Analysis

https://doi.org/10.1038/s41893-019-0222-5

Examining different recycling processes for

lithium-ion batteries
Rebecca E. Ciez   1,4 and J. F. Whitacre1,2,3*

Finding scalable lithium-ion battery recycling processes is important as gigawatt hours of batteries are deployed in electric vehi-
cles. Governing bodies have taken notice and have begun to enact recycling targets. While several battery recycling processes
exist, the greenhouse gas emissions impacts and economic prospects of these processes differ, and could vary by specific bat-
tery chemistry. Here we use an attributional life-cycle analysis, and process-based cost models, to examine the greenhouse gas
emissions, energy inputs and costs associated with producing and recycling lithium-ion cells with three common cathode chem-
istries: lithium nickel manganese cobalt oxide (NMC-622), lithium nickel cobalt aluminium oxide and lithium iron phosphate.
We compare three recycling processes: pyrometallurgical and hydrometallurgical recycling processes, which reduce cells to
elemental products, and direct cathode recycling, which recovers and reconditions ceramic powder cathode material for use in
subsequent batteries—retaining a substantial fraction of the energy embodied in the material from their primal manufacturing
process. While pyrometallurgical and hydrometallurgical processes do not significantly reduce life-cycle greenhouse gas emis-
sions, direct cathode recycling has the potential to reduce emissions and be economically competitive. Recycling policies should
incentivize battery collection and emissions reductions through energetically efficient recycling processes.

T
he electric vehicle market has grown quickly; sales were neg- an analysis of the life-cycle carbon emissions associated with direct
ligible before 2010 and have grown to over 2 million vehicles recycling of lithium manganese oxide (LMO) cathode materials.
sold by 20161. If manufacturers meet their 2020 production However, LMO is more unstable and less energy dense than other
targets, annual production capacity would be on the order of at cathodes18, and therefore uncommon in today’s electric vehicles
least 40 GWh yr−1, or 200,000 tonnes of lithium-ion battery cathode unless blended with other cathodes19. The analysis also assumed
material annually2,3. With this production scale, and the number of that 100% of the cathode would be recovered and did not consider
batteries that will be retired, automakers have expressed an interest the cost relative to conventional cathode manufacturing.
in a range of end-of-life scenarios for spent packs and battery cells, Using a Monte Carlo simulation of an attributional life-cycle
including recovering materials from used battery packs4,5. Lithium- analysis, we compare the GHG emissions from manufacturing
ion batteries also contain high-value and energy-intensive cathode and recycling batteries using pyrometallurgical, hydrometallurgi-
materials, including nickel and cobalt. As lithium-ion batteries have cal and direct cathode recycling methods for three cathode chem-
become more prevalent, several jurisdictions have imposed regu- istries: nickel manganese cobalt oxide (NMC; specifically, one
lations requiring their safe disposal: the European Union requires with 60% nickel, 20% manganese and 20% cobalt (NMC-622)),
45% of rechargeable batteries to be collected, and half of those bat- nickel cobalt aluminium oxide (NCA) and iron phosphate (LFP).
teries must be recycled6; China is restricting battery disposal in NMC and NCA batteries account for most of the storage capacity
landfills7; and some states (New York, California and Minnesota) in light-duty electric vehicles in the United States and are eligible
have rechargeable battery recycling mandates (http://www.call2re- for Chinese government subsidies7, while LFP cells are common
cycle.org/recycling-laws-by-state/; see also ref. 8). However, these in electric buses. We examine both cylindrical (18650 and 20720)
policies do not stipulate how these batteries should be recycled, and pouch cells because they are both represented in the US elec-
or whether the recycling process must result in a net reduction in tric vehicle fleet19. We consider three electricity grid mixes: the US
greenhouse gas (GHG) emissions. average, a low emitting grid (the Northwest Power Pool (NWPP)
Today, both the facilities and processes to recycle lithium-ion bat- North American Electric Reliability Corporation subregion) and
teries are limited. Most efforts to improve recycling have focused on a higher-emitting grid (the ReliabilityFirst Corporation/Michigan
recovering valuable metals, especially cobalt9–15. Pyrometallurgical (RFCM) North American Electric Reliability Corporations subre-
and hydrometallurgical recycling facilities are operating at scale, gion). These two subregions are also of interest because of existing
but require significant energy inputs for extensive processing of electric vehicle battery manufacturing facilities, where new recy-
the recovered material. Recently, direct physical processes16, which cling facilities could be co-located. Location also affects shipping
recover battery materials as a mixture, rather than as individual distances and transportation options, and we conduct a paramet-
metals, have been developed. These direct physical processes take ric analysis to assess the relative role of transportation. Since direct
advantage of the limited battery chemistries and pack formats of cathode recycling is a nascent process, we examine how the cath-
electric vehicles to target the known cathode materials—the sec- ode yield rate affects GHG emissions. Using a process-based cost
ond-largest contributor to material costs and GHG emissions of model, we determine the cost of traditional cathode manufacturing
battery manufacturing, as shown in Fig. 1. Dunn et al.17 conducted and find breakeven recovery costs for direct cathode recycling to be

1
Department of Engineering and Public Policy, Carnegie Mellon University, Pittsburgh, PA, USA. 2Department of Materials Science and Engineering,
Carnegie Mellon University, Pittsburgh, PA, USA. 3Wilton E. Scott Institute for Energy Innovation, Carnegie Mellon University, Pittsburgh, PA, USA.
4
Present address: Andlinger Center for Energy and the Environment, Princeton University, Princeton, NJ, USA. *e-mail: [email protected]

148 Nature Sustainability | VOL 2 | FEBRUARY 2019 | 148–156 | www.nature.com/natsustain

NaTuRe SusTainabiliTy Analysis
42 kg CO2e kWh−1 US$186 kWh−1 LFP cells, which are less energy dense than the other chemistries
100%
considered and do not contain nickel or cobalt (these produce more
Overhead (11%) GHG emissions than LFP inputs), the GHG emissions impact of the
cathode material inputs contributes a smaller fraction to the total
manufacturing emissions. The cell-manufacturing process accounts
for larger percentages of CO2e emissions, and for cylindrical and
Cell Operating pouch LFP cells, the manufacturing GHG emissions for a factory
80% manufacturing expenses
(40%) (18%)
in the NWPP are significantly lower than for a facility located in
RFCM. There are no scenarios where the emissions of manufactur-
ing LFP cells in RFCM are lower than manufacturing them in the
NWPP (as shown in Supplementary Figs. 9–12).
As Table 1 shows, there is variation in the literature regarding
GHG emissions and energy consumption per kg of battery pro-
60% Transportation (1%) Capital
duced. While our results lie towards the lower end of the range of
expenses the results of the existing literature, drawing direct conclusions is
(30%) difficult based on different assumptions about the energy consumed
during different processes, the impact of plant utilization in other
studies, and the evolution of cell designs and materials since 2010,
Cathode
(30%)
increasing the specific energy of cathode materials (and cells over-
40% all)20,21. Supplementary Table 1 provides the same information on a
Cathode
(8%) per kWh basis.
Anode (3%)
Cell recycling GHG emissions
Binder, solvent and
electrolyte (5%) Fig. 3a shows the CO2e emissions associated with each kg of cylin-
Current collectors and drical battery manufactured and recycled using pyrometallurgi-
20% separators (7%) cal, hydrometallurgical and direct cathode recycling methods for
Other
material the US average power grid. For both the hydrometallurgical and
Other material
(29%) (33%) direct recycling processes, we assume that any materials not recov-
Cell ered through the recycling process (for example, transition metals,
hardware (18%) lithium or cathode materials) or as scrap (cell hardware and current
collectors) are incinerated. Fig. 3b shows the medians and 95% con-
0% fidence intervals for the GHG emissions avoided when using each
GHG emissions per Cost per kWh recycling process. Fig. 3c,d shows the results for pouch cells. For
kWh NMC battery NMC battery
LFP cells, all of the recycling methods and cell formats considered
Fig. 1 | Manufacturing cost and CO2e emissions for NMC cylindrical cells. result in net increases in CO2e emissions when accounting for the
Cathode materials are a substantial contributor to overall costs and CO2e GHG emissions from incinerating non-valuable materials. The iron
emissions for manufacturing NMC cylindrical cells. precursor materials used in LFP cathodes is more energetically effi-
cient to mine and emits lower levels of GHGs per kg and per kWh
compared with the nickel, cobalt and manganese precursors of the
competitive. Finally, we comment on how the recycling processes other cathodes (see Supplementary Table 17), resulting in a smaller
and their impacts align with recycling regulations. GHG emissions offset from the recovered materials that is insuf-
ficient to offset the energy and GHG emissions associated with the
Cell manufacturing GHG emissions recycling processes considered.
Fig. 2 reports the GHG emissions resulting from battery manu- For NMC and NCA cells, there is a median reduction in GHG
facturing, using CO2e values and 20-year warming potentials. Cell emissions from both hydrometallurgical and direct cathode recy-
input materials are a substantial contributor to manufacturing GHG cling. However, these reductions are only statistically significant
emissions, especially when facilities utilize lower-emitting electric- for pouch cells recycled via a direct cathode recycling process.
ity. Plots of the GHG emissions and input energy required per kg For pyrometallurgical recycling, the GHG emissions offsets are
of battery and per kWh of battery are provided in Supplementary a combination of the output slag used as a cement input substi-
Figs. 1–8. For NMC cylindrical cells, if the facility is in the RFCM tute and the alloy of high-value metals recovered (nickel, copper,
subregion, which is higher emitting than the US average, materials iron and cobalt). For both nickel- and cobalt-containing cathodes,
account for 50% of total GHG emissions per kg of battery, while the median result is a net increase in GHG emissions, and these
they account for 60% of the total if the facility is in the lower-emit- increases are statistically significant except for cylindrical NCA
ting NWPP subregion. Fig. 2 also shows that our baseline trans- cells, where the combination of cathode materials that are recov-
portation assumptions have a minimal impact on GHG emissions erable through the pyrometallurgical process (manganese is not
per kg of battery produced. Although there is a difference in the recovered, but nickel and cobalt are) and the cell hardware result
median GHG emissions depending on the grid assumptions, the in some ambiguity (22% of the scenarios result in net reductions in
statistical significance of these differences varies depending on the CO2e emissions per kg, and these scenarios occur when the recycled
cathode material. Fig. 2 presents the medians and 95% confidence content of the nickel and cobalt offset is low and embodied emis-
intervals for GHG emissions based on the Monte Carlo simulations. sions are high). Supplementary Figs. 13–16 show that pyrometal-
For NMC and NCA cells with high specific energy capacity, at a lurgical recycling results in net increases in GHG emissions on all
95% confidence interval, there is no statistically significant differ- three electricity grids and when considering the results per kWh
ence between the CO2e emissions for any of the grids. Cumulative of battery recycled, whereas with hydrometallurgical and direct
distribution plots of the total manufacturing GHG emissions are recycling reductions in GHG emissions are possible for NMC and
available in Supplementary Figs. 9–12, which show that there are NCA cells—but the reductions are uncertain. LFP recycling always
significant differences at the 90% confidence interval. However, for results in increases in GHG emissions. Similarly, pyrometallurgical

Nature Sustainability | VOL 2 | FEBRUARY 2019 | 148–156 | www.nature.com/natsustain 149

Analysis NaTuRe SusTainabiliTy

a NMC NCA LFP b NMC NCA LFP

12 12

10 10
Emissions (kg CO2e kg battery−1)

8 8

6 6
Manufacturing

4 4 Transport

Cathode
materials
2 2
Other
materials
0 0
S
PP

S
PP

S
PP

S
PP

S
PP

S
PP

M
U

U
FC

FC

FC

FC

FC

FC
W

W
R

R
N

N
Fig. 2 | Cell manufacturing emissions. a,b, Median CO2e emissions per kg of cell during the manufacturing of NMC, NCA and LFP cylindrical (a) and
pouch cell batteries (b), using US, NWPP and RFCM average grid emissions data. Error bars represent 95% confidence intervals. Manufacturing includes
the cathode material preparation process.

recycling results in net increases in energy consumption, whereas However, since pyrometallurgical recycling results in net increases
hydrometallurgical and direct recycling have the potential to in GHG emissions, the cathode yield rate from a direct cathode
reduce energy consumed for NMC and NCA cells (as shown in recycling process can be lower and still result in lower GHG emis-
Supplementary Figs. 17–20). sions increases than pyrometallurgical recycling.
Fig. 3 assumes that the direct recycling process recovers 100%
of the cathode material, but we are interested in determining how Impact of transportation
much cathode material must be recovered to result in GHG emis- Transportation of the input materials and collection of spent batter-
sions reductions. Since the pyrometallurgical and hydrometal- ies is responsible for a small fraction of both the total energy input
lurgical processes do not have significant environmental benefits, into the process and the overall GHG emissions. In total, transpor-
we compared the GHG emissions reductions from a direct cath- tation only accounts for 0.33 kg CO2e per kg battery—roughly 3.5%
ode recycling method over not recycling the manufactured cells. of the total CO2e emissions when using a pyrometallurgical process,
Holding our cell construction and materials input assumptions and 4% when using a hydrometallurgical process or direct cathode
fixed, we compare the avoided CO2e emissions, as the yield rate recycling method.
for the direct cathode recycling process varies from 0 to 100%. As Of the 0.33 kg of CO2e, roughly 70% is attributed to battery col-
Fig. 4 shows, using direct cathode recycling for NMC cells offers lection. Because of strict regulations on battery shipping, they are
slightly larger emissions savings than using the direct cathode recy- transported on trucks (the highest-emitting form of transporta-
cling process for NCA cells. No amount of LFP recovered through tion)22. Without knowing the specific collection point of spent bat-
direct recycling is sufficient to offset the GHG emissions that result teries, and assuming that a few facilities would be co-located near
from the recycling process or incineration of other waste products, existing electric vehicle manufacturing plants, we also assumed that
for either cylindrical or prismatic cells. For cylindrical NMC cells the collection distance travelled was 2,500 miles. Doubling the emis-
(Fig. 4a), 78% (53–115%) of the cathode material must be recovered sions factors, or both the emissions factors and distances travelled,
to result in a net CO2e emissions reduction with 95% confidence. has a relatively small impact on the total emissions per kg. Doubling
For NCA cells, 88% (58–133%) of the cathode would need to be the emissions factors increases CO2e emissions per kg battery by
recovered. For pouch cells, 59% (49–74%) of NMC cathodes or 4%, while doubling both the emissions factors and distances trav-
63% (48–87%) of NCA cathodes must be recovered. Supplementary elled increases the CO2e emissions per kg battery by 12%.
Figs. 21–28 show the breakeven emissions and energy input savings
for different direct cathode recycling yield rates on a per kg and per Breakeven recovery cost
kWh basis, for all three electric grid options. Our process-based cost model of traditional cathode material
The yield rate of cathode material recovered from direct recy- manufacturing from purchased materials shows that, at economies
cling required for direct cathode recycling to have a smaller GHG of scale, the costs of manufacturing 1 kg of NCA, NMC or LFP
impact than pyrometallurgical or hydrometallurgical recycling is cathode are US$24 (US$16–43), US$21 (US$14–40) and US$20
dependent on the cell format and chemistry. Supplementary Table 2 (US$13–43), respectively. The at-scale cost of relithiating the cath-
summarizes the breakeven cathode yield rates for direct cathode ode materials recovered from a direct cathode recycling process is
recycling to outperform hydrometallurgical and pyrometallurgical approximately US$6 kg−1 (US$4–13) for both NMC and NCA cath-
recycling of cylindrical cells. Supplementary Table 3 contains the odes, assuming that 50% of the lithium is replaced, and economies
same data for pouch cells. Since the hydrometallurgical recycling of scale are reached (depicted in Fig. 5) at approximately 1,500 met-
process offers median reductions in CO2e emissions, the recovery ric tonnes yr−1—enough for roughly 1 GWh of cells (well within
rate of cathode material from a direct recycling process required current manufacturing capacities). Since LFP fabrication does not
to result in more CO2e offsets is high (>40% across all metrics). include premixing and drying steps to combine transition metals,

150 Nature Sustainability | VOL 2 | FEBRUARY 2019 | 148–156 | www.nature.com/natsustain

NaTuRe SusTainabiliTy Analysis
Table 1 | Comparison of GHG emissions and energy inputs to cell manufacturing between this and other studies
Reference Cathode chemistry Specific energy GHG emissions (kg CO2e) and input energy (MJ) per kg of battery
(kWh kg−1)
Materials Manufacturing Total
Ref. 54
LMO 0.114 5 kg 0.98 kg 6 kg
86 MJ 18 MJ 100 MJ
Ref. 55 LFP (pouch cells) 0.100 7.7 kg 9.5 kg 17 kg
– – –
Ref. 56 NCM 0.112 16 kg 6 kg 22 kg
130 MJ 80 MJ 210 MJ
LFP 0.088 16 kg 6.1 kg 22 kg
130 MJ 80 MJ 210 MJ
Ref. 17 LMO (pouch cells) 0.130 4.8 kg 0.27 kg 5.1 kg
72 MJ 2.7 MJ 75 MJ
Ref. 57 Average 0.090 7.2 kg 3 kg 10 kg
110 MJ 55 MJ 160 MJ
Ref. 57; cells only Average 0.150 11 kg 0.09 kg 11 kg
160 MJ 1.4 MJ 160 MJ
Ref. 58 NCM (pouch cells) 0.110 6.8 kg 11 kg 18 kg
– – –
Ref. 58; cells only NCM (pouch cells) 0.174 9.2 kg 19 kg 28 kg
– – 100 MJ
Ref. 59 LMO/NCM (pouch cells) 0.080 6.1 kg 5.2 kg 11 kg
– – 120 MJ
Ref. 59; cells only LMO/NCM (pouch cells) 0.140 4 kg 9.1 kg 13 kg
– – –
This study (US average NMC (cylindrical) 0.210 5.2 kg (4.4–6.3 kg) 3.6 kg (3.0–4.2 kg) 8.9 kg (7.9–9.9 kg)
power mix, with 95% 69 MJ (57–80 MJ) 49 MJ (41–57 MJ) 120 MJ (108–130 MJ)
confidence intervals)
NCA (cylindrical) 0.190 5.2 kg (4.1–6.3 kg) 4.9 kg (4.1–45.8 kg) 10.3 kg (9.1–11.5 kg)
69 MJ (57–80 MJ) 59 MJ (50–70 MJ) 130 MJ (117–143 MJ)
LFP (cylindrical) 0.100 4.8 kg (3.7–6.0 kg) 3.5 kg (2.9–4.1 kg) 8.5 kg (7.4–9.4 kg)
64 MJ (50–77 MJ) 48 MJ (40–56 MJ) 114 MJ (101–126 MJ)
NMC (prismatic) 0.270 5.8 kg (5.1–6.8 kg) 3.9 kg (3.4–4.4 kg) 9.8 kg (8.9–10.8 kg)
69 MJ (57–80 MJ) 49 MJ (41–57 MJ) 120 MJ (108–130 MJ)
NCA (prismatic) 0.260 5.8 kg (5.1–6.8 kg) 5.3 kg (4.7–6.0 kg) 11.3 kg (10.2–12.4 kg)
69 MJ (57–80 MJ) 59 MJ (50–70 MJ) 130 MJ (117–143 MJ)
LFP (prismatic) 0.19 5.4 kg (4.5–6.5 kg) 3.9 kg (3.4–4.4 kg) 9.5 kg (8.4–10.6 kg)
64 MJ (50–77 MJ) 48 MJ (40–56 MJ) 114 MJ (101–126 MJ)

the cost of relithiation approaches the cost of traditional LFP cath- Relithiation costs also depend on how much lithium carbonate
ode manufacturing. is required. Supplementary Fig. 17 shows both the cost of relithia-
A major contributor to the uncertainty in the cost of relithia- tion and the breakeven cost for direct recycling to be profitable as
tion is the uncertainty bounds on the cost of lithium carbonate. We the percentage of lithium required during relithiation varies from 0
have assumed an upper bound of US$25 kg−1—the cost of extracting to 100%. In practice, the maximum amount of lithium that would
lithium from seawater23. However, this is a global cost and would need to be added during reprocessing is 60%, as the crystal struc-
affect both the cost of manufacturing new cathode materials and tures of these NCA and NMC cathode materials are unstable if more
relithiating recovered cathode material. As such, the uncertainty than 60% of the lithium is lost, so they are rarely (if ever) charged to
bounds on the cost at which recycling is economically competitive beyond this point24. Since LFP fabrication does not include a sepa-
with traditional cathode manufacturing are lower. If NMC cathode rate mixing step to combine transition metals, and the other cathode
material can be recovered for US$15 kg−1 (US$10–27) or less, NCA material inputs are very inexpensive, as the percentage of lithium
cathode can be recovered at US$19 kg−1 (US$12–31) and LFP cath- added approaches 100%, the process approaches the total cost of
ode can be recovered for US$3.90 kg−1 (US$2.60–10.00), the refunc- LFP cathode manufacturing, narrowing the potential profitability
tionalized cathode materials could be produced at the same cost as of a direct cathode recovery process, as shown in Supplementary
traditional cathode manufacturing methods, making direct cathode Fig. 29. Supplementary Figs. 30 and 31 show that the variation in
recycling economically competitive. This breakeven cost would lithium added during reprocessing has no significant impact on the
have to include the cost of collecting and transporting the cells, in GHG emissions associated with direct recycling for cylindrical and
the absence of other policies to defray the cost. prismatic cells.

Nature Sustainability | VOL 2 | FEBRUARY 2019 | 148–156 | www.nature.com/natsustain 151

Analysis NaTuRe SusTainabiliTy

a c
15 NMC NCA LFP 15 NMC NCA LFP

Emissions (kg CO2e kg battery−1)

10 10

5 5

0 0
al ical

D l
ct

al ical

al

ct

al cal

al

ct

al ical

al

ct

al ical

al

al

al

ct
a

c
ic

re

ic
ire

ic
ire

ic
ire

ic
ire

et gic

ic
ire
i
m lurg

rg

m lurg

rg

m urg

rg

m lurg

rg

m lurg

rg

rg
i

D
r
lu

lu

lu

lu

lu

ro allu

lu
ll
l

al
yd eta

ta

ta

yd eta

ta
et

et

et

et

et

t
H me

H me

H me

H me
m

m
ro

ro

ro

ro

ro

ro

ro

ro

ro

ro

ro
Py

Py

yd

Py

yd

Py

Py

yd

Py

yd
H

H
b d
4 NMC NCA LFP 4 NMC NCA LFP
(kg CO2e kg battery−1)
Emissions avoided

2 2

0 0

–2 –2

–4 –4
al

al

ct

al ical

D l
ct

al ical

D l
ct

al

al

ct

al ical

D l
t

al

al

ct
a

c
ic

ic
ire

ic
ire

ic
ire

ic

ic
ire

ic
ire

et gic

ic
ire
rg

rg

m lurg

rg

m lurg

rg

rg

rg

m lurg

rg

rg
D

D
r
lu

lu

lu

lu

lu

lu

lu

ro allu

lu
al

al

al

al

al
yd eta

ta

yd eta
et

et

et

et

et

et

et

t
H me

H me
m

m
ro

ro

ro

ro

ro

ro

ro

ro

ro

ro

ro
Py

yd

Py

Py

yd

Py

yd

Py

Py

yd
H

Fig. 3 | Battery recycling emissions. a–d, Medians and 95% confidence intervals for CO2e emissions per kg less the CO2e offsets from recovered materials
(a,c), and net CO2e emissions avoided by using each recycling process (b,d) for cylindrical (a,b) and pouch battery manufacturing and recycling processes
(c,d). All processes use US average electricity grid data.

Discussion lead acid recycling process is also simple and profitable27. In con-
Although there is potential for both economic benefits and GHG trast, lithium-ion batteries have no overarching federal regulation,
emissions reductions from a direct cathode recycling process over a and the recycling requirements vary by state. Even states with lith-
pyrometallurgical or hydrometallurgical process, there is some con- ium-ion battery recycling policies focus efforts on recycling small
trast between these environmental and economic objectives. The cells, such as those found in consumer electronics8. Implementing
significant manganese content in NMC-based cells increases the consistent regulations for automotive lithium-ion batteries across
potential environmental benefit from using a direct cathode recy- state lines, and adopting policies such as refundable deposits with
cling process, but makes manufacturing the cathode from raw input battery purchases, could increase collection rates. Non-refundable
materials less expensive, narrowing the potential window for eco- deposits could help to fund battery collection and transportation
nomic viability. For NCA cells, which have a higher percentage of to recycling facilities, easing the financial burden on recyclers and
high-value metals (cobalt and nickel), the cost associated with man- allowing more financial resources to be devoted to recovering
ufacturing the cathodes from raw materials is higher, increasing the higher-value materials.
potential payoff from a direct cathode recycling process, especially The European Union also has regulations requiring lead acid
because of recent increases in cobalt commodity prices (tradingeco- automotive batteries to be recycled, although collection rates have
nomics.com/commodity/cobalt; see also ref. 25). However, existing lagged behind the United States28. They also have requirements for
recycling pathways for nickel and aluminium, outside of battery lithium-ion battery recycling. The European Union battery direc-
recycling, reduce the net GHG offsets possible from direct recycling tive requires 50% by weight of the total battery content be recycled6.
of NCA cells. For pyrometallurgical recycling, much of the cathode material, in
While recycling programmes for lead acid automotive batteries addition to some current collector and cell canister material, is out-
have been successful (especially in the United States, where 99% of put as part of the metal alloy. The other non-combustible cell com-
lead acid batteries are recycled)26, the prognosis for lithium-ion bat- ponents are output as part of slag that can be repurposed as cement
teries is less clear; in the US, federal law requires lead acid batteries input materials. As such, combined, the metal alloy and slag mate-
to be recycled, and many states have enacted exchange mandates: rial satisfy the requirement that 50% by weight of battery materials
sellers must accept old batteries at the time of new purchases. The be recycled to be compliant with European Union regulations, even

152 Nature Sustainability | VOL 2 | FEBRUARY 2019 | 148–156 | www.nature.com/natsustain

NaTuRe SusTainabiliTy Analysis
a though pyrometallurgical recycling usually results in net increases
2.5 in GHG emissions. Our analysis also does not consider any poten-
tial downstream ecological impacts that could result from using the
2.0
slag material as a cement input29.
1.5 For hydrometallurgical and direct cathode recycling, ensuring
Emissions avoided (kg CO2e kg battery−1)

compliance with European Union regulations depends on addi-

1.0
tional recycling of cell materials, despite the more substantial envi-
0.5 NMC ronmental benefits. Assuming a 100% yield, the cathode material
NCA recovered during direct recycling of cylindrical cells accounts for
0
0 20 40 60 80 100
approximately 30% of cell mass—less than the 50% required. Others
b have noted similar mismatches between the environmental impact
2.5 of recycling and the mass requirements of manufacturing for other
waste streams30. Recovering and recycling other components, such
2.0
as current collectors (~10% of cell mass) and cell canisters (~30%
1.5 of cell mass), in addition to direct recycling of the cathode, could
NMC
NCA bring direct the recycling process into compliance. Selling them as
1.0
scrap metal could also be another income source from the direct
0.5 recycling process. Whether this combination of processes—poten-
tially undertaken at multiple facilities by multiple actors—would be
0
0 20 40 60 80 100 considered compliant depends on how the European Union direc-
Cathode yield rate (%) tive defines the boundaries of recycling facilities and the yield rate
of the direct recycling process.
Fig. 4 | Avoided emissions of direct cathode recycling with different
cathode yield rates. a,b, Medians and 95% confidence intervals for net Recycling and future market demand for metals. The demand for
CO2e emissions avoided per kg battery when combining manufacturing many metals used in lithium-ion batteries is expected to continue to
with direct cathode recycling over using no recycling method after increase in the coming decades31,32, with potential implications for
manufacturing. The GHG offsets for cylindrical cells (a) are lower than the economic viability and environmental benefits of pyrometallur-
for pouch cells (b). Recycling LFP cells does not result in net reductions in gical and direct cathode recycling. Lithium and cobalt prices have
GHG emissions because of the incineration of non-valuable materials and been particularly volatile. Our cost assessments for cathode produc-
the lack of nickel and cobalt precursors. tion and relithiation account for the wide uncertainty in future lith-
ium prices. However, the breakeven costs between relithiation and
traditional cathode manufacturing are not as sensitive to changes
in the cost of lithium, as both processes would be affected by global
lithium price shocks.
Cobalt prices have also been volatile, roughly doubling between
Annual production volume (GWh) 2016 and 2017 (tradingeconomics.com/commodity/cobalt; see
NMC also ref. 33). However, because cobalt is a fraction of both NMC-
0 0.5 1.0 1.5
622 and NCA cathodes, our model shows that this change in price
0 0.5 1.0 1.5
NCA only increased the cost per kg of the NMC or NCA cathode by
~US$3 kg−1 (roughly 12.5%). Ongoing research is focusing on fur-
LFP
0 0.5 1.0 ther reducing the cobalt content in NMC cathodes, reducing the
sensitivity to cobalt price shocks further. Regardless of cobalt con-
50 tent, the increase in the cost of traditional manufacturing increases
the breakeven cost for direct cathode recycling to be competitive.
40 Beyond the economic prospects of recycling lithium-ion batteries,
these processes could also serve as an alternative to complex global
Cost (US$ kg cathode−1)

supply chains potentially with unethical labour practices. Banza

30 Lubaba Nkulu et al.34 found that populations surrounding artisanal
mines in the Democratic Republic of the Congo have elevated levels
of cobalt in their urine and blood, potentially adversely affecting the
20
LFP heart, lungs, blood and thyroid, and as Olivetti et al.35 and Habib et
al.36 discuss, the Democratic Republic of the Congo holds a large
10 stake of control in the global market. The concentration of cobalt
NCA resources, along with the geopolitical instability of the Democratic
NMC
Republic of the Congo, results in price instability and the potential
0
0 500 1,000 1,500 2,000 2,500 for short-term supply gaps.
Annual production volume (tonnes) Increasing demand for metals will also result in less favour-
able sources being mined at lower efficiencies, requiring addi-
Fig. 5 | Relithiation costs. Estimated relithiation costs per kg of cathode tional energy and producing more emissions31, but also potentially
material for NMC, NCA and LFP cells as production volumes vary and 50% shifting the amount of material recycled through other pathways.
of lithium is replaced. Shaded regions show the most and least optimistic These currently unused recycling methods for non-battery materi-
cost estimates. The process reaches economies of scale at roughly 1,500 als would also probably consume more energy than the recycling
tonnes of annual production. Corresponding production volumes in GWh methods currently in use. These changes would shift both the upper
of capacity produced are provided on the secondary x-axis for NMC, NCA bound and the peak for the triangular distributions we used to assess
and LFP cathodes. LFP costs are higher because there is no premixing of the emissions associated with input materials such as cobalt and
transition metals before the cathode is lithiated. nickel, thereby increasing the emissions and energy offsets for both

Nature Sustainability | VOL 2 | FEBRUARY 2019 | 148–156 | www.nature.com/natsustain 153

Analysis NaTuRe SusTainabiliTy

recycling processes. Increasing the underlying GHG emissions and with existing recycling infrastructure, such as nickel and steel, we have included
energy consumed to produce the input materials would increase the triangular distributions of the potential emissions, with the minimum GHG
emissions from existing recycling processes, the maximum GHG emissions from
offsets from all of the recycling processes, but does not guarantee mining operations and the peak of the distribution from the current assumption
that all of the recycling processes would result in net reductions in based on the mix of recycled and virgin sources.
GHG emissions or embodied energy. Hydrometallurgical and direct For the energy consumed at the battery manufacturing and recycling
cathode recycling offsets would still be larger than pyrometallurgi- facilities, we assess the damages from natural gas consumption using GREET
2016 assumptions for a kiln using North American shale gas. For grid electricity
cal recycling, and they can better capitalize on grid electricity emis-
emissions, we use Environmental Protection Agency Emissions & Generation
sions reduction trends. Recycling LFP cells, which lack the metals Resource Integrated Database 2014 data46. Specifically, we consider the emissions
such as cobalt and nickel that are expensive and energy intensive to associated with using a US average grid, the NWPP subregion grid and the
produce, does not offer net reductions in GHG emissions for any of RFCM subregion grid. These subregions were selected based on known electric
the processes we consider, and the low breakeven cost for recover- vehicle manufacturing locations, and the emissions assumptions are summarized
in Supplementary Table 19. We use average emissions assumptions because the
ing cathode materials (US$3.50 kg−1) limits the economic viability processes modelled operate continuously.
of LFP recycling compared with other cathode chemistries.
Transportation assumptions. We also use the location assumptions to create
Methods rough estimates of the shipping distances for raw materials and the collection
We used a Monte Carlo simulation (with 5,000 scenario runs) to conduct an distances for cell recycling. Estimates for the distances travelled were calculated
attributional life-cycle assessment of the materials, manufacturing and recycling using information about both where materials are produced and tools that calculate
of lithium-ion batteries. We focused our analysis on just the GHG emissions and distances between major ports. For steel and graphite, we assumed that the
energy consumption of these steps, and did not include the emissions or energy materials would be shipped from Shanghai to Long Beach (for NWPP) or Newark
associated with their use in electric vehicles. Supplementary Fig. 32 outlines the (RFCM) (https://www.searates.com), and then would be shipped via rail for
manufacturing, pyrometallurgical, hydrometallurgical and direct cathode recycling 600 miles each (approximately the distance between Newark and Detroit, or
processes modelled. Long Beach and Reno). Most of the bauxite used in the United States is imported
The cylindrical cell bill of materials for each simulation was constructed from Jamaica (GREET 2016), so we calculated the distance from Kingston to
based on literature estimates of cell contents for two cell dimensions: 18650 Newark, and then the shipping distance by rail). Aluminium, copper, nickel,
(18 mm diameter, 65 mm tall) and 20720 (20 mm diameter, 72 mm tall)37. The cobalt and manganese are all produced in North America, and the United States
uncertainty in cylindrical cell dimensions is captured in triangular distributions Geological Survey has maps with locations of these production facilities (https://
with the upper, lower and best estimates of the mass of each cell component. mrdata.usgs.gov/mineral-resources/copper-smelters.html and ref. 33). Distances
Additional information about the cell material inputs and dimensions is provided between these facilities and approximate locations were calculated and attributed
in Supplementary Tables 4, 6 and 8. The bills of materials for the pouch cells to rail shipping. For the other input chemicals, production is not tied to a specific
were drawn from BatPaC (version 3.1; issued 28 June 2018; http://www.cse.anl. location, and we assumed 500 miles by rail. For all of the inputs, we assumed
gov/batpac/). Unlike cylindrical cells, the external dimensions of pouch cells are 50 miles of road transport to reach the final destination. The distances assumed
variable, and can be optimized based on cell chemistry to meet specified power for different modes of transportation (water, rail and road; https://www.searates.
and energy storage targets. We included both low- and high-energy capacity cells com and https://mrdata.usgs.gov/mineral-resources/copper-smelters.html) and
in this analysis, and the details of the input materials are listed in Supplementary damages per tonne mile are listed in Supplementary Tables 20 and 21. We also
Tables 3, 5 and 9. For each simulation in the analysis, a cell was constructed from conducted a sensitivity analysis on these estimates to determine the impact of
the distributions of potential dimensions, and the cell energy storage capacity shipping on overall emissions and energy consumption, doubling the distance
and total mass was calculated and used to normalize the analysis per kg of battery travelled as well as the GHG emissions and energy consumption per tonne mile.
produced, or per kWh of battery capacity produced.
For the NMC and NCA cathodes, we considered two types of inputs to produce Cell collection assumptions. Without a specified location for a recycling facility,
the cathode precursors: nitrates and sulfates. Supplementary Tables 10–13 detail or a specified point of origin for the collected batteries, we conservatively estimated
the required inputs for both pathways38. These precursors are then combined that the batteries would have to travel 2,500 to be recycled (roughly the distance
with lithium carbonate and calcined to produce the final cathode material (details between New York and San Francisco), and that this transportation would be
included in Supplementary Tables 14 and 16). For LFP, there is no separate entirely by truck. This is in line with US Department of Transportation regulations
mixing and drying step to combine multiple transition metal precursors; iron that specify this mode of transportation for all lithium-ion batteries, along with
sulfate, phosphoric acid and a lithium source are reacted and the mixture is dried. the labelling and taping of terminals to prevent contact between terminals during
Supplementary Table 15 lists the material inputs for LFP fabrication. shipping8. We used the same GHG and energy consumption per tonne mile from
The pyrometallurgical process, as outlined in a Umicore patent39, involves road transportation as for the shipping of raw materials (listed in Supplementary
firing the battery materials with slag, limestone, sand and coke to produce a Table 20), and conducted the same parametric analysis of doubling both the
metal alloy, which can be separated into its constituent materials, and a slag that distance travelled and the GHG emissions and energy consumed per tonne mile.
can be repurposed to make cement. Both hydrometallurgical and direct cathode
recycling processes begin with a discharging and disassembly step, where external Recycling process emissions, energy consumption and offsets. The
cell hardware can be removed and recycled separately. For hydrometallurgical pyrometallurgical process is based on the process detailed in US Patent 7,169,206
recycling, the cathode material is then crushed to remove the current collector. B2 (ref. 39), where batteries are mixed with other input materials (listed in
The cathode material is then filtered and calcined, and the remaining powder is Supplementary Table 15) and heated to produce a metal alloy and slag material.
leached to produce transition metal and lithium solutions. Direct cathode recycling Process and heat inputs (listed in Supplementary Table 16) are fixed regardless of
involves extracting the electrolyte using liquid or supercritical CO240–43, then the specific cell chemistry or format. Estimates of the energy consumed for the
reducing the size of the recovered components and separating out the cathode leaching processes to separate the metal slag are drawn from Dunn et al.44. We
materials16,40,44. For hydrometallurgical and direct recycling, we assume that also calculated the stoichiometric CO2 emissions produced by the combustion of
the materials not recovered through the recycling process, or collected as scrap the graphitic carbon anode, polyvinylidene difluoride (PVDF) electrode binder
(current collectors and cell hardware) are incinerated. (PVDF), electrolyte and other plastics in the cells as part of the pyrometallurgical
For each process considered, we present the resulting emissions in CO2e using process. The GHG emissions from the combustion of these materials were also
20-year warming potentials and energy consumption in MJ (https://greet.es.anl. included in the GHG emissions totals for hydrometallurgical and direct cathode
gov/; see also ref. 45). We also use two different functional units: kg of battery recycling, as these materials would need to be disposed of in some way if not
manufactured (including cell electrode materials, separators and packaging) and directly recovered.
kWh of battery manufactured (including all cell components). We include an Hydrometallurgical recycling, as outlined by Dunn et al.44, consists of five
analysis across both functional units for relevance in different sectors. Most LCA steps: discharge and disassembly; crushing of the cathode material to remove
studies of battery manufacturing consider the emissions or energy consumed per the current collector; filtration and calcination of the cathode active material;
kg, while in the energy storage and electric vehicle markets, the storage capacity in grinding; and leaching to produce solutions of lithium and transition metals. The
kWh is the most relevant parameter when considering the size of a battery. electricity consumption assumption for the discharge and disassembly step of
0.034 MMBtu tonne of battery−1 comes from the Dunn et al44. assumption for the
LCA assumptions. The energy inputs and carbon emissions for the cell material total electricity required during the formation cycle of battery cells. For the cathode
inputs were modelled using GREET 2016 (with a database update on 19 January crushing and soaking, we assume that NMP is recovered and reused. As is the case
2017; https://greet.es.anl.gov/), often including a range of assumptions depending in battery manufacturing, recovery of the solvent is necessary to minimize the
on the recycled content of the source materials. Supplementary Tables 17 and 18 manufacturing costs47. Process heat is provided by a natural gas boiler with 80%
list the specific assumptions and distributions used to model these contributions to efficiency. At a temperature change of 75 K, the heat capacities (Cp) of N-methyl-2-
the overall GHG emissions and energy resources for cell manufacturing. For inputs pyrrolidone (NMP), the cathode material and the aluminium current collector are

154 Nature Sustainability | VOL 2 | FEBRUARY 2019 | 148–156 | www.nature.com/natsustain

NaTuRe SusTainabiliTy Analysis
2.1, 1.0 and 0.9 J gK−1, respectively, and the mass of NMP is assumed to be equal to recycling and any transportation costs for battery collection. Transportation could
the mass of the current collector. The electricity and energy consumption values be particularly expensive because of the Department of Transportation regulations
for the filtration and calcination step are listed in Supplementary Table 25, and we on shipping class 9 hazardous materials22.
assume that the PVDF binder and electrolyte are combusted during this process. Constructing a process-based cost model requires determination of the number
Electricity consumption data for the grinding and leaching steps are also provided of manufacturing lines required for each step in a manufacturing process, to satisfy
in Supplementary Table 25. The leaching steps also require solvents and binding a specified annual production volume. The number of lines required is calculated
agents. We assume that the H2O2 added is equal to 2.5 times the mass of lithium to by dividing the annual production volume (Vannual) by the product of the batch size
be recovered from the cathode; 15 g of citric acid is added per kg of lithium to be (xbatch), cycle time (tcycle) and amount of time the process runs per year (which is
recovered; and the sodium carbonate is stoichiometrically balanced to precipitate a function of the days per year (ndays), shifts worked per day (nshifts) and length of
lithium carbonate (0.5 moles NaCO2: 1 mole lithium)44. The GHG emissions and a shift (xhours) less paid (xpb) and unpaid breaks (xub), adjusted for any unplanned
embodied energy assumptions for these additives were sourced from EcoInvent48 downtime (xdown)), as shown in equation (1). The annual production volume must
and GREET 2016, and are listed in Supplementary Table 27. The transportation also be increased to account for any lost yield (γloss) during the manufacturing step.
distance assumptions are listed in Supplementary Table 21. The lines required are rounded up to the nearest integer to ensure that annual
The direct cathode recycling process consists of four main steps: (1) discharge production demands are met. Once the number of production lines required for
and disassembly; (2) electrolyte extraction; (3) size reduction; and (4) final each step is established, the capital equipment, factory area required and hours
separation. We assumed that the energy consumed during the disassembly and of labour are calculated. Both building and capital expenses are annualized.
discharge was the same as during this step in a hydrometallurgical process. The Assumptions about equipment, labour, material and energy costs are detailed in
primary driver of electricity consumption for the electrolyte extraction step is Supplementary Tables 20–21, and a spreadsheet detailing the calculations for the
the energy used to compress the CO2 used as a solvent. Recent research40–43 has cathode materials considered is included in the Supplementary Information.
focused on the optimum combination of CO2, duration of the extraction process
and additional solvents, but we use the assumptions from Grutzke et al.40. They use  Vannual 
liquid CO2 at a flow rate of 15 ± 3 l min−1 for 50 min, along with a 3:1 acetonitrile ylr =  
 (1)
 xbatchtcyclendaysnshifts(xhours−x pb−x ub)(1−xdown)(1−γloss) 
(ACN):propylene carbonate mixture at a rate of 0.5 l min−1 for 20 min. We use
a triangle distribution to estimate the mass of CO2 consumed based on these
uncertainty bounds and then scale this by the cell volume treated. To estimate the
energy consumed to produce the liquid CO2, we also use a triangle distribution, Code availability
with a peak of 0.0981 kWh kg−1 (ref. 40), and bounds of 0.0719 kWh kg−1 (from Sample MATLAB code for the recycling model described here is available at https://
Zahid et al.43) and 0.1609 kWh kg−1 (from Dunn et al.44). For both the size reduction github.com/rciez2125/batteryRecycling. Sample calculations for the process-
and final separation steps, we use the estimates provided by Dunn et al.44, but with based cost model of cathode manufacturing are available in the Supplementary
triangular distributions to represent the uncertainty to the functional unit. We use Information files.
a triangular distribution to divide the 0.22 MMBtu for the size reduction step by
the total cell mass, the cell mass that is present at that step and the mass of cathode
material. We use the same functional units to divide the 0.02 MMBtu assumed Data availability
for the final separation step. The GHG emissions and energy associated with the The authors declare that the data used as model inputs supporting the findings of
other solvents (ACN and propylene carbonate) were based on information from this study are available within the paper and its Supplementary Information files.
EcoInvent49 and GREET 2016, respectively. Details are provided in Supplementary Additional questions about the data supporting the findings of this study can be
Tables 26 and 27. Shipping distance assumptions for the ACN and propylene directed to the corresponding author.
carbonate solvents are listed in Supplementary Table 21.
To determine the GHG emissions reductions from recycling, we calculate the
GHG emissions offset by the final products. For pyrometallurgical recycling, this Received: 24 February 2018; Accepted: 7 January 2019;
includes copper, iron, nickel, cobalt and cement slag. The total amount of each Published online: 11 February 2019
material present in the smelted material is calculated based on cell inputs, and the
amount of each metal recovered is calculated based on the yield rates from the
process patent (listed in Supplementary Table 16). The emissions offsets from the References
recovered metals are listed in Supplementary Table 17. For hydrometallurgical 1. Global EV Outlook 2017: Two Million and Counting (International Energy
recycling, this includes the lithium carbonate recovered, along with the other Agency, Clean Energy Ministerial & Electric Vehicles Initiative, 2017).
transition metals (nickel, cobalt, manganese and iron) depending on cell 2. Ayre, J. GM aiming for 500,000 ‘new energy vehicle’ sales per year by 2025.
chemistry. For the direct cathode recycling process, the emissions and energy Clean Technica https://cleantechnica.com/2017/07/10/gm-aiming-500000-
offsets include the sum of the embodied energy and emissions of cell hardware new-energy-vehicle-sales-per-year-2025/ (2017).
recycled (cell canisters and current collectors), the cathode precursor materials, 3. Randall, T. Here’s how Elon Musk gets Tesla to 500,000 cars a year by 2020.
the transportation of these materials and the energy consumed during the cathode Bloomberg https://www.bloomberg.com/news/articles/2015-08-10/here-s-how-
precursor drying step. The offsets are then scaled based on the yield rate of the elon-musk-takes-tesla-to-500-000-cars-in-five-years (2015).
cathode recovery process. 4. Field, K. Yes, Tesla recycles all of its spent batteries & wants to do more in
the future. Clean Technica https://cleantechnica.com/2018/06/07/yes-tesla-
Cost model. We use a process-based cost model to estimate the cost to produce recycles-all-of-its-spent-batteries-wants-to-do-more-in-the-future/ (2018).
cathode materials from raw inputs. This modelling technique originated in the 5. Stringer, D. & Ma, J. Where 3 million electric vehicle batteries will go when
electronics industry, where it was used to inform design decisions by estimating they retire. Bloomberg https://www.bloomberg.com/news/features/2018-06-27/
manufacturing costs50. The method has since been adopted for product design where-3-million-electric-vehicle-batteries-will-go-when-they-retire (2018).
more broadly51,52 and has been used to examine battery manufacturing costs19,53. 6. Directive 2006/66/EC of the European Parliament and of the Council of 6
Process-based cost models disaggregate the many factors that come together September 2006 on batteries and accumulators and waste batteries and
during a manufacturing process into discrete units that increase with production accumulators and repealing Directive 91/157/EEC http://data.europa.eu/eli/
volumes. Labour, capital equipment and input materials are allocated for each step dir/2006/66/oj (European Parliament & Council of the European
in the manufacturing process, with the possibility of sharing resources between Union, 2006)
steps where appropriate. Production volumes at each step are adjusted based on 7. Ritchie, M. & Wu, D. China EV battery boom set to turn ugly as subsidies
yield rates of the downstream processes. This type of bottom-up modelling makes revised. Bloomberg https://www.bloomberg.com/news/articles/2018-07-18/
it possible to make more granular comparisons between process technologies, china-ev-battery-boom-set-to-turn-ugly-after-subsidies-revised (2018).
determine the production volumes required to reach economies of scale and 8. Richa, K., Babbitt, C. W. & Gaustad, G. Eco‐efficiency analysis of a
identify the inputs and processes that contribute the most to product costs. lithium‐ion battery waste hierarchy inspired by circular economy. J. Ind. Ecol.
These characteristics mean that process-based cost modelling can provide more 21, 715–730 (2017).
substantive reasons for why technology costs are falling, and because process 9. Xu, J. et al. A review of processes and technologies for the recycling of
models can be adapted for technologies that do not currently exist, it offers a lithium-ion secondary batteries. J. Power Sources 177, 512–527 (2008).
distinct advantage other some other approaches in some cases30. 10. Hanisch, C., Haselrieder, W. & Kwade, A. in Glocalized Solutions for
We leverage knowledge from previous work examining battery manufacturing Sustainability in Manufacturing (eds Hesselbach, J. & Herrmann, C.) 85–89
costs to focus on the manufacturing costs of the cathodes from raw materials, (Springer, Berlin & Heidelberg, 2011).
and contrast this with the costs associated with the cathode materials from a 11. Wang, R.-C., Lin, Y.-C. & Wu, S.-H. A novel recovery process of metal values
direct cathode recycling process. The reprocessing of the cathode materials from from the cathode active materials of the lithium-ion secondary batteries.
the direct cathode recycling process is essentially the same as the lithiation step Hydrometallurgy 99, 194–201 (2009).
of cathode manufacturing, so we can isolate the breakeven cost of recovering 12. Shin, S. M., Kim, N. H., Sohn, J.-S., Yang, D. H. & Kim, Y. H. Development
the cathode material. Without another source of income (for example, a fee for of a metal recovery process from Li-ion battery wastes. Hydrometallurgy 79,
recycling service) this breakeven cost would need to include both the costs of direct 172–181 (2005).

Nature Sustainability | VOL 2 | FEBRUARY 2019 | 148–156 | www.nature.com/natsustain 155

Analysis NaTuRe SusTainabiliTy
13. Swain, B., Jeong, J., Lee, J.-C., Lee, G.-H. & Sohn, J.-S. Hydrometallurgical 42. Mönnighoff, X. et al. Supercritical carbon dioxide extraction of electrolyte
process for recovery of cobalt from waste cathodic active material generated from spent lithium ion batteries and its characterization by gas
during manufacturing of lithium ion batteries. J. Power Sources 167, chromatography with chemical ionization. J. Power Sources 352,
536–544 (2007). 56–63 (2017).
14. Lee, C. K. & Rhee, K.-I. Reductive leaching of cathodic active materials from 43. Zahid, U., An, J., Lee, U., Choi, S. P. & Han, C. Techno-economic assessment
lithium ion battery wastes. Hydrometallurgy 68, 5–10 (2003). of CO2 liquefaction for ship transportation. Greenh. Gases 4, 734–749 (2014).
15. Contestabile, M., Panero, S. & Scrosati, B. A laboratory-scale lithium-ion 44. Dunn, J. B., Gaines, L., Barnes, M., Sullivan, J. L. & Wang, M. Material and
battery recycling process. J. Power Sources 92, 65–69 (2001). Energy Flows in the Materials Production, Assembly, and End-of-Life Stages of
16. Sloop, S. Giga Life Cycle: Manufacture of Cells from Recycled EV Li-ion the Automotive Lithium-Ion Battery Life Cycle (Argonne National Laboratory,
Batteries (OnTo Technology, 2015). Argonne, 2014).
17. Dunn, J. B., Gaines, L., Sullivan, J. & Wang, M. Q. Impact of recycling on 45. IPCC. Climate Change 2007: The Physical Science Basis (eds Solomon, S.
cradle-to-gate energy consumption and greenhouse gas emissions of et al.) (Cambridge Univ. Press, Cambridge, 2007).
automotive lithium-ion batteries. Environ. Sci. Technol. 46, 46. Emissions & Generation Resource Integrated Database (eGRID) https://www.
12704–12710 (2012). epa.gov/energy/egrid (United States Environmental Protection Agency,
18. Schipper, F. et al. Review—recent advances and remaining challenges for accessed 10 June 2016).
lithium ion battery cathodes. J. Electrochem. Soc. 164, A6220–A6228 (2016). 47. Nelson, P. A., Gallagher, K. G., Bloom, I. & Dees, D. W. Modeling the
19. Ciez, R. E. & Whitacre, J. F. Comparison between cylindrical and prismatic Performance and Cost of Lithium-Ion Batteries for Electric-Drive Vehicles
lithium-ion cell costs using a process based cost model. J. Power Sources 340, (Argonne National Laboratory, Argonne, 2012).
273–281 (2017). 48. Althaus, H. J. et al. in Life Cycle Inventories of Chemicals (eds Althaus, H.-J. et al.)
20. Ellingsen, L. A.-W., Majeau-Bettez, G. & Strømman, A. H. Comment on “The 1–10 (Swiss Centre for Lifecycle Inventories, Dübendorf, 2007).
significance of Li-ion batteries in electric vehicle life-cycle energy and 49. Sutter, J. in Life Cycle Inventories of Petrochemical Solvents 1–10 (Swiss Centre
emissions and recycling’s role in its reduction” in Energy & Environmental for Lifecycle Inventories, Dübendorf, 2007).
Science. J. Ind. Ecol. 19, 518–519 (2015). 50. Ong, N. S. Manufacturing cost estimation for PCB assembly: an activity-
21. Placke, T., Kloepsch, R., Dühnen, S. & Winter, M. Lithium ion, lithium metal, based approach. Int. J. Prod. Econ. 38, 159–172 (1995).
and alternative rechargeable battery technologies: the odyssey for high energy 51. La Trobe-Bateman, J. & Wild, D. Design for manufacturing: use of a
density. J. Solid State Electrochem. 21, 1939–1964 (2017). spreadsheet model of manufacturability to optimize product design and
22. Hazardous Materials; Transportation of Lithium Batteries https://www. development. Res. Eng. Des. 14, 107–117 (2003).
federalregister.gov/documents/2014/08/06/2014-18146/hazardous-materials- 52. Bloch, C. & Ranganathan, R. Process-based cost modeling. IEEE Trans.
transportation-of-lithium-batteries (Federal Register, 2014). Compon. Hybrids Manuf. Technol. 15, 288–294 (1992).
23. Els, F. CHARTS: giant gap between future lithium supply, demand. Mining. 53. Sakti, A., Michalek, J. J., Fuchs, E. R. H. & Whitacre, J. F. A techno-economic
com http://www.mining.com/charts-chasm-between-future-lithium-supply- analysis and optimization of Li-ion batteries for light-duty passenger vehicle
demand-28137/ (2014). electrification. J. Power Sources 273, 966–980 (2014).
24. Idemoto, Y. & Matsui, T. Thermodynamic stability, crystal structure, and 54. Notter, D. A. et al. Contribution of Li-ion batteries to the environmental
cathodic performance of Lix(Mn1/3Co1/3Ni1/3)O2 depend on the synthetic impact of electric vehicles. Environ. Sci. Technol. 44, 6550–6556 (2010).
process and Li content. Solid State Ionics 179, 625–635 (2008). 55. Zackrisson, M., Avellán, L. & Orlenius, J. Life cycle assessment of lithium-ion
25. Wild, F., Riseborough, J. & Wilson, T. Glencore buys out billionaire with $1 batteries for plug-in hybrid electric vehicles—critical issues. J. Clean. Prod.
billion Congo mining deal. Bloomberg https://www.bloomberg.com/news/ 18, 1519–1529 (2010).
articles/2017-02-13/glencore-said-to-agree-on-gertler-buyout-in-960-million- 56. Majeau-Bettez, G., Hawkins, T. R. & Strømman, A. H. Life cycle
deal (2017). environmental assessment of lithium-ion and nickel metal hydride batteries
26. Advancing Sustainable Materials Management: 2013 Fact Sheet https://www. for plug-in hybrid and battery electric vehicles. Environ. Sci. Technol. 45,
epa.gov/sites/production/files/2015-09/documents/2013_advncng_smm_fs.pdf 4548–4554 (2011).
(United States Environmental Protection Agency, 2015). 57. Lithium-Ion Batteries and Nanotechnology for Electric Vehicles: A Life Cycle
27. Gaines, L. The future of automotive lithium-ion battery recycling: charting a Assessment https://www.epa.gov/sites/production/files/2014-01/documents/
sustainable course. Sustain. Mater. Technol. 1–2, 2–7 (2014). lithium_batteries_lca.pdf (United States Environmental Protection
28. Recycling and Reuse: Batteries and Accumulators: European Union Directive Agency, 2012).
https://archive.epa.gov/oswer/international/web/pdf/200806_batteries_eu_ 58. Ellingsen, L. A.-W. et al. Life cycle assessment of a lithium-ion battery vehicle
directive.pdf (United States Environmental Protection Agency, 2008). pack. J. Ind. Ecol. 18, 113–124 (2013).
29. Chaurand, P. et al. Environmental impacts of steel slag reused in road 59. Kim, H. C. et al. Cradle-to-gate emissions from a commercial electric
construction: a crystallographic and molecular (XANES) approach. J. Hazard. vehicle Li-ion battery: a comparative analysis. Environ. Sci. Technol. 50,
Mater. 139, 537–542 (2007). 7715–7722 (2016).
30. Sullivan, J. L. & Gaines, L. A Review of Battery Life-Cycle Analysis: State of
Knowledge and Critical Needs (Argonne National Laboratory, Argonne, 2010).
31. Elshkaki, A., Graedel, T. E., Ciacci, L. & Reck, B. K. Copper demand, supply, Acknowledgements
The authors thank C. Samaras, M. Mauter and J. Michalek for discussions on life-
and associated energy use to 2050. Glob. Environ. Change 39, 305–315 (2016).
cycle analysis and the framing of the research findings. This material is based on work
32. Ciacci, L., Harper, E. M., Nassar, N. T., Reck, B. K. & Graedel, T. E. Metal
supported by the National Science Foundation Graduate Research Fellowship under
dissipation and inefficient recycling intensify climate forcing. Environ. Sci.
grant number DGE 1252522. Any opinions, findings, conclusions and recommendations
Technol. 50, 11394–11402 (2016).
expressed in this material are those of the authors(s) and do not necessarily reflect the
33. Mineral Commodity Summaries 2018 https://minerals.usgs.gov/minerals/pubs/
views of the National Science Foundation.
mcs/2018/mcs2018.pdf (US Department of the Interior & US Geological
Survey, 2018).
34. Banza Lubaba Nkulu, C. et al. Sustainability of artisanal mining of cobalt in Author contributions
DR Congo. Nat. Sustain. 1, 495–504 (2018). R.E.C. designed the research with input from J.F.W. R.E.C. conducted the majority of the
35. Olivetti, E. A., Ceder, G., Gaustad, G. G. & Fu, X. Lithium-ion battery supply analysis and wrote most of the paper. J.F.W. made significant contributions to the analysis
chain considerations: analysis of potential bottlenecks in critical metals. Joule and editing of the paper.
1, 229–243 (2017).
36. Habib, K., Hamelin, L. & Wenzel, H. A dynamic perspective of the
geopolitical supply risk of metals. J. Clean. Prod. 133, 850–858 (2016). Competing interests
37. Nakura, K., Ariyoshi, K., Yoshizawa, H. & Ohzuku, T. Characterization of The authors declare no competing interests.
lithium insertion electrodes and its verification: prototype 18650 batteries
consisting of LTO and LAMO. J. Electrochem. Soc. 162, A622–A628 (2015).
38. Wang, Z., Benavides, P., Dunn, J. B. & Cronauer, D. C. Development of Additional information
GREET Catalyst Module (Argonne National Laboratory, 2015). Supplementary information is available for this paper at https://doi.org/10.1038/
39. Cheret, D. & Santen, S. Battery recycling. US patent 7,169,206 B2 (2007). s41893-019-0222-5.
40. Grützke, M. et al. Extraction of lithium-ion battery electrolytes with liquid Reprints and permissions information is available at www.nature.com/reprints.
and supercritical carbon dioxide and additional solvents. RSC Adv. 5, Correspondence and requests for materials should be addressed to J.F.W.
43209–43217 (2015).
41. Nowak, S. & Winter, M. The role of sub- and supercritical CO2 as “processing Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
solvent“ for the recycling and sample preparation of lithium ion battery published maps and institutional affiliations.
electrolytes. Molecules 22, 403 (2017). © The Author(s), under exclusive licence to Springer Nature Limited 2019

156 Nature Sustainability | VOL 2 | FEBRUARY 2019 | 148–156 | www.nature.com/natsustain