3. ON-LINE SOLID-PHASE EXTRACTION–GAS
3.2. Optimization of Solid-Phase Extraction–Gas
3.2.1. Stationary Phase for Trace Enrichment
In SPE–GC, the SPE part of the procedure is in most cases only used to effect trace enrichment and no attempt is made to improve selectivity (cf. below). Con- sequently, in actual practice only two types of sorbent are used, hydrophobic C18-bonded silicas and highly hydrophobic copolymers. According to abundant literature information, breakthrough volumes of nonpolar and medium-polar ana- lytes, and even many polar analytes, on SPE cartridges packed with these sor- bents, and with dimensions of 10 to 20 mm length⫻1 to 4.6 mm ID, are much higher than 10 ml. Polystyrene copolymers typically provide 20- to 30-fold more retention than the alkyl-bonded silicas. In other words, it is safe to state that the sorption step of the SPE process will not cause a noticeable loss of analytes. In addition, because of the reliable information already available in the literature, there is no need for the analytical chemist tackling a new problem to start with a rather time-consuming collection of experimentally determined breakthrough data. For a more detailed discussion, the reader is referred to texts such as that by Barcelo´ and Hennion [32].
In recent years, cartridge holders containing one or a few small (diameter 3 to 4.6 mm, thickness 0.5 mm) membrane extraction disks have been recom- mended as an alternative to conventional precolumns or cartridges for both LC and GC [33,34]. The disks contain approximately 90 wt% of a hydrophobic sor- bent held in a polytetrafluoroethylene (PTFE) mesh, and can be loaded at fairly high speed. Drying with nitrogen at room temperature proceeds rapidly. The com- mercial Empore (3M, St. Paul, MN) extraction disks, which have a diameter of 47 mm, are often recommended for field studies. It will be clear that a single 47-mm ID disk—after having been used for field sampling—can be used for a large number of 4.6-mm-diameter-based SPE–GC analyses!
Abundant experimental evidence shows that analyte desorption from the loaded cartridges or disks (an aspect that is equally important for the final recov- ery to be obtained as is the sorption step), is easily achieved with less than 100 àl of methyl or ethyl acetate, applied at a flow rate of 50 to 100àl/min. This is
168 Hankemeier and Brinkman
well within the conditions of ordinary LVI procedures and does not present any technical problems.
So far, hydrophobic sorbents have been used in essentially all SPE–GC procedures. One main reason why the additional selectivity provided by modified SPE sorbents, which is frequently studied in SPE–LC, is less popular in GC- based analyses, no doubt is the much higher efficiency of conventional GC com- pared with LC separations. Still, two studies have been devoted to the on-line combination of immunoaffinity SPE (IASPE) and GC [35,36]. In IASPE, desorp- tion from an antibody-loaded precolumn has to be carried out with, typically, several milliliters of methanol–water (95:5, v/v). Since it is impossible to intro- duce such a solution directly into the GC part of the system, it is on-line diluted with an excess of HPLC-grade water and the mixture led through a conventional C18-bonded silica precolumn as trapping column, as in RPLC–GC. The gain in breakthrough volume of the analytes, due to increased retention caused by the considerably decreased modifier percentage, easily outweighs the volume in- crease. Consequently, the analytes are quantitatively trapped on this second pre- column. Desorption from this trapping column and the further procedure are as for conventional SPE–GC. The method was applied to steroid hormones in 5 to 25 ml of urine. The detection limit of 19-β-nortestosterone was about 0.1 àg/L (FID detection). A similar approach, which combined an antibody-loaded first, and a copolymer-packed second precolumn was used for the determination of triazines in river water, wastewater, and orange juice. The detection limits were about 10 ng/L when 10-ml samples were analyzed using nitrogen–phosphorus detector (NPD) detection. It is interesting to add that, although IASPE–GC has not yet been combined with MS detection, setting up such a system is not ex- pected to cause any technical problems. We shall then have an instrumental setup that permits highly selective analyte isolationand structure-based identification or, in other words, identification and confirmation, in one run.
3.2.2. Removal of Water by Drying
Problems caused by the presence of water in the retention gap can be overcome by inserting a drying cartridge or drying the SPE cartridge for about 15 to 30 minutes with nitrogen gas (at ambient temperature). Nitrogen drying has the ad- vantage that it is a well-known and simple procedure and is, therefore, generally preferred. Volatile compounds such as chlorobenzene are not lost to a significant extent [37]. Drying of copolymers is distinctly more rapid, and has a somewhat more reliable outcome than that of C18-bonded silicas. The insertion of a drying cartridge containing silica or sodium sulfate between the SPE cartridge and the GC part of a system is an interesting alternative to reduce the drying time. Both silica and sodium sulfate can be reused many times if they are regenerated be-
On-Line Sample Preparation for Water Analysis 169
tween runs by (external electrical) heating. Since analyte losses appear to be negligible for a wide variety of compounds, e.g., triazines, alkylbenzenes, chloro- benzenes, and chlorophenols, even at the trace level, the use of a drying cartridge is a viable approach [38–40]. Recently, the design of the drying cartridge was improved to enable higher temperatures during regeneration, and volatile analytes up to tetrachloroethylene could be included (methyl acetate as desorption solvent) [41]. Although molecular sieves and sodium sulfate have a higher drying capac- ity, silica was found to be the best choice in actual practice.
3.2.3. Solid-Phase Extraction to Gas Chromatography Transfer
When using the SPE cartridges described above, the analytes are generally trans- ferred with 50 to 100àl of organic solvent (preferably ethyl or methyl acetate) from the cartridge to the GC. This volume is required to desorb the analytes and to prevent memory effects due to adsorption of analytes in the transfer capillary.
If relatively volatile analytes are included in the set of target compounds that has to be determined, on-column interfacing is the preferred technique if the sample extract is not too dirty [42,43]. The main problem is that with highly contaminated samples, such as wastewater, the retention gap easily looses its performance: distorted peak shapes and/or lower analyte responses for, espe- cially, the more polar analytes, can already show up after a few GC runs. To maintain the quality of the analyses, the retention gap should be heated prior to starting the temperature program of the analytical column. This can be done by putting the retention gap in a separate GC oven [44,45] or by wrapping it with heating wire [46]. Recently, the retention gap was placed in a low-weight oven in the GC oven itself and no loss of performance was observed after more than 200 on-line SPE–GC–MS analyses of river water samples [47].
Recent studies recommend the use of a PTV injector in on-line SPE–GC of highly contaminated samples [48]. Exchange of the packed liner of such an injector is straightforward and takes little time. The main drawback is that the separation of volatile analytes from the (desorption) solvent is less satisfactory than with a retention gap [43]. Staniewski et al. used a PTV injector as interface in on-line SPE–GC and 50àl of ethyl acetate for desorption [49]. Several herbi- cides were determined in water at about the 1-àg/L level with recoveries of 20 to 90%.
If volatile analytes do not play a prominent role in the samples to be ana- lyzed, a loop-type interface—which is rather easy to use—can be recommended [50,51]. In order to somewhat increase the application range at the volatile end, Noij et al. [50] desorbed the analytes with 500àl of methyltert-butyl ether–
ethyl acetate (90:10, v/v), and the eluate was injected together with 50àl ofn-
170 Hankemeier and Brinkman
decane as cosolvent into the GC. The most volatile analyte included in the study was mevinphos. Another approach to extend the application range of loop-type injections is to use an Autoloop interface (Interchro, Bad Kreuznach, Germany), which essentially consists of an SPE cartridge and two loops for storing organic solvent, which are mounted on a 14-port valve [52–55]. Solvent transfer for de- sorption to the gas chromatograph is achieved by the carrier gas, which can be diverted via a 6-port valve. The solvent in the first loop effects phase swelling of the retaining precolumn to increase the application range to more volatile ana- lytes. The solvent in the second loop is used to desorb the analytes and transfer them to the retention gap. Because a rather long retention gap and a total transfer volume of about 200àl are used, a transfer temperature of 90°C can be used, which is lower than that usually applied with ethyl acetate.
As has repeatedly been mentioned, one of the incentives of on-line SPE–
GC is that the total analyte-containing fraction is transferred to the GC column.
There are, however, a few studies in which this was not done. For example, Ballesteros et al. [56,57] desorbed the analytes from a copolymer SPE cartridge with 100àl of ethyl acetate and, after homogenization of the extract by means of a mixing coil, injected 5àl of the eluent via a loop and a capillary in a splitless injector. When 50-ml water samples were analyzed by GC–FID, detection limits of 0.7 to 1àg/L were achieved forN-methylcarbamates and their phenolic degra- dation products. The procedure is elegant, but one should keep in mind that using a (small) aliquot of the total sample will, in most instances, cause a considerable loss of detectability expressed in concentration units.
Finally, analyte desorption and, consequently, SPE-to-GC transfer, can also be performed without any organic solvent being used, that is, by thermal desorp- tion (TD). The two options are SPETD, discussed later in this chapter, and SPME, discussed in Chapter 8.
3.2.4. At-Line Operation
Sample enrichment by SPE or LLE and GC analysis can also be integrated into one setup by at-line coupling: the sample extract is transferred from the sample preparation module to the gas chromatograph via, e.g., an autosampler vial using an ASPEC (Gilson, Villiers-le-Bel, France) or a PrepStation (Hewlett-Packard, Palo Alto, CA). The main disadvantage of most of the published procedures is that, after elution of the SPE cartridge and collection of the solvent in a vial, only an aliquot is injected. In some studies, such as those on organochlorines and pyrethroids in surface water [58,59] and benzodiazepines in plasma [60], 100 to 200àl were injected out of 2 to 5 ml. In another paper, only 1àl out of 1 ml was injected to determine barbiturates in urine [61]. Not surprisingly, detec- tion limits in the latter PrepStation–GC–MS study were in the sub- to low-mg/
L range. Similar sensitivity problems were encountered [62] when serum had to
On-Line Sample Preparation for Water Analysis 171
be analyzed and, by other workers, in a PrepStation-based study on the determina- tion of organic acids [63].
In view of the above, it is interesting to look at another study [64,65], which was directed at improving the performance of the PrepStation–GC setup by intro- ducing several modifications, viz., (1) increasing the aqueous sample volume from 1.5 to 50 ml, (2) using 50-àl ‘‘at-once’’ on-column LVI rather than 1-àl injections, and (3) decreasing the desorption volume to 300àl by reducing the amount of sorbent in the SPE cartridge. The redesigning was markedly success- ful: an overall 300-fold improvement had been calculated, and a 150- to 300- fold improvement was observed in actual practice. An initial problem was that, as a result of the increased sensitivity, the PrepStation was found to be less inert than expected: several interferences from impurities extracted from the septa and also the commercial cartridges showed up. A cartridge made from stainless steel and polychlorotrifluoroethylene and a 2-needle system were constructed to elimi- nate these interferences. Several micropollutants were detected in 50 ml of (unfil- tered) river water at the 0.2 to 400 ng/L level using full-scan MS acquisition (Fig. 4).
In summary, with carefully designed at-line sample preparation–GC sys- tems, analyte detectability can be made similar to that in on-line SPE–GC set- ups. However, interferences due to contamination and analyte losses will always be more serious.
3.2.5. Medium-polar and Nonpolar Analytes
Most conventional GC and also LVI–GC and SPE–GC procedures have been designed for the determination of the medium-polar compound range, which com- prises analytes from, typically, simazine andN,N-dimethylaniline to trichloroben- zene and dibutylphthalate. With such sets of compounds, it is highly unlikely that any sort of technical problems will be encountered. If the range of analytes has to be extended to include really nonpolar compounds, such as organochlorine pesticides, ethion, or bromophos-ethyl, one should add 20 to 30 vol% of methanol to the aqueous sample to prevent adsorption of these analytes to the inner walls of capillaries and valves [66,67]. It will be clear that, in the end, a situation may arise in which too wide a range of analytes has to be determined. One can then of course compromise with regard to the recoveries on either the polar or nonpolar end. However, for a more robust operation, it is recommended to carry out two separate runs, with conditions optimized for the former (purely aqueous sample) and the latter (modifier addition) group, respectively. For example, the addition of 30 vol% of methanol to surface water samples did not interfere with the deter- mination of a set of nonpolar and medium-polar pesticides. However, the most polar analyte in the test set, dimethoate, was largely lost [66].
172 Hankemeier and Brinkman
Figure 4 Full-scan PrepStation–GC–MS of 50 ml of Meuse River water (B) without and (A) with spiking with 37 micropollutants at the 0.18àg/L level; 100àl out of the 300àl extract were injected. The inserts show the reconstructed-ion chromatograms of two characteristic masses of 1,3-dichlorobenzene (C), 2-methylthiobenzothiazole (D), and Musk G and T (E). Compounds detected (0.2 to 430 ng/l) in the nonspiked river water:
2, 1,3-dichlorobenzene; 2′, 1,4-dichlorobenzene; 2″, 1,2-dichlorobenzene; 3, acetophe- none; 4, decamethyl-cyclopentasiloxane; 6, naphthalene; 10, isoquinoline; 11, 2-methyl- quinoline); 12, 2,4,7,9-tetramethyl-5-decyne-4,7-diol; 13, dibenzofuran; 14, triisobutyl phosphate; 15, N,N′-diethyl-3-methylbenzamide; 16, 2,2,4-trimethylpentane-1,3-dioldi- isobutyrate; 17, diethyl phthalate; 18, 2-methylthiobenzothiazole; 20, tetraacetylethylene- diamine; 21, tributyl phosphate; 22, ethyl citrate; 23, desethylatrazine; 25, simazine; 26, atrazine; 27, tris(2-chloroethyl)phosphate; 28, N-butylbenzenesulfonamide; 30, tris(2- chloroisopropyl)phosphate; 31, Musk G; 33, Musk T. (From Ref. 65.)
On-Line Sample Preparation for Water Analysis 173
3.2.6. Volatile Analytes
Recently, our understanding of the processes involved in the on-column LVI–
GC [68–71] and on-line SPE–GC [37] analysis of volatile compounds has im- proved. Due to the pressure drop along the solvent film in the retention gap, which occurs when an SVE is used, solvent evaporation takes place not only at the rear end, but also along the whole length, and even at the front of the solvent film. Loss of volatiles can therefore be severe, especially in SPE–GC, because these analytes are mainly present in the front part of the desorption solvent. To improve performance, after conventional sample loading and drying with nitro- gen, one should introduce some pure organic solvent, the so-called presolvent, into the retention gap prior to the actual desorption (using the lower-boiling methyl rather than ethyl acetate). A solvent film will now have been formed before the (volatile) analytes arrive as a result of the SPE-to-GC transfer, and will ensure their retention also during evaporation of the solvent film. The bene- ficial effect is vividly illustrated in Table 4: with some 30àl of presolvent, ana-
Table 4 Dependence of Analyte Recoveries of On-Line SPE–GC Transfer on Amount of Presolventa
Recoveries (%) for a presolvent volume of
Compound 0àl 10àl 20àl 30àl
Volatile
Monochlorobenzene 5 7 70 97
p/m-Xylene 7 8 72 95
Styrene 22 34 86 100
o-Xylene 9 14 85 99
Methoxybenzene 40 66 95 100
o-Chlorotoluene 33 54 94 96
Semi-volatile
Benzaldehyde 70 102 100 103
1,2-Dichlorobenzene 62 89 95 97
Indene 64 94 96 101
Nitrobenzene 88 100 95 100
Naphthalene 89 96 95 99
High-boiling
Methylnaphthalene 95 94 95 96
Acenaphthene 99 100 99 101
Metolachlor 99 100 102 99
aPure methyl acetate introduced as presolvent into the GC prior to desorption with 50àl methyl acetate [37].
bRecovery values above 80% are shown in bold print.
174 Hankemeier and Brinkman
lytes as volatile as monochlorobenzene are quantitatively recovered even at the 0.5àg/L level [37]. Recently, it was found that satisfactory recoveries (70 to 90%) can be obtained for volatile analytes such as monochlorobenzene and the xylenes with a mere 20 to 25 àl of methyl acetate even without the use of a presolvent. Preconditions are the use of an 0.53 mm ID retention gap and closure of the SVE during SPE-to-GC analyte transfer [47].
3.2.7. Self-Controlled Setup
When using an on-column interface for the SPE-to-GC transfer, until recently, the injection speed and the timing of the start of the transfer and the SVE closure had to be determined prior to analysis. It has now been demonstrated that the SVE closure can also be performed in an automated fashion [68]. As an example, Figure 5 shows the carrier gas (helium) and solvent vapor flow (FID response) profiles during a 54-àl on-column injection of ethyl acetate into a 0.53-mm ID retention gap. At the start of the injection, the helium flow sharply decreases and the solvent flow sharply increases, while at the end of the evaporation process the helium flow sharply increases and the solvent flow decreases. The sharp in-
Figure 5 Helium flow rate and solvent peak profile for injections of ethyl acetate into a 0.53-mm ID retention gap. Injection time, 20 sec; injection speed, 160àl/min; evapora- tion rate, 130àl/min; 10àl were left as solvent film in the retention gap at end of injection.
Helium flow is measured by means of a flow meter in the carrier gas tubing, and the solvent flow by means of an FID at the end of the retention gap. (From Ref. 68.)
On-Line Sample Preparation for Water Analysis 175
crease of the carrier gas flow (or, better, of its first derivative) observed when evaporation is complete, can be used to trigger SVE closure. For 30-àl injections ofn-alkanes (C7to C20) inn-hexane under partially concurrent solvent evapora- tion (PCSE) conditions, no analytes were lost with the automated procedure when compared with (the conventional) closing of the SVE 0.1 minute prior to comple- tion of the evaporation.
Automation not only makes preoptimization superfluous, but also improves the robustness of the SPE–GC procedure (and any LVI–GC procedure in gen- eral): if the evaporation time slightly changes due to, e.g., small changes of the injection speed or injection volume, the SVE will still be closed just in time without undue loss of volatiles or a significant change of the solvent peak width at the detector. The latter aspect is important when working with a mass-selective detector, because the delay time for switching on the filament can now be kept constant. As additional advantages, the repeatability of the retention times of volatile analytes is improved [47] and the capacity of the retention gap is signifi- cantly larger when the SVE is closed at the last possible moment [71].
As regards desorption of the SPE, the introduction of the sorption solvent into the retention gap will cause a similar decrease of the helium flow as depicted in Figure 5 for the start of the sample introduction. This allows the introduction of the desired volume of desorption solvent without the need of (re)assessing the proper timing of the start of the introduction when using, e.g., a new type of SPE cartridge. The transfer is stopped after the preprogrammed delay time by switch- ing the transfer valve [47].
The final parameter to be optimized in SPE–GC using an on-column inter- face is the injection speed. As regards this aspect, recent research [47] has shown that (1) with an optimized desorption-plus-transfer-line-flushing strategy only 20 to 25àl of methyl acetate are required per run, and (2) the closure of the SVE at the very end of the evaporation process considerably increases the capacity of the retention gap (cf. above). Both improvements allow significant reduction of the amount of solvent evaporated during injection without using too long a reten- tion gap. This implies that the injection speed will, in any case, be higher than the evaporation rate—which makes injection speed optimization superfluous!
There is now no parameter left that has to be optimized when exchanging the retention gap or the SPE cartridge. In other words, one has a truly ‘‘self- controlled system.’’