ProductBlogAlgorithmsAboutSign inSign up

Research Replication

Three-Qubit Grover Search: Eight Phase Oracles

Examine a selected part of the three-qubit Grover experiment: eight phase oracles, one amplification step, saved count distributions and finite-shot comparisons.

Historical local simulator results · Public package pending

How closely do the output populations of eight three-qubit search circuits follow the ideal one-iteration prediction?

Source paper: Figgatt and colleagues, Complete 3-Qubit Grover Search on a Programmable Quantum Computer (2017).

Read the source paper →

What this replication investigates

This is an adapted logical subset of the paper: three data qubits, all eight single-solution phase oracles, and one Grover iteration. Each oracle marks a different three-bit state.

The reference omits the Boolean-oracle and ancilla variants, two-solution cases, ion-native compilation, and gate or process characterization. It is not a complete reproduction of the paper’s hardware study.

Method and parameters

  • Prepare a uniform superposition over eight candidates, mark one candidate with a phase oracle, then apply one diffusion step.
  • Repeat for all eight marked states, with 4,096 shots per circuit: 32,768 saved shots in total.
  • Calculate average marked-state success probability and squared statistical overlap with the ideal classical output populations.
  • Compare all 64 oracle/outcome probabilities using a simultaneous Hoeffding screen at α = 0.05, with a union bound.

Results and comparison

Scroll the table horizontally to see every results column.

Count-derived observations from eight historical local count maps
QuantityIdeal predictionSaved local observation
Mean marked-state success25/32 = 78.125%77.63671875%
Each unmarked population1/32 = 3.125%Compared separately across all eight oracles
Mean squared statistical overlap1 for matching populations0.9994776603
Marked state 111 success78.125%3,126/4,096 = 76.318359375%

One iteration gives marked amplitude 5/(2√8) and unmarked amplitude 1/(2√8), hence probabilities 25/32 and 1/32. This is the selected paper case, rather than a claim that one iteration always maximizes success.

All 64 ideal probabilities fall inside the declared conservative simultaneous intervals. This screen detected no discrepancy; it does not establish distribution equivalence or prove circuit correctness. The 111 marked-state residual is about −2.80 theoretical standard errors, so not every oracle is within one or two standard errors of theory.

Squared statistical overlap compares measured populations. It is not quantum state or process fidelity and does not recover phases or coherences.

What the evidence establishes

The evidence is eight preserved local count maps labelled local-quantumflow. They do not contain authenticated hosted job IDs or the historical executable version.

The current report reconstructs counts, theoretical probabilities and comparison metrics. Preparation hashes link the report inputs and generated outputs; they cannot establish which source version produced historical counts.

Limitations

  • No current hosted run, public runnable Capsule or independent-team reproduction is established by these saved results.
  • The paper’s 25% classical comparator has a specific one-query-and-guess resource budget; it is not an unrestricted classical bound.
  • A fixed eight-item demonstration does not establish a practical or asymptotic speedup, hardware performance or a device noise budget.
  • The sampling screen assumes independent shots within each fixed circuit and excludes systematic error, drift and correlated shots.

Access and reproduction

Read the original paper and the scoped comparison here. The experiment’s hosted execution and public reproduction package remain separate publication steps.

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’s learning circuit searches four candidates with two qubits and ideally returns 11 on every shot. This research reference uses eight candidates and eight separate three-qubit phase oracles; its ideal marked-state success is 78.125%. They are different experiments.

Try the two-qubit Grover learning example →

Explore the research collection →

References

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.