Advances in electron backscatter diffraction

Alessandro Previero; Alex Foden; Thomas Benjamin Britton

REVIEW 3 major objections 5 minor 1 cited by

Advances in electron backscatter diffraction

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 This paper argues that a two-stage indexing pipeline—unsupervised principal component analysis followed by refined template matching—recovers smoother orientation gradients within grains and across grain boundaries, giving better EBSD…

desk verdict A competent EBSD pipeline demo whose central claim about better grain-boundary orientations is not backed by accuracy data; worth reviewing with requests for quantitative validation. read the letter →

arxiv 1908.04860 v1 pith:EN6HZBTL submitted 2019-08-12 physics.comp-ph cond-mat.mtrl-sci

Alex Foden , Alessandro Previero , Thomas Benjamin Britton This is my paper · ORCID

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

keywords electronbackscatterdiffractionEBSDtemplatematchingprincipalcomponentanalysismachinelearningdirectdetectorsphericalindexingtransfer

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 recent software and hardware advances can extract more from electron backscatter diffraction (EBSD) patterns than conventional band detection. Its central demonstration is a two-stage indexing pipeline: an unsupervised principal component analysis labels each grain with a representative orientation, then a refined template matching step, using FFT-based cross-correlation with iterative interpolation, updates every pixel's orientation starting from that label. The result is smoother orientation gradients within and between labels, which the authors say yields better measurements of orientation near grain boundaries. The paper also shows that a single direct-electron-detector silicon pattern, symmetry-overlapped on the sphere, can form an experimental reference sphere, and that a transfer-learned convolutional neural network trained only on dynamical simulations can separate easily confused phases such as FCC, BCC, and diamond cubic.

What carries the argument

The load-bearing machinery is the two-stage indexing workflow of PCA-MSA plus refined template matching. In the first stage, the diffraction patterns are arranged as variables, decomposed into principal components, and rotated into representative patterns, so each grain receives one orientation label. In the second stage, a refined template matching routine, based on FFT-based cross-correlation with an iterative interpolation step, searches a small library of templates around that starting label to update the orientation of every pixel in the map. Supporting machinery includes spherical reprojection, in which a direct-electron-detector pattern is symmetrically overlapped on the sphere using the crystal's 24 cubic symmetry rotations to build a high-quality experimental reference sphere, and a transfer-learned convolutional neural network that classifies phases from simulated dynamical patterns.

What would settle it

Take a sample with a known orientation gradient and an independently measured grain boundary misorientation, run the PCA-MSA-plus-refined pipeline, and compare the recovered near-boundary orientations with the independent measurement; the central claim fails if the refinement does not reduce the error relative to the labels alone.

Watch

Extended reading notes

Core claim

The central claim of the paper is that combining the PCA-MSA method with refined template matching improves the precision of EBSD orientation indexing. PCA-MSA compresses the full set of diffraction patterns into a small number of common components, gives each region a single representative orientation label, and the refined template matching then re-indexes every pixel starting from that label, recovering smooth orientation gradients within and between labels. On a deformed iron data set this yields a better measurement of orientations near grain boundaries than labels alone. The same case-study logic extends to two further demonstrations: a silicon pattern recorded on a direct electron detector, matched to a dynamical simulation and symmetrically overlapped on the sphere to build an experimental reference sphere, and a convolutional neural network trained on simulated six-phase patterns, whose layer activations echo edge detection, band detection, and zone-axis structure and which can separate phases that conventional EBSD analysis tends to confuse.

Load-bearing premise

The load-bearing premise is that dynamical simulations reproduce experimental diffraction contrast faithfully enough that a simulated reference pattern and simulated training patterns can stand in for real detector data.

Editorial extensions

If this is right

On large maps, the two-stage pipeline offers a computationally cheaper route to high-precision orientation maps: a small number of correlations per grain labels the map, and only the refinement step re-indexes every pixel.
An experimental reference sphere built from a single direct-electron-detector pattern can serve as a template source for spherical indexing, reducing the need for dynamical simulations in cases where simulations are hard, such as minerals with variable composition.
A convolutional neural network trained only on simulated patterns can classify phases that conventional EBSD confuses, such as FCC iron versus copper versus diamond-cubic silicon.
The CNN's layer activations show a correspondence to classic EBSD features, suggesting the network learns physically meaningful contrast rather than arbitrary features.
Open-source indexing routines and open hierarchical data formats make the whole workflow inspectable and testable, which increases confidence in the resulting microstructural maps.

Reading between the lines

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

Because the refined template matching operates on raw patterns rather than EBSD-specific features, the same PCA-plus-refinement recipe should transfer to other diffraction-imaging modalities, such as 4D-STEM, where virtual apertures are already replacing fixed detectors.
The visual improvement near grain boundaries should be tested quantitatively against an independent orientation or strain measurement on a sample with known boundary misorientations; the paper shows smoother maps but does not report a numerical error metric.
The CNN activation maps suggest that the network's early layers rediscover the edge and band features that conventional band-detection analysis computes explicitly, so one could design compact deterministic descriptors for phase classification rather than treating the network as a black box.
Combining the experimental reference sphere with the CNN classifier could remove the dependence on simulation accuracy: real detector patterns could supply training data that includes actual detector contrast.

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

Short version: this is a readable demonstration paper, not a validation paper. The one claim that sounds like a result — that PCA-MSA plus refined template matching gives better orientations near grain boundaries — is unsupported by any accuracy metric. The rest is a competent tour of existing components.

What's actually new: linking the PCA-MSA labelling of Wilkinson et al. to the FFT cross-correlation refinement of Foden et al., applying transfer learning with AlexNet to phase classification from dynamical EBSD simulations, and projecting a direct-electron-detector silicon pattern onto a symmetry-augmented spherical reference. None of the pieces is new in isolation, but the combinations are reasonable and the paper does a service by showing them in one place with open data and open code. The DED vs dynamical simulation match (Figure 4) is visually decent, and the CNN activation maps are a genuinely instructive way to see what the network is keying on.

The soft spots are real but mostly in the gap between demonstration and claim. Section 6 states the refined template matching "results in a better measurement of the orientations near grain boundaries." The evidence is a colour map in Figure 3 where the refined labels look smoother. Smoothness is not accuracy; an interpolation-based refinement will always reduce pixel-to-pixel orientation differences even if it moves orientations away from the truth. There is no comparison against a ground truth (e.g. synthetic patterns with known orientations), no comparison against PCA-MSA alone or dictionary indexing, and no angular error statistics. Also, in Section 5 the text says "8 potential phases" and then lists six; the learning rates (3, 6, 9, 18, 24) are ambiguous without units or scale; and the CNN is trained only on simulations, so it is unclear how it transfers to experimental patterns (the authors do acknowledge this indirectly in the Discussion). These are fixable with numbers or softened wording.

The citation pattern is fine: heavy reliance on the authors' own AstroEBSD and prior template matching is expected in a methods demonstration, and they cite competing approaches (dictionary indexing, spherical indexing) fairly. The paper is honest about many limitations, which makes the one overstated claim stand out more.

Bottom line: this is a useful, competent demonstration for people working in EBSD who want to see the pieces assembled. It deserves peer review at an applied microscopy or materials-characterisation venue, but the referee should push for quantitative validation of the refinement claim and corrections to the machine-learning hyperparameters and phase-count mismatch.

Referee Report

3 major / 5 minor

Summary. The manuscript presents four case-study demonstrations of recent developments in EBSD analysis: an open-source AstroEBSD workflow on deformed interstitial-free steel; a combination of PCA-based multivariate statistics with refined template matching for orientation indexing; spherical reprojection of a direct-electron-detector silicon pattern matched to a dynamical simulation; and CNN-based phase classification using transfer learning with AlexNet. The stated aim is to show that open algorithms, open data formats, and combinations of statistical, template-matching, and machine-learning approaches can accelerate development of microstructural insight. The central claim appears in Section 6, where the PCA-MSA plus refinement route is said to give 'a better measurement of the orientations near grain boundaries'.

Significance. If the demonstrated workflows perform as claimed, the paper would be a valuable community resource: it builds on open-source tools (AstroEBSD), uses a publicly archived data set, connects several recent EBSD analysis strands, and shows a route from direct detector patterns to spherical reference data and transfer-learned classification. The strengths are the reproducibility-oriented presentation and the clear identification of software components. However, the significance is conditional because the main quantitative claims are not yet supported: there are no accuracy figures, baselines, or error analyses for the central orientation-refinement claim, and the machine-learning demonstration lacks the details needed to judge its validity. The paper is closer to a collection of demonstrations than a self-contained methods validation.

major comments (3)

[Section 6, Figure 3] The central claim that combining PCA-MSA with refined template matching 'results in a better measurement of the orientations near grain boundaries' is not supported by the evidence shown. The only supporting evidence is the colourized orientation map in Figure 3, which shows smoother orientation gradients after refinement; smoothness is not accuracy, and the iterative interpolation step in the refinement will tend to reduce pixel-to-pixel orientation differences even when the true orientation error increases. No angular error statistics, no ground-truth or independent reference (e.g., HR-EBSD, dictionary indexing, or simulated patterns with a known orientation field), and no baseline comparison to PCA-MSA alone or to conventional Hough indexing are provided. This is the load-bearing claim of the paper and needs a quantitative demonstration.
[Section 5, Figure 5] The phase classification demonstration is under-specified and internally inconsistent. The text says '8 potential phases' but then lists only six phases, and the learning rates used for optimization are given as '3, 6, 9, 18 and 24' (five values) without specifying which quantity they refer to, how they were varied, or how the optimum was selected. The confusion matrix is displayed but no overall accuracy, per-class accuracy, training/validation/test split, number of patterns per phase, or data augmentation is reported. Without these details, the claim of 'extremely impressive results' cannot be evaluated.
[Section 4, Figure 4] The spherical EBSD demonstration is qualitative only. The match between the experimental DED silicon pattern and the dynamical simulation is described as 'sensible alignment' of band centres with great circles, and the overlap after symmetry operations is said to show 'inadvertent blurring', but no quantitative metric (e.g., normalized cross-correlation before and after reprojection, pattern-centre error, or coverage fraction) is reported. Since the pattern centre and detector distance are free parameters in the projection, the demonstration does not establish that the resulting reference sphere is accurate.

minor comments (5)

[Abstract and Section 1] The phrase 'rapidly to develop' in the abstract should be 'rapidly develop', and 'multivariant statistics' in the introduction should be 'multivariate statistics'.
[Section 2] Reference [19] is cited for HDF5, but the entry appears to be a dictionary-indexing paper; a proper citation for the HDF5 format is needed.
[Section 5] The list of phases in Section 5 contains six entries, not eight; the number of phases should be corrected and used consistently throughout the section.
[Figure 1 caption] The caption contains 'purple..' and the phrase 'the maps are in μm' is slightly awkward; the figure caption should be cleaned up.
[Figure 4 caption] The sentence 'The units of these patterns are in the gnomonic coordinate system' is unclear; coordinates are not units, and the intended meaning should be restated.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper's demonstrations are empirical case studies using cited prior methods, with no prediction reducing to its input by definition or fit.

full rationale

The paper is a set of demonstration case studies rather than a derivation chain. Section 3 uses PCA-MSA (Wilkinson et al.) and the authors' own refined template matching (Foden et al.) to produce an orientation map; the displayed 'smoother orientation gradients' are an observed output of the refinement, not a quantity defined to equal the input. The discussion phrase 'better measurement of the orientations near grain boundaries' is an unsupported empirical assertion (no ground-truth or independent accuracy baseline is reported), but an unsupported claim is not a circular one unless the claimed output is definitionally or statistically forced by the input. Section 4 aligns an experimental silicon pattern to a dynamical simulation and projects it onto a sphere; this is a comparison, not a prediction derived from fitted parameters. Section 5 trains a CNN on dynamically simulated patterns and shows a confusion matrix; even granting that the evaluation protocol is underspecified (no explicit held-out split is stated), the text does not rename a fitted training accuracy as an independent prediction. The heavy citation of the authors' own prior software (AstroEBSD) and methods (refined template matching, spherical indexing) is load-bearing only in the sense that those methods are the tools being showcased; the present paper does not invoke a uniqueness theorem or an unverified self-citation to close a derivation. Therefore there are no circular steps to report.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central demonstrations depend on the accuracy of dynamical simulations, the calibration of the experimental geometry, the choice of unsupervised clustering hyperparameters, and the transferability of a photo-trained CNN. None are independently validated in this paper.

free parameters (3)

Number of principal components retained in PCA-MSA
The PCA step reduces diffraction patterns to a limited number of common components (Section 3), but the number of components is not stated; the quality of labels depends on this hyperparameter.
CNN learning rates = 3, 6, 9, 18, 24 (as written)
Section 5 states that learning rates were optimized for the six-phase classification with these values; the values are ambiguous (likely a typo or scaling issue) and the final chosen rate is unclear.
Pattern centre and detector distance for spherical projection = estimated from geometry
Section 4 uses simple geometry [32] requiring an accurate pattern centre and detector distance; these experimental calibration parameters affect the alignment of the reference sphere.

assumptions (4)

domain assumption Dynamical simulations (Bruker DynamicS) accurately represent experimental EBSD contrast.
Used in Section 4 to match a silicon pattern and in Section 5 to generate training templates for the CNN; inaccurate simulations would invalidate both the sphere alignment and phase classification.
domain assumption The pattern centre is known accurately for the DED detector geometry.
Section 4 reconstructs a portion of the diffraction sphere using simple geometry that requires the pattern centre; errors cause blurring or misalignment.
domain assumption PCA-MSA labels are a sufficient starting point for refined template matching to converge.
Section 3 uses the PCA-MSA labels as initial guesses; if these labels are wrong, the refinement step may not recover the true orientation.
domain assumption Transfer learning from ImageNet-trained AlexNet applies to EBSD pattern classification.
Section 5 reuses AlexNet; this assumes generic low-level image features are useful for diffraction patterns.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Advances in electron backscatter diffraction." pith.science (2026). https://pith.science/paper/EN6HZBTL

@misc{pith2026190804860,
  author       = {Pith},
  title        = {Pith review of: Advances in electron backscatter diffraction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EN6HZBTL}},
  note         = {Machine review of arXiv:1908.04860}
}

read the original abstract

We present a few recent developments in the field of electron backscatter diffraction (EBSD). We highlight how open source algorithms and open data formats can be used to rapidly to develop microstructural insight of materials. We include use of AstroEBSD for single pixel based EBSD mapping and conventional orientation mapping; followed by an unsupervised machine learning approach using principal component analysis and multivariate statistics combined with a refined template matching method to rapidly index orientation data with high precision. Next, we compare a diffraction pattern captured using direct electron detector with a dynamical simulation and project this to create a high quality experimental "reference diffraction sphere". Finally, we classify phases using supervised machine learning with transfer learning and a convolutional neural network.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Comparison of Kikuchi Diffraction Geometries in Scanning Electron Microscope
cond-mat.mtrl-sci 2024-11 conditional novelty 6.0 of 10

All four SEM Kikuchi diffraction geometries can reconstruct a full experimental diffraction sphere of aluminum, with transmission geometries preserving sharper higher-order details and reflection geometries limited by...

Reference graph

Works this paper leans on

47 extracted references · 47 canonical work pages · cited by 1 Pith paper

[1]

Mater Today 15 366-376

Wilkinson, AJ and Britton, TB 2012. Mater Today 15 366-376

work page 2012
[2]

J Microsc-Oxford 213 214-224

Dingley, D 2004. J Microsc-Oxford 213 214-224

work page 2004
[3]

Ultramicroscopy 184

Mingard, KP, Stewart, M, Gee, MG, Vespucci, S, and Trager -Cowan, C 2017 . Ultramicroscopy 184

work page 2017
[4]

Phys Rev B 92

Vespucci, S, Winkelmann, A, Naresh-Kumar, G, Mingard, KP, Maneuski, D, Edwards, PR, Day, AP, O'Shea, V, and Trager-Cowan, C 2015. Phys Rev B 92

work page 2015
[5]

Phys Rev Lett 111

Wilkinson, AJ, Moldovan, G, Britton, TB, Bewick, A, Clough, R, and Kirkland, AI 2013 . Phys Rev Lett 111

work page 2013
[6]

US3069654A

Hough, PVC 1962. US3069654A

work page 1962
[7]

Mater Sci Tech Ser 12 837-843

Lassen, NCK 1996. Mater Sci Tech Ser 12 837-843

work page 1996
[8]

Metallurgical Transactions A 24 819-831

Adams, BL, Wright, SI, and Kunze, K 1993. Metallurgical Transactions A 24 819-831

work page 1993

Show all 47 references

[9]

Britton, TB, Tong, VS, Hickey, J, Foden, A, and Wilkinson, AJ, J Appl Crystallogr. 2018

2018
[10]

Mater Sci Tech Ser 22 1271 - 1278

Wilkinson, AJ, Meaden, G, and Dingley, DJ 2006. Mater Sci Tech Ser 22 1271 - 1278

2006
[11]

Ultramicroscopy 106 307-313

Wilkinson, AJ, Meaden, G, and Dingley, DJ 2006. Ultramicroscopy 106 307-313

2006
[12]

J Microsc-Oxford 233 290-301

Villert, S, Maurice, C, Wyon, C, and Fortunier, R 2009. J Microsc-Oxford 233 290-301

2009
[13]

Ultramicroscopy 113 171-181

Maurice, C, Driver, JH, and Fortunier, R 2012. Ultramicroscopy 113 171-181

2012
[14]

Ultramicroscopy 111 1395-1404

Britton, TB and Wilkinson, AJ 2011. Ultramicroscopy 111 1395-1404

2011
[15]

Ultramicroscopy 114 82-95

Britton, TB and Wilkinson, AJ 2012. Ultramicroscopy 114 82-95

2012
[16]

Scripta Mater 62 639-642

Britton, TB, Birosca, S, Preuss, M, and Wilkinson, AJ 2010. Scripta Mater 62 639-642

2010
[17]

Ultramicroscopy 110 1443-1453

Britton, TB, Maurice, C, Fortunier, R, Driver, JH, Day, AP, Meaden, G, Dingley, DJ, Mingard, K, and Wilkinson, AJ 2010. Ultramicroscopy 110 1443-1453

2010
[18]

23 212-213

Singh, S, Ram, F, and Graef, MD 2017. 23 212-213

2017
[19]

Microscopy and Microanalysis 23:S1 213-232

Singh, S, Ram, F, and de Graef, M 2017. Microscopy and Microanalysis 23:S1 213-232

2017
[20]

Microsc Microanal 21 739-52

Chen, YH, Park, SU, Wei, D, Newstadt, G, Jackson, MA, Simmons, JP, De Graef, M, and Hero, AO 2015. Microsc Microanal 21 739-52

2015
[21]

Model Simul Mater Sc 1-22

Singh, S and De Graef, M 2016. Model Simul Mater Sc 1-22

2016
[22]

Ultramicroscopy 196 88-98

Wilkinson, AJ, Collins, DM, Zayachuk, Y, Korla, R, and Vilalta -Clemente, A 2019 . Ultramicroscopy 196 88-98

2019
[23]

ArXiv arXiv:1807.11313

Foden, A, Collins, D, Wilkinson, AJ, and Britton, TB 2018. ArXiv arXiv:1807.11313

2018 arXiv
[24]

Journal of Microscopy 230 472-486

Day, AP 2008. Journal of Microscopy 230 472-486

2008
[25]

ArXiv arXiv:1810.03211

Hielscher, R, Bartel, F, and Britton, TB 2018. ArXiv arXiv:1810.03211

2018 arXiv
[26]

Winkelmann, A, Cios, G, Tokarski, T, Nolze, G, Hielscher, R, and Kociel, T 2019. ArXiv

2019
[27]

Ultramicroscopy 135 136-141

Britton, TB, Jiang, J, Clough, R, Tarleton, E, Kirkland, AI, and Wilkinson, AJ 2013 . Ultramicroscopy 135 136-141

2013
[28]

Appl Phys Lett 110

Vespucci, S, Naresh-Kumar, G, Trager-Cowan, C, Mingard, KP, Maneuski, D, O'Shea, V, and Winkelmann, A 2017. Appl Phys Lett 110

2017
[29]

Vespucci, S, Winkelmann, A, Mingard, K, Maneuski, D, O'Shea, V, and Trager -Cowan, C
[30]

Journal of Microscopy 267 330-346

Winkelmann, A, Nolze, G, Vespucci, S, Naresh -Kumar, G, Trager-Cowan, C, Vilalta - Clemente, A, Wilkinson, AJ, and Vos, M 2017. Journal of Microscopy 267 330-346

2017
[31]

Ultramicroscopy 184 242-251

Mingard, KP, Stewart, M, Gee, MG, Vespucci, S, and Trager -Cowan, C 2018 . Ultramicroscopy 184 242-251

2018
[32]

Mater Charact 117 113-126

Britton, TB, Jiang, J, Guo, Y, Vilalta -Clemente, A, Wallis, D, Hansen, LN, Winkelmann, A, and Wilkinson, AJ 2016. Mater Charact 117 113-126

2016
[33]

Ultramicroscopy 107 414-421

Winkelmann, A, Trager -Cowan, C, Sweeney, F, Day, AP, and Parbrook, P 2007 . Ultramicroscopy 107 414-421

2007
[34]

Adv Eng Mater 5 601-606

Schwarzer, RA and Sukkau, J 2003. Adv Eng Mater 5 601-606

2003
[35]

Commun Acm 60 84-90

Krizhevsky, A, Sutskever, I, and Hinton, GE 2017. Commun Acm 60 84-90

2017
[36]

Journal of Microscopy 245 245-251

Keller, RR and Geiss, RH 2012. Journal of Microscopy 245 245-251

2012
[37]

MRS Bulletin

Gianola, DS, Britton, TB, and Zaefferer, S 2019. MRS Bulletin

2019
[38]

Microscopy and Microanalysis 20

Orphus, C, Ericius, P, Sarahan, M, Czarnik, C, and Ciston, J 2014 . Microscopy and Microanalysis 20

2014
[39]

ArXiv arXiv:1907.07363

Muscarella, L, Hutter, E, Sanchez, S, Dieleman, CD, Savenjie, TJ, Hagfeldt, A, Saliba, M, and Ehhrler, B 2019. ArXiv arXiv:1907.07363

2019 arXiv
[40]

Ultramicroscopy 191 44-50

Vermeij, T and Hoefnagels, JPM 2018. Ultramicroscopy 191 44-50

2018
[41]

Scripta Mater 162 266-271

Vermeij, T, De Graef, M, and Hoefnagels, J 2019. Scripta Mater 162 266-271

2019
[42]

Ultramicroscopy 195 85-92

Ruggles, TJ, Bomarito, GF, Qiu, RL, and Hochhalter, JD 2018. Ultramicroscopy 195 85-92

2018
[43]

Phys Rev B

Winkelmann, A, Britton, TB, and Nolze, G 2019. Phys Rev B

2019
[44]

Science 364

Holm, EA 2019. Science 364

2019
[45]

Appl Optics 33 3127-3134

George, N and Wang, SG 1994. Appl Optics 33 3127-3134

1994
[46]

arXiv arXiv:1902.03682

Kauffmann, K, Zhu, C, Rosengarten, AS, Maryanovsky, D, Harrington, TJ, Marin, E, and Vecchio, KS 2019. arXiv arXiv:1902.03682

2019 arXiv
[47]

Acta Mater 170 118- 131

Shen, YF, Pokharel, R, Niz olek, TJ, Kumar, A, and Lookman, T 2019 . Acta Mater 170 118- 131

2019

Pith tools

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

[1] [1]

Mater Today 15 366-376

Wilkinson, AJ and Britton, TB 2012. Mater Today 15 366-376

work page 2012

[2] [2]

J Microsc-Oxford 213 214-224

Dingley, D 2004. J Microsc-Oxford 213 214-224

work page 2004

[3] [3]

Ultramicroscopy 184

Mingard, KP, Stewart, M, Gee, MG, Vespucci, S, and Trager -Cowan, C 2017 . Ultramicroscopy 184

work page 2017

[4] [4]

Phys Rev B 92

Vespucci, S, Winkelmann, A, Naresh-Kumar, G, Mingard, KP, Maneuski, D, Edwards, PR, Day, AP, O'Shea, V, and Trager-Cowan, C 2015. Phys Rev B 92

work page 2015

[5] [5]

Phys Rev Lett 111

Wilkinson, AJ, Moldovan, G, Britton, TB, Bewick, A, Clough, R, and Kirkland, AI 2013 . Phys Rev Lett 111

work page 2013

[6] [6]

US3069654A

Hough, PVC 1962. US3069654A

work page 1962

[7] [7]

Mater Sci Tech Ser 12 837-843

Lassen, NCK 1996. Mater Sci Tech Ser 12 837-843

work page 1996

[8] [8]

Metallurgical Transactions A 24 819-831

Adams, BL, Wright, SI, and Kunze, K 1993. Metallurgical Transactions A 24 819-831

work page 1993