aimdesign

Lab OBF Material Simulators

This directory contains skeleton Python modules for two simulators designed to evaluate new materials under the Open Bench Format (OBF) framework. These simulators provide a starting point for building decision‑quality and physics‑based analyses of candidate materials. They are meant to live under the lab.aimdesign.ai environment, but can be run locally for development and testing.

Simulators

1. Decision Quality Simulator (decision_quality_simulator.py)

This simulator models the campaign‑level process of evaluating a new material. It focuses on the value of information provided by OBF evidence versus ad‑hoc evidence. The high‑level loop is:

1. Generate candidate materials with intrinsic properties and metadata completeness. The current implementation uses simple random distributions; real implementations should draw from known distributions or sampling plans.
2. Propagate candidates through process stages. Each stage (e.g., synthesis, casting/B‑stage, storage/handling, lamination/cure, via formation, plating/adhesion, pattern/yield, reliability) can introduce variability and failure modes. The simulator applies simple transformations to candidate properties at each stage.
3. Evaluate evidence quality. Two evidence modes are modeled: ad‑hoc (with missing metadata and bias sources) and OBF (with lower missingness and reference runs for normalization).
4. Make decisions based on simulated outcomes, tracking metrics such as time‑to‑decision, false positives/negatives and value of information per metadata field.

The simulator returns both raw run data and summary statistics so you can assess the benefit of OBF adoption and identify which metadata fields contribute most to decision quality.

2. OTV Physics Simulator (otv_physics_simulator.py)

This simulator models the screening‑level physics of candidate materials for the OBF Tier 0–2 gates. It contains placeholder implementations for the following modules:

• M1 – Dielectric constant/loss (m1_dielectric_loss): models electromagnetic response of the material (e.g., effective permittivity, loss tangent) given frequency, conductor roughness and copper state.
• Moisture uptake (m2_moisture_uptake): models diffusion and solubility, optionally with temperature/humidity dependence.
• M4 – Adhesion retention (m4_adhesion_retention): models bond strength and degradation under environmental cycling.
• M2 – Via formation/yield (m2_via_chain_yield): models the compatibility of the material with via formation, desmear and plating, producing yield and resistance distributions.
• M3 – CAF/SIR (m3_caf_sir): models electrochemical hazard (ion content, moisture, bias) and predicts surface insulation resistance failures.

Each module returns simulated outputs, which are passed through a gate logic replicating OBF Tier 0–2 outcome definitions. The simulator aggregates module outputs into a data structure matching the OBF dataset template, so you can validate synthetic runs with your existing analysis scripts.

Next Steps

These modules provide interfaces and documentation rather than complete models. To make the simulators realistic you will need to:

• Replace placeholder probability distributions with distributions informed by your experimental data.
• Implement physics‑based models or surrogate models calibrated to experiments (e.g., using scipy, pint and material libraries).
• Extend the gating logic to match the latest OBF specification.
• Build a simple CLI or web API for running parameter sweeps and visualizing results.

Contributions are welcome! Use these skeletons as a foundation for your lab’s simulators and adapt them to the specific needs of your materials and processes.