Middle East Technical University

Chemical Engineering Department

ChE 352 – Math. Modeling in ChE (1)
Instructor: Dr. Necip B. Üner
Assistant: Uğur Sökmen
Date: 26/12/2023

Active session - IV
Regeneration of silica gel beads
A study on the solution of PDEs by separation of variables

Silica gels are commonly used as desiccants to remove moisture. They constitute a major
portion of the materials used in chromatographic separations as well. They are often made in the
form of spherical beads (Figure 1a) and usually with a porous structure (Figure 1b and 1c) .

Figure 1. Silica gel beads. (a) A desiccant bag and the silica beads inside it. (b) Scanning electron
microscope image of a silica bead with a diameter of ~2 mm. (c) Higher magnification image that shows
the porous structure of a bead [1].

Silica gel beads change their color when they absorb moisture, so they are visible indicators
of the moisture they have absorbed (Figure 2). Volumetric absorption of water proceeds with water
rapidly adsorbing on the surface and then diffusing on the surface into the porous structure. The
overall transport process can be described approximately by Fickian diffusion with an effective
diffusivity.
Absorption of moisture in silica gel is easily reversible. By heating the silica gel up to
approximately 140 °C, almost all of the water can be removed from the gel. In this active session,
we will focus on the regeneration of saturated silica gel beads. We will consider the case where
the beads are put on a tray and placed in a convection oven (Figure 3). We will simplify the
geometry and model a single sphere that dries uniformly over its entire surface. The goal is to
elucidate the role of convective mass transfer on drying rate.

1
Figure 2. Moisture content in silica gel beads. Check the file on ODTÜClass to see the colors.

Figure 3. Regeneration. Realistic case vs the model problem.

a) The differential equation and the boundary conditions. Develop a shell balance and
derive a parabolic PDE that describes unsteady-state mass transfer in an initially
saturated silica bead. Assume thermal equilibrium inside. D is the effective diffusivity
of water in the porous structure. At the outer surface of the bead, there is convective
drying. Assume that the concentration of water vapor in the surrounding atmosphere is
approximately zero. How would you write the boundary conditions? Are the PDE and
BCs linear and homogeneous?

b) Normalization. Make the PDE and boundary conditions dimensionless by normalizing

your variables. How do you scale time? Use θ = C/C* for the dimensionless
concentration and Bim = kmR/D to make the convective boundary condition
dimensionless. km is the convective mass transfer coefficient and C* is the saturation
moisture content.

c) Separation of variables. Apply separation of variables and develop the general

equations for G(ξ) and F(τ). What are the boundary conditions for the ODE for G? How
can you solve for G? See Spiegel 24.23 and 24.24 to write the general solution in terms
of trigonometric functions.
2
 d) Eigenvalues and eigenfunctions: Evaluate the boundary conditions for G and derive
the following relation that yields the eigenfunctions1. What are the eigenfunctions?
n
tan ( n ) =
1 − Bim

e) Eigenvalues for large and small Bim (Optional). Knowing that λ1 is smaller than but
very close to π, show that when Bim is large, the above eigenvalue equation reduces to2:
n
n =
1+ 1
Bim
Similarly, show that for small Bim, the first eigenvalue is:
1  3Bim

f) Constants An and the final solution: Apply orthogonality criterion to the initial value.
Derive the following relation for An (see the Appendix), to complete your solution:
4 Bim sin ( n )
An =
n  2n − sin ( 2n ) 

g) Average dimensionless moisture content. Convert the averaging integral into

dimensionless form and provide a relation for <θ>.

h) Strong convective drying. Consider the case where convection is very strong (km is
very large). This will lead to the boundary condition C = 0 at r = R. Starting from the
general solution for G, re-solve the problem. Derive the expressions for λn, An, and
finally for θ.

i) Average dimensionless moisture content - Strong convective drying. Similar to

what you did in part (g), find the average dimensionless concentration for the strong
convective drying case.

j) Lumped analysis. By using a macroscopic balance, derive a lumped drying model for
<θ> valid for small Bim.

k) Comparison of results and the effect of Bim on BCs and lumping. Using the Excel
sheet on ODTÜClass3 that contains λn and An for various Bim in the general solution of
part (f), plot <θ> with respect to the logarithm of dimensionless time, log(τ), for the

1
See the first MATLAB example for a root-solving algorithm. However, there is no need for you to do the computations
for this problem. A table of the eigenfunctions are provided. See part (k) and its footnote.
2
Give a look at how tan(x) looks like. Note that the following properties apply:
tan ( x ) = tan ( x +  ) ; tan ( x ) = tan ( − x ) ; tan ( x )  x
The latter comes from the Maclaurin series expansion, and it is valid for small x.
3
Active session 4 - lambda_n & A_n.xlsx
3
 given Bim numbers. Add the plot of the expression in part (i) and the plot of the
expression in part (j) for Bim = 2 and 0.1. Considering the fact that the solution in part
(g) is general for any drying rate, approximately at which Bim number does the solution
in part (i) work? Similarly, when does the solution in part (j) work?

l) Calculation of drying time. At ~140 °C, the effective diffusivity of water in silica gel
beads is roughly 1×10-8 m2s-1 [2]. For a 3 mm-diameter bead, calculate the minimum
time it would take for an initially saturated bead to lose 95% of its moisture content4.

m) Effective diffusivity. The value of D depends on structure. For example, in part (l), the
value of D was given for beads with 2 nm average pore size and 500 m2/g surface area.
See the beads Figure 4. Which one would have a higher D?

Figure 4. Electron microscope image of silica beads with different structures [1].

References
[1] I. Novák and D. Berek, “Structural inhomogeneities in wide-pore silica gels,” Journal of
Chromatography A, vol. 665, no. 1, pp. 33–36, Apr. 1994, doi: 10.1016/0021-9673(94)87028-4.
[2] C.-C. Ni and J.-Y. San, “Measurement of apparent solid-side mass diffusivity of a water vapor–silica
gel system,” International Journal of Heat and Mass Transfer, vol. 45, no. 9, pp. 1839–1847, Apr.
2002, doi: 10.1016/S0017-9310(01)00291-5.

Appendix: Useful Integrals5

1
sin ( n ) − n cos ( n ) 1
2n − sin ( 2n )
  sin ( n ) d =  sin (   ) d =
2

n 4n
2 n
0 0

4
If you use the expression you derived in part (j) you will see that this time would be approximately 40 times longer
when Bim = 0.01.
5
The integral sheet previously uploaded to ODTÜClass will also be useful in part (h).

4
s dc
=- Rm -B!r
c
-

S
& (T,5) =
F(5)G(T)
L CENG-
25
alt)
G(t))
GaFT
/Th a (F(T)

Fe
F(5)
-
=> now d!v!de
e
aT
dT 52 F(5) G(T)

&
·
1

G(5) 52 aT
1st equat!on 2nd equat!on

constant
Y
x2 F(5) DIFIT)
2) -

= => by !ntegrat!on Coexp(-x25) =

F (5)
aT

1) alt2@PT) &
+ 1272G(5) = 0 Bessel
at
D S

= 1

2 j 2 a x2 =
=
12
p = = =

=
1/2

where = =
x

[C Jua (x 5)
!
e = .
+ 25- 2(X 5)] .

( C (x)]
1
G =
s!n(x) +

YE

f(as!n(xt)
P = +...