pith. sign in

arxiv: 2510.02141 · v1 · submitted 2025-10-02 · 🪐 quant-ph · cond-mat.str-el

Quantum speed-up for solving the one-dimensional Hubbard model using quantum annealing

Kunal Vyas , Fengping Jin , Hans De Raedt , Kristel Michielsen This is my paper

Pith reviewed 2026-05-18 10:29 UTC · model grok-4.3

classification 🪐 quant-ph cond-mat.str-el
keywords quantum annealingHubbard modelquantum speed-upone-dimensional latticeBethe ansatzground statequantum simulationmany-body physics
0
0 comments X

The pith

Quantum annealing simulations reveal a substantial speed-up over Bethe-ansatz methods for finding ground states of the half-filled one-dimensional Hubbard model up to 40 qubits.

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

The paper investigates quantum annealing as a way to solve the one-dimensional Hubbard model, a key description of interacting electrons in condensed matter physics. Using gate-based quantum computer simulations of the annealing process, the authors examine systems up to 40 qubits and track how annealing time scales to reach the ground state. For half-filled cases they report a clear quantum speed-up relative to classical algorithms that rely on the Bethe-ansatz equations. This matters because the Hubbard model underpins understanding of phenomena such as magnetism and superconductivity, and demonstrating an advantage even for an exactly solvable model tests whether quantum methods can scale usefully for many-body problems.

Core claim

The authors perform gate-based quantum computer simulations of quantum annealing for the Hubbard Hamiltonian. They study the scaling of required annealing time for obtaining the ground state and find that for the half-filled cases considered, there is a substantial quantum speed-up over algorithms based on the Bethe-ansatz equations.

What carries the argument

Quantum annealing dynamics applied to the one-dimensional Hubbard Hamiltonian and simulated on a gate-based quantum computer to extract ground-state success probability and time scaling.

If this is right

  • For half-filled one-dimensional Hubbard models the annealing time grows more slowly than the time needed by Bethe-ansatz solvers as system size increases.
  • The same annealing approach may be applied to other integrable lattice models where exact classical solutions exist but become costly.
  • Observed scaling up to 40 qubits suggests the method remains viable for modestly larger systems before hardware limits intervene.
  • Success on an exactly solvable model provides a benchmark for testing quantum algorithms on non-integrable extensions of the Hubbard model.

Where Pith is reading between the lines

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

  • If the speed-up survives on real hardware it would motivate hybrid quantum-classical workflows that use annealing only for the hardest parameter regimes of correlated-electron models.
  • The result raises the question whether similar annealing schedules can be tuned for two-dimensional Hubbard systems where no exact classical solution is known.
  • A direct comparison of the extracted ground-state wave functions against Bethe-ansatz results could quantify how much of the speed-up comes from quantum coherence versus classical preprocessing.

Load-bearing premise

Classical simulation of the quantum annealing dynamics accurately reproduces the success probability and scaling that physical quantum hardware would produce, and the observed scaling up to 40 qubits continues without degradation for larger systems.

What would settle it

Implementing the same annealing schedule on actual quantum hardware for a half-filled chain beyond 40 sites and measuring either no speed-up relative to Bethe-ansatz run times or a breakdown in success probability scaling.

Figures

Figures reproduced from arXiv: 2510.02141 by Fengping Jin, Hans De Raedt, Kristel Michielsen, Kunal Vyas.

Figure 1
Figure 1. Figure 1: Dispersion relation and the ground state [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: ∆E as a function of TA for a linear annealing schedule with U/tH = 4. gates is 2(L − 1) + 2(L − 1) + 2 × L = 6L − 4. The total number of gates required to perform quantum annealing for a given TA is therefore (TA/τ )(18L − 4). In summary, the product-formula algorithm allows us to perform gate-based quantum annealing simulations by a circuit the size of which scales linearly with system size. The amount of… view at source ↗
Figure 3
Figure 3. Figure 3: ∆E as a function of TA for a linear annealing schedule with (a) U/tH = 8 and (b) U/tH = 16. The crosses indicate onsets ϵ(L), which are plotted as a function of L in the inset. machine at the J¨ulich Supercomputing Center. One time step of the quantum annealing algorithm of a 40-qubit system requires 356 gates. For the largest an￾nealing time considered (TA = 40) and a time step of τ = 0.025, the number of… view at source ↗
Figure 4
Figure 4. Figure 4: shows the scaling of α(L) for U/tH = 4, 8, 16. The parallel lines are suggestive of a consistent scaling with L dependence. For U/tH = 16, α(L) ∝ L 1.29 . (34) Combining these results to extrapolate the scaling of re￾quired TA to arbitrarily large L, we find from Eq. (32) that, TA < T′ A ∝ L 0.995 . (35) To calculate the ground state energy in practice re￾quires performing measurements on the quantum de￾vi… view at source ↗
Figure 5
Figure 5. Figure 5: Quantum circuit for implementing a Givens [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Circuit that prepares the ground state of [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗
Figure 8
Figure 8. Figure 8: The computation time TBA to solve the Bethe-ansatz equations as a function of L for Hubbard Hamiltonians with 2N↓ = N = L and U = 4tH. We are interested in finding the ground state energy of the half-filled Hubbard model (2N↓ = N = L) for a range of system sizes L. We thus calculate E0(L) = E0(L, N = L, N↓ = L/2) using Eq. (4) and the obtained roots {k1, k2 . . . kL} [PITH_FULL_IMAGE:figures/full_fig_p010… view at source ↗
Figure 9
Figure 9. Figure 9: (a) ∆E as a function of TA and (b) α(L) as a function of L for a sinusoidal annealing schedule with U/tH = 4. Appendix C: Sinusoidal schedule We repeat the simulations for the annealing schedule with a sinusoidal dependence on s. Namely, H(s) = − t X L ⟨i,j⟩,σ (c † iσcjσ + c † jσciσ) + sin(πs − π 2 ) + 1 2 U X L i ni↑ni↓ . (C1) For such a schedule with H˙ (0) = H˙ (1) = 0, the depen￾dence of the bound in E… view at source ↗
read the original abstract

The Hubbard model has occupied the minds of condensed matter physicists for most part of the last century. This model provides insight into a range of phenomena in correlated electron systems. We wish to examine the paradigm of quantum algorithms for solving such many-body problems. The focus of our current work is on the one-dimensional model which is integrable, meaning that there exist analytical results for determining its ground state. In particular, we demonstrate how to perform a gate-based quantum computer simulation of quantum annealing for the Hubbard Hamiltonian. We perform simulations for systems with up to 40 qubits to study the scaling of required annealing time for obtaining the ground state. We find that for the half-filled cases considered, there is a substantial quantum speed-up over algorithms based on the Bethe-ansatz equations.

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

3 major / 2 minor

Summary. The manuscript describes gate-based classical simulations of quantum annealing applied to the one-dimensional Hubbard model. Simulations are performed for systems up to 40 qubits, with a focus on half-filled cases. The central claim is that the required annealing time exhibits a substantial quantum speed-up relative to classical algorithms based on the Bethe-ansatz equations for obtaining the ground state.

Significance. A verified polynomial or better scaling advantage for quantum annealing over the Bethe-ansatz solver on the 1D Hubbard model would be noteworthy for quantum algorithms targeting integrable many-body systems, especially if the advantage persists on hardware. The work also supplies concrete numerical data on annealing-time scaling up to 40 qubits, which could serve as a benchmark if the simulation fidelity and extrapolation are rigorously established.

major comments (3)
  1. The abstract and simulation description provide no quantitative details on the annealing schedule (e.g., functional form of s(t), minimum gap encountered, or chosen total time T), error metrics for the final state fidelity, or the precise implementation and runtime scaling of the Bethe-ansatz baseline solver. Without these, it is impossible to assess whether the reported speed-up is robust or sensitive to post-hoc parameter choices.
  2. For the 1D Hubbard model the single-particle gap closes as ~1/L. The manuscript reports results only up to 40 qubits and does not analyze or bound the growth of the required annealing time T once diabatic transitions become probable for larger L. This leaves open the possibility that the observed scaling degrades exponentially beyond the simulated range.
  3. Exact classical simulation of the time-dependent Schrödinger equation is infeasible at 40 qubits; the work therefore necessarily employs tensor-network, truncated-basis, or other approximate methods. No fidelity benchmarks, convergence tests, or error bounds relative to ideal coherent evolution are supplied, undermining that the simulated success probabilities match those expected on physical hardware.
minor comments (2)
  1. Notation for the Hubbard Hamiltonian parameters (U, t) and the mapping to the qubit Hamiltonian should be stated explicitly in the methods section for reproducibility.
  2. Figure captions should include the precise system sizes, filling factors, and number of independent runs used to extract the scaling data.

Simulated Author's Rebuttal

3 responses · 1 unresolved

We thank the referee for their thorough review and valuable feedback on our manuscript. We have carefully considered each comment and provide detailed responses below. Where appropriate, we have revised the manuscript to address the concerns raised.

read point-by-point responses

  1. Referee: The abstract and simulation description provide no quantitative details on the annealing schedule (e.g., functional form of s(t), minimum gap encountered, or chosen total time T), error metrics for the final state fidelity, or the precise implementation and runtime scaling of the Bethe-ansatz baseline solver. Without these, it is impossible to assess whether the reported speed-up is robust or sensitive to post-hoc parameter choices.

    Authors: We agree that providing these quantitative details is essential for evaluating the robustness of our results. In the revised version of the manuscript, we have expanded the Methods section to include the specific functional form of the annealing schedule s(t), the minimum energy gap observed in our simulations, and the chosen total annealing times T for different system sizes. We have also added fidelity metrics comparing the final state to the known ground state for smaller systems where exact results are available, and to Bethe-ansatz solutions for larger ones. Furthermore, we now describe the numerical implementation of the Bethe-ansatz solver, including its runtime scaling, which we find to be polynomial in system size. These additions clarify that the reported speed-up is not sensitive to arbitrary parameter choices but follows from systematic simulations. revision: yes

  2. Referee: For the 1D Hubbard model the single-particle gap closes as ~1/L. The manuscript reports results only up to 40 qubits and does not analyze or bound the growth of the required annealing time T once diabatic transitions become probable for larger L. This leaves open the possibility that the observed scaling degrades exponentially beyond the simulated range.

    Authors: We acknowledge the referee's concern regarding the closing of the gap and the potential for diabatic transitions at larger system sizes. Our simulations up to 40 qubits demonstrate a favorable scaling of the annealing time that provides a substantial speed-up over the Bethe-ansatz approach for the half-filled cases. While we have not performed simulations beyond 40 qubits due to computational limitations of the classical simulation methods, we have added a discussion in the revised manuscript on the implications of the gap closing and why the observed polynomial-like scaling is expected to persist based on the adiabatic condition and the nature of the quantum annealing process for this model. However, a rigorous bound for arbitrarily large L would require further theoretical analysis or larger-scale simulations, which are beyond the scope of the current work. revision: partial

  3. Referee: Exact classical simulation of the time-dependent Schrödinger equation is infeasible at 40 qubits; the work therefore necessarily employs tensor-network, truncated-basis, or other approximate methods. No fidelity benchmarks, convergence tests, or error bounds relative to ideal coherent evolution are supplied, undermining that the simulated success probabilities match those expected on physical hardware.

    Authors: We thank the referee for highlighting this important point about the approximate nature of our classical simulations. The simulations were carried out using tensor-network methods to approximate the time evolution. In the revised manuscript, we have included detailed information on the tensor-network implementation, along with fidelity benchmarks against exact results for smaller system sizes (up to 20 qubits), convergence tests with respect to the bond dimension, and error bounds on the success probability. These additions demonstrate that the approximation errors are controlled and that the reported success probabilities are reliable indicators of what would be observed in an ideal coherent quantum annealing process on hardware. revision: yes

standing simulated objections not resolved
  • Rigorous bound on the annealing time scaling for system sizes significantly larger than 40 qubits where diabatic transitions may dominate.

Circularity Check

0 steps flagged

No significant circularity in the derivation of the claimed quantum speed-up

full rationale

The paper's central claim rests on classical simulations of quantum annealing dynamics for the 1D Hubbard model (up to 40 qubits) and a direct comparison of the observed annealing-time scaling against the known computational cost of solving the Bethe-ansatz equations for half-filled instances. The Bethe-ansatz solver is an independent, externally established classical method for the integrable model and is not derived from or fitted within the paper's own equations or simulation outputs. No self-definitional steps, fitted parameters renamed as predictions, load-bearing self-citations, or ansatz smuggling appear in the reported chain; the scaling result is presented as an empirical observation from the simulations rather than a tautological consequence of the inputs. The derivation therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no explicit free parameters, axioms, or invented entities are identifiable from the provided text.

pith-pipeline@v0.9.0 · 5665 in / 1027 out tokens · 30413 ms · 2026-05-18T10:29:43.377473+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

51 extracted references · 51 canonical work pages · 1 internal anchor

  1. [1]

    Initial state preparation circuit

    work page

  2. [2]

    Circuit for performing dynamics of quantum an- nealing In the following sections, we describe how we construct these circuits to perform quantum annealing for fermion systems on a digital quantum computer. 4 A. Jordan-Wigner transformation The first step is to transform the fermion Hamiltonian into a qubit Hamiltonian comprising Pauli matrices. To preserv...

    work page

  3. [3]

    Hubbard, Proceedings of the Royal Society of London

    J. Hubbard, Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences276, 238 (1963)

    work page 1963

  4. [4]

    F. H. L. Essler, H. Frahm, F. G¨ ohmann, A. Kl¨ umper, and V. E. Korepin,The One-Dimensional Hubbard Model (Cambridge University Press, Cambridge, 2005)

    work page 2005

  5. [5]

    E. H. Lieb and F. Y. Wu, Physical Review Letters20, 1445 (1968)

    work page 1968

  6. [6]

    Bethe, Zeitschrift f¨ ur Physik71, 205 (1931)

    H. Bethe, Zeitschrift f¨ ur Physik71, 205 (1931)

    work page 1931

  7. [7]

    Andrei, inLow-Dimensional Quantum Field Theories for Condensed Matter Physicists, Series in Modern Con- densed Matter Physics, Vol

    N. Andrei, inLow-Dimensional Quantum Field Theories for Condensed Matter Physicists, Series in Modern Con- densed Matter Physics, Vol. Volume 6 (WORLD SCIEN- TIFIC, 1995) pp. 457–551

    work page 1995

  8. [8]

    Schulz, Journal of Physics C: Solid State Physics18, 581 (1985)

    H. Schulz, Journal of Physics C: Solid State Physics18, 581 (1985)

    work page 1985

  9. [9]

    L. K. Grover, Physical Review Letters79, 325 (1997)

    work page 1997

  10. [10]

    Peruzzo, J

    A. Peruzzo, J. McClean, P. Shadbolt, M.-H. Yung, X.-Q. Zhou, P. J. Love, A. Aspuru-Guzik, and J. L. O’Brien, 12 Nature Communications5, 4213 (2014)

    work page 2014

  11. [11]

    C. Cade, L. Mineh, A. Montanaro, and S. Stanisic, Phys- ical Review B102, 235122 (2020)

    work page 2020

  12. [12]

    Wecker, M

    D. Wecker, M. B. Hastings, and M. Troyer, Physical Re- view A92, 042303 (2015)

    work page 2015

  13. [13]

    J. S. Van Dyke, G. S. Barron, N. J. Mayhall, E. Barnes, and S. E. Economou, PRX Quantum2, 040329 (2021)

    work page 2021

  14. [14]

    W. Li, M. Okyay, and R. I. Nepomechie, Journal of Physics A: Mathematical and Theoretical55, 355305 (2022)

    work page 2022

  15. [15]

    Kadowaki and H

    T. Kadowaki and H. Nishimori, Physical Review E58, 5355 (1998)

    work page 1998

  16. [16]

    Farhi, J

    E. Farhi, J. Goldstone, S. Gutmann, J. Lapan, A. Lund- gren, and D. Preda, Science292, 472 (2001)

    work page 2001

  17. [17]

    Quantum Computation by Adiabatic Evolution

    E. Farhi, J. Goldstone, S. Gutmann, and M. Sipser, Quantum Computation by Adiabatic Evolution (2000), arXiv:quant-ph/0001106

    work page internal anchor Pith review Pith/arXiv arXiv 2000

  18. [18]

    Martoˇ n´ ak, G

    R. Martoˇ n´ ak, G. E. Santoro, and E. Tosatti, Physical Review E70, 057701 (2004)

    work page 2004

  19. [19]

    Brooke, D

    J. Brooke, D. Bitko, T. F., Rosenbaum, and G. Aeppli, Science284, 779 (1999)

    work page 1999

  20. [20]

    Munoz-Bauza and D

    H. Munoz-Bauza and D. Lidar, Physical Review Letters 134, 160601 (2025)

    work page 2025

  21. [21]

    R. D. Somma, D. Nagaj, and M. Kieferov´ a, Physical Re- view Letters109, 050501 (2012)

    work page 2012

  22. [22]

    Albash and D

    T. Albash and D. A. Lidar, Physical Review X8, 031016 (2018)

    work page 2018

  23. [23]

    Mehta, F

    V. Mehta, F. Jin, H. De Raedt, and K. Michielsen, Phys- ical Review A105, 062406 (2022)

    work page 2022

  24. [24]

    T.-J. Hsu, F. Jin, C. Seidel, F. Neukart, H. D. Raedt, and K. Michielsen, Communications in Computational Physics26, 928 (2019)

    work page 2019

  25. [25]

    Farhi and A

    E. Farhi and A. W. Harrow, Quantum Supremacy through the Quantum Approximate Optimization Algo- rithm (2019), arXiv:1602.07674 [quant-ph]

    work page arXiv 2019

  26. [26]

    R. P. Feynman, International Journal of Theoretical Physics21, 467 (1982)

    work page 1982

  27. [27]

    Smith, M

    A. Smith, M. S. Kim, F. Pollmann, and J. Knolle, npj Quantum Information5, 106 (2019)

    work page 2019

  28. [28]

    Gross and I

    C. Gross and I. Bloch, Science357, 995 (2017)

    work page 2017

  29. [29]

    Bernien, S

    H. Bernien, S. Schwartz, A. Keesling, H. Levine, A. Om- ran, H. Pichler, S. Choi, A. S. Zibrov, M. Endres, M. Greiner, V. Vuleti´ c, and M. D. Lukin, Nature551, 579 (2017)

    work page 2017

  30. [30]

    Harris, Y

    R. Harris, Y. Sato, A. J. Berkley, M. Reis, F. Al- tomare, M. H. Amin, K. Boothby, P. Bunyk, C. Deng, C. Enderud, S. Huang, E. Hoskinson, M. W. Johnson, E. Ladizinsky, N. Ladizinsky, T. Lanting, R. Li, T. Med- ina, R. Molavi, R. Neufeld, T. Oh, I. Pavlov, I. Perminov, G. Poulin-Lamarre, C. Rich, A. Smirnov, L. Swenson, N. Tsai, M. Volkmann, J. Whittaker, ...

    work page 2018

  31. [31]

    Wecker, M

    D. Wecker, M. B. Hastings, N. Wiebe, B. K. Clark, C. Nayak, and M. Troyer, Physical Review A92, 062318 (2015)

    work page 2015

  32. [32]

    Jiang, K

    Z. Jiang, K. J. Sung, K. Kechedzhi, V. N. Smelyanskiy, and S. Boixo, Physical Review Applied9, 044036 (2018)

    work page 2018

  33. [33]

    Zhang, M.-H

    J. Zhang, M.-H. Yung, R. Laflamme, A. Aspuru-Guzik, and J. Baugh, Nature Communications3, 880 (2012)

    work page 2012

  34. [34]

    Born and V

    M. Born and V. Fock, Zeitschrift f¨ ur Physik51, 165 (1928)

    work page 1928

  35. [35]

    J. J. Sakurai and J. Napolitano, Modern Quantum Me- chanics (2020)

    work page 2020

  36. [36]

    Jansen, M.-B

    S. Jansen, M.-B. Ruskai, and R. Seiler, Journal of Math- ematical Physics48, 102111 (2007)

    work page 2007

  37. [37]

    Jordan and E

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

    work page 1928

  38. [38]

    Suzuki, Proceedings of the Japan Academy, Series B 69, 161 (1993)

    M. Suzuki, Proceedings of the Japan Academy, Series B 69, 161 (1993)

    work page 1993

  39. [39]

    H. F. Trotter, Proceedings of the American Mathematical Society10, 545 (1959)

    work page 1959

  40. [40]

    Suzuki, Communications in Mathematical Physics51, 183 (1976)

    M. Suzuki, Communications in Mathematical Physics51, 183 (1976)

    work page 1976

  41. [41]

    Huyghebaert and H

    J. Huyghebaert and H. D. Raedt, Journal of Physics A: Mathematical and General23, 5777 (1990)

    work page 1990

  42. [42]

    Suzuki, Journal of Mathematical Physics26, 601 (1985)

    M. Suzuki, Journal of Mathematical Physics26, 601 (1985)

    work page 1985

  43. [43]

    De Raedt, Computer Physics Reports7, 1 (1987)

    H. De Raedt, Computer Physics Reports7, 1 (1987)

    work page 1987

  44. [44]

    De Raedt, K

    K. De Raedt, K. Michielsen, H. De Raedt, B. Trieu, G. Arnold, M. Richter, T. Lippert, H. Watanabe, and N. Ito, Computer Physics Communications176, 121 (2007)

    work page 2007

  45. [45]

    De Raedt, F

    H. De Raedt, F. Jin, D. Willsch, M. Willsch, N. Yoshioka, N. Ito, S. Yuan, and K. Michielsen, Computer Physics Communications237, 47 (2019)

    work page 2019

  46. [46]

    Kesselheim, A

    S. Kesselheim, A. Herten, K. Krajsek, J. Ebert, J. Jitsev, M. Cherti, M. Langguth, B. Gong, S. Stadtler, A. Mozaf- fari, G. Cavallaro, R. Sedona, A. Schug, A. Strube, R. Kamath, M. G. Schultz, M. Riedel, and T. Lippert, JUWELS Booster – A Supercomputer for Large-Scale AI Research (2021), arXiv:2108.11976 [cs]

    work page arXiv 2021

  47. [47]

    Morita and H

    S. Morita and H. Nishimori, Journal of Mathematical Physics49, 125210 (2008)

    work page 2008

  48. [48]

    Esslinger, Annual Review of Condensed Matter Physics1, 129 (2010)

    T. Esslinger, Annual Review of Condensed Matter Physics1, 129 (2010)

    work page 2010

  49. [49]

    Hensgens, T

    T. Hensgens, T. Fujita, L. Janssen, X. Li, C. J. Van Diepen, C. Reichl, W. Wegscheider, S. Das Sarma, and L. M. K. Vandersypen, Nature548, 70 (2017)

    work page 2017

  50. [50]

    Tarruell and L

    L. Tarruell and L. Sanchez-Palencia, Comptes Rendus. Physique19, 365 (2018)

    work page 2018

  51. [51]

    B. W. Reichardt, inProceedings of the thirty-sixth an- nual ACM symposium on Theory of computing, STOC ’04 (Association for Computing Machinery, New York, NY, USA, 2004) pp. 502–510

    work page 2004