50
Anal. Chem. 1992, 6 4 , 50-55
Atomic Emission Detection for Supercritical Fluid Chromatography Using a Moderate-Power Helium Microwave-Induced Plasma Gregory K. Webster' and Jon W. Carnahan* Department of Chemistry, Northern Illinois University, DeKalb, Illinois 60115 The coupling of mlcrobore packed column supercrltlcal fluld chromatography (SFC) and a moderate-power (ca. 500 W) hellum mlcrowavelnduced plasma (MIP) for element-selectlve detectlon of nonmetals Is presented. The moderate power plasma was used to allevlate the slgnal reductlon characterlstlc with lncreaslng SFC moblle-phase flows w e n In earlier, low-power SFC-MIP systems. For the deiermlnatlon of chlorine and sulfur, the near-Infrared atomic emlsslon llnes of 837.6 and 921.3 nm were used, respectively. For the determination of carbon and hydrogen, the uttravlolet-vlslble (UV-vis) atomlc emission llnes of 247.8 and 486.1 nm were used, respectively. The UV-VIS reglon Illustrated reduced signal-to-noise ratlos due to molecular band Interference from both CO, and N20 mobile phases. The near-IR reglon exhlblted molecular band problems wlth an N20moblle phase, only. Detectlon l l m b of 0.8 ng/s for sulfur using a CO, moMle phase ( S / N = 3) are reported with a shot noise llmlted system.
INTRODUCTION Initial successes have been demonstrated coupling supercritical fluid chromatography (SFC) to low-power (40-150W) helium plasma detectors (1-8). Non-metal element selective detection has been achieved for both capillary (1,3,4) and packed column (5, 7) SFC systems. Unfortunately, though successful as initial studies, the low-power systems have had their drawbacks. It has been shown that the SFC mobile phase causes increases in detection limits (3) and decreases in analyte line emission intensity (2,4,6-8) and signal-tenoise ratios (5). Previous work has also cited impedance tuning problems (I, 5) with SFC pressure increases. In a comparison study (7), work with the moderate-power microwave-induced plasma (MIP) demonstrated improved mobile-phase tolerance to low-power SFC-MIP systems. Under high mobile-phase flow rate conditions, the moderate-power SFC-MIP system exhibited significant resistance to both emission signal reduction and impedance tuning problems. In this paper, preliminary characterizations of packed column SFC-MIP using a moderatepower plasma are presented. Initial system figures of merit and examples of multielement analysis are provided as the foundation for continued development of helium plasma atomic emission detection for SFC. EXPERIMENTAL SECTION The SFC-MIP system used for this study is illustrated in Figure 1. Each component is discussed further in the next few paragraphs. SFC System. The pump used for this study was a Lee Scientific Beta pump (Salt Lake City, UT) from a Model 501 SFC instrument. The flow of the SFC mobile phase was directed from *To whom correspondence may be addressed. 'Present address: A. L. Laboratories,Animal Health Division, 400 State St., Chicago Heights, IL 60411.
the pump to a Rheodyne Model 7520 injector (ChromTech,Apple Valley, MN) with a 0.2-& rotor. The injector was connected with a 10 cm X 0,010 in. i.d. Keystone Scientific Slipfree connector No. 31410 (Bellefonte, PA) to a White Associates (Pittsburgh, PA) 150 X 1 mm Deltabond Phenyl 5-pM microbore packed column. The column was housed in a Varian Aerograph No. 204-1C GC from Varian Associates (Palo Alto, CA). The oven temperature for this study was 100 OC. The mobilephase pressure was maintained using an integral packed column restrictor from White Associates (Pittsburgh, PA). The volumetric flow rate range for these restridom was listed as 15-20 ml/min at 55 atm of COz. SFC-MIP Interface. The end of the integral restrictor was placed directly in the analyte channel of the plasma torch as with the earlier SFC-MIP systems. Because of mobile-phase freezing due to JouleThomson cooling, the tip of the restridor was placed as near to the plasma as possible without melting. This distance was approximately 1 cm. The plasma torch used for this study was typical of the moderate-power systems used in our laboratory (9-13). As was discussed in ref 7, Nichrome wire used to alleviate JouleThomson cooling could not be placed inside this torch for it would disturb the helium tangential flow pattern. For this reason, the outside of the torch was wrapped with resitively heated Nichrome wire. The restrictor was centered in the analyte flow channel of the plasma torch with a 1/16 in. stainless steel tube. The stainless tube was centered with Teflon tape wrapping, thereby centering the restrictor and directing the analyte to the center of the plasma. The use of this metal tube also helped conduct the heat of the Nichrome wire-improving upon the Nichrome wire heating technique to avoid Joule-Thomson freezing. Helium Plasma System. The moderate-power plasma torch was inserted in a TNl0resonant cavity similar to that described by Michlewicz et al. (14). The copper resonator had an internal diameter of 88 mm and a depth of 11mm. Impedance matching was performed using adjustable position tuners, as detailed by Haas et al. (15),with parts from an externalthree-stub tuner from Maury Microwave Corp. (Cucamonga, CA). The microwave generator was a 500 W,2450 MHz Micro-Now Model 420B (Micro-NowCorp., Skokie, IL)with a 7 / 8 in. 0.d. foam dielectric Heliax LDF7-50 cable with U5W male connectors (AndrewCorp., Orland Park, IL). Optimum helium flow rates were found to be 22 L/min for plasma support and 0.25 L/min in the analyte flow channel. The emission from the plasma was focused upon a 1-m all-silica bifurcated bundle (Part No. 3B-3.2-1.0)from General Fiber Optics (Cedar Grove, NJ). The fiber optic was used for simultaneous monitoring with two spectrometers: (1) a MPD 850 organic analyzer (Applied Chromatography Systems Ltd.,Bedfordshire, England) fixed slit polychromator and (2) a Model 340E (Spex Industries, Edion, NJ) scanning monochromator. The MPD 850 was used to monitor carbon, hydrogen, and chlorine UV-vis wavelengths at 247.86 (second order), 486.13, and 479.45 nm, respectively. At all three channels, an RCA 1P28Ph4T (Lancaster, PA) detector was used. Carbon and hydrogen emission were monitored with the PMT biased at -960 V. Chlorine emission was monitored with the PMT biased at -1020 V. Oxygen emission at 777.19 nm was monitored to focus the optics of the MPD 850 analyzer. An RCA 1P28 PMT biased at -900 V was used for this channel as well. The Spex monochromator had a 0.34 m focal length. Entrance and exit slits were set at 300 pm. This spectrometer was used for non-metal emission detection in the near-IR spectral region. The monochromator was equipped with a Model 12006-1000X (Spex Industries, Edison, NJ) 1200 groove/"
0003-2700/92/0364-0050$03.00/00 1991 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 64, NO. 1, JANUARY 1, 1992 12
51
A
87 -
6-
5-
4-
--
1
- ,
Flgure 1. SFC-MIP system diagram for the bifurcated fiber optic
system. grating blazed at lo00 nm and a Hamamatsu R406 (Middlesex, NJ) PMT biased at -1200 V. The PMT was powered by a Keithley Instruments high-voltage power supply (Cleveland, OH). The MPDs built-in ampwiers were used in these experiments, offering output ranges of 10 mV and 1 V. The 10-mV range was used for this study. The output of the Spex system PMT was converted to voltage, fiitered, and amplified by an i / v converter and amplifier built at NIU. These voltage outputs were conneded to a home-built, multichannel voltage gain device for computer data acquisition or an Omniscribe Model B5117-5 (Houston Instruments, Austin, TX) strip chart recorder. The output of the voltage gain device was direded to a MetraByte DAS-8 (Taunton, MA) analog-to-digitalconverter (ADC) installed in a Kaypro 286 AT (MicroSolutions,DeKalb, IL) system. The 12-bit ADC has a range of 10 V (-5 to +5 V), or approximately 2.5mV resolution. Labtech Acquire Version 1.2.2 (LaboratoryTechnologies Corp., Wilmington, MA) was used to collect and store the signals in real time. This program was confiied to acquire data at 10 points/s at each channel. The data were manipulated and plotted using LIB, Release 2.2 (LotusDevelopment Corp., Cambridge, MA) and a nine-pin Panasonic KX-1080i or KX-109li (Secaucus, NJ) dot matrix printer. Computer analysis of the data for peak area determination and data smoothing was performed with standard boxcar and trapezoid rule algorithms written with Microsoft Quickbasic Version 4.0 (Redmond, WA). All computer analysis was preceded with a nine-point sliding boxcar. Reagents. The certified ACS grade tetrahydrofuran was obtained from Fisher Scientific (Fair Lawn,NJ). The research grade hexachloropropene, 1,1,2,2-tetrachloroethane,tetrachloroethylene, 1,3-dichlorobenzene,and carbon disulfide were obtained from Aldrich (Milwaukee, WI). 2-Methylbenzothiazole was purchased from E. H. Sargent (Chicago, IL). The decane was obtained from Eastman Kodak (Rochester,NY). The undecane was obtained from Matheson, Coleman, and Bell (Norwood, OH). The Neodene 12,14,16,and 18 samples were from Shell Chemical Co. (Geismar, LA) and the sulfolane was obtained from Phillips Petroleum (Bartlesville, OK). All chemicals were used as received. Helium, rated at 99.995% pure, was from Rockford Industrial Welding Supply (Rockford, IL). The nitrous oxide and carbon dioxide were SFC grade from Scott Specialty Gases (Troy, MI).
RESULTS AND DISCUSSION Choice of Optical Conditions. The use of the bifurcated fiber optic allowed the use of two spectrometers,which allowed for simultaneous analysis in both the UV-vis and near-IR spectral regions. The use of a diffuse, higher powered plasma with the fiber optic had one drawback-the system was detector noise limited. In short, the magnitude of the noise was identical whether radiation from the plasma was observed or the spectrometer entrance slit was blocked. This effect was also seen in the work of Gehlhausen (16). The entrance and exit slits of the MPD 850 system were fiied at 75 pm. Thus,the PMTs were biased at slightly higher than normal voltages (Le. 1020 V for chlorine at 479.45 nm). The Spex spectrometer had adjustable entrance and exit slits. The slit width chosen for this work was 300 pm. Using these
50
C at 247.8nm12ndl I
I
,
70
I
I
90
I
I
I
110
I
130
,
150
TIME(S)
FIgm 2. SFC-MIP of tetrachloroethane (1 pg of Cl) in THF monitored at 247.8 (second order), 479.5, and 837.6 nm. The system used an N20mobile phase, an injection volume of 0.2 pL, isobaric analysis at 80 atm, and a temperature of 100 'C. All chromatographic, plasma, and spectroscopic parameters are detailed in the text.
CI at 837.6nm C at 247.8nm 12061
0
100
200
300
4m
TIME(S)
Flgure 3. SFC-MIP of four chlorinacontaining compounds (1 pg of CI each) and two n-alkanes (20 pg of C each) in THF. Carbon emission was monitored in the second order of 247.8 nm, chlorine emission was monitored at 837.6 nm, and hydrogen emisslon was monitored at 486.1 nm. All condltlons were ldmtical to those in Flgure 2.
widths, the band-pass of the monochromator was 0.75 nm. Chlorine emission in the near-IR region was monitored at 837.6 nm. Although a slightly less intense chlorine emission line than the 911.2-nm line, this region was free of spectral overlap. Chlorine-Selective Detection. Zhang et al. (4) detailed how monitoring chlorine emission a t 479.5 nm provided poor signal-to-noise ratios with SFC-MIP analysis. This observation was concluded to be an effect of excessive background noise in that region. Their work also suggests this phenomenon to be characteristic of low-power plasmas. With the moderate power system, it was found that a 1-pg sample of chlorine gave intense emission at the C1 I emission a t the 837.6-nm line and not at the C1 I1 emission line at 479.5 nm (Figure 2). In Figure 3, the chromatographic separation in N20 of four chlorine-containing compounds (1pg of chlorine each) and two n-alkanes (20 pg of carbon each) is shown. The solvent, tetrahydrofuran, gave an intense emission peak at the carbon, hydrogen, and chlorine channel. The latter signal is not due to chlorine atomic emission, but to CN molecular band emission. While the number of applications in the near-IR spectral region is increasing in atomic spectrometry due to its low number of molecular band interferences as compared to the UV-vis, spectral interference becomes evident at sensitive settings. This problem is true, particularly for nitrogen-containing SFC mobile phases such as NzO, in the
52
ANALYTICAL CHEMISTRY, VOL. 64, NO. 1, JANUARY 1, 1992 10
10
I
9-
THF
6-
-z
6-
-I
1,1,2,2-Ttb~~hlof~Ihane
2-Melhylbenz0Ihiaz0Ie
7-
Sulfolane
6-
m-Dkhbrobnzens
4 3
-
CI al637.6nm
1 Decane
THF
Undecane
0 , 0
100
400
300
200
1
H at 466.1 nm 0
,
0
/
,
l
l
120
60
40
l
160
l
l
200
240
TIME(S)
TIME(S)
Flgure 4. SFC-MIP of four chlorine-containing compounds (1 pg of Ci each) and two n-alkanes (20 pg of C each) in THF. Chlorine emission was monitored at 837.6 nm, and hydrogen emisslon was monitored at 486.1 nm. All conditions were identical to those in Figure 2, except a C02 mobile phase was used.
Figure 6. SFC-MIP of two sulfur-containing compounds (1 pg of S each) in THF. Ail chromatographic, plasma, and spectroscopic parameters are identical to Figure 5, except a COP mobile phase was used.
I1
8 -
l2
7 6 5 4 -
32 -
1 "
0 )
C
,
, 40
,
, 80
,
,
,
120
, 160
,
, 200
,
,
I
0
V
I
-
40
,
I
80
I
I
120
1
I
160
I
I
200
I
t
240
t
I
280
TIME(S)
240
TIME(S)
Flgure 5. SFC-MIP of two sulfur-containing compounds (1 pg of S each) in THF. Sulfur emission was monitored at 921.3 nm, and hydrogen emission was monitored at 486.1 nm. The system used an N20 mobile phase, an injection volume of 0.2 pL, isobaric analysis at 200 atm, and a temperature of 100 O C . All chromatographic,plasma, and spectroscopic parameters are detailed in the text.
presence of large amounts of carbon-containing species. In this case, the solvent is present at a high concentration. After solvent elution, selective detection between the chlorinecontaining analytes and non-chlorine containing n-alkanes was achieved. Intense emission from dichlorobenzenewas detected at the chlorine emission wavelength, and no interference was detected from the decane and undecane peaks. Figure 4 shows the SFC analysis of the same mixture and conditions, except C02is the mobile phase. Carbon emission is not monitored due to the carbon-containing mobile phase. With the use of C02as the mobile phase, no CN interference was detected from the solvent at 837.6 nm. Selective detection of chlorine-containing and non-chlorine species is enhanced with this choice of mobile ph&e. Sulfur-Selective Detection. Because of ita environmental and industrial importance and the large number of sulfurcontaining compounds that cannot be separated by GC, sulfur-selectiveSFC detection was characterized at 921.3 nm. Figures 5 and 6 illustrate isobaric SFC analysis at 200 atm of two sulfur-containing compounds in tetrahydrofuran with N20and GO2mobile phases, respectively. Hydrogen emission at 486.1nm was used to monitor solvent elution. The analytes, each containing 1 pg of sulfur, were not detected at this wavelength, due the higher pressures required for the elution of the analytes. Higher pressures led to elevated mobile-phase
Flgure 7. SFC-MIP of a Necdene mixture (5% each) In THF. An isobaric analysis at 135 atm of N 2 0 was used. All other chromatographic, plasma, and spectroscopic parameters are detaHed in the text.
flows into the plasma, increasing molecular band interference from N2in the 486.1-nm region. As expected, higher pressure decreased the elution time of the peaks. As well, higher pressures led to more Gaussian-shaped peaks. Note that intense peaks are present in Figure 5 at the 921.3-nm sulfur channel during tetrahydrofuran elution. This response is due to CN band emission from the nitrogen of the mobile phase and the carbon from the solvent. The separation of the analytes in COz is not as good with this mobile phase, but as shown in Figure 6, no molecular band emission at 921.3 nm for tetrahydrofuran was detected with the C02mobile phase. In ref 3,4, and 7, low-power plasma systems began to show a loss in sensitivity at mobile-phase pressures of 160-200 atm. The moderate-power SFC-MIP exhibits a high tolerance to changing mobile-phase pressures from 150 to 300 atm. Within experimental error, the peak areas from the sulfur-selective detection of sulfolane at 921.3 nm did not change significantly in the pressure range 150-300 atm. These results indicatethat the analyst should be able to avoid the variable response factors in the work of Zhang et al. (4) and inferred from the data of Skelton et al. (3). Carbon- and Hydrogen-Selective Detection. A model mixture of Shell Neodene samples was examined. These samples are a-unsaturated olefins in the Clz to C18range. These samples were analyzed with isobaric and pressureprogramming conditions. Isobaric analyses were recorded via a computer. Pressure programs were generated with the Model 501 SFC system software (Lee Scientific, Salt Lake City, UT).
ANALYTICAL CHEMISTRY, VOL. 64, NO. 1, JANUARY 1, 1992
59
SFC-MIP of SHELL NEODENE MIXTURE Mobile Phase: N20Temperature: 100' C Pressure: 125 atm for 15 sec 15 atmlmln ramp Wavelength: 486.1 nm-H(I) Power: 500W He Flow: 22Umln support 0.25 Umln analyte
THF
12
0.0 4
12
16
14
I
16
SHELL NEODENE SAMPLE
Flgwe 9. C/H ratios for each Neodene sample. Carbon emlaskn was monitored at 247.8 nm (second order), and hydrogen emission was monitored at 486.1 nm. The Isobaric analysls for Neodene 12 was at 120 atm, Neodene 14 was at 130 atm, "e16 was at 150 atm, and Neodene 18 at 165 atm. All other chromatographic,plasma, and spectroscoplc parameters are identical to Flgure 8.
"
0.15
-4 >
"
0.10
z" '
12
&/ I 0
I
I
60
120
180
Flgure 8. SFC-MIP of a Neodene mixture (5% each) In THF. Pressure programming was used with the Initial pressure at 125 atm for 15 s, followed by a 15 atm/min pressure ramp. Hydrogen emission at 486.1 nm was monitored with a full scale of 0.1 V. All other chromatographic,plasma, and spectroscopic parametersare detailed in the text.
Figure 7 is an isobaric analysis of a 5% (v/v) mixture of each of the Neodene 12-18 samples. The range of boiling points and molecular masses from the major components are approximately 400-600 OF and 168-252 amu, respectively. The mobile phase for Figure 7 was NzO. Associated with the 247.8nm carbon emission wavelength was a great deal of noise, due to molecular band emission. This interference is most likely from Nz and NO band emission. Hydrogen-selective detection was monitored as well, giving much better signalto-noise ratios. The Neodene 16 and 18 peaks are broad due to the low pressure used. The chromatography was improved significantly with pressure programming, as shown in Figure 8. The mixture was analyzed with a 15 atm/min ramp, starting 15 s after injection from an initial pressure of 125 atm to a maximum pressure of 170 atm. Hydrogen-selective detection at 486.1 nm was used. Pressure programming improved the peak shapes of the analysis, with the last eluting peak (Neodene 18) more Gaussian in shape. The baseline found with pressure programming in a moderate-power SFC-MIP also appears not to rise as dramatically with the ramp, as in refs 1 and 5. In Figure 9, the carbon/hydmgen peak area ratios from the N20mobile-phase isobaric analysis of Neodene 12 at 120 atm, Neodene 14 at 130 atm, Neodene 16 at 150 atm, and Neodene 18 at 165 atm are plotted against carbon number (equivalent
14
18
16
0.05
-
1"
SHELL NEODENE SAMPLE CIH Ratio
.t- Noise at 2478nm
4
N o w a1 466 1nm
Figure 10. Noise at 247.8 nm (second order) and 486.1 nm during the C/H ratio analysis. The average CIH ratio for each Neodene sample was also plotted.
to the Neodene number). Except for Neodene 18, the peak areas, measured at different pressures, are not significantly different. This plot again demonstrates that the varying response factors in earlier, low-power systems have been effectively eliminated. The C/H ratio does decrease for Neodene 18. The reason for the smaller value may be seen by examination of Figure 10. In this figure, the average C/H ratio, the noise at second order of 247.8 nm (carbon), and the noise at 486.1 nm (hydrogen) are plotted for the pressure used with each Neodene sample. The noise was calculated as the standard deviation of the baseline for greater than 50 data points. Noise at the carbon emission wavelength for the Neodene 18 analysis was much higher than for the preceding analytes. Thus,the error for Neodene 18 in Figure 10 exceeds those of the other values. It is likely that this noise is caused by the high flow of NzO at the pressure of Neodene 18 elution. Consequently, the apparent decreasing C/H ratio of Figure 9 may not be real, but be a function of measurement error. The SFC-MIP analysis of the Neodene samples and a C02 mobile phase was performed. Because of the mobile phase, carbon emission was not monitored. Hydrogen-selective detection at 468.1 nm was used. The initial pressure was 130 atm, held for 15 8 , and ramped at 20 atm/min. Although not shown, the resultant chromatogram was nearly identical to that of Figure 8. System Figures of Merit. Table I lists the detedion limits ( S I N = 3) for the moderate-power SFC-MIP. Noise was defined as the standard deviation of 10 points along the chromatographic baseline taken at 1-s intervals. The detedion limits for this study were established with both NzO and COz mobile phases at 80 atm for chlorine using tetrachloroethylene, except for the 919.2-nm value, in which the later eluting
54
ANALYTICAL CHEMISTRY, VOL. 64, NO. 1, JANUARY 1, 1992
Table 1. Detection Limits" for a Moderate-Power SFC-MIP
Table 111. Selectivity in N20
System
S C1 C1
H C
ng
A, nm
anal*
ngls
ng
" S I N = 3;shot noise limited. *NA = not applicable.
Table 11. Comparison of SFC-MIP Detection Limits (SIN = 3) for Chlorine and Sulfur Using COz plasma
power,
design
W
analyte
A,nm
sulfatronO RPDb
110 100
sulfur
921.3 837.6 921.3 837.6 921.3 837.6 921.3
nm
ng/s
12.9 0.8 7.0 0.5 921.3 16.6 1.4 17.8 1.5 837.6 364.6 34.5 143.1 16.3 919.2 91.4 4.4 39.6 2.6 486.1 66.9 NAb NA 247.8 (second order) 1379
helium
wavelength,
coz
NzO
reported detection limit
921.3 837.6 919.2
anal*
sulfur chlorine chlorine
chlorine Sulfur
TMoloC
70
chlorine
TMo1od
500
sulfur chlorine Sulfur
"Reference 1: converted from SIN = 2, capillary column system. bReference 3 converted from SIN = 2, capillary column system. Reference 4 converted from SIN = 2,capillary column svstem. dThis work shot noise limited, packed column system.
1,1,2,2-tetrachloroethanewas used. The detection limits were established with both N20 and C02 mobile phases at 80 atm for sulfur using carbon disulfide, and at 100 atm for hydrogen using undecane. The detection limit for carbon was established with N20 at 100 atm using undecane. Detection limits are comparable with both mobile phases, except for the case of carbon. In general, detection limits for S and C1 were in the nanogxam per second range. Due to the addition of SFC mobile phase to the plasma, these valuea are somewhat greater than those obtained with GC-MIP systems. Comparisons of the reported C1 and S detection limits with SFC-MIP using a C02 mobile phase are found in Table 11. Though more sensitive when common elements are compared than the previous work in packed column SFC-MIP (5),the detection limits are not as sensitive as those reported with capillary column SFC-MIP systems (I, 3,4). The previous packed column work used UV-vis emission lines. Improvement can be expected with this system with applications of the near-IR region spectral region. Comparing the detection limit for sulfur (500pg/s at SIN = 3) to the 180 pg/s (converted from SIN = 2 to S I N = 3) detection limit reported by Zhang et al. (4)indicates that the moderate-power SFC-MIP system is approaching the detection limitsof capillary column SFC-MIP systems using low powers with reduced sample injection constraints. On the basis of noise source calculations, 1 order of magnitude improvement in light throughput to the chlorine PMT at 837.6 nm would half the detection limit. Such a case may be possible by redesigning this system without the fiber optic. With more light reaching the PMT, not only would the detection limits approach the range established for the capillary SFC-MIP systems (1, 3, 4 ) but also the mobile phase induced signal reduction characteristics seen in the low-power systems of ref 3, 4,6,and 7 could be avoided. A calibration plot for chlorine at 837.6 nm was constructed in the concentration range from 1 to 100 N.At 80 alm of COz, the correlation coefficient for the plot was 0.9935. The log-log slope for the tetrachloroethylene analysis was 1.04, indicating acceptable linearity. More severe conditions were chosen for
selectivity (ht)
460 180 40
640 260 20
the sulfur calibration plot at 921.3 nm. Using a COz mobile phase at 200 atm, a calibration more applicable to standard SFC analysis was achieved. Using sulfolane as the source for sulfur-seledive detection, the plot had a correlation coefficient of 0.9939 and a log-log slope of 1.09 in the mass range 1-100 fig of S. Again, acceptable linearity resulted. Table III illustrates the selectivity of chlorine to carbon and sulfur to carbon using an N20 mobile phase at selected wavelengths. Using the equation SXJC
37.5 pg/s 174.Opgls 89.4 pg/s 62.7 pg/s 180.0 pg/s 1.5 ng/s 0.5 ng/s
selectivity (area)
= mcIx/mxIc
where S is the selectivity, m is the mass,and I is the emission intensity, values are reported for selectivity are based upon both peak height and peak area. The subscripts C and X indicate carbon and analyte, respectively. The interference problem is a result of CN molecular band emission. The decreased selectivity at 919.2 nm for chlorine correlates directly to the poor selectivity as compared to 837.6-nmC1 I emission.
CONCLUSIONS With low detection limits, a non-metal atomic emission detector which is much more tolerant to SFC mobile-phase changes has been implemented. The packed column SFCMIP system using a moderate-power plasma provided promising resulta for carbon-, hydrogen-, chlorine-, and sulfur-selective detection. Sulfur detection limits as low as 0.8 ng/s were reported (SIN = 3) for a shot noise limited system. Improvements to the o p t i d system should reduce these limits to the range established by the low-power, capillary column systems. Using the carbon emission line in the near-IR spectral region (940.6nm) should improve upon its limit of detection. Current work is underway focusing upon improving the emission signal and detection. Because the moderate power plasma has expanded the practical range of SFC-MIP analysis, further applications for atomic emission detection for SFC are assured. ACKNOWLEDGMENT We thank Larry Gregerson and Ed Hyland of Northern Illinois University for their help in the construction of the various devices and torches used in this work, Charlea Caldwell of Northem IllinoisUniversity for the design and construction of the voltage gain device, and Curt White of White Associates for helpful SFC discussions. Registry NO.Clz, 7782-50-5;S,7704-34-9;C,7440-44-0;Hz, 1333-74-0;tetrachloroethylene, 127-18-4;1,1,2,2-te~achloroethane, 79-34-6;m-dichlorobenzene, 541-73-1; hexachloropropene, 188871-7;Zmethylbendhiamle,120-752;sulfolane, 126-33-0;Neodene 12, 112-41-4;Neodene 14, 1120-36-1;Neodene 16, 629-73-2; Neodene 18,112-88-9. REFERENCES (1) Luffer, D. R.; Galante, L. J.; David, P. A.; Novotny, M.; HieftJe,G. M. Anal. Chem. 1088. BO, 1385. (2) Galante. L. J.; Selby, M.; Luffer, D. R.; Hleftje. 0. M.; Novotny, M. Anal. Chem. 1988. 60, 1370. (3) Skelton, R. J.; Farnsworth, P. B.; Markldes, K. E.; Lee, M. L. Anal. chem.ises. 81, 181s. 14) Zhana. L.: Camahan. J. W.: Wlnans. R. E.: Nelll. P. H. Anal. Chem. l091,~Ss; 212. ' (5) Motley, C. 0.;Long, 0. L. J . Anal. At. Specttom. 1990, 5 , 477. (6) Webstw, 0. K.; Camahan, J. W. Appl. Specfrapc. 1980, 44, 1020. .
I
Anal. Chem. 1902, 64, 55-60 (7) Webster, G. K.; Carnahan, J. W. Appl. Spectrosc. 1991, 45. (8) Rlvke, 6.; Mermet, J. M.; Derauz, D. J. J. Anal. At. Spectrom. 1988,
3,551. (9) Michlewicz, K. G.; Carnahan, J. W. Anal. Chem. 1985, 57, 1092. (IO) Michlewicz, K. G.;Carnahan, J. W. Anal. chlm. Acta 1988, 786, 275. (11) Mlchlewicz, K. G.;Carnahan. J. W. Anal. Chem. 1988, 58, 3122. (12) Oehlhausen, J. M.; Carnahan, J. W. Anal. Chem. 1989, 67, 674. (13) Wu, M.; Carnahan, J. W. Appl. Spectrmc. 1990, 44, 673. (14) Mlchlewicz, K. G.; Urh, J. J.; Carnahan, J. W. Spectrochlm. Acta 1985, 408, 493. (15)Haas, D. L.; Carnahan. J. W.; Caruso, J. A. Appl. Spectrosc. 1983,
37, 82.
55
(16) Gehlhausen, J. M.; Carnahan. J. W. Anal. Chem. 1991, 63, 2430.
RECEIVED for review May 24,1991. Accepted October 8,1991. This work was financially supported in part by a Grant-in-Aid of Research from Sigma Xi and a Northern Illinois University Graduate School Dissertation Completion Award. Presented in part at the 42nd Pittsburgh Conference and Exposition on Analytical Chemistry and Applied Spectroscopy, Chicago, 1991, Paper No. 537.
Planar Waveguide Immunosensor with Fluorescent Liposome Amplification Steven J. Choquette,* Laurie Locascio-Brown, and Richard A. Durst'
National Institute of Standards and Technology, Gaithersburg, Maryland 20899
A regenerable planar waveguide lmmunosensor for the cHnkal analyte theophylline has been developed. Regeneration Is accomplished under flow conditions uslng a moderate afflnity antibody, and muitlple analyses can be performed with a single wavegulde sensor. Sensors capable of more than 15 sequential measurements have demonstrated better than 10% precision. The use of theophylllnelabeledllposomes In this Competitive immunoassay provides 1 order of magnitude greater dgnal enhancement over theophylline derivatlzed with fluoresceln.
INTRODUCTION Planar optical waveguides are typically thin dielectric films (0.1-10 pm) supported on a rigid substrate of lower refractive index. In the thicker (1 mm), geometrically similar total internal reflection (TIR) element, the propagating light makes discrete contacts with the dielectric interface. Light guided by a planar or fiber waveguide, however, is more accurately described as a continuous energy distribution along the path of propagation ( I ) . The potential advantage of planar and fiber waveguides over TIR elements is the increased optical path length of the evanescent wave. The advantages of planar versus fiber waveguides are increased durability, potential for miniaturization, and ease of fabrication using a variety of materials and methods. As a result, the optical properties of planar guides may be potentially engineered for a particular chemical measurement, an attribute that is currently not enjoyed by contemporary fiber-based sensors. The first planar waveguide sensors to be developed were based upon attenuation of the guided beam (absorption spectrometry) (2-4). Interest in these sensors waned with the development of inexpensivefiber components until the early 1980s when planar waveguides were "rediscovered" for applications in Raman spectroscopy (5,6). Applications of planar waveguides to absorption spectrometry are continuing (7-11) and planar waveguide fluorescence techniques are currently under investigation (12, 13). Recent novel applications of planar waveguides include refractive index sensors based on
* To whom correspondence should be addressed.
Present address: Cornell University, NYAES, Geneva, NY 14456.
input grating couplers (14) and integrated optic Mach Zehnder interferometers (15). These sensors respond to general optical phenomena (Le. refractive index and color) and, as such, lack chemical specificity. Attachment of antibodies to an optical sensor surface however can provide the chemical specificity required for analytical measuremeats. Sutherland et al. (16)were among the first to demonstrate both slab and cylindrical total internal reflection fluorescence (TIRF) immunosensors. Highly sensitive fluorescence immunosensorsemploying evanescent wave and distal tip excitation were then extended to multimode and single-mode fibers by Vo Dinh (In,Tromberg and Sepaniak (I@, and Bright (19),among others. Planar waveguides employing immunospecific reactions have also recently been reported with evanescent fluorescence excitation (20, 21). Quantitative analytical measurements made with immunosensors require either regeneration of the antibody or calibration based on a series of single-use devices. High-affinity antibodies are typically used because the analytical sensitivity of the assay is related to the affinity constant. However, facile dissociation of the antigen-antibodycomplexes formed with such antibodies is not favored, making regeneration of the intact antibody difficult to achieve. The use of chaotropic reagents to dissociate the antigen-antibody complex has often resulted in significant loss of antibody activity when used with immunosensors, with the recent work of Bright (19) being a notable exception. As a result, the majority of immunosensors are single use devices and calibration curves are obtained using multiple disposable sensors. The higher dissociation rates exhibited by moderate affinity antibodies (K,< lo8) make regeneration of complexed antibodies feasible using mass action in a flow system. Recent investigations in this laboratory, using a flow injection immunoanalysis (FIIA)system (221, have demonstrated quantitative regeneration of a surface-immobilized antibody in an affinity column. Flowing antigen-free buffer over the antibody-antigen complex for a period of time greater than 10 half-lives ( N 15min) of the dissociation rate constant can lead to quantitative regeneration of the immobilized antibody. We have adapted this approach to optical waveguide sensors and developed an immunosensor for theophylline in a competitive assay using antigen-tagged liposomes. Liposomes are spherical phospholipid structures that possess an internal cavity. For this assay, the aqueous interior
0003-2700/92/0364-0055$03.00/00 1991 American Chemical Society