Role of anthropogenic mineral circularity in addressing dual challenges of resource supply and waste management in global photovoltaic development

Abstract

Anthropogenic mineral circularity offers a synergistic solution to dual challenges of resource supply and waste management in photovoltaic development. However, the global supply potential of secondary materials remains underexplored, limiting future informed decision-making. Here, we present the global analysis of secondary material supply potential in five photovoltaic technologies under different energy scenarios, using the tailored modeling framework. Results show that cumulative material demand and waste generation are projected to reach 705–1879 megatonnes and 238–529 megatonnes, respectively, by 2050. With the circularity strategy, the annual supply ratio of secondary materials is expected to increase from 3.3% in 2020 to 43.4%–101.6% by 2050, with silver and tellurium potentially in surplus. Additionally, circularity brings 6.6%–55.0% decrease in metal criticality, 321–700 billion US Dollars economic potential, and 697.0–1546.1 megatonnes CO₂-equivalent emissions reductions. This study quantifies anthropogenic mineral circularity’s role in photovoltaic development and provides insights for energy transition.

Introduction

Solar photovoltaic (PV) energy is widely acknowledged as a clean and abundant energy source. Its global capacity has surged from 1.2 Gigawatt (GW) in 2000 to 1053.1 GW in 2022—an over 800-fold increase^1,2. PV capacity is projected to continue its upward trajectory, with the share of electricity generated from solar energy expected to surge from 5.5% in 2023 to become the world’s predominant source of electricity by 2050³. This rapid expansion will inevitably drive both the construction and eventual decommissioning of related infrastructure, thereby intensifying concerns over mineral resource demands and waste management challenges⁴. On the one hand, the depletion of natural minerals, combined with increasing material needs for PV deployments, threatens the sustainability of material supply⁵. On the other hand, unmanaged PV waste risks environmental damage⁶. These challenges, in turn, may constrain the development of the PV industry.

Anthropogenic mineral circularity from PV waste offers a synergistic solution to simultaneously mitigate virgin material demand and address environmental concerns. (Supplementary Fig. 6)⁷. While recycling technologies are developing, the potential of anthropogenic minerals circularity to meet material demands remains underexplored, which is a crucial step toward informed and predictive decision-making on future resource supply and waste management^8,9. Meanwhile, improvements in PV technologies and conversion efficiency drive changes in market share, material composition, and weight-to-power ratios, thus resulting in dynamic shifts in material demand and waste generation that complicate supply potential assessments¹⁰. To clarify the supply potential of anthropogenic minerals circularity in mitigating the demand for natural resources, it is essential to determine the material demand for new PV deployments and the anticipated generation of PV waste. Consequently, many studies have investigated these aspects. For instance, research has been conducted on critical materials for major low-carbon technologies¹¹, metal demand for electrical grid systems in wind power and PV systems¹², material demand for global silicon-based PV modules¹³, resource demands for multi-terawatt-scale perovskite PV tandem¹⁴, and metal production demand for rapid PV deployment¹⁵. Additionally, PV waste generation has been estimated in various countries, including Spain¹⁶, India¹⁷, Mexico¹⁸, Italy¹⁹, Australia²⁰, China^21,22, and Organization for Economic Co-operation and Development (OECD) nations²³.

However, much of the existing research has primarily focused on either material demand or waste generation, with few studies integrating both to assess the supply potential of anthropogenic minerals^13,24. Furthermore, most of these studies are geographically confined to the national level, with few addressing the issue from a global perspective. In terms of scope, the research has focused narrowly on PV modules and overlooked Balance of System (BOS) components, which can lead to underestimated findings^20,25. Some studies also restrict their analysis to a select few materials, without considering the full range of relevant materials²⁶. From a methodological standpoint, projections of future PV deployment have typically relied on gray system theory or machine learning approaches, which may inadequately capture the influence of evolving policy frameworks^22,24. Although policy roadmaps are considered in some studies, they are frequently outdated and not aligned with current plans¹⁵. In addition, clarifying the benefits of anthropogenic mineral circularity, including metal criticality mitigation, economic potential, and carbon reduction, is essential to motivate stakeholders throughout the PV value chain to embrace recycling practices.

To bridge these gaps, we evaluate the global potential of secondary materials supply in alleviating virgin materials demand under three energy scenarios developed by the International Energy Agency (IEA) up to 2050. Initially, material demand for PV installations and waste generation from decommissioned systems are quantified using a stock-based dynamic material flow analysis (MFA) model. Subsequently, the supply potential of secondary materials and the net material demand are assessed, factoring in current recycling rates for all embedded materials. Additionally, we calculate the benefits in metal criticality, carbon emissions, and economic potential from anthropogenic mineral circularity. Finally, sensitivity analysis is conducted not only to evaluate the robustness of the results but also to identify potential optimization pathways for further reducing the material demand. This analysis quantifies the role of anthropogenic mineral circularity in simultaneously addressing resource supply challenges and bulk PV waste, offering insights for PV development and the energy transition.

Results and discussion

Dynamic material flows analysis

We conducted a dynamic MFA based on the framework in Fig. 1. Using historical annual installed capacities (Fig. 2a)²⁷ and future deployment from the IEA’s Stated Policies Scenario (SPS), Sustainable Development Scenario (SDS), and Net-Zero Emissions Scenario (NES), interpolated trends are presented in Fig. 2b. Then, combining with technology market shares (Supplementary Fig. 1) and lifetime distributions modeled using a Weibull function (Fig. 2c)²⁸, we estimate annual inflows and outflows for each PV technology.

Fig. 1: System boundary and model framework of this study.

a Schematic of the system boundary, covering five PV module (m-Si, p-Si, a-Si, CdTe, CIGS) and BOS components (mounting, cabling, inverters). b Model framework of this study: (1) dynamic materials flow analysis; (2) secondary material supply potential assessment; (3) assessment of metals criticality alleviation, economic potential, and carbon emission reduction benefits; and (4) sensitivity analysis of key parameters. m-Si monocrystalline silicon, p-Si polycrystalline silicon, a-Si amorphous silicon, CdTe cadmium telluride, CIGS copper indium gallium selenide, IEA International Energy Agency, GW/y gigawatt/year.

Fig. 2: Dynamic stocks and flows of global PV capacities.

a Global historical annual installed capacities by different types of PV technologies from 2000 to 2022. b Future possible global PV capacities stocks toward 2050 under three International Energy Agency’s energy scenarios (SPS, SDS, and NES). c Failure probability distribution of PV modules under different loss scenarios. d–f Newly installed capacities for PV power under SPS, SDS, NES in the early-loss scenario (the newly installed capacities in the regular-loss scenario are provided in Supplementary Fig. 8a–c). g, h Annual decommissioned capacities under the SDS for early-loss and regular-loss scenario (The decommissioned capacities under the SPS and NES are provided in Supplementary Fig. 9). i Cumulative decommissioned capacities under three energy scenarios (shaded areas represent the range between early-loss and regular-loss scenarios). The three energy scenarios are defined as follows: SPS Stated Policies Scenario, SDS Sustainable Development Scenario, NES Net-Zero Emissions Scenario. The five photovoltaic (PV) technologies include: m-Si monocrystalline silicon, p-Si polycrystalline silicon, a-Si amorphous silicon, CdTe cadmium telluride, CIGS copper indium gallium selenide, GW denotes gigawatt.

As Fig. 2d–f, inflows under SPS and SDS grow steadily, reaching ~255 GW and ~656 GW by 2050, respectively. The difference reflects the different policy ambitions¹², where SPS aligns with current commitments, while SDS assumes more aggressive global efforts toward sustainable development. In contrast, NES shows a peak of ~700 GW in 2035, followed by a decline to ~500 GW by 2050. This front-loaded pattern reflects urgent decarbonization needs to limit warming to 1.5 °C¹, requiring rapid PV expansion in the 2020s–2030 s. The latter decline results from market saturation, as cumulative capacity approaches the level needed to meet net-zero electricity demand. We also note the sharp difference between 2022 and 2023 data, especially under the SPS. This is because 2022 reflects actual data, while 2023 is based on scenario interpolation. To validate the model, we compared the projected 2023 value with the actual installed capacity (346 GW)²⁷, which closely matches the NES projection (350 GW), indicating consistency with recent deployment trends.

In addition to capacity expansion, replacements due to system failures also drive installations. To underscore the considerable potential of recycled materials in alleviating future resource demands, this study emphatically discusses the results of early-loss scenarios. This allows us to assess the optimistic outcome of recycling strategies. The differences between early-loss and regular-loss scenarios are further discussed in the following section to provide a more comprehensive perspective. Supplementary Fig. 8d–i provides detailed insights into the contributions of both replacement and expansion-driven installations. Under all three energy scenarios, outflow capacities exhibit exponential growth for both the early-scenario and regular-loss scenario (Fig. 2g–h, Supplementary Fig. 9).

By integrating capacity flows with dynamic weight-to-power ratios and evolving material intensities^28,29, we derive the corresponding material flows, as illustrated in Fig. 3. When translating installed capacities into material weight, a noticeable deceleration in the growth of total material inflows emerges (Fig. 3a). This slowdown is particularly evident under SPS, where annual material inflows maintain ~20 megatonnes (Mt) despite the continuous expansion of installed capacity. This trend is primarily driven by technological development in crystalline silicon (c-Si) PV cell architectures, transiting from aluminum back surface field to passivated emitter and rear contact and tunnel oxide passivated contact technologies^13,30. These efforts have yielded more efficient modules, thereby reducing the material demand per watt (Supplementary Fig. 2a). This progress not only reduces overall material demand but also underscores the need for continued development in recycling technologies to enhance material efficiency and enable design-for-recyclability. In contrast, material outflows continue to grow exponentially due to the decommissioning of earlier PV systems, and projects' annual outflows are expected to rise from 0.6 Mt in 2020 to 15.9 Mt, 25.1 Mt, and 44.8 Mt in 2050 for SPS, SDS, and NES, respectively. They also demonstrate that material demand and waste generation are highly dependent on future deployment scenarios, highlighting the importance of flexible, forward-looking recycling infrastructure planning²⁸.

Fig. 3: Corresponding materials demands (inflows) and decommissioned materials (outflows) associated with PV system under the early-loss scenario for three energy scenarios.

a Annual total materials inflows and outflows. b Total materials flow by different PV system components per period. c–e Materials flow per period of base materials, precious metals, and critical metals. Materials flow of hazardous metals (Pb, Cd), semiconductor and metalloid materials (Si, Se), and other metals (Mn, Ti, Mo, Cr, Ni, Sn, Mg, Zn) are provided in Supplementary Fig. 10. f Cumulative flow under three energy scenarios by 2050. g Proportion of cumulative outflows by material type by 2050. a–f Positive values represent inflows, while negative values represent outflows. The three energy scenarios are defined as follows: SPS Stated Policies Scenario, SDS Sustainable Development Scenario, and NES Net-Zero Emissions Scenario. The five photovoltaic (PV) modules include: m-Si monocrystalline silicon photovoltaic module, p-Si polycrystalline silicon photovoltaic module, a-Si amorphous silicon photovoltaic module, CdTe cadmium telluride photovoltaic module, and CIGS copper indium gallium selenide photovoltaic module. Mt denotes megatonnes.

As Fig. 3b, modules contribute 57% of system mass, with BOS components accounting for the rest. c-Si module accounts for ~80–90% of the module-related weight throughout the study period. Figure 3c–e and Supplementary Fig. 10 illustrate the material flows. Notably, the demand trajectories of precious and critical metals diverge from those of other material categories. In c-Si modules, Silver (Ag) demand has significantly declined due to improved screen-printing and narrower front-side metallization³¹. In thin-film technologies, improved deposition methods like high-precision sputtering and co-evaporation have reduced tellurium (Te), indium (In), gallium (Ga), and tin (Se) use by enabling thinner absorber layers and better material efficiency³². There is another interesting finding that Ag outflows are projected to level off after 2035, which is unlike the continued exponential rise in other materials. This trend reflects the sharp decline in manufactured Ag content since around 2010²⁹, resulting in less recycled Ag expected from modules reaching end-of-life after 2035. This calls for short-term policy incentives to scale up Ag recovery before the resource opportunity diminishes. As summarized in Fig. 3f, cumulative inflows by 2050 are projected to reach 706–1879 Mt inflows and 238–529 Mt outflows. Base materials and silicon (Si), which account for over 99% of total weight, should remain a key focus of recycling efforts. At the same time, targeted recycling technologies for high-value and high-impact materials such as Ag and Te should be prioritized, even though they represent only 0.04% of the total content.

Secondary material supply potential

Based on quantified material inflows and outflows, and integrating the recovery ratios (Supplementary Table 14) from current recycling technologies, Fig. 4 displays the supply ratio of secondary materials and the resulting net material demand after accounting for secondary supply. Under the SDS (Fig. 4a), net demand for most materials levels off or declines after 2035, highlighting the role of recycling in easing resource pressures. Results for SPS and NES are shown in Supplementary Fig. 11.

Fig. 4: Supply ratio of secondary materials and net material demand after secondary materials supply.

a Supply ratio of secondary materials and annual net material demand after secondary materials supply under early-loss scenario in SDS for different materials. In each panel, the line chart (left axis) represents the supply ratio (same scale across all materials except Ag and Te), while the bar chart (right axis) represents net material demand (with varying scales). b Supply ratio of secondary materials and annual net material demand under the early-loss scenario for total materials in the SPS and NES. c Supply ratios and net material demand under the regular-loss scenario for total materials in the SPS, SDS, and NES. The three energy scenarios are defined as follows: SPS stated policies scenario, SDS sustainable development scenario, NES Net-Zero Emissions Scenario. Mt denotes megatonne, and Kt denotes kiloton.

Unlike most materials, Ag and Te exhibit distinct temporal dynamics. For Ag, the secondary supply ratio surges and peaks between 2035 and 2040, temporarily exceeding demand and resulting in a state of oversupply. This is primarily attributed to the early-phase retirement of c-Si modules with high Ag content, coupled with relatively high recovery efficiencies. This near-term surplus could buffer against geopolitical supply risks, given the concentration of primary Ag production. It also presents an opportunity to strategically stockpile recovered Ag to hedge against future price volatility and supply disruptions. However, it should be noticed that recyclers could flood regional markets, temporarily depress Ag prices and disrupt mining economies. To avoid future shortages, PV manufacturers are advised to lock in access to recycled silver before the current surplus fades after 2040. In contrast, the supply ratio of Te steadily increases, supported by reduced material intensity and the end-of-life recovery from cadmium telluride (CdTe) modules. Surplus conditions are expected after 2045 across all scenarios, reflecting its longer-term surplus potential. Considering Te’s high criticality and limited primary supply sources, Closed-loop Te recovery reduces reliance on geopolitically concentrated primary supplies³³. CdTe manufacturers could even pivot from “module sellers” to “material suppliers”, selling excess Te to electronics sectors.

Base materials-like iron (Fe), copper (Cu), aluminum (Al), and glass, remain key recycling targets, but low recovery rates for polymers remain a significant barrier³⁴. While methods such as pyrolysis and solvent-based delamination are emerging^9,35, they still face trade-offs in terms of efficiency, cost, and environmental impact. Accelerating research and development in this area is essential to improve overall recycling yields and close material loops more comprehensively. Materials such as Si, Se, In, and Sn also demonstrate strong secondary supply potential due to anticipated reductions in material intensity and improved end-of-life recovery processes. In contrast, metals with low-content (manganese (Mn), titanium (Ti), and molybdenum (Mo), etc.) currently exhibit low supply ratios for low recovery rates, primarily due to their trace presence and the high costs associated with selective separation³⁶. Nevertheless, co-recycling strategies, where these minor elements are recovered alongside more abundant or valuable metals within integrated hydrometallurgical or pyrometallurgical processes, may offer a more practical solution.

Under different energy transition scenarios, we observe different trends in supply ratios and net material demand (Fig. 4b). By 2050, annual supply ratios reach 70.5% in SPS, 43.4% in SDS, and 101.6% in NES (Fig. 4b). Net material demand varies accordingly: decreasing to 6 Mt in SPS, peaking around 30 Mt in SDS by 2035 before stabilizing, and falling to near-zero levels by 2050 in NES. These outcomes highlight the growing importance of circular supply chains in offsetting primary demand, especially under aggressive decarbonization. Our comparison of regular-loss and early-loss scenarios (Fig. 4c) further reinforces the resilience of circular supply systems: while early-loss accelerates material outflows in the short-term, the long-term differences diminish, suggesting that end-of-life assumptions have a limited impact on mid-to-late century material security. This underlines the need for forward-looking investment in secondary material recovery, irrespective of short-term module failure patterns.

The impact of geospatial heterogeneity

We further analyze dynamic material flows and secondary material supply potential across six continents to assess regional disparities and geographical impacts. This continental-level resolution was chosen to balance model granularity with global coverage and data availability. While we acknowledge that national-level differences within each continent exist, our approach captures major regional trends and provides a scalable framework that can be refined with country-level data in future studies. The in-use PV capacities (stock) of the six continents are predicted by combining energy planning with polynomial fitting, as shown in Supplementary Fig. 12 (for the SPS), Supplementary Fig. 13 (for the SDS), and Supplementary Fig. 14 (for the NES). Following this, the calculation for each continent follows the global procedure, with results in Fig. 5, Supplementary Fig. 15, and Supplementary Fig. 16. Given our focus on geographical factors, we analyzed total material flows without differentiating material types.

Fig. 5: Annual and cumulative total materials flow, net demand, and the supply ratio of secondary material across six continents under early-loss scenario in the SDS.

a Annual inflow, outflow, net demand, and supply ratio of secondary materials across six continents. In each panel, inflows, outflows, and net demand are shown on the left vertical axis (with panel-specific scales), while the supply ratio is indicated by the line chart on the right vertical axis (with varying scales). b Cumulative material outflows by 2050 across six continents. Darker colors represent higher cumulative outflows. c Cumulative net material demand and supply ratios by 2050 across six continents. Darker colors indicate higher cumulative net demand, while larger circles represent higher cumulative supply ratios. SDS refers to the Sustainable Development Scenario. Mt denotes megaton. The details about the other two energy scenarios are provided in Supplementary Figs. 15 and  16.

Results highlight that Asia dominates both PV deployment and recycling needs, largely due to its high energy demand, extensive industrial base, and ambitious renewable energy targets, particularly in countries such as China and India^17,25. In contrast, Africa exhibits the lowest share in both inflows and outflows, reflecting its relatively underdeveloped PV market and limited deployment pace, which in turn constrains future recycling potential³⁷. Under different energy scenarios, regional net material demand follows divergent pathways. Under the SPS, net demand declines gradually in all regions, in alignment with the global trend of steady but moderate PV deployment and recycling development. Under the NES, all continents exhibit an initial rise followed by a decline in net demand, indicating an early surge in PV installation driven by aggressive decarbonization goals, followed by an increasing contribution of secondary material supply. The SDS presents a transitional trajectory, with Oceania and South America displaying a distinctive rise-then-fall pattern in net demand, while Asia, Europe, North America, and Africa largely follow the global trend. This unique pattern in Oceania and South America may be attributed to their rapid mid-term adoption of PV technologies driven by favorable policy shifts and international investment, followed by significant decommissioned output as their installed capacities mature²⁰.

Overall, these findings underscore the critical need for region-specific strategies for PV deployment and end-of-life management that consider infrastructure readiness, economic capacity, and deployment maturity. Asia must prioritize the development of large-scale and technologically mature recycling infrastructure to address its substantial future waste flows. Africa requires targeted investment and institutional support to improve both deployment and recycling capacity. Meanwhile, the rapidly increasing PV adoption in South America and Oceania highlights the urgency of establishing early-stage recycling systems in tandem with deployment efforts to avoid future waste bottlenecks and support circular material use. Moreover, future research should incorporate detailed nation-level assessments, particularly within Asia. Such granularity will be essential for designing actionable, region-appropriate strategies that accelerate the transition to a circular PV economy.

Multiscale benefits of anthropogenic mineral circularity

In this section, we assess the potential incentives for PV system recycling from three key perspectives: metal criticality alleviation, economic potential, and carbon emission reduction. These benefits may serve as important drivers for stakeholders across the PV value chain to adopt recycling practices. The corresponding results are presented in Fig. 6.

Fig. 6: Benefits of global secondary materials supply from decommissioned PV systems.

a Overall metal criticality comparison with and without recycling across the three energy scenarios. b Overall reduction in criticality due to recycling, compared to the no-recycling case, across the three energy scenarios. c Annual economic potential under three energy scenarios. d Annual economic potential of different PV materials under the SDS. e Annual carbon emission reduction potential across the three energy scenarios. f Annual carbon emission reductions attributed to different components under the SDS. c, e shaded areas represent the range between early-loss and regular-loss lifetime scenarios. The three energy scenarios are defined as follows: SPS stated policies scenario, SDS sustainable development scenario, NES net-zero emissions scenario. The five photovoltaic (PV) technologies include: m-Si monocrystalline silicon, p-Si polycrystalline silicon, a-Si amorphous silicon, CdTe cadmium telluride, and CIGS copper indium gallium selenide. Mt denotes megaton.

Based on the three-dimensional metal criticality assessment framework (Supplementary Fig. 17a), the results are displayed in a three-dimensional “criticality space” (Supplementary Fig. 17b–e). As Fig. 6a, the introduction of recycling significantly reduces both the Vulnerability-to-Supply Risk and Environmental Impact dimensions, thereby alleviating overall metal criticality. It indicates that for high-risk materials such as Ag, In, and Te, sourcing them from urban mines could help bypass geopolitical chokepoints³³. Notably, Fe and Al experience substantial reductions in criticality, with declines exceeding 50%, even in these materials without high criticality. This highlights the benefit of recycling for bulk materials with high mass fractions in PV systems. Moreover, the extent of criticality reduction varies slightly across energy scenarios (SPS, SDS, NES), the overall trend remains consistent, which means that recycling universally contributes to criticality mitigation regardless of future PV deployment pathways (Fig. 6b).

Despite resource price fluctuations, the economic potential of PV system recycling is expected to grow significantly. As shown in Fig. 6c, under three energy pathways, it increases from 0.6–1.3 billion US Dollars (USD) in 2020 to USD 11.5–25.9 billion by 2050. The SDS and NES projects are USD 18.2–41.0 billion and USD 32.5–73.2 billion in 2050, respectively. Among all materials, base and precious metals generate the highest economic returns, with Al, Ag, and Cu contributing approximately 40.6%–50.3%, 10.1%–35.5%, and 12.3%–25.2%, respectively (Fig. 6d). The pronounced contribution of Al and Cu underscores the importance of bulk materials with high mass fractions, and Ag’s high-value highlights the role of precious metals in recycling profitability. However, the pace of economic potential growth will slow after 2035, which is primarily attributed to the declining content of high-value metals, especially Ag and critical metals (Te, In, and Ga) in newly installed PV modules^31,38. This shift raises concerns over the long-term financial viability of purely market-driven recycling systems. To ensure sustained recycling momentum, especially post-2035, policy incentives such as recycling subsidies, recovery quotas, or minimum recycled content standards may be necessary. At the same time, manufacturers should invest in technologies that improve recovery yields for low-content, high-value materials and consider modular design to enhance disassembly efficiency.

On the environmental side, recycling offers considerable carbon mitigation benefits. Under three energy pathways, the potential from secondary materials increases from 1.3–1.8 Mt CO₂-equivalent in 2020 to 42.5–175.6 Mt by 2050 (Fig. 6e). As Fig. 6f, c-Si module recycling contributes over 80% of the total carbon emission reduction potential. c-Si recycling typically involves sequential steps: mechanical disassembly to recover the aluminum frame, delamination to separate glass and expose solar cells, chemical leaching and etching to dissolve metals and recover silicon, and final metal extraction³⁹. Delamination and leaching stages account for 60–80% of the total energy consumption and emissions⁴⁰. This highlights the importance of technological upgrades and the development of greener reagents to reduce the environmental burden in these stages. In parallel, the development and deployment of recyclable encapsulants could significantly improve recyclability and potentially reduce recycling-related emissions. Moreover, recycled glass and silicon require further purification and remanufacturing into solar glass and solar-grade silicon, both of which are energy-intensive and carbon-intensive⁴⁰. Improving the initial purity of recovered materials can help streamline these downstream processes, reduce secondary emissions, and improve the overall environmental performance of the recycling system. Finally, integrating renewable energy into recycling operations can further lower carbon footprints.

In sum, the strategic value of PV recycling goes beyond material recovery. It offers a viable pathway to reduce critical material risks, support industrial competitiveness, and contribute to decarbonization goals.

Sensitivity analysis

The sensitivity analysis is conducted by adjusting key parameters, including market share, weight-to-power ratio, lifetime, and recycling rate, as detailed in Supplementary Table 16. This analysis is designed not only to evaluate the robustness of the results but also to identify potential optimization pathways for further reducing the material net demand of PV systems. The analysis is conducted under the early-loss scenario in the SDS framework, and the outcomes are presented in terms of the total materials to capture overall trends rather than material-specific differences. Annual results are shown in Supplementary Fig. 18, and cumulative results are provided in Supplementary Table 17. Although the projected expansion of third-generation PV technologies is optimistic, its impact on material outflows is limited. However, it significantly reduces material inflows and net demand. Changes in the weight-to-power ratio directly affect material demand due to their linear relationship with material intensities, while the secondary supply ratio remains stable as proportional inflows and outflows are preserved. Lifetime changes significantly affect material outflows, with variations of up to 30%, while the impact on inflows is relatively minor (<10%). Net material demand remains largely unchanged. Finally, the recycling rate has no impact on material inflows or outflows. Their influence on the secondary supply ratio and net material demand is also minor, typically within 1%–3%.

Based on the above analysis, several strategies can be proposed to further reduce the material net demand of PV systems. Accelerating the adoption of third-generation PV technologies offers a promising avenue, not only due to their high efficiency and lightweight characteristics, but also their decreasing reliance on some critical materials with limited reserves and elevated supply risks⁴¹. Although base materials like glass remain indispensable, the reduced use of elements such as Ag, Te, In, and Ga represents a significant advantage from a material security perspective. Reducing the weight-to-power ratio through lightweight design also proves effective, as it directly lowers material input per unit of capacity. Development in module structure, including frameless designs, thin-glass encapsulation, and lighter mounting systems, can support this goal²⁰. However, given that projected improvements in material intensity have already been factored into our baseline, the scope for further reduction through this pathway may be constrained. Extending the operational lifetime of PV systems shows a limited short-term impact. Nonetheless, in the long term, particularly as global PV stock begins to stabilize, this measure can meaningfully reduce replacement-driven demand⁴². While improving the recycling rate has minimal influence on near-term total material demand, its importance will increase over time as the volume of decommissioned PV systems rises and the pace of new deployment slows⁴². Moreover, selectively enhancing the recovery of materials with currently low recycling rates, such as polymers or trace metals, can yield targeted benefits by increasing the availability of specific secondary materials.

Policy implications

While our study demonstrates the potential of material recycling to address resource shortages and environmental impacts, its effectiveness depends largely on policy frameworks for PV waste collection. The collection stage, where decommissioned PV modules are gathered and transported for recycling, is key to enabling resource circularity. Therefore, we analyzed the regulatory frameworks of the top 10 countries by installed PV capacity (see Supplementary Table 20).

In Europe, the European Union (EU) has introduced PV-specific end-of-life directives that mandate recycling and set strict recovery targets, achieving rates above 95% in countries like Germany, Spain, and Italy. In contrast, China, the largest PV market, still lacks clear national regulations, unified infrastructure, and financial incentives, resulting in low recycling rates. The United States faces similar challenges due to the absence of federal regulation and inconsistent state-level policies, keeping the recycling rate below 10%. Japan and India have issued preliminary policies, but enforcement remains limited. Our projections show that under a strong policy framework like the EU’s, a 95% global collection rate could meet 41.2%–96.5% of material demand by 2050. However, with China’s low collection rates, this contribution drops to just 4.3%–10.2%. These findings underscore the vital role of policy in enabling large-scale PV circularity.

Driven by supportive policies and technological development, several successful PV recycling programs have emerged⁴³. In Europe, PV CYCLE achieves recovery rates above 90% by efficiently separating glass, aluminum, silicon, and precious metals. In the United States, First Solar operates a closed-loop process for CdTe modules, recovering up to 90% of semiconductor materials and 95% of glass. These examples demonstrate the technical and operational feasibility of PV recycling and support the practical relevance of our findings. Based on our findings, we propose the following targeted policy recommendations to promote a global circular PV economy.

The first is strengthening regulatory mandates. Our results reveal substantial regional disparities in both circularity potential and policy enforcement. In countries with high projected secondary supply, particularly in Asia and North America, regulatory mandates remain weak. To address this gap, governments should establish regional recycling targets aligned with future secondary material availability. Clear national policies are needed to implement or strengthen Extended Producer Responsibility (EPR) schemes, assigning collection and recycling obligations to manufacturers and importers. Minimum recovery rates should also be set for key materials with high economic value and supply risk, such as Ag, Cu, Al, Cd, and Te. In addition, consistent national regulations on module classification, collection logistics, and recycling standards are essential to improve overall system efficiency. The second is deploying targeted financial incentives. Despite the high economic potential identified for secondary materials, early-stage collection and recycling still face significant financial barriers. Relying solely on strict mandates may increase compliance costs and discourage industry participation⁴⁴. To overcome these challenges, direct financial support, such as subsidies for certified recyclers, is recommended to offset operational expenses. Deposit refund schemes can further encourage end-users to return decommissioned modules to formal collection systems, improving recovery rates. The third is building institutional and infrastructure capacity. Our projections indicate a sharp rise in end-of-life PV modules after 2030, particularly in Asia and North America. To prepare for this, governments should invest in regional recycling infrastructure, ideally through public–private partnerships, to match growing waste volumes and avoid processing delays. Preventing overcapacity and ensuring a stable raw material supply are also key to long-term sustainability²⁵. In parallel, establishing standards and certification for recovered materials, along with recycled content requirements in public procurement, can boost market demand and support recycler viability. Strengthening international coordination on recycling standards, data sharing, and cross-border trade is equally critical to support global circularity.

Comparison with previous studies and future outlook

In the context of three global energy scenarios from the IEA and utilizing a stock-based dynamic material model, this study estimates the global secondary supply potential of anthropogenic minerals from decommissioned PV systems to meet the material demand for future PV deployments across five PV technologies, including PV modules and BOS components (mounting, cabling, and inverters). Our findings reveal that anthropogenic mineral circularity holds potential to reduce reliance on raw materials, particularly under the scenario characterized by early and rapid deployment (NES). Notably, Ag and Te exhibit a surplus secondary supply potential, and given their high criticality, they should be prioritized in recycling efforts. Furthermore, we identify multiscale benefits to encourage stakeholders across the PV value chain to adopt recycling strategies. Overall, this work offers a pathway to mitigate material constraints on long-term energy transitions.

To validate our findings, we compared our results with existing studies. As shown in Supplementary Fig. 19, our estimated inflows and outflows of PV capacity align with those reported by Liang et al.¹¹, but are higher than those in Chen et al.¹², mainly due to different end-of-life assumptions. We apply a Weibull distribution to reflect realistic failure rates, while Chen et al. use a normal distribution. For material demand, we compare our projections for key materials, including Ag, Te, and In, with previous studies (see Supplementary Table 19). Our estimates for Ag and In are lower than those by Xu et al. due to their assumption of more aggressive PV deployment¹³. Our results are also consistent with those of Fahimi Bandpey et al.⁴⁵, though slightly lower as we account for evolving material intensities. Compared to the IEA⁴⁶, our material demand is higher, primarily because we adopt more conservative assumptions regarding future changes in material intensity. Regarding PV waste, the IEA projects decommissioned waste in the NZE to rise from 0.2 Mt in 2021 to about 4 Mt by 2030, 50 Mt by 2040, and over 200 Mt by 2050⁴⁷. Our estimates are higher, as the IEA excludes BOS components, which constitute about 45% of system mass. These comparisons support the robustness of our findings. To provide further context, we compare our cumulative demand and decommissioned amounts by 2050 with global mineral production levels in 2023 (Supplementary Table 21). Critical metals, especially Te, exhibit the highest ratios, with demand ranging from 2016% to 5469% and decommissioned amounts from 986% to 2453% across the three scenarios. Ag also shows substantial ratios, with demand ranging from 325% to 746% and decommissioned amounts from 262% to 539%. Significant ratios are also observed for Al, glass, Cu, and Si. These results highlight the substantial mineral consumption of the PV industry and underscore the importance of recycling anthropogenic minerals to reduce the need for primary resource extraction.

This study acknowledges several limitations that warrant further investigation. First, the analysis focuses on five mainstream PV technologies, excluding others such as Gallium Arsenide solar cells, which may lead to an underestimation of demand for certain materials like Ga. The potential growth of third-generation PV technologies also introduces uncertainty, although their current market share is minimal and only limited decommissioning is expected within the study period. While dynamic material intensities are considered for key elements (Cd, Te, Cu, In, Ga, Se) in thin-film PV, the weight-to-power ratio is assumed constant, unlike in the c-Si PV analysis. It potentially led to an overestimation of materials demand in thin-film systems. Moreover, the study does not account for future raw material supply, which may cause deviations in estimating the secondary material supply potential, particularly for critical materials with constrained geological availability or subject to geopolitical risks. Finally, our study adopts static recovery rates for different materials, without accounting for potential improvements over time driven by technological development. Addressing these limitations in future work will improve the precision of material flow projections and inform the development of more resilient markets for recovered PV materials.

Methods

System boundary and model framework

The system boundary definition and calculation modeling framework of our study are illustrated in Fig. 1. PV systems include both grid-connected and off-grid systems. A PV system comprises various components, including modules that absorb and convert sunlight into electricity, a solar inverter that transforms the output from direct current to alternating current, and mounting and cabling to establish a functional system⁴⁸. Additionally, off-grid PV systems include rechargeable batteries to store surplus energy for use at night. However, grid-connected PV systems are currently the prevalent form, accounting for more than 95% of the market share (Supplementary Fig. 7)⁴⁹. Therefore, this study covers five widely used PV technologies, including monocrystalline silicon (m-Si), polycrystalline silicon (p-Si), amorphous silicon (a-Si), cadmium telluride (CdTe), and copper indium gallium selenide (CIGS), as well as the BOS components (mounting, cabling, and inverters), excluding storage batteries and grid system cabling. The technologies, components, and materials investigated in our study are detailed in Fig. 1a.

A prospective dynamic stock-based model combining MFA is used to simulate the material flows associated with PV systems, thereby quantifying material demand, decommissioned materials, and supply potential of secondary materials in a stepwise manner⁵⁰. The detailed calculation model framework is illustrated in Fig. 1b. (1) Annual installed PV capacity is calculated based on energy scenarios from IEA and market share forecasts, and PV system lifetimes are modeled using a Weibull distribution to estimate decommissioned capacities. After that, material flows (required material and decommissioned material amount) are determined by combining the dynamic weight-to-power ratio and the dynamic material intensity data. (2) Based on the yield of existing recycling technologies, the supply ratio of secondary materials and net material demand are determined. (3) Subsequently, the metal criticality alleviation potential, economic potential and carbon emission reduction potential of recycling secondary materials from decommissioned PV systems are further assessed. (4) The final step involves conducting a sensitivity analysis to evaluate how key parameters affect the uncertainty of the projected results and identify potential optimization pathways for further reducing the material net demand.

Energy scenarios and the stock-based dynamic material model

The estimates for future PV development in our study are based on the energy scenarios developed by the IEA. Three main scenarios are used to capture a range of possibilities for PV development: SPS, SDS, and NES⁵¹. Among these, the SPS is based on current stated policies and represents a more conservative outlook. The SDS provides a balanced approach with achievable goals for sustainable energy development, whereas the NES sets ambitious goals for energy transition, aiming to achieve net-zero carbon dioxide emissions by 2050. Detailed descriptions of these three energy scenarios and their future deployment plans are available in the Supplementary Note 2. Based on these scenarios, we employ interpolation methods to project the annual operational capacity of PV systems.

To calculate annual newly installed PV capacities and annual scrapped PV capacities, a stock-based dynamic MFA is subsequently applied. The relationship between the stocks (in-use PV capacities), inflows (annual newly installed PV capacities), and outflows (annual decommissioned PV capacities) can be expressed as a convolution⁵², as shown in Eq. (1):

$${I}_{t}={S}_{t}-{S}_{t-1}+{O}_{t}$$

(1)

Where I_t (GW) is the inflows, representing annual newly installed capacities in year t; S_t (GW) is the stock, representing in-use capacities in year t, and it is based on future PV development trends from energy scenarios provided by EIA. O_t (GW) is the outflows, representing annual decommissioned PV capacities in year t.

The inflows of specific PV technologies are further determined by the projected market shares of each technology (Eq. (2)), as provided by the International Technology Roadmap for Photovoltaics (ITRPV) and relevant literatures^20,30. The projected shares and corresponding descriptions are presented in Supplementary Method 1.

$${I}_{t,i}={I}_{t}\times {{{{\rm{MS}}}}}_{{{{\rm{t}}}},{{{\rm{i}}}}}$$

(2)

Where, I_t,i (GW) is the annual newly installed capacities of the i-th PV technology in year t; O_t,i (GW) is the annual decommissioned PV capacities of the i-th PV technology in year t; MS_t,i (%) is the market share forecast of the i-th PV technology.

Furthermore, the outflows account for the decommissioning of PV system capacities upon reaching their assumed end-of-life targets due to various failures. In this study, the lifetime of the PV module, a core component of the PV system, is modeled using a Weibull distribution with an average lifespan of 20–30 years (Eq. (3))²⁸. The Weibull distribution is a widely accepted method for modeling PV module lifetime, and the parameters used are aligned with values commonly adopted in existing literature^20,21,24,25. While recent field observations and accelerated aging tests suggest that current modules may operate beyond 30 years^53,54, lower energy output, higher operation and maintenance costs, and increased safety risks make continued use economically unjustifiable. Specifically, the baseline average lifetime of 30 years reflects industry consensus. PV module waste is assessed by considering early and regular loss, with waste generation expected to fall between these projections^28,55. PV module lifetime is split into two categories: Condition A (systems installed before 2013) assumes a 20.84-year lifespan due to lower quality and lack of regulation, while Condition B (systems installed after 2014) assumes a 30-year lifespan due to improved standards. Weibull distribution parameters are detailed in Supplementary Table 13^28,55, and Fig. 2c shows failure probability under different loss scenarios. It is also important to acknowledge that degradation rates and module lifespans can vary significantly due to technological development, as well as differing environmental and operational conditions. To account for these uncertainties, a sensitivity analysis is conducted by adjusting the lifetime parameters accordingly. The expected lifetime of mounting and cabling is projected to exceed 30 years, potentially reaching up to 40 years⁵⁶. However, since PV modules are decommissioned once their lifespan ends, these auxiliary components are assumed to have the same lifespan as the PV modules. The inverter, in contrast, has an estimated lifetime of approximately 10−15 years, modeled using a uniform distribution ranging from 10 to 15 years⁵⁷. The outflows calculation is detailed in Eq. (4).

$${f}_{(x)}=\frac{\beta }{\alpha }{\left(\frac{x}{\alpha }\right)}^{\beta -1}{e}^{-{\left(\frac{x}{\alpha }\right)}^{\beta }}$$

(3)

$${O}_{t,i}={\sum}_{j={t}_{0}}^{j=t}({I}_{j,i}\times {f}_{(\;j-{t}_{0})})$$

(4)

Where f_(x) the Weibull distribution function, representing the failure probability after x years of use. α is the scale factor, and β is the shape factor. I_t,i and O_t,i (GW) are the outflows, representing annual newly installed capacities and annual decommissioned PV capacities in year t.

Based on the inflow and outflow capacities of PV systems discussed above, the prospective inflows and outflows for different materials can be calculated by multiplying the weight-to-power ratio and material intensity, as shown in Eqs. (5) and (6). However, technological development has significantly altered PV module design and material composition over time. For example, the adoption of thinner silicon wafers and higher-efficiency cell architectures has reduced material demand per watt. In addition, developments such as bifacial glass-glass modules, frameless designs, and novel interconnection methods have further shifted material compositions^20,58. Therefore, incorporating dynamic weight-to-power ratios^28,59 and material intensities^29,31,38 provides a more accurate representation of evolving resource demand and end-of-life material flows. The detailed calculation procedures for these dynamic parameters are provided in Supplementary Method 2.

$${{{{\rm{IM}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}={\sum}_{i=1}^{5}({I}_{t,i}\times {R}_{t,i}\times {{{{\rm{MI}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}},{{{\rm{i}}}}})$$

(5)

$${{{{\rm{OM}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}={\sum}_{j={t}_{0}}^{j=t}({{{{\rm{IM}}}}}_{{{{\rm{j}}}},{{{\rm{m}}}}}\times {f}_{(\;j-{t}_{0})})$$

(6)

Where IM_t,m (tonne) and OM_t,m (tonne) represent the mass of inflows and outflows of the m-th material in year t, respectively; R_t,i represents the weight-to-power ratio of the i-th PV technology in year t. MI_t,m,i (%) is the material intensity of the m-th material for the i-th type for PV technology in year t.

Indicators for the supply potential of secondary materials

In this section, we define the annual supply ratio, cumulative supply ratio, annual net demand, and cumulative net demand to evaluate the potential of secondary materials in meeting future resource requirements. The corresponding calculation formulas are provided in Eqs. (7)–(14). The annual and cumulative supply ratios represent the proportion of material demand met by recycled materials in a given year and cumulatively by 2050, respectively. Furthermore, the annual and cumulative net demands refer to the remaining demand for primary raw materials after accounting for secondary material supply, both annually and cumulatively through 2050. We conduct a comprehensive assessment of the supply potential for each material used in PV systems. Additionally, we aggregate all materials to evaluate the overall trend in secondary material supply potential under different energy scenarios. These indicators consider the material-specific recycling rates of recycling technologies, as detailed data and sources provided in Supplementary Table 14. Given the high recycling rates in current technologies and the uncertainty surrounding future changes, we applied static values and conducted sensitivity analysis to assess the uncertainty of this assumption.

$${{{{\rm{ASR}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}=\frac{{{{{\rm{OM}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}\times {{{{\rm{RR}}}}}_{{{{\rm{m}}}}}}{{{{{\rm{IM}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}}$$

(7)

$${{{{\rm{ASR}}}}}_{{{{\rm{t}}}}}=\frac{{\sum }_{m=1}^{23}\left({{{{\rm{OM}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}\times {{{{\rm{RR}}}}}_{{{{\rm{m}}}}}\right)}{{\sum }_{m=1}^{23}{{{{\rm{IM}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}}$$

(8)

$${{{{\rm{CSR}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}=\frac{\left(\mathop{\sum }_{j=2011}^{t}{{{{\rm{OM}}}}}_{{{{\rm{j}}}},{{{\rm{m}}}}}\right)\times {{{{\rm{RR}}}}}_{{{{\rm{m}}}}}}{\mathop{\sum }_{j=2011}^{t}{{{{\rm{IM}}}}}_{{{{\rm{j}}}},{{{\rm{m}}}}}}$$

(9)

$${{{{\rm{CSR}}}}}_{{{{\rm{t}}}}}=\frac{\mathop{\sum }_{j=2011}^{t}\left(\left(\mathop{\sum }_{m=1}^{23}{{{{\rm{OM}}}}}_{{{{\rm{j}}}},{{{\rm{m}}}}}\right)\times {{{{\rm{RR}}}}}_{{{{\rm{m}}}}}\right)}{\mathop{\sum }_{j=2011}^{t}\mathop{\sum }_{m=1}^{23}{{{{\rm{IM}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}}$$

(10)

$${{{{\rm{AND}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}={{{{\rm{IM}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}-{{{{\rm{OM}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}\times {{{{\rm{RR}}}}}_{{{{\rm{m}}}}}$$

(11)

$${{{{\rm{AND}}}}}_{{{{\rm{t}}}}}=\mathop{\sum }_{m=1}^{23}{{{{\rm{AND}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}$$

(12)

$${{{{\rm{CND}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}=\mathop{\sum}_{j=2011}^{t}{{{{\rm{AND}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}$$

(13)

$${{{{\rm{CND}}}}}_{{{{\rm{t}}}}}={\sum}_{j=2011}^{t}{\sum}_{m=1}^{23}{{{{\rm{AND}}}}}_{{{{\rm{j}}}},{{{\rm{m}}}}}$$

(14)

Where IM_t,m (tonne) and OM_t,m (tonne) are the mass of inflows and outflows for the m-th material in year t, respectively; ASR_t,m (%) and ASR_t (%) are the annual supply ratio of the m-th material and total materials in year t, respectively; RR_m (%) is the recovery ratio of the m-th material in the recycling process (detailed sources are listed in Supplementary Table 14); CSR_t,m (%) and CSR_t (%) are the cumulative supply ratios for the m-th material and for total materials in year t, respectively; AND_t,m (tonne) and AND_t (tonne) are the annual net demand for the m-th material and for total materials in year t, respectively; CND_t,m (tonne) and CND_t (tonne) are the cumulative net demand for the m-th material and for the total materials in year t, respectively.

Assessment of the benefits of secondary material supply

To encourage specific stakeholders to adopt desired behaviors and actively participate in recycling, we further quantitatively calculate three key drivers for PV waste recycling: the potential benefits to metal criticality, economic potential, and carbon emission reduction potential.

In this study, we adapt the metal criticality assessment from the Yale Criticality methodology to compare the metal criticality alleviation potential between the cases with and without recycling^33,60. As shown in Supplementary Fig. 17, each metal is positioned in a three-dimensional “criticality space” defined by supply risk (SR), vulnerability-to supply restriction (VSR), and environmental implications (EI)³³. Different indicators are available for users from various organizational types, allowing for flexibility in assessment methodology. This study focuses on the global level and long-term outlook. Consequently, SR considers geological, technological, and economic factors, VSR accounts for importance, substitutability, and sustainability, while EI encompasses impacts on human health and ecosystems. Detailed definitions and calculation methods for these indicators are provided in Supplementary Method 3. Each indicator is scored from 0 to 100, with higher scores indicating higher criticality. The overall criticality is calculated using Eq. (15)⁶¹.

$$C=\sqrt{\frac{{{{{\rm{SR}}}}}^{2}+{{{{\rm{VSR}}}}}^{2}+{{{{\rm{EI}}}}}^{2}}{3}}$$

(15)

The economic potential and carbon emission reductions associated with the global supply of secondary materials are also assessed in this study.

For economic potential, this analysis focuses on estimating the economic potential of secondary materials rather than the direct economic benefits of the recycling processes themselves. The goal is to identify which materials offer the highest potential value and to provide insights for prioritizing material recovery⁷. Additionally, this assessment reflects how the economic potential of various materials may evolve over time, thereby supporting long-term resource planning and decision-making. Given the scope of this study, we do not conduct detailed techno-economic analyses. Such analyses typically require highly granular and location-specific data, including capital expenditures, labor costs, energy use, taxation policies, and regulatory environments. These factors can vary widely across countries and regions, and applying generalized or static assumptions in a global context would introduce significant uncertainty and could lead to misleading conclusions. Therefore, instead of adopting a detailed techno-economic modeling approach, we estimate the economic potential of recycled materials by integrating macro-scale material outflow projections with corresponding market prices of individual materials. The detailed calculation of economic potential is shown in Eq. (16).

We applied life cycle assessment to quantify the carbon emission reduction coefficients associated with the recycling of various PV modules and BOS components at the micro level, as detailed in Supplementary Method 4. These coefficients were then combined with the projected material flow data to estimate the global carbon mitigation potential enabled by secondary material recovery. The detailed calculation process is presented in Eq. (17).

$${{{{\rm{EBP}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}={{{{\rm{OM}}}}}_{{{{\rm{t}}}},{{{\rm{m}}}}}\times {P}_{m}$$

(16)

$${{{{\rm{CRP}}}}}_{{{{\rm{t}}}},{{{\rm{i}}}}}={{{{\rm{OM}}}}}_{{{{\rm{t}}}},{{{\rm{i}}}}}\times {C}_{t,i}$$

(17)

Where EBP_t,m (USD) and CRP_t,i (kg CO₂-equivalent) are the global economic benefits potential of the m-th material and carbon emission reduction potential of recycling the i-th components in year t; P_m (USD·tonne^-1) and C_t,i (kg CO₂-equivalent·tonne⁻¹) represent the market prices for the m-th material and the carbon emission reduction from recycling the i-th components in year t, as presented in Supplementary Table 15 and Supplementary Method 4.

Sensitivity analysis

A large volume of data is required in the calculation process of the stock-based dynamic material flow model, which inevitably introduces uncertainties due to potential input errors and future variability. To address this issue, we perform a sensitivity analysis to assess how changes in key parameters influence the uncertainty of projected outcomes, which is a commonly adopted approach in dynamic material flow studies. Specifically, we examine the sensitivity of four critical parameters: market share, weight-to-power ratio, lifetime, and material recovery rate. Although many previous studies have assessed the impact of installation capacity²¹, our study accounts for this factor more comprehensively by incorporating multiple global energy scenarios.

The analysis of market share focuses on the potential expansion of third-generation PV technologies. According to previous research, their combined market share could reach up to over 20% by 2050 under optimistic scenarios²⁰. However, it should be noted that it is optimistic because considerable uncertainty remains regarding their long-term stability and commercial viability. For the weight-to-power ratio, we assume a ±10% variation to reflect potential uncertainties. For the lifetime, we adjust the average lifetime of the PV module from 30 years to 25 and 35 years to reflect potential differences in durability across technologies. For the recycling rate, we consider both a static 10% increase and a dynamic improvement trajectory. A detailed description of the parameter ranges and assumptions is provided in Supplementary Table 16. All sensitivity analyses are conducted under the early-loss scenario in the SDS, and the results are presented in terms of total materials to illustrate overall trends rather than material-specific variations.