Pith. sign in

REVIEW 3 major objections 7 minor 25 references

Ground-State Preparation of the Fermi-Hubbard Model on a Quantum Computer with 2D Topology via Quantum Eigenvalue Transformation of Unitary Matrices

T0 review · 3 major / 7 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read Fermionic swap networks let a 9-qubit grid prepare the 2x2 Fermi-Hubbard ground state with at least 0.99 overlap in noiseless simulation.

desk verdict A careful, honest QETU implementation study for a 2x2 Fermi-Hubbard toy model; the main limitations are disclosed by the authors themselves, so the result is a useful engineering benchmark rather than an overclaimed breakthrough. read the letter →

arxiv 2411.18535 v1 pith:U3MWW3AM submitted 2024-11-27 quant-ph

classification quant-ph PACS 03.67.Ac71.10.Fd
keywords Fermi-HubbardmodelQETUground-statepreparationfermionicswapnetworksTrotterdecomposition2Dqubittopologynoisesimulationquantumeigenvaluetransformation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to show that QETU, a recently introduced algorithm for preparing ground states without expensive block-encoding, can be implemented on realistic 2D grid hardware for the 2x2 Fermi-Hubbard model. It does this by turning the controlled forward-and-backward time evolution operator into a circuit of nearest-neighbour gates using fermionic swap networks, on a 3x3 grid of nine qubits. In noiseless statevector simulations, a single second-order Trotter step plus a polynomial filter of degree about 45 to 50 produces a state whose overlap with the exact ground state is at least 0.99. The paper also quantifies noise sensitivity, showing that hardware fidelity requirements are the limiting resource, and identifies the need for an initial state with guaranteed overlap with the ground state as the key practical bottleneck. Because the Fermi-Hubbard model is a standard testbed for strongly correlated electrons, a hardware-native route to its ground state is a concrete step toward practical quantum simulation.

What carries the argument

The central object is the oracle $V = \mathrm{diag}(e^{i\Delta t H}, e^{-i\Delta t H})$, the controlled forward and backward time evolution operator that QETU needs. Instead of controlling the time evolution directly, the paper builds V from a Trotter-Suzuki decomposition and anticommuting Pauli strings K1 and K2 that flip each term's sign, so only Pauli gates are controlled by the ancilla qubit. A fermionic swap network routes the fermionic modes so that every interacting orbital becomes a nearest neighbour at some point in the circuit; the 2x2 lattice needs only two distinct mappings, and the center-placed ancilla becomes adjacent to every orbital. The filter itself is a polynomial approximation F(x) to the shifted sign function, with rotation angles found by optimization, applied through the alternating QETU circuit.

What would settle it

Run a noiseless statevector simulation of the published circuit with $u=t=1$, the chosen initial state, one second-order Trotter step, and polynomial degree $d=50$; if the final overlap is materially below 0.99, the central efficiency claim is not reproducible as stated.

Watch

Extended reading notes

Core claim

Using the 2x2 Fermi-Hubbard model as a toy system, the authors claim that QETU ground-state preparation can be made efficient on a 3x3 grid of qubits. The Hamiltonian is split into an onsite interaction H1 and two hopping halves H2 and H3; two fermion-to-qubit mappings, with the ancilla qubit placed in the center of the grid, bring every required pair of orbitals into adjacency so that all gates act between neighboring qubits. The controlled forward and backward time evolution operator V is realized without controlling the Trotter factors: Pauli strings K1 and K2 that anticommute with the Hamiltonian terms flip the sign of the evolution, and a single second-order Trotter step suffices because the spectrum shift multiplies the simulation time by a small constant c1. With a polynomial approximation of the shifted sign function of degree around 45 to 50, the final state has overlap at least 0.99 with the exact ground state in noiseless simulation; the same circuit estimates the ground-state energy, and noise simulations show the hardware accuracy needed for practical use.

Load-bearing premise

Everything rests on the initial state having a large enough overlap with the ground state; the authors pick $|1001\rangle|--++\rangle$ empirically because random initialization cannot guarantee that overlap, and they expect the requirement to become harder for larger systems.

Editorial extensions

If this is right

  • For the 2x2 Fermi-Hubbard model, one second-order Trotter step is enough: the Hamiltonian spectrum shift shrinks the effective time step so the Trotter error stays around $5\times 10^{-3}$, and higher polynomial degree rather than more Trotter steps is what raises the overlap.
  • The controlled time evolution operator V can be implemented with nearest-neighbour two-qubit gates on a 3x3 grid using native iSWAP and CPhase gates, so the algorithm maps directly onto currently available superconducting qubit topologies.
  • Ground-state energy estimation needs only three measurement circuits, one for the onsite terms and two for the hopping terms after basis rotations, which keeps the measurement overhead small.
  • Even with error mitigation based on fermion-number conservation and the ancilla state, today's depolarizing noise levels prevent accurate energy estimates for this toy model, meaning gate fidelity is the limiting resource.
  • The empirically chosen initial state $|1001\rangle|--++\rangle$ is necessary; random initialization does not reliably give the overlap that the filter needs.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The center-ancilla placement is a 2x2 special case; for larger lattices the ancilla will not be adjacent to every orbital in two mappings, so the controlled Pauli strings will require additional swap rounds and the gate-count advantage may not carry over.
  • The same swap-network construction for V could be reused for other two-dimensional fermionic Hamiltonians whose two-body terms have the same structure, since the local decomposition only relies on iSWAP and CPhase gates.
  • Because the polynomial filter is independent of the hardware layout, the paper's trade-off between Trotter depth and polynomial degree can be attacked from either side; adaptive or factorized polynomial filters are natural next steps.
  • A practical large-scale version would need a separate, cheap initial-state preparation method, since the overlap assumption is the only ingredient in the paper that is not backed by a general construction.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 7 minor

Summary. The paper presents a hardware-oriented implementation of the QETU algorithm for ground-state preparation of the 2×2 Fermi-Hubbard model. The authors construct the controlled forward and backward time-evolution operator using a fermionic swap network tailored to a 9-qubit 3×3 grid, decompose all gates into a native gate set, and benchmark the resulting circuit in noiseless and noisy simulations. The main numerical result is that, with a single second-order Trotter step and a polynomial filter of degree around 45–50, the prepared state has overlap at least 0.99 with the exact ground state in the noiseless case. A noise model with depolarizing and measurement errors is used to estimate the required hardware quality, and the authors conclude that current superconducting devices are not yet accurate enough for this protocol.

Significance. If taken as an implementation study of a specific 2×2 instance, the paper is a useful and honest contribution. It provides explicit circuit decompositions into a realistic native gate set, makes the code available, and gives a candid noise-resource analysis. The main value is in showing how fermionic swap networks can reduce the cost of the controlled time evolution needed by QETU on a small toy model. However, the demonstrated protocol is conditional on two pieces of information that would not be available in an actual application: an empirically chosen initial state with a guaranteed overlap γ ≥ 0.09102, and spectral parameters (μ, λ_min, λ_max) obtained from exact diagonalization. The paper also does not provide a scaling analysis for larger lattices. These caveats are partly acknowledged in Section VIII, but the abstract and concluding claims should be scoped accordingly.

major comments (3)
  1. [Section VII, Fig. 6] The reported overlap |⟨ψ_final|ψ_0⟩|² ≥ 0.99 is obtained with the initial state |ψ_init⟩ = |1001⟩|--++⟩, which is selected empirically to guarantee γ ≥ 0.09102. Section VIII concedes that random initialization cannot guarantee this overlap and that scalable initial-state preparation is indispensable. Since QETU is an eigenspace filter, its success is entirely conditional on the initial state containing a non-negligible ground-state component; the paper provides no argument that γ does not decay exponentially with system size. The abstract and concluding claims should therefore be framed as a proof-of-principle for the 2×2 instance rather than as a general ground-state preparation method for the Fermi-Hubbard model.
  2. [Section IV E and Section VIII] The circuit parameters c1, c2, and μ are computed from λ_min and λ_max obtained by exact diagonalization of the Hamiltonian. Consequently, the numerical results in Figs. 6 and 7 demonstrate the performance of the circuit conditional on spectral information that would not be available in an actual application. The paper should state this limitation prominently, e.g., in the abstract, and clarify that the reported experiments validate the circuit implementation rather than the complete QETU ground-state-solving protocol including the binary search over μ.
  3. [Section VIII] The sentence 'we showed that ... we can efficiently realize a controlled forward and backward time evolution operator of the Fermi-Hubbard model on a 2D lattice' is stronger than what the paper demonstrates. The construction exploits the 2×2-specific placement of the auxiliary qubit in the center of the grid and the fact that only two fermion-to-qubit mappings are needed; the paper itself notes that these optimizations do not generalize and that finding optimal Pauli strings for larger systems is nontrivial. The claim should be restricted to the 2×2 instance, or the paper should provide a concrete resource estimate (gate count and depth scaling) for general 2D lattices.
minor comments (7)
  1. [Eqs. (16) and (19)] There are apparent typos in the spin indices and duplicated hopping terms: in Eq. (16) the term (a†_{3,↓}a_{4,↓} + a†_{3,↓}a_{4,↓}) appears twice, and in Eq. (19) 'Y_{1,↑}Y_{2,↓}Z_{1,↓}Z_{2,↓}' should presumably read 'Y_{1,↑}Y_{2,↑}Z_{1,↓}Z_{2,↓}'. These should be corrected.
  2. [Section IV C] The text contains a typo: 'The fist part of the kinetic hopping term' should be 'The first part'.
  3. [Section V] The phrase 'prior the the measurements' should be 'prior to the measurements'.
  4. [Section VII] The sentence 'Both circuits were initialized with the same initial state' is ambiguous; Section V describes three measurement circuits, so the text should say 'all three circuits'.
  5. [Fig. 7] The axes of the magnified inset are not clearly labeled; please specify that the horizontal axis is the polynomial degree d and the vertical axis is the estimated ground-state energy ⟨ψ0|H|ψ0⟩.
  6. [Eq. (33)] The matrix U is written with a bare 'H' in the lower-right block, which is ambiguous; clarify that this denotes the Hadamard gate or write out the explicit matrix entries.
  7. [Appendix B, Eq. (B4)] The notation 'hop_{1,3} · Z_2' is nonstandard and could be misread as a product of operators; please clarify that the resulting operator is the hopping term with a Z-string on the intermediate qubit.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the circuit construction and QETU demonstration are benchmarked against exact diagonalization with explicitly acknowledged inputs, not derived from the target outputs.

full rationale

This is an implementation and feasibility study for the 2x2 Fermi-Hubbard model, not a first-principles prediction. The central technical content—constructing the controlled forward/backward time-evolution operator via fermionic swap networks and then applying the QETU filter—is derived from the Hamiltonian terms, the Jordan-Wigner mapping, and the Trotter decomposition. The reported final-state overlap and energy estimates are obtained by simulating the actual circuit and measuring the prepared state, not by fitting those outputs. The initial state |psi_init> = |1001>|--++> is explicitly described as empirically selected to ensure gamma >= 0.09102, and the spectral parameters lambda_min, lambda_max, and mu are taken from exact diagonalization; Section VIII openly states that 'we assume knowledge of the parameter µ obtained by direct diagonalization' and that scalable initial-state preparation is indispensable. These are acknowledged limitations on the algorithm's standalone predictive power, but they do not make the derivation circular: QETU transforms a state with a known nonzero overlap into a state with higher overlap, and the simulation result is a nontrivial function of the polynomial degree and Trotter error. The only self-citation ([23], by one of the present authors) appears in the outlook as a suggested scalability improvement and is not load-bearing. No equation is defined in terms of the quantity it is claimed to predict, and no fitted parameter is renamed as an independent result. The paper's claims are explicitly scoped to the 2x2 toy model and validated against external exact diagonalization, so the appropriate finding is no significant circularity.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The implementation rests on standard QETU and fermion-to-qubit machinery. The paper's own choices are the empirical initial state, the unspecified spectral margin η, and the specific Pauli strings K1 and K2. No new physical entities are introduced.

free parameters (2)
  • initial state |ψ_init⟩ = |1001⟩|--++⟩
    Chosen empirically in Section VII to ensure overlap γ ≥ 0.09102 with the ground state; the reported fidelities depend on this choice.
  • spectral margin η = not specified in text
    The spectrum shift in Eq. (30) requires choosing η to keep eigenvalues in [η, π−η]; the paper does not state the value used, which affects c1, c2, and the polynomial constraints.
assumptions (5)
  • domain assumption QETU algorithm correctly implements F(cos(H/2)) when V = diag(e^{iΔtH}, e^{-iΔtH}) is available (Corollary 17 of [1]).
    The paper relies on the QETU theorem for its filtering and circuit construction; it is cited from [1] and not re-derived.
  • standard math Jordan-Wigner transformation maps fermions to qubits with the anti-commutation relations preserved (Appendix A).
    Standard fermion-to-qubit mapping used throughout the Hamiltonian decomposition.
  • standard math Second-order Trotter-Suzuki decomposition (Eq. 21) approximates the time evolution with error O(Δt^3).
    Used to implement e^{-iΔtH} as a product of H1, H2, H3 exponentials.
  • ad hoc to paper The Pauli strings K1 and K2 in Eq. (28) anticommute with the respective Hamiltonian terms as claimed, and the fermionic swap network identity in Appendix B holds.
    The paper states these relations in Section IV.D and Appendix B but does not provide a full proof for every term; they are load-bearing for the controlled time evolution.
  • domain assumption Exact spectrum knowledge: μ, λ_min, λ_max are obtained by direct diagonalization and used to build the filter and shift the Hamiltonian.
    The paper assumes this knowledge in the numerical experiments and flags it in the Discussion as a limitation requiring binary search for unknown spectra.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Ground-State Preparation of the Fermi-Hubbard Model on a Quantum Computer with 2D Topology via Quantum Eigenvalue Transformation of Unitary Matrices." pith.science (2026). https://pith.science/paper/U3MWW3AM

@misc{pith2026241118535,
  author       = {Pith},
  title        = {Pith review of: Ground-State Preparation of the Fermi-Hubbard Model on a Quantum Computer with 2D Topology via Quantum Eigenvalue Transformation of Unitary Matrices},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/U3MWW3AM}},
  note         = {Machine review of arXiv:2411.18535}
}
abstract

Quantum computing holds immense promise for simulating quantum systems, a critical task for advancing our understanding of complex quantum phenomena. One of the primary goals in this domain is to accurately approximate the ground state of quantum systems. The Fermi-Hubbard model, particularly, is of profound interest due to its implications for high-temperature superconductivity and strongly correlated electron systems. The quantum eigenvalue transformation of unitary matrices (QETU) algorithm offers a novel approach for ground state estimation by utilizing a controlled Hamiltonian time evolution operator, circumventing the resource-intensive block-encoding required by previous methods. In this work, we apply the QETU algorithm to the $2 \times 2$ Fermi-Hubbard model, presenting circuit simplifications tailored to the model and introducing a mapping to a 9-qubit grid-like hardware architecture inspired by fermionic swap networks. We investigate how the selection of a favorable hardware architecture can benefit the circuit construction. Additionally, we explore the feasibility of this method under the influence of noise, focusing on its robustness and practical applicability.

Figures

Figures reproduced from arXiv: 2411.18535 by the authors.

Figure 1
Figure 1. FIG. 1. Circuit of the QETU algorithm using a controlled forward and backward time evolution operator, adapted from [1]. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Left: Plot of the step function [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. (a) Depiction of the mapping of the 2 [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: FIG. 4. A visualization of the quantum circuit that realizes one step of the second-order Trotter decomposition of the controlled [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: The quantum circuit for one Trotter step is de [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 5
Figure 5. Figure 5: FIG. 5. The convergence of the Trotterization and the depth [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6. Results of the numerical simulations in the ideal, [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7. Estimation of the ground-state energy of the Fermi-Hubbard model with [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 15 canonical work pages

  1. [1]

    Y. Dong, L. Lin, and Y. Tong, Ground-state prepara- tion and energy estimation on early fault-tolerant quan- tum computers via quantum eigenvalue transformation of unitary matrices, PRX Quantum 3, 040305 (2022)

  2. [2]

    Gily´ en, Y

    A. Gily´ en, Y. Su, G. H. Low, and N. Wiebe, Quantum singular value transformation and beyond: exponential improvements for quantum matrix arithmetics, in Pro- ceedings of the 51st Annual ACM SIGACT Symposium on Theory of Computing , STOC 2019 (Association for Computing Machinery, New York, NY, USA, 2019) p. 193–204

  3. [3]

    G. H. Low and I. L. Chuang, Hamiltonian Simulation by Qubitization, Quantum 3, 163 (2019)

  4. [4]

    J. M. Martyn, Z. M. Rossi, A. K. Tan, and I. L. Chuang, Grand unification of quantum algorithms, PRX Quan- tum 2, 040203 (2021)

  5. [5]

    Schuch and J

    N. Schuch and J. Siewert, Natural two-qubit gate for quantum computation using the XY interaction, Phys. Rev. A 67, 032301 (2003)

  6. [6]

    D. M. Abrams, N. Didier, B. R. Johnson, M. P. da Silva, and C. A. Ryan, Implementation of XY entangling gates with a single calibrated pulse, Nat. Electron. 3, 744–750 (2020)

  7. [7]

    Hubbard, Electron correlations in narrow energy bands, Proc

    J. Hubbard, Electron correlations in narrow energy bands, Proc. R. Soc. Lond. A 276, 238–257 (1963)

  8. [8]

    I. D. Kivlichan, J. McClean, N. Wiebe, C. Gidney, A. Aspuru-Guzik, G. K.-L. Chan, and R. Babbush, Quantum simulation of electronic structure with linear depth and connectivity, Phys. Rev. Lett. 120, 110501 (2018)

Show all 25 references
  1. [9]

    Jordan and E

    P. Jordan and E. Wigner, ¨Uber das paulische ¨ aquivalenzverbot, Zeitschrift f¨ ur Physik47, 631 (1928)

  2. [10]

    Ortiz, J

    G. Ortiz, J. E. Gubernatis, E. Knill, and R. Laflamme, Quantum algorithms for fermionic simulations, Phys. Rev. A 64, 022319 (2001)

  3. [11]

    Y. Dong, X. Meng, K. B. Whaley, and L. Lin, Effi- cient phase-factor evaluation in quantum signal process- ing, Phys. Rev. A 103, 042419 (2021)

  4. [12]

    J. Wang, Y. Dong, and L. Lin, On the energy landscape of symmetric quantum signal processing, Quantum6, 850 (2022)

  5. [13]

    Wille, A

    R. Wille, A. Lye, and R. Drechsler, Optimal swap gate insertion for nearest neighbor quantum circuits, in 2014 19th Asia and South Pacific Design Automation Confer- ence (ASP-DAC) (Singapore, 2014) pp. 489–494

  6. [14]

    Somma, G

    R. Somma, G. Ortiz, J. E. Gubernatis, E. Knill, and R. Laflamme, Simulating physical phenomena by quan- tum networks, Phys. Rev. A 65, 042323 (2002)

  7. [15]

    D. C. McKay, C. J. Wood, S. Sheldon, J. M. Chow, and J. M. Gambetta, Efficient z gates for quantum comput- ing, Phys. Rev. A 96, 022330 (2017)

  8. [16]

    Hatano and M

    N. Hatano and M. Suzuki, Finding exponential prod- uct formulas of higher orders, in Quantum Annealing and Other Optimization Methods , edited by A. Das and B. K. Chakrabarti (Springer Berlin Heidelberg, Berlin, Heidelberg, 2005) pp. 37–68

  9. [17]

    Dong and L

    Y. Dong and L. Lin, QET-U, https://github.com/ qsppack/QETU (2022)

  10. [18]

    C. Cade, L. Mineh, A. Montanaro, and S. Stanisic, Strategies for solving the fermi-hubbard model on near- term quantum computers, Phys. Rev. B 102, 235122 (2020)

  11. [19]

    Qiskit contributors, Qiskit: An open-source framework for quantum computing (2023)

  12. [20]

    T. R. M¨ uller, Ground-state preparation of the 2 × 2 fermi-hubbard model via QETU, https://github.com/ thilomueller/QETU (2024)

  13. [21]

    Georgopoulos, C

    K. Georgopoulos, C. Emary, and P. Zuliani, Modeling and simulating the noisy behavior of near-term quantum computers, Phys. Rev. A 104, 062432 (2021). 15

  14. [22]

    M. A. Nielsen and I. L. Chuang, Quantum Computa- tion and Quantum Information: 10th Anniversary Edi- tion (Cambridge University Press, 2010) pp. 378–379

  15. [23]

    Karacan, Y

    E. Karacan, Y. Chen, and C. B. Mendl, Enhancing scal- ability of quantum eigenvalue transformation of unitary matrices for ground state preparation through adaptive finer filtering (2024), arXiv:2401.09091 [quant-ph]

  16. [24]

    J. M. Martyn, Z. M. Rossi, K. Z. Cheng, Y. Liu, and I. L. Chuang, Parallel quantum signal processing via polyno- mial factorization (2024), arXiv:2409.19043 [quant-ph]

  17. [25]

    K¨ okc¨ u, D

    E. K¨ okc¨ u, D. Camps, L. B. Oftelie, W. A. De Jong, R. Van Beeumen, and A. F. Kemper, Algebraic com- pression of free fermionic quantum circuits: Particle cre- ation, arbitrary lattices and controlled evolution, in 2023 IEEE International Conference on Quantum Computing and ...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.