Are future recycling benefits misleading? Prospective life cycle assessment of lithium‐ion batteries

Life cycle assessment (LCA) quantifies the whole‐life environmental impacts of products and is essential for helping policymakers and manufacturers transition toward sustainable practices. However, typical LCA estimates future recycling benefits as if it happens today. For long‐lived products such as lithium‐ion batteries, this may be misleading since there is a considerable time gap between production and recycling. To explore this temporal mismatch problem, we apply future electricity scenarios from an integrated assessment model—IMAGE—using “premise” in Brightway2 to conduct a prospective LCA (pLCA) on the global warming potential of six battery chemistries and four recycling routes. We find that by 2050, electricity decarbonization under an RCP2.6 scenario mitigates production impacts by 57%, so to reach zero‐carbon batteries it is important to decarbonize upstream heat, fuels, and direct emissions. For the best battery recycling case, data for 2020 gives a net recycling benefit of −22 kg CO2e kWh−1 which reduces the net impact of production and recycling from 71 to 49 kg CO2e kWh−1. However, for recycling in 2040 with decarbonized electricity, net recycling benefits would be nearly 75% lower (−6 kg CO2e kWh−1), giving a net impact of 65 kg CO2e kWh−1. This is because materials recycled in the future substitute lower‐impact processes due to expected electricity decarbonization. Hence, more focus should be placed on mitigating production impacts today instead of relying on future recycling. These findings demonstrate the importance of pLCA in tackling problems such as temporal mismatch that are difficult to capture in typical LCA.


Temporal mismatch in life cycle assessment and misleading recycling credits?
multifunctionality problem-how should impact contributions be assigned to these outputs?One method is substitution, which assumes that the process outputs leaving the system displace another process (Heijungs et al., 2021).Therefore, impacts associated with the displaced process are avoided and credits can be assigned to show potential impact reductions outside the system.This can be particularly beneficial for highly recyclable products with high production impacts as credits from recycled materials can lead to lower net impacts compared to production alone.For example, the recycling of lithium-ion batteries can reduce their net global warming potential by over a third compared to their production alone (Kallitsis et al., 2022;Mohr et al., 2020).However, for batteries used in electric vehicles, their end-of-life recycling occurs many years after production.Typical LCA of long-lived products does not account for this and instead models end-of-life based on current data as if recycling occurs in present-day systems.
This creates uncertainty in the end-of-life impacts and claimed recycling credits that could be misleading.
To illustrate this problem, consider this hypothetical scenario: (a) Battery production in the year 2020 emits 10 kg CO 2 e kg −1 ; (b) Battery recycling in that year emits 1 kg CO 2 e kg −1 and recovers 50% of cobalt sulfate (CoSO 4 ) by battery mass; (c) Primary production of CoSO 4 emits 8 kg CO 2 e kg −1 .In 2020, the total life cycle impact of a 1 kg battery is 7 kg CO 2 e consisting of 10 kg CO 2 e during production, 1 kg CO 2 e during recycling and −4 kg CO 2 e credits for CoSO 4 recovery.It could be said that recycling reduces net impact by 30% from 10 kg CO 2 e (during production) to 7 kg CO 2 e (post-recycling).However, the battery will last much longer than a year.Assuming battery lifetime to be 10 years (i.e., end-of-life in 2030) and a decarbonization extent of 50% by 2030, battery recycling and CoSO 4 primary production will now emit 0.5 and 4 kg CO 2 e kg −1 in 2030, respectively.Here, the total life cycle impact of a 1 kg battery is 8.5 kg CO 2 e consisting of 10 kg CO 2 e during production in 2020, 0.5 kg CO 2 e during recycling in 2030, and −2 kg CO 2 e credits for CoSO 4 recovery in 2030.In contrast to the expected 30% reduction in net impact, this is now a 15% reduction from 10 to 8.5 kg CO 2 e.In this case, the recycling benefits have been overestimated which could be problematic.First, it may disincentivize impact mitigation from upfront production if it is interpreted that future recycling "offsets" these impacts.Second, perceived future recycling benefits may be a decisive factor between products.However, when the potential overestimation of future recycling benefits is considered, this could lead to alternative product decisions.These challenges are framed as the temporal mismatch problem which has not been investigated for recycling credits.This is likely due to the reliance on historic data and difficulty in forecasting future impacts (Bisinella et al., 2021).
However, emerging prospective LCA (pLCA) methods that use integrated assessment models (IAM) for future scenario modeling (Beltran et al., 2020) present an opportunity to investigate the effects of temporal mismatch on recycling credits.

Prospective methods toward an integrated approach
In recent years, a new branch of pLCA has emerged that combines external models to forecast future life cycle inventories (LCI) (Bisinella et al., 2021).In particular, scenarios from IAMs have been applied to LCI databases (Beltran et al., 2020;Sacchi et al., 2022).IAMs are cost-optimization models that forecast future scenarios, such as changes in regional electricity mixes to meet the 2 • C global warming target by 2050, based on socioeconomic and technology development narratives called "Shared Socioeconomic Pathways (SSPs)" (Pauliuk et al., 2017;Riahi et al., 2017).
For example, to address temporal mismatch problems between foreground and background LCI in future scenario modeling, Beltran et al. (2020) incorporated IMAGE (van Vuuren et al., 2017) electricity scenarios under several SSPs into ecoinvent v3.3 using Wurst and Brightway2 (Mutel, 2017).They showed that considering electricity scenarios in the background LCI had a significant influence on LCA outcomes for electric vehicles finding that under ambitious mitigation scenarios up to 2050, climate change impacts could be altered by 80% compared to when impacts are calculated without background changes.Subsequently, Sacchi et al. (2022) introduced "premise," a streamlined methodology to produce databases for pLCA using several IAM scenarios that consider multi-sector transformations such as electricity, fuels, and cement.The use of IAMs in pLCA has been growing, with applications in the automotive sector (Cox et al., 2018(Cox et al., , 2020)), energy systems (Astudillo et al., 2019;Vandepaer et al., 2020), metal supply systems (Harpprecht et al., 2021;van der Meide et al., 2022), and water supply (Baustert et al., 2022).
IAMs provide an explorative method for future production and end-of-life activities that could evaluate the temporal mismatch problem illustrated in Section 1.1.In particular, lithium-ion batteries are an interesting case study because of their long lifespan in electric vehicles and the popular discussion of recycling pathways (Chen et al., 2019).LCA of batteries for electric vehicles has been widely studied where they can nearly double the global warming potential of electric vehicle production compared to typical combustion vehicle production (Das, 2022;Shafique et al., 2022;Yang et al., 2021).Half of the impact of battery production arises from direct energy use (heat and electricity) with the rest occurring from upstream production for cathode materials such as CoSO 4 and nickel sulfate (NiSO 4 ) (Kelly et al., 2020;McManus, 2012;Winjobi et al., 2022).
Perhaps surprisingly, most studies exclude end-of-life stages for batteries, and when considered, can lack clarity in methodological choices; have limited and undisclosed data; and considerably vary in the influence of end-of-life on outcomes (Ellingsen et al., 2017;Nordelöf et al., 2019;Pellow et al., 2020).Some studies have considered foreground scenario analysis (Chordia et al., 2021;Degen & Schütte, 2022;Kelly et al., 2020) but only one IAM-applied pLCA has been conducted on lithium-ion batteries that considered future background changes although end-of-life and temporal mismatch were not accounted for (Xu et al., 2022).Thus, the future production and end-of-life environmental impacts of batteries remain to be explored.In response, this study introduces the potential effects of overlooking temporal mismatch in LCA using prospective methods to investigate: Foreground diagram within the study system boundaries for the NMC111 battery with pyrometallurgical recycling and key exchanges that are grouped in the contribution analyses.Blue-active materials specific to the battery chemistry.Red-total heat required across all production processes.Orange-total electricity required across all production processes.Purple-other materials used that are not specific to the battery chemistry.Green-recycling inputs.Patterned green-recovered materials solved by substitution.
1. What is the global warming potential of lithium-ion battery production now and into the future?
2. What are the recycling benefits of lithium-ion batteries for global warming potential?
3. How does considering temporal mismatch affect future recycling benefits for global warming potential?

Prospective life cycle assessment
The LCA framework followed ISO 14040/14044 (British Standards Institute, 2021).The study goal was to conduct a pLCA to investigate the temporal mismatch effects on lithium-ion battery production and recycling.For simplification, a single impact category (global warming potential) and prospective transformation (2020-2050 electricity decarbonization from IMAGE) were considered.The system boundaries considered a cradleto-grave scope where the battery foreground system encompassed the unit processes of precursor production, active material production, cell production, and recycling (Figure 1).Potential stages such as electric vehicle pack-level production, use, and remanufacturing were not considered given the study goal.The key battery function is to store electrical energy; therefore, the functional unit (FU) selected was the production and recycling of a battery that can hold 1 kWh of energy capacity.China was selected as the geographical scope because it accounts for 79% of global lithium-ion battery production (Placek, 2022).Given the study aim and the infancy of lithium-ion battery recycling markets, recycling was also assumed to take place in China.Background unit processes such as raw materials and energy production was considered using ecoinvent v3.8 (Wernet et al., 2016).The pLCA adopted an attributional framework where multifunctionality from recycled products was solved by substitution assuming an open-loop process with closed-loop recycling procedures (British Standards Institute, 2012).
Most electric vehicle batteries are warranted for 8 years (or 100,000 miles) but are expected to last much longer (Department for Transport, 2022), especially with recent advances in lifespan (Aiken et al., 2022).Hence, it was assumed that most batteries reach their end-of-life 10−30 years after production.The selected life cycle impact assessment (LCIA) method was IPCC 2013, GWP100a with adjusted characterization factors by premise (Section 2.3).The LCI and LCIA calculations were conducted using Brightway2 in Jupyter Notebooks (Python), with outcomes exported to Excel for more detailed evaluation (available in Supporting Information S1).During interpretation, the present and future impact of battery production was explored with a contribution analysis.In the contribution analysis, total foreground electricity and total foreground heat used across all production activities were grouped, respectively (Figure 1).Key chemistry-specific material inputs across all production stages were grouped into active materials.Non-chemistry-specific inputs (mostly from cell production) were grouped into other materials.The recycling process-impacts from materials, energy, and emissions-and recycling credit-avoided burdens from recovered materials displacing primary production-were also distinguished.The effects of temporal mismatch were then investigated by setting battery production impacts to occur in the baseline year 2020 while end-of-life recycling impacts were varied to occur in 2020, 2030, 2040, and 2050 (Section 3.2).All numerical results are available in Supporting Information S3 and Supporting Information S4.

Foreground LCI for lithium-ion batteries
Six lithium-ion batteries were selected based on industry-standard chemistries consisting of lithium manganese oxide (LMO), lithium iron phosphate (LFP), nickel manganese cobalt (varying stoichiometric ratios of NMC111, NMC622, and NMC811), and nickel cobalt aluminum (NCA).
To provide a consistent dataset (available in Supporting Information S2), the foreground LCI was compiled predominantly from publications and software from Argonne National Laboratories (Dai et al., 2019;Winjobi et al., 2020).Further additions from the literature were made where necessary, such as infrastructure and transportation, but these typically have minimal impact contributions (Crenna et al., 2021;Peters & Weil, 2018).
Precursor production was required for NMC and NCA chemistries while not needed for LFP and LMO (Winjobi et al., 2020).Subsequently, all batteries required active material production to form the lithium active cathode materials before final cell production (Dai et al., 2018;Majeau-Bettez et al., 2011;Notter et al., 2010).Materials required to produce the variety of active cathode materials included CoSO 4 , NiSO 4 , lithium carbonate (Li 2 CO 3 ) or lithium hydroxide (LiOH), iron sulfate (FeSO 4 ), and manganese sulfate (MnSO 4 ) which were linked to their respective ecoinvent v3.8 activities.Each battery cell was characterized by its specific energy-electrical energy stored per unit mass-which is key to the FU.This ranged from 0.174 to 0.251 kWh kg −1 for LFP and NCA, respectively (Winjobi et al., 2020).Next, the cell percentage mass compositions were characterized (Dai et al., 2019).This includes the electrochemical substances of the active cathode material at the positive aluminum current collector, as well as the graphite at the negative copper current collector.The electrochemical materials are mixed with carbon black and polyvinylidene fluoride (PVDF) that, respectively, aid electrical conductivity and binding of the mixtures.Since PVDF was not available in ecoinvent v3.8, a recent inventory was used (Hu et al., 2022).The electrolyte serving as the electrochemical medium is a mixture of lithium hexafluorophosphate (LiPF 6 ), ethylene carbonate (EC), and dimethyl carbonate (DMC).Polypropylene (PP), polyethylene (PE), and polyethylene terephthalate (PET) were used in the separator and packaging.The energy consumption for cell production ranged from 31.7 to 42.3 MJ kg −1 for LFP and NMC811, respectively, and in all cases was assumed to be provided by 82.4% natural gas and 17.6% electricity, encompassing various assembly line and drying activities (Dai et al., 2017;Winjobi et al., 2020).For the end-of-life stage, four recycling routes were modeled, encompassing various inputs and material recovery rates (Dai & Winjobi, 2019;Dai et al., 2019): • Pyrometallurgical recycling: Batteries are shredded and sent to a smelter to burn off the electrolyte and plastics.Carbon and aluminum are oxidized which subsequently reduces key metals such as cobalt and nickel in a matte, while a slag makes up the rest of the materials.The key metals from the matte are then recovered using acid leaching, solvent extraction, and precipitation.• Inorganic and organic hydrometallurgical recycling: Consisting of an inorganic and organic route which both have identical recovery rates but different inputs.Batteries are shredded and undergo a low-temperature calcination process that burns off some organic compounds.Then, several physical processes separate plastic and aluminum, copper, and steel as metal scraps.Acid leaching and solvent extraction processing recover cobalt, nickel, manganese, and lithium compounds.
• Direct recycling: Batteries are perforated and undergo supercritical CO 2 extraction to recycle electrolyte solvents and salts.The rest of the battery goes through several physical separation processes to recover original materials such as the active cathode material that can be relithiated to restore the electrochemical properties.
These recycling routes have different levels of technology maturity (Dai & Winjobi, 2019;Dai et al., 2019).While pyrometallurgical and inorganic hydrometallurgical routes are commercialized processes, the organic hydrometallurgical route is an active research area and direct recycling is in its infancy at laboratory scale.It is assumed that compounds recovered by solvent extraction and precipitation steps (Dai & Winjobi, 2019;Dai et al., 2019), such as CoSO 4 , NiSO 4 , and MnSO 4 , have high purity (Chen et al., 2019).Therefore, for substitution, these are assumed to displace their original primary production activities.An exception is made for graphite where the recovered material is assumed to be below the required grade for batteries, hence, displacing a lower-grade primary production process.Copper, aluminum, and iron recovered in the slag are assumed to be low grade, hence, lower-grade primary production activities are selected to be displaced.

Background LCI by application of IMAGE electricity scenarios to ecoinvent
This section summarizes the key steps in background LCI transformations (available in Supporting Information S1).The present study applied SSP2-"Middle of the Road"-electricity scenarios from IMAGE using premise (Sacchi et al., 2022) (Figure 2).SSP2 depicts medium challenges to climate change mitigation and adaption, representing slow and uneven socioeconomic progress toward sustainable development (Riahi et al., 2017).
Two SSP2 scenarios were used, which each represent the development of regionalized electricity production and markets from 2020 to 2050 in 10-year steps.The first scenario was Base, depicting a "slow" decarbonization approach to the electricity markets.The second was RCP2.6, representing a "fast" decarbonization approach for electricity markets to limit global warming to 2 • C. Figure 3 provides an example of how electricity production mixes in China change under the SSP2 scenarios.In Jupyter Notebooks (Python), the ecoinvent v3.8 database was imported to serve as the background LCI.The premise package was then applied to transform the electricity markets and production activities in ecoinvent representing Base and RCP2.6 scenarios.Subsequently, a new version of ecoinvent was created every 10 years from 2020 to 2050 for both Base and RCP2.6 The approach overview that encompasses the integration of IMAGE scenarios to ecoinvent background life cycle inventories (LCI) using premise.Multiple background systems are created that are applied to the battery foreground LCI.Brightway2 is then used to conduct the prospective life cycle assessment and assess the effects of temporal mismatch.
F I G U R E 3 China electricity production mixes for Base and RCP2.6 scenarios based on SSP2 from IMAGE.In Base mixes, while the distribution of fossil-fuel technology increases compared to low-carbon technologies, the increase of natural gas and decrease in coal generation still leads to a lower grid carbon intensity.This is due to natural gas having a significantly lower global warming potential compared to coal generation.In RCP2.6 mixes, the grid carbon intensity reduces substantially due to the uptake of low-carbon technology such as biomass, solar, wind, and nuclear and the rapid reduction of coal generation.Additionally, carbon capture and storage (CCS) technology is also introduced, particularly to the remaining significance of natural gas.The underlying data for this figure can be found in Supporting Information S3.
scenarios.Next, the foreground LCI for battery production and recycling was formatted, imported, and matched to each version of ecoinvent.Thus, multiple foreground LCIs for each year and scenario were created.To account for net negative emission technology flows (e.g., input biogenic CO 2 ) in the LCIA, a premise function was used to adjust for these characterization factors for IPCC 2013, GWP100a.Finally, Brightway2 was used to conduct the pLCA computations, evaluate the future impacts of batteries, and investigate the effects of temporal mismatch.

F G U R 4
Total global warming potential of battery production for Base and RCP2.6 scenarios from 2020 to 2050.Across all batteries, the Base scenario sees minor reductions in impact due to small decarbonization in the China electricity grid carbon intensity.Expectedly, the RCP2.6 scenario sees major impact reductions due to high decarbonization in the electricity grid carbon intensity.The underlying data for this figure can be found in Supporting Information S3.

Present and future global warming potential of lithium-ion battery production
Figure 4 shows the global warming potential of battery production for each 10-year time step, from 2020 to 2050, for both Base and RCP2.6 scenarios.To summarize key impact contributions in present-day batteries, the Base 2020 scenario is discussed.Across all batteries, LMO shows the lowest impact while NMC111 is the highest.One impact determiner is the battery specific energy where greater values lead to less material and energy required to fulfil the functional unit.For example, NMC811 has a significantly greater specific energy (0.248 kWh kg −1 ) than LFP (0.174 kWh kg −1 ) which leads to a lower impact.However, despite LMO having much lower specific energy (0.184 kWh kg −1 ) over NMC811, LMO is still lower in impact.These variations can be investigated further in the contribution analysis (Figure 5).The contribution analysis suggests that foreground electricity and active materials are key impact determiners.For example, foreground electricity contribution shows significant variation, ranging from 21% to 39% in NMC111 and LFP, respectively.This results from varying foreground electricity requirements across all production stages.For example, the total foreground electricity requirement for LFP is substantially greater (31 kWh kg −1 ) compared to NMC111 (21 kWh kg −1 ).The contribution of active materials also shows considerable variation ranging from 22% to 49% in LMO and NMC111, respectively.A large factor is the NiSO 4 and CoSO 4 required in precursor production for the nickel-based chemistries which results in a high embedded impact from their background processes.Since CoSO 4 has a higher impact over NiSO 4 , the reduced use of it from NMC111 to NMC622, NCA, and NMC811 results in these batteries having a 20%, 22%, and 28% lower impact, respectively.Therefore, the dominant differences in total production impact depend on the battery specific energy, foreground electricity requirements, and inputs for active materials.
Looking into the future global warming potential of battery production from 2020 to 2050, the Base scenario shows small reductions of up to 12% due to a modest reduction in electricity grid carbon intensity (Figure 4).In comparison, the RCP2.6 scenario for battery production shows greater impact reductions of 38%−57% resulting from the high decarbonization of grid electricity.However, the magnitude of impact reductions varies across batteries which can shift the rank order.For example, by 2050 LFP shows a 54% reduction compared to only 43% in NMC811, which results in LFP having a lower impact than NMC811.The varying impact reductions are determined by two factors.First, high relative contributions of foreground electricity to total impact lead to greater reductions due to only electricity decarbonization being modeled such as in LFP and LMO (Figure 5).Second, the higher background contribution of electricity in active materials also leads to greater impact reductions.For example, by 2050, active materials in LMO show a 62% impact reduction versus 23% in NCA. Figure 6 shows the varying degrees of decarbonization for battery materials displaying that the impact of LMO materials such as manganese (III) oxide and oxygen experience 49% and 111% reductions, respectively.In contrast, the CoSO 4 and NiSO 4 materials used in NCA show lesser 17% and 15% reductions, respectively.This is due to the greater impact contributions of heat, fuels, and direct emissions that cannot be directly decarbonized by electricity.Across all batteries, other materials show considerable 39%−43% impact reductions (Figure 5).This is due to materials such as PVDF, copper, and Li 2 CO 3 reducing in impact by 87%, 68%, and 43%.In contrast, the aluminum current collector used in all batteries displays a minor 9% reduction due to the heat use and direct CO 2 emissions from background processes that cannot be decarbonized by electricity.In summary, the future RCP2.6 scenarios highlight that in addition F I G U R E 5 Global warming potential contribution analysis of battery production for RCP2.6 scenario from 2020 to 2050.Across all batteries, foreground electricity shows substantial impact reductions which are expected.The impact reductions in active materials vary depending on the contribution of electricity in upstream background processes.Other materials also show relatively significant impact reductions primarily from decarbonized background processes.The underlying data for this figure can be found in Supporting Information S3.
F I G U R E Percentage of global warming potential reduction of battery materials from ecoinvent v3.8 market activities for RCP2.6 scenario from 2020 to 2050.The difference in impact reductions is a result of the varying electricity contributions in upstream background processes.The underlying data for this figure can be found in Supporting Information S3.
to present-day impact, the future impact reduction potential should be considered based on electricity contributions in both the foreground and background that could lead to alternative product choices such as in the case of LFP and NMC811.Moreover, it is highlighted that electricity decarbonization is not a whole-system solution to directly decarbonize many materials, hence, further impact reductions require deep decarbonization strategies for heat, fuels, direct emissions, and other processes.

3.2
The effect of temporal mismatch on future recycling benefits and direct recycling achieve net impact reductions of 22%−28%, 25%−37%, 30%−41%, and 43%−49%, respectively.Primarily, these reductions result from the recovery of high-impact CoSO 4 and NiSO 4 .The increasing net benefits between the recycling routes are primarily due to more materials being recovered which leads to greater recycling credits.For example, while both pyrometallurgical and hydrometallurgical recycling have the same recovery rates for NiSO 4 and CoSO 4 , hydrometallurgical recycling can recover materials such as aluminum, Li 2 CO 3 , and electrolyte, yielding greater recycling credits.Direct recycling is most beneficial due to the recovery of the active cathode material which avoids active material production leading to the greatest recycling credits.LFP and LMO also benefit from direct recycling, showing net impact reductions of 32% and 20%, respectively.In contrast, pyrometallurgical recycling of LFP and LMO shows a 21% and 27% increase in net impact, respectively, with inorganic hydrometallurgical recycling showing small increases too.This is the result of only low-impact materials recovered such as copper and MnSO 4 that have small recycling credits which do not outweigh the recycling process impacts (e.g., electricity, chemical use, and direct CO 2 emissions).Overall, where LFP and LMO show lower impact over certain nickel-based chemistries when considering battery production, the addition of recycling routes shows favorable net impacts for nickel-based chemistries due to the higher net recycling benefits.But, as done in typical LCA, these results are based on current data (representing 2020) for both production and recycling even though recycling is expected to occur in the future.However, the IMAGE electricity scenarios can now explore the future impact of recycling processes and future production processes that recycled materials will displace (Figures 4-6).Thus, the potential effects of production and recycling temporal mismatch can be investigated.
Using direct recycling as an example, Figure 8 considers battery production to occur in 2020 while recycling is varied to occur in subsequent years based on the RCP2.6 scenario for electricity.It becomes apparent that when recycling is considered to occur in a decarbonized future, the global warming potential reductions of recycling credits outweigh those of the recycling process, leading to underestimated net impacts.This is because compared to the 2020 baseline, materials recovered in the future displace production processes with lower impact due to decarbonized electricity.For example, assuming that LFP is recycled in 2020 leads to net recycling benefits of −22.4 kg CO 2 e kWh −1 which reduces the net impact by 32% from 71.0 to 48.6 kg CO 2 e kWh −1 .However, for LFP being recycled in 2040 under the RCP2.6 scenario, the net recycling benefits are instead −5.7 kg CO 2 e kWh −1 .In turn, this now reduces the net impact by 8% from 71.0 to 65.3 kg CO 2 e kWh −1 .The baseline approach overestimated net recycling benefits by 75% which led to an underestimated net impact by 34%.These effects are largest for LFP due to the substantial contribution of electricity to its foreground and background impact, although the direct recycling of other batteries still sees 23%−33% underestimated net impacts.For the nickel-based chemistries, a similar trend is seen for the pyrometallurgical, inorganic hydrometallurgical and organic hydrometallurgical routes where the net recycling benefits are overestimated by 0%−8%, 10%−13%, and 19%−20%, respectively (available in Supporting Information S4).This leads to underestimated net impacts by 0%−3%, 3%−7%, and 9%−13%, respectively.It is shown that in most cases, accounting for temporal mismatch leads to production having a greater percentage contribution to net impact than initially thought, emphasizing that more attention is needed to mitigate production impacts directly rather than relying on future recycling benefits that are ambiguous.
In other cases, where pyrometallurgical and inorganic hydrometallurgical recycling of LFP and LMO caused greater net impact (Figure 7), considering temporal mismatch shows overestimated net impacts as the impact-inducing recycling process is decarbonized in the future (available in Net global warming potential for battery production and direct recycling, with production occurring in 2020 and recycling in subsequent years under the RCP2.6 scenario.When the assumed date of recycling is moved into the future, the recycling benefits are reduced due to the increasingly decarbonized production processes that are displaced.These reduce at a greater magnitude compared to the recycling process which also decarbonizes.The underlying data for this figure can be found in Supporting Information S3. Supporting Information S4). Figure 8 also reveals that batteries may be further differentiated from each other and lead to changes in their rank order.For example, where the 2020 baseline net impact for LMO (43.7 kg CO 2 e kWh −1 ) and NMC111 (44.6 kg CO 2 e kWh −1 ) is similar, if direct recycling is considered to occur in 2040, the net impact of LMO (53.8 kg CO 2 e kWh −1 ) becomes increasingly lower than NMC111 (58.0 kg CO 2 e kWh −1 ).In another example, for organic hydrometallurgical recycling, the 2020 baseline net impact of LMO (53.9 kg CO 2 e kWh −1 ) is greater than NMC111 (52.3 kg CO 2 e kWh −1 ) (available in Supporting Information S4).However, considering recycling to occur in 2040 under RCP2.6,LMO (54.6 kg CO 2 e kWh −1 ) becomes lower than NMC111 (57.5 kg CO 2 e kWh −1 ).These effects are due to the temporal mismatch effects having an increased influence on nickel-based chemistries which have greater percentage contributions of net recycling benefits to their net impact.In both instances, LMO becomes a favored option over NMC111 showing that considering temporal mismatch can result in alternative product choices.

DISCUSSION
The global warming potential for current lithium-ion battery production mostly aligns with similar studies.Our results for nickel-based chemistries range 63−88 kg CO 2 e kWh −1 with the greatest and lowest impacts presented in NMC111 and NMC811, respectively.Studies using similar foreground data and high-carbon electricity in China for battery production align with these results of 68−100 kg CO 2 e kWh −1 (Kelly et al., 2020;Winjobi et al., 2022;Xu et al., 2022) with differences due to factors such as our study using a more recent ecoinvent v3.8 compared to v3.6 (Xu et al., 2022).However, it should be highlighted that our impact differences between nickel-based chemistries are more pronounced.Similar studies show that NMC111 is no more than 10% higher impact compared to NMC811 (Winjobi et al., 2022;Xu et al., 2022).In our case, NMC111 has a nearly 40% higher impact than NMC811.This is because we use the "market for cobalt sulfate-CN" activity introduced to ecoinvent v3.8 which has a notably greater global warming potential of 28.9 kg CO 2 e kg −1 compared to 2.6-9.6 kg CO 2 e kg −1 from other sources used (Winjobi et al., 2022;Xu et al., 2022).LMO being the lowest impact battery of 54 kg CO 2 e kWh −1 is expected (Crenna et al., 2021) and LFP having an impact of 71 kg CO 2 e kWh −1 is also expected (Xu et al., 2022).Our findings support that hotspots in battery production impacts are foreground electricity and active materials such as NiSO 4 and CoSO 4 (Chordia et al., 2021;Winjobi et al., 2022;Xu et al., 2022).Hence, we underpin that achieving lowimpact batteries requires maximizing specific energy to reduce the overall material and electricity demands (e.g., NMC811) and avoiding the use of high-impact CoSO 4 and NiSO 4 (e.g., LMO).
For the future global warming potential of lithium-ion battery production, our findings show some differences from a similar pLCA-IAM study (Xu et al., 2022).We applied the SSP2-RCP2.6scenario for electricity decarbonization from IMAGE revealing that by 2050, battery production impacts reduce by 38%−57%.In comparison, Xu et al. (2022) used the SSP2-PkBudg1100 scenario (similar to RCP2.6) for electricity decarbonization from REMIND to find that battery production impacts reduce more substantially by 48%−67%.A key difference is in the IAM electricity grid mix outputs for China in 2050.The SSP2-PkBudg1100 scenario from REMIND yields an electricity mix with 80% from low-carbon sources (e.g., hydro, nuclear, solar, and so on) and 20% from high-carbon natural gas.In contrast, our study using the SSP2-RCP2.6scenario from IMAGE yields 52% from lowcarbon sources, 26% from high-carbon natural gas, and 22% from fossil-fuel generation with carbon capture and storage.Hence, from 2020 to 2050, our study sees a lower 85% decarbonization in foreground electricity for China as opposed to the 92% decarbonization seen by Xu et al. (2022).Still, both studies agree that impact reductions depend on the relative contribution of foreground electricity in production and background electricity for input materials.While Xu et al. (2022) focus more on future metal supply scenarios, our study delves deeper into showing that despite the electricity mix decarbonizing substantially by 2050, a range of battery materials show less significant effects such as NiSO 4 having a relatively small 15% impact reduction.This uncovers the increased role of heat, fuels, and direct emissions where decarbonizing electricity mixes can be insignificant.As a result, we see that LMO and LFP-which have greater electricity contributions across their value chain-experience the most substantial future impact reductions.By 2050, this leads to LFP becoming favored over NMC811 showing the importance of weighing both presentday global warming potential and future decarbonization potential.Perhaps more importantly, we reiterate that deep decarbonization of product supply chains needs mitigation strategies for heat, fuels, and direct emissions which may not decarbonize with low-carbon electricity unless directly electrified.
Most recycling routes for lithium-ion batteries appear to reduce the net global warming potential compared to production especially for nickelbased chemistries since the recovery of NiSO 4 and CoSO 4 avoids their high-impact production.For example, our study shows hydrometallurgical recycling of nickel-based batteries can reduce net impacts by 25%−41% which aligns with similar studies that suggest net impact reductions of up to 25% (Mohr et al., 2020) and 38% (Kallitsis et al., 2022).On the other hand, pyrometallurgical and hydrometallurgical recycling of LFP and LMO has minimal benefits due to the recovery of low-impact materials which do not substantially outweigh the recycling process impacts.In general, direct recycling achieves significant net impact reductions across all batteries because of the considerable benefits of active cathode material recovery, although this method is in its infancy (Chen et al., 2019).Nonetheless, where considering battery production alone led to LFP and LMO being favored over some nickel-based chemistries, the addition of recycling leads to the nickel-based chemistries now having more favorable net impacts.Therefore, it could be argued that in the present day, the net impact of nickel-based chemistries is more favorable due to their high recycling benefits despite higher production impacts.However, this can be a misjudged interpretation as current recycling benefits rely on recovered materials displacing production processes based on present-day impacts despite batteries reaching their end-of-life 10−30 years later.
Our study is the first to explore the effects of this temporal mismatch by using the RCP2.6 electricity decarbonization scenario from IMAGE to set production to occur in 2020 and recycling to occur in 2030, 2040, or 2050 where recycled materials expect to displace production processes with reduced impacts.Consequently, we find that net recycling benefits are being overestimated in most cases by up to 75% which can underestimate net impacts by up to 34%.As a result, where the addition of recycling initially made the net impacts for nickel-based chemistries favorable over LFP and LMO, considering temporal mismatch finds that net impacts are comparable and, in some cases, demonstrating that the net impact of LMO becomes favored over NMC111.Two wider implications can be drawn from this.First, with the expectation that product value chains are decarbonizing, future recycling benefits for the global warming potential of long-lived products are being overestimated in typical LCA approaches.This is because recycled materials will displace future processes that are expected to have a lower impact compared to present-day processes.In consequence, net impacts are being underestimated, and the contribution of present-day production becomes greater to life cycle impacts.Therefore, mitigation should focus on present-day production given that relying on mitigation from future recycling benefits is uncertain and potentially misleading.Second, accounting for the potential overestimation of future recycling benefits when comparing products can lead to alternative product choices from a global warming potential perspective.Hence, for long-lived products, considering recycling benefits as a decisive factor is cautioned since these are again, uncertain, and potentially misleading.Both points reveal that aggregating production and recycling into net impacts can be a misrepresentative metric and should be distinguished in practice.
Nonetheless, we invite successive work to build on our limitations.First, the outcomes were based only on electricity decarbonization scenarios from IMAGE where it was identified that other IAMs and other decarbonization scenarios such as for heat and fuels could yield other outcomes.
The prospective assessment can also be expanded to consider improvements in battery technology such as potential changes in specific energies or how future recycling technologies could mature as organic hydrometallurgical and direct recycling processes are early development processes prone to scaling effects.Furthermore, only the global warming potential category was assessed where temporal mismatch effects for other impact categories may have different outcomes.For example, while the net global warming potential for batteries was found to be underestimated, for resource-based categories, the outcome may be the opposite if resource scarcity is expected to increase in the future.Last, these concepts should be expanded to other long-lived products where temporal mismatch for recycling or other unit processes may be an important determiner of LCA outcomes such as passenger vehicles or construction materials.

CONCLUSIONS
This is the first pLCA-IAM study to assess the effects of temporal mismatch in long-lived products.Using premise to generate future electricity scenarios from IMAGE for ecoinvent v3.8, the current and future global warming potential for production of six lithium-ion batteries and four recycling routes was evaluated to investigate the temporal mismatch effects when recycling is considered to occur in the future.Current batteries with the lowest production impacts demonstrate the best balance of (1) high specific energies to reduce overall material and energy demands (e.g., NMC811) and ( 2) avoiding the use of high-impact materials such as CoSO 4 (e.g., LMO).Future battery production impacts reduce up to 57% by 2050 under the RCP2.6 scenario with the greatest reductions achieved in batteries that have high contributions of electricity in both their foreground and background systems.Achieving further impact reductions in battery supply chains requires deep decarbonization strategies for heat, fuels, and direct emissions which will not be significantly influenced by low-carbon electricity unless directly electrified.Based on data for today's recycling processes, recycling appears to be an effective strategy to reduce the net global warming potential for battery production, particularly in nickel-based batteries.However, recycling will occur in the future and our pLCA-IAM shows that if electricity is decarbonized, recycling benefits in 2040 will be substantially lower (up to 75%) leading to higher net impacts (up to 34%).Consequently, this reveals that relying on future recycling to reduce net global warming potential from production is misleading, disincentivizing impact mitigation in production, and potentially leading to alternative product choices where recycling may be a decisive factor.Therefore, production and potential recycling benefits should be distinguished and cautioned while the greater focus should be placed on mitigating the impacts of present-day production provided future recycling benefits are uncertain.

Figure 7
Figure7reveals that the addition of recycling routes to present-day batteries is beneficial in most cases, showing reduced net global warming potential compared to production alone.For nickel-based chemistries, pyrometallurgical, inorganic hydrometallurgical, organic hydrometallurgical,