Anal. Chem. 2002, 74, 2930-2937
Articles
Development and Testing of a Detection Method for Liquid Chromatography Based on Aerosol Charging Roy W. Dixon*
Chemistry Department, California State University, Sacramento, 6000 J Street, Sacramento, California 95819-6057 Dominic S. Peterson†
Chemistry Department, New Mexico Institute of Mining and Technology, Socorro, New Mexico 87801
Aerosol-based detection methods for HPLC in which HPLC effluent is converted to an aerosol and detected optically have been employed in the past. This paper describes a new aerosol-based detection method for HPLC, which we name aerosol charge detection. This detection method also involves generation of an aerosol but with aerosol detection by charging aerosol particles and measuring the current from the charged particle flux. A commercial electrical aerosol size analyzer was used for the aerosol detection. The constructed detector was tested using flow injection analysis with water as the mobile phase, and the signal response was found to be linear for sodium sulfate over the concentration ranges of 0.2-100 µg mL-1 using one of the nebulizers. Minimum mass and concentration detection limits using the more efficient nebulizer were estimated to be 0.2 ng and 10 ng mL-1, respectively. Behavior for most of the other compounds tested was similar with some differences in sensitivity. Testing the detector using reversed phase HPLC for glucose gave a range of linear response and detection limits that were similar to the flow injection analysis studies. Under most HPLC conditions, the noise will primarily be a function of solvent impurities; however, the electrical aerosol size analyzer allows the removal of small charged particles to improve the signal-to-noise ratio. The need for universal detection methods for liquid chromatography led to the development of evaporative light scattering detection (ELSD) in the late 1970s and 1980s.1-4 This detection method is useful for both detection of compounds without chromophores and detection of compounds separated with gradient elution, providing advantages over ultraviolet absorption * Corresponding author. Fax: 916-278-4986. E-mail:
[email protected]. † Current address: Lawrence Berkeley National Lab, and Chemistry Department, University of California, Berkeley, CA 94720. (1) Charlesworth, J. M. Anal. Chem. 1978, 50, 1414. (2) Macrae, R.; Dick, J. J. Chromatogr. 1981, 210, 138. (3) Stolyhwo, A.; Colin, H.; Guiochon, G. J. Chromatogr. 1983, 265, 1. (4) Mourey, T. H.; Oppenheimer, L. E. Anal. Chem. 1984, 56, 2427.
2930 Analytical Chemistry, Vol. 74, No. 13, July 1, 2002
detection and refractive index detection, respectively.5 With ELSD, detection involves nebulization of a liquid stream into a gas stream, evaporation of the droplets, and detection of the resultant aerosol by light scattering.1 The theory and applications of this method have been reviewed recently.6 This paper investigates using similar nebulization and evaporation steps, but with detection of the resultant aerosol through electrical charging of the aerosol. Provided that the analyte is not very volatile, ELSD is a universal detection method. As with many aerosol processes, light scattering depends on the size of the aerosol particles, which is given by the equation
()
dp ) d d
C Fp
1/3
(1)
where Fp is the density of the particle (given by the density of the analyte), C is the concentration, dp is the particle diameter, and dd is the droplet diameter.5 It has been noted that as concentration decreases, the size of particles can decrease from the size range where light scattering is efficient (diameters on the order of hundreds of nm for the Mie scattering regime) to particle sizes much smaller than light where very little incident light is scattered (in the Rayleigh scattering regime with diameters below ∼100 nm).6-8 This limitation has led to moderate sensitivity, with lower detection limits for instruments typically on the order of 0.1-1 ppm.6,9-11 To overcome the moderate sensitivity of ELSD, an additional step of condensation of vapor on the analyte aerosol particles can be added between the aerosol production and the light-scattering (5) Stolyhwo, A.; Colin, H.; Guiochon, G. Anal. Chem. 1985, 57, 1342. (6) Koropchak, J. A.; Magnusson, L.-E.; Heybroek, M.; Sadain, S.; Yang, X.; Anisimov, M. Adv. Chromatogr. 2000, 40, 275. (7) Guiochon, G.; Moysan, A.; Holley, C. J. Liq. Chromatogr. 1988, 11, 2547. (8) Allen, L. B.; Koropchak, J. A. Anal. Chem. 1993, 65, 841. (9) Righezza M.; Guiochon, G. J. Liq. Chromatogr. 1988, 11, 1967. (10) Henry, C. Anal. Chem. 1997, 18, 5561A. (11) Snyder, L. R.; Kirkland, J. J.; Glajch, J. L. Practical HPLC Method Development, 2nd ed.; John Wiley & Sons: New York, 1997; p 81. 10.1021/ac011208l CCC: $22.00
© 2002 American Chemical Society Published on Web 05/22/2002
detection to increase the particle size leading to greater light scattering.8 Detection of aerosols using condensation nucleation methods has been applied in the determination of trace residue in solutions12,13 and in the development of condensation nucleation light-scattering detection (CNLSD) for application to liquid-based separation methods.8,14,15 The huge increase in the size of particles from condensation results in outstanding inherent sensitivity for CNLSD;8 however, its application to HPLC has been limited by nonvolatile contaminants in solvents, which produce a large number of detectable particles.15 Despite this problem, lower detection limits around 10 ng mL-1 have been reported for various HPLC methods,6 with subpicogram-per-milliliter detection limits using flow injection analysis (FIA).8 CNLSD also has been applied for submicrogram-per-milliliter detection of analytes separated by capillary electrophoresis,14,16,17 size exclusion chromatography,18 and ion chromatography19,20 and has been reviewed recently.6,21 In addition to optical methods for detecting particles, numerous electrical methods of analyzing aerosols are known and have been reviewed recently by Flagan.22 Electrical methods for analyzing aerosols have been used for sizing aerosols22,23 and for measuring aerosol surface area.24 Small charged particles are detected with good sensitivity by collecting them on an air filter electrically connected to an electrometer.22 Aerosol charging with nebulization and evaporation for trace residue analysis of solvents has also been described previously.25 In this paper, we investigate using aerosol charging to develop a new detection method, which we name aerosol charge detection (ACD). The aerosol charge detection process consists of production steps similar to those in ELSD and CNLSD but with the aerosol detection step performed by aerosol charging using an electrical aerosol analyzer (EAA). The EAA is a commercial instrument that was built in the 1970s for sizing aerosols and is described in detail by Liu and Pui.23 The EAA operates by charging particles as they pass near a region of positive corona discharge, by removing particles of larger mobility (which is inversely related to particle size) through attraction to a charged rod of negative charge, and by detecting the charged particles with a filter/ electrometer. The objective of this work was to demonstrate an ACD system utilizing the EAA that would perform competitively with ELSD and CNLSD. Because aerosol charging is more sensitive, as compared to light scattering for small particles (diameters less than 100 nm), ACD offers inherent advantages over ELSD. Although the EAA (12) Kinney, P. D.; Pui, D. Y. H.; Liu, B. Y. H.; Kerrick, T. A.; Blackford, D. B. J. Inst. Environ. Sci. 1995, 38, 27. (13) Xu, M.; Chang Chien, S. Y.; Wang, H. C. J. Inst. Environ. Sci. 1996, 38, 27. (14) Lewis, K. C.; Dohmeier, D. M.; Jorgenson, J. W.; Kaufman, S. L.; Zarrin, F.; Dorman, F. D. Anal. Chem. 1994, 66, 2285. (15) Allen, L. B.; Koropchak, J. A.; Szostek, B. Anal. Chem. 1995, 67, 659. (16) Szostek, B.; Koropchak, J. A. Anal. Chem. 1996, 68, 2744. (17) Szostek, B.; Zajac, J.; Koropchak, J. A. Anal. Chem. 1997, 69, 2955. (18) Koropchak, J. A.; Heenan, C. L.; Allen, L. B. J. Chromatogr. 1996, 736, 297. (19) Sadain, S. K.; Koropchak, J. A. J. Liq. Chromatogr. 1999, 22, 799. (20) Sadain, S. K.; Koropchak, J. A. J. Chromatogr. 1999, 844, 111. (21) Koropchak, J. A.; Sadain, S.; Yang, X.; Magnusson, L.-E.; Heybroek, M.; Anisimov, M.; Kaufman, S. L. Anal. Chem. 1999, 71, 386A. (22) Flagan, R. C. Aerosol Sci. Technol. 1998, 28, 301. (23) Liu, B. Y. H.; Pui, D. Y. H. J. Aerosol Sci. 1975, 6, 249. (24) Matter, U.; Siegmann, H. C.; Burtscher, H. Environ. Sci. Technol. 1999, 33, 1946. (25) Blackford, D. B.; Quant, F. R.; Kerrick, T. A.; Sem, G. J.; Havir, D. D. U.S. Patent 5,098,657, 1992.
is less sensitive than condensation nuclei counters (CNCs) used in CNLSD,22 ACD has some potential advantages over CNLSD. With aerosol charging, no optical components are needed. Chromatographic detectors based on electrical, as opposed to optical, detection methods are often more economical. In addition, CNLSD has used additional components, as compared to ELSD, thus adding to instrument complexity. This has included condensers for removal of solvent vapor following the evaporation stage and diffusion screens for removal of small particles.13,14 In aerosol charge detection, removal of solvent vapor is not needed, provided that the aerosol detection component is heated so that condensation does not occur. Although the EAA used in these experiments was not heated, a related commercial aerosol instrument, the diffusion charger, is heated.26 The diffusion screens used in CNLSD have been shown to improve signal linearity and sensitivity when solvent impurities are higher.14,15 For ACD, there may be less need for removal of small particles for improving linearity because, because as will be discussed in the next paragraph, the signal depends on the particle size over a wide range of particle sizes. The potential on the charged rod in the EAA can be easily changed to select a size for removal of small particles if needed, whereas diffusion screens must be changed manually. When the EAA is operated under normal conditions, the charge imparted to particles in the corona discharge region depends primarily on the particle size. The sensitivity of the instrument as a function of particle size is given by eq 2, which is derived from data in Liu and Pui23
for dp < 10 nm, S ) 2.3 × 10-16dp6.6 for dp > 10 nm, S ) 1.61 × 10-10dp1.11
(2)
where S is the sensitivity in units of fA m3 particles-1, and dp is in nm. The sensitivity given for dp 10 nm, Sm )
4.4 × 105 3.6 dp Fp
3.01 × 1011 -1.89 dp Fp
(3)
The maximum sensitivity per particle mass occurs at particle diameters of ∼10 nm. Although it is possible to use equations for instrument sensitivity along with equations for the size distribution of droplets generated by a pneumatic nebulizer to predict the response, this is complicated by a number of factors, such as the loss of droplets to walls in the spray chamber and the possible (26) Matter Engineering, AG. Document, Operating Instructions for Diffusion Charging Particle Sensor Type LQ1-DC, 2001, p 2. (27) Adachi, M.; Kousaka, Y.; Okuyama, K. J. Aerosol Sci. 1985, 16, 109.
Analytical Chemistry, Vol. 74, No. 13, July 1, 2002
2931
Figure 2. Detailed diagram of the capillary nebulizer and spray chamber.
Figure 1. Flow diagram of the aerosol charge detection system.
charging of droplets from spray electrification.28 Thus, prediction of the response will not be discussed here. For this study, the EAA was used under normal operating conditions with some optimization of the variables associated with nebulization and evaporation of droplets. Future work aimed at improving the aerosol detector in ACD should investigate either modification of the EAA, construction of an aerosol detector specific for ACD, or use with an aerosol instrument currently being produced commercially (e.g., the diffusion charger instrument for measurement of aerosol surface area24). EXPERIMENTAL SECTION The EAA used for aerosol detection was a TSI (St. Paul, MN) model 3030. It was operated using the specified flow rates (4.0 L min-1 sample flow, 1 L min-1 charger sheath flow, and 50 L min-1 total flow). The analogue signal from the EAA was connected to integrators (Hewlett-Packard, Avondale, PA, models 3390A and 3396A) or a personal computer with an analogue-to-digital board and software (Justice Laboratory Software, Palo Alto, CA). A Sage Instruments (Cambridge, MA) model 355 syringe pump was used in initial continuous flow and FIA studies. Rheodyne (Cotati, CA) model 7125 or model 7010 injection valves with 20-µL injection loops were used for all FIA and HPLC injections. For application of this detector with later FIA studies and HPLC, a Waters (Milford, MA) pump (M6000A) was used. An Applied Biosystems (Foster City, CA) model 785A absorbance detector was used for dual detector studies. Separations were performed on a 4.6 × 150 mm Alltima C18 column (Alltech Associates, Deerfield, IL). A flow diagram for the ACD is shown in Figure 1. Nitrogen was used as the nebulizer gas and was heated both prior to and following nebulization in an oven, which was a temperaturecontrolled HPLC column heater. Copper tubing (6.35-mm o.d.) was used for preheating and for the drift tube in the oven, and the oven was maintained at 50 °C for the studies discussed. The nebulizers and spray chamber used are described in greater detail in the following paragraph and are shown in Figure 2. Following the spray chamber, bends in the tubing were gradual to minimize (28) Reist, R. C. Introduction to Aerosol Science; Macmillan Publishing Co.: New York, 1980; p 140.
2932 Analytical Chemistry, Vol. 74, No. 13, July 1, 2002
aerosol loss between the spray chamber and the EAA. Initial FIA studies indicated some instability in the response due to droplet formation in the tube between the spray chamber and the oven. To reduce these instabilities, this line was heated using a resistive wire and a temperature controller (model 16010, Love Controls, Michigan City, IN). Just before the EAA, a “tee” connector was placed in the line with an absolute filter in the branch line to separate the analytical aerosol from the ambient aerosol while maintaining the inlet pressure of the EAA near ambient pressure. The exhaust air leaving the EAA was pulled through a pump and routed to a hood. A home-built concentric capillary nebulizer (referred to as capillary nebulizer), shown in Figure 2, and a Meinhard glass nebulizer (Santa Ana, CA) were used. The capillary nebulizer was constructed from a stainless steel 1.59 mm o.d. × 0.76 mm i.d. outer tube and a fused-silica capillary (0.43 mm o.d. × 0.32 mm i.d.) inner tube. A 1.59-mm stainless steel tee connector was used with a fused-silica adapter for allowing the fused-silica capillary to pass through the tee. The fused-silica capillary was connected to 1.59-mm HPLC tubing using a connector with another fusedsilica adapter. The spray chamber was made from a 25 mm o.d. × 162 mm length piece of glass tubing. The capillary nebulizer was attached to the spray chamber using a 25-mm compression fitting with the outer tube to the nebulizer and a drain tube soldered in place. The Meinhard nebulizer was attached to the spray chamber using a 6.35-mm compression fitting that was soldered to a 25-mm compression fitting. Tygon tubing, with a loop to restrict air flow, was used to connect the drain tube to a waste bottle. The spray chamber was tilted upward at an angle from inlet to outlet, because this resulted in the most uniform nebulization. The outlet side of the spray chamber was connected to 6.35-mm o.d. Teflon tubing using another modified 25-mm compression fitting. A short 6.35mm o.d. tube was soldered to the compression fitting for connecting with the Teflon tubing. Also soldered to the center of the 25-mm compression fitting was a 6.35-mm compression fitting to allow movement of the impacting surface (impactor in Figure 2) in the spray chamber. The impactor was positioned roughly 80 mm away from the nebulizer to reduce some of the larger droplets during some of the experiments and was used for the experiments described. Deionized water purified with Barnstead (Dubuque, IA) cartridges was used. Ammonium sulfate, sodium sulfate, and adipic acid were purchased from Aldrich (Milwaukee, WI).
Adenosine, nicotinamide, and uracil were from Sigma (St. Louis, MO); glucose (dextrose) was from Spectrum (Redondo Beach, CA); and sulfanilamide was from Baker (Phillipsburg, NJ). The reagents used for preparing calibration curve standards had a minimum purity of at least 99%. HPLC grade acetonitrile and methanol (Fisher Scientific, Pittsburgh, PA) had listed residues of 1.0 mL min-1; however, response (peak height) under FIA for low concentration standards was found to be close to a maximum at a flow rate of 1.0 mL min-1 for the Meinhard nebulizer, with the capillary nebulizer giving maximum response at a flow rate of ∼1.5 mL min-1. Most of the development work was performed with the capillary nebuilzer. Ideally, a nebulizer will give a maximum response with gas flow rates near the EAA flow rate (4.0 L min-1) to avoid loss of sensitivity due to venting of excess aerosol (at higher nebulizer flow rates) or dilution of the aerosol (at lower nebulizer flow rates). The capillary nebulizer was operated at flow rates of 4.8 L min-1 for the HPLC experiments and at 5.4 L min-1 for the continuous flow and FIA experiments. These flow rates were close to those of maximum response for low concentrations and result in predicted velocities at the nebulizer nozzles of 260 and 290 m s-1, respectively. Past research on optimization of
Figure 3. Aerosol signal (a) and number (b) size distributions for nebulization of sodium sulfate solutions of concentration 0.2 (triangles), 2 (diamonds), and 20 µg mL-1 (squares) obtained with capillary (open symbol) and Meinhard nebulizers (0.2 and 2 µg mL-1 concentrations only, filled symbol). The unit for the signal is fA or mV, and the units for the number concentration are particles m-1. The 3σ uncertainty on the signal is shown by the bars.
nebulizers for ELSD has shown that subsonic velocities tend to produce large droplets and lead to poorer signal-to-noise ratios.5 Because a spray chamber was used to remove large droplets in this study, the main effect of using subsonic velocities appeared to be a reduced response without affecting the noise greatly. Switching the nebulizer to the Meinhard nebulizer resulted in a greater response. The greater response is probably due to a smaller area between the two capillaries at the nebulizer tip, resulting in higher velocities when run under similar flow rates. The Meinhard nebulizer was run at a gas flow rate of 4.0 L min-1, which was the approximate flow rate of maximum response, and this corresponded to a velocity of ∼1040 m s-1. In aerosol-based detectors, the particle size distribution and analyte transport efficiencies are two important quantities that affect detector response. In a few experiments, the EAA was used to acquire size distributions by changing the “cut-size” or minimum size for passing charged aerosol particles to the detector.23 Figure 3 shows the signal and number size distributions for 0.2 and 2 µg mL-1 sodium sulfate solutions nebulized with the capillary and Meinhard nebulizers as well as a 20 µg mL-1 Analytical Chemistry, Vol. 74, No. 13, July 1, 2002
2933
Figure 4. Flow injection analysis response traces for 0.1, 0.2, and 0.5 µg mL-1 sodium sulfate using water as a mobile phase with a liquid flow rate of 1.0 mL min-1 and a gas flow rate of 5.4 L min-1. Data were smoothed using a Hamming filter with time constant of 2 s.
sodium sulfate solution nebulized with the capillary nebulizer. The number size distribution was calculated using the specified conversion parameters.23 The maximum signal in all cases occurred in the 10-32-nm size range, which is the region of maximum sensitivity on a particle mass basis. With both nebulizers, the fraction of the signal coming from larger particles increased with increasing concentration, as expected. A greater fraction of both the signal and the number of particles also was observed to come from larger particles for the capillary nebulizer relative to the Meinhard nebulizer at the same concentrations. This is expected, because the lower gas velocity at the nebulizer tip should produce larger droplets.7 However, the Meinhard nebulizer was observed to produce more large particles than the capillary nebulizer. This may be due to a greater loss of large droplets from the capillary nebulizer in the spray chamber. By using the number distribution and assuming spherical particles with a density of 2.68 g cm-3 (the density of sodium sulfate29), a mass distribution was calculated. The total mass concentration reaching the EAA was then calculated summing the aerosol mass over all sizes, and this mass concentration was used with the aqueous concentrations and the gas and liquid flow rates to estimate the analyte transportation efficiency. The analyte transport efficiency was estimated to be ∼0.7% for the capillary nebulizer and ∼4% for the Meinhard nebulizer. These values may be higher than the true values, because water associated with hygroscopic growth probably contributes to the measured aerosol size and the calculated aerosol mass. Flow Injection Analysis Tests. Tests were conducted in the FIA mode with sodium sulfate and with other compounds for determination of calibration curves and variability of response with the compound type. In the calibration study, sodium sulfate standards with concentrations ranging from 0.05 to 100 µg mL-1 were injected multiple times over several hours using the capillary nebulizer. Figure 4 shows typical FIA response traces for 0.1, 0.2, and 0.5 µg mL-1 sodium sulfate standards injected in triplicate. The background signal for water ranged from 12 to 20 mV, with 2934
Analytical Chemistry, Vol. 74, No. 13, July 1, 2002
Figure 5. Calibration curve showing peak height (mean value and 3σ uncertainty) and versus concentration for sodium sulfate standards using FIA using the capillary nebulizer (lower line) and using the Meinhard nebulizer (upper line). The equations for the least-squares fit lines are y ) 48.1x1.03 (r2 ) 0.9968) and y ) 997x0.94 (r2 ) 0.9938) for the capillary and Meinhard nebulizers, respectively, using the individual data values (not the mean values plotted).
an unfiltered 3σ noise level of ∼1.7 mV. Reproducibility based on the variability of peak height of the standards was decent, with an average relative standard deviation of 7% for standards g1 µg mL-1. Better reproducibility was obtained over shorter time periods. A 2-s Hamming filter was found to reduce the noise to ∼1.4 mV while decreasing peak heights and increasing peak widths by about 5%. A detection limit of 0.03 µg mL-1 was estimated by using a least-squares linear regression fit on the peak heights observed for the 0.05-1.0 µg mL-1 standards and the filtered noise. Using unfiltered data, peak widths at half-height of 4.0 s were observed, indicating dispersion from the 1.2-s injection plug. Tailing was observed in the peaks, probably as a result of convective dispersion of the analyte aerosol following nebulization. Effects of this dispersion on HPLC use will be discussed in the HPLC section. The calibration curve obtained from sodium sulfate standards ranging from 0.2 to 100 µg mL-1 using the capillary nebulizer is shown as the lower line in Figure 5. The response was fit to the equation
y ) Axb
(4)
using least-squares regression where y is the peak height and x is the concentration. On the basis of the b coefficient and r2 values being close to one, the response was found to be linear. A linear response was not expected, since aerosol charging does not depend directly on the aerosol mass (or diameter to the third power). ELS detection is nonlinear at parts-per-milion level concentrations with fit exponents being >1 for calibrations.6,7,15 On the other hand, CNLSD has been found to give linear calibration for concentration below 10 ppm and over a concentra(29) Weast, R. C., Astle, M. J., Eds.; CRC Handbook of Chemistry and Physics, 61st ed.; CRC Press: Boca Raton, FL, 1980.
Table 1. Exponential Fit Equation Terms for Calibration Curves for Five Compounds compd
A terma
b terma
r2
ammonium sulfate adipic acidc glucose uracil sodium sulfate
85 64 55.1 43 48.1
1.09 1.08 1.00 1.04 1.03
0.994 0.994 0.997 0.998 0.997
density of solid29,b 1.77 1.36 1.56 2.68
a A and b term from eq 4. b Given for anhydrous solid with units of g cm-3. c The 0.5 and 1 µg mL-1 standards were excluded because peak heights were much lower than given by the line.
tion range similar to that shown here when used with diffusion screens.8,15 Because the response per particle mass for aerosol charging decreases for particles >10 nm, it is expected that as concentrations are increased beyond some point, exponents
