The need for a critical mineral demand model incorporating technical change

Introduction

Studying the effects of technical change on critical mineral demand and supply in the context of the low-carbon energy transition is an important and open area of research. Despite the crucial role played by these minerals in low-carbon technologies, long-term demand projections remain uncertain due to intricate interactions between drivers of technical change. In this writeup, I lay out what a framework that studies the effects of technical change on critical mineral demand would look like, how it can be developed, and what are its potential use cases.

There is a need to develop a critical mineral demand framework that incorporates endogenous technical change driven by phase shifts, as well as improvements in mineral intensities, energy efficiencies, recycling rates, life spans, and second-life proportions. Such a framework would need to be driven by a multi-sector computable general equilibrium (CGE) model of the low-carbon production network with critical minerals as inputs. Moreover, the framework could potentially be enabled by material flow analysis (MFA) of these minerals embodied in crucial energy technologies and machine learning (ML) based parameter learning from critical mineral data. Such a framework, once developed, will be able to provide policy-relevant insights on opportunities and risks relating to the critical minerals value chain under alternative technical change scenarios.

Minerals such as lithium, cobalt, nickel, graphite, and rare earth elements (REEs) play a crucial role in supporting low-carbon technologies essential for addressing climate change, including electric vehicles, wind turbines, solar photovoltaics (PVs), and grid-scale energy storage. These minerals are classified as critical due to their significance for economic and energy security, with their value chains vulnerable to disruptions. The International Energy Agency (IEA) projects a substantial increase in demand for these critical minerals in the coming decade [1], driven by the growing adoption of low-carbon technologies and the decarbonization scenarios outlined by the Intergovernmental Panel on Climate Change (IPCC) [2]. For instance, canada has recognized the importance of these minerals and developed a critical minerals strategy [3] and a Canadian minerals and metals plan (CMMP) [4]; however, long-term demand projections are subject to significant uncertainties from complex interactions between drivers of technical change. Several models have estimated the critical minerals needed for deploying low-carbon technologies across alternate decarbonization scenarios [5-10]. The design of these critical mineral demand models usually involves considering four key elements [11]: (a) anticipated deployment of future low-carbon technologies, (b) existing and prospective low-carbon sub-technologies, (c) material intensities associated with sub-technologies, and (d) effect of technical change.

The accuracy of critical mineral demand estimates is significantly influenced by the specific technologies considered in the assessments, baking in assumptions about sub-technologies, deployment shares, and related infrastructure [11]. Inconsistencies in these assumptions across different models lead to varying conclusions regarding which critical minerals a country should invest in. These inconsistencies also extend to assumptions about the future deployment shares of specific low-carbon sub-technologies and improvements in mineral intensities, energy efficiencies, recycling rates, life spans, and second-life proportions. The alternate futures imagined through these assumptions significantly impact critical mineral demand projections [11] and are all related to technical change. Therefore, the central issue is that many of the current models focus mainly on supply shortages, and while this is important, it is crucial to model technical change better, as it has a pervasive influence on every stage of modeling.

The modern canonical version of directed technical change (DTC) theory [12], when initially introduced by Acemoglu in 2002, focused on labor versus capital augmenting technical change, as dictated by the opposing effects of market size and price. Later, this theory was extended to analyze clean and dirty production factors [13]. Since then, endogenous technical change has been modeled using the DTC framework to study many low-carbon settings, including the electricity sector [14], mineral extraction efficiency improvements [15], energy efficiency improvements [16-18], and path dependence [19]. Some models focus on critical minerals in CGE models [20,21]; however, none study them as production factors in a low-carbon production network [22] with endogenously modeled technical change. Therefore, there is a need to develop a framework that factors in endogenous technical change and its drivers in the low-carbon production network and study equilibrium effects on critical mineral input factors. To fully understand the impact of endogenizing technical change, it is also crucial to contrast the findings of this model and traditional MFA [23,24] with network equilibrium effects and scenario-based exogenous technical change. Such a model is of immediate relevance to critical mineral strategy and infrastructure development.

A Potential Framework

A promising critical mineral demand framework would combine endogenous technical change and critical mineral data-driven parameter learning within CGE models [25] of the low-carbon production network, with critical minerals as primary input factors. Top-down CGE models capture the interdependence between sectors, making them suitable for this context compared to bottom-up models. While technical change can be incorporated exogenously or endogenously in CGE models, an endogenous approach using R&D investment towards different directions of technical change in an endogenous production network offers advantages. The endogenous production network could be constructed by allowing the production of several low-carbon technologies, combining critical minerals and an endogenous subset of other low-carbon sub-technologies as inputs [22]. Initial production network configuration would be informed by MFA of low-carbon technologies and concepts from the physical stocks and flows framework (PSFF) [24]. Meanwhile, learnable parameters of the endogenous production network could be estimated from critical mineral and low-carbon production data using ML models that handle time and network dependency [26, 27], offering more sophistication than simple OLS regression.

This framework would enable studying equilibrium effects on critical mineral demand and supply under multiple potential futures of technical change. These futures include technological phase shifts and improvements in mineral intensities, energy efficiencies, recycling rates, life spans, and second-life proportions. Equilibrium effects could be studied by endogenizing these directions of technical change through R&D investment in technology efficiency, durability, and recycling technology. Comparing results with traditional MFA incorporating scenario-based exogenous technical change would illuminate the effects of endogenizing technical change.

The framework could also explore implications of technical change on current critical mineral infrastructure development and strategic policy worldwide. For instance, the Canadian critical minerals strategy [3] does not fully account for various technical change scenarios, representing a vulnerability for the Canadian low-carbon transition. As an exmaple, an efficiency improvement in lithium-iron-phosphate (LFP) batteries compared to lithium-ion batteries with nickel-manganese cobalt (NMC) cathodes could lead to drastic reductions in demand for cobalt, manganese, and nickel and a significant increase in demand for copper [11]. The proposed framework would help governments navigate the vast and uncertain technical change landscape, addressing such vulnerabilities.

Methods

The low-carbon production network could be created by considering a wide range of low-carbon technologies, including hydroelectric, wind turbines, solar PVs, transmission lines, zero-emission vehicles (ZEVs), grid-scale energy storage, coal and natural gas with carbon capture and storage (CCS), hydrogen fuel cells, and direct air capture (DAC), among others. These technologies would comprise their constituent technologies, such as motors and batteries for ZEVs, electrolyzers for hydrogen fuel cells, and adsorbents for CCS and others. Each of these constituent technologies would comprise current and promising sub-technologies. For instance, batteries would mainly include lithium-ion, solid-state, lithium-sulfur, sodium-ion, lithium-iron-phosphate, iron-air, and zinc-based. The critical minerals required to make these sub-technologies would serve as inputs to the sub-technologies. The configuration of this low-carbon production network would need to be reconciled with established CGE models like GCAM [25], reduced according to the needs to serve as a starting point for the production network simulations.

The initial quantities of the critical minerals could be estimated (in kg/MW) by looking at the current technology and sub-technology mix that contributes to the total energy demand and conducting MFA using open-source tools like the open dynamic material systems model (ODYM) [23] and techniques from PSFF [24]. ODYM allows consideration of the product, component, material, and chemical element levels simultaneously and can support extensions related to efficiencies and lifetimes [23]. Importantly, no assumptions should be made about the future deployment shares of different technologies and sub-technologies. Instead, they need to be altered by endogenously modeling three drivers of technical change, technology efficiency, technology durability, and recycling technology, through the surrogate of R&D investment in these drivers.

Knowledge capital has been previously integrated into CGE models to capture endogenous technical change, with R&D investment as a surrogate for measuring knowledge capital [14,28]. Similarly, R&D investment could be used as an input in the production function to connect knowledge capital accumulation within the low-carbon production network with the dynamics of technical change. Three major drivers of technical change would need to be considered, including technology efficiency, technology durability, and recycling technology, which would enable modeling of most phenomena we care about. Technology efficiency is directly related to mineral intensity, where an increase in technology efficiency means a lesser quantity of minerals required to meet total energy demand. Furthermore, when one technology becomes more efficient than the other, a technological phase shift occurs in the production network as a reaction to the “positive technology shock” [22]. How the equilibrium production network changes due to various technological phase shifts and the secondary effects of these phase shifts, for instance, through reduced prices, is an interesting direction to explore. Mechanisms need to be designed to identify these phase shifts before they occur, serving as lamp posts that prepare us for altered states of the low-carbon production network. The other two drivers, technology durability and recycling technology are connected. Technology durability is related to the life spans of technologies and whether they will be suitable for second-life applications [24], and recycling technology dictates the quantity of primary mined critical minerals required for a particular sub-technology. Therefore, a sub-technology can be made from primary mined critical minerals, recycled critical minerals, or second-life technologies, resulting in a sum constraint. A similar sum constraint exists for the mix of various sub-technologies equaling total energy demand, which grows according to known energy scenarios like the ones by the IEA [1].

Moving away from the conventional ordinary least squares (OLS) regression for estimating the learnable parameters of the endogenous production network, more sophisticated machine learning (ML) models designed to handle time series data within graph/network structures could be adopted. Specifically, graph neural networks (GNNs) [26] and their time series variants [27] could be employed. These ML models would need to be calibrated using historical low-carbon production data, available critical mineral data, and estimated critical mineral data from MFA. Moreover, these ML models could work in tandem with the production network CGE model and estimate network variables crucial for the functioning of the CGE model.

The equilibrium dynamics of the endogenous low-carbon production network could be compared with traditional MFA. The critical minerals required in the future could be estimated by assuming total energy demand according to the IEA scenarios [1] and assuming the mix of low-carbon technologies and sub-technologies that will meet that demand [1]. The equilibrium dynamics of this production network could be studied at various future time points, with technical change modeled through transitions between alternative combinations of technology efficiencies, technology durabilities, and recycling rates. These findings will need to be juxtaposed with earlier results obtained by endogenously incorporating drivers of technical change to gain insights into the impact of endogenization.

The results from these studies could be used to provide recommendations that guide critical mineral infrastructure development and strategic policy. These recommendations will be particularly crucial in addressing vulnerabilities related to technical change scenarios within existing national critical minerals strategies. Furthermore, they could help governments navigate the complex technical change landscape. It’s important to note that such a research project does not investigate the impacts of the shifting geopolitical landscape [29], alternate government policies [30], and uncertainties in production data [31], which are also factors that are highly relevant to estimating future critical minerals demand.

Conclusion

The suggested framework would be the first to integrate endogenous technical change into a CGE model of the low-carbon production network using critical minerals as primary input factors. It would also be one of the first to estimate the learnable parameters of the endogenous production network using an ML model that can handle time and network dependency. The designed framework would allow researchers to study equilibrium effects on critical mineral demand and supply under multiple potential futures of technical change. Moreover, the comparison studies with traditional MFA and scenario-based exogenous technical change could provide a foundation for researchers studying technical change in the low-carbon critical minerals space.

The current critical mineral strategies do not account for multiple potential futures of technical change, including technological phase shifts and improvements in mineral intensities, energy efficiencies, recycling rates, life spans, and second-life proportions. The analyis resulting from developed models would help governments navigate a vast and uncertain technical change landscape and fix limitations in their existing critical minerals modeling frameworks and strategies. Furthermore, if the resulting modeling platform is made open-source, it would benefit academic researchers, industry professionals, government personnel, and policy experts. It would enable rapid experimentation, scenario analysis, and computational modeling, and allow modelers to probe questions not addressed here, thus furthering exploration and knowledge democratization.

References

1. IEA. (2021). The Role of Critical Minerals in Clean Energy Transitions. Available online: https://www.iea.org/reports/the-role-of-critical-minerals-in-clean-energy-transitions.
2. Rogelj, Joeri, et al. (2018). Mitigation pathways compatible with 1.5 C in the context of sustainable development. In Global warming of 1.5 C, pp. 93-174. Intergovernmental Panel on Climate Change.
3. Natural Resources Canada. (2022). The Canadian Critical Minerals Strategy. Available online: https://www.canada.ca/en/campaign/critical-minerals-in-canada/canadian-critical-minerals-strategy.
4. Natural Resources Canada. (2019). Canadian Minerals and Metals Plan. Available online: https://natural-resources.canada.ca/sites/nrcan/files/CMMP/CMMP_The_Plan-EN.pdf.
5. Hund, Kirsten, et al. (2020). Minerals for Climate Action: The Mineral Intensity of the Clean Energy Transition. Climate Smart Mining. The World Bank.
6. U.S. Department of Energy. (2021). U.S. Department of energy - critical materials strategy. Available online: https://www.energy.gov/eere/ammto/2021-doe-critical-materials-strategy.
7. Liang, Yanan, et al. (2022). Material requirements for low-carbon energy technologies: a quantitative review. In Renewable and Sustainable Energy Reviews, 161, 112334.
8. Rietveld, Elmer, et al. (2019). Global Energy Transition and Metal Demand.” In An Introduction and Circular Economy Perspectives. TNO.
9. Watari, Takuma, et al. (2019). Total material requirement for the global energy transition to 2050: A focus on transport and electricity. In Resources, Conservation and Recycling, 148, 91-103.
10. Månberger, André, and Björn Stenqvist. (2018). Global metal flows in the renewable energy transition: Exploring the effects of substitutes, technological mix and development. In Energy Policy, 119, 226-241.
11. Calderon, J. L., et al. (2024). Critical mineral demand estimates for low-carbon technologies: What do they tell us and how can they evolve?. In Renewable and Sustainable Energy Reviews, 189, 113938.
12. Acemoglu, Daron. (2002). Directed technical change. In The review of economic studies, 69.4, 781-809.
13. Acemoglu, Daron, et al. (2012). The environment and directed technical change. In American economic review, 102.1, 131-166.
14. Jiang, Hong-Dian, et al. (2023). The role of technical change in low-carbon transformation and crises in the electricity market: A CGE analysis with R&D investment. In Energy Economics, 125, 106897.
15. Song, Yi, et al. (2023). Technological change in critical metallic mineral sub-sectors and its impacts on mineral supply: Evidence from China. In Resources Policy, 85, 103850.
16. Haas, Christian, and Karol Kempa. (2018). Directed technical change and energy intensity dynamics: structural change vs. energy efficiency. In The Energy Journal, 39.4.
17. Casey, Gregory. (2022). Energy efficiency and directed technical change: implications for climate change mitigation.
18. Hassler, John, Per Krusell, and Conny Olovsson. (2021). Directed technical change as a response to natural resource scarcity. In Journal of Political Economy, 129.11, 3039-3072.
19. Acemoglu, Daron, et al. (2023). Green innovation and the transition toward a clean economy.
20. Wang, Peng, et al. (2022). Critical mineral constraints in global renewable scenarios under 1.5° C target. In Environmental Research Letters, 17.12, 125004.
21. Deetman, Sebastiaan, et al. (2018). Scenarios for demand growth of metals in electricity generation technologies, cars, and electronic appliances. In Environmental science & technology, 52.8, 4950-4959.
22. Acemoglu, Daron, and Pablo D. Azar. (2020). Endogenous production networks. In Econometrica, 88.1, 33-82.
23. Pauliuk, Stefan, and Niko Heeren. (2020). ODYM—An open software framework for studying dynamic material systems: Principles, implementation, and data structures. In Journal of Industrial Ecology, 24.3, 446-458.
24. CSIRO. (2021). Known unknowns: the devil in the details of energy metals demand. Available online: https://www.csiro.au/en/work-with-us/industries/mining-resources/resourceful-magazine/issue-22/known-unknowns-the-devil-in-the-details-of-energy-metals-demand
25. Calvin, Katherine, et al. (2019). GCAM v5. 1: representing the linkages between energy, water, land, climate, and economic systems. In Geoscientific Model Development, 12.2, 677-698.
26. Zhou, Jie, et al. (2020). Graph neural networks: A review of methods and applications. In AI open, 1, 57-81.
27. Jin, Ming, et al. (2023). A survey on graph neural networks for time series: Forecasting, classification, imputation, and anomaly detection. arXiv preprint arXiv:2307.03759.
28. Wang, Ke, Can Wang, and Jining Chen. (2009). Analysis of the economic impact of different Chinese climate policy options based on a CGE model incorporating endogenous technological change. In Energy policy, 37.8, 2930-2940.
29. Lee, Jordy, et al. (2020). Reviewing the material and metal security of low-carbon energy transitions. In Renewable and Sustainable Energy Reviews, 124, 109789.
30. Ge, Jianping, and Yalin Lei. (2018). Resource tax on rare earths in China: Policy evolution and market responses. Resources Policy, 59, 291-297.
31. McNulty, Brian A., and Simon M. Jowitt. (2021). Barriers to and uncertainties in understanding and quantifying global critical mineral and element supply. In IScience, 24.7.