🌡️ Thermodynamics · Statistical Mechanics
📅 Березень 2026⏱ ≈ 10 хв читання🟡 Середній

What Is Entropy? — Finally Understood

Entropy is one of the most misused words in science. It is not simply "disorder." It is a precisely defined quantity that measures the number of ways a system can be arranged to look the same — and understanding it explains why time has a direction.

1. Intuition: Why Eggs Don't Unscramble

Drop a glass of red wine into a tank of water. The wine disperses. Wait a billion years: the wine will not spontaneously reassemble. Why? The laws of physics at the particle level are time-reversible — the equations work equally well forwards and backwards. So why does the wine only spread?

The answer is probability. There are vastly more ways for the wine molecules to be spread throughout the tank than for them to be clumped in one corner. A system randomly exploring its possible configurations almost certainly moves toward the most numerous ones — the spread-out states.

This is entropy: the logarithm of the number of available configurations (microstates).

2. Thermodynamic Entropy

Clausius defined entropy operationally in 1865, before anyone understood atoms. When a system absorbs a small amount of heat dQ reversibly at absolute temperature T, its entropy increases by:

Clausius definition (1865) dS = δQrev / T

S = entropy (J/K)
δQrev = infinitesimal reversible heat absorbed (J)
T = absolute temperature (K)

Heating something increases its entropy. The hotter the system already is, the smaller the entropy increase from the same heat input — because at higher temperatures the system already has many accessible energy states.

Carnot efficiency: The maximum efficiency of any heat engine operating between temperatures Tcold and Thot is η = 1 − Tcold/Thot. This limit follows directly from entropy: the entropy shed to the cold reservoir must be at least as much as was absorbed from the hot one.

3. Boltzmann's Statistical Entropy

In 1877, Boltzmann connected Clausius's thermodynamic entropy to probability. His tombstone famously bears the equation:

Boltzmann's entropy formula S = kB ln W

kB = 1.380649 × 10⁻²³ J/K (Boltzmann constant)
W = number of microstates consistent with the macrostate
ln = natural logarithm

Why a logarithm? Because entropy must be additive (entropy of two independent systems = sum of their entropies), but if you combine two systems each with W₁ and W₂ microstates, the combined system has W₁ × W₂ microstates. Logarithm converts multiplication to addition: ln(W₁·W₂) = ln W₁ + ln W₂.

4. Microstates and Macrostates

A macrostate is what we observe: temperature, pressure, volume. A microstate specifies the position and velocity of every molecule. Many microstates can produce the same macrostate.

Consider 4 molecules in a 2-compartment box:

Left | Right Microstates W Probability
4 | 0 1 6.25%
3 | 1 4 25%
2 | 2 6 37.5% ← most likely
1 | 3 4 25%
0 | 4 1 6.25%

With 4 molecules, the all-in-one-side configuration is plausible. With 6 × 10²³ molecules (Avogadro's number), the probability of spontaneous separation is so small it has never and will never occur in the lifetime of the universe.

5. The Second Law of Thermodynamics

The Second Law states: the total entropy of an isolated system never decreases over time.

Macroscopic view

Heat flows from hot to cold. Gas expands to fill a vacuum. Ordered structures decay. Processes are irreversible.

Statistical view

The system wanders among microstates with equal probability. High-entropy (more numerous) macrostates are visited almost always.

The Second Law is statistical, not absolute. In principle, an egg could unscramble — it is just so astronomically unlikely that it never happens. At the nanoscale, temporary decreases in entropy occur (Brownian motion, fluctuation theorems).

Entropy and life: Living organisms maintain low internal entropy by exporting entropy to the environment (eating food, emitting heat). The Second Law is not violated — local order is paid for by global disorder increase. The Earth as a whole exports entropy via thermal radiation to space.

6. Information Entropy (Shannon)

In 1948, Claude Shannon independently derived the same mathematical form for communication theory. Shannon entropy of a probability distribution {p₁, p₂, …, pₙ} is:

Shannon entropy H = − Σ pᵢ log₂(pᵢ) (bits)

Or equivalently: H = − Σ pᵢ ln(pᵢ) / ln(2)

Maximum entropy: all outcomes equally likely (H = log₂ n)
Minimum entropy: one outcome certain (H = 0)

Shannon entropy measures information content or surprise. A fair coin flip has H = 1 bit. A biased coin with P(heads)=0.9 has H ≈ 0.47 bits — less uncertainty, less information gained from each flip.

This is not a coincidence: Boltzmann and Shannon entropy are the same mathematical object. Physical entropy measures uncertainty about which microstate a system is in; information entropy measures uncertainty about which symbol was sent.

Compression: Shannon entropy sets the theoretical minimum bits needed to encode a source. ZIP/DEFLATE, JPEG, MP3 all approach this limit through different statistical models of the data.

7. The Arrow of Time

The microscopic laws of physics (Newton, Maxwell, Schrödinger) are time-symmetric — they work the same forwards and backwards. Yet the world has a clear past → future direction: coffee cools, memories form, causes precede effects.

Entropy explains this: the past had lower entropy. Why? The universe started in an extremely ordered (low-entropy) state — the Big Bang. The second law is just statistics: we are still relaxing toward maximum entropy, and "forward in time" is the direction of increasing entropy.

Maxwell's Demon (1867): James Clerk Maxwell imagined a tiny demon opening a door between gas compartments to let fast molecules in and slow molecules out — decreasing entropy for free. The resolution came in 1961 (Landauer): the demon's memory must eventually be erased, and that erasure dissipates at least kBT ln 2 of energy per bit. Information erasure is irreversible and generates entropy.

8. Common Misconceptions