# Techniques: Entropy Density Benchmarking of Near-Term Quantum Circuits

## What the paper contributes (one paragraph)
The paper introduces entropy density accumulation as a benchmark for noisy quantum processing units, designed to bridge circuit-level and application-level benchmarking. The methodology has three steps: build a simple heuristic model of how Renyi-2 entropy density accumulates in a variational circuit, validate and refine it against real hardware (Rigetti's Aspen-M-3 superconducting QPU), then use it to bound the circuit volume above which quantum advantage is unattainable for a given optimisation problem. The authors derive a global depolarising heuristic model that links circuit width, depth, and local gate-noise parameters to a single global depolarising probability, and fit it to both simulation and experiment. Entropy density is estimated experimentally using the classical shadows protocol with Pauli measurements. Combining the model with the Stilck Franca framework for MAX-CUT yields a circuit-size threshold that is provably tighter than prior work, without running the QPU on the target application.

## Techniques offered
- **Entropy density benchmarking** - a benchmarking methodology that monitors Renyi-2 entropy density accumulation in a circuit to bridge circuit-level and application-level protocols and decide whether quantum advantage is reachable.
  - guarantee: bridges circuit-level and application-level benchmarking and yields a conservative threshold on attainable quantum advantage (Renyi-2 substitution into the von Neumann analysis remains a valid lower bound)
  - quote: "We believe entropy density benchmarking provides an excellent bridge between circuit-level and application-level benchmarking protocols, enabling us to estimate the circuit size above which quantum advantage is unattainable for solving a given problem of interest."
- **Global depolarizing heuristic model of entropy accumulation** - an analytical model approximating local depolarising noise by a single global depolarising channel, linking circuit size (width, depth) and local gate-error rates to the purity and Renyi-2 entropy of the output.
  - guarantee: provides a heuristic lower bound to experimental entropy accumulation, making downstream circuit-volume predictions conservative and valid
  - quote: "Interestingly, a global depolarizing model provides a good heuristic description of the entropy accumulation behaviour at considered noise levels and for small systems without having to average over the whole family of circuits (variational gate parameters); it achieves a good fit already for a given specific circuit."
- **Circuit-size threshold for quantum advantage** - combines the entropy-accumulation model with the Stilck Franca application-level bound to certify a depth or gate-count threshold beyond which a noisy QPU cannot beat a classical solver, without running the device on the application.
  - guarantee: a certified classical-superiority bound on circuit depth and volume, proven analytically tighter than the prior Stilck Franca threshold for any c and p2
  - quote: "Using this simple model for entropy accumulation, we have shown how to impose more stringent bounds than previously shown in \cite{stilck_franca_limitations_2021} on the circuit size above which quantum advantage is lost for hardware-efficient VQA circuits run on a typical superconducting NISQ device."
- **Renyi-2 entropy (purity) estimation via classical shadows** - uses the classical shadows protocol with random single-qubit Pauli measurements to estimate the purity and second-order Renyi entropy of a circuit's output state on NISQ hardware.
  - guarantee: statistical sampling guarantee, an epsilon-accurate purity estimate with success probability at least 1 minus delta, at a cost scaling as O(ln(1/delta) 4^n / epsilon^2) Pauli measurements
  - quote: "By noting that the second-order Renyi entropy of a quantum state, $S^{(2)}(\rho)$, is a function of the purity $\Tr[\rho^2]$ ... one can use the classical shadows protocol to build an estimate for the second-order Renyi entropy"

## Where it could apply
- Target primitive(s): VQE, QAOA, Trotterised simulation, and hardware-efficient variational circuits more generally (the paper centres on a hardware-efficient VQA ansatz and MAX-CUT QAOA)
- Target application group(s): platforms_tooling (benchmarking and characterisation), optimization (MAX-CUT and combinatorial optimisation), chemistry_materials (the threshold technique is stated to translate to quantum-Hamiltonian problems)
- Code family / hardware assumptions: superconducting NISQ devices (demonstrated on Rigetti Aspen-M-3); no error-correcting code assumed; method is noted to need extension for entropy-decreasing errors such as photon loss on photonic platforms

## Caveats
The classical-shadows purity estimation is not scalable: the global SWAP observable forces an exponential O(4^n) Pauli sampling cost, so it is practical only for a handful of qubits, with the SWAP test left as the alternative at larger sizes. The global depolarising model is a heuristic lower bound that misses amplitude damping, coherent errors, and crosstalk, so a residual gap to experiment remains even after adding a T1-relaxation refinement. The certified threshold is conservative and generous to the QPU rather than tight.