The Second Law and Entropy
Film a pendulum swinging and you cannot tell whether the movie is running forwards or backwards — the
laws of mechanics look the same both ways. But film a cup of coffee, and you know instantly. Coffee
cools; it never spontaneously reheats itself by drawing warmth from the room. Smoke spreads out; it
never gathers back into the cigarette. Eggs scramble; they never unscramble. There is a direction to
time — an arrow — and nothing in the first law explains it, because energy is
conserved whichever way the film runs. The quantity that does know which way is forwards is
called entropy, and the rule it obeys, the second law of
thermodynamics, is arguably the deepest statement in all of physics.
This is where the module's threads tie together. The
Carnot ceiling,
the wasted heat, the impossibility of a perfect engine — all of them turn out to be one law wearing
different costumes. That law is: entropy never decreases.
Entropy as heat-per-temperature
Clausius gave entropy its first precise, thermodynamic definition. When a small amount of heat
\mathrm{d}Q flows into a system reversibly (gently, always near
equilibrium) at absolute temperature T, the system's entropy
S changes by
-
\mathrm{d}S = \frac{\mathrm{d}Q_{\text{rev}}}{T}, \qquad \text{so for heat at constant } T,\quad \Delta S = \frac{Q}{T}.
-
The unit is joules per kelvin (\text{J/K}). Like internal energy,
entropy is a state function: it depends only on the state, not the path.
-
The same joule of heat carries more entropy when it enters a cold system (small
T) than a hot one. Temperature is the "exchange rate" between heat and
entropy.
That last point is the secret engine of the whole subject. When heat Q
leaves a hot body at T_h and arrives at a cold body at
T_c, the hot body loses entropy Q/T_h but the
cold body gains a larger entropy Q/T_c (because
T_c < T_h). The total goes up. Heat flowing from hot to
cold always increases the entropy of the universe — which is exactly why it happens, and its reverse
never does.
The second law: the arrow of time
-
The entropy of an isolated system never decreases. For any process,
\Delta S_{\text{universe}} \ge 0.
-
It stays constant (\Delta S = 0) only for an ideal
reversible process; every real, irreversible process increases it.
Two famous everyday-language versions say the same thing, and each forbids a specific "obvious"
machine that would run the film backwards:
-
Clausius statement: heat cannot flow spontaneously from a colder body to a hotter
one. (A fridge can move heat that way — but only by doing work, which raises the total
entropy anyway.)
-
Kelvin(–Planck) statement: no cyclic engine can take heat from a single reservoir
and turn all of it into work. There is no perfect heat engine; you always need a cold
reservoir to dump entropy into.
These look like statements about engines and fridges, but they are the same law that says coffee cools
and smoke spreads. Ban the perfect fridge and you have banned the un-spreading of smoke.
Why? Entropy is really counting
Clausius told us what entropy does; Boltzmann told us why. His insight — carved on
his gravestone — is that entropy measures how many microscopic arrangements
(microstates) correspond to what we see (a macrostate):
S = k_B \ln W,
where W is the number of ways the molecules can be arranged to give the
same overall appearance, and k_B is Boltzmann's constant. A tidy,
low-entropy state (all the gas in one corner) can be made in very few ways; a spread-out, high-entropy
state (gas filling the box) can be made in overwhelmingly more. So a system drifts toward high entropy
for a reason as simple as flipping coins: the "half heads, half tails" outcome dominates simply
because there are astronomically more ways to make it.
The chart shows this for a box of 20 molecules: the number of arrangements with
n of them on the left. The half-and-half split isn't slightly more
likely — it towers over the "all on one side" states. With 10^{23}
molecules the peak becomes so needle-sharp that any measurable departure from uniform spreading simply
never happens. That is the second law, born from pure counting.
Worked examples
Example 1 — melting ice. It takes
Q = 334\ \text{J} to melt one gram of ice at
0\,^\circ\text{C} = 273\ \text{K}, and it melts at constant temperature. The
entropy gained by the water is
\Delta S = \frac{Q}{T} = \frac{334}{273} \approx 1.22\ \text{J/K}.
Liquid water is more disordered than ice, so its entropy is higher — as the count confirms.
Example 2 — heat flowing hot to cold. A large hot reservoir at
T_h = 400\ \text{K} passes Q = 1000\ \text{J} to a
cold reservoir at T_c = 300\ \text{K}. The net entropy change of the universe
is
\Delta S = \frac{Q}{T_c} - \frac{Q}{T_h} = \frac{1000}{300} - \frac{1000}{400} = 3.33 - 2.50 = +0.83\ \text{J/K}.
Positive, as the second law demands. Run the heat the other way (cold to hot) and
\Delta S would be negative — forbidden.
Example 3 — a reversible cycle. A Carnot engine absorbs
Q_h/T_h of entropy at the hot side and rejects exactly
Q_c/T_c at the cold side. Because it is reversible these are equal,
so \Delta S_{\text{universe}} = 0. This is the very condition
Q_c/Q_h = T_c/T_h that gave the Carnot efficiency — the two chapters are one
law.
Life, crystals, snowflakes, a tidied room — order appears all the time. Does that break the second
law? No, and the escape clause is the word isolated. Entropy can happily
decrease in one place as long as it increases by more somewhere else. You tidy your room
(local entropy down), but your body burns food and radiates heat, raising the universe's entropy far
more than the tidying lowered it. The Earth is not isolated: it bathes in a stream of concentrated,
low-entropy sunlight (a few high-energy photons from a very hot Sun) and radiates away a flood of
cold, high-entropy infrared (many low-energy photons to cold space). Life is powered by that entropy
gradient. So the intricate order of a living planet is "paid for" many times over by the entropy the
Sun exports into the dark. The universe as a whole still runs relentlessly downhill; we are just
clever little eddies on the way down.
Three pitfalls. First, "entropy = messiness" is a loose analogy that misleads:
entropy is precisely k_B \ln W, a count of microstates, and a "messy" desk
vs a "tidy" one has nothing to do with it. A better mental image is how spread out the energy and
matter are. Second, the definition \mathrm{d}S = \mathrm{d}Q_{\text{rev}}/T
uses the heat along a reversible path; for an irreversible process you must invent a
reversible route between the same endpoints (entropy is a state function, so that's allowed) — the
actual irreversible heat over T would give too small a value. Third, that
T is absolute (kelvin), always. Finally, the second law
forbids the total entropy of an isolated system from dropping; it says nothing against a
part of the system becoming more ordered, provided the rest more than compensates.