From Amorphous to Amorphous-Crystalline Mixed-Phase Boron Nitride: Evolution of the Thermal and Elastic Properties

Bartlomiej Graczykowski; Catherine Journet; Catherine Marichy; Clivia M. Sotomayor Torres; Daniel Capolat Palomar; Emigdio Chavez-Angel; Javier Rodriguez Viejo; Jiaqi Yang; Marianna Sledzinska; Onurcan Kaya

REVIEW 3 major objections 5 minor 3 references

Amorphous boron nitride films conduct heat below 0.5 W/mK, and their heat flow and stiffness track deposition temperature.

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 · deepseek-v4-flash

2026-08-01 05:31 UTC pith:KDV5TMEL

load-bearing objection Useful first systematic dataset on CVD a-BN thermal and elastic properties; the absolute kout values hinge on assumed interface conductances, but the trends and structural story are solid. the 3 major comments →

arxiv 2607.22185 v1 pith:KDV5TMEL submitted 2026-07-24 cond-mat.mtrl-sci

From Amorphous to Amorphous-Crystalline Mixed-Phase Boron Nitride: Evolution of the Thermal and Elastic Properties

Jiaqi Yang , Thomas Souvignet , Onurcan Kaya , Peng Xiao , Emigdio Chavez-Angel , Daniel Capolat Palomar , Zuzanna Ewa Kaczmarska , Bartlomiej Graczykowski

show 6 more authors

Javier Rodriguez Viejo Clivia M. Sotomayor Torres Catherine Marichy Catherine Journet Stephan Roche Marianna Sledzinska

This is my paper

classification cond-mat.mtrl-sci

keywords amorphous boron nitridechemical vapor depositioncross-plane thermal conductivityfrequency-domain thermoreflectanceYoung's modulusBrillouin light scatteringGreen-Kubo molecular dynamicsstructural ordering

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 tries to establish that the temperature at which a boron nitride film is grown by chemical vapour deposition acts as a single control knob for its atomic-scale ordering, and that this ordering in turn determines two practically important properties: how well the film conducts heat across its thickness and how stiff it is. The authors show that a fully amorphous film grown at 800 °C conducts less than 0.5 W/mK and has a Young's modulus of about 7.5 GPa, whereas films grown at 900–1000 °C, which contain embedded BN nanocrystallites, conduct up to ~1.5 W/mK and stiffen to about 53 GPa. If these numbers hold, amorphous BN is one of the softest and least heat-conducting dielectrics available, which matters for nanoelectronic interconnects and coatings where heat dissipation and mechanical flexibility are both needed. The paper also uses simulations to argue that the trends come from the bonding topology (sp² vs sp³ character, B–B vs B–N bonds) and from the presence of oxygen, hydrogen and carbon impurities, which can be used as further tuning knobs.

Core claim

The central claim is that short-range order — not just crystallinity — controls thermal and mechanical response in CVD boron nitride films. In the fully amorphous phase (grown at 800 °C), cross-plane thermal conductivity kout is below 0.5 W/mK and shows a mild thickness dependence (about 0.1–0.3 W/mK between 10 and 40 nm), which the authors attribute to heat carried by non-propagating vibrational modes (diffusons and propagons) rather than by long-wavelength phonons. As the growth temperature rises to 1000 °C, the films become a mix of amorphous matrix and BN nanocrystallites, and kout rises to ~1.5 W/mK while Young's modulus rises from 7.5 to 53 GPa. The authors support the structural inter

What carries the argument

The central mechanism is the deposition-temperature-controlled evolution of the bonding network: from a fully amorphous, low-density, partially oxidized network with many homonuclear B–B and N–N bonds and sp-hybridized sites, to a mixed amorphous–crystalline network with sp²-bonded BN crystallites. The paper ties this structural order to thermal transport through the vibrational density of states and the participation of extended vs localised modes, and to elastic response through the density and defect content visible in the Rayleigh surface-wave dispersion. The measurement chain — frequency-domain thermoreflectance for heat flow, micro-Brillouin light scattering with finite-element fitting

Load-bearing premise

The absolute values of the film's thermal conductivity rest on assuming the gold–BN and BN–silicon interface conductances are 20 and 30 MW/m²K; if those literature values do not apply to these rough, partially oxidised films, the reported kout, especially for the thinnest films, could shift beyond the quoted error bars.

What would settle it

Measure the same films with a different transducer metal (e.g., aluminium instead of gold) so that the metal/BN interface conductance changes by a known amount; if the extraction still returns kout <0.5 W/mK for the 800 °C film and the same thickness trend, the claim is robust. Alternatively, perform time-domain thermoreflectance with a much wider frequency range to independently separate the film conductance from the two interface conductances.

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

If this is right

If a-BN indeed conducts <0.5 W/mK cross-plane, it is competitive with or better than SiO₂, Al₂O₃ and TiO₂ as a heat-shy dielectric, but with the advantage of a substantially lower deposition temperature.
The systematic rise of kout with film thickness (0.1→0.3 W/mK from 10 to 40 nm) implies that vibrational modes with mean free paths of tens of nanometres contribute to heat flow even in the amorphous state, so thickness is a design parameter for a-BN thermal management.
The Young's modulus of ~7.5 GPa for a-BN means it is highly flexible and fracture-resistant, making it suited for protective coatings on flexible electronics, while the 1000 °C film's 53 GPa is closer to bulk h-BN's out-of-plane stiffness.
Non-stoichiometry and oxygen contamination naturally present in CVD a-BN lower its heat conduction, so stoichiometry control or intentional carbon doping (predicted to raise k up to ~15% carbon) offers a practical tuning route.

Where Pith is reading between the lines

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

The paper leaves implicit that the thickness dependence of kout in a-BN, if real, makes it an outlier among amorphous glasses; this could mean a-BN's amorphous network still hosts partially propagating modes, a claim worth testing with inelastic scattering or by measuring the same films at lower temperatures.
The assumed interface thermal conductances (20 and 30 MW/m²K) are the weakest point: for 10–40 nm films with k of order 0.1–0.5 W/mK, the film's own thermal resistance is comparable to the assumed interface resistances, so an independent measurement of the Au/BN and BN/Si interfaces (e.g., by varying the transducer metal) would either confirm or overturn the absolute kout values.
A testable extension suggested by the simulations is to grow a-BN with controlled carbon content near 15% and measure kout; the predicted non-monotonic peak would directly test the bonding-topology explanation against a simple mass-mismatch picture.

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

Useful first systematic dataset on CVD a-BN thermal and elastic properties; the absolute kout values hinge on assumed interface conductances, but the trends and structural story are solid.

read the letter

Quick take: this is a usable experimental paper on CVD-grown amorphous boron nitride, and the genuinely new part is the temperature series across the amorphous-to-mixed-phase transition with both thermal conductivity and elastic modulus. If you work on BN dielectrics or nanoscale thermal management, you want this reference.

What the paper does well: the structural characterization is thorough and converging — FTIR, Raman, HRTEM, EELS, XPS all point the same way, and the BLS+FEM elastic moduli are a nice addition. The MD simulations are independent of the FDTR data — computed from a chosen potential and structure generation, not fitted to the measured k — so there is no circularity. The qualitative ordering (800 °C < 900 °C < 1000 °C in both k and stiffness) is credible and probably robust.

The soft spot is the FDTR interface assumption. The paper fixes G(Au/BN) = 20 and G(BN/Si) = 30 MW/m²K from literature. For a 10–40 nm film with k around 0.1–0.3 W/mK, the film resistance and the interface resistance are the same order of magnitude. That means the absolute kout values — and especially the 0.1→0.3 W/mK thickness trend — are not uniquely determined by the data. If the actual interfaces are rougher or more oxidized than the literature surfaces, which is plausible for this a-BN, then the thinnest-film value could shift well beyond the quoted error bars. The sensitivity analysis that would settle this is in the SI, which I cannot see, so I would flag this as a real condition on the central quantitative claim rather than a fatal flaw.

The MD "good quantitative agreement" overstates things. Stoichiometric MD gives about 1.4 W/mK; the experiment is 0.1–0.3 W/mK. Impurity/porosity corrections can bring the simulation down to maybe 0.5–0.7 W/mK, still several times above the data. The MD is a reasonable mechanism story, but it is not a close quantitative match.

Only three growth temperatures back the "systematic" trend, which is fine for a first report but limits the claim. No data/code repository is provided, which matters here given the interface-assumption sensitivity.

Who should read this: anyone doing thermal modeling of BN-based dielectrics or designing a-BN coatings. It deserves a serious referee. My recommendation: send it to peer review, ask for the sensitivity analysis on G to be brought into the main text, for the G uncertainties to be propagated into the quoted k values, and for the MD agreement claim to be softened.

Referee Report

3 major / 5 minor

Summary. The paper reports a multi-technique study of CVD-grown boron nitride films deposited at 800–1000 °C, tracking the structural evolution from fully amorphous BN (a-BN) to mixed amorphous–crystalline and polycrystalline phases. Using FDTR, the authors extract an ultralow cross-plane thermal conductivity for a-BN (<0.5 W/mK for 10–40 nm thick films) with a slight thickness dependence, rising to ~1.5 W/mK with increasing crystalline order. Micro-Brillouin light scattering combined with COMSOL FEM yields Young's moduli of 7.5 ± 0.7 GPa (800 °C) and 53 ± 5 GPa (1000 °C). Green–Kubo molecular dynamics simulations are used to rationalize the thermal trends in terms of stoichiometry, hybridization, voids, and impurity content (C, H, O). The central claim is that growth temperature controls short-range order, which in turn controls heat conduction and stiffness.

Significance. If the quantitative values are correct, the work establishes a-BN as an ultralow-thermal-conductivity, ultralow-modulus dielectric whose properties are tunable by deposition temperature, which is significant for thermal management in nanoelectronics and for flexible coating applications. The paper's strengths are its comprehensive structural characterization (FTIR, Raman, XPS, TEM/SAED, EELS), the inclusion of BLS reference measurements on exfoliated h-BN that validate the elastic methodology, and MD simulations that are not fitted to the FDTR data. The systematic structural-to-property correlation is internally consistent. However, the absolute thermal conductivity values and the thickness trend rest on an interface-conductance assumption that needs stronger support, and the elastic extraction assumes isotropy for a polycrystalline film.

major comments (3)

[§3, Fig. 3a]
[§4, Fig. 4]
[§5, 'Good quantitative agreement']

minor comments (5)

[Abstract]
[Fig. 3b caption]
[§2, Fig. 2c]
[§4, first paragraph]
[References]

Circularity Check

0 steps flagged

No significant circularity: FDTR, BLS/FEM, and MD are independent measurements/simulations; fixed interface conductances are a stated assumption, not a self-referential prediction.

full rationale

The paper's main quantitative claims are measured quantities extracted by standard inverse analysis: FDTR fits a three-layer thermal model to the phase lag, BLS/FEM fits elastic constants to surface-wave dispersion, and Green-Kubo MD computes thermal conductivity from melt-quenched structural models. None of these stages fits a parameter to the quantity it is later said to explain. In FDTR, G12 = 20 and G23 = 30 MW/m2K are fixed from literature (refs 23,24) before fitting kout; this is an external parameter assumption, not a reduction of the predicted kout to a fitted value. The MD simulations use XPS-derived B/N ratios and impurity levels as inputs, but their output k values are not tuned to the FDTR data, and the paper does not adjust the interatomic potential or simulation protocol to reproduce the measured 0.1-0.5 W/mK values. The BLS/FEM Young's modulus determination is likewise an independent fit to measured dispersion and is benchmarked against literature h-BN elastic constants. Self-citations (Hong et al. Nature 2020 for a-BN dielectric properties; Sledzinska et al. and Graczykowski et al. for MoS2 thermal/elastic comparisons) are background or comparative benchmarks and do not carry the derivation. The fixed-interface-conductance assumption is a legitimate experimental-uncertainty concern because film and interface resistances are of similar magnitude, but it is not a circularity: kout is not defined in terms of G12/G23, nor are those values fitted from the same data. Thus no load-bearing step reduces to its own inputs.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 0 invented entities

The central quantities (kout, E) come from model-based fits with assumed inputs (interface conductances, heat capacity, isotropic elasticity), and the trend rationalization inherits the MD potential and structure-protocol choices. No new physical entities are introduced: 'propagons' and 'diffusons' are prior-literature vibrational categories, and the impurity knobs are existing chemical species. The ledger shows roughly even balance between what the paper contributes (measurements) and what it borrows from literature assumptions (interface values, potential, protocol).

free parameters (4)

Interface thermal conductances G12, G23 in FDTR model = 20 MW/m²K (Au/BN), 30 MW/m²K (BN/Si)
Fixed as "conservative literature-based values" (§3) rather than measured; for 10-40 nm films the interface resistance is the same order as the film resistance, so the extracted kout values and thickness trend depend on them.
FDTR volumetric heat capacity of a-BN = not stated in main text
The three-layer FDTR fit requires cp as input or output; the value used for a-BN is not given in the main text even though it trades against k in the fit.
Melt-quench cooling rate in MD structure preparation = 1 K/ps
Chosen by hand (Materials and Methods, SI); amorphous network topology, density, void fraction, and hence computed k all depend on this rate, which is orders of magnitude faster than experimental cooling.
Isotropic elastic inputs for COMSOL FEM of BLS dispersion = not stated in main text
The films (including the mixed-phase 1000 °C film with embedded crystallites) are modeled as isotropic effective media; Poisson's ratio, density, and roughness inputs are not specified in the main text.

axioms (5)

domain assumption The FDTR three-layer analytical heat-conduction model (refs 21-24) adequately describes the Au/BN/Si stack with isotropic cross-plane conduction and literature heat capacities.
Invoked in §3 for all kout extractions; standard but unverified for a-BN, whose measured k approaches the model's resolvability limit in 10 nm films.
domain assumption Literature interface-conductance values transfer to these specific CVD-grown interfaces.
The Au/BN and BN/Si interfaces in this work are partially oxidized (B-O bonds, XPS), boron-rich (B/N 1.2-1.4), and possibly rough; the actual G values are unknown.
domain assumption The empirical interatomic potential used in Green-Kubo MD accurately describes B-N, B-O, N-H, B-C, N-C bonding and anharmonicity.
All simulated k values (Fig. 5c-d) inherit the potential's accuracy; no validation against ab initio or experimental structure factors is shown in the main text (details deferred to SI).
ad hoc to paper Melt-quench structures at 1 K/ps with the chosen impurity distributions are representative of CVD-grown a-BN.
The amorphous network, density (2.21 → 1.93 g/cm³), nanovoid formation, and impurity placement are protocol-dependent and are asserted, not validated, to match the deposited films.
domain assumption XPS surface composition (B/N 1.2-1.4 and O/H/C contamination) represents the bulk of the film used to reconcile MD with FDTR.
XPS probes near-surface regions; the impurity percentages that close the MD-experiment gap are not quantified in the main text (Table S1, SI).

pith-pipeline@v1.3.0-alltime-deepseek · 10887 in / 20083 out tokens · 201028 ms · 2026-08-01T05:31:31.952963+00:00 · methodology

0 comments

read the original abstract

Amorphous boron nitride (aBN) is a promising dielectric and protective coating, yet its nanoscale heat dissipation and elastic response remain poorly quantified. Here we synthesize a variety of BN thin films by borazine-based chemical vapor deposition (800-1000 C) and study the temperature-driven structural transition from fully amorphous networks to mixed amorphous-crystalline films with embedded BN nanocrystallites.Frequency-domain thermoreflectance data show an ultralow, thickness-dependent cross-plane thermal conductivity for aBN (kout < 0.5 W m-1 K-1 for 10-40 nm), which increases systematically with crystalline order up to 1.5 W m-1 K-1. Micro-Brillouin light scattering and finite-element modelling reveal a concomitant stiffening, with Young's modulus rising from 7.5 +/- 0.7 GPa (800 C) to 53 +/- 5 GPa (1000 C). Green-Kubo molecular dynamics simulations rationalize these trends via bonding topology and vibrational transport, and highlight how oxygen, hydrogen and carbon impurities and composition provide practical knobs to further tune the thermal and mechanical responses in BN films for improving nano-electronics, interconnects and coating applications.

Figures

Figures reproduced from arXiv: 2607.22185 by Bartlomiej Graczykowski, Catherine Journet, Catherine Marichy, Clivia M. Sotomayor Torres, Daniel Capolat Palomar, Emigdio Chavez-Angel, Javier Rodriguez Viejo, Jiaqi Yang, Marianna Sledzinska, Onurcan Kaya, Peng Xiao, Stephan Roche, Thomas Souvignet, Zuzanna Ewa Kaczmarska.

**Figure 5.** Figure 5: Molecular Dynamics simulations. (a) Change in bonding environment and atomic hybridization according to stoichiometry of a-BN Inset: atomic structure of stoichiometric aBN. (b) Vibrational density of states as a function of B/N ratio. (c) Thermal conductivity as a function of ratio between boron and nitrogen. (d) Thermal conductivity as a function of concentration of contaminant atoms. Overall, the MD res… view at source ↗

discussion (0)

Reference graph

Works this paper leans on

3 extracted references · 1 canonical work pages

[1]

Dielectric Permittivity, Conductivity and Breakdown Field of Hexagonal Boron Nitride

(1) Pierret, A.; Mele, D.; Graef, H.; Palomo, J.; Taniguchi, T.; Watanabe, K.; Li, Y .; Toury, B.; Journet, C.; Steyer, P.; Garnier, V .; Loiseau, A.; Berroir, J.-M.; Bocquillon, E.; Fève, G.; V oisin, C.; Baudin, E.; Rosticher, M.; Plaçais, B. Dielectric Permittivity, Conductivity and Breakdown Field of Hexagonal Boron Nitride. Mater. Res. Express 2022, ...

doi:10.1088/2053-1591/ac4fe1 2022
[2021]

(16) Ye, F.; Liu, Q.; Xu, B.; Feng, P

https://doi.org/10.1088/2053-1583/ac2e51. (16) Ye, F.; Liu, Q.; Xu, B.; Feng, P. X.-L.; Zhang, X. Ultra-High Interfacial Thermal Conductance via Double hBN Encapsulation for Efficient Thermal Management of 2D Electronics. Small 2023, 19 (12), 2205726. https://doi.org/10.1002/smll.202205726. (17) Biswas, A.; Alvarez, G. A.; Li, T.; Christiansen-Salameh, J....

arXiv 2053
[2024]

(10) Weber, M.; Coy, E.; Iatsunskyi, I.; Yate, L.; Miele, P.; Bechelany, M

https://doi.org/10.1088/2515-7639/ad561e. (10) Weber, M.; Coy, E.; Iatsunskyi, I.; Yate, L.; Miele, P.; Bechelany, M. Mechanical Properties of Boron Nitride Thin Films Prepared by Atomic Layer Deposition. CrystEngComm 2017, 19 (41), 6089–6094. https://doi.org/10.1039/C7CE01357D. (11) Hong, S.; Lee, C.-S.; Lee, M.-H.; Lee, Y .; Ma, K. Y .; Kim, G.; Yoon, S...

work page doi:10.1088/2515-7639/ad561e 2017

[1] [1]

Dielectric Permittivity, Conductivity and Breakdown Field of Hexagonal Boron Nitride

(1) Pierret, A.; Mele, D.; Graef, H.; Palomo, J.; Taniguchi, T.; Watanabe, K.; Li, Y .; Toury, B.; Journet, C.; Steyer, P.; Garnier, V .; Loiseau, A.; Berroir, J.-M.; Bocquillon, E.; Fève, G.; V oisin, C.; Baudin, E.; Rosticher, M.; Plaçais, B. Dielectric Permittivity, Conductivity and Breakdown Field of Hexagonal Boron Nitride. Mater. Res. Express 2022, ...

doi:10.1088/2053-1591/ac4fe1 2022

[2] [2021]

(16) Ye, F.; Liu, Q.; Xu, B.; Feng, P

https://doi.org/10.1088/2053-1583/ac2e51. (16) Ye, F.; Liu, Q.; Xu, B.; Feng, P. X.-L.; Zhang, X. Ultra-High Interfacial Thermal Conductance via Double hBN Encapsulation for Efficient Thermal Management of 2D Electronics. Small 2023, 19 (12), 2205726. https://doi.org/10.1002/smll.202205726. (17) Biswas, A.; Alvarez, G. A.; Li, T.; Christiansen-Salameh, J....

arXiv 2053

[3] [2024]

(10) Weber, M.; Coy, E.; Iatsunskyi, I.; Yate, L.; Miele, P.; Bechelany, M

https://doi.org/10.1088/2515-7639/ad561e. (10) Weber, M.; Coy, E.; Iatsunskyi, I.; Yate, L.; Miele, P.; Bechelany, M. Mechanical Properties of Boron Nitride Thin Films Prepared by Atomic Layer Deposition. CrystEngComm 2017, 19 (41), 6089–6094. https://doi.org/10.1039/C7CE01357D. (11) Hong, S.; Lee, C.-S.; Lee, M.-H.; Lee, Y .; Ma, K. Y .; Kim, G.; Yoon, S...

work page doi:10.1088/2515-7639/ad561e 2017