Entropy and the Direction of Spontaneous Change

Drop an ice cube into warm tea, and it always melts. It never spontaneously refreezes while the tea gets hotter. Something in nature has a direction, and entropy is the quantity that tracks it. This article shows how the entropy of the universe, system plus surroundings, provides an exact, quantitative criterion for spontaneity. This matters because energy conservation alone cannot explain why processes run one way. The first law is silent on direction. It permits heat to flow from cold to hot just as happily as hot to cold. The second law fixes this by introducing total entropy, and demanding that it never decrease. You will see how the Clausius inequality leads to a total-entropy criterion. You will learn to compute $\Delta S_{total}$ for real processes. You will see why $\Delta S_{total} > 0$ signals irreversibility, and $\Delta S_{total} = 0$ signals reversibility. Two full numerical examples anchor the ideas.

Key takeaways

Spontaneous processes increase the total entropy of system plus surroundings; this is the operational Second Law criterion.
$\Delta S_{total} = \Delta S_{sys} + \Delta S_{surr} \geq 0$ , with equality only for reversible changes.
A process can lower the system's own entropy, provided the surroundings' entropy rises even more.
The Clausius inequality, $dS \geq dq/T_{ext}$ , is the differential root of this criterion.
Free energies ( $G$ , $A$ ) later repackage this same criterion using system-only variables at constant $T$ , $p$ .
Key idea: Spontaneity is decided by the sign of $\Delta S_{universe}$ , never by $\Delta S_{system}$ alone.

The core insight

A process proceeds spontaneously in a given direction if and only if it increases the combined entropy of the system and its surroundings; the system's own entropy may rise, fall, or stay constant, but the sum $\Delta S_{sys} + \Delta S_{surr}$ can never be negative for a real, physically achievable change.

The physical picture

Consider two blocks of metal, one hot, one cold, brought into thermal contact. Heat flows spontaneously from hot to cold. Energy is conserved: the heat lost by the hot block exactly equals the heat gained by the cold block. Yet the reverse process, heat flowing from cold to hot with no other change, would also conserve energy. It never happens. The First Law cannot explain the asymmetry; a second, independent principle is needed. That principle is entropy. When heat $dq$ leaves the hot block at temperature $T_h$ , its entropy changes by $dq/T_h$ , a negative quantity since heat is leaving. When the same heat enters the cold block at $T_c < T_h$ , its entropy rises by $dq/T_c$ , a positive quantity of larger magnitude because $T_c$ is smaller. The total, treating hot block as "system" and cold block as "surroundings" for bookkeeping, is

\Delta S_{total} = -\frac{dq}{T_h} + \frac{dq}{T_c} > 0

because $1/T_c > 1/T_h$ . The asymmetry between the two directions is now explicit: heat flowing hot-to-cold gives $\Delta S_{total} > 0$ , while the reverse would give $\Delta S_{total} < 0$ , which the Second Law forbids. This is the microscopic seed of the criterion built formally below from the Clausius inequality. Crucially, the system's entropy alone is not decisive. Many spontaneous processes, water freezing, a gas condensing, a protein folding, actually decrease the system's entropy. What guarantees spontaneity is that the surroundings' entropy increases by at least as much, so the sum is never negative.

Figure

Heat flow between two blocks: entropy lost by the hot block is smaller in magnitude than entropy gained by the cold block

The second figure shows the same criterion plotted more abstractly, tracking $\Delta S_{total}$ as a reaction coordinate advances, illustrating why the maximum of total entropy marks equilibrium.

Figure

Total entropy versus extent of reaction: spontaneous change climbs toward the entropy maximum at equilibrium

The core relations

dS_{univ} = dS_{sys} + dS_{surr} \geq 0

where: equality holds only for a reversible process; strict inequality signals an irreversible, spontaneous process.

\Delta S_{sys} \geq \int \frac{dq}{T_{ext}}

where: $T_{ext}$ is the temperature of the surroundings at the boundary where heat is exchanged (the Clausius inequality).

\Delta S_{surr} = -\frac{q_{sys}}{T_{surr}}

where: valid when the surroundings are a large reservoir at constant $T_{surr}$ , so heat exchange with it is effectively reversible from its perspective.

\Delta S_{total} = \Delta S_{sys} - \frac{q_{sys}}{T_{surr}} \geq 0

where: this is the working spontaneity criterion for a process at constant surroundings temperature $T_{surr}$ .

Derivation AFrom the Clausius Inequality to the Total-Entropy Criterion

Goal. Starting from the Clausius inequality, which you already know relates heat and entropy for any cyclic or one-way process, derive the statement that total entropy change is non-negative, and show equality holds exactly at reversibility.

Step 1. Recall the Clausius inequality. For any process taking a system from state 1 to state 2, exchanging heat $dq$ with surroundings at the boundary temperature $T_{ext}$ ,

dS_{sys} \geq \frac{dq}{T_{ext}}

This inequality is already established from the Kelvin statement of the Second Law applied to a Carnot-engine argument; here it is only the starting point.

Step 2. Model the surroundings as a reservoir. Take the surroundings to be large enough that absorbing or releasing heat does not change their temperature appreciably. Because the surroundings experience heat transfer reversibly, in the sense that their own internal relaxation is fast compared to the process, their entropy change is exact:

dS_{surr} = \frac{dq_{surr}}{T_{surr}} = -\frac{dq_{sys}}{T_{surr}}

The minus sign appears because heat leaving the system, $dq_{sys}$ , is heat entering the surroundings, $dq_{surr} = -dq_{sys}$ . Since the surroundings are a reservoir, $T_{ext} = T_{surr}$ .

Step 3. Add the two entropy changes. Substitute $T_{ext} = T_{surr}$ into Step 1 and rearrange:

dS_{sys} \geq \frac{dq_{sys}}{T_{surr}}

dS_{sys} + \left(-\frac{dq_{sys}}{T_{surr}}\right) \geq 0

dS_{sys} + dS_{surr} \geq 0

Step 4. Identify equality with reversibility. The Clausius inequality becomes an equality, $dS_{sys} = dq/T_{ext}$ , precisely when the process is carried out reversibly, meaning infinitesimally slowly with the system always in internal equilibrium and only infinitesimal temperature differences driving heat flow. In that limit, Step 3 gives $dS_{total} = 0$ : a reversible process leaves total entropy unchanged. Any finite temperature gradient, friction, unrestrained expansion, or other irreversibility makes the original inequality strict, $dS_{sys} > dq/T_{ext}$ , and propagates through to $dS_{total} > 0$ .

Step 5. Integrate for a finite change. Integrating both sides from initial to final states,

\Delta S_{total} = \Delta S_{sys} + \Delta S_{surr} \geq 0

Boxed result.

\boxed{\Delta S_{total} \geq 0 \text{, with equality if and only if the process is reversible}}

Worked examples

Example 1 — Foundational: irreversible expansion of an ideal gas into vacuum. One mole of an ideal gas, initially confined to a 10.0 L bulb at 298 K, is allowed to expand freely (Joule expansion) into an evacuated 20.0 L bulb, doubling its volume with no external pressure resisting the expansion and no heat exchanged with the surroundings, since the whole apparatus is rigid and thermally isolated. System entropy change. Entropy is a state function, so $\Delta S_{sys}$ depends only on initial and final states, not on the (irreversible) path actually taken. Use the reversible isothermal path formula:

\Delta S_{sys} = nR\ln\frac{V_2}{V_1} = (1.00 \text{ mol})(8.314 \text{ J K}^{-1}\text{mol}^{-1})\ln\frac{20.0}{10.0}

\Delta S_{sys} = (8.314 \text{ J K}^{-1})(0.6931) = 5.76 \text{ J K}^{-1}

Surroundings entropy change. Since the process is adiabatic, $q_{actual} = 0$ (no heat crosses into the surroundings, because the container is rigid and insulated, so no work is done either). Therefore

\Delta S_{surr} = -\frac{q_{actual}}{T_{surr}} = -\frac{0}{298 \text{ K}} = 0

Total entropy change.

\Delta S_{total} = 5.76 \text{ J K}^{-1} + 0 = 5.76 \text{ J K}^{-1}

Since $\Delta S_{total} > 0$ , the free expansion is confirmed spontaneous and irreversible, consistent with everyday experience: gas never spontaneously recompresses into half its volume. Note the key subtlety: $\Delta S_{surr}$ is computed from the actual heat exchanged ( $q = 0$ here), not from any hypothetical reversible path, because the surroundings genuinely receive zero heat.

Example 2 — Applied: freezing of supercooled water at $-10\,^\circ\text{C}$ . One mole of liquid water is supercooled to $-10\,^\circ\text{C}$ (263 K) and freezes irreversibly at that temperature, releasing heat to surroundings held at 263 K. Data: enthalpy of fusion at the normal melting point (273 K), $\Delta H_{fus}(273\text{ K}) = 6008$ J/mol; heat capacities $C_{p,m}(\text{liquid}) = 75.3$ J K $^{-1}$ mol $^{-1}$ , $C_{p,m}(\text{solid}) = 37.7$ J K $^{-1}$ mol $^{-1}$ . Strategy. Build a reversible three-step path between the same initial and final states to get $\Delta S_{sys}$ : (i) warm supercooled liquid water from 263 K to 273 K, (ii) freeze reversibly at 273 K, (iii) cool the solid from 273 K back to 263 K.

Step (i):

\Delta S_1 = C_{p,m}(\text{liq})\ln\frac{273}{263} = (75.3)\ln(1.0380) = (75.3)(0.03731) = 2.81 \text{ J K}^{-1}

Step (ii):

\Delta S_2 = \frac{-\Delta H_{fus}}{273 \text{ K}} = \frac{-6008 \text{ J}}{273 \text{ K}} = -22.01 \text{ J K}^{-1}

Step (iii):

\Delta S_3 = C_{p,m}(\text{sol})\ln\frac{263}{273} = (37.7)\ln(0.96337) = (37.7)(-0.03731) = -1.41 \text{ J K}^{-1}

Sum:

\Delta S_{sys} = 2.81 - 22.01 - 1.41 = -20.6 \text{ J K}^{-1}

Heat released to surroundings. Compute the actual enthalpy change along the same three-step path (heat capacities at constant pressure give $q_p = \Delta H$ ):

\Delta H_1 = 75.3 \times 10 = 753 \text{ J}, \quad \Delta H_2 = -6008 \text{ J}, \quad \Delta H_3 = 37.7 \times (-10) = -377 \text{ J}

\Delta H_{total} = 753 - 6008 - 377 = -5632 \text{ J}

Since the actual process occurs at constant pressure with the same $T_i, T_f$ , this is the heat released by the system, $q_{sys} = -5632$ J, so the surroundings absorb $+5632$ J at 263 K:

\Delta S_{surr} = -\frac{q_{sys}}{T_{surr}} = -\frac{(-5632)}{263} = 21.4 \text{ J K}^{-1}

Total.

\Delta S_{total} = -20.6 + 21.4 = +0.8 \text{ J K}^{-1}

Total entropy rises by a small but positive amount, confirming that supercooled water freezing at $-10\,^\circ\text{C}$ is spontaneous, even though the system's own entropy decreases substantially (order increases as liquid becomes solid). The surroundings gain more entropy from absorbing the latent heat than the system loses. This is the everyday resolution of the apparent paradox that ice (an "ordered" solid) forms spontaneously.

Limiting cases & common traps

Situation	Correct treatment
Reversible process	$\Delta S_{total} = 0$ exactly; system and surroundings entropy changes are equal and opposite.
Irreversible, isolated system	$\Delta S_{surr} = 0$ (no heat crosses boundary), so spontaneity requires $\Delta S_{sys} > 0$ alone.
Exothermic reaction lowering $\Delta S_{sys}$	Still spontaneous if heat released raises $\Delta S_{surr}$ enough; check the sum, not $\Delta S_{sys}$ alone.
Using $q_{rev}$ for the surroundings	Wrong unless the surroundings' process is reversible; use the actual heat transferred, $q_{sys}$ , with $\Delta S_{surr} = -q_{sys}/T_{surr}$ .
Small system, non-reservoir surroundings	Formula $\Delta S_{surr} = -q_{sys}/T_{surr}$ assumes surroundings stay at constant $T$ ; recompute with heat capacities if not.

The chemist behind it

Rudolf Clausius introduced the term "entropy" in 1865, deliberately choosing a word related to the Greek for "transformation" to parallel "energy." Building on Sadi Carnot's earlier analysis of heat engines and Kelvin's statement of the Second Law, Clausius formulated the inequality that bears his name and stated the Second Law in its most quoted form: the entropy of the universe tends to a maximum. His insight, that entropy could increase even while local order increased elsewhere, laid the foundation for the total-entropy spontaneity criterion used throughout thermodynamics and, later, statistical mechanics.

Origins · 1 minRudolf Clausius and how this was discovered

Rudolf Clausius

1822–1888 · German physicist who coined "entropy" and formulated the Clausius inequality underlying the Second Law

Ask the tutor: Can you walk me through how this total-entropy criterion transforms into the Gibbs energy criterion at constant temperature and pressure?

Why isn't ΔS_system enough to decide spontaneity on its own?

Because the system can trade entropy with its surroundings. A system's entropy can fall, as in freezing or condensation, while heat released to the surroundings raises their entropy by an even larger amount. The First and Second Laws together track energy and entropy across the whole universe, system plus surroundings, not the system in isolation. Only the sum, $\Delta S_{total}$ , is guaranteed non-negative for any real process; the system term alone carries no such guarantee.

How is ΔS_surroundings actually calculated in practice?

For surroundings large enough to act as a constant-temperature reservoir, $\Delta S_{surr} = -q_{sys}/T_{surr}$ , where $q_{sys}$ is the actual heat transferred to the system (so heat leaving the system enters the surroundings with the opposite sign) and $T_{surr}$ is the reservoir's temperature. This works even for an irreversible process in the system, because the reservoir itself experiences the heat transfer reversibly, since its temperature does not change appreciably.

Does ΔS_total = 0 mean nothing happens?

No. It means the process, if it happens, is thermodynamically reversible: an idealized, infinitely slow path where the system passes through a continuous sequence of equilibrium states. Real processes always proceed at finite rates and involve some irreversibility, friction, finite temperature gradients, unrestrained expansion, so $\Delta S_{total} > 0$ for anything that actually occurs at a measurable rate. The reversible limit is a useful benchmark, not a literal description of real change.

How does this total-entropy criterion connect to Gibbs energy later?

At constant temperature and pressure, the criterion $\Delta S_{total} \geq 0$ can be rewritten entirely in terms of system properties by substituting $\Delta S_{surr} = -\Delta H_{sys}/T$ . This yields $\Delta G = \Delta H - T\Delta S \leq 0$ for the system alone. Gibbs energy is simply a repackaging of the same total-entropy criterion, convenient because it avoids explicitly computing surroundings' properties.

What role does the Clausius inequality play here, and how does it relate to entropy being a state function?

The Clausius inequality, $dS \geq dq/T_{ext}$ , is the differential statement from which the total-entropy criterion is derived, as shown in the derivation above. Because entropy is a state function (a prerequisite result), $\Delta S_{sys}$ can always be computed along any convenient reversible path connecting the same initial and final states, even when the actual process is irreversible, which is exactly the technique used in both worked examples.