School of Physics and Astronomy Junior Honours Thermodynamics

Lecture TOPIC 9 1 (Finn: 5.6, 5.8, 5.9, 5.10) Synopsis: The Second Law of Thermodynamics (review) and the Central Equation of Thermodynamics. Entropy of an ideal gas. Entropy, probability and disorder. The Principle of Increasing Entropy: corollary Imagine the system of interest plus surroundings to be enclosed by an adiabatic wall (eg a wall enclosing some large volume of surroundings providing a practical realisation of a reservoir). Then for the ensemble dq = 0 S 0. The entropy of a thermally isolated ensemble comprising the system plus surroundings can never decrease. This will be used later to examine conditions for equilibrium under different sets of circumstances. The Central Equation of Thermodynamics This combines the rst and second laws for innitesimal reversible processes in differential form: dU = dQ + dW dU = TdS - PdV

GJA 2013-2014

This equation involves only state variables. It can therefore always be integrated to relate changes of state variables independent of the path of integration, even when those changes occur via irreversible processes. It effectively replaces an irreversible path with an equivalent reversible path along which the integrals can be done. Owing to its usefulness and generality it is called the central equation of thermodynamics and is also usefully written: T dS = dU + PdV If applied to a system which is not a simple uid, the equation has to be augmented to allow for other variables. For electric charge (denoted by Z ) driven through a system by an emf E , the equation becomes dU = T dS PdV + E dZ . For a one dimensional system like a rubber band under tension F , changes in length L dominate, so the the equation becomes dU = T dS + F dL. Entropy of an ideal gas For an Ideal gas U = U (T ), so that dU = Cv dT . The central equation for dS, or, more conveniently, for the molar entropy ds then gives: dT dv ds = cv +R T v where the Ideal gas equation has been used to re-express the second term. Integrating the above expression gives: s = cv ln T + R ln v + s0 with s0 the constant of integration. In a Joule, or free expansion of an ideal gas, U does not change, nor therefore does T . However, because v changes, there must be a change in entropy s. For 1 mole of ideal gas, the change in entropy during a free expansion is: s f si = cv ln T f + R ln v f + s0 (cv ln Ti + R ln vi + s0 ) = R ln
1 le

vf vi

topic09.pdf

An example, heating water on a stove Question: Calculate the change in entropy of water heated from 20o C to 100o C on a stove. method 1: as per topic 08. A reversible path is constructed involving a series of heat baths between initial and nal temperatures Ti and T f with the water brought into contact with each bath in turn. The thermodynamic denition of entropy relates changes of entropy to reversible heat transfers, which can be applied along this alternative path between Ti and T f :
Tf

S =
Ti

dqR = T

Tf Ti

Cp dT T

The change of entropy of the water is the same as in the irreversible process that actually occurs since entropy is a state function and the initial and nal equilibrium states match. method 2: Brute force solution applying the central equation:
f

S =
i

dU + T

f i

p dV T

We know only Ti and T f and pi = p f , but not Ui and U f or Vi and V f . Therefore we need to expand dU and dV in terms of {p,T}.
f U T

dT +
p

U p

dp
T

S =
i

+
i

p T

V T

dT +
p

V p

dp
T

pot is open to atmosphere, so the natural choice of path is p = const = pi = p f for which d p = 0.
Tf U T

+p
p

V T

S =
Ti

Tf

(U + pV ) T

dT =
Ti

Tf

H T

dT =
Ti

Tf

dT =
Ti

Cp dT T

method 3: The water is heated at constant pressure so it is likely to be more convenient to work with H=U+PV rather than U, which combined with the central equation gives dH = dU + pdV + V d p = (T dS pdV ) + pdV + V d p = T dS + V d p dH = T dS + V d p The above is an equivalent form of the central equation. We can integrate this choosing the path p =const to get dH dS = p=const p=const T Writing H = H (T , P) dH = dH | p = Cp dT S =
Tf Ti H T

dT +
p

H p

d p = Cp dT +

H p

d p. Thus along the path p =const

Cp dT T

as found previously.

In methods (2) & (3) we have not to had to think explicitly about an equivalent reversible path for the irreversible process as done in method (1). The central equation deals with this implicitly. Another example, electrical work Consider an emf E driving a current I through a resistor R. The power delivered is E I = I 2 R. Suppose the resistor, regarded as the system, completely lls a box with adiabatic walls. The power is dissipated irreversibly in the resistor, which therefore heats up. 2

From the 1st law the change in internal energy of the system in time interval t is U = (q = 0)+ w = I 2 Rt (assuming the volume is xed so that no mechanical work is done). We can relate changes of U to changes of temperature for a constant volume process with dU = CV dT |V const . The change in temperature is then: I 2 Rt =
Tf

CV dT = CV T
Ti

where the last equality follows if CV can be approximated as constant (independent of temperature) T = I 2 Rt /CV . What is the change of entropy? method 2 The change of entropy can be calculated with the central equation dS = dU /T |V const dS = T V CV dT /T |V const S = Ti f C T dT with T f calculated above. method 1 The same answer can be found by thinking of a reversible process leading to the same change of state variables between the initial and nal equilibrium states, in this case a change of temperature T . This could be achieved by putting the resistor successively in thermal contact with a sequence of heat reservoirs, starting with one at Ti and ending with one at T f giving:
Tf Tf

S =
Ti

dqR /T =
Ti

Cv dT /T

ie we have considered an alternative process where heat enters the system reversibly but no work of any kind is done.

The connection between Entropy, Counting, and Probability

a preview of subject matter to be developed in next terms statistical mechanics course The Second Law shows that the total entropy of an isopartition lated system must increase and is maximised at equilibrium. In statistical physics we expect a system to change from a

less probable initial state to a more probable nal state. This V V suggests there is a relationship between probability and engas vacuum tropy. As an example, consider the Joule expansion of a large number of atoms, N, of an ideal gas (see gure). Imbreak mediately after the partition is removed the gas is initially partition ? ? all on the left hand side of the container, whereas in the nal state the gas has an equal density everywhere. Denote the 2V left hand side of the container by A and the right hand side gas everywhere by B with VA = VB = V for simplicity. A long time after the partition has been removed and assuming the system is free to explore all possible states available to it (respecting energy conservation), the probability that any one molecule is in A is 1/2. The chance that all the molecules are in A is (1/2)N . This is the probability of nding the system in its initial state a long time after the wall has been removed. Denote the total number of possible states available with energy E and all the particles in A, to be A (N , VA , E ). Denote the total number of states with energy E, N particles, and volume VA + VB with the particles located anywhere within volume VA + VB as A+B (N , VA+B , E ). Assume that all the accessible microstates are equally likely. Then A (N , V, E ) 1 N = = (1) A+B (N , 2V, E ) 2 Suppose that there is a relationship between the entropy and the number, , of microstates available to the system; S = f () where f is some arbitrary function. If we consider 2 isolated systems we know that the total entropy is the sum of the entropies of each state, S = S1 + S2 , whereas the total number of microstates available grows much faster, as the product 1 2 . Therefore we require that, f (1 2 ) = f (1 ) + f (2 ) (2)

Only a logarithm can achieve this Formally: differentiate EQN 2 rst with respect to 1 and then with respect 2 (eg by changing the volume of system 1 and then the volume of system 2) we have: f (1 2 )2 = f (1 ) (differentiating EQN 2 w.r.t. 1 ) (differentiating EQN 3 w.r.t. 2 ) (3) (4) (5) (6) (7) (8)

f (1 2 )1 2 + f (1 2 ) = 0

= f (x)x + f (x) = 0 (substituting 1 2 = x in EQN 4) d (x f (x)) = = 0 dx kB (kB is an arbitrary constant at this stage) = f (x) = x = f (x) = kB log(x) + const

Thus we see that if S and are related the relationship must be of the form S = kB log() + S0 . Setting the constant S0 to zero denes a natural zero for the entropy; if the state is unique (eg the system is in a quantum ground state) the entropy is zero. As the temperature is increased more states are accessible and the entropy increases (we will return to this when we discuss the 3rd law). Call S dened in this way the statistical entropy: S = kB log() Denition of statistical entropy for xed E (9)

Taking the logarithm of our expression for an Ideal gas, EQN 1, we have: SA+B SA = kB Nlog(2) (10)

SA+B is the statistical entropy of the gas freely occupying the whole volume, which is the nal equilibrium state of the system. SA is the statistical entropy when the gas is restricted to be in volume VA , which is the initial state of the system. The expression is identical to the expression for the change in the thermodynamic entropy (page 2) if we identify kB NA = R (where NA is the number of atoms in a mole). kB is the Boltzmann constant.

General denition of statistical entropy

It is often more convenient to work with systems at constant temperature than at constant energy. The collection of states available is known as a canonical ensemble. A micro-canonical ensemble refers to the collection of states with a xed total energy. It will be proved next term that a more general statistical denition of entropy that is also applicable to the canonical ensemble is given by: S = kB

pi log( pi )

general statistical denition of entropy

(11)

The sum is over all microstates accessible to the system. pi is the probability that the system is in microstate i. If we specialise to the micro-canonical ensemble where the total energy is xed, all the microstates have a-priori equal probability. If there are possible states, pi = 1/ is the probability of the system being in any one of them. Substituting this into the above expression for the statistical entropy we get the expression for the entropy of a micro-canonical ensemble given in EQN 9.

Heat and work in the statistical picture

The mean energy of a system can be written as probability of a given energy that energy: < E >= i pi i where i is the energy of each (quantum mechanical) microstate. The differential of E is, d < E >= d pi i + pi d i
i i

The rst sum on the right hand side can be interpreted as heat supplied to the system and the second sum can be interpreted as the work done. < E > corresponds to the internal energy, U (proof in Stat.Mech.).

Note
THE REST OF THIS TERMS THERMODYNAMICS COURSE IS NOT DEPENDENT ON THE ABOVE STATISTICAL INTERPRETATIONS. THEY ARE GIVEN HERE AS A POINT OF REFERENCE TO HELP DEVELOP A BETTER INTUITION WHEN WORKING WITH ENTROPY. THE RELATIONSHIP BETWEEN STATISTICAL MECHANICS AND THERMODYNAMICS WILL BE DEVELOPED MORE RIGOUROUSLY NEXT TERM.