High Speed Gradient Elution Reversed-Phase Liquid Chromatography of Bases in Buffered Eluents Part I: Retention Repeatability and Column Reequilibration

Abstract

We studied the run-to-run repeatability of the retention times of both non-ionizable and basic compounds chromatographed using buffered eluents. The effect of flow rate, organic modifier and other additives, buffer type/concentration, stationary phase type, batch-to-batch preparation of the initial eluent, gradient time, sample type and intra-day changes on retention repeatability were examined. We also assessed the effect of column storage solvent conditions on the inter-day repeatability. Although retention repeatability is strongly influenced by many parameters (flow rate, solvent compressibility compensation, precision of temperature control, and buffer/stationary phase type), our primary finding is that with a reasonable size column (15 cm by 4.6 mm (i.d.)) two column volumes of reequilibration with initial eluent suffices to provide acceptable repeatability (no worse than 0.004 min) for both non-ionizable and basic analytes under a wide variety of conditions. Under ideal conditions (e.g. the right buffer, flow rate, etc.) it is possible to obtain truly extraordinary repeatability often as good as 0.0004 minutes. These absolute fluctuations in retention translate to worst case changes in resolution of 0.2 units and average changes of only 0.02 units.

1. Introduction

There have been many investigations of speed in chromatography [1–11] to reach the ultimate goal of a separation: obtaining an acceptable separation in a reasonable time. The cycle time (injection-to-injection), which is the sum of the gradient development time (t_G) and the reequilibration time (t_re), sets the through-put of gradient elution methods[12]. Although one can decrease t_G and maintain resolution by using smaller particles [13], the pressure limitation of conventional instrumentation (~ 40 MPa) sets the upper limit of speed. To solve this problem, Jorgenson and coworkers [14–16], and Lee and coworkers [17–20] have studied ultra-high pressure liquid chromatography with instrumentation that can withstand significantly higher backpressures.

Although there are many ways to minimize t_G without sacrificing resolution, once the gradient time is set the only way to further increase the speed of gradient elution is to minimize reequilibration time. The current rule of thumb for reequilibrating a column is to flush it with at least 10 column volumes of initial eluent before beginning the next run [21]. Dorsey and coworkers suggested that the addition of a small amount of n-propanol to the eluent can substantially decrease the reequilibration time [22,23]; however, they did not quantify the state of column equilibration as a function of the reequilibration time [23]. Furthermore, Dorsey assumed that the retention of acetone, a relatively unretained species, accurately represents the state of column equilibration under various conditions [22]. Our recent work shows that this assumption is often wrong [12].

To clarify the term “reequilibration”, we define two distinct states of reequilibration: repeatable equilibrium and full equilibrium. Repeatable equilibrium occurs at a value of t_re which suffices to provide excellent run-to-run reproducibility in retention (typically a standard deviation of < 0.002 min for 4 replicates) based on the experimental conditions used (a 150 × 4.6 mm column with 5 µm particles at 1 mL/min). This degree of reproducibility is probably acceptable for all but the most exacting work (vide infra). On the other hand, full equilibrium under gradient occurs at a value of t_re where the retentions of all peaks no longer statistically change as t_re is increased. This can take quite a bit longer than the state of reproducible equilibrium. Although others have investigated column equilibration phenomena [21–31], we believe our previous study was the first to distinguish between these two different states of reequilibration [12].

It should be evident that high reproducibility under isocratic elution requires that the column has reached a state of full i.e. thermodynamic, equilibrium. It is more than possible that reproducible retention, even very reproducible retention, can be obtained in gradient elution chromatography under certain conditions even if a state of full equilibrium with the initial eluent has not been achieved. One can easily imagine that if the final eluent, say one which very readily wets the pores and establishes thermodynamic equilibrium of the mobile and stationary phases, effectively re-initializes the column then all that is needed to achieve reproducible retention is that the instrument deliver a very reproducible gradient.

Previously, based on a test mixture of non-ionizable solutes in unbuffered eluents, we found that repeatable equilibrium was achieved in at most two column volumes of reequilibration under all conditions investigated [12]. However, when full equilibrium was required with flushing limited to 1–2 column volumes it was absolutely essential that both 1% (v/v) n-butanol be added to the eluent and that the system flush-out volume be minimized; when this was not done it was under some conditions necessary to flush the column with more than 20 column volumes of initial eluent.

Although gradient analysis of non-ionizable solutes in unbuffered eluents requires very little time to reequilibrate the system, other work indicated that column equilibration, under isocratic conditions, measured with ionized compounds in buffered eluents could require several hours [24,32]. Thus, we here investigated the reequilibration time needed to achieve repeatable retention in gradient chromatography for cationic solutes in buffered eluents. We also studied the effect of flow rate, organic modifier, buffer type/concentration, stationary phase, n-butanol concentration, batch of the initial eluent and sample mixture on the intra-day repeatability and the effect of the storage solvent on the inter-day repeatability. In a second companion study we will report on the reequilibration time required for full equilibrium. The results in that study are rather different. Specifically we find that one must add n-butanol to the initial eluent to rapidly achieve a state of full equilibrium. Such is not necessary to achieve repeatable equilibrium. However, to minimize confusion in these two studies we decided at the outset to keep conditions as similar as possible.

The degree to which retention variability is acceptable is strongly dependent on the application. Some may be willing to accept a variability of 0.010 min for routine analytical work. However, when the data is subjected to chemometric analysis and the powerful trilinearity condition is imposed, high repeatability becomes very valuable if not absolutely vital. In this case, the potential advantage of obtaining excellent repeatability in retention (< 0.002 min) is that data pre-processing steps such as parametric time warping, dynamic time warping (DTW) or correlation optimized warping (COW) procedures [33–35] can be simplified or sometimes avoided before performing chemometric curve resolution [36]. Although such time warping algorithms show great potential for aligning chromatograms with poor repeatability (i.e. > 0.01 min), the results of time warping are highly dependent on the sample and experimental conditions. 2DLC is another case where extremely high repeatability presents an advantage because one needs to align the chromatograms with high precision to construct the 2D picture from a long series of 1D runs.

Another advantage of obtaining excellent repeatability is that the potential to identify components using their retention will be improved considerably [37]. Furthermore, the ability to obtain excellent repeatability with a short reequilibration time is of great importance to increase sample throughput in fields such as forensic toxicology and drug discovery [38–44]. Obviously, the ability to obtain excellent repeatability with minimal reequilibration time for basic compounds separated in buffered eluents has practical implications in many fields and applications.

2. Experimental

2.1 Instrumentation

All chromatographic experiments were conducted using an HP 1100 chromatographic instrument controlled by version A.10.01 Chemstation software (Agilent Technologies; Palo Alto, CA). The HP 1100 was equipped with a low pressure mixing chamber, autosampler, block heater, quaternary pump, and variable wavelength UV detector. The dwell volume of the HP 1100 instrument, including all tubing required to connect the column, was determined to be 0.90 mL at 1 mL/min using the technique found in chapter 8 of reference [45]. The flush-out volume of the HP 1100 was determined to be 2.5 mL (99%) using the technique described previously [12]. The flow rate was checked using a 10 mL volumetric flask and a stopwatch, and was determined to be consistently accurate to within 1% of the set point.

Temperature was monitored at the column outlet using a thermistor (part # 44008) from Omega Engineering, Inc. (Stamford, CT). The thermistor was wrapped in a piece of copper metal soldered to 5 cm of pre-cut stainless steel tubing (0.007”) obtained from Upchurch Scientific, Inc. (Oak Harbor, WA). An OA-2 OP-AMP Designer (E & L Instruments, Inc.; Derby, CT) was used to supply a 5 V input to a Wheatstone bridge circuit containing the appropriate resistances to output a 0 – 1 V signal around temperatures of 40°C. The analog signal was recorded in Chemstation using an HP 35900E A/D Interface.

2.2 Reagents

All solutes were of reagent grade or better and were used as obtained from the manufacturer without further purification. Uracil, acetone, pheniramine, methapyrilene, chlorpheniramine, N-benzyl formamide, 3-pentanone, alprenolol, protriptyline, promethazine, perphenazine, nortriptyline, amitriptyline, thioridazine, meclizine, a homologous series of nitroalkanes (nitromethane, nitroethane and nitropropane) and a homologous series of alkylphenones (acetophenone, propiophenone, butyrophenone and valerophenone) were obtained from Aldrich (Milwaukee, WI). These solutes were diluted into one sample using a 10/89.9/0.1 (v/v/v) acetonitrile/water/trifluoroacetic acid eluent. The concentration of uracil, acetone, N-benzylformamide, 3-pentanone and the nitroalkanes was 1 mg/mL; the concentration of alprenolol, thioridazine and the alkylphenones were 300 µg/mL, 30 µg/mL and 5 µg/mL, respectively. The concentration of all other solutes was 150 µg/mL. Uracil was used to measure the kinetic dead volume of the column. This solute mixture was used to collect repeatability data unless otherwise noted.

A trypsin digestion protocol from Michrom BioResources, Inc. (Auburn, CA) was slightly modified and used to prepare a sample of tryptic peptides from bovine serum albumin (BSA). Ammonium bicarbonate was purchased from Fisher (FairLawn, NJ) and iodoacetic acid was purchased from Avocado Research Chemicals Ltd. (Heysham, Lancashire). TPCK treated trypsin, tris(2-carboxyethyl)phosphine hydrochloride and BSA (Cohn fraction V, product number A-2153) protein were obtained from Sigma (St. Louis, MO).

The eluent reservoirs and filtration apparatus glassware were scrupulously cleaned, rinsed with water then acetonitrile, and dried using nitrogen before use. The organic co-solvents in this study were used as obtained from the manufacturer; HPLC grade acetonitrile was obtained from Aldrich, HPLC grade methanol and tetrahydrofuran were obtained from Mallinckrodt Baker, Inc. (Phillipsburg, NJ), and ACS grade n-butanol was obtained from Fisher. HPLC grade water was obtained in-house from a Barnstead Nanopure Deionizing system (Dubuque, IA). This water was boiled to remove carbon dioxide and cooled to room temperature before use. Trifluoroacetic acid (ReagentPlus 99%), formic acid (88%) and phosphoric acid (85%) were obtained from Aldrich. Triethylamine was obtained from Spectrum Chemical MFG. Corp. (Gardena, CA).

All eluents were prepared gravimetrically (+/−0.01 g) based on the density [46] at room temperature (25 °C) of acetonitrile (MeCN), methanol (MeOH), tetrahydrofuran (THF), n-butanol (BuOH), trifluoroacetic acid (TFA), formic acid and water (H₂O) where eluent composition is reported as the v/v ratio. Solvents were made by first adding BuOH to the organic modifier followed by dilution with water; components of the buffer were added last. Each triethylamine phosphate (TEAP) buffer was prepared in pure water first to determine the amounts of phosphoric acid and triethylamine required for a specific TEAP concentration (reported as the phosphoric acid concentration) and pH. The true pH of the eluent was not measured after addition of organic modifier. The eluents were stirred magnetically until they reached room temperature. All eluents were passed through a 0.45 µm nylon filtration apparatus from Lida Manufacturing Inc. (Kenosha, WI) immediately before use. These eluents were not degassed to any extent beyond the degassing that occurred during filtration.

2.3 Columns

All columns were of dimension 15 cm × 4.6 mm i.d. packed with 5 µm narrow pore (i.e. 80 Å) particles. An SB-C₁₈ column was used for most of the study but an XDB-C₁₈ column was also used; both columns were gifts from Agilent technologies.

2.4 Chromatographic Conditions

Detection was performed at 254 nm, the column thermostat was set to 40.0 °C and 5 µL injections of sample were made. The instrument was programmed to deliver a gradient from 100% channel A to 100% channel B in 10.00 mL followed by a hold at 100% channel B for 1.00 mL and then a step change back to 100% channel A; flow rates of 1.00 mL/min, 2.00 mL/min and 3.00 mL/min were used. The instrument was flushed with 100% channel A for a desired reequilibration time before ending the run (i.e. stopping data collection and beginning data analysis). The time between the end and beginning of two consecutive runs in the sequence (i.e. injector cycle time) was approximately one minute; this additional reequilibration of the column is included in the reported reequilibration times/volumes below. However, the time/volume required to flush-out the instrumentation to redeliver initial eluent to the column (see section 2.1 above) is not included in the reequilibration time/volume reported. The full equilibrium scheme described previously was used to collect repeatability data [12]. We gathered data for eight test reequilibration volumes (typically 3, 4, 6, 8, 11, 16, 26 and 32 mL) and used a control reequilibration volume of 15 mL (roughly 10 column volumes). The data analysis procedure used to calculate the repeatability is described in ref. [12].

3. Results/Discussion

3.1 Intra-Day Repeatability Studies

Fig. 1 shows a typical separation of the mixture of non-ionizable solutes and basic drugs on a SB-C₁₈ column using an acetonitrile/water eluent with 1% (v/v) n-butanol and a 0.1% (v/v) TFA buffer. To minimize measurement bias, the solute mixture was prepared so that both the cationic and neutral solutes were dispersed across the entire chromatographic space.

Fig. 1.

Representative gradient elution separation of the non-ionizable solute and basic drug mixture. The baseline signal was subtracted from the chromatogram. Conditions: 15 cm × 4.6 mm I.D. column packed with 5 mm SB-C₁₈ 80 Å particles; eluent A is 1/10/88.9/0.1 BuOH/ACN/H2O/TFA (v/v/v/v); eluent B is 1/90/8.9/0.1 BuOH/ACN/H2O/TFA (v/v/v/v); 100/0 to 0/100 to 0/100 A/B in 10 mL to 11 mL at 2 mL/min; V_D = 0.90 mL; 40 °C; 5 µL injection; 254 nm detection; 10 mL reequilibration volume. Solutes: 1) uracil; 2) solvent A; 3) solvent B; 4) acetone; 5) pheniramine; 6) nitromethane; 7) methapyrilene; 8) chlorpheniramine; 9) N-benzylformamide; 10) nitroethane; 11) 3-pentanone; 12) alprenolol; 13) protriptyline; 14) acetophenone; 15) promethazine impurity; 16) promethazine; 17) nitropropane; 18) perphenazine; 19) nortriptyline; 20) amitriptyline; 21) propiophenone; 22) thioridazine; 23) butyrophenone; 24) meclizine; 25) valerophenone; x) unknown impurities

3.1.1 Repeatability as a function of the reequilibration time

The retention mechanism of basic drugs separated in buffered eluents are sometimes more complicated than that of non-ionizable solutes. Specifically, the retentions of basic drugs under certain conditions are determined by both reversed-phase and ion exchange (i.e. Coulombic) interactions which result in a mixed-mode mechanism (see refs. [47–51]). Obviously, ion exchange is not possible when using non-ionizable solutes. Therefore, we anticipated that the repeatability of the basic drugs might be worse than that of the non-ionizable solutes at least under certain conditions. Fig. 2 clearly shows that both solute types show similar repeatability with a SB-C₁₈ column and an acetonitrile-water eluent with 0.1% (v/v) TFA. Additionally, there is no trend in repeatability with reequilibration time (data not shown) provided that the column is flushed with at least one column volume of the initial eluent. Therefore, we will report the median repeatability (i.e. either the median of the repeatability for a solute of all reequilibration times (n = 16) used or the median of the repeatability for all solutes and reequilibration times) for the remainder of this study unless otherwise noted. Although a repeatability of 0.003 min is quite acceptable, some experimental conditions explored showed statistically very significant effects on the repeatability. Therefore, we studied the effect of a number of key experimental parameters on repeatability.

Fig. 2.

Median repeatability in retention time versus solute retention time. The median repeatability (n = 4) of all reequilibration times used for the non-ionizable solutes (●) and basic drugs (○) is reported. The error bars represent the standard deviation (n = 16) of the median repeatability from all reequilibration times used. Other conditions are described in Fig. 1.

3.1.2 Effect of buffer composition and type on the repeatability of basic drugs

Table 1 shows that for the basic drugs the 0.1% (v/v) TFA, 0.5% (v/v) TFA and 0.1% (v/v) formic acid buffers gave repeatability as good as the neutral solutes; poorer but still quite acceptable repeatability was observed with TEAP buffers. Furthermore, we found that the repeatability for the neutral solutes was always < 0.001 min and independent of the TEAP buffer composition (data not shown). Fig. 3 shows that the repeatability of the more strongly retained basic drugs was worse than the weakly retained basic drugs for all TEAP buffer compositions used. Increasing the concentration of phosphoric acid (i.e. 42 mM TEAP at pH = 3.0 or 50 mM TEAP at pH = 2.0) improved the repeatability but it is clear that some property of the TEAP buffer lowered the repeatability under these conditions. The effect is not due to the pH per se as the TFA and formic acid buffers are not more acidic than the pH 2.0 TEAP buffer.

Table 1.

The effect of the buffer type/concentration on the repeatability of basic solutes.^a

Buffer	Repeatabilitya (min)
0.1% (v/v) TFA	0.0005 ± 0.0004
0.5% (v/v) TFA	0.0005 ± 0.0002
0.1% (v/v) Formic Acid	0.0006 ± 00005
16 mM TEAP (pH = 3.0)	0.0020 ± 00005
42 mM TEAP (pH = 3.0)	0.0009 ± 00002
50 mM TEAP (pH = 2.0)	0.0011 ± 00003

^a.The median repeatability (n = 4) of the basic drugs at all reequilibration times used is reported. The median repeatability (n = 4) of the non-ionizable solutes at all reequilibration times used in each buffer was ~ 0.0005 min. Other conditions are described in Fig. 1.

Fig. 3.

Median repeatability in retention time of the basic drugs versus solute retention volume and TEAP buffer concentration. The median value of the repeatability (n = 4) from all reequilibration times used for the 16 mM TEAP pH = 3.0 (●), 42 mM TEAP pH = 3.0 (○) and 50 mM TEAP pH = 2.0 (▼) buffer compositions is reported. Other conditions are described in Fig. 1.

The rate of change of the pK_a of formic acid with temperature (0.002 pK_a units/°C) and that of the first pK_a of phosphoric acid (0.006 pKa units/°C) differ by a factor of three [52]. No corresponding data for TFA were found but dilute TFA is almost fully dissociated so there would be very little change in pH in any case. All of the pHs used in this study were so low that all the basic solutes were fully protonated regardless of the pH. Therefore, we believe that thermally induced changes in the interactions of the bases, not changes in the degree of ionization of the solutes, with some of the ionized silanols on the SB-C₁₈ phase are responsible for the poorer repeatability observed in TEAP buffers. To test this hypothesis, we measured the repeatability of the basic drugs in a TEAP buffer using a second but closely related column which we felt would have fewer available silanol sites.

3.1.3 Repeatability of basic drugs in a 50 mM TEAP at pH = 2.0 buffer as a function of the stationary phase type

XDB phases are based on the same silica substrate as SB phases, however, XDB is end-capped. We hypothesized that the pH equilibration of XDB-C₁₈ might be faster than of SB-C₁₈ and thus possibly more reproducible. Table 2 indicates that retention on XDB-C₁₈ is considerably more repeatable than on SB-C₁₈. To be sure of this, we performed a two-way ANOVA test (with α = 0.05) to compare the repeatability of the basic compounds (i.e. average value for the solute from all reequilibration times used) on each column. For just the basic solutes we found that there is a statistically significant difference in repeatability between XDB-C₁₈ and SB-C₁₈ using a 50 mM TEAP (pH = 2.0) buffer (F (30.65) > F_critical (4.60)) but that repeatability was not dependent on retention (F (1.51) > F_critical (2.48)). Thus, the stationary phase and buffer type can affect the repeatability of basic compounds. However, it is clear that even the worst combination of stationary phase and buffer type may still provide acceptable repeatability (< 0.004 min).

Table 2.

The effect of the stationary phase on the repeatability^a of basic solutes using a TEAP buffer.^b

Stationary Phase	Repeatabilitya (min)
SB-C18	0.0011 ± 00003
XDB-C18	0.0006 ± 00003

^a.The median repeatability (n = 4) of the basic drugs at all reequilibration times used is reported. The median repeatability (n = 4) of the non-ionizable solutes at all reequilibration times used on each stationary phase was ~ 0.0005 min.

^b.A 50 mM TEAP buffer at pH = 2.0 was used; other conditions are described in Fig. 1.

3.1.4 Effect of flow rate on repeatability

We examined the effect of flow rate on retention repeatability (see Fig. 4); note that the gradient time was simultaneously adjusted to hold the gradient slope constant to avoid changing selectivity [53,54]. We were rather surprised to see that compared to 1 mL/min, the higher flow rates gave better retention time repeatability (< 0.002 min) for the intermediately to highly retained solutes regardless of their retention volume (see Fig. 4A). The more weakly retained solutes, i.e. those near the dead time, gave excellent retention time repeatability (i.e. ~ 0.0005 min) at all three flow rates. This indicates that the trend in repeatability versus flow rate is not due to imprecision in the flow rate as the flow rate is the main variable for manipulating the retention time of weakly retained species. Furthermore, the similar retention time repeatability at 2 and 3 mL/min suggests that limitations in the instrumentation and data analysis software do not allow better values of the repeatability to be observed.

Fig. 4.

Median repeatability in retention versus solute retention volume and flow rate. The repeatability is calculated using the retention time (A) and retention volume (B). The median value of the repeatability (n = 4) from all reequilibration times used at flow rates of 1 mL/min (●), 2 mL/min (○) and 3 mL/min (▼) is reported. Additional solutes not identified in Fig. 1 (i.e. impurities) are included in this plot. Other conditions are described in Fig. 1.

In Fig. 4B we examined the effect of flow rate on the retention volume (V_R) repeatability. When plotting the data in this fashion, the trends observed in Fig. 4A, although less definite, remain the same: at 1ml/min repeatability worsens for the more highly retained solutes whereas the at the higher flow rates repeatability is nearly independent of retention volume; this is most evident at 3 mL/min. Based on the data in Fig. 4, we were compelled to perform additional experiments to investigate why higher flow rates gave better retention time repeatability and nearly constant repeatability independent of retention. In Appendix I, we show that the increased backpressure resulting from the higher flow rates or changes in the mixing process as a function of the flow rate cannot explain the enhanced repeatability at higher flow rates. However, Appendix II shows that run-to-run variation in the average column temperature has an important effect on the retention time. More specifically, the flow rate interacts with the cycle time of the column heater such that the use of higher flow rates results in more accurate control of the column temperature and improved repeatability. We further explore the effect of column temperature on gradient retention time in Appendix III to clarify the results in Fig. 4. Appendix IV shows that higher flow rates provide acceptable repeatability for different organic modifiers whereas lower flow rates might require optimization of the solvent compressibility compensation factor.

3.1.5 Repeatability as a function of fraction of n-butanol

Previous work showed that addition of 1% (v/v) n-butanol to the eluent unequivocally reduces the time required to establish full column equilibrium but the effect of butanol on the time needed to achieve repeatable equilibrium was not investigated [12]. Therefore, we tested the effect of n-butanol on retention repeatability and what is required to achieve it. Table 3 shows that addition of n-butanol in amounts less than 3% (v/v)) has no effect on how long the column must be flushed to achieve repeatable equilibrium. This is not to say that it has no effect on the extent of flushing needed for full equilibration. We recommend the addition of 1% (v/v) n-butanol only under those rare circumstances where fast, full column equilibration is required. Of course, one must do this during method development to take into account the effect of butanol on retention. Unfortunately, adding even only 1% n-butanol decreases the gradient range by increasing the initial eluent strength. Thus, the resolution of separations requiring water-rich initial eluents might worsen by using n-butanol.

Table 3.

The effect of the n-butanol concentration on the repeatability^a.

Eluent System	Repeatabilitya (min)
0/10/89.9/0.1 to 0/90/9.9/0.1 BuOH/MeCN/H2O/TFA (v/v/v/v)	0.0005 ± 00003
1/10/88.9/0.1 to 1/90/8.9/0.1 BuOH/MeCN/H2O/TFA (v/v/v/v)	0.0005 ± 00004
3/8/88.9/0.1 to 3/88/8.9/0.1 BuOH/MeCN/H2O/TFA (v/v/v/v)	0.0006 ± 00003

^a.The median repeatability (n = 4) of all solutes and all reequilibration times used is reported. Other conditions are described in Fig. 1.

3.1.6 Repeatability among different batches of initial eluent

When a new batch of initial eluent is prepared, a small change in the eluent composition might have a significant effect on the retention time and thus repeatability. We chose not to investigate the repeatability related to the final eluent as the initial eluent has a larger effect on the selectivity and absolute value of the retention time, especially for the weakly retained solutes. Instead of measuring changes in repeatability caused by intentional changes in the initial eluent composition, we investigated the repeatability available from different batches of initial eluent prepared gravimetrically. Table 4 shows that we obtained superb repeatability with four batches of the initial eluent prepared gravimetrically when the precision of weighting each component (4.05 g/38.83 g/445.00 g/0.78 g BuOH/MeCN/H₂O/TFA) is ± 0.01 g. Under no circumstances do we suggest adjusting the pH by adding acids or bases and monitoring changes with a pH meter if one desires good reproducibility of bases when using different batches of eluent.

Table 4.

Average retention time of each solute^a using four batches of the initial eluent on the same day.^b

Solute	Initial Eluent Batchc				Average (min)	STDEV (min)
	1	2	3	4
uracil	0.7281	0.7282	0.7284	0.7278	0.7281	0.0002
solvent A	0.7769	0.7769	0.7770	0.7767	0.7769	0.0001
solvent B	0.9785	0.9791	0.9790	0.9784	0.9788	0.0003
acetone	1.0759	1.0766	1.0771	1.0763	1.0765	0.0005
pheniramine	1.5408	1.5420	1.5416	1.5410	1.5413	0.0006
nitromethane	1.7052	1.7060	1.7059	1.7052	1.7056	0.0005
methapyrilene	1.9044	1.9054	1.9061	1.9051	1.9053	0.0007
chlorpheniramine	2.2552	2.2550	2.2561	2.2556	2.2555	0.0005
N-benzylformamide	2.3865	2.3859	2.3869	2.3869	2.3865	0.0005
nitroethane	2.6776	2.6767	2.6771	2.6777	2.6773	0.0005
3-pentanone	2.7427	2.7432	2.7443	2.7438	2.7435	0.0007
alprenolol	3.3133	3.3118	3.3130	3.3136	3.3129	0.0008
protriptyline	3.3480	3.3462	3.3470	3.3478	3.3472	0.0008
acetophenone	3.4391	3.4376	3.4384	3.4393	3.4386	0.0008
impurity	3.5248	3.5237	3.5256	3.5258	3.5250	0.0009
promethazine	3.6171	3.6152	3.6166	3.6174	3.6166	0.0010
nitropropane	3.6725	3.6705	3.6717	3.6726	3.6718	0.0010
perphenazine	3.7758	3.7737	3.7753	3.7762	3.7752	0.0011
nortriptyline	3.8691	3.8669	3.8683	3.8693	3.8684	0.0011
amitriptyline	3.9355	3.9333	3.9346	3.9356	3.9348	0.0011
propiophenone	4.2038	4.2025	4.2045	4.2049	4.2039	0.0010
thioridazine	4.3838	4.3815	4.3828	4.3835	4.3829	0.0010
butryophenone	4.7069	4.7057	4.7078	4.7080	4.7071	0.0011
meclizine	4.7529	4.7507	4.7521	4.7528	4.7521	0.0010
valerophenone	5.1461	5.1445	5.1467	5.1471	5.1461	0.0011

^a.The average retention time (min) of four runs was used and the standard deviation (n = 4) was < 0.001 min for all solutes.

^b.Other conditions are described in Fig. 1.

^c.The initial eluent was prepared gravimetrically by adding each component (4.05 g/38.83 g/445.00 g/0.78 g BuOH/MeCN/H₂O/TFA) with a precision of ± 0.01 g.

3.1.7 Effect of the sample on the repeatability

Fig. 5 is a convincing illustration of the ability to achieve extraordinary reproducibility in gradient elution chromatography as shown by the nine superimposed chromatograms of the set of tryptic digest peptides of BSA and the associated “blow-ups.” Obviously, outstanding repeatability in retention time and peak shape (i.e. peak area and height) is possible after only about 1.5 column volumes of reequilibration; the reproducibility of a number of the well formed, larger peaks spaced throughout the chromatogram is summarized in Fig. 6.

Fig. 5.

Nine overlaid chromatograms of the BSA tryptic peptide sample separated over a short time with minimal column reequilibration. We show the (A) entire chromatogram and associated “blow-ups” of the (B) beginning and (C) near-end of the separation. The column was reequilibrated with 1.5 column volumes of initial eluent before the next run was begun. No data analysis was performed at the end of the run to shorten the reequilibration time. Conditions: Eluent A: 5/94/9/0.1 MeCN/H₂O/TFA; Eluent B: 60/39.9/0.1 MeCN/H₂O/TFA; 100/0 to 20/80 A/B in 12 min, hold at 20/80 A/B for 1 min then back to 100/0 A/B; the run was ended at 14 min; 2 mL/min; V_D = 0.90 mL; 220 nm detection; 15 µL injection; 40 °C.

Fig. 6.

Repeatability (n = 9) of well formed peaks from the BSA tryptic peptide separation described in Fig. 5.

In the work shown in Fig. 5 we did not add any n-butanol to the eluent because preliminary work showed that slightly longer reequilibration times (approximately 2.2 column volumes) were required to reach a state of repeatable equilibrium for those peaks eluting between 0.8 and 1.3 minutes. Furthermore, we obtained very poor standard deviations (as bad as 0.030 min, see Fig. 7) for these peaks. Additionally, the peak height reproducibility for solutes eluting between 0.8 and 1.3 minutes in Fig. 7 was not as good as that observed in Fig. 6. However, all peaks eluting both before and after this interval were quite repeatable (< 0.002 min) and not dependent on the presence of n-butanol or on the reequilibration time provide that the column was flushed with at least 0.5 columns volumes of initial eluent. This indicates that a “critical region” where the repeatability of analytes is strongly influenced by the reequilibration time exists. It is possible that the critical region results due to incomplete column equilibration as discussed on pages 386–391 in ref. [55]. We believe the location and presence of a “critical region” is highly dependent on the separation conditions. Obviously, it can only be observed if the sample contains species which elute in the critical region.

Fig. 7.

Evidence for a “critical region” with poor repeatability for the BSA tryptic peptide mixture. A reequilibration of 1.6 (A) or 2.2 (B) column volumes was used at the end of the gradient. Conditions: Eluent A: 1/5/93.9/0.1 BuOH/MeCN/H₂O/TFA; Eluent B: 1/60/38.9/0.1 BuOH/MeCN/H₂O/TFA; other conditions are described in Fig. 5.

We found one situation were the addition of n-butanol to the eluent actually impeded, albeit only slightly, the speed of gradient elution (see Figs. 5–6). To test for the presence of a “critical region,” we recommend comparing the repeatability obtained after five column volumes of reequilibration to the repeatability obtained after a shorter time. The “critical region” will be discussed in more detail in a subsequent study [56].

3.1.8 Repeatability of selectivity and resolution

Despite the excellent intra-day retention time/volume repeatability under gradient conditions, another question remains: does the resolution of adjacent peaks significantly change from run-to-run? Using the SB-C₁₈ column with an acetonitrile-water-0.1% (v/v) TFA eluent system at 2 mL/min, we measured the average baseline peak width (W) of analytes in the mixture of non-ionizable solutes and basic drugs to be about 0.051 min. Also, we found that the worst variability in the retention time difference between two adjacent peaks was 0.005 min whereas the average value was nearly ten-fold lower (i.e. 0.00065 min). We were not surprised to discover that the value of 0.005 min was for a solute pair eluted within the “critical region” at the shortest reequilibration time used and that its variability improved dramatically at slightly longer reequilibration times. Using the general resolution equation (see eqn. 1 where t_R is the retention time and t_R,2 > t_R,), it is evident that the resolution does not

$$R_s=\frac{2·\left(t_{R,2}-t_{R,1}\right)}{W_1+W_2}$$ (1)

change by more than 0.1 units in the worst case and that changes of 0.01 units are more common. Even when the buffer with the poorest reproducibility (16 mM TEAP at pH = 3.0)), the resolution only changed at most by 0.2 units and the average changes were about 10-fold smaller. Thus, it appears that conditions that provide excellent retention repeatability also minimize changes in resolution from run-to-run. Under conditions where the resolution is robust (i.e. R_s = 2.0), changes in resolution on the order of 0.2 would be inconsequential; however, when resolution is only marginally adequate (say R_s = 1.0) such changes would have serious consequences especially if the peaks were quite different in size.

3.2 Inter-Day Repeatability Studies

Fig. 8 shows the intra-day (12 runs) and inter-day (6 days) repeatability for representative solutes for a column which was stored in the initial eluent. Because the first one or two runs in a day were collected before the column and instrument were ready (i.e. thermally equilibrated and/or properly conditioned), the retention time for these runs were outliers from the otherwise satisfactory intra-day repeatability (< 0.001 min). However, obvious day-to-day retention time drift occurred for all but the most weakly retained solutes (i.e. k’ < 2).

Fig. 8.

Inter-day retention time of representative solutes when the column is stored in the initial eluent. After the instrument was able to perform an injection but not necessarily thermally equilibrated or properly conditioned, twelve runs were performed on each day. The column was flushed with 60 mL of the initial eluent at 2 mL/min and then removed from the instrument at the end of each day. Other conditions are described in Fig. 1.

Stopping the instrument causes small changes in instrument components (i.e. pump seals) such that the instrument performance drifts with time. For example, the pump seals expand during use and may permanently change from day-to-day due to the stress of starting and stopping the instrument. Another explanation for this retention drift is that organic modifier is lost from the initial and final eluent over time despite efforts taken to seal the eluent bottles. Although we do not fully understand what caused these small changes in retention, we believe that a retention index scheme [57–60] will improve the day-to-day repeatability.

Exposure of a given column to a variety of eluents and storage solvents are inevitable. Fig. 9 shows that the storage solvent (see Table 5) has a complex affect on the day-to-day variations in analyte retention. For example, the day-to-day retention times of all solutes vary non-systematically when the column was stored in any liquid other than the initial eluent. Furthermore, upon cycling between different eluents (i.e. organic modifiers or buffers) we noticed that persistent shifts in the retention time of the basic drugs and non-ionizable solutes of up to 0.1 min and 0.01 min, respectively, took place. These are clearly significantly higher shifts in inter-day retention time compared to the results in Fig. 8 where the column was stored in the initial eluent. However, there were no significant changes in the intra-day repeatability. Therefore, we very strongly recommend the practice of dedicating columns to specific assays and their storage in the initial eluent when the highest possible retention repeatability is required.

Fig. 9.

Inter-day retention time of representative solutes when the column is stored in various solvents. Twelve runs were performed each day as described in Fig. 6. The column was flushed as described in Fig. 6 and the storage solvents used are listed in Table 5.

Table 5.

Initial and final solvent used to store the column from day-to-day.

Day	Initial Solventa	Storage Solventb
1	Eluent Ac	ACN
2	ACN	45/55 ACN/H2O (v/v)
3	45/55 ACN/H2O (v/v)	Eluent Bc
4	Eluent Bc	MeOH
5	MeOH	ACN
6	ACN	Eluent Ac
7	Eluent Ac	MeOH
8	MeOH	Eluent Ac

^a.The initial solvent was the storage solvent from the previous day.

^b.The column was flushed at 2 mL/min for 60 minutes with the indicated solvent before being removed from the instrument and stored overnight.

^c.These eluents are described in Fig. 1.

4. Conclusions

The major conclusion of this work indicates that in the majority of the conditions examined here at most only two column volumes of reequilibration to the initial eluent were required to obtain acceptable repeatability (< 0.004 min) for both non-ionizable solutes and basic compounds. However, we found that many variables significantly affect the repeatability. The following summarizes the important conclusions from this work.

• Buffers whose pK_a are less dependent on temperature and endcapped stationary phases show decidedly better intra-day repeatability for basic compounds.

• The presence of a “critical region” in the chromatogram, which is strongly dependent on the sample type and experimental conditions, may require somewhat longer reequilibration times (e.g. three column volumes) before a state of repeatable equilibrium is established.

• Conditions which provided the worst repeatability encountered here (> 0.004 min) did not result in changes in resolution by more than 0.2 units which is trivial for most work unless the resolution is low.

• Use of higher flow rates (e.g. > 2 mL/min) improves the precision of column temperature control, which in turn improves retention repeatability.

• Improvements in the inter-day repeatability will result when new liter-scale batches of the initial eluent are gravimetrically preparing by adding the components with a precision of ± 0.01 g.

• Storing the column in the initial eluent overnight and dedicating columns for use with a specific eluent dramatically improve the inter-day repeatability.

Appendix I: Repeatability as a function of the system backpressure and pumping system parameters

Fig. 4 suggests that the increased system backpressure at higher flow rates may have improved the repeatability. Therefore, we added columns at the outlet of an HP 1040A detector equipped with a micro flow cell (1.7 µL) (which can withstand higher backpressures) to adjust the backpressure and maintain a 1 mL/min flow rate. Table A1 shows that repeatability worsens at higher backpressures which means that the higher backpressure at higher flow rates cannot account for the improvement in repeatability.

Table A1.

The effect of pressure added to the instrument^a on the repeatability^b.

Pressurea (bar)	Repeatabilityb (min)
0	0.0011 ± 0.0005
130	0.0016 ± 0.0006
240	0.0022 ± 0.0006

^a.To tolerate additional back-pressure, we used an HP 1040A detector and micro flow cell instead of the VWD detector on the HP 1100. The backpressure was varied by attaching a column to the HP 1040 detector outlet. The instrument was operated at 1 mL/min and provided a backpressure of 90 bar without any column attached to the detector outlet. Other conditions are described in Fig. 1.

^b.The median repeatability (n = 6) of all solutes is reported.

Other parameters beside the system backpressure change at higher flow rates. For example, the stroke volume increases to compensate for solvent compression changes. Likewise, one may adjust the solvent compressibility compensation factor although this parameter does not change with the flow rate. Regardless, these pumping system parameters might affect the repeatability. Fig. A1A shows that the stroke volume has a negligible effect on the repeatability at 1 mL/min. Conversely, Fig. A1B shows that the solvent compressibility compensation factor does affect the repeatability to some extent. Unfortunately, neither result explains why the repeatability improves at higher flow rates.

Fig. A1.

Effect of the pumping system parameters on the repeatability. We varied the A) the stroke volume from 20 µL (), 33 µL (), 50 µL (), 70 µL (), 90 µL () and 100 µL () and B) solvent compressibility compensation factor from 0·10⁻⁶ bar (), 25·10⁻⁶ bar (), 50·10⁻⁶ bar (), 60·10⁻⁶ bar (), 70·10⁻⁶ bar (), 80·10⁻⁶ bar () and 100·10⁻⁶ bar () to obtain the median, average, maximum and minimum repeatability (n = 6) of all solutes at 1 mL/min. The default value of the stroke volume (33 µL at 1 mL/min) and solvent compressibility compensation factor (100·10⁻⁶ bar) were used when varying the other pumping system parameter. Other conditions are described in Fig. 1.

Appendix II: Why does the repeatability vary with flow rate?

Another variable that might affect the repeatability at different flow rates is the precision of the column temperature. Therefore, we monitored the column temperature at various flow rates (and scaled gradient times) using a thermistor (see Fig. A2). The decrease in column temperature in Fig. A2 is due to the lower viscosity of eluents that contain more acetonitrile and due to the concomitant decrease in column heating (i.e. there is less frictional heating). When the gradient ends, the initial, higher viscosity eluent begins to percolate through the column and the column temperature increases.

Fig. A2.

Temperature as a function of time for gradients with different flow rates and scaled gradient times. The temperature response (in mV) was measured using a thermistor and Chemstation software (see Experimental section) and converted into a temperature value with a calibration curve (a change of 1 mV corresponds to a temperature change of 0.029 °C). A series of twelve gradient runs were performed at each combination of flow rate and gradient time. The twelve runs at F = 1 mL/min, t_G = 10 min and F = 2 mL/min, t_G = 5 min were performed twice. Other conditions are described in Fig. 1.

Fig. A3 plots the retention time of a select solute versus the average column temperature of the run at each flow rate. Obviously, column temperature is varying less at higher flow rates. This shows that higher flow rates improve the column thermostat’s tightness of control. Thus, lower flow rates provide poor repeatability due to column temperature drift. A higher power demand at higher flow rates decreases the thermostat cycle time and intrinsically improves the column temperature precision. Also, the shorter gradient times at higher flow rates may further inhibit temperature “wander.”

Fig. A3.

The relationship of retention time and column temperature at different flow rates and scaled gradient times. The average column temperature during one gradient run (see Fig. A3) is reported. The retention time of the same solute eluting at each condition is plotted. Other conditions are described in Fig. 1.

More thermostating trends are evident in Fig. A2. For example, the absolute temperature and change in temperature throughout the run increase at higher flow rates (and shorter gradient times) due to the generation of more heat in less time. Also, there is more short-term variation in temperature at lower flow rates. We believe the column thermostat senses a decrease in temperature as the acetonitrile concentration increases, which triggers heating, but the overpowering effect of a decreasing eluent viscosity again lowers the temperature. It is important to note that the column thermostat only measures the temperature of the pre-heated eluent entering the column, not the column temperature itself. Thus, the thermostat cannot recognize that the column temperature increases at higher flow rates due to more frictional heating; this is why the column temperature increases beyond the set point at higher flow rates. Overall, we believe it is best to use higher flow rates and shorter gradient times to decrease temperature drift and improve retention repeatability while appropriately decreasing the thermostat set point.

Appendix III: The effect of temperature on retention time

To determine the change in retention as a function of temperature, we varied the column temperature from 25 to 55 °C and recorded the retention time of representative solutes in the mixture. Fig. A4 shows the predicted change in retention time resulting from a 0.1 °C change in column temperature. Obviously, a small fluctuation in temperature can cause the retention time to change by 0.002 min. Also, the highly retained solutes are more sensitive to changes in temperature which agrees with the repeatability trend at 1 mL/min in Fig. 4.

Fig. A4.

Change in retention time caused by a 0.1 °C change in the column temperature. The retention time of the non-ionizable solutes and basic drugs at 40 °C is plotted on the X-axis. Linear fits of the gradient retention time as a function of the temperature (25, 30, 35, 38, 39, 40, 41, 42, 45, 50 and 55 °C) were generated at 1 mL/min for each solute using the conditions described in Fig. 1.

Appendix IV: Repeatability as a function of organic modifier

Here we compare the repeatability of acetonitrile (MeCN), methanol (MeOH) and tetrahydrofuran (THF) at 1 mL/min and 2 mL/min (see Table A2). Although the basic drugs and non-ionizable solutes exhibit similar repeatability with each organic modifier (data not shown), the repeatability is worst with MeOH at 1 mL/min. However, the repeatability at 2 mL/min was excellent (~ 0.0005 min) and independent of the organic modifier. As the viscosity of THF-H₂O and MeOH-H₂O mixtures are similar yet higher than MeCN-H₂O mixtures in the range of 10% to 70% (v/v) organic modifier [61,62], this solvent property cannot explain the trends in Table A2. However, mixtures of MeOH-H₂O have higher surface tensions compared to mixtures of MeCN-H₂O or THF-H₂O [63].

Table A2.

The effect of the organic modifier and flow rate on the repeatability.^a

Organic Modifier	Flow Rate
	1 mL/min	2 mL/min
Acetonitrile	0.0014 ± 0.0008	0.0005 ± 0.0004
Methanol	0.0030 ± 0.0019	0.0007 ± 0.0006
Tetrahydrofuran	0.0016 ± 0.0012	0.0006 ± 0.0004

^a.The median repeatability (n = 4) of all solutes and all reequilibration times used is reported. Other conditions are described in Fig. 1.

Many pumping systems allow the user to change the solvent compressibility compensation factor to optimize the eluent mixing process but the default value is often optimal for MeCN-H₂O eluents. Therefore, we attempted to optimize this parameter for MeOH-H₂O eluents at 1 mL/min (see Fig. A5). Fig. A5 shows that the default value of the compressibility compensation factor (100 · 10⁻⁶ bar) gives poor repeatability whereas other values provide similar repeatability as MeCN-H₂O and THF-H₂O eluents (see Table A2 and Fig. A5). Unfortunately, there is no clear trend of repeatability as a function of the compressibility compensation factor, which makes optimization of this parameter more difficult. However, it appears that the instrument accounts for differences in solvent compressibility more effectively at higher flow rates (in terms of repeatability; see Table A2).

Fig. A5.

Effect of the solvent compressibility compensation factor on repeatability using a MeOH-H₂O eluent (see Table 1). Solvent compressibility compensation factors of 10·10⁻⁶ bar (), 25·10⁻⁶ bar (), 40·10⁻⁶ bar (), 50·10⁻⁶ bar (), 60·10⁻⁶ bar (), 70·10⁻⁶ bar (), 80·10⁻⁶ bar (), 90·10⁻⁶ bar (), 100·10⁻⁶ bar (), 115·10⁻⁶ bar (), 130·10⁻⁶ bar () and 150·10⁻⁶ bar () were used. The median, average, maximum and minimum repeatability (n = 6) of all solutes at 1 mL/min is reported. Other conditions are described in Fig. 1.