{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp09_archetype_breadth150_20260804","cell_id":"search_space_pruning__chemistry_materials","arm":"BREADTH_PROBE_ONE_SHOT","candidate_id":"search_space_pruning__chemistry_materials__P1","proposal_index":1,"version":0,"title":"Auditable Branch Pruning for Solid-State Electrolyte Substitution Search","problem":"A solid-state electrolyte research team has enumerated a branching candidate space comprising crystal-structure prototypes, charge-balancing elemental substitutions, site occupancies, and processing windows. The resulting set is too large for available synthesis and characterization capacity. Researchers therefore narrow it informally, but composition branches that are charge-impossible, prohibited by laboratory safety rules, or incapable of meeting mandatory processing limits can survive into costly evaluation, while unconventional but feasible branches can disappear without a recorded reason.","actors":["Solid-state electrolyte chemists who define candidate families and interpret measurements","Computational materials scientists who enumerate substitutions and calculate feasibility bounds","Laboratory manager who allocates synthesis and characterization capacity","Environmental health and safety officer who owns prohibited-material constraints","Principal investigator who approves pruning criteria and exceptions"],"observable_state":"The team can inspect an enumeration in which each candidate is traceable to a prototype, substitution branch, site assignment, nominal composition, and processing window. The experiment queue exceeds available capacity; some queued candidates later fail elementary charge-balance, permitted-element, precursor-compatibility, or process-temperature checks; and spreadsheet shortlists do not consistently record which candidates were excluded, why, or how they could reenter.","consequence":"Scarce synthesis and characterization effort is spent on candidates whose infeasibility could have been detected earlier, comparison remains stalled by the candidate count, and undocumented narrowing creates an unmeasured risk that feasible chemistry families are prematurely abandoned.","affected_objective":"Use a fixed synthesis and characterization budget to examine a tractable, chemically diverse set of solid-electrolyte candidates that could still satisfy mandatory safety, charge-balance, processing, and performance requirements.","intervention":"Represent the enumeration as a tree whose upper levels are structure prototype and substitution family and whose lower levels are site occupancy, stoichiometry, and processing window. Apply a governed branch-and-bound screen: hard-prune a branch only when every descendant violates an exact constraint or when a conservative bound shows that no descendant can reach a mandatory threshold; treat uncertain model predictions as prioritization signals rather than exclusions. Retain stratified representatives from uncertain and pruned families for false-negative review, record a machine-readable certificate for every exclusion, and reopen branches when measurements, constraints, or bound uncertainty change.","structural_mapping":[{"archetype_element":"Search Space Definition","domain_realization":"All enumerated combinations of approved structure prototypes, elemental substitutions, site occupancies, stoichiometries, and processing windows for the electrolyte campaign."},{"archetype_element":"Candidate or Region Representation","domain_realization":"A tree in which a branch denotes all descendants of a prototype-and-substitution choice and a leaf denotes one composition and processing-window candidate."},{"archetype_element":"Objective or Success Criterion","domain_realization":"Retain candidates capable of satisfying the campaign's frozen mandatory safety, charge-balance, processing, and measured-property requirements within the available assay budget."},{"archetype_element":"Constraint Filter","domain_realization":"Exact checks for charge neutrality, permitted elements and precursors, stoichiometric validity, equipment temperature limits, and declared incompatibilities."},{"archetype_element":"Feasibility Test","domain_realization":"Interval or conservative descendant-level bounds determine whether any member of a branch could still satisfy each mandatory requirement."},{"archetype_element":"Pruning Rule","domain_realization":"Exclude an entire branch only when an exact constraint fails for every descendant or a documented conservative bound rules out every descendant; otherwise retain or merely defer it."},{"archetype_element":"Diversity Preservation Rule","domain_realization":"Keep at least one reviewable representative from each surviving prototype and substitution family, plus a stratified holdout from excluded uncertain regions."},{"archetype_element":"False-Negative Review","domain_realization":"A blinded reviewer examines holdout exclusions against later-known measurements and checks whether any qualifying or near-boundary candidate was removed."},{"archetype_element":"Reentry or Exception Path","domain_realization":"The principal investigator may restore a branch from the immutable enumeration when new measurements widen a bound, a constraint changes, or an exclusion certificate is challenged."},{"archetype_element":"Pruning Audit Trail","domain_realization":"Each exclusion stores the branch identifier, rule, input values, uncertainty assumption, rule version, timestamp, and accountable reviewer."},{"archetype_element":"Pruning Owner","domain_realization":"The principal investigator owns scientific criteria, while the environmental health and safety officer retains authority over prohibited-material rules."},{"archetype_element":"Criteria Update Trigger","domain_realization":"A known-feasible holdout exclusion, systematic near-threshold error, revised equipment limit, changed safety rule, or new measurement method triggers rule review and possible reentry."}],"mechanism_mapping":[{"mechanism_slug":"branch_and_bound","role":"Uses conservative descendant-level bounds to eliminate a whole substitution branch only when it cannot meet a mandatory requirement.","counterfactual_removal":"Without branch-level bounds, the system could filter individual leaves but would not remove large regions early enough to make the combinatorial search tractable."},{"mechanism_slug":"constraint_filtering","role":"Removes candidates that fail exact chemical, safety, or processing requirements before synthesis.","counterfactual_removal":"Without explicit constraint filtering, elementary infeasibilities would continue into downstream calculations or laboratory work."},{"mechanism_slug":"sample_audit_of_exclusions","role":"Tests stratified samples of excluded branches for false negatives and model-family bias.","counterfactual_removal":"Without exclusion audits, apparent efficiency could coexist with undetected removal of feasible or atypical chemistry families."}],"causal_chain":["Prototype, substitution, occupancy, and processing choices create a candidate tree larger than the laboratory can evaluate exhaustively.","Exact constraints and conservative bounds identify branches whose descendants cannot satisfy mandatory requirements.","Branch-and-bound and constraint filtering remove those branches before synthesis while uncertain branches are deferred rather than deleted.","The remaining candidate set becomes small enough for detailed calculation, synthesis planning, and comparison under the fixed assay budget.","Diversity retention prevents the surviving set from collapsing to one familiar chemistry family.","Exclusion certificates, holdout audits, and reentry triggers expose false negatives and allow pruning decisions to be reversed when their assumptions fail."],"baseline":"Researchers manually shortlist individual compositions in spreadsheets using expert judgment and selected computed properties, then synthesize candidates sequentially. Candidate families may be dropped implicitly, hard constraints and preferences are not consistently separated, and exclusions lack standardized audit or reentry records.","nearest_rivals":["Design of experiments samples a composition or process space efficiently, but it does not necessarily certify and remove entire infeasible branches or preserve an exclusion ledger.","Bayesian optimization ranks the next candidate using an updated surrogate model, whereas this intervention reserves hard pruning for exact constraints or conservative impossibility bounds and treats uncertain predictions as priorities.","High-throughput synthesis evaluates more candidates in parallel, reducing evaluation bottlenecks without eliminating predictably infeasible regions before physical work.","Standalone thermodynamic or property screening scores candidates, but does not by itself supply the hard-versus-soft decision rule, stratified exclusion audit, accountable ownership, and reentry path specified here."],"remaining_contrastive_claim":"The claim to test is structural: representing electrolyte substitutions as branches and permitting exclusion only through exact constraints or conservative descendant-level bounds, coupled to diversity holdouts and reversible exclusion records, can make the campaign search tractable in a way that candidate ranking or faster testing alone does not. This is not a claim of novelty, prevalence, demand, or effect magnitude.","authority_safety":{"decision_authority":"The principal investigator may approve scientific pruning rules and restore branches; the environmental health and safety officer alone may approve or revise prohibited-material constraints. The software may recommend and record exclusions but has no authority to order synthesis or waive safety requirements.","authorized_first_step":"Run a read-only retrospective evaluation on one completed campaign using frozen pre-synthesis descriptors and hidden outcome labels; do not change an active experiment queue.","excluded_actions":["Autonomously ordering or executing chemical synthesis","Overriding environmental health and safety exclusions","Hard-pruning a branch solely from an uncertain property prediction or familiarity heuristic","Deleting the original enumeration, measurements, or exclusion records","Presenting a retained candidate as safe or effective without the laboratory's ordinary validation"],"halt_rollback":"Halt if a hard-pruning rule excludes a known qualifying holdout, lacks a reproducible certificate, or cannot distinguish an exact constraint from a predictive preference. Disable the implicated rule, restore all affected branches from the immutable enumeration, and require joint scientific and safety review before reuse."},"negative_tests":{"strongest_counterevidence":"Electrolyte performance may depend on emergent defects, metastable phases, interfaces, and processing history that are not conservatively bounded from prototype and nominal composition. If these effects make descendant-level bounds unreliable, branch pruning would create unjustified confidence and could remove the useful region.","problem_falsifier":"The problem is falsified for the selected campaign if the full candidate set can be evaluated within the existing budget and safety envelope, or if retrospective records show that early-detectable infeasibility is not consuming meaningful downstream review or laboratory capacity.","intervention_falsifier":"The intervention is falsified if, under frozen rules and blinded outcomes, it hard-prunes any known qualifying candidate without flagging that branch for holdout review and reentry, or if valid conservative bounds cannot eliminate branches beyond what the exact leaf-level baseline already removes.","risks":["Proxy constraints could favor familiar crystal prototypes and erase feasible atypical families.","Uncertainty may be understated, turning soft predictions into false impossibility certificates.","Coarse branch representation may hide a feasible descendant inside an otherwise weak family.","Safety or equipment rules may become stale while remaining encoded as hard exclusions.","Maintaining bounds, audits, and holdouts may shift substantial effort upstream rather than reduce total evaluation burden.","Repeated exceptions could recreate the original overload or encourage inconsistent rule application."]},"next_evidence_step":"Select one completed electrolyte-enumeration campaign capped at 500 candidates. Before revealing outcomes, freeze the search tree, exact constraints, conservative-bound method, diversity rule, and baseline shortlist procedure. Apply both procedures using only information available before synthesis; then reveal outcomes and record candidate and branch counts retained, known qualifying candidates retained, chemistry-family coverage, certificate reproducibility, review effort, and every false-negative or reentry event. Conduct no new synthesis and stop after this single retrospective audit.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"No other experiment candidates or proposals were inspected. This candidate is derived solely from the supplied archetype and domain card and realizes pruning as reversible, branch-level exclusion in a crystal-substitution search.","revision_record":{"parent_version":null,"progress_targets_addressed":[],"conceptual_changes":[],"operational_changes":[],"evidence_changes":[],"claim_changes":[]}}