Computation of Phase and Chemical
EquiIibrium
Port IV: Approach to Chemical Equilibrium
The Rand algorithm for calculation of compositions in phase and chemical
equilibrium is extended to permit specificationsof extents of reactions that cannot
be taken at equilibrium. Pseudo-Gibbs free energies of formation are introduced
to permit specification of temperature approaches. The extended algorithm is
demonstrated for methane reforming.
SCOPE
In Parts I-IZI of this series, we considered methods for cal-
culating compositions in phase and chemical equilibrium. Part
TV presents modifications when some of the reactions cannot be
taken at equilibrium. In these cases, it is possible to minimize
Gibbs free energy given “approaches to equilibrium” such as
the extents of reactions and “temperature approaches to
equilibrium. ”
Smith (1980) terms methods for minimization that adjust
extents of reaction “stoichiometric methods” and methods that
adjust moles of species in each phase “non-stoichiometric
methods.” In this paper, we show that when one or more
reactions cannot be taken at equilibrium, the stoichiometry of
an independent set of reactions must be specified, whether
CONCLUSIONS AND SIGNIFICANCE
(I) The extended Rand algorithm works well to compute
compositions in chemical equilibrium, given specifications for
extent or temperature approach of reactions that cannot be
taken at equilibrium.
(2) This technique adds stoichiometry through extents of
reaction and pseudo-Gibbs free energy of formation, while
retaining the advantages of methods that adjust moles of
species j in phase 1 to minimize Gibbs free energy. These
advantages include smaller matrices with fewer operations
when: (a)there are few condensed species; and (b) the number
of independent chemical reactions is large compared to the
number of reactions with specifications for extent. Further-
more, in Part 11of this series, a phase-splitting algorithm has
COMPUTATION OF CHEMICAL AND PHASE EQUILIBRIUM
Compositions in chemical and phase equilibrium are at the
minimum Gibbs free energy:
oM)1-1541-81-4822-0466-$2.00 0 T h e American Institute of Chemical Engineers,
1981.
Page 466 May, 1981
CHARLES W. WHITE, 111
WARREN D. SEIDER
Department of Chemical Engineering
University of Pennsylvania
Philadelphia, PA 19104
extents of reaction or moles of species in each phase are ad-
justed. Hence, Smiths distinction does not apply.
We concentrate on methods that adjust moles of species in
each phase where it is not obvious how to handle approach
specifications. In particular, we illustrate a procedure for ex-
tending the Rand method (White et al., 1958), an implementa-
tion of Newton’s method. When all reactions can be taken at
equilibrium, Smith (1980) favors methods that adjust extents of
reaction (stoichiometric coefficients are required), but there
are good reasons for retaining methods that adjust moles of
species in each phase (stoichiometric coefficients are not re-
quired). These are discussed in the section on computation of
Chemical and Phase Equilibrium.
been coupled to the Rand method and shown to improve its
efficiency and reliability when the phase distribution is uncer-
tain at equilibrium. To our knowledge, phase-splitting al-
gorithms have not been coupled to minimization methods that
adjust extent of reaction i in phase 1. In principle, this coupling
should work well and is worthy of experimentation to deter-
mine its advantages and disadvantages.
(3)When one or more reactions cannot be taken at equilib-
rium, Smith’s distinction between stoichiometric and
non-stoichiometric methods does not apply. Both require spec-
ification of stoichiometric coefficients for R independent re-
actions. The distinction lies, however, in the use of extents of
reaction or mole numbers as iteration variables.
)“I j=S+l / = I
subject to mass balance constraints, where C is the number of
chemical species, P is the number of mixed phases (vapor,
liquid,. solid), S is the number of condensed species (which
AlChE Journal (Vol. 27, No. 3)
appear in only one pure phase, are normally solid, and d o not
distribute amongst other phases), n,l is the moles of compound j
in phase 1, and G,I is the chemical potential of compoundj in
phase 1 When the number of independent chemical reactions
equals C - p, where p is the rank of the atom matrix (mlh),the
mass balances can be expressed as atom balances:
bk =
S
,=I
mjknf + cc L P
,=s+1 1=1
m,~njl k=l, . ,E
where bh I S the gram-atoms of element k , mJh is the number of
atoms of element k in compoundj, and E is the number of
elements Methods that adjust nf and n,,to minimize G subject
to mass balances are termed non-stoichiometric by Smith
(1980)
Alternatively, extents of reactions can be introduced:
I n,' = nj0
R
+ c Vij& j = 1, , , , , s
where R is the number of independent chemical reactions (nor-
mally C - p), is the extent of the reaction i, n: is the moles of
compound j in the feed mixture and vij are the stoichiometric
coefficients of compound j in reaction i given by the atom
balances MT E~ = 0. Usually, K-values are introduced as a "no-
tational dgviCe" (Smith, 1980), that is, for reaction i:
K 1. = -A G ! L R T ~ ~ (4)
and at equilibrium
j=1
where uj is the activity of speciesj. Methods that adjust ti to
minimize G are termed stoichiometric by Smith (1980); that is,
they require knowledge of stoichiometric coefficients for an
independent set of reactions.
Features of Stoichiometric Methods
Smith (1980) presents an extensive review with emphasis 011
problem5 in metallurgy when P = 1 and S is uncertain and
concludes that stoichiometric methods are advantageous be-
cause they avoid singular coefficient matrices when n: ---* 0. In
addition, these methods d o not require equality constraints in
the form of mass balances; for example, atom balances in Eq. 2.
But, the inequality constraints, nJ 2 0, can be more difficult to
satisfy when adjusting 8.
Also, it can be difficufi to guess 5. Even with knowledge of the
S
reaction mechanism, estimates fo; can differ significantly from
the solution at equilibrium.
When P > 1, nested iteration algorithms are commonly used
to compute compositions in phase and chemical equilibrium.
Typically, t l ,i = 1, . . . , R , are adjusted in an outer loop and for
S.
each N,[ are computed in an inner loop; that is, compositions
in phase equilibrium are determined in an inner loop.
Many such algorithms exist and, hence, it is difficult to quan-
tify computational effort. In o n e class, Newton or quasi-Newton
methods give quadratic convergence for the inner and outer
loops. When these methods are applied to systems with large C
and R , evaluation and inversion of lacobian matrices or updates
to approximate I-' consume most operations. Hence, in the
outer loom, the :umber of operations increases with Rm', where
1 < $,, when R is sufficiently large. In the inner loop, the
number of operations increase linearly with C - S and Pmz,
where 1 < $*.
Smith discusses the possibility of adjusting tll,i = 1, . . . , R,
and 1 = 1, . , . , P . Since,
R j = S + 1,. . . ,C
nfl = fl; + 2 Yij& 1 = ?,. . . , P
i=1
AlChE Jwnwl (Vol. 27, No. 3)
an iterative inner loop to compute compositions in phase
equilibrium is unnecessary. However, no such method has been
implemented to our knowledge, probably because the number
of operations would increase with [RPl03, where 1 < 4 3 , when
using Newton or quasi-Newton methods.
(2)
Features of Non-Stoichiometric Methods
Seider and coworkers (1980) in another review emphasize
problems in chemical processing where P is uncertain and S = 0.
They review three methods that evaluate and invert a matrix of
partial derivatives or approximate its inverse to adjust n,!.These
are: (1)quadratic programming (Gautam and Seider, 1979a); (2)
Powell's quasi-Newton method (George e t al., 1976); (3)variable
metric projection (Castillo and Grossmann, 1980). In these
cases, the number of operations increase with [(C - S)P]"*,
where 1 < &.
The Rand method (White et al., 1958)is also an implementa-
tion of Newton's method. But, its coefficient matrix has S + P +
E rows and columns which are independent of C - S , as illus-
trated in the following review of the Rand method.
A quadratic Taylor Series approximates the Gibbs free energy
at 8, a vector of moIe numbers in close proximity to the vector of
guessed mole numbers, 9:
-
N is computed at minimum Q subject to the atom balance
constraints (Eq. 2). An unconstrained objective function, using
Lagrange multipliers, v k ( k = 1, . . . , E )
+ 9?T
F{NJ = Q{8} Th(hk - m,aNC
h=l J= 1
is minimized:
to.give linear equations in _N and v k . Substituting NI =
N,,, this set is reduced to S + P +E linear equations
J=Sfl
and solved by L - Ufactorization to give NS, h'i,and Th. Then, N,,
are computed such that NJl 2 0. Part I gives details of a unidirec-
tional search that satisfies these inequality constraints.
Operations for matrix inversion increase with (S + P + E)+5,
where 1 < &. Some elements of the coefficient matrix involve
summation over species from S +1 to C. These operations and
those to compute N,, increase linearly with C-S. Hence, for
large C and small E , the Rand method requiresfewer operations
than other Newton methods with larger coefficient matrices.
When the phase distribution is uncertain at equilibrium,
Gautam and Seider (1979b) couple a phase-splitting algorithm
with the Rand method to improve its efficiency and reliability.
The algorithm permits very poor guesses for phase distribution;
for example, all vapor or all liquid. Vapor or liquid phases are
added as necessary during calculations to minimize Gibbs free
energy. The search space is small and the phase-splitting al-
gorithm has been successful in avoiding local and constrained
minima. In principle, phase-splitting algorithms could be
coupled to algorithms that adjust =$,I, but none have been tested
to our knowledge.
REACTIONS NOT TAKEN A T EQUlLlBRlUM
In many reaction systems, one or more reactions cannot be
taken at equilibrium. With as few as one reaction not taken at
equilibrium concentrations at minimum Gibbs free energy de-
pend upon the stoichiometry of R independent reactions. Con-
sider, for example, the system 0 - 0 2 - 0 3 with R = C - p = 3 -
&y,1981 Page467
1 = 2 independent reactions; e.g., matrix is selected arbitrarily, with the restriction that it be ofRl
0 + o2z O3 (R1)
rank.

20 * Oe (R2)
Pseudo-Gibbs Free Energy of Formation
If R l proceeds tR1kmol, the moles of O3 are known. Whereas,
with the reactions When a common standard state is used for species j in all
phases:
0 + O2 e o3 (RS)
3 0 Z O3 (R3)

the moles of O2are established. Hence, compositionsdepend on This requires one Gibbs free energy of formation for each
the reaction path given by the stoichiometry of all R reactions. species j ; that is, C values. But, compositions in chemical
And, hence, when one or more reactions cannot be taken at equilibrium can be computed with Eqs. 4 and 5, which require
equilibrium, Smith's distinction between stoichiometric and AGP for each reaction i; that is, R values. As indicated by
non-stoichiometric methods does not apply. Both require spec- Sorensen and Stewart (1980), this is t h e minimum ther-
ification of stoichioinetric coefficients for R independent re- modynamic data required. Hence, C Gibbs free energies of
actions. The distinction lies, however, in the use of -[ or _N as g
formation are not required and any vector which satisfies
iteration variables.
Given the extent of a reaction that cannot be taken at equilib-
rium, [!,, Eq. 5 does not apply for i = n . Or, given the tempera-
can be used. Since go is a vector of dimension R and 2 is a
ture approach to equilibrium for reaction n, AT,,, Eq. 4 is
vector of dimension C , an infinite number o f g exist. Unlike Go,
replaced by:
the elements of g have no physical significance, but have a
K,, = ,-bGn/BTn similar function. Hence, we call them pseudo-Gibbs free ener-
(9)
gies of formation.
where T,, = T,,/, + AT,,, with AT,, > 0 for exothermic reactions Use o f 3 is required when one or more K-values are specified.
and AT,, < 0 for endothermic reactions. When extents of re-
actions are adjusted to minimize Gibbs free energy and K-values
are used as a notational device, minor modifications are neces- Temperature Approach to Chemical Equilibrium
sary, as shown by White (1979) in his FORTRAN program, In Eq. 9, the Gibbs free energy of reaction, AGE, is evaluated
EQUILK. Note that for these methods, specification of an ex- at T,,(= Ts,,, + AT,,) using:
tent of reaction reduces computational effort since it reduces the
number of iteration variables.
In methods that adjust moles of species in each phase, it is not J=I
obvious how to handle specifications for extents and tempera-
ture approaches of reactions that cannot be taken at equilibrium. where GP is the Gibbs free energy of formation of species j from
We have developed techniques that we illustrate as an extension its elements in their standard state at T,, and, usually, P,,,. But,
of the Rand method in the remainder of this section. T,, differs from T,,,,,, giving rise to different standard states for
reactions not taken at equilibrium and those taken at equilib-
rium.
Extents for Reactions Not Taken at Equilibrium Since the Rand method requires a single standard state for
species j in all reactions, we compute pseudo-Gibbs free ener-
Given [,!, to adapt the Rand method, it is necessary t o ex- gies of formation as follows:
press [,, explicitly as a function of the iteration variables, "j,
(1) AGPI%T, is computed for each reaction i, using T , = T,,,,
j = 1, . . . , S and N / ( = C )1
j=S+l
Nj/ , = 1, . . . , P . The result- for reactions taken at equilibrium and T , = T,, for reactions not
taken at equilibrium. In matrix notation:
ing equations, in residual form, are added to the objective
hGoI%T
- - = v Go/%T (17)
function, Eq. 7 , multiplied by Lagrange multipliers.
To isolate t,,for the reactions not taken at equilibrium, Eq. 3 (2) As previously, let vRbe any R x R nonsingular submatrix
is written in matrix notation of v.Then, Gibbs f r e e e n e r g y of reaction for R reactions is
-
N = _ N " +=
v T- [ (10) related to pseudo-Gibbs free energy of formation for R species
P
where N j = N,C, j = 1, . . . , S, and h'j = Nj,, j = S + 1, AGoI%T = ER ?3~l%Tstl~
-~ (18)
/= 1
. . . , C. Eq. SO cannot be solved for because Eis not a square
v
matrix. Since is an R X C matrix, extents of-R reactions are or
related to moles of C species. Let vRbe any R x R nonsingular 9?,qR/%T,s,,,s= @-R ) - I AGoI%T
submatrix of?. Then extents of R reactions are related to moles (19)
of R species (&): A(AGOI%T)
=-

ZR = _N"R f y.'$
- (11) Eq. 19 gives $3, for R species with i a j taken as zero for the
or remaining C - R species.
Normally, chemical potentials for species j in phase I are:
= ($?-'(&R
-
- &'g)
(12)
= & W R - Ha
and, for each reaction not taken at equilibrium: where GP, is Gibbs free energy of formation in the standard state
for s p e c i e s j in phase 1. With P > 1, as Gibbs free energy is
e,, = c A,,,";
R P

c A,,,[z NJr
$

- Np] f - A!'] (13) minimized, compositions of phases are computed at equilibrium

J=l /=I
with Ts,l,7and P,,,,. It is possible to have different approach
For convenience in notation, the S condensed or solid species temperatures for each solution phase 1 and use Eq. 19 to give
are included within the subset of R species. Otherwise, the Z R pseudo-Gibbs free energy for species j in phase I, %,[.

Page 468 May, 1981 AlChE Journal (Vol. 27, No. 3)

MokrofRodvct~ System for Process Engineering-Evans et al., 1979). The ob-
jective function, Eq. 7 is augmented with Eqs. 13 and 21:
It is less common to specify the moles of compound j in the
product stream, rather than t,$or ATn. But, this specification is
possible and requires a constraint equation of the form:
Nj-N$=O j = 1,.. . , S
S
P
Nj - 2 N j l = 0 j = S + 1 , .. . , C
1=1

EXTENDED RAND ALGORITHM

C' P ,
The Rand Algorithm has been extended in the FORTRAN + 2 ?TJ"(Nj"- 2 Njq) (22)
program GIBBS and has been installed in ASPEN (Advanced jr=S+l /=1

"j"
"t, T J ,

j ' = 1,
k = 1,. . . , E n=l,.. . ,R' . . . , j" = S + 1, . . . , C'
S'
k = l r' B , , . . . B,, n P

j=S+l
E
1=1

P
j = l
A,,, sjj,

S
n = l D , , . . . DIR'

R' DRpl. . . DRtRr

j' = 1

S'
j" = S +1
C'
P r) P

Right Hand Side

P ,' P

Figure 1. Coefficient rnotrix.

AlCkE Journal (Vol. 27, No. 3) May, 1981 Page 469

 TABLE 1. FEED TO METHANE REFORMER TABLE2.CHEMICALEQUILIBRIUM

Species Mole fraction Feed, Guesses, Solution at G,,,,,,,

Species inol in ol in 01
HJO 0 8407
H, 0 0156 CH, 15.14 15.14 0.875
CH, 0 1283 H,O 84 07 84.07 61.821
CLHO 0 0061 S, 0.59 0.59 0.59
C&L 0 0027 HL 0.20 0.20 50.980
C,Hic 0 0007 CO 0 0.01 6.282
K, 0 0058 CO, 0 0.01 7,983
GIRT -1871.076 - 1963.1983
FIRT -1870.459 - 1963.1983

TABLE 3. THERMOPHYSICAL
PROPERTYDATA+

Heat capacity coefs, *

Species G'IBT HVRT 1 b C d

CH4 -5.7326 - 8.4408 1.924X l o 4 5.209X101 1.197x lo-' -1.131 X

H,O -25.780 -27.271 3.222X10' 1.923 1 . 0 5 5lo-'
~ -3.594X10-6
N2 0 0 3.113X104 - 1.356X 10' 2.678~ lo-' - 1.167x lo-'
H1 0 0 2.713X104 9.268 - 1.380X lo-' 7.640x10-6
co - 15.480 - 12.465 3.085~10~ - 1.285 x 10' 2.787x10-' - 1.271X 10+
c02 -44.473 -44.374 1.978x l o 4 7.339x 101 -5.598XlW' 1 . 7 1 4 1~O P

*Where c) = u + h l + cl' A dTJ with T in 'K

tValuer taken from Reid et al. (1977)

E TRAJECTORIESOF G
T A ~ L 4. AND F FOR CASES 1 AND 3 TABLE 5 TEMPERATUREAPPROACHTO CHEMICAL EQUILIBRIUM

Iteratioii GIRT FIRT P\c~rltlo Glle55f25, Solution at G,,,,,,

Sp e a e s w9
21 mol mol
Case 1
0 -1871.0762 - 1870.4593 CH, -0 68240 15 14 1382
1 - 1876.4287 - 1876.4287 H,O 0 84 07 62 465
2 -1897.3813 -1897.3813 h, 0 0 59 0 59
3 -1936.6702 '1936.6702 H2 0 0 20 49 321
4 -1962.7549 - 1962,7549 CO 0 0 01 5 912
5 -1963.1887 - 1963.1887 CO , -0 046885 0 01 7 847
6 - 1963.1982 -1963.1982 GIRT -57 9771 -142 1963
7 - 1963.1983 - 1963.1983
8 - 1963.1983 - 1963.1983
9 - 1963.1983 - 1963.1983 Unlike methods that adjust t,,specification o f t i increases the
Case 3 computational effort for methods that adjust njl by introducing
0 - 1963.1983 - 1962.8274 an equality constraint (Eq. 13). Qualitatively, the operations
1 - 1963.0346 - 1963.0346 required for the extended Rand algorithm increase with (S P +
2
3
- 1963.0346
- 1963.0346
-1963,0346
- 1963.0346
+ E + R')"6, where 1 < &. Hence, this approach is advanta-
geous when few of the independent chemical reactions have
extents specified.
where R' is the number of reactions with t,,given, S' is the
number of solid (condensed) species with moles in product
RESULTS
given, and C'-S is,the number of species in solution phases with
moles in product given. At the minimum: The augmented Rand Algorithm was tested for the steam
reforming system described by Rase (1977) in Case Study 111.
dF - aF - dF - d F - d F - aF - 0 (23) Feed gas enters a tubular reformer at 637°K and 1.235MPa with
ah': ah',, d~,! d~jr dry composition in Table 1. A direct-fired furnace is used to supply
j = 1,.. . , C the heat required by this endothermic system which involves
I = 1,., , , P fast reactions adequately modelled by
k = 1,.. . , E CH, + HZO + C O + 3H2 (24)
n = 1 , . . . ,R'
j ' = 1, . . . , S' CO + H2O 8 C O : + H;! (25)
j " = S + 1, . . . , C' The reactants are heated to 1067°K with C,H,, C3H8,and C,H,,
Substituting NI = ZNj,, we have P + E + R' +
S ' C' linear+ cracked to CH, i n the inlet portion of the tube. Under these
equations to be solved for X;, Nl, T ~ - n,,,
, njt and n,..The conditions, the compositions leaving the converter satisfy
coefficient matrix is shown in Figure 1.
Care must be taken to give consistent specifications for the
approach to equilibrium. If'the extent of a reaction is specified, it
is redundant to specify the moles of a species confined to this and the water-gas shift reaction is said to be at equilibrium. The
reaction. Similar redundancies must be avoided. compositions a l s o satisfy

Page 470 May, 1981 AlChE Journal (Vol. 27, NO.3)

 BR = vector of moles of species j - R elements
P = n u m b er of solution phases; vapor, liquid, or solid so-
lutions
but with t h e methanation K-value evaluated using a tempera- P,,.,,,,.
= pressure of system
ture variously reported as 15-25°C below t h e actual (Rase, 1977). Q = quadratic approximation to Gibbs free energy
Three sample calculations w e r e carried out for this system: R = n u m b er of independent chemical reactions
(1) C h e m i c d Equilibrium. Compositions were computed for R’ = n u m b er of reactions with specifications for t,,
both reactions taken in equilibrium at 1067°K and 1.235 MPa. % = universal gas constant
Fe e d and guessed compositions and compositions a t equilib- S = n u m b er of solid (condensed) species; occur in a pur e
rium, after 9 iterations, are shown in Table 2. Thermophysical phase and do not distribute among other phases
propert!- data for the ideal gas mixture is given in Table 3 . And, S’ = n u m b er of solid (condensed) species with specifica-
the trajectories of G/%T and Fl%T as a function of iteration tions for moles in product
number are tabulated in Table 4. Guessed compositions differ T,, = temperature for reaction n (= Ts,,s + AT,,)
from zero to avoid underflow when evaluating logarithms. After T,,,, = temperature of reacting mixture; i.e., system temper-
1 iteration GI%T and Fl%T agree to greater than 8 significant ature
figures. AT,, = temperature approach to equilibrium for reaction n
(2) Temperature Approach to Equilibrium. Compositions were yj = mole fraction of species j in vapor
computed for the conditions given by Rase (1977) with an ap-
proach temperature of -25°C. Feed and guessed compositions
Greek Symbols
are in Table 2. Pseudo-Gibbs free energies of formation are
shown in Table 5 , along with compositions at minimum Gibbs Vl, = stoichiometric coefficient of species j in reaction i
free energy, after Siterations of the Rand Method. Rasedoes not -
U
-
= stoichiometric coefficient matrix with elements v,,
report product compositions, b u t indicates comparable percent = R x R submatrix of - with rank R
carbon converted. = extent of reaction i
(3) Extent of Reaction. 13.7583 mollh of methane w e r e con- t,, = extent of reaction n not taken at equilibrium
verted by the gasification reaction in t h e second calculation ~ = vector of extents of reaction with R elements
above. To test the algorithm’s ability to move from lower to n = Lagrange multiplier
higher G i t h s free energy, guessed compositions w e r e taken to p = rank of atom matrix
be an equili1)riuin (Table 2). Only three iterations were required
to achieve convergence at th e solution given in Table 5. Table 4
shows that GI%T increases, whereas FI%T decreases to a Superscripts
minimum as t h e extent of reaction specification is satisfied. C = condensed (solid) species
0 = initial
ACKNOWLEDGMENT T = matrix transpose

The helpful comments of R. Gautam and E. B. Dussan V. are appre-

ciated Partial financial support for this work was provided by the U.S. Subsvints
Department of Energy under Contract No. E(49-18)-2295,Task Order
No. 9. the ASPEN Project. i = reaction counter
j = species counter
NOMENCLATURE k = element counter
1 = solution phase counter
= activity of species j fl = reaction not taken at equilibrium
= (.XI-’
-
with elements Aij
= gram-atoms of element k
= n u m b er of chemical species LITERATURE CITED
= n u m b er of species in solution phases with specifica-
tions for moles in product + S Evans, L. B., et al., “ASPEN: An Advanced System for Process Engi-
= ideal gas heat capacity, J/Kmol “ K neering,” accepted for publication in Comp. and Chem. Eng. (1979).
Gautam, R., and W. D. Seider, “Calculation of Phase and Chemical
= number of elements or atoms Equilibrium. Part I: Local and Constrained Minima in Gibbs Free
51 = fugacity of species j in phase Z Energy,” AZChEJ., 25, 6 (1979a).
A = fugacity of species j in standard state for phase 1 Gautam, R . , and W. D. Seider, “Calculation of Phase and Chemical
F = unconstrained objective function for minimization Equilibrium. Part 11: Phase-Splitting,” AZChE I., 25, 6, (197913).
= Gibbs free energy Gautam, R., and W. D. Seider, “Calculation of Phase and Chemical
= partial m o l d Gibbs free energy for species j in phase 1 Equilibrium. Part 111: Electrolytic Solutions,” AZChE J . , 25, 6
(chemical potential) (197%).
= Gibbs free energy of formation of species j from its Rase, H. F., Chemical Reactor Design for Process Plants, Vol. 2 : Case
elements in the standard state Studies and Design Data, Wiley (1977).
Reid, R. C., Prausnitz, J. M . , and T. M. Sherwood, The Properties of
= vector of pseudo-Gibbs free energy of formation at T,,,,v
Gases and Liquids, 3rd e d . , McGraw-Hill (1977).
R elements
and PS,,,.; Seider, W. D., R. Gautam, and C. W. White, 111, “Computation of
= Gibbs free energy of reaction i with all species in their Phase and Chemical Equilibrium-A Review,” ACS Symposium
standard states Series No. 124 Computer Applications to Chemical Process Design
Hj” = enthalpy of formation of species j from its elements and Simulation, ed., R. 6. Squires and G. V. Reklaitis (1980).
L-’ = inverse of Jacobian matrix Smith, W. R., “The Computation of Chemical Equilibria in Complex
Systems,” ZEC Fund., 19, 1 (1980).
K, = equilibrium constant for reaction i
= number of atoms of element k in species j Sorensen, J. P., and W. E. Stewart, “Structural Analysis of Multicom-
= atom matrix with elements mjk ponent Reaction Models: Part I, Systematic Editing of Kinetic and
Thermodynamic Values,” AZChE I., 26, 1 (1980).
= moles of species j White, 111, C . W., ASPEN Technical Documentation, ASPEN Project,
= moles of species j in p u r e solid (condensed) phase Dept. of Chem. Eng., M.I.T. (1979).
= moles of species j in solution phase 1 White, W. B., S. M. Johnson, and G. B. Dantzig, “Chemical Equilib-
= moles of species j in feed rium in Complex Mixtures,” J . Chem. Phys., 28, 751 (1958).
= moles of solution phase 1
= vector of moles of species j - C elements Manuscript receivedAugust4.1980; revision recetved ondacceptedAugust28,19Bo.

AlChE Journal (Vol. 27, No. 3) May, 1981 Page 471