Based on J. Crank, The Mathematics of Diffusion, 2nd ed., Clarendon Press, Oxford, 1975, Section 5.3.3 (eqs. 5.33-5.35, Table 5.1, Fig. 5.7)
Physical problem (Crank, §5.3.3):​A cylinder of radius a (occupying r < a) is immersed in a well-stirred bath of limited volume, whose cross-section is A (excluding the cylinder). The solute concentration in the bath is always uniform and equals C0 initially; the cylinder is initially free of solute. The total amount of solute
M
t
taken up by the cylinder after time t, as a fraction of the amount
M
∞
after infinite time, is (Wilson 1948)​​
​​and α = A/(π a^2) is the ratio of the volumes of solution and cylinder (α = A/(π a^2 K) if a partition coefficient K applies). α is tied to the final fractional uptake of solute by the cylinder through​​
​​Crank tabulates the first roots
q
n
for several values of α (Table 5.1) to help evaluate (5.33), and Fig. 5.7 shows
M
t
/
M
∞
against
Dt/
2
a
for final uptakes of 0, 30, 50, 70 and 90 percent. The series converges slowly for small
Dt/
2
a
, for which Crank gives alternative small-time expressions (5.36)-(5.41). The concentration inside the cylinder is given by (5.42), and the same equations describe desorption into a well-stirred solution when (5.35) is replaced by (5.43).
This notebook computes the roots
q
n
of (5.34) as a function of a single parameter
m
, where
m=α/(1+α)=1-
M
∞
/(AC0)
is the fraction of solute left in the bath at equilibrium (
m=0
means α = 0, i.e. 100% uptake;
m=1
means α = Infinity, i.e. no uptake). Dividing (5.34) by (1 + α) gives the equivalent form used in the first cell (eq1):
​
m q J0(q) + 2 (1 - m) J1(q) = 0
​
Each root branch
q
n
(m)
is a curve in the (m, q) plane. Rather than solving for
q
at each
m
separately, the notebook traces the whole curve parametrically: eq2 imposes arc-length parametrisation, m'(s)^2 + q'(s)^2 = 1, and NDSolve (function sol[k]) integrates the pair {m[s], q[s]} from the starting point m = 0, q = BesselJZero[1, k] (where J1(q) = 0, the α = 0 limit), for k = 1, 2, 3, ... This avoids the difficulty of root finding near turning points and gives all branches q_1, q_2, ... as smooth interpolating functions.
​
The list mus = {1, 9/10, ..., 1/10, 0} contains the values of m corresponding to the final uptakes 0%, 10%, ..., 100%. These are converted to the α values via α = m/(1 - m) (the first one being Infinity), and the interpolating solutions are then evaluated at each m to produce a table of α against q_1 ... q_6, in the same format as Crank's Table 5.1. The roots can then be used directly in the series (5.33) and (5.42) to reproduce the uptake curves of Fig. 5.7.
​
In[]:=
eq1=mqBesselJ[0,q]+2(1-m)BesselJ[1,q]==0/.m->m[s]/.q->q[s]
Out[]=
2BesselJ[1,q[s]](1-m[s])+BesselJ[0,q[s]]m[s]q[s]0
In[]:=
eq2=m'[s]^2+q'[s]^2==1
Out[]=
2
′
m
[s]
+
2
′
q
[s]
1
In[]:=
sol[k_]:=Quiet@NDSolve[{eq1,eq2,m[0]==0.0,q[0]==BesselJZero[1,k],q'[0]==0.1,m'[0]==0.1},{m,q},{s,-10,10},AccuracyGoal->5]
In[]:=
mus={1,9/10,4/5,7/10,3/5,1/2,2/5,3/10,1/5,1/10,0}
Out[]=
1,
9
10
,
4
5
,
7
10
,
3
5
,
1
2
,
2
5
,
3
10
,
1
5
,
1
10
,0
In[]:=
alpha=N[Join[{Infinity},Drop[-mus,1.]/(Drop[mus,1]-1)],{5,4}]
Out[]=
{∞,9.000,4.000,2.333,1.500,1.000,0.667,0.429,0.250,0.111,0.×
-4
10
}
In[]:=
tbl[k_]:=Flatten[Quiet@Table[FindRoot[(m[s]/.sol[k])==mus[[i]],{s,-1,0}],{i,1,Length@mus}],2]
In[]:=
tblf=Table[tbl[k],{k,1,6}]
Out[]=
{{s-1.75652,s-1.62371,s-1.48463,s-1.3379,s-1.18191,s-1.01491,s-0.835197,s-0.641693,s-0.434937,s-0.218502,s0.},{s-1.88804,s-1.78039,s-1.66926,s-1.55265,s-1.42724,s-1.28758,s-1.12493,s-0.926278,s-0.67516,s-0.360382,s0.},{s-1.9711,s-1.86788,s-1.76299,s-1.65527,s-1.54252,s-1.42039,s-1.28007,s-1.10387,s-0.858255,s-0.493076,s0.},{s-2.02935,s-1.92759,s-1.82489,s-1.72054,s-1.61311,s-1.49957,s-1.37299,s-1.21681,s-0.991447,s-0.610992,s0.},{s-2.07298,s-1.97188,s-1.87018,s-1.7674,s-1.6626,s-1.55362,s-1.43519,s-1.29336,s-1.08891,s-0.713234,s0.},{s-2.10728,s-2.00653,s-1.90536,s-1.80345,s-1.70012,s-1.59378,s-1.48041,s-1.3485,s-1.16192,s-0.800819,s0.}}
In[]:=
tbl2=Transpose@Join[{alpha},Table[First[q[s]/.Quiet@sol[j]/.(tblf[[j]][[i]])],{j,1,6},{i,1,11}]]
Out[]=
{{∞,2.40483,5.52008,8.65373,11.7915,14.9309,18.0711},{9.000,2.4922,5.55988,8.67929,11.8103,14.9458,18.0833},{4.000,2.58883,5.60829,8.71087,11.8337,14.9643,18.0987},{2.333,2.69618,5.66821,8.75082,11.8634,14.9879,18.1183},{1.500,2.81587,5.74378,8.80278,11.9026,15.0192,18.1443},{1.000,2.9496,5.84112,8.87272,11.9561,15.0624,18.1803},{0.667,3.0989,5.96923,8.9709,12.0334,15.1255,18.2334},{0.429,3.26454,6.14073,9.11565,12.1529,15.2255,18.3188},{0.250,3.44549,6.37095,9.33965,12.3543,15.4031,18.4754},{0.111,3.63743,6.66936,9.69067,12.721,15.7646,18.8215},{0.×
-4
10
,3.83171,7.01559,10.1735,13.3237,16.4706,19.6159}}
In[]:=
TableForm[tbl2,TableHeadings->{None,{Style["α",Darker@Green,14],Style["q1",Darker@Red,14],Style["q2",Darker@Red,14],Style["q3",Darker@Red,14],Style["q4",Darker@Red,14],Style["q5",Darker@Red,14],Style["q6",Darker@Red]}},TableSpacing->{3,3}]
Out[]//TableForm=
α
q1
q2
q3
q4
q5
q6
∞
2.40483
5.52008
8.65373
11.7915
14.9309
18.0711
9.000
2.4922
5.55988
8.67929
11.8103
14.9458
18.0833
4.000
2.58883
5.60829
8.71087
11.8337
14.9643
18.0987
2.333
2.69618
5.66821
8.75082
11.8634
14.9879
18.1183
1.500
2.81587
5.74378
8.80278
11.9026
15.0192
18.1443
1.000
2.9496
5.84112
8.87272
11.9561
15.0624
18.1803
0.667
3.0989
5.96923
8.9709
12.0334
15.1255
18.2334
0.429
3.26454
6.14073
9.11565
12.1529
15.2255
18.3188
0.250
3.44549
6.37095
9.33965
12.3543
15.4031
18.4754
0.111
3.63743
6.66936
9.69067
12.721
15.7646
18.8215
0.×
-4
10
3.83171
7.01559
10.1735
13.3237
16.4706
19.6159
In[]:=
Show[Table[ParametricPlot[Evaluate[{m[s],q[s]}/.sol[j]],{s,-2.2,0},AspectRatio->1,Frame->True,GridLines->Automatic,PlotStyle->Hue[j/7]],{j,1,6}],​​Table[ListPlot[Table[First[{m[s],q[s]}/.Quiet@sol[j]/.(tblf[[j]][[i]])],{i,1,11}],PlotStyle->{Hue[j/7],PointSize[0.0175]}],{j,1,6}],PlotRange->All,FrameLabel->{Style[m[s],14],Style[q[s],14]}]
Out[]=