We develop an integrated modelling framework to project end-of-life (EOL) photovoltaic (PV) waste generation across 32 global regions and evaluate the environmental and economic benefits of alternative solar panel recycling strategies. The framework is designed to support evidence-based policy development and combines material price modelling, the Global Change Analysis Model (GCAM), dynamic material flow analysis (MFA), life-cycle assessment (LCA), and life-cycle costing (LCC) (Extended Data Fig. 1 and Supplementary Table 3).
The framework includes three interconnected components. First, long-term price trajectories for critical PV materials are generated and incorporated into GCAM to simulate regional solar PV deployment under alternative socioeconomic and climate scenarios. GCAM electricity-generation outputs are then converted into installed PV capacity and used in a dynamic MFA model to estimate regional EOL PV waste. Second, projected PV waste streams are coupled with LCA and LCC models to quantify the climate impacts and economic performance of different recycling technologies. Third, a multidimensional scenario framework covering PV decommissioning pathways, recycling technologies, international waste trade and subsidy policies is applied to assess how market and policy conditions influence regional solar panel recycling outcomes. Detailed model parameters are provided in Supplementary Tables 4–9.
Critical material price model for solar PV recycling
To assess how critical material price uncertainty affects PV deployment and recycling, we simulate long-term price trajectories for four solar-relevant materials: copper, aluminium, silver and silicon. The material price model builds on previous studies7,51,52 and incorporates historical price trends, future demand growth and the potential for material substitution. The resulting price pathways are introduced as exogenous inputs into GCAM to determine solar PV deployment under alternative climate targets.
The purpose of the model is not to provide precise price forecasts. Instead, it generates plausible scenario-based price trajectories that support the evaluation of long-term PV deployment, decommissioning and recycling pathways under material market uncertainty.
The model is based on dynamic market equilibrium. In the long term, material prices are determined by the marginal cost of the new supply required to satisfy demand. When existing mining capacity is insufficient, additional mining projects are developed. Consequently, material prices are governed by the marginal cost of newly installed mining capacity.
The modelling process comprises three steps.
Step 1: Projecting material demand. Future critical material demand is determined by exogenous demand growth and price-responsive substitution effects:
$${Q}^{t+1}={Q}^{t}(1+g+\Delta {p}^{t}\times \varepsilon )$$
(1)
Here, Qt is material demand in period t; g is the annual exogenous demand growth rate (Supplementary Table 4); ε is the price elasticity of demand; and Δpt is the annual price change rate:
$$\Delta {p}^{t}=\frac{{p}^{t}-{p}^{t-1}}{{p}^{t-1}}$$
(2)
The initial price, p0, is specified exogenously using the global average price of each material in 2020.
Step 2: Determining new mining capacity and material prices. The additional mining capacity required to satisfy future demand is calculated as follows:
$${O}^{t+1}={(Q}^{t+1}-{Q}^{t})+{L}^{t}-({R}^{t+1}-{R}^{t})$$
(3)
Here, Ot+1 is the newly required annual production capacity in period t + 1; Lt represents supply losses caused by mine closures and resource depletion in period t; and Rt is secondary material supply in period t.
The marginal cost of new mining capacity is represented by an incentive cost curve:
$${I}^{t}({O}^{t})={a}^{t}+{b}^{t}{O}^{t}$$
(4)
In this equation, at represents the minimum marginal cost of new mining capacity in period t, while bt represents the rate at which marginal costs increase because of declining ore grades, more complex investment conditions and other supply constraints.
The equilibrium material price in year t is determined by the marginal cost of the most expensive required new capacity:
$${P}_{t}={I}_{t}({O}_{t})$$
(5)
Step 3: Updating mining cost structures. The mining incentive cost curve changes over time because of ore-grade depletion, technological progress and changes in operating costs51. These effects are represented by the cost-adjustment factor n:
$$n=\frac{(1+e)\times (1-{t}_{g})}{(1-l)}$$
(6)
Here, e is the annual growth rate of operating costs, including energy, labour, water, reagents and environmental compliance; tg is the annual cost-reduction rate associated with technological progress; and l is the annual ore-grade depletion rate.
The intercept of the operating mine cost curve, at+1, is updated as the minimum of the previous period’s operating cost adjusted by nt and the operating cost of the least-cost mine opened in period t + 1, after subtracting annualized capital costs:
$${a}^{t+1}=\min ({a}^{t}\times n,{a}^{t}-{c}_{a})$$
(7)
The upper bound of the operating cost curve, Ct+1(1), representing the operating cost of the most expensive active mine in period t + 1, is updated as:
$${C}^{t+1}(1)=\max ({p}^{t},{p}^{t+1}-{c}_{a})$$
(8)
This formulation captures two market mechanisms. First, mines with operating costs above the previous market price, pt, are assumed to exit. Second, new mines enter only when their operating costs are covered by the current material price net of annualized capital costs, pt+1 − ca. The maximum of these values ensures that the operating cost frontier reflects both mine closures and new capacity investment.
The slope of the operating cost curve is then calculated as:
$${b}^{t+1}=\frac{{C}^{t+1}(1)-{a}^{t+1}}{{Q}^{t+1}}$$
(9)
Further information about the model structure and assumptions is provided in our previous study7.
Solar material price scenarios
To represent uncertainty in future critical material prices and its effects on solar PV deployment and decommissioning, we develop alternative price scenarios by varying parameters related to mining productivity and material substitution.
Mining productivity is represented by tg − e, which captures the difference between reductions in mining costs caused by technological progress and increases in operating costs. Material substitution is represented by the price elasticity parameter ε in equation (1), which measures the responsiveness of demand to price changes and the potential shift toward alternative materials.
Parameter values are selected from ranges reported in the literature, as detailed in Supplementary Table 4. For both mining productivity and substitution elasticity, low and high values are classified using the midpoint of each reported range as the threshold. These assumptions produce two contrasting scenarios:
-
(1)
Low-price scenario: assumes rapid technological progress in mining and strong material substitution, leading to faster reductions in mining costs and more moderate long-term material price growth.
-
(2)
High-price scenario: assumes limited mining productivity improvements and weak material substitution, resulting in tighter supply conditions and higher long-term prices.
Price trajectories for copper, aluminium, silver and silicon are simulated under both scenarios. The resulting pathways are shown in Extended Data Fig. 2.
GCAM model for global solar PV deployment
The Global Change Analysis Model (GCAM) is an integrated assessment model linking water, energy, land use, socioeconomic and climate systems. GCAM is widely used in international and national energy and climate scenario assessments52,53.
The model represents a broad range of electricity-generation technologies, including solar photovoltaics and wind power. Renewable resources are region-specific and non-tradable between regions. Technology deployment is determined through a multistage process that combines technology costs, resource supply curves and market-based technology selection.
Linking material prices to electricity generation in GCAM
Incorporating material price changes into PV capital costs
To quantify the effect of critical material price changes on solar PV deployment, simulated material prices are translated into technology-specific capital cost adjustments. PV capital cost in year t is defined as:
$${C}_{t}={C}_{\mathrm{BLS},t}+{\Delta C}_{t}$$
(10)
Here, \({C}_{\text{BLS},t}\) is the baseline PV capital cost under GCAM’s default assumptions54, and ΔCt is the cost increase associated with changes in material prices:
$${\Delta C}_{t}=\sum _{m}{\mathrm{MI}}_{m}\times ({P}_{{mt}}-{P}_{mt,\mathrm{BLS}})$$
(11)
Here, MIm is the material intensity of material m in solar PV; Pmt is the simulated price of material m in year t; and \({P}_{mt,\mathrm{BLS}}\) is the corresponding baseline price.
Total electricity-generation technology cost
GCAM calculates the total cost of an electricity-generation technology as7:
$${C}_{\mathrm{total}}=t({C}_{t})+\sum _{j}{p}_{j}+\sum _{k}{g}_{k}-\sum _{l}{v}_{l}\,$$
(12)
In this equation, t(Ct) represents capital and fixed operating costs; pj is the marginal cost of input resource j; gk is the cost of greenhouse gas emissions; and vl is the value of secondary outputs. For solar technologies, resource costs are determined using the renewable resource supply curve in equation (15).
Technology choice and solar PV market shares
GCAM determines electricity-generation technology shares using a relative-cost logit formulation7:
$${s}_{j}=\frac{{\alpha }_{j}{C}_{\mathrm{total},j}^{\gamma }}{{\sum }_{j=1}^{N}{\alpha }_{j}{C}_{\mathrm{total},j}^{\gamma }}$$
(13)
Here, αj is the technology-specific share weight, Ctotal,j is the total cost of technology j, and γ is the logit exponent that determines the sensitivity of technology shares to cost differences. Technologies with lower relative costs receive larger market shares.
Electricity generation from technology j is calculated as:
$${Q}_{j}^{e}={s}_{j}\times D$$
(14)
Here, D is total electricity demand under the relevant shared socioeconomic pathway (SSP).
Renewable resource supply curves
GCAM uses region- and technology-specific resource supply curves to represent differences in renewable resource availability. The cumulative generation potential available at or below marginal cost p is:
$${Q}^{e}(p)=\text{MaxSubResource}\frac{{p}^{\text{curveExponent}}}{{\text{MidPrice}}^{\text{curveExponent}}+{p}^{\text{curveExponent}}}$$
(15)
MaxSubResource is the maximum exploitable resource potential, MidPrice is the cost at which half of this potential becomes available, and curveExponent controls the steepness of the supply curve.
For solar PV, GCAM assumes an abundant and effectively flat resource base. This reflects the high solar potential of many regions and the modular scalability of PV systems. Under this assumption, marginal resource costs do not rise substantially with deployment, allowing critical material price shocks to affect PV deployment primarily through capital costs.
Historical GCAM electricity-generation outputs were cross-validated against official IRENA statistics to assess historical accuracy and backcasting performance. Detailed comparisons are provided in Supplementary Note 1 and Supplementary Fig. 3.
Converting solar electricity generation into installed PV capacity
GCAM electricity-generation outputs are converted into installed solar PV capacity to estimate future PV waste. Generation reported in exajoules (EJ) is first converted into gigawatt-hours (GW h) using the conversion factor ω = 277,778.8. Installed capacity is then calculated by dividing annual electricity generation by annual operating hours and the regional capacity factor:
$${\mathrm{IC}}_{i,t}=\frac{{\mathrm{EJ}}_{i,t}\times \omega }{{\mathrm{CF}}_{i}\times T}$$
(16)
Here, \({\mathrm{IC}}_{i,t}\) is installed PV capacity in region i and year t; EJi,t is annual solar electricity generation; CFi is the regional capacity factor; and T is the number of hours per year, set to 8,760.
Integrated socioeconomic and climate scenarios
GCAM generates internally consistent energy, economic, land-use and climate pathways under alternative socioeconomic and climate policy assumptions. We combine five SSPs (SSP1–SSP5) with three Representative Concentration Pathways (RCP2.6, RCP4.5 and RCP6.0) to construct a comprehensive scenario set for evaluating future solar PV deployment53.
For presentation purposes, scenario combinations are classified into commonly used benchmark pathways and supplementary stress-test pathways according to their use in the literature56. This classification does not change the model simulations or parameterization. Complete scenario descriptions are provided in Supplementary Table 3 and Supplementary Note 2.
All simulations use GCAM v.8.2. Regional solar electricity-generation trajectories are extracted for each scenario and aggregated into four income groups using the regional classification in Supplementary Table 1. Additional model documentation is available in the GCAM v.8.2 Model Overview.
Projecting future end-of-life PV waste generation
Future solar panel waste is estimated using a dynamic material flow analysis (MFA) model. MFA is widely used to track material stocks and flows over time and to estimate waste generation from solar energy systems57,58.
The primary MFA input is annual PV capacity in use, which combines historical installed capacity with future capacity additions. Historical installed capacity from 2000 to 2024 is updated using 2025 IRENA statistics59 (Supplementary Tables 10–13). Extending the time series to 2000 improves calibration of early PV decommissioning volumes. The main analysis covers 2020–2060.
Future PV capacity is obtained from GCAM projections at five-year intervals and interpolated annually using cubic spline interpolation. PV waste is estimated in two stages: first, annual PV capacity inflows and outflows are calculated; second, decommissioned capacity is converted into waste mass.
Step 1: Calculating PV capacity inflows and outflows. PV module lifetimes are represented using a Weibull lifetime distribution58:
$${\mathrm{outflow}}_{i}$
(17)
$${\rm{F}}(t-{t}^{{\prime} })=1-\exp \left[-{\left(\frac{t-{t}^{{\prime }}}{T}\right)}^{\beta }\right]$$
(18)
$${\mathrm{inflow}}_{i}$
(19)
Here, t is the year from 2000 to 2060; t′ is the PV installation year; and t − t′ is the module service time. F(t − t′) is the cumulative Weibull distribution function, T is the average PV module lifetime of 30 years, and β is the shape parameter.
We use a regular-loss scheme that assumes no premature module failure during the service life. The Weibull shape parameter is therefore set to β = 5.3759 (ref. 60). The variables outflowi(t), stocki(t) and inflowi(t) represent decommissioned capacity, installed capacity in use and newly added capacity, respectively. Initial stock is assumed to equal first-year inflow.
Step 2: Converting decommissioned PV capacity into waste mass. Decommissioned capacity is converted into PV waste mass using time-varying module weight-to-power ratios:
$${\mathrm{PVWaste}}_{i}$
(20)
Here, perPVton(t) is the PV module weight-to-power ratio in period t, obtained from IRENA reports60 (Supplementary Table 14). Because projections are available only to 2050, the 2050 ratio is held constant through 2060.
Annual PV inflow, outflow and in-use stock trajectories for 2020–2060 are presented in Supplementary Figs. 4–6. The analysis focuses on crystalline-silicon (c-Si) PV modules, which represented more than 90% of global PV installations after 201261. Other PV technologies are excluded. Detailed assumptions are provided in Supplementary Note 3.1.
Economic cost–benefit analysis of PV recycling
The economic feasibility of recycling EOL PV panels is evaluated using a life-cycle cost (LCC) framework aligned with the LCA system boundary23,62. Recycling costs are estimated for 2020 in US dollars per tonne of PV waste, adjusted over time for technological learning and economies of scale, and multiplied by projected regional PV waste volumes. Economic benefits are calculated from the market value of recovered secondary materials.
The unit total recycling cost, uTC, includes stage-specific and whole-process costs. Stage-specific costs are variable costs incurred during individual processing stages, whereas whole-process costs cover expenses across the complete PV recycling chain.
The 2020 unit recycling cost is:
$${\mathrm{uTC}}_{0}=\mathop{\sum }\limits_{\mathrm{ss}=1}^{n}{\mathrm{uC}}_{\mathrm{ss}}+{\mathrm{uC}}_{{\rm{w}}}$$
(21)
$${\mathrm{uC}}_{{\rm{w}}}={\mathrm{uC}}_{{\mathrm{l}}}+{\mathrm{uC}}_{{\mathrm{m}}}+{\mathrm{uC}}_{{\mathrm{f}}}+{\mathrm{uC}}_{\mathrm{ope}}+{\mathrm{uC}}_{\mathrm{opp}}$$
(22)
Here, uCss represents stage-specific costs for collection, transportation, dismantling, technical treatment and disposal. uCw includes labour, management, fixed-asset depreciation, operation and maintenance, and opportunity costs.
Because globally differentiated PV recycling cost data are limited, we use an internally consistent cost-benefit inventory for China as the reference baseline56. Region-specific labour, collection and transportation costs are estimated separately. Other components, including dismantling, technical treatment and operating expenses, are adjusted using purchasing power parity (PPP):
$${\mathrm{uC}}_{i,0}=\left(\frac{{\mathrm{uC}}_{\mathrm{CN},0}}{{\mathrm{PPP}}_{\mathrm{CN},0}}\right)\times \frac{{\mathrm{PPP}}_{i,0}}{{\mathrm{ER}}_{i,0}}$$
(23)
Here, uCi,0 is the estimated 2020 unit cost for region i; uCCN,0 is the Chinese baseline cost; PPPCN,0 and PPPi,0 are the PPP conversion factors for China and region i; and ERi,0 is the regional market exchange rate.
Collection costs are estimated using previous research63. Labour costs are based on national per-capita gross national income data from the United Nations Trade and Development Data Hub64, assuming a PV recycling capacity of 100 tonnes per worker per year62.
Transportation costs include domestic and international maritime transport. Domestic maritime costs are estimated using equation (23). International shipping costs are based on bilateral maritime distances and a unit freight cost of US$0.1 per nautical mile per twenty-foot equivalent unit65. Each container is assumed to carry a maximum of 30 tonnes in accordance with ISO standards66.
Maritime distances are derived from a global port-network model67,68. Shortest navigable port-to-port routes are calculated between all trading-region pairs. The 10th percentile of the route-distance distribution is used as the representative bilateral distance, and a correction factor of 1.3 is applied to approximate real-world sailing conditions. Regional costs are aggregated to the 32 GCAM regions using arithmetic means. Regional 2020 unit costs are shown in Extended Data Fig. 9.
PV recycling costs are expected to decline over time because of technological learning and economies of scale23,43. We therefore use a modified learning curve with an irreducible cost floor:
$${\mathrm{uTC}}_{t}={\mathrm{uTC}}_{0}\times [{C}_{\min \mathrm{\_ratio}}+(1-{C}_{\mathrm{min\_ratio}})\times {(1-\mathrm{LR})}^{t-{t}_{0}}]$$
(24)
Here, uTCt is the unit recycling cost in year t; uTC0 is the 2020 cost; t0 is the base year; and LR is the learning rate. The parameter Cmin_ratio represents irreducible costs associated with energy, reagents and labour. Based on evidence that global PV costs declined by up to 87% between 2010 and 202471, Cmin_ratio is set to 0.13.
Total recycling costs in year t are calculated as:
$${\mathrm{TC}}_{t}={\mathrm{uTC}}_{t}\times {\mathrm{PVWaste}}_{t}\times {(1+\pi )}^{t-{t}_{0}}$$
(25)
Here, π is the annual inflation rate calibrated using the global average inflation rate from 2000 to 202472.
Recycling benefits are calculated from recovered aluminium, glass, silver, copper and silicon. Benefits depend on material prices, recovery efficiencies and the volume of decommissioned PV modules:
$${B}_{t}={P}_{{mt}}\times {\mathrm{uR}}_{m}\times {\mathrm{PVWaste}}_{t}\times {(1+\pi )}^{t-{t}_{0}}$$
(26)
Here, Pmt is the simulated price of material m in year t, and uRm is the recovered quantity per unit of PV waste. Costs and material prices are first estimated in constant 2020 US dollars. Nominal values are calculated by applying inflation.
Net economic benefit and unit net benefit are calculated as:
$${\mathrm{NB}}_{t}=({B}_{t}-{\mathrm{TC}}_{t})$$
(27)
$${\mathrm{UNB}}_{t}=\frac{{\mathrm{NB}}_{{t}}}{{\mathrm{PVWaste}}_{t}}$$
(28)
Detailed cost and benefit accounting procedures are provided in our previous study62.
Climate benefits of recycling end-of-life PV panels
We quantify the climate benefit of solar panel recycling as the net greenhouse gas (GHG) emissions avoided by recycling 1 tonne of EOL c-Si PV modules:
$${\mathrm{CB}}_{s,i,c,t}={\mathrm{RB}}_{s,i,c,t}-{\mathrm{RG}}_{s,i,c,t}$$
(29)
Here, CBs,i,c,t is the net climate benefit of recycling technology s in region i, under climate pathway c and year t; RB is the avoided GHG emissions generated by recycling; and RG is the GHG emissions released by the recycling process.
Recycling benefits include avoided emissions from replacing virgin materials with recovered materials of equivalent quality and quantity. These avoided emissions include primary extraction, refining and manufacturing. Where applicable, energy recovery from polymeric fractions, such as backsheets and encapsulants, also offsets emissions from conventional energy production21,75.
Recycling burdens include emissions from electricity consumption, auxiliary inputs and transportation throughout the EOL treatment chain, including the transport of waste modules, intermediate fractions and residual materials76,77.
The functional unit is 1 tonne of EOL c-Si PV panels, consistent with the LCA framework62,78. The system boundary extends from PV module collection to the recovery of secondary materials and energy and is shown in Supplementary Fig. 7.
Because electricity emission intensity changes over time and differs among regions and climate pathways, electricity-related LCA emissions are adjusted using region-, pathway- and year-specific emission factors:
$${\mathrm{EF}}_{i,c,t}={\alpha }_{i,c,t}\times {\mathrm{EF}}_{i,c,\mathrm{base}}$$
(30)
Here, EFi,c,t is the electricity-sector emission factor in region i, under climate pathway c and year t; EFi,c,base is the 2020 baseline emission factor; and αi,c,t is the adjustment coefficient.
The resulting climate benefit is:
$${\mathrm{CB}}_{s,i,c,t}=({\alpha }_{i,c,t}\times {\mathrm{RB}}_{s}^{\mathrm{elec}}+{\mathrm{RB}}_{s}^{\mathrm{non}\text{-}\mathrm{elec}})-({\alpha }_{i,c,t}\times {\mathrm{RC}}_{s}^{\mathrm{elec}}+{\mathrm{RC}}_{s}^{\text{non-elec}})$$
(31)
Here, RBselec and RBsnon-elec are electricity-related and non-electricity-related recycling benefits, respectively. RCselec and RCsnon-elec represent electricity-related and non-electricity-related recycling burdens.
Carbon dioxide accounts for more than 70% of total GHG emissions and is the primary contributor to the global warming potential of PV recycling, particularly through transport fuel use, polymer incineration and coal-based electricity generation38,77,78,79. Accordingly, the temporal evolution of regional GHG emission factors is assumed to follow the carbon-emission trajectories projected by GCAM. Process inventories are provided in Supplementary Tables 16–18 and ref. 62.
Solar PV recycling technology scenarios
We develop an integrated framework to model the evolution of PV recycling technologies across regions and over time. The framework combines scenario design, a Logit-based technology choice model and region-specific constraints80.
The model accounts for differences in low-carbon technology investment, technical capacity, environmental regulation and projected PV waste generation. Mechanical, thermal and chemical recycling technologies coexist and compete, while their market shares evolve according to regional conditions.
Four recycling technology scenarios are considered: business as usual (BAU), economic priority, carbon priority and technology diffusion.
Under the BAU scenario, each region’s recycling technology structure remains unchanged throughout the study period. Technology shares are fixed at their 2020 levels and do not respond to changes in economic performance or climate benefits. Base-year shares are calibrated using recycling patent distributions and facility locations reported by the IEA31 (Supplementary Tables 19 and 20).
Under the economic-priority scenario, technology selection is driven by net economic benefit, defined as revenue from recovered materials minus recycling costs. Market shares evolve in response to relative profitability. Regional differences in investment capacity are also included because low-carbon technology investment and research and development remain concentrated in developed regions81,82,83.
Mechanical recycling is less capital-intensive but generally has lower recovery efficiency for high-value metals. It therefore produces lower net profits per tonne than thermal and chemical recycling. As a result, mechanical recycling retains larger and more persistent market shares in many middle- and low-income regions than in high-income regions.
The carbon-priority scenario represents climate-driven technology selection. Technology attractiveness is determined by the GHG reduction benefit of recycling. Regional differences in regulatory ambition and policy enforcement are represented using a climate policy stringency index based on the OECD Environmental Policy Stringency Index84. Higher policy stringency shifts market shares toward technologies with greater carbon mitigation performance.
The technology diffusion scenario follows historical PV patent diffusion patterns from 1970 to 2022. Because PV waste generation is projected to increase sharply after 204058, technology diffusion occurs in four stages: baseline shares before 2030; early adoption from 2030 to 2040; accelerated expansion from 2040 to 2050; and maturation from 2050 to 2060. Recipient regions gradually converge toward the technology structures of leading regions. Stage-specific parameters are calibrated using historical PV patent data85. Further assumptions are provided in Supplementary Note 3.4.
Logit model for PV recycling technology choice
In the economic-priority and carbon-priority scenarios, recycling technology shares are calculated using a Logit-based choice model that allows mechanical, thermal and chemical recycling technologies to coexist and compete within each region80.
For region i and year t, technology shares satisfy:
$$\sum _{s}{\mathrm{Share}}_{s,i,t}=1$$
(32)
The market share of technology s is:
$$\begin{array}{c}{\mathrm{Share}}_{s,i,t}={p}_{s,i}\exp ({\gamma }_{i}\times {V}_{s,i,t})/\sum _{s}{p}_{\text{s},i}\exp ({\gamma }_{i}\times {V}_{\text{s},i,t})\end{array}$$
(33)
Here, ps,i is the 2020 base-year weight of technology s in region i, representing existing infrastructure and path dependence. Vs,i,t is the scenario-specific performance indicator: net economic benefit in the economic-priority scenario or carbon reduction benefit in the carbon-priority scenario. γi represents regional investment capacity or climate policy stringency. Higher values concentrate market shares more strongly on the best-performing technology.
International PV waste trade scenarios
End-of-life PV module management generally follows two pathways: domestic recycling or outsourced recycling. Extended producer responsibility (EPR) regulations in regions such as the European Union and South Korea require producers to support the collection, transport and recycling of decommissioned PV modules10.
Globalized PV supply chains may further increase cross-border EOL module flows. In 2017, imported modules accounted for 76.89% of newly installed PV capacity86. Regions that rely heavily on imported modules but lack domestic recycling capacity may therefore export EOL PV modules to technologically advanced recycling regions.
Formal PV recycling capacity is currently concentrated in a limited number of regions, including China, the United States, the European Union, Japan and South Korea87. Many regions in the Global South face limited industrial capacity, technical expertise and regulatory enforcement. EOL modules may consequently be exported or processed through informal recycling channels, which can create environmental and occupational health risks40,41,88.
We examine four international PV recycling trade scenarios: local recycling, EPR-oriented trade, expanded global trade and regional trade. In this study, transboundary flow refers specifically to the international shipment of intact EOL PV modules. Detailed assumptions are provided in Supplementary Note 3.5.
Under the local recycling scenario, regions with established recycling technologies process all domestically generated EOL modules within their borders. International PV waste trade is prohibited. Regions without recycling technologies cannot recycle and receive no recycling benefits. This scenario provides a benchmark for evaluating the effects of domestic recycling capacity and international trade.
The EPR-oriented trade scenario represents a producer-responsibility system. EOL modules are exported to the regions responsible for their original production according to a return-to-producer rule and historical bilateral PV module trade routes86. Producer regions conduct recycling, pay treatment costs and retain revenue from recovered materials. Historical trade structures remain unchanged through 2060.
The expanded global trade scenario allows all regions with PV recycling technologies to import EOL modules. Regions without recycling capacity export their waste to eligible recycling regions. Waste allocation is based on economic globalization and trade openness indices89, reflecting trade barriers, logistics capacity and institutional readiness. Trade structures remain fixed through 2060.
The regional trade scenario limits international PV waste trade to five continental groups: Africa, Asia, Europe, the Americas and Oceania. Regions without recycling technologies export EOL modules to recycling-capable regions within the same continent. Import allocation is based on relative globalization and trade openness indices. Intercontinental trade is prohibited, reflecting potential logistical, regulatory and political constraints.
Measuring regional inequality in PV recycling benefits
To evaluate how PV recycling trade scenarios affect distributional equality, we use the cross-regional dispersion of realized economic and climate benefits as a transparent proxy for equality.
For each trade scenario j, the variance of regional benefits across the 32 modelled regions is calculated as:
$${\sigma }_{\mathrm{sc}}^{2}=\frac{1}{n}\mathop{\sum }\limits_{i}^{n}{({\mathrm{EB}}_{i,\mathrm{sc}}-\overline{{\mathrm{EB}}_{\mathrm{sc}}})}^{2}$$
(34)
Here, EBi,sc is the economic or climate benefit received by region i under scenario sc, and \(\overline{{\mathrm{EB}}_{\mathrm{sc}}}\) is the mean benefit across n = 32 regions. A larger variance indicates a more unequal distribution.
Variances are normalized to [0, 1] using min–max normalization:
$${\sigma }_{\mathrm{sc}}^{2* }=\frac{{\sigma }_{\mathrm{sc}}^{2}-{\sigma }_{\min }^{2}}{{\sigma }_{\max }^{2}-{\sigma }_{\min }^{2}}$$
(35)
A normalized variance of 0 represents the most equal distribution, while 1 represents the greatest inequality. This indicator measures disparities in realized benefits but does not capture procedural equity or broader environmental justice considerations.
PV recycling subsidy scenarios
We develop alternative subsidy scenarios to encourage PV waste recycling and reduce excessive geographic concentration of recycling activities. The scenarios represent direct fiscal support and market-based carbon incentives commonly discussed in the solar and electronic-waste recycling literature73,90,91,92.
Five subsidy scenarios are evaluated: no subsidy, continuous subsidy, declining subsidy, low carbon price and high carbon price. The no-subsidy case is the reference scenario. Continuous and declining subsidies reduce the effective cost of recycling, whereas low- and high-carbon-price scenarios provide financial incentives based on avoided GHG emissions.
Cost-based subsidy scenarios
The continuous subsidy scenario assumes sustained government support throughout the project period. The subsidy is proportional to total recycling costs, including capital expenditure and operating expenditure:
$${S}_{i,t}^{\mathrm{cont}}={w}_{i,t}\times {\mathrm{TC}}_{i,t}^{\mathrm{total}}$$
(36)
Here, \({\mathrm{TC}}_{i,t}^{\mathrm{total}}\) is total recycling cost in region i and year t, while \({w}_{i,t}\) is the subsidy rate. Because region-specific policy data are limited, subsidy rates are sampled from 0–15%, consistent with the upper range reported for circular-economy investment programmes92. Both cost-based subsidy scenarios use 1,000 Monte Carlo simulations.
The declining subsidy scenario represents temporary support that decreases after recycling becomes economically viable90. Subsidies initially match the continuous subsidy rate and then decline annually by a fixed proportion after the break-even year:
$${S}_{i,t}^{\mathrm{decl}}={w}_{i,t}\times {\mathrm{TC}}_{i,t}^{\mathrm{total}}$$
(37)
$${w}_{i,t}=\left\{\begin{array}{c}{w}_{i,0}\times {(1-\delta )}^{(t-{t}_{i}^{* })},t > {t}_{i}^{* }\\ {w}_{i,0},t\le {t}_{i}^{* }\end{array}\right.$$
(38)
Here, δ = 0.10 is the annual subsidy reduction rate, \({t}_{i}^{* }\) is the break-even year, and \({w}_{i,0}\) is the 2020 subsidy rate.
Carbon-price-based PV recycling subsidies
The low- and high-carbon-price scenarios are based on policy pathways proposed by the International Energy Agency: the Stated Policies Scenario (STEPS) and Announced Pledges Scenario (APS), respectively95.
Under the high-carbon-price scenario, regions with net-zero commitments adopt carbon prices consistent with the APS pathway. Under the low-carbon-price scenario, carbon prices equal 30–80% of APS levels and apply only to regions that have implemented or formally announced carbon-pricing mechanisms, including Canada, South Korea, China and the European Union.
Carbon-based subsidies are proportional to avoided emissions from PV recycling:
$${S}_{i,t}^{\mathrm{carbon}}={P}_{i,t}^{{\mathrm{CO}}_{2}}\times {\mathrm{TEI}}_{i,t}$$
(39)
Here, \({P}_{i,t}^{\text{C}{{\rm{O}}}_{2}}\) is the regional carbon price, and \({\mathrm{TEI}}_{i,t}\) is the avoided carbon emissions generated by PV recycling. Carbon price data up to 2030 are obtained from the World Bank96, while values after 2030 are sourced from the IEA95.
Including subsidies in PV recycling net benefits
Subsidies are included in annual net benefits either by reducing effective recycling costs or by providing additional carbon-related revenue. The subsidized net benefit for region i in year t is:
$$\mathrm{SUB}\_{\mathrm{NB}}_{i,t,\mathrm{sc}}={S}_{i,t,\mathrm{sc}}+{B}_{i,t,\mathrm{sc}}^{\mathrm{total}}-{\rm{T}}{{\rm{C}}}_{i,t,\mathrm{sc}}^{\mathrm{total}}$$
(40)
Here, \({B}_{i,t,\mathrm{sc}}^{\mathrm{total}}\) is revenue from recovered materials; Si,t,sc is the subsidy; and \({{\rm{TC}}}_{i,t,\mathrm{sc}}^{\mathrm{total}}\) is total recycling cost, including capital and operating expenditures.
Evaluating the equality effects of PV recycling subsidies
We assess how subsidy policies affect regional disparities in PV recycling benefits using two indicators: variance in unit net benefits and the maximum–minimum benefit gap. Both indicators are calculated before and after subsidies for each scenario and year.
Pre-subsidy variance is calculated as:
$${\sigma }_{\mathrm{sc},t}^{2,\mathrm{pre}}=\frac{1}{N}\mathop{\sum }\limits_{i=1}^{N}{({\mathrm{UNB}}_{i,t,\mathrm{sc}}-{\bar{\mathrm{UNB}}}_{\mathrm{sc},t})}^{2}$$
(41)
Here, UNBi,sc,t is the unit net benefit of region i before subsidies, \({\bar{\mathrm{UNB}}}_{\mathrm{sc},t}\) is the regional mean, and N = 32 is the number of regions.
After subsidies, unit net benefits are recalculated as:
$$\mathrm{SUB}\_{\mathrm{UNB}}_{i,\mathrm{sc},t}=\frac{\mathrm{SUB}\_{\mathrm{NB}}_{i,\mathrm{sc},t}}{{\mathrm{PVWaste}}_{i,\mathrm{sc},t}}$$
(42)
Post-subsidy variance is then:
$${\sigma }_{\mathrm{sc},t}^{2,\mathrm{post}}=\frac{1}{N}\mathop{\sum }\limits_{i=1}^{N}{(\mathrm{SUB}\_{\mathrm{UNB}}_{i,\mathrm{sc},t}-{\overline{\mathrm{SUB}\_\mathrm{UNB}}}_{\mathrm{sc},t})}^{2}$$
(43)
The pre- and post-subsidy benefit gaps are:
$${\mathrm{Gap}}_{\mathrm{sc},t}^{\mathrm{pre}}={\max }_{i}({\mathrm{UNB}}_{i,\mathrm{sc},t})-{\min }_{i}({\mathrm{UNB}}_{i,\mathrm{sc},t})$$
(44)
$${\mathrm{Gap}}_{\mathrm{sc},t}^{\mathrm{post}}={\max }_{i}(\mathrm{SUB}\_{\mathrm{UNB}}_{i,\mathrm{sc},t})-{\min }_{i}(\mathrm{SUB}\_{\mathrm{UNB}}_{i,\mathrm{sc},t})$$
(45)
These metrics show whether subsidies reduce or increase disparities in regional PV recycling benefits.
Uncertainty and sensitivity analysis
We conduct uncertainty and sensitivity analyses to test the robustness of the modelling results and principal conclusions (Supplementary Note 4, Supplementary Table 23 and Supplementary Fig. 9).
The analysis examines PV module lifetime, material intensity, inflation and carbon price assumptions. Alternative lifetime and material-intensity scenarios are used to represent potential technological improvements. Both longer module lifetimes and lower material intensity substantially reduce future PV waste generation.
Macroeconomic uncertainty is assessed using alternative high- and low-inflation pathways. Inflation changes have only a limited effect on estimated PV recycling net benefits. We also test alternative carbon price growth rates for 2050–2060, particularly under the high-carbon-price subsidy scenario. High carbon prices consistently increase inequality in regional recycling benefits, confirming the robustness of the principal findings.
Reporting summary
Additional information about the research design is available in the Nature Portfolio Reporting Summary linked to this article.
Source: www.nature.com