Research Replication
H₂ VQE: A Two-Qubit Energy Curve
Explore finite-shot energy measurement and classical optimization at thirteen molecular geometries, with an exact comparator and explicit accuracy limits.
Historical local simulator results · Public package pending
Can a one-parameter variational circuit track the ground-state energy of a reduced molecular-hydrogen model across a bond-length grid?
Source paper: O’Malley and colleagues, Scalable Quantum Simulation of Molecular Energies (2016).
What this replication investigates
This finite-shot local adaptation uses a two-qubit Hamiltonian at 13 selected molecular geometries from the paper’s rounded coefficient table. It demonstrates energy measurement and a hybrid quantum-classical optimization loop.
The comparator is exact diagonalization of the same rounded 4×4 Hamiltonian, not a new ab initio molecular calculation or a quantum phase-estimation run. The paper’s hardware and error-robustness comparisons are outside this reference’s scope.
Method and parameters
- Use the paper’s STO-6G model and two-qubit reduction, with a single variational parameter.
- Measure the Z, X and Y bases independently, using 4,096 shots per basis setting for each energy evaluation.
- Optimize with COBYLA, with a configured budget of 60 objective evaluations per geometry; the saved run used 20–29 plus a fresh final measurement.
- Reconstruct final energies from the counts and compare with independent eigenvalue checks for the same rounded Hamiltonian.
Results and comparison
Scroll the table horizontally to see every results column.
| Bond length (Å) | VQE estimate (Ha) | Exact model (Ha) | Signed error (mHa) |
|---|---|---|---|
| 0.20 | 0.151174 | 0.144210 | +6.963 |
| 0.75 | −1.143964 | −1.145599 | +1.635 |
| 2.85 | −0.943460 | −0.942973 | −0.487 |
Across all 13 points, the maximum absolute error is 0.007146 Ha and the mean absolute error is 0.003061 Ha. All points pass the repository’s 0.03 Ha engineering tolerance. This tolerance is not a confidence interval or a claim of chemical accuracy.
Only 4 of 13 final estimates are within the paper’s 0.0016 Ha chemical-accuracy benchmark; nine are outside it. The lowest sampled grid value is at 0.75 Å, which is not a fitted equilibrium geometry.
The saved run contains 978 measured circuits and 4,005,888 shots. The report separates optimizer objective calls from the three measurement circuits needed per energy evaluation, including the fresh final measurement.
What the evidence establishes
These results come from preserved local JSON. No current hosted Marqov job, public Capsule or separate-team rerun has been verified for this experiment.
The reference independently reconstructs final energies, preserves covariance among the Z-basis Hamiltonian terms, and checks circuit totals. Final standard errors describe measurement sampling only, excluding optimizer selection, device bias, coefficient rounding and run-to-run variation.
Limitations
- The saved dataset has no recorded source revision, simulator version, execution date or reproducible seed; current input hashes do not retroactively fill those gaps.
- Finite-shot energy estimates can lie below the exact eigenvalue. The variational bound applies to the true expectation, not every sampled estimate.
- Successful optimizer termination does not prove a global minimum. The saved history lacks trial parameters, so intermediate state preparation cannot be reconstructed.
- The reference does not reproduce the physical molecule, the paper’s full hardware experiment or its quantum phase-estimation comparison.
Access and reproduction
The paper and this model-level scientific overview are public. A verified hosted workflow and publicly reusable experiment package are still being prepared.
The experiment's source package and runnable Capsule are not yet public. This page describes the scoped reference and its evidence; it does not provide a one-click reproduction.
Learning example or research replication?
The editor example prepares a fixed state and has no Hamiltonian, energy evaluation or optimizer. This research reference measures a molecular model’s energy and optimizes a parameter across thirteen geometries. The learning snippet is not an H₂ VQE calculation.
Inspect the Fixed-Angle Ansatz learning example →
For a smaller local workflow with new teaching samples, follow the separate H₂ VQE: Measure Energy and Optimize a Parameter learning tutorial. It does not reproduce the retained results above.
Explore the research collection →
References
- O’Malley and colleagues: Scalable Quantum Simulation of Molecular Energies (2016)
- Published article: Physical Review X 6, 031007 (2016)
Scientific scope and retained results reviewed with the Marqov Replications project on October 16, 2026. Each reference is a separate experiment from the editor's learning examples.