Is quantum computing just
quantum mechanics?

Not the same thing, but not a different theory either. The word means exactly what it means in physics, down to the letter.

You'll be able to say what separates quantum computing from quantum mechanics, and track a qubit on the Bloch sphere.

The rulebook and the machine

Quantum mechanics is the rulebook. Worked out in the 1920s, it describes how very small things behave, and they behave nothing like the objects we have intuitions about. A particle can be in a blend of two states at once. Looking at it changes it. Two particles can be linked so that neither has a definite state on its own. None of this is a metaphor or a gap in our knowledge. Its most stringently tested piece, quantum electrodynamics, predicts the electron's magnetic moment, and prediction and measurement agree to about one part in a trillion. That is among the most precisely checked predictions anyone has ever made about anything.

For more than fifty years, that rulebook was used mainly to explain things: why the sun shines, why metals conduct, why chemistry works. The machine came later. In the early 1980s a few people asked a stranger question: not "what does nature do?" but "what could I compute, if I built something that runs on these rules by design?"

The difference lives in the question. A quantum computer adds no new physics: it is a very carefully built quantum-mechanical system whose behaviour you have chosen, so that the answer you want falls out at the end. Same rulebook. New use.

One thing worth saying plainly, because the word is so abused: in "quantum computing," quantum means quantum. It is the same word physicists use, pointing at the same phenomena in a real refrigerated chip. It is not true of "quantum healing" or a quantum-branded shampoo, where the word is pure decoration. "Quantum leap" is the interesting middle case: it does come from something real (an electron jumping between energy levels) but popular use inverted the meaning into "an enormous change," when the actual jump is discrete and indivisible (the system cannot make a smaller, gradual change), which is what "quantum" meant in the first place. Quantum computing is the case where the label stayed literal.

One state, two questions

Here is the whole relationship in a single object. Below is a Bloch sphere, the standard picture of one qubit. Every pure state of a single qubit is a point on that surface, and every quantum gate is a rotation of it.

Press the gates and watch the arrow move. Then read the two panels underneath: the same state, described the way a physicist would and the way a programmer would. Neither panel is a translation of the other; they are the same fact, asked about differently.

Now let it decohere

The sphere above is the perfect case: every gate is a clean rotation and the arrow never leaves the surface. A real qubit is not left alone — it leaks energy to its surroundings and its phase smears out — so the arrow spirals inward. Two numbers set the clock: T₁, the time to fall back toward |0⟩, and T₂, the shorter time for the phase (the part that does the computing) to wash out. Set them, pick a starting state, and press play.

The same postulates, renamed

The clearest way to see that quantum computing is not a separate theory is to lay the two vocabularies side by side. Every item in the right column is the item on its left: the same mathematical object, given a job title.

That fourth row deserves a correction that the popular version always skips. It is often said that 2ⁿ amplitudes are why quantum computers beat classical ones. That can't be the full explanation, because there is a large family of quantum circuits (stabilizer circuits, built from a specific gate set) that live in exactly the same exponential space and yet can be simulated efficiently on a laptop (the Gottesman–Knill theorem). Enormous state space is necessary for a quantum advantage; it is nowhere near sufficient. Whatever the advantage is, it is subtler than counting amplitudes.

Nothing was added. The postulates were written down by 1932 and quantum computing uses them unchanged, which is why a physics graduate can read a quantum-algorithms paper without learning new physics, only new goals.

So what did quantum computing add? Three things, none of them physical laws. First, the circuit model: a discipline of building any evolution out of a small reusable gate set. Second, complexity: the machinery for asking how the cost of a computation grows with problem size, which is a computer-science question that physics simply had never posed about physical systems. Third, and most usefully, error correction: the realisation that a fragile quantum state can be protected by encoding it across many physical carriers. That last one has since become a physics tool in its own right.

Where does interference come in? A quantum computer is not a machine that "tries every answer at once"; that phrasing is the most common way the subject is oversold. Superposition alone gets you nothing, because measurement hands back a single random outcome. The work is done by interference: a good algorithm arranges the amplitudes of wrong answers to cancel and the right ones to reinforce, so that the final measurement is likely to land where you want. Designing that cancellation is the actual craft, and it is why useful quantum algorithms are rare and hard to find rather than automatic.

The part the sphere cannot draw

The widget above is one qubit, and one qubit is as far as that picture can go. Entanglement needs two, and it is where the difference between quantum mechanics and any classical story stops being a matter of interpretation and becomes something you can win a bet with.

Here is that bet, in its cleanest form: the CHSH game, from Clauser, Horne, Shimony and Holt in 1969. Two players, Alice and Bob, are separated and cannot communicate. A referee sends each of them a random bit. Each must answer with a bit of their own. They win the round if their two answers differ only when both received a 1.

They may agree on any strategy beforehand. Here is the theorem: if the world is local and classical (Alice's answer depending only on her own bit and whatever they agreed in advance) then no strategy on earth wins more than 75% of rounds. No amount of cleverness gets past it, because the ceiling is a theorem. Sharing an entangled pair lifts the ceiling to 85.4%. Play both and watch.

This exact statistic has a job outside the game. Swap the referee's win condition for "do these two measurement settings still correlate the way only entanglement can explain," and the same 75%-classical-ceiling-vs-higher-observed-rate logic is the security proof behind the Ekert91 quantum key distribution protocol, covered in full on the post-quantum cryptography page, including why national cybersecurity agencies still recommend against relying on it in place of the mathematics that page is about. The game is also a game in the sense of the Feasible Region's game-theory topic 32, which computes all 16 classical strategies and the quantum value exactly.

The formal statement, and its caveats

Take the standard postulates. A pure state of a register is a unit vector |ψ⟩ ∈ ℋ ≅ (ℂ²)^⊗n, dim = 2ⁿ. Closed evolution over time t is |ψ(t)⟩ = U(t)|ψ(0)⟩ with U = exp(−iHt/ℏ) for Hamiltonian H = H^†. A projective measurement in basis {|k⟩} returns k with probability |⟨k|ψ⟩|², leaving the state in |k⟩.

A quantum circuit is nothing but a factorisation of one such U into a product of local unitaries:

The Solovay–Kitaev theorem says any single-qubit unitary can be approximated to accuracy ε by O(log^c(1/ε)) gates from a fixed universal set closed under inverses (c < 4; see Dawson & Nielsen, quant-ph/0505030). So "programming" a quantum computer means choosing an H you can physically implement whose induced U factorises into gates you can run. It is engineering constrained by the Hamiltonian your hardware can produce: the physics is a given, not a variable.

The complexity layer is the new part. BQP is the class of decision problems solvable by a uniform family of polynomial-size quantum circuits with error ≤ 1/3. The known containments are

and here is the caveat the field's marketing usually omits: not one of those inclusions has been proven strict. (Not that they all should be: most complexity theorists expect P = BPP outright, on derandomisation grounds. The point is that the chain is a ladder of open questions, not a ladder of established gaps.) In particular, BQP ≠ BPP has never been demonstrated, and proving it would imply P ≠ PSPACE, settling a famous open problem along the way. Shor's algorithm factors in polynomial time (quant-ph/9508027), but "exponential speedup" there is measured against the best known classical algorithm, not against a proven classical lower bound; no one has ruled out a fast classical factoring algorithm. What we do have is proven separation relative to an oracle, and the two standard results are not the same size, which is worth stating precisely rather than lumping them together. Bernstein & Vazirani's recursive Fourier sampling gives a superpolynomial gap; Simon's problem gives a fully exponential one (Bernstein & Vazirani, SIAM J. Comput. 26, 1411 (1997); Simon, SIAM J. Comput. 26, 1474 (1997)). To those add, in 2018, an unconditional separation between constant-depth quantum circuits and constant-depth classical circuits of bounded fan-in (Bravyi, Gosset & König, Science 362, 308; arXiv 1704.00690). Those are real theorems about restricted models, not the general claim.

The loop runs backwards too. The traffic has not been one-way from physics into computing. Treating entanglement as a quantifiable resource, and stabiliser codes as objects in their own right, changed how physicists work: the toric code is simultaneously an error-correcting code and an exactly solvable model of topological order (Kitaev, quant-ph/9707021); entanglement entropy is now a standard diagnostic of quantum phases; and tensor-network methods born in quantum-information language are routine in condensed-matter numerics. In high-energy theory, the Ryu–Takayanagi result ties entanglement entropy to geometry (hep-th/0603001), and quantum error correction has become an explicit part of how the AdS/CFT correspondence is understood (Almheiri, Dong & Harlow, arXiv 1411.7041). So the accurate picture is a cycle rather than a hierarchy: physics supplied the laws, computing supplied a new set of questions about them, and the answers changed physics.

This is also where this site's subject sits. AI decoding quantum errors is not a new physical principle either: it is a statistical-inference technique aimed at a problem that quantum mechanics created and cannot solve on its own.

References

Go deeper