A four-qubit universe: validated digital simulation of 1+1D lattice QED quenches on IBM Kingston

Baseline-validated digital quantum simulation of four-site Schwinger-model quench dynamics and background-field response on IBM Kingston Field Value Author Amit Brahmbhatt Affiliation Quantum Clarity LLC ORCID 0009-0003-5882-8476 Report date 2026-09-25 Version 1.2 Resource type Technical report / computational research record Abstract We report a validation-led digital quantum simulation of real-time dynamics in a four-site lattice Schwinger model, a (1+1)-dimensional formulation of quantum electrodynamics [1]. The study had three stages executed on the IBM Quantum processor ibm_kingston. First, an 18-PUB control experiment tested previously proposed “anomaly” observables against independently calculated noiseless baselines. Hardware agreed with the correct baselines to a maximum absolute deviation of 0.084, whereas the discarded analytic expectations differed by as much as 1.062. This negative result established that the earlier apparent anomalies were baseline errors rather than evidence of new physics. Second, we prepared the bare staggered vacuum on four qubits and evolved it under a Gauss-law-reduced Schwinger Hamiltonian with open boundaries, lattice spacing a=1, fermion mass m=0.5, and gauge coupling g=1. A nine-time-point quench reproduced the predicted non-monotonic staggered-magnetization trajectory. The maximum absolute hardware deviation was 0.262 from the noiseless first-order Trotter circuit and 0.240 from exact evolution; the largest deviations occurred in the deepest circuits. Third, we scanned a background-field parameter alpha over five values and six times. The hardware qualitatively reproduced both preregistered trends: increasing field suppressed the early-time rise of the pair-density proxy, while stronger fields sustained a larger late-time density. A repeated zero-field subset agreed across two jobs within 0.036 at every shared time. The three jobs comprised 57 circuits/PUBs, 116,736 requested shots, and 38 seconds of reported QPU usage. These results demonstrate an end-to-end workflow for small lattice-gauge simulations on noisy hardware: correct the classical baseline first, verify two independent Hamiltonian constructions, quantify Trotter error before submission, audit several days of backend-calibration history, freeze the selected physical qubit path into every hardware script, and interpret only features that survive those controls. A five-day, 12-hour-cadence audit selected ibm_kingston path [107,106,105,117]; the same mapping was then used for the control, zero-field quench, and field scan. The experiment does not measure the continuum Schwinger tunneling rate, probe electron substructure, or establish new particle physics. Keywords: Schwinger model; lattice gauge theory; quantum simulation; IBM Quantum; Qiskit; real-time dynamics; digital quantum simulation; background electric field; NISQ; validation Plain-language summary The Schwinger model is a simplified version of quantum electrodynamics in one space dimension and one time dimension. It is often used as a compact testbed for studying how matter and electric fields interact. Here, four qubits represented four points on a tiny spatial lattice. The qubits began in a simple “empty” reference pattern. Quantum gates then approximated how that state would change under the model. We measured a quantity called staggered magnetization. In this encoding, a value of -1 is the starting vacuum pattern; movement toward zero indicates more pair-like fermionic excitations on the lattice. This is an indirect model observable, not a detector counting real electrons and positrons. Three hardware jobs were used. The first was a sanity check. It showed that an earlier apparent anomaly came from incorrect expected formulas, not from the quantum processor. The second job then reproduced the predicted rise, partial return, and renewed rise of the excitation proxy after a sudden quench. The third changed a background-field parameter and found the same two qualitative effects predicted before hardware execution: stronger field slightly suppressed the early response but sustained much more excitation at later times. Before spending QPU time, we also examined five days of IBM calibration records at 12-hour intervals. Rather than choosing the least busy machine or trusting one favorable calibration, a script compared connected four-qubit paths using two-qubit-gate quality, readout quality, coherence relative to expected idle exposure, and stability over time. It selected the connected Kingston path [107,106,105,117]. We then locked the four model sites to those four physical qubits in all three hardware jobs. This made the hardware choice part of the experimental protocol rather than an after-the-fact explanation of the results. The important result is not a discovery of new physics. It is that a carefully validated, four-qubit lattice-field calculation retained its principal dynamical features on a real superconducting quantum processor despite gate noise. The work also illustrates why simulation baselines and negative controls must be established before interpreting unusual hardware output. 1. Research objective and scope The project asked whether a small, gate-based quantum processor could reproduce controlled real-time features of an interacting lattice gauge theory after a sudden quench, and whether those features changed predictably when a background electric-field parameter was varied. The final investigation addressed three questions: Baseline validity: Did the original electron-probe expectations survive exact, noiseless reconstruction? Zero-field quench: Could ibm_kingston reproduce the predicted non-monotonic evolution of a four-site Schwinger-model pair-density proxy? Background-field response: Could the processor preserve the predicted early- and late-time field-dependent trends without increasing the entangling-gate count? The scope is deliberately narrow. This is a finite, four-site digital simulation at dimensionless parameters, not a continuum extrapolation. It measures dynamics of an encoded model, not physical pair production in the IBM device. 2. Model and observable We used the staggered-fermion Hamiltonian for the Schwinger model with open boundaries. Gauss's law was used to eliminate explicit gauge-link degrees of freedom, leaving a spin Hamiltonian suitable for qubits. Hamiltonian: H(α) = [1/(2a)] · Σ(n = 0 to N − 2) [σₙ⁺σₙ₊₁⁻ + σₙ⁻σₙ₊₁⁺] + (m/2) · Σ(n = 0 to N − 1) (−1)ⁿσₙᶻ + (ag²/2) · Σ(n = 0 to N − 2) (Lₙ + α)² where Lₙ = Σ(k = 0 to n) [σₖᶻ + (−1)ᵏ]/2 The parameters used in every Schwinger run were: Lattice size: N=4 sites mapped to four qubits. Boundary condition: Open. Lattice spacing: a=1. Mass: m=0.5. Gauge coupling: g=1.0. Initial state: Bare staggered vacuum |1,0,1,0> in Qiskit qubit order. Background field: alpha=0 for the main quench; alpha={0,0.5,1.0,1.5,2.0} for the scan. The measured observable was staggered magnetization: M(t) = (1/N) · Σ(n = 0 to N − 1) (−1)ⁿ⟨σₙᶻ(t)⟩ For the prepared bare vacuum, M(0)=-1. Movement toward zero corresponds to increased occupation relative to that reference and is used here as a pair-density proxy. Because the initial state is not generally the interacting ground state, the observed dynamics combine quench energy, interactions, finite-size effects, and the background field. Consequently, the scan is interpreted as field reshaping of quench-induced pair-like dynamics, not as a direct measurement of the continuum Schwinger tunneling law. 3. Validation-first methodology 3.1 Reconstructing the original control baselines Before running the lattice-gauge model, the earlier electron-probe circuits were rebuilt in noiseless Qiskit Aer statevector simulation. This exposed three problems: An ad hoc formula predicted nonzero correlators where the simulated circuit predicted zero. A presumed random bit-match ratio of 0.5 was actually 0.0873 for the specified noiseless circuit. A tomography string-ordering error targeted qubits 3 and 2 rather than qubits 0 and 1 because Qiskit Pauli labels are little-endian; the error created a spurious -1 value in one YZ channel. A corrected hardware control then compared Kingston measurements with the reconstructed noiseless values. This step changed the interpretation from “possible anomaly” to a conventional null result before any Schwinger-model claim was made. 3.2 Independent Hamiltonian checks The zero-field Hamiltonian was constructed twice: as a dense matrix from tensor products of single-site operators; and independently as a SparsePauliOp. The two constructions agreed to machine precision (max|dH|=0). For the field-dependent model, the independent constructions agreed after removal of a physically irrelevant identity contribution, which changes only the global phase. The checks also caught and corrected: a Qiskit little-endian initial-state ordering error; a Q_j^2 expansion error in the electric term; label indices incorrectly treated as qubit indices in the field scan; and incorrect coefficients in a hand-simplified field-dependent expansion. 3.3 Exact and Trotter baselines Exact time evolution was obtained by diagonalizing the finite Hamiltonian. Digital circuits used first-order Trotterization. One-body Z terms were implemented as RZ rotations. Terms acting on the same qubit pair were exponentiated jointly and decomposed into two-qubit gates; every synthesized two-qubit unitary was checked to fidelity greater than 1-10^-9 against its target matrix. Combining the XX+YY interaction on each bond reduced the logical two-qubit cost at t=4 from 144 to 80 CX gates. A second-order formula was tested but was less favorable in the selected time-step regime. The chosen zero-field operating point used eight first-order steps at t=4, with a maximum noiseless magnetization error of 0.0552 relative to exact evolution. For the 30-point field grid, the maximum noiseless Trotter deviation was 0.0428. All final hardware-circuit statevectors matched the stored Trotter baselines within 5×10^-7 before submission. 3.4 Calibration-snapshot audit and path selection Physical-qubit selection was treated as a pre-experiment characterization step, not as an ad hoc choice made after seeing results. The ibm_backend_path_audit.py workflow queried five days of backend-calibration history at 12-hour intervals and enumerated connected four-qubit paths across the available backends. Each candidate was evaluated on: Estimated two-qubit contribution: Error accumulated over the path edges expected to carry the model's nearest-neighbor interactions. Readout contribution: Measurement-error exposure across the four selected qubits. Idle/coherence contribution: Estimated idle exposure relative to each qubit's T2. Temporal stability: Cross-snapshot coefficient of variation, excursion of the latest calibration from its recent history, and the fraction of unfavorable snapshots. Queue depth was reported separately because it affects waiting time, not circuit fidelity. This prevented a short queue from making a less stable path appear scientifically preferable. The audit itself submitted no quantum job. It emitted versioned CSV and JSON records plus preflight QASM, with SHA256 hashes to pin the audited artifacts. The 2026-09-24 audit selected ibm_kingston and the connected path [107,106,105,117] as the robust four-qubit candidate. The result was operationalized as follows: Logical lattice sites 0 through 3 were mapped in order to physical qubits 107, 106, 105, and 117. The model's three nearest-neighbor bonds were therefore placed along the audited connected path. The mapping was fixed in the hardware scripts rather than allowing a later run to choose a different physical quartet. The same path was reused for the corrected control, the nine-time-point zero-field quench, and the 30-circuit background-field scan. Exact and Trotter checks were completed before submission; IBM-connected dry runs then recorded transpiled depth and two-qubit count on the frozen layout. This discipline reduced two avoidable confounds: choosing a path from a single favorable calibration snapshot, and changing the physical qubits between related jobs. It does not eliminate drift during queueing or execution, and the per-snapshot calibration table should accompany the report in the Zenodo archive for independent reanalysis. 3.5 Hardware execution The model jobs used IBM's SamplerV2 interface [4]; the control used EstimatorV2 with resilience level 0. Every circuit/PUB used 2,048 requested shots. In the field scan, changing alpha modified only one-qubit Z-rotation coefficients. At each time, all five field settings retained the same two-qubit count after transpilation on the frozen path. Thus, the field-dependent comparison held the selected physical qubits and entangling-gate count constant while varying the Hamiltonian coefficient of interest. 4. Hardware execution record The three IBM jobs used 38 seconds of reported QPU time in total. Queue and wall-clock waiting are not included in that total. Job Workload QPU time daqv9pbt55cs738sb9a0 18 PUBs 12 s daqvesakqcdc73abc07g 9 circuits 7 s daqvm05b46qs73a1sfp0 30 circuits 19 s Total 57 units 38 s At 2,048 requested shots per circuit/PUB, the aggregate requested sampling was 116,736 shots: 36,864 in the control, 18,432 in the zero-field quench, and 61,440 in the field scan. Job 1: corrected hardware control IBM job ID: daqv9pbt55cs738sb9a0 Backend: ibm_kingston Workload: 18 Estimator PUBs × 2,048 shots Reported QPU time: 12 seconds Resilience level: 0 The corrected control produced: Maximum deviation from Aer baseline: 0.084. Maximum deviation from discarded formula: 1.062. Parity-control ZZZZ: +0.916, -0.011, and +0.943 at theta=0, pi, and 2pi, respectively. Zero-baseline checks: all nine deeper quick-probe observables were within ±0.062 of zero. Result: The hardware tracked the correctly simulated quantum-mechanical baselines within device-level error. The prior anomaly interpretation was rejected. This is a scientifically important negative result: it demonstrates that a mistaken reference model can produce an apparent hardware anomaly even when the processor behaves conventionally. Job 2: zero-field Schwinger quench IBM job ID: daqvesakqcdc73abc07g Backend: ibm_kingston Workload: 9 Sampler circuits × 2,048 shots Reported QPU time: 7 seconds Time grid: t={0,0.5,1.0,1.5,2.0,2.5,3.0,3.5,4.0} The transpiled circuits ranged from depth 2 with no two-qubit gates at t=0 to depth 423 with approximately 125 two-qubit gates at t=4. t Hardware M Trotter M Exact M 0.0 -0.979 -1.000 -1.000 0.5 -0.767 -0.824 -0.830 1.0 -0.462 -0.470 -0.495 1.5 -0.269 -0.294 -0.336 2.0 -0.374 -0.465 -0.499 2.5 -0.608 -0.791 -0.793 3.0 -0.680 -0.942 -0.920 3.5 -0.573 -0.789 -0.779 4.0 -0.354 -0.498 -0.532 The maximum absolute deviation was 0.262 from the noiseless Trotter result and 0.240 from exact evolution. Hardware preserved the main morphology: a rise from the vacuum value, a first turning point near t=1.5, a return toward more negative magnetization near t=3, and another rise by t=4. The largest residuals appeared in the deepest-circuit region from t=2.5 through t=3.5. Job 3: background-field scan IBM job ID: daqvm05b46qs73a1sfp0 Backend: ibm_kingston Workload: 30 Sampler circuits × 2,048 shots Reported QPU time: 19 seconds Field grid: alpha={0,0.5,1.0,1.5,2.0} Time grid: t={0,0.5,1.0,1.5,2.5,3.0} The maximum transpiled depth was 320. Two-qubit counts were 0,13,29,45,77,93 across the six times and were identical for every alpha, isolating the field scan from a gate-count confound. The job completed successfully after approximately 1,009 seconds of wall-clock queue and execution time. An initial local parser expected register c, while the returned DataBin used c0; results were recovered from the completed job without resubmission, and the parser was made register-name-agnostic. alpha = 0.0 t Hardware M Trotter M Exact M 0.0 -0.984 -1.000 -1.000 0.5 -0.756 -0.824 -0.830 1.0 -0.443 -0.470 -0.495 1.5 -0.258 -0.294 -0.336 2.5 -0.594 -0.791 -0.793 3.0 -0.644 -0.942 -0.920 alpha = 0.5 t Hardware M Trotter M Exact M 0.0 -0.984 -1.000 -1.000 0.5 -0.739 -0.824 -0.832 1.0 -0.426 -0.496 -0.525 1.5 -0.309 -0.380 -0.417 2.5 -0.576 -0.751 -0.774 3.0 -0.526 -0.723 -0.761 alpha = 1.0 t Hardware M Trotter M Exact M 0.0 -0.984 -1.000 -1.000 0.5 -0.752 -0.824 -0.836 1.0 -0.484 -0.531 -0.563 1.5 -0.386 -0.445 -0.468 2.5 -0.325 -0.349 -0.371 3.0 -0.205 -0.217 -0.260 alpha = 1.5 t Hardware M Trotter M Exact M 0.0 -0.982 -1.000 -1.000 0.5 -0.766 -0.824 -0.842 1.0 -0.530 -0.575 -0.608 1.5 -0.479 -0.499 -0.506 2.5 -0.127 -0.109 -0.113 3.0 -0.071 -0.012 -0.021 alpha = 2.0 t Hardware M Trotter M Exact M 0.0 -0.983 -1.000 -1.000 0.5 -0.758 -0.824 -0.849 1.0 -0.553 -0.624 -0.657 1.5 -0.489 -0.555 -0.552 2.5 -0.160 -0.138 -0.136 3.0 -0.109 -0.086 -0.081 Across all 30 points, the maximum absolute hardware-to-Trotter deviation was 0.298 at alpha=0, t=3.0. This was the same deep-circuit regime in which the zero-field job showed its largest residual. Preregistered trend checks Two field-dependent signatures were identified from the exact and Trotter baselines before hardware submission. Early-time suppression at t=1.0: As alpha increased from 0 to 2, hardware values were [-0.443,-0.426,-0.484,-0.530,-0.553]. With a 0.10 tolerance for small local violations, the sequence retained the predicted decreasing trend. The first two points are nearly tied and differ by only 0.017. Late-time sustained density at t=3.0: Hardware values were [-0.644,-0.526,-0.205,-0.071,-0.109]. The large overall movement toward zero with stronger field reproduced the predicted sustained-density trend. The final two points show a small reversal of 0.038, below the predefined 0.10 tolerance. The correct claim is therefore qualitative trend reproduction, not strict point-by-point monotonicity. Cross-job reproducibility The alpha=0 portion of Job 3 independently repeated six time points from Job 2. Absolute differences were 0.005,0.011,0.019,0.011,0.014,0.036 at t=0,0.5,1.0,1.5,2.5,3.0, respectively. Every repeated point agreed within 0.036 across the two submissions. This supports repeatability of the measured trajectory over the tested interval, while not replacing formal uncertainty estimation. 5. Findings and significance 5.1 Primary findings The control rejected the anomaly hypothesis. Kingston followed the corrected noiseless circuit baselines, while the older formulas were demonstrably wrong. This protects the later study from building a physical story on an invalid reference. The zero-field quench retained its dynamical shape. The processor reproduced the non-monotonic change in staggered magnetization predicted by exact and Trotter evolution, even though late-time amplitudes were compressed toward zero. Both field-dependent signatures survived hardware execution. Increasing alpha suppressed the early response and sustained substantially greater late-time density in the encoded observable. The zero-field trajectory was repeatable. Shared points from two independent jobs agreed within 0.036. The calibration audit became an experimental control. A five-day history, rather than one current snapshot, selected the connected Kingston path [107,106,105,117]; freezing that layout across all three jobs limited hardware-mapping variation between the control and model runs. Field variation did not add entangling gates. Because all field settings had equal two-qubit counts at fixed time on the same physical path, the observed field dependence cannot be attributed simply to giving high-alpha circuits more entangling operations. 5.2 Scientific significance The result is a compact demonstration that real-time lattice-gauge dynamics can remain recognizable on present-day noisy hardware when the experiment is designed around explicit validation. The value lies in the workflow as much as in the curves: proposed signal --> exact reconstruction --> independent Hamiltonian cross-check --> circuit verification --> multi-snapshot hardware-path audit --> frozen-layout dry run --> real-device run --> bounded interpretation The snapshot discipline is significant because backend properties are time-dependent. A path that looks attractive in one calibration can be atypical or unstable. Scoring both estimated error and recent variation, then freezing the chosen mapping before QPU submission, turned calibration history into a prespecified design input. Reusing the path did not make the three jobs noise-identical, but it removed arbitrary remapping as an obvious explanation for their agreement. This sequence prevented two categories of false claim. First, the control showed that a dramatic-looking discrepancy can come entirely from a bad expected formula or qubit-ordering error. Second, exact analysis showed that a bare-vacuum quench in the field-dependent Hamiltonian cannot be equated directly with the continuum vacuum-tunneling rate. Within the stated four-site model, the hardware data support a more modest but useful conclusion: the background field changed the timing and persistence of pair-like fermionic excitations in the same qualitative manner predicted by noiseless evolution. This adds a superconducting-qubit realization to the broader body of few-qubit and non-equilibrium Schwinger-model quantum simulations [2,3] and provides a reproducible case study in evidence-bounded NISQ experimentation. 5.3 What was not established This work does not establish any of the following: discovery of new particles, forces, or electron substructure; observation of physical electron-positron pairs inside the processor; measurement of the continuum Schwinger pair-production rate; a continuum, thermodynamic, or large-volume result; quantum advantage over classical computation; proof that hardware noise alone explains every residual; strict monotonicity of every field-scan point. The four-qubit model is classically tractable. Its purpose here is controlled validation of a digital quantum-simulation pipeline and preservation of qualitative dynamics on hardware. 6. Error sources and limitations Finite-size effects: Four sites are sufficient for a proof-of-workflow but not for continuum physics. Boundary effects and discrete spectra strongly shape the dynamics. State preparation: The bare staggered vacuum is easy to prepare but is not generally the interacting ground state, especially as the background field changes. The initial energy injected by the quench is part of the observed response. Trotterization: First-order product formulas introduce algorithmic error. The worst pre-hardware magnetization error was 0.0552 on the zero-field trajectory and 0.0428 on the field grid. Device noise: Deeper circuits showed larger deviations in the zero-field runs, consistent with accumulated gate and readout errors. No calibrated noise model was fitted, so noise consistency is not proof of a complete causal explanation. Sampling uncertainty: Each point used 2,048 shots. The report does not include per-point confidence intervals or covariance estimates. The cross-job comparison is empirical repeatability, not a substitute for error bars. No mitigation in the control: The baseline-control job explicitly used resilience level 0. The model-job scripts did not implement a separate mitigation study. Single backend and qubit path: All hardware evidence comes from one IBM backend and one four-qubit path. Independent replication on another device or path remains necessary for a broader robustness claim. Limited observables: Staggered magnetization is an aggregate proxy. Site-resolved occupation, charge separation, electric-field energy, Gauss-law violation, and vacuum persistence were not all measured on hardware. Coarse trend grid: Five field values and six times resolve the broad response but not fine dynamical structure. The trend criteria included a 0.10 tolerance and therefore do not imply exact monotonicity. 7. Reproducibility record Software and numerical checks The exact-baseline work used Qiskit 2.5.2 and qiskit-aer 0.17.2 with NumPy/SciPy statevector and dense-matrix calculations. The exact software package versions used by the remote Runtime client were not preserved in the hardware-result transcript and should be added to a future archived environment lockfile. The analysis pipeline contains the following logical components: ibm_backend_path_audit.py: five-day, 12-hour-cadence backend characterization; candidate-path enumeration; error-and-stability scoring; and versioned CSV/JSON/QASM output with SHA256 hashes. The audit recommended ibm_kingston path [107,106,105,117] without consuming QPU time. simulate_baselines.py and baselines_results.json: reconstruction of the original probe baselines. kingston_baseline_check.py: corrected control submitted as Job 1. schwinger_baseline.py and schwinger_exact_baselines.json: exact Hamiltonian, charge-sector checks, quench trajectories, and vacuum persistence. schwinger_trotter.py, schwinger_trotter2.py, and schwinger_trotter_opt.json: circuit construction and operating-point selection. schwinger_kingston.py: zero-field hardware run submitted as Job 2. schwinger_efield_baseline.py and schwinger_efield_exact.json: exact field-dependent trajectories. schwinger_efield_trotter.py and schwinger_efield_trotter.json: field-grid circuit validation. schwinger_efield_kingston.py: field-scan hardware submission. schwinger_efield_recover.py: recovery and analysis of completed Job 3 after a classical-register naming mismatch. Hardware-result files The hardware scripts generated timestamped JSON outputs on the execution host. The recorded filenames are: kingston_baseline_check_20260924_211343.json schwinger_kingston_20260924_212648.json schwinger_efield_kingston_20260925_091658.json For a complete Zenodo deposit, archive these JSON files together with the scripts, exact/Trotter baseline files, the path-audit CSV/JSON/preflight-QASM bundle and its SHA256 manifest, a dependency lockfile, and this report. IBM job IDs are included for provenance but may require the submitting account to retrieve full provider-side records. 8. Conclusion Across three IBM Quantum jobs, this project moved from attempted anomaly detection to a validated lattice-field simulation. The initial control demonstrated that the previously suspected anomalies were artifacts of incorrect formulas and qubit ordering. After that correction, independently checked exact and Trotter calculations defined a four-site Schwinger-model quench that could be tested on hardware. The zero-field run preserved the expected non-monotonic magnetization dynamics, with the largest quantitative errors occurring in the deepest circuits. The background-field scan then reproduced two predicted qualitative effects: suppression of the early-time response and persistence of pair-like density at late times. The repeated zero-field subset agreed across jobs within 0.036. The strongest defensible conclusion is: A four-qubit superconducting processor reproduced the principal qualitative features of baseline-validated Schwinger-model quench dynamics and their reshaping by a background-field parameter, within finite-size, sampling, Trotter, and hardware-noise limitations. The result is methodologically meaningful because it shows how rigorous baselines, explicit failure correction, and conservative interpretation can turn a small NISQ experiment into a reproducible scientific record without overstating what the device measured. 9. Suggested next experiments Add uncertainty estimates: Bootstrap shot counts or repeat jobs to provide confidence intervals for each M(t,alpha) point. Measure local structure: Record site-resolved occupations, charge separation, and electric-field observables to test the physical interpretation of the late-time density. Check gauge consistency: Quantify Gauss-law violations after transpilation and hardware execution. Replicate across hardware: Repeat the same frozen circuits on another backend and on a second qubit path. Test mitigation: Compare raw results with readout mitigation and zero-noise extrapolation under a prespecified analysis plan. Increase system size: Move from four sites to six or more while tracking depth, finite-size effects, and classical-verification cost. Prepare an interacting vacuum: Compare the bare-vacuum quench with an adiabatically or variationally prepared approximation to the interacting ground state. Archive calibrations: Deposit the backend calibration snapshots and path-audit artifact used to select [107,106,105,117]. 10. Suggested Zenodo metadata Title: Baseline-validated digital quantum simulation of four-site Schwinger-model quench dynamics and background-field response on IBM Kingston Creator: Amit Brahmbhatt — Quantum Clarity LLC — ORCID 0009-0003-5882-8476 Publication date: 2026-09-25 Version: 1.2 Resource type: Technical report Suggested description: This deposit documents three IBM Quantum jobs comprising a corrected hardware baseline control, a four-site zero-field Schwinger-model quench, and a five-value background-field scan. It includes a five-day calibration-snapshot audit that selected and froze the physical path [107,106,105,117], exact and Trotter validation, circuit and hardware parameters, full reported observables, job identifiers, limitations, and a plain-language summary. The principal result is qualitative reproduction of baseline-validated quench dynamics and two field-dependent trends on ibm_kingston, without claims of continuum-rate measurement or new physics. Keywords: Schwinger model; lattice quantum electrodynamics; digital quantum simulation; IBM Quantum; Qiskit; NISQ; quench dynamics; background field; staggered fermions; validation License: MIT License 11. License MIT License Copyright (c) 2026 Amit Brahmbhatt / Quantum Clarity LLC Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. References Schwinger, J. “On Gauge Invariance and Vacuum Polarization.” Physical Review 82, 664–679 (1951). DOI: 10.1103/PhysRev.82.664. https://inspirehep.net/literature/113 Martinez, E. A., Muschik, C. A., Schindler, P., et al. “Real-time dynamics of lattice gauge theories with a few-qubit quantum computer.” Nature 534, 516–519 (2016). DOI: 10.1038/nature18318. https://www.nature.com/articles/nature18318?error=cookies_not_supported&code=5664f358-cb3d-4dd5-a98c-0d39d32be96b de Jong, W. A., et al. “Quantum simulation of non-equilibrium dynamics and thermalization in the Schwinger model.” arXiv:2106.08394 (2021). https://arxiv.org/abs/2106.08394v1 IBM Quantum. “SamplerV2.” IBM Quantum Documentation. https://quantum.cloud.ibm.com/docs/en/api/qiskit-ibm-runtime/sampler-v2 Data integrity statement All hardware numbers in this report are transcribed from the recorded outputs of the three named IBM jobs. The audit description and selected path are drawn from the recorded 2026-09-24 ibm_backend_path_audit.py workflow; exact per-snapshot values are not reproduced here and should be preserved in the accompanying audit bundle. Exact and Trotter values come from the saved classical baseline artifacts. No unrecorded hardware values have been inferred. Values in result tables are rounded to three decimal places, matching the console output. The reported 38-second total is the sum of the IBM portal's per-job QPU usage values supplied for this study; it excludes queue time and classical preprocessing.