Journal of Chromatography A, 1516 (2017) 89–101

Contents lists available at ScienceDirect

Journal of Chromatography A
journal homepage: www.elsevier.com/locate/chroma

Full length article

Influence of different packing methods on the hydrodynamic stability

of chromatography columns
M. Dorn, F. Eschbach, D. Hekmat ∗ , D. Weuster-Botz
Institute of Biochemical Engineering, Technical University of Munich, Boltzmannstr. 15, 85748 Garching, Germany

a r t i c l e i n f o a b s t r a c t

Article history: It is well known that packing non-uniformity may cause peak asymmetry and limit the performance of
Received 16 March 2017 packed-bed chromatographic columns. However, understanding of the reasons leading to packing non-
Received in revised form 11 July 2017 uniformity is still limited. Therefore, the effect of different column packing methods, i.e. dynamic axial
Accepted 6 August 2017
compression (DAC), flow packing, and combinations of both on the hydrodynamic packing heterogeneity
Available online 8 August 2017
and stability of packings composed of polymer-based compressible porous resins with a mean diameter
of 90 ␮m was investigated experimentally as well as in-silico. Deterministic Euler-Lagrange modeling
Keywords:
of a small chromatographic column with a diameter of 9.6 mm and a bed height of 30 mm was applied
Axial hydrodynamic dispersion
CFD-DEM modeling
by coupling Computational Fluid Dynamics (CFD) and the Discrete Element Method (DEM). Interparti-
Column packing method cle micromechanics as well as the fluid-particle and particle-wall interactions were taken into account.
Hydrodynamic stability Experiments and simulations revealed substantial non-uniformity of compression force transmission
Integral porosity deviation and axial packing density distribution during both dynamic axial compression and flow packing which
Packing heterogeneity was related to wall support and interparticle friction. By combining both packing methods sequentially
(dynamic axial compression followed by flow packing or vice versa), the compression forces were more
homogeneous resulting in improved packing procedures. Repeated alternating application of flow pack-
ing and DAC (the so-called hybrid packing method) resulted in the most homogeneous packing density
distribution and the highest packing stability which was kept nearly constant during long-term oper-
ation with cyclic hydrodynamic load. The hydrodynamic stability of the chromatographic column was
evaluated by calculating the integral porosity deviation and packing induced flow velocity dispersion.
The hybrid packing method gave the best results for both parameters.
© 2017 Elsevier B.V. All rights reserved.

1. Introduction of the packing after some period of operation as a direct result

of progressive settling of the packing particles and a reduction
During the last decades, considerable effort has been devoted of the packing porosity [29–31]. It is well known that packing
to the improvement of chromatographic performance [1–4], the homogeneity is required for a uniform sample concentration and
fundamental thermodynamics of the process [e.g. 5–7], as well as flow velocity distribution. Knox and co-workers [32,33] showed
methods maximizing the production rate or minimizing the costs that packings of dry packed glass beads are radially inhomoge-
[e.g. 8–12]. Further effort has been devoted to the understanding neous being more permeable at the column wall region. Later, Eon
and prediction of column hydrodynamics, i.e. the correlation of [34] found out that radial compression of the packing markedly
pressure drop and flow velocity [13–24] as well as the influence improved column performance. In contrast to this, Baur et al. [35]
of different buffer solutions on the resin swelling characteristics demonstrated that the column wall region in wide bore columns
[25–28]. However, a fundamental analysis of the hydrodynamic is less permeable than the column core region. Similar observa-
column behavior on the particle scale remains challenging. Fur- tions were made by Farkas et al. [29] using porous silica packings.
thermore, the accessibility of relevant parameters which influence Guiochon [36] pointed out later that axial and radial inhomogene-
column hydrodynamics and determine the degree of packing ity depends on the packing method applied. It was shown that the
homogeneity is still limited. Voids are known to appear at the top local fluid velocity and local HETP (Height Equivalent to Theoreti-
cal Plate) values vary across the column and over time for axial and
radial as well as for slurry compressed columns [37]. Hence, it is
established knowledge that the chromatographic packing exhibits
∗ Corresponding author.
intrinsic consolidation dynamics during column operation which is
E-mail address: hekmat@lrz.tum.de (D. Hekmat).

http://dx.doi.org/10.1016/j.chroma.2017.08.019
0021-9673/© 2017 Elsevier B.V. All rights reserved.
90 M. Dorn et al. / J. Chromatogr. A 1516 (2017) 89–101

governed by particle rearrangement and migration as well as by the 2. Definition of chromatographic column parameters
influence of the column wall. In order to maintain long-term sta-
bility of the chromatographic packing, dynamic compression needs In the following, the most relevant parameters are defined that
to be applied forcing the packing material in a position that resists were used to characterize the chromatographic packing.
further movement.
Ideally, the packing would be hexagonally close packed [38], all 2.1. Packing density, porosity, and permeability
particles being in contact with their maximum number of neigh-
bor particles. Then, all flow channels from the column inlet to the The total porosity ␧t of a gravity settled packing consisting of
outlet would have the same volume and length [39]. However, porous spherical particles is defined as the ratio of the volume of
such a packing configuration is impossible to obtain due to sev- void space Vv to the bulk volume of the packing V which is the sum
eral reasons. First, the packing particles are not monodisperse and of the solid volume vs and the void volume Vv
by having a size distribution, a hexagonally close packing config- Vv Vv
uration is impossible [39]. Second, the impenetrable column wall εt = = (1)
V Vs + Vv
prohibits a close packing configuration at the wall and establishes a
degree of packing heterogeneity near the column wall region which As the void space is the sum of the interparticle or external void
is a geometrical wall effect [40,41]. Third, the wall and the particles volume Ve and the internal particle pore volume Vi the total packing
represent a frictional system in which random loose packing con- porosity is also given by
figurations are more likely to occur during packing processes [38]. Ve + Vi
Hence, during the packing process, interparticle friction prevents εt = = εe + εi (2)
V
the bed from becoming closely packed but at the same time fric-
where ␧e and ␧i are the external or interparticle and the intraparti-
tion is required to keep the bed in place, once consolidated [39].
cle porosity, respectively [45]. The packing density or the fraction
As a consequence, different packing methods result in quite differ-
of the column occupied by the solid particle skeleton is then
ent mechanical behavior of the packed bed and in a different local
stress distribution. Vs V − Vv
= = = 1 − εt (3)
Flow packing is a widely accepted method for the consolidation V V
of chromatographic packed beds as it causes suspended particles to During packing compression, the void volume is reduced and
pack more quickly and uniformly than during gravity settling [42]. the packing density increases. The compression of the packed bed is
The latter leads to a classification of particles where the larger par- described by the compression factor ␭ as the ratio of bed subsidence
ticles settle first and the smaller particles gradually settle above the and initial packing height according to
larger particles. However, the reproducibility of flow packing meth-
ods was reported to be only fair [36]. Dynamic axial compression ho − h
␭= (4)
(DAC) is reported to have several advantages over the flow packing h0
method by reducing particle size segregation due to gravity settling where h0 is the initial gravity settled height and h is the compressed
and providing a constant compression pressure on the whole pack- bed height [19]. A previous study of the compression of packed beds
ing. Moreover, the desired level of compression may be achieved consisting of soft agarose particles [44] revealed that the average
by only one compression step [42]. However, axial compression particle retained its spherical shape and did not show any signifi-
results in a significant increase in particle stress when the piston cant compression or deformation. Hence, as a first approximation,
penetrates the surface of the packing leading to fragmentation and the reduction in void volume is mainly caused by a reduction of the
particle breakage [36,43,44]. external particle porosity ␧e . For large packing compression, as usu-
Due to the restricted experimental accessibility of the param- ally applied during column packing, Keener et al. [46] defined the
eters influencing the dynamic column packing behavior, the packing compression as  = −du/dz = (h0 − h)/h. The packing density
theoretical understanding of the complex consolidation mechan-  and the packing compression factor are then correlated by
ics of packed beds of compressible particles is still limited. Hence,
little work was dedicated so far to the fundamental analysis and  = (1 − εo )(1 + ) (5)
model-based prediction of column packing dynamics by explicit
where ␧0 is the external particle porosity under gravity settled
consideration of single particle behavior as well as the interpar-
packing conditions [46].
ticle micromechanics coupled to the fluid flow. Therefore, in the
present work, a systematic analysis of common column packing
2.2. Variation of axial packing density profiles
methods of soft compressible resins will be presented regard-
ing its influence on the dynamic packing compression-relaxation
The variation of packing density profiles of columns, which were
behavior, hydrodynamic packing long-term stability, and packing
packed by different packing procedures, was quantified by common
quality in terms of eddy dispersion that, to our knowledge, has
statistical measures [47]. Hence, as an indicator of the inhomogene-
not been published before. Using a three-dimensional determin-
ity of the axial packing density distribution, we define the average
istic modeling approach describing the hydrodynamic interaction
deviation of the axial packing density from the mean packing den-
of the compressible chromatographic packing with the fluid phase,
sity as
the intrinsic packing properties such as particle stress and force

transmission will be determined which up to now have not been 1 n 2
accessible. For this purpose, a simplified chromatographic system Smean = (i − <  >) , (6)
n i=1
using pure water as the fluid phase will be considered in absence
of any salt or tracer substance which might have an influence where n is the number of column sections in which the packing
on the mechanical and hydrodynamic column packing behavior. density was determined, i is the packing density in section i and
Hence, it was not the aim of the present work to investigate sepa- <> is the mean packing density of the column. As an example,
ration performance per se. The expected findings will give valuable Smean = 0 corresponds to a column with a constant packing density
information on hydrodynamic phenomena observed during col- in axial direction.
umn operation and will show that packing quality can be enhanced The variation of the axial packing density profiles during column
using a proper column packing method. operation was described in analogy by the average deviation of the
 M. Dorn et al. / J. Chromatogr. A 1516 (2017) 89–101 91

packing density profile during operation from the initial packing top of the column according to the guidelines of the flow pack-
density profile right after column packing according to ing method described by Bemberis et al. [42]. The packing fluid
 flow rate was adjusted to 1.6 L h−1 , being equal to a superficial
1 n  2 flow velocity of 2210 cm h−1 for the given column geometry. This
Sinit = ˚i − i,init . (7)
n i=1 superficial flow velocity was 65% of the critical superficial flow
velocity (approx. 3400 cm h−1 ) which was within the range of rec-
Here, i,init is the packing density of section i after column packing.
ommended superficial flow velocities for flow packing of 63–90%
according to literature [42]. Flow packing by pumping the resin
3. Materials and methods slurry into the column from the bottom or lower sides of the col-
umn, which is known as the “Pack-In-Place” method [42], was not
3.1. Column, chromatography material, and fluid phase used in this work as this method would rule out the axial section-
ing of the column. However, sedimentation of small amounts of
The experimental analysis of column packing methods was particles required for column sectioning might result in a slightly
carried out in a novel pressure-resistant micro-chromatography different particle size distribution compared to gravity settling
®
column. The column corpus was made of a 3 mL B Braun Omnifix from stirred slurry as described in [42] which represents a limi-
Luer Lock Solo polypropylene syringe (Carl Roth GmbH, Karl- tation of the present study. The third packing method was flow
sruhe, Germany) with an inner diameter of 9.6 mm and a tube packing to reach 10% compression followed by DAC to reach a final
length of 60 mm. The resin was retained by a 3 mm thick porous compression of 20%. The fourth packing method was DAC com-
polyethylene frit with a pore size <20 ␮m and a negligible pressure pression by 10% followed by flow packing to reach again a final
drop (BEKOlut GmbH & Co. KG, Hauptstuhl, Germany). Preliminary compression of 20%. The fifth packing method was incremental
experiments were performed using the agarose-based resins Affi- alternating application of flow compression and DAC (the so-called
Gel 102 and Affi-Gel Blue Gel (kindly provided by Bio-Rad, Munich, hybrid packing method). Flow packing was applied first by adjust-
Germany). All other experiments were performed at a controlled ing the fluid velocity to achieve a 2.5% compression of the packing.
constant temperature of 20 ◦ C. The chromatographic resins were The fluid flow was then stopped and further 2.5% compression
® ®
CM Sepharose 6 FF and Blue Sepharose 6 FF (kindly provided was achieved by DAC. This was repeated three times resulting in
®
by GE Healthcare Europe, Munich, Germany). CM Sepharose 6 FF a final bed compression of 20%. As all packing procedures were
(resin SEP) is a weak cation exchange resin made of spherical 6% adjusted to an overall packed bed compression of 20%, the adapter
cross-linked agarose. According to the manufacturer, the particle was set to identical heights assuring identical initial conditions for
diameters in wet state are within the range of 45–165 ␮m. The further analysis of hydrodynamic packing stability and homogene-
®
mean diameter in wet state is given at 90 ␮m. Blue Sepharose 6 FF ity. Like in industrial applications, the packed columns were stored
®
is Cibacron Blue 3G dye coupled to Sepharose 6 FF . The particle overnight after packing allowing the internal packing stress to be
size distribution was measured using a standard optical microscope equilibrated.
type Axioplan and evaluated by ImageJ (results not shown). As We defined a relative hydrodynamic load of RHL = 1 to be
fluid phase, deionized water was used with 10 mM NaOH added equal to the flow rate applied during the flow packing method in
to prevent contamination by microorganisms. order to achieve a 20% packing compression (1.6 L h−1 ). For flow
packed columns using SEP resins, it is recommended to not exceed
3.2. Column packing methods and operation RHL = 0.75 during operation [48] and we used this value as a refer-
ence load for comparison of the packing methods. Hydrodynamic
Columns were prepared by gravity settling of the CM Sepharose stability of the different packings was analyzed by operation at
®
6 FF (SEP) particles. During this process, thin layers of Blue stepwise increase of hydrodynamic load by RHL = 0.25 every 20 min
®
Sepharose 6 FF were inserted as described elsewhere [44]. A simi- up to a maximum of RHL = 1.25 (approx. 81% of the critical super-
lar concept to visually separate axial sections of a chromatography ficial flow velocity). Long-term packing behavior was analyzed by
column was provided earlier by Shalliker et al. [39]. This resulted repetition of 25 simulated load/elution/equilibration cycles. During
in an optical partitioning of the sedimented packed bed into seven each simulated cycle, the fluid velocity was adjusted to RHL = 0.75
equally sized sections (see Fig. 1). Preliminary experiments of the lasting 10 min and RHL = 0.05 lasting 10 min, respectively.
same kind as described above using the agarose-based resin Affi-
®
Gel 102, which has a refractive index similar to water, and thin 3.3. Computational method
®
layers of Affi-Gel Blue Gel revealed that the radial differences in
packing compression were negligible compared to the axial pack- 3.3.1. CFD-DEM approach
ing compression. Due to the material similarities of Sepharose and The coupling of Computational Fluid Dynamics (CFD) with the
Affi-Gel and the comparability of the experimental procedures, we Discrete Element Method (DEM) is a powerful and deterministic
assumed that the radial differences in packing compression were modeling technique for the simulation of large-scale disperse fluid-
negligible compared to the axial packing compression in the case of particle systems [49–53]. The Euler-Lagrange coupled CFD-DEM
Sepharose as well. As a result, the axial packing compression pro- approach relies on the solution of volume-averaged Navier-Stokes
files could be obtained by simply measuring the change in height equations to calculate the motion of the fluid phase as well its
of each section after packing as described previously [44]. properties in the presence of a secondary particulate phase.
The bed height of the gravity settled packing was adjusted to
30 mm. After gravity settling, the bed was further consolidated ∂ (f ˛f )
+ ∇ · (f ˛f uf ) = 0 (8)
and compressed by 20% according to five different column packing ∂t
methods given in Fig. 1. Each packing experiment was performed Eq. (8) describes the continuity equation for a compressible
in a separate column. The resulting final bed height was 24 mm in fluid, where f is the fluid density, ␣f is the volume fraction occu-
all cases. pied by the fluid, and uf is the fluid velocity. The momentum
The first packing method applied was dynamic axial compres- conservation equation can be written as
sion (DAC). Consolidation of the particle packing was achieved
mechanically by moving the column plunger downward man- ∂ (f ˛f uf )
+ ∇ · (f ˛f uf uf ) = −˛f ∇ p − Rf,p + ∇ · (˛f  f ) , (9)
ually. The second packing method was flow packing from the ∂t
92 M. Dorn et al. / J. Chromatogr. A 1516 (2017) 89–101

Fig. 1. Overview of the different column packing methods. The methods were applied to a sequentially gravity settled unconsolidated particle bed being composed of CM
® ®
Sepharose 6 FF and thin layers of Blue Sepharose 6 FF . The arrow indicates the direction of flow during operation of the column after packing. The column diameter was
9.6 mm and the packing height was 30 mm (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

were p is the pressure and ␶f is the stress tensor for the fluid phase. be described either by linear (Hooke’s law) or non-linear (Hertzian
Rf,p is the volumetric particle-fluid interaction force, i.e. the force contact) contact laws [57,58]. The linear contact model was used for
exerted on the fluid phase by the moving particle phase [49,54]. this study. The DEM simulations were performed using the open-
The motion and interaction of each single particle is calculated by source software package LIGGGHTS [55].
the Lagrangian DEM by explicitly solving the particle’s trajectory. For solving the CFD governing equations, a pressure-based
The force balance for the i-th particle reads solver using the well-known Pressure Implicit with Splitting of
Operator (PISO) algorithm is used [59]. The coupling of CFD
mi ẍi = Fi,n + Fi,t + Fi,f + Fi,b , (10) and DEM code is applied within the CFDEM coupling framework
where mi is the mass of particle i, ẍi its acceleration, and Fi,n and (www.cfdem.com) [54]. All simulations were performed on a DELL
Fi,t are the normal and tangential particle–particle contact forces Precision computer (Intel Xeon E5-2699-v3 Processor, 18 cores at
[55,56]. Fi,f is the fluid drag force exerted on the particles. Other 2.3 GHz, 64 GB RAM).
body forces like gravity, electrostatic, and hydrophobic interactions
are summed up in Fi,b . 3.3.2. Packing simulation procedure
The momentum balance of particle i is described by A packing of 20k discrete particles with an experimentally
dωi acquired particle size distribution of SEP resin was simulated using
Ii = ri,c × Fi,t + Ti,r (11) the CFDEM coupling framework in a pseudo two-dimensional sim-
dt
ulation domain. The size of the simulation domain was 10 mm,
where Ii is the moment of inertia of particle i, ␻i is its angular veloc-
0.6 mm and 40 mm in x-, y- and z-direction, respectively, with
ity, and ri is its radius [55,56]. In DEM, the particles are assumed
frictional walls in x- and z-direction and periodic boundaries in
to be spherical. Hence, effects due to non-sphericity of real par-
y-direction. This was equivalent to the size of the laboratory col-
ticles are considered by an additional torque Ti,r . The sphericity
umn. Inlet and outlet were set on the z-axis, fluid flow and gravity
assumption allows the calculation of the translational and angular
were in negative z-direction. The pseudo two-dimensional domain
accelerations by corresponding force and momentum balances. In
geometry was found to give results similar to a three-dimensional
this work, a so-called soft-particle DEM model was applied where
cylinder geometry which is computationally much more elabo-
the particles are allowed to overlap slightly. The repulsive normal
rate to simulate (results not shown). In order to further speed
force Fn can be derived from the spatial overlap ␦n and the nor-
up the calculation, a coarse-graining of 3 was applied [60–62].
mal relative velocity un by a simple Kelvin-Voigt spring-dashpot
The particle Young’s modulus, Poisson’s ratio, and coefficient of
model with spring and damper in parallel [54]
restitution were set to 35 MPa, 0.2 and 0.5, respectively, resulting
Fn = −kn ın + cn un . (12) from single particle compression experiments carried out earlier
(results not shown). The particle-wall and interparticle friction
The magnitude of the tangential contact force due to shear stress was set to 0.5 and 0.1, respectively, which were identified in a
is given by parameter study to give the best matching to the measurements
  
t (results not shown). The initial simulated packing was achieved
Ft = min |kt ut dt + ct ut |, Fn , (13) by a random distribution of the particles and subsequent gravity
tc,0 settling.

where ut is the relative tangential velocity of the particles in con-

tact [54]. kn , kt and cn , ct are the normal spring, tangential spring, 3.3.3. Calculation of integral porosity deviation
and damping coefficients, respectively. The tangential contact force The integral porosity deviation (IPD) [63] is a measure for the
model also holds for particle-wall contact and is truncated to ful- trans-column contribution to eddy dispersion that may result from
fil the well-known Coulomb friction criterion. The tangential force radial local porosity distributions. The IPD calculates the integral
is limited to a maximum force of ␮Fn until sliding begins with ␮ local porosity deviations from the packing bulk porosity ␧bulk under
being the coefficient of friction. The contact of two particles can consideration of all porosity inequalities that a tracer would expe-
 M. Dorn et al. / J. Chromatogr. A 1516 (2017) 89–101 93

rience when diffusing from the column wall towards the packing also observed during column packing at high pressures using the
center as [63] downward packing method [69].
 R On order to analyze the micromechanical compression effects
IPD = (εr (r) − εbulk ) dr. (14) on the particle scale, a numerical study of packed bed compres-
0 sion was performed. A pseudo two-dimensional packing of 2200
discrete particles with an experimentally acquired particle size dis-
3.3.4. Axial hydrodynamic dispersion simulation tribution of SEP particles was compressed in-silico by nearly 30%
The heterogeneity of a packed bed consisting of frictional parti- by DAC (compression force Fmech = 0.48 N) as well as by flow pack-
cles is quantified by transient hydrodynamic dispersion of the fluid ing with a high flow rate (u0 = 2763 cm h−1 ). This compression was
phase flowing through the packed bed. The axial hydrodynamic higher than in the laboratory experiment, but chosen to make the
dispersion (neglecting any diffusion effects) ␴z 2 is determined differences more clearly. The size of the simulation domain was 10
according to mm, 0.4 mm and 10 mm in x-, y- and z-direction limited by periodic
1 N
boundaries in y-direction and frictional walls in x- and z-direction.
 z 2 (t) = (vz,i (t)− < vz (t) >)2 , (15) Fig. 3 shows the interparticle connection forces in a packing dur-
N i=1
ing DAC of (a) particles with a low friction coefficient of ␮ = 0.1, (b)
as the second central statistical moment of the local fluid veloc- particles with a high friction coefficient of ␮ = 0.5 and (c) and dur-
ity calculated in each CFD cell [47]. N is the number of CFD cells, ing flow packing (␮ = 0.1). Particle-wall friction was set to 0.5 in all
vz,i (t) is the fluid flow velocity in axial direction calculated in the cases. Further particle properties were set to the values given in the
ith CFD cell at time t, and <vz (t)> is the fluid flow velocity in axial methods section. The interparticle connection forces form a force
direction at time t averaged over all CFD cells. Neglecting any diffu- chain network [70] which is scaled and colored according to the
sion effects, this parameter focuses solely on the contribution of the magnitude of compression pressure of the individual particles. Dur-
flow velocity differences to hydrodynamic dispersion and, hence, ing the DAC method (Fig. 3a, b), high particle compression forces
gives a reasonable measure of the packing heterogeneity. A similar are located in the upper packing region close to the adapter which
approach was made by Tallarek and co-workers [64,65]. is represented by the plate in Fig. 3a,b. It can be noticed that the par-
ticle compression pressure decayed with increasing distance from
4. Results and discussion the adapter.
Moreover, it can be seen by comparing Fig. 3a,b that the axial
4.1. Influence of friction forces on the column packing process force transmission is influenced by the friction between the par-
ticles and between the particles and the column wall as well. In
Fig. 2 exemplarily shows the measured axial packing density the packing with lower friction (Fig. 3a), the compression force
profiles of wet and semi-dry SEP resin packed by DAC. On the left, decayed over a shorter span of particles compared to the packing
a schematic drawing of the laboratory column is given which was with higher friction (Fig. 3b) which is in analogy with the mea-
divided into seven equally sized sections (see Fig. 1). By measuring sured packing density profiles of the wet and semi-dry packed
the individual height of each section during the column packing columns shown in Fig. 2. Furthermore, shorter and more horizon-
process, the packing density in each section was calculated accord- tally oriented forces chains can be identified in the system with
ing to Eq. (5). The gravity settled packed bed porosity was not lower friction. This observation indicates that the forces were trans-
experimentally accessible. Therefore, a value of 0.38 was assumed mitted increasingly in radial direction (along the x-direction). In
according to literature data [e.g. 45,66–68]. The column was packed contrast to this, longer force chains are visible in the system with
using wet particles as well as semi-dry particles. In the former pack- higher friction due to stronger axial force transmission (along the
ing process, the particles were first gravity settled from resin slurry z-direction). On the basis of these findings, the difference in com-
and afterwards compressed mechanically by lowering the plunger. pression behavior of the wet material and the semi-dry material
The column inlet valve was then closed so that excess fluid left the can be explained. During the packing compression of the wet DAC
column through the outlet. The particles of the packing were still packing, the interparticle fluid may act as a lubricant decreasing
surrounded by fluid. In the latter packing process, the interparticle friction. Hence, the particles can slide more easily against each
fluid was drained by gravity first. Then, most of the fluid contained other promoting particle rearrangement. As a consequence, the
in the interparticle void space was removed by a nearly stagnant axial compression force is distributed more evenly in radial direc-
flow of about one column volume of air through the column. This tion among the individual particles and finally dissipated to the
resulted in a semi-dry gravity settled packed bed. However, the column wall. The packing compression behavior of the semi-dry
residual moisture content of the resin particles was not quanti- packing is governed by higher friction between the particles which
fied because this parameter was not experimentally accessible. prevented them from sliding against each other. This resulted in an
The semi-dry packed bed was then compressed mechanically up increased force transmission in axial direction.
to 12.5% by lowering the plunger by 20 mm min−1 . The packing Flow packing, in contrast, led to packing compression behavior
procedure was repeated five times and the experimental error was which differed noticeably from the mechanical packing compres-
determined to be less than 10%. It can be seen that both packings sion method. High particle compression pressures are located in
exhibited a different compression behavior. Compressing the wet the lower packing sections close to the column outlet as it can
packing material resulted in a homogeneous packing density in be seen in Fig. 3c. Moreover, the force distribution is much more
axial direction in sections 2–6. Column section 1, which is clos- heterogeneous in axial as well as in radial direction. Fewer force
est to the adapter, showed the highest packing density indicating chains transmitting high forces can be identified. Compared to the
that this section was compressed the most. However, section 7 also DAC packing method, wall support is clearly less present during
showed a higher packing density than the middle sections. As this flow compression. Here, the force chains did not show any signif-
section is opposite to the adapter and closest to the column out- icant contact to the column wall. During DAC, the force chains are
let, the compression was most likely induced by drag forces of the noticeably connected to the outer particle layer closest to the col-
outflowing fluid. In contrast, the compression of the semi-dry pack- umn wall which indicates a strong influence of the wall effect. As
ing material resulted in a more pronounced axial packing density pointed out by Shalliker et al. [39], friction prevents the particles
gradient. The packing density decreased from section 1 to section from being homogeneously packed but at the same time, friction
7 by about 8%. Pronounced axial packing density gradients were is required to stabilize and keep the packed bed in place. As the
94 M. Dorn et al. / J. Chromatogr. A 1516 (2017) 89–101

Fig. 2. Measured axial packing density profiles resulting from dynamic axial compression (DAC) of (a) wet chromatographic material and (b) semi-dry material. In the latter,
the interparticle fluid was removed first.

Table 1 variation of the packing density profile during the reference hydro-
Deviations Smean and Sinit of the different column packing methods during a relative
dynamic load of RHL = 0.75, respectively.
hydrodynamic load of RHL = 0.75.
The initial packing profiles given in Fig. 4 refers to the axial
Deviation Smean Sinit packing density profiles right after the column packing procedure
DAC 0.04 0.07 was completed. Fig. 4a shows the axial packing density profiles
Flow packing 0.04 0.02 after DAC packing. After overnight storage, no remarkable change
DAC and flow packing 0.06 0.03 in the packing density profile was observed, only the uppermost
Flow packing and DAC 0.05 0.03
section relaxed. However, the profile changed significantly dur-
Alternating flow packing and DAC 0.03 0.02
ing increasing hydrodynamic load. The packing density in section
1 was decreased and the lower packing sections were considerably
packing compression behavior depends on the small-scale particle compressed. As can be seen in Table 1, this resulted in the largest
dynamics, the packed bed behavior can be influenced by the way deviation from the initial profile of Sinit = 0.07.
the column is packed. Particle sliding and rearrangement as well Fig. 4b shows the results of a column packed by flow pack-
as the wall effect can be intensified or reduced by the application ing. During the overnight storage, the packing density profile
of different column packing strategies for the purpose of increased changed from the initial packing state (blue curve) to a force-
hydrodynamic packing stability. equilibrated state where the packing density increased near the top
of the column and decreased near the bottom of the column. We
assume that the compression energy which was stored to a large
4.2. Comparison of different column packing methods at different
extent in (visco)-elastic particle compression of the lower pack-
hydrodynamic loads
ing sections caused a compression of the upper packing sections
during the overnight force equilibration time period. However,
Based on the findings of the analysis of packed bed properties
this equilibrated packing density profile changed with increasing
during dynamic axial compression and flow compression, different
hydrodynamic load. When the hydrodynamic load was increased,
column packing methods were analyzed experimentally regarding
the packing density profiles reversed again leading to increased
hydrodynamic packing stability and axial packing homogeneity.
compression of the lower packing sections and to a relaxation of
Fig. 4 shows the axial packing density profiles of the laboratory
the upper packing section. Substantial changes in the packing den-
column packed according to five different packing strategies. The
sity profile were observed when reaching the hydrodynamic load
overall packing compression was 20% in all experiments.
during column packing (RHL = 0.0) and even more when exceeding
Table 1 shows the deviations Smean and Sinit as measures of
this load (RHL = 1.25).
the inhomogeneity of the axial packing density profile and of the
 M. Dorn et al. / J. Chromatogr. A 1516 (2017) 89–101 95

The following packing strategies were combinations of the pre-

vious discussed flow packing and DAC. As a consequence, the
hydrodynamic compression behavior of the columns packed by
these methods is comparable to flow packed or DAC packed
columns leading to a relaxation of the upper packing sections and
to a compression of the lower packing sections. The column packed
by DAC and subsequent flow compression (Fig. 4c) showed a minor
deviation of the initial packing density profile up to hydrodynamic
loads of RHL = 0.75 (Sinit = 0.03). Hence, the hydrodynamic packing
stability was increased by combining DAC and the flow packing
method. When the applied hydrodynamic load was increased up
to RHL = 1.25, a pronounced increase of packing density in the low-
ermost section was observed.
By switching the sequence of DAC and flow packing (Fig. 4d), a
different hydrodynamic packing behavior was obtained. Compared
to profiles shown in Fig. 4c, the packing was more homogeneous
(Smean = 0.05 instead of 0.06) whereas the average variation of the
packing density profile during load was identical (Sinit = 0.03).
The best hydrodynamic stability was achieved by an alternating
repetition of flow packing and DAC as indicated in Fig. 4e. Up to rela-
tive hydrodynamic loads of RHL = 0.75, the packing showed a more
homogeneous packing compression in axial direction (Smean = 0.03)
and the smallest variations in its axial packing density profile (Sinit
= 0.02) compared to the other packing strategies.

4.3. Measured column behavior during long-term cyclic operation

Three column packing methods, DAC, flow packing, and alter-

nating flow packing and DAC (the hybrid packing method) were
analyzed with regard to the long-term hydrodynamic stability of
the packing. For this purpose, the packed columns were oper-
ated for 25 sequential simulated load/elution/equilibration cycles
in which the packing was compressed and relaxed alternatingly.
The axial packing density profiles were measured during each cycle
(Fig. 5).
The variation of the packing density profile during operation Sinit
is given in Fig. 6 as the deviation from the initial packing density
profile.
Fig. 5a shows axial packing density profiles resulting from the
DAC packing method. The initial packing profile was obtained by
lowering the adjuster with a rate of 30 mm min−1 leading to a pack-
ing compression of 20%. This caused an exponentially decreasing
initial packing density profile from column top to bottom. During
cyclic operation, the packing density decreased in the uppermost
packing section and increased in the lower packing section. No
considerable variation of the axial packing density profiles was
observed from the 15th cycle to the 25th cycle as can be seen in
Fig. 6 by a nearly constant deviation from the initial packing den-
sity profile of Sinit ≈ 0.043. Hence, under the applied hydrodynamic
load, the packing reached a stable packing density profile fast dur-
ing cyclic operation. Flow packing resulted in a pronounced initial
axial packing density profile in which the lowermost section 7 was
compressed as mentioned above. The packing top section was less
compressed. During cyclic operation, the packing density profile
varied noticeably as it can be seen from Fig. 5b and from Fig. 6 by
a steadily increasing deviation of the packing density profile from
the initial packed state. The packing top section was compressed
Fig. 3. Simulation of the force transmission in a frictional particle packed bed. The gradually and the packing bottom section was relaxed gradually.
force chain network is colored and scaled according to the magnitude of compression It can be deduced that hydrodynamic stability was not achieved
force. Mechanical compression of (a) wet packing with low friction, ␮ = 0.1, (b) semi-
within the 25 compression/relaxation cycles. This may be due to
dry packing with increased friction, ␮ = 0.5 and (c) flow compression of packing,
␮ = 0.1.
two reasons: i) flow packing led to a highly compressed packing
bottom region, which, as a direct consequence of the interparti-
cle as well as particle-wall friction effects, is difficult to relax; ii)
the direction of fluid drag forces and packing relaxation was oppo-
site. During flow packing, the major part of the compression energy
was stored in the highly compressed lower packing sections. Dur-
96 M. Dorn et al. / J. Chromatogr. A 1516 (2017) 89–101

Fig. 4. Measured variations of the axial packing density profiles of the laboratory column during different hydrodynamic loads applied after packing. The column was packed
according to five different column packing methods (a) DAC, (b) flow packing, (c) DAC (10% packing compression) followed by flow packing (10% packing compression), (d)
flow packing (10% packing compression) followed by DAC (10% packing compression), (e) alternating 2.5% flow packing followed by 2.5% DAC performed 4 times.

ing cyclic operation, fluid drag pushed the particles downwards in the same direction during DAC packing leading to a faster force
while the particles were pushed upwards during packing relax- equilibration of the column. The hybrid packing method resulted
ation. This slowed the process of force equilibration of the column. in a similar hydrodynamic long-term stability as the straight DAC
In contrast to this, fluid drag forces and packing relaxation acted packing method as shown in Fig. 5c. Only very small changes of the
 M. Dorn et al. / J. Chromatogr. A 1516 (2017) 89–101 97

Fig. 5. Measured long-term axial packing density profiles of columns packed by (a) DAC, (b) flow packing and (c) hybrid packing method. Standard deviations were less than 1%.

the long-term packing density profile. However, this optimization

was beyond the scope of the present work.

4.4. Simulated column behavior during long-term cyclic

operation

It is difficult to correlate the changes of the axial packing den-

sity to column performance, i.e. flow distribution and packing
asymmetry [71–73]. Radial variations of the packing configuration
and porosity distribution are known to have a serious effect on
often observed packing and peak asymmetry [71]. The results of
this study showed that axial packing density variations become
increasingly important when the packing exhibits a pronounced
compression/relaxation behavior. In particular if the packing shows
a wide particle size distribution, migration of smaller particles may
be enforced with pronounced axial variations of the packing com-
pression/relaxation behavior. Furthermore, differences between
the wall friction and the interparticle friction may lead to differ-
ences of the packing relaxation behavior of the near-wall packing
region compared to the inner packing region. Either leads to radial
variations of the packing properties and is highly influenced by
the axial variations and packing compression/relaxation behav-
ior. In order to analyze this, we carried out computer simulations
Fig. 6. Visualization of the variation of the packing density profiles from measure-
ments for the three packing methods during cyclic operation via parameter Sinit .
of the packing behavior during hydrodynamic load and evaluated
the effect of the different column packing methods on the packing
asymmetry and fluid flow distribution. For this purpose, we simu-
packing density were measured in the uppermost and lowermost lated the column packing based on the gravity settled packed bed
packing section whereas in the middle sections, the packing density according to three different column packing methods. The packed
was unchanged. Compared to the DAC and flow packed column, this column was then operated by 10 simulated cycles using the same
column showed the least differences between the initial and the fluid velocity as in the experiments. This amount of cycles was
long-term packing density profile (Sinit ≈ 0.02). As a consequence, found to be sufficient as no further changes were obtained with
we deduce that alternating combination of flow packing and DAC ongoing simulated cyclic operation. The computational effort of the
is the most appropriate column packing method which distributes simulations for each cyclic column operation was about 10 h on a
the compression forces more homogeneously among the particles 16 core Intel Xeon E5 processor (2.3 GHz).
and avoids regions of high particle compression stress as observed Fig. 7 shows the simulated axial packing density profiles of
in DAC and flow packed columns. Furthermore, the ratio of mechan- columns packed by (a) DAC, (b) flow packing, and (c) the hybrid
ical compression to flow compression of this packing method can be method. The overall packing compression was 20% as in the exper-
optimized in order to align the initial packing density profile with iments. The simulated axial packing density profiles reproduced the
98 M. Dorn et al. / J. Chromatogr. A 1516 (2017) 89–101

Fig. 7. Simulation of the axial packing density profiles during long-term operation of columns packed by (a) DAC, (b) flow packing and (c) hybrid packing method.

measured packing behavior qualitatively although the model was

not able to reflect the measured axial packing density gradients as
pronounced. DAC packing resulted in an exponentially decreasing
initial packing density profile from column top to bottom. Already
after the fourth simulated operational cycle, the packing density
profiles showed no further variation leading to a constant deviation
of Sinit ≈ 0.05. The simulated flow packed column (Fig. 7b) in con-
trast showed minor changes in the axial packing density profile in
nearly all operation cycles similar to the measurements. The lower
packing sections relaxed and the upper packing sections were com-
pressed as a result of force relaxation. This led to an increasing
deviation of the packing density profile from the initial packed
state from Sinit = 0.013 to Sinit = 0.019. As in the experiments, the
simulated behavior of the column packed by the hybrid method
exhibited the least axial deviation from the initial packing den-
sity profile (Sinit ≈ 0.014). The packing was force-equilibrated right
after the first simulated cycle and kept being hydrodynamically
stable during further simulation. The deviations of the simulated
packing density profiles were lower than the deviations of the mea-
sured profiles as a consequence of the inability of the model to
reproduce the pronounced packing density gradients. This is most
probably due to the fact that in the DEM model, particle–particle
interactions are described according to a simplified linear Hookean
behavior. However, the dynamic behavior of the packed beds was
still reproduced by the model in a sufficient manner.

4.5. Analysis of simulated radial fluctuations of the axial flow

velocity

Based on the simulations, a detailed analysis of the packing Fig. 8. Simulated radial fluctuations of the axial flow velocity of the (a) DAC, (b) flow
heterogeneity in terms of packing asymmetry and fluid flow distri- packed column, and (c) column packed by the hybrid method. The local axial flow
bution was possible. Fig. 8 shows the simulated radial fluctuations velocity was normalized to the average inner packing axial flow velocity.
of the axial fluid flow velocity obtained by the different column
packing methods during cyclic operation. The local axial flow veloc-
ity uz was normalized to the average packing axial flow velocity. column wall was consistent with results of Knox and co-workers
According to the boundary conditions, the axial flow velocity at [32,33]. In our simulations, the wall effect was prominent up to a
the wall is zero. Close to the column wall, the simulation results distance from the column wall of about four average particle diam-
showed an about 1.6-fold higher flow velocity than the average eters. The differences of the flow velocity distributions of the three
flow velocity. This effect of increased flow velocity in vicinity of the different packings during simulated operation were small but visi-
 M. Dorn et al. / J. Chromatogr. A 1516 (2017) 89–101 99

Fig. 10. Calculated axial hydrodynamic dispersion in columns packed by

the three different packing methods during operation lasting 10 simulated
load/elution/equilibration cycles. Mean values are given by the dash-dotted lines.

Fig. 9. Simulated radial porosity distribution of the (a) DAC, (b) flow packed column, the porosity profile over a distance of two particle diameters from
and (c) column packed by the hybrid method as a function of the mean particle the column wall, whereas only one oscillation was observed for
diameter dp . The integral porosity deviation (IPD) is given exemplary for the 10th the other two packings. This indicates a higher ordered packing
simulated operation cycle.
structure in the column wall region for the DAC packed column.
However the porosity profiles varied slightly during ongoing oper-
ble. The DAC packed column showed the largest variations of axial ation cycles for both DAC and flow packed columns (Fig. 9a, b). The
flow velocity in radial directions with a 1.65-fold increased velocity hybrid packing method (Fig. 9c) resulted in the least variations of
in the vicinity of the column wall. the packing porosity profile. This method also resulted in the lowest
The particles are unable to form a dense packing configuration IPD of 0.72 after the 10th simulated cycle and therefore showed the
against the rigid flat surface of the column wall. Hence, the first least deviations of local porosity from the packing bulk porosity. In
particle layer with contact to the column wall is highly ordered contrast, the IPD values were 0.79 for the DAC packed column and
and differs from subsequent layers, because the interstitial void 0.87 for the DAC packed column which indicated a higher degree
between the column wall and the first layer of particles cannot be of hydrodynamic dispersion.
partially occupied by other particles [63]. This results in a higher
packing porosity and packing permeability in the column wall 4.6. Evaluation of axial hydrodynamic dispersion
region. The subsequent particle layers exhibit an increasing ran-
dom packing configuration toward the packing bulk [63]. Fig. 9 The observed porosity oscillations as well as the deviations of
shows the geometrical wall effect resulting in oscillating poros- the local porosity from the packing bulk porosity cause the radial
ity profiles. The mean particle diameter of the PSD was chosen variations of the axial fluid flow velocity as shown in Fig. 8. To evalu-
as reference diameter. As can be seen, the porosity oscillations ate the variations of the axial fluid flow velocity, the overall packing
were reduced significantly after a distance of about two particle induced axial fluid dispersion ␴z 2 was calculated as the second cen-
diameters to the column wall. Hence, based on radial profiles of tral statistical moment of the local fluid flow velocity according to
the porosity and axial velocity, the region from a distance of four Eq. (15) for 10 simulated load/elution/equilibration cycles (Fig. 10).
particle diameters towards the column center was defined to be ␴z 2 was calculated regardless of thermodynamic molecular dif-
the packing bulk region which can be characterized by the packing fusion effects and can be seen as a direct measure of packing
bulk porosity ␧bulk . To obtain a quantitative measure of the pack- quality in terms of eddy dispersion. The average hydrodynamic
ing heterogeneity and contribution of the local packing porosity dispersion values for DAC and flow packing were 0.72 cm2 s−2 and
distribution to trans-column eddy dispersion, the integral poros- 0.76 cm2 s−2 , respectively. The hybrid column packing method led
ity deviation (IPD) parameter was calculated according to Eq. (14) to a more homogeneous packing porosity flow velocity distribution
[63]. This calculation was performed over the whole packing cross in axial direction which is indicated by the least axial hydrody-
section (equivalent to 48 dp ). Based on the radial porosity distri- namic dispersion with an average value of 0.63 cm2 s−2 . Hence,
bution given in Fig. 9, the bulk porosity ␧bulk was calculated in the calculated packing induced dispersion confirmed what was
a region ranging from 4 dp from the column wall to the column expected on the basis of the calculated IPDs. Moreover, pronounced
axis. As illustrated in Fig. 9, DAC, flow packing and the hybrid pack- fluctuation of the packing dispersion during cyclic operation was
ing method resulted in different local porosity distributions. The observed for all three packings. These fluctuations were attributed
DAC packed column (Fig. 9a) exhibited two visible oscillations of to radial variations of the axial flow velocities (see Fig. 8) which are
100 M. Dorn et al. / J. Chromatogr. A 1516 (2017) 89–101

caused by the variations of the axial packing density profiles (see Dorn by the TUM Graduate School, Technical University of Munich,
Fig. 7). The column packed by the flow packing method shows a Germany, is gratefully acknowledged.
steadily decreasing hydrodynamic dispersion from the 7th opera-
tion cycle on which indicates that the packing was still engaged in
the force equilibration process. This observation was in accordance References
with above mentioned statement that the column had not reached
[1] A.S. Rathore, R.M. Kennedy, J.K. O’Donnell, I. Bemberis, O. Kaltenbrunner,
a stable consolidated state during cyclic operation (see Fig. 6). Qualification of a chromatographic column, Biopharm. Int. 16 (2003) 30–40.
[2] E. Boschetti, Advanced sorbents for preparative protein separation purposes,
J. Chromatogr. A 658 (1994) 207–236.
5. Conclusions [3] A. Jungbauer, Chromatographic media for bioseparation, J. Chromatogr. A
1065 (2005) 3–12.
[4] G. Carta, A. Jungbauer, Protein Chromatography − Process Development and
A systematic modeling-based analysis of different column pack- Scale-up, Wiley-VCH, Weinheim, 2010.
ing methods was provided focusing on axial packing heterogeneity [5] J. Knox, H. Pyper, A framework for maximizing throughput in preparative
and hydrodynamic long-term packing stability. Macroscopic pack- liquid chromatography, J. Chromatogr. A 363 (1986) 1–30.
[6] L.R. Snyder, J.W. Dolan, Preparative high-performance liquid chromatography
ing experiments revealed that the local axial packing density and under gradient conditions III. Craig simulations for heavily overloaded
consequently the hydrodynamic packing stability vary consider- separations, J. Chromatogr. A 484 (1989) 437–450.
ably for different column packing strategies as well as during [7] H. Schmidt-Traub, Preparative Chromatography, Wiley-VCH, Weinheim,
2005.
cyclic column operation. Dynamic axial compression (DAC) led to [8] S. Golshan-Shirazi, G. Guiochon, Theory of optimization of the experimental
a significant increase in packing density near the top of the col- conditions of preparative elution chromatography: optimization of the
umn whereas flow packing resulted in increasing packing density column efficiency, Anal. Chem. 61 (1989) 1368–1382.
[9] S. Golshan-Shirazi, G. Guiochon, Optimization of the experimental conditions
towards the column outlet. Using coupled CFD-DEM modeling, the
in preparative liquid chromatography with touching bands, J. Chromatogr.
measured packing behavior was reproduced qualitatively provid- 517 (1990) 229–256.
ing valuable information regarding intrinsic packing properties that [10] A. Felinger, G. Guiochon, Comparison of maximum production rates and
up to now have been inaccessible. It was shown that interparticle optimum operating/design parameters in overloaded elution and
displacement chromatography, Biotechnol. Bioeng. A 43 (1993) 134–147.
friction as well as friction between particles and the column wall [11] T.M. Larson, J. Davis, H. Lam, J. Cacia, Use of process data to assess
determine how the compression force is transmitted through the chromatographic performance in production-scale protein purification
packing. While friction played a minor role during flow packing at columns, Biotechnol. Prog. 19 (2003) 485–492.
[12] G. Guiochon, The limits of the separation power of unidimensional column
high flow velocities, it became noticeable during DAC. Here, the liquid chromatography, J. Chromatogr. A 1126 (2006) 6–49.
wall support and particle rearrangement constrained axial com- [13] M.R. Ladisch, G.T. Tsao, Theory and practice of rapid liquid chromatography at
pression force transmission and force equilibration. The resulting moderate pressure using water as eluent, J. Chromatogr. 166 (1978) 85–100.
[14] M.R. Ladisch, Bioseparations Engineering, Wiley, New York, 2001.
inhomogeneous particle stress distribution negatively affected the [15] K.A.S. Jönsson, B.T.L. Jönsson, Fluid flow in compressible porous media I:
hydrodynamic packing stability. Repeated alternating application steady state conditions, AIChE J. 38 (1992) 1340–1348.
of flow packing and DAC (the hybrid packing method) was shown [16] K.A.S. Jönsson, B.T.L. Jönsson, Fluid flow in compressible porous media II:
dynamic behavior, AIChE J. 38 (1992) 1349–1356.
to enhance compression force transmission into the packed bed [17] K.C.E. Östergren, A.C. Trägårdh, Numerical study of two-dimensional
and to reduce particle stress during the column packing process. compaction, flow, and dispersion in chromatographic column, Numer. Heat
Hence, the axial packing density gradient was less pronounced Transfer Part A 32 (1997) 247–265.
[18] K.C.E. Östergren, A.C. Trägårdh, G.G. Enstad, J. Mosby, Deformation of a
and stayed nearly constant during cyclic hydrodynamic load in
chromatographic bed during steady state liquid flow, AIChE J. 44 (1998) 2–12.
comparison to the other packing methods investigated. Moreover, [19] J. Stickel, A. Fotopoulos, Pressure-flow relationships for packed beds of
the hybrid packing method resulted in a packing with the lowest compressible chromatography media at laboratory scale, Biotechnol. Prog. 17
integral porosity distribution (IPD) and hydrodynamic dispersion (2001) 744–751.
[20] R.N. Keener, L.E. Maneval, E.J. Fernandez, Toward a robust model of packing
(␴z 2 ). Analysis of the packed bed homogeneity in terms of radial and scale-up for chromatographic beds: 1. Mechanical compression,
flow and porosity distribution confirmed the measured results and Biotechnol. Prog. 20 (2004) 1146–1158.
demonstrated that the observed dynamics had a noticeable effect [21] R.N. Keener, L.E. Maneval, E.J. Fernandez, Toward a robust model of packing
and scale-up for chromatographic beds: 2. Flow packing, Biotechnol. Prog. 20
on column hydrodynamic stability and packing quality in terms of (2004) 1159–1168.
eddy dispersion. [22] R.N. Keener, E.J. Fernandez, L.E. Maneval, R.A. Hart, Advancement in the
As the present investigation required the use of a simplified modeling of pressure-flow for the guidance of development and scale-up of
commercial-scale biopharmaceutical chromatography, J. Chromatogr. A 1190
chromatographic system in absence of any salt or tracer substance, (2008) 127–140.
future work will extend towards the measurement and numerical [23] D. Hekmat, R. Mornhinweg, G. Bloch, Y. Sun, P. Jeanty, M. Neubert, D.
simulation of component diffusion as a contribution to the over- Weuster-Botz, Macroscopic investigation of the transient hydrodynamic
behavior of preparative packed chromatography beds, J. Chromatogr. A 1218
all dispersion in chromatographic packings. This will enable the (2010) 944–950.
evaluation of the column performance in terms of local HETP in [24] D. Hekmat, M. Kuhn, V. Meinhardt, D. Weuster-Botz, Modeling of transient
combination with the dynamics of compressible packings which flow through a viscoelastic preparative chromatography packing, Biotechnol.
Prog. 29 (2013) 958–967.
so far has not been investigated deterministically.
[25] M. Sarker, G. Guiochon, Consolidation of the packing material in
chromatographic columns under dynamic axial compression. III. Effect of the
nature of the packing solvent on the consolidation and performance of axial
Acknowledgements compression columns, J. Chromatogr. A 741 (1996) 165–173.
[26] B.G. Yew, E.C. Drumm, G. Guiochon, Mechanics of column beds: I. Acquisition
The authors thank Christoph Goniva, DCS Computing GmbH, of the relevant parameters, AIChE J. 49 (2003) 626–641.
[27] E. Müller, J.-T. Chung, Z. Zhang, A. Sprauer, Characterization of the mechanical
Linz, Austria, for fruitful discussions regarding the CFDEM soft- properties of polymeric chromatographic particles by micromanipulation, J.
ware. The kind provision of CM Sepharose 6FF and Blue Sepharose Chromatogr. A 1097 (2005) 116–123.
6FF resins by GE Healthcare Europe, Munich, Germany, and of [28] M.A. Shadday, A one-dimensional model of down-flow through a swelling
packed porous bed, Chem. Eng. Sci. 61 (2006) 2688–2700.
Affi-Gel 102 and Affi-Gel Blue Gel resins by Bio-Rad Laborato- [29] T. Farkas, J.Q. Chambers, G. Guiochon, Column efficiency and radial
ries GmbH, Munich, Germany, is acknowledged. The authors also homogeneity in liquid chromatography, J. Chromatogr. A 679 (1994) 231–245.
thank Eva Pichlmeier and Benjamin Münch, Institute of Biochemi- [30] T. Yun, G. Guiochon, Modeling of radial heterogeneity in chromatographic
columns - columns with cylindrical and symmetry and ideal model, J.
cal Engineering, Technical University of Munich, Germany, for their Chromatogr. A 672 (1994) 1–10.
valuable contribution. Funding by the German Research Foundation [31] T. Yun, G. Guiochon, Visualization of the heterogeneity of column beds, J.
(DFG project HE 2482/5-1/2) and the additional support of Martin Chromatogr. A 760 (1997) 17–24.
 M. Dorn et al. / J. Chromatogr. A 1516 (2017) 89–101 101

[32] J.H. Knox, J.F. Parcher, Effect of column to particle diameter ratio on the [53] K. Schmidt, S. Ripperger, S. Antonyuk, 3D DEM-CFD Simulation of the
dispersion of unsorbed solutes in chromatography, Anal. Chem. 41 (1969) formation of a fixed bed of activated carbon grains, air and toluene
1599–1606. permeation and toluene adsorption, Chem. Ing. Technol. 88 (3) (2016)
[33] J.H. Knox, G.R. Laird, P.A. Raven, Interaction of radial and axial dispersion in 307–312.
liquid chromatography in relation to the infinite diameter effect, J. [54] C. Goniva, C. Kloss, N.G. Deen, J.A.M. Kuipers, S. Pirker, Influence of rolling
Chromatogr. A 122 (1976) 129–145. friction on single spout fluidized bed simulation, Particuology 10 (2012)
[34] C.H. Eon, Comparison of broadening patterns in regular and radially 582–591.
compressed large-diameter columns, J. Chromatogr. A 149 (1978) 29–42. [55] C. Kloss, C. Goniva, A. Hager, S.S. Amberger. Pirker, Models, algorithms and
[35] J.E. Baur, E.W. Kristensen, R.M. Wightman, Radial dispersion from commercial validation for opensource DEM and CFD-DEM, Prog. Comput. Fluid Dyn/ 12
high-performance liquid chromatography columns investigated with (2012) 140–152.
microvoltammetric electrodes, Anal. Chem. 60 (1988) 2334–2338. [56] C. Schilde, C.F. Burmeister, A. Kwade, Measurement and simulation of
[36] G. Guiochon, T. Farkas, H. Guan-Sajonz, J.-H. Koh, M. Sarker, B.J. Stanley, T. micromechanical properties of nanostructured aggregates via
Yun, Consolidation of particle beds and packing of chromatographic columns, nanoindentation and DEM-simulation, Powder Technol. 259 (2014) 1–13.
J. Chromatogr. A 762 (1997) 83–88. [57] A. di Renzo, F.P. di Mayo, Comparison of contact force models for the
[37] G. Guiochon, M. Sarker, Consolidation of the packing material in simulation of collisions in DEM-based granular flow codes, Chem. Eng. Sci. 59
chromatographic columns under dynamic axial compression. I. Fundamental (2004) 525–541.
study, J. Chromatogr. A 704 (1995) 247–268. [58] T. Pöschel, T. Schwager, Computational Granular Dynamics, Springer, Berlin,
[38] J. Jin, H.A. Makse, A first-order phase transition defines the random close 2005.
packing of hard spheres, Physica A 389 (2010) 5362–5379. [59] R. Issa, Solution of the implicitly discretised fluid flow equations by operator
[39] R.A. Shalliker, V. Wong, B.S. Broyles, G. Guiochon, Visualization of bed splitting, J. Comput. Phys. 62 (1986) 40–65.
compression in an axial compression liquid chromatography column, J. [60] C.H. Rycroft, M.Z. Bazant, G.S. Grest, J.W. Landry, Dynamics of random
Chromatogr. A 977 (2002) 213–223. packings in granular flow, Phys. Rev. E 73 (2006) 1–8.
[40] R.A. Shalliker, B.S. Broyles, G. Guiochon, Physical evidence of two wall effects [61] M. Sakai, M. Abe, Y. Shigeto, S. Mizutani, H. Takahashi, A. Viré, J.R. Percival, J.
in liquid chromatography, J Chromatogr. A 888 (2000) 1–12. Xiang, C.C. Pain, Verification and validation of a coarse grain model of the
[41] S. Bruns, D. Stoeckel, B.M. Smarsly, U. Tallarek, Influence of particle properties DEM in a bubbling fluidized bed, Chem. Eng. J. 244 (2014) 33–43.
on the wall region in packed capillaries, J. Chromatogr. A 1268 (2012) 53–63. [62] D.S. Nasato, C. Goniva, S. Pirker, C. Kloss, Coarse graining for large scale DEM
[42] I. Bemberis, A. Noyes, V. Natarajan, Column packing for process-scale simulations of particle flow −an investigation of contact and cohesion
chromatography: guidelines for reproducibility, Biopharm Int. 2 (July) (2003) models, Procedia Eng. 102 (2015) 1484–1490.
23–30. [63] S. Bruns, J.P. Grinias, L.E. Blue, J.W. Jorgenson, U. Tallarek, Morphology and
[43] M. Sarker, A.M. Katti, G. Guiochon, Consolidation of the packing material in separation efficiency of low-aspect-ratio capillary ultrahigh liquid
chromatographic columns under dynamic axial compression. II. Consolidation chromatography columns, Anal. Chem. 84 (2012) 4496–4503.
and Breakage of several packing materials, J. Chromatogr. A 719 (1996) [64] S. Khirevich, A. Deneyko, A. Höltzel, A. Seidel-Morgenstern, U. Tallarek,
275–289. Statistical analysis of packed beds the origin of short-range disorder, and its
[44] M. Dorn, D. Hekmat, Simulation of the dynamic packing behavior of impact on eddy dispersion, J. Chromatogr. A 1217 (2010) 4713–4722.
preparative chromatography columns via discrete particle modeling, [65] A. Daneyko, D. Hlushkou, S. Khirevich, U. Tallarek, From random sphere
Biotechnol. Prog. 32 (2016) 363–371. packing to regular pillar arrays: analysis of transverse dispersion, J.
[45] B.J. Stanley, M. Sarker, G. Guiochon, Consolidation of the packing material in Chromatogr. A 1257 (2012) 98–115.
chromatographic columns under dynamic axial compression. IV. Mechanical [66] P. DePhillips, A.M. Lenhoff, Pore size distributions of cation-exchange
properties of some packing materials, J. Chromatogr. A 741 (1996) 175–184. adsorbents determined by inverse size-exclusion chromatography, J.
[46] R.N. Keener, L.E. Maneval, K.C.E. Östergren, E.J. Fernandez, Mechanical Chromatogr. A 883 (2000) 39–54.
deformation of compressible chromatographic columns, Biotechnol. Prog. 18 [67] D.E. Cherrak, M. Al-Bokari, E.C. Drumm, G. Guiochon, Behavior of packing
(2002) 587–596. materials in axially compressed chromatography columns, J. Chromatogr. A
[47]. P. Sabine, C. Plumpton, Statistics, Palgrave Mcmillan, UK, London, 1985. 943 (2002) 15–31.
[48] G.E. Healthcare, Ion Exchange Chromatography: Principles and Methods, [68] B.D. Bowes, H. Koku, K.J. Czymmek, A.M. Lenhoff, Protein adsorption and
Handbook, 2016 (Retrieved February 17, 2017 from) https://www. transport in dextran-modified ion-exchange media I. Adsorption, J.
gelifesciences.com/gehcls images/GELS/Related%20Content/Files/ Chromatogr. A 1216 (2009) 7774–7784.
1466001762444/litdoc11000421 20161014021056.pdf. [69] V. Wong, R.A. Shalliker, G. Guiochon, Evaluation of the uniformity of
[49] Z.Y. Zhou, S.B. Kuang, K.W. Chu, A.B. You, Discrete particle simulation of analytical-size chromatography columns prepared by the downward packing
particle-fluid flow: model formulation and their applicability, J. Fluid Mech. of particulate slurries, Anal. Chem. 76 (2004) 2601–2608.
661 (2010) 482–510. [70] H.A. Makse, D.L. Johnson, L.M. Schwartz, Packing of compressible granular
[50] Lennart Fries, Sergiy Antonyuk, Stefan Heinrich, Daniel Dopfer, Stefan Palzer, materials, Phys. Rev. Lett. 48 (2000) 4160–4163.
Collision dynamics in fluidised bed granulators: a DEM-CFD study, Chem. Eng. [71] Q.S. Yuan, A. Rosenfeld, T.W. Root, D.J. Klingenberg, E.N. Lightfoot, Flow
Sci. 86 (2013) 108–123. distribution of chromatographic columns, J. Chromatogr. A 831 (1999)
[51] Vitalij Salikov, Sergiy Antonyuk, Stefan Heinrich, Vinayak S. Sutkar, Niels G. 149–165.
Deen, J.A.M. Kuipers, Characterization and CFD-DEM modelling of a prismatic [72] A. Williams, K. Taylor, K. Dambuleff, O. Persson, R.M. Kennedy, Maintenance
spouted bed, Powder Technol. 270 (2015) 622–636. of column performance at scale, J. Chromatogr. A 944 (2002) 69–75.
[52] Vinayak S. Sutkar, Niels G. Deen, Amit V. Patil, Vitalij Salikov, Sergiy [73] M.A. Teeters, I. Quiñones-García, Evaluating and monitoring the packing
Antonyuk, Stefan Heinrich, J.A.M. Kuipers, CFD-DEM model for coupled heat behavior of process scale chromatography columns, J. Chromatogr. A 1069
and mass transfer in a spout fluidized bed with liquid injection, Chem. Eng. J. (2005) 53–64.
288 (2016) 185–197.