Illustrative · unadmitted · revision 1
Adaptive recurrent mixed-halide perovskite operator
This is a proposed definition for criticism. It is not an experiment run, result, evidence claim, canonical admission, inventory record, or procurement request.
- Stable identifier
adaptive-recurrent-mixed-halide-perovskite- Kind
- hybrid
- Intent
- characterization
Exact target
- external-reference ·
adaptive-recurrent-mixed-halide-perovskite — operator-supplied adaptive recurrent mixed-halide perovskite proposal
Minimal decisive protocol
Closed-loop adaptive operator test
Apply a fixed sequence of optical inputs to a mixed-halide perovskite loop while an explicit controller adapts one declared parameter, then distinguish recurrent state adaptation from photochemical drift.
- Measure baseline response and halide-state proxy.
- Run fixed-input feedback blocks with controller logs.
- Repeat with adaptation disabled and randomized input order.
- Fit a bounded state-space model to raw traces.
Decision rule: An adaptive recurrent operator survives only if held-out sequence prediction and state reset tests distinguish it from drift and memoryless controls.
Controls
- ablated: Disable the adaptive controller. (controller update; Adaptive improvement should disappear.)
- matched: Reset optical and thermal state between blocks. (state history; Persistent history dependence must follow the declared state variable.)
Observables and units
Observables| Observable | Units | Measurement | Uncertainty reporting |
|---|
| held-out sequence prediction error | normalized | state-space prediction of output trace | block bootstrap across independent sequences |
|---|
| halide-state proxy | spectral nm | spectral response and imaging proxy | calibration-aware interval |
|---|
Calibration and repetitions
- spectrometer wavelength: reference lamp and fixed spectral lines — wavelength residual below declared threshold (nm)
Replicate unit: independent feedback sequence; minimum 24, independent 6.
Randomization: sequence and adaptation schedule randomized
Stopping rule: complete held-out blocks
Uncertainty: photochemical drift, thermal drift, spectral calibration, controller latency; hierarchical block model with reset and adaptation controls; retain controller logs and raw spectra
Success and falsifiers
Success
- The adaptive loop improves held-out sequence prediction beyond disabled and reset controls. (prediction-error reduction >= predeclared margin fraction)
Falsifiers
- A drift-only model explains the same sequence and state dependence without adaptive updates. (model residual difference <= predeclared margin normalized residual)
Confounds and raw artifacts
- thermal history is mistaken for adaptive state: thermal sensor and reset blocks; mitigation: include temperature as a measured nuisance variable
- closed-loop-trace · CSV and spectral stack · retain raw sensor, controller, and input streams
Nonclaims
- No recurrent computing result is claimed.
- The fixture does not claim perovskite stability, adaptation mechanism, or useful scale.
Provenance boundary
The citation supplies mixed-halide material context only; it does not support adaptive recurrence or computation.
Agent-readable JSON · History · Provenance · Back to experiments