S. M.

Blinder* 1

Mathematical Methods in
University of Michigan
Ann Arbor
Elementary Thermodynamics

Thermodynamics is the study of the
laws which govern transformations of matter and
energy during physical and chemical changes. These
laws have their most rigorous and concise formulation
in terms of certain specialized mathematical constructs,
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
view some of the more useful results from the theory
of functions of several independent variables.
Consider first a single-valued function z of two in-
dependent variables x and y. The functional relation,
written z z(x,y), expresses the fact that

f(x,y) or 2
= =

including partial derivatives, linear differential forms, for a pair of numbers x, y (perhaps within a restricted
Downloaded via INDIAN INST OF TECH PATNA on April 23, 2020 at 04:53:13 (UTC).

and line integrals. This paper contains a condensed
treatment of some of the mathematical techniques
relevant to the study of thermodynamics. It has been
the writer’s experience in teaching physical chemistry
and thermodynamics courses that prior familiarity
with this mathematics greatly facilitates the student’s
subsequent progress. For then he should be able to
concentrate more on physical principles, unencumbered
range) there exists a unique value for 2. Geometri-
cally, the functional relation may be represented by a
surface in three-dimensional space.4 If one of the in-
dependent variables is constrained to a constant value,
z can be considered a function of the other variable
alone. Partial derivatives can then be defined:
/hf \ ^ lim z(x
+ Ax,y) -

z(x,y)
j
\bx)y /u—o Ax I
by irrelevant mathematical difficulties. In addition, and \ (1)
the student will become more aware of the distinction
/dz\ z(x,y + Ay) -

z(x,y) I
between physical inference and mathematical manipula- .

tion in various steps of a derivation. W* "

a"™0 Ay J

Mathematics provides a compact language for con- The subscripts x or y can be omitted, but only if there

veying quantitative ideas. A mathematical symbol is no ambiguity in what is being kept constant. A par-
can express something which would take many pages to tial derivative, like an ordinary derivative, can be
say in words. By use of mathematics, a train of ab- interpreted geometrically as the instantaneous slope of
stract thought can be broken down into a series of
short manageable steps. Otherwise, a complex logical
sequence might be beyond comprehension.
The stress in our approach is on intuitive conceptual-
ization rather than on mathematical rigor. The
reader is referred elsewhere for more rigorous treatment
of the topics covered.2

Partial Derivatives

Thermodynamic systems are characterized by certain

quantitative physical variables, such as pressure,
volume, temperature, internal energy, and entropy.
Experience has shown that it is not always possible
to vary such quantities at will, but that specification
of some results in definite values for others. Mathe- Figure 1. Graphical representation of partial derivative. The curved
matically, such a situation is expressed by saying that surface represents z = f(x,y) in the first quadrant. The horizontally and
functional relations exist among the variables. The vertically cross-hatched planes are y constant, and x = constant,
—

respectively. Lines ab and cd are drawn tangent to the intersection curves;

methods of partial differentiation then become es- (dz/dx)y is then equal to the slope of ob and (dz/dy)x is equal to the
pecially appropriate.3 In this first section we shall re- slope of cd.

a curve (Fig. 1). Thus (dz/dx) represents the slope of y

the curve cut from the surface z f(x,y) by a plane y = =

constant; similarly (dz/dy)x represents the slope in a

Adapted from Chapter 1 of a physical chemistry textbook to be plane x constant.
=

published by Macmillan Co., New York. Partial derivatives are evaluated by the rules for
1
Guggenheim Fellow, 1965-66, at University College, London.
See, for example, Widder, D. V., “Advanced Calculus,”
2

Prentice-Hall Inc., Englewood Cliffs, N. J., 1947.

3
See any book on elementary calculus, for example, Granville,
W. A., Smith, P. E.. and Longley, W. R., “Elements of Differ- 1
For example, z ax + by
= is represented by a plane and 2 =

ential and Integral Calculus,” Ginn and Co., Boston, Mass., 1957. (x1 + y2)1'2 by a sphere.
Volume 43, Number 2, February 1966 / 85
ordinary differentiation, treating the appropriate vari- + (12)
ables as constants. For examples, the pressure P of one
mole of an ideal gas is given by Extension of the above theory to functions of more
P RT/V (2) = than two variables is straightforward. For a function
where V is the volume, T is the absolute temperature,
u of r variables, u f (xj, x2, ..., xr), there are r first
=

partial derivatives <Ju/dxh du/dx2, ..., du/dxr. The

and R is a universal constant. Treating P as the total differential is given by
dependent variable, we have
(£>P/dF)r -RT/V* (3) = du =

&copy;dxi
+
(£) ***+
(ibP/bT)v R/V (4) =

+ (13)
Since the partial derivatives are also functions of the &copy;

independent variables, they may themselves be differ- Other properties have obvious analogies and need not
entiated to yield second (and higher) derivatives.
be remarked upon further.
These are written
Several useful relationships among partial derivatives
(Fz _
J> /dz\
°r
rd /dz \ "I can be obtained from (12). Suppose, for example, that
dx2 dx \dx/ Ldx \dx/jJ„ z(x>y) constant. Then dz
=
0 and =

(5)
d fbz\ fd /dz\ "I

(IWd),*-0
d2z
01 6 C‘
bybx
~~

by \dx/ Ldi/ \dx)„_L’ (14)

When the function and its derivatives are single-valued Rearranging we find
and continuous, the order of differentiation in the mixed
derivatives is immaterial and dy _
_
(bz/bx)v
dx (bz/by)x
d2z b/z_
=
(6 But the ratio of dy to dx means, in this instance, (d,?//-
bxby bybx
dx)2, sincez is constrained to a constant value. Thus
This is always the case in thermodynamic applications. we arrive at the important identity
For the above example.
(d2P/dF2)r -2 RT/V3 (7)
/a= (bz/bx)v
=
\bx), (bz/by)t
(d2P/dP2)r =
0 (8)
This relationship often enables us to evaluate physical
/b(bP/bV)T\ (b(bP/bT)v\ R
_

quantities not directly obtainable. To illustrate,

^ bT )v \ bV )T F2
/bT\ (bP/bV) T RT/V2 T P
Products of partial derivatives can be manipulated in
the same way as products of ordinary derivatives, pro-
W/p (bP/bT)v R/V V R

vided that the same variables are held fixed. Two use- which agrees with (6T/'dV)P calculated directly, after
ful identities are solving the ideal gas equation for T. Making use of
eqn. (11), eqn. (15) can be rearranged to
mm. -

m. bx
(17)
and &copy;,(!),( bz )v
In the preceding discussion, the relationship among
(S).(g).-1 -
(g).
=

<sk <">
the three variables x, y, z has been given in explicit
Thus far we have considered changes in z(x,y) brought form, i.e., z f(x,y). Solving for x or y, one could
=

about by changing only one of the independent varia- alternatively write x g(y,z) or y h{z,x). Thus the
= =

bles at a time. The more general case involves simul- ideal gas equation could be written in any of three
taneous variation of x and y. This could be represented ways:
by the slope of the surface z f(x,y) along a direction =
RT
P
RT
T
PV
V (18)
not, in general, parallel to either coordinate axis. For a P’
’
V R
function of a single independent variable y /(x), the =

increment in y brought about by an infinitesimal change In many applications it is more desirable to express
the functional relationship among x, y, and z in a more
in x is given by dy (dy/dx)dx. When 2 =
f(x,y), the =

increment in z wrought by simultaneous infinitesimal symmetrical fashion. We accordingly write F(x,y,z)

changes in x and y is given by the total differential5
11
(for example, PV RT 0). In this way, the
—
—

choice of dependent and independent variables is left

6
This can be shown as follows. Let free. One might choose x as the dependent variable
Az z(x + Ax, y + Ay)
=
z(x, y) -

in one application and y in another, without needing

Adding and subtracting the quantity z(x, y + Ay) and inserting to reformulate the functional relationship. One would
the factors Ax/Ax and Ay/Ay, we have then say, for example, that z is an implicit function of x
f z(x + A x, y + Ay) z(x, y + Aj/)-| —

and y through the relation F(x,y,z)

|_--- 0.
, =
AZ =
Ax +
J Implicit functions can be applied to the evaluation
A y) z(x, y + -

of derivatives. Consider the total differential of F(x,y,

Ay
j~z(x, )/)j z) using eqn. (13).
Passing to the limit Ax —
0, Ay —*
0, the two bracketed quantities
become partial derivatives by eqn. (1), while the increments
dF dy + dz —
0 (19)
Ax, Ay, Az become differentials dx, dy, dz.

86 / Journal of Chemical Education

This vanishes since F is a constant ( 0), thus dF 0. = —
The following relations among the alternative in-
Suppose now that we require (dy/dx)z. Since z is being dependent variables are easily proved
held constant, we may drop the term in dz. Thus

dy = 0 (20)
@).&copy;.
+
®).®).-1
(£).(s).-<-(*).(£).-» <33>

Another useful formula involves a “mixed” pair of

(dy\ (bF/bx)v,1
(21)
independent variables, wherein we take z
_

\bx), (bF/by)s,x h(x,u). =

This method is useful when it is impossible or in- Dividing eqn. (12) by dx at constant u, we obtain1
convenient to solve the implicit relation for y or x.
To illustrate, suppose we wish to evaluate (dV/dT)P (!). (!).+®).(2).
=
<33>

for a gas obeying Dieterici’s equation of state or applying eqn. (15)

P(V b)e“IRTV RT (22) =
/bz\
(bz\ (bu/bx)y
—

(bz\ _
_
,

where a and b are two additional empirical constants. \d*/„ \ds/a \by) x(bu/by)x
Eqn. (22) cannot be solved in closed form for either Thus if the internal energy E is given as a function of T
V or T. But defining and V, (dE/bT) P can easily be evaluated from
F(P,V,T) = P(V -

b)e‘IRTV -

RT = 0 (23)
we have*5 (I),=(iX+Qf MS), <»>

We conclude this section with a brief discussion of

(bF/bT)v,P R
P%^r-a e°IRTV (24)
= - -

homogeneous functions. A function f(x\, ., x,) is . .

said to be “homogeneous of degree N” if

(bF/bV)T.P = P
[l -

a(^r~2&)]
e«'Rry (25)
/(X*t, ..\xr) =
\Nf(Xl, ..xr) (36)
Whence by (21), after simplification using (22),
where X arbitrary number. Thus/(x,y) x2 + y2,
is an =

(dV\
\bTjp
__
(bF/bT)v,P
\
a
/( RT
\
TV) l VT B
a_\
V1)
f(x,y) =
(x2 + y2)^1, and f(x,y) = y/x are homogeneous
(bF/bV)T.P -

of degrees 2, 1, and 0, respectively. Euler’s theorem

(26)
on homogeneous functions states that
In most thermodynamic applications, there are alter-
native choices of the independent variables even if the Y m *r) (37)
17
dependent variable is fixed. We next consider some
formulas arising from transformation of variables. when f(xi, ..., xr) is a homogeneous function of degree
Suppose z f(x,y) but x
=
x(u,v) and y y(u,v). =
—

N. To prove (37), we define a new set of independent

If eqn. (12) is divided by du and constant v is speci- variables
fied, we obtain the “chain rule” for two independent
variables Ui \n, i -
1. . . r.

Next we differentiate each side of (36) with respect to

(s)." (a).(S).+ (!),(!), ™ X. The left-hand side becomes, using the chain rule
(29),
Similarly,
bf_ y bj/bUi
(£).=(! ).&copy;).+(S).(8). ™ bX
=

bui bX
(38)

For a function of r independent variables u(xi, x2, while the right-hand side gives simply
xr), the chain rule generalizes to
bu _
bu bXi
(20)
^ =
XkH-'fix,, ..
., xr) (30)

btj bXi btj

Equating (38) and (39), noting that du,/b\ xs, and =

setting X 1, we obtain Euler’s theorem, eqn. (37).

where the tj, j r, are an alternative set of in-
=
1. =
. .

dependent variables. It is seen that z could equally Homogeneous functions of degree 0 and 1 are impor-
well be regarded as a function of the new variables u and tant in thermodynamics, especially in the study of
v, namely z g(u,v). The total differential can be ex-
= multicomponent systems. Corresponding to the vari-
pressed as ables xt are usually the component mole numbers nt.
Thermodynamic quantities homogeneous of degree
*= + m 0 are known as intensive variables. These do not
(M),du &copy;„dv
change in value when the mole numbers are increased
after writing the total differentials dx and dy in terms in proportion, i.e.,
of du and dv, and eliminating among eqns. (12), (27),
and (28). /(Xrei, • •
X?ir) =
f(nh . .
nr) (40)

Examples of intensive variables are temperature, pres-

6
This method of evaluating derivatives is entirely equivalent, sure, density, viscosity, and index of refraction.
in tiiis instance, to differentiating eqn. (23) with respect to T
at constant P, then solving for (bV/bT)P, This, in effect, re-
traces the steps in the derivation of eqn. (21). 1
Note the importance of the subscripts here.

Volume 43, Number 2, February 1 966 / 87

 Thermodynamic quantities homogeneous of degree 1
are known as extensive variables. These have values dF-(lf),4- +
(!),* m
proportional to the amount of matter in the system, But since this differential expansion is unique, eqns.
i.e., (46) and (47) must be equal and we can identify
g(\nu ..., \nr) =
Xg(ni, ...,nr) (41)
«*»> -(£),• "->»>-(I). (4S)
Examples of extensive variables are mass, volume, in-
ternal energy, entropy, and, of course, the mole num- Furthermore,
bers themselves. The ratio of two extensive variables
dX d^F_ OF d*F_
is again intensive, for if (
_ _

dy dydx’ dx dxdy
nr)
gi(ni..., nr) (42) but as discussed earlier, mixed second derivatives of
k(ni, ...,
=

gt(ni..., »r) well-behaved functions are independent of the order of

then differentiation [eqn. (6)]. This leads to Euler’s
gi(\n„ ..., \nr) Xgl(»l, .
, Tlr) reciprocity relation
h(\ni, ’ '
nr)
gz(Xrai, • •
Xwr) Xgrj(a„ . .
.,
=
h(ni, .
.n,) (43) (if), (if). -

Thus density (mass/volume) is intensive, as are various
molar quantities, e.g., molar volume (volume/number
of moles).
Euler’s theorem is useful in relating extensive
thermodynamic variables to the corresponding partial
molal properties. To illustrate, the volume of a mix-
ture of two liquids at a fixed temperature and pressure
depends on the mole numbers ni and n2, V V(n\,n2).
By Euler’s theorem, eqn. (37), with N 1,
as a necessary condition for exactness. It is also a suffi-
cient condition and can thus be used as a test for exact-
ness. That is to say, if eqn. (50) applies to X(x,y) and
Y(x,y), then eqn. (45) is an exact differential and there
exists some function F(x,y)9 which satisfies eqn. (46).
Note however that the reciprocity criterion neither
requires nor identifies the function F(x,y).
=
To illustrate, let us test the differential expression
=

y dx Y x dy (51)
dV dF V
V— »1 +,
V— «2 (44) for exactness. Here X(x,y) y and Y(x,y) x, and = =
drii dn-t
bX/by bY/bx=
1. Thus by eqn. (50), (51) must
=

The quantities bV/bni and dV/dn2 (more precisely represent the total differential of some function of x
(dV/6n1)T,p,„!! and (bV/bn2)r.p,n-) are_known as partial and y. The latter is easily identified by inspection as
molal volumes (usually abbreviated Vi and ¥ >) The
F(x,y) xy since d{xy)
=
y dx + x dy. Again, con-
=

first of these represents, for example, the change in sider the differential
volume per mole when a small amount of component 1 is
added to a mixture of n, moles of component 1 plus n2 ~{RT/P*)dP + (R/P)dT (52)
moles of component 2. Other extensive variables like- The reciprocity condition is again fulfilled since
wise have partial molal analogs, the partial molal free
energy (7; being especially important in the derivation
of thermodynamic principles.
&(-5)],-D&
and eqn. (52) is found to represent the total derivative
Differential Expressions of V(P,T) RT/P. =

Differential quantities of the type Two differential expressions for which the reciprocity
test fails are
dq{x,y) =
X(x,y)dx + Y{x,y)dy (45)
y dx —
x dy (54)
known as Pfaff differential expressions, are of central
and
importance in thermodynamics. Two cases are to be (RT/P)dP -

RdT (55)
distinguished: (1) in which there exists some function
F(x,y) for which eqn. (45) is the total differential, and These are consequently inexact differentials. There
(2) in which there exists no function of x and y which exist no functions which have total differentials (54) or
yields (45) upon differentiation. In case (1), dq is said (55).
to be an exact differential (or complete differential or An inexact differential expression X dx + Y dy
(bX/by ^ bY/bx) can be converted into an exact one
perfect differential) and we can write
by means of an integrating factor M(x,y). In that case
dF(x,y) =
X(x,y)dx + Y(x,y)dy (46)
M(X dx + Y dy) becomes exact, i.e.,
In case (2), dq is called an inexact (incomplete, im-
d(MX) d{MY)
perfect) differential.® dy
=

“dS- (°6)
A differential expression can be tested for exactness
without explicit reference to the function F(x,y). Con- For example, (54) can be converted into an exact differ-
sider first the case in which eqn. (45) is exact and F(x,y) ential by choosing M(x,y) l/:c2. It is easily verified
=

is known. We can then write the total differential

8
As a proof of sufficiency, the function
8
Some authors use the notation dq or Dq when the differential F(x,y) fX(x,y)dx + fY{x,y)dy f fd{x/dy)dxdy
= —

is inexact but we shall not make this distinction. satisfies eqn. (48) provided that eqn. (50) holds.

88 / Journal of Chemical Education

that The solutions (63) are here expressed as implicit func-
, y dx —

x dy y dx 1 tional relations among the xt. Each solution can be rep-

-dy
,

dq = -—-
= - -

(57)
resented by a surface in ?’-dimensional space. Cara-
satisfies (50) and that it is the total differential of thcodory’s principle is now stated without proof:10
f(x,y) =
—y/x. Alternatively, the integrating factor Let/(xi, . .
Xr) = constant be one of a family of solutions to
1 /y2 converts (54) to d(x/y). Likewise, multiplication
by P turns (55) into the exact differential (52). An
— the differential equation
*==1
^ Xtdxt = 0 and let each solution be

integrating factor can always be found for an inexact represented by a surface. If there exist points P'{xY, xr’) ..

differential expression in two independent variables. and P"(xi", ., xr") in r-dimensional space which cannot be con-
. .

Evidently, the choice of M{x,y) is not unique. In fact, nected by some one of these surfaces, then the differential expres-
M(x,y) times any function of F(x,y) is also an integrat- integrating factor.
sion dQ —

ing factor.
The concept of exact differential is especially impor-
tant in the formulation of the first and second laws of
thermodynamics. Thus, the first law postulates that
dE =
dQ + dW is an exact differential even though dQ,
the increment of heat gained by a system, and dW, the
increment of work done on a system, are individually in-
exact. This constitutes a definition of the internal
energy E. The second law postulates that l/T, the re-
ciprocal of the absolute temperature, is an integrating
factor for dQ, Thus, dS
dQ/T is exact, which de-
^2 Xidxt possesses an
In the thermodynamic application of Carath^odory’s
principle, dQ represents an increment of heat gained by a
thermodynamic system, each point P', P”, etc., repre-
sents a possible state of the system, and each surface
represents a manifold of states accessible one from
another by adiabatic processes (those in which no heat
is transferred). The experimental fact that there exist
thermodynamic states inaccessible by adiabatic means
from a given state thus implies the existence of an
=

fines the entropy S. integrating factor for dQ. This leads to a definition of
Many of the chemical applications of thermodynamics entropy and to a compact statement of the second law.
involve more than two independent variables. The To conclude this section, we describe a method for
linear differential expression in r independent variables modifying a differential expression in order to change its
is written independent variables. Such transformations are of
considerable importance in chemical thermodynamics.
Consider a differential
dq{xj, ...,Xr)= J] Xi(xi, . .
xr)dxi (58)
>‘= 1
df(x,y) =
X(x,y)dx + Y(x,y)dy (64)
where each of the functions X( depends on some or all
r and define a function
of the xi. Recalling that the total differential of a
g f ~
Xx (65)
function F(xi, ..., xr) of these r independent variables
—

is given by The differential of g is given by

dg —

df —

Xdx —

xdX (66)
(59)
>•=i °x‘ Substituting (64) for df we obtain
dg
—

Xdx 4- Ydy —
Xdx —

xdX (67)
we obtain, when (58) is exact,
The first and third terms cancel to give
dF
X< = 1"‘r (60)
§? dg =
Ydy -
xdX (68)

From (GO) wc obtain, in analogy with (50), r(r l)/2 The differential (68) is appropriate for a function g g =

reciprocity relations (one for each pair i,j) (y,X) dependent upon y and X as independent variables.
Analogously one could define
dXi dXj 1.. .r
. .

dx j (61) h=f Yy (69)

dxt
-

Conversely, the reciprocity conditions (61) constitute a and obtain

test for exactness. dh —

Xdx —

ydY (70)
As in the ease of two independent variables, an in-
tegrating factor M(xh ..xT) can convert an inexact showing that h h(x,Y). It is also possible to carry
=

differential expression into an exact one. In contrast out a transformation in which both independent vari-
to the former case, however, an integrating factor does ables are changed, for
not always exist. A criterion for the existence of an =/ m —

Xx —

Yy (71)
integrating factor for eqn. (58) forms the basis of has a total differential
Caratheodory’s formulation of the second law of thermo-
dynamics. Before stating this important principle, du =
—xdX ydY (72)
—

we observe that the first order partial differential equa-

tion showing that u u(X,Y). The above are examples of
=

Legendre transformations, in which one or more of the

functions Xt in eqn. (62) replace the corresponding xt
dq =
x*d i = 0 (62)
=
» 1

(known as a Pfaff differential equation) possesses a

10
For a proof, see Mahgenau, H., and Murphy, G. M., “Math-
family of solutions of the form ematics of Physics and Chemistry,” If. Van Nostrand & Co.,
f(xi, ..., xT) =
constant (63) Inc., Princeton, N. J., 1943, sect. 2.18.
Volume 43, Number 2, February 1966 / 89
as independent variables.11 The prescription for these line integrals we shall review briefly the theory of
transformations, as exemplified by (65), (69), and (72), ordinary (Riemann) integration for functions of a single
is to define a new function g by subtracting from the independent variable. The definite integral of a con-
original function / products of the original and the new tinuous function f(x) is defined by the limit:
independent variables. r*b n
The most important thermodynamic application of I f(x)dx = lim lim ^ f(xi)Ax{ (80)
Legendre transformations is based upon Clausius’s dif- J a «-> » all i=\
ferential relation, where Axt The geo-
=
xt-i, x<) a, and xn b. = =
—

dE =
TdS -

PdV (73) metrical interpretation of (80), of an integral as an

which is actually a compact statement of the com- area, is illustrated in Figure 2.
Another fundamental property of the definite in-
bined first and second laws of thermodynamics. The
internal energy E is evidently a natural function of S tegral, connecting the differential and integral calculus
is given by
and V. If a natural function of S and P "were required,
one would define the enthalpy
(81)
H E + PV (74)=

for then where

dH = dE + PdV + VdP = TdS + VdP (75)
(82)
On the other hand, transformation to a function of T
and V would be accomplished by defining the Helmholtz This follows easily when the definition of the derivative
function F(xj) —

F(xs-1)
lim f(xt) (83)
A E TS Ajcf—*0 A Xi
= —

(76)

giving is substituted into (80). We have then

dA = dE -

TdS -

SdT = -SdT -

PdV (77)
E IVixd -

Ffe-dl =
F(b) -
F(a) (84)
The remaining choice of independent variables, T and P,
leads to the Gibbs function or free energy which establishes (81). Also, using (82) in (81).
G E + PV TS =
H TS = A + PV (78) *b

f
= - -

dF(x) =
F(b) -

F(a) (85)
with total differential
dG = -SdT -f VdP (79) which expresses the definite integral entirely in terms of
The four functions, E, II, A, and G are known collec- the boundary values of a function.
tively as thermodynamic potentials. General condi- An extension of the concept of integration to the case
tions for chemical equilibrium can be formulated in terms of a function of more than one independent variable
of them. involves continuous summation of a differential ex-
pression dq{x,y) along a specified curve. For the case of
Line Integrals two independent variables, we define a line integral as
Finite quantities known as line integrals (or follows (Fig. 3)
curvilinear integrals) derive from differential expressions
by integration along curves in space. Before taking up \X(x,y)dx + Y(x,y)dy ]
=

x'y
11
The student familiar with analytical mechanics will recognize
that the Lagrangian and the Hamiltonian functions are connected
lim
n— ® all
lim
A#i—>-0
lim
all Ayi— ()
E lX(.Xi,yi)Axi + Y(xi,yi)&yi] (86)

by a Legendre transformation. The Lagrangian depends upon

coordinates and velocities, and the Hamiltonian on coordinates where all the points xt, y{ lie oil a continuous curve C
and momenta. connecting x', y'(x0,y0) with x", y"(xn,yn)- Possibly the
best known example of a line integral occurs in me-
chanics, where the work done on a particle is defined as

Figure 2. Geometrical representation of the definite integral. The cross-

hatched vertical strip has an area f(x;)Ax;. The sum of the n strips with
r = 0, ..., n —
i approximates the shaded area and approaches it in
the limit all Axj — 0. According to definition (80), the definite
6
integral J* f(x)dx equals the shaded area. Figure 3. The line integral, as defined in equation (86).

90 / Journal of Chemical Education

the line integral of force along a trajectory Along path B we obtain analogously,
Aw fF dr f[Fxdx + Fvdy + Fidz]
= =
(87)
AWb RT0 In §
t o
-

R(To --
2’„) (94)
A line integral, like a Riemann integral, depends on
the sense (direction) of integration. From the defini- To evaluate XWC, we eliminate T using the equation of
tion (86) it follows that the linear patii C
r*xft,y" f*x',y' Ti -

To
dq(x,y) = -
I dq(x,y) (88) (95)
| P0 ~

x',yr CJ x",y"
whence
It should however be noted that, in contrast to a
Riemann integral, a line integral is not represented in
any obvious way by an area.
T ~T- +
ibAfy -
p-> (96)

and
It is readily verified that (86) reduces to a Riemann
To
integral (80) when the path of integration is parallel to dT = dP (97)
Pi -

Po
either coordinate axis. For example, along the linear
path y y0 =
constant, (86) becomes
—
Thus
/*X*
AWc
-/,.']?R+(s^:)(p-ft>]-
=
Aqc X(x,y„)dx (89)
J
dP
(since dy 0). In general, when the curve
= G repre-
sents a functional relationship '

=
RT0 In (98)
y =
gix) (90)

y can be eliminated between (86) and (90) to yield a RjToPo -

7\Pp) P,
Riemann integral Pi -

Po Po

Comparing (93), (94), and (98) we see that

X[x,g(x)] + Y[x,g{x)\ dx
J (91)
=
Age
| ^j- AWa 7^ AWb 7* AWc- (99)

The limits on y are automatically fulfilled since y' =

Thus the line integral depends in general on the path of
cj(x') and y" g{x"). Alternatively one could elim-
= integration.
inate x between (86) and (90) to obtain a Riemann If, on the other hand, we integrate the differential
integral over y. expression (52)
RT
aV dP (100)
J Ta.Po
' r

along the same three paths, we find

B AVA AVB AVc
-^iTi.pi
= = =
-J1 - =
Vo -

Vo (101)
To.pi[- "

A
!
i
[using V from the ideal gas law (2) ]. In fact, eqn. (100)
evaluated along any path between 7V,.P0 and 7V Pi gives
h,po
the same result. The independence of path stems from
the fact that we are integrating an exact differential
(52) [whereas (92) contains the inexact differential (55) ].
Figure 4. Three possible paths between the states Tq,Pq and T\,Pi of a
To prove the above statement, let dq in (86) be the
thermodynamic system. total differential of a function F(x,y). Then using (48)
we can write
AF(xi,yi) -

X(xi,yi)Axi + Y{xi,yi)Ayi (102)

As an illustration, we integrate the differential ex-
where A F(xo,yf) means F(xt,yt) F{xi-i,yi-o). Sub-
pression (55) along the three paths shown in Figure 4.
—

These line integrals stituting (102) in (86) we find

/* X"#*
•Ti.Pi
^
-r Aqc dF{x,y) lira AF(xi,yd
=
AIV =
I dP Ft dT't (92)
i
—

( bJ x',yf i-1
JTo.Po
F{x\y“) F{x’,y’) (103) = -

represent the work done on one mole of an ideal gas in

reversible expansion or compression. Along path A, independent of the path C. Note the correspondence
between eqns. (103) and (85). In both instances, the
(92) reduces to the sum of two Riemann integrals: value of the integral depends only on its end points.
T i,.Po
dp Conversely, it can be shown that, if a line integral is
f AA
AW =

JJ
I RdT[ +
-

I
T(,,Po
independent of path, the differential must be exact.
Of particular importance are line integrals around
rF'\fdp-RdT\=RTl Ip.r
Ti,P\

JTo.Po
1 ’ f
r
closed paths, in which case the initial and final points
may be thought to coincide. For cyclic paths, the
rTi integral sign is usually written <f>, The closed curve
R
J To
dT =a
RTi In y ~

R{T, -

To) (93)
is bjr convention traversed in the counterclockwise

Volume 43, Number 2, February 1966 / 91

 y

V
Figure 6. Representation of work as area.8

X
rFiV i

Figure 5. Line integral around closed path, as defined in equations (105) AWC - -

I PdV = -
I P(V)dV (109)
and (1 06). CJ PoVo J Vo

sense. If dq(x,y) is an exact differential, then where the path C determines P as a function of V.
Since the right hand side is a Itiemann integral it can be
c<f> dq(x,y) «*
cj> | X(x,y)dx + Y(x,y)dy} = 0 (104)
represented by an area (Fig. 6).
To summarize several of the interrelationships dealt
for an arbitrary closed path C. This is easily proved,
with in this section, for a Pfaff differential expression
for dividing C into two segments C' and C" connecting
x',y' with x",y" (Fig. 5), we have dq(x,y) =
X(x,y)dx + Y(x,y)dy
r x*.y" rx'y any of the following statements implies the other two:
cy dq = I dq + I dq (105) (1) There exists a function F{x,y) whose total derivative
C*y ’x*,y' C*J vx".y"
is dq(x,y) (dq is an exact differential); (2) (dX/d;/)* =

Both since both integrals on the right hand side are (dF/dx)j, (Euler’s reciprocity relation); and (3)cjf
independent of path by eqn. (103), and since one is the dq(x,y) 0 for an arbitrary closed path C.
=

negative of the other by eqn. (88), we obtain eqn. (104). Many of the mathematical abstractions encountered
When dq(x,y) is inexact, however, the cyclic integral in this paper have obvious analogs in thermodynamic
is, in general, different from zero. terms. A point in the xy plane can represent the state of
A final connection between differential expressions a thermodynamic system having two degrees of freedom.
and line integrals is provided by Green’s theorem12 The independent variables can be chosen from among P,
V, T, E, S, etc. A function F(x,y) corresponds to
c<j>
[X(x,y)dx + Y{x,y)dy\ =

another thermodynamic variable. It is called a func-

tion of state since it is uniquely determined by the
/£[( §).-(£)>*<“> state of the system. An exact differential represents an
increment in a function of state whereas an inexact
where the right hand side represents a double Itiemann differential represents an increment in a thermodynamic
integral over the area S enclosed by the curve C. If quantity not a function of state. A curve on the xy
Euler’s reciprocity condition (50) holds for dq(x,y),
plane can represent an equilibrium process, that is, a
then the right hand side of (106) vanishes and (104)
sequence of states between some initial and final states.
follows. Conversely if the left hand side of (106) A line integral represents the change in a thermo-
vanishes for arbitrary C, then the integrand of the
dynamic quantity in the course of a process. If that
right hand side must vanish and (50) follows. quantity is a function of state, the line integral is in-
By Green’s theorem, certain cyclic line integrals can dependent of path. This is entirely reasonable since
be represented by areas. To illustrate, a differential the change in a function of state is simply the differ-
element of work done on a system in a reversible proc- ence between its initial and final values. A line integral
ess is given by
around a closed path represents a cyclic process, one
dW = -

PdV (107) which returns the system to its initial state. Clearly,
such a process leaves a function of state unchanged.
using eqn. (106) with P and V as independent variables
(noting that the coefficient of dP is 0), we find, for the
work done in a cyclic reversible process 1S
The work done in traversing the path C' from FoWo to Pi,Vi
is equal to the negative of the area under C’
dPdV
A Wc
c§ PdV (108)
= -
= -

f
'Pi, Vi
A Wc' PdV
The right hand side is clearly equal to the negative of J Pv,Vi
C'J P,

the area S enclosed by the path C. Note that the non- The work done in returning to the original state via path C" is
vanishing of (108) shows dW to be an inexact differ- equal to the area under C.
ential. Actually, since the dP term is lacking in eqn. o nPuV!
(107), AW can be represented by an area even for a non- PdV = PdV
J'Po.V
Pl.Vi C"J Po,T'ro
cyclic process. Thus we may write
The net work done in the cyclic process is given by

For proof see, for example, Kaplan, W., “Advanced Cal-

12 A Wc =
AWc' + A Wc’
culus,” Addison-Wesley Publishing Co., Inc., Reading, Mass., and is equal to the area enclosed by the cyclic path. The last
1952, p. 239ff. result also follows from Green’s theorem (106).

92 / Journal of Chemical Education