pith. sign in

arxiv: 2606.04771 · v1 · pith:KQDR4TFXnew · submitted 2026-06-03 · 🪐 quant-ph · cond-mat.str-el

Pushing the Classical Frontier of 1D Fermi-Hubbard Quench Dynamics Beyond Current Quantum Simulations

Roman Rausch , Sukhbinder Singh , Saeed S. Jahromi , Augustine Kshetrimayum , Roman Orus This is my paper

Pith reviewed 2026-06-28 06:14 UTC · model grok-4.3

classification 🪐 quant-ph cond-mat.str-el
keywords Fermi-Hubbard modelquench dynamicsTDVPtensor networksclassical simulationGPU accelerationquantum advantageU(1) SU(2) symmetry
0
0 comments X

The pith

Classical TDVP simulations converge 1D Fermi-Hubbard quench dynamics to t=7 at bond dimensions up to 62000

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

The paper shows that classical simulations of quench dynamics in the one-dimensional Fermi-Hubbard model can reach full convergence across the full time window up to t=7 by combining the complete U(1)×SU(2) symmetry with GPU-accelerated tensor contractions at bond dimensions up to 62000. This includes rigorous verification of the high-entanglement regime between t=5.2 and 6 that earlier classical runs left unresolved. At bond dimensions matching the prior classical benchmark the GPU run finishes in roughly 100 minutes, which lowers the claimed quantum speedup from 3000 times to about 36 times. A sympathetic reader cares because the work supplies a new, higher classical reference point against which quantum hardware performance must be measured.

Core claim

Exploiting the full U(1)×SU(2) symmetry of the Fermi-Hubbard Hamiltonian combined with GPU-accelerated tensor contractions, we reach bond dimensions up to χ≈62,000 on four NVIDIA H200 GPUs to achieve fully converged results for the quench dynamics across the entire simulation window, including rigorous certification of the high-entanglement regime t∈[5.2,6], and advance the classical frontier to t=7.

What carries the argument

TDVP algorithm with full U(1)×SU(2) symmetry reduction and GPU tensor contractions that enable bond dimensions up to 62000

If this is right

  • Fully converged results are now available for the high-entanglement regime t∈[5.2,6].
  • Simulation at bond dimension comparable to the earlier classical benchmark completes in approximately 100 minutes on four GPUs.
  • The claimed 3000× quantum advantage is reduced to roughly 36×.
  • The verified classical window is extended to t=7, past the quantum hardware experiment.

Where Pith is reading between the lines

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

  • Symmetry reduction combined with GPU scaling may allow classical verification of still later times or modestly larger lattices in the same model.
  • Ongoing classical advances of this kind require quantum advantage claims to be re-benchmarked against the current best classical reference rather than against older runs.
  • The same symmetry-exploiting TDVP strategy could be tested on related Hubbard-like models to check whether the performance gain generalizes.

Load-bearing premise

The TDVP truncation at the reported bond dimensions together with the symmetry reduction produces results converged to the exact dynamics within the error tolerance needed to certify the high-entanglement regime.

What would settle it

A run at bond dimension substantially larger than 62000 that produces observables or entanglement values in t∈[5.2,6] differing beyond the reported error bars would show the claimed convergence is incomplete.

Figures

Figures reproduced from arXiv: 2606.04771 by Augustine Kshetrimayum, Roman Orus, Roman Rausch, Saeed S. Jahromi, Sukhbinder Singh.

Figure 1
Figure 1. Figure 1: Main results. (a) TDVP time evolution of [PITH_FULL_IMAGE:figures/full_fig_p015_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Classical benchmarks for the Q-CTRL quantum simulation results of Fig. 3 in [PITH_FULL_IMAGE:figures/full_fig_p016_2.png] view at source ↗
read the original abstract

Establishing quantum advantage requires comparison against the best achievable classical simulation. The Q-CTRL team recently simulated quench dynamics of the one-dimensional Fermi-Hubbard model on an IBM processor, completing a $L=60$ evolution to time $t=6$ in under three minutes and claiming a $3000\times$ speedup over classical Time-Dependent Variational Principle (TDVP) simulation at bond dimension $\chi=4096$. Their classical benchmark required over 160 hours on a CPU cluster, failed to converge in the high-entanglement regime $t\in[5.2,6]$, and left the most challenging window of the experiment unverified. Here, we push the boundaries of classical simulation by exploiting the full $\mathrm{U}(1)\times\mathrm{SU}(2)$ symmetry of the Fermi-Hubbard Hamiltonian combined with GPU-accelerated tensor contractions. Reaching bond dimensions up to $\chi\approx62{,}000$ on four NVIDIA H200 GPUs -- among the largest ever achieved in TDVP simulations and fifteen times larger than Q-CTRL's classical baseline -- we achieve fully converged results across the entire simulation window, including rigorous certification of the previously unresolved high-entanglement regime $t\in[5.2,6]$. We further advance the classical frontier to $t=7$, which lies beyond the quantum hardware experiment and any previously verified classical evolution of the full wavefunction. At the bond dimension comparable to Q-CTRL's best classical run, our GPU implementation completes in $\sim\!100$ minutes, directly reducing the claimed $3000\times$ quantum advantage to $\sim\!36\times$. These results substantially narrow the quantum-classical performance gap and establish a new standard for tensor-network benchmarking of large-scale quantum simulations.

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 / 1 minor

Summary. The paper reports a GPU-accelerated, symmetry-reduced TDVP simulation of quench dynamics in the 1D Fermi-Hubbard model on L=60 sites. By reaching bond dimensions up to χ≈62,000 on four NVIDIA H200 GPUs, the authors claim fully converged results across the full time window, including rigorous certification of the previously unresolved high-entanglement regime t∈[5.2,6], extension of the classical frontier to t=7, and reduction of the Q-CTRL quantum-advantage claim from 3000× to ∼36× at comparable bond dimension.

Significance. If the convergence and certification claims are substantiated by explicit numerical evidence, the work would establish a substantially higher classical benchmark for tensor-network simulations of large-scale quantum dynamics, directly impacting assessments of quantum advantage in quench protocols and demonstrating the practical gains from symmetry exploitation and GPU tensor contractions.

major comments (2)
  1. [Abstract and numerical-results section] Abstract and § on numerical results: the central claim of 'fully converged results' and 'rigorous certification' of the high-entanglement regime t∈[5.2,6] at χ≈62,000 rests on the assumption that TDVP truncation error has fallen below the tolerance needed to certify observables against quantum hardware. No explicit χ-doubling data (e.g., stabilization of local densities, currents, or entanglement entropy between χ=31k and χ=62k) or independent error bounds are referenced; without such verification the certification remains unestablished.
  2. [performance-comparison section] § on performance comparison: the reduction of the quantum advantage to ∼36× is obtained by comparing the new GPU run at χ comparable to Q-CTRL's χ=4096 baseline. Because the manuscript does not demonstrate that the χ=4096 run itself has converged in the t∈[5.2,6] window, the factor-of-36 claim inherits the same unverified truncation error and cannot be used to certify the performance gap.
minor comments (1)
  1. [methods] Notation for the symmetry-reduced bond dimension and the precise definition of the truncation tolerance should be stated explicitly in the methods section to allow direct reproduction of the reported χ values.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thorough review and for highlighting the need for explicit verification of our convergence claims. We address each major comment below with clarifications and commitments to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract and numerical-results section] Abstract and § on numerical results: the central claim of 'fully converged results' and 'rigorous certification' of the high-entanglement regime t∈[5.2,6] at χ≈62,000 rests on the assumption that TDVP truncation error has fallen below the tolerance needed to certify observables against quantum hardware. No explicit χ-doubling data (e.g., stabilization of local densities, currents, or entanglement entropy between χ=31k and χ=62k) or independent error bounds are referenced; without such verification the certification remains unestablished.

    Authors: We agree that the current manuscript does not explicitly reference or display χ-doubling data or independent error bounds in the main text or abstract. The certification in the submitted version relies on internal checks of observable stabilization at the highest bond dimensions reached, but these were not presented with the requested detail. In the revised manuscript we will add explicit χ-doubling plots and tables in the numerical-results section, showing that local densities, currents, and entanglement entropy change by less than 10^{-4} between χ=31,000 and χ=62,000 throughout t∈[5.2,6], together with a brief discussion of the resulting truncation-error estimate. revision: yes

  2. Referee: [performance-comparison section] § on performance comparison: the reduction of the quantum advantage to ∼36× is obtained by comparing the new GPU run at χ comparable to Q-CTRL's χ=4096 baseline. Because the manuscript does not demonstrate that the χ=4096 run itself has converged in the t∈[5.2,6] window, the factor-of-36 claim inherits the same unverified truncation error and cannot be used to certify the performance gap.

    Authors: The ∼36× figure is a direct wall-clock-time comparison at fixed bond dimension χ=4096 and does not rely on that run being converged; it simply quantifies the speedup of our symmetric GPU implementation over the CPU baseline used by Q-CTRL at the same χ. The convergence certification is provided separately by the higher-χ runs. We will revise the performance-comparison section to state this distinction explicitly and to note that the original Q-CTRL quantum-advantage claim was made against a non-converged classical reference at χ=4096. revision: partial

Circularity Check

0 steps flagged

No circularity: results are direct outputs of new large-scale TDVP computations

full rationale

The paper reports new numerical results from GPU-accelerated TDVP simulations of the Fermi-Hubbard model at bond dimensions up to χ≈62,000. The central claims concern computational achievement, convergence via internal numerical checks at high χ, and direct comparison to prior quantum and classical runs. No steps reduce by the paper's own equations to fitted inputs, self-definitions, or self-citation chains; the simulation outputs are independent of any such loops. The provided text contains no load-bearing self-citations, ansatzes smuggled via prior work, or renaming of known results as derivations.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The claim rests on the domain assumption that the Fermi-Hubbard model possesses exploitable U(1)×SU(2) symmetry that reduces the effective Hilbert space without altering the dynamics, plus the standard mathematical assumption that TDVP with finite bond dimension converges to the exact evolution as chi increases.

free parameters (1)
  • bond dimension chi
    Convergence parameter chosen to reach the reported accuracy; values up to 62000 are selected to certify results.
axioms (2)
  • domain assumption The 1D Fermi-Hubbard Hamiltonian commutes with total particle number and total spin operators, allowing block-diagonal tensor-network representations.
    Invoked to justify the symmetry reduction that enables higher chi on available hardware.
  • standard math TDVP time evolution with finite bond dimension approximates the exact Schrödinger dynamics with controllable error that vanishes as chi→∞.
    Standard assumption of tensor-network methods; required for the convergence certification claim.

pith-pipeline@v0.9.1-grok · 5869 in / 1442 out tokens · 29943 ms · 2026-06-28T06:14:02.982589+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

36 extracted references · 28 canonical work pages · 2 internal anchors

  1. [1]

    G. S. Hartnett,et al., Quantum simulation of 1D Fermi-Hubbard model using a quan- tum computer (2025),https://arxiv.org/abs/2605.04025, arXiv:2605.04025v1

    Pith/arXiv arXiv 2025

  2. [2]

    Quantum Computing in the NISQ era and beyond

    J. Preskill, Quantum Computing in the NISQ era and beyond.Quantum2, 79 (2018), doi:10.22331/q-2018-08-06-79,https://doi.org/10.22331/q-2018-08-06-79

    work page internal anchor Pith review doi:10.22331/q-2018-08-06-79 2018

  3. [3]

    A. J. Daley,et al., Practical quantum advantage in quantum simulation.Nature607, 667–676 (2022), doi:10.1038/s41586-022-04940-6

    work page doi:10.1038/s41586-022-04940-6 2022

  4. [4]

    T. A. Chowdhury, V. Korepin, V. R. Pascuzzi, K. Yu, Quantum utility in simulating the real-time dynamics of the Fermi–Hubbard model using superconducting quantum computers.Applied Physics Reviews13(1), 011434 (2026), doi:10.1063/5.0306069, https://doi.org/10.1063/5.0306069

    work page doi:10.1063/5.0306069 2026

  5. [5]

    C. Cade, L. Mineh, A. Montanaro, S. Stanisic, Strategies for solving the Fermi- Hubbard model on near-term quantum computers.Phys. Rev. B102, 235122 (2020), doi:10.1103/PhysRevB.102.235122

    work page doi:10.1103/physrevb.102.235122 2020

  6. [6]

    S. Stanisic,et al., Observing ground-state properties of the Fermi-Hubbard model using a programmable quantum computer.Nature Communications13, 5743 (2022), doi:10.1038/s41467-022-33335-4

    work page doi:10.1038/s41467-022-33335-4 2022

  7. [7]

    Ignacio and Osborne, Tobias J

    J. Haegeman,et al., Time-Dependent Variational Principle for Quantum Lattices. Phys. Rev. Lett.107, 070601 (2011), doi:10.1103/PhysRevLett.107.070601,https: //link.aps.org/doi/10.1103/PhysRevLett.107.070601

    work page doi:10.1103/physrevlett.107.070601 2011

  8. [8]

    doi:10.1103/PhysRevB.94.165116 , url =

    J. Haegeman, C. Lubich, I. Oseledets, B. Vandereycken, F. Verstraete, Unifying time evolution and optimization with matrix product states.Phys. Rev. B94, 10 165116 (2016), doi:10.1103/PhysRevB.94.165116,https://link.aps.org/doi/10. 1103/PhysRevB.94.165116

    work page doi:10.1103/physrevb.94.165116 2016

  9. [9]

    doi:10.1016/j.aop.2019.167998 , url =

    S. Paeckel,et al., Time-evolution methods for matrix-product states.Annals of Physics411, 167998 (2019), doi:10.1016/j.aop.2019.167998

    work page doi:10.1016/j.aop.2019.167998 2019

  10. [10]

    Tindall, M

    J. Tindall, M. Fishman, E. M. Stoudenmire, D. Sels, Efficient Tensor Network Simulation of IBM’s Eagle Kicked Ising Experiment.PRX Quantum5, 010308 (2024), doi:10.1103/PRXQuantum.5.010308,https://link.aps.org/doi/10.1103/ PRXQuantum.5.010308

    work page doi:10.1103/prxquantum.5.010308 2024

  11. [11]

    Patra, S

    S. Patra, S. S. Jahromi, S. Singh, R. Or´ us, Efficient tensor network sim- ulation of IBM’s largest quantum processors.Phys. Rev. Res.6, 013326 (2024), doi:10.1103/PhysRevResearch.6.013326,https://link.aps.org/doi/10. 1103/PhysRevResearch.6.013326

    work page doi:10.1103/physrevresearch.6.013326 2024

  12. [12]

    H.-J. Liao, K. Wang, Z.-S. Zhou, P. Zhang, T. Xiang, Simulation of IBM’s kicked Ising experiment with Projected Entangled Pair Operator.arXiv e-printsarXiv:2308.03082 (2023), doi:10.48550/arXiv.2308.03082

    work page doi:10.48550/arxiv.2308.03082 2023

  13. [13]

    Tindall, A

    J. Tindall, A. F. Mello, M. Fishman, E. M. Stoudenmire, D. Sels, Dynamics of dis- ordered quantum systems with two- and three-dimensional tensor networks.Science 392(6800), 868–872 (2026), doi:10.1126/science.adx2728,https://www.science. org/doi/abs/10.1126/science.adx2728

    work page doi:10.1126/science.adx2728 2026

  14. [14]

    Design Initiative for a 10 TeV pCM Wakefield Collider,

    A. Kshetrimayum, S. S. Jahromi, S. Singh, R. Or´ us, Quantum Advantage: a Tensor Network Perspective.arXiv e-printsarXiv:2603.18825 (2026), doi:10.48550/arXiv. 2603.18825. 11

    work page internal anchor Pith review doi:10.48550/arxiv 2026

  15. [15]

    Fishman, S

    M. Fishman, S. R. White, E. M. Stoudenmire, The ITensor software library for tensor network calculations.SciPost Phys. Codebasesp. 4 (2022), doi:10.21468/ SciPostPhysCodeb.4,https://scipost.org/SciPostPhysCodeb.4

    2022

  16. [16]

    P. W. Anderson, Random-Phase Approximation in the Theory of Superconductivity. Physical Review112(6), 1900–1916 (1958), doi:10.1103/PhysRev.112.1900

    work page doi:10.1103/physrev.112.1900 1900

  17. [17]

    Zhang, Pseudospin symmetry and new collective modes of the Hubbard model

    S.-C. Zhang, Pseudospin symmetry and new collective modes of the Hubbard model. Physical Review Letters65(1), 120–122 (1990), doi:10.1103/PhysRevLett.65.120

    work page doi:10.1103/physrevlett.65.120 1990

  18. [18]

    Singh, R

    S. Singh, R. N. C. Pfeifer, G. Vidal, Tensor network decompositions in the presence of a global symmetry.Phys. Rev. A82, 050301(R) (2010), doi:10.1103/PhysRevA.82. 050301,https://link.aps.org/doi/10.1103/PhysRevA.82.050301

    work page doi:10.1103/physreva.82 2010

  19. [19]

    Singh, R

    S. Singh, R. N. C. Pfeifer, G. Vidal, Tensor network states and algorithms in the presence of a global U(1) symmetry.Phys. Rev. B83, 115125 (2011), doi:10.1103/ PhysRevB.83.115125,https://link.aps.org/doi/10.1103/PhysRevB.83.115125

    work page doi:10.1103/physrevb.83.115125 2011

  20. [20]

    Singh, G

    S. Singh, G. Vidal, Tensor network states and algorithms in the presence of a global SU(2) symmetry.Phys. Rev. B86, 195114 (2012), doi:10.1103/PhysRevB.86.195114, https://link.aps.org/doi/10.1103/PhysRevB.86.195114

    work page doi:10.1103/physrevb.86.195114 2012

  21. [21]

    Schmoll, S

    P. Schmoll, S. Singh, M. Rizzi, R. Or´ us, A programming guide for tensor net- works with global SU(2) symmetry.Annals of Physics419, 168232 (2020), doi:https: //doi.org/10.1016/j.aop.2020.168232,https://www.sciencedirect.com/science/ article/pii/S0003491620301664

    work page doi:10.1016/j.aop.2020.168232 2020

  22. [22]

    I. P. McCulloch, From density-matrix renormalization group to matrix product states. J. Stat. Mech.: Theory Exp.2007, P10014 (2007), doi:10.1088/1742-5468/2007/10/ P10014. 12

    work page doi:10.1088/1742-5468/2007/10/ 2007

  23. [23]

    A. Weichselbaum, Non-abelian symmetries in tensor networks: A quantum sym- metry space approach.Annals of Physics327(12), 2972–3047 (2012), doi:https: //doi.org/10.1016/j.aop.2012.07.009,https://www.sciencedirect.com/science/ article/pii/S0003491612001121

    work page doi:10.1016/j.aop.2012.07.009 2012

  24. [24]

    Rausch, S

    R. Rausch, S. Singh, Symmetric Tensor Networks on GPUs (2026), manuscript in preparation

    2026

  25. [25]

    ITensor Developers, Running on GPUs—ITensors.jl documentation (2025),https: //itensor.github.io/ITensors.jl/dev/RunningOnGPUs.html, accessed: May 25, 2026

    2025

  26. [26]

    Miller,et al., Simulation of Fermionic circuits using Majorana Propagation (2025), https://arxiv.org/abs/2503.18939

    A. Miller,et al., Simulation of Fermionic circuits using Majorana Propagation (2025), https://arxiv.org/abs/2503.18939

    arXiv 2025

  27. [27]

    Menczer, O

    A. Menczer, O. Legeza, Massively Parallel Tensor Network State Algorithms on Hy- brid CPU-GPU Based Architectures.Journal of Chemical Theory and Computation 21, 1572–1587 (2025), doi:https://doi.org/10.1021/acs.jctc.4c00661

    work page doi:10.1021/acs.jctc.4c00661 2025

  28. [28]

    Menczer, O

    A. Menczer, O. Legeza, Boosting the effective performance of massively parallel tensor network state algorithms on hybrid CPU-GPU based architectures via non-Abelian symmetries (2023),https://arxiv.org/abs/2309.16724

    arXiv 2023

  29. [29]

    A. Menczer,et al., Parallel Implementation of the Density Matrix Renormalization Group Method Achieving a Quarter petaFLOPS Performance on a Single DGX-H100 GPU Node.Journal of Chemical Theory and Computation20(19), 8397–8404 (2024), doi:10.1021/acs.jctc.4c00903,https://doi.org/10.1021/acs.jctc.4c00903

    work page doi:10.1021/acs.jctc.4c00903 2024

  30. [30]

    F. Pan, H. Gu, P. Springer, X. Li, Parallelizing Large-Scale Tensor Network Contrac- tion on Multiple GPUs (2026),https://arxiv.org/abs/2606.01852. 13

    Pith/arXiv arXiv 2026

  31. [31]

    Schollw¨ ock, The density-matrix renormalization group in the age of matrix product states.Annals of Physics326(1), 96–192 (2011), doi:https://doi.org/10

    U. Schollw¨ ock, The density-matrix renormalization group in the age of matrix product states.Annals of Physics326(1), 96–192 (2011), doi:https://doi.org/10. 1016/j.aop.2010.09.012,https://www.sciencedirect.com/science/article/pii/ S0003491610001752

    2011

  32. [32]

    Van Damme, L

    M. Van Damme, L. Vanderstraeten, J. Haegeman, F. Verstraete, Tangent-space meth- ods for truncating uniform matrix product states.SciPost Phys. Core4, 004 (2021), doi:10.21468/SciPostPhysCore.4.1.004

    work page doi:10.21468/scipostphyscore.4.1.004 2021

  33. [33]

    M. Yang, S. R. White, Time-dependent variational principle with ancillary Krylov subspace.Phys. Rev. B102, 094315 (2020), doi:10.1103/PhysRevB.102.094315

    work page doi:10.1103/physrevb.102.094315 2020

  34. [34]

    A. J. Dunnett, A. W. Chin, Efficient bond-adaptive approach for finite-temperature open quantum dynamics using the one-site time-dependent variational principle for matrix product states.Phys. Rev. B104, 214302 (2021), doi:10.1103/PhysRevB.104. 214302

    work page doi:10.1103/physrevb.104 2021

  35. [35]

    C. Lubich,From Quantum to Classical Molecular Dynamics: Reduced Models and Nu- merical Analysis, Zurich Lectures in Advanced Mathematics (European Mathematical Society (EMS), Z¨ urich) (2008), doi:10.4171/067

    work page doi:10.4171/067 2008

  36. [36]

    truncated weight

    P. Calabrese, J. Cardy, Evolution of entanglement entropy in one-dimensional sys- tems.J. Stat. Mech.: Theory Exp.2005, P04010 (2005), doi:10.1088/1742-5468/ 2005/04/P04010. 14 Figure 1: Main results. (a) TDVP time evolution of⟨n46↑⟩up tot= 8 forL= 60 fermions withϵtol = 10−4. Our U(1)×U(1) curve atχ= 4096 reproduces the Q-CTRL iTensor calculation and con...

    work page doi:10.1088/1742-5468/ 2005