Matrix Product Evolution: A Method for Simulating Quantum Circuits Using Tensor Networks

Haruyuki Kawabe; Minoru Nagai; Synge Todo; Tsuyoshi Okubo

REVIEW 3 major objections 5 minor 53 references

Matrix Product Evolution: A Method for Simulating Quantum Circuits Using Tensor Networks

T0 review · 3 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read The paper introduces Matrix Product Evolution (MPE), a circuit simulation tensor train organized along circuit depth, and claims post-selection reduces its truncation error relative to MPS-based simulation.

desk verdict A competent methods paper with a useful temporal contraction idea, but the post-selection advantage is not established because the MPS baseline's conditioning protocol is unspecified. read the letter →

arxiv 2608.03472 v1 pith:AIPJAH6N submitted 2026-08-04 quant-ph

Haruyuki Kawabe , Minoru Nagai , Tsuyoshi Okubo , Synge Todo This is my paper

classification quant-ph

keywords matrixproductevolutiontensornetworksquantumcircuitsimulationpost-selectiontemporalentanglementbonddimensiontruncationzip-upcontractionIsingtime

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 introduces Matrix Product Evolution (MPE), a tensor-train representation of a quantum circuit built along the time direction rather than along the qubit index. It treats circuit simulation as the contraction of multiple temporal tensor trains, using a zip-up procedure that canonicalizes and truncates temporal bonds. The central claim is that this reorientation changes how post-selection interacts with approximation: fixing output qubits reduces the temporal bond-dimension bounds, and therefore lowers truncation error at a fixed bond dimension. In 14-qubit Ising-Trotter circuits with 12 of 14 output qubits fixed, the paper reports consistently lower infidelity for MPE than for MPS at equal maximum bond dimension, while in random circuits with product initial states MPS remains more accurate. The paper positions MPE as a complementary contraction order whose performance depends on the temporal singular-value structure of the circuit.

What carries the argument

The central object is the MPE: a tensor train arranged along the temporal direction, where temporal bonds encode correlations between earlier and later parts of the circuit evolution. Three mechanisms carry the argument: exact temporal compression, which packs nontrivial local blocks toward earlier time steps and merges them only when the spatial bond dimension stays at most four; the zip-up contraction, a backward LQ canonicalization followed by a forward SVD sweep that truncates temporal bonds to a maximum chi after each spatial merge; and the post-selection rank bound, which replaces a factor 2^j by 2^{j-p} when p of j merged qubits have fixed outcomes, thereby reducing the effective degr

What would settle it

Run the same 14-qubit Ising circuits at a fixed bond dimension near 100, count the expected number of circuit preparations needed to observe the fixed 12-qubit outcome, and implement the MPS baseline with post-selection applied only at the final projection rather than during evolution. If MPS then matches or exceeds MPE in per-shot fidelity normalized by sampling cost, the claimed post-selection advantage is an artifact of the conditioning protocol; if MPE still wins, the temporal-bond reduction is structural.

Watch

Extended reading notes

Core claim

The paper establishes MPE as a tensor train indexed by circuit depth: each row of the circuit tensor network is treated as the evolution history of a qubit or small subsystem, and neighboring rows are merged by a zip-up contraction. The key structural result is that post-selection directly shrinks temporal rank bounds. After j qubits are merged and p of them are fixed, the bond-dimension bound for a product initial state changes from min{4^k, 4^{d-k-1} 2^j, 2^j} to min{4^k, 4^{d-k-1} 2^{j-p}, 2^j}; for a highly entangled initial state the relevant bound becomes min{4^{d-k-1} 2^{j-p}, 2^j}. When that term dominates, post-selection reduces the bond dimension by 2^{-p}. The paper then reports n

Load-bearing premise

The load-bearing premise is that comparing fidelities inside the subspace selected by fixing 12 of 14 output qubits is a fair benchmark, even though reaching that subspace in a random circuit requires sampling about 2^12 times, and the MPS baseline's handling of post-selection is not specified.

Editorial extensions

If this is right

Post-selection reduces the MPE truncation error at a fixed bond dimension: fixing p of j merged qubits shrinks the temporal rank bound by 2^{-p} when that term dominates, so fewer singular values need to be kept.
MPE is a complement to MPS, not a universal replacement: in random circuits with product initial states and no post-selection, MPS truncation error grows more slowly than MPE.
The accuracy of MPE is governed by temporal singular-value decay rather than spatial entanglement: for Ising Trotter circuits with Rx angles near multiples of pi, temporal compression leaves nearly one time slice and MPE is almost exact even at small chi.
The method has controlled cost: zip-up contraction of an MPE of depth d costs O(d chi^3), merging n MPEs costs O(n d chi^3), and memory scales as O(d chi^2).
The post-selection effect is structural: fixing final measurement outcomes removes temporal degrees of freedom in MPE, while it does not directly constrain the spatial bonds of an MPS, explaining the observed accuracy gap in post-selected settings.

Reading between the lines

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

Beyond the paper, MPE should transfer naturally to circuits with mid-circuit measurements: each classical outcome can be treated as an additional post-selection, shrinking temporal bonds at the point of measurement; the paper names this as future work, and the rank-bound mechanism gives a concrete reason to expect it.
A hybrid contraction order that starts spatially in low-entanglement regions and switches to temporal MPE near post-selected output qubits could inherit advantages of both; this follows from the paper's own view of MPE as one admissible contraction order of a common two-dimensional tensor network.
A testable extension is to benchmark MPE on measurement-based resource states, where temporal bonds represent correlation history; if the post-selection bound is the operative effect, fixed-chi fidelity should improve with the number of measured qubits.
The sampling cost of finding the post-selected outcome is not included in the paper's fidelity comparison; an end-to-end accounting that multiplies per-shot error by the expected number of shots would separate a structural advantage from a conditioning artifact.

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.

Desk Editor's Note

read the letter

Colleague,

The paper is worth a look for anyone working on tensor-network simulation of circuits. The core idea is to organize the circuit tensor network along the time axis rather than the qubit axis, represent each qubit row as a tensor train (MPE), and then contract rows spatially with a zip-up procedure. That is a clean, workable construction, and the authors are honest that MPE is not a general improvement over MPS: without post-selection, MPS wins (Figs. 5, 7). The interesting claim is that post-selection lowers the temporal bond-dimension bound by a factor 2^{-p} in MPE, giving an accuracy advantage in the post-selected, highly entangled regime (Figs. 8, 12). If that is robust, it is a genuine new tool for measurement-based or post-selected circuits.

What is good: the representation is clearly explained, the temporal compression is exact, the zip-up contraction is standard enough to follow, and the authors do not oversell. They explicitly call MPE complementary and point out where it fails (flat singular-value spectra).

The soft spots, in decreasing seriousness. First, the numerical comparison for post-selection is under-specified. The paper never states how the MPS baseline imposes the post-selection: does it project at the very end of the evolution, or does it condition during the sweep? That matters. If the MPS baseline only projects at the end, all truncations are made without knowledge of the conditioning, so the comparison stacks the deck in MPE's favor. It could shrink or vanish with a matched MPS conditioning protocol. The probability of the chosen outcomes (~2^-12) also is not discussed; rare outcomes are likely to have smaller norms and different error properties. Second, the bond-dimension bounds in Sec. II E/F are asserted with a short plausibility argument, not derived; the paper needs a rigorous derivation or at least a careful statement of assumptions before the numbers should be trusted. Third, the zip-up SVD singular values are called "Schmidt coefficients of the corresponding bipartition" - that is only exactly true before any truncation; after truncation they are the Schmidt coefficients of the approximate state, which is standard but should not be presented as exact optimality. Finally, the relation to earlier time-axis tensor-train and influence-matrix constructions (refs 28-30) is mentioned only in passing; the authors should say explicitly what is new relative to those.

My verdict is conditional: the framework is plausible and the structural post-selection bound is a nice observation, but the central practical claim currently rests on a comparison whose protocol is ambiguous. I would send this to a serious referee because the idea is of interest and the weaknesses are fixable; I would ask for a matched-control experiment and a rigorous derivation of the bounds. I would not cite it for the post-selection advantage until that is done.

Recommendation: engage with it in review, not as a settled result.

Referee Report

3 major / 5 minor

Summary. The paper introduces Matrix Product Evolution (MPE), a tensor-train representation of a quantum circuit organized along the temporal (depth) direction rather than the spatial (qubit) direction. Circuits are encoded as a rectangular tensor network, pre-compressed in time by relocating local tensor blocks, and then contracted along the spatial direction using a zip-up procedure with SVD truncation to a maximum bond dimension χ. The authors derive structural bounds on the temporal bond dimensions, including a post-selection-modified bound, and present numerical comparisons against MPS-based simulation for 14-qubit random circuits and Trotterized Ising-model time evolution. They report that MPE is generally less accurate than MPS for random circuits, but that under post-selection on all but two qubits MPE consistently achieves lower infidelity for both random circuits with highly entangled initial states and Ising evolution circuits.

Significance. If the central claims hold, the paper makes a useful contribution by identifying a genuinely different contraction order for circuit simulation and by showing analytically and numerically that post-selection can reduce temporal bond-dimension growth in that order. The structural bounds in Sec. II E/F are parameter-free statements about the representation, and the numerical claims are benchmarked against exact state-vector fidelity rather than against the method's own fit quantities. The paper is also appropriately cautious: it repeatedly states that MPE is a complement to, not a replacement for, MPS simulation, and it explicitly identifies regimes where MPE fails. However, the headline post-selection advantage rests on a comparison whose MPS baseline is not fully specified. Because the paper's main numerical conclusion is that MPE outperforms MPS specifically in post-selected settings, the absence of a matched post-selection protocol in the MPS baseline is a load-bearing issue that must be resolved before the claim can be accepted.

major comments (3)

[Sec. III A/B, Figs. 8 and 12] The central numerical claim—that MPE yields lower infidelity than MPS under post-selection—depends critically on how the MPS baseline imposes post-selection. The manuscript never states whether the MPS simulation applies the projection operators only at the end of the evolution, or whether it interleaves projections at the output and recompresses to the same χ, or conditions the evolution in some other way. If the MPS baseline evolves the full state and projects only at the final step, then post-selection cannot reduce the spatial bond dimensions that were truncated during the evolution, while MPE builds the outcome constraints into the temporal contraction. The observed gap would then reflect an asymmetry in the conditioning protocol, not a structural advantage of the MPE representation. Please specify the exact MPS post-selection algorithm, and ideally add a matched control experiment
[Sec. II F] The post-selection bond-dimension bounds, e.g. min{4^k, 4^{d-k-1} 2^{j-p}, 2^j}, are asserted rather than derived. For a claim that is central to the paper's main advantage, the derivation should be explicit. In particular, the text should define precisely what k, j, and p index at the point of the bound, explain how fixing the final outcomes of p of the j merged qubits reduces the second term from 4^{d-k-1}2^j to 4^{d-k-1}2^{j-p}, and state the conditions under which this term actually determines the minimum. As written, the reduction only holds 'whenever this term determines the minimum,' but the numerical sections assert a broad accuracy improvement without verifying that this condition is met for the circuits and depths used. Please either prove the bound step by step or add numerical verification that the relevant term is the active one in Figs. 8 and 12.
[Sec. II D] The claim that 'the singular values obtained from each SVD coincide with the Schmidt coefficients of the corresponding bipartition' and that truncation 'yields the optimal low-rank approximation for the corresponding bipartition' is only exactly true for a given canonical tensor train before any truncation has been performed. In a zip-up sweep, after a truncation at an earlier bond, the tensor train has been modified, so subsequent SVDs are taken with respect to the current, already approximated tensor train, not the exact MPE. The optimality is therefore local to each step in the sweep, not global for the full contraction. This is standard in MPS/DMRG practice, but the manuscript should state that the truncation is quasi-optimal in the sweep sense, and should not present the equality with Schmidt coefficients of the original bipartition without qualification.

minor comments (5)

[Sec. II E] The computational cost O(d χ^3) assumes that local tensor dimensions are of order χ. Please spell out how the MPO bond dimension (bounded by 4) and the spatial open indices enter the prefactor, and whether the estimate holds during the intermediate stages when multiple qubits have been merged.
[Fig. 3] The symbols U, V, S, L, Q are used for several different objects across panels (1)–(14). The caption would be easier to follow if the roles of U and U′, V and V′, and the propagation directions were stated explicitly.
[Sec. III A] For the random circuits, the paper reports results for depths 6–14 only. Please state whether the effective depth after temporal compression is used when plotting against 'circuit depth,' since the preprocessing step changes the actual depth entering the MPE contraction.
[Sec. III B] The heat map in Fig. 12 would benefit from a color scale that saturates explicitly at machine precision; several entries appear as -16.00, making it hard to distinguish exact results from values that are merely close to zero.
[General] The manuscript would be strengthened by a short pseudocode or algorithm box for the zip-up contraction and for the post-selection handling in both MPE and MPS, since the current prose description leaves room for ambiguity.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the MPE bond-dimension bounds are analytic counting statements, and the numerical claims are benchmarked against exact state-vector fidelity, not re-fitted inputs.

full rationale

The paper's central claims are: (i) a tensor-train representation along circuit depth whose temporal bond dimensions obey the analytic bounds min{4^k, 4^{d-k-1} 2^j, 2^j}, with post-selection replacing 2^j by 2^{j-p}; (ii) a zip-up contraction with SVD truncation; and (iii) numerical comparisons against exact state-vector fidelity. The bounds are derived from the local tensor dimensions and from fixing boundary indices; they are not fitted to the simulation outputs and do not define the reported infidelities. The numerical results are measured as 1 - |<φ|φ0>|^2 against an independently obtained exact state, so the observed post-selection improvement is a benchmarked observation rather than a quantity that the method's own construction forces by definition. The paper also explicitly disclaims a universal advantage over MPS, stating that the effect 'reflects a structural feature of the MPE-based representation rather than a general advantage over MPS-based methods.' There is no load-bearing self-citation: the cited zip-up and MPO techniques are standard external references, and no uniqueness theorem is imported from the authors' prior work. The strongest concern—that the MPS baseline's post-selection protocol is not fully specified—is a potential experimental-design or correctness issue, but it is not circularity: it does not make any prediction equal to its inputs by construction. No circular step can be exhibited from the paper's own equations or citations, so the appropriate score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

The central claims rest on the truncation parameter chi, the compression threshold, and the extreme post-selection scenario; the axioms are standard tensor-network domain assumptions plus the approximate validity of local truncation after earlier truncations. No physical entities are introduced; MPE is an algorithmic construct.

free parameters (3)

Maximum temporal bond dimension chi = 40, 100, 120
Truncation threshold for SVDs in the zip-up sweeps; the main accuracy/cost control and the parameter swept in all comparisons.
Temporal-compression merge threshold = 4
Merges in temporal compression are accepted only when the resulting spatial bond dimension is at most 4 (Sec. II C); chosen because circuit MPO bond dimensions are at most 4.
Number of post-selected qubits p = 12 of 14 (all but boundary qubits)
All post-selection experiments fix outcomes on all but the two boundary qubits; this extreme conditioning drives the reported MPE accuracy gains and is chosen by the authors rather than derived.

assumptions (4)

domain assumption SVD truncation of each local bipartition is a good global approximation during zip-up, because the singular values are claimed to equal the Schmidt coefficients of that bipartition (Sec. II D).
Valid for the exact canonical form before any truncation; after the first truncation the network is approximate, so the optimality statement is an assumption used to justify the accuracy claims.
domain assumption Gates can be encoded as MPOs of bond dimension at most 4, so all spatial bond dimensions inside local tensor blocks are bounded by 4 (Sec. II C).
Holds for the nearest-neighbor two-qubit gates used here; limits the generality of the compression criterion to such gates.
domain assumption Randomly generated two-qubit gates, produced by QR-decomposing matrices with uniform complex entries, are representative random circuit instances (Sec. III A).
The observed singular-value spectra and the accuracy comparisons depend on this gate distribution.
domain assumption Highly entangled initial states can be represented by an MPS truncated to a fixed bond dimension and renormalized, and this faithfully mimics a saturated MPS simulation (Sec. III A).
The truncation error of the initial state is inherited by both methods and is not reported separately.

invented entities (1)

Matrix Product Evolution (MPE) tensor-train representation
purpose: Organizes the circuit tensor network along the temporal (depth) direction so that conditioning information such as post-selection can reduce internal bond dimensions.
MPE is a new formal object, not a physical entity. Its value is assessed only through the paper's numerical benchmarks; there is no external falsifiable handle beyond the algorithm's performance, so independent evidence is not provided within the paper.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Matrix Product Evolution: A Method for Simulating Quantum Circuits Using Tensor Networks." pith.science (2026). https://pith.science/paper/AIPJAH6N

@misc{pith2026260803472,
  author       = {Pith},
  title        = {Pith review of: Matrix Product Evolution: A Method for Simulating Quantum Circuits Using Tensor Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AIPJAH6N}},
  note         = {Machine review of arXiv:2608.03472}
}

read the original abstract

Classical simulation of quantum circuits is an essential tool in quantum information science, but its applicability is constrained by the exponential growth of the Hilbert space and the entanglement structure of quantum states. In this work, we introduce Matrix Product Evolution (MPE), a tensor-train representation of quantum circuits constructed along the circuit depth rather than along the qubit index. Within this formulation, the simulation of a quantum circuit is modeled as the contraction of multiple MPE tensors. We develop an efficient contraction strategy based on a zip-up procedure to carry out this contraction in practice. We investigate the numerical behavior of this MPE-based contraction framework through simulations of random quantum circuits and the time evolution of a quantum many-body state. Our results characterize the growth of temporal bond dimensions, clarify how post-selection modifies the contraction cost and approximation accuracy, and identify regimes in which depth-oriented tensor-network contractions provide a useful complement to standard MPS-based simulation approaches.

Figures

Figures reproduced from arXiv: 2608.03472 by the authors.

**Figure 1.** FIG. 1. Overview of the tensor-network representation and contraction strategy underlying the MPE-based simulation for [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗

**Figure 2.** FIG. 2. Procedure of the temporal compression. (1) Each [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** FIG. 3. Procedure for zipping up two MPEs. The process consists of three parts. In (1)–(7), tensors in adjacent MPEs are [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (7 more)

**Figure 4.** Figure 4: FIG. 4. Example of a quantum circuit composed of randomly [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

**Figure 5.** Figure 5: FIG. 5. Approximation errors of the MPS- and MPE-based [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

**Figure 7.** Figure 7: FIG. 7. Approximation errors of the MPS- and MPE-based [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

**Figure 8.** Figure 8: FIG. 8. Approximation errors of the MPS- and MPE-based [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]

**Figure 9.** Figure 9: FIG. 9. A single Trotter step of the quantum circuit used to [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]

**Figure 10.** Figure 10: FIG. 10. Approximation errors of the MPS- and MPE-based [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]

**Figure 12.** Figure 12: FIG. 12. Heat map of the infidelity (on a base-10 logarithmic [PITH_FULL_IMAGE:figures/full_fig_p008_12.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

53 extracted references · 24 canonical work pages

[1]

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

1982
[2]

Lloyd, Science273, 1073 (1996)

S. Lloyd, Science273, 1073 (1996). 10

work page 1996
[3]

M. A. Nielsen and I. L. Chuang,Quantum Computation and Quantum Information(Cambridge University Press, 2000)

2000
[4]

Preskill, Quantum2, 79 (2018)

J. Preskill, Quantum2, 79 (2018)

2018
[5]

A. W. Harrow and A. Montanaro, Nature549, 203 (2017)

work page 2017
[6]

Vidal, Physical Review Letters91, 147902 (2003)

G. Vidal, Physical Review Letters91, 147902 (2003)

work page 2003
[7]

Vidal, Physical Review Letters93, 040502 (2004)

G. Vidal, Physical Review Letters93, 040502 (2004)

work page 2004
[8]

Vidal, Physical Review Letters98, 070201 (2007)

G. Vidal, Physical Review Letters98, 070201 (2007)

2007

Show all 53 references

[9]

Or´ us and G

R. Or´ us and G. Vidal, Physical Review B78, 155117 (2008)

2008
[10]

S. R. White, Physical Review Letters69, 2863 (1992)

1992
[11]

S. R. White, Physical Review B48, 10345 (1993)

1993
[12]

¨Ostlund and S

S. ¨Ostlund and S. Rommer, Physical Review Letters75, 3537 (1995)

1995
[13]

Rommer and S

S. Rommer and S. ¨Ostlund, Physical Review B55, 2164 (1997)

1997
[14]

Schollw¨ ock, Reviews of Modern Physics77, 259 (2005)

U. Schollw¨ ock, Reviews of Modern Physics77, 259 (2005)

2005
[15]

Schollw¨ ock, Annals of Physics326, 96 (2011)

U. Schollw¨ ock, Annals of Physics326, 96 (2011)

2011
[16]

Verstraete, V

F. Verstraete, V. Murg, and J. I. Cirac, Advances in Physics57, 143 (2008)

2008
[17]

P´ erez-Garc ´ ıa, F

D. P´ erez-Garc ´ ıa, F. Verstraete, M. M. Wolf, and J. I. Cirac, Quantum Information and Computation7, 401 (2007)

2007
[18]

Or´ us, Annals of Physics349, 117 (2014)

R. Or´ us, Annals of Physics349, 117 (2014)

2014
[19]

Or´ us, Nature Reviews Physics1, 538 (2019)

R. Or´ us, Nature Reviews Physics1, 538 (2019)

2019
[20]

Zwolak and G

M. Zwolak and G. Vidal, Physical Review Letters93, 207205 (2004)

2004
[21]

Paeckel, T

S. Paeckel, T. K¨ ohler, A. Swoboda, S. R. Manmana, U. Schollw¨ ock, and C. Hubig, Annals of Physics411, 167998 (2019)

2019
[22]

Haegeman, C

J. Haegeman, C. Lubich, I. Oseledets, B. Vandereycken, and F. Verstraete, Physical Review B94, 165116 (2016)

2016
[23]

I. L. Markov and Y. Shi, SIAM Journal on Computing 38, 963 (2008)

2008
[24]

Y. Shi, L. Duan, and G. Vidal, Physical Review A74, 022320 (2006)

2006
[25]

M. J. Bremner, R. Jozsa, and D. J. Shepherd, Proceed- ings of the Royal Society A467, 459 (2011)

2011
[26]

Bravyi and D

S. Bravyi and D. Gosset, Physical Review Letters116, 250501 (2016)

2016
[27]

Pednault, J

E. Pednault, J. A. Gunnels, G. Nannicini, L. Horesh, and R. Wisnieff, arXiv (2019), 1910.09534

2019 arXiv
[28]

Lerose, M

A. Lerose, M. Sonner, and D. A. Abanin, Phys. Rev. X 11, 021040 (2021)

2021
[29]

Giudice, G

G. Giudice, G. Giudici, M. Sonner, J. Thoenniss, A. Lerose, D. A. Abanin, and L. Piroli, Phys. Rev. Lett. 128, 220401 (2022)

2022
[30]

Foligno, T

A. Foligno, T. Zhou, and B. Bertini, Phys. Rev. X13, 041008 (2023)

2023
[31]

I. V. Oseledets, SIAM Journal on Scientific Computing 33, 2295 (2011)

2011
[32]

Fishman, S

M. Fishman, S. R. White, and E. M. Stoudenmire, Sci- Post Physics Codebases4(2022)

2022
[33]

Gray, Journal of Open Source Software3, 819 (2018)

J. Gray, Journal of Open Source Software3, 819 (2018)

2018
[34]

S. V. Isakov, D. Kafri, O. Martin, C. V. Heid- weiller, W. Mruczkiewicz, M. P. Harrigan, N. C. Ru- bin, R. Thomson, M. Broughton, K. Kissell, D. Garcia, J. R. McClean, H. Neven, and R. Babbush, arXiv (2021), 2111.02396

2021 arXiv
[35]

E. M. Stoudenmire and S. R. White, New Journal of Physics12, 055026 (2010)

2010
[36]

Bayraktaret al., in2023 IEEE International Con- ference on Quantum Computing and Engineering (QCE) (2023) pp

H. Bayraktaret al., in2023 IEEE International Con- ference on Quantum Computing and Engineering (QCE) (2023) pp. 1050–1061

2023
[37]

Aaronson, Proceedings of the Royal Society A461, 3473 (2005)

S. Aaronson, Proceedings of the Royal Society A461, 3473 (2005)

2005
[38]

Raussendorf and H

R. Raussendorf and H. J. Briegel, Physical Review Let- ters86, 5188 (2001)

2001
[39]

H. J. Briegel and R. Raussendorf, Physical Review Let- ters86, 910 (2001)

2001
[40]

M. A. Nielsen, Reports on Mathematical Physics57, 147 (2006)

2006
[41]

D. E. Browne and H. J. Briegel, inLectures on Quantum Information(Wiley-VCH, 2006)

2006
[42]

Jozsa, inQuantum Information Processing, NATO Science Series, III: Computer and Systems Sciences, Vol

R. Jozsa, inQuantum Information Processing, NATO Science Series, III: Computer and Systems Sciences, Vol. 199 (IOS Press, 2006) pp. 137–158

2006
[43]

Boixo, S

S. Boixo, S. V. Isakov, V. N. Smelyanskiy, R. Babbush, N. Ding, Z. Jiang, M. J. Bremner, J. M. Martinis, and H. Neven, Nature Physics14, 595 (2018)

2018
[44]

Aruteet al., Nature574, 505 (2019)

F. Aruteet al., Nature574, 505 (2019)

2019
[45]

I. L. Markov, A. Fatima, S. V. Isakov, and S. Boixo, arXiv (2018), 1807.10749

2018 arXiv
[46]

Villalonga, S

B. Villalonga, S. Boixo, B. Nelson, C. Henze, E. Rieffel, R. Biswas, and S. Mandr` a, npj Quantum Information5, 86 (2019)

2019
[47]

Huang, F

C.-N. Huang, F. Zhang, M. Newman, J. Cai, X. Gao, Z. Tian, J. Wu, H. Yu, B. Yuan, M. Szegedy, Y. Shi, and J. Chen, arXiv (2020), 2005.06787

2020 arXiv
[48]

Pan and P

F. Pan and P. Zhang, Physical Review Letters128, 030501 (2022)

2022
[49]

H¨ aner and D

T. H¨ aner and D. S. Steiger, inProceedings of SC17 (2017)

2017
[50]

Smelyanskiy, N

M. Smelyanskiy, N. P. Sawaya, and A. Aspuru-Guzik, arXiv (2016), 1601.07195

2016 arXiv
[51]

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

1959
[52]

Suzuki, Journal of Mathematical Physics32, 400 (1991)

M. Suzuki, Journal of Mathematical Physics32, 400 (1991)

1991
[53]

A. Li, S. Stein, S. Krishnamoorthy, J. R. Wendt, M. P. Frank, and M. Martonosi, ACM Transactions on Quan- tum Computing1, 8:1 (2020)

2020

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

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

1982

[2] [2]

Lloyd, Science273, 1073 (1996)

S. Lloyd, Science273, 1073 (1996). 10

work page 1996

[3] [3]

M. A. Nielsen and I. L. Chuang,Quantum Computation and Quantum Information(Cambridge University Press, 2000)

2000

[4] [4]

Preskill, Quantum2, 79 (2018)

J. Preskill, Quantum2, 79 (2018)

2018

[5] [5]

A. W. Harrow and A. Montanaro, Nature549, 203 (2017)

work page 2017

[6] [6]

Vidal, Physical Review Letters91, 147902 (2003)

G. Vidal, Physical Review Letters91, 147902 (2003)

work page 2003

[7] [7]

Vidal, Physical Review Letters93, 040502 (2004)

G. Vidal, Physical Review Letters93, 040502 (2004)

work page 2004

[8] [8]

Vidal, Physical Review Letters98, 070201 (2007)

G. Vidal, Physical Review Letters98, 070201 (2007)

2007