The silver learning curve for photovoltaics and projected silver demand for net‐zero emissions by 2050

The clean energy transition could see the cumulative installed capacity of photovoltaics increase from 1 TW before the end of 2022 to 15–60 TW by 2050, creating a significant silver demand risk. Here, we present a silver learning curve for the photovoltaic industry with a learning rate of 20.3 ± 0.8%. Maintaining business as usual with a dominance of p‐type technology could require over 20% of the current annual silver supply by 2027 and a cumulative 450–520 kt of silver until 2050, approximately 85–98% of the current global silver reserves. A rapid transition to higher efficiency tunnel oxide passivated contact and silicon heterojunction cell technologies in their present silver‐intensive forms could increase and accelerate silver demand. As we approach annual production capacities of over 1 TW by 2030, addressing the silver issue requires increased efforts in research and development to increase the silver learning rate by 30%, with existing silver‐lean and silver‐free metallisation approaches including copper plating and screen‐printing of aluminium and copper.


| INTRODUCTION
In 2022, the world reached a cumulative photovoltaic (PV) installed capacity of 1 TW, 1 accounting for >4% of worldwide electricity demand. 2,3 However, techno-economic roadmaps [4][5][6] predict that to fulfil the Paris Climate Agreements to mitigate climate change, between 15 TW 6 and >60 TW 2,7 need to be installed by 2050. Annual growth rates for PV installations of 23-30% are required 2,8,9 to reach at least 15 TW by 2050, which the industry has consistently demonstrated for decades. 2 Details of projected scenarios can be found in Figure S1 and Table S1.
The global energy transition shifts material requirements from fossil fuels to different materials such as metals. As the land area requirement of coal mining is large, the energy transition can greatly reduce the requirement 10,11 despite the more intensive metal requirements, such as copper 12,13 increase sixfold by 2040. 14 While environmental impacts go beyond the directly affected area, the mining area correlates with overall impacts. 15,16 Present mainstream PV does not use rare-earth elements, but abundant metals such as aluminium and copper, required for multiple clean energy technologies. 17,18 The rapid growth of the PV market is not an option but a necessary step to mitigate climate change. However, such growth can lead to new challenges regarding material consumption. Concerns of surging material demands from PV production were raised for terawatt-Brett Hallam and Moonyong Kim contributed equally to the manuscript. level deployment of PV in 2008. 19 Based on the current rate of PV production, a number of studies and reports have highlighted the concern of increasing material demand, which will greatly impact supply chains and the long-term sustainability of PV manufacturing. [20][21][22][23][24][25] Currently, silver is the most critical metal posing price and supply risks when PV production expands. 9,19 In 2020, PV used approximately 12.7% of annual silver production, 18,26 despite the fact that only $3.2-8 g/m 2 of a PV module is needed. Many previous studies have highlighted that the current estimated silver consumption is too high to allow sustainable terawatt-scale production. [27][28][29] One particular study by Goldschmidt et al. highlighted material consumption learning rates (LRs) for PV including silver. 30 However, it was not clear what data were used to establish the LR of approximately 20%, and the study did not consider the impact of cell technology on silver consumption or the impact of a transition of the industry towards high-efficiency cell technologies on silver consumption by the PV industry.
While the potential for recovering silver from PV modules is significant, the current low collection and recovery rates, coupled with the 20-30% per annum growth rate of the PV industry and 25-year module lifetime, mean that recycled silver from PV modules can contribute only marginally to the silver supply for PV for quite some time.
All parts of the PV systems and modules can be dismantled mechanically, 31,32 recovered chemically 33,34 or via electrowinning. 35 The PV industry has room to facilitate recycling and optimise the design for reuse. Currently, there is a very limited recycling industry for PV modules because the number of end-of-life modules is still too small. 19 This is a general problem with technologies that are scaled up, such as batteries: Recycling technology lags behind until end-of-life volumes become sufficient to ensure a profitable business. We should also bear in mind that even some established technologies have low recycling rates, such as plastics, which makes a policy directed towards the circular economy urgently necessary.
The PV community is working on addressing the silver issue, by substitution with more abundant metals such as copper and aluminium. Copper has a similar conductivity as silver and is therefore a practical substitute, but there are still processing and reliability challenges to be solved before mass production. 2 Aluminium has a lower conductivity and challenges for forming a contact with n-type silicon without shunting the device, due to the alloying of aluminium and silicon during the high-temperature metallisation firing process to form a local p-type Al-doped silicon region. They are nevertheless feasible and a question of production costs and complexity.
In this work, we present a silver learning curve for PV based on the current industry's global silver consumption and module production, to project silver demand under different growth scenarios towards 2050. We consider the impact of cell technology and projected technology market shares on silver requirements by transitioning from p-type technology (e.g., passivated emitter and rear contact [PERC]) to n-type technology (e.g., tunnel oxide passivated contact [TOPCon] and silicon heterojunction [SHJ]) and the subsequent impact on global silver supply and reserves. The results show that the current rate of reduction in silver consumption is not sufficient to avoid increasing silver demand from the PV industry and that the transition to high-efficiency technologies including TOPCon and SHJ could greatly increase silver demand, posing price and supply risks.

| The silver learning curve for global PV deployment
As a whole, the PV industry has demonstrated a remarkable reduction in silver consumption over the past 10 years from a value 51.8-65.1 mg/W in 2010 to $19.5 mg/W in 2020 (see Figure 1A). A key driver for this reduction was manufacturing cost. Silver accounts for approximately 60% of the non-wafer cost 2 and 5-10% of the module manufacturing cost. For the emerging TOPCon and SHJ cell technologies (see Table S2), the cost of silver metallisation is even higher. Predictions of technology-dependent silver consumption per cell (CPC) F I G U R E 1 (A) Silver learning curve for the photovoltaic industry with silver consumption based on global reported silver use by the PV industry and global installed PV capacity also highlighting key global PV deployment scenarios. (B) Historical and projected silver consumption as a function of year under different scenarios along with predicted values of different PV technology from the 2021 ITRPV (IRV21). 2 For full details of scenarios for (A), see Table S1. are given annually in the International Technology Roadmap for Photovoltaic (ITRPV), with values expected to halve over the next decade 2 through gradual improvements in printing technology (see Table S2).
However, a key limitation of such projections is that they fail to fully account for the learning that comes from manufacturing an immense cumulative number of solar cells. 30 In 2021, we estimate that there were approximately 30 billion solar cells fabricated globally. Figure 1A shows the silver learning curve for global PV deployment, with the silver consumption (mg/W) reducing by 20.3 ± 0.8% for every doubling of cumulative installed capacity, consistent with estimations in reference, 30 although with a substantially higher pre-factor (see Table S5). This value is based on the last 10 years of data since 2010 (LR Recent ). When we consider all historical values, the long-term silver LR for the PV industry (LR All ) is estimated to be 18.7 ± 1.3%, which is slightly lower than the recent values. If the industry continues along the LR Recent learning curve in Figure 1A, for the broad electrification scenario in 2050, the estimated silver consumption would be in the range of 5.3 ± 0.5 mg/W. For PERC, if only using silver for metal fingers, this translates to a relative power loss of <0.36% rel with current pastes. If also requiring silver for busbars/tabbing regions, this equates in an excessive power loss of 2.76% rel . For TOPCon and SHJ, this would undesirably lead to greater power loss (12.9-24.9% rel ). Realising this lower limit without power losses will be challenging and require significant changes in cell metallisation and/or interconnection methods.
From this LR, we can estimate the silver consumption as a function of year under different industry growth scenarios (see Figure 1B).

| Technology-dependent silver consumption for solar cells
Although increased solar cell efficiencies present an immediate opportunity for lower silver consumption, silver consumption is primarily driven by the solar cell technology choice ( Figure 1B). For the industry-dominating p-type PERC cell (see Figure S4b), based on the first silicon solar cell to reach 25% efficiency, 36 silver is only required for the front n-type contact, while the rear p-type contact is formed using low-cost and abundant aluminium, with small amounts of silver for interconnection tabs, 37 providing an estimated module-level silver consumption of 14.4-15.7 mg/W in 2020 (see Table 1).
Emerging next-generation high-efficiency n-type TOPCon and SHJ solar cell technologies, with record efficiencies of 25.5% 41 and 26.3% 42 for two-sided contact devices, respectively, have a substantially higher requirement for silver. The current industrial implementation of TOPCon uses silver for the rear n-type contact as well as silver/Al for the front p-type contact to balance between optical and electrical resistive losses, which results in a silver consumption of 20.4-26.0 mg/W, 30-80% higher than that of PERC. SHJ solar cells use a low-temperature silver paste for both contacts with silver consumption reported in the range of 30.3-37.4 mg/W, more than double that of PERC (see Figure 2).

| Projected future cumulative silver demand in the PV industry
Due to the higher potential efficiencies of TOPCon and SHJ than PERC (see Table S2), these n-type technologies are expected to have It is also important to understand the impact of PV's silver consumption on global silver reserves. Figure 3 shows the cumulative sil-

| Annual silver demand projections for the PV industry
The future annual silver demand required by the PV industry will greatly depend on prior cumulative installed PV capacity through accumulated learning, technology choice and annual PV installed (see Figure 4). As shown in Figure 4A, only the most conservative scenario in the ITRPV (IRV21 Low) with $160 GW/year in 2030, which is less than the installed capacity in 2021, 44  the PV modules will hinder the material availability from recycling. To maintain silver demand within the PV industry to less than 10 kt/year ($43% annual silver supply), the silver LR must accelerate substantially to $30% and even higher at 30-40% for a shift towards silverintensive n-type technologies (see Figure 4B).

| DISCUSSION
The PV industry's average silver consumption in 2020 of 19.5 mg/W is approximately 20-27% higher than the estimate for PERC, despite PERC having more than 80% market share. 2  will not be feasible if also requiring silver busbars and tabbing regions. 37 Even for PERC, a 5-mg/W goal may impose challenges with reliability due to finger breakages with the small cross-sectional areas required. 37,47 Modified interconnection approaches to reduce silver consumption could also potentially introduce new material challenges such as bismuth. 37 To reduce the impact of global PV deployment on silver reserves, we must increase R&D investment for innovation with silver-lean PV technologies. Currently, a key driver for innovation in the PV sector is the manufacturing cost. For PV manufacturers, increasing solar cell efficiency is one generic method used to reduce manufacturing cost in terms of $/W, with follow-on benefits for the levelised cost of electricity through financial savings in balance-of-system components.
However, manufacturing costs and efficiency can also overlap with material consumption. Increasing efficiencies can reduce material consumption per unit of power (CPP) (mg/W), along with reducing the associated social and environmental impacts across the value chain of PV deployment, even for abundant materials such as copper, aluminium and steel.
The average yearly silver price has increased by 57-60% since 2019, 48 which can increase the cost of PV. To mitigate this risk, manufacturers can focus on adopting technologies that allow reduced silver consumption. The transition from the previous industry-dominating ptype aluminium back-surface field (Al-BSF) to PERC is a great example of this, in increasing cell efficiencies from 20% to over 23% without requiring additional silver (see Figure S3 and to accelerate the LR substantially to greatly reduce silver demand on the path to net zero by 2050 and innovate to provide feasible solutions towards silver reduction. One area of innovation is to improve silver utilisation by taking into account spatially varying resistive losses in devices. 37 To avoid any decrease in the silver LR of the PV industry as a whole, any major deployment of silver-intensive screen-printed ntype TOPCon and SHJ technologies must be balanced by a substantial deployment of silver-free or silver-lean TOPCon and SHJ solar cells. For screen-printed TOPCon cells, silver consumption could be greatly reduced by replacing the silver/Al p-type contact by a pure Al contact, similar to that of PERC. For SHJ solar cells, the existing low-temperature silver paste has a lower conductivity than hightemperature pastes used for PERC and TOPCon, which therefore requires more silver to achieve similar resistance. Innovation for these solar cells could focus on improving the conductivity of lowtemperature silver pastes. Alternatively, the use of copper pastes could greatly reduce silver consumption for both the front and rear contacts. However, despite pure copper exhibiting a similar bulk resistivity as pure silver, the conductivity of copper screen-printing pastes is substantially lower than that of silver pastes. One emerging approach is using silver-coated copper pastes, allowing a reduction in silver consumption by 30-50%. 49 of manufacturing should be based on copper plating. It could be argued that increasing the material supply or the reserve can overcome material sustainability issues. However, although reserve-to-production ratios for silver have been roughly constant since the 1950s, 20 typically the most economical resources to develop have already been mined and newly opened resources are more energy intensive to mine (e.g., are deeper or the ore grade is reduced). 20,57,58 This will 58 lead to a higher cost and a higher embodied energy (carbon footprint) of metal production, with a negative impact on deploying low-cost and sustainable PV. 59 Therefore, reducing material consumption is not only important to avoid shortage and cost reduction but also to minimise the increase of embodied energy per material. While it can be debated that using clean and renewable energy can compensate for the higher embodied energy for extraction, the energy should be used to decarbonise the electricity and not to compensate for the undesirable increase in embodied energy.
In the longer term, we must ensure that the recycling of PV panels recovers silver. With appropriate levels of recycling, and a stable long-term capacity of PV production, the embedded silver in solar panels may sustain future long-term PV production beyond 2050.

| Historical and projected silver demand in existing reports
The annual global silver consumption from the PV industry was obtained from the Silver Institute's 2020 report on the role of silver in  Table S1).
Estimates of the annual installed PV capacity were also used from the IEA PVPS.
A number of different scenarios for future PV deployment were considered in this work with cumulative installed capacities ranging from 2.02 to 63.4 TW by 2050 (see Table S1)  Table S1.
Where data for scenarios are not presented each year, such as for the four scenarios listed in the 2021 ITRPV (with data provided in

5-year increments), a linear interpolation of the annual PV installed
capacity is made between the provided data points.
A limited sensitivity analysis of the impact of growth rate of the PV industry on annual silver demand is performed by changing the growth rate of hypothetical scenarios with a fixed cumulative installed capacity by 2050 (see Figure S6).

| Conversion from yearly energy yield to a nominal power output of PVs
For scenarios listing either the nominal PV installed capacity additions or total PV energy yield, the conversion between nominal power output of PV and annual energy yield was assuming 1.2 PWh/TWp. For other scenarios, the published value is also included.

| Silver CPC, module and unit of power
From the annual PV capacity additions ( Figure S1a) and global consumption of silver from the PV industry, 26,44 the silver CPP in terms of mg/W was estimated using Equation (1), representing system-level silver consumption data for the PV industry: where ASD is the annual silver demand and APC is the annual PV capacity installed.
Technology-dependent silver consumption was obtained from a Equation (2).
where CPC is the silver CPC and P Cell is the power of a PV cell. Equation (3).
where n Cell is the number of equivalent full cells, CPC is the consumption of silver per cell and P Module is the power of a PV module.

| Silver learning curve for the global PV industry
The equation for the LR of silver consumption within the PV industry is given by Equation (4).
where A is a pre-factor for the value of the silver consumption at 1 TW of cumulative production; P Total is the cumulative installed capacity in terawatts; and LR Ag is the LR in per cent, representing the percentage of reduction in silver CPP, every time P Total doubles.
Curve fitting was used with Equation (4) where i is the cell technology (PERC, TOPCon or SHJ) and CPP i,2020 is the silver CPP of different technologies, which is the averaged value from Table 1 for each technology.
In this work, 'Industry' represents the PV industry in its entirety.
The Industry silver consumption is calculated using Equation (1) and the values from reported annual silver demand 26,44 and the annual PV production. Values for PERC, TOPCon and SHJ represent reported values for individual technologies in the ITRPV 2 and other sources. [38][39][40] 'N-type' represents cases with a transition of the PV industry to a domination of n-type technology (see Figure S4a). This assumes that two thirds of n-type technology market share is TOP-Con and one third is SHJ, consistent with forecasts in PV Magazine. 40 A value for overall n-type (A n-type ) = A PERC Â p-type% + n-type% (2/3 Â A TOPCon + 1/3 Â A SHJ ). p-type% and n-type% are the % of the PV market share from Figure S4a.
Curve fitting was used with Equation (4) in Origin Pro to determine the pre-factor and LR for the historical data for the global silver usage by the PV industry and cumulative installed capacity by minimising the root-mean-square error.
To calculate the cumulative silver demand (Ag Total ), which accounts for production in each year, n, and annual demand, Equation (6)

| Technology-dependent silver consumption learning curves
For estimating technology-dependent learning curves, the premium in silver consumption for the TOPCon and SHJ solar cell technologies over PERC was calculated from the silver CPC listed in historical ITRPV reports with the first data point in each respective report, along with future projections beyond this year. Similarly, the stabilised efficiency reported in the respective ITRPV reports is used to estimate the power at the cell level by taking into account the device area. Stabilised module efficiencies and module powers for the respective technologies are also obtained from the respective ITRPV reports.
From this, silver consumption is estimated in terms of mg/W. Average and standard deviations for the premiums of silver consumption for TOPCon and SHJ over PERC are then calculated accounting for all historical and predictions available at the module level. Table S2 shows the historical and projected silver consumption at the cell and module level. For Figure 4, the premiums in silver consumption were made above the global silver consumption of the PV industry, given the historical dominance of p-type technologies.