Anal. Chem. 2008, 80, 6358–6364
Surface Area of Reversed-Phase HPLC Columns A. Giaquinto,† Zhaoxia Liu,‡ Andrew Bach,‡ and Yuri Kazakevich*,† Department of Chemistry and Biochemistry, Seton Hall University, 400 South Orange Avenue, South Orange, New Jersey 07079, and Novartis Pharmaceuticals, One Health Plaza, East Hanover, New Jersey 07936 Six reversed-phase columns from different manufacturers were characterized in terms of adsorbent geometry (e.g., pore volume, surface area, column void volume, and interparticle volume). Measurement of the surface area of chemically modified silica-based adsorbents is discussed together with the methods for the determination of the amount of adsorbent in the column. The behavior of nearly ideal chromatographic systems was studied. Retention factors of alkylbenzenes in acetonitrile/water and methanol/water systems were compared with surfacespecific retention factors. The distribution of conventional retention factor values for the same analyte among the six columns using identical chromatographic conditions exceeded 35%, while the relative standard deviation of surface-specific retention factors was on the level of 3%. The most studied aspect of HPLC research has concerned the impact of variations in key system parameters (e.g., mobile-phase composition, temperature, and pH) on the chromatographic performance.1–3 Adsorbent surface chemistry or the variation of surface interactions by chemical modification of the surface of the solid porous material is another intensively developed aspect of HPLC.4,5 It is interesting to note that, after more than 30 years of intensive research, we can describe in functional form how an analyte’s retention can vary with changes in mobile-phase composition, pH, temperature, and buffer concentration.6 However, the use of these functions generally does not work for predicting the retention of a different analyte. While a functional description is possible for the mobile-phase-related parameters, the explanation of the influence of adsorbent parameters on analyte’s retention is still only phenomenological. For example, a comparison of an analyte’s retention on different reversed-phase columns with the same chromatographic conditions (mobile phase, pH, temperature, etc.) usually shows significantly different results, even for adsorbents with identical surface chemistry, such as octadecyl-modified HPLC adsorbents.7 * To whom correspondence should be addressed. E-mail:
[email protected]. † Seton Hall University. ‡ Novartis Pharmaceuticals. (1) Schoenmakers, P. J.; Billet, H. A. H.; de Galan, L. J. Chromatogr. 1979, 185, 179–195. (2) Neue, U. D.; Phoebe, C. H. J.; Tran, K.; Cheng, Y.; Lu, Z. J. Chromatogr., A 2001, 925, 49–67. (3) Nikitas, P.; Pappa-Louisi, A. J. Chromatogr., A 2001, 971, 47–60. (4) Kirkland, J. J. J. Chromatogr., A 2004, 1060, 9–21. (5) Neue, Uwe D. HPLC Columns, Wiley-VCH, New York, 1997. (6) Kazakevich, Y. J. Chromatogr., A 2006, 1126, 232–243.
6358
Analytical Chemistry, Vol. 80, No. 16, August 15, 2008
Traditional representation of HPLC retention is based on the retention factor,8 which in turn is associated with the thermodynamic equilibrium constant.9 Retention factor is a convenient dimensionless parameter that is routinely used for comparing the retention properties of different columns.10 The expression that is used to associate the retention factor with the thermodynamic equilibrium constant is dependent on the model that is selected for the description of the retention process. The two most commonly used models, adsorption and partitioning, are different in principle. Both models have been compared and contrasted in numerous publications,11 and both describe analyte retention in a very similar mathematical form (simplified),
VR ) Vm + Vs
dcs(c) (partitioning) dc
(1)
VR ) V0 + S
dΓ(c) (adsorption) dc
(2)
where VR is the retention volume, Vm is the volume of the mobile phase, Vs is the volume of the stationary phase, and dcs/dc is the derivative of the partitioning isotherm, usually assumed to be a constant at low analyte concentrations; Vo is the volume of the liquid phase in the column, S is the total surface area of the adsorbent in the column, and dΓ/dc is the derivative of the analyte excess adsorption isotherm. The first expression states that reduced retention is proportional to the volume of the stationary phase, while the second expression indicates its proportionality to the adsorbent surface. Let us consider two exaggerated hypothetical cases: (a) 1 mL of stationary-phase volume over a 1-m2 adsorbent surface and (b) 1 mL of stationary-phase volume over a 100-m2 adsorbent surface. The analyte retention profiles for each of these situations will be significantly different. In case a, the retention will be probably very low with significant tailing as a result of slow analyte mass transfer in the bulky (1 µm thick) stationary phase, while in case b we expect significantly better retention with an acceptable peak shape, since the thickness of the adsorbed layer is only 10 nm. In partitioning-based descriptions of chromatographic retention, the thickness of the stationary-phase parameter is essentially (7) Snyder, L. R.; Kirkland, J. J.; Glajch, J. L. Practical HPLC Method Development; Wiley: New York, 1997; p 194. (8) Horvath, Cs.; Melander, W. Reversed-Phase Chromatography. In High Performance Liquid Chromatography Advances and Perspectives; Horvath, Cs., Ed.; Academic Press: New York, 1980; Vol. 2, pp 113-303. (9) Kazakevich, Y.; LoBrutto, R. HPLC for Pharmaceutical Scientists; Wiley: New York, 2007. (10) Colin, H.; Guiochon, G. J. Chromatogr. 1977, 141, 289–312. (11) Vailaya, A.; Horvath, C. J. Chromatogr., A 1998, 829, 1–27. 10.1021/ac800772s CCC: $40.75 2008 American Chemical Society Published on Web 07/23/2008
disregarded, which leads to significant discrepancies in the correlation of retention factors between different columns packed with adsorbents of similar chemistry but different geometric characteristics. Correlation of the retention factors for different columns may be also misleading since the capacity factor represents the adjusted retention volume related to the column void volume, while the retention volume is actually proportional to the adsorbent surface area.6 For example, different columns may have similar void volumes but significantly different surface areas, and the retention of the same analyte on these column could be significantly different while surface-specific retention may actually be the same. Kiselev12 was the first to introduce the idea of using surfacespecific retention, and subsequently, Kovats13 also argued for the use of surface-specific retention parameters. For example, most of the C18-type stationary phases for HPLC, although slightly different from each other, show very similar methylene selectivity, which indicates that their hydrophobicity is essentially the same.14 In order to use surface-specific retention parameters, one must determine how much adsorbent is packed into the column and how much of the actual surface is available. In 2001, Rustamov15 introduced a nondestructive method for determining the amount of the packing material in a column. In that study, it was shown that the specific pore volume of a modified adsorbent that is measured using low-temperature nitrogen adsorption (LTNA) is equivalent to the specific pore volume value that is calculated from the difference of the column void volume and interparticle volume divided by the adsorbent mass in the column. Thus,
mads )
Vo - Vip Vp
(3)
where mads is the mass of adsorbent in the column, V0 is the column void volume, Vip is the column interparticle volume, and Vp is the specific pore volume of adsorbent measured by LTNA. Methodologies for the measurement of the true column void volume and interparticle volume are also described in the same paper.15 The easiest method for measuring the column void volume involves the determination of the retention of deuterated methanol (CD3OD) eluted with the flow of pure HPLC grade methanol (CH3OH). The determination of the interparticle column volume is based on the measurement of the retention of high molecular weight polymers that are completely excluded from the adsorbent porous space as they elute through the column. The retention of these polymers, which are eluted with a mobile phase that excludes their interaction with the adsorbent surface (polystyrenes in tetrahydrofuran is a good choice), essentially represents the column interparticle volume. A correction should be made for the volume of the polymer molecules themselves, which is done by extrapolating the retention dependence of each polymer on the cubic root of the polymer mass to the value of zero mass.15 The determination of the adsorbent specific surface value, which is exposed to the analyte as it migrates through the column, (12) Kiselev, A. V. J. Chromatogr. 1970, 49, 84–129. (13) Foti, G.; Belvito, M. L.; Alvarez-Zepeda, A.; Kovats, E. sz. J. Chromatogr., A 1993, 630, 1–20. (14) Jaroniec, M. J. Chromatogr., A 1993, 656, 37–50. (15) Rustamov, I.; Farkas, T.; Ahmed, A.; Chan, F.; LoBrutto, R.; McNair, H. M.; Kazakevich, Y. V. J. Chromatogr., A 2001, 913, 49–63.
Figure 1. Correction of the nitrogen molecular area for adsorbents with alkyl-modified surfaces.
presents another complex problem. The standard method for the measurement of the surface area of porous materials is the application of the Brunauer, Emmet, and Teller (BET) theory to the nitrogen adsorption isotherm.16 BET theory allows for the calculation of the amount of nitrogen molecules in the dense monolayer (nm, monolayer capacity) that are adsorbed on the measured surface from the experimental adsorption isotherm. This value of monolayer capacity is then multiplied by the area that a nitrogen molecule occupies on the adsorbent surface and divided by the mass of the adsorbent sample used in actual adsorption measurements. Thus,
S)
nmω ma
(4)
where S is the adsorbent specific surface, ω is the nitrogen molecular area, and ma is the mass of adsorbent sample. The nitrogen molecular area is usually assumed to be 16.2 Å2.17 However, this value has been a subject of significant skepticism over the past 40 years.18 Buyanowa and Karnaukhov argued that nitrogen adsorption is less localized on hydrophobic surfaces due to weaker adsorbate-surface interactions and has more freedom for lateral movement, thus effectively occupying a larger area of the surface.18 Therefore the use of the 16.2 Å2 value would lead to an underestimation of the actual surface area of hydrophobic surfaces, since there would be significantly fewer nitrogen molecules adsorbed in the monolayer. Bass and Bratt19 provided the surface area values for both unmodified silica and for the same silica modified with different alkylsilanes, which are plotted in Figure 1. As one can see, the unmodified silica has a surface area of 420 m2/g, and the surface area of the same silica modified with trimethylchlorosilane is at least 100 m2/g lower. Other adsorbents modified with longer alkyl chains show a gradual decrease in surface area, in an almost linear manner, as a function of the number of carbon atoms in the bonded chain. Extrapolation of this line to the y-axis results in an approximate value of 330 m2/g, which is 90 m2 lower than the (16) Brunauer, S.; Emmet, P. H.; Teller, E. J. Am. Chem. Soc. 1938, 60, 309– 319. (17) Gregg, S. J.; Sing, K. S. W. Adsorption, Surface Area and Porosity; Academic Press: London, 1976. (18) Buyanova, N. E.; Zagrafskaya, R. V.; Karnaukhov, A. P.; Shepelina, A. S. Kinetica i Cataliz 1983, 24, 1011–1023. (19) Bass, J. L.; Sands, B. W.; Bratt, P. W. Silanes, Surf. Interfaces 1986, 1, 267-282.
Analytical Chemistry, Vol. 80, No. 16, August 15, 2008
6359
value that was calculated using the standard BET method with a value of 16.2 Å2 for the nitrogen molecular area. As mentioned above, nitrogen molecules are expected to occupy larger areas on a hydrophobic surface. Thus, it is logical to assume that the ratio of the measured BET area of a silica sample (420 m2) to the extrapolated area (330 m2) will be inversely proportional to the ratio of nitrogen’s molecular area on a polar silica surface to its area on a hydrophobic surface. SN2(16.2) S(extrap)
)
ω(hydroph) 16.2
(5)
The value of the area occupied by nitrogen molecule on a hydrophobic surface estimated from expression 5 is equal to 20.6 Å2, which is essentially equivalent to the area (20.5 Å2, on average) estimated by Buyanova.18 If we multiply the surface area values given by Bass et al.19 for his modified adsorbents by the ratio of 20.5/16.2, which is essentially the correction for the higher nitrogen molecular area, we obtain a nice correlation between the original silica and corresponding modified silicas as shown in Figure 1 (corrected data). There is no longer a significant drop in the surface area between the silica and the silica modified with the smallest possible ligand. In addition, there is a gradual linear decrease in the surface area with an increase in the ligand length, which is logical due to the filling of the inner pore space with organic moieties. The problems associated with the evaluation of the actual geometric parameters of different commercial columns and a comparison of the surface corrected retention data for these columns are discussed in this paper. EXPERIMENTAL SECTION Materials. All solvents used in this study were HPLC grade from Pharmco (Phillipsburg, PA). Water was purified using a Milli-Q system (Millipore, El Paso, TX). Equipment. Nitrogen adsorption measurements were performed on an Omnisorb model 100CX system (Omnisorb) using the static adsorption mode with full equilibration of each adsorption point. Adsorption and desorption isotherms were measured for each adsorbent. Raw data from the adsorption system were transferred into MathCad 12.0 software and a homemade software template was used for the calculation of the adsorbent pore volume, surface area, BET C-constant, and pore size distribution. Determinations of the column interparticle volume were made on an HP model 1050 (Agilent, New Castle, DE) system with high molecular weight polystyrenes (194K, 382K, 410K 860K, and 994K), all of which had a narrow molecular weight distribution (r less than 1.05) from Polymer Laboratories using tetrahydrofuran (HPLC grade) as the mobile phase. Column void volumes were measured on HP model 1050 (Agilent) system equipped with a 3900E analog interface, a PE LC-30 refractive index detector (Perkin-Elmer Wellesley, MA), and Chemstation software. Chromatographic retention of model compounds was measured on an Agilent model 1100 system with diode array detector (Agilent). Columns. Six commercial columns with different geometric and surface chemistry characteristics were used in this study. 6360
Analytical Chemistry, Vol. 80, No. 16, August 15, 2008
Table 1. Column Parameters name Halo-C18 Allure-C18 Xterra-C18 YMC Pack Pro C18 Gemini C18 XBridge C18
manufacturer Advanced Materials Technology Restek Waters YMC Phenomenex Waters
dimensions (mm)
particle size (µm)
50 × 4.6
2.7
150 × 4.6 150 × 3 50 × 4.6 100 × 2 150 × 3
5.0 3.5 3.0 3.0 3.5
Names, dimensions and other parameters provided by manufacturers are shown in Table 1. In this set, we include columns of a principally different nature. Allure and YMC columns are conventional, fully porous, silicabased porous materials modified with octadecylsilanes. Halo is also a silica-based material, but it is not fully porous. Instead, its ideally spherical and very uniform particles consist of a solid (nonporous) silica core (1.7 µm in diameter) that is surrounded by a 0.5-µm thick porous silica layer. This particle design presumably increases surface accessibility by decreasing the molecular traveling distance inside porous particles, while it also decreases the material surface area. Xterra is a composite material made by the cosynthesis of a silica matrix using tetraethoxysilane and methyltriethoxysilane, which allows for the inclusion of methyl groups in the body of silica-based material. The presence of methyl groups on the silica surface results in the decrease of the total number of silanols. Thus, after chemical modification of the surface, fewer silanols remain. XBridge is a further development of the Xterra concept, where methyl groups are substituted with ethyl bridges, which presumably should increase the strength of the base material and result in even fewer unreacted silanols on the surface.20 Gemini is another example of a composite material, where regular porous silica is used as a base core and a layer of organicembedded silica is synthesized on the top of the silica core. This particle design should allow for the silica-like mechanical properties of base material while similarly decreasing the number of residual silanols.22 One column of each type was washed with methanol and unpacked into tared vials (one vial per unpacked adsorbent). Each vial with adsorbent was then dried in a vacuum oven at 60 °C for 24 h and then at 80 °C for 1 h to allow for complete evaporation of the solvent. Each vial was then cooled, weighed, and placed back into the oven for 1 h, then cooled, and weighed again. This procedure was repeated until a constant weight was obtained. These unpacked materials were then used for nitrogen adsorption and TGA measurements. RESULTS AND DISCUSSION Nitrogen Adsorption. Each of the unpacked adsorbents was carefully collected and dried to constant weight to determine the actual amount of adsorbent (Table 3) that was packed into each column by the manufacturer. The surface area, pore volume, and (20) Waters Inc. http://www.waters.com. (21) Phenomenex Inc. http://www.phenomenex.com. (22) Kazakevich, Y. V.; LoBrutto, R.; Chan, F.; Patel, T. J. Chromatogr. 2001, 913, 75–87. (23) Bocian, S.; Felinger, A.; Buszewski, B. Chromatographia online, Jan 2008.
Table 2. Adsorbent Geometric Parameters from Full (Adsorption and Desorption) Nitrogen Adsorption Isotherms adsorbent
S (at 16.2 Å2), m2/g
S corr (20.5/16.2), m2/g
Vpore, mL/g
Rpore (max), Å
Rpore (median), Å
R from 2V/S
distrib assymmetry
C BET const
Halo-C18 Allure-C18 Xterra-C18 YMC-C18 Gemini-C18 XBridge-C18
72 230 115 201 170 120
91 291 146 254 215 152
0.134 0.335 0.501 0.586 0.531 0.471
27.0 26.7 44.9 52.7 42.4 74.7
27.0 23.8 58.6 46.5 46.5 69.7
29.4 23.0 68.9 46.1 49.4 62.0
1 0.89 1.3 0.88 1.1 0.93
20.3 21.3 20.6 20.6 22.1 21.7
Table 3. Column Geometric Parameters adsorbent
Vo, mL
Halo-C18 Allure-C18 Xterra-C18 YMC-C18 Gemini-C18 XBridge-C18
0.442 1.381 0.740 0.595 0.232 0.71
Vip, mads mads RSD, S (total) mL (measured), g (calc), g % (calc), m2 0.34 0.88 0.40 0.32 0.13 0.41
0.771 1.53 0.681 0.461 0.189 0.652
0.761 1.493 0.677 0.464 0.183 0.645
1.3 2.4 0.5 0.7 2.4 0.7
70 445 99 117 44 99
pore size distribution (Table 2) were obtained by low-temperature nitrogen adsorption measurements of full (adsorption and desorption) isotherms (Figure 2). All of the isotherms shown in Figure 2 have a very small increase at low relative pressure values, which indicates minimal nitrogen adsorption interactions with hydrophobic surface. The hysteresis loops for the Halo and Allure adsorbents are also at relatively low p/ps values, which indicate that their pore sizes are smaller when compared to the other adsorbents. The last column in Table 2 lists the C-constant values of the BET equation, which were calculated for each of the adsorbents in this study. All these values are between 20 and 22, which, according to Karnaukhov,18 correspond to the adsorption of nitrogen molecules on a polyethylenelike surface or a surface covered mainly with methylene (CH2) groups. Buyanova suggested that nitrogen will occupy a molecular area of ∼20.5 Å2 on that surface type. The surface area values corrected for this larger area value for molecular nitrogen are shown in column 3 of Table 2. In earlier publications,15,22 the use of the base silica surface for the calculation of excess adsorption isotherms and other surface-specific parameters was suggested. This approach is appropriate for the comparison of the chromatographic and adsorption behavior of analytes on adsorbents made with the same base silica and when chemical surface modification did not alter significantly the adsorbent geometry. In a recent paper, Buszewski23 suggested the use of a cylindrical pore model for the determination of the actual surface area of modified adsorbents, which assumes the applicability of the following relationship between pore volume, surface area, and pore radius, Vp R ) S 2
to the values that are determined when the cylindrical pore model is assumed, but the more asymmetric the pore size distribution the greater the difference between these values. The minimum relative standard deviation of the differences between the pore radii calculated using the 2V/S relationship and the median pore radii obtained from the pore size distribution is 35%. The Allure and YMC modified silicas show the best correlation with the cylindrical pore model, even though their pore size distributions are not the most symmetric ones. Interestingly, the Xterra and Xbridge adsorbents have the most asymmetric distributions and also show the highest differences in calculated and measured pore radii.
Figure 2. Full nitrogen (adsorption and desorption) isotherms for all studied adsorbents.
(6)
where the ratio of the specific pore volume (Vp) to the specific surface area (S) should be equal to half of the pore radius (R). Figure 3 shows the pore size distributions for the adsorbents in this study. The maximum and median values of the pore radius and the pore radius calculated using eq 6 for each of these adsorbents are listed in columns 5-7, respectively, of Table 2. As shown by these data, the median pore radius values are closer
Figure 3. Pore size distribution of studied adsorbents. Analytical Chemistry, Vol. 80, No. 16, August 15, 2008
6361
Table 4. Alkylbenzene Retention Volumes (mL) 70/30 acetonitrile/water adsorbent Halo-C18 Allure-C18 Xterra-C18 YMC-C18 Gemini-C18 XBridge-C18
70/30 methanol/water
benzene
toluene
ethylbenzene
benzene
toluene
ethylbenzene
0.814 3.778 1.250 1.238 0.469 1.254
1.027 5.137 1.518 1.579 0.574 1.551
1.308 6.886 1.863 2.025 0.711 1.938
1.122 5.748 1.796 1.658 0.622 1.674
1.688 9.323 2.518 2.507 0.891 2.415
2.491 14.235 3.543 3.695 1.259 3.44
Table 5. Retention Factors 70/30 acetonitrile/water adsorbent Halo-C18 Allure-C18 Xterra-C18 YMC-C18 Gemini-C18 XBridge-C18 RSD, %
70/30 methanol/water
benzene
toluene
ethylbenzene
benzene
toluene
ethylbenzene
0.842 1.738 0.689 1.081 1.022 0.766 33.9
1.324 2.722 1.051 1.654 1.474 1.185 35.1
1.959 3.99 1.518 2.403 2.065 1.73 35.7
1.122 5.748 1.796 1.658 0.622 1.674 33.7
1.688 9.323 2.518 2.507 0.891 2.415 35.8
2.491 14.235 3.543 3.695 1.259 3.44 36.5
The most symmetric pore size distributions are for the Halo and Gemini adsorbents. These significant deviations from the cylindrical pore model may be due to the pore connectivity (networking effect) and indicate that it is better to avoid the use of the pore radius value in any adsorbent geometry calculations. Column Characterization. The void volume of each column was measured by using the retention of deuterated methanol eluted with pure methanol, and the interparticle volume was measured using an inverse GPC method. Table 3 shows the chromatographic characterization data for all columns including the measured and calculated weight of packing material in each column. The mass of adsorbent in the column was calculated according to the procedure described in ref 15 and using eq 3. The actual surface areas of each column were calculated using the determined mass of adsorbent in the column and the adsorbent surface area (corrected for the increased molecular area of nitrogen on a hydrophobic surface (20.5/16.2)) and are listed in the last column of Table 3. As one can see, there is a significant difference among these columns in the total surface available for interaction with analytes. The relative standard deviation (RSD) of the ratio between the column void volumes to the empty column volume is only 12% for all of the columns in this study, whereas the deviation of the ratio between the column surface area and the empty column volume is 30%. Significant variation in the total surface area available in a column can be considered as an important factor in influencing the deviation of the retention of the same analyte on different columns that have the same surface chemistry. To verify the influence of the adsorbent surface area on retention, we determined and compared the retention volumes of three simple analytes, which were run on chromatographic systems that were purposely selected to avoid any secondary interactions. In this case, we determined the retention of benzene, toluene, and ethylbenzene on the columns in our study using two mobile phases; 70 acetonitrile/30 water and 70 methanol/30 water. The retention volumes of benzene and the two alkylbenzenes are shown in Table 4 for all columns. In addition, the retention factors for the three analytes were then calculated and are shown in Table 6362
Analytical Chemistry, Vol. 80, No. 16, August 15, 2008
5 together with the relative standard deviation for each analyte on all columns. The variation in retention factor is on the level of 35% for all of the analytes. In 1970, Kiselev12 described the chromatographic retention process on the basis of the theory of adsorption from solutions and argued for the use of surface-specific retention factors. In a more recent article, Kovats13 introduced the “aerial retention volume” as the reduced retention volume related to the adsorbent surface area. He argued that this parameter represents the slope of the excess adsorption isotherm of an analyte at a given eluent composition and is a model-independent characteristic of analyte retention. We will designate this parameter as the “surfacespecific” retention factor, which essentially represents the fraction of the analyte retention associated with 1 square meter of the adsorbent surface. This parameter could be calculated as,
ks )
VR - V0 Stot
(7)
where VR is the analyte retention volume, V0 is the column void volume, and Stot is the total surface area of the adsorbent in the column. We would not argue that this parameter is modelindependent and immediately represent the analyte adsorption equilibrium, as stated by Kovats,13 since the studied chromatographic system is essentially a three-component system and the analyte adsorption isotherm is not a line but a surface. Thus, the slope of the isotherm is dependent on the mutual reaction of the analyte and organic modifier to the equilibrium disturbance introduced by the injection of the analyte. As it is also related to the proper measurement of the total adsorbent surface area, this parameter is undoubtedly better suited for the comparison of different columns with similar surface chemistry. The illustration of this statement is shown in Table 6, where the surface-specific retention factors are shown for all of the studied analytes. As one can see, the relative standard deviation of the retention parameters for the same analyte at the same conditions among different columns has decreased 9-10 times, which is clearly visible
Figure 4. Column-to-column variation of surface-specific retention factors (left) and conventional retention factors (right). Table 6. Surface-Specific Retention Factors (µL/m2) 70/30 acetonitrile/water adsorbent Halo-C18 Allure-C18 Xterra-C18 YMC-C18 Gemini-C18 XBridge-C18 RSD, %
70/30 methanol/water
benzene
toluene
ethylbenzene
benzene
toluene
ethylbenzene
5.30 5.39 5.15 5.48 5.77 5.51 3.5
8.33 8.44 7.86 8.39 8.32 8.52 2.5
12.33 12.37 11.35 12.20 11.66 12.44 3.4
9.68 9.81 10.67 9.07 9.49 9.77 4.9
17.74 17.84 17.97 16.31 16.04 17.28 4.4
29.17 28.87 28.33 26.44 24.99 27.66 5.3
in Figure 4 where retention factors (right pane) and surface-specific retention (left pane) are shown graphically for each column. Bonding Density Evaluation. The authors of ref 23 also suggested the use of bonding density value for the calculation of the modified adsorbent surface area. Bonding density is a very important adsorbent characteristic because it indicates the degree of hydrophobicity of the corresponding packing material. However, this statement could be questionable for modern composite materials, such as Xterra, XBridge, and Gemini, since the base silica itself already contains alkyl moieties and thus it is less hydrophilic and accepts a lower degree of chemical modification. In addition, these composite materials also could not be analyzed by the end-user for verification of the carbon loading and bonding density parameters. Halo, Allure, and YMC materials are based on regular base silica, and TGA analysis in an air environment was used to obtain the weight loss and to collect a sufficient amount of base silica material for subsequent nitrogen adsorption analysis. The direct correlation of the TGA and CHN analyses was previously shown by Fadeev24 for different types of ligands bonded on the surface of the silica. It was shown previously that a bonded C18 ligand occupies ∼600 Å3 on the surface of silica,15 which is in good agreement with the molecular volume of eicosan (595 Å3) that was calculated using ACDLabs (Toronto, Canada) software. These values could be used for the estimation of bonding density from the difference between the pore volumes of unmodified and modified adsorbent. The pore volume of the base silica minus the volume of bonded layer is equal to the pore volume of modified adsorbent per gram of base silica. Vp(mod.) Vp(SiO2) - νligdbSSiO2 ) f
(8)
(24) McElwee, J.; Helmy, R.; Fadeev, A. Y. J. Colloids Interface Sci. 2005, 285, 551–556.
Table 7. Manufacturer’s Data on Materials Surface Chemistry manufacturer’s data S (base)
db
C%
db (calc)
150 na 334
3.5 na na
naa 27 15.9
na 3.6 2.5
Halo-C18 Allure-C18 YMC-C18 a
na, not available.
The conversion of the pore volume per gram of base silica to the measured specific pore volume of modified silica could be done using a correction factor:15 f)
1 1 + dbSSiO2MWlig
(9)
where db is the bonding density, S is the surface of base silica, and MWlig is the molecular weight of the bonded ligand. Solving these two equations for db we obtain db )
Vp(SiO2) - Vp(mod.) SSiO2(νlig + Vp(mod)MWlig)
(10)
Another way to calculate the bonding density is based on the weight loss values obtained from TGA experiments, and could be done with the following expression, db )
w% mwligSSiO2(100 - w % )
(11)
where w% is the percent of weight loss, mwlig is the weight of the cleaved portion of the ligand (283 g/mol), and SSiO2 is the surface Analytical Chemistry, Vol. 80, No. 16, August 15, 2008
6363
Table 8. Comparison of the Bonding Density Values Calculated Using Different Methods
Halo-C18 Allure-C18 YMC-C18
surface area, m2/g
pore volume (modified), m2/g
pore volume (base silica), mL/g
weight loss, w%
db (from pore volume), µmol/m2
db (from weight loss), µmol/m2
132 460 351
0.134 0.335 0.586
0.275 1.12 1.12
8 32 18
2.6 3.7 2.7
2.5 3.6 2.3
area of silica. In our calculations, we assume that all of the organic moieties were completely burned off into the air, while the silica atoms were left on the surface. All TGA experiments were performed by heating at 10 °C/min up to a final temperature of 550 °C. Subsequent BET analysis of the materials that remained after TGA showed C-constant values on the level of 50, which indicates a partial dehydroxylation of the silica surface. These calculations can be only considered as an estimate, since the degree of dehydroxylation is unknown, although it probably does not introduce a very significant error. The characteristics of the packed materials that were provided by the manufacturers are shown in Table 7. In the last column of Table 7, the bonding density values that were calculated using the manufacturers’ data are shown. For the Allure material, the surface area measured after TGA was used in the calculation, since manufacturer did not provide this information. The adsorbent pore volume, measured surface area, and bonding density calculated using eqs 10 and 11 are shown in Table 8. As could be seen from comparison of the last two columns of Table 8, both methods gave consistent results. Bonding density values for Allure and YMC materials are also close to the values that were calculated from the manufacturer’s data, while for the Halo material, the given value is much higher than the value calculated from the decreased pore volume and the TGA weight loss.
of different reversed-phase columns under nearly ideal conditions, and in the absence of secondary equilibria effects. The calculation of surface-specific retention factors requires proper determination of the specific surface area of the packing material and the amount of packing material in the column. The specific surface area of chemically modified silica can be determined using standard lowtemperature nitrogen adsorption measurements and by the application of BET theory. However, for C18-type modified (hydrophobic) surfaces, the surface area values should be calculated using 20.5 Å2 (instead of 16.2 Å2) for the value of the nitrogen molecular area. The amount of adsorbent in the column can be determined from the difference between the column void volume and interparticle volume related to the specific pore volume of modified adsorbent. The everyday practical use of surface-specific retention factors could significantly simplify column selection and column comparison; although, for general chromatographers, these advantages will be realized only if column manufacturers regularly provide information about the adsorbent surface area and pore volume of the chemically modified surface.
CONCLUSIONS We demonstrated that surface-specific retention factors are more suitable for the comparison of chromatographic behavior
Received for review April 17, 2008. Accepted June 11, 2008.
6364
Analytical Chemistry, Vol. 80, No. 16, August 15, 2008
ACKNOWLEDGMENT This work was supported by Civilian Research and Development Foundation, grant RUC2-2873-PE-07.
AC800772S
