Note 9: Methodologies For the Quantification Of Purge and Trap Thermal Desorption and Direct Thermal Desorption Analyses

INTRODUCTION

This manuscript describes methodologies which are useful for quantitative analyses involving purge and trap thermal desorption (P&T-TD) or direct thermal desorption (DTD) using the SIS Short Path Thermal Desorber and sample isolation accessories (1). Quantification techniques used for the determination of volatile organic compounds in water or wastewater involving P&T-TD have been extensively reviewed. These procedures form the basis of the US-EPA water and wastewater testing program and are outlined in great detail in methods #524 and 624 (2). Since this methodology has been extensively validated, it will not be reviewed in this manuscript. Rather, this discussion will address quantification problems which arise in the analysis of non-routine matrices such as foods, polymers, soils, sludges and other samples.

Analyte recoveries during purge and trap experiments can vary widely due to matrix effects, purging efficiency (volatility), purge cell design, sparging rate, choice of adsorbent, isolation temperature and many other factors. For these reasons, quantification of P&T-TD and DTD experiments are best carried out using various spiked internal standard methodologies which may offer some degree of correction for variable analyte extraction efficiency or recovery. Each individual sample matrix and analyte presents unique problems which must be overcome in order to achieve accurate and precise quantification. Analytical strategies must be formulated on an individual basis and are not always universally applicable. Intensive methods development and validation studies are required for each individual sample matrix and analyte combination to guarantee accuracy and precision in the measurements. In our laboratory, we routinely employ several different internal standard methods to achieve quantification of P& T-TD and DTD analyses. These include the standard addition, internal standard, surrogate internal standard and stable isotope labeled internal standard methods. These methods are described in this manuscript and are illustrated with examples from current research projects.

The first stage in quantitative methods development begins with designation of target compounds or analytes and choice of analysis mode. A feasibility evaluation should be conducted to determine whether or not the chosen analytes are amenable to the techniques of P&T-TD or DTD and if indeed these methodologies are the best way to conduct the investigation. The target compounds must be of sufficient volatility so that they can be efficiently sparged from the matrix and trapped on an adsorbent (P&T-TD) or be delivered directly into the GC injector using the technique DTD. In general, candidates for DTD include solid samples with low moisture contents (< 5%) and relatively high volatile organic contents (.01 - 10% w/w). Sample sizes in the range of 1-500 mg are common in this technique. However, factors such as injector split ratio and GC column loading capacity must be considered in sizing samples. Small samples are often cumbersome and difficult to accurately weigh, but are ideal for analytical separations using narrow diameter capillary columns (0.25-0.32 mm i.d.) with low loading capacities. Some compromise can be achieved by using high split ratios (100:1), which may enable larger samples to be analyzed. Maximum sample loads in the DTD technique are attainable using packed or wide bore capillary column columns, although some loss of resolution may be realized. Moisture contents of DTD samples must be low enough so they do not cause freezing and plugging of the GC column flow when operating in cryofocusing mode. Once again, high split ratios or packed and wide bore column operation will tolerate higher moisture content matrices. The advantages of the DTD technique, when permissible, are numerous. The sample requirement is small and high sensitivity is achieved. In addition, low molecular weight polar species which are not efficiently trapped by Tenax® and other common adsorbents are directly transferred to the GC for analysis and are not lost. Samples analyzed successfully by this technique include polymers, spices, candy, chewing gum, extruded snack foods, soil, wood, pine needles and other natural products. This technique has been described in detail in a previous newsletter article (3). Sample matrices which contain high moisture contents and low concentrations of flavor are best analyzed by the technique P&T-TD. In this technique, the dynamic headspace purge and trap step achieves a concentration factor, while allowing water and other low molecular weight polar species to escape the adsorbent trap. The choice of adsorbent also governs the range of analytes which can be effectively trapped. Tenax-TA resin is the most widely promoted adsorbent with various forms of activated graphitized carbons serving equally well. The particle size distribution of the activated carbons will often effect the breakthrough volumes of various analytes with smaller particles and pore sizes being more efficient for trapping highly volatile low molecular weight species. Glass beads are often used as traps to condense high boiling oligomers outgassed from polymers in high temperature desorption studies. Silica gel is also useful for trapping low molecular weight polar compounds and gasses; however, it has the disadvantage of trapping water, as well. Often, the ideal adsorbent trap may be a mixed bed design with any number and combination of these adsorbents. The next consideration is then the sample size and purge cell design. SIS manufactures a temperature controlled purge and trap apparatus specially designed for solid sample matrices as well as a range of liquid sparging vessels which can be fitted with heating accessories. In some cases, individual users may wish to construct purge and trap vessels of their own particular design. Adapters are readily available from SIS to accommodate these special requirements. In any event, it is often advantageous to set up the particular P&T-TD or DTD experiment and perform a qualitative analysis before attempting quantification. These preliminary experiments can be used to optimize the analysis with regard to sample sizing, isolation or desorption temperatures, sparge gas flow rate, chromatographic conditions and other parameters. Once all of the isolation and analysis parameters have finally been worked out on a qualitative basis, a quantification strategy must then be devised and validated. The following discussion will address specific types of internal standard methods of quantification with examples of how they have been applied in specific research projects.

STABLE ISOTOPE LABELED INTERNAL STANDARD METHODOLOGY

The stable isotope labeled internal standard method is the most accurate and precise method available for quantifying P&T-TD and DTD analyses. In this technique, a stable isotope labeled analogue is obtained for each target compound. These isotope labeled standards are then spiked into the sample matrix, isolated and analyzed. The ratio of analyte to stable isotope incorporated standard is then determined and compared to an off-line calibration or linearity curve to achieve quantification. Since their chemical and physical properties are almost exactly identical to those of the analyte, they are self correcting with respect to extraction efficiency, recovery and detector response. Of course, this method of quantification requires a mass spectrometer as a detector, since the isotope labeled standards may co-elute at the same retention time as the analyte and separation is achieved by mass chromatography or selected ion monitoring. The most common isotopes used for this purpose are 2H (deuterium) and 13 C. Companies  (as MSD Isotopes and Isotech) stock thousands of stable isotope analogues, particularly those of environmental or biochemical interest. Custom synthesis services are also available upon special request. Before using a stable isotope internal standard, they must be analyzed and screened for isotope purity. Only standards with isotope purity greater than 99% are useful and for some trace level determinations purities of 99.99% are required. Next, a linear calibration curve should be constructed using a constant level of internal standard (concentration equal to the spike level in the matrix) and a varying amount of analyte. The variable analyte concentrations must cover a dynamic range sufficient to encompass the levels expected to be present in the actual sample matrices. The curve is normally constructed by plotting the ratio of analyte peak area/stable isotope labeled I.S. peak area versus concentration of analyte. The line equation for this curve can then be used to calculate an analyte concentration from the experimentally observed analyte/I.S. peak area ratio. The primary advantage of this methodology is the ability to achieve excellent accuracy and precision in the quantification. The disadvantages are the need for a mass spectrometer as detector and the high cost often associated with custom synthesis and purification of stable isotope labeled standards.

We routinely apply this method of quantification for conducting measurements of benzene, toluene and xylenes (BTOX analysis) in food products and or environmental samples. In these experiments, we spike d-6 benzene, d-8 toluene and d-10 xylene into the sample at a level approximating the concentrations of proteo analytes in the matrix. We prepare the spiking standards in methanol so that the desired spike level can be delivered to the product in a volume of 1-10 ul. Methanol is an ideal solvent for spiking standards used in P&T-TD and DTD experiments, since it is not retained by Tenax or graphitized carbon adsorbents. Therefore, the methanol passes directly through the adsorbent trap during the purge and trap step and does not cause a large peak to appear in the chromatogram which may often obscure data. The spike should be quantitatively delivered to the matrix once sealed in the purge apparatus using solvent flush technique; however, the volume of methanol used to deliver the spike should not exceed 25 ul. We have observed decreased recoveries of volatiles in purge and trap experiments when large volumes (1.0 ml or greater) of low molecular weight polar solvents are present in the matrix during isolation. Presumably, the large volume of solvent passing through the adsorbent can actually cause premature elution of components, especially the highly volatile species.

Figure 1 - GC-MS Ion Current and Mass Chromatogram of Cracker Sample Spiked With Deuterated Internal Standards For Measurement of Benzene, Toluene and Total Xylene Concentrations (BTOX Analysis)

The results of a typical P&T-TD analysis using stable isotope spiked internal standards is shown in Figure #1. This is a mass chromatogram derived from a store bought cracker product in which we were conducting a BTOX analysis. The ion masses plotted are m/z 84, 78, 100, 92 and 106 corresponding to the molecular ions of d-6 benzene, benzene, d-8 toluene, toluene and o, m, p-xylenes (no d-10 xylenes were used in this analysis). The lower trace is integrated ion current and m/z 91 corresponding to tropylium ion is also displayed. The exact spike level of d-6 benzene and d-8 toluene in this experiment was 13.2 and 14.4 ppb w/w, respectively. This analysis was conducted by P&T isolation from 10 grams of cracker crumbs using the SIS sample collection oven at 60 degrees C with Tenax/Carbotrap adsorbent with 30 minutes of sparging using nitrogen at a rate of 40 ml/minute. The adsorbent trap was then thermally desorbed at 250 degrees C for 5 minutes with subsequent analysis by GC-MS. The peak eluting just before 12 minutes yielding the large response in the m/z 84 channel is d-6 benzene. This peak is immediately followed by benzene which shows up in m/z channel 78. Similarly, d-8 toluene elutes at r.t. 16:30 at m/z 100 and is immediately followed by toluene at m/z 92. As can be seen from the data, modern analytical GC capillary columns actually achieve some separation of deutero versus proteo compounds with the deutero analogue eluting first. The ratios of deutero/proteo compounds are then used to calculate the exact concentration of each individual analyte. In this particular instance, the benzene and toluene levels in the crackers were determined to be 125 and 52 ppb, respectively. The levels of total xylenes were estimated to be 30 ppb using d-8 toluene as reference.

SURROGATE INTERNAL STANDARD METHODOLOGY

A surrogate internal standard is a matrix spiked internal standard which has chemical and physical relatedness to the analytes of interest so that extraction efficiency, recovery and detector response characteristics are similar. The surrogate standard must be a compound not normally present in the matrix. For instance, standard EPA methodology utilizes surrogate standards in its water and wastewater analysis protocols (2). They make use of fluorinated compounds which are not normally found in nature or as environmental contaminants. The EPA surrogates are available as neutral, acidic, phenolic and basic species to simulate partition behavior in conventional pH controlled extraction protocols or in purge and trap experiments. The advantages of this type of quantification are that surrogates can usually be obtained in high purities and low cost. They are useful for GC as well as GC-MS studies since the surrogates can be separated on the basis of retention time from the target compounds. Additionally, surrogates are user selectable from a wide range of chemical and physical properties which cater to specific applications. For instance, we have recently quantified vanillin and other phenolic compounds in vanilla beans by DTD using 2, 6-dimethoxyphenol as a surrogate internal standard (4). Vanillin is 4-hydroxy-3-methoxybenzaldehyde and many of the character flavor compounds of vanilla have similar phenolic functionality, so 2, 6-dimethoxyphenol was considered a good surrogate for purposes of quantification. In these analyses, vanillin samples were placed into preconditioned desorption tubes containing a 2 cm bed volume of Tenax-TA adsorbent and were then spiked with 1000 ppm of I.S.. The Tenax-TA adsorbent bed is utilized for conducting DTD analyses so that the I.S. spike and sample volatiles are not lost in the initial purge step (used to remove air) prior to thermal desorption. We constructed a calibration curve by spiking blank desorption tubes using a constant 10 ug of 2, 6-dimethoxyphenol I.S. and a varying level of vanillin (log and half-log serial dilutions from 1000-1.0 ug). A linear calibration curve was obtained with a 0.999 correlation coefficient. Excellent analytical precision was obtained. The ratio from the midpoint of the calibration curve was 0.81 (mean, n=5) with a standard deviation of 0.02 and a 2.47 coefficient of variation. Indeed, this level of precision is equal to that obtained by conventional direct injection into the GC using a syringe. Results such as these are not unusual for DTD analyses, since the technique is akin to direct injection except that the desorption tube essentially becomes a sacrificial injection port. Volatile and semivolatile components of the sample are outgassed and delivered directly into the GC injector, with nonvolatile residues remaining in the desorption tube. Figure #2 is a typical ion current chromatogram of a vanilla bean sample analyzed by DTD-GC-MS showing the I.S. , 2, 6-dimethoxyphenol, vanillin and numerous other flavor compounds. Figure #3 is the calibration curve used for the vanillin determinations. The curve was of sufficient dynamic range to encompass the vanillin concentrations detected in all samples analyzed. Vanillin contents of 1.94% observed by DTD-GC-FID and DTD-GC-MS analyses were found to be in excellent agreement with the value of 2.0% determined using traditional (but much more time consuming) extraction, isolation and HPLC analysis.

Figure 2 - GC-MS Ion Current and Mass Chromatogram of Vanilla Bean Sample Analyzed By DTD-GC-MS, Showing Internal Standard, Vanillin and Numerous Other Flavor Compounds

Figure 3 - Calibration Curve Generated For Quantification of Vanillin In Vanilla Beans Ising DTD-GC-MS With 2,6-Dimethoxyphenol Internal Standard

INTERNAL STANDARD METHODOLOGY

The internal standard method of quantification is useful for P&T-TD and DTD analyses, although it is somewhat less accurate than stable isotope or surrogate standard strategies. The technique is similar to those previously described with the exception that the internal standard is not incorporated into the matrix during isolation. Rather, it is spiked into the adsorbent traps, after a sample has been isolated or perhaps is injected directly into the GC at the time of desorption. Therefore, no correction for extraction efficiency or recovery is achieved using this technique. However, it serves as a useful means of quantifying the levels of components present on the adsorbent traps. As in the previous examples, it is necessary to calculate detector response factors or calibration curves for each I.S. relative to the analytes of interest. Many times this is not possible, since analytical standards may not be available for the target compounds. For instance, in flavor investigations it is not unusual to generate chromatograms containing several hundred compounds. It is impractical or impossible to obtain standards and determine response factors for each and every compound observed. Often, it suffices to obtain a semi-quantitative estimate of the concentration for each compound by simple peak area comparisons assuming a response factor equal to one. We use this method to estimate the levels of outgassing products from polymers in our polymer thermal desorption studies. Figure #4 is a GC-MS ion current and mass chromatogram derived from a typical DTD polymer desorption analysis using d-10 anthracene as an internal standard. In this experiment, approximately 500 mg of silicone rubber tubing was placed into a desorption tube, spiked with 0.5 ppm of d-10 anthracene and subjected to DTD analysis at 250 degrees C for 10 minutes. The lower trace is the integrated ion current profile, while the upper trace is a mass chromatogram for m/z 188 (molecular ion of d-10 anthracene). Semi-quantitative estimates for the levels of thermally desorbed products were obtained by simple peak area comparisons to that of the d-10 internal standard. Compounds identified included processing solvents, cross linking agents, plasticizers, antioxidants, polymerization accelerators and dimethylpolysiloxane oligomers. The DTD analysis employed was compared to solvent extraction methodology outlined by the FDA in the Code of Federal Regulations manual. The DTD analyses consistently yielded higher recoveries and less artifact formation than did the standard FDA methods. For instance, many of the acids observed in the DTD analyses were detected as ethyl esters in the FDA protocol, since the extraction regime used ethanol as solvent. Also, the ethanol failed to efficiently extract the polysiloxane oligomer components since they are very hydrophobic. Characterization by ethanol extraction alone would have missed these important potential migrants.

Figure 4 - GC-MS Ion Current and Mass Chromatogram For a Polymer Desorption Investigation Analyzed By DTD-GC-MS Using D-10 Anthracene Internal Standard. Upper Trace is Mass Chromatogram For Ion Mass 188, Which is the Molecular Ion of D-10 anthcene.

We have recently used internal standard methodology to quantify breath analysis data in human subjects ingesting raw and dehydrated garlic preparations (5). In this investigation, adsorbent traps were spiked with 100 ng of 2, 5-dimethylthiophene to serve as I.S. and subjects were required to exhale one liter of breath through the adsorbent traps at various time intervals after ingesting garlic. Preliminary qualitative breath analysis studies using GC-MS indicated several characteristic sulfur containing allylic compounds to be present on the breath. We then chose 2, 5-dimethylthiophene to serve as I.S. and constructed calibration curves for each specific analyte. This was accomplished by spiking the adsorbent traps with internal standard and various concentrations of target compounds. The traps were then analyzed by thermal desorption as usual and the calibration data was derived from peak area integrations. Figure #5 is a typical concentration curve obtained from diallylsulfide relative to 2, 5-dimethylthiophene I.S.. The curve is linear with a dynamic range covering four orders of magnitude and has a correlation coefficient of 1.0. The curve was constructed by plotting the mean values of five replicate determinations at five different concentrations. Excellent analytical precision was obtained. Typical ratios from the low, middle and high ends of the calibration were 0.059 +/- 0.003, 6.45 +/- 0.03 and 12.75 +/- 0.35 (n=5), respectively. Concentrations of garlic-borne phytochemicals on the breath of subjects were then calculated using the line equations from the calibration curves for each individual compound. The ideal gas law was used to convert the data into ppm v/v. Figure #6 shows the breath elimination curves from a subject after ingesting raw garlic. As was stated previously, this particular internal standard approach at quantification is considered to be accurate in terms of calculating the actual concentrations of various compounds on the Tenax adsorbent traps with no provision or correction for extraction efficiency or recovery. However, we could not conceive of any way to provide for matrix spiking of the breath. We are confident that the adsorbent traps were quantitative in the ability to trap the garlic-borne phytochemicals from the breath, since no odor was evident at the outlet of the traps.

Figure 5 - Calibration Curve For Diallylsulfide Relative to 2,5-Dimethylthiophene Internal Standard. Data Was Produced By Spiking Adsorbent Traps Followed By Thermal Desorption GC-FID Analysis.

Figure 6 - Elimination Curves of Garlic-Borne Allylic Compounds On the Breath of Human Subjects Following Ingestion of Raw Garlic

STANDARD ADDITION METHOD OF QUANTIFICATION

The standard addition method of quantification is a useful means of quantifying P&T-TD analyses. In this technique, the matrix to be analyzed is split into approximately five portions and four of the sample aliquots are spiked with increasing levels (concentrations greater than those expected in the sample) of the target compounds. One portion of the sample is left unspiked. All samples are then subject to P&T-TD analysis. The goal is to form a linear detector response concentration curve for the four high level spiked samples. The linear curve is then extrapolated down to the low concentration level to obtain quantification of the target compounds in the unspiked sample which must be present at low levels for the principal to be valid. If the target compound concentration in the unspiked sample is sufficiently low relative to the spiked samples, then the error in the measurement will be low. However, if high levels of the analytes are already present in the matrix, the error in such extrapolations may be high and would require mathematical correction. In this case, an alternative would be to secure control matrices of the sample being tested which do not contain the target species and use these for spiking purposes. We routinely conduct off-odor/flavor investigations on food products which are tainted with solvents from floor finishing preparations, paints, pesticides and from numerous other sources. We have found that the best way to quantify such contamination levels is to actually spike the offending species (if available) into a control (not contaminated) product at various levels to develop a concentration curve. The suspect samples are then analyzed and the level of contamination is determined from this curve. In a cracker sample contaminated with mineral spirits from a warehouse painting operation, we observed a cluster of approximately 50 hydrocarbon peaks in its chromatogram when analyzed by P&T-TD. We simply summed the area integration of the contaminating species to obtain a total. Next, we spiked control crackers with 0.1, 1.0, 10.0 and 100.0 ppm of the mineral spirits then analyzed the spiked samples by P&T-TD using the exact same conditions as the suspect product. We then sum the area integrations for the mineral spirit peaks in each of the spiked samples and plot these values versus spike level in ppm to obtain a calibration curve. The level of contamination in the suspect product is then readily calculated from this curve. This methodology is very accurate since the extraction efficiency, recovery and detector response need not be quantitative to achieve useful data. The disadvantages are that many samples need to be analyzed and analytes or target compounds must be secured. Furthermore, for best results control matrices must be available for the spiking experiments.

SUMMARY

In summary, quantitative measurements by P&T-TD and DTD are best accomplished by first conducting qualitative measurements to optimize sample sizing, isolation, chromatographic and detection parameters. Next, a decision must be made regarding which technique of quantification is most suitable for the particular analysis. A quantitative analytical strategy is then devised and implemented. Methods development is performed to work out details such as spiking and recovery, building calibration curves, adjusting dynamic range and evaluating accuracy and precision. Finally, the methodology is challenged and validated using authentic or fortified samples.

ACKNOWLEDGMENT

We acknowledge the Center for Advanced Food Technology (CAFT) Mass Spectrometry Lab facility for providing instrumentation support for these projects. CAFT is an initiative of the New Jersey Commission of Science and Technology. We also thank Scientific Instrument Services, Inc. of Ringoes, New Jersey for donation of the Short Path Thermal Desorption instrument used in these studies. This is NJAES publication #D-10569-3-91.

REFERENCES

Hartman, T.G., Karmas, K., Chen, J., Shevade, A., Deagro, M. and Hwang, H. -I.; Determination of Vanillin, Other Phenolic Compounds and Flavors in Vanilla Beans by Direct Thermal Desorption - Gas Chromatography and Gas Chromatography - Mass Spectrometric Analysis, In ÒPhenolic Compounds in Food and HealthÓ, ACS Symposium Series, chapter 3, Eds. C.-T Ho and M. -T. Huang, In Press (1991).

Hartman, T.G., Lech, J. and Rosen, R.T.; Breath Analysis of Garlic-Borne Phytochemicals in Human Subjects by Combined Adsorbent Trapping, Short Path Thermal Desorption GC-MS, Proceedings of the 39th ASMS Conference on Mass Spectrometry and Allied Topics, pg. 641 (1991).

Manura, J.J.; Direct Thermal Analysis Using the Short Path Thermal Desorption System, Mass Spec Source, XIV, No. 2 (1991).

Manura, J.J., Overton, S.V., Baker, C.W. and Manos, J.N.; Design and Operation of the Short Path Thermal Desorption System, Mass Spec Source, XIII, No. 4 (1990).

Methods for the Determination of Organic Compounds in Drinking Water, Environmental Monitoring Systems Laboratory Office of Research and Development, US-EPA, Cincinnati, Ohio 45268 (1988).