Wavelength Modulation Diode Laser Atomic Absorption Spectrometry in Modulated Low-pressure Helium Plasmas for Element-selective Detection in Gas Chromatography* Plenary Lecture ALEKSANDR ZYBIN CHRISTOPH SCHNURER-PATSCHAN AND KAY NIEMAXt Institute of Physics University of Hohenheim Garbenstrasse 30 0-70599 Stuttgart Germany A new technique for element-selective detection in gas chromatography is reported. It is based on wavelength modulation diode laser atomic absorption spectrometry ( WM-LAAS) in modulated low-pressure dc or microwave- induced plasmas (MIP). The double modulation laser atomic absorption spectrometry (DM-LAAS) eliminates not only flicker noise from the laser as well as from the plasma but also etalon effects which limit the detection in WM-LAAS. DM-LAAS allows the measurement of absorbances of about lo-'.The analytical power of the technique is demonstrated by the analysis of haloform samples applying DM-LAAS of metastable chlorine and bromine atoms which are generated in the low-pressure plasmas by dissociation and excitation. Since the species are dissociated completely the signals reflect the relative element concentration in the molecules which allows calibration by internal standards. The detection limits are very low. For example 3s detection limits were found to be below 3 ng ml-' for species such as CHC13 or CC14 applying splitless injection of 0.5 pl samples and DM-LAAS of chlorine in the MIP. The detection limits found in the dc plasma were slightly higher. Keywords Diode laser atomic absorption spectrometry; modulation spectrometry element-selective detection; gas chroma tograph y The signal-to-noise ratio in absorption measurements can be improved significantly if lasers whose wavelength can be modulated with high frequency are used as radiation sources.' This has been demonstrated in atomic absorption spectrometry with semiconductor diode lasers in graphite tube atomizers,' analytical flames3 and low-pressure dc plasmas4 Depending on the radiation power of the fundamental or after frequency doubling in a non-linear crystal of the second harmonic laser wavelength absorbances of 10-4-10-6 have been measured.The reduction of the detection limits in elemental analysis promised improvements of the analytical figures of merit in chromatography with element-selective detection. Recently we have successfully coupled wavelength modulation laser atomic absorption spectrometry (WM-LAAS) of chromium in an analytical flame with high-performance liquid chromatography (HPLC) for the speciation of Cr"' and CrV'.5 Detection limits of about 1 ng ml-' were obtained.Sensitive detection of molecular compounds such as * Presented at the 1995 European Winter Conference on Plasma tTo whom correspondence should be addressed. Spectrochemistry Cambridge UK January 8-13 1995. Journal of Analytical Atomic Spectrometry C2CI2F CCl CHF and 02 has been shown by the WM-LAAS of metastable chlorine fluorine and oxygen atoms which were generated by dissociation and excitation in a low- power dc argon or helium pla~ma.~ In a further paper,6 this technique was coupled with gas chromatography (GC) in order to speciate chlorinated hydrocarbons by the measurement of chlorine.Since other halogens can also be measured by diode lasers element-selective measurements of halogens allow not only the discrimination between halogenated and non- halogenated species but also the halogens. In the preliminary experiment on coupling GC with WM-LAAS a low-pressure microwave-induced plasma (MIP) was used. Although the WM-LAAS signal showed the expected strength low detection limits could not be obtained because of slow fluctuations of the background absorption which could only be partly sup- pressed by the largest possible time constant of 0.1 s. The detection limits for different chlorinated hydrocarbons were only of the order of 1 pg ml-' or 80 pg s-'. The present paper reports on significant improvements of the analytical figures of merit of the coupling of GC and WM-LAAS in low-pressure plasmas (MIP as well as dc plasma).The improvements are the result of the additional modulation of the absorption by switching the plasmas on and off with a frequency of a few kHz. The absorption has to be measured at the sum or difference frequency of the second harmonic of the laser and the first harmonic of the plasma modulation frequency. The modulation of the plasma not only increases the population density in the metastable states but also suppresses wavelength-dependent changes of the laser intensity measured with the detector such as interference effects in the optical path (etalon effects).' It is shown that double modulation laser atomic absorption spectrometry (DM-LAAS) allows absorbances near to the theoretical detection limit given by the shot noise to be measured.EXPERIMENTAL The experimental arrangement for WM-LAAS in a modulated MIP is shown in Fig. 1. The radiation of a semiconductor laser diode (HL 8314 Hitachi or LT15 Sharp; line widths x 20 MHz; power 30 mW) was collimated by a large aperture lens and attenuated by an optical filter to about 0.8 mW. The diameter of the laser beam in the absorption volume was about 1 mm. The attenuation by a filter was necessary to avoid optical saturation of the strong absorption lines in the low- pressure helium plasmas. The MIP was operated in a quartz capillary (diameter 2 mm) placed in a Beenaker-type resonator.The plasma length was about 15cm at optimum operating Journal of Analytical Atomic Spectrometry September 1995 Vol. 10 563Modulator Modulator Microwave Mixer power * I I qy Lens u . Sample injection - u 11 1 Pump Gas - *Y (haloform test solutions from Chromatographie Service) and of CCl and CHCI (both extra pure quality from Merck) in pentane were prepared for the measurements. The concen- trations of the species in the haloform test solutions are listed in Table 1 as well as the chlorine and bromine concentrations in the solutions. RESULTS AND DISCUSSION Photodiode Characterization of the Helium MIP chromatograph I - I L Gas mixture Fig. 1 gas chromatography by DM-LAAS in a low-pressure plasma Experimental arrangement for element-selective detection in condition.The absorption was measured by a low-noise silicon photodiode. The absorption signal was processed by a lock-in amplifier (Stanford Instruments RS 830DS) and stored by a personal computer. When measurements in the low-pressure dc plasma (DCP) were performed the MIP shown in Fig. 1 was replaced by a 15 cm long plasma tube (diameter 3 mm) completely made of Pyrex glass. Near the windows the tube ends were widened in order to have space for large area disc electrodes (diameter 10mm) which minimized the sputtering at a given current. The plasma length was about 13 cm. For the measurement of chlorinated hydrocarbons the laser diode powered by a commercial driver (Melles Griot 56 DLD 403) was tuned to the chlorine absorption line at 837.60nm ( 3s23p44s ,PSi2 + 3s23p44p ,Do7/!) which has a large oscillator strength [f=0.39 (ref.7)]. Bromine compounds were measured by absorption of the Br 827.24nm line (4s24p45s ,P5/? +4s24p45p To our knowledge there are no data available on the oscillator strength of the Br line. However it is expected to be of the same order of magnitude as thef-value of the Cl 837.60 nm line. The wavelength of the laser diode was modulated sinusoidi- The properties of the double modulation scheme were studied by absorption of the C1 827.60nm line using mixtures of C2C1,F and helium as the plasma gas. If the laser is locked to the centre of the line and the plasma is modulated with frequency fi the laser intensity with and without plasma is measured. The output signal of the lock-in amplifier represents the specific element absorption and the non-selective absorp- tion of the plasma which was about 2% in our case.We observed an increase of the specific absorption signal with increasing plasma modulation frequency in the range from a few Hz to about 5 kHz. This behaviour is due to strong changes of the population density in the C1 metastable state during the periods the MIP was burning. The moment the plasma was switched on by the rectangular pulse the absorption reached a much larger value than in the continuous mode and decreased rapidly with time while the plasma was switched on. The time constant of this process was about 10 ps. The increase of the modulation frequency shortened the period of the burning plasma therefore on average a larger absorption was meas- ured.Most probably the high population density at the beginning of each plasma period is due to a smaller collisional deactivation of the metastable levels at a relatively low gas temperature. With time the gas temperature increases and depopulation processes become stronger until the plasma reaches asymptotically the temperature of the stationary case. When the modulation frequency was larger than 5 kHz the time between the plasma shortages and the following re-ignitions was too short for a significant cooling of the gas. Therefore the population densities at the beginning of each plasma period were smaller than for lower modulation ally with a frequency of f l = 11.5 kHz using a commercial power supply (Wavetek FG-5000). To obtain maximum signals modulation amplitudes of 7 and 5 pm were applied for the measurement of the chlorine and bromine lines respectively.A microwave generator (Feuerbacher GMW 24-303 DR) produced radiation at 2.45 GHz with a maximum power of 300 W. Typical powers of about 120 W were applied. The frequencies. Optima absorption signals were obtained at gas flow rates between 3 and 5 ml min-'. The optimum gas pressure was 70 hPa. The absorption signal increased with microwave power between 80 and 200 W. This behaviour is mainly due to an increase of the plasma length with power as was found in our earlier paper.6 However 120 W was chosen as the operation reflected power was about 2 W. The helium MIP was operated between 10 and 100 hPa. A second Wavetek power supply was used for external modulation of the MIP.The plasma was switched on and off by a square wave function. The modulation frequency of the MIP was 5 kHz. The modulation frequency f2 of the plasma was mixed with the second harmonic of the laser modulation frequency to obtain a reference frequency of 2fi -f2 = 18 kHz for the lock-in amplifier. A stabilized power supply (Fug HCN 140-3500) was used for the dc helium plasma. The gas pressure was 40 hPa and the current about 15 mA at a discharge voltage of 900 V. As in the case of the MIP a 5 kHz square wave function was applied to pulse the plasma. The gas chromatograph (Shimadzu GC-14A) was equipped with a fused silica column from Chromatographie Service (type FS-SE-54-CB-1; length 50 m; inner diameter 0.32 mm). Helium was used not only as power because the plasma became unstable at larger powers.As mentioned above the non-selective background absorp- tion in the plasma modulation measurements can be suppressed by additional modulation of the laser wavelength. A spectrum of the chlorine line measured at the frequency 2f1 -f2 is shown as trace 1 in Fig. 2. The concentration of C2C12F4 was 0.4 ppb. The scan time was 30 s and the time constant 3 s. Trace 2 is the blank measured in helium without C,Cl,F admixtures. The difference between the maximum and one of the minima of the second harmonic line profile was recorded as the analytical signal with a large time constant (50 s). The data are presented as full squares in Fig. 3 giving a calibration curve for C2C12F,. The total dynamic range not shown in Fig. 3 was about five orders of magnitude.The 3s detection limit of C2C12F was found to be about 60 ppt. The blank level was about 30 ppt (n = lo) found in preconcentration measurements the plasma gas but also as the carrier gas in the chromatograph (flow rate z 4 ml min-I). Liquid samples of 0.5 pl were introduced manually with a syringe into the splitless injector of the instrument. Two solutions of halocarbons in pentane where C2C12F molecules in a 2 1 volume were condensed in a cold trap and then released by heating to the helium gas flow. The blank was mainly due to contamination of the capillary wall and the tubes of the gas handling system. However there 564 Journal of Analytical Atomic Spectrometry September 1995 Vol. 10Table1 concentrations in 1 and 2 respectively Concentrations of different species in the haloform solutions used.The columns headed chlorine and bromine give the CI and Br Species Dichloromethane (CH,Cl,) Trichloromethane (CHCl,) 1,1,l-Trichloroethane (C,H,Cl,) Tetrachloromethane (CCI,) Trichloroethane (C,HCl,) Bromodichloromethane (CHBrC1,) Dibromochloromethane( CHBr,CI ) Tetrachloroethane (C,Cl,) Tribromomethane (CHBr,) Sample 1/ pg ml-' 200 5 1 0.25 2.5 1 1.5 0.6 4.5 Chlorine/ pg ml-' 4.4 0.79 0.23 2.03 0.42 0.26 0.54 167 - Sample 2/ pg ml-' 158.4 14.9 13.4 15.9 14.6 20.0 24.5 16.2 28.9 Bromine/ pg ml-' - ~ - ~ 9.8 18.8 27.4 - 2 GHz - I I I Optical frequency .- Fig. 2 Spectrum 1 2f chlorine absorption line measured at a concen- tration of 400 ppt C2Cl,F in helium by DM-LAAS and spectrum 2 blank .-1 ,-' /.,m ' I' 7 ,P' -1 3s detection limit 0.01 C2C12F4 in helium (ppb) Fig. 3 gases in the MIP Calibration curve for C2C1,F4 in helium measured with mixed was also a contribution by chlorine-containing polyatomic species in the ultra-pure helium used. It is interesting to note that about 0.05% of all chlorine atoms are in the 3p44s 4P5,2 metastable state at optimum conditions in the continuously operating MIP. These values were evaluated from the absorption spectra taking into account the total number density of chlorine atoms in the plasma the integrated absorption and the oscillator strength of the 837.60nm line. The fraction of metastable atoms can be increased significantly (> 1 YO) by pulsed operation. Coupling of Gas Chromatography and DM-LAAS in the MIP Since the chromatographic spikes were relatively broad at the optimum gas pressure of 70 hPa the helium pressure had to be reduced to 20 hPa to obtain narrow signals. A chromato- gram of haloform solution 1 diluted by a factor of 30 and measured by DM-LAAS of chlorine is shown in Fig.4. Note 1i I / 0 $0 120 180 240 3bO 360 420 Time/s Fig. 4 Chromatogram of a haloform solution measured by DM-LAAS of chlorine in the MIP. Species and concentrations 1 CH,Cl (6.7 pg ml-'); 2 CHCl (150 ng ml-I); 3 C,H,Cl (33 ng ml-'); 4 CCI (8 ng ml-'); 5 C,HCl (80 ng ml..'); 6 CHBrCI (33 ng ml-'); 7 CHBr,C1(50 ng ml-I); and 8 C2C14 (20 ng ml-I) the very low absorbance which can be measured by DM-LAAS. The numbered peaks in the chromatogram correspond to species listed in Table 1.At the start the temperature of the chromatograph was 25°C and increased with a rate of 10 "C min-l up to 105 "C. The start time of the chromatogram was 110 s after injection. Since the concentration of the first component (dichloromethane) was much higher than of the following species the gain of the lock-in amplifier had to be changed. In order to minimize the influence of the solvent to the plasma the plasma was switched on after the major volume of the solvent had passed the capillary of the MIP and before the first species arrived. In our case the plasma was switched on with a delay of about 23 s. Except for the first component where the plasma was still slightly influenced by the solvent the relative signals of the different chlorinated hydrocarbons reflect the chlorine concen- trations given in Table 1.On the basis of six chromatograms the relative chlorine ratios could be reproduced within k 5%. This means that within experimental uncertainty the chlorine atoms were completely dissociated from the species and the population density of the metastable state was proportional to the chlorine density. This can only happen if the plasma does not change its parameters when the species are in the plasma. The correlation between element concentration and signal gives an important advantage of element-selective detec- tion by DM-LAAS in GC i.e. the possibility of calibration by internal standards. Fig. 5 shows calibration curves for solutions with trichloro- methane and tetrachloromethane. The data were obtained by integrated absorbance measurement.The detection limits were 2.8 and 2.5 ng ml-I for CHC13 and CCl respectively. Similar Journal of Analytical Atomic Spectrometry September 1995 Vol. 10 565Concentration/ng ml-' 0 1 I I ' I Fig. 5 Calibration curves for CHCl and CCl measured chromato- graphically by DM-LAAS of chlorine in the MIP J (D 1 x 1 0 ~ ~ ? 2 1x1027 c m 1x101; 13 a l X l O O - to measurements of C2C12F4 in helium the dynamic range of the chromatographic measurements of CHC13 and CC14 was about six orders of magnitude. Taking into account the injected sample volume the detection limit of chlorine was about 0.12 pg s-' or 1.2 pg. 0 ; ; I t Characterization of the dc Helium Plasma The properties of the dc helium plasma were studied with mixtures of C2C12F4 and helium.Similar to that for the MIP the absorption signal had a maximum at a plasma modulation frequency,f2 of about 5 kHz. The laser wavelength modulation frequency fl was again 11.5 kHz. Therefore the same refer- ence frequency ( 2f1 -f2 = 18 kHz) for the lock-in amplifier could be used as in the measurements with the MIP. The dependence of the chlorine line absorption on gas pressure for five different flow rates is displayed in Fig. 6. While the signal was almost independent of pressure for small flow rates e.g. 1.3 ml min-' it increased with pressure for higher flow rates. Unfortunately we could not go far beyond 50 hPa because our current modulator was limited to 1 kV. Higher gas pressures required voltages beyond 1 kV. Fig.7 shows the absorption signal dependence on the gas flow measured at 40 hPa the pressure which is also used in gas chromatography. A calibration curve for C2Cl,F4 in helium measured at 40 hPa and 20 ml rnin-' is given in Fig. 8. The time constant was 50s. A detection limit of about 85ppt was found. This value is comparable to the detection limit obtained with the MIP (60 ppt). Much stronger memory effects were observed in the dc plasma than in the MIP. After measurements of high C2C12F 5 60- c - c .- 50- .- - 2 40- $ 30- 1.3 2 Y - c 2 20 $ . a I O - 13 27 ml min-' 13 ml min-' Fig.6 Chlorine line absorption versus gas pressure measured at different flow rates by DM-LAAS in the DCP Fig.7 Chlorine line absorption versus gas flow rate at 40 hPa by DM-LAAS in the DCP 1 lxlo4i xl o - ~ /3s detection limit I ' ~ " ' ' ~ ' 1 ' " ~ ~ " ~ ' " ' 1 - 3 - ' """.' * ' * * * 0.1 1 l b lb0 1600 10 C2CI2F4 in He (ppb) 100 Fig. 8 Calibration curve for C2C1,F in helium obtained with mixed gases in the DCP concentration the discharge had to be cleaned by flushing pure helium through the tube at high plasma current. Coupling of Gas Chromatography and DM-LAAS in the dc Plasma For optimum coupling of gas chromatography with the DCP an auxiliary helium gas flow of 10ml min-' was used with the flow rate of the gas chromatograph (4ml min-'). The pressure in the plasma tube was 40 hPa. The memory effects were reduced significantly by the additional gas flow. A chromatogram obtained by DM-LAAS of chlorine with halo- form sample 1 (see Table 1) diluted by a factor of 20 is displayed in the upper part of Fig.9. A time constant of 1 s was applied. The calibration curve of CCl obtained by integrated absorbance measurements is given in Fig. 10. The detection limit was 5 ng ml-'. Taking into account the sample volume injected into the gas chromatograph the detection limit for chlorine was about 0.25 pg s-' or 2.5 pg. These data are slightly higher than the values measured in the MIP. However the DCP has an economic advantage. The instrumentation to run a dc plasma is less expensive than for an MIP. As in the experiments with the MIP within small statistical error bars the ratios of the peaks reflected the ratios of the chlorine concentrations in the species. Deviations were only found for the first component where the plasma was still influenced by the solvent.However this has also been observed in the measurements with the MIP. The lower chromatogram in Fig. 9 was measured by element- selective detection of bromine injecting a 20-fold diluted sample volume of the haloform test sample 2 (see Table 1). Bromine peaks were observed at the time when species CHBrCl and CHBr,Cl also gave chlorine signals with sample 1. The peak 566 Journal of Analytical Atomic Spectrometry September 1995 Vol. 101 0 . shows a hyperfine structure splitting with four major compo- nents. The component which was used for absorption rep- resents only about 30% of the total line absorption. Furthermore the population densities in the metastable state and the oscillator strength of the analyte line should be different than for chlorine.In sample 2 the concentration ratios of bromine in CHBrCI2 CHBr,Cl and CHBr are 1 1.84 2.74. These ratios can be found by inspection of the bromine signals in the lower chromatogram in Fig. 9. ' - 1 1 I Timels Fig.9 Upper trace chromatogram of diluted haloform sample 1 measured by DM-LAAS of chlorine in the DCP 1 CH2C1 (10pgml-'); 2 CHC1 (225 ngml-I); 3 C,H,CI (50ngml-'); 4 CCll (12 ngml-'); 5 C,HCl (120 ng ml-I); 6 CHBrCl (50 ng ml-'); 7 CHBr,Cl (75 ngml-I); and 8 C,Cl (30ngml-'). Lower trace chromatogram of diluted haloform sample 2 measured by DM-LAAS of bromine in the DCP 6 CHBrC1 (0.49pgml-'); 7 CHBr,Cl (0.92 pg ml-'); and 9 CHBr (1.37 pg ml-I) 'oool A 1 1000 10600 [ C C I ~ I in pentane/ng mi-' Fig. 10 Calibration curves for CC1 measured chromatographically by DM-LAAS of chlorine in the DCP 9 is due to tribromomethane which does not contain chlorine.Only very few measurements have been made with bromine. Furthermore these measurements were made at the optimum conditions found for chlorine. It is possible that the optimum plasma parameters for Br are slightly different. The preliminary detection limit found for bromine (4.5 pg s-' or 45 pg) was about a factor of 18 higher than for chlorine in the dc helium plasma. The difference is partly due to the higher relative atomic mass of bromine (factor of 2.2 in comparison with Cl) and the fact that the bromine line at 827.24nm Comparison with Other Element-specific GC detectors There are many papers on element-specific detectors based on optical emission spectrometry (OES) of different kinds of plasmas.References to these papers can be found in the literature (e.g. refs. 8-10). Depending on the kind of plasma and the lines used in emission spectrometry the best detection limits for chlorine are typically about 7 pg s-'. This value is much larger than the detection limit for chlorine obtained by DM-LAAS in the MIP and the DCP (0.12 and 0.25 pg s-' respectively). The preliminary detection limit of bromine by DM-LAAS in the dc helium plasma (4.5 pg s-') is comparable to the detection limits by OES in different plasmas!-'0 The best reported detection limits of Br by OES are about 10 pg s-1. REFERENCES 1 Silver J. A. Appl. Opt. 1992 31 707. 2 Schnurer-Patschan C. Zybin A. Groll H. and Niemax K. J. Anal. At. Spectrom. 1993 8 1103. 3 Groll H. Schnurer-Patschan C. Kuritsyn Yu. and Niemax K. Spectrochim. Acta Part B 1994 49 1463. 4 Zybin A Schniirer-Patschan C. and Niemax K. Spectrochim. Acta Part B 1993 48 1713. 5 Groll H. Schaldach G. Berndt H. and Niemax K. Spectrochim. Acta Part B in the press. 6 Schnurer-Patschan C. and Niemax K. Spectrochim. Acta Part B in the press. 7 Wiese W. L. Smith M. W. and Miles B. M. Atomic Transition Probabilities US National Bureau of Standards Washington DC vol. 11 1969. 8 Uden P. C. Element-Specijic Chromatographic Detection by Atomic Emission Spectrometry ACS Symposium Series 479 American Chemical Society Washington DC 1992. 9 Long G. L. Ducatte G. R. and Lancaster E. D. Spectrochim. Acta. Part B 1994 49 75. 10 Skelton R. J. Markides K. E. Lee M. L. and Farnsworth P. B. Appl. Spectrosc. 1990 44 853. Paper 5/01 6270 Received March 14 1995 Accepted May 15 1995 Journal of Analytical Atomic Spectrometry September 1995 Vol. 10 567