Physics Based Observers for Real Time Anode Overpotential Estimation
Real-time physics-based observers prevent lithium plating by dynamically modulating fast-charge current to maintain positive anode intercalation potential.

Interphase
Fast charging pushes lithium-ion negative electrodes toward thermodynamic thresholds where side reactions occur. Inside a closed commercial cell, the terminal voltage recorded across the tabs reflects the sum of positive electrode potential, negative electrode potential, and total internal ohmic drops. Terminal voltage masks internal polarization.
When current flows into a graphite anode, the local electrical potential of the graphite matrix drops relative to the surrounding liquid electrolyte. This difference defines the anode potential Ea, calculated as the equilibrium open-circuit potential minus overpotential contributions from kinetic, ohmic, and mass-transport limitations. If local overpotential suppression drives the anode potential below zero volts against the reference lithium couple Li/Li+, intercalation into the host graphite structure ceases to be the sole favored reaction.
Metallic lithium deposition begins on the graphite surface, initiating rapid capacity loss and potential dendrite growth.
Lithium plating destroys graphite anodes. Avoiding this failure mode while maximizing charging speed requires real-time knowledge of internal interfacial potentials. Because internal three-electrode probes cannot survive commercial cell manufacturing or long-term field operation, battery management systems must calculate anode overpotential through physical state estimation rather than direct sensing.
The total anode overpotential aggregates three distinct physical phenomena: charge-transfer overpotential governed by Butler-Volmer kinetics at the solid-electrolyte boundary, ohmic potential drops across the solid-electrolyte interphase layer and pore electrolyte, and concentration polarization resulting from slow solid-phase lithium diffusion inside graphite particles.
At 4C fast charge and 25°C, solid-state diffusion polarization accounts for more than 60 percent of total anode potential suppression in 20-micrometer graphite particles.

Mechanistic Origin of Anode Intercalation Potentials
During charge, electrons move through the copper current collector into graphite particles while lithium ions cross the porous separator. The local potential difference between the solid graphite surface φs and the adjacent liquid electrolyte φe determines the driving force for intercalation. The net thermodynamic anode potential Eanode relative to Li/Li+ is defined by:
Eanode = Ua(thηs) + ηct + Δφsei
where Ua(thηs) represents the open-circuit potential of graphite evaluated at the surface stoichiometry thηs = cs,surf / cs,max, ηct is the activation overpotential derived from the Butler-Volmer current density equation, and Δφsei represents the voltage drop across the resistive surface film. As charging current increases, ηct shifts negative and cs,surf approaches saturation, driving Ua(thηs) down toward zero volts. Once Eanode drops below zero volts, the reaction overpotential for lithium plating turns positive, initiating phase transformation from ionic lithium to metallic lithium on particle surfaces.

Kinetic and Mass Transport Constraints during Fast Charge
High current rates establish steep concentration profiles within the liquid electrolyte channels. Diffusion gradients restrict fast charging. Solid particles develop internal concentration fronts.
The liquid-phase ionic concentration ce drops inside the electrode pores furthest from the separator, increasing liquid-phase ohmic polarization φe. Simultaneously, solid-phase mass transport is constrained by the solid diffusion coefficient Ds. When spherical graphite particles face rapid charge pulses, lithium ions accumulate on the outer particle boundary faster than solid diffusion transports them toward the particle core. This surface concentration buildup drops the local open-circuit potential Ua(thηs) long before the average particle stoichiometry reaches full capacity.
| C-Rate | Ohmic Drop (mV) | Charge-Transfer (mV) | Solid Diffusion (mV) | Net Anode Potential (mV vs Li/Li+) |
|---|---|---|---|---|
| 1.0C | 12.4 | 18.6 | 22.1 | 66.9 |
| 2.0C | 24.8 | 31.2 | 48.5 | 15.5 |
| 3.0C | 37.2 | 42.8 | 79.3 | -19.3 |
| 4.0C | 49.6 | 53.1 | 114.6 | -97.3 |
Standard cell procurement contracts append Clause 8.4, which voids fast-charging warranty coverage if the battery management system fails to maintain predicted graphite intercalation potential above zero millivolts versus reference lithium.

Mesh
Continuous partial differential equations describing concentration dynamics demand significant computational memory. The classic Pseudo-Two-Dimensional model formulated by Doyle, Fuller, and Newman captures solid-phase particle diffusion, liquid-phase mass transport, charge conservation, and electrochemical kinetics using coupled non-linear partial differential equations. Executing these full continuous domain models in real time on standard automotive microcontrollers with core clock speeds under 200 MHz is impossible.
Model order reduction saves microcontrollers. Truncation introduces high frequency phase shifts.
Physics-based state observers rely on discrete order reduction techniques that project continuous spatial continuum variables onto low-dimensional state-space matrices. Spatial discretization across the anode thickness dimension and the radial particle dimension transforms partial differential equations into a set of ordinary differential equations. Preserving physical parameters such as electrode thickness, porosity, particle radius, and temperature-dependent diffusion coefficients allows reduced-order models to maintain accurate voltage predictions while operating within real-time computational constraints.
Model order reduction that sacrifices liquid-phase concentration dynamics corrupts overpotential estimates precisely when high C-rate pulses demand maximum precision.

Reduction of Continuum Transport Equations for Real Time Execution
Standard Pseudo-Two-Dimensional models employ coupled balance equations across spatial dimensions. Solid-phase diffusion governed by Fick’s second law in spherical coordinates requires solving concentration distributions across radial coordinates r at every spatial location x through the electrode thickness. Spectral modal decomposition and Padé approximation techniques reduce spherical solid-phase diffusion to low-order polynomial expressions.
Approximating radial concentration distributions using parabolic or cubic profiles eliminates the radial spatial grid entirely, replacing radial diffusion partial differential equations with two state variables per electrode volume element: average concentration cs,avg and surface concentration cs,surf.
Liquid-phase transport reduction uses spectral volume discretization to approximate electrolyte concentration ce(x,t) across discrete boundary nodes. Truncating spatial modes reduces state dimension from over fifty continuum nodes to fewer than twelve state variables, bringing execution time per time-step below five milliseconds on 32-bit floating-point processors.

State Space Formulations for Microcontroller Integration
Low-order matrix approximations translate continuous radial solid diffusion into discrete differential-algebraic structures. The resulting state equations adopt standard linear state-space forms modified by non-linear output mapping functions:
- Extract the coupled partial differential equations governing solid-phase radial diffusion and liquid-phase concentration transport.
- Apply Chebyshev polynomial projection or spectral modal expansion across the radial particle domain.
- Truncate high-frequency spatial eigenvalues while retaining modes contributing to low-frequency polarization response.
- Discretize the electrolyte transport equations along the electrode thickness axis using finite volume approximation.
- Construct linear state-space system matrices representing transient solid surface concentrations and liquid potential drops.
It remains unclear whether polynomial state truncation maintains mathematical stability across extreme temperature transitions where electrolyte transport properties undergo non-linear phase shifts.

Tracking
Real-time state estimation relies on feedback loops to correct errors between model outputs and measured terminal voltage. Open-loop model execution suffers from parameter drift, sensor bias, and unmodeled thermal effects. An observer couples the reduced-order physical model with closed-loop gain matrices, adjusting internal state variables dynamically based on the residual error between measured cell tab voltage and model-predicted terminal voltage.
Extended Kalman Filters, Unscented Kalman Filters, and Sliding Mode Observers represent the dominant algorithmic structures for real-time state estimation.
Parameter drift compromises observer accuracy. Temperature gradients distort state tracking. As cells cycle, internal physical properties shift away from baseline laboratory calibration values.
The solid diffusion coefficient Ds decays as mechanical microcracking disrupts diffusion paths within graphite grains, while the solid-electrolyte interphase resistance Rsei grows through continuous solvent reduction. A physics-based observer operating with fixed initial parameters produces expanding overpotential estimation errors over cycle life, eventually predicting safe anode potentials while the physical electrode enters plating territory.
Observer feedback algorithms correct state trajectories by mapping measured voltage errors back into state variable space. Because terminal voltage is an aggregate output, the observer uses partial sensitivity derivatives of terminal voltage with respect to individual state variables to distribute state corrections accurately.
Uncompensated temperature gradients between cell core and surface introduce state estimation errors larger than the voltage safety margin required to prevent lithium deposition.

Observer Architecture and Closed Loop State Correction
Extended Kalman filters linearize non-linear electrochemical differential equations at every sampling step. The system state vector xk includes solid surface concentration, liquid concentration across discrete zones, and core cell temperature. The state update sequence follows standard prediction and correction stages:
xk|k-1 = f(xk-1, uk-1)
Pk|k-1 = Fk-1 Pk-1 Fk-1T + Qk
Kk = Pk|k-1 HkT (Hk Pk|k-1 HkT + Rk)-1
xk|k = xk|k-1 + Kk (yk – g(xk|k-1, uk))
where Fk-1 is the Jacobian matrix of partial derivatives of state transition function f, Hk is the output matrix linearized around current state predictions, Qk is process noise covariance, Rk is sensor measurement noise covariance, and Kk represents the calculated Kalman gain. The estimated surface concentration cs,surf extracted from state vector xk|k directly yields real-time anode open-circuit potential Ua(thηs), enabling real-time computation of Eanode.

Parameter Drift and Aging Adaptation Metrics
Solid-electrolyte interface growth consumes cyclable lithium ions while thickening the resistive surface layer. Tracking anode overpotential accurately over two thousand cycles requires adaptive observer architectures that identify parameter drift concurrently with state tracking. Dual Extended Kalman Filters run two parallel estimation loops: one tracking fast state variables on millisecond scales, and one updating slow parameter variables such as active lithium inventory and effective diffusion coefficients on minute scales.
- Diffusion Coefficient Mismatch causes the estimated particle surface concentration to lag behind physical ion accumulation during aggressive current transients.
- Electrolyte Conductivity Suppression reduces predicted ionic mobility inside electrode pores, causing unexpected potential collapse near the current collector.
- Solid Interphase Resistance Growth creates unmodeled ohmic potential drops across aging cycles, shifting calculated reaction kinetics away from actual values.
- Active Material Loss Bias alters the stoichiometry mapping between measured cell state of charge and actual local graphite intercalation state.
Pack vendors frequently cite unpredictable solid-phase diffusion coefficient degradation across batch variations as the primary reason their internal state estimators produce divergence under fast charge.

Probe
Direct measurement of internal negative electrode potential inside a commercial cell remains impossible without physical modification. Validating observer estimates requires laboratory test cells modified to include isolated reference electrodes. Inserting a third electrode creates a thermodynamic reference point against which the absolute working potential of the graphite anode is measured during high-rate charge and discharge cycles.
Reference electrodes validate mathematical predictions. Gold wire probes require baseline calibration.
Laboratory validation protocols combine three-electrode transient measurements with operando diagnostic techniques. Placing a micro-reference electrode inside the separator layer allows engineers to isolate anode overpotential directly, verifying observer accuracy under identical temperature, current, and state-of-charge profiles. Operando optical cell testing and differential voltage analysis provide supplementary evidence of metallic lithium formation, confirming the exact overpotential threshold where plating initiates.
- Reference Electrode Drift Calibration identifies shift in baseline half-cell potential before acquiring high-frequency charge transient data.
- Operando Microscopy Alignment matches optical reflectivity shifts on graphite surfaces directly with modeled threshold crossings.
- Differential Capacity Signature Cross Check confirms the onset of post-charge lithium stripping peaks against observer-predicted overpotential drops below zero volts.

Reference Electrode Architectures for Laboratory Validation
Three-electrode pouch cells incorporate metallic lithium wire insulated within porous sleeves. Inserting lithium foil micro-wires or gold-coated copper wires between separator sheets provides stable reference potential signals without disrupting internal ionic flux patterns. Gold wire reference electrodes alloyed with lithium form a stable biphasic gold-lithium reference phase (LiAu), exhibiting less chemical drift than pure metallic lithium in organic carbonate electrolytes over extended cycling.
Acquiring accurate transient data requires positioning the reference tip precisely within the central active area of the electrode stack. Edge placements introduce geometric current distribution distortion, artificially elevating measured overpotentials relative to central particle conditions.
How Does Operando Testing Verify Observer Potential Estimates?
Optical cells equipped with transparent quartz windows allow direct visual observation of metallic lithium phase changes. As local anode potential drops below zero volts vs Li/Li+, dark grey graphite transitions to golden stage-one intercalation (LiC6), followed rapidly by the appearance of bright silver metallic lithium deposits. Synchronizing high-speed optical video with real-time observer logging validates the exact overpotential margin where the observer predicts Eanode le 0 mV.
Supply agreements mandating fast-charging warranty coverage rely on observer accuracy verification against calibrated reference electrodes across the operational thermal envelope.
| Validation Method | Measured Parameter | Accuracy Boundary (mV) | Implementation Setting |
|---|---|---|---|
| Three-Electrode Li-Wire | Direct Intercalation Potential | ± 2.0 | Laboratory Pouch Cell |
| Gold-Wire Biphasic Probe | Stable Half-Cell Potential | ± 1.5 | Automated Test Bench |
| Operando Optical Microscopy | Surface Reflectance Phase Shift | ± 5.0 | Visual Diagnostic Cell |
| Post-Test ICP-OES Teardown | Harvested Plated Lithium Mass | ± 12.0 | Post-Mortem Destructive Laboratory |
Observer models validated against reference electrode cells under isothermal bench conditions lose fidelity when deployed into high-density pack assemblies with uneven thermal dissipation.

Protocol
Closed-loop fast charging algorithms dynamically adjust incoming current based on predicted anode surface overpotentials. Traditional charging algorithms follow rigid step-current or constant-current constant-voltage profiles designed conservatively to avoid worst-case plating conditions. These static methods extend charging times unnecessarily during low-risk operating windows, while failing to protect the cell when cold temperatures or parameter aging compress kinetic margins.
Anode-overpotential-controlled charging algorithms replace static voltage limits with real-time control loops that throttle current to maintain negative electrode potential precisely at a target safety margin above zero millivolts.
Adaptive current profiles preserve kinetic margins. Uncontrolled overpotential triggers cell failure. Algorithm execution time dictates controller choice.
Plating onset accelerates capacity loss. Implementing physics-based overpotential observers directly alters the landed economics of high-capacity storage systems by expanding usable fast-charge power windows without shortening total cycle life.

Adaptive Charging Dynamics for Plating Prevention
Traditional constant-current constant-voltage charging profiles rely on conservative fixed voltage limits. Physics-based observer control replaces fixed current steps with continuous closed-loop modulation. The controller monitors predicted anode potential Eanode(t) at every sampling interval.
If Eanode(t) approaches the control threshold Eset (typically set between +5 mV and +15 mV vs Li/Li+ to account for estimation uncertainty), a proportional-integral feedback loop reduces charge current I(t) according to:
I(t) = I(t-Δ t) + Kp (Eanode(t) – Eset) + Ki int (Eanode(τ) – Eset) dτ
This dynamic throttling maintains maximum permissible current density throughout the charging process, accelerating 10% to 80% state-of-charge duration while holding the graphite anode safely out of the lithium deposition regime.

Commercial Value and Amortized Storage Economics
Integrating real-time estimation algorithms into battery management microcontrollers extends active cycle life. Consider a 100 kWh battery pack operating in commercial transport duty, assuming an initial cell cost of 110 USD per kWh ($11,000 total pack cell cost) and a baseline lifespan of 1,200 cycles when subjected to standard 3C fast-charging protocols. At 80 percent average depth of discharge, this baseline configuration delivers 96,000 kWh of total throughput over its operational life, yielding an amortized cell cost of 0.1146 USD per delivered kWh.
Implementing a physics-based overpotential observer on an upgraded microcontroller (adding 4.00 USD per pack in component cost) allows dynamic current throttling that holds the graphite potential precisely at 5 mV versus Li/Li+ without dipping into plating territory. Under identical thermal and duty conditions, cycle life expands to 2,600 cycles before reaching the 80 percent capacity end-of-life limit. Total delivered energy increases to 208,000 kWh.
The amortized cell cost drops to 0.0529 USD per delivered kWh, saving 0.0617 USD for every kWh delivered across the lifetime of the energy storage system.
Failure to constrain anode potential estimation error below five millivolts results in localized metallic lithium accumulation, triggering rapid capacity degradation, warranty forfeiture, and premature battery pack replacement.




