Visual walkthrough — Entropy change in irreversible processes — always - 0

We build up to this master statement, the Clausius inequality: $\oint \frac{\delta Q}{T_{\text{surr}}}\le 0$ Don't worry about what a single symbol means yet — every one gets earned below.

---

Step 1 — What is a "state" and what is a "process"?

WHAT. Picture a gas trapped in a box. At any instant we can measure two numbers about it: how hot it is (temperature $T$) and how squeezed it is (volume $V$). One dot on a chart of $T$ against $V$ is one state of the gas — one complete "snapshot".

WHY. Before we can talk about change, we need to fix what changes between. A process is just a trip: the gas moves from a starting dot $A$ to a finishing dot $B$. The whole derivation is a claim about that trip.

PICTURE. Two dots, $A$ and $B$, and a wiggly line joining them — the path the gas actually took.

---

Step 2 — Two kinds of trip: reversible and irreversible

WHAT. Some trips can be run backwards and leave no trace — rewind the film and nothing looks wrong. Those are reversible. Others cannot: shove the piston hard, let gas rush into vacuum, rub with friction — rewind that film and it looks absurd. Those are irreversible.

WHY. The whole inequality hinges on this fork. Reversibility is the knife-edge where the law reads "$=$"; every real trip falls to the irreversible side where it reads "$>$". See Reversible vs irreversible processes.

PICTURE. A smooth path (reversible: system always in balance, tiny steps) versus a jagged path (irreversible: big shoves, gradients).

---

Step 3 — Defining entropy on the gentle path

WHAT. Along a reversible trip only, we define a running score $S$ called entropy by adding up $dS=\frac{\delta Q_{\text{rev}}}{T}$

WHY this formula and not something simpler? We want a number whose total change depends only on the endpoints $A$ and $B$, never on the wiggles between. Heat $\delta Q$ alone fails that test — how much heat you exchange depends wildly on the path. But dividing each heat puff by the temperature $T$ at which it crosses turns out to give a quantity that only cares about $A$ and $B$. That is a small miracle, and it's why the $T$ in the denominator is not optional — see Entropy as a state function.

Term by term:

• $dS$ — the tiny change in the entropy score.
• $\delta Q_{\text{rev}}$ — the tiny heat puff, measured on a reversible (gentle) path.
• $T$ — the absolute temperature (in kelvin) at which that puff crosses the wall. The subscript "rev" is a promise: this formula is only licensed on a reversible path.

PICTURE. A gentle path from $A$ to $B$ chopped into puffs $\delta Q_1,\delta Q_2,\dots$, each divided by the local $T$, and stacked into a bar that is the total $\Delta S$.

---

Step 4 — The only law we may assume: the Clausius inequality

WHAT. For any process that returns to where it started (a cycle, the loop symbol $\oint$), experiment and the Second Law of Thermodynamics give us exactly one weapon: $\oint \frac{\delta Q}{T_{\text{surr}}}\le 0$

WHY. We are not allowed to assume $\Delta S_{\text{univ}}\ge 0$ — that is what we're trying to prove. The Clausius inequality is the one honest starting brick. Everything else is deduced.

Term by term:

• $\oint$ — "add up all the way around a closed loop", i.e. over a full cycle.
• $\delta Q$ — heat puff crossing the boundary (the real one, could be irreversible).
• $T_{\text{surr}}$ — the temperature of the surroundings where that puff crosses. Not the system's temperature! When things are irreversible these two differ, and which $T$ we use matters.
• $\le 0$ — equals $0$ only if the whole loop was reversible; strictly less than $0$ if any part was irreversible.

PICTURE. A closed loop in the $T$–$V$ chart, arrows showing heat in (red) and heat out (blue), with the running sum $\oint \delta Q/T_{\text{surr}}$ landing at or below zero.

---

Step 5 — The clever cycle: irreversible out, reversible back

WHAT. Take any real (irreversible) trip $A\to B$. Now imagine returning $B\to A$ along a gentle reversible path. The two legs joined together make a closed loop — a cycle.

WHY. Our only weapon (Step 4) fires on cycles. A single one-way trip is not a cycle, so we manufacture one by gluing a reversible return trip onto our real trip. The reversible return is where Step 3's definition of $S$ will bite.

PICTURE. The jagged red irreversible leg $A\to B$, then the smooth green reversible leg $B\to A$, closing the loop.

---

Step 6 — Apply the weapon and split the loop

WHAT. Fire the Clausius inequality on our manufactured loop. The loop-sum splits into the two legs: $\underset{\text{leg 1: the real trip}}{\underbrace{{\int }_A^B\frac{\delta Q_{\text{irr}}}{T_{\text{surr}}}}}+\underset{\text{leg 2: gentle return}}{\underbrace{{\int }_B^A\frac{\delta Q_{\text{rev}}}{T}}}\le 0$

WHY. "Around the whole loop" just means "leg 1 then leg 2". We split so we can recognise leg 2 as an entropy change (Step 3).

The recognition. By Step 3, the reversible leg is an entropy difference — and it runs $B\to A$, so it comes out negative: ${\int }_B^A\frac{\delta Q_{\text{rev}}}{T}=S_A-S_B=-\Delta S_{\text{system}}$

Term by term of that line:

• ${\int }_B^A$ — summing along the reversible return leg.
• $S_A-S_B$ — because we walk from $B$ back to $A$, we pick up $A$'s score minus $B$'s score.
• $-\Delta S_{\text{system}}$ — that's just $-(S_B-S_A)$, the negative of the system's forward entropy change.

PICTURE. The loop-sum equation drawn as two stacked bars: the red leg-1 bar plus the green leg-2 bar (pointing down, since it's $-\Delta S$) must not rise above zero.

---

Step 7 — Rearrange into the master result

WHAT. Move the $-\Delta S_{\text{system}}$ to the other side: $\Delta S_{\text{system}}\ge {\int }_A^B\frac{\delta Q_{\text{irr}}}{T_{\text{surr}}}$

WHY. The right-hand side is heat entering the system, scored at the surroundings' temperature. But the heat the system gains is exactly the heat the surroundings lose: $\delta Q_{\text{irr}}^{\text{sys}}=-\delta Q^{\text{surr}}$. The surroundings are a huge reservoir, so they always exchange heat reversibly at their steady $T_{\text{surr}}$. That means the right-hand integral is precisely $-\Delta S_{\text{surr}}$: ${\int }_A^B\frac{\delta Q_{\text{irr}}}{T_{\text{surr}}}=-\Delta S_{\text{surr}}$

Substitute and move it over: $\Delta S_{\text{system}}+\Delta S_{\text{surr}}\ge 0⟹\boxed{\Delta S_{\text{univ}}\ge 0}$

Term by term:

• $\Delta S_{\text{system}}$ — the system's entropy change (can be negative!).
• $\Delta S_{\text{surr}}$ — the surroundings' entropy change.
• Their sum — the universe's entropy change — is what can never drop.

PICTURE. Two bars — system and surroundings — that individually can point up or down, but whose combined height is always $\ge 0$.

---

Step 8 — Where does the extra come from? Entropy production $\sigma$

WHAT. Rewrite the inequality as an equality by naming the surplus: $d{S}_{\text{system}}=\frac{\delta Q}{T_{\text{surr}}}+\underset{\ge 0}{\underbrace{\delta \sigma }}$

WHY. The gap between "$\ge$" and "$=$" isn't mysterious — it's a real, non-negative quantity $\sigma$ called entropy production. Every finite gradient (a temperature gap, a pressure gap, friction, free expansion) manufactures some $\sigma >0$. Only when every driving push is infinitesimally gentle does $\sigma =0$ and the process become reversible.

Term by term:

• $\frac{\delta Q}{T_{\text{surr}}}$ — entropy transported in with the heat.
• $\delta \sigma \ge 0$ — entropy created inside, out of nothing, by irreversibility. Never negative.

PICTURE. A pipe: entropy flows in through the heat ($\delta Q/T_{\text{surr}}$), and an extra spring inside the box adds $\delta \sigma$ — the box's entropy rises by the sum.

---

The one-picture summary

Everything above, compressed: manufacture a cycle (irreversible out, reversible back), fire the Clausius inequality, recognise the return leg as $\Delta S$, and read off that the system-plus-surroundings entropy can only sit still (reversible) or rise (real).

Recall Feynman retelling — the walkthrough in plain words

We started with just two ideas: a snapshot of a system (a state) and a trip between snapshots (a process). Some trips are gentle enough to rewind (reversible); most real ones aren't (irreversible). On the gentle trips we invented a running score — entropy — by adding up each puff of heat divided by the temperature it crossed at, and that score turned out to depend only on where you start and stop, not the wiggles in between.

Then we grabbed the one law we're allowed: go around any loop, add up heat-over-temperature, and you never get a positive number. To use it on a one-way real trip, we cheated cleverly — we glued a gentle return trip onto it to make a loop. The return leg, being gentle, is an entropy change (that's what entropy is). The loop law then squeezes out the punchline: the system's entropy change plus the surroundings' entropy change can never be negative.

Finally we named the surplus: every real gradient — a hot thing touching a cold thing, a gas bursting into vacuum, a squeak of friction — makes a little extra entropy out of nothing. Shrink the gradients to zero and you make none (reversible, the knife-edge). Make them finite and you always make some. You can never un-make it. That "can only go up" is why spilled milk never leaps back into the cup.

Recall Quick self-test

• Why glue a reversible leg onto the irreversible trip? ::: To form a cycle, so we can use $\oint \delta Q/T\le 0$.
• Which temperature scores the surroundings? ::: $T_{\text{surr}}$, not the system's $T$.
• What is $\sigma$ and its sign? ::: Entropy produced inside by irreversibility; $\sigma \ge 0$ always.
• When is $\Delta S_{\text{univ}}=0$? ::: Only in the reversible limit, every gradient infinitesimal.

---

Related: Second Law of Thermodynamics · Clausius inequality · Entropy as a state function · Reversible vs irreversible processes · Free expansion of an ideal gas · Statistical interpretation of entropy (Boltzmann S = k ln W) · Arrow of time · Carnot cycle and efficiency