Chromatographic Determination of Metallic ... - ACS Publications

collect metal ions from the column components and the instrument hardware. Stainless .... Ramia Al Bakain , Isabelle Rivals , Patrick Sassiat , Di...
0 downloads 0 Views 203KB Size
Anal. Chem. 1999, 71, 1885-1892

Chromatographic Determination of Metallic Impurities in Reversed-Phase HPLC Columns H. Engelhardt* and T. Lobert

Institute for Instrumental and Environmental Analysis, University of the Saarland, Im Stadtwald, D-66123 Saarbruecken, Germany

A new test procedure is proposed for the characterization of packed RP columns with respect to their metal content. It is based on the peak asymmetry of 2,2′-bipyridyl, which can be directly correlated to the metal content of the stationary phase. Since an unbuffered eluent (methanol/ water, 49:51 w/w) is used, the influence of the simultaneously present silanol groups on peak asymmetry is assessed via 4,4′-bipyridyl, which exhibits no chelating activity. A metal factor, MF, is introduced, which is the ratio of the peak asymmetries of 2,2′-bipyridyl and 4,4′bipyridyl multiplied by 100. The superiority of this test compared to others is demonstrated via various commercially available columns. It is also demonstrated that columns continuously collect metal ions from the column components and the instrument hardware. Stainless steel nets and frits are the principal source of the metallic contamination. But even in totally metal-free LC systems, the stationary phase collects metal ions from the HPLC grade eluent components. The introduction of high-purity silica for HPLC packings has been one of the most successful improvements in the recent years. It is believed that the impurities incorporated onto the silica surface are often responsible for poor chromatography. Metal ions are considered as the primary impurities in silica and reversedphase (RP) packings. They may stem from the preparation process of the silica gel, may be introduced during the synthesis of bonded phases, or may be accumulated during chromatographic operation from the mobile phase or stainless steel components of the equipment or column.1,2 The influence of metallic impurities on the chromatographic properties of stationary phases is still a topic of intensive discussion. The metal impurities are strong adsorption sites for complexing solutes. Chelating analytes can bind directly to metal ions on the silica surface, causing poor peak shape that affects resolution and impairs analytical sensitivity.3 Metal ions may also increase the acidity of adjacent silanol groups, resulting in tailing of basic compounds or even in irreversible adsorption despite end capping of the silanols.3 The presence of metal ions also influences the bonding of stationary phases, such as C18, to the silica surface. This can affect the uniformity of the surface coverage with alkyl groups, leaving (1) Iler, R. K. The chemistry of Silica; Wiley: New York, 1979. (2) Unger, K. K. Porous Silica; Elsevier: Amsterdam, 1979. (3) SGE Inc. Solutions; June 1997. 10.1021/ac981198x CCC: $18.00 Published on Web 04/02/1999

© 1999 American Chemical Society

more residual silanols and metal ions themselves available to influence analytes susceptible to their effects. A high-purity silica can minimize these effects, eliminating any visible tailing of problematic analytes.3 This was the reason for the preparation of pure silicas for the new generation of stationary phases. Especially in protein separation, the influence of metal impurities was reported to be more important than that of the silanols4 and the growing application of HPLC in biochemistry raised concerns regarding the metal content of stationary phases and has even led to the development of metal-free chromatographic instruments and columns.5 It has always been controversial whether this is a necessity or just a marketing ploy. Chromatographic grade silicas of the older types usually contain between 0.1 and 0.3 wt % metallic impurities.6 It has been reported that ∼20 different elements can be present in the ppm range and 15 additionally in the ppb range.7 The presence of metals on the silica surface is best indicated by analyzing the elution performance of strong chelates. The metal influence on the separation of hop constituents has been clearly demonstrated8 and has led to a chromatographic test for metallic impurities based on these components. Later, a test was described using 2,4pentanedione because of its better availability.9 Others used different chelates such as dihydroxyanthraquinones, Hinokitiol (4-isopropylcycloheptatrienolone), or aromatic dihydroxy compounds, where only those with 1,2- or 1,2,3-substituents form complexes.10-12 Recently, a test was described comparing the plate counts of 2,3- and 2,7-dihydroxynaphthalene (DERT test, dihydroxynaphthalene efficiency ratio test).13 Other complexing agents such as 8-hydroxyquinoline (oxine), diphenhydramine, and the copper oxinate chelate had been proposed,14-16 but because of their high basicity, they may also interact strongly with surface (4) Koyama, J.; Nomura, J.; Ohtsu, Y.; Nakata, O.; Takahashi, M. Chem Lett. 1990, 4, 687. (5) Mason, A. Z. Chromatogram 1989, (Nov), 7. (6) Nawrocki, J. Chromatographia 1991, 31, 177. (7) Verzele, M.; de Potter, M.; Ghysels, J. J. High Resolut. Chromatogr. Chromatogr. Commun. 1979, 3, 151. (8) Verzele, M.; Dewaele, C. J. Chromatogr. 1981, 217, 399. (9) Verzele, M.; Dewaele, C. Chromatographia 1984, 18, 84. (10) Ohtsu, Y.; Shiojima, Y.; Okumura, T.; Koyama, J.-K.; Nakamura, K.; Nakata, O. J. Chromatogr. 1989, 481, 147. (11) Tanaka, O.; Kimata, K.; Araki, T. J. Chromatogr. 1992, 594, 87. (12) Law, B.; Chan, P.F. J. Chromatogr. 1991, 552, 429. (13) Vespalec, R.; Neca, J. J. Chromatogr. 1983, 281, 35. (14) Euerby, M. R.; Johnsson, C. M.; Rushin, I. D.; Sakunthala, D. A. S. J. Chromatogr., A 1995, 705, 219. (15) Ohhira, M.; Ohmura, F.; Hanai, T. J. Liq. Chromatogr. 1989, 12, 1065.

Analytical Chemistry, Vol. 71, No. 9, May 1, 1999 1885

Table 1. Commercially Available Columns That Were Investigated

Figure 1. Adsorption of Fe3+ ions on bare silica (for details see Experimental Section).

Figure 2. Chromatogram of acetylacetone. Conditions: column, RP18; Fe content, 90 ppm (no surface coating); eluent, methanol/ water 80:20 (w/w).

silanols, and differentiation between silanophilic and metallophilic influences may be ambiguous. In the course of the development of test procedures for the EU project “standard reference column” the various tests described for metallic impurities in chromatographic columns were compared with respect to their applicability and reliability even for the determination of trace metallic impurities. Our aim was to develop a simple chromatographic test procedure that would be applicable to commercially available packed columns. A standard silica purified initially by a prolonged acid wash was coated with different amounts of iron as one of the main and most probable contaminants. The iron content was determined by independent standard analytical methods. Using a standard procedure, these silicas were modified with octadecylsilane, and the columns prepared were characterized with respect to their metal content by the various tests described. Because the influence of the metal content should not be masked by buffer components, the test procedures were performed in aqueous methanolic eluents. EXPERIMENTAL SECTION Chemicals. All buffer chemicals and solvents used were of HPLC grade. All chemicals were obtained from Fluka except 2,3(16) Moriyama, H.; Anegayama, M.; Komiya, K.; Kato, Y. J. Chromatogr., A 1995, 691, 81.

1886 Analytical Chemistry, Vol. 71, No. 9, May 1, 1999

and 2,7-dihydroxynaphthalene (Sigma Chemical Co., Steinheim, Germany), 3-phenylacetylacetone (Avocado Chemicals, Heysham, UK), and dimethyloctadecylchlorosilane (ABCR, Karlsruhe, Germany). Water was purified via a Millipore MilliQ (Bedford, MA) filtration device. pH Measurement. A standard pH meter with a combination glass and reference electrode was calibrated prior to use with standard solutions of pH 4 and 7. Preparation of the Stationary Phases. Silica with an average pore diameter of 10 nm, a surface area of 320 m2/g, and particle diameter of 10 µm (IMPAC Bischoff, Leonberg, Germany) was used exclusively. For the preparation of the iron-modified silicas, a method was used that is described elsewhere.17 The silanization was performed by an in-house standard procedure.18 Determination of the Iron Content. The iron content was determined by ICP-OES measurements after dissolution of the silica with a mixture of 50:50 v/v% hydrofluoric acid/nitric acid (silica purified by an acid wash was coated with Fe3+ ions of different concentrations). For the acid-washed silica, an iron content of 90 ppm was obtained. Later tests showed that iron could barely be detected in the acid-washed silica by chromatographic tests. It can be assumed that this iron is totally included in the silica matrix and is not accessible at the surface. (17) Deng, Z.; Zhang, J. Z.; Ellis, A. B.; Langer, S. H. J. Liq. Chromatogr. 1993, 16, 1083. (18) Engelhardt, H.; Dreyer, B.; Schmidt, H. Chromatographia 1982, 16, 11.

Figure 3. Influence of the mobile phase Column and sample as in Figure 2: eluent A, methanol/water 49:51 (w/w); eluent B, methanol/1 mM sodium acetate, adjusted to pH 7, 49:51 (w/w).

Figure 4. Peak asymmetry of aromatic 1,2-dihydroxy compounds on RP 18 columns with different iron contents.

Chromatography. A Hewlett-Packard 1090 HPLC system equipped with a diode array detector was used. Data acquisition and integration was controlled by a Hewlett-Packard Chem Station. Unless otherwise stated, all characterization tests had the following parameters in common: a mobile phase of methanol/ water (49:51 w/w), a flow rate of 1 mL/min, detection at 254 nm, a run time up to 30 min, and a column oven temperature of 40 °C. Commercially available columns (summarized in Table 1) were used for comparison in the optimized test procedure.

RESULTS AND DISCUSSION Preparation of Iron-Coated Silica and RP 18 Phases. Figure 1 shows the adsorption of iron from solution. A typical Langmuir-type adsorption isotherm was observed. A linear relationship between the amount of iron present in solution and that adsorbed on the surface was obtained up to a concentration of 500 ppm. An iron content of 90 ppm was measured for the initial silica. The silicas coated with iron were subsequently reacted with

a monofunctional octadecylsilane to prepare RP stationary phases. When the iron content of the silica is below 1000 ppm, the amount of carbon bound is independent of its iron content. Typically, 18.7 ( 0.5% carbon could be bound. With an iron content of 1152 ppm the amount was reduced to 17.5%, and with a silica containing 1407 ppm iron coated on the surface only 14.5% carbon was bound under otherwise identical reaction conditions. The test for silanophilic properties19,20 showed no significant differences with these stationary phases. p-Ethylaniline was eluted with identical asymmetry factors from all these stationary phases. The stationary phases thus prepared with different iron contents were tested by the different test procedures described in the literature. The tests should reveal the influence of a minor iron content. All the stationary phases with an iron content of less than 700 ppm showed identical hydrophobic properties, the differences in solute retention being attributable solely to the influence of metal ions. (19) Engelhardt, H.; Arangio, M. GIT Spez. Chromatogr. 1996, 2, 54. (20) Engelhardt, H.; Arangio, M.; Lobert, T. LC-GC 1997, 15/9, 856.

Analytical Chemistry, Vol. 71, No. 9, May 1, 1999

1887

Figure 5. Influence of the iron content of RP 18 phases on the retention, peak asymmetry, and peak area of 2,2′-bipyridyl.

Figure 6. Standardization of the test procedure with 2,2′-bipyridyl and 4,4′-bipyridyl columns: RP 18, different iron contents; mobile phase, methanol/water 80:20 (w/w).

Test with 1,3-Diketones (β-Diketones). Acetylacetone (2,4pentanedione) was the first test solute proposed by Verzele for the characterization of RP columns for trace metal content.9 The peak shape and the differences in peak areas should be a measure of metal contamination. Surprisingly, the chromatograms obtained with this compound were independent of the iron content of the stationary phase. A typical chromatogram is shown in Figure 2. The ratio of the peak areas is not a function of the iron content of the stationary phase, but depends on the presence of the ketoenol tautomers in the solution. As is customary in HPLC, the sample solution should always be prepared in the eluent mixture. Figure 3 compares the chromatograms obtained with the same column. With buffered eluents, the keto-enol equilibrium is attained very rapidly, so that only a single peak is observed. The ratio of the two peaks and their shape in the unbuffered solution depend more or less on the keto-enol interconversion in the sample solution and during chromatography and cannot be correlated to the metal content of the stationary phases. 1888 Analytical Chemistry, Vol. 71, No. 9, May 1, 1999

Besides 2,4-pentanedione, other diketones such as 3-phenylacetylacetone, benzoylacetone, and dibenzoylmethane were included in this study. For all these solutes, peak shape and elution behavior could always be attributed to the keto-enol tautomerism in the sample solution. Consequently, the chromatographic results depended on several parameters including solvent composition, temperature, time of sample storage, pH value, etc. The metal content of the stationary phase may also influence the chromatogram, but its influence could not be elucidated reproducibly. Test with Aromatic 1,2-Dihydroxy Compounds. Another test recommended for the determination of the surface metal content of stationary phases is the elution behavior of aromatic di- or trihydroxy compounds.10-12 Only solutes with vicinial hydroxyl groups can complex metal ions and should, therefore, be eluted with a larger peak asymmetry than the isomers with the hydroxyl groups in other positions. Vicinial hydroxy derivatives of benzene and naphthalene have been used with our ironmodified RP columns. As can be seen in Figure 4, 1,2-dihydroxy-

Table 2. Optimized Test Conditions compounds 4,4′-bipyridyl 2,2′-bipyridyl mobile phase methanol/water 49:51 (w/w) detection UV 254 nm injection vol 10 µL

concn 0.25 g/L 1.00 g/L flow 1.0 mL/min temp 40 °C run duration up to 30 min

Figure 7. Differentiation between silanophilic and metallophilic activity on RP 18 columns (peak 1, 4,4′-bipyridyl; peak 2, 2,2′bipyridyl): column A, low silanophilic activity, low metal content; column B, high silanophilic activity, high metal content; column C, low silanophilic activity, high metal content.

benzene (catechol) gives only asymmetrical peaks on stationary phases with very high iron content. 2,3-Dihydroxynaphthalene was the only test solute that showed a linear correlation between the iron content of the RP column and peak asymmetry. The efficiency of this test procedure is, however, inadequate because peak asymmetry increased only by a factor of 1.4 when the iron content increased by a factor of 5. As stated above, the surface concentration of the C18 groups and, hence, the silanophilic properties of these stationary phases are identical. As will be discussed later, this test procedure failed when commercially available stationary phases with different bonded group density were compared (see Figure 10). Test with 2,2′-Bipyridyl. The ability of 2,2′-bipyridyl to complex metal ions having charges higher than 2, especially with iron, is well-known in inorganic chemistry.21 Consequently, its retention behavior must also depend on the metal content of a stationary phase. As can be seen in Figure 5, several chromatographic parameters depend on the metal content. There is an almost linear dependence of the peak asymmetry of 2,2′-bipyridyl on the iron content of the stationary phase. Also the retention of the tracer (k value) increases with increasing iron content. The peak area decreases, demonstrating the increasing amount of irreversible adsorption of the solute. It should be mentioned that this test was performed in an unbuffered eluent. Since 2,2′-bipyridyl has two basic nitrogen atoms and a pKa value of 4.6, it may also interact with surface silanols resulting in a similar observation of peak tailing. To be able to differentiate (21) Cotton, F. A.; Wilkinson, G. Advanced Inorganic Chemistry, 5th ed.; Wiley: New York, 1988.

Figure 8. Influence of the buffer on the metal on an RP 18 phase with a high metal content: (A) unbuffered mobile phase; (B) buffered with 1 mM phosphate buffer pH 7.

between metallophilic and silanophilic interactions, 4,4′-bipyridyl was used as a standard. This solute has approximately the same pK-value and similar basic nitrogen atoms but does not exhibit any chelating and metal complexing activity.22 As can be seen in Figure 6, neither the k value nor the peak asymmetry of 4,4′-bipyridyl is affected by the iron content of the stationary phase, whereas 2,2′-bipyridyl exhibits a strong dependence, especially of peak asymmetry, on the iron content. 4,4′-Bipyridyl interacts as a base with the surface silanols in this unbuffered eluent mixture. The presence of unshielded surface silanols causes an asymmetric peak form for both 2,2′bipyridyl and 4,4′-bipyridyl. A ratio of >1 of the peak asymmetry of both isomers is an indication of additional metal activity of the stationary phase. Symmetrical peaks for 4,4′-bipyridyl but asymmetrical ones for 2,2′-bipyridyl can be expected for stationary phases with low silanophilic activity, but with additional metal impurities concentrated on the surface. The reliability of this test procedure for measuring metal activity independent of silanophilic activity is demonstrated in Figure 7 for three different commercially available stationary phases. In panel A, the elution behavior of both bipyridyls is (22) Lobert, T.; Engelhardt, H. Poster presented at the Analytica 1996, Munich.

Analytical Chemistry, Vol. 71, No. 9, May 1, 1999

1889

Figure 9. Correlation between metal content and the asymmetry of 2,2′-bipyridyl and tert-butylcatechol (see Table 2).

satisfactory, as is expected for an excellent column with low silanophilic and metalophilic activities. In the panel B, the elution behavior of the two test solutes on a column having high silanophilic activity and a high metal content is shown. 4,4′Bipyridyl shows serious tailing due to interaction with the surface silanols. The peak of 2,2′-bipyridyl is barely detectable due to irreversible adsorption. The elution behavior on a column with low silanophilic activity but a high metal content is shown in panel C. It should be mentioned that the sample injected was identical in all three cases. The concentrations of both solutes in the text mixture were adjusted so that 2,2′-bipyridyl is eluted as an measurable peak, even from columns with a high metal content. Because irreversible adsorption of 4,4′-bipyridyl is negligible and because of its higher molar absorptivity, its concentration in the test mixture is only one-fourth of that of 2,2′-bipyridyl. Additionally, at this ratio, both peaks are eluted with approximately the same peak heights on columns with a low metal content in order to obtain direct visual information about the metal contamination of the column. For the chromatograms shown in Figure 7, 2.5 µg of 4,4′-bipyridyl but 10 µg of 2,2′-bipyridyl was injected. The amount injected, especially for 2,2′-bipyridyl has an influence on peak shape and k value when columns with a high metal content have to be compared. The amount of sample injected has been kept constant through all these comparison studies. The optimized test conditions are summarized in Table 2. The test mixture proved to be stable over 36 months when stored under nitrogen. For the metal activity test, the same eluent composition was used as for the standard RP column test.19,20 An unbuffered eluent is used to enable differences in stationary phases with respect to their silanophilic activity to be revealed. There have been discussions on the use of buffered or unbuffered eluents in column characterization. Every buffer or salt added to the eluent changes the silanophilic properties and improves column behavior. Because the bipyridyls are also basic solutes, their elution behavior is affected as well when buffered instead of unbuffered eluents are 1890 Analytical Chemistry, Vol. 71, No. 9, May 1, 1999

Table 3. Metal Content of RP 18 Phases Determined by ICP-AES metal content (ppm) RP columns: SymC18 Kro100C18 IntODS3 NucC18 HypODS ParODS1 LisphRP18

Al