摘要:

Tutorial Review Theoretical considerations of chemical reactions in micro-reactors operating under electroosmotic and electrophoretic control Paul D. I. Fletcher,* Stephen J. Haswell and Vesselin N. Paunov Department of Chemistry, University of Hull, Hull, UK HU6 7RX. E-mail: p.d.fletcher@chem.hull.ac.uk Received 6th May 1999, Accepted 7th July 1999 Summary of contents 1 Introduction 2 Electroosmotic flow and electrophoresis in micro-reactor manifolds 3 Micro-reactor manifold configuration 4 Voltage conditions for loading, flow and injection 5 Control of chemical reactions under electroosmotic and electrophoretic flow 6 Example calculations 7 Conclusions and future outlook 8 Appendix 8.1 Numerical algorithm for the solution of eqns.(11)–(13) 8.2 Symbols 9 References 1 Introduction Since the introduction of the concepts of micro flow injection analysis (mFIA) and micro total analytical systems (mTAS) nearly a decade ago;1–3 few researchers in the field of analytical science can fail to have been impressed by their impact, particularly in the area of DNA diagnostics.4–10 From the literature one is able to trace the pioneering developments of fabrication11–19 through detection20–26 and separation23,24,27–32 to sample preparation,5,33–35 culminating, for example, in a recent paper by Waters et al.34 which describes a fully integrated mTAS device for DNA characterisation.Whilst the majority of mFIA and mTAS studies have been focused on their application as capillary electrophoresis (CE) separation systems, 23,24,27–32 the opportunity exists to extend such microreactor technology into the concept of ‘Lab-on-a-Chip’.36–39 In this approach, the possibility exists of using a microfabricated system for the full characterisation of a wide range of chemical processes.Realisation of this goal requires a better understanding of the fluidics of chemically reacting systems in microreactors. Whilst hydrodynamically pumped systems have been described in the literature,40–43 it has been the application of electrokinetic based fluidic pumping that has dominated previous studies.44–46 This clear trend can be attributed to Paul Fletcher is Professor of Physical Chemistry at the University of Hull.He leads the Surfactant Science Group within the Department of Chemistry and his research interests include micro-reactors and the surface and colloid chemistry of surfactant systems such as micelles, monolayers, microemulsions, emulsions and foams and reactions in complex media. He has published over 100 papers in these areas.E-mail: P.D.Fletcher@chem.hull.ac.uk Stephen Haswell is Professor of Analytical Science at the University of Hull. His current research activities are in the areas of micro-reactor and mFIA development, microwave enhanced reaction chemistry, trace elemental speciation and process analysis. He is author of over 100 research papers, a number of books and patents and is widely known nationally and internationally for his enthusiastic lectures.For a number of years one of the underlying principles of Professor Haswell’s research has been to break down the sectorial walls which exist in science, in particular, the integration of analytical science with main line chemistry, physics, engineering and biology. Many of these ideals are encompassed in the development of micro-chemical processes and devices, part of which is reflected in this current review.E-mail: S.J.Haswell@chem.hull.ac.uk Vesselin N. Paunov is a Postdoctoral Fellow at the Department of Chemical Engineering of the University of Delaware. His research interests include surface forces and stability of foams and emulsions, wetting and spreading, capillary forces between colloidal particles, electrokinetic phenomena, phase behaviour and microstrucutre of complex fluids, and computer modelling of physicochemical processes and micro-reactors.He has published over 25 papers in these areas. E-mail: paounov@che.udel. edu Analyst, 1999, 124, 1273–1282 1273factors such as the experimental simplicity in achieving electroosmotic flow, i.e., no moving parts, and minimal backpressure effects, with the added dimension of superimposed electrophoretic separations. It follows, therefore, that a good basic understanding of the nature and capability of electrokinetic based devices is fundamental to the design and development of future applications. A number of theoretical and experimental studies of hydrodynamic and electrokinetic flow within mFIA and mTAS, including channel switching and velocity profile control, have been reported.47–52 In this tutorial review, we describe the basic theoretical considerations governing liquid phase chemical reactions in micro-reactor manifolds using electrokinetic based fluidics.The calculations described demonstrate how the voltages applied may be used to control both the spatial (i.e., lengthways along the channel long axis) and temporal evolution of chemical components and reaction products.The combination of spatial and temporal control of reactions, realisable in such microreactor manifolds (but not, for example, in microtitre wells), offers many potential advantages such as identifying the optimum detector position and the best point at which to perform reagent additions in catalytic systems. The purpose of this work is to review the quantitative theoretical basis for this type of control and to provide illustrative calculations to guide the design and development of novel micro-reactor systems.The paper is organised as follows. First, the basic principles of electroosmotic flow (EOF) and electrophoresis are described. We then consider a specific micro-reactor manifold configuration and show how the operating voltages can be adjusted to control the loading, injection and flow phases necessary to investigate analytical type chemical reactions.The next section details the equations describing the spatial and temporal evolution of chemical reactions under EOF and electrophoretic control. Numerical results are then presented which illustrate the main features of the behaviour of chemical reactions under voltage control. Finally, the conclusions and the outlook for the future are discussed. 2 Electroosmotic flow and electrophoresis in micro-reactor manifolds Fabrication of mFIA or mTAS micro-reactor manifolds involves creating a network of micron sized channels in a solid substrate surface using either wet etch, laser ablation, embossing, micromachining or microlithography techniques.11–19,53–56 Suitable substrates include materials such as glass, oxidised silicon and various plastics which support EOF.In most fabrication procedures a top cover is then bonded to the substrate using anodic or fusion bonding.53,56 Holes drilled through the top cover allow connection to the channels and also form the reagent reservoirs.Voltages to drive EOF are applied through electrodes placed within the reservoirs. With this form of fabrication the channels approximate to a rectangular crosssection with depths in the range 10–200 mm, widths of 50–200 mm and lengths in the centimetre range, as shown in Fig. 1. The reagent reservoirs are typically 1 mm in diameter and 1 cm in depth. A plan view of an entire micro-reactor manifold (as used in recent analytical studies57,58) is shown in Fig. 1. For the purposes of this paper, we shall consider parameters appropriate to glass micro-reactor manifolds containing aqueous solutions of reactant species X and Y that react to form a product Z which can be detected colorimetrically. For pure electroosmotic flow within a channel, the velocity profile across the channel is uniform except for the region very close to the channel wall.50,59,60 The thickness over which the velocity is non-uniform is of the order of the Debye length and is in the nanometer range.The linear liquid velocity far from the walls due to electroosmosis vos is60 v E os = - ee z h 0 (1) where E is the electric field (equal to the voltage divided by the distance between the electrodes for channels of uniform resistance per unit length), e is the relative permittivity of the liquid, e0 is the permittivity of free space, z is the zeta potential of the channel/liquid interface and h is the liquid viscosity.The negative sign indicates that when z is negative, the diffuse charge in the liquid is positive and so the liquid flow is towards the negative electrode. The volumetric flow rate due to electroosmotic flow is Vos = Achannelvos, where Achannel is the cross-sectional area of the channel. The electric current I transported by the liquid is proportional to vos according to I A v = channel os 0 hl ee z 0 (2) where l0 is the electrical conductivity of the liquid.Eqn. (2) neglects the possibility of surface conduction, which can be a complicating factor under some conditions.60 It can be seen that the EOF is primarily controlled by the zeta potential at the channel wall/solution interface. For aqueous solutions in glass channels the zeta potential varies from zero at pH Å 2 to about 2100 mV at pH 7.61–64 At fixed pH, the magnitude of z decreases with increasing concentration of most common electrolytes.61–64 It should be noted that the presence of species such as cationic surfactants, which adsorb strongly at the glass/water interface, can strongly influence z and hence the EOF.65 The zeta potential is also sensitive to the nature of the glass and its treatment.61 For aqueous solutions around pH 7 (for which z is around 2100 mV) with E of the order of 100 V cm21, vos is of the order of mm s21.Fig. 1 Plan view of the basic micro-reactor configuration (not to scale). The reservoir diameters are typically 1 mm, the channel widths are typically 100 mm and the side length of the microreactor is typically 2 cm.The middle diagram shows channel BC filled with Y. The bottom diagram shows the experimental configuration modelled here in which a stream of X containing a slug of Y (shown in 3D view in inset) is moving by EOF towards reservoir D. 1274 Analyst, 1999, 124, 1273–1282Note that pure electroosmotic flow is only obtained in the absence of a pressure difference DP across the tube.If DP is not zero, one obtains a combination of electroosmotic and pressure driven flow. Since pressure driven flow (in a cylindrical channel) shows a parabolic velocity profile, a flat velocity profile across the channel is only obtained when DP is negligibly small. Experimentally, non-negligible pressure differences may be caused by a difference in liquid levels between the inlet and outlet reservoirs (Dhres), by Laplace pressure differences resulting from the curved liquid menisci in the inlet and outlet reservoirs or by obstruction within the channel leading to a back-pressure.The Laplace pressure change across the liquid menisci within the reservoirs is equal to 2g/r, where g is the liquid/air surface tension and r is the radius of curvature of the liquid meniscus. As in capillary rise phenomena,66 the radius of curvature of the liquid meniscus depends on both the radius of the reservoir containing the surface (rr) and the contact angle q made by the liquid with the reservoir wall according to r = rr/ cosq.For pure water of tension 72 mN m21 making a contact angle of 0° within a cylindrical reservoir of radius 1 mm, the Laplace pressure is approximately 140 Pa. The magnitude of this Laplace pressure, equivalent to the hydrostatic pressure exerted by a column of water of approximately 14 mm in height, can be significant in considerations of flow within microreactors. Within a micro-reactor, a non-zero DP from Laplace effects arises only when the Laplace pressure differences across the inlet and outlet reservoirs are not equal.Since the liquid menisci within the inlet and outlet reservoir are normally similar, DP values from Laplace effects are generally expected to be considerably smaller than the value quoted above for a single meniscus. For a cylindrical reservoir containing an electrode, the meniscus shape is complex and the Laplace pressure will depend on the positioning of the electrode within the reservoir, the contact angles of the liquid with the electrode and the reservoir wall in addition to the tension.The value of DP arising from Laplace pressure differences can be minimised by matching reservoir diameters and electrode positioning for the inlet and outlet reservoirs as far as possible. The Laplace pressure can be reduced to zero if the reservoir diameter is made sufficiently large such that the liquid surface contains a flat region.This situation applies when the reservoir radius is much greater than the length scale over which the liquid meniscus is curved, i.e., the capillary length equal to Ag/Drg, where Dr is the density difference between the liquid and air and g is acceleration due to gravity. In order to ensure that ‘pure’ EOF is obtained within a channel, it is necessary to consider the limits of DP within which the pressure driven component of the total flow can be considered negligible relative to that from EOF.In the case that Laplace pressures and channel obstruction effects are absent, i.e., DP arises only from hydrostatic pressure resulting from a difference in reservoir liquid height Dhres (DP =DhresDrg), this can be estimated as follows. We consider the magnitude of Dhres sufficient to produce a pressure driven volumetric flow rate Vpress equal to Vos. We estimate Vpress for laminar flow within a cylindrical channel of ‘effective’ radius reff such that preff 2 = Achannel and equate this with Vos: V h gr l V V l press res eff channel os 0 channel = = = pD r h ee z h 4 8 (3) where lchannel is the channel length held between the reservoirs and V is the voltage applied between the reservoirs.Rearrangement of eqn. (3) shows that, for Vpress to be less than 10% of Vos, then Dh V gr res 0 eff 2 @ ee x r 10 (4) Inspection of eqn. (4) shows that for many mTAS operating conditions described in the literature, Dhres may have to be less than 1 mm of water pressure in order to suppress pressure driven flow.Depending on the conditions (e.g., whether the feed reservoir height is greater or less than the destination reservoir), the pressure driven flow may either accelerate or retard the EOF. For either acceleration or retardation, pressure driven flow will perturb the flat velocity profile expected for ‘pure’ EOF. Disturbances of EOF by pressure effects have been demonstrated experimentally by Boer et al.46 Lack of proper control of these small pressure differences is expected to lead to irreproducible and erratic experimental results.In addition to EOF, charged species within the channels move under the influence of the electric field by electrophoresis. The electrophoretic velocity vph of a species is given by67 v zeED kT ph = (5) where z is the number of electronic charges on the species (positive for cations, negative for anions), e is the electronic charge (magnitude only), D is the diffusion coefficient, k is the Boltzmann constant and T is the absolute temperature.The total velocity of a particular species is simply the vector sum of that due to the electroosmosis and electrophoresis, i.e., vtotal = vph + vos. For aqueous solutions in a glass channel (where z is negative), a positive value of vos signifies movement towards the negative electrode. In this situation, the electrophoretic velocities of cations (z positive) are increased relative to vos whereas the velocities of anions are decreased.For common small ions, the magnitudes of vph and vos, both of which scale with E, are generally similar and in the mm s21 range. 3 Micro-reactor manifold configuration In this section we present explicit calculations for the simple micro-reactor manifold configuration shown as a plan view in Fig. 1. It consists of four reservoirs (A, B, C and D) each containing an electrode and connected by etched channels.The lengths of the different channel sections are specified here by reference to the letters marking the corner points as shown in Fig. 1. We shall consider a bimolecular, reversible reaction which, in analytical terms, could be the formation of a chromogenic complex: X + Y " Z (6) The reaction has a forward, second-order rate constant kf and reverse, first-order rate constant kr. For the configuration shown, detection of the reaction progress is provided by optical absorbance measurements along the EH channel section situated between fibre optics connected to a spectrophotometer.Although the theoretical results presented here refer to the specific configuration of Fig. 1, the calculation approach may be easily modified to apply to a very wide range of manifold designs with more complex channel labyrinths and different detection systems. The reactant species X and Y are introduced into the microreactor as follows. Initially, all reservoirs and channels are filled with solvent.The solution of reactant X is introduced into reservoir A and a suitable voltage is applied across the electrodes within the reservoirs A and D to fill the AD channel with X (the ‘loading’ phase). Reactant Y is introduced into reservoir B and a voltage across BC is used to fill the BC channel (the ‘injection’ phase) as shown in Fig. 1 (middle). For the experimental situation to be modelled, a voltage is reapplied across AD which mobilises the stream of X, now containing a slug of Y, in the EH channel (the ‘flow’ phase, Fig. 1, bottom). In fact, as will be discussed in detail in the next section, suitable voltages across both AD and BC must be applied simultane- Analyst, 1999, 124, 1273–1282 1275ously for the loading, flow and injection phases in order to achieve a ‘clean’ injection. The progress of the reaction of the slug of Y within the flowing stream of X is then monitored by the fibre optic spectroscopic detection systems and is considered in detail in the calculations presented later. 4 Voltage conditions for loading, flow and injection As discussed by Seiler et al.,49 dc circuit analysis (using Kirchhoff’s rules) can be used to predict the variation with applied voltages of the electrical currents, and hence EOFs, in the different channel sections of a manifold. The manifold configuration of Fig. 1 can be represented as the equivalent dc circuit shown in Fig. 2. The circuit consists of two voltage sources VAD and VBC (supplied by the electrode pairs in reservoirs AD and BC, respectively) connected by the appropriate channel sections which form resistance elements RIJ, where the subscripts signify the particular channel section.The overall circuit contains two loops which both contain RFG. We assume here that all channel sections have a uniform crosssectional area and zeta potential and all contain liquid of identical conductivity. Under these conditions, easily achievable with solutions containing low concentrations of reacting species in a relatively high concentration of inert electrolyte, the resistance of a channel section is proportional to its length.We note that the approach could be extended to the more complex case where the resistance per unit length of the channel is not constant in different parts of the manifold. Neglecting surface conductivity, the electrical currents (proportional to the EOF velocity as discussed earlier) in the arms of loops 1 and 2 are I1 and I2, respectively.The current in the FG channel section (common to both loops) is I1 + I2. Summing the product of current and resistances around each loop gives the following pair of equations: VAD = (RAE + REF + RGH + RHD + RFG)I2 + RFGI1 = Rloop 2I2 + RFGI1 VBC = (RCG + RFB + RFG)I1 + RFGI2 = Rloop1I1 + RFGI2 (7) Solving the simultaneous equations and rearranging yields the following expressions for I1 and I2: I V V R R R R R I V V R R R R R 1 BC AD FG loop 2 loop 1 FG 2 loop 2 AD BC FG loop 1 loop 2 FG 2 loop 1 = - - = - - / / / / 2 (8) The current through the FG channel section is I1 + I2.Eqn. (8) allows the currents, and hence the EOF, in each channel section to be calculated for any voltages provided that the resistances of the different channel sections are known. For the loading and flow phases, we require EOF (i.e., finite current I2) between reservoirs A and D with zero EOF between reservoirs B and C (i.e., current I1 equal to zero) to avoid contamination of one reactant stream with the other.Similarly, for the injection phase, finite current I1 and zero current I2 are required. Inspection of eqn. (8) shows that, in order to obtain zero I1, the voltages VAD and VBC must obey the relationship VBC = VADRFG/Rloop 2 = VADlengthFG/lengthloop 2 (9) The second equality is valid in the case that the resistance per channel length is constant. Similarly, to obtain zero I2, we require VAD = VBCRFG/Rloop 1 = VBClengthFG/lengthloop 1 (10) Hence, in order to obtain a ‘clean’ injection of a slug of reactant Y into a stream of X, both voltages VAD and VBC must be switched synchronously between the values required [and calculated using eqns. (9) and (10)] for the loading, injection and flow phases.This highlights the necessity for automated computer control of the applied voltages in micro-reactor devices. It is, of course, possible to operate the micro-reactor such that the contents of reservoirs A and B are made to flow into the detection channel EH in different ratios controlled by the applied voltages (as opposed to the load, inject, flow sequence described above).Fig. 3 shows the ratio I2/I1 for different ratios of the applied voltages VAD/VBC at constant VBC. Since the ratio of electrical currents is equal to the ratio of electroosmotic flow rates, the plot demonstrates that the mixing ratio of the flowing streams can be varied continuously by adjustment of the voltage ratio.(It should be noted that the current and voltage ratios in Fig. 3 are invariant with the absolute magnitudes of either current or voltages.) Fig. 3 also shows a similar plot in which the voltage VAD is held constant. These calculations demonstrate that, in principle, the applied voltages can be used to vary Fig. 2 The dc circuit equivalent to the micro-reactor configuration of Fig. 1. The rectangular boxes are the resistance elements arising from the channel sections marked by the subscripts. Fig. 3 Variation of I2/I1 with VAD/VBC (a) and I1/I2 with VBC/VAD (b). The calculations are for the micro-reactor configuration of Fig. 1 (assuming constant resistance per channel length) for channel lengths AE = CG = FB = HD = 10 mm, EF = FG = 5 mm and GH = 20 mm. The plots show the voltage ratios required (for this particular configuration) to obtain zero EOF in the injection circuit during loading/flow and zero EOF in the loading/flow circuit during injection. 1276 Analyst, 1999, 124, 1273–1282continuously the ratio of concentrations of X and Y entering the detection channel.This type of micro-reactor manifold control can therefore be used, for example, to determine calibration data in analytical systems53 or to investigate the concentration dependence of reaction kinetics under voltage control without refilling of the reservoirs. It should be noted that this analysis, although revealing the basic principles, is somewhat simplistic in that hydrodynamic effects associated with the flows across manifold junctions and surface conductivity effects are neglected. More sophisticated modelling, as described, for example, in refs. 47, 48, 51 and 52, show that complex flow patterns around manifold junctions may significantly modify slug profiles obtained by the injection procedure. Additionally, non-zero surface conductivity would require modification of the equations presented here. 5 Control of chemical reactions under electroosmotic and electrophoretic flow We take as the starting point a channel containing species X with a rectangular slug of reactant Y as would be obtained from a perfect load–inject–flow sequence (shown in Fig. 1). We assume here that species movement within the channel is controlled only by EOF and electrophoresis, i.e., that pressure driven flow is absent. Under these conditions, the species’ velocity profiles across the channel are flat (except for the region very close to the channel wall).Hence all concentrations vary only in the direction of the channel long axis x but are uniform in both orthogonal directions across the channel. We consider a section of channel of length 2a centred at x = 0 and extending from x = 2a to x = +a. For the purpose of the numerical calculations, we adopt a moving coordinate system such that xlab = x + vost, where xlab is the x coordinate relative to the laboratory, vos is the linear electroosmotic velocity and t is time.Initially the channel contains species X at concentration CX 0 with a rectangular slug of species Y at concentration CY 0. The slug of Y has a width of 2b and is initially centred at x = 0. As described above, X and Y can react reversibly to form product Z with forward rate constant kf (second order) and reverse rate constant kr (first order). The initial conditions are stated as follows: 2b @ x @ b: CX = 0, CY = CY 0 at t = 0 ± b @ x @ ±a: CX = CX 0, CY = 0 at t = 0 2a @ x @ a: CZ = 0, at t = 0 (11) Species X, Y and Z have diffusion coefficients DX, DY and DZ and move (relative to x = 0) with electrophoretic velocities vphX, vphY and vphZ, respectively.We assume that all diffusion coefficients are invariant with concentration and that all thermal effects (arising, for example, from heats of reaction) are negligible. The concentrations of X, Y and Z are functions of both time and x according to the following set of equations:68 ¶ ¶ ¶ ¶ ¶ ¶ ¶ ¶ ¶ ¶ ¶ ¶ ¶ ¶ ¶ ¶ ¶ ¶ C t D C x k C C k C v C x C t D C x k C C k C v C x C t D C x k C C k C v C x Y X X X 2 f X Y r Z phX X Y Y 2 f X Y r Z phY Y Z Z Z 2 f X Y r Z phZ Z = - + - = - + - = + - - 2 2 2 (12) These equations correspond to the simple reaction scheme of eqn.(6) but more complex reaction schemes are easily incorporated. The situation described here for micro-reactors is identical (from a theoretical point of view) with that of electrophoretically mediated microanalysis (EMMA) described both experimentally and theoretically by Regnier et al.69–72 Additionally, a related set of equations have been used recently to describe the elution characteristics of species undergoing a first-order reaction in a capillary electrophoresis system.73 The boundary conditions are taken to be that the concentrations of X, Y and Z are unperturbed from their initial values at x = ±a, i.e., CX(x = ±a) = CX 0, CY(x = ±a) = CZ(x = ±a) = 0 (13) The use of these boundary conditions restricts the analysis to conditions such that the zone of reaction is far from the channel ends.For the manifold configuration shown in Fig. 1, this is valid since we wish to simulate the concentration changes occurring in the detection channel section GH before the reaction zone moves round the corner into section HD. The series of eqns. (11)–(13) is solved numerically to obtain plots of CX, CY and CZ versus xlab for different times.The numerical algorithm, outlined in the Appendix, was implemented in a Visual Basic program running in EXCEL on a PC. 6 Example calculations We first model the time evolution of the concentration profiles for a reaction in which all species X, Y and Z are uncharged. All other conditions are specified in the legend of Fig. 4. In this case, all reaction species move together in the channel with the electroosmotic velocity vos and mixing of the reactants occurs only by inter-diffusion between the X stream and slug of Y.Fig. 4 shows three ‘snapshots’ of the concentration profiles where it can be seen that product Z is formed only at the trailing and leading edges of the slug of Y where diffusional inter-mixing gives finite concentrations of both X and Y. Because the time Fig. 4 Calculated concentration profiles for CX (solid line), CY (dashed line) and product CZ (solid line) for time = 0 (a), 15 (b) and 30 s (c).The parameters are CX 0 = 1 mM, CY 0 = 0.9 mM, CZ 0 = 0 mM, DX = 1 31029 m2 s21, DY = 0.7 3 1029 m2 s21, DZ = 0.5 3 1029 m2 s21, vos = 0.5 mm s21, kf = 1000 l mol21 s21, kr = 0 s21, slug width (2b) = 5 mm and all electrophoretic velocities set equal to zero (i.e., all species are uncharged). Analyst, 1999, 124, 1273–1282 1277required for reactant diffusion across the width of the slug of Y is long relative to the time the slug takes to traverse the detection channel EH, the extent of product formation is low.The behaviour of charged reactant species is very different. We simulate the case in which X and Z both bear a positive charge and Y is uncharged. Within the electric field, the velocities of X and Z are accelerated relative to vos whereas Y moves with velocity vos. As seen in Fig. 5, the difference in electrophoretic velocities of the different species causes a displacement of the slug of Y relative to the ‘gap’ in the concentration profile of X. This gives a greatly increased mixing of X and Y (with concomitant formation of Z) at the trailing edge of the Y slug.The extent of product formation within the detection time is therefore greatly increased relative to that for the case of uncharged reagents. Changing the signs of the charges on both X and Z from positive to negative (calculations not shown) causes the product formation to occur at the leading (rather than the trailing) edge of the slug.We note here that it is not necessary for the species X and Y to have different sign charges to induce displacement of the slug of Y relative to the ‘gap’ in X. Even with the same (non-zero) charges, X and Y will have different electrophoretic velocities, and thus show displacement, so long as their diffusion coefficients are different [see eqn. (5)]. For the case in which X and Z bear charges, and Y is neutral as in Fig. 5, the extent of product formation is largely controlled by the relative rate of displacement of the concentration profiles of X and Y.In turn, this is determined by their relative electrophoretic velocities which, like vos, scale with the applied electric field. Fig. 6 shows the effect of increasing the electric field which is modelled by increasing vos whilst maintaining vph for the different species at constant ratios relative to vos. The ordinate of Fig. 6 shows the integral of the product concentration profile, integrated over the detection channel length.(For an optically absorbing species Z with the absorbance detection configuration of Fig. 1, the measured absorbance signal is proportional to this integral.) It can be seen that the applied voltage can be used to control the extent of product formation. The minimum value of product formation is obtained with zero voltage when the extent of product formation is determined only by inter-diffusion of X and Y without the aid of electrophoretic displacement of the concentration profiles.Sufficiently high applied voltage produces virtually complete displacement and complete conversion of Y to product. As seen above, the time required for complete reaction of the Y slug is determined largely by the time taken for the slug to be displaced from the ‘gap’ in the concentration profile of X and should therefore decrease as the slug width is decreased. Simulation of this effect is shown in Fig. 7, where it can be seen that 100% product conversion can be achieved (at a particular applied voltage) by reducing the width sufficiently. Obviously, in the simulation shown, the total amount of product formed reduces as the initial width of the Y slug is decreased. However, it would of course be possible to inject multiple slugs of Y, so as to increase the total amount of product formation simultaneously with increasing the percentage conversion. We note here that techniques to produce very narrow slugs (mm) have been demonstrated experimentally.74 At sufficiently small widths, diffusion alone would ensure complete reactant mixing and conversion of the Y slug to product.We next examine the effects of varying the forward rate constant of the chemical reaction. As seen above, for the concentrations used in the simulation with kf = 1000 l mol21 s21, the extent of chemical reaction is largely controlled by the time required for inter-mixing of the reagents.Under Fig. 5 Calculated concentration profiles for the conditions of Fig. 4 except that vphX = 0.5, vphY = 0 and vphZ = 0.25 mm s21. This corresponds to the species X and Z both bearing a positive charge (which serve to accelerate their total motion) and Y being uncharged. Fig. 6 Variation of the integral of the product concentration profile with time for different vos equal to 0, 0.5 and 1 mm s21 for the curves in ascending order. The electrophoretic velocities were held at vphX = vos, vphY = 0 and vphZ = vos/2 with other conditions as for Fig. 5. The horizontal dashed line corresponds to total conversion to product. Fig. 7 Variation of percentage conversion of Y with time for (in ascending order) initial width of the slug of Y equal to 10, 5, 2 and 1 mm. All other conditions were as for Fig. 5. 1278 Analyst, 1999, 124, 1273–1282these conditions, increasing kf gives virtually no change in the extent of product formation (Fig. 8) since chemical reaction is already faster than the inter-mixing of X and Y.Reducing kf to lower values causes product formation to decrease as the ratedetermining step switches from mixing to the chemical reaction step. Concentration profiles for reactions where product formation is controlled either by mixing or by chemical reaction are compared in Fig. 9. We note here that the transition from mixing rate control to chemical reaction rate control may be induced either by changing the forward rate constant or by changing the reactant concentrations since forward reaction rate is equal to the product kfCXCY.The simulations highlight the importance of a number of time-scales in considering second-order chemical reactions in micro-reactor manifolds. Definitions of the time-scales appropriate to the manifold configuration discussed here are as follows: tdiffusion = b2/D (14) tchemical = 1/kfCX 0 (15) tdisplacement = 2b/|vphX 2 vphY| (16) tdetection = ldetection/(vos + vphY) (17) where tdiffusion is the time required for inter-diffusional mixing of X and Y across the slug of Y (width 2b) to occur. In this context, D is the mean of DX and DY.tchemical is the time required for chemical reaction between X and Y (under conditions when CX 0 > CY 0). tdisplacement is the time required to displace completely the slug of Y from the ‘gap’ in the concentration profile of X. tdetection is the time spent by the slug of Y within the detection channel of length ldetection.Consideration of the relative magnitudes of these times allows a crude prediction of the behaviour of a chemical reaction within a microreactor system. Virtually complete conversion to product is expected when tdetection > tdisplacement (or tdiffusion) and tchemical. The extent of product formation is controlled by the applied voltage (by control of vph) when tdetection < tdisplacement and tdisplacement > tchemical. Under these conditions, the product formation is insensitive to the chemical reaction rate and the initial concentration of X but is controlled by the applied voltage.When tdetection < tchemical and tchemical > tdisplacement or tdiffusion, product formation is sensitive to the chemical rate constant, the reactant concentrations and the applied voltage. These considerations apply to the simple load–inject–flow voltage control sequence described earlier. However, as demonstrated elegantly by Regnier’s group in the context of EMMA, more complex voltage control sequences can be used to control the extent of reaction.69–72 Finally, we consider the effect of introducing reversibility into the chemical reaction.Under conditions when the reaction reaches its final, equilibrium extent of product formation before exiting the detection channel section, the final value reached decreases with increasing kr as shown in Fig. 10. In this situation, when displacement and chemical reaction are complete, the final extent of product formation is primarily controlled by the equilibrium constant K ( = kf/kr) and CX 0.Fig. 8 Variation of integral CZdx with time for (in ascending order) kf = 10, 100 and 1000 l mol21 s21. The initial width of the slug of Y was 2 mm and all other conditions were as for Fig. 5. The open circles were calculated for kf = 10 000 l mol21 s21 and correspond to the fast reaction limit under these conditions. The horizontal dashed line corresponds to total conversion to product.Fig. 9 Concentration profiles of X, Y (dashed line) and Z after 20 s for kf = 1000 (a), 100 (b) and 10 l mol21 s21 (c). All other conditions were as for Fig. 8. Fig. 10 Variation of integral CZdx with time for kf = 1000 l mol21 s21 and (in ascending order) kr = 3, 1, 0.3 and 0 s21. All other conditions were as for Fig. 8. The horizontal dashed line corresponds to total conversion to product. Analyst, 1999, 124, 1273–1282 1279The simulations illustrate that product formation in microreactor manifolds may (under different conditions) be sensitive to the applied voltage, the chemical rate constants, concentrations, diffusion constants and species charge.In principle, micro-reactor investigation of reactions can yield information on all these physico-chemical properties. For the simulations, the range of input parameters were chosen to be realistic for the type of reaction that may be studied. For example, in aqueous solution at pH 7 and 25 °C, the complex formation reaction Ni2+ + PADA gives NiPADA2+ has kf = 1300 l mol21s21 and kr = 0.1 s21 where PADA is pyridine-2-azo-p-dimethylaniline.75,76 Ni2+ and the NiPADA complex both have a charge of +2 whereas PADA is uncharged. The value of D for Ni2+ is 1.25 3 1029 m2 s21.77 Since PADA has a larger molecular volume than Ni2+, it is expected to have a lower diffusion coefficient (approximately half).Similarly, the complex is expected to have a D value lower than that of either reactant.Hence the physicochemical properties of the Ni–PADA reaction are such that the simulation parameters should correspond approximately to somewhere between those of the top two curves of Fig. 10. For the Ni–PADA reaction, the molar absorptivity of the complex product is approximately 32 000 l mol21 cm21 at the wavelength corresponding to maximum absorption.75 Using the spectrophotometric detection configuration described here, the integrated product concentration profile would produce a large absorbance signal, easily detectable with good precision. 7 Conclusions and future outlook The theoretical principles and calculations described in this tutorial review provide the basis for understanding the behaviour of chemical reactions within micro-reactor manifolds with electrokinetic flow control. The aim has been to provide principles to guide the design and development of such systems and the main conclusions are as follows: 1.EOF is determined primarily by the zeta potential of the channel/solution interface and gives a uniform velocity profile across the channel except very close (nm) to the channel wall. Non-uniform velocity profiles may be caused by pressure gradients arising from unequal reservoir heights, Laplace pressure effects resulting from the liquid menisci within the reservoirs and non-uniformity of the cross-sectional areas and zeta potential of the channels.Non-uniform zeta potentials may arise owing to specific adsorption of reagents in different channel sections or when the different channel sections are constructed of different materials. These complicating factors require careful experimental control in order to obtain accurate, reproducible results in micro-reactor systems. 2. Analysis of the dc circuit equivalent to the micro-reactor configuration allows the proper calculation of the voltages required for a ‘clean’ injection of a reactant slug into a stream of a second reactant. 3. The temporal and spatial evolution of a chemical reaction under EOF and electrophoretic control is determined primarily by the relative magnitudes of tdiffusion, tchemical, tdisplacement and tdetection. Proper adjustment of the relative magnitudes of these different time-scales allows the extent of product conversion to be controlled by the voltages applied to the micro-reactor device. The potential power of micro-reactor manifolds lies in the fact that complex channel labyrinths can be accommodated within a small device and that they allow the investigation of chemical reactions to be made under computer control.From the results described here, measurement of the extent of product conversion under different voltage conditions should, in principle, yield quantitative information on reaction rate parameters and charge/diffusion properties of the reactant species. It is technically feasible to construct micro-reactor manifolds in which many reactions could be investigated either in sequence or simultaneously using automated computer control.Such a development would go some way towards realising the ‘Lab-on-a-Chip’ concept and would provide a quantum leap in the rate of accumulation of physico-chemical information for analytical, general chemical, biochemical and catalytic reactions. Analogously to a conventional electronic chip, the main function of such a device would be rapid gathering and processing of chemical information.In this paper, we have discussed a simple homogeneous liquid phase reaction, but many other possibilities can be envisaged. The areas of homogeneous and heterogeneous catalysis, in particular, could benefit from the high speed, high throughput experimentation possibilities of micro-reactors. In this connection, it has recently been demonstrated that glass frits may be incorporated within a micro-reactor channel.78 Further, a palladium catalyst supported on such a frit has been used successfully to catalyse reactions within a manifold under EOF control.79 It was demonstrated that the spatial and temporal control of reactants could be used to deliver a first reagent to a catalyst surface followed by a second reagent after a controllable time period.This type of detailed control is generally impossible (at least within the short time-scales achievable in micro-reactors) in the usual situation of stirring a reagent mixture over a slurry of catalyst.In analytical terms, such an approach could mean controlling the output from a chemiluminescent reaction accurately at a specific detector site.80 Some of these possibilities are currently being pursued further in the authors’ laboratories. 8 Appendix 8.1 Numerical algorithm for the solution of eqns. (11)–(13) We use the fact that the diffusion–reaction eqns. (12) have virtually similar form: ¶ ¶ ¶ ¶ ¶ ¶ C t v C x D C x S x t k k k k k k k + = + = ph X,Y,Z) 2 2 ( , ) ( (A1) where Sk(x,t) are the source terms which depend on x and t through the rates of the respective chemical reactions.In our particular case of chemical reaction, eqn. (6), we have SX = SY =2SZ = krCZ(x,t) 2 kfCX(x,t)CY(x,t) (A2) To find the evolution of the concentration profiles, Ck(x,t), with time we use a semi-implicit Crank–Nicholson method for integration of eqns. (A1).81,82 For the convenience of readers we give here the details of the numerical scheme applied to eqns.(A1). To avoid overburden, we omit the subscript k in our further notations. The space and time are discretised as follows: xi = (i 2 1)Dx, tj = (j 2 1)Dt, i = 1, 2, . . ., n; j = 1, 2, . . . (A3) where Dx and Dt are the spatial and time steps. Since eqns. (A1) are highly non-linear owing to the source terms, an appropriate linearisation is needed. In our calculations we apply the Crank– Nicholson scheme as follows: ¶ ¶ ¶ ¶ ¶ ¶ ¶ ¶ ¶ ¶ C t D C x C x v C x C x S x t i i j i j i j i j i j ÊË �� = Ê Ë Á � � � + Ê Ë Á � � � È Î ÍÍ .° .. - ÊË �� È Î ÍÍ Ï Ì Ô ÓÔ + ÊË �� . ° .. �. Ô .Ô + + + 1 2 2 2 1 2 2 1 , , , , , [ ( , )] ph (A4) where the non-linear terms, S, are estimated by using only information from the previous time step (j) or from the initial 1280 Analyst, 1999, 124, 1273–1282conditions (j = 0). We remark that the coupling between the three eqns. (A1) comes from the source terms. That is why the resulting equations after the above discretisation are semidecoupled in the framework of a single time step.Thus, the substitution of the partial derivatives in eqn. (A4) by finite difference approximations ¶ ¶ ¶ ¶ ¶ ¶ C t C C t O t C t C C x O x C x C C C x O x i i j i j i i j i j i i j i j i j ÊË �� = - + ÊË �� = - + Ê Ë Á � � � = - + + + + - + - , , , ( ) ( ) ) 1 1, 1, 1, 1, ( D D D D D D 2 2 2 2 2 2 2 (A5) gives a system of linear equations describing each of the profiles in the next time (j + 1) step: b(1 + a)Ci + 1,j + 1 2 (1 + 2b)Ci,j + 1 + b(1 2 a)Ci 2 1, j + 1 = gi, i = 2, .. ., n 2 1 (A6) where a b g b a b b a = = = - + - - - - - + v x D D t x S t C C C i i j i j i j i j D D D D 2 , , ( ) ( ) ( ) , , , , 2 1 1 2 1 2 1 1 (A7) The implication of the boundary conditions, eqns. (11), requires that Ci,j + 1 = Cn,j + 1 = C0 (A8) where C0 = C0 X for X profile and C0 = 0 for Y and Z profiles. Initially, stepwise profiles are used as an initial condition, according to eqns.(11). We solve the three-diagonal system of linear eqns. (A6) by using the Thomas algorithm81 to obtain the new concentration profiles (j + 1), which then are used to calculate gi for the next time step, etc. Once the spatial step-size Dx has been selected, the time step Dt is controlled to maintain the stability of the numerical method. We should stress that the accuracy of this numerical scheme decreases when the chemical reaction is much faster than the respective diffusion process (tdiffusion > > tchemical).In this case we recommend the use of the fully implicit Crank–Nicholson scheme.81 8.2 Symbols A–D reservoirs a half-length of channel Achannel cross-sectional area of channel b initial half-width of rectangular slug of reactant Y CX concentration of species X CX 0 initial concentration of species X DX diffusion coefficient of species X E–H channel corners E electric field e electronic charge g acceleration due to gravity i, j indices for space and time steps used in the numerical calculations I1 electrical current in loop 1 k Boltzmann constant K equilibrium constant for reaction kf second-order forward rate constant kr first-order reverse rate constant lchannel channel length ldetection length of detection channel section r radius of curvature of liquid meniscus in reservoir RAE resistance across channel section AE Rloop 1 sum of resistances across channel sections comprising loop 1 S(x,t) source terms T absolute temperature t time tchemical time required for chemical reaction tdetection time required for Y slug to traverse the detection channel section tdiffusion time required for diffusion across the rectangular slug of Y tdisplacement time required for displacement of slug of Y from ‘gap’ in X concentration profile VAD applied voltage across reservoirs A and D Vos EOF driven volumetric flow rate vos linear electroosmotic flow velocity vphX electrophoretic velocity of species X Vpress pressure driven flow rate x co-ordinate along channel normalised with respect to vos X, Y, Z reactant species xlab coordinate along channel in laboratory coordinates z number of electronic charges on an ionic species Dhres height difference between reservoirs DP hydrostatic pressure difference between reservoirs e relative permittivity e0 permittivity of free space g liquid/air surface tension h viscosity of liquid in the channel l0 electrical conductivity of liquid r liquid density z zeta potential of the channel/liquid interface 9 References 1 A.Manz, D. J. Harrison, E. Verpoorte, J. C. Fettinger, H. Lüdi and H. M. Widmer, Chimia, 1991, 45, 103. 2 A. Manz, D. J. Harrison, E. Verpoorte and H. M. Widmer, Adv. Chromatogr., 1993, 33, 1. 3 A. Manz, C. S. Effenhauser, N. Barggraf, E. J. M. Verpoorte, D. E. Raymond and H. M. Widmer, Analusis, 1994, 22, M25. 4 A. T. Woolley, and R. A. Mathies, Anal.Chem., 1995, 67, 3676. 5 A. T. Woolley, D. Hadley, P. Ladre, A. J. deMello, R. A. Mathies and M. A. Northrup, Anal. G. F. Sensabaugh and R. A. Mathies, Anal. Chem., 1997, 69, 2181. 7 A. T. Woolley and R. A. Mathies, Proc. Natl. Acad. Sci. USA, 1994, 91, 11348. 8 S. C. Jacobson and J. M. Ramsey, Anal. Chem., 1996, 68, 720. 9 P. C. Simpson, D. Roach, A. T. Woolley, T. Thorsen, G. F. Sensabaugh and R. A. Mathies, Proc. Natl. Acad.Sci. USA, 1998, 95, 2256. 10 R. F. Service, Science, 1998, 282, 399. 11 A. Manz, E. J. M. Verpoorte, C. S. Effenhauser, N. Barggraf, D. E. Raymond and H. M. Widmer, Fresenius’ J. Anal. Chem., 1994, 348, 567. 12 H. Y. Wang, R. S. Foote, S. C. Jacobson, J. H. Schneibel and J. M. Ramsey, Sens. Actuators B, 1997, 45, 199. 13 Z. H. Fan and D. J. Harrison, Anal. Chem., 1994, 66, 177. 14 D. J. Harrison, K. Fluri, N. Chiem, T. Tang and Z. Fang, Sens. Actuators B, 1996, 33, 105. 15 D.J. Harrison, P. G. Glavina and A. Manz,, Sens. Actuators B, 1993, 10, 107. 16 M. A. Roberts, J. S. Rossier, P. Bercier and H. Girault, Anal.Chem., 1997, 69, 2035. 17 L. E. Locasico, Anal. Chem., 1997, 69, 4783. 18 W. Ehrfeld and H. Lehr, Radiat. Phys. Chem., 1995, 45, 349. 19 D. C. Duffy, J. Cooper McDonald, O. J. A. Schueller and G. M. Whitesides, Anal. Chem., 1998, 70, 4974. 20 H.-J. Ache, in Micro Total Analytical Systems, ed. A. van den Berg and P. Bergveld, Kluwer, Dordrecht, 1995, p. 47. Analyst, 1999, 124, 1273–1282 128121 R. V�okel, H. P. Herzig, P. Nussbaum and R. D�andliker, Opt. Eng., 1996, 35, 3323. 22 E. Verpoorte, A. Manz, H. Lüdi, A. E. Bruno, F. Maystre, B. Krattiger, H. M. Widmer, B. H van der Schoot and N. F. de Rooij, Sens. Actuators B, 1992, 6, 66. 23 D. J. Harrison, K. Fluri, K. Seiler, Z. Fan, C. S. Effenhauser and A. Manz, Science, 1993, 261, 895. 24 S. C. Jacobson and J. N. Ramsey, Anal. Chem., 1996, 68, 720. 25 C. S. Effenhauser, G.J. M. Bruin, A. Paulus and M. Ehrat, Anal. Chem., 1978, 69, 3451. 26 A. T. Woolley and R. A. Mathies, Anal. Chem., 1995, 67, 3676. 27 D. J. Harrison, A. Manz, Z. Fan, H. Lüdi and H. M. Widmer, Anal. Chem., 1992, 64, 1926. 28 N. Chiem and D. J. Harrison, Anal. Chem., 1997, 69, 373. 29 D. E. Raymond, A. Manz and H. M. Widmer, Anal. Chem., 1996, 66, 2858. 30 C. S. Effenhauser, A. Manz and H. M. Widmer, Anal. Chem., 1995, 67, 2284. 31 S. C. Jacobson and J. M. Ramsey, Anal.Chem., 1997, 67, 3212. 32 S. C. Jacobson, R. Hergenröder, L. B. Koutny and J. M. Ramsey, Anal. Chem., 1994, 66, 1114. 33 S. C. Jacobson, L. B. Koutny, R. Hergenröder, A. W. Moore and J. M. Ramsey, Anal. Chem., 1994, 66, 3472. 34 L. C. Waters, S. C. Jacobson, N. Kroutchinina, J. Khandurina, R. S. Foote and J. M. Ramsey, Anal. Chem., 1998, 70, 5172. 35 M. U. Kopp, A. J. deMello and A. Manz, Science, 1998, 280, 1046. 36 C. S. Effenhausaer and A. Manz, Am. Lab., 1994, 26, 15. 37 G. Blankenstein and U. D. Larsen, Biosens. Bioelectron., 1998, 13, 427. 38 S. Sundberg, A. Copfsill, R. Nagle, S. Gallagher, C. Chow, G. Wada, T. Nikiforov and J. W. Parce, Am. Lab., 1998, 30, 22. 39 R. Kniss, Am. Lab., 1998, 30, 40. 40 P. Graveson, J. Branebjerg and O. S. Jensen, J. Micromech. Microeng., 1993, 3, 168. 41 S. Shoji and M. Esashi, J. Micromech. Microeng., 1994, 4, 157. 42 P. K. Dasgupta and S. Liu, Anal. Chem., 1994, 66, 1792. 43 R. Zengerle, J. Ulrich, S. Kluge, M.Richter and A. Richter, Sens. Actuators A, 1995, 50, 81. 44 A. Manz, C. S. Effenhauser, N. Burggraf, D. J. Harrison, K. Seiler and K. Fluri, J. Micromech. Microeng., 1994, 4, 257. 45 P. H. Paul, D. W. Arnold and D. J. Rakestraw, in Micro Total Analysis Systems ’98, ed. D. J. Harrison and A. van den Berg, Kluwer, Dordrecht, 1998, p. 49. 46 G. Boer, A. Dodge, K. Fluri, B. H. van der Schoot, E. Verpoorte and N. F. de Rooij, in Micro Total Analysis Systems ’98, ed. D. J. Harrison and A.van den Berg, Kluwer, Dordrecht, 1998, p. 53. 47 S. C. Jacobson, R. Hergenröder, L. B. Koutny, R. J. Warmack and J. M. Ramsey, Anal. Chem., 1994, 66, 1107. 48 S. C. Jacobson, R. Hergenröder, A. W. Moore and J. M. Ramsey, Anal. Chem., 1994, 66, 4127. 49 K. Seiler, Z. H. Fang, K. Fluri and D. J. Harrison, Anal. Chem., 1994, 13, 3485. 50 P. H. Paul, M. G. Garguilo and D. J. Rakeshaw, Anal. Chem., 1998, 70, 2459. 51 S. V. Ermakov, S. C. Jacobson and J. M. Ramsey, in Micro Total Analysis Systems ’98, ed.D. J. Harrison and A. van den Berg, Kluwer, Dordrecht, 1998, p. 149. 52 N. A. Patankar and H. H. Hu, Anal. Chem., 1998, 70, 1870. 53 S. J. Haswell, Analyst, 1997, 122, 1R. 54 L J. Kricka and P. Wilding, in Handbook of Clinical Automation, Robotics and Optimisation, ed. G. J. Kost and J. Welsh, John Wiley & Sons, New York, 1996, p. 45. 55 G. T. A. Kovacs, Micromachined Transducers Sourcebook, WCB/ McGraw-Hill, Boston, . 56 G. T. A. Kovacs, K. Petersen and M. Albin, Anal. Chem., 1996, 68, 407A. 57 G. N. Doku and S. J. Haswell, Anal. Chim. Acta, 1999, 382, 1. 58 G. M. Greenway, S. J. Haswell and P. Petsul, Anal. Chim. Acta, 1999, 387, 1. 59 C. L. Rice and R. Whitehead, J. Phys. Chem., 1965, 69, 4017. 60 R. J. Hunter, Zeta Potential in Colloid Science, Academic Press, London, 1981. 61 J. Jednacak, V. Pravdic and W. Haller, J. Colloid Interface Sci., 1974, 49, 16. 62 G. R. Wiese, R. O. James and T. W. Healy, Discuss. Faraday Soc., 1971, 52, 302. 63 A. Vernhet, M. N. Bellon-Fontaine and A. Doren, J. Chim. Phys., 1994, 91, 1728. 64 R. A. Van Wagenen and J. D. Andrada, J. Colloid Interface Sci., 1979, 76, 305. 65 M. P. Sidorova, I. A. Savina and L. E. Ermakova, Colloid J., 1996, 53, 380. 66 See, for example, A. W. Adamson and A. P. Gast, Physical Chemistry of Surfaces, John Wiley and Sons, New York, 5th edn., 1997, p. 10. 67 See, for example, P. W. Atkins, Physical Chemistry, Oxford University Press, Oxford, 5th edn., 1994, p. 840. 68 See, for example, W. Jost, Diffusion in Solids, Liquids and Gases, Academic Press, New York, 1952. 69 B. J. Harmon, D. H. Patterson and F. E. Regnier, Anal. Chem., 1993, 65, 2655. 70 B. J. Harmon, I. Leesong and F. E. Regnier, Anal. Chem., 1994, 66, 3797. 71 F. E. Regnier, D. H. Patterson and B. J. Harmon, Trends Anal. Chem., 1995, 14, 177. 72 D. H. Patterson, B. J. Harmon and F. E. Regnier, J. Chromatogr., A, 1996, 732, 119. 73 S. N. Semenov, Anal. Commun., 1998, 35, 411. 74 J. B. Knight, A. Vishwanath, J. P. Brody and R. H. Austin, Phys. Rev. Lett., 1998, 80, 3863. 75 M. A. Cobb and D. N. Hague, J. Chem. Soc., Faraday Trans. 1, 1972, 68, 932. 76 A. D. James and B. H. Robinson, J. Chem. Soc., Faraday, Trans. 1, 1978, 74, 10. 77 V. N. M. Lobo and J. L. Quaresma, Handbook of Electrolyte Solutions, Part B, Elsevier, Amsterdam, 1989. 78 P. D. Christensen, S. W. P. Johnson, T. McCreedy, V. Skelton and N. G. Wilson, Anal. Commun., 1998, 35, 341. 79 V. Skelton, S. J. Haswell, G. M. Greenway, P. Styring and D. O. Morgan, Chem. Eur. J., submitted for publication. 80 G. M. Greenway, L. Nelstrop and S. N. Port, Anal. Chim. Acta, in press. 81 See, for example, W. H. Press, S. A. Teukolsky, W. T. Vetterling and B. P. Flannery, Numerical Recipes in FORTRAN, Cambridge University Press, Cambridge, 2nd edn., 1992. 82 V. N. Paunov, K. D. Danov, N. Alleborn, H. Raszcilier and F. Durst, Chem. Eng. Sci., 1998, 53, 2839. Paper 9/03624E 1282 Analyst, 1999, 124, 1273–

ISSN:0003-2654    

DOI:10.1039/a903624e

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Solid-phase microextraction for the determination of pethidine and methadone in human urine using gas chromatography with nitrogen–phosphorus detection Seung-Woon Myung,*a Seungki Kim,a Joon-Ho Park,a Myungsoo Kim,a Jong-Chul Leeb and Taek-Jae Kimb a Bioanalysis and Biotransformation Research Center, Korea Institute of Science and Technology, P.O. Box 131, Cheongryang, Seoul 130-650, South Korea. E-mail: swmyung@kist.re.kr b Department of Chemistry, Kyonggi University, Suwon, Kyonggi 440-760, South Korea Received 26th May 1999, Accepted 5th July 1999 A simple and rapid analytical method is presented for the determination of pethidine (meperidine) and methadone in human urine using solid-phase microextraction (SPME) and gas chromatography with nitrogen–phosphorus detection (GC-NPD).After the analytes had been partitioned between an extracting phase and the aqueous sample matrix, the needle of the coating fiber assembly was injected directly into the GC injector.The analytes were thermally desorbed in the heated injector (240 °C) and subsequently separated and detected by the GC-NPD system. The factors influencing the SPME method, such as the salt (NaCl) effect (15%), pH (pH 11), and equilibration time (30 min), were optimized. The calibration graphs for urine samples showed a good linearity. The detection limit was below 1 ng ml21 for both drugs. Introduction Pethidine (meperidine) and methadone, which have nearly the same pKa, are analgesic drugs widely used in the relief of pain.Pethidine can be used in any situation where an opioid analgesic is required. Methadone is also used in the treatment of opioid abstinence syndromes and heroin users.1 Athletes abuse these drugs as a means of enhancing performance in competition. The International Olympic Committee (IOC) Medical Commission and other sports federations have defined these drugs as prohibited classes of pharmacological agents.2 Therefore, a rapid and highly sensitive method for their determination in biological fluids is required in anti-doping tests, toxicology, and clinical and forensic fields.In general, these drugs have been determined by chromatographic methods following liquid– liquid extraction (LLE)3–5 or solid-phase extraction (SPE).6–8 Although these methods may be successful in extracting the drugs from biological fluids, the large amount of organic solvent used in the extraction procedure causes problems with regard to health and the environment.Moreover, these methods are time-consuming and tedious and often require preconcentration of the extract prior to instrumental analysis. Solid-phase microextraction (SPME)9,10 allows the simultaneous extraction and preconcentration of analytes from a sample matrix. This methodology has been widely used for the determination of various compounds such as amphetamines,11 homocysteine and related compounds,12 tricyclic antidepressants, 13 organic acids,14 and organophosphorus insecticides.15 The method is simpler and more rapid than both LLE and SPE, and does not require solvents and preconcentration in the pretreatment step.The coated fiber is immersed directly in the urine sample until equilibration of the analytes adsorbed on to the coating phase is reached. The analytes are then thermally desorbed into the injection port of a gas chromatograph, and subsequently determined. The purpose of this study was to establish the use of SPME for the determination of pethidine and methadone in human urine samples by gas chromatography with nitrogen–phosphorus detection (GC-NPD).Experimental Chemicals Pethidine and methadone bases were obtained from Dr. W. Schanzer, Deutsche Sporthochschule, Cologne, Germany. Diphenylamine [internal standard (ISTD)] was purchased from Sigma (St. Louis, MO, USA). Stock solutions (1000 mg ml21) of these compounds were prepared in methanol. Working solutions were prepared by dilution of the stock solutions with methanol.Sodium chloride and potassium hydroxide were from Kanto Chemical (Tokyo, Japan). Other chemicals used were of analytical-reagent grade. Instrumentation A HP 5890 Series II gas chromatograph (Hewlett-Packard, Palo Alto, CA, USA) with a nitrogen–phosphorus detector was used as for method development and optimization. Nitrogen was used as the carrier gas at a flow rate of 1 ml min21. The initial column temperature was 100 °C for 0.5 min, and was then increased to 300 °C at a rate of 20 °C min21, and held for 3 min. The temperatures of the injector and detector were 240 and 280 °C, respectively. For desorption of the adsorbed solutes, the injector temperature was held constant at 240 °C, and the splitting ratio was 1+10.The other gases were optimized with the following flows: make-up gas (N2, 30 ml min21), air (170 ml min21), and H2 (3 ml min21). SPME fibers The SPME devices and fiber assemblies were purchased from Supelco (Bellefonte, PA, USA).The fibers purchased were coated with either 85 mm polyacrylate (PA), or 100 mm polydimethylsiloxane (PDMS). All fibers were conditioned in the hot injection port of the gas chromatograph according to instructions provided by the supplier. Analyst, 1999, 124, 1283–1286 1283Analytical procedure To obtain the optimized conditions, standard aqueous solutions (5 mg ml21 of both stimulants) were prepared by spiking into 4 ml vials.To prevent any disturbance of the salt solubility and adsorption by the fiber of the solvent (MeOH) of the spiked solution, the solvent was completely removed by a stream of N2. The vial was filled with 3 ml of distilled water, and the pH was adjusted using 5 M KOH after which the salt was added. For the experiments concerning application, 15 ml of diphenylamine solution (ISTD, 100 mg ml21) were added. The fiber was immersed in the sample with stirring at room temperature.After equilibration had been achieved, the fiber was directly transferred into the GC injection port and desorbed for 1 min. All experiments concerning the optimized parameters were performed in triplicate. Optimization The equilibration time (or immersion time) between the coated fiber and the aqueous solution, the pH of the sample solution, and the ionic strength (salt effect), which have an effect on the extraction from the sample matrix, were investigated.The ionic strength was controlled by the addition of various amounts of NaCl to attain a solution containing 5, 10 or 15% (m/m). Because the organic solvent (methanol) used as a working solution to prepare the standard aqueous solution including diphenylamine (ISTD) reduced the solubility of NaCl in the aqueous solution and prevented the extraction of the solutes, the solvent was removed from the vial by a N2 stream. To the distilled water (or human urine), KOH and NaCl were added subsequently.The investigated pH levels were 7, 9, 11 and 13. The pH was adjusted by the addition of 5 M KOH. The equilibration time was studied by directly immersing the coated fiber in the aqueous solution for the intended time (5, 15, 30, 45 or 60 min). The experiment for the selection of an appropriate fiber type was performed using two commercially available fiber coatings, viz., 100 mm PDMS and 85 mm PA. Since the pH and salt in combination generally enhance the affinity of the coated fiber, the effects of the pH and added salt concentration in combination were investigated. The precision and accuracy were determined with three extractions in the concentration range 1–12 mg ml21 for pethidine and 0.05–0.8 mg ml21 for methadone. The detection limits were calculated as the concentration of analytes in the sample based on a signal-tonoise ratio (S/N) of three.Results and discussion In order to obtain the optimum conditions for the simultaneous determination of the two drugs using the SPME method, the effects of several parameters were investigated.As a preliminary experiment, a comparison of the extraction efficiency using a 100 mm PDMS fiber for mid- to non-polar semi-volatiles and a 85 mm PA fiber for polar semi-volatiles was performed. Fig. 1 shows the extraction efficiency according to the fiber type. This experiment was carried out on the control sample (no salt addition and no adjustment of pH) and a comparison of the extraction efficiency was performed by comparing the peak area obtained from the GC traces.Both pethidine and methadone showed a higher affinity for the 100 mm PDMS coated fiber than for the 85 mm PA coated fiber. The peak area obtained with the 100 mm PDMS fiber for both drugs was about eight times higher than that obtained with PA. Hence, the experiment to find the optimum extraction conditions was performed on the 100 mm PDMS coated fiber. The equilibration time was determined by plotting the peak area of the GC trace according to the immersion time of the SPME fiber in the aqueous sample.The experiment was performed by extracting the drugs from distilled water at pH 7.0 and 0% NaCl with magnetic stirring agitation and use of the 100 mm PDMS coated SPME fiber. The peak area increased up to 30 min of immersion and remained constant after 45 min (Fig. 2). Hence, the equilibration time for the two drugs was defined as 30 min.This immersion time was suitable as regards subsequent analysis by GC because the time of the oven temperature program plus the time taken to reach the initial temperature from the maximum oven temperature was about 25 min. Table 1 shows the effects of pH and ionic strength (salt effect) on extraction efficiency. The peak area obtained from the GC trace for control samples of the same concentration, at neutral pH (pH 7) and without salt addition, was compared with that for samples in which the pH was controlled by the addition of 5 M KOH and the ionic strength adjusted by the addition of NaCl in Fig. 1 Comparison of the sensitivity of different fiber coatings for pethidine and methadone. Fig. 2 Effect of the immersion time on the extraction of pethidine and methadone using 100 mm PDMS coated SPME fiber. Table 1 Effect of salt concentration and pH in combination Drug pH NaCl (%) Pethidine Methadone 0 1.00 1.00 5 0.44a 0.31a pH 7 10 0.49 0.41 15 0.26 0.26 0 10.18 4.68 5 11.72 1.89 pH 9 10 13.15 2.13 15 15.23 3.83 0 10.90 19.94 5 23.41 20.61 pH 11 10 28.54 12.41 15 45.36 33.50 0 6.31 18.27 5 13.90 13.78 pH 13 10 27.31 9.78 15 28.38 12.95 a Values correspond to the peak area relative to that of the control sample (pH 7.0 and 0% NaCl). 1284 Analyst, 1999, 124, 1283–1286combination (Table 1). When the concentration of NaCl at pH 7 (no adjustment of pH) was increased, the extracted amount of the drugs decreased compared with the control sample. As expected from the theory, increases in the amount extracted are frequently observed when the salt concentration is increased.Sometimes, however, when analytes are in the dissociated form, a decrease in the amount extracted is observed.16 However, at the other pH levels (pH 9, 11, and 13), as the salt concentration increased, the extraction efficiency increased. These results can be explained by the fact that the activity coefficient of ionic species in water increases with an increase in the ionic strength of the sample.Sample pH affects the dissociation equilibrium in aqueous media. When the pH is increased, the acid–base equilibrium shifts towards the neutral species of basic compounds present in the sample and, therefore, there is an increase in the amount extracted. For total conversion of species to neutral forms, the pH should be at least two units above the pKa of a given basic analyte (pKa + 2).16 The extracted amount of pethidine (pKa = 8.7),17 which has a higher pKa than methadone, showed a larger increase than that of methadone (pKa = 8.3).17 The pH optimization should include experimental verification of the expected results, since adding a buffer to the sample modifies the matrix, which results in distribution constant changes.As expected, the maximum affinity for the coated fiber was obtained for a sample of pH 11 containing 15% NaCl. The amount of pethidine and methadone extracted on to the coated fiber was increased by 45 and 33 times, respectively, as compared with the control sample (pH 7 and no salt addition). Hence, these conditions (pH 11, 15% NaCl concentration) were found to be the optimum for determining pethidine and methadone by the SPME method.Validation and applications Table 2 summarizes the accuracy and precision of the determination of the two drugs by GC-NPD using the SPME method. The experiments were performed under the conditions of pH 11, 15% salt concentration, and 30 min immersion time.The detection limit was evaluated using a S/N of three. The detection limits of pethidine and methadone were below 1 ng ml21. Chiarotti and Marsili18 reported the determination of methadone in urine using SPME, but they used GC-MS for quantification and the detection limit was 20 ng ml21. The within-day relative standard deviations (RSDs) (n = 5) measured at the working range of 1–12 mg ml21 for pethidine and 0.05–0.8 mg ml21 for methadone were 1.8–5.9 and 4.0-8.7%, respectively.For the quantification of the drugs in addicts’ urine, the equations for the calibration graphs were: y = 0.385x + 0.9415 for pethidine and y = 0.0021x + 0.0137 for methadone. The correlation coefficients (r2) of the calibration graphs were 0.9948 for pethidine and 0.9985 for methadone. Two human urine samples, known to be positive for stimulant drugs, were analyzed by the developed method. Although the sensitivities were different, the two drugs could be determined simultaneously with minimal interference. A typical GC trace for an addict’s urine sample is presented in Fig. 3. Pethidine and its major metabolite (norpethidine) were detected at 6.16 and 6.27 min, respectively, and diphenylamine (ISTD) was detected at 5.27 min [Fig. 3(a)]. Methadone and EDDP, which is one of the major metabolites of methadone, were detected at retention times of 8.24 and 7.68 min, respectively. The amounts of pethidine and methadone found in the addict’s urine were 6.63 ± 0.16 mg ml21 (RSD = 2.3%) and 0.42 ± 0.02 mg ml21 (RSD = 4.2%), respectively.Conclusion The developed SPME method for the determination of pethidine and methadone appeared to be useful for analyzing addicts’ urine or administered urine of athletes. The optimum conditions using a 100 mm PDMS coated fiber were a 30 min immersion time, pH 11, and 15% NaCl concentration in combination. The SPME method is simple and highly sensitive and does not require the use of organic solvents. These factors make it attractive for rapid screening in anti-doping tests in sport and for clinical and forensic use.References 1 A. G. Gilman and L. S. Goodman, The Pharmacological Basis of Therapeutics, Macmillan, New York, 7th edn., 1985, pp. 513–519. 2 IOC Medical Commission, Medical Code and Explanatory Document, 1996, ch. II. 3 N. Chikhi-Chorfi, C. Pham-Huy, H. Galons, N. Manuel, W. Lowenstein, J.-M. Warnet and J.-R.Claude, J. Chromatogr., B: Biomed. Appl., 1998, 718, 278. 4 J. F. Furness, A. Weekes, M. Clench, K. Wolff, I. Barnes, W. M. Alistair and M. Cooke, J. Liq. Chromatogr., 1994, 17, 4431. 5 A. M. Bermejo, C. Remuinan, P. Fernandez and M. J. Tabernero, Anal. Lett., 1998, 31, 2645. 6 M. J. Bogusz, R.-D. Maier, K.-D. Kruger and U. Kohls, J. Anal. Toxicol., 1998, 22, 549. 7 M. E. Alburges, W. Huang, R. L. Foltz and D. E. Moody, J. Anal. Toxicol., 1996, 20, 362. 8 S. Rudaz and J.-L.Veuthey, J. Pharm. Biomed. Anal., 1996, 14, 1271. 9 C. L. Arthur and J. B. Pawliszyn, Anal. Chem., 1990, 62, 2145. Table 2 Accuracy and precision of the determination of pethidine and methadone. Found amount Analyte Spiked amount [Mean ± s (% RSD, n = 5)] Pethidine 1.00 0.57 ± 0.03 (5.8) (mg ml21) 3.00 3.39 ± 0.06 (1.8) 6.00 6.18 ± 0.28 (4.5) 9.00 9.05 ± 0.32 (3.5) 12.00 11.8 ± 0.28 (2.4) Methadone 50 54.4 ± 3.98 (7.3) (ng ml21) 100 114 ± 9.86 (8.6) 200 188 ± 16.3 (8.7) 400 390 ± 23.0 (5.9) 800 812 ± 32.6 (4.0) Fig. 3 GC-NPD traces obtained from an addict’s urine sample using the SPME fiber: (a) pethidine positive and (b) methadone positive. Analyst, 1999, 124, 1283–1286 128510 C. L. Arthur, L. M. Killam, K. D. Buchholz and J. B. Pawliszyn, Anal. Chem., 1992, 64, 1960. 11 S.-W. Myung, H.-K. Min, S. Kim, M. Kim, J.-B. Cho and T.-J. Kim, J. Chromatogr., B: Biomed. Appl., 1998, 716, 359. 12 S.-W. Myung, M. Kim, H.-K. Min, E.-A. Yoo and K.-R. Kim, J. Chromatogr., B: Biomed. Appl., 1998, 727, 1. 13 X.-P. Lee, T. Kumazawa, K. Sato and O. Suzuki, J. Chromatogr. Sci., 1997, 35, 302. 14 T. J. Clark and J. E. Bunch, J. Chromatogr. Sci., 1997, 35, 209. 15 S. Magic, A. B. Boland, K. Jinno and J. B. Pawliszyn, J. Chromatogr., 1996, 736, 219. 16 J. Pawliszyn, Solid Phase Microextraction: Theory and Practice, Wiley–VCH, New York, 1997, pp. 130–131. 17 A. C. Moffat, Clarke’s Isolation and Identification of Drugs, Pharmaceutical Press, London, 1986. 18 M. Chiarotti and R. Marsili, J. Microcol. Sep., 1994, 6, 577. Paper 9/04235K 1286 Analyst, 1999, 124, 1283–1286

ISSN:0003-2654    

DOI:10.1039/a904235k

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Determination of organochlorine pesticide residues in foods using solid-phase extraction clean-up cartridges Ruey-an Doong* and Chen-yu Lee Department of Nuclear Science, National Tsing Hua University, Hsinchu, 30043, Taiwan. E-mail: radoong@ins.nthu.edu.tw Received 6th April 1999, Accepted 27th July 1999 Fourteen organochlorine pesticide residues in fatty foods were determined using a simple and rapid procedure based on solid-phase extraction (SPE) clean-up cartridges with octadecyl (C18)-bonded porous silica, a tandem C18 and Florisil column, Alumina-N and Florisil.A Florisil cartridge eluted with 12 ml petroleum ether–ethyl ether (95 + 5) was the most efficient clean-up procedure capable of eliminating the matrix interference and satisfying the agreed acceptable recovery for the large numbers of organochlorine pesticides in nine kinds of foods having different fat contents. Average recoveries of organochlorine pesticides in shellfish, fish and meats ranged from 77 to 105%, 84 to 98% and 85 to 107%, respectively.In addition, analysis of a certified Standard Reference Material (SRM 1945) verified the satisfactory performance of Florisil clean-up cartridge. This SPE method not only yielded comparable results for nonfatty foods, but also provided a reliable separation and quantification of organochlorine pesticides for analyzing a large number of foods with a wide range of fat content. Introduction Organochlorine pesticides (OCPs) have been extensively applied in recent decades against vegetal pests and vector-borne diseases.Owing to their high lipophilic properties, organochlorine pesticides (particularly those having high fat content) have contaminated the environment and food chain. Analysis of organochlorine pesticides in food samples usually comprises three steps: extraction, clean-up and analysis, of these, clean-up is the most laborious. A variety of different clean-up methods have been used to analyze organochlorine pesticides.1,2 Solidphase extraction (SPE) with a conventional adsorption column to remove impurities is a widely used environmental analysis procedure. Several sorbents, including octadecyl (C18)-bonded porous silica, silica gel and Florisil, have been used to trap the analytes.3–5 However, these traditional methods are timeconsuming, require large volumes of solvents, and result in significant quantities of solvent waste and reagent costs.Recent developments of commercially available solid-phase extraction cartridges have significantly reduced analytical time, reagent costs and hazardous waste production. Studies have shown that organochlorine pesticides can be extracted from the nonfatty sample matrix using acetonitrile.6,7 However, recoveries of these compounds are diminished when extracting organochlorine pesticides from a fatty fish.Related investigations used a tandem C18 and Florisil column, eluted with 2% ethyl ether–petroleum ether, to modify the original SPE method for use with fatty fish.8 To determine the presence of organochlorine pesticides in vegetable oils, butterfat, egg and cheese, other investigators indicated that cartridges of Alumina- N and Florisil are optimal in terms of retaining the interfering compounds, which remain in the extract.9,10 Nevertheless, no comprehensive information is available for the clean-up of high-fat foods, such as seafood and meats.Although a clean-up method based on normal-phase liquid chromatography (NPLC) using an LC column packed with 3 mm silica was developed for separating OCPs in fatty matrices,11 an appropriate and efficient SPE clean-up procedure should also be developed to determine the organochlorine pesticide residues in total diet.The developed clean-up procedure could then be applied to the determination of organochlorine pesticide residues in both fatty and nonfatty foods. This investigation compares the results of a comparative study of four different SPE procedures for the clean-up of organochlorine pesticide residues in nine kinds of fatty and two kinds of nonfatty samples.Octadecyl (C18)-bonded porous silica, a tandem C18 and Florisil, and Alumina-N and Florisil cartridges are selected as the clean-up cartridges as they are the most frequently used in the literature. In addition, a Standard Reference Material (SRM) 1945 Whale Blubber was certified to validate the clean-up procedure.Experimental Reagents and materials Acetone and acetonitrile were purchased from Tedia (HPLC/ Spectro grade, Fairfield, OH, USA). n-Hexane and ethyl ether were obtained from Mallinckrodt Co. (nanograde, Phillipsburg, NJ, USA). Petroleum ether was purchased from Reidel-de-Ha�en (pestanal grade, Seelze, Germany). The SPE cartridges, C18 (octadecyl, 17% C), Florisil (100/120 mesh) and Alumina-N (60/325 mesh), were all purchased from Supelco (Bellefonte, PA, USA).The pesticide standards, aldrin, a-HCH, b-HCH, lindane, d- HCH, dieldrin, a-endosulfan, b-endosulfan, endrin, heptachlor, heptachlor epoxide, p,pA-DDE, p,pA-DDT, and p,pA-DDD, were purchased from Supelco. The concentrations were 2000 mg ml21. The SRM 1945 Whale Blubber which contains a- HCH, lindane, p,pA-DDE, p,pA-DDT, and p,pA-DDD was purchased from NIST (Gaithersburg, MD, USA). All other reagents were analytical grade. Extraction For recovery experiments, shrimp and oyster were selected as the food matrices to evaluate the matrix effect on clean-up procedures with different SPE cartridges.The edible portions of Analyst, 1999, 124, 1287–1289 1287shrimp and oyster were dried with a vacuum drier at 250 °C and extracted by Soxhlet extraction with 250 ml hexane for l6 h. The clean-up procedures for the four different SPE procedures are given in Table 1. The eluents were concentrated to about 1 to 2 ml on a rotary evaporator and then transferred to 10 ml glass tubes.The solvent in the glass tube was evaporated off to near dryness under a gentle stream of nitrogen and was quantified to 1 ml. GC determination Concentrations of organochlorine pesticides in the extracts were separated and quantified with gas chromatography equipped with an ECD and a PTE-5 fused silica capillary column (30 m 3 0.32 mm 3 0.25 mm). The column temperature programmed at 140 °C, was increased to 200 °C at 15 °C min21, held for 2 min, ramped to 250 °C at 2 °C min21, and held for 2 min.The injector and detector temperatures were maintained at 250 °C and 300 °C, respectively. The column and make-up gas flow rates were 1.73 ml min21 (linear velocity, 25.2 cm s21) and 35 ml min21, respectively. Pentachloronitrobenzene and decachlorobiphenyl were selected as the internal standards to confirm the retention time of the organochlorine pesticides. Also, identified peaks were confirmed by GC-ECD using an SPB-608 fused silica capillary column (30 m 3 0.53 mm 3 0.5 mm).The SPB-608 column was maintained at 100 °C, heated to 200 °C at a rate of 6 °C min21, programmed to 250 °C at a rate of 8 °C min21, and held for 15 min. The correlation coefficients (R) of the calibration curves for the organochlorine pesticides were all greater than 0.998. The limits of detection (LOD) for the organochlorine pesticides ranged from 0.54 to 4.8 mg kg21. Results and discussion The recoveries of fourteen organochlorine pesticide residues in oyster and shrimp from four different SPE clean-up procedures were performed to evaluate the most suitable clean-up procedure.When a C18 cartridge was used for the clean-up, only twelve organochlorine pesticides were detected. A better recovery of organochlorine pesticides was achieved when using a tandem C18 and Florisil column eluted with acetonitrile. Similar to the results of the C18 cartridge, however, more of the matrix components are present in the extracts. For the Alumina- N and Florisil cartridges, all of the fourteen spiked organochlorine pesticides were detected.Acceptable recoveries and low standard deviations demonstrated that the Florisil cartridge can effectively eliminate matrix components and provide a reliable quantification of organochlorine pesticides. To further validate the Florisil cartridge on the clean-up of different kinds of food, nine kindsf food with different fat contents, including three shellfish, three fish and three meats were tested.Table 2 lists the recoveries of the organochlorine pesticides in seafood using a Florisil SPE cartridge. The average recoveries obtained for the organochlorine pesticide residues in shellfish, ranging from 77 to 105%, were within a satisfactory range. However, the standard deviation for the recovery ranged from 1 to 36%. Aldrin had a high variation of recovery. This finding may be attributed to the different matrices of shellfish used in this study.The shrimp is a crustacean, whereas oyster and clam are molluscs. The physicochemical property of aldrin may be another possibility. The vapor pressure of aldrin is 7.5 3 1025 mmHg, which is higher than those of the other pesticides. Therefore, aldrin might evaporate during the preconcentration procedures. A better clean-up efficiency was observed when changing food matrix to fish. The average Table 1 The SPE cartridges and their operation conditions used in this study Item C18 cartridge Alumina cartridge Florisil cartridge Tandem column Component Octadecyl-bonded porous silica Alumina oxide (60/325 mesh) Magnesium silicate (100/120 mesh) C18 + Florisil Capacity 500 mg per 6ml 2000 mg per 12ml 1000 mg per 6 ml 500 mg (C18)/1000 mg (Florisil) Conditioning 6 ml petroleum ether 6 ml acetone 12 ml acetonitrile 6 ml petroleum ether 6 ml petroleum ether 6 ml petroleum ether 12 ml acetonitrile Eluent 6 ml acetonitrile 12 ml petroleum ether–ethyl ether (95 + 5) 12 ml petroleum ether–ethyl ether (95 + 5) 6 ml acetonitrile Table 2 The recovery (mean ± s, %) of OCPs in seafood using Florisil cartridge with 12 ml petroleum ether–ethyl ether (95 + 5) (n = 3) Shellfish Fish OCPs LODa/mg kg21 Shrimp Oyster Clam Average Roach Salmon Shark Average a-BHC 0.78 88 ± 4 88 ± 2 107 ± 3 94 ± 12 88 ± 3 84 ± 3 99 ± 5 90 ± 8 b-BHC 1.04 82 ± 4 76 ± 1 99 ± 15 86 ± 12 88 ± 3 80 ± 4 90 ± 9 86 ± 6 Lindane 1.97 96 ± 4 94 ± 1 95 ± 10 95 ± 1 88 ± 2 91 ± 3 104 ± 3 94 ± 9 d-BHC 0.54 94 ± 5 96 ± 2 108 ± 3 99 ± 8 97 ± 2 91 ± 3 105 ± 5 98 ± 7 Heptachlor 1.08 86 ± 4 80 ± 2 99 ± 5 88 ± 10 97 ± 2 78 ± 3 91 ± 3 87 ± 9 Heptachlor epoxide 1.53 83 ± 4 68 ± 6 98 ± 1 83 ± 15 90 ± 1 77 ± 4 93 ± 3 86 ± 8 a-Endosulfan 1.16 68 ± 2 68 ± 2 94 ± 2 77 ± 15 84 ± 4 75 ± 4 88 ± 5 82 ± 7 b-Endosulfan 0.94 77 ± 1 75 ± 2 88 ± 3 80 ± 7 87 ± 3 71 ± 4 92 ± 5 83 ± 11 Aldrin 4.83 63 ± 6 122 ± 3 129 ± 20 105 ± 36 97 ± 3 61 ± 7 89 ± 4 82 ± 19 Dieldrin 1.84 81 ± 4 82 ± 2 102 ± 5 88 ± 12 94 ± 3 78 ± 4 96 ± 5 89 ± 10 Endrin 0.78 87 ± 2 89 ± 2 89 ± 4 88 ± 1 88 ± 2 78 ± 3 85 ± 2 84 ± 5 p,pA-DDD 1.44 91 ± 4 86 ± 2 122 ± 6 100 ± 20 89 ± 4 83 ± 3 99 ± 6 90 ± 8 p,pA-DDE 1.27 89 ± 3 76 ± 2 90 ± 6 85 ± 8 85 ± 4 82 ± 5 99 ± 5 89 ± 9 p,pA-DDT 0.76 87 ± 4 87 ± 3 103 ± 12 92 ± 10 81 ± 2 82 ± 2 97 ± 4 87 ± 9 a LOD: limit of detection. 1288 Analyst, 1999, 124, 1287–1289recoveries of organochlorine pesticides in three kinds of fish ranged from 82 to 98%.The standard deviation, ranging from 5 to 19%, was lower than those for shellfish. Moreover, the fish species did not significantly differ in terms of recovery. Good recoveries of organochlorine pesticide residues in meats were obtained. The average recoveries shown in Table 3 ranged from 84 to 107% with a standard deviation of 2–20%. The fat contents of the meat ranged between 13% (for chicken) and 39% (for pork). Notably, the analytical results for the recoveries correlate well with reference reports. This finding suggests that the clean-up procedure used here can eliminate the matrix interference of the high-fat foods.In addition, the Florisil clean-up procedure was evaluated with nonfatty food (Table 3). The mean recoveries of organochlorine pesticides in cereal ranged from 71 to 121%, demonstrating that the Florisil SPE cartridge is appropriate for the clean-up and can provide a reliable quantification of organochlorine pesticides in different kinds of foods having different fat contents.Following the recommended clean-up procedure, the SRM 1945 Whale Blubber was analyzed. Table 4 summarizes the analysis results and the recommended value of each organochlorine pesticide. The certified values of a-BHC, lindane, p,pADDE, p,pA-DDD and p,pA-DDT were 4.25, 0.87, 116.8, 34.9 and 64.3 mg kg21, respectively. The average analyzed concentrations of those compounds were 4.39, 0.89, 123.1, 26.8 and 56.6 mg kg21, respectively. This corresponds to recoveries of 77–103% with a standard deviation ranging from 3 to 19%.The high standard deviation for lindane may be because the certified value is close to the detection limit. Based on the results of the recoveries and standard deviations, the recommended procedure is satisfactory for the clean-up of organochlorine pesticide residues in high-fat samples. Recently, an automated system comprising a single mass spectrometer (MS) combined with online SPE-GC and SPE-LC was found to be a powerful tool for the detection of microcontaminants in aqueous samples.12,13 Therefore, the combination of the recommended clean-up procedure with on-line SPE-LC/GC-MS may be practically applied for the determination of trace level OCPs in fatty foods.Acknowledgements The author would like to thank the National Science Council, R. O. C. for financial support of this investigation under Contract No. NSC 88-2113-M-007-035. References 1 Y.C. Ling and H. C. Teng, J. Chromatogr. A, 1997, 790, 153. 2 M. Ashraf-Khorassani, L. T. Taylor and F. K. Schweighardt, J. Agric. Food. Chem., 1996, 44, 3540. 3 W. C. Quayle, I. Jepson and I. A. Fowlis, J. Chromatogr. A, 1997, 773, 271. 4 S. M. Lee, M. L. Papathakis, H. C. Feng, G. F. Hunter and J. E. Carr, Fresenius’ J. Anal. Chem., 1991, 339, 376. 5 G. H. Tan, Analyst, 1992, 117, 1129. 6 M. Clower and G. Alvarez, Laboratory Information Bulletin 2564, U.S. Food and Drug Adminstration, Division of Field Science, HFC- 140, Rockville, MD, 1981. 7 Food and Drug Administration, Pesticide Analytical Manual, Food and Drug Administration, Washington, DC, 3rd edn., 1994, vol. I, sect. 304E4. 8 F. J. Schenck, L. Calderon and L. V. Podhorniak, J. AOAC Int., 1996, 79, 1209. 9 A. Bentabol and M. Jodral, Pestic. Sci., 1995, 44, 177. 10 C. Nerin, A. R. Tornes, C. Domeno and J. Cacho, Fresenius’ J. Anal. Chem., 1995, 352, 364. 11 G. R. van der Hoff, A.C. van Beuzekom, U. A. T. Brinkman, R. A. Baumann, and P. van Zoonen, J. Chromatogr. A, 1996, 754, 487. 12 A. J. H. Louter, A. C. Hogenboom, J. Slobodnik, R. J. J. Vreuls and U. A. T. Brinkman, Analyst, 1997, 122, 1497. 13 C. Hidalgo, J. V. Sancho, A. Roinavarro and F. Hernandez, Chromatographia, 1998, 47, 596. Paper 9/02722J Table 3 The recovery (mean ± s, %) of OCPs in meat and cereals using Florisil cartridge with 12 ml petroleum ether–ethyl ether (95 + 5) (n = 3) Meats Cereals OCPs Pork Beef Chicken Average Rice Wheat flour Average a-BHC 106 ± 6 109 ± 5 92 ± 4 102 ± 9 94 ± 1 98 ± 4 96 ± 3 b-BHC 88 ± 6 99 ± 5 87 ± 5 91 ± 6 102 ± 9 88 ± 10 101 ± 10 Lindane 96 ± 7 104 ± 2 95 ± 4 98 ± 5 62 ± 1 96 ± 1 79 ± 24 d-BHC 85 ± 5 95 ± 22 104 ± 6 95 ± 9 88 ± 5 126 ± 4 107 ± 27 Heptachlor 105 ± 6 93 ± 4 89 ± 4 96 ± 9 96 ± 6 113 ± 2 111 ± 12 Heptachlor epoxide 95 ± 7 69 ± 2 87 ± 3 84 ± 13 100 ± 11 125 ± 13 113 ± 18 a-Endosulfan 95 ± 4 79 ± 3 83 ± 3 86 ± 8 79 ± 7 63 ± 7 71 ± 12 b-Endosulfan 94 ± 3 83 ± 3 80 ± 3 86 ± 8 74 ± 8 78 ± 2 76 ± 3 Aldrin 110 ± 6 125 ± 14 85 ± 3 107 ± 20 88 ± 4 106 ± 7 97 ± 13 Dieldrin 102 ± 3 106 ± 9 105 ± 4 104 ± 2 111 ± 8 93 ± 7 102 ± 13 Endrin 95 ± 3 96 ± 3 77 ± 4 89 ± 11 85 ± 4 90 ± 1 87 ± 4 p,pA-DDD 79 ± 3 96 ± 7 90 ± 3 88 ± 11 91 ± 8 80 ± 6 83 ± 8 p,pA-DDE 107 ± 4 109 ± 7 75 ± 4 97 ± 19 71 ± 1 78 ± 15 75 ± 3 p,pA-DDT 100 ± 3 64 ± 1 91 ± 3 85 ± 19 106 ± 9 135 ± 6 121 ± 21 Table 4 The certified and analyzed values of OCPs in SRM 1945 (n = 4) Analyzed concentration/mg kg21 OCPs Certified concentration/ mg kg21 Range Average Recovery (%) RSD (%) a-BHC 4.25 3.92–4.73 4.39 103 10 Lindane 0.87 0.7–1.16 0.89 103 19 p,pA-DDE 116.8 119–134 123.1 105 9 p,pA-DDD 34.9 24.7–28.3 26.8 77 6 p,pA-DDT 64.3 54.7–58.1 56.6 88 3 Analyst, 1999, 124, 1287–1289 1289

ISSN:0003-2654    

DOI:10.1039/a902722j

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

HPLC-APCI-MS with calibration based on stable isotope-labelled internal standards for the quantification of carbonyls in air samples Gabriela Zurek,a Heinrich Luftmannb and Uwe Karst*a a Anorganisch-Chemisches Institut, Abteilung Analytische Chemie, Westf�alische Wilhelms-Universit�at M�unster, Wilhelm-Klemm-Str. 8, D-48149 M�unster, Germany b Organisch-Chemisches Institut, Westf�alische Wilhelms-Universit�at M�unster, Corrensstr. 40, D-48149 M�unster, Germany Received 2nd June 1999, Accepted 13th July 1999 A method for the selective determination of formaldehyde and acetaldehyde in air samples based on HPLC-APCI-MS with calibration using stable isotope-labelled standards was developed.d2-Formaldehyde-2,4-dinitrophenylhydrazone and d4-acetaldehyde-2,4-dinitrophenylhydrazone were synthesized from 2,4-dinitrophenylhydrazine and the respective deuterated aldehydes. The labelled derivatives were used as internal calibration standards in the quantification of formaldehyde and acetaldehyde in complex matrices, including automobile exhaust.Co-elution of the labelled and non-labelled derivatives in HPLC provides for identical mass spectrometric conditions, thus significantly reducing the relative standard deviation of the internal calibration approach compared with external calibration with non-labelled derivatives. Hydrazine reagents have been established as highly efficient derivatizing agents for the determination of aldehydes and ketones in recent years.Among these, 2,4-dinitrophenylhydrazine (DNPH) is the most widely used substance.1–3 Carbonyls react with the reagent in acidic media with the formation of the respective hydrazones as shown below: Low limits of detection and high selectivity are major demands for a separation and detection technique for DNPH derivatives. In most cases, they are separated by means of reversed-phase liquid chromatography and detected by UV/vis spectrophotometry.1-4 Although characterized by a larger number of theoretical plates per column, GC with flame ionization detection5,6 is only rarely used with the DNPH method owing to the low volatility of the derivatives and their possible decomposition.Additionally, the formed E- and Zisomers5 of the hydrazones are separated, thus leading to a more difficult assignment of the peaks. Micellar electrokinetic chromatography (MEKC)7 has been introduced as a separation technique with DNPH derivatization, but the LOD is too high according to Beer’s law.To achieve the desired selectivity, HPLC with mass spectrometric detection appears to be a very attractive approach. Olson and Swarin8 were the first to use an LC-MS system with a moving belt interface and chemical ionization for the DNPH derivatives. The use of a particle beam interface in this respect has been described.9 However, these ionization techniques have been replaced by atmospheric pressure ionization methods owing to their better detection limits.10 Recently, K�olliker and co-workers11 published a procedure for the identification of carbonyls based on HPLC with atmospheric pressure chemical ionization mass spectrometry (APCI-MS) in the negative ion mode.The method is characterized by limits of detection in the lower nanomolar range, with good possibilities of identifying single carbonyls from a complex mixture based on tandem MS in the ion trap instrument used. However, the authors did not describe the quantification of aldehydes or ketones. We present here a method for the quantification of the carbonyls in complex samples based on HPLC-APCI-MS with calibration using stable isotope-labelled internal standards.Experimental Chemicals All chemicals were purchased from Aldrich Chemie (Steinheim, Germany) in the highest quality available. As solvent for LC, acetonitrile gradient grade from Merck (Darmstadt, Germany) was used. Air sampling was performed with coated DNPH cartridges (LpDNPH S10) from Supelco (Deisenhofen, Germany).HPLC-MS instrumentation The HPLC-MS system from Shimadzu (Duisburg, Germany) consisted of the following components: SCL-10Avp controller unit, DGU-14A degasser, two LC-10ADvp pumps, SUS mixing chamber (0.5 mL), SIL-10A autosampler, CTO-10ASvp column oven, SPD-10AV UV/vis detector, LCMS QP8000 single Analyst, 1999, 124, 1291–1295 1291quadrupole mass spectrometer with atmospheric pressure ionization and Class 8000 Version 1.01 software. HPLC conditions All separations were performed using a Merck LiChroSpher RP-18EC column in ChromCart cartridges of the following dimensions: particle size, 5 mm; pore size, 100 Å; length, 125 mm; and id, 2 mm.An acetonitrile–water binary gradient at a flow rate of 0.3 mL min21 with the following profile was used: t/min: 0.03 1 15 18 19 24 24.5 [CH3CN] (%): 49 49 100 100 49 49 stop The injection volume was 10 mL, the oven temperature was 23 °C and the detection wavelengths were 300 and 360 nm.MS conditions All MS measurements were recorded using APCI in the negative ion mode under the following conditions: nebulizer gas flow (N2), 2 L min21; probe voltage, 23 kV; temperature of the APCI probe, 300 °C; curved desolvation line (CDL) voltage, 220 V; CDL temperature, 250 °C; deflector voltage, 235 V; and detector gain, 1.5 kV. For scan mode measurements, a mass range from m/z 100 to 400 was chosen and the integration time was 1.2 s.For selected ion monitoring (SIM) measurements, the integration time was 1 s. All variations or further specifications to these parameters are mentioned in the respective analyses. Dependence of MS spectra on the deflector voltages A 0.1 mmol L21 solution of dinitrophenyl hydrazone (DNPhydrazone) standards in acetonitrile was separated by HPLC. The mass spectra were recorded for varying deflector voltages (25, 215, 225, 235, 255 and 275 V) in the scan mode.Several DNPhydrazones were investigated, including those of formaldehyde, acetaldehyde, acetone and other saturated, unsaturated and aromatic aldehydes and ketones. Linear range of MS detection A calibration curve for DNPhydrazone standards in acetonitrile was recorded three times in the range 5 3 1028–1 3 1024 mol L21. The mixture contained the standards of formaldehyde, acetaldehyde, acetone, butynal, methacrolein, cyclohexanone, p-tolualdehyde, hexanal and 2-ethylhexanal.MS detection was carried out in the SIM mode using the following time program of the SIM traces: t/min: 3.5–5.0 5.0–6.4 6.4–7.8 7.8–10.0 10.0–13.5 13.5–15.5 m/z: 209 223 237 247; 249 277; 279; 299 307 Synthesis and characterization of deuterated DNPhydrazones All hydrazones were prepared according to the procedure of Behforouz et al.12 The products were recrystallized from ethanol. d2-Formaldehyde DNPhydrazone: MS (EI, 70 eV), m/z 212 (86%, M+.), 180 (22%), 152 (7%), 122 (16%), 79 (100%); 1H NMR (CDCl3, TMS), d 7.97 (d 1H, H6), 8.50 (dd 1H, H5), 9.13 (d 1H, H3), 11.13 (s 1H, NH); IR (KBr), 3309, 3108, 3088, 1674, 1619, 1528, 1336, 1314, 927, 907, 831 cm21, UV (CH3CN), lmax = 349 nm; e(lmax) = 19 200 L mol21 cm21, elemental analysis Ccalc = 39.6%; Cfound = 39.7%; Ncalc = 26.4%; Nfound = 26.0%; mp, 176 °C.d4-Acetaldehyde DNPhydrazone: MS (EI, 70 eV), m/z 228 (100%, M+.), 180 (6%), 152 (14%), 122 (17%); 1H NMR (CDCl3, TMS), d 7.95 (d 1H, H6), 8.28 (dd 1H, H5), 9.11 (d 1H, H3), 11.04 ppm (s 1H, NH); IR (KBr), 3289, 3105, 3086, 1623, 1588, 1518, 1310, 934, 924, 916 cm21; UV (CH3CN), lmax = 358 nm; e(lmax) = 20 100 L mol21 cm21; elemental analysis, Ccalc = 42.1%, Cfound = 42.3%; Ncalc = 24.6%; Nfound = 24.4%; Mp, 168 °C.Recovery rates using internal standards Solutions of the deuterated and non-deuterated standards were mixed in different ratios [1 3 1025: 1 3 1026, 3 3 1026: 1 3 1026, 1 3 1026: 1 3 1026, 1 3 1026: 3 3 1026 and 1 3 1026: 1 3 1025 (all mol L21)] each in triplicate, and analyzed by HPLC-MS using the following time program for the SIM traces: t/min: 3.5–5.0 5.0–6.4 m/z: 209, 211 223, 227 Every analysis was carried out three times.The peak areas of the SIM traces were integrated separately. The recovery rate of the non-deuterated compound was calculated based on the peak area of the deuterasampling and sample preparation The sampling volume of different car exhaust samples was 1.5 L at a flow rate of 0.5 L min21 using a personal sampling pump (Buck, Orlando, FL, USA).Every cartridge was equipped with a backup cartridge to ensure that the exhaust was only sampled on the first cartridge. The car was stationary, the machine was under idling conditions and the engine had been running for approximately 5 min. The cartridges were eluted the same day with 10 mL of acetonitrile and left to rest for another 20 min. Analysis of car exhaust samples Each sample was analyzed in three different ways: (a) scan mode, (b) SIM mode with adapted time program and (c) SIM mode with addition of deuterated standards.For sample preparation for (c), 900 mL of each sample were mixed with 100 mL of a deuterated standard solution at three different concentrations (1023, 1024 and 1025 mol L21). This mixture was then injected into the HPLC system and analyzed using the same time program as for the recovery rates. UV detection was carried out for every sample in parallel.UV quantification was performed using the chromatographic conditions described by P�otter and Karst.4 Results and discussion The main aim of this work was the development of a robust HPLC-MS method for the quantification of aldehydes and ketones. The first experiments were carried out with a single quadrupole mass spectrometer to investigate which ionization mode is most suitable for the detection of the carbonyl DNPH derivatives.While there was almost no detection possible with APCI and electrospray ionization (ESI) in the positive ion mode or with ESI in the negative ion mode using a standard ESI probe, we were able to confirm the results of K�olliker et al.11 that APCI in the negative ion mode is suitable as ionization technique. Depending on the conditions, either the [M2H]2 peak or the m/ z = 182 peak was the base peak in the mass spectra for all of the more than 30 hydrazones investigated. Tandem MS measurements with a triple quadrupole instrument equipped with a 1292 Analyst, 1999, 124, 1291–1295nanospray interface in the negative ion mode confirmed that [M2H]2 is not the precursor of the m/z 182 peak.The generation of the m/z 182 peak can be explained by electron capture and subsequent cleavage of an R1R2CN radical. The fragmentation scheme is illustrated. Further frequently observed ions are m/z 152 and 122, which are caused by the loss of two NO neutral molecules from the m/z 182 ion. The dependence of the relative peak intensity on the deflector voltage with other conditions held constant (see Experimental section) is presented in Fig. 1 for acetaldehyde 2,4-DNPhydrazone.An increase in the deflector voltage results in an additional acceleration of the ions. Owing to higher energy collisions with neutral molecules, more fragmentation is induced. Using this collision induced dissociation (CID), it is possible to obtain structural information on an analyte in a single quadrupole mass spectrometer if only the analyte is present.This last demand is in many cases met by the chromatographic separation. As it is intended to perform quantification of the hydrazones by their SIM traces, a deflector voltage of 235 V is selected for quantification. Under these conditions, [M2H]2 is the base peak and only slight fragmentation to the m/z 182 peak is observed. To quantify the hydrazones, their standard solutions were investigated by triple LC-MS analysis with quantification via time programmed SIM traces for the respective derivatives.Four different concentrations of each derivative were evaluated. As the MS detector is connected in series with a UV/vis detector, the respective peaks were also evaluated by UV spectrophotometry at l = 360 nm. The relative standard deviations (RSDs, n = 3) obtained by both detection techniques are presented in Table 1. It can be seen that UV detection leads to negligible RSDs ( < 1%), especially for the two higher concentrations.Even at concentrations only slightly above the limit of quantification, the RSDs are very low. In contrast, mass spectrometric detection leads to higher RSDs which in some cases even exceed 10%. It should be noted that these data were obtained under simulated routine conditions, which means that samples of different content and concentrations were run together with some blanks from an autosampler without any tuning in between.The data prove that external calibration in LC-MS with calibration curves for the hydrazones will lead to significantly lower reproducibility than UV detection. This is due to changes in the mass spectrometric conditions over longer times. To overcome this problem, stable isotope-labelled hydrazones were synthesized as standards for internal calibration for the two most important aldehydes, formaldehyde and acetaldehyde. This is based on the idea that both isotope-labelled and analyte hydrazone should co-elute in reversed-phase chromatography, thus offering identical MS conditions for both compounds.With known concentration of the stable isotopelabelled standards, the concentration of the non-labelled analytes could simply be re-calculated from the comparison of the SIM trace peak areas of the two substances. Synthesis was carried out according to Behforouz et al.12 in non-deuterated solvents. The hydrazones of d2-formaldehyde and d4-acetaldehyde were obtained according to the reaction mentioned above (see Introduction). Table 1 Variation of the relative standard deviation (n = 3) for different concentrations using UV detection at 360 nm and MS detection for the respective SIM trace of the [M2H]2 c/mol L21 1027 mol L21 1026 mol L21 1025 mol L21 1024 mol L21 UV MS UV MS UV MS UV MS Formaldehyde 7.0 10.3 3.2 11.0 0.1 1.7 0.1 12.6 Acetaldehyde 3.5 4.5 0.5 13.1 0.6 1.7 0.1 5.1 Acetone 9.8 10.5 2.5 12.6 0.1 5.3 0.2 7.8 Butynal 1.7 17.2 1.2 18.2 0.3 2.3 0.2 4.4 Methacrolein 9.4 17.3 2.1 11.8 0.4 1.9 0.3 2.3 Cyclohexanone 14.8 9.9 1.7 7.5 0.5 7.1 0.1 3.7 p-Tolualdehyde 4.6 11.0 0.8 6.3 0.4 10.1 0.1 3.0 Hexanal 7.6 7.4 0.7 1.8 0.7 9.6 0.1 11.5 2-Ethylhexanal 0.9 10.0 0.9 13.6 0.3 4.8 0.3 2.6 Fig. 1 APCI-MS spectra of acetaldehyde DNPhydrazone as a function of deflector voltage. Analyst, 1999, 124, 1291–1295 1293No deuterium exchange by hydrogen was observed even in protic solutions, as the derivatives precipitated immediately under these conditions.The derivatives were stable towards exchange reactions. It should be noted, however, that the deuterated acetaldehyde tends to exchange protons completely in aqueous solutions within a few hours. The mass spectra of the deuterated derivatives under the conditions described above are shown in Fig. 2. A closer look at these mass spectra proves that the d2- formaldehyde hydrazone ([M2H]2: m/z 211) contains approximately 1.5% d1-formaldehyde hydrazone ([M2H]2: m/z 210), while the d4-acetaldehyde hydrazone ([M2H]2: m/z 227) contains approximately 4.5% d3-acetaldehyde hydrazone ([M2H]2: m/z 226).In both cases, the impurities can be traced back to incompletely deuterated precursors. The impurities were compensated for mathematically in all calculations of concentrations. Fig. 3 proves that almost complete co-elution is observed in reversed-phase HPLC for both pairs of labelled and nonlabelled compounds.Traces A and B are for the non-labelled and the labelled formaldehyde derivatives, respectively, and traces C and D represent the non-labelled and labelled acetaldehyde derivatives, respectively. The recoveries of the non-labelled hydrazones were determined for different mixing ratios (see Experimental). The recovery of formaldehyde hydrazone was 95–100%, and that of acetaldehyde hydrazone was almost 100% with an RSD of 1–3% (n = 3). It is known from the literature that nitrogen dioxide from automobile exhaust may interfere with the DNPH method4,13 owing to co-elution of 2,4-dinitrophenylazide (DNPA) formed in its reaction with DNPH.We therefore investigated car exhaust samples with respect to their content of different aldehydes and compared the quantification of selected aldehydes using UV/vis detection with MS detection based on calibration with the ipe-labelled internal standards. In Fig. 4, the UV traces at l = 360 and 300 nm and the SIM traces at m/z 197 ([M2H]2 of DNPH), m/z 182 (DNPH and all hydrazones), m/z = 209 ([M2H]2 of formaldehyde DNPhydrazone), m/z 223 ([M2H]2 of acetaldehyde DNPhydrazone), m/z 235 ([M2H]2 of acrolein DNPhydrazone) and m/z 237 ([M2H]2 of acetone DNPhydrazone and propanal DNPhydrazone) of a diesel fuelled car exhaust sample (car B) are presented.Significant amounts of the hydrazones of formaldehyde, acetaldehyde, acetone, acrolein and propanal and also DNPA and DNPH themselves are detected and identified by their mass spectra and retention times.It is obvious that only small peaks at m/z 182 are observed for all hydrazones, as explained above. The DNPA exhibits two strong peaks at m/z 181 and 182 and coelutes with formaldehyde DNP hydrazone. The m/z 181 peak can be rationalized by radical anion formation and subsequent loss of N2, whilst the m/z 182 peak which exceeds the 13C isotopic peak of m/z 181 might originate from the same reaction but with an uptake of hydride ion.However, none of these mass peaks interfere with the determination of any of the investigated aldehydes. The identitiy of DNPA is confirmed by the retention time and the high absorption at 300 nm. In case of the C3-aldehydes, here acetone and acrolein, MS detection is useful for revealing co-elutions. It should be noted that the mass spectral peaks are delayed by approximately 8 s compared with the UV peaks, as the mass detector is connected in series behind the UV detector.A series of different automobile exhaust samples were investigated regarding their formaldehyde and acetaldehyde Fig. 2 APCI-MS spectra of the DNPhydrazones of d2-formaldehyde (A) and d4-acetaldehyde (B). Fig. 3 MS chromatogram of a mixture of the hydrazones of formaldehyde (A), m/z 209; d2-formaldehyde (B), m/z 211; acetaldehyde (C), m/z 223; and d4-acetaldehyde (D), m/z 227 displayed as extracted SIM traces. Fig. 4 UV and MS chromatograms (displayed as SIM traces) of a diesel fuelled car exhaust real sample (car B). 1294 Analyst, 1999, 124, 1291–1295concentrations. Samples were taken and analyzed as described above. For UV quantification, a gradient was applied in such a way that no co-elution of DNPA and formaldehyde 2,4-DNPhydrazone was observed.4 UV and mass spectrometric detection with calibration using internal stable isotope-labelled standards were performed. In Table 2, comparative data for this investigation for five different car exhaust samples are provided.All diesel fuelled cars (A–C) exhibit extremely high concentrations of NOx (see Fig. 4). Some aromatic aldehydes are observed for the regular fuelled cars (D and E). The regular fuelled car without catalyst shows by far the highest formaldehyde concentrations. It is obvious that the UV and mass spectrometric data correlate well, with a difference between them of less than 1% and 9% (for one individual sample) for all samples.In conclusion, the new method may be considered as very reliable and robust. Its broader application to further aldehydes is currently limited by the availability of the respective stable isotope-labelled carbonyl compounds. Acknowledgements Financial support by the European Union (”Aldehydes” project, SMT4-CT97-2190) is gratefully acknowledged. G.Z. thanks the Stiftung der Deutschen Wirtschaft (Berlin, Germany) for a scholarship. References 1 R. H. Beasley, C. E. Hoffmann, M. L. Rueppel and J. W. Worley, Anal. Chem., 1980, 52, 1110. 2 K. Kuwata, M. Uebori and H. J. Yamasaki, J. Chromatogr. Sci., 1979, 17, 264. 3 D. Grosjean and K. Fung, Anal. Chem., 1982, 54, 1221. 4 W. P�otter and U. Karst, Anal. Chem., 1996, 68, 3354. 5 M. G�org�enyi, H. Van Langenhove and Z. Kir�aly, J. Chromatogr. A, 1995, 693, 181. 6 Y. Hoshika and Y. Takata, J. Chromatogr., 1976, 120, 379. 7 S.-L. Zhao, T.-Y. Dai, Z. Liu, F.-S. Wei, H.-F. Zou and X.-B. Xu, Chemosphere, 1997, 35, 2131. 8 K. L. Olson and S. J. Swarin, J. Chromatogr., 1985, 333, 337. 9 E. Grosjean and D. Grosjean, Int. J. Environ. Anal. Chem., 1995, 61, 47. 10 W. M. A. Niessen and A. P. Tinke, J. Chromatogr., 1995, 703, 37. 11 S. K�olliker, M. Oehme and C. Dye, Anal. Chem., 1998, 70, 1979. 12 M. Behforouz, J.-L. Bolan and M. S. Flynt, J. Org. Chem., 1985, 50, 1186. 13 U. Karst, N. Binding, K. Cammann and U. Witting, Fresenius’ J. Anal. Chem., 1993, 345, 48. Paper 9/04393D Table 2 Formaldehyde and acetaldehyde concentrations in car exhaust real samples using UV detection at 360 nm and MS detection with addition of internal standards (A–C, diesel fuelled; D, regular fuelled with catalyst; E, regular fuelled without catalyst) c(ppm) Car A Car B Car C Car D Car E UV MS UV MS UV MS UV MS UV MS Formaldehyde 9.19 8.41 5.00 5.18 2.70 2.64 3.70 4.00 17.02 16.43 Acetaldehyde 6.69 6.65 4.63 4.62 1.09 1.13 3.42 3.62 3.60 3.73 Analyst, 1999, 124, 1291&ndash

ISSN:0003-2654    

DOI:10.1039/a904393d

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

GC-MS determination of steroids related to androgen biosynthesis in human hair with pentafluorophenyldimethylsilyl–trimethylsilyl derivatisation Man Ho Choi and Bong Chul Chung* Bioanalysis and Biotransformation Research Center, KIST, P.O. Box 131, Cheongryang, Seoul, Korea. E-mail: bcc0319@kist.re.kr Received 17th May 1999, Accepted 13th July 1999 An efficient method for the simultaneous determination of eight steroids, androstenedione, dihydrotestosterone (DHT), dehydroepiandrosterone (DHEA), testosterone, androsterone, etiocholanolone, progesterone and pregnenolone, in human hair by gas chromatography-mass spectrometry (GC-MS) using d3-testosterone as internal standard is described.The method involves alkaline digestion, liquid–liquid extraction and subsequent conversion to mixed pentafluorophenyldimethylsilyl–trimethylsilyl (flophemesyl–TMS) derivatives for sensitive analysis in the selected ion monitoring (SIM) mode. This method showed good overall repeatability and reproducibility of 4.88–11.24 and 3.19–9.58%, respectively.For the first time, the quantification of DHT, DHEA and pregnenolone in human hair has been achieved by GC-MS, testosterone was also quantified. The detection of four steroids in hair samples was possible in the concentration range 0.12–8.45 ng g21. The other four steroids, androstenedione, androsterone, etiocholanolone and progesterone, were not detected. The detection limits for SIM of the steroids varied in the range 0.02–0.5 ng g21, and the SIM responses were linear with correlation coefficients varying from 0.991 to 0.996 for most of the steroids studied.The concentrations of the four steroids detected were different in male and female hair samples. 1. Introduction Steroids related to androgen biosynthesis regulate a variety of biological functions, including reproductive and adaptive responses. Alhough principally synthesized by the adrenal gland, gonads, liver and placenta, the brain is a further site of steroid synthesis and metabolism.Usually, the measurement of endogenous and exogenous steroids is achieved in urine or plasma.1,2 In contrast with urine and plasma, the benefit of analysis in hair is that it can show the effect of long-term exposure. To our knowledge, only a few studies have been published in which some anabolic steroids and testosterone have been discovered as androgenic steroids in human hair by gas chromatography-mass spectrometry (GCMS).These involved the additional purification steps, such as solid-phase extraction or fractional liquid chromatography.3–5 Generally, the derivatisation of steroids to mono- or trimethylsilyl ether (TMS) derivatives has been the most common approach.6,7 Selecting the best and most appropriate choice of derivatisation is of primary concern not only for GC properties, but also for the selection of useful ions for trace quantification in the selected ion monitoring (SIM) mode.Therefore, we introduced mixed pentafluorophenyldimethylsilyl–trimethylsilyl (flophemesyl–TMS) derivatisation, which had been confirmed in a previous study.8 The significant advantages of this technique include the formation of an intense molecular ion and the minimization of background noise without additional purification steps. The objective of this work was to determine rapidly and sensitively the steroid contents of hair without additional purification steps using GC-MS, because so far no MS assay has been described for the detection of steroids in hair.Such a method would give the possibility of elucidating their metabolism and improving the detection of xenobiotic compounds in hair by trace analysis. 2. Experimental 2.1. Chemicals Androstenedione (androst-4-ene-3,17-dione), progesterone (pregn-4-ene-3,20-dione), androsterone (5a-androstan-3a-ol- 17-one), etiocholanolone (5b-androstan-3a-ol-17-one), DHEA (dehydroepiandrosterone; 5-androsten-3b-ol-17-one), DHT (dihydrotestosterone; 5a-androstan-17b-ol-3-one), testosterone (androst-4-en-17b-ol-3-one) and pregnenolone (pregen-5-ene- 3b-ol-20-one) were purchased from Sigma (St.Louis, MO, USA). The internal standard (IS) was [16,16,17-2H3]testosterone, obtained from the Cologne Laboratory (Institute of Biochemistry, German Sports University, Cologne, Germany). Pentafluorophenyldimethylchlorosilane (flophemesyl chloride), N-methyl-N-trimethylsilyltrifluoroacetamide (MSTFA), ammonium iodide (NH4I) and dithioerythritol (DTE) were purchased from Sigma. 2.2. Preparation of standard solution A stock standard solution of eight steroids and d3-testosterone was prepared at concentrations of 0.1 mg ml21 in methanol. The stock standard solution was used to prepare a working standard solution of varying concentrations (0.01–1.0 mg ml21) in methanol. 2.3. Derivatisation Flophemesyl chloride (50 ml) was added to the residue and the mixture was allowed to stand at room temperature for 15 min.After the excess reagent had been evaporated under a stream of nitrogen at 70 °C, the trimethylsilylating reagent [40 ml of Analyst, 1999, 124, 1297–1300 1297MSTFA–NH4I–DTE (1000+4+2 v/w/w)] was added to the residue and the mixture was heated at 60 °C for 15 min. Approximately 2 ml of the flophemesyl–TMS derivatised sample solution was injected into the GC-MS system. Mixed standard solutions in the range 0.01–1.0 ng g21 were prepared for determining the detection limits of the overall procedure. Calibration standard solutions were obtained by spiking suitable amounts of each of these mixed standard solutions into 1 ml of water.Mixtures containing absolute amounts of several concentrations of each steroid were pretreated, derivatised and analysed in triplicate. To prepare a calibration curve, the ratio of the peak area of steroids to that of d3-testosterone as IS was plotted versus the concentration of the steroids in the calibration standard solution, and a least-squares linear regression analysis was performed.Values of unknown concentrations in hair were determined from the regression line of this calibration curve. 2.4. Sample preparation and pre-treatment Human scalp hair from five males and five females (aged 28–44 years) was collected during haircutting. To prevent contamination with steroids from sweat, sebum, etc., the hair was washed first with acetone and then with methanol.After drying at 60 °C, the hair samples were cut into short lengths of about 1–2 mm and amounts of 200 mg were weighed into glass test-tubes. The pre-treatment was modified with alkaline digestion, and liquid– liquid extraction is well established for testosterone detection in hair and urine.5,8 A 10 ml volume of a d3-testosterone solution (10 ng ml21) and 1 ml of 1 mol l21 NaOH were added, and the solution was heated at 80 °C for 1 h. Then 1 ml of 0.1 mol l21 phosphate buffer (pH 7.0) was added and the pH was adjusted to 10–11 by adding 0.3 ml of 2 mol l21 HCl and with 5 ml of pentane.The mixture was mechanically shaken (10 min) and centrifuged (2400 rpm, 5 min), and the organic phase was transferred to a test-tube. The organic layer was evaporated to dryness with a rotary evaporator. The residue was dried in a vacuum desiccator over P2O5–KOH for at least 30 min before the derivatisation procedure. 2.5. Evaluation of repeatability and reproducibility The repeatability of the chromatographic analysis was determined in a day by 10 replicate 2 ml injections of a mixture of derivatised standard solution at the 1 ppm level.The reproducibility for hair samples was examined every other day (n = 3) by a 2 ml injection of 10 derivatised extracts obtained from fortified water samples at 10 and 50 ppb levels. 2.6. Gas chromatography-mass spectrometry The GC-MS system (Model 5973 mass-selective detector combined with a Model 6890 Plus gas chromatograph, Hewlett- Packard, Avondale, PA, USA) was used in both scan and SIM modes.The electron energy was 70 eV and the ion source temperature was 230 °C. The gas chromatograph was equipped with a 17 m 3 0.2 mm id 3 0.11 mm film thickness capillary column coated with a cross-linked methylsilicone gum phase (Hewlett-Packard). The carrier gas was helium at a column head pressure of 121 kPa. The split (1+10) method of injection was used.The temperature programme was as follows: the initial temperature was 220 °C (2 min), increased at 4 °C min21 to 240 °C (held for 5 min), then at 15 °C min21 to a final temperature of 310 °C (held for 3.33 min). 2.7. Data acquisition In the SIM mode for quantification, a molecular peak ion for each steroid except pregnenolone was selected as a quantitative Table 1 GC-MS data of eight steroids as bis-TMS and flophemesyl–TMS derivatives m/z (relative intensity, %) Steroid Derivativea [M]+ [M 2 15]+ [M 2 72]+ [M 2 90]+ [M 2 105]+ [M 2 167]+ [M 2 225]+ [M 2 257]+ Androstenedione 1 430 (100) 415 (18.3) 358 (1.4) — 325 (7.5) — — — Progesterone 1 458 (41.2) 443 (36.7) 386 (4.3) — 353 (5.3) — — Androsterone 2 586 (43.1) 571 (62.2) — 496 (1.8) 481 (2.2) — 361 (3.1) 329 (20.8) Etiocholanolone 2 586 (49.4) 571 (50.8) — — 481 (2.4) — 361 (1.7) 329 (22.6) DHEA 2 584 (59.9) 569 (51.0) — — 479 (5.3) — 359 (2.0) 327 (14.2) DHT 2 586 (47.1) 571 (8.8) — 496 (3.6) 481 (2.0) — — 329 (4.1) Testosterone 2 584 (92.6) 569 (9.7) 512 (4.1) — 479 (2.5) 417 (3.6) 359 (4.3) 327 (9.8) Pregnenolone 2 612 (16.3) 597 (100) — 522 (9.4) 507 (3.8) — — 355 (7.8) a (1) Bis-TMS and (2) flophemesyl–TMS derivatives.The derivatives were analysed on an Ultra-1 capillary column (17 m 3 0.2 mm i.d. 3 0.11 mm film thickness), with the injector and transfer line temperatures set at 280 and 300 °C, respectively. In the scanning mode, the mass range was m/z 50–650 at a rate of 1.42 scans s21.Table 2 GC-SIM-MS data for the eight steroids Regression linec Steroid RRTa Selected ion (m/z)b Detection limit/ng g21 Calibration range/ng g21 a b Linearityd Androstenedione 0.39 430, 415, 358 0.1 0.5–50 0.0108 0.0005 0.992 Progesterone 0.59 458, 443, 386 0.5 0.5–50 0.0143 0.0011 0.991 Androsterone 0.71 586, 571, 329 0.02 0.5–50 0.0127 0.0007 0.995 Etiocholanolone 0.88 586, 571, 329 0.02 0.5–50 0.0133 0.0008 0.995 DHEA 0.69 584, 569, 327 0.2 1.0–20 0.0052 0.0008 0.996 DHT 0.96 586, 571, 329 0.1 0.1–10 0.0079 0.0008 0.996 Testosterone 1.00 584, 569, 512 0.2 1.0–20 0.0279 0.2155 0.992 Pregnenolone 1.14 612, 597, 522 0.2 0.5–50 0.0125 0.0003 0.996 a Retention time relative to that of d3-testosterone (13.03 min).b Quantitative ions are underlined. c a = Slope = relative mass response = mean peak area ratio of steroid 3 mass of IS/mass of steroid; b = y-intercept. d Linearity is represented by the linear correlation coefficients for the calibration curves. 1298 Analyst, 1999, 124, 1297–1300ion (m/z 430, 458, 584, 586 and 597). The start time for ion monitoring was programmed from 4.1 to 18.0 min to set up one group of 16 ions to be monitored. A dwell time of 150 ms and a relative electron multiplier voltage of 600 V were applied for each ion monitored. 2.8. Calculation The detection limit for each steroid was calculated based on the weight giving a signal three times the peak-to-peak noise of the background signal.A least-squares regression analysis was performed on the measured peak area ratios against the increasing weight ratios of steroids to IS in order to obtain linearity of the SIM responses and to plot calibration curves for the quantitative measurement of steroids. 3. Results 3.1. Mass spectral analysis The eight steroids have two ionizable positions with keto and hydroxyl groups. In order to stabilize the compounds and improve their GC properties, an initial effort was made to examine the mixed derivatisation method using flophemesyl chloride and MSTFA–NH4I–DTE mixture.The six steroids that were displayed confirmed the formation of flophemesyl–TMS derivatives with good GC-MS properties, as noted in our previous report.8 In contrast, androstenedione and progesterone formed bis-TMS derivatives. Flophemesyl chloride did not lead to the formation of enol ethers and greatly hindered the hydroxyl groups to react.9 The eight steroid derivatives were subjected to GC-MS analysis and the electron impact MS data are summarized in Table 1.Similar to our previous report,8 the intense peaks of six flophemesyl–TMS derivatives were either M+ or [M 2 15]+ ions. Also, in both cases androstenedione and progesterone were identified. 3.2. Validation of the method To confirm the peak identities, the present method was designed so that three characteristic ions for each steroid were selected on the basis of their mass fragmentation.Each peak was identified by ratios matched with derivatised steroid standards. When the calibration curve and sensitivity were examined, linear responses for each compound were obtained with correlation coefficients and detection limits varying from 0.991 to 0.996 and from 0.02 to 0.5 ng g21, respectively (Table 2). The repeatability was evaluated for 1 ppm derivatised mixed standards of eight steroids and the reproducibility was determined with extracts that were fortified with standards of eight steroids at 10 and 50 ppb levels in water.The peak areas of selected ions (quantitative ions) were obtained for the eight steroids. They were quantitated by the ratio of the peak area from the fortified sample to that from the corresponding internal standard (d3-testosterone), and absolute values were calculated (Table 3). 3.3. Screening of steroids in human hair The present GC-SIM-MS technique provides a sensitive method for the quantification of steroids in 200 mg of hair.Using this method, at least four steroids, DHT, DHEA, testosterone and pregnenolone, were quantitated. The selected ion chromatogram demonstrates the usefulness of our method in rapid screening for four steroids in scalp hair samples (Fig. 1). Each steroid detected in 200 mg of hair from five male and five female samples was quantitatively determined (Table 4). 4. Discussion We have achieved the quantification of steroids related to androgen biosynthesis in human scalp hair using GC-MS.Fig. 1 Selected ion chromatograms of a water blank extract (A), a fortified hair extract concentrated at the 5 ppb level in hair (B) and a male hair extract (C) separated on Ultra-1 (17 m 3 0.2 mm id 3 0.11 mm film thickness) fused-silica capillary column. Peaks: 1 = androstenedione; 2 = progesterone; 3 = androsterone; 4 = etiocholanolone; 5 = DHEA; 6 = DHT; 7 = testosterone; 8 = pregnenolone. Table 3 Repeatability and reproducibility for the eight steroids Reproducibility (%) Steroid Repeatability (%) 10 ppb 50 ppb Androstenedione 11.24 8.34 7.27 Progesterone 5.92 6.39 6.28 Androsterone 4.88 5.76 6.83 Etiocholanolone 7.45 5.82 4.89 DHEA 8.11 4.22 6.16 DHT 5.90 5.48 9.12 Testosterone 7.39 3.19 9.12 Pregnenolone 6.76 7.86 6.22 All samples were individually prepared as 10 replicates.They were analysed in a day for repeatability and every other day (n = 3) for reproducibility. Analyst, 1999, 124, 1297–1300 1299Highly sensitive detection methods are required to identify steroids in hair, owing to their low concentrations. Derivatisation to TMS derivatives has been the common approach for the determination of steroids, but the mass increment provided by trimethylsilylation is small.Therefore, mixed flophemesyl– TMS derivatives have been advocated as a better choice for SIM at the higher mass of M+ or [M 2 15]+ ions. In contrast with previous studies3–5 on related topics, the most significant advantage of this technique is the minimization of the background noise level without additional purification steps such as solid-phase extraction with a C18 cartridge or Sephadex LH-20.The sensitivity achieved with mixed flophemesyl–TMS derivatisation was excellent for GC-MS assay using a massselective detector. In the present work, the detection of DHEA, DHT, testosterone and pregnenolone in hair samples was possible in the concentration range 0.12–8.45 ng g21.In the case of testosterone, the amounts detected in hair agree with those reported by other workers.4,5 Androstenedione, androsterone, etiocholanolone and progesterone were not detected in this work. On comparing the concentrations of the four steroids detected in male and female hair, a significantly higher concentration of pregnenolone in female hair could be detected. In conclusion, we have developed a method for the detection of steroids in human hair.This study may be the starting point for further studies in the field of steroid analysis that could lead to diverse applications in biomedical research. References 1 M. H. Choi, B. C. Chung, W. Lee, U. C. Lee and Y. Kim, Rapid Commun. Mass Spectrom., 1999, 13, 376. 2 M. H. Choi, J. Y. Kim and B. C. Chung, Anal. Lett., 1999, 32, 1313. 3 D. Thieme, J. Grosse, H. Sachs and R. K. Mueller, in Recent Advances in Doping Analysis: Proceedings of the 16th Cologne Workshop on Dope Analysis, ed. W. Schänzer, H. Geyer, A. Gotzman, U. Mareck-Engelke and S. Rauth, Sport und Buch Strauss, Cologne, 1998, pp. 9–29. 4 M. J. Wheeler, Y. B. Zhong, A. T. Kicman and S. B. Coutts, J. Endocrinol., 1998, 159, R5. 5 C. Scherer, U. Wachter and S. A. Wudy, Analyst, 1998, 123, 2661. 6 W. Schänzer and M. Donike, Anal. Chim. Acta, 1993, 275, 23. 7 H. Geyer, W. Schänzer, U. Schindler and M. Donike, in Recent Advances in Doping Analysis: Proceedings of the 13th Cologne Workshop on Dope Analysis, ed. M. Donike, H. Geyer, A. Gotzman and U. Mareck-Engelke, Sport und Buch Strauss, Cologne, 1996, pp. 215–229. 8 M. H. Choi, J. Y. Kim and B. C. Chung, Analyst, 1999, 124, 675. 9 E. D. Morgan and C. F. Poole, J. Chromatogr., 1975, 104, 351. Paper 9/03912K Table 4 Results of GC-MS quantification of the four steroids detected in human scalp hair Male hair/ng g21 Female hair/ng g21 Steroid Median Range Mean ± s Median Range Mean ± s DHEA 5.19 2.67–5.84 4.57 ± 0.79 2.18 1.02–3.54 2.25 ± 0.59 DHT 0.50 0.38–0.83 0.57 ± 0.11 0.21 0.12–0.30 0.21 ± 0.04 Testosterone 2.51 2.03–2.54 2.36 ± 0.13 1.41 0.88–1.66 1.32 ± 0.19 Pregnenolone 1.82 0.83–3.75 2.13 ± 0.70 17.09 8.45–38.60 21.38 ± 7.32 1300 Analyst, 1999, 124, 1297–1300

ISSN:0003-2654    

DOI:10.1039/a903912k

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Response characteristics of optical sensors for oxygen: models based on a distribution in to or kq Andrew Mills† Department of Chemistry, University of Wales Swansea, Singleton Park Swansea, UK SA2 8PP Received 18th March 1999, Accepted 12th July 1999 The features of two popular models used to describe the observed response characteristics of typical oxygen optical sensors based on luminescence quenching are examined critically. The models are the ‘two-site’ and ‘Gaussian distribution in natural lifetime, to,’ models.These models are used to characterise the response features of typical optical oxygen sensors; features which include: downward curving Stern–Volmer plots and increasingly non-first order luminescence decay kinetics with increasing partial pressures of oxygen, pO2. Neither model appears able to unite these latter features, let alone the observed disparate array of response features exhibited by the myriad optical oxygen sensors reported in the literature, and still maintain any level of physical plausibility.A model based on a Gaussian distribution in quenching rate constant, kq, is developed and, although flawed by a limited breadth in distribution, r, does produce Stern-Volmer plots which would cover the range in curvature seen with real optical oxygen sensors. A new ‘log-Gaussian distribution in to or kq’ model is introduced which has the advantage over a Gaussian distribution model of placing no limitation on the value of r.Work on a ‘log-Gaussian distribution in to’ model reveals that the Stern-Volmer quenching plots would show little degree in curvature, even at large r values and the luminescence decays would become increasingly first order with increasing pO2. In fact, with real optical oxygen sensors, the opposite is observed and thus the model appears of little value. In contrast, a ‘log-Gaussian distribution in ko’ model does produce the trends observed with real optical oxygen sensors; although it is technically restricted in use to those in which the kinetics of luminescence decay are good first order in the absence of oxygen.The latter model gives a good fit to the major response features of sensors which show the latter feature, most notably the [Ru(dpp)3 2+(Ph4B2)2] in cellulose optical oxygen sensors. The scope of a log-Gaussian model for further expansion and, therefore, application to optical oxygen sensors, by combining both a log-Gaussian distribution in ko with one in to is briefly discussed.Introduction The detection and monitoring of oxygen is important in many areas,1,2 including: environmental studies (ambient levels in the atmosphere or dissolved in river, lake or sea waters), patient monitoring (blood gas or breath analysis) and many industrial processes (as diverse as fermentation or sewage treatment). As a consequence, sensing for oxygen is a very important area in analytical chemistry; an area dominated for nearly 4 decades by electrochemical devices, most notably the Clark cell, an oxygen membrane polarographic device.2 In the last decade, the optical oxygen sensor has emerged as a real competitor to the Clark cell, replacing, as it does, costly wires with cheap fibre optics, eliminating electrical interference effects and removing the need for expensive and timeconsuming sensor preparation.1,3 Optical sensors for oxygen are cheap enough to be made disposable; they are also robust and capable of very remote, but rapid sensing. The potential of optical oxygen sensors has prompted many research groups to work on their development in recent years.And yet, despite the intensity of this research, there remain key features of the response characteristics of optical oxygen sensors that have eluded a unifying physical model.4 Such a model, if it is to be had, would provide a better understanding of the underlying processes involved and enable the generation of accurate calibration curves from only a few experimental points.A unifying model would also lead to improvements in sensor development, moving from its current, strongly empirical-based approach to, eventually, a designer-based status, in which optical oxygen sensors are designed and made with specific response features, in keeping with their different areas of application. The majority of optical oxygen sensors work on the principle of the quenching of the luminescence of a dye by oxygen; the higher the level of oxygen the lower the level of luminescence. 1,3 The lumophores that have been used for this purpose are many and varied and include ruthenium(ii) diimine complexes, most notably tris(4,7-diphenyl-1,10-phenanthroline) ruthenium(ii), [Ru(dpp)3 2+], and Pt(ii) and Pd(ii) porphyrins. 1 In its simplest form, an optical oxygen sensor can be generated using a solution of any suitable lumophore (i.e., one that is quenched by oxygen).Under such conditions, the lumophore is usually dispersed homogeneously throughout the solution and the quenching of its luminescence fits a simple Stern–Volmer equation,1 i.e. to/t = Io/I = 1 + KSV.pO2 (1) where Ksv is the Stern–Volmer constant and (t and to) and (I and Io) are the lifetimes and intensities of luminescence in the presence and absence of oxygen, respectively; the latter at a partial pressure of pO2. The Stern–Volmer constant is, of course, a composite parameter equal to the product of the quenching rate constant, kq, and the natural lifetime of the lumophore, to, i.e., KSV = kq.to.The above crude example of an oxygen optical sensor, although simple to make and with easily modeled response features, is not really a practical device. Instead, with most optical oxygen sensors that can be construed as practical, the † Present address: Department of Pure and Applied Chemistry, University of Strathclyde, Thomas Graham Building, 295 Cathedral Street, Glasgow, UK G1 1XL.E-mail: A.Mills@strath.ac.uk Analyst, 1999, 124, 1301–1307 1301lumophore is dispersed on and/or in a solid substrate, such as a sol–gel,5 metal oxide6 or, most often, a polymer.1,7 In such systems, the Stern–Volmer intensity plot, i.e., Io/I versus pO2, often exhibits a reproducible and characteristic negative deviation from linearity; the degree of curvature varying from sensor to sensor with no apparent logical cause. In the latter cases, the lumophore is usually the same and it is simply the encapsulating medium that is altered.Fig. 18–11 illustrates a selection of such Stern–Volmer intensity plots. With the many sensors that exhibit such curvature, lifetime studies also show that the kinetics of the luminescence decays in the presence, and often in the absence, of oxygen are usually no longer simple first order and so cannot be adequately described by the single parameters, t or to, respectively. Several models12–22 have been proposed to describe and explain the downwardly curved Stern–Volmer intensity plots associated with most optical oxygen sensors.Most of these models attribute the latter feature to the micro-heterogeneous nature of the lumophore/encapsulating medium type sensor; a feature that manifests as a distribution of spectroscopically different sites. In this paper the problems, as well as positive features, of the two major models used for interpreting and predicting trends in the response characteristics of optical oxygen sensors, namely, the ‘two-site’12–14 and the ‘Gaussian distribution in t’17–19 models, are identified.From the understanding gained, a third model, the ‘log-Gaussian’ model is introduced. In the first of two papers, the effect of the predicted response characteristics of an optical oxygen sensor in which there is a log-Gaussian distribution in to or kq is examined. The predictions are tested using the results from real optical sensors for oxygen and the results discussed.Experimental Most of the experimental data used to test the various models discussed in this paper were taken from the referenced literature. The exception is the results for the tetraphenyl borate salt of tris(4,7-diphenyl-1,10-phenanthroline) ruthenium(ii), i.e., [Ru(dpp)3 2+(Ph4B2)2], encapsulated in the polymer, cellulose acetate. The method of preparation of [Ru(dpp)3 2+(Ph4B2)2] is described elsewhere.23 The typical solution used for casting films comprised a mixture of (a) a solution containing 2 mg of [Ru(dpp)3 2+(Ph4B2)2] dissolved in 1 mL acetone and (b) 10 g of a 20% m/v solution of cellulose acetate in acetone.The final film solution was cast onto cut microscope glass slides (dimensions: 4 30.9 30.1 cm) through a 100 mm thick brass template with a rectangular hole (0.5 3 1 cm) and then allowed to dry in a desiccator for at least 3 h. The typical film thickness was 25 mm.Luminescence decay measurements were performed using a Nd/YAG Spectron (Rugby, UK) nano-second laser with an Applied Photophysics (London, UK) laser kinetic spectrophotometer monitoring system. The average trace from 32 shots of laser excitation was collected on a Gould Instruments (Ilford, UK) OS4072 digital storage oscilloscope and transferred to a microcomputer for kinetic analysis. Unless stated otherwise, all illustrated Stern– Volmer plots are based on intensity, rather than lifetime, measurements.The two-site model12–14 In all the models which are described in this paper, the general assumption is made that there is a distribution of quenching sites within the encapsulating medium and that at any one type of site, i, the lumophore molecules are quenched by oxygen so that their lifetime, ti, is given by the expression: ti = to,i/(1 + to,i.kq,ipO2,i) (2) where, to,i is the lifetime of the lumophore in the absence of oxygen and kq,i is the rate constant for quenching the lumophore by oxygen at a local partial pressure of pO2,i.As the name suggests, in the two site model it is assumed that there are only two types of quenching sites present in the encapsulation medium, each resulting in a characteristic natural lifetime and quenching rate constant for the lumophore molecules at those sites, i.e., to,1 and k1 and to,2 and k2, respectively. In this model it is also assumed that there is no difference in the partial pressures of oxygen at the two sites and that pO2,i is proportional to the ambient partial pressure of oxygen, pO2, if not equal to it.Unless stated otherwise it is assumed throughout this paper that pO2i = pO2. Under such conditions, it can be shown that the modified Stern–Volmer equation has the following form: I I f k p f k p o o,1 2 o,2 2 O O = + + - + 1 1 1 1 1 1 1 2 ( ) ( ) /( ) t t (3) Where, f1 is the fraction of the overall total number of sites that have a natural lifetime to,1 and a quenching rate constant k1.The above equation has been used extensively to describe the downward curving Stern–Volmer plots associated with most optical oxygen sensors. Table 1 provides the values of KSV1 ( = k1to,1) and KSV2 ( = k2.to,2) which give a good fit to the Stern- Volmer plots illustrated in Fig. 1. The problem with the two site model is the lack of consistency between the values of f1, KSV1 and KSV2, obtained from the Stern-Volmer plot of the intensity data, and the Fig. 1 Stern–Volmer plots of Io/I versus pO2 for the following optical oxygen sensors: [Pd(ii)OEPK] in PVC8 (8), [Pt(dpp)(CN2)2] in silicone rubber9 (:), [Ru(dpp)3 2+(DS2)2] in silicone rubber10 (5) and [Ru(dpp)3 2+(Ph4B2)2] in cellulose acetate (µ), this work. The solid lines have been calculated using eqn. (19) and the optimised values of r and Kmdl are given in Table 1. Table 1 Characteristics of some typical optical oxygen sensors Lumophorea Encapsulation medium f1 KSV1/Torr21 KSV2/Torr21 Kmdl/Torr21 r Ref.[Pd(ii)OEPK] Polystyrene 0.152 2.7 0.142 0.22 1.12 8 [Pt(dpp)(CN2)2] Silicone rubber (RTV 732; GE) 0.893 0.0588 0.00135 0.086 3.15 9 [Ru(dpp)3 2+(DS-)2] Silicone rubber (E4; Wacker) 0.74 0.0346 0.00161 0.0188 2.5 10 [Ru(dpp)3 2+(BPh4 -)2] Cellulose acetate 0.498 0.00134 0.00809 0.0033 1.4 11, this work a OEPK: octaethylporphyrin ketone; DS2: dodecyl sulfate; ph: 1,10-phenanthroline. 1302 Analyst, 1999, 124, 1301–1307luminescence decay results obtained with the same optical oxygen sensor.The luminescence decay of a multi-exponential system, i.e., the I(t) versus t profile, is described by: I(t) = ·aiexp(2t/ti) (4) where ai is the pre-exponential factor associated with the luminescence decay of sites i. Demas and his coworkers17 have introduced the feature of a pre-exponential weighted mean lifetime, tM, defined as follows: tM = · aiti/· ai (5) It follows that to,M = · aito,i/· ai. These workers proved17 that in the absence of static quenching: Io/I = to,M/M (6) In all reported examples of optical oxygen sensors eqn.(6) appears to hold, i.e., there appears to be no evidence of static quenching with any optical oxygen sensor. In the two-site model, eqn. (5) reduces to: tM = a1t1 + (1 2 a1)t1 (7) assuming, for simplicity, · ai = 1. If the two-site model provides a true description of the response features of an oxygen sensor, then the value of f1 (from the Stern–Volmer plot Io/I (or to,M/tM) versus pO2) should be equal to that of a1 at all values of pO2.This prediction is usually not fulfilled with most optical oxygen sensors that have been fitted to a two site model. For example, Demas et al.12. noted, for the [Ru(bpy)3 2+(ClO42)2] in RTV-118 silicone rubber oxygen sensor, that a1 varies from 0.16–0.31 over the pO2 range 0–760 Torr (1 Torr = 133.322 Pa), whereas from the Stern–Volmer plot the value of f1 appears to be 0.4. Apparent dramatic changes in a1 with increasing pO2 have also been noted for different optical oxygen sensors by others.10,11,20 Although the luminescence decay curves exhibited by many optical oxygen sensors can be fitted to two exponentials, some appear to require three,12 introducing the idea of a three site model.However, as with the two-site model, with such systems the variations in ti and ai with increasing pO2 are inconsistent with any physically plausible cause. Thus, the two- and threesite models are used nowdays simply as useful good quality fitting routines.Table 224–26 lists the typical, apparently very disparate response features, which have been observed with optical oxygen sensors. It can be seen that in the absence of oxygen some optical oxygen sensors exhibit first order decay kinetics, but that these change to multi-exponential with the introduction of oxygen. With other sensors the kinetics of luminescence decay, with or without oxygen present, require two or three exponentials in order to be described adequately.In general, the more exponentials needed to describe the decay kinetics, the more curved the Stern–Volmer plot. In order to be useful, any model should ideally be able to embrace all these features and still have a reasonable physical basis. A Gaussian distribution model Although the two- and three-site models of the response features of optical oxygen sensors have been rejected as models with any physical meaning, it is a mistake to abandon the principle of a multi-site model completely.It seems logical to wonder what the response features of an optical oxygen sensor would be if it comprised a much bigger distribution in quenching sites; a distribution which can be described by a simple statistical function? Demas and his co-workers17–19 were the first to address this question with their work on a model based on a ‘Gaussian distribution in to,i’; kq,i and pO2,i remaining constant and equal to kq and pO2, respectively.For a Gaussian distribution, the number of sites i, i.e., ni, is related to the modal number of sites, nmdl, (i.e., the most frequently occurring type of site in the distribution) by the expression: ni/nmdl = exp(2x2) (8) where x is related to the value of to,i associated with sites i and that associated with the modal number of sites, i.e., to,mdl, through the expression: x = {(to,i/to,mdl) 2 1}/r (9) where r is a measure of the spread of the distribution in the natural radiative lifetime of the lumophore, to,i.Although Demas and his co-workers17–19 focussed on a model based on a Gaussian distribution in to,i, it follows that in a model in which there is, instead, a Gaussian distribution in kq,i; with to,i and pO2,i invariant for all sites and equal to to and pO2, respectively, that: x = {(kq,i/kq,mdl) 2 1}/r (10) Note that the effect of a Gaussian distribution in pO2,i, with to,i and kq,i constant for all sites, is indistinguishable from, and will produce the same trends as, a Gaussian distribution in kq,i; with to,i and pO2,i invariant for all sites; thus, to avoid repetition, only the latter case will be considered throughout this work. For a Gaussian distribution in to,i the variation in the fraction of sites i compared with the modal number of sites, i.e., ni/nmdl, as a function of the ratio to,i/to,mdl can be calculated using eqns.(8) and (9), given a value for the breadth of the distribution, r.A similar exercise can be performed for a Gaussian distribution in kq,i using eqns. (8) and (10). The two profiles are the same for any given value of r, as illustrated by the results in Fig. 2 which show the calculated variation in ni/nmdl as a function of the ratio to,i/to,mdl (or kq,i/kq,mdl) for values of r spanning the range 0.1–2. The important feature to note from this work is that it has to be assumed that to,i (or, kq,i) cannot be < 0, i.e., x cannot be < 21/r.Thus, as the value of r is increased, the original Gaussian distribution becomes increasingly distorted, see Fig. 2, and the model accordingly becomes more detached from any likely physical reality. This latter feature led Demas and his co- Table 2 Response features of typical oxygen optical sensors Lifetime studies: No. of exponentials Examples Ref. Stern–Volmer plot N2 O2 Lumophore Encapsulating medium 24 Linear 1 2 PtOEPK PVC 24 Linear 2 2 PtOEPK PS 24 Slightly curved 1 2 PdOEPK PVC 24 [Ru(dpp)3 2+(ClO42)2] Poly(dimethyl siloxane) diacetoxy polymer 25 [Ru(dpp)3 2+(Ph4B2)2] Cellulose acetate 11, 23, this work Curved 2 2 [Ru(bpy)3 2+(ClO42)2] RTV-118 silicone rubber 12 [Ru(dpp)3 2+(ClO42)2] Polystyrene 26 [Ru(dpp)3 2+(ClO42)2] PVC 20 Very curved 3 3 [Ru(phen)3 2+(ClO42)2] RTV-118 silicone rubber 12 [Ru(dpp)3 2+(ClO42)2] Ethyl cellulose or polystyrene 20 Analyst, 1999, 124, 1301–1307 1303workers18,19 to limit their modeling to r values � 0.35 and is a clear drawback to using a Gaussian distribution in to,i or, kq,i.For a Gaussian distribution in to,i it can be shown, using eqns. (2), (6), (8) and (9), that the ratio of the total unquenched to total quenched luminescence intensity, i.e., Io(total)/I(total), is related to pO2 through the expression: I I x x x x x k p x x o o,mdl o,mdl q o,mdl 2 (total) (total) exp(– d exp(– O d = Ú + + + + - • - • Ú Ú t r t r t r r r ( ) ) { ( ) ) /[ ( )]} / / 1 1 1 1 2 1 2 1 (11) For a Gaussian distribution in kq,i it can be shown that the ratio of the total unquenched to total quenched luminescence intensity, i.e., Io(total)/I(total), is related to pO2 through the expression: I I x x x k p x x o q,mdl o 2 (total) (total) exp(– d exp(– O d = + + - • - • Ú Ú 2 1 2 1 1 1 ) { ) /[ ( )]} / / r r t r (12) It is useful at this stage to define, for a Gaussian distribution in ti, a unitless parameter, q, which is related directly to the partial pressure of oxygen, pO2, i.e., q = kqto,mdlpO2 (13) Eqn.(13) allows a series of Stern–Volmer plots (Io(total)/ I(total) versus q) to be calculated using eqn. (11) and different values of r. In making these calculations, it was assumed initially that 80 different sites existed spanning the range x = 24 to x = 4, in steps of Dx = 0.1. The number of sites was reduced usually with the additional necessary condition that only sites with x � 21/r can exist, i.e., to,i (or, kq,i) cannot be < 0.Additional work showed that increasing the range in x and/ or decreasing the step size in 3 increased the computational time with little effect on the overall results. The results of these calculations are illustrated in Fig. 3(a) and are very similar to those reported by Demas et al.18 for the same model. Most notable amongst these results is that apparent high degree of linearity in the Stern–Volmer plots even when the breadth of the distribution is so large as to make the model unrealistic, i.e.> 0.35. This finding further undermines the usefulness of the model, since from Fig. 1 and Table 2 it is clear that many optical oxygen sensors produce Stern–Volmer plots with pronounced curvature. Demas and his co-workers17–19 were able to get round the lack of curvature exhibited by their initial Gaussian model by suggesting that there may be more than one Gaussian distribution, both narrow, but each with a different kq value.Under these conditions most of the curved Stern–Volmer plots associated with optical oxygen sensors can be generated, although, as with the two-site model to which it is closely related, it does not appear that the bimodal Gaussian model has any physical significance. It is useful also to examine what effect a Gaussian distribution in ki will have on an optical oxygen sensor. This can be done using the same approach as above, but this time using eqn.(12) and defining q = kq,mdltopO2. The results of this work are illustrated in Fig. 3(b) and are strikingly dissimilar to those in Fig. 3(a) in that the degree of curvature exhibited by the plots in the former figure increases markedly with increasing value of r. In the Stern-Volmer plots illustrated in Fig. 3(b) the range of curvature is well within that observed for most optical oxygen sensors. And yet this curvature may simply arise from the distortion of the Gaussian distribution with increasing value of r; a distortion that makes its suitability, as a physical model of optical oxygen sensors, unlikely.However, the results are encouraging enough to merit the investigation of other possible statistical distributions and this has led to the following work with log-Gaussian distributions. A log-Gaussian distribution model The log-Gaussian distribution is no stranger to heterogeneous systems. It has been used to fit successfully kinetic data from semiconductor surface states, colloidal semiconductors and fluorescent probes attached to biological membranes.27,28 A log-Gaussian distribution in the rate constant for a process is usually taken as a reflection of a Gaussian distribution in activation free energies for that process.If there is a log-Gaussian distribution in to,i, i.e.: rx = ln(to,i/to,mdl) (14) Alternatively, if there is a log-Gaussian distribution in ko,i, i.e.: rx = ln(ko,i/ko,mdl) (15) For a log-Gaussian distribution in to,i, the fraction of sites i compared with the modal number of sites, i.e., ni/nmdl, as a Fig. 2 Plot of the calculated ratio of fraction of sites i compared to the modal number of sites, i.e., ni/nmdl, { = exp(2x2)} as a function of the ratio to,i/to,mdl or kq,i/kq,mdl, calculated using a Gaussian distribution model, i.e. eqns. (8) and (9), or (8) and (10), respectively. Each profile corresponds to a different value of r, i.e., from inside-out r = 0.1, 0.5, 1, 1.5, 2, respectively.Fig. 3 Model generated Stern–Volmer quenching plots of Io(total)/I(total) versus q, calculated for a Gaussian distribution in: (a) natural lifetime of the lumophore, i.e., to,i, using eqns. (11) and (13) with r = (from bottom to top): 0, 1, 1.5, 2, 2.5 and 3, respectively; or (b) oxygen quenching rate constant, kq,i, using eqn. (12) and q = kq,mdl.to.pO2, with r = (from top to bottom): 0, 0.5, 0.65, 1, 2, and 3, respectively. 1304 Analyst, 1999, 124, 1301–1307function of the ratio to,i/to,mdl can be calculated using eqns.(8) and (14), for any value of r. A similar exercise can be performed for a Gaussian distribution in kq,i using eqns. (8) and (15) Fig. 4 illustrates the results of this work, i.e., a plot of ni/nmdl versus to,i/to,mdl (or kq,i/kq,mdl) for a range of r values from 0.1 to 2. From these results it can be seen that, unlike a Gaussian distribution, a log-Gaussian distribution has the attractive feature of generating to,i/to,mdl (or ko,I/ko,mdl) values that are � 0, whatever the value of r or x, cf.Fig. 4 and Fig.om the equations for a log-Gaussian distribution in to,i it follows that the ratio of the total unquenched to total quenched luminescence intensity, i.e., Io(total)/I(total), is related to q [related directly to pO2, cf. eqn. (13)] through the expression: I I x x x x x x x o(total) (total) exp( exp(– d exp( exp(– d = + -• • -• • Ú Ú r r q r ) ) { ) ) /[ exp( )]} 2 2 1 (16) Using eqn.(16) and different values of r, a series of Stern– Volmer plots (Io(total)/I(total) versus q) were generated and the results of this work are illustrated in Fig. 5(a). As with a Gaussian distribution in to,i it appears that only when the distribution in site natural lifetimes is large, i.e., r > 1, does the Stern–Volmer plot begin to show some signs of curvature. For a log-Gaussian distribution in to,i at any time t after an initial pulse of excitation light, the number of sites i present is given by the expression: ni,t = niexp2 [(1/to,i + kqpO2)t] (17) If we define a normalised time unit, t*, as t* = t/to,mdl, then it is also possible to predict what the luminescence decay profiles will look like with increasing q (i.e., increasing pO2) using the following expression: I I x x x x x x t t q r t r * * )exp [ exp( )) * exp )] ) = -• • -• • = - + Ú Ú 0 2 2 1 exp(– (– d exp(– d (18) Fig. 6 illustrates the results of a series of calculations using eqn.(18), of the It*/It*=o versus t* profiles for a log-Gaussian distribution in to,i with increasing q, with r set at 1. The insert diagram illustrates the first-order plot of the data and shows that as q, i.e., the partial pressure of oxygen, is increased, the decay kinetics become increasingly better first-order. This effect is typical of a model in which there is a distribution in to,i, but not in kq,i, which is set at a constant value kq. However, it is also a feature that is not usually seen with real optical oxygen sensors. As noted previously, cf.Table 2, with real optical sensors the decay kinetics, if anything, become less first order with increasing pO2. The lack of a marked degree of curvature in the model-generated Stern–Volmer plots with increasing r value and the predicted increase in the first order nature of the luminescence decay kinetics with increasing pO2 makes it unlikely that a log-Gaussian distribution in to,i is a useful model for the response features of any optical oxygen sensors reported to date.From the equations for a log-Gaussian distribution in kq,i it follows that the ratio of the total unquenched to total quenched luminescence intensity, i.e., Io(total)/I(total), is related to q ( = kq,mdl.to.pO2 = Kmdl.pO2)) through the expression: I I x x x x x (total) (total) exp(– d {exp(– d = + -• • -• • Ú Ú 2 2 1 ) ) /[ exp( )]} q r (19) Fig. 4 Plot of the calculated ratio of fraction of sites i compared to the modal number of sites, i.e., ni/nmdl, { = exp(2x2)} as a function of the ratio to,i/to,mdl or kq,i/kq,mdl, calculated using a log-Gaussian distribution model, i.e., eqns (8) and (14), or (8) and (15), respectively.Each profile corresponds to a different value of r, i.e., from inside-out r = 0.1, 0.5, 1, 1.5, 2, respectively. Fig. 5 Model generated Stern–Volmer quenching plots of Io(total)/I(total) versus q, calculated for a log-Gaussian distribution in: (a) natural lifetime of the lumophore, i.e., to,i, using eqns.(16) and (13) with r = (from bottom to top): 0, 1, 1.5, 1.75, and 2, respectively; or (b) oxygen quenching rate constant, kq,i, using eqn. (19) and q = kq,mdl.topO2, with r = (from top to bottom): 0, 1, 1.5, 2, 2.5 and 3, respectively. Fig. 6 Model-generated transient luminescence decay profiles for an optical oxygen sensor which exhibits a log-Gaussian distribution in to,i. The profiles were calculated using eqn.(18), for a sensor with r = 1 and q (from right to left) = 0, 0.3, 1, 3, 6 and 10. The insert diagram illustrates the firstorder plots of the kinetic data shown in the main diagram. Analyst, 1999, 124, 1301–1307 1305Using eqn. (19) and different values of r, a series of Stern– Volmer plots (Io(total)/I(total) versus q) were generated and the results of this work are illustrated in Fig. 5(b). In contrast to the profiles generated based on a log-Gaussian distribution in to,i , cf.Fig. 5(a), a log-Gaussian distribution in kq,i generates Stern- Volmer profiles of increasing curvature with increasing r. Indeed, in a previous paper4 eqn. (19) was used to fit a number of Stern–Volmer plots, of widely differing curvature, reported for real optical oxygen sensors. The model requires only two parameters (r and Kmdl) to provide a good fit, rather than the usual three or more associated with most other models. However, the model does not embrace all the kinetic features of those real optical sensors; features which are summarised in Table 2.Most notably, a log-Gaussian distribution in kq,i model assumes that there is no distribution in to,i so that at q = 0, i.e., pO2 = 0, the overall luminescence decay should fit perfectly first-order kinetics. In practice, only a few optical oxygen sensors show this feature. Fig. 1 illustrates Stern–Volmer data for some of the reported optical oxygen sensors which appear to show good first-order decay kinetics in the absence of oxygen, and increasingly worse first order kinetics with increasing pO2; the results are summarised in Table 1.The solid lines in Fig. 1 are the lines of best fit calculated using eqn. (19) and the optimised values for r and Kmdl given in Table 1. Simple methods for finding the values of r and Kmdl that give the best fit of the model-generated curves to the real Stern–Volmer data are given elsewhere.4. If the log-Gaussian distribution in kq,i model is really appropriate to such systems, although they be limited in number, it must also be able to predict accurately the decay kinetics exhibited by such systems.Unfortunately, very few details have been reported on the decay kinetics of the optical oxygen sensors listed in Table 1 and illustrated in Fig. 1. Thus, in order to test the model, a kinetic study was carried out on one of the optical oxygen sensors which gives a curved Stern– Volmer plot and exhibits good first-order luminescence decay kinetics at pO2 = 0, but not at pO2 ì 0; the optical oxygen sensor was [Ru(dpp)3 2+(Ph4B2)2] in non-plasticised cellulose acetate.11,23 The results of the kinetic study and the Stern– Volmer plot of the lifetime data are illustrated by the data points in Fig. 7 and its insert diagram, respectively. For a log-Gaussian distribution in kq,i at any time t after an initial pulse of light the number of sites i present is given by the expression: ni,t = niexp[2(1/to + kq,ipO2)t] (20) It follows that It*/It*=0 versus t* profiles can be generated for any value of q using the expression: I I x x x x x t t q r t * * )exp [ exp( )) *] ) = -• • -• • = - + Ú Ú 0 2 2 1 exp(– d exp(– d (21) In order to convert the model generated It*/It*=0 versus t* plot for any value of q and r, into one which can be compared with the data for a real optical sensor, optimised values for to and Kmdl are needed.The former converts the unitless time parameter, t* to real time, t, (t* = t/to) and the latter converts the unitless oxygen concentration parameter, q, to real pO2 (q = KmdlpO2). For the simple optical system under test, a value for to can be gleaned from a first-order plot of the luminescence versus time profile recorded in the absence of oxygen; the resulting gradient of the straight line = 21/to.Using this approach to was found to be 5.85 ms for the [Ru(dpp)3 2+(Ph4B2)2] in cellulose acetate optical sensor.Optimised values for Kmdl and r can be obtained from the Stern–Volmer plot of the real data, either in the form of Io/I or to,M/tM versus pO2. The latter requires finding the values of Kmdl and r which, when used with eqn. (19) of the model, generates a curve of best fit to the real Stern–Volmer data associated with the optical oxygen sensor under test. From Table 2 it can be seen that the optimised values for Kmdl and r for the [Ru(dpp)3 2+(Ph4B2)2] in cellulose acetate oxygen sensor were found to be 0.0033 Torr21 and 1.4, respectively.Using these values, the log-Gaussian distribution in kq,i model eqn. (19) and q = KmdlpO2 were used to generate the solid line in the insert diagram in Fig. 7 which provides a good fit to the observed Stern–Volmer plot of the lifetime data for the [Ru(dpp)3 2+(Ph4B2)2] in cellulose acetate optical oxygen sensor. The optimised parameters for to (5.85 ms), Kmdl (0.0033 Torr21) and r (1.4) were used in eqn.(21) to generate the solid lines luminescent decay profiles for the [Ru(dpp)3 2+(Ph4B2)2] in cellulose acetate oxygen sensor illustrated in the main diagram of Fig. 7 and recorded for a range of different pO2 levels; the fits appear reasonable and encouraging. Conclusion The ‘two-site’ and ‘Gaussian distribution in natural lifetime, to,’ models, used to describe the observed response characteristics of typical oxygen optical sensors based on luminescence, although good fitting procedures (especially the former), do not generate fitting parameters which have any likely physical reality.Ideally, a model should be able to unite all the response features of typical optical oxygen sensors, including downward curving Stern–Volmer plots and increasingly non-first order luminescence decay kinetics with increasing partial pressures of oxygen. A new ‘log-Gaussian distribution in to or kq’ model is introduced which has the advantage over a Gaussian distribution model of placing no limitation on the value of r.However, a ‘log-Gaussian distribution in to’ model generates Stern– Volmer quenching plots that show little degree in curvature, even at large r values and luminescence decays that become increasingly first order with increasing pO2. In fact, with real optical oxygen sensors, the opposite is observed and thus the model appears of little value. In contrast, a ‘log-Gaussian distribution in ko’ model does produce the trends observed with real optical oxygen sensors; although its use is restricted to Fig. 7 Transient decay profiles (5) recorded for a [Ru(dpp)3 2+(Ph4B2)2] in cellulose acetate oxygen optical sensor when exposed to the following (top to bottom) partial pressures of oxygen: 0, 160, 456 and 760 Torr, respectively. The solid lines are ‘log-Gaussian distribution in kq,i’ model generated decay profiles, calculated using eqn. (21) and to = 5.85 ms, Kmdl = 0.0033 Torr21 and r = 1.4. The insert diagram illustrates the Stern– Volmer plot based on lifetime measurements (5), i.e., to,M/tM versus pO2, for the [Ru(dpp)3 2+(Ph4B2)2] in cellulose acetate oxygen optical sensor.In this insert diagram the solid line is the ‘log-Gaussian distribution in kq,i’ model generated curve, calculated using eqn. (19) with Kmdl = 0.0033 and r = 1.4. 1306 Analyst, 1999, 124, 1301–1307those in which the kinetics of luminescence decay are good first order in the absence of oxygen, such as the [Ru(dpp)3 2+(Ph4B2)2] in cellulose optical oxygen sensor.It appears likely that a combination of a log-Gaussian distribution in ko with one in to will produce a model that is relevant to a much broader range of optical oxygen sensor, and this forms the basis of the next paper in this series. Acknowledgement The author thanks Ms. F. C. Williams for providing the data on the cellulose acetate based optical sensor. References 1 A. Mills, Platinum Met. Rev., 1997, 41, 115. 2 M. L. Hitchman, Measurement of Dissolved Oxygen, John Wiley, Geneva, 1978, ch. 1. 3 O. S. Wolfbeis, in Fibre Optic Chemical Sensors, ed. O. S. Wolfbeis, Vol II, CRC Press, Boca Ranton, FL, 1991, ch. 10. 4 A. Mills, Sens. Actuators B, 1998, 51, 69. 5 C. McDonagh, B. D. MacCraith and A. K. McEvoy, Anal. Chem., 1998, 70, 45. 6 E. R. Carraway, J. N. Demas and B. A. DeGraff, Langmuir, 1991, 7, 2991. 7 J. R. Bacon and J. N. Demas, Anal. Chem., 1987, 59, 2780. 8 P. Hartmann and W. Trettnak, Anal. Chem., 1996, 68, 2615. 9 W. W-S. Lee, K-Y. Wong and X-M Li, Anal. Chem., 1993, 65, 255. 10 I. Klimat and O. S. Wolfbeis, Anal. Chem., 1995, 67, 3160. 11 H. N. McMurray, P. Douglas, C. Busa and M. S. Garley, J. Photochem. Photobiol., 1994, 80, 283. 12 E. R. Carraway, J. N. Demas, B. A. Degraff and J. R. Bacon, Anal. Chem., 1991, 63, 337. 13 L. Sacksteder, J. N. Demas and B. A. DeGraff, Anal. Chem., 1993, 65, 3480. 14 J. N. Demas, B. A. DeGraff and W. Xu, Anal. Chem., 1995, 67, 1377. 15 X-M Li and K-Y. Wong, Anal. Chim. Acta, 1992, 262, 27. 16 X-M Li, F-C. Ruan and K-Y. Wong, Analyst, 1993, 118, 289. 17 E. R. Carraway, J. N. Demas and B. A. DeGraff, Anal. Chem., 1991, 63, 332. 18 J. N. Demas and B. A. DeGraff, SPIE, 1992, 1681, 2. 19 J. N. Demas and B. A. DeGraff, Sens. Actuators, 1993, 11, 35. 20 S. Draxler, M. E. Lippitsch, I. Klimant, H. Kraus and O. S. Wolfbeis, J. Phys. Chem., 1995, 99, 3162. 21 M. L. Bossi, M. E. Daraio and P. F. Aramendia, J. Photochem. Photobiol., 1999, 120, 15. 22 P. M. Gewehr and D. T. Delpy, Med. Biol. Eng. Comput., 1994, 659. 23 A. Mills and F. C. Williams, Thin Solid Films, 1997, 306, 163. 24 P. Hartmann and W. Trettnak, Anal. Chem., 1996, 68, 2615. 25 W. Xu, R. C. McDonough, B. Langsdorf, J. N. Demas and B. A. DeGraff, Anal. Chem., 1994, 66, 4133. 26 P. Hartmann, M. J. P. Leiner and M. E. Lippitsch, Anal. Chem., 1995, 67, 88. 27 W. J. Albery, P. N. Bartlett, C. P. Wilde and J. R. Darwent, J. Am. Chem. Soc., 1985, 107, 1854. 28 G. T. Brown, J. R. Darwent and P. D. I. Fletcher, J. Am. Chem. Soc., 1985, 107, 6446. Paper 9/02153A Analyst, 1999, 124, 1301–1307 1307

ISSN:0003-2654    

DOI:10.1039/a902153a

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Response characteristics of optical sensors for oxygen: a model based on a distribution in to and kq Andrew Mills† Department of Chemistry, University of Wales Swansea, Singleton Park, Swansea, UK SA2 8PP Received 18th March 1999, Accepted 12th July 1999 The typical response characteristics of most optical oxygen sensors include downward curving Stern–Volmer plots and often multi-exponential luminescence decay profiles in the absence and presence of oxygen. The wide range of different response features exhibited by optical oxygen sensors is usually attributed to different degrees of heterogeneity in the sensor films. This heterogeneity is described in this paper by a log-Gaussian distribution in the natural luminescent lifetime of the oxygen-quenchable lumophore, to, and the quenching rate constant, kq.A ‘log-Gaussian distribution in to and kq’ model is used to generate theoretical response profiles which exhibit the same disparate range of features as real optical oxygen sensors.However, unlike other models, such as the ‘two-site’ model, the ‘log-Gaussian distribution in to and kq’ model generates model parameter values which are physically plausible and consistent at all partial pressures of oxygen, pO2. The model is used to fit successfully the Stern–Volmer plots and luminescent decay profiles reported for a number of different optical oxygen sensors. For each of the real sensors examined, the values of the model parameters KSV,mdl, r1, r2 and to,mdl which give the best fit to the observed data are reported.The latter data allows prediction of the response features of the associated optical oxygen sensor at any value of pO2 Introduction Most optical oxygen sensors are based on luminescence quenching by oxygen, with the lumophore entrapped in a polymer medium.1 New optical oxygen sensors, based on this principle, are being developed all the time, such is the growing interest in this area of analysis.2 And yet, despite all this work, there does not exist a coherent model for the observed response characteristics of such optical sensors.These response characteristics include downward-curving Stern–Volmer plots (Io/I vs. pO2) and multi-exponential luminescent decay profiles;4–10 response characteristics that vary markedly from sensor to sensor. In general, it is assumed that these response features are due to some degree of heterogeneity in the lumophoreencapsulated type oxygen sensor.A major goal in optical oxygen sensor research is to find a model that provides an adequate description of the nature and degree of the heterogeneity of the system.3 Such a model should be able to embrace the whole gamut of response characteristics exhibited by optical oxygen sensors and still be on all occasions physically plausible. A good illustration of the typical response features of an optical oxygen sensor is provided by the work of Daraio and coworkers11 on dodecyl sulfate (DS2) salts of ruthenium(ii) diimine complexes in polydimethylsiloxane (silicone rubber, E- 4 from Wacker).Fig. 1 illustrates the luminescence decay profiles observed by these workers for ruthenium tris(4,7- diphenyl-1,10-phenanthroline), i.e., [Ru(dpp)3]2+ in the absence and presence of 100% oxygen. The insert diagram illustrates the first order plots of this kinetic data, both of which are curved. Thus, the luminescence decay profiles recorded in the absence or presence of oxygen for this optical oxygen sensor do not fit first order kinetics.Also in Fig. 1 is the Stern–Volmer plot of the data derived from a study of the variation of the luminescence intensity of the sensor film with oxygen partial pressure, pO2. As is common amongst optical sensors in which silicone rubber is used as an encapsulating medium, the Stern–Volmer plot exhibits a marked degree of downward curvature. Although models based on a ‘log-Gaussian distribution in to or kq’10 do not appear able to embrace all the response features exhibited by real optical oxygen sensors, there is the possibility that a ‘log-Gaussian distribution in to and kq’ model will.In this paper, the characteristics of this latter model are explored and the results tested using a range of data reported for real optical oxygen sensors based mainly on the lumophore [Ru(dpp)3]2+. Unless stated otherwise, all illustrated Stern–Volmer plots are based on intensity, rather than lifetime, measurements.† Present address: Department of Pure and Applied Chemistry, University of Strathclyde, Thomas Graham Building, 295 Cathedral Street, Glasgow, UK G1 1XL. E-mail: A.Mills@strath.ac.uk Fig. 1 Response features of the [Ru(dpp)3 2+(DS2)2] in silicone rubber (E- 4, Wacker) oxygen sensor reported by Daraio and coworkers.11 The main left hand side diagram illustrates a reconstruction (without scatter) of the observed decay profiles recorded for the sensor in the absence (solid line) and presence (broken line) of 100% oxygen.The insert diagram illustrates the first order plots of these two sets of decay data. The right hand side diagram illustrates the Stern–Volmer plot based on intensity measurements reported11 for this sensor. The solid line was calculated using eqns. (7) and (8) of the ‘log-Gaussian distribution in to and kq’ model and the optimum fit data reported in Table 1 for the sensor, vide infra.Analyst, 1999, 124, 1309–1314 1309A log-Gaussian distribution in to and kq model12–15 In a log-Gaussian model it is assumed that the heterogeneity within the system manifests as a log-Gaussian distribution in rate. In the case of optical oxygen sensors, the response features are controlled by two kinetic processes, namely: (i) intramolecular deactivation of the electronically excited state of the lumophore, a measure of which is the reciprocal of the natural lifetime of the excited state, to, and (ii) inter-molecular quenching of the excited state by oxygen, a measure of which is kq·pO2.In a log-Gaussian distribution model,12–15 the number of sites i, i.e. ni, is related to the modal number of sites, nmdl, by the expression: ni/nmdl = exp(2x2) (1) In the ‘log-Gaussian distribution in to and kq’ model it is assumed that there is a log-Gaussian distribution in to,i, i.e.: r1 x = ln(to,i/to,mdl) (2) where r1 is a measure of the breadth of the distribution with respect to to,i, and to,mdl is the natural lifetime associated with the modal number of sites.In this model it is also assumed that there is a log-Gaussian distribution in kq,i, i.e.: r2 x = ln(kq,i/kq,mdl) (3) where r1 is a measure of the breadth of the distribution with respect to kq,i, and kq,mdl is the quenching rate constant by oxygen (partial pressure = pO2) of the modal number of sites. The model is not prescriptive in the relative signs of r1 and r2, thus, the value of r2 can be of the same sign or opposite to that of r1.However, it could be argued that it is more likely that the values of r1 and r2 for any real system will have opposite signs, since this will mean that sites with quenching rate constants that are larger than kq,mdl will be associated with lumophore molecules with natural lifetimes which are shorter than to,mdl and vice versa. The alternative situation, namely, where r1 and r2 have the same sign, results in a scenario in which quenching rate constants that are larger than kq,mdl will be associated with lumophore molecules with natural lifetimes which are larger than to,mdl, and vice versa, and this intrinsically appears to be a less likely situation.In such a system, for any type of site, i, Io,i/Ii = 1 + to,i kq,i pO2 (4) The observed overall variation in sensor luminescence intensity in the absence and presence of oxygen, i.e., Io(total) and I(total), respectively, are related by the following expression: Io(total)/I(total) = ·Io,i/·Ii (5) Thus, it follows from eqns.(3), (4) and (5) that: Io(total)/I(total) = ·Io,i/·{Io,i/[1 + to,i kq,i pO2]} (6) It can be shown7 that for any site i, Io,i ª toi ni, thus, using this expression, with eqns. (1) and (6), the following Stern–Volmer equation for a ‘log-Gaussian distribution in to and kq’, with respect to x, can be derived: I I x x x x x x x o ( ) ( ) exp( )exp( ) {exp( )exp( ) /[ exp({ } )]} total total d d = - - + + -• • -• • Ú Ú r r q r r 1 2 1 2 1 2 1 (7) where q = KSV,mdl pO2 = to,mdl kq,mdl pO2 (8) and KSV,mdl is the Stern–Volmer constant for the modal number of sites, nmdl.Eqn. (7) allows the Stern–Volmer plot Io/I versus q (the latter parameter is proportional to pO2) for the ‘log-Gaussian distribution in to and kq’ model to be calculated, for any given set of values of r1 and r2. All the model-generated results reported in this paper derived from calculations based on the assumption that the log-Gaussian distribution consists of 80 different sites, spanning the range x = 24 to x = 4 in steps of Dx = 0.1.Additional work showed that increasing the range in x and/or decreasing the step size in x increased the computational time with little effect on the overall results. Fig. 2 illustrates the model predicted Stern–Volmer plots for the cases when r1 is fixed (at 0.5) and r2 is varied and vice versa; the plots were generated using eqn.(7). In generating these curves it was decided for simplicity, and for reasons outlined above, to set all values of r2 negative. The modelpredicted Stern–Volmer plots illustrated in Fig. 2 show that whereas a log-Gaussian distribution in kq causes a significant degree of downward curvature, a log-Gaussian distribution in to does not. Since most optical oxygen sensors exhibit some degree of curvature in their Stern–Volmer plots, the results derived from the log-Gaussian model appear to suggest that the curvature is largely due to a distribution in kq, rather than to.From the log-Gaussian model results illustrated in Fig. 2 it appears that with an increasing distribution in kq the optical sensor becomes less sensitive towards oxygen, implying that the sites with kq < kq,mdl have a much greater effect on the overall observed luminescence than those sites for which kq > kq,mdl. In contrast, from the same results, with an increasing distribution in to the optical sensor appears to become more sensitive towards oxygen, implying that the sites with to > to,mdl have a much greater effect on the overall observed luminescence than those sites for which to < to,mdl.One of the problems associated with applying eqn. (8) to fit the Stern–Volmer data associated with real optical oxygen sensors is the fact that there are three variables, i.e., KSV,mdl, r1 and r2. For easier, more reliable curve-fitting, it would be better if one or more of these variables could be gleaned from other experiments and this can be achieved from a study of the sensor’s luminescence decay profiles, recorded in the absence and presence of oxygen.For a ‘log-Gaussian distribution in to and kq’, at any time t after an initial pulse of light, the number of electronically excited sites i still present is given by the expression: ni,t = ni exp[2(1/ti + kq,i pO2) t] (9) where ni is defined by eqn. (1). If the normalised time unit, t* is defined as t* = t/to,mdl, then combining eqns.(1)–(3) and (9) Fig. 2 ‘Log-Gaussian distribution in to and kq’ model generated Stern– Volmer curves, calculated using eqn. (7), in the form Io/I versus q, where q is ª pO2 [see eqn. (8)]. The left hand side diagram illustrates a typical set of Stern–Volmer plots in which the value of r1 is fixed at 0.5 and r2 is (from top to bottom): 0, 20.5, 21, 21.5, 22 and 23. The right hand side diagram illustrates a typical set of Stern–Volmer plots in which the value of r2 is fixed at 20.5 and r1 is (from bottom to top): 0, 0.5, 1, 1.5, 2 and 3. 1310 Analyst, 1999, 124, 1309–1314together allows the following expression for the modelpredicted normalised luminescence decay profile to be generated for the situation where oxygen is present: I I x x x x x x t t q r r t r * * exp( )exp [( exp({ } )) *exp( )] exp( ) = -• • -• • = - + + - - Ú Ú 0 2 1 2 1 2 1 - d d (10) When oxygen is absent from the system, i.e., q = 0, eqn.(10) reduces to the following expression: I I x x x x x t t t r * * exp( )exp [ *exp( )] exp( ) = -• • -• • = - - - Ú Ú 0 2 1 2 - d d (11) The above equation is identical in form to eqn. (5) in ref. (12), a study of the kinetics associated with semiconductor photocatalysts. In the latter study, a simple evaluation procedure involving Simpson’s rule was used. These workers12 also noted the occurrence of a common intersection point on their decay curves.Fig. 3 illustrates the model-predicted decay curves, calculated using eqn. (11), i.e., It*/It*=o versus t* for pO2 = 0 (i.e., q = 0), for a range of different r1 values. Interestingly, from the results of this work, at It*/It*=o = 0.429, t* = 0.831, regardless of the value of r1. Thus, a value of to,mdl can be calculated from a ‘real It/It=o versus t’ luminescence decay profile, recorded for a real optical oxygen sensor under test in the absence of oxygen, simply by measuring, from the real decay profile, the time, t0.429, at which It/It=0 = 0.429, since to,mdl = t0.429/0.831 (12) Once a value for to,mdl is known for a particular sensor then any model-generated ‘It*/It*=o versus t*’ plot can be converted to a more useful form, i.e., ‘model It/It=o versus t’ plot, by converting all the model t* values into real time, t, values by multiplying them by to,mdl.Once a model-generated ‘It*/It*=o versus t*’ plot has been converted into its ‘model It/It=o versus t’ plot form, comparison with the ‘real It/It=o versus t’ plot is simple and through an iterative process, a value of r1 which gives an optimal fit to the real data can be obtained.The above procedure provides best fit values for to,mdl and r1, however, in order to obtain a complete description by the model of the response features of the real sensor, values for r2 and KSV,mdl are also required. One way is to find the values of r2 and KSV,mdl that give the best model fit to the observed Stern– Volmer data for the sensor, i.e.Io/I versus pO2, as calculated using eqns. (7) and (8). Alternatively, values for r2 and KSV,mdl can be obtained by fitting the model to any observed normalised decay profile for the sensor, i.e., It/It=o versus t, at any oxygen level > 0 Torr, using eqns. (8) and (10). In practice, it is easier to find the optimal values for r2 and KSV,mdl by fitting eqns. (7) and (8) to the observed Stern–Volmer plot for the sensor.Once optimised values for the parameters: r1, r2, KSV,mdl and to,mdl have been obtained for a real sensor, then the model can be used to describe its response characteristics, such as its Io/I value or decay profile, at any value of pO2. In addition, the optimised values of each of these parameters gives an insight into the physical nature of the optical oxygen sensor system. For example, small values for r1 and r2 indicate little variation in lumophore lifetime and quenching rate constant between the various sites in the sensor film, whereas large values indicate large variations.Sensors with similar, modal quenching values, i.e., similar KSV,mdl values, may have quite different overall sensitivities towards oxygen, simply due to different distributions in r1 and r2. The key difference between the ‘log- Gaussian distribution in to and kq’ model, and all the others which have gone before it, is that the former appears to provide a coherent, physically plausible model that unifies the observed, often complex, luminescence decay kinetics associated with any and every optical oxygen sensor with its respective, and often curved, Stern–Volmer plot. Each of the optimised values of the fitting parameters will provide valuable information about the micro-environment in the sensor film and help build up a better understanding of the factors which control the response features exhibited by optical oxygen sensors.In the following sections, various ‘distribution in to and kq’ scenarios are considered for the ‘log-Gaussian distribution in to and kq’ model and, where possible, examples from the literature are identified and the model tested using the associated reported data.No distribution in to and kq ( r1 = r2 = 0); the simplest case If there is no heterogeneity in an optical oxygen sensor with respect to to and kq, then, according to the model, r1 and r2 will both have a value of zero. This situation is usually only ever attained if the lumophore is dispersed homogeneously throughout the quenching medium; e.g., as is usually the case when the lumophore is in solution. Under these conditions quenching of the luminescence by oxygen in such a ‘sensor’ would be expected to fit the standard Stern–Volmer equation, i.e.: to/t = Io/I = 1 + KSV pO2 (13) where KSV = kq.to.When oxygen-quenchable lumophores are encapsulated in polymers and sol–gels to create oxygen sensor films usually an element of heterogeneity is introduced and this simple ‘homogeneous kinetics’ case no longer applies.No distribution in to but a distribution in kq ( r1 = 0, r2 ­ 0) The typical characteristics of an optical sensor for oxygen which shows evidence of little or no distribution in to, but some evidence of a distribution in kq, are firstly, the luminescence Fig. 3 ‘Log-Gaussian distribution in to and kq’ model generated normalised luminescence decay curves for pO2 = 0, calculated using eqn.(11) and the following different r1 values (from the bottom, left to right): 0.2, 0.7, 1, 1.3, 1.6 and 2, respectively. Note the common intersection of all the decay curves at It*/It*=0 = 0.429 and t* = 0.831. Analyst, 1999, 124, 1309–1314 1311decay observed in the absence of oxygen gives an excellent fit to first-order kinetics, but the decays in the presence of oxygen do not. In the latter case the decays give increasingly worse fits to first order kinetics with increasing pO2.Secondly, the Stern– Volmer plot is either linear or slightly curved. The above situation is a special case for the ‘log-Gaussian distribution in to and kq’ model; a case in which r1 = 0 and the modulus of r2, |r2|, is > 0. In fact, in the case where r1 = 0, the sign of r2 is immaterial to the model predictions because of the nature of the equations. This special case was considered in the previous article10 in this two-part series; the model was used to fit data from appropriate examples taken from the literature.Some of these examples and the best-fit-to-the-model values of KSV,mdl and r2 are given in Table 1.11,16–20 A distribution in to and kq ( r1 and r2 ­ 0) The major response characteristics of most optical oxygen sensors can be summarised as follows: (i) the observed luminescence decays recorded for the sensor in the absence or presence of oxygen do not give good fits to first order kinetics and (ii) the Stern–Volmer plot is usually curved.The model predicts that such characteristics are typical of optical oxygen sensors in which there are distributions in both to and kq, i.e., r1 > 0 and |r2|, is > 0. In order to test the model, luminescence decay data and Stern–Volmer plots for various optical oxygen sensors were gleaned from the literature and an attempt was then made to fit the model to the observed data; the results of this work are as follows. Lippitsch and coworkers19 reported the luminescent decay profile of an optical oxygen sensor, [Ru(dpp)3 2+(ClO42)2] encapsulated in polystyrene, in the absence of oxygen.In addition, these workers reported the Stern–Volmer plot for this sensor. The results of this work are illustrated in Fig. 4. The main diagram comprises data points, with error bars (to show the extent of the scatter on the decay data), taken from the luminescence decay profile reported by Lippitsch and coworkers19 in the absence of oxygen.The solid line in this figure, is the ‘log-Gaussian distribution in to and kq’ model decay curve, generated using eqns. (11) and (12), and values of to and r1 of 2.68 ms and 1.2, respectively. The insert diagram in Fig. 4 illustrates the Stern–Volmer plot of the data points reported by Lippitsch and coworkers19 for the same [Ru(dpp)3 2+(ClO42)2] in polystyrene optical oxygen sensor and the solid line is the Stern–Volmer plot predicted by the model, using eqns.(7) and (8) and optimised values of KSV,mdl, r1 and r2 of 0.00265 Torr21, 1.2 and 22.2, respectively (1 Torr = 133.322 Pa). Thus, the ‘log-Gaussian distribution in to and kq’ model appears to be able to fit the observed data for the [Ru(dpp)3 2+(ClO42)2] in polystyrene sensor very well. Once the values of to,mdl, KSV,mdl, r1 and r2 have been determined for an optical oxygen sensor, then the model will allow the sensor’s Io/I value or its normalised decay profile to be predicted for any value of pO2.As noted earlier, Daraio and coworkers11 have reported the luminescent decay profiles of a [Ru(dpp)3 2+(DS2)2] in silicone rubber (E-4, Wacker) oxygen sensor, in the absence and presence of oxygen, see Fig. 1. These workers11 typically were able to begin recording the decay profiles 0.3 ms after the initial pulse of light and fitted the observed decay data to a stretched exponential, i.e.: It = Io.exp(2B At) (14) where B is a constant which depends upon pO2.However, when the results of this work are fitted to eqn. (14), the extrapolated values of Io appear to vary with pO2,21 which makes little physical sense, and, thus, the model should be considered, just like the ‘two-site’ model, only as a useful fitting routine. In the main part of Fig. 5(a), data points associated with the decay profiles observed by Daraio and coworkers11 for their [Ru(dpp)3 2+(DS2)2] in silicone rubber optical oxygen sensor, in the absence and presence of oxygen, are illustrated; the insert diagram illustrates the first order plot of the decay profiles in the main diagram.Fig. 5(b) illustrates the Stern–Volmer plot of the data points reported by Daraio and coworkers11 for their sensor. In both Figs. 5(a) and 5(b), the solid lines correspond to the ‘log- Gaussian distribution in to and kq’ model predicted curves, calculated using the following optimised values of KSV,mdl, r1, r2 and to,mdl of: 0.0019 Torr21, 1.36, 2.64 and 1.32 ms, respectively. Once again the ‘log-Gaussian distribution in to and kq’ model appears able to fit the response features exhibited by an optical oxygen sensor.It is interesting, but, at present, by no means obvious why, the values of r1 and r2 have the same sign, implying that, in the E-4 silicone rubber matrix, sites which have a high kq also have a long to, and vice versa. Finally, Leiner and coworkers20 have reported a substantial body of kinetic data concerning the luminescent decay profiles of a [Ru(dpp)3 2+(ClO2)2] in polystyrene oxygen sensor. This kinetic data is based on fits to the observed decay profiles based on a ‘two-site’ model and includes: k1, k2 and to,M/tM 7 data for the sensor at various pO2 levels, spanning the range 0–727 Torr.This kinetic information was used in the present work to Fig. 4 Luminescence decay profile, recorded in the absence of oxygen, reported by Lippitsch and coworkers19 for their optical oxygen sensor: [Ru(dpp)3 2+(ClO42)2] encapsulated in polystyrene.The error bars are provided to give some idea of the scatter on the original reported decay trace19. The insert diagram illustrates the Stern–Volmer plot, based on intensity measurements, reported for this sensor.19 In both diagrams, the solid lines were calculated using the appropriate ‘log-Gaussian distribution in to and kq’ model eqns., i.e., eqns. (11) and (12) for the luminescence decay profile and eqns.(7) and (8) for the Stern–Volmer profile, and the optimum fit data reported in Table 1 for the sensor. Fig. 5 (a) Luminescence decay profiles, recorded in the absence (5) and presence (-) of oxygen, reported by Daraio and coworkers11 for their [Ru(dpp)3 2+(DS2)2] in silicone rubber (E-4, Wacker) optical oxygen sensor. Note that these workers were only able to record the decay 0.3 ms after the initial flash. The insert diagram illustrates the first order plot of the data in the main diagram.(b) The Stern–Volmer plot based on intensity measurements reported for this sensor.11 In both diagrams, the solid lines were calculated using the appropriate ‘log-Gaussian distribution in to and kq’ model equations, and the optimum fit data reported in Table 1 for the sensor. 1312 Analyst, 1999, 124, 1309–1314generate a series of normalised decay curves spanning the pO2 range 0–727 Torr. Thus, a different decay curve was generated for each different value of pO2 selected, with each curve defined by 20 calculated ‘data’ points.The results are illustrated in the main diagram in Fig. 6. In addition, Leiner and coworkers20 reported two sets of Stern–Volmer plot data for their [Ru- (dpp)3 2+(ClO2)2] in polystyrene optical oxygen sensor; one set based on luminescence intensity measurements (Io/I data, open circles) and the other based on lifetime measurements (to,M/tM data; filled triangles); these data are illustrated in the insert diagram in Fig. 6. In both the main and the insert diagrams in Fig. 6, the solid lines correspond to the ‘log-Gaussian distribution in to and kq’ model predicted curves, calculated using the following optimised values of KSV,mdl, r1, r2 and to,mdl of: 0.0022 Torr21, 0.4, 21.0 and 5.63 ms, respectively. Once again the ‘log-Gaussian distribution in to and kq’ model appears able to fit the response features exhibited by an optical oxygen sensor. The results of the fitting of the ‘log-Gaussian distribution in to and kq’ model to the response features exhibited by real optical oxygen sensors are summarised in Table 1.It is interesting to note from the data contained in this table that for the [Ru(dpp)3]2+ based sensors, the use of dodecylsulfate as the lumophore’s counter anion reduces the lifetime (to,mdl) of the complex from its usual value of ca. 5.3 ms in aqueous solution to 1–2 ms. implying some anion quenching ability. In contrast, the ClO42 or BPh42 anions do not appear to alter significantly the natural lifetime of the quencher, i.e., to,mdl is ca. 5.7 ms. BPh42 appears to have greater potential than ClO42 as a counter ion. For example, in contrast to ClO42, it not only has a negligible lumophore quenching ability, but it is also very lipophilic, and so capable of solubilising hydrophilic lumophore cations in lipophilic polymers. Conclusion The ‘log-Gaussian distribution in to and kq’ model appears capable of providing a physically plausible rationale for the whole range of disparate response features exhibited by optical oxygen sensors.The latter response features, with the simple interpretations offered by the ‘log-Gaussian distribution in to and kq’ model are summarised in Table 2. The model offers not only a qualitative rationale for the response features exhibited by an optical oxygen sensor, but also, from the few examples considered in this paper, provides a quantitative interpretation of these features, provided sufficient data is available to determine the sensor’s values for the parameters KSV,mdl, r1, r2 and to,mdl.The latter data enables the model to then predict the sensor’s Io/I value or its normalised decay profile at any value of pO2. The values of KSV,mdl, r1, r2 and to,mdl for a sensor can Fig. 6 Luminescence decay profiles, recorded at different ambient pO2 levels, reported by Leiner and coworkers20 for their [Ru(dpp)3 2+(ClO2)2] in polystyrene optical oxygen sensor.The different decay curves (top to bottom) correspond to the following different pO2 levels: 0, 142, 359 and 727 Torr, respectively. The insert diagram illustrates the Stern–Volmer plots based on intensity (2) and lifetime (:) measurements reported for this sensor.20 In both diagrams, the solid lines were calculated using the appropriate ‘log-Gaussian distribution in to and kq’ model equations, i.e., eqns. (8), (10)-(12) for the luminescence decay profiles and eqns.(7) and (8) for the Stern–Volmer profile, and the optimum fit data reported in Table 1 for the sensor. Table 1 Characteristics of some typical optical oxygen sensors Lumophorea Encapsulation medium KSV,mdl/ Torr21 r1 r2 to/ms Ref. [Pd(II)OEPK] Polystyrene 0.22 0 1.12 — 16 [Pt(dpp)(CN2)2] Silicone rubber (RTV 732; GE) 0.086 0 3.15 — 17 [Ru(dpp)3 2+(BPh42)2] Cellulose acetate 0.0033 0 1.4 5.85 10, 18 [Ru(dpp)3 2+(ClO42)2] Polystyrene (M: 250 000) 0.00265 1.2 22.2 2.68 19 [Ru(dpp)3 2+(DS2)2] E-4 (Wacker) silicone rubber 0.0019 1.36 2.64 1.32 11 [Ru(dpp)3 2+(ClO42)2] Polystyrene (M: 240 000) 0.0022 0.4 21 5.63 20 a OEPK: octaethylporphyrin ketone; DS2: dodecyl sulfate; ph: 1,10 phenanthroline.Table 2 Typical response features of oxygen optical sensors and their interpretation using a ‘log-Gaussian distribution in to and kq’ model Lifetime studies: no. of exponentials Stern–Volmer plot N2 O2 ‘Log-Gaussian distribution in to and kq’ model interpretation Linear 1 2 An excellent fit to first-order kinetics in the absence of O2 is indicative of there being little or no distribution in to in the sensor film, i.e., r1 = 0 .The need for two exponentials to fit the luminescence decay data in the presence of O2 indicates that there is a distribution in kq. However, the apparent linearity of the Stern–Volmer plot indicates that the distribution is not large, i.e., typically, the modulus value of r2 lies in the range: 0 > |r2| > 0.5 Linear 2 2 The need for two exponentials to fit the luminescence decay data in the absence and presence of O2 indicates that there is a distribution in both to and kq, i.e., r1 and r2 are both non-zero. The linearity of the Stern–Volmer plot indicates that the distribution in kq is not large, i.e., 0 > |r2| > 0.5 Slightly curved 1 2 r1 = 0 and, typically, |r2| > 0.75 Curved 2 2 Typically, r1 > 0.5 and, |r2| > 1 Very curved 2 or 3 3 Typically, r1 > 1 and |r2| > 1; r1 and r2 may also have the same sign, i.e., sites with large to values have large kq values, and vice versa Analyst, 1999, 124, 1309–1314 1313be obtained from either of the following sources of experimental data: (i) the combination of the Stern–Volmer plot and the luminescence decay in the absence of O2, or (ii) the luminescence decays in the absence and presence of O2.The ‘log-Gaussian distribution in to and kq’ model, as it stands, may not be appropriate if the encapsulating system is clearly biphasic, e.g., in a silicone rubber, or any polymer, incorporating a silica filler, as is the case with some RTV silicone rubbers.However, such a bi-phasic system would reasonably be expected to fit a model comprising two log-Gaussian distributions in to and kq. References 1 A. Mills, Platinum Met. Rev., 1997, 41, 115. 2 O. S. Wolfbeis, in Fibre Optic Chemical Sensors, ed. O. S. Wolfbeis, Vol II, CRC Press, Boca Ranton, FL, 1991, ch. 10. 3 A. Mills, Sen. Actuators B, 1998, 51, 69. 4 E. R. Carraway, J. N. Demas, B. A. Degraff and J. R. Bacon, Anal. Chem., 1991, 63, 337. 5 L. Sacksteder, J. N. Demas and B. A. DeGraff, Anal. Chem., 1993, 65, 3480. 6 J. N. Demas, B. A. DeGraff and W. Xu, Anal. Chem., 1995, 67, 1377. 7 E. R. Carraway, J. N. Demas and B. A. DeGraff, Anal. Chem., 1991, 63, 332. 8 J. N. Demas and B. A. DeGraff, SPIE, 1992, 1681, 2. 9 J. N. Demas and B. A. DeGraff, Sens. Actuators, 1993, 11, 35. 10 A. Mills, Analyst, 1999, 124, 1301. 11 M. L. Bossi, M. E. Daraio and P. F. Aramendia, J. Photochem. Photobiol., 1999, 120, 15. 12 W. J. Albery, P. N. Bartlett, C. P. Wilde and J. R. Darwent, J. Am. Chem. Soc., 1985, 107, 1854. 13 G. T. Brown, J. R. Darwent and P. D. I. Fletcher, J. Am. Chem. Soc., 1985, 107, 6446. 14 A. Mills, P. Douglas and A. Green, J. Photochem. Photobiol. A, 1990, 53, 127. 15 A. Mills, P. Douglas, A. Green and G. Williams, Electrochemistry in Colloids and Dispersions, ed. R. A. Mackay and J. Texter, VCH Publishers, Inc., New York, 1992, ch. 29. 16 P. Hartmann and W. Trettnak, Anal. Chem., 1996, 68, 2615. 17 W. W.-S. Lee, K.-Y. Wong and X.-M. Li, Anal. Chem., 1993, 65, 255. 18 H. N. McMurray, P. Douglas, C. Busa and M. S. Garley, J. Photochem. Photobiol., 1994, 80, 283. 19 S. Draxler, M. E. Lippitsch, I. Klimant, H. Kraus and O. S. Wolfbeis, J. Phys. Chem., 1999, 99, 3162. 20 P. Hartmann, M. J. P. Leiner and M. E. Lippitsch, Anal. Chem., 1995, 67, 88. 21 M. E. Daraio, personal communication. Paper 9/02155H 1314 Analyst, 1999, 124, 1309–1314

ISSN:0003-2654    

DOI:10.1039/a902155h

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Evaluation of an immunobiosensor for the on-site testing of veterinary drug residues at an abattoir. Screening for sulfamethazine in pigs G. Andrew Baxter*, Mary C. O’Connor, Simon A. Haughey, Steven R. H. Crooks and Christopher T. Elliott Veterinary Sciences Division, Department of Agriculture for Northern Ireland, Stoney Road, Stormont, Belfast, UK BT4 3SD. E-mail: a.baxter@qub.ac.uk Received 24th May 1999, Accepted 23rd July 1999 A study was conducted to determine the feasibility of performing “on-site” screening for sulfamethazine (SMT), at an abattoir, using a rapid immunobiosensor method.This involved transfer of the biosensor technology and an assay developed in the laboratory, to the cold, humid conditions of a modern pig-processing factory. A pre-determined threshold limit of 0.4 mg ml21 SMT in bile was used to identify the likelihood that corresponding tissue samples contained SMT concentrations in excess of the European maximum permissible residue limit of 0.1 mg kg21.Bile samples containing SMT concentrations above the threshold limit were deemed positive and the corresponding kidney and muscle samples were sent to the laboratory for HPLC analysis. The robustness of the biosensor instrumentation in the harsh operating conditions was monitored throughout the project. The performance of the assay, on-site, was assessed by the regular inclusion of QA samples and by the submission of control ‘SMT-positive’ pigs to the abattoir.Sampling procedures, identification and traceability were also under scrutiny. During the project, 337 (9.35%) of the total kill were tested for SMT residues, representing 75% of all producers submitting pigs for slaughter. Twelve animals, including the ten controls, gave positive bile results. HPLC analysis confirmed SMT residues in all 12 kidneys (11 in excess of the permissible level). Ten muscle samples also contained violative SMT levels. Throughout the project, the biosensor performed reliably, with no adverse reaction of any mechanical or electrical components. The SMT assay also performed reliably.This is the first report of a biosensor being used for ‘on-site’ drug screening. Introduction Sulfonamides are widely used in intensive pig production for therapeutic and prophylactic reasons.1 Despite a recommended withdrawal period, sulfonamide residues are commonly detected in edible pig tissues. This may be due to non-adherence of the withdrawal period prior to slaughter, cross-contamination between medicated and non-medicated feeds in mills, or contact between treated and non-treated pigs.2 In Europe, an acceptable maximum residue limit (MRL) of 0.1 mg kg21 total sulfonamide has been set by the Veterinary Products Committee.3 In Northern Ireland (NI), sulfonamide residues are monitored as part of the NI Pig Testing Scheme.2 This includes a two stage test programme, where initially a screening test is used to sift large numbers of samples for potential positives.This is followed by confirmatory analysis of samples identified by the screening method as requiring further investigation. Producers presenting positive pigs are then targeted for on-farm investigation and intensive sampling. Positive carcasses are removed from the supply chain at the producer’s expense. Together, this approach provides an efficient and cost effective means of controlling drug residues in pig products intended for human consumption.Since its introduction in 1988, the frequency of violative sulfonamide residues has dramatically decreased from 6.8% to less than 0.5%.2 A common approach employed in drug screening is to use fluid matrices (serum, urine, bile) as predictors of sulfonamide levels in tissue.4–6 This removes the need for time-consuming extraction procedures. In NI, to screen for sulfamethazine (SMT) residues, bile is tested using a rapid biosensor assay.Using a predetermined threshold limit of 0.4 mg ml21, this method gives an accurate indication that the corresponding tissue sample contains sulfamethazine residues in excess of the MRL.7 Bile samples with sulfamethazine concentrations in excess of 0.4 mg ml21 are deemed positive and the corresponding tissue sample is forwarded for confirmatory analysis by HPLC.8 Consumer awareness with regard to food safety issues has increased in recent years. Demand for quality produce, free from drug residues, adds pressure on not only the food industry, but also on regulatory authorities to increase testing.The best means of providing this requirement is currently subject to debate. It is clear, however, that the increased workloads will increase costs. The biosensor method has the potential for handling increased sample numbers, but is still reliant on the samples being collected at the abattoir and then transported to the laboratory. Transferring the technology out of the laboratory and into the abattoir would reduce the costs involved in sample transfer and security.An ideal scenario would be an ‘on-line’ instrument with automated sampling. However, such technology is not available at present. The aim of this study was to take a commercially available biosensor instrument (Biacore AB) and subject it to the cold, humid conditions of an abattoir, by installing it ‘on-line’ and performing the testing immediately after sample collection.A small, family run factory (Stevenson & Co. Ltd.) which processes approximately 180 pigs per hour and kills on four days of each week, cooperated in the project. Goals were set to test approximately 10% of the total kill and at least one pig from each producer submitting animals for slaughter. Another aim was to determine if all results could be made available before the carcasses left the chill rooms for processing. During the study, high priority was given to sample identification and Analyst, 1999, 124, 1315–1318 1315traceability. The robustness and reliability of the biosensor and assays were closely monitored in the harsh environment of the abattoir.Experimental Equipment An optical biosensor (BIACORE 1000, Biacore AB, Uppsala, Sweden) was installed in the abattoir, in close proximity to the kill line (Fig. 1). Instrument performance was monitored for 3 d prior to the commencement of sample analysis to allow the mechanical and electrical components time to equilibrate to the environment. BIACORE Control 1.2 software was used for instrument operation and BIAevaluation 1.0 software for data handling.Chemicals and reagents Sulfamethazine reference standard, min. 99%, was obtained from Sigma Chemical Co. (Poole, Dorset, UK). All other chemicals were purchased from Sigma Chemical Co. Sensor chips (Type CM5), amine coupling kit and HBS buffer (0.01 M HEPES buffer, pH 7.4, 0.15 M sodium chloride, 3 mM EDTA, 0.005% polysorbate 20 v/v) were obtained from Biacore AB (Uppsala, Sweden).Bile was obtained from pigs, reared in an experimental unit under sulfonamide free conditions, for use as a negative control and in preparation of a standard curve. Antibody production and cross-reactivity To prepare the immunogen, SMT was covalently coupled to bovine thyroglobulin (BTG) via the carbodiimide method and purified by extensive dialysis.9 A polyclonal antibody was raised in a rabbit (Harlam UK Ltd, Oxon, UK) by subcutaneous injection with 1 mg of the SMT-BTG immunogen, as a waterin- oil emulsion with Freund’s adjuvant.10 Booster injections were given fortnightly.Test bleeds were obtained from the marginal ear vein and monitored for anti-SMT activity. When the titre had stabilised the antiserum was collected. The crossreactivity profile was determined in the presence of bile using the biosensor method. Immobilisation of SMT to sensor chip surface Sensor chips (Type CM5, research grade) were coated with sulfamethazine by the method of Sternesjo et al.11 Briefly, the carboxymethyldextran surface was activated by the injection of 35 ml of a 1+1 mixture of 0.05 M N-hydroxysuccinimide (NHS)/ 0.2 M 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC), at a flow rate of 5 ml min21. 35 ml of a SMT solution (1 mg ml21 in 10 mM sodium acetate containing 20% dimethylformamide) was then injected over the ‘activated’ surface. Unreacted sites were then ‘capped’ by the injection of 35 ml 1 M ethanolamine, pH 8.5. Sample test kits To minimise sample preparation, SMT test kits were prepared in advance of the abattoir trial.Test vials contained 285 ml anti- SMT diluted 1 in 300 with HBS buffer. Each kit also contained SMT calibration standards, ranging from 0–1.0 mg ml21 diluted with HBS buffer, for generation of a calibration curve, a 0.4 mg ml21 SMT control, bile known to be sulfonamide free and regeneration solution (100 mM sodium hydroxide containing 10% acetonitrile). SMT positive incurred material (‘control pigs’) Ten pigs were fed SMT at a therapeutic level of 100 mg kg21 for 14 d.Five animals were sent to the abattoir 24 h after being withdrawn from the drug. The five remaining pigs were slaughtered 48 h after withdrawal. On arrival at the abattoir these ‘control pigs’ were stamped with a unique mark and identified to sampling staff. Bile, kidney and muscle samples were taken before the carcasses were removed from the kill line.Throughout the exercise, the origin of these pigs remained unknown to the biosensor operator. Sampling criteria Producers were identified by their unique number (‘slap mark’), stamped onto the shoulder of each pig. The first pig from each producer and every additional tenth animal were identified for testing using a numbered tag placed over the hind leg at sampling point A (Fig. 1). When the tagged pigs reached the evisceration section, bile samples were collected by vacutainer syringe from the gall bladder at sampling point B and forwarded to the biosensor operator.Kidney samples, from the tagged pigs, were then collected at sampling point C before the carcasses passed into the chill rooms. At all times the numbered tag was visible, permitting the identification of carcasses screened as positive by biosensor analysis. Muscle samples from these pigs were then collected. Biosensor assay configuration and conditions The biosensor method is an inhibition assay and detects the SMT antibody when it binds to the prepared sensor surface.A fixed concentration of the SMT antibody is mixed with the sample prior to injection. Any SMT present in the sample will bind to the antibody and subsequently inhibit it from then binding to the surface of the sensor. The higher the concentration of SMT in the sample, the higher the level of inhibition and hence the lower the response of the biosensor. The analysis cycle is recorded in the form of a sensogram (Fig. 2). Report points are recorded before and after each analysis cycle. The surface is then regenerated, ready for the next sample. 15 ml bile was pipetted into a kit vial containing antibody, mixed and placed in the autosampler rack. The biosensor then injected 10 ml of this, over the sensor chip surface at a flow rate of 25 ml min21. Report points were recorded before and after the sample injection. The surface was then prepared for the next Fig. 1 Floor plan of the abattoir. A = animal tagging area, B = bile sampling area, C = kidney sampling area. 1316 Analyst, 1999, 124, 1315–1318sample with a 25 ml injection of regeneration solution. At the start of each day, a calibration curve in negative bile was constructed. The 0.4 mg ml21 control was introduced at random throughout testing. Once a positive bile sample was identified a muscle sample was collected from the appropriate tagged carcass, if available, from the chill room. On return to the laboratory, the corresponding kidney and muscle samples from any animals screened positive, plus a selection of samples deemed as negative, were subjected to HPLC analysis.Results and discussion The cross-reactivity profile of the antibody (R105) was determined in the presence of bile using the biosensor method. Results are shown in Table 1. The antibody showed high crossreactivity with N4-acetylsulfamethazine (110%), the main metabolite of SMT. During the 4 d study the factory processed 3603 pigs submitted by 96 different producers.Bile was sampled from 337 (9.4%) animals representing 72 (75%) of producers who submitted pigs for slaughter. Table 2 shows the correlation between the biles screened positive and their corresponding kidney and muscle samples. A total of 12 bile samples (3.5%), which included the ten ‘control pigs’, screened positive ( > 0.4 mg ml21) for SMT.Confirmatory analysis of the kidneys showed all 12 (100%) to contain SMT residues above the limit of detection (0.05 mg g21) of the HPLC procedure. Violative concentrations ( > 0.1 mg g21) of SMT were confirmed in 11 (91.7%) of the kidneys. Confirmatory analysis of the corresponding muscle samples showed 10 to contain violative SMT levels. Results for two of the bile samples screened positive were not available before the corresponding carcasses had been removed from the factory.As a result muscle samples could not be taken from these animals. Additional HPLC analysis was performed on 28 (8.3%) kidney samples, whose corresponding biles had concentrations lower than the threshold limit, to determine the false negative rate. SMT residues were not detected in any of these samples. A 0.4 mg ml21 SMT control was introduced at random throughout testing. A mean concentration of 0.383 mg ml21 (n = 15) was achieved with an RSD of 9.5%. This figure is within the RSD range of 4.8–15.5% determined for this method under laboratory conditions.4 From the analysis of known negatives (n = 20), the limit of detection (mean + 3s) and determination (mean + 6s) of the assay were calculated as 0.03 and 0.05 mg ml21 compared with the laboratory values of 0.023 and 0.041 mg ml21. These values are well below the threshold limit of 0.4 mg ml21.The reproducibility of the calibration curve over the 4 d is shown in Fig. 3. Drug screening methods should be simple, rapid and inexpensive.These performance attributes should be accompanied by a low incidence rate of false positives and minimal false negatives. When the initial cost of the instrument is removed, the biosensor method fulfills these criteria. Sample preparation is minimal with most of the savings being made in operator and analysis time. Taking the biosensor to the abattoir also reduces the costs involved in sample collection, processing and transfer to the laboratory.One main advantage of ‘on-site’ testing is that carcasses containing violative residues can be removed from the production line immediately. During the study, the Biacore instrument was in place at the abattoir, beside the kill line. Despite the cold, humid conditions of this environment, the machine performed reliably at all times. Adverse reaction to any mechanical or electrical components Fig. 2 Typical sensogram with report points recorded before and after sample injection.Table 1 Cross-reactivity profile of R105, anti-sulfamethazine R105 anti-sulfamethazine % cross-reaction Sulfamethazine 100 N4-acetylsulfamethazine 110 Sulfadiazine < 0.01 N4-acetylsulfadiazine < 0.01 Sulfamerazine 8 Sulfathiazole < 0.01 Sulfaquinoxaline < 0.01 Sulfamethoxypyridazine < 0.01 Sulfatroxazole < 0.01 Penicillin G < 0.01 Chlortetracycline < 0.01 Streptomycin < 0.01 Dapsone < 0.01 Table 2 Biosensor screening of 337 bile samples for sulfamethazine identified 12 pigs as likely to contain violative tissue residues (correlation of the bile positives with the confirmatory analysis of corresponding kidney and muscle samples is shown) Screening test Bile (BIA)/ Confirmatory analysis/mg g21 (MRL > 0.1 mg g21) mg ml21 (Positive > 0.4 mg ml21) Kidney (HPLC) Muscle (HPLC) > 1.00 1.90 0.97 > 1.00 1.30 0.46 > 1.00 1.95 1.02 > 1.00 1.23 0.59 > 1.00 2.05 1.13 0.797 0.22 NSa > 1.00 1.11 0.51 > 1.00 0.85 0.44 > 1.00 2.02 0.75 > 1.00 0.68 0.31 > 1.00 1.30 0.72 0.470 0.06 NSa a NS = no sample.Fig. 3 Reproducibility of calibration curve, n = 4 d. Analyst, 1999, 124, 1315–1318 1317were not encountered either during this period or after the instrument was returned to the laboratory. The original aims for sampling were to collect and test approximately 10% of the total kill and at least one pig from every producer submitting pigs for slaughter.Sample coordination and testing ran smoothly and a figure of 9.4% was achieved. The attempt to test at least one pig from each producer proved more difficult to achieve and only 75% of producers were sampled. This sampling target was not fully met due to the variation in the numbers of pigs each producer submitted and to the difficulty in reading the unique ‘slap mark’ on each pig. Some producers submitted up to 200 pigs daily, while others submitted only one or two. It was generally the pigs from the latter category that caused problems due to their mixing with larger batches of pigs.A further aim of the present study was to determine if the analytical results generated could be traced to individual pig carcasses. The tagging system employed was able to cope with this demand provided results were generated within several hours of sampling. However, toward the end of each day, the generation of results fell further behind the speed of the kill line.At the end of each day, the backlog of samples was assayed overnight. The carcasses corresponding to these samples had often been transported to the boning halls and the tags removed. Another factor was that some carcasses went immediately into chilled container lorries, for transport to outside processors. Tags from these animals were removed on entering the lorry and as soon as the container was full, it left the abattoir. From these experiences, we can conclude that the biosensor analysis must become faster in order to keep up with the speed of the kill line.In terms of assay performance, all ‘positive control pigs’, whose origin remained unknown to the biosensor operator, were correctly identified during testing. In addition, two ‘routine’ animals (0.6% of total animals tested) screened positive, with one (0.3%) of the corresponding kidney samples confirmed as containing violative levels of SMT. Based on this limited population, a false positive rate of 0.3% is obtained. This is slightly higher than the 0.14% reported for over 2000 samples tested under laboratory conditions by Crooks et al.7 One possible explanation for the occurrence of false positives is the high cross-reactivity of the antibody with the metabolite, N4- acetyl-SMT, which may be present in pig bile in high concentrations.The HPLC confirmatory method only detects the parent sulfonamide. More importantly, no false negatives were generated.In this first attempt at technology transfer from the laboratory to the abattoir, many important lessons have been learned. The biosensor offers great potential for this area of drug screening and proved itself to be robust enough to withstand the harsh conditions under which it was being operated. Sample traceability could be achieved, however, the speed of analysis needs to be increased to keep up with the kill lines and prevent screened positives leaving the abattoir. To achieve this the development of a biosensor, specifically for high throughput drug screening is necessary.12 Acknowledgements Sincere thanks are extended to the staff of Stevenson & Co.Ltd. for their help and co-operation during the project and to Sanj Kumar, Biacore AB, for the loan of the BIACORE 1000 instrument. We also acknowledge Mark Druce, Field Service Engineer, Biacore AB, for his help with the installation of the biosensor. Thanks are also extended to the Drug Residues Laboratory Staff for their assistance. References 1 P. W. Saschenbrecher and N. A. Fish, Can. J. Compar. Med., 1980, 44, 338. 2 W. J. McCaughey, J. D. G. McEvoy and B. McCartan, The Pig J., 1997, 39, 105. 3 Anabolic, Anthelmintic and Antimicrobial Agents, Food surveillance Paper No. 22, HM Stationery Office, London, 1987, pp. 19–20. 4 V. W. Randecker, J. A. Reagen, R. E. Engel, D. L. Soderberg and J. E. McNeal, J. Food Prot., 1987, 50, 115. 5 W. Haasnoot, G. O. Korsrud, G. Cazemier, F. Maneval, H. Keukens and J. Nouws, Food Addit. Contam., 1996, 13, 811. 6 T. L. Fodey, S. R. H. Crooks, C. T. Elliott and W. J. McCaughey, Analyst, 1997, 122, 165. 7 S. R. H. Crooks, G. A. Baxter, M. C. O’Connor and C. T. Elliott, Analyst, 1998, 123, 2755. 8 W. J. McCaughey, C. T. Elliott, J. N. Campbell, W. J. Blanchflower and D. A. Rice, Ir. Vet. J., 1990, 43, 127. 9 J. R. Fleeker and L. J. Lovett, J. Assoc. Off. Anal. Chem.,1985, 88, 172. 10 W. J. McCaughey, C. T. Elliott and S. R. H. Crooks, Food Addit. Contam., 1990, 7, 259. 11 A. Sternesjo, C. Mellgren and L. Bjorck, Anal. Biochem., 1995, 226, 175. 12 http://www.slv.se/foodsense Paper 9/04122B 1318 Analyst, 1999, 124, 1315–1318

ISSN:0003-2654    

DOI:10.1039/a904122b

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Porphyrins as carriers in poly(vinyl chloride)-based membrane potentiometric sensors for histamine Mohammad K. Amini,* Said Shahrokhian and Shahram Tangestaninejad Chemistry Department, University of Esfahan, Esfahan, 81744, Iran Received 4th May 1999, Accepted 8th July 1999 Iron(iii) and manganese(iii) tetraphenylporphyrins were explored as electroactive materials for preparing polymeric membrane-based sensors selective to histamine. The membranes, incorporating 30% poly(vinyl chloride) (PVC), 60% dioctyl phthalate (DOP), and 10% iron or manganese tetraphenylporphyrin, coated on the surface of graphite electrodes, exhibited near-Nernstian responses in the concentration range from 1 3 1026 to 1 3 1021 mol dm23 with detection limits of about 5 3 1027 mol dm23 histamine, and response times of < 30 s.The electrodes show high selectivity towards histamine over several amino acids and common inorganic anions and can be used in a wide pH range. They were applied to the determination of histamine in synthetic serum samples.Introduction One of the most important recognition elements that can be utilised in the development of potentiometric sensors involves specific metal–ligand interactions.1 In this regard, metalloporphyrins have been demonstrated to be highly selective carriers for the preparation of potentiometric sensors.2–6 Owing to their ligand discriminating ability, these compounds appear to be one of the most promising classes of compound to be used as the active material in ion-selective electrodes (ISEs).The methods described in the literature for the measurement of histamine are mainly chromatographic methods. They include derivatization with fluorescent reagents followed by chromatographic separation,7,8 high-performance capillary electrophoresis with fluorescence detection,9 gas chromatography, 10 thin-layer chromatography and high-performance liquid chromatography.11,12 Other alternatives include, enzyme isotope assay,13 and carbon fibre-based electrochemical techniques. 14 Potentiometric detection based on ion-selective sensors, as a simple method, offers great advantages such as speed and ease of preparation and procedures, wide dynamic range, and low cost. These characteristics have inevitably led to sensors for several ionic species, and the list of available electrodes has grown substantially over the last few years. A very interesting development of potentiometric sensors is in the construction of electrodes that respond selectively to biological compounds.Imidazole and its derivatives, such as histamine and histidine, are known to coordinate to the metal centre of porphyrins.15–18 Therefore, we were prompted to investigate the potentiometric behaviour of iron and manganese porphyrins towards these compounds. Potentiometric sensors prepared by coating polymeric films containing electroactive species on metallic or graphite conductors, 19,20 of any convenient geometric size and shape (i.e.wire, disc, cylinder, thin film, etc.), have been shown to be very effective for a wide variety of inorganic and organic cations and anions.21,22 Although the relevant processes that occur at membrane–solid contact interfaces are still not fully understood, electrodes of this type have become popular because they are very simple, durable, inexpensive, and capable of reliable response in a wide concentration range.1,23 These types of electrode have a much higher mechanical resistance, and as such they have been used as potentiometric detectors in capillary electrophoresis.23,24 In this work, we examine the utility of plasticized poly(vinyl chloride) (PVC)-based membranes of 5,10,15,20-tetraphenyl( porphyrinato)iron(iii) chloride [Fe(TPP)Cl], and 5,10,15,20-tetraphenyl(porphyrinato)manganese(iii) chloride [Mn(TPP)Cl], coated on the surface of graphite electrodes, for the detection of histamine, a chemical messenger important in the nervous and immune systems.The importance of histamine in this role makes it desirable to develop chemical probes for its monitoring.14 Histamine is well known as a neurohormone that triggers a variety of allergic reactions.25 This biogenic amine is responsible for a wide variety of physiological and pathological effects, and has been extensively studied as one of the fundamental chemical mediators of neural, secretory and musculotropic action.26 It exists widely in many foods and biological materials,27,28 and its level has been shown to be a good index of food decomposition.9 Experimental Reagents Fe(TPP)Cl, Mn(TPP)Cl, dioctyl phthalate (DOP) and PVC of high relative molecular mass were used as received from Aldrich.Tetrahydrofuran (THF), reagent grade from Merck, was distilled prior to use. All other chemicals were of analyticalreagent grade from Merck, and were used without further purification.All solutions were prepared with de-ionised, distilled water. The pH adjustments were made with 0.05 mol dm23 phosphate buffer, pH 7.0. Electrode preparation Membrane solutions were prepared by dissolving Mn(TPP)Cl or Fe(TPP)Cl together with DOP and PVC, to give a total mass of 200 mg, in about 5 ml of THF. Graphite electrodes (3 mm diameter and 10 mm long) were prepared from spectroscopicgrade graphite. A shielded copper wire was glued to one end of the graphite rod with epoxy resin, and the electrode was sealed into the end of a PVC tube of about the same diameter by epoxy resin.The working surface of the electrode was polished with fine alumina slurries on a polishing cloth, sonicated in distilled water and dried in air. The polished electrode was dipped into the membrane solution, and the solvent was evaporated. A Analyst, 1999, 124, 1319–1322 1319membrane was formed on the graphite surface, and it was allowed to set overnight. For a comparative study, a membrane containing no active component was also prepared.Potential measurement The electrodes were rinsed with water and stored for at least 24 h in an aqueous solution of histamine (0.05 mol dm23) for conditioning. The coated graphite electrodes containing Fe(TPP)Cl or Mn(TPP)Cl were used as the measuring electrodes in conjunction with a saturated calomel electrode (SCE). The potential measurements were carried out at 25.0 ± 0.5 °C on a digital pH/millivoltmeter (Jenway, Model 3305) by setting up the following cell assembly: Hg, Hg2Cl2, KCl (satd.) · sample solution | membrane | graphite electrode The pH of the sample solutions was monitored simultaneously with a conventional glass pH electrode.Calibration graphs were constructed by plotting the potential, E versus the logarithm of the concentration of histamine at constant pH. Results and discussion The Fe(TPP)Cl- and Mn(TPP)Cl-based membranes coated on the surface of graphite electrodes generated stable potential responses in aqueous solutions containing histamine after conditioning for about 24 h in a 0.05 mol dm23 histamine solution.Table 1 gives the data obtained with membranes having various ratios of different constituents. The potential responses of all the membrane sensors were studied in a wide range of concentrations of histamine solution and the results are illustrated in Fig. 1. The membrane without the active material (membrane A) exhibits insignificant selectivity towards histamine.The electrodes containing between 5 and 10 wt.-% of Fe(TPP)Cl or Mn(TPP)Cl ionophores exhibited linear responses in the range from 1 3 1026 to 1 3 1021 mol dm23, and can therefore be used for the determination of histamine in this concentration range. As with many carrier-modified membrane electrodes, the total potentiometric response of the electrode towards histamine is dependent on the concentration of the metalloporphyrin incorporated within the membrane.As can be seen in Fig. 1 and Table 1, increasing levels of porphyrins result in membranes that display larger slopes and lower detection limits. Using 10 wt.-% of Mn(TPP)Cl or Fe(TPP)Cl in the membrane yielded electrodes with near-Nernstian slopes towards histamine. Therefore, membranes C and E were chosen for subsequent work. A maximum slope of +56 mV per decade of histamine concentration was observed for electrode C with 10 wt.-% of Mn(TPP)Cl. The optimum responses of the electrodes were tested after conditioning for different periods of time in 0.05 mol dm23 histamine.The slopes obtained using 24 h of conditioning were closer to the theoretical slopes expected on the basis of the Nernst equation. Longer conditioning times produced no further improvements in response. The optimum conditioning solution was determined to have a concentration of about 0.05 mol dm23. The pH dependence of the membrane electrodes was tested over the pH range 2–10 at a 5 3 1023 mol dm23 histamine concentration, and a typical plot for electrode C is shown in Fig. 2. The potentials are fairly constant in the pH range 4–9. Beyond this range, a gradual drift was observed. The observed drift at lower pH values is due to protonation of the imidazole site of histamine and its release from the membrane. The behaviour of the electrode at high pH can be explained in terms of the increased interference from hydroxide ions.29 Values of the detection limit, defined as the concentration of histamine obtained when extrapolating the linear region of the calibration graph to the baseline potential, for different electrodes are given in Table 1.As can be seen, electrodes C and E show detection limits of about 7 3 1027 and 5 3 1027 mol dm23, respectively. The detection limits obtained with these electrodes are much lower than the usual range of dose of histamine injected as a diagnostic agent for testing several functions such as the capacity of gastric glands and pheochromocytoma test.30 The proposed sensors may also be applied to the analysis of foods as an index of food decomposition. The detection limits obtained with the potentiometric sensors are comparable to that obtained in a recent paper on the determination of histamine in foods by high-performance capillary electrophoresis.9 For analytical applications, the response time of a sensor is an important factor.The response time of the electrodes, tested by measuring the time required to achieve a steady potential (within ±1 mV), was < 30 s and was sustained for at least 10 min over the entire concentration range.The detection system is very stable, and could be used over a period of 1 month without any significant change in response characteristics being observed. The potential readings were highly reproducible within ±0.5 mV at several histamine concentrations. Potentiometric selectivity coefficients (Khis,j), describing the preference by the membrane for an interfering ion j relative to histamine, were determined by the separate solution method31 from potential measurements on solutions containing 1 3 1023 mol dm23 histamine and the interfering species.The potentiometric selectivity coefficients of electrodes C and E, based on Mn(TPP)Cl and Fe(TPP)Cl carriers, are summarised in Table 2. The selectivity coefficients obtained at lower concentrations of histamine and higher concentrations of the interfering species were in good agreement with these values.The selectivity coefficient patterns clearly indicate that both electrodes are selective to histamine over several amino acids and common anions. A typical selectivity pattern presented by electrode C towards several amino acids, imidazole and histamine is as follows: histamine 9 imidazole > histidine > tryptophan > lysine > tyrosine > serine > aspartic acid > cysteine > arginine > glycine > phenylalanine > alanine > methionine.This electrode responded in the following order of preference to a number of anions: histamine > SCN2 > N32 > ClO42 > I2 > CH3COO2 > salicylate > NO22 > Br2 > Cl2 > NO32 > F2 > HCO32 > H2PO42 > C2O4 22 > SO4 22. Most of these Table 1 Composition of membranes and their potentiometric response properties in histamine-selective electrodes Composition of electrode (%) Detection Slope/ Linear range/ limit/ Electrode PVC DOP Ionophore mV per decade mol dm23 mol dm23 A 34.5 65.5 — — — — B 32.9 61.5 5.6a 52 1 3 1026–5 3 1022 1 3 1026 C 30.0 60.0 10a 56 1 3 1026–1 3 1021 7 3 1027 D 32.8 62.0 5.2b 46 1 3 1026–5 3 1022 1 3 1026 E 30.0 60.0 10b 52 1 3 1026–1 3 1021 5 3 1027 a Mn(TPP)Cl carrier.b Fe(TPP)Cl carrier. 1320 Analyst, 1999, 124, 1319–1322ions would be expected to interfere seriously with classical ionexchanger type membrane sensors.The potentiometric response of metalloporphyrin-doped membrane electrodes is believed to be based on the coordination of the analyte anion as an axial ligand to the metal centre of the carrier molecule. Changing the central metal of these complexes strongly influences the potentiometric selectivities.4 The electrode based on Fe(TPP)Cl (electrode E) shows a different selectivity order for amino acids and anions. This electrode is more selective to histamine, in comparison with that based on Mn(TPP)Cl (electrode C), with respect to N32, CH3COO2, salicylate, NO22, Br2, Cl2, NO32, F2, HCO32, C2O4 22, and SO4 22.However, the selectivity of the Fe(TPP)Clbased electrode with respect to the most lipophilic anions such as perchlorate, thiocyanate and iodide, is worse than that of the Mn(TPP)Cl system. The selective bonding of histamine to these complexes is thought to be the origin of the high selectivity of these electrodes for histamine over other compounds.It is well known that aromatic amines coordinate more strongly than aliphatic amines to metalloporphyrins.16 The extra stabilisation of the metal–nitrogen bond may arise from involvement of the p* orbital of the aromatic amines in some degree of p-back-bonding. It has been observed that, in the absence of steric interactions, imidazoles generally form more stable complexes with metalloporphyrins than do pyridines of the same s donor strength.32,33 The stronger interactions of imidazoles with metalloporphyrins, as compared with those of aliphatic amines and pyridines, have been related to the more favourable bond angle of the five-membered imidazole ring, which acts to remove much of the steric interaction with the hydrogens of the adjacent carbons.32 The imidazole selectivity sequence of the present sensors at the working pH of 7.0, histamine 9 imidazole > histidine (Table 2), may be the result of several factors such as protonation of the imidazole ring, charge carried by the ligand, and steric effects of the imidazole side chain.16 The charge of the ligand is a determining factor in coordination-based potentiometric sensors. Many coordination-based potentiometric sensors have been developed so far; however, they do not respond to neutral ligands in principle.Therefore, imidazole, histamine or histidine can be detected by such sensors only when the molecules are charged, although imidazole and histidine exist predominantly as neutral molecules at neutral pH.On the other hand, the protonated imidazole ring cannot coordinate to the metal centre, because no lone pair of electrons is available for the coordination.15 Since the pKa of imidazole is 7.0,34 about 50% of its molecules are protonated at neutral pH, and, therefore, cannot coordinate to the metal centre in the porphyrin ring. Although the unprotonated imidazole can coordinate to the metal centre, it does not contribute to the potential response of the electrode.For histidine, the pKa values for the amino, carboxyl, and imidazole groups are 9.2, 1.8 and 6.0, respectively.34 Therefore, histidine, existing mainly (about 90%) as a zwitterion at neutral pH, also does not contribute to the potential response of the electrode system. The bulky side chain of histidine also interferes with the coordination of this compound. At neutral pH, histamine exists as a mixture of base, monocation and dication in a ratio of about 1+96+3.34 Therefore, about 96% of histamine can contribute to the potential response of the electrodes under neutral conditions.To Fig. 1 Potentiometric response of the coated graphite electrodes to histamine. (A) Without the active material, (B) 5.6% Mn(TPP)Cl, (C) 10% Mn(TPP)Cl, (D) 5.2% Fe(TPP)Cl, (E) 10% Fe(TPP)Cl, all at pH 7. Fig. 2 pH response of the Mn(TPP)Cl-based membrane electrode. Table 2 Selectivity coefficients of the Fe(TPP)Cl- and Mn(TPP)Cl-based membranes (electrodes C and E) using the separate solution method with a histamine concentration of 131023 mol dm23 Ki,j pot Ki,j pot Interferent, j C E Interferent, j C E d,l-Alanine 2.10 3 1024 1.19 3 1023 Azide 4.97 3 10–1 9.57 3 1023 l-Arginine 2.57 3 1024 6.71 3 1024 Bromide 1.57 3 1022 7.33 3 1023 l-Aspartic acid 4.58 3 1024 3.78 3 1024 Carbonate 1.44 3 1023 1.05 3 1023 l-Cysteine 2.68 3 1024 3.31 3 1024 Chloride 7.50 3 1023 1.19 3 1023 Glycine 2.47 3 1024 1.05 3 1023 Fluoride 1.57 3 1023 6.14 3 1024 l-Histidine 4.39 3 1023 3.31 3 1023 Iodide 5.40 3 1022 2.42 3 10–1 Imidazole 5.62 3 1023 4.92 3 1023 Nitrate 2.37 3 1023 7.33 3 1024 l-Lysine 7.81 3 1024 3.61 3 1024 Nitrite 1.78 3 1022 2.77 3 1023 d,l-Methionine 1.78 3 1024 7.02 3 1024 Oxalate 1.18 3 1024 5.62 3 1025 l-Phenylalanine 2.28 3 1024 7.67 3 1024 Perchlorate 8.14 3 1022 4.92 3 10–1 l-Serine 5.18 3 1024 2.77 3 1024 Phosphate 4.97 3 1024 5.62 3 1024 Tryptophan 1.93 3 1023 2.89 3 1024 Salicylate 1.85 3 1022 1.14 3 1022 l-Tyrosine 6.91 3 1024 4.92 3 1024 Sulfate 8.14 3 1025 2.65 3 1025 Acetate 2.47 3 1023 1.36 3 1023 Thiocyanate 6.91 3 10–1 1.25 Analyst, 1999, 124, 1319–1322 1321summarise, in comparison with histidine and imidazole, histamine has a much larger contribution to the potential response of the electrode regarding the charge and protonation at pH 7.0.In aqueous solution at neutral pH, the Fe(iii) and Mn(iii) tetraphenylporphyrins in the membranes are expected to be sixcoordinate of the type M(TPP)Cl·H2O,29 where M stands for the metal centre. A simplified reaction mechanism for the potentiometric response of the proposed sensors can therefore be suggested as the reaction of the metal centres with protonated histamine (Im-NH3 +) to form M(TPP)Cl·Im-NH3 +.The departing ligand is assumed to be a water molecule. This results in a positive change in the potential of the electrodes, as indicated by the positive slopes of the calibration graphs.Although the above discussion is qualitative, it could be of significance for the design of this type of sensor. It is of particular significance that both electrodes displayed high selectivity over several amino acids and particularly histidine, making possible the use of the electrodes for the determination of histamine in biological samples. In this regard, experiments were performed to determine the feasibility of using electrodes C and E to measure histamine in a synthetic serum sample.The composition of the synthetic serum is listed in Table 3. The concentration of each component was chosen to match its normal level in human serum.35 Recovery studies were conducted with the sample containing various amounts of histamine. The concentration of histamine was chosen to be lower than the experimental LD50 in laboratory animals (630 mg kg21, about 6 3 1023 mol dm23).34 The results of recovery studies are summarised in Table 4.A good recovery was observed, indicating that the constituents of the synthetic serum sample do not interfere significantly with the detection of histamine. The proposed electrode seems to provide an alternative device for the direct determination of histamine in biological samples. References 1 R. S. Hutchins and L. G. Bachas, Anal. Chem., 1995, 67, 1654. 2 D. Gao, J. Gu, R. Q. Yu and G. D. Zheng, Anal. Chim. Acta, 1995, 302, 263. 3 A. Jyo, R. Minakami, Y. Kanda and H. Egawa, Sens. Actuators B, 1993, 13–14, 200. 4 N. A. Chaniotakis, S. B. Park and M. E. Meyerhoff, Anal. Chem., 1989, 61, 566. 5 H. Abe and E. Kokufuta, Bull. Chem. Soc. Jpn., 1990, 63, 1360. 6 D. Ammann, M. Huser, B. Kräutler, B. Rusterholz, P. Schulthess, B. Lindemann, E. Halder and W. Simon, Helv. Chim. Acta, 1986, 69, 849. 7 C. M. J. J. van Haaster, W. Engels, P. J. M. R. Lemmens, G. Hornstra and G. J. van der Vusse, J. Chromatogr., 1993, 617, 233. 8 D. Egger, G. Reisbach and L. Hultner, J. Chromatogr. B, 1994, 662, 103. 9 S. Oguri, S. Watanabe and S. Abe, J. Chromatogr. A, 1997, 790, 177. 10 H. Mita, H. Yasueda and T. Shida, J. Chromatogr, 1980, 221, 1. 11 N. Seiler, Methods Biochem. Anal., 1970, 18, 259. 12 T. B. Jensen and P. D. Marley, J. Chromatogr. B, 1995, 670, 199. 13 J. R. Vane, Br. J. Pharmacol., 1964, 20, 340. 14 K. Pihel, S. Hsieh, J. W. Jorgenson and R. M. Wightman, Anal. Chem., 1995, 67, 4514. 15 T. Tatsuma and D.A. Buttry, Anal. Chem., 1997, 69, 887. 16 T. Tatsuma and T. Watanabe, Anal. Chem., 1992, 64, 143. 17 M. Rougee and D. Brault, Biochemistry, 1975, 14, 4100. 18 D. K. White, J. B. Cannon and T. G. Taylor, J. Am. Chem. Soc., 1979, 101, 2443. 19 H. Freiser, J. Chem. Soc. Faraday Trans. 1, 1986, 82, 127. 20 Y. K. Lee, J. T. Park, C. K. Kim and K. J. Whang, Anal. Chem., 1986, 58, 2101. 21 P. Schnierle, T. Kappes and P. C. Hauser, Anal. Chem., 1998, 70, 3585. 22 P. B. Bühlmann, S.Yajima, K. Tohda, K. Umezawa, S. Nishizawa and Y. Umezawa, Electroanalysis, 1995, 7, 811. 23 Y. Yang, Y. Bi, M. Liu, J. Fu and Z. Xi, Microchem. J., 1997, 55, 348. 24 B. L. De Backer and L. J. Nagels, Anal. Chem., 1996, 68, 4441. 25 L. M. Lichtenstein, Sci. Am., 1993, 269, 117. 26 M. A. Beaven, N. Engl. J. Med., 1976, 294, 30. 27 E. Werle and A. Raub, Biochem. Z., 1948, 318, 538. 28 J. L. Mietz and E. Karmas, J. Food Sci., 1977, 42, 155. 29 N. A. Chaniotakis, A. M. Chasser, M. E. Meyerhoff and J. T. Groves, Anal. Chem., 1988, 60, 185. 30 E. A. Swinyard, in Remington’s Pharmaceutical Sciences, ed. E. R. Gennaro, Mack Publishing, Pensylvania, 1990, pp. 1123–1124. 31 IUPAC Recommendations for Nomenclature of Ion-Selective Electrodes, Pure Appl. Chem., 1994, 66, 2527; IUPAC Selectivity Coefficients for Ion-Selective Electrodes: Recommended Methods for Reporting KAB Values, Pure Appl. Chem., 1995, 67, 507. 32 F. A. Walker, J. Am. Chem. Soc., 1973, 95, 1150. 33 H. M. Marques, O. Q. Munro and M. L. Crawcour, Inorg. Chim. Acta, 1992, 196, 221. 34 Dictionary of Organic Compounds, Chapman and Hall, New York, 6th ed., 1996, vol 4. 35 C. P. Pau and G. A. Rechnitz, Anal. Chim. Acta, 1984, 160, 141. Paper 9/03500A Table 3 Composition of synthetic serum Compound mol dm23 Compound mol dm23 d,l-Alanine 4.1 3 1024 d,l-Methionine 3.4 3 1025 l-Arginine 2.1 3 1024 l-Phenylalanine 1.6 3 1024 l-Aspartic acid 8.8 3 1024 l-Serine 1.2 3 1024 l-Cysteine 5.1 3 1025 l-Tyrosine 8.1 3 1025 Glycine 1.4 3 1024 d,l-Tryptophan 6.9 3 1025 l-Histidine 1.1 3 1024 NaHCO3 7.9 3 1023 l-Lysine 2.0 3 1024 NaCl 8.7 3 1022 Table 4 Recovery of histamine added to synthetic serum samples Electrode C Electrode E Amount added/1024 mol dm23 Amount founda/1024 mol dm23 Recovery (%) Amount founda/1024 mol dm23 Recovery (%) 4.62 4.44 ± 0.12 96.1 4.68 ± 0.09 101.3 11.1 10.9 ± 0.3 98.2 11.2 ± 0.4 101 17.4 16.7 ± 0.5 96.0 17.0 ± 0.4 96 a Average of three determinations ± S. 1322 Analyst, 1999, 124, 1319–1322

ISSN:0003-2654    

DOI:10.1039/a903500a

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Minium; FT-Raman non-destructive analysis applied to an historical controversy Howell G. M. Edwards,*a Dennis W. Farwell,a Emma M. Newtona and Fernando Rull Perezb a Chemical and Forensic Sciences, University of Bradford, Bradford, UK BD7 1DP b Cristalografia y Mineralogia, Facultad de Ciencias, Universidad de Valladolid, Prado de la Magdalena S/N, 47011 Valladolid, Spain Received 21st May 1999, Accepted 13th July 1999 The term minium was applied to both cinnabar (HgS) and red lead (Pb3O4) pigments in antiquity; in Roman times, minium was reserved for mercury(ii) sulfide but was applied increasingly to lead tetraoxide by the Renaissance.Confusion in the interpretation of ancient recipes for pigment mixtures is inevitable and is compounded by the practice of adulteration of mercury(ii) sulfide with lead tetraoxide for economic or artistic reasons. In this study, the Raman spectra of mixtures of cinnabar and red lead were recorded and used to determine the composition of several pigment mixtures from mediaeval artwork.The potential of the method for the non-invasive quantitative analysis of pigment mixtures involving red lead and cinnabar is thereby demonstrated. 1. Introduction The analysis of pigments used in artwork can provide four pieces of information which can assist in the scientific attribution of paintings:1 (i) the use made by the artist of pigments and pigment mixtures including the ‘layering’ of different pigment compositions to produce specific colour effects or tones; (ii) the restoration of damaged paintings and the matching of original pigments with modern versions; this includes the assessment of pigment changes with time; (iii) the conservation of artwork demands a knowledge of the effects of heat, humidity, UV radiation and atmospheric pollution on pigments; (iv) the authentication of a work of art and the determination of the extent of later restoration programmes depend critically on a knowledge of the pigments used in antiquity and their dates of introduction and of common usage.In most cases, the pigments used in antiquity are referred to by authors and historians in contemporary manuscripts or texts, and nomenclature can change with time; sometimes there is confusion where similar names are used to describe more than one pigment. A classic example of this is provided by minium, a naturally occurring mineral which is of importance since it was used by Roman artists over 2000 years ago and adopted extensively through mediaeval and Renaissance periods by many cultures.2 There is much historical confusion and controversy over the use of the term minium in accounts of artists’ pigments. Pliny the Elder3 in the first century ad applied minium to cinnabar, mercury(ii) sulfide (HgS), a naturally occurring form of vermilion, whereas minium secondarium was reserved for red lead, lead tetraoxide (Pb3O4).Theophrastus identified cinnabar with minium, but also applied this term to red lead; true minium was often then called cenobrium.Further complication arises with the use of the term vermilion, which originally was applied to synthetic mercury(ii) sulfide, interchangeably with cinnabar. Because of stringent Roman controls on cinnabar mining and its expense (ca. 40–70 sesterces per pound in the first century),4 adulteration of cinnabar with red lead was common, and the name minium was also adopted for this mixture.However, by the Middle Ages and Renaissance periods, minium was applied to red lead alone.5 Later translation of the pigment name gave a risk of even further confusion as the terms Saturnine red, red lead oxide, orange lead and red ore of lead were all used diversely to describe lead tetraoxide; loose translations provided further problems, e.g., red ore of lead is not lead tetraoxide at all and is now known to be crocoite (PbCrO4), lead chromate! Orange lead has alternatively been described as rouge vermilion to add to the confusion.In the Far East, Chinese manuscripts refer to the manufacture of red lead and to the use of lead cinnabar during the Han dynasty (200 bc–200 ad).6 Normally, the identification of minium as either cinnabar or red lead is not difficult chemically,7 but problems can arise under difficult sampling conditions or when adulteration has occurred: Cennine8 mentions the use of red lead over vermilion to achieve a special decorative effect, and a painting by Renoir in 1876 contains vermilion adulterated with red lead.9 In this study, the non-destructive characterisation of minium, red lead and cinnabar was undertaken using Raman microscopy and details are provided for the quantitative evaluation of adulterated cinnabar, i.e., minium–minium secondarium. To test the application of the method, the composition of adulterated cinnabars in mediaeval polychrome paintings is determined. 2. Experimental 2.1 Samples Specimens of cinnabar [mercury(ii) sulfide], red lead, litharge [lead(ii) oxide] and lead(ii) sulfide were obtained from Aldrich Chemical (Gillingham, Dorset, UK) and were used without further purification.Ancient samples of cinnabar and red lead pigments were obtained from the following sources: (i) Convento de la Peregrina, Sahagun, Spain; a 13th Century chapel on the Camino de Santiago, an important pilgrim’s route from the Pyrenees to Santiago de Compostela in Galicia.The samples were taken from important frescoes executed in an Islamic style.10 (ii) Church of SS Cosmo and Damien, Basconcillos del Tozo, Spain; the discovery of important wall paintings from the 14th Century which had been hidden by a reredos was made only 3 years ago.11 Analyst, 1999, 124, 1323–1326 1323The sites of these mediaeval samples are available in the form of Supplementary Information.† It is important to emphasize that both churches had been closed for many years and this featured strongly in the unrestored nature of the wall paintings and frescoes. 2.2 Raman spectroscopy Raman spectra were excited using 1064 nm radiation from an Nd3++YAG laser and a Bruker IFS66 instrument with an FRA 106 Raman module attachment (Bruker, Karlsruhe, Germany). The sample ‘footprint’ at the focus of the laser beam was about 100 mm in diameter. Spectra were recorded using laser powers of about 40 mW or lower to avoid sample decomposition and about 1000 scans were effected at 8 cm21 spectral resolution to enhance the signal-to-noise ratios.The spectral response was calibrated against a white light source and wavenumbers are correct to ±1 cm21 or better for sharp, strong features. 2.3 Cinnabar–red lead mixtures A series of specially prepared pigment mixtures containing cinnabar and red lead in the molar ratios of 0, 10, 20, … 100% were examined to provide the basis for quantification of the composition. From these mixtures, it was possible to deduce the following parameters: (i) the limit of Raman spectroscopic detection of adulteration of cinnabar by red lead; and (ii) the precision of quantitative determination of ancient pigment samples of unknown composition.Because of the very different relative molecular masses of red lead (685.6) and cinnabar (232.7), an equimolar mixture (50+50) actually corresponds closely to a 3+1 w/w mixture. From the calibration plots of relative band intensities of the pigment mixtures against species concentration it becomes possible for the first time using Raman spectroscopy to evaluate the composition of red lead and cinnabar mixtures in ancient wall paintings.Mixtures of known composition of cinnabar and red lead separately with potassium bromide were also made and analysed using Raman spectroscopy to determine the relative molar scattering factors of cinnabar and red lead. 3. Results and discussion The Raman spectra of cinnabar and red lead are shown in Fig. 1 and the vibrational band wavenumbers are given in Table 1. It can be seen that the strong bands at 254 and 547 cm21 in cinnabar and red lead, respectively, representing n(HgS) and n(PbO) vibrational modes, are good candidates for the quantitative determination of composition in the pigment mixtures. In reality, however, the situation is complicated by two factors: (i) the band at 228 cm21 in the Raman spectrum of red lead causes some interference to the measurement of band areas of the n(HgS) mode at 254 cm21, particularly in pigment mixtures which have a relatively small content of cinnabar; (ii) the more intense Raman scattering of the Hg–S bonds compared with the Pb–O bonds, bestows a significantly larger molar scattering coefficient on cinnabar.The significance of this can be appreciated by examination of the 50+50 mixture (1+1 molar ratio) of red lead and cinnabar, where the selected band area ratios of HgS to Pb3O4 are about 60+1 for I 254 HgS+I 547 Pb3O4 (designated Irel here).A Raman spectral stack plot is shown in Fig. 2 for 10–90 mol% pigment mixtures of red lead and cinnabar in 10% steps over the wavenumber range 50–650 cm21. All spectra were normalised to the I 254 HgS intensity in the 50+50 sample. The increase in band intensity of the 547 cm21 band due to n(PbO) stretching in lead tetraoxide with increase in the red lead component in the mixtures is evident. Also, the interference of the weaker feature due to d(PbO2) at 228 cm21 on the band intensity of the n(HgS) mode in cinnabar at 254 cm21 is seen especially for mixture compositions which are low in cinnabar content.For this reason, deconvolution of the band areas in the wavenumber range 190–310 cm21 was undertaken to extract the I 254 HgS parameter for each mixture. A plot of the relative band intensity (area) of the 547 and 254 cm21 bands representing the red lead in the mixtures as a function of the mixture composition is a smooth curve [Fig. 3(a)]; a similar plot for the relative band intensity (area) at 254 and 547 cm21 representing the cinnabar in the mixtures as a function of the composition [Fig. 3(b)] is also a curve, but the Irel values are larger. † Available as Electronic Supplementary Material; see http://www.rsc.org/ suppdata/an/1999/1323. Fig. 1 FT-Raman spectra of (a) red lead and (b) cinnabar. Excitation at 1064 nm, 1000 scans at 8 cm21 resolution; wavenumber range, 50–650 cm21.Table 1 Vibrational wavenumbers in the Raman spectra of cinnabar and red lead Compound n/cm21 Cinnabar (HgS) 85 vw, 105 vw, 254 vs,a 284 w, 344 m Red lead (Pb3O4) 121 vs, 150 mw, 228 w, 312 mw, 389 m, 455 vw, 547 msa a Bands selected for quantitative studies in mixtures of red lead and cinnabar. Fig. 2 Stack plot of FT-Raman spectra of mixtures of red lead and cinnabar in the composition range 90–10 mol% Pb3O4 (from top) in steps of 10 mol%.The spectra have been normalised to the 254 cm21 band intensity of the 50+50 (50 mol%) composition. Bands used in quantitative Raman spectroscopic determination are 547 cm21 for Pb3O4 and 254 cm21 for HgS. Conditions as in Fig. 1. 1324 Analyst, 1999, 124, 1323–1326These calibration curves are useful for the quantitative, nondestructive analysis of red lead–cinnabar mixtures in samples of ancient pigments. This application will be illustrated here for 12 mediaeval samples taken from several sites in Spain.The relative molar scattering coefficient, Jrel = 62, for cinnabar and red lead can be evaluated from the relative band intensities of the 254 and 547 cm21 bands in the spectrum of the equimolar mixture (50+50). The effect of the significantly larger molar scattering coefficient of cinnabar over that of red lead is clearly seen in Fig. 4, which shows the Raman spectrum of an equimolar mixture of red lead and cinnabar; the band at 547 cm21 in Pb3O4 is significantly weaker in intensity than that at 254 cm21 in HgS; the contribution of the weaker band at 228 cm21 to the band area of the 254 cm21 feature is not evident in this spectrum.We deduce from measurements of the band intensities of the two-component mixtures that the detection limit of red lead under our experimental conditions is ~ 1 mol% and that of cinnabar ~ 0.1 mol%, representing about 30 and 0.5 mg, respectively, per gram of mixture. Mediaeval paint samples Twelve samples of red paint pigments were analysed quantitatively; the samples had been collected as part of a larger, comprehensive examination of ancient wall paintings in a variety of Spanish sites.The mediaeval wall paintings in the Convento de la Peregrina, Sahagun (site 1) and SS Cosmo y Damien at Basconcillos del Tozo (site 2) are very important historically and archaeologically since they occur in churches which were closed due to pestilence in late mediaeval times and which have only been discovered in recent years.The historical record, therefore, confirms that restoration has not been undertaken previously; this is an important factor in the analysis of wall paintings as unrecorded application of cleaning agents or liberal re-touching could affect the contemporary interpretation. Some selected sample Raman spectra from the mediaeval wall paintings at the Convento de la Peregrina, Sahagun, are shown stack plotted in Fig. 5, where the large diversity of composition of the cinnabar and red lead mixtures employed in mediaeval times at these sites can be seen.Using Irel measurements similar to those undertaken for the construction of the calibration plots described above, it is now possible to determine quantitatively for the first time the cinnabar and red lead composition in each of these samples. Details of individual sample measurements are provided in Table 2, from which the Fig. 3 Plots of relative band intensity Irel against molar percentage mixture composition for (a) Pb3O4 and (b) HgS.Irel is I 547 Pb3O4+I 254 HgS for (a) and I 254 HgS+I 547 Pb3O4 for (b). Fig. 4 FT-Raman spectrum of 50+50 (50 mol%) Pb3O4–HgS mixture showing the relative weakness in intensity of the 547 cm21 band for Pb3O4 compared with that at 257 cm21 for HgS. Fig. 5 FT-Raman spectra of mediaeval pigments from the Convento de la Peregrina. (a) ‘Window’ 7, lhs; (b) ‘window’ 3, rhs; (c) ‘window’ 3, upper edge sill; (d) ‘window’ 5, lower sill; (e) ‘window’ 7, rhs.Wavenumber range, 50–650 cm21. The relative intensities of the red lead band at 547 cm21 and the cinnabar band at 254 cm21 should be noted, reflecting the large range of compositions in the mediaeval pigments. Table 2 Cinnabar–red lead compositions in mediaeval pigments from two mediaeval sources of artwork, as determined by Raman spectroscopy using the calibration plots defined in this paper Source of pigment sample Cinnabar (%) Site 1: Sahagun, Convento de la Peregrina— Sample 1: ‘window’ 7, lhs 82 Sample 2: ‘window’ 3, rhs 35 Sample 3: ‘window’ 3, upper edge 9 Sample 4: ‘window’ 5, upper 12 Sample 5: ‘window’ 5, centre 0.5 Sample 6: ‘window’ 5, lower 100 Sample 7: ‘window’ 6, rhs 5 Sample 8: ‘window’ 7, rhs 75 Sample 9: ‘window’ 7, lower edge 0.5 Sample 10: ‘window’ 10, arch 19 Site 2: Basconcillos del Tozo, SS Cosmo y Damien— Sample 1: angel trumpet 36 Sample 2: Christ’s cloak 100 Analyst, 1999, 124, 1323–1326 1325following conclusions may be drawn.(i) There is a large variation in composition between the 12 mediaeval samples of red pigment studied here. At the Sahagun site the samples show a cinnabar composition ranging between 0.5 and 100%, whereas the two samples from Basconcillos del Tozo have a cinnabar composition of 36 and 100%. (ii) The wall paintings at Sahagun consist of geometric designs with an Islamic influence which are similar to those exhibited at the Alhambra in Granada, yet the church under study has always been in a Christiancontrolled part of Spain.There seems to be little reason for the adulteration of the cinnabar here, except perhaps for the economic factor which has been recognized in other cultures. Even so, two samples contained high levels of cinnabar in their composition (samples 1 and 8) and sample 6 was pure, unadulterated cinnabar as determined spectroscopically. Perhaps there was some special significance in sample location on the frescoes which determines the cinnabar content; certainly, only for the heavily adulterated red lead samples (e.g., samples 3, 4, 5, 7 and 9) is there any colour difference observed visually.From our results it is apparent that the adulteration of the cinnabar was not undertaken in batch preparation of pigment at Sahagun and it is possible that several artisans were involved in the decoration, each of whom had their own procedures for mixing their colour palettes.(iii) The analysis of the two samples from the Church at Basconcillos del Tozo reveals an intriguing piece of information about the hierarchical use of pigments which hitherto has only been noticed in the historiated initials of valuable mediaeval and Renaissance manuscripts. Sample 1 from an angel’s trumpet consists of heavily adulterated cinnabar, whereas sample 2 from Christ’s cloak is pure cinnabar, as determined spectroscopically. The use of pure, expensive pigments only for the most important religious figures of Saints, of Christ and of the Virgin Mary has been alluded to previously.Here is the first example of this that we have found in mediaeval wall paintings. The successful demonstration of the Raman spectroscopic method in the determination of mixture compositions in ancient samples of pigments has established for the first time a protocol for the extension of this type of work to other systems involving cinnabar and red lead mixtures.Acknowledgements We are grateful to the British Council and Spanish Ministry of Science for the award of an ‘Accion Integrada’ grant during the tenure of which this work was carried out. Also, we thank the Director of Culture, Junta de Castille y Leon, for permission to sample the paintings at Convento de la Peregrina, Sahagun, and the assistance of D. Jose Valdavida Lobo for permission to study the wall paintings at Basconcillos del Tozo.References 1 Artists’ Pigments: a Handbook of Their History and Characteristics, ed. R. L. Feller, Cambridge University Press, Cambridge, 1986, vol. 1. 2 E. W. Fitzhugh, in Artists’ Pigments: a Hanbook of Their History and Characteristics, ed. R. L. Feller, Cambridge Univesity Press, Cambridge, 1986, vol. 1, p. 109. 3 The Elder Pliny’s Chapters on Chemical Subjects, Part 1, ed. K. C. Bailey, Notes and Translation by K. C. Bailey, E. Arnold & Co., London, 1929. 4 S. Rozenberg, in Proceedings of the International Workshop on Roman Wall Painting, Fribourg, ed. H. Béarat, M. Fuchs, M. Maggetti and D. Paunier, Institute of Mineralogy and Petrography, Fribourg University, Pérolles, 1997, p. 63. 5 H. C. Hoover and L. H. Hoover, translation of De re Metallica, by Agricola, 1556, New York, 1950. 6 E. H. Schafer, The Early History of Lead Pigments and Cosmetics in China, T’oung Pao, 1955, pp. 44 and 413. 7 N. Porat, in Art and Architecture, Masada V: the Y. Yadin Excavations, 1963–65—Final Reports, ed. G. Foerster, Jerusalem, 1995, p. 224. 8 D. V. Thompson, Jr., Il Libro dell’arte; the Craftsman’s Handbook of Cennino d’Andrea Cennini, Oxford University Press, Oxford, 1932. 9 M. H. Butler, in Paintings by Renoir, ed. J. Maxon, Chicago, 1973, p. 208. 10 F. Rull Perez, H. G. M. Edwards, A. Rivas and L. Drummond, J. Raman Spectrosc., 1999, 30, 301. 11 H. G. M. Edwards, D. W. Farwell, F. Rull Perez and S. Jorge Villar, J. Raman Spectrosc., 1999, 30, 307. Paper 9/04083H 1326 Analyst, 1999, 124, 1323–1326

ISSN:0003-2654    

DOI:10.1039/a904083h

出版商:RSC    

年代:1999

数据来源: RSC