Changelog
What's new in each version of the REEL LCI Database
Coverage Expansion
- 477 unit process datasets (up from 363) across 18 component categories
- JEDEC-landscape memory waves: HBM4 stacks (8/12/16-hi), GDDR7 graphics DRAM, DDR5 module variants (CUDIMM, CSODIMM, MRDIMM, CAMM2) with companion ASICs, SOCAMM/SOCAMM2 system-on-module tiers, and UFS 5.0
- 3D NAND extended to 321-layer (three-deck, string-stacked); DRAM 1α/1β/1γ coverage
- Photonics: silicon photonics PIC, single-mode optical fiber (OS2/OM4), 100–400G optical transceivers (QSFP28/QSFP-DD), LC and MPO fiber connectors
- Specialty wafers: GaN-on-Si, SiC, MEMS, CIS, InP laser, GaAs VCSEL
- Data-center infrastructure: Open Compute MGX rack assemblies, busbars, manifolds, PDUs, fan walls, and 180nm BCD power management
Cleanroom HVAC — Recipe-Driven Reframe
- Cleanroom heating and cooling is now modeled as an ISO-6 ballroom plus per-tool minienvironments, with the tool fleet derived from each model's own process recipe rather than a single facility average
- This differentiates facility overhead node-by-node instead of assigning one value per fab class — an advanced-logic wafer and a mature-node wafer no longer inherit identical cleanroom overhead
- Built on the v1.0 fab-overhead rebuild: a fab-type taxonomy anchored to publicly disclosed reference-fab cleanroom areas, a SEMI S23-aligned utility envelope, and explicit make-up-air and central-exhaust energy flows
Gas Abatement — Fleet-Allocation Reframe
- Point-of-use gas-abatement electricity is now a flat per-wafer fleet allocation (sized to tool and pump count, running continuously) rather than a per-etch-pass charge — correcting a structural overstatement for etch-intensive processes
- Abatement fuel combustion is surfaced as a background ecoinvent link, so users can carry the upstream natural-gas supply and combustion through their own impact assessment
Redundancy-Aware NAND Die Yield
- Production NAND tolerates array defects through bad-block / spare-block redundancy and error-correcting codes, so good-die counts now reflect that recovery rather than a naive defect-only model
Upstream Linkage (ecoinvent 3.12)
- Elementary and technosphere flows remapped to ecoinvent 3.12 nomenclature for downstream linking
- Full upstream-linkage coverage retained: every flow carries an ecoinvent link, an internal dataset reference, or an explicit not-in-ecoinvent flag
Air & Water Emissions
- Water-borne emissions are now captured and carried through to every dataset
- Reactive process gases are mapped to their downstream emitted species using documented chemistry rules
- Residual fluorinated-gas emissions differentiate by abatement technology (thermal, plasma, or wet) and its destruction efficiency
- Ultrapure-water demand recalibrated for advanced-node wafers, aligned to disclosed fab water data
Waste Classification
- Plating sludges (Cu/Ni/Sn) are now classified to their correct waste-treatment route
- Copper trim and scrap from leadframes is classified as recoverable copper; hazardous organic waste is routed to incineration with energy recovery
Provenance & Validation
- Every input and output in a dataset traces to a documented source, preserved through cross-model aggregation
- Each dataset flags which of its outputs are products versus waste, so downstream users can allocate impacts correctly
- Per-category process-completeness review across wafers, memory wafers, specialty wafers, PCBs, connectors, and cables
- Each release is validated against published industry benchmarks
Methodology Report
- Methodology Report v1.0 (PDF) — to be published
- Documents the recipe-driven HVAC and fleet-allocation abatement reframes, redundancy-aware NAND yield, the ecoinvent 3.12 migration, and expanded JEDEC memory coverage
Initial Release
- 360+ LCI datasets across 18 component categories
- Complete semiconductor wafer fabrication models (3nm through 180nm)
- Advanced packaging: CoWoS, HBM TSV stacking, fan-out WLP, flip chip
- Passive components: MLCCs, power inductors, ferrite beads
- PCB models: 4-layer through 40-layer, HDI, flex, rigid-flex
- System-level assemblies: server, laptop, SSD, PSU, cooling
Data & Access
- Each dataset built from source-documented process inventories, with uncertainty ranges on inventory values
- Datasets available as structured exports on Circa, with upstream-database (ecoinvent) links
- Methodology report (v0.1) published
- Available for browsing on Circa.ai