Separation Science and Technology

ISSN: 0149-6395 (Print) 1520-5754 (Online) Journal homepage: http://www.tandfonline.com/loi/lsst20

Solution of the Diffusion Equations in a Gas

Centrifuge for Separation of Multi component
Mixtures

Chuntong Ying , Zhixiong Guo & Houston G. Wood

To cite this article: Chuntong Ying , Zhixiong Guo & Houston G. Wood (1996) Solution of
the Diffusion Equations in a Gas Centrifuge for Separation of Multi component Mixtures,
Separation Science and Technology, 31:18, 2455-2471, DOI: 10.1080/01496399608001059

To link to this article: http://dx.doi.org/10.1080/01496399608001059

Published online: 22 Aug 2006.

Submit your article to this journal

Article views: 42

View related articles

Citing articles: 15 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=lsst20

Download by: [University of Waterloo] Date: 12 November 2016, At: 09:46

SEPARATION SCIENCE AND TECHNOLOGY, 31(18), pp. 2455-2471, 1996

Solution of the Diffusion Equations in a Gas Centrifuge

for Separation of Multicomponent Mixtures

CHUNTONG YING and ZHIXIONG GUO

DEPARTMENT OF ENGINEERING PHYSICS
TSINGHUA UNIVERSITY
BEIJING. PEOPLE'S REPUBLIC OF CHINA

HOUSTON C. WOOD
DEPARTMENT OF MECHANICAL. AEROSPACE AND NUCLEAR ENGINEERING
UNIVERSITY OF VIRGINlA
CHARLOTTESVILLE, VIRGINIA. USA

ABSTRACT

The demand for stable isotopes in physical and chemical research and in medical
diagnostics is growing, and the gas centrifuge process is able to provide large
quantities of stable isotopes. A set of diffusion equations describing separation in
a gas centrifuge for a multicomponent mixture is established. These equations
involve general diffusion coefficients. Using the radial averaging method and the
simplified diffusion transport vector for a multicomponent isotopic mixture, non-
linear partial differential equations are transformed to a set of nonlinear ordinary
differential equations. An iteration method for the solution is presented. The rela-
tionship between the separation factor and the mass difference. yu = y$'-'"i, is
shown to be in agreement with both the computational and the experimental results
with very high precision.

Key Words. Gas centrifuge; Diffusion equation; Multicompo-

nent mixture; Stable isotopes; Separation

1. INTRODUCTION

Over the past years, many countries have developed the gas centrifuge
to separate the binary mixture of 235UFgand 238UF6for producing uranium
2455

Copyright 0 1996 by Marcel Dekker, Inc.

2456 YING, GUO, AND WOOD

enriched in the fissionable isotope 235Ufor fuel in nuclear reactors. Re-

cently, with the growing demand for stable isotopes in physical and chemi-
cal research and in medical diagnostics, the use of the gas centrifuge pro-
cess has made it possible to produce many different isotopes, especially
when large quantities are needed. Many countries [such as the United
States (1, 2), Russia ( 3 - 9 , and China (7)] and organizations [URENCO
(6)] have reported their activities in the public literature in the field of
multicomponent separation by gas centrifuge.
In a recent paper, Wood et al. (8) discussed multi-isotope separation in
a gas centrifuge. In that paper they solved Onsager's pancake equation
for the countercurrent flow field and the diffusion equation for each iso-
tope. This solution method was connected to an algorithm which optimizes
the centrifuge's performance. In the analysis, the diffusion coefficient was
assumed to be the same throughout the centrifuge, a good assumption for
gases with large molecular weights and small differences in the molecular
weights of the isotopes. However, when these conditions are not met,
variations in the diffusion coefficient may be important.
The object of the present paper is to establish a general set of diffusion
equations describing the separation phenomena in a gas centrifuge for
multicomponent mixtures. The diffusion equations in a gas centrifuge for
multicomponent mixtures are different from those for binary mixtures
because general diffusion coefficients are involved. These equations are
a set of nonlinear partial differential equations. Using the radial averaging
method (9- 12) and the simplified diffusion transport vector for multicom-
ponent isotopic mixtures (13), nonlinear ordinary differential equations
are obtained. An iteration method of the solution is presented. The compu-
tational and experimental results show the separation factor, yG,may be
expressed as yii = $''J-~I, where yo is the overall separation factor for
the unit mass difference, and M i and Mj are the molecular weights of the
ith and the jth component, respectively. This relationship is compared
with the experimental data and is found to agree with very high precision.

2. DIFFUSION EQUATIONS AND DIFFUSION

COEFFICIENTS FOR MULTICOMPONENT MIXTURES

For a mixture of n components, the diffusion transport vector Ji of the

ith isotope is (14)

DQdj + DTVln i = 1,2, ..., n (2.1)

where p is the density of the mixture, Ciis the concentration of the ith
DIFFUSION EQUATIONS IN A GAS CENTRIFUGE 2457

component, Mi is the molecular weight of the - ith component, is the

average molecular weight of the mixture, i.e., M = cY=,
MiCi, D, are
the general multicomponent diffusion coefficients (GMDC), and d, is the
diffusion driving force, which could be written as (15)
n

M P k= I
(2.2a)
j = 1, 2, ..., n
II

cd,=O (2.2b)
J= I

where p is the pressure, FAis the external body force per unit mass of
the kth component, and pk is the density of the kth component. If we
consider the process gas is rotating in the gas centrifuge with angular
velocity 0, then F, = 0%.Because p = EiLlpk, the last two terms in
Eq. (2.2a) are eliminated and we obtain

dj = VC' + C, i 3
1- 4 Vlnp; j = 1 , 2, ..., n (2.3)

The GMDC are determined by (14)

(2.4b)

here 6, is the Kronecker delta and 9 ik is the binary diffusion coefficient

(15)
+
T3I2J ( M ~ Mk)/(2M;Mk)
9 & = 2.628 x lo-' [m2/sl (2.5)
-
pU&.n:L ')
where p (atm) is the pressure, (T;k (lo-" m) is the molecular interaction
diameter; Szj;.') is the integral of interaction for molecular mass transfer;
M i and M k (mole) are the molecular weights of the ith and kth component,
respectively; and T (K) is the temperature. 9iik does not depend on the
concentration Ci or Ck, but from Eq. (2.4) one finds that Dij is a function
of the concentration.
Using Eqs. (2.1) to (2.5) to obtain the diffusion transport vector in the
gas centrifuge, we assume that the term of thermal diffusion DTVln T in
Eq. (2.1) is negligible.
2458 YING, GUO, AND WOOD

As for the diffusion coefficients, Levin and Ying (13) showed that if
any transformation such as
n..= D . . + A? (2.6)
gives a new value of the diffusion coefficient &, the diffusion transport
vector of Ji is unchanged. This is because
n n

2 AFdj = A,*
j= 1 j = 1
dj = 0 (2.7)

Reference 13 also shows that for an isotopic mixture with large molecu-

-
lar weights and a small difference in the molecular weights of the isotopes,
the diffusion coefficient CiD$ 9i(8ii - Mi/@) is diagonally dominant.
where

Then the approximate diffusion transport vector Ji in the gas centrifuge

for an isotopic mixture is obtained as

(2.9)

The diffusion equations in steady state are a set of mass conversation

equations for each component in the mixture. They are

i:
V. pV = Ci +
1
Ji = 0;
n-1
i = 1 , 2, ..., n - I (2.10a)

c,, = 1 - c c;
i= 1
(2. lob)

where V is the velocity of the mixture. The first term in Eq. (2.10a) is the
convection vector of the ith component and the second term is the diffu-
sion transport vector. By substituting Eq. (2.9) into Eq. (2.10), a set of
concentration equations is obtained:

- M ; )C;]
(2.1la)
i = 1, 2, ...) n - 1
DIFFUSION EQUATIONS IN A GAS CENTRIFUGE 2459

n-1
c n =1 - 2 cj
i= 1
(2.11b)

where V , is the radial component of the velocity and V, is the axial compo-
nent of the velocity. The boundary conditions are as follows:
a. There is no radial transport at the rotor wall and on the axis.

at r = r,,
aci + -a2r, (-
- M - MJCi = 0 (2.12a)
ar RT

at r = 0 ,
ac;
-=0 (2.12b)
ar
where r, is the radius of the rotating cylinder.
b. The axial transport over the end caps equals the withdrawal flow
rate.

i = 1, 2, ..., n

i = 1 , 2, ..., n
where ZH is the length of the gas centrifuge. F is the feed flow rate, and
6 is the “cut,” i.e., the product flow rate equals OF.
In addition, the feed concentration CF of the ith component is related
to product concentration CP and waste concentration CY of the ith compo-
nent by the overall balance equation for the ith component:
CF = ecp + (1 - 0)CW; i = 1 , 2, ..., n (2.14)
It is obvious that Eq. (2.1 1) are a set of nonlinear partial differential equa-
tions. The coefficient B i and average molecular weight R in Eq. (2.1la)
are dependent on the concentration.

3. RADIAL AVERAGING APPROXIMATION METHOD FOR

THE SOLUTION OF THE DIFFUSION EQUATIONS
An averaged concentration of the ith isotope ci is defined as

Ci2rrdr; i = 1 , 2, ..., n (3.1)

The variable cjdepends only on the axial coordinate z.

2460 YING, GUO, AND WOOD

The stream function + is defined as

+(r, z ) = J pVz2nrdr (3.2)
0

Then we have
a+/& = 2nrpVZ (3.3)
The radial convection term pV,(dCildr) in Eq. (2.11) can be neglected
because the radial component V , of the velocity is predominant over the
axial component V , only in the very thin Ekman layers near the end caps.
The diffusion term - p 9 i(d2Cj18z2)is negligible also. Integrating Eq.
(2.1la) over r, we obtain

_
dCi -
- LR'r (-
-_.

M - M;)C; + -- di' j r pVzrldrl

ar RT r p 9 i dz 0
(3.4)
- _-LR2r (-
M - M;)C; + --.
+ dz; i = 1, 2 , ..., n
RT 2 ~ r p 9dz~'
We introduce the net axial flow flux of the ith component PT,as

Pi* = lora+ (Jiz pCiV,

M
2nrdr; i = 1, 2, ..., n (3.5)

The net axial flow flux of mixture, P*, is

P* = c Pi*
j
I1

- 1
(3.6)

Using integration by parts, we obtain

(3.7)

Substituting Eq. (3.7) into Eq. ( 3 . 3 , we have

i = 1, 2 , ...) n
Using Eq. (3.4)from Eq. (3.8),we obtain
DIFFUSION EQUATIONS IN A GAS CENTRIFUGE 2461

Equation (3.9) is a set of concentration equations.

We define the following parameters:

(3.10)

Using a procedure similar to that used by Soubbaramayer (12) for the

binary mixture, the concentration equations in the enriching section are
obtained:
dCi
(1 + r Z i ) =~ (2eiYli + (ppi)Ci - cppiCIp; i = 1. 2, ..., n - 1
n- 1.

c n =1 - c ci
..

i= 1
(3.11)

where s = z/ra. We drop the overbar of C from Eq. (3.11).

The concentration equations in the stripping section are
dCi
(1 + YZAZ = ( 2 ~ i Y l i- Vwi)Ci + VwiCiw; i = 1, 2, ..., n - I

n-1

cn=l-cci (3.12)
i= I

The coefficients ei, Y l i , Yzi, ( p p i , and cpwi are dependent on concentration.

Equations (3.11) and (3.12) are two sets of nonlinear ordinary differential
equations. Before we start to .solve Eqs. (3.11) and (3.12), we need to
know the velocity distribution in the gas centrifuge, i.e., the V , or +. The
velocity field can be obtained in a variety of ways, such as with the On-
sager pancake model reported by Wood and Morton (16).However, here
we use a simplified model for the purpose of illustration. We assume that
+ has the following pattern (17):
+(c, q) = Rw[e-b‘S - (1 + b 1 5 ) e - ~ ” ~ 1 [ 4 q-( lq)12’3
(3.13)
+ Rs[ewbz5- (1 + b2~)e-2b2c]e-2T
where

The term with R w represents the wall-driven pattern; the term with Rs
represents the mechanical-driven pattern.
The constants bl and b2 vary for different process gases, and we take
12 5 b l A 2 5 25, 8 5 b2A2 5 15, and b l A 2 > b2A2 > 7 . 2 .
+,
When parameters F , 0, and Cp are given, using our iteration method
we obtain the solution of Eqs. (3.11) and (3.12), i.e., the concentration
distribution of each component in the gas centrifuge and the Cf, CW. The
solution of the concentration C j k f i )for the ( k + 1)th iteration in the
enriching section from Eq. (3.11) is

i = 1 , 2, ...) n - 1 (3.14)
n-1
cp+1)= 1 - Ci(k+l)
i= I

where

Ciois the concentration of the ith component at the feed point in the gas
centrifuge, S f = Z f / r a ,and ZF is the feed position.
The solution of the concentration Ci(k+i)for the ( k + 1)th iteration in
the stripping section from Eq. (3.12) is
DIFFUSION EQUATIONS IN A GAS CENTRIFUGE 2463

i = 1, 2 , ..., n - 1 (3.15)

where

From Eq. (3.14) we obtain

C-p + l )
- 1
Ci&k+1) - (3.16)
exp[ --B$?(s’)]
exp[ - B$?(SH)] + cp $? I,, SH

1 + YF’
ds’

From Eq. (3.15) we obtain

The right sides of Eqs. (3.16) and (3.17) are known. Combining Eqs. (3.16),
(3.17), and (2.141, the C ? ( k + l )CW(k+’),
, and Cibk+l)are obtained. At the
same time, we obtain the concentration distribution Cjk+I)for the ( k +
1)th iteration in the gas centrifuge. The criterion of the convergence of
the iteration is
m;x - Ci’k’

1-
IC$k+l)

i= 1 C i ( k + 1)

where E is the allowable error.

We calculated many examples, and some of them are shown in Figs. 1
to 4. The figures show the concentration distribution along the axis. Figure
1 shows the concentration of 234U,235U,and 236Uincreasing along the
axis from the bottom to the top. The concentration of 238Uis nearly con-
stant through the whole gas centrifuge because the concentration of 238U
in the feed flow is very large, 0.9925. Figure 2 shows a similar result, only
the dominant isotope is the light one instead of the heavy one. Figure 3
shows the concentration distribution of the tungsten isotopes in the gas
centrifuge. Figure 4 gives the results for Os04 separation. Note that the
maximum concentration of “0s is not at the ends of the gas centrifuge.
For long gas centrifuges, this phenomenon often appears.
2464 YING, GUO, AND WOOD

ctc,
FIG. I The concentration distribution of UF6 in the gas centrifuge.

FIG. 2 The concentration distribution of SF6 in the gas centrifuge.

DIFFUSION EQUATIONS IN A GAS CENTRIFUGE 2465

cjc,
FIG. 3 The concentration distribution of WF6 in the gas centrifuge.
2466 YING, GUO, AND WOOD

Head C,'

Feed C;'

Tails C,!"
FIG. 5 A schematic of a gas centrifuge.

4. RELATIONSHIP BETWEEN SEPARATION FACTORS

AND MASS DIFFERENCE

There are several definitions of the separation factors for multicompo-

nent mixtures, and we use the following definitions (see Fig. 5 ) :

The computational results show that the following relationship holds with
very high precision:
yij = yo - M ,
Mj
(4.2)

1.60

1.20

0.80

0.40

0.00
0.0 1.0 2.0 3.0 4.0 5.0
d. mij
FIG. 6 The relationship between In y i and (M, - M,).
DIFFUSION EQUATIONS IN A GAS CENTRIFUGE 2467

5 .OO

4.00

3.00

2.00

1.oo

0.00
0.0 2.0 4.0 6.0 8:O 10.0
A mij
FIG. 7 The relationship between In yij and ( M j - M i ) .

where yo is the overall separation factor with unit mass difference. yo

depends on the flow pattern in the gas centrifuge, the size and the operat-
ing parameters, etc.
The relationship between In yij and (Mi- Mi) for the computational
results is shown in Figs. 6 and 7. In the figures the points are the calculated
results. Almost all of them are on the straight lines which represent the
relationship.
Some authors, such as Von Halle (18) and Raichuraet al. (19), define the
separation factors differently. They made assumptions for the relationship
between their separation factor and mass difference. Here, we use the
definition of the separation factor as (4. I), and we find that the relationship
of (4.2) is kept with very high ?recision. This is useful for the calculation
of the concentration distribution in a cascade of gas centrifuges. Our work
about the cascade theory will be published in the near future.

5. COMPARISON OF THE EXPERIMENTAL RESULTS

WITH THE RELATIONSHIP

A variety of isotopes have been enriched using gas centrifuges in the

laboratory of Tsinghua University, and some experimental results are
shown in Figs. 8 to 10. In Figs. 9 and 10 the feed concentrations are
different because they were obtained from different areas.
We checked our experimental results and the results published by Rob-
erts (1) and Szady (2) with the relationship (4.2). The correlation coeffi-
cients r and the overall separation factors for unit mass difference, yo,
are calculated. When the absolute value of the correlation coefficient Irl
is close to unity, it means the relationship (4.2) agrees with the experimen-
2468 YING, GUO, AND WOOD

5 Yo
32 95.96
33 0.686
34 3.35
36 <o.os
32 95.02
33 0.75
34 4.21
36 <0.05 5 %
32 92.06
33 0.96
34 6.98
36 <0.05

FIG. 8 Sulfur isotopes were separated in a gas centrifuge. SF6 was the process gas.

187 1.69 0s Yo
I
188 13.39 184 0.019
189 16.14
190 26.15 187 0.94
192 40.94 188 8.99
DIFFUSION EQUATIONS IN A GAS CENTRIFUGE 2469

0s Yo
184 0.06
t 186 3.77
187 3.32
188 21.79
0s 189 21.40
184 0.027 190 26.61
I86 192 23.03
I87
188
189 16.12 0s %
I90 26.34 184 0.01
192 41.26 186 0.94
L 187 1.15
188 10.57
189 14.69
190 26.29
192 46.35

FIG. 10 Osmium isotopes were separated in a gas centrifuge. OsO4 was the process gas.

tal data. If Irl = 1, it means all the data are completely satisfied with the
relationship (4.2).
Table 1 lists the correlation coefficients for six experimental samples.
The first two columns, i.e., sulfur and osmium, are calculated using the
experimental data obtained at Tsinghua University; the other four col-
umns are calculated using the experimental data from References I and
2. Most of the correlation coefficients are greater then .99 which confirms
that relationship (4.2) is a good approximation for the separation factors.
yo is different for these six samples.

TABLE 1
The Correlation Coefficients r for the Experimental Data
Elements S 0s Cr S Kr Xe
Process gas SFs oso4 CrOzFz SFs Kr Xe
r .9997 .9983 ,9991 ,9960 .995 I .9895
YO
1.47 1.40 2.43 2.06 6.09 2.28
2470 YING, GUO, AND WOOD

6. CONCLUSIONS

A theory for the separation of multicomponent mixtures based on gen-

eral diffusion coefficients has been developed. The method of radial aver-
aging has been used to reduce the partial differential equations to ordinary
differential equations, and an iterative method has been used to obtain
solutions to these equations. A simplified model of the countercurrent
flow has been used here to calculate the concentration distributions for
a variety of isotopic mixtures. The relationship between the separation
factor and mass difference, yo = y F M 1 ,is determined from the results
of these calculations and compared to experimental results obtained at
Tsinghua University and at Oak Ridge, Tennessee. The formula is found
to be in good agreement.
In the future we plan to couple this separation theory with Onsager’s
pancake model for the centrifuge flow field. This will allow direct compari-
sons with the work reported by Wood et al. (8), and it will allow studies
in which the isotopes do not necessarily have large molecular weights or
small differences in molecular weights.

REFERENCES

1. W. L. Roberts, “Gas Centrifugation of Research Isotopes,” Nucl. Instrum. Methods

Phys. Res., A282, 271-276 (1989).
2. A. J. Szady, “Enrichment of Chromium Isotopes by Gas Centrifugation,” Ibid., A282,
277-280 (1989).
3. V. D. Borisevich, G . A. Potapov, G. A. Sulaberidze, andV. A. Chuzhinov, “Multicom-
ponent Isotope Separation in Cascades with Additional External Flows,” in Proceed-
ings of the Fourth Workshop on Separation Phenomena in Liquids and Gases (C. Ying,
Ed.), Tsinghua University, Beijing, China, 1995.
4. V. E. Fillipov and L. Yu. Sosnin, “Modeling of Gas Flow and Separation Process of
Multicomponent Mixture of Isotopes in Countercurrent Centrifuge with Internal Input
of Feed,” Zbid.
5. V. D. Borisevich, E. V. Levin, S. V. Yupatov, and E. M. Aisen, “Numerical Investiga-
tion of the Separation of Sulfur Isotopes in a Single Gas Centrifuge,“ At. Energy, 76(6),
454-458 (1994).
6. E. Ratz, E. Coester, and P. deJong, “Production of Stable Isotopes by Gas Centri-
fuge,” in Proceedings of the International Symposium on Synthesis and Applications
of Isotopes and Isotopically Labeled Compounds, Toronto, September 3-7, 1991.
7. C. Ying and Z. Guo, “Some Characteristics for Multicomponent Isotope Separation,”
in Proceedings of the Fourth Workshop on Separation Phenomena in Liquids and
Gases (C. Ying, Ed.), Tsinghua University, Beijing, China, 1995.
8. H. G. Wood, T. C. Mason, and Soubbaramayer, “Multi-Isotope Separation in a Gas
Centrifuge Using Onsager’s Pancake Model,” Sep. Sci. Technol., 31(9), 1185-1213
(1996).
9. R. L. Hoglund. J. Shacter, and E. Von Halle, “Diffusion Separation Method,” in
Encyclopedia of Chemical Technology, 30(13), 2631-2657 (1979).
DIFFUSION EQUATIONS IN A GAS CENTRIFUGE 2471

10. E. Von Halle, “The Countercurrent Gas Centrifuge for the Enrichment of U-235,” in
Proceedings 70th Annual Meeting AIChE, New York, 1977.
1 1 . K. Cohen, The Theory of Isotope Separafion, McGraw-Hill, New York, NY, 1952.
12. Soubbaramayer, “Centrifugation,” in Uranium Enrichment (S. Villani, Ed.), Springer-
Verlag, New York, NY, 1979.
13. E. V. Levin and C. Ying, “Diffusion Transport Vector for MulticomponentGas Separa-
tion in Ultracentrifuge,” Sep. Sci. Technol., 30(18), 3445-3458 (1995).
14. C. F. Curtiss, “Symmetric Gaseous Diffusion Coefficients,” J. Chem. Phys., 49(7),
29 17-2919 (1968).
15. J. 0. Hirschfelder, C. F. Curtiss, and R. B. Bird, The Molecular Theory ofGases and
Liquids, Wiley, New York, NY, 1954.
16. H. G. Wood and J. B. Morton, “Onsager Pancake Approximation for the Fluid Dynam-
ics of a Gas Centrifuge,” J. Fluid Mech., 101(1), 1-31 (1980).
17. D. R. Olander, “The Theory of Uranium Enrichment by the Gas Centrifuge,” Prog.
Nucl. Energy, 8(1), 1-33 (1981).
18. E. Von Halle, “Multicomponent Isotope Separation in Matched Abundance Ratio Cas-
cade of Stages with Large Separation Factor,” in Proceedings ofthe First Workshop
on Separation Phenomena in Liquids and Gases, Darmstadt, Germany, July 20-23,
1987.
19. R. C. Raichura, M. A. M. Al-Janabi, and G. M. Langbein, “Some Aspects of the
Separation of Multi-isotope Mixtures with Gas Centrifuge,” in Proceedings of the
Second Workshop on Separation Phenomena in Liquids and Gases (P. Louvet, P.
Noe, and Soubbaramayer, Eds.), Versailles, France, July 10-12, 1989.

Received by editor December 22, 1995