Addressing the Scientific Gaps Between Life Cycle Thinking and Multi-Criteria Decision Analysis for the Sustainability Assessment of Electric Vehicles’ Lithium-Ion Batteries

Abstract

Electric vehicles can substantially lower the overall carbon footprint of the transportation sector, and their batteries become key enablers of widespread electrification. Although high capacity and efficiency are essential for providing sufficient range and performance in electric vehicles, they can be compromised by the need to lower costs and environmental impacts and retain valuable materials. In the present work, multi-criteria decision analysis was adopted to assess the sustainability of different lithium-ion batteries. Life cycle carbon emissions and toxicity, material criticality, life cycle costs, specific energy, safety, and durability were considered in the analysis as key parameters of the transition to electric mobility. A subjective approach was chosen for the weight attribution of the criteria. Although certain alternatives, like lithium nickel cobalt manganese oxide (NCM) and lithium nickel cobalt aluminum oxide (NCA), outweigh others in specific energy, they lack in terms of safety, material preservation, and environmental impact. Addressing cost-related challenges is also important for making certain solutions competitive and largely accessible. Overall, while technical parameters are crucial for the development of lithium-ion batteries, it is equally important to consider the environmental burden, resource availability, and economic factors in the design process, alongside social aspects such as the ethical sourcing of materials to ensure their sustainability.

1. Introduction

As urban areas face significant challenges due to environmental pollution and poor air quality largely driven by the transportation sector, the European Union has set ambitious targets for emissions reduction. Electrifying the transportation sector is a major area of research and innovation for mitigating these emissions. Since electric vehicles (EVs) have no direct tailpipe emissions, when charged with electricity from renewable sources, they can substantially lower the overall carbon footprint of transportation. High-capacity, efficient batteries are essential for EVs to provide sufficient range and performance, making them viable alternatives to conventional vehicles [1]. As battery costs decrease and their efficiency improves, they become key enablers of widespread electrification, helping to reduce reliance on fossil fuels and support a more sustainable energy future [2]. Assessing lithium-ion batteries (LIBs) based on sustainability criteria is crucial due to their growing role. First of all, as the demand for EVs increases, so does the need for large-scale battery production, which can severely impact the environment and lead to a scarcity of valuable resources [3,4]. In addition, when considering the expansion of LIBs, it is essential to examine technical parameters, as advancements in this field enable improved performance, efficiency, and safety. Economic factors drive affordability and scalability and should then be included to ensure a holistic assessment. By evaluating LIBs against sustainability criteria, their long-term viability can be assured, minimizing negative environmental impacts and promoting responsible production practices that support a circular economy [5].

In the current study, five alternative LIBs were assessed against sustainability criteria using multi-criteria decision analysis (MCDA). MCDA is a tool for finding the best possible solution or ranking of alternatives in complex decision-making problems in which different quantitative and qualitative criteria should be considered. MCDA has been applied in multiple areas of expertise ranging from economics and engineering to medicine. To date, several MCDA methods have been developed that differentiate in the complexity of algorithms, the methods for criteria weighting, the representation of preferences evaluation criteria, data aggregation type, etc. [6]. In the present study, the ranking of the alternatives was performed through two MCDA methods. The first is the preference ranking organization method for the enrichment of evaluations (PROMETHEE), which is a widely used outranking decision-making method for solving complex MCDA problems. There are several modules of PROMETHEE. PROMETHEE I provides a partial ranking of the alternatives by comparing them in pairs and determining a preference index for each pair based on the weights of the criteria. PROMETHEE II extends it by providing a complete ranking of the alternatives [7]. PROMETHEE III offers ranking based on intervals, while PROMETHEE IV represents the continuous case. PROMETHEE V extends the method to cases where a subset of alternatives under a set of constraints is needed. Finally, PROMETHEE GAIA allows for the graphical display of the alternatives and the criteria [8,9]. In general, PROMETHEE has the advantage of being more robust by avoiding inversion compared to other outranking methods, such as ELimination Et Choix Traduisant la REalité (ELECTRE). In addition, it consists of simple calculations, has flexibility in the preference function selection, and has a strong adaptability to the decision environment [10]. The second MCDA method for the ranking of the alternatives is the Analytic Hierarchy Process (AHP), which organizes a problem into a hierarchical structure and uses pairwise comparisons to rate the different options on a scale from 1 to 9 [11].

Four categories of criteria were selected to assess the different LIB technologies. Each category aligns with the central pillars of sustainability, ensuring a comprehensive approach. The first category of criteria is linked to the environmental pillar and consists of two sub-criteria: life cycle greenhouse gas emissions and toxicity. While the majority of studies rely on literature data to assess environmental aspects, the current study employs the Life Cycle Assessment (LCA) to model environmental impacts. For this purpose, global warming potential (GWP) and toxicity are quantified for the total life cycle of a battery pack, from raw material extraction to recycling and final disposal. Toxicity is considered not only in terms of the toxic components within the batteries but also across the entire life cycle, encompassing all stages of its life. Studying material criticality is equally essential due to the reliance of LIBs on specific raw materials, such as lithium, cobalt, and nickel. Understanding such issues allows for better management of supply chains, promotes the development of alternative materials, and encourages recycling and reuse strategies. Additionally, it supports the creation of policies that can mitigate the risks of material shortages [12]. For the economic pillar, the total cost of ownership and the residual value of the batteries at the End-of-Life (EoL) stage were evaluated. Finally, as the technical performance of batteries plays a vital role in their expansion, specific energy, lifetime, and safety features were considered in the analysis. Five LIBs were identified as the alternatives for the analysis: lithium nickel cobalt manganese oxide (NCM), lithium iron phosphate (LFP), lithium nickel cobalt aluminum oxide (NCA), lithium manganese oxide (LMO), and lithium cobalt oxide (LCO).

The weights of criteria in MCDA problems play a significant role in the assessment of the alternatives and thus major attention should be paid to the weighting method. There are three basic concepts for weighting the criteria: subjective, objective, and hybrid. The subjective approach depends on the experience and knowledge of experts and specialists with extensive expertise in the subject under study. Experts often use pair-based comparisons to determine the weight of each criterion [10,13]. In the present analysis, a subjective approach was adopted to determine the weights of the criteria. A questionnaire of pairwise comparisons was handed out to a pool of experts with extensive expertise in electric mobility applications, adopting the AHP model with a scale from 1 to 9. AHP was selected as it can potentially offer experts a better understanding of the problem by breaking it down into smaller ones and by enabling pairwise comparisons. The questionnaire consisted of pairwise comparisons among the categories of criteria and their sub-criteria.

Few studies have assessed batteries for EVs following a multi-criteria approach. Loganathan et al. [14], used a weighted sum model to evaluate several alternatives based on specific power, energy density, price, reliability, and safety. Ecer et al. [15] used the Borda Count and Copeland methods to rank different EVs, incorporating criteria linked to the performance of the battery. Battery capacity, full charge time, and quick charge time were considered in their study. Liaqat et al. [16] evaluated energy storage technologies for EVs. The evaluation was based on different criteria that included costs, technical properties, compatibility, technological maturity, environment, health, and safety. These criteria and their sub-criteria were used to create an AHP model. In their work, the alternatives included different battery types (LIBs, lead acid, vanadium redox flow, sodium–nickel chloride, nickel metal hydride, lithium–sulfur, nickel–cadmium), supercapacitors, hydrogen fuel cell storage, flywheel storage, and superconducting magnetic energy storage. Tajik et al. [13] developed a hybrid MCDA method based on subjective, objective, and combined weighting methods, including Simple Additive Weighting, Criterion Impact Loss, the Technique for Order Preference by Similarity to Ideal Solution, the Measurement of Alternatives and Ranking according to Compromise Solution, Additive Ratio Assessment, and the Combined Compromise Solution to assess different cathode materials for LIBs. The cathode materials assessed were LCO, lithium nickel oxide, NCA, NCM, LMO, and LFP. The performance of the alternatives was evaluated through the Final Ranking of Alternatives and the Copeland method. The Data Envelopment Analysis model was used in their study to evaluate the performance of the different cathode materials. The alternatives were assessed against nine criteria: cell voltage, power, energy density, safety, lifetime, discharge capacity, stability, cost, and toxic components. Baars et al. [17] proposed an integrated model and framework to assess the environmental, technical, and socio-economic aspects of LIBs for EVs. The methodology was used to calculate how material design strategies can potentially improve the cost, carbon footprint, material criticality, and energy density of the batteries. In their study, Multi-Objective Optimization was the proposed method to solve the model. Niu et al. [10] conducted a comprehensive evaluation of different EVs from a consumer perspective, based on PROMETHEE II. The weighting methods included large-scale group decision-making and the entropy-based method. In their study, parameters related to the battery part of the EVs, like driving range, charging time, battery capacity, and energy consumption per kilometer, were evaluated. Correlations between the criteria were also identified.

This study takes a little step forward in the existing knowledge by combining LCA and MCDA to assess different battery technologies for EVs, integrating a group decision-making approach driven by expert insights. The use of LCA ensures full comparability of the environmental impacts linked to different alternatives, by establishing uniform system boundaries, functional units, impact assessment methods, and assumptions.

2. Materials and Methods

2.1. Selection of Alternatives

As the market of electric mobility is under a dynamic shift in battery technologies, the selection of battery types that will serve as the alternatives in the study was guided by their current level of use in EVs and the expectation that they will remain integral to these applications in the future or even their potential use in EVs. As the demand for higher-performance and safer batteries grows, assessing all potential chemistries ensures a comprehensive understanding of the best options for different segments of the EV market [18]. Five distinct LIB technologies were identified as the focus of this study.

The NCM cathode is a leading choice for EVs, primarily due to its high specific energy. It offers a well-balanced combination of cost, safety, and performance, making NCM cathodes suitable for a broad spectrum of EVs in the mainstream market segment. The stoichiometry of nickel, cobalt, and manganese can be adjusted to meet the specific demands of various applications.

LFP batteries combine a long lifespan and a high level of safety. Despite their relatively low specific energy, they are particularly dominant in the market segment for affordable vehicles, where mass production is the key. Moreover, these batteries are suitable candidates for off-grid and grid-connected power supply systems [18].

NCA batteries, due to their high specific energy and power density, deliver long-range and superior performances, making them ideal for high-performing luxury vehicles in the premium segment. They are typically composed of 80% nickel, 15% cobalt, and 5% aluminum. Particularly, the NCA-955 variant, characterized by its low cobalt content, is widely used in vehicles that achieve ranges of over 400 km.

LMO cathodes, often combined with NCM, are another noteworthy technology in the EV industry. They are known for their high thermal stability and thus safety, which is a critical factor for mobility applications. Although they have low specific energy, they are cost-effective and have fast-charging capabilities, making them an attractive option for certain EV applications, particularly in urban environments where range requirements are lower. Assessing LMO batteries is essential for understanding their potential contribution to hybrid battery systems where different chemistries are combined to optimize overall performance [18,19].

LCO batteries, despite being one of the older lithium-ion chemistries, remain relevant in the context of electric mobility due to their high energy density and proven performance in a range of products. Assessing LCO batteries for EV applications is essential as they offer significant advantages in terms of range and efficiency of the vehicle. Additionally, understanding the limitations of LCO batteries, such as their relatively lower thermal stability and higher cost, is crucial in evaluating their feasibility in the evolving market of EVs.

2.2. Selection of Criteria

2.2.1. Technical Criteria

The criteria used in the present analysis are depicted in Figure 1. As efforts to achieve carbon neutrality increase, replacing internal combustion engine vehicles with EVs presents significant challenges for battery technology, particularly in terms of energy content, safety, and longevity. High specific energy expressed in Wh stored per mass of battery cell is essential for extending the driving range [20]. The specific energy for each of the alternatives was calculated based on the general methodology proposed by Son et al. [21], as described by Equations (1)–(3).

Mass loading of cathode = Material density·Material thickness

(1)

Cell capacity = Area of electrode·Mass loading of cathode·Active material ratio·Gravimetric capacity·Number of layers

(2)

Specific energy = Cell capacity·Nominal voltage/Total weight of the cell

(3)

A pouch configuration was the selected format for the cells. Each cell consists of 8 pairs of cathode and anode layers, as presented in Figure 2. Graphite was considered for the anode. The data for gravimetric capacities of the cathode materials and nominal voltages are provided in

Table 1. Gravimetric capacity and nominal voltage of the alternative LIB technologies [20,21,22].

Cathode	Gravimetric Capacity (mAh/g)	Nominal Voltage (V)
NCM	175	3.8
LFP	150	3.3
NCA	160	3.6
LMO	100	3.8
LCO	165	3.8

Abbreviations: LCO: lithium cobalt oxide; LFP: lithium iron phosphate; LMO: lithium manganese oxide; NCA: lithium nickel cobalt aluminum oxide; NCM: lithium nickel cobalt manganese oxide.

. The full methodology adopted for specific energy calculation and the data used are provided in the Supplementary Materials.

Moreover, ensuring the safety of batteries is critical since it directly impacts both personal and property protection, but should also not compromise energy storage capabilities. The safety of LIBs is highly influenced by the chemistry of the cathode material and its properties and is indicated by the thermal runaway temperature. Thermal runaway represents the most severe safety concern for LIBs. The phenomenon arises from side reactions involving electrolyte, cathode, and anode reactions, electrode surface interactions, and lithium plating. Such reactions are typically triggered by mechanical, thermal, or electrical stress. A thermal runaway is primarily caused by the failure of the separator and the release of oxygen from the cathode [23].

Long lifespan is another crucial aspect as it may lower the cost of maintenance and replacement [20]. Lifetime data are expressed in full cycles of charge and discharge. The life of LIBs in EVs practically reaches its end when the capacity drops to 80% of the initial value. Even if recent studies suggest more flexible limits, in the current analysis, the data gathered correspond to 80% state of health to ensure the satisfaction of driver mobility [24].

2.2.2. Environmental Criteria

Understanding the environmental impacts of LIBs through a lifecycle perspective is critically important in the broader context of sustainability. As EVs are promoted as a solution for the reduction in greenhouse gas emissions from transportation, it is essential to evaluate the full lifecycle impact of their batteries. GWP allows us to quantify the greenhouse gas emissions associated with battery production, use, and disposal, providing a clear picture of the actual environmental benefits of EVs compared to conventional vehicles. This information is crucial for identifying areas where improvements can be made in battery technology and manufacturing processes to further minimize their carbon footprint. Additionally, understanding the GWP of EV batteries supports informed decision-making by policymakers, industry stakeholders, and consumers, ensuring that the shift to electric mobility truly contributes to global efforts to mitigate climate change [25].

In addition, assessing the toxicity of LIBs is crucial for understanding their full environmental impact. Toxicity is primarily linked to the manufacturing processes and depends strongly on the materials used. Nickel and cobalt, in particular, lead to higher toxicity in mining, production, and EoL stages. Toxic waste may also be discharged during the manufacturing of the cathode, anode, collector, and battery management system [2,26]. The inappropriate management of EoL batteries may also lead to aquatic toxicity [27]. Understanding these toxic effects is essential for developing safer battery technologies, improving recycling processes, and formatting regulations accordingly.

These two impact categories serve as the environmental criteria in the present analysis and were quantified for all alternative technologies through LCA. The system boundaries of the analysis were determined through a cradle-to-grave approach involving all stages of the life cycle from material mining to EoL management. The functional unit was set to 1 kWh of energy delivered by the battery pack. ReCiPe 2016 Midpoint (Hierarchist approach) was the chosen method for life cycle impact assessment. OpenLCA software (version 1.11.0) was used for the implementation of LCA.

ReCiPe methodology uses characterization factors that translate the emitted pollutants into potential damage to ecosystems. Toxicity in the ReCiPe Midpoint methodology is assessed across three distinct levels: marine water, freshwater, and terrestrial, measured in 1,4-Dichlorobenzene (DCB) equivalent points. In order to combine the damage in these three levels and acquire a single indicator for ecotoxicity, the impacts on marine water, freshwater, and terrestrial were unified. For this shift, the midpoint-to-endpoint conversion factors provided by the ReCiPe method, as shown in

Table 2. Midpoint-to-endpoint conversion factors of ReCiPe method [28].

Midpoint	Conversion Factor
Freshwater ecotoxicity	6.95 × 10−10 species·yr/kg 1,4-DCB * eq
Marine ecotoxicity	1.05 × 10−10 species·yr/kg 1,4-DCBeq
Terrestrial ecotoxicity	5.39 × 10−8 species·yr/kg 1,4-DCBeq

* Dichlorobenzene.

, were applied. Specifically, the midpoint impacts were multiplied by their respective conversion factors and summed up to obtain the total ecotoxicity [28].

Background data for the LCA were acquired through the Ecoinvent database [29]. The life cycle inventory for the stages of production and recycling was based on literature data [25,26,30,31,32,33,34,35,36] (see Tables S4–S15 in the Supplementary Materials). The use phase was modeled for two driving cycles to represent typical vehicle driving. The Urban Dynamometer Driving Schedule (UDDS), commonly referred to as the “LA4” or “the city test”, is representative of light-duty vehicle testing in city driving conditions and the Highway Fuel Economy Test (HWFET) represents highway driving conditions [37]. City conditions were assumed for 55% of the driving range and the rest of the cycle was assumed to be under highway conditions. The methodology was adapted from Deng et al. [38] to acquire the required energy and battery mass for one full cycle of charge and discharge. For the calculation of the required battery mass for one full cycle, the specific energy as calculated through Equations (1)–(3) was applied. For the conversion of the specific energy from cell to pack level, a coefficient of 0.77 was used [39]. An extended description of the methodology is provided in the Supplementary Materials.

2.2.3. Economic Criteria

As the EV market grows, minimizing the total cost of LIBs becomes crucial for making them more accessible. The total cost can be broken down into capital expenditure (CAPEX) and operational expenditure (OPEX). CAPEX refers to the initial investment required for battery production and integration into EVs, while OPEX encompasses the ongoing costs associated with battery charging, maintenance, and replacement if needed. Understanding these economic factors is essential for both manufacturers and consumers, as they directly influence the affordability and overall cost-effectiveness of EVs [40].

The estimation of CAPEX was conducted through the cost analysis tool offered by the openLCA software and Ecoinvent database. The estimation of OPEX per kWh delivered by the battery pack includes the calculation of the energy that is required for charging the vehicle. The required energy was calculated based on the methodology explained in Section 2.2.2. The cost for vehicle charging was calculated for the Greek electricity cost per kWh [41].

The remaining value of the batteries at their EoL was considered in the analysis as a factor for economic value recovery. Batteries with significant residual value can be recycled or repurposed, recovering valuable materials such as lithium, cobalt, and nickel. This can offset the costs of new battery production. Additionally, EoL batteries with sufficient capacity can be repurposed for less demanding applications, such as stationary energy storage, extending their useful life [36]. In the present study, the residual value is represented by the cost that recyclers are required to pay to obtain the batteries and is presented in

Table 3. Residual value of the alternatives at EoL [42].

Alt.	Residual Value (€/kg)
NCM	0
LFP	−2.09
NCA	0
LMO	−0.95
LCO	1.9

Abbreviations: LCO: lithium cobalt oxide; LFP: lithium iron phosphate; LMO: lithium manganese oxide; NCA: lithium nickel cobalt aluminum oxide; NCM: lithium nickel cobalt manganese oxide.

. This cost depends on the battery’s chemistry and is, in general, higher for batteries with a higher cobalt content. Cases where recyclers are paid to recycle certain battery types are indicated by negative values.

2.2.4. Material Criticality

As the demand for EVs continues to rise, addressing material criticality is crucial for ensuring that battery production can keep pace with market growth without causing environmental degradation or economic instability [12]. The material criticality of the available LIBs was assessed by adapting the methodology of Wentker et al. [19]. In their study, a single-supply risk score was formulated for each element. The first indicator involved is supply reduction expressed by reserve and resource depletion time and recycling rate. The second indicator considers the risk of an increase in demand for a certain technology by examining by-product dependency, technology demand in the future, and substitutability of materials. Market concentration is evaluated by means of country concentration and global supply concentration. The fourth indicator refers to political stability, policy perception, and general political regulation in each country, as perceived by mining corporations [19].

According to their study, manganese and nickel have the strongest concerns for reserve depletion time, followed by cobalt and iron, based on the current knowledge of available reserves and rates of extraction. Nickel is also critical in terms of the availability of resources. Recycling risk is severe for phosphorus and lithium. By-product dependency is a key issue for cobalt and lithium as they have a very limited supply. Substitutability is a major bottleneck for almost all elements of LIBs. In particular, batteries without cobalt have lower energy density and reduced cyclic stability, making them impractical for EV applications, due to their increased size and weight. In terms of market concentration, cobalt, lithium, and phosphorus exhibit the most concentrated market structures. In addition, cobalt and phosphorus are more dependent on regions with elevated political risks. While political regulation risk is relatively low for most elements, cobalt stands out with the highest policy perception risk.

2.3. Determination of the Weights of Criteria

A questionnaire of pairwise comparisons, as presented in Figure 3, was handed out to a pool of experts on electric mobility in order to determine the weights of each criterion. The pool of experts was built to encompass professionals from a diverse range of sectors critical to the sustainability of electric mobility, including LCA, energy policy, transportation infrastructure, battery technology, sustainable transportation, automotive industry, and energy systems. This diverse representation allows for a holistic approach to addressing sustainability concerns across the life cycle of EV batteries.

Following the collection of the answers, the decision matrix $A_{ij}$ was structured as described by Equation (4):

$$A_{ij}=\left(\begin{array}{ccc}{a}_{11}& \cdots & a_{1j}\\ ⋮& \ddots & ⋮\\ a_{i1}& \cdots & a_{ij}\end{array}\right)$$

(4)

where $a_{ij}$ is the importance of criterion $i$ against criterion $j$. In matrix $A_{ij}$, the elements above the diagonal represent direct comparisons, and the elements below the diagonal are their reciprocals, ensuring consistency in the matrix [13]. The same process is followed for all the responses in the questionnaire, resulting in a number of decision matrixes equal to the number of experts. The average importance of each criterion was then calculated by applying the aggregation of individual judgments, as described in Figure 4, and using the geometric mean value $B_{ij}$, given by Equation (5):

$$B_{ij}=\sqrt[n]{a_{ij1}·a_{ij2}\dots ..{·a}_{ijn}}$$

(5)

where $a_{ijn}$ is the importance of criterion $i$ against criterion $j$ according to expert $n$ and $n$ is the number of experts.

As inconsistencies may arise in the context of pairwise comparisons, the consistency rate (CR) of matrix $A_{ij}$ is calculated, through Equation (6):

$$CR={\displaystyle \frac{CI}{RI}}$$

(6)

where $RI$ is a random consistency index related to the dimension of matrix $A_{ij}$ and is extracted from

Table A1. Values of Random Consistency Index [43].

Dimension	RI
1	0
2	0
3	0.5799
4	0.8921
5	1.1159
6	1.2358
7	1.3322
8	1.3952
9	1.4537
10	1.4882

and $CI$ is the consistency index calculated through Equation (7):

$$C{\rm I}={\displaystyle \frac{{\lambda }_{max}\left(A_{ij}\right)-c}{c-1}}$$

(7)

where ${\lambda }_{max}$ is the principal eigenvalue of matrix $A_{ij}$ and $c$ is the number of criteria.

The lower the $CR$, the more consistent the pairwise comparison matrices are considered. In general, if  $CR$ ≤ 0.1, then  $A_{ij}$ is acceptably consistent. In other cases, the pairwise comparison matrix needs to be readjusted, or the transitivity rule may be enforced [43]. The transitivity rule states that if criterion A is preferred over criterion B, and B is preferred over criterion C, then criterion A should logically be preferred over C. In pairwise comparisons, transitivity ensures consistency in the preferences expressed by decision-makers by reducing the comparisons, leading to more reliable outcomes.

Finally, the weights originally assigned on a scale from 1 to 9 were converted to absolute values for the PROMETHEE method. These values are the elements of the normalized eigenvector of matrix $A_{ij}$.

2.4. Ranking of the Alternatives

For the ranking of the alternatives, two MCDA methods, AHP and PROMETHEE, were deployed to enhance the robustness and reliability of the results and provide a more comprehensive understanding of the problem. Analytic Hierarchy Process software by SpiceLogic (version 4.2.6) [44] and Visual PROMETHEE software (version 1.4.0.0, Academic Edition) [45] were employed for the analysis.

2.4.1. AHP Methodology

In AHP, in order to structure the decision matrix $A_{km}$with the scores of the alternatives, all values acquired in the data collection stage were converted to a Saaty’s scale from 1 to 9, through Equation (8):

$$W_{1-9, k}=\left(\left(W_k-W_{\mathrm{m}\mathrm{i}\mathrm{n}}\right){\displaystyle \frac{8}{\left(W_{\mathrm{m}\mathrm{a}\mathrm{x}}-W_{\mathrm{m}\mathrm{i}\mathrm{n}}\right)}}\right)+1,$$

(8)

where $W_{1-9, k}$ is the score of alternative $k$ on a scale from 1 to 9, $W_k$ is the absolute score of the alternative, and $W_{\mathrm{m}\mathrm{i}\mathrm{n}}$ and $W_{\mathrm{m}\mathrm{a}\mathrm{x}}$, are the minimum and maximum values among all alternatives against a certain criterion. In cases where a minimum value responds to the optimal option, values $W_{\mathrm{m}\mathrm{i}\mathrm{n}}$ and $W_{\mathrm{m}\mathrm{a}\mathrm{x}}$ are reversed.

The pairwise comparison matrix $A_{km}$ is structured for the alternatives through Equation (9):

$$A_{km}=\left(\begin{array}{ccc}{a}_{11}& \cdots & a_{1\mathrm{m}}\\ ⋮& \ddots & ⋮\\ a_{\mathrm{k}1}& \cdots & a_{km}\end{array}\right)$$

(9)

where $a_{km}$ is the score of alternative $k$ against $m$ and $a_{mk}$ is the score of alternative $m$ against $k$. The elements above the diagonal represent direct comparisons, and the elements below the diagonal are their reciprocals, ensuring consistency in the matrix [13]. The eigenvalue method is then used to obtain the priority vectors of the alternatives. Finally, the combination of the criteria weights and scores is achieved through the weighted sum model. For each criterion, the score of each alternative is multiplied by the criterion’s weight, and products are summed to obtain the total score, by which the alternatives are ultimately ranked.

2.4.2. RPOMETHEE Methodology

In PROMETHEE I, after the alternatives and criteria are selected, a preference function $P(a_k,a_m)$ that expresses the preference of alternative $a_k$ over alternative $a_m$ shall be selected. The preference function chosen for the selected criteria is the V-shape, in which preference increases gradually with the difference. The preference function $P(a_k,a_m)$ describing the V-shape is given by Equation (10).

$$P(a_k, a_m)=\left\{\begin{array}{c}0, d\le 0\hfill \\ {\displaystyle \frac{d}{p}}, 0<d\le p\hfill \\ 1, d>p\hfill \end{array}\right.$$

(10)

where $d$ is the deviation between the performance of alternative $a_k$ over alternative $a_m$ on a particular criterion and $p$ is the preference threshold which indicates the difference at which one alternative is strictly preferred over another. The larger the function value is, the bigger the difference between alternatives. In particular, when $P\left(a_k,a_m\right)=0$, then $a_k \mathrm{and} a_m$ are indifferent, while if $P\left(a_k,a_m\right)=1$, then $a_k$ is strictly preferential to $a_m$.

For each pair of alternatives $a_k,a_m$ and for each criterion $j$, the preference index ${\pi }_j(a_k,a_m)$ is then calculated. The preference index is a weighted sum of the preference functions, reflecting the importance of each criterion:

$${\pi }_j(a_k,a_m)=w_j·P_j\left(a_k,a_m\right)$$

(11)

where $w_j$ is the weight of criterion $j$ and $P_j\left(a_k,a_m\right)$ is the result of the preference function between alternatives $a_k$ and $a_m$ for criterion $j$. The aggregated preference index $\pi (a_k,a_m)$ for each pair of alternatives is then calculated by summing up the weighted preference indices across all criteria:

$$\pi \left(a_k,a_m\right)=\sum _{j=1}^c{w}_j·P_j(a_k,a_m)$$

(12)

The positive flow ${\phi }^{+}\left(a_k\right)$, which measures how much alternative $a_k$ is preferred over all other alternatives and the negative flow ${\phi }^{-}\left(a_k\right)$ which measures how much all other alternatives are preferred over alternative $a_k$ are calculated through Equations (13) and (14), accordingly:

$${\phi }^{+}\left(a_k\right)={\displaystyle \frac{1}{t-1}}{\sum }_{m=1}^t\pi (a_k,a_m)$$

(13)

$${\phi }^{-}\left(a_k\right)={\displaystyle \frac{1}{t-1}}{\sum }_{m=1}^t\pi (a_m,a_k)$$

(14)

where $a_m$ represents all alternatives apart from $a_k$ and $t$ is the total number of alternatives. Based on the positive and negative flow scores, a partial ranking of the alternatives is then created. An alternative $a_k$ is considered to outrank alternative $a_m$ if ${\phi }^{+}\left(a_k\right)>{\phi }^{+}\left(a_m\right)$ and ${\phi }^{-}\left(a_k\right)<{\phi }^{-}\left(a_m\right)$.

PROMETHEE II extends PROMETHEE I by providing a complete ranking of the alternatives. PROMETHEE II combines the positive flow ${\phi }^{+}\left(a_k\right)$ and the negative flow ${\phi }^{-}\left(a_k\right)$ into a single net flow score $\phi \left(a_k\right)$ through Equation (15):

$$\phi \left(a_k\right)={\phi }^{+}\left(a_k\right)-{\phi }^{-}\left(a_k\right)$$

(15)

Using the net flow scores $\phi \left(a_k\right)$, all alternatives can be ranked. The higher the net flow, the better the alternative. If ${\phi (a}_k)$is a positive value, then alternative $a_k$ is important, while if ${\phi (a}_k)$is a negative value, alternative $a_k$ is unimportant [7].

3. Results

3.1. LCA Results

Figure 5 shows the relative results of LCA for the different LIBs, regarding their environmental impacts. Climate change is measured in CO₂-eq and freshwater, marine, and terrestrial ecotoxicity are measured in 1,4-DCB-eq. For each indicator, the maximum impact caused by an alternative is set to 100% and the impacts of the other LIBs are displayed in relation to this result. The comparative assessment reveals that LMO is the most harmful technology across all four impact categories. This can be attributed to its insufficient lifetime and low specific energy, which both lead to a larger battery pack delivering the same energy as other technologies over its lifespan. The need for more materials leads to greater environmental impacts, as well as an increased burden in terms of waste and disposal. LCO is the second most harmful technology. Its negative impact is primarily driven by a combination of an inadequate lifetime and its significant reliance on cobalt. Cobalt mining has well-documented toxicity concerns. This makes LCO a less sustainable choice compared to alternatives that either use less harmful materials or have better durability. NCM and NCA exhibit similar levels of environmental damage across all impact categories, but NCM is slightly less harmful, especially in terms of toxicity. This is connected primarily to the elevated specific energy of NCM compared to NCA. Lastly, the performance of LFP stands out as particularly positive. It exhibits the lowest impact across all categories, which can largely be attributed to its long operational lifetime and minimal content in harmful materials. Also, the minimal need for replacements results in reduced resource extraction and waste production, making it a more sustainable option. In summary, the key factors influencing the comparative assessment of environmental impacts include lifetime, specific energy, and material composition, with technologies that perform better in these areas generally demonstrating lower environmental impact.

3.2. Performance of the Alternatives and Criteria Weights

Table 4. Performance of the alternatives with respect to the specified criteria.

	Specific Energy (Wh/kg)	Safety (°C)	Lifetime (Cycles)	GWP (CO2-eq)	Toxicity (Species·Year)	Total Cost (€/kWh)	Residual Value (€/kWh)	Material Criticality (1 to 9)
	Max	Max	Max	Min	Min	Min	Max	Min
Alt.
NCM	274	210	1500	1.11	3.92	0.34	0	6
LFP	204	270	2500	1.03	2.63	0.3	−0.005	1
NCA	237	150	1500	1.09	4.6	0.35	0	6
LMO	156	250	750	2.08	25.65	1.11	−0.076	3
LCO	258	150	500	1.4	12.72	0.74	0.041	9

Abbreviations: LCO: lithium cobalt oxide; LFP: lithium iron phosphate; LMO: lithium manganese oxide; NCA: lithium nickel cobalt aluminum oxide; NCM: lithium nickel cobalt manganese oxide.

offers an overview of the values each alternative assumes for each criterion, highlighting whether a maximum (max) or minimum (min) value is desirable for the criterion. These raw values represent the performance of each alternative with respect to the specified criteria before any weighting is applied. The values are extracted through the methodological processes described in Section 2.2.1, Section 2.2.2, Section 2.2.3 and Section 2.2.4.

The weights of the criteria categories, along with the weights of the sub-criteria within each category, are presented in Figure 6, following the aggregation of the individual judgments of experts. It is worth noting that technical criteria and material criticality are the top-rated categories, followed by environmental, and lastly economic factors. Based on the weights of the criteria, the calculated CR is 0.0217, which is significantly below the threshold of 0.1. This indicates a high level of consistency in the responses provided by the experts, thereby eliminating the need to enforce transitivity rules.

3.3. Final Ranking of the Alternatives

Figure 7a shows the attributes of the alternatives before incorporating the weights of the criteria. LFP outperforms the rest in lifetime, safety, material criticality, and environmental performance. On the other hand, NCM and NCA appear to have inferior specific energy and an overall good environmental performance and low cost. Their high specific energy and adequately good lifetime drastically lower the required mass of battery and replacement rate, respectively. LCO has a high residual value due to its increased content in cobalt, which results in considerable environmental damage and total cost. Lastly, LMO has the lowest specific energy and lifetime, and poor performance in almost all criteria.

Combining these attributes with the weights of the criteria and keeping in mind the above analysis, the weighted attributes of the alternatives as determined by AHP are presented in Figure 7b. Overall, the LFP battery appears to be the best option for combining a moderate performance among the selected criteria with an outstanding performance in terms of material criticality and lifetime. NCM and NCA are two highly evaluated alternatives, especially for their specific energy. However, their high content of valuable materials like nickel and cobalt and the fact that experts judged material criticality to be of profound importance lowers their overall score. LMO demonstrates strong performance in terms of material criticality and safety; however, its shortcomings in lifetime and specific energy undermine its overall effectiveness for EV applications. The frequent replacement requirements not only drive up costs but also impose a considerable environmental burden. LCO exhibits poor performance across all criteria, with its high cobalt content positioning it as the least favorable option. When taking into account its short lifetime, the already considerable environmental impacts are amplified, and LCO’s ranking falls even further.

Table 5. Positive, negative, and net flow scores of the alternatives.

Alt.	${\mathsf{\phi }}^{\mathsf{+}}$	${\mathsf{\phi }}^{\mathsf{-}}$	$\mathsf{\phi }$
LFP	0.8717	0.0833	0.7883
NCM	0.4973	0.3123	0.1849
NCA	0.3705	0.4245	−0.0541
LMO	0.3182	0.6659	−0.3477
LCO	0.1937	0.7652	−0.5715

Abbreviations: LCO: lithium cobalt oxide; LFP: lithium iron phosphate; LMO: lithium manganese oxide; NCA: lithium nickel cobalt aluminum oxide; NCM: lithium nickel cobalt manganese oxide.

summarizes the positive, negative, and net flow scores of each alternative as determined by applying PROMETHEE I and II methods. The positive flow score represents the degree to which an alternative is preferred over others, reflecting its overall dominance. Conversely, the negative flow score indicates the extent to which the alternative is dominated by others. The net flow score provides a comprehensive ranking, highlighting the overall desirability of each alternative. It is evident that LFP is the dominant option, as it has the highest positive, the lowest negative, and the highest net flow scores.

A visual representation of the flow scores is provided in Figure 8. Figure 8a represents the partial ranking of the alternatives based on PROMETHEE I and the positive and negative flow scores. Figure 8b illustrates the complete ranking of the alternatives based on PROMETHEE II, which combines the positive and negative flow scores into a single net flow score. LFP has the highest ranking, achieving a net flow score of 0.7883 among all other alternatives. Far below is the NCM battery scoring 0.1849, followed by the NCA technology. It is worth noting that NCA has a negative net flow score, which according to PROMETHEE indicates an undesirable option. Similarly to the results of the AHP method, LMO and LCO scored at the bottom of the ranking.

A better understanding of the scores of each technology is offered in Figure 9, in which the performance of each alternative for all criteria is illustrated. Particularly for the LFP battery, the combination of low material criticality, long lifetime, low toxicity impacts and total cost, as well as good safety and low GWP, brings it to the top of the ranking. Residual value and specific energy, on the other hand, are the two primary drawbacks of this technology. The NCM battery performs well in all criteria. High specific energy is the most favorable characteristic, but the great significance of material criticality given by the experts brings it down to a lower preference. Similarly, NCA lags in terms of critical materials used in it, as well as safety which brings it below average in the full ranking. On the contrary, LMO excels in terms of material criticality and safety but falls behind in environmental impacts and cost. Its low specific energy also remains an issue. Finally, the biggest bottleneck in the potential extensive use of LCO in mobility applications is the number of critical materials used in it. A comprehensive representation of the results by GAIA is given in Figure 10.

3.4. Sensitivity Analysis

The extent to which variations in the weights of the criteria may potentially influence the result was evaluated through sensitivity analysis. The stability level of criteria was used to observe the stability of the ranking and identify the most influential criteria, as illustrated in Figure 11. In Figure 11, the horizontal axis represents the criterion’s weight, while the vertical axis represents the net flow score. Each alternative’s line shows how the net flow score changes with the varying weight of the criterion. At the left point, the criterion weight is 0%, and at the right point, the alternatives are ranked based on this single criterion. The point where each alternative’s line intersects the vertical axis indicates its overall ranking. The two dashed vertical lines show the range of weights within which the ranking remains unchanged.

The most unstable criteria are specific energy and residual value. It is evident that if the weight of specific energy increases from 7% to 19.12%, the LCO battery outweighs LMO. A further increase to 35% would lead to NCM coming to the top of the ranking. The results are also sensitive to the weight attributed to residual value after the EoL of batteries. Only a slight increase from 4% to 13.7% would again bring the LCO technology above LMO in the final ranking, which is explained by the high residual value of the certain technology due to cobalt content. The most stable criteria are GWP (77.81%), lifetime (41.38%), total cost (39.29%), and toxicity (39.29%). This means that the final ranking would only be altered if major increases in the weights of these factors would happen.

4. Discussion

This study explored the assessment of LIBs for automotive applications guided by sustainability criteria and the judgments of electric mobility experts through group decision-making. The application of both AHP and PROMETHEE in the analysis yielded similar results. This outcome is attributed to the consistency of how both methods handle the evaluation of alternatives based on the criteria and the assigned weights. AHP derives rankings by structuring decision-making into a hierarchy and using pairwise comparisons to determine relative importance, while PROMETHEE employs outranking techniques to compare alternatives directly across criteria. Despite their methodological differences, the alignment in their results occurs because both approaches ultimately rely on the same set of criteria, weights, and input data to guide the decision-making process. This high consistency underscores the robustness of the evaluation and enhances the credibility of the findings.

In general, the findings indicate that LFP, which is a technology that does not use scarce materials like cobalt or nickel and additionally is characterized by an exceptionally long lifetime, is a leading choice among all technologies. LFP’s criticality risk lies only upon the supply risk of iron. As mentioned above, LFP is characterized by extensive durability, ensuring not only high reliability but also maintaining higher resale value for the vehicle as well as minimizing the need for replacements. LFP also ensures advanced safety, holding the higher thermal runaway temperature. This technology is held back by its modest specific energy which leads to a larger battery mass required to deliver the same energy as other cathodes. Overall, the combination of a long lifetime and exceptional safety in addition to its low content in critical materials that not only would raise its cost but also make the supply chain unstable makes this technology ideal for EVs.

NCM and NCA cathodes are also good options for mobility applications. The remark of these technologies is their high specific energy. Although they share similar technical characteristics related to the performance of the EV, NCM offers elevated safety and that is the main characteristic that places it higher. The main limiting factor for these two technologies is their content in critical materials. Nickel, manganese, and cobalt are strong resource concerns due to their estimated depletion time. Nickel is also critical in terms of the availability of resources. Cobalt exhibits the most concentrated market structure and is dependent on regions with significant political risks and policy perception risks.

LMO and LCO are at the bottom of the hierarchy for EV applications. LMO utilizes non-critical materials and offers good safety, but lags in the other technical specifications. It is particularly marked by insufficient durability, which increases the total cost of ownership and amplifies environmental impact. As LMO is an option that does not contain any hazardous or expensive materials, it can be combined with other cathodes, like NCM and NCA, to achieve a combination of their top-rated characteristics.

LCO is defined by its high specific energy, but several other factors limit its extensive use in EVs. The primary limiting factor is the high content in cobalt, which brings this alternative to a disadvantageous place in terms of material criticality, toxicity effects, and high cost. LCO is a cathode that should take advantage of its superior specific energy and be used in applications whose primary requirement is extremely low weight and relatively low needs in durability over time. Suitable applications are wearable technology and medical devices, portable electronics and medical equipment, drones, and e-bikes.

5. Conclusions

The results of the above analysis underscore that the widespread adoption of a technology relies broadly on the availability and sustainability of the materials it requires. Utilizing abundant materials is crucial for ensuring reliable production, reducing costs, and enabling broader accessibility. This approach not only supports ethical and sustainable development but also fosters resilience and adaptability in a rapidly changing global landscape. On the contrary, technologies that depend on scarce materials with rather questionable and unstable supply chains might face limitations to scalability and long-term adoption.

It is worth noting that despite the contributions of this research, certain areas remain underexplored. Firstly, emerging technologies that have not yet reached commercialization were not considered in the analysis as the data are difficult to retrieve. Future research could explore the multidimensional assessment of such emerging battery technologies, like lithium–sulfur and lithium–oxygen. These are very promising technologies in the field of electric mobility due to their exceptional energy content. Apart from graphite, next-generation materials for the anode of the battery should be further examined as well. Novel materials, like silicon nanotubes and nanowires or lithium metal, may improve the overall performance of the battery, but to fully understand their potential, they must be evaluated from a multi-criteria perspective, including environmental, economic, and other critical factors. LCA combined with MCDA could deepen the understanding of any bottlenecks that need to be faced to make these options widely available. Future research should also focus on the social aspects of the supply chain of batteries for EVs. By addressing this issue, the full viability of these technologies can be ensured.

It must also be emphasized that the scope of the study is limited to assessing the sustainability of alternative LIBs; however, a comprehensive analysis would encompass additional end-user-related criteria. Factors such as charging time, cycling stability, and vehicle performance play a crucial role in determining the overall feasibility and practicality of these alternatives. By integrating these aspects, a more holistic evaluation could be achieved, providing insights that better align with the needs and expectations of end users.

Finally, while this study provides valuable insights into sustainability criteria weighting, the reliance on expert judgment introduces certain limitations, particularly in terms of subjectivity. Future studies could complement this approach by incorporating quantitative analysis tools to validate the expert-based results. Such integration could help bridge qualitative expert insights with data-driven methodologies.