active-learning

What problem does it solve?

Orchestrates closed-loop autonomous experimental optimization to accelerate materials synthesis and process tuning, removing manual trial-and-error and bridging experiments with computational models. It enables rapid discovery of synthesis conditions and process parameters by coordinating campaign setup, experiment suggestion, automated characterization, and iterative model updates.

Core Features & Use Cases

• ARROWS campaign orchestration: Initialize thermodynamically-guided synthesis campaigns, enumerate balanced precursor sets, rank by driving force, and suggest discrete experiments for solid-state synthesis targets.
• Automated XRD integration: Analyze XRD patterns with CNN-based phase ID and Rietveld refinement and feed phase and weight-fraction outputs directly into the campaign recorder.
• Bayesian optimization loops: Define continuous, discrete, and categorical parameter spaces, collect observations, build GP surrogates, and suggest experiments using EI/UCBe/PI acquisition strategies.
• Result recording & knowledge extraction: Record experimental outcomes, learn pairwise reaction rules, update campaign state files, and export learned reactions and summaries.
• Hybrid strategies & workflows: Combine ARROWS and BO (ARROWS→BO, BO→ARROWS, or parallel) to exploit thermodynamic guidance and surrogate-based fine-tuning in complex problems.
• Use cases: Synthesizing known oxide phases with XRD validation, optimizing thin-film deposition parameters for conductivity, and screening precursor combinations for perovskites.

Quick Start

Start an ARROWS campaign to synthesize BaTiO3 from BaCO3 and TiO2 across 600–900°C and request the first experiment suggestion.