In next-generation battery development, a persistent bottleneck slows progress: a deep and silent divide separating material scientists, electrochemical modelers, and BMS (Battery Management System) circuit designers. While material scientists optimize solid diffusion properties and electrolyte conductivity, and modelers construct partial differential equations (PDEs), circuit engineers are left utilizing extremely simplified equivalent circuit models (ECMs) with little to no physical foundation.
Having previously worked as a semiconductor process and device engineer before entering the battery field, this disconnect feels highly unnatural. In the semiconductor industry, bridging device physics to circuit simulation is a standard and robustly automated process. This article explores why we must bridge this gap for lithium-ion batteries and how we can use the Single Particle Model with Electrolyte (SPMe) to generate SPICE-compatible equivalents directly from materials and structure parameters.
- Bridging Physics to Circuit Simulators: Translates non-linear electrochemical PDE physics (SPMe) into robust equivalent subcircuits (`.subckt`) compatible with standard circuit tools like LTspice.
- Coupled Electrochemical Overpotentials: Accurately represents SOC-dependent open-circuit voltage curves, Butler-Volmer charge-transfer kinetics, and electrolyte diffusion concentration dynamics via behavioral voltage sources and RC networks.
- BMS & Thermal Co-Simulation: Empowers power electronics engineers and BMS developers to simulate realistic cell physical dynamics and temperature dependencies during early R&D phases before physical cell fabrication.
1. Lessons from LSI: The Standard of Model Extraction
In LSI (Large Scale Integration) design, transistor developers do not hand over raw physical measurements or raw material specs to chip designers. Instead, device engineers perform systematic parameter extraction to represent LSI element behaviors (such as I-V and C-V curves) in standard models like BSIM (Berkeley Short-channel IGFET Model). These models are delivered directly to circuit designers as SPICE model cards.
This allows circuit engineers to simulate millions of transistors in complex configurations with complete trust that the electrical responses align with the physical limitations of the fabrication process. In the battery industry, however, this pipeline is virtually non-existent. BMS designers are usually forced to purchase cells, perform their own transient load tests, and fit mathematical curve equations (e.g., empirical RC parallel pairs) that completely ignore the physical structure—such as electrode thicknesses, active material volume fractions, and temperature-activated diffusivities.
When every device buyer is forced to reconstruct heuristic models individually, development slows down. If material properties and battery structural designs could be mapped straight into equivalent circuits, BMS designers could run simulations on battery cells that are still in the R&D phase, creating a powerful feedback loop.
2. The Unique Hardship of Battery Modeling
Admittedly, translating a battery cell to SPICE is considerably more difficult than characterizing a static silicon junction. Batteries exhibit dynamic, state-dependent behaviors that SPICE simulators were not originally designed to handle natively:
- Non-linear open-circuit potentials: The equilibrium voltage depends heavily on local stoichiometry ($x$, $y$) which varies as lithium ions diffuse in and out of the particles.
- Transient transport dynamics: Lithium-ion diffusion in the solid particles ($D_s$) and liquid phase ($D_e$) introduces long memory effects and delayed voltage drops that cannot be perfectly represented by a few discrete capacitors and resistors.
- Arrhenius temperature sensitivity: Kinetics (exchange currents $k$) and diffusion scale non-linearly with temperature, requiring dynamic activation energy calculations.
3. The Limit of Two-Terminal Measurements: Why 4-RC Cannot Be Easily Measured Directly
In standard battery testing, engineers connect a cell to a cycler or an EIS (Electrochemical Impedance Spectroscopy) analyzer using just two terminals (the positive and negative contacts). The resulting measurement is a lumped sum of all internal overpotentials combined:
- Anode kinetics + Cathode kinetics
- Anode solid-phase diffusion + Cathode solid-phase diffusion
- Electrolyte transport + contact resistances
This creates a mathematically ill-posed inverse problem. When fitting equivalent circuit models to terminal responses, the optimizer cannot differentiate whether a slow transient response comes from solid diffusion inside the anode particles or the cathode particles, or if a fast voltage drop is caused by charge-transfer kinetics at the positive or negative electrode interface. As a result, standard fits collapse these parameters into generic, mathematically-lumped RC blocks that lack physical meaning, limiting their predictive accuracy at different currents, states of charge, or temperatures.
To physically separate anode and cathode contributions, experimentalists must construct a complex three-electrode cell containing a reference electrode (e.g., lithium metal). This process is highly labor-intensive, unstable over long cycles, and practically impossible to perform on commercial production cells without destructive teardown.
This is where the power of physics-based models shines. Inside our SPMe simulation, the anode potential ($V_n(t)$) and cathode potential ($V_p(t)$) are calculated independently as separate field variables. By treating the simulation as a virtual three-electrode measurement, we can isolate and extract the individual parameters for four distinct, physically meaningful RC systems. Rather than lumped cell-level parameters, we build two separate 2-RC models (one for each electrode) that connect in series to form the complete 4-RC subcircuit (plus the ohmic resistance $R_0$):
- Anode Charge Transfer ($R_{1,n}, C_{1,n}$): Fast kinetics at the negative electrode interface (part of anode 2-RC).
- Anode Solid Diffusion ($R_{2,n}, C_{2,n}$): Slow particle-level diffusion inside the negative electrode (part of anode 2-RC).
- Cathode Charge Transfer ($R_{1,p}, C_{1,p}$): Fast kinetics at the positive electrode interface (part of cathode 2-RC).
- Cathode Solid Diffusion ($R_{2,p}, C_{2,p}$): Slow particle-level diffusion inside the positive electrode (part of cathode 2-RC).
4. The Extraction Logic: How Physical Parameters Translate to SPICE
The SPMeToSPICE Web Tool automates this mapping using a robust extraction algorithm. Rather than relying on simple mathematical approximations, the tool uses the following logic flow to compile the LTspice-ready subcircuit:
Step 1: Virtual HPPC Characterization
First, the tool runs a "Virtual HPPC (Hybrid Pulse Power Characterization)" protocol on the physics-based SPMe model. The simulation divides the SOC range (10% to 100%) into discrete steps. At each step, the model is subjected to a brief, highly sensitive 10-second discharge/charge pulse followed by a rest period. Crucially, the internal potentials of the anode and cathode are monitored independently at microsecond intervals.
Step 2: Dual-Electrode 2-RC Exponential Fitting
During the pulse and relaxation phases, the transient voltage curves for each electrode ($V_n(t)$ and $V_p(t)$) are isolated. The instantaneous step represents the ohmic contribution, while the transient curve for each electrode is fitted to a 2-RC (two-time-constant) exponential relaxation model using non-linear least squares:
$V_{trans}(t) = I \cdot R_1 \cdot (1 - e^{-t / \tau_1}) + I \cdot R_2 \cdot (1 - e^{-t / \tau_2})$
Here, the two time constants $\tau_1 = R_1 C_1$ (fast charge-transfer kinetics) and $\tau_2 = R_2 C_2$ (slow solid-phase diffusion) are extracted for the negative and positive electrodes individually. Resolving this separately for both electrodes yields the base resistance and capacitance tables as functions of SOC at the reference temperature (25°C). Connecting the anode 2-RC and cathode 2-RC in series forms the complete cell 4-RC model.
Step 3: Multi-Temperature Arrhenius Mapping
To capture thermal behavior, the tool repeats the virtual HPPC simulation at a lower temperature (typically 0°C). By comparing the extracted resistance tables at 25°C and 0°C, the algorithm extracts the exact Activation Energy ($E_a$) for each individual resistor using the Arrhenius equation:
$R(T) = R(T_{ref}) \cdot \exp\left( \frac{E_a}{R_g} \left( \frac{1}{T} - \frac{1}{T_{ref}} \right) \right)$
This step ensures that the SPICE model does not just use generic temperature coefficients, but inherits the precise thermodynamic sensitivities (e.g., diffusivity and kinetic activation energies) defined in the physical model.
Step 4: Compilation into LTspice Netlist
Finally, the extracted data tables (SOC vs. $R_0, R_1, C_1, R_2, C_2$ for both electrodes) and the calculated activation energies ($E_a$) are written into a standard LTspice subcircuit format (`.SUBCKT`). The temperature dependence is computed dynamically during circuit simulation using LTspice’s behavioral sources, scaling the resistances on the fly according to the voltage applied to the model's TEMP pin.
5. SPMe to SPICE: Creating the Bridge
The Single Particle Model with Electrolyte (SPMe) provides the perfect mathematical compromise. It simplifies the DFN (Doyle-Fuller-Newman) model by assuming uniform reaction distribution across the electrodes, yet it still captures key physical constraints—including liquid-phase diffusion and ohmic electrolyte resistance.
By discretizing the radial diffusion equation within the active material spheres and the concentration gradients in the liquid electrolyte, we can construct a set of equivalent network equations that SPICE can natively solve. The mapping converts physical phenomena into electrical equivalents:
| Physical Phenomenon | Governing Variable / Parameter | SPICE Equivalent Element |
|---|---|---|
| Solid-Phase Diffusion | Spherical diffusion equation ($D_s$) | RC Ladder Network (discretized transmission line representing diffusion resistance/capacity) |
| Electrochemical Kinetics | Butler-Volmer equation ($j_0$, $\alpha$) | Non-linear Voltage-Controlled Current Sources (B-sources) |
| Liquid-Phase Concentration & Ohmic Losses | Electrolyte diffusivity ($D_e$) & conductivity ($\kappa_e$) | Serial resistors and voltage offsets linked to state-dependent lookup blocks |
| Open Circuit Potentials | Positive/Negative OCP tables ($U_p(y)$, $U_n(x)$) | State-of-Charge (SoC) integration capacitors feeding non-linear voltage sources |
6. Empowering Designers with SPMeToSPICE
To demonstrate this concept, we developed the SPMeToSPICE Web Tool. This application takes material specifications (like electrode thickness, active material fractions, and solid diffusivity) and generates a ready-to-run SPICE subcircuit netlist (`.sub` file).
By simply importing the netlist into standard simulators like LTspice, circuit designers can immediately analyze:
- Dynamic cell voltage variations during high-frequency pulse charging.
- BMS thermal management behavior, as internal heat generation calculations are linked directly to kinetic and electrolyte overpotentials.
- True physical limits of cells, protecting BMS hardware from over-discharge under transient loads.
We hope this simple bridging tool sparks interaction across disciplines. When materials, simulations, and circuit design speak the same language, battery development will move at a pace closer to the semiconductor revolution.
Convert Physical Models to SPICE Netlists Instantly
Export your battery designs, material parameters, and structure layouts into functional LTspice-ready equivalent circuits using our web tool.
Try SPMe to SPICE Tool →