WorksPhysics and Astronomy
Nuclear scattering phase shifts with factorized geometric-time RODEO on a quantum processor
Abstract
We reconstruct elastic -wave neutron-proton () phase shifts from trapped spectra measured on IBM Aachen using a compressed RODEO circuit. For a schematic square-well interaction represented by classically constructed four-dimensional effective Hamiltonians, four positive-energy levels at five trap strengths supply twenty inputs to classical modified effective range expansion (MERE) extrapolation to free space. The six-cycle factorized geometric-time RODEO implementation (FG-R6) combines ancilla reuse, exact query-phase separation, post-compilation binding, and six numerically optimized geometric evolution times. Numerical cycle-count tests support this choice within the adopted local spectral tolerances. At the same three-qubit width and equal shot budgets, FG-R6 reduces median compiled depth and two-qubit-gate count by 33.1% and 37.2% relative to ten-cycle dynamic direct RODEO (direct R10), increases the median fitted amplitude, and lowers median finite-shot energy uncertainty from 0.610 to 0.534 keV. For this dataset's primary fit on the 0.1-30.0 MeV grid, the maximum central phase-shift deviation from exact-energy MERE decreases from to , despite a larger root-mean-square (RMS) trapped-energy deviation. The full-grid gain arises mainly in the low-energy extrapolation region; direct R10 has slightly smaller residuals relative to the same reference on the 10.0-30.0 MeV common energy-interpolation subset. The FG-R6 central curve differs from the analytical square-well solution by at most . Fit-form sensitivity remains appreciable and is assessed separately from finite-shot uncertainty. This reduced-space benchmark connects RODEO circuit compression to nuclear continuum observables and shows why circuit performance must be assessed through the scattering reconstruction and its energy range, not spectral RMS errors alone.
PLB_FULL_RODEO.pdf
Introduction
Elastic-scattering phase shifts probe nuclear interactions in the continuum. A unified description of bound states and scattering requires nuclear correlations and appropriate asymptotic boundary conditions, a continuing challenge despite advances in ab initio reaction theory [1, 2, 3]. Weak harmonic confinement allows bound-state methods to address scattering by replacing the continuum with discrete, normalizable states [4, 5]. For short-range interactions in an uncoupled elastic channel, the modified effective range expansion (MERE) relates these trap-dependent levels to free-space scattering, including finite-confinement corrections. For waves, extrapolation to vanishing confinement yields the effective-range function . The eigenenergies of the confined system can be estimated using quantum eigensolvers, while the subsequent fitting in energy and extrapolation in confinement strength are performed classically [6].
The physical significance of the trapped-spectrum approach lies not in reproducing isolated eigenvalues, but in connecting quantum-computed discrete spectra to continuum observables [4, 5, 6]. Scattering phase shifts encode the interaction over a continuous energy interval and determine experimentally accessible quantities such as cross sections, resonance positions, and widths. Their reconstruction combines information from several trapped levels and confinement strengths, so spectral errors can be amplified, suppressed, or reorganized by the subsequent energy interpolation and zero-confinement extrapolation. Consequently, the performance of a quantum eigensolver cannot, in general, be assessed from level-by-level energy errors alone. Carrying the measured spectrum through the complete trapped-spectrum analysis therefore provides a more physically relevant benchmark of whether a resource-reduced quantum circuit preserves the information required for nuclear scattering.
The RODEO algorithm filters energy components through repeated ancilla-controlled time evolution and postselection [7, 8]. Wang et al. used it to extract phase shifts for model neutron–proton () and neutron–alpha () scattering from trapped spectra in noiseless simulations [6]. Complementary hardware studies employed variational eigensolvers and quantum subspace expansion in finite oscillator spaces [9], or real-time evolution [10]. The trapped-spectrum approach requires controlling both the cost of estimating several positive-energy levels and the propagation of their errors to the free-space phase shift.
Email addresses: [email protected] (Jubin Park), [email protected] (Myung-Ki Cheoun)
Our recent benchmark of a single trapped level [11] implemented the same ideal ten-cycle RODEO filter with twelve active qubits in a static circuit on IonQ Forte-1 and three in an ancilla-recycled dynamic circuit on IBM Aachen, reducing circuit width by 75.0%. Mid-circuit measurement and reset recycle one ancilla [12], making the auxiliary-qubit requirement independent of the cycle count and leaving more qubits for the nuclear system, but retaining the same controlled-evolution sequence. The interleaved IBM scan reproduced the trapped energy and its finite-confinement MERE input within finite-shot uncertainties, although independent runs showed additional variation. That calculation supplied one finite-confinement input; a free-space extraction requires several levels and trap strengths.
Here we retain the three-qubit register but shorten the evolution sequence using the six-cycle factorized geometric-time RODEO implementation (FG–R6), as shown in Figure 1. It combines the established separation of Hamiltonian evolution and the query-energy phase [6, 7] with geometric time sampling [13], using six evolution times selected for the present scan window. A reference-energy circuit is compiled once per target, with the query offset subsequently bound to an ancilla phase. Retrospective cycle-count and simulator tests support the six-cycle choice within the adopted local spectral tolerances, without establishing a minimum cycle count.
Figure 1. Representative cycles of the ancilla-recycled RODEO circuits. (a) Dynamic direct R10 uses ten cycles and compiles the query-dependent controlled evolution separately for each query energy. (b) FG–R6 uses six geometric evolution times and a reference-energy circuit with the query-dependent ancilla phase bound after compilation. Both use two system qubits and one recycled ancilla. M/R denotes measurement after each cycle and reset before the next; no reset follows the final cycle. System qubits are measured only at the end, and classical ancilla records are omitted for clarity. The displayed cycle is included in the total of cycles; trial-state preparation is performed only once. Success requires all recorded ancilla outcomes to be zero, irrespective of the terminal system bits .
For each implementation, we estimate four positive-energy levels at five trap strengths from raw IBM measurements, providing twenty spectral inputs to the classical MERE analysis. The four-dimensional Okubo–Lee–Suzuki (OLS) effective Hamiltonians are constructed classically from the reference eigensystems of the schematic -wave interaction of Ref. [6]. Exact diagonalization, the analytical scattering solution, and noiseless and calibration-derived noisy simulations provide references. Finite-shot energy uncertainties are propagated to the phase shift separately from sensitivity to the extrapolation form.
Relative to dynamic R10, FG–R6 reduces median compiled depth and two-qubit-gate count by approximately one third. For the present dataset and primary MERE fit, its maximum central phase-shift deviation from exact-energy MERE is smaller on the 0.1–30.0 MeV grid despite a larger root-mean-square (RMS) trapped-energy deviation. This gain arises mainly in the low-energy extrapolation region; direct R10 has slightly smaller maximum and RMS phase residuals on the 10.0–30.0 MeV common energy-interpolation subset. The comparison links circuit compression to a nuclear continuum observable and shows why spectral errors must be assessed through the reconstruction and its energy range, rather than by their RMS alone.
Trapped-spectrum formulation
We consider a schematic -wave square well of Refs. [6, ?]. In the center-of-mass frame and in natural units (), the relative Hamiltonian is
where is the relative kinetic energy and MeV is the neutron–proton reduced mass. The interaction is a square-well potential, for and zero otherwise, with MeV and fm. The square well provides an analytical scattering reference. Energies are measured from the free threshold, and the trap strengths are MeV.
The reference spectrum is obtained by classical diagonalization in a radial oscillator basis with , corresponding to 301 basis states in the -wave channel . The basis frequency MeV is distinct from the trap strength . At each , an OLS transformation [14, 15] yields a effective Hamiltonian preserving the four lowest positive eigenvalues of the reference matrix to numerical precision. Each reduced Hamiltonian is encoded on two system qubits. The twenty reference energies range from 8.855 to 35.340 MeV (see Table S1 in the Supplemental Material).
| (MeV) | ||||
| 3.6 | 8.855470 | 16.643000 | 24.213795 | 31.691304 |
| 3.7 | 9.129551 | 17.134039 | 24.916082 | 32.602105 |
| 3.8 | 9.404443 | 17.625943 | 25.619273 | 33.513827 |
| 3.9 | 9.680129 | 18.118693 | 26.323347 | 34.426445 |
| 4.0 | 9.956592 | 18.612272 | 27.028285 | 35.339939 |
Table S1. Positive reference energies (MeV), retained by the four-dimensional OLS Hamiltonians. Displayed values are rounded.
At each trapped eigenenergy , we evaluate [4, 5, 6]
With , the notation denotes a finite-confinement quantity, not the free-space effective-range function, since the trap also perturbs the interaction region. Numerical values of both and the momenta in fm are obtained by dividing their natural-unit values by MeV fm.
Following Ref. [6], we fit the four points at each by
Evaluating these five polynomials at a common scattering energy , we extrapolate their leading dependence to zero confinement:
Here in natural units and estimates , with and expressed in the same units. The principal inverse tangent selects the continuous branch used for both reconstructed and analytical phases. Both stages use unweighted least squares and are repeated for the energy-bootstrap samples; fit-form sensitivity is assessed separately.
Energy interpolation and trap removal are distinct. Relative to the reference inputs, all five energy fits extrapolate below 8.855 MeV, while interpolation and extrapolation coexist up to 9.957 MeV. Above this boundary, all five interpolate within the analysis range; trap removal remains an extrapolation throughout. On a uniform 300-point energy grid spanning 0.1–30.0 MeV with steps of 0.1 MeV, the common interpolation subset therefore spans 10.0–30.0 MeV and contains 201 grid points. Interval comparisons restrict the same reconstructed curves without refitting.
Applying the primary prescription to the exact reference energies defines . Its maximum absolute difference from the analytical square-well phase on the analysis grid is 0.1411° below 8.855 MeV and 0.0502° above it, decreasing to 0.0372° on the common-interpolation subset. The full-grid maximum relative difference is 0.1128%. These discrepancies characterize the fixed reference spectrum and primary reconstruction, not quantum processing unit (QPU) uncertainties or a complete extrapolation-error budget.
Factorized geometric-time RODEO
For and an initial state , the ideal probability of an all-zero ancilla record is [6, 7]
where . Both implementations use two system qubits and one recycled ancilla [11]. The ancilla is measured after each cycle and reset only between cycles; all-zero records define success. The direct R10 reference uses ten fixed Gaussian-quantile times with scale and separately compiles the controlled for each query energy.
The FG–R6 circuit combines query-phase separation with a shorter sequence (Figure 1). Using the established RODEO phase separation [6, 7],
where , the scalar factor becomes the ancilla phase . The complete parameterized circuit is compiled once per target at fixed , and is bound afterward for each query. The identity is exact: at identical times, the numerical check gives an absolute matrix-element difference no larger than . This check is distinct from the approximation involved in changing the time sequence.
Following geometric time sampling [13], we use
with and . At fixed , these parameters were optimized against the isolated R10 response on 1,001 points over keV. The maximum response difference is , and noiseless scans including all four levels give a maximum fitted-center offset of 0.0853 keV in the keV query window. The objective, bounds, and optimization settings and the full time sequences are provided in § 6.1.1 and Table S4 of the Supplemental Material, respectively. As will be discussed in § 1.4, retrospective tests of the unchanged hardware sequence support this scan-specific choice but do not establish a minimum cycle count.
| Cycle | Direct R10 | FG–R6 |
| 1 | -34.54192616598 | 36.50539373580 |
| 2 | -21.76510117937 | 30.44067631215 |
| 3 | -14.16428475412 | 25.38350308032 |
| 4 | -8.09172979456 | 21.16648861613 |
| 5 | -2.63888828396 | 17.65005558608 |
| 6 | 2.63888828396 | 14.71781493101 |
| 7 | 8.09172979456 | – |
| 8 | 14.16428475412 | – |
| 9 | 21.76510117937 | – |
| 10 | 34.54192616598 | – |
Table S4. Fixed production times (MeV), rounded for display. Equations (S10) and (S11) define the sequences; no fresh random-time averaging is used.
Each scan is centered on the classically known target energy . We select the computational-basis state whose diagonal Hamiltonian element is nearest , giving squared target overlaps of 0.855–0.922. The seven offsets 0.0, , , keV, with the central query repeated to check success-probability consistency, give circuits at 500 shots each. Both Direct R10 and FG–R6 use on ibm_aachen, without gate or measurement twirling, runtime dynamical decoupling, or count debiasing.
The production controls comprise the fixed-time formula, ideal Aer at 2,000 shots per circuit, and calibration-derived noisy Aer at 500 shots [16]; the latter only approximates processor noise. The additional cycle-count study, detailed in Secs. S8 and S9 of the Supplemental Material, uses separate noiseless and synthetic-noise simulations. No additional IBM measurements were performed for the retrospective cycle-count study.
Each isolated peak is fitted by binomial maximum likelihood to
where estimates the trapped energy and is a fitted probability scale, not a state fidelity. Repeated central counts are pooled without drift correction. We generate 2,000 parametric binomial replicas per energy fit, refitted by approximate profile weighted least squares, and independently sample the twenty centers in 1,500 MERE replicas. These finite-shot uncertainties are conditional on the peak model and exclude cross-target covariance and run-to-run drift; details of the fitting and uncertainty propagation are given in § 7.1 and § 7.2 of the Supplemental Material.
Results
All 320 circuit results were retrieved without missing or unexpected identifiers, yielding twenty fitted levels and 80,000 shots per method. As shown in Figure 2, FG–R6 reduces the median compiled depth from 459 to 307 and the median two-qubit-gate count from 113 to 71 relative to direct R10, corresponding to reductions of 33.1% and 37.2%, respectively. Both implementations use three active qubits [11]; the reduction is from ten controlled evolutions to six, not in circuit width. Mid-circuit measurement and reset still introduce latency and errors [12].
Figure 2. Compiled resources on ibm_aachen for dynamic direct R10 and FG–R6. Circuit depth, two-qubit-gate count, and total operation count are shown as medians over 160 circuits per implementation after compilation to the backend instruction set architecture (ISA). Both implementations use three active qubits.
Retrospective cycle-count tests support the six-cycle choice. With the original optimization objective, the tested and candidate sequences exceed the adopted 0.10 keV noiseless center-error tolerance, whereas satisfies it, including the tested scan-center shifts of up to 10.0 keV (maximum error 0.0926 keV). Its isolated response matches R10 within on the matching grid. An sequence separately retuned using the known spectrum also passes the local tests; FG–R6 is therefore a validated accuracy–resource compromise, not a minimum-cycle solution. The cycle-count, off-target, and simulator comparisons are given in Secs. S8 and S9 of the Supplemental Material.
Table 1 separates spectral and scattering metrics. At equal hardware shot budgets, FG–R6 gives a larger median fitted amplitude and smaller finite-shot energy uncertainty, but its RMS center offset increases from 0.793 to 1.648 keV. Improved precision therefore does not imply improved absolute energy accuracy. Repeated central queries differ by more than two nominal binomial standard errors for 11 of 20 direct-R10 targets and 1 of 20 FG–R6 targets (Table S3 of the Supplemental Material). No drift correction or uncertainty rescaling is applied. The original ideal-Aer controls give RMS offsets of 0.219 and 0.132 keV at 2,000 shots per circuit, while calibration-derived noisy Aer gives 0.491 and 0.493 keV at 500 shots. The latter does not reproduce the larger hardware offsets. These data-source comparisons, the finite-confinement MERE inputs, and representative energy and trap-removal fits are provided in Table S6, Figure S2, and Figure S3 and Figure S4 of the Supplemental Material, respectively.
| Quantity | Direct R10 | FG–R6 |
| Median | 0.064 | 0.036 |
| Maximum | 0.106 | 0.102 |
| Maximum | 3.373 | 3.314 |
| Targets with | 11/20 | 1/20 |
| Targets with | 2/20 | 1/20 |
Table S3. Original IBM repeated-center consistency over twenty targets per method. The counts are descriptive, with no multiple-comparison correction.
Figure S2. Original FG–R6 finite-confinement inputs from four data sources, each with four levels at five trap strengths. These enter the two-stage MERE fit, not independent continuum measurements.
Figure S3. First MERE stage: archived quadratic fits at MeV for original FG–R6 sources. Coefficients use in MeV and in fm; curves are least-squares fits, not arbitrary splines.
Figure S4. Second MERE stage: fixed- fits versus yield at zero trap. The five points are energy-polynomial evaluations, not extra QPU measurements. The energy step extrapolates at 5.0 MeV and interpolates at 20.0 MeV; trap removal extrapolates in both. IBM intervals and shading are pointwise 16–84% ranges from the original 1,500 archived MERE draws. Symbols at denote the fitted intercepts , not additional spectral inputs.
| Source | Method | Shots | RMS | Max. | Median | Median | ||
| Local formula | Direct R10 | – | 0.0237 | 0.0788 | – | 0.8930 | 0.00345 | 1.054 |
| Local formula | FG–R6 | – | 0.0334 | 0.0853 | – | 0.8947 | 0.01742 | 1.072 |
| Local sampled | Direct R10 | 2000 | 0.2154 | 0.6043 | 0.2023 | 0.8908 | 0.15175 | 3.723 |
| Local sampled | FG–R6 | 2000 | 0.1807 | 0.4299 | 0.1993 | 0.8960 | 0.02478 | 3.310 |
| Ideal Aer | Direct R10 | 2000 | 0.2188 | 0.4000 | 0.2056 | 0.8923 | 0.09559 | 1.437 |
| Ideal Aer | FG–R6 | 2000 | 0.1316 | 0.2934 | 0.2015 | 0.8987 | 0.11817 | 4.524 |
| Calibration Aer | Direct R10 | 500 | 0.4913 | 0.9110 | 0.5000 | 0.7039 | 0.52103 | 8.805 |
| Calibration Aer | FG–R6 | 500 | 0.4930 | 1.2165 | 0.4576 | 0.7763 | 0.53408 | 11.390 |
| IBM QPU | Direct R10 | 500 | 0.7934 | 1.6380 | 0.6099 | 0.5371 | 0.84226 | 20.119 |
| IBM QPU | FG–R6 | 500 | 1.6476 | 2.5517 | 0.5342 | 0.6520 | 0.16347 | 6.753 |
Table S6. Original production summary. Energy entries are in keV, phase entries in degrees. uses the full 300-point grid; is defined in Eq. (31). Dashes mark quantities inapplicable to deterministic probabilities. Shots are per circuit.
Figure 3(a) compares the two IBM reconstructions with the analytical solution and the exact-energy MERE reference. Figure 3(b) shows the corresponding residuals . On the full 0.1–30.0 MeV grid, the maximum absolute residual decreases from 0.842° to 0.163°, an 80.6% reduction, and the phase RMS residual decreases from 0.305° to 0.137°. The two maximum residuals occur at 1.2 and 6.1 MeV, respectively, below the lowest trapped input. The maximum deviation from the analytical solution also decreases, from 0.981° to 0.156°.
Figure 3. Free-space phase shifts reconstructed from raw IBM measurements. (a) Direct R10 and FG–R6 results from twenty trapped levels per method, compared with the analytical square-well solution and exact-energy MERE. (b) Residuals relative to the exact-energy MERE reference, . Shading denotes pointwise 16–84% finite-shot bootstrap intervals, excluding fit-form sensitivity, cross-target covariance, and run-to-run drift. Symbols identify reconstructed curves, not independent continuum measurements. The dotted line at MeV marks the onset of common energy interpolation; trap removal remains an extrapolation throughout. Maxima use MeV.
The full-grid gain is driven by the larger low-energy deviation of R10, not by uniformly improved phase shifts. On the 10.0–30.0 MeV common-interpolation subset, direct R10 has smaller maximum and RMS residuals relative to exact-energy MERE: 0.137° and 0.102°, compared with 0.159° and 0.129° for FG–R6. The maximum deviations from the analytical solution are instead similar, 0.160° and 0.156°. These subset metrics restrict the same reconstructed curves; no refitting or additional continuum measurements are involved.
The full-grid reversal shows that the RMS spectral error alone need not rank the accuracy of a reconstructed scattering observable. All twenty level shifts enter both the energy and trap-strength fits in Eqs. (3) and (4). The level-resolved offsets listed in Table S7 of the Supplemental Material are predominantly positive for FG–R6 and sign-changing for direct R10. This pattern does not establish an error-cancellation mechanism or identify the hardware noise responsible for the simulator–QPU difference.
| R10: | F6: | |||
| 3.6 | 8.855470 | +0.153 ± 0.576 | +1.838 ± 0.506 | |
| 3.6 | 16.643000 | +1.142 ± 0.591 | +1.371 ± 0.521 | |
| 3.6 | 24.213795 | +0.450 ± 0.651 | +1.682 ± 0.539 | |
| 3.6 | 31.691304 | +0.254 ± 0.593 | +1.766 ± 0.539 | |
| 3.7 | 9.129551 | +0.595 ± 0.615 | +1.271 ± 0.506 | |
| 3.7 | 17.134039 | −1.207 ± 0.643 | +2.552 ± 0.530 | |
| 3.7 | 24.916082 | +0.306 ± 0.612 | +1.276 ± 0.535 | |
| 3.7 | 32.602105 | +0.079 ± 0.613 | +2.433 ± 0.546 | |
| 3.8 | 9.404443 | −1.055 ± 0.589 | +0.987 ± 0.498 | |
| 3.8 | 17.625943 | −0.358 ± 0.599 | +1.629 ± 0.534 | |
| 3.8 | 25.619273 | +0.704 ± 0.656 | +2.035 ± 0.575 | |
| 3.8 | 33.513827 | −0.975 ± 0.618 | +0.980 ± 0.552 | |
| 3.9 | 9.680129 | +0.090 ± 0.605 | +1.209 ± 0.487 | |
| 3.9 | 18.118693 | −1.638 ± 0.643 | +1.716 ± 0.497 | |
| 3.9 | 26.323347 | −0.653 ± 0.665 | +1.311 ± 0.503 | |
| 3.9 | 34.426445 | −0.074 ± 0.620 | +2.178 ± 0.543 | |
| 4.0 | 9.956592 | +0.749 ± 0.584 | −0.068 ± 0.498 | |
| 4.0 | 18.612272 | +0.856 ± 0.586 | +1.648 ± 0.562 | |
| 4.0 | 27.028285 | −1.066 ± 0.566 | +0.624 ± 0.571 | |
| 4.0 | 35.339939 | +0.920 ± 0.608 | +2.168 ± 0.537 |
Table S7. Original IBM spectral inputs: reference energies in MeV; signed offsets and finite-shot errors in keV. Thus . Reproduction uses full-precision files, not these rounded values.
Extrapolation sensitivity remains appreciable. Across the six combinations of energy-polynomial orders one to three and trap fits with or without an term, the maximum half-range decreases from to on the full grid and from to on the common-interpolation subset. This spread is a fit-form diagnostic, not a calibrated truncation uncertainty or a finite-shot interval. All energy fits extrapolate below 8.855 MeV, and trap removal remains an extrapolation at every energy. The finite-shot bands omit cross-target covariance and run-to-run drift. Smaller central residuals therefore do not establish a corresponding reduction in total uncertainty.
| Quantity | Direct R10 | FG–R6 |
| Trapped-energy estimates: twenty levels | ||
| RMS (keV) | 0.793 | 1.648 |
| Median (keV) | 0.610 | 0.534 |
| Median | 0.537 | 0.652 |
| Phases: –30.0 MeV | ||
| 0.842 | 0.163 | |
| 0.305 | 0.137 | |
| 0.981 | 0.156 | |
| 0.429 | 0.401 | |
| 20.12 | 6.75 | |
| Phases: –30.0 MeV | ||
| 0.137 | 0.159 | |
| 0.102 | 0.129 | |
| 0.160 | 0.156 | |
| 0.074 | 0.065 | |
| 3.19 | 1.12 |
Table 1. Raw IBM benchmark: twenty trapped levels and 80,000 shots per method. Here , where is exact-energy MERE with the primary fit, and is the analytical square-well phase. Phase entries are in degrees. For phase quantities, maxima and RMS values are evaluated on an energy grid with 0.1 MeV spacing: 300 points over the full interval (0.1–30.0 MeV) and 201 over the common interpolation subset (10.0–30.0 MeV). Trap removal remains an extrapolation. Standard deviations describe finite-shot uncertainty only. is the maximum pointwise half-range over six fit forms, not a confidence interval.
The calculation extends the simulator study of Ref. [6] and our single-level static–dynamic benchmark [11] to complete hardware-derived spectral inputs for classical free-space phase-shift reconstruction, complementing quantum-subspace approaches [9]. The resource reduction and scattering results validate the combined FG–R6 implementation in the present four-dimensional OLS benchmark, not quantum advantage or many-body scalability. The hardware data do not separately determine the effects of phase compilation and time sampling, or establish independent-run reproducibility.
Conclusions
We have reconstructed elastic -wave phase shifts for a schematic interaction by combining IBM spectral measurements with classical MERE extrapolation. Each implementation supplies four positive-energy levels at five trap strengths from classically constructed four-dimensional OLS Hamiltonians, extending our single-level ancilla-recycling benchmark [11] to all twenty spectral inputs for free-space scattering.
The FG–R6 circuit retains three active qubits and combines exact query-phase separation, post-compilation binding, and six geometric evolution times. Retrospective numerical tests support this choice within the adopted local spectral tolerances, without establishing a minimum cycle count or universal optimum. Relative to dynamic direct R10, it reduces median compiled depth and two-qubit-gate count by 33.1% and 37.2%, respectively, with a larger median fitted amplitude and smaller median finite-shot energy uncertainty.
For the present dataset and primary fit on the 0.1–30.0 MeV grid, the maximum central phase-shift deviation from exact-energy MERE decreases from 0.842° to 0.163°, despite a larger RMS energy offset underestimated by the calibration-derived simulator. The full-grid gain arises mainly in the low-energy extrapolation region. On the 10.0–30.0 MeV common energy-interpolation subset, direct R10 instead has slightly smaller maximum and RMS residuals relative to the same reference. The RMS spectral error alone therefore need not rank scattering accuracy, which also depends on the reconstruction and energy interval.
A natural next step is to extend the present -wave benchmark to resonant higher-partial-wave scattering. Neutron– scattering provides a particularly suitable case because its low-energy and channels generate the unbound ground-state and excited-state resonances of , whose phase shifts and resonance parameters are well constrained by experimental -matrix analyses [2] [17]. Reconstructing these two channels would test whether the trapped-spectrum method can recover not only phase shifts but also resonance energies, widths, and the associated spin–orbit splitting. This extension requires channel-resolved trapped Hamiltonians and the higher-partial-wave effective-range function , while retaining the same RODEO-based spectral-estimation framework. A subsequent application to proton– scattering would introduce the long-range Coulomb interaction and require a Coulomb-modified trap quantization condition. Comparing the and systems would then provide a controlled route from the present short-range -wave demonstration to spin-dependent, resonant, and Coulomb-distorted nuclear scattering on quantum hardware.
Fit-form sensitivity remains appreciable, and finite-shot bands exclude cross-target covariance and run-to-run drift. These central-value comparisons do not establish a reduction in total uncertainty. Separate hardware controls and independent repetitions would clarify the contributions of compilation, time sampling, and execution variability. The present benchmark thus provides a starting point for these extensions to larger model spaces, realistic interactions, and few-body reaction dynamics.
Data Availability
The raw QPU counts, job identifiers, transpiled circuits, processed data, and analysis scripts supporting this work are available from the corresponding authors upon reasonable request.
Acknowledgments
This work was supported by the National Research Foundation of Korea (NRF) through grants funded by the Korean government (Ministry of Science and ICT, MSIT): Grant Nos. RS-2026-25487837 and RS-2018-NR031074 (M.-H.M.); Grant Nos. RS-2021-NR060129, RS-2024-00460031, and RS-2025-16071941 (M.-K.C.); and Grant No. RS-2025-00513410 (E.-J.H.). Additional support was provided through NRF Grant Nos. RS-2025-24533596 and RS-2025-25400847, funded by MSIT and the Ministry of Education, respectively (J.-B.P.). Quantum computing cloud resources were provided by the Korea Institute of Science and Technology Information (KISTI) through an NRF grant funded by the Korean government (MSIT) (Grant No. RS-2025-24534879). This research was also supported by the “Quantum Information Science R&D Ecosystem Creation” program through the NRF, funded by the Korean government (MSIT) (Grant No. 2020M3H3A1110365).
Supplemental Material for “Nuclear scattering phase shifts with factorized geometric-time RODEO on a quantum processor”
Myeong-Hwan Mun; Jubin Park; Myung-Ki Cheoun; Eunja Ha
The production dataset comprises the original IBM measurements for the dynamic direct ten-cycle RODEO implementation (direct R10) and the six-cycle factorized geometric-time implementation (FG–R6), together with their local and simulator controls. The retrospective cycle study adds classical and simulator tests only: it neither retunes the archived hardware R6 times nor replaces measured energies.
Suppl_PLB_FULL_RODEO.pdf
Hamiltonian and reference trapped spectrum
The schematic -wave neutron–proton () square well of Refs. [1, 2] provides an analytical benchmark, not a realistic interaction. With and ,
, .
The trap acts only on relative motion; energies are measured from the free threshold. Classical diagonalization uses (301 -wave states), basis frequency , and 500-point Gauss–Legendre quadrature over the well. The external frequencies are distinct from .
Let contain projections of the four lowest positive-energy reference eigenvectors onto the first four oscillator states. The hermitian Okubo–Lee–Suzuki (OLS) construction [3, 4] is
It preserves those finite-reference-space energies (Table S1) and is encoded in two system qubits. The original matrix agrees with its independent archive within per element; this reconstruction check is not an infinite-basis uncertainty or a quantum solution of the 301-state problem.
With lengths in fm, define
where . Matching the radial waves gives and
The quadrant-aware argument equals the recorded two-argument arctangent. fixes the shared continuous branch , not a correction to quantum processing unit (QPU) data. Reported phases are in degrees.
Trial states and fixed query grid
The implemented diagonal-nearest rule is
In the ordered basis , the choices are 00, 11, 10, 01 for levels 1–4 at every trap. These are OLS labels, not nucleon occupations; the rule does not explicitly maximize exact overlap (Table S2).
| (MeV) | : | : | : | : |
| 3.6 | 0.897683 | 0.883041 | 0.855202 | 0.867051 |
| 3.7 | 0.902214 | 0.894193 | 0.867914 | 0.871462 |
| 3.8 | 0.906169 | 0.904422 | 0.880093 | 0.875233 |
| 3.9 | 0.909560 | 0.913712 | 0.891722 | 0.878369 |
| 4.0 | 0.912403 | 0.922046 | 0.902786 | 0.880876 |
Table S2. Squared target overlaps for the diagonal-nearest trial states ; these bits label OLS basis states, not nucleon occupations.
The requested sequence around the known is
The central query is intentionally repeated to test success-probability consistency. At 500 shots per occurrence, the eight entries give 4,000 shots per target and 80,000 per method. Pooling the repeated center gives seven fit points: 1,000 center shots
Email addresses: [email protected] (Jubin Park), [email protected] (Myung-Ki Cheoun)
and 500 elsewhere. This is a local scan, not blind spectral discovery. For the two center occurrences,
Table S3, checked against counts, shows discrepancies beyond binomial sampling for several pairs. No drift correction or error rescaling is applied: the shot band remains conditional on pooled probabilities. These are probability-consistency checks, not separate energy fits or timing-resolved drift measurements. Requested order does not establish execution chronology, and the batches in Ref. [2] are not repeats of this twenty-level campaign.
RODEO schedules, factorization, and numerical design
For , the ideal response is [5, 1]
The R10 times are fixed standard-normal quantiles with the scale of Ref. [1], not fresh Gaussian draws:
The unchanged production R6 configuration is
Table S4 lists both sequences, used for all targets.
Exact phase separation and circuit records
Controlled evolutions use numerical exponentials of , with no product-formula approximation to the matrix. The established phase separation [1, 5] is
The scalar becomes a relative ancilla phase . FG–R6 compiles the complete target circuit with unbound, then assigns it per query; direct R10 recompiles per query. At identical times the check is
This is an elementwise discrepancy, not a spectral norm, hardware fidelity, or equivalence of R10 and R6.
Both circuits use two system qubits and one ancilla. Each cycle applies Hadamard–controlled evolution–query phase (where separate)–Hadamard–ancilla measurement, storing a fresh classical bit. There are measurements and resets, with terminal system readout after the last cycle. All entries execute fully; success is an all-zero record, irrespective of the terminal system bits.
Optimization at fixed cycle count
For geometric sampling [6], set , (), and . The loss is
Bounds are and . Differential evolution uses seed 20,260,812, population multiplier 15, tolerance , 1200 maximum iterations, and final polishing. The time-sum penalty acts on both sides of its chosen value; it models neither QPU duration nor a physical resolution limit. MeV detunings ensure dimensionless .
Production optimizes and stores parameters at fixed , without searching . The loss uses no off-target eigenenergies; the full spectrum enters subsequent validation with inherited limits
Stored R6 gives on the matching grid and maximum four-level center error 0.0853 keV over the actual keV queries. These are local design tests, not universal nuclear tolerances; Sec. S8 gives the retrospective cycle comparison.
Original IBM execution and simulation controls
Both production methods share Table S5’s layout and shot budget. Gate/measurement twirling, Runtime dynamical decoupling, fractional-gate mode, and count debiasing are off. The submitted circuits are archived in Qiskit’s portable binary QPY format and retain the backend-specific mid-circuit measurement instruction. Only temporary Aer copies replace measure_2 with standard measurement while retaining calibration-derived noise; this does not reproduce the full measurement channel or timing effects.
| Item | Recorded value |
| Backend | ibm_aachen |
| Instance | 0_2026_HU-eu |
| Layout (, , ) | [114, 119, 113] |
| Targets / query entries / methods | 20/8/2 |
| QPU circuits / shots per circuit | 320/500 |
| Shots per method / total shots | 80 000/160 000 |
| Retrieved task identifiers | 320/320 |
| Median depth: R10 / FG–R6 | 459/307 |
| Median two-qubit gates: R10 / FG–R6 | 113/71 |
| Median size: R10 / FG–R6 | 667/435 |
| Qiskit / Aer / IBM Runtime | 2.5.1/0.17.2/0.48.0 |
| Compilation level / seed | 3/20260816 |
Table S5. Original production metadata. Resource medians use 160 circuits per method; software versions describe the recorded environment.
Local formula gives exact reduced-Hamiltonian probabilities; local sampled draws 2000 binomial shots per circuit; ideal Aer is noiseless gate-level simulation at 2000 shots; calibration Aer uses the archived IBM noise model at 500 shots; and IBM QPU means 500 raw Sampler shots [7]. The two IBM implementations have equal shot budgets; the shot counts are not uniform across data sources. Agreement or disagreement with approximate Aer noise does not identify coherent, crosstalk, drift, or reset mechanisms.
Peak fitting, repeated queries, and bootstrap
The full response (S9) is approximated near one target by
without an additive background. is a probability scale, not state fidelity. At query energy , let denote the pooled all-zero ancilla count from shots. The binomial maximum-likelihood estimate (MLE) minimizes
with , omitting constant binomial factors. Bounds are and . L-BFGS-B uses 5000 maximum iterations, tolerance , and probability clipping to in logarithms. ; the Pearson diagnostic has five nominal degrees of freedom (seven points, two parameters), with no uncertainty rescaling.
For each of 2,000 parametric replicas, , where is evaluated at the original maximum-likelihood fit. Approximate weighted-least-squares (WLS) refitting uses fixed weights
and profiles the amplitude on 1001 centers :
After clipping to its bounds, the minimum weighted residual selects ; parabolic interpolation at an interior minimum refines the 0.05 keV grid within one step. The reported center remains the original binomial MLE, not the bootstrap mean.
Independent profile-binomial refits of the two simulator fits with optimizer termination warnings reproduce their centers within , leaving the rounded tabulated results unchanged. For FG–R6 in the additional simulator study, exact-MLE checks of 12 bootstrap replicas per target give absolute differences from the profile-WLS centers below 0.041 keV. These checks concern the additional simulations, not a replica-by-replica validation of the original QPU analysis.
Modified effective range expansion (MERE) reconstruction and uncertainty definitions
At each exact or fitted trapped energy, the finite-confinement input is
This is in , as is in Eq. (12). Writing does not make it a free-space observable: the trap perturbs the interaction region. Four points at each give an unweighted quadratic fit,
Evaluate all five polynomials at the same , then fit unweighted:
The principal arctangent gives for , equivalent to . Variable rescaling affects numerical conditioning, not the weights or model. Neither following one level to zero trap nor evaluating (S20) at arbitrary replaces this fixed-energy procedure (Figs. S2–S4).
Each of the 1,500 MERE replicas independently draws one stored center for each of the twenty targets, reevaluates Eq. (28) at the sampled energies, and repeats both fits. Pointwise 16–84% quantiles and standard deviations are conditional finite-shot errors; cross-target hardware covariance, batch drift, and fit-form uncertainty are excluded. Different energies on a curve share those twenty inputs and are correlated, not independent measurements.
On , all five energy fits extrapolate below 8.855470 MeV (88 points); support is mixed until 9.956592 MeV (11 points); common interpolation covers 10.0–30.0 MeV (201 points). All interval metrics restrict the same curves without refitting, and trap removal always extrapolates. Exact-energy MERE differs from the analytical reference by at most , , and in these three regions; the maximum above the lowest input is and the full-grid relative maximum is . These measure the fixed reference and primary prescription, not total theoretical or QPU uncertainty.
Six variants combine energy degrees 1, 2, 3 with trap powers or . Using a common phase convention,
Figure S1. Original production scan at MeV, . Solid curves show exact fixed-time responses; dashed curves use archived single-peak fits. Error bars are binomial standard errors with pooled central queries; displayed IBM center errors are finite-shot bootstrap uncertainties. No refitting was performed.
All six are retained (Figure S5). This is a deterministic sensitivity diagnostic, not a confidence/credible interval or calibrated truncation error, and is not added in quadrature to shot errors. A cubic through four energies has no residual degree of freedom and can amplify extrapolation sensitivity.
Figure S5. Original IBM FG–R6 fit-form sensitivity: all six energy/trap variants are retained and the primary fit is identified. This envelope is not the finite-shot band of the main-text phase plot. Legend labels specify the energy-polynomial degree and the basis functions retained in the trap fit. The vertical dotted line at MeV marks the onset of common energy interpolation; trap removal remains an extrapolation throughout.
Original production performance and hardware inputs
Table S6 keeps the original sources separate. Even noiseless full-spectrum data can yield nonzero single-peak bias because Eq. (24) omits other states. Local sampling isolates shot noise; IBM F6 has larger amplitude and better median shot precision but larger energy RMS than D10, not reproduced by calibration Aer. The signs in Table S7 establish neither covariance nor error cancellation. These twenty fitted energies supply each IBM curve; none is replaced by an exact eigenvalue or sampled-subspace result.
The full-grid exact-MERE residual maxima, 0.842° (R10) and 0.163° (F6), occur at 1.2 and 6.1 MeV, below all inputs. On the common-interpolation subset, R10 has slightly smaller maximum and RMS reference residuals, while maxima against the analytical solution are similar, 0.160°/0.156° (Table S8). Model half-ranges are 20.12°/6.75° on the full grid and 3.19°/1.12° on the subset. Thus the full-grid central improvement is not uniform or a total-error reduction.
| (MeV) | Method | Max. | RMS | Max. | Max. | |
| 0.1–30.0 | Direct R10 | 0.842 | 0.305 | 0.981 | 0.429 | 20.12 |
| 0.1–30.0 | FG–R6 | 0.163 | 0.137 | 0.156 | 0.401 | 6.75 |
| 10.0–30.0 | Direct R10 | 0.137 | 0.102 | 0.160 | 0.074 | 3.19 |
| 10.0–30.0 | FG–R6 | 0.159 | 0.129 | 0.156 | 0.065 | 1.12 |
Table S8.
Full-grid analytical residual maxima (R10/F6) are 0.138/0.158° for local response, 0.237/0.258° for ideal Aer, 0.660/0.396° for calibration Aer, and 0.981/0.156° for IBM (Figure S6). They use a different reference from ; maxima cannot be added or subtracted to convert between them.
Figure S6. Original production FG–R6 residuals relative to the analytical square-well phase. The main-text residual instead uses exact-energy MERE. These sources are distinct from the retrospective synthetic-noise study. The vertical dotted line marks the onset of common energy interpolation at MeV.
Retrospective justification of the six-cycle sequence
The unchanged production R6 is compared with fixed- designs using the same loss, bounds, and one optimizer seed. The retuned R4 ( MeV, ) was separately selected using the known four-level spectrum, unlike the isolated-response loss. This adds no hardware data. Deterministic probabilities at seven queries are fitted by profile least squares, not finite-shot binomial MLE. Anchor shifts of −10, −5, 0, 5, 10 keV leave the Hamiltonian and true target fixed; errors are measured from that target, not the shifted anchor.
Under the original loss, returned R4/R5 fail the inherited 0.10 keV center limit, whereas R6 passes even at shifted anchors and matches the local R10 response much more closely (Table S9). Retuned R4 also passes: no minimum-cycle or global-optimality claim follows from these single-seed tests. R8 improves several ideal measures at greater cycle cost. Tolerances were not tightened after inspecting shorter candidates.
| Sequence | (keV) | (keV) | |||
| R10 reference | – | – | 0 | 0.07882 | 0.08623 |
| R4, original loss | 31.542746 | 1.010000 | 0.12303 | 0.13505 | |
| R4, retuned | 31.771753 | 1.015000 | 0.08986 | 0.09577 | |
| R5, original loss | 35.000916 | 1.139906 | 0.31481 | 0.36376 | |
| R6, production | 36.505394 | 1.199231 | 0.08533 | 0.09255 | |
| R7, original loss | 37.232115 | 1.228648 | 0.07222 | 0.08738 | |
| R8, original loss | 37.712886 | 1.246506 | 0.03128 | 0.03996 |
Table S9.
Local resolution is retained:
R10/R6 second moments are 3879.8622/3879.7596 MeV; the median isolated-peak Fisher lower bounds for the actual query allocation and overlaps are 0.402008/0.402010 keV. These describe ideal local precision, not hardware errors.
Declared off-target domains
Each non-target gap is broadened by the ±40 keV query range plus 10 keV padding per side; overlapping intervals are merged in detuning_domains.json. Positive intervals span approximately 7.4275–25.4333 MeV with gaps; evenness of covers negative detuning. With grid spacing at most MeV,
Bounds add when comparing two filters. These are enclosures only on the declared domains; no mandatory off-target tolerance was enabled. The weighted non-target contribution at is
It differs from and from conditional-state infidelity (Table S10). Figure S7 shows local response and recurrences. R6 is a tested accuracy–resource compromise for this spectrum/window; other overlaps, Hamiltonians, or required scattering accuracy can require different times or . Factorization does not single out six cycles.
Figure S7. Retrospective filter comparison. Local agreement does not ensure off-target agreement. The 0–1 MeV panel shows recurrences, not actual interlevel spacings or a suppression certificate; Table S10 uses the declared spectral intervals.
| Sequence | ||
| R10 reference | 0.05675 | 0.001265 |
| R4, original loss | 0.25514 | 0.008530 |
| R4, retuned | 0.54326 | 0.007250 |
| R5, original loss | 0.38644 | 0.014574 |
| R6, production | 0.31651 | 0.003184 |
| R7, original loss | 0.42230 | 0.021778 |
| R8, original loss | 0.04347 | 0.000792 |
Table S10.
Table S8: Original IBM phase metrics (degrees), as in the main text. . The full and common-interpolation subsets contain 300 and 201 points from the same curves, without refitting.
Table S9: Retrospective cycle tests with the original loss. is the matching-grid response difference; and are maximum four-level center errors (keV) at the nominal and shifted anchors. is in MeV. Retuned R4 uses the known spectrum.
Table S10: Ideal off-target diagnostics on the declared domains: transmission upper bound and weighted non-target probability at central queries. Neither is a conditional-state infidelity; no mandatory off-target tolerance was imposed.
Additional simulator comparisons and statistical checks
All sampled retrospective sources use 500 shots per occurrence on the same targets/queries, unlike the original ideal/local-sampled controls. The methods are D10/F10 (direct/factorized R10), D6/F6 (direct/factorized with production R6 times), F4C (spectrum-informed R4), F5O/F7O (optimized R5/R7), and F6S (signed R6). R4O error only classical tests. F6S multiplies the R6 times by ; cosine squares preserve ideal all-zero probabilities, not necessarily the full accepted-state operator. It is neither an IBM modification nor established mitigation.
Exploratory SGC–R6 files describe a different spectrum-guided sequence.
Independent density and synthetic-noise models
The complete three-qubit density-matrix propagation includes controlled evolution, phase, measurement and reset, agreeing with the fixed-time formula. Local sampling draws binomial ideal counts. Cycle-level noise uses with per cycle, ancilla assignment error 0.015, and reset excitation 0.005. The accepted branch is the recorded all-zero branch, including assignment errors, not an ad hoc attenuation.
Offline Aer compilation uses level 3, seed 20260912, and without IBM topology. Synthetic noise uses depolarizing parameters on SX/X and on CX, symmetric readout error 0.01, and X after reset with probability 0.005; has no added stochastic error. These are assumed dimensionless parameters, not measured infidelities or IBM calibration. Recorded versions are Python 3.11.15, NumPy 2.4.6, SciPy 1.17.1, Qiskit 2.5.1, Aer 0.17.2, Runtime 0.48.0; base seed 20260912 and task/source seeds are archived. Energy/MERE replica counts remain 2000/1500.
F6 improves amplitude and shot precision in both assumed noise models, not all phase errors (Table S11); synthetic-Aer is versus for D10. Identical-time direct/factorized probabilities agree ideally, while shot estimates differ. Six cycles lower offline gate–shot cost, but phase factorization alone does not (Table S12); these costs do not replace IBM topology-mapped resources.
Independent shot realizations and method contrasts
One hundred independent noiseless shot realizations show comparable D10/F6 energy precision (Table S13), not hardware repeatability. For on the full grid, independent-target bootstrap contrasts all span zero (Table S14); central rankings are not universal accuracy or reproducibility claims.
| Method | RMS | Med. | Med. | |||
| Deterministic density matrix | ||||||
| D10 | 10 | 0.0237 | – | 0.8930 | 0.00345 | 1.054 |
| F10 | 10 | 0.0237 | – | 0.8930 | 0.00345 | 1.054 |
| D6 | 6 | 0.0334 | – | 0.8947 | 0.01742 | 1.072 |
| F6 | 6 | 0.0334 | – | 0.8947 | 0.01742 | 1.072 |
| F4C | 4 | 0.0449 | – | 0.8954 | 0.07909 | 1.054 |
| F5O | 5 | 0.1389 | – | 0.8968 | 0.07488 | 3.997 |
| F7O | 7 | 0.0283 | – | 0.8966 | 0.03093 | 1.769 |
| F6S | 6 | 0.0334 | – | 0.8947 | 0.01742 | 1.072 |
| Local binomial sampling | ||||||
| D10 | 10 | 0.3662 | 0.4060 | 0.8907 | 0.77757 | 3.914 |
| F10 | 10 | 0.3083 | 0.4015 | 0.8909 | 0.16326 | 6.523 |
| D6 | 6 | 0.4544 | 0.4048 | 0.8913 | 0.27730 | 5.784 |
| F6 | 6 | 0.3580 | 0.4002 | 0.8989 | 0.11661 | 9.036 |
| F4C | 4 | 0.4060 | 0.4024 | 0.8917 | 0.15754 | 6.641 |
| F5O | 5 | 0.4265 | 0.3993 | 0.8978 | 0.21506 | 5.761 |
| F7O | 7 | 0.3319 | 0.4031 | 0.8939 | 0.08197 | 3.321 |
| F6S | 6 | 0.4002 | 0.4061 | 0.8876 | 0.08009 | 25.788 |
| Cycle-level assumed noise | ||||||
| D10 | 10 | 0.4797 | 0.5153 | 0.6829 | 0.62365 | 6.449 |
| F10 | 10 | 0.4666 | 0.5106 | 0.6912 | 0.57843 | 6.271 |
| D6 | 6 | 0.6373 | 0.4637 | 0.7642 | 0.14399 | 65.307 |
| F6 | 6 | 0.4069 | 0.4659 | 0.7674 | 0.15785 | 8.031 |
| F4C | 4 | 0.3612 | 0.4455 | 0.8094 | 0.33733 | 6.056 |
| F5O | 5 | 0.4688 | 0.4555 | 0.7918 | 0.76957 | 7.910 |
| F7O | 7 | 0.4835 | 0.4751 | 0.7437 | 0.27558 | 41.057 |
| F6S | 6 | 0.3784 | 0.4603 | 0.7659 | 0.05048 | 11.036 |
| Ideal Aer (500 shots) | ||||||
| D10 | 10 | 0.4237 | 0.4013 | 0.8930 | 0.27816 | 1.473 |
| F10 | 10 | 0.4228 | 0.3995 | 0.9009 | 0.38658 | 3.190 |
| D6 | 6 | 0.4723 | 0.3996 | 0.8901 | 0.31652 | 8.923 |
| F6 | 6 | 0.4770 | 0.4033 | 0.8885 | 0.06947 | 12.480 |
| F4C | 4 | 0.2619 | 0.4022 | 0.8957 | 0.24655 | 3.329 |
| F5O | 5 | 0.3618 | 0.4007 | 0.9005 | 0.32664 | 22.162 |
| F7O | 7 | 0.3543 | 0.3991 | 0.8985 | 0.27549 | 17.752 |
| F6S | 6 | 0.4329 | 0.4010 | 0.8949 | 0.19736 | 11.633 |
| Synthetic noisy Aer (500 shots) | ||||||
| D10 | 10 | 0.3460 | 0.5159 | 0.6810 | 0.21115 | 12.370 |
| F10 | 10 | 0.6144 | 0.5119 | 0.6775 | 0.52271 | 11.054 |
| D6 | 6 | 0.4961 | 0.4636 | 0.7645 | 0.60839 | 70.771 |
| F6 | 6 | 0.5239 | 0.4677 | 0.7591 | 0.39983 | 64.448 |
| F4C | 4 | 0.4232 | 0.4398 | 0.8098 | 0.22516 | 9.192 |
| F5O | 5 | 0.5547 | 0.4556 | 0.7961 | 0.61483 | 46.785 |
| F7O | 7 | 0.3861 | 0.4803 | 0.7425 | 0.29006 | 2.596 |
| F6S | 6 | 0.3735 | 0.4646 | 0.7642 | 0.09890 | 9.670 |
Table S11. Retrospective simulator results, separate from production (Table S6). Energy entries are in keV. All sampled sources use 500 shots per occurrence; deterministic quantities have no shot error. and are full-grid degrees. The flagged simulator fits were independently checked without changing the rounded entries (Sec. S5).
Table S13: One hundred independent noiseless shot realizations: mean twenty-level energy RMS (keV) and the distribution (degrees) of full-grid . These are not hardware repeats.
| Method | Mean RMSE | Median | 16–84% range |
| D10 | 0.4056 | 0.2044 | [0.0868, 0.4059] |
| F6 | 0.4027 | 0.1774 | [0.0718, 0.4061] |
Table S13.
Table S14: Conditional bootstrap contrast (degrees); positive favors F6. Quantiles exclude run-to-run hardware variability.
| Source | Median | 2.5% | 97.5% |
| Ideal Aer | 0.0776 | −0.4939 | 0.7046 |
| Cycle-level noise | 0.3121 | −0.4883 | 1.1058 |
| Synthetic noisy Aer | −0.0558 | −0.9031 | 0.6962 |
Table S14.
High-order extrapolation sensitivity
Synthetic-Aer F6 reaches at MeV (Fig. S8): primary and cubic-energy/ phases are and , with intercepts and . This is a change in , not merely a phase-branch plotting error; exact inputs with the latter model give . For , the trap weights are
Their norms 6.03/96.70, combined with cubic energy extrapolation, explain strong sensitivity to small spectral perturbations. On the common interpolation subset, the synthetic F6 half-range remains versus primary residual . The full spread is retained, not removed, and is distinct from the original IBM 6.75°.
| Method | Median depth | Mean 2Q gates | 2Q gate–shots |
| D10 | 202.0 | 63.6 | 5 088 000 |
| F10 | 208.5 | 63.6 | 5 088 000 |
| D6 | 130.0 | 40.0 | 3 200 000 |
| F6 | 128.5 | 40.0 | 3 200 000 |
| F4C | 92.0 | 28.1 | 2 248 000 |
| F5O | 110.5 | 33.0 | 2 640 000 |
| F7O | 147.0 | 44.3 | 3 544 000 |
| F6S | 129.0 | 40.0 | 3 200 000 |
Table S12. Offline compilation in the basis, without IBM topology. Gate–shots sum two-qubit gates times 500 shots over 160 entries; this is not execution time.
Figure S8. Retrospective synthetic noisy-Aer F6 fit-form sensitivity, not IBM production. All six saved curves and both references are shown. The half-range reaches at MeV, separately from shot uncertainty and hardware repeatability. Legend labels specify the energy-polynomial degree and the basis functions retained in the trap fit. The vertical dotted line marks the onset of common energy interpolation at MeV.
References
- [1]P. Wang, W. Du, W. Zuo, J. P. Vary, Nuclear scattering via quantum computing, Phys. Rev. C 109 (2024) 064623, doi:10.1103/PhysRevC.109.064623.DOI
- [2]M.-H. Mun, J. Park, M.-K. Cheoun, E. Ha, From twelve to three active qubits: Ancilla-recycled rodeo filtering for trapped neutron–proton scattering, arXiv:2608.26569 (2026).
- [3]S. Ōkubo, Diagonalization of Hamiltonian and Tamm–Dancoff equation, Prog. Theor. Phys. 12 (1954) 603–622, doi:10.1143/PTP.12.603.DOI
- [4]K. Suzuki, S. Y. Lee, Convergent theory for effective interaction in nuclei, Prog. Theor. Phys. 64 (1980) 2091–2106, doi:10.1143/PTP.64.2091.DOI
- [5]K. Choi, D. Lee, J. Bonitati, Z. Qian, J. Watkins, Rodeo algorithm for quantum computing, Phys. Rev. Lett. 127 (2021) 040505, doi:10.1103/PhysRevLett.127.040505.DOI
- [6]M. Patkowski, O. Ayylildiz, K. Hunt, N. Jansen, D. Lee, Improved RODEO algorithm performance for spectral functions and state preparation, arXiv:2602.05978 (2026).
- [7]A. Javadi-Abhari et al., Quantum computing with Qiskit, arXiv:2405.08810 (2024).
- [8]T. Luu, M. J. Savage, A. Schwenk, J. P. Vary, Nucleon-nucleon scattering in a harmonic potential, Phys. Rev. C 82 (2010) 034003, doi:10.1103/PhysRevC.82.034003.DOI
- [9]X. Zhang, Extracting free-space observables from trapped interacting clusters, Phys. Rev. C 101 (2020) 051602(R), doi:10.1103/PhysRevC.101.051602.DOI