Low-Temperature Performance and Durability of Electric Vehicle Battery Cells Under Isothermal Conditions

Abstract

Electric vehicle (xEV) battery durability significantly impacts the long-term operation, consumer satisfaction, and market adoption of xEVs. As driving range diminishes over time, it affects vehicle service life and lifecycle GHG emissions. Measuring the full service life of xEV batteries in laboratory tests presents technical and logistical challenges, necessitating representative measurements for parameterizing numerical models. These models are crucial for predicting long-term performance and rely on high-quality experimental data. While performance and aging trends under extreme temperatures are documented, cell thermal contact conditions suitable for direct model input are not well characterized. This study investigates lithium-ion cells from three xEV types, cycled at constant currents from C/40 to 1C, at temperatures between −15 °C and +45 °C, over 1000 cycles in a multi-year campaign. Stable isothermal cell temperatures were achieved using custom-built liquid immersion baths with forced convection, highlighting fundamental electrochemical behaviors by decoupling complex self-heating not typically monitored in air environments. The data inform and validate physics-based models on temperature-dependent performance and durability, providing operational limits to enhance cell and battery thermal management design and educate xEV consumers about conditions affecting performance, range, and durability.

1. Introduction

Unlike the main propulsion components in internal combustion engine vehicles, battery aging negatively impacts EV performance and range. For hybrid EVs, this results in reduced fuel economy that has a profound effect on greenhouse gas calculations and predictions [1].

As the number of xEVs on public roads continues to increase, more information is required to determine their durability under various usage patterns. The Canadian winter climate offers additional challenges. It has been shown that decreased solid-state Li diffusivity, low electrolyte conductivity, and sluggish kinetics in lithium-ion batteries at low temperatures cause significant power loss and a dramatic reduction in driving range [2]. There have been several other adverse effects documented, including capacity loss, power loss, life degradation, safety hazards, and charging difficulty [3]. There is a body of literature on cold climate driving range, but these do not address low-temperature cell performance directly [4,5]. Numerous studies have looked at performance losses at low temperatures in commercial lithium-ion cells [6,7,8,9].

It is difficult to experimentally test xEV cells under realistic conditions for the full duration of their service lifetime endpoint, which is presently well over 10 years. The final service life of an xEV battery is usually denoted by a threshold limit of remaining usable capacity (ex. 75%). The date when this threshold is reached depends on a multitude of factors, such as cell design, operating limits, battery thermal management, charging protocols, driver behaviors, and environmental conditions. Accelerated aging testing is well known to be effective for only some high-temperature cell aging mechanisms [10] (e.g., solid electrolyte interface formation), which may not be representative of aging in low-temperature climates, thus raising questions about its equivalency to other dynamic operation conditions [11]. Many dominant parasitic side reactions follow the Arrhenius equation (increases exponentially with temperature) [10,12]. However, at low temperatures, cell aging is predominately caused by lithium plating [13,14,15,16,17], which is inhomogeneous [18] and accumulates over time [7,19]. Aging effects result from a complex combination of these temperature-dependent mechanisms [20,21].

Predicting service lifetime through numerical modeling remains an essential means of providing lifetime estimates [22,23]. The accuracy of physics-based numerical models relies on the fidelity of the aging behaviors under different usage patterns [24]. Many battery simulations run with highly irregular duty cycles, such as U.S. Environmental Protection Agency (EPA) drive cycles, where battery loads vary dramatically on a second-by-second basis, and a high degree of temperature dependent fidelity is essential for plausible forecasting [17]. Simulation accuracy is also highly dependent on the experimental data used for training [2]. Parameter identifiability is a critical challenge in both electrochemical battery models and equivalent circuit models [25], and there is a high sensitivity to model parameterization [26]. Coarse temperature estimates have been shown to overestimate pack lifetimes by 10 to 20% [27]. With model input and validation data that are decoupled from cell/pack thermal behaviors (heat capacities, thermal management strategies, etc.), a numerical model can pull data from various temperature states and overlay the specific thermal conditions of each usage scenario [28,29].

Experimentally testing xEV lithium-ion cells at low temperatures is technically challenging and requires careful experimental design. First, low temperatures and high-current operation are well-known safety risks [30]. The low-temperature fast charging of lithium-ion batteries is known to cause lithium plating, which can lead to an increased risk of thermal runaway [31]. Tests involving any low-temperature charging protocols must be performed within appropriate enclosures that can contain/mitigate such events. Special care and maintenance must be taken to avoid condensation and frost build-up on electrical connections to the cells, which may increase contact resistance and influence the electrochemical measurements. Perhaps the most critical detail that affects the accuracy and applicability of the experimental data collected is the thermal contact conditions of the cells under test. Cells in operation will generate internal heat through internal resistance losses (joule heating), which will raise their effective operating temperatures. As currents increase, the heat generated increases through ohmic heating. There is an important distinction between an environment’s set-point temperature, the cell surface temperature, and the cell’s internal temperature, which is rarely considered or controlled in experimental studies. In conventional forced air convection environments with cells installed in open air, as current increases, heat transfer rate limitations allow large thermal gradients to develop, resulting in increased and less stable cell temperatures. These types of test environments have been referred to as isoparbolic cell temperature conditions [32]. The absolute temperature increase depends on a multitude of factors, including the cell chemistry, physical construction, state of charge (SOC), current load, and the chosen thermal management system (in a laboratory or in a vehicle) [33]. Jaguemont et al. demonstrate these effects, leading to an increase in cell temperature of +14 °C over a 1C discharge at a set point of −20 °C [8]. Wu et al. present experimental data showing the cell surface temperature on Lithium Nickel Cobalt Aluminum Oxide (NCA) based 18,650 cells increasing by +10 °C, +23 °C, and +37 °C at the end of 1C, 2C, and 3C discharges, respectively. In any low-temperature durability study, the effects of self-heating are apparent when cell potential increases during the early stages of discharge. In these such cases, although a cell may start at a sub-zero temperature initially, the average cell behavior will be more representative of much warmer cell conditions.

These self-heating behaviors have led some studies to conclude that self-heating caused by moderate/high current rates improves capacity and power retention during low-temperature experiments [9,34]. Ecker et al. demonstrate the variability in self-heating extent across different cell types and found that cells with more self-heating are less prone to lithium plating [7]. These conclusions are true for the specific thermal boundary conditions of the cell level experiment; however, the improvements are subjective to each cell’s heat dissipation, which can vary by manufacturing differences and cell age [35], and are not easily relatable to the thermal configuration of an arbitrary battery module design [36,37]. Thermal trajectories caused by self-heating occurring in laboratory cell level experiments are not equivalent cells in vehicle use because of differences in cell heat transfer conditions (ex. cell surrounding materials, vehicle thermal management strategies, etc.).

Tests designed to control the cell surface temperature isothermally eliminate these temperature fluctuations and enable a high-fidelity determination of cell properties and performance as a function of a set-point temperature. While it is unconventional in classical electrochemical testing, the data collected under these conditions are ideal for use in numerical models where rapidly changing battery thermal environments maintain real-time accuracy for thermally dependent parameters that drive electrochemical phenomena. Landini et al. highlight the strongly dissimilar performance in Lithium Cobalt Oxide (LCO) cells tested under isothermal and controlled environmental temperatures (i.e., isoperibolic) conditions in the 0 °C−40 °C range [32,38]. The authors emphasize the need for more robust isothermal datasets to be universally comparable and not dependent on random transient heat transfer conditions in isoperibolic environments [32]. This study aims to expand on the work of Landini et al. to include three xEV (Lithium Nickel Manganese Cobalt Oxide (NMC)-based) cells, studying them at sub-zero temperatures over a multi-year test campaign.

The purpose of this study is to generate an extensive and wide-ranging dataset on EV cell performance and aging that can be used to support the parameterization of battery numerical models. The data also stand as cell performance benchmarks at absolute temperature reference points, which can be used toward the design of optimized thermal management systems. A distinguishing attribute of this study was the use of custom-built isothermal baths, which enable high-precision control of cell surface temperatures. Testing under isothermal conditions allows for the fundamental understanding of the performance and endurance at low temperatures since complicating heat accumulation factors, which introduce errors in temperature attribution, have been effectively removed here.

This work contributes to a broader team effort that uses this battery data in a suite of battery simulation models, which span the range of complexity from full 3D electrochemical transport models to equivalent circuit representations used for long-term durability assessments for various automotive transportation sector applications [22,39,40]. A key strength of these battery operation models is the accompanying empirical battery degradation model developed from battery durability experiments such as those reported in this paper. The experimental approach isolates the various degradation drivers and has enabled the development of a model that can be applied over a short time step (1 s or less) and applied to the instantaneous battery state accounting for current level, temperature, SOC, and a duration of operation effect.

2. Methods

Experimental tests were conducted on different xEV cell types and at various charge/discharge currents and isothermal environment temperatures to assess the cell’s performance and durability under these conditions.

2.1. Cell Selection

For this study, three xEV cell types were selected based on

• The availability of EV packs such that test cells could be extracted;
• The prevalence of the respective EVs within the Canadian market;
• The diversity in physical size, format, and capacity of the chosen cells and the EV architecture from which they were extracted.

The following table identifies the three cell types selected and lists their key parameters.

2.2. Performance Evaluation

The purpose of performance rate mapping of EV cells was to

• Provide an overview of baseline performance;
• Estimate the general state of health of the test cells;
• Establish limits of operation.

Performance evaluation tests were performed for every possible combination of the three parameters listed in

Table 1. Test matrix for selected EV cells.

EV Type	PHEV (Plug-in Hybrid Electric Vehicle)				HEV (Conventional Hybrid Electric Vehicle)				BEV (Pure Battery Electric Vehicle)
Vehicle make	Chevrolet Volt				Toyota Prius				Nissan Leaf
Model year	2013				2013				2014
OEM battery thermal management system	Indirect liquid				Forced air				Passive + resistive heating
Past usage history	Cells were extracted from EVs with less than 2000 km traveled and stored outdoors in a shipping container in Ottawa, Canada, for 3 years.
Cell format	Pouch				Prismatic				Pouch
Cell mass (g)	445				721				900
Cell surface to volume ratio (cm−1)	378				106				298
Principle chemistry (anode/cathode)	Graphite/ NMC (1:1:1): LiMnO4 (60:40) blend				Graphite/ NMC (1:1:1)				Graphite/ NMC (6:2:1): LiMnO4 (40:60) blend
Rated discharge capacity (Ah)	15				21.5				33
Voltage limits 1 (V)	3.0–4.15				3.0–4.15				2.5–4.2
Charge/discharge current (C-rate)	C/40	C/4	C/2	1C	C/40	C/4	C/2	1C	C/40	C/4	C/2	1C
Charge/discharge current (A)	0.375	3.75	7.5	15	0.538	5.38	10.75	21.5	0.825	8.25	16.5	33
Environment temperatures (°C)	For each cell and each C-rate
	−15	−5			5	15		25	35		45

¹ Suggested voltage limits are based on the cell’s ranges quoted by the US Department of Energy reports.

(3 cell types, 4 symmetric charge/discharge current, 7 temperatures, for a total of 84 tests). Due to limitations in equipment availability, it was favored to have wider and more resolute temperature and current ranges than to perform duplicate tests at each configuration. For each configuration, an individual fresh cell was electrochemically charged at constant current and discharged at constant current between the listed voltage limits, with 15 min rest periods at the end of both charge and discharge. This constitutes one complete cycle.

Lithium-ion cells that have been dormant in storage for months or years will yield irregular capacity for the first few cycles. Generally, the discharge capacities of aged cells “recover” (increase) over the first few cycles after dormancy. Thus, preconditioning consisted of 20 cycles to establish a steady state response, free of transient effects caused by prolonged dormancy [41].

2.3. Durability Evaluation

The purpose of long-term endurance cycling of EV cells was to

• Determine an absolute value for useful service life (if possible in testing timeframe) under laboratory controlled conditions;
• Establish the cell aging trends at various operating conditions, which can inform numerical models.

Durability evaluation tests were identical to performance evaluation (see Figure 1), except they were continued for a total of 1000 charge/discharge cycles, equivalent to 1000 full useable range trips in the associated xEV, or nearly 3 years of one full SOC range trip with daily driving. A total of 1000 cycles was necessary to develop sufficient usage that would advance cell aging to a state of health where durability could be assessed with measurement certainty. Durability evaluation tests were performed on every input parameter permutation in Table 1, with the exception of C/40 currents since they would be impractical to complete because each test would take over 9 years (total of 63 tests).

2.4. Testing Apparatus

A unique feature of this study is the use of in-house-built isothermal baths to maintain EV cell surface temperatures at the desired set points. EV cells are placed inside flexible, thin rubber-lined sleeves, which are suspended and submerged into a large volume (60 L) of heat transfer fluid (50/50 ethylene glycol–water). The heat transfer fluid is maintained at isothermal temperatures by a forced convection flow from an external process chiller. The chiller is capable of providing 500 W of cooling at a −20 °C set point. The batteries were encased in tight fitting 0.8 mm thick rubber membranes for immersion into the cooling baths, thereby providing very direct thermal contact with the heat transfer fluid, which resulted in both excellent heat extraction and high thermal stability (detailed description of this equipment and EV cell heat generation will be the subject of a forthcoming paper).

Figure 2 shows the inside of the custom-built isothermal baths and battery cell sleeves for testing large EV-sized battery cells.

Calibrations were conducted on the isothermal baths to determine the rise in cell surface temperature under worst-case conditions (continuous charging and discharging at 1C rate on the largest capacity cell type (BEV) at −15 °C). The maximum temperature difference between the cell’s surface temperature and the bath set point was observed to be +0.3 °C during an 8 h period.

One physical limitation of testing with an isothermal cell surface is that some extent of internal cell temperature gradients will always persist [42,43] because of the thermal conduction limit of the cell [44]. Some research utilizes internal thermocouples [43,45] to further study these internal thermal gradients; however, this approach is impractical for testing hundreds of commercial cells needed to develop a robust dataset, and the absolute internal temperatures that develop are still greatly reduced compared to conventional air-cooled environments.

Electrical connections were made to each cell using the standard 4-wire connection measurement technique, using separate pairs of current-carrying and voltage-sensing wires, which allows for battery test equipment to compensate for resistance losses along the current carrying wires. The battery test equipment used to collect and log data from these experiments has a calibrated capacity measurement accuracy of +/−0.02 Ah for the current range used for PHEV cells and +/−0.05 Ah for the current range used for HEV and BEV cells.

3. Results

Note: This section presents the measurements from the battery test equipment and describes and justifies their representation format. A detailed analysis and discussion of these results are performed in the subsequent Section 4.

The discharge capacity of a battery refers to the amount of electric charge a battery can deliver under specified conditions. In this case, the fully charged battery cells are discharged at a constant current to a cutoff voltage, which is set by the manufacturer, while recording the time taken to reach this voltage. The discharge capacity (in amp-hours) equals the discharge current (in amps) times the discharge time (in hours).

The normalized discharge capacity for each test was calculated by dividing the measured discharge capacity logged by the battery testing equipment at the given current rate and temperature condition by each cell’s rated capacity given by the manufacturer at nominal conditions (see Table 1). The normalized discharge capacity representation allows for the direct comparison of performance and durability trends across all cells of different types, sizes, and capacities. It is also a convenient format to compare performance and durability trends to a static reference value for each cell type. The raw measured values for discharge capacity can be back-calculated by multiplying the normalized values by the rated values for each cell type listed in Table 1.

Table 2. Baseline normalized discharge capacities for all test cases.

Cell Type	C-Rate	Temperatures (°C)
		−15	−5	5	15	25	35	45
PHEV	C/40	90%	95%	97%	100%	100%	100%	100%
	C/4	63%	72%	84%	91%	88%	96%	98%
	C/2	46%	72%	68%	84%	88%	93%	94%
	1C	0%	12%	49%	68%	79%	88%	88%
HEV	C/40	92%	94%	94%	95%	96%	96%	96%
	C/4	82%	87%	91%	93%	95%	95%	96%
	C/2	75%	84%	88%	92%	93%	94%	94%
	1C	65%	70%	79%	88%	87%	86%	93%
BEV	C/40	91%	97%	99%	96%	98%	98%	99%
	C/4	69%	84%	85%	93%	97%	98%	98%
	C/2	47%	60%	74%	87%	91%	95%	93%
	1C	20%	24%	56%	70%	80%	84%	86%

shows the baseline normalized discharge capacities during the performance evaluation of all test cases following the 20th charge/discharge cycle.

The PHEV, HEV, and BEV performance data contained in Table 2 are visualized in the three surface plots in Figure 3, Figure 4, and Figure 5, respectively.

Durability is assessed by comparing capacity lost over extended usage and time. The final normalized discharge capacities remaining after 1000 test cycles are reported in

Table 3. Final normalized discharge capacities for all test configurations after 1000 cycles.

Type	C-Rate	Temperatures (°C)
		−15	−5	5	15	25	35	45
PHEV	C/4	59%	71%	84%	88%	88%	85%	69%
	C/2	37%	72%	67%	82%	85%	85%	73%
	1C	0%	5%	49%	68%	75%	83%	65%
HEV	C/4	71%	83%	86%	87%	88%	86%	83%
	C/2	60%	70%	83%	87%	77%	87%	85%
	1C	52%	59%	73%	82%	77%	68%	85%
BEV	C/4	68%	83%	85%	86%	88%	85%	68%
	C/2	46%	48%	70%	78%	82%	81%	67%
	1C	15%	12%	51%	51%	65%	70%	69%

. Capacity lost is the measure of the initial capacity minus the final capacity. Again, capacity loss is normalized by dividing by each cell’s rated capacity.

Capacity loss rate is commonly represented over a number of cycles or time. Many numerical models for estimating battery durability include a linear combination of cycling fade and calendar fade aging behaviors, but during continuous current experimental tests, both time and cycle number are proportionally related through the C-rate. Capacity loss per cycle can be misleading during high-temperature operation. For example, the capacity lost during the same number of cycles exposed to 45 °C was found to be greater at C/4 than 1C because at C/4, the cells will spend approximately 4 times longer exposed to the extreme environmental temperature. Therefore, capacity loss is represented as a function of time spent at a given condition (temperature, current passed) rather than per cycle.

Table 4. Average normalized capacity lost per day for each case.

Type	C-Rate	Temperatures (°C)
		−15	−5	5	15	25	35	45
PHEV	C/4	0.02%	<0.01%	<0.01%	0.01%	0.02%	0.04%	0.09%
	C/2	0.11%	<0.01%	<0.01%	<0.01%	0.02%	0.05%	0.13%
	1C	INF	0.28%	<0.01%	<0.01%	0.05%	0.05%	0.27%
HEV	C/4	0.04%	0.03%	0.02%	0.02%	0.02%	0.02%	0.04%
	C/2	0.11%	0.09%	0.03%	0.03%	0.10%	0.04%	0.05%
	1C	0.40%	0.34%	0.18%	0.07%	0.12%	0.23%	0.08%
BEV	C/4	<0.01%	<0.01%	<0.01%	0.02%	0.03%	0.04%	0.10%
	C/2	0.01%	0.11%	0.04%	0.06%	0.05%	0.08%	0.15%
	1C	0.12%	0.19%	0.06%	0.12%	0.18%	0.17%	0.20%

shows the normalized capacity loss divided by the test time (in days) continuously exposed to each test condition.

It is important to emphasize that days here refers to continuous laboratory controlled tests, consisting of full SOC cycling at the listed currents and temperatures conditions, with only 10 min rests between charge/discharge. These test days are greatly accelerated compared to even the most severe in-vehicle usage scenarios and thus cannot be directly compared to service life days. For example, at 1C charge/discharge rates, the battery test equipment can perform at least 12 full CC charge/discharge cycles on a cell per test day (more if cell capacity has diminished), whereas the same number of charge/discharges would be impossible to achieve through a single day of vehicle operation. Therefore, these values provide estimates of the relative capacity loss only while operating over the shorter time interval in which these conditions apply.

The PHEV, HEV, and BEV durability data contained in Table 4 are visualized in the three surface plots in Figure 6, Figure 7, and Figure 8, respectively.

Following these durability evaluation results, there was interest to determine whether the capacity reductions observed at low temperatures were conditional on low temperatures or permanently lost. To answer this question, an additional 20 cycle performance evaluations were performed on select cells that endured 1000 cycles at cold temperatures. These performance evaluations were conducted at 25 °C on select cells that had experienced long-term exposure to −5 °C and −15 °C, with the intent to determine the distribution of reversible and irreversible capacity loss. The difference between the capacity observed during post-tests at 25 °C and the final capacity observed during long-term exposure at low temperature can be considered as the reversible capacity loss, and these results are shown in

Table 5. Final normalized discharge capacities at specified set-point temperature vs. post-test normalized discharge capacities at 25 °C.

Type	C-Rate	Final Normalized Discharge Capacities at −15 °C	Post-Test Normalized Discharge Capacities at 25 °C	Normalized Capacity Recovered	Final Normalized Discharge Capacities at −5 °C	Post-Test Normalized Discharge Capacities at 25 °C	Normalized Capacity Recovered
PHEV	C/4	59%	No data	No data	71%	87%	+16%
	1C	0%	80%	+80%	5%	65%	+60%
HEV	C/4	71%	No data	No data	83%	87%	+4%
	1C	52%	60%	+8%	59%	68%	+9%
BEV	C/4	68%	No data	No data	83%	97%	+14%
	1C	15%	76%	+61%	12%	74%	+62%

.

4. Discussion

According to Table 2, at a C/40 C-rate, all cells report >90% of their rated discharge capacity across the entire temperature range. While this very low current rate is impractical for any vehicle application, the capacity retention is of note, considering the age of the cells and the duration spent in unregulated dormancy. The C/40 data also highlights the fundamental temperature dependence of electrochemical processes and unbiased electrical inefficiencies (ex. resistance losses) that occur at higher currents. All cells report an 8 to 10% capacity decrease caused by exposure to −15 °C.

The most apparent observation from the baseline performance evaluation dataset (Table 2) is that all cells have a significant performance decrease, especially at high rates, as temperatures decrease. All tests performed at −15 °C with currents above C/4 or performed at −5 °C with currents of 1C provided less capacity than the common end-of-life criteria (75% rated capacity), indicating they would not meet requirements if these conditions were included in performance standards. The HEV cells fared the best under extreme cold conditions. Its performance is likely related to the HEV application it is used in, which requires higher power densities and simplistic thermal management strategies, necessitating a larger range of operating temperatures for normal vehicle use. In addition, the cell’s hard-case prismatic physical format has thicker walls and a lower surface-to-volume ratio than the other two types (see Table 1), which may allow greater internal heat retention during operation and less surface area to dissipate it, even during isothermal conditions. The PHEV cell at −15 °C and 1C had no available capacity during cycling. Demanding conditions form a distinct operational threshold, which is apparent in the steep downward spikes in Figure 3 and Figure 5 for certain combinations of low temperatures and high power levels. Where this threshold occurs in the variable space is dependent on the cell design (e.g., electrode chemistry, additives, etc.) and can be manipulated by a properly designed thermal management system (e.g., cell heating blankets, etc.). However, operation and storage in this region should still be avoided since the control effort required to heat cells into the nominal BMS-required temperature range requires energy, which comes at the cost of driving range at times when the EV is not connected to a charger [46]. For the BEV cells tested, some estimate cabin and battery heating can shorten the driving range by as much as 30–40% in sub-zero temperatures [30].

Some amount of the discharge capacity reductions observed during low-temperature operation is not permanently lost. Rather, a portion of the missing capacity is considered conditionally inaccessible or reversible. The majority is restored when returned to nominal temperatures, as shown in Table 5. Thus, extended exposure to low temperature in the vehicle at nominal SOC will not cause severe damage to the cell as long as the cells/pack is warmed up slowly before use.

It should also be considered that the discharge capacity is only half of the story. The inability to charge a cell to full capacity also contributes to the unavailability of capacity during discharge, and it becomes more apparent at low temperatures. One limitation of this study was that experiments were performed with only a constant current (CC) charge at the indicated current rate, where CC charging was applied until the cells reached their specified voltage limits. The rationale for this selection was based on a continuous-use scenario, such as long-duration highway driving with charging en route without significant rest or idling, where it would be impractical to trickle charge (constant voltage (CV) charging) for possibly several hours to achieve the complete remaining portion of capacity. It has also been shown in the literature that performing CV charge during lithium plating conditions leads to increased aging [7]. Cells with high impedance caused by dormancy, aging, or exposure to low temperatures will require a larger portion of their capacity to be delivered through the CV charge. Naturally, if this CV portion of energy is not inputted during charge, it is not available to be extracted during discharge, affecting the reported discharge capacities.

Table 4 indicates that a cell’s available capacity is disproportionally reduced at higher power rates in all cases due to resistance growth with age. Two distinct capacity-reducing peaks are apparent in Figure 6. The root cause of the peak, which occurs at high temperatures (≥15 °C), is related to well-known temperature-dependent parasitic side reactions (e.g., Solid Electrolyte Interphase (SEI) growth) within lithium-ion cells [47]. The reaction rate of these side reactions increases according to an Arrhenius equation, approximately doubling with every increase of +10 °C.

The root cause of the peak, which occurs at low temperatures (≤5 °C), is related to lithium plating [48]. At low temperatures, the capacity loss rate is coupled with other effects under durability testing since the cells are often operating with less than 75% of their initial rated capacity, as discussed earlier. Irreversible (permanent) capacity loss on low-temperature cells is better evaluated by comparing post-test performance evaluations when these cold-cycled cells have been brought up to +25 °C (Table 5). Much of the inaccessible capacity is restored (>60%) for the PHEV and BEV cells. The PHEV cell at −15 °C provided no meaningful capacity over 1000 cycle attempts; yet, when brought back up to +25 °C, it provided the same capacity as the baseline at that temperature. This cell was perfectly preserved, despite being exposed to −15 °C temperatures for 600 h. The same cell type at −5 °C showed the most capacity decrease for this type according to Table 4, and most of this capacity loss was irreversible according to Table 5. Interestingly, the HEV cell type, which was the best performer at sub-zero temperatures, shows the greatest extent of irreversible loss, which may be attributed to the cell design considerations stated above. These trends suggest that the aging effects are a stronger function of the total current passed at low temperatures rather than solely time exposure to low temperatures.

Parametrization of Battery Models

To demonstrate the utility and application of the capacity measurements made in this study, a state-of-the-art capacity fade approach is introduced in this section, which was parameterized using these data. The comprehensive methodology and findings for each predictive numerical simulation can be found within each referenced article. Capacity loss or fade is normally expressed in amp-hours, occurring at nano- to micro-level quantities over a time step and accumulating to significant percentage levels over extended time periods. Explicitly, the expression developed is

$${\mathrm{cap}}_{\mathrm{fade}}={\mathrm{cap}}_{\mathrm{fade}}^{\mathrm{REF}} ∆t{ F}_{\mathrm{CUR}}\left[I\left(∆t\right)\right] F_{\mathrm{DOD}}\left[\mathrm{DOD}\right] F_T\left[T,I\right] F_{\mathrm{DUR}}[∆\mathrm{DOD}]$$

(1)

In this equation, ${\mathrm{cap}}_{\mathrm{fade}}^{\mathrm{REF}}$ is a reference degradation rate determined from experiments and normally given in Ah/s, Δt is the time step, I is the current, T is the temperature, DOD is the depth of discharge in the cell, ΔDOD is the extent of DOD swing in effect, and F_i terms are factors applied to modify the degradation rate according to how the various inputs differ from the reference conditions. More complete details are provided in [39,49]. In the present work, the measurements presented here would provide cell-specific means to establish the reference capacity fade, ${ F}_{\mathrm{CUR}}$ and $F_T$ terms in Equation (1). An experimental approach using variable amplitude cycles has been developed by the authors to provide the F_DOD and F_DUR terms and is outlined elsewhere [49].

Applying Equation (1) in an electric vehicle usage simulation enables tracking battery state of health over the course of a battery pack’s lifetime by summing micro-level capacity losses for 1 s or shorter time steps over a service lifetime that can be upwards of 20 years. This approach absolves the simulations of requiring linkages to prescribed duty cycles (i.e., EPA drive cycles) so that virtual real-time load demands can be treated with simulations. This approach has enabled leading-edge technology assessments via model forecasting for subjects, such as using EVs as taxis, and emerging technical solutions for heavy-duty long-haul road transport [50,51].

5. Conclusions

Cells from three different EV types were electrochemically cycled across a broad range of currents and temperatures over 1000 cycles, simulating 1000 full-range trips for each xEV. The resulting data were used to create detailed operational and aging maps. Cell surface temperatures were maintained between −15 °C and +45 °C, providing a benchmark for performance and durability under these thermal conditions.

The experimental results showed that cells retained over 90% capacity at low charge/discharge rates (C/40) across all temperatures. However, capacity was significantly reduced at sub-zero temperatures (−5 °C, −15 °C) and high current rates (>C/4), falling short of service life criteria. For the PHEV and BEV cells tested, most capacity loss was reversible at room temperature, but the HEV cells experienced greater irreversible losses at 1C rates, with capacity reducing to 60–68% after cycling at sub-zero temperatures compared to 25 °C. At +45 °C, capacity loss was more related to the time spent at this temperature rather than the number of cycles. Conversely, at sub-zero temperatures, capacity loss was more influenced by the total current passed and was greater at higher power rates.

The impact of this research is an improved understanding of low-temperature aging behaviors and fundamental operating limits of xEV battery cells, especially useful for the characterization and validation of predictive numerical models used to extrapolate xEV battery durability. The results can inform approaches to enhance battery durability in cold climates, such as battery durability standards and testing methods. It highlights important considerations for conducting experimental electrochemical tests at low temperatures, such as cell self-heating and the use of nominal temperature capacity checks to distinguish irreversible capacity loss from conditional decreases. Finally, the results can be utilized to inform and improve cell and battery thermal management system design.

There are limitations that should be noted. First, finite battery test equipment availability meant that test duplication was not performed, and repeatability was not assessed. Duplicate or triplicate tests are recommended to improve the statistical confidence of the dataset and remove outliers. The battery cell types that were tested were limited to the availability of supply. Testing the latest generation of EV cells (such as nickel-rich NMC cathodes) would be ideal; however, there is an intrinsic time lag in obtaining and extracting EV cells and then executing multi-year aging experiments. Finally, experiments were performed with symmetric and constant current charge and discharge cycles, which lump charge and discharge behaviors together. Tests designed with asymmetric charge and discharge portions allow the ability to isolate the influences of each.

Future research directions will involve comparing EV cells extracted from model year 2020 vehicles with various usage histories and comparing effects of power function cycling (such as standard level 2 charging and drive cycle based loading) with conventional constant current electrochemical cycling. These tests will be conducted under isothermal conditions. The methodologies developed here can be applied to newer or advanced chemistry EV cells.