物理 · Physics

What Is a Physical Law

Day 1 · 2026 · Phase A — the Skeleton of Classical Mechanics
Before learning any specific physics, get clear on what the word "law" even means — and why symmetry is the deepest single idea in all of physics.
Most people think physics is a pile of formulas. The deepest physicists see it completely differently: a law is a compression of regularity, it holds scale by scale, it often grows out of symmetry, and whether it is "the law of the universe" or "a map we drew" is still an open question. This issue teaches no specific phenomenon — it just installs a lens. From force to the quantum to the cosmos, you'll look through it. Four cards: symmetry→conservation, effective theory & scale, reduction vs emergence, and what a "law" really is.

Symmetry → Conservation Noether, 1918

Symmetry · 1918
Intuition Run the same experiment today and tomorrow — same result. This is time-translation symmetry: shift "when" and the physics is unchanged. In 1918 the mathematician Emmy Noether found something astonishing: behind every "shift-it-and-nothing-changes" hides a quantity that never changes. Shift in time → energy is conserved. Shift in space → momentum is conserved. Not a coincidence — a theorem.
Mechanism Noether's theorem: every continuous symmetry of a system corresponds to a conserved quantity Q (energy, say), satisfying
dQdt = 0
Read as "the rate of change of Q is exactly zero." Here d means "a tiny change": dQ is a tiny change in Q, dt is a short slice of time, and their ratio dQ/dt is "how fast Q changes over time" (its derivative). Setting it to the number 0 (zero) means Q does not change with time at all.
The correspondence is exact:
Continuous symmetry Conserved quantity Time translation (any when) Energy E Space translation (any where) Momentum p Rotation (any direction) Angular momentum L
Each "change-it-and-nothing-changes" corresponds to one quantity that is "conserved."
The counterintuitive point We usually assume we first observe energy conservation and then explain it. Noether flips the causation: energy is conserved because the laws of physics do not change over time. A conservation law is not an extra "rule" the universe decrees — it is the shadow a symmetry casts. That upgrades "why is energy conserved" from an empirical regularity to a near-geometric necessity.
Cross-domain · AI Inductive bias / equivariance in machine learning is the same wisdom: CNNs are translation-equivariant, equivariant nets are rotation-equivariant — the physics intuition "symmetry → fewer parameters" baked into architecture. Symmetry = a free prior; physics and deep learning both use it to shrink the search space.
In one line: a conservation law is the shadow of a symmetry.
Think: if some quantity stopped being conserved in the very early universe, which symmetry would you go looking for as broken?
Look for which symmetry broke. Energy non-conservation ↔ broken time-translation symmetry — exactly the case in an expanding universe: spacetime itself changes with time (the scale factor grows), so cosmological "total energy" really isn’t strictly conserved. A lost conservation law always points to a failed symmetry.

Effective Theory & Scale Effective Theory

Scale · Effective theory
Intuition A bridge engineer never uses quark equations. Newtonian mechanics is actually "wrong" (relativity and quantum are more accurate), yet it builds bridges and launches rockets just fine. How can a wrong theory still work? Because each scale has its own good-enough laws — you can get things right at one layer without knowing the layer below.
Mechanism This is the idea of an effective theory: at a given length/energy scale, the finer microscopic detail is "averaged out," leaving only that scale's effective degrees of freedom and effective laws. Nature is layer-wise approximable — each layer is insensitive enough to the one below, and that is exactly what makes science possible.
small large scale quarks / particles QFT · Standard Model atoms / molecules quantum mechanics · chemistry materials / fluids condensed matter · thermo bridges / planets Newtonian ← good enough galaxies / cosmos general relativity · cosmology
One universe, ruled by different "effective laws" at different scales — Newton never fails inside his own box.
Why it matters This explains why physics is even possible: if you had to understand quarks before computing a planet's orbit, science couldn't move. It also clarifies what "wrong" means — Newton isn't wrong, he's valid within his scale and fails beyond it. No law claims to apply at all scales.
Cross-domain · Engineering You write Python without minding transistors, use distributed systems without minding every TCP byte — layered abstraction is the lifeblood of software. Physics's "effective theory" and engineering's "abstraction layer" are the same organizing principle: each layer exposes a good-enough interface and hides the details below.
In one line: there is no "universal law," only "the right law at the right scale."
Think: is a large model's scaling law a kind of "effective theory"? What detail has it averaged out?
Yes. It averages away individual weights, specific data points, and training trajectories, keeping only the regularity among macro quantities — "parameters · data · compute → loss" — just as thermodynamics ignores single molecules and speaks of temperature and pressure. Scaling laws are deep learning’s thermodynamics.

Reduction vs Emergence More is Different

Emergence · More is Different
Intuition Solve the equations of water molecules all the way down — can you "read off" whirlpools, wetness, even life? No. In 1972 the physicist P. W. Anderson wrote four words — "More is Different": pile up many parts and wholly new levels and laws appear, ones you can't simply deduce from the rule for a single part.
Mechanism This is emergence. Temperature, phase transitions, superconductivity, consciousness — all collective phenomena, appearing only in the "many," and often insensitive to microscopic detail (that's universality — the "Phase Transitions & Criticality" issue works it out via the renormalization group). The key distinction: reductionism (everything decomposes to basic particles) is true — but "constructionism" (rebuild everything from particles and predict all higher phenomena) often fails.
The counterintuitive point "Knowing all the basic laws = understanding everything" is an illusion. The second half of physics (condensed matter, complex systems) is precisely about emergence. It is the core of this site's "Condensed Matter & Emergence" theme, and the philosophical bedrock of complexity science and consciousness research.
Cross-domain · Physics / Neuro / AI / Society The same thing recurs in four fields — a property absent in the parts appears once there are enough of them:
  • Physics: a single water molecule has no "wetness" or "whirlpool"; "liquid" and "freezing" exist only for trillions together;
  • Neuro: a single neuron just fires — no "seeing red"; perception and consciousness emerge from ~86 billion neurons acting together;
  • AI: a single weight has no "language understanding"; multi-step reasoning and in-context learning appear "suddenly" once model scale crosses a threshold (the "emergent abilities" debate);
  • Society: no single car is "in a traffic jam" — congestion is a phase transition of the flow (individuals unchanged, the collective state flips); prices, language, fashion the same.
So "More is Different" is one and the same proposition across physics, neuroscience, AI, and society.
In one line: reductionism holding does not mean constructionism holds.
Think: is an LLM's "emergent ability" the same kind of emergence as a physical phase transition — or just an artifact of how we measure?
Debated — probably not the same. Physical transitions have rigorous order parameters and scaling laws; many "emergent abilities" have been argued to be a measurement artifact — a hard pass/fail metric creates the jump, while a smooth metric shows steady growth. Test: real emergence should be metric-independent.

What a "Law" Really Is The Ontology of a Law

Philosophy of science · Ontology
Intuition Is a law a rule the universe "obeys," or our compressed description of regularity? A planet doesn't first "read" Newton's equations and then decide where to go. So is a law an external decree, or a map we drew?
Mechanism · two views ① Descriptivism (Humean): a law is just the simplest compression of existing regularities, with no "compulsion" of its own — the best summary of the universe's patterns.
② Nomic realism: laws are real, governing structures; the universe "is thus because of them."
Above both hovers Wigner's puzzle of "the unreasonable effectiveness of mathematics": why does abstract math describe physics so precisely?
Why it matters This is no idle metaphysics — it shapes what you mean by "why." When you ask "why is energy conserved," do you want decree-style compulsion, or a symmetry-style compression? The deeper physics goes, the more our "laws" look like highly compressed maps rather than the universe's statute book.
Cross-domain · Philosophy / AI This is the old question in the philosophy of science — are laws discovered or invented; it echoes the adage "the map is not the territory." It is also isomorphic to an ML creed: a model is not truth, it is a useful compression — all laws and models are compressions, differing only in how well they compress and where they fail.
In one line: a law is a compression, not a decree.
Think: if a "law" could never be violated and never be checked outside observation, does it differ from a definition?
Almost none — which is descriptivism’s sharpest point. A "law" that makes no falsifiable prediction collapses into a rule about word usage (a definition), not a claim about the world. Falsifiability is the watershed between a law and a definition.

Going Deeper

Could symmetry be something we "impose" on nature? We love symmetry — is nature really symmetric, or do we just look at the symmetric bits?
A healthy doubt — but physics answers the opposite way. Nature is full of spontaneous symmetry breaking: a magnet is equivalent in all directions when hot (symmetric), yet on cooling "chooses" one magnetization direction (broken). Precisely because we can name "which symmetry broke, and what order it produced," symmetry is a real starting point, not after-the-fact prettification. The existence of breaking is the strongest evidence that symmetry was real (the "Symmetry Breaking & Order" issue goes into this).
If "More is Different," then even a found "Theory of Everything (TOE)" — what could it deliver, and what not?
First, what a TOE is: a Theory of Everything is the ultimate theory that would unify all of nature's four fundamental forces (gravity, electromagnetism, the strong and weak nuclear forces) and every basic particle into one framework. The Standard Model already unifies the latter three; only gravity has resisted merging with quantum mechanics — string theory and loop quantum gravity are attempts, none experimentally confirmed.
Even if a TOE were found, it would give the deepest "alphabet" — a great reductionist victory; but it would not automatically tell you how turbulence flows, how proteins fold, or how consciousness arises. Those are emergent-level questions; the answers live in how the "many" are organized, not in the particle list. A TOE is both an ending and a beginning: it settles "what is at the bottom," yet barely touches half of physics (emergence).
If each scale has its own laws, does a "bottom layer" really exist — or is it turtles all the way down?
An open question. One possibility: a fundamental scale exists (e.g. the Planck scale), below which there is no "more microscopic." Another: effective theories nest forever, always a next layer, and "fundamental" just means "the deepest we've probed." Intriguingly, the effective-theory framework doesn't need the answer — it's designed so that "whatever is below, this layer stays self-consistent," which is both its power and its humility.
Wigner's puzzle: is math's effectiveness a property of the universe, or an adaptation our brains evolved?
Both camps have defenders. Universe side: the world has mathematical structure and we're discovering it (mathematical Platonism). Brain side: our math grew out of experience with this world, so of course it "happens" to fit — like a shoe fits because it was made to the foot. Perhaps the truth is in between: math is a human tool, but it works because the universe really does have compressible regularity. No consensus — but worth chewing on as you learn each law.

Further Reading