01 Executive summary
A small set of substance classes accounts for nearly all of the data-quality risk on the non-energy material side.
Most public attention on the semiconductor industry's environmental footprint focuses on two well-characterized areas: grid electricity consumed at the fab, and direct atmospheric emissions of fluorinated process gases used in etch and chamber cleaning. Both are tracked rigorously by industry working groups, foundries, and consortia. Grid electricity is the dominant single contributor in every fab model; the direct-gas slice, once abatement is applied, is a far smaller share of cradle-to-gate GWP than its public profile suggests. Both have well-developed mitigation pathways: renewable power purchase agreements for the first, abatement equipment with gas-dependent destruction efficiencies of 95–99%+ for the second.
This paper is about everything else.
Across the REEL LCI v1.0 database, covering 14 wafer technology nodes from 3 nm to 180 nm and 42 packaging architectures from wirebond QFN to HBM3 2.5D stacking, the burden of materials and chemicals (excluding electricity and direct gas emissions) accounts for roughly 14–17% of cradle-to-gate GWP at the advanced logic nodes and 19–23% at the mature ones. In both cases that is most of everything which is not grid electricity, because the direct fluorinated-gas slice has become very small. In packaging the absolute material burden per part is far smaller but the proportional share is much more variable, from about 5% of a fan-out wafer-level package to over 99% of a gold-wirebonded plastic leaded chip carrier. This is the slice of the inventory where public emission-factor data is weakest, where industry transparency lags furthest behind, and where downstream LCAs are most forced to rely on proxies, secondary sources, and modeled estimates.
This paper maps that landscape. It identifies which materials and chemicals show up as hotspots in wafer fabrication and advanced packaging, what publicly available emission-factor data exists for each today, and where the data quality is weakest. It draws on the REEL LCI v1.0 screening data, the available peer-reviewed literature, ecoinvent v3.10 and v3.11 release notes, vendor disclosures, and recent academic theses. It is a landscape report rather than a call for contributions. Its readers, primarily LCA and sustainability teams at semiconductor materials, chemicals, gas, and equipment suppliers, can self-identify where their organization is positioned to move the public data forward.
The hotspot rankings presented here are based on industry-typical REEL models. Real fabs vary substantially in equipment, recipes, abatement, sourcing, and utility mix; a specific fab's hotspot ordering and the relative weight of each emission factor may differ from the screening below. See Section 8 for the full caveat on within- and between-fab variation.
The rest of this paper details where these gaps sit and what is known about each.
02 Reading guide for supplier LCA teams
Find your portfolio. Each row points to the relevant archetype and tier.
If you are an LCA, sustainability, or product-stewardship professional at a supplier company, the entry point most relevant to your portfolio is below.
Each archetype block in Section 5 names what publicly available data would have most improved REEL's coverage. Whether and how a supplier engages is their call; the landscape map above is what we can offer.
03 Scope and method
What this paper covers, what it excludes, and how "best available" is judged.
The REEL LCI database v1.0 includes 14 wafer technology nodes spanning advanced FinFET (3 nm) through mature CMOS (180 nm), and 42 packaging architectures across wirebond, flip-chip, fan-out, wafer-level, and advanced 2.5D / 3D stacking. Each model is built from publicly available data (peer-reviewed papers, vendor specifications, patents, government reports, and industry roadmaps) and produces inventory data (mass and energy flows), not characterized impact results.
For the analysis behind this paper, REEL's v1.0 models were screened using AR6 GWP100 characterization factors. The screening identified the largest contributors to cradle-to-gate GWP for each model. This paper focuses on a specific subset of those contributors: materials and chemicals, excluding two categories that are well-characterized elsewhere.
Materials and chemicals: in scope
- Bulk gases supplied to the fab (industrial N₂, H₂, O₂, He, Ar): the gas itself, as distinct from its emission.
- Specialty process gases (etchants, dopants): the upstream production burden, distinct from any direct atmospheric emission of the same molecule.
- Wet-clean chemicals at electronics grade (NH₄OH, HCl, H₂SO₄, HF, H₂O₂).
- Photoresists and ancillaries (resists, developers, BARC, edge-bead removers).
- CMP slurries and pad consumables.
- Plating chemistries: base salts (KAu(CN)₂, CuSO₄) and the proprietary additive packages.
- Bulk and electrodeposited metals at electronics or assembly grade (Au, Cu, Ag, Cu-Mo).
- Substrate organics and films (EMC, polyimide, ABF, dry film resist, BCB).
- Single-crystal silicon and other wafer substrates.
Out of scope
Grid electricity at the fab. The dominant single contributor to advanced logic wafer GWP. Reported by foundries, regulated under various jurisdictional carbon accounting rules, and addressed by an active renewables-procurement effort. Its emission factor is determined by grid mix, itself extensively characterized.
Direct atmospheric emissions of process gases. Fluorinated gases (SF₆, NF₃, C₄F₈, C₄F₆, CF₄, C₂F₆) and other high-GWP species that escape abatement. The industry tracks these through the IPCC 2019 Refinement Vol. 3 Ch. 6 methodology and WSC PFC reduction agreements. Abatement systems achieve destruction efficiencies of 95–99% or higher, depending on the gas and abatement technology.
Excluding these two categories isolates the slice this paper is about: the production-phase emission factors of the materials and chemicals consumed in fabrication and assembly.
One further exclusion: on-site utilities with no external supplier. Ultrapure water (UPW) is produced at the fab from incoming water, not procured from a supplier and not meaningfully transportable. Its production burden is on-site electricity (reverse osmosis, electro-deionization, UV, and pumping), already inside the fab-electricity slice above and metered directly by fabs. With no external supplier to characterize, UPW sits outside this paper's supplier-facing scope; the per-m³ figure a bottom-up modeler wants is a fab-disclosure question, not a supplier-data gap.
What "best available" means
A strict source hierarchy. Peer-reviewed publications first, then vendor EPDs, then industry-consortium reports (SEMI, JEITA, IPC, EU Horizon), then corporate sustainability disclosures with quantitative inventory data, then government LCI databases (USLCI, ELCD), then the ecoinvent v3.10 / 3.11 datasets (licensed, and used in REEL only for flow mapping, never as a substitute for a verified primary source).
Other licensed databases (Sphera/GaBi, IDEMAT-paid) are noted only where their existence is methodologically relevant; their values are never cited.
A note on the word "public." In this paper a "public" emission factor is one that can be derived from publicly accessible sources: open literature, vendor disclosures, government datasets, or a database any practitioner can obtain without an institutional license. "Public" here means accessible, not free or open-source. REEL-modeled values cited below are published in the REEL LCI Database, which any practitioner can procure on an affordable, pay-as-you-go basis: accessible without the institutional licensing typical of proprietary LCA databases.
Provenance labels used throughout this paper
Open-access journal article or thesis with a verifiable DOI containing a cradle-to-gate LCI.
Vendor EPD, consortium-published LCI, or government dataset built from semiconductor-industry primary data.
Standard background database entry (ecoinvent or comparable) at industrial-bulk grade; not semi-specific.
A Generic public entry plus REEL's documented industrial-to-electronics-grade purification multiplier from the Higgs/Boyd baseline.
Modeled by REEL from precursor chemistry, mass balance, and engineering estimates because no peer-reviewed or vendor source exists publicly. Published in the REEL LCI Database v1.0 (to be published).
Substance not yet modeled in any public source within the search scope of Section 8.
For the construction of REEL LCI v1.0 inventories (DQI scoring, uncertainty propagation, mass-balance closure, and elementary flow nomenclature) readers are directed to the REEL LCI Methodology Report v1.0 (to be published), which accompanies this paper.
04 The 80 / 20 picture
Materials and chemicals are the remaining 14–23%, and exactly the slice with the weakest public coverage.
Across REEL LCI v1.0's GWP screening, three categories account for the whole of every model's cradle-to-gate GWP: grid electricity at the fab or assembly site, direct atmospheric emissions of fluorinated process gases, and the production-phase burden of all other materials and chemicals. The first leads in every fab model; in many assembled packages, materials and chemicals lead instead. The second, after abatement, is the smallest of the three at almost every model screened, and it is absent entirely from assembly-only packages.
For logic wafers, grid electricity dominates: roughly three-quarters to 85% of cradle-to-gate GWP across the 14 nodes screened. That proportional band is remarkably stable from 3 nm to 180 nm even though the absolute electricity burden falls by about a factor of five over the same span, so the mature-node share sits slightly below, not above, the advanced-node share. Direct fluorinated-gas emissions, after abatement, range from 0.4% of cradle-to-gate GWP at 3 nm and 5 nm to 7.6% at 7 nm, and settle at roughly 2–4% across the mature nodes. Materials and chemicals, the focus of this paper, account for the rest: 14–17% at 3 nm through 14 nm and 19–23% at the mature nodes, which is 65–97% of everything that is not grid electricity (Figure 1).
For packaging architectures the absolute material burden is much smaller per part, and the proportional share is far more variable: from about 5% of a fan-out wafer-level package to over 99% of a plastic leaded chip carrier. Wirebond packages sit at the top of that range, and the gold bonding wire alone is 69–93% of total package GWP in the TQFP, TSOP, SSOP, and TSSOP exports. In the FCBGA family the substrate laminate and its copper lead the non-energy burden, with the copper plating bath a smaller 7–22% of it. In wafer-level and fan-out packages the copper plating solution is generally the single largest non-energy contributor, at roughly 43–85% of non-energy GWP, while TMAH developer stays below 4% in every wafer-level export screened.
Figure 1. Five representative products from REEL LCI v1.0's GWP screening, normalized to share of cradle-to-gate GWP. The direct fluorinated-gas segment is a sliver at the leading edge and does not appear at all in the assembly-only package rows, which leaves materials and chemicals as effectively the whole of the non-electricity burden. That slice (the rightmost segment) is exactly where the public emission-factor base is weakest. The HBM3 row is the 8-high memory stack itself (core dies, base logic die, and TSV stacking and assembly), not a full 2.5D module. Segment labels are rounded to whole percent; product-level values are in the v1.0 dataset release.
Why this matters for emission-factor work
The non-energy material slice is exactly the slice with the weakest public emission-factor coverage. This is the case for two independent reasons.
Proprietary opacity. Many of the most chemically specialized substances (specialty process gases, CMP slurries, photoresists, plating additives) are produced by a small number of vendors whose formulations are trade secrets. Production-phase emission factors are rarely published because doing so would expose proprietary process information.
Missing purification energy. Many substances that are in public databases are present only at industrial bulk grade. The transformation from industrial to electronics grade (purification to parts-per-trillion impurity tolerances) carries a substantial additional energy burden that has not been quantified publicly in over a decade.
These two failure modes define the landscape that the rest of this paper maps.
05 Hotspots by supplier archetype
Seven supplier archetypes that account for the majority of the non-energy material burden.
For each archetype the structure is the same: where its products appear as hotspots, what the chemistry or function is, what the current best public emission factor is, and what is known to be approximate or generic. The order is approximately by leverage: the higher up an archetype is, the more REEL's downstream wafer and package totals would shift if its emission factors improved.
§ 5.1
Specialty process gases: etch, clean, and deposition
SubstancesNF₃, SF₆, CF₄, C₂F₆, C₄F₈, C₄F₆, GeH₄, F₂, SiH₄
- Where they appear
- Every advanced-node wafer in REEL's screening. NF₃ production in particular is a top-five non-energy material contributor at 3 nm, 5 nm, and 7 nm, and stays inside the top eight at every node down to 180 nm. SF₆ production ranks lower on the wafer side, but it leads the non-energy material burden in HBM TSV stacking, and CoWoS 2.5D shows an anomalously high SF₆-production share (about 14% of its non-energy GWP) owing to fluorinated chemistry in the silicon interposer fab.
- Why they matter
- These gases enable the high-aspect-ratio etching, selective film removal, and chamber cleaning that define advanced logic and DRAM manufacturing. Direct atmospheric escape is heavily regulated and abated; the production-phase burden is not. Synthesis typically involves multi-step fluorination chemistry with substantial energy inputs for distillation and purification to the 5N (99.999%) to 6N (99.9999%) purity grades required by semiconductor users.
- Best public EF
- REEL-modeled, derived from open-literature sources on production chemistry, vendor energy-intensity claims, and engineering estimates. No vendor EPDs were located for any of the major specialty gas producers (Air Liquide, Linde Electronics, Resonac, Kanto Denka, SK Materials, Merck Electronics) covering the production phase of these gases at electronics grade. Thompson (2020), a University of Idaho master's thesis, models NF₃ at a single Idaho-fab scope and is the only academic source identified with a public LCI for an electronics-grade specialty gas.
- Data-quality flag
-
REEL-modeled Single-supplier reliance for the NF₃ academic baseline; modeled approaches for the others.
Vendor "low-GWP" alternative gases (C₄F₈-linear, C₃F₄, C₄F₇N) are being developed but their production-phase emission factors are also not publicly disclosed.
§ 5.2
CMP slurries
SubstancesOxide CMP slurry (silica- or ceria-based), copper CMP slurry, tungsten CMP slurry, plus CMP pads and conditioner disks.
- Where they appear
- Every wafer node with CMP; that is, every wafer node in REEL's screening. Also prominent in HBM TSV stacking and in advanced packaging interposers.
- Why they matter
- CMP is the workhorse planarization step in modern fabs. Slurries are complex aqueous suspensions of nanoscale abrasives plus proprietary additive packages (oxidizers, chelating agents, surfactants, pH modifiers). Per-wafer slurry consumption ranges from 50 to 400 mL per CMP step, and a modern advanced-node wafer passes through 20–50 CMP steps. The production-phase emission factor of the slurry (driven by abrasive calcination, hydrothermal synthesis, and ball milling) has not been quantified in any public database.
- Best public EF
- REEL-modeled from precursor chemistry. Venkataswamy et al. (2024, ACS Sustainable Chemistry & Engineering) provides the most current qualitative review of CMP consumable manufacturing impacts, identifying slurry and pad production as particularly energy- and water-intensive, but it is a perspectives paper that does not produce a synthesized per-kg-CO₂e emission factor for any specific formulation. No supplier EPDs were located from Merck Electronics, CMC Materials (Entegris), Fujimi, or Resonac.
- Data-quality flag
-
REEL-modeled Venkataswamy et al. (2024) is a useful qualitative reference for defending the proxy approach but is not itself a drop-in emission factor.
§ 5.3
Photoresists and ancillaries
SubstancesArF immersion photoresist, EUV photoresist (organic chemically amplified and emerging metal-oxide), bottom anti-reflective coating (BARC), edge-bead remover (EBR), tetramethylammonium hydroxide (TMAH) developer.
- Where they appear
- Every patterning step at every node. Relevance scales with the mask count of the technology: an advanced logic node may use 60–80 photomask layers, each requiring resist coat, exposure, develop, and strip.
- Why they matter
- Photoresist formulations are among the most tightly guarded trade secrets in the materials supply chain. The supplier set is concentrated (JSR, Tokyo Ohka Kogyo, Shin-Etsu Chemical, Sumitomo Chemical, Fujifilm Electronic Materials, DuPont Electronics & Imaging). Quantitative cradle-to-gate emission factors are not published for any commercial resist product. Several suppliers acknowledge using LCA internally for product development but do not release the underlying inventory data.
- Best public EF
-
Photoresists (ArF, EUV, BARC, EBR). REEL-modeled or proxied. No public LCI was located within search scope.
TMAH developer. Fuentes et al. (2023, ACS Sustainable Chemistry & Engineering) provides a peer-reviewed cradle-to-gate LCI for TMAH synthesis. Although the paper's primary subject is magnetite-nanoparticle production, the authors modeled their own TMAH product system in OpenLCA 1.11 against ecoinvent 3.5 background data, building it from trimethylamine plus methyl chloride to tetramethylammonium chloride, then reacted with potassium hydroxide. Inventory data is in the open Supporting Information.
- Data-quality flag
-
Near-complete void Photoresists. TMAH covered by a single modeled inventory at ecoinvent 3.5 vintage.
§ 5.4
Plating chemistry: substrate and bumping
SubstancesPotassium gold cyanide (KAu(CN)₂) for gold electroplating; copper sulfate baths plus organic additive packages (suppressors, accelerators, levelers) for damascene and bump plating; ENIG and ENEPIG chemistries for substrate finishing.
- Where they appear
- Copper plating baths run across the whole flip-chip, wafer-level, and fan-out set: 7–22% of non-energy GWP in the FCBGA family, and the single largest non-energy contributor in most wafer-level and fan-out exports at 43–85%. Gold-flash plating is narrower but heavier where it lands. KAu(CN)₂ reaches about 34% of non-energy GWP in the 2.5D silicon-interposer export and ranges from a few percent to about 47% across the ENIG-finished board and substrate exports, the high end being the fine-line and flexible boards. It does not appear in the FCBGA family at all. ENIG is the standard finish for most modern PCBs and many semiconductor substrates.
- Why they matter
- Base metal salts (KAu(CN)₂, CuSO₄) appear in standard public databases, but the proprietary additive packages (accelerators, suppressors, levelers) that determine deposition kinetics and final film properties are not characterized in any open dataset. Drag-out losses, additive degradation, and bath disposal further complicate the cradle-to-gate picture.
- Best public EF
-
Base salts. Generic public (ecoinvent v3.10 / v3.11); adequate for the salt itself but not for the proprietary additive blend.
ENIG / ENEPIG. Wagih, Bainbridge et al. (2024, IEEE Journal of Microwaves) provides a peer-reviewed LCA of ENIG-finished microwave PCBs with explicit modeling of gold and nickel mass as a function of trace coverage. Finds gold has the highest impact per unit mass owing to mining and refining; reducing gold-plated area is the simplest decarbonization lever.
Additive packages. No public LCI identified. The EU GENESIS project (Horizon Europe 101194246, May 2025 – April 2028, 58 partners, CEA coordinator) includes plating-chemistry LCAs in scope; data not yet public.
- Data-quality flag
-
Bounded open question A reasonable expectation is that GENESIS will produce open public LCIs in the 2027–2028 window. Until then, REEL models additives as proxy fractions of bath chemistry.
§ 5.5
Bonding gold and substrate metallization
SubstancesBonding wire gold (4N–6N purity, drawn to 15–25 μm diameter); electronics-grade copper for substrate metallization; copper-molybdenum alloys for high-reliability and power packaging; silver die-attach paste (sintered nano-silver and conventional); leadframe copper.
- Where they appear
- Bonding wire gold is the single largest non-energy contributor in the gold-wirebond packages, accounting for 69–93% of total package GWP in TQFP, TSOP, SSOP, and TSSOP exports in REEL's screening, and rising to 94–99% in the gold-wirebonded DIP and PLCC exports (the copper-wire QFN sizes carry almost none of it). Substrate copper appears across the entire FCBGA family. Silver die-attach paste appears in QFN and power packages.
- Why they matter
- The substances themselves are in standard public databases (ecoinvent has gold, copper, silver, molybdenum at generic industrial grades), but the energy delta to electronics-grade purity and to the specific metallurgical processes used in semiconductor assembly (wire drawing for bonding gold, electrodeposition for substrate copper, sintering for nano-silver) has not been quantified publicly in over a decade. Boyd (2012, Life-Cycle Assessment of Semiconductors) summarizes earlier work by Higgs (Intel) showing that purification from industrial to semiconductor grade can carry a 20% to 1000% additional carbon footprint depending on the substance. That figure is itself vintage and has not been updated in any open public source.
- Best public EF
-
Bonding gold. Generic public (ecoinvent gold mining and refining), with an internal REEL multiplier for the wire-drawing and purification delta. Wang (2024, KTH master's thesis) examines the carbon footprint of semiconductor products with explicit attention to gold as a backend hotspot but does not propose an updated production-phase emission factor.
Cu-Mo substrate alloys. Generic public; International Copper Association (ICA, 2022) and International Molybdenum Association (IMOA, 2018) provide mass-allocated bulk metal LCIs.
Silver die-attach paste. Generic public (ecoinvent silver) with REEL modeling for paste formulation. No public LCI located for sintered nano-silver formulations.
- Data-quality flag
-
Vintage delta Industrial-to-semi-grade purification penalty unrefreshed since ~2012.
For high-volume wirebond products, even a modest correction to bonding-gold EF would shift package-level totals materially.
§ 5.6
Bulk gases and utilities
SubstancesN₂, O₂, H₂, He, Ar; compressed dry air (CDA).
- Where they appear
- Every export. N₂ alone is a top-five non-energy contributor in 44 of the 56 wafer-and-packaging exports screened.
- Why they matter
- Per-unit emission factors are small but ubiquity creates large aggregate exposure. The methodologically relevant distinction is on-site fab generation (cryogenic air separation built into the fab utility plant) versus merchant supply (trucked-in liquid); energy intensities differ substantially between the two, and neither is characterized at semiconductor-grade specificity in public datasets.
- Best public EF
-
Bulk industrial gases. Generic public (ecoinvent v3.10 cryogenic air separation datasets; v3.11 added improved gallium datasets but did not target bulk gases).
- Data-quality flag
-
Broad exposure Modest per-unit factor. Leverage is in aggregate effect across all products.
§ 5.7
Substrate organics and films
SubstancesAjinomoto Build-up Film (ABF); polyimide (commonly PMDA/ODA-based); benzocyclobutene (BCB); epoxy mold compound (EMC); dry film resist.
- Where they appear
- ABF in every modern flip-chip CPU and GPU substrate. Polyimide in fan-out wafer-level packaging and wafer-level chip-scale packages. EMC in every encapsulated package.
- Why they matter
- These are highly engineered composite materials whose generic-material proxies in standard databases substantially misrepresent the true cradle-to-gate footprint. Epoxy mold compound is typically more than 80% silica filler by weight; using a generic epoxy emission factor overestimates the petrochemical burden while underestimating the mineral-processing burden. Polyimide synthesis from PMDA and ODA precursors is well-documented in chemistry literature but not in public LCI databases. ABF in particular is a near-monopoly product from Ajinomoto Fine-Techno with no public product-level disclosure.
- Best public EF
-
ABF. REEL-modeled / proxied as silica-filled epoxy plus PET release film plus protective overfilm. Ajinomoto Group publishes corporate-level sustainability disclosures but no product-level EPD or cradle-to-gate LCI for ABF. No peer-reviewed open-source LCA covers it.
Polyimide. Generic public (ecoinvent polyimide); adequate for chemistry but not for semiconductor-grade purity or thin-film processing. Zhang, Bainbridge et al. (2024, Scientific Reports) provides a comparative LCA of FR4, PET, paper, and a degradable substrate but does not cover polyimide.
EMC. Generic public (ecoinvent epoxy resin) plus REEL modeling for the silica filler fraction.
BCB and dry film resist. REEL-modeled / generic-public proxies. Major suppliers (Asahi Kasei, Toray) acknowledge running internal LCAs but do not publish.
- Data-quality flag
-
Single-supplier hotspot ABF: a textbook case of no public data. Polyimide is a confirmed gap.
06 The emission-factor landscape
A single table of best-available public emission factors, by substance.
The following table presents the best-available public emission-factor source for each material and chemical hotspot identified in REEL LCI v1.0's GWP screening. Substances are grouped by supplier archetype as in Section 5, plus two additional groups (wet-clean chemicals and the silicon substrate, introduced in Section 3.3) that fall outside the seven archetypes.
Before the per-substance detail, the matrix below is the at-a-glance read: each substance class placed against the best public provenance available for it today. Read it left to right; the further right a class sits, the thinner the public foundation.
reviewed
specific
public
Boyd
modeled
LCI
Figure 2. Each substance class placed at its best-available public provenance; dot color encodes leverage tier. The Industry-specific public column (tinted) is empty by design — no vendor EPD, consortium LCI, or government dataset built from semiconductor-industry primary data was located for any class. The detailed per-substance table follows.
| Substance | Best-available public emission factor | Provenance | Vintage | Gap statement |
|---|---|---|---|---|
| Specialty process gases | ||||
| NF₃ production | Thompson (2020), Univ. of Idaho M.S. thesis | Peer-reviewed (open) | 2020 | Only academic source publicly available; single-fab scope; not vendor-validated. |
| C₄F₈ production | REEL LCI Database v1.0 (modeled) | REEL-modeled (published) | 2026 | Vendor EPDs absent; modeling required. |
| C₄F₆ production | REEL LCI Database v1.0 (modeled) | REEL-modeled (published) | 2026 | Same as C₄F₈. |
| GeH₄ production | REEL LCI Database v1.0 (modeled) | REEL-modeled (published) | 2026 | Specialty epi gas; no merchant LCI. |
| F₂ production | REEL LCI Database v1.0 (modeled) | REEL-modeled (published) | 2026 | On-site generated; no merchant EF. |
| SF₆ production (upstream) | REEL LCI Database v1.0 (modeled) for electronics grade | REEL-modeled (published) | 2026 | Despite SF₆'s dominance, no vendor EPD covers semi-grade production. |
| SiH₄ production | ecoinvent v3.10 | Generic public | 2023 (v3.10) | Bulk-chemical entry; lacks high-purity processing delta. |
| CMP slurries | ||||
| Oxide CMP slurry silica / ceria |
Venkataswamy et al. (2024, ACS Sus. Chem. Eng.); qualitative review | Peer-reviewed (open) | 2024 | No quantified per-kg EF; perspectives paper identifying slurry and pad production as energy- and water-intensive. |
| Copper CMP slurry | REEL LCI Database v1.0 (modeled) | REEL-modeled (published) | 2026 | Li et al. (2025) covers mechanics, not LCA. |
| Tungsten CMP slurry | REEL LCI Database v1.0 (modeled) | REEL-modeled (published) | 2026 | No quantified public EF; modeled from precursor chemistry. |
| Photoresists and ancillaries | ||||
| ArF immersion resist | REEL LCI Database v1.0 (modeled) | REEL-modeled (published) | 2026 | All major suppliers acknowledge internal LCA; nothing published. |
| EUV resist organic + metal-oxide |
REEL LCI Database v1.0 (modeled) | REEL-modeled (published) | 2026 | Emerging chemistry; no baseline. |
| BARC / EBR | REEL LCI Database v1.0 (modeled) | REEL-modeled (published) | 2026 | - |
| TMAH developer | Fuentes et al. (2023, ACS Sus. Chem. Eng.) | Peer-reviewed (open) | 2023 · ecoinvent 3.5 bg | Single-paper modeled inventory; not vendor-validated. |
| Plating chemistry | ||||
| KAu(CN)₂ potassium gold cyanide |
ecoinvent KAu(CN)₂ | Generic public | 2023 (v3.10) | Additive packages not in the dataset. |
| Cu plating bath base | ecoinvent CuSO₄ | Generic public | 2023 (v3.10) | Same; additives proprietary. |
| ENIG / ENEPIG | Wagih, Bainbridge et al. (2024, IEEE J. Microwaves) | Peer-reviewed (open) | 2024 | First open LCA with explicit gold + nickel mass model; microwave-PCB scope. |
| Additive packages accelerator, suppressor, leveler |
REEL LCI Database v1.0 (modeled) | REEL-modeled (published) | 2026 | EU GENESIS project (CORDIS 101194246) covers this; data expected 2028. |
| Bonding metals and substrate metallization | ||||
| Bonding wire Au semi-grade |
ecoinvent generic gold + Higgs / Boyd purification penalty | Generic + Higgs/Boyd | 2012 baseline | Industrial-to-semi delta not updated publicly since ~2012. |
| Cu (electronics-grade) | ICA (2022) generic + REEL modeling | Generic public | 2022 | Bulk LCI sound; electrodeposition delta not characterized. |
| Cu-Mo alloy substrate |
ICA + IMOA (2018) | Generic public | 2018–2022 | Bulk mass-allocation only. |
| Silver die-attach paste | ecoinvent generic silver | Generic public | 2023 (v3.10) | No public LCI for sintered nano-silver. |
| Bulk gases and utilities | ||||
| N₂ / O₂ / Ar cryogenic air separation |
ecoinvent v3.10 air-separation datasets | Generic public | 2023 (v3.10) | Merchant supply only; on-site fab generation differs. |
| H₂ (semiconductor-grade) | REEL LCI Database v1.0 (modeled) for electronics grade | REEL-modeled (published) | 2026 | Merchant H₂ in ecoinvent; semi-grade purification delta absent. |
| Compressed dry air | ecoinvent compressed air | Generic public | 2023 (v3.10) | Bulk only. |
| He (merchant) | ecoinvent v3.10 | Generic public | 2023 (v3.10) | Recovered from natural-gas processing, not air separation; merchant supply only. |
| Substrate organics and films | ||||
| Ajinomoto Build-up Film ABF |
REEL LCI Database v1.0 (modeled) | REEL-modeled (published) | 2026 | Single-supplier monopoly; modeled via silica-filled epoxy proxy. |
| Polyimide PMDA / ODA |
ecoinvent polyimide | Generic public | 2023 (v3.10) | Zhang / Bainbridge 2024 covers other substrates but not PI. |
| Epoxy mold compound EMC |
ecoinvent generic epoxy + REEL filler modeling | Generic public | 2023 (v3.10) | > 80% filler by mass; generic epoxy substantially mis-states burden. |
| Benzocyclobutene BCB |
REEL LCI Database v1.0 (modeled) | REEL-modeled (published) | 2026 | REEL-modeled. |
| Dry film resist | REEL LCI Database v1.0 (modeled) | REEL-modeled (published) | 2026 | Asahi Kasei / Toray run internal LCA, not published. |
| Wet-clean chemicals (electronics grade) | ||||
| H₂O₂ | ecoinvent industrial + Higgs / Boyd | Generic + Higgs/Boyd | 2012 baseline | Same pattern as bonding gold. |
| NH₄OH | ecoinvent + Higgs / Boyd | Generic + Higgs/Boyd | 2012 baseline | Purification delta vintage. |
| HCl | ecoinvent + Higgs / Boyd | Generic + Higgs/Boyd | 2012 baseline | - |
| H₂SO₄ | ecoinvent + Higgs / Boyd | Generic + Higgs/Boyd | 2012 baseline | - |
| HF dilute, electronics |
ecoinvent + Higgs / Boyd | Generic + Higgs/Boyd | 2012 baseline | - |
| Substrate | ||||
| Silicon wafer single-crystal CZ |
ecoinvent v3.10 | Generic public | 2023 (v3.10) | Electronics-specific CZ dataset in background DBs; refinements marginal. |
Note: ecoinvent vintages are cited as v3.10 / v3.11, the releases current during REEL LCI v1.0 development. These are not the latest release; ecoinvent v3.12 is now current, and the datasets referenced here are expected to carry forward into v3.12 with refreshed background data.
6.1 · What the table reveals
Across the table's thirty-seven substance rows, only three carry a peer-reviewed open-access source that is genuinely industry-specific: Thompson (2020) for NF₃ (narrow, single-fab scope), Fuentes et al. (2023) for TMAH (modeled, not vendor-validated), and Wagih, Bainbridge et al. (2024) for ENIG / ENEPIG (explicit gold + nickel mass model). A fourth, Venkataswamy et al. (2024) for CMP consumables, provides a qualitative perspectives review but no quantified emission factor.
Notably, not a single row qualifies for the "Industry-specific public" label: no vendor EPD, no consortium-published LCI, and no government dataset built from semiconductor-industry primary data was located for any substance in the table.
Every other row falls into one of three categories: REEL-modeled (published), + Higgs / Boyd, or Generic public. In other words: for most semiconductor-specific materials and chemicals, REEL LCI itself is the most representative emission-factor source available to practitioners, built from precursor chemistry because no peer-reviewed or vendor inventory exists. Section 7 addresses where additional contributions would have the largest leverage.
07 Where better data would move the needle
Leverage tiers: substances grouped by how much fixing them would change REEL's downstream totals.
The substance rows in Section 6 are not all equally important to fix. The tiers below reflect modeled leverage on REEL's downstream wafer and package totals; that is, the answer to the question: if a credible public emission factor for this substance became available next year, how much would it change the numbers we publish?
Tier 1 · Highest leverage
No public LCI exists today. Improvement here moves every dependent wafer and package total.
Substances where REEL currently models the upstream from precursor chemistry and engineering estimates, and which appear as top-three non-energy contributors in multiple advanced products.
- Specialty process gases: production-phase LCINF₃, C₄F₈, C₄F₆, GeH₄, F₂, and electronics-grade SF₆. Single largest cluster of avoidable uncertainty in advanced-node wafer LCIs.
- CMP slurries: quantified per-kg EFOxide, copper, and tungsten formulations. Venkataswamy et al. (2024) provides the qualitative basis; a quantified open inventory would close the gap.
- Photoresist upstreamArF and EUV (organic and metal-oxide). Currently an absolute data void. Suppliers run internal LCAs but do not publish.
- Ajinomoto Build-up Film (ABF)Single-supplier monopoly product with no public emission factor. Modeling via silica-filled-epoxy proxy is the current workaround.
- Polyimide (semiconductor-grade)Zhang, Bainbridge et al. (2024) covers FR4, PET, paper, and a degradable substrate but does not extend to polyimide.
- Other proprietary films: BCB and dry film resistThe same no-public-LCI pattern as ABF and polyimide at smaller leverage; REEL models both from precursor chemistry.
Tier 2 · High leverage
Generic data exists. The semi-grade processing delta has not been characterized publicly in over a decade.
Improvement would shift specific product families substantially but not the entire database.
- Bonding wire gold69–93% of total package GWP in the thin gold-wirebond packages (TQFP, TSOP, SSOP, TSSOP). Semi-grade purification and wire-drawing delta unrefreshed since the early 2010s (Boyd 2012, summarizing Higgs).
- Plating-bath additive packagesSuppressors, accelerators, levelers in Cu damascene and gold electroplating baths. EU GENESIS expected to publish around 2028.
- Sintered nano-silver die-attach pasteGeneric silver in ecoinvent; nano-silver paste formulations not characterized.
- Epoxy mold compound (EMC)EMC is > 80% silica filler by weight; a quantified semi-grade EMC inventory would correct package-level totals across virtually all encapsulated products.
- Semiconductor-grade silane (SiH₄)Bulk-chemical dataset in ecoinvent; the high-purity processing delta is uncharacterized.
- TMAH: vintage updateFuentes et al. (2023) at ecoinvent 3.5 background. A vendor-validated or more recent academic update would tighten the picture.
- ENIG / ENEPIG: broader scopeWagih, Bainbridge et al. (2024) within a microwave-PCB scope. Generalization across the full FCBGA family would extend utility.
- Electronics-grade wet-clean chemicalsH₂O₂, NH₄OH, HCl, H₂SO₄, HF. The Higgs / Boyd 20–1000% purification penalty is well-known qualitatively but has not been quantified publicly in over a decade.
Tier 3 · Broad exposure
Per-unit factors are small but ubiquity creates large aggregate exposure across the database.
Improvement is broad but proportionally small.
- Industrial N₂, O₂, Arecoinvent provides adequate merchant-supply factors. On-site fab generation differs in energy intensity; a published delta would refine on-site-vs-merchant comparisons.
- Helium (merchant supply)Recovered from natural-gas processing rather than air separation; the ecoinvent merchant dataset is adequate; minor per-unit effect.
- Semiconductor-grade H₂Merchant H₂ in ecoinvent; the electronics-grade purification delta is not.
- Compressed dry air (CDA)Bulk-utility factor; widely used; minor per-unit effect.
- Silicon wafer (single-crystal CZ)ecoinvent covers this adequately at the substrate level; refinements would be marginal.
7.1 · The form of improvement that would help, by tier
Tier 1 needs
Peer-reviewed publications or open consortium-published LCIs covering production-phase chemistry, mass balance, and energy inputs. Vendor EPDs would be valuable but may not be forthcoming for proprietary formulations. Modeled inventories, built transparently from public literature on synthesis chemistry, are an acceptable substitute and could be contributed by academic groups, supplier sustainability teams (publishing methodology even if the formulation remains proprietary), or industry-funded research projects similar to EU GENESIS.
Tier 2 needs
Vendor EPDs that quantify the semi-grade processing delta. The chemistry is largely known; what is missing is the producer's primary data on purification energy, drag-out losses, and waste-stream burden. Suppliers in this tier (Tanaka, Heraeus, Atotech / MKS, Henkel, etc.) are in a strong position to publish without exposing formulation IP.
Tier 3 needs
On-site fab disclosures. Foundries already disclose total energy and water consumption; an additional breakdown by utility-plant function (bulk-gas generation, abatement) would close most of the Tier 3 gap.
08 Limitations and open questions
Scope boundaries readers should keep in mind.
Within- and between-fab variation
REEL LCI v1.0 represents industry-typical models for each technology node and package architecture, built bottom-up from public sources via a "virtual factory" methodology. Actual fabs differ in equipment vintage, process recipes, abatement systems, chemical sourcing, on-site utility generation, and water/energy mix, and the same fab varies over time as recipes and tools change. A specific fab's hotspot ranking may therefore differ from REEL's screening: some emission factors will be more relevant for that fab, others less so, and the relative weight of the materials-and-chemicals slice may shift up or down depending on the fab's electricity mix and abatement performance. Suppliers reading this paper should weigh the hotspot rankings against the actual fab mix of their customers rather than treating them as the universal ordering for every product.
GWP100 only
All analysis is conducted using AR6 GWP100 characterization factors. Other impact categories are not addressed: water scarcity (AWARE), acidification, eutrophication, human toxicity, ecotoxicity, and abiotic resource depletion. Several of the hotspots identified here have very different rankings in non-GWP categories. Water scarcity in particular concentrates much of the impact on bulk and ultrapure water flows, a dimension this paper's GWP scope does not address.
v1.0 coverage
REEL LCI v1.0 covers the dominant technology nodes and packaging architectures, but several specialty domains are not yet fully screened: compound semiconductors (GaAs, GaN, SiC), photonics (silicon photonics PIC, fiber optics), MEMS, image sensors, and analog / RF discretes. Some hotspots that are minor in advanced logic and HBM may rise in importance in these specialties (e.g., gallium, indium, rare earths).
Direct-gas emissions are out of scope by construction
Fluorinated process gas emissions to atmosphere are dominant in absolute GWP terms for advanced nodes and are well-characterized elsewhere. They are excluded for scope reasons: the public-data gap sits elsewhere. Any reader integrating this paper's findings into a full life cycle inventory should treat them as a separate, well-developed line item.
Bonding gold and recycled content
Bonding-wire gold figures in standard databases assume primary gold mining and refining. Real semiconductor assembly supply chains use varying fractions of recycled gold (from scrap wire, plating sludge, and end-of-life recovery). Recycled gold has a substantially lower emission factor; the actual aggregate-weighted footprint is therefore lower than the generic gold dataset suggests for most products. This is a known limitation of every public bonding-gold model.
Vintage of academic baselines
Several references this paper relies on are now over a decade old (Boyd 2012, Higgs ~2009 reviewed in Boyd). The 20–1000% industrial-to-electronics-grade purification penalty is a well-known framing but has not been refreshed in the open public literature. The actual delta today may be smaller (modern facilities are more energy-efficient) or larger (modern electronics-grade purity targets are stricter); without a contemporary public study, it is not knowable from public sources.
Search scope
The literature and disclosure scan behind this paper covered open-access peer-reviewed journals, indexed theses, foundational LCA monographs (Boyd 2012), vendor sustainability and ESG reports, consortium publications (SEMI, JEITA, IPC), IPCC methodology volumes, government LCI databases (USLCI, ELCD), and the ecoinvent v3.10 and v3.11 release notes. Licensed databases not openly accessible to the public are noted only where their existence is methodologically relevant.
09 References
Primary sources cited in this paper.
Each entry includes a DOI or URL; sources are openly accessible except where noted.
Peer-reviewed publications and theses
- Boyd, S. B. (2012). Life-Cycle Assessment of Semiconductors. Springer. doi:10.1007/978-1-4419-9988-7.
- Fuentes, O. P., Cruz, J. C., Mignard, E., Sonnemann, G., & Osma, J. F. (2023). Life cycle assessment of magnetite production using microfluidic devices: moving from the laboratory to industrial scale. ACS Sustainable Chemistry & Engineering. doi:10.1021/acssuschemeng.2c06875.
- Li, X., Zeng, R., Shi, X., & Lin, Y. (2025). A review of the Cu chemical mechanical planarization process in hybrid bonding technology. Journal of Electronic Packaging, 147(3). doi:10.1115/1.4068883.
- Thompson, M. (2020). Nitrogen Trifluoride and Furfural Supply Chains: LCA and TEA. University of Idaho M.S. Thesis. verso.uidaho.edu.
- Venkataswamy, R., Trimble, L., McDonald, A., Nevers, D., Vazquez Bengochea, L., et al. (2024). Environmental impact assessment of chemical mechanical planarization consumables: challenges, future needs, and perspectives. ACS Sustainable Chemistry & Engineering, 12(32), 11841–11855. doi:10.1021/acssuschemeng.4c03195.
- Wagih, M., Bainbridge, A., et al. (2024). Environmental life-cycle assessment of wireless RF systems: a comparative sustainability analysis and a microwave engineers' guide to LCA. IEEE Journal of Microwaves. doi:10.1109/JMW.2024.3455575.
- Wang, X. (2024). Carbon Footprint of Semiconductor Products. KTH Royal Institute of Technology M.S. Thesis. urn.kb.se.
- Zhang, T., Bainbridge, A., et al. (2024). Life cycle assessment of circular consumer electronics based on IC recycling and emerging PCB assembly materials. Scientific Reports. doi:10.1038/s41598-024-79732-1.
Industry, consortium, and government references
- IPCC AR6 WG1 (2021). Climate Change 2021: The Physical Science Basis. doi:10.1017/9781009157896.
- IPCC (2019). 2019 Refinement to the 2006 IPCC Guidelines, Volume 3 Chapter 6 (Electronics Industry Emissions). ipcc-nggip.iges.or.jp.
- International Copper Association (2022). GHG Measurement of Copper Production. internationalcopper.org.
- International Molybdenum Association (2018). LCI Update Summary Report. imoa.info.
Foundry operational disclosures (fab water)
- TSMC (2025). 2024 Sustainability Report. esg.tsmc.com.
- Samsung (2022). Corporate Sustainability Report 2022. semiconductor.samsung.com.
- Intel (2017). Corporate Responsibility Report 2017. csrreportbuilder.intel.com.
Database and consortium references (descriptive)
- ecoinvent (2024). Version 3.10 and 3.11 release notes. ecoinvent.org.
- EU GENESIS Project (Horizon Europe, CORDIS 101194246). cordis.europa.eu.
- Semiconductor Climate Consortium (SEMI). semi.org.
- Imec Sustainable Semiconductor Technologies and Systems (SSTS) and imec.netzero. netzero.imec-int.com.
REEL LCI
- REEL LCI Methodology Report v1.0 (2026, to be published). reellci.com.
- REEL LCI Database v1.0 (2026, to be published). reellci.com/datasets/.
Appendix A Substance index
One-line entry per substance in the Section 6 landscape table.
| Substance | Formula / abbrev. | CAS | Primary use | See | Tier |
|---|---|---|---|---|---|
| Ajinomoto Build-up Film | ABF | - (composite) | Flip-chip substrate dielectric | 5.7 / 6 | Tier 1 |
| Ammonium hydroxide | NH₄OH | 1336-21-6 | SC-1 wet clean | 3.3 / 6 | Tier 2 |
| ArF immersion photoresist | - | - (formulation) | Patterning, 193 nm immersion | 5.3 / 6 | Tier 1 |
| Argon | Ar | 7440-37-1 | Inert ambient, sputter carrier | 5.6 / 6 | Tier 3 |
| Benzocyclobutene (DVS-bis-BCB) | BCB | 117732-87-3 | Advanced packaging dielectric | 5.7 / 6 | Tier 1 |
| Bonding wire gold | Au (4N–6N) | 7440-57-5 | Wire bonding | 5.5 / 6 | Tier 2 |
| Bottom anti-reflective coating | BARC | various | Anti-reflective underlayer for patterning | 5.3 / 6 | Tier 1 |
| C₄F₆ (hexafluoro-1,3-butadiene) | C₄F₆ | 685-63-2 | Selective dielectric etch | 5.1 / 6 | Tier 1 |
| C₄F₈ (octafluorocyclobutane) | C₄F₈ | 115-25-3 | Polymer-forming dielectric etch | 5.1 / 6 | Tier 1 |
| CMP slurry (copper) | - | various | Cu damascene polishing | 5.2 / 6 | Tier 1 |
| CMP slurry (oxide, silica / ceria) | - | various | STI, ILD planarization | 5.2 / 6 | Tier 1 |
| CMP slurry (tungsten) | - | various | W contact polishing | 5.2 / 6 | Tier 1 |
| Compressed dry air | CDA | - | Pneumatics, blow-off | 5.6 / 6 | Tier 3 |
| Copper (electronics-grade) | Cu | 7440-50-8 | Damascene, substrate metallization | 5.5 / 6 | Tier 2 |
| Copper-molybdenum alloy | Cu-Mo | - | Substrate, high-reliability | 5.5 / 6 | Tier 2 |
| Copper plating bath (base) | CuSO₄ + additives | various | Electrodeposition | 5.4 / 6 | 2 (additives), 3 (salt) |
| Dry film resist | - | various | Substrate patterning | 5.7 / 6 | Tier 1 |
| Edge-bead remover | EBR | various | Resist edge removal | 5.3 / 6 | Tier 1 |
| ENIG / ENEPIG | - | various | Substrate / PCB surface finish | 5.4 / 6 | Tier 2 |
| Epoxy mold compound | EMC | various | Package encapsulation | 5.7 / 6 | Tier 2 |
| EUV photoresist | - | various | Patterning, 13.5 nm | 5.3 / 6 | Tier 1 |
| F₂ (fluorine) | F₂ | 7782-41-4 | Chamber clean (NF₃ alternative) | 5.1 / 6 | Tier 1 |
| GeH₄ (germane) | GeH₄ | 7782-65-2 | SiGe epitaxy | 5.1 / 6 | Tier 1 |
| Helium | He | 7440-59-7 | Carrier gas, leak detection | 5.6 / 6 | Tier 3 |
| Hydrochloric acid (electronics) | HCl | 7647-01-0 | SC-2 wet clean | 3.3 / 6 | Tier 2 |
| Hydrofluoric acid (dilute, electronics) | HF | 7664-39-3 | Oxide etch, wet processing | 3.3 / 6 | Tier 2 |
| Hydrogen (semiconductor-grade) | H₂ | 1333-74-0 | Epi reduction, annealing | 5.6 / 6 | Tier 3 |
| Hydrogen peroxide (electronics) | H₂O₂ | 7722-84-1 | SC-1, SPM, SC-2 cleans | 3.3 / 6 | Tier 2 |
| KAu(CN)₂ (potassium gold cyanide) | KAu(CN)₂ | 13967-50-5 | Gold electroplating | 5.4 / 6 | 2 (additives), 3 (salt) |
| Nitrogen (industrial) | N₂ | 7727-37-9 | Inert ambient, purge | 5.6 / 6 | Tier 3 |
| Nitrogen trifluoride | NF₃ | 7783-54-2 | Chamber clean | 5.1 / 6 | Tier 1 |
| Oxygen | O₂ | 7782-44-7 | Oxidation, ashing | 5.6 / 6 | Tier 3 |
| Polyimide (PMDA / ODA) | - | various | RDL, flex substrate | 5.7 / 6 | Tier 1 |
| Silane | SiH₄ | 7803-62-5 | Si and dielectric deposition | 5.1 / 6 | Tier 2 |
| Silicon wafer (single-crystal) | Si | 7440-21-3 | Substrate | 3.3 / 6 | Tier 3 |
| Silver (die-attach paste, sintered nano) | Ag | 7440-22-4 | Die attach | 5.5 / 6 | Tier 2 |
| Sulfur hexafluoride | SF₆ | 2551-62-4 | Etch (Si, SiN, W, MEMS DRIE), chamber clean | 5.1 / 6 | Tier 1 |
| Sulfuric acid (electronics) | H₂SO₄ | 7664-93-9 | SPM, piranha clean | 3.3 / 6 | Tier 2 |
| Tetramethylammonium hydroxide | TMAH | 75-59-2 | Resist developer | 5.3 / 6 | Tier 2 |
Appendix B Provenance taxonomy
Definitions and assignment rules for the six provenance labels.
The six provenance labels used in this paper are defined as follows. The rule for assigning each label is given in plain language.
Peer-reviewed (open)
The substance has a cradle-to-gate emission factor or production-phase LCI published in an open-access peer-reviewed journal article, conference proceeding indexed in IEEE / ACM / ACS / Nature platforms, or a publicly accessible university thesis. The DOI or URN resolves and the source is independently verifiable. The cited content directly supports the emission-factor claim. A tangential mention, a methodology reference, or a single quantitative footnote does not qualify.
Rule. Assigned when (a) the source exists publicly, (b) it contains a cradle-to-gate or production-phase LCI for the substance, and (c) the LCI is sufficient to compute or look up an emission factor.
Industry-specific public
The substance is covered by a vendor EPD (an Environmental Product Declaration formally registered with an EPD program operator), a consortium-published LCI built from semiconductor-industry primary data, or a government LCI dataset (USLCI, ELCD) that is industry-specific. The source is openly accessible to the public.
Rule. Assigned when (a) the source is from a semiconductor-industry actor or government, (b) the data is primary or based on primary industry input, and (c) it is openly accessible without a license fee.
Generic public
The substance is in a standard background LCI database (ecoinvent v3.10 or later, which is licensed but ubiquitous in LCA practice, or the free USLCI and ELCD), but only at industrial-bulk grade or at a generic chemistry that does not reflect the semiconductor-grade purification, processing, or formulation. The dataset is adequate for upstream chemistry but does not capture the semi-specific delta.
Rule. Assigned when (a) the substance is in a standard background database, (b) the available dataset is generic or only partially industry-specific, and (c) the semi-grade processing delta is not separately characterized in any public source.
Generic public + Higgs / Boyd
Same as Generic public, but where REEL applies an additional multiplier to estimate the industrial-to-electronics-grade purification energy penalty. The multiplier draws on the work of Higgs (Intel, ~2009) as summarized in Boyd (2012, Life-Cycle Assessment of Semiconductors), which reports purification penalties ranging from 20% to 1000% additional carbon footprint depending on the substance. This label flags rows where the underlying delta has not been refreshed in any public source for more than a decade.
Rule. Assigned when (a) Generic public would otherwise apply, and (b) REEL's model adds a documented purification multiplier from the Higgs / Boyd baseline because the semi-specific delta is meaningful but not separately characterized in current public sources.
REEL-modeled (published)
The substance has been modeled by REEL from precursor chemistry, mass balance, and engineering estimates and is published in the REEL LCI Database v1.0 (to be published). Applied when no peer-reviewed or vendor source exists publicly but REEL's bottom-up model is documented in the database release.
Rule. Assigned when (a) no Peer-reviewed (open) or Industry-specific public source exists for the substance, and (b) REEL publishes a modeled inventory for that substance in v1.0 or later.
No public LCI identified within search scope
A diligent search of peer-reviewed open-access literature, vendor EPDs, consortium publications, government databases, and ecoinvent release notes did not locate a qualifying public emission factor for the substance. REEL models it from precursor chemistry or applies a proxy from a related substance.
Rule. Assigned when (a) the prior five labels do not apply, and (b) the search scope described in Section 8 was applied without success. The conservative wording is "within search scope" rather than "does not exist" because new sources may have been published or may be in non-indexed locations.