Neura Parse
HomeBlog

Observable-adapted, inversion-free quantum estimation of nuclear electroweak cross sections

A research gap for computing experimentally weighted nuclear observables directly from quantum moments, without reconstructing the full response function first.

June 18, 202614 min readNeura Parse Research
Scientific visualization of a nuclear response workflow with quantum circuit traces and moment bars

Scientific visualization of a nuclear response workflow with quantum circuit traces and moment bars

Literature horizon

Core baselines

Separated error bands

First physics benchmark

The open problem is not another nucleus on a quantum computer. It is the end-to-end combination of a realistic Hamiltonian, a physical electroweak probe, direct weighted-observable estimation, certified algorithmic uncertainty, and a resource comparison against reconstruction-first methods.

The literature now contains several important pieces: Chebyshev reconstruction with rigorous finite-resolution bounds, Fourier moment computation on IBM hardware with mitigation, a 2026 O-19 quantum-classical response calculation with realistic internucleon interactions, and a 2026 proof of concept showing that some flux-averaged neutrino cross sections can be obtained from Euclidean-response moments without reconstructing the full response.

What still appears open, as of the June 18, 2026 targeted scan, is the complete combination: a realistic nuclear Hamiltonian, a physical electroweak system-probe operator, direct computation of the experimentally weighted observable, a certified algorithmic error budget, and a resource comparison against reconstruct-then-integrate baselines.

That is a sharper gap than simply finding a nucleus that has not been simulated yet. The scientific object is the whole estimator and its uncertainty contract.

This is not a priority claim. Before submission, the gap should be rechecked through INSPIRE, Web of Science, Scopus, and direct citation chasing.

For a ground state |Psi_0>, Hamiltonian H, and physical excitation operator O(q), the response function is usually written as an energy-resolved spectral object. Traditional workflows reconstruct that function and then integrate it against a kinematic or flux weight.

The observable needed by an experiment is often narrower: a known weighted integral over energy transfer. By the spectral theorem, that weighted integral can be written as an expectation value of a function of the Hamiltonian.

R_O(omega,q) = sum_n |<Psi_n|O(q)|Psi_0>|^2 delta[omega - (E_n - E_0)]

Sigma_W(q) = integral W(omega,q) R_O(omega,q) d omega

Sigma_W(q) = <Psi_0| O(q)^dagger W(H - E_0,q) O(q) |Psi_0>
  • The reconstruction-first problem asks for a high-resolution function R_O(omega,q).
  • The observable-adapted problem asks only for W(H - E_0,q) at the accuracy required by the measured cross section.
  • The second problem can be smaller, but only if polynomial degree, shot budget, state preparation, and physical weights are designed together.

After rescaling the relevant Hamiltonian spectrum to [-1, 1], the experimental weight can be approximated by a Chebyshev, Fourier, piecewise-polynomial, or rational representation. A first implementation can start with Chebyshev polynomials because the response-function literature already has strong error-control tools.

The physical source state is prepared from the probed ground state. The quantum device estimates Hamiltonian moments in that source state. The classical side computes the weight coefficients and combines them into the cross-section estimator.

W(a x + b,q) ~= p_K(x) = sum_{k=0}^K c_k(q) T_k(x)

|Omega> = O(q)|Psi_0> / sqrt(S_O(q))
S_O(q) = <Psi_0|O(q)^dagger O(q)|Psi_0>

Sigma_W(q) ~= S_O(q) sum_{k=0}^K c_k(q) mu_k(q)
mu_k(q) = <Omega|T_k(H_scaled)|Omega>
  • Chebyshev coefficients are not chosen only for approximation quality; they should also control variance and circuit cost.
  • If moment estimates are approximately independent, shot allocation should scale with |c_k| and the moment standard deviation.
  • With correlated estimators, the uncertainty should be propagated through the covariance matrix c^T C_mu c.

The identity above is useful, but it is not enough. The algorithmic contribution should be the optimization loop that chooses polynomial degree, coefficients, and measurement allocation under a target observable-level error budget.

This turns the problem from response reconstruction into cost-aware estimator design. The polynomial is judged by the cross section it produces, not by a generic pointwise fit to a response curve.

minimize   C_quantum(K, {c_k}, {N_k})
subject to epsilon_poly + epsilon_stat + epsilon_prep + epsilon_sim + epsilon_noise <= epsilon_target

Var(Sigma_W) ~= S_O^2 sum_k c_k^2 sigma_k^2 / N_k
or
Var(Sigma_W) = S_O^2 c^T C_mu c

A realistic electroweak operator contains one-body and two-body pieces. For the first paper, the responsible scope is a physical one-body charge, longitudinal, or E1 operator. Two-body current implementation can be analyzed as a scaling extension rather than forced into the first benchmark.

Because the operator is generally nonunitary, the source state can be prepared through LCU or block-encoding. The metrics to report should include the LCU normalization, post-selection success probability, ancilla count, two-qubit or logical T-gate cost, and leakage from the intended symmetry sector.

  • Prepare coefficients for O(q) as a linear combination of implementable unitaries.
  • Measure or bound the source norm S_O(q) separately.
  • Check particle number, angular momentum, and center-of-mass contamination.
  • Reserve leading two-body currents for a second-stage resource and uncertainty analysis.

The minimum viable paper should not chase fault-tolerant advantage. It should prove that the estimator produces physical, normalized, uncertainty-calibrated results and show where it is cheaper or not cheaper than reconstruction-first methods.

A sensible ladder starts with synthetic spectra and small Hermitian matrices, then moves to deuteron E1 photodisintegration or longitudinal response, then to H-3, He-3, or He-4 with a physical one-body current.

  • Baseline 1: exact diagonalization for small systems.
  • Baseline 2: reconstruct then integrate with Chebyshev, LIT, or GIT reconstruction.
  • Baseline 3: real-time Fourier correlation methods.
  • Baseline 4: direct weighted-observable estimation with the proposed W(H) moment combination.
  • Report total cost as state preparation plus Hamiltonian calls plus shots plus post-selection plus classical post-processing.

The article should not hide every error source inside a single bar. Algorithmic uncertainty can be certified or calibrated at the observable level. Physics-model uncertainty should be treated through EFT order, basis truncation, interaction ensembles, current operators, and low-energy constants.

The strongest reporting format is two-layered: one interval for algorithmic confidence and one credible band for physics modeling.

epsilon_alg = epsilon_poly + epsilon_shot + epsilon_state + epsilon_Hamiltonian + epsilon_operator + epsilon_mitigation

epsilon_physics = epsilon_basis + epsilon_EFT + epsilon_current + epsilon_LEC

Report: Sigma_W +/- Delta_algorithmic with separate -Delta_physics/+Delta_physics
  • Polynomial truncation should have a deterministic or validated upper bound.
  • Finite-shot statistics should be tested with coverage experiments.
  • Noise-mitigation bias should be calibrated and reported as a possible systematic.
  • EFT and basis errors should be described as physics uncertainty, not algorithmic certification.

The direct estimator will not win everywhere. If many different W functions must be queried after the fact, a reconstructed response can amortize its cost. The useful output is therefore a crossover map: which weights, resolutions, and query counts make direct estimation attractive.

The minimum success criteria should be predetermined: a physical one-body probe, deuteron benchmark, noiseless total algorithmic error near or below 2 percent, 95 percent intervals with credible coverage under noisy simulations, and reproducible code plus small Hamiltonian matrices.

  • If sharp windows cause high polynomial degree, move to smoothed bins, Jackson damping, piecewise Chebyshev, or rational approximations.
  • If source preparation is expensive, use symmetry grouping, low-rank operator factorizations, amplitude amplification, or a composite O^dagger W(H)O encoding.
  • If hardware circuits are too deep, use hardware only to validate low-degree moments and keep the main claim algorithmic.

The target observable is a weighted cross section, not a fully reconstructed response curve.

The estimator should co-optimize polynomial approximation and quantum shot allocation.

Physical source-state preparation is a central contribution, not an implementation detail.

Algorithmic confidence intervals and nuclear-model credible intervals should remain separate.

The honest result is a crossover map against exact, reconstruction-first, and Fourier baselines.