Pith. sign in

REVIEW 3 major objections 4 minor 95 references

Nonequilibrium relaxation exponentially delays the onset of quantum diffusion

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Nonequilibrium polaron relaxation reaches diffusion only exponentially slowly in time and system size.

desk verdict The exponential finite-size scaling the paper sells is an artifact of its own STL-GME propagator; the useful parts are the method itself and the 1D/2D transport comparison. read the letter →

arxiv 2411.17021 v1 pith:IGM5N5GY submitted 2024-11-26 physics.chem-ph cond-mat.mtrl-sciquant-ph

Srijan Bhattacharyya , Thomas Sayer , Andrés Montoya-Castillo This is my paper · ORCID
classification physics.chem-phcond-mat.mtrl-sciquant-ph
keywords dispersiveHolsteinmodelpolarontransportspace-andtime-localGMEnonequilibriumrelaxationsubdiffusionthermodynamiclimithierarchicalequationsofmotion
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

Polarons—charges dragging a local lattice deformation behind them—are central to energy transport in many materials. This paper claims that after a sudden excitation, a polaron in the dispersive Holstein model does not settle into ordinary diffusive motion until exponentially large times and system sizes: the transport exponent $\alpha$ in $\mathrm{MSD}\propto t^{\alpha}$ approaches 1 only asymptotically, with $1-\alpha_{\max}$ falling linearly on a log scale as the lattice grows to 1500 sites and the dynamics run to 1.5 ns. To see this, the authors use a generalized master equation whose memory kernel is local in both time and space, so that a short, small-lattice numerically exact simulation can be zero-padded into a thermodynamically large one. They also find that dilute-limit transport in 1D quantitatively determines transport in 2D after dimensional renormalization, and that anisotropic 2D energy landscapes localize polaron motion along specific directions. If the claims hold, finite-size simulations cannot be casually extrapolated to the diffusive regime, and experimentally observed polaron motion may be probing a very long transient rather than asymptotic transport.

What carries the argument

The load-bearing object is the space- and time-local generalized master equation (STL-GME), a generator $U(r,t)$ that updates site populations according to $C(r,t+\delta t)=\int dr'\, U(r-r',t)\,C(r',t)$, where $r$ is the separation between the initial injection site and the measurement site. The generator is taken to be finite in time—with a lifetime $\tau_U=90$ fs—and finite in space—with a cutoff $d_U=4r_0$—both measured against a 10-site hierarchical equations of motion (HEOM) reference. Because $d_U<N/2$, the small-lattice generator contains no finite-size contamination, and translational invariance lets one zero-pad it to 1500 sites in 1D and $30\times30$ sites in 2D; after $\tau_U$ the generator is a constant matrix, so propagation beyond that time is simple repeated matrix multiplication. The spatial locality is justified by a light-cone picture: by the time the memory dies, a quasiparticle moving at the characteristic speed has traveled at most $d_U\sim\tau_U V$, so transitions beyond that distance can be set to zero.

What would settle it

Directly compute the memory kernel $U(r,t)$ from a larger exact reference, say 30 sites run well past $\tau_U=90$ fs, and check whether transitions beyond $r=4r_0$ or times beyond $\tau_U$ are genuinely negligible; alternatively, compare the STL-GME populations on sites farther than $d_U$ from the injection site against exact dynamics on the same lattice. If non-negligible tails appear or the error grows with propagation time, the exponential $1-\alpha_{\max}$ scaling is an artifact of the truncation.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is that $\alpha_{\max}$, the maximum of the scaling exponent $\alpha$ before finite-size reflection sets in, approaches unity exponentially slowly as a function of system size $N$ and of the finite-size reflection time $\tau_R$, for every phonon decorrelation rate $\gamma$ studied. A 10-site chain gives $\alpha_{\max}\approx0.78$ near 200 fs, a 30-site chain $\alpha_{\max}\approx0.95$ near 2 ps, and a 100-site chain $\alpha_{\max}\approx0.99$ near 20 ps; extending to 1500 sites and 1.5 ns shows $\log(1-\alpha_{\max})$ decreasing linearly, i.e., diffusion is reached only asymptotically in the thermodynamic limit. The same space-time-local GME applied to homogeneous 2D lattices up to $30\times30$ shows that the dimensionality-renormalized $d\mathrm{MSD}/dt$ and $\alpha$ agree quantitatively with the 1D results, although the full 2D density is not exactly the outer product of the 1D density; the difference is orders of magnitude smaller than the populations and decays over time. In 2D periodic lattices with two site energies, the paper finds that the relative hopping strengths direct polaron localization: a small inter-column hopping confines the polaron along the vertical axis, while a larger cross-hopping restores an elliptical spread.

Load-bearing premise

The assumption that carries the whole extrapolation is that the memory kernel's cutoffs—$d_U=4r_0$ in space and $\tau_U=90$ fs in time, fixed from a 10-site, sub-picosecond HEOM reference—stay quantitatively exact when the generator is zero-padded to 1500 sites and stepped for 1.5 ns.

Editorial extensions

If this is right

  • Nonequilibrium polaron transport in the dispersive Holstein model remains subdiffusive for times that grow exponentially with system size, so a finite-lattice simulation underestimates the difficulty of reaching diffusion.
  • Equilibrium linear-response diffusion constants describe the asymptotic limit, not the nonequilibrium relaxation: the system can appear subdiffusive on every experimentally practical timescale even though its equilibrium transport is diffusive.
  • In the dilute limit, 1D relaxation quantitatively fixes the dimensionality-renormalized $\frac{1}{2N_d}\frac{d\mathrm{MSD}}{dt}$ and $\alpha$ in 2D homogeneous lattices, so 2D transport can be built from a 1D outer product without loss of accuracy in the transport metrics.
  • Anisotropic 2D energy landscapes direct polaron flow: lowering one hopping pathway confines the polaron along one axis, while raising the cross-hopping restores an elliptical distribution—predictions accessible to spatiotemporal microscopy.
  • The STL-GME construction extends beyond polarons to other short-range lattice models, including spin relaxation, qubit crosstalk, topological magnons, and thermal conduction.

Reading between the lines

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

  • The exponential onset suggests that measured polaron mobility will depend on the observation window, so two experiments on the same material could report different effective diffusion constants simply by probing different times after excitation.
  • If the space-local memory picture is correct, other short-range carrier-phonon models with similar local couplings may also show exponentially slow onsets of diffusion in their nonequilibrium relaxation.
  • The cutoff $d_U$ could shift with temperature, reorganization energy, or phonon frequency; mapping where the exponential regime begins would give a testable boundary between practically diffusive and asymptotically diffusive behavior.
  • The predicted direction-dependent polaron densities in anisotropic 2D lattices could be tested by preparing polarons at specific sites and imaging the spread, since the two hopping arrangements give sharply different shapes (a localized stripe versus an ellipse).
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 / 4 minor

Summary. The paper presents a space- and time-local generalized master equation (STL-GME) method for simulating polaron transport in the dispersive Holstein model. The method constructs a memory kernel U(t) from small-lattice HEOM reference dynamics, truncates it in space and time, and then extends the generator to much larger lattices by zero padding and translational invariance, propagating with the fixed matrix U(τU) for long times (Eq. M2). The central claim is that the maximum of the scaling exponent α (defined through MSD ∝ t^α) approaches unity exponentially slowly as a function of lattice size, so that the onset of diffusion is delayed to exponentially large times and system sizes. The paper also reports that 1D transport, via an outer-product construction, quantitatively reproduces 2D transport coefficients, and that anisotropic energy landscapes in periodic 2D lattices direct polaron motion.

Significance. If the exponential-delay claim were correct, it would challenge the standard extrapolation of finite-lattice simulations to the thermodynamic limit and would have broad implications for the interpretation of nonequilibrium polaron transport. The STL-GME idea of exploiting joint temporal and spatial memory locality is attractive and the small-system agreement between GME and HEOM (errors below 1e-7 for τU and 5e-6 for dU) is credible. However, the central claim is not merely unsupported; it is mathematically incompatible with the paper's own propagation rule, Eq. (M2), as detailed below. The 2D transport and localization results are of interest but do not rescue the headline claim, which is the basis for the paper's stated significance.

major comments (3)
  1. [Polaron transport; Methods, Eq. (M2)] The exponential decay of 1−α_max with N is contradicted by the STL-GME propagation rule itself. For t > τU = 90 fs, the dynamics are generated by repeated application of the fixed, finite-range, translationally invariant matrix U(τU) via C(t+nδt) = [U(τU)]^n C(t=τU). Any such matrix has Fourier eigenvalues of the form λ(q) = 1 − D q^2 δt + O(q^4), where D > 0 is the diffusion constant. Consequently, on an infinite lattice MSD(t) = 2Dt + c, which gives 1−α(t) = c/(2Dt) + O(t^−2); on a ring of N sites the slowest nonconstant mode has relaxation time τ_R ∼ N^2/D, so α_max is reached at t ∼ N^2/D and 1−α_max ∼ const/N^2. This is an algebraic, not exponential, dependence on N. The authors provide no derivation of the exponential form and no independent large-N check; the log-linear behavior in Fig. 2(d)–(e) is therefore most plausibly an artifact of the 10-site reference, the zero-padding/renormalization in Methods C, or the choice of the α_max sampling window. Please provide a direct test: plot log(1−α_max) versus log N for the STL-GME with the fixed generator U(τU) and report the slope; if it is not −2, explain explicitly how the generator violates the small-q expansion.
  2. [Methods B and Figs. S1–S2] The memory cutoffs τU = 90 fs and dU = 4r0 are determined by comparing GME predictions against the same small-system HEOM reference from which U was constructed. This establishes self-consistency on the 10-site system, but it does not validate the zero-padded extension to M = 1500 sites propagated to 1.5 ns. The claim that the kernel is exactly zero beyond dU and exactly constant after τU is a strong truncation assumption; if the true kernel has slowly decaying tails, or if truncation error accumulates under repeated powers [U(τU)]^n, the extrapolated α_max(N) is uncontrolled. The paper repeatedly labels the STL-GME results as 'exact' (Abstract, Introduction, Outlook), yet no exact or independent large-N benchmark is provided. This is load-bearing because the central quantitative claim rests entirely on the extrapolated large-N dynamics.
  3. [Fig. 2(d)–(e)] The exponential fits lack error bars, fit ranges, and residuals. The text states that for γ = 400 cm^−1 the asymptotic behavior is evident for N = 600–1000, while for γ = 1000 cm^−1 it only arises for N = 1000–1500, but each curve contains only a handful of points and no statistical uncertainty is reported. Given the algebraic prediction of Eq. (M2) described above, the authors should show whether the data actually follow 1−α_max ∼ N^−2 or demonstrate, with a concrete mathematical argument, that U(τU) produces an exponentially small spectral gap that depends on N. Without such evidence, the exponential-delay claim is unsupported.
minor comments (4)
  1. [Methods D] In the paragraph on periodic 2D lattices, 'another form blue site' should read 'another from blue site'.
  2. [Fig. S5 caption] The caption begins with 'PPolaron', which should be 'Polaron'.
  3. [Fig. 2(d) caption and main text] The phrase 'logarithmic behavior of 1−αmax ... scales linearly' is ambiguous; please specify the axes of the inset and state explicitly that the fit is log(1−αmax) versus N.
  4. [Methods B and Fig. S2] The error threshold is given as 1e-7 per element for τU in Methods B, but Fig. S2 and the surrounding text state a threshold of 5e-6 per element for dU. Please clarify whether these are different thresholds and why.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the exponential-delay claim is an extrapolated simulation output, not a re-labeled fit or a self-citation chain.

full rationale

The paper's derivation chain is not circular. The STL-GME generator is defined from a 10-site HEOM reference by Eq. M1, and the cutoffs tau_U and d_U are chosen from error plateaus against that same reference (Methods B, Figs. S1-S2). This is a calibration step: the paper's headline prediction, the exponential approach of alpha_max to unity with increasing system size, is not used in the fit, and the N=30 HEOM comparison (Fig. 2c, Fig. S2) provides an external check. The large-lattice dynamics is then an output of propagating the zero-padded, translation-invariant generator by Eq. M2, so the exponential-delay statement is a computed consequence of the model plus the method, not an input redefined as an output. Self-citations [43,50] supply the STL-GME and the earlier subdiffusion observation, but the present paper re-derives the locality cutoffs against HEOM benchmarks and does not invoke an unverified uniqueness theorem to force its choice. Whether a local finite-range generator can actually produce exponential rather than algebraic finite-size scaling is a physical correctness and validation question, not a circularity: the prediction is not equivalent to its construction by definition, and no fitted parameter is renamed as the predicted scaling law.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim rests on the space-time locality of the GME kernel (delegated to the companion preprint), the Markovian freeze of the generator after tauU, and HEOM convergence parameters; the 2D results additionally rely on a high-temperature approximation and the n-particle truncation. No free parameters are fitted to the headline scaling, but tauU and dU are selected from error plateaus, and the exponential rate is fitted to the output.

free parameters (3)
  • generator lifetime tauU = 90 fs for gamma=400 cm^-1 (from error plateau)
    Chosen so the time-local GME reproduces HEOM below a 1e-7 threshold; the long-time extrapolation freezes the generator at this value.
  • characteristic memory distance dU = 4 r0 for gamma=700 cm^-1 (from error plateau)
    Chosen as the spatial cutoff beyond which generator elements are set to zero; controls the light cone and finite-size onset in the extrapolated dynamics.
  • exponential decay rate of 1-alpha_max with N = reported slopes from log-linear fits in Fig. 2(d)
    The headline claim that the onset is exponential rests on these fitted slopes; no error bars or theoretical derivation are provided.
assumptions (5)
  • ad hoc to paper The GME memory kernel U(s,t) is local in space with cutoff dU and local in time with cutoff tauU.
    Central enabling assumption; justified heuristically in 'Argument for Space-Local Memory' via a light-cone picture and delegated to the companion preprint ref [43].
  • ad hoc to paper After tauU, the generator is exactly constant, so propagation is by a fixed rate matrix [U(tauU)]^n.
    Used in Eq. M2; this Markovian freeze is what forces the long-time MSD to be affine and makes the claimed exponential scaling internally inconsistent.
  • domain assumption HEOM with convergence parameters (L=10, K=1 in 1D; L=26, K=0 in 2D) and the n-particle approximation is exact for the model.
    Methods A; HEOM is exact only in the converged limit, and the 2D simulations use the high-temperature approximation K=0 plus the n-particle truncation.
  • standard math Translational invariance of homogeneous lattices allows reconstruction of the full correlation matrix from a single reference simulation.
    Invoked in Methods D; standard for homogeneous lattices, and reasonable for the periodic 2D cases considered.
  • ad hoc to paper The 1D outer product quantitatively reproduces 2D transport coefficients.
    Claimed from one parameter set (Fig. 3, Table I in SI); the paper notes the exact 2D density differs from the outer product, so this is an empirical simplification rather than a proven identity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Nonequilibrium relaxation exponentially delays the onset of quantum diffusion." pith.science (2026). https://pith.science/paper/IGM5N5GY

@misc{pith2026241117021,
  author       = {Pith},
  title        = {Pith review of: Nonequilibrium relaxation exponentially delays the onset of quantum diffusion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IGM5N5GY}},
  note         = {Machine review of arXiv:2411.17021}
}
read the original abstract

Predicting the exact many-body quantum dynamics of polarons in materials with strong carrier-phonon interactions presents a fundamental challenge, often necessitating one to adopt approximations that sacrifice the ability to predict the transition from nonequilibrium relaxation to thermodynamic equilibrium. Here, we exploit a recent breakthrough that generalizes the concept of memory beyond its conventional temporal meaning to also encompass space. Specifically, we leverage our discovery that the dynamics of observables in systems with local couplings satisfy Green's functions with kernels that are local in time and space. This enables us to employ the dynamics of small lattices over short times to predict the dynamics of thermodynamically large lattices over arbitrarily long timescales while circumventing the deleterious impacts of finite-size effects. We thus interrogate the \textit{exact} nonequilibrium formation and migration of polarons in one- (1D) and two-dimensional (2D) systems, revealing that their motion approaches diffusive transport only asymptotically in time and system size. We also compare transport in 1D and 2D systems to investigate the effect of dimension in polaron migration physics, illustrating how energy variations can cause localization -- a phenomenon observable via current microscopy experiments.

Figures

Figures reproduced from arXiv: 2411.17021 by the authors.

Figure 1
Figure 1. FIG. 1. Schematic of the non-Markovian generator [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Dependence of the scaling exponent, [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. Transport properties of dispersive Holstein polarons in 2D with [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: FIG. 4. Polaron transport in 2D 30 [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

95 extracted references · 74 canonical work pages

  1. [1]

    Balhorn, Q

    L. Balhorn, Q. MacPherson, K. C. Bustillo, C. J. Takacs, A. J. Spakowitz, and A. Salleo, Proceedings of the Na- tional Academy of Sciences 119, e2204346119 (2022)

    2022

  2. [2]

    V. Bhat, C. P. Callaway, and C. Risko, Chemical Re- views 123, 7498 (2023)

    2023

  3. [3]

    Fratini, M

    S. Fratini, M. Nikolka, A. Salleo, G. Schweicher, and H. Sirringhaus, Nature Materials 19, 491 (2020)

    2020

  4. [4]

    D. H. L. Tjhe, X. Ren, I. E. Jacobs, G. D’Avino, T. B. E. Mustafa, T. G. Marsh, L. Zhang, Y. Fu, A. E. Man- sour, A. Opitz, Y. Huang, W. Zhu, A. H. Unal, S. Hoek, V. Lemaur, C. Quarti, Q. He, J.-K. Lee, I. McCul- loch, M. Heeney, N. Koch, C. P. Grey, D. Beljonne, S. Fratini, and H. Sirringhaus, Nature Materials (2024), 10.1038/s41563-024-01953-6

    work page doi:10.1038/s41563-024-01953-6 2024

  5. [5]

    Pastor, M

    E. Pastor, M. Sachs, S. Selim, J. R. Durrant, A. A. Bakulin, and A. Walsh, Nature Reviews Materials 7, 503 (2022)

    2022

  6. [6]

    Wang, X.-B

    H.-M. Wang, X.-B. Liu, S.-Q. Hu, D.-Q. Chen, Q. Chen, C. Zhang, M.-X. Guan, and S. Meng, Science Advances 9, eadg3833 (2023)

    2023

  7. [7]

    Emin, Polarons, 1st ed

    D. Emin, Polarons, 1st ed. (Cambridge University Press, 2012)

    2012

  8. [8]

    Franchini, M

    C. Franchini, M. Reticcioli, M. Setvin, and U. Diebold, Nature Reviews Materials 6, 560 (2021)

    2021

Show all 95 references
  1. [9]

    T. Zhu, J. M. Snaider, L. Yuan, and L. Huang, Annual Review of Physical Chemistry 70, 219 (2019)

    2019

  2. [10]

    N. S. Ginsberg and W. A. Tisdale, Annual Review of Physical Chemistry 71, 1 (2020)

    2020

  3. [11]

    Delor, H

    M. Delor, H. L. Weaver, Q. Yu, and N. S. Ginsberg, Nature Materials 19, 56 (2020)

    2020

  4. [12]

    S. J. Yoon, Z. Guo, P. C. Dos Santos Claro, E. V. Shevchenko, and L. Huang, ACS Nano 10, 7208 (2016)

    2016

  5. [13]

    Z. Guo, Y. Wan, M. Yang, J. Snaider, K. Zhu, and L. Huang, Science 356, 59 (2017)

    2017

  6. [14]

    Chernikov, T

    A. Chernikov, T. C. Berkelbach, H. M. Hill, A. Rigosi, Y. Li, B. Aslan, D. R. Reichman, M. S. Hybertsen, and T. F. Heinz, Physical Review Letters 113, 076802 (2014)

    2014

  7. [15]

    Holstein, Annals of Physics 8, 343 (1959)

    T. Holstein, Annals of Physics 8, 343 (1959)

    1959

  8. [16]

    W. P. Su, J. R. Schrieffer, and A. J. Heeger, Physical Review Letters 42, 1698 (1979)

    1979

  9. [17]

    Fr¨ ohlich, Advances in Physics3, 325 (1954)

    H. Fr¨ ohlich, Advances in Physics3, 325 (1954)

    1954

  10. [18]

    Grabert, Projection Operator Techniques in Nonequi- librium Statistical Mechanics, Springer Tracts in Modern Physics, Vol

    H. Grabert, Projection Operator Techniques in Nonequi- librium Statistical Mechanics, Springer Tracts in Modern Physics, Vol. 95 (Springer Berlin Heidelberg, Berlin, Hei- delberg, 1982)

    1982

  11. [19]

    Nakajima, Progress of Theoretical Physics 20, 948 (1958)

    S. Nakajima, Progress of Theoretical Physics 20, 948 (1958)

    1958

  12. [20]

    Zwanzig, The Journal of Chemical Physics 33, 1338 (1960)

    R. Zwanzig, The Journal of Chemical Physics 33, 1338 (1960)

    1960

  13. [21]

    Mori, Progress of Theoretical Physics 33, 423 (1965)

    H. Mori, Progress of Theoretical Physics 33, 423 (1965)

    1965

  14. [22]

    K. S. Schweizer, The Journal of Chemical Physics 91, 5802 (1989)

    1989

  15. [23]

    A. P. Chatterjee and R. F. Loring, The Journal of Chem- ical Physics 101, 1595 (1994)

    1994

  16. [24]

    J. C. Dyre, Physical Review Letters 58, 792 (1987)

    1987

  17. [25]

    L. M. Janssen and D. R. Reichman, Physical Review Let- ters 115, 205701 (2015)

    2015

  18. [26]

    M. K. Nandi, A. Banerjee, C. Dasgupta, and S. M. Bhat- tacharyya, Physical Review Letters 119, 265502 (2017)

    2017

  19. [27]

    J. P. Garrahan, Physica A: Statistical Mechanics and its Applications 504, 130 (2018)

    2018

  20. [28]

    C. Liu, G. Biroli, D. R. Reichman, and G. Szamel, Phys- ical Review E 104, 054606 (2021)

    2021

  21. [29]

    Bidhoodi and S

    N. Bidhoodi and S. P. Das, Journal of Statistical Me- chanics: Theory and Experiment 2023, 063301 (2023)

    2023

  22. [30]

    C. Ayaz, L. Tepper, F. N. Br¨ unig, J. Kappler, J. O. Daldrop, and R. R. Netz, Proceedings of the National Academy of Sciences 118, e2023856118 (2021)

    2021

  23. [31]

    S. Cao, A. Montoya-Castillo, W. Wang, T. E. Markland, and X. Huang, The Journal of Chemical Physics 153, 014105 (2020)

    2020

  24. [32]

    Vroylandt, L

    H. Vroylandt, L. Gouden` ege, P. Monmarch´ e, F. Pietrucci, and B. Rotenberg, Proceedings of the National Academy of Sciences 119, e2117586119 (2022)

    2022

  25. [33]

    S. Cao, Y. Qiu, M. L. Kalin, and X. Huang, The Journal of Chemical Physics 159, 134106 (2023)

    2023

  26. [34]

    A. J. Dominic, T. Sayer, S. Cao, T. E. Markland, X. Huang, and A. Montoya-Castillo, Proceedings of the National Academy of Sciences 120, e2221048120 (2023)

    2023

  27. [35]

    A. J. Dominic, S. Cao, A. Montoya-Castillo, and X. Huang, Journal of the American Chemical Society 145, 9916 (2023)

    2023

  28. [36]

    B. A. Dalton, C. Ayaz, H. Kiefer, A. Klimek, L. Tepper, and R. R. Netz, Proceedings of the National Academy of Sciences 120, e2220068120 (2023)

    2023

  29. [37]

    Maniscalco and F

    S. Maniscalco and F. Petruccione, Physical Review A 73, 012111 (2006)

    2006

  30. [38]

    Barnes, L

    E. Barnes, L. Cywi´ nski, and S. Das Sarma, Physical Review Letters 109, 140403 (2012)

    2012

  31. [39]

    De Vega and D

    I. De Vega and D. Alonso, Reviews of Modern Physics 89, 015001 (2017)

    2017

  32. [40]

    Campaioli, J

    F. Campaioli, J. H. Cole, and H. Hapuarachchi, PRX Quantum 5, 020202 (2024)

    2024

  33. [41]

    W. C. Pfalzgraff, A. Kelly, and T. E. Markland, The Journal of Physical Chemistry Letters 6, 4743 (2015)

    2015

  34. [42]

    Y. Yan, M. Xu, Y. Liu, and Q. Shi, The Journal of Chemical Physics 150, 234101 (2019)

    2019

  35. [43]

    Space-local memory in generalized master equations: Reaching the thermodynamic limit for the cost of a small lattice simulation,

    S. Bhattacharyya, T. Sayer, and A. Montoya-Castillo, “Space-local memory in generalized master equations: Reaching the thermodynamic limit for the cost of a small lattice simulation,” (2024), arxiv:2411.08598 [physics]

    2024 arXiv

  36. [44]

    G. D. Mahan, Many-Particle Physics (Springer US, Boston, MA, 2000)

    2000

  37. [45]

    Cheng and R

    Y.-C. Cheng and R. J. Silbey, The Journal of Chemical Physics 128, 114713 (2008)

    2008

  38. [46]

    M. B. Qarai, R. Ghosh, and F. C. Spano, The Journal of Physical Chemistry C 125, 24487 (2021)

    2021

  39. [47]

    Reticcioli, U

    M. Reticcioli, U. Diebold, G. Kresse, and C. Franchini, in Handbook of Materials Modeling, edited by W. Andreoni 7 and S. Yip (Springer International Publishing, Cham,

  40. [48]

    Mousavi and M

    H. Mousavi and M. Bagheri, Physica E: Low-dimensional Systems and Nanostructures 44, 1722 (2012)

    2012

  41. [49]

    Nematiaram and A

    T. Nematiaram and A. Troisi, The Journal of Chemical Physics 152, 190902 (2020)

    2020

  42. [50]

    Bhattacharyya, T

    S. Bhattacharyya, T. Sayer, and A. Montoya-Castillo, The Journal of Physical Chemistry Letters 15, 1382 (2024)

    2024

  43. [51]

    Weiss, Quantum Dissipative Systems (WORLD SCI- ENTIFIC, 2012)

    U. Weiss, Quantum Dissipative Systems (WORLD SCI- ENTIFIC, 2012)

    2012

  44. [52]

    Q. Shi, L. Chen, G. Nan, R.-X. Xu, and Y. Yan, The Journal of Chemical Physics 130, 084105 (2009)

    2009

  45. [53]

    K. Song, S. Bai, and Q. Shi, The Journal of Chemical Physics 143, 064109 (2015)

    2015

  46. [54]

    Sayer and A

    T. Sayer and A. Montoya-Castillo, The Journal of Chem- ical Physics 158, 014105 (2023)

    2023

  47. [55]

    L. M. Carneiro, S. K. Cushing, C. Liu, Y. Su, P. Yang, A. Alivisatos, and S. R. Leone, Nature Materials 16, 819 (2017)

    2017

  48. [56]

    C. S. Ahart, J. Blumberger, and K. M. Rosso, Physical Chemistry Chemical Physics 22, 10699 (2020)

    2020

  49. [57]

    H. Li, M. Guo, Z. Zhou, R. Long, and W.-H. Fang, The Journal of Physical Chemistry Letters 14, 2448 (2023)

    2023

  50. [58]

    Takahashi and R

    H. Takahashi and R. Borrelli, Journal of Chemical The- ory and Computation (2024), 10.1021/acs.jctc.4c00711

    2024 doi

  51. [59]

    Bhattacharyya, T

    S. Bhattacharyya, T. Sayer, and A. Montoya-Castillo, Chemical Science 15, 16715 (2024)

    2024

  52. [60]

    M. B. Isichenko, Reviews of Modern Physics 64, 961 (1992)

    1992

  53. [61]

    Piraud, L

    M. Piraud, L. Pezz´ e, and L. Sanchez-Palencia, New Jour- nal of Physics 15, 075007 (2013)

    2013

  54. [62]

    Manzano, C

    D. Manzano, C. Chuang, and J. Cao, New Journal of Physics 18, 043044 (2016)

    2016

  55. [63]

    J. A. Tulyagankhodjaev, P. Shih, J. Yu, J. C. Russell, D. G. Chica, M. E. Reynoso, H. Su, A. C. Stenor, X. Roy, T. C. Berkelbach, and M. Delor, Science 382, 438 (2023)

    2023

  56. [64]

    Troisi, Advanced Materials 19, 2000 (2007)

    A. Troisi, Advanced Materials 19, 2000 (2007)

    2007

  57. [65]

    Noriega, A

    R. Noriega, A. Salleo, and A. J. Spakowitz, Proceedings of the National Academy of Sciences 110, 16315 (2013)

    2013

  58. [66]

    G. M. Akselrod, P. B. Deotare, N. J. Thompson, J. Lee, W. A. Tisdale, M. A. Baldo, V. M. Menon, and V. Bulovi´ c, Nature Communications5, 3646 (2014)

    2014

  59. [67]

    Guzelturk, T

    B. Guzelturk, T. Winkler, T. W. J. Van De Goor, M. D. Smith, S. A. Bourelle, S. Feldmann, M. Trigo, S. W. Teit- elbaum, H.-G. Steinr¨ uck, G. A. De La Pena, R. Alonso- Mori, D. Zhu, T. Sato, H. I. Karunadasa, M. F. Toney, F. Deschler, and A. M. Lindenberg, Nature Materials 20,...

    2021

  60. [68]

    L. R. V. Buizza and L. M. Herz, Advanced Materials 33, 2007057 (2021)

    2021

  61. [69]

    H. Liu, A. Wang, P. Zhang, C. Ma, C. Chen, Z. Liu, Y.-Q. Zhang, B. Feng, P. Cheng, J. Zhao, L. Chen, and K. Wu, Nature Communications 14, 3690 (2023)

    2023

  62. [70]

    M. Kang, S. W. Jung, W. J. Shin, Y. Sohn, S. H. Ryu, T. K. Kim, M. Hoesch, and K. S. Kim, Nature Materials 17, 676 (2018)

    2018

  63. [71]

    Schott, U

    S. Schott, U. Chopra, V. Lemaur, A. Melnyk, Y. Olivier, R. Di Pietro, I. Romanov, R. L. Carey, X. Jiao, C. Jellett, M. Little, A. Marks, C. R. McNeill, I. McCulloch, E. R. McNellis, D. Andrienko, D. Beljonne, J. Sinova, and H. Sirringhaus, Nature Physics 15, 814 (2019)

    2019

  64. [72]

    P. Zhao, K. Linghu, Z. Li, P. Xu, R. Wang, G. Xue, Y. Jin, and H. Yu, PRX Quantum 3, 020301 (2022)

    2022

  65. [73]

    P. A. McClarty, Annual Review of Condensed Matter Physics 13, 171 (2022)

    2022

  66. [74]

    Y. Xia, D. Gaines, J. He, K. Pal, Z. Li, M. G. Kanatzidis, V. Ozoli¸ nˇ s, and C. Wolverton, Proceedings of the Na- tional Academy of Sciences 120, e2302541120 (2023)

    2023

  67. [75]

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

    U. Schollw¨ ock, Annals of Physics326, 96 (2011). 8 METHODS A. Details of Exact dynamics We have used the hierarchical equations of motion (HEOM) [52] as an exact solver to perform the dynam- ics of the dispersive Holstein system at finite temper- atures. HEOM maps the dissipa...

    2011

  68. [76]

    3) using var- ious candidate lifetimes, τ

    We generate TL-GME dynamics (Eq. 3) using var- ious candidate lifetimes, τ . Choosing a lifetime im- plies that we treat all elements of U (t) as constant for t ≥ τ . We then compute the error between the exact dynamics and those generated by the TL- GME and quantify the error...

  69. [77]

    Choos- ing a specific memory distance means we set all the elements in the generator beyond that range set to

    Similarly, we employ the generator U TL(t) and con- struct the space-local (SL) GME dynamics with dif- ferent characteristic memory distance dU . Choos- ing a specific memory distance means we set all the elements in the generator beyond that range set to

  70. [78]

    The distance at which the error minimizes we choose as our char- acteristic memory distance dU

    Similarly, we compute the error between the exact dynamics and those generated by SL-GME with different dU and plot the error as a function of the proposed cutoff distance. The distance at which the error minimizes we choose as our char- acteristic memory distance dU . See Fig...

  71. [79]

    Method 1: Redistribute the discarded elements among the remaining elements after truncation. This can be mathematically expressed as: U [i, j; d < dU ] = U [i, j; d < dU ] + 1 NdU X j U [i, j; d > dU ], ∀i, (M3) where NdU is the number of elements in U that remain after spatia...

  72. [80]

    (M4) For 1D simulations, we employed Method 1

    Method 2: Renormalize the generator U at each time step following spatial truncation: U [i, j; d < dU ] = U [i, j; d < dU ]P j U [i, j; d < dU ] , ∀i. (M4) For 1D simulations, we employed Method 1. We used Method 2 in 2D simulations due to its superior ability to conserve popu...

  73. [81]

    M1 in the dimension of the reference calculation

    Construct U TL(t) from Cref (t) using Eq. M1 in the dimension of the reference calculation

  74. [82]

    Identify the lifetime τU by following the procedure in Methods Sec. B

  75. [83]

    Identify the characteristic distance, dU using the procedure outlined in step 2 in Methods Sec. B. To do this, see below for the protocol on choosing a distance cutoff dU by truncating elements of the generator that connect sites beyond dU

  76. [84]

    Reshape the generator matrix U [N, N; t] at all times up to τU from a representation where x and y indices are collapsed onto a single index to a 5- tensor, where each dimension spans its original in- dices, UτU [N, N; t] → UτU [Nx, Ny; Nx, Ny; t]

  77. [85]

    For a homogeneous 2D lattice, consider the Nx × Ny submatrix defined by focusing on a particu- lar initial excitation position (i.e., the second in- dices in the 5-tensor), say the 2D grid origin, UτU [Nx, Ny; 0, 0; t] to implement spatial truncation

  78. [86]

    For the UτU [Nx, Ny; 0, 0; t] submatrix, calculate the dis- tance d between the initial point (0, 0) and all sites (i, j) where carriers are measured as d = p (i2 + j2) ∀ i, j

    Implement the spatial truncation based on the characteristic distance, dU . For the UτU [Nx, Ny; 0, 0; t] submatrix, calculate the dis- tance d between the initial point (0, 0) and all sites (i, j) where carriers are measured as d = p (i2 + j2) ∀ i, j. (M6) Then, set all eleme...

  79. [87]

    M4 to the generator elements UτU [Nx, Ny; 0, 0; t] to build ˜UτU [Nx, Ny; 0, 0; t]

    Apply the norm-conserving renormalization scheme in Eq. M4 to the generator elements UτU [Nx, Ny; 0, 0; t] to build ˜UτU [Nx, Ny; 0, 0; t]

  80. [88]

    Exploit translational symmetry to populate the modified generator ˜UτU [Nx, Ny; Nx, Ny; t] by iter- ating steps 4-5 over all initial conditions, ( k, l) in ˜UτU [Nx, Ny; k, l; t]

  81. [89]

    Reshape the 5-tensor generator after spatial trun- cation into matrix ˜UτU [N, N, t] by collapsing the x and y indices of the initial and final conditions to a compound single index as dictated by Eq. M5

  82. [90]

    As in the 1D case, one can use Eq

    Employ ˜U [N, N; t] to generate 2D-GME dynamics. As in the 1D case, one can use Eq. 3 for time t ≤ τU and Eq. M2 for t > τU to propagate CGME(t)

  83. [91]

    B for a range of proposed distance cutoffs

    Calculate the error between CGME(t) and Cref (t) using the procedure in Methods Sec. B for a range of proposed distance cutoffs

  84. [92]

    S2 in the supplementary information)

    Plot the ||L||2 error as a function of the distance cutoff to identify the characteristic memory dis- tance, dU , where the error converges to a predefined threshold (similar to Fig. S2 in the supplementary information). Following these previous steps, one can construct CGME(t...

  85. [93]

    Start with the norm-conserving generator in reshaped space, ˜U [Nx, Ny; 0, 0; t] for all times t ∈ [0, τU ] 10 (see step 7 of the previous protocol)

    Augment the dimension of the generator. Start with the norm-conserving generator in reshaped space, ˜U [Nx, Ny; 0, 0; t] for all times t ∈ [0, τU ] 10 (see step 7 of the previous protocol). Expand the size of each spatial dimension, taking Nx → Mx and Ny → My and add zeros to ...

  86. [94]

    Following steps 8 and 9 of the previous protocol to construct the extended generator ˜U ex[Mx, My; Mx, My; t] → ˜U ex τU [M, M; t], where M = Mx × My

  87. [95]

    For the 2D homogeneous lattice in Fig

    Employ the generator ˜U ex[M, M; t] to construct CGME(t) for extended M × M lattice (see step 10 in the above protocol). For the 2D homogeneous lattice in Fig. 3, we employ the polaron dynamics on an 8 × 8-site lattice as the ref- erence simulation to predict the dynamics of a...

Pith tools

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