Seeking Guidance on Implementing Coupled Density, Grain Growth, and Viscous Stress Kernels for Liquid Phase Sintering in MOOSE

[princevalliyil]

Dear MOOSE Team,

I am developing a custom MOOSE-based model to simulate liquid phase sintering by coupling the evolution of density (ϕ), grain size (G), and viscous stress. The goal is to capture microstructural densification through strain-rate-driven porosity evolution and temperature-dependent grain growth.

The model involves the following components:

ComputeSinterViscousStress – computes the viscous stress and provides the volumetric strain rate (ε̇ₖₖ) as an internal property.

DensityEvolutionSource – evolves the density field using a backward Euler scheme:
ϕn+1−ϕn−Δt(−ϕn+1ε˙kk​)=0
GrainGrowthSource – updates the grain size using an Arrhenius-type law:
Gn+1−Gn−Δtk0​G−2exp(−QG​/(RT))=0, where T is obtained from a user-defined temperature function.
ViscosityCoupled – defines temperature- and grain-size-dependent viscosity and bulk modulus as:
η=ηd​exp(Qd​/(RT))(G/G0​), K=0.6η
I have implemented these as custom Kernel and Material classes (DensityEvolutionSource, GrainGrowthSource, ComputeSinterViscousStress, and ViscosityCoupled) in my application lps1App. The code compiles successfully, but the simulation either diverges or crashes at runtime.

Before proceeding with detailed debugging, I would appreciate confirmation on whether this approach and implementation pattern aligns with MOOSE conventions.
I have attached the source files for all four classes for reference. Any feedback or best-practice suggestions for this type of coupled sintering model would be greatly appreciated.

Thank you for your time and support.

Best regards,
Prince Joseph
Senior Research Fellow, HBNI
Indira Gandhi Centre for Atomic Research
GrainGrowthSource.cpp
DensityEvolutionSource.cpp
ComputeSinterViscousStress.cpp
ViscosityCoupled.cpp
inputfile.pdf

[princevalliyil]

This is a follow-up to my earlier message regarding the implementation of custom kernels and materials for liquid phase sintering in MOOSE.

I would like to clarify that the objective of the current implementation is to develop a viscous constitutive model that captures the coupled evolution of density, grain growth during the liquid phase sintering process. The model is intended to study the deformation of the body after sintering

[GiudGiud]

Hello

I took a quick look. A few things I saw that do not match existing MOOSE techniques:

• the commitState() routine in the material properties. We don't need to store the current state into the old state manually. MOOSE does that for you.
• the time derivative term needs to be treated separately (in both kernels) in a TimeKernel, not in the same kernel. This is so that the time integrator system knows how to compute higher order time schemes, but it could have unexpected consequences not to follow this pattern.
• the off-diagonal Jacobian terms are not implemented. This could impede convergence