Qubit

Definition: The basic unit of quantum information, analogous to a classical bit, but able to exist in a combination of 0 and 1 simultaneously through superposition rather than being strictly one or the other.

How It Works

A classical bit is definitely 0 or definitely 1. A qubit’s state is a vector in a two-dimensional complex space, written |ψ⟩ = α|0⟩ + β|1⟩, where α and β are complex numbers called amplitudes.
The amplitudes aren’t probabilities themselves. |α|² gives the probability of measuring 0, |β|² gives the probability of measuring 1, and normalization requires |α|² + |β|² = 1.
Measuring a qubit collapses it to a definite classical value. The superposition is destroyed the instant a measurement is taken, and you cannot inspect α and β directly, only the statistics of many repeated measurements reveal them.
A single qubit’s state can be pictured as a point on the surface of the Bloch sphere, with the north and south poles representing pure |0⟩ and |1⟩ and every other point a distinct superposition.
Physically realized on several competing hardware platforms: superconducting circuits, trapped ions, photonic qubits, and neutral atoms. Each trades off coherence time, gate speed, and qubit-to-qubit connectivity differently.
N qubits together describe a single state in a 2^N-dimensional space, not N independent 2-dimensional spaces. This exponential scaling is why simulating even a modest 50-qubit system is out of reach for classical supercomputers.
A qubit differs from a probabilistic classical bit (like a weighted coin) because its amplitudes can interfere, negative and positive amplitudes can cancel each other out, a phenomenon with no classical bit analogue.
The no-cloning theorem forbids making an independent copy of an unknown qubit state. This is not an engineering limitation, it’s a mathematical consequence of quantum mechanics being linear, and it shapes how quantum error correction and quantum communication protocols have to work.
A qubit’s state is manipulated by quantum gates, unitary operations that rotate the state vector on the Bloch sphere without collapsing it, chained together to form a quantum circuit.
Reading out a qubit’s value requires dedicated readout hardware, on superconducting chips this is typically a microwave resonator coupled to each qubit, and the readout itself introduces error that has to be characterized and corrected for separately from gate errors.
A “logical qubit” is an abstraction built from many physical qubits using a quantum error-correcting code, such as the surface code, so that the logical information survives even though individual physical qubits keep decohering.

Under the Hood

A qubit’s state is written as:

|ψ⟩ = α|0⟩ + β|1⟩,   |α|² + |β|² = 1

On the Bloch sphere, the same state is parameterized by two angles:

|ψ⟩ = cos(θ/2)|0⟩ + e^(iφ) sin(θ/2)|1⟩

Given: a qubit prepared as |ψ⟩ = (1/√2)|0⟩ + (1/√2)|1⟩. Step: measure it in the computational basis. Answer: P(0) = |1/√2|² = 0.5, P(1) = 0.5. One run of the measurement returns one definite bit, the 50/50 split only shows up statistically across many repeated preparations and measurements.

Given: two independent qubits, both initialized to |0⟩. Step: write the combined system as a tensor product. Answer: |00⟩ = |0⟩⊗|0⟩ is a 4-dimensional vector [1,0,0,0]ᵀ over the basis {|00⟩,|01⟩,|10⟩,|11⟩}. In general, n qubits require 2^n amplitudes to describe: 10 qubits need 1024, 50 qubits need over 10^15.

Given: a candidate state with α = 0.6 and β = 0.8. Step: check whether it’s a valid quantum state. Answer: 0.6² + 0.8² = 0.36 + 0.64 = 1.0, so it’s normalized and valid. Measuring it gives 0 with 36% probability and 1 with 64% probability.

Given: a fully classical-looking qubit state, |ψ⟩ = 1|0⟩ + 0|1⟩. Step: apply a Hadamard gate, which maps |0⟩ to (1/√2)(|0⟩ + |1⟩). Answer: the qubit leaves its definite state and enters equal superposition, this single-gate operation is the standard way circuits initialize superposition before running an algorithm.

Given: a system of 3 qubits, all initialized to |0⟩, after each is put through a Hadamard gate. Step: compute the resulting state. Answer: the state becomes an equal superposition of all 8 basis states, (1/√8)(|000⟩+|001⟩+…+|111⟩), each with probability 1/8 ≈ 12.5% on measurement. This is exactly the starting point Grover’s Algorithm uses before its search iterations begin.

Given: a real superconducting qubit with a measured T1 (energy relaxation) time of 100 microseconds. Step: compare that to a two-qubit gate that takes roughly 200 nanoseconds to execute. Answer: roughly 500 such gates could in principle run before the qubit’s stored energy decays away, in practice usable circuit depth is lower once gate errors and T2 dephasing are also accounted for, which is why algorithms are judged by circuit depth as much as by qubit count.

Why It Matters

Qubits are the entire reason quantum computers can potentially solve certain problems dramatically faster than classical computers: superposition and entanglement together let algorithms exploit a structured combination of possibilities before collapsing to one answer.
Qubit count alone is a poor proxy for capability. A processor’s usable power depends on gate fidelity, coherence time, and connectivity between qubits as much as raw count.
The distinction between a noisy physical qubit and a fault-tolerant logical qubit drives most near-term hardware roadmaps, since useful algorithms like Shor’s Algorithm need error-corrected logical qubits, not raw physical ones.
Understanding qubits is the prerequisite for understanding every other quantum computing concept: Superposition, Entanglement, and quantum gates are all defined in terms of what they do to qubit states.
Government and industry roadmaps (IBM’s, for example) are measured largely in qubit quality metrics like quantum volume and error-per-gate, not raw qubit counts, precisely because raw counts alone tell you little about what the machine can actually run.
Cloud-accessible qubits let researchers and students run real experiments on physical hardware today, without owning a dilution refrigerator, which has accelerated algorithm research even while fault-tolerant machines are still years away.

Common Pitfalls

Thinking of a qubit as “a bit that’s secretly both 0 and 1 at once” in a literal sense. The state is a genuine probability amplitude, not a hidden classical value waiting to be revealed.
Assuming more qubits automatically means more useful computation. A 1000-qubit chip with short coherence times and noisy gates can be less useful than a 50-qubit chip with high-fidelity gates.
Confusing “qubit” with “logical qubit.” Current hardware runs noisy physical qubits directly; a fault-tolerant logical qubit needs many physical qubits for error correction, often estimated in the hundreds to low thousands per logical qubit.
Believing a single qubit stores more classical information than one bit. Even though its state needs two complex numbers to describe, measurement only ever yields one classical bit, this is a consequence of the no-cloning theorem and Holevo’s bound.
Forgetting that amplitudes can be negative or complex, which enables interference. Treating α and β as if they were plain probabilities themselves (rather than probability amplitudes) misses how quantum algorithms cancel out wrong answers.
Expecting a qubit to be a drop-in performance upgrade for arbitrary classical code. Quantum advantage only shows up for specific problem structures, most everyday programs get no benefit from running on a qubit-based machine.
Assuming coherence time and gate fidelity are fixed hardware properties rather than moving targets. Both improve yearly across every major platform, so comparisons between vendors go stale fast.
Assuming a qubit needs to be exotic or mysterious to reason about mathematically. Once you accept the vector-and-amplitude formalism, qubit behavior follows ordinary linear algebra, unitary matrices multiplying state vectors, with no extra mysticism required.

Comparison

	Classical bit	Qubit	Qutrit
Possible states	0 or 1	Superposition of \|0⟩ and \|1⟩	Superposition of 3 basis states
Description needed	Single value	Complex amplitude pair (α, β)	Complex amplitude triple
Measurement result	Always deterministic	Probabilistic, collapses to 0 or 1	Probabilistic, collapses to one of 3
N-unit state space size	N (linear)	2^N (exponential)	3^N (exponential)
Physical noise sensitivity	Negligible	Extremely high	Extremely high
Can be perfectly copied	Yes	No, no-cloning theorem	No
Error correction overhead	None needed	Hundreds to thousands of physical qubits per logical qubit	Similar, higher-dimensional codes still maturing

Physical Implementations

Platform	Qubit carrier	Typical coherence time	Notable vendors
Superconducting	Josephson junction circuit loop	Microseconds to ~1 millisecond	IBM, Google, Rigetti
Trapped ion	Individual charged ion in an EM trap	Seconds or longer	IonQ, Quantinuum
Photonic	Polarization or path of a single photon	Effectively unlimited, but hard to store	Xanadu, PsiQuantum
Neutral atom	Individual atom held in optical tweezers	Milliseconds to seconds	QuEra, Pasqal
Topological (experimental)	Anyons whose braiding encodes information	Theoretically very high, still unproven at scale	Microsoft

Example

IBM’s Eagle and Heron processors use superconducting transmon qubits, cooled in a dilution refrigerator to roughly 15 millikelvin, colder than deep space. Google’s Sycamore chip, used in the 2019 Quantum Supremacy claim, also used superconducting qubits arranged in a 2D grid, and its 2024 successor Willow demonstrated error rates that dropped as more physical qubits were added to a logical qubit, a long-sought milestone.

IonQ and Quantinuum instead trap individual ions, such as ytterbium or calcium atoms, with electromagnetic fields and manipulate their internal energy levels with laser pulses. Trapped-ion qubits tend to have longer coherence times than superconducting ones but slower gate operations, a tradeoff that shapes which algorithms each platform is best suited to run today.

Cloud access to real qubit hardware is available today through platforms like IBM Quantum Experience and Amazon Braket, letting developers run small circuits on physical qubits rather than only simulating them classically.

Microsoft’s Azure Quantum takes a different bet entirely, pursuing topological qubits based on anyons, a still-experimental approach aimed at qubits that are intrinsically more resistant to decoherence, at the cost of being far harder to fabricate reliably.

Referenced by

Dig deeper

Spot an error in this note?