- Email: [email protected]
Headspace gas analysis liquid desorption of headspace volatiles trapped on activated carbon open tubular traps
Journal of Chromatography, 402 (1987) 95403 Elsevier Science Publishers B.V., Amsterdam -
Printed in The Netherlands
CHROM. 19 607
HEADSPACE
GAS ANALYSIS
LIQUID DESORPTION OF HEADSPACE VOLATILES VATED CARBON OPEN TUBULAR TRAPS*
TRAPPED ON ACTI-
B. V. BURGER* and ZENDA MUNRO Laboratory for Ecological Chemistry, University of Stellenbosch. Stellenbosch 7600 (South Africa) (Received March 24th, 1987)
SUMMARY
The liquid desorption of headspace volatiles from Grob-Habich activated carbon open tubular traps with small volumes of solvent and capillary column gas chromatographic analysis of the resulting solutions in an on-column type of procedure, was found to be an excellent alternative to thermal desorption which, in the case of thermally labile compounds, may lead to the formation of artefacts. The influence of the solvent volume and its flow-rate through the traps on the efficiency with which the adsorbed material is desorbed and transported to the capillary column, were investigated. It was found that practically quantitative recovery of trapped terpenes could be achieved with 5 ~1 of methylene chloride at a flow-rate of 0.5 &s. The efficiency of liquid desorption and the simplicity of the on-column liquid desorption technique was illustrated in a headspace gas determination of an imitation fruit drink.
INTRODUCTION
As an extension of the pioneering work of Grob and Habich’ on the development of short open tubular capillary traps for headspace gas analysis, we have recently introduced long fused silica traps for applications requiring quantitative trapping of highly volatile componenW. Although the instrumentation required for the implementation of the latter technique is relatively simple, finding and assembling the few pieces of equipment may not be equally simple in all laboratories. In comparison, the method of Grob and Habich is straightforward, requiring only the short capillary trap and a piece of shrinkable PTFE. The analysis, furthermore, requires no special manipulative skills. It is therefo,re not unlikely that the Grab-Habich method will be preferred, especially in the exploratory stages of the development of a headspace-analytical method for a particular analytical problem, the Burger-Munro procedure only being invoked in cases where highly volatile and thermally labile
* Dedicated to the memory of Professor Kurt Grob. 0021-9673/87/$03.50
@
1987 Elsevier Science Publishers B.V.
96
B. V. BURGER,
Z. MUNRO
compounds, for example, have to be quantitatively trapped, or where sampling has to be carried out over extended periods of time. A serious drawback of activated carbon as adsorbent in short capillary traps, appears to be its catalytic activity. It is possible to overcome this problem to a certain extent by subjecting the carbon open tubular trap (COT) to temperature-programmed desorption, using an initial injector temperature at which the adsorbed material does not undergo any catalytic reaction, and a final temperature just high enough to effect quantitative desorption. However, since temperature-programmed desorption is essentially a much slower process than isothermal desorption at a high temperature, it will only produce acceptably sharp gas chromatographic (GC) peaks if the desorbed volatiles can be thermally focussed on the capillary column. During a series of courses on capillary chromatography recently held in our laboratory, Prof. Kurt Grob suggested that whereas thermal desorption produces satisfactory results in work with film traps (FTs) and should therefore be reserved for this purpose, COTS may be more amenable to liquid desorption procedures, and he prompted us to investigate this method in more detail 3. In this paper the results are given of experiments carried out to establish general guidelines for the liquid desorption of material from Grob-Habich COTS. EXPERIMENTAL
GC analyses were carried out with a Carlo Erba Fractovap 4160 gas chromatograph (Carlo Erba, Milan, Italy) equipped with a flame ionization detector, using helium as carrier gas. A glass capillary column (30 m x 0.3 mm I.D.), coated with OV-61-OH and having a l-m retention gap was used for analytical separations. Quantitative determinations of the material desorbed from the COT were carried out by splitless injection of solutions of the compounds under investigation, at concentrations of 0.5 mg/ml in methylene chloride (Merck, residue analysis grade). Quantitation was done with a Hewlett-Packard 3385A Lab Automation System (Hewlett-Packard, Palo Alto, CA, U.S.A.). The COTS were prepared by melting activated carbon particles (10-35 pm) into the inside surface of Pyrex capillaries (0.3 mm I.D. x 0.8 mm O.D.) in such a manner that the resulting traps with a standard length of 80 mm had an uncoated section of ca. 20 mm which was used for connection to the capillary column. Some oxidation of the activated carbon takes place during the preparation of the COTS, leaving a residue of basic inorganic salts. To eliminate any additional catalytic activity, this material was removed by rinsing the finished traps with 20% nitric acid, followed by repeated washings with distilled water until neutral. Gas samples were made up by injecting ca. 1 ~1 of the respective compounds onto some purified glass wool in a lOO-ml screw-capped bottle filled with nitrogen which had been purified through a column (200 mm x 30 mm) of activated charcoal. All glass-ware and the glass wool were purified at 500°C in an oven before use. The PTFE-lined gaskets used in the screw-capped bottles were cleaned in a well-ventilated oven at ,lOO”Cfor 12 h. In experiments requiring accurate control of the flow-rate of solvent through the trap, the solvent-handling syringe was installed in an infusion pump (Sage Instruments, White Plains, NY, U.S.A.).
HEADSPACE GAS ANALYSIS OF VOLATILES
97
Manipulation
For headspace sampling the PTFE-lined rubber gasket of the sample-containing bottle was pierced with two glass capillaries (0.5 mm I.D. x 1.2 mm O.D.). Of these two capillaries, the shorter one with its lower tip reaching to the neck of the bottle, was used to introduce purified nitrogen into the bottle. The COT was connected with shrinkable PTFE to the second capillary which reached almost to the bottom of the bottle. A bubble flow meter connected to the other end of the trap allowed the measurement of the volume of headspace gas sampled through the trap. Sampling was carried out at a rate of 4-8 ml/min at a room temperature of ca. 20°C. On-column liquid desorption was carried out as follows. The glass capillary column was installed in the gas chromatograph with its one end connected to the detector. The other end of the column with the retention gap was supplied with a ferrule and a ferrule-supporting ring. The detector flame was ignited and as soon as a stable base line had been obtained, the front end of the column was connected to a loaded COT. An air plug of approximately 30 ~1, followed by the volume of solvent to be used for liquid desorption, was sucked into a 50-~1 syringe. The syringe needle was connected to the COT with a shrinkable PTFE connection, whereafter the solvent and air plugs were slowly pushed through the trap, the solvent extracting the adsorbed material from the COT and the air plug transporting the solvent ca. 30 cm into the first coil of the column, in order to prevent it from being pushed back, out of the column. The trap with its PTFE connection was removed, the carrier gas turned on, the column connected to the injector as quickly as possible, and the analysis started. Although the injector temperature is of minor importance in the liquid desorption procedure, a relatively high temperature was nevertheless selected to avoid tailing of high-boiling compounds which could result from the retention of traces of such compounds on the section of the column inside the injector. The COT was reactivated after completion of each analysis by a short thermal desorption of ca. 10 min along the lines described by Grob and Habich’ by using the injector of a gas chromatograph as heating device. The application of liquid desorption to the determination of compounds with different polarities and volatilities was demonstrated by determinations carried out on gas mixtures containing I-butanol, 1-pentanol, 1-hexanol, 1-heptanol, I-octanol and methyl octanoate. Using a gas mixture containing the thermally labile terpenes, limonene, y-terpinene and a-pinene (0.8-1.0 ,ul of each terpene on some glass wool in 100 ml of nitrogen), the results obtained by on-line liquid desorption of these terpenes were compared to those obtained by thermal desorption according to the procedure of Grob and Habich’. The terpene-containing gas mixture was also used to determine the influence of the volume of the solvent on the desorption and removal of adsorbed volatiles from the COT. Finally the applicability of the solvent desorption technique in general headspace-analytical practice was demonstrated by a headspace gas determination of an imitation passion-fruit soft drink. RESULTS AND DISCUSSION
It could be argued that the capacity of the Grob-Habich COTS is too low for the determination of highly volatile compounds and that, for this reason, they cannot be used for headspace gas determinations which require sampling over long periods
B. V. BURGER,
98
2. MUNRO
of time. This shortcoming can, however, be rectified to a certain extent by merely increasing the particle size of the activated carbon used in the traps. We have, for instance, found that a four-fold increase in the particle size of the activated carbon used in fused silica traps, resulted in a 27-fold increase in capacity of the trap for n-pentane2. Unfortunately, increasing the particle size of the adsorbent invariably results in higher desorption temperatures being required for rapid and quantitative desorption of the trapped volatiles, resulting in a corresponding increase in the catalytic activity of the activated carbon. Although thermal desorption is undoubtedly the most convenient desorption technique, liquid desorption was seen as a possible solution to the problems arising from this relationship between capacity and desorption temperature. A crucial question is, however, how effectively adsorbed material can be desorbed by the solvent and transported onto the capillary column. It is clear that traces of the desorbed material could be expected to remain in the trap, to be removed by a second solvent plug if the desorption step is repeated. In an exploratory investigation of the efficiency with which relatively volatile polar and non-polar compounds are desorbed and transported onto the capillary column by small volumes of a suitable solvent, a headspace gas determination was carried out on a gas sample containing a number of primary alcohols and a methyl ester. These compounds were trapped on a COT at a gas flow-rate of 8 ml/mm, whereafter the trapped material was desorbed with 6 ,ul of methylene chloride at a flow-rate of 0.5 pi/s, i.e. a solvent/ adsorbent contact time of 12 s for each solvent plug. The recovered material was subjected to quantitation using a mixture of the same alcohols and esters as external standards. Second and third liquid desorptions, using the same desorption and quantitation procedure, were carried out to determine the efficiency of this procedure. The results of these experiments are summarised in Table I and show practically quantitative recovery under these experimental conditions. In contrast to the sharply eluting peaks obtained by a conventional GC analysis of the three thermally labile terpenes, a-pinene, limonene and y-terpinene (Fig. lA), thermal desorption of these terpenes from a COT at 250°C results in considerable peak broadening, tailing and the production of artefacts (Fig. 1B). Increasing the desorption temperature to 310°C slightly improved the peak shapes, but this
TABLE I RECOVERY
OF VOLATILES
FROM A COT BY THREE SUCCESSIVE LIQUID DESORI’TIONS
The results are given in pg. The flow-rate was ca. 0.5 pi/s.
compou?ldr
I-Butanol 1-Pentanol 1-Hexanol 1-Heptanol 1-O&m01 Methyl octanoate
Desorptions with 6-pl aliquots of methylene chloride I
II
III
10.27 12.27 9.60 5.85 1.99 1.23
0 0 0 0.001 0.002 0
0 0 0 0 0 0
HEADSPACE GAS ANALYSIS OF VOLATILES
I
99
C
+
L- ;
t
L
e il:
5
Fig. 1. GC analyses of thermally labile terpenes. Column, 30 m x 0.3 mm I.D. glass, 0.3 q OV-61-OH; temperature programme, 30 to 180°C at 4”C/min; carrier gas, helium at 28.6 cm/s. (A) Split injection of the mixture of terpenes dissolved in methylene chloride. Injector, 80°C; detector, 220°C. 1 = a-Pinene; 2 = limoaene; 3 = y-terpinene. Headspace analyses were carried out by trapping 3-ml samples of a gas mixture containing these three terpenes on a COT and comparing thermal and liquid desorption in order to demonstrate the superiority of the latter desorption technique for the determination of thermally labile compounds. (B) Thermal desorption. Desorption temperature, 250°C (5 min). (C) Thermal desorption. Desorption temperature, 31O’C (2 mitt). (D) Liquid desorption with 5 ~1 of methylene chloride at a flow-rate of 0.5 pi/s.
advantage was offset by increased artefact production (Fig. 1C). As expected, no artefacts were formed when liquid desorption was used, and better peak shapes were obtained, although some peak broadening, most likely due to column overloading, was observed (Fig. 1D). The influence of the solvent plug volume on the recovery of adsorbed material from a COT was investigated in more detail by employing a gas sample containing the same three terpenes mentioned above. These apolar compounds were preferred to more polar compounds such as, for example, a mixture of alcohols, to avoid involving the influence of the polarity of the trapped volatiles in the investigation of this aspect. If liquid desorption is carried out with decreasing volumes of solvent, a stage will be reached beyond which a further decrease will result in increasing quantities of the trapped material being left in the trap itself or in the PTFE connection. A second or even third rinsing with such a small volume of solvent would then be needed to effect quantitative removal of the material from the trap. It is clear from the results shown in Table II that if large quantities of the three terpenes are desorbed from the COT with 10 ~1 of methylene chloride, traces of the three compounds could still be detected in the third lo-p1 washing with this solvent. On the other hand, however, the first desorption with 10 ~1 of solvent already removed 99.7% of the adsorbed material, a result which is far better than the sampling accuracy that could
loo
B. V. BURGER, 2. MUNRO
TABLE II RECOVERY OF TEWENES
FROM A COT BY THREE SUCCESSIVE LIQUID DESORPTIONS
The results are given in fig with the corresponding percentage recovery in brackets. Compounds
Ginene Limonene y-Terpinene
Desorption with IO-$ aliquots of methylene chloride* I
II
III
664 (99.7) 422 (99.7) 478 (99.7)
1.7 (0.3) 1.2 (0.3) 1.2 (0.3)
0.1 (0.02) 0.08 (0.02) 0.06 (0.01)
Desorption with 5-yl aliquots of methylene chloride*
a-Pinene Limonene y-Terpinene
I
II
III
0.42 (98.8) 0.25 (98.4) 0.27 (98.5)
0.005 (1.2) 0.004 (1.6) 0.004 (1.5)
0 0 0
* Flow-rate was ca. 0.7 pll/s. ** Flow-rate was ca. 0.5 d/s.
normally be achieved in headspace gas analysis. Whereas headspace-analytical work normally requires the determination of volatiles in relatively small quantities, a very large quantity of material was collected on the trap in this experiment and consequently the three terpenes were eluted as very broad peaks at an attenuation of x 1024. Certain column types or applications such as GC-mass spectrometry analysis, may require volumes of solvent smaller than the 10 ~1 used in this experiment. Further experiments were therefore carried out to determine the efficiency with which much smaller quantities of the terpenes could be desorbed with 5 ~1 of methylene chloride. The results in Table II show that it is still possible to obtain an acceptable recovery with this volume of solvent. Since a certain volume of solvent is retained by the activated carbon in the trap, only about 4.5 ~1 of the solvent actually reaches the capillary column. In those cases where it is unavoidable to desorb the trapped material with a sufiiciently small volume of solvent, the desorbed material and solvent can be collected in a Reati-Vial (Pierce, Rockford, IL, U.S.A.) to be analysed in the conventional manner. The recovery of material from the COT is also expected to be affected by the flow-rate at which the solvent is pushed through the trap, a slow-moving plug being more effective than the same volume of solvent which moves rapidly through the trap. This may be partly due to the fact that the desorption of material from the activated carbon is not an instantaneous process, and partly to mixing of the solvent resulting from its rapid movement over the rough activated carbon surface. Whereas a sufficiently long solvent plug will compensate for any adverse effects arising from uneven plunger movement and solvent turbulence, the effect of these factors may become noticeable if short solvent plugs are used. As could be expected when these considerations are taken into account, attempts to improve recovery of the material from a COT by moving the solvent plug back and forth a few times failed, as this
HEADSPACE
GAS ANALYSIS
101
OF VOLATILES
promoted mixing of the solvent. In the experiments of which the results arc reported in Table II, solvent flow&rates corresponding to solvent/adsorbent contact times of 10-15 s, produced quite satisfactory results. That the flow-rate of the solvent is not an extremely critical factor, as far as recovery of material from LICOT is concerned, was illustrated in an experiment in which the volatiles from 40 ml of the headspace gas of an imitation fruit drink were
I
1
L.
A:
First
desorption
Second
desorption
B:
First
desorption
Second
desorption
.a-
Fig. 2. Demonstration of the itiuence of solvent flow-rate on the recovery of the headspace volatiles of an imitation passion-fruit cool drink from a COT. Column, 27 m x 0.3 mm I.D. glass, 0.3 m OV-61OH; temperature programme, 30 to 220°C at Z”C/min; carrier gas, helium at 28.6 cm/s; attenuation, x 64, headspace gas sample size, 40 ml. (A) First and second desorptions carried out with 6 fl of methylene chloride at a flow-rate of 0.3 d/s. (B) First desorption with 6 fl of methylene chloride at 3 ccl/s; second desorption with 6 ~1 at 0.3 PI/S.
102
B. V. BURGER, Z. MUNRO
trapped on a COT and desorbed at widely different solvent flow-rates. In the first of these headspace gas determinations quantitative recovery of the adsorbed volatiles was achieved by pushing 6 ~1 of methylene chloride through the trap with an infusion pump at a flow-rate of 0.3 pi/s. The gas chromatograms obtained for the first and second liquid desorptions, both carried out at 0.3 @/s, are given in Fig. 2A. The determination was then repeated using a flow-rate of 3 pi/s for the first and 0.3 pi/s for the second desorption. The resulting gas chromatograms are shown in Fig. 2B. According to quantitative data obtained from these chromatograms, 13-20% of the individual volatile components remained on the COT at the unreasonably high flow-rate used in the first desorption. This result was to be expected if it is taken into consideration that a flow-rate of 3 pi/s corresponds to a solvent/adsorbent contact time of only 2 s. It has to be pointed out that parameters such as the particle size of the activated carbon and its activity, the length of the COT, the type and quantity of the material adsorbed on the trap and the solvent selected for desorption, may to a certain extent influence the efficiency with which material will be recovered from a COT by liquid desorption. It would therefore be advisable to determine which volume of a particular solvent is required for quantitative removal of volatiles from a COT every time a new batch of activated carbon is used to prepare a COT or if one of the other parameters is changed. From the results of the present investigation, a volume of 5-6 ~1 of methylene chloride used at a flow-rate of approximately 0.5 pi/s appears to be a good starting point for such a determination. CONCLUSIONS
The two main advantages of activated carbon as an adsorbent in headspace gas analysis are its high capacity and its thermal stability which eliminates the possibility that background peaks, due to thermal decomposition of the adsorbent, may appear in the gas chromatogram of the desorbed headspace volatiles. In some cases, however, the advantages of the thermal stability of activated carbon cannot be utilized, as the one important disadvantage of this adsorbent, viz. its catalytic activity, is accentuated if the adsorbed volatiles are to be thermally desorbed. In the present study, Prof. Kurt Grob’s idea3 of using solvent desorption as the preferred desorption method in headspace-analytical work with COTS, was found to be a simple but nevertheless quite elegant solution to the problem of the catalytic activity of activated carbon at elevated temperatures. Even in inexperienced hands excellent results can be obtained as long as a few simple precautions are taken, such as using tightly fitting PTFE connections, avoiding large gaps between trap and column, and using high-purity solvents. We feel that this method deserves to be added to the repertoire of methods employed in laboratories which are involved in headspace-analytical work. ACKNOWLEDGEMENTS
The support of this work by the Foundation for Research Development (Pretoria, South Africa), and the University of Stellenbosch (Stellenbosch, South Africa) is gratefully acknowledged. We thank Dr. I. M. Moodie (Medical Research Council, Cape Town, South Africa), for reading the manuscript.
HEADSPACE GAS ANALYSIS OF VOLATILES REFERENCES 1 K. Grob and A. Habich, J. Chromutogr., 321 (1985) 45. 2 B. V. Burger and Z. Munro, J. Chromotogr., 370 (1986) 449. 3 K. Grob, personal communication.
103
Printed in The Netherlands
CHROM. 19 607
HEADSPACE
GAS ANALYSIS
LIQUID DESORPTION OF HEADSPACE VOLATILES VATED CARBON OPEN TUBULAR TRAPS*
TRAPPED ON ACTI-
B. V. BURGER* and ZENDA MUNRO Laboratory for Ecological Chemistry, University of Stellenbosch. Stellenbosch 7600 (South Africa) (Received March 24th, 1987)
SUMMARY
The liquid desorption of headspace volatiles from Grob-Habich activated carbon open tubular traps with small volumes of solvent and capillary column gas chromatographic analysis of the resulting solutions in an on-column type of procedure, was found to be an excellent alternative to thermal desorption which, in the case of thermally labile compounds, may lead to the formation of artefacts. The influence of the solvent volume and its flow-rate through the traps on the efficiency with which the adsorbed material is desorbed and transported to the capillary column, were investigated. It was found that practically quantitative recovery of trapped terpenes could be achieved with 5 ~1 of methylene chloride at a flow-rate of 0.5 &s. The efficiency of liquid desorption and the simplicity of the on-column liquid desorption technique was illustrated in a headspace gas determination of an imitation fruit drink.
INTRODUCTION
As an extension of the pioneering work of Grob and Habich’ on the development of short open tubular capillary traps for headspace gas analysis, we have recently introduced long fused silica traps for applications requiring quantitative trapping of highly volatile componenW. Although the instrumentation required for the implementation of the latter technique is relatively simple, finding and assembling the few pieces of equipment may not be equally simple in all laboratories. In comparison, the method of Grob and Habich is straightforward, requiring only the short capillary trap and a piece of shrinkable PTFE. The analysis, furthermore, requires no special manipulative skills. It is therefo,re not unlikely that the Grab-Habich method will be preferred, especially in the exploratory stages of the development of a headspace-analytical method for a particular analytical problem, the Burger-Munro procedure only being invoked in cases where highly volatile and thermally labile
* Dedicated to the memory of Professor Kurt Grob. 0021-9673/87/$03.50
@
1987 Elsevier Science Publishers B.V.
96
B. V. BURGER,
Z. MUNRO
compounds, for example, have to be quantitatively trapped, or where sampling has to be carried out over extended periods of time. A serious drawback of activated carbon as adsorbent in short capillary traps, appears to be its catalytic activity. It is possible to overcome this problem to a certain extent by subjecting the carbon open tubular trap (COT) to temperature-programmed desorption, using an initial injector temperature at which the adsorbed material does not undergo any catalytic reaction, and a final temperature just high enough to effect quantitative desorption. However, since temperature-programmed desorption is essentially a much slower process than isothermal desorption at a high temperature, it will only produce acceptably sharp gas chromatographic (GC) peaks if the desorbed volatiles can be thermally focussed on the capillary column. During a series of courses on capillary chromatography recently held in our laboratory, Prof. Kurt Grob suggested that whereas thermal desorption produces satisfactory results in work with film traps (FTs) and should therefore be reserved for this purpose, COTS may be more amenable to liquid desorption procedures, and he prompted us to investigate this method in more detail 3. In this paper the results are given of experiments carried out to establish general guidelines for the liquid desorption of material from Grob-Habich COTS. EXPERIMENTAL
GC analyses were carried out with a Carlo Erba Fractovap 4160 gas chromatograph (Carlo Erba, Milan, Italy) equipped with a flame ionization detector, using helium as carrier gas. A glass capillary column (30 m x 0.3 mm I.D.), coated with OV-61-OH and having a l-m retention gap was used for analytical separations. Quantitative determinations of the material desorbed from the COT were carried out by splitless injection of solutions of the compounds under investigation, at concentrations of 0.5 mg/ml in methylene chloride (Merck, residue analysis grade). Quantitation was done with a Hewlett-Packard 3385A Lab Automation System (Hewlett-Packard, Palo Alto, CA, U.S.A.). The COTS were prepared by melting activated carbon particles (10-35 pm) into the inside surface of Pyrex capillaries (0.3 mm I.D. x 0.8 mm O.D.) in such a manner that the resulting traps with a standard length of 80 mm had an uncoated section of ca. 20 mm which was used for connection to the capillary column. Some oxidation of the activated carbon takes place during the preparation of the COTS, leaving a residue of basic inorganic salts. To eliminate any additional catalytic activity, this material was removed by rinsing the finished traps with 20% nitric acid, followed by repeated washings with distilled water until neutral. Gas samples were made up by injecting ca. 1 ~1 of the respective compounds onto some purified glass wool in a lOO-ml screw-capped bottle filled with nitrogen which had been purified through a column (200 mm x 30 mm) of activated charcoal. All glass-ware and the glass wool were purified at 500°C in an oven before use. The PTFE-lined gaskets used in the screw-capped bottles were cleaned in a well-ventilated oven at ,lOO”Cfor 12 h. In experiments requiring accurate control of the flow-rate of solvent through the trap, the solvent-handling syringe was installed in an infusion pump (Sage Instruments, White Plains, NY, U.S.A.).
HEADSPACE GAS ANALYSIS OF VOLATILES
97
Manipulation
For headspace sampling the PTFE-lined rubber gasket of the sample-containing bottle was pierced with two glass capillaries (0.5 mm I.D. x 1.2 mm O.D.). Of these two capillaries, the shorter one with its lower tip reaching to the neck of the bottle, was used to introduce purified nitrogen into the bottle. The COT was connected with shrinkable PTFE to the second capillary which reached almost to the bottom of the bottle. A bubble flow meter connected to the other end of the trap allowed the measurement of the volume of headspace gas sampled through the trap. Sampling was carried out at a rate of 4-8 ml/min at a room temperature of ca. 20°C. On-column liquid desorption was carried out as follows. The glass capillary column was installed in the gas chromatograph with its one end connected to the detector. The other end of the column with the retention gap was supplied with a ferrule and a ferrule-supporting ring. The detector flame was ignited and as soon as a stable base line had been obtained, the front end of the column was connected to a loaded COT. An air plug of approximately 30 ~1, followed by the volume of solvent to be used for liquid desorption, was sucked into a 50-~1 syringe. The syringe needle was connected to the COT with a shrinkable PTFE connection, whereafter the solvent and air plugs were slowly pushed through the trap, the solvent extracting the adsorbed material from the COT and the air plug transporting the solvent ca. 30 cm into the first coil of the column, in order to prevent it from being pushed back, out of the column. The trap with its PTFE connection was removed, the carrier gas turned on, the column connected to the injector as quickly as possible, and the analysis started. Although the injector temperature is of minor importance in the liquid desorption procedure, a relatively high temperature was nevertheless selected to avoid tailing of high-boiling compounds which could result from the retention of traces of such compounds on the section of the column inside the injector. The COT was reactivated after completion of each analysis by a short thermal desorption of ca. 10 min along the lines described by Grob and Habich’ by using the injector of a gas chromatograph as heating device. The application of liquid desorption to the determination of compounds with different polarities and volatilities was demonstrated by determinations carried out on gas mixtures containing I-butanol, 1-pentanol, 1-hexanol, 1-heptanol, I-octanol and methyl octanoate. Using a gas mixture containing the thermally labile terpenes, limonene, y-terpinene and a-pinene (0.8-1.0 ,ul of each terpene on some glass wool in 100 ml of nitrogen), the results obtained by on-line liquid desorption of these terpenes were compared to those obtained by thermal desorption according to the procedure of Grob and Habich’. The terpene-containing gas mixture was also used to determine the influence of the volume of the solvent on the desorption and removal of adsorbed volatiles from the COT. Finally the applicability of the solvent desorption technique in general headspace-analytical practice was demonstrated by a headspace gas determination of an imitation passion-fruit soft drink. RESULTS AND DISCUSSION
It could be argued that the capacity of the Grob-Habich COTS is too low for the determination of highly volatile compounds and that, for this reason, they cannot be used for headspace gas determinations which require sampling over long periods
B. V. BURGER,
98
2. MUNRO
of time. This shortcoming can, however, be rectified to a certain extent by merely increasing the particle size of the activated carbon used in the traps. We have, for instance, found that a four-fold increase in the particle size of the activated carbon used in fused silica traps, resulted in a 27-fold increase in capacity of the trap for n-pentane2. Unfortunately, increasing the particle size of the adsorbent invariably results in higher desorption temperatures being required for rapid and quantitative desorption of the trapped volatiles, resulting in a corresponding increase in the catalytic activity of the activated carbon. Although thermal desorption is undoubtedly the most convenient desorption technique, liquid desorption was seen as a possible solution to the problems arising from this relationship between capacity and desorption temperature. A crucial question is, however, how effectively adsorbed material can be desorbed by the solvent and transported onto the capillary column. It is clear that traces of the desorbed material could be expected to remain in the trap, to be removed by a second solvent plug if the desorption step is repeated. In an exploratory investigation of the efficiency with which relatively volatile polar and non-polar compounds are desorbed and transported onto the capillary column by small volumes of a suitable solvent, a headspace gas determination was carried out on a gas sample containing a number of primary alcohols and a methyl ester. These compounds were trapped on a COT at a gas flow-rate of 8 ml/mm, whereafter the trapped material was desorbed with 6 ,ul of methylene chloride at a flow-rate of 0.5 pi/s, i.e. a solvent/ adsorbent contact time of 12 s for each solvent plug. The recovered material was subjected to quantitation using a mixture of the same alcohols and esters as external standards. Second and third liquid desorptions, using the same desorption and quantitation procedure, were carried out to determine the efficiency of this procedure. The results of these experiments are summarised in Table I and show practically quantitative recovery under these experimental conditions. In contrast to the sharply eluting peaks obtained by a conventional GC analysis of the three thermally labile terpenes, a-pinene, limonene and y-terpinene (Fig. lA), thermal desorption of these terpenes from a COT at 250°C results in considerable peak broadening, tailing and the production of artefacts (Fig. 1B). Increasing the desorption temperature to 310°C slightly improved the peak shapes, but this
TABLE I RECOVERY
OF VOLATILES
FROM A COT BY THREE SUCCESSIVE LIQUID DESORI’TIONS
The results are given in pg. The flow-rate was ca. 0.5 pi/s.
compou?ldr
I-Butanol 1-Pentanol 1-Hexanol 1-Heptanol 1-O&m01 Methyl octanoate
Desorptions with 6-pl aliquots of methylene chloride I
II
III
10.27 12.27 9.60 5.85 1.99 1.23
0 0 0 0.001 0.002 0
0 0 0 0 0 0
HEADSPACE GAS ANALYSIS OF VOLATILES
I
99
C
+
L- ;
t
L
e il:
5
Fig. 1. GC analyses of thermally labile terpenes. Column, 30 m x 0.3 mm I.D. glass, 0.3 q OV-61-OH; temperature programme, 30 to 180°C at 4”C/min; carrier gas, helium at 28.6 cm/s. (A) Split injection of the mixture of terpenes dissolved in methylene chloride. Injector, 80°C; detector, 220°C. 1 = a-Pinene; 2 = limoaene; 3 = y-terpinene. Headspace analyses were carried out by trapping 3-ml samples of a gas mixture containing these three terpenes on a COT and comparing thermal and liquid desorption in order to demonstrate the superiority of the latter desorption technique for the determination of thermally labile compounds. (B) Thermal desorption. Desorption temperature, 250°C (5 min). (C) Thermal desorption. Desorption temperature, 31O’C (2 mitt). (D) Liquid desorption with 5 ~1 of methylene chloride at a flow-rate of 0.5 pi/s.
advantage was offset by increased artefact production (Fig. 1C). As expected, no artefacts were formed when liquid desorption was used, and better peak shapes were obtained, although some peak broadening, most likely due to column overloading, was observed (Fig. 1D). The influence of the solvent plug volume on the recovery of adsorbed material from a COT was investigated in more detail by employing a gas sample containing the same three terpenes mentioned above. These apolar compounds were preferred to more polar compounds such as, for example, a mixture of alcohols, to avoid involving the influence of the polarity of the trapped volatiles in the investigation of this aspect. If liquid desorption is carried out with decreasing volumes of solvent, a stage will be reached beyond which a further decrease will result in increasing quantities of the trapped material being left in the trap itself or in the PTFE connection. A second or even third rinsing with such a small volume of solvent would then be needed to effect quantitative removal of the material from the trap. It is clear from the results shown in Table II that if large quantities of the three terpenes are desorbed from the COT with 10 ~1 of methylene chloride, traces of the three compounds could still be detected in the third lo-p1 washing with this solvent. On the other hand, however, the first desorption with 10 ~1 of solvent already removed 99.7% of the adsorbed material, a result which is far better than the sampling accuracy that could
loo
B. V. BURGER, 2. MUNRO
TABLE II RECOVERY OF TEWENES
FROM A COT BY THREE SUCCESSIVE LIQUID DESORPTIONS
The results are given in fig with the corresponding percentage recovery in brackets. Compounds
Ginene Limonene y-Terpinene
Desorption with IO-$ aliquots of methylene chloride* I
II
III
664 (99.7) 422 (99.7) 478 (99.7)
1.7 (0.3) 1.2 (0.3) 1.2 (0.3)
0.1 (0.02) 0.08 (0.02) 0.06 (0.01)
Desorption with 5-yl aliquots of methylene chloride*
a-Pinene Limonene y-Terpinene
I
II
III
0.42 (98.8) 0.25 (98.4) 0.27 (98.5)
0.005 (1.2) 0.004 (1.6) 0.004 (1.5)
0 0 0
* Flow-rate was ca. 0.7 pll/s. ** Flow-rate was ca. 0.5 d/s.
normally be achieved in headspace gas analysis. Whereas headspace-analytical work normally requires the determination of volatiles in relatively small quantities, a very large quantity of material was collected on the trap in this experiment and consequently the three terpenes were eluted as very broad peaks at an attenuation of x 1024. Certain column types or applications such as GC-mass spectrometry analysis, may require volumes of solvent smaller than the 10 ~1 used in this experiment. Further experiments were therefore carried out to determine the efficiency with which much smaller quantities of the terpenes could be desorbed with 5 ~1 of methylene chloride. The results in Table II show that it is still possible to obtain an acceptable recovery with this volume of solvent. Since a certain volume of solvent is retained by the activated carbon in the trap, only about 4.5 ~1 of the solvent actually reaches the capillary column. In those cases where it is unavoidable to desorb the trapped material with a sufiiciently small volume of solvent, the desorbed material and solvent can be collected in a Reati-Vial (Pierce, Rockford, IL, U.S.A.) to be analysed in the conventional manner. The recovery of material from the COT is also expected to be affected by the flow-rate at which the solvent is pushed through the trap, a slow-moving plug being more effective than the same volume of solvent which moves rapidly through the trap. This may be partly due to the fact that the desorption of material from the activated carbon is not an instantaneous process, and partly to mixing of the solvent resulting from its rapid movement over the rough activated carbon surface. Whereas a sufficiently long solvent plug will compensate for any adverse effects arising from uneven plunger movement and solvent turbulence, the effect of these factors may become noticeable if short solvent plugs are used. As could be expected when these considerations are taken into account, attempts to improve recovery of the material from a COT by moving the solvent plug back and forth a few times failed, as this
HEADSPACE
GAS ANALYSIS
101
OF VOLATILES
promoted mixing of the solvent. In the experiments of which the results arc reported in Table II, solvent flow&rates corresponding to solvent/adsorbent contact times of 10-15 s, produced quite satisfactory results. That the flow-rate of the solvent is not an extremely critical factor, as far as recovery of material from LICOT is concerned, was illustrated in an experiment in which the volatiles from 40 ml of the headspace gas of an imitation fruit drink were
I
1
L.
A:
First
desorption
Second
desorption
B:
First
desorption
Second
desorption
.a-
Fig. 2. Demonstration of the itiuence of solvent flow-rate on the recovery of the headspace volatiles of an imitation passion-fruit cool drink from a COT. Column, 27 m x 0.3 mm I.D. glass, 0.3 m OV-61OH; temperature programme, 30 to 220°C at Z”C/min; carrier gas, helium at 28.6 cm/s; attenuation, x 64, headspace gas sample size, 40 ml. (A) First and second desorptions carried out with 6 fl of methylene chloride at a flow-rate of 0.3 d/s. (B) First desorption with 6 fl of methylene chloride at 3 ccl/s; second desorption with 6 ~1 at 0.3 PI/S.
102
B. V. BURGER, Z. MUNRO
trapped on a COT and desorbed at widely different solvent flow-rates. In the first of these headspace gas determinations quantitative recovery of the adsorbed volatiles was achieved by pushing 6 ~1 of methylene chloride through the trap with an infusion pump at a flow-rate of 0.3 pi/s. The gas chromatograms obtained for the first and second liquid desorptions, both carried out at 0.3 @/s, are given in Fig. 2A. The determination was then repeated using a flow-rate of 3 pi/s for the first and 0.3 pi/s for the second desorption. The resulting gas chromatograms are shown in Fig. 2B. According to quantitative data obtained from these chromatograms, 13-20% of the individual volatile components remained on the COT at the unreasonably high flow-rate used in the first desorption. This result was to be expected if it is taken into consideration that a flow-rate of 3 pi/s corresponds to a solvent/adsorbent contact time of only 2 s. It has to be pointed out that parameters such as the particle size of the activated carbon and its activity, the length of the COT, the type and quantity of the material adsorbed on the trap and the solvent selected for desorption, may to a certain extent influence the efficiency with which material will be recovered from a COT by liquid desorption. It would therefore be advisable to determine which volume of a particular solvent is required for quantitative removal of volatiles from a COT every time a new batch of activated carbon is used to prepare a COT or if one of the other parameters is changed. From the results of the present investigation, a volume of 5-6 ~1 of methylene chloride used at a flow-rate of approximately 0.5 pi/s appears to be a good starting point for such a determination. CONCLUSIONS
The two main advantages of activated carbon as an adsorbent in headspace gas analysis are its high capacity and its thermal stability which eliminates the possibility that background peaks, due to thermal decomposition of the adsorbent, may appear in the gas chromatogram of the desorbed headspace volatiles. In some cases, however, the advantages of the thermal stability of activated carbon cannot be utilized, as the one important disadvantage of this adsorbent, viz. its catalytic activity, is accentuated if the adsorbed volatiles are to be thermally desorbed. In the present study, Prof. Kurt Grob’s idea3 of using solvent desorption as the preferred desorption method in headspace-analytical work with COTS, was found to be a simple but nevertheless quite elegant solution to the problem of the catalytic activity of activated carbon at elevated temperatures. Even in inexperienced hands excellent results can be obtained as long as a few simple precautions are taken, such as using tightly fitting PTFE connections, avoiding large gaps between trap and column, and using high-purity solvents. We feel that this method deserves to be added to the repertoire of methods employed in laboratories which are involved in headspace-analytical work. ACKNOWLEDGEMENTS
The support of this work by the Foundation for Research Development (Pretoria, South Africa), and the University of Stellenbosch (Stellenbosch, South Africa) is gratefully acknowledged. We thank Dr. I. M. Moodie (Medical Research Council, Cape Town, South Africa), for reading the manuscript.
HEADSPACE GAS ANALYSIS OF VOLATILES REFERENCES 1 K. Grob and A. Habich, J. Chromutogr., 321 (1985) 45. 2 B. V. Burger and Z. Munro, J. Chromotogr., 370 (1986) 449. 3 K. Grob, personal communication.
103