Automatic Column Coupling System To Operate Chromatographic

Dec 30, 2009 - Chromatographic Supports Closer To Their Kinetic. Performance Limit and To Enhance Method. Development. Deirdre Cabooter,† Wim ...
0 downloads 0 Views 2MB Size
Download PDF
Anal. Chem. 2010, 82, 1054–1065

Automatic Column Coupling System To Operate Chromatographic Supports Closer To Their Kinetic Performance Limit and To Enhance Method Development Deirdre Cabooter,† Wim Decrop,‡ Sebastiaan Eeltink,†,‡ Remco Swart,‡ Mario Ursem,‡ Franc¸ois Lestremau,§ and Gert Desmet*,† Department of Chemical Engineering, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussels, Belgium, Dionex Corporation, Abberdaan 114, 1046 AA, Amsterdam, The Netherlands, and Pfizer Global Research and Development, Ramsgate Road, Sandwich, Kent, CT139NJ, U.K. A new hardware solution is proposed that allows one to automatically change the length of a chromatographic bed. The setup is based on the serial coupling of chromatographic columns using two rotor-stator valves (with N positions, N + 1 ports). Despite the use of a prototype setup requiring rather long connection tubing, only 9% loss in efficiency is observed for compounds with a retention factor above 4 compared to the efficiency expected on the basis of the individual column results. It has been demonstrated for a number of isocratic and gradient separations that the system allows one to realize considerable analysis time savings by adapting the total column length to the specific sample requirements and/ or to the stage of method development wherein one is working. During method development, a separation on a short column length can first be used to rapidly gain insight into the composition of the sample, leaving fewer runs to be done on a column of maximal length (offering efficiencies that are inaccessible with individual column systems). The ease with which information can be obtained on columns of different lengths can furthermore be exploited for screening purposes to detect coeluting components in a stage wherein they still appear completely unresolved (i.e., have a resolution well below Rs ) 0.5). Method development (MD) for high-pressure liquid chromatography (HPLC) is one of the most important and timedemanding jobs in an analytical facility. Traditionally, MD is carried out on a single column with a fixed length by varying the mobile phase composition and/or the temperature to improve the resolution of the separation.1-3 Because a large diversity of * To whom correspondence should be addressed. Phone: (+) 32 (0)2 629 32 51. Fax: (+) 32 (0)2 629 32 48. E-mail: [email protected]. † Vrije Universiteit Brussel. ‡ Dionex Corporation. § Pfizer Global Research and Development. (1) Neue, U. HPLC Columns. Theory, Technology and Practice; Wiley-VCH: New York, 1997; pp 43-48. (2) Neue, U.; Grumbach, E.; Mazzeo, J.; Van Tran, K.; Wagrowske-Diehl, D. Handbook of Analytical Separations, Bioanalytical Separations; Elsevier: Amsterdam, New York, 2003; pp 185-214.

1054

Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

separation problems exists there are, however, many cases wherein the employed column is either too long (leading to unnecessary long analysis times) or too short (leading to an incomplete separation). In a general analytical lab where methods are being developed for a large variety of different samples with unknown composition, it would therefore be very advantageous to dispose of a column system with variable column length. This would allow saving precious analysis and MD time by tailoring the column length to precisely match the separation task at hand. In the past, it has already been demonstrated on several occasions (and as early as in the 1970s4) how strong the efficiency of a separation can be increased by coupling several columns in series.5-11 With modern instruments offering up to 1000 bar of inlet pressures, coupled columns of, for example, 4 × 10 cm 1.7 µm particles, 4 × 15 cm 2.7 µm porous shell particles, and even 6 × 15 cm 3.5 µm particles, can easily be conceived of. Despite the fact that it opens the road to achieve very high efficiencies in analysis times of a few tens of minutes using existing commercial columns, the use of coupled column systems has still not found widespread use in industry. This is partly due to the fact that the coupling of columns requires some skilled manual interference, as new connection pieces need to be made every time another column is coupled to the system to ensure a good fit between the ferrules and the column. Also, this manual coupling of columns takes some time, requiring the experiments to be interrupted, and the column heating compartment to be opened in order to insert a new/additional column. In order to deal with the disadvantages that arise from manual column coupling, the present study discusses a new hardware (3) Snyder, L. R.; Glajch, J. L.; Kirkland, J. J. Practical HPLC Method Development, 2nd ed.; Wiley-Interscience: New York, 1997. (4) Scott, R. P. W.; Kucera, P. J. Chromatogr. 1979, 169, 51–72. (5) Lestremau, F.; de Villiers, A.; Lynen, F.; Cooper, A.; Szucs, R.; Sandra, P. J. Chromatogr. 2007, 1138, 120–131. (6) Lestremau, F.; Cooper, A.; Szucs, R.; David, F.; Sandra, P. J. Chromatogr. 2006, 1109, 191–196. (7) McCalley, D. V. J. Chromatogr. 2008, 1193, 85–91. (8) Cunliffe, J. M.; Maloney, T. D. J. Sep. Sci. 2007, 30, 3104. (9) Cabooter, D.; Lestremau, F.; Lynen, F.; Sandra, P.; Desmet, G. J. Chromatogr. 2008, 1212, 23–34. (10) Cabooter, D.; Lestremau, F.; de Villiers, A.; Broeckhoven, K.; Lynen, F.; Sandra, P.; Desmet, G. J. Chromatogr. 2009, 1216, 3895–3903. (11) McCalley, D. V.; Brereton, R. G. J. Chromatogr. 1998, 828, 407–420. 10.1021/ac902404v  2010 American Chemical Society Published on Web 12/30/2009

solution wherein several columns are coupled through rotor-stator valves with a custom-made connection groove pattern. Rotor-stator valves can nowadays be machined with internal volumes in the nanoliter range, ensuring that they do not lead to a significant extracolumn band broadening. Moreover, valves sustaining up to 1000 bar are available, making high-pressure operations possible. This approach allows an automatic switching between different column lengths, without the need to make new connections or arrest the experiments when switching to a different column length. As a proof-of-principle approach, two rotor-stator valves (6 + 1 central port) and a return capillary are used in the present study to automatically switch the column length between one, two, three, and four column segments, operated at a maximum pressure of 800 bar. EXPERIMENTAL SECTION Chemicals and Columns. A standard mixture consisting of acetanilide and eight alkylphenones (acetophenone, propiophenone, butyrophenone, benzophenone, valerophenone, hexanophenone, heptanophenone, and octanophenone), dissolved in a concentration of 50 mg/mL in 70/30 acetonitrile/water (ACN/ H2O) (v/v), was obtained from Dionex Corporation (Amsterdam, The Netherlands), and diuretic standards (bendroflumethiazide, benzthiazide, bumetanide, and probenecid) were obtained from Sigma-Aldrich (Steinheim, Germany). Stock concentrations of 5 mg/mL were prepared in methanol (MeOH) for each of the diuretics, and these were further diluted to 0.1 mg/mL in 75/25 MeOH/H2O (v/v). A tryptic digest of bovine cytochrome c (Dionex Corp.) was dissolved in a concentration of 0.1 mg/mL in 95/5 H2O/ACN (v/v). Uracil (Sigma-Aldrich) was used as unretained marker. Absorbance values were measured at 254 nm for the alkylphenones, at 285 nm for the diuretics, and at 214 nm for cytochrome c with a sample rate of 50 Hz. ACN, formic acid, and trifluoroacetic acid (TFA) were purchased from Biosolve (Valkenswaard, The Netherlands). All solvents were HPLC gradient grade or better. Ammonium acetate and MeOH were from Sigma-Aldrich. Water was prepared inhouse using a Milli-Q gradient (Millipore, Billerica, MA) water purification system. Acclaim rapid separation liquid chromatography (RSLC) C18 columns (2.1 mm × 100 mm) with a particle size of 2.2 µm were from Dionex Corp. (Sunnyvale, CA). The total porosity of the columns was derived from the holdup volume of the unretained marker (uracil) and was on average εtot ) 0.61. The permeability of the columns was determined with Darcy’s law.12 Apparatus and Methodology. All experiments were performed on an Ultimate 3000 system (Dionex Corp.) equipped with a high-pressure gradient pump, an autosampler, a thermostatted forced-air oven (TCC 3000 RS) with a maximum temperature of 110 °C, and a variable wavelength detector with a flow cell of 45 nL (1 cm path length). The dwell volume of the system was 600 µL, and the maximum pressure was 800 bar. Chromeleon software (12) Coulson, J. M.; Richardson, J. F. Chemical Engineering Volume 2sParticle Technology and Separation Processes; Pergamon Press: Oxford, U.K., 1999. (13) Li, J.; Carr, P. W. Anal. Chem. 1997, 69, 2530–2536. (14) Desmet, G.; Clicq, D.; Gzil, P. Anal. Chem. 2005, 77, 4058. (15) Desmet, G.; Gzil, P.; Clicq, D. LC · GC Eur. 2005, 7, 403–409. (16) Desmet, G.; Cabooter, D. LC · GC Eur. 2009, 22, 70–77.

was used for system operation and data evaluation (Dionex, Munchen, Germany). The thermostatted column compartment was equipped with two high-pressure multiposition switching valves with a pressure limit of 1000 bar (TitanHT, HT715-000) from Rheodyne (Rhonert Park, CA). The stator consisted of six peripheral ports with one central port and had a port-to-port volume of 300 nL. The rotor was custom-made in order to allow the valves to be used in six different positions. The valves were operated with the Chromeleon software. The connection tubing used to connect the columns with the switching valves was PEEKsil tubing (SGE Analytical Science, Melbourne, Australia) with a diameter of 75 µm. Each piece of tubing between valve and column had a fixed length of 300 mm. The return capillary between ports 5 and 3 of the right and left valve, respectively, had a length of 700 mm. PEEKsil tubing is resistant up to pressures of 1034 bar, in the pH range of 0-10, and has a maximum tolerance of 75 ± 3 µm on its inner diameter. To determine the possible loss associated with the use of the switching valves and tubing connecting the columns, the efficiency obtained on the different column lengths was both determined with the switching valves and by manually coupling the columns with short stainless steel connection pieces with a diameter of 127 µm and a length of 65 mm (Viper connections, Dionex Corp.), without using the valves. For these experiments a mobile phase of 75/25 ACN/H2O (v/v) was used at a flow rate of 0.25 mL/ min and a temperature of 30 °C. van Deemter data were measured for probenecid (0.1 mg/ mL dissolved in the mobile phase) on a single column at flow rates ranging between 0.01 and 0.7 mL/min. The operating temperature was 30 °C, and the mobile phase consisted of 75/25 MeOH/H2O v/v. The variances were calculated using the peak width at half height. The viscosity of the mobile phase was calculated according to Li and Carr.13 These van Deemter data were consecutively transformed into kinetic plots for a maximum pressure of 750 bar.14,15 Fixed length kinetic plot curves were constructed for column lengths of 10, 20, 30, and 40 cm by using the following expressions:16 N ) L/H

(1)

t0 ) L/u0

(2)

RESULTS AND DISCUSSION System Characterization. Schematic Setup of the “Automatic Column Coupler” (ACC). A possible strategy to couple columns and to change the column length in an automatic way is shown schematically in Figure 1. In this approach, four different column segments are coupled via two rotor-stator valves with one central and six peripheral ports. With a custom-made connection groove pattern, it becomes possible to change the total column length (ranging between one and four coupled column segments), while the different columns can still be tested individually. In the figure, the position in which the valves should be switched to obtain a certain number of coupled segments is shown just above the valves using the code Lx-Ry (L stands for left valve and R for right valve, the numbers x and y represent the number of the port to which the central port is connected). For example, to test each column individually, the valves should be positioned in the Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

1055

Figure 1. Schematic representation of some possible configurations of the column coupler: (a) single column use, (b) two coupled columns, with return capillary (red) in order to direct the flow toward the detector, (c) three coupled columns, and (d) four coupled columns. The arrows indicate the direction of the flow.

following way: L1-R1 to test column 1, L6-R2 for column 2, L5-R3 for column 3, and L4-R4 for column 4. As can also be seen from Figure 1, it is necessary that the columns can be operated in both directions of flow. Normally, this is no problem when the frits used on both the inlet and outlet have the same pore size. In some cases, however, the inlet frit has a larger pore size than the outlet frit (in order to avoid blockage at the inlet). This type of column is obviously not suited to be operated with the automatic column coupler (ACC). The columns used in this setup had identical frits on the inlet and outlet and performed equally well in both flow directions. 1056

Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

Furthermore, when an even number of column segments is to be used, it is necessary to use a return capillary (shown in red on the scheme) in order to direct the flow back to the detector. This obviously leads to an increase in total backpressure and system volume, especially when the distance between the two valves is considerable. Practical Implementation of the Automatic Column Coupler. To practically implement the ACC, the different column segments were coupled via PEEKsil tubing to the two multiposition valves of the oven compartment. Because the valves in this unit are positioned quite remote from each other (about 50 cm), a

Table 1. Permeability Values and Plate Counts Obtained for Heptaphenone (Retention Factor k = 4.0) on the Four Individual Columnsa

column column column column

1 2 3 4

plate count

permeability (m2)

1.88 × 10 1.59 × 103 1.63 × 103 1.84 × 103

7.60 × 10-15 6.70 × 10-15 6.80 × 10-15 6.80 × 10-15

3

a Shown plate counts are the average out of three different injections. Mobile phase 75/25 ACN/H2O (v/v); F ) 0.25 mL/min; T ) 30 °C.

significant length of connection tubing was necessary to connect all the columns with the valves. Although much more compact designs can be conceived of, it was preferred to work with existing, and commercially available, equipment in the present proof-ofprinciple phase. To assess the potential efficiency losses associated with the different connection pieces used in the present setup, their volumes were calculated and compared with the total volume of the columns. Taking also the contribution of the port-to-port volumes into account, the relative contribution of the tubing volume was 3.1% compared to the single-column volume, 3.2% for two columns, 2.1% for three columns, and 2.3% for four columns. The relative volume of the system does not decrease with increasing column length when considering one versus two columns and three versus four columns. This is because the twoand four-column setups require the use of a rather long return capillary (with a volume of 3.1 µL) in order to direct the flow back to the detector, increasing the total tubing volume significantly. Comparison of the Performance of Manually Coupled and Automatically Coupled Columns. To evaluate the possible loss in efficiency that can be associated with the coupling of columns, the efficiencies obtained on the columns coupled with the ACC were compared with the efficiencies obtained on a (manually) coupled column system (using short length connection pieces) as well as with the efficiency of the individual columns. The efficiency of the individual columns was determined without using the valves by injecting an alkylphenone mixture and assessing the plate counts for every compound. Table 1 shows the plate counts and permeabilities that were obtained for heptaphenone (k ) 4.0) on each individual column. The columns were subsequently manually coupled with short pieces of connection tubing (0.127 mm × 65 mm). The efficiencies were again determined using the alkylphenone mixture. Figure S-1 (see the Supporting Information) shows the efficiencies that were obtained on the manually coupled columns for the same component (red squares). These increased nearly linear with increasing column length. As a reference, the dashed line that has been added to Supporting Information Figure S-1 shows the efficiencies that can be expected if there would be no coupling losses or high-pressure effects, and assuming every column would have an efficiency equal to the average efficiency of the four columns. The deviation from this linear behavior is maximally 9% and certainly partly due to the column-to-column variabilities noted in Table 1. Possibly also some undesirable viscous heating effects on the band broadening already occur,17-19 as the four-column system is operated near the maximal pressure of the system (800 bar; all different column lengths were tested at equal flow rate). The observation that no

significant losses in efficiency result from the manual coupling of columns corresponds with other studies.6,7 Subsequently, the efficiency of the automatically coupled columns (i.e., with the two rotor-stator valves intersected as depicted in Figure 1) was determined with the same alkylphenone mixture. As can be seen from Supporting Information Figure S-1 the presence of the valves had a completely negligible effect on the efficiency for compounds with k ) 4.0 and for all column lengths, as the black triangles nearly perfectly coincide with the red squares. Fully similar observations were made for the later eluting compounds, whereas for the earlier eluting compounds (k < 4), the overall losses were somewhat larger, as these are more prone to extracolumn band broadening. For a compound with k ) 1.4, for example, a maximum loss in efficiency of 8% was observed for the ACC compared with the manually coupled setup. The connection tubing also inevitably led to an increase in system pressure. The pressure drop over the columns was compared for the manually coupled setup and the ACC setup. The ACC led to pressure drops that were 10% higher for the single column, 15% higher for the two- and three-segment setups, and 20% higher for the four coupled columns. This larger pressure drop on the ACC system is due to the large amount of narrow (75 µm i.d.) connection tubing and can drastically be reduced by using shorter connection capillaries. Some Possible Applications of the ACC System. Application 1: Adapting the Column Length to the Efficiency Requirements of the Sample. A system with automatically variable column length could be advantageous in laboratories where often the same column is used to separate samples with different complexity or difficulty. In this case, a lot of analysis time can be saved by adapting the column length in such a way that it delivers a resolution or a peak capacity that is as close as possible to the needs of the sample under consideration, and nothing more.16 This is illustrated in Figure 2, parts a and b, for the separation of two simple mixtures of diuretic compounds (sample 1 ) bendroflumethiazide, benzthiazide, bumetanide, and probenecid; sample 2 ) same as sample 1, but without bendroflumethiazide). The upper chromatograms in Figure 2, parts a and b, show the best performance obtainable on a fixed column length system (here taken as a coupled column system consisting of two 10 cm long columns), whereas the bottom chromatograms in Figure 2, parts a and b, show the best possible separations obtainable on a variable column length system. Considering sample 1 (Figure 2a), the efficiency of the two-column system is just high enough to achieve a baseline separation (Rs ) 1.5) of the critical pair when the column is operated at its optimal flow rate (F ) 0.06 mL/ min). Trying to speed up the separation of sample 1 on the fixed column length system is not possible, as any increase of the flow rate leads to a reduction of the resolution of the critical pair. However, by having the possibility to increase the column length, one can operate the column system in a kinetically more favorable regime (i.e., at a higher pressure and deeper into the C-term regime where the time needed per theoretical plate is known to be smaller than near the optimal flow rate16,20). Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

1057

Figure 2. Chromatograms obtained for a diuretic mixture on columns of different lengths. (a) Sample 1: two columns (20 cm) operated at the optimum flow rate (F ) 0.06 mL/min) (upper chromatogram) and three columns (30 cm) operated at the maximum pressure (750 bar; F ) 0.22 mL/min) (lower chromatogram). (b) Sample 2: two columns (20 cm) operated at the maximum pressure (750 bar, F ) 0.30 mL/min) (upper chromatogram) and one column (10 cm) operated at the maximum pressure (750 bar, F ) 0.65 mL/min) (lower chromatogram). The compounds are for sample 1 (1) bendroflumethiazide, (2) benzthiazide, (3) bumetanide, (4) probenecid and for sample 2 (1) benzthiazide, (2) bumetanide, (3) probenecid. The mobile phase was 75/25 v/v MeOH/H20 (with 0.1% of formic acid), the operating temperature was 30 °C, and injection volume was 0.5 µL.

This is shown in the bottom chromatogram of Figure 2a, where a slightly better critical pair resolution (Rs ) 1.7) is obtained in less than half of the time needed with the fixed column length system. 1058

Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

If now a less complex sample needs to be separated (as is the case for sample 2), the advantage of a system that can readily switch to a shorter column length is even more straightforward. Whereas the fixed length system still needs 4 min to complete

the separation at the maximal pressure (Figure 2b, upper chromatogram), a time that cannot be reduced since the system is already operated at maximal pressure, the variable column length system can complete the separation in less than 1 min when applying the same maximal pressure and switching to its one segment position (Figure 2b, bottom chromatogram). In this case, a factor of 4 in analysis time is gained since only half of the column length is required and it can be operated at a flow rate that is twice as large. The kinetic advantage of a system at hand that can readily switch between different column lengths can be represented in a more general way by looking at the fixed length (black curves) and free length (red enveloping curve) kinetic plot curves shown in Figure S-2 (see the Supporting Information). These were established on the basis of an experimental van Deemter curve measured on a single column for one of the four components in the diuretics mixture (probenecid). Despite the fact that this component is quite specific and related to the presently considered application, the shape of the resulting curves is characteristic for any kind of chromatographic system displaying a normal van Deemter curve (examples: Figure 5 of Desmet et al.21 or Figures 2-4 of Desmet and Cabooter16). Supporting Information Figure S-2 clearly shows that the availability of more than one column length dramatically broadens the range of efficiencies wherein the available packing type is operated close to its kinetic performance limit. The latter is given by the free length kinetic plot (red curve) and represents the lower analysis time limit for any possible value of the required efficiency. Arbitrarily setting the criterion for a well-optimized separation such that it should lie within less than a factor of 1.5 (or 50%) of this kinetic performance limit, Supporting Information Figure S-2 shows that, whereas the 20 cm fixed column length system only satisfies this criterion in the N ) 12 000 to the N ) 21 000 plate range, the variable column length system has an operational range going from N ) 3500 to N ) 46 000 wherein it can offer separation times that lie within 50% of the theoretical minimum (corresponding to the red kinetic plot curve shown in Supporting Information Figure S-2) of the employed particle type. Application 2: Detection of Coeluting Peaks. Another application that comes within direct reach with an ACC system is the possibility to compare the evolution of the separation quality with the column length as a means to reveal additional information about the composition of the sample. The most significant piece of additional information one could obtain by switching to other (i.e., longer) column lengths would of course be the emergence of one or more additional peaks in the chromatogram. However, the increase in efficiency that is needed to split one peak in two or more is very high (it takes, for example, a 16-fold increase in efficiency to go from a resolution of Rs ) 0.2 to Rs ) 0.8) so that this event will be rather rare (at least as long as the change in column length is not accompanied by changes in selectivity, see further below). However, one does not need a peak that actually splits to detect a coeluting doublet or triplet of (17) (18) (19) (20) (21)

Neue, U.; Kele, M. J. Chromatogr., A 2007, 1149, 236–244. Desmet, G. J. Chromatogr., A 2006, 1116, 89. Guiochon, G. J. Chromatogr., A 2006, 1126, 6. Poppe, H. J. Chromatogr., A 1997, 778, 3–21. Desmet, G.; Clicq, D.; Nguyen, D. T.-T.; Guillarme, D.; Rudaz, S.; Veuthey, J.-L.; Vervoort, N.; Torok, G.; Cabooter, D.; Gzil, P. Anal. Chem. 2006, 78, 2150–2162.

components. When assuming that a peak consisting of two overlapping compounds (respectively, with retention times tR1 and tR2 and variances σ12 and σ22) has a total width of wp ) (tR2 - tR1) + 2σt2 + 2σt1

(3)

and assuming that σ1 ≈ σ2, both approximately corresponding to an efficiency of Ntheor, the observed efficiency of this doublecomponent peak (Nobs) will be given by

Nobs )

tR2 σt22

)

t02(1 + k)2 k2 - k1 t0 t0 + (1 + k) 4 √Ntheor

[(

]

)

2

(4)

Assuming further that tR2 ≈ tR1 ≈ tR and k2 ≈ k1 ≈ k, the ratio of the efficiencies of a monocomponent single peak (Ntheor) and the double-component peak (Nobs) is then given by Ntheor ) Nobs

[

√Ntheor 4

]

k (R - 1) +1 1+k R

2

) [Rs + 1]2

(5)

wherein Rs is the resolution of the coeluting pair. Subsequently solving eq 5 for Rs then finally yields

Rs ) -1 +



Ntheor Nobs

(6)

Equation 6 shows that if it would be possible to detect a deviation of 20% between the expected and the actually obtained plate number with a sufficient confidence (Nobs ) 0.8Ntheor), the presence of a peak doublet could already be detected from a resolution of Rs ) 0.12 on. If it would only be possible to significantly detect a deviation of 40%, one would still be able to detect a coeluting pair with a Rs of 0.29. And for a deviation of 50%, the critical Rs is 0.41. These resolution values are significantly lower than the minimum resolution that is required to see two peaks split (Rs ) 0.50). To illustrate how the aforementioned method works, Figure 3 shows the chromatograms of the separation of an alkylphenone mixture on different column lengths. First, using a single column (Figure 3a), the mixture could be resolved into nine different compounds. Keeping all conditions of mobile phase composition and flow rate the same, and only changing the column length to two coupled columns, the chromatogram in Figure 3b is obtained. This chromatogram still has nine peaks but now provides the necessary information to scout for anomalies in the evolution of the band broadening with the column length. This information is represented in Figure 4, showing a plot of observed over theoretically expected plate number (Nobs,i/Ntheor,i), with Ntheor,i being the expected plate number for a certain component i on two columns and calculated as 2 times the Nobs,i value obtained for component i on one column. Although most peaks, and especially the early eluting peaks, display a clear deviation from unity (a deviation that is readily explainable from the fact that the two-column system is less prone to extracolumn band broadening than the single-column system, which holds especially for the early eluting peaks), the plot clearly Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

1059

Figure 3. Chromatograms obtained for an alkylphenone mixture on (a) a single column and (b) two columns coupled with the column coupler: flow rate ) 0.25 mL/min; mobile phase, 90/10 v/v ACN/H2O; detection wavelength, 245 nm; injection volume, 0.5 µL. The retention times and plate counts of the different compounds are shown.

reveals a strong deviation for peak no. 5, where Nobs,i only reaches some 60% of the expected value, whereas the other peaks all have an Nobs,i/Ntheor,i value lying above unity. The fact that peak no. 5 indeed consists of a doublet pair could be confirmed by further increasing the column length from 2 to 4 units (leading to L ) 40 cm) where now indeed peak no. 5 splits and reveals the presence of a second component (Figure S-3, see the Supporting Information). Obviously, the use of anomalies in peak width as a means to detect the presence of coeluting peaks will not work as efficiently 1060

Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

when the two coeluting components have a significantly different contribution to the detected signal (as, for example, encountered in pharmaceutical impurity profiling applications). To improve the detectability of such coeluting pairs, one could resort to measuring the peak width at 4σ or even 5σ instead of at half height as was done in the present example, but even this approach will eventually have its limitations (as eventually any chromatographic system or technique). The proposed method is most reliable if the peak widths and peak heights on the different column length systems are compared

Figure 4. Ratio of observed over theoretically expected efficiencies for each of the analytes in the alkylphenone mixture. The dashed line corresponds to Nobs,i/Ntheor,i ) 1. See text for calculation of Nobs,i.

at constant flow rate and under identical mobile phase conditions (so as to maintain the retention factor constant). This condition is most stringent under gradient conditions, where one should not only keep the same gradient steepness tG/t0 (t0 increases proportionally with the total column length at constant flow rate; hence, tG needs to be multiplied with the same factor) but also the dwell time has to be adapted to the column length under consideration (this was applied in the example discussed in Figures 5 and 6 shown further on). Another interference might occur when the selectivity slightly varies between the different column length systems. These selectivity changes can be caused by column-to-column variabilities, although state-of-the-art columns are nowadays reliable in terms of selectivity. Another source of selectivity changes is the use of very high pressures.22 In the coupled column system, these high pressures are inevitably needed to operate the longest column sequence at the same flow rate as the shorter column sequence. A slight inaccuracy in the dwell-volume correction might also cause shifts in selectivity. As these shifts in selectivity can either bring the two peaks in a coeluting pair closer to each other or further apart, the small changes in selectivity that might accompany the change in column length might magnify, but also moderate or even reverse, the observed peak broadening anomalies. The latter implies that an Nobs,i/Ntheor,i value lying well above unity (for example, reaching a value of 2 or more, see Figure 6 further on) can also be considered as an indication of component coelution. Although the above arguments imply that the conclusions drawn from the comparison of the peak width at different column lengths are not full-proof, the test anyhow provides an important “warning sign” for the occurrence of coelution of overlapping peaks that cannot be obtained when separations are only optimized on a fixed column length system. This information can then be used in subsequent runs, preferably invoking new selectivities. (22) Fallas, M. M.; Neue, U. D.; Hadley, M. R.; McCalley, D. V. J. Chromatogr., A 2008, 1209, 195–205.

Similar information can also be retrieved by studying the evolution of the peak heights. A comparison of the chromatograms in parts a and b of Figure 3 clearly illustrates this, as peak no. 5 clearly undergoes a much stronger height decrease than the other peaks. Of course, peak heights are more subject to variability and less sensitive to the presence of a low-concentration coeluting component than, for example, the peak width at 5σ. Nevertheless, it certainly makes sense to include the relative peak height variation as an additional source of information to confirm observed anomalies in the efficiency variation and hence to increase the confidence with which coeluting component pairs or triplets can be detected. Application 3: Enhancing Method Development. By combining the advantages discussed in the two previous sections, the possibility to automatically change the efficiency can be used to enhance the MD of samples with an unknown composition. To investigate this, the separation of a tryptic digest of cytochrome c has been studied. In this application, the ACC was used under gradient conditions. In this case, a direct comparison between chromatograms obtained on columns with different lengths is only possible when the components have the same effective retention factor. To ensure this, one has to modify the gradient time tG so that the same tG/t0 is maintained on all column lengths. In addition, an extra delay time needs to be introduced for each extra column that is added, to ensure that the gradient reaches the sample compounds at exactly the same relative distance in the column sequence. Practically, the additional dwell time that needs to be added is equal to the product of the system’s dwell time and the number of column segments minus one. Two cases have been compared: fixed length MD and variable length MD. In both cases, no a priori knowledge about the composition of the sample was used. The sample was resolved using an MD strategy wherein combinations of two pHs (pH ) 2.1 and pH ) 6.8) and two gradient composition ranges (∆φ ) φstart - φend ) 0.95 - 0.05 ) 0.90 and ∆φ ) 0.95 - 0.13 Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

1061

Figure 5. Separation of a cytochrome c mixture on (a) one column and a gradient time of 10 min and (b) four columns and a gradient time of 40 min. The mobile phase concentration was varied from 95/5 buffer/ACN (v/v) to 5/95 buffer/ACN (v/v) at a flow rate of 0.2 mL/min. The pH of the buffer was 2.1 (H2O with 0.05% TFA).

) 0.82) were used for a gradient time of tG ) 10 min. The ∆φ ) 0.90 case was also applied for both pHs and a gradient time of tG ) 45 min. The fixed length MD was carried out without column coupling, i.e., on a single-column system. Figure S-4 (see the Supporting Information) shows the best separation that was obtained on the single-column system (pH ) 2.1, ∆φ ) 0.9 and tG ) 45 min). Under these conditions, the sample is resolved in 12 main peaks, spread over a sufficiently broad elution window running from an effective retention factor k ) 4 to k ) 17 (with k defined as k ) (tR - t0)/t0). The total analysis time was about 18.5 min. The variable column length MD started with the same set of tG ) 10 min scouting runs as used in the fixed length MD case. After this initial set of scouting runs (for which the case with 1062

Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

pH ) 2.1 and ∆φ ) 0.9 yielded the broadest elution window and the largest number of peaks), the four-column system (total length ) 40 cm) was used to run two tG ) 40 min runs at different pH values, starting with the gradient conditions (pH ) 2.1 and ∆φ ) 0.9) that gave the best separation in the scouting runs on the short column system (Figure 5a). Doing so, the increased efficiency of the four-column system readily revealed (see Figure 5b) a 12th and a 13th component (i.e., one more than in the fixed column length MD), as peaks d and i each split in two. When comparing the resolution of the two new critical pairs d, e and i, j, it can be deduced from Figure 5 that peak i splits to a much larger extent (from unresolved to Rs = 2.0) than peak d (from unresolved to Rs = 0.65). It is therefore very likely that peak i has mainly split due to changes in selectivity (most

Table 2. Total Analysis Time Needed To Resolve the Largest Possible Amount of Peaks on the Column Length(s) under Considerationa MD system

individual analysis times (min) ) ) ) ) ) )

) ) ) ) ) )

) ) ) ) ) )

pH pH pH pH pH pH

min min min min min min

190

12

variable column length (equilibration time ) 40 min for a 40 cm column) (additional dwell time ) 9 min)

same as fixed column length + pH ) 2.1, ∆φ ) 0.90, tG ) 40 min pH ) 6.8, ∆φ ) 0.90, tG ) 40 min

368

13

∆φ ∆φ ∆φ ∆φ ∆φ ∆φ

0.90, 0.90, 0.82, 0.82, 0.90, 0.90,

tG tG tG tG tG tG

10 10 10 10 45 45

no. of peaks resolved

fixed column length (single column) (equilibration time ) 10 min per run)

a

2.1, 6.8, 2.1, 6.8, 2.1, 6.8,

total analysis time (min)

Flow rate ) 0.2 mL/min; sample ) cytochrome c digest; T ) 30 °C.

probably caused by temperature changes due to viscous heating occurring because of the high pressure needed in the four-column system) and not just due to an increase in efficiency. In the present example, the MD was hence helped by an unexpected change in selectivity between the one- and the four-column systems. The separation into 13 main components was, however, perfectly reproducible. Further changing the pH or ∆φ on the four-column system did not reveal any additional peaks. Table 2 compares the total analysis time used in the singlecolumn case (fixed column length MD) with that used in the variable column length MD case. Obviously the total time is shorter on the fixed column length case, which is partly due to the larger equilibration times that are needed for longer columns, but this does not weigh up against the fact that the variable column length system revealed the presence of a 13th peak. Of course, one could argue that the 13 peaks would also have been revealed with a fixed column length system consisting of four columns. This is certainly true, but the total MD time would then have been 4 times longer than in the single-column system, considering the analyses and the equilibration would take 4 times longer, and would hence have amounted up to 760 min (4 × 190 min), which is twice as large as in the variable column length MD case. The above illustrates the advantage of the flexibility of the variable column length system, as one can run all basic scouting runs (in which gradient range, gradient time, and pH are tuned to obtain a sufficiently large elution window) on the short column system, hence minimizing the time needed for these scouting runs. The best conditions can then be applied in the high-efficiency system, where the analysis times are longer but where one can benefit from the increased efficiency to further refine the separation. In the present example, only the pH and the mobile phase starting composition have been changed, but it goes without saying that also the use of ternary mixtures or changes in ionic strength and temperature could be included in the initial scouting runs. Obviously, the use of an ACC system will not make the use of computer simulation software23-30 that is typically used in MD superfluous. It will simply help generating more information about (23) Krisko, R. M.; McLaughlin, K.; Koenigbauer, M. J.; Lunte, C. E. J. Chromatogr., A 2006, 1122, 186–193. (24) Jupille, T.; Snyder, L.; Molnar, I. LC · GC Eur. 2002, 15, 2–6. (25) Molnar, I. J. Chromatogr., A 2002, 965, 175–194.

the sample, especially useful when dealing with samples of unknown composition. After having been reassured that none of the other combinations of pH and gradient times led to more than 13 peaks, we also used the technique discussed in Application 2 to investigate whether the peaks that were present in both the single- and the four-column system did not mask any other coeluting compounds. To collect a maximal amount of information, an additional run on the two column length system (20 cm) was conducted as well. Figure 6 shows how the obtained Nobs,i/Ntheor,i values vary as a function of the component number for the three possible length combinations (4/1, 2/1, and 4/2). Anomalies are clearly present for peaks d and i, and these are precisely the two peaks that split when going from one to four columns (compare parts a and b of Figure 5). This information was, however, already visible from the chromatograms themselves so that the most important information that is obtained from Figure 6 is that the Nobs,i/Ntheor,i ratios for peaks b and c also deviate significantly from unity. This means that peaks b and c can be suspected of peak overlap and should therefore be analyzed in more detail, preferably by switching to other selectivities. Another observation from Figure 6 is that peaks a, f, g, h, k, l, m display a normal bandbroadening pattern. This still does not exclude that these peaks consist of multiple components, but it nevertheless allows us to conclude that it will never be possible to split these peaks with the presently employed stationary phase, certainly not in a practically feasible analysis time. Having a Nobs,i/Ntheor,i ratio close to unity, one can, namely, be certain that, if these peaks would be composed out of more than one component, the separation resolution certainly lies below Rs ) 0.1, even at the highest employed efficiency of the four-column system (offering roughly 60 000 theoretical plates under isocratic conditions). Further MD work on the same phase can hence be considered as futile. (26) Xiao, K. P.; Xiong, Y.; Liu, F. Z.; Rustum, A. M. J. Chromatogr., A 2007, 1163, 145–156. (27) Hewitt, E. F.; Lukulay, P.; Galushko, S. J. Chromatogr., A 2006, 1107, 79– 87. (28) Euerby, M. R.; Petersson, P.; Campbell, W.; Roe, W. J. Chromatogr., A 2007, 1154, 138. (29) Subirats, X.; Bosch, E.; Roses, M. J. Chromatogr., A 2006, 1121, 170–177. (30) Fekete, S.; Fekete, J.; Molnar, I.; Ganzler, K. J. Chromatogr., A 2009, 1216, 7816–7823.

Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

1063

Figure 6. Nobs,i/Ntheor,i ratios for the compounds in the cytochrome c mixture: [ Nobs,i ) plate counts on four columns, Ntheor,i ) 4 times the plate counts on one column; 9 Nobs,i ) plate counts on four columns, Ntheor,i ) 2 times the plate counts on two columns; 2 Nobs,i ) plate counts on two columns, Ntheor,i ) 2 times the plate counts on one column. The dashed line corresponds to Nobs,i/Ntheor,i ) 1.

In a final step, and using the newly obtained information that the sample contained 13 main compounds instead of only 12, the column length and gradient times were further optimized to improve the separation. Doing so, the best possible separation was obtained on a column length of 30 cm using a gradient time of 20 min (Figure S-5, see the Supporting Information). In this case, all peaks were baseline-separated in an analysis time of 14 min. CONCLUSIONS An automatic column coupling system, offering the possibility to change the number of involved segments, can be realized using two multiposition valves and a specific rotor design. This possibility brings ruggedness and automation to the process of column coupling and can bring the major advantage of coupled column systems (i.e., their increased efficiency) within reach of daily practice. Although efficiency can certainly not replace selectivity to enhance MD or the speed of routine analysis, the possibility offered by a variable column length system to change the efficiency over a broad range of values is a key asset in finding the minimal analysis time solution and can also be used to gain additional information about the sample by switching to column lengths that are not immediately accessible when using a singlecolumn system. In addition, the flexibility of a variable column length system is also ideally suited to speed up MD. Performing the initial scouting gradients (initial and final gradient composition, gradient range, gradient steepness, pH, ionic strength, temperature) on the shortest segment, a wealth of information about the sample can be obtained in the shortest possible time. Subsequently switching to the longest available column length, the use of the mobile phase and gradient parameters yielding the maximal number of resolved peaks on the shorter column length then allows us to use the available high efficiency to check for the presence of additional components and/or potentially coeluting components. This can be done by tracing 1064

Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

for splitting peaks (resolution above Rs ) 0.5) in the chromatograms obtained at maximal efficiency or by tracing for anomalous changes in peak width or peak height when comparing chromatograms obtained on the different lengths (allowing us to detect coeluting peaks at a resolution well below Rs ) 0.5). The additional information about the composition of the sample that is obtained in this way can then be used to further optimize the separation on the maximal column length system (by further changing the mobile phase composition, the temperature, and the gradient parameters) or (when the sample is deemed to be completely resolved) by reducing the number of segments again to perform the separation closer to the kinetic optimum. These advantages are achieved at the expense of an increased cost (price of columns and investment in two additional rotor-stator valves). The possibility of running many different condition runs in a preprogrammed sequence, however, makes up for this. Furthermore, if one decides to benefit from the large efficiency offered by coupled column systems, one anyhow needs to buy the full set of columns. In comparison to manually coupled column systems, the rotor-stator coupling method also offers the possibility to regularly change the column that is subjected to the highest pressure, thus allowing one to increase the lifetime of the columns. To include selectivity one could use different automatically coupled column systems in parallel, one for each type of stationary phase. Evidently, it is also possible to combine different phases in the same train of column segments. The wealth of opportunities offered by this type of system will be investigated in a future study. A moderating remark concerning the shown applications is that the proposed method to detect coeluting pairs will be much less efficient when the coeluting pair has strongly differing areas. It should also be realized that gathering MD information on systems with different lengths might be troubled by the

fact that changes in length might induce changes in selectivity (different column lengths, for example, require different pressures, and pressure differences might lead to selectivity changes). These selectivity changes can either work at one’s disadvantage or at one’s advantage. In the latter case, it helps reveal the presence of additional compounds, as was the case for peaks i and j in the example in Application 3.

for the kind donation of the columns and the Rheodyne valves.

ACKNOWLEDGMENT D.C. is a postdoctoral fellow of the Research Foundation, Flanders (FWO Vlaanderen). Dionex is acknowledged

Received for review December 7, 2009.

SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.

October

23,

2009.

Accepted

AC902404V

Analytical Chemistry, Vol. 82, No. 3, February 1, 2010

1065