Journal of Chromatography A, 1490 (2017) 133–137

Contents lists available at ScienceDirect

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

Empirical correlations for axial dispersion coefficient and Peclet

number in fixed-bed columns
Seyed Omid Rastegar a , Tingyue Gu b,∗
a
Department of Engineering, Chemical Engineering Group, University of Kurdistan, Sanandaj, Iran
b
Department of Chemical and Biomolecular Engineering, Ohio University, Athens, OH 45701, USA

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

Article history: In this work, a new correlation for the axial dispersion coefficient was obtained using experimental data
Received 3 November 2016 in the literature for axial dispersion in fixed-bed columns packed with particles. The Chung and Wen
Received in revised form 7 February 2017 correlation, the De Ligny correlation are two popular empirical correlations. However, the former lacks
Accepted 13 February 2017
the molecular diffusion term and the latter does not consider bed voidage. The new axial dispersion coef-
Available online 14 February 2017
ficient correlation in this work was based on additional experimental data in the literature by considering
both molecular diffusion and bed voidage. It is more comprehensive and accurate. The Peclet number
Keywords:
correlation from the new axial dispersion coefficient correlation on the average leads to 12% lower Peclet
Axial dispersion coefficient
Peclet number
number values compared to the values from the Chung and Wen correlation, and in many cases much
Empirical correlation smaller than those from the De Ligny correlation.
Mass transfer © 2017 Elsevier B.V. All rights reserved.
Liquid chromatography
Adsorption

1. Introduction
Axial dispersion in a fixed-bed packed with particles is impor-
tant in mass transfer in many applications in areas such as
adsorption, liquid chromatography and hydrology [1]. Axial dis-
persion in a fixed-bed column is caused by molecular diffusion and
turbulent mixing around particles [2,3]. A model for dispersion pro-
cesses in porous media is usually based on convective-diffusion
and the mean solute concentration in the bulk-fluid phase (Cb ), the
mean interstitial velocity of fluid (), the source term (Sb ), the axial
dispersion coefficient (Db ) and time (t) [2,4]. Peclet (Pe) number
for nondimensionalization in analytical and numerical solutions is
sometimes used in axial dispersion models. It reflects the char-
acteristic ratio of convective transport rate to diffusive transport
rate. The Pe term is sometimes dropped for simplicity. However,
this practice often leads to abnormally sharp or shockwave pro-
files [5]. In numerical solutions, it is beneficial to retain the Pe term
for solution stability. In liquid chromatography simulation, some
researchers [3,5,6] have noticed that the PeL (Peclet number based
on the bed length) values estimated from existing empirical corre-
lations tend to be very large and yield much sharper concentration
profiles than experimental profiles. Gu and Zheng [7] artificially
used PeL = 1000 for simulation to fit experimental data because the
estimated PeL was too large and caused numerical difficulties. Thus,
it is desirable to revisit the estimation of PeL .
The popular Chung and Wen correlation [8] was first published
in 1968 by correlating large amounts of experimental data. It did
not consider the molecular diffusion because it suggested that Db
was usually one or two orders of magnitude higher than molec-
ular diffusivity [3]. In 1970, De Ligny [1] incorporated molecular
diffusion that comes from Langer et al. [9] based on the turbulent
mixing term from Giddings’ random-walk analysis. However, his
correlation used a fixed bed voidage of εb = 0.4, which limits its
application. Other researchers [10–12] published additional axial
dispersion coefficient (Db ) correlations using experimental data.
However, these new correlations did not gain popularity due to var-
ious reasons such as increased complexity [13] or no improvement
in accuracy.
It is desirable to obtain an updated axial dispersion coefficient
correlation (Db ) that considers both molecular diffusion coefficient
(Dm ) and bed voidage (εb ) based on the large amount of experimen-
tal data used by Chung and Wen and additional published data in
the literature. A new correlation based on more experimental data
can improve accuracy in the estimation of PeL .

∗ Corresponding author.
E-mail address: gu@ohio.edu (T. Gu).

http://dx.doi.org/10.1016/j.chroma.2017.02.026
0021-9673/© 2017 Elsevier B.V. All rights reserved.
134 S.O. Rastegar, T. Gu / J. Chromatogr. A 1490 (2017) 133–137

2. Mathematical correlation 1
Chung and Wen
Miller and King
The primary mechanisms that contribute to axial dispersion in Ebach and White
a fixed bed are molecular diffusion and turbulent mixing due to Carins and Prausnitz
merging and splitting of flows around the particles [14]. It is now Hennico, Jacques and vermeuien
generally agreed that the effects of diffusion and convection are Liles and Geankoplis
“coupled” in the axial dispersion of matter by fluid flow through Kubo et al.

εbPep*
fixed beds [15]. Inglezakis
Eq. (1) represents the classical correlation that combines the This work (Eq. 13)
contribution of eddy diffusion and molecular diffusion using the
coupling theory in the absence of mass transfer between a mobile
phase and stationary phase [14,15],
1 D  Dm
= b = + (1)
Pep dp  1 + CDm /(dp ) (dp ) 0.1
0.001 0.01 0.1 1 10 100 1000 10000
where ,  and C are dimensionless coefficients, depending on the Re

geometry of the column packing and the flow rate, Pep is the Pe
Fig. 1. Correlating experimental data to obtain Eq. (12).
number based on the particle diameter (vdp /Db ), dp the particle
diameter, v the interstitial velocity [9,16]. Eq. (1) can be rearranged
to yield Eq. (2), Thus, the new Db correlation in this work was obtained using Eq.
dp  (10) inspired by Eq. (6),
Db = Dm + (2)
1 + CDm /(dp )
Db = 0.7Dm + Db∗ (10)
Eq. (2) can be nondimensionalized using dimensionless Pep ,
Reynolds number defined by (Re = 2Rp vεb /) using the superficial Db ∗ is shown as the second term on the right hand side of Eq.
velocity and Schmidt number defined by Sc = /(Dm ) as shown in (11) based on Eq. (6),
Eq. (3). Note that in mass transfer correlations, Re is often defined
2Rp vεb
using superficial velocity (εb ) instead of interstitial velocity (). Db = 0.7Dm + (11)
x1 + x2 Rex3
1 ε  
= b + εb C
(3) in which the Db ∗ term adopted Eq. (4) form. The correlation param-
Pep ReSc 1 + ReSc
eters x1 , x2 , x3 were obtained through experimental data fitting
Based on Eq. (2), one of the popular correlations is the Chung and using MATLAB (Version R2014a).
Wen empirical correlation after dropping the molecular diffusion
term by considering only eddy diffusion [8]:
3. Results and discussions
2Rp vεb  −3 3

Db = 10 < Re < 10 (4)
0.2 + 0.011Re0.48 3.1. Peclet number correlation

where Rp is particle radius. Using the Pep number, Eq. (4) converts
Eq. (12) was obtained from fitting experimental data sets in
to Eq. (5) as follows,
the literature using MATLAB (R2014a). This is a new empirical
εb Pep = 0.2 + 0.011Re0.48 (5) relationship between Peclet number and Reynolds number, which
considers eddy diffusion while ignoring molecular diffusion.
Note that Pep here uses the Db in Eq. (4) that considers only eddy
diffusion while ignoring molecular diffusion (i.e., Dm = 0). εb Pep∗ = 0.18 + 0.008Re0.59 (12)
Based on Eq. (2), De Ligny [15] obtained the following empiri-
cal correlation for Db for liquid flow in packed beds with spherical It is customary in chemical engineering literature to present dis-
particles, persion data as graphs of the logarithm of the Peclet number vs.
the logarithm of the Reynolds number. Fig. 1 shows the double-log
5Rp v
Db = 0.7Dm + (6) plot of experimental data [17–23] and Eq. (12). In the creeping flow
1 + 4.4Dm /(Rp v) for obtaining a conservative upper estimate for the convective dis-
Using Pep , Eq. (6) yields Eq. (7), persion coefficient, the Miller and King [17] data and also Ebach
and White [18] data including experiment with rings, Berl saddles,
0.7Dm 1 Intalox saddles were used. Additionally, there were some reports
1/Pep = + (7)
2Rp v 0.4 + 1.76Dm /(Rp v) for the creeping flow regime in the literature but some of them
The molecular diffusion coefficient (Dm ) in Eq. (7) can be esti- were deemed not adoptable because they used narrow columns or
mated based on the solute’s molecular weight (MW ) using the had large errors in the measured dispersion coefficients [10].
following correlation from Polson [14]: Fig. 2 shows the Chung and Wen correlation [8] and its compar-
  ison with the same experimental data in Fig. 1. Figs. 1 and 2 show
Dm m2 · s−1 = 2.74 × 10−9 (MW )−1/3 (MW > 1000) (8) that the new correlation fitted the experimental data slightly better
than the Chung and Wen correlation in both lower and higher Re
In this work, the following correlation form was first used to regions.
update the Chung and Wen correlation for the fix bed condition, It should be noted that Chung and Wen correlation were based
εb Pep∗ = x1 + x2 Rex3 (9) on both fixed bed and fluidized bed data in the literature, while the
new empirical correlation in this work aimed at finding a better cor-
where Pep ∗ = vdp /Db ∗ and Db ∗ is the axial dispersion coefficient relation for fixed-bed columns. The new correlation is particularly
that considers only eddy diffusion and ignores molecular diffusion. suited for liquid chromatography modeling.
 S.O. Rastegar, T. Gu / J. Chromatogr. A 1490 (2017) 133–137 135

Table 1
The correlations for the estimation of axial dispersion coefficient in fixed-bed columns.

Authors Correlation Ref.

Hiby et al. (1962) 1/Pep = √0.65 + 0.67εb /(ReSc) [24]

1+7 εb /(ReSc)
Chung and Wen εb Pep = 0.2 + 0.008Re0.48 [8]
(1968) Or
0.48
PeL = 2RL p 0.2+0.011Re
ε b
0.7Dm 1
De Ligny 1/Pep = 2Rp v
+ 0.4+1.76Dm /(Rp v)
[1]
(1970)  2

Koch and Brady (1985) 1/Pep = εb 43 + 6 (1 − εb ) ln(ReSc) + ReSc
1
[12]
 ReSc −0.262
Han et al. (1985) Pe = 4.36 [25]
1−εb
  4˛21 (1 − εb )

ReSc 2 Re2 Sc 2 3 εb
1/Pep = (1 − p) + p(1 − p) exp − −1 +
Gunn (1987) 4˛21 (1 − εb ) 16˛41 (1 − εb )
2 p(1 − p)ReSc 1.4ReSc [11]
24
p = 0.17 + 0.33 exp(− ), ˛1 = 2.405
 1−ε 1/6 Re
Athalye et al. (1992) Pep = ReScb [10]
This work 1/Pep =
0.7Dm εb
2Rp v
+
0.18+0.008Re0.59
0.7Dm
Or 1/PeL = Lv
+
2Rp εb
L 0.18+0.008Re0.59

1 If the Dm term in Eq. (16) is negligible, Eq. (16) reduces to the

Chung and Wen following equation,
Miller and King
Ebach and White L 0.18 + 0.008Re0.59
PeL = (17)
Carins and Prausnitz 2Rp εb
Hennico, Jacques and vermeuien
Liles and Geankoplis In liquid chromatography involving proteins with large molec-
ular weight (MW ) values, the molecular diffusivity (Dm ) value
ε bPe p*

Kubo et al.
Inglezakis calculated from Eq. (8) is very small. Thus it can be dropped from
Chung and Wen correlation Eq. (13). Furthermore, the particle diameter based Re value is usu-
ally less than 1 in liquid chromatography [2,4]. By dropping both
Dm and Re the new Db correlation in Eq. (13) can be simplified to
Eq. (17),

Rp vεb
Db = (Re < 1) (18)
0.1 0.09
0.001 0.01 0.1 1 10 100 1000 10000
Re This simplifies the new PeL correlation in Eq. (17) to Eq. (19),

Fig. 2. Chung and Wen correlation vs. experimental data. 0.09L

PeL = (Re < 1) (19)
Rp εb
Eq. (12) can be modified to include molecular diffusion (Dm ) as
The PeL values calculated from Eq. (19) are 10% smaller than what
follows based on the coupling theory and using value of 0.7 for 
Chung and Wen correlation when its Re term is negligible.
in Eq. (2),

2Rp vεb
Db = 0.7Dm + (13) 3.2. Effects of mass transfer parameters on PeL number
0.18 + 0.008Re0.59
This new Db correlation, which considers both molecular diffu- Finding the sensitivities of the parameters in the new PeL corre-
sion and eddy diffusion, leads to Eq. (14) by using Pep , lation in Eq. (16) is useful because it can tell which parameters are
relatively more important. Table 2 shows the effect of the different
0.7Dm εb parameters on the axial dispersion coefficient (Db ) and the Peclet
1/Pep = + (14)
2Rp v 0.18 + 0.008Re0.59 number based on the bed length (PeL ) obtained from the Chung
and Wen correlation, the De Ligny correlation and the new corre-
The particle diameter based Peclet number Pep is frequently
lation. Note that in liquid chromatography modeling, Re is usually
used in axial dispersion coefficient correlations. However, liquid
less than 1.
chromatography modeling needs the bed length-based Peclet (PeL ).
Pep can be converted to PeL using follow relationship,
3.2.1. Effects of Re number
L
PeL = vL/Db = Pep (15) The effects of Re number on the Db and PeL values were shown in
2Rp Table 2. It was investigated in Re values from 0.01–100 by varying
where L is column length (m). This leads to a new PeL correla- interstitial velocity when the other parameters were kept constant.
tion that considers both eddy diffusion and molecular diffusion as Although, the 1 < ReSc < 200 range is typically of interest in liquid
shown in Eq. (16), chromatographic practice [10], but some researchers considered
both laminar and turbulent regimes. Table 2 shows that the new
0.7Dm 2Rp εb correlation yields on the average 12% lower than PeL than the Chung
1/PeL = + (16)
Lv L 0.18 + 0.008Re0.59 and Wen correlation.
136 S.O. Rastegar, T. Gu / J. Chromatogr. A 1490 (2017) 133–137

3.2.2. Effect of bed voidage
Bed voidage (εb ) in the axial dispersion correlation is a key
parameter. In liquid chromatography modeling, it influences the
mass balances in breakthrough curves and retention times of eluted
peaks considerably. Bed voidage depends on the particle size, shape
and how they are packed. Its effect was investigated from 0.35 to 0.6
when the other parameters were kept constant as shown in Table 2.
The Db and PeL values obtained from the De Ligny correlation do not
change with bed voidage because εb is not an input parameter. This
inflexibility in the De Ligny correlation causes some large errors as
shown in Table 2.
3.2.3. Effect of molecular diffusion (Dm )
Molecular diffusion can play an important role on the axial
dispersion coefficient and thus the PeL number under certain con-
ditions. According to Eq. (8), increasing protein molecular weights
(MW ) results in decreasing molecular diffusion coefficient (Dm ). The
molecular diffusivity values for biological molecules fall between
10−8 m2 s−1 to 10−11 m2 s−1 [2,14] that is covered in Table 2. For
large Dm values, the molecular diffusion term can be significant
and even dominating for the axial dispersion coefficient (Db ). The
Chung and Wen correlation ignores Dm variation while the new
correlation responds to its variation in Table 2. A larger Dm led to
an increased Db value, and a decreased PeL value for the new cor-
relation. Furthermore, results in Table 2 show the PeL values for
different Dm values obtained from the new correlation are all lower
than those from the other correlations.
3.2.4. Effect of bed height (L)
According to Eq. (15), an increasing bed height increases the
Peclet number based on the bed length (PeL ), proportionally. In
Table 2, this trend exists as expected in the three correlations com-
pared in Table 2. Increasing L leads increased PeL that gives sharper
peaks in liquid chromatography modeling. In liquid chromatogra-
phy modeling, bed height is directly related to the plate height in
the column [10].
3.2.5. Effect of particle radius (Rp )
The effects of particle radius on Db and PeL were studied in the
particle radius range of 10–100 ␮m in Table 3. The new correlation
is suitable for liquid chromatography modeling with small particles
(e.g., 10 ␮m or 20 ␮m in radius) and large particles (“big beads”)
that have diameters reaching 100 ␮m. In all the cases in Table 3,
PeL values from the new correlation are lower than those from the
De Ligny and the Chung and Wen correlations. This is favorable in
liquid chromatography modeling because a less stiff PeL leads to
better numerical stability.
3.3. Graphical comparison of different Pep correlations
The comparison (Re ≤ 1) of the new Pep correlation in Eq. (14),
the De Ligny correlation in Eq. (7), the Chung and Wen correlation
in Eq. (5) and the Gunn correlation [11] in Table 1 with two sets of
experimental data (Sc = 730) from Miller and King [17] and Ebach
and White [18] is exhibited in Fig. 3. Fig. 3 shows that the new
correlation is much closer to experimental data than the De Ligny
correlation when Re < 0.1. In Fig. 3, the Pep values from the new
correlation are roughly 10% smaller than those from the Chung and
Wen correlation. Fig. 3 also shows that the Gunn correlation departs
from the other correlations when Re is significantly less than 1.

Table 2
Effects of different parameters on the Db and PeL values (Rp = 20 ␮m, / = 1 × 10−6 m2 s−1 ).

Re εb Dm L(m) Db (m2 s−1 ) PeL

(m2 s−1 )

De Ligny Chung and Wen This work De Ligny Chung and Wen This work Eq. (16)
Eq. (13)

0.01 0.45 1E-09 0.15 4.050E-08 4.970E-08 5.613E-08 2057 1676 1486
0.1 5.351E-07 4.911E-07 5.499E-07 1557 1697 1515
1 5.534E-06 4.739E-06 5.319E-06 1505 1758 1566
10 5.553E-05 4.288E-05 4.736E-05 1500 1943 1759
100 5.555E-04 3.330E-04 3.321E-04 1500 2502 2509
0.01 0.35 1E-09 0.15 4.050E-08 3.869E-08 4.380E-08 2057 2154 1903
0.4 4.050E-08 4.420E-08 4.995E-08 2057 1885 1668
0.5 4.050E-08 5.521E-08 6.224E-08 2057 1509 1339
0.6 4.050E-08 6.621E-08 7.452E-08 2057 1258 1118
0.01 0.45 1E-08 0.15 1.820E-08 4.970E-08 6.239E-08 4578 1676 1336
1E-09 4.050E-08 4.970E-08 5.613E-08 2057 1676 1486
1E-10 5.351E-08 4.970E-08 5.546E-08 1557 1676 1503
1E-11 5.535E-08 4.970E-08 5.540E-08 1505 1676 1504
0.01 0.45 1E-09 0.1 4.050E-08 4.970E-08 5.613E-08 1371 1117 990
0.15 4.050E-08 4.970E-08 5.613E-08 2057 1676 1486
0.2 4.050E-08 4.970E-08 5.613E-08 2743 2235 1981
0.25 4.050E-08 4.970E-08 5.613E-08 3429 2794 2476
0.3 4.050E-08 4.970E-08 5.613E-08 3353 4115 2971

Table 3
Effects of different particle sizes on the Db and PeL values (L = 0.15 m, εb = 0.45, / = 1 × 10−6 m2 s−1 , Dm = 1 × 10−9 m2 s−1 ).

Rp Re Db (m2 s−1 ) PeL

(␮m)

De Ligny Chung and Wen This work Eq. (13) De Ligny Chung and Wen This work
Eq. (16)

10 0.005 1.620E-08 2.489E-08 2.842E-08 5143 3347 2932

20 0.01 4.050E-08 4.970E-08 5.613E-08 2057 1676 1486
50 0.025 1.206E-07 1.239E-07 1.389E-07 691 673 600
100 0.05 2.581E-07 2.468E-07 2.764E-07 323 338 302
 S.O. Rastegar, T. Gu / J. Chromatogr. A 1490 (2017) 133–137
Fig. 3. Comparison between different correlations and experimental data.
4. Conclusion
This work presented a new empirical correlation for the axial
dispersion coefficient in fixed beds that incorporates both molecu-
lar diffusion and bed voidage. It used more experimental data than
the popular Chung and Wen correlation. It improved the Chung
and Wen correlation by introducing molecular diffusion and the De
Ligny correlation by introducing bed voidage. The new correlation
fitted the experimental data well. It yields PeL values significantly
smaller (on the average 12% smaller) than those from the Chung
and Wen correlation, and in many cases much smaller than those
from the De Ligny correlation for different bed voidage, particle
diameter, molecular diffusion coefficient, bed height and Re num-
ber values. The lower PeL leads to better numerical stability in liquid
chromatography modeling.
References
[1] C.L. De Ligny, Coupling between diffision and convection in radial dispersion
of matter by fluid flow through packed beds, Chem. Eng. Sci. 25 (1970)
1177–1181.
[2] J.M.P.Q. Delgado, A critical review of dispersion in packed beds, Heat Mass
Transf. 42 (2006) 279–310.
[3] T. Gu, Mathematical Modeling and Scale-Up of Liquid Chromatography: with
Application Examples, second ed., Springer, 2015.
137
[4] J.M.P.Q. Delgado, Longitudinal and transverse dispersion in porous media,
IchemE 85 (2007) 1245–1252.
[5] S. Ghosal, Z. Chen, Nonlinear waves in capillary electrophoresis, Bull. Math.
Biol. 72 (2010) 2047–2066.
[6] Z. Li, Y. Gu, T. Gu, Mathematical modeling and scale-up of size-exclusion
chromatography, Biochem. Eng. J. 2 (1998) 145–155.
[7] T. Gu, Y. Zheng, A study of the scale-up of reversed-phase liquid
chromatography, Sep. Purif. Technol. 15 (1999) 41–58.
[8] S.F. Chung, C.Y. Wen, Longitudinal dispersion of liquid flowing through fixed
and fluidized beds, AIChE J. 14 (1968) 857–866.
[9] G. Langer, A. Roethe, K.P. Roethe, D. Gelbin, Heat and mass transfer in packed
beds-III, Axial mass dispersion, Int. J. Heat. Mass. Transf. 21 (1978) 751–759.
[10] A.M. Athalye, S.J. Gibbs, E.N. Lightfoot, Predictability of chromatographic
protein separations. Study of size-exclusion media with narrow particle size
distributions, J. Chromatogr. A. 589 (1992) 71–85.
[11] D.J. Gunn, Axial and radial dispersion in fixed beds, Chem. Eng. Sci. 42 (1987)
363–373.
[12] D.L. Koch, J.F. Brady, Dispersion in fixed Beds,J, Fluid. Mech. 154 (1985)
399–427.
[13] R. Pfeffer, J. Happel, An analytical study of heat and mass transfer in
multiparticle systems at low Reynolds numbers, AIChE J. 10 (1964) 605–611.
[14] T. Gu, G. Iyer, K.S.C. Cheng, Parameter estimation and rate model simulation
of partial breakthrough of bovine serum albumin on a column packed with
large Q Sepharose anion-exchange particles, Sep. Purif. Technol 116 (2013)
319–326.
[15] C.L. De Ligny, The contribution of eddy diffusion and of the macroscopic
mobile phase velocity profile to plate height in chromatography, J.
Chromatogr. A. 49 (1970) 393–401.
[16] J.C. Giddings, Eddy diffusion in chromatography, Nature 184 (1959) 357–358.
[17] S.F. Miller, C.J. King, Axial dispersion in liquid flow through packed beds,
AIChE J. 12 (1966) 767–773.
[18] E.A. Ebach, R.R. White, Mixing of fluids flowing through beds of packed solids,
AIChE J. 4 (1958) 161–169.
[19] E.J. Cairns, J.M. Prausnitz, Longitudinal mixing in packed beds, Chem. Eng. Sci.
12 (1960) 20–34.
[20] A. Hennico, T., Vermeulen, G. Jacques, Longitudinal dispersion in packed
extraction columns, UCRL10696, 1963.
[21] A.W. Liles, C.J. Geankpolis, Axial diffusion of liquids in packed beds and end
effects, AIChE J. 6 (1960) 591–595.
[22] K. Kubo, T. Aratani, A. Mishima, T. Yano, Photographic observation of flow
pattern in voids of packed-bed of spheres, J. Chem. Eng. Jpn. 11 (1978)
405–407.
[23] V. Inglezakis, Design of Ion Exchange Column Using Natural Minerals as
Packing Materials, PhD Thesis, National Technical University of Athens,
Athens, Greece, 2002.
[24] J.W. Hiby, Longitudinal and transverse mixing during single phase flow
through granular beds, Proc. Symp. London, 1962.
[25] N.W. Han, J. Bhakta, R.G. Carbonell, Longitudinal and lateral dispersion in
packed beds: effect of column length andparticle size distribution, AIChE J. 31
(1985) 277–288.