{"schema_version":1,"research_id":"eoa_inverse_innovation_exp06_external_evaluation_20260803","source_assessment_id":"representation_independent_interface_contract__chemistry_materials:P1:v0","cell_id":"representation_independent_interface_contract__chemistry_materials","search_queries":["site:optimade.org specification structures OPTIMADE interoperability materials databases","OPTIMADE common API materials databases interoperability paper 2021","site:iucr.org CIF core dictionary atom sites symmetry official","periodic crystal structure equivalence canonicalization algorithm materials database identifier paper","pymatgen StructureMatcher documentation primitive_cell scale attempt_supercell comparator official","spglib documentation crystal symmetry cell standardize official","MatID materials identifier crystal structures equivalence paper","NOMAD materials identifiers periodic structures MatID normalized structure interoperability","site:pymatgen.org pymatgen.analysis.structure_matcher StructureMatcher documentation","site:materialsproject.github.io pymatgen StructureMatcher fit anonymous official docs","site:nomad-lab.eu MatID material identifier algorithm structure identity","materials hash identifier crystalline structure canonical equivalence MatID paper","XtalComp open source duplicate crystal structures 2012 DOI official full text","site:bls.gov Occupational Employment Wage Statistics software developers May 2025 annual mean wage","site:bls.gov OEWS materials scientists annual mean wage 2025","BLS employer costs employee compensation private industry 2026 benefits percentage"],"sources":[{"source_id":"S1","title":"Identifying duplicate crystal structures: XtalComp, an open-source solution","publisher":"Computer Physics Communications / Elsevier","url":"https://www.sciencedirect.com/science/article/pii/S0010465511003699","source_class":"PRIMARY_RESEARCH","publication_date":"2012-03-01","accessed_at":"2026-08-03","claims_supported":["A periodic crystal has infinitely many valid unit-cell descriptions, making representation-level equality unreliable.","Numerical noise, periodic boundaries, degenerate lattices, and absence of a canonical origin create practical duplicate-detection problems.","XtalComp is an implemented representation-insensitive structure-comparison system and therefore close prior art."]},{"source_id":"S2","title":"pymatgen.core package: structure_matcher module","publisher":"Materials Project / pymatgen","url":"https://pymatgen.org/pymatgen.core.html","source_class":"OFFICIAL_PRODUCT_DOCUMENTATION","publication_date":"2026-07-27 documentation release","accessed_at":"2026-08-03","claims_supported":["pymatgen already groups periodic structures using configurable lattice, site, and angular tolerances, primitive-cell reduction, scaling, and species comparators.","Structure matching can ignore selected species or perform anonymous species mappings, showing that equivalence is policy-dependent.","Existing production-grade primitives make a sandbox comparator technically feasible but substantially overlap the proposed equivalence operation."]},{"source_id":"S3","title":"Spglib conventions of standardized unit cell","publisher":"Spglib project","url":"https://spglib.readthedocs.io/en/latest/definition.html","source_class":"OFFICIAL_PRODUCT_DOCUMENTATION","publication_date":"date not stated","accessed_at":"2026-08-03","claims_supported":["Spglib implements cell standardization through change-of-basis transformations and exposes primitive-cell operations.","Its standardized cell is not fully unique, even after crystallographic conventions, supporting the need to distinguish abstract structure from canonical encoding.","Spglib supplies an available canonicalization comparator but not the proposed multi-implementation behavioral contract."]},{"source_id":"S4","title":"OPTIMADE API specification v1.3.0","publisher":"Open Databases Integration for Materials Design consortium","url":"https://www.optimade.org/specification/latest/","source_class":"STANDARD","publication_date":"version 1.3.0; date not stated","accessed_at":"2026-08-03","claims_supported":["OPTIMADE already specifies versioned, machine-readable structures entries, identifiers, lattice vectors, site positions, species, and extensibility rules.","The standard permits provider-dependent underlying representations and provider-specific properties.","OPTIMADE standardizes exchange and retrieval but does not require physical-equivalence semantics, opacity, immutability, or one black-box behavioral oracle across implementations."]},{"source_id":"S5","title":"OPTIMADE, an API for exchanging materials data","publisher":"Scientific Data / Springer Nature","url":"https://www.nature.com/articles/s41597-021-00974-z","source_class":"PRIMARY_RESEARCH","publication_date":"2021-08-12","accessed_at":"2026-08-03","claims_supported":["Materials-database providers explicitly reported interoperability and client-maintenance problems caused by differing and evolving APIs, taxonomies, semantics, and structure representations.","Providers including major materials databases jointly designed OPTIMADE, and multiple leading databases and software packages implemented it.","The consortium and participating database operators are identifiable potential adopters or standardization partners, although no source expresses commitment to this stronger contract."]},{"source_id":"S6","title":"Core CIF dictionary","publisher":"International Union of Crystallography","url":"https://www.iucr.org/resources/cif/dictionaries/browse/cif_core1","source_class":"STANDARD","publication_date":"date not stated","accessed_at":"2026-08-03","claims_supported":["CIF already standardizes crystallographic fields for fractional coordinates, symmetry operations, translations, units, and related metadata.","CIF is a mature exchange-format rival but does not itself define representation-independent object equality, substitution acceptance, mutation behavior, or behavioral conformance across software implementations."]},{"source_id":"S7","title":"NOMAD — Materials science data, managed and shared","publisher":"NOMAD / FAIRmat","url":"https://nomad-lab.eu/nomad-lab/","source_class":"OFFICIAL_ORGANIZATION_DATA","publication_date":"date not stated","accessed_at":"2026-08-03","claims_supported":["NOMAD publicly expresses a need to extract structured data from more than 60 formats and provide one unified, machine-actionable representation across heterogeneous codes and workflows.","NOMAD and FAIRmat are identifiable infrastructure operators with adjacent adoption authority and relevant workflow expertise.","The page supports interoperability demand but does not document demand for the proposed opaque immutable structure contract or its conformance suite."]},{"source_id":"S8","title":"Employer Costs for Employee Compensation — March 2026","publisher":"U.S. Bureau of Labor Statistics","url":"https://www.bls.gov/news.release/pdf/ecec.pdf","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"2026-06-12","accessed_at":"2026-08-03","claims_supported":["Private-industry professional occupations cost employers about $72.99 per hour in wages and benefits in March 2026.","Private-industry benefits averaged 30.1 percent of total compensation, supporting resource-equivalent labor estimates rather than salary-only estimates.","The published compensation data support order-of-magnitude cost bands but not vendor quotes or institution-specific budgets."]}],"problem_evidence":{"support":"STRONG","rationale":"Primary research directly documents that many unit-cell descriptions can encode the same periodic structure and that numerical tolerances, periodicity, lattice choices, and origin choices complicate duplicate detection. OPTIMADE independently documents semantic, representational, and maintenance friction across materials databases. What remains unmeasured is how often the candidate's specific client failures—cache misses, identifier changes, snapshot breakage, and parser-substitution regressions—occur in current production repositories.","source_ids":["S1","S5"]},"stakeholder_evidence":{"support":"MODERATE","rationale":"The OPTIMADE consortium, its participating database providers, and NOMAD/FAIRmat are identifiable operators that have publicly expressed and acted on a need for interoperable materials data and unified access across representations. This establishes credible adopter classes and adjacent authority, but no source requests, funds, or commits to the proposed immutable abstract type, behavioral equivalence rules, or conformance oracle.","source_ids":["S5","S7"]},"prior_art":{"proximity":"ADJACENT_PRIOR_ART","closest_analogues":[{"name":"XtalComp","similarity":"Directly distinguishes a crystal's underlying periodic structure from its many unit-cell representations and implements duplicate detection robust to periodic and numerical representation differences.","remaining_difference":"It is a comparison library, not a versioned opaque abstract data type governing construction, typed failures, provenance, serialization, side effects, and substitutability across all parsers and stores.","source_ids":["S1"]},{"name":"pymatgen StructureMatcher","similarity":"Provides mature, configurable periodic-structure matching with primitive reduction, scaling, species comparators, and explicit numerical tolerances.","remaining_difference":"It exposes one library's data model and matching policy; it is not an implementation-neutral contract or shared black-box certification suite, and its policy options underscore rather than resolve the scientific choice of equivalence.","source_ids":["S2"]},{"name":"Spglib cell standardization","similarity":"Normalizes crystal cells under symmetry and basis transformations and can produce primitive or standardized cells.","remaining_difference":"It is a canonicalization and symmetry tool; its own documentation notes residual non-uniqueness, and it does not specify client-visible behavioral laws, provenance, errors, or substitution acceptance.","source_ids":["S3"]},{"name":"OPTIMADE v1.3","similarity":"A versioned, implementation-flexible standard for materials-database APIs and structure fields, already implemented across multiple providers.","remaining_difference":"It standardizes retrieval shape and selected field semantics, not physical-equivalence behavior, immutable operations, representation opacity, or one semantic oracle for interchangeable structure implementations.","source_ids":["S4","S5"]},{"name":"IUCr Core CIF dictionary","similarity":"A mature standard vocabulary for crystallographic cells, sites, coordinates, symmetry operations, translations, and units.","remaining_difference":"It defines an exchange representation rather than representation-independent equality, invariants across operations, no-mutation errors, or parser/store substitutability.","source_ids":["S6"]},{"name":"NOMAD unified materials representation","similarity":"Normalizes many heterogeneous file formats into unified, structured, machine-actionable materials data.","remaining_difference":"It is an operational data platform and schema ecosystem, not evidence of an independently implemented behavioral ADT contract with preregistered metamorphic acceptance laws.","source_ids":["S7"]}],"distinctive_claim_remaining":"Under a preregistered ordered-periodic scope and tolerance policy, two independently structured adapters that share no atom ordering, storage layout, or canonicalization implementation will satisfy the same public-only oracle for identity/equivalence, invariant summaries, typed failures, immutability, and provenance across representation-only transformations, while producing fewer semantic divergences and unpromised observable leaks than (a) the current array/serializer baseline, (b) a spglib/pymatgen canonicalization-only comparator, and (c) an OPTIMADE/CIF schema-only round trip. This is falsified by any preregistered law violation, tolerance-induced non-transitive fixture triple, provenance loss, required exposure of hidden representation, or failure to improve divergence/leakage counts over the comparators.","confidence":"MODERATE"},"implementation_evidence":{"support":"MODERATE","rationale":"Existing standards and open implementations supply parsers, schemas, symmetry normalization, structure matching, versioning patterns, and unified materials-data infrastructure, so a two-adapter read-only prototype is technically credible. The difficult unresolved work is scientific rather than basic software construction: selecting equivalence observables, proving or bounding tolerance behavior, preserving distinctions such as magnetism and chirality, and preventing a finite suite from over- or under-specifying behavior. No external source demonstrates the proposed end-to-end contract.","source_ids":["S2","S3","S4","S6","S7"]},"scores":{"meaningful_impact":{"score":4,"rationale":"If the stated failures are prevalent, consistent identity, safer parser substitution, and clearer provenance would materially improve computational-materials workflows. Production prevalence and realized effect size are unmeasured.","source_ids":["S1","S5"]},"stakeholder_pull":{"score":3,"rationale":"Major database operators have acted on closely related interoperability needs, but there is no expressed demand or commitment for this specific behavioral contract.","source_ids":["S5","S7"]},"incremental_advantage":{"score":2,"rationale":"The proposal adds a unified behavioral and substitution boundary, but most constituent capabilities already exist in XtalComp, pymatgen, spglib, CIF, OPTIMADE, and NOMAD. Superiority over combining those tools is untested.","source_ids":["S1","S2","S3","S4","S6","S7"]},"distinctiveness_plausibility":{"score":3,"rationale":"A public-only, multi-implementation semantic oracle covering errors, immutability, provenance, leakage, and structure invariances is a testable remaining distinction, though it is largely an integration and governance claim rather than a new equivalence algorithm.","source_ids":["S1","S2","S4"]},"technical_implementability":{"score":4,"rationale":"Open libraries and standards provide most required parsing, symmetry, matching, schema, and validation primitives. Scientific policy and tolerance semantics remain material technical risks.","source_ids":["S2","S3","S4","S6"]},"adoption_authority_feasibility":{"score":3,"rationale":"Database platform owners and scientific-method leads are identifiable authorities, and a read-only sandbox is low-authority. Cross-platform standardization or production identifier migration would require partner commitments not evidenced here.","source_ids":["S5","S7"]},"evidence_readiness":{"score":3,"rationale":"The problem and adjacent solutions are externally documented, and the proposed benchmark is bounded. Candidate-specific benefit, repository prevalence, and adapter conformance require live empirical work.","source_ids":["S1","S2","S3","S5"]},"safety_net_benefit":{"score":4,"rationale":"Copied fixtures, isolated adapters, no production writes, preregistered stopping rules, and discardable sandbox artifacts make the first test reversible. Mislabeling scientific identity remains a risk if results escape the sandbox.","source_ids":["S2","S3","S4"]},"scalability":{"score":3,"rationale":"One parameterized suite could scale across implementations, but the contract and fixture space grow sharply when disorder, defects, magnetism, surfaces, trajectories, and domain-specific provenance enter scope.","source_ids":["S2","S4","S6","S7"]}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"10K_TO_50K","scope":"Preregister the version-0 contract; curate 12 copied fixtures and 60 seeded variants; build two minimal sandbox adapters; run 200 seeded sequences and three comparators; document divergences and leakage.","confidence":"MODERATE","assumptions":["Approximately 250-550 professional hours split between a materials scientist and software engineer.","Existing open-source libraries and non-sensitive fixtures are reused.","No production integration, procurement, or external security review is included.","BLS professional compensation is used as a resource-equivalent baseline, with modest compute and review overhead."],"source_ids":["S2","S3","S8"]},"initial_deployment_startup":{"band_2026_usd":"50K_TO_250K","scope":"Harden the contract and test harness for one materials-data platform, implement two to four production-quality adapters, add CI, provenance validation, documentation, and scientific governance.","confidence":"MODERATE","assumptions":["Roughly 0.5-1.5 resource-equivalent professional FTE-years.","Existing database identifiers remain unchanged during startup.","Open-source dependencies remain suitable and no large licensing expense is incurred.","Scope stays limited to fully ordered three-dimensional periodic structures."],"source_ids":["S2","S3","S4","S8"]},"operational_launch":{"band_2026_usd":"250K_TO_1M","scope":"Production rollout across a platform and several downstream consumers, including shadow comparisons, migration tooling, monitoring, incident rollback, user documentation, and scientific sign-off.","confidence":"LOW","assumptions":["Approximately 2-5 professional FTE-years plus infrastructure and coordination.","Launch includes multiple parser/store/analysis integrations but not a field-wide standardization campaign.","No destructive database deduplication is performed automatically.","Cost could exceed the band if legacy-client inventories or identifier migrations are extensive."],"source_ids":["S4","S5","S7","S8"]},"annual_recurring":{"band_2026_usd":"50K_TO_250K","scope":"Contract stewardship, version review, fixture expansion, dependency updates, conformance CI, leakage audits, user support, and scientific adjudication of disputed equivalence cases.","confidence":"MODERATE","assumptions":["Approximately 0.5-1.5 professional FTE-years annually.","Compute requirements remain modest relative to labor.","Major scope extensions such as disorder, magnetism, surfaces, and trajectories are separately funded."],"source_ids":["S4","S7","S8"]}},"verified_pipeline_gates":{"externally_supported_problem":{"status":"YES","reason":"Independent primary sources document non-unique periodic representations, duplicate-detection difficulty, semantic incompatibility, and client maintenance burden.","source_ids":["S1","S5"]},"externally_credible_adopter_or_authorizer":{"status":"YES","reason":"OPTIMADE database providers and NOMAD/FAIRmat are identifiable infrastructure operators that have implemented adjacent interoperability solutions and possess relevant adoption authority. Their commitment to this proposal is not established.","source_ids":["S5","S7"]},"distinct_testable_incremental_claim":{"status":"YES","reason":"The remaining claim compares one public behavioral oracle across independently represented adapters against array/serializer, canonicalization-only, and schema-only comparators, with observable divergence, leakage, provenance, and invariance outcomes.","source_ids":["S1","S2","S3","S4","S6"]},"bounded_next_evidence_step":{"status":"YES","reason":"The proposed test is limited to 12 copied fixtures, five seeded variants per fixture, two adapters, 200 operation sequences, three comparators, and preregistered falsifiers without production changes.","source_ids":["S1","S2","S3"]},"no_unresolved_safety_or_authority_stop":{"status":"YES","reason":"For the sandbox only, copied fixtures, read-only execution, discardable adapters, and exclusion of production identifiers and simulations bound the risk. Production deployment remains outside this gate.","source_ids":["S4","S7"]},"credible_cost_scope_and_range":{"status":"YES","reason":"All four bands state included work and FTE-equivalent assumptions and are calibrated to March 2026 BLS employer compensation data. Launch uncertainty remains high because no partner architecture or repository inventory was available.","source_ids":["S8"]}},"next_evidence_step":"With one materials-data platform partner and a designated scientific-method lead, preregister the ordered-periodic abstract state, observables, tolerance policy, and exclusions before inspecting adapter differences. Use 12 copied fixtures spanning simple and skewed cells, multiple species, wrapped boundary sites, alternative conventional/primitive descriptions, and near-tolerance cases. Generate five seeded representation-only variants per fixture: atom permutation, origin shift, periodic wrapping, unit conversion, and admissible cell re-expression. Implement an array-backed adapter and an independently structured map/graph-backed adapter. Run fixed assertions and 200 seeded operation sequences against both. Compare results with the existing array/serializer behavior, a spglib-plus-pymatgen canonicalization path, and an OPTIMADE/CIF schema round trip. Measure law violations, pairwise and three-way equivalence consistency, identity/cache divergence, invariant-summary disagreement, typed-error agreement, provenance retention, unpromised observable leaks, runtime, and memory. Stop and reject version 0 if any declared meaning-preserving transformation changes a preregistered observable; any fixture triple demonstrates tolerance-induced non-transitivity; provenance is lost; either adapter needs atom order or format-specific fields to classify a fixture; or the contract does not reduce semantic divergences or leakage relative to the best comparator. This step requires live coding and testing and cannot be resolved by further bounded web search.","blocking_evidence":["A measured dependency inventory from real parser, database, cache, descriptor, and adapter repositories showing the prevalence and severity of representation coupling.","Results from the preregistered two-adapter benchmark, including comparator performance and deliberately broken-implementation mutation tests.","Scientific approval that the version-0 abstract state and tolerance policy preserve all distinctions relevant to the selected ordered-periodic fixtures.","A partner's explicit authorization and adoption criteria for any work beyond the read-only sandbox.","Evidence about performance and operational compatibility under representative production workloads."],"research_disposition":"PARTNERED_RESEARCH_PROGRAM","world_novelty_boundary":"This eight-source open-web evaluation measures neither world novelty nor patentability, freedom to operate, market size, or realized impact. It establishes only that close component-level prior art and standards exist and that the proposed combined behavioral-contract claim was not demonstrated in the reviewed sources; unreviewed publications, repositories, standards work, patents, and proprietary systems may contain closer matches.","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_version":0,"controller_recommendation":{"action":"STOP_EMPIRICAL_RESEARCH_NEEDED","repairable":false,"material_progress_observed":true,"progress_targets":["Secure a materials-data platform partner and scientific-method approver for the sandbox.","Measure representation-coupling prevalence in real workflow repositories and dependency inventories.","Execute the preregistered two-adapter benchmark with array/serializer, canonicalization-only, and schema-only comparators.","Demonstrate deterministic, scientifically approved tolerance behavior, including explicit tests for non-transitivity and omitted material distinctions.","Quantify semantic divergence reduction, provenance retention, leakage, runtime, and memory before considering production adoption."],"reason":"External evidence verifies the underlying representation and interoperability problem, identifies credible adjacent adopters, and reveals substantial but not complete prior art. The remaining advantage is an empirical systems claim: that a preregistered behavioral contract outperforms existing matching, canonicalization, and schema approaches across independent adapters without scientific information loss. Resolving it requires repository fieldwork and live adapter testing, so additional bounded web search cannot complete the evaluation."},"proposal_index":1}