Crystallographic Challenges in Microscopy of Multidomain Spinel Materials

Colin Ophus; Gerbrand Ceder; Mary Scott; Ninon Scherz; Shashwat Anand; Tara P. Mishra; Tucker Holstun

REVIEW 3 major objections 4 minor 7 references

Crystallographic Challenges in Microscopy of Multidomain Spinel Materials

T0 review · 3 major / 4 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read STEM-HAADF imaging along [110] cannot detect a class of antiphase boundaries between spinel domains in δ-DRX cathodes.

desk verdict Worth reading and worth refereeing: the enumeration of invisible spinel-variant boundaries is a real contribution, but the undetectability claim outruns the projected-potential imaging model and needs a multislice check before it can stand as a categorical statement. read the letter →

arxiv 2604.17561 v2 pith:ZRXQ7RVR submitted 2026-04-19 cond-mat.mtrl-sci

Ninon Scherz , Shashwat Anand , Colin Ophus , Tucker Holstun , Mary Scott , Tara P. Mishra , Gerbrand Ceder This is my paper · ORCID

classification cond-mat.mtrl-sci

keywords STEM-HAADFantiphaseboundariesspinelvariantsdisorderedrocksaltcathodedelta-DRXFourierfilteringprojectionoverlapdomaindetection

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 asks what atomically resolved STEM-HAADF imaging can actually reveal about the nanometer-scale spinel domains that form when a disordered rocksalt cathode material (δ-DRX) transforms to a spinel-like ordered structure. Simulating all 28 pairings of the eight crystallographically distinct spinel variants, it shows that every interface falls into one of four Fourier-filtered fringe profiles: continuous in both spinel superlattice frequencies, discontinuous in both, or discontinuous in only one. One profile—interfaces between variants related by a twofold rotation, such as A1–A2—leaves the boundary completely undetectable along the standard [110] zone axis. The paper further argues that 'disordered' or 'layered' regions seen in real micrographs can be projections of ordered spinel domains overlapping across slanted {100} boundaries, not true remnant disorder. If correct, single-axis imaging systematically misses a class of domain boundaries and overestimates disorder, which matters for understanding the electrochemical performance of Mn-rich cathodes.

What carries the argument

The key machinery is the crystallographic relationship among the eight spinel variants and their projection onto the [110] zone axis. In spinel, the octahedral sublattice splits into occupied 16d and vacant 16c sites; along [110] these form pure 16d columns, pure 16c columns, and mixed columns, and the eight variants differ by rotations and translations that rearrange these columns. Fourier filtering of simulated STEM-HAADF images isolates two spinel superlattice frequencies (indexed as (-11-1)- and (-111)-type), and the continuity or discontinuity of these fringes across a slanted {100} interface defines the four boundary classes. The slanted {100} interface, intersecting [110] at 45°, is t

What would settle it

Perform multislice STEM-HAADF simulations of an A1–A2 boundary at experimental thicknesses (e.g., 50–100 nm) with full dynamical scattering and a realistic detector; if the (-11-1)- or (-111)-type filtered fringes show a detectable shift across the interface, the claim that this class is undetectable along [110] is falsified. Alternatively, image a sample containing a known A1/A2 boundary along [110] and see whether the boundary appears.

Watch

Extended reading notes

Core claim

The central claim is that, viewed along [110], the eight spinel variants collapse into four projected motifs, and the 28 possible antiphase boundaries between them produce exactly four detectable Fourier-filtered profiles. Boundaries between variants in the same rotational pair (A1/A2, B1/B2, C1/C2, D1/D2) preserve both (-11-1)- and (-111)-type fringes and are invisible in atomically resolved STEM-HAADF; all other pairings are detectable, with eight disrupting both fringes, eight only the (-11-1)-type, and eight only the (-111)-type. The paper also claims that apparent excess disorder and layered-like contrast in prior STEM-HAADF studies of δ-DRX can arise from projection overlap across low-

Load-bearing premise

The undetectability classification rests on the Methods 4.5 simulation of HAADF contrast as the squared projected Mn potential convolved with a Gaussian, which neglects dynamical diffraction, channeling, detector geometry, and thickness; if real scattering changes fringe continuity, the set of 'invisible' boundaries could change.

Editorial extensions

If this is right

Single-zone-axis STEM-HAADF will systematically undercount domain boundaries: roughly one seventh of variant-pair types (A1/A2, B1/B2, C1/C2, D1/D2) are invisible along [110], so domain-size and morphology statistics derived from such images are biased.
Regions that look disordered or layered in atomic-resolution images of δ-DRX can be fully ordered spinel domains seen through overlapping slanted boundaries; apparent remnant disorder should not be interpreted as real DRX without complementary diffraction evidence.
Even for visible boundaries, multiple variant pairings share the same Fourier-filtered profile, so exact variant identity cannot be assigned from [110] images alone; boundaries, not domains, are the only reliable unit of characterization.
Diffraction-based methods such as SEND and XRD remain necessary to quantify remnant disorder, while STEM-HAADF is best used for domain size and morphology once its blind spots are known.

Reading between the lines

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

A direct test of the undetectability claim would be multislice simulations of the same interfaces at realistic thicknesses: the paper's Gaussian-squared-potential proxy may miss dynamical contrast that either reveals or further hides a boundary.
If variant fractions in δ-DRX are roughly equal, roughly one seventh of all antiphase boundaries are invisible; correcting published domain-size distributions for this factor could shift quantitative conclusions about ionic transport and phase-transformation suppression.
The same projection-overlap logic should apply to other topotactic transformations that preserve a parent lattice while reducing symmetry, such as perovskite-to-infinite-layer reductions; boundaries there may also be hidden when the beam direction aligns with the preserved sublattice.
Combining multi-axis imaging or ptychography with the paper's Table 1 classification could identify which variant pairs actually dominate δ-DRX, testing whether all eight variants form with roughly equal probability.

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 the first systematic map of what [110] STEM-HAADF plus Fourier filtering can and cannot see for all 28 spinel-variant interfaces in δ-DRX, and the invisible-boundary class is a useful warning that should change how people read filtered HAADF images of these cathodes. The classification itself is clean and parameter-free: given the eight variants from ref 33, the four Fourier profiles follow from octahedral-column bookkeeping, and Table 1 is coherent. That is real credit, not repackaging—ref 33 gave the variant framework and {100} energies, not the [110] projection signatures. The self-citation is not a problem.

The soft spots are in the imaging model, not the crystallography. The HAADF simulation is a projected Mn potential, squared, then Gaussian-blurred (Methods 4.5). That is a thin-slice projection proxy: no dynamical diffraction, no channeling, no detector geometry. For the four undetectable pairs (A1/A2 etc.), the two variants differ only by depth ordering of pure 16c/16d columns along the beam, so 'invisible' is baked into the projection by construction. The abstract's 'leaves the boundary undetectable in atomically resolved electron microscopy' overstates what a projected-potential calculation can establish, especially since the experimental lamella is ~100 nm while the simulated slab is ~3.4 nm. The authors note the projected interface width is thickness-dependent but do not test whether detectability is thickness-dependent. A multislice sensitivity check on a few representative pairs, or at least explicit hedging in the abstract, would close the gap.

Two smaller points. Table 1 is about {100} interfaces only; the extension to other planes is asserted from symmetry, and the Discussion itself adds a caveat for boundaries perpendicular to the beam. The abstract's 'each domain interface' is broader than the evidence. And the experimental check is qualitative: one image, four marked regions; it supports the scheme but does not validate it. No code or data files are shipped, which is annoying but not disqualifying.

Net: the central geometric classification is sound and likely correct for ideal {100} boundaries; the load-bearing imaging claim needs strengthening before it should appear as a categorical statement. This paper deserves a serious referee. Send it out, and ask for multislice validation and a more careful abstract.

Referee Report

3 major / 4 minor

Summary. The paper analyzes single-zone-axis STEM-HAADF imaging of δ-DRX (disordered rocksalt derived spinel) materials, focusing on the eight crystallographic variants of spinel ordering that form on the rocksalt lattice. The authors construct {100} antiphase boundary models between every pair of the eight variants, simulate HAADF-like images using a projected Mn potential squared and Gaussian blurred, and then apply Fourier filtering to isolate two spinel superlattice frequencies indexed near the (111) and parent-lattice (222) positions. They find that all 28 variant pairs fall into four discontinuity profiles: continuous in both frequencies, discontinuous in both, or discontinuous in only one. Four pairs (Ai–Ai, Bi–Bi, Ci–Ci, Di–Di) are reported as producing no discontinuity, making the boundary 'undetectable' in STEM-HAADF along [110]. The paper argues that this explains apparent remnant disorder and layered-like contrast in experimental images as projection overlap from slanted domain boundaries, and warns that single-axis HAADF systematically undercounts domain boundaries in δ-DRX.

Significance. The crystallographic enumeration and the four-profile taxonomy are internally consistent and provide a useful conceptual framework for interpreting spinel-domain contrast in the [110] projection. The complete simulation of all 28 variant pairs and the explicit connection to Fourier filtering are valuable strengths. If the 'undetectable boundary' claim holds for real STEM-HAADF imaging, the paper identifies an important and non-obvious limitation that could affect many experimental studies of Mn-rich cathodes and analogous topotactic systems. However, the central quantitative claim about detectability in actual electron microscopy rests on a projection-only imaging proxy; the paper's own method section describes it as 'projected electrostatic potential calculations with Gaussian blurring.' This gap needs to be addressed before the abstract-level claim can be accepted as a statement about real STEM-HAADF rather than about a simplified model.

major comments (3)

The 'undetectable' classification for A1–A2, B1–B2, C1–C2, and D1–D2 is obtained from an imaging model that squares the projected Mn potential and convolves it with a Gaussian (σ=3 pixels). For these pairs, the two variants differ only by the depth ordering of pure 16c/16d columns along the beam direction; in any strictly projected model they are mathematically identical. Thus the undetectability is built into the model rather than demonstrated for real STEM-HAADF, where dynamical diffraction, channeling, finite detector geometry, and sample thickness can break the depth degeneracy. The simulated slab is only c=34.24545 Å along [110], whereas the experimental lamella is ~100 nm. Please either (a) restrict the central claim to 'projected-potential imaging' and adjust the Abstract and Conclusions accordingly, or (b) add multislice simulations for representative boundaries—at least one 'und
The statement that 'the results summarized in Table 1 are, in principle, applicable to interfaces beyond the {100} family, since the Fourier discontinuity arises from the symmetry relationship between variants rather than from the specific plane along which the boundary forms' is asserted without derivation or simulation. The Fourier discontinuity depends on the relative projected offset of the two variant lattices, and for a general boundary plane (or a boundary with local relaxations) this offset need not be the same as for the specifically constructed slanted {100} boundaries. Only {100} boundaries were simulated; the abstract's unqualified 'each domain interface' is therefore broader than the evidence. Please either add a second boundary orientation (e.g., {111} or {110}) to substantiate the generality, or explicitly label this as a hypothesis and qualify the abstract.
The binary 'continuous/discontinuous' classification in Table 1 is based on visual inspection of filtered fringe images ('interfaces were classified as continuous or discontinuous depending on whether the filtered fringe phase and intensity remained coherent across the boundary'). No quantitative criterion (phase-shift threshold, fringe-amplitude jump, automated metric) or uncertainty estimate is provided, so the reproducibility of the four-profile taxonomy from the simulations is not established. Given that the entire paper rests on this classification, a quantitative definition would strengthen confidence; at minimum, the authors should state whether the classification was performed blinded and report the robustness of the category assignments to the Gaussian blur width and mask parameters.

minor comments (4)

Typo: 'was subsequentely tiled' should be 'was subsequently tiled'.
The statement that a {100} boundary viewed along [110] intersects the slab at 45° is clear, but the projected-width dependence on thickness is mentioned only qualitatively. A quantitative statement about the simulated thickness effect would help the reader assess the relevance to the ~100 nm experimental lamella.
The suppression of the DRX-like frequency within regions of clear spinel ordering is a processing choice that could affect the apparent localization of the DRX-like fringes at the interface. The description says this is done 'to enable visualization' but does not specify the suppression amplitude or criterion; please give the exact procedure or a reference to a previous implementation.
Ref. 33 (the authors' own prior work) is the source of the eight-variant enumeration and the low-energy {100} boundary selection, and this is correctly acknowledged in the text. The dependence on that prior work is substantive; consider citing the relevant sections or equations of Ref. 33 more specifically in §2.1 and §2.2.

Circularity Check

1 steps flagged · score 5.0 of 10

The 'undetectable boundary' class is the projection model restating its own depth-averaging assumption; the other three Fourier profiles are genuine computed results.

self definitional [Section 2.1 (variant pairing definition); Table 1 rows A1–A2, B1–B2, C1–C2, D1–D2; Methods 4.5 (HAADF proxy)]
"In the [110] projection, the eight variants collapse into four projected motifs, each shared by a pair within a family: (A1,A2), (B1,B2), (C1,C2), and (D1,D2), differing only by the relative depth of atoms along the beam direction. These depth differences do not affect HAADF contrast in a simple projection, making the i=1 and i=2 variants within each pair indistinguishable in [110]."

Methods 4.5 defines the simulated HAADF image as the squared [110]-projected Mn potential convolved with a Gaussian (σ=3 pixels), so depth ordering along the beam is erased by construction. Section 2.1 defines the i=1/i=2 pairs as differing only by atom depth, with identical projected motifs. Consequently the four 'No/No/No' rows of Table 1, and the Abstract's headline that such boundaries are 'undetectable in atomically resolved electron microscopy,' restate the projection model's defining assumption rather than a derived scattering result: the boundary is invisible because the model cannot see depth. The prediction is equivalent to the input. The paper hedges once ('in a simple projection') but Table 1 and the Abstract present the model's built-in degeneracy as a simulated detection limi

full rationale

The four-Fourier-profile classification is largely a parameter-free geometric computation: for 24 of 28 variant pairings the fringe-discontinuity assignments are derived here from the constructed {100} interface supercells and the Fourier filtering of the simulated images, with no fitted parameters and no experimental target that could force the outcome. The experimental image in Figure 2 is used only for qualitative comparison, not as a fitting target. The eight-variant enumeration and the low-energy {100} interface choice are taken from the authors' own prior work (ref 33), a self-citation, but that prior work is an independently calculated group-theory/DFT result and the present paper regenerates the variant structures ab initio in Methods 4.3; this does not reduce the imaging classification to the citation. The genuine circular element is confined to the 'undetectable' class: the pairs A1–A2, B1–B2, C1–C2, D1–D2 are defined as projection-identical structures, and the HAADF proxy is defined as a function of the depth-integrated potential, so Table 1's 'No' detection entries and the claim that such boundaries are invisible in real atomically resolved STEM-HAADF follow by construction from the two definitions. Real dynamical diffraction, channeling, and thickness effects, which the projected-potential-plus-Gaussian model omits, could in principle break the depth degeneracy; the paper acknowledges projection-only limitations ('would not affect HAADF contrast in a simple projection') and disclaims quantitative width predictions, but the unhedged abstract/Table 1 statement goes beyond what the model can establish. Score 5 reflects partial circularity of one central claim class, with substantial independent content in the other three profiles.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The paper's central claim rests on the prior self-cited crystallographic framework (eight variants, {100} low-energy boundaries), the projected-potential HAADF proxy, and an unproved generalization from {100} to arbitrary boundary planes. No new physical entities are introduced.

free parameters (3)

Interface tilt m and position b (Eq. 1) = not stated; chosen to center interface
Hand-selected geometric parameters defining the slanted {100} partition; set the projected interface width but should not alter the four-profile classification.
Gaussian blur sigma in simulated HAADF = 3 pixels
Ad hoc smoothing of squared projected potential (Methods 4.5); affects fringe contrast/visibility and therefore the binary detectability calls.
Fourier mask radii/positions = adjusted between experimental and simulated datasets
Methods 4.6 says filtering parameters were tuned per dataset; the discontinuity classification depends on these choices, so they are effectively free parameters.

assumptions (5)

standard math Eight spinel variants arise from the Fm-3m to Fd-3m symmetry reduction (Section 2.1; ref 33).
Used without proof; basis for the entire enumeration.
domain assumption {100} interfaces are the lowest-energy domain boundaries for most variant pairings (Section 2.2; ref 33).
Limits simulations to one boundary plane; if real particles favor other planes, the four-profile classification may not apply.
domain assumption Squared projected potential convolved with a Gaussian reproduces STEM-HAADF contrast sufficiently for detecting fringe discontinuities (Methods 4.5).
Neglects dynamical scattering; not validated against multislice or experimental standards.
ad hoc to paper Fourier discontinuity profiles depend only on variant symmetry, not on the boundary plane (Discussion).
Asserted 'in principle' without proof; needed for the abstract's 'each domain interface' wording.
domain assumption Li and oxygen can be omitted from simulations because HAADF contrast is dominated by Mn (Methods 4.5).
Valid for Z-contrast in first approximation, but oxygen columns could modulate the projected potential; the paper does not quantify this.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Crystallographic Challenges in Microscopy of Multidomain Spinel Materials." pith.science (2026). https://pith.science/paper/ZRXQ7RVR

@misc{pith2026260417561,
  author       = {Pith},
  title        = {Pith review of: Crystallographic Challenges in Microscopy of Multidomain Spinel Materials},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZRXQ7RVR}},
  note         = {Machine review of arXiv:2604.17561}
}

read the original abstract

Electron microscopy techniques are instrumental in the characterization of energy storage materials, with atomic resolution images providing the detailed structural features that are needed to understand their properties. Atomically resolved electron microscopy techniques have been routinely used to study the microstructure in high performing Mn-based oxide cathodes, which often contain spinel-like ordering. Here, we evaluate STEM-HAADF imaging and subsequent Fourier filtering as tools for characterizing {\delta}-DRX spinel domains and their antiphase boundaries, which play a central role in the material's electrochemical performance. Using electron microscopy simulations and recent theoretical insight into the structural makeup of {\delta}-DRX, we attempt to characterize the crystallographic spinel variants which occur in its multi-domain structure. We show that each domain interface, arising from pairings among eight distinct variants, can be categorized into one of four Fourier filtered profiles, one of which leaves the boundary undetectable in atomically resolved electron microscopy when viewed along the preferred [110] zone axis. Our results also suggest that the appearance of seemingly disordered or layered-like regions might actually arise from low energy domain boundaries which are slanted relative to the [110] viewing direction. Our findings highlight the need for careful interpretation of atomic-resolution micrographs of phase transitions, where local reordering drives transformations from higher to lower symmetry structures while maintaining lattice coherence.

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

7 extracted references · 1 linked inside Pith

[110]

zone axis. Taken together, these findings reveal sub- tle crystallographic challenges inherent to the microscopy of multi-domain systems and are broadly relevant to other materials that require precise characterization of domain boundaries on coherent lattices34–36. 2 Results 2.1 Crystallographic variants formed during symme- try reducing phase transforma...
[2]

The sites from the16cand16d sub-lattices corresponding to each variant of spinel are shown with white and purple respectively

on the parent rock-salt lattice. The sites from the16cand16d sub-lattices corresponding to each variant of spinel are shown with white and purple respectively. The sites beneath the top surface layer are hidden by the surface sites. The encircled sites represent the column of sites beneath the surface (extending into the paper) that either belong entirely...
[3]

In this view, the previously hidden sub- surface cation sites are visible, and the “pure”16ccolumns project as empty

projection as Figure 1b but without marking the white vacancy sites. In this view, the previously hidden sub- surface cation sites are visible, and the “pure”16ccolumns project as empty . In the [110] projection, the eight variants collapse into four projected motifs, each shared by a pair within a family: (A1,A2), (B1,B2), (C1,C2), and (D1,D2), dif- + P ...
[4]

direction. 8 | 1–15 + P V S O B M / B N F < Z F B S > < W P M > 3 Discussion We employ structural modeling and projected electrostatic potential calculations with Gaussian blurring to simulate STEM-HAADF contrast, illustrating how antiphase bound- aries inδ-DRX appear under specific projection conditions. This work defines the limits of single-zone-axis S...
[5]

pure16c" and “pure16d

zone axis. As shown in Figure S1, other low-index axes (e.g., [100] and [111]) result in overlapping projec- tions of spinel16cand16dWyckoff sites; these mixed- occupancy columns are indistinguishable from the DRX phase. Higher-order axes, such as [211], are impractical due to the extreme alignment precision required for atomic resolution. In contrast, th...
[6]

direction (c-axis of supercell containing the inter- face), yielding a 2D projected potential map. To generate an intensity map representative of Z-contrast imaging suitable for Fourier analysis, the projected poten- tial was squared and subsequently convolved with a Gaus- sian filter usingσ=3pixels. The resulting image was nor- malized to the range[0,1]p...
[7]

projection and to evaluate domain boundary discon- tinuities. For both experimental and simulated datasets, the result- ing micrographs were Fourier transformed using a 2D FFT, and the complex fourier spectrum was analyzed to identify frequency components corresponding to parent DRX lattice periodicities and spinel superlattice periodicities. Circular rec...

arXiv 2025

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.

[1] [110]

zone axis. Taken together, these findings reveal sub- tle crystallographic challenges inherent to the microscopy of multi-domain systems and are broadly relevant to other materials that require precise characterization of domain boundaries on coherent lattices34–36. 2 Results 2.1 Crystallographic variants formed during symme- try reducing phase transforma...

[2] [2]

The sites from the16cand16d sub-lattices corresponding to each variant of spinel are shown with white and purple respectively

on the parent rock-salt lattice. The sites from the16cand16d sub-lattices corresponding to each variant of spinel are shown with white and purple respectively. The sites beneath the top surface layer are hidden by the surface sites. The encircled sites represent the column of sites beneath the surface (extending into the paper) that either belong entirely...

[3] [3]

In this view, the previously hidden sub- surface cation sites are visible, and the “pure”16ccolumns project as empty

projection as Figure 1b but without marking the white vacancy sites. In this view, the previously hidden sub- surface cation sites are visible, and the “pure”16ccolumns project as empty . In the [110] projection, the eight variants collapse into four projected motifs, each shared by a pair within a family: (A1,A2), (B1,B2), (C1,C2), and (D1,D2), dif- + P ...

[4] [4]

direction. 8 | 1–15 + P V S O B M / B N F < Z F B S > < W P M > 3 Discussion We employ structural modeling and projected electrostatic potential calculations with Gaussian blurring to simulate STEM-HAADF contrast, illustrating how antiphase bound- aries inδ-DRX appear under specific projection conditions. This work defines the limits of single-zone-axis S...

[5] [5]

pure16c" and “pure16d

zone axis. As shown in Figure S1, other low-index axes (e.g., [100] and [111]) result in overlapping projec- tions of spinel16cand16dWyckoff sites; these mixed- occupancy columns are indistinguishable from the DRX phase. Higher-order axes, such as [211], are impractical due to the extreme alignment precision required for atomic resolution. In contrast, th...

[6] [6]

direction (c-axis of supercell containing the inter- face), yielding a 2D projected potential map. To generate an intensity map representative of Z-contrast imaging suitable for Fourier analysis, the projected poten- tial was squared and subsequently convolved with a Gaus- sian filter usingσ=3pixels. The resulting image was nor- malized to the range[0,1]p...

[7] [7]

projection and to evaluate domain boundary discon- tinuities. For both experimental and simulated datasets, the result- ing micrographs were Fourier transformed using a 2D FFT, and the complex fourier spectrum was analyzed to identify frequency components corresponding to parent DRX lattice periodicities and spinel superlattice periodicities. Circular rec...

arXiv 2025