SEI="none" with "partially reversible" lithium plating raises ZeroDivisionError during parameter processing

[dikshant1103-beep]

Description

I believe this is a validation issue rather than a solver issue.

When using the OKane2022 parameter set, the option combination

• SEI: "none"
• lithium plating: "partially reversible"

is accepted during model construction but fails during parameter processing with a bare ZeroDivisionError before any simulation begins.

The OKane2022 dead-lithium decay model is defined as SEI-limited and divides by the SEI thickness, which is identically zero when the SEI submodel is disabled.

The resulting exception contains no message and the traceback terminates inside expression_tree/binary_operators.py, making it difficult to determine that the failure is caused by an incompatible option combination rather than by the solver or the user's model.

Steps to reproduce

1. Install PyBaMM 25.12.2.
2. Run the script below.
3. Observe that parameter processing raises ZeroDivisionError before any simulation begins.

Minimal reproduction

import pybamm

model = pybamm.lithium_ion.SPMe(
    options={"lithium plating": "partially reversible", "SEI": "none"}
)
sim = pybamm.Simulation(model, parameter_values=pybamm.ParameterValues("OKane2022"))
sim.solve([0, 300], initial_soc=0.5)   # ZeroDivisionError (no message)

Root cause

pybamm/input/parameters/lithium_ion/OKane2022.py, in
SEI_limited_dead_lithium_OKane2022():

gamma_0 = pybamm.Parameter("Dead lithium decay constant [s-1]")
L_sei_0 = pybamm.Parameter("Initial SEI thickness [m]")

gamma = gamma_0 * L_sei_0 / L_sei      # L_sei == 0 when SEI option is "none"

L_sei is the SEI thickness variable, which is identically zero when the SEI
submodel is disabled, so pybamm.divide raises during
ParameterValues.process_model.

Scope

Tested with ParameterValues("OKane2022"), SEI: "none":

lithium plating option	result
"none"	ok
"reversible"	ok
"partially reversible"	ZeroDivisionError
"irreversible"	ok

• Reproduces identically with SPMe and DFN — it occurs at parameter
  processing, so it is model independent.
• OKane2022 is the only shipped parameter set containing lithium-plating
  parameters, so this is reachable by any user who turns SEI off while keeping
  partially-reversible plating.

Expected behaviour

The option combination should be rejected during model construction or parameter processing with a clear pybamm.OptionError explaining that "partially reversible" lithium plating requires an active SEI submodel when using the OKane2022 parameterization.

The current behaviour is difficult to diagnose because the exception is raised deep inside parameter processing rather than during option validation.

Actual behaviour

ZeroDivisionError

with an empty message, raised from expression_tree/binary_operators.py via
parameters/parameter_substitutor.py, with no reference to the option
combination that caused it.

Environment

• PyBaMM 25.12.2 (pip, site-packages)
• Python 3.13.11
• Linux 6.8.0-136-generic, x86_64, glibc 2.35
• numpy 2.0.2, scipy 1.17.1, casadi 3.6.7

[karkikamal098]

Confirmed on main (26.7.1.1.dev20+g6e5f8be9a), so this isn't specific to 25.12.2. I also reproduced your option matrix exactly none, reversible and irreversible all process fine with SEI: "none"; only partially reversible fails.

One correction to the scope, though this is broader than OKane2022. Four shipped parameter sets assign SEI_limited_dead_lithium_OKane2022 to "Dead lithium decay rate [s-1]", and all four raise ZeroDivisionError with SEI: "none" and partially reversible plating:

• OKane2022
• Ecker2015
• OKane2022_graphite_SiOx_halfcell (with working electrode: "positive")
• Ecker2015_graphite_halfcell (with working electrode: "positive")

I'd like to take this. I have a working patch locally implementing the OptionError you describe, raised from the incompatibility checks in BatteryModelOptions.__init__ alongside the existing option-conflict checks, plus regression tests in test_base_battery_model.py. The unit suites for the model and parameter modules still pass.

One design question before I open the PR. Given it affects four parameter sets, I've written the check at the option level it rejects partially reversible + SEI: "none" regardless of parameter set, on the grounds that partially reversible plating is defined in terms of an SEI thickness at all. The alternative is to keep it parameter-set-specific and only fail for sets that use the SEI-limited decay function, leaving the combination available to anyone supplying their own "Dead lithium decay rate [s-1]". Do you have a preference? The option-level check is simpler and catches the failure earlier, but it does close off a combination a user could in principle parameterise themselves.

Implementation note: for a full cell the plating option gets rewritten to a tuple (('partially reversible', 'none')) while SEI stays the scalar "none", since the tuple conversion skips "none" so the check needs to handle both forms or it won't fire on full cells.

[DrSOKane]

Hi Kamal, thank you so much for offering to take this.

The BatteryModelOptions class sets {"SEI": "constant"} as the default option if {"lithium plating": "partially reversible"} is passed to the options dictionary and no "SEI" option is specified, but manually specifying {"SEI": "none", "lithium plating": "partially reversible"} will indeed trigger a ZeroDivisionError.

An option-level check would be best. If the user wishes to run a model with partially reversible lithium plating but no SEI, they can still do so by setting {"SEI": "constant", "SEI film resistance": "none", "lithium plating": "partially reversible"}, which will create an SEI that has a finite thickness but no resistance, and therefore no effect on the voltage. (Source: I have done it before.)

Thanks again, Simon