Pith. sign in

REVIEW 2 major objections 2 minor 22 references

Automatic differentiation applied to two-dimensional THz signals retrieves time-dependent scattering rate, plasma frequency and resonance frequency at sub-pulse-width resolution.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.3

2026-06-25 20:23 UTC pith:SJ3NROTA

load-bearing objection The paper gives a practical AD+JAX route to pull time-dependent Drude-Lorentz parameters out of 2D THz waveforms at sub-pulse resolution, but uniqueness rests on numerical examples rather than a derived guarantee. the 2 major comments →

arxiv 2606.25479 v1 pith:SJ3NROTA submitted 2026-06-24 cond-mat.str-el

Automatic-differentiation-enabled dynamic parameter retrieval with sub-pulse-width resolution

Huaiyue Peng , Yuchen Lin , Fu Deng , Xiaoyue Zhou , Jingdi Zhang This is my paper
classification cond-mat.str-el
keywords THz-TDSautomatic differentiationinverse problemtime-dependent parametersfull-waveform inversionnon-equilibrium dynamicssub-pulse resolution
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Standard analysis in time-resolved terahertz spectroscopy fails when material properties change faster than the probe pulse, rendering the usual conductivity unreliable. The paper shows that the full two-dimensional time-domain waveform E(t_g, t_pp) together with automatic differentiation supplies enough information to invert uniquely for three well-defined time-dependent quantities: the scattering rate γ(t), plasma frequency ω_p(t) and resonance frequency ω0(t). Gradient-based optimization (Adam plus L-BFGS) makes the inversion stable against noise and pulse shape errors. A reader cares because the method recovers ultrafast non-equilibrium dynamics that conventional response-function approaches cannot access.

Core claim

By leveraging automatic differentiation on the two-dimensional time-domain signal E(t_g, t_pp), the inverse problem of retrieving the time-dependent scattering rate γ(t), plasma frequency ω_p(t) and resonance frequency ω0(t) can be solved uniquely at sub-pulse-width resolution, even though the optical conductivity becomes unreliable when dynamics occur on or below the probe-pulse timescale; the resulting optimization via JAX is robust to experimental noise and probe distortions.

What carries the argument

Full-waveform inversion framework that uses automatic differentiation on the measured 2D signal E(t_g, t_pp) to optimize the three time-dependent parameters γ(t), ω_p(t) and ω0(t).

Load-bearing premise

The two-dimensional measured signal E(t_g, t_pp) encodes enough independent information to determine uniquely the three time-dependent parameters despite the finite duration of the probe pulse.

What would settle it

Forward-simulating E(t_g, t_pp) from the retrieved γ(t), ω_p(t) and ω0(t) and finding systematic, noise-exceeding mismatches with the original measured 2D waveform.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Dynamics on timescales shorter than the probe pulse become accessible in photoinduced non-equilibrium experiments.
  • The retrieved parameters remain physically meaningful and well-defined even when conventional conductivity spectra are not.
  • Gradient optimization yields stable results in the presence of realistic experimental noise and pulse distortions.
  • The same framework is validated by both self-consistent numerical tests and actual ultrafast THz measurements.

Where Pith is reading between the lines

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

  • The method may transfer to other pulsed spectroscopies that face analogous pulse-width limits on the observable.
  • If additional independent signal dimensions can be recorded, further time-dependent parameters could be retrieved.
  • Open implementation of the JAX-based optimizer would allow direct testing on existing THz datasets.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. The manuscript presents an automatic-differentiation-enabled full-waveform inversion framework for time-resolved THz-TDS. Using the 2D measured field E(t_g, t_pp) and gradient-based optimization (Adam + L-BFGS in JAX), it retrieves the time-dependent parameters γ(t), ω_p(t), and ω0(t) under a Drude-Lorentz model at sub-pulse-width resolution, where the conventional conductivity σ(ω, t_pp) becomes unreliable.

Significance. If the uniqueness and robustness claims hold, the approach would allow extraction of well-defined physical observables during ultrafast non-equilibrium dynamics that standard response-function analysis cannot access. The AD/JAX implementation provides a concrete strength in computational efficiency and reproducibility for optimization against experimental noise and pulse distortions.

major comments (2)
  1. [Abstract] Abstract and central claim: the assertion that the inverse problem can be 'uniquely solved' is supported only by numerical self-consistent benchmarks and experimental examples; no analytic argument (e.g., injectivity of the Fréchet derivative of the forward map or full-rank condition on the discretized Jacobian) is supplied to establish that distinct trajectories {γ(t), ω_p(t), ω0(t)} cannot map to identical E(t_g, t_pp) within noise.
  2. [Methodology (forward model)] Forward-model section: the recovery is performed under an assumed Drude-Lorentz response; the manuscript does not derive or test the conditions under which the three-parameter mapping remains injective after convolution with finite probe width, nor does it supply exclusion criteria or noise-dependent error bounds that would falsify uniqueness.
minor comments (2)
  1. [Notation] Notation: define t_g and t_pp explicitly on first use and maintain consistent subscripting throughout the equations and figures.
  2. [Results figures] Figures: include quantitative uncertainty estimates (e.g., standard deviation across noise realizations) on the retrieved parameter traces in the experimental benchmarks.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for their positive assessment of the significance of our work and for the constructive major comments. We respond to each point below, and propose revisions where appropriate.

read point-by-point responses

  1. Referee: [Abstract] Abstract and central claim: the assertion that the inverse problem can be 'uniquely solved' is supported only by numerical self-consistent benchmarks and experimental examples; no analytic argument (e.g., injectivity of the Fréchet derivative of the forward map or full-rank condition on the discretized Jacobian) is supplied to establish that distinct trajectories {γ(t), ω_p(t), ω0(t)} cannot map to identical E(t_g, t_pp) within noise.

    Authors: We agree that the manuscript does not provide an analytic proof of uniqueness. Our claim is based on the observation that the optimization procedure, initialized from different starting points, consistently recovers the ground-truth parameters in self-consistent numerical tests even in the presence of noise and pulse distortions, and yields physically consistent results on experimental data. A rigorous mathematical analysis of the injectivity of the nonlinear forward operator is a substantial undertaking that lies beyond the scope of this primarily methodological paper in condensed-matter physics. We will revise the abstract to replace 'uniquely solve' with 'reliably retrieve' to better reflect the evidence presented. revision: partial

  2. Referee: [Methodology (forward model)] Forward-model section: the recovery is performed under an assumed Drude-Lorentz response; the manuscript does not derive or test the conditions under which the three-parameter mapping remains injective after convolution with finite probe width, nor does it supply exclusion criteria or noise-dependent error bounds that would falsify uniqueness.

    Authors: The forward model incorporates the convolution with the finite probe pulse width through the full time-domain simulation of E(t_g, t_pp). While we do not derive general analytic conditions for injectivity, the numerical benchmarks explicitly include the effects of finite pulse width and demonstrate successful recovery. We will add a new subsection discussing the robustness of the retrieval under varying noise levels and probe distortions, including examples where the optimization fails to converge, thereby providing practical exclusion criteria based on convergence behavior and residual error. revision: yes

standing simulated objections not resolved
  • A complete analytic proof of uniqueness via Fréchet derivative injectivity or Jacobian rank analysis.

Circularity Check

0 steps flagged

No circularity; retrieval is numerical optimization against external data.

full rationale

The paper frames the core contribution as an optimization procedure (Adam + L-BFGS via JAX/AD) that fits the three time-dependent parameters γ(t), ω_p(t), ω0(t) to the measured 2D waveform E(t_g, t_pp). No equation or step defines any recovered quantity in terms of itself or renames a fitted output as an independent prediction. No self-citation is invoked to establish uniqueness or to smuggle an ansatz; the forward model is the standard Drude-Lorentz response convolved with the probe, and the inverse step is purely data-driven fitting. The absence of an analytic injectivity proof is a completeness issue, not a circularity issue. The derivation chain therefore remains self-contained against external measured waveforms.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Review performed on abstract only; no explicit free parameters, invented entities, or additional axioms beyond the core domain assumption of unique invertibility are stated.

axioms (1)
  • domain assumption The two-dimensional time-domain signal E(t_g, t_pp) contains sufficient information to uniquely determine the time-dependent parameters γ(t), ω_p(t), ω0(t) at sub-pulse resolution
    This premise is required for the inversion framework to succeed and is asserted in the abstract without derivation.

pith-pipeline@v0.9.1-grok · 5777 in / 1292 out tokens · 29331 ms · 2026-06-25T20:23:21.450252+00:00 · methodology

0 comments
read the original abstract

Time-resolved terahertz time-domain spectroscopy (THz-TDS) is a phase-sensitive tool in condensed matter physics for tracking photoinduced non-equilibrium dynamics of low-energy elementary excitations. However, the measured response function, optical conductivity $\sigma(\omega,t_{pp})$, becomes unreliable in reporting the state of matter when material properties drastically change on a timescale comparable to or less than the probe pulse duration, obscuring the sub-pulse-width dynamics. To resolve this issue, we present a full-waveform inversion framework inspired by the multi-dimensional retrieval philosophy of frequency-resolved optical gating (FROG). By leveraging the automatic differentiation (AD) technique and the two-dimensional time-domain signal $E(t_{g},t_{pp})$, we show one can uniquely solve the inverse problem, at the sub-pulse-width resolution, of retrieving physical observables that are still well-defined, i.e., time-dependent scattering rate $\gamma(t)$, plasma frequency $\omega_\mathrm{p}(t)$ and resonance frequency $\omega_0(t)$, while the response functions are not. Further optimization by gradient-based routines (Adam + L-BFGS) via JAX makes the method exceptionally robust against experimental noise and probe pulse distortions. The validity of the AD-enabled methodology is benchmarked both by a self-consistent numerical approach and by experimental data from real ultrafast THz spectroscopy measurements.

Figures

Figures reproduced from arXiv: 2606.25479 by Fu Deng, Huaiyue Peng, Jingdi Zhang, Xiaoyue Zhou, Yuchen Lin.

Figure 1
Figure 1. Figure 1: FIG. 1. (a) Schematic of OPTP measurement: a femtosecond pump pulse (red) excites the sample; a broadband THz probe [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Probe-pulse independence. (a) THz probe pulses [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Representative test with experimental data on WTe [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

22 extracted references

  1. [1]

    D. N. Basov, R. D. Averitt, D. Van Der Marel, M. Dres- sel, and K. Haule, Electrodynamics of correlated electron materials, Reviews of Modern Physics83, 471 (2011)

    2011

  2. [2]

    Zhang and R

    J. Zhang and R. D. Averitt, Dynamics and control in complex transition metal oxides, Annual Review of Ma- terials Research44, 19 (2014)

    2014

  3. [3]

    Dressel and G

    M. Dressel and G. Gr¨ uner,Electrodynamics of solids: op- tical properties of electrons in matter(Cambridge univer- sity press, 2002)

    2002

  4. [4]

    Averitt, G

    R. Averitt, G. Rodriguez, A. Lobad, J. Siders, S. A. Trug- man, and A. J. Taylor, Nonequilibrium superconductivity and quasiparticle dynamics in YBa 2Cu3O7−δ, Physical Review B63, 140502 (2001)

    2001

  5. [5]

    Zhang, J

    J. Zhang, J. Yong, I. Takeuchi, R. L. Greene, and R. D. Averitt, Ultrafast terahertz spectroscopy study of a Kondo insulating thin-film SmB6: Evidence for an emer- gent surface state, Physical Review B97, 155119 (2018)

    2018

  6. [6]

    J. T. Kindt and C. A. Schmuttenmaer, Theory for deter- mination of the low-frequency time-dependent response function in liquids using time-resolved terahertz pulse spectroscopy, J. Chem. Phys.110, 8589 (1999)

    1999

  7. [7]

    M. C. Beard, G. M. Turner, and C. A. Schmuttenmaer, Transient photoconductivity in GaAs as measured by time-resolved terahertz spectroscopy, Phys. Rev. B62, 15764 (2000)

    2000

  8. [8]

    R. D. Averitt, G. Rodriguez, J. L. Siders, S. A. Trugman, and A. J. Taylor, Conductivity artifacts in optical-pump THz-probe measurements of YBa 2Cu3O7, Journal of the Optical Society of America B17, 327 (2000)

    2000

  9. [9]

    Nienhuys and V

    H.-K. Nienhuys and V. Sundstr¨ om, Intrinsic complica- tions in the analysis of optical-pump, terahertz-probe ex- periments, Phys. Rev. B71, 235110 (2005)

    2005

  10. [10]

    Ashoka, H

    A. Ashoka, H. Tam, S. Heffernan, T. Buber, L. Brus, A. Bakulin, and A. Musser, Extracting quantitative di- electric properties from pump-probe spectroscopy, Nat. Commun.13, 1437 (2022)

    2022

  11. [11]

    Trebino, K

    R. Trebino, K. W. DeLong, D. N. Fittinghoff, J. N. Sweetser, M. A. Krumb¨ ugel, B. A. Richman, and D. J. Kane, Measuring ultrashort laser pulses in the time- frequency domain using frequency-resolved optical gat- ing, Rev. Sci. Instrum.68, 3277 (1997)

    1997

  12. [12]

    D. J. Kane, Real-time measurement of ultrashort laser pulses using principal component generalized projections, IEEE J. Sel. Top. Quantum Electron.4, 278 (1998)

    1998

  13. [13]

    R. W. Gerchberg and W. O. Saxton, A practical algo- rithm for the determination of phase from image and diffraction plane pictures, Optik35, 237 (1972)

    1972

  14. [14]

    J. R. Fienup, Phase retrieval algorithms: a comparison, Appl. Opt.21, 2758 (1982)

    1982

  15. [15]

    A. G. Baydin, B. A. Pearlmutter, A. A. Radul, and J. M. Siskind, Automatic differentiation in machine learning: a survey, J. Mach. Learn. Res.18, 1 (2018)

    2018

  16. [16]

    D. P. Kingma and J. Ba, Adam: A method for stochastic optimization, in3rd International Conference on Learn- ing Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings(2015)

    2015

  17. [17]

    D. C. Liu and J. Nocedal, On the limited memory BFGS method for large scale optimization, Math. Program.45, 503 (1989)

    1989

  18. [18]

    S. S. Schoenholz and E. D. Cubuk, JAX, M.D.: A frame- work for differentiable physics, inAdvances in Neural In- formation Processing Systems, Vol. 33 (2020)

    2020

  19. [19]

    Y. Hu, L. Anderson, T.-M. Li, Q. Sun, N. Carr, J. Ragan- Kelley, and F. Durand, Difftaichi: Differentiable pro- gramming for physical simulation, in8th International Conference on Learning Representations (ICLR)(2020)

    2020

  20. [20]

    Raissi, P

    M. Raissi, P. Perdikaris, and G. E. Karniadakis, Physics- informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations, J. Comput. Phys.378, 686 (2019)

    2019

  21. [21]

    X. Zhou, F. Deng, Y. Gao, Y. Wu, Y. Chan, S. Li, N. Wang, J. Liu, and J. Zhang, Light-induced percolative electronic phase transition in the type-II Weyl semimetal WTe2, Phys. Rev. B112, L161107 (2025)

    2025

  22. [22]

    C. C. Homes, M. N. Ali, and R. J. Cava, Optical proper- ties of the perfectly compensated semimetal WTe2, Phys. Rev. B92, 161109 (2015)

    2015