{"id":"01ea4d7b-a75d-4e62-bfd1-4d9364bbec7a","arxiv_id":"2607.16899","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Quaternion composition is rewritten for direct use in a crystal (non-Cartesian) basis; the new 'cross tensor' equals the scaled inverse of the metric tensor.","lead":"This paper shows that rotation quaternions can be multiplied directly in a crystal's own skewed coordinate system, using only the lattice metric tensor, with no Cartesian round-trip. If it holds, electron-backscatter-diffraction software could become simpler and vendor-convention-free, though the paper's own speed numbers show only marginal gains.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"EBSD efficiency claim rests on unverified assumption that spherical indexing outputs orientations in the fractional-coordinate crystal basis; paper's own FLOP table shows crystallographic product is costlier per operation.","rationale":"The reader's verdict is CONDITIONAL, and my stress-test agrees with the identified weak point. The central algebraic claim—equations (18)–(20) for the crystallographic quaternion product—is correct: I verified that the cross tensor X = √det(M) M⁻¹ is exactly the transformation that maps the coordinate-wise cross product to the physical cross product in a non-orthogonal basis, and the product law is simply the standard quaternion product expressed covariantly. No internal inconsistency or circular reasoning appears in the derivation. The load-bearing weakness is in the application to EBSD. The paper's motivation and conclusion rest on an efficiency comparison that is (a) internally inconsistent in its FLOP counts ('45 FLOPs (9 M and 6 A)' sums to 15), (b) unbenchmarked, and (c) dependent on the assumption that spherical indexing returns orientations directly in the lattice basis. That assumption is not proven and is contrary to typical EBSD practice, where orientation output is a rotation in a Cartesian frame. If that assumption fails, the crystallographic-quaternion pipeline does not avoid the conversion cost, and the paper's own Table 1 shows the per-product cost is higher than the Cartesian route. The metric 'independence of convention' claim is also overstated, but it is secondary. Since the algebra is sound and the application can be salvaged by removing or properly verifying the efficiency claims, the CONDITIONAL verdict remains appropriate: the paper should not be rejected, but the application claims require verification or substantial qualification.","tokens_in":14481,"tokens_out":13434,"duration_ms":120347,"concrete_test":"Inspect the orientation output of a spherical-indexing EBSD pipeline (e.g., the open-source implementation from Lenthe et al., 2020, or commercial software like EDAX OIM or Oxford AZtec). Check whether the returned orientation is expressed as coordinates in the lattice basis (fractional coordinates, as required by the paper's crystallographic quaternions) or as a rotation matrix in a Cartesian frame. If the latter, the paper's Section 4b assumption fails and the efficiency advantage disappears; alternatively, benchmark end-to-end FLOPs per pixel for both pipelines to verify whether the crystallographic path is actually faster.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's motivating application (Section 4) claims crystallographic quaternions outperform Cartesian quaternions in EBSD because spherical indexing returns orientations 'directly in the crystal basis', avoiding a conversion to Euler angles (Section 4b). This premise is not substantiated and is likely false: spherical-indexing implementations (e.g., Lenthe et al., 2020) typically output a rotation matrix in a Cartesian frame attached to the crystal, not coordinates in the non-orthogonal lattice basis a,b,c. If the output is Cartesian, then using crystallographic quaternions requires an additional structure-tensor conversion to enter the crystal basis, eliminating the claimed savings. Even taken at face value, the paper's own Table 1 shows the crystallographic product (46 FLOPs) is more expensive than the Cartesian product plus axis transformation (43 FLOPs), so the entire advantage hinges solely on the unbenchmarked and internally inconsistent pipeline accounting (the text claims '45 FLOPs (9 M and 6 A)', which sums to 15, not 45). Thus the central practical claim is load-bearing on an unverified and likely incorrect premise about the native output space of spherical indexing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a generalization of unit quaternions to non-orthogonal crystallographic bases. It introduces a 'cross tensor' X = [b×c, c×a, a×b], proves X = sqrt(det M) M^{-1} for the metric tensor M, and uses M and X to write the composition law for quaternions directly in the crystal basis (Eqs. 18-20), together with analogous Rodrigues and vector-rotation formulas (Eqs. 24-25). The final section argues that this construction could make EBSD orientation calculations more efficient by avoiding Cartesian structure-tensor round trips.","tokens_in":14692,"tokens_out":26391,"duration_ms":241094,"significance":"If the stated qualifications are added, the paper gives a correct and useful coordinate formula: the central identity X = sqrt(det M) M^{-1} and the product law (18)-(20) reproduce the standard rotation composition in fractional coordinates, and the derivation in the appendices is self-contained and not circular. The extension to Rodrigues vectors is a useful by-product. The practical EBSD claim, however, is not established, and some stated properties of the metric tensor are overstated; these issues do not invalidate the core algebra but require substantial revision of Section 4.","major_comments":[{"comment":"The efficiency comparison is not reliable. The paper counts a 3×3 symmetric matrix-vector product as 6M+3A, but symmetry reduces storage, not arithmetic; the operation count is 9M+6A. With correct counts, the crystallographic product costs at least 58 FLOPs, not the reported 46. In addition, item (b) states 'additional 45 FLOPs (9 M and 6 A)', which is arithmetically inconsistent (9+6=15), and even accepting the paper's counts the comparison is 46 vs 43 in favor of the Cartesian route. The claimed net advantage therefore rests entirely on the unsubstantiated assertion that spherical indexing returns orientations 'directly in the crystal basis'; the cited implementations return orientation matrices in an orthogonal crystal frame. This section needs a corrected benchmark or should be reframed as a qualitative discussion.","section":"§4 and Table 1"},{"comment":"The statement that the metric tensor is 'independent of any convention' is incorrect. The matrix M is the Gram matrix of the chosen basis vectors (a,b,c); its entries change with the unit-cell setting, axis order, and handedness. What is convention-independent is the underlying bilinear form, not the matrix used in Eqs. (19)-(20). Since the product law is written in a particular crystal basis, it is basis-dependent in the same way as any coordinate formula. This overstatement should be corrected.","section":"§4(d), §2.1"},{"comment":"The identity X = sqrt(det M) M^{-1} is valid only for a right-handed basis. The derivation uses a* = (b×c)/V with V = det(a,b,c) and then sets V = sqrt(det M), conflating signed determinant with positive volume. For a left-handed basis, the correct cross tensor is X = -sqrt(det M) M^{-1}; the metric tensor alone does not encode the orientation of the basis. The manuscript should either explicitly assume the standard right-handed crystallographic setting or carry a signed volume throughout. Without this, the claim that the product law depends uniquely on the metric is not general.","section":"Eq. (16), §2.2"}],"minor_comments":[{"comment":"The text calls M the transformation 'from the reciprocal basis to the direct basis', but the derivation v* = M v shows that M maps direct-basis coordinates to reciprocal-basis coordinates, i.e. M = [B_c -> B_c^*]. Please correct the wording.","section":"§2.1"},{"comment":"The split of FLOP counts for the Cartesian quaternion product is inconsistent: '4 M and 4 A' plus '12 M and 6 A' sums to 16M+10A, not the stated 16M+12A. The correct split is 4M+3A (angular) and 12M+9A (vector).","section":"§4"},{"comment":"The sentence 'with c_i = cos(p_jk) and c_i = sin(p_jk)' should read 'and s_i = sin(p_jk)'.","section":"Appendix 1"},{"comment":"The text says 'u_ij = -u_ji' immediately after defining u_ij as an angle between vectors; if the angles are ordinary positive angles, the relation should be u_ij = u_ji. If signed angles are intended, this should be stated.","section":"Appendix 2"},{"comment":"The code is described as 'available upon request'; for reproducibility, a permanent repository or supplementary material would be preferable.","section":"§3"}],"recommendation":"major_revision","confidential_remarks":"The core mathematical result appears correct and publishable after revision; the main weakness is Section 4, where the FLOP accounting is internally inconsistent and the EBSD advantage rests on an unverified assumption about the output space of spherical indexing. The novelty is moderate but adequate for a crystallographic computing venue. I recommend major revision rather than rejection because the algebra can stand once the overclaims are fixed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the central algebra is correct, and I verified it independently; but the 'cross tensor' is just the scaled inverse metric and the EBSD efficiency argument is not supported by the paper's own FLOP comparison. The paper is worth a careful referee, but mainly as a mathematical note, not as a performance claim.\n\nThe genuinely useful part is equations (18)-(20): quaternion composition written directly in fractional coordinates using the metric tensor and X = sqrt(det M) M^{-1}. The derivation in Appendices 1 and 2 is self-contained, and the spherical laws are proved from the metric. I spot-checked eq (16) on an orthorhombic cell and it gives the correct physical cross products. There are no free parameters and no circularity.\n\nThe soft spots are in the surrounding claims. The 'new' cross tensor is, by the paper's own eq (16), simply sqrt(det M) M^{-1} - the standard reciprocal/dual basis construction. Calling it a new tensor overstates the novelty. The EBSD claim hinges on an unbenchmarked pipeline: the paper's Table 1 shows 46 FLOPs for the crystallographic product vs 43 for Cartesian product plus basis transform, so all advantages come from the asserted savings in converting Kikuchi data to Euler angles. That premise is dubious - spherical indexing implementations typically output orientations in a Cartesian frame attached to the crystal, not in the fractional-coordinate basis. The paper offers no evidence otherwise. There are also internal inconsistencies in the FLOP arithmetic (the text says '45 FLOPs (9 M and 6 A)' where the sum is 15), and the statement that the metric tensor is 'independent of any convention' is misleading since its components depend on the unit-cell setting. Code is not shipped; 'available upon request' is not a substitute.\n\nThese problems don't sink the mathematics. The quaternion product in fractional coordinates is correct and could be useful in crystallographic computing, especially for readers comfortable with the metric-tensor formalism. But the efficiency argument should be either benchmarked or removed.\n\nI'd send it to peer review - a crystallographic journal referee can judge whether the reformulation earns its place - but the author should be asked to fix the FLOP table and either defend or drop the spherical-indexing premise. I wouldn't cite it in my own work on the basis of this version.","headline":"Correct algebra, but the novelty is a repackaged inverse metric and the EBSD efficiency claim is not supported by the paper's own numbers.","tokens_in":15260,"tokens_out":2412,"would_cite":false,"duration_ms":26031,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Quaternion rotations can be multiplied directly in the crystal basis using a metric-tensor product law.","keywords":["quaternions","rotations","metric tensor","cross tensor","crystal basis","electron backscatter diffraction","non-Cartesian coordinates","composition law"],"falsifier":"Take a triclinic cell, encode two rotations as crystallographic quaternions, multiply them by the proposed formula, convert the result to a Cartesian rotation matrix, and compare with the product of their Cartesian matrix forms; any mismatch would refute the product law.","tokens_in":14249,"feed_emoji":"💎","tokens_out":5623,"duration_ms":54079,"temperature":0.7,"pith_summary":"The paper aims to show that the standard quaternion product law, which composes 3D rotations, can be rewritten so that it works directly in the fractional-coordinate basis of a crystal, without transforming to a Cartesian frame. The key move is to replace the ordinary scalar product with the metric tensor and the ordinary cross product with a new 'cross tensor' that equals sqrt(det M) times the inverse of the metric tensor. This yields a product formula that depends only on the lattice parameters and lets orientations be encoded and multiplied in the crystal's own coordinates. The author argues this could make electron backscatter diffraction (EBSD) orientation calculations faster and convention-free.","feed_headline":"Rotations compose right in the crystal basis, no Cartesian detour","feed_subtitle":"A cross tensor built from the metric tensor lets quaternion products run in fractional coordinates","key_machinery":"The cross tensor X = sqrt(det(M)) M⁻¹, a linear map that turns the ordinary coordinate cross product into the geometric cross product in the crystal basis. It is the cross-product counterpart of the metric tensor and, together with M, allows the scalar and vector parts of the quaternion product to be evaluated entirely in non-Cartesian coordinates.","core_discovery":"At the paper's center is a closed-form product law for rotations encoded as quaternions whose axes are written in the fractional coordinates of an arbitrary crystal cell. For two such quaternions q1 = c1 + s1 u~1 and q2 = c2 + s2 u~2, the composition is q3 = q1q2 = c3 + s3 u~3, with c3 = c1c2 − s1s2 (u~1ᵀ M u~2) and s3 u~3 = c2 s1 u~1 + c1 s2 u~2 + s1 s2 X (u~1 × u~2), where M is the metric tensor of the lattice and X = sqrt(det(M)) M⁻¹. Because X is a function of M alone, the whole composition rule is set by the lattice parameters. The paper argues that this removes the need to shuttle between crystal and Cartesian coordinates during rotation composition.","pith_inferences":["The FLOP advantage hinges on the unverified assumption that spherical indexing can deliver orientations directly in crystal coordinates; if that fails, the efficiency argument collapses even though the product law itself is unaffected.","Since X and M depend only on the phase, they can be precomputed once per material, making the per-product cost effectively that of a few matrix-vector products, a point the paper leaves implicit.","The construction suggests a general recipe: any rotation formula written with ordinary dot and cross products can be lifted to a non-Cartesian basis by substitution (dot becomes M-dot, cross becomes X-cross), potentially simplifying other crystallographic calculations."],"forward_implications":["EBSD orientation maps could be processed end to end in fractional coordinates, eliminating structure-tensor conversions per pixel.","Symmetry recognition in orientation maps simplifies because symmetry axes have coordinates 0 or ±1 in the crystal basis.","The same construction transplants to other rotation representations that rely on dot and cross products, such as compact axis-angle forms.","The computational cost per product is 46 FLOPs with the crystallographic formula, versus 28 for the Cartesian formula alone, but the paper argues the full pipeline favors the crystallographic route."],"fun_headline_variants":["Quaternion product in crystal basis, no Cartesian detour","Crystallographic quaternions: product law from metric tensor alone","Direct rotation composition in fractional coordinates via cross tensor","Quaternion composition in crystal basis: cross tensor replaces Cartesian","Metric tensor alone sets crystallographic quaternion product"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The practical case for using these quaternions in EBSD rests on an unbenchmarked cost assumption: converting Kikuchi-derived data into Euler angles costs roughly 45 extra floating-point operations per pixel, and spherical indexing can supply orientations directly in crystal coordinates.","fun_headline_variants_meta":{"raw":{"variants":["Quaternion product in crystal basis, no Cartesian detour","Crystallographic quaternions: product law from metric tensor alone","Direct rotation composition in fractional coordinates via cross tensor","Quaternion composition in crystal basis: cross tensor replaces Cartesian","Metric tensor alone sets crystallographic quaternion product"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001014,"raw_usage":{"total_tokens":4134,"prompt_tokens":773,"completion_tokens":3361,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":517,"completion_tokens_details":{"reasoning_tokens":3292}},"tokens_in":517,"tokens_out":3361,"duration_ms":22160,"temperature":1.0,"reasoning_tokens":3292,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T19:38:28.665927+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a triclinic cell, encode two rotations as crystallographic quaternions, multiply them by the proposed formula, convert the result to a Cartesian rotation matrix, and compare with the product of their Cartesian matrix forms; any mismatch would refute the product law.","supporting_citations":[],"review_version":1}