Commercial-stage assessments typically provide quantified insights to environmental impact and economic performance questions for a technology in a specific setting. However, studies for early-stage technologies often fail to answer these questions at the same level of detail, as many options for the technology remain yet to be defined, and large parts of the specific settings are still unknown.

When defining the goal and scope of early-stage studies, LCA and TEA practitioners must manage the stakeholders' expectations and consider this knowledge gap. To manage stakeholder expectations, practitioners need to adapt the goals for early-stage studies not to conventional TRL 9 or commercial operations but for the most pressing questions at the current early development stage. The available data determines whether and in what detail a question can be explored (e.g., what yields are required to compete with incumbent technologies costs or which process unit is the key contributor to environmental impact). For example, a stage-gate process, which reduces the number of technology options at each gate, requires the adaptation of goals to support decision-making at each stage. Typical stage-gate decisions for a chemical CCU technology might be selecting catalysts at TRL 3, choosing equipment types at TRL 4, or narrowing the range of operating conditions at TRL 5.

In addition, practitioners need to define functional units, reference flows, and system boundaries. Functional units serve as the relative basis for the comparison and shall reflect all functions of the CCU process. Reference flows measure the outputs from processes required to fulfill the function expressed by the functional unit (ISO, 2006a, Amd1: 2020). The system boundaries specify the processes within the assessment's scope, which are fulfilling the function (Zimmermann et al., 2020). The goal and scope should be defined iteratively between practitioners and stakeholders (ISO, 2006b; European Commission - Joint Research Center, 2010; Zimmermann et al., 2020; Sylvia et al., 2021, Amd2: 2020, 2006, Amd1: 2020). As best practice for this iterative process, Bergerson et al. provide a list of questions to discuss with stakeholders during goal and scope definition, helping manage expectations (Bergerson et al., 2020).

With increasing maturity and data availability, practitioners can also increase the goal's level of detail. While the screening of various reactions at low TRL requires a simplification of assumptions (e.g., only considering the stoichiometry of the primary reaction with complete conversion and perfect selectivity), the comparison of options for equipment types at medium TRL requires specific models for each type and expected operation conditions (Roh et al., 2020; Zimmermann et al., 2021). Thus, LCAs and TEAs can be conducted several times during the R&D process—for example, at each stage-gate, the goal and scope can be refined, asking more detailed questions and providing more specific evaluations at each stage. An assessment at the next higher TRL is a new iteration and can build on the previous results. For example, while the research question of an assessment of a CO₂ capture technology at a low TRL may focus on the general feasibility and potential, it becomes more detailed at a mid TRL, such as specifying availability, costs, and environmental impacts of different CO₂ sources, and even further detailed at a high TRL, for example, discussing the equipment alternatives for one specific capture process.

Typically, research questions aim to identify a technologies' relative benefits and drawbacks. This identification requires comparing the assessed technology to a reference or benchmark technology, and practitioners need to select the benchmark technology according to the function and market of the assessed technology to ensure comparability.

One complication for early-stage assessments arises when the potential benchmark technologies are also under development. It is paramount to have information on the final function and the related reference flow, markets, and customer segments of the assessed technology and the benchmark for selecting a future benchmark technology. However, such data is subject to change, not yet defined in detail, or missing entirely at early development stages. For example, CO₂-based methanol can serve as a chemical intermediate, combustion fuel, or medium for electricity storage and release; potential benchmark technologies for fuels in the shipping market could be hydrogen, ammonia, methane, bio-oils, or also electric propulsion. Another complication for early-stage assessments arises from co-product use, as recycling and disposal strategies may deviate between the assessed technology and a benchmark. For example, producing e-crude from CO₂ by Fischer-Tropsch synthesis provides several products, including fuel gas, fuel oils, and waxes, which can each replace products in different industries and thus potentially each require the selection of a separate benchmark.

At the start of each assessment iteration, identifying and updating the function and market is highly encouraged, as this information is one of the cornerstones for conducting an assessment (ISO, 2006a, Amd1: 2020); if not or only rudimentary available, we recommend setting up a joint identification process with the commissioners and stakeholders of the study (Sylvia et al., 2021). When selecting the benchmark, three considerations are key:

First, the benchmark should reflect the goal of the low TRL study. One typical goal at low TRLs is replacing an existing fossil technology with a CCU technology. In this case, the existing technology represents the so-called “marginal technology” and, as such, serves as the benchmark technology. However, besides the assessed CCU technology, there might be additional and better-performing technology alternatives. Ensuring that the assessment captures the competitiveness of the CCU technology to both existing and future technology alternatives, we recommend including benchmarks reflecting the marginal and future technology. Therefore, practitioners should encourage stakeholders to explore which technology options might exist and how they could be included in the comparison. For example, a methanol company plans to replace a conventional steam methane reforming process with a production based on CO₂ from industrial flue gas. In this case, fossil methane production serves as benchmark technology, but other CCU or bio-based methanol production pathways could also be of interest for the comparison. Another typical goal at low TRLs is to identify the best-performing unit operations. The unit operations should be compared to each other, including all related impacts or costs. For example, an R&D team needs to decide between a separation via distillation vs. a separation via extraction. The selection of a unit process has implications for the overall process that the R&D team needs to consider—the extraction, for example, requires the separation and recycling of the solvent.

Second, practitioners must ensure that both processes fulfill the same function(s) as defined by the functional unit at each study iteration. If the products of the benchmark technology and the assessed CCU technology are chemically identical (e.g., CO₂-based methanol and fossil methane-based methanol), they will serve the same function, and the system boundary can be limited to cradle-to-gate (Kleinekorte et al., 2020; Zimmermann et al., 2020). In this case, product quality (e.g., purity) must be the same for both technologies.

Third, the benchmark should reflect the technology expected to be replaced when the novel CCU technology enters the market. Skone et al. recently developed recommendations to enable a harmonized selection of benchmarks at low TRL and suggest using the best-in-class technology based on greenhouse gas emissions (Skone et al., 2019).

If practitioners identify multiple functions or markets, we suggest a prioritization based on stakeholder engagement, or conducting a series of assessments, showing the pros and cons for each function or market separately as described by Wender et al. (Wender et al., 2014). If practitioners identify multiple functional units, we suggest prioritizing or conducting a series of assessments, which helps evaluate the sensitivity toward those methodological choices and interpret the results (Mendoza Beltran et al., 2020). Last, while it may be apparent to LCA and TEA practitioners that the lacking information at early maturity limits the evaluation, this limitation needs to be communicated clearly to stakeholders, especially in the interpretation and discussion sections.

A central challenge when assessing early-stage technologies is that, in many cases, data necessary for creating an inventory or calculating impacts, costs, or revenues is missing. Reasons for missing data can be manifold. Typical examples are critical reaction and process data are not yet measured or modeled, technical performance has not yet been demonstrated in the specific function, environmental impact data are not yet characterized in standard LCA databases, or costs and revenue data are outside of budget reach. Practitioners can estimate missing inventory, impact, cost, and revenue data for all maturity stages to overcome the data gaps (Parvatker and Eckelman, 2019).

Each assessment iteration should start with a round of stakeholder engagement. Establishing, updating, and validating the block flow diagram during the stakeholder engagement is critical. Practitioners can use this information to estimate mass and energy balances data and build the inventory (Tsoy et al., 2020). At TRL 2, the reaction's stoichiometry and enthalpy, the theoretical foundation for any mass and energy balance, provide the first data points. At TRL 3, practitioners can enrich the mass and energy balance with available information such as yield and reaction conditions (Althaus et al., 2007; Meys et al., 2020; Roh et al., 2020; Zimmermann et al., 2021). At TRL 4 and higher, practitioners can estimate missing data points from expert interviews or common industry values, such as a standard purge rate or typical energy consumption of a specific separation process, as information about the planned unit operations becomes available (Piccinno et al., 2016).

Practitioners can use the calculated mass and energy flows for both LCA and TEA—chemical engineering textbooks provide typical values and heuristics (Perry et al., 1997; Towler and Sinnott, 2008). If found significant, practitioners can estimate missing data, employing heuristics, expert interviews, machine learning tools, common industry values, or values from similar markets (Zimmermann and Schomäcker, 2017; Roh et al., 2020; Zimmermann et al., 2020; Sayyed Ahmad et al., 2022).

Process simulation can be applied if sufficient information and technical expertise are available. Process simulation does require the specification of a process flow diagram, but this can also be hypothetical at early stages. However, practitioners must make many assumptions, for which they must involve subject matter experts. The R&D community used process simulation already for a large variety of CCU technologies, such as the individual conversions of CO₂ to CO (van Bavel et al., 2020), methanol (Pérez-Fortes et al., 2016b; Blumberg et al., 2017), formic acid (Pérez-Fortes et al., 2016a), DME (Michailos et al., 2019; Ostovari et al., 2020) or carbonates (Ostovari et al., 2020) and the combination of individual conversion steps to value chains for chemicals (Kätelhön et al., 2019; Roh et al., 2019; Lee et al., 2021), and plastics (Meys et al., 2021).

During the impact assessment phase, characterization factors help to quantify environmental impacts. As CCU technologies might use or produce novel materials (e.g., CO₂-based organic polymers, filling materials, or concrete), those materials might not be well-characterized. In case of lacking impact assessment methods or characterization factors for novel materials, practitioners can follow chapter 6.7.4 of the ILCD handbook (European Commission - Joint Research Center, 2010). First, practitioners should evaluate the relevance of missing characterization factors assuming a conservative value or realistic worst-case value (e.g., based on expert knowledge or scientific literature). Practitioners can disregard the missing characterization factor if applying a characterization factor to the related elementary flow does not lead to relevant changes in an impact category. Nevertheless, the ILCD handbook recommends reporting missing characterization factors.

However, if applying a characterization factor leads to relevant changes, the contribution of the elementary flow is significant. In this case, the ILCD handbook recommends finding a more accurate and precise value for the missing characterization factor (European Commission - Joint Research Center, 2010). Molecular structure-based impact prediction or machine learning models can provide estimates of characterization factors (Roy et al., 2015; Dearden, 2016). Nevertheless, deriving characterization factors at a high quality is unrealistic for conventional LCA project timeframe and budget, as this requires deep expertise in life cycle impact assessment (LCIA) and specialized knowledge and data in the relevant chemical and environmental science issues. In practice, practitioners, and stakeholders must accept the limitations arising from a missing characterization factor and report them and their implications in the interpretation section.

Last, practitioners need to clarify whether and how the assessment can address the research question if data is missing. If the missing data makes answering a research question impossible, stakeholders and practitioners need to adapt it to the available data or pause the study until research and development have generated the missing data.

Not only data availability is a central challenge, but also the representativeness of available data. Data measured or calculated at low TRLs does not likely represent the final process accurately, as changes of unit operations, equipment, and the inclusion of efficiency measures such as heat integration or material recycling are common steps in research and development. But how can practitioners compare early-stage technologies with commercial-stage technologies representatively?

The conventional approach for comparing across different TRLs entails forecasting a technology's characteristics at an early stage to a commercial stage and comparing the predicted features to the incumbent or benchmark technologies (Tsoy et al., 2020). This approach is further called a “full-scope” assessment.

A second possible approach for comparing representatively across different TRLs is taking the reverse perspective from forecasting: “backcasting.” With backcasting, we mean simplifying the characteristics of a commercial-stage incumbent or benchmark technology to data typically available at an early stage and then comparing them to the ones of the early-stage technology in focus. As this approach uses technology maturity as a scope, it is further called a “maturity-scope” assessment (Buchner et al., 2018; Zimmermann et al., 2021).

For maturity-scope assessments that backcast commercial-stage data to early stages, shortcut methods provide suitable toolsets (Roh et al., 2020; Tsoy et al., 2020; Zimmermann et al., 2021). Shortcut methods differ from conventional, “full-scope” methods taking the shortcut of limiting scopes and leaving out intricate details, allowing simple comparisons. Shortcut methods are versatile when data availability is limited, such as at low maturity, or for reducing assessment effort, such as when comparing many alternatives (Zimmermann et al., 2021). When using maturity-scope assessments, practitioners should first estimate the best-case performance to check whether environmental impact reduction or economic benefits are possible at all. Practitioners could then add scenarios estimating a realistic case evaluating the robustness of comparisons, and investigating if a unit process may have a relevant contribution to the overall process.

Practitioners can also combine the two described forecasting and backcasting approaches: the commercial-stage data is backcasted, and early-stage technology's data forecasted but only to the next maturity stage. This hybrid approach allows for an evaluation of the R&D process by providing information on the technology's maximum potential and room for improvement (Zimmermann et al., 2021). For example, the assessment of a CCU technology producing Dimethyl Carbonate finds an energy efficiency of 72% at TRL 2; after further development of that technology to TRL 4, a new assessment iteration finds an energy efficiency of 51% (Zimmermann et al., 2021). While a slight decrease of efficiency from including material and energy losses to the assessment scope is to be expected when moving from a theoretical concept to a preliminary process design, such substantial decrease shows that additional factors contribute to the efficiency losses, which the R&D team needs explore in a subsequent design iteration.

Assessments at early stages have broader probability distributions and larger error bars than assessments at late stages, making recommendations in assessments at early stages much less specific and clear (Mendoza Beltran et al., 2018). Uncertainty itself is propagated from data, models, and choices to LCA and TEA results and can therefore not be avoided; the key is to derive meaningful recommendations also under uncertainty. Wynne defined four types of uncertainty: “risk”—known system parameters and known probabilities, “uncertainty”—known system parameters and unknown probability distributions, “ignorance”—unknown system parameters and unknown probability distribution, and “indeterminacy”—future development is inherently undetermined (Wynne, 1992). At high maturity, information on inventories, impact categories, characterization factors, and costs are generally available; ignorance and indeterminacy are typically low. Practitioners can analyze and quantify risk and uncertainty using conventional approaches, such as Monte-Carlo simulations, focusing on the “known unknowns.”

At low maturity, conventional approaches reach their limits, as “unknown unknowns” largely contribute to uncertainty (van der Giesen et al., 2020). For example, the timeline until market entry at early stages is still long, and the future market conditions are unknown. To deal with ignorance and indeterminacy, practitioners can consider multiple plausible futures scenarios, and involve a panel of stakeholders with most likely opposing views (van der Giesen et al., 2020). Practitioners can limit the impacts of uncertainty and increase the clarity of recommendations already during the goal and scope definition: The goal and scope of the study could be limited to identify promising options or deselect unpromising options. This way, practitioners can derive meaningful recommendations based on relative comparisons or best-case assumptions, reducing the amount of data required (Ravikumar et al., 2021a). Heijungs provides a more detailed discussion on uncertainty assessment for comparative decision-making (Heijungs, 2021).

Practitioners can further increase the clarity of recommendations by analyzing the uncertainty factors' impact on the final results and their correlations with sensitivity analysis. Ideally, practitioners exclude uncertainty factors with minimal impact on results or correlated factors affecting the assessed technology's and benchmark's results in the same way (Heijungs et al., 2019). Available methods include an anticipatory decision-driven approach (Ravikumar et al., 2018), stochastic multi-attribute analysis, and global or moment-independent sensitivity analysis (Borgonovo et al., 2012; Cucurachi et al., 2016; Ravikumar et al., 2021b).

Lastly, practitioners need to check whether the uncertainty in the results allows answering the research questions or if the goal and scope need to be adapted or the study interrupted until the uncertainty is reduced by research and development.

Required resources and times may depend on the purpose of the technology assessment. Data quality requirements are not critical when stakeholders request ex-ante comparative LCAs and TEAs for screening technology pathways. Such a study could be completed relatively quickly at a lower level of detail based on the “backcasting” approach described in 3.4 or existing publicly available studies with some effort in terms of harmonization. In contrast, practitioners need to apply a robust and comprehensive methodology and high-quality data for prescriptive LCA and TEA, as they generally require some level of certification.

When LCA and TEA are required for funding or developing new policies, technology developers and agencies often wait until the end of the development stage to engage with assessment experts, with the expectation to confirm the advantages of economic and environmental impacts with a high level of confidence. Data inventory collection and curing are time and resource-intensive, and conducting comprehensive LCAs and TEAs is, therefore, a time-consuming process. At low TRLs, the availability of primary data is limited, and gathering data would take a long time. Since assessment time is critical to support decision-making in the R&D process, LCA and TEA practitioners should consider trade-offs between the level of detail and the required analysis time to provide support in a timely fashion.

Simplified LCA and TEA (“shortcut”) methods (also see 3.4) can provide an effective and efficient solution for technology assessment at low TRLs (Beemsterboer et al., 2020). Streamlined approaches can be helpful to screen technology options or assess low TRL technologies. For example, the “Life-cycle Screening of Emerging Technologies” method (LiSET) developed by Hung et al. provides an LCA method that allows practitioners to extend the assessment scope from screening systematically and streamlined LCA to full LCA (Hung et al., 2020; Moni et al., 2020). Finally, a close collaboration between LCA and TEA practitioners, the R&D team, and further stakeholders can reduce the number of assessment iterations and overall assessment effort.