See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/346461477

Large-scale automotive battery cell manufacturing: Analyzing strategic and

operational effects on manufacturing costs

Article in International Journal of Production Economics · November 2020

DOI: 10.1016/j.ijpe.2020.107982

CITATIONS READS

104 7,157

5 authors, including:

Fabian Duffner Lukas Mauler

University of Münster University of Münster
9 PUBLICATIONS 1,213 CITATIONS 6 PUBLICATIONS 439 CITATIONS

SEE PROFILE SEE PROFILE

Marc Wentker Jens Leker

University of Münster University of Münster
8 PUBLICATIONS 563 CITATIONS 156 PUBLICATIONS 4,664 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Fabian Duffner on 29 November 2020.

The user has requested enhancement of the downloaded file.

 Int. J. Production Economics 232 (2021) 107982

Contents lists available at ScienceDirect

International Journal of Production Economics

journal homepage: http://www.elsevier.com/locate/ijpe

Large-scale automotive battery cell manufacturing: Analyzing strategic and

operational effects on manufacturing costs
Fabian Duffner a, b, *, Lukas Mauler a, b, Marc Wentker a, Jens Leker a, c, Martin Winter c, d
a
Institute of Business Administration at the Department of Chemistry and Pharmacy (IfbM), University of Münster, 48148, Münster, Germany
b
Porsche Consulting GmbH, 74321, Bietigheim-Bissingen, Germany
c
Helmholtz-Institute Münster (HIMS), 48149, Münster, Germany
d
MEET Battery Research Center, Institute of Physical Chemistry University of Münster, 48149, Münster, Germany

A R T I C L E I N F O A B S T R A C T

Keywords: Cost-efficient battery cell manufacturing is a topic of intense discussion in both industry and academia, as battery
Battery costs are crucial for the market success of electrical vehicles (EVs). Based on forecasted EV growth rates, battery
Automotive cell manufacturers are investing billions of dollars in new battery cell plants. Whether these billion-dollar in­
Manufacturing
vestments are economically viable depends on the materialization of forecasted EV growth rates and company-
Cost
Optimization
specific competitive market positions. For both, cost-efficient battery cell manufacturing is key for success. To
Process-based cost modelling ensure cost-efficient battery cell manufacturing, transparency is necessary regarding overall manufacturing costs,
their cost drivers, and the monetary value of potential cost reductions. Driven by these requirements, a cost
model for a large-scale battery cell factory is developed. The model relies on the process-based cost modelling
technique (PBCM) and includes more than 250 parameters. Based on this cost model, directions are provided,
how minimum costs can be achieved reflecting current and future state of technology. Further, it is outlined
which process steps and cost elements have the greatest impact on total cost and should thus be focused within
future cost reduction activities.

1. Introduction Gigafactory) have made battery-cell cost optimization relevant for both
science and industry. Triggered by this, some optimizations have
The battery manufacturing industry is forecast to be one of the fastest already been achieved, mainly based on new materials and innovative
growing production industries through 2030. Especially driven by the cell chemistries (Liu et al., 2010; Placke et al., 2017; Scrosati and Gar­
expanded production of electrical vehicles (EVs) with the overall goal of che, 2010; Wagner et al., 2013; Winter et al., 2018). Although it has
minimizing vehicular CO2 and NO2 emissions, annual global lithium-ion received less attention, battery cell manufacturing has also improved,
battery capacity demand is expected to increase from 160 GWh cell with notable results in the last two decades. However, due to numerous
energy in 2018 to >1000 GWh cell energy in 2030. By 2030, this in­ consecutive process steps, the interaction of these steps and the high
crease will have triggered investments of more than $100 billion and number of individual process parameters, it can be assumed that there is
generate annual revenue of more than $100 billion within battery potential for further optimization. The realization of this potential will
manufacturing industry (Avicenne Energy, 2019). To transform these require a deep understanding of the individual production process steps,
investments into sustainable business models, cost efficient battery process parameters, and their impact on cost (Kwade et al., 2018).
manufacturing is the key factor as it is the prerequisite to make EVs With regard to costs, cost models with extended capabilities to
competitive compared to internal combustion engine vehicles (Pollet analyze cost drivers and to simulate a large number of parameters can be
et al., 2012; Sierzchula et al., 2014; Wu et al., 2015). the key for success to achieve the necessary transparency (Qian and
The need to produce cost-efficient batteries, the launch of the first Ben-arieh, 2008). Driven by this requirement, numerous studies have
mass-market EVs (e.g. Tesla Model 3), and initial investments worth dealt with the modelling of battery costs (Berg et al., 2015; Nelson et al.,
several billion dollars for the first battery-cell factories (e.g. Tesla’s 2019; Petri et al., 2015; Schmuch et al., 2018; Vaalma et al., 2018;

* Corresponding author. Institute of Business Administration at the Department of Chemistry and Pharmacy (IfbM), University of Münster, 48148, Münster,
Germany.
E-mail address: [email protected] (F. Duffner).

https://doi.org/10.1016/j.ijpe.2020.107982
Received 30 March 2020; Received in revised form 11 August 2020; Accepted 3 November 2020
Available online 11 November 2020
0925-5273/© 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

Fig. 1. Process steps for the manufacture of a lithium-ion pouch battery cell in a large-scale factory.

Wentker et al., 2019). However, most of these models focus on material extensive parameter foundation, the presented model and the associated
using surcharge rates to forecast processing costs (Duffner et al., 2020b). model architecture provides an optimal starting point to translate
While such approaches are suitable to forecast rough costs for a new further battery innovations into costs. This is, especially in the highly
product with less effort, their capabilities to translate new innovations dynamic battery environment, a valuable capability as it provides the
comprehensively into costs are limited (Niazi et al., 2006). basis for cost-optimal and data-driven decision making. Summarizing,
Addressing these limitations, within this paper, a model is presented the results of this paper contribute to evaluate the technological po­
that can translate the academic discussion related to process in­ tential of lithium-ion batteries and support the materialization of this
novations, material & design-innovations, and location alternatives into potential which is relevant for both, science and industry.
costs. Therefore, first, cost estimation techniques are reviewed to eval­
uate suitability for the presented requirements. Finding that bottom-up 2. Background
techniques and especially the process-based cost modelling technique
fits best, a model for battery manufacturing relying on more than 250 2.1. Battery design and manufacturing
parameters is proposed. Based on this model, cost driver analysis within
process steps, cost elements and parameter categories is provided. Automotive traction battery systems consist of battery modules and
Further, a current state and a future cost level is introduced by trans­ battery cells that are connected and controlled by a battery management
lating the associated parameter sets from literature into costs. system. The cells are a crucial component as they significantly influence
The main innovations of this study are as follows: First, a current and the performance and cost of the whole system (Kwasi-Effah and Rabc­
future cost level is presented that is derived by linking an established zuk, 2018; Nelson et al., 2019). The major constituents of a lithium-ion
cost estimation technique (PBCM) with the current battery specific battery cell, which is currently the state-of-the-art technology (Aalder­
discussion related to process optimization, material & design optimi­ ing et al., 2019), are the cathode (positive electrode) (Arinicheva et al.,
zation and location alternatives. This analytical approach provides the 2020; Whittingham, 2004) and the anode (negative electrode) (Andre
most reliable and up to date basis to evaluate the cost potential of et al., 2017) as well as the separator and the electrolyte. Lithium tran­
lithium-ion batteries. Second, by presenting the most comprehensive sition metal oxides (LiMO2) with transition metals (M) such as nickel,
cost driver analysis within lithium-ion battery manufacturing (scenario- cobalt, and manganese (NMC) are the most widely used class of positive
related parameters, process steps and cost elements) guidance is pro­ active materials (Andre et al., 2015; Myung et al., 2017; Schmuch et al.,
vided to set focus in future cost optimization activities. Third, due to its 2018). Carbonaceous materials, in particular synthetic and natural

2
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

Fig. 2. Classification of cost estimation techniques including key advantages, limitations and examples for techniques.

graphite, as well as amorphous carbons are mostly used as negative g. NMP), and conductive additives (e.g. carbon), are batch-wise mixed in
active material (Blomgren, 2017; Schmuch et al., 2018; Scrosati et al., a planetary mixer for several hours to produce a cathode slurry. The
2015). As electrolyte solvents, a mixture of cyclic and linear organic target of this process step is to achieve the desired homogeneity and
carbonates with lithium hexafluorophosphate as conductive salt (typi­ viscosity of the slurry ensuring the electrode’s subsequent electro­
cally ~1 mol L− 1) are used (Schmuch et al., 2018; Xu, 2014). The used chemical performance and adhesion to the current collector foil (Dreger
microporous separator is typically based on polyolefins such as poly­ et al., 2015). In current cathode chemistries, this adhesion is provided by
propylene (Lee et al., 2014). For current collectors, thin sheets of PVdF binders whose processing relies on the toxic and teratogen solvent
aluminium and copper are used for the cathode and the anode, respec­ NMP. For anode production, NMP has already been replaced by water
tively. Conductive electrode additives are usually carbon-based. With which could also be implemented in cathode processing. This reduces
regard to battery cell production, numerous consecutive process steps the drying effort due to a lower boiling point (water 100 ◦ C vs. NMP
are required. The manufacturing processes outlined in the following 203 ◦ C), eliminates the complex and costly solvent recovery process,
represent large-scale production of a lithium-ion pouch cell, as pre­ decreases associated material cost (water < $0,02 L− 1 vs. NMP $1-3 L− 1)
sented in Kwade et al. (2018). This production can be divided into three and results in more sustainable manufacturing (lower CO2 emissions,
value-adding superordinate main processes: electrode production, cell less toxic materials) (Bresser et al., 2018a). The Mixing process is fol­
production, and cell conditioning. Other processes are also necessary to lowed by the four continuous and interconnected process steps Coating,
support the execution of these three value-adding processes (e.g. Drying, Calendering and Slitting, which can be summarized as
inter-process material handling). Fig. 1 shows the individual process roll-to-roll processes. These four adjacent processes should be linked and
steps. hence share the roll-to-roll working speed as an essential process
In electrode production, anodes and cathodes are produced. Anode parameter. State-of-the-art electrode processing has already reached
and cathode production are spatially separated, but basically the same working speeds of 25–50 m min− 1 (Kwade et al., 2018). Within the
process steps are followed. For the sake of simplification, the process Coating process step, thin metal carrier foils (e.g. aluminum) are coated
steps for cathode production only are described here. First, the cathode on both sides with the active material slurry and directly dried to solidify
components, namely the active material, typically NMC622 in state-of- the slurry by evaporating the solvent. As the working width of the coater
the-art (Schmuch et al., 2018), polymer binder (e.g. PVdF), solvent (e. (up to 1500 mm) exceeds the width of common single electrodes, the

Fig. 3. Battery-cell-specific process-based cost modelling (PBCM) framework.

3
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

Table 1 (Ciez and Whitacre, 2017).

Product characteristics of battery cells considered for Base and Optimistic To enable electrode production, cell production and cell condition­
Scenario. ing additional supporting processes are necessary: ensuring a clean at­
Cell Design Specifications Base Optimistic Unit mosphere required for cell production through the use of a dry room,
Energy 193 211 Wh cell− 1
recycling the NMP solvent used for cathode production, ensuring proper
Energy density 265 296 Wh kg− 1 material handling through all process steps, and managing the receipt of
Format Pouch purchased materials and the shipping of finished cells (Nelson et al.,
Chemistry NMC622 ||G NMC811 ||G 2012).
Number of two-sided electrodes
1
Cathode 30 20 pieces cell−
Anode 30 20 pieces cell− 1 2.2. Cost estimation and battery cost modeling
Coating Thickness
Cathode 65 100 μm Cost estimation involves the forecasting of costs of activities that
Anode 74 120 μm have not yet been carried out at the time the estimation is conducted
Width of electrodes*
Cathode 93 mm
(H’midaa et al., 2006; Shehab and Abdalla, 2002). Over the last decades,
Anode 95 mm a variety of different cost-estimation techniques have been developed
Length of electrodes* that researchers categorize using various criteria (Cavalieri et al., 2004;
Cathode 294 mm Hueber et al., 2016; Niazi et al., 2006; Qian and Ben-Arieh, 2008b;
Anode 296 mm
Shehab and Abdalla, 2001, 2002; Zhang et al., 1996). Referencing the
* based on ISO cell format Pouch VIFB-/99/300
approach published by Hueber et al. (2016), the techniques can be
categorized as intuitive, analogical, parametric or bottom-up.
electrodes are cut to the desired width after Calendering by roll knives. Intuitive techniques rely on the estimator’s knowledge and experi­
In the Coating and Calendering processes the targeted electrode thick­ ence. Within these techniques, rough, somehow subjective and arbitrary
ness is manufactured, which strongly effects cell properties. Current results can be generated without greater effort but they are neither
cathode thicknesses of high-energy cells range from 65 to 80 μm comprehensible nor repeatable for a third person (Niazi et al., 2006).
(Schmuch et al., 2018). Increasing electrode thickness results in higher Analogical techniques link historical cost data to a new product using
energy densities but has a decreasing effect on rate capability and regression models or neural network approaches. These techniques rely
therefore power density, which also applies vice versa (Zheng et al., on the assumption that similar products have similar costs (Curran et al.,
2012). Further, electrode thickness has an impact on cost, as thicker 2004). Parametric cost estimation techniques use product-specific cost
electrodes result in a higher share of active material in the cell and functions to predict costs. The cost functions can consist of several pa­
hence, lower material cost for non-active components can be achieved rameters or variables, like part weight or part size (International Society
(Patry et al., 2015). Finally, within the last process step of electrode of Parametric Analysts, 2018). Within the bottom-up techniques a
production, the electrodes are dried under vacuum to further reduce product is first broken down into its individual components. The com­
moisture before they are transferred to a dry room (Kaiser et al., 2014; ponents are then broken down into resources and processes required to
Nelson et al., 2019). produce them, with costs assigned to each (Ben-Arieh and Qian, 2003).
To produce battery cells in a z-folded format, which is a state-of-the- Fig. 2 presents the categories and summarizes the key advantages and
art electrode stacking order for lithium-ion pouch cells produced in large limitations per category. Further research on intuitive techniques can be
scale, the anodes and cathodes are first cut into single sheets. Second, found in Ficko et al. (2005), Rush and Roy (2001), Sakti et al. (2017),
the separator is fed as an endless folded band, and the anodes and Shehab and Abdalla (2002), on analogous techniques in Cavalieri et al.
cathodes are alternately inserted into the interstitial space. This is a (2004), Hagen et al. (2015), Verlinden et al. (2008), Wang (2007), on
highly automated process which is currently conducted at an operating parametric techniques in Duverlie and Castelain (1999), Nelson et al.
speed of 60 sheets min− 1 (Mooy, 2019). Third, within Contacting, in­ (2019), Patry et al. (2015) and on bottom-up techniques in Ciez and
ternal contacts between the anode, cathode, and separator assembly are Whitacre (2017), Cooper and Kaplan (1988), Ficko et al. (2005), Field
created by welding. Subsequently, the assembly is inserted into the et al. (2007), Schulze et al. (2012).
housing (e.g. pouch). After insertion, the cell is filled with electrolyte (e. Relying on these techniques and driven by the relevance of battery
g. LiPF6) and closed (Kwade et al., 2018; Tagawa and Brodd, 2009). costs, various battery cost models have been developed within the last
Cell conditioning begins during the Formation and Final sealing years. Most of these models focus on the calculation of material costs
process step. During Formation, the cell is charged for the first time, (Berg et al., 2015; Petri et al., 2015; Schmuch et al., 2018; Wentker et al.,
followed by discharge and further charging cycles at different charging 2019). A detailed description and analysis of these models can be found
rates. This procedure currently takes several days (Wood et al., 2015). in the review article published by Duffner et al. (2020). However, to
The Formation process is crucial for cell performance and safety, as it date, across this literature, no battery cost study is available that neither
builds up the solid electrolyte interphase (Winter 2009), which protects deals with the state-of-the-art nor with further optimized future pa­
the graphite-anode from adverse ongoing reactions with the electrolyte rameters, that have already been reported within literature, compre­
(Arora, 1998). During the Formation procedure, gas is generated within hensively. This paper will address this gap by translating current and
the cell. Applied external pressure causes the cell to expel the gas, and future parameters characteristics into costs. Therefore, a more
the cell is finally sealed. Lastly, an Aging procedure is conducted that comprehensive lithium-ion battery manufacturing cost model, which
takes up to several weeks and consists of storing the cells under relies on more than 250 calculation parameters, is presented.
controlled conditions and performing several quality measurements to
detect non-standard properties such as short circuits. (Michaelis et al., 3. Cost model
2016; Tagawa and Brodd, 2009; Verma et al., 2010). Only cells that
fulfill quality requirements in Final Control can be sold according to To develop the cost model presented within this study, PBCM is used,
their original purpose. If a cell does not fulfill these requirements it is which is a bottom-up technique that calculates manufacturing costs
stated as end-of-line scrap. The later a process step takes place in the analytically based on technical and operational parameters. Its reliance
value chain the more sensitive it is to scrap cost as more and more value on technical parameters makes the technique powerful, especially for
has already been added in previous process steps (Kwade et al., 2018). predicting costs for unexplored technologies, as technical data is usually
Hence, low end-of-line scrap rates are crucial for a competitive cell more easily available then historical cost data. Further, it generates
production. Typically, 95% of finished cells fulfill quality requirements transparency in regard to which parameters contribute the most to the

4
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

total cost and it enables the monetary quantification of potential manufacturing processes and to answer various research questions,
parameter improvements. These properties make the method highly including those focused on the evaluation of alternative processes, ma­
suitable for industries based on unexplored technologies that are facing terials and concepts or the evaluation of process improvements (Ciez
high cost pressure and therefore, anticipate numerous parameter opti­ and Whitacre, 2017; Farooq et al., 2018; Fuchs et al., 2006; Johnson and
mizations (Field et al., 2007; Fuchs et al., 2006; Nadeau et al., 2010). Kirchain, 2009a, 2009b; Nadeau et al., 2010; Sakti et al., 2015).
Based on its mentioned strengths, PBCM has been used for multiple The PBCM framework was introduced by Field et al. (2007). As

Table 2
Cost categories, related parameters, parameter descriptions and sources of parameter specifications.
Cost Parameter description Source
category

Overarching PVSalable Required number of annual salable units Derived from Michaelis et al. (2018)
PCEnergy Energy content per unit Wentker et al., 2019a
r Annual discount rate Ciez and Whitacre (2017)
DPY Operating days per year Nelson et al. (2019); Schnell et al. (2020)
CTotal Total unit cost Calculated (see Supplementary information)
j
PVEffective Gross number of units produced in process Calculated (see Supplementary information)
step j
j
CProcess Cost for process step j Calculated (see Supplementary information)

Variable cost j
CVariable Variable cost for process step j Calculated (see Supplementary information)

Cje Unit cost for cost element e ∈{Material, Labor, Calculated (see Supplementary information)
Energy}
ACe
j Annual cost for cost element e ∈{Material, Calculated (see Supplementary information)
Labor, Energy}

• Material Mj,Material Net mass of the material required BatPaC, 2018, Wentker et al., 2019a
x Machine-specific scrap losses for process step j Nelson et al. (2019), Ciez and Whitacre (2017)
Scrapj,Mat. Material type-specific scrap losses for process Nelson et al. (2019), Ciez and Whitacre (2017)
step j
UMaterial Unit cost of materials Nelson et al. (2019), Wentker et al., 2019a

• Labor NLaborers Number of laborers required per machine Nelson et al. (2019), 2012; Sakti et al. (2015); Schünemann (2015); expert discussions
NOS Number of shifts per day Ciez and Whitacre (2017)
OHS Operating hours per shift Ciez and Whitacre (2017)
UB Unpaid breaks hours per shift Fuchs et al. (2006)
PB Paid breaks hours per shift Derived from Fuchs et al. (2006); Sakti et al. (2015)
APOT Annual paid operating time Calculated (see Supplementary information)
ULabor Unit cost of labor Eurostat (2019)

• Energy SREnergy Surcharge rate for energy Ciez and Whitacre (2017); Sakti et al. (2015)

Fixed cost j
CFixed Fixed cost for process step j Calculated (see Supplementary information)
j
Ce Unit cost for cost element e ∈{Machine, Calculated (see Supplementary information)
Building, Maintenance, Overhead}
ACe
j Annual cost for cost element e ∈{Machine, Calculated (see Supplementary information)
Building, Maintenance, Overhead}

• Machine CT Cycle time Derived from Kaiser et al. (2014); Knoche (2017); Kwade et al. (2018); Mao et al. (2018); Nelson
et al. (2019), 2012; Sakti et al. (2015); Schünemann (2015); Tagawa and Brodd (2009); Wood et al.
(2015); Yoshio et al. (2009)
UD Unplanned downtime Sakti et al. (2015)
LMachine Useable lifetime of machines Nelson et al. (2019); Schnell et al. (2020)
UMachine Unit cost of machines Nelson et al. (2019), 2012; Sakti et al. (2015); Schünemann (2015); expert discussions
RjMachine Annual allocated machine costs for process Calculated (see Supplementary information)
step j
j
NMachine Number of machines required for process step j Calculated (see Supplementary information)

reqMT j Annual required machine time for process step Calculated (see Supplementary information)
j
availMTj Annual available operating time of a machine Calculated (see Supplementary information)

• Building FPMachine Footprint per machine Nelson et al. (2019), 2012; Sakti et al. (2015); Schünemann (2015); constructive design
LBuilding Useable lifetime of buildings PwC (2020)
UBuilding Unit cost of buildings Turner and Townsend (2018); ECC, 2019
RjMachine Annual allocated building costs for process Calculated (see Supplementary information)
step j

• Mainten. SRMainten. Surcharge rate for maintenance Ciez and Whitacre (2017); Sakti et al. (2015)

• Overhead SROverhead Surcharge rate for overhead Ciez and Whitacre (2017); Sakti et al. (2015)

5
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

shown in Fig. 3, it consists of three interconnected sub-models: the The model parameters related to the categories Technical Observa­
process model, operations model, and financial model. The process tions, Operating Conditions, and Factor Prices are listed in Table 2. For
model transforms product characteristics (e.g. size, shape, and material) reasons of clarity, these parameters have been classified according to
into technical parameters (e.g. cycle time, machine capacity, downtime, their cost category (e.g. variable, fixed) and their associated cost
and rejection rate) based on engineering, scientific, and technological element (e.g. material, labor).
principles. The operations model derives the overall resource re­
quirements (e.g. equipment, footprint, labor, and energy) based on 4. Scenario-based analysis
operating conditions (e.g. working days/year, shifts/day and hours/­
shift) and technical parameters. The financial model adds factor prices The target of the scenario-based analysis is to identify the current
to resources, which results in the manufacturing costs for a specific battery cost level by initializing the process-based cost model with state-
amount of a specific product (Field et al., 2007; Nadeau et al., 2010). of-the-art large-scale parameter specifications and to forecast a future
Since battery manufacturing comprises of a large number of indi­ cost level by translating most relevant innovations reported in academic
vidual and complex process steps, all of which mutually influence each discussion into costs. Therefore, two scenarios are defined, a Base Sce­
other, not all of the engineering, scientific, and technological principles nario representing the state-of-the-art in large-scale battery
involved have been studied in a holistic way in efforts to translate manufacturing and a future-oriented Optimized Scenario. For the Base
product characteristics into technical parameters. To address the infor­ Scenario, the battery literature is surveyed regarding characteristics that
mation gap for technical parameters, we rely on empirical data. To map represent both, the state-of-the-art production technology and materials
the battery-specific approach in the PBCM framework, we enlarged the and designs that are currently in use for large-scale production. Further,
framework by adding an additional input parameter category, namely a typical high-cost country for battery manufacturing is assumed as
technical observations, as shown in Fig. 3. plant location. For the Optimized Scenario, a categorized approach is
taken that classifies reported innovations from literature into process-
3.1. Model architecture related, material & design-related and location-related simulation pa­
rameters. For process-related simulation parameters, based on the
The presented model architecture operationalizes the PBCM tech­ resulting cell cost calculation in the Base Scenario, a literature review for
nique for the manufacture of battery cells. Similar architecture de­ the most cost-driving process steps is conducted and related process
scriptions have been introduced for other technologies (see e.g., Johnson innovations are identified. For material & design-related simulation
and Kirchain, 2009a). The model architecture used for the presented parameters, based on the resulting cell cost calculation in the Base
cost model is adapted, as it must ideally meet the specific requirements Scenario, a literature review for the most cost-driving component is
of battery-cell cost modelling. In the following, the most important conducted and reported material and design innovations are identified.
variable definitions and calculation rules are introduced. The complete For location-related simulation parameters, a typical low-cost country
model architecture can be found in Appendix A. for battery manufacturing is assumed as plant location, respective
The total unit cost CTotal is calculated by summing up the cost of all country-specific characteristics are taken from literature and integrated
process steps of battery cell manufacturing (see Fig. 1). The cost for in the parameter set. Finally, in order to ensure inter-scenario consis­
j j
process step j, CProcess , can be divided into variable CVariable and fixed tency of the underlying parameters taken from various sources, both
j j parameter sets have been discussed with industry experts. In the
CFixed . The variable cost CVariable includes the cost elements for Material
j j j j
following, each simulation parameter is described, including its effect on
CMaterial , Labor CLabor , and Energy CEnergy whereas the fixed cost CFixed the cell cost calculation, its value in both scenarios and potential chal­
j j
includes the cost elements for Machine CMachine , Building CBuilding , lenges in implementing respective improvements.
j j
Maintenance CMaintenance , and Overhead COverhead . The mathematical re­
lations are shown in equations (1)–(4), where n is the total number of 4.1. Process-related simulation parameters
considered process steps and j is the number of the specific process step.
∑
n [1] Increased roll-to-roll working speed
CTotal = j
CProcess (1)
The speed at which the so-called roll-to-roll processes (Coating &
j=1

Drying, Calendering, and Slitting) are performed depends on material

j
CProcess j
= CVariable + j
CFixed (2)
properties, process competencies and machine capabilities. An increase
in working speed decreases the time needed to produce the electrodes
j
CVariable j
= CMaterial j
+ CLabor j
+ CEnergy (3)
required for one battery cell. This effect is represented in the cost model
with a reduced cycle time for these process steps (CTj=Coating&Drying ,
j
CFixed j
= CMachine j
+ CBuilding j
+ CMaintenance j
+ COverhead (4)
CTj=Calendering , CTj=Slitting ). Where the cycle time is in general calculated
process step specific, for the process steps Coating & Drying, Calen­
3.2. Input parameters dering and Slitting the identical parameters are used. Namely, working
width of machine, working speed, cathodes/anodes per cell, width of
The presented cost model consists of more than 250 parameter cathode/anode, length of cathode/anode are used. A reduced cycle time
characteristics from the Product Characteristics, Technical Observa­ reduces the number of machines needed to produce a target volume of a
tions, Operating Conditions, and Factor Prices categories. The origin of product (NM). While a higher working speed of the roll-to-roll process
the data for the specific parameters used in the model is described in the steps increases the machine capabilities required, the cost-increasing
following. effects must also be considered. In particular, this means higher unit
Product Characteristics were mainly defined using the CellEst bat­ investments and larger machine footprints for all roll-to-roll processes.
tery cost model by Wentker et al., 2019. It proposes the selected cell Within this increase, additional capabilities are crucial for process step
dimensions, the pouch cell format and different cell chemistries, j = Coating&Drying, as a higher number of electrodes must be dried in a
reflecting a battery cell for a vehicle that is purely electric. For some certain time. This is technically achieved by increasing the length of the
parameters that are not considered in CellEst (e.g. process-related ma­ dryer that is part of the Coating & Drying machine unit. Accordingly, a
terial cost like NMP), data from BatPac has been included in the analysis. linear cost function for machine unit investments and the machine
Table 1 lists the product characteristics used for this study. footprint of the dryer unit must be assumed. A parameter value of 25 m

6
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

min− 1 is characterized as state-of-the-art and therefore applied in the Within the Formation procedure, the cell is charged and discharged
Base Scenario (Kwade et al., 2018). Regarding future developments, it is several times using specific charging rates to build up the solid elec­
reported that a coating speed of up to 100 m min− 1 is feasible and is trolyte interphase (Winter 2009). Both, the number of charge and
therefore set for the Optimized Scenario (Schmitt, 2015). To materialize discharge cycles required and the rate at which they can be executed
this higher coating speed, the bead pressure and low-flow limit (C-rate), defines the duration of the formation procedure within process
(maximum speed at a given film thickness, or the minimum film thick­ step j = Formation&Final Sealing. Optimizing these parameters results in
ness at a given speed, at which the coating bead remains stable) and its a cycle time reduction for this process step (CTj=Formation&Final Sealing ). For
associated parameters must be controlled especially to avoid film the Base Scenario, reported as state-of-the-art, we assume a typical
break-ups and to ensure film uniformity as basis to not deteriorate scrap formation procedure with 3.5 charge/discharge cycles and C-rates
rate and/or cell performance (especially unfavorable current distribu­ ranging from 0.05 to 0.5 C resulting in a formation time of 75 h (Wood
tions in operating cell and reduced power density due to increased cell et al., 2015). In the Optimized Scenario, a reduced formation protocol is
volume) (Carvalho and Kheshgi, 2000; Romero et al., 2004; Schmitt followed with 1.5 charge/discharge cycles and a C-rate of 0.5 C that
et al., 2014). totals in 11 h (Mao et al., 2018). As the graphite anodes are thermo­
dynamically unstable against the electrolyte, an optimal solid electro­
[2] Usage of alternative solvent lyte interphase layer is the basis to protect the graphite-anode from
adverse ongoing reactions with the electrolyte (Arora, 1998) which
As described in section 2.1, water is currently used as solvent to would have negative effects on cell capacity due to continuous elec­
produce active anode material slurry and teratogen and toxic NMP (N- trolyte decomposition, and graphite exfoliation (Mao et al., 2018;
methyl-2-pyrrolidone) is used to produce active cathode material slurry. Winter 2009). Thus, when further reducing charging cycles and/or
With regard to future, it is reported that water can also be used as solvent increasing C-rates, the product specific fulfillment of the solid electro­
to produce the active cathode material slurry (Bresser et al., 2018a; Du lyte interface requirements must be considered. To reduce formation
et al., 2017; Ibing et al., 2019; Wood et al., 2015). Using water instead of time while fulfilling these requirements at the same time, optimized
NMP results in three beneficial main effects: (1) The solvent recovery charging protocols (Mao et al., 2018) as well as optimized electrode
process can be completely omitted as it is only necessary due to the surfaces and electrolyte compositions (Winter 2009) can be key to
js=Solvent revovery
properties of the NMP (ACSupporting = 0). (2) The dryer length of the success.
Coating & Drying unit can be reduced as the evaporation rate of water is
twice as high as that of NMP (Wood et al., 2015). This reduction leads to [5] Decreased end-of-line scrap rate
j=Coating & Drying
reduced machine unit investments (UMachine ) and a decreased
j=Coating & Drying
Battery manufacturing is very cost sensitive to the scrap produced
machine footprint (FPMachine ). (3) The material unit cost for NMP due to the high number of process steps and the high share of material
(Uz=NMP
Material ) can be replaced by the lower cost of water(Uz=Water
Material ). Hence, a costs. The end-of-line scrap rate (xj=Aging&Final Control ) indicates the per­
state-of-the-art NMP-process is assumed for the Base Scenario and an centage of rejected parts identified during process step j =
aqueous cathode coating process for the Optimized Scenario, respec­ Aging&Final Control. The rate depends on the process quality of the in­
tively. However, a prerequisite for the substitution of NMP is the ability dividual upstream process steps as well as the ability to detect defective
to control the negative effects of using water: (1) Risk of metal disso­ parts at an early stage and exclude them from further production pro­
lution in water which can result in capacity losses if ions cannot be cess. In general, if scrap can be reduced, a lower number of cells must be
intercalated back to the cathode active material, (2) Risk of reduced produced to reach a certain target quantity of salable products. Typi­
cycle life due to higher slurry surface tension and lower adhesion cally, 5% of finished cells do not fulfill quality requirements, hence this
strengths between slurry and current collector, (3) Risk of reaction be­ rate is applied for the state-of-the-art Base Scenario. For the Optimized
tween the cathode alkaline water solution and the processing machines Scenario we assume the lower bound of 1% as end-of-line scrap (Ciez
and the collector foils (Bresser et al., 2018a; Du et al., 2017; Ibing et al., and Whitacre, 2017). There are in general two strategies to further
2019; Wood et al., 2015). reduce the end-of-line scrap rate: (1) Increase process quality within the
process steps along the whole value chain for which an in-depth un­
[3] Increased stacking speed derstanding of the process steps, its process parameters as well as its
interaction is of utmost relevance (Kwade et al., 2018), (2) The early
Within the state-of-the-art z-folding procedure, the stacking speed is detection of scrap components, as the later a process step takes place in
defined as the speed at which the electrodes (anodes and cathodes) are the value chain the more value is lost when identifying a defect. To
positioned in the zigzag fold separator during process step j = Stacking. materialize strategy (1), data-driven approaches (Turetskyy et al.,
The speed reached depends on the process competencies and the capa­ 2020), and for materialization of strategy (2), advanced quality man­
bilities of the stacking machine. An increase in stacking speed reduces agement concepts (Schnell and Reinhart, 2016) are currently discussed
the cycle time of the process step Stacking (CTj=Stacking ). For the Base as promising approaches.
Scenario, a state-of-the-art parameter value of 60 electrode sheets
picked and placed min− 1 is applied (Mooy, 2019). In recent literature,
4.2. Material and design-related simulation parameters
higher stacking speeds of or even exceeding 180 sheets min− 1 can be
observed (Sakti et al., 2015; Schnell et al., 2020; Schünemann, 2015).
[6] Increased electrode thickness
Therefore, this value is set for the Optimized Scenario. The main chal­
lenge to materialize higher stacking speeds is the increase of machine
Electrode thickness within battery cells can vary depending on the
capabilities to accelerate the highly automated pick-and-place operation
target product specifications of the battery cell. Using thick electrodes
while keeping sheet positioning and orientation accuracy. This accuracy
requires a fewer number of electrodes to achieve the target cell energy
is crucial for cell performance (especially capacity and safety), as it
(Ibing et al., 2019; Patry et al., 2015; Singh et al., 2016). Reducing the
determines the area coverage of anode and cathode sheets and prevents
number of electrodes has three cost-effective impacts: (1) a reduced
the physical contact of the electrode as basis to avoid short circuits
cycle time of the roll-to-roll processes (CTj=rtr ), the cutting process
(Mooy, 2019).
(CTj=Cutting ), and the stacking process (CTj=Stacking ), (2) a reduced mass of
Cathode foil Anode foil
[4] Accelerated formation procedure non-active electrode materials required (MMaterial ; MMaterial ), (3) an
increase in the dryer length of the Coating & Drying machine unit, since a

7
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

Table 3
Names of simulation parameter, parameter categories, parameter characteristics, and initial sources.
Number and name of simulation Parameter category Unit Base Opti- Source
parameter mized
1
[1] Increased roll-to-roll working Process-related Working speed in m min− 25 100 Base: Kwade et al. (2018)
speed Optimized: Schmitt (2015)
[2] Usage of alternative solvent Process-related Type of solvent NMP Water Base: Ciez and Whitacre (2017); Nelson et al. (2019),
2012; Sakti et al. (2015)
Optimized: Bresser et al. (2018a); Du et al. (2017);
Wood et al. (2015)
[3] Increased stacking speed Process-related Electrode sheets picked and 60 180 Base: Mooy (2019)
placed min− 1 Optimized: Schnell et al. (2020); Schünemann (2015)
[4] Accelerated formation procedure Process-related Number of charge/discharge 3.5 1.5 Base: Wood et al. (2015)
cycles Optimized: Mao et al. (2018)
c-rate 0.05–0.5 0.5
[5] Decreased end-of-line scrap rate Process-related Scrap rate in % 5 1 Base & Optimized: Ciez and Whitacre (2017)
[6] Increased electrode thickness Material & design- Coating thickness Cathode in 65 100 Base: Schmuch et al. (2018)
related μm Optimized: Zheng et al. (2012)
Coating thickness Anode in 78 120 Base & Optimized: Derived from CellEst, Wentker et al.,
μm 2019a
[7] Material change from NMC622 to Material & design- Type of material NMC NMC Base: Schmuch et al. (2018)
NMC811 related 622 811 Optimized: Wentker et al. (2019)
[8] Decreased unit cost for labor Location-related Unit costs for labor in $ h− 1 38 10 Base & Optimized: Eurostat (2019)
[9] Decreased unit cost for building Location-related Unit costs for buildings in $ 2345 1292 Base & Optimized: Turner and Townsend (2018); ECC,
space m− 2 2019
[10] Increased number of operating Location-related Operating days year− 1 300 360 Base: Nelson et al. (2019);
days Optimized: Schnell et al. (2020)
[11] Increased depreciation period Location-related Useful life of machines in 6 8 Base: Nelson et al. (2019)
for machines years Optimized: Schnell et al. (2020)
[12] Increased depreciation period Location-related Useful life of buildings in 25 50 Base & Optimized: PwC, 2020
for buildings years

Fig. 4. Cost reduction per simulation parameter (single simulation-parameter approach) in $ kWh− 1 @ 35 GWh annual factory capacity; [n] number of single
simulation parameters; Categorical affiliation: Process-related, Material & design-related, Location-related.

higher quantity of material must be dried per electrode (anode and Scenario a more cost-effective thicker electrode of 100 μm (Zheng et al.,
cathode) in the same time period. This increase leads to higher machine 2012) is assumed. Producing electrodes with a thickness of 100 μm and
unit investments (UMachine
j=Coating& Drying
) and increased machine footprints above is technical feasible. However, as mentioned, the optimum
j=Coating&Drying thickness of the electrodes depends on the target product specifications
(FPMachine For the Base Scenario, a common electrode thickness
).
of the battery cell. Thin electrodes are advantageous for products that
of 65 μm (Schmuch et al., 2018) is applied, whereas for the Optimized
are sensitive to durability and power. Thick electrodes are used for

8
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

still are under consideration) as a large-scale battery factory location.

For the Base Scenario, we choose Germany as plant location. This is
firstly due to the fact that several battery manufacturers (e.g. SAFT,
Northvolt) have decided to localize here (Michaelis et al., 2018) and
secondly it shows comparatively high labor cost levels ($38 h− 1, Euro­
stat, 2019). For the Optimized Scenario, we choose Hungary as plant
location, as there are already plants in operation or under construction
(e.g. Samsung SDI, SK Innovation) (Michaelis et al., 2018) and labor
costs are low within European comparison ($10 h− 1, Eurostat, 2019).

[9] Decreased unit costs for building space

Unit costs for building space (UBuilding ) depend on the location and
building requirements. Building requirements are fixed by the product
being produced and thus do not vary for a battery cell plant. Location
dependency, on the other hand, is variable and thus relevant for the
study, since different locations are considered for a battery cell factory.
To remain consistent with the labor cost scenarios introduced above,
building cost levels of Germany ($2345 m− 2) and Hungary ($1292 m− 2)
Fig. 5. Cost reduction per category (categorized simulation-parameter are considered for the Base Scenario and the Optimized Scenario,
approach) in $ kWh− 1 @ 35 GWh annual factory capacity.
respectively (ECC, 2019; Turner and Townsend, 2018).1

products that are more sensitive to energy density and costs. Conse­ [10] Increased number of operating days
quently, increasing electrode thickness without changing cell di­
mensions (as in the here presented model) results in altered battery A plant’s operating days per year (DPY) mainly depend on the fac­
characteristics such as increased energy density and decreased power tory’s production strategy. While some shut down days per year are
density. As the present study is focused on costs, simultaneous variations necessary for maintenance or due to public holiday, especially factories
in cell specifications are accepted (see Table 1). using highly automated and cost-intensive manufacturing equipment
are encouraged to maximize the number of operating days, as this results
[7] Material change from NMC622 to NMC811 in an increased annual available operating time of machines (availMT).
Therefore, the number of machines required to produce a target volume
Since cathode active materials represent the most cost-driving cell (NM) decreases. A parameter value of 300 days year− 1 has been set in
components (cell cost share > 30% in the Base Scenario) and they the Base Scenario to limit fixed assets to a reasonable extent (Nelson
currently represent a bottleneck for cell performance (Schmuch et al., et al., 2019). For the Optimized Scenario a value of 360 days year− 1 was
2018), significant research effort has been spent on their development. taken into account that has been used in recent literature (Schnell et al.,
An improvement in their inherent characteristics, such as specific ca­ 2020).
pacity or crystallographic density, results in a higher energy content per
unit PCEnergy , thereby reducing material input quantities required to [11] Increased depreciation period for machines
produce a certain amount of energy output and hence reducing cell cost
(Schmuch et al., 2018). While NMC622 is considered a state-of-the-art The time span in which a machine is depreciated is represented by its
cathode active material (Schmuch et al., 2018) and is therefore useful life, LMachine , which depends on its technical durability and the
applied in the Base Scenario, one strategy to enhance energy content is length of time it can economically produce products with the required
to increase the nickel share. Hence, the use of NMC811 with a higher performance. In general, depreciation periods are longer in low cost
nickel content is assumed for the Optimized Scenario (Wentker et al., countries (e.g. Hungary) compared to high cost countries (e.g. Germany)
2019). The resulting product characteristics for both scenarios such as (PwC, 2020). Battery related literature report periods range from 6 to 8
cell energy, energy density and number of electrodes per cell are pre­ years. Therefore, we used 6 years (Nelson et al., 2019) for the Base
sented in Table 1. However, when enhancing the nickel share, the cobalt Scenario and 8 years (Schnell et al., 2020) for the Optimized Scenario.
and manganese shares are reduced simultaneously. As the nickel and the
cobalt provide structure stability within the cathode active material, [12] Increased depreciation period for buildings
which is primarily responsible for cell properties like thermal stability or
cycle life, strategies must be developed to address these challenges to
avoid limitations in cell performance (Andre et al., 2015). Promising 1
To survey the location-specific factor cost of building, we use a two-stage
approaches were already reported in literature related for example to
approach as no data are directly available. Turner and Townsend (2018) pub­
structural design (core–shell structure, concentration gradient, etc.) and
lished absolute construction costs for some European countries, and ECC (2019)
intrinsic structure optimization (Wang et al., 2020). published percentage values of construction costs for all EU-28 countries.
Hence, we defined the United Kingdom as the baseline since, among EU
4.3. Location-related simulation parameters countries, Turner and Townsend published the most comprehensive data for
this country, including the differentiation of costs in the United Kingdom ac­
[8] Decreased unit costs for labor cording to region, industry affiliation, and type of building. Since cell produc­
tion involves many consecutive and complex production steps (Kwade et al.,
2018), which are partly performed under increased environmental re­
The unit costs for labor (ULabor ) depend on the location, the required
quirements (Ahmed et al., 2017; Nelson et al., 2015; Yoshio et al., 2009), we
competence level of workers, and the industry. The required level of use values from the industrial high-tech factory/laboratory category to calcu­
competence and industry are fixed by the product and are thus not late an average cost based on the region-specific values in the United Kingdom.
variable. Location dependency, on the other hand, is variable and thus Using this baseline value from Turner and Townsend and the percentage values
relevant for this study, since different locations are considered for a from EEC, we calculate the construction cost per square meter for each Ger­
battery cell factory. In Europe, several countries have been chosen (or many and Hungary.

9
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

1
Fig. 6. Cost walk from the Base Scenario to the Optimized Scenario based on cost elements (overall simulation-parameter approach) in $ kWh− @ 35 GWh annual
factory capacity.

The time span in which a building is depreciated is represented in its parameter approach). Thereby, a cost walk at the cost-element level is
useful life, LBuilding , which depends on its durability and production re­ presented from the total costs of the Base Scenario to the total costs of
quirements regarding the shop floor layout and environment. Produc­ the Optimized Scenario. The total costs are $106 kWh− 1 in the Base
tion requirements can change over time, even if the same type of product Scenario and $64 kWh− 1 in the Optimized Scenario. This corresponds to
is manufactured. Again, to remain consistent with the building cost a cost reduction of 40%.
scenarios, Germany as a high cost country is chosen for the Base Sce­ The cost walk shows that the presented simulation parameters
nario (depreciation period of 25 years for buildings, PwC, 2020) and reduce, in particular, the cost elements Material, Machine and Labor.
Hungary as a low cost country is chosen for the Optimized Scenario The improvement in the cost element Material mainly results from
(depreciation period of 50 years for buildings, PwC, 2020). Increased electrode thickness and Material change from NMC622 to
To enable an overarching discussion of cost effects and to derive a NMC811 that both induce an increased energy content per cell, thereby
total cost level, beside the evaluation of the single simulation parameters reducing material input quantities required to produce a certain amount
and the evaluation of the described scenarios, an evaluation based on of energy output. For cost elements Machine and Labor, this reduction is
parameter categories (Process-related, material & design related, and regarding that, on the one hand, in the Optimized Scenario, fewer factor
location-related simulation parameters) is conducted. quantities (less machinery and less working hours per unit of output) are
Table 3 lists the simulation parameters, the associated characteris­ necessary to produce the target volume, and, on the other hand, the
tics, their sources and the associated parameter category. factor price for labor is lower. Looking at the cost reductions by cost
element at the single-parameter level reveals that all simulation pa­
5. Results and discussion rameters have a cross-cost-element effect. Nevertheless, the level of the
effect of each parameter differs. The simulation parameters for Increased
The following section presents and discusses the results of the study, electrode thickness and Material change from NMC622 to NMC811 affect
which are based on the presented model architecture (see Section 3.1), each of the eight cost elements. On the other hand, the simulation pa­
the presented input parameters (see Section 3.2), the presented simu­ rameters Decreased unit cost for labor, Decreased unit cost for building
lation parameters (see Section 4) and a targeted annual production ca­ space, and Increased depreciation period for buildings each affect only two
pacity of 35 GWh. of the eight cost elements. When looking at the superordinate catego­
Fig. 4 shows the cost reduction per simulation parameter (single rization level, it is striking that all simulation parameters from the
simulation-parameter approach) in $ kWh− 1. The highest values result process-related and material & design-related categories have a strong
from the two material and design-related optimizations Increased elec­ cross-cost element, as they affect at least six of the eight cost elements
trode thickness at $12.9 kWh− 1 and Material change from NMC622 to considered. On the other hand, the simulation parameters of the
NMC811 at $8.9 kWh− 1. This is followed by the process-related opti­ location-related categories influence only two or three of the eight cost
mization Accelerated formation procedure at $6.9 kWh− 1 and the elements considered.
location-related optimization Decreased unit cost for labor at $6.3 kWh− 1. Although all parameter characteristics within the Optimized Sce­
The value of each of the remaining simulation parameters is < $5 nario are derived from literature, there is an uncertainty to which extend
kWh− 1. the assumed parameter optimizations can be achieved within future
Fig. 5 shows the results based on the categorized simulation- large-scale manufacturing. To evaluate this uncertainty a sensitivity
parameter approach. The results reveal that the highest cost re­ analysis is conducted and can be found in Appendix C.
ductions, at $20.3 kWh− 1 result from the process-related category. This Transforming these overall cost results into vehicle-level figures and
is followed by the material & design-related and location-related cate­ linking them to revenues and profits illustrates the importance of cost-
gories, with cost reductions of $19.4 kWh− 1 and $14.9 kWh− 1, respec­ efficient battery production. Multiplying the average energy content of
tively. The result that process-related innovations have the highest a battery for a mid-range vehicle, 60 kWh (Schmuch et al., 2018), with
impact on cell cost indicates that cost-optimal cell production depends the cost per kWh from the Base Scenario and the cost per kWh from the
in particular on company-and factory-specific process competence. Optimized Scenario results in a cost difference of approximately $2500
Compared to this process competence, the often-discussed location- per vehicle, thereby in the case of full materialization, significantly
driven parameters (Brodd and Helou, 2013; Duffner et al., 2020a), impacting the economic success of EVs.
which are represented by the location-related category, are of minor Fig. 7 shows the percentage distribution of the total costs among the
importance. Especially in the context of labor, this is because batteries cost elements for the two overall scenarios (overall simulation-
are manufactured using highly automated and thus less labor-intensive parameter approach). In both the Base Scenario and the Optimized
processes. Scenario, the cost elements that drive costs the most are Material and
Fig. 6 shows the results of the overall scenarios (overall simulation- Machine. The sharp increase of the Material cost share in the Optimized

10
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

Fig. 7. Percentage of costs per cost element for the Base Scenario and Optimized Scenario @ 35 GWh annual factory capacity.

Fig. 8. Cost share between process steps in % for the Base Scenario and Optimized Scenario @ 35 GWh annual factory capacity.

Scenario results from the disproportionately low decrease of material striking that the proportion of the cost element Labor is 2% in the
cost compared to other cost elements. This is due to the fact that only Optimized Scenario, compared to 8% in the Base Scenario. This effect is
three out of twelve examined simulation parameters have an impact on mainly driven by the simulation parameter Decreased unit cost for labor as
material costs. In contrast, Machine cost decreases overproportionately it has a high absolute value and mainly affects the cost element Labor,
since it is optimized by nine simulation parameters. In addition, it is whereas the other simulation parameters with a high value, as

11
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

described, have a much stronger cross-cost-element effect. Production (32%), Cell Production (37%), and Cell Conditioning (31%).
The high ratio of the cost elements Material (77% in the Optimized In the Optimized Scenario, the dominance of the Mixing process and the
Scenario) and Material-Scrap (6% in the Optimized Scenario) to total reduction in Formation & Final Sealing lead to a cost shift from Cell
costs show that large-scale battery-cell production is highly sensitive to Conditioning (20%) to Electrode Production (44%). The process cost
net material input quantities, scrap rates and costs of purchased mate­ share of Cell Production remains at the same magnitude (36%).
rials. From a materials-related point of view, measures to optimize cell Taking all the results into account, for cost reduction in optimized
chemistry that focus on using fewer and more cost-effective materials large-scale battery cell factories, the focus should be on the process steps
are appropriate to reduce both Material and Material-Scrap costs. From Mixing, Coating & Drying, Stacking, Formation & Final sealing and Aging &
a process-related point of view, the focus has to be set on Scrap as its cost Final Control. For this purpose, in the following, process-step-specific
impact is amplified by the large number of interlinked and complex measures for cost reduction are described that complement the previ­
process steps in cell production. Thus, measures to produce a smaller ously described cost-reduction measures.
number of process-step-specific defective parts and measures to identify
defective produced parts as soon as possible so that no additional costs • Mixing: Measures to reduce the quantity of material to be mixed per
arise from subsequent process steps are suitable to reduce Material- salable cell, such as reducing the solvent quantity; measures to
Scrap costs. reduce mixing time, such as optimizing process specification; and
In addition to the cost elements Material and Material-Scrap, the cost measures to optimize batch capacity. In addition, from a cost
element Machine should be the focus of future cost-reduction activities, perspective, promising approaches are taken in science and industry
as this cost element accounts for the second largest share in the Opti­ to switch from batch-wise mixing to continuous mixing (Bühler
mized Scenario at 8%. Both operational and strategic measures can Group, 2017; Dreger et al., 2015).
contribute to reducing costs from this cost element. From an operational • Coating & Drying: Measures to further increase working speed;
point of view, measures that contribute to the production of specific measures to increase working width; measures to reduce solvent
quantities of acceptable parts with fewer machines are suitable. These quantity up to complete elimination (dry coating); measures to
include measures to reduce cycle time, increase the annual operating reduce the number of electrodes, such as increasing the coating
time (e.g. by reducing unplanned downtime), or to reduce the scrap rate. thickness.
From a strategic point of view, cost reductions can be achieved by • Stacking: Measures to increase the pick and place speed of the
extending the useful life of machines. This would be conceivable if a cell electrodes; measures to reduce positioning time during cell ex­
manufacturer, after a period of producing state-of-the-art cells change; measures to reduce the number of electrodes, such as
(regarding quality and performance), decides not to scrap deprecated increasing the layer thickness.
machines but continue producing battery cells with lower performance • Formation & Final Sealing: Optimized cycle programs to reduce
and quality requirements. cycle time. Therefore, measures to reduce number of cycles or to
Fig. 8 shows the process steps that drive costs the most in the Base increase the charging rate per cycle are appropriate.
Scenario and the Optimized Scenario. To increase comparability with • Aging & Final Control: Measures to reduce aging duration due to
the focus on operative process step execution, Material and Material- improved quality forecast methods.
Scrap have been excluded for evaluation. This approach was chosen
because material costs, as described above, make up a high proportion of Fig. 9 compares the results of the Base Scenario and the Optimized
the cost of battery cells and are for calculation purposes allocated to a Scenario with the reported cell costs across battery-related literature.
specific process step. This allocation is correct from a calculation point Therefore, a comprehensive literature review of reported cell cost esti­
of view but not ideal for comparability of the process steps, since the mations has been conducted.2 Firstly, as cathode chemistry strongly
process steps in which cost-intensive materials are used would exhibit influences cell cost, values for NMC have been taken into account
disproportionately high costs, even though the added material costs are wherever possible to allow for a more precise comparison with the re­
not directly related to execution of the cell making process. sults of the present study. Secondly, as not all publications report cost
In the Base Scenario, the three most cost-driving process steps are based on cell level (e.g. on pack level), corrective factors have been
Formation & Final Sealing (25%), Stacking (22%) and Coating & Drying calculated (derived from Nelson et al., 2019) and applied to the origi­
(12%). Due to their cost-driving properties, these process steps have nally reported cost values. The underlying data, associated sources and
been focused within past research. The reported optimization has been corrective factors are included in Appendix D. The plot of the reported
considered within the Optimized Scenario, which reduces their share values over the years of publication shows a continuously decreasing
significantly. cost trend that is characterizing for technologies that are undergoing
Formation loses its cost dominance to Mixing as a result of the mass industrialization. Reported cost estimations range from far above
Accelerated formation procedure and plays only a subordinate role in the $200 kWh− 1 in 2015 (Sakti et al., 2015) to below $100 kWh− 1 in 2018
Optimized Scenario. Mixing process costs get most relevant as they are (Schmuch et al., 2018). Furthermore, a tendency to a lower spread be­
only decreased by one optimization of the process-related category tween values can be noticed throughout the years which might result
(Decreased End-of-line Scrap). The process steps Stacking and Coating & from increasingly existent and accessible empirical evidence in industry.
Drying however, remain cost drivers in the Optimized Scenario despite The comparison of the derived value of our state-of-the-art oriented Base
the optimized parameter assumptions of a tripled stacking speed from 60 Scenario ($106 kWh− 1, horizontal dotted line) with literature values
to 180 sheets min− 1 (simulation parameter: Increased stacking speed), a shows that it is in line with recently reported values albeit it lies on the
quadrupled roll-to-roll speed from 25 to 100 m min− 1 (simulation
parameter: Increased roll-to-roll speed) and a reduction of the cathode
dryer length by replacing water with NMP (simulation parameter: Usage
2
of alternative solvent). It should also be noted that in the Optimized To survey battery cost literature, 14 publications between 2014 and 2019
Scenario, the cost-driving process steps Mixing, Coating & Drying and have been reviewed. Wherever possible, reported values regarding NMC have
been taken into account to allow for a more precise comparison with the pre­
Stacking are supplemented by the process step Aging & Final Control. As
sent study. If multiple values have been reported (e.g. due to different sce­
this step is no roll-to-roll process and no specific parameter optimization
narios) an average of those values has been calculated. As values have not
has been simulated, its cost share is increasing compared to the Base always been reported based on cell level, these have been normalized by
Scenario. corrective factors derived from Nelson et al. (2019): Reported pack cost have
Interestingly, in the Base Scenario, process costs are almost evenly been multiplied by 0.75 and reported cell material cost have been multiplied by
shared between the three superordinate main processes Electrode 1.31 to derive a cell-based cost value.

12
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

Fig. 9. Comparison of costs for NMC based battery cells reported in literature over time.

lower end of the range. In contrast, the value of the Optimized Scenario model that can be used as basis for data-driven decision making within
($64 kWh− 1, horizontal solid line) is significantly lower than historically location decisions, make or buy decisions and to evaluate process and
published cost estimations. The comparison with the most recent liter­ material alternatives. (3) We introduce and prove the feasibility of a new
ature shows that costs of both, the Base Scenario and the Optimized and lower cost level which, if materialized, will have a positive effect on
Scenario, are significantly lower than those reported for example by Ciez the EV penetration rate. Accordingly, it should be considered in future
and Whitacre (2017). The main reason for this is that Ciez and Whitacre research studies and product planning of automotive original equipment
used manufacturing parameters based on a much smaller factory size manufacturers cost optimizations.
(0.87 GWh annual production capacity). Although simulation for larger The presented model has several limitations that need to be
factory sizes is conducted, cost decrease is limited as the manufacturing addressed by future research. First, it is limited to lithium-ion battery
parameters are not adapted. Compared with Nelson et al. (2019), the technology, which is currently the most beneficial battery technology
values are in a similar range, with costs for the Base Scenario above the for automotive applications (Betz et al., 2019; Bresser et al., 2018b;
BatPaC range and costs for the Optimized Scenario below the BatPaC Placke et al., 2017; Schmuch et al., 2018). However, there are other
range. In summary, we introduce with the Optimized Scenario a lower technologies such as lithium metal, solid-state, sodium-ion, lithium||
cost level for the manufacturing of battery to literature and give di­ sulfur, or metal||air batteries with promising advantages for some
rections which technical parameters reported in literature must there­ characteristics such as costs or safety (Eftekhari and Kim, 2018; Tan
fore be industrialized in large-scale. et al., 2017; Zhang et al., 2017) and also for other applications than
automotive. These new technologies are still in the development stage,
6. Conclusion and it will take years until they are produced in large scale, but the
presented approach can be adapted to them. Second, the selection of
This study at hand successfully applies the process-based cost- simulation parameters is focused on the process-related, material &
modelling technique to the manufacture of battery cells. Accordingly, design-related and location-related parameters that drive cost most. This
the study contributes to the research fields of both process-based cost cross-category approach gives a comprehensive indication of current
modelling and battery technology. and future cost levels within battery manufacturing but makes no claim
The PBCM research field is complemented by the application of the to completeness within each single category. As described in Section 5,
technique to a complex technology whose manufacturing process con­ for example, there are also future process improvement possibilities for
sists of a variety of individual, complex, and interdependent process the process steps Mixing and Aging & Final Control, which are not
steps and whose engineering, scientific, and technological principles considered in the Optimized-Scenario. Third, although the used pa­
have not been fully researched. Therefore, we enlarge the PBCM rameters are derived from literature, are validated by industry experts,
framework introduced by Field et al. (2007). and are translated into costs by using an established cost estimation
In the field of battery research, we make several contributions (1) We technique, the calculated costs are not yet validated by empirical data of
give directions how further cost reductions within large-scale a battery cell manufacturer what offers together with the other
manufacturing can be achieved (Optimized Scenario) and which cost mentioned limitations new perspectives for future research.
drivers within process steps, cost elements and parameter categories
should be focused for future cost reduction activities. (2) We introduce a

13
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

Declaration of competing interest related to this manuscript.

The authors of this article states that there are no conflicts of interest

Appendix

Appendix A. Detailed cost model

Overall
The following describes the calculation of the variables, which have an overarching character as they are used as basis for the calculation of various
j
cost elements. The first variable of this type is the annual gross number of units PVEffective that must be produced within process step j to reach the
required number of annual salable units, PVSalable . Therefore, the logic must consider that a process step must compensate for the scrap losses of the
subsequent process steps. This means that process steps located early in the process chain must produce more units than later steps. For the last process
j
step within the presented process chain, the gross number of units required, PVEffective , is calculated as shown in equation (5). For the other process
steps, the gross numbers of units required are calculated as shown in equation (6), wherexj is the machine-specific scrap losses for process step j.
PVSalable
j
PVEffective =( ) (j = n) (5)
1 − xj

j+1
PVEffective
j
PVEffective =( ) (j < n) (6)
1 − xj

A second key variable is the number of required machines, Nj , at process step j. This is calculated by the process-step-specific annual operating time
required to produce the target volume, reqMT j , and the annual available operating time of the machine, availMT j , at process step j. The result is
rounded to the next highest whole number, as it is assumed in the present study that machines may be underutilized as they are not used to produce
other products.
reqMT j
j
NMachine = (7)
availMT j
The annual time required to produce the target volume, reqMTj , at process step j is calculated as the product of the associated cycle timeCT j , which
j
is the time interval after one unit is produced, and the associated gross number of cell equivalents required, PVEffective .
j
reqMT j = CT j x PVEffective (8)

The annual available operating time of a machine, availMT , at process step j is calculated according to equation (9), where DPY is the operating
j

days per year, NOS is the number of shifts per day, OHS is the operating hours per shift, UB is the unpaid breaks hours per shift, PB is the paid break
hours per shift, and UDj is the process-specific unplanned downtime hours per shift− 1.
( )
availMT j = DPY x NOS x OHS − UB − PB − UDj (9)

The costs presented in this study are on a kWh cell energy basis. Therefore, the annual cost for each cost element, ACjElement, must be divided by the
number of annual salable units (number of cells), PVSalable , an the energy content per unit (kWh cell − 1) PCEnergy, as shown in equation (10). Element is
used to differentiate between the cost elements Material, Labor, Energy, Machine, Building, Maintenance, and Overhead.
j
ACElement
j
CElement = (10)
PVSalable x PCEnergy

Machines and buildings are capital goods used over a specific period. In the present model, this is considered by distributing the investment in
capital goods such as machines and buildings over their life cycles. In addition, opportunity costs are considered by means of a capital recovery factor.
These costs arise because capital is tied up in machines and buildings and thus cannot be used to generate alternative revenue. The annual allocated
costs Rje for process step j are calculated as shown in equation (11), where Uje is the unit cost for a machine or building, e is used to differentiate between
machine and building, r is the annual discount rate, and Lje is the useable lifetime in years.
j
r(1 + r)Le
Rje = Uej j (11)
Le
(1 + r) − 1

Material
j
The calculation of annual material costs, ACMaterial , for process step j considers that several material types z may be used within process step j.
j
Therefore, annual material costs, ACMaterial , are calculated as the sum of the corresponding annual material-type-specific material costs, ACzj, Material , as
shown in equation (12), where z is a specific material type and d is the total number of material types used within process step j.
∑
d
j
ACMaterial = ACj,z Material (12)
z=1

14
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

The annual material costs for a material type z within a process step j, ACzj,Material , are calculated using equation (13), where Mzj,Material is the net mass
of material type z required, within process step j, to produce one unit; UMaterial
z
is the unit cost of material type z (usually currency weight− 1);
j
Scrapzj,Material is the percentage of specific losses of additional material type z within process step j; and PVEffective is the process-step-specific annual gross
number of units required.
1
z
ACj,Material total
z
= Mj,Material z
x UMaterial x j
x PVEffective (13)
1 − Scrapzj,Material

To ensure a holistic calculation and enable computation of the cost elements for Energy CjEnergy and Overhead CjOverhead , total annual material costs,
ACzj,Material total ,
were calculated in the present cost model as shown in equation (13). For evaluation, we focused on the part of the annual material costs
that are potentially influenced by production, which, in principle, is the part induced by scrap losses. Annual material-scrap costs, ACzj,Material scrap , are
calculated as shown in equation (14). This approach ensures the focus of the evaluation remains on the costs relevant to manufacturing. Otherwise, the
results would be overshadowed by the part of material costs that is not potentially influenced by production, as this is dominant within battery cells.
z
ACj,Material scrap
z
= Mj,Material z
x UMaterial j
x Scrapzj,Material x PVEffective (14)

Labor
j
As shown in equation (15), the labor costs for process step j are calculated as the product of the number of machines required for that step, NMachine ,
j j
the number of laborers required per machine, NLaboreres per Machine , the unit cost of labor, ULabor (usually in currency per unit time), and the annual paid
operating time, APOT j .
j
ACLabor j
= NMachine j
x NLaboreres j j
per Machine x ULabor x APOT ; (j ∕
= Supporting) (15)

Labor costs for process step j = Supporting are calculated differently, since this process step cannot be described on a machine basis. Nevertheless,
the same calculation logic is used, referring to reference bases (e.g., annual GWh plant capacity) instead of machines. As process step j = Supporting
consists of various sub-processes with different characteristics, the reference bases were built to be specific to each sub-process step. For each reference
base and sub-process step, the necessary number of laborers was determined. The following reference bases were used: annual factory capacity for the
sub-process - Inter-process material handling, Control lab, and Receiving and shipping including scrap recycle; annual amount of NMP solvent used for the
sub-process step -Solvent recovery; and footprint required for the processes to be carried out under dry room conditions (cell production) for the sub-
process step -Dry room management. Based on this, the labor costs of process step j = Supporting comprise the sum of the labor costs of the sub-
js
processes, as shown in equation (16), where NRBs is the number of reference bases required; NjsLaboreres per RB is the number of laborers required per
reference base; UjLabor is the unit cost of labor (usually in currency unit time− 1); APOTjs is the annual paid operating time; js is a specific sub-process of
process step j = Supporting, and ns is the total number of sub-processes of process step j = Supporting.
∑
ns
j
ACLabor = js
NRBs js
x NLaboreres js js
per RB x ULabor x APOT ; (j = Supporting) (16)
js

Machine
j
As shown in equation (17), the annual machine costs ACMachine for process step j ∕
= Supporting are calculated as the product of the annualized
j j
equivalent of the machine investment RMachine and the number of machines NMachine required to produce the process-step-specific gross target number
j
of produced units, PVEffective .
j
ACMachine j
= NMachine x RjMachine ; (j ∕
= Supporting) (17)
As described in the section on the labor cost element, process step j = Supporting cannot be described on a machine basis. Therefore, the cost
element for this process step, ACj=Supporting
Machine , is calculated using the described reference bases, where NjsRBs is the number of reference bases required;
RjsRB Machine is the annualized equivalent of the machine investment per reference base; js is a specific sub-process of process step j = Supporting; and ns is
the total number of sub-processes of process step j = Supporting. The following equation describes the calculation:
∑
ns
j=Supporting
ACMachine = js
NRBs x RjsRB Machine ; (j = Supporting) (18)
js

Building
j
As shown in equation (19), the annual building costs for process step j ∕
= Supporting, ACBuilding , are calculated as the product of the annualized
equivalent of building investment, RBuilding ($ m− 2), the number of machines required to produce the process-step-specific gross target number of cell
j

j j
equivalents, NMachine , and the footprint per machine, FPMachine (m2 machine− 1).
j
ACBuilding = RjBuilding x NMachine
j
x FPjMachine ; (j ∕
= Supporting) (19)

As shown in equation (20), the building costs for process step j = Supporting, CjBuilding are again calculated using the reference bases, where NjsRBs is
the number of reference bases required;FPjsRB is the footprint per corresponding reference base; js is a specific sub-process of process step j = Supporting;

15
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

and ns is the total number of sub-processes of process step j = Supporting.

∑
d
j=Supporting
CBuilding = RjBuilding x NRBs
js
x FPjsRB ; (j = Supporting) (20)
z=1

Energy, maintenance, and overhead

j j j
As shown in equations (21)–(23), the annual costs for Energy ACEnergy , Maintenance ACMaintenance , and Overhead ACOverhead for process step j are
j
calculated based on the percentage of surcharge rates SR. The surcharge rates are applied to known cost elements (e.g., Machine ACMachine ). Both the
surcharge rates SR and associated application bases are specific for the production of battery cells.
( j )
j
ACEnergy = ACLabor j
+ ACMaterial x SREnergy (21)

j
ACMaintenance j
= ACMachine x SRMaintenance (22)
( )
j
ACOverhead j
= ACMachine j
+ ACBuilding j
+ ACMaintenance x SROverhead (23)

Appendix B. Overlapping phenomena between simulation parameters

Fig. 10. Pairwise combination of simulation parameters and resulting effect on cell cost.

Appendix C. Sensitivity analysis3

3
For each simulation-parameter the improvement between Base Scenario and Optimized Scenario has been calculated. To account for uncertainty regarding the
extend of achievement has been varied. (1) Only 80% or (2) 120% of improvement can be materialized. The resulting impacts on cell costs are displayed in the
tornado plot. [2] Usage of Alternative solvent & [7] Material change from NMC622 to NMC811 have been excluded from the analysis as the parameter value can
either be yes or no in both cases

16
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

Fig. 11. Tornado plot showing the sensitivity of simulation-parameters.

Appendix D. Reported cell cost estimations in literature

Table 4
Source, year of publication and reported battery cost

Source Year Cathode Cost Reported cost [$ kWh− 1] Average value [$ Corrective Factor Resulting cell cost [$
material Base kWh− 1] [-] kWh− 1]

Gallagher et al. 2014 NMC Pack 178; 289 234 0.75 175
Patry et al. 2015 NMC Cell 306; 231; 231; 239; 247 251 – 251
Sakti et al. 2015 NMC Pack 545; 325; 265; 230 341 0.75 256
Berg et al. 2015 NMC Cell 224 224 – 224
Nelson et al. 2015 NMC Pack 188 188 0.75 141
Wood et al. 2015 NMC Pack 503; 370 437 0.75 327
Schünemann 2015 NMC Cell 189; 173; 163; 156; 189; 188; 188; 187; 292; 203 – 203
267; 245
Petri 2015 NMC Cell 122 122 1.31 159
mat.
Ciez and 2017 NMC Cell 244; 182; 174 200 – 200
Whitacre
Ahmed et al. 2017 NMC Pack 148; 155; 139; 146 147 0.75 110
Berckmans et al. 2017 NMC Pack 432; 300; 293 342 0.75 256
Schmuch et al. 2018 NMC 811 Cell 79; 68 73 1.31 96
mat.
Wentker et al. 2019 NMC 811 Pack 179; 167; 154; 143; 163; 152; 139; 129; 134; 142 0.75 106
125; 113; 103
Nelson et al. 2019 NMC 811 Cell 127; 109; 106; 103; 132; 105; 101 112 – 112

Funding sources

This work was partly funded and supported by the Bundesministerium für Bildung und Forschung (BMBF) and the Ministerium für Wirtschaft,
Innovation, Digitalisierung und Energie des Landes Nordrhein-Westfalen within the project BenchBatt [03XP0047A].

References Arinicheva, Y., Wolff, M., Lobe, S., Dellen, C., Fattakhova-Rohlfing, D., Guillon, O.,
Böhm, D., Zoller, F., Schmuch, R., Li, J., Winter, M., Adamczyk, E., Pralong, V.,
2020. Ceramics for electrochemical storage. In: Advanced Ceramics for Energy
Aaldering, L.J., Leker, J., Song, C.H., 2019. Analysis of technological knowledge stock
Conversion and Storage. Elsevier, pp. 549–709.
and prediction of its future development potential: the case of lithium-ion batteries.
Arora, P., 1998. Capacity fade mechanisms and side reactions in lithium-ion batteries.
J. Clean. Prod. 223, 301–311. https://doi.org/10.1016/j.jclepro.2019.03.174.
J. Electrochem. Soc. 145, 3647. https://doi.org/10.1149/1.1838857.
Ahmed, S., Nelson, P.A., Gallagher, K.G., Susarla, N., Dees, D.W., 2017. Cost and energy
Avicenne Energy, 2019. The Rechargeable Battery Market and Main Trends 2018-2030.
demand of producing nickel manganese cobalt cathode material for lithium ion
BatPaC, 2018. BatPaC Version 3.1 [WWW Document]. http://www.cse.anl.gov/batpac/
batteries. J. Power Sources 342, 733–740. https://doi.org/10.1016/j.
download.php (accessed 5.8.19).
jpowsour.2016.12.069.
Ben-Arieh, D., Qian, L., 2003. Activity-based cost management for design and
Andre, D., Hain, H., Lamp, P., Maglia, F., Stiaszny, B., 2017. Future high-energy density
development stage. Int. J. Prod. Econ. 83, 169–183. https://doi.org/10.1016/S0925-
anode materials from an automotive application perspective. J. Mater. Chem. 5,
5273(02)00323-7.
17174–17198. https://doi.org/10.1039/C7TA03108D.
Berg, E.J., Villevieille, C., Streich, D., Trabesinger, S., Novák, P., 2015. Rechargeable
Andre, D., Kim, S.-J., Lamp, P., Lux, S.F., Maglia, F., Paschos, O., Stiaszny, B., 2015.
batteries: grasping for the limits of chemistry. J. Electrochem. Soc. 162,
Future generations of cathode materials: an automotive industry perspective.
A2468–A2475. https://doi.org/10.1149/2.0081514jes.
J. Mater. Chem. 3, 6709–6732. https://doi.org/10.1039/C5TA00361J.

17
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

Betz, J., Bieker, G., Meister, P., Placke, T., Winter, M., Schmuch, R., 2019. Theoretical Johnson, M., Kirchain, R., 2009a. Quantifying the effects of parts consolidation and
versus practical energy: a plea for more transparency in the energy calculation of development costs on material selection decisions: a process-based costing approach.
different rechargeable battery systems. Adv. Energy Mater 9, 1803170. https://doi. Int. J. Prod. Econ. 119, 174–186. https://doi.org/10.1016/j.ijpe.2009.02.003.
org/10.1002/aenm.201803170. Johnson, M., Kirchain, R., 2009b. Quantifying the effects of product family decisions on
Blomgren, G.E., 2017. The development and future of lithium ion batteries. material selection: a process-based costing approach. Int. J. Prod. Econ. 120,
J. Electrochem. Soc. 164, A5019–A5025. https://doi.org/10.1149/2.0251701jes. 653–668. https://doi.org/10.1016/j.ijpe.2009.04.014.
Bresser, D., Buchholz, D., Moretti, A., Varzi, A., Passerini, S., 2018a. Alternative binders Kaiser, J., Wenzel, V., Nirschl, H., Bitsch, B., Willenbacher, N., Baunach, M., Schmitt, M.,
for sustainable electrochemical energy storage-the transition to aqueous electrode Jaiser, S., Scharfer, P., Schabel, W., 2014. Prozess- und Produktentwicklung von
processing and bio-derived polymers. Energy Environ. Sci. 11, 3096–3127. https:// Elektroden für Li-Ionen-Zellen. Chem. Ing. Tech. 86, 695–706. https://doi.org/
doi.org/10.1039/c8ee00640g. 10.1002/cite.201300085.
Bresser, D., Hosoi, K., Howell, D., Li, H., Zeisel, H., Amine, K., Passerini, S., 2018b. Knoche, T., 2017. Elektrolytbefüllung Prismatischer Lithium-Ionen-Zellen. Technische
Perspectives of automotive battery R&D in China, Germany, Japan, and the USA. Universität München.
J. Power Sources 382, 176–178. https://doi.org/10.1016/j.jpowsour.2018.02.039. Kwade, A., Haselrieder, W., Leithoff, R., Modlinger, A., Dietrich, F., Droeder, K., 2018.
Brodd, R.J., Helou, C., 2013. Cost comparison of producing high-performance Li-ion Current status and challenges for automotive battery production technologies. Nat.
batteries in the U.S. and in China. J. Power Sources 231, 293–300. https://doi.org/ Energy 3, 290–300. https://doi.org/10.1038/s41560-018-0130-3.
10.1016/j.jpowsour.2012.12.048. Kwasi-Effah, C.C., Rabczuk, T., 2018. Dimensional analysis and modelling of energy
Bühler Group, 2017. Efficient Battery Electrode Slurry Production. density of lithium-ion battery. J. Energy Storage 18, 308–315. https://doi.org/
Carvalho, M.S., Kheshgi, H.S., 2000. Low-flow limit in slot coating: theory and 10.1016/j.est.2018.05.002.
experiments. AIChE J. 46, 1907–1917. https://doi.org/10.1002/aic.690461003. Lee, H., Yanilmaz, M., Toprakci, O., Fu, K., Zhang, X., 2014. A review of recent
Cavalieri, S., Maccarrone, P., Pinto, R., 2004. Parametric vs. neural network models for developments in membrane separators for rechargeable lithium-ion batteries. Energy
the estimation of production costs: a case study in the automotive industry. Int. J. Environ. Sci. 7, 3857–3886. https://doi.org/10.1039/C4EE01432D.
Prod. Econ. 91, 165–177. https://doi.org/10.1016/j.ijpe.2003.08.005. Liu, C., Li, F., Ma, L.-P., Cheng, H.-M., 2010. Advanced materials for energy storage. Adv.
Ciez, R.E., Whitacre, J.F., 2017. Comparison between cylindrical and prismatic lithium- Mater. 22, E28–E62. https://doi.org/10.1002/adma.200903328.
ion cell costs using a process based cost model. J. Power Sources 340, 273–281. Mao, C., An, S.J., Meyer, H.M., Li, J., Wood, M., Ruther, R.E., Wood, D.L., 2018.
https://doi.org/10.1016/j.jpowsour.2016.11.054. Balancing formation time and electrochemical performance of high energy lithium-
Cooper, R., Kaplan, R.S., 1988. How Cost Accounting Distorts Product Costs (Manag. ion batteries. J. Power Sources 402, 107–115. https://doi.org/10.1016/j.
Account). jpowsour.2018.09.019.
Curran, R., Raghunathan, S., Price, M., 2004. Review of aerospace engineering cost Michaelis, S., Maiser, E., Kampker, A., Heimes, H.H., Lienemann, C., Wessel, S.,
modelling: the genetic causal approach. Prog. Aero. Sci. 40, 487–534. https://doi. Thielmann, A., Sauer, A., Hettesheimer, T., 2016. Batterie-Produktionsmittel 2030 -
org/10.1016/j.paerosci.2004.10.001. Update 2016. Aachen.
Dreger, H., Bockholt, H., Haselrieder, W., Kwade, A., 2015. Discontinuous and Michaelis, S., Rahimzei, E., Kampker, A., Heimes, H., Lienemann, C., Offermanns, C.,
continuous processing of low-solvent battery slurries for lithium nickel cobalt Kehrer, M., Thielmann, A., Hettesheimer, T., Neef, C., Kwade, A., Haselrieder, W.,
manganese oxide electrodes. J. Electron. Mater. 44, 4434–4443. https://doi.org/ Rahlfs, S., Uerlich, R., Bognar, N., 2018. Roadmap Batterie-Produktionsmittel 2030.
10.1007/s11664-015-3981-4. Mooy, R., 2019. Beitrag zur Produktivitätssteigerung in der Vereinzelung, Positionierung
Du, Z., Rollag, K.M., Li, J., An, S.J., Wood, M., Sheng, Y., Mukherjee, P.P., Daniel, C., und Orientierung von Elektrodenfolien durch eine kontinuierliche
Wood, D.L., 2017. Enabling aqueous processing for crack-free thick electrodes. Materialbewegung. Disseration. Technische Universität Berlin.
J. Power Sources 354, 200–206. https://doi.org/10.1016/j.jpowsour.2017.04.030. Myung, S.T., Maglia, F., Park, K.J., Yoon, C.S., Lamp, P., Kim, S.J., Sun, Y.K., 2017.
Duffner, F., Krätzig, O., Leker, J., 2020a. Battery plant location considering the balance Nickel-rich layered cathode materials for automotive lithium-ion batteries:
between knowledge and cost: a comparative study of the EU-28 countries. J. Clean. achievements and perspectives. ACS Energy Lett 2, 196–223. https://doi.org/
Prod. 121428 https://doi.org/10.1016/j.jclepro.2020.121428. 10.1021/acsenergylett.6b00594.
Duffner, F., Wentker, M., Greenwood, M., Leker, J., 2020b. Battery cost modeling: a Nadeau, M., Kar, A., Roth, R., Kirchain, R., 2010. A dynamic process-based cost modeling
review and directions for future research. Renew. Sustain. Energy Rev. 127, 109872. approach to understand learning effects in manufacturing. Int. J. Prod. Econ. 128,
https://doi.org/10.1016/j.rser.2020.109872. 223–234. https://doi.org/10.1016/j.ijpe.2010.07.016.
Duverlie, P., Castelain, J.M., 1999. Cost estimation during design step: parametric Nelson, P.A., Ahmed, S., Gallagher, K.G., Dees, D.W., 2019. Modeling the Performance
method versus case based reasoning method. Int. J. Adv. Manuf. Technol. 15, and Cost of Lithium-Ion Batteries for Electric-Drive Vehicles. THIRD EDITION
895–906. https://doi.org/10.1007/s001700050147. [WWW Document]. http://www.cse.anl.gov/batpac/download.php (accessed
ECC, 2019. European Construction Costs: Cost Index [WWW Document]. http://construc 5.25.19).
tioncosts.eu/cost-index/ (accessed 5.10.19). Nelson, P.A., Ahmed, S., Gallagher, K.G., Dees, D.W., 2015. Cost savings for
Eftekhari, A., Kim, D.-W., 2018. Sodium-ion batteries: new opportunities beyond energy manufacturing lithium batteries in a flexible plant. J. Power Sources 283, 506–516.
storage by lithium. J. Power Sources 395, 336–348. https://doi.org/10.1016/j. https://doi.org/10.1016/j.jpowsour.2015.02.142.
jpowsour.2018.05.089. Nelson, P.A., Gallagher, K.G., Bloom, I., Dees, D.W., 2012. Modeling the Performance
Eurostat, 2019. Strompreise nach Art des Benutzers [WWW Document]. https://ec.eu and Cost of Lithium-Ion Batteries for Electric-Drive Vehicles [WWW Document],
ropa.eu/eurostat/tgm/table.do?tab=table&init=1&language=de&pcode=ten0011 second ed.. http://www.cse.anl.gov/batpac/files/BatPaC ANL-12_55.pdf (accessed
7&plugin=1 (accessed 5.10.19). 5.8.19).
Farooq, M.A., Roth, R., Kirchain, R., 2018. Lightweighting technologies: analyzing Niazi, A., Dai, J.S., Balabani, S., Seneviratne, L., 2006. Product cost estimation: technique
strategic and economic implications of advanced manufacturing processes. Int. J. classification and methodology review. J. Manuf. Sci. Eng. 128, 563. https://doi.
Prod. Econ. 206, 268–279. https://doi.org/10.1016/j.ijpe.2018.10.003. org/10.1115/1.2137750.
Ficko, M., Drstveň, I., Brezǒ, M., Balǐ, J., Vaupotic, B., 2005. Prediction of total Patry, G., Romagny, A., Martinet, S., Froelich, D., 2015. Cost modeling of lithium-ion
manufacturing costs for stamping tool on the basis of. CAD-model of finished battery cells for automotive applications. Energy Sci. Eng. 3, 71–82. https://doi.org/
product, 165, 1327–1335. https://doi.org/10.1016/j.jmatprotec.2005.02.013. 10.1002/ese3.47.
Field, F., Kirchain, R., Roth, R., 2007. Process cost modeling: strategic engineering and Petri, R., Giebel, T., Zhang, B., Schünemann, J.H., Herrmann, C., 2015. Material cost
economic evaluation of materials technologies. J. Occup. Med. 59, 21–32. https:// model for innovative li-ion battery cells in electric vehicle applications. Int. J. Precis.
doi.org/10.1007/s11837-007-0126-0. Eng. Manuf. - Green Technol. 2, 263–268. https://doi.org/10.1007/s40684-015-
Fuchs, E.R.H., Bruce, E.J., Ram, R.J., Kirchain, R.E., 2006. Process-based cost modeling 0031-x.
of photonics manufacture: the cost competitiveness of monolithic integration of a Placke, T., Kloepsch, R., Dühnen, S., Winter, M., 2017. Lithium ion, lithium metal, and
1550-nm DFB laser and an electroabsorptive modulator on an InP platform. alternative rechargeable battery technologies: the odyssey for high energy density.
J. Lightwave Technol. 24, 3175–3186. https://doi.org/10.1109/JLT.2006.875961. J. Solid State Electrochem. 21, 1939–1964. https://doi.org/10.1007/s10008-017-
H’midaa, F., Martin, P., Vernadata, F., 2006. Cost estimation in mechanical production : 3610-7.
the Cost Entity approach applied to integrated product engineering. Int. J. Prod. Pollet, B.G., Staffell, I., Shang, J.L., 2012. Current status of hybrid, battery and fuel cell
Econ. 103, 17–35. https://doi.org/10.1016/j.ijpe.2005.02.016. electric vehicles: from electrochemistry to market prospects. Electrochim. Acta 84,
Hagen, M., Hanselmann, D., Ahlbrecht, K., Maça, R., Gerber, D., Tübke, J., 2015. 235–249. https://doi.org/10.1016/j.electacta.2012.03.172.
Lithium-sulfur cells: the gap between the state-of-the-art and the requirements for PwC, 2020. Worldwide Tax Summaries Online [WWW Document]. http://taxsummaries.
high energy battery cells. Adv. Energy Mater 5. https://doi.org/10.1002/ pwc.com/ID/tax-summaries-home (accessed 3.4.20).
aenm.201401986. Qian, L., Ben-arieh, D., 2008. Parametric Cost Estimation Based on Activity-Based
Hueber, C., Horejsi, K., Schledjewski, R., Hueber, C., Horejsi, K., Schledjewski, R., 2016. Costing : A Case Study for Design and Development of Rotational Parts 113,
Review of cost estimation : methods and models for aerospace composite pp. 805–818. https://doi.org/10.1016/j.ijpe.2007.08.010.
manufacturing Review of cost estimation : methods and models for aerospace Qian, L., Ben-Arieh, D., 2008. Parametric cost estimation based on activity-based costing:
composite manufacturing. Adv. Manuf. Polym. Compos. Sci. 2, 1–13. https://doi. a case study for design and development of rotational parts. Int. J. Prod. Econ. 113,
org/10.1080/20550340.2016.1154642. 805–818. https://doi.org/10.1016/j.ijpe.2007.08.010.
Ibing, L., Gallasch, T., Schneider, P., Niehoff, P., Hintennach, A., Winter, M., Romero, O.J., Suszynski, W.J., Scriven, L.E., Carvalho, M.S., 2004. Low-flow limit in slot
Schappacher, F.M., 2019. Towards water based ultra-thick Li ion battery electrodes – coating of dilute solutions of high molecular weight polymer. J. Nonnewton. Fluid
a binder approach. J. Power Sources 423, 183–191. https://doi.org/10.1016/j. Mech. 118, 137–156. https://doi.org/10.1016/j.jnnfm.2004.03.004.
jpowsour.2019.03.020. Rush, C., Roy, R., 2001. Expert judgement in cost estimating: modelling the reasoning
International Society of Parametric Analysts, 2018. Parametric Estimating Handbook, process. Concurr. Eng. 9, 271–284. https://doi.org/10.1177/
fourth ed. 1063293X0100900404.

18
F. Duffner et al. International Journal of Production Economics 232 (2021) 107982

Sakti, A., Azevedo, M.L., Fuchs, E.R.H., Michalek, J.J., Gallagher, K.G., Whitacre, J.F., Turner and Townsend, 2018. International Construction Market Survey [WWW
2017. Consistency and robustness of forecasting for emerging technologies: the case Document]. Int. Constr. Mark. Surv. 2018. https://www.turnerandtownsend.com/m
of Li-ion batteries for electric vehicles. Energy Pol. 106, 415–426. https://doi.org/ edia/3352/international-construction-market-survey-2018.pdf (accessed 12.10.18).
10.1016/j.enpol.2017.03.063. Vaalma, C., Buchholz, D., Weil, M., Passerini, S., 2018. A cost and resource analysis of
Sakti, A., Michalek, J.J., Fuchs, E.R.H., Whitacre, J.F., 2015. A techno-economic analysis sodium-ion batteries. Nat. Rev. Mater. 3, 18013. https://doi.org/10.1038/
and optimization of Li-ion batteries for light-duty passenger vehicle electrification. natrevmats.2018.13.
J. Power Sources 273, 966–980. https://doi.org/10.1016/j.jpowsour.2014.09.078. Verlinden, B., Duflou, J.R., Collin, P., Cattrysse, D., 2008. Cost estimation for sheet metal
Schmitt, M., 2015. Slot die coating of lithium-ion battery electrodes. Karlsruhe Institute parts using multiple regression and artificial neural networks: a case study. Int. J.
of Technology. https://doi.org/10.5445/KSP/000051733. Prod. Econ. 111, 484–492. https://doi.org/10.1016/j.ijpe.2007.02.004.
Schmitt, M., Scharfer, P., Schabel, W., 2014. Slot die coating of lithium-ion battery Verma, P., Maire, P., Novák, P., 2010. A review of the features and analyses of the solid
electrodes: investigations on edge effect issues for stripe and pattern coatings. electrolyte interphase in Li-ion batteries. Electrochim. Acta 55, 6332–6341. https://
J. Coating Technol. Res. 11, 57–63. https://doi.org/10.1007/s11998-013-9498-y. doi.org/10.1016/j.electacta.2010.05.072.
Schmuch, R., Wagner, R., Hörpel, G., Placke, T., Winter, M., 2018. Materials for Wagner, R., Preschitschek, N., Passerini, S., Leker, J., Winter, M., 2013. Current research
automotive Batteries : perspective on performance and cost of lithium-based trends and prospects among the various materials and designs used in lithium-based
rechargeable batteries. Nat. Energy 3. https://doi.org/10.1038/s41560-018-0107-2. batteries. J. Appl. Electrochem. 43, 481–496. https://doi.org/10.1007/s10800-013-
Schnell, J., Knörzer, H., Imbsweiler, A.J., Reinhart, G., 2020. Solid versus liquid—a 0533-6.
bottom-up calculation model to analyze the manufacturing cost of future high- Wang, Q., 2007. Artificial neural networks as cost engineering methods in a collaborative
energy batteries. Energy Technol. 8, 1901237. https://doi.org/10.1002/ manufacturing environment. Int. J. Prod. Econ. 109, 53–64. https://doi.org/
ente.201901237. 10.1016/j.ijpe.2006.11.006.
Schnell, J., Reinhart, G., 2016. Quality management for battery production: a quality Wang, X., Ding, Y., Deng, Y., Chen, Z., 2020. Ni-Rich/Co-Poor layered cathode for
gate concept. Procedia CIRP 57, 568–573. https://doi.org/10.1016/j. automotive Li-ion batteries: promises and challenges. Adv. Energy Mater. 10,
procir.2016.11.098. 1903864. https://doi.org/10.1002/aenm.201903864.
Schulze, M., Seuring, S., Ewering, C., 2012. Int . J . Production Economics Applying Wentker, M., Greenwood, M., Leker, J., 2019. A bottom-up approach to lithium-ion
activity-based costing in a supply chain environment. Int. J. Prod. Econ. 135, battery cost modeling with a focus on cathode active materials. Energies 12, 504.
716–725. https://doi.org/10.1016/j.ijpe.2011.10.005. https://doi.org/10.3390/en12030504.
Schünemann, J.H., 2015. Modell zur Bewertung der Herstellkosten von Whittingham, M.S., 2004. Lithium batteries and cathode materials. Chem. Rev. 104,
Lithiumionenbatteriezellen. Technische Universität Braunschweig. 4271–4302. https://doi.org/10.1021/cr020731c.
Scrosati, B., Garche, J., 2010. Lithium batteries: status, prospects and future. J. Power Winter, M., 2009. The solid electrolyte interphase – the most important and the least
Sources 195, 2419–2430. https://doi.org/10.1016/j.jpowsour.2009.11.048. understood solid electrolyte in rechargeable Li batteries. Z. Phys. Chem. 223,
Scrosati, B., Garche, J., Tillmetz, W., 2015. Advances in Battery Technologies for Electric 1395–1406. https://doi.org/10.1524/zpch.2009.6086.
Vehicles. Elsevier. Winter, M., Barnett, B., Xu, K., 2018. Before Li ion batteries. Chem. Rev. 118,
Shehab, E., Abdalla, H., 2001. Manufacturing cost modelling for concurrent product 11433–11456. https://doi.org/10.1021/acs.chemrev.8b00422.
development. Robot. Comput. Integrated Manuf. 17, 341–353. https://doi.org/ Wood, D.L., Li, J., Daniel, C., 2015. Prospects for reducing the processing cost of lithium
10.1016/S0736-5845(01)00009-6. ion batteries. J. Power Sources 275, 234–242. https://doi.org/10.1016/j.
Shehab, E.M., Abdalla, H.S., 2002. A design to cost system for innovative product jpowsour.2014.11.019.
development. Proc. Inst. Mech. Eng. Part B J. Eng. Manuf. 216, 999–1019. https:// Wu, G., Inderbitzin, A., Bening, C., 2015. Total cost of ownership of electric vehicles
doi.org/10.1243/09544050260174201. compared to conventional vehicles: a probabilistic analysis and projection across
Sierzchula, W., Bakker, S., Maat, K., Van Wee, B., 2014. The influence of financial market segments. Energy Pol. 80, 196–214. https://doi.org/10.1016/j.
incentives and other socio-economic factors on electric vehicle adoption. Energy Pol. enpol.2015.02.004.
68, 183–194. https://doi.org/10.1016/j.enpol.2014.01.043. Xu, K., 2014. Electrolytes and interphases in Li-ion batteries and beyond. Chem. Rev.
Singh, M., Kaiser, J., Hahn, H., 2016. A systematic study of thick electrodes for high 114, 11503–11618. https://doi.org/10.1021/cr500003w.
energy lithium ion batteries. J. Electroanal. Chem. 782, 245–249. https://doi.org/ Yoshio, M., Brodd, R.J., Kozawa, A., 2009. Lithium-Ion Batteries. Springer New York,
10.1016/j.jelechem.2016.10.040. New York, NY. https://doi.org/10.1007/978-0-387-34445-4.
Tagawa, K., Brodd, R.J., 2009. Production Processes for Fabrication of Lithium-Ion Zhang, X., Xie, H., Kim, C.-S., Zaghib, K., Mauger, A., Julien, C.M., 2017. Advances in
Batteries, pp. 181–194. https://doi.org/10.1007/978-0-387-34445-4. lithium—sulfur batteries. Mater. Sci. Eng. R Rep. 121, 1–29. https://doi.org/
Tan, P., Jiang, H.R., Zhu, X.B., An, L., Jung, C.Y., Wu, M.C., Shi, L., Shyy, W., Zhao, T.S., 10.1016/j.mser.2017.09.001.
2017. Advances and challenges in lithium-air batteries. Appl. Energy 204, 780–806. Zhang, Y.F., Fuh, J.Y.H., Chan, W.T., 1996. Feature-based cost estimation for packaging
https://doi.org/10.1016/j.apenergy.2017.07.054. products using neural networks. Comput. Ind. 32, 95–113. https://doi.org/10.1016/
Turetskyy, A., Thiede, S., Thomitzek, M., von Drachenfels, N., Pape, T., Herrmann, C., S0166-3615(96)00059-0.
2020. Toward data-driven applications in lithium-ion battery cell manufacturing. Zheng, H., Li, J., Song, X., Liu, G., Battaglia, V.S., 2012. A comprehensive understanding
Energy Technol. 8, 1900136. https://doi.org/10.1002/ente.201900136. of electrode thickness effects on the electrochemical performances of Li-ion battery
cathodes. Electrochim. Acta 71, 258–265. https://doi.org/10.1016/j.
electacta.2012.03.161.

19

View publication stats