Pith. sign in

REVIEW 3 major objections 5 minor 8 references

Resolving the Nucleation Stage in Atomic Layer Deposition of Hafnium Oxide on Graphene

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 Atomic-scale imaging shows hafnium oxide nucleates on adventitious carbon contamination from ambient air, which acts as an unintentional seeding layer.

desk verdict Strong microscopy and a plausible, well-hedged carbon-seeding story, but the causal claim rests on an unproven assumption; deserves review and a careful revision. read the letter →

arxiv 1909.00712 v1 pith:XXTPY4CU submitted 2019-09-02 cond-mat.mtrl-sci

Bernhard C. Bayer , Adrianus I. Aria , Dominik Eder , Stephan Hofmann , Jannik C. Meyer This is my paper · ORCID
classification cond-mat.mtrl-sci
keywords atomiclayerdepositiongraphenehafniumoxidenucleationadventitiouscarbonscanningtransmissionelectronmicroscopysubstrate-assistedALDhigh-kdielectric
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

Using atomically resolved electron microscopy, this paper tries to establish why hafnium oxide nucleates where it does when it is grown by atomic layer deposition (ALD) on graphene that has not been transferred off its copper growth catalyst. The central claim is that under realistic, non-ultra-high-vacuum conditions the oxide does not nucleate on clean graphene basal planes: it starts on adventitious carbon contamination picked up from ambient air, which therefore acts as an unintentional seeding layer. The paper also argues that the copper support participates catalytically in a graphene-layer-dependent way, with monolayer graphene acquiring far more HfO$_2$ than bilayer or few-layer regions, and that graphene layer steps act as chemically reactive nucleation lines. If this is right, ambient-air exposure history is a first-order control variable in ALD coating of two-dimensional materials, alongside precursor chemistry and temperature.

What carries the argument

The decisive object is the atomic-number-contrast HAADF-STEM image: carbon (Z=6) is dark, hafnium (Z=72) is bright, and the graphene lattice can be resolved, so each nucleation site can be classified as sitting on bare graphene, on adventitious carbon, or on the HfO$_2$ clusters themselves. The analytical step that carries the argument is the areal-fraction comparison encapsulated in $\theta_{\mathrm{contamination}} + \theta_{\mathrm{oxide}} = 1 - \theta_{\mathrm{graphene}}$, together with the observation that $\theta_{\mathrm{graphene}}$ stays nearly constant while $\theta_{\mathrm{oxide}}$ changes strongly. That combination rules out the competing "HfO$_2$ attracts carbon" scenario and identifies the carbon as the pre-existing seed.

What would settle it

Image the identical graphene region before and after ALD, or after in-situ removal of the adventitious carbon: if HfO$_2$ appears on regions that were atomically clean before deposition, or if carbon is absent underneath freshly nucleated clusters, the "carbon attracts HfO$_2$" scenario is wrong; a companion test is an UHV-clean, ambient-free ALD run, which should strongly suppress carbon-correlated nucleation.

Watch

Extended reading notes

Core claim

The paper reports direct atomic-scale HAADF-STEM evidence that ALD HfO$_2$ on chemical-vapor-deposited graphene nucleates preferentially on pre-existing adventitious carbon rather than on atomically clean graphene. The authors explicitly weigh two scenarios, "HfO$_2$ attracts carbon" versus "carbon attracts HfO$_2$", and conclude for the latter from the visual spatial continuity of a carbon network underneath the oxide and from measured areal fractions: the bare-graphene fraction stays roughly constant (about 17–24%) across regions that differ strongly in HfO$_2$ coverage and across ALD runs with 6 versus 16 cycles, whereas the oxide coverage changes from about 54% on monolayer graphene to about 20% on bilayer graphene. They attribute the layer-number dependence to the copper catalyst's electronic influence decaying through additional graphene layers. The as-deposited HfO$_2$ is amorphous and, under the electron beam, crystallizes to metastable cubic/tetragonal polymorphs rather than the equilibrium monoclinic phase, and no epitaxial relationship with the graphene lattice appears.

Load-bearing premise

The carbon-seeding conclusion assumes the carbon seen under and around the HfO$_2$ was already on the graphene before deposition and formed a continuous layer underneath the oxide, rather than accumulating at HfO$_2$ sites during or after ALD or being redistributed by the electron beam.

Editorial extensions

If this is right

  • Ambient air exposure before or during ALD should be treated as a deliberate process parameter, because it controls where and how densely HfO$_2$ nucleates.
  • Repeatedly interrupting ALD with air exposure should incrementally contaminate still-clean graphene and thereby homogenize subsequent oxide coating, matching the earlier Al$_2$O$_3$ result cited in the paper.
  • On Cu-supported graphene, monolayer regions will coat more completely than bilayer or few-layer regions, so the areal uniformity of the graphene layer number sets an upper bound on oxide film uniformity.
  • Graphene layer steps and folds will collect thicker oxide and may act as leakage or contact points in devices, since they are preferential nucleation lines.
  • Thin ALD HfO$_2$ films on graphene start amorphous and, when crystallized, favor metastable cubic/tetragonal phases, which is relevant because those phases have higher dielectric constants than monoclinic HfO$_2$.

Reading between the lines

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

  • Editorial inference: the carbon-seeding mechanism is likely to transfer to other ALD oxide/2D-material combinations where ambient exposure is not controlled, since the paper's cited XPS work already connects it to graphite and MoS$_2$; a controlled comparative study would make the generality quantitative.
  • Editorial inference: if contamination is the true seed, deliberately depositing a uniform carbonaceous or oxygen-functional adlayer before ALD could convert a source of run-to-run irreproducibility into a reproducible nucleation-density knob.
  • Editorial inference: the cleanest falsification the authors could not perform in situ is a same-region before/after comparison; with a heating stage or in-situ cleaning in the microscope one could strip the adventitious carbon, deposit ALD cycles, and check whether HfO$_2$ then nucleates elsewhere or not at all.
  • Editorial inference: the electron-beam crystallization to cubic/tetragonal HfO$_2$ may not reproduce furnace annealing, so the polymorph claim for device processing should be checked with conventional thermal anneals before relying on the high-k phases.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper reports an aberration-corrected STEM study of the nucleation stage of HfO2 grown by substrate-assisted ALD directly on CVD graphene that remains on its Cu growth foil. The authors identify several factors governing nucleation: a graphene-layer-dependent catalytic participation of the Cu support, preferential nucleation at graphene layer steps, a Volmer–Weber growth mode with amorphous as-deposited clusters, e-beam-induced crystallization to metastable cubic/tetragonal HfO2, and no preferential nucleation at graphene grain boundaries. The central mechanistic claim is that atmospheric adventitious carbon contamination acts as an unintentional seeding layer: HfO2 nucleates preferentially on pre-existing carbon deposits rather than on clean graphene basal planes. This claim is supported by spatial correlation between HfO2 clusters and carbon-like contrast, by approximately constant bare-graphene areal fractions across regions with very different HfO2 coverages, and by a comparison of 16-cycle and 6-cycle ALD samples together with a non-ALD control.

Significance. If the carbon-seeding conclusion holds, it is practically important for ALD process design on 2D materials, since adventitious carbon is ubiquitous in non-UHV processing. The paper's strengths include the use of polymer-free sample preparation, atomic-resolution and element-specific STEM imaging, direct quantification of areal coverages, the observation of non-equilibrium HfO2 polymorphs, and several control observations (non-ALD sample, grain-boundary comparison, layer-step comparison). The manuscript contains no fitted parameters or circular derivations, and the imaging evidence for layer-dependent Cu assistance and step-edge nucleation is visually compelling. However, the central causal inference about pre-existing carbon is presently supported by a stated unproven assumption and by single-image statistics, so the significance of the paper depends on strengthening that link.

major comments (3)
  1. [Results and Discussion, paragraph following Fig. 4d] The conclusion that scenario (ii) 'Carbon attracts HfO2' is correct rests entirely on the assumption that 'the adventitious carbon adsorbates are a continuous layer underneath the HfO2 deposits', which the authors introduce as a plausible assumption based on visual spatial continuity. The measured quantity θ_graphene alone cannot distinguish this scenario from scenario (i) 'HfO2 attracts carbon', because carbon that accumulates on top of or at the edges of HfO2 clusters after ALD is not counted as bare graphene and therefore does not affect θ_graphene. The relation θ_contamination + θ_oxide = 1 − θ_graphene is thus compatible with both scenarios. A direct test of carbon pre-existence is needed, for example by characterizing the graphene surface before ALD on identically prepared samples, by performing ALD on samples with controlled levels of ambient exposure, or by atomic-scale elemental mapping of the carbon/HfO2 interface to show that the carbon layer lies beneath the HfO2 rather than on top of it.
  2. [Results and Discussion, Fig. 4b–d and surrounding text] The areal coverage fractions are reported as single values with no error bars and no statement of how many independent images or sample regions were analyzed: θ_oxide,monolayer ≈ 54%, θ_oxide,bilayer ≈ 20%, θ_graphene,monolayer ≈ 17%, θ_graphene,bilayer ≈ 22%, θ_oxide,6 cycles ≈ 15%, θ_graphene,6 cycles ≈ 24%, and θ_graphene,no ALD ≈ 25–30%. The central argument that θ_graphene is 'roughly constant' despite strongly varying θ_oxide is the quantitative linchpin of the carbon-seeding conclusion, but differences of 5–8 percentage points between the quoted values are not distinguishable from measurement noise without sampling statistics. The 6-cycle versus 16-cycle comparison also uses different sample regions and lacks paired statistics, so it cannot robustly corroborate the constancy of θ_graphene.
  3. [Results and Discussion, paragraph preceding Fig. 4] The authors note that e-beam exposure can induce contamination dewetting and/or attraction of additional contamination, and they state that such changes were minimal for their imaging conditions, but no quantitative evidence is provided. Because the observed spatial correlation between HfO2 clusters and carbon-like contrast is the primary evidence for the seeding mechanism, beam-induced carbon redistribution during STEM imaging could in principle produce the same apparent correlation even if HfO2 had nucleated on clean graphene. The authors should report low-dose control images or otherwise quantify the stability of the carbon distribution under the exact scan conditions used for the correlation analysis.
minor comments (5)
  1. [Fig. 4c caption and text] The phrase 'a 5 few layer graphene region' in the Fig. 4c description is unclear; it presumably means a few-layer graphene region of five layers, and this should be stated explicitly.
  2. [Results and Discussion, paragraph following Fig. 2d] The text states the HAADF intensity 'scales linearly with specimen thickness' and also 'provides materials contrast which is dependent on Z with Z~1.64'. The notation Z~1.64 is ambiguous; it should be written as Z^1.64 to indicate the power-law dependence.
  3. [SAED phase identification, Fig. 3d and Supporting Fig. S3] The phase assignment to cubic and/or tetragonal HfO2 is made on a qualitative presence/absence basis. The authors should clarify whether the data can distinguish between cubic and tetragonal, and if not, state that the two phases are indistinguishable at the available resolution.
  4. [Methods, transfer procedure] The polymer-free direct transfer is cited to reference 40 (Regan et al.), but it is worth stating explicitly that no polymer support layer is used in that method, since the absence of polymer residues is a key premise of the paper.
  5. [Throughout] Several minor grammatical issues appear (e.g., 'the salient features observed in image labelled' in the Fig. 2d caption, and 'under conditions' instead of 'under our conditions' in the Conclusions). A careful proofreading pass is recommended.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the central nucleation claim is an experimental inference, not a definitional or fitted tautology.

full rationale

No circular step meets the required bar of reducing to the paper's own inputs by construction or by renaming a fitted parameter as a prediction. The central claim that adventitious carbon acts as an unintentional seeding layer for ALD HfO2 on graphene is inferred from HAADF-STEM spatial correlation, measured bare-graphene fractions across regions with different HfO2 coverage, a 6-cycle versus 16-cycle comparison, and no-ALD control samples. These are independent observations rather than definitions. The relation θ_contamination + θ_oxide = 1 – θ_graphene is an accounting identity among classified image areas, but the paper does not use that identity alone to force the 'Carbon attracts HfO2' conclusion; the conclusion explicitly relies on the stated auxiliary assumption that carbon is a continuous layer underneath the HfO2 deposits. That assumption is an evidentiary/causal inference and could be wrong, but it is not circular in the sense that the conclusion is assumed in the premise. The substrate-assisted ALD mechanism is attributed to prior work by overlapping authors (refs 28 and 34), but it is also directly tested here through the measured graphene-layer-dependent HfO2 coverage in Fig. 4 and Supporting Figures S4/S5, so the self-citation is not load-bearing. No uniqueness theorem is imported from the authors' prior work, no ansatz is smuggled in solely via citation, and no known empirical pattern is merely renamed. The paper is self-contained against its own STEM/TEM/SEM evidence; weaknesses in the continuity assumption belong to correctness risk, not circularity.

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

The central claims rest on standard microscopy contrast assumptions, prior work on substrate-assisted ALD, and the presence of ambient carbon contamination. The carbon-seeding conclusion additionally depends on the stated but untested assumption that carbon seen near HfO2 is continuous underneath and pre-existed ALD. No free parameters are fitted, since the paper is observational and contains no equations.

assumptions (4)
  • domain assumption HAADF STEM intensity is proportional to specimen thickness and scales with atomic number (Z^1.64) for the studied materials.
    Used to identify HfO2 clusters (high Z) and carbon contamination (low Z) in Fig. 2 and to quantify coverage fractions. Standard STEM imaging principle, but the specific Z-power and absence of diffraction contrast in the thin amorphous clusters are assumed.
  • domain assumption The Cu catalyst's electronic properties emanate through monolayer graphene and partake in the ALD reactions, with diminished contribution through bilayer or few-layer graphene.
    Invoked to explain the higher HfO2 coverage on monolayer graphene versus bilayer or few-layer graphene. This is based on prior substrate-assisted ALD work (refs 28, 34, 39) and not directly measured in this paper.
  • domain assumption Adventitious carbon contamination is present on non-UHV processed graphene before ALD and appears as a continuous layer underneath the HfO2 deposits.
    The paper infers pre-ALD presence from the roughly constant bare-graphene fraction across differing HfO2 coverages and from visual continuity of the carbon network. The continuity and pre-existence are stated as plausible assumptions, not directly verified.
  • domain assumption The thin low-Z contrast near HfO2 clusters is adventitious carbon, not polymer residue or beam-induced carbon redistribution.
    Polymer residues are avoided by the transfer method, and the authors note e-beam changes were minimal. However, no spectroscopic confirmation of the chemical nature of the carbon was performed at the imaged locations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Resolving the Nucleation Stage in Atomic Layer Deposition of Hafnium Oxide on Graphene." pith.science (2026). https://pith.science/paper/XXTPY4CU

@misc{pith2026190900712,
  author       = {Pith},
  title        = {Pith review of: Resolving the Nucleation Stage in Atomic Layer Deposition of Hafnium Oxide on Graphene},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XXTPY4CU}},
  note         = {Machine review of arXiv:1909.00712}
}
read the original abstract

The integration of two-dimensional (2D) materials with functional non-2D materials such as metal oxides is of key importance for many applications, but underlying mechanisms for such non-2D/2D interfacing remain largely elusive at the atomic scale. To address this, we here investigate the nucleation stage in atomic layer deposition (ALD) of the important metal oxide HfO2 on chemical vapor deposited graphene using atomically resolved and element specific scanning transmission electron microscopy (STEM). To avoid any deleterious influence of polymer residues from pre-ALD graphene transfers we employ a substrate-assisted ALD process directly on the as grown graphene still remaining on its Cu growth catalyst support. Thereby we resolve at the atomic scale key factors governing the integration of non-2D metal oxides with 2D materials by ALD: Particular to our substrate-assisted ALD process we find a graphene-layer-dependent catalytic participation of the supporting Cu catalyst in the ALD process. We further confirm at high resolution the role of surface irregularities such as steps between graphene layers on oxide nucleation. Employing the energy transfer from the scanning electron beam to in situ crystallize the initially amorphous ALD HfO2 on graphene, we observe HfO2 crystallization to non-equilibrium HfO2 polymorphs (cubic/tetragonal). Finally our data indicates a critical role of the graphene's atmospheric adventitious carbon contamination on the ALD process whereby this contamination acts as an unintentional seeding layer for metal oxide ALD nucleation on graphene under our conditions. As atmospheric adventitious carbon contamination is hard to avoid in any scalable 2D materials processing, this is a critical factor in ALD recipe development for 2D materials coating. Combined our work highlights several key mechanisms underlying scalable ALD oxide growth on 2D materials.

Figures

Figures reproduced from arXiv: 1909.00712 by the authors.

Figure 1
Figure 1. Schematic illustration of the hypothesized governing factors in substrate-assisted ALD34 of HfO2 on graphene. Results and Discussion We grow graphene by CVD on Cu catalyst foils49,50 using conditions that lead to a continuous polycrystalline graphene film of predominantly monolayer graphene with an appreciable amount of bilayer and few layer graphene islands. This film structure allows us to readily study the depend… view at source ↗
Figure 4
Figure 4. (a) HAADF STEM image of ALD HfO2 deposits (16 cycles) over the transition from a graphene bilayer region (left) to a graphene monolayer region (right). The insets show HAADF data at higher magnification confirming the turbostratically stacked bilayer graphene lattice (left) and the monolayer graphene lattice (right), respectively. (b) False-color coded recalculation of [PITH_FULL_IMAGE:figures/full_fig_p014_4.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

8 extracted references · 8 canonical work pages

  1. [1]

    (1) Kim, S.; Nah, J.; Jo, I.; Shahrjerdi, D.; Colombo, L.; Yao, Z.; Tutuc, E.; Banerjee, S. K. Realization of a High Mobility Dual -Gated Graphene Field-Effect Transistor with Al 2 O 3 Dielectric. Appl. Phys. Lett. 2009, 94, 062107. (2) Radisavljevic, B.; Radenovic, A.; Brivio, J.; Giacometti, i V.; Kis, A. Single -Layer MoS2 Transistors. Nat. Nanotechnol...

    work page 2009

  2. [2]

    Monoclinic HfO 2 is the thermodynamically most stable HfO2 polymorph whereas cubic and tetragonal HfO 2 are only metastable non-equilibrium polymorphs.S3–S6 S6 Supporting Figure S4. Scanning electron microscopy (SEM) images of HfO 2 deposits from substrate-assisted ALD arrested in the nucleation stage (16 cycles) on CVD graphene at various magnifications ...

    work page 2007

  3. [8]

    Amorphous Hafnium Oxide Films from High Rate Room Temperature Deposition. Appl. Phys. Lett. 2011, 98, 252903. (S6) Li, F. M.; Bayer, B. C.; Hofmann, S.; Speakman, S. P.; Ducati, C.; Milne, W. I.; Flewitt, A. J. High -Density Remote Plasma Sputtering of High -Dielectric-Constant Amorphous Hafnium Oxide Films. Phys. Status Solidi B 2013, 250, 957–967. (S7) ...

    work page 2011

  4. [30]

    Amorphous Hafnium Oxide Films from High Rate Room Temperature Deposition. Appl. Phys. Lett. 2011, 98, 252903. (61) Li, F. M.; Bayer, B. C.; Hofmann, S.; Speakman, S. P.; Ducati, C.; Milne, W. I.; Flewitt, A. J. High -Density Remote Plasma Sputtering of High -Dielectric-Constant Amorphous Hafnium Oxide Films. Phys. Status Solidi B 2013, 250, 957–967. (62) ...

    work page 2011

  5. [35]

    R.; Bayer, B

    (6) Meyer, J.; Kidambi, P. R.; Bayer, B. C.; Weijtens, C.; Kuhn, A.; Centeno, A.; Pesquera, A.; Zurutuza, A.; Robertson, J.; Hofmann, S. Meta l Oxide Induced Charge Transfer Doping and Band Alignment of Graphene Electrodes for Efficient Organic Light Emitting Diodes. Sci. Rep. 2014, 4,

    work page 2014

  6. [218]

    A Direct Transfer of Layer -Area Graphene

    (40) Regan, W.; Alem, N.; Alemán, B.; Geng, B.; Girit, Ç.; Maserati, L.; Wang, F.; Crommie, M.; Zettl, A. A Direct Transfer of Layer -Area Graphene. Appl. Phys. Lett. 2010, 96, 113102. (41) Krivanek, O. L.; Chisholm, M. F.; Nicolosi, V.; Pennycook, T. J.; Corbin, G. J.; Dellby, N.; Murfitt, M. F.; Own, C. S.; Szilagyi, Z. S.; Oxley, M. P.; Pantelides, S. ...

    work page 2010

  7. [4781]

    I.; Kidambi, P

    (64) Aria, A. I.; Kidambi, P. R.; Weatherup, R. S.; Xiao, L.; Williams, J. A.; Hofmann, S. Time Evolution of the Wettability of Supported Graphene under Ambient Air Exposure. J. Phys. Chem. C 2016, 120, 2215–2224. (65) Miller, D.; Biesinger, M.; McIntyre, N. Interactions of CO2 and CO at Fractional Atmosphere Pressures with Iron and Iron Oxide Surfaces: O...

    work page 2016

  8. [5380]

    -Q.; Xu, A

    (7) Wang, W.- S.; Wang, D.- H.; Qu, W.- G.; Lu, L. -Q.; Xu, A. -W. Large Ultrathin Anatase TiO2 Nanosheets wi th Exposed {001} Facets on Graphene for Enhanced Visible Light Photocatalytic Activity. J. Phys. Chem. C 2012, 116, 19893–19901. (8) Kim, H. G.; Lee, H.-B.-R. Atomic Layer Deposition on 2D Materials. Chem. Mater. 2017, 29, 3809–3826. (9) Vervuurt,...

    work page 2012

Pith tools

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