Shranjeno v:
| Glavni avtor: | |
|---|---|
| Format: | Recurso digital |
| Jezik: | angleščina |
| Izdano: |
Zenodo
2026
|
| Teme: | |
| Online dostop: | https://doi.org/10.5281/zenodo.19478241 |
| Oznake: |
Označite
Brez oznak, prvi označite!
|
Kazalo:
- <h1>Y⊗Z Stabilizer Validation on IBM Kingston: Backend-Adapted Continuation at 116-Qubit Scale (29 Modules, 87.34% Average Fidelity)</h1> <p><strong>Author:</strong> Amit Brahmbhatt<br><strong>Organisation:</strong> Quantum-Clarity LLC<br><strong>Date:</strong> April 8, 2026<br><strong>IBM Job ID:</strong> d7bf7t65nvhs73a39fvg (publicly verifiable)<br><strong>Backend:</strong> ibm_kingston (IBM Heron r3, 156 qubits, heavy-hex topology)<br><strong>Predecessor record:</strong> DOI 10.5281/zenodo.18498540 (February 5, 2026)</p> <h2>Plain Language Summary</h2> <p><strong>What we found:</strong></p> <p>We ran the same Y⊗Z stabilizer experiment that first appeared in our February 2026 Zenodo record, this time on IBM's Kingston processor — a different, newer IBM backend — and it worked. Across 29 independent four-qubit modules spanning 116 qubits, the experiment returned an average fidelity of 87.34%, peak Y⊗Z correlation of 98.49%, and Z-orthogonality success of 96.15%, clearing the 85% validation threshold. The full run completed in 107.9 seconds of QPU time, faster than projected, because we rewrote the circuit in Kingston's native gate vocabulary (CZ) rather than the original CX form. The underlying logical method was preserved exactly. Only the physical gate representation changed.</p> <p><strong>Why this matters:</strong></p> <p>Our February records showed two things: the Y⊗Z parity-triangle method is a sensitive probe of real hardware error structure, and that error structure on IBM superconducting processors is not the simple independent noise that many quantum error correction models assume — it is topology-dependent, temporally correlated, and environmentally persistent. The Kingston result matters because it shows the same logical method can be transplanted onto a different current-generation backend without modification to its mathematical action, and still return strong results at scale. This is evidence that the QuantaCore approach is not tuned to one specific chip or topology. It is a method that works with the physical structure of quantum hardware generally, and that can be adapted to new backends through standard gate decomposition while preserving its scientific and legal identity. This dated, DOI-indexed continuation strengthens the public record of the method on a current-generation IBM backend.</p> <h2>Abstract</h2> <p>We present the results of a 116-qubit Y⊗Z stabilizer validation experiment executed on IBM Kingston (ibm_kingston, Heron r3, 156-qubit heavy-hex processor) on April 8, 2026. The experiment deployed 29 non-overlapping four-qubit modules using dynamic coupling-map-based layout discovery, with no hardcoded qubit assignments. The circuit was adapted to Kingston's native CZ gate set using a standard unitary-preserving decomposition, without changing the underlying logical method. Fidelity scoring used the same three-observable estimator protocol (Primary YZ, Secondary YZ, Z-orthogonality) established in the predecessor record (DOI: 10.5281/zenodo.18498540).</p> <p>Results across 29 modules: average fidelity 87.34% ± 5.52%, range 63.62%–93.23%, peak Y⊗Z correlation 98.49%, average Y⊗Z correlation 90.71%, Z-orthogonality success 96.15%. Validation threshold (≥85%) was met. Observed QPU execution time was 107.9 seconds, lower than the initial projection of 145 seconds, consistent with efficient native-gate adaptation on the Kingston backend.</p> <p>One module (Module 18, layout [100, 102, 101, 116]) produced an anomalous result (fidelity 63.62%), with primary Y⊗Z channel collapse to 0.417 while the secondary channel remained normal at 0.678 — a dissociation pattern consistent with localised decoherence or miscalibration on a specific qubit or edge rather than global circuit failure. This pattern is consistent with the topology-dependent, spatially localised error structure characterised in the February 6, 2026 predecessor campaign.</p> <p>The secondary-to-primary Y⊗Z suppression ratio (B/A) measured 0.731 on Kingston, compared to 0.525 on ibm_marrakesh in earlier work — a 39% reduction in B-channel suppression that may reflect the CZ-native circuit rewrite, hardware differences, or both. This cross-backend variation in the suppression ratio is a finding warranting controlled follow-up.</p> <p>This record establishes the third stage of a continuous QuantaCore research program: method disclosure (February 5, 2026), non-Markovian physical diagnosis (February 6, 2026), and backend-adapted operational continuation at scale (April 8, 2026, this record).</p> <h2>Keywords</h2> <p>Y⊗Z stabilizers, quantum error characterisation, IBM Kingston, Heron r3, CZ-native circuits, non-Markovian noise, topology-dependent errors, NISQ hardware, basis migration, quantum reliability, prior art, QuantaCore, Q-HAL, superconducting qubits, parity-triangle test, hardware-adaptive quantum circuits</p> <h2>1. Experimental Background and Continuity</h2> <p>This record is the third in a sequence of connected QuantaCore experimental disclosures. The first record (DOI: 10.5281/zenodo.18498540, February 5, 2026) established the Y⊗Z parity-triangle consistency test as a method for probing topology-dependent error structure on superconducting hardware, demonstrating 4.86σ and 3.76σ deviations from independence assumptions on IBM Heron r2 (ibm_fez). The second record (February 6, 2026) extended that finding into a full non-Markovian reliability campaign, reporting spatial correlations at 4.86σ, temporal memory effects at 2.8σ, and environmental persistence at 3.6σ, with a characteristic timescale of approximately 30 microseconds surviving full qubit reset and repreparation.</p> <p>The present record demonstrates that the same logical method, adapted to CZ-native gate constraints through a standard unitary-preserving decomposition, executes successfully across 29 modules and 116 qubits on IBM Kingston — a different current-generation IBM backend. The method is unchanged; only its physical representation differs.</p> <h2>2. Backend Characterisation</h2> <p><strong>Backend:</strong> ibm_kingston<br><strong>Processor family:</strong> IBM Heron r3<br><strong>Qubit count:</strong> 156<br><strong>Topology:</strong> Heavy-hex lattice (confirmed by coupling map analysis)<br><strong>Native 2Q gate:</strong> CZ (not ECR as on ibm_marrakesh/ibm_fez)<br><strong>Native gate set:</strong> {rz, sx, x, id, cz, measure, reset, delay, if_else}<br><strong>Average T1:</strong> 280.3 µs<br><strong>Average T2:</strong> 175.0 µs<br><strong>Average readout error:</strong> 2.33%</p> <p>Kingston is Heron r3 heavy-hex, not the Nighthawk square lattice. Full per-qubit and per-pair hardware characterisation data is included in the accompanying hardware JSON files.</p> <h2>3. Circuit Implementation</h2> <h3>3.1 Circuit adaptation</h3> <p>The circuit was adapted to Kingston's CZ-native gate set using a standard unitary-preserving decomposition. The implementation preserves the same logical method disclosed in the earlier QuantaCore record (DOI: 10.5281/zenodo.18498540). Full circuit details are subject to U.S. patent protection and are not disclosed here.</p> <h3>3.2 Layout discovery</h3> <p>Rather than hardcoded qubit arrays (as used in prior Fez experiments), Kingston layouts were discovered dynamically from the live coupling map at runtime. Layouts were selected using a hardware-aware scoring procedure designed to reduce routing overhead and favour high-quality local connectivity, with no qubit assignments fixed in advance.</p> <h3>3.3 Observables</h3> <p>Three Pauli observables per module, consistent with predecessor record:</p> <table> <thead> <tr> <th>Observable</th> <th>Label</th> <th>Role</th> </tr> </thead> <tbody> <tr> <td>YZII</td> <td>Primary YZ</td> <td>Direct Y⊗Z correlation on qubits 0,1</td> </tr> <tr> <td>IIYZ</td> <td>Secondary YZ</td> <td>Y⊗Z correlation on qubits 2,3</td> </tr> <tr> <td>ZIII</td> <td>Z-orthogonality</td> <td>Isolation of Z-basis from Y⊗Z subspace</td> </tr> </tbody> </table> <p>Total PUBs submitted: 87 (29 modules × 3 observables).</p> <h2>4. Results</h2> <h3>4.1 Overall statistics</h3> <table> <thead> <tr> <th>Metric</th> <th>Value</th> </tr> </thead> <tbody> <tr> <td>Modules</td> <td>29</td> </tr> <tr> <td>Total qubits</td> <td>116</td> </tr> <tr> <td>Average fidelity</td> <td>87.34% ± 5.52%</td> </tr> <tr> <td>Fidelity range</td> <td>63.62% – 93.23%</td> </tr> <tr> <td>Average Y⊗Z correlation</td> <td>90.71%</td> </tr> <tr> <td>Peak Y⊗Z correlation</td> <td>98.49% (Module 28, Q7-8-9-10)</td> </tr> <tr> <td>Z-orthogonality success</td> <td>96.15%</td> </tr> <tr> <td>Validation status</td> <td>SUCCESS (≥85% threshold)</td> </tr> <tr> <td>QPU time</td> <td>107.9 seconds</td> </tr> </tbody> </table> <h3>4.2 Fidelity distribution</h3> <table> <thead> <tr> <th>Range</th> <th>Modules</th> <th>%</th> </tr> </thead> <tbody> <tr> <td>60–65%</td> <td>1</td> <td>3.4%</td> </tr> <tr> <td>65–70%</td> <td>0</td> <td>0.0%</td> </tr> <tr> <td>70–75%</td> <td>0</td> <td>0.0%</td> </tr> <tr> <td>75–80%</td> <td>1</td> <td>3.4%</td> </tr> <tr> <td>80–85%</td> <td>4</td> <td>13.8%</td> </tr> <tr> <td>85–90%</td> <td>14</td> <td>48.3%</td> </tr> <tr> <td>90–95%</td> <td>9</td> <td>31.0%</td> </tr> </tbody> </table> <p>27 of 29 modules (93.1%) returned fidelity ≥ 80%. 23 of 29 modules (79.3%) cleared the 85% threshold individually.</p> <h3>4.3 Module 18 anomaly</h3> <p>Module 18 (layout [100, 102, 101, 116], P2 = qubit 101) produced a fidelity of 63.62%, a 23.7 percentage-point deviation below the experiment mean. The dissociation pattern is diagnostically significant: primary Y⊗Z collapsed to 0.417 (the lowest value on the chip) while secondary Y⊗Z remained at 0.678 (within normal range), and Z-orthogonality leaked to 0.275. This channel-specific collapse, with the secondary channel intact, is inconsistent with a global circuit failure and instead points to a localised decoherence or miscalibration event on a specific qubit or edge — most likely the 101–116 coupling. This pattern is consistent with the spatially localised, topology-dependent error hotspots characterised in the February 6 predecessor campaign.</p> <h3>4.4 B/A suppression ratio</h3> <p>The ratio of secondary to primary Y⊗Z correlation magnitude (B/A) was 0.731 on Kingston, compared to 0.525 on ibm_marrakesh in the Marrakesh campaign. This 39% reduction in B-channel suppression may reflect the CZ-native circuit rewrite altering secondary basis preparation, hardware calibration differences between backends, or both. Isolating these contributions requires a matched control experiment with identical circuits on both backends.</p> <h2>5. IP and Prior Art Statement</h2> <p>The Y⊗Z stabilizer method implemented in this experiment is protected under U.S. patent application (USPTO Serial No. 63/952,786). This dataset constitutes experimental validation of that method on IBM Kingston hardware, extending the public record established in DOI 10.5281/zenodo.18498540 to a second current-generation IBM backend at larger scale.</p> <p>This record is the third stage of a continuous QuantaCore research program. The predecessor records predate recent industry disclosures on quantum cryptographic vulnerability timelines, and this continuation further strengthens the dated, DOI-indexed public record of the method.</p> <h2>6. Reproducibility and Verification</h2> <p><strong>IBM Quantum job ID:</strong> d7bf7t65nvhs73a39fvg<br>All results are independently verifiable by any IBM Quantum account holder with access to ibm_kingston job history.</p> <p><strong>Software environment:</strong></p> <pre><code>qiskit >= 1.0 qiskit-ibm-runtime >= 0.20 numpy </code></pre> <p><strong>Replication:</strong> The experiment can be independently assessed through the public IBM Quantum job record. Implementation-specific scripts and selection logic are proprietary and are not included in this dataset.</p> <p><strong>Files in this record:</strong></p> <ul> <li><code>yz_kingston_29mod_116q_20260408_173541.json</code> — complete per-module results, layouts, and statistics</li> <li><code>ibm_kingston_hardware_20260408_170917.json</code> — IBM Kingston hardware characterisation: per-qubit and per-pair properties extracted at time of experiment</li> <li><code>ibm_kingston_optimization_report_20260408_170917.json</code> — derived topology and connectivity analysis</li> <li><code>ibm_kingston_coupling_map_20260408_170917.png</code> — visualisation of the Kingston coupling map, nodes colour-coded by T1 coherence time</li> </ul> <h2>7. Acknowledgements</h2> <p>IBM Quantum for hardware access and QPU credits on ibm_kingston. The IBM Quantum team for continued platform support during the post-Classic platform transition.</p> <h2>Citation</h2> <p>Brahmbhatt, A. (2026). <em>Y⊗Z Stabilizer Validation on IBM Kingston: Backend-Adapted Continuation at 116-Qubit Scale (29 Modules, 87.34% Average Fidelity)</em>. Zenodo. https://doi.org/10.5281/zenodo.[PENDING]</p> <p><strong>Predecessor record:</strong><br>Brahmbhatt, A. (2026). <em>Experimental Evidence of Topology-Dependent Error Correlations in Y⊗Z Stabilizer Measurements on IBM's 156-Qubit Heron Processor (4.86σ)</em>. Zenodo. https://doi.org/10.5281/zenodo.18498540</p> <p><em>© 2026 Amit Brahmbhatt, Quantum-Clarity LLC. Data: CC BY 4.0. Code: MIT.</em></p>