Electron Channelling Contrast SEM Imaging of Twist Domains in Transition Metal Dichalcogenide Heterostructures

Andrew Pollard; Astrid Weston; Evan Tillotson; James Howarth; James McHugh; Matthew Lindley; Nick Clark; Roman Gorbachev; Samuel Sullivan-Allsop; Sarah J. Haigh

REVIEW 3 major objections 6 minor 15 references

Electron Channelling Contrast SEM Imaging of Twist Domains in Transition Metal Dichalcogenide Heterostructures

T0 review · 3 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 Conventional SEM, via electron channelling contrast, can map reconstructed twist domains in TMD heterostructures even when they are sealed beneath an encapsulation layer.

desk verdict A practical, experimentally strong SEM imaging method for twist domains; the mechanistic model has a fitting soft spot that needs disclosure and softer claims. read the letter →

arxiv 2411.16248 v1 pith:HQZUJOVA submitted 2024-11-25 cond-mat.mtrl-sci

Evan Tillotson , James McHugh , James Howarth , Teruo Hashimoto , Nick Clark , Astrid Weston , Vladimir Enaldiev , Samuel Sullivan-Allsop

show 7 more authors

William Thornley Wendong Wang Matthew Lindley Andrew Pollard Vladimir Falko Roman Gorbachev Sarah J. Haigh

This is my paper · ORCID

classification cond-mat.mtrl-sci PACS 68.37.Hk

keywords twistronicstransitionmetaldichalcogenideselectronchannellingcontrastimagingscanningmicroscopymoirésuperlattice3RstackingdomainshBNencapsulationMichelson

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 argues that an ordinary scanning electron microscope, using electron channelling contrast, can image the reconstructed stacking domains of twisted transition-metal dichalcogenide bilayers, including domains buried under a hexagonal boron nitride encapsulation layer. The central experimental claim is that this non-destructive method works on as-produced samples on conventional substrates and does not require knowing the crystal orientation in advance. To explain why domain brightness reverses between detectors and with sample tilt, the paper combines elastic Bragg scattering between the two layers with a Monte Carlo model of inelastic secondary-electron generation and attenuation. If the picture is right, the twistronics community gains a cheap, widely available tool for correlating local domain architecture with device behaviour.

What carries the argument

The load-bearing object is a two-layer interference model for elastically Bragg-scattered partial waves, expressed as $I_n = 2A^2(1 + \cos(\mathbf{G}_n \mathbf{r}_0 - \Delta\varphi_n))$, where $\Delta\varphi_n = d(k_0 - \sqrt{k_0^2 - G_n^2})$ is the phase difference accumulated because the top and bottom layers sit a distance $d$ apart, and the interlayer offset $\mathbf{r}_0$ is what distinguishes the XM' and MX' stackings. This predicts which Bragg reflections carry domain contrast and at which scattering angles they appear. A separate semi-classical Monte Carlo model treats inelastic signal as secondary electrons generated in proportion to local electron density and attenuated along their escape path using empirical inelastic mean-free-path data. Combining the two scattering channels with an empirically determined weighting reproduces the measured angular dependence of STEM detector contrast, including the inversion near 20 degrees.

What would settle it

Measure the STEM detector domain contrast versus collection angle for a twisted bilayer whose stacking assignment is known from an independent probe, with the elastic/inelastic mixing weight fixed by the scattering calculation alone rather than fitted to the same images; if the zero-contrast crossing does not fall near 20 degrees, the proposed balance of elastic and inelastic scattering is wrong.

Watch

Extended reading notes

Core claim

For freely suspended twisted MoS2 bilayers at 1500 V, low-angle transmitted electrons give negative domain contrast dominated by the {1-100} Bragg reflection, while high-angle transmitted and backscattered electrons give positive contrast dominated by inelastic scattering; the zero crossing sits near 20 degrees scattering angle. The sign of contrast is therefore not intrinsic to a domain but depends on which scattering population the detector collects. On bulk substrates the reflected signals behave similarly for both low- and high-energy detectors, because the substrate adds positive elastic and inelastic contributions from electrons crossing the bilayer in reverse; on ultra-thin supports the ETD signal can appear inverted. A specimen tilt near 21 degrees multiplies contrast by roughly four and reverses its sign, and encapsulation by 3.5 nm of hBN reduces contrast by about half while leaving domains visible. The paper concludes that optimum imaging is obtained by maximising collected signal while avoiding the angular and tilt conditions where contrast inverts.

Load-bearing premise

The calculated position of the contrast inversion depends on an empirically chosen weighting that mixes the elastic and inelastic signals to match the measured contrast values, so the theory's prediction near 20 degrees is not fully independent of the data it explains.

Editorial extensions

If this is right

Twist domains in finished, encapsulated devices can be checked non-destructively, so domain architecture can be correlated with measured electronic properties of the same device.
Detector choice matters: annular backscattered-electron detectors give consistent positive contrast at zero tilt and do not require rotating the sample toward the detector, while ETD images can mislead unless the tilt and substrate regime is understood.
Tilting the sample by about 21 degrees can quadruple domain contrast, and the accompanying sign flip provides an internal consistency check that the observed contrast really comes from the stacking domains.
The same BSE channelling approach should map domains in WS2 with higher contrast than MoS2, and remains usable for lower-layer thicknesses up to at least three layers.

Reading between the lines

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

Beyond the paper, a robust inversion angle would give a scattering-based fingerprint of stacking registry that does not depend on image histogram calibration.
Beyond the paper, the ETD result implies that in atomically thin samples the conventional secondary-electron detector signal is largely reporting channelled higher-energy electrons, so ETD micrographs of such samples should not be read as pure surface topography.
Beyond the paper, the same combined elastic and inelastic contrast logic should apply to other layered systems with two distinct registries, such as twisted graphene on hBN, wherever a low-index Bragg reflection falls inside the accessible detector angle.
Beyond the paper, serial imaging after plasma cleaning showed contrast recovery, suggesting that quantitative contrast versus contamination thickness could turn ECCI into a calibrated probe of surface cleanliness.

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.

Desk Editor's Note

read the letter

Bottom line: this is a useful, mostly convincing experimental paper. The SEM ECCI method for mapping twist domains in TMDs is a real practical advance, and the paper does a thorough job of showing when it works and when it doesn't. The theoretical story is more fragile than the abstract implies, but that doesn't sink the main result.

What's new here is the systematic scope. Earlier work had shown ECCI on a few twisted 2D samples, but this paper maps contrast across detector type and angle, sample tilt, substrate, encapsulation, layer thickness, and two material systems, and demonstrates transfer across several SEM platforms. The encapsulation result is the one that people will actually use: being able to see twist domains non-destructively in a finished device is a genuine capability gain.

The experimental core is solid. Contrast values come with propagated errors, the sample-flip control confirms the domain assignments, and the qualitative picture—negative elastic contrast at low scattering angles, positive inelastic contrast at high angles—is a useful organizing framework. The EsB detector working without tilt is a practical point that deserves emphasis.

The soft spot is the 'empirically determined weighting' in Figure 2d. One (or a few) fitted numbers connect the elastic and inelastic contributions, and the paper does not report their values, the fitting criterion, or any sensitivity test. Because the experimental data themselves contain the zero crossing, locating the predicted inversion at ~20° is at least partly constructed, not derived. The abstract's 'reveal' overstates the predictive strength. To their credit, the authors admit the high-angle discrepancy and mark the substrate ETD explanation as a hypothesis, so this is an omission of transparency rather than a hidden fatal flaw. A referee should ask for the weight to be reported, a sensitivity sweep, and more careful wording in the abstract and conclusions.

Who is this for: anyone doing twistronics device characterization. The method is cheap, fast, and non-destructive, and the paper gives enough detail to reproduce it. It deserves a serious referee, not a desk reject. With the fitting made transparent and the claims softened, I'd take it as published.

Referee Report

3 major / 6 minor

Summary. The manuscript demonstrates that scanning electron microscope electron channelling contrast imaging (SEM ECCI) can reveal reconstructed twist domains in twisted TMD heterostructures, including samples sealed beneath hBN encapsulation, on bulk substrates, and for thicker TMD stacks. The experimental work is extensive: it quantifies domain contrast with propagated errors for multiple detectors (EsB, InL, ETD, and annular STEM segments), maps contrast versus scattering angle, studies the dependence on working distance, accelerating voltage, tilt, rotation, substrate, and encapsulation, and validates the approach on several SEM platforms and on both MoS2 and WS2. The paper also presents a theoretical model that combines elastic Bragg scattering and a semi-classical Monte Carlo treatment of inelastic scattering to explain the observed sign of domain contrast and its inversion near 20° scattering angle. The central mechanistic claim, however, rests on an 'empirically determined weighting' that is not described or justified, and the theoretical account of tilt-dependent contrast is absent. The experimental demonstration is convincing and likely useful, but the theoretical explanation is not yet at the level implied by the abstract.

Significance. If the experimental claims are accepted, this is a practical and widely accessible characterization route for twist domains: it is non-destructive, works on conventional substrates, and crucially works under hBN encapsulation, which is the geometry relevant to real devices. The systematic quantification of contrast (Michelson contrast with propagated errors) across detectors, tilts, thicknesses, and instruments is a genuine strength. The paper also shows that the method does not require precise azimuthal alignment at zero tilt, and it quantifies contrast degradation from encapsulation and contamination. The theoretical modelling is less convincing: the elastic-scattering analysis is ab initio and parameter-light, but the inelastic model is semi-classical and is combined with the elastic contribution using an unspecified empirical weight that is evidently chosen to match the very data it is used to explain. Consequently the claimed prediction of the ~20° contrast-inversion angle is not independent. The strength of the paper is therefore primarily experimental; the mechanistic explanation needs substantial reframing or additional validation.

major comments (3)

[Domain contrast in suspended samples; Fig. 2d; SI Section 3.2] The theoretical points in Fig. 2d are obtained 'by combining elastic and inelastic scattering contributions with an empirically determined weighting', but the manuscript does not state the value of the weight, how many weights were adjusted, the fitting criterion, or any sensitivity analysis. Since the elastic contribution at 1500 eV is negative for all relevant angles (SI Fig. S3.4d and S3.5) and the inelastic contribution has a trough near 16° and peaks near 22° and 30° (Fig. 2c), a single weight chosen to reproduce the experimental low-angle and high-angle contrast magnitudes can place the zero crossing at ~20° essentially by construction. The abstract's claim that the calculations 'reveal' contrast inversions at specified detector scattering angles is therefore not a parameter-free prediction. Please report the fitted weighting, show the sensitivity of the predicted inversion angle to that weight, or fix it a priori from the model rather than from the data being explained.
[Domain contrast as a function of sample tilt; Fig. 4; SI Section 3.2.4] The abstract also claims that the calculations reveal contrast inversions 'at specified detector scattering angles and sample tilts', but no theoretical calculation of contrast versus specimen tilt is presented. Figure 4 shows experimental tilt dependence (inversion near 11°, optimum near 21°), and SI Fig. S3.16 shows emission intensity versus azimuthal angle for fixed emission angles, which is not the same as a tilt series. The theoretical model in Fig. 2c treats scattering angle relative to the incident beam, not the sample tilt angle. Either add a calculation that predicts the tilt dependence or remove 'sample tilts' from the abstract's mechanistic claim.
[Conclusions; SI Sections 3.2 and 3.2.4] The inelastic scattering model is semi-classical and omits core-electron ionization and plural scattering, which are invoked to explain the discrepancy at high scattering angles. The paper does not show that the inelastic model, without the empirical weighting, reproduces any independent observable (e.g., absolute angular distributions or energy dependence of the contrast). Given that the contrast inversion is the key mechanistic result, the model should be validated against data other than the single fitted curve in Fig. 2d, or the theoretical claim should be softened to a qualitative consistency check rather than a quantitative prediction.

minor comments (6)

[Conclusions] 'preferrable' should be 'preferable'.
[SI Section 3.1] Typo: 'c4ontributions' should be 'contributions'.
[Throughout] The contrast metric is spelled 'Michaelson' in several places but 'Michelson' in the SI; please standardize.
[SI Section 2] The main text defines contrast using IXM' and IMX' while the SI uses I1 and I2; please align the notation so the reader can map between the two.
[Figure 4 caption] 'there is a contrast is reversal at ~11°' is ungrammatical; should read 'there is a contrast reversal at ~11°'.
[Domain contrast in suspended samples] The explanation for the absence of ETD contrast mentions the need to include high-energy inelastic core-electron scattering for a full treatment; this limitation should be stated in the main text near the model description, not only in the SI.

Circularity Check

1 steps flagged · score 6.0 of 10

The predicted ~20° contrast-inversion angle in Fig. 2d is partly calibrated to the data: the theory curve is produced by combining elastic and inelastic contributions with an unreported 'empirically determined weighting,' so the mechanistic claim is not fully independent.

fitted input called prediction [Results, 'Domain contrast in suspended samples', paragraph after Fig. 2c and Fig. 2d caption]
"To understand the unusual contrast behaviour, and specifically the zero-contrast inversion point at ~20°, we have performed theoretical calculations of both elastic and inelastic scattering contributions to domain contrast. ... It is therefore possible to reproduce a good match to the experimentally observed STEM detector contrast behavio ur for scattering angle s up to θ~35° by combining elastic and inelastic scattering contributions with an empirically determined weighting, illustrated by the black dots in Figure 2d."

At 1500 eV the elastic contribution is negative for all relevant scattering angles (SI Fig. S3.4d and S3.5), while the inelastic contribution in Fig. 2c already contains the sign-changing structure: a trough near ~16°, positive peaks near ~22° and ~30°. A single empirical weight in the sum controls the relative amplitude of these two curves and therefore where the combined contrast crosses zero. Choosing the weight to match the experimental STEM2 and STEM5 contrast values places the zero crossing near the experimentally observed ~20° essentially by construction.

full rationale

The core experimental results are self-contained: ECCI images are measured directly, Michelson contrasts are quantified from pixel histograms, and the encapsulation, thickness, and tilt trends are direct observations. The domain-type labelling (XM' vs MX') is inherited from the authors' previous work [32], but that is a prior experimental assignment supplemented by the sample-flip test, rather than a load-bearing derivation of the new imaging claim. The only substantial circularity is in the theoretical contrast-inversion mechanism. The black calculated points in Fig. 2d are obtained by summing independently computed elastic and inelastic angular curves with an 'empirically determined weighting.' Because the elastic contribution is single-signed at 1500 eV and the inelastic contribution already contains the trough-to-peak sign change, the empirical weight, not the calculation alone, determines the location of the predicted zero crossing. Without the fitted weight value or a sensitivity analysis, the claimed prediction of a ~20° inversion is partially calibrated to the same STEM detector data it is used to explain. This weakens the mechanistic claim in the abstract but does not undermine the experimental demonstration that SEM ECCI can image twist domains, including under encapsulation. Score 6 reflects partial, not total, circularity.

Assumptions & free parameters 1 free parameters · 6 assumptions · 0 invented entities

The central imaging demonstration rests on sample fabrication and SEM operation, while the explanatory model rests on standard scattering approximations plus one fitted weighting to match the observed detector-angle dependence. No new physical entities are introduced.

free parameters (1)

empirical weighting for elastic/inelastic combination = not specified
Used to combine elastic and inelastic scattering contributions to reproduce the experimental STEM contrast versus angle in Figure 2d; the paper calls it an 'empirically determined weighting' without giving a value or uncertainty.

assumptions (6)

domain assumption Secondary electron generation and attenuation are both proportional to local electron density along the emission path (Eq. 9, SI Section 3.2.3).
This semi-classical Beer-Lambert model is a simplification; it omits core electron ionization and plural scattering, which the authors note become important at high angles (SI Section 3.2.4).
domain assumption The TPP-2M semiempirical formula gives the inelastic mean free paths for MoS2, WS2, and hBN in the 50-2000 eV range (Eq. 8, SI Section 3.2.2).
Borrowed from prior literature; the paper validates against experimental IMFP data.
standard math Elastic scattering is described in the Born approximation with Thomas-Fermi atomic potentials (Eq. 7, SI Section 3.1).
A standard approximation for fast electron diffraction; the paper acknowledges its limits.
domain assumption The reflected elastic signal from the bilayer is negligible (four orders of magnitude smaller) compared to the transmitted signal (SI Figure S3.7).
Used to ignore reflected elastic contributions in the model; assumed to hold at 1500 eV.
domain assumption Domain identity (which stacking is XM' versus MX') follows from previous work by the same group (ref [32]).
The assignment of positive/negative contrast signs relies on this external identification; the paper uses it to label contrast polarity.
domain assumption The ETD detector signal for supported samples originates from low-energy secondary electrons generated by higher-energy channelling electrons, not from channelling of the low-energy electrons themselves.
Supported by escape-depth arguments and the parallel reduction of ETD and EsB contrast under encapsulation (Figure 5), but the bulk-substrate ETD contrast inversion is explicitly a hypothesis.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Electron Channelling Contrast SEM Imaging of Twist Domains in Transition Metal Dichalcogenide Heterostructures." pith.science (2026). https://pith.science/paper/HQZUJOVA

@misc{pith2026241116248,
  author       = {Pith},
  title        = {Pith review of: Electron Channelling Contrast SEM Imaging of Twist Domains in Transition Metal Dichalcogenide Heterostructures},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HQZUJOVA}},
  note         = {Machine review of arXiv:2411.16248}
}

read the original abstract

Twisted 2D material heterostructures provide an exciting platform for investigating new fundamental physical phenomena. Many of the most interesting behaviours emerge at small twist angles, where the materials reconstruct to form areas of perfectly stacked crystal separated by partial dislocations. However, understanding the properties of these systems is often impossible without correlative imaging of their local reconstructed domain architecture, which exhibits random variations due to disorder and contamination. Here we demonstrate a simple and widely accessible route to visualise domains in as-produced twisted transition metal dichalcogenide (TMD) heterostructures using electron channelling contrast imaging (ECCI) in the scanning electron microscope (SEM). This non-destructive approach is compatible with conventional substrates and allows domains to be visualised even when sealed beneath an encapsulation layer. Complementary theoretical calculations reveal how a combination of elastic and inelastic scattering leads to contrast inversions at specified detector scattering angles and sample tilts. We demonstrate that optimal domain contrast is therefore achieved by maximising signal collection while avoiding contrast inversion conditions.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

15 extracted references · 15 canonical work pages

[1]

K. Kim, M. Yankowitz, B. Fallahazad, S. Kang, H. C. P. Movva, S. Huang, S. Larentis, C. M. Corbet, T. Taniguchi, K. Watanabe, S. K. Banerjee, B. J. Leroy, E. Tutuc, Nano Lett. 2016, 16, 1989

work page 2016
[2]

C. A. Schneider, W. S. Rasband, K. W. Eliceiri, Nat. Methods 2012, 9, 671

work page 2012
[3]

Tanuma, C

S. Tanuma, C. J. Powell, D. R. Penn, Proceedings, Annu. Meet. Electron Microsc. Soc. Am. 1996, 54, 130

work page 1996
[4]

Shinotsuka, S

H. Shinotsuka, S. Tanuma, C. J. Powell, D. R. Penn, Surf. Interface Anal. 2019, 51, 427

work page 2019
[5]

Shimizu, D

R. Shimizu, D. Ze-Jun, Reports Prog. Phys. 1992, 55, 487

work page 1992
[6]

Tanuma, C

S. Tanuma, C. J. Powell, D. R. Penn, Acta Phys. Pol. A 1992, 81, 169

work page 1992
[7]

Krawczyk, M

M. Krawczyk, M. Pisarek, R. Szoszkiewicz, A. Jablonski, Materials (Basel). 2020, 13, 1

work page 2020
[8]

P. S. Neu, M. Šiškins, E. E. Krasovskii, R. M. Tromp, S. J. Van Der Molen, Phys. Rev. B 2023, 107, 75427

work page 2023

Show all 15 references

[9]

H. Y. Chang, A. Alvarado, T. Weber, J. Marian, Nucl. Instruments Methods Phys. Res. Sect. B Beam Interact. with Mater. Atoms 2019, 454, 14

2019
[10]

Sutter, E

P. Sutter, E. Sutter, APL Mater. 2014, 2, DOI 10.1063/1.4889815

2014 doi
[11]

S. Luo, D. C. Joy, Scanning Microsc. 1987, 1, 951

1987
[12]

T. Liu, I. Temprano, D. A. King, S. M. Driver, S. J. Jenkins, Chem. Commun. 2015, 51, 537

2015
[13]

L. D. Landau, E. M. Lifshitz, Quantum Mechanics - Non-Relativistic Theory, Pergamon Press, ISBN: 0750635398, 1981

1981
[14]

T. I. Andersen, G. Scuri, A. Sushko, K. De Greve, J. Sung, Y. Zhou, D. S. Wild, R. J. Gelly, H. Heo, K. Watanabe, T. Taniguchi, P. Kim, H. Park, M. D. Lukin, Nat. Mater. 2021, 20, 480

2021
[15]

A. Rupp, J. Göser, Z. Li, I. Bilgin, A. Baimuratov, A. Högele, ArXiv: 2306.14198 2023, 24

2023 arXiv

Pith tools

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

[1] [1]

K. Kim, M. Yankowitz, B. Fallahazad, S. Kang, H. C. P. Movva, S. Huang, S. Larentis, C. M. Corbet, T. Taniguchi, K. Watanabe, S. K. Banerjee, B. J. Leroy, E. Tutuc, Nano Lett. 2016, 16, 1989

work page 2016

[2] [2]

C. A. Schneider, W. S. Rasband, K. W. Eliceiri, Nat. Methods 2012, 9, 671

work page 2012

[3] [3]

Tanuma, C

S. Tanuma, C. J. Powell, D. R. Penn, Proceedings, Annu. Meet. Electron Microsc. Soc. Am. 1996, 54, 130

work page 1996

[4] [4]

Shinotsuka, S

H. Shinotsuka, S. Tanuma, C. J. Powell, D. R. Penn, Surf. Interface Anal. 2019, 51, 427

work page 2019

[5] [5]

Shimizu, D

R. Shimizu, D. Ze-Jun, Reports Prog. Phys. 1992, 55, 487

work page 1992

[6] [6]

Tanuma, C

S. Tanuma, C. J. Powell, D. R. Penn, Acta Phys. Pol. A 1992, 81, 169

work page 1992

[7] [7]

Krawczyk, M

M. Krawczyk, M. Pisarek, R. Szoszkiewicz, A. Jablonski, Materials (Basel). 2020, 13, 1

work page 2020

[8] [8]

P. S. Neu, M. Šiškins, E. E. Krasovskii, R. M. Tromp, S. J. Van Der Molen, Phys. Rev. B 2023, 107, 75427

work page 2023