Characterization of a Range of Alkyl-Bonded Silica HPLC Stationary

David A. Barrett,* Vikki A. Brown, Martyn C. Davies, and P. Nicholas Shaw. Department of Pharmaceutical Sciences, University Park, University of Notti...
0 downloads 0 Views 538KB Size
Anal. Chem. 1996, 68, 2170-2178

Characterization of a Range of Alkyl-Bonded Silica HPLC Stationary Phases: Correlation of Quantitative Surface Analysis Data with the Retention Behavior of Neutral, Acidic, and Basic Solutes David A. Barrett,* Vikki A. Brown, Martyn C. Davies, and P. Nicholas Shaw

Department of Pharmaceutical Sciences, University Park, University of Nottingham, Nottingham NG7 2RD, U.K.

An investigation was made of the correlation between quantitative surface analytical data obtained by XPS and static SIMS and the chromatographic performance of a range of n-alkyl-bonded silica (C1-C18) packing materials. A series of acidic, basic, and neutral solutes was used to study the retention behavior. For comparison, analysis of bulk total percentage carbon (%C) and alkyl surface density of the bonded silica particulates were also included. Significant correlations were observed, in the majority of cases, between the retention factor (k) and the XPS C:Si atomic ratio, which was similar to that obtained between k and the bulk %C or k and the bonded alkyl chain length. Similar significant correlations were also obtained between k and the static SIMS alkyl:Si ion peak area ratios. XPS alkyl:Si atomic ratios were calculated as an estimate of alkyl surface coverage of the silica support, and these correlated well with the surface density calculated from the bulk %C and the surface area of the packing material. The XPS alkyl:Si ratio also demonstrated a significant correlation with the peak asymmetry factor derived for basic solutes. These studies confirm that both XPS and static SIMS can generate surface chemical data from chromatography particulates, which has direct relevance to the prediction of chromatographic behavior. We believe that these techniques will prove to be effective tools to assist in the characterization of chromatographic supports and stationary phases. In reversed-phase HPLC, the relationship between the chemical nature of the bonded phase and its separation behavior has been studied in considerable detail by a wide range of techniques.1-4 The techniques applied include the chromatographic behavior of probe solutes, spectroscopic studies of the bonded phase, bulk analysis of the chemical content, and the indirect measurement of surface functionalities. Most of these techniques used to study the support material and/or the stationary phase do not provide a direct and quantitative measurement of the chemical nature of the chromatographic surface. Solid state NMR and diffuse reflectance IR spectroscopy can provide detailed structural information relating to the bulk chromatographic silica materials, such (1) Sander, L. C.; Wise, S. A. Crit. Rev. Anal. Chem. 1987 18, 299-415. (2) Dorsey, J. G.; Cooper, W. T. Anal. Chem. 1994 66, 857A-866A. (3) Nawrocki, J.; Buszewski, B. J. Chromatogr. 1988 449, 1-24. (4) Rutan, S. C.; Harris, J. M. J. Chromatogr. A 1993, 656, 197-215.

2170 Analytical Chemistry, Vol. 68, No. 13, July 1, 1996

as the relative proportion of different types of silanols or alkyl groups, but these techniques give a measurement of the bulk chemical composition rather than the chemistry of the chromatographically relevant surface. It is generally accepted that interactions between the solute and stationary phase take place exclusively on the outer few nanometers of the chromatographic support, and hence it can be argued that surface-specific measurements have the greatest relevance to understanding the separation behavior of a chromatographic material. The most widely used reversed-phase HPLC materials are based on microparticulate porous silica and are commercially available with bonded alkyl chain lengths ranging between C1 and C18. These materials have an array of surface functionalities which may influence chromatographic performance, the most important of which are the length and surface density of the bonded alkyl chains, varying amounts and types of unreacted silanols, and a wide range of trace metals. It is proposed that, to understand fully the role of these surface substituents, highly surface-sensitive analysis techniques should be employed which can provide both quantitative and structural information relating to the chromatographic interaction surface of the bonded silica particles. The complementary surface analytical techniques of X-ray photoelectron spectroscopy (XPS) and static secondary ion mass spectrometry (SIMS) have the potential to provide such information but have seen relatively little use in this area of application. In a previous study,5 we have reported the use of XPS and static SIMS to characterize the surface chemistry of a range of alkyl-bonded silica materials. XPS, which yields surface information to a depth of 5-6 nm, provides a quantitative elemental analysis of the silica surface. Static SIMS, operated under conditions to give a sampling depth of 1-3 nm, gives structural information relating to the bonded alkyl groups, the silanols, and the more abundant trace metals. To date, few quantitative studies have been performed on the potential correlation between surface analytical data and the chromatographic behavior of bonded silica chromatography packing materials. The study reported here evaluates the correlation between quantitative surface analytical data obtained by XPS and static SIMS and the chromatographic behavior of a series of acidic, basic, and neutral solutes on eight monofunctional n-alkyl-bonded (5) Brown, V. A.; Barrett, D. A.; Shaw, P. N.; Davies, M. C.; Ritchie, H. J.; Ross, P.; Paul, A. J.; Watts, J. F. Surf. Interface Anal. 1994, 21, 263-273. S0003-2700(95)00829-8 CCC: $12.00

© 1996 American Chemical Society

Table 1. Physical Properties of the Alkyl-Bonded Silicas

samplea

no. of carbon surface chain bulk atoms coverageb length 2 %C in chain (µmol/m ) (nm)

base silica 0.06 trimethylsilyl silica (C1) 2.14 dimethylethyl silica (C2) 2.59 dimethyl-n-propyl silica (C3) 3.00 dimethyl-n-butyl silica (C4) 3.62 dimethyl-n-octyl silica (C8) 5.86 dimethyl-n-dodecyl silica (C12) 8.14 dimethyl-n-octadecyl silica (C18) 10.64

0 3 4 5 6 10 14 20

3.41 3.09 2.87 2.88 2.80 2.78 2.54

0.30 0.43c 0.56 0.69 1.20 1.73 2.45

a From the following: Roumeliotis, P.; Unger, K. K. J. Chromatogr. 1978, 149, 211-224. b Calculated from bulk carbon determination and surface area of the base silica. c Estimated on the basis of 0.13 nm increment for each additional -CH2- group.

silicas with alkyl chain lengths between C1 and C18. For comparison, data related to the analysis of bulk total percentage carbon (%C) and surface coverage are also included. EXPERIMENTAL SECTION Chemicals. Sodium dihydrogen phosphate, disodium hydrogen phosphate, potassium dihydrogen phosphate, and methanol (all HPLC grade) were obtained from Fisons plc (Loughborough, UK). Deionized, purified water was produced using an Elgastat water purification system (Elga Ltd., Cheltenham, UK). Chromatographic grade microporous silica was provided by Shandon HPLC, and n-dimethylalkylchlorosilanes were obtained from Fluka Chemicals Ltd. (Dorset, UK). All other chemicals were of analytical grade or better. All solvents required for moisturesensitive reactions were dried before use. Chemical reactions were performed in inert conditions under an atmosphere of dry nitrogen. Preparation of the Spherical Base Silica. A single batch (1.0 kg) of microparticulate base silica was prepared by the gelation of a colloidal silica under controlled conditions. The method of preparation has been described in detail.6 The pore size and pore volume of the silica particles were determined using mercury porosimetry (Micrometrics 9300 analyzer), and the surface area was measured by BET single-point analysis (Micrometrics Flowsorb 2300). The base silica material had a pore volume of 0.72 cm3/g, a pore size of 130 Å, a surface area of 174 m2/g, and an average particle size of 4.57 µm. Synthesis of the Alkyl-Bonded Silica Materials. A series of 5 µm alkyl-bonded silicas was prepared from a single batch of the hydrated Hypersil base silica using the appropriate ndimethylalkylsilyl chlorides with alkyl chain lengths of between C1 and C18. Monofunctional chlorosilanes were used in the reaction, and the subsequent n-alkyl-bonded silica materials were not end capped. Details of the synthetic method have been published.6 The final product was oven-dried at 80 °C prior to use. Using this method, a series of n-alkyl-bonded silicas was synthesized, as shown in Table 1. Chromatographic Measurements. The eight alkyl-bonded silica packing materials were slurry-packed into 10 cm × 4.6 mm i.d. stainless steel columns using acetone as a packing and slurry (6) Barrett, D. A.; Brown, V. A.; Shaw, P. N.; Davies, M. C.; Ritchie, H. J.; Ross, P. J. Chromatogr. Sci. 1996, 34, 146-156.

Figure 1. Structures of test solutes used for chromatographic evaluation.

solvent, at a pressure of 8000 psi. Prior to chromatographic testing, the columns were conditioned with methanol/water (70: 30). All the columns were subjected to rigorous quality control testing to ensure that the chromatographic performance was within the criteria accepted for a high-efficiency reversed-phase packing material. The reduced plate heights of these columns, based on a standard neutral probe compound and a particle size of 4.57 µm, ranged between 2.5 and 2.8, which is typical of columns of this type. The liquid chromatographic system consisted of two Gilson 303 HPLC pumps, a Gilson 116 ultraviolet detector, a Gilson 231/ 401 autoinjector, and Gilson 714 integration and data handling software (Gilson Medical Instruments, Villiers-le-Bel, France). The column temperature was controlled by a column oven set at 30 °C (Jones Chromatography Ltd., Mid Glamorgan, UK). All mobile phases were degassed by helium sparge before use. Three series of test solutes were chosen, comprising neutral, weakly acidic, and weakly basic compounds of pharmaceutical relevance: 4-substituted biphenyls, disubstituted barbiturates, and 3-substituted pyridines, respectively. The substituents on each of the above test series were selected to include a wide range of lipophilicities. Details of all the test compounds are shown in Figure 1. The test solutes, including acids and bases, were chosen to be predominantly in their un-ionized form at the pH of analysis. The stability of the test solutes under the conditions of storage and analysis was established to ensure that degradation products did not interfere with the chromatographic analysis. For the barbiturate standards, the mobile phase was pH 3.0 aqueous potassium phosphate buffer (0.005 M)/methanol (68: 32), and the detection wavelength was 210 nm. For the biphenyl standards, the mobile phase was water/methanol (39:61), and the detection wavelength was 260 nm. For the pyridine standards, the mobile phase was pH 7.0 aqueous sodium phosphate buffer (0.005 M)/methanol (87:13), and the detection wavelength was 205 nm. The standard solutions of all the test compounds were made up in the appropriate HPLC mobile phase at a concentrations of 50 µg/mL for the biphenyls and 100 µg/mL for the barbiturates Analytical Chemistry, Vol. 68, No. 13, July 1, 1996

2171

and pyridines. An injection volume of 20 µL was used for analysis of all the test solutions, and the flow rate was 1.0 mL/min. The retention behaviors of the compounds in the neutral, acidic, and basic test mixtures were characterized as follows. Retention factor (k) and peak asymmetry factor were measured from replicate (n ) 6) injections of each test compound on each column evaluated. The retention factors and asymmetry factors were calculated as follows:

retention factor (k) ) (tR - tM)/tM

where tM is the mobile phase holdup time and tR is the total retention time of the solute. The holdup time was taken as the time from injection to the moment when the trace solvent disturbance crossed the baseline. tM was found to be 0.8 min at a flow rate of 1.0 mL/min using a 100 mm × 4.6 mm column in all the mobile phases used. The peak asymmetry factor was calculated according to Kirkland et al.7 Elemental Analysis and Surface Coverage. Elemental analysis data of percentage carbon were obtained for each of the bonded phase materials using an EC-12 carbon determinator (Leco Corp., St. Joseph, MI). The percentage of carbon present in the sample was calculated to a precision of (0.04% and was used to calculate the alkyl surface density according to the following equation:

surface coverage (µmol/m2) )

%C × 106 1200nA

where %C is expressed as w/w, n is the number of carbons in the bonded phase, and A is the surface area of the substrate (m2/ g). The percentage carbon values and the surface coverage data are reported in Table 1. XPS Analyses. The XPS data were obtained using a VG Scientific ESCALAB Mk II electron spectrometer, employing Al KR X-rays (hν ) 1486.6 eV). The X-ray gun was operated at 10 kV and 34 mA giving 340 W, and spectra were recorded at an electron take-off angle of 45°, providing a sampling depth of around 5 nm. The silica powders were slurried in methanol (Fisons plc, Loughborough, UK; HPLC grade) and dropped onto a copper foil (Aldrich Chemical Co. Ltd., Dorset, UK; 99+% purity, 0.127 mm thickness)-covered sample stub. A survey spectrum (0-1000 eV) and high-resolution spectra of the Cu 2p3 (925-950 eV), C 1s (280-310 eV), O 1s (525-550 eV), Si 2p (95-120 eV), Cl 2p3 (190-215 eV), Fe 2p3 (700-740 eV), and Na 1s (1065-1090 eV) regions were recorded for each sample. The analyzer was operated in fixed transmission mode with a pass energy of 50 eV for survey scans, 100 eV for Cu 2p3 narrow scans (not included in quantification), and 20 eV for other narrow scans. Peak positions were corrected to C 1s at 285.0 eV. Estimation of the Si:C and Si:O surface atomic ratios involved the calculation of integrated areas using Schofield atomic sensitivity factors, corrected for the analyzer transmission function of the spectrometer. Static SIMS Analyses. TOF-SIMS spectra were obtained using a VG IX23S instrument, based on a Poschenrieder design (7) Kirkland, J. J.; Yau, W. W.; Stoklosa, H. J.; Dilks, C. H. J. Chromatogr. Sci. 1977, 15, 303-316.

2172 Analytical Chemistry, Vol. 68, No. 13, July 1, 1996

Figure 2. X-ray photoelectron spectroscopy survey scan from n-dimethyloctyl (C8)-bonded silica.

and equipped with a pulsed liquid metal ion source. A 30 keV Ga+ primary ion beam (ion flux density ) 5 × 10-11 A cm-2) was focused onto an area of 0.6 mm × 0.6 mm, and the resultant secondary ions were accelerated to 5 keV for the analysis by applying a bias to the sample. Charge compensation was achieved using a pulsed electron flood source, operating at an effective electron energy of 250 eV and an electron flux density of 1 × 10-8 A cm-2. Mounting of the samples for static SIMS analyses involved lightly pressing the silicas into indium foil (Johnson Matthey Materials Technology UK, Royston, UK; grade 1, 1 mm thick), dusting the silica powders onto double-sided adhesive tape, or placing the silica powder in a sample dish. Positive and negative secondary ion spectra were collected in the region m/z 0-200. A flood gun counteracted surface charging during negative ion collection. The static SIMS technique is highly surface-specific and hence is very easily influenced by contributions from adsorbed hydrocarbon contaminants. In view of this, special care was taken in the sample preparation and storage to minimize any adventitious surface contamination which might produce erroneous or misleading results. The success of the these measures was confirmed by the lack of contaminants seen on the SIMS spectra due to common surface contaminants such as phthalate plasticizers. A plot of m/z versus intensity was produced for each sample, allowing the identification of trace elements and fragments attributable to base silica and alkyl chains. The numbers of counts per second were recorded, for each ion, as peak areas over a range (0.5 m/z. The relative ion intensity was calculated using (A/B) × 100, where A is the ion peak area and B is the total sum of all the positive ion peak areas for that sample. RESULTS AND DISCUSSION XPS Measurements. Figure 2 shows a typical XPS survey spectrum of C8-bonded silica, indicating the major elements which were identified on the surface. The %C and the C:Si atomic ratio determined by XPS demonstrated a linear correlation with the %C determined by bulk analysis and with the number of carbons in the silane reagent (R2 > 0.99, p < 0.001), as shown in Table 2. This is in agreement with previous reports of XPS analysis of a series of ODS-bonded silicas.8,9 However, the previously reported nonlinearity of the above XPS plots at the higher alkyl chain lengths, attributed to an attenuation of the Si substrate signal by (8) Linton, R. W.; Miller, M. L.; Maciel, G. E.; Hawkins, B. L. Surf. Interface Anal. 1985, 7, 196. (9) Miller, M. L.; Linton, R. W.; Bush, S. G.; Jorgenson, J. W. Anal. Chem. 1984, 56, 2204-2210.

Table 2. XPS Atomic Ratios from Alkyl-Bonded Silicas XPS data sample

Si:O

C:Si

alkyl:Si

base silica C1 C2 C3 C4 C8 C12 C18

2.04 2.27 1.82 1.92 1.96 1.82 1.67 1.69

0.09 0.20 0.25 0.27 0.29 0.49 0.75 1.02

0.067 0.063 0.054 0.048 0.049 0.054 0.051

the carbon-containing overlayer,9 did not occur in our studies, and an excellent correlation was obtained over the full range of C1C18 alkyl chain lengths studied (Table 2). The linearity of the XPS plots may be attributed to the similarity of the surface density of all the alkyl-bonded silicas used in this investigation. The small percentage of carbon found on the unbonded silicas was attributed to adventitious contamination and did not make a major contribution to the bonded alkyl group measurements. The surface specificity of the XPS technique can be exploited to estimate the surface coverage of bonded groups at the chromatographically important outer few nanometers of the support surface. It has been established above that quantitative results for %C or C:Si ratios are obtainable by XPS; hence, these ratios reflect the alkylsilane reaction with the surface silanol groups of the silica support. To provide a measure of surface density of the bonded alkyl groups from the XPS data, an alkyl:Si ratio was calculated by dividing the XPS C:Si ratio by the number of carbon atoms per attached alkyl group (Table 2). Similar data have been obtained from other XPS studies8,10 of octadecylsilyl and trimethylsilyl bonded phases but not for the range of alkyl chain lengths used in this study. The alkyl:Si ratio gives an indication of the number of bonded alkyl groups per silicon atom and hence provides a realistic estimate of the surface coverage determined from the outer 5 nm of the chromatographic particulate. A weak linear correlation (Figure 3; R2 ) 0.708, p ) 0.017) was obtained between the surface coverage determined from bulk physical properties of the silica (bulk %C and surface area), but there were marked deviations from linearity at the lower alkyl:Si ratios. The reason for this deviation is not clear, but it may be associated with the limitations in precision of the XPS technique when measuring the very small differences in surface carbon content between the C1, C2, and C3 alkyl-bonded materials. The alkyl:Si ratio reflects the extent of bonding of the alkylsilanes with the silanol groups of the silica surface. This ratio represents an alternative approach to the calculation of alkyl surface coverage and has advantages compared with the traditional approach of determining the surface area and the bulk %C of a packing material. The bulk analysis approach is difficult to use for packing materials which have elements common to the bonded phase and the support matrix, whereas surface analysis is applicable to a wide variety of stationary/support matrixes. By taking the inverse of the XPS alkyl:Si ratios (Table 2), it is possible to estimate the number of silicon atoms present on the surface for each bonded alkyl group. Between 15 and 22 silicon atoms were present for each alkyl group for the range of bonded groups from C1 to C18. Whereas the %C measured by XPS is (10) Miller, M. L.; Linton, R. W. Anal. Chem. 1985, 57, 2314.

Figure 3. Correlation between surface coverage of the bonded silicas determined from bulk %C and the XPS-determined alkyl:Si ratio.

due entirely to the alkyl chain, the %Si represents several different surface species: siloxane internal bonds (Si-O-Si), surface silanol groups (SiOH and SiO-, unreacted silanols), and silane chains (SiOSi(CH3)2-alkyl). Thus, the absolute ratio of silicon atoms per alkyl group is much lower than would be anticipated on the basis of a knowledge of the bulk surface coverage (Table 2). The sampling depth of the XPS instrument under the operating conditions used is ∼5 nm, and since the distance between silicon atoms in the silica has been calculated to be about 0.5 nm, the Si signal is representative not of the outermost molecular layer of silica but of a combination of the outer surface and several of the underlying layers. An additional factor is that each surface-bonded alkyl group has an additional Si in the silane, thus increasing the Si signal. Hence, for each surface alkyl group detected, there is a contribution from subsurface signals from the silicon atoms, which overestimates the real surface atomic ratio. However, the ratio of alkyl groups to Si atoms is more important than the absolute values and can provide quantitative information on the bonded phase and the underlying silica surface. Provided that comparative data are available from bonded materials with a range of ligand lengths or surface densities, it is possible to obtain useful measurements from the chromatographically relevant surface. Unbonded silicas would be predicted to give an Si:O stoichiometric ratio of 1:2 (i.e., SiO2). This was confirmed by our XPS data giving an Si:O ratio of SiO2.04 for the base silica examined. The bonded silicas would be expected to produce an Si:O stoichiometry of less than 1:2 because of the surface-bonded silane contributing an additional Si signal and due to the preferential attenuation (see later) of the O signal by the bonded alkyl layer. Such a trend was observed in the XPS data, and in general, as the alkyl chain length was increased, the Si:O ratio decreased to a minimum of SiO1.7 for the C8 and C18 chains, with the one exception of the C1-bonded material, which had an Si:O ratio of SiO2.27 (Table 2). The increased Si signal with the longer alkyl chain lengths can be explained by the following two factors: (1) As the alkyl chain length increases, there is a diminution of the photoelectron signals from the Si and O. Since the O 1s photoelectron inelastic mean free path in an organic film is Analytical Chemistry, Vol. 68, No. 13, July 1, 1996

2173

estimated to be 2.3 nm compared with the Si 2p photoelectron inelastic mean free path of 3.0 nm,11 the O signal will be attenuated more. The length of the bonded alkyl group overlayer ranges from 0.3 (C1) to 2.45 nm (C18) (Table 1) and hence will contribute to this attenuation. (2) The bonding of the silane groups adds one Si atom per chain. Hence it would be expected that the Si:O ratio would be increased in the alkyl-bonded silicas. The unusually high O 1s signal in the C1 ABS material cannot be adequately explained by the arguments above and may be due to adsorption of water molecules to the silica surface. The presence of trace elements on the surface of alkyl-bonded silicas can have profound effects on the chromatographic behavior of these materials, but such species have not previously been analyzed by surface analytical techniques. Attempts were made to determine the level of trace metals on the range of bonded silica samples using XPS. Although it was possible to detect the predominant trace elements (Na, K, Al, Cl) in the base silica, it was not possible to detect other trace elements, such as zinc, cobalt, nickel, iron, chromium, magnesium, copper, titanium, and calcium, which have been measured in this batch of silica by a bulk analysis method.12 In addition, the increasing length of the bonded alkyl group resulted in a reduction of the signal from the trace elements, such that sodium was not detectable with alkyl chain lengths of C2 and above. XPS is insufficiently sensitive when compared with other elemental analysis techniques which can measure the bulk metal content of a material (such as ICP-AES). XPS is hence unable to quantify the full range of trace elements on the silica surface. Static SIMS Measurements. Static SIMS, as applied to chromatography particulates, is primarily of use in the confirmation of chemical identity of ligands bonded to the surface of the support material. Data obtained on the structural identification of the alkyl ligands from the range of alkyl bonded materials used in this study have been presented elsewhere.5 However, data from static SIMS can also be interpreted to provide a quantitative measure of the ratios between fragments of the matrix and surfacebonded species. To obtain meaningful results from the SIMS raw ion intensities, it is necessary to use a normalization procedure which involves scaling the response of the peaks of interest to an ion (or set of ions) representative of the bonded alkyl phase. In this study, the total sum of all the raw positive ion peak areas from m/z 0 to 100 was used as the reference set of ions for each alkyl-bonded silica sample. The relative ion intensity was then obtained by dividing the peak area of the ion of interest by the m/z 0-100 total peak area. This method of comparing ion intensities has been used in SIMS and FAB-MS analysis to provide a semiquantitative measure of the variation in surface substituents between different samples.13 The data presented here have been selected for their potential relevance to the chromatographic separation process and consist of ratios of the ion intensities of alkyl- to silica-based species. Several of the key low-mass fragments of the bonded silica surface have multiple assignments, which have been characterized in a previous study,5 and this has been accounted for in the selection of predominantly Si- or alkylbased ions for correlation with the chromatographic results (Table 3). The alkyl species were carefully chosen to be representative of the entire range of alkyl ligands from a minimum of 3 or 4 (11) Seah, M. P.; Dench, W. A. Surf. Interface Anal. 1979, 1, 2-11. (12) Simko, S. J.; Miller, M. L.; Linton, R. W. Anal. Chem. 1985, 57, 2448. (13) Watson, R. C.; Davies, M. C.; Shaw, P. N.; Barrett, D. A., unpublished data.

2174 Analytical Chemistry, Vol. 68, No. 13, July 1, 1996

Table 3. Major Ions and Assignments Obtained from Static SIMS Analysis of Alkyl-Bonded Silicasa

a

m/z

ion assignment

main contribution

27 28 29 41 43 44 45 55 57 59

C2H3 Si+ SiH+/ 29Si+/C2H5+ C3H5+ SiCH3+/C3H7+ SiO+ SiOH+ C4H7+ SiC2H5+/C4H9+ (CH3)2SiH+

alkyl Si Si/alkyl alkyl Si/alkyl Si Si alkyl alkyl Si/alkyl

+/Al+

Compiled from original data and ref 5.

Figure 4. Positive ion TOF-SIMS spectrum (m/z 0-100) recorded for n-dimethyloctyl (C8)-bonded silica. Table 4. TOF-SIMS Alkyl:Si Ion Peak Area Ratios m/z bonded silica chain length 27/28 41/28 57/28 45/44 27/44 C1 C2 C3 C4 C8 C12 C18

0.064 0.198 0.296 0.217 0.512 0.736 0.805

0.073 0.152 0.226 0.136 0.477 0.881 1.159

0.066 0.208 0.101 0.083 0.176 0.373 0.665

3.011 1.952 2.352 2.949 3.046 2.402 2.297

0.412 1.048 2.117 3.221 6.389 9.232 8.878

41/44

57/44

0.466 0.802 1.612 2.022 5.954 11.049 12.784

0.423 1.099 0.724 1.235 2.194 4.671 7.338

carbons to 22 carbons; hence, only fragments containing up to four carbons were assessed. The peaks at m/z 27, 41, and 57 were chosen as characteristic of the bonded alkyl chain, attributed predominantly to C2H3+, C3H5+, and C4H9+. Higher mass fragments containing C5 and above would be applicable to a narrower range of bonded alkyl groups but have not been used in this analysis. The predominantly silicon-based peaks were m/z 28 and 44, representing Si+ and SiOH+, respectively. Positive ion TOF-SIMS spectra (m/z 0-100) were recorded for all the alkyl-bonded silicas. A typical spectrum (m/z 0-100) recorded for a C8-bonded silica is shown in Figure 4. From the raw ion peak areas taken from these spectra, a series of relative ion intensities of alkyl:Si peaks were calculated (Table 4). These ion ratios gave an excellent correlation with the data obtained by XPS, thus confirming the quantitative nature of the data obtained from the SIMS analysis. Certain ions had intensity patterns which did not correlate well with the alkyl chain length, and these were found to be dominated by contributions from substrate inorganic ions and were not used in the analysis. In the static SIMS analyses, there was a marked attenuation of the Si+ and SiOH+ substrate signal as the the length of the

Figure 8. Correlation between retention factors of the pyridine (basic) solutes and the XPS-determined C:Si atomic ratio. 0, 3-Ethyl (R2 ) 0.980); b, 3-bromo (R2 ) 0.938); 4, 3-chloro (R2 ) 0.938); 9, 3-amino (R2 ) 0.923); 2, pyridine (R2 ) 0.912); O, 3-acetyl (R2 ) 0.898). Figure 5. Effect of bonded alkyl chain length on the normalized SiOH+ ion peak intensity.

Figure 6. Correlation between the retention factors of the biphenyl (neutral) solutes and the XPS-determined C:Si atomic ratio. b, 4-Ethyl (R2 ) 0.993); 9, 4-bromo (R2 ) 0.993); 0, 4-methyl (R2 ) 0.995); 2, biphenyl (R2 ) 0.985); 4, 4-acetyl (R2 ) 0.987); O, 4-hydroxymethyl (R2 ) 0.951).

Figure 7. Correlation between the retention factors of the barbiturate (acidic) solutes and the XPS-determined C:Si atomic ratio. 0, Allyl1-methylbutyl (R2 ) 0.981); 2, ethylisopentyl (R2 ) 0.975); b, ethylbutyl (R2 ) 0.971); 4, ethylphenyl (R2 ) 0.976); 9, diethyl (R2 ) 0.873) (C1-C8 alkyl chain length).

bonded alkyl chain increased (Figure 5). These data clearly indicate the surface shielding effect of the longer alkyl chains. Correlation between XPS Surface Analysis and Chromatographic Behavior. A detailed examination of the chromatographic retention behavior of the test solutes has been reported elsewhere,6 including the numerical data for k, peak asymmetry factors, and plate numbers. Hence, these data will not be reported

here. This section is concerned with the evaluation of potential correlations between the surface analytical data and the chromatographic retention behavior. The chromatographic retention factors of each series of test solutes were compared with the XPS C:Si atomic ratio for each of the alkyl-bonded silicas (C1-C18). The neutral biphenyl solutes showed a strong linear correlation between the XPS C:Si atomic ratio and retention factor (R2 ) 0.95-0.99 for each solute, p < 0.0004) (Figure 6). However, for the acidic barbiturate solutes and the basic pyridine solutes, the linear relationship was not as strong over the C1-C18 range. The barbiturate retention factors showed little change above an XPS C:Si ratio of 0.5 (equivalent to a C8 chain length), although they increased linearly up to this value (R2 ) 0.87-0.98, p < 0.0167) (Figure 7). This effect can be explained by the mechanism of interaction between the solutes and the alkyl-bonded stationary phase and is discussed elsewhere.6 The retention behavior of the pyridine solutes followed the pattern of the neutral biphenyl solutes, with a linear correlation between the XPS C:Si ratios and retention factor (R2 ) 0.90-0.98, p < 0.012) (Figure 8), although several of the solutes showed deviations from a linear relationship. These results establish that a quantitative relationship exists between the XPS-determined surface elemental composition of a bonded silica packing material and the chromatographic behavior of a range of test solutes on the same material. The XPS C:Si atomic ratio plots are similar in shape to those obtained from bulk carbon analysis or alkyl chain length of the bonded silica packing materials.6 Previous work on the XPS analysis of alkyl-bonded silicas has examined the different surface coverages of an octadecyl (C18) alkyl ligand.8 In this work, Miller et al.8 demonstrated that the XPS C:Si ratios of C18-bonded silica materials did not correlate in the expected manner with retention data, which may have been due to the relatively low surface coverage used in their materials. Our studies have observed the effect of increasing alkyl chain length while maintaining an approximately constant surface coverage and have shown an extremely good correlation with chromatographic retention data. The present study indicates the usefulness of XPS atomic ratios for the quantitation of surface ligands bonded to chromatographic materials and suggests that the technique will be equally useful in providing quantitative data on ligands bonded to a broad range of chromatographic materials, including both inorganic and polymer-based support materials. Analytical Chemistry, Vol. 68, No. 13, July 1, 1996

2175

a

b

Figure 9. Correlation between XPS-determined alkyl:Si ratio derived from C1-C18 alkyl-bonded silicas and peak asymmetry values for 3-aminopyridine chromatographed on the same materials.

The extent of the surface coverage of the bonded groups is likely to have a direct effect on the ability of solutes to interact with the underlying surface of the silica. With alkyl-bonded silica materials, it has been observed that relatively small differences in the alkyl surface coverage can lead to large differences in chromatographic behavior, but the influence of the alkyl chain length in this process is not well understood.3 The presence of accessible and unreacted silanol groups on the underlying silica surface are known to influence retention behavior, often in a detrimental manner. These unreacted silanol groups are acidic and hence carry a negative charge when ionized. It is well established that ionized basic solutes will interact strongly with such accessible acidic silanol groups on the surface of bonded silicas, producing an undesirable increase in chromatographic retention. The masking of the residual silanols by the bonded alkyl groups may reduce these unwanted interactions, but the masking effect is likely to vary, depending on the alkyl chain length and surface coverage. The use of XPS alkyl:Si ratios for each packing material gives a surface measurement of ligand density, which may be correlated with chromatographic data reflecting secondary interactions between the solute and the packing material. Several of the basic pyridine solutes showed strong secondary interactions with the stationary phases, as indicated by severe peak tailing, suggesting that residual silanols were accessible under chromatographic conditions. The magnitude of these secondary interactions was inferred from the peak asymmetry factor, which was readily calculated from the chromatographic analysis.6 Plots of XPS alkyl:Si ratio versus the peak asymmetry factors determined for the basic pyridine solutes demonstrated a weak linear correlation for 3-aminopyridine, pyridine, and 3-ethylpyridine. However, only the most ionized of these, the 3-aminopyridine, showed a statistically significant linear correlation (R2 ) 0.703, p < 0.018; Figure 9). The correlation of the same chromatographic data with the traditional calculation of alkyl surface coverage (in µmol/m2) determined from the bulk %C data did not give statistically significant results for any of these basic solutes: 3-aminopyridine (R2 ) 0.348, not significant), pyridine (R2 ) 0.272, not significant), and 3-ethylpyridine (R2 ) 0.0915, not significant). Although the correlation seen was relatively weak, the greater 2176 Analytical Chemistry, Vol. 68, No. 13, July 1, 1996

c

Figure 10. Correlation between the retention factors of the biphenyl (neutral) solutes and the TOF-SIMS m/z 27/44, (a) 41/44 (b), and 57/44 (c) peak area ratios. (a) b, 4-Ethyl (R2 ) 0.866); 9, 4-bromo (R2 ) 0.867); 0, 4-methyl (R2 ) 0.877); 2, biphenyl (R2 ) 0.928); 4, 4-acetyl (R2 ) 0.937); O, 4-hydroxymethyl (R2 ) 0.980). (b) b, 4-Ethyl (R2 ) 0.974); 9, 4-bromo (R2 ) 0.973); 0, 4-methyl (R2 ) 0.978); 2, biphenyl (R2 ) 0.982); 4, 4-acetyl (R2 ) 0.992); O, 4-hydroxymethyl (R2 ) 0.974). (c) b, 4-ethyl (R2 ) 0.981); 9, 4-bromo (R2 ) 0.980); 0, 4-methyl (R2 ) 0.977); 2, biphenyl (R2 ) 0.949); 4, 4-acetyl (R2 ) 0.932); O, 4-hydroxymethyl (R2 ) 0.860).

significance with the XPS data compared with bulk data suggests that surface analysis can predict at least one critical aspect of chromatographic behavior. It should be noted that the relationship between asymmetry factor and surface coverage will not necessarily be a linear function. The relatively small differences in surface coverage of the packing materials studied are not sufficient to define this relationship, and further experimental data will be required using a single, standard bonded alkyl chain length with a much wider range of surface coverages. Correlation between Static SIMS Ion Peak Area Ratios and Chromatographic Retention. The normalized peak area ratios of predominant alkyl-based fragments (m/z 27, 41, and 57) and silica-based fragments (m/z 28, 44) were plotted against retention factors of the neutral, acidic, and basic solutes on the

a

a

b

b

c c

Figure 11. Correlation between the retention factors of the barbiturate (acidic) solutes and the TOF-SIMS m/z 27/44 (a), 41/44 (b), and 57/44 (c) peak area ratios. (a) 0, Allyl-1-methylbutyl (R2 ) 0.986); 2, ethylisopentyl (R2 ) 0.989); b, ethylbutyl (R2 ) 0.985); 4, ethylphenyl (R2 ) 0.982); 9, diethyl (R2 ) 0.909). (b) 0, Allyl-1methylbutyl (R2 ) 0.977); 2, ethylisopentyl (R2 ) 0.966); b, ethylbutyl (R2 ) 0.952); 4, ethylphenyl (R2 ) 0.954); 9, diethyl (R2 ) 0.808). (c) 0, Allyl-1-methylbutyl (R2 ) 9.12); 2, ethylisopentyl (R2 ) 0.909); b, ethylbutyl (R2 ) 0.909); 4, ethylphenyl (R2 ) 0.918); 9, diethyl (R2 ) 0.820).

packing materials. Good correlations were obtained for the majority of these plots, demonstrating a strong linear relationship between the static SIMS relative ion intensity ratios and the retention factor. The retention factors of the neutral biphenyl solutes showed a linear correlation with the SIMS ion peak area ratios for m/z 27/28, 41/28, 57/28, 27/44, 41/44, and 57/44 (R2 ) 0.72-0.99 for each solute, p < 0.015) (typical examples shown in Figures 10-12). There was some deviation from linearity at the higher alkyl chain lengths (for example, with m/z 27/44 and 41/44). The barbiturate retention factors increased linearly up to a chain length of C8 and demonstrated good linear correlation with all the SIMS ion peak area ratios (R2 ) 0.56-0.99, p < 0.05) (typical examples shown in Figures 10-12), apart from the m/z 57/28 correlation, which was not significant. The retention correlation of the

Figure 12. Correlation between retention factors of the pyridine (basic) solutes and the TOF-SIMS m/z 27/44 (a), 41/44 (b), and 57/ 44 (c) peak area ratios. (a) 0, 3-Ethyl (R2 ) 0.899); b, 3-bromo (R2 ) 0.947); 4, 3-chloro (R2 ) 0.941); 9, 3-amino (R2 ) 0.862); 2, pyridine (R2 ) 0.797); O, 3-acetyl (R2 ) 0.822). (b) 4, 3-Ethyl (R2 ) 0.962); 9, 3-bromo (R2 ) 0.937); b, 3-chloro (R2 ) 0.934); 2, 3-amino (R2 ) 0.902); O, pyridine (R2 ) 0.893); 0, 3-acetyl (R2 ) 0.843). (c) 0, 3-ethyl (R2 ) 0.952); b, 3-bromo (R2 ) 0.862); 4, 3-chloro (R2 ) 0.865); 9, 3-amino (R2 ) 0.910); 2, pyridine (R2 ) 0.953); O, 3-acetyl (R2 ) 0.853).

pyridine solutes was also highly significant over the entire range of m/z ratios examined (R2 ) 0.73-0.96, p < 0.014) (typical examples shown in Figures 10-12). The shape of the plots was similar to those obtained from the XPS C:Si ratios versus retention plots, indicating a good agreement between the two surface analytical techniques. CONCLUSIONS Our studies have confirmed that it is possible to obtain a correlation between surface analytical and chromatographic data generated from a range of alkyl-bonded silica particulates. XPS and static SIMS can provide quantitative data relating to both the support matrix and the surface-bonded ligands which are relevant to the chromatographic behavior of the materials. In addition, XPS alkyl:Si ratios provide a surface analytical equivalent of measuring ligand density, which correlates well with both the bulk Analytical Chemistry, Vol. 68, No. 13, July 1, 1996

2177

determination of surface density and the degree of secondary interactions of chromatographic solutes with the chromatographic surface. These studies confirm that both XPS and static SIMS are techniques that can generate data which are potentially useful in the prediction of chromatographic behavior and hence will prove effective tools to assist in the design and development of chromatographic supports and stationary phases. ACKNOWLEDGMENT This work was supported by a grant from the UK Science and Engineering Research Council as part of the Separation Processes

2178

Analytical Chemistry, Vol. 68, No. 13, July 1, 1996

Initiative. The authors thank Mr. P. Ross and Dr. H. Ritchie of Shandon HPLC for the provision of chromatography materials and are grateful to Dr. J. F. Watts of Surrey University, Guildford, UK, and Dr. A. J. Paul of CSMA Ltd., Manchester, UK, for their assistance in the XPS and TOF-SIMS studies, respectively. Received for review August 15, 1995. Accepted March 19, 1996.X AC950829+ X

Abstract published in Advance ACS Abstracts, May 1, 1996.