Thermodynamics and Entropy — Why Time Moves Forward

You can unscramble a sentence but not an egg. You remember yesterday but not tomorrow. Heat flows from hot coffee into a cold room, never the reverse. All three are consequences of one of the deepest principles in physics: entropy increases. Thermodynamics is the science of energy, heat, and work, but its biggest gift to human understanding is an explanation for why time only runs one way.

The Four Laws

Thermodynamics rests on four statements, numbered in the order they were recognized as fundamental rather than the order they were discovered. Each one closes off a different kind of "free lunch" that engineers spent centuries trying to build.

Law	Statement	Consequence
Zeroth	If object A is in thermal equilibrium with object B, and B is in equilibrium with C, then A is in equilibrium with C	Defines temperature as a real, comparable quantity; makes thermometers meaningful
First	Energy is conserved: the change in internal energy equals heat added minus work done (ΔU = Q − W)	No machine can produce more energy than it consumes; rules out "perpetual motion of the first kind"
Second	The total entropy of an isolated system never decreases	Sets the arrow of time; rules out "perpetual motion of the second kind" (a machine that converts heat entirely into work with no waste)
Third	The entropy of a perfect crystal approaches a constant (often taken as zero) as temperature approaches absolute zero	Absolute zero (0 K, −273.15 °C) can be approached but never fully reached in a finite number of steps
Table: The four laws of thermodynamics and what each one forbids.

The "zeroth" law was named after the fact, in the early twentieth century, once physicists realized the other three all silently depend on it — you cannot even define temperature without it, so it had to be logically first even though it was written down last.

What Entropy Actually Is

Entropy entered physics in 1865, when the German physicist Rudolf Clausius coined the term while trying to formalize the Second Law mathematically. His definition was macroscopic: dS = dQ/T, the heat exchanged by a system divided by the temperature at which the exchange happens. That equation is exactly what an engineer needs to size a steam engine, but it does not tell you what entropy is.

Ludwig Boltzmann answered that question two decades later. For a system that can occupy any of W equally likely microscopic arrangements (called microstates) while still looking identical at the macroscopic level, Boltzmann showed that entropy is simply:

S = k ln W

where k ≈ 1.380649 × 10⁻²³ joules per kelvin is what we now call Boltzmann's constant, fixed to that exact value by the 2019 redefinition of the SI base units. Entropy, in this view, is a count. It is the logarithm of the number of ways a system's molecules can be shuffled around while producing the same macroscopic picture — the same temperature, pressure, and volume you'd measure with a thermometer and a gauge.

A brand-new deck of cards, sorted by suit and rank, has exactly one arrangement that counts as "sorted." A shuffled deck can land in any of 52! ≈ 8.07 × 10⁶⁷ distinct orders. Every single one of those 8 × 10⁶⁷ shuffles is equally likely — the deck isn't being pulled toward disorder by any force. It just has astronomically more ways of being disordered than of being sorted, so a random shuffle almost always lands there.

That card analogy is the whole idea in miniature. Nothing in physics forbids a shuffled deck from spontaneously reassembling itself into suit-and-rank order, or a broken glass from reassembling itself off the floor. Both are allowed by the same laws that govern every other collision of atoms. They are just so overwhelmingly outnumbered by the disordered arrangements that no one has ever seen it happen, and — statistically — no one ever will, not in the lifetime of the universe. This is the sense in which "entropy increases" is really a statement about probability at an enormous scale, not a separate force pushing systems around.

It's worth being precise about a second analogy people reach for: entropy is not quite the same thing as "disorder" in casual conversation. A more accurate mental model is spread. A drop of ink in a glass of water starts concentrated in one place — few molecular arrangements look like "all the ink in one corner." As it diffuses, the number of arrangements that look like "ink evenly spread through the glass" is vastly larger, so that's where the system ends up. The ink isn't becoming "messier" so much as becoming more evenly distributed across the space of possibilities available to it.

Heat Engines and the Carnot Limit

A heat engine is any device that takes in heat Qₕ from a hot reservoir at temperature Tₕ, converts a fraction of it into useful work W, and dumps the remainder Qᶜ into a cold reservoir at temperature Tᶜ. Its efficiency is simply η = W / Qₕ — the fraction of the heat you put in that you got back as usable work.

In 1824, a 28-year-old French military engineer named Sadi Carnot published Réflexions sur la puissance motrice du feu (Reflections on the Motive Power of Fire), asking a question no one had answered rigorously: is there a hard ceiling on how efficient a heat engine can ever be, regardless of what working fluid it uses? Carnot showed there was, and derived it before the First or Second Laws were even formally written down. Only 600 copies of his book were printed, and Carnot died of cholera at 36 without seeing his work recognized, but the American Society of Mechanical Engineers now designates it a historic engineering landmark for laying the theoretical foundation of thermodynamics itself.

The limit he found, now called the Carnot efficiency, is:

η_Carnot = 1 − Tᶜ / Tₕ

with both temperatures measured in kelvin. This ceiling comes directly from the Second Law: some heat must always be exhausted to the cold reservoir to keep total entropy from decreasing, and the only way to raise the ceiling is to widen the temperature gap the engine works across.

Concrete numbers make this less abstract. A gasoline engine burning fuel at roughly 900 K and exhausting to roughly 300 K air has a theoretical Carnot ceiling near 67%; real automotive engines land closer to 25–35% because of friction, incomplete combustion, and the fact that a real cycle can't run infinitely slowly the way Carnot's idealized reversible cycle does. Modern combined-cycle natural gas power plants, which pair a gas turbine with a steam turbine to recapture waste heat, are the practical champions of this trade-off: manufacturers report thermal efficiencies above 62% under optimal conditions, according to GE Vernova's HA-class turbine specifications, by running hotter and recycling exhaust heat that older single-cycle plants simply vented. No design gets to 100%, because that would require Tᶜ = 0 K, and the Third Law says absolute zero is unreachable.

Refrigerators and heat pumps run this same cycle in reverse: instead of letting heat flow downhill from hot to cold and skimming off work, they spend work to force heat uphill from cold to hot. Their performance is measured by coefficient of performance (COP) rather than efficiency, and COP can exceed 1 — a household refrigerator typically moves 2 to 4 units of heat for every unit of electrical work it consumes, which is why heat pumps are more energy-efficient for building heating than resistive electric heaters that convert electricity to heat one-for-one. This isn't a violation of the Second Law. The compressor's electrical work still adds its own entropy to the universe, and that addition is always larger than the entropy the refrigerator locally subtracts from its own interior. You're not cheating the ledger, just paying a different account.

Why can't we build a perpetual motion machine?

A "perpetual motion machine of the first kind" would violate energy conservation by producing more work than the energy it consumes — ruled out by the First Law. A "perpetual motion machine of the second kind" would convert heat entirely into work with no waste heat and no rise in entropy anywhere — ruled out by the Second Law. Patent offices, including the US Patent and Trademark Office, have rejected perpetual motion machine applications for over a century on exactly this basis, and none has ever been demonstrated to work.

The Arrow of Time

Here's the puzzle that makes thermodynamics more than an engineering footnote. The equations governing individual particles — Newton's laws, Maxwell's equations of electromagnetism, the Schrödinger equation of quantum mechanics — are all time-symmetric. Film a single collision between two billiard balls, run the footage backward, and it looks like a perfectly valid physical event. There is nothing in the underlying microscopic laws that says time has to run one direction.

Yet on the macroscopic scale, time obviously has a direction. Eggs break but don't unbreak. Perfume disperses through a room but never spontaneously reconcentrates into the bottle. So where does that one-way arrow come from, if the underlying rules don't demand it?

The leading answer, first sketched by Boltzmann and developed further by physicists including Roger Penrose, is called the Past Hypothesis: the universe began in an extraordinarily low-entropy state near the Big Bang, and has been climbing toward higher entropy ever since. Time's arrow, on this view, just is the direction of entropy increase. We remember the past and not the future because the past was lower-entropy and left behind the physical records — photographs, fossils, the arrangement of neurons in your brain — that memory depends on; the future hasn't generated those records yet.

Penrose has estimated just how special that initial low-entropy state had to be: in Cycles of Time he calculates the odds at around 1 part in 10^(10^123) — a number so large that writing it in ordinary decimal notation is itself physically impossible within the observable universe. Why the universe started in such an improbable state is one of cosmology's genuinely open problems, tied up with questions covered in our explainer on why time moves slower near gravity and in our companion piece on entropy and the arrow of time.

Is the Second Law a hard rule or just very likely?

The Second Law is statistical, not absolute. For a macroscopic system of roughly 10²³ particles — a mole of gas, say — the odds of a spontaneous, observable entropy decrease are so vanishingly small that it has never once been recorded happening. On the scale of a handful of particles, tiny, temporary entropy dips are both allowed and measurable, which is exactly what fluctuation theorems in statistical mechanics describe.

Information and Shannon Entropy

In 1948, Claude Shannon, working at Bell Labs, defined a completely different-sounding quantity for measuring the information content of a message:

H = −Σ pᵢ log pᵢ

summed over every possible symbol, weighted by its probability pᵢ. Shannon called it "entropy" partly on the advice of the mathematician John von Neumann, who reportedly told him the term would guarantee he'd win any argument since nobody really understood what entropy meant anyway — an anecdote widely repeated in histories of information theory, including Scientific American's retrospectives on Shannon's work.

The name turned out to be more than a joke. Shannon's formula has exactly the same mathematical shape as Boltzmann's S = k ln W, up to a choice of constant and logarithm base. That's not coincidence: both are counting the number of equally likely possibilities a system could be in, whether those possibilities are molecular arrangements in a gas or characters in a message. Modern physics treats thermodynamic entropy and information entropy as two expressions of the same underlying quantity, and that equivalence turns out to have teeth, as the next section shows.

Maxwell's Demon and Landauer's Principle

In 1867, James Clerk Maxwell proposed a thought experiment designed to break the Second Law. Imagine a tiny, intelligent "demon" guarding a trapdoor between two chambers of gas at equal temperature. The demon lets fast-moving molecules through in one direction only, and slow ones through in the other, without doing any obvious work itself. Over time, one chamber gets hotter and the other colder — a temperature difference created from nothing, apparently decreasing the system's entropy for free.

It took nearly a century to fully resolve the paradox. In 1961, the IBM physicist Rolf Landauer showed that the demon's own memory is the missing piece: to sort molecules, the demon has to remember which ones it has already let through, and erasing that memory is not free. Landauer's principle states that erasing one bit of information in an environment at temperature T must dissipate at least kT ln 2 of heat into the surroundings — roughly 2.9 × 10⁻²¹ joules per bit at room temperature. When the demon eventually resets its memory to keep working, that erasure generates exactly enough entropy to balance the books, and the Second Law survives intact.

This wasn't just a thought experiment forever. In 2012, a team led by Antoine Bérut at the CNRS in France built a physical system — a single colloidal particle trapped by a laser — and measured the heat released when a bit of information stored in its position was erased. Their result, published in Nature, matched Landauer's predicted minimum almost exactly, confirming experimentally that information really is physical. Every computer that erases a memory register, every hard drive that overwrites a file, pays this same minimum thermodynamic tax, which is one reason chip designers researching reversible computing care about Landauer's limit as a hard floor on energy-efficient computation.

Black Holes: Entropy at the Extreme

If you want to see entropy pushed to its most extreme and strangest form, look at black holes. In the early 1970s, the physicist Jacob Bekenstein proposed something that seemed absurd at first: that a black hole has entropy, and that entropy is proportional not to its volume, as you'd expect for ordinary matter, but to the area of its event horizon. Stephen Hawking initially resisted the idea, then proved it himself by showing black holes emit faint thermal radiation — now called Hawking radiation — which gives them an actual temperature and confirms they really do carry entropy.

The resulting Bekenstein-Hawking formula is:

S = kc³A / (4Gℏ)

where A is the horizon's surface area, c is the speed of light, G is Newton's gravitational constant, and ℏ is the reduced Planck constant. Because this formula scales with area rather than volume, a black hole turns out to be the most entropy-dense object physics knows of — a useful clue in the ongoing effort to unify gravity with quantum mechanics, since it suggests that whatever the true microscopic description of spacetime is, it behaves in some sense like information smeared across a surface rather than packed through a volume. This "holographic" idea, first proposed by Gerard 't Hooft and developed by Leonard Susskind, remains an active area of theoretical research rather than settled physics, but the entropy-area relationship itself is well established and central to modern black hole thermodynamics, as summarized in the Scholarpedia entry on Bekenstein-Hawking entropy. For a sense of scale, see our explainer on the biggest black holes ever found — the largest known, at tens of billions of solar masses, carry entropy that dwarfs every other object in the observable universe combined.

Life as Local Entropy Reversal

If entropy always increases overall, how does life — visibly ordered, self-repairing, self-replicating — exist at all? The physicist Erwin Schrödinger, better known for the equation that bears his name, addressed this directly in his 1944 book What Is Life?, coining the term "negative entropy" (often shortened to negentropy) for what living organisms consume to stay organized. Plants absorb relatively low-entropy sunlight and radiate higher-entropy infrared heat back into space. Animals eat low-entropy, chemically complex food and excrete high-entropy waste — carbon dioxide, water, and body heat.

A living body is, moment to moment, a local decrease in entropy. That decrease is always paid for by a much larger entropy increase somewhere in its surroundings, ultimately traced back to the Sun, which converts low-entropy nuclear potential energy into high-entropy radiated heat and light at a staggering rate. The Second Law is never violated by a living cell, a growing plant, or a snowflake forming in a cloud; the accounting is just outsourced to a larger system. Think of it like a household budget: you can run a surplus in one account as long as a connected account runs an equal or larger deficit, and the total across both never goes positive.

Does a growing crystal or a refrigerator break the Second Law?

No. Both are local, ordered decreases in entropy that are more than offset by a larger entropy increase elsewhere — released latent heat for the crystal, electrical work converted to waste heat for the refrigerator's compressor. The Second Law only constrains the total entropy of an isolated system, never any single piece of it in isolation.

What People Get Wrong About Entropy

The single most common misconception is treating entropy as a synonym for "mess" or "chaos" in the everyday sense, as if there's some force actively working to wreck things. There isn't. Entropy increase is a statistical drift toward the overwhelmingly more probable arrangements, not a directed push. A tidy desk doesn't get messy because entropy "wants" it to; it gets messy because there are vastly more ways for objects on a desk to be scattered than to be neatly stacked, and small random disturbances — a draft, a bump, someone reaching for a pen — sample from that larger space of possibilities far more often than not.

A second common error is assuming the Second Law forbids order from ever appearing anywhere, which would make snowflakes, crystals, and living organisms mysterious violations rather than ordinary physics. As covered above, local order is entirely compatible with global entropy increase, provided the books balance elsewhere.

A third mix-up: people sometimes treat "entropy" and "energy" as interchangeable, since both show up in the same equations. They are not the same kind of quantity. Energy is conserved — the First Law says the total amount never changes. Entropy is not conserved; the Second Law says it only ever increases (or stays flat, in the idealized limit of a fully reversible process). A system can have exactly the same total energy before and after a process while its entropy has still increased, because that energy has become less useful for doing work — more spread out, more evenly distributed, harder to extract as directed motion.

Finally, some people assume evolution or the emergence of complex life somehow contradicts the Second Law, since organisms become more structurally complex over time. This confuses the entropy of a local, open system (Earth's biosphere, which constantly exchanges energy with the Sun and radiates waste heat to space) with the entropy of an isolated system, which the Second Law actually governs. Earth is not thermodynamically isolated, so nothing about biological complexity violates the law.

The Heat Death

Run the Second Law forward as far as physics currently allows, and you arrive at the heat death of the universe, sometimes called the "Big Freeze." If the universe keeps expanding and entropy keeps climbing without limit, every usable temperature gradient eventually flattens out. No temperature differences means no more heat engines, no chemistry driven by energy gradients, and no life as we understand it — just a cold, dilute, maximally spread-out state.

Current cosmological models, built on measurements of dark energy from missions like NASA's WMAP and ESA's Planck satellite, point toward an accelerating expansion that makes this outcome look likely on timescales of roughly 10¹⁰⁰ years, though such extreme extrapolations carry real uncertainty since they depend on the long-term behavior of dark energy, which is not yet fully understood. Along the way, even black holes eventually evaporate via Hawking radiation, a process so slow that a solar-mass black hole would take on the order of 10⁶⁷ years to fully evaporate. What's left after that, in these models, is an extremely thin, extremely cold sea of photons and particles, with no structure left to speak of.

FAQs

Q: Does entropy ever decrease locally? A: Yes. A refrigerator decreases the entropy inside its cold compartment, and a growing crystal decreases its own internal entropy. But the total entropy of system plus surroundings always increases or stays flat — it never drops when you add up the full ledger.

Q: Is the Second Law a hard law or a statistical tendency? A: It's statistical, not logically absolute. For a system of roughly 10²³ molecules, the probability of a spontaneous, observable entropy decrease is so small it has never been recorded happening, even though nothing in the underlying physics strictly forbids it.

Q: Why can't we build a 100%-efficient engine? A: Because some heat must be dumped to a cold reservoir to keep total entropy from decreasing. The Carnot limit, η = 1 − Tᶜ/Tₕ, is the mathematical statement of exactly how much heat has to be wasted, and it only reaches 100% if the cold reservoir is at absolute zero, which is unreachable.

Q: Does the universe's expansion affect entropy? A: Yes. As space expands and matter spreads out, the maximum possible entropy the universe could have also grows. The universe's actual entropy is climbing, but the theoretical ceiling is climbing too, which is part of why heat death hasn't been reached yet despite roughly 13.8 billion years of expansion.

Q: What's the connection between entropy and disorder? A: "Disorder" is a useful but imprecise analogy. More precisely, entropy counts the number of microscopic arrangements consistent with a given macroscopic state. A pile of bricks and a finished wall contain the same atoms, but the wall corresponds to far fewer possible arrangements, so it has lower entropy.

Q: Do black holes have entropy? A: Yes, and an enormous amount of it. The Bekenstein-Hawking formula shows black hole entropy is proportional to the area of the event horizon, not its volume, which makes black holes the most entropy-dense objects known to physics.

Q: Does life violate the Second Law? A: No. Living systems are open, not isolated — they take in low-entropy energy (sunlight, food) and expel high-entropy waste (heat, carbon dioxide). The local order of a living body is always paid for by a larger entropy increase in its surroundings.

Q: What is Maxwell's demon and why doesn't it break thermodynamics? A: It's an 1867 thought experiment about a hypothetical being that sorts fast and slow molecules to create a temperature difference for free. Landauer's principle resolves the paradox: erasing the demon's memory of which molecules it sorted requires a minimum amount of heat dissipation that exactly restores the entropy balance.

Q: Could time run backward in another universe? A: If a hypothetical universe began in a high-entropy state instead of a low-entropy one, entropy could still increase from its own starting point while looking, from our reference frame, like it was becoming more ordered — which would appear to run backward. This remains speculative and untestable with current physics, but it is not mathematically forbidden.

Q: Is entropy the same thing as heat? A: No. Heat is a transfer of energy due to a temperature difference; entropy is a measure of how many microscopic arrangements are consistent with a system's macroscopic state. They are related — Clausius originally defined entropy change as heat transferred divided by temperature — but they measure fundamentally different things.

The takeaway

Entropy is not a mysterious force dragging the universe toward ruin. It's a counting problem: systems drift toward the macroscopic states that have overwhelmingly more microscopic ways of being arranged, and that statistical drift is powerful enough to set the direction of time itself, cap the efficiency of every engine ever built, and even give black holes a temperature. Local order, whether it's a snowflake, a refrigerator, or a living cell, is never a violation of the Second Law, only a loan against a larger entropy increase paid somewhere else in the universe.

---

Read next:

• Entropy and the Arrow of Time, Explained
• Why Time Moves Slower Near Gravity
• Biggest Black Holes Ever Found

Sources

• Boltzmann, L. (1877). On the Relationship between the Second Law and Probability Theory.
• Clausius, R. (1865). On Several Forms of the Fundamental Equations of the Mechanical Theory of Heat.
• Carnot, S. (1824). Réflexions sur la puissance motrice du feu. Discussed via ASME Engineering History Landmark.
• GE Vernova. HA-Class Gas Turbine combined-cycle efficiency specifications. gevernova.com
• Shannon, C. E. (1948). A Mathematical Theory of Communication. Bell System Technical Journal.
• Schrödinger, E. (1944). What Is Life? Cambridge University Press.
• Penrose, R. (2010). Cycles of Time: An Extraordinary New View of the Universe. Bodley Head.
• Landauer, R. (1961). Irreversibility and Heat Generation in the Computing Process. IBM Journal of Research and Development.
• Bérut, A. et al. (2012). Experimental verification of Landauer's principle linking information and thermodynamics. Nature 483, 187–189. nature.com/articles/nature10872
• Bekenstein, J. D. (1973). Black Holes and Entropy. Physical Review D 7, 2333.
• Hawking, S. W. (1975). Particle Creation by Black Holes. Communications in Mathematical Physics 43, 199–220.
• Scholarpedia. Bekenstein-Hawking Entropy. scholarpedia.org/article/Bekenstein-Hawking_entropy
• NASA WMAP and ESA Planck mission results on cosmic expansion and dark energy. map.gsfc.nasa.gov, esa.int/Science_Exploration/Space_Science/Planck