"""Generate public-safe wet-clean and surface-prep teaching datasets. The generated files are conceptual, precomputed replays for SemiAgora pages. They intentionally avoid chemical concentrations, process times, tool settings, facility instructions, and qualified process-window claims. """ from __future__ import annotations import json from pathlib import Path ROOT = Path(__file__).resolve().parent SOURCES = [ { "label": "Stanford SNF Pre-Diffusion Clean", "url": "https://snfguide.stanford.edu/taxonomy/term/712/all/feed?order=field_equipment_name&sort=desc", "note": "Public SNF guide page describing SC1, SC2, and optional HF dip roles for organics, particles, metals, and native oxide context.", }, { "label": "Stanford SNF Standard Clean 1", "url": "https://snfguide.stanford.edu/guide/chemicals/acids/standard-clean-1", "note": "Public SNF chemical page anchoring SC1 as a particle and surface-preparation clean before growth furnaces.", }, { "label": "Stanford SNF HMDS", "url": "https://snfguide.stanford.edu/guide/chemicals/primers/hmds", "note": "Public SNF guide page describing HMDS as a photoresist adhesion promoter.", }, { "label": "NIST wet chemical cleaning publication", "url": "https://www.nist.gov/publications/wet-chemical-cleaning-plasma-oxide-grown-heated-001-inp-surfaces", "note": "Public NIST publication page anchoring wet cleaning as surface preparation before epitaxy, metal/dielectric deposition, and diffusion.", }, ] LIMITS = [ "No recipe, concentration, bath timing, tool sequence, facility procedure, or endpoint-control instruction is included.", "No wafer-release, contamination-control, metrology qualification, safety approval, or qualified process-window claim is made.", "All curves are educational teaching proxies and must be replaced by facility-approved measurements before process decisions.", "Material compatibility, pattern sensitivity, queue time, and safety review must be handled by qualified local procedures.", ] def write_json(filename: str, payload: dict) -> None: (ROOT / filename).write_text(json.dumps(payload, indent=2) + "\n", encoding="utf-8") def monotonic_decrease(values: list[float]) -> bool: return all(values[i] >= values[i + 1] for i in range(len(values) - 1)) def monotonic_increase(values: list[float]) -> bool: return all(values[i] <= values[i + 1] for i in range(len(values) - 1)) def common_payload(experiment_id: str, title: str, model_boundary: str) -> dict: return { "schema": "semiagora.process-simulation.v1", "experiment_id": experiment_id, "title": title, "execution_mode": "precomputed-only", "model_boundary": model_boundary, "limitations": LIMITS, "sources": SOURCES, } def wet_removal() -> None: cases = [ { "label": "incoming_surface", "organic_residue_percent": 100.0, "particle_index": 100.0, "metal_contamination_index": 100.0, "native_oxide_proxy_nm": 1.2, "review_note": "Incoming state before any public-safe conceptual clean.", }, { "label": "solvent_degrease_proxy", "organic_residue_percent": 42.0, "particle_index": 88.0, "metal_contamination_index": 96.0, "native_oxide_proxy_nm": 1.2, "review_note": "Organic residue falls, but particle and metal evidence is still needed.", }, { "label": "sc1_particle_clean_proxy", "organic_residue_percent": 18.0, "particle_index": 28.0, "metal_contamination_index": 74.0, "native_oxide_proxy_nm": 1.8, "review_note": "Particle proxy improves while oxide state remains part of the evidence.", }, { "label": "sc2_metal_clean_proxy", "organic_residue_percent": 14.0, "particle_index": 24.0, "metal_contamination_index": 22.0, "native_oxide_proxy_nm": 1.9, "review_note": "Metal contamination proxy improves, but this does not remove native oxide.", }, { "label": "rinse_dry_control_proxy", "organic_residue_percent": 12.0, "particle_index": 20.0, "metal_contamination_index": 20.0, "native_oxide_proxy_nm": 1.9, "review_note": "Final public-safe state still needs watermark, particle, metal, and oxide checks.", }, ] organics = [case["organic_residue_percent"] for case in cases] particles = [case["particle_index"] for case in cases] metals = [case["metal_contamination_index"] for case in cases] payload = common_payload( "SA-PROC-CLEAN-WET-REMOVAL-001", "Wet clean contamination removal", "Educational wet-clean replay using normalized contamination proxies; not a wet-bench recipe or measured wafer result.", ) payload.update( { "axes": ["conceptual clean stage"], "cases": cases, "derived_metrics": { "organic_residue_reduction_percent": round(100 - cases[-1]["organic_residue_percent"], 2), "particle_index_reduction_percent": round(100 - cases[-1]["particle_index"], 2), "metal_index_reduction_percent": round(100 - cases[-1]["metal_contamination_index"], 2), "native_oxide_final_proxy_nm": cases[-1]["native_oxide_proxy_nm"], }, "verification": { "case_count": len(cases), "organic_decreases": monotonic_decrease(organics), "particle_decreases": monotonic_decrease(particles), "metal_decreases": monotonic_decrease(metals), "native_oxide_not_removed_by_sequence": cases[-1]["native_oxide_proxy_nm"] >= cases[0]["native_oxide_proxy_nm"], "all_cases_precomputed": True, }, } ) write_json("process-wet-clean-contamination-removal-web-v1.json", payload) def native_oxide_window() -> None: exposure = [0, 1, 2, 3, 4, 5] oxide_nm = [1.8, 1.15, 0.58, 0.22, 0.08, 0.03] contact_angle = [18, 42, 65, 78, 84, 87] reoxidation = [0.05, 0.18, 0.34, 0.52, 0.68, 0.78] queue = [0.08, 0.15, 0.29, 0.46, 0.64, 0.76] samples = [ { "normalized_exposure_index": index, "native_oxide_remaining_nm": oxide, "contact_angle_proxy_deg": angle, "reoxidation_sensitivity": risk, "queue_time_risk": queue_risk, } for index, oxide, angle, risk, queue_risk in zip(exposure, oxide_nm, contact_angle, reoxidation, queue) ] payload = common_payload( "SA-PROC-CLEAN-DHF-NATIVE-OXIDE-001", "Dilute HF native-oxide window", "Educational native-oxide window replay using normalized exposure; not a chemical process setting, qualified HF procedure, or measured wafer result.", ) payload.update( { "axes": ["normalized exposure index"], "samples": samples, "derived_metrics": { "oxide_after_index_3_nm": oxide_nm[3], "hydrophobic_contact_angle_index_4_deg": contact_angle[4], "reoxidation_sensitivity_index_5": reoxidation[5], }, "verification": { "sample_count": len(samples), "oxide_decreases": monotonic_decrease(oxide_nm), "hydrophobicity_increases": monotonic_increase(contact_angle), "reoxidation_risk_increases": monotonic_increase(reoxidation), "queue_risk_increases": monotonic_increase(queue), }, } ) write_json("process-dilute-hf-native-oxide-window-web-v1.json", payload) def hmds_prime() -> None: cases = [ { "label": "humid_no_prime", "moisture_index": 1.0, "contact_angle_proxy_deg": 32, "adhesion_score": 0.32, "defect_risk": 0.42, }, { "label": "dehydrate_only", "moisture_index": 0.35, "contact_angle_proxy_deg": 52, "adhesion_score": 0.58, "defect_risk": 0.25, }, { "label": "balanced_hmds_prime", "moisture_index": 0.12, "contact_angle_proxy_deg": 76, "adhesion_score": 0.91, "defect_risk": 0.08, }, { "label": "overprime_or_delay_risk", "moisture_index": 0.18, "contact_angle_proxy_deg": 88, "adhesion_score": 0.74, "defect_risk": 0.26, }, ] payload = common_payload( "SA-PROC-CLEAN-HMDS-ADHESION-001", "HMDS adhesion-prime readiness", "Educational HMDS adhesion readiness replay using normalized surface-state proxies; not a lithography track recipe or measured wafer result.", ) payload.update( { "axes": ["surface-state case"], "cases": cases, "derived_metrics": { "best_adhesion_case": "balanced_hmds_prime", "balanced_adhesion_score": 0.91, "overprime_risk_delta": round(cases[-1]["defect_risk"] - cases[2]["defect_risk"], 2), }, "verification": { "case_count": len(cases), "best_adhesion_case_is_balanced": max(cases, key=lambda x: x["adhesion_score"])["label"] == "balanced_hmds_prime", "moisture_lower_after_dehydrate": cases[1]["moisture_index"] < cases[0]["moisture_index"], "hydrophobicity_increases_with_prime": cases[2]["contact_angle_proxy_deg"] > cases[1]["contact_angle_proxy_deg"], "risk_increases_for_overprime": cases[-1]["defect_risk"] > cases[2]["defect_risk"], }, } ) write_json("process-hmds-adhesion-prime-web-v1.json", payload) def clean_sequence_queue() -> None: cases = [ { "downstream_intent": "oxidation_or_diffusion_preclean", "surface_target": "hydrophilic controlled oxide", "evidence_need": ["particle trend", "metal contamination proxy", "native oxide state", "furnace compatibility review"], "risk_flags": ["metal carryover", "particle redeposition", "facility qualification gap"], "sequence_fit_score": 0.82, }, { "downstream_intent": "deposition_preclean", "surface_target": "low residue and known oxide condition", "evidence_need": ["organic residue proxy", "particle trend", "watermark check", "film nucleation note"], "risk_flags": ["residue-driven nucleation shift", "queue-time drift", "material compatibility gap"], "sequence_fit_score": 0.78, }, { "downstream_intent": "lithography_hmds", "surface_target": "dry hydrophobic adhesion-ready surface", "evidence_need": ["dehydration state", "contact-angle proxy", "adhesion score", "delay-to-coat note"], "risk_flags": ["moisture return", "overprime risk", "resist compatibility gap"], "sequence_fit_score": 0.88, }, { "downstream_intent": "contact_or_surface_sensitive", "surface_target": "controlled native oxide or hydrogen-terminated state", "evidence_need": ["native oxide proxy", "contact resistance proxy", "queue-time risk", "metrology boundary"], "risk_flags": ["reoxidation", "surface damage", "unqualified contact claim"], "sequence_fit_score": 0.73, }, ] payload = common_payload( "SA-PROC-CLEAN-SEQUENCE-QUEUE-001", "Clean sequence decision queue", "Educational clean-sequence queue map for evidence planning; not a wet-bench flow, tool route, or facility-approved traveler.", ) payload.update( { "axes": ["downstream intent"], "cases": cases, "derived_metrics": { "case_count": len(cases), "total_evidence_needs": sum(len(case["evidence_need"]) for case in cases), "average_sequence_fit_score": round(sum(case["sequence_fit_score"] for case in cases) / len(cases), 3), }, "verification": { "case_count": len(cases), "all_cases_have_evidence_need": all(len(case["evidence_need"]) >= 4 for case in cases), "all_cases_have_risk_flags": all(len(case["risk_flags"]) >= 3 for case in cases), "scores_are_public_planning_proxies": True, }, } ) write_json("process-clean-sequence-queue-web-v1.json", payload) def main() -> None: wet_removal() native_oxide_window() hmds_prime() clean_sequence_queue() if __name__ == "__main__": main()