Journal of Chromatography A, 1365 (2014) 212–218
Contents lists available at ScienceDirect
Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma
Temperature effects in supercritical fluid chromatography: A trade-off between viscous heating and decompression cooling Ruben De Pauw a , Konstantin 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 14 July 2014 Received in revised form 5 September 2014 Accepted 8 September 2014 Available online 17 September 2014 Keywords: Supercritical fluid chromatography Viscous heating Decompression cooling Temperature Pressure Coupled columns
a b s t r a c t The study of radial and axial temperature profiles always has been an area interest both in liquid chromatography (LC) and supercritical fluid chromatography (SFC). Whereas in LC always an increase in temperature is observed due to the dominance of viscous heating, in SFC, especially for low modifier content, a decrease in temperature is found due to the much larger decompression cooling. However, for higher modifier content and higher operating pressure, the temperature effects become a trade-off between viscous heating and decompression cooling, since in SFC the latter is a strong function of operating pressure and mobile phase composition. At a temperature of 40 ◦ C and for neat CO2 , the effect of decompression cooling and viscous heating cancel each other out at a pressure 450 bar. This pressure decreases almost linearly with volume fraction of methanol to 150 bar at 25 vol%. As a result, both cooling and heating effects can be observed when operating at high back pressure, large column pressure drops or high modifier content. For example at a back pressure of 150 bar and a column pressure drop of 270 bar decompression cooling is observed throughout the column. However at 300 bar back pressure and the same pressure drop, the mobile phase heats up in the first part of the column due to viscous heating and then cools in the second part due to decompression cooling. When coupling columns (2.1 mm × 150 mm, 1.8 m fully porous particles) at very high operating pressure (e.g. 750 bar for 8 vol%), the situation is even more complex. E.g. at a back pressure of 150 bar and using 8 vol% methanol, viscous heating is only observed in the first column whereas only decompression cooling in the second. Further increasing the inlet pressure up to 1050 bar resulted in no excessive temperature differences along the column. This implies that the inlet pressure of SFC instrumentation could be expanded above 600 bar without additional band broadening caused by excessive radial temperature differences. © 2014 Elsevier B.V. All rights reserved.
1. Introduction In liquid chromatography (LC), the presence and consequences of radial and axial temperature profiles were discussed four decades ago and were extensively studied on a theoretical and experimental basis by Poppe et al. [1–4]. The problem received a renewed attention in the last decade with the introduction of ultrahigh pressure liquid chromatography (UHPLC) [5–10]. In UHPLC, a significant temperature increase of the mobile phase is observed and expected due to viscous heating. Since large axial temperature profiles lead to radial temperature profiles and thus radial velocity differences [11], a maximal pressure limit for commercial
∗ Corresponding author. Tel.: +32 26293781; fax: +32 26293248. E-mail address: [email protected] (K. Broeckhoven). http://dx.doi.org/10.1016/j.chroma.2014.09.022 0021-9673/© 2014 Elsevier B.V. All rights reserved.
instrumentation of roughly 1200–1300 bar appeared to be established, which was recently increased to 1500 bar. In supercritical fluid chromatography (SFC) similar examinations were performed by several authors, showing that excessive performance loss can be expected when the outlet pressure and temperature are near the critical point [12,13]. Recent papers showed that these performance losses were mainly due to radial temperature profiles as a result of significant decompression cooling [14,15]. Even though in SFC and LC opposite temperature gradient profiles are observed, the same basic principles are valid and determine the resulting temperature gradient. Two main effects can be identified: viscous heating and decompression cooling. Whereas viscous heating is the heating of the mobile phase due to friction with the column bed, decompression of the mobile phase leads to a decrease in temperature since energy is absorbed to overcome attractive forces between molecules of the mobile phase [16]. These two
R.D. Pauw et al. / J. Chromatogr. A 1365 (2014) 212–218
213
effects can be combined into a single equation, in the ideal case of an adiabatic column (no heat transfer to environment) [5] Tadi = −(1 − ˛ · T ) ·
P Cp,V
(1)
where Tadi , T, P and Cp,V are respectively the temperature change in adiabatic conditions, the temperature, pressure drop over the column and the specific volumetric heat capacity of the mobile phase. Please note that the gradients are defined from inlet to outlet, hence P is always smaller than 0 and T > 0 when a temperature increase is observed. ˛ is the coefficient of thermal expansion and is related to the density, via [17,18] 1 ˛=−
∂ ∂T
(2) P
The first term in Eq. (1), i.e. −P/Cp,V , is positive (P = Pout − Pin < 0), leading to a temperature increase and is thus related to the viscous heating, whereas the second term ((˛ · TP)/Cp,V ) opposes this increase (˛ usually has a positive sign). The latter term is thus the decompression cooling. Since ˛ is a function of pressure, temperature and mobile phase composition (volume-based, ), a resulting increase or decrease in temperature will depend on the experimental conditions. In case of a thermostatted column, an isothermal assumption is more valid (column wall at fixed temperature). The temperature change (averaged out over the cross-section of the column) can then be written as [9,19] Tiso = −(1 − ˛ · T ) ·
us · P · R2 8 · rad · L
(3)
with us , L, R and rad respectively the superficial velocity, the column length, column inner radius and the radial thermal conductivity of the packed bed. The amplitude of the radial temperature difference is found by changing the factor of 8 by 4 in Eq.(3). In this work, thermal gradients across columns are further investigated in SFC conditions, up to the pressure limits of novel SFC instrumentation (600 bar) and beyond (up till 1050 bar). In addition, the influence of the column diameter and oven type (still-air versus forced-air) is considered. 2. Experimental 2.1. Column, tubing and chemicals Methanol (LC–MS grade) was purchased from Biosolve (Valkenswaard, The Netherlands), CO2 was purchased from Air Liquide (Paris, France). HILIC RRHD columns, packed with 1.8 m bare-silica Zorbax particles, were kindly provided by Xiaoli Wang (Agilent Technologies, Little Falls, USA) with different column lengths (10 and 15 cm) and with different IDs (2.1 and 3 mm). 2.2. Instrumentation and conditions The SFC system used in the study was a modified Agilent G4301A-based SFC system with an extended column pressure range up to 1050 bar in combination with a thermostatted column compartment, autosampler with a 1.2 L injection loop (used in full loop mode) and a DAD-detector with a 1.7 L flow cell. For all the experiments the back pressure regulator was set at 150 bar (unless otherwise specified) and the oven temperature at 40 ◦ C. Other oven temperatures may give different quantitative results but will yield the same qualitative results.
Fig. 1. The term ‘(1 − ˛ · T)’ (see Eq. (1)) as a function of pressure and mobile phase composition (methanol as modifier). Using the data for ˛ provided by NIST REFPROP [21] for a temperature of 40 ◦ C. The black line denotes where viscous heating and decompression cooling cancel each other or where (1 − ˛ · T) = 0.
2.3. Temperature measurements Temperature measurements were performed on the columns and connection capillaries using chromal-alumel thermocouples (K-type thermocouple wire) positioned as indicated in Fig. 2. To provide a perfect thermal contact, the two separate wires of the thermocouple were point-welded on top of each other and directly onto the column or capillaries at the measurement spot location using a device built in-house. The almost perfect thermal contact achieved by point-welding the thermocouple on the column resulted in a very accurate measurement of the temperature on the surface of the column, without the need for thermal tempering [20], as was shown experimentally in the supplementary information. The voltages were read out using a TBX-68T Isothermal Terminal Block (National Instruments, USA) and a NI 435X (National Instruments, USA) PCI card. The local temperature inside the thermal block was measured using a thermistor. Small deviations in measured temperatures between the different thermocouples were compensated by a single-point calibration near the working point. In order to check the accuracy of the thermocouples, the read out was verified using an ice bath (0 ◦ C), boiling water (100 ◦ C) and a measurement at lab temperature (using a normal mercury thermometer as reference), e.g. at 0 ◦ C the measured deviations were maximal 0.1 ◦ C. The software used for the temperature read out was VI Logger (LabVIEW 6i, National Instruments). The measurement frequency was 1 Hz [9]. At this frequency, the noise on the measured temperature values was around ±0.1 ◦ C (see supplementary information). As a result of the single-point calibration and experimental verification, the experimentally measured temperatures are expected to be accurate within at least ±0.5 ◦ C, which is sufficient as the goal of this study is mainly to investigate the relative variation in temperature along the column. The used thermostatted compartment was a still-air oven. Due to heat losses from the oven to the surroundings, temperature differences were measured inside the oven. To minimize this effect on the results, the oven was further optimized by placing thermal insulation inside the oven and thus reducing heat loss to the environment (e.g. the switch valve was replaced with thermal insulation). This diminished temperature differences inside the oven to maximal 1 ◦ C. 3. Results and discussion 3.1. Possible temperature gradients in SFC For both the adiabatic or isothermal case, the sign of the term ‘1 − ˛ · T’ in Eqs. (1) and (3) determines the direction of the temperature gradient. In Fig. 1, this term is plotted as a function of pressure
214
R.D. Pauw et al. / J. Chromatogr. A 1365 (2014) 212–218
Fig. 2. Schematic representation of the placement of the thermocouples (TC) on the column and capillary walls. TCs 3 and 5 were placed at roughly 5 mm from the column fitting towards the center.
and vol% modifier methanol in CO2 at 40 ◦ C. As can be seen from this figure, a significant decompression cooling effect is expected for neat CO2 at pressures below 300 bar, i.e. at 150 bar and 300 bar respectively, (1 − ˛ · T) ≈ −2 and −0.33. As the pressure increases, the value of ˛ decreases leading to an increasing importance of viscous heating with respect to decompression cooling. As a result, the overall cooling effect vanishes at a pressure of 440 bar, as the term (1 − ˛ · T) turns positive. Beyond this pressure, a heating effect will be observed even for neat CO2 . When a modifier is added to CO2 , the resulting mixture converges toward a more liquid-like behavior. As a result, the pressure at which the term changes sign decreases from 440 to 360 bar when going from neat CO2 to 8 vol% methanol and drops to 150 bar at roughly 25 vol% in a more or less linear fashion. Above 25 vol% methanol the term is positive over the entire studied pressure range (P > 150 bar) and thus only heating should be expected as resulting total effect. The isoline where 1 − ˛ · T = 0 is plotted in Fig. 1 in black to illustrate this. As shown in literature there are infinite couples of temperature and pressure where heating and cooling cancel each other out, e.g. at 30 ◦ C and 50 ◦ C, the term (1 − ˛ · T) changes sign at 389 bar and 443 bar for neat CO2 [17]. It is therefore important to notice that the observations in this work are valid for an oven temperature of 40 ◦ C. Whereas it can be expected that qualitatively the observations remain similar around this temperature (e.g. the 30–50 ◦ C range), the strong temperature dependency of the mobile phase, especially for neat CO2 and at low pressure, can result in quantitatively very different results at lower (10 ◦ C) or higher (100 ◦ C) temperatures, as shown in [16]. Interesting to note is that mature instrumentation, with a pressure limit of 450 bar, operates in a region generally governed by decompression cooling (below the isoline). However, current stateof-the-art instrumentation with a pressure limitation of 600 bar expands toward the region where both cooling and heating effects may be observed, and hence counters each other. At a even higher pressure range (1000 bar), a new area may be found where heating effects could be dominating. 3.2. Temperature gradients by changing the flow rate Temperature profiles are measured by point-welding thermocouples (TC) on the outer column surface [9]. Fig. 2 shows the locations of the different thermocouples used and the corresponding numbers which will be used later on in the study. Note that the location of the TC is not equidistant due to geometrical limitations of the column. They were placed in such a way that the temperature was measured on the inlet capillary (TC 1), column end fittings (TC 2 and 6), two on the column close to the end fittings (TC 3 and 5) and one in the middle of the column (TC 4). In addition, temperature measured on the column wall, though not exactly representing the temperature in the column bed still gives a good indication (temperature gradients will be larger in the column) [15,18,22]. The influence of changing the flow rate (8 vol% methanol) is shown as a function of the average column pressure in Fig. 3 for a 2.1 mm × 150 mm HILIC RRHD column. With increasing flow rate, the inlet pressure increases and thus also the average column pressure. First of all, the constant temperature value of TC 1 illustrates that in the experimental set-up the mobile phase temperature is not
Fig. 3. Temperature profiles at different flow rates (from 0.6 to 3.9 ml/min in steps of 0.3 ml/min) plotted versus the average column pressure measured on the 2.1 mm × 150 mm column with a 8 v% methanol mobile phase and a back pressure of 150 bar. Placement of the thermocouples as shown in Fig. 2.
influenced by the increased flow rate or pump pressure. In addition, from 0.6 to 1.8 ml/min TCs 5 and 6 show a decreasing temperature profile. This is a result of the increase in pressure drop with flow rate and since (1 − ˛ · T) < 0, a decompression cooling effect is observed (cf. Eq. (1)). However, a monotonously increasing temperature profile is observed in the middle and front of the column (TCs 3 and 4). This is a consequence of the fact that at higher pressures the term (1 − ˛ · T) switches sign and thus leads to an overall heating effect. Logically, this effect increases with increasing flow rate as the local pressure in the column increases (cf. Fig. 1). Due to a significant heating in the first part of the column, it turns out that at higher flow rates (>1.8 ml/min) the outlet temperature (TCs 5 and 6) starts to increase. This is probably a combined effect of (1) heat flow from the front and middle of the column toward the back via the column wall and (2) the flow of a warmer mobile phase toward the end of the column (it is heated in the front). In Fig. 4b, the data is transformed to provide an axial temperature profile along the column for the three selected flow rates, as denoted by the ellipses in Fig. 3. At a flow rate of 3.9 ml/min, the increasing temperature can clearly be seen in the front and the middle of the column. After TC 4, a cooling effect is observed due to the larger decompression effect at lower pressures near the end of the column. 3.3. Influence of the mobile phase composition In Fig. 4, the column temperature profiles at 0.6, 2.4 and 3.9 ml/min are plotted for 0 and 8 vol%, and at 0.6, 2.4 and 3.3 ml/min for 20 vol%. As expected from the observation of Fig. 1, the temperature change is the largest for neat CO2 and the smallest for 20 vol% methanol at low flow rate. At this flow rate, 1 − ˛ · T is the largest in absolute values (smaller than 0) for neat CO2 and thus leads to the largest decompression cooling. The slight temperature increase near the end of the column (TCs 5 and 6) is due to the more efficient heat transfer from the still-air oven to the large surface end fitting (TC 6). This is no longer observed for the 2.4 ml/min case (Fig. 4) as the total heat transfer to the column from the oven is, in a first approach, constant and the mobile phase mass flow, which needs to be heated, is much larger. For the highest flow rate (3.9 ml/min), viscous heating is clearly dominant for the first part of the column (P > 450 bar) causing an increase in temperature, followed by a decrease in the second part of the column where the decompression cooling effect is larger. As expected from Fig. 1, the observed temperature increase in the first part of the column increases with higher modifier content, and leads to a decrease in cooling effect near the end (see Fig. 4a and c). In addition, since less decompression cooling takes place at the end of the column for the higher methanol fraction cases, less heat will be conducted away
R.D. Pauw et al. / J. Chromatogr. A 1365 (2014) 212–218
215
Fig. 4. Influence of flow rate (0.6, 2.4 and 3.9 ml/min) on the column temperature profile for different compositions: (a) 0 v%, (b) 8 v%, (c) 20 v% methanol (note that here the maximal flow rate was 3.3 ml/min). Placement of the thermocouples as shown in Fig. 2. Note: the connecting lines do not represent experimental or fitting data but were added to guide the eye.
from TC 4 via the column wall. For 2.4 ml/min, an intermediate behavior between the 0.6 and 3.9 ml/min cases is observed.
composition for a CO2 /methanol mixture, it is highly dependent on the operating conditions which of the two effects will be dominant. 3.5. Temperature gradient over two coupled columns
3.4. Influence of the back pressure As changing the flow rate not only increases the average pressure but also the pressure drop across the column, it is difficult to isolate the effect of pressure on the thermal expansion coefficient using this approach. Changing the back pressure at a fixed flow rate, however, is a convenient way to isolate the effect of pressure. Fig. 5 shows the influence of the back pressure on the temperature distribution along the column at a constant flow rate of 1.8 ml/min for the three studied compositions. Increasing the back pressure leads to an increase in column pressure drop of only 7% for neat CO2 from 150 to 300 bar back pressure, as a result of the higher mobile phase viscosity at higher pressure. When increasing the back pressure for neat CO2 (Fig. 5a) to 300 bar, the temperature is almost constant along the column axis. Under these conditions, the decompression cooling is almost entirely canceled out by the viscous heating. For increasing fractions of methanol (Fig. 5b and c), a clear decrease of the decompression cooling is observed at 150 bar, whereas a dominance of the viscous heating effect is observed at higher back pressures. For example for 20 vol% at 300 bar, the temperature at TC 5 is roughly 42.6 ◦ C, whereas only 40.1 ◦ C for 8 vol% and 39.1 ◦ C for neat CO2 . Fig. 5 hence clearly shows that in SFC, indeed both decompression cooling and viscous heating can be observed. Since the thermal expansion coefficient is a strong function of inlet pressure and
Whereas previous sections dealt with only one column, using coupled columns, which e.g. is done to achieve high efficiencies or combine different selectivities [23–27], leads toward even more elaborate temperature profiles. Fig. 6 shows the temperature profile along two coupled 2.1 mm × 150 mm columns for 0, 8 and 20 vol% methanol at a flow rate of 1.8 ml/min and a back pressure of 150 bar. The locations on the second column are denoted by a but are otherwise the same as for the first column. As can be seen, the decompression cooling effect is the largest for neat CO2 at the end of the second column. The first column however exhibits a more or less constant temperature profile. This is due to the fact that the outlet pressure of the first column is 418 bar, which is in the region of almost no net heating or cooling effects for neat CO2 (1 − ˛ · T ≈ 0). Adding modifier, however, results in a more complicated temperature profile. For 20 vol% methanol, the first column now shows a strong increase in column temperature toward the end. Due to these elevated pressures, the mobile phase leans toward LC-like behavior (e.g. dominant viscous heating). However, the temperature gradient is smaller than what would be expected in LC. In the second column there is clearly a combined effect of viscous heating in the front of the column with a decompression cooling near the end of the column. The coupling of the column also inherently causes a partial ‘reset’ of the temperature effect between the columns [9,23], due
216
R.D. Pauw et al. / J. Chromatogr. A 1365 (2014) 212–218
Fig. 5. Temperature profile over a column for different back pressures and different compositions: (a) 0 v%, (b) 8 v%, (c) 20 v% methanol. Placement of the thermocouples as shown in Fig. 2.
to the very efficient intermediate heat exchange at the connection capillary having a larger surface-to-volume ratio (as illustrated by the 40 ± 1 ◦ C temperature for TC 1 ). The fact that the observed temperature gradients are rather small, indicates that SFC separations could be performed at high inlet pressures (high column pressure drops), without experiencing excessive band broadening as a result of thermally induced velocity gradients inside the column. The high pressure drops would of course only be reached when using long coupled columns and in this case it appears that the occurring temperature gradients are largely eliminated in the connections. The operation at these elevated pressures and resulting higher mobile phase density will
46 44
however induce several other pressure related effects (e.g. increasing viscosity, decreasing molecular diffusion and retention) which might affect chromatographic performance. 3.6. Influence of the column ID Fig. 7 compares two 15 cm HILIC RRHD columns with different IDs at a constant column pressure drop and used in a still-air oven: 2.1 mm (full symbols) and 3 mm (open symbols). As can be seen, only small differences in the temperature are measured for a wider column. This illustrates that the thermal behavior is predominantly
0 v% 8 v% 20 v%
T (°C)
42 40 38 36 34
1
2
3
4
5
6 1' 2' Position
3'
4'
5'
6'
Fig. 6. Temperature profile over two coupled columns at different amounts of methanol modifier: position on column as shown in Fig. 2, position on second column is denoted by . At a flow rate of 1.8 ml/min and a back pressure of 150 bar.
Fig. 7. Influence of the column ID for 15 cm HILIC RRHD columns at a constant pressure drop for three different compositions: 2.1 mm (full symbols) and 3 mm (open symbols). Inlet pressure is equal to 520, 540 and 649 bar for 0, 8 and 20 v% MeOH respectively, for a flow rate of 2.4 ml/min for the 2.1 mm ID and 4.7 ml/min for the 3 mm ID column, and with a back pressure of 150 bar.
R.D. Pauw et al. / J. Chromatogr. A 1365 (2014) 212–218
44
T (°C)
42
4. Conclusions
0 v% 8 v% 20 v%
40
38
36
34
1
217
2
3 4 Position
5
6
Fig. 8. Influence of the oven type for a 3 mm × 150 mm HILIC RRHD column for three different compositions: still-air (full symbols) and forced-air (open symbols). Same conditions as in Fig. 7.
adiabatically controlled with only limited effect of the heat flux to/from the environment. Whereas for 0 vol% the temperature decrease due to decompression cooling is larger at the end of the column for the 3 mm ID column, also the temperature increase due to the viscous heating in the middle of the column is larger for 20 v% MeOH for the wider column. This larger temperature gradient for the larger ID columns is the result of a more adiabatic behavior (less isothermal) of the larger ID column due to a smaller heat flow from column to oven relative to its volume. This is also in agreement with Eq. (3), showing a decrease in temperature gradient with column radius. For both column IDs the temperature changes are relatively small at these extended pressure drops (e.g. an inlet pressure of 650 bar for 20 vol%).
3.7. Still- versus forced-air oven The temperature profiles are compared for two different oven types on a 3 mm × 150 mm column: still-air (full symbols) and forced-air (open symbols). The forced-air oven was created by placing a ventilator inside a still-air oven. The TCs were covered by a small amount of insulation foam in order to measure the column wall temperature without being disturbed by the forced-air convection. The ventilator was placed at the location of the switch valve in the oven, the column was placed central in the oven (in front of ventilator) in order to have the convective air flows fully surrounding it. The results are shown in Fig. 8. As expected, when applying forced air the temperature profiles are more smooth and the system converges more toward the isothermal case. For neat CO2 , the temperature drop at the end of the column due to decompression cooling is reduced whereas the heating effect for 20 v% is almost completely removed. The effect of the ventilator is even more apparent for the 20 v% case at TC 4, as the ventilator was placed near the center of the column. The use of the ventilator had only a very minor effect on column performance (results not shown), which is a result of the rather small temperature gradient observed in this study, i.e. at an operating temperature of 40 ◦ C and a back pressure of 150 bar. It is however not straightforward to choose an optimal operating mode (still versus forced air) based on this observation. In LC, retention is mainly determined by the temperature of the mobile phase, whereas in SFC both temperature and mobile phase density play an important role [22]. As a result, the increase in temperature (lower retention) along the column can in some cases offset the decrease in density (higher retention), making the choice in optimal operating mode difficult to generalize.
Temperature gradients in both liquid chromatography and supercritical fluid chromatography have been of interest due to the fact that excessive radial temperature profiles can cause excessive extra-band broadening. Whereas in LC, a temperature increase is expected and observed due to the dominant viscous heating, it is commonly known that a temperature decrease takes place under most SFC conditions due to decompression cooling. However, the same fundamental principles apply for both techniques. Both decompression cooling and viscous heating take place in both techniques and it is the trade-off between them that determines the resulting temperature profile. Whereas the thermal expansion coefficient of typical LC solvents is relatively small, this is not the case for SFC, where it is a strong function of pressure, temperature and mobile phase composition. For neat CO2 at 40 ◦ C, a temperature decrease due to the larger decompression cooling effect prevails below 450 bar or thus in the operating range of mature instrumentation. When going to higher pressure (e.g. the pressure rating of state-of-the-art instruments, i.e. 600 bar), viscous heating becomes more relevant. The pressure at which decompression cooling and viscous heating cancel each other out decreases almost linearly from 450 bar for neat CO2 to 150 bar for 25 vol% methanol. At low back pressure (150 bar), for pressure drops up to 300 bar, decompression cooling is observed throughout the column. However, when increasing the back pressure to 300 bar, the mobile phase heats up in the first part of the column due to the dominance of viscous heating and then cools again due to the decompression cooling at the end of the column. When coupling columns to achieve high efficiencies at very high operating pressure (e.g. 750 bar inlet pressure for 8 vol%), the situation is even more complex. E.g. at a back pressure of 150 bar and using 8 vol% methanol, viscous heating is only observed in the first column whereas only decompression cooling in the second. Whereas the combination of viscous heating and decompression cooling gives rise to complex temperature profiles along a column, increasing the inlet pressure up to 1050 bar resulted in no excessive temperature differences measured along the column (max. 5 ◦ C). This implies that the inlet pressure of SFC instrumentation could be expanded above 600 bar without additional band broadening caused by excessive temperature differences. Acknowledgements R.D.P. gratefully acknowledges a research grant from the Research Foundation – Flanders (FWO Vlaanderen). The authors gratefully acknowledge Xiaoli Wang for the kind donation of the HILIC RRHD columns. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.chroma. 2014.09.022. References [1] I. Halasz, R. Endele, J. Asshauer, Ultimate limits in high-pressure liquid chromatography, J. Chromatogr. 112 (1975) 37. [2] Cs. Horvath, H.-J. Lin, Ultimate limits in high-pressure liquid chromatography, J. Chromatogr. 149 (1978) 43. [3] H. Poppe, J.C. Kraak, J.F.K. Huber, J.H.M. Van den Berg, Temperature gradients in HPLC columns due to viscous heat dissipation, Chromatographia 14 (1981) 515. [4] H. Poppe, J.C. Kraak, J. Chromatogr. 282 (1983) 399. [5] M. Martin, G. Guiochon, Ultimate limits in high-pressure liquid chromatography, J. Chromatogr. A 1090 (2005) 16.
218
R.D. Pauw et al. / J. Chromatogr. A 1365 (2014) 212–218
[6] F. Gritti, G. Guiochon, Measurement of the axial and radial temperature profiles of a chromatographic column: influence of thermal insulation on column efficiency, J. Chromatogr. A 1138 (2007) 141. [7] U.D. Neue, M. Kele, Performance of idealized column structures under high pressure, J. Chromatogr. A 1149 (2007) 236. [8] M.M. Fallas, M.R. Hadley, D.V. McCalley, Practical assessment of frictional heating effects and thermostat design on the performance of conventional (3 m and 5 m) columns in reversed-phase high-performance liquid chromatography, J. Chromatogr. A 1216 (2009) 3961. [9] K. Broeckhoven, J. Billen, M. Verstraeten, K. Choikhet, M. Dittmann, G. Rozing, G. Desmet, Towards a solution for viscous heating in ultra-high pressure liquid chromatography using intermediate cooling, J. Chromatogr. A 1217 (2010) 2022–2031. [10] A. de Villiers, H. Lauer, R. Szucs, S. Goodall, P. Sandra, Influence of frictional heating on temperature gradients in ultra-high-pressure liquid chromatography on 2.1 mm I.D. columns, J. Chromatogr. A 1113 (2009) 84–91. [11] F. Gritti, G. Guiochon, Optimization of the thermal environment of columns packed with very fine particles, J. Chromatogr. A 1216 (2009) 1353–1362. [12] P.A. Mourier, M.H. Caude, R.H. Rosset, The dependence of reduced plate height on reduced velocity in carbon dioxide supercritical fluid chromatography with packed columns, Chromatographia 23 (1987) 21. [13] T.A. Berger, J.F. Deye, Effect of mobile phase density gradients on efficiency in packed column supercritical fluid chromatography, Chromatographia 30 (1990) 57. [14] D.P. Poe, J.J. Schroden, Effects of pressure drop, particle size and thermal conditions on retention and efficiency in supercritical fluid chromatography, J. Chromatogr. A 1216 (2009) 7915–7926. [15] K. Kaczmarski, D.P. Poe, G. Guiochon, Numerical modeling of the elution peak profiles of retained solutes in supercritical fluid chromatography, J. Chromatogr. A 1218 (2011) 6531–6539. [16] 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 (2014), http://dx.doi.org/ 10.1016/j.chroma.2014.08.084.
[17] K. Kaczmarski, D. Poe, G. Guiochon, Numerical modeling of elution peak profiles in supercritical fluid chromatography. Part I – elution of an unretained tracer, J. Chromatogr. A 1217 (2010) 6578–6587. [18] D.P. Poe, D. Veit, M. Ranger, K. Kaczmarski, A. Tarafder, G. Guiochon, Pressure, temperature and density drops along supercritical fluid chromatography columns. I. Experimental results for neat carbon dioxide and columns packed with 3- and 5-micron particles, J. Chromatogr. A 1250 (2012) 105–114. [19] J.H. Knox, I. Grant, Chromatographia 24 (1987) 135. [20] D.C. Ginnings, Precision Measurement and Calibration, Selected National Bureau of Standards Papers on Heat, NBS Special Publication 300, vol. 6, 1970. [21] REFPROP Program, Version 9.1, National Institutes of Sciences and Technology, Physical and Chemical Properties Division, Boulder, CO, USA, 2013. [22] A. Tarafder, J. Hill, P. Iraneta, Effects of Pressure and Temperature on the Retention and Selectivity of Supercritical Fluid Chromatography, presented at the 41st conference on High Pressure Liquid Chromatography and Related Techniques (HPLC2014), May 11–15, New Orleans, Louisiana, USA, 2014. [23] R. De Pauw, B. Degreef, H. Ritchie, S. Eeltink, G. Desmet, K. Broeckhoven, Extending the limits of operating pressure of narrow-bore column liquid chromatography instrumentation, J. Chromatogr. A 1347 (2014) 56–62. [24] 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. [25] F. Lestremau, A. de Villiers, F. Lynen, A. Cooper, R. Szucs, P. Sandra, High efficiency liquid chromatography on conventional columns and instrumentation by using temperature as a variable: kinetic plots and experimental verification, J. Chromatogr. A 1138 (2007) 120–131. [26] A. Vaast, J. De Vos, K. Broeckhoven, M. Verstraeten, S. Eeltink, G. Desmet, Maximizing the peak capacity using coupled columns packed with 2.6 m core-shell particles operated at 1200 bar, J. Chromatogr. A 1256 (2012) 72–79. [27] A. de Villiers, F. Lestremau, R. Szucs, S. Gelebart, F. David, P. Sandra, Evaluation of ultra performance liquid chromatography. Part I. Possibilities and limitations, J. Chromatogr. A 1127 (2006) 60–69.
Contents lists available at ScienceDirect
Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma
Temperature effects in supercritical fluid chromatography: A trade-off between viscous heating and decompression cooling Ruben De Pauw a , Konstantin 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 14 July 2014 Received in revised form 5 September 2014 Accepted 8 September 2014 Available online 17 September 2014 Keywords: Supercritical fluid chromatography Viscous heating Decompression cooling Temperature Pressure Coupled columns
a b s t r a c t The study of radial and axial temperature profiles always has been an area interest both in liquid chromatography (LC) and supercritical fluid chromatography (SFC). Whereas in LC always an increase in temperature is observed due to the dominance of viscous heating, in SFC, especially for low modifier content, a decrease in temperature is found due to the much larger decompression cooling. However, for higher modifier content and higher operating pressure, the temperature effects become a trade-off between viscous heating and decompression cooling, since in SFC the latter is a strong function of operating pressure and mobile phase composition. At a temperature of 40 ◦ C and for neat CO2 , the effect of decompression cooling and viscous heating cancel each other out at a pressure 450 bar. This pressure decreases almost linearly with volume fraction of methanol to 150 bar at 25 vol%. As a result, both cooling and heating effects can be observed when operating at high back pressure, large column pressure drops or high modifier content. For example at a back pressure of 150 bar and a column pressure drop of 270 bar decompression cooling is observed throughout the column. However at 300 bar back pressure and the same pressure drop, the mobile phase heats up in the first part of the column due to viscous heating and then cools in the second part due to decompression cooling. When coupling columns (2.1 mm × 150 mm, 1.8 m fully porous particles) at very high operating pressure (e.g. 750 bar for 8 vol%), the situation is even more complex. E.g. at a back pressure of 150 bar and using 8 vol% methanol, viscous heating is only observed in the first column whereas only decompression cooling in the second. Further increasing the inlet pressure up to 1050 bar resulted in no excessive temperature differences along the column. This implies that the inlet pressure of SFC instrumentation could be expanded above 600 bar without additional band broadening caused by excessive radial temperature differences. © 2014 Elsevier B.V. All rights reserved.
1. Introduction In liquid chromatography (LC), the presence and consequences of radial and axial temperature profiles were discussed four decades ago and were extensively studied on a theoretical and experimental basis by Poppe et al. [1–4]. The problem received a renewed attention in the last decade with the introduction of ultrahigh pressure liquid chromatography (UHPLC) [5–10]. In UHPLC, a significant temperature increase of the mobile phase is observed and expected due to viscous heating. Since large axial temperature profiles lead to radial temperature profiles and thus radial velocity differences [11], a maximal pressure limit for commercial
∗ Corresponding author. Tel.: +32 26293781; fax: +32 26293248. E-mail address: [email protected] (K. Broeckhoven). http://dx.doi.org/10.1016/j.chroma.2014.09.022 0021-9673/© 2014 Elsevier B.V. All rights reserved.
instrumentation of roughly 1200–1300 bar appeared to be established, which was recently increased to 1500 bar. In supercritical fluid chromatography (SFC) similar examinations were performed by several authors, showing that excessive performance loss can be expected when the outlet pressure and temperature are near the critical point [12,13]. Recent papers showed that these performance losses were mainly due to radial temperature profiles as a result of significant decompression cooling [14,15]. Even though in SFC and LC opposite temperature gradient profiles are observed, the same basic principles are valid and determine the resulting temperature gradient. Two main effects can be identified: viscous heating and decompression cooling. Whereas viscous heating is the heating of the mobile phase due to friction with the column bed, decompression of the mobile phase leads to a decrease in temperature since energy is absorbed to overcome attractive forces between molecules of the mobile phase [16]. These two
R.D. Pauw et al. / J. Chromatogr. A 1365 (2014) 212–218
213
effects can be combined into a single equation, in the ideal case of an adiabatic column (no heat transfer to environment) [5] Tadi = −(1 − ˛ · T ) ·
P Cp,V
(1)
where Tadi , T, P and Cp,V are respectively the temperature change in adiabatic conditions, the temperature, pressure drop over the column and the specific volumetric heat capacity of the mobile phase. Please note that the gradients are defined from inlet to outlet, hence P is always smaller than 0 and T > 0 when a temperature increase is observed. ˛ is the coefficient of thermal expansion and is related to the density, via [17,18] 1 ˛=−
∂ ∂T
(2) P
The first term in Eq. (1), i.e. −P/Cp,V , is positive (P = Pout − Pin < 0), leading to a temperature increase and is thus related to the viscous heating, whereas the second term ((˛ · TP)/Cp,V ) opposes this increase (˛ usually has a positive sign). The latter term is thus the decompression cooling. Since ˛ is a function of pressure, temperature and mobile phase composition (volume-based, ), a resulting increase or decrease in temperature will depend on the experimental conditions. In case of a thermostatted column, an isothermal assumption is more valid (column wall at fixed temperature). The temperature change (averaged out over the cross-section of the column) can then be written as [9,19] Tiso = −(1 − ˛ · T ) ·
us · P · R2 8 · rad · L
(3)
with us , L, R and rad respectively the superficial velocity, the column length, column inner radius and the radial thermal conductivity of the packed bed. The amplitude of the radial temperature difference is found by changing the factor of 8 by 4 in Eq.(3). In this work, thermal gradients across columns are further investigated in SFC conditions, up to the pressure limits of novel SFC instrumentation (600 bar) and beyond (up till 1050 bar). In addition, the influence of the column diameter and oven type (still-air versus forced-air) is considered. 2. Experimental 2.1. Column, tubing and chemicals Methanol (LC–MS grade) was purchased from Biosolve (Valkenswaard, The Netherlands), CO2 was purchased from Air Liquide (Paris, France). HILIC RRHD columns, packed with 1.8 m bare-silica Zorbax particles, were kindly provided by Xiaoli Wang (Agilent Technologies, Little Falls, USA) with different column lengths (10 and 15 cm) and with different IDs (2.1 and 3 mm). 2.2. Instrumentation and conditions The SFC system used in the study was a modified Agilent G4301A-based SFC system with an extended column pressure range up to 1050 bar in combination with a thermostatted column compartment, autosampler with a 1.2 L injection loop (used in full loop mode) and a DAD-detector with a 1.7 L flow cell. For all the experiments the back pressure regulator was set at 150 bar (unless otherwise specified) and the oven temperature at 40 ◦ C. Other oven temperatures may give different quantitative results but will yield the same qualitative results.
Fig. 1. The term ‘(1 − ˛ · T)’ (see Eq. (1)) as a function of pressure and mobile phase composition (methanol as modifier). Using the data for ˛ provided by NIST REFPROP [21] for a temperature of 40 ◦ C. The black line denotes where viscous heating and decompression cooling cancel each other or where (1 − ˛ · T) = 0.
2.3. Temperature measurements Temperature measurements were performed on the columns and connection capillaries using chromal-alumel thermocouples (K-type thermocouple wire) positioned as indicated in Fig. 2. To provide a perfect thermal contact, the two separate wires of the thermocouple were point-welded on top of each other and directly onto the column or capillaries at the measurement spot location using a device built in-house. The almost perfect thermal contact achieved by point-welding the thermocouple on the column resulted in a very accurate measurement of the temperature on the surface of the column, without the need for thermal tempering [20], as was shown experimentally in the supplementary information. The voltages were read out using a TBX-68T Isothermal Terminal Block (National Instruments, USA) and a NI 435X (National Instruments, USA) PCI card. The local temperature inside the thermal block was measured using a thermistor. Small deviations in measured temperatures between the different thermocouples were compensated by a single-point calibration near the working point. In order to check the accuracy of the thermocouples, the read out was verified using an ice bath (0 ◦ C), boiling water (100 ◦ C) and a measurement at lab temperature (using a normal mercury thermometer as reference), e.g. at 0 ◦ C the measured deviations were maximal 0.1 ◦ C. The software used for the temperature read out was VI Logger (LabVIEW 6i, National Instruments). The measurement frequency was 1 Hz [9]. At this frequency, the noise on the measured temperature values was around ±0.1 ◦ C (see supplementary information). As a result of the single-point calibration and experimental verification, the experimentally measured temperatures are expected to be accurate within at least ±0.5 ◦ C, which is sufficient as the goal of this study is mainly to investigate the relative variation in temperature along the column. The used thermostatted compartment was a still-air oven. Due to heat losses from the oven to the surroundings, temperature differences were measured inside the oven. To minimize this effect on the results, the oven was further optimized by placing thermal insulation inside the oven and thus reducing heat loss to the environment (e.g. the switch valve was replaced with thermal insulation). This diminished temperature differences inside the oven to maximal 1 ◦ C. 3. Results and discussion 3.1. Possible temperature gradients in SFC For both the adiabatic or isothermal case, the sign of the term ‘1 − ˛ · T’ in Eqs. (1) and (3) determines the direction of the temperature gradient. In Fig. 1, this term is plotted as a function of pressure
214
R.D. Pauw et al. / J. Chromatogr. A 1365 (2014) 212–218
Fig. 2. Schematic representation of the placement of the thermocouples (TC) on the column and capillary walls. TCs 3 and 5 were placed at roughly 5 mm from the column fitting towards the center.
and vol% modifier methanol in CO2 at 40 ◦ C. As can be seen from this figure, a significant decompression cooling effect is expected for neat CO2 at pressures below 300 bar, i.e. at 150 bar and 300 bar respectively, (1 − ˛ · T) ≈ −2 and −0.33. As the pressure increases, the value of ˛ decreases leading to an increasing importance of viscous heating with respect to decompression cooling. As a result, the overall cooling effect vanishes at a pressure of 440 bar, as the term (1 − ˛ · T) turns positive. Beyond this pressure, a heating effect will be observed even for neat CO2 . When a modifier is added to CO2 , the resulting mixture converges toward a more liquid-like behavior. As a result, the pressure at which the term changes sign decreases from 440 to 360 bar when going from neat CO2 to 8 vol% methanol and drops to 150 bar at roughly 25 vol% in a more or less linear fashion. Above 25 vol% methanol the term is positive over the entire studied pressure range (P > 150 bar) and thus only heating should be expected as resulting total effect. The isoline where 1 − ˛ · T = 0 is plotted in Fig. 1 in black to illustrate this. As shown in literature there are infinite couples of temperature and pressure where heating and cooling cancel each other out, e.g. at 30 ◦ C and 50 ◦ C, the term (1 − ˛ · T) changes sign at 389 bar and 443 bar for neat CO2 [17]. It is therefore important to notice that the observations in this work are valid for an oven temperature of 40 ◦ C. Whereas it can be expected that qualitatively the observations remain similar around this temperature (e.g. the 30–50 ◦ C range), the strong temperature dependency of the mobile phase, especially for neat CO2 and at low pressure, can result in quantitatively very different results at lower (10 ◦ C) or higher (100 ◦ C) temperatures, as shown in [16]. Interesting to note is that mature instrumentation, with a pressure limit of 450 bar, operates in a region generally governed by decompression cooling (below the isoline). However, current stateof-the-art instrumentation with a pressure limitation of 600 bar expands toward the region where both cooling and heating effects may be observed, and hence counters each other. At a even higher pressure range (1000 bar), a new area may be found where heating effects could be dominating. 3.2. Temperature gradients by changing the flow rate Temperature profiles are measured by point-welding thermocouples (TC) on the outer column surface [9]. Fig. 2 shows the locations of the different thermocouples used and the corresponding numbers which will be used later on in the study. Note that the location of the TC is not equidistant due to geometrical limitations of the column. They were placed in such a way that the temperature was measured on the inlet capillary (TC 1), column end fittings (TC 2 and 6), two on the column close to the end fittings (TC 3 and 5) and one in the middle of the column (TC 4). In addition, temperature measured on the column wall, though not exactly representing the temperature in the column bed still gives a good indication (temperature gradients will be larger in the column) [15,18,22]. The influence of changing the flow rate (8 vol% methanol) is shown as a function of the average column pressure in Fig. 3 for a 2.1 mm × 150 mm HILIC RRHD column. With increasing flow rate, the inlet pressure increases and thus also the average column pressure. First of all, the constant temperature value of TC 1 illustrates that in the experimental set-up the mobile phase temperature is not
Fig. 3. Temperature profiles at different flow rates (from 0.6 to 3.9 ml/min in steps of 0.3 ml/min) plotted versus the average column pressure measured on the 2.1 mm × 150 mm column with a 8 v% methanol mobile phase and a back pressure of 150 bar. Placement of the thermocouples as shown in Fig. 2.
influenced by the increased flow rate or pump pressure. In addition, from 0.6 to 1.8 ml/min TCs 5 and 6 show a decreasing temperature profile. This is a result of the increase in pressure drop with flow rate and since (1 − ˛ · T) < 0, a decompression cooling effect is observed (cf. Eq. (1)). However, a monotonously increasing temperature profile is observed in the middle and front of the column (TCs 3 and 4). This is a consequence of the fact that at higher pressures the term (1 − ˛ · T) switches sign and thus leads to an overall heating effect. Logically, this effect increases with increasing flow rate as the local pressure in the column increases (cf. Fig. 1). Due to a significant heating in the first part of the column, it turns out that at higher flow rates (>1.8 ml/min) the outlet temperature (TCs 5 and 6) starts to increase. This is probably a combined effect of (1) heat flow from the front and middle of the column toward the back via the column wall and (2) the flow of a warmer mobile phase toward the end of the column (it is heated in the front). In Fig. 4b, the data is transformed to provide an axial temperature profile along the column for the three selected flow rates, as denoted by the ellipses in Fig. 3. At a flow rate of 3.9 ml/min, the increasing temperature can clearly be seen in the front and the middle of the column. After TC 4, a cooling effect is observed due to the larger decompression effect at lower pressures near the end of the column. 3.3. Influence of the mobile phase composition In Fig. 4, the column temperature profiles at 0.6, 2.4 and 3.9 ml/min are plotted for 0 and 8 vol%, and at 0.6, 2.4 and 3.3 ml/min for 20 vol%. As expected from the observation of Fig. 1, the temperature change is the largest for neat CO2 and the smallest for 20 vol% methanol at low flow rate. At this flow rate, 1 − ˛ · T is the largest in absolute values (smaller than 0) for neat CO2 and thus leads to the largest decompression cooling. The slight temperature increase near the end of the column (TCs 5 and 6) is due to the more efficient heat transfer from the still-air oven to the large surface end fitting (TC 6). This is no longer observed for the 2.4 ml/min case (Fig. 4) as the total heat transfer to the column from the oven is, in a first approach, constant and the mobile phase mass flow, which needs to be heated, is much larger. For the highest flow rate (3.9 ml/min), viscous heating is clearly dominant for the first part of the column (P > 450 bar) causing an increase in temperature, followed by a decrease in the second part of the column where the decompression cooling effect is larger. As expected from Fig. 1, the observed temperature increase in the first part of the column increases with higher modifier content, and leads to a decrease in cooling effect near the end (see Fig. 4a and c). In addition, since less decompression cooling takes place at the end of the column for the higher methanol fraction cases, less heat will be conducted away
R.D. Pauw et al. / J. Chromatogr. A 1365 (2014) 212–218
215
Fig. 4. Influence of flow rate (0.6, 2.4 and 3.9 ml/min) on the column temperature profile for different compositions: (a) 0 v%, (b) 8 v%, (c) 20 v% methanol (note that here the maximal flow rate was 3.3 ml/min). Placement of the thermocouples as shown in Fig. 2. Note: the connecting lines do not represent experimental or fitting data but were added to guide the eye.
from TC 4 via the column wall. For 2.4 ml/min, an intermediate behavior between the 0.6 and 3.9 ml/min cases is observed.
composition for a CO2 /methanol mixture, it is highly dependent on the operating conditions which of the two effects will be dominant. 3.5. Temperature gradient over two coupled columns
3.4. Influence of the back pressure As changing the flow rate not only increases the average pressure but also the pressure drop across the column, it is difficult to isolate the effect of pressure on the thermal expansion coefficient using this approach. Changing the back pressure at a fixed flow rate, however, is a convenient way to isolate the effect of pressure. Fig. 5 shows the influence of the back pressure on the temperature distribution along the column at a constant flow rate of 1.8 ml/min for the three studied compositions. Increasing the back pressure leads to an increase in column pressure drop of only 7% for neat CO2 from 150 to 300 bar back pressure, as a result of the higher mobile phase viscosity at higher pressure. When increasing the back pressure for neat CO2 (Fig. 5a) to 300 bar, the temperature is almost constant along the column axis. Under these conditions, the decompression cooling is almost entirely canceled out by the viscous heating. For increasing fractions of methanol (Fig. 5b and c), a clear decrease of the decompression cooling is observed at 150 bar, whereas a dominance of the viscous heating effect is observed at higher back pressures. For example for 20 vol% at 300 bar, the temperature at TC 5 is roughly 42.6 ◦ C, whereas only 40.1 ◦ C for 8 vol% and 39.1 ◦ C for neat CO2 . Fig. 5 hence clearly shows that in SFC, indeed both decompression cooling and viscous heating can be observed. Since the thermal expansion coefficient is a strong function of inlet pressure and
Whereas previous sections dealt with only one column, using coupled columns, which e.g. is done to achieve high efficiencies or combine different selectivities [23–27], leads toward even more elaborate temperature profiles. Fig. 6 shows the temperature profile along two coupled 2.1 mm × 150 mm columns for 0, 8 and 20 vol% methanol at a flow rate of 1.8 ml/min and a back pressure of 150 bar. The locations on the second column are denoted by a but are otherwise the same as for the first column. As can be seen, the decompression cooling effect is the largest for neat CO2 at the end of the second column. The first column however exhibits a more or less constant temperature profile. This is due to the fact that the outlet pressure of the first column is 418 bar, which is in the region of almost no net heating or cooling effects for neat CO2 (1 − ˛ · T ≈ 0). Adding modifier, however, results in a more complicated temperature profile. For 20 vol% methanol, the first column now shows a strong increase in column temperature toward the end. Due to these elevated pressures, the mobile phase leans toward LC-like behavior (e.g. dominant viscous heating). However, the temperature gradient is smaller than what would be expected in LC. In the second column there is clearly a combined effect of viscous heating in the front of the column with a decompression cooling near the end of the column. The coupling of the column also inherently causes a partial ‘reset’ of the temperature effect between the columns [9,23], due
216
R.D. Pauw et al. / J. Chromatogr. A 1365 (2014) 212–218
Fig. 5. Temperature profile over a column for different back pressures and different compositions: (a) 0 v%, (b) 8 v%, (c) 20 v% methanol. Placement of the thermocouples as shown in Fig. 2.
to the very efficient intermediate heat exchange at the connection capillary having a larger surface-to-volume ratio (as illustrated by the 40 ± 1 ◦ C temperature for TC 1 ). The fact that the observed temperature gradients are rather small, indicates that SFC separations could be performed at high inlet pressures (high column pressure drops), without experiencing excessive band broadening as a result of thermally induced velocity gradients inside the column. The high pressure drops would of course only be reached when using long coupled columns and in this case it appears that the occurring temperature gradients are largely eliminated in the connections. The operation at these elevated pressures and resulting higher mobile phase density will
46 44
however induce several other pressure related effects (e.g. increasing viscosity, decreasing molecular diffusion and retention) which might affect chromatographic performance. 3.6. Influence of the column ID Fig. 7 compares two 15 cm HILIC RRHD columns with different IDs at a constant column pressure drop and used in a still-air oven: 2.1 mm (full symbols) and 3 mm (open symbols). As can be seen, only small differences in the temperature are measured for a wider column. This illustrates that the thermal behavior is predominantly
0 v% 8 v% 20 v%
T (°C)
42 40 38 36 34
1
2
3
4
5
6 1' 2' Position
3'
4'
5'
6'
Fig. 6. Temperature profile over two coupled columns at different amounts of methanol modifier: position on column as shown in Fig. 2, position on second column is denoted by . At a flow rate of 1.8 ml/min and a back pressure of 150 bar.
Fig. 7. Influence of the column ID for 15 cm HILIC RRHD columns at a constant pressure drop for three different compositions: 2.1 mm (full symbols) and 3 mm (open symbols). Inlet pressure is equal to 520, 540 and 649 bar for 0, 8 and 20 v% MeOH respectively, for a flow rate of 2.4 ml/min for the 2.1 mm ID and 4.7 ml/min for the 3 mm ID column, and with a back pressure of 150 bar.
R.D. Pauw et al. / J. Chromatogr. A 1365 (2014) 212–218
44
T (°C)
42
4. Conclusions
0 v% 8 v% 20 v%
40
38
36
34
1
217
2
3 4 Position
5
6
Fig. 8. Influence of the oven type for a 3 mm × 150 mm HILIC RRHD column for three different compositions: still-air (full symbols) and forced-air (open symbols). Same conditions as in Fig. 7.
adiabatically controlled with only limited effect of the heat flux to/from the environment. Whereas for 0 vol% the temperature decrease due to decompression cooling is larger at the end of the column for the 3 mm ID column, also the temperature increase due to the viscous heating in the middle of the column is larger for 20 v% MeOH for the wider column. This larger temperature gradient for the larger ID columns is the result of a more adiabatic behavior (less isothermal) of the larger ID column due to a smaller heat flow from column to oven relative to its volume. This is also in agreement with Eq. (3), showing a decrease in temperature gradient with column radius. For both column IDs the temperature changes are relatively small at these extended pressure drops (e.g. an inlet pressure of 650 bar for 20 vol%).
3.7. Still- versus forced-air oven The temperature profiles are compared for two different oven types on a 3 mm × 150 mm column: still-air (full symbols) and forced-air (open symbols). The forced-air oven was created by placing a ventilator inside a still-air oven. The TCs were covered by a small amount of insulation foam in order to measure the column wall temperature without being disturbed by the forced-air convection. The ventilator was placed at the location of the switch valve in the oven, the column was placed central in the oven (in front of ventilator) in order to have the convective air flows fully surrounding it. The results are shown in Fig. 8. As expected, when applying forced air the temperature profiles are more smooth and the system converges more toward the isothermal case. For neat CO2 , the temperature drop at the end of the column due to decompression cooling is reduced whereas the heating effect for 20 v% is almost completely removed. The effect of the ventilator is even more apparent for the 20 v% case at TC 4, as the ventilator was placed near the center of the column. The use of the ventilator had only a very minor effect on column performance (results not shown), which is a result of the rather small temperature gradient observed in this study, i.e. at an operating temperature of 40 ◦ C and a back pressure of 150 bar. It is however not straightforward to choose an optimal operating mode (still versus forced air) based on this observation. In LC, retention is mainly determined by the temperature of the mobile phase, whereas in SFC both temperature and mobile phase density play an important role [22]. As a result, the increase in temperature (lower retention) along the column can in some cases offset the decrease in density (higher retention), making the choice in optimal operating mode difficult to generalize.
Temperature gradients in both liquid chromatography and supercritical fluid chromatography have been of interest due to the fact that excessive radial temperature profiles can cause excessive extra-band broadening. Whereas in LC, a temperature increase is expected and observed due to the dominant viscous heating, it is commonly known that a temperature decrease takes place under most SFC conditions due to decompression cooling. However, the same fundamental principles apply for both techniques. Both decompression cooling and viscous heating take place in both techniques and it is the trade-off between them that determines the resulting temperature profile. Whereas the thermal expansion coefficient of typical LC solvents is relatively small, this is not the case for SFC, where it is a strong function of pressure, temperature and mobile phase composition. For neat CO2 at 40 ◦ C, a temperature decrease due to the larger decompression cooling effect prevails below 450 bar or thus in the operating range of mature instrumentation. When going to higher pressure (e.g. the pressure rating of state-of-the-art instruments, i.e. 600 bar), viscous heating becomes more relevant. The pressure at which decompression cooling and viscous heating cancel each other out decreases almost linearly from 450 bar for neat CO2 to 150 bar for 25 vol% methanol. At low back pressure (150 bar), for pressure drops up to 300 bar, decompression cooling is observed throughout the column. However, when increasing the back pressure to 300 bar, the mobile phase heats up in the first part of the column due to the dominance of viscous heating and then cools again due to the decompression cooling at the end of the column. When coupling columns to achieve high efficiencies at very high operating pressure (e.g. 750 bar inlet pressure for 8 vol%), the situation is even more complex. E.g. at a back pressure of 150 bar and using 8 vol% methanol, viscous heating is only observed in the first column whereas only decompression cooling in the second. Whereas the combination of viscous heating and decompression cooling gives rise to complex temperature profiles along a column, increasing the inlet pressure up to 1050 bar resulted in no excessive temperature differences measured along the column (max. 5 ◦ C). This implies that the inlet pressure of SFC instrumentation could be expanded above 600 bar without additional band broadening caused by excessive temperature differences. Acknowledgements R.D.P. gratefully acknowledges a research grant from the Research Foundation – Flanders (FWO Vlaanderen). The authors gratefully acknowledge Xiaoli Wang for the kind donation of the HILIC RRHD columns. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.chroma. 2014.09.022. References [1] I. Halasz, R. Endele, J. Asshauer, Ultimate limits in high-pressure liquid chromatography, J. Chromatogr. 112 (1975) 37. [2] Cs. Horvath, H.-J. Lin, Ultimate limits in high-pressure liquid chromatography, J. Chromatogr. 149 (1978) 43. [3] H. Poppe, J.C. Kraak, J.F.K. Huber, J.H.M. Van den Berg, Temperature gradients in HPLC columns due to viscous heat dissipation, Chromatographia 14 (1981) 515. [4] H. Poppe, J.C. Kraak, J. Chromatogr. 282 (1983) 399. [5] M. Martin, G. Guiochon, Ultimate limits in high-pressure liquid chromatography, J. Chromatogr. A 1090 (2005) 16.
218
R.D. Pauw et al. / J. Chromatogr. A 1365 (2014) 212–218
[6] F. Gritti, G. Guiochon, Measurement of the axial and radial temperature profiles of a chromatographic column: influence of thermal insulation on column efficiency, J. Chromatogr. A 1138 (2007) 141. [7] U.D. Neue, M. Kele, Performance of idealized column structures under high pressure, J. Chromatogr. A 1149 (2007) 236. [8] M.M. Fallas, M.R. Hadley, D.V. McCalley, Practical assessment of frictional heating effects and thermostat design on the performance of conventional (3 m and 5 m) columns in reversed-phase high-performance liquid chromatography, J. Chromatogr. A 1216 (2009) 3961. [9] K. Broeckhoven, J. Billen, M. Verstraeten, K. Choikhet, M. Dittmann, G. Rozing, G. Desmet, Towards a solution for viscous heating in ultra-high pressure liquid chromatography using intermediate cooling, J. Chromatogr. A 1217 (2010) 2022–2031. [10] A. de Villiers, H. Lauer, R. Szucs, S. Goodall, P. Sandra, Influence of frictional heating on temperature gradients in ultra-high-pressure liquid chromatography on 2.1 mm I.D. columns, J. Chromatogr. A 1113 (2009) 84–91. [11] F. Gritti, G. Guiochon, Optimization of the thermal environment of columns packed with very fine particles, J. Chromatogr. A 1216 (2009) 1353–1362. [12] P.A. Mourier, M.H. Caude, R.H. Rosset, The dependence of reduced plate height on reduced velocity in carbon dioxide supercritical fluid chromatography with packed columns, Chromatographia 23 (1987) 21. [13] T.A. Berger, J.F. Deye, Effect of mobile phase density gradients on efficiency in packed column supercritical fluid chromatography, Chromatographia 30 (1990) 57. [14] D.P. Poe, J.J. Schroden, Effects of pressure drop, particle size and thermal conditions on retention and efficiency in supercritical fluid chromatography, J. Chromatogr. A 1216 (2009) 7915–7926. [15] K. Kaczmarski, D.P. Poe, G. Guiochon, Numerical modeling of the elution peak profiles of retained solutes in supercritical fluid chromatography, J. Chromatogr. A 1218 (2011) 6531–6539. [16] 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 (2014), http://dx.doi.org/ 10.1016/j.chroma.2014.08.084.
[17] K. Kaczmarski, D. Poe, G. Guiochon, Numerical modeling of elution peak profiles in supercritical fluid chromatography. Part I – elution of an unretained tracer, J. Chromatogr. A 1217 (2010) 6578–6587. [18] D.P. Poe, D. Veit, M. Ranger, K. Kaczmarski, A. Tarafder, G. Guiochon, Pressure, temperature and density drops along supercritical fluid chromatography columns. I. Experimental results for neat carbon dioxide and columns packed with 3- and 5-micron particles, J. Chromatogr. A 1250 (2012) 105–114. [19] J.H. Knox, I. Grant, Chromatographia 24 (1987) 135. [20] D.C. Ginnings, Precision Measurement and Calibration, Selected National Bureau of Standards Papers on Heat, NBS Special Publication 300, vol. 6, 1970. [21] REFPROP Program, Version 9.1, National Institutes of Sciences and Technology, Physical and Chemical Properties Division, Boulder, CO, USA, 2013. [22] A. Tarafder, J. Hill, P. Iraneta, Effects of Pressure and Temperature on the Retention and Selectivity of Supercritical Fluid Chromatography, presented at the 41st conference on High Pressure Liquid Chromatography and Related Techniques (HPLC2014), May 11–15, New Orleans, Louisiana, USA, 2014. [23] R. De Pauw, B. Degreef, H. Ritchie, S. Eeltink, G. Desmet, K. Broeckhoven, Extending the limits of operating pressure of narrow-bore column liquid chromatography instrumentation, J. Chromatogr. A 1347 (2014) 56–62. [24] 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. [25] F. Lestremau, A. de Villiers, F. Lynen, A. Cooper, R. Szucs, P. Sandra, High efficiency liquid chromatography on conventional columns and instrumentation by using temperature as a variable: kinetic plots and experimental verification, J. Chromatogr. A 1138 (2007) 120–131. [26] A. Vaast, J. De Vos, K. Broeckhoven, M. Verstraeten, S. Eeltink, G. Desmet, Maximizing the peak capacity using coupled columns packed with 2.6 m core-shell particles operated at 1200 bar, J. Chromatogr. A 1256 (2012) 72–79. [27] A. de Villiers, F. Lestremau, R. Szucs, S. Gelebart, F. David, P. Sandra, Evaluation of ultra performance liquid chromatography. Part I. Possibilities and limitations, J. Chromatogr. A 1127 (2006) 60–69.
