Dwight Kahng — Independent Researcher
Paper: A 15-Qubit Decoherence-Reversal Protocol for Benchmarking Mid-Circuit Measurements and Environmental Imprinting (March 18, 2026)
arXiv: [link to be added]
Zenodo: [link to be added]

This repository contains the full implementation of a 15-qubit Loschmidt-echo protocol designed to benchmark mid-circuit measurement (MCM) performance on near-term quantum hardware. The protocol is intentionally restricted to a state-vector tractable regime (15 qubits), providing an exactly computable ideal baseline (F_max) against which noisy hardware recovery can be directly compared.

The protocol consists of three stages:

1. Forward scrambling — a 15-cycle unitary U applied to a 6-qubit system entangled with a 5-qubit artificial environment
2. Mid-circuit QND probe measurement — 4 probe qubits measure coarse-grained system observables as a proxy for environmental imprinting
3. Loschmidt-echo reversal — exact logical inverse U† is applied; recovery fidelity on the 11-qubit target register is measured

A surrogate noise model calibrated to published Quantinuum H2-1 specifications provides expected recovery thresholds. A strict 2σ decision rule classifies hardware results as Consistent, Anomalous, or Inconclusive.

The two scripts use different classical register layouts by design:

• Local simulation uses three named registers (cp, cs, ce) — readable, Qiskit-native, compatible with Aer
• Azure submission uses a single flat 15-bit register — required for deterministic bitstring concatenation in the Quantinuum JSON payload

Both scripts implement identical quantum physics. The register difference is purely a classical infrastructure adaptation. See Section 3.1 of the paper for details.

*The "Anomalous Decoherence" classification reflects a statistically significant gap between the idealized local surrogate model and the full-stack emulator, consistent with unmodeled compilation/routing overhead. It does not indicate unexpected decoherence on physical hardware. This is a successful demonstration of the protocol's diagnostic sensitivity to model mismatch.

Noise profiles calibrated to published Quantinuum H2-1 specifications:

pip install qiskit qiskit-aer

For Azure submission only:

pip install azure-quantum[qiskit] azure-identity

python local_simulation_v48_master.py

No Azure account required. Runs full F_max computation, Table 2 sweep on local Aer simulator, and a Phase VI diagnostic test reproducing the anomalous decoherence classification from the observed emulator result. Expect approximately 30 minutes on a standard Colab instance.

export AZURE_QUANTUM_RESOURCE_ID="<your-resource-id>"
export AZURE_QUANTUM_TENANT_ID="<your-tenant-id>"
python azure_quantinuum_submission_v48_public.py

You will be prompted to authenticate via browser (device code flow).

Standard backend.run() via the Azure QIR compiler was found to mishandle multi-shot mid-circuit measurements on the Quantinuum backend. This script bypasses the QIR compiler by:

1. Compiling the circuit locally to OpenQASM 2.0 via qasm2.dumps()
2. Submitting via the bare-metal Job.from_input_data() API with Honeywell format tags
3. Injecting input_params={"count": hardware_shots} directly into the raw payload

In this project, the bare-metal submission path was required to obtain correct multi-shot behavior on the Quantinuum backend.

Estimated cost: ~221 eHQC for a 400-shot run on quantinuum.sim.h2-1e (5 eHQC base fee + ~0.54 eHQC/shot).

If you use this protocol or code, please cite:

Kahng, D. (2026). A 15-Qubit Decoherence-Reversal Protocol for Benchmarking
Mid-Circuit Measurements and Environmental Imprinting.
arXiv: [to be added]

MIT License. See LICENSE file for details.

AI Use Disclosure: The associated manuscript incorporates structural and drafting assistance from large language models. The author remains fully responsible for all content, theoretical claims, and scientific accuracy.