Qubits & States
Q1What is it?
A qubit is a two-level quantum system — the smallest thing quantum mechanics lets you store information in. A classical bit is a switch: at any moment it reads 0 or 1, and that value is the state. A qubit's state is not a value. It's a vector:
The complex numbers α and β are probability amplitudes. Measure the qubit in the computational basis and you get 0 with probability |α|² and 1 with probability |β|² — and the state jumps to match the answer. That asymmetry is the whole subject in one sentence: the state is described by amplitudes; measurement produces probabilities. Everything strange and useful about quantum computing lives in the gap between those two.
For a single qubit there is a picture worth keeping. Any pure state can be written with just two real angles,
so the states live on the surface of a sphere — the Bloch sphere — with |0⟩ at the north pole, |1⟩ at the south, equal superpositions on the equator, and the relative phase φ sweeping around it. Gates become rotations of this sphere; choosing a measurement basis means choosing an axis.
The caveat that took me a while to appreciate: the sphere is a complete picture of one qubit in a pure state, and nothing more. Mixed states fall into the sphere's interior and need density matrices; several qubits together need a state space the sphere can't draw at all. The picture is a ladder — climb it, then know when to let go.
Q2Why does it exist?
Because probabilities aren't enough, and here is the shortest experiment that proves it. Take a qubit in |0⟩ and apply a Hadamard gate:
Measure now and you get a fair coin — 0 half the time, 1 half the time. So suppose the qubit were merely a random bit. Applying H again would be randomizing a coin flip: still a coin flip. But that's not what happens. Apply H a second time, without measuring in between, and:
The amplitudes headed for |1⟩ arrive with opposite signs and cancel; the ones headed for |0⟩ reinforce. No amount of probability bookkeeping predicts that — you need amplitudes that carry phase and can interfere. That is why the qubit is defined the way it is, and interference is the resource every quantum algorithm is built to spend. The mental model I'm carrying forward into algorithms:
"A qubit is 0 and 1 at the same time" — useful as a first rung, but it's the coin-flip model, and the coin-flip model just failed the two-Hadamard test. Better: a qubit holds a coherent superposition, described by complex amplitudes whose phases matter.
Q3How does it connect?
Upstream, the qubit is linear algebra wearing physics: states are unit vectors, gates are unitary matrices, and amplitudes live in the complex plane, with probability theory waiting at the measurement. Sideways, it's the object the neighboring nodes act on — superposition is the structure of the state itself, measurement converts it into classical information, and gates rotate it. Downstream is where the field opens up: tensor products glue qubits together, n of them spanning 2n basis states,
and joint states appear that cannot be factored into a state per qubit — entanglement, which is where the sphere picture finally breaks and the real machinery begins.
One caution belongs on this node and not later: having 2n amplitudes does not mean a quantum computer runs 2n classical computations in parallel. You measure once and get one outcome. Advantage, when it exists, comes from choreographing interference so wrong answers cancel — which is the Q2 story, scaled up.
Q4Where is it used?
Everywhere — it's the atom of the field, so the honest answer is "see every other node." The sharper question is where the abstraction earns its keep. "Two-level quantum system" is a spec, not a device: superconducting circuits, trapped ions, neutral atoms, photons, and semiconductor spins all implement it, and the mathematics upstairs doesn't care which. That indifference is what lets algorithm designers and hardware physicists work the same problem from opposite ends.
The gap between the mathematical qubit and its physical implementation is also where all the trouble lives: decoherence, control error, measurement error, and eventually the whole Noise & Error Correction map — that entire territory is the story of defending this node's clean abstraction against physics. It's why a vendor's qubit count is the least interesting number on the spec sheet: the mathematical qubit is what they're counting, the physical qubit is what they have, and the distance between the two is what they're selling.
To make it concrete: build a one-qubit circuit in Qiskit, apply X, Y, Z, and H, and watch the counts. Then run the two-Hadamard experiment from Q2 yourself — it's three lines, and it's the cheapest genuine quantum surprise available.
My open questions
- What is the most useful mental model for phase — geometry, clocks, or something better?— chasing
- Why exactly does the Bloch sphere completely characterize a single-qubit pure state?— can compute, can't yet feel
- What changes, conceptually, when a state vector becomes a density matrix?— next node
- How should entanglement change the way I think about an individual qubit?— open
- Which properties of the ideal qubit survive contact with a noisy physical one?— bridges to the noise map
Explore further
- 3Blue1Brown, Essence of Linear Algebra — visual grammar for vectors, transformations, and complex numbers— intuition
- Nielsen & Chuang, Quantum Computation and Quantum Information, ch. 1–2— rigor
- Qiskit — one qubit, four gates, and the two-Hadamard experiment— hands on
- Bloch Cubes — hands-on Bloch-sphere learning: videos, a buildable cube, even Bloch Cube Chess— play