Impact of the Infrared Cutoff on Structure Formation in Tsallis Holographic Dark Energy

Biswajit Das

REVIEW 1 major objections 7 minor 1 cited by

Defects cut COF strength 40% while stiffness holds

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 · glm-5.2

2026-07-05 00:37 UTC pith:TEKNKYGS

load-bearing objection Solid MLIP-for-COFs paper with a real gap in finite-size treatment for defective-structure thermal conductivities the 1 major comments →

arxiv 2604.21490 v2 pith:TEKNKYGS submitted 2026-04-23 astro-ph.CO gr-qc

Impact of the Infrared Cutoff on Structure Formation in Tsallis Holographic Dark Energy

Biswajit Das This is my paper

classification astro-ph.CO gr-qc

keywords covalent organic frameworksmachine learning interatomic potentialsthermal conductivitymechanical propertiesdefectsMACEnon-equilibrium molecular dynamicsYoung's modulus

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.

This paper develops a machine learning interatomic potential called QCOF, built on the MACE architecture and trained on density functional theory data, to enable large-scale molecular dynamics simulations of covalent organic frameworks (COFs) containing structural defects at sizes exceeding 40,000 atoms. The authors use QCOF to compute thermal conductivity and mechanical stress-strain responses for two COFs, CTF-1 and COF-LZU1, each with point defects and grain boundaries at roughly 0.9% defect concentration. The central finding is a split between stiffness and strength: the 2D Young's modulus remains virtually unchanged by low defect densities, consistent with prior predictions that elastic moduli are insensitive to dilute disorder, yet the breaking strength drops by up to 40% from even a single defect in a large supercell, because defects act as stress concentrators that nucleate fracture. For thermal transport, the impact of defects depends on framework stiffness: the stiff CTF-1 shows 10-30% reductions in thermal conductivity from defects and grain boundaries, while the more compliant COF-LZU1 is nearly unaffected. The authors attribute this to the fact that stiff frameworks support longer phonon mean free paths, giving defects more scattering to disrupt, whereas compliant frameworks already have short mean free paths and additional scattering from a localized defect barely changes the overall heat flow. This interpretation is supported by atomic displacement data showing that a single defect in CTF-1 induces long-range structural perturbations across the full system width, while in COF-LZU1 the distortion remains localized.

Core claim

The paper's central object is the QCOF model, a compact invariant MACE-based machine learning interatomic potential (48 descriptor dimensions, no equivariance, 4 Angstrom cutoff) trained on over 36,000 DFT-computed conformations of carbon nitride nanosheets. The core discovery enabled by this model is that in defective COFs, the mechanical and thermal consequences of defects decouple from each other and from defect density in a framework-stiffness-dependent manner: Young's modulus is insensitive to dilute defects, breaking strength is sharply reduced (up to 40%) because defects concentrate stress, and thermal conductivity reduction depends on whether the pristine framework is stiff enough to

What carries the argument

The mechanism carrying the argument is the relationship between framework stiffness, phonon mean free path, and defect scattering. Stiff COFs like CTF-1 have longer phonon mean free paths in their pristine state, so defects introduce significant additional scattering and long-range structural perturbations, reducing thermal conductivity. Compliant COFs like COF-LZU1 already have short mean free paths and localized thermal fluctuations, so a defect adds little incremental scattering. For mechanical properties, the mechanism is the distinction between linear-response elasticity, which averages over the extended network and is insensitive to isolated defects, and fracture, which is initiated at

Load-bearing premise

The thermal conductivity values for defective CTF-1 and COF-LZU1 are computed at a single system size (approximately 28,000-45,000 atoms) without the finite-size extrapolation applied to the C2N validation system, even though the paper acknowledges that the NEMD method is known to be sensitive to system size due to phonon scattering at thermostat boundaries. If finite-size effects differ between pristine and defective structures, the reported absolute thermal conductivity

What would settle it

If finite-size extrapolation of the defective CTF-1 and COF-LZU1 systems reveals systematically different convergence behavior than the pristine C2N validation, the reported thermal conductivity values could shift enough to alter the conclusion about framework stiffness governing defect sensitivity.

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

If this is right

COF membranes or thin films may retain adequate stiffness for normal operation even with synthesis-induced defects, but remain vulnerable to catastrophic fracture under high stress, shifting design priorities from maximizing crystallinity to suppressing crack nucleation sites.
The finding that defect impact on thermal conductivity depends on framework stiffness rather than defect density alone suggests that thermal transport in COFs can be tuned by engineering compliance independently of defect engineering.
The QCOF model's ability to simulate 140,000-atom systems at roughly six GPU-days per nanosecond makes it suitable for high-throughput property screening of extended network materials, not just single-structure studies.
The split between stiffness and strength at low defect concentrations implies that standard elastic-property benchmarks (Young's modulus) are insufficient for assessing mechanical reliability of defective COFs; fracture stress measurements are the relevant metric.

Where Pith is reading between the lines

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

The thermal conductivity results for defective CTF-1 and COF-LZU1 were computed at a single system size without finite-size extrapolation, unlike the C2N validation where five sizes were used. If finite-size effects differ in defective structures, particularly given that defects introduce additional phonon scattering, the reported absolute kappa values could be systematically biased, though the re
The QCOF model was trained exclusively on carbon-nitrogen-hydrogen systems and cannot describe oxygen-containing defects, which are common in real COFs. Extending to oxygen-containing linkages would likely require retraining, and the transferability to such chemistries is untested.
The defects considered are fully interconnected (no dangling linkers), whereas experimental grain boundaries often involve missing connections. The authors note this would likely produce more pronounced property degradation, suggesting the reported 40% strength reduction may be a lower bound for real materials.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

Solid MLIP-for-COFs paper with a real gap in finite-size treatment for defective-structure thermal conductivities

read the letter

The paper you're looking at is actually the Szewczyk et al. COF/MLIP paper, not the cosmology paper in the arXiv abstract — there's a metadata mismatch on the arXiv listing. The scientific content is the materials work. Here's my read of it. The main contribution is a MACE-based MLIP (QCOF) trained specifically for carbon-nitride COFs, with systematic hyperparameter optimization and honest benchmarking against general-purpose MACE models and their fine-tuned variants. The 48-0-4 architecture lands on a sensible Pareto front of speed, memory, and accuracy. The C2N thermal conductivity validation is convincing: they did a five-size extrapolation (37–150 nm) and landed at 88.9 W/K/m, matching prior DFT-BTE (82.2) and MTP-NEMD (85.5) results. The large-scale defective COF simulations (>40k atoms) are genuinely new, and the qualitative finding — that stiff CTF-1 is much more sensitive to defects than compliant COF-LZU1 — is physically reasonable and supported by the atomic displacement analysis showing long-range perturbations in CTF-1 vs. localized distortions in COF-LZU1. The mechanical results (Young's modulus nearly unchanged at low defect concentration, but breaking strength drops up to 40%) are also solid and consistent with prior coarse-grained work. The stress-test concern about finite-size extrapolation is the real soft spot, and it lands. The authors themselves demonstrate that NEMD κ for C2N varies by roughly a factor of 4 across system sizes before extrapolation, yet the defective CTF-1 and COF-LZU1 thermal conductivities are computed at a single system size (34×10 unit cells, ~28–45k atoms). This matters because the defective-system atom counts are comparable to the smallest C2N system, where κ was far below the extrapolated value. Defects also change phonon scattering physics, so the finite-size dependence could differ from the pristine case in ways that confound the relative comparison. The absolute κ values and percentage reductions for defective structures could be systematically biased. That said, the qualitative trend — CTF-1 more defect-sensitive than COF-LZU1 — is probably robust, given the independent displacement analysis and the physical argument about mean free paths. The other limitations (fully interconnected defects, no shipped code/models, COF-LZU1 force RMSE degradation to ~400 meV/Å) are real but the authors acknowledge most of them honestly. This is a competent paper that does what it sets out to do, with one methodological gap that a referee should push on. I'd recommend accepting for serious peer review. The referee should ask for at least a limited finite-size check on one or two defective structures to confirm the qualitative trend holds.

Referee Report

1 major / 7 minor

Summary. This manuscript develops a machine learning interatomic potential (QCOF) based on the MACE architecture, trained on DFT data for carbon nitride nanosheets, and applies it to study thermal and mechanical properties of defective covalent organic frameworks (COFs). The QCOF model (48-0-4 invariant architecture) is benchmarked against general-purpose MACE models and fine-tuned variants, showing favorable trade-offs between accuracy, speed, and memory footprint. The model is validated on C2N thermal conductivity (88.9 W/K/m, matching literature values of 82-86 W/K/m) and then applied to defective CTF-1 and COF-LZU1 systems at scales exceeding 40,000 atoms. The key findings are that defects reduce thermal conductivity in stiff CTF-1 by 10-30% but leave compliant COF-LZU1 nearly unaffected, while mechanical breaking strength drops up to 40% at low defect concentrations despite Young's modulus remaining virtually unchanged.

Significance. The manuscript makes a solid contribution to the MLIP-for-materials literature by demonstrating a practical, quantum-informed potential capable of simulating defective COFs at experimentally relevant scales. The systematic benchmarking of architectures (Table 1, Fig. 2) and the Pareto-front analysis of speed, memory, and accuracy are valuable for the community. The C2N thermal conductivity validation against independent DFT-BTE and MTP-NEMD results (Refs. 55, 56) provides a credible external check. The physical interpretation—that defect impact on thermal transport correlates with framework stiffness—is plausible and supported by the atomic displacement analysis (Fig. 4c). The stress-strain results showing a split between stiffness and strength at low defect concentrations are a useful design insight.

major comments (1)

§Results (Thermal transport properties of defective COFs): The defective-structure thermal conductivities for CTF-1 and COF-LZU1 are computed at a single system size (34×10 unit cells, ~28-45k atoms) without finite-size extrapolation. The authors themselves demonstrate that NEMD is 'known to be sensitive to system size due to scattering of phonons on thermostat region boundaries' and perform a five-size extrapolation (37-150 nm) for the C2N validation, where κ varies from 23 to 56 W/K/m before converging to ~89 W/K/m — a factor of ~4 variation. No such extrapolation is performed for the defective COF systems. This is load-bearing for the central claim that defects reduce CTF-1 κ by 10-30% while leaving COF-LZU1 nearly unchanged, because: (1) the defective-system simulation size likely falls in the regime where finite-size effects are most severe; (2) finite-size effects could differ for

minor comments (7)

The abstract and title in the submission metadata reference 'Tsallis Holographic Dark Energy' (arXiv:2604.21490, astro-ph.CO), but the manuscript text is entirely about MLIPs for defective COFs (arXiv:2604.21494, cond-mat.mtrl-sci). This appears to be a metadata mismatch that should be corrected.
Fig. 4b: The y-axis label and units are unclear in the provided text. Please ensure all figure axes are clearly labeled with units.
§Results (Exploring elastic properties): The footnote explaining the Young's modulus calculation from Lin et al. is helpful but could be integrated more cleanly into the main text to avoid confusion about which elastic property is being reported.
§Results (Prediction of thermal transport properties): The statement 'κ values obtained with ReaxFF are consistently higher than those predicted by QCOF' would benefit from a brief discussion of whether this discrepancy is expected based on known ReaxFF limitations for these systems.
Table 1: The energy RMSE values are reported in meV but the header says 'energy RMSE [meV]'. Please verify that the units are per-atom or per-structure and state this explicitly.
§Discussion: The statement that 'defects matter most when the pristine framework is stiff enough to support longer phonon mean free paths' is a reasonable interpretation but is not directly supported by phonon mean free path calculations. A brief acknowledgment of this gap would strengthen the argument.
The reference to Ref. [46] (the training dataset) appears to be a prior work by the authors. Please ensure this is clearly disclosed as a self-citation and that the dataset is accessible.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for a careful and constructive report. The referee correctly identifies that our central claim about differential defect impact on thermal conductivity in CTF-1 versus COF-LZU1 rests on single-size NEMD simulations, despite our own demonstration that finite-size effects are severe in the C2N validation (factor of ~4 variation before convergence). We address this point below.

read point-by-point responses

Referee: §Results (Thermal transport properties of defective COFs): The defective-structure thermal conductivities for CTF-1 and COF-LZU1 are computed at a single system size (34×10 unit cells, ~28-45k atoms) without finite-size extrapolation. The authors themselves demonstrate that NEMD is 'known to be sensitive to system size due to scattering of phonons on thermostat region boundaries' and perform a five-size extrapolation (37-150 nm) for the C2N validation, where κ varies from 23 to 56 W/K/m before converging to ~89 W/K/m — a factor of ~4 variation. No such extrapolation is performed for the defective COF systems. This is load-bearing for the central claim that defects reduce CTF-1 κ by 10-30% while leaving COF-LZU1 nearly unchanged, because: (1) the defective-system simulation size likely falls in the regime where finite-size effects are most severe; (2) finite-size effects could differ for

Authors: The referee raises a valid and important concern. We acknowledge that the absence of finite-size extrapolation for the defective COF systems is a gap in our analysis, and we agree that it weakens the quantitative precision of the reported κ reductions (10–30% for CTF-1). We will address this in revision through the following measures: (1) We will perform additional NEMD simulations at 2–3 larger system sizes for both pristine and defective CTF-1 and COF-LZU1 to assess whether the relative κ reduction converges or changes with system length. (2) We will add an explicit discussion of this limitation in the revised manuscript, noting that while the pristine and defective systems are compared at identical sizes (so that absolute finite-size effects partially cancel in the ratio), the referee is correct that finite-size effects could differ between pristine and defective structures—particularly if defects alter the phonon mean free path distribution. (3) We will temper the quantitative claims accordingly, framing the 10–30% reduction as a preliminary estimate pending full finite-size analysis rather than a converged result. We note that the qualitative interpretation—that defect impact correlates with framework stiffness—is independently supported by the atomic displacement analysis (Fig. 4c), which shows long-range structural perturbations in stiff CTF-1 versus localized perturbations in compliant COF-LZU1. This mechanistic evidence does not depend on the thermal conductivity extrapolation and provides a physical rationale for the observed trend. However, we agree that the quantitative magnitude of the effect requires the additional simulations described above to be reported with confidence. revision: yes

Circularity Check

0 steps flagged

No significant circularity; training data from authors' prior DFT dataset (Ref 46) is standard MLIP practice and not load-bearing for the central defect-property predictions, which are validated against external benchmarks.

full rationale

The paper's derivation chain is: (1) DFT training data from Ref 46 (same-group prior work) → (2) QCOF MLIP trained on that data → (3) QCOF validated against external benchmarks (DFPT phonon dispersions, C2N thermal conductivity from independent groups in Refs 55/56, ReaxFF comparisons, Young's modulus from Lin et al. Ref 61) → (4) QCOF used to predict thermal/mechanical properties of defective COFs. The self-citation of Ref 46 for the training dataset is standard MLIP development practice — generating your own DFT reference data is not circular because the DFT calculations are independent of the downstream property predictions. The central claims (defects reduce CTF-1 κ by 10–30%, COF-LZU1 κ nearly unchanged, breaking strength drops up to 40%) are emergent outputs of NEMD and tensile simulations, not fitted quantities. No equation or prediction reduces to its inputs by construction. The C2N validation (κ = 88.9 W/K/m vs. literature 82–85.5 W/K/m) provides genuine external falsification. The only minor concern is that the training dataset and some validation structures come from the same group, but this does not constitute circularity in the sense of a prediction being forced by its inputs. Score 2 reflects the minor self-citation for training data, which is not load-bearing for the central claims.

Axiom & Free-Parameter Ledger

8 free parameters · 4 axioms · 1 invented entities

The axiom ledger captures the key assumptions and parameters. The most consequential ad hoc assumption is that defects are fully interconnected, which the authors themselves flag as not matching many experimental situations.

free parameters (8)

Descriptor dimensionality D = 48
Selected by hyperparameter optimization to balance accuracy, speed, and memory (Table 1).
Cutoff radius rc = 4 Å
Selected after screening showed no improvement beyond 4 Å due to periodic unit cell dimensions.
Equivariance E = 0 (invariant)
Invariant model chosen over equivariant for computational efficiency; the 48-0-4 model was selected as QCOF despite 128-1-4 having lower force RMSE.
Training epochs = 500
Number of training epochs for final models.
Loss weights (stage 1) = E:1, F:100, S:1
Two-stage weighting scheme with default weights.
Loss weights (stage 2) = E:1000, F:100, S:10
Second stage weights applied after 80% of training.
Defect concentration = ~0.9% (1 per 113 pores)
Single defect per NVE region in thermal simulations; single defect per 10×17 supercell in mechanical simulations.
Assumed monolayer thickness = 3.35 Å
Used to convert 2D stress to 3D GPa units; standard graphene thickness assumption applied to COFs.

axioms (4)

domain assumption DFT at PBE+D3 level provides accurate reference energies and forces for carbon nitride nanosheets
The entire MLIP training pipeline depends on DFT reference quality. PBE+D3 is standard for 2D materials but is an approximation, not exact.
domain assumption Non-equilibrium MD with the direct method yields reliable thermal conductivity after finite-size extrapolation
Invoked in Methods (Thermal transport properties). Standard assumption in the field, but the defective-structure simulations skip the extrapolation step.
ad hoc to paper Defects modeled as fully interconnected (no dangling linkers) are representative enough to draw conclusions about real defective COFs
Stated in Discussion: 'the defects and grain boundaries considered here are fully interconnected; that is, no dangling linkers are present.' Authors acknowledge this assumption does not hold for many experimental COFs.
domain assumption The MACE architecture with local atomic descriptors captures the relevant physics of COF interatomic interactions
Local cutoff of 4 Å means only short-range interactions are modeled. Long-range electrostatics and dispersion are not explicitly included.

invented entities (1)

QCOF model independent evidence
purpose: Machine learning interatomic potential for COF simulation
Validated against external C2N thermal conductivity, DFPT phonon dispersions, and compared with MTP, ReaxFF, and MACE-OFF24. The model is a specific trained instance of the public MACE architecture, not a new physical entity.

pith-pipeline@v1.1.0-glm · 18377 in / 4536 out tokens · 100055 ms · 2026-07-05T00:37:30.939911+00:00 · methodology

0 comments

read the original abstract

We investigate the viability of Tsallis holographic dark energy (THDE) models, focusing on the role of the infrared (IR) cutoff in the growth of cosmic structures. Considering two commonly used choices of the cutoff, the particle horizon and the future event horizon, we analyze the evolution of linear matter perturbations and compute the growth factor, growth rate, and the observable $f\sigma_8(z)$. These predictions are compared with observational data from redshift-space distortion measurements. We find that the growth history is highly sensitive to the choice of IR cutoff. Models based on the future event horizon are consistent with observational data and can provide a fit comparable to, or slightly better than, the $\Lambda$CDM model for suitable values of the Tsallis parameter $\delta$. In contrast, models constructed using the particle horizon generally fail to reproduce the observed growth of structure. These results demonstrate that the viability of THDE models depends crucially on the choice of IR cutoff and highlight the importance of structure formation as a stringent test of generalized holographic dark energy scenarios.

Review history (2 revisions) →

discussion (0)

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Phantom-Divide Crossing in Barrow-Tsallis Holographic Dark Energy with a Scale-Dependent Barrow Exponent
gr-qc 2026-07 conditional novelty 5.0

For generalized-entropy holographic dark energy with a future-event-horizon cutoff, the phantom-divide crossing on the kinematic branch is unique, quintessence-to-phantom, and fixes the local entropy-scaling dimension...

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages · cited by 1 Pith paper · 1 internal anchor

[1]

Data-Driven Thermal and Mechanical Modeling of Defective Covalent Organic Frameworks

Data-Driven Thermal and Mechanical Modeling of Defective Covalent Organic Frameworks Aleksander Szewczyk1, Leonardo Medrano Sandonas 1,*, David Bodesheim 1, Bohayra Mortazavi2,3, and Gianaurelio Cuniberti 1,4,5,6,* 1Institute for Materials Science and Max Bergmann Center of Biomaterials, TUD Dresden University of Technology, 01062, Dresden, Germany 2Insti...

work page internal anchor Pith review Pith/arXiv arXiv 2026
[2]

Data-Driven Characterization of (A)Chiral 2D Polymers

In this approach, the simulation cell containing a single layer of the studied COF under periodic boundary conditions was divided into four regions. Two regions (hot and cold reservoirs) were supplied with a constant heat 10/15 �� Yuan, S.et al.Covalent organic frameworks for membrane separation.Chem. Soc. Rev.��, 2665–2681, 10.1039/ C8CS00919H (2019). ��...

work page doi:10.1021/acscentsci.0c00463 2019

[1] [1]

Data-Driven Thermal and Mechanical Modeling of Defective Covalent Organic Frameworks

Data-Driven Thermal and Mechanical Modeling of Defective Covalent Organic Frameworks Aleksander Szewczyk1, Leonardo Medrano Sandonas 1,*, David Bodesheim 1, Bohayra Mortazavi2,3, and Gianaurelio Cuniberti 1,4,5,6,* 1Institute for Materials Science and Max Bergmann Center of Biomaterials, TUD Dresden University of Technology, 01062, Dresden, Germany 2Insti...

work page internal anchor Pith review Pith/arXiv arXiv 2026

[2] [2]

Data-Driven Characterization of (A)Chiral 2D Polymers

In this approach, the simulation cell containing a single layer of the studied COF under periodic boundary conditions was divided into four regions. Two regions (hot and cold reservoirs) were supplied with a constant heat 10/15 �� Yuan, S.et al.Covalent organic frameworks for membrane separation.Chem. Soc. Rev.��, 2665–2681, 10.1039/ C8CS00919H (2019). ��...

work page doi:10.1021/acscentsci.0c00463 2019