Sorptive sample preparation a review

Transcription

1 Anal Bioanal Chem (2002) 373 :3 22 DOI /s SPECIAL ISSUE PAPER E. Baltussen C. A. Cramers P. J. F. Sandra Sorptive sample preparation a review Received: 21 November 2001 / Revised: 15 February 2002 / Accepted: 20 February 2002 / Published online: 9 April 2002 Springer-Verlag 2002 E. Baltussen ( ) Notox B.V., Hambakenwetering 7, 5231DD, s Hertogenbosch, The Netherlands erik.baltussen@notox.nl C.A. Cramers Eindhoven University of Technology, Laboratory of Instrumental Analysis, P.O. Box 513, 5600 MB Eindhoven, The Netherlands P.J.F. Sandra Research Institute for Chromatography, Kennedypark 20, B-8500, Kortrijk, Belgium Abstract Most sample-enrichment procedures currently available rely on adsorption of the analytes of interest by a suitable adsorbent material. Although good performance can be obtained for many practical problems, in some cases the applicability of adsorptive sample preparation falls short, particularly for the enrichment of polar and/or high-molecular-weight compounds, especially in combination with thermal desorption. Because of the very strong retention of adsorbent materials, undesired effects such as incomplete desorption and artifact formation are observed. Polar solutes are easily adsorbed but readily undergo surface-catalyzed reactions and on desorption yield compounds different than those originally sampled. Highmolecular-weight compounds cannot be desorbed because of extremely strong interactions with the adsorbent and their low volatility. To overcome some of these problems sample-preparation techniques based on polydimethylsiloxane sorption have been developed over the past 15 years. In contrast with adsorptive trapping, sorption is based on dissolution of the analytes in a liquid polymeric material. This is a much more inert means of solute retention which overcomes some of the limitations encountered when working with adsorbents. In this contribution, the basic principles of sorption, the different instrumentation used, and applications of the technique will be reviewed. The review covers the sorptive sample-preparation techniques, open-tubular trapping (OTT), solid-phase microextraction (SPME), gum-phase extraction (GPE), equilibrium gum-phase extraction (EGPE), and stir-bar-sorptive extraction (SBSE). Because of the nature of sorptive sample-preparation techniques, which perform particularly well in combination with thermal desorption, this review focuses strongly on gas chromatography as the means of chemical analysis. Keywords Sorptive sample preparation Gas chromatography Thermal desorption Polydimethylsiloxane Introduction Most of the sample-preparation techniques currently available rely on trapping of the analytes of interest from the sample (gas, liquid, or solid) by an adsorbent material; this is followed by desorption and (chromatographic) analysis. Adsorbents are porous materials with a high internal surface area (typically m 2 g 1 ) and the analytes are temporarily stored on the adsorbent surface. After analyte trapping and matrix removal the trapped analytes can be released by extraction with a small amount (typically milliliters) of an organic solvent. An aliquot (typically microlitres) of this extract is subsequently injected into the analytical instrument. Although this approach works quite successfully, it is likely to result in poor sensitivity because only a fraction of the sample is used. Overcoming this sensitivity limitation is the topic of much research into sample-preparation. Possible solutions include online combination of extraction with liquid chromatography and injection of large volumes into the analytical system (i.e. large-volume injection in gas chromatography). As an alternative to liquid desorption, thermal desorption under an inert gas stream is increasingly being used. Thermal desorption can be coupled rather conveniently to a gas chromatograph and the (heated) carrier gas is used for thermal desorption. When cryogenic focusing is employed, quantitative transfer of the analytes trapped on the adsorbent material to the chromatographic column is possible; this results in a considerable increase in sensitivity compared with liquid desorption. Consequently, thermal

2 4 desorption can be a very attractive alternative to classical procedures involving liquid desorption. It must be noted, however, that if analysis is to be successful the analytes subjected to thermal desorption must be thermally stable, otherwise decomposition will occur. Because volatile and thermally stable analytes are amenable to GC analysis, thermal desorption is, in practice, only used in combination with gas chromatography. Many standardized methods are available for enrichment of analytes from gaseous and liquid samples on adsorbent cartridges followed by liquid desorption and injection into a chromatographic system. Unfortunately, relatively few standardized analytical methods take advantage of the high sensitivity of thermal desorption, because although a wide variety of adsorbent materials is available, none is universally suitable for thermal desorption. Many adsorbents (i.e. most inorganic adsorbents) interact too strongly with the trapped analytes, requiring very high desorption temperatures, which lead to degradation reactions. This limits the use of inorganic adsorbents to highly volatile apolar analytes only. Organic adsorbents, on the other hand, often lead to poor blanks because of thermal decomposition of the material itself. Most adsorbents also have significant catalytic activity, even at low temperatures; this prevents their use for the enrichment of chemically labile compounds. This is, of course, another highly undesirable effect that often prevents the application of organic adsorbents. These problems with the thermal sorption of adsorbents have prompted several research groups to focus on another class of material sorption materials. In contrast with adsorbents, sorption (dissolution or partitioning) materials are a group of polymeric materials that are above their glass transition point (T g ) at all the temperatures employed. In this temperature range sorbents are in a gumlike or liquid-like state and behave similarly to organic solvents. Sorbents are, in principle, homogeneous, non-porous materials in which the analytes can dissolve. The analytes do not, therefore, undergo real (temporary) bonding with the material but are retained by dissolution. In this contribution the state-of-the-art in sorptive sampling and thermal desorption will be reviewed. First some basic concepts, illustrating the different approaches used for sorptive sampling, are presented. This is followed by a detailed overview of the different techniques described in the literature open-tubular trapping (OTT), solid-phase microextraction (SPME), gum-phase extraction (GPE), equilibrium gum-phase extraction (EGPE), and stir-bar-sorptive extraction (SBSE). In this contribution, emphasis will be on the high-sensitivity sorptive techniques GPE, EGPE, and SBSE; the older (and less sensitive) techniques OTT and SPME are described in much less detail and the reader is referred to the literature for more detailed information. Basic concepts The primary aim of all sample-preparation methods is transfer of the analytes of interest from their original surroundings (sample matrix) into a form more suitable for introduction into the analytical instrument. This can be achieved by many different techniques all of which have their strengths for specific analytes and analyte matrix combinations. Usually the sample is placed in direct contact with the extraction phase (extractant) to accomplish transfer of the analytes into the extractant. Subsequently the extractant can be processed further or, occasionally, it can be introduced directly into the analytical device. Several basic sample preparation concepts are described in forthcoming sections. Static sampling In static sampling techniques all the extractant is in contact with all the sample during extraction, i.e. neither the sample nor the extractant is renewed. Static techniques rely on diffusion of the sample analytes into the extractant with the ultimate goal of reaching equilibrium between both phases. Selection of the extractant phase is usually based on the so-called like like principle a substance will always have the highest affinity for a phase with properties similar to those of the substance itself. This means that if an apolar compound is to be extracted from a polar matrix, an apolar extractant should be used. It should be noted that mixing procedures such as stirring, shaking, or sonication are often used to promote diffusion of analytes from the sample into the extractant. This, however, only affects the time required for equilibration; it does not affect the equilibrium itself or other properties of the static process. The most important factor governing static extraction is the distribution constant (K) which is defined as: K = C E = m E V S = m E β (1) C S m S V E m S where C S is the concentration of analyte in the sample, in g L 1 ; C E is the concentration of analyte in the extractant, in g L 1 ; m S is the mass of analyte remaining in the sample, in g; m E is the mass of analyte in the extractant phase, in g; V E is the volume of the extractant, in L; and V S is the volume of sample, in L. β is the phase ratio of the static extraction system and is defined as V S /V E. In subsequent equations, the total mass of analyte in the system is defined as m tot (m E +m S ). Rewriting Eq. (1) leads to a more useful expression, that of the extraction efficiency (η= m E /m tot ): η = 1 β K + 1 (2) The extraction efficiency is usually expressed as a percentage and as such is generally known as the recovery. From Eq. (2) it is important to note that the only two terms affecting the recovery of an analyte are β and K. For very large partitioning constants the numerator becomes unity, leading to 100% recovery. Very large phase ratios (small volume of extractant relative to the sample volume) lead to a large numerator and, consequently, to low

3 recovery. In practice, K is often a more or less fixed constant which depends mainly on properties of the analyte and the characteristics of the sample and extractant phases. β is chosen by selecting the phase volumes applied, usually to ensure high recovery with a minimum amount of extractant. It should be noted, however, that under static conditions extraction is never complete and that some of the analytes always remains in the sample no matter how large K is or how small β is. Although static sampling can be an easy, reliable, and straightforward technique, because it relies on the equilibrium distribution of compounds rather than on exhaustive extraction care should be taken to ensure the distribution constant (K) is equal in all experiments, including calibration and sample extraction. Although this seems to be a simple requirement, in practice it is often not so. In chemical equilibria, temperature has a dominant effect on equilibrium and distribution constants, so careful control of the temperature, often within 1 2 C is necessary. In the laboratory this requirement is easily met; it can be problematic in field-sampling applications. As has been pointed out above, a distinction between adsorption and sorption extraction phases must be made. Sorption phases (including all organic solvents, water, ideal gases, and polymeric materials at a temperature above their glass transition point) retain solutes purely by dissolution (at sufficiently low concentrations). The analytes partition into the bulk of these phases where they can diffuse freely throughout the sorbent. They therefore experience the bulk properties of the sorbent and, as long as the total amount of sorbed compounds is less than 1%, these bulk properties do not change significantly with concentration. High concentration levels of this kind are seldom found in practice and static sorptive extraction is, therefore, usually a very reliable approach. Static extraction performed with an adsorbing phase is, unfortunately, much more complicated. Here, the analytes are retained on an active surface containing a fixed number of adsorptive sites. The equilibrium reached is that between analytes present in the sample and those adsorbed on the adsorbent surface. This means that if the sample concentration is high, all the adsorptive sites are occupied and increasing the concentration of the sample will no longer lead to an increase in the amount of compound adsorbed; this is, of course, a highly undesirable effect. Analyte concentration levels are, however, usually low enough to circumvent this effect, so if a single compound is adsorbed from an otherwise clean sample this is not a real problem. When several analytes are adsorbed simultaneously this might, however, become problematic. Not only do the different analytes compete for the same adsorptive sites but matrix compounds present at relatively high concentrations that are of no interest in a particular analysis (e.g. salts, humic acids, proteins) can block adsorptive sites leading to unpredictable and irreproducible results. The application of adsorbents in static sampling is, therefore, limited to clean and dilute samples. In special circumstances, particularly if the sample is a solid, the sample itself might also have adsorbing properties, preventing the reliable use of static sampling techniques. Dynamic sampling Whereas in static sampling mixing, stirring, and other dynamic processes are solely a means of promoting faster equilibration, dynamic sampling procedures essentially require that these basic dynamic processes ensure complete extraction. In dynamic sampling all the extractant is not immediately brought into contact with all the sample. Many dynamic sampling techniques resemble chromatography in that they also are based on the use of a stationary phase (often the extractant) and a moving, mobile phase (often the sample), see Fig. 1. Gaseous or liquid samples are usually pumped through the extractant that can, for example, be a packed bed. The analytes will be retained in the packed bed and, consequently, the concentration of analyte in the sample will decrease through the bed. Initially the concentration of analyte in the outgoing sample phase will be zero and sampling is usually stopped when the first analyte of interest starts to elute from the trap. This is called breakthrough sampling and will be discussed in this section. It is, however, also possible to continue sampling beyond the breakthrough point until all analytes are in equilibrium with the extractant. This is a relatively new technique, called equilibrium sampling ; it will be illustrated by discussion of EGPE. The most important property in breakthrough sampling is the breakthrough volume, which determines the maximum volume of sample that can be passed through the trapping device before analytes are no longer sufficiently retained. It is important to state that there is no single def- Fig. 1 Principle of dynamic breakthrough sampling. Top, the sample is pumped through the extraction trap; middle, the concentration profiles in the sample and in the extractant; bottom, analyte concentration in the outgoing (extracted) sample as a function of the volume sampled 5

4 6 inition of breakthrough volume; rather the breakthrough volume depends on the acceptable loss of analyte. The acceptable loss of analyte is usually taken as 5 to 10%. A general equation for calculation of the breakthrough volume is: V B = V 0 (1 + k) f (N, b) (3) where V B is the breakthrough volume, in L; V 0 is the void volume of the trap, in L; k is the retention factor; and f(n,b) is a function taking into account the number of theoretical plates in the trap (N) and the acceptable breakthrough loss (b). The retention factor is the same as that defined in chromatography and corresponds to K/β, with β defined here as V 0 /V E. For strongly retained compounds (k>>1), Eq. (3) reduces to: V B = V E K f (N, b) (4) This equation represents the conventional concept of breakthrough the amount of stationary phase multiplied by the capacity factor, corrected for the speed of sampling. Both V E and K are usually readily determined from simple experiments, so all that must be found is an expression for f(n,b). In the literature there has been some controversy about the definition of the breakthrough factor, as will be illustrated below. Definition of the breakthrough factor (b) In the literature two definitions of the breakthrough loss or breakthrough factor (b) are used side by side. The first, and most commonly applied, breakthrough factor is based on the momentary loss of analyte; this will be referred to as the differential breakthrough factor (b D ). The second definition of breakthrough is based on the total loss of analyte and will be referred to as the integral breakthrough factor (b I ). Mathematically, the differential breakthrough factor is defined as: b D = C O,V (5) C 1 where C O,V is the concentration of analyte in the outgoing sample at a certain sampled volume (V) and C I is the concentration of analyte in the original sample. This is presented graphically in Fig. 2 where the 10% breakthrough factor is shown on the curve. Although the differential definition can easily be calculated from chromatographic theory, for real-life sampling this is not often very useful, because the value of b D only represents the momentary loss at the point sampling is stopped, the breakthrough volume. The momentary analyte loss increases with increasing volume sampled, however, as is apparent from Fig.2; this means that although at the predicted breakthrough volume the momentary analyte loss will be b D, the overall amount of analyte lost will be less than b D. If, therefore, a sample loss of 10% is acceptable (a b D of 0.1), at the calculated breakthrough volume the actual overall analyte loss will always be less than 10%. Not only does this lead to reduced sensitivity, because a larger volume could have been sampled, it is also impossible to correct for sample losses, because the exact analyte loss it not known. Fig. 2 Illustration of the definition of the differential breakthrough factor (b D ). Shown here is the 10% breakthrough volume under differential conditions Fig. 3 Illustration of the definition of integral breakthrough. Shown here is the 10% breakthrough volume under integral conditions The arguments presented above have led to the adoption of the integral breakthrough factor, b I, which is defined as: b I = V B 0 V B 0 C O,V dv C I dv = V B 0 C O,V dv C I V B (6) As is apparent from Fig. 3, the integral definition of the breakthrough volume is based on the amount of analyte lost from the trap relative to the total amount of analyte sampled. At the breakthrough volume under integral breakthrough conditions the amount of analyte lost is exactly equal to b I. The predicted breakthrough volume under integral conditions thus represents the true maximum volume that can be sampled before a predetermined portion of analyte (b I ) is lost. In practice care should be taken when comparing different means of calculation of f(n,b), because b can be either a differential- or integral-based term. If, moreover, breakthrough factors are determined experimentally, integral rather than differential breakthrough factors are determined. The use of b I rather than b D is, therefore, usually recommended. Differential breakthrough factors This model for analyte breakthrough was initially developed by Werkhoven-Goewie [1] and assumes that ana-

5 lytes elute from the extraction column as Gaussian-shaped bands; this is valid for sufficiently large plate numbers. The expression then found for f(n,b), from Eq. (4), is: f (N, b) = 1 a G (b D ) (7) N b D a G (b D ) 0.1% % % % % where a G is a constant which depends on the accepted breakthrough. Values for a G are shown in Eq. (7). Figure 4 shows a graph of the dependence of f(n,b) on N for different values of b D. It is clear that at any b D a non-negative breakthrough volume is obtained only at a certain, critical plate number (N crit =a G2 ). In practice this implies that at very low plate numbers immediate breakthrough is expected. A sample loss of 10% before sampling is even started is, however, very unrealistic and is, in fact, an artifact of Gaussian theory, which is valid at relatively high plate numbers only. Care should, therefore, be taken when using values predicted by Eq. (7), which should, preferably, be used only if f(n,b) is predicted to be in excess of 0.5 (50%). With the theory presented above it is possible to calculate breakthrough volumes all that is required is an expression for calculation of plate numbers. For capillary (open tubular) traps this can be the Golay equation, which gives an exact theoretical description of the plate number; for packed columns semi-empirical equations such as those proposed by van Deemter and Knox can be used. Integral breakthrough factors The problems associated with the differential breakthrough theory described above have led researchers to develop alternative, more realistic, expressions for f(n,b). Lövkvist and Jönsson [2] compared several dedicated equations for breakthrough curves at low plate numbers Fig.4 f(n,b) predicted from the differential Gaussian breakthrough model (Eq. 7). Five curves are drawn for different breakthrough fractions (b D ) Table 1 Terms of Eq. (8) as a function of the breakthrough level (b I ) b I a L,0 =(1 b I ) 2 a L,1 a L,2 0.5% % % % % Fig. 5 f(n,b) predicted from the integrated Lövkvist breakthrough model (Eq. 8). Five curves are drawn for breakthrough fractions (b I ) of 0.5 to 10% and suggested the following expression, which they found to be valid for strongly retained compounds. ( f (N, b) = a L,0 + a L,1 N + a ) 1 L,2 2 (8) N 2 where a L,0, a L,1, and a L,2 are constants which depend on the breakthrough factor (b I ). Values for a L,0 through a L,2 are listed in Table 1. Breakthrough curves predicted by use of Eq. (8) are shown in Fig. 5. It is clear that non-negative breakthrough volumes are predicted at any plate number. For high plate numbers f(n,b) can become larger than unity; this is indicative of a breakthrough volume in excess of the retention volume. This is not an artifact of Eq. (8) but is a result of the definition of the integral breakthrough factor. For a trap with an infinite plate number (N= ) no analyte is lost before the retention volume is reached. At V R, therefore, the breakthrough factor is still zero. For sample volumes in excess of the retention volume the amount of analyte sampled at the trap inlet is equal to the amount lost at the trap outlet, hence the breakthrough factor increases from this point on. The authors of Eq. (8) recommend the use of their breakthrough expression with a modified Knox equation [3, 4] for prediction of plate numbers: h r = 3υ 1 / υ (9) υ where h r is the reduced plate height and υ is the reduced velocity in the trapping column. These reduced properties are defined as: h r = H d p = L N d p υ = u d p D m (10) 7

6 8 where H is the plate height, in m; d p is the diameter of packing particles, in m; L is the length of the trapping column, in m; u is the superficial linear velocity in the trap, in m s 1 and D m is the diffusion constant in the mobile phase, in m 2 s 1. Eqs. (9) and (10) are valid only when the pressure drop over the packed bed can be neglected. Adsorptive sample preparation Adsorbents for thermal desorption can be divided into three categories. The first includes inorganic carbon-based materials such as carbon blacks, carbon molecular sieves, and activated carbon. These materials usually have very high affinity for organic compounds and are most often used for gaseous samples. Carbon molecular sieves can be used to retain C 2 C 3 hydrocarbons, and even methane, whereas activated carbon and carbon blacks are more suited to less volatile analytes. Because of their inorganic nature, these materials can be heated to high temperatures ( C) without degradation. The second category includes inorganic materials based on silica and alumina. Silica-type materials can be used unaltered to trap analytes from gaseous and liquid samples and are generally more suited to larger molecules than are inorganic carbon-based adsorbents. These materials are thermally stable up to C. The surfaces of these materials can also be covered with organic groups, resulting in materials such as octadecylsilica (ODS), which enable very successful enrichment of liquid samples. The thermal stability of organic-coated silicas is, however, poor, because at elevated temperatures (>100 C) the organic groups tend to be expelled from the surface. The third category is the polymeric adsorbents. This is the largest and most diverse group and comprises many commonly used materials, e.g. Tenax and Chromosorb. Most are synthetic and consist of polymers of building blocks such as styrene. One of the most important drawbacks of this type of material is that (especially on heating) depolymerization occurs, releasing monomeric units and reaction products thereof. These, unfortunately, include many of the target analytes, e.g. styrene and benzene. At moderate temperatures, at which adsorbent degradation is not very pronounced, small quantities of emitted compounds can easily lead to false-positive results. Tenax, the most commonly used organic adsorbent, is particularly notorious for its background of benzaldehyde, acetophenone, benzophenone, and other aldehydes and ketones. In practice maximum temperatures range from 150 C for some Chromosorbs to 350 C for Tenax, although traces of water or oxygen strongly promote degradation and can lead to significant deterioration of the background. The adsorbent surface always contains active groups (adsorptive sites) that can interact with the analytes and bond them to the surface. Depending on the nature of adsorbent and analyte the interaction can range from very weak van der Waals-type bonding to very strong ionic interaction. The strength of the interaction also determines the desorption process required. Desorption with liquid can break strong adsorbent analyte interactions whereas thermal desorption can overcome relatively weak van der Waals- type interactions only. Adsorptive sampling is, therefore, most often combined with liquid desorption. Thermal desorption is prone to poor recoveries, even at very high temperatures. These high temperatures might result in degradation of the adsorbent materials or promote catalytic breakdown of the trapped analytes; this is most pronounced for polar analytes. Adsorptive sampling is, therefore, not often used in combination with high-sensitivity thermal desorption. Applications of adsorptive sampling thermal desorption are almost completely limited to very apolar analytes, e.g. hydrocarbons such as alkanes, alkenes, and aromatics. There is clearly an urgent need for alternative techniques that enable extension of the range of applicability of thermal desorption to more polar compounds. The rest of this review evaluates sorptive sample preparation as an ideal means of overcoming the limitations of adsorptive sampling. Sorptive sample preparation Sorptive materials (or sorbents) are a group of polymeric materials with a glass transition temperature (T g ) below the temperature at which the material is used during the sampling storage desorption process. Although, initially, this might seem a trivial requirement, the consequences are enormous. At temperatures above their T g polymeric materials no longer behave as solid materials but assume a gum-like, or even liquid-like, state with properties, e.g. diffusion and distribution constants, similar to those of organic solvents. Sorbents are, in principle, homogeneous, non-porous materials in which analytes can actually dissolve. The analytes do not, therefore, undergo real (temporary) bonding with the material but are retained by dissolution. It should be noted that all sorptive materials work only in the sorption regime above their glass transition point. This means that on cooling any sorbent loses the sorptive mechanism below its glass transition point and is then turned into an adsorbent with a low specific surface area. The most commonly used sorbent is the apolar polydimethylsiloxane (PDMS), a 100% methyl-substituted siloxane polymer commonly used as a stationary phase in gas chromatography. Its structure is shown in Fig. 6. This material is so popular because it is very inert, reducing the risk of losses of unstable and/or polar analytes by irreversible adsorption or by catalytic (surface) reaction. Retention data for many compounds can be found in the literature. In addition, PDMS synthesis is relatively simple and leads to very reproducible properties, and consistency between manufacturers. Its degradation products are, moreover, very well known and can easily be identified by mass spectrometry. These advantages and the lack of availability of other materials as stable, reproducible, and inert as PDMS account for its widespread use in sorptive sample-

7 Fig. 6 Structures of sorbents commonly used for sample enrichment. Glass transition temperatures [5]: polydimethylsiloxane ( 125 C) and polybutylacrylate ( 54 C) preparation techniques. Alternative materials, e.g. the polar poly(butyl)acrylates, are used for more polar analytes that have a low affinity for PDMS and consequently do not partition very well into this material. The mechanical stability of sorbents is usually provided by means of crosslinking, which ensures that the extraction phase will retain its shape, even at elevated temperatures. It is essential to realize that when sorbents are used preconcentration of analytes occurs by sorption of the analytes into the polymeric liquid phase instead of adsorption on to a solid adsorbent surface. Different sorptive enrichment procedures are based on the different approaches and geometries in which the sorbent is used for sample extraction. Four techniques, all of which have been well described in the literature, can be distinguished. The first, opentubular trapping, is the oldest technique and employs a (thick film) capillary GC column for sampling. The second technique, solid-phase microextraction (SPME) is based on use of a PDMS-coated fiber which, when not in use, is protected by being withdrawn within the needle of a syringe-like device. The third technique, gum-phase extraction (GPE), is based on a bed packed with sorbent material. The applicability of this technique has been illustrated in the breakthrough mode and by use of the novel equilibrium sampling approach (EGPE). The fourth, and latest, sorptive technique is stir-bar-sorptive extraction (SBSE) which is based on static extraction of liquid samples with a sorbent-coated stir-bar. Open-tubular trapping Although open-tubular traps coated with adsorbent particles have been used [6, 7], focus has been on sorbentcoated capillaries, because of their favorable characteristics (which are similar to those of capillary GC columns). Occasionally short capillary traps, which can be desorbed in the injector (e.g. PTV or split/splitless) of a gas chromatograph, have been used [8]. More commonly, coated lengths of fused silica columns with an inner diameter of mm are employed; the typical film thickness is µm [9] but occasionally high-capacity open-tubular traps with extremely thick films of 100 µm [10] or even 165 µm [11] are preferred. Although films up to 15 µm can be prepared in short columns by procedures commonly employed for the preparation of thick film capillary columns [12], the use of a low-molecular-weight methylsiloxane polymer was suggested by Bicchi et al. [13], because it enables the preparation of 15-µm films in capillary traps up to 5 m long. Immobilization of the stationary phase is performed by addition of dicumyl peroxide to the coating solution. Film stability is further enhanced by treatment with azo-t-butane. Both agents effect crosslinking upon heating. It has been found that conditioning for one week at 250 C is required if a perfectly immobilized and stable film is to be obtained. Classical coating techniques are not suitable for thicker films, because the deposited film will quickly rearrange into droplets owing to drainage and Rayleigh instability [14]. An approach that enables preparation of traps containing very thick films and circumvents the problems associated with dynamic coating has been described by Roeraade et al. [10, 15, 16]. The film was fixed by heat-accelerated crosslinking of a suitable pre-polymer in a process in which the column is pulled through an oven at the same speed as evaporation of the coating liquid. In this way stable films up to 100 µm thick could be obtained. Traps with even thicker films can be produced by means of an innovative process described by Burger et al. [11, 17]. Instead of using coating solutions or pre-polymers these authors used (crosslinked) polydimethylsiloxane tubing (0.65 mm outer diameter, o.d. and 0.3 mm inner diameter, i.d.), which was transferred into a fused-silica capillary by use of liquid nitrogen. This procedure facilitates the production of very thick films. The advantage of this approach is that the capillary obtained is very stable with a low background profile and favorable sorption characteristics. Open-tubular trapping gaseous samples Open-tubular traps have been applied to the analysis of gaseous samples by many groups [8, 9, 10, 11, 18]. Most commonly traps up to 1 3 m long are used for the retention of gaseous analytes. Sampling is usually performed by means of a vacuum pump attached to the outlet of the open-tubular trap. The reverse, pushing the sample through a trap by means of a pump is not a very good approach, because this can easily lead to contamination of the sample by compounds released by the pump or to alteration of the composition of the sample as a result of (ad)sorption of analytes inside the pump. This limitation to sampling by suction implies that restriction of flow by the opentubular trap should not be excessive, otherwise a very low flow rate will result, leading to excessively long sampling times. For high capacity (expressed as K V E, Eq. 8), on the other hand, a long open-tubular trap is required, especially for volatile compounds. In practice, a compromise must be found between sample capacity and sampling speed. Open-tubular traps have been successfully employed for a range of gaseous samples including wine headspace [19], plant volatiles [9, 13, 20], pheromones [21], and environmental air samples [22, 23]. For a wide variety of compounds, e.g. alkanes, aromatics, esters, and alcohols, good performance was obtained at trace concentration lev- 9

8 10 els, illustrating the favorable properties of sorbents. For gaseous samples, in particular, the use of very-thick-film traps seems essential, because sufficient trapping capacity from samples up to ca. one liter is possible with these traps, even for the most volatile analytes. The disadvantage of using films up to 200 µm is that desorption is slow and cryofocusing of the thermally released analytes before injection on to the analytical column becomes essential. Cryotrapping can be performed either in a separate device or on-column [24] with liquid carbon dioxide or nitrogen; this results in full utilization of the efficiency of the analytical column. By use of open-tubular traps coated with 80-µm PDMS films Blomberg and Roeraade [25] demonstrated the viability of OTT for collection of fractions of compounds eluting from a capillary GC column. Over extended time periods and from multiple GC runs compounds were trapped quantitatively on the OTT. Recovery of the collected volatile compounds was accomplished either by thermal desorption or by extraction of the OTT with pentane. Complete recoveries could be obtained by either method. The use of open-tubular traps for enrichment of samples for high-speed narrow-bore capillary gas chromatography was described by Pham-Tuan et al. [26] In contrast with the normal sampling procedure, in which sampling is stopped before breakthrough of the analytes, sampling was continued until the trap was saturated with analyte and had reached equilibrium with the sample. In equilibrium sampling the open-tubular trap was fully saturated with sample before thermal desorption. The trap was subsequently heated for desorption and only a small part (time-sliced injection) was transferred to the analytical column. This enabled the use of open-tubular enrichment without the need for cryotrapping or other focusing techniques. This can be a substantial advantage in field application of (micro) GC, when cryofocusing is very impractical; the use of sorptive sampling preparation can, therefore, be extended to this field also. Unfortunately, heartcutting injection results in comparatively low sensitivity, particularly in combination with the thermal conductivity detector (TCD) of the micro-gc; this results in a very limited range of application. can be tolerated [27]. An HPLC pump can be used for effective delivery of the sample, free from contamination, to the capillary trap so that pressure limitations are not important, as was observed for gaseous samples. Although open-tubular traps with films <15 µm can be used successfully, low breakthrough volumes often result, because of the small amount of stationary phase present. The use of thick-film traps [28], with films up to 165 µm, seems more promising, because they enable the retention of analytes from larger sample volumes. Thin-film traps can be used for sample volumes up to 2.5 ml if the stationary phase is swollen by absorption of chloroform. This has been demonstrated by Mol et al. for analytes ranging from the apolar toluene to the more polar dimethylphenol and chloroaniline [29]. Kaiser and Rieder [30] described an OTT technique which used the same capillary for both analyte enrichment and chromatographic separation. This was achieved by backflushing the capillary between these two steps and cryotrapping the analytes at the head of the column. Mol et al. [31] described the use of an open-tubular trapping column as a means of phase switching in on-line reversed phase LC GC. By use of the breakthrough theory described above (Eqs. 3, 4, 5, 6, 7, 8) in combination with the Golay equation for determination of the number of theoretical plates in the open tubular column, the conditions for OTT were calculated and optimized. As an application, the 16 EPA priority PAH were trapped from an LC effluent on an open-tubular trap. After brief drying the analytes were desorbed with a small amount (80 µl) of hexane which was subsequently injected in to a large-volume-injection (LVI) PTV GC system. Equilibrium enrichment for aqueous samples analogous with that used for gaseous samples has been described by Aguilar et al. [32]. Again the sample was passed through the trap until equilibrium was reached between the sample and coated stationary phase. The applicability of the system was illustrated for very polar and small analytes which cannot easily be determined quantitatively by any other technique. Analysis of amines at low µg L 1 levels in aqueous samples is illustrated in Fig. 7. The past four years has seen renewed interest in opentubular trapping as a technique for enrichment of aqueous Open-tubular trapping liquid samples Open-tubular trapping can be an attractive alternative to classical techniques for the enrichment of aqueous samples. The main advantage of OTT over alternative techniques is that complete removal of water from the trap can be achieved by purging the capillary with a short plug of gas. Long drying times, often needed for solid-phase extraction, are not required. The main disadvantage of OTT is, however, is its low retentive power for the trapping of compounds from aqueous samples, particularly very polar compounds that do not partition strongly into the stationary phase. Also, because of the low diffusion coefficient of compounds in the liquid phase, the flow rate during sampling is rather critical and only very low flow rates Fig. 7 Equilibrium enrichment of a sample of river water containing ethylamine (1), pentylamine (2), and hexylamine (3), each at a concentration of 2 µg L 1. The sample volume was 40 ml and the desorption temperature 275 C. Published in Journal of Chromatography A [32]

9 samples. Under the name in-tube solid-phase microextraction (ITSPME) several groups have published several interesting applications. It should be noted, however, that this technique is not the same as solid-phase microextraction, which is described below. Surprisingly, all publications on ITSPME employ adsorptive stationary phases for sample extraction, probably because the small amount of stationary phase that can be coated inside an OTT requires the use of highly adsorptive materials. When combined with the equilibrium sampling approach chosen in all ITSPME applications, particular attention must be devoted to matrix effects and competitive adsorption problems. Although ITSPME does not qualify as a sorptive enrichment technique it is still briefly described below, because of its strong analogy with OTT. In the original publication of Eisert and Pawliszyn [33], Omegawax was used to extract several phenylurea pesticides from aqueous samples. Although this Carbowax-type stationary phase is often used as a sorptive stationary phase in gas chromatography, it has true sorptive properties only above its glass transition point (50 60 C). At the temperature of sampling, which was not reported but was probably room temperature, this material is essentially an adsorbent material. The adsorptive nature of Omegawax might also explain the relatively slow rate at which equilibrium was reached. This is yet another illustration of the limitations of (non-porous) adsorptive materials in sample preparation. Analysis of phenylurea herbicides by ITSPME HPLC UV was demonstrated. Detection limits were approximately 10 µg L 1. By use of more or less the same system, combined with mass spectrometric detection, β-blockers and metabolites were successfully monitored at low ng ml 1 levels in urine and serum [34]. Use of a polypyrrole extraction column resulted in improved performance for these compounds, with detection limits below 0.1 ng ml 1 [35]. Analysis of the drug ranitidine at the low- to sub-ng L 1 level has been achieved by ITSPME on an Omegawax column in combination with liquid desorption LC MS [36]. Reliable quantitation of ranitidine in serum over the range ng ml 1, with within- and between-day variation of 2.5 and 6.2%, respectively, was demonstrated. Tan et al. [37] have described the use of ITSPME with a poly(ethylene glycol) extraction capillary combined with liquid desorption GC analysis for the analysis of phenols and BTX compounds (benzene, toluene, xylene and related compounds) from aqueous samples. The performance of the technique under practical sampling conditions cannot, unfortunately, be assessed, because the concentration levels investigated were unrealistically high (between 10 and 40 mg L 1 ). In a recent paper Mullett et al. [38] reported the use of a molecularly imprinted polymeric material for the selective determination of propranolol in biological samples. Although recovery was good, because of the high affinity of the adsorbent for the target analytes, improved selectivity of the material for the target compound was not demonstrated. Solid-phase microextraction 11 Solid-phase microextraction (SPME) is a powerful and innovative extraction procedure introduced by Arthur and Pawliszyn in 1990 [39]. SPME employs a fused-silica fiber with an outer diameter of, typically, 150 µm which is coated with an (ad)sorbent layer 5 to 100 µm thick. This fiber can simply be inserted into a gaseous or aqueous sample for analyte extraction and into the heated zone of a gas chromatograph injector for desorption [40]. The small size of the SPME fiber and its cylindrical shape enable it to fit inside the needle of a syringe-like device. The SPME fiber is attached to the syringe plunger and this arrangement can be used to expose the fiber for extraction or desorption and to retract the fiber for storage and piercing of injector and sample vial septa. The latter is necessary because the coated fused-silica fiber has a very low mechanical strength and cannot, as such, be inserted directly through septa. Here the SPME literature is reviewed briefly to identify the strengths and weaknesses of the technique. For a more detailed and complete overview the reader is referred to the literature [41]. An interesting feature of sorptive extraction on PDMS, that partition constants between sample and fiber can be estimated from literature data, has been well illustrated for SPME of both gaseous [42] and aqueous [43] samples. Linear temperature-programmed retention indices [44] have been used to estimate equilibrium constants between the PDMS fiber and a gaseous sample, on the basis of the assumption that the behavior of solutes in a gas chromatographic column is similar to that of the same solutes in the gas phase SPME coating equilibrium. By comparison of retention indices determined on a capillary GC column and with the SPME extraction device the validity of this concept was confirmed. A combination of the gas PDMS equilibrium constant and Henry s law (water gas equilibrium) was used to yield a value for the PDMS water distribution coefficient [45]. This approach can be used for many volatile solutes, for which a Henry s law constant is readily available. For semi-volatiles compounds the PDMS water distribution constant can be approximated by the octanol water partition coefficient [46, 47, 48]. Numerous applications to the determination of pesticides and other priority pollutants in aqueous samples have been described in the literature; a brief selection is listed in Table 2. Reported detection limits cover a very wide concentration range, from as low as 0.01 ng L 1 (ppt) to as high as 9 mg L 1 (ppm). This is partly because of the different analytical systems used low-sensitivity FID, high-sensitivity ECD, and ion-trap detection (ITD) and other forms of mass spectrometry. More important, the polarity of the target compounds also differs widely from very apolar (PAH, PCB) to polar (some pesticides); this has a substantial effect on the extent to which analytes partition from the polar water matrix into the SPME fiber. Most compounds can be monitored below the desired 1 µg L 1 (ppb) level for surface water. The limit for drinking water analysis (0.1 µg L 1 ) cannot, unfortunately, be reached for many analytes, not even if high-sensitivity GC ECD or

10 12 Table 2 Overview of the performance of typical SPME procedures for extraction of pesticides and other priority pollutants from aqueous samples VOC, volatile organics; VHOC, volatile halogenated organic compounds; PAH, polyaromatic hydrocarbons, PCB, polychlorinated biphenyls, ITD, ion-trap detector; FID, flameionization detector; NPD, nitrogen phosphorus detector; ECD, electron-capture detector a no film thickness mentioned b phenols were derivatized with acetic anhydride before analysis Compounds SPME fiber Sample Technique Detection limit volume VOC, VHOC [49] PDMS, 100 µm 50 ml GC ITD µg L 1 PAH, PCB [50] PDMS, 15 µm 40 ml GC ITD low ng L 1 N-herbicides [51] Acrylate, 95 µm 4 ml GC FID µg L 1 GC NPD µg L 1 GC ITD ng L 1 N,P-Pesticides [52] PDMS, 100 µm 4 ml GC NPD µg L 1 GC ITD µg L 1 Cl-pesticides [53] PDMS, 100 µm 35 ml GC FID µg L 1 GC ECD µg L 1 GC MS µg L 1 P-insecticides [54] Acrylate, 85 µm 4 ml GC FID µg L 1 GC NPD µg L 1 GC ITD µg L 1 P-pesticides [55] PDMS, 100 µm 3 ml GC NPD µg L 1 Acrylate, 85 µm µg L 1 Triazine herbicides [55] Acrylate, 85 µm 3 ml GC NPD µg L 1 2,6-Dinitroaniline herbicides [55] Acrylate, 85 µm 3 ml GC NPD µg L 1 P-Pesticides [56] PDMS, 100 µm 3 ml GC NPD µg L 1 Acrylate, 85 µm µg L 1 Cl-Pesticides [57] PDMS, 100 µm 110 ml GC ECD ng L 1 Anilines [58] PDMS DVB a 5 ml GC FID µg L 1 Phenolic compounds b [59] Acrylate, 95 µm 40 ml GC FID µg L 1 GC MS µg L 1 GC MS is applied. This lack in sensitivity is the most important disadvantage of SPME and is partly because the sorbents have significantly lower analyte capacity than typical adsorbents. This, combined with the extremely small amount of sorbent coated on to the SPME fiber, up to 0.5 µl, makes overcoming the sensitivity problem hard to achieve. As a result sorption SPME is today, almost 10 years after its introduction, still not as widely accepted as deserved, despite its clear instrumental advantages, simplicity, and low cost. To improve the capacity of SPME fibers, several new SPME coatings have been introduced [58, 60]. These include materials such as copolymers of PDMS with divinylbenzene (PDMS DVB) and Carbowax (PDMS WAX) and physical mixtures of PDMS with adsorbents such as Carboxen. Although these materials do indeed, have significantly increased trapping capacity, an important drawback is that the true sorption mechanism is lost, because these materials are no longer pure polymeric sorbents. Carbowax, for example, is used below its glass transition temperature (ca. 70 C) and Carboxen is an inorganic adsorbent. Application of these materials in static sampling is likely to lead to irreproducible results, because adsorption of matrix compounds (salts, humic acids, proteins, etc.) will compete with the target analytes for available adsorbent sites. This complicates, even prohibits, reliable quantitation in SPME. In a study on the extraction of benzodiazepines from biological fluids [61] several SPME fibers, PDMS, PDMS DVB, acrylate and WAX DVB, were compared. It was found that the performance of the PDMS fiber in the extraction of the polar benzodiazepines was very poor, because these analytes do not partition strongly into the apolar PDMS bulk. The other fibers, which are more polar in nature, were able to extract a significantly larger amount of analyte. The highest recoveries were observed for the WAX DVB fiber, closely followed by the acrylate fiber, which yielded recoveries at least half those for the WAX DVB fiber for all compounds. The authors preferred the WAX DVB fiber, because of its greater recoveries, and obtained detection limits of the order of mg L 1 (ppm) from 1 2 ml samples by use of ion-trap mass spectrometry. Although addition of salt was found to have a positive effect on recovery (salting-out effect), the effect of (high concentrations of) other matrix compounds (e.g. proteins) was not investigated, even though this also is likely to have a pronounced effect. If the acrylate sorbent phase had been used these effects would have played a much less important role and, therefore, the acrylate fiber might probably have been preferred, because of its favorable sorption, rather than adsorption, characteristics. An alternative to SPME sampling directly in the aqueous phase is SPME extraction of compounds present in the headspace of the sample headspace-spme. This was described theoretically by Zhang and Pawliszyn [62] and by Ai [63]. In headspace-spme volatilized analytes are extracted and concentrated in the SPME coating; this can have several advantages over direct SPME extraction in the liquid phase. For analytes that partition strongly into the SPME fiber equilibration times can be reduced sub-