Reference

Glossary

Terminology from every tutorial, in one alphabetical reference. Click any entry with a linked tutorial to jump into the deep dive. Entries with a marker have a deeper definition.

Terms: 130
With deep definitions: 14

A

Amplitude ·: The complex coefficient of a basis state in a quantum superposition; its squared magnitude is the Born-rule probability.
For |ψ⟩ = Σ αx|x⟩ with Σ|αx|² = 1, αx is the amplitude associated with computational-basis state |x⟩. Amplitudes interfere — adding them before squaring lets quantum algorithms cancel wrong answers and amplify right ones, the source of every quantum speedup.
→ See: Born rule
Amplitude amplification ·: Generalized form of Grover's algorithm that boosts the amplitude of any 'good' subspace from sin θ to near 1 in O(1/√p) iterations.
Brassard, Høyer, Mosca, Tapp (2002). The 'quadratic speedup primitive' — most quadratic quantum-vs-classical wins (Grover, counting, Monte Carlo via amplitude estimation) are amplitude-amplification instances.
→ See: Grover tutorial · Amplitude Amplification in the Zoo
Amplitude estimation: Quantum algorithm to estimate the success probability p of a circuit A to precision ε in O(1/ε) calls — quadratic improvement over O(1/ε²) classical Monte Carlo. → See: Amplitude Estimation in the Zoo
Ancilla: A helper qubit introduced to make an otherwise irreversible computation reversible, then (usually) uncomputed back to |0⟩ for reuse.
Ansatz ·: A parameterized quantum circuit used as a variational model — e.g., in VQE, QAOA, or QML classifiers.
Choosing the ansatz family is the dominant design decision in variational QC: hardware-efficient ansätze are easy to run but suffer barren plateaus; problem-inspired ansätze (UCCSD, ADAPT-VQE) are harder to run but train better.
→ See: VQE
Approximate QFT (AQFT): Quantum Fourier Transform with the smallest controlled-rotation angles dropped — saves depth at the cost of approximation error that QPE tolerates.
Arbitrary-state preparation ·: The problem of preparing an arbitrary n-qubit state from |0⟩^n. Generic preparation requires O(2^n) gates — a major obstacle for QML and any algorithm assuming QRAM.
Lower bounds (Plesch-Brukner 2011) and matching constructions show 2^n is essentially tight for unstructured states. Structured states (matrix product states, computational basis states, sparse states) admit much cheaper preparation.
→ See: The data-loading bottleneck

B

Barren plateau ·: A pathology of variational quantum algorithms: gradients vanish exponentially with qubit count, making optimization intractable.
McClean, Boixo, Smelyanskiy, Babbush, Neven (Nature Comms 2018). The variance of partial derivatives ∂⟨H⟩/∂θ for sufficiently random parameterized circuits scales as 2^{-O(n)}. The result is foundational — mitigations (local cost functions, warm starts, problem-structured ansätze) help at small scale but don't fix the asymptotic.
→ See: Barren plateaus
Bell state: One of the four maximally entangled two-qubit states (|Φ±⟩, |Ψ±⟩). The 'hello world' of entanglement. → See: Entanglement
Bell's theorem: Any local-hidden-variable theory must obey the CHSH inequality |⟨S⟩| ≤ 2; quantum mechanics violates it up to 2√2 (Tsirelson bound). Experimentally confirmed (Aspect 1982; Hensen et al. 2015 loophole-free). → See: Bell's theorem
Bloch sphere ·: Geometric representation of a single-qubit state as a point on the surface of a unit sphere; mixed states sit inside.
|ψ⟩ = cos(θ/2)|0⟩ + e^{iφ}sin(θ/2)|1⟩. North pole = |0⟩, south pole = |1⟩, equator = computational-basis-equal superpositions with phase φ. Mixed states lie on the interior; rotations under unitary gates trace closed paths on the sphere.
→ See: The Bloch sphere
Block encoding ·: A unitary U whose top-left block (after projecting on a known ancilla state) is a scaled copy of a matrix A — the input model for QSVT and modern Hamiltonian simulation.
If U = [[A/α, *],[*, *]] then U is an (α, m, ε)-block encoding of A where m is the number of ancillas and ε bounds the error. Replaces 'sparse-matrix oracle' as the dominant matrix-access abstraction in 2026.
→ See: Block encoding · Block encoding in the Zoo
Born rule: The postulate that the probability of measuring outcome |x⟩ from |ψ⟩ = Σ αx|x⟩ equals |αx|².
Boson sampling: Sampling task for n indistinguishable photons in a linear interferometer; classically hard under permanent-of-Gaussians conjecture. → See: Boson sampling in the Zoo
BQP: Bounded-error Quantum Polynomial time — the class of problems solvable on a quantum computer in polynomial time with ≤1/3 error probability. Contains BPP; relation to NP unknown.

C

Cat state: Superposition of two macroscopically distinguishable states; in qubit contexts often (|00...0⟩ + |11...1⟩)/√2 (GHZ state on n qubits).
CHSH inequality: |⟨A0B0⟩ + ⟨A0B1⟩ + ⟨A1B0⟩ - ⟨A1B1⟩| ≤ 2 (classical), ≤ 2√2 (quantum). The standard test of Bell nonlocality. → See: Bell's theorem and CHSH
Circuit depth: The number of sequential gate layers in a quantum circuit. Limited by coherence time × gate speed on real hardware.
Clifford gate ·: A gate in the group generated by H, S, and CNOT. Circuits made entirely of Clifford gates are efficiently classically simulable (Gottesman-Knill).
Clifford gates normalize the Pauli group: CPC† is a Pauli string for any Pauli P. Universal quantum computation requires at least one non-Clifford gate (typically T = diag(1, e^{iπ/4})), and that gate's count (T-count) dominates fault-tolerant cost.
→ See: Clifford group
CNOT / CX: Controlled-NOT — flips the target qubit iff the control is |1⟩. The canonical two-qubit entangling gate.
Coherence time (T₁, T₂) ·: T₁ is energy-relaxation time; T₂ is dephasing time. T₂ ≤ 2T₁ always. Set the maximum useful circuit depth on real hardware.
T₁ measures how fast an excited qubit relaxes to ground. T₂ measures how fast superposition phase decoheres. Spin-echo can extend T₂ toward 2T₁ but cannot exceed it. Trapped-ion qubits have T₁ in seconds; transmons in tens to hundreds of microseconds.
→ See: Noise and decoherence
Color code: Topological QEC code on a 3-colorable lattice; distinguished by admitting transversal Clifford+T gates on certain 3D variants.
Continuous-variable QC (CV): Quantum computation with infinite-dimensional 'qumode' systems (light fields, mechanical modes). Used in Xanadu's photonic platform and Gaussian boson sampling.
Controlled-U: A two-register gate: applies U to the target iff the control is |1⟩. Generalizes CNOT (where U = X) and CZ (where U = Z). → See: Controlled-U synthesis
Cross-entropy benchmarking (XEB): Fidelity metric for random circuit sampling: estimate fidelity from the ratio of measured-vs-ideal output probabilities, sample by sample.

D

Density matrix (ρ) ·: A positive-semidefinite trace-1 operator generalizing state vectors to mixed states. Rank 1 ⇔ pure state.
For an ensemble {(pi, |ψi⟩)}, ρ = Σ pi |ψi⟩⟨ψi|. All physical predictions (expectation values, evolution under channels, measurement statistics) generalize naturally. Tr(ρ²) ≤ 1 with equality iff ρ is pure.
→ See: Density matrices
Dequantization ·: Producing a classical algorithm matching a quantum algorithm's asymptotic complexity — e.g., Ewin Tang's recommendation-system result (2018).
Tang's insight: many QML 'exponential speedups' assumed QRAM input access. Given classical sample-and-query access to the same data, classical algorithms can match the quantum asymptotic, with polynomial overhead. Most claimed exponential QML speedups have been dequantized.
→ See: Tang dequantization · Tang's dequantization in the Zoo · See Paper Verdicts
Deutsch-Jozsa: First quantum algorithm with provable exponential separation in the deterministic oracle model. Constant-vs-balanced in one query. → See: Deutsch-Jozsa
Dilation: Embedding a non-unitary operation (channel, measurement) into a unitary on a larger Hilbert space. Stinespring dilation is the canonical construction.
Distillation: Process for producing low-error 'magic states' (T-eigenstates) from many noisy ones. Dominant cost of fault-tolerant non-Clifford operation. → See: Magic state distillation

E

Eastin-Knill theorem: No QEC code admits a continuous group of transversal logical gates that is universal. Forces non-trivial constructions (magic-state distillation, code switching) to get a universal fault-tolerant gate set. → See: Eastin-Knill theorem
Eigenstate: A state |ψ⟩ such that U|ψ⟩ = λ|ψ⟩ for some scalar λ. For unitary U, |λ| = 1.
Entanglement: A quantum state that cannot be written as a tensor product of single-qubit states. The resource behind most quantum speedups. → See: Entanglement tutorial
Entanglement entropy: von Neumann entropy S(ρA) = -Tr(ρA log ρA) of the reduced state on subsystem A. Quantifies entanglement for pure bipartite states.
Error correction (QEC): Encoding one logical qubit into many physical qubits so errors can be detected and fixed. Surface code is the dominant scheme. → See: Surface code
Error mitigation: Post-processing techniques (zero-noise extrapolation, probabilistic error cancellation, Clifford data regression) that reduce bias in expectation values without full QEC. NISQ-era workhorse.

F

Fault-tolerant quantum computing (FTQC): Regime where QEC suppresses logical error rate exponentially with code distance. Requires per-operation error below the threshold (~1% for surface code).
Fidelity ·: A measure of how close two quantum states or operations are. 1 = identical; 0 = orthogonal. Gate error = 1 − fidelity.
For pure states F(|ψ⟩, |φ⟩) = |⟨ψ|φ⟩|². For mixed states F(ρ, σ) = (Tr√(√ρ σ √ρ))². Process fidelity for gates uses average-over-input-states or worst-case (diamond-norm) variants.
Flag qubit: An ancilla qubit used during stabilizer measurement to detect when an error during the measurement itself would have caused a bad correction. Key to small-distance fault tolerance.
Fusion-based quantum computing (FBQC): Photonic architecture using small resource states fused together via Bell measurements; the model PsiQuantum builds on. Tolerates ~10% loss before logical errors blow up.

G

Gate-set tomography (GST): Self-consistent characterization of a complete gate set — recovers Pauli transfer matrices for every gate without assuming any reference is perfect. → See: Gate-set tomography
Gauge transformation (in QEC): Stabilizer that is *not* enforced — the code's logical state is invariant under it. Subsystem codes have non-trivial gauge groups; Bacon-Shor and color codes use this for cheaper measurement.
Gaussian boson sampling (GBS): Variant of boson sampling using squeezed-light inputs. The setting Xanadu's Borealis and USTC's Jiuzhang use.
GKP state: Gottesman-Kitaev-Preskill encoding of a qubit into a continuous-variable mode. Discretizes phase space; tolerates small displacement errors. Key to bosonic-code FTQC (PsiQuantum, AWS, Yale).
Grover's algorithm ·: Quantum unstructured search in O(√N) queries, provably optimal.
Grover (1996). Quadratic, not exponential — and real-world search has structure classical solvers exploit. Best treated as a primitive (use it inside amplitude amplification) rather than a 'database speedup'.
→ See: Grover tutorial · Grover in the Zoo

H

Hadamard (H): Maps |0⟩ → |+⟩ and |1⟩ → |−⟩. The gate that creates superposition out of basis states.
Hamiltonian simulation: Implementing the time-evolution operator e^{-iHt} for a Hamiltonian H. Trotter / qubitization / QSVT are the main approaches. → See: Hamiltonian simulation · Hamiltonian Simulation in the Zoo
Hardware-efficient ansatz: Parameterized circuit using only the native gates and connectivity of a given device. Trains badly at scale (barren plateaus) but runs deep on noisy hardware.
Harvest now, decrypt later: Attack model: adversary records encrypted traffic today and decrypts it once a Shor-capable quantum computer exists. Motivates near-term PQC migration. → See: HNDL · PQC threat model
Heron: IBM's tunable-coupler superconducting chip family (133/156 qubits). Heron r2 (2024) is the highest-quality IBM device available on the free tier.
HHL algorithm: Quantum algorithm to output |x⟩ ∝ A^{-1}|b⟩ for sparse well-conditioned A in Õ(log N · κ · s · 1/ε). Subject to four caveats that rarely all apply (Aaronson 2015). → See: HHL linear systems · HHL in the Zoo
Hidden subgroup problem (HSP): Generalization of period-finding (Shor) and graph-isomorphism. Solved efficiently on quantum computers for abelian groups; non-abelian case has resisted progress.
Holographic algorithm: Classical algorithm inspired by matchgate / fermionic tensor structures; can simulate certain quantum circuits efficiently. Sometimes blurs the 'classical / quantum' boundary on specific problems.

I

Imaginary-time evolution: Use of the non-unitary operator e^{-Hτ} to project onto the ground state. Implemented variationally (McLachlan, time-dependent variational principle) or via QSVT. → See: Imaginary-time evolution
Ising model: Hamiltonian H = -Σ Jij Zi Zj - Σ hi Zi. Native to D-Wave annealers; the canonical 'easy-to-state, hard-to-solve' combinatorial-optimization formulation.
Iterative QPE (IQPE / Kitaev's QPE): Phase-estimation variant that uses a single ancilla qubit and classical post-processing, instead of a t-qubit ancilla register + inverse QFT. NISQ-friendlier than textbook QPE.

J

Jordan-Wigner transformation ·: Standard mapping from fermion creation/annihilation operators to Pauli strings on qubits, used in quantum chemistry simulations.
ci = (Π_{j<i} Zj) · (Xi - iYi)/2. The Z-string locality is poor — Bravyi-Kitaev mapping reduces it to O(log n). Fenwick-tree refinements bring further improvements.

K

Ket (|ψ⟩): Dirac notation for a state vector in a complex Hilbert space.
Kraus operators: A decomposition of a quantum channel ε(ρ) = Σ Kk ρ Kk† with Σ Kk†Kk = I. The general language of noisy quantum evolution.

L

Lattice surgery: Technique for performing logical operations on surface-code qubits by merging and splitting patches. Lower overhead than transversal gates for most logical operations.
LCU (Linear Combination of Unitaries): PREP / SELECT / PREP† construction to apply a linear combination Σ αi Ui of unitaries probabilistically. Backbone of modern Hamiltonian simulation. → See: LCU · LCU in the Zoo
Logical qubit: A qubit encoded across many physical qubits using an error-correcting code. The surface code uses ~d² physical qubits for one logical qubit at distance d.

M

Magic state: An ancilla state that, when consumed by a circuit, enables a non-Clifford operation (typically the T gate). Costs T-count × distillation overhead in FTQC.
Magic state distillation: Process for preparing low-error T-eigenstates (ancillas for non-Clifford gates) by sacrificing many noisy ones. Dominant cost in fault-tolerant computing. → See: Magic state distillation
Majorana zero mode: Non-Abelian quasi-particle at the ends of certain topological superconducting wires. The physical substrate Microsoft bets on for topological qubits. The 2018 Nature paper claiming detection was retracted in 2021. → See: See Paper Verdicts
Measurement basis: The set of eigenstates against which a measurement projects. Z-basis = {|0⟩, |1⟩}; X-basis = {|+⟩, |−⟩}; Y-basis = {|i⟩, |−i⟩}.
Measurement-based QC (MBQC): Computation by adaptive single-qubit measurements on a fixed entangled resource (cluster) state. Equivalent in power to the circuit model. Native to photonic and some neutral-atom platforms.
ML-DSA (FIPS 204): NIST-standardized lattice-based digital signature — replaces RSA/ECDSA/Ed25519.
ML-KEM (FIPS 203): NIST-standardized lattice-based key encapsulation — replaces RSA/ECDH in PQC migrations. → See: ML-KEM in practice

N

NISQ: Noisy Intermediate-Scale Quantum era — roughly 50–10,000 qubits without full error correction (Preskill 2018). Today.
No-cloning theorem: There is no unitary U with U|ψ⟩|0⟩ = |ψ⟩|ψ⟩ for arbitrary |ψ⟩. Forces all quantum communication and error correction to live with this constraint. → See: No-cloning theorem
Noise channel: A completely-positive trace-preserving (CPTP) map describing how a quantum state is corrupted between operations. Depolarizing, dephasing, amplitude-damping are the canonical models.

O

OpenQASM 3: Industry-standard quantum assembly language. Supports parameterized circuits, classical control flow, mid-circuit measurement, gate decompositions. → See: OpenQASM and real hardware
Oracle: A black-box unitary Uf |x⟩|y⟩ = |x⟩|y ⊕ f(x)⟩ that lets a quantum algorithm query a Boolean function f reversibly.

P

Parameter-shift rule: Exact identity for derivatives of expectation values in parameterized circuits: ∂⟨H⟩/∂θ = ½[⟨H⟩θ+π/2 − ⟨H⟩θ−π/2]. Makes gradient-based QML training work on real hardware. → See: Parameter-shift rule
Pauli matrices: The three 2×2 matrices X, Y, Z that generate all single-qubit rotations and form the backbone of quantum error-correction stabilizer codes.
Phase estimation (QPE): Recovers the eigenvalue phase φ of a unitary given an eigenvector, using O(1/ε) controlled applications to reach precision ε. → See: QFT + QPE · QPE in the Zoo
Phase kickback: When an oracle acts on |−⟩ in the target register, the output f(x) appears as a phase (−1)^{f(x)} on the input register. Central to Deutsch-Jozsa, Grover, Shor.
Photonic quantum computing: Quantum computing using single photons (KLM-style), squeezed-light modes (CV), or fusion-based architectures (FBQC). Room-temperature, naturally networkable; loss is the dominant practical limit. → See: Photonic QC
Post-quantum cryptography (PQC): Cryptographic primitives (lattice, code, hash, isogeny based) believed secure against quantum computers with Shor's algorithm.
Probabilistic error cancellation (PEC): Error-mitigation technique using a noise model to construct a quasi-probability distribution that cancels expected noise. Exponential sample-cost overhead — limits practical scale.
Projector: An operator P with P² = P, corresponding to a measurement outcome; for |φ⟩, P = |φ⟩⟨φ|.
Purification: Embedding a mixed state ρ on system A as a pure state |ψ⟩ on A ⊗ B such that ρ = Tr_B(|ψ⟩⟨ψ|). Always exists; the dilation theorem in disguise.

Q

QAOA: Quantum Approximate Optimization Algorithm (Farhi-Goldstone-Gutmann 2014). p alternating problem+mixer layers; classical-optimizer-tuned parameters. No demonstrated advantage on real workloads. → See: QAOA · QAOA in the Zoo
QFT — Quantum Fourier Transform: The quantum analog of the discrete Fourier transform, implemented in O(n²) gates. The engine of Shor, QPE, and many algorithms. → See: QFT tutorial · QFT in the Zoo
qLDPC code: Quantum low-density parity-check codes. Constant-rate alternatives to surface code; promise much lower qubit overhead for large-scale FTQC. → See: qLDPC codes
QRAM: Quantum random-access memory — hypothetical device for preparing |ψ⟩ = Σ √pi |i⟩|xi⟩ in O(log N) time. Required for most claimed exponential QML speedups; no scalable QRAM exists.
QSVT: Quantum Singular Value Transformation. Applies polynomial transformations of singular values of block-encoded matrices. Unifies HHL, AE, Grover, Hamiltonian simulation. → See: QSVT · QSVT in the Zoo
Quantum advantage: Demonstrated practical superiority of a quantum computer over the best classical algorithm on a specific task. Distinct from 'quantum supremacy' (any task) and 'quantum utility' (vague).
Quantum channel: A CPTP map ε : ρ → ε(ρ) describing physical evolution including noise. Most general quantum operation.
Quantum kernel: Kernel function K(xi, xj) = |⟨φ(xi)|φ(xj)⟩|² with quantum feature map φ. Used in quantum SVMs; no exponential speedup on natural data. → See: Quantum kernels · Quantum kernels in the Zoo
Quantum supremacy: Originally Preskill's term for any task on a quantum computer that classical computers cannot perform in feasible time. Now mostly rebranded as 'quantum advantage' due to social/political baggage of the original term.
Quantum volume (QV): IBM-defined benchmark: largest m such that the device can run random m×m circuits with >2/3 heavy-output probability. Single-number quality metric; less commonly cited in 2026 than per-gate error.
Quantum walks: Quantum analog of classical random walks on a graph. Backbone of element distinctness, spatial search, and several QSVT-based algorithms. → See: Quantum walks · Quantum walks in the Zoo
Qubit: The quantum unit of information: α|0⟩ + β|1⟩ with |α|² + |β|² = 1.

R

Randomized benchmarking (RB): Self-consistent gate-error characterization: apply random Clifford sequences of varying length, fit decay constant to extract average error per Clifford. Insensitive to SPAM. → See: Randomized benchmarking
Reverse adiabatic: Adiabatic evolution backward from a known eigenstate of the target Hamiltonian to a trivial Hamiltonian — useful for state preparation in variational algorithms.
Rydberg blockade: Strong dipole-dipole interaction between nearby Rydberg-excited atoms preventing simultaneous excitation. Mechanism for two-qubit gates in neutral-atom systems (QuEra, Atom, Pasqal).

S

Sampling, classical: Generating bitstrings from a target distribution. The complexity-theoretic basis for boson sampling and random-circuit-sampling 'advantage' claims.
Shor's algorithm: Factors N-bit integers in time polynomial in N, using quantum phase estimation on a modular-exponentiation unitary. → See: Shor tutorial · Shor in the Zoo
Solovay-Kitaev theorem: Any single-qubit unitary can be approximated to ε by a sequence of O(log^c(1/ε)) gates from a finite universal gate set. Constructive — gives a practical decomposition algorithm. → See: Solovay-Kitaev
Stabilizer: A Pauli operator that leaves a codeword unchanged. Surface codes are defined by mutually commuting X- and Z-stabilizers.
Stinespring dilation: Theorem that every CPTP map can be implemented as a unitary on a larger system plus a partial trace. The mathematical foundation of quantum-channel theory.
Sub-system code (gauge code): QEC code with a non-trivial gauge group: stabilizers that don't have to be measured. Bacon-Shor and topological color codes use this for cheaper syndrome extraction.
Superconducting qubits: Charge / flux / phase / transmon variants. Transmons dominate today (IBM, Google, Rigetti); fab uses standard semiconductor processes at sub-Kelvin temperatures. → See: Transmon qubits
Superposition: A linear combination α|0⟩ + β|1⟩ (or higher-dimensional analog) of basis states. → See: Superposition tutorial
Surface code: 2D topological error-correcting code on a qubit lattice with local stabilizer measurements. Current experimental leader (see Willow). → See: Surface code
Syndrome extraction: Measuring the stabilizer generators of a QEC code to detect (without disturbing) which error occurred. Decoder uses the syndrome to apply a correction.

T

T gate: The π/8 phase gate. Non-Clifford. Its count (T-count) dominates fault-tolerant-circuit cost.
T-count / T-depth: Total number of T gates / depth of T-gate layers in a circuit. Primary cost metric for fault-tolerant compilation; entire research subfield optimizes for these. → See: Toffoli decomposition and T-count
Tensor network: Classical representation of quantum states (MPS, PEPS, MERA) factorizing into low-rank tensors. Extremely effective for low-entanglement systems; major classical competition for many QC tasks.
Threshold theorem: Below a code-dependent physical error threshold (~1% for surface code), increasing code distance reduces logical error rate exponentially.
Toffoli (CCNOT): Three-qubit reversible AND gate. Universal for classical reversible computation; costs 6 CNOTs + 7 T-gates in the surface code.
Topological qubit: Qubit encoded in topological degrees of freedom (anyons, Majorana modes) that are robust to local perturbations by topology. Microsoft's long-running bet.
Transmon: Charge-insensitive superconducting-qubit variant by Koch et al. (2007). The dominant superconducting-qubit design — IBM, Google, Rigetti all use transmons.
Trapped-ion qubits: Qubits encoded in hyperfine or optical levels of trapped atomic ions (Ca⁺, Yb⁺, Ba⁺). Long T₁ (seconds), low gate error, all-to-all connectivity. Quantinuum, IonQ, AQT. → See: Trapped-ion QC
Trotter error: First-order Trotter formula (Π e^{-iHk Δt})^r approximates e^{-iHt} with error O(t²/r) for r steps. Higher-order Suzuki recursion gives O((t/r)^{p+1}) for p-th order.
Twirling: Averaging a noise channel over a unitary 2-design (e.g., Clifford group) to convert arbitrary noise into a depolarizing channel. Used in randomized benchmarking and Pauli twirling for error mitigation.

U

UCCSD ansatz: Unitary Coupled Cluster Singles-and-Doubles. Physics-motivated chemistry ansatz for VQE; expressive but high-depth (O(N⁵) gates for N orbitals).
Unitary: A linear operator U with U†U = UU† = I. Every non-measurement quantum operation is a unitary.
Universal gate set: A set of gates that approximates every unitary to arbitrary precision in a bounded number of applications. {H, T, CNOT} is the canonical fault-tolerant universal set.

V

Variational quantum algorithm (VQA): Hybrid quantum-classical algorithm: parameterized circuit on hardware, classical optimizer in the loop. VQE, QAOA, and most QML methods are VQAs.
Von Neumann entropy: S(ρ) = -Tr(ρ log ρ). Quantum analog of Shannon entropy; quantifies mixedness and bipartite entanglement.
VQE: Variational Quantum Eigensolver — hybrid classical/quantum algorithm for estimating ground-state energies of quantum systems. → See: VQE tutorial · VQE in the Zoo

W

Warm start: Initialization strategy for variational algorithms: seed parameters from a classical heuristic (e.g., Goemans-Williamson rounding for QAOA on MaxCut) to escape barren-plateau-prone random initialization. → See: Warm-start strategies
Willow: Google's 105-qubit superconducting chip (Dec 2024) — first experimental demonstration of below-threshold surface-code error correction. → See: Surface code + Willow

X

XEB (Cross-entropy benchmarking): Fidelity metric: log-cross-entropy of measured-vs-ideal sample probabilities. Standard for random-circuit-sampling experiments.

Y

Y2Q: Shorthand for the quantum apocalypse — the year a Shor-capable quantum computer breaks RSA. Commonly used in PQC migration planning.

Z

Zero-noise extrapolation (ZNE): Error-mitigation technique: run circuits at intentionally amplified noise levels, extrapolate to zero noise. Polynomial sample overhead; modest practical impact at NISQ scale.
ZX-calculus: Graphical language for quantum circuits. Rewriting rules let you simplify, verify, and optimize circuits at the diagram level. → See: ZX-calculus