Journal of Chromatography A, 1403 (2015) 132–137
Contents lists available at ScienceDirect
Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma
Understanding and diminishing the extra-column band broadening effects in supercritical fluid chromatography Ruben De Pauw a , Konstantin Shoykhet (Choikhet) b , Gert Desmet a , Ken Broeckhoven a,∗ a b
Vrije Universiteit Brussel, Department of Chemical Engineering (CHIS-IR), Pleinlaan 2, 1050 Brussels, Belgium Agilent Technologies Europe, Hewlett-Packard-Strasse 8, 76337 Waldbronn, Germany
a r t i c l e
i n f o
Article history: Received 31 March 2015 Received in revised form 7 May 2015 Accepted 7 May 2015 Available online 19 May 2015 Keywords: Supercritical fluid chromatography Detector cell Extra-column band broadening Extra-column volume Sample solvent Injection volume
a b s t r a c t Supercritical fluid chromatography, where a low-viscosity mobile phase such as carbon dioxide is used, proves to be an excellent technique for fast and efficient separations, especially when sub-2 m particles are used. However, to achieve high velocities when using these small particles, and in order to stay within the flow rate range of current SFC-instruments, narrow columns (e.g. 2.1 mm ID) must be used. Unfortunately, state-of-the-art instrumentation is limiting the full separation power of these narrower columns due to significant extra-column band broadening effects. The present work identifies and quantifies the different contributions to extra-column band broadening in SFC such as the influence of the sample solvent, injection volume, extra-column volumes and detector cell volume/design. When matching the sample solvent to the mobile phase in terms of elution strength and polarity (e.g. using hexane/ethanol/isopropanol 85/10/5 vol%) and lowering the injection volume to 0.4 L, the plate count can be increased from 7600 to 21,300 for a low-retaining compound (k = 2.3) on a 2.1 mm × 150 mm column (packed with 1.8 m particles). The application of a water/acetonitrile mixture as sample solvent was also investigated. It was found that when the volumetric ratio of water/acetonitrile was optimized, only a slightly lower plate count was measured compared to the hexane-based solvent when minimizing injection and extra-column volume. This confirms earlier results that water/acetonitrile can be used if water-soluble samples are considered or when a less volatile solvent is preferred. Minimizing the ID of the connection capillaries from 250 to 65 m, however, gives no further improvement in obtained efficiency for early-eluting compounds when a standard system configuration with optimized sample solvent was used. When switching to a state-of-the-art detector design with reduced (dispersion) volume (1.7–0.6 L), an increase in plate count is observed (from 11,000 to 14,000 plates on a 2.1 mm × 100 mm column with 1.8 m particles for k = 3) even when 250 m tubing was used. Using this detector cell and decreasing the ID of the tubing from 250 to 120 m resulted in an additional increase to 17,300 plates. Further decreasing the tubing ID (e.g. 65 m) appeared to have no observable influence on the obtained plate count. © 2015 Elsevier B.V. All rights reserved.
1. Introduction In order to achieve fast and efficient separations in chromatography, it is important to decrease the resistance for mass transfer. Among several approaches, using a low-viscosity fluid, such as fluidic CO2 , results in very high diffusion coefficients and thus allows the use of higher flow rates without significant efficiency loss [1]. As a result, very fast and efficient separations can be performed in supercritical fluid chromatography (SFC) [2–7]. For example, a plate count of 182,000 plates can be achieved for a void time of 5.8 min
∗ Corresponding author. Tel.: +32 26293781; fax: +32 26293248. E-mail address: [email protected] (K. Broeckhoven). http://dx.doi.org/10.1016/j.chroma.2015.05.017 0021-9673/© 2015 Elsevier B.V. All rights reserved.
when the pressure limitation of state-of-the-art instrumentation (600 bar) is considered [8]. Although this makes carbon dioxide an attractive solvent for chromatography, working with a highly compressible mobile phase has some drawbacks of its own. First of all, a back pressure above the critical pressure (72 bar) is required to avoid the formation of gas bubbles in the detector cell. Typically, even higher back pressures are used (120–150 bar) to operate in a region sufficiently far from the critical point, i.e. far from the region of high compressibility of the mobile phase. Otherwise, the large changes in mobile phase density can cause extensive decompression cooling which results in a loss in performance and significant variation in velocity and retention along the column occurs due to the large variation in density [9–13]. Another consequence is the inability to
R. De Pauw et al. / J. Chromatogr. A 1403 (2015) 132–137
precisely match the sample solvent with the mobile phase, as this would require the sample to be compressed in a CO2 -based mobile phase. Depending on the column ID and the injected volume, this mismatch between sample solvent and mobile phase may lead to significant extra-column band broadening especially in preparative scale separations [16]. Several studies highlighted the effect of sample solvents on performance in SFC but even with this knowledge, issues with extra-column band broadening still remain for narrow columns. In general it was found that apolar solvents such as heptane allow to minimize peak distortion, whereas another study found that acetonitrile in combination with water could be used if water-soluble samples are considered [14,15,17,18]. A recent investigation of Perrenoud et al. showed that, due to the significant extra-column variance in SFC, the optimal column geometry would be a 3 mm ID column in the case of small particles and short columns. This ID would be a perfect trade-off between minimizing extra-column band broadening effects and still being able to achieve sufficiently high linear velocities for e.g. sub-2 m particles [6]. A previous study about speed-resolution limits in SFC also illustrated important extra-column band broadening effects for early-eluting compounds on short columns with a 2.1 mm ID format, which were largely resolved when using very long columns (e.g. 500 mm) [8]. Nonetheless, the origin and relative contributions of extracolumn band broadening in SFC are not entirely understood or quantified even though it follows the same basic principles as e.g. in liquid chromatography (LC). The potential sources are: • Sample solvent: even though the use of hexane/ethanol (EtOH)/isopropanol (IPA), where IPA is added to allow better mixing of EtOH and hexane [6], as a sample solvent provides a reasonable viscosity and retention match with the CO2 /methanol (MeOH) mobile phase, it is not perfect and may still lead to a distortion of the sample band at the column inlet. In addition, a water (H2 O)/acetonitrile (ACN) solvent will also be assessed since it may be of practical use when testing water-soluble samples [17]. • Injection volume: SFC-injectors generally work via the fixed-loop principle where the injection loop is generally large (5 or 10 L). If partial-loop injection is considered, the rest of the loop is generally filled with the mobile phase co-solvent (or another wash solvent). At the moment of injection, a large amount of mismatching solvent hence enters the column with the sample, which may lead to peak distortion. • Extra-column volumes: in SFC, generally long tubing with higher ID’s are used (170 m instead of the more common 120 or even 75 m in LC), which could lead to severe extra-column band broadening. In addition, the geometry of the extra-column volumes (e.g. detector design) may also lead to an additional contribution due to the presence of stagnant zones. The present work investigates and quantifies the relative contributions of these potential band broadening sources for current state-of-the-art SFC instrumentation and explores some new ways to reduce efficiency losses for low-retained compounds. 2. Experimental 2.1. Column, tubing and chemicals Methanol, hexane and isopropanol (LC-MS grade) were purchased from Biosolve (Valkenswaard, Netherlands), CO2 was purchased from Air Liquide (Paris, France). Test components such as testosterone, chlorthalidone, bendroflumethiazide, altizide and -estradiol were kindly provided by Deirdre Cabooter (Laboratory
133
of Pharmaceutical Analysis, KU Leuven, Belgium). The samples were dissolved in a mixture of ethanol (EtOH), isopropanol (IPA) and hexane (except for the assessment of the water/acetonitrile solvent). The IPA was added to allow mixing of EtOH with hexane [6]. In order to inject different sample volumes, a 20 cm fused-silica 50 m capillary, a 7 cm, 120 m and a 11 cm, 170 m stainless steel capillary were used, to achieve, respectively, a 0.4, 0.8 and 2.5 L injection volume (instead of the 5 L sample loop intended for the standard configuration of the instrument). All the reported extra-column volumes are calculated, based on the nominal dimensions given by the manufacturers. These nominal dimensions were evaluated by measuring the total extra-column volume under LC conditions and a deviation of maximally 15% was found. 2.2. Instrumentation and conditions The SFC-system used in the study was a modified Agilent G4301A-based SFC system in combination with a thermostatted column compartment, autosampler used in full-loop mode and two flow cells with a dispersion volume of 1.7 L (DAD G1315C with a G1314-60082 flow cell) and 0.6 L (Agilent 1290 Infinity DAD G4212A with a G4212-60038 flow cell). Zorbax HILIC RRHD columns (2.1 mm ID, 150 mm and 100 mm, 1.8 m fully porous particles) were used in the current study. The columns were kindly provided by Xiaoli Wang (Agilent Technologies, Little Falls, USA). The oven temperature was set at 40 ◦ C, 8 v% methanol as modifier was used and the back pressure was set at 150 bar, unless otherwise specified. The detector was set at an acquisition rate of 100 Hz and a wavelength of 230 nm. Due to the high acquisition rate, the observed peak widths and variances were not affected, even for very the narrow, early-eluting peaks. The shown data are average values of 3 repeats and a maximum relative standard deviation for the plate count of 6% was found. 3. Results and discussion 3.1. Possibilities with current state-of-the-art SFC instrumentation The first goal is to investigate the possibilities to reduce peak broadening in state-of-the-art SFC systems. For this purpose we have • varied the sample composition. • varied the injection volume by using 0.4 and 2.5 L loops. • varied pre- and post-column volumes by choosing smaller ID tubing or switching the preheater (e.g. 1.6 L instead of 3 L preheater intended for the standard configuration of the used instrument). 3.1.1. Sample solvent and extra-column volume To study the influence of extra-column volume, only the volume before the column (extra-precolumn volume) was changed by: changing the preheater (1.6 or 3 L) and the connection capillary from injector to preheater (120 or 170 m). This resulted in two different extra-precolumn volumes of 4.4 L (20 cm, 120 m connection capillary from injector to preheater, 1.6 L preheater) and 7.4 L (17 cm, 170 m connection capillary, 3 L preheater). The same 5 cm long 120 m connection capillary from preheater to column was kept for both configurations. The injection volume was varied by using either the 0.4 or the 2.5 L loop in the overfill mode. Whereas commercial SFC-systems are typically supplied with a 5 L-loop, smaller loops can be used but it is not recommended on a regular basis as it may lead to system errors during the washing or filling phase due to an increased
134
R. De Pauw et al. / J. Chromatogr. A 1403 (2015) 132–137
Fig. 1. Apparent plate count as function of sample solvent concentration (volumetric EtOH-concentration, , mixed with hexane and IPA), with the EtOH/IPA ratio always equal to 2, for different injection volumes (Vinj ) and extra-precolumn volumes (VEC ): (a) testosterone, low-retaining compound (k = 2.3), (b) Altizide, high-retaining compound (k = 10.8). Measured on a 2.1 × 150 mm Zorbax HILIC RRHD column (packed with 1.8 m particles) at a flow rate of 1 mL/min.
pressure drop. In addition, a 50 m ID tubing generates a significant pressure drop during the separation which may result in a cyclic load on the system when switching the injection valve. However, for this study on injection volumes in SFC, it provided very low and reliable injection volumes. The last important parameter, the composition of the sample solvent, was studied as sample solvents in SFC are never matched to the mobile phase (in terms of viscosity and retention matching) as it is impractical to use a compressed CO2 /MeOH mixture [14,15]. To investigate the effect of sample solvent, the volumetric fraction of EtOH was changed while keeping the volume ratio EtOH/IPA equal to 2. This was achieved by using the following sample solvents: 10% EtOH (with 5% IPA and 85% hexane), 20% EtOH (with 10% IPA and 70% hexane) and 40% EtOH (with 20% IPA and 40% hexane) [6]. In Fig. 1 the apparent plate counts are given for a low and high-retaining compound (k = 2.3 or 10.8). For the low-retaining compound (Fig. 1a), variation of the extra-precolumn volume between 4.4 and 7.4 L ( versus ) did not influence the observed plate count when using an injection volume of 2.5 L. The effect of the injection volume, however, is large for high EtOH-fractions (20% and above) ( versus ). Increasing the EtOH-content to 40% results in a significant decrease in plate count for the three different cases and the effect is most pronounced for the largest injection volume. When decreasing the EtOH-content, the plate count for the smallest injection volume ( ) levels off to a maximum value. The two other cases also appear to converge towards this limit at low EtOH-content. This leveling off indicates that a further reduction of the EtOH-fraction will not improve the performance and that the
reduction of other extra-column band broadening contributions (such as the detector cell) is required. For the high-retaining compound, Fig. 1b, a similar behavior is observed but the absolute values of the plate count are higher since extra-band broadening is less significant for higher retaining compounds eluting in broader peaks. However, an additional effect can now be seen for 40% EtOH. It appears that increasing the precolumn volume leads to an improvement in plate count. Whereas with the small extra-precolumn volume ( ) an efficiency of only 8000 plates is observed, the plate count increases to 14,000 if the precolumn volume is increased to 7.40 L (standard deviation for these measurements is equal to 300 plates). This suggests that at high EtOH-content, the higher extra-precolumn volume allows a better mixing of the sample with the mobile phase and results in a more compatible sample solvent when entering the column. In Fig. 2, a similar study is performed for a H2 O/ACN sample solvent. In both configurations, the same 120 m tubing from injector to preheater was used and only the injection volume and preheater volume were changed (e.g. extra-precolumn volume equal to 4.4 L and 5.8 L when using a 1.6 and 3 L preheater respectively). From Fig. 2b, it is clear that decreasing the injection volume significantly improves the performance. For larger injection volumes and a strongly retained compound, the efficiency was highest with a fraction of 40% water in the sample solvent. For the lowretaining compound (Fig. 2a) no clear trend with ACN volume fraction could be observed and the curves for the two injection volumes even cross, resulting in an apparent maximum at 20% water for the 0.4 L injection and 40% for the 2 L injection. Near this maximum, nearly the same performance is found as for the
Fig. 2. Apparent plate count as function of sample solvent concentration (volumetric H2 O-concentration, , mixed with ACN) for two different injection volumes (Vinj ) and extra-precolumn volumes (VEC ): (a) testosterone, low-retaining compound (k = 2.2). (b) altizide, high-retaining compound (k = 12.1). Same column and flow rate as in Fig. 1.
R. De Pauw et al. / J. Chromatogr. A 1403 (2015) 132–137
Fig. 3. Influence of post-column tubing ID (65 ( ), 170 ( ) and 250 m ( )) on extra-column band broadening at different flow rates: 1 mL/min (full symbols) and 2 mL/min (open symbols). The experiments were performed on 2.1 × 100 mm Zorbax HILIC column, back pressure at 150 bar, injection volume of 0.8 L with a sample solvent of 20/10/70 EtOH/IPA/hexane. The used compounds (in order of elution) are aspirine, testosterone, -estradiol, bendroflumethiazide and chlorthalidone.
optimized EtOH/IPA/hexane-solvent (20,000 instead of 21,000 plates). The occurrence of these local maxima indicates that besides excessive elution strength of the sample solvent, other factors, such as a mild form of viscous fingering due to a viscosity mismatch, phase separation of H2 O and CO2 or other effects that affect mixing of the mobile phase with the sample solvent in the precolumn tubing play a role in extra-column dispersion [16]. As the solubility and retention properties differ from one sample to the other, the optimal sample solvent composition will vary, but is worthwhile to investigate from a performance perspective, as shown in Fig. 2. The applicability of water/ACN as solvent is interesting as it allows to separate water-based samples (such as urine samples) which would be impossible with a hexane-based sample solvent [17]. In addition, the volatility of H2 O/ACN is much lower than the hexanebased solvent and may be more convenient for extended sample storage times.
3.1.2. The effect of solely tubing ID for current state-of-the-art SFC instrumentation In the previous section, the ID of the connection tubing was changed together with the employed preheater (1.7 and 3.5 L) to change the extra-column band broadening. As a result, the individual effect of the tubing ID is difficult to isolate. In Fig. 3, the effect of the tubing ID is shown for 65, 170 and 250 m tubing. The tubing length was 15 cm and it was placed between the column and detector inlet. All other parameters (such as injection volume, sample solvent, precolumn volume/mixing) were held constant as indicated in the figure caption. Two flow rates were investigated: 1 mL/min (around the optimal velocity for the column and mobile phase used, full symbols in figure) and 2 mL/min (C-term dominated operation mode, open symbols in figure). Since 2 mL/min is twice the optimal flow rate, the maximal achievable plate count is lower. Additionally, due to a higher average column pressure, lower retention factors are observed. As can be seen, reducing the tubing ID for a standard 1.7 L detector cell appears to have only a minor influence on the extra-column band broadening. The reason for this may be (1) higher molecular diffusion coefficients in SFC leading to a decrease in extra-column band broadening in capillaries, allowing the use of larger ID as compared to e.g. in LC, (2) the effect is masked by a larger contribution such as the design/volume of the detector cell.
135
Fig. 4. Pressure drop versus flow rate over the detector cell at a back pressure of 130 bar: 0.6 L (
) and 1.7 L (
).
3.2. Modifications to current-state-of-the-art SFC instrumentation for using 2.1 mm columns with UHPLC-efficiency In order to achieve separation efficiencies close to 90% of the column efficiency for weakly-retained compounds, further modifications need to be made to the SFC-system. The last main extracolumn contribution is the influence of the detector. Unfortunately, low-dispersion volume detector cells with sufficiently high pressure rating are currently not available for use in SFC. The currently available flow cells for SFC-systems are e.g. 8.4 L (Waters), 4 L (Jasco), 2 L (John Morris Scientific) and 1.7 L (Agilent Technologies). However, for LC detector cells are available in a wide range of different and smaller internal volumes and path lengths. It was therefore investigated whether a low-dispersion (V() = 0.6 L) UHPLC flow cell could be used under SFC conditions. Although the recommended maximum pressure rating of this cell is 70 bar, it was assumed that the actual breaking pressure would be higher and can be used in SFC at moderate back pressures (e.g. 130 bar). It is however important to stress that this might lead to a reduced life time and/or failure of the flow cell, even under the conditions mentioned here. In addition, care was taken to minimize sudden changes in pressure, allowing the use of the 0.6 L cell for a limited time to investigate the influence of the detector cell. All experiments
), 120 ( ), 170 ( ) and 250 m ( )) on Fig. 5. Influence of tubing ID (65 ( extra-column band broadening using a 0.6 L flow cell at 1 mL/min. For reference the 1.7 L flow cell with 250 m tubing is given as well (+). The experiments were performed on 2.1 × 100 mm Zorbax HILIC column, back pressure at 130 bar, injection volume of 0.8 L consisting of 10/5/85 EtOH/IPA/hexane. The used compounds (in order of elution) are aspirine, testosterone, -estradiol and chlorthalidone.
136
R. De Pauw et al. / J. Chromatogr. A 1403 (2015) 132–137
Fig. 6. (a) Separation of (1) aspirine, (2) testosterone, (3) -estradiol, (4) bendroflumethiazide (5) chlorthalidone at a flow rate of 1 mL/min on a 2.1 × 100 mm Zorbax HILIC RRHD column using a 1.7 L flow cell, a 3 L preheater and 250 m tubing at a back pressure of 130 bar. (b) Separation of (1) aspirine, (2) testosterone, (3) -estradiol, (4) altizide and (5) chlorthalidone at a flow rate of 1 mL/min on a 2.1 × 100 mm Zorbax HILIC RRHD column using a 0.6 L flow cell, a 1.6 L preheater and 120 m tubing at a back pressure of 130 bar. The injection volume was equal to 0.8 L for both configurations.
with the 0.6 L detector cell were performed at a back pressure of 130 bar. The extra-column band broadening experiments for the 1.7 L detector cell at 130 bar were compared with the results at 150 bar. Whereas absolute values for e.g. the retention factor changed between the different set-ups (because of the inevitable differences in average column pressure), the same trends of plate count with retention factor were observed (results not shown). 3.2.1. Smaller detector cells: behavior under turbulent flow conditions In a recent paper, it was shown that turbulent flow conditions occur in the inlet tubing and/or flow cell of the given 1.7 L detector cell in SFC [19]. Given the lower dispersion volume of the 0.6 L detector cell, and its longer path length (10 mm vs. 6 mm for a 1.7 L), it can be assumed that the internal cell diameter is roughly 1.5–2 times smaller, resulting in higher mobile phase velocities and thus turbulent flow conditions may be more readily developed (since the Reynolds number, defined as Re = (4 · F/ · D2 )(D/), where F, , and D are respectively the flow rate, density and viscosity of the mobile phase and the inner diameter of the tubing, increases with decreasing internal diameter D). As the detector cell is located downstream of the column, such a narrower cell could give rise to an unexpected increase in the real column outlet pressure. Fig. 4 shows the pressure drop over both the 1.7 and 0.6 L detector cells as a function of flow rate for a back pressure of 130 bar. Due to turbulent flow conditions in the cell and/or connection capillaries both flow cells show a non-linear increase in pressure drop with increasing flow rate. The pressure drop for the 0.6 L flow cell is, however, only 6 bar higher at 3 mL/min than the 1.7 L cell. These results shows that, at least from the perspective of pressure drop losses, potential miniaturization of detector cells for use in SFC-instrumentation at flows up to several mL/min is possible. However, care has to be taken to avoid an excessive increase in detector noise due to the turbulent flow conditions inside the detector cell [19]. 3.2.2. Enhancing efficiency by reducing flow cell dispersion volume Figs. 1 and 3 showed that reduction of extra-column band broadening by changing the ID of connection tubing and sample solvent has its natural limits (or can have only a limited effect), and it is impractical to inject smaller amounts than 0.4 L. In Fig. 5, the influence of the detector cell volume (and also cell design/variance) is shown. Switching from the 1.7 L to the 0.6 L cell and using 250 m connection tubing (before and after column, total length of 50 cm), the plate count increases from 11,000 to 14,000 for a
low-retaining compound (k = 3). Optimizing the ID of the connection tubings to 120 m results in a further increase to 17,300 plates. Decreasing the ID further to 65 m appeared to result in a slightly lower plate count for low-retaining compounds ( data points lie below ). This could be due to the decreasing precolumn volumes, which allows less mixing of the sample solvent as observed in Fig. 1. Also the Reynolds number in the tubing increases from 1800 to 3100 when decreasing the tubing ID from 120 to 65 m. This brings the flow condition in the transition region between laminar and turbulent flow, where extra-column band broadening in capillaries reaches a maximum [20]. It also needs to be noted that the first peak (k = 1.1) was asymmetrical and thus the plate number shown (calculated via half-height method) may not be a good approximation for the extra-column band broadening. Fig. 6 shows a comparison of the initial system configuration (Fig. 6a: 1.7 L detector cell and 250 m tubing) and the improved system (Fig. 6b: 0.6 L and 120 m tubing) at a back pressure of 130 bar. Whereas using a 2.1 × 100 mm column, 90% of the efficiency (extrapolation towards very high k ) could only be reached at k = 8 for the non-optimized system, it is reached at roughly k = 3.0 when the system is optimized. Although this is not yet the desired UHPLC-efficiency, it is a substantial improvement compared to current SFC instrumentation. It should be noted that this relative position of the early-eluting compounds versus this apparent plateau inherently assumes that the performance is not strongly affected by retention factor and/or sample compound [6]. An additional advantage of the low-dispersion volume cell is that the noise level decreases from 1 mAU (for the 1.7 L cell) to only 0.1 mAU for the 0.6 L cell. This may be due to a different design of the cell itself (optical path and optical fiber) as well as due to a different detector hardware. 4. Conclusions The contributions to extra-column band broadening from sample solvent, injection volume, extra-column volumes and detector cell volume/design were determined for current state-of-the art SFC instrumentation using short, narrow-bore columns (e.g. 2.1 × 100 mm). Improving the match of sample solvent in terms of viscosity, elution strength and polarity to the mobile phase (e.g. using hexane/ethanol/isopropanol 85/10/5 vol%) and lowering the injection volume to 0.4 L allowed to increase the plate count from 7600 to 21,300 for a low-retaining compound (k = 2.3). Minimizing the ID of connection capillaries from 250 to 65 m appeared to have no significant influence when using a standard 1.7 L flow cell. A water/acetonitrile sample solvent was also investigated and an optimal water/acetonitrile ratio around 20–40% was found for the
R. De Pauw et al. / J. Chromatogr. A 1403 (2015) 132–137
used compounds. At this composition, the plate count was only 5% lower compared to the hexane-based solvent (if the injection and extra-column volumes were minimized), showing the suitability of this sample solvent for water-soluble samples and having the additional advantage that the volatility is much lower compared to the hexane-based solvent. When using a modern UHPLC detector cell (G4212-60038) versus a high-pressure SFC detector cell (G1314-60082) with large dispersion volume, a significant improvement in plate count is observed (e.g. an increase from 11,000 to 14,000 plates on a 2.1 × 100 mm column for k = 3) even when 250 m tubing is used. When using the smaller detector cell, the system could be further optimized by reducing extra-column volumes. Going from 250 to 120 m ID tubing resulted in a plate count increase from 14,000 to 17,300 (k = 3). Further decreasing the tubing ID (e.g. 65 m) appeared to have no further influence on the extra-column band broadening. Due to the more readily development of turbulent flow conditions in the narrower flow path of low dispersion flow cells an unexpected increase in column outlet pressure could occur. However, pressure drop measurements over the detector showed a 6 bar higher pressure drop at 3 mL/min for the 0.6 L detector cell versus the 1.7 L one. Effective usage of short and narrow-bore columns in SFC can be enabled by the reduction of injection volume and more thorough matching the properties of the sample solvent to the eluent fluid. For further improvements, low volume/variance detector cells with a high pressure rating need to be available. In this case, a further reduction in connection tubing volume can provide additional benefits. Acknowledgements R.D.P. (grant number: 11H6113N) and K.B. (1.5.201.15N) gratefully acknowledge research grants from the Research Foundation Flanders (FWO Vlaanderen). Xiaoli Wang (Agilent Technologies, Little Falls, USA) is kindly acknowledged for the gift of the chromatographic columns. References [1] A. Grand-Guillaume Perrenoud, J.L. Veuthey, D. Guillarme, Comparison of ultra-high performance supercritical fluid chromatography and ultra-high performance liquid chromatography for the analysis of pharmaceutical compounds, J. Chromatogr. A 1266 (2012) 158–167.
137
[2] C. Bouigeon, D. Thiebaut, M. Caude, Long packed column supercritical fluid chromatography: influence of pressure drop on apparent efficiency, Anal. Chem. 68 (1996) 3622–3630. [3] C. Sarazin, D. Thiebaut, P. Sassiat, J. Vial, Feasibility of ultra high performance supercritical neat carbon dioxide chromatography at conventional pressures, J. Sep. Sci. 34 (2011) 2773–2778. [4] E. Lesellier, L. Fougere, D.P. Poe, Kinetic behaviour in supercritical fluid chromatography with modified mobile phase for 5 m particle size and varied flow rates, J. Chromatogr. A 1218 (2011) 2058–2064. [5] E. Lesellier, Efficiency in supercritical fluid chromatography with different superficially porous and fully porous particles ODS bonded phases, J. Chromatogr. A 1228 (2012) 89–98. [6] A. Grand-Guillaume Perrenoud, C. Hamman, M. Goel, J.-L. Veuthey, D. Guillarme, S. Fekete, Maximizing kinetic performance in supercritical fluid chromatography using state-of-the-art instruments, J. Chromatogr. A 1314 (2013) 288–297. [7] E. Lesellier, A. Latos, A.L. de Oliveira, Ultra high efficiency/low pressure supercritical fluid chromatography with superficially porous particles for triglyceride separation, J. Chromatogr. A 1327 (2013) 141–148. [8] R. De Pauw, K. Choikhet, G. Desmet, K. Broeckhoven, Exploring the speedresolution limits of supercritical fluid chromatography at ultra-high pressures, J. Chromatogr. A 1374 (2014) 247–253. [9] J. Zauner, R. Lusk, S. Koski, D.P. Poe, Effect of the thermal environment on the efficiency of packed columns in supercritical fluid chromatography, J. Chromatogr. A 1266 (2012) 149–157. [10] A. Tarafder, G. Guiochon, Extended zones of operations in supercritical fluid chromatography, J. Chromatogr. A 1265 (2012) 165–175. [11] R. De Pauw, K. Choikhet, G. Desmet, K. Broeckhoven, Temperature effects in supercritical fluid chromatography: a trade-off between viscous heating and decompression cooling, J. Chromatogr. A 1365 (2014) 212–218. [12] A. Tarafder, P. Iraneta, G. Guiochon, K. Kaczmarski, D.P. Poe, Estimations of temperature deviations in chromatographic columns using isenthalpic plots: I. Theory for isocratic systems, J. Chromatogr. A 1366 (2014) 126–135. [13] S.O. Colgate, T.A. Berger, On axial temperature gradients due to large pressure drops in dense fluid chromatography, J. Chromatogr. A 1385 (2015) 94–102. [14] J.N. Fairchild, J.F. Hill, P.C. Iraneta, Influence of sample solvent composition for SFC separations, LCGC N. Am. 31 (2013) 326–333. [15] V. Abrahamsson, M. Sandahl, Impact of injection solvents on supercritical fluid chromatography, J. Chromatogr. A 1306 (2013) 80–88. [16] Y. Dai, G. Li, A. Rajendran, Peak distortions arising from large-volume injections in supercritical fluid chromatography, J. Chromatogr. A 1392 (2015) 91–99. [17] L. Novakova, A. Grand-Guillaume Perrenoud, R. Nicoli, M. Saugy, J.-L. Veuthey, D. Guillarme, Ultra high performance supercritical fluid chromatography coupled with tandem mass spectrometry for screening of doping agents: I. Investigation of mobile phase and MS conditions, Anal. Chim. Acta 853 (2015) 637–646. [18] T.A. Berger, Characterization of a 2.6 m Kinetex porous shell hydrophilic interaction liquid chromatography column in supercritical fluid chromatography with a comparison to 3 m totally porous silica, J. Chromatogr. A 1218 (2011) 4559–4568. [19] R. De Pauw, K. Choikhet, G. Desmet, K. Broeckhoven, Occurrence of turbulent flow conditions in supercritical fluid chromatography, J. Chromatogr. A 1361 (2014) 277–285. [20] O. Levenspiel, Chemical Reaction Engineering, 3rd ed., J. Wiley, New York, 1999.
Contents lists available at ScienceDirect
Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma
Understanding and diminishing the extra-column band broadening effects in supercritical fluid chromatography Ruben De Pauw a , Konstantin Shoykhet (Choikhet) b , Gert Desmet a , Ken Broeckhoven a,∗ a b
Vrije Universiteit Brussel, Department of Chemical Engineering (CHIS-IR), Pleinlaan 2, 1050 Brussels, Belgium Agilent Technologies Europe, Hewlett-Packard-Strasse 8, 76337 Waldbronn, Germany
a r t i c l e
i n f o
Article history: Received 31 March 2015 Received in revised form 7 May 2015 Accepted 7 May 2015 Available online 19 May 2015 Keywords: Supercritical fluid chromatography Detector cell Extra-column band broadening Extra-column volume Sample solvent Injection volume
a b s t r a c t Supercritical fluid chromatography, where a low-viscosity mobile phase such as carbon dioxide is used, proves to be an excellent technique for fast and efficient separations, especially when sub-2 m particles are used. However, to achieve high velocities when using these small particles, and in order to stay within the flow rate range of current SFC-instruments, narrow columns (e.g. 2.1 mm ID) must be used. Unfortunately, state-of-the-art instrumentation is limiting the full separation power of these narrower columns due to significant extra-column band broadening effects. The present work identifies and quantifies the different contributions to extra-column band broadening in SFC such as the influence of the sample solvent, injection volume, extra-column volumes and detector cell volume/design. When matching the sample solvent to the mobile phase in terms of elution strength and polarity (e.g. using hexane/ethanol/isopropanol 85/10/5 vol%) and lowering the injection volume to 0.4 L, the plate count can be increased from 7600 to 21,300 for a low-retaining compound (k = 2.3) on a 2.1 mm × 150 mm column (packed with 1.8 m particles). The application of a water/acetonitrile mixture as sample solvent was also investigated. It was found that when the volumetric ratio of water/acetonitrile was optimized, only a slightly lower plate count was measured compared to the hexane-based solvent when minimizing injection and extra-column volume. This confirms earlier results that water/acetonitrile can be used if water-soluble samples are considered or when a less volatile solvent is preferred. Minimizing the ID of the connection capillaries from 250 to 65 m, however, gives no further improvement in obtained efficiency for early-eluting compounds when a standard system configuration with optimized sample solvent was used. When switching to a state-of-the-art detector design with reduced (dispersion) volume (1.7–0.6 L), an increase in plate count is observed (from 11,000 to 14,000 plates on a 2.1 mm × 100 mm column with 1.8 m particles for k = 3) even when 250 m tubing was used. Using this detector cell and decreasing the ID of the tubing from 250 to 120 m resulted in an additional increase to 17,300 plates. Further decreasing the tubing ID (e.g. 65 m) appeared to have no observable influence on the obtained plate count. © 2015 Elsevier B.V. All rights reserved.
1. Introduction In order to achieve fast and efficient separations in chromatography, it is important to decrease the resistance for mass transfer. Among several approaches, using a low-viscosity fluid, such as fluidic CO2 , results in very high diffusion coefficients and thus allows the use of higher flow rates without significant efficiency loss [1]. As a result, very fast and efficient separations can be performed in supercritical fluid chromatography (SFC) [2–7]. For example, a plate count of 182,000 plates can be achieved for a void time of 5.8 min
∗ Corresponding author. Tel.: +32 26293781; fax: +32 26293248. E-mail address: [email protected] (K. Broeckhoven). http://dx.doi.org/10.1016/j.chroma.2015.05.017 0021-9673/© 2015 Elsevier B.V. All rights reserved.
when the pressure limitation of state-of-the-art instrumentation (600 bar) is considered [8]. Although this makes carbon dioxide an attractive solvent for chromatography, working with a highly compressible mobile phase has some drawbacks of its own. First of all, a back pressure above the critical pressure (72 bar) is required to avoid the formation of gas bubbles in the detector cell. Typically, even higher back pressures are used (120–150 bar) to operate in a region sufficiently far from the critical point, i.e. far from the region of high compressibility of the mobile phase. Otherwise, the large changes in mobile phase density can cause extensive decompression cooling which results in a loss in performance and significant variation in velocity and retention along the column occurs due to the large variation in density [9–13]. Another consequence is the inability to
R. De Pauw et al. / J. Chromatogr. A 1403 (2015) 132–137
precisely match the sample solvent with the mobile phase, as this would require the sample to be compressed in a CO2 -based mobile phase. Depending on the column ID and the injected volume, this mismatch between sample solvent and mobile phase may lead to significant extra-column band broadening especially in preparative scale separations [16]. Several studies highlighted the effect of sample solvents on performance in SFC but even with this knowledge, issues with extra-column band broadening still remain for narrow columns. In general it was found that apolar solvents such as heptane allow to minimize peak distortion, whereas another study found that acetonitrile in combination with water could be used if water-soluble samples are considered [14,15,17,18]. A recent investigation of Perrenoud et al. showed that, due to the significant extra-column variance in SFC, the optimal column geometry would be a 3 mm ID column in the case of small particles and short columns. This ID would be a perfect trade-off between minimizing extra-column band broadening effects and still being able to achieve sufficiently high linear velocities for e.g. sub-2 m particles [6]. A previous study about speed-resolution limits in SFC also illustrated important extra-column band broadening effects for early-eluting compounds on short columns with a 2.1 mm ID format, which were largely resolved when using very long columns (e.g. 500 mm) [8]. Nonetheless, the origin and relative contributions of extracolumn band broadening in SFC are not entirely understood or quantified even though it follows the same basic principles as e.g. in liquid chromatography (LC). The potential sources are: • Sample solvent: even though the use of hexane/ethanol (EtOH)/isopropanol (IPA), where IPA is added to allow better mixing of EtOH and hexane [6], as a sample solvent provides a reasonable viscosity and retention match with the CO2 /methanol (MeOH) mobile phase, it is not perfect and may still lead to a distortion of the sample band at the column inlet. In addition, a water (H2 O)/acetonitrile (ACN) solvent will also be assessed since it may be of practical use when testing water-soluble samples [17]. • Injection volume: SFC-injectors generally work via the fixed-loop principle where the injection loop is generally large (5 or 10 L). If partial-loop injection is considered, the rest of the loop is generally filled with the mobile phase co-solvent (or another wash solvent). At the moment of injection, a large amount of mismatching solvent hence enters the column with the sample, which may lead to peak distortion. • Extra-column volumes: in SFC, generally long tubing with higher ID’s are used (170 m instead of the more common 120 or even 75 m in LC), which could lead to severe extra-column band broadening. In addition, the geometry of the extra-column volumes (e.g. detector design) may also lead to an additional contribution due to the presence of stagnant zones. The present work investigates and quantifies the relative contributions of these potential band broadening sources for current state-of-the-art SFC instrumentation and explores some new ways to reduce efficiency losses for low-retained compounds. 2. Experimental 2.1. Column, tubing and chemicals Methanol, hexane and isopropanol (LC-MS grade) were purchased from Biosolve (Valkenswaard, Netherlands), CO2 was purchased from Air Liquide (Paris, France). Test components such as testosterone, chlorthalidone, bendroflumethiazide, altizide and -estradiol were kindly provided by Deirdre Cabooter (Laboratory
133
of Pharmaceutical Analysis, KU Leuven, Belgium). The samples were dissolved in a mixture of ethanol (EtOH), isopropanol (IPA) and hexane (except for the assessment of the water/acetonitrile solvent). The IPA was added to allow mixing of EtOH with hexane [6]. In order to inject different sample volumes, a 20 cm fused-silica 50 m capillary, a 7 cm, 120 m and a 11 cm, 170 m stainless steel capillary were used, to achieve, respectively, a 0.4, 0.8 and 2.5 L injection volume (instead of the 5 L sample loop intended for the standard configuration of the instrument). All the reported extra-column volumes are calculated, based on the nominal dimensions given by the manufacturers. These nominal dimensions were evaluated by measuring the total extra-column volume under LC conditions and a deviation of maximally 15% was found. 2.2. Instrumentation and conditions The SFC-system used in the study was a modified Agilent G4301A-based SFC system in combination with a thermostatted column compartment, autosampler used in full-loop mode and two flow cells with a dispersion volume of 1.7 L (DAD G1315C with a G1314-60082 flow cell) and 0.6 L (Agilent 1290 Infinity DAD G4212A with a G4212-60038 flow cell). Zorbax HILIC RRHD columns (2.1 mm ID, 150 mm and 100 mm, 1.8 m fully porous particles) were used in the current study. The columns were kindly provided by Xiaoli Wang (Agilent Technologies, Little Falls, USA). The oven temperature was set at 40 ◦ C, 8 v% methanol as modifier was used and the back pressure was set at 150 bar, unless otherwise specified. The detector was set at an acquisition rate of 100 Hz and a wavelength of 230 nm. Due to the high acquisition rate, the observed peak widths and variances were not affected, even for very the narrow, early-eluting peaks. The shown data are average values of 3 repeats and a maximum relative standard deviation for the plate count of 6% was found. 3. Results and discussion 3.1. Possibilities with current state-of-the-art SFC instrumentation The first goal is to investigate the possibilities to reduce peak broadening in state-of-the-art SFC systems. For this purpose we have • varied the sample composition. • varied the injection volume by using 0.4 and 2.5 L loops. • varied pre- and post-column volumes by choosing smaller ID tubing or switching the preheater (e.g. 1.6 L instead of 3 L preheater intended for the standard configuration of the used instrument). 3.1.1. Sample solvent and extra-column volume To study the influence of extra-column volume, only the volume before the column (extra-precolumn volume) was changed by: changing the preheater (1.6 or 3 L) and the connection capillary from injector to preheater (120 or 170 m). This resulted in two different extra-precolumn volumes of 4.4 L (20 cm, 120 m connection capillary from injector to preheater, 1.6 L preheater) and 7.4 L (17 cm, 170 m connection capillary, 3 L preheater). The same 5 cm long 120 m connection capillary from preheater to column was kept for both configurations. The injection volume was varied by using either the 0.4 or the 2.5 L loop in the overfill mode. Whereas commercial SFC-systems are typically supplied with a 5 L-loop, smaller loops can be used but it is not recommended on a regular basis as it may lead to system errors during the washing or filling phase due to an increased
134
R. De Pauw et al. / J. Chromatogr. A 1403 (2015) 132–137
Fig. 1. Apparent plate count as function of sample solvent concentration (volumetric EtOH-concentration, , mixed with hexane and IPA), with the EtOH/IPA ratio always equal to 2, for different injection volumes (Vinj ) and extra-precolumn volumes (VEC ): (a) testosterone, low-retaining compound (k = 2.3), (b) Altizide, high-retaining compound (k = 10.8). Measured on a 2.1 × 150 mm Zorbax HILIC RRHD column (packed with 1.8 m particles) at a flow rate of 1 mL/min.
pressure drop. In addition, a 50 m ID tubing generates a significant pressure drop during the separation which may result in a cyclic load on the system when switching the injection valve. However, for this study on injection volumes in SFC, it provided very low and reliable injection volumes. The last important parameter, the composition of the sample solvent, was studied as sample solvents in SFC are never matched to the mobile phase (in terms of viscosity and retention matching) as it is impractical to use a compressed CO2 /MeOH mixture [14,15]. To investigate the effect of sample solvent, the volumetric fraction of EtOH was changed while keeping the volume ratio EtOH/IPA equal to 2. This was achieved by using the following sample solvents: 10% EtOH (with 5% IPA and 85% hexane), 20% EtOH (with 10% IPA and 70% hexane) and 40% EtOH (with 20% IPA and 40% hexane) [6]. In Fig. 1 the apparent plate counts are given for a low and high-retaining compound (k = 2.3 or 10.8). For the low-retaining compound (Fig. 1a), variation of the extra-precolumn volume between 4.4 and 7.4 L ( versus ) did not influence the observed plate count when using an injection volume of 2.5 L. The effect of the injection volume, however, is large for high EtOH-fractions (20% and above) ( versus ). Increasing the EtOH-content to 40% results in a significant decrease in plate count for the three different cases and the effect is most pronounced for the largest injection volume. When decreasing the EtOH-content, the plate count for the smallest injection volume ( ) levels off to a maximum value. The two other cases also appear to converge towards this limit at low EtOH-content. This leveling off indicates that a further reduction of the EtOH-fraction will not improve the performance and that the
reduction of other extra-column band broadening contributions (such as the detector cell) is required. For the high-retaining compound, Fig. 1b, a similar behavior is observed but the absolute values of the plate count are higher since extra-band broadening is less significant for higher retaining compounds eluting in broader peaks. However, an additional effect can now be seen for 40% EtOH. It appears that increasing the precolumn volume leads to an improvement in plate count. Whereas with the small extra-precolumn volume ( ) an efficiency of only 8000 plates is observed, the plate count increases to 14,000 if the precolumn volume is increased to 7.40 L (standard deviation for these measurements is equal to 300 plates). This suggests that at high EtOH-content, the higher extra-precolumn volume allows a better mixing of the sample with the mobile phase and results in a more compatible sample solvent when entering the column. In Fig. 2, a similar study is performed for a H2 O/ACN sample solvent. In both configurations, the same 120 m tubing from injector to preheater was used and only the injection volume and preheater volume were changed (e.g. extra-precolumn volume equal to 4.4 L and 5.8 L when using a 1.6 and 3 L preheater respectively). From Fig. 2b, it is clear that decreasing the injection volume significantly improves the performance. For larger injection volumes and a strongly retained compound, the efficiency was highest with a fraction of 40% water in the sample solvent. For the lowretaining compound (Fig. 2a) no clear trend with ACN volume fraction could be observed and the curves for the two injection volumes even cross, resulting in an apparent maximum at 20% water for the 0.4 L injection and 40% for the 2 L injection. Near this maximum, nearly the same performance is found as for the
Fig. 2. Apparent plate count as function of sample solvent concentration (volumetric H2 O-concentration, , mixed with ACN) for two different injection volumes (Vinj ) and extra-precolumn volumes (VEC ): (a) testosterone, low-retaining compound (k = 2.2). (b) altizide, high-retaining compound (k = 12.1). Same column and flow rate as in Fig. 1.
R. De Pauw et al. / J. Chromatogr. A 1403 (2015) 132–137
Fig. 3. Influence of post-column tubing ID (65 ( ), 170 ( ) and 250 m ( )) on extra-column band broadening at different flow rates: 1 mL/min (full symbols) and 2 mL/min (open symbols). The experiments were performed on 2.1 × 100 mm Zorbax HILIC column, back pressure at 150 bar, injection volume of 0.8 L with a sample solvent of 20/10/70 EtOH/IPA/hexane. The used compounds (in order of elution) are aspirine, testosterone, -estradiol, bendroflumethiazide and chlorthalidone.
optimized EtOH/IPA/hexane-solvent (20,000 instead of 21,000 plates). The occurrence of these local maxima indicates that besides excessive elution strength of the sample solvent, other factors, such as a mild form of viscous fingering due to a viscosity mismatch, phase separation of H2 O and CO2 or other effects that affect mixing of the mobile phase with the sample solvent in the precolumn tubing play a role in extra-column dispersion [16]. As the solubility and retention properties differ from one sample to the other, the optimal sample solvent composition will vary, but is worthwhile to investigate from a performance perspective, as shown in Fig. 2. The applicability of water/ACN as solvent is interesting as it allows to separate water-based samples (such as urine samples) which would be impossible with a hexane-based sample solvent [17]. In addition, the volatility of H2 O/ACN is much lower than the hexanebased solvent and may be more convenient for extended sample storage times.
3.1.2. The effect of solely tubing ID for current state-of-the-art SFC instrumentation In the previous section, the ID of the connection tubing was changed together with the employed preheater (1.7 and 3.5 L) to change the extra-column band broadening. As a result, the individual effect of the tubing ID is difficult to isolate. In Fig. 3, the effect of the tubing ID is shown for 65, 170 and 250 m tubing. The tubing length was 15 cm and it was placed between the column and detector inlet. All other parameters (such as injection volume, sample solvent, precolumn volume/mixing) were held constant as indicated in the figure caption. Two flow rates were investigated: 1 mL/min (around the optimal velocity for the column and mobile phase used, full symbols in figure) and 2 mL/min (C-term dominated operation mode, open symbols in figure). Since 2 mL/min is twice the optimal flow rate, the maximal achievable plate count is lower. Additionally, due to a higher average column pressure, lower retention factors are observed. As can be seen, reducing the tubing ID for a standard 1.7 L detector cell appears to have only a minor influence on the extra-column band broadening. The reason for this may be (1) higher molecular diffusion coefficients in SFC leading to a decrease in extra-column band broadening in capillaries, allowing the use of larger ID as compared to e.g. in LC, (2) the effect is masked by a larger contribution such as the design/volume of the detector cell.
135
Fig. 4. Pressure drop versus flow rate over the detector cell at a back pressure of 130 bar: 0.6 L (
) and 1.7 L (
).
3.2. Modifications to current-state-of-the-art SFC instrumentation for using 2.1 mm columns with UHPLC-efficiency In order to achieve separation efficiencies close to 90% of the column efficiency for weakly-retained compounds, further modifications need to be made to the SFC-system. The last main extracolumn contribution is the influence of the detector. Unfortunately, low-dispersion volume detector cells with sufficiently high pressure rating are currently not available for use in SFC. The currently available flow cells for SFC-systems are e.g. 8.4 L (Waters), 4 L (Jasco), 2 L (John Morris Scientific) and 1.7 L (Agilent Technologies). However, for LC detector cells are available in a wide range of different and smaller internal volumes and path lengths. It was therefore investigated whether a low-dispersion (V() = 0.6 L) UHPLC flow cell could be used under SFC conditions. Although the recommended maximum pressure rating of this cell is 70 bar, it was assumed that the actual breaking pressure would be higher and can be used in SFC at moderate back pressures (e.g. 130 bar). It is however important to stress that this might lead to a reduced life time and/or failure of the flow cell, even under the conditions mentioned here. In addition, care was taken to minimize sudden changes in pressure, allowing the use of the 0.6 L cell for a limited time to investigate the influence of the detector cell. All experiments
), 120 ( ), 170 ( ) and 250 m ( )) on Fig. 5. Influence of tubing ID (65 ( extra-column band broadening using a 0.6 L flow cell at 1 mL/min. For reference the 1.7 L flow cell with 250 m tubing is given as well (+). The experiments were performed on 2.1 × 100 mm Zorbax HILIC column, back pressure at 130 bar, injection volume of 0.8 L consisting of 10/5/85 EtOH/IPA/hexane. The used compounds (in order of elution) are aspirine, testosterone, -estradiol and chlorthalidone.
136
R. De Pauw et al. / J. Chromatogr. A 1403 (2015) 132–137
Fig. 6. (a) Separation of (1) aspirine, (2) testosterone, (3) -estradiol, (4) bendroflumethiazide (5) chlorthalidone at a flow rate of 1 mL/min on a 2.1 × 100 mm Zorbax HILIC RRHD column using a 1.7 L flow cell, a 3 L preheater and 250 m tubing at a back pressure of 130 bar. (b) Separation of (1) aspirine, (2) testosterone, (3) -estradiol, (4) altizide and (5) chlorthalidone at a flow rate of 1 mL/min on a 2.1 × 100 mm Zorbax HILIC RRHD column using a 0.6 L flow cell, a 1.6 L preheater and 120 m tubing at a back pressure of 130 bar. The injection volume was equal to 0.8 L for both configurations.
with the 0.6 L detector cell were performed at a back pressure of 130 bar. The extra-column band broadening experiments for the 1.7 L detector cell at 130 bar were compared with the results at 150 bar. Whereas absolute values for e.g. the retention factor changed between the different set-ups (because of the inevitable differences in average column pressure), the same trends of plate count with retention factor were observed (results not shown). 3.2.1. Smaller detector cells: behavior under turbulent flow conditions In a recent paper, it was shown that turbulent flow conditions occur in the inlet tubing and/or flow cell of the given 1.7 L detector cell in SFC [19]. Given the lower dispersion volume of the 0.6 L detector cell, and its longer path length (10 mm vs. 6 mm for a 1.7 L), it can be assumed that the internal cell diameter is roughly 1.5–2 times smaller, resulting in higher mobile phase velocities and thus turbulent flow conditions may be more readily developed (since the Reynolds number, defined as Re = (4 · F/ · D2 )(D/), where F, , and D are respectively the flow rate, density and viscosity of the mobile phase and the inner diameter of the tubing, increases with decreasing internal diameter D). As the detector cell is located downstream of the column, such a narrower cell could give rise to an unexpected increase in the real column outlet pressure. Fig. 4 shows the pressure drop over both the 1.7 and 0.6 L detector cells as a function of flow rate for a back pressure of 130 bar. Due to turbulent flow conditions in the cell and/or connection capillaries both flow cells show a non-linear increase in pressure drop with increasing flow rate. The pressure drop for the 0.6 L flow cell is, however, only 6 bar higher at 3 mL/min than the 1.7 L cell. These results shows that, at least from the perspective of pressure drop losses, potential miniaturization of detector cells for use in SFC-instrumentation at flows up to several mL/min is possible. However, care has to be taken to avoid an excessive increase in detector noise due to the turbulent flow conditions inside the detector cell [19]. 3.2.2. Enhancing efficiency by reducing flow cell dispersion volume Figs. 1 and 3 showed that reduction of extra-column band broadening by changing the ID of connection tubing and sample solvent has its natural limits (or can have only a limited effect), and it is impractical to inject smaller amounts than 0.4 L. In Fig. 5, the influence of the detector cell volume (and also cell design/variance) is shown. Switching from the 1.7 L to the 0.6 L cell and using 250 m connection tubing (before and after column, total length of 50 cm), the plate count increases from 11,000 to 14,000 for a
low-retaining compound (k = 3). Optimizing the ID of the connection tubings to 120 m results in a further increase to 17,300 plates. Decreasing the ID further to 65 m appeared to result in a slightly lower plate count for low-retaining compounds ( data points lie below ). This could be due to the decreasing precolumn volumes, which allows less mixing of the sample solvent as observed in Fig. 1. Also the Reynolds number in the tubing increases from 1800 to 3100 when decreasing the tubing ID from 120 to 65 m. This brings the flow condition in the transition region between laminar and turbulent flow, where extra-column band broadening in capillaries reaches a maximum [20]. It also needs to be noted that the first peak (k = 1.1) was asymmetrical and thus the plate number shown (calculated via half-height method) may not be a good approximation for the extra-column band broadening. Fig. 6 shows a comparison of the initial system configuration (Fig. 6a: 1.7 L detector cell and 250 m tubing) and the improved system (Fig. 6b: 0.6 L and 120 m tubing) at a back pressure of 130 bar. Whereas using a 2.1 × 100 mm column, 90% of the efficiency (extrapolation towards very high k ) could only be reached at k = 8 for the non-optimized system, it is reached at roughly k = 3.0 when the system is optimized. Although this is not yet the desired UHPLC-efficiency, it is a substantial improvement compared to current SFC instrumentation. It should be noted that this relative position of the early-eluting compounds versus this apparent plateau inherently assumes that the performance is not strongly affected by retention factor and/or sample compound [6]. An additional advantage of the low-dispersion volume cell is that the noise level decreases from 1 mAU (for the 1.7 L cell) to only 0.1 mAU for the 0.6 L cell. This may be due to a different design of the cell itself (optical path and optical fiber) as well as due to a different detector hardware. 4. Conclusions The contributions to extra-column band broadening from sample solvent, injection volume, extra-column volumes and detector cell volume/design were determined for current state-of-the art SFC instrumentation using short, narrow-bore columns (e.g. 2.1 × 100 mm). Improving the match of sample solvent in terms of viscosity, elution strength and polarity to the mobile phase (e.g. using hexane/ethanol/isopropanol 85/10/5 vol%) and lowering the injection volume to 0.4 L allowed to increase the plate count from 7600 to 21,300 for a low-retaining compound (k = 2.3). Minimizing the ID of connection capillaries from 250 to 65 m appeared to have no significant influence when using a standard 1.7 L flow cell. A water/acetonitrile sample solvent was also investigated and an optimal water/acetonitrile ratio around 20–40% was found for the
R. De Pauw et al. / J. Chromatogr. A 1403 (2015) 132–137
used compounds. At this composition, the plate count was only 5% lower compared to the hexane-based solvent (if the injection and extra-column volumes were minimized), showing the suitability of this sample solvent for water-soluble samples and having the additional advantage that the volatility is much lower compared to the hexane-based solvent. When using a modern UHPLC detector cell (G4212-60038) versus a high-pressure SFC detector cell (G1314-60082) with large dispersion volume, a significant improvement in plate count is observed (e.g. an increase from 11,000 to 14,000 plates on a 2.1 × 100 mm column for k = 3) even when 250 m tubing is used. When using the smaller detector cell, the system could be further optimized by reducing extra-column volumes. Going from 250 to 120 m ID tubing resulted in a plate count increase from 14,000 to 17,300 (k = 3). Further decreasing the tubing ID (e.g. 65 m) appeared to have no further influence on the extra-column band broadening. Due to the more readily development of turbulent flow conditions in the narrower flow path of low dispersion flow cells an unexpected increase in column outlet pressure could occur. However, pressure drop measurements over the detector showed a 6 bar higher pressure drop at 3 mL/min for the 0.6 L detector cell versus the 1.7 L one. Effective usage of short and narrow-bore columns in SFC can be enabled by the reduction of injection volume and more thorough matching the properties of the sample solvent to the eluent fluid. For further improvements, low volume/variance detector cells with a high pressure rating need to be available. In this case, a further reduction in connection tubing volume can provide additional benefits. Acknowledgements R.D.P. (grant number: 11H6113N) and K.B. (1.5.201.15N) gratefully acknowledge research grants from the Research Foundation Flanders (FWO Vlaanderen). Xiaoli Wang (Agilent Technologies, Little Falls, USA) is kindly acknowledged for the gift of the chromatographic columns. References [1] A. Grand-Guillaume Perrenoud, J.L. Veuthey, D. Guillarme, Comparison of ultra-high performance supercritical fluid chromatography and ultra-high performance liquid chromatography for the analysis of pharmaceutical compounds, J. Chromatogr. A 1266 (2012) 158–167.
137
[2] C. Bouigeon, D. Thiebaut, M. Caude, Long packed column supercritical fluid chromatography: influence of pressure drop on apparent efficiency, Anal. Chem. 68 (1996) 3622–3630. [3] C. Sarazin, D. Thiebaut, P. Sassiat, J. Vial, Feasibility of ultra high performance supercritical neat carbon dioxide chromatography at conventional pressures, J. Sep. Sci. 34 (2011) 2773–2778. [4] E. Lesellier, L. Fougere, D.P. Poe, Kinetic behaviour in supercritical fluid chromatography with modified mobile phase for 5 m particle size and varied flow rates, J. Chromatogr. A 1218 (2011) 2058–2064. [5] E. Lesellier, Efficiency in supercritical fluid chromatography with different superficially porous and fully porous particles ODS bonded phases, J. Chromatogr. A 1228 (2012) 89–98. [6] A. Grand-Guillaume Perrenoud, C. Hamman, M. Goel, J.-L. Veuthey, D. Guillarme, S. Fekete, Maximizing kinetic performance in supercritical fluid chromatography using state-of-the-art instruments, J. Chromatogr. A 1314 (2013) 288–297. [7] E. Lesellier, A. Latos, A.L. de Oliveira, Ultra high efficiency/low pressure supercritical fluid chromatography with superficially porous particles for triglyceride separation, J. Chromatogr. A 1327 (2013) 141–148. [8] R. De Pauw, K. Choikhet, G. Desmet, K. Broeckhoven, Exploring the speedresolution limits of supercritical fluid chromatography at ultra-high pressures, J. Chromatogr. A 1374 (2014) 247–253. [9] J. Zauner, R. Lusk, S. Koski, D.P. Poe, Effect of the thermal environment on the efficiency of packed columns in supercritical fluid chromatography, J. Chromatogr. A 1266 (2012) 149–157. [10] A. Tarafder, G. Guiochon, Extended zones of operations in supercritical fluid chromatography, J. Chromatogr. A 1265 (2012) 165–175. [11] R. De Pauw, K. Choikhet, G. Desmet, K. Broeckhoven, Temperature effects in supercritical fluid chromatography: a trade-off between viscous heating and decompression cooling, J. Chromatogr. A 1365 (2014) 212–218. [12] A. Tarafder, P. Iraneta, G. Guiochon, K. Kaczmarski, D.P. Poe, Estimations of temperature deviations in chromatographic columns using isenthalpic plots: I. Theory for isocratic systems, J. Chromatogr. A 1366 (2014) 126–135. [13] S.O. Colgate, T.A. Berger, On axial temperature gradients due to large pressure drops in dense fluid chromatography, J. Chromatogr. A 1385 (2015) 94–102. [14] J.N. Fairchild, J.F. Hill, P.C. Iraneta, Influence of sample solvent composition for SFC separations, LCGC N. Am. 31 (2013) 326–333. [15] V. Abrahamsson, M. Sandahl, Impact of injection solvents on supercritical fluid chromatography, J. Chromatogr. A 1306 (2013) 80–88. [16] Y. Dai, G. Li, A. Rajendran, Peak distortions arising from large-volume injections in supercritical fluid chromatography, J. Chromatogr. A 1392 (2015) 91–99. [17] L. Novakova, A. Grand-Guillaume Perrenoud, R. Nicoli, M. Saugy, J.-L. Veuthey, D. Guillarme, Ultra high performance supercritical fluid chromatography coupled with tandem mass spectrometry for screening of doping agents: I. Investigation of mobile phase and MS conditions, Anal. Chim. Acta 853 (2015) 637–646. [18] T.A. Berger, Characterization of a 2.6 m Kinetex porous shell hydrophilic interaction liquid chromatography column in supercritical fluid chromatography with a comparison to 3 m totally porous silica, J. Chromatogr. A 1218 (2011) 4559–4568. [19] R. De Pauw, K. Choikhet, G. Desmet, K. Broeckhoven, Occurrence of turbulent flow conditions in supercritical fluid chromatography, J. Chromatogr. A 1361 (2014) 277–285. [20] O. Levenspiel, Chemical Reaction Engineering, 3rd ed., J. Wiley, New York, 1999.
