992 LOGE VOLUME (6 NUMBER 11 NOVEMBER 1908 LE Troubleshooting A Case Study ‘Michael D. Nelson and John W. Dolan Gradient Background Peaks — When peaks appear in a blank gradient, isolating the problem source can be a challenge. ne problem chromatographers com: ‘monly encounter when performing gradient liquid chromatography (LC) ‘uns is spurious peaks in blank base- ines. Because these peaks occur ‘when no sample is injected, the peaks originate from the solvents or system components, Recently we encountered excessively noisy baselines in one of our methods. The process we used to isolate the problem souzce serves asa good case study in ‘isolating gradient contaminants ‘THE METHOD We performed the separation on a 150 mm 446 mm, 5-yum d, Inensil ODS-3 column (MetaChem Technologies, Torrance, Califor. ni) operated ata temperature of 30 °C and a flow rate of 1.5 mL/min, The LC system comprised two Shimadzu LC-10AD pumps (high-pressure mixing), a model SIL-10A, autosampler, and a model SPD-10A variable- ‘wavelength detector operated ata detection ‘wavelength of 215 nm (all components from Shimadzu Scientific Instrumeets, Columbia, Maryland). Gradients of 0-1 00% B were run lover 15 min, with a 5 min bold at 100%: « 10-min equilibration period was used berween runs unless otherwise noted, For uns using buffer, the A solvent comprised 5:95 (viv) acetonile-buffer and B was 80:20 (vis) Aacetonivle-buffer. A pH17.0 10 mM phos- ate buffer was prepared by blending 10 mM ‘monobasic potassium phosphate and 10 mM basic potassium phosphate. In some exper ‘ments, water was substituted for buffer, as noted, We used HPLC- grade solvents for all experimonts; HPLC grade water was, ‘generated in-house using a Mill-C water- purification system (Millipore, Bedford, Massachusets atacoresponss (nV) Time in) FGURE 1: ‘equllbcation (uaper. See text for otver contons. Exvaneous peaks observed ina blank gradent. 10min equlioration (lawer) vs. 30min ‘THE PROBLEM Figure | llustates the problem we encoun- tered, The lower trace shows a blank gradient (0 injection) using the standard conditions sated above. Background peaks as large as TmV (1 x 10- absorbance units [AUD are allowable in HPLC-grade water. Te lower chromatogram of Figure | generates numer- ‘us peaks approximately twice this size. ‘These peaks commonly originate from the A solvent through on-column concentration, {in which nonpolar materials from the A sol- vont accumulate atthe head ofthe column using equilibration and then are eluted as ‘aks by the gradient. One easy check fo this problem isto extend the equilibration time be- tween runs — the extended equilibration time increases the amount of material that collects atthe head of the column and yields larger peaks inthe gradient The upper trace of Fi tre shows the result ofa 30 min equiibra- tion between runs. The peaks roughly tiple in size, confirming that the peaks originate inthe Asolvent, We rarely have problems with our HPLC-grade wate, 0 we suspected thatthe phosphate buffer was te problem source. This problem also gave us an excuse to compare the backgrounds generated by different brands of utter. PHOSPHATES COMPARED We purchased mono- and ditasic phosphate salts from Four vendors and prepared buffers v0 186 GE VELL 6 NBER (f NOWEMBER 1598 as described above. To facilitate comparison ‘of background peaks, a 30-min equilibration period was used. Figure 2 shows the four blank chromatograms, We were surprised to sce very similar chromatograms for al four ‘brands, Each product generated unacceptably large background peaks. Asa check, we used {ster blank (bottom trace) in which phos- phate was replaced by purified water. Several ofthe larger peaks persisted in the water, for ‘example the peaks near 12,13, 16, and 17 ‘in, The fact that all the buffers had similar Appearances and thatthe peaks were much lagee than expected led us to suspect that the ‘extra peaks originated from inadvertent conta- ‘mination during buffer preparation. ‘SUSPECTED SOURCES Several huffer preparation steps have the potential for contaminating the solvents ‘These possibilities included glassware contact, :nerofilation, pH adjustment, and degassing, ‘We chose to simplify the system and use water instead of buffer forthe isolation steps. Dur- ‘ng normal buffer preparation, the buffer was TTA 6 a first guess, analysts should assume that a single source is responsible for any given LC prob- lem. exposed to six separate pieces of glassware, so laxsware was tested by switling purified in six 600-ml. beakers. Filtered water was prepared by passing it throvgh 3 0.45-um ‘membrane filler three times. We checked the ‘pH-adjustment step by placing a str bar in 2 beaker of purified water and stirring it for 10 ‘min with the pH meter probe immersed inthe water, Degassing by helium sparging was used. only forthe degassing test. Fr each test, we voided exposure tothe ther sourees if poss: ble «for example, only one batch of solvent was degassed, but all solvents required atleast 4 minimum exposure (0 glassware), Figure 3 shows chromatograms for the var- us water weatments (30-min equilibration) ‘The degassed and filtered water are very simi= far and have the fewest contaminant peak. Dirty glassware appears tobe the source of ‘most ofthe smaller peaks. We were surprised fo see two lange peaks at approximately 13, ‘and 14 min that were present only when the ‘solvent was exposed tothe pH meter probe, DIRTY GLASSWARE [Because the test above suggested that some of the contaminants originated from the glass- ‘ware, we thoroughly investigated the dish- ‘washing procedure. All glassware is washed bby hand using a commercial laboratory dish ‘soap, rinsed sx times with tap water, then six times with deionized water. Most glassware is baked dry ina laboratory oven, We first sus= to 5 0 Time FIGURE 2: Blank gracints from four phosphate bufers (upper traces). Lowest chromatogram uses putfed water insta of buffer, Al runs included 3 30min equltretion. x0 Ipottom). See tat for other conditions to 5 2 Tine iy FIGURE 3: Chromatograms resulting fom use of suspected contamination sources, including pass ‘ware contact (upper, solvent fitration (next to top), pH meter probe (nex to bottom), degassing pected hat the wrong soap dluton was used. land we Were right — the typical snk full of dishes contained approximately 10% ofthe recommended detergent. Esen this reduced detergent concenation appeared to leave unacceptable resides ‘We found that rinsing the glassware an ad- itional ive times before use eliminated most ofthe spurious peaks. In addition, we changed the pl adjustment procedure so that aliquots of ber were remaved from the main batch to chock the pH so thatthe pH meter probe swas nt place in the lk ber. Figure 4 compares phosphate gradient using the extra fines and external pH adjustment (lower ‘race) withthe original problem gradient (up- per trace) of Figure 1 pH METER ‘The previous test showed thatthe glassware and pH meter contributed to contamination. ‘When extraclean glassware was used for the ‘reparation of buffer and the pH meter probe ‘vas inserted in the solvent for 30 min, we ob tained an unacceptable baseline, as illustrated inthe upper trace of Figure S. Compare this resul othe lower trace in whic the same starting buffer was adjusted using the external Tf20 1” 6 2 Tine nin} FIQURE 4: Comparison of chromatograms generated using contaminated buffer top) and buter prepared with extraclean glassware and no excosureto the pH meter probe (hottom). (bottom, 0 15 2 Tie (ri) Chromatograms generated usig butfer exposed for 30 min ta the pH meter pro (top) and using butter prepared with extraciean glassware and no exposure to the oH meter crobe pH meter probe technique, The pH meter probe obviously is a major contamination ‘THE CULPRITS EXPOSED ‘Asa final st of tess, we examined the conta ‘mination sources more closely. When unused (or example, between tess), the pH meter probe is kepe immersed in a commescial pH 7 calibration buffer. We diluted this buffer 100- fold with purified water and used it instead of ‘phosphate buffer. The lower tave of Figure 6 ‘hows a blank gradient for ths sample. The large peak at 4 min normally is not a problem: ‘the ether small peaks later in the run may or ‘may not corel with problem peaks inthe normal praient In any event, no large peaks appear to contribute to the general problem. ‘Our second experiment used a 100-fold dlloton ofthe saturated potassium chloride solution used to fil the pHi meter probe. The upper trae of Figure 6 shows a gradient using996 L666 VOLUUKE (6 HUMBER 11 NOVEMBER 1988 upper trace of Figure 6 shows a gradient using ‘his solution instead of batfer. The large peak ‘near 14 min is one of the peaks that consis- ‘ently appears in problem gradient runs (some retention time shifts are observed between days fortis peak because of slightly different LLC conditions). The potassium chloride ap ears to contribute significantly to the buffer contamination problem. Finally, soapy dishwater was diluted [1 ‘wih puified water and SO pL was injected ico the system, Figure 7 compares the cheo- :matogram For soapy water (upper trace) to a (dirty) blank phosphate gradient (lower trace). (Note thatthe upper trace is attenuated five- {old over the lower one.) Cleanly several ofthe peaks inthe contaminated phosphate correlate ‘with peaks from the soapy water CONCLUSIONS This case study shows one example of how analysts can use a stepwise technique for iso- lating the problem source. From the current sudy, we ean draw the following conclusions: ‘In our procedure, six tap water sinses fol- Towed by six deionized water rinses is insul- BS rious peaks in bank baselines occurred when no sample was injected, so they originated from the solvents or system components. ficient to remove soap residues that create unwanted peaks inthe present method. We adjusted the procedure to use two 10-fold rinses and now obtain satisfactory results, # The 0.45-yum filter used for mobile phases does not appear to contaminate the solvents, ‘© Degassing by vigorous helium sparging for 5-10 min does not appear to contaminate the solvents. ‘* Although the storage solution for the pH meter probe does contain potential contami- nants, they do not appear to transfer to the ‘mobile-phase buffer, The lack of contami- nants is likely due to our practice of rinsing the probe thoroughly before each use, ‘The saturated potassium chloride soletion contained inthe pH meter probe contributes significant contamination when the probe is ‘immersed in the buffer during preparation. By checking the pH externa tothe bulk Dutfer (and then discarding the tested aliquot), we were able to determine buifer ‘pH without contaminating the bulk buffer. ‘The experiments described here used UV detection at 215 nm. Many ofthe problem peaks probably would not be of concern at higher wavelengths (for example, 250 nm), ‘Asa first guess, analysts should assume that a single source is responsible for any given LC problem. Systematic problem isola- Frou 6 >i Chromatograns gener 10 a 2 Tine rin} ed using mobile phase prepared trom possi- contanination sources, including the pH 7.0 cal ation staniard (lomer) and ‘and phosphate buffer (lower trace) : : i. ote that the upper trace is fivefold less tion then can be used to find the problem source by changing one variable ata time ‘uni the problem is eKiminated. Although the present problem was caused by two separate sourees and we did not alsvays change just one thing ata time during isolation, stepwise prob lem isolation le us quickly tothe sources. asasaseK Michael D. nelson is @ chenist at LC Resources Inc.'s laboratory in Mottin- ville, Oregon, "LE Troubleshooting” editor John W. Dolan is president of LC Resources Inc of Walnut Creek, California, and a member of LC+G¢'s editorial advisory board, Direct correspondence about this columa to "L¢ Troubleshooting, * Ue+0C, 859 Willanette Sereet, Sugene,