng ml21 and a determination limit of 54.6 ng ml21 were obtained.The detection limit according to Clayton considers the probability of false positive and false negative values, the detection limit being 18.8 ng ml21. In order to study the precision of the method, a series of 10 solutions of 800 ng ml21 of dipyridamole were measured on the Table 1 Results from statistical analysis of data; least squares regression with replicates Intercept on ordinate (a) 21.705 3 1023 SD of intercept on ordinate (sa) 5.3 3 1023 Slope (b) 1.249 3 1023 SD of slope (sb) 5.8 3 1026 SD of regression (syx) 1.3 3 1022 Coefficient of determination (r2) 0.9996 Confidence interval of intercept on ordinate 9.447 3 1023, 21.286 3 1022 Confidence interval of slope 1.262 3 1023, 1.237 3 1023 Slope without intercept on ordinate 1.248 3 1023 Table 2 ANOVA test: linearity test Source of variation SS* DF* MS* Fexp.Ftheoretical Due to regression 8.2804 1 8.2804 Set means about the line (lack of fit) 2.6916 3 1024 4 6.7290 3 1025 0.31 3.26 Within line (pure error) 2.6363 3 1023 12 2.1969 3 1024 Total 8.2833 17 0.4873 * SS = Sum of squares of deviation; DF = degrees of freedom; MS = SS DF.Fig. 6 Set of room temperature phosphorescence spectra. Dipyridamole concentration: 1, 1600; 2, 1200; 3, 800; 4, 400; 5, 200; and 6, 100 ng l21. 256 Analyst, March 1997, Vol. 122same day. By applying the IUPAC definition, the mean standard deviation of replicates was 3.88 ng ml21 and the relative error 1.1%, whereas if error propagation is assumed, the SE obtained was 2.59 ng ml21 and the relative error 0.71% (95% confidence level).In this case the relative error for replicates is less on applying the SE of the regression line, owing to the classical hyperbolic shape of the confidence and dispersion bands of the true line which are closest at the mean concentration, this being very near to 800 ng ml21, where the standard deviation is reduced to the standard deviation of the mean, and therefore, the relative error is less.To estimate concentration values at greater distances from the mean concentration of the regression line, the error propagation method gives the greatest values for the relative error. The accuracy of the method was established by testing the analytical signal corresponding to the concentrations of the calibration line. As can be seen from the RSDs in Table 3, the values obtained by applying the error propagation theory confirm that, to calculate the confidence interval of a measurement, the error propagation method may be used.The determination of dipyridamole by RTP was compared with the currently accepted method.37,43 In this, 1 ml of the problem solution is extracted at pH 10 with 8 ml of hexane– isoamyl alcohol (95 + 5) and the organic phase used directly for measurement of the fluorescence intensity (lex = 405 and lem = 495 nm).A comparison study of 10 samples of dipyridamole was performed by applying least squared pair analysis40 with the concentration calculated in reverse by means of the currently accepted method and the proposed method. As both methods should give the same concentrations for the same samples, a zero intercept on the ordinate and a slope of units must be obtained. Fig. 7 shows the 95% confidence region for the true slope and intercept estimated.As can be observed, the point corresponding to the zero intercept and unit slope falls within the joint confidence region. This means that the intercept is not significantly different from zero and the slope is not significant different from unity. Consequently, the accuracies of proposed and currently accepted method are not significantly different. Applications and Interference Study The recommended procedure was applied satisfactorily to the determination of dipyridamole in the Spanish pharmaceutical products that contain this vasodilator agent in different proportions (Persantin 100 mg and Asasantin 75 mg).The assay results expressed as a percentage of the nominal contents resulting from the average of three determinations of three different tablets were 94.67% for Persantin and 96.75% for Asasantin, the standard deviations being 0.58% and 1.37%, respectively. The recoveries agree well enough with the nominal content and the precision is quite satisfactory The specificity of this determination was studied by adding related drugs to the pharmaceutical preparation of dipyridamole and other drugs with intrinsic fluorescence and testing to see if the added drugs cause interference in the dipyridamole quantification.In the first case, dipyridamole (1600 ng ml21) was spiked with saccharose, lactose and glucose (20 mg ml21). No interferences were found. In the second, the added drugs were aspirin, atenolol, spironolactone and canrenoic acid (1 mg ml21), amiloride, furosemide and triamterene (25 ng ml21), aspirin and nadolol (2 mg ml21), metoprolol (80 ng ml21) and caffeine, quinidine and quinine (1.6 mg ml21).Except for the furosemide, no significant variation of the analytical signal was observed from the value expected when dipyridamole is present alone. Nevertheless, interferences were observed if the concentrations of these drugs were increased. Conclusions A method for direct phosphorimetric determination of the vasodilator agent dipyridamole in pharmaceutical preparations has been developed.The determination can be performed by measuring phosphorescence intensity within an emission wavelength of 616 nm after excitation at 313 nm with excellent repeatability and sensitivity. Owing to the high selectivity of the phosphorimetric methods, the determination of dipyridamole by RTP shows no significant interferences and is suitable for its determination in tablets.Other components of the tablets do not interfere in the phosphorescence spectra of dipyridamole. An exhaustive statistical analysis was applied to the calibration graph, including least squares regression and ANOVA. The regression line shows homoscedasticity. The validity of the overall least squares regression is proved by the ANOVA test, the variation of group means about the line, which means the lack of fit, not being significantly different from the variation within groups (pure error).Therefore, the model chosen is an adequate description of the true relationship between phosphorescence intensity and dipyridamole concentration. The proposed method was compared with the currently accepted method and the accuracies were not significantly different. The authors gratefully acknowledge financial support from Direcci�on General de Investigaci�on Cient�ýfica y T�ecnica (Project No. PB 94-0743). References 1 Moye, H. A., and Winefordner, J.D., J. Agric. Food Chem., 1965, 13, 516. 2 Inman, E. L., Jurgensen, A., Jr., and Winefordner, J. D., Analyst, 1982, 107, 538. 3 McGlynn, S. P., Neely, B. T., and Neely, C., Anal. Chim. Acta, 1963, 28, 472. Table 3 RSDs obtained by applying error propagation theory Xmi Ymi RSD (%) 100 0.1217 7.32 200 0.2441 3.58 400 0.5008 1.72 800 1.005 0.842 1200 1.495 0.591 1600 1.996 0.493 Fig. 7 Comparison between the proposed method and the currently accepted method. The ellipse is the 95% confidence region for the true slope and the intercept on the ordinate estimated from the overall least squares regression performed with the concentration calculated in reverse using both methods.The point (a, b) is the centre of the ellipse corresponding to the true intercept and estimated slope. The point (0, 1) means zero intercept and unit slope. Analyst, March 1997, Vol. 122 2574 Winefordner, J. D., and Tin, M., Anal. Chim. Acta, 1964, 31, 239. 5 Becker, R. S., Theory and Interpretation of Fluorescence and Phosphorescence, Wiley, New York, 1969. 6 Guilbault, G. G., Practical Fluorescence: Theory , Methods and Techniques, Marcel Dekker, New York, 1973. 7 Parker, C. A., Photoluminescence of Solutions, Elsevier, Amsterdam, 1968. 8 Schulman, E. G., Fluorescence and Phosphorescence Spectroscopy: Physicochemical Principles and Practice, Pergamon Press, Oxford, 1968. 9 Wehry, E. L., Modern Fluorescence Spectroscopy, Plenum Press, New York, 1976. 10 Roth, M., J. Chromatogr., 1967, 30, 276. 11 Schulman, E. M., and Walling, C., J. Phys. Chem., 1973, 77, 902. 12 Wellons, S. L., Paynter, R. A., and Winefordner, J. D., Spectrochim. Acta, Part A, 1974, 30, 2133. 13 Vo-Dinh, T., Room Temperature Phosphorimetry for Chemical Analysis, Wiley, New York, 1984. 14 Hurtubise, R. J., Solid Surface Luminescence Analysis: Theory, Instrumentation and Applications, Marcel Dekker, New York, 1981. 15 Vo-Dinh, T., and Winefordner, J. D., Appl. Spectrosc.Rev., 1977, 13, 261. 16 Parker, R. T., Freedlander, R. S., and Dunlap, R. P., Anal. Chim. Acta, 1980, 119, 189. 17 Su, S. Y., and Winefordner, J. D., Talanta, 1982, 28, 713. 18 Dalterion, R. A., and Hurtubise, R. J., Anal. Chem., 1983, 55, 1084. 19 Wandruszka, R. M. A., and Hurtubise, R. J., Anal. Chim. Acta, 1977, 93, 331. 20 Niday, G. J., and Seybold, P. G., Anal. Chem.,1978, 50, 1577. 21 Cline Love, L. J., Skrilec, M., and Habarto, J. G., Anal. Chem., 1980, 52, 754. 22 Skrilec, M., and Cline Love, L. J., Anal. Chem., 1980, 52, 1559. 23 D�ýaz Garc�ýa, M. E., and Sanz-Medel, A., Anal. Chem., 1986, 58, 1436. 24 Babary, M., Elsayed, M. A. H., and Sabry, S. M., Acta Pharm. Jugosl., 1990, 40, 513. 25 Li, H., Shenyang Yaoxueyuan Xuebao, 1984, 1, 239. 26 Shumeiko, V. A., Farm. Zh., 1977, 3, 88. 27 Hu, J., Sun, H., and Xu, Y., Huasue Shijie, 1987, 28, 19. 28 Umapathi, P., Parimoo, P., and Thomas, S. K., Indian Drugs, 1994, 31, 489. 29 Babary, M. H., Elsayed, M.A. H., Abdel-Hay, M. H., and Mohamed, S. M., Anal. Lett., 1989, 22, 1643. 30 Korany, M. A., and Haller, R., Anal. Chem., 1982, 65, 144. 31 Issa, A. S., Mahrous, M. S., Abdel Salam, M. A., and Soliman, N., Talanta, 1987, 34, 670. 32 Ganescu, I., Preda, M., Papa, I., and Vladoianu, A., Zentralbl. Pharm., Pharmakother. Laboratoriumsdiagn., 1988, 127, 577. 33 Mahrous, M. S., Issa, A. S., Abdel Salam, M. A., and Soliman, N., Anal. Lett., 1986, 19, 901. 34 Sane, R. T., Nayak, V. G., Naik, N.R., and Gupte, D. D., Indian Drugs, 1983, 20, 334. 35 Shao, Z., Pan, Z., and Gan, Z., Zhongguo Yaoxue Zazhi, 1991, 26, 225. 36 Wilczynska-Wojtulewicz, I., Chem. Anal. (Warsaw), 1977, 22, 805. 37 Steyn, J. M., J. Chromatogr., 1979, 164, 487. 38 Murillo, J. A., and Ala�n�on, A., Comput. Chem., 1993, 17, 34. 39 Massart, D. L, Vandeginste, B. G. M., Deming, S. N., and Kaufman, L., Chemometrics: a Textbook, Elsevier, Oxford, 1988. 40 Lark, P. D., Craven, B. R., and Bosworth, R.C. L., The Handling of Chemical Data, Pergamon Press, Oxford, 1968, ch. 4. 41 Miller, J. N., Analyst, 1991, 116, 3. 42 Gary, L., and Winefordner, J. D., Anal. Chem., 1983, 55, 712. 43 Clarke’s Isolation and Identification of Drugs, ed. Moffat, A. C., Pharmaceutical Press, London, 1986, pp. 562–563. Paper 6/06511B Received September 23, 1996 Accepted December 4, 1996 258 Analyst, March 1997, Vol. 122 N N N N N N N N CH2CH2OH CH2CH2OH HOCH2CH2 HOCH2CH2 Phosphorimetric Determination of Dipyridamole in Pharmaceutical Preparations Jos�e A.Murillo Pulgar�ýn*, Aurelia Ala �n�on Molina and Pablo Fern�andez L�opez Department of Analytical Chemistry and Foods Technology, University of Castilla La Mancha, 13071 Ciudad Real, Spain Room temperature phosphorescence was applied to the determination of dipyridamole in pharmaceutical preparations. The response was linear in the concentration range 100–1600 ng ml21. The use of phosphorescence enhancers such as thallium(I) nitrate (external heavy atom), sodium dodecyl sulfate (microemulsion stabilizer) and sodium sulfite (deoxygenation agent) was studied and optimized to obtain maximum sensitivity and adequate selectivity.The determination was performed in 0.026 M sodium dodecyl sulfate, 0.0156 M thallium nitrate and 0.02 M sodium sulfite. The pH value was 11.5, adjusted by adding srescence was totally developed in 15 min, after that the intensity was measured at lex = 303 nm and lem = 616 nm.The recovery of the method was tested on commercial formulations containing dipyridamole. The recoveries obtained were 94.67 ± 0.58% for Persantin and 96.75 ± 1.37% for Asasantin. The overall least squares regression method was applied to find the most exact straight line that fits the experimental data. The detection limit according to the error propagation theory was 16.4 ng ml21. The repeatability and relative standard deviation were also determined according to this theory.Keywords: Dipyridamole; pharmaceuticals; phosphorimetry; room temperature phosphorescence Phosphorimetry is a selective method for measuring different organic compounds, such as pesticides,1 PAHs2,3 and some drugs.4 Although room temperature fluorimetry is usually more sensitive than phosphorimetry, the broad band of spectra for different compounds limits the selectivity. Therefore, phosphorimetry is used because of its better selectivity over room temperature fluorescence and absorption spectrometry,5–9 since not all molecules that fluoresce will phosphoresce, often completely eliminating the spectral interference, and the phosphorescence is shifted to a less crowded spectral region.However, the difficulties associated with low temperature phosphorescence (LTP) make this technique unpopular. Owing to these difficulties, LTP has not been used extensively for the identification and determination of many compounds.Since its discovery in 1967 by Roth,10 room temperature phosphorimetry (RTP) has attracted great interest and has become a practical technique for the detection of many organic compounds.10–18 As the sample must be adsorbed on an inert substrate such as filter-paper,19,20 the technique has the disadvantages of cumbersome sample preparation, critical drying requirements and high phosphorescent background intensity from the filter-paper substrate.RTP can also be observed from many organic compounds in liquid solutions incorporating the phosphors into organized media such as cyclodextrins, micellar systems and microcrystalline media or by using sensitized RTP. In 1980, Cline Love and co-workers21,22 first investigated micelle-stabilized RTP and studied several PAHs. In a micellar solution the analytes included in the micellar assembly are apparently protected from the quenchers present in the solution. Observation of RTP in a micellar solution usually requires the presence of a heavy atom.It is placed as a counter ion outside the micelle, thus being in proximity to the hydrophobic molecules associated with the micelle. The high local concentration of the heavy atoms produces an efficient spin–orbit coupling that can diminish the fluorescence and increase the phosphorescence. Furthermore, the micelles can effectively screen molecules in the excited triplet state from the action of potential quenchers present in the bulk water phase.However, phosphorescence is not observed unless oxygen is removed, as it is a very efficient quencher that easily penetrates the micelles. D�ýaz Garc�ýa and Sanz-Medel23 have proposed the substitution of the troublesome deoxygenation with nitrogen by chemical deoxygenation with sodium sulfite, thus avoiding foam formation and ensuring more permanent protection of the solution against contamination with oxygen. In the presence of sulfite, however, the phosphorescence is not immediately observed.The oxygen in the bulk water phase is removed first, followed by that in the micelle pseudophases as it diffuses out. Equilibrium is achieved in a few minutes. In the spite of the interest in observing phosphorescence in micellar solutions, only a few analytical procedures have used this technique. In this work, the appropriate experimental conditions to obtain a reproducible and maximum phosphorescence signal, when sulfite is used to eliminate the oxygen from the micellar solutions, were studied.Dipyridamole (2,2A,2AA,2AAA-{(4,8-dipiperidinopyrimino[5,4- d]pyrimidine-2,6-diyl)dinitrilo}tetraethanol) is shown in Fig. 1. It is an intensely yellow crystalline powder. Its solution gives a yellowish blue fluorescence. It is almost insoluble in water. Dipyridamole is a vasodilator agent that is widely used in medicine. After oral administration, it stimulates a rise in the blood flow through the coronary circulation, providing more blood to the myocardium.This effect has been used in certain sports to increase energy production, such as via ATP molecules. Therefore, this vasodilator agent is classified, in doping terms, as a stimulant. Stimulants have been consumed in sports to increase efficiency and decrease tiredness. Stimulants include other substances that intensify the attention and could increase competitiveness and aggressiveness. Nevertheless, the uncontrolled use of such drugs could cause the loss of mental power and also have serious secondary effects that could cause grave danger to health. Fig. 1 Structure of dipyridamole. Analyst, March 1997, Vol. 122 (253–258) 253There have been various reports describing the spectrophotometric determination of dipyridamole in pharmaceutical tablets.24–34 In comparison with the method proposed here, these methods show low sensitivity and selectivity. The need for derivatization reactions31–34 to increase the selectivity makes these methods tedious, low accuracy is achieved and the detection limits are higher than when direct spectrophotometry is used.Only three methods have utilized the highly fluorescent nature of this drug.35–37 Although Shao et al.35 developed a direct spectrofluorimetric determination, the method has low sensitivity. The methods of Wilczynska-Wojtulewicz36 and Steyn37 require various preliminary procedures such as extraction and separation by means of TLC.We propose the application of RTP to determine dipyridamole in pharmaceutical tablets. The main advantages of using RTP over spectrophotometric and fluorimetric methods for this purpose are the high selectivity and sensitivity achieved, together with a detection limit of < 20 ng ml21 compared with approximately 2 mg ml21 and 100 ng ml21 for spectrophotometric and spectrofluorimetric methods, respectively. Experimental Apparatus All phosphorimetric measurements were performed on an Aminco Bowman Series 2 luminescence spectrometer, connected to software which runs on the OS2 operating system.The instrument utilizes a 7 W integral pulsed xenon lamp for phosphorescence measurements. The gated photomultiplier tube detection includes a unique masking method for detection in < 200 ms after the initiation of the flash lamp. Quartz glass cuvettes with pathlengths of 1.0 3 1.0 cm were used. Thermostatic equipment and a Crison Model 2001 pH meter with a glass–saturated calomel combination electrode were also used.Software The AB2 program allows the operation of the instrument to obtain excitation and emission spectra, total phosphorescence spectra and time traces, such as decay curves and time resolved curves. The Ftotal program38 was used to generate fluorescence and phosphorescence contour spectra. Statistical analysis was performed by means of a program developed by us, which has an option menu that includes all the procedures mentioned in this paper.Reagents All experiments were performed with analytical reagent grade chemicals, pure solvents and Milli-Q-purified water. Sodium dodecyl sulfate (SDS) was obtained from Sigma (St. Louis, MO, USA) and thallium(i) nitrate and sodium sulfite from Merck (Darmstadt, Germany). A stock standard solution of dipyridamole (Aldrich, Milwaukee, WI, USA) (100.0 mg disolved 100 ml of 0.1 m SDS) was diluted to prepare working standard solutions.The stock standard solution of dipyridamole was stored and protected from the light and maintained below 5 °C. Under these conditions, it was stable for at least 2 weeks. The working standard solutions of dipyridamole were stable for at least 2 d at room temperature. Stock standard solutions of 0.1 m SDS and 0.3 m thallium(i) nitrate were used. A 0.25 m sodium sulfsolution was prepared daily. Procedure For the preparation of the calibration graph, an aliquot of dipyridamole standard solution was pipetted into a 25 ml calibrated flask, then SDS as necessary to give a 0.026 m concentration, 2.5 ml of 0.025 m sodium hydroxide solution, 1.3 ml of 0.3 m thallium(i) nitrate solution and 2.0 ml of 0.25 m sodium sulfite were added.The solution was diluted to volume with water and shaken. After the flask had stood for 15 min in a thermostat at 20 °C, a portion of the solution was transferred into a phosphorescence cuvette and the RTP measured at an excitation wavelength of 303 nm and an emission wavelength of 616 nm.For the analysis of Persantin (Boehringer Ingelheim, Barcelona, Spain), 10 tablets were weighed and for the analysis of Asasantin (Boehringer Ingelheim) ten capsules (which contained inside dipyridamole in the form of minute tablets) were weighed. In all instances, the solid was powdered, homogenized and about 0.1 g was taken for analysis. Suitable dilutions were made with 0.1 m SDS. In all instances, the excipients were not soluble in SDS, so after shaking and maintaining in an ultrasonic bath for 5 min the solutions had to be centrifuged.Results and Discussion Spectral Characteristics Fig. 2 shows the total phosphorescence spectrum (solid line) and the total fluorescence spectrum (broken line) of dipyridamole. They are identical in form, varying only in the location of the emission wavelength of maximum intensity. In the Fig. 3(a) the phosphorescence (solid line) and fluorescence (broken line) excitation spectra at the emission wavelength of Fig. 2 Total phosphorescence spectrum (solid line) and total fluorescence spectrum (broken line) of dipyridamole.Fig. 3 Phosphorescence (solid line) and fluorescence (dashed line) spectra of dipyridamole. Photomultiplier voltage 800 V in phosphorescence and 410 V in fluorescence. (a) Excitation spectra; and (b) emission spectra. 254 Analyst, March 1997, Vol. 122maximum intensity are shown. Both excitation spectra are coincident, displaying two broad peaks at excitation wavelengths of 303 and 411 nm.The coincidence of the phosphorescence and fluorescence excitation spectra means an identical excitation process in both cases. This is justified in theory by the light absorption promoting an electron from the ground electronic state to the first and second excited singlets. The peak at 303 nm shows a higher luminescence intensity than that at 411 nm and the former excitation wavelength was therefore chosen for phosphorescence measurements.Fig. 3(b) shows the phosphorescence (solid line) and fluorescence (broken line) emission spectra. The fluorescence spectrum gives a band with a characteristic emission wavelength of 488 nm, corresponding to the transition from the first excited singlet to the ground state. The phosphorescence spectrum presents a band at 616 nm corresponding to the transition from the excited triplet to the singlet ground state.The fluorescence and phosphorescence spectra show different characteristic emission wavelengths. This difference of 128 nm gives an idea of the non-radiant energy that is lost in the intersystem crossing and the subsequent vibrational relaxation to the lowest vibrational level of the excited triplet. The phosphorescence lifetime of dipyridamole is approximately 1324 ms. This is the time required for the population of the excited triplet state to decrease to 1/e of its original value after the excitation source has been turned off.The lifetime is a means of considering the luminescence process in terms of rates. Fluorescence lifetimes are typically of the order of 1–20 ns. As phosphorescence is a spin forbidden process, phosphorescence lifetimes are considerably longer, generally ranging from milliseconds to seconds. Consequently, the phosphorescence lifetime is a measure of the forbiddenness of singlet–triplet transitions in a given molecule.Factors Affecting Phosphorescence Chemical variables were optimized to obtain maximum phosphorescence sensitivity and adequate selectivity. Dipyridamole is almost insoluble in water, so it was necessary to use an organic solvent. Owing to the micellar properties of SDS it was used to dissolve dipyridamole; further, the semirigid structure of the solution favours the development of phosphorescence. The extremely high sensitivity of the triplet state to quenching by oxygen requires deoxygenation of the sample.Owing to the high efficiency of this quenching process, it is necessary to eliminate totally the oxygen in the micellar solution. The method proposed for sample deoxygenation is based on the redox reaction of sulfite with molecular oxygen to product sulfate.23 It was observed that a concentration of Na2SO3 of 0.02 m was required to eliminate oxygen completely from the solutions, which was evident from the intensity of the phosphorescence signal.This decreases at higher concentrations of Na2SO3. The decrease in the signal for higher concentrations of Na2SO3 has been interpreted as the displacement of thallium(i) from the micelle because of the high concentration of sodium in the solution.23 An Na2SO3 concentration of 0.02 m was chosen. The influence of pH on phosphorescence intensity was studied by adding different amounts of H2SO4 and NaOH to a dipyridamole solution. The phosphorescence is not significant at pH values up to 9.5.In Fig. 4 phosphorescence intensity is plotted versus the pH of the solution and, as can be readily observed, dipyridamole has maximum intensity at pH 11.5. Phosphate buffer was first used to adjust this pH value, but the phosphorescence intensity diminished. This is a quenching situation due to the formation of exciplexes. The explanation for this behaviour is that an excited dipyridamole molecule associates with the phosphate in the ground state. Consequently, the phosphorescence intensity of excited dipyridamole diminishes.This pH was subsequently adjusted by adding 2.5 ml of 0.025 m NaOH. It was also necessary to establish how the phosphorescence intensity varies with changes in thallium concentration. Fig. 5(a) shows this relationship for thallium concentrations between 0.0024 and 0.0168 m. As can be readily observed, the phosphorescence intensity increases with increase in thallium concentration. For concentrations of thallium above 0.0168 m precipitation is observed.For lower concentrations of thallium, the heavy atom effect is very sensitive. When the thallium salt is not present the phosphorescence intensity disappears. A concentration of thallium of 0.0144 m gives high intensity without problems of precipitation. The effect of SDS concentration was investigated by preparing samples with SDS concentrations between 0.014 and 0.074 m. As Fig. 5(b) shows, phosphorescence intensity of dipyridamole diminishes as the SDS concentration increases.For SDS concentrations lower than 0.014 m precipitation occurred. A 0.022 m concentration of SDS was therefore selected, giving good sensitivity while being sufficient to dissolve the dipyridamole. From the above two experiments, the phosphorescence intensity is maximum with SDS and thallium concentrations in Fig. 4 Variation of phosphorescence intensity of dipyridamole with pH. The pH was varied from 9.5 to 12.5 by adding sodium hydroxide. Fig. 5 (a) Effect of thallium(i) concentration on phosphorescence intensity at an SDS concentration of 0.022 m; (b) effect of SDS concentration on phosphorescence intensity at a thallium(i) concentration of 0.0144 m; and (c) effect of SDS and thallium(i) on phosphorescence intensity at an SDS to thallium(i) ratio of 1.667. Analyst, March 1997, Vol. 122 255the proportion of 1.667, and Fig. 5(c) shows the phosphorescence intensity versus SDS concentration with this proportion maintained.As can be observed, the phosphorescence intensity increases as the concentration of SDS increases, reaching a constant value at SDS concentrations above 0.025 m. Concentrations of SDS of 0.026 m and of thallium(i) 0.0156 m were selected as adequate. At concentrations of SDS above 0.032 m precipitation occurred. Another factor that affects the phosphorescence intensity is temperature, the phosphorescence intensity decreasing when the temperature increased from 15 to 79 °C.This decrease is measured by plotting the relative signal increment versus temperature, i.e., the intensity at each temperature minus the intensity at the lowest temperature, divided by the intensity of the higher temperature and multiplied by 100. This relationship shows a linear behavior, the slope (which is the temperature coefficient of dipyridamole) being 1.36% °C21, the intercept on the ordinate 104.4% and the coefficient of the determination r2 = 0.995.The influence of dipyridamole concentration on the phosphorescence intensity was studied under the above conditions. The phosphorescence intensity reaches a constant value at dipyridamole concentrations above 20 mg ml21. The dipyridamole concentration range for a linear relationship between phosphorescence intensity and concentration was found to be up to 1600 ng ml21. The inner filtering effect is significant for concentrations of dipyridamole above 1600 ng ml21, and increases as the dipyridamole concentration increases.Consequently, calibration was performed for dipiridamole concentrations up to 1600 ng ml21 with three replicates per point. Determination of Dipyridamole Under the operating conditions outlined above, we propose a method to determine dipyridamole by direct measurement of phosphorescence intensity with an emission wavelength of 616 nm and an excitation wavelength of 303 nm in the concentration range 100–1600 ng ml21. The delay time required was 120 ms with the photomultiplier tube masked.The gate time appropriate for this delay time was 800 ms and the detector voltage was 740 V. Both excitation and emission bandpasses were 16 nm and the scan rate was 5 nm s21. A calibration graph was constructed with three replicates per point. Fig. 6 shows the average emission spectra of the calibration concentrations. The proposed method was evaluated by a statistical analysis of the experimental data by fitting the overall least squares line according to y = a + bx.39,40 Table 1 gives the results from the statistical analysis.The calibration line presents homoscedasticity (standard residuals have a uniform variance) and therefore it is not necessary to weight the phosphorescence intensity values according to the mean standard deviation. In order to test the linearity of the overall least squares regression, the ANOVA test was performed.39 The results are given in Table 2. When the 95% confidence region for true slope and intercept40 estimated is represented, the zero intercept on the ordinate falls within the joint confidence region.This means that the intercept on the ordinate is not significantly different from zero. Therefore, the relationship between phosphorescence intensity and dipyridamole concentration is proportional. The confidence interval for the corresponding slope is 1.257 3 1023–1.238 3 1023. Nevertheless, the significance of the intercept was tested by applying Student’s t-test (Student’s t-test is more restrictive than the confidence region of the true slope and intercept).The experimental t-value obtained was 0.324, which is less than theoretical value of 2.12. If the theory of error propagation is considered, the values of the detection and determination limits are consistent with the reliability of the blank measurements and the signal measurements of the standards.41,42 In this case a detection limit of 16.4 ng ml21 and a determination limit of 54.6 ng ml21 were obtained.The detection limit according to Clayton considers the probability of false positive and false negative values, the detection limit being 18.8 ng ml21. In order to study the precision of the method, a series of 10 solutions of 800 ng ml21 of dipyridamole were measured on the Table 1 Results from statistical analysis of data; least squares regression with replicates Intercept on ordinate (a) 21.705 3 1023 SD of intercept on ordinate (sa) 5.3 3 1023 Slope (b) 1.249 3 1023 SD of slope (sb) 5.8 3 1026 SD of regression (syx) 1.3 3 1022 Coefficient of determination (r2) 0.9996 Confidence interval of intercept on ordinate 9.447 3 1023, 21.286 3 1022 Confidence interval of slope 1.262 3 1023, 1.237 3 1023 Slope without intercept on ordinate 1.248 3 1023 Table 2 ANOVA test: linearity test Source of variation SS* DF* MS* Fexp. Ftheoretical Due to regression 8.2804 1 8.2804 Set means about the line (lack of fit) 2.6916 3 1024 4 6.7290 3 1025 0.31 3.26 Within line (pure error) 2.6363 3 1023 12 2.1969 3 1024 Total 8.2833 17 0.4873 * SS = Sum of squares of deviation; DF = degrees of freedom; MS = SS DF.Fig. 6 Set of room temperature phosphorescence spectra. Dipyridamole concentration: 1, 1600; 2, 1200; 3, 800; 4, 400; 5, 200; and 6, 100 ng l21. 256 Analyst, March 1997, Vol. 122same day. By applying the IUPAC definition, the mean standard deviation of replicates was 3.88 ng ml21 and the relative error 1.1%, whereas if error propagation is assumed, the SE obtained was 2.59 ng ml21 and the relative error 0.71% (95% confidence level).In this case the relative error for replicates is less on applying the SE of the regression line, owing to the classical hyperbolic shape of the confidence and dispersion bands of the true line which are closest at the mean concentration, this being very near to 800 ng ml21, where the standard deviation is reduced to the standard deviation of the mean, and therefore, the relative error is less.To estimate concentration values at greater distances from the mean concentration of the regression line, the error propagation method gives the greatest values for the relative error. The accuracy of the method was established by testing the analytical signal corresponding to the concentrations of the calibration line. As can be seen from the RSDs in Table 3, the values obtained by applying the error propagation theory confirm that, to calculate the confidence interval of a measurement, the error propagation method may be used.The determination of dipyridamole by RTP was compared with the currently accepted method.37,43 In this, 1 ml of the problem solution is extracted at pH 10 with 8 ml of hexane– isoamyl alcohol (95 + 5) and the organic phase used directly for measurement of the fluorescence intensity (lex = 405 and lem = 495 nm). A comparison study of 10 samples of dipyridamole was performed by applying least squared pair analysis40 with the concentration calculated in reverse by means of the currently accepted method and the proposed method. As both methods should give the same concentrations for the same samples, a zero intercept on the ordinate and a slope of units must be obtained.Fig. 7 shows the 95% confidence region for the true slope and intercept estimated. As can be observed, the point corresponding to the zero intercept and unit slope falls within the joint confidence region.This means that the intercept is not significantly different from zero and the slope is not significant different from unity. Consequently, the accuracies of proposed and currently accepted method are not significantly different. Applications and Interference Study The recommended procedure was applied satisfactorily to the determination of dipyridamole in the Spanish pharmaceutical products that contain this vasodilator agent in different proportions (Persantin 100 mg and Asasantin 75 mg).The assay results expressed as a percentage of the nominal contents resulting from the average of three determinations of three different tablets were 94.67% for Persantin and 96.75% for Asasantin, the standard deviations being 0.58% and 1.37%, respectively. The recoveries agree well enough with the nominal content and the precision is quite satisfactory The specificity of this determination was studied by adding related drugs to the pharmaceutical preparation of dipyridamole and other drugs with intrinsic fluorescence and testing to see if the added drugs cause interference in the dipyridamole quantification.In the first case, dipyridamole (1600 ng ml21) was spiked with saccharose, lactose and glucose (20 mg ml21). No interferences were found. In the second, the added drugs were aspirin, atenolol, spironolactone and canrenoic acid (1 mg ml21), amiloride, furosemide and triamterene (25 ng ml21), aspirin and nadolol (2 mg ml21), metoprolol (80 ng ml21) and caffeine, quinidine and quinine (1.6 mg ml21).Except for the furosemide, no significant variation of the analytical signal was observed from the value expected when dipyridamole is present alone. Nevertheless, interferences were observed if the concentrations of these drugs were increased. Conclusions A method for direct phosphorimetric determination of the vasodilator agent dipyridamole in pharmaceutical preparations has been developed. The determination can be performed by measuring phosphorescence intensity within an emission wavelength of 616 nm after excitation at 313 nm with excellent repeatability and sensitivity.Owing to the high selectivity of the phosphorimetric methods, the determination of dipyridamole by RTP shows no significant interferences and is suitable for its determination in tablets. Other components of the tablets do not interfere in the phosphorescence spectra of dipyridamole. An exhaustive statistical analysis was applied to the calibration graph, including least squares regression and ANOVA.The regression line shows homoscedasticity. The validity of the overall least squares regression is proved by the ANOVA test, the variation of group means about the line, which means the lack of fit, not being significantly different from the variation within groups (pure error). Therefore, the model chosen is an adequate description of the true relationship between phosphorescence intensity and dipyridamole concentration. The proposed method was compared with the currently accepted method and the accuracies were not significantly different.The authors gratefully acknowledge financial support from Direcci�on General de Investigaci�on Cient�ýfica y T�ecnica (Project No. PB 94-0743). References 1 Moye, H. A., and Winefordner, J. D., J. Agric. Food Chem., 1965, 13, 516. 2 Inman, E.L., Jurgensen, A., Jr., and Winefordner, J. D., Analyst, 1982, 107, 538. 3 McGlynn, S. P., Neely, B. T., and Neely, C., Anal. Chim. Acta, 1963, 28, 472. Table 3 RSDs obtained by applying error propagation theory Xmi Ymi RSD (%) 100 0.1217 7.32 200 0.2441 3.58 400 0.5008 1.72 800 1.005 0.842 1200 1.495 0.591 1600 1.996 0.493 Fig. 7 Comparison between the proposed method and the currently accepted method. The ellipse is the 95% confidence region for the true slope and the intercept on the ordinate estimated from the overall least squares regression performed with the concentration calculated in reverse using both methods.The point (a, b) is the centre of the ellipse corresponding to the true intercept and estimated slope. The point (0, 1) means zero intercept and unit slope. Analyst, March 1997, Vol. 122 2574 Winefordner, J. D., and Tin, M., Anal. Chim. Acta, 1964, 31, 239. 5 Becker, R. S., Theory and Interpretation of Fluorescence and Phosphorescence, Wiley, New York, 1969. 6 Guilbault, G. G., Practical Fluorescence: Theory , Methods and Techniques, Marcel Dekker, New York, 1973. 7 Parker, C. A., Photoluminescence of Solutions, Elsevier, Amsterdam, 1968. 8 Schulman, E. G., Fluorescence and Phosphorescence Spectroscopy: Physicochemical Principles and Practice, Pergamon Press, Oxford, 1968. 9 Wehry, E. L., Modern Fluorescence Spectroscopy, Plenum Press, New York, 1976. 10 Roth, M., J. Chromatogr., 1967, 30, 276. 11 Schulman, E. M., and Walling, C., J. Phys. Chem., 1973, 77, 902. 12 Wellons, S. L., Paynter, R. A., and Winefordner, J. D., Spectrochim. Acta, Part A, 1974, 30, 2133. 13 Vo-Dinh, T., Room Temperature Phosphorimetry for Chemical Analysis, Wiley, New York, 1984. 14 Hurtubise, R. J., Solid Surface Luminescence Analysis: Theory, Instrumentation and Applications, Marcel Dekker, New York, 1981. 15 Vo-Dinh, T., and Winefordner, J. D., Appl. Spectrosc. Rev., 1977, 13, 261. 16 Parker, R. T., Freedlander, R. S., and Dunlap, R. P., Anal. Chim. Acta, 1980, 119, 189. 17 Su, S. Y., and Winefordner, J. D., Talanta, 1982, 28, 713. 18 Dalterion, R. A., and Hurtubise, R. J., Anal. Chem., 1983, 55, 1084. 19 Wandruszka, R. M. A., and Hurtubise, R. J., Anal. Chim. Acta, 1977, 93, 331. 20 Niday, G. J., and Seybold, P. G., Anal. Chem.,1978, 50, 1577. 21 Cline Love, L. J., Skrilec, M., and Habarto, J. G., Anal. Chem., 1980, 52, 754. 22 Skrilec, M., and Cline Love, L. J., Anal. Chem., 1980, 52, 1559. 23 D�ýaz Garc�ýa, M. E., and Sanz-Medel, A., Anal. Chem., 1986, 58, 1436. 24 Babary, M., Elsayed, M. A. H., and Sabry, S. M., Acta Pharm. Jugosl., 1990, 40, 513. 25 Li, H., Shenyang Yaoxueyuan Xuebao, 1984, 1, 239. 26 Shumeiko, V. A., Farm. Zh., 1977, 3, 88. 27 Hu, J., Sun, H., and Xu, Y., Huasue Shijie, 1987, 28, 19. 28 Umapathi, P., Parimoo, P., and Thomas, S. K., Indian Drugs, 1994, 31, 489. 29 Babary, M. H., Elsayed, M. A. H., Abdel-Hay, M. H., and Mohamed, S. M., Anal. Lett., 1989, 22, 1643. 30 Korany, M. A., and Haller, R., Anal. Chem., 1982, 65, 144. 31 Issa, A. S., Mahrous, M. S., Abdel Salam, M. A., and Soliman, N., Talanta, 1987, 34, 670. 32 Ganescu, I., Preda, M., Papa, I., and Vladoianu, A., Zentralbl. Pharm., Pharmakother. Laboratoriumsdiagn., 1988, 127, 577. 33 Mahrous, M. S., Issa, A. S., Abdel Salam, M. A., and Soliman, N., Anal. Lett., 1986, 19, 901. 34 Sane, R. T., Nayak, V. G., Naik, N. R., and Gupte, D. D., Indian Drugs, 1983, 20, 334. 35 Shao, Z., Pan, Z., and Gan, Z., Zhongguo Yaoxue Zazhi, 1991, 26, 225. 36 Wilczynska-Wojtulewicz, I., Chem. Anal. (Warsaw), 1977, 22, 805. 37 Steyn, J. M., J. Chromatogr., 1979, 164, 487. 38 Murillo, J. A., and Ala�n�on, A., Comput. Chem., 1993, 17, 34. 39 Massart, D. L, Vandeginste, B. G. M., Deming, S. N., and Kaufman, L., Chemometrics: a Textbook, Elsevier, Oxford, 1988. 40 Lark, P. D., Craven, B. R., and Bosworth, R. C. L., The Handling of Chemical Data, Pergamon Press, Oxford, 1968, ch. 4. 41 Miller, J. N., Analyst, 1991, 116, 3. 42 Gary, L., and Winefordner, J. D., Anal. Chem., 1983, 55, 712. 43 Clarke’s Isolation and Identification of Drugs, ed. Moffat, A. C., Pharmaceutical Press, London, 1986, pp. 562–563. Paper 6/06511B Received September 23, 1996 Accepted December 4, 1996 258 Analyst, Marc

ISSN:0003-2654    

DOI:10.1039/a606511b

出版商:RSC    

年代:1997

数据来源: RSC