3 More applications of derivatives

3.1 Exact & inexact differentials in thermodynamics

So far we have been discussing total or “exact” differentials
µ ¶ µ ¶
∂u ∂u
du = dx + dy, (1)
∂x y ∂y x

but we could imagine a more general situation

du = M (x, y)dx + N (x, y)dy. (2)

¡ ∂u ¢ ³ ´
If the differential is exact, M = ∂x y and N = ∂u ∂y x . By the identity of mixed
partial derivatives, we have
µ ¶ µ 2 ¶ µ ¶
∂M ∂ u ∂N
= = (3)
∂y x ∂x∂y ∂x y

Ex: Ideal gas pV = RT (for 1 mole), take V = V (T, p), so

µ ¶ µ ¶
∂V ∂V R RT
dV = dT + dp = dT − 2 dp (4)
∂T p ∂p T p p

Now the work done in changing the volume of a gas is

RT
dW = pdV = RdT − dp. (5)
p
Let’s calculate the total change in volume and work done in changing the system
between two points A and C in p, T space, along paths AC or ABC.
1. Path AC:
dT T2 − T1 ∆T ∆T
= ≡ so dT = dp (6)
dp p2 − p1 ∆p ∆p
T − T1 ∆T ∆T
& = ⇒ T − T1 = (p − p1 ) (7)
p − p1 ∆p ∆p
so (8)
R ∆T R ∆T R ∆T
dV = dp − 2 [T1 + (p − p1 )]dp = − 2 (T1 − p1 )dp (9)
p ∆p p ∆p p ∆p
R ∆T
dW = − (T1 − p1 )dp (10)
p ∆p

1
 T
(p2,T2)
C

(p,T)

(p1,T1)
A B

Figure 1: Path in p, T plane for thermodynamic process.

Now we can calculate the change in volume and the work done in the process:

Z ¯p
∆T 1 ¯¯ 2 T2 p 1 − T1 p 2
V2 − V1 = dV = R(T1 − p1 ) ¯ = R (11)
AC ∆p p p1 p1 p2
Z ¯p2
∆T ¯ T2 p 1 − T 1 p 2 p 2
W1 − W2 = pdV = R(T1 − p1 ) ln p¯¯ = R ln (12)
AC ∆p p2 − p1 p1 p1

2. Path ABC: Note along AB dT = 0, while along BC dp = 0.

Z µ ¶ µ ¶ Z p2 µ ¶ Z T2 µ ¶
∂V ∂V ∂V ∂V
V2 − V1 = dT + dp = dp + dT
ABC ∂T p ∂p T p1 ∂p T T1 ∂T p
Z p2 Z T2
−RT1 R T 2 p 1 − T1 p 2
= dp + dT = R (13)
p1 p2 T1 p2 p1 p2
Z p2 µ ¶ Z T2 µ ¶
∂V RT ∂V p2
W2 − W1 = p dp + dT = −RT1 ln + R(T2 − T1 )
p1 ∂p T T1 V ∂T p p1
Note the change in V is independent of the path – the volume is characteristic
of a point (p, T ) in equilibrium – but the work done in the process is not! What’s
the difference? In the system with p, T as independent variables, dV is an exact
differential, while dW is not. How can you see the difference? Go back and examine
the forms we had for dV and dW in (4) and (5). In the case of dV , we had
R RT
dV = M dp + N dT, with M = and N = − 2 (14)
p p
∂M R ∂N R
=− 2 = − 2, (15)
∂p p ∂T p
2
which is indeed exact, whereas
RT
dW = M 0 dp + N 0 dT M0 = R ; N0 = − (16)
p
∂M 0 ∂N 0 R
=0 =− , (17)
∂p ∂T p
is not. This is a demonstration (we won’t use the word proof) that for a for a
thermodynamic process involving changes in the p − T plane, the volume of the
system is a ”state variable”, i.e. (for 1 mole of gas) it simply depends on what p and
T are; if you have specified p, T , you know the volume of the system. The change
in volume between two points will therefore always be V2 −V1 independent of which
path is chosen. The work done in the same process is however not independent of
the path of integration.

3.2 Maxima/minima problems with constraints (Lagrange

multipliers)
I’m going to make a mathematical detour before coming back to thermodynamics,
in order to give you some tools you need to solve homework problems. In physics
we often need to find the extrema of a function of several variables subject to a
constraint of some pkind. In math, a simple example would be the distance function
in 3-space, d = x2 + y 2 + z 2 . Of course the minimum of this function over all
3-space is just 0, achieved at x = y = z = 0. But suppose we were to look
for the minimum of d over the ”constrained” set of points defined by the plane
x − 2y − 2z = 3? The usual way to proceed is often the simplest, if it works:
express y = y(x), then set dy/dx = 0 and solve. But sometimes one can’t solve
for y(x) explicitly, so one can try the method of implicit differentials (see Boas ch.
4), or use an elegant technique called the method of Lagrange multipliers (Joseph
Lagrange (1736-1813) was a French mathematical physicist).
The idea is, if you can express the problem in terms of the minimization of a
function f (x, y, . . . ), together with a constraint g(x, y, . . . ) = 0, to imagine you
are solving the unconstrained problem of finding the minimum of F (x, y, . . . ; λ) =
f (x, y, . . . ) + λg(x, y, . . . ). Then you can simply treat λ as an additional variable,
and minimize with respect to it as well. In the process of solving the problem, you
eliminate λ from the solution.
Ex: Coming back to our problem with the plane, let’s first make our life a bit
easier by recalling that if we minimize d0 = x2 + y 2 + z 2 , the square root will also

3
be a minimum. So define

F = x2 + y 2 + z 2 + λ(x − 2y − 2z − 3) (18)

and set all the derivatives equal to zero:

∂F
= 2x + λ = 0 (19)
∂x
∂F
= 2y − 2λ = 0 (20)
∂y
∂F
= 2z − 2λ = 0 (21)
∂z
∂F
= x − 2y − 2z − 3 = 0. (22)
∂λ
Note the equation ∂F/∂λ = 0 is always just the constraint equation itself. Now
we have a problem with 4 equations and 4 unknowns, and it can be solved. The
usual idea is to eliminate the constraint as quickly as possible. The first eqn. tells
us that λ = −2x, so

2y + 4x = 0 (23)
2z + 4x = 0 (24)
x − 2y − 2z − 3 = 0, (25)

which we can solve to find (x, y, z) = (1/3, −2/3, −2/3).

3.3 Differentiation of integrals

Leibnitz’ rule for differentiating integrals:
Z v(x) Z v(x) µ ¶
d dv du ∂f
f (x, t)dt = f (x, v) − f (x, u) + dt. (26)
dx u(x) dx dx u(x) ∂x

Ex :
Z 2x
ext
I = dt (27)
x t
x·2x Z 2x
dI e ex·x
= ·2− ·1+ ext dt (28)
dx 2x x x
2 2 2
= (e2x − ex ) (29)
x

4
3.4 Laws of thermodynamics
Let’s come back to the idea of exact and inexact differentials in thermodynamics.
Here’s another example of an inexact differential:

dQ = cp dT + Λp dp, (30)
|{z} |{z} |{z}
heat absorbed heat capacity const. p latent heat. (31)

We showed already that the work done is also not an exact differential, however
the combination of the two is:

dU ≡ dQ − dW (32)
" µ ¶# · µ ¶ ¸
∂V ∂V
= cp − p dT + Λp − p dp (33)
∂T p ∂p T

This is called the internal energy, and this equation is sometimes referred to as
the 1st law of thermodynamics, expressing energy conservation, i.e. the change
in internal energy of a gas in a thermodynamic cycle goes either into heating the
system (dQ is the infinitesimal heat absorbed by the system is the or into doing
work (done by the system).
Another exact differential is dS = dQ/T .
Remark 1: sometimes you will see the notation dQ ¯ and dW¯ for infinitesimal
heat absorbed and work done. This is just a more careful notation to remind you
that they are inexact differentials.
Remark 2: since the work done in a thermodynamic process depends on the
path, it is really nonsense what I wrote above “W2 − W1 ”. You are to think of this
as the work performed by the system in the process of going 1 → 2, but W2 and W1
have no independent meaning, since they are not characteristic of a macrostate.

3.5 Legendre transformation

If for a function f (x, y) we have the differential df = pdx + qdy, where p and q
are equal to ∂f /∂x and ∂f /∂y, respectively, we might ask the question, how do we
make a simple change of variables to a new function g(x, q) with q one of the the
independent variables? We simply make use of the fact that

d(f − qy) = df − dq y − q dy = p dx + q dy − dq y − q dy = p dx − y dq. (34)

This function is by definition associated with an exact differential dg. You explored
this on the HW for various thermodynamic quantities.