Thermionic detection in microcolumn liquid chromatography

Quantitative reproducibility study with automated microcolumn liquid chromatography. Hernan J. Cortes , Jeffrey R. Larson , Gerald M. McGowan. Journal...
0 downloads 0 Views 960KB Size
2298

Anal. Chem. 1983, 55,2296-2302

Thermionic Detection in Microcolumn Liquid Chromatography V. L. McGuffin’ and Milos Novotny* Department of Chemistry, Indiana University, Bloomington, Indiana 47405

A dual-flame thermionic detector has been developed that Is compatible wlth mlcrocolumn llquld chromatography. The total mlcrocolumn effluent was concentrlcally nebullzed and aspirated dlrectly into a prlmery hydrogen-air dlffuslon flame. Phosphorus and nltrogen compounds were then selectlvely detected by monltorlng the conductlvlty of the secondary flame In the presence of a rubidlum glass bead. The sensltlvlty for phosphorus compounds was 2 X lo-’‘ g/s at the maxlmum of the Gaussian peak, and response was llnearly related to concentratlon over at least 3 decades.

The importance of selective detection has been well established in gas chromatography (GC) with such devices as the electron capture, flame photometric, thermionic, and microwave-induced plasma detectors (I). Molecules which contain the requisite heteroatoms or functional groups may be detected and quantitated with minimal interference in highly complex samples or in the presence of coeluting solutes. The excellent sensitivity and specificity of the aforementioned detectors would be equally valuable in high-performance liquid chromatography (HPLC);however, their implementation has been hindered by numerous technological problems. Specifically, the continuous introduction of aqueous and organic mobile phases at the typical flow rates encountered in HPLC (1mL/min) causes a severe disturbance of the selective ionization or spectral emission of the detected species. Both direct effluent introduction and a transport device have been utilized to interface HPLC with such detehrs, but, for various reasons, neither approach has been entirely successful (2-4). The recent introduction of miniaturized HPLC systems may provide an alternative solution to these technologicalproblems. The total microcolumn effluent, on the order of microliters per minute, may be directly introduced to a flame- or plasma-based detector with minimal disruption of the chemical and physical phenomena responsible for solute detection. Previously, both flame ionization (5, 6) and flame emission (7) have been successfully employed as detectors for microcolumn HPLC. In this investigation, a thermionic detector has been modified for the selective detection of phosphorusand nitrogen-containing compounds separated by microcolumn liquid chromatography. The thermionic detector was first described for gas chromatography by Karmen and Giuffrida in 1964 (8). The GC column effluent was combusted in a hydrogen-air diffusion flame and the flame conductivity was monitored in the presence of an alkali salt. The collector electrode was negatively polarized with respect to the alkali source, and the alkali cation was believed to be the detected species. According to Sevcik (9),thermal energy was responsible for ionization of the alkali metal, thereby creating a standing ion current in the flame. This ion current was enhanced or diminished in the presence of certain heteroatoms due to electron-capture processes, formation of thermally stable products, or formation of products with low ionization potential. Since the alkali was Current address: Departmentof Chemistry, Stanford University, Stanford, California 94305.

continuously consumed in the detection process, sensitivity was gradually reduced and the lifetime of the alkali source was extremely short. Kolb and Bischoff (10) later described a “flameless” thermionic detector in which the collector was positively polarized with respect to the alkali source. The proposed mechanism for this detector configuration involved the formation during combustion of a radical species, such as CN. for nitrogen compounds and PO. or POz. for phosphorus compounds, which was capable of abstracting an electron from the alkali metal to form a stable anion. The anion was subsequently detected at the collector electrode, and the alkali cation was recaptured by the negatively charged alkali source. Additional improvements in the design and operation of thermionic detectors for gas chromatography have been reviewed by Brazhnikov and co-workers (11). Thermionic detection has also been described for liquid chromatography using a transport interface to facilitate solvent evaporation (I2,13). Direct introduction of the HPLC effluent to the sensitized flame has not as yet been demonstrated, presumably due to solvent interferences. A dual-flame thermionic detector has been developed in our laboratory which minimizes solvent interferences by spatially separating the fundamental flame processes, such as nebulization, desolvation, and decomposition, from the analytical measurement of ion current. This detector appears to be extremely promising for the selective detection of organic phorphorus and nitrogen compounds separated by microcolumn liquid chromatography. EXPERIMENTAL SECTION Thermionic Detector. The dual-flame thermionic detector, illustrated in Figure 1, is of the type conceived by Kolb and Bischoff (10) and was constructed in-house from a modified commercial instrument (Part No. 023-0301,Perkin-Elmer Corp., Norwalk, CT). The stability of the thermionic detector was dependent upon the carefully regulated flow of high-purity flame gases. The gas flow rates were controlled by single-stage regulators at the cylinder and by calibrated rotameters (Model R-2-15-D-CO,Brooks Instrument Division, Emerson Electric Co., Hatfield, PA). Molecular sieves (Type 3A and 4A) were used to remove trace contaminants from the air and nitrogen cylinders. Hydrogen, the fuel gas, and nitrogen, a noncombustible nebulizing gas, were premixed in a baffle and flowed through a stainless steel flame jet. Air was supplied to the primary diffusion flame through a fritted stainless steel disk at the base of the flame jet. A glass or fused-silica capillary (50 pm i.d., 10 cm length), connected at one end to the HPLC microcolumn with PTFE tubing, was inserted through the burner base and was extended approximately 1 mm above the top of the flame jet. The total microcolumn effluent was nebulized by the concentric flow of gases over the capillary orifice and was aspirated directly into the primary diffusion flame. The combustion products from the primary flame were combined with additional fuel and were swept into the analytical flame, which was also supplied with air by diffusion. The analytical flame was not allowed to ignite and burn freely; the best response was obtained when combustion was carefully controlled in the region near the alkali source. When the analytical flame was accidentally ignited,a large increase in background ion current was observed.

0003-2700/S3/0355-2296$01.50/00 1983 American Chemical Society

ANALYTICAL CHEMISTRY, VOL. 55, NO. 14, DECEMBER 1983

COLLECTOR ELECTRODE RUBIDIUM BEAD ANALYTICAL FLAME

HYDROGEN AND AIR INLETS

I 1 @&

PRIMARY FLAME

AIR INLET CAPILLARY HYDROGEN AND NITROGEN INLETS

Figure 1. Schematic diagram of the dual-flame thermionic detector for microcolumn liquid chromatography.

1

CAPILLIRY COLM

Figure 2. Schematic diagram of the chromatographic system, thermionic detector, and ancillary electronic equipment.

An alkali bead (rubidium nitrate), fabricated according to the procedure of Lubkowitz et al. (14),was positioned 1-2 mm directly above the analytical flame jet. This bead was electricallyheated by a constant-current source and negatively polarized by a high-voltage power supply (Model 240A, Keithley Instruments, Inc., Cleveland, OH). The electricalconductivity of the analytical flame was monitored with a cylindrical collector electrode,located 6 mm above the flame jet. The resulting ion current was amplified with an electrometer (Part No. 045-0989, Perkin-Elmer Corp., Norwalk, CT), fiitered to remove high-frequency noise, and finally displayed on a strip-chart recorder (Model 355, Linear Instruments Corp., Irvine, CA). A block diagram of the complete analytical system is shown in Figure 2. Chromatographic System. Column Preparation. The detector performance was evaluated with several types of HPLC microcolumns: small-bore packed columns, packed capillaries, and open-tubular columns. Small-bore packed microcolumns were prepared from fusedsilica capillary tubing of 0.2 mm i.d. and 1-2 m length. The tubing was packed under high pressure (400 atm) with an acetonitrile slurry of the reversed-phase chromatographic material (Spherisorb 55 ODs, Phase Sep, Hauppauge, NY) (15, 16). Packed capillary columns were prepared according to the previously described procedure (17,18). Standard-bore Pyrex glass tubing was packed with a silica adsorbent (30 pm LiChrosorb Si-100,E. Merck Reagents, Darmstadt, G.F.R.) and, subsequently,

2297

was extruded to an inner diameter of approximately 70 pm with a commericalglass-drawingapparatus (Model GDM-lB, Shimadzu Seisakusho Ltd., Kyoto, Japan). An octadecylsilane stationary phase (Petrarch Systems, Inc., Levittown, PA) was then chemically bonded in situ to the silica substrate (19). Open-tubular fused-silica capillaries (Scientific Glass Engineering, Inc., Victoria, Australia) of 30-60 pm i.d. were also tested to ascertain compatibility with the flame-based detector. The open-tubular columns were employed without further surface modification. HPLC Instrumentation. A high-pressure syringe pump (Model 8500, Varian Instrument Division, Palo Alto, CA) was operated under the conditions of constant pressure and provided uniform solvent flow through the HPLC microcolumns with minimal pulsation. A direct sampling method was utilized for the optimization and characterization of the thermionic detector. This injection method employed a short length of stainless steel or fused-silicacapillary tubing, containing 0.1-0.2 pL of the sample solution, which was connected to the microcolumn with PTFE tubing (20). For the determination of detector variance, a low-volume, “heart-cut” sampling system was employed (21). A variable-wavelength UV-absorbance detector (Model Uvidec 10011, Jasco, Inc., Tokyo, Japan) was modified to permit oncolumn detection through the fused-silica capillary columns. The illuminated volume of this detector with 30 pm i.d. tubing was approximately 3 nL, and the variance was calculated to be 0.7 nL2. The response and variance of the UV-absorbance detector were compared with those of the dual-flame thermionic detector. Reagents. The model solutes utilized to characterize the thermionic detector for phosphorus, nitrogen, sulfur, and halogen-selective response were of reagent-grade purity and were obtained from the Aldrich Chemical Co. (Milwaukee, WI). Organophosphorus pesticide standards were “qualitativegrade” and were purchased from the Anspec Co. (Ann Arbor, MI). Diazinon (phosphorothioic acid, 0,O-diethyl 0-[6-methyl-2-(1methylethyl)-4-pyrimidinyl]ester; 98% purity), Ethion (phosphorodithioic acid, S,S’-methylene O,O,O’,O’-tetraethylester; 95% purity), Guthion (phosphorodithioic acid, 0,O-dimethyl S-[(4oxo-1,2,3-benzotriazin-3(4H)-yl)methyl]ester;99% purity), and Zolone (phosphorodithioic acid, S-[ (6-chloro-2-oxo-3(2H)-benzoxazolyl)methyl] 0,O-diethyl ester; 98% purity) were employed in this investigation. Organic compounds containing hydroxyl and amine functional groups were derivatized to incorporate phosphorus using either 0,O-dimethyl phosphorochloridothioate (Orgmet, Inc., E. Hampstead, NH) (22-24) or 0,O-diethyl phosphorochloridate (Aldrich) (25). The compounds derivatized in this manner included several hydroxysteroids (Sigma Chemical Co., St. Louis, MO) and aliphatic alcohols (Matheson Coleman & Bell, Norwood, OH). The phosphorothioate derivatives of estra-l,3,5(1O)-triene-3,17P-diol,4-pregnen-21-01-3,2O-dione, cyclohexylamine, and 1,3,5-benzenetriol were obtained through the courtesy of Karl Jacob, Ludwig-Maximilians-Universitat Munchen, G.F.R. All organic solvents utilized in this investigation were A.C.S. spectral grade (Fisher Scientific Co., Fair Lawn, NJ); water was deionized, distilled, and finally redistilled from alkaline permanganate. RESULTS AND DISCUSSION Detector Optimization. There are many variables, some subtle and some obvious, that influence the sensitivity and selectivity of thermionic detectors (11). For example, the shape and position of the collector electrode are known to influence detector response, as is the general design of the flame housing proper. Furthermore, the bead composition, size, shape, and position can significantly affect the detector response. These structural parameters were maintained constant insofar as possible throughout this preliminary investigation. The thermionic detector was nominally optimized in the phosphorus-selective mode by independent linear variation of the gas flow rates, bead-heating current, bead potential, and flame-jet potential to obtain the maximum signal-to-noise (RMS) ratio. This optimization was performed

ANALYTICAL CHEMISTRY, VOL. 55, NO. 14, DECEMBER 1983

2298

,

120

,

~

,

1

~

'

~

~

I

"

'

'

I

L

'081 HyDRoGEN4 :\ 88

28

~

~

"

"

"

"50 "

"

"

1

I88 ' "

FLOU RATE

'

'

I60

t

'

208

(nL/HIN>

Flgure 3. Thermionic detector response as a function of hydrogen and nitrogen flow rates to the primary flame: column, packed capillary column (70 pm i.d., 5 m length) prepared with 30 pm LiChrosorb Si-100; mobile phase, methanol (40 atm); solute, trimethyl phosphate (50 ng of phosphorus).

I80

I60

FLOU RATE

288

250

308

(mL/tlIN)

Figure 5. Thermionic detector response as a function of hydrogen and air flow rates to the analytical flame. Chromatographic conditions are described in Figure 3.

a highly oxidizing flame with a fuel-to-oxidant ratio of 0.6 under optimum conditions. This suggests that combustible organic solvents served as an auxiliary fuel and played an important role in the flame chemistry. This conclusion was substantiated by the faint but distinct odor of formaldehyde that was present whenever methanol was employed as the mobile phase. The gas flow rates to the analytical flame did not alter the detector signal intensity, but had a significant effect upon the background noise level. The noise level increased with an increase in fuel flow or with a decrease in oxidant flow. Hence, the signal-to-noise ratio was improved at low hydrogen flow rates and at high air flow rates, as demonstrated in Figure 5. These results indicated that the principal effect of the analytical flame gases was to maintain the alkali bead at a cool temperature. This conclusion was later confirmed in an investigation of the effect of bead temperature upon detector response. The bead temperature was controlled by varying the applied current from 1.5 to 3.5 A. Although the signal intensity remained relatively constant, the background noise increased exponentially with a linear increase in bead-heating current. Consequently, the signal-to-noise ratio decreased dramatically with increasing bead temperature. These results directly conflict with those reported previously for the thermionic detector in gas chromatography, wherein the response increased with bead temperature (26,27). It is expected that the addition of combustible organic solvents substantially increased the operating temperature of both the primary flame and the alkali source. Therefore, the alterations in detector response with bead current and with gas flow rates, which seem to conflict with previous reports, are most likely a consequence of the combustible mobile phases employed for liquid chromatography. It was also necessary to optimize the applied potential between the bead and the collector electrode, which influenced the efficiency of ion collection. The potential of the bead with respect to the collector electrode was varied from 0 to -300 V, and the optimum signal-to-noise ratio was obtained at -275 V. The potential of the analytical flame jet with respect to the bead was also varied to ascertain its effect upon the detector response. The highest sensitivity was obtained when the flame jet was at the same potential as the alkali bead. However, the best signal-to-noiseratio was obtained when the flame jet was grounded. By grounding the flame jet, the background ion current created by combustion processes in the primary flame was substantially reduced. The background noise level of the detector was also improved through the use of a Faraday cage to minimize environmental interferences. Detector Characterization. Detection Limits and Linear

k-

I 680

80

FLOU RATE

808

CmL/HIN>

Figure 4. Thermionic detector response as a function of air flow rate to the primary flame. Chromatographic conditions are described In Flgure 3. Solute was trimethyl phosphate (50 ng and 1 ng of phosphorus).

by using trimethyl phosphate as a nonretained model solute and methanol as the mobile phase. The capillary and flame jet dimensions were adjusted for proper nebulization as described previously (7).It was deemed necessary to add a noncombustible gas, such as nitrogen or helium, to the primary flame in order to further improve the nebulization efficiency. As indicated in Figure 3, the nebulizing gas did not seem to have a significant effect upon the signal-to-noise ratio and could be added in flow rates of 110 to 170 mL/min with equivalent results. The spatial separation of the primary and secondary flames permitted the independent optimization of flame conditions for the combustion and analytical measurement processes. In the primary flame, the flow rate of fuel gas was varied between 60 and 110 mL/min, as illustrated in Figure 3. It was difficult to sustain the flame at hydrogen flow rates less than 50 mL/min when solvent was aspirated. Consequently, the decrease in signal-to-noise ratio at low flow rates primarily reflects the higher base line noise resulting from flame instability. At higher flow rates, the signal intensity decreased consistently with increasing hydrogen flow rates, whereas the noise level remained relatively constant. It was therefore desirable to employ the lowest hydrogen flow rate possible consistent with stable detector operation (ca. 75 mL/min). The air flow rate to the primary flame was varied between 250 and 800 mL/min, as illustrated in Figure 4. At flow rates up to 600 mL/min, the oxidant had a significant effect upon the signal intensity, whereafter an increase in the noise level was predominant. The primary combustion flame was thus

ANALYTICAL CHEMISTRY, VOL. 55, NO. 14, DECEMBER 1983

Dynamic Range. The response of the thermionic detector was characterized under optimum conditions using trimethyl phosphate as a nonretained model solute and methanol as the mobile phase. The detection limits were established by using the method previously described (7). The minimum detectable quantity of phosphorus was determined to be 500 pg, and the corresponding mass flux at the maximum of the Gaussian peak was 22 pg/s. This detection limit was measured at a signal-to-noise (RMS)ratio of 7, which corresponded to the 99.9% confidence level. The response of the thermionic detector was compared with that of the UV-absorbancedetector having a flow cell of 30 pm i.d. and 3 nL volume. The signal-to-noise ratio of the two detectors was equivalent for the pesticide Diazinon, whose molar absorptivity was estimated from the literature to be 4300 L/(mol cm) at 254 nm (28). Thus, the thermionic detector compared favorably with UVabsorbance detection for solutes of moderate absorptivity. The response of the thermionic detector was linearly related to the injected solute mass from the detection limit to at least 800 ng of phosphorus (50 ng/s). Therefore, the linear dynamic range was greater than lo3, as is typical of this detector in gas chromatography as well (21). Selectivity. The selectivity of the thermionic detector for phosphorus-containingcompounds was also investigated with trimethyl phosphate as a model solute and methanol as the mobile phase. This measurement entailed a comparison of the base line displacement caused by the introduction of the mobile phase at a known flow rate with the displacement caused by the injected solute. For a freshly prepared alkali bead, the detector did not response measurably to the mobile phase, and selectivity was greater than lo6. As the bead aged, however, the detector response to the solvent increased, whereas the response to phosphorus remained constant. After the bead had been continuously operated for 3 weeks, the selectivity for phosphorus compounds was reduced to 4 X lo6, and after 6 weeks the selectivity was further reduced to 8 X lo4. Therefore, the detector selectivity for phosphorus compounds was reduced with bead age, whereas the sensitivity remained relatively constant. The average lifetime of an alkali source for this detector was 2 or 3 months; hence, the continuous introduction of an organic mobile phase did not substantially reduce the bead lifetime. The effect of solute structure and volatility on the response of the thermionic detector was investigated by using homologous model compounds. Triethyl thiophosphate, triethyl phosphate, triethyl phosphite, and triethylphosphine were employed to ascertain whether the chemical structure of the phosphorus group influenced detector response. There was no statistically significant difference in detector response for these solutes at several concentration levels. The effect of substituent chain length and volatility was studied using trimethyl, triethyl, and tributyl phosphates as model solutes. For this series of compounds, detector response was also indistinguishable at several concentration levels. Further investigation with compounds of higher molecular weight, such as phospholipids, sugar phosphates, and nucleotide phosphates, revealed that some discrimination does occur on the basis of volatility. It should be possible to improve sensitivity for compounds of low volatility by increasingthe temperature of the primary flame; however, the background noise level would be simultaneously increased. Therefore, it would be desirable to reoptimize flame conditions if solutes of low volatility were to be routinely analyzed. Effect of Mobile-Phase Composition and Flow Rate. As mentioned previously, the separation of the analytical measurement from the fundamental combustion processes in the thermionic detector significantly reduced the background ion current created by the solvent. Consequently, the dual-flame

2299

detector was compatible with a wide range of organic solvents, including methanol, ethanol, 2-propanol, acetone, ethyl acetate, diethyl ether, and hexane. These solvents were utilized in the mobile phase in concentrations ranging from 0 to 100% with little or no loss in sensitivity relative to the response obtained with pure methanol. The small variations in detector response were closely correlated to the physical properties of the individual solvents. For example, solvents with low viscosity and surface tension, such as diethyl ether and acetone, were more efficiently nebulized and caused a slight increase in detector sensitivity. Other solvents with higher viscosity or surface tension, such as 2-propanol or ethyl acetate, correspondingly reduced the detector sensitivity. However, none of the organic solvents containing only carbon, hydrogen, and oxygen exhibited chemical or ionization interferences in the dual-flame thermionic detector. In contrast, water, acetonitrile, and methylene chloride caused a substantial increase in the background ion current. Presumably, this was due to the formation of OH., CN., and C1. radical species, respectively, which could interact with the alkali bead to form the corresponding anions and, subsequently, be detected at the collector electrode. These solvents could only be tolerated in concentrations up to 1 6 2 0 % in the mobile phase. It seems intuitively evident and has been experimentally verified (11,29)that the thermionic detector is a mass-sensitive device; that is, it responds to the instantaneous mass flux of solute into the flame. However, it was also reported by Lubkowitz et al. (30)that this detector may be concentration-sensitive under certain circumstances. It was, therefore, desirable to investigate the influence of mobile-phase flow rate on the response to the thermionic detector for liquid chromatography. For a standard injection of trimethyl phosphate, a linear increase in peak height was observed with increasing mobile-phase velocity. Thus, the thermionic detector behaved in the predicted manner for a mass-sensitive device as long as the mass flux of phosphorus remained within the linear dynamic range. Mobile-phase velocities of at least 10 pL/min could be readily tolerated without extinguishing the flame or depositing soot on the alkali bead or collector electrode. However, the detector noise normally increased with mobile-phase flow rate, and the best response was obtained for flow rates less than 5 pL/min. Detector Dispersion. Extracolumn contributions to the total system variance must be very small in microcolumn HPLC in order to preserve the high column efficiency. If the microcolumn effluent is introduced directly into a flame-based detector without benefit of connecting tubing or a nebulization chamber, then the apparent dead-volume of the detection system will be quite small and primarily of temporal origin. In the dual-flame thermionic detector, there are several possible sources of temporal band dispersion: the nebulization time and the exponential contribution caused by solute adhesion to the capillary tip; the residence time of solutes in the flame; the time required for solute interaction with the bead and collector electrode; and the time constants of the amplifier (0.15 s), low-pass filter (1s), and chart recorder (0.5 s). The combined time constant of the ancillary electronic devices is expected to be the most significant factor among those cited. The apparent dead-volumeof the flame-baseddetector was estimated in the following manner. An open-tubular fusedsilica capillary (30 pm i.d., 14.3 m length) was inserted through the on-column UV-absorbance detector and then directly into the primary flame of the thermionic detector. The opposite end of this open-tubular column was connected to a lowvolume, “heart-cut” injection system (21). The pesticide Diazinon was injected as a nonretained model solute, and the resulting chromatographic peak was simultaneously acquired

2300

ANALYTICAL CHEMISTRY, VOL. 55, NO. 14, DECEMBER 1983

2

AI

1

d-

F P 0

20

40

TIME

Crn 1 n >

60

0

20

40

TIME

Crnln>

60

Figure 8. Chromatogram of organophosphorus pesticides: column, small-bore fused-silica column (0.2 mm i.d., 1 m length) packed with 5 p m Spherisorb ODS; mobile phase, 8 5 % aqueous methanol (1.6 pL/mln): solutes (1) solvent and phosphorus-containing impurities, (2) Guthion, 8 9 ng of phosphorus, (3) Zolone, 71 ng of phosphorus, (4) Ethion, 144 ng of phosphorus: detectors (a) UV-absorbance detector, 254 nm, (8) thermionic detector.

from the UV-absorbance and thermionic detectors. The variance of each peak was calculated by computer from the second statistical moment (31-33). The variance of the flame-based detector proper was then calculated by subtracting the total variance of the peak measured by the UV-absorbance detector (1.06 X lo6 nL2) from that measured by the thermionic detector (1.22 X lo3 nL2). It was also necessary to subtract the variance of the connecting tube between the two detectors, which was calculated from the Taylor equation (34)to be 23 nL2. Therefore, the variance of the flame-based detector was estimated to be 137 nL2, and the apparent dead-volume (uv)was 11.7 nL. It is evident from these approximate measurements that the variance contributed by the flame-based detector is extremely small. These detectors are, therefore, ideally suited for application to microcolumn HPLC and seem to be compatible with small-bore packed columns, packed capillaries, and open-tubular columns as well. Applications. Many synthetic industrial products, as well as their precursors and byproducts, contain phosphorus, among which pesticides, fire retardants, surfactants, and fertilizers are representative examples. The high toxicity or slow degradation of such compounds frequently necessitates their detection and quantitation in complex environmental samples. To demonstrate the potential of the thermionic detector for industrial applications, a mixture of several organophosphoruspesticides was separated on a reversed-phase, small-bore packed column using 85% aqueous methanol as the mobile phase. The response of the thermionic detector was compared with that of the UV-absorbance detector at 254 nm (75 pm path length). As illustrated in Figure 6, the UVabsorbance detector exhibited good sensitivity for Guthion

I 0

20

40

TIME

60

Cm In)

Flgure 7. Separation of Ethion and its hydrolysis products. Chromatographic conditions are described In Figure 7. Solutes are as follows: (1) solvent and inorganic phosphate; (2-6) hydrolysls products; (7) Ethion; total mass 165 ng of phosphorus.

(e = lo3L/(mol cm) at 254 nm (28))but much lower response to Zolone and Ethion. In contrast, the thermionic detector exhibited excellent sensitivity and selectivity for all pesticides, and response was nearly independent of solute structure. Thus, the thermionic detector can provide sensitivity comparable to that of UV-absorbance detectors for solutes of moderate molar absorptivity and superior response to organophosphorus compounds without chromophores. As a further demonstration of the practical utility of this detector, the decomposition products of Ethion were studied under controlled experimental conditions. The pesticide was hydrolyzed in a methanol-water solution for 1 week at room temperature, and the resulting mixture was analyzed by microcolumn HPLC with thermionic detection. The high sensitivity and selectivity characteristic of this detector are clearly evident in this chromatogram (Figure 7 ) , as the total sample mass corresponds to only 165 ng of phosphorus. In combination with other detectors, the thermionic device can potentially provide confirmatory evidence in the chromatographic analysis of organophosphoruscompounds in complex matrices. Deriuatization Methods. Chemical modification of solute structure is commonly employed in chromatographicanalyses for two distinct purposes: to increase solute volatility by derivatizing certain polar functional groups and to enhance solute detectability by incorporating specific elements or groups that are amenable to selective detection. Both of these approaches have been employed to some advantage with the flame-based detectors for microcolumn liquid chromatography. Some molecules contain phosphorus in the relatively involatile acid form, most notable are such biochemically important compounds as phospholipids, nucleotide phosphates, and sugar phosphates. Esterification of the phosphoric acid group can substantially increase the volatility of such com-

ANALYTICAL CHEMISTRY, VOL. 55, NO. 14, DECEMBER 1983

2301

phylline, quinoline, and diphenylamine. Although the magnitude of response was not identical, prominent peaks were observed for all solutes when injected at the 200-ng level. These preliminary results are especially promising because they were obtained under the optimum conditions established for phosphorus-selective detection. With specific optimization for nitrogen response, currently under way in our laboratory, the thermionic detector should provide a reasonable degree of sensitivity and selectivity for such compounds. Potential applications in the nitrogen-selective mode are numerous and include the determination of nitrogen-containing drugs and their metabolites in physiological fluids, polynuclear heteroaromatic compounds in energy-relatedsamples, and alkaloids in plant materials. The response of the thermionic detector to halogen compounds was briefly investigated by use of chloro-, bromo-, and iodobenzene as model compounds. Similarly, sulfur response was investigated with dimethyl sulfate and sodium dodecyl sulfate as test probes. Little or no selective response was observed for any of these models solutes under the optimum conditions established herein.

2

4

CONCLUSIONS 1

..

L ~

0

20

TIME

40

60

r-

80

Cmln>

Figure 8. Chromatogram of the diethyl phosphate derivatives of aliphatic alcohols: column, small-bore fused-silica column (0.2 mm I.d., 1.2 m length) packed with 5 pm Spherisorb ODs; mobile phase, methanol (0.9 pL/min); solutes, (1) solvent, (2) dodecyl alcohol derivative, 77 ng of phosphorus, (3) tetradecyl alcohol derivative, 65 ng of phosphorus, (4) hexadecyl alcohol derivative, 58 ng of phosphorus.

pounds, thereby improving their detectability. Since silylating reagents are known to degrade the alkali source through the formation of nonvolatile alkali silicates (35), methylation is the preferable alternative. We have utilized diazomethane reagent to esterify the phosphoric acid groups in diverse molecules and generally observed a response enhancement with the thermionic detector. This derivatization methodology may become important for marginally detectable compounds such as phospholipids (36). In order to expand the range of detectable substances with the thermionic detector, derivatization methods have also been employed to incorporate phosphorus into various classes of compounds. Jacob and co-workers (22-24),Deo and Howard (%),and others (37-39) have utilized the phosphorochloridates to derivatize organic molecules containing hydroxyl or amine functional groups for gas chromatographic analysis. We have further established that such derivatives are sufficiently stable in aqueous solution to permit analysis by liquid chromatography as well (7). An exemplary chromatogram is shown in Figure 8, wherein the diethyl phosphate derivatives of several aliphatic alcohols were separated on a reversed-phase microcolumn using methanol as the mobile phase. Such derivatization procedures can potentially provide a means to selectively label and detect a wide range of organic compounds. Selective Detection of Other Elements. Although the applications demonstrated herein have been limited to the detection of phosphorus-containing compounds, selective response for additional heteroatoms has also been investigated. The thermionic detector response was evaluated with a wide variety of nitrogen-containing organic compounds, including tetramethylthiourea, pyridine, pyrazine, caffeine, amino-

In this investigation, a thermionic detector has been developed and characterized for use with microcolumn liquid chromatography. The dual-flame design of this detector permits the separation of fundamental flame processes, which can be the predominant source of background noise, from the analytical measurement of ion current. After independent optimization of each flame, the thermionic detector exhibited excellent stability, sensitivity, and selectivity for both nitrogen and phosphorus compounds. As this prototype detector was constructed from an obsolete commercial instrument and was only nominally optimized,further improvements in sensitivity should be feasible. As a further consequence of its dual-flameconstruction, the thermionic detector was compatible with a wide range of organic solvents. In general, those solvents that contained only carbon, hydrogen, and oxygen could be utilized in all proportions in the mobile phase with little or no resulting loss in sensitivity. In contrast, water, acetonitrile, and methylene chloride caused a significant increase in background ion current and could be tolerated only at low concentrations. However, a sufficient variety of mobile phase is available to accomplish important separations using both normal- and reversed-phase liquid chromatography. Yet another advantage of the flame-based thermionic detector is its extremely small dead-volume. This device was found to have an apparent dead-volume of 11.7 nL, with much of this dispersion directly attributable to the time constant of the electronic instruments. Consequently, the thermionic detector was compatible with small-bore packed columns, as well as the more instrumentally demanding packed capillary and open-tubular columns. The thermionic detector described herein and the flame photometric detector described previously (7),appear to be particularly well suited for this application and are believed to be the predecessors of other useful flame- and plasma-based detectors for microcolumn liquid chromatography.

ACKNOWLEDGMENT The authors acknowledge the technical assistance of John Dorsett and Robert Ensman in the construction of the thermionic detector. Fused-silica capillary tubing was generously provided by Kenneth Mahler and Ernest Dawes of Scientific Glass Engineering,Inc. The phosphorus-containing derivatives of estra-1,3,5(lO)-triene-3,17&diol,4-pregnen-21ol-3,20-dione, cyclohexylamine, and 1,3,5-benzenetriolwere obtained through the courtesy of Karl Jacob, Ludwig-Maxi-

2302

Anal. Chem. 1983, 55, 2302-2309

milians-Universitat Munchen, G.F.R. LITERATURE CITED ( I ) Ettre, L. S. J . Chromatogr. Scl. 1978, 76, 396-417. (2) Julln, B. G.; Vandenborn, H. W.; Kirkland, J. J. J . Chromatcgr. 1975, 7 72, 443-453. (3) Locke, D. C.; Dhlngra. B. S.; Baker. A. D. Anal. Chem. 1982, 54, 447-450. (4) Willmott, F. W.; Doiphln, R. J. J . Chromatogr. Scl. 1974, 72, r- -i ~ .~- -. 7nn (5) Krejci, M.; Tesarik, K; Rusek, M.; Pajurek, J. J . Chromatogr. 1981, 278, 167-178. (6) McGuffln, V. L.; Novotny, M. unpublished research, Indlana Unlversb, Bloomington, IN, 1982.(7) McGuffln, V. L.; Novotny, M. Anal. Chem. 1981, 53, 946-951. (8) Karmen, A.; Gluffrlda, L. Nature (London) 1984, 207, 1204-1205. (9) Sevcik, J. Chromatographla 1973, 6, 139-148. (10) Kolb, 6.; Blschoff, J. J . Chromatogr. Scl. 1974, 72, 625-629. (11) Brazhnikov, V. V.; Gur'ev, M. V.; Sakodynsky, K. I.Chromatogr. Rev. 1970, 72, 1-41. (12) Slais, K.; Krejci, M. J . Chromatogr. 1974, 97, 181-186. (13) Compton, B. J.; Purdy, W. C. J . Chromatogr. 1979, 769, 39-50. (14) Lubkowitz,J. A.; Semonlan, B. P.; Galobardes, J.; Rogers, L. B. Anal. Chem. 1978, 50, 672-676. (15) Yang. F. J. J . Chromatogr. 1982, 236,265-277. (16) Gluckrnan, J. C.; Hlrose, A.; McGuffln, V. L.; Novotny, M. Chromatographla 1983, 77, 303-309. (17) Tsuda, T.; Novotny, M. Anal. Chem. 1978, 50, 271-275. (18) McGuffin, V. L.; Novotny, M. J . Chromatogr. 1983. 255, 381-393. (19) Hirata, Y.; Novotny, M.; Tsuda, T Ishii, D. Anal. Cbem. 1979, 57, 1807-1809. (20) Hirata, Y.; Novotny, M. J . Chromatogr. 1979, 786, 521-528. (21) McGuffln, V. L.; Novotny, M. Anal. Chem. 1983, 55, 560-583. (22) Jacob, K.; Vogt, W.; Knedel, M. Lleb/gs Ann. Chem. 1979, 878-865. (23) Jacob, K.; Falkner, C.; Vogt, W. J . Chromatogr. 1978, 767, 67-75. (24) Jacob, K.; Maler, E.; Schwertfeger, G.; Vogt, W.; Knedel, M. Blomed. Mass Spectrom. 1978, 5 , 302-311.

-

(25) Deo, P. G.; Howard, P. H. J . Assoc. Off. Anal. Chem. 1978, 67, 2 10-2 13. (26) Semonian, B. P.; Lubkowitz, J. A,; Rogers, L. B. J . Chromatogr. 1978, 757, 1-10, (27) Patterson, P. L. J . Chromafogr. 1978, 767, 381-397. (28) Gore, R. C.; Hannah, R. W.; Pattacini, S. C.; Porro, T. J. J . Assoc. Off. Anal. Chem. 1971, 54, 1040-1082. (29) Hartmann, C. H. Anal. Chem. 1971, 4 3 , 113A-125A. (30) Lubkowitz, J. A.; Glajch, J. L.; Semonlan, B. P.; Rogers, L. B. J . Chromatogr. 1977, 733, 37-47. (31) Sternberg, J. C. "Advances In Chromatography"; Giddings, J. C., Keiler, R. A., Eds.; Marcel Dekker: New York, 1966; Vol. 2; pp 205-270. (32) Kirkland, J. J.; Yau, W. W.; Stoklosa, H. J.; Diiks, C. H. J . Chromatogr. Sci. 1977, 15, 303-316. (33) Chesier, S. N.; Cram, S. P. Anal. Cbem. 1971, 4 3 , 1922-1933. (34) Taylor, G. R o c . R . SOC.London, Ser. A 1953, 279, 186-203. (35) Bayer, F. L. J . Chromatogr. Scl. 1977, 15, 580-581. (36) McGuffln, V. L. Ph.D. Dlssertatlon, Indiana Universlty, Bloomlngton, IN, 1983. (37) Bowman, M. C.; Beroza, M. J . Assoc. Off. Anal. Cbem. 1967, 50, 926-933. (38) McCallum, N. K. J . Chromatogr. Sci. 1973, 7 7 , 509-514. (39) Heenan, M. P.; McCallum, N. K. J . Chromatogr. Scl. 1974, 72, 89-90.

RECEIVED for review April 4,1983. Accepted August 25,1983. This research was supported by the Department of Health and Human Services, Grant No. GM 24349. V.L.M. was the recipient of a full-year graduate fellowship bestowed by the American Chemical Society, Division of Analytical Chemistry, and sponsored by the Upjohn Co. Preliminary results of this research were presented at the Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, Atlantic City, NJ, 1982.

Pressure Dependence of Diffusion Coefficient and Effect on Plate Height in Liquid Chromatography Michel Martin and Georges Guiochon* Ecole Polytechnique, Laboratoire de Chimie Analytique Physique,' 91128 Palaiseau, France

The dlffuslon coefflclent of solutes In solvents Is a function of pressure, the relative varlatlon being of the order of lo4 bar-'. Thus, the reduced linear velocity varies along the column and the average plate height Is not equal to the local plate helght at column outlet. The dtfference between these plate helgMs Increases markedly wlth Increasing velocltles. Accordingly, losses In plate number would appear at large velocities when coupling several columns, but they would be lnslgnlflcant around the velocity corresponding to the mlnlmum plate height because of a compensation effect. Because of this pressure effect on dlffuslon coefficient, the values of the coefflclents of the plate height equation derived from a least-squares flt of the data differ slgnlflcantly from their local values. Errors exceedlng 30 and 100% can be observed on the values of A and C, respectively. This effect of the longitudinal pressure gradlent Is parallel to the thermal effect. I t may be somewhat reduced by the axial temperature gradient but Interacts wlth the radlal temperature effect, 80 as to Increase their common contrlbutlon to band broadening.

Pressure can influence in many ways the peak dispersion in a chromatographiccolumn. The most obvious effect of inlet pressure comes from the fact that it determines the mobile phase velocity in the column and, hence, changes the plate

height for the solute. Such an effect has been extensively studied and is described in almost every text on chromatography (1,2). If however, the column is operated with a given mobile phase at a fixed average velocity, but with different average pressures, one will observe a change in the plate height value of the solute due mainly to the pressure dependence of its diffusion coefficient. Such an effect is accounted for in gas chromatography through pressure correction factors. Pressure effects on mobile phase parameters are quite often considered as negligible in liquid chromatography (LC), because the compressibility and the pressure dependence of viscosity of liquids are relatively weak. Typical values are 104/bar for the former and 104/bar for the latter. Thus, when operating a chromatographic column under a pressure drop of 100 bars (approximately 1500 psi), the change in the viscosity of the mobile phase reaches 10% between the inlet and the outlet of the column. Such a variation must be taken into account for accurate calculations of the chromatographic behavior, as indicated in a previous study about the effect of pressure on the retention time and the retention volume of an inert compound in LC (3). Pressure dependences of density and viscosity tend to increase, respectively, the elution volume and the elution time of an unretained solute above the values expected for a noncompressiblesolvent of constant viscosity. Beside these effects on an inert compound, pressure can thermodynamicallyaffect the distribution coefficients and the capacity ratios of retained solutes, offering the possibility of

0003-2700/83/0355-2302$01.50/00 1983 American Chemical Society