ARC data without Arrhenius fitting is decorative — the kinetic parameters are what you need to simulate propagation, and the simulation is what you need to prove AIS-156 compliance without burning a pack.
- Accelerating Rate Calorimetry (ARC) under adiabatic conditions is the standard method to extract Arrhenius kinetic parameters {Ea, A, ΔH} per reaction stage — these are the required inputs for thermal runaway propagation simulation.
- The Frank-Kamenetskii criticality condition shows larger cells reach thermal runaway at lower temperatures; a 280 Ah prismatic cell enters runaway sooner than a 21700 cell of the same chemistry, requiring cell-specific ARC characterisation.
- CFD propagation simulation using ARC kinetics costs a fraction of physical pack tests and enables design iteration before committing to a test build — physical tests then validate, not discover.
- AIS-156 Phase 2 compliance requires co-optimisation of BMS detection speed (<30 s), inter-cell thermal barrier selection, and venting path geometry — no single element is sufficient alone.
- HF yield of 10–200 mg per Wh of cell energy from LiPF₆ decomposition creates an acute toxicology risk for first responders; Indian fleet operators must brief local emergency services and ensure SCBA availability.
This is the rigorous engineering article on thermal runaway. The focus is on quantitative methods: ARC testing methodology and data interpretation, Arrhenius parameter extraction, Frank-Kamenetskii criticality analysis, CFD-based propagation simulation for pack certification, and the specific design choices that determine whether a pack passes AIS-156 Phase 2 with margin or barely passes with 4 minutes 58 seconds of warning.
ARC Testing: Methodology and Interpretation
Accelerating Rate Calorimetry (ARC) is the standard method for characterising the thermal runaway kinetics of battery cells and materials. The instrument operates under adiabatic conditions — it precisely heats the chamber to match the sample temperature so that no heat flows between sample and environment. Under these conditions, any self-heating is exclusively from the sample's exothermic reactions.
Standard ARC protocol for a full cell:
Charge to target SOC (typically 100% and 50% SOC separately). Rest 4 hours at 25°C. Record initial OCV. Mount in ARC calorimeter with calibrated heat capacity.
ARC heats cell by 5°C steps, holds 30 min, then seeks self-heating. If dT/dt < 0.02°C/min, proceed to next step. If > 0.02°C/min, ARC switches to exotherm-following mode.
ARC matches heater power to maintain adiabatic conditions. Records T(t) and dT/dt during self-heating cascade.
Temperature at which dT/dt first exceeds 0.02°C/min consistently. This is the onset of Stage 1 (SEI decomposition onset).
For each reaction stage, fit ln(dT/dt) vs 1/T to extract Ea and pre-exponential A per stage.
Integrate the temperature-time curve with the calorimeter's heat capacity calibration to get total ΔH per stage and overall.
Interpreting ARC data for design:
A typical NMC 811 cell at 100% SOC shows:
- Onset at ~80–90°C (SEI Stage 1)
- First acceleration at ~130°C (electrolyte + separator Stage 2)
- Rapid acceleration at ~175°C (cathode Stage 3 onset)
- Maximum dT/dt at ~250–350°C: 50–500°C/min depending on cell size
- Total heat release: 400–700 J/g of cell mass
A typical LFP cell at 100% SOC shows:
- Onset at ~85–100°C (SEI + electrolyte)
- No distinct cathode decomposition peak below 270°C
- If heated further: slow decomposition begins at 260–270°C with dT/dt of 5–20°C/min
- Total heat release: 200–400 J/g
The Arrhenius kinetic parameters {Ea, A, ΔH} for cathode decomposition are SOC-dependent because the degree of lithium delithiation determines how reactive the cathode is at elevated temperature. An NMC 811 cell at 80% SOC has a Stage 3 onset approximately 15–20°C higher than the same cell at 100% SOC. If your pack's BMS limits charge to 90% max SOC, using 100% SOC ARC data in your propagation model is conservative but may over-size required thermal barriers — adding cost and mass. The correct approach is to run ARC at the maximum operating SOC of the specific pack design and use those parameters for compliance simulations.
Extracting Arrhenius Parameters
For first-order reaction kinetics, the self-heating rate in the adiabatic ARC has the form:
dT/dt = (ΔH × A × c₀) / Cp × (1 - (T - T₀)/ΔT_ad) × exp(-Ea/RT)
# ARC test data: self-heating rate vs temperature (Arrhenius model)
import numpy as np
def arrhenius_self_heating(T_K: float, A: float, Ea_J: float) -> float:
"""
Self-heating rate model: dT/dt = A * exp(-Ea / R*T)
A: pre-exponential [deg C/min]
Ea: activation energy [J/mol]
R: gas constant 8.314 J/mol/K
"""
R = 8.314
return A * np.exp(-Ea_J / (R * T_K))
# Fit parameters for NMC811 at 4.3V (from published ARC data)
# Ea ~= 100-150 kJ/mol for SEI decomposition
A_fit = 1e12 # pre-exponential (high for fast reaction)
Ea_fit = 130e3 # J/mol
print(f"{'Temp (deg C)':>14} {'dT/dt (deg C/min)':>20}")
print("-" * 36)
for T_C in range(80, 220, 10):
T_K = T_C + 273.15
rate = arrhenius_self_heating(T_K, A_fit, Ea_fit)
bar = "#" * min(int(rate * 5), 30)
print(f"{T_C:>14} {rate:>18.4f} {bar}")where ΔT_ad = ΔH × c₀ / Cp is the adiabatic temperature rise, T₀ is the onset temperature, and c₀ is the initial reactant concentration.
For the linear approximation in the early exotherm (small conversion):
ln(dT/dt) ≈ ln(ΔH × A × c₀ / Cp) - Ea/RTPlotting ln(dT/dt) vs 1/T gives a straight line with slope -Ea/R. The intercept gives ln(A). This Arrhenius fit is performed separately for each stage identified in the ARC data, producing a set {Ea_i, A_i, ΔH_i} for each stage i.
These parameters are the direct input to pack-level thermal models — specifically, they define the volumetric heat generation rate of a cell in runaway as a function of temperature and degree of reaction.
ARC characterisation must be performed at the operating SOC of the cell, not just at 100%. An NMC 811 cell at 80% SOC has a Stage 3 onset ~15–20°C higher than at 100% SOC. A pack designed with a 90% max SOC limit should use 90% SOC ARC data for propagation simulations — using 100% SOC data is conservative but may over-size the required thermal barriers for the actual operating condition.
Frank-Kamenetskii Criticality for Pack Geometry
The FK criticality condition determines the temperature at which an exothermic body (cell or group of cells) will self-accelerate to thermal runaway without additional external heat input. For a slab geometry (approximately correct for prismatic cells):
δ = (Ea × Q₀ × δ_char²) / (RT₀² × λ) = 0.88 [criticality threshold]where Q_0 = ΔH × A × exp(-Ea/RT₀) is the heat generation rate per unit volume at the reference temperature T₀, δ_char is the half-thickness of the slab, and λ is the thermal conductivity.
Critical implications for pack design:
- Larger cells reach criticality at lower temperatures. A 280 Ah prismatic cell has δ_char 3–5× larger than a 50 Ah prismatic cell. If both cells have the same chemistry and therefore the same Ea and Q_0, the larger cell enters runaway at a lower temperature. Fleet operators using large-format Chinese LFP cells (280 Ah, 302 Ah) should use the cell-specific Ea/Q_0 values from ARC testing for their specific cell format.
- Pack thermal conductivity directly affects criticality. Higher in-plane thermal conductivity (from cooling plates or thermally conductive TIM) raises the effective λ in the FK equation, increasing the critical temperature threshold. This is why cooling system quality matters even for TR onset, not just for temperature management during normal operation.
The FK criticality condition includes the thermal conductivity λ of the surrounding medium in the denominator. Higher effective thermal conductivity — from cooling plates, thermally conductive TIM, or conductive barrier materials — increases the critical temperature threshold at which a cell transitions to self-accelerating runaway. This means cooling system quality directly influences where runaway initiates, not just how fast it progresses. Conversely, insulating intumescent barriers placed between cells reduce λ locally, which lowers the threshold — but intumescent materials expand at high temperature and change their k dramatically, so propagation models must use temperature-dependent material properties, not room-temperature datasheet values.
CFD-Based Propagation Modelling
For AIS-156 Phase 2 compliance, a physical propagation test is required. But simulation is essential for design iteration before the physical test — burning a pack to failure costs ₹15–50 lakh and months of test time per iteration.
Propagation simulation workflow:
# Propagation modelling: finite difference thermal model
import numpy as np
def simulate_propagation_1d(n_cells: int = 10,
dx_m: float = 0.025, # cell pitch 25mm
dt_s: float = 0.1,
t_total_s: float = 600.0,
trigger_cell: int = 0) -> list:
"""
1D finite difference model of cell-to-cell thermal propagation.
Returns list of (time_s) when each cell reaches TR onset (185 deg C).
"""
k_module = 1.5 # W/m·K effective thermal conductivity
rho_Cp = 2.1e6 # J/m3·K (volumetric heat capacity)
alpha = k_module / rho_Cp # thermal diffusivity
T = np.ones(n_cells) * 25.0 # initial temperature [deg C]
T[trigger_cell] = 600.0 # trigger cell already in runaway
propagation_times = [None] * n_cells
propagation_times[trigger_cell] = 0.0
t = 0.0
while t < t_total_s:
T_new = T.copy()
for i in range(1, n_cells - 1):
if propagation_times[i] is not None:
T_new[i] = max(T_new[i], 600.0) # cell is burning
else:
laplacian = (T[i+1] - 2*T[i] + T[i-1]) / dx_m**2
T_new[i] += alpha * dt_s * laplacian
if T_new[i] >= 185.0 and propagation_times[i] is None:
propagation_times[i] = t
T = T_new
t += dt_s
return propagation_times
times = simulate_propagation_1d()
print("Cell-to-cell propagation times (trigger = Cell 0):")
for i, t in enumerate(times):
if t is not None:
print(f" Cell {i}: {t:.0f} s ({t/60:.1f} min)")
else:
print(f" Cell {i}: did not reach TR in simulation window")- Cell thermal model: 3D FEM model of cell with ARC-derived heat generation function Q_gen(T, degree_of_reaction). Typically a lumped-parameter model for the cell interior, finite element for the cell surface.
- Inter-cell heat transfer model: Thermal conductivity of the TIM (Thermal Interface Material) or air gap between cells. For intumescent sheets, the thermal conductivity changes with temperature (expands and reduces k at elevated T) — this must be measured separately.
- Boundary conditions: Pack enclosure thermal mass and conductivity, ambient temperature, cooling system heat removal capability.
- Trigger: Apply ARC-equivalent temperature boundary condition to one cell (simulating a nail penetration or overcharge trigger). Track heat spread to adjacent cells.
- Pass criterion: Time from first cell reaching Stage 3 to any cell in the passenger-compartment-adjacent zone reaching 200°C (proxy for smoke/fire hazard) must exceed 5 minutes.
| Design Element | Effect on Propagation Time | Implementation Cost |
|---|---|---|
| Intumescent sheet between cells (1 mm) | +2–4 min | ₹200–500 per cell pair |
| Aerogel sheet (5 mm) | +4–8 min | ₹800–1,500 per cell pair |
| LFP vs NMC 811 chemistry | +15–30 min (LFP baseline) | Addressed at cell selection |
| 200 W/K cooling system vs 100 W/K | +1–3 min | Cooling system capex |
| Module-level venting channel | Directs hot gas, +1–2 min | ₹100–300 per module |
| BMS gas detection (CO/HF) triggering max cooling | +0.5–1.5 min | ₹500–1,000 per pack |
Thermal barrier materials must be tested at actual temperature cycles and compression levels, not just at room temperature. Intumescent materials have temperature-dependent k. Aerogel can degrade under the compressive forces applied in a pack (0.5–2 MPa typical). TIM (thermal interface material) k is specified at controlled pressure — real pack assembly pressure variation of ±20% causes significant k variation. All propagation simulations must use field-validated material properties, not supplier datasheet values at standard conditions.
AIS-156 Phase 2 Compliance Architecture
AIS-156 Phase 2 requires that a single cell thermal runaway event does not cause:
- Explosion (pressure release from passenger compartment)
- Fire visible in or near passenger compartment
- Within 5 minutes of onset detection by the BMS warning system
Compliance path:
# AIS-156 Amendment 2 -- thermal propagation test pass/fail criteria
import json as _json
def evaluate_ais156_propagation(test_data: dict) -> dict:
"""
Evaluate thermal propagation test results against AIS-156 Amendment 2.
Pass criteria (per AIS-156:2023):
1. No explosion within 5 minutes of TR trigger
2. No fire external to battery pack within 5 minutes
3. Cell-to-cell propagation alarm must activate before pack fire
4. Vehicle must retain structural integrity for occupant egress (1 min)
"""
results = {}
# Check 1: No explosion
results["no_explosion"] = not test_data.get("explosion_detected", False)
# Check 2: External fire timing
ext_fire_t = test_data.get("external_fire_time_s", 9999)
results["external_fire_timing"] = ext_fire_t > 300 # > 5 min
# Check 3: Warning before fire
alarm_t = test_data.get("alarm_trigger_time_s", 9999)
results["alarm_before_fire"] = alarm_t < ext_fire_t
# Check 4: Egress time (structural integrity)
struct_fail_t = test_data.get("structural_failure_time_s", 9999)
results["egress_time_met"] = struct_fail_t > (ext_fire_t + 60)
overall_pass = all(results.values())
results["OVERALL"] = "PASS" if overall_pass else "FAIL"
return results
# Example test result
test_result = {
"explosion_detected": False,
"external_fire_time_s": 420, # 7 minutes -- exceeds 5 min limit
"alarm_trigger_time_s": 85, # alarm at 1.4 min
"structural_failure_time_s": 620, # structure holds 3.3 min after fire
}
results = evaluate_ais156_propagation(test_result)
print("AIS-156 Amendment 2 Compliance Check:")
print("=" * 42)
for criterion, passed in results.items():
status = "PASS" if passed is True else ("FAIL" if passed is False else str(passed))
print(f" {criterion:<35} {status}")Full ARC data at operating SOC and temperatures. Extract {Ea_i, A_i, ΔH_i} per stage. Separate LFP and NMC data required if mixed chemistry.
3D model with inter-cell barriers, venting paths, cooling system. Simulate 5 trigger scenarios: overcharge, nail, overtemperature, external fire, internal short.
Measure TIM/intumescent k vs temperature at actual pack compression. Validate against simulation input assumptions.
Specify detection algorithm (temperature rate-of-rise + gas sensor minimum). Measure detection time from ARC-equivalent trigger to BMS Level 3 alert. Must be <30 seconds for 5-minute compliance to be achievable.
Nail penetration on worst-case cell in pack (highest SOC, highest temperature, nearest to passenger-adjacent zone). Instrument with thermocouples on all cells, gas sampling in pack headspace, and passenger compartment temperature sensor. Record time from cell TR to passenger zone hazard.
ARC data, simulation methodology, material characterisation, BMS algorithm specification, and physical test data. Submit to ARAI/ICAT for AIS-156 type approval.
AIS-156 Phase 2 requires that a thermal runaway event in one cell does not produce a passenger-compartment fire or explosion within 5 minutes of onset. For this to be achievable, the BMS must detect the event and trigger its maximum cooling and isolation response fast enough that the propagation cascade — which depends on cell spacing, barrier material quality, and chemistry — has insufficient time to reach the passenger zone. Empirically, with the best available LFP barrier designs and a well-optimised propagation architecture, BMS detection must occur within 30 seconds of Stage 2 onset to ensure 5-minute compliance. This is why gas sensing (CO/HF detection at Stage 2) is architecturally superior to temperature-only detection, which typically triggers 60–120 seconds later.
Field Failure Investigation Framework
When a TR incident occurs in the field, the investigation must distinguish:
Abuse-caused: External physical damage, overcharge from faulty charger, severe overdischarge. Evidence: pack enclosure deformation, BMS fault log showing limit exceedance before TR, charger communication records.
Manufacturing defect: Internal contamination creating a short circuit from first use. Evidence: TR occurring without any measurable abuse trigger, multiple cells failing in same lot, contamination particles in post-failure residue analysis.
Degradation-induced: Cell degraded to the point where normal use constitutes abuse for its remaining condition. Evidence: high cycle count, capacity significantly below nominal, elevated internal resistance history in BMS logs.
Indian commercial EV TR incidents in 2022–2023 were disproportionately in the 2W segment and traced primarily to non-certified cell supply (counterfeit or off-spec cells) and BMS designs without adequate current limiting during fast charge. The 4W and commercial vehicle segment with AIS-156-certified packs had significantly fewer incidents per vehicle. The certification process works — the gap is enforcement and supply chain integrity in the lower-cost segment.
Key Takeaways
- ARC testing under adiabatic conditions provides Arrhenius parameters {Ea, A, ΔH} per reaction stage — the required inputs for propagation simulation. ARC data without kinetic parameter extraction is not useful for compliance modelling.
- The Frank-Kamenetskii criticality condition shows that larger cells reach thermal runaway at lower temperatures than smaller cells of the same chemistry. Large-format prismatic cells (280 Ah) need their specific FK analysis, not data extrapolated from smaller cylindrical formats.
- CFD propagation simulation using ARC kinetic parameters is the cost-effective path for design iteration before physical tests. Physical testing validates the simulation — it does not replace the design process.
- AIS-156 Phase 2 compliance requires co-optimisation of BMS detection speed (<30 s to Level 3 for 5-minute compliance), inter-cell barrier material k and expansion characteristics, and venting path design. No single element is sufficient alone.
- HF toxicology from LiPF₆ decomposition (10–200 mg/Wh) is a significant firefighter safety concern. Indian fleet operators must brief local emergency services on HF hazard and SCBA requirements before deploying large EV fleets in specific operating areas.
Part of the thermal Series
Frequently Asked Questions
What is the difference between ARC testing and DSC for thermal characterisation?
How do you extract Arrhenius parameters from ARC data?
What does the Frank-Kamenetskii criticality condition mean for pack geometry?
What are the key design decisions for AIS-156 Phase 2 compliance?
What is the HF toxicology concern in thermal runaway events and how does it affect firefighting protocol?
References
- Huang, P. et al. (2016) — Thermal runaway behavior during overcharge for large-format lithium-ion batteries with different packaging patterns, Journal of Energy Storage
- Golubkov, A.W. et al. (2014) — Thermal runaway of commercial 18650 Li-ion batteries with LFP and NCA cathodes, RSC Advances
- AIS 156 Amendment 4 — Battery Pack Safety Requirements for Electric Vehicles in India
- UN ECE Regulation No. 100 — Safety Requirements for Electric Powertrains