Main

Aluminium, the second most-used metal after steel, is integral to various industries1,2, including clean energy infrastructures3,4,5,6 such as photovoltaic panels7,8,9 and electric vehicles10,11, which are driving increased aluminium demand12. Owing to its high chemical reactivity, the energy required to produce aluminium can be up to ten times greater per tonne than that for crude steel13,14,15. The primary method for aluminium production is the Hall–Héroult process16,17, involving the dissolution of aluminium oxide in molten salts at ~960 °C and the application of electrical current to facilitate the reaction18.

Recent advancements have focused on optimizing cell designs19,20,21, electrode configurations22,23,24,25,26 and operational adaptability27,28,29 to reduce the energy intensity of the Hall–Héroult process30. However, the benefits of reduced energy intensity are offset by the increased use of fossil fuel-based captive power plants, which have reduced the share of non-fossil energy in the aluminium industry from 60% to 33% over the past three decades14. Additionally, the electrolysis process emits not only CO2 but also perfluorocarbons (PFCs), which have a substantially higher global warming potential than CO2 (refs. 31,32).

Given the growing demand for aluminium, technological innovations in production are important. Mitigation efforts at the facility level are crucial because of their direct impact on the production processes adopted. We developed a smelter-based bottom-up global aluminium GHG emissions inventory (CEADs-AGE), which includes 249 aluminium smelters and 280 associated fossil fuel-based captive power units. This inventory covers a wide range of technologies and configurations, providing a detailed assessment of emissions at the facility level. Our study utilizes the latest smelter survey data to compile GHG emissions inventory, identifying patterns and presenting tailored mitigation strategies for global aluminum smelters. Detailed methodology and data descriptions are provided in the Methods section.

Technologies and emission patterns

In 2021, global primary aluminium smelting capacity reached 81 Mt yr−1 and with total emissions of 651 MtCO2e, comprising 82% electricity-related emissions and 18% process-related emissions (Supplementary Fig. 1). Global aluminium smelters are increasingly adopting higher amperage, with smelters using currents >400 kA now accounting for 64% of smelting capacity in facilities <20 years old (Fig. 1b). The adoption of higher amperage technology is aimed at improving current efficiency and reducing energy intensity. Smelters with currents of [200, 300), [300, 400), [400, 500) and [500, 600) kA exhibit 4%, 5%, 10% and 11% lower energy intensity, respectively, compared to those <200 kA (purple dots in Fig. 1b). The 600+ kA cell, a breakthrough technology in the last decade, can reduce substantially smelting energy intensity to 12,630 kWh per tonne of aluminium, as demonstrated by the Weiqiao (Shandong) smelter in China, one of the most energy-efficient smelters today (Fig. 1b). Currently, 90% of the smelting capacity using amperages >400 kA is in China, making it the most energy-efficient country in terms of aluminium production (Fig. 1a).

Fig. 1: Technology and emission patterns of global aluminium smelters and their fossil fuel-based captive power units in 2021.
figure 1

a, Geographical distribution of aluminium smelters and fossil fuel-based captive power units. Icons may overlap because of the close proximity of many units. b, Energy intensity of smelters categorized by different age groups within each amperage category. c, Distribution of captive power units by installed capacity (left) and generation technology type (right), across different age groups. d, Electricity generation emission factors for coal-based (left, by unit) and gas-based (right, with the error bars showing maximum and minimum emissions intensity range from literatures) captive power units, categorized by generation technology. The average energy intensity of each smelter amperage group and the average electricity emission factor of each generation technology are shown by purple dots in b and d, respectively.

Full size image

In 2021, we identified 91 GW of operational captive thermal power capacity (Fig. 1c), supporting 49% global aluminium production. This share has increased by 12% over the past decade, up from 37% in 2012. Of this capacity, 70 GW was built after 2010, meaning that these units are still relatively young. The majority (85%) of total captive power capacity is coal-based, primarily subcritical units, which accounts for 51% of the total captive power capacity (Fig. 1c). Supercritical and ultrasupercritical units, which operate at higher pressures and temperatures for increased thermal efficiency33, make up 21% and 13% of captive generation capacity, respectively, and have 13–25% lower emission factors than subcritical technologies (Fig. 1d). Supercritical and ultrasupercritical units have seen increased adoption in China, particularly in the (6, 12] years age group. In India, 85% of aluminium capacity is supported by coal-based subcritical captive power units, combined with [300, 400) kA cell configurations, placing India at the top of the ranking for aluminium smelting emissions intensity (Extended Data Fig. 1). Gas-based captive power units constitute 15% of the total capacity, with 92% located in the Middle East. Notably, 90% of gas-fired power units use combined cycle gas turbine (CC) technology, which offers 12–15% higher efficiency than standard gas turbine (GT) technology by using waste heat for additional electricity generation (Fig. 1d).

Smelters in different regions, each operating with distinct cell designs and energy sources, face unique mitigation challenges, underscoring the need for tailored mitigation strategies for each facility.

Mitigation technology packages

The future GHG emissions of a smelter are influenced by the adopted mitigation technology, electricity source and the output level of the smelter34,35,36 (Extended Data Fig. 2).

Choice of mitigation technology for a smelter depends on the readiness level of the technology, the current cell configuration and the local climate mitigation ambitions. For example, the Alma smelter in Canada could be retrofitted with inert anode technology by 2030, leveraging the 450 kA industrial application programme led by its parent company, Rio Tinto37, and Canada’s goal of achieving carbon neutrality by 2050. In contrast, the Lanzhou aluminium smelter in China may not begin inert anode retrofitting until after 2040, as there is currently no evidence that its parent company, Chinalco, has mastered inert anode technology at an industrial level, and China’s carbon neutrality target is set for 2060 (subscenario E2; Methods).

The electricity source of a smelter is determined by its fossil fuel-based captive power plants and the regional electricity grid. Smelters typically prioritize self-generated electricity when available. For instance, the Aditya smelter in India may continue relying on its 6 × 150 MW coal-based supercritical power units, which began operation in 2014, until their decommissioning in 2044, assuming a 30 year lifespan38,39 (default subscenario for captive power plant lifetime, L30; Methods). After decommissioning, the Aditya smelter will transition to purchasing electricity from the regional power grid (Extended Data Fig. 3).

The production output of a smelter is influenced by its current capacity and the regional total output. Regional production is allocated to smelters on the basis of their energy efficiency, with the most efficient smelters prioritized until the demand for regional production is met. If existing capacity cannot satisfy the demand, new capacity will be constructed (Extended Data Fig. 3; Methods).

We compiled a key mitigation technology inventory for each of the 249 aluminium smelters, providing essential groundwork for differentiated mitigation packages (Fig. 2). This inventory aligns the compatibility of these technologies with the infrastructure of existing smelters. It includes essential retrofit technologies to optimize thermal balance, renew electrode configurations and enhance operational adaptability within current smelters. The inventory also introduces four cutting-edge cell designs for new smelting capacities and two electricity decarbonization strategies for all smelters. Each technology is documented in detail, with key features such as energy or emissions reduction potential, technology readiness level and technology suitability and compatibility (Supplementary Section 3).

Fig. 2: Key mitigation technologies for aluminium smelters.
figure 2

The shades of yellow represent the technology readiness level. The horizontal and vertical axes list the same set of technologies. Compatibility among technologies is indicated by colour: light green denotes compatibility, light pink denotes incompatibility, light blue indicates that the technologies can be used independently but are recommended to be combined for greater emissions reduction and dark blue indicates that the technology on the vertical axis is a prerequisite for the one on the horizontal axis. The numbers in the white areas represent the energy-saving potential of a retrofit technology, the energy intensity of new cell designs (in kWh kg−1 Al) or the emissions reduction potential (in %) of a specific technology. aDepend on the source of waste heat. bProcess-related emission reduction. cReduce cell smelting energy intensity to this value. dDepend on the specific situation of captive power plants and local power grid.

Full size image

We analyse how various strategies, and their combinations, could influence the emissions trajectory by treating each scenario as a combination of subscenarios (Methods). We considered five socioeconomic subscenarios by combining different shared socioeconomic pathways (SSPs) and representative concentration pathways (RCPs), including SSP360 (SSP 3–RCP 6.0), SSP260 (SSP 2–RCP 6.0), SSP245 (SSP 2–RCP 4.5), SSP126 (SSP 1–RCP 2.6) and SSP119 (SSP 1–RCP 1.9). Each pathway represents a distinct developmental trajectory characterized by higher gross domestic product (GDP), lower population growth rates and increasing shares of renewable energy. These factors affect aluminium production projections and electricity generation, thereby influencing the GHG emissions trajectory.

In addition, we introduce four sets of technology application subscenarios, which include three subscenarios considering different lifetimes of captive power plants (L); two commercialized technology subscenarios (C); two inert anode and wettable cathode technology subscenarios (E); and two innovative cells design subscenarios (I).This framework results in a comprehensive analysis encompassing 5 × 3 × 2 × 2 × 2 = 120 possible scenarios (Extended Data Fig. 2, orange part). From these, five technology narrative pathways are selected P1–P5 (Table 1). They are progressively related, with each one modifying part of the subscenarios, illustrating the emissions reduction potential of different mitigation strategies by cross-comparing them. Among them, P1 represents a business-as-usual (BAU) scenario without any additional mitigation strategies, while P5 represents the most aggressive mitigation scenario, in which the aluminium sector adopts all available strategies (see Methods for description of selected technology narrative pathways).

Table 1 Technology narrative pathway definition
Full size table

Emissions intensity decline of existing smelters

The emissions reduction potentials of various technology categories in four typical aluminium smelters under technology narrative pathways P1 and P5 are illustrated in Fig. 3 (Extended Data Fig. 4 provides details for P2, P3 and P4). Commercially available energy-saving technologies can achieve a 15–25% efficiency improvement for smelters, depending on their current cell configurations. However, the emissions reduction potential varies substantially among smelters, as it is closely tied to the source of electricity used. For smelters that predominantly rely on renewable energy sources, the reduction potential is limited. For example, the Alma smelter in Quebec, Canada, which already uses decarbonized electricity (with an electricity emission factor of ~0.001 kgCO2e kWh−1), shows no emissions reduction potential from energy-saving technologies (Fig. 3a, right). In contrast, energy-saving measures are crucial for smelters that depend on fossil fuel-based self-generated electricity, particularly in the short term. For instance, the Aditya smelter in India could achieve a reduction of 2.1 tCO2e t−1 Al by 2040 through the retrofitting of energy-saving technologies (Fig. 3d, right).

Fig. 3: Technology emissions reduction contributions of four typical smelters.
figure 3

a–d, The contribution of different technologies to reducing the emissions intensity of smelter Alma, Canada (a); Aldel, the Netherlands (b); Lanzhou Aluminium, China (c); and Aditya, India (d), under narrative pathways P1 (SSP245, L30, C1, E1, I1) (left) and P5 (SSP119, L20, C2, E2, I2) (right).

Full size image

Long-lifetime fossil fuel-based captive power plants can lock in high emissions intensity for smelters over the next 10–20 years, preventing them from benefiting from the gradual decarbonization of regional electricity grids. As shown in Fig. 3, smelters such as Lanzhou Aluminium in China and Aditya in India, which currently operate with coal-based captive power plants, experience substantial emissions intensity reductions only when their captive power plants are decommissioned and they transition to grid electricity. This shift could deliver a reduction of 5.0–10.5 tCO2e t−1 Al for smelters supported by fossil fuel-based captive power plants by reducing electricity emission factors by 50–100% (Fig. 3c,d). Conversely, smelters such as Aldel in the Netherlands, which do not rely on fossil fuel-based captive power plants, could see decreasing emissions intensity from 2020 to 2050 without additional mitigation efforts due to the increasing penetration of renewable energy in the regional grid (Fig. 3b, left).

Achieving carbon-free aluminium production requires the retrofitting of smelters with inert anode technology, which could reduce emissions intensity by 1.3–1.8 tCO2e t−1 Al. The Alma smelter in Canada may be the first to benefit from such retrofitting, thanks to the inert anode demonstration projects by Alcoa and Rio Tinto (Fig. 3a, right). However, in regions where technology innovation is limited and the pursuit of carbon neutrality is delayed, smelters may postpone adopting technologies that are still in the demonstration or development stage. For example, by 2050, despite the decommissioning of six coal-based captive power units after a 20 year lifespan, the Aditya smelter in India may still face 1.8 tCO2e t−1 Al unmitigated process-related emissions under pathway P5 (Fig. 3d, right).

Global primary aluminium industry emissions pathways

If the global aluminium industry maintains its current technology levels, without adopting additional mitigation technologies and with the regional electricity structure and production level following the SSP245 scenario, total emissions could reach 675 Mt yr−1 by 2050, with 22.4 Gt of cumulative emissions from 2020 to 2050 (P1, Fig. 4). Although cumulative emissions during this period could decrease to 15.2 Gt if the regional electricity structure follows the SSP119 scenario (P2), this would still result in total emissions of 203 Mt yr−1 and a global average emissions intensity of 2.0 tCO2e t−1 Al by 2050. This scenario would lead to slow mitigation progress before 2045, as many existing captive power plants continue operating without abatement (P2, Fig. 4).

Fig. 4: GHG emissions pathways of the aluminium industry.
figure 4

a–g, Total GHG emissions from the primary aluminium smelting sector in global (a), Europe (b), North America (c), China (d), India (e), Middle East (f) and Africa (g). The light shaded area represents the range of highest and lowest emissions pathways under the five SSP–RCP scenarios. The bold lines depict the five technology narrative pathways. The bar chart represents the emissions structure under P5, illustrating the most climate-ambitious scenario for the global aluminium industry, which adopts all possible technological strategies to mitigate GHG emissions.

Full size image

Retiring fossil fuel-based captive power plants a decade earlier than planned could result in an additional cumulative emissions reduction of up to 3.4 Gt (P2 versus P3, Fig. 4a), accounting for 28% of the total cumulative emissions reduction from P1 to P5. Reducing the lifespan of captive power plants from 30 to 20 years would require 26% of currently operating captive power units to cease operations by 2030 and 53% to be shut down between 2030 and 2035. This strategy would lead to cumulative emissions reductions of 2.9 Gt for China and 0.3 Gt for India during 2020–2050 (P2 versus P3, Fig. 4d,e).

While power decarbonization through abated fossil fuel-based captive power plants and green grid is necessary, it is not sufficient to achieve carbon-free aluminium production. Additional cumulative emissions reductions of 1.4–5.1 Gt require technological innovations within the aluminium sector itself (Supplementary Fig. 3a–c). Enhancing energy efficiency by retrofitting existing smelters and adopting innovative cell designs for new smelters could contribute to 0.5–4.2 Gt cumulative emissions reduction, depending on the electricity source used (P3 versus P4; Fig. 4a and Supplementary Fig. 3a,c).

Implementing inert anode and wettable cathode technologies is crucial to reducing process-related emissions, which currently account for 18% of total GHG emissions from the primary aluminium industry. As electricity sources become more decarbonized, this percentage will increase, making process-related emissions mitigation even more critical. These technologies could achieve 0.8–0.9 Gt in cumulative emissions reduction under different output scenarios (P4 versus P5; Fig. 4a and Supplementary Fig. 3b). This requires retrofitting at least 18% of smelting capacity with inert anode technology by 2040, increasing to 67% by 2050. The high readiness level of inert anode technology in Europe and North America allows for its adoption in these regions before 2035, supported by several pilot projects40,41,42. However, future expansions of primary aluminium production are likely to be concentrated in developing regions such as India, the Middle East and Africa, driven by rapidly growing domestic demand. Given the current lack of a strong foundation for technology R&D in these areas, and with commitments to achieving carbon neutrality extending beyond 2050, new smelting capacity in these developing regions is unlikely to adopt inert anode technology extensively until after 2040 (Fig. 4e–g).

In regions where self-generated electricity is the primary source for aluminium production, China is expected to experience a rapid decrease in emissions across all scenarios, primarily due to the retirement of fossil fuel-based captive power plants and the transition to a cleaner energy generation structure (Fig. 4d). However, other emerging economies face substantial mitigation challenges as their aluminium production grows. Under the BAU scenario (P1), total emissions in India, the Middle East and Africa are projected to increase by 287%, 48% and 314%, respectively, by 2050 compared to 2020 levels (P1, Fig. 4e–g). This highlights the critical need to focus on technological choices for capacity expansion in these emerging economies.

In the short term, key mitigation strategies for aluminium producers in India and the Middle East include halting the construction of new fossil fuel-based captive power plants and planning the retirement of existing ones. In the long term, as electricity generation becomes largely decarbonized, process-related emissions will become the dominant concern, including in Europe and North America (Fig. 4b,c). Therefore, future mitigation efforts in these regions should prioritize technological innovations that address process-related emissions.

Discussion and conclusion

By developing the CEADs-AGE, we identified the sources of emissions from the global primary aluminium industry from the smelter level upwards. This allowed tailored differentiated mitigation technology packages for global aluminium smelters based on their current cell amperage and electricity sources.

The historical total GHG emissions uncertainty for smelters ranges from 5.3% to 45.3%, with 88% of smelters having uncertainty <20% (Supplementary Fig. 14). This uncertainty may stem from variations in surveyed activity levels, captive power plants and the adopted electricity and process-related emission factors (Supplementary Section 6.1). To improve future accounting accuracy, it is crucial to enhance the spatial and temporal resolution of monitoring smelter electricity structures and the electrolysis process, especially in less developed regions.

Sensitivity analysis indicates that projected cumulative emissions from 2020 to 2050 could vary by −15.5% to +5.6%, depending on the schedule43 of technology diffusion, energy-saving potential and assumptions about scrap recycling and manufacturing yield44 (Supplementary Section 7). Regional primary aluminium output levels may be influenced by the patterns of trade in intermediate products, final goods and recycled scrap. Our study has not yet fully integrated trade and recycling factors into the model, which could introduce fluctuations in regional-level projections. Additionally, new climate pledges, policies, research and pilot projects could alter the pathways of the aluminium sector. Although our study modelled a broad range of SSP–RCP scenarios and diverse technology portfolios, it remains essential to track new trends and continuously incorporate them into future pathway modelling.

Our study focuses solely on GHG emissions from primary aluminium smelting, which accounts for 64–80% of the cradle-to-grave emissions of the entire aluminium production process45,46,47,48. Other studies have explored strategies such as electrifying or switching to hydrogen-based boilers and calciners to mitigate emissions from alumina refineries17,49, decarbonizing the casting process50 and pursuing sustainable bauxite mining practices51,52. Future mitigation efforts should encompass the entire supply chain to ensure a systematic approach to decarbonization.

Methods

Global aluminium GHG emissions inventory

To construct CEADs-AGE, we began with data from BAIINFO (http://www.baiinfo.com/) based on a smelter survey, which provides monthly data (2018–2021) and yearly data (2012–2017) on smelter capacity, production, amperage configuration, energy intensity and the proportion of self-generated and purchased electricity (the proportion data available only for smelters in China). We then acquired a coal- and gas-based power plant inventory from the Global Energy Monitor (https://globalenergymonitor.org/), which identified 594 captive power units associated with the aluminium industry. By comparing the geographical locations and ownership information of smelters and captive power units, we established the affiliations. After excluding captive power plants linked to alumina refineries and those no longer in operation, we identified 280 operating captive power plant units in 2021.

GHG emissions from primary aluminium production are categorized into four main types: (1) CO2 emissions from self-generated electricity; (2) CO2 emissions from grid-purchased electricity; (3) process-related CO2 emissions; and (4) process-related PFCs (including CF4 and C2F6) emissions. The total emissions of an aluminium smelter are the sum of these four types:

$${E}_{s,y}={{\rm{EE}}_{\mathrm{S}}}_{s,y}+{{\rm{EE}}_{\rm{G}}}_{s,y}+{{E}_{\rm{{PC}}}}_{s,y}+{{E}_{\rm{{PFC}}}}_{s,y}$$
(1)

Where subscript s and y represents the smelter and year, respectively. E, EES, EEG, EPC and EPFC represent total CO2 equivalent emissions, self-generated electricity CO2 emissions, grid-purchased electricity CO2 emissions, process-related CO2 emission and PFCs emission (in CO2e), respectively.

CO2 emissions from self-generated and grid-purchased electricity are estimated using equations (2) and (3), respectively:

$${{\rm{EE}}_{\rm{S}}}_{s,y}={\rm{AD}}_{s,y}\times {\rm{EI}}_{s,y}\times {S}_{s,y}{\times {{\rm{EF}}_{\rm{S}}}}_{s,y}$$
(2)
$${{\rm{EE}}_{\rm{G}}}_{s,y}={\rm{AD}}_{s,y}\times {\rm{EI}}_{s,y}\times \left(1-{S}_{s,y}\right){\times {{\rm{EF}}_{\rm{G}}}}_{s,y}$$
(3)

where ADs,y represents activity data (here referring to production) of smelter s in year y; EIs,y represents energy intensity of smelter s in year y (in kWh t−1 Al); Ss,y represents the share of self-generated electricity in the energy supply of smelter s in year y (in %); and EFSs,y and EFGs,y represent self-generated and grid-purchased electricity emission factor of smelter s at year y, respectively (in kgCO2 kWh−1).

The activity data, AD, and electricity energy intensity, EI, for smelters are obtained directly from the smelter survey.

For smelters in China, our survey data include the proportion of self-generated electricity. However, for smelters outside China, this information is not available. In these cases, we assume that smelters prioritize using electricity from their captive power plants and only purchase electricity from the local grid if the captive power capacity is insufficient to meet electricity demand. For smelters without captive power plants, we assume that all electricity is sourced from the local grid.

The electricity emission factor of each coal-based captive power plant (\({{\rm{EF}}_{\rm{S}}}_{s,y}\)) is obtained from Global Energy Monitor53. The electricity emission factor of gas-based power plants is obtained from the literature54,55,56,57,58,59. For the purchased grid electricity emission factor (\({{\rm{EF}}_{\rm{G}}}_{s,y}\)), we prioritized data released by government and further collected subnational grid electricity emission factors for Australia60, Canada61, China62 and the United States63 to replace national-average value since these countries have large territories with several relatively independent power grids; therefore, subnational value could reflect regional variations in electricity supply and emission patterns. For countries where government-released data are unavailable, particularly for less developed countries, we use third-party evaluations, including (1) Carbon Footprint64; (2) Climate Transparency65; and (3) Low-Carbon Power66. Different data sources are cross-compared and either take the average of the three or exclude outliers to calculate the average value for these countries. Supplementary Section 2 describes the details of electricity emission factor collection.

The process-related CO2 emission from primary aluminium production comes from the consumption of carbon anode during electrolysis:

$$\rm{2{Al}}_{2}{O}_{3}+3C\to 4{Al}+3{{CO}}_{2}$$
(4)

We use the IPCC-recommended tier 1 method for process-related CO2 emission estimation31:

$${{E}_{\rm{PC}}}_{s,y}={\rm{AD}}_{s,y}\times {{\rm{EF}}_{\rm{PC}}}_{s,y}$$
(5)

where \({{\rm{EF}}_{\rm{PC}}}_{s,y}\) represents emission factor of process-related CO2.

The process-related PFCs emission comes from the anode effect during smelting. Liquid alumina must be added to the cells periodically to maintain a stable alumina concentration level since alumina is consumed during the electrolysis. Too low/high alumina concentration would result in the gaseous electrical insulation layer around carbon anode and increase of cell voltage, leading to the anode gas composition change from CO2 to CO and PFCs31, which is known as the anode effect. We used the IPCC-recommended tier 1 method31 for process-related PFCs emission estimation as follows:

$${{E}_{\rm{PFC}}}_{s,y}={\rm{AD}}_{s,y}\times {{\rm{EF}}_{\rm{PFC}}}_{s,y}$$
(6)

where \({{\rm{EF}}_{\rm{PFC}}}_{s,y}\) represents emission factor of process-related PFCs (in tCO2e t−1 Al).

The process-related CO2 emission factors are from International Aluminium Institute (IAI)47. The PFCs emission factors for smelters in China are from our previous study67, which adjusts the emission factor according to technology structure in China. The PFCs emission factors for smelters beyond China are from IAI47. Supplementary Table 2 lists all the adopted process-related emission factors.

Emissions projection of global aluminium smelters

Similar to GHG emissions accounting framework stated above, the future emissions projection of each smelter is also based on seven parameters under three categories: mitigation technology (includes energy intensity, process-related PFCs and CO2 emission factors), electricity source (includes share of self-generated electricity, emission factor for self-generated and grid-purchased electricity) and smelter production (Extended Data Fig. 2). Here we set a series of subscenario combinations to explore the possible future trajectory of each smelter.

We examined five socioeconomic development subscenarios by combining various shared SSPs and RCPs to provide a detailed picture of potential futures by linking socioeconomic development with climate forcing. This impacts aluminium production projections and electricity generation pathways, thereby influencing the GHG emissions trajectory of primary aluminium production. The five typical SSP–RCP combinations are:

  • SSP360: SSP3–RCP6.0. High population growth but low GDP growth. Relatively high fossil fuel-based with slow progress in renewables.

  • SSP260: SSP2–RCP6.0. Moderate GDP and population growth. Relatively high fossil fuel-based with slow progress in renewables.

  • SSP245: SSP2–RCP4.5. Moderate GDP and population growth. A balanced energy mix with gradual growth in renewables.

  • SSP126: SSP1–RCP2.6. High GDP growth, low population growth. Rapid transition to renewables, substantial reduction in fossil fuel.

  • SSP119: SSP1–RCP1.9. High GDP growth, low population growth. Near-total renewable energy dominance, aggressive decarbonization.

    The five SSP–RCP combination pathways are simulated using GCAM v.6.0, a widely used integrated assessment model68,69,70,71. GCAM v.6.0, released in 2022, divides the world into 32 regions, uses 2015 as the base year and projects the future in 5 year time steps72. Besides the five SSP–RCP subscenarios (SSP), we also introduce four technology application subscenario sets, including:

  • Three scenarios considering different lifetimes of captive power plants (L).

  • Two commercialized technology subscenarios (C).

  • Two inert anode and wettable cathode technology subscenarios (E).

  • Two innovative cells design subscenarios (I).

Each subscenario set is used to determine one or several parameters. Extended Data Fig. 2 summarizes how each subscenario impacts the parameters and thus the GHG emissions from the aluminium smelter. We treat each scenario as the different combinations of the individual subscenarios (SSP, L, C, E, I) such that we will have 5 × 3 × 2 × 2 × 2 = 120 possible scenarios.

Scenario settings

Socioeconomic pathway

The GHG emissions of each smelter are influenced by three macro-socioeconomic factors: (1) electricity cost, (2) energy mix of electricity generation and (3) primary aluminium production (Extended Data Fig. 3) as outlined below.

Decision on new captive power plants

The decision to equip newly built smelting capacity with captive power plants is based on a cost comparison between fossil fuel power generation (mainly coal and gas) and grid-purchased electricity. Smelters typically opt for the cheaper electricity source. Detailed modelled electricity costs are provided in Supplementary Figs. 8–12.

Grid-purchased electricity emission factors

For smelters without captive power plants, the emission factor for grid-purchased electricity depends on the regional electricity generation mix. We assume that the grid electricity emission factor follows the same trend for all smelters within the same region. As a result, we apply the regional electricity emission factor trend to each smelter in the corresponding region. The modelled trend is shown in Supplementary Fig. 7.

Smelter-level output

Smelters adjust their production to meet the demand for regional primary aluminium production. If the current operating capacity is insufficient to meet this demand, new capacity must be built. Achieving a 100% capacity factor is challenging as a result of maintenance, energy supply fluctuations and operational inefficiencies, leading to actual operating capacity being lower than installed capacity. Therefore, the new-built smelting capacity of region r in year y is calculated as follows:

$$\begin{array}{l}{\rm{New}}\_{\rm{Capacity}}_{r,y}=0\,({\rm{if}}:{\rm{Capacity}}_{r,y-1} U_f\ge {\rm{Output}}_{r,y})\\ {\rm{New}}\_{\rm{Capacity}}_{r,y}=({\rm{Output}}_{r,y}-{\rm{Capacity}}_{r,y-1}U_f)/U_f\,\\\qquad\qquad\qquad\qquad\;\;({\rm{if}}:{\rm{Capacity}}_{r,y-1}U_f < {\rm{Output}}_{r,y})\end{array}$$
(7)

Where \({U}_{f}\) represents maximum utilization factor of smelters. Currently, global average utilization factor is 85%; we assume this capacity utilization factor will keep in the projection. Total aluminium smelting capacity of region r in year y can be obtained as:

$${\rm{Capacity}}_{r,y}={\rm{Capacity}}_{r,y-1}+{\rm{New}}\_{\rm{Capacity}}_{r,y}$$
(8)

We allocate the regional primary aluminium output to each smelter according to both its efficiency and capacity. If the regional output reaches the maximum operating capacity, then each smelter should operate at its maximum utilization factor. Otherwise, the smelters in the region are ranked from lowest to highest in terms of energy intensity, and this ranking is used to prioritize production to the most efficient smelters. AO represents the ascending order. The order of smelters in terms of energy intensity from lowest to highest is as follows:

$${\bf{s}}={\rm{AO}}({\rm{EI}}_{s,y})$$
(9)

The aluminium output of the nth ranked smelter in region r and year y can be written as:

$$\begin{array}{l}{\rm{AD}}_{n,y}={\rm{Capacity}}_{n,y}U_f\,({\rm{if}}:{\rm{Output}}_{r,y}\ge \mathop{\sum }\limits_{s=1}^{n}{\rm{Capacity}}_{s,r,y}U_f)\\ {\rm{AD}}_{n,y}={\rm{Output}}_{r,y}-\mathop{\sum }\limits_{s=1}^{n-1}{\rm{AD}}_{s,r,y}\,\left({\rm{if}}:{\rm{Output}}_{r,y} > \mathop{\sum }\limits_{s=1}^{n-1}{\rm{Capacity}}_{s,r,y}U_f\wedge \right.\\\qquad\qquad\left.{\rm{Output}}_{r,y} < \mathop{\sum }\limits_{s=1}^{n}{\rm{Capacity}}_{s,r,y}U_f\right)\\ {\rm{AD}}_{n,y}=0\,({\rm{if}}:{\rm{Output}}_{r,y} < \mathop{\sum }\limits_{s=1}^{n-1}{\rm{Capacity}}_{s,r,y}U_f)\end{array}$$
(10)

The future aluminium output growth is driven by GDP, income elasticities and price elasticities in GCAM:

$${{\rm{Output}}}_{r,t}={\rm{Output}}_{r,t-1}{\left(\frac{{Y}_{r,t}}{{Y}_{r,t-1}}\right)}^{\alpha }{\left(\frac{{P}_{r,t}}{{P}_{r,t-1}}\right)}^{\beta }\left(\frac{{N}_{r,t}}{{N}_{r,t-1}}\right)$$
(11)

where Y is the per capita GDP, P is the price of industry service (here, mainly primary aluminium price), N is the population and α and β are income and price elasticities, respectively.

The modelled regional primary production output can be found in Supplementary Fig. 6. Supplementary Fig. 23 gives the comparison of primary aluminium production projection with existing literature.

Captive power plant lifetime

The lifetimes of captive power plants determine when an existing unit will cease operation, thereby affecting the share of self-generated electricity. Previous studies indicate that the typical lifetime of thermal generation units ranges from 30 to 50 years38,39, with the widest reported range being 20–60 years73. To evaluate the impact of shorter or longer operational periods of captive power plants on the GHG emissions of the aluminium sector, we define three subscenarios:

  • L40: lifetime = 40 years.

  • L30: lifetime = 30 years.

  • L20: lifetime = 20 years.

Commercialized technology

The implementation of commercialized energy-saving technologies will influence the energy intensity of each smelter. Figure 2 illustrates the range of energy-saving potential for each technology, while Supplementary Section 3 provides a detailed description. We define two subscenarios for commercialized technologies:

  • C1: without the application of commercialized technologies.

  • C2: with the application of commercialized energy-saving technologies for retrofitting existing smelters. The technology diffusion follows an S-shaped logistic curve (Supplementary Section 4) and we apply the average energy-saving potential of each technology.

Inert anode and wettable cathode

The adoption of inert anode and wettable cathode technology will result in process-related CO2 and PFCs emissions being reduced to zero (Supplementary Sections 3.14 and 3.15). We define two subscenarios for the application of this technology:

  • E1: without the application of inert anode and wettable cathode technology

  • E2: with the application of inert anode and wettable cathode technology to all smelters, once the necessary technology readiness level is reached

Here, E represents the retrofitting of electrodes (anode and cathode). The timing for when a smelter can be retrofitted with inert anode and wettable cathode technology depends on the technology readiness level of the parent company and the climate ambition of the region. The diffusion rate of the technology follows an S-shaped logistic curve (Supplementary Section 4).

Innovative cell design

The application of innovative cell design will affect the energy intensity of newly built smelting capacity (Supplementary Section 3). We define two subscenarios for the adoption of innovative cell design:

  • I1: without the application of innovative cell design

  • I2: with innovative cell design applied to new smelting capacity once the technology reaches the necessary readiness level

Description of selected technology narrative pathways

Pathway P1 (SSP245, L30, C1, E1, I1) is a BAU scenario. Global primary aluminium production levels and regional electricity generation pathways are based on the SSP245 scenario, which is widely used as a baseline in research74,75,76 and aligns with current global development trends77,78,79. No additional mitigation technologies are adopted within the aluminium sector and the lifetime of captive power plants follows the default value of 30 years.

Pathway P2 (SSP119, L30, C1, E1, I1) is based on P1 but shifts the aluminium production projection and regional electricity pathways to SSP119. It represents a scenario where the aluminium sector does not undertake any additional mitigation efforts, with emissions decreasing as a result of the decarbonization of grid power.

Pathway P3 (SSP119, L20, C1, E1, I1) builds on P2; this pathway shortens the lifespan of captive power plants from 30 to 20 years.

Pathways P4 and P5 each build on the previous one by incorporating additional mitigation strategies. Specifically, they adopt commercialized technologies and innovative cell design to improve smelter energy efficiency, as well as inert anode and wettable cathode technologies to mitigate process-related emissions. Pathway P5 (SSP119, L20, C2, E2, I2) represents the most aggressive mitigation scenario, where the aluminium sector adopts all possible strategies, including both current and future technologies, early retirement of captive power plants and a high penetration of renewable energy in the regional electricity generation mix, as outlined in SSP119.