Pith. sign in

REVIEW 4 major objections 4 minor 53 references

The QPE measurement distribution is exactly a closed-form Dirichlet kernel, so a shallow variational circuit can be trained to mimic QPE without simulating it.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-01 08:55 UTC pith:XXEY5EM6

load-bearing objection The Dirichlet-target training trick is real, but the energy-recovery headline is circular: the target is built from the exact FCI energy, so the 'recovered' energy is just the input reference read back out. the 4 major comments →

arxiv 2607.20943 v1 pith:XXEY5EM6 submitted 2026-07-23 quant-ph cs.ETcs.LGphysics.chem-ph

An Analytically Trained Variational Surrogate for Quantum Phase Estimation on NISQ Hardware

Mousumi Kundu , Ashish Kumar Patra , Anurag K. S. V. , Ruchika Bhat , Sai Shankar P. , Alok Shukla , Jaiganesh G This is my paper
classification quant-ph cs.ETcs.LGphysics.chem-ph PACS 03.67.-a
keywords quantum phase estimationDirichlet kernelvariational quantum circuitsmolecular ground-state energyNISQ hardwarehydrogen moleculedynamical decouplingchemical accuracy
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper claims that the probability distribution measured by quantum phase estimation can be replaced by a closed-form Dirichlet kernel computed classically from the exact ground-state energy, the ancilla qubit count, and the evolution time. That replacement removes the exponential cost of simulating QPE to generate training data for a variational surrogate circuit. A shallow single-layer variational circuit trained against this analytical target on the hydrogen molecule reproduces the distribution's dominant peak and sidelobes, and the recovered energy lands within the chemical-accuracy threshold on real NISQ hardware. A reader should care because the result points to a scalable route for turning QPE-style spectral information into hardware-amenable circuits, provided a reference energy is available.

Core claim

The paper's central claim is that the QPE measurement distribution over ancilla bitstrings is exactly specified by the squared Dirichlet kernel once the phase is known, and in a chemistry setting the phase is determined by the FCI ground-state energy, the ancilla count, and the time parameter. The training target for a variational surrogate can therefore be computed entirely classically in time proportional to the number of bitstrings, with no simulated QPE circuit. The authors demonstrate on H2, after symmetry tapering, that a linear-entangler RY-CZ ansatz with a single layer, trained with COBYLA on this target, yields a hardware output distribution whose dominant bitstring identifies the s

What carries the argument

The closed-form squared Dirichlet kernel, P_gt(k;φ) = (1/N²)[sin(πNδ_k)/sin(πδ_k)]² with N = 2^n_a and δ_k = φ − k/N, evaluated from the exact ground-state energy, the ancilla count, and the evolution time. This object carries the argument because it converts the full QPE measurement statistics into a cheap classical table of probabilities, so the variational circuit can be trained to mimic QPE without ever running or simulating the QPE circuit. The load-bearing identity is that the measured bitstring probabilities of textbook QPE are exactly this kernel.

Load-bearing premise

The load-bearing premise is that the exact ground-state energy of the molecule is known in advance; the paper itself notes that such a reference will generally not be available for molecules where quantum advantage is sought, so without it the analytical target cannot be built.

What would settle it

Take a molecule for which the exact energy is unknown but a good approximate energy exists, run the pipeline with the approximate energy as the phase input, and check whether a single-layer linear-entangler variational circuit trained on the resulting Dirichlet target recovers the true energy within chemical accuracy on hardware. A deviation at or above the chemical-accuracy threshold would show the target's dependence on the exact reference energy is load-bearing in the regime the paper aims at.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Training a QPE surrogate no longer requires classically simulating the QPE circuit; the target is available in closed form for any system where the ground-state energy, ancilla count, and evolution time are known.
  • A single-layer linear-entangler variational circuit is sufficient to reproduce the dominant QPE peak and its decaying sidelobes on NISQ hardware for H2, with the recovered energy within chemical accuracy.
  • For this task, deeper circuits (p = 1 to 5) monotonically hurt hardware performance, and the reduced RY-CZ ansatz matches the RY-RZ-CZ ansatz while using half the rotation parameters.
  • Applying XpXm dynamical decoupling consistently degraded distributional agreement in every tested configuration, so the paper recommends architectural choices over pulse-level DD for this circuit family.
  • Because the surrogate reproduces sidelobes as well as the peak, the learned circuit carries QPE's spectral structure, not just the most likely bitstring.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The method's practical reach is bounded by availability of the exact ground-state energy; the paper itself notes that for molecules where quantum advantage is sought this reference will generally be unknown. A natural test is to feed coupled-cluster or other approximate energies into the same pipeline and measure how reference error propagates to the recovered energy.
  • The same analytical replacement should apply to any eigenphase of a unitary, so the framework could be extended to excited states or to non-chemical unitaries by replacing the single-peak target with a multi-peak Dirichlet mixture.
  • The surprising dynamical-decoupling result suggests a testable hardware prediction: on circuits with deliberately padded idle windows, DD should reverse its effect, so comparing the same ansatz with and without inserted idle time would isolate whether the harm comes from pulse calibration or from lack of idle space.
  • The O(2^n_a) evaluation of the target is classically trivial for small ancilla counts but grows in bitstring count; if n_a grows large, sampling the Dirichlet kernel probabilistically rather than enumerating all outcomes is an obvious scaling extension.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper proposes a variational surrogate for quantum phase estimation (QPE) in which a shallow VQC is trained to reproduce the QPE measurement distribution. The training target is not obtained by simulating the QPE circuit, but from a closed-form Dirichlet kernel evaluated using the exact FCI ground-state energy, the ancilla count n_a, and the evolution time τ. The authors apply the method to H₂ in a symmetry-tapered STO-3G Hamiltonian, conduct a four-stage hardware study on IBM Quantum devices comparing entangler topologies, circuit depth, ansatz variants, and dynamical decoupling, and report that the trained VQC reproduces the dominant QPE bitstring and recovers the ground-state energy within chemical accuracy (|ΔE| = 1.786×10⁻⁴ Ha). They argue that this removes the exponential classical simulation bottleneck of prior QPE-surrogate approaches.

Significance. If the central claim were valid, the paper would offer a way to train QPE surrogates without QPE circuit simulation, reducing the computational cost of generating training data. The hardware experiments are systematic: four distributional metrics are used, multiple ansätze and depths are compared, and the side-lobe analysis of Section 4.7 is a useful diagnostic that goes beyond dominant-bitstring counting. The paper is also unusually candid in its Section 5 limitations, explicitly acknowledging that the FCI reference energy needed to build the target will generally be unavailable for molecules where quantum advantage is sought. However, this admission undermines the paper's headline claim. The energy 'recovered' in Eq. (21) is the same EFCI inserted into Eq. (18), so the reported chemical accuracy is a property of the input, not of the learning procedure. The framework is best viewed as a distribution-learning benchmark on a known answer, not as a scalable method for estimating unknown molecular energies. Given the central claim, the significance is therefore substantially lower than the abstract suggests.

major comments (4)
  1. [§3.2, Eqs. (18)–(22)] The energy-recovery claim is circular. Eq. (18) defines Eshift = EFCI − c0 − Enuc, Eq. (19) converts this to a phase, and Eq. (20) builds the target distribution Pgt(k; φ) from that phase. The dominant bitstring k* is then read off from the same Pgt, and Eq. (21) inverts k* to obtain Etotal. By construction Etotal ≈ EFCI up to the discretization error of the 2^na grid. Table 2's |ΔE| = 1.786×10⁻⁴ Ha is therefore a rounding artifact of the chosen na and τ applied to the input EFCI, not a measure of the VQC's estimation ability. Any ansatz that successfully outputs the precomputed bitstring 00101010011 will appear 'chemically accurate' regardless of whether it has learned any physics. The energy-recovery claim must be removed or thoroughly reframed.
  2. [§5, Limitations] The paper itself states: 'the reference energy needed to construct Pgt(k; φ) will generally not be available in practice' for larger molecules. This is a load-bearing limitation, not a peripheral one. Without access to EFCI the training target cannot be built, so the method cannot estimate unknown ground-state energies. The abstract's conclusion that the framework constitutes 'a scalable, hardware-efficient paradigm for QPE-based molecular energy estimation on NISQ devices' is unsupported by the presented evidence. The method is currently a surrogate-training benchmark for a known energy, not an energy-estimation algorithm.
  3. [§2.3 and §3.5] The claim that the training pipeline is 'fully classical' and requires 'no quantum circuit simulation at any stage' is overstated. Section 3.5 describes COBYLA training in which the VQC is executed on AerSimulator in both ideal and noisy modes; these are quantum circuit simulations of the VQC. What the paper avoids is simulation of the QPE circuit, not all quantum circuit simulation. Moreover, evaluating Pgt over all 2^na bitstrings is O(2^na), which is exponential in the ancilla count; replacing QPE simulation with FCI evaluation does not remove exponential scaling — it relocates it to the classical exact diagonalization used to obtain EFCI.
  4. [§4.5, Table 2] Table 2 reports Etotal as '–' and |ΔE| as '<1.6×10⁻³' for the hardware configurations, without actual numerical values. Since the energy estimate is fully determined by the dominant bitstring, these entries add no information beyond the bitstring counts already reported. The table should either report the measured bitstring statistics with uncertainty (e.g., bootstrap confidence intervals on the dominant-bitstring probability) or explicitly state that the energy value is the discretized input EFCI. As presented, the table gives the misleading impression that an energy was independently estimated on hardware.
minor comments (4)
  1. [§2.2 vs §3.2] The notation for the phase φ is inconsistent. In Eq. (2), φ = 2^na θ is the rescaled phase, while in Eq. (19) φ = |Eshift|τ/(2π) is the normalized phase θ. Eq. (20) then treats φ as θ. Please use distinct symbols (e.g., φ for rescaled phase and θ for normalized phase) throughout, and check the factor 2^na in Eqs. (7), (19), and (21) accordingly.
  2. [§5, first paragraph of Limitations] The sentence beginning 'For larger molecules and basis sets where the FCI energy itself scales exponentially ... meaning that the reference energy needed to construct Pgt(k; φ) will generally not be available in practice' is grammatically incomplete and difficult to parse. Please revise.
  3. [§3.8] The Laplace smoothing ε_smooth = 10⁻¹⁰ is applied 'to zero-count bins' in the JSD computation, but the JSD formula in Eq. (14) uses M = (P+Q)/2, and if either P or Q has a zero bin, the KL term may still diverge unless P and Q are both smoothed. Please state explicitly how the smoothing is applied to both distributions and whether it affects the reported metric values significantly.
  4. [§4.7, Figure 11] The single-bit-flip sidelobe analysis is interesting, but the claim that these outcomes 'have no analytical basis for appearing in the Dirichlet target' is not fully justified: any finite-shot sampling of the target distribution can populate such bins. A statistical significance test (e.g., comparing observed counts to the binomial confidence interval from the Dirichlet expectation) would strengthen the claim that the Δ(b) = −16 peak is a hardware artifact rather than statistical fluctuation.

Circularity Check

3 steps flagged

Energy recovery is circular: the Dirichlet target is built from EFCI (Eqs. 18-20), so the 'recovered' energy (Eqs. 21-22) is the input EFCI up to discretization; Section 5 concedes EFCI is unavailable for useful molecules.

specific steps
  1. fitted input called prediction [Section 3.2, Eqs. (18)-(20)]
    "Eshift =E FCI −c 0 −E nuc, ... φ= |Eshift|τ/2π ... Pgt(k;φ) = 1/N^2 [sin(πNδ_k)/sin(πδ_k)]^2, δ_k = φ − k/N, N=2^na"

    The training target is defined directly from the exact FCI energy. Its peak k* = round(Nφ) is therefore determined by EFCI before any circuit is run. The VQC is trained by minimizing distance to this same target (Eq. 8), so successfully outputting the precomputed dominant bitstring is a self-consistency check, not an independent energy estimate.

  2. self definitional [Section 3.2 / 4.5, Eqs. (21)-(22) and Table 2]
    "The recovered ground-state energy from the dominant QPE bitstring k ∗ is computed as Etotal =− 2π φ ∗ /τ +c 0 +E nuc ... |∆E|=|E FCI −E total|."

    Etotal is obtained by inverting φ* = k*/N, where k* is the dominant bitstring of the distribution that was constructed from EFCI in Eq. (20). Thus Etotal is just EFCI up to the rounding imposed by na and tau. Table 2's 1.786e-4 Hartree error is a discretization artifact of the input energy, not a measured prediction.

  3. other [Section 5, Conclusion (limitations paragraph)]
    "the reference energy needed to construct Pgt(k;φ) will generally not be available in practice"

    The paper itself concedes that in the regime where quantum advantage is sought, the FCI energy required to build the target is classically intractable. This confirms that the framework cannot estimate unknown molecular energies; it benchmarks a VQC against a known answer. The energy-recovery claim therefore reduces to its own input.

full rationale

The analytical replacement of QPE simulation by the Dirichlet kernel is a legitimate, non-circular contribution: Eq. (2) is the standard QPE measurement distribution, and the four-metric VQC training experiments on real hardware are genuine and could have failed. There is no load-bearing self-citation or imported uniqueness theorem. The circularity is confined to the energy-recovery framing: Pgt(k;phi) is constructed from EFCI via Eqs. (18)-(20), the VQC is trained to match that target, and the reported energy is read off from the matched distribution's dominant bitstring via Eq. (21). Consequently the headline '|Delta E| = 1.786e-4 Ha' is the quantization error of the input EFCI, not an independent estimate, and the paper's own Section 5 states that EFCI will generally be unavailable for molecules of interest. Score 7 rather than 10 because the distribution-learning results are substantive and non-circular; only the energy-estimation claim reduces by construction.

Axiom & Free-Parameter Ledger

3 free parameters · 5 axioms · 0 invented entities

The framework rests on the textbook QPE distribution, standard quantum-chemistry reductions, and an empirically asserted expressibility of a shallow VQC. No new physical entities are introduced; the main free choices are tau, n_a, and the learned VQC parameters.

free parameters (3)
  • Evolution time tau = 1.0 a.u.
    Chosen manually a priori; together with EFCI it sets the phase phi and hence the target peak position. The paper admits no principled selection procedure and that tau was fixed using knowledge of EFCI (Sections 3.2 and 5).
  • Ancilla count n_a = 11
    Design choice determining target resolution and distribution support (2048 bins); not derived from a required precision. All experiments use this value.
  • VQC parameters theta (COBYLA-optimized) = not reported
    All M variational parameters are fitted to the target distribution during training; the final values are not given, so the hardware results cannot be independently replayed.
axioms (5)
  • standard math QPE measurement distribution equals the squared Dirichlet kernel (Eqs. 2, 7, 20).
    Textbook result from Nielsen & Chuang and Cleve et al.; used to define P_gt. Accepted as correct within the standard QPE formalism.
  • domain assumption Symmetry-tapered H2 STO-3G Hamiltonian is exactly a single-qubit operator c0 I + c1 Z + c2 X preserving the FCI ground-state energy (Eq. 17).
    Relies on Bravyi et al. tapering and PySCF integrals; standard for minimal-basis H2, but not a general scalable reduction.
  • domain assumption FCI energy EFCI from PySCF is exact and available for H2/STO-3G.
    H2 in a minimal basis is classically exactly diagonalizable. For larger molecules this assumption fails, as the paper concedes in Section 5.
  • domain assumption Single-layer linear-entangler RY-CZ VQC can represent the sharply peaked Dirichlet target under hardware noise.
    No expressibility proof is given; support is purely empirical at p=1 on this one target. This is load-bearing for the surrogate claim.
  • domain assumption COBYLA with 100,000 shots per iteration converges to a distribution close to the global optimum.
    Standard optimizer assumption; no convergence analysis or parameter reproducibility is provided.

pith-pipeline@v1.3.0-alltime-deepseek · 28030 in / 14914 out tokens · 152961 ms · 2026-08-01T08:55:29.940771+00:00 · methodology

0 comments
read the original abstract

Quantum Phase Estimation (QPE) is a foundational algorithm for molecular ground-state energy estimation, but its deep circuit requirements make direct hardware execution impractical on Noisy Intermediate-Scale Quantum (NISQ) devices. We present an analytically grounded variational surrogate framework in which a shallow Variational Quantum Circuit (VQC) is trained to reproduce the QPE measurement distribution without any quantum circuit simulation. The training target is computed entirely classically via the Dirichlet kernel, evaluated directly from the Full Configuration Interaction (FCI) ground-state energy, the ancilla qubit count, and the time evolution parameter, eliminating the exponentially scaling simulation bottleneck of prior surrogate approaches. We apply this framework to the hydrogen molecule (H$_2$) with a symmetry-tapered Hamiltonian, conducting a four-stage experimental investigation on IBM Quantum hardware. Stage 1 compares linear and full entangler topologies for the $R_Y$-$R_Z$-$CZ$ ansatz, with and without XpXm Dynamical Decoupling (DD), across four distributional metrics (Hellinger distance, fidelity error, total variation distance, Jensen-Shannon divergence), identifying the linear entangler as optimal. Stage 2 varies VQC layers ($p=1$ to $5$) for the linear-entangler ansatz, identifying single-layer depth as optimal under hardware noise. Stage 3 applies this configuration to the reduced $R_Y$-$CZ$ ansatz, comparing ideal and noisy simulator-trained parameters. A supplementary noise analysis at $p \in \{8,64\}$ characterizes the depth-dependent interplay between circuit depth and DD effectiveness. The framework enables faithful QPE mimicry using a linearly scaling VQC, recovering the ground-state energy within the chemical accuracy threshold (1 kcal/mol), constituting a scalable, hardware-efficient paradigm for QPE-based molecular energy estimation on NISQ devices.

Figures

Figures reproduced from arXiv: 2607.20943 by Alok Shukla, Anurag K. S. V., Ashish Kumar Patra, Jaiganesh G, Mousumi Kundu, Ruchika Bhat, Sai Shankar P..

Figure 1
Figure 1. Figure 1: Training workflow for the analytically-trained VQC surrogate. Step 1: the H2 molecular Hamiltonian is simulated classically and the exact ground-state energy EFCI is obtained via Full Configuration Interaction. Step 2: EFCI, together with the ancilla count na and evolution time τ, is substituted directly into the analytical Dirichlet kernel to obtain the target QPE measurement distribution Pgt(k;φ) — no qu… view at source ↗
Figure 2
Figure 2. Figure 2: VQC ansatz architectures used as QPE surrogates (p = 1 layer, na = 4 qubits shown for illustration). (a) RY -RZ-CZ ansatz with the full entangler: each repeated layer applies RY and RZ rotations to every qubit followed by a CZ gate between every pair of qubits, ending in a final untied RY -RZ rotation block. (b) RY -RZ-CZ ansatz with the linear entangler, where CZ gates connect only nearest-neighbor qubits… view at source ↗
Figure 3
Figure 3. Figure 3: Convergence of the cost function L (θ ) during COBYLA optimization of the RY -RZ-CZ VQC (linear entangler, p = 1) trained on the analytical Dirichlet distribution. The cost decreases monotonically, confirming stable convergence of the variational optimization. 0 20000 40000 60000 80000 100000 Counts for Dominant Bitstring "00101010011" Dirichlet QPE Ideal Sim Noisy Sim Real (Noisy Trained) Real + DD (Noisy… view at source ↗
Figure 4
Figure 4. Figure 4: Dominant bitstring 00101010011 counts for the RY -RZ-CZ ansatz at p = 1 layer, comparing the full entangler (pink) and linear entangler (purple) across all experimental cases (Nshots = 100,000). The linear entangler consistently achieves higher counts on real hardware across all configurations, with the gap being most pronounced under real hardware execution (linear: 91,104 vs full: 58,281 for noisy-traine… view at source ↗
Figure 5
Figure 5. Figure 5: Metric heatmap (Hellinger distance, fidelity error, TVD, JSD) comparing the RY -RZ-CZ ansatz with full entangler (top, orange) and linear entangler (bottom, blue) at p = 1 layer, across all experimental cases. Lower values indicate better agreement with the Dirichlet target. The linear entangler achieves uniformly lower metric values across all hardware configurations, confirming its superior noise resilie… view at source ↗
Figure 6
Figure 6. Figure 6: Dominant bitstring 00101010011 counts for the RY -RZ-CZ ansatz with linear entangler across layers p ∈ {1,2,3,4,5}, comparing the analytical Dirichlet reference, ideal simulation, noisy simulation, real hardware with ideal￾and noisy-trained parameters, and their DD-augmented variants (Nshots = 100,000). The ideal simulation closely tracks the Dirichlet reference across all layers. On real hardware, all con… view at source ↗
Figure 7
Figure 7. Figure 7: Detailed metric heatmap (Hellinger distance, fidelity error, TVD, JSD) for the RY -RZ-CZ ansatz with linear entangler, showing results for each layer p ∈ {1,2,3,4,5} under all experimental cases. Each colored block represents a distinct experimental scenario. Lower metric values (darker color within each block) indicate better agreement with the Dirichlet target. Metrics degrade monotonically with increasi… view at source ↗
Figure 8
Figure 8. Figure 8: shows the dominant bitstring counts for the RY -CZ ansatz at p = 1 across all four non-DD experimental cases: ideal simulation, noisy simulation, real hardware with noisy-trained parameters, and real hardware with ideal-trained parameters. 0 20000 40000 60000 80000 100000 Counts Dirichlet QPE (Analytical) Ideal VQC Real VQC (Ideal Sim Trained) Noisy VQC Real VQC (Noisy Sim Trained) 98890 98933 91499 76627 … view at source ↗
Figure 9
Figure 9. Figure 9: presents the metric heatmaps for both ansatze at p = 1 with the linear entangler, evaluated across all four non-DD experimental cases. Hellinger Fidelity Error TVD JSD Metrics Ideal Noisy Real + Noisy Tr. Real + Ideal Tr. Ideal Noisy Real + Noisy Tr. Real + Ideal Tr. Experimental Cases 0.071 0.005 0.007 0.062 0.331 0.110 0.230 0.289 0.188 0.035 0.083 0.164 0.177 0.031 0.074 0.156 0.054 0.003 0.005 0.047 0.… view at source ↗
Figure 10
Figure 10. Figure 10: Distributional metric heatmap (Hellinger distance, fidelity error, TVD, JSD) for the RY -CZ ansatz with linear entangler at p = 1 layer, evaluated across ideal simulation, noisy simulation, real hardware with noisy-trained parameters, and real hardware with ideal-trained parameters. The ideal-trained real hardware configuration achieves the lowest metric values (Hellinger = 0.176, fidelity error = 0.031, … view at source ↗
Figure 11
Figure 11. Figure 11: Measured outcome probabilities P(b) (log scale) for the four real-hardware VQC configurations (RY -RZ-CZ and RY -CZ ansatze, ideal- and noisy-trained, linear entangler, p = 1) against the analytical Dirichlet reference, plotted against bit-value offset ∆(b) from the dominant peak 00101010011 over the window |∆(b)| ≤ 28. The Dirichlet distribution decays smoothly with |∆(b)|, while the hardware distributio… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

53 extracted references · 1 canonical work pages

  1. [1]

    Nielsen, M. A. & Chuang, I. L.Quantum Computation and Quantum Information(Cambridge University Press, Cambridge, UK, 2000). 2.Preskill, J. Quantum computing in the NISQ era and beyond.Quantum2, 79, DOI: 10.22331/q-2018-08-06-79 (2018). 3.Steane, A. Quantum computing.Reports on Prog. Phys.61, 117–173, DOI: 10.1088/0034-4885/61/2/002 (1998). 4.Kitaev, A. Y ...

    Pith/arXiv arXiv 2000

  2. [5]

    & Mosca, M

    Cleve, R., Ekert, A., Macchiavello, C. & Mosca, M. Quantum algorithms revisited. InProceedings of the Royal Society of London A, vol. 454, 339–354, DOI: 10.1098/rspa.1998.0164 (1998)

    arXiv 1998

  3. [6]

    Shor, P. W. Algorithms for quantum computation: discrete logarithms and factoring. InProceedings of the 35th Annual Symposium on Foundations of Computer Science, 124–134, DOI: 10.1109/SFCS.1994.365700 (IEEE, 1994)

    arXiv 1994

  4. [7]

    W., Hassidim, A

    Harrow, A. W., Hassidim, A. & Lloyd, S. Quantum algorithm for linear systems of equations.Phys. Rev. Lett.103, 150502, DOI: 10.1103/PhysRevLett.103.150502 (2009)

    doi:10.1103/physrevlett.103.150502 2009

  5. [8]

    D., Love, P

    Aspuru-Guzik, A., Dutoi, A. D., Love, P. J. & Head-Gordon, M. Simulated quantum computation of molecular energies. Science309, 1704–1707, DOI: 10.1126/science.1113479 (2005)

    doi:10.1126/science.1113479 2005

  6. [9]

    Babbush, R.et al.Low-depth quantum simulation of materials.Phys. Rev. X8, 011044, DOI: 10.1103/PhysRevX.8.011044 (2018)

    doi:10.1103/physrevx.8.011044 2018

  7. [10]

    M., Wecker, D

    Reiher, M., Wiebe, N., Svore, K. M., Wecker, D. & Troyer, M. Elucidating reaction mechanisms on quantum computers. Proc. Natl. Acad. Sci.114, 7555–7560, DOI: 10.1073/pnas.1619152114 (2017). 24/30

    doi:10.1073/pnas.1619152114 2017

  8. [11]

    & Barzen, J

    Leymann, F. & Barzen, J. The bitter truth about gate-based quantum algorithms in the NISQ era.Quantum Sci. Technol.5, 044007, DOI: 10.1088/2058-9565/abae7d (2020)

    doi:10.1088/2058-9565/abae7d 2058

  9. [12]

    Bharti, K.et al.Noisy intermediate-scale quantum algorithms.Rev. Mod. Phys.94, 015004, DOI: 10.1103/RevModPhys. 94.015004 (2022)

    doi:10.1103/revmodphys 2022

  10. [13]

    & Gambetta, J

    Temme, K., Bravyi, S. & Gambetta, J. M. Error mitigation for short-depth quantum circuits.Phys. Rev. Lett.119, 180509, DOI: 10.1103/PhysRevLett.119.180509 (2017)

    doi:10.1103/physrevlett.119.180509 2017

  11. [14]

    Endo, S., Cai, Z., Benjamin, S. C. & Yuan, X. Hybrid quantum-classical algorithms and quantum error mitigation.J. Phys. Soc. Jpn.90, 032001, DOI: 10.7566/JPSJ.90.032001 (2021)

    doi:10.7566/jpsj.90.032001 2021

  12. [15]

    & Benjamin, S

    Li, Y . & Benjamin, S. C. Efficient variational quantum simulator incorporating active error minimization.Phys. Rev. X7, 021050, DOI: 10.1103/PhysRevX.7.021050 (2017)

    doi:10.1103/physrevx.7.021050 2017

  13. [16]

    & Zeng, W

    Giurgica-Tiron, T., Hindy, Y ., LaRose, R., Mari, A. & Zeng, W. J. Digital zero noise extrapolation for quantum error mitigation.arXiv preprintarXiv:2005.10921(2020)

    Pith/arXiv arXiv 2005

  14. [17]

    & Almudever, C

    Lao, L. & Almudever, C. G. Timing and resource-efficient quantum circuit compilation.arXiv preprintarXiv:2209.07005 (2022)

    Pith/arXiv arXiv 2022

  15. [18]

    Cerezo, M.et al.Variational quantum algorithms.Nat. Rev. Phys.3, 625–644, DOI: 10.1038/s42254-021-00348-9 (2021)

    doi:10.1038/s42254-021-00348-9 2021

  16. [19]

    R., Romero, J., Babbush, R

    McClean, J. R., Romero, J., Babbush, R. & Aspuru-Guzik, A. The theory of variational hybrid quantum-classical algorithms. New J. Phys.18, 023023, DOI: 10.1088/1367-2630/18/2/023023 (2016)

    doi:10.1088/1367-2630/18/2/023023 2016

  17. [20]

    Commun.5, 4213, DOI: 10.1038/ncomms5213 (2014)

    Peruzzo, A.et al.A variational eigenvalue solver on a photonic quantum processor.Nat. Commun.5, 4213, DOI: 10.1038/ncomms5213 (2014)

    doi:10.1038/ncomms5213 2014

  18. [21]

    Kandala, A.et al.Hardware-efficient variational quantum eigensolver for small molecules and quantum magnets.Nature 549, 242–246, DOI: 10.1038/nature23879 (2017)

    doi:10.1038/nature23879 2017

  19. [22]

    Reports986, 1–128, DOI: 10.1016/j.physrep.2022.08.003 (2022)

    Tilly, J.et al.The variational quantum eigensolver: A review of methods and best practices.Phys. Reports986, 1–128, DOI: 10.1016/j.physrep.2022.08.003 (2022)

    doi:10.1016/j.physrep.2022.08.003 2022

  20. [23]

    & Lin, C.-H

    Liu, C.-Y ., Chen, K.-C. & Lin, C.-H. A. Learning quantum phase estimation by variational quantum circuits. In2024 International Joint Conference on Neural Networks (IJCNN), 1–8, DOI: 10.1109/IJCNN60899.2024.10651206 (2024)

    arXiv 2024

  21. [24]

    M., Mezzacapo, A

    Bravyi, S., Gambetta, J. M., Mezzacapo, A. & Temme, K. Tapering off qubits to simulate fermionic Hamiltonians.arXiv preprintarXiv:1701.08213(2017)

    Pith/arXiv arXiv 2017

  22. [25]

    T., Richard, M

    Seeley, J. T., Richard, M. J. & Love, P. J. The Bravyi-Kitaev transformation for quantum computation of electronic structure.The J. Chem. Phys.137, 224109, DOI: 10.1063/1.4768229 (2012)

    doi:10.1063/1.4768229 2012

  23. [26]

    Neue Begründung der Theorie quadratischer Formen von unendlichvielen Veränderlichen.J

    Hellinger, E. Neue Begründung der Theorie quadratischer Formen von unendlichvielen Veränderlichen.J. für die reine und angewandte Math.136, 210–271 (1909)

    1909

  24. [27]

    Endres, D. M. & Schindelin, J. E. A new metric for probability distributions.IEEE Transactions on Inf. Theory49, 1858–1860, DOI: 10.1109/TIT.2003.813506 (2003)

    arXiv 2003

  25. [28]

    & Lloyd, S

    Viola, L. & Lloyd, S. Dynamical suppression of decoherence in two-state quantum systems.Phys. Rev. A58, 2733–2744, DOI: 10.1103/PhysRevA.58.2733 (1998)

    doi:10.1103/physreva.58.2733 1998

  26. [29]

    & Knill, E

    Viola, L., Lloyd, S. & Knill, E. Universal control of decoupled quantum systems.Phys. Rev. Lett.83, 4888–4891, DOI: 10.1103/PhysRevLett.83.4888 (1999). 30.Zanardi, P. Symmetrizing evolutions.Phys. Lett. A258, 77–82, DOI: 10.1016/S0375-9601(99)00365-5 (1999)

    work page doi:10.1103/physrevlett.83.4888 1999

  27. [31]

    & Ekert, A

    Mosca, M. & Ekert, A. The hidden subgroup problem and eigenvalue estimation on a quantum computer. InQuantum Computing and Quantum Communications, vol. 1509 ofLecture Notes in Computer Science, 174–188, DOI: 10.1007/ 3-540-49208-9_15 (Springer, 1999)

    1999

  28. [32]

    K., Hastings, M

    Wecker, D., Bauer, B., Clark, B. K., Hastings, M. B. & Troyer, M. Progress towards practical quantum advantage in quantum chemistry.Phys. Rev. A92, 062318, DOI: 10.1103/PhysRevA.92.062318 (2015)

    doi:10.1103/physreva.92.062318 2015

  29. [33]

    McArdle, S., Endo, S., Aspuru-Guzik, A., Benjamin, S. C. & Yuan, X. Quantum computational chemistry.Rev. Mod. Phys. 92, 015003, DOI: 10.1103/RevModPhys.92.015003 (2020)

    doi:10.1103/revmodphys.92.015003 2020

  30. [34]

    W.et al.Qubitization of arbitrary basis quantum chemistry leveraging sparsity and low rank factorization

    Berry, D. W.et al.Qubitization of arbitrary basis quantum chemistry leveraging sparsity and low rank factorization. Quantum3, 208, DOI: 10.22331/q-2019-12-02-208 (2019). 25/30

    doi:10.22331/q-2019-12-02-208 2019

  31. [35]

    & Tong, Y

    Lin, S.-H. & Tong, Y . Heisenberg-limited ground-state energy estimation for early fault-tolerant quantum computers via rejection sampling.PRX Quantum3, 010318, DOI: 10.1103/PRXQuantum.3.010318 (2022)

    doi:10.1103/prxquantum.3.010318 2022

  32. [36]

    Kimmel, S., Low, G. H. & Yoder, T. J. Robust calibration of a universal single-qubit gate set via robust phase estimation. Phys. Rev. A92, 062315, DOI: 10.1103/PhysRevA.92.062315 (2015)

    doi:10.1103/physreva.92.062315 2015

  33. [37]

    & Granade, C

    Wiebe, N. & Granade, C. Efficient Bayesian phase estimation.Phys. Rev. Lett.117, 010503, DOI: 10.1103/PhysRevLett. 117.010503 (2016). 38.Helgaker, T., Jørgensen, P. & Olsen, J.Molecular Electronic-Structure Theory(Wiley, Chichester, UK, 2000)

    doi:10.1103/physrevlett 2016

  34. [39]

    Sim, S., Johnson, P. D. & Aspuru-Guzik, A. Expressibility and entangling capability of parameterized quantum circuits for hybrid quantum-classical algorithms.Adv. Quantum Technol.2, 1900070, DOI: 10.1002/qute.201900070 (2019)

    doi:10.1002/qute.201900070 2019

  35. [40]

    R., Boixo, S., Smelyanskiy, V

    McClean, J. R., Boixo, S., Smelyanskiy, V . N., Babbush, R. & Neven, H. Barren plateaus in quantum neural network training landscapes.Nat. Commun.9, 4812, DOI: 10.1038/s41467-018-07090-4 (2018)

    doi:10.1038/s41467-018-07090-4 2018

  36. [41]

    & Coles, P

    Cerezo, M., Sone, A., V olkoff, T., Cincio, L. & Coles, P. J. Cost function dependent barren plateaus in shallow parametrized quantum circuits.Nat. Commun.12, 1791, DOI: 10.1038/s41467-021-21728-w (2021). 42.Biamonte, J.et al.Quantum machine learning.Nature549, 195–202, DOI: 10.1038/nature23474 (2017)

    doi:10.1038/s41467-021-21728-w 2021

  37. [43]

    & Fiorentini, M

    Benedetti, M., Lloyd, E., Sack, S. & Fiorentini, M. Parameterized quantum circuits as machine learning models.Quantum Sci. Technol.4, 043001, DOI: 10.1088/2058-9565/ab4eb5 (2019)

    doi:10.1088/2058-9565/ab4eb5 2058

  38. [44]

    Y .-C.et al.Variational quantum circuits for deep reinforcement learning.IEEE Access8, 141007–141024, DOI: 10.1109/ACCESS.2020.3010470 (2020)

    Chen, S. Y .-C.et al.Variational quantum circuits for deep reinforcement learning.IEEE Access8, 141007–141024, DOI: 10.1109/ACCESS.2020.3010470 (2020)

    arXiv 2020

  39. [45]

    Powell, M. J. D. A direct search optimization method that models the objective and constraint functions by linear interpolation.Adv. Optim. Numer. Analysis51–67, DOI: 10.1007/978-94-015-8330-5_4 (1994)

    doi:10.1007/978-94-015-8330-5_4 1994

  40. [46]

    & Fujii, K

    Mitarai, K., Negoro, M., Kitagawa, M. & Fujii, K. Quantum circuit learning.Phys. Rev. A98, 032309, DOI: 10.1103/ PhysRevA.98.032309 (2018)

    2018

  41. [47]

    & Killoran, N

    Schuld, M., Bergholm, V ., Gogolin, C., Izaac, J. & Killoran, N. Evaluating analytic gradients on quantum hardware.Phys. Rev. A99, 032331, DOI: 10.1103/PhysRevA.99.032331 (2019)

    doi:10.1103/physreva.99.032331 2019

  42. [48]

    Bhattacharyya, A. K. On a measure of divergence between two statistical populations defined by their probability distributions.Bull. Calcutta Math. Soc.35, 99–109 (1943). 49.Jozsa, R. Fidelity for mixed quantum states.J. Mod. Opt.41, 2315–2323, DOI: 10.1080/09500349414552171 (1994)

    doi:10.1080/09500349414552171 1943

  43. [50]

    Levin, D. A. & Peres, Y .Markov Chains and Mixing Times(American Mathematical Society, Providence, RI, 2017), 2nd edn. 51.Tsybakov, A. B.Introduction to Nonparametric Estimation(Springer, New York, 2009)

    2017

  44. [52]

    & Leibler, R

    Kullback, S. & Leibler, R. A. On information and sufficiency.Annals Math. Stat.22, 79–86, DOI: 10.1214/aoms/ 1177729694 (1951)

    doi:10.1214/aoms/ 1951

  45. [53]

    Anurag, K. S. V .et al.Resource estimation for vqe on small molecules: Impact of fermion mappings and hamiltonian reductions.J. Comput. Chem.47, e70379, DOI: https://doi.org/10.1002/jcc.70379 (2026)

    doi:10.1002/jcc.70379 2026

  46. [54]

    & Wigner, E

    Jordan, P. & Wigner, E. P. Über das Paulische äquivalenzverbot.Zeitschrift für Physik47, 631–651, DOI: 10.1007/ BF01331938 (1928). 55.Hahn, E. L. Spin echoes.Phys. Rev.80, 580–594, DOI: 10.1103/PhysRev.80.580 (1950)

    doi:10.1103/physrev.80.580 1928

  47. [56]

    Carr, H. Y . & Purcell, E. M. Effects of diffusion on free precession in nuclear magnetic resonance experiments.Phys. Rev. 94, 630–638, DOI: 10.1103/PhysRev.94.630 (1954). 57.Qiskit contributors. Qiskit: An open-source framework for quantum computing, DOI: 10.5281/zenodo.2573505 (2023)

    doi:10.1103/physrev.94.630 1954

  48. [58]

    & Lidar, D

    Pokharel, B., Anand, N., Fortman, B. & Lidar, D. A. Demonstration of fidelity improvement using dynamical decoupling with superconducting qubits.Phys. Rev. Lett.121, 220502, DOI: 10.1103/PhysRevLett.121.220502 (2018)

    doi:10.1103/physrevlett.121.220502 2018

  49. [59]

    J., Stewart, R

    Hehre, W. J., Stewart, R. F. & Pople, J. A. Self-consistent molecular-orbital methods. I. use of Gaussian expansions of Slater-type atomic orbitals.The J. Chem. Phys.51, 2657–2664, DOI: 10.1063/1.1672392 (1969)

    doi:10.1063/1.1672392 1969

  50. [60]

    Sun, Q.et al.PySCF: the Python-based simulations of chemistry framework.Wiley Interdiscip. Rev. Comput. Mol. Sci.8, e1340, DOI: 10.1002/wcms.1340 (2018). 61.Harris, C. R.et al.Array programming with NumPy.Nature585, 357–362, DOI: 10.1038/s41586-020-2649-2 (2020). 26/30

    doi:10.1002/wcms.1340 2018

  51. [62]

    Methods17, 261–272, DOI: 10.1038/s41592-019-0686-2 (2020)

    Virtanen, P.et al.SciPy 1.0: Fundamental algorithms for scientific computing in Python.Nat. Methods17, 261–272, DOI: 10.1038/s41592-019-0686-2 (2020)

    doi:10.1038/s41592-019-0686-2 2020

  52. [63]

    Hunter, J. D. Matplotlib: A 2D graphics environment.Comput. Sci. & Eng.9, 90–95, DOI: 10.1109/MCSE.2007.55 (2007)

    doi:10.1109/mcse.2007.55 2007

  53. [64]

    Bravyi, S., Sheldon, S., Kandala, A., Mckay, D. C. & Gambetta, J. M. Mitigating measurement errors in multiqubit experiments.Phys. Rev. A103, 042605, DOI: 10.1103/PhysRevA.103.042605 (2021). 27/30 Supplementary Information A Noise Analysis of VQC Circuit Depth and Dynamical Decoupling As a supplementary investigation beyond the main results of this work, ...

    doi:10.1103/physreva.103.042605 2021