Preparation and Evaluation of Slurry-Packed Liquid Chromatography

Fused silica capillary liquid chromatography columns with inner diameters ... Analytical Chemistry 0 (proofing), .... In-Depth Characterization of Slu...
0 downloads 0 Views 393KB Size
Anal. Chem. 1996, 68, 1212-1217

Preparation and Evaluation of Slurry-Packed Liquid Chromatography Microcolumns with Inner Diameters from 12 to 33 µm Showchien Hsieh and James W. Jorgenson*

Department of Chemistry, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599-3290

Fused silica capillary liquid chromatography columns with inner diameters between 12 and 33 µm were slurry packed with 5 µm octadecylsilane-modified silica particles. Column efficiencies and van Deemter coefficients were compared. A linear decrease of the A term as column diameter was decreased was the most significant contributor to a lower overall plate height at the optimum velocity. Several groups have successfully fabricated packed microcolumns with inner diameters from 44 to 300 µm.1-6 The minimum reduced plate height (hmin) reported was between 2.2 and 3 for particles of 5 µm. Kennedy and Jorgenson7 prepared columns from 21 to 50 µm i.d. packed with 5 µm particles. They found that, for the 21 µm i.d. diameter columns, hmin decreases to 1.0 for unretained compounds and to 1.5 for retained compounds with a capacity factor of 2.7. This paper will discuss the preparation and evaluation of columns with inner diameters from 33 to12 µm packed with 5 µm silica-based reversed-phase particles. Column performance was evaluated by collecting plate height and flow velocity data. These data were then plotted as van Deemter curves. The magnitudes of the A, B, and C terms were evaluated to determine reasons for performance variations as the column inner diameter changes. The columns were tested with a mixture of norepinephrine (NE), 3,4-dihydroxybenzylamine (DHBA), and dopamine (DA) using carbon fiber electrochemical detection. EXPERIMENTAL SECTION Fabrication of Columns. Fused silica capillaries (Polymicro Technologies Inc., Phoenix, AZ) of 12, 17, 23, and 33 µm i.d. were cut to the desired length and used as columns. Capillary inner diameter measurements were carried out on the basis of the mercury resistance method developed by Guthrie et al.8 The frit for each column was formed by tapping the end of the capillary into a pile of 5 µm spherical silica particles, thus forcing some into the end of the column. To make room for carbon fiber detection electrode placement, the particles were forced up the (1) Karlsson, K. E.; Novotny, M. Anal. Chem. 1988, 60, 1662-1665. (2) Takeuchi, T.; Ishii, D. J. J. Chromatogr. 1982, 238, 409-418. (3) Yang, F. J. J. Chromatogr. 1982, 236, 265-277. (4) Gluckman, J. C.; Hirose, A.; Mcguffin, V. L.; Novotny, M. Chromatographia 1983, 17, 303-309. (5) Deweerdt, M.; Dewaele, C.; Verzele, M. J. High Resolut. Chromatogr. Chromatogr. Commun. 1987, 10, 553. (6) Hoffman, S.; Blomberg, L. Chromatographia 1987, 24, 416-420. (7) Kennedy, R. T.; Jorgenson, J. W. Anal. Chem. 1989, 61, 1128-1135. (8) Guthrie, E. J.; Jorgenson, J. W.; Knecht, L. A.; Bush, S. G. J. High Resolut. Chromatogr. Chromatogr. Commun. 1983, 6, 566-567.

1212 Analytical Chemistry, Vol. 68, No. 7, April 1, 1996

Figure 1. Size distribution of YMC ODS-AQ C-18 particles. (A) Size distribution of unfiltered particles. (B) Size distribution of particles that have been filtered.

column with a push rod. The push rod was made by heating and pulling a 360 µm o.d. fused silica capillary into a fine strand. The strand was cut to fit into the inner diameter of a particular column. For the 33 µm columns, a 25 µm tungsten wire was used as the push rod. After sufficient silica particles were pushed 0.5-1 mm into the column, the frit was sintered by passing the column through a match flame. The slurry of packing material was prepared in ratios from 0.003 to 0.017 g/mL using 5 µm YMC ODS-AQ C-18 silica particles (YMC, Wilmington, NC) and hexane as the slurry liquid. It was observed that the YMC particles had a range in distribution of sizes, as shown in Figure 1. Therefore, to reduce the chance of larger particles plugging the small columns, the slurry was filtered through an 8 µm polycarbonate filter (Costar, Cambridge, MA) before use in packing. The success rate of packing these small columns increased after filtering the slurry. The slurry was sonicated for 10 min before transfer into a high-pressure stainless steel reservoir. The open end of the fused silica capillary was placed inside a high-pressure stainless steel reservoir containing the slurry. The packing apparatus was operated inside a shielded 0003-2700/96/0368-1212$12.00/0

© 1996 American Chemical Society

Figure 2. Chromatogram from a 12 µm i.d. column of the three test analytes.

hood in case of accidental release of high pressure. An Altex Model 110A pump supplied isopropyl alcohol to the reservoir and forced the slurry into the column. An initial pressure of 70 bar (1000 psi) was used until a few centimeters of the column was packed. The pressure was then increased to 140 bar (2000 psi) until the column was fully packed. After packing, the column was rinsed with isopropyl alcohol, followed by deionized water before switching to mobile phase. Measurement of the Particle Diameter. The particle size distribution was determined so that the mean diameter of the packing material could be used in column evaluation. Particles were measured by light microscopy using JAVA image analysis software (Jandel Scientific, San Rafael, CA). Particles were measured both before and after filtration through the polycarbonate filter. Samples of 300 particles of each type were measured, and the distributions for each type are shown in Figure 1. The distribution data show that filtering eliminated particles larger than 8 µm. The unfiltered particles had a larger population of larger diameters, with an average size of 5.87 µm and a standard deviation of 1.06. The average size of the filtered particles was 5.44 µm, with a standard deviation of 1.27. The mean value of 5.44 µm was used for all calculations because only filtered particles were used to pack the microcolumns. A picture of the microcolumn was taken with an electron microscope. A small portion of a few millimeters was cut off from the end of the capillary column and mounted on a plate. This was then sputtered with a mixture of gold and palladium before being scanned by the electron microscopy. COLUMN TEST CONDITIONS The mobile phase used was 75 mM phosphate buffer at pH 3. The columns were evaluated with a mixture of NE, DHBA, and DA, each at a concentration of 1 × 10-4 or 5 × 10-4 M. These analytes were chosen because of their retention time distributions, with NE being the least retained (capacity factor, 0.4) and DA the most retained (capacity factor, 3.2). These three compounds also yield Gaussian peaks with the YMC ODS-AQ stationary phase used. The capacity factor, k′, was determined from the equation,

k′ ) (tr - t0)/t0

(1)

where tr is the retention time of the test analyte and to is the dead

time. For the YMC ODS-AQ stationary phase, electroactive anions, which are typically used for dead time markers in the reversed phase, have been found to be slightly retained. However, we found that a dip (Figure 2) occurs in every chromatogram. This dip occurs shortly before any anion peak and is present in the chromatogram when only the mobile phase was injected. Since this dip reflects the dead time better than any slightly retained anion, it was used as the dead time marker. Diffusion coefficients (Dm) for the three test compounds in the mobile phase were measured by determining their dispersion in an open tube.9 The Dm values (in cm2/s ( SD) for the three compounds are (6.23 ( 0.04) × 10-6 (n ) 17) for NE, (6.97 ( 0.11) × 10-6 (n ) 28) for DHBA, and (6.65 ( 0.05) × 10-6 (n ) 26) for DA. CHROMATOGRAPHIC SYSTEM Injections and chromatographic runs were performed on a system similar to the open tubular LC system developed in this laboratory.10 However, instead of using gas pressure, an LC pump (Model 222B, SSI, State College, PA) was used to supply the mobile phase to the packed microcolumn. DETECTION A 9 µm carbon fiber, ranging in length from 0.5 to 1 mm, was used as an electrochemical detector. The fiber was cut specifically to ensure that the majority of the carbon fiber was inside the column and very close to the frit. A voltage of +0.6 V vs Ag/ AgCl was applied to the carbon fiber working electrode so that the test compounds were completely oxidized. The signal current was amplified by a Model 427 Keithley current amplifier (Keithley Instrument, Inc., Cleveland, OH) and fed directly into a 16-bit analog-to-digital converter. Data acquisition and manipulation were performed with a Hewlett Packard Vectra 386/25 computer. The rise time of the current amplifier was set at 100 ms for the faster runs and 300 ms for the slower runs, which correspond to time constants, τ, of 45 and 140 ms, respectively. The data collection rate on the computer was varied for each run so that there were at least 25 points across 6σ of each peak. The data were not subjected to any digital filtering. (9) St. Claire, R. L., III. Ph.D. Thesis, University of North Carolina at Chapel Hill, 1986. (10) Jorgenson, J. W.; Guthrie, E. J. J. Chromatogr. 1983, 255, 335-348.

Analytical Chemistry, Vol. 68, No. 7, April 1, 1996

1213

COLUMN EVALUATION All the peaks were analyzed by a statistical moments program developed in-house. To eliminate any operator bias in arbitrarily defining the beginning and end of a peak, an iterative marking process is used to mark the limits of integration. This is done as follows. After marking the beginning and end of a peak, the program calculates the first moment (m1) and the second central moment (mc2), which is the temporal variance, σ2. The standard deviation value, σ, which is obtained by taking the square root of the temporal variance, σ2, is used in marking the peak’s limits of integration. This time, the peak is marked such that the integration limits are now (3σ from the center (m1) of the peak. The program then calculates new values of σ2 and σ. If the new σ2 value differs from the previous σ2 by more than 1%, the statistical moments calculation is done again using the new standard deviation value to mark the peak (3σ away from the peak center. When the change in variance becomes less than 1%, the process is terminated. The final values for the first moment and temporal variance, after correction for extracolumn variances, are used to calculate the number of theoretical plates, N. Before the retention time and σ values of a peak are used to calculate N, the value for σ must be revised for extracolumn contributions to measured band broadening. The variance due to injection has been evaluated11 on the basis of the theory devised by Sternberg.12 The injector contribution, σ12, can be written as

σ12 ) t2/K2

(2)

σtot2 ) σ12 + σ22 + σ32 + σ42

The total extracolumn temporal variance contribution was subtracted from the second central statistical moment, which was experimentally measured as previously described. The number of theoretical plates for a particular peak was then calculated from the following expression:

N ) tr2/σcol2

2

2

σ2 ) L /12

(3)

where L is the length of the carbon fiber. This expression can be divided by u2, where u is the mobile phase linear velocity, to convert the variance to time units. The variance due to the time constant of the current amplifier, σ32, is τ2, where τ is the time constant of the current amplifier. Since data were collected using a computer, the variance due to the data collection rate, σ42, can be expressed as

σ42 ) (1/2πf)2

(4)

where f is the frequency of data collection in units of points per second. The total extracolumn band broadening variances, σtot2, is the sum of all the variances described above, where (11) Kennedy, R. T.; Jorgenson, J. W. Anal. Chem. 1988, 60, 1521-1524. (12) Sternberg, J. C. In Advances in Chromatography, Volume 2; Giddings, J. C., Keller, R. A., Eds.; Marcel Dekker: New York, 1966; pp 205-277.

1214

Analytical Chemistry, Vol. 68, No. 7, April 1, 1996

(6)

where tr is the retention time (m1) of the peak and σcol2 is the temporal variance caused by the column after subtracting the extracolumn variance. The contribution of extracolumn variances due to injection ranges from 0.01% to 8% of the total, with a typical value of 0.5%. The contribution due to the detection, σ22, σ32, and σ42, which was typically less than 0.001%, was insignificant. The plate height, H, was then obtained from the expression

H ) L/N

(7)

where L is the length of the column and N is the number of theoretical plates obtained from the method described above. The performance of a chromatography column can be evaluated by examining the effects of mobile phase velocities on band broadening. This was achieved by fitting plate height (H) and flow velocity (u) data to the van Deemter equation, where

H ) A + B/u + Cu where t is the length of time of the injection and K is a unitless number which is characteristic of the injector’s input function or profile. K has been measured to be 4.4 in our laboratory,11 and each injection has been timed, so the variance due to injection can be determined. As described before, the detection was done using a carbon fiber microelectrode inserted into the end of the column.7 The contribution from the microelectrode to the spatial variance, σ22, can be expressed as

(5)

(8)

By using reduced parameters such as reduced plate height, h, and reduced velocity, v, where

h ) H/dp

(9)

v ) udp/Dm

(10)

and dp is the particle diameter, one can effectively normalize the measured efficiencies by accounting for the effects of particle diameter and diffusion coefficients. Thus, comparison of the quality of packing in different columns under different test conditions can be achieved. The peaks were measured to study the h vs v curves for at least eight different velocities, where at least three different velocities were less than the optimum velocity. For each column, two measurements of plate height were obtained at each linear velocity, and both were used in all subsequent calculations. The h vs v curves were fit to the van Deemter equation using nonlinear regression analysis provided by the software program Igor (Wavemetrics, Inc., Lake Oswego, OR). The linear fit for the van Deemter terms versus column inner diameter was also done using Igor. The average of the two van Deemter coefficient values for two columns with the same inner diameter was used to obtain the best linear fit versus column inner diameter. RESULTS AND DISCUSSION Figure 2 shows a chromatogram that was obtained from a 12 µm column. This particular chromatogram was obtained at a reduced velocity of about 10, well above the optimum velocity. This chromatogram is typical of the chromatograms obtained and

Figure 3. van Deemter plots of norepinephrine obtained from two 12 µm i.d. columns. Table 1. Coefficients of Reduced Parameter van Deemter Equation for Norepinephrine, 3,4-Dihydroxybenzylamine, and Dopamine column i.d. (µm)

k′

A

χ2

coeff of determ (r2)

33 33 23 23 17 17 12 12

0.43 0.43 0.41 0.41 0.40 0.40 0.33 0.33

0.63 0.67 0.40 0.58 0.31 0.40 0.22 0.19

0.051 0.068 0.073 0.047 0.446 0.269 0.031 0.019

0.985 0.964 0.952 0.971 0.887 0.924 0.989 0.985

33 33 23 23 17 17 12 12

1.56 1.56 1.54 1.54 1.54 1.54 1.13 1.13

3,4-Dihydroxybenzylamine 0.55 1.94 0.11 0.094 0.53 1.99 0.10 0.063 0.45 1.80 0.09 0.086 0.43 2.21 0.10 0.381 0.38 1.44 0.08 0.483 0.42 1.92 0.13 0.255 0.26 2.29 0.13 0.067 0.33 1.32 0.13 0.189

0.992 0.973 0.981 0.968 0.917 0.971 0.990 0.966

33 33 23 23 17 17 12 12

3.27 3.27 3.24 3.24 2.64 2.64 2.35 2.35

Dopamine 1.99 0.09 2.46 0.09 2.01 0.09 2.43 0.10 2.16 0.12 2.24 0.13 2.07 0.15 1.70 0.15

0.950 0.979 0.963 0.978 0.963 0.964 0.967 0.988

0.72 0.61 0.41 0.52 0.35 0.39 0.35 0.22

B

C

Norepinephrine 1.32 0.10 1.30 0.08 1.34 0.09 1.11 0.09 1.47 0.09 1.41 0.12 1.53 0.11 1.27 0.12

0.585 0.540 0.186 0.284 0.466 0.384 0.050 0.072

Figure 4. Graphs of A, B, and C term values versus column inner diameter for norepinephrine.

shows that a good signal-to-noise ratio was achieved. The van Deemter plots for two different 12 µm diameter columns are compared in Figure 3. The data fit to the van Deemter equation is indicated by the χ2 values or the coefficient of determination (r2) in Table 1. The smaller the χ2 value is, the better the fit.13 The data can also be correlated with the values obtained from

the van Deemter fit at each linear velocity to yield a linear leastsquares fit regression.14 A coefficient of determination of 1 indicates a perfect correlation between the experimental plate height data and the fitted data. The van Deemter plots in Figure 3 show that hmin approaches 1.0 near the optimum velocity for

(13) Box, G. E. P.; et al. Statistics for Experimenters, 1st ed.; John Wiley and Sons, Inc.: New York, 1977; pp 118, 634.

(14) Press, W. H.; et al. Numerical Recipes in C, 1st ed.; Cambridge University Press: New York, 1988; p 503.

Analytical Chemistry, Vol. 68, No. 7, April 1, 1996

1215

Figure 5. Graphs of A, B, and C term values versus column inner diameter for 3,4-dihydroxybenzylamine.

NE (k′ ) 0.33). For the more retained analytes, DHBA (k′ ) 1.13) and DA (k′ ) 2.35), the hmin values near the optimum velocity approach 1.2 and 1.3, respectively. These hmin values are lower than the hmin values obtained by Kennedy and Jorgenson7 for the retained compound, resorcinol (k′ ) 2.7), on a 21 µm i.d. column. Figures 4-6 show the variance of A, B, and C terms as column inner diameter changes for NE, DHBA, and DA, respectively. For the three test analytes, the A term shows a good linear relationship with respect to the column inner diameter, as indicated by the slopes of the lines and their coefficients of determination. This demonstrates that the eddy diffusion (A) term becomes smaller as the inner diameter decreases. From the graphs and the intercept values, one can see that eddy diffusion approaches zero as the inner diameter approaches zero. A term values that decrease with decreasing column inner diameter can be explained by a combination of effects. In large diameter columns, there are two distinct regions of packing, a loosely packed “wall” region and a more tightly packed “core” region. Differences in mobile phase velocity and chromatographic phase ratio exist between these two regions. This phenomenon was described in the past as the “wall effect” by Knox and Parcher15 and was observed in microcolumns by Kennedy and Jorgenson.7 The presence of two distinct regions of different packing density will result in band broadening for two reasons. First, the velocity of mobile phase flow will be faster in the less densely packed wall region than in the core region. Analyte bands will undergo dispersion based on the fact that individual molecules (15) Knox, J. H.; Parcher, J. F. Anal. Chem. 1969, 41, 1599-1606.

1216 Analytical Chemistry, Vol. 68, No. 7, April 1, 1996

Figure 6. Graphs of A, B, and C term values versus column inner diameter for dopamine.

will spend different amounts of time in the two regions. Analytes spending greater portions of time in the wall region will tend to elute earlier. Second, the capacity factor for analytes will be lower in the less densely packed wall region than in the core region. Again, analytes spending greater portions of time in the wall region will tend to elute earlier. Notice that these two effects do not compensate for each other but, instead, are additive in their peak broadening effects. According to Knox and Parcher,15 as the column diameter-to-particle diameter ratio decreases below 6, the core region disappears and the packing structure is dominated by the remaining loosely packed wall region. The packing structure becomes, in effect, more homogeneous. This eliminates the two previously described causes of band broadening. The relative openness of packing can be seen in Figure 7, where, since only two particles can fit across the 12 µm column, the core has the same open space as near the wall. The experimental results tend to corroborate these effects. Table 1 shows that k′ tends to decrease as column inner diameter decreases, with the biggest change occurring from 17 to 12 µm. Another contribution to the increased efficiency of small inner diameter columns is the ability for analytes to diffuse rapidly across the diameter of the column. For a 12 µm column, an analyte with a diffusion coefficient of 6 × 10-6 cm2/s (similar to the analytes in this study) would require a mean time of 0.03 s to diffuse from the center of the capillary to the wall (assuming free diffusion). For a 50 µm column, the mean time would be 0.5 s. The more rapidly an analyte is able to diffuse back and forth across the column’s cross section, the better the analyte can average out any remaining cross-column differences in flow and retention,

Figure 7. Electron microscope picture of a 12 µm i.d. column packed with 5 µm particles.

yielding sharper analyte bands. Thus, not only should smaller diameter columns have more uniform packing density, but also, the importance of regions of different packing density is diminished by the ability of analyte to rapidly diffuse across the entire column cross section. Figures 4-6 show that the longitudinal diffusion contribution to the plate height (the B term) is apparently not a significant function of column inner diameter. This is indicated by the near zero slope of the B term versus column inner diameter and the low coefficient of determination value (r2) from the linear regression analysis. Since the B term does not decrease with inner diameter, its effect is insignificant to the improvement in plate height observed in the columns studied here. For the three analytes, the B term is less than 2 for the least retained compound, NE, and closer to 2 for the more retained analytes, DHBA and DA. These results can be explained from the fact that, once a molecule is in the stationary phase, it can also undergo longitudinal diffusion. The extent of this depends on how long the molecule is retained and its diffusion coefficient in the stationary phase.

The C term (resistance to mass transfer) has an apparent weak negative dependence on column diameter. The C term depends largely on the analyte’s diffusion coefficient and the particle diameter of the packing material. There is, however, a slight decrease in the C term as the column diameter is increased. This can best be explained by the increase in packing density with increasing column diameter. As the packing density increases, the distance between particles decreases, resulting in faster mass transfer in the mobile phase, and thus a lower resistance to mass transfer in the mobile phase. The data supporting this trend are most clear in the case of DA, the most retained analyte. Evaluation of these slurry-packed microcolumns clearly shows the impact of column inner diameter on plate height values. As column inner diameter is reduced, higher separation efficiency is achieved. It is important to note that smaller diameter columns have special problems as well as advantages associated with their size. As column diameter decreases, for instance, the sensitivity of on-column UV absorbance detection is reduced due to the shorter optical path length and increasing problem with stray light. On the other hand, in the case of our electrochemical detection system, the sensitivity is actually enhanced with smaller diameter columns, since there is a shorter distance the analyte has to diffuse to the carbon fiber electrode. Smaller diameter columns also require smaller sample volumes and lower mobile phase flow rates. This can require special approaches and devices for sample injection and mobile phase delivery, but it is also a decided advantage when one is analyzing samples as small as a single cell. ACKNOWLEDGMENT Financial support for the work was provided by National Institutes of Health under Grant GM-39515. We thank Hewlett Packard for the gift of the Vectra 386-25 computer. Received for review July 10, 1995. Accepted December 4, 1995.X AC950682M X

Abstract published in Advance ACS Abstracts, February 15, 1996.

Analytical Chemistry, Vol. 68, No. 7, April 1, 1996

1217