Introduction to Modern Liquid Chromatography, Third Edition part 91 pps

856 TROUBLESHOOTING SCH 3 H H HO F O S(O)CH 2 CH 3 S 30 343230 R 32 34 34 36 R R S S time (min) (c)(b)(a) Figure 17.22 Separation of tipredane epimers. Conditions: (a, b) 100 × 4.6-mm Hypersil ODS column; 29/32/62% acetonitrile/pH-7.2 buffer in 0/10/20 min; 1.5 mL/min; 26 ◦ C. (c) 150 × 3.9-mm Resolve C 18 column and similar, but not identical gradient conditions. (a) Injection of S-epimer; (b)injectionofR-epimer;(c) injection of S-epimer. Adapted from [31, 32]. slow, the sample may degrade while the sample transits through the column, resulting in a distorted peak—the result of two (or more) distinct molecular structures passing through the column, with the ratio of their concentrations changing during the separation [30, 31]. An example of both fast and slow sample reactions is provided in Figure 17.22 [31, 32] for gradient separations of two tipredane epimers (structure shown in the figure) under different conditions. In Figure 17.22a, the pure S-epimer was injected, and peaks for both the R- and S-epimers are observed in the chromatogram (i e, reaction of S-epimer to R). Because the two peaks are sharp and well-separated, the reaction of R to S must have occurred prior to significant elution through the column. The injection of the pure R-epimer (Fig. 17.22b) shows a similar, but reduced conversion to the alternate epimer. The two separations of (Fig. 7.22a) and (Fig. 7.22b) were each carried out on a Hypersil ODS column. When the column was changed to Resolve C 18 , the separation of an injection of the pure S-epimer in Figure 17.22c was obtained. In this case a characteristic ‘‘saddle’’ is observed between the two peaks, indicating that the sample reaction occurred more slowly—during the separation, rather than primarily during sample injection. If degradation is suspected, this can often be confirmed by changing the chromatographic conditions (temperature, pH, etc.) to speed or slow the rate of degradation. For example, increasing or decreasing column temperature will usually speed or slow the rate of sample reaction, with a predictable effect on peak shape. For the sample of Figure 17.22 it was found that a higher temperature accelerates sample reaction, while a higher mobile-phase pH slows the reaction. 17.4.6 Interpretation of System Performance Tests HPLC system performance tests were described in detail in Section 3.10.1 and summarized briefly in Section 17.2.1. Of particular importance for identifying hardware problems are the gradient performance test (Section 3.10.1.2) and the additional system tests (Section 3.10.1.3). We recommend that these tests be run every 6 to 12 months on each HPLC system to ensure that optimal equipment performance is obtained. A summary of failed performance test symptoms and 17.4 COMMON SYMPTOMS OF HPLC PROBLEMS 857 solutions is given in Table 17.11. The following discussion uses case-study examples to illustrate some of the problems that can be identified using these tests. 17.4.6.1 Interpretation of Gradient Performance Tests The gradient performance test described in Section 3.10.1.2 contained the following elements: • gradient linearity • dwell-volume determination • gradient step-test • gradient proportioning valve (GPV) test These tests often are run as a set, and the results of one test are related to the results of other tests. At other times the results of a test run suggest running another of the tests. The examples below illustrate the interrelationships of the tests, as well as the results of failed tests. The discussion is organized as a set of five case studies of problems that were highlighted as a result of the gradient performance tests. Each example is followed to completion so as to show how the various tests apply to real problems. Case 1. The gradient linearity test was described in Section 3.10.1.2. This test comprises replacing the column with a piece of capillary tubing and running a linear gradient from 100% water to 0.1% acetone-water, monitored at 265 nm. The typical result in Figure 3.26 shows a delay at the beginning, corresponding to the dwell-volume, followed by a linear transition to 100% B, with slight rounding at the ends of the gradient. Visual inspection of the plot usually is sufficient to determine linearity; the plot can also be printed and a line can be drawn next to the curve for reference, as shown in Figure 17.23 (dashed line). In this case [33] the overall plot was linear, but at about 25%, 50%, and 75%B there were slight offsets in the plot. These results suggest that the mobile-phase proportioning process generated an error at each of these points in the curve. 0 2 4 6 8 10 12 14 16 time ( min ) Figure 17.23 Plot for a linear gradient, using a system with faulty proportioning valves. Arrows show deviations from linearity; dashed line drawn below plot for reference. Gradient 0–100%B in 15 min at 1 mL/min; Solvents: A = water, B = 0.1% acetone in water; detection UV 265 nm. Adapted from [33]. 858 TROUBLESHOOTING The next step was to examine the gradient step-test. The step-test is also described in Section 3.10.1.2 and a portion of a successful test is shown in Figure 3.27. The step-test comprises a series of steps of varying mobile-phase composition 0–100% B, using the same acetone-water mobile phase described above. The recommended steps are 10%B increments between 0 and 100%B, plus two additional steps at 45 and 55%B. For the present example, the latter series of steps (Fig. 17.24a) passed the acceptance criteria of ±1% B from the target values [33]; nevertheless, deviations from linearity were obvious in the linearity test (Fig. 17.23). Next, the step-test was repeated in the vicinity of the questionable results but re-programmed for 1% steps. The results for the 45–55%B test region are shown in Figure 17.24b, where all steps look normal except the 50–51% step (arrow), where the step size is obviously too small. This same type of deviation was observed at 25–26% B and 75–76% B. The HPLC system in this case study relied on low-pressure mixing (Section 3.5.2.2), so a gradient-proportioning valve (GPV) test (Section 3.10.1.2) was the next step in problem isolation. This test relies on alternate steps of water compared to acetone-water using various combinations of the 4 solvent supply lines (lines A and B are water, C and D are acetone-water), with a normal test result appearing as in Figure 3.28. For the present case, the results of Figure 17.25 [33] were observed. The maximum acceptable deviation between the highest and lowest plateaus is 5%, whereas Figure 17.25 has a 12% deviation between the first two steps. Because the A-solvent is common between the two steps, it is a likely source of the problem. The suspected problem’s source was first a partially blocked solvent inlet frit or solvent transfer tube between the reservoir and the proportioning valve; however, a siphon test (Section 3.21) showed that there were no significant restrictions in the inlet frit 02040 time (min) ( b) 1% steps (a) 10% steps Figure 17.24 Gradient step-test results for HPLC system of Figure. 17.23. (a) Steps of 0, 10, 20, 30, 40, 45, 50, 55, 60, 70, 80, 90, and 100% B; (b) upper trace is 45–55% in 1% steps. Arrow showing ‘‘short’’ step between 50 and 51%B. Adapted from [33]. 17.4 COMMON SYMPTOMS OF HPLC PROBLEMS 859 01020 time ( min ) 50/50 A/B 90/10 A/C A/D B/C B/D Figure 17.25 Gradient proportioning valve test for HPLC system of Figure 17.23. Baseline is generated by 50:50 A:B; the remaining plateaus are 90:10 A:C, A:D, B:C, and B:D from left to right. Solvents: A = B = water; C = D = 0.1% acetone in water. Adapted from [33]. or connecting tubing. Other potential corrective actions were to degas the mobile phase (Section 3.3), sonicate the check valves (17.2.5.4), and replace the pump seals (Section 3.5.1), but these did not correct the problem. Several attempts were made to adjust the proportioning algorithm in the control software, and although improvements were observed, the problem persisted. Finally, the proportioning valve assembly was replaced, with a step-test result of 0.9% maximum deviation—well within the 5% limit [33]. A less severe, but more common gradient-linearity-test failure occurs when a linear 0–100%B gradient is programmed but appears as a segmented 0/50/100%B gradient, with a slightly different slope for the 0–50% segment rather than the 50–100% segment. Usually the controlling software is at fault, and it can be adjusted for some HPLC systems; consult the pump service manual or the manufacturer for specific recommendations. Case 2. An example of a more dramatic failure of the gradient step-test and gradient linearity test is shown in Figure 17.26. In this case the operator was unable to obtain reproducible retention times [34]. A gradient step-test (Fig. 17.26a)and a linearity test (Fig. 17.26b) were run, with obviously unacceptable results. The cause was suspected to be trapped air bubbles in the pumping system because occasional pressure fluctuations were observed. Thorough purging of the system with degassed solvent accompanied by tapping on each component of the system with a screwdriver handle (to dislodge bubbles adhering to internal surfaces) resulted in a series of bubbles in the waste stream. The two tests were rerun, with acceptable results. (See Section 17.2.5.1 for further hints on removing entrapped air.) Case 3. In the final example of a failed gradient step-test, the method worked well, with acceptable retention time, precision, accuracy, and resolution. However, when the gradient step-test was run, the results of Figure 17.27 were obtained [34]. It can be seen that a small secondary step is located between each major 860 TROUBLESHOOTING 0102030 01020 time ( min ) (a) (b) Figure 17.26 Results for an unacceptable (a) gradient step-test, and (b) gradient linearity test. Adapted from [34]. step (note the small step marked by the arrow between the 10 and 20% B major steps). In the process of eliminating likely causes, the autosampler was replaced with a manual injector (module substitution, Section 17.3.5), at which time the problem disappeared. Replacement of two stainless-steel frits within the autosampler corrected the problem. It was not clear why these blocked frits generated the secondary steps of Figure 17.27. In retrospect, it may be that the frits controlled flow through a flow-bypass channel that is used in some autosampler designs to minimize pressure pulses to the column [35–37]. In such designs part of the mobile phase flow bypasses the injection valve so that flow is not shut off when the injection valve is rotated (for additional information, see p. 238 of [38]). One of the authors has observed retention time and peak width problems when the flow through such a passage was disturbed; a disturbance in the gradient can also result from such partial blockage. Case 4. In some cases a failed gradient linearity test can reflect the inappropriate use of the HPLC instrument rather than an instrument failure per se. It was noted in Section 3.10.1.2 that rounding of the gradient occurs at its beginning and end (Fig. 3.26). This rounding is normally minor and unlikely to affect the separation. However, when the gradient-volume (= t G F) is comparable to or 17.4 COMMON SYMPTOMS OF HPLC PROBLEMS 861 0 102030405060 time (min) 10%B 20%B Figure 17.27 Unacceptable step test showing small secondary steps between each major step. Adapted from [34]. smaller than the dwell-volume, this rounding becomes more pronounced, so as to create serious distortion of the gradient. An example is shown in Figure 17.28 for two different gradients [39]. A low-pressure-mixing system (V D ≈ 1mL)was used with a 100 × 2.1-mm i.d. column at 0.2 mL/min, and the gradient reached the detector at ≈5 minutes (see dashed traces of expected gradient profiles at the column outlet in Fig. 17.28a, b). A 9.9-minute gradient (t G F = 9.9 × 0.2 ≈ 2mL) generated the expected linear profile with little apparent distortion, as seen in the solid line of Figure 17.28a. This gradient was followed by a 1-minute isocratic hold and a 0.1-minute step gradient back to the initial conditions. In this case the gradient-volume (2 mL) was significantly larger than the dwell-volume of 1.0 mL. For Figure 17.28b, a steep, 1.4-minute gradient was run, followed by a 3.5-minute isocratic hold and a 0.1-minute step back to the initial conditions. The gradient volume t G F = 1.4 × 0.2 ≈ 0.3mL,whichismuchsmallerthanthe dwell-volume; severe distortion of the gradient can be predicted in this case, as observed in Figure 17.28b. Although it may be possible to generate reproducible gradients under the conditions of Figure 17.28b on one instrument, it is unlikely that such a steep-gradient method will transfer to a second instrument without problems. Further evidence of severe distortion can be seen in both Figure 17.28a, b during the re-equilibration phase (slow return of gradient to starting %B). For more details on the effects of gradient rounding on separation, see pp. 393–396 of [18]. Case 5. The gradient dwell-volume can be determined from the same exper- iment used to check gradient linearity (Section 3.10.1.2 and Fig. 3.26). The effect of the dwell-volume on the separation is discussed in Section 9.2.2.4. Differences in dwell-volume among different gradient HPLC systems are one of the primary reasons that gradient methods are difficult to transfer from one system to another. There are two common effects that are observed when a method is run on systems with 862 TROUBLESHOOTING 0105152025 time (min) 06931215 time (min) (a) (b) %B %B Figure 17.28 Gradient distortion due to excessive gradient rounding. (a)Gradientof 20/20/50/50/20%B at times 0.0/0.1/10/11/11.1 min (little rounding); (b) 20/20/40/40/20%B at 0.0/0.1/1.5/5/5.1 min (excessive rounding). Solid lines show water/water–acetone gradient trace; dashed lines show gradient program (vertically offset for clarity). Data of [39]. different dwell-volumes, as illustrated in Figure 17.29. In this case a 10–40% B gradi- ent was run over 12 minutes at a flow rate of 1 mL/min. The system of Figure 17.29a had a dwell-volume of 1 mL, whereas the system in 17.29b had a dwell-volume of 3 mL. These differences in dwell-volume translate into the effective gradients shown by the dashed lines in Figure 17.29, with a 1- and 3-minute delay before the gradient reaches the column in Figure 17.29a and b, respectively. With later-eluting solutes for which the k-value at the start of the gradient (k 0 ) is sufficiently large, the primary effect of a dwell-volume difference is a shift in retention times equivalent to the difference in dwell-time. This is seen as the increase in retention of peaks 3 to 9 by 2 minutes [(3 mL—1 mL)/1 mL/min = 2 min] in Figure 17.29b. However, for peaks that elute early in the chromatogram, and especially for ‘‘irregular’’ solutes (Section 2.5.2.2), a change in relative peak spacing (and α) may also occur. This is illustrated by peaks 1 and 2 (note the loss in resolution for these peaks in Fig. 17.29b, despite an expected increase in resolution for early peaks and a greater dwell-volume (Section 9.2.2.3). One way of looking at this is as follows: For Figure 17.29a, early peaks migrate for 1 minute under isocratic conditions and 2 to 3 minutes under gradient conditions, whereas in Figure 17.29b, there is 3 minutes of isocratic migration plus 2 to 3 minutes of gradient elution. This change in the ratio of isocratic/gradient migration results in a difference in effective values of k (k ∗ ) and peak spacing. With later peaks the initial isocratic migration is insignificant, so only a retention-time shift is observed due to the delay of the gradient arriving at the head of the column. There are several ways to compensate for differences in dwell-volume for two HPLC systems. For the example of Figure 17.29, there are three possible solutions, assuming that the method developed on system of Figure 17.29a is transferred to system 17.29b. If the HPLC equipment is capable, programming an injection delay is the simplest solution. The gradient in Figure 17.29b would be started and the 17.4 COMMON SYMPTOMS OF HPLC PROBLEMS 863 (b) (a) Time (min) 14121086420 14121086420 Figure 17.29 Effect of varying gradient dwell-volume on retention and selectivity. (a)1-mL dwell volume system; (b) 3-mL dwell volume system. 100 × 4.6-mm column; 10–40%B gra- dient in 12 minutes at 1 mL/min. injection would be programmed to occur 2 minutes after the gradient started. This would have the effect of shifting the chromatogram 2 minutes to the right relative to the dashed gradient overlay, so the sample would see exactly the same gradient as for Figure 17.29a. Unfortunately, not all HPLC systems have the capability of making a delayed injection. A second approach is to develop a maximum dwell-volume method. This requires advanced knowledge of the dwell-volume of system 17.29b. The dwell-volume of the initial system is adjusted to equal the largest dwell volume of any system to which the method will be transferred. In the present case, system 17.29a would be adjusted by adding 2 min (2 min at 1 mL/min = 2 mL) of isocratic hold at the beginning of each run so the total effective dwell-volume would be 3 mL. When the method is transferred, this additional hold would be dropped and the resulting gradients would be identical for both systems. If several systems with different dwell-volumes are to be used, the method would be developed with a dwell-volume equivalent to the largest dwell volume and the combination of isocratic hold and true dwell-volume would then be adjusted so that the effective dwell-volume is the same in all cases. A third possibility is to adjust the initial %B for system 17.29b. This involves a combination of starting the gradient at a higher %B and changing the times of the isocratic and gradient segments. See Section 5.2.1.3 of [18] for an example of this technique. A final technique for compensating for dwell-volume differences applies when the initial system has a larger dwell-volume than the new system, for example if the present method was 864 TROUBLESHOOTING developed on Figure 17.29b and transferred to Figure 17.29a.Inthiscasean isocratic hold of 2 minutes can be added to the gradient of Figure 17.29a, giving an equivalent delay time for the start of each gradient. This approach is closely related to the maximum-dwell-volume technique discussed above. A more detailed discussion of adjusting for gradient dwell-volume differences can be found in Section 5.2.1.3 of [18]. 17.4.6.2 Interpretation of Additional System Tests Section 3.10.1.3 described several additional tests that should be run on a regular basis: • flow-rate check • pressure bleed-down test • retention reproducibility test • peak-area reproducibility test If the results of the flow-rate check exceed ±2% of the flow-rate setting, the source of the problem should be investigated. Lower than normal flow is most common. First check for leaks (Section 17.4.1, Table 17.3). Other likely causes of subnormal flow rate are bubbles in the pump, faulty check valves, and worn pump seals. Sometimes the flow rate for organic solvents will be different than that for pure water with the same pump and settings, because of the greater compressibility of organic solvents. Solvent compressibility is usually ignored by the end user, but many pumps have adjustments to compensate for solvent compressibility (see the pump operator’s manual). It is rare that the flow rate is higher than set (with the exception of extreme compressibility adjustments). First check that the settings were made correctly and repeat the test. Consult the pump operator’s manual for instructions on how to adjust the flow-rate calibration if the problem persists. For an additional discussion of flow-rate problems, see Section 17.4.3.1 and the general discussion of pump operation in Section 3.5. The pressure bleed-down test is a check of the pump’s ability to hold pressure under static conditions. If the bleed-down test (Section 3.10.1.3) results in a pressure loss of > 15% in 10 minutes, the cause should be identified and corrected. Because it is based on blocking the pump outlet tubing, the bleed-down test reflects the integrity of the pressure-limiting pump component farthest downstream (closest to the blocked outlet tubing). First check for leaks between the pump and the location of the plugged outlet tubing. If the pump uses an outlet check valve(s), this is the most likely point of failure. If the pump does not use an outlet check valve(s), a component further upstream is responsible for the problem—failure may be due to the pump seal, inlet check valve (dual-piston pumps, Section 3.5.1.1), or intermediate check valve (accumulator-piston pumps, Section 3.5.1.2). Replace any questionable pump seals and sonicate (Section 17.2.5.4) or replace questionable check valves. Retention-time reproducibility generally should be better than ±0.05 minutes (1 S.D.), but some methods may exhibit poorer reproducibility. If the questionable retention-precision values are for a method that has been run before, check the results against historic values (Section 17.2.4), and run the retention reproducibility test (Section 3.10.1.3, Table 3.5). If no leaks are present (Section 17.4.1, Table 17.3), the problem is most likely related to the pump and/or mobile-phase mixing process. Run 17.5 TROUBLESHOOTING TABLES 865 the gradient performance test (Section 3.10.1.2) and isolate the problem (Section 17.4.6.1). Peak-area reproducibility will vary according to the application, with accept- able values ranging from ≤ 1–2% (content assays) to 15–20% (trace analysis), so it is best to compare test results with the past performance of the method (Section 17.2.4). When the standard peak-area reproducibility test (Section 3.10.1.3) exceeds 0.5–1%, check for autosampler problems. Make sure that any vents or vent-needles are not blocked. Check the sample needle for partial block- age or damage. Check any seals or syringes in the sampling system. Check the needle-to-injection-valve seal. Ensure that the detector time-constant is not set too slow. See Section 3.6 and the autosampler operator’s manual for more ideas. 17.5 TROUBLESHOOTING TABLES Most of the troubleshooting tables from this chapter are gathered in this section for convenience in simultaneously referring to more than one table. Various approaches can be used to identify a problem. First, consult the outline at the beginning of the chapter for the appropriate section containing topics of interest. Alternatively, use Table 17.2 as a guide to the present section. Table 17.2 lists the major symptoms likely to be encountered and cross-references other tables in this section as well as the appropriate section of text that will be useful. The remaining tables contain more specific information to help isolate specific problems. For the most part, these are organized like an outline, with the left-hand column giving a high-level symptom (e.g., the location of a leak in Table 17.3); as you move to the right across the table, each column gives more detailed information about isolation and possible sources of the problem. The right-hand column usually contains suggested solutions and often provides additional cross-references to discussions of the problem. The tables alone may be sufficient to isolate and solve a problem, or you may need to refer to the associated text material—depending on the specific problem and your level of experience. Table 17.2 HPLC Problem Symptoms Problem Symptom More Information Leaks Table 17.3, Section 17.4.1 Pressure Table 17.4, Section 17.4.2 Retention time Tables 17.5, 17.6, Section 17.4.3 Peak area Tables 17.7, Section 17.4.4 Baseline drift Table 17.8, Section 17.4.5.1 Baseline noise Table 17.9; Section 17.4.5.2 Peak shape Table 17.10 Sections 17.4.5.3, 2.4.2 Failed system performance tests Table 17.11, Section 17.4.6 . 100% water to 0.1% acetone-water, monitored at 265 nm. The typical result in Figure 3.26 shows a delay at the beginning, corresponding to the dwell-volume, followed by a linear transition to 100%. sampling system. Check the needle -to- injection-valve seal. Ensure that the detector time-constant is not set too slow. See Section 3.6 and the autosampler operator’s manual for more ideas. 17.5. are observed in the chromatogram (i e, reaction of S-epimer to R). Because the two peaks are sharp and well-separated, the reaction of R to S must have occurred prior to significant elution through