Pressure drop characteristics of poly(high internal phase emulsion) monoliths

Journal of Chromatography A, 1144 (2007) 48–54 Pressure drop characteristics of poly(high internal phase emulsion) monoliths I. Junkar a , T. Koloini b , P. Krajnc c , D. Nemec a , A. Podgornik a,∗ , A. Štrancar a a BIA Separations d.o.o., Teslova 30, SI-1000 Ljubljana, Slovenia Faculty of Chemistry and Chemical Technology, University of Ljubljana, Aškerčeva 5, SI-1000 Ljubljana, Slovenia c Faculty of Chemistry and Chemical Engineering, University of Maribor, Smetanova 17, SI-2000 Maribor, Slovenia b Available online 18 January 2007 Abstract Today, monoliths are well-accepted chromatographic stationary phases due to several advantageous properties in comparison with conventional chromatographic supports. A number of different types of monoliths have already been described, among them recently a poly(high internal phase emulsion) (PolyHIPE) type of chromatographic monoliths. Due to their particular structure, we investigated the possibility of implementing different mathematical models to predict pressure drop on PolyHIPE monoliths. It was found that the experimental results of pressure drop on PolyHIPE monoliths can best be described by employing the representative unit cell (RUC) model, which was originally derived for the prediction of pressure drop on catalytic foams. Models intended for the description of particulate beds and silica monoliths were not as accurate. The results of this study indicate that the PolyHIPE structure under given experimental condition is, from a hydrodynamic point of view, to some extent similar to foam structures, though any extrapolation of these results may not provide useful predictions of pressure versus flow relations and further experiments are required. © 2007 Elsevier B.V. All rights reserved. Keywords: Chromatographic monoliths; PolyHIPEs; Pressure drop 1. Introduction Chromatographic separation is preferred to be fast and giving high purity of the end product at low pressure drop on the column [1–3]. Conventional packed beds used in liquid chromatography have limitations due to the indispensable compromise between the column efficiency and the pressure drop. Smaller particle size is used in packed beds to achieve better efficiency, which is associated with high pressure drop on the column [4]. To overcome this problem, monolithic stationary phases which exhibit high performance at relatively low pressure drop and are made of a single piece of highly porous solid, have been introduced [3–5]. To describe the pressure drop on chromatographic columns, different models have been proposed, such as the Kozeny–Carman (KC) or, for higher porosities, the Happel’s correlation [3]. The problem of both correlations is that they ∗ Corresponding author. Tel.: +386 1 426 56 49; fax: +386 1 426 56 50. E-mail address: [email protected] (A. Podgornik). 0021-9673/$ – see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.chroma.2007.01.003 were originally developed for description of particulate beds and their extrapolation to the monolithic beds might be questionable. To overcome this limitation other models were developed for particular monolithic structures. Meyers and Liapis used a pore-network modelling approach wherein a number of socalled flow nodes are interconnected by the cylindrically shaped pores with variable diameters [6–8]. An improvement of this approach was introduced by Bryntesson [9] and enables to assign finite volume to both the bonds and sites and can therefore more appropriately describe various monolithic structures. These models require accurate estimation of connectivity what might be a challenging task. Tallarek et al. introduced equivalent particle dimension for silica monoliths [10–12]. It is calculated by the dimensionless scaling of macroscopic fluid behaviour, i.e. the hydrodynamic permeability and the hydrodynamic dispersion in both types of material: particulate and monolithic. This approach is more suitable for comparison of silica monoliths with the conventional chromatographic supports, rather than predicting the pressure drop a priori from the monolith structure. On the other hand, a model for the prediction of pressure drop I. Junkar et al. / J. Chromatogr. A 1144 (2007) 48–54 on silica monoliths based on structural properties was developed by the computational fluid dynamics simulations using the Navier–Stokes equations [13–15]. In addition to the models applied on chromatographic supports, pressure drop has also been extensively studied for solid catalysts. This goes particularly for the model predicting the pressure drop on catalytic foams characterised by porosity even well above 90% [16]. Despite many different approaches, flow properties of monoliths are still poorly understood and a general model linking the observed pressure drop to the specific pores structure is still lacking [3,13,16–21]. Recently, poly(high internal phase emulsion) (PolyHIPE) chromatographic monoliths, characterised by significantly different structure to any chromatographic monoliths developed so far, have been introduced [22]. PolyHIPE monoliths are prepared by polymerising the continuous phase of a high internal phase emulsion (a concentrated emulsion with a high volume fraction of the discontinuous phase) [23]. The morphology of PolyHIPE monoliths features larger voids (due to the droplets of the discontinuous phase) and interconnecting pores (sometimes referred to as “windows”) due to the shrinkage during the polymerisation [24]. If interconnecting pores would be extremely large, structure would resemble to some extent the structure of silica monoliths. However, when the interconnecting pores are small as in our case, the structure is more closed and PolyHIPE structure is rather unique. Due to potential implementation of PolyHIPE monoliths for chromatographic applications, it would be useful to predict pressure drop from their structural properties. It was our goal to investigate whether any of different simple models for pressure drop prediction described in the literature for beds of different structures can be used to estimate pressure drop on PolyHIPE monoliths. 2.1.2. Happel’s correlation [25] P = 18 × ηvL (1 − ε) dp2  3 + 2(1 − ε)5/3 3 − 9/2(1 − ε)1/3 + 9/2(1 − ε)5/3 − 3(1 − ε)2 2.1.3. Tetrahedral skeleton column (TSC) model [13]   55ηvL 1 − ε 1.55 P = ds2 ε ε (2) (3) The TSC model was developed for silica monoliths assuming tetrahedral skeleton structure and is appropriate for porosities above 70%. The model is a simplification of reality, but it has been considered the best way to gain insight into a complex relation between the internal structure of a monolithic column and its flow resistance [13]. 2.1.4. Representative unit cell (RUC) model [16] The RUC model was originally made for catalytic foams assuming a unit cell shown in Fig. 1. It is devised to provide a simple isotropic configuration whose porosity may be changed without altering the nature of the configuration [16]. It can be used in a similar manner for any porosity value, particularly for high porosities, and is described by Eq. (4).   ηvL 36χ(χ − 1) P = 2 (4) d ε2 2.1. Pressure drop modelling Models described in the literature mainly correlate flow characteristics such as pressure drop, superficial velocity, length of the porous media and their structural characteristics. Generally, structural characteristics are given according to their particle diameter and bed porosity. The models developed for particular structure include explicit (e.g. RUC model) or implicit (constants in other selected models) tortuosity factors. The following models were selected due to their simplicity and frequent implementation in chromatography or catalysis. (1) The KC equation with the constant of 180 is well validated for the bed of packed spheres and is limited to a small range of porosities up to 60%.  The Happel’s correlation can be used for the whole range of porosities from 0 to 100% and was originally developed for spherical particles. 2. Experimental 2.1.1. Kozeny–Carman (KC) correlation 180ηvL(1 − ε)2 P = ε3 ds2 49 Fig. 1. Cubic representative unit cell for RUC model [15]. 50 I. Junkar et al. / J. Chromatogr. A 1144 (2007) 48–54 Tortuosity (χ) derived from the structure in Fig. 1 is expressed as [17]: √   2  3 4π 1 1 9 − 8ε −1 8ε − 36ε + 27 = + cos + cos χ 4ε 2ε 3 3 (9 − 8ε)3/2 (5) The special feature characterised by this model is that both the particle diameter and the pore diameter can be calculated (Eq. (6)). This is particularly useful in our case since pore diameters of PolyHIPE monoliths are easier to determine then particle diameters. The value d used in Eq. (4) represents characteristic length d and is correlated with particle diameter or solid dimension ds , namely in Eq. (6):   2 ε 2 ds = 1 − d2 (6) χ According to the authors [16], the observable pore diameter can be calculated from Eq. (7) as follows: Dp = 3ε d χ (7) All four types of models were applied to evaluate the most appropriate prediction of pressure drop on high porosity PolyHIPE monoliths at given conditions. 2.2. PolyHIPE monoliths PolyHIPE (glycidyl methacrylate-co-ethyleneglycol dimethacrylate) monoliths of different nominal porosities 60, 70, 75 and 80% (scanning electron microscopy (SEM) pictures shown in Fig. 2) were prepared by polymerisation of continuous phases of high internal phase emulsions [22]. They were cut into disks of 3 mm in thickness and 12 mm in diameter, in order to fit them into the Convective Interaction Media (CIM) disk housings produced by BIA Separations (BIA Separations, Ljubljana, Slovenia). 2.3. Instrumentation An HPLC system (Knauer, Berlin, Germany) was built by Pump 64 (Knauer), a variable-wavelength UV–vis detector (Knauer) with a 10 mm optical path set to 280 nm, set response time of 0.1 s, connected by means of 0.25 mm i.d. capillary tubes and HPLC hardware/software (Knauer). A digital pressure Fig. 2. SEM pictures of PolyHIPE monoliths with different nominal porosities at 3500× magnification: (a) nominal porosity 60%, (b) nominal porosity 70%, (c) nominal porosity 75% and (d) nominal porosity 80%. 51 I. Junkar et al. / J. Chromatogr. A 1144 (2007) 48–54 gauge Digibar by HBM (Darmstadt, Germany) was connected to the system. 2.4. Dynamic porosity measurement The dynamic porosity was determined by pulse response experiments. Tracer, 1 mg/ml myoglobin dissolved in 20 mM Tris–HCl buffer, pH 7.4, was injected into the HPLC system. The applied flow rate was 1 ml/min. The retention time was obtained from the tracer peak by calculating the first momentum [26]. The experiment was performed once using the HPLC system with connected CIM housing containing PolyHIPE monolithic disk and once using the HPLC system with connected empty CIM housing. Pore volume was determined by subtracting both retention times. Since the flow rate was 1 ml/min, dynamic porosity was calculated by dividing the retention time difference (monolithic disk total volume being 0.34 ml). The same experiment was also performed using NaNO3 as a tracer, providing the same results for porosity, which demonstrates that no size exclusion was present for myoglobin measurements. Fig. 3. SEM picture showing typical structure of PolyHIPE monolith. Void and interconnecting pores are marked. 2.5. Pressure drop experiments Pressure drop measurements were carried out by placing each PolyHIPE monolith into a CIM housing and pumping deionised water through the column at different flow rates up to 4.5 ml/min. Before the pressure measurements, the pulse response experiments described in Section 2.4 were performed to confirm that the entire mobile phase flows through the PolyHIPE monolithic disk. Pressure was measured at the column inlet. The column was opened with the outlet exposed to the atmospheric pressure, making the measured pressure drop equal to the pressure drop on the CIM column. To calculate the pressure drop on the monolith itself, pressure drop on an empty CIM housing was subtracted. 2.6. Measurements of pore diameter from SEM SEM pictures produced by a scanning electron microscope (JEOL T300, Tokyo, Japan) were analysed to determine the pore diameter of PolyHIPE monoliths. We were able to observe that the monoliths have a complex inner structure with two types of pores. Void pores have an oval shape and are connected through interconnecting pores (marked as Dp ), which are much smaller and have an irregular form (Fig. 3). Fig. 4. SEM picture of PolyHIPE monolith with nominal porosity of 75% under 20,000× magnification. The size of the interconnecting pores was estimated from SEM pictures by counting them and assigning them a size in the form of a circle diameter which could be drawn into the pore. This data served as the basis for calculation of the average pore diameter (Table 1). At least 30 pores were measured for each monolith. At a higher magnification (Fig. 4), it is evident that the void pore wall is actually assembled from small particles. It seems to be a reasonable assumption that this low permeable wall contributes significantly to the pressure drop, prompting us to estimate the size of the particles forming this wall. Size was estimated assuming a spherical particle form (Table 1). Table 1 Measured structural properties of PolyHIPE monoliths PolyHIPE porosity SEM picture and mercury porosity measurements Nominal porosity (%) Dynamic porosity (%) Static porosity (%) SEM ds (nm) SEM Dp (␮m) Mercury porosimetry Dp (␮m) 60 70 75 80 57.7 69 72.5 81 57.2 66.1 73 82.4 184 144 105 132 0.454 0.521 0.395 0.461 0.684 0.583 0.347 0.63 NP is nominal porosity and determines percentage of aqueous phase in emulsion. Dynamic porosity was calculated from a pulse response experiments while static porosity was measured by mercury porosimetry. Pore and particle diameter were determined from scanning electron microscopy (SEM) pictures and pore diameter was in addition measured with mercury porosimetry. 52 I. Junkar et al. / J. Chromatogr. A 1144 (2007) 48–54 2.7. Measurements of static porosity and pore size with mercury porosimetry Static porosity and pore size in the range of 14–20,000 nm was determined using mercury porosimetry (PASCAL 440, ThermoQuest Italia, Rodano, Italy). A piece of the monolith was cut to an approximate weight of 0.1 g and completely dried prior to measurement. In the case of PolyHIPE structure, interpretation of the pore size data requires, understanding of the measurement principle [22]. The pore size peak is obtained at the pore size where most of the mercury is able to penetrate into the sample. Since void pores are surrounded by a wall permeable through interconnecting pores, mercury is able to penetrate into the void pore only when the pressure is high enough to allow mercury penetration through the interconnecting pores. The pore size measurements therefore give the information about the size of the interconnecting pores but the differential volume which should be assigned to void pores. 3. Results and discussion Structure of the PolyHIPE monoliths resembles to an inverse structure of the conventional packed bed with void pores instead of particles (Fig. 3). Void pores are linked through the interconnecting pores, which might be seen as an analogy of contacts between the particles. As a result of this peculiar structure, it is not obvious what kind of correlation would properly describe the pressure drop. Experimental values of the pressure drop data from PolyHIPE monoliths are shown in Fig. 5. This figure features an interesting observation, namely that pressure drop does not correlate with the porosity. In particular, the monolith featuring 70% porosity exhibits comparable pressure drop to the monolith with 80% porosity, while pressure drop featured by the monolith with 75% porosity is significantly higher. Such behaviour has to be attributed to the size of interconnecting pores, which are the largest for monoliths with 70% porosity (Table 1). On the other hand, monoliths with 80% porosity have somewhat smaller interconnecting pores but higher porosity, which seems to compensate for the pressure drop. Data for all monoliths show a linear relation between the pressure drop and the flow rate (Fig. 5), which indicates that the bed is not compressible and that a laminar flow regime is present. Mathematical correlations Fig. 5. The dependence of flow rate on pressure drop for PolyHIPE monolith with different nominal porosities. described in Section 2.1 may thus be applied. Therefore, it is possible to estimate predicted particle diameter from the measured pressure drop data using Eqs. (1)–(4), including the pore diameter from the RUC model. To apply these models, both pressure drop data and the porosity should be measured. Rough estimation of the porosity can be made using a percentage of aqueous phase in the polymerisation mixture. It was proven for PolyHIPE methacrylate monoliths that this value corresponds closely to the static porosity [22] what was also reconfirmed by mercury porosimetry data (Table 1). However, for a proper estimation of pressure drop, dynamic porosity is required and this was subsequently measured using a tracer (Table 1). We can see that values of dynamic porosity are in close agreement to the values of static porosity, indicating that all pores are accessible to the mobile phase during operating conditions (small differences should probably be accounted due to experimental error). Dynamic porosity and pressure drop data were used from an estimation of the structural properties obtained from Eqs. (1) to (4). Since none of the applied mathematical models were developed for the PolyHIPE structure, values predicted from the models were compared to the measured structural data obtained from SEM pictures and mercury porosimetry (Table 1) in order to find which model most accurately describes the pressure drop on the PolyHIPE structure. Results in Table 2 summarise particle diameter estimated by KC, Happel’s, TSC and RUC models by using experimental data. As expected, particle diameters estimated with KC model, which is appropriate only for porosities up to 60%, differ significantly from the values estimated from SEM pictures. On the other hand, the Happel’s model, which could be applied to a much higher porosity, also seems an inappropriate model, as it still anticipates a very high particle diameter value. Therefore, neither the KC nor the Happel’s correlation, developed for a particulate bed, properly describe the pressure drop on PolyHIPEs. On the contrary, TSC or RUC models, both developed for monolithic structures featuring high porosity more similar to PolyHIPE, show a better correlation. Indeed, the prediction of particle size from these models is much closer to the values estimated from SEM. Averaged particle diameter calculated from TSC model shows around 20% deviation from SEM values, while particle diameter calculated from RUC model shows better agreement (Table 2). Deviation of particle diameter between SEM values and RUC model is below 10% for porosities up to 75%, though a much higher difference was observed for 80% porosity. One possible reason might be that following examination of several additional monolith samples with 80% porosity, few much larger void pores than average were observed (data not shown). Because of their low number, however, they were not considered in the pore diameter estimation, which may be a reason for a lower pressure drop than anticipated from the particle size based on SEM estimates. Next to particle diameter, the RUC model also enables the prediction of pore diameter implementing Eq. (6). Calculated values are in good agreement with the values measured from SEM and mercury porosimetry (Table 2). We can conclude that pore size estimated from the RUC model is appropriate for the descrip- 53 I. Junkar et al. / J. Chromatogr. A 1144 (2007) 48–54 Table 2 Particle and pore diameter calculated by various models from experimental pressure drop data measured on PolyHIPE monoliths having different porosities (see Fig. 5) PolyHIPE Calculated averaged particle diameter ds from different models and deviation from measured ds values from SEM Nominal porosity (%) KC model (nm) Happel model (nm) TSC model (nm) Percentage difference from SEM (%) RUC model (nm) Percentage difference from SEM (%) RUC model Dp (␮m) Percentage difference from SEM [%] 60 70 75 80 377.3 311.4 190.8 148.4 377.4 332.5 209.9 181.1 223.7 180.7 131.2 113.7 21.6 25.5 24.9 13.9 179.8 158.8 99.5 82.6 2.3 10.3 5.2 37.4 0.38 0.464 0.325 0.368 16.3 10.9 17.7 20.1 tion of PolyHIPE structure, particularly taking into account the significant differences between pore size estimated from SEM and mercury porosimetry. Results of this study show that the flow pattern through PolyHIPE monoliths might be to some extend similar to the flow pattern through catalytic foams. Significant discrepancy between predicted pressure drop and measured one for other models is expected since KC and Happel’s models were derived for beds of spherical particles while TSC model in its present form is specific for the tortuosity and surface to volume ratio of the tetrahedral skeleton. Though the structure of PolyHIPE monoliths significantly differs also from structure of catalytic foams it seems that RUC model rather accurately predicts pressure versus flow rate for PolyHIPE monoliths, at least under tested experimental conditions. However, it should be noted that even RUC model might give misleading results for PolyHIPE monoliths under different experimental conditions. To verify this new PolyHIPE monoliths with different structure (in terms of porosity, pore size distribution, nanoparticle diameter) should be prepared and tested. Unfortunately, preparation of PolyHIPE monoliths with precisely defined structural properties is at this stage still rather challenging task and further improvement in reproducibility is needed. 4. Conclusions PolyHIPE monoliths exhibit a very complex structure, which differs significantly from conventional packed beds. As they are a rather new type of monolith with regard to chromatography applications, no model correlating pressure drop with structural properties has yet been developed. We demonstrated that the application of model developed for correlating pressure drop on catalytic foams predicts reasonably well pressure drop for tested PolyHIPE monoliths if their structural properties are known. RUC model was developed for structure significantly different from that of PolyHIPE monoliths, therefore, one should be aware of the limitation of this model as it is not necessary that it would give good prediction for pressure drop on PolyHIPE monoliths at different conditions. Therefore, any extrapolation of these results might be misleading. To demonstrate general implementation, further experiments on PolyHIPE monoliths with different structure are required. Based on this premise, further optimisation of PolyHIPE monoliths would be possible. Calculated averaged pore diameter Dp and deviation from SEM 5. Nomenclature d ds dp Dp Fv L v microscopic characteristic length in RUC model (m) solid diameter (m) pore section hydraulic diameter (m) interconnecting pore diameter (m) volumetric flow rate (m3 /s) bed length (m) superficial velocity (m/s) Greek symbols P pressure drop (Pa) χ pore structure tortuosity ε effective porosity η viscosity (Pa s) References [1] F. Švec, T.B. Tehnikova, Z. Deyl (Eds.), Monolithic Materials: Preparation, Properties and Applications, Elsevier, Amsterdam, 2003, p. 137. [2] S. Ghose, S.M. Cramer, J. Chromatogr. A 928 (2001) 13. [3] I. Mihelič, D. Nemec, A. Podgornik, T. Koloini, J. Chromatogr. A 1065 (2005) 59. [4] H. Minakuchi, K. Nakanishi, N. Soga, N. Ishizuka, N. Tanaka, J. Chromatogr. A 762 (1997) 135. [5] G. Guiochon, Cs. Horvath (Eds.), High Performance Liquid Chromatography—Advances and Perspectives, vol. 2, Academic Press, New York, 1980, p. 1. [6] J.J. Meyers, A. Liapis, J. Chromatogr. A 827 (1998) 197. [7] J.J. Meyers, A. Liapis, J. Chromatogr. A 852 (1999) 3. [8] A. Liapis, J.J. Meyers, O.K. Crosser, J. Chromatogr. A 865 (1999) 13. [9] L.M. Bryntesson, J. Chromatogr. A 945 (2002) 103. [10] F.C. Leinweber, D. Lubda, K. Cabrera, U. Tallarek, Anal. Chem. 74 (2002) 2470. [11] U. Tallarek, F.C. Leinweber, A. Seidel-Morgenstern, Chem. Eng. Technol. 25 (2002) 1177. [12] F.C. Leinweber, U. Tallarek, J. Chromatogr. A 1006 (2003) 207. [13] N. Vervoort, P. Gzil, G. Baron, G. Desmet, Anal. Chem. 75 (2003) 843. [14] N. Vervoort, P. Gzil, G. Baron, G. Desmet, J. Chromatogr. A 1030 (2004) 177. [15] A. Bakker, A.H. Haidari, E.M. Marshall, Chem. Eng. Prog. (December) (2001) 30. [16] P.D. Plessis, A. Montillet, J. Comiti, J. Legrand, Chem. Eng. Sci. 49 (1994) 3545. [17] P.D. Plessis, Heat and Mass Transfer in Porous Media, Elsevier, Amsterdam, 1992, p. 249. [18] J.G. Fourie, J.P. Du Plessis, Chem. Eng. Sci. 57 (2002) 2781. 54 I. Junkar et al. / J. Chromatogr. A 1144 (2007) 48–54 [19] N. Tanaka, H. Kobayashi, N. Ishizuka, H. Minakuchi, K. Nakanishi, K. Hosoya, T. Ikegami, J. Chromatogr. A 965 (2002) 35. [20] I.F. Macdonald, M.S. El-Sayed, K. Mow, F.A.L. Dullien, Ind. Eng. Chem. Fundam. 18 (1979) 199. [21] R. Hahn, A. Jungbauer, Anal. Chem. 72 (2000) 4853. [22] P. Krajnc, N. Leber, D. Štefanec, S. Kontrec, A. Podgornik, J. Chromatogr. A 1065 (2005) 69. [23] D. Barby, Z. Haq, US Patent 4,522,953 (1985). [24] N.R. Cameron, D.C. Sherrington, Adv. Polym. Sci. 126 (1996) 163. [25] J. Happel, AIChE J. 4 (1985) 197. [26] R.H. Perry, D.W. Green, Perry’s Chemical Engineering Handbook, seventh ed., McGraw-Hill, New York, 1997.