{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp12_substrate_denial72_20260805","research_id":"eoa_inverse_innovation_exp12_light_screen_20260805","cell_id":"authority_mentor_relationship_anchoring__computer_science","search_lanes":{"direct_problem_and_intervention":{"queries":["memristor crossbar analog self programming differential pair reference conductance copy circuit","memristor conductance transfer reference device closed loop analog programming no ADC","analog memory cloning circuit memristor copy conductance reference","resistive memory write verify variability contact resistance nonlinear I-V crossbar calibration"],"source_ids":["SRC1","SRC2","SRC3"],"no_result_note":null},"synonyms_and_historical_terms":{"queries":["copying resistor value into memristive device circuit","memristor-to-memristor state copy analog conductance cloning","memristive transfer matrices contact resistance","in-memory mirroring memristor cloning source destination analog"],"source_ids":["SRC2","SRC4"],"no_result_note":"Historical searches found resistance copying, reference-resistor tuning, transfer-matrix analysis, and binary in-memory mirroring, but no opened source demonstrating post-disconnection transfer of a complete multi-point analog response from one qualified compute tile to another."},"products_practices_and_standards":{"queries":["memristor array write-verify analog conductance tuning practice","4K memristor analog-grade passive crossbar tuning device variation","reference resistor array memristor programming patent","RRAM crossbar calibration parasitic resistance nonlinear current voltage"],"source_ids":["SRC1","SRC2","SRC3"],"no_result_note":"Relevant research practice and a patent were found; no directly applicable official standard or first-party product documentation for whole-tile analog-response cloning was located in the bounded search."},"component_combination":{"queries":["same current reference resistor memristor differential amplifier programming pulse","self-terminating analog memristor programming fixed resistor target","memristor crossbar contact resistance transfer matrix","device-to-device variation nonlinear I-V analog crossbar write verify"],"source_ids":["SRC1","SRC2","SRC3","SRC4"],"no_result_note":null}},"sources":[{"source_id":"SRC1","title":"Programming arbitrary analog conductance states of memristors in one step","publisher":"Frontiers Media SA","url":"https://www.frontiersin.org/journals/nanotechnology/articles/10.3389/fnano.2026.1866642/full","source_type":"PRIMARY_RESEARCH","claims_supported":["Memristor conductance programming is complicated by nonlinear ion-migration, filamentary, and thermal mechanisms and by cycle-to-cycle and device-to-device variation.","Conventional high-precision programming predominantly uses write-verify cycles with conversion and control peripherals.","An experimentally tested analog feedback circuit compares a memristor with a fixed reference resistor, converts imbalance into programming voltage, and self-terminates as the target is approached.","The demonstrated loop used fabricated Ta/HfO2 devices and fixed resistors rather than a second qualified compute tile."]},{"source_id":"SRC2","title":"US8416604B2 — Method of implementing memristor-based multilevel memory using reference resistor array","publisher":"United States Patent and Trademark Office record via Google Patents","url":"https://patents.google.com/patent/US8416604B2/en","source_type":"OTHER","claims_supported":["The patent applies identical currents to a reference-resistor node and a selected memristor, compares their voltages, and applies learner-side programming pulses in the polarity that reduces the difference.","Pulse width is proportional to mismatch and comparison repeats until the selected memristor approaches the reference resistance.","The disclosed implementation includes differential amplifiers, sample-and-hold circuits, comparators, pulse-width modulators, and a fixed reference-resistor array rather than a read-isolated qualified memristive tile."]},{"source_id":"SRC3","title":"4K-memristor analog-grade passive crossbar circuit","publisher":"Nature Communications","url":"https://www.nature.com/articles/s41467-021-25455-0","source_type":"PRIMARY_RESEARCH","claims_supported":["A fabricated 64-by-64 passive crossbar required forming and automated write-verify tuning to place devices near target conductances.","Device-to-device switching-threshold variation and half-select disturbance degrade tuning accuracy.","The work demonstrates that programmed target values and realized array conductances differ and that repeated tuning is established practice, but it does not transfer behavior directly from another physical tile."]},{"source_id":"SRC4","title":"Memristive Transfer Matrices","publisher":"arXiv","url":"https://arxiv.org/abs/1004.0041","source_type":"PRIMARY_RESEARCH","claims_supported":["Memristor crossbars can implement multiple-input, multiple-output transfer matrices.","Electrical analysis of those transfer matrices includes parasitic and contact resistances, supporting the premise that realized transfer behavior is not determined solely by ideal crosspoint conductances.","The source analyzes transfer behavior but does not disclose reference-to-learner nonvolatile adaptation."]}],"problem_evidence":{"status":"PARTLY_SUPPORTED","finding":"The underlying problem is visible: device-to-device variation, nonlinear current-voltage behavior, parasitic/contact resistance, half-select disturbance, and programming error can make a physical crossbar's realized response depart from nominal targets. The retained sources do not directly report the narrower replacement scenario—two nominally equivalent qualified tiles producing different bounded transfer matrices—so that extrapolation is plausible but not directly demonstrated.","source_ids":["SRC1","SRC3","SRC4"]},"closest_prior_art":[{"name":"Reference-resistor-array memristor write-in circuit (US8416604B2)","source_ids":["SRC2"],"overlap":"Uses matched physical excitation, analog differential comparison, mismatch-proportional signed programming pulses, repeated learner-only adjustment, and convergence toward a physical reference without first requiring a digitized conductance target.","remaining_difference":"Its authority is a selected fixed-resistor node and its objective is copying a scalar resistance level; it does not use a qualified compute tile's contemporaneous multi-point current-voltage behavior, reproduce a full transfer envelope, or validate against an independently held-out second tile."},{"name":"Self-terminating analog feedback programming against fixed resistors","source_ids":["SRC1"],"overlap":"Directly converts analog imbalance between a fixed reference branch and a writable memristor into state-changing voltage that diminishes near convergence; fabricated-device experiments support the physical-loop concept.","remaining_difference":"The target remains a discrete fixed resistor, the demonstration programs individual conductance levels, and the reported circuit does not test whole-tile response transfer after uncoupling or agreement with a secondary reference."},{"name":"Automated write-verify tuning of an analog-grade 4K crossbar","source_ids":["SRC3"],"overlap":"Addresses device variation and tunes a physical crossbar toward specified analog targets, then exercises vector-matrix multiplication.","remaining_difference":"Targets are represented externally and tuning uses automated write-verify practice; no second physical tile supplies the instantaneous local setpoint."}],"prior_art_disposition":"ADJACENT_PRIOR_ART","contrastive_claim_remaining":"Under matched multi-point voltage and temperature exposure, a read-isolated qualified memristive tile—not a fixed resistor, stored code, or computed model—can serve as the contemporaneous physical setpoint for signed learner-only programming, such that after complete uncoupling the learner retains the declared analog transfer envelope and agrees with an independently held-out reference better than an equal-energy fixed-standard trim or sham coupling.","contrastive_claim_falsifier":"The claim is falsified if equal-energy fixed-standard trimming or sham pairing attains the same retained two-reference envelope; if direct pairing only matches the training tile; if mismatch does not decrease consistently under bounded pulses; if performance fails at untrained operating points or after disconnection; or if the reference measurably changes.","gates":{"adequate_source_search":{"status":"PASS","rationale":"The bounded search covered direct wording, copying and historical terminology, established tuning practices and patent records, and combinations involving differential feedback, transfer matrices, parasitics, and device variation. Exactly four opened sources from four publisher contexts were retained, including three primary-research sources.","source_ids":["SRC1","SRC2","SRC3","SRC4"]},"supported_problem":{"status":"PASS","rationale":"Although the exact spare-tile replacement case was not directly demonstrated, primary evidence supports every important physical contributor: device variation, nonlinear response, programming error, disturbance, and parasitic/contact resistance. This is sufficient for partly supported problem evidence.","source_ids":["SRC1","SRC3","SRC4"]},"distinct_testable_claim":{"status":"PASS","rationale":"The remaining distinction is narrower than the known reference-resistor feedback loop and has observable discriminators: retained multi-point transfer after disconnection, agreement with a held-out tile, and superiority to equal-energy fixed-reference and sham controls.","source_ids":["SRC1","SRC2","SRC3"]},"bounded_next_test":{"status":"PASS","rationale":"A small probe-station study with two read-only references, four sacrificial learners, matched within-learner blocks, fixed energy and safety ceilings, physical disconnection, raw current-voltage endpoints, and one retention interval is bounded and directly tests the remaining contrast.","source_ids":["SRC1","SRC3"]},"no_obvious_safety_or_authority_stop":{"status":"PASS","rationale":"No categorical safety or authority prohibition is apparent for the proposed sacrificial-coupon experiment. Reference write rails are absent, learner energy is clamped, thermal and divergence cutoffs are specified, production hardware is excluded, and final acceptance remains outside the experiment's authority.","source_ids":[]}},"screen_survival":true,"world_novelty_boundary":"This bounded public-web screen establishes neither world novelty nor patentability. It found close prior art for analog mismatch-driven memristor programming against physical resistor references, leaving only the narrower device-to-device, multi-point transfer-envelope, post-uncoupling, and held-out-reference claim for testing; it does not establish market size, expert acceptance, superiority, or realized value."}