Li demand projection
The concerns for Li availability are driven by expected demand growth associated with the significant increase in the LIB market. Hence, an understanding of application areas and their roles in major global demand trends is of greatest importance. In 2015, Li demand is about 34.6 kt. This spreads over EVs (14%), stationary ESSs (around 1%), traditional battery markets (25%) and non-battery applications (60%)16. A soaring demand for battery application over the last few years, with Li consumption share reaching more than 60%, was credited for the substantial increase in Li consumption to ~49 kt in 201931. This trend is expected to continue in the coming years.
The estimate for stationary battery capacity for power applications up to 2050 is based on the results from LUT University energy system transition research5,6. The results estimate that battery capacity increases along with the growing penetration of RE to ~47.8 TWhcap6 in 2050. The second driving factor is the global increase of TPED. According to the United Nation’s long-term target of global equity32, a world population of 11.2 billion people33 living at a European level of welfare will require 40 MWhth primary energy per capita34 by the end of the century. Consequently, battery storage demand is scaled to 200 TWhcap by 2100 (Supplementary Fig. 1) because the total electricity demand by the year 2100 may be at least four times the electricity demand of the year 205034. This gross estimate provides the basis of the potential upper limit of stationary LIB demand. Driven by cost advantages, mobile LIBs used in EVs are assumed to serve a second life (the remaining half life) as stationary batteries. Individual providers already claim its practicality35,36. Furthermore, other non-Li battery systems, e.g., vanadium redox flow or sodium–sulphur batteries, could share the stationary battery market.
In 2016, there are around one billion light duty vehicles (LDVs) on the road37. Following ICCT trends, this figure increases to 3.05 billion in 205037 (Supplementary Fig. 2). Thus, almost ten billion people34 would have 0.3 LDVs/capita. This factor still exceeds today’s global share of 0.1633,38, but does not reach the European level of 0.48 LDVs/capita38. EV penetration, a term referring to the combined share of EVs (BEV plus plug-in hybrid electric vehicle (PHEV)) as a percentage of total LDV stock, is widely discussed in society. Due to its promise to become the least cost solution for transport modes28, it is expected to enjoy an increasing market share though the projected rates of its increase varies depending on the source13,28,29,30,39,40,41,42,43,44. Despite major differences in the range of their projections, EV scenarios can be grouped into two. The first group assumes that EV sales shares continue to grow following a fast s-curve with proper policies, due to the dire need to decarbonise the transport sector28,29,30,40,41,44. For instance, these sources estimate global EV sales share to be 14%28 by 2025; 40%40, 48%41 and 50%30 by 2030; 100%29 by 2050. These scenario projections suggest a faster rate of sales share increase even if one may be faster/slower than the other by <5 years. In addition to the specific projection discussed above, some of these studies present additional scenarios. The low scenario of Hummel et al.28 expects 5.5% by 2025, suggesting a s-curve which lags by few years (~3 years). Such a delay has little impact on the result as presented below. The low scenarios of DNV GL leave several polluting cars on the roads by 205029. The second group of scenarios associates the achievable EV shares strictly to the deployment of charging infrastructure, battery markets and ongoing regulatory policy changes, etc. Thus, they perform an in-depth analysis of what needs to be achieved by various actors together with their EV projections. Typically, they have projections with lower market shares as compared to projections by the previous group. EV sales share of up to 33%42 (IEA, 30@2030 scenario) by 2030, 57%13 by 2040 and 66%43 by 2050 was estimated. IEA also presents another scenario (named New Policies Scenario42 (NPS)), which projects 15% global EV sales share by 2030. However, it should be noted that their NPS was created to analyse the sales growth based on existing policies and recent EV updates by ignoring commitments and potential improved policies. In this study, we adopted two EV sales share projections up to 2050 in order to closely evaluate the impact of the likely possibilities foreseen by both groups of forecasters. For the Best Policy Scenarios, we assumed 49% and 86% EV sales share by 2030 and 2050, respectively (Supplementary Table. 1), in agreement with Khalili et al.44. Though this target is lower than the trends assumed in the other studies, such as DNV GL29, it is considered to be a suitable target to achieve the required emission reduction in the transport sector to keep global temperature rise at ~1.5 °C by 2100 compared to the pre-industrial age. The corresponding numbers in terms of effective EV penetration, which is calculated as a BEV equivalent for the entire EV stock, are 18.4% and 79.8%, respectively (Supplementary Fig. 4 and Supplementary Table 1). Similarly, for the second scenario the share of newly sold vehicles is assumed to be 33% and 65% EVs by 2030 and 2050, respectively, in agreement with trends that the second group projects based on current EV initiatives and recent EV updates. Because of the encouraging trend in the industry and policy arena, we assume that EV growth may not be much lower than the scenarios of the second group. Thus, we excluded very pessimistic EV growth projections, for which sufficient data are also available16,17,18,19,20,21,22,23,24,25,26,27. Moreover, any lessons to be obtained by including additional low scenarios can be understood from the results presented in this paper. Because the various studies referred to above also assume different LDV stocks13,28,29,30,39,40,41,42,43,44 by 2050, this study also includes 2 billion final LDV stock by 2050 for a low demand scenario, while applying both EV shares increases assumptions. Note that LDV stock was assumed to remain constant for all years after 2050 for both the cases, though intuition suggests an increase. The remaining share of the 100% renewable LDV transport is assumed to be provided by alternative concepts like power-to-liquid, biofuels, power-to-gas, EV with new battery chemistry or fuel cell electric vehicles based on hydrogen. However, it could be noted that extreme scenarios may still rely on some fossil ICEs by 2050, particularly for the trend that follows the second EV projection. The EV market is assumed to be constituted by BEV and PHEV. The assumed average battery capacity per vehicle is set to 60 kWh for BEV and 15 kWh for PHEV, which will have 8 years of lifetime serving in EVs16,28,45. At the end of their life, these batteries will be used for an additional 8 years as stationary batteries before entering a recycling loop. The assumed 16-year service time is lower than the 20-year lifetime that the industry provides for stationary batteries in agreement with Turcheniuk et al.46.
Though recycling LIB is still under development, reports show that ~97,000 tonnes of LIBs were recycled globally in 201847. The slow development is due to economic reasons48 and lack of regulations, as well as challenging technical processes and collection procedures. The Li recycling efficiency was set to 95% based on recent technological development47,48,49,50,51,52,53, while the collection rate was set to grow from ~45% at present based on global data47 to 99% by 2050 (see ‘Methods’ and Supplementary Fig. 5).
Whenever efficiency improvements of LIBs fit to the requirement of city and intercity buses, electric bikes and scooters as well as medium duty vehicles (delivery trucks) and heavy trucks will use batteries. A growth curve is applied to provide the projection for these types of applications, which are assumed to require 50 TWh (average) battery capacity in 2100 (Supplementary Fig. 6).
To convert the battery capacity to the equivalent Li requirement, a long-term estimate of Li intensity per storage capacity of ~130 g/kWhcap16 is applied uniformly up to 2100, which is at the bottom range of literature data (Supplementary Table 2). However, future research should employ insights from presently missing Li intensity learning curve when such data are available.
In addition to these quantitatively dominating applications, Li is also used in varying industrial applications, as batteries and for non-battery use. Based on current annual demand16, the corresponding Li demand is calculated with a CAGR of 3% and 2%, respectively, following a recent global economic growth trend54 (Supplementary Figs. 7 and 8).
This study creates eight demand variations (Supplementary Fig. 9) by combining relevant factors, which are used to create 18 scenarios together with 4 supply scenarios to be discussed in the following section. While addressing the low demand cases, our scenario definitions lean towards investigating the possible challenges of the aspired transition to sustainable energy systems, for which at present EVs are the best candidate to meet the climate change mitigation targets in the transport sector.
Li supply from resources to production output
In 2016, global Li supply is 38 kt31 (Supplementary Table 3). Because of its very-high chemical reactivity, Li has no elemental occurrence in nature, but can be mainly found in ionic compounds like oxides or chlorides26. These are enriched either in ores as minerals (Supplementary Fig. 10) or in salt solutions as brines. Both major types of deposits differ in geological formation, extraction and process technology, associated costs and time, sustainability as well as size and dispersion. Furthermore, Li is dissolved in oceans as an almost ‘unlimited’ resource. Due to poor maturity of the extraction techniques and expensive production costs, seawater extraction is not expected in the near future20,26.
To focus on strategic and long-term aspects, this study is limited to examining resources that are geologically confirmed without considering the restrictions concerning socio-economical exploitation or current state of technology. The latest data from the United States Geological Survey (USGS) indicate total resources of 80 Mt Li31. However, an in-depth literature review reveals the subjective, non-transparent and imprecisely defined character of resource estimation. Figures ranging from 3020 to 95 Mt Li26 differ by more than a factor of three (Supplementary Table 4). Due to these divergences, this study uses four scenarios covering one low (26 Mt Li), one medium (41 Mt), one high (56 Mt) and one very-high (73 Mt) resource value (see ‘Methods’ and Supplementary Table 5). The lowest number covers the range of proven reserves26,55 and describes a worst-case situation, where no additional resources are exploited. Both next higher assessments, in turn, assume the potential extractable mineral deposits. Notice that, as shown in Fig. 1, all deposit costs are lower than the price of industrial grade Li2CO3, suggesting their economic viability depends on time. The 41 Mt reserve estimate is based on the higher range of the proven mineral reserves, which is below the red line, and as shown by the yellow line. The value of 56 Mt corresponds to the more optimistic reserve quantity, which assumes that all reserves could provide their estimated high resource potential. The very-high reserve covers the range of some very high, but due to missing rationale, rather unrealistic estimates25,26,27.
Geographically, Li deposits are distributed rather unequally on a global scale (Supplementary Fig. 11). Effects on social and political interests as well as economic trading are important17. Because to bring these resources out of the ground, exploitation must pay off; the marketable price must exceed extraction costs. For a long-term assessment of the latter, the concept of the cumulative availability curve56 is used (see ‘Methods’ and Fig. 1), which specifies the amount of resources being available at certain costs. In theory, society must extract the next more expensive deposits as the demand for the resource increases. Because of a less energy-intensive extraction process, brine deposits are generally cheaper. Their extent, particularly in large South American ‘Salars’, determines how long these low-cost resources are available. But respective to the current lithium carbonate (Li2CO3) prices57 and the continuing demand, increased extraction costs may not restrict the availability of Li.
However, time has its own constraints. In the build-up phase, so-called greenfield projects must go through resource discovery, several stages of feasibility studies, facility construction and production start-up. This usually takes one to two decades16,20. After, the process time along the value chain determines the flow rate of fresh material into society. The lead time of Li2CO3 appears to be uncritical for mineral deposits (e.g., 5 days for spodumene treatment17) but becomes a limiting factor for brines. Relying on solar irradiation, the evaporation process is not constant throughout the year and takes 1–2 years17. Even perfect conditions as found at Salar de Atacama delay the production for at least 12 months20. The Li supply system implies a certain moment of inertia.
To quantitatively assess Li supply, the inflow of virgin material expressed by the production volume per year is modelled by applying logistic-growth-based bell-shaped curves following Vikströom et al.26 (see ‘Methods’ and Supplementary Fig. 11). The bell-shaped curve is fit to the historical production data to choose the curve that produces short-term projections in agreement with recent developments. However, because fitting to the present production trends leads to a sharp rise in supply as compared to the demand projections, we enforce a criterion that production around mid-century is not larger than 10% of the annual base case demand. Yet for the high and very-high production scenario, some years see an over production as high as 30% for the same demand due to the demand curve that has a shape of a roller coaster. (Supplementary Fig. 9). As opposed to the production estimated by Vikströom et al.26, this study is much more comprehensive because of included resource scenarios.
Critical dynamics of Li
As a first step, we present the comparison of supply and demand (see ‘Methods’ and Fig. 2) on a yearly basis. Figure 3 reveals highly critical dynamics using various scenarios. Figure 3a reveals that Li production shows a good balance with annual fresh Li demand in the near-term for the medium production for almost all policy scenarios. However, Li supply and demand balance start to show strong demand scenario dependence around 2030. Focusing on scenarios related to EV shares, it can be seen that for the Best Policy Scenarios (BPS 3b LDV) the observed good balance of Li demand and supply extends to ~2050, when it reaches a time the market started to experience a large deficit that lasts for the remaining half of the century. The inflow of virgin material and the increase in recycling is not sufficient to supply the important transition years for most part of the second half of the century. The supply and demand balance was found to show large surplus from 2030 up to 2050 for other EV-related Li demand as can be seen from the curves corresponding to the three scenarios, namely BPS 2b LDV LD, CPS 3b LDV and CPS 2b LDV. In particular, the CPS 2b LDV scenario resulted in larger surplus, which extended up to 2053 when the deficit dominating the remaining part of the century begins. The appearance of such an early deficit following two decades of large surplus is obviously not a good representation of the real market operation for these scenarios. In reality, while a short-lived large surplus may be possible, the long-term market should show stable demand and supply balance, except for the case of not manageable deficits caused by resource scarcity. The reported special surplus occurred because the production was modelled solely focusing on BPS 3b LDV demand in order to simplify inter-scenario comparison rather than producing production scenarios that fit to each demand curve. This is what led to an extended large surplus for lower demand scenarios as observed in these cases. To address this gap, we assumed that the surplus will be accumulated for later use (Supplementary Figs. 13–15). In this circumstance, the year the deficit occurs moves to a later year, though the amount of the shift depends on the demand scenario. For these three scenarios, the shift is larger. Specifically, the deficit moves from 2054 to 2077 for the CPS 2b LDV demand scenario. Three of the remaining four curves, which correspond to battery lifetime in second use (represented by the remaining two curves presenting variation in lifetime of secondary batteries), and vehicle-to-grid (V2G) integration, closely follow the trend corresponding to the BPS 3b LDV demand with some minor differences depending on the demand scenario. For instance, the integration of V2G technology40,58 relieves the power sector. Hence, less second-life material is required and more Li can go directly to recycling because of BEV integration into the grid. This retention of Li in the loop pulls the corresponding graph a little up. The last remaining scenario, namely the BPS 3b LDV LR demand, was an exception in generating deficits well before 2040 that got worse for the remaining part of the century as shown by the drastic deficit. This is strong evidence of the dire need for the establishment of an efficient Li recycling system. Under the high production conditions given in Fig. 3b, all scenarios maintain similar trends to the corresponding scenarios discussed in Fig. 3a, but the additional supply pulls up the curves with a little delay of the start of the deficit in the second half of the century. Accumulating the surpluses also shifted the time when this deficit starts for all scenarios as shown in Supplementary Fig. 14. Specially, the CPS 2b LDV experiences no deficit throughout the century for this supply. Such a significant change also shows that the deficits are mainly due to transport sector Li demand.
The foregoing result shows that the balance between supply and demand depends on the presence of well-established recycling systems, Li production rates, achievement of V2G integration and the ability to limit LDV stock growth without compromising the transportation services that society needs to function efficiently. This finding has important lessons regarding future policy directions that should be pursued in order to achieve a sustainable transport sector that can conform to the sector’s emission target of keeping temperature rise at about 1.5 °C above pre-industrial levels by 2100. The above result clearly shows that scenarios that can conform to the stated climate target and improved transport equity will definitely result in serious Li supply deficits over the next century. On the contrary, low demand scenarios, such as the CPS 2b LDV or lower, achieve a balanced Li supply and demand throughout the century. However, studies show that such scenarios definitely compromise the climate change target59. To solve this conflict, global concerted effort is required to acquire commitment to enforce the mix of the following policies across the globe: (i) develop transport services that could reduce the dependence on LIB (reduce number of LDV) by promoting improved public transportation, shared rides and other possible solutions; (ii) establish and maintain efficient recycling system; (iii) improve LIB technology to reduce material demand per battery capacity; (iv) develop new battery chemistries or other sustainable transportation options that will reduce the demand for LIB.
In addition, we present the impact of deviations of production as given in Fig. 3c. The effect of different supply estimates is not visible until 2028 because all production projections are fitted to historical production. After 2030, however, the curves start to diverge. At low resources, a continuous deficit runs through the entire century. The aforementioned effects of TPED increase reflect a clear dip. Very-high resources, in turn, enable partly significant surpluses around 2050. For this scenario, the observed deficit can be covered by accumulating these surpluses to achieve the supply–demand balance throughout the century.
Material flow of Li during this century
In a second step, the availability of Li is examined at a century level in order to clarify the capability of the estimated resource potentials to cover the demand dynamics over longer time periods. Figure 4 reveals one rather clear message: the penetration of storage systems based on LIBs results in a prospective availability constraint of Li during this century. Nearly all considered scenarios run into a—varyingly strong—deficit. However, at high supply condition, demand of the CPS 2b LDV and BPS 2b LDV scenarios could be matched by the available resource throughout the century.
The very-high resource condition was the only case that can match the BPS 3b demand scenario (one of the highest demand) and thus can meet all other demand scenarios throughout the century. The depletion year given on top of each bar in Fig. 4 also shows that the fewer the resources, the earlier they are depleted and the higher the resulting deficit.
Now, let us examine the material flow using the BPS 3b LDV scenario to understand the process. The deficits in the BPS 3b LDV scenarios (except in the very-high supply condition) are a consequence of the cumulative base case demand of 68.03 Mt of fresh Li until the year 2100. The way of this material flow within the system can be traced in Fig. 5, which presents a hypothetical scenario by assuming that the required fresh Li is available (a condition that is possible for the very-high resource scenario only). This chart is of greatest importance as major correlations of application areas and energy sectors become clear. The bulk of inflowing fresh material is used for BEVs and—a little less—other transport applications. This huge stock continues in a flow towards second-life material and feeds almost the entire stationary sector. Only a tiny amount of net demand remains to be supplied by fresh material. All spent batteries go to recycling—irrespective of whether directly after first use or after second use. The rest of the material, which is currently in use, forms the fictional volume of Li in stock. Applying the base case demand, this amounts to 51.29 Mt in the reporting year 2100. The difference of 16.74 Mt leaves the system and is lost. This drain consists of losses due to collection rate and recycling efficiency less than unity as well as all industrially used material, which is not recovered.
In consideration of the significance of annual dynamics and the size of the cumulative recycling loop, the return flow of secondary material back to society is one fundamental part of the model. A well-established and highly efficient recovery system is essential to maintain Li in circulation to augment the supply shortage. The consequences of a less efficient recycling system can be seen from the corresponding deviation that shows vast deficits at the end of this century. At medium resources, Li deposits are already depleted in 2055.