Part 2 · Chapter 20

Entropy and the Second Law of Thermodynamics

Why time runs one way, why a shattered egg never reassembles, why no engine can be perfect, and why air never spontaneously crowds into one corner of the room — all written in a single quantity that only ever grows

Fundamentals of Physics Prof. Mithun Mondal Reading time ≈ 65 min
i What you'll learn
  • Why irreversible processes run one way only, and that their direction is set by the entropy change, not by energy.
  • The definition of entropy change, \(\Delta S = \int_{i}^{f} dQ/T\), that entropy is a state function, and for an ideal gas \(\Delta S = nR\ln(V_{f}/V_{i}) + nC_{V}\ln(T_{f}/T_{i})\).
  • The second law in a closed system: \(\Delta S \ge 0\) (= for reversible, > for irreversible).
  • How a Carnot engine works and its efficiency ceiling \(\varepsilon_{C} = 1 - T_{L}/T_{H}\), with \(\varepsilon = W/|Q_{H}|\) for any engine.
  • Why no perfect engine or perfect refrigerator can exist, and the coefficient of performance \(K_{C} = T_{L}/(T_{H}-T_{L})\).
  • The statistical meaning of entropy through microstates and multiplicity, \(W = N!/(n_{1}!\,n_{2}!)\), and Boltzmann's \(S = k\ln W\).
Section 20-1

What Is Physics?

Time has a direction — the direction in which we age. We live among one-way processes: an egg dropped on the floor, a pizza baked, a car driven into a lamppost, a beach eroded by waves. These are irreversible — they cannot be undone by some small tweak of their surroundings. One goal of physics is to understand why time has a direction and why such processes never run backward.

This question may seem far from practical life, yet it sits at the very heart of every engine, because it sets the ultimate limit on how well an engine can run. The key to it all is a single quantity: entropy.

Section 20-2

Irreversible Processes and Entropy

The one-way character of nature is so familiar we rarely notice it — but it is deeply strange. If you wrapped your hands around a hot mug and your hands grew colder while the coffee grew hotter, you would be astonished. Yet energy would be perfectly conserved either way. Pop a balloon and the helium spreads through the room; it never spontaneously regathers into the balloon's shape — though energy conservation would permit it.

So energy conservation does not pick the direction of an irreversible process. Something else does — the change in entropy \(S\).

The Entropy Postulate
If an irreversible process occurs in a closed system, the entropy S of the system always increases — it never decreases.

Unlike energy, entropy is not conserved. For irreversible processes it grows, which is why the change in entropy is called the arrow of time. An exploding popcorn kernel marks the forward direction of time; the reverse — fragments reassembling into the kernel — would lower entropy, and so it never happens.

Section 20-3

Change in Entropy

We define the entropy change of a system going from state \(i\) to state \(f\) in terms of the heat exchanged and the temperature at which it flows:

Definition of entropy change (and the isothermal special case)
\[ \Delta S = S_{f} - S_{i} = \int_{i}^{f}\frac{dQ}{T} \qquad\qquad \Delta S = \frac{Q}{T}\;\text{(isothermal)} \]
The SI unit is the joule per kelvin (J/K). Because \(T\) is always positive, the sign of \(\Delta S\) follows the sign of \(Q\). When the temperature change is small, \(\Delta S \approx Q/T_{\text{avg}}\).

There is a catch. During an irreversible process like a free expansion, the gas has no well-defined pressure or temperature midway through — there is no path to integrate along. The way out rests on entropy being a state function: the change \(\Delta S\) depends only on the endpoints, not the route. So we replace the messy irreversible process with any convenient reversible process linking the same two states, and compute \(\Delta S\) for that.

🔁
The reversible-replacement strategy
To find ΔS for an irreversible process, replace it with any reversible process between the same initial and final states, and apply ΔS = ∫dQ/T.

For an ideal gas taken reversibly from \((V_{i},T_{i})\) to \((V_{f},T_{f})\), integrating the first law gives a result that depends only on the endpoints:

Entropy change of an ideal gas
\[ \Delta S = nR\ln\frac{V_{f}}{V_{i}} + nC_{V}\ln\frac{T_{f}}{T_{i}} \]
No particular reversible path was assumed in the derivation, so this holds for every path between the two states — direct proof that \(\Delta S\) is a state function. For the free expansion (constant \(T\)), the second term vanishes and \(\Delta S = nR\ln(V_{f}/V_{i})\).
Section 20-4

The Second Law of Thermodynamics

A puzzle: reverse a reversible isothermal expansion and the gas alone loses entropy (\(Q < 0\)). Does that break the entropy postulate? No — that postulate applies only to closed systems undergoing irreversible processes. Include the reservoir, and the reservoir gains exactly the entropy the gas loses, so for the enlarged closed system \(\Delta S = 0\). This widens the postulate into the full law:

📈
The Second Law of Thermodynamics
ΔS ≥ 0  (closed system)

The equals sign applies to reversible processes, the greater-than sign to irreversible ones. Entropy may fall in part of a closed system, but there is always an equal or larger increase elsewhere, so the total never decreases. Real processes — with friction and turbulence — are always irreversible to some degree, so the entropy of a real closed system always rises. Constant-entropy processes are idealizations.

You can feel entropy in a rubber band. Unstretched, its polymer chains are coiled in a disordered, high-entropy tangle. Stretch it and you uncoil and align them, lowering the entropy. Since \(F = -T\,dS/dx\) and \(dS/dx\) is negative, the inward pull you feel is the chains' tendency to return to their more disordered, higher-entropy state. The restoring force of rubber is, quite literally, an entropy force.
Section 20-5

Entropy in the Real World: Engines

A heat engine extracts energy as heat and does useful work. At its core is a working substance (water in a steam engine, a fuel-air mix in a car) that must operate in a cycle, returning to each state again and again. Just as the ideal gas illuminates real gases, an ideal engine illuminates real ones. The benchmark is the Carnot engine — the most efficient engine possible, conceived by Sadi Carnot in 1824, before either the first law or entropy was known.

Each cycle, the Carnot engine absorbs \(|Q_{H}|\) from a hot reservoir at \(T_{H}\), dumps \(|Q_{L}|\) to a cold reservoir at \(T_{L}\), and delivers work \(W\). Its cycle is two isothermal steps (the only steps with heat transfer) joined by two adiabatic steps. On a p–V diagram the enclosed area is the net work; on a temperature–entropy diagram the cycle is a clean rectangle.

First law over a cycle, and the entropy condition
\[ W = |Q_{H}| - |Q_{L}| \qquad\qquad \frac{|Q_{H}|}{T_{H}} = \frac{|Q_{L}|}{T_{L}} \]
Over a complete cycle the working substance returns to its start, so \(\Delta E_{\text{int}} = 0\) and \(\Delta S = 0\). The entropy gained at \(T_H\) must equal the entropy lost at \(T_L\), which forces the ratio above. Since \(T_H > T_L\), necessarily \(|Q_H| > |Q_L|\): more heat is drawn in than dumped out.
Section 20-6

The Efficiency of a Carnot Engine

An engine's thermal efficiency is "what we get" (work) over "what we pay for" (heat absorbed). Combining that with the cycle relations above gives the famous Carnot limit:

⚙️
Efficiency: any engine, and the Carnot ceiling
ε = W / |Q_H|  ·  ε_C = 1 − T_L/T_H

The temperatures are in kelvins. Because \(T_{L} > 0\) and \(T_{H}\) is finite, \(\varepsilon_{C} < 1\) always — no engine can be 100% efficient. A perfect engine (\(|Q_{L}| = 0\), all heat turned to work) would need \(T_{L} = 0\) or \(T_{H} \to \infty\), both impossible.

🚫
Second law (Kelvin–Planck form)
No process can have, as its sole result, the complete conversion of heat from a reservoir into work.

In short: there are no perfect engines. A real car engine that could reach the Carnot limit between its temperatures might hit ~55%; in practice it manages ~25%. A nuclear plant's Carnot ceiling might be ~40%, with ~30% achieved. No design trick beats the limit set by \(\varepsilon_{C}\).

The ceiling is for Carnot engines only. An ideal Stirling engine also runs between \(T_{H}\) and \(T_{L}\), but its isotherms are joined by constant-volume steps rather than adiabats — so heat is exchanged in all four strokes, and its efficiency is lower than a Carnot engine's between the same temperatures. The Carnot value is the absolute best; nothing does better.
Section 20-7

Real Engines and Refrigerators

One can prove no real engine beats a Carnot engine between the same temperatures. Suppose an engine X claimed \(\varepsilon_{X} > \varepsilon_{C}\). Couple it to drive a Carnot refrigerator; the combination would transfer heat from cold to hot with no net work — a perfect refrigerator, which the second law forbids. The only flawed assumption is the claim itself, so \(\varepsilon_{X} \le \varepsilon_{C}\).

A refrigerator runs the cycle in reverse: work \(|W|\) is done to pump heat \(|Q_{L}|\) from cold to hot. Its figure of merit is the coefficient of performance — "what we want" (heat removed) over "what we pay for" (work):

Coefficient of performance (any refrigerator, and Carnot)
\[ K = \frac{|Q_{L}|}{|W|} \qquad\qquad K_{C} = \frac{T_{L}}{T_{H} - T_{L}} \]
Typical room air conditioners have \(K \approx 2.5\); household fridges \(K \approx 5\). Notice \(K\) is larger when the two reservoirs are closer in temperature — which is why heat pumps work best in mild climates and struggle in deep cold.
🧊
Second law (Clausius form)
No process can have, as its sole result, the transfer of heat from a cold reservoir to a hot one.

A perfect refrigerator running with no work input would give the cold reservoir entropy \(-|Q|/T_{L}\) and the hot one \(+|Q|/T_{H}\), for a net \(\Delta S < 0\) — a violation. In short: there are no perfect refrigerators. If you want yours to run, you must plug it in.

Section 20-8

A Statistical View of Entropy

There is a second, deeper definition of entropy — by counting. Imagine \(N\) indistinguishable molecules split between the two halves of a box, \(n_{1}\) on the left and \(n_{2}\) on the right. A configuration is a split like (4, 2); a microstate is one specific arrangement of the actual molecules. The number of microstates in a configuration is its multiplicity:

Multiplicity of a configuration
\[ W = \frac{N!}{n_{1}!\;n_{2}!} \]
For six molecules, the even (3, 3) split has the most microstates (20 of 64), while the all-on-one-side splits have just 1 each. The fundamental assumption: all microstates are equally probable. So the configuration with the most microstates is the one the system spends the most time in.
📊
Boltzmann's entropy equation
S = k ln W

Entropy is the logarithm of the number of microstates, scaled by \(k = 1.38\times10^{-23}\,\mathrm{J/K}\). The logarithm is natural: entropies of two systems add while their probabilities multiply, and \(\ln ab = \ln a + \ln b\). This formula — engraved on Boltzmann's tombstone — and the thermodynamic \(\int dQ/T\) give the same entropy change. For huge factorials, use Stirling's approximation \(\ln N! \approx N\ln N - N\).

Why the air never crowds into one corner. For just 100 molecules, the even 50–50 split outnumbers the all-on-one-side configuration by about \(10^{29}\) to 1; counting those microstates at one per nanosecond would take ~200 times the age of the universe. For a real mole (\(\sim 10^{24}\) molecules) the lopsided configurations are so vanishingly rare that you can breathe easy — the air will never abandon your half of the room. Equilibrium is simply the overwhelmingly most probable arrangement.
Worked Examples

Putting It to Work

1 Two blocks coming to thermal equilibrium

Problem. Two identical copper blocks (\(m = 1.5\,\mathrm{kg}\), \(c = 386\,\mathrm{J/kg\cdot K}\)), one at 60 °C and one at 20 °C, are placed in contact in an insulated box and reach 40 °C. Find the net entropy change.

Solution. Replace the irreversible mixing with two reversible steps, each \(\Delta S = mc\ln(T_{f}/T_{i})\), in kelvins (333 K, 313 K, 293 K).

ΔS = mc ln(T_f/T_i) for each block, then sum
\[\begin{aligned} \Delta S_{L} &= (1.5)(386)\ln\frac{313}{333} \approx -35.86\,\mathrm{J/K} \\ \Delta S_{R} &= (1.5)(386)\ln\frac{313}{293} \approx +38.23\,\mathrm{J/K} \\ \Delta S &= -35.86 + 38.23 \approx +2.4\,\mathrm{J/K} \end{aligned}\]

The hot block loses less entropy than the cold block gains, because the same heat is more "valuable" (lower \(T\)) where it arrives. The net is positive — exactly as the second law demands for an irreversible process.

2 Entropy change of a free expansion

Problem. One mole of nitrogen (ideal) is confined to the left half of a container; the stopcock opens and the volume doubles freely. Find \(\Delta S\).

Solution. A free expansion is irreversible, so substitute a reversible isothermal expansion (same \(T_{i} = T_{f}\)): \(\Delta S = nR\ln(V_{f}/V_{i})\).

ΔS = nR ln(V_f/V_i)
\[ \Delta S = (1.00)(8.31)\ln 2 \approx 5.76\,\mathrm{J/K} \]

Positive, as required. Section 8 derives the very same \(nR\ln 2\) by counting microstates with \(S = k\ln W\) — two utterly different routes to one answer.

3 A Carnot engine: efficiency, power, entropy

Problem. A Carnot engine runs between 850 K and 300 K, doing 1200 J of work per 0.25 s cycle. Find (a) efficiency, (b) power, (c) \(|Q_{H}|\), (d) \(|Q_{L}|\), and (e) the entropy changes of the working substance.

Solution. Use \(\varepsilon_{C} = 1 - T_{L}/T_{H}\), then \(P = W/t\), \(|Q_{H}| = W/\varepsilon\), \(|Q_{L}| = |Q_{H}| - W\), and \(\Delta S = Q/T\).

ε = 1 − T_L/T_H; P = W/t; |Q_H| = W/ε; |Q_L| = |Q_H| − W
\[\begin{aligned} \varepsilon &= 1 - \tfrac{300}{850} \approx 0.647 \;(65\%) \qquad P = \tfrac{1200}{0.25} = 4.8\,\mathrm{kW} \\ |Q_{H}| &= \tfrac{1200}{0.647} \approx 1855\,\mathrm{J} \qquad |Q_{L}| = 1855 - 1200 = 655\,\mathrm{J} \\ \Delta S_{H} &= \tfrac{1855}{850} \approx +2.18\,\mathrm{J/K} \qquad \Delta S_{L} = \tfrac{-655}{300} \approx -2.18\,\mathrm{J/K} \end{aligned}\]

The two entropy changes cancel — the net per cycle is zero, exactly as a state function over a closed cycle must be.

4 An impossibly efficient engine

Problem. An inventor claims an engine that is 75% efficient operating between the boiling and freezing points of water. Possible?

Solution. Compare with the Carnot ceiling between \(T_{H} = 373\,\mathrm{K}\) and \(T_{L} = 273\,\mathrm{K}\).

ε_C = 1 − T_L/T_H
\[ \varepsilon_{C} = 1 - \frac{273}{373} \approx 0.268 \;(27\%) \]

The best possible efficiency between those temperatures is about 27%. A real engine must do worse still, so a claimed 75% is impossible — the inventor is mistaken.

5 Microstates and multiplicity

Problem. For 100 indistinguishable molecules in a box, compare the multiplicities of the (50, 50) and (100, 0) configurations.

Solution. Apply \(W = N!/(n_{1}!\,n_{2}!)\) to each.

W = N! / (n₁! n₂!)
\[ W_{50,50} = \frac{100!}{50!\,50!} \approx 1.01\times10^{29} \qquad W_{100,0} = \frac{100!}{100!\,0!} = 1 \]

The even split is more likely than the all-on-one-side split by a factor of about \(10^{29}\). And 100 is a tiny sample — for a mole the lopsided arrangements are effectively impossible, which is why gases sit at equilibrium.

6 Free expansion via Boltzmann's equation

Problem. Show that doubling the volume of \(n\) moles of ideal gas in a free expansion gives \(\Delta S = nR\ln 2\) — using statistical mechanics.

Solution. Initially all \(N\) molecules are on one side (\(W_{i} = 1\)); finally they spread over both (\(W_{f} = N!/[(N/2)!]^{2}\)). Apply \(S = k\ln W\) with Stirling's approximation.

ΔS = S_f − S_i = k ln W_f − k ln W_i
\[ S_{f} = k\big[N\ln N - 2(\tfrac{N}{2})\ln\tfrac{N}{2}\big] = Nk\ln 2 = nR\ln 2 \]
\[ \Delta S = nR\ln 2 - 0 = nR\ln 2 \]

Identical to the thermodynamic result of Example 2 (using \(Nk = nR\)). The microscopic counting and the macroscopic \(\int dQ/T\) are two faces of one quantity.

Review

Chapter Summary

Irreversibility

One-way processes set time's arrow. Their direction is fixed by entropy, not energy.

Entropy change

\(\Delta S = \int_{i}^{f} dQ/T\); isothermal \(\Delta S = Q/T\). A state function.

Ideal-gas entropy

\(\Delta S = nR\ln(V_{f}/V_{i}) + nC_{V}\ln(T_{f}/T_{i})\) — depends only on endpoints.

Second law

\(\Delta S \ge 0\) for a closed system (= reversible, > irreversible).

Engines

\(\varepsilon = W/|Q_{H}|\); Carnot \(\varepsilon_{C} = 1 - T_{L}/T_{H}\). No perfect engine.

Refrigerators

\(K = |Q_{L}|/|W|\); Carnot \(K_{C} = T_{L}/(T_{H}-T_{L})\). No perfect fridge.

Carnot cycle

Two isotherms + two adiabats; \(W = |Q_{H}| - |Q_{L}|\), \(|Q_{H}|/T_{H} = |Q_{L}|/T_{L}\).

Multiplicity

\(W = N!/(n_{1}!\,n_{2}!)\); all microstates equally probable.

Boltzmann entropy

\(S = k\ln W\); \(\ln N! \approx N\ln N - N\) (Stirling).

Practice

Problems

Take \(R = 8.31\,\mathrm{J/mol\cdot K}\), \(k = 1.38\times10^{-23}\,\mathrm{J/K}\), and water's specific heat \(4187\,\mathrm{J/kg\cdot K}\) (copper 386, lead 128). Always work in kelvins. For irreversible processes, build a reversible path between the same endpoints; for engines and fridges, lean on \(\varepsilon_{C} = 1 - T_{L}/T_{H}\) and the first law over a cycle.

  1. Suppose 4.00 mol of an ideal gas expands reversibly and isothermally from \(V_{1}\) to \(2.00V_{1}\) at 400 K. Find (a) the work done and (b) the entropy change. (c) If the expansion were reversible and adiabatic instead, what is \(\Delta S\)?
  2. A 2.50 mol sample of ideal gas expands reversibly and isothermally at 360 K until its volume doubles. What is the entropy increase?
  3. Find (a) the heat absorbed and (b) the entropy change of a 2.00 kg copper block whose temperature rises reversibly from 25.0 °C to 100 °C.
  4. What is the entropy change of a 12.0 g ice cube that melts completely in water just above the freezing point? (Heat of fusion 333 kJ/kg.)
  5. A 50.0 g copper block at 400 K is placed with a 100 g lead block at 200 K in an insulated box. Find (a) the equilibrium temperature and (b) the entropy change of the two-block system.
  6. In an insulated container, 200 g of aluminum (\(c = 900\,\mathrm{J/kg\cdot K}\)) at 100 °C is mixed with 50.0 g of water at 20.0 °C. Find (a) the equilibrium temperature and (b) the net entropy change.
  7. A Carnot engine absorbs 52 kJ and exhausts 36 kJ as heat each cycle. Find (a) its efficiency and (b) the work done per cycle.
  8. A Carnot engine has an efficiency of 22.0% and operates between reservoirs differing by 75.0 C°. Find the (a) lower and (b) higher reservoir temperatures.
  9. A Carnot engine whose cold reservoir is at 17 °C has an efficiency of 40%. By how much must the hot-reservoir temperature be raised to push the efficiency to 50%?
  10. How much work must a Carnot refrigerator do to move 1.0 J as heat (a) from 7.0 °C to 27 °C, and (b) from −73 °C to 27 °C?
  11. A heat pump (a Carnot engine in reverse) keeps a building at 22 °C while it is −25.0 °C outside, delivering 7.54 MJ per hour. At what rate must work be done to run it?
  12. A Carnot air conditioner moves heat from a 70 °F room to 96 °F outdoors. For each joule of electrical work, how many joules are removed from the room?
  13. Construct a table like the six-molecule table for a box of eight indistinguishable molecules: list the configurations, their multiplicities, and the total number of microstates.
  14. A box holds \(N = 50\) molecules. Find (a) the multiplicity of the central (25, 25) configuration, (b) the total number of microstates, and (c) the percentage of time the system spends in the central configuration.
Tip: three habits carry most of this chapter. For any entropy calculation, first ask "is this reversible?" — if not, invent a reversible path between the same endpoints and integrate \(dQ/T\) along it. For engines and refrigerators, the temperatures in \(\varepsilon_{C}\) and \(K_{C}\) must be in kelvins, and every real device falls short of the Carnot value. For the statistical view, remember that entropy is just \(k\ln W\), so the most probable configuration — the one with the most microstates — is equilibrium.