846 TROUBLESHOOTING pulsations (Section 17.4.2.3) or leaks (Section 17.4.1). In the case of Figure 17.10 isolation of the HPLC system from the electrical circuit, as well as pump and detector troubleshooting, could not solve the problem. Rather, movement of the system away from its original location reduced the frequency of the noise, and with sufficient distance eliminated it. Although the source of noise was never definitively identified, the noise was attributed to interference from an electronic air filter [21]. Artifact or ‘‘ghost’’ peaks in blank gradients represent a special type of long-term noise. An example of this is shown in Figure 17.11a for a gradient of 5–80% ACN-phosphate buffer (pH-7) in 15 minutes, with a hold at 80%B [7]. Ideally the baseline should be free of peaks in this blank gradient (the baseline drift is caused by differences in absorbance of the A- and B-solvents as discussed above). The most likely source of peaks in blank gradients is contamination of the A-solvent, since these contaminants tend to concentrate at the head of the column during equilibration between runs—followed by their elution during the gradient. A simple way to confirm A-solvent contamination is to increase the equilibration time between runs (flowing A-solvent). If the contaminants arise from the A-solvent, all the peaks should increase roughly in proportion to the increased equilibration time (i.e., a larger volume of A-solvent, with an increase of collected contaminants). In the present example, the 10-minute equilibration of Figure 17.11a was extended to 30 minutes (Fig. 17.11b) and the gradient was repeated. It can be seen that the peaks are each about three times larger, so contamination of the A-solvent is confirmed. Further isolation of the problem identified the pH-meter probe as the source of contamination in this example [7]. Figure 17.12 compares results for a blank run made with buffer prepared by dipping the pH probe in the buffer to adjust the pH (Fig. 17.12a) with results from the use of buffer made without contact with the pH probe (Fig. 17.12b). Additional examples of gradient ghost peaks originating 048121620 time (min) 0 5 10 15 20 absorbance (mAU) (a) (b) Figure 17.11 Effect of impurities in the A-solvent on a gradient chromatogram: blank gradi- ent runs after (a) 10-minute and (b) 30-minute equilibration. C 18 column; gradient 5–80% ACN–10-mM phosphate buffer (pH-7) in 15-minute plus 5-minute hold at 80%; UV detec- tion at 215 nm. Adapted from [7]. 17.4 COMMON SYMPTOMS OF HPLC PROBLEMS 847 4 8 12 16 0 5 10 15 20 absorbance (mAU) time ( min ) (a) (b) Figure 17.12 Comparison of (a) contaminated buffer (same conditions as Fig. 17.12a), and (b) buffer prepared with extra-clean glassware and no exposure to the pH probe. All other conditions as in Figure 17.12. Adapted from [7]. from water and other reagents can be found in Section 5.5.4 of [18]. An excellent discussion of artifact peaks in gradients also can be found in [6]. 17.4.5.3 Peak Shape Problems The ideal chromatographic peak is a symmetrical, Gaussian curve. Deviations from an ideal peak shape can be quantified by the peak tailing factor or peak asymmetry (Section 2.4.2), as illustrated in Figure 2.16a. Peak shape problems may or may not be accompanied by abnormal retention times. Deviations from symmetry can be classified as: • tailing peaks • fronting peaks • broad peaks • split or distorted peaks These irregular peak shapes are discussed below, with a corresponding summary of symptoms, causes, and solutions in Table 17.10. Section 2.4.2 also contains a detailed discussion of some of the causes of peak shape irregularities. Peak tailing is the most common peak-shape problem. New-column specifica- tions often allow peaks with tailing factors TF ≤ 1.2, so a small amount of peak tailing should be considered as normal. Although regulatory agency guidelines [22] allow TF ≤ 2 for pharmaceutical methods, peaks with TF< 1.5 are preferred. Peak tailing tends to increase over time, due to deterioration of the column; when TF > 2 is observed, action should be taken to reduce peak tailing (e.g., replace the column). If all peaks are severely tailing, see the discussion below for split or distorted peaks. If the early peaks tail more than later peaks, extra-column peak broadening may be the source of the problem. This is illustrated in Figure 17.13, where the tailing factor ranges from TF ≈ 2.5 for the first peak to TF ≈ 1.2 for the last peak. If 848 TROUBLESHOOTING 024 Time (min) 68 Figure 17.13 Peak tailing from extra-column effects. TF ≈ 2.5 for first peak, TF ≈ 1.5forlast peak. extra-column peak broadening is suspected, reduce extra-column volume, such as by using shorter lengths of smaller i.d. tubing and ensuring that all connections are made properly (Sections 2.4.1, 3.9). In the past a common cause of peak tailing was the strong interaction of ionized basic compounds BH + with ionized silanols–SiO − on the column (Section 7.3.4.2): BH + + SiO − K + ⇔ BH + SiO − + K + However, this problem is becoming less frequent with the use of less-acidic columns made from type-B silica (Section 5.2.2.2). Apart from a change to a type-B column, peak tailing of this kind can also be reduced by changes in the mobile phase. Silanol ionization and tailing decrease as the pH is lowered, while solute ionization and tailing decrease for a pH pK a for the solute. Increased ion-pairing of the solute BH + or addition of the competing ion triethylamine (in the protonated form) to the mobile phase can be effective in reducing peak tailing. An increase in ionic strength can also reduce peak tailing, although this is usually less effective than are other changes in the mobile phase. However, these changes in column or mobile phase need to be addressed during method development; otherwise, the method may require re-validation. An example of the effect of a change in the mobile phase on peak tailing is shown in Figure 17.14 for the analysis of 4 proteins by gradient elution with a TFA/ACN mobile phase on 3 columns of different purity silica (column A, high purity [i.e., type-B]; B, intermediate purity; C, low purity [type-A]). When 0.1% TFA is used in the mobile phase, there is little difference between the chromatograms observed on the three columns (Fig. 17.14a–c). However, when the TFA concentration is reduced to 0.01% TFA, the proteins show strong interaction with the lower purity columns, as exhibited by greater peak tailing (Fig. 17.14d–e), and longer retention times for the low-purity column (f ). The effect of insufficient TFA becomes more pronounced as the column is changed from high-purity (d) to intermediate-purity (e) to low-purity (f) silica, because of greater silanol ionization. TFA is used as an ion-pairing reagent for proteins (Sections 7.4, 13.4.1.2), so the advantage of higher TFA concentration may be due to increased ion-pairing, although a corresponding decrease of pH and increase of ionic strength may also contribute to better peak shape. 17.4 COMMON SYMPTOMS OF HPLC PROBLEMS 849 Figure 17.14 Influence of trifluoroactic acid concentration on peak tailing for columns of varying silica purity. (a, d) High-purity silica (column A); (b, e) intermediate purity silica (col- umn B); (c, f) low-purity silica (column C). (a-c) 0.1% Trifluoroacetic acid (TFA); (d-f) 0.01% TFA. C 18 , 30-nm pore silica particles, 5–70% ACN/TFA gradients. Sample: A, ribonuclease A; B, cytochrome C; C, holo-transferin, D, apomyoglobin. Data courtesy of Advanced Chro- matography Technologies (ACT). Injection of too large a mass of sample can result in mass overload of the column. Mass overload can occur for one or more peaks in the chromatogram, and peak tailing then takes on a right-triangle appearance with a concurrent reduction in retention time as the mass on column is increased, as seen in Figure 17.15a (right to left, 0.01 to 5 μg [23]). Mass overload is confirmed if dilution of the sample or injection of a smaller sample volume gives a longer retention time and a reduction in peak tailing. Fully ionized compounds exhibit mass overload for sample weights about 50-fold smaller than for other compounds (Section 15.3.2.1); for a 4.6-mm-i.d. column, mass overload and peak tailing occur for 1 > μg of an ionized solute, as opposed to about 50 μg of a neutral solute. Peak fronting is much less common in RPC than is peak tailing. As with peak tailing, a small amount of peak fronting can be tolerated; many column manufacturers’ specifications allow for some peak fronting: TF ≥ 0.9. Peak fronting has been attributed to temperature problems with ion pairing [24, 25], but these reports are for older type-A, low-purity silica columns and do not seem to be prevalent with type-B, higher purity columns. Usually fronting is attributed to a void in the column (column collapse), and the result can be quite dramatic, as shown in Figure 17.16b [26]. In this case the C 18 column was operated at pH-9, above its recommended operating pH. Sample analysis proceeded with normal peak shape (Fig. 17.16a)for≈500 injections, then suddenly, from one injection to the next, 850 TROUBLESHOOTING (a)(b) (d )(c) 4.41 min 4.73 min 8.66 min 9.05 min Figure 17.15 Examples of peak distortion. (a) Right-triangle peak shape as a result of a too-large sample weight; right to left 0.01–5 μg nortryptiline; 150 × 4.6-mm i.d. (3.5-μm) XTerra MS C18 column; 28% ACN pH-2.7 (20-mM formic acid); adapted from [23]. (b) Flat-topped peak characteristic of either injection volume overload or detector overload. (c) Distortion due to injection of too large a volume of too strong an injection solvent (30 μLof 100% ACN injected in a 18% ACN mobile phase); (d) same conditions as (c), except mobile phase used as injection solvent. Conditions for (c, d): 250 × 4-mm Lichrosorb RP-18, 18:81:1 ACN-water-acetic acid mobile phase; adapted from [29]. the peak began to front (Fig. 17.16b). No change in peak area was observed, so data collection for the portion of the run with the fronting peak could be used for quantitative purposes. Column flushing and other restorative measures could not regenerate the column, so the column was discarded. Such a pattern of failure is typical for this method and is attributed to a collapse of the column bed due to dissolution of the silica at high pH. For this method, it was deemed more prudent to replace the column each time it failed rather than go to the time and expense to redevelop and revalidate the method for a more stable column. Other potential causes of fronting include limited analyte solubility in the mobile phase, a tendency of analyte molecules to aggregate, and conformational changes in the analyte molecules. Broad peaks may be a precursor of split or distorted peaks, so the discussion of these peak types (later in this section) should be consulted if the suggestions below do not solve the problem. Broad peaks have a calculated plate number N that is significantly less than the specified value for a new column. Because columns are initially tested by the manufacturer with ‘‘ideal’’ samples and conditions, the resulting plate number N may be significantly larger than for ‘‘real’’ samples and conditions. It is best to measure the plate number (or peak width for gradient elution) on a new column and one or more sample solutes for each method; this 17.4 COMMON SYMPTOMS OF HPLC PROBLEMS 851 (a) (b) 0 2 4 Time ( min ) Figure 17.16 Peak fronting due to column collapse. (a) Normal chromatogram; (b)peak fronting after ≈500 samples run at pH-9 with a C 18 column. Adapted from [26]. can serve as a reference value for the method. Acceptable peak broadening will vary from one application to the next, and a reference value should be determined as part of the system suitability test. As a rule of thumb, if N for a routine assay with a new column is less than about 75% of N under the manufacturer’s test conditions, the cause for this lower value of N should be investigated. As the column ages, a reduction in plate number will occur, so it is normal for peaks to become increasingly broad over the lifetime of the column. If peak broadening is observed after 500 or more injections, this is the likely cause. Flushing the column with strong solvent (e.g., 20–30 mL ACN or MeOH) may improve peak shape, but column replacement is often the most expedient corrective action. If a chromatogram exhibits a single broad peak and the surrounding peaks are narrower, the single broad peak is likely a late-eluting peak from the prior injection. Late elution can be confirmed by extending the run to allow sufficient time for the peak to elute in the correct run (Section 17.4.4.1, Fig. 17.5). Extra-column effects may generate tailing peaks as in Figure 17.13, with some peak broadening. Reduce extra-column effects by using shorter lengths of narrower diameter tubing to connect the column to the autosampler and detector. Signal processing problems also can result in broad peaks. If a detector’s time constant (noise filter) is used, it should be no larger than 1/10 the width of the narrowest peak of interest. For example, a 1-second time constant is suitable for a 10-second wide peak (measured at the baseline), but 5-second time constant may broaden the peak excessively. In a similar manner the data rate of the data system should be sufficiently fast to collect a minimum of 20 data points across the peak (Section 11.2.1.1). With some detectors, such as MS, peak smoothing can be used 852 TROUBLESHOOTING to improve the appearance of the peak. However, excessive smoothing can broaden peaks. Too large an injection volume can result in broader peaks that may even appear to be flat-topped—as in Figure 17.15b. For injection in mobile phase as the injection solvent, the injection volume should be no more than ≈15% of the volume of the first peak of interest (Sections 2.6.1, 3.6.3, 15.3.2.2, Table 3.3) for a 5% increase in peak width (5% loss in resolution). If sample solvents are used that are more than ≈10% weaker than the mobile phase, larger injection volumes may be possible (Sections 2.6.1, 3.6.3). Flat-topped peaks are also characteristic of detector overload. At low concentrations, the detector response will increase in the normal manner for an increase in analyte concentration (e.g., the front and back of a peak), but when the detector is overloaded, no increase in response is seen for an increase in concentration—the peak appears with a flat top. Split or distorted peaks can appear for just one peak, several, or all the peaks in a chromatogram. Examples of split peaks throughout the chromatogram are shown in Figure 17.17. Split or distorted peaks for all peaks in the chromatogram is a classic symptom of a blocked frit or (less commonly) a column void, and often this is accompanied by an increase in pressure. The effects of a blocked frit or column void are illustrated in Figure 17.18. Figure 17.18a represents the inlet of a normal column with a frit in place; when the sample is injected, all portions of the sample stream (arrows) arrive at the top of the column at the same time. The chromatographic separation thus starts for the entire front edge of the sample at the same time. When the frit is partially blocked (Fig. 17.18b), the sample stream is distorted such that some portions of the sample reach the head of the column late—with peak tailing. A void at the head of the column (Fig. 17.18c) may also cause peak distortion, but usually a strongly fronting peak is the result (as in Figure 17.16b). Because these distortions happens before any chromatographic separation has taken place, each peak is distorted in the same way as the peaks migrate through the column. Reversal of the column is the best way to flush particulate matter from the top of the inlet frit and restore normal peak shapes. After reversing the column, flow 20 to 30 mL of solvent through the column to waste (not the detector), then reconnect the column and leave it in the reverse-flow direction (the problem may recur if the column is returned to its original direction). For example, the chromatogram of Figure 17.19a was observed with all peaks doubled [27]. Injection of a reference Figure 17.17 Examples of similar peak distortion for all peaks in the chromatogram, attributed in each case to a partially blocked column-inlet frit. 17.4 COMMON SYMPTOMS OF HPLC PROBLEMS 853 (a)(b) (c) Figure 17.18 Peak distortion at the column inlet. Arrows represent flow streams for sample components as they enter the column. (a) Normal flow of sample through column-inlet frit during injection. (b) Distortion of sample stream due to partially blocked column-inlet frit. (c) Distortion of sample stream due to void at head of column. 042681210 Time (min) (a) (b) Figure 17.19 Split peaks attributed to a partially blocked column-inlet frit. (a) Splitting observed for all peaks in a normal sample; (b) splitting for reference standard. Adapted from [27]. 854 TROUBLESHOOTING standard also resulted in a split peak (Fig. 17.19b), but reversal of the column corrected the problem. Permanent column-reversal is not recommended for some columns. For example, some 3-μm particle columns use a 0.5-μm porosity frit on the outlet end and a 2-μm frit on the inlet end. The 2 μm is not fine enough to hold the 3-μm particles in the column, but a brief back-flushing usually will cause no harm if the column is returned to the normal flow direction. If in doubt, consult the column care-and-use instructions for flow-direction limitations. In the past it was common to replace the column inlet frit when blockage was suspected, but today’s columns may be permanently damaged when the column end-fitting is removed, so this technique is no longer recommended. Use of a 0.5-μm porosity in-line filter (Section 3.4.2.3), sample filtration or centrifugation, and/or better sample cleanup may prevent the recurrence of blocked column frits. Inadequate control of the column temperature can cause distorted peaks, especially for peaks later in the chromatogram. In the example of Figure 17.20a,the mobile phase is preheated to the column temperature (56 ◦ C), and acceptable peak shapes are observed. In Figure 17.20b the temperature of the mobile phase entering the column was 39 ◦ C (i.e., 17 ◦ C lower than the column temperature because of inadequate pre-heating of the mobile phase). As a result peaks are broader than those in Figure 17.20a, and later peaks are visibly distorted. Peak broadening and distortion in Figure 17.20b are caused by a radial temperature-gradient along and across the column, as illustrated in Figure 17.21. In Figure 17.21a, the incoming mobile phase and the column are in thermal equilibrium, so the mobile-phase temperature is the same throughout the column, and the sample band (shown as dots) travels at a uniform velocity down the column. When cooler mobile phase is introduced into the column (Fig. 17.21b), the mobile phase at the center of the column is cooler (and 0 (a) (b) 2 A B C D E F G A B C D E F G 4 6 8 10 12 14 16 02468 Time ( min ) Time (min) 10 12 14 16 Figure 17.20 Effect of temperature mismatch of incoming mobile-phase and column. (a) Inlet solvent 56 ◦ C; oven 56 ◦ C (matched temperatures). (b) Inlet solvent 39 ◦ C; oven 56 ◦ C (unmatched temperatures). Sample: A, uracil; B, nitroethane; C, phthalic acid; D, 3,5-dimethylaniline; E, 4-chloroaniline; F, 3-cyanobenzoic acid; and G, 1-nitrobutane. 150 × 4.6-mm (5 μm) Zorbax SB-C 18 column; 90/10 50 mM potassium phosphate (pH-2.6)/ACN at 2.0 mL/min. Adapted from [28]. 17.4 COMMON SYMPTOMS OF HPLC PROBLEMS 855 56ºC 56ºC 56ºC 56ºC 56ºC 56ºC 56ºC 39ºC 45ºC 50ºC 56ºC (a) (b) Figure 17.21 Band broadening due to thermal effects. (a) Ideal case, no thermal effects; (b) effect of incoming mobile phase that is at a lower temperature than the column. Adapted from [28]. more viscous) than that at the walls. Then molecules of the sample near the center flow more slowly through the column than at the column wall, so the sample band widens and can become distorted. Because the average temperature in the column is lower when cooler mobile phase is introduced, retention times also increase—with some changes in selectivity (note that peaks D and E appear in reversed order in Fig. 17.21a, b). The solution to the problem of Figure 17.20 is better control of mobile-phase temperature, specifically by pre-heating the incoming mobile phase (Section 3.7.1). When the temperature of the incoming mobile phase is within ±6 ◦ C of the column, peak distortion is not likely to occur [28]. Three causes of misshapen peaks are related to sample injection: distortion from too strong a sample solvent, broadening from too large an injection volume (see discussion under broad peaks), and tailing from too large an injection mass (see the discussion above of mass overload). If the sample solvent used for the injection is too strong, the solvent will not be diluted quickly enough by the mobile phase, so part of the sample may travel too quickly through the column before it is fully diluted. This can cause peak distortion, especially of early peaks in the chromatogram, as illustrated in Figure 17.15c. Use of a sample solvent that is too strong can reduce retention time. In the chromatogram of Figure 17.15c [29], a 30-μLofsample diluted in 100% ACN was injected with a mobile phase of 18% ACN—which is much too strong an injection solvent for this injection volume. The first peak is badly distorted, and the second peak is broadened. The retention times are also shorter than normal for both peaks. This problem can be corrected either by (1) diluting the sample so that the solvent is no stronger than the mobile phase, and/or (2) reducing the injection volume to ≤10 μL (for a 150 × 4.6-mm column; smaller volumes for smaller volume columns). When the sample of Figure 17.15c was diluted in mobile phase, a 30-μL injection gave normal peak shape and retention times (Fig. 17.15d) [29]. Sample solvent effects are discussed further in Sections 2.6 and 3.6.3.2. Another cause of distorted peaks is degradation or chemical change of an analyte as it passes through the column, when a compound is not stable under the chromatographic conditions. If degradation takes place rapidly at the head of the column, possibly catalyzed by the metal frit of the column, the decomposed sample is chromatographed without further change in sample composition, yielding original and reacted peaks of normal appearance. However, if the rate of decomposition is . way to flush particulate matter from the top of the inlet frit and restore normal peak shapes. After reversing the column, flow 20 to 30 mL of solvent through the column to waste (not the detector),. column, peak distortion is not likely to occur [28]. Three causes of misshapen peaks are related to sample injection: distortion from too strong a sample solvent, broadening from too large an injection. peaks in a chromatogram. Examples of split peaks throughout the chromatogram are shown in Figure 17.17. Split or distorted peaks for all peaks in the chromatogram is a classic symptom of a blocked