In the development of lithium ion batteries, safety design and evaluation play an important role in preventing problems such as ignition due to thermal runaway. We use simulation technologies such as COMSOL Multiphysics to understand the various phenomena that affect lithium ion batteries and to estimate battery safety. This paper describes a modeling methodology for testing the safety of chemical reaction heating in lithium ion batteries.
If the heat release is greater than the heat generated by external and internal heating, the battery will be thermally stable. If heat release is less, then temperatures rise steadily and lead to thermal runaway.
Our study focused on heating tests using thermal analysis to evaluate thermal runaway conditions. Three types of heating are considered:
If the heat release is greater than the heat generated by external and internal heating, the battery will be thermally stable. If heat release is less, then temperatures rise steadily and lead to thermal runaway. In the heating test simulations, external heating is supplied using an oven.
Figure 1. DSC measurements of heat reaction of active materials
When modeling internal heating by chemical reactions, several physical phenomena must be accounted for. The first is thermal degradation of the separators and electrolyte, which affects conductivity. Second is the negative electrode-electrolyte reaction, which involves multiple reaction processes that cannot be described using a single reaction formula. Here, the reaction is divided into two parts: the solid- electrolyte interface (SEI) reaction and the negative electrode-electrolyte reaction through the SEI. Finally, the positive electrolyte reaction is included in the model.
“ COMSOL Multiphysics seems to be the ideal platform for battery analysis.”
Table 1. Analysis conditions (18650 cylindrical battery).
To obtain the reaction heat model, we ran a series of Differential Scanning Calorimeter (DSC) measurements at a constant rate of temperature rise for each of the chemical reactions to obtain parameter fitting. Figure 1 shows an example of DSC measurements for a 1-hour temperature elevation process (5°C/min) where the positive electrode is LiCoO2, the negative electrode is carbon, and the electrolyte is a mixture of EC (ethylene carbonate) and DEC (diethyl carbonate).
From the results of DSC measurements in Figure 1, the heat generation rate coefficient — i.e., the amount of heat generation per unit volume and unit time — and its dependence on temperature T were drawn.
Figure 2. Analysis of reacted ratio under temperature elevation process (5°C/min) using reaction rate formula.
Figure 2 shows the reacted ratio of each material under the temperature elevation process, analyzed using the reaction rate formula determined by peak fitting from the experiments. The first reaction to occur was in the SEI and the negative electrode, followed by the electrode-electrolyte reaction in the temperature range 100°C to 200°C. The thickness of the SEI suppresses the rate of the negative electrode reactions, but the rate still increases as the reaction progresses (150°C to 200°C), where the reaction then participates in a direct reaction with the electrolyte and is greatly increased (200°C – 250°C).
We conducted a suite of simulations using the obtained reaction rate formula. The battery analyzed was an 18650 cylindrical battery with a LiCoO2 positive electrode, a carbon negative electrode, and a mixture of EC and DEC electrolyte.
For these simulations, the material property values for density, specific heat, heat transfer coefficient within the electrode surface, and other materials were the average of measured values for the positive electrode plate, negative electrode plate, and separators. For the area perpendicular to the electrode surface, we used the thermal diffusivity coefficient of the laminated film. We express heat release from the surface of the battery as the sum of heat transfer and thermal radiation. The heat transfer coefficient depends on atmospheric conditions, and thermal radiation depends on the material of the surface. We determined the coefficients of heat transfer and thermal radiation by fitting the change in surface temperature measured in oven tests. Table 1 lists the analysis conditions.
Figure 3. Change in battery surface temperature over time in heating test simulations at target temperatures: 145°C, 153°C, and 155°C.
Figure 3 shows the results of the heating test simulations. The atmospheric temperature was raised to and maintained at various target temperatures. At a target temperature of 145°C, we observed some self-heating, but the temperature stabilized and did not produce a thermal runaway. At 155°C, however, thermal runaway occurred and the surface temperature increased sharply. At a target temperature of 153°C, the surface temperature increased but thereafter stabilized. Yet, as the chemical reaction progressed, surface temperature increased by more than 100°C, which we judged to effectively constitute thermal runaway.
Figure 4. Internal temperature at start of thermal runaway for target temperature of 155°C (190–270°C) (above) and isosurface of negative electrode reacted ratio (below).
Figure 4 shows the internal temperature distribution and negative electrode reacted ratio distribution at the time when thermal runaway commenced at the target temperature of 155°C. The central portion of the battery was thehottest, and the temperature difference between the ends and center of the battery reached 80°C.
Results from the ARC tests with the same type of battery showed that self-heating was observed from 73°C onwards and thermal runaway began at 150°C. These results indicate that the temperature predictions for thermal runaway obtained in the simulation were valid.
Figure 5. Change over time of temperature distribution (color contours) and isosurfaces of reacted ratio of the negative electrode during thermal runaway for an internal short circuit heat generation of 20W (upper row) and 100W (lower row).
Figure 5 shows an extended safety study where we simulated a calamity resulting from an internal short circuit. Shown over a period of time are the temperature distributions during thermal runaway and the isosurfaces of reacted ratio of the negative electrode at 20W and 100W heat sources, respectively, from the short circuit. For several tens of seconds, a wide reaction zone was observed to move from the vicinity of the nail towards the end of the battery.
COMSOL Multiphysics seems to be the ideal platform for battery analysis, which requires the analysis of complex physical phenomena on different scales, such as the modification of chemical reaction model formulae, the application of integral boundary conditions for current distribution analysis, and the analysis of different physical phenomena for each domain. By being able to model all the phenomena related to heating and cooling a lithium ion battery properly, we are then able to extend our investigations to possible catastrophes.
Kobelco Research Institute, Inc. has devised not only the macro safety test simulations described in this paper but multilevel modeling and simulations, including charge and discharge cycle test, ionic transport within the electrodes, and nano-simulations of electrode surface reactions, using COMSOL Multiphysics and several other applications relating to molecular dynamics and ab-initio molecular dynamics. We use simulation technology with validation tests and measurements from material design and selection to battery design and evaluation.
Tatsuya Yamaue, PhD, conducts his research at the Kobelco Research Institute, Inc.’s Engineering Mechanics Division, Technology Headquarters.