pith. sign in

arxiv: 2606.24922 · v1 · pith:4KI6WWJ5new · submitted 2026-06-20 · 🪐 quant-ph · cond-mat.mtrl-sci

Constraint-Aware Quantum Optimization of Defect Configurations in Doped ZrO2: XY-Mixer QAOA and Grover Adaptive Search

Huajing Song This is my paper

Pith reviewed 2026-06-26 11:56 UTC · model grok-4.3

classification 🪐 quant-ph cond-mat.mtrl-sci
keywords quantum optimizationQAOAGrover Adaptive Searchdefect configurationsZrO2QUBOconstraint preservationmaterials design
0
0 comments X

The pith

Constraint-preserving XY-mixer QAOA confines sampling to feasible defect configurations in doped ZrO2 and achieves 86% probability on near-optima at depth p=3.

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

The paper constructs a 24-variable QUBO model from machine-learned energies for doped ZrO2 that enforces exactly two rare-earth cation substitutions and one oxygen vacancy. It then applies a constraint-preserving XY-mixer QAOA that never leaves the 448 allowed configurations and places 86% of its probability mass within 1 meV of the best energy already at circuit depth 3. A second pathway builds a fault-tolerant Grover Adaptive Search oracle with built-in feasibility checks, using 324 logical qubits and 36,000 to 43,000 Toffoli gates per iteration. Resource estimates indicate that restricting the search to the feasible subspace can cut the total Toffoli cost by as much as 240 times compared with searching the full occupation space.

Core claim

The central claim is that the XY-mixer QAOA, by construction, confines all sampling to the feasible subspace of the 24-variable QUBO and places 86% of the probability mass within 1 meV of the MACE optimum at depth p = 3, while the constrained Grover Adaptive Search provides explicit resource counts of 324 high-level logical qubits and 3.6 to 4.3 times 10 to the 4 Toffoli gates per iteration, with an idealized amplification factor suggesting up to 240 times Toffoli reduction.

What carries the argument

The constraint-preserving XY-mixer Hamiltonian in QAOA, which generates transitions only between feasible bit strings that satisfy the exact two-substitution and one-vacancy rules.

If this is right

  • The QUBO surrogate reproduces held-out MACE energies with R-squared of 0.997.
  • QAOA at depth 3 recovers near-optimal configurations without needing post-selection or penalty terms.
  • The Grover oracle implements feasibility checking, fixed-point arithmetic, and phase kickback within the stated gate counts.
  • Feasible-space amplification yields up to 240 times reduction in total Toffoli cost relative to the full 2 to the 24 space.

Where Pith is reading between the lines

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

  • The same mixer construction could be applied to other materials problems with similar cardinality constraints on defect numbers.
  • Resource estimates assume ideal fault-tolerant hardware; real-device noise would require additional error-correction overhead not accounted for here.
  • Direct experimental validation of the lowest-energy configurations predicted by the QUBO would test whether the surrogate landscape matches physical behavior.
  • Hybrid workflows combining the QAOA warm-start with classical refinement might further reduce quantum resource demands.

Load-bearing premise

The 24-variable QUBO fitted to the MACE-MPA-0 dataset serves as a faithful surrogate for the true energy landscape of the constrained defect configurations.

What would settle it

Running exact diagonalization or additional first-principles calculations on configurations outside the training set and checking whether the QUBO ranking of energies matches the true ranking.

Figures

Figures reproduced from arXiv: 2606.24922 by Huajing Song.

Figure 1
Figure 1. Figure 1: End-to-end constraint-aware workflow. Starting from a ZrO [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Binary occupation encoding for the fixed-composition ZrO [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: QUBO-predicted versus MACE single-point energies for all 448 feasible configurations. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Ranked corrected-QUBO energies over the 448 feasible configurations, annotated with [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Penalty QAOA feasible probability as a function of penalty strength. Feasibility is low [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Constraint-preserving XY-mixer QAOA: probability mass on useful low-energy regions as [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: First-order noise survival factor η versus QAOA depth under several two-qubit-fidelity profiles. The two-qubit gate count dominates the error budget, producing a depth–noise trade-off between ideal concentration and noisy survival [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Construction of the constrained GAS phase oracle from occupation variables: term-flag [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Concrete full-space Grover/GAS baseline versus the idealized feasible-space amplification [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗
read the original abstract

Quantum optimization offers a route to searching the large defect-configuration spaces that arise in materials design. We develop an end-to-end, constraint-aware quantum optimization workflow for composition-defect search in a doped ZrO2 thermal-barrier-coating (TBC) material system, using a MACE-MPA-0 energy dataset to fit a 24-variable QUBO over 8 cation-occupation and 16 oxygen-vacancy variables with exactly two rare-earth substitutions and one oxygen vacancy, yielding 448 feasible configurations. The QUBO surrogate reproduces the MACE energies with held-out R2 = 0.997 (full-data R2 = 0.999, RMSE = 17 meV). We validate two complementary quantum pathways against exact enumeration: a constraint-preserving XY-mixer QAOA that confines sampling to the feasible subspace and places 86% of probability mass within 1 meV of the MACE optimum at depth p = 3, and a fault-tolerant constrained Grover Adaptive Search oracle with explicit fixed-point arithmetic, branch-safe comparison, feasibility checking, and phase kickback. Across threshold cases, the validated oracle uses 324 high-level logical qubits, or 352 to 358 with conservative clean-ancilla v-chain accounting, and requires 3.6 to 4.3 x 104 Toffoli gates per Grover (GAS) iteration. An idealized feasible-space amplification estimate suggests up to a 240x reduction in total Toffoli cost relative to the full 224 occupation space, providing a resource-estimation bridge between materials-informed QUBO modeling, constraint-aware QAOA, and fault-tolerant threshold search.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript develops an end-to-end constraint-aware quantum optimization workflow for defect-configuration search in doped ZrO2, fitting a 24-variable QUBO (8 cation + 16 oxygen-vacancy variables, exactly two rare-earth substitutions and one vacancy) to MACE-MPA-0 energies over the 448 feasible configurations. It reports held-out R²=0.997 for the surrogate and validates two quantum algorithms against exact enumeration: an XY-mixer QAOA that confines sampling to the feasible subspace and places 86% probability mass within 1 meV of the MACE optimum at depth p=3, plus a fault-tolerant constrained Grover Adaptive Search oracle using 324 logical qubits and 3.6–4.3×10⁴ Toffoli gates per iteration, with an idealized 240× Toffoli reduction from feasible-space amplification.

Significance. If the QUBO surrogate is shown to be faithful for optimization (not merely regression), the work supplies a concrete materials-to-quantum pipeline with explicit resource estimates that could inform near-term and fault-tolerant algorithm design for constrained combinatorial search in materials. The exact-enumeration validation of both QAOA and GAS, together with the constraint-preserving mixer and oracle construction, are positive features that strengthen the algorithmic claims.

major comments (2)
  1. [Abstract / QUBO construction] Abstract and QUBO construction paragraph: the central claim that QAOA performance on the QUBO informs real defect energies rests on the surrogate being faithful in the low-energy tail. The reported held-out R²=0.997 is a global regression metric; no explicit check is provided that the QUBO minimum coincides with the MACE minimum or that the lowest-energy configurations are correctly ranked. Because QAOA samples are post-evaluated on MACE, any systematic bias in the low-energy region would render the 86% probability-mass figure an artifact of the fit rather than evidence of materials-relevant performance.
  2. [GAS resource estimation] GAS resource paragraph: the Toffoli counts (3.6–4.3×10⁴ per iteration) and qubit counts (324 logical, 352–358 with v-chain) are presented as validated, yet the manuscript does not supply the explicit circuit decomposition or the precise fixed-point arithmetic and branch-safe comparison subroutines used to obtain these numbers. Without those details or a reference to a machine-checked implementation, the 240× reduction estimate cannot be independently verified.
minor comments (2)
  1. [Abstract] The abstract states “full-data R²=0.999, RMSE=17 meV” but does not specify the train/test split sizes or whether the split respects the constraint structure; this should be stated explicitly.
  2. [QUBO construction] Notation for the 24-variable QUBO (8 cation + 16 vacancy) is introduced without an equation defining the precise mapping from occupation variables to the QUBO coefficients; adding this equation would improve reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback and positive evaluation of the manuscript's contributions. We address each major comment below with targeted revisions to strengthen the claims.

read point-by-point responses

  1. Referee: [Abstract / QUBO construction] Abstract and QUBO construction paragraph: the central claim that QAOA performance on the QUBO informs real defect energies rests on the surrogate being faithful in the low-energy tail. The reported held-out R²=0.997 is a global regression metric; no explicit check is provided that the QUBO minimum coincides with the MACE minimum or that the lowest-energy configurations are correctly ranked. Because QAOA samples are post-evaluated on MACE, any systematic bias in the low-energy region would render the 86% probability-mass figure an artifact of the fit rather than evidence of materials-relevant performance.

    Authors: We agree that an explicit verification of the low-energy tail is necessary to support the claim. Since the feasible space is fully enumerated (448 configurations), we will add a direct comparison in the revised manuscript (new table or figure) confirming that the QUBO minimum matches the MACE minimum and that the lowest-energy configurations are correctly ranked by the surrogate. This addresses the potential for bias in the relevant region. revision: yes

  2. Referee: [GAS resource estimation] GAS resource paragraph: the Toffoli counts (3.6–4.3×10⁴ per iteration) and qubit counts (324 logical, 352–358 with v-chain) are presented as validated, yet the manuscript does not supply the explicit circuit decomposition or the precise fixed-point arithmetic and branch-safe comparison subroutines used to obtain these numbers. Without those details or a reference to a machine-checked implementation, the 240× reduction estimate cannot be independently verified.

    Authors: The manuscript describes the oracle components (explicit fixed-point arithmetic, branch-safe comparison, feasibility checking, and phase kickback), but we acknowledge that full circuit decompositions are not provided. In revision we will add detailed pseudocode and gate-level breakdowns of the key subroutines to the supplementary material, enabling independent verification of the reported Toffoli and qubit counts. revision: yes

Circularity Check

0 steps flagged

No circularity; algorithmic claims validated by exact enumeration on external MACE energies

full rationale

The paper constructs a QUBO surrogate by fitting to the external MACE-MPA-0 dataset and reports regression metrics (held-out R2=0.997). It then runs QAOA and GAS on the QUBO Hamiltonian but validates performance by exact enumeration over all 448 feasible configurations, comparing sampled bitstrings directly to MACE-computed energies and the MACE optimum. This external benchmark (MACE energies, full enumeration) is independent of the QUBO fit and of any self-citation chain. No step reduces a claimed prediction or uniqueness result to a fitted parameter or prior self-citation by construction. The derivation chain for the quantum algorithms and resource estimates is therefore self-contained against the stated external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The workflow depends on the MACE potential being an accurate proxy and on the chosen constraints correctly capturing physical feasibility; the QUBO fit introduces fitted coefficients whose accuracy is asserted by R2 but not independently derived.

free parameters (1)
  • QUBO coefficients for 24 variables
    Fitted to reproduce MACE-MPA-0 energies for cation occupations and oxygen vacancies
axioms (2)
  • domain assumption MACE-MPA-0 machine-learning potential yields reliable energies for the doped ZrO2 defect system
    Used as ground truth to fit the QUBO surrogate
  • domain assumption Exactly two rare-earth substitutions and one oxygen vacancy define the physically relevant feasible configurations
    Reduces search space to 448 configurations

pith-pipeline@v0.9.1-grok · 5839 in / 1542 out tokens · 38512 ms · 2026-06-26T11:56:12.303778+00:00 · methodology

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

25 extracted references · 16 canonical work pages · 3 internal anchors

  1. [1]

    Padture, Maurice Gell, and Eric H

    Nitin P. Padture, Maurice Gell, and Eric H. Jordan. Thermal barrier coatings for gas-turbine engine applications.Science, 296(5566):280–284, 2002. doi: 10.1126/science.1068609

    work page doi:10.1126/science.1068609 2002

  2. [2]

    Effective Field Theory.Annu

    David R. Clarke and Carlos G. Levi. Materials design for the next generation thermal barrier coatings.Annual Review of Materials Research, 33:383–417, 2003. doi: 10.1146/annurev. matsci.33.011403.113718

    work page doi:10.1146/annurev 2003

  3. [3]

    Padture, Paul G

    Jie Wu, Xuezheng Wei, Nitin P. Padture, Paul G. Klemens, Maurice Gell, Eugenio García, Pilar Miranzo, and Maria I. Osendi. Low-thermal-conductivity rare-earth zirconates for potential thermal-barrier-coating applications.Journal of the American Ceramic Society, 85(12):3031– 3035, 2002. doi: 10.1111/j.1151-2916.2002.tb00574.x

    work page doi:10.1111/j.1151-2916.2002.tb00574.x 2002

  4. [4]

    A Computational Study of Yttria-Stabilized Zirconia: I. Using Crystal Chemistry to Search for the Ground State on a Glassy Energy Landscape

    Yanhao Dong, Liang Qi, Ju Li, and I-Wei Chen. A computational study of yttria-stabilized zirconia: I. using crystal chemistry to search for the ground state on a glassy energy landscape. Acta Materialia, 127:73–84, 2017. doi: 10.1016/j.actamat.2017.01.006. arXiv:1701.05089

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1016/j.actamat.2017.01.006 2017

  5. [5]

    Kovács, Gregor N

    Ilyes Batatia, Dávid P. Kovács, Gregor N. C. Simm, Christoph Ortner, and Gábor Csányi. MACE: Higher order equivariant message passing neural networks for fast and accurate force fields. InAdvances in Neural Information Processing Systems (NeurIPS), 2022. arXiv:2206.07697

    arXiv 2022

  6. [6]

    A foundation model for atomistic materials chemistry

    Ilyes Batatia, Philipp Benner, Yuan Chiang, Alin M. Elena, Dávid P. Kovács, Janosh Riebesell, et al. A foundation model for atomistic materials chemistry.The Journal of Chemical Physics, 163(18):184110, 2025. doi: 10.1063/5.0297006. MACE-MPA-0; arXiv:2401.00096

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1063/5.0297006 2025

  7. [7]

    Ising formulations of many NP problems

    Andrew Lucas. Ising formulations of many np problems.Frontiers in Physics, 2:5, 2014. doi: 10.3389/fphy.2014.00005

    work page doi:10.3389/fphy.2014.00005 2014

  8. [8]

    July 2022

    Fred Glover, Gary Kochenberger, Rick Hennig, and Yu Du. Quantum bridge analytics I: A tutorial on formulating and using QUBO models.Annals of Operations Research, 314(1):141– 183, 2022. doi: 10.1007/s10479-022-04634-2. Extended version of arXiv:1811.11538

    work page doi:10.1007/s10479-022-04634-2 2022

  9. [9]

    Effect of anisotropy on small magnetic clusters.Phys

    Alfred Hucht, Sanjubala Sahoo, Shreekantha Sil, and Peter Entel. Effect of anisotropy on small magnetic clusters.Phys. Rev. B, 84:104438, 2011. doi: 10.1103/PhysRevB.84.104438

    work page doi:10.1103/physrevb.84.104438 2011

  10. [10]

    A quantum approximate optimization algorithm.arXiv preprint, 2014

    Edward Farhi, Jeffrey Goldstone, and Sam Gutmann. A quantum approximate optimization algorithm.arXiv preprint, 2014. arXiv:1411.4028

    Pith/arXiv arXiv 2014

  11. [11]

    Grover adaptive search for constrained polynomial binary optimization.Quantum, 5:428, 2021

    Austin Gilliam, Stefan Woerner, and Constantin Gonciulea. Grover adaptive search for constrained polynomial binary optimization.Quantum, 5:428, 2021. doi: 10.22331/ q-2021-04-08-428. arXiv:1912.04088

    arXiv 2021

  12. [12]

    Algorithms 12(2):34, ://dx.doi.org/10.3390/a12020034

    Stuart Hadfield, Zhihui Wang, Bryan O’Gorman, Eleanor G. Rieffel, Davide Venturelli, and Ru- pak Biswas. From the quantum approximate optimization algorithm to a quantum alternating operator ansatz.Algorithms, 12(2):34, 2019. doi: 10.3390/a12020034

    work page doi:10.3390/a12020034 2019

  13. [13]

    Fuchs, Kjetil Olsen Lye, Halvor Møll Nilsen, Alexander J

    Franz G. Fuchs, Kjetil Olsen Lye, Halvor Møll Nilsen, Alexander J. Stasik, and Giorgio Sartor. Constraintpreservingmixersforthequantumapproximateoptimizationalgorithm.Algorithms, 15(6):202, 2022. doi: 10.3390/a15060202. arXiv:2203.06095. 17

    work page doi:10.3390/a15060202 2022

  14. [14]

    Generalizedclusterdescriptionofmulticomponent systems.Physica A: Statistical Mechanics and its Applications, 128(1):334–350, 1984

    J.M.Sanchez, F.Ducastelle, andD.Gratias. Generalizedclusterdescriptionofmulticomponent systems.Physica A: Statistical Mechanics and its Applications, 128(1):334–350, 1984. doi: 10.1016/0378-4371(84)90096-7

    work page doi:10.1016/0378-4371(84)90096-7 1984

  15. [15]

    Automatingfirst-principlesphasediagramcalculations

    AxelvandeWalleandGerbrandCeder. Automatingfirst-principlesphasediagramcalculations. Journal of Phase Equilibria, 23:348–359, 2002. doi: 10.1361/105497102770331596

    work page doi:10.1361/105497102770331596 2002

  16. [16]

    Ridge Regression: Appli- cations to Nonorthogonal Problems

    Arthur E. Hoerl and Robert W. Kennard. Ridge regression: Biased estimation for nonorthog- onal problems.Technometrics, 12(1):55–67, 1970. doi: 10.1080/00401706.1970.10488634

    work page doi:10.1080/00401706.1970.10488634 1970

  17. [17]

    Lov K. Grover. A fast quantum mechanical algorithm for database search. InProceedings of the 28th Annual ACM Symposium on Theory of Computing (STOC), pages 212–219, 1996. doi: 10.1145/237814.237866

    work page doi:10.1145/237814.237866 1996

  18. [18]

    Tight bounds on quantum search- ing.Fortschritte der Physik, 46(4-5):493–505, 1998

    Michel Boyer, Gilles Brassard, Peter Høyer, and Alain Tapp. Tight bounds on quantum search- ing.Fortschritte der Physik, 46(4-5):493–505, 1998. arXiv:quant-ph/9605034

    Pith/arXiv arXiv 1998

  19. [19]

    A quantum algorithm for finding the minimum.arXiv preprint, 1996

    Christoph Dürr and Peter Høyer. A quantum algorithm for finding the minimum.arXiv preprint, 1996. arXiv:quant-ph/9607014

    Pith/arXiv arXiv 1996

  20. [20]

    Quantum Networks for Elementary Arithmetic Operations

    Vlatko Vedral, Adriano Barenco, and Artur Ekert. Quantum networks for elementary arith- metic operations.Physical Review A, 54(1):147–153, 1996. doi: 10.1103/PhysRevA.54.147. arXiv:quant-ph/9511018

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1103/physreva.54.147 1996

  21. [21]

    Cuccaro, Thomas G

    Steven A. Cuccaro, Thomas G. Draper, Samuel A. Kutin, and David Petrie Moulton. A new quantum ripple-carry addition circuit.arXiv preprint, 2004. arXiv:quant-ph/0410184

    Pith/arXiv arXiv 2004

  22. [22]

    Bennett, Richard Cleve, David P

    Adriano Barenco, Charles H. Bennett, Richard Cleve, David P. DiVincenzo, Norman Margolus, Peter Shor, Tycho Sleator, John A. Smolin, and Harald Weinfurter. Elementary gates for quantum computation.Physical Review A, 52(5):3457–3467, 1995. doi: 10.1103/PhysRevA.52

    work page doi:10.1103/physreva.52 1995

  23. [23]

    arXiv:quant-ph/9503016

    Pith/arXiv arXiv

  24. [24]

    Nielsen and Isaac L

    Michael A. Nielsen and Isaac L. Chuang.Quantum Computation and Quantum Information. Cambridge University Press, 10th anniversary edition, 2010

    2010

  25. [25]

    Fixed-point grover adaptive search for quadratic binary optimization problems.IEEE Transactions on Quantum Engineering, 5, 2024

    Ákos Nagy, Jaime Park, Cindy Zhang, Atithi Acharya, and Alex Khan. Fixed-point grover adaptive search for quadratic binary optimization problems.IEEE Transactions on Quantum Engineering, 5, 2024. arXiv:2311.05592. 18

    arXiv 2024