[Bug]: Chen2020_composite cannot run a thermal model (per-phase electrode density) #5877

Activity

1. rtimms

   added

   bugSomething isn't workingSomething isn't working

   on Oct 8, 2026
2. rtimms commented on Oct 8, 2026

   

   rtimms

   on Oct 8, 2026

   ContributorAuthor

   More actions

   Workaround for the lumped model: use "use lumped thermal capacity": "true" and give a cell-level "Cell heat capacity [J.K-1.m-3]" (volumetric). The lumped model then never reads the electrode or phase densities, so Chen2020_composite builds. The parameter set doesn't include a cell heat capacity, so you need to add one:

   import pybamm
   
   parameter_values = pybamm.ParameterValues("Chen2020_composite")
   parameter_values.update(
       {"Cell heat capacity [J.K-1.m-3]": 2.5e6}, check_already_exists=False
   )
   model = pybamm.lithium_ion.SPM(
       {
           "particle phases": ("2", "1"),
           "thermal": "lumped",
           "use lumped thermal capacity": "true",
       }
   )
   pybamm.Simulation(model, parameter_values=parameter_values).build()  # builds

   (2.5e6 is a placeholder; use the cell's measured volumetric heat capacity.)

   So the bug remains for:

   • "thermal": "x-full" and the pouch-cell thermal models, which resolve temperature in space, need the per-domain densities, and reject "use lumped thermal capacity".
   • "thermal": "lumped" without the option (the default). It still builds the effective heat capacity from the per-domain densities and fails with the KeyError above.
3. Fstarnb commented on Oct 13, 2026

   

   Fstarnb

   on Oct 13, 2026

   More actions

   Could a maintainer clarify the intended contract for composite-electrode thermal properties in #5877? Should a composite parameter set provide effective electrode-level density, heat capacity, and conductivity, or should PyBaMM derive them from per-phase properties? If deriving them, which volume fractions and constituents should the mixture include—for example, active materials only, or also binder, conductive additive, and pore electrolyte? The issue’s active-phase weighting option may not define the whole-electrode thermal properties needed by the thermal model.
4. rtimms commented on Oct 20, 2026

   

   rtimms

   on Oct 20, 2026

   ContributorAuthor

   More actions

   Good question, @Fstarnb. The contract is that thermal properties are electrode-level effective properties. The thermal models treat each electrode as one homogenised layer, and "Negative electrode density [kg.m-3]", "... specific heat capacity [J.kg-1.K-1]" and "... thermal conductivity [W.m-1.K-1]" describe that whole layer, whatever it contains. That means active material(s), binder, conductive additive and the pore electrolyte, if the source measured or estimated it that way. PyBaMM doesn't derive them from per-phase values. A mixture rule over the active phases alone would leave out binder, carbon and electrolyte. A full rule would need volume fractions and properties PyBaMM doesn't model. So a composite parameter set should supply the electrode-level values, as it does for any other electrode, and anyone who wants a mixture rule can compute it and pass the result (optionally as a function of temperature).

   The per-phase "Primary: / Secondary: Negative electrode density" entries in Chen2020_composite and Bonkile2024 are a mistake: no model reads them. #5934 gives both sets an electrode-level "Negative electrode density [kg.m-3]" (1657, the graphite electrode value from Chen2020; the silicon is 1.5% by volume) and removes the per-phase ones, so these sets run with any thermal option.