Pith. sign in

REVIEW 5 major objections 6 minor 7 references

Chirality-Dependent Kinetics of Single-Walled Carbon Nanotubes from Machine-Learning Force Fields

T0 review · 5 major / 6 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 Nanotube chirality is not fixed when the cap nucleates; post-nucleation defect kinetics at the cobalt–carbon interface shifts the population toward smaller, near-armchair tubes.

desk verdict A serious MLFF-MD study that reframes chirality selection as post-nucleation defect kinetics, but the unvalidated force field and fits to the same data keep it from being definitive. read the letter →

arxiv 2411.19764 v1 pith:AFYNJQNZ submitted 2024-11-29 cond-mat.mes-hall cond-mat.mtrl-sciphysics.comp-ph

Sida Sun , Shigeo Maruyama , Yan Li This is my paper · ORCID
classification cond-mat.mes-hallcond-mat.mtrl-sciphysics.comp-ph
keywords carbonnanotubechiralitymachine-learningforcefielddefectkineticsmicrokineticmodelingcobaltcatalystedgepatternsmoleculardynamicschirality-selectivegrowth
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 argues that the chiral structure of a growing carbon nanotube is not locked in when the tube cap nucleates; it continues to be rewritten by defect kinetics at the nanotube–catalyst interface. Using molecular dynamics driven by a cobalt–carbon machine-learning force field, the authors grow nanotubes on a cobalt catalyst at 1500 K and extract the rates of C2 addition and etching from a new '0-1' edge-pattern representation. They fit these rates into a microkinetic model that reproduces both the observed edge-pattern distributions and the power-law lifetime distribution of pentagon defects. The conclusion is that pentagon and heptagon defects—grouped into $\ell=2,3,4$ classes by their local environment—heal at different chirality-dependent rates, and this differential healing shifts the chirality distribution toward smaller diameters and near-armchair tubes. If correct, this makes defect healing, not cap design, the practical lever for chirality-selective nanotube growth.

What carries the argument

The load-bearing machinery is a microkinetic model of edge and defect reactions built on a binary-necklace representation of nanotube edges. In that representation, each edge is encoded as a circular string of '0' and '1' beads (the two orientations of adjacent deg-2/deg-3 vertex pairs), which removes the ambiguity of the traditional armchair/zigzag labeling; all length-preserving reactions are C2 addition ('01' $\rightarrow$ '10') and C2 etching ('10' $\rightarrow$ '01') on a strongly connected reaction network. Four rate parameters—base rate $k_0$, a chemical-potential factor, an armchair-edge modifier, and an armchair/zigzag-junction modifier—are fitted to MLFF-MD edge-pattern statistics and reproduce the observed distributions. The defect part of the model classifies pentagons by the number $\ell=2,3,4$ of hexagons they share, and fits healing rates $k_{d2}=104\ \mathrm{ns}^{-1}$, $k_{d3}=17.7\ \mathrm{ns}^{-1}$, and $k_{d4}=3.63\ \mathrm{ns}^{-1}$; solving the resulting master equation reproduces the power-law lifetime statistics of defects. These chirality-dependent healing rates are what carry the shift in the chirality distribution.

What would settle it

Compute the healing barriers of the three pentagon-defect classes at a cobalt–carbon interface with a method independent of the fitted force field; if $\ell=4$ defects do not heal several times slower than $\ell=3$ defects, the predicted shift toward near-armchair chiralities and the claimed composition of the power-law lifetime distribution collapse.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the chirality distribution observed after five hexagon layers differs drastically from the distribution immediately after nucleation, and the difference is created by defect kinetics. The initial 'zero-layer' distribution is roughly consistent with kinetic nucleation from cap configurational degeneracy, but pentagon defects form immediately after nucleation and either heal or become encapsulated; the two diameter-control mechanisms—tilted-interface etching (type I) and heptagon compensation (type II)—then either shrink or expand the tube circumference and switch the chiral index. Because $\ell=4$ pentagon defects heal much more slowly than $\ell=2$ or $\ell=3$ defects, and because $\ell=4$ defects are relatively more common at small chiral angles, the net effect is a shift toward smaller diameters and near-armchair chiralities. The authors conclude that the initial cap chirality is only a boundary condition and that the formation and resolution of defects drive the chirality distribution toward a steady state.

Load-bearing premise

The entire argument rests on the in-house cobalt–carbon machine-learning force field being accurate for the specific C2 addition/etching and defect-healing events at 1500 K, a regime validated only indirectly through bulk properties rather than by direct barrier comparison.

Editorial extensions

If this is right

  • The initial cap chirality is only a boundary condition: the five-layer chirality distribution is produced by post-nucleation defect kinetics, so cap-nucleation energies alone cannot predict the final chirality.
  • The model attributes the near-armchair preference to the slower healing of $\ell=4$ pentagon defects, not to equilibrium edge energies or to cap selection.
  • Diameter and chirality are controlled by the same defect kinetics: pentagon and heptagon formation and resolution shrink or expand the tube circumference and can switch the chiral index.
  • Applying the same microkinetic workflow to other catalyst elements or alloys should reveal different defect-healing rates and, potentially, different chirality distributions, making defect kinetics a design target.

Reading between the lines

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

  • An implication the authors leave implicit is that a catalyst engineered to accelerate $\ell=4$ defect healing, or to suppress heptagon compensation, should narrow the chirality distribution more effectively than a catalyst surface tuned for epitaxial cap matching.
  • If the defect-kinetic picture is right, time-resolved observations of a single growing tube should show discretized diameter jumps corresponding to type-I tilted-interface etching and type-II heptagon compensation, rather than a fixed chirality inherited from nucleation.
  • The model's growth-mode crossover could be tested independently by measuring the contact angle of the nanotube edge on the catalyst and checking whether the fitted armchair/zigzag junction free energy changes sign at the same diameter.
  • The defect model predicts that the power-law tail of pentagon-defect lifetimes is composed of at least two distinct healing channels; a single-exponential experimental lifetime histogram would be hard to reconcile with the claimed $\ell=3$/$\ell=4$ rate separation.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The paper reports a combined machine-learning force field (MLFF) development, large-scale molecular dynamics (MD) growth simulations of single-walled carbon nanotubes (SWCNTs) on a Co55 catalyst at 1500 K, and microkinetic modeling of edge and defect dynamics. A new binary-necklace representation of nanotube edge patterns is introduced, and a four-parameter kinetic growth model is fitted to steady-state edge pattern statistics. The authors further classify pentagon defects by the number of surrounding hexagons (l = 2, 3, 4) and fit a four-rate defect model to defect lifetime histograms. The central claim is that post-nucleation defect kinetics, rather than cap nucleation alone, controls the chirality distribution, shifting it toward smaller diameters and near-armchair chiralities via the so-called diameter control mechanisms (type-I tilted-interface etching and type-II heptagon compensation).

Significance. If the central claim is correct, the paper offers a new mechanistic picture in which chirality selection is a kinetic, defect-driven process at the nanotube-catalyst interface, and it provides a modeling workflow (binary necklace representation, microkinetic growth model, defect lifetime model) that could be transferred to other catalyst systems. The proposed edge representation is a genuinely useful formal contribution, as it resolves ambiguities in the conventional A-Z edge label. The paper also explicitly reproduces and explains a power-law defect lifetime distribution, which is a key observation from earlier MLFF-MD work. However, the strength of the conclusions is currently limited by two intertwined issues: the MLFF is not validated on the specific defect reaction barriers that govern the purported rate ordering, and the central model reproductions are fits to the same statistics used to determine their parameters.

major comments (5)
  1. [Edge pattern statistics from MLFF-driven MD simulations] The MLFF is validated only through global properties (equations of state, melting point, 11-fold CV RMSE of 8.4 meV/atom and 0.28 eV/Å). The central conclusion that l=4 pentagon defects heal much slower than l=3 defects (k_d3=17.7 ns^-1, k_d4=3.63 ns^-1 in the chirality preference section) rests on the relative accuracy of the MLFF for these defect healing channels at the Co-C interface. At 1500 K, a barrier error of only 0.1-0.2 eV changes rates by a factor of ~2-7, and a 0.5 eV error by ~40. The training set is not demonstrated to sample the rare defect configurations or transition states for C2 addition/etching and pentagon healing at the interface. Without a targeted comparison of MLFF against DFT for the l=2,3,4 defect reaction barriers and the growth-relevant C2 addition/etching barriers, the rate ordering that drives the near-armchair preference is not established.
  2. [SWCNT edge patterns and reaction networks; Chirality preference from defect kinetics] The growth model parameters (k0, lambda_mu, lambda_A, lambda_AZ) are fitted per chirality to the steady-state edge pattern distributions, and the fitted predictions are then plotted against those same distributions (Fig. 2b and 3b). Likewise, the defect model rates (k0, k_d2, k_d3, k_d4) are fitted to the defect lifetime histograms in Fig. 7a and then shown to reproduce them in Fig. 7b. These reproductions are therefore not independent validations of the models; they are consistency checks of the fitting procedure. An independent predictive test is needed, such as holding out a subset of trajectories or predicting a different observable (e.g., the zero-layer chirality distribution or defect occurrence ratios as a function of chiral angle) without refitting.
  3. [Chirality preference from defect kinetics, Figure 5] The comparison between the zero-layer and five-layer chirality distributions is based on 210 MD simulations, with many chiralities represented by only one or two tubes (e.g., (12,4): 1, (13,1): 1, (13,4): 2 in the Fig. 5a caption). The claimed shift toward smaller diameters and near-armchair chiralities could be dominated by a few trajectories that undergo the shrinking-cone or type-I/II mechanisms. The paper should provide Poisson error bars per chirality and a statistical significance test (e.g., Kolmogorov-Smirnov on the diameter or chiral-angle distributions) to support the claim that the shift is robust rather than a small-sample artifact.
  4. [SWCNT edge patterns and reaction networks; Chirality preference from defect kinetics] The growth model explicitly excludes anomalous edges longer than (n+m) and therefore excludes (n,0) and (n,1) chiralities, as stated in the text: "the edges with lengths longer than (n+m) ... are temporarily not considered. Therefore, chiralities such as (n,0) and (n,1) are effectively excluded from our discussions." However, the defect model's l=4 defects are described as hosted on anti-zigzag sites in anomalous edges, and the slow healing of these defects is used to explain the near-armchair preference (Fig. 7d-f). There is a tension between excluding anomalous edges from the growth model and relying on them for the defect-mediated chirality claim. The authors should either extend the growth model to include anomalous edges or clearly demonstrate that the exclusion does not bias the conclusions drawn from the defect model.
  5. [Edge pattern statistics from MLFF-driven MD simulations, Figures 3 and 4] The paper states that the quality of the growth-model fit decreases with diameter and that for the two (7,6) cases the edge-pattern statistics are inconsistent between trajectories. The extracted A|Z junction free energy, Delta_G_A|Z, crosses to negative values for large diameters (Fig. 4a), which is attributed to 'insufficient statistics' or a growth-mode crossover. Because the diameter-dependent parameters (Delta_G^‡ and the chemical potential shift in Fig. 4b-c) are used to argue for the diameter shrinking mechanism and the preference toward smaller diameters, the reliability of the fit for the larger-diameter chiralities is load-bearing. The authors should quantify the fit uncertainty, e.g., by bootstrap or by reporting the number of observed transitions per edge pattern, and show that the large-diameter trends are not fitting artifacts.
minor comments (6)
  1. [SWCNT edge patterns and reaction networks, Figure 1] The mapping between the '0-1' representation and the 'A-Z' representation is introduced in the text but is difficult to follow; a small table listing the ten (6,3) edge patterns, their '0-1' strings, and their corresponding 'A-Z' labels would greatly improve readability.
  2. [SWCNT edge patterns and reaction networks, Eq. (1)] In the Polya enumeration formula, the summation variable d is not explicitly stated to run over the divisors of gcd(n,m); also, the text says 'permutating' where 'permuting' is meant.
  3. [Chirality preference from defect kinetics, Figure 5a caption] The caption states that three chiralities are not included, but does not say why; a brief explanation (e.g., the detaching case) would be useful for interpreting the distribution.
  4. [Chirality preference from defect kinetics] The value mu_C_demo = 0.17 eV used in Fig. 4c to illustrate the constant-chemical-potential scenario is introduced without explaining its origin or whether it is a representative experimental value.
  5. [Edge pattern statistics from MLFF-driven MD simulations, objective function] The objective function J combines a cross-entropy term (dimensionless) with quadratic terms weighted by w1 = w2 = 1 ns; a sentence clarifying the choice of units and the normalization of the incidence counts would make the fitting procedure more transparent.
  6. [Conclusion] The conclusion appropriately hedges the main claim with 'might depend largely on the defect kinetics,' but the abstract states more definitively that 'the formation and resolution of defects drive the chirality distribution toward a steady state'; the paper would benefit from a more explicit statement of which parts are established vs. hypothesized.

Circularity Check

2 steps flagged · score 6.0 of 10

Growth and defect models are fitted to the same MD statistics they are claimed to reproduce; the near-armchair preference inherits a fitted rate ordering.

  1. fitted input called prediction [Figure 1 caption; 'SWCNT edge patterns and reaction networks' and Figures 2b-3b]
    "The green (blue) arrows show the possible C2 addition (etching) reactions, and their arrow widths denote the rates which are fitted to the statistics from the machine-learning force field-driven molecular dynamics (MLFF-MD) simulation. (b) Reaction incidences and the steady-state distributions from the MLFF-MD simulation (left) and the model prediction (right)."

    The 'model prediction' is the steady-state solution p_inf of the master equation whose four rate parameters were fitted to the same edge-pattern distribution and reaction incidences. The paper's objective J = -sum p_inf,i ln p_hat,i + w1(f-f_hat)^2 + w2(r-r_hat)^2 is minimized against those same observables, so the plotted agreement between bars and fitted lines is enforced by construction. Describing this as a 'prediction' presents a fit as an independent validation.

  2. fitted input called prediction ['Chirality preference from defect kinetics' section, Figure 7b]
    "By solving this defect model numerically with rejection-free kinetic Monte Carlo simulations, the power-law behavior can be reproduced as in Figure 7b. ... The rate constants are fitted with Bayesian optimization, being k0= 1.63 ns−1, kd2= 104 ns−1, kd3= 17.7 ns−1, and kd4= 3.63 ns−1."

    The four defect-model rates are fitted to the MD lifetime histograms shown in Figure 7a, and the 'reproduced' power-law curve in Figure 7b is generated from those fitted rates. The reproduction is therefore tautological. The subsequent chirality argument that 'the faster healing rate of l = 3 defects in our system causes the negative correlation between the defect lifetimes and the chiral angles (Figure 7d), resulting in the preference toward near-armchair chiralities' relies on the fitted ordering kd3 > kd4, so the central chirality conclusion inherits a fitted input rather than being independently predicted.

full rationale

The paper is transparent about fitting, but it repeatedly frames fitted model outputs as predictions or reproductions. The growth model's four parameters are fitted per chirality to the steady-state edge-pattern distribution and the reaction incidence/growth rate, so the 'model prediction' in Figures 1b/2b/3b is the same data re-expressed through the fitted master equation. The defect model's four rates are fitted to the lifetime histograms, making the 'power-law behavior reproduced' in Figure 7b a consequence of the fit rather than evidence for the model. The central claim that chirality preference arises from defect kinetics is partially circular because the key rate ordering kd3 > kd4 is itself a fitted parameter from the same trajectories. However, several major results are independent of these fits: the 'zero-layer' versus 'five-layer' chirality distribution shift, the observed type-I/type-II diameter-control mechanisms, and the chirality-dependent occurrence ratios of l=3/l=4 defects come directly from MLFF-MD trajectories. The MLFF validation gap (no defect-specific barrier check) is a correctness risk, not a circularity, and the self-citations to prior Maruyama-group work are not load-bearing. Overall, the paper contains two clear fitted-input-called-prediction episodes, giving partial circularity rather than full derivation-by-definition.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The central claim rests on a fitted machine-learned potential, a four-parameter per-chirality kinetic model, and a four-rate defect model. The MLFF is a domain assumption rather than an axiom; the rate-factor model and the linear defect chain are ad hoc constructions introduced for this paper. The steady-state master equation is standard mathematics. No new physical entities are introduced.

free parameters (4)
  • Per-chirality growth-model parameters (k0, lambda_mu, lambda_A, lambda_AZ) = not tabulated; derived free energies in Figure 4
    Fitted per chirality to minimize objective J against MLFF-MD edge-pattern distributions, total reaction incidence, and growth rate. These parameters carry the model's reproduction of the observed kinetics.
  • Defect-model rates (k0, kd2, kd3, kd4) = 1.63, 104, 17.7, 3.63 ns^-1
    Fitted with Bayesian optimization to the observed defect lifetime distributions; used to infer healing-rate contrasts among l=2,3,4 defects and their chirality dependence.
  • Bell-Evans-Polanyi transition-state coordinate alpha = 0.5 (assumed)
    Set by hand, not fitted; converts lambda_mu into delta_mu and lambda_A or lambda_AZ into free energies. Changing alpha would rescale the reported thermodynamic quantities.
  • Co-C MLFF model parameters (DeepPot-SE weights) = trained on 6,523 structures; RMSE 8.4 meV/atom energy, 0.28 eV/Angstrom force
    The entire MD dataset, and therefore all edge-pattern and defect statistics, is generated by this fitted potential. The weights are not released.
assumptions (6)
  • domain assumption The home-built Co-C MLFF trained on 6,523 DFT structures approximates ab initio accuracy for C addition, etching, and defect kinetics at 1500 K.
    All MD statistics and fitted models inherit this. Validation is limited to equations of state and cobalt melting point; there is no direct kinetic benchmark against DFT.
  • domain assumption SWCNT edge growth is dominated by C2 addition and etching over anti-armchair sites or A-edges; length-changing anomalous edge reactions are negligible for the chiralities considered.
    The reaction networks are built only from '10' to '01' and '01' to '10'; near-zigzag (n,0) and (n,1) tubes, where anomalous edges are crucial, are excluded.
  • standard math The master equation has a unique steady state because the addition and etching networks are strongly connected, and the steady-state distribution describes the MD edge-pattern statistics.
    Uses the Fernengel-Drossel result for finite-state Markov chains; assumes sufficient sampling so observed incidences equal steady-state fluxes.
  • ad hoc to paper Rate constants factor as k0 times lambda_mu^Delta_i_mu times lambda_A^Delta_i_A times lambda_AZ^Delta_i_AZ, with lambda_A > 1 and lambda_AZ near 1, and alpha = 0.5 in the Bell-Evans-Polanyi principle.
    This is the assumed four-parameter rate model with no derivation from first principles; the fitted lambda values are then interpreted as thermodynamic free energies.
  • domain assumption Carbon deposition at one atom per 2 ns corresponds to a constant, negative excess chemical potential, and the simulated mu_C is adaptive; extrapolation to experiments assumes delta_mu = c0 / d_t^2 - mu_C with mu_C = 0.17 eV.
    Used to argue that in real constant-mu_C conditions the diameter distribution narrows; this is an extrapolation beyond the simulated ensemble.
  • ad hoc to paper Pentagon defects can be classified by l=2,3,4 shared hexagons and their kinetics modeled by a linear chain of hexagon addition and etching reactions at equal rates k0 with defect healing rates kd2, kd3, kd4.
    The defect model is deliberately crude and ignores geometry; it is fitted to lifetime distributions and then used to explain chirality preference.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Chirality-Dependent Kinetics of Single-Walled Carbon Nanotubes from Machine-Learning Force Fields." pith.science (2026). https://pith.science/paper/AFYNJQNZ

@misc{pith2026241119764,
  author       = {Pith},
  title        = {Pith review of: Chirality-Dependent Kinetics of Single-Walled Carbon Nanotubes from Machine-Learning Force Fields},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AFYNJQNZ}},
  note         = {Machine review of arXiv:2411.19764}
}
read the original abstract

The origin of the chirality of single-walled carbon nanotubes (SWCNTs) has been a long-standing dispute. Molecular dynamics (MD) simulations driven by machine-learning force fields (MLFF), which can study the interface dynamics under near ab-initio accuracy, provides a powerful technique to reveal the formation mechanism of SWCNTs. Here, we develop a cobalt-carbon MLFF and perform growth simulations on a cobalt catalyst to investigate the chirality preference of the growth of SWCNTs under the vapor-liquid-solid (VLS) regime. Through microkinetic modeling, we reproduce the observed growth and defect kinetics, demonstrating their dependence on the chirality. It is observed that while the initial chirality assignment is likely related to the configurational degeneracy of the nanotube caps, pentagon defects immediately form and resolve after nucleation. Such processes, which we name as diameter control mechanisms, not only control the diameter toward an optimum but also shift the chirality distribution drastically. Our work therefore offers a microkinetic modeling workflow for the chirality-dependent kinetics of the SWCNTs, highlighting the important contribution of the defect kinetics to the chirality origination.

Figures

Figures reproduced from arXiv: 2411.19764 by the authors.

Figure 1
Figure 1. “0-1” and “A-Z” edge pattern representations and reaction networks of a (6,3) single-walled carbon nanotube (SWCNT). (a) Construction of the reaction network. The edge atoms and bonds are colored by the type of atomic pair they are in, the pair “23” (“32”) in violet (blue). This nota￾tion is further simplified by denoting “23” (“32”) with “0” (“1”) or violet (blue) beads on a necklace. After assigning “10” to A-edge… view at source ↗
Figure 2
Figure 2. Edge pattern statistics and predictions from the growth model. (a) The 17 selected single-walled carbon nanotubes (SWCNTs) with small￾er diameters among the 34 selected ones grown on the Co55 catalyst using our developed machine learning force field (MLFF). These SWCNTs are allowed to evolve for at least ten layers of defect-free hexagons for detailed analysis. Pink (grey) spheres are cobalt (carbon) atoms, and pent… view at source ↗
Figure 3
Figure 3. Similar to [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Thermodynamic quantities derived from the fitted models. (a) A-edge free energies and A|Z junction free energies. (b) Relative activation free energy of the baseline reaction. (c) Chemical potential (violet line) and excess chemical potential (arrows), showing that all…
Figure 5
Figure 5. Figure 5: Chirality distributions and observed diameter control mechanisms that resolve small circumferences of the nanotube-catalyst interface. (a) “Five-layer” distribution, taken at five defect-free hexagon layers for all cases, except one case that has finally detached from …
Figure 6
Figure 6. Figure 6: (a) Mean defect lifetime distribution of the 34 trajectories. Error bars denote the range, and the dashed line is fitted against a pow￾er-law function. (b) Three classes of pentagon defects classified by their number of shared edges l = 2,3,4 with the hexagons. system,…
Figure 7
Figure 7. Figure 7: Statistics of the pentagon defects, including lifetimes and occurrences. (a) Lifetime distribution of the pentagon defects, classified by their surrounding hexagons 𝑙 = 2,3,4 when they form. (b) Fitted predictions of the lifetime distributions from our defect model. Ba…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

7 extracted references · 7 canonical work pages

  1. [1]

    Single-Shell Carbon Nanotubes of 1-Nm Diameter

    (1) Iijima, S.; Ichihashi, T. Single-Shell Carbon Nanotubes of 1-Nm Diameter. Nature 1993, 363 (6430), 603–605. (2) Reich, S.; Li, L.; Robertson, J. Epitaxial Growth of Carbon Caps on Ni for Chiral Selectivity. Phys. Status Solidi B 2006, 243 (13), 3494–3499. (3) Reich, S.; Li, L.; Robertson, J. Control the Chirality of Carbon Nanotubes by Epitaxial Growt...

    work page 1993

  2. [31]

    DeePMD-Kit: A Deep Learn- ing Package for Many-Body Potential Energy Representation and Molecular Dynamics

    (23) Wang, H.; Zhang, L.; Han, J.; E, W. DeePMD-Kit: A Deep Learn- ing Package for Many-Body Potential Energy Representation and Molecular Dynamics. Comput. Phys. Commun. 2018, 228, 178–184. (24) Zeng, J.; Zhang, D.; Lu, D.; Mo, P.; Li, Z.; Chen, Y.; Rynik, M.; Huang, L.; Li, Z.; Shi, S.; Wang, Y.; Ye, H.; Tuo, P.; Yang, J.; Ding, Y.; Li, Y.; Tisi, D.; Ze...

    work page 2018

  3. [1906]

    R.; Yakobson, B

    (8) Ding, F.; Harutyunyan, A. R.; Yakobson, B. I. Dislocation Theory of Chirality-Controlled Nanotube Growth. Proc. Natl. Acad. Sci. U.S.A. 2009, 106 (8), 2506–2509. (9) Magnin, Y.; Amara, H.; Ducastelle, F.; Loiseau, A.; Bichara, C. Entropy-Driven Stability of Chiral Single-Walled Carbon Nano- tubes. Science 2018, 362 (6411), 212–215. (10) Förster, G. D....

    work page 2009

  4. [2024]

    (17) Bachilo, S

    http://arxiv.org/abs/2302.09264 (accessed 2024-08-04). (17) Bachilo, S. M.; Balzano, L.; Herrera, J. E.; Pompeo, F.; Resasco, D. E.; Weisman, R. B. Narrow (n,m)-Distribution of Single-Walled Carbon Nanotubes Grown Using a Solid Supported Catalyst. J. Am. Chem. Soc. 2003, 125 (37), 11186–11187. (18) Wang, H.; Wang, B.; Quek, X.-Y.; Wei, L.; Zhao, J.; Li, L...

    work page arXiv 2024

  5. [4892]

    S.; Bets, K

    (5) Penev, E. S.; Bets, K. V.; Gupta, N.; Yakobson, B. I. Transient Ki- netic Selectivity in Nanotubes Growth on Solid Co–W Catalyst. Nano Lett. 2018, 18 (8), 5288–5293. (6) Bets, K. V.; Penev, E. S.; Yakobson, B. I. Janus Segregation at the Carbon Nanotube–Catalyst Interface. ACS Nano 2019, 13 (8), 8836–8841. (7) Penev, E. S.; Artyukhov, V. I.; Yakobson,...

    work page 2018

  6. [6750]

    Growth Modes of Single-Walled Carbon Nanotubes on Cata- lysts

    (28) Yang, F.; Zhao, H.; Li, R.; Liu, Q.; Zhang, X.; Bai, X.; Wang, R.; Li, Y. Growth Modes of Single-Walled Carbon Nanotubes on Cata- lysts. Sci. Adv. 2022, 8 (41), eabq0794. (29) Tersoff, J. Energies of Fullerenes. Phys. Rev. B 1992, 46 (23), 15546–15549. (30) Zhang, R.; Zhang, Y.; Zhang, Q.; Xie, H.; Qian, W.; Wei, F. Growth of Half-Meter Long Carbon N...

    work page 2022

  7. [9653]

    Molecular Dynamics of Chirality Definable Growth of Single-Walled Carbon Nanotubes

    (14) Yoshikawa, R.; Hisama, K.; Ukai, H.; Takagi, Y.; Inoue, T.; Chiashi, S.; Maruyama, S. Molecular Dynamics of Chirality Definable Growth of Single-Walled Carbon Nanotubes. ACS Nano 2019, 13 (6), 6506–6512. (15) Hedman, D.; McLean, B.; Bichara, C.; Maruyama, S.; Larsson, J. A.; Ding, F. Dynamics of Growing Carbon Nanotube Interfaces Probed by Machine Le...

    work page 2019

Pith tools

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