Attention Is Not All You Need for Diffraction

Determining crystal symmetry from powder x-ray diffraction is a central problem in materials characterization, yet multiple space groups produce indistinguishable patterns, and automated methods plateau at low accuracy on real data. The open question is the recoverable symmetry information is lost. We cast symmetry determination as decoding through a noisy measurement channel and measure its capacity directly. On our real-data benchmark, a Bayes-optimal decoder that is given the pattern of systematically absent reflections recovers 67 % Top-1 at the benchmark's class distribution, yet end-to-end model accuracy is only ∼ 10 % . An explicit two-stage decomposition localizes nearly all of this gap to identifying truly absent reflections in noisy, degraded diffraction patterns. Once the absence pattern is known, mapping it to the correct symmetry contributes little additional error. To probe this limit with a strong decoder, we introduce a physics-informed transformer with an explicit sin 2 θ coordinate channel, physics-aware positional encoding, and multitask outputs that separate geometric rule learning from holistic pattern recognition. We train the model on a synthetic-distribution curriculum chosen to reduce train-test overlap risk. This model is competitive with or stronger than published baselines and still hits the same wall, showing that the limit is structural. We also show that several recent state-of-the-art models share this wall. Mapping residual errors onto the directed acyclic graph of maximal symmetry-reducing () subgroups shows physically structured errors that are local on the hierarchy and biased toward lower-symmetry descendants, a signature of erased absence cues. The same hierarchy makes classical Pawley verification tractable by bounding the search. Further improvements in automated symmetry determination will require more reliable recovery of systematic absences from experimental data.

Authors

Institutions

Publication Details

Journal
PRX Intelligence
Published
2026-11-17
DOI
https://doi.org/10.1103/88gm-4gjx
Citations
1
Primary Topic
X-ray Diffraction in Crystallography
Type
article
Field-Weighted Citation Impact
0.87

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Attention Is Not All You Need for Diffraction

William Ratcliff, Abhishek Shetty, Harshita Dwarcherla, Elizabeth Baggett et al.
1 citations
PRX Intelligence
X-ray Diffraction in Crystallography
0.87
article

Attention Is Not All You Need for Diffraction

William Ratcliff, Abhishek Shetty, Harshita Dwarcherla, Elizabeth Baggett, Edward G. Friedman, Derrick Chan-Sew, Paul Kienzle, Vanellsa Acha
article en
1 citations

Abstract

Determining crystal symmetry from powder x-ray diffraction is a central problem in materials characterization, yet multiple space groups produce indistinguishable patterns, and automated methods plateau at low accuracy on real data. The open question is the recoverable symmetry information is lost. We cast symmetry determination as decoding through a noisy measurement channel and measure its capacity directly. On our real-data benchmark, a Bayes-optimal decoder that is given the pattern of systematically absent reflections recovers 67 % Top-1 at the benchmark's class distribution, yet end-to-end model accuracy is only ∼ 10 % . An explicit two-stage decomposition localizes nearly all of this gap to identifying truly absent reflections in noisy, degraded diffraction patterns. Once the absence pattern is known, mapping it to the correct symmetry contributes little additional error. To probe this limit with a strong decoder, we introduce a physics-informed transformer with an explicit sin 2 θ coordinate channel, physics-aware positional encoding, and multitask outputs that separate geometric rule learning from holistic pattern recognition. We train the model on a synthetic-distribution curriculum chosen to reduce train-test overlap risk. This model is competitive with or stronger than published baselines and still hits the same wall, showing that the limit is structural. We also show that several recent state-of-the-art models share this wall. Mapping residual errors onto the directed acyclic graph of maximal symmetry-reducing () subgroups shows physically structured errors that are local on the hierarchy and biased toward lower-symmetry descendants, a signature of erased absence cues. The same hierarchy makes classical Pawley verification tractable by bounding the search. Further improvements in automated symmetry determination will require more reliable recovery of systematic absences from experimental data.

PRX IntelligenceVol. 1(2)
Boston College (US), National Institute of Standards and Technology (US), NIST Center for Neutron Research (US), University of Maryland, College Park (US), Carnegie Mellon University (US), University of California, Berkeley (US)
National Science Foundation, University of Texas at Austin, National Institute of Standards and Technology, Division of Materials Research
Openalex Percentile: Top 36%
X-ray Diffraction in Crystallography
0.87
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon