Thermodynamics of mixtures: Functions of mixing and excess functions

R. Nieto, M. C. González, and F. Herrero

Energetic Engineering and Fluid Mechanics Department, E. T. S. I. I. Madrid, U. P. M., Spain
共Received 27 February 1999; accepted 2 July 1999兲
Applying thermodynamics to realistic systems requires a knowledge of the thermodynamic
properties of mixtures. Functions of mixing and excess functions provide a useful approach. The
concepts are simple and their application straightforward, but students often fail to apply them
correctly when they are given only a theoretical explanation. We discuss some typical mistakes and
some problems we have found useful for overcoming them. © 1999 American Association of Physics
Teachers.

I. INTRODUCTION particles of every component has the same meaning as the

derivative for one-component closed systems:
Because most real world systems are mixtures, thermody-

冉 冊
C
namics would be an empty theoretical framework with few ⳵U
applications without the ability to calculate the properties of dU⫽T dS⫺ P dV⫹ 兺
i⫽1 ⳵Ni
dN i , 共4兲
mixtures. Fortunately, there is a simple and useful way to S,V,N j⫽i

兺冉 冊
calculate the thermodynamic properties of mixtures based on C
knowing the thermodynamic properties of the pure sub- ⳵G
stances involved and the functions of mixing. dG⫽⫺S dT⫹V d P⫹ dN i . 共5兲
i⫽1 ⳵Ni T, P,N j⫽i
For pure substances, simple models are useful for calcu-
lating the properties of real substances, for example, the ideal The partial derivatives in Eqs. 共4兲 and 共5兲 can be shown to be
gas model and its corrections. The corresponding model for the same by using the relation G⫽U⫹ PV⫺TS. Each de-
mixtures is the ideal mixture 共the Lewis–Randall model兲 and
rivative is the chemical potential of component i and is de-
the excess function.
Because the study of mixtures is not common in the un- noted by ␮ i . 7–9
dergraduate physics curriculum, we will first give a brief From Eq. 共3兲, any extensive thermodynamic property X
summary of the formalism.1,2 Our treatment will be based on can be obtained as
the number of particles rather than the number of moles to X⫽X 共 P,T,N 1 ,...,N C 兲 . 共6兲
facilitate the connection with statistical mechanics.3 Al-
though our approach is straightforward, many students fail to The total derivative of X can be written as
apply the concepts when they are given only a theoretical

冉 冊 冉 冊 冉 冊
C
explanation. For this reason we also give some selected ⳵X ⳵X ⳵X
problems. dX⫽
⳵T
dT⫹
⳵P
d P⫹ 兺
i⫽1 ⳵Ni
dN i .
P,N j T,N j T, P,N j⫽i

II. DEFINITIONS 共7兲

10
The generalized work W can be expressed as We call x i the partial property x of component i:

dW⫽ 兺j y j,ex dX j , 共1兲 x i⫽ 冉 冊

⳵X
⳵Ni T, P,N j⫽i
. 共8兲

where y j,ex are the generalized external forces, and X j are the Note that this symbolic expression is not a single definition,
conjugate properties known as metric variables.6 A typical but a set of definitions, because X may be replaced by V, H,
example is the work due to a change of volume, where the U, S, etc.
generalized force is the external pressure P ex and the metric If all the N i ’s are multiplied by ␭ at constant T and P, the
variable is the volume V. Systems may have more than one value of X would be multiplied by ␭ because X is extensive.
metric variable, but for simplicity, we will consider systems Applying the Euler theorem for homogeneous functions, we
that have only the volume. For these systems, the internal have
energy U(S,V,N) of a pure substance as a function of the
C
entropy S, volume V, and number of particles N provides a
complete description of the equilibrium state of the system. X⫽ 兺 x iN i .
i⫽1
共9兲
Such an equation is known as a characteristic equation.
Characteristic equations may be given for the other ther- By comparing the definition of the partial derivative of the
modynamic potentials such as the enthalpy H, the Helmholtz Gibbs free energy, g i , with the definition of the chemical
free energy F, and the Gibbs free energy G. For a multicom- potential, we see that
ponent system of C components, we have
␮ i ⫽g i , 共10兲
U⫽U 共 S,V,N 1 ,...,N C 兲 , 共2兲
and from Eq. 共9兲,
G⫽G 共 T, P,N 1 ,...,N C 兲 . 共3兲
C C
From Eqs. 共2兲 and 共3兲 we can obtain the corresponding Gibbs
equations by noting that a derivative at constant number of
G⫽ 兺 g i N i ⫽ i⫽1
i⫽1
兺 ␮ iN i . 共11兲

1096 Am. J. Phys. 67 共12兲, December 1999 © 1999 American Association of Physics Teachers 1096
A comparison of Eqs. 共11兲 and 共3兲 shows that a knowledge Lewis–Randall ideal mixtures, even at very low tempera-
of the chemical potential of all the pure substances in the tures. Because the Lewis–Randall ideal mixture is very use-
mixture as a function of P, T, and 兵 N j 其 is equivalent to the ful, it has become customary to refer to it simply as the
characteristic equation. ‘‘ideal mixture.’’
We define a function of mixing X (M ) as the value of X for From the definition of ␮ LR,i in Eq. 共18兲, it follows imme-
the multicomponent system minus the value of the same diately that
property in the pure substances which compose it at the same C C
value of P and T. So, if X is any extensive property, X (M ) is
given by
(M )
G LR ⫽ 兺 共 ␮ LR,i ⫺ ␮ pure,i 兲 N i ⫽NkT i⫽1
i⫽1
兺 n i ln n i , 共19兲

C C C C
X (M ) ⫽ 兺
i⫽1
x iN i⫺ 兺
i⫽1
x pure,i N i ⫽ 兺
i⫽1
共 x i ⫺x pure,i 兲 N i . 共12兲 (M )
S LR ⫽⫺Nk 兺 n i ln n i ,
i⫽1
共20兲

The subscript pure denotes the pure substance at the same (M )

V LR ⫽U LR
(M )
⫽H LR
(M )
⫽0, 共21兲
value of P and T; x is the property X per particle. From its (M )
definition, X (M ) is a function of state of the mixture. Because where N is the total number of particles. The quantity S LR
the mixing function is linear 共it is simply a difference兲, it comes from the term 1/N! in the partition function for the
permutes with other linear operators such as derivatives. pure substance.15
Hence, from Properties of real fluids 共especially, the enthalpy and en-

冉 冊
tropy兲 for given values of T and P often are calculated in
⳵G two stages: the property is first calculated for the substance
V⫽ , 共13兲
⳵P T, 兵 N j 其
using the ideal gas model; then the departure function is
added, where the departure function is the difference be-
it follows immediately that tween the property of the real fluid and the same property if

V (M ) ⫽ 冉 冊 ⳵ G (M )
⳵P T, 兵 N j 其
. 共14兲
the substance were an ideal gas at T and P. Although this
approach may look strange, it has many advantages, because
it can be shown that departure functions may be calculated
from the knowledge of only the P – V – T equation of state.
Similarly, we obtain the important relations
In a way similar to the departure function for a pure sub-
S (M ) ⫽⫺ 冉 冊
⳵ G (M )
⳵T P, 兵 N j 其
, 共15兲
stance, we define the excess function X (E) as the difference
between the value of the function X for the mixture and the
value X LR it would have if the mixture were ideal. From the

U (M ) ⫽⫺T 2 冉 冊 ⳵ 共 G (M ) /T 兲
⳵T P, 兵 N j 其
, 共16兲
definition of the function of mixing and the values obtained
for the ideal mixture, Eqs. 共19兲–共21兲, it follows that
X (E) ⫽X (M ) ⫺X LR
(M )
, 共22兲
H (M )
⫽U (M )
⫹ PV (M )
. 共17兲 C
From the definition of a function of mixing, it is clear that it G (E) ⫽G (M ) ⫺NkT 兺 n i ln n i , 共23兲
accounts for two effects: 共1兲 particles of component i are i⫽1

distinguishable from particles of component j which makes C

the correction 1/N! in the partition function for Maxwell– S (E) ⫽S (M ) ⫹Nk 兺 n i ln n i , 共24兲
Boltzmann statistics inappropriate;11 and 共2兲 the interaction i⫽1

between particles of different components might differ from
the interactions between particles of the same component. A
function of mixing X (M ) can be determined by measuring the
variation of the property X in a closed system in any process
that begins with the pure components at T and P, and ends
with the mixture at T and P. 12
We define the Lewis–Randall 共LR兲 ideal mixture13 as one
in which for every component i,
␮ LR,i ⫽ ␮ pure,i ⫹kT ln n i ,
where k is Boltzmann’s constant, and n i is the molar fraction
共or particle fraction兲 of component i in the mixture.14 This
definition of the Lewis–Randall ideal mixture implies that
the interaction between the particles will be the same as if
they were all the same component. This assumption implies
that the validity of the model is limited to components that
are chemically similar and have similar molecular sizes. For
ideal gas mixtures, for which the interactions between par-
ticles vanish, the model is always valid. That is, ideal gas
mixtures are a special case of the Lewis–Randall ideal mix-
ture. Also, mixtures of isotopes of an element are very nearly
and V (E) ⫽V (M ) , U (E) ⫽U (M ) , and H (E) ⫽H (M ) . The nonzero
values of the excess functions are due to the fact that inter-
actions between particles of different components are differ-
ent from the interactions between particles of the same
component.16 Practical calculations of vapor–liquid and
liquid–liquid equilibria in multicomponent systems are usu-
ally done with the help of models for the excess Gibbs func-
tion, because a very accurate estimation of the liquid behav-
ior is needed for such calculations.17 Excess functions may
共18兲 be calculated by statistical mechanics methods18 or measured
experimentally.19
III. PROBLEMS
Students frequently find it difficult to realize that any ex-
tensive property, for example, the volume of a given mixture
at T and P, should differ generally from the same property
for the pure components at the same T and P.
Problem 1. Needed are a quantity of granular material of
two different sizes 共for example, popcorn and rice兲, and three
graduated vessels, which we label 1, 2, and 3. The volumes

1097 Am. J. Phys., Vol. 67, No. 12, December 1999 Nieto, González, and Herrero 1097
of vessels 1 and 2 should each be less than the volume of
vessel 3. Put a quantity of one material in vessel 1, shake
gently and measure the volume V 1 of the material. Put a
quantity of the other material in vessel 2 and measure its
volume V 2 . Now, place the contents of vessel 1 and vessel 2
into vessel 3. Mix up well and measure the volume V 3 .
Compare V 3 to V 1 ⫹V 2 .
Explanation: the smallest particles tend to fill the spaces
between the largest ones so that the difference V 3 ⫺V 1 ⫺V 2
will be negative. Best results are obtained with particles of
very different sizes which are not too compact. Of course,
molecules are not grains, so they will not behave in such a
simple way. Nevertheless, the volume of mixing is not zero
in general and more frequently is negative.
The name ‘‘function of mixing’’ is misleading, because it
appears to be related to a process of mixing. Of course, under
certain circumstances the variation of a function of state in a
process of mixing may be equal to the function of mixing.
For example, if two pure substances are mixed at constant P
and T, the change in volume will be ⌬V⫽V (M ) . However,
the function of mixing describes the equilibrium state of a
mixture independent of how the mixture was obtained, and
therefore is a function of state of the mixture.
We have tried changing the name in the classroom to
‘‘function of mixture,’’ but this change did not work.20 It
seems that there is a natural tendency to make this mistake.
Problem 2 treats a change of state of a mixture when no
mixing process is specified.
Problem 2. Suppose that the mixture of two liquids 1 and
2 at pressure P has an excess molar Gibbs function given by
冉 冊
g (E) ⫽ a⫹
b
n n ,
T 1 2
共25兲
where a and b are constants, and n 1 and n 2 are the molar
fractions. 关Many mixtures have excess Gibbs free energies
that depend on temperature and composition in a way that
agrees qualitatively with Eq. 共25兲.21 Quantitative agreement
usually requires a far more complicated mathematical form.兴
The constant pressure specific heats of the pure liquids,
c pure,1 and c pure,2 , are constant over the conditions of interest.
A mixture with n 1 ⫽0.5 is heated from T 1 to T 2 at constant
pressure P. Find the change of the mean molar entropy.22
Solution: the excess molar entropy and molar entropy of
mixing are
冉 冊
⳵ g (E) b
s (E) ⫽⫺ ⫽ n n , 共26兲
⳵T P,n 1
T2 1 2
s (M ) ⫽s LR
(M )
共 n 1 兲 ⫹s (E) 共 n 1 ,T 兲 , 共27兲
⌬s⫽n 1 关 s pure,1共 P,T 2 兲 ⫺s pure,1共 P,T 1 兲兴
⫹n 2 关 s pure,2共 P,T 2 兲 ⫺s pure,2共 P,T 1 兲兴 共28兲
⫹s (M ) 共 n 1 ,T 2 兲 ⫺s (M ) 共 n 1 ,T 1 兲 . 共29兲
(M )
Because s LR depends only on composition, which is con-
stant, ⌬s LR ⫽0, and therefore ⌬s (M ) ⫽⌬s (E) . Hence,
(M )
1098 Am. J. Phys., Vol. 67, No. 12, December 1999
T2
⌬s⫽ 共 n 1 c pure,1⫹n 2 c pure,2兲 ln ⫹s (E) 共 n 1 ,T 2 兲 ⫺s (E) 共 n 1 ,T 1 兲
T1
共30兲
冉 冊
1 T2 1 1 1
⫽ 共 c pure,1⫹c pure,2兲 ln ⫹ b 2 ⫺ 2 . 共31兲
2 T1 4 T2 T1
For an ideal mixture, the variation of the enthalpy 共and the
heat exchanged in this system兲 are the same as when the
components were separated, but this identity is no longer
true for real mixtures. The reason is that the interaction be-
tween particles depends on temperature.
The goal of Problem 3 is to clarify which mixing pro-
cesses are directly related to functions of mixing.
Problem 3. At fixed T 0 and P 0 , N molecules of liquid A
occupy a volume V pure,A , and N molecules of liquid B oc-
cupy a volume V pure,B . Liquid A and liquid B are then mixed
adiabatically at constant P 0 . The mixture is found to be at
temperature T 1 ⬍T 0 and occupies a volume V 1 . The mixture
is then heated to T 0 at constant P 0 , and it is observed that
the mixture absorbs a heat Q. The volume of the mixture at
the end of this process is V 2 . Find H (M ) , V (M ) , U (M ) , H (E) ,
V (E) , and U (E) for this mixture.
Solution: From the definition of the function of mixing,
V (M ) ⫽V 2 ⫺V pure,A ⫺V pure,B . The process is isobaric, and
therefore ⌬H 0→2 ⫽Q, and because ⌬H 0→2 ⫽H (M ) ⫽Q,
U (M ) ⫽H (M ) ⫺ P 0 V (M ) ⫽Q⫺ P 0 共 V 2 ⫺V pure,A ⫺V pure,B 兲 . 共32兲
From the definition of the excess functions,
H (E) ⫽H (M ) ⫺H LR
(M )
⫽H (M ) ⫽Q,
V (E) ⫽V (M ) ⫺V LR
(M )
⫽V (M ) ⫽V 2 ⫺V pure,A ⫺V pure,B ,
U (E) ⫽H (E) ⫺ P 0 V (E) ⫽Q⫺ P 0 共 V 2 ⫺V pure,A ⫺V pure,B 兲 .
This problem is a direct application of the definitions, and we
have found that all students 共at least, those who have paid
attention兲 obtain the correct answer after a short time. But
before we began to use this problem in our classes, we found
that many students confused a function of mixing with the
change of a function in any process of mixing.
Frequently, students forget that there are many relations
between thermodynamic functions. Also, students often link
functions of mixing to the isothermal, isobaric mixing pro-
cess, instead of remembering that they are just functions of
state. Problem 4 illustrates these points. We will use the
abbreviation Ẋ⫽dX/dt, where X is any variable. 共This no-
tation is used routinely in the analysis of open systems in
steady flow.兲
Problem 4. A stream of Ṅ 1 ⫽1 mol/s of pure component 1
at T 1 ⫽50 °C, and another stream Ṅ 2 ⫽2 mol/s of pure com-
ponent 2 at T 2 ⫽70 °C, both at pressure P, are added to an
adiabatic vessel. The mixture is continuously extracted at
pressure P and temperature T 3 . At pressure P, it is known
that
g (E) ⫽1000n 1 n 2 J/mol, 共33兲
and the pure components have constant pressure specific
heats c pure,1⫽50 J/mol K and c pure,2⫽70 J/mol K in the range
of conditions considered. Find T 3 and the entropy generation
per second ⌬Ṡ in this process.
Solution: From Eq. 共5兲, it follows that
Nieto, González, and Herrero 1098
 s (E) ⫽⫺ 冉 冊
⳵ g (E)
⳵T P,n 1
⫽0, 共34兲
the absolute value of h (E) is greater for a negligible amount
of water diluted in acid, the temperature increase will also be
greater. In fact, when water is added to pure sulfuric acid, the
which implies that h (E) ⫽g (E) ⫽h (M ) . We also have resulting temperature rise under the usual circumstances will
2 cause local boiling or sputtering. When the acid is added to
s (M ) ⫽s (E) ⫹s LR
(M )
⫽s LR
(M )
⫽⫺R 兺 n i ln n i .
i⫽1
共35兲
the water, there will be a smaller increase in temperature, so
adding the acid slowly is enough to prevent boiling.

Applying the first law for the open system (Q̇⫽Ẇ⫽0), we IV. SUMMARY
have The study of the properties of mixtures can be done with
Ḣ 1 ⫹Ḣ 2 ⫽Ḣ 3 . 共36兲 not much effort by using the functions of mixing and excess
functions. As with other concepts of thermodynamics, the
It follows that standard terminology, adopted for historical reasons, is
Ṅ 1 h pure,1共 P,T 1 兲 ⫹Ṅ 2 h pure,2共 P,T 2 兲 somewhat misleading and contributes to misconceptions.

⫽Ṅ 1 h pure,1共 P,T 3 兲 ⫹Ṅ 2 h pure,2共 P,T 3 兲 ⫹Ḣ (E) 共37兲

1
M. Modell and R. C. Reid, Thermodynamics and its Applications
3 ,
共Prentice–Hall, New Jersey, 1983兲, 2nd ed., pp. 98–115, 175–213.
2
where the subscript 0 means pure component as before. H. C. Van Ness and M. M. Abbott, Classical Thermodynamics of Non-
electrolyte Solutions 共McGraw–Hill, New York, 1982兲.
Ṅ 1 c pure,1T 1 ⫹Ṅ 2 c pure,2T 2 ⫺1000 J/mol Ṅ 1 Ṅ 2 / 共 Ṅ 1 ⫹Ṅ 2 兲 3
G. S. Rushbrooke, Introduction to Statistical Mechanics 共Clarendon, Ox-
T 3⫽ ford, 1949兲, pp. 169–185 and 218–234.
Ṅ 1 c pure,1⫹Ṅ 2 c pure,2 4
J. S. Rowlinson and F. L. Swinton, Liquids and Liquids Mixtures 共Butter-
worths, London, 1982兲, 3rd ed., pp. 87–129.
⫽61.23 °C. 共38兲 5
M. L. McGlashan, Chemical Thermodynamics 共Academic, London, 1979兲,
pp. 139–156. See pp. 240–292 for an interesting discussion on experimen-
From the second law for open systems, the entropy genera- tal methods.
tion per unit time is 6
G. H. Wannier, Statistical Physics 共Wiley, New York, 1966兲, pp. 135–
138.
⌬Ṡ⫽Ṡ 3 ⫺Ṡ 2 ⫺Ṡ 1 , 共39兲 7
See for example, F. Reif, Fundamentals of Statistical and Thermal Physics
and therefore 共McGraw–Hill, New York, 1965兲, Sec. 8.7.

冉 冊
T3 T3 1 2
⌬Ṡ⫽Ṅ 1 c pure,1 ln ⫹Ṅ 2 c pure,2 ln ⫺R Ṅ 1 ln ⫹Ṅ 2 ln
T1 T2 3 3
⫽13.96 W/K.
Sometimes students ask why they should care about func-
tions of mixing. Problem 5 gives an example of their impor-
tance.
Problem 5. In the chemistry laboratory, you are asked to
prepare a dilute solution of sulfuric acid in water. You have
two glasses with the needed quantities of pure sulfuric acid
and water. You know that the excess specific enthalpy for the
sulfuric acid–water mixture is always negative, with a value
of ⫺730 J/g for sulfuric acid infinitely dilute in water, and
⫺2250 J/g for water infinitely dilute in sulfuric acid, both at
room pressure and temperature. How should you mix the
liquids? 共Do not try to do so until you are sure of the answer,
or you risk getting burned.兲
Solution: You should add the sulfuric acid very slowly to
the water 共not the water to the acid兲. The process will be
isobaric at room pressure P 0 , so
8
J. Kestin and J. R. Dorfman, A Course in Statistical Thermodynamics
共Academic, New York, 1971兲, p. 204.
9
The chemical potential per particle ␮ i is related to the chemical potential
per mole, ␮ i,mole by ␮ i ⫽ ␮ i,mole /N a , where N a is Avogadro’s number.
共40兲 Some textbooks use the notation ␮ i for the chemical potential per mole.
10
Standard textbooks on classical thermodynamics use the term ‘‘partial
molar property,’’ but we are dealing with partial derivatives with respect
to the number of particles, not the number of moles. Ref. 3, p. 193 uses the
term partial property.
11
See for example, Ref. 7, Section 7.3, Ref. 6, pp. 167–170, and Ref. 3, pp.
169–187.
12
See for example, Ref. 5, pp. 34–35, 69–71.
13
Henry’s model is sometimes considered to be ideal and is useful for very
dilute solutions. A discussion of this model may be found in Ref. 6, pp.
366–383. We do not consider this model because it is more difficult than
the Lewis–Randall model, and is of interest mainly to chemical engineers
14
and chemists.
This definition is the same as that used in Ref. 4, p. 93.
15
The entropy of mixing of a mixture of ideal gases is calculated from
statistical thermodynamics in for example, Ref. 8, pp. 263–268, or Ref. 5,
p. 227.
16
Mixtures with H (E) ⫽0 are called athermal mixtures, and those with S (E)
⫽0, regular mixtures. Most real mixtures are neither athermal nor regular.
A study of regular solutions can be found in Ref. 3, pp. 287–309.
17

冕
W⫽ P 0 dV⫽ P 0 ⌬V⫽⌬ 共 PV 兲 ,
Q⫽⌬U⫹W⫽⌬U⫹⌬ 共 PV 兲 ⫽⌬H.
Thus, the enthalpy change will be equal to the heat ex-
changed with the surroundings. If the process were also iso-
thermal, then by the definitions of enthalpy of mixing and
excess enthalpy,
⌬H⫽H (M ) ⫽H (E) .
For the process to be isothermal, heat must be removed, be-
cause the excess enthalpy is negative. The real process will
be more nearly adiabatic, because heat flow is a very slow
process, so there will be an increase in temperature. Because
See for example, Ref. 2, pp. 262–400, 424–430, and J. P. Novák, J.
Matouš, and J. Pick, Liquid–Liquid Equilibria 共Elsevier, Amsterdam,
共41兲 1987兲.
18
See for example, Ref. 5, pp. 252–259, or for a more detailed exposition,
共42兲 Ref. 4, Chaps. 7 and 8.
19
See for example Ref. 5, pp. 34–35, 69–71, 242–252, or Ref. 4, pp. 133–
190, which also has many experimental results and references.
20
We use ‘‘function of mixing’’ here because it is standard usage. Ref. 6, pp.
109–110 uses ‘‘entropy of mixture,’’ a far better name.
21
See, for example, R. C. Reid, J. M. Prausnitz, and B. E. Poling, The
Properties of Gases and Liquids 共McGraw–Hill, New York, 1987兲, 4th
共43兲 ed., pp. 251–264.
22
Problems 2–4 have been translated 共with only small changes兲 from R.
Nieto, J. M. Lacalle, M. C. González, J. Honduvilla, A. Teijeiro, F. Her-
rero, and J. Turet, Cuestiones de Termodinámica 共Sı́ntesis, Madrid, 1998兲,
Chap. 7, and were formerly used in final exams of our course on thermo-
dynamics at E.T.S.I.I.M.

1099 Am. J. Phys., Vol. 67, No. 12, December 1999 Nieto, González, and Herrero 1099