Quantum Volume Metric: Benchmark & Beyond

• Quantum Volume Metric is a scalable, architecture-agnostic measure that quantifies a quantum device’s capacity to execute random circuits with high fidelity using the heavy-output probability threshold.
• It encompasses volumetric generalizations (QV-k) to capture varying circuit depths, thus broadening its relevance to diverse algorithms such as quantum chemistry and Shor’s algorithm.
• Incorporating hardware performance, error mitigation techniques, and simulation of quantum geometric invariants, the metric informs performance across superconducting, photonic, and other quantum architectures.

The quantum volume metric provides a scalable, architecture-agnostic quantification of the computational capabilities or “size” of quantum computing devices and, separately, the geometric and topological structures in fields such as condensed matter physics, quantum gravity, and quantum information. The term has distinct, context-specific characters—most notably, as a randomized-circuit benchmark for noisy quantum hardware (NISQ devices), as a volume measure in geometric quantum state manifolds, and as a holographic complexity dual in AdS/CFT. Across these domains, the quantum volume metric encodes how dimensionality, error rates, topology, and experimental protocol interact to bound or enable quantum computational or structural complexity.

1. Quantum Volume for Quantum Computing Devices

Quantum Volume (QV), as established in Cross et al. and subsequent literature, is a single-number metric that quantifies the largest random quantum circuit of equal width and depth that a quantum device can implement with high-fidelity—specifically, such that the observed heavy-output probability (HOP) exceeds 2/3. The operational definition is as follows (Cross et al., 2018, Wack et al., 2021, Niekerk et al., 2024, Baldwin et al., 2021, Pelofske et al., 2023, LaRose et al., 2022):

1. For an $n$-qubit quantum processor, random circuits of width $n$ and depth $n$ are constructed. Each layer consists of randomly paired qubits and independent Haar-random SU(4) gates; between layers, random permutations (for routing) are applied.
2. For each circuit, the ideal output probabilities $p(z)$ are evaluated (classically, via simulation) for all $2^n$ bitstrings $z$.
3. The heavy-output set $H$ is defined for each circuit as those $z$ with $p(z) >$ median($p(z)$).
4. The quantum processor executes many such random circuits and records, for each, the fraction of measurement outcomes in $H$.
5. Averaged over circuits, the device is said to "pass" for size $n$ if mean HOP $\geq 2/3$ (precisely: often the criterion is mean HOP $> 2/3$ by $2\,\sigma$ statistical margin).
6. The quantum volume is defined as $V_Q = 2^{n_\text{max}}$, where $n_\text{max}$ is the largest $n$ passing the test (Cross et al., 2018, Wack et al., 2021).

Table: Quantum Volume Measurement Protocol (Condensed)

Step	Description	Reference
Random Circuit	Square $n \times n$ layers, Haar-random SU(4), random pairing	(Cross et al., 2018)
Heavy Output Test	$H$ = outcomes with $p(z) >$ median; HOP averaged over circuits	(Niekerk et al., 2024)
Success Criterion	Mean HOP $> 2/3$ (possibly plus $2\,\sigma$)	(Baldwin et al., 2021)
Metric	$V_Q = 2^{n_\text{max}}$	(Miller et al., 2022)

In the noise-free limit, the ideal heavy-output probability is not 1 but converges to $(1+\ln 2)/2 \approx 0.8466$ for Haar-random states (Niekerk et al., 2024, Baldwin et al., 2021). Increasing circuit width or depth causes the HOP to decay toward $1/2$ in the presence of noise, making $2/3$ an effective threshold in NISQ regimes.

2. Volumetric Generalizations and Quantum Volumetric Classes

The standard QV metric is restricted to "square" circuits (depth = width), but this does not reflect the circuit-depth scaling of most important quantum algorithms. To address this, the concept of Quantum Volumetric Classes (QV-$k$) was introduced (Miller et al., 2022, Rodenburg, 31 Jan 2025). Here, each class matches a distinct asymptotic depth scaling:

• QV-1: depth $\sim n$ (original metric, e.g., shallow circuits)
• QV-2: depth $\sim n^2$ (quantum chemistry, data-hiding)
• QV-3: depth $\sim n^3$ (Shor's algorithm, quantum simulation)
• QV-4/5: depth $\sim n^4$, $n^5$ (complex algorithms, high-order simulation)

For a given $k$, the metric is

$$\text{QV-}k = \max_{n \le n_\mathrm{max}}\, \min\bigl[n,\,d(n)^{1/k}\bigr], \quad d(n) \sim n^k.$$

Thus, QV-$k$ encodes the maximal number of qubits on which a device can execute random circuits of depth $n^k$ with HOP exceeding 2/3, making the metric relevant for a broader spectrum of algorithmic complexity (Miller et al., 2022, Rodenburg, 31 Jan 2025).

Device limitations due to connectivity, gate error, and error correction are analytically tied to QV-$k$ as follows (Rodenburg, 31 Jan 2025):

• Effective error per SU(4) gate: $\epsilon \simeq 7\,\epsilon_1 + 3\,\epsilon_2$ (with $\epsilon_1$, $\epsilon_2$ single-, two-qubit errors).
• Layer error, routing overhead, and the connectivity exponent $m$ enter as $\epsilon_\mathrm{eff} = n^m \epsilon$; $m = 0$ (all-to-all), $0.5$ (2D grid), $1$ (1D).
• $n_\mathrm{opt} = \epsilon_\mathrm{eff}^{-1/(k+1)}$, then QV-$k = \min(n_\mathrm{max},n_\mathrm{opt})$.
• For error-corrected devices, logical error and code overhead are included in the effective error and qubit count.

3. Quantum Volume and Error Mitigation

Quantum Volume is a "full-stack" metric: both hardware improvements (longer coherence, lower error) and software (compilation, error mitigation) can increase measured QV. Recent work has established that error mitigation (e.g., zero-noise extrapolation, Richardson's method) can strictly raise the effective QV without modifying hardware (LaRose et al., 2022, Pelofske et al., 2023).

• Zero-noise extrapolation is applied to the expectation value of heavy-output projectors.
• Error-mitigated HOP is computed via extrapolation over stretched-noise circuits; if the zero-noise HOP crosses 2/3 for a higher $n$, the "effective quantum volume" (eQV) increases accordingly (Pelofske et al., 2023).
• Results on IBM five-qubit devices confirm that ZNE increases eQV by one to two circuit sizes over unmitigated QV (LaRose et al., 2022).

This establishes QV's role as a system-wide metric sensitive to all layers of the computation stack (physical, compilation, measurement, data movement, classical feedback, and error mitigation).

4. Quantum Volume in Alternative Quantum Architectures

The QV benchmarking framework has been adapted for non-circuit-based and alternative quantum computing platforms:

• Photonic and MBQC architectures: In measurement-based photonic processors, physical noise models (e.g., loss, finite squeezing in GKP states) are analytically mapped to logical error channels. The effective QV is computed via the same heavy-output criterion, but per-gate errors are parametrized by device- and architecture-specific noise processes (Zhang et al., 2022).
• Room-temperature NV centers and neutral atoms: QV is used as a unifying benchmark for registers based on NV centers with all-to-all connectivity (Jaeger et al., 2024), incorporating randomized benchmarking-extracted fidelities into a simulated noise model.
• QV is also used for classical quantum circuit simulators, quantifying not fidelity but simulation time at target QV (circuit size) (Niekerk et al., 2024).

5. Quantum Volume Metric in Quantum Geometry and Topology

In the context of condensed matter, quantum geometry, and quantum field theory, “quantum volume” can refer to the volume form induced by the quantum metric (Fubini-Study metric) on parameter spaces such as the Brillouin zone or twist-angle tori (Mera et al., 2021).

• For a map $P : T^2 \to \mathrm{Gr}_r(\mathbb{C}^n)$ defined by the Fermi projector, the pullback of the canonical metric gives a quantum metric $g_{ij}(k)$ on $T^2$.
• The quantum volume is $$V_g(T^2) = \int_{T^2} \sqrt{\det g_{ij}(k)}\,\mathrm{d}^2k$$.
• A fundamental result is the topological lower bound $V_g(T^2) \geq \pi |\mathcal{C}|$, where $\mathcal{C}$ is the first Chern number (topological charge) of the occupied bundle; equality obtains when the pullback admits a Kähler structure (Mera et al., 2021).
• This minimal volume condition is physically tied to the stability of fractional Chern insulator states—flatness and isotropy of the quantum metric and Berry curvature maximize the stability of such correlated phases.

6. Quantum Volume Metric in AdS/CFT and Quantum Information

In holographic duality and quantum gravity, the quantum information metric $G_{\lambda\lambda}$ (fidelity susceptibility) in a boundary CFT is proposed to be dual to the maximal bulk time-slice volume $V_\mathrm{max}$ in asymptotically AdS spacetimes (Sinamuli et al., 2016).

• The path-integral quantum information metric quantifies the quadratic response of ground-state fidelity to marginal perturbations.
• In the BTZ/CFT$_2$ or planar AdS/CFT correspondence,

$$G_{\lambda\lambda} = c\,\frac{V_\mathrm{max}}{L^{d+1}},$$

with $c$ a normalization constant set by the CFT two-point function.

• For topological quotients (geons), the relation is modified by a factor of 4: $c_\mathrm{geon} = 4c_\mathrm{BH}$, reflecting the sensitivity of the quantum information metric to spacetime topology (Sinamuli et al., 2016).
• This geometric “quantum volume” metric is physically distinct from the circuit-based QV but reflects a related principle: complexification and accessibility of the quantum state space.

7. Limitations, Caveats, and Outlook

• Quantum Volume, especially the standard QV-1, targets random-circuit performance and may not directly predict real-world algorithm execution, particularly for highly structured circuits (Niekerk et al., 2024, Baldwin et al., 2021).
• The metric is sensitive to all layers from hardware error rates to compiler routing, but its value can be "gamed" via aggressive optimization of only the test circuit shape.
• Generalized volumetric classes (QV-$k$) and adaptations to different architectures (e.g., MBQC, qudit systems) improve representativity but complicate cross-platform comparisons (Miller et al., 2022, Rodenburg, 31 Jan 2025).
• In geometric/topological contexts, the quantum volume is a structural invariant, strongly linked to curvature, holomorphicity, and topological charge, directly influencing stability and ideality in correlated quantum phases (Mera et al., 2021).
• In quantum gravity, the quantum information metric’s bulk-boundary correspondence provides a diagnostic for quantum complexity and topology.

Quantum volume metrics, in all contexts, encapsulate the tension between size, error, topology, and control, providing an operational bound on the “effective capacity” of quantum systems—from NISQ processors to topological phases and holographic spacetimes.