You are on page 1of 5

Full Paper

Fluid Dynamics in Monolithic Adsorbents: Phenomenological


Approach to Equivalent Particle Dimensions
By Ulrich Tallarek, Felix C. Leinweber, and Andreas Seidel-Morgenstern*

Due to the complex, often sponge-like structure of monolithic adsorbents it is difficult to define appropriate constituent units
that characterize the hydrodynamics of the material, or to determine relevant shape and size distribution factors comparable to
those for spherical particles in (particulate) fixed beds. Based on a phenomenological analysis of the friction factor (Reynolds
number relation and the longitudinal dispersivity ± Peclet number dependence for random sphere packings) we derive
characteristic lengths (i.e., equivalent particle dimensions) for a monolith with regard to its hydraulic permeability and
dispersion originating in stagnant zones. Equivalence to the hydrodynamic behavior in ªreferenceº sphere packings is
established by dimensionless scaling of the respective data for the monolithic structure. This phenomenological approach, which
is simply based on liquid flow and stagnation in a porous medium, can successfully relate hydrodynamic properties of the
monolith to that of particulate beds.

1 Introduction major principles underlying the advanced processing of a high


surface area into appropriate structures, (i.e. suitable perme-
The recent development of rigid monolithic support ability, heterogeneity and associated mass transfer resis-
structures has contributed to many technological and envi- tances) remain identical, for example the physical mecha-
ronmental processes, including selective high surface area nisms responsible for dispersion. These involve molecular
materials (e.g., microcellular foams, nanostructured films, diffusion in the interconnected pore network, a mechanical
hierarchically-ordered composites, molded polymers) for contribution due to velocity fluctuations in the flow field
emission control of industrial and automobile gases, high induced by the randomly distributed solid phase, nonmecha-
throughput catalysis and (bio)medical screening, chromato- nical dispersion associated with the no-slip condition at the
graphic separations, or as an important component in super- solid-liquid interface (in general), and the particular access to
capacitors and sensors [1±10]. By controlling chemical intraparticle or intraskeleton stagnant zones in which diffu-
composition, pore and surface structure, as well as physical sion remains the dominating transport mechanism [13]. While
properties down to the nanometer scale it is possible to equivalence between different geometries concerning the
produce materials with tailored optical, mechanical, electrical convection, diffusion and reaction is well recognized for
or hydrodynamic properties and increased selectivity for particulate systems [14], a direct comparison with monolithic
adsorption/reaction and catalytic processes. The introduction structures is less obvious due, not least, to the complex shape
of hierarchical order in the interconnected pore network and of the skeleton domain. Although monoliths are considered as
independent manipulation of the contributing sets of pores a new generation of adsorbents, appropriate characteristic
[11] allows the design of highly permeable media and lengths that competitively relate their hydrodynamic proper-
optimization of solute transport to and from the active surface ties (i.e. flow and stagnation of the mobile phase) to those of
sites. Compared to conventional particulate fixed-bed adsor- particulate fixed beds are not readily available (Fig. 1).
bers monolithic materials combine high selectivity and
capacity, mobile phase velocity and mass transfer efficiency
in a unique manner [4,6,12].
3 Materials and Methods
Particulate beds were prepared with either nonporous or
completely porous spherical C18-silica particles, the latter
2 Problem Formulation having mesopores of 30 nm average diameter. These packings
were compared to a C18-silica monolith with bimodal pore
While the solid phase in random sphere packings is
size distribution (Fig. 2) purchased from Merck KGaA
discontinuous, monoliths have a continuous rigid skeleton.
(Darmstadt, Germany). Relevant column, particle and pack-
Although differences between monolithic and particulate
ing characteristics are summarized in Tab. 1. The mobile phase
stationary phases are obvious with regard to morphology, the
in all experiments was a water/acetonitrile (50:50 v/v) mixture
containing 0.1 % trifluoroacetic acid. Thus, following tracer
±
pulse injection, the elution of analytes (acetone, angiotensin II
[*] Dr. rer. nat. U. Tallarek (author to whom correspondence should be
addressed), Dipl.-Chem. F. C. Leinweber, Prof. Dr.-Ing. A. Seidel- and insulin) occured under nonretained conditions. The
Morgenstern, Lehrstuhl für Chemische Verfahrenstechnik, Institut longitudinal dispersion coefficients were determined by
für Verfahrenstechnik der Otto-von-Guericke-Universität Magdeburg,
Universitätsplatz 2, D-39106 Magdeburg, Germany (E-mail: ulrich.
analysis of residence time distributions with an exponentially
tallarek@vst.uni-magdeburg.de). modified Gaussian function [15].

Chem. Eng. Technol. 25 (2002) 12, Ó 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 0930-7516/02/1212-1177
0930-7516/02/1212-1177 $ 17.50+.50/0
$ 17.50+.50/0 1177
Full Paper

RANDOM SPHERE PACKING MONOLITHIC STRUCTURE 4 Results and Discussion


a)
NONPOROUS
d) In this communication we use a simple, general, phenom-
PARTICLE NONPOROUS
SKELETON enological approach for analyzing single-phase incompressi-
ble flow through porous media with a continuous solid phase.
It relies on equivalent particle (sphere) dimensions which
b) Lstag characterize the corresponding behaviour in a particulate bed.
POROUS
LAYER
e) Equivalence is obtained by the dimensionless scaling of
POROUS
SKELETON
macroscopic flow behaviour (pore space permeability and
?
longitudinal dispersion) for both types of media, without the
equiv. Lstag need for a direct geometrical translation of their constituent
c)
Lstag units. In this approach, the Reynolds number Re = usfLflow/m
COMPLETELY
POROUS (defined with respect to the superficial velocity, usf) is used to
? analyse the resistance of the porous medium to liquid flow, and
the Peclet number Pe = uavLstag/Dm (defined with respect to
Lflow equiv. Lflow the average velocity of an analyte through the bed, uav) is used
Figure 1. Characteristic lengths for hydraulic permeability and hydrodynamic
in the analysis of longitudinal dispersion. Dm is the diffusivity
dispersion in a bed of spheres and monoliths. While the particle diameter is of the analyte in the mobile phase and m its (kinematic)
useful to define a region impermeable for flow, the thickness of its porous layer viscosity. Most important in the present context, Lflow and
may be used to address dispersion originating in stagnant zones of a sphere
packing. Apparently, Lflow and Lstag are different. Lstag are appropriate lengths of the medium characterizing
flow permeability and dispersion originating in stagnant
zones, respectively (cf. Fig. 1)1).

4.1 Hydraulic Permeability

Based on a general phenomenological analysis, the resis-


tance of (incompressible) random close packings of spherical
particles to liquid flow can be expressed by the following
relationship between dimensionless variables (with, as usual,
c) the mean sphere diameter dp for Lflow) [16]
 
DP dp usf dp
ˆ F ˆ f Re ˆ ; e inter ; qi ; w (1)
ru2sf L m i

where DP/L is pressure drop over length L of the bed (in the
macroscopic flow direction), r is the density of the liquid, and
qi and wi are parameters for particle shape and size
distribution. Eq. (1) may be much simplified when the
deviation from a uniform spherical shape and differences in
particle size distribution, surface roughness, interparticle
porosity (einter) and bed structure for different sphere
packings may be considered insignificant. Eq. (1) then reduces
to a simple unique relation between friction factor F and Re
[16,17].
Figure 2. Morphology of the silica-based monolith. a) SEM picture showing the
macroporous interskeleton pore space with dmacro = 1.9 lm, and b) the ±
mesoporous skeleton, dmeso = 25 nm. c) Discrete bimodal pore size distribution. 1) List of symbols at the end of the paper.

Table 1. Characteristic dimensions and parameters of the porous media. dcol is the column diameter and L denotes the bed length. etotal, einter and eintra are the total,
interparticle (interskeleton) and intraparticle (intraskeleton) porosities of the porous medium, respectively [12]. Aspec is the specific and Arel the volume-weighted
surface area.

Fixed bed L [mm] dcol [mm] dp [lm] etotal einter eintra Aspec [m2 g±1] Arel [m2 cm±3] Lflow [lm] Lstag [lm]
Spheres (nonporous) 53 4.6 3.0 0.36 0.36 ± 0.6 0.7 3.0 ±
Spheres (porous) 55 2.0 6.4 0.75 0.37 0.60 76 65 6.4 3.2
Monolithic structure 100 4.6 ± 0.92 0.72 0.70 147 39 9.5 1.0

1178 Ó 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 0930-7516/02/1212-1178 $ 17.50+.50/0 Chem. Eng. Technol. 25 (2002) 12
Full Paper

Fig. 3 demonstrates, as then expected, a superposition of which, in contrast to Eq. (1), includes features of both the
these data for different sphere packings having similar einter, qi convection-dominated interparticle and diffusion-limited
and wi. The linear dependence of F on Re represents the intraparticle pore space. Eq. (2) can also be simplified for
validity of Darcy's law. An equivalent permeability length packed beds with comparable einter, qi, wi and structure,
(Lflow) for the monolithic structure, i.e., a particle diameter dp resulting in a relationship between the dispersivity D and Pe
for a fictitious bed (with similar einter, qi, wi, and structure as almost as simple as that between F and Re. The remaining
the other two sphere packings) is obtained by dimensionless difficulty in Eq. (2) is associated with an evaluation of the
scaling of the pressure drop ± flow rate dependence for analyte diffusivity in stagnant zones (Dstag), as in general it
monoliths to the F ± Re relation characterizing particulate depends on the morphology of the intraparticle pore space.
beds (Fig. 3). It is achieved by adjusting dp which gives Lflow = However, when Dstag is similar the dispersion data for
9.5 lm in the present case. The fact that the data for porous different particulate beds and monoliths are expected to
and nonporous sphere packings coincide by means of their collapse on a unique curve by means of the characteristic
mean dp shows that the intraparticle pore network of these dimension Lstag. This is demonstrated by the data in Tab. 2 and
porous particles does not contribute measurably to macro- Fig. 4 which show Dstag is very similar for the porous particles
scopic flow. The actual macro-to-mesopore diameter ratio in and monolith used in this work. For completely porous
the sphere packing (and monolith) is of the order of 100, thus, spheres Lstag = rp is easily measured (cf. Fig. 1), and has been
hydraulic permeability of the intraparticle (and intraskeleton) used to calculate Pe in Fig. 4.
mesopore space is some 104 times lower than that of the As with the permeability data (Fig. 3), the next step consists
interparticle (and interskeleton) macroporous network, and of dimensionless scaling, this time of the monoliths dispersion
the assumption that the former constitutes a stagnant zone data to those obtained with the fixed bed of porous spheres,
seems to be justified. giving Lstag = 1.0 lm (Fig. 4). The equivalence in this case
states that dispersion in the monolith associated with liquid
holdup in the intraskeleton pore space resembles the
contribution from stagnant zones (characterized by the same
Dstag) in a bed of spherical particles with porous layer of
thickness Lstag. It is important to note that, if liquid holdup
exists in a monolith and dominates dispersion, nonporous
spheres are not appropriate for evaluating an equivalence
based on this particular (nonmechanical) contribution.

Figure 3. Friction factor vs. Reynolds number for the liquid in beds of
(non)porous spheres, with dp as Lflow in Re (Fig. 1). Dimensionless scaling for
the monolith gives Lflow = 9.5 lm. Re = 0.1 corresponds to a pressure drop DP
over the monolith (L = 100 mm) of about 15 MPa.

4.2 Hydrodynamic Dispersion


Figure 4. Longitudinal dispersivity vs. Peclet number for nonretained analytes in
a bed of porous spheres, with rp as Lstag in Pe (Fig. 1). Dimensionless scaling for
The spatial dimension of stagnant zones is often used as the monolith gives Lstag = 1.0 lm.
characteristic length for hydrodynamic dispersion in macro-
scopically homogeneous, microscopically disordered media
like random packings of porous spheres because liquid
holdup, i.e., diffusion-limited mass transfer in these regions, 5 Conclusions
dominates dispersion for Pe >> 1 [13]. Similar to Eq. (1) we
may write for the longitudinal dispersion coefficient DL It is possible to phenomenologically translate the macro-
  scopic hydraulic permeability and longitudinal dispersion
DL uav L stag Dstag characteristics (dominated by the holdup contribution) of a
ˆD ˆ f Pe ˆ ; einter ; qi ; wi ; (2)
Dm Dm Dm monolith with bimodal pore size distribution and hierarchi-

Chem. Eng. Technol. 25 (2002) 12, Ó 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 0930-7516/02/1212-1179 $ 17.50+.50/0 1179
Full Paper

Table 2. Physical properties of the analytes. RG is the radius of gyration and Dstag denotes the effective diffusion coefficient of analyte molecules in the (intraparticle
or intraskeleton) stagnant zone. Dstag = eintraKpDm/sintra where sintra = eintra + 1.5 (1 ± eintra) is the tortuosity factor [18] and Kp characterizes hindered pore diffusion
depending on the molecule size [19].

Dstag [10±10 m2 s±1]


Analyte Mw [g mol±1] RG [nm] Dm [10±10 m2 s±1]
Porous particle Monolith skeleton
Acetone 58.1 n.a. 12.8 6.6 8.1
Angiotensin II 1046 < 0.2 3.1 1.5 1.9
Insulin 5807 0.54 1.4 0.7 0.8

cally-structured pore network into distinct particle dimen- Dstag [m2 s±1] effective diffusion coefficient in
sions of what is most conventionally understood by a stagnant zone, Dstag = eintraKpDm/sintra
ªrandom-closeº sphere packing. The actual ratio of equivalent usf [m s±1] superficial mobile phase velocity,
lengths, R hp = Lflow/Lstag, may be used as a parameter usf = Fv/Acol
characterizing the hydrodynamic performance (hp) of a uav [m s±1] average analyte velocity through the
monolith. While this material provides relatively large flow- bed, uav = usf/etotal
through pores (Lflow = 9.5 lm), it only needs a comparatively Fv [ml min±1] volumetric flow rate
short diffusion length (Lstag = 1.0 lm) to combine the high Re [±] Reynolds number
permeability with a large surface area. With regard to R hp, the Pe [±] Peclet number
monolithic column resembles a fixed bed of (solid core-porous dp [m] mean diameter of spherical particles
shell) spheres with dp = Lflow and a porous layer of thickness Kp [±] hindrance factor concerning pore-level
Lstag (cf. Fig. 1). Thus, the volume-weighted surface area for a diffusion
monolith can much larger than for a bed of core-shell spheres F [±] friction factor
with the same R hp, and may be the origin of the superior L stag [m] characteristic length for liquid holdup
performance displayed by the monolith in many applications. in porous medium
Vice versa, completely porous particles provide a higher Lflow [m] characteristic length for hydraulic
surface area than the core-shell particles, which is still of the permeability
same order as that of the monolith, but in this case R hp = 2 qi [±] parameter(s) of particle size
(while R hp » 10 for the monolith). Thus, due to the high distribution
porosity (etotal > 0.9) and its unique, rigid structure that can R hp [±] hydrodynamic performance factor,
withstand high pressure, this type of monolith offers sub- R hp = Lflow/Lstag
stantial adsorption capacity (see Aspec and Arel in Table 1) for Vcol [m3] column volume
fast and efficient liquid phase separations. Vinter [m3] interparticle (interskeleton) pore
volume
Vintra [m3] intraparticle (intraskeleton) pore
volume
Acknowledgements Vsolid [m3] volume of particle (skeleton) solid

We gratefully acknowledge support of this work by the


Deutsche Forschungsgemeinschaft (DFG) under grant SE Greek symbols
586/7±1.
Received: June 12, 2002 [CET 1628] einter [±] interparticle (interskeleton) porosity,
einter = Vinter/Vcol
eintra [±] intraparticle (intraskeleton) porosity,
eintra = Vintra/(Vintra + Vsolid)
etotal [±] total porosity
Symbols used
wi [±] particle shape parameter
m [m2 s±1] kinematic viscosity of mobile phase
Acol [m2] column cross-sectional area
sintra [±] intraparticle (intraskeleton) tortuosity
Arel [m2 m±3] relative surface area of particulate bed
factor, sintra = eintra + 1.5 (1 ± eintra)
(monolith)
r [kg m±3] density of mobile phase
Aspec [m2 kg±1] specific surface area
dcol [m] column diameter
DL [m2 s±1] longitudinal dispersion oefficient
Dm [m2 s±1] analytes free molecular diffusion References
coefficient in mobile phase [1] R. M. Heck, R. J. Farrauto, Catalytic Air Pollution Control: Commercial
D [±] reduced dispersivity, D = DL/Dm Technology, John Wiley & Sons, New York 1995.

1180 Ó 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 0930-7516/02/1212-1180 $ 17.50+.50/0 Chem. Eng. Technol. 25 (2002) 12
Full Paper

[2] R. A. Dunbar, J. D. Jordan, F. V. Bright, Anal. Chem. 1996, 68, 604. [11] K. Nakanishi, R. Takahashi, T. Nagakane, K. Kitayama, N. Koheiya,
[3] F. Kapteijn, J. J. Heiszwolf, T. A. Nijhuis, J. A. Moulijn, CatTech. 1999, H. Shikata, N. Soga, J. Sol-Gel Sci. Technol. 2000, 17, 191.
3, 24. [12] F. C. Leinweber, D. Lubda, K. Cabrera, U. Tallarek, Anal. Chem. 2002,
[4] F. Svec, J. M. J. FrØchet, Ind. Eng. Chem. Res. 1999, 38, 34. 74, 2470.
[5] D. Josic, A. Buchacher, J. Biochem. Biophys. Methods 2001, 49, 153. [13] D. Kandhai, D. Hlushkou, A. G. Hoekstra, P. M. A. Sloot, H. Van As,
[6] N. Tanaka, H. Kobayashi, K. Nakanishi, H. Minakuchi, N. Ishizuka, U. Tallarek, Phys. Rev. Lett. 2002, 88, article no. 234501.
Anal. Chem. 2001, 73, 420A. [14] Z. P. Lu, M. M. Dias, J. C. P. Lopes, G. Carta, A. E. Rodrigues, Ind.
[7] A. Kirschning, C. Altwicker, G. Dräger, J. Harders, N. Hoffmann, Eng. Chem. Res. 1993, 32, 1839.
U. Hoffmann, H. Schönfeld, W. Solodenko, U. Kunz, Angew. Chem. Int. [15] V. B. Di Marco, G. G. Bombi, J. Chromatogr. A 2001, 931, 1.
Ed. 2001, 40, 3995. [16] H. Rumpf, A. R. Gupte, Chem. Ing. Tech. 1971, 43, 367.
[8] J. L. Williams, Catal. Today 2001, 69, 3. [17] I. F. Macdonald, M. S. El-Sayed, K. Mow, F. A. L. Dullien, Ind. Eng.
[9] P. Sepulveda, J. R. Jones, L. L. Hench, J. Biomed. Mater. Res. 2002, 59, Chem. Fundam. 1979, 18, 199.
340. [18] M. Suzuki, J. M. Smith, Chem. Eng. J. 1972, 3, 256.
[10] H. Pröbstle, C. Schmitt, J. Fricke, J. Power Sources 2002, 105, 189. [19] H. Brenner, L. J. Gaydos, J. Colloid Interface Sci. 1977, 58, 312.

_______________________

Chem. Eng. Technol. 25 (2002) 12, Ó 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 0930-7516/02/1212-1181 $ 17.50+.50/0 1181

You might also like