Integration of Atomic Layer Epitaxy Crystalline Ga2O3 on Diamond for Thermal Management

Jingjing Shi; Karl D. Hobart; Marko J. Tadjer; Mark S. Goorsky; Samuel Graham; Tatyana Feygelson; Tingyu Bai; Virginia D. Wheeler; Zhe Cheng

REVIEW 3 major objections 5 minor 3 references

Integration of Atomic Layer Epitaxy Crystalline Ga2O3 on Diamond for Thermal Management

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read Directly grown Ga2O3 on diamond conducts heat across the interface about ten times better than a weakly bonded Ga2O3-diamond interface; a clean interface measures 179 MW/m2K.

desk verdict First ALD-grown Ga2O3 on diamond with high measured TBC, but the quantitative claims need uncertainty analysis before they are relied upon. read the letter →

arxiv 1908.08665 v1 pith:P65HXHFS submitted 2019-08-23 physics.app-ph cond-mat.mtrl-sci

Zhe Cheng , Virginia D. Wheeler , Tingyu Bai , Jingjing Shi , Marko J. Tadjer , Tatyana Feygelson , Karl D. Hobart , Mark S. Goorsky

show 1 more author

Samuel Graham

This is my paper · ORCID

classification physics.app-phcond-mat.mtrl-sci

keywords galliumoxideatomiclayerdepositiondiamondthermalboundaryconductancetime-domainthermoreflectancemanagementinterfacebondingwidebandgapsemiconductors

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

This paper tries to show that a scalable, covalently bonded interface between gallium oxide and diamond can remove heat much more effectively than the weakly bonded interfaces used in earlier device demonstrations. By growing crystalline Ga2O3 with atomic layer deposition directly on single-crystal diamond and measuring with time-domain thermoreflectance, the authors find a thermal boundary conductance of 179 MW/m2K for a clean interface, about ten times the 17 MW/m2K reported for Van der Waals bonded Ga2O3-diamond. They also report that gallium-rich and oxygen-rich surface pretreatments lower the boundary conductance by about 20 percent, consistent with interface chemistry controlling heat transport. The films themselves have very low thermal conductivity, near the amorphous limit, which makes the high boundary conductance a necessary but not sufficient condition for thermal management.

What carries the argument

The measurement engine is time-domain thermoreflectance (TDTR), in which a modulated pump laser heats an aluminum transducer and a delayed probe laser reads the surface temperature decay; the data are fit with an analytical heat transfer solution that yields the thermal conductivity of the Ga2O3 film and the Ga2O3-diamond thermal boundary conductance as parameters. The comparison baseline is the previously reported Van der Waals bonded interface from exfoliated Ga2O3, and the structural companion is cross-sectional TEM showing 10-20 nm grains and abrupt interfaces. The pretreatments - an ultra-clean surface, a gallium-rich surface, and an oxygen-rich surface - are the controlled variable that tests whether interface chemistry changes conductance.

What would settle it

Take the same 30 nm sample and analyze the TDTR data with a sensitivity or uncertainty analysis over the film conductivity and interface conductance, or grow films at several thicknesses (e.g., 30, 60, 120 nm) with identical pretreatments; if the best-fit interface conductance drifts with film thickness, the reported 179 MW/m2K value is not uniquely determined by the data.

Watch

Extended reading notes

Core claim

The central discovery is that direct ALD growth of nanocrystalline Ga2O3 on diamond produces an atomically abrupt, void-free interface whose thermal boundary conductance reaches 179 MW/m2K, roughly an order of magnitude above the 17 MW/m2K of exfoliated, Van der Waals bonded Ga2O3 on diamond. In the same set of samples, pretreating the diamond surface with gallium or oxygen before growth reduces the conductance to about 136-139 MW/m2K, about 20 percent lower, which the authors attribute to interface chemical states rather than structural differences. The measured thermal conductivity of the ALD films is only 1.5-1.8 W/mK, close to the amorphous minimum, because the 10-20 nm grains scatter phonons heavily; the paper's point is therefore that the interface no longer dominates the thermal bottleneck, even though the film still does.

Load-bearing premise

The paper assumes that the heat flow through the film and the heat flow across the interface can be told apart from one laser-heating measurement, but it does not show how much each parameter can move without changing the fit.

Editorial extensions

If this is right

A clean, covalently bonded Ga2O3-diamond interface removes the interfacial heat bottleneck that limits exfoliated devices, so the main remaining thermal resistance in such a stack is the Ga2O3 film itself.
Surface pretreatments matter: if a Ga-rich or O-rich surface is required for electrical or epitaxial reasons, the expected boundary conductance is roughly 20% lower than on an ultra-clean surface.
ALD integration offers a scalable path to Ga2O3-on-diamond thermal management, in contrast to mechanical exfoliation.
Because measured film conductivity is near the amorphous limit, device designs that reduce the Ga2O3 thickness or replace it with higher-quality material would benefit most from the high interface conductance.
Room-temperature surface-activated bonding of pre-grown Ga2O3 layers to diamond or SiC should also produce high boundary conductance if covalent bonds across the interface are the controlling factor.

Reading between the lines

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

A direct test would vary Ga2O3 thickness while keeping pretreatment fixed; if the extracted interface conductance is truly intrinsic, it should be thickness-independent, whereas fitting degeneracy would reveal itself as systematic drift.
The paper's 'clean interface' TBC of 179 MW/m2K is likely close to the practical upper bound for Ga2O3-diamond, so further gains in heat extraction will have to come from raising the film conductivity or using thinner films, not from larger interface conductance.
The same ALD-on-diamond recipe could be applied to other low-conductivity ultra-wide-bandgap oxides, such as AlGaO or InGaO alloys, to test whether the bonding argument generalizes.

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.

Desk Editor's Note

read the letter

Here's the short version: this is the first ALD growth of crystalline Ga2O3 directly on diamond, and the thermal boundary conductance numbers are the first for a covalently bonded Ga2O3–diamond interface. The central qualitative claim — strong interfacial bonding gives an order-of-magnitude higher TBC than weak van der Waals bonding — is plausible and consistent with prior work. The quantitative values, however, are less solid than the abstract implies, because the TDTR analysis never shows that film conductivity and interface conductance are separable for a 30 nm film.

What the paper does well: the materials work is clean, the TEM is convincing, and the comparison of the measured film conductivity (1.5–1.76 W/m·K) to the amorphous limit is a useful benchmark. The method section is standard TDTR, and the authors cite the relevant prior work, including their own, appropriately.

The soft spots are real. The film's thermal resistance t/k is about 20 m²K/GW, three to four times the interface resistance claimed (≈5.6–7.4 m²K/GW). With the film dominating the in-series resistance, a TDTR fit that extracts k and TBC simultaneously needs a sensitivity or correlation analysis to show the two are unique; the paper does not provide one. There are no error bars, and each condition is one sample. The reported 20% reduction for Ga-rich and O-rich pretreatments could be fitting artifact. The comparison to the 17 MW/m²K vdW baseline also depends on the absolute accuracy of both fits, which is not established.

If I had to bet, I'd say the qualitative conclusion survives: strong interface bonds increase TBC, and the clean ALD interface is likely much better than vdW transfer. But the headline 179 MW/m²K and the 10× factor should be presented as provisional until the uncertainty analysis is done.

This deserves peer review — the integration route is new and the thermal management problem is important. A solid revision would include repeat samples, error bars, and a sensitivity analysis showing the parameter separation. Then the claim will be in a shape people can rely on.

Referee Report

3 major / 5 minor

Summary. The manuscript reports ALD growth of Ga2O3 on single-crystal diamond with different surface pretreatments and TDTR measurements of the resulting film thermal conductivity and Ga2O3-diamond thermal boundary conductance (TBC). The principal claims are: (i) ALD-Ga2O3 films are nanocrystalline with very low thermal conductivity around 1.5 W/m-K, close to amorphous-limit models; (ii) the ultra-clean Ga2O3-diamond interface has TBC of 179 MW/m2-K, about ten times larger than the 17 MW/m2-K previously reported for van der Waals bonded exfoliated Ga2O3-diamond interfaces; and (iii) Ga-rich and O-rich pretreatments reduce the TBC by about 20% (136 and 139 MW/m2-K). TEM images show abrupt interfaces without voids and grain sizes of 10-20 nm. The authors conclude that covalent interfacial bonding strongly enhances heat transport and that ALD integration is a promising thermal-management route for Ga2O3 devices.

Significance. If the quantitative claims are correct, the paper provides an important data point for Ga2O3 thermal management: a scalable ALD route to Ga2O3-on-diamond with an order-of-magnitude larger TBC than vdW-bonded interfaces. The comparison of the measured film conductivity with the Cahill and diffuson minimum-thermal-conductivity models is appropriate, and the TEM evidence for abrupt, void-free interfaces supports the qualitative conclusion that strong interfacial bonding improves heat transfer. The manuscript is concise and the experimental design is straightforward. However, the main quantitative contribution is not yet fully supported because the paper reports no uncertainty analysis, uses one sample per condition, and does not demonstrate that the TDTR fits can uniquely separate the film conductivity from the interface TBC for these thin, low-conductivity films.

major comments (3)

[Table 1; RESULTS AND DISCUSSION] No uncertainties are reported for any of the fitted thermal property values, and each growth condition is represented by a single sample. The claim that Ga-rich and O-rich pretreatments reduce the TBC by about 20% (179 MW/m2-K versus 136 and 139 MW/m2-K) is a central conclusion of the paper, but with no error bars or repeat measurements the differences could plausibly be within experimental scatter. Please provide uncertainty propagation from the TDTR fits and, if possible, repeat measurements or an explicit statement of the run-to-run repeatability of this measurement geometry.
[SAMPLES AND METHODS; Table 1] The TDTR analysis fits both the Ga2O3 film thermal conductivity and the Ga2O3-diamond TBC from films only 28-30 nm thick, but the paper does not demonstrate that these parameters are uniquely separable. With k ≈ 1.5 W/m-K, the film contributes a series thermal resistance of t/k ≈ 19-20 m2K/GW, while the claimed interface resistance is 1/TBC ≈ 5.6-7.4 m2K/GW; the film therefore contributes roughly three to four times more resistance than the interface. A sensitivity analysis (for example, derivatives of the TDTR signal with respect to k and TBC, or a correlation matrix) is needed to show that the fitted k and TBC values are not trading against each other. The authors should also report the TDTR model inputs, including Al transducer thermal properties, diamond substrate properties, spot sizes, and modulation frequencies, so that the parameter separation can be independently assessed.
[INTRODUCTION; RESULTS AND DISCUSSION] The headline comparison of 179 MW/m2-K with the 17 MW/m2-K van der Waals interface relies on the absolute accuracy of both values, but the baseline of 17 MW/m2-K is cited only to a conference abstract (Ref. 9) with no methodological details. Please provide the supporting measurement details for that baseline or cite a fuller account; otherwise the tenfold-improvement claim is not securely anchored.

minor comments (5)

[Title; Abstract; SAMPLES AND METHODS] The title and abstract describe 'Atomic Layer Epitaxy' while the body consistently describes atomic layer deposition with alternating TMG and oxygen plasma pulses; please clarify whether the growth qualifies as ALE or adjust the title to avoid overstating the epitaxial character.
[Fig. 2 caption] The caption says 'Cahill mode of minimum thermal conductivity' and should read 'Cahill model'; the caption also begins with a lowercase letter and should be capitalized.
[REFERENCES] Reference 4 has an incomplete volume/page format ('Appl. Phys. Lett. 15 (8)') and reference 9 is a conference abstract with incomplete bibliographic information; both need to be completed.
[RESULTS AND DISCUSSION (Fig. 3 discussion)] The text refers to 'focus ion beam (FIB)' preparation; this should be 'focused ion beam'.
[RESULTS AND DISCUSSION; Table 1] The text refers to 'Table I' in one place and 'Table 1' elsewhere; please use consistent table numbering.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the Ga2O3-diamond TBC values are new TDTR measurements, and the central comparisons are external benchmarks rather than fitted inputs.

full rationale

The paper's load-bearing claims are the measured TBC values (179, 136, and 139 MW/m2K) and film thermal conductivities, obtained by fitting TDTR signals with a standard analytical heat-transfer solution. These outputs are not defined in terms of the paper's conclusions; they are independent experimental results. The comparison against amorphous-limit thermal conductivity models (Cahill and diffuson) is a benchmark calculation, not a fit to the measured values. The order-of-magnitude comparison with the Van der Waals bonded Ga2O3-diamond interface is taken from a separate experimental report (ref. 9), not constructed from the present model. Reliance on the authors' prior TDTR methodology and on material property inputs from their earlier papers is ordinary method reuse, and no load-bearing step reduces by construction to its own inputs. The absence of an explicit sensitivity analysis for separating film conductivity and interface TBC in the thin, low-thermal-conductivity films (SAMPLES AND METHODS, Table 1) is a measurement-uncertainty/correctness concern rather than a circularity concern: the paper does not define the reported TBC as an input or derive it from an assumed film conductivity. No circular step can therefore be quoted from the text.

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

The paper does not introduce ad hoc fit parameters or invented entities. The central claim relies on the TDTR model assumptions, the approximation of amorphous DOS by single-crystal DOS, the grain-boundary-scattering interpretation, and the inferred covalent nature of the interface. These are domain assumptions, not free parameters.

assumptions (4)

domain assumption The TDTR analytical heat transfer solution accurately represents the layered Al/Ga2O3/diamond system and yields unique values for film thermal conductivity and interface TBC.
The paper states the signal is fitted with an analytical heat transfer solution (SAMPLES AND METHODS) but does not provide a sensitivity or uniqueness analysis, which is necessary to separate the two unknown thermal properties for a 30 nm low-conductivity film.
domain assumption The density of states of single crystal beta-Ga2O3 approximates that of the amorphous/nanocrystalline phase for the minimum thermal conductivity models.
Stated in the text: 'We use the single crystal phase to approximate the density of states of amorphous phase because they are usually similar' (RESULTS AND DISCUSSION, Eq. 2 discussion). This is a reasonable approximation but unverified for Ga2O3.
domain assumption The low thermal conductivity of the ALD films is dominated by phonon grain boundary scattering and not by other defects or measurement artifacts.
The paper attributes the about 1.5 W/mK value to 10 to 20 nm grains seen in TEM; no alternative contributions such as porosity, impurities, or FIB damage are quantified.
domain assumption The interface in Samp2 is covalently bonded rather than vdW bonded, based on atomically abrupt TEM images and the comparison to exfoliated interfaces.
The claim that strong bonds cause high TBC rests on TEM showing good contact and the absence of voids, but bonding chemistry is not directly measured.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Integration of Atomic Layer Epitaxy Crystalline Ga2O3 on Diamond for Thermal Management." pith.science (2026). https://pith.science/paper/P65HXHFS

@misc{pith2026190808665,
  author       = {Pith},
  title        = {Pith review of: Integration of Atomic Layer Epitaxy Crystalline Ga2O3 on Diamond for Thermal Management},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/P65HXHFS}},
  note         = {Machine review of arXiv:1908.08665}
}

read the original abstract

Ga2O3 has attracted great attention for electronic device applications due to its ultra-wide bandgap, high breakdown electric field, and large-area affordable substrates grown from the melt. However, its thermal conductivity is significantly lower than that of other wide bandgap semiconductors, which will impact its ability to be used in high power density applications. Thermal management in Ga2O3 electronics will be the key for device reliability, especially for high power and high frequency devices. Similar to the method of cooling GaN-based high electron mobility transistors by integrating it with high thermal conductivity diamond substrates, this work studies the possibility of heterogeneous integration of Ga2O3 with diamond for thermal management of Ga2O3 devices. In this work, Ga2O3 was deposited onto single crystal diamond substrates by ALD and the thermal properties of ALD-Ga2O3 thin films and Ga2O3-diamond interfaces with different interface pretreatments were measured by TDTR. We observed very low thermal conductivity of these Ga2O3 thin films due to the extensive phonon grain boundary scattering resulting from the nanocrystalline nature of the Ga2O3 film. However, the measured thermal boundary conductance (TBC) of the Ga2O3-diamond interfaces are about 10 times larger than that of the Van der Waals bonded Ga2O3 diamond interfaces, which indicates the significant impact of interface bonding on TBC. Furthermore, the TBC of the Ga-rich and O-rich Ga2O3-diamond interfaces are about 20% smaller than that of the clean interface, indicating interface chemistry affects interfacial thermal transport. Overall, this study shows that a high TBC can be obtained from strong interfacial bonds across Ga2O3-diamond interfaces, providing a promising route to improving the heat dissipation from Ga2O3 devices.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

3 extracted references · 3 canonical work pages

[1]

Higashiwaki and G

1 M. Higashiwaki and G. H. Jessen, Appl. Phys. Lett. 112, 060401 (2018). 2 S. Pearton, J. Yang, P. H. Cary IV, F. Ren, J. Kim, M. J. Tadjer, and M. A. Mastro, Appl. Phys. Rev. 5, 011301 (2018). 3 P. Jiang, X. Qian, X. Li, and R. Yang, Appl. Phys. Lett. 113, 232105 (2018). 4 Z. Cheng, N. Tanen, C. Chang, J. Shi, J. McCandless, D. Muller, D. Jena, H. G. Xing, and S

work page 2018
[2]

Graham, Appl. Phys. Lett. 15 (8), (2019). 5 Z. Cheng, L. Yates, J. Shi, M. J. Tadjer, K. D. Hobart, and S. Graham, APL Materials 7, 031118 (2019). 6 S. B. Reese, T. Remo, J. Green, and A. Zakutayev, Joule, 3, 4 (2019). 7 J. Noh, M. Si, H. Zhou, M . J. Tadjer, and D. Y. Peide, 76th Device Research Conference (DRC), pp. 1-2. IEEE, (2018). 8 H. Zhou, K. Maiz...

work page 2019
[3]

Foley, S

Zivasatienraj, B. Foley, S. T. Pantelides, and L. F. Piper, Phys. Rev. Mater. 3, 025002 (2019). 14 14 J. T. Gaskins, G. Kotsonis, A. Giri, C. T. Shelton, E. Sachet, Z. Cheng, B. M. Foley, Z. Liu, S. Ju, and M. S. Goorsky, Nano Lett. 18, 12, 7469, (2017). 15 P. Jiang, X. Qian, and R. Yang, J. of Appl. Phys. 124, 161103 (2018). 16 F. Mu, Z. Cheng, J. Shi, S...

work page 2019

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Higashiwaki and G

1 M. Higashiwaki and G. H. Jessen, Appl. Phys. Lett. 112, 060401 (2018). 2 S. Pearton, J. Yang, P. H. Cary IV, F. Ren, J. Kim, M. J. Tadjer, and M. A. Mastro, Appl. Phys. Rev. 5, 011301 (2018). 3 P. Jiang, X. Qian, X. Li, and R. Yang, Appl. Phys. Lett. 113, 232105 (2018). 4 Z. Cheng, N. Tanen, C. Chang, J. Shi, J. McCandless, D. Muller, D. Jena, H. G. Xing, and S

work page 2018

[2] [2]

Graham, Appl. Phys. Lett. 15 (8), (2019). 5 Z. Cheng, L. Yates, J. Shi, M. J. Tadjer, K. D. Hobart, and S. Graham, APL Materials 7, 031118 (2019). 6 S. B. Reese, T. Remo, J. Green, and A. Zakutayev, Joule, 3, 4 (2019). 7 J. Noh, M. Si, H. Zhou, M . J. Tadjer, and D. Y. Peide, 76th Device Research Conference (DRC), pp. 1-2. IEEE, (2018). 8 H. Zhou, K. Maiz...

work page 2019

[3] [3]

Foley, S

Zivasatienraj, B. Foley, S. T. Pantelides, and L. F. Piper, Phys. Rev. Mater. 3, 025002 (2019). 14 14 J. T. Gaskins, G. Kotsonis, A. Giri, C. T. Shelton, E. Sachet, Z. Cheng, B. M. Foley, Z. Liu, S. Ju, and M. S. Goorsky, Nano Lett. 18, 12, 7469, (2017). 15 P. Jiang, X. Qian, and R. Yang, J. of Appl. Phys. 124, 161103 (2018). 16 F. Mu, Z. Cheng, J. Shi, S...

work page 2019