Electrical and electronic equipment (EEE) is characterised by complex material systems and highly specialised manufacturing processes. Although finished equipment such as computers, displays, communication devices, servers and industrial electronics can be described relatively easily at product level, their environmental performance depends on upstream components whose production is considerably more difficult to characterise. Semiconductor devices, printed circuit boards, displays, magnetic storage systems and optical sensors combine high-purity materials, thin films, multilayer structures, intensive chemical processing and manufacturing infrastructures that can contribute substantially to their life-cycle impacts. Life cycle assessment (LCA) provides an established framework to quantify these impacts, but the reliability of the results remains dependent on the quality and representativeness of the underlying life cycle inventory (LCI). For rapidly evolving electronic technologies, this dependency matters because technological development can outpace the renewal of environmental datasets.

The difficulty is therefore not limited to the availability of environmental information. It also concerns its temporal, technological and geographical representativeness. Environmental inventories may describe manufacturing technologies that have since changed, while apparently comparable studies may use different boundaries, electricity mixes, scaling approaches, or assumptions about manufacturing losses and waste treatment. These issues have previously been discussed in relation to the limitations of environmental assessments of digital equipment. At the component level, however, the problem becomes more fundamental: a generic environmental factor may no longer capture the physical and technological parameters that drive differences between products.

EEE products: why better environmental data matter

---

The environmental assessment of EEE relies strongly on a relatively small number of upstream component families. Semiconductor inventories illustrate this dependence particularly well. Boyd's work and the inventories subsequently incorporated into widely used databases established an important methodological basis for semiconductor LCA, but later assessments have frequently extended or scaled earlier inventories rather than replacing them with independent primary measurements. Pirson et al. (2023) show that this genealogy creates difficulties when representing technological evolution because semiconductor fabrication has changed substantially in process complexity, lithography, device architecture and manufacturing infrastructure. Agreement between several environmental studies therefore does not necessarily indicate independent confirmation when their underlying inventories derive from the same primary sources.

The magnitude of uncertainty is also visible outside semiconductor manufacturing. Published estimates for the cradle-to-gate global warming potential of printed circuit boards range from approximately 18.6 kg CO₂-eq/m² in Ozkan et al. to several hundred kilograms CO₂-eq/m² in more recent modelling. Across the values compiled in the source material, the difference between the lower and upper estimates approaches a factor of 27. The range reflects differences in system boundaries, board configurations, electricity assumptions, process representation and upstream material modelling rather than simple statistical variation. Under these conditions, the environmental difference attributed to the choice of dataset can exceed the difference produced by the design modification being evaluated. The problem is therefore as much methodological as numerical: environmental inventories must retain enough information to distinguish the technical characteristics that determine both material composition and manufacturing requirements.

Introducing naKnow

---

We are part of naKnow, supported under the France 2030 ECONUM programme managed by ADEME. The work addresses environmental information for five component families that recur throughout digital equipment: semiconductor wafers and their derivatives, printed circuit boards, display panels, hard disk drives and optical sensors. Rather than treating these components through fixed average factors alone, the approach combines an open expert knowledge base with parametric environmental models and experimental characterisation. The objective is to preserve the relationship between the physical component, its manufacturing route and the environmental inventory so that datasets can be revised as technologies and evidence evolve.

The scientific approach combines literature analysis, process-level modelling and laboratory characterisation within the same modelling framework. Literature provides technical descriptions of manufacturing sequences, material systems, process inputs, and reported environmental performance, but it also reveals important data limitations. Some technologies are documented primarily through historical inventories, while more recent information may be available only at factory level or through aggregated manufacturer reporting. Process modelling therefore reconstructs relationships between technical parameters and environmental flows rather than applying a single factor to an entire component family. The distinction between determinant and scalable parameters is particularly useful in this context. A determinant parameter changes the technological configuration or determines whether a particular material or process is present, whereas a scalable parameter modifies the quantity required within that configuration. Surface finish, transistor technology or recording architecture can therefore define different technological archetypes, while board area, copper thickness, die area or panel dimensions govern scaling within those archetypes.

Laboratory characterisation provides an independent constraint on the material part of these models. The sampling strategy is built from modelling hypotheses so analytical measurements can test whether the selected technical descriptors explain the observed composition. The procedure therefore links parameter selection, sample identification, dismantling, mass balance, elemental analysis and model correction. This is particularly important because environmental models can remain internally consistent while being physically inconsistent with the component being represented. If two components classified using the same modelling parameters show substantially different measured compositions, the discrepancy identifies missing explanatory variables rather than an analytical inconvenience. The laboratory work was explicitly designed to obtain composition data capable of correcting or calibrating the models and to maintain traceability of the components, their origin and their technical characteristics throughout the analytical procedure.

Learnings from the naKnow project

---

Manufacturing Processes: A Key Driver of Environmental Performance

The results reinforce the importance of manufacturing processes in the environmental performance of EEE. Semiconductor fabrication is an extreme example because the final mass of a die contains little information about the sequence of operations required to produce it. Lithography, deposition, etching, cleaning, ion implantation, chemical mechanical planarization and metrology are repeated according to device architecture and process generation. More advanced technologies can increase mask counts, cleaning requirements, metallisation levels and the number of deposition–etch sequences even when individual features become physically smaller. Schmidt et al. (2012), Kuo et al. (2022) and Ruberti (2022) similarly identify technological generation, processing intensity, yield and manufacturing conditions as important drivers. Environmental performance cannot therefore be inferred directly from miniaturisation. A reduction in material per functional unit can coexist with increased process demand per unit wafer area.

Process intensity is not confined to semiconductor production. PCB fabrication requires imaging, lamination, drilling, copper deposition, electroplating, etching, surface finishing and wastewater treatment, while display manufacturing can combine vacuum deposition, photolithography and tightly controlled production environments. Advanced optical sensors add wafer thinning, stacking and hybrid bonding operations that are absent from conventional area-scaled semiconductor models. Although the technological details differ, the methodological consequence is similar: generic component categories omit information needed to explain environmental variation. A useful LCI must therefore retain parameters that modify the manufacturing route rather than describing technological differences only through component mass or external dimensions.

Energy emerges as a major cross-cutting factor, but its importance varies with the processes included and the assumed electricity system. The PCB literature compiled in the source material places electricity at approximately 45–86% of global warming potential in different manufacturing studies. Amasawa et al. reported facility energy as 41.5% of the production impact considered for AMOLED manufacturing. Semiconductor fabrication adds substantial electricity requirements from cleanroom conditioning, vacuum equipment, plasma generation, thermal processes and the repeated operation of fabrication tools. Manufacturing geography consequently becomes part of the technological description rather than a secondary contextual variable. Weppe et al. (2024), for example, represent CMOS image-sensor fabrication through a model in which manufacturing energy is directly coupled with the carbon intensity of the electricity supply. Two nominally equivalent components can therefore have different climate-change impacts when produced within different electricity systems.

Consumables create a related but less visible contribution because many manufacturing inputs are not present in the finished component. Acids, etchants, photoresists, solvents, cleaning agents, ultrapure water and process gases can be consumed in substantial quantities while leaving little or no material signature in the product. Subtractive manufacturing makes this distinction particularly important. Copper removed during PCB etching has already incurred the environmental burden of extraction, refining and foil production before entering the waste-treatment system. Semiconductor fabrication repeatedly deposits, patterns, cleans and removes materials, meaning that the inventory of the finished device systematically understates the gross material throughput of its manufacture. Environmental modelling must therefore account for both the material incorporated into the component and the material consumed to create it.

The fate of these process inputs introduces additional uncertainty. Wastewater treatment and abatement determine whether substances are captured, transformed or emitted, and their environmental consequences depend on the chemical form and receiving compartment. This is especially relevant for fluorinated gases in semiconductor manufacturing, where climate impacts depend strongly on utilisation, destruction, and removal efficiencies. Site-level environmental reporting does not fully resolve the problem because factories generally report electricity, water, wastes and emissions for the facility as a whole, whereas LCI requires attribution to products or process modules. Allocation from site to process therefore constitutes an additional modelling assumption that should remain explicit.

Yield interacts with all these factors. Material inputs, electricity, consumables and facility requirements associated with rejected units must ultimately be allocated to the successful output. Yield losses occurring after numerous processing steps therefore carry a larger environmental burden than losses occurring early in production. This becomes particularly relevant for manufacturing systems with long sequences of repeated processes. Consequently, the environmental effects of process complexity, electricity intensity, and yield are coupled rather than independent, and modelling them as isolated parameters can underestimate their combined influence.

Materials have an impact

---

Low-Mass Materials with High Environmental Impacts

The prominence of manufacturing processes does not reduce the importance of material composition. The relative importance of materials and processes changes according to the environmental indicator considered. Electricity and direct process emissions can dominate climate change, ionising radiation, or primary energy demand, whereas extracting and refining comparatively small quantities of metals can become decisive for abiotic resource depletion and can also contribute significantly to toxicity-related indicators. A climate-only interpretation can therefore create the impression that materials are secondary when, under another indicator, the ranking of environmental drivers is reversed. This distinction matters for EEE because component architectures combine high-mass structural materials with low-mass functional materials whose environmental intensities can differ by several orders of magnitude.

The first PCB analyses provide direct evidence of the limits of generic material assumptions. Two bare four-layer FR-4 boards declared with ENIG surface finish contained approximately 24.76 wt% and 11.38 wt% copper, respectively, in the prepared samples. The factor of approximately 2.2 cannot be explained by layer count or substrate designation because these descriptors were nominally identical. Copper coverage, routing density, internal planes, actual stack-up, plating and panel configuration remain plausible explanatory variables. The result does not imply that generic PCB inventories are unusable, but it demonstrates that commonly available descriptors are insufficient to determine composition at the accuracy required for parametric modelling. Similarly, the analysis of populated boards shows that solder, connectors, shields, and mounted components introduce additional metals and require separating bare-board composition from assembly-related contributions.

Material significance also cannot be inferred from mass fraction alone. Gold may occur in very small quantities but contribute disproportionately to resource-related impacts because of upstream extraction and refining. In such cases, plated area and coating thickness are more meaningful modelling parameters than total board mass. The same principle extends to indium, gallium, tantalum, rare-earth elements and platinum-group metals used in different electronic technologies. Yeom et al. (2018), for example, reported substantial differences in metal content and resource-depletion potential between the OLED and LCD systems examined. The value of this observation is not to extrapolate a numerical ratio to all displays, but to show that a change in technology can change the material system itself and therefore alter environmental indicators independently of changes in component mass.

A similar distinction applies to magnetic storage. Hard disk drives contain large quantities of structural metals, but some resource-related impacts originate from much smaller quantities of functional materials in magnets, recording layers and read/write systems. The modelling consequence is that mass-based allocation alone can misrepresent environmental relevance. We have already examined these questions in detail through the CEDaCI project, particularly in relation to data-centre equipment, material recovery and circularity. For further information on HDDs and their material systems, the CEDaCI work provides the appropriate technical background rather than repeating the complete analysis here.

Material characterisation as a mechanism to improve the LCI

Elemental characterisation by inductively coupled plasma optical emission spectrometry (ICP-OES) provides one way to constrain these material inventories experimentally. With the prepared sample mass and preparation losses known, the analytical result can be converted from concentration to absolute elemental mass, providing a physically measured inventory against which model assumptions can be evaluated. ICP-OES does not, however, determine layer sequence, layer thickness, plated area, organic composition, chemical speciation or manufacturing history. Its role is therefore to correct the model rather than build the inventory from scratch. Record structural information before homogenisation, and use complementary analytical approaches for polymeric and organic fractions. We have documented the ICP-OES preparation and analytical procedure in detail in the open knowledge base of naKnow , including dismantling, thermolysis, milling, digestion and instrumental analysis.

The broader implication is that material and process inventories should not be treated as competing explanations of environmental performance. The material inventory describes what is incorporated into the component, whereas the process inventory describes the additional resources required and emissions generated to produce that material configuration. Subtractive operations make the distinction especially visible because part of the environmental burden belongs to material that was produced but does not remain in the finished object. Conversely, elemental analysis can identify material differences that a generic process dataset cannot infer. A robust EEE inventory therefore requires both descriptions and an explicit relationship between them.

The role of WeLOOP

---

WeLOOP's role is concentrated at this interface between physical characterisation and environmental modelling. The analytical work begins with the parameters that the model intends to represent. Sampling methods are then defined so that laboratory measurements can test those parameters, while component origin, technical characteristics, preparation losses and analytical conditions are documented to preserve traceability. The resulting elemental measurements are not treated as standalone composition data. They are compared with theoretical relationships and existing inventory assumptions to determine whether the model parameters explain the observed variation and, where possible, to calibrate those relationships. This approach follows the intended purpose of the laboratory work: obtaining component-composition data that can correct or calibrate environmental models rather than building an empirical catalogue disconnected from the modelling framework.

This distinction is central to improving EEE environmental data. Additional datasets alone do not resolve the methodological problem if their underlying technological assumptions remain implicit. Environmental inventories need to distinguish measured information from inferred information, retain the parameters that explain technological variation and remain sufficiently transparent to permit revision when new evidence becomes available. Process modelling provides information that cannot be recovered from the finished object, including electricity use, process chemistry, manufacturing losses, and treatment. Laboratory characterisation provides physical evidence that process descriptions alone cannot reliably provide. Together, they allow environmental models to be tested against the component they claim to represent and provide a stronger basis for LCA-based ecodesign of electrical and electronic equipment.