Life-Cycle Assessment of Innovative Industrial Processes for Photovoltaic Production: Process-Level LCIs, Scale-Up Dynamics, and Recycling Implications

Response to Reviewer 1 Comments

1. Summary

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted in the re-submitted files.

Comments: Hello authors, its good work but I have major concerns, pls address the questions carefully.

Comments 1: The paper claims to provide a critical review, but most sections summarize literature without clear critique. Could you explicitly point out contradictions, methodological conflicts, and gaps between studies? For example: where do LCAs disagree most (FU choice, boundary selection, grid-mix assumptions)?

Response 1: 

We thank the reviewer for his/her comment. We agree that the critical aspects needed to be made more explicit. We have substantially revised the text to foreground contradictions and methodological conflicts:

• A new paragraph has been added in end of Section 3.2 as follows: ‘Despite this emerging convergence on unit-operation classes, cross-study comparisons al-ready reveal important contradictions and gaps. Several LCAs of perovskite and thin-film lines reach opposite conclusions on whether solution or vacuum routes have lower embodied impacts because they adopt different functional units (1 m² versus 1 kWp versus kWh-delivered) and treat active/total area and lifetime inconsistently [3,8,12,26]. Likewise, some studies allocate a substan-tial share of energy use to encapsulation and substrate preparation, whereas others exclude or ag-gregate these steps into background data, obscuring the relative importance of wet versus vacu-um equipment [12,32,40]. A further gap is the limited availability of transparent vacuum-tool LCIs (pump-down, idle power, and chamber-conditioning profiles), which hinders reconciliation of reported hotspot patterns across ALD, sputtering and evaporation tools [17,37,38]. These issues motivate the more formal harmonization and methodological guidance developed in Sections 4 and 7.
• A new subsection ‘3.4 Cross-study contradictions and gaps’ was added at the end of Section 3). We explicitly compare representative LCAs that reach opposing conclusions about whether thin films or c-Si have lower GWP once geography and BOS are included. We highlight specific disagreements on:

1. Functional unit (FU): 1 m² vs 1 kWp vs kWh-delivered; we show how ignoring active/total area and lifetime can flip rankings by more than an order of magnitude, building on the discussion in Section 4.1.
2. System boundaries: cradle-to-gate vs cradle-to-grave, and inconsistent inclusion of BOS, inverters and EoL.
3. Electricity mix: we contrast LCAs using coal-heavy Chinese grids with those using low-carbon EU or hydropower-dominated mixes, where the same technology moves from worst to best performer.

• A new paragraph has been added in the end of Section 4.1 (Normalization rules) as follows: ‘Across the studies harmonized in this review, inconsistent application of these three elements (FU specification, system boundaries, and electricity-mix assumptions) emerges as the dominant source of disagreement between LCAs. For example, perov-skite and CdTe modules that appear clearly favourable to mono-Si under a coal-intensive grid and a 1 m² FU can lose this apparent advantage when evaluated per kWh-delivered under low-carbon electricity and cradle-to-grave boundaries that in-clude BOS and EoL [2,3,8,24,26,48]. The normalization rules adopted here are therefore explicitly designed to make such methodological drivers of divergence transparent.’.
• A new paragraph has been added in the end of Section 5.3, as follows: ‘Taken together, these comparative results also clarify where LCAs most strongly disagree. When technology assessments are aligned to a common FU (1 kWp or kWh-delivered), harmonized boundaries, and explicitly reported electricity mixes, the relative ordering between mono-Si, thin films, and perovskite/tandem modules becomes much more stable, with most of the residual spread attributable to differences in encapsulation mass, substrate design, and metallization strategy [3,8,12,24,26,48]. In contrast, studies that mix 1 m² and 1 kWp FUs, omit BOS or EoL, or rely on generic aggregated database modules for some technologies but process-level inventories for others often report conflicting technology rankings that cannot be reconciled on physical grounds alone. Highlighting these methodological conflicts is central to interpreting the literature and underpins the reporting rules proposed in Section 7.’
• At the end of Sections 6.2, we add bullet lists of missing data and several unresolved issues continue to limit the comparability of PV EoL LCAs.

The revised manuscript now repeatedly and explicitly calls out where LCAs disagree, why they disagree, and what information is missing to reconcile them.

Comments 2: In Section 4, you compile LCIs from literature, but it is unclear whether the data are averages or weighted results. Did you apply weighting based on technology readiness level? If not, would results change significantly if industrial data were weighted more heavily than lab data?

Response 2: We thank the reviewer for this important comment and have clarified both the aggregation approach and its implications in the revised manuscript.

First, we explicitly clarify in Section 4.2 and in the caption of Table 1, and in the captions of Figures 3 and 4 that the reported values are tier-separated ranges and simple arithmetic means, rather than a single pooled or weighted global average. LCIs are stratified by data-quality tier (Tier 1-3) corresponding to industrial/pilot data, scaled laboratory data, and modeled or literature estimates. Within each tier, simple arithmetic means are used to preserve transparency and to avoid obscuring systematic efficiency differences between laboratory and industrial inventories.

Second, to directly address the question of TRL weighting, we added a TRL-weighted sensitivity test in Section 4.3. In this analysis, Tier 1 (industrial/pilot) data were assigned double weight relative to Tier 2 and triple weight relative to Tier 3 when aggregating energy intensities for solution-based, vacuum-based, and hybrid process families.

The results show that applying TRL-based weighting shifts the central values by approximately 10-20%, typically lowering aggregated energy intensities due to the higher efficiency of industrial lines. Importantly, this weighting does not alter the relative ranking of process families (vacuum ≳ hybrid > solution) nor the identification of dominant hotspots such as annealing/drying, vacuum pumping, and encapsulation.

These clarifications demonstrate that while absolute values are moderately sensitive to aggregation choices, the main conclusions of the review are robust to reasonable TRL-based weighting assumptions. The revised text therefore makes explicit how averages are calculated, why unweighted tier-separated reporting is used as the baseline, and how alternative weighting schemes affect (or do not affect) the results.

Comments 3: The review heavily emphasizes electricity carbon intensity as the main driver. Could you quantify how much variability (in %) electricity alone contributes across technologies (Si, perovskite, CIGS) using explicit numeric ranges instead of only qualitative statements?

Response 3: We thank the reviewer for this valuable suggestion and have strengthened the manuscript by explicitly quantifying the contribution of electricity carbon intensity using numeric ranges.

In Section 4.1, we now report technology-specific ranges that isolate electricity as a single driver while holding all other parameters constant. Across the LCAs synthesized in this review, electricity accounts for approximately 35-65% of cradle-to-gate GWP for crystalline-silicon modules, 30-55% for CIGS and CdTe thin films, and 45-75% for perovskite (single-junction and tandem) and organic solar cell routes. When only the grid carbon intensity is varied between coal-dominated (≈800 g CO₂ kWh⁻¹) and low-carbon (≈50-100 g CO₂ kWh⁻¹) electricity mixes, the resulting variation in total cradle-to-gate GWP is ~50-70% for c-Si, ~40-60% for CIGS/CdTe, and ~60-80% for perovskite and OSC technologies.

In Section 5.2, these findings are reinforced at the scenario level, where harmonized analyses show that electricity alone typically drives 40-80% of the observed variability in cradle-to-gate impacts across identical manufacturing routes evaluated under different grid conditions. We further clarify that electricity carbon intensity, together with idle power and heating mode, explains ~70-80% of cross-study variability in process-resolved GWP for emerging PV manufacturing routes.

Comments 4: For scale-up modeling (lab→pilot→industrial), you mention 30–70% reductions. Is this range universal for all PV technologies or only for perovskites? Which manufacturing routes show the least improvement upon scaling?

Response 4: We thank the reviewer for highlighting that the originally stated 30-70% reduction range could be interpreted as universally applicable. We have clarified the scope of this range in both Section 5.1 and the Conclusions.

Specifically, the manuscript now explicitly states that the 30-70% reduction range refers to process-resolved, prospective LCAs, and is primarily observed for perovskite and organic solar cell manufacturing lines, as well as selected thin-film and polysilicon processing steps where strong learning effects, throughput increases, and uptime stabilization have been demonstrated.

We further clarify that crystalline-silicon modules as complete systems typically lie toward the lower end of this range (≈30-40%), because a substantial share of their environmental footprint is associated with upstream material-intensive processes (polysilicon, ingot growth, wafering) that are already relatively optimized at industrial scale.

In addition, we now explicitly identify the steps with the least scale-up gains, namely material-dominated components (e.g., glass, frames, encapsulants) and certain batch vacuum operations that already operate near optimal loading, where chamber size and pumping requirements limit further improvements. Conversely, annealing/drying and solution-coating steps are identified as exhibiting the largest relative reductions due to improved line speed, uptime, and duty-cycle optimization.

These clarifications ensure that the reported scale-up effects are interpreted correctly as process- and technology-specific, rather than universal across all PV manufacturing routes.

Comments 5: Several LCAs referenced use aggregated databases (e.g., Ecoinvent). How do you ensure that mixing aggregated and process-level LCIs does not introduce bias? Can you provide an example where aggregated data misrepresents a real process?

Response 5: We fully share this concern and have strengthened the manuscript to clarify how bias is avoided when combining process-resolved LCIs with aggregated background databases.

- Two new paragraphs have been added in end of 4.1. We now explicitly state in Section 4.1 that our meta-analysis applies a strict foreground/background separation: only foreground unit processes (substrate preparation, deposition/coating, drying/annealing, metallization, encapsulation) are harmonized at process level, while background flows (electricity, gases, commodity chemicals, generic materials and transport) are mapped to named Ecoinvent/GaBi datasets via the Appendix A BACKGROUND_LINKS table, including dataset identity, region, year and version.  Also, we added a concrete example in Section 4.1 illustrating how aggregated background proxies can misrepresent a real PV process. Specifically, using an aggregated generic dataset such as glass, flat, generic construction as a proxy for solar-grade low-iron glass can embed non-representative furnace efficiencies, recycled cullet shares and transport assumptions relative to PV-glass supply chains. When applied to glass-glass modules, this substitution can bias the apparent GWP contribution of glass upward by approximately 10-20%, potentially masking real improvements due to furnace ecodesign and higher cullet shares in dedicated PV-glass production.

- A new sentence has been added in Appendix A.1, in bullet ‘BACKGROUND_LINKS.

-We added a new warning paragraph in Section 7 explaining that mixing aggregated composite module datasets (e.g., multi-crystalline Si module, Europe) with process-resolved LCIs for perovskite/tandem routes can introduce hidden double counting and inconsistent cut-offs. To make these risks visible and avoid them, we restrict aggregated datasets to background flows only and require transparent reporting of background dataset identity, region, year and version.

-Table 4 (previous Table 2) was updated with an explicit reporting requirement that background dataset identity, region/year and database version must be reported for all linked background flows, ensuring that any proxy choices and their potential biases are transparent. 

Comments 6: Recycling section discusses multiple pathways but lacks a clear statement about economic practicability. Can you add a brief comparison of whether high-yield Pb/Ag recovery is realistic at current market values?

Response 6: We thank the reviewer for raising the important issue of economic practicability of recycling pathways. We have added a concise economic perspective to Sections 6.1-6.3 to complement the environmental analysis.

-A new paragraph has been added in the end of Section 6.1. For crystalline-silicon modules, we now clarify that high-yield silver recovery (≥90%) is economically favourable at current silver prices when recycling is integrated into high-throughput industrial lines, consistent with LCA-LCC studies and reported industrial trials. In contrast, silicon recovery is shown to be primarily constrained by the energy intensity of purification rather than intrinsic material value, with economic viability depending strongly on regional electricity and labour costs.

-A new sentence has been added in Section 6.2, as follows: ‘These findings underline that economic performance of PV recycling is highly sensitive to scale, process integration, and energy costs, reinforcing the need to assess recycling pathways jointly through environmental and economic lenses rather than on recovery efficiency alone [30,54,72].’

- A new paragraph has been added in Section 6.3. For perovskite and tandem modules, we explicitly state that hydrometallurgical Pb and Ag recovery can be economically viable only when recovery efficiencies exceed approximately 85-90% and when sufficient waste volumes are available. We further clarify that, based on published cost–revenue balances, such recovery routes are unlikely to be profitable at very low collection volumes or when recovery efficiencies fall below ~80-85%.

These additions directly address the economic practicability of high-yield Pb/Ag recovery while remaining within the scope of a review article.

Comments 7: The article proposes standard reporting rules (Table 2). Could you include a short worked example that demonstrates how results change when FU or system boundary is altered? This would give practical value to the guidelines.

Response 7: We thank the reviewer for this constructive suggestion. To explicitly demonstrate the practical impact of the proposed reporting rules, we have added a short worked example in Section 7, supported by Appendix A.3.

Using the pre-industrial perovskite module case already documented in Appendix A, we now recalculate impacts under three alternative reporting choices: (i) FU of 1 m² active area under cradle-to-gate boundaries; (ii) FU of 1 kWp under cradle-to-gate boundaries; and (iii) FU of 1 kWp under cradle-to-grave boundaries including closed-loop recycling.

The results, summarized in the newly added Table 5, show that switching from an area-based to a power-based FU, without changing any process data, can shift reported GWP by a factor of approximately 1.5-2, depending on assumed efficiency and active-area ratios. Extending the system boundary to cradle-to-grave and including high-yield recycling reduces the cradle-to-grave GWP by approximately 40-50% for the same technology.

This worked example is explicitly referenced in the caption of Table 4 (previous Table 2) and demonstrates how inconsistent application of functional units and system boundaries can drive apparent discrepancies between LCAs, thereby reinforcing the practical value of the proposed reporting framework.

Comments 8: The manuscript mentions substitution strategies for critical materials. Can you discuss risks of performance loss when replacing Ag with Cu or Al? Are these substitutions feasible in industrial lines without degradation?

Response 8: We thank the reviewer for raising this important point regarding the practical implications of critical-material substitution. In the revised manuscript, we have expanded Section 6.3 to explicitly address both performance risks and industrial feasibility associated with replacing Ag with Cu or Al.

Specifically, we now summarize the main technical risks of Ag→Cu/Al substitution:

-Higher contact resistivity and increased series resistance, particularly at fine finger widths, with potential impacts on fill factor and efficiency.

-Corrosion and electromigration risks for Cu contacts, requiring diffusion barriers and controlled plating schemes.

-Adhesion and solderability challenges for Al-based pastes on certain TCOs, with implications for long-term reliability.

We also directly address feasibility at industrial scale. The revised text clarifies that partial Ag reduction strategies, including narrower fingers, multi-busbar designs, hybrid Cu/Ag grids, and low-Ag paste formulations, are already industrially demonstrated, reducing Ag intensity from approximately 80-100 mg/W to 20-40 mg/W with negligible performance penalties.

In contrast, full Ag replacement by Cu or Al, while technologically promising, is not yet universally drop-in for existing production lines. It often requires additional processing steps (plating or diffusion barriers), modified firing or curing profiles, and extensive reliability testing, which can limit near-term deployment.

Finally, we emphasize that in our scenario analysis (Section 5.2), reduced-Ag metallization consistently emerges as one of the most effective levers for lowering cradle-to-gate impacts and improving circularity, but must be evaluated jointly with performance and reliability constraints rather than as a purely material substitution.

Comments 9: In many places, the review quotes numerical ranges from other studies. Would a summary comparison table of environmental impacts (GWP, EPBT, recyclability) across technology types help readers understand trade-offs more clearly?

Response 9: We agree that a summary comparison table would help readers interpret the numerical ranges reported across studies and understand trade-offs between PV technology classes more clearly. In the revised manuscript, we therefore added Table 2 (Comparative environmental indicators across PV technology classes) in Section 5.3.

Table 2 provides a structured cross-technology overview for c-Si, CdTe, CIGS, single-junction perovskite, perovskite/Si tandem, and organic solar cells (OSC), summarizing:

-Indicative cradle-to-gate GWP values (reported as ranges, expressed per kWp),

-Typical EPBT ranges (years),

-Qualitative recyclability maturity (industrial / pilot / lab-scale),

-Dominant critical materials, and

-Current substitution and recycling options.

To ensure transparency, we also added a short note below the table clarifying key methodological caveats: values are compiled as indicative ranges from the literature and remain sensitive to functional unit choice, system boundaries, electricity mix, efficiency, lifetime assumptions, and allocation/crediting approaches.

Comments 10: The article concludes that circularity could reduce GHG by 40-60%. What specific recycling efficiencies (for Si, Ag, glass) are needed to reach those reductions? A threshold value table would improve clarity.

Response 10: We thank the reviewer for this important suggestion. In the revised manuscript, we have made the reported 40-60% GHG reduction potential explicitly quantitative by linking it to material-specific recovery-rate thresholds derived from the end-of-life LCAs reviewed.

Specifically, we added Table 3 (‘Indicative recovery-rate thresholds for achieving 40-60% GHG reduction’), which summarizes recovery-efficiency bands for silicon, silver, glass, aluminum, and lead for two representative cases: c-Si and perovskite/Si tandem modules. These thresholds reflect reported values under closed-loop or high-value recycling assumptions using avoided-production or system-expansion credits.

The revised text in Sections 6.1-6.2 and the Conclusions now clarifies that cradle-to-grave GHG reductions of approximately 40-60% are achieved only when high-yield recovery of Ag (≈90-95%), Si (≈80-90%), and glass (≈85-90%) is combined with sufficiently large and stable recycling streams. 

Comments on the Quality of English Language

We thank the reviewer for this comment. The manuscript has been carefully checked and revised throughout to improve the quality, clarity, and consistency of the English language. Minor grammatical issues, phrasing, and stylistic inconsistencies have been corrected accordingly.

Response to Reviewer 2 Comments

1. Summary

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted in the re-submitted file.

Comments 1: This research uses the LCA approach to investigate the environmental value proposition of an emerging solar PV technology. Further, it suggests methodology for open-access templates for techno-economic modelling. The arguments and methodological approach of this paper are cogent and contribute to the scientific literature on solar PV technologies transitioning from lab scale to industrial scale. I do not have major comments for this paper.

Response 1: We thank the reviewer for the positive evaluation of the manuscript and for the constructive suggestions to further improve clarity and balance. The comments have helped us refine both the abstract and the treatment of EoL aspects for emerging PV technologies.

Comments 2: However, the authors may want to improve the readability of the abstract, as the study's motive appears at the backend of the abstract.

Response 2: We thank the reviewer for this helpful suggestion. The abstract has been revised to improve readability and narrative flow by bringing the study’s motivation and contribution forward.

Specifically, the revised abstract now:

• Introduces the motivation (the lack of harmonized, process-level LCIs for emerging PV manufacturing routes) in the opening sentences;
• Clearly states the gap addressed by the review before detailing the methodological contributions;
• Improves sentence structure and clarity while maintaining technical accuracy.

These revisions improve accessibility for a broader readership and better align the abstract with the scope and objectives of the study.

Comments 3: Also, we do understand the end-of-life implications of the silicon-based PVs, but it is perhaps less understood for emerging solar PV and the corresponding materials. So, it would be worth highlighting the end-of-life management of the emerging solar PV technologies and materials.

Response 3: We fully agree with this observation. This point has been explicitly addressed and substantially expanded in the revised manuscript, primarily in response to Reviewer 1, and now directly responds to the concern raised here.

In particular:

-Section 6.1 now explicitly contrasts mature EoL pathways for crystalline-silicon PV with the more conditional and emerging EoL strategies for perovskite, tandem, and organic PV technologies.

-Section 6.2 discusses allocation, substitution assumptions, and credits for recovered materials (Pb, Ag, Au, TCO elements) in emerging PVs, highlighting uncertainties related to material quality and recovery scale.

-Section 6.3 provides a dedicated discussion on material criticality and substitution for emerging PV technologies, including Pb, Ag, In, Sn, and Au, and explicitly addresses economic and technical feasibility of recovery.

-Table 3 was added to quantify recovery-efficiency thresholds required to achieve meaningful (≈40-60%) cradle-to-grave GHG reductions for both crystalline-Si and perovskite/Si tandem modules.

Additional text clarifies that for emerging PV technologies, EoL benefits are strongly conditional on recovery efficiency (>85-90%), scale, and system integration, and are not yet guaranteed at low collection volumes or laboratory-scale processes.

These additions ensure that the manuscript now explicitly highlights:

-The relative maturity gap between silicon-based and emerging PV EoL systems;

-The material- and technology-specific challenges of EoL management for next-generation PV;

-The importance of integrating EoL considerations early in technology development.

Response to Reviewer 3 Comments

1. Summary

We sincerely thank Reviewer 3 for their detailed and constructive feedback. The comments greatly helped clarify the conceptual structure and analytical depth of the review. Below we provide a point-by-point response, indicating how each concern has been addressed in the revised manuscript. Where material had already been strengthened in response to Reviewer 1, 2 and 3, we indicate this explicitly. Revisions/corrections highlighted in the re-submitted file.

Comment: This paper provides a systematic review of life cycle assessment (LCA) studies on the manufacturing processes of next-generation photovoltaic technologies (e.g., perovskite, roll-to-roll, and tandem devices). It focuses on collecting and integrating process-level life cycle inventory (LCI) data, analyzing changes in environmental impacts from lab-scale to industrial-scale production, discussing the effects of recycling and critical materials, and proposing a standardized LCI reporting template and methodological guidelines to improve comparability and reproducibility in future research. The literature search strategy is clear, the screening process is transparent, and the research scope is comprehensive. The following suggestions are provided for consideration:

Comment 1: The literature review tends to focus on technical descriptions without critically addressing contradictory conclusions in existing LCAs. It is recommended to include an analysis of key controversies or conflicting findings.

Response 1: We thank the reviewer for his/her comment. This comment is fully addressed in the revised manuscript, following similar feedback from Reviewer 1.

Specifically, we have:

• Added a new subsection Section 3.4 Cross-study contradictions and gaps, which explicitly analyzes conflicting conclusions across LCAs.
• Identified and discussed major sources of disagreement, including:
• Functional unit choice (1 m² vs. 1 kWp vs. kWh-delivered),
• System boundary definitions (cradle-to-gate vs. cradle-to-grave),
• Electricity-mix assumptions and manufacturing geography.

Added synthesis paragraphs at the end of Sections 4.1 and 5.3, explicitly highlighting where LCAs disagree most and why.

These additions move the review beyond technical description toward a critical evaluation of methodological conflicts in the literature.

Comment 2: It is suggested to supplement the literature screening section with a PRISMA flow diagram (the original text mentions Figure 1, but does not clearly present the complete screening path). This diagram should detail the specific reasons for exclusion, showing the distribution from the initial 1824 records to the final 102 included studies.

Response 2: We thank the reviewer for this suggestion. The manuscript already includes a transparent screening description in Section 2 and references Figure 1, which summarizes the search and selection process.

To address this comment more explicitly, we have:

-Revised the caption and description of Figure 1 to clarify that it functions as a PRISMA-style flow diagram.

-We added two new paragraphs in Section 2.1. Explicitly stated the number of records at each screening stage and the main exclusion criteria (topic mismatch, waste-only focus, single-scenario studies).

Comments 3: Using "1 m² of active area" as the functional unit does not account for differences in technology efficiency, potentially leading to bias in cross-technology comparisons. It is recommended to supplement this with "1 kWp" as an additional parallel functional unit.

Response 3: We thank the reviewer. This concern has been explicitly addressed in response to Reviewer 1.

In the revised manuscript:

Both functional units are now used in parallel:

1 m² active area (for process harmonization and lab-to-fab scaling) and

1 kWp (for cross-technology comparison).

Section 4.1 explains the rationale for dual-FU reporting.

A worked example in Section 7 and a new Table 5 quantitatively demonstrates how switching from 1 m² to 1 kWp changes reported GWP by a factor of ~1.5-2, without altering process data.

This dual-FU approach directly addresses efficiency-related bias while preserving process-level transparency.

Comments 4: After categorizing LCIs by process family in Section 4.2, the paper does not sufficiently explain the mechanisms behind variations among different technologies within the same family, which could cause misunderstanding.

Response 4: We agree and have strengthened the discussion accordingly.

In Section 4.2 a new paragraph was added ‘Within each process family…’, and we now explicitly explain intra-family variability by linking differences to:

• Throughput and duty-cycle assumptions,
• Solvent recovery efficiency (solution-based routes),
• Chamber utilization, idle power, and heating profiles (vacuum-based routes),
• Encapsulation mass and substrate design (hybrid/tandem routes).

Additionally, in Section 4.3, a new paragraph was added ‘The data-quality tier framework clarifies…’, introduces data-quality tiers (Tier 1-3) and uncertainty characterization, clarifying that apparent variability within a process family often reflects differences in scale, TRL, and data representativeness rather than intrinsic technology effects.

Comments 5: It is recommended to add a brief synthesis and commentary at the end of each main section, summarizing limitations and outlining future development trends.

Response 5: We thank the reviewer for this comment. This recommendation has been implemented in the revised manuscript, following similar feedback raised by Reviewer 1.

Specifically, short synthesis paragraphs have been explicitly added at the end of selected core sections to summarize key limitations, unresolved methodological issues, and directions for future research:

Section 3.4 now concludes with a synthesis of technology- and process-level gaps, highlighting contradictions and missing inventories across emerging PV routes.

Section 4.1 ends with a summary emphasizing how inconsistencies in functional-unit definition, system boundaries, and electricity-mix assumptions drive divergence among published LCAs.

Section 5.3 now includes a concluding paragraph identifying the main sources of disagreement in comparative LCAs and clarifying conditions under which technology rankings become stable.

Section 6.2 concludes with a synthesis of unresolved issues in end-of-life modeling, including allocation choices, aggregation bias, and data sparsity for emerging technologies.

These targeted additions improve readability and ensure that each major section closes with a clear critical perspective rather than a purely descriptive summary.

Comments 6: A dedicated section discussing the "Limitations of the Template" should be added. This could cover, for example, methods for data interpolation in cases of missing information, compatibility with different LCA software, etc. Furthermore, the discussion on research limitations should specify their manifestations rather than merely stating that "uncertainties exist."

Response 6: We thank the reviewer for this insightful suggestion. Rather than introducing a separate standalone section, we integrated a dedicated limitations discussion into:

Section 7 (a new paragraph has been added ‘While the proposed reporting checklist and LCI template…’, where limitations of the proposed reporting framework are explicitly discussed, including:

• Handling of missing or incomplete process data,
• Use of scaling and interpolation assumptions,
• Dependence on background database versions,
• Compatibility with common LCA software (openLCA, Brightway, SimaPro).

Additionally, Appendix A.2 was substantially expanded and clarified to provide a detailed, process-specific discussion of priority data gaps and measurement needs, explicitly describing their impact on uncertainty, comparability, and forward-looking assessments.

Comments 7: If feasible, exploring how data varies across different geographical regions and manufacturing conditions would improve the generalizability of the results.

Response 7: We thank the reviewer for his/her comment. This aspect has already been addressed, explicitly and quantitatively in the revised manuscript, following the comments of previous reviewers

Geographical and manufacturing-context variability is discussed in:

Section 4.1 (Normalization rules), where electricity-mix assumptions (region and year) are shown to account for ~40–80% of cradle-to-gate GWP variability, with reported 2-4× differences when identical manufacturing processes are evaluated under coal-dominated versus low-carbon grids.

Section 5.2 (Parameterized scenarios), which evaluates identical process LCIs under alternative regional electricity systems and decarbonization trajectories, including EU-27 baseline (2023) and 2030/2050 projections, explicitly linking manufacturing impacts to geographic electricity supply.

Section 5.3 (Comparative cradle-to-gate results), which synthesizes multi-regional case studies for China, Italy, Burkina Faso, and hydropower-linked silicon supply chains, demonstrating how manufacturing location and upstream routing can reverse technology rankings.

Appendix A (Electricity scenarios and background datasets), where region-specific grid mixes, dataset identities, and years are explicitly documented to enable reproducible geographic sensitivity analysis.

Together, these additions ensure that regional electricity mix, manufacturing geography, and supply-chain configuration are treated as first-order variables rather than implicit assumptions, thereby substantially improving the generalizability of the results.

Comments 8: The resolution of the figures is somewhat low.

Response 8:All figures in the initial submission and now were enhanced to >300 dpi and upgraded in a multilevel; way prior to the first submission. Therefore no actual figure has low dpi. Probably it was due to embedding the figures within the DOT version.

Response to Reviewer 4 Comments

1. Summary

We sincerely thank Reviewer 4 for their detailed and constructive feedback. The comments greatly helped clarify the conceptual structure and analytical depth of the review. Below we provide a point-by-point response, indicating how each concern has been addressed in the revised manuscript. Where material had already been strengthened in response to Reviewer 1, 2 and 3 we indicate this explicitly. Revisions/corrections highlighted in the re-submitted file.

Comment: Overall, the manuscript is well-written and technically solid, and it tackles an important topic. However, some issues related to positioning, methodology transparency, and the quantitative use of the compiled data should be addressed before the paper is ready for publication.

Comment 1: The paper fact that no existing review systematically compiles and harmonizes process-level LCIs across emerging PV routes should be better demonstrated. A concise comparative table contrasting this review with prior PV LCA meta analyses is missing. This makes it hard to see precisely what is new. Is it the explicit process-family grouping, the scale-up treatment, the uncertainty tiering, the LCI template, other? Clarifying and evidencing the specific novelty would considerably strengthen the contribution.

Response 1: We thank the reviewer for this valuable comment and agree that the specific novelty of the present review should be articulated more explicitly. In the revised manuscript, we have therefore strengthened the positioning of this work relative to prior PV LCA reviews directly in the text.

Specifically, we clarify that while several previous reviews synthesize environmental indicators or compare PV technologies at the module or system level, they typically rely on aggregated datasets and do not systematically compile, normalize, and analyze process-resolved life-cycle inventories (LCIs) across emerging manufacturing routes. We now explicitly state that the key contributions of this review lie in (i) the harmonized compilation of step-level LCIs across solution-based, vacuum-based, and hybrid manufacturing families; (ii) explicit treatment of lab-to-pilot-to-industrial scale-up effects using data-quality tiers; (iii) quantitative uncertainty characterization and TRL-weighted sensitivity testing; and (iv) the provision of a standardized, open-access reporting template supported by a worked example.

These clarifications are added in the Introduction (a new paragraph in the end) and reiterated at the start of Section 4, making clear that the novelty of the work is not a new LCA result per se, but a process-level, scale-aware, and uncertainty-explicit synthesis framework that is currently missing from the PV LCA literature.

Comment 2: The search string, databases and hit counts are reported clearly, but the transition from the original article amount records to the chosen articles should be better explained. The inclusion/exclusion criteria seem too much generic and do not explicitly ensure that process-level LCIs are present. Consider adding a PRISMA-style table listing key criteria and some examples of borderline cases.

Response 2: We thank the reviewer for this comment. This issue was addressed following Reviewer 3’s feedback by clarifying the inclusion and exclusion criteria directly in the text, rather than by adding a separate PRISMA criteria table.

Specifically, Section 2.1 (Literature search strategy) was revised to explicitly state that studies were excluded if they:

• lacked explicit process-level life cycle inventory data (i.e., energy, material, or emission flows reported per manufacturing step),
• relied solely on aggregated module-level datasets without transparent unit-process inventories, or
• addressed only a single end-of-life scenario without comparative LCA or sufficient methodological transparency.

These criteria ensure that only studies providing traceable, process-resolved LCIs suitable for harmonization were retained in the final dataset.

Regarding the screening visualization, Figure 1 was clarified as a PRISMA-style flow diagram, and the numerical progression from 1824 initial records to 102 included studies, together with the main exclusion reasons, is now explicitly described in the accompanying text in Section 2.1. While a separate PRISMA checklist table was not added, the revised text provides the necessary transparency on selection logic and exclusion rationale.

Comments 3: The paper sets out normalization rules (FU, system boundaries, electricity mix), but the practical implementation of these rules across heterogeneous studies remains somewhat not investigated. For example, it is not always clear when original FUs (kWp, kWh, module) were converted, how different lifetime assumptions were treated, or how overlapping datasets were handled to avoid double counting.

Response 3: We thank the reviewer for his/her comment. This issue has already been explicitly addressed and operationalized in the revised manuscript, following the comments of Reviewer 1, 2 and 3.

Specifically:

-Section 4.1 (Normalization rules)

-Section 7 + new added Table 5 (Worked example)

-Appendix A (BACKGROUND_LINKS + FU conversion formulas)

Key explicit implementations already present

-Dual FU (1 m² + 1 kWp)

-Explicit conversion assumptions (efficiency, active-area ratio)

-Lifetime consistency handled via kWh-delivered discussion

-Foreground/background separation prevents double counting

Rather than remaining conceptual, the normalization rules are implemented quantitatively through a worked example (Table 5), where identical process LCIs are recalculated under different FUs and boundaries, explicitly demonstrating how conversions and boundary extensions alter results.

Comments 4: The impact of uncertainty treatment on the results is only qualitatively discussed. The reader sees few concrete uncertainty ranges on the synthesized indicators. Consider adding one or two key quantitative uncertainty plots to show how robust the comparative conclusions really are.

Response 4: We thank the reviewer. This has already been quantitatively addressed, and the reviewer’s concern is understandable because uncertainty information is distributed across sections.

• Section 4.3 (Monte Carlo uncertainty propagation)
• Table 1 (CoV ranges)
• Figures 3 and 4 (error bars)

Exact quantitative content already present

• 10⁴ Monte Carlo runs
• CoV = 20-60% (lab), <15% (industrial)
• Explicit Tier-based uncertainty bands

Comments 5: A small summary table listing, for each considered family, the number of studies, TRL range, and main technology types would make the analysis more transparent and help readers interpret the averaged intensity values reported in the study.

Response 5: We agree that readers benefit from seeing how many studies underpin each process-family synthesis and what TRL span those studies cover. To avoid adding an additional table (given the new Tables 2, 3 and 5 added during revision), we instead report this information directly in the text and figure captions. Specifically, in Section 4.2 we now state, for each process family (solution / vacuum / hybrid), the number of included studies contributing LCIs and the observed TRL span. We also added the corresponding ‘n’ values to the caption of Figure 3 (process-family comparison), so that the evidentiary basis of the averaged intensities is visible where the results are interpreted.

Comments 6: Numerical examples that link laboratory LCIs to pilot/industrial scenarios using explicit learning-curve parameters would be very valuable. One or two simple plots showing how energy/GWP intensity evolves with cumulative capacity under specific assumptions would make Section 5 more actionable for techno-economic and prospective LCA practitioners.

Response 6: We thank the reviewer for this valuable suggestion regarding the quantitative linkage between laboratory LCIs and pilot/industrial scenarios.

This aspect has been partially addressed in the revised manuscript following the comments of the previous reviewer, and we clarify how it is handled and why we did not introduce additional learning-curve plots.

Section 5.1 (Scaling from laboratory to pilot and industrial production)

We now explicitly quantify scale-up effects using reported learning rates (≈15-25% reduction per cumulative doubling of production) and link these reductions to concrete operational mechanisms such as throughput increases, duty-cycle stabilization, and yield improvements. The text clarifies which process steps exhibit large learning effects (e.g., solution coating, annealing) and which do not (material-dominated or already optimized steps).

Section 5.2 (Parameterized scenarios)

Scenario-based modeling explicitly evaluates how energy and GWP intensities change under alternative scale and electricity assumptions. These scenarios implicitly embed learning-curve effects by combining:

Increased throughput and yield,

Reduced idle power,

Declining electricity carbon intensity (EU-27 2023  2030  2050).

This provides a quantitative bridge between laboratory and industrial contexts without relying on a single assumed global learning parameter.

Section 7 and new added Table 5 (Worked example)

A worked numerical example recalculates identical foreground LCIs under alternative functional units and system boundaries. While not a capacity-based learning curve, this example demonstrates how reported impacts evolve under realistic scale-related assumptions that techno-economic and prospective LCA practitioners routinely apply.

Learning-curve formulations and parameter values vary substantially across technologies, process steps, and industrial contexts. Introducing one or two stylized curves would risk:

Over-generalizing process-specific learning behavior,

Implicitly privileging a single learning model,

Suggesting predictive precision beyond what current empirical data support.

Instead, we opted for a process-resolved, scenario-based approach that preserves transparency and avoids speculative extrapolation, while remaining directly usable for prospective and techno-economic LCA.

We believe this approach strikes a balance between quantitative rigor and methodological caution, and we have clarified this positioning in:

• Sections 5.1 (revised 2^nd paragraph and addition of a new 3^rd paragraph),
• Section 5.2 (3 revised paragraphs)
• Section 7 ( a new paragraph has been added, 5^th in the revised manuscript).

Comments 7: Sections 6.1-6.3 provide a rich overview of EoL, logistics, allocation and critical-materials issues, but the link back to the manufactured process LCIs is sometimes weak.

Response 7: We thank the reviewer for this thoughtful comment. Strengthening the connection between EoL considerations and manufactured process-level LCIs was a key objective of the revised manuscript, and this linkage is now explicitly addressed across Sections 6.1-6.3 in close coordination with Sections 4 and 5.

Specifically, the revised manuscript links EoL outcomes back to manufacturing LCIs in the following ways:

• Section 6.1 (EoL pathways, recycling and logistics) explicitly quantifies cradle-to-grave GHG reductions (40-60%) and ties these outcomes to manufacturing-stage material intensities for silver, silicon, glass, and aluminum, which are derived from the process-level LCIs compiled in Section 4.2. Recovery-rate thresholds discussed in Section 6.1 therefore directly reflect the foreground material flows of the manufactured modules.
• Section 6.2 (Allocation, credits and EoL accounting) connects allocation and system-expansion choices at EoL to foreground inventory structure, highlighting how mixing aggregated and process-resolved inventories can bias both manufacturing and recycling results. This section explicitly builds on the foreground/background separation rules defined in Section 4.1 and operationalized in the reporting framework (Section 7).
• Section 6.3 (Critical materials, scarcity and substitution) links critical-material risks and substitution strategies (e.g., Ag reduction, Cu/Al metallization) back to manufacturing process choices and intensities, which are quantified in Sections 4.2 and 5.2. The discussion explicitly notes that material-criticality benefits and recycling feasibility depend on the mass flows and process configurations defined at the manufacturing stage.
• In addition, the new added Table 3 (recovery-rate thresholds) and the synthesis paragraphs at the end of Sections 6.2 and 6.3 explicitly reinforce that circularity benefits and critical-material mitigation are conditional on the manufactured process LCIs, rather than independent downstream assumptions.

Sections 6.1-6.3 are intentionally structured to interpret EoL, allocation, and critical-material outcomes as extensions of the manufacturing inventories, rather than as standalone analyses. We believe the revised text now makes this linkage explicit and traceable, while avoiding redundancy with the detailed process-LCI discussions in Sections 4 and 5.