Preconditioning for a Cahn-Hilliard-Navier-Stokes model for morphology formation in organic solar cells

Pelin Çiloglu, Carmen Tretmans, Roland Herzog, Jan-Frederik Pietschmann, Martin Stoll

Introduction

Organic photovoltaics (OPVs) represent a promising frontier in renewable energy, offering lightweight, flexible, and cost-effective alternatives to traditional silicon-based solar cells. The efficiency of these devices is fundamentally linked to their microscopic structure, or morphology, which forms during the manufacturing process. Typically, OPVs are produced by depositing a liquid blend of a polymer (electron donor), a non-fullerene acceptor (NFA), and a solvent onto a substrate. As the solvent evaporates, the materials undergo phase separation, creating an intricate interwoven network that determines the cell's ability to convert light into electricity.

Predicting and optimizing this morphology through experimental trial and error is prohibitively expensive and time-consuming. Consequently, high-fidelity computer simulations are essential. However, modeling the simultaneous effects of fluid flow, phase separation, and evaporation presents a massive computational challenge. This paper introduces a robust numerical framework and a highly efficient preconditioned iterative solver designed to handle the complex mathematical systems generated by these simulations.

The Physical Model

The researchers developed a four-component phase-field model to track the volume fractions of the polymer, NFA, solvent, and surrounding air. This multiphysics approach couples several fundamental equations:

  • Navier-Stokes Equations: These model the macroscopic fluid dynamics and movement of the blend.
  • Cahn-Hilliard Equations: These describe the thermodynamic process of phase separation as the polymer and NFA spontaneously organize into distinct regions.
  • Allen-Cahn Equation: This is utilized to model the evaporation of the solvent into the air at the top boundary of the domain.

To ensure numerical stability during complex 2D and 3D simulations, the authors addressed a common mathematical hurdle: the logarithmic free-energy function typically used in Flory-Huggins theory. While physically accurate, this function often causes numerical crashes. The authors replaced it with a fitted polynomial "double-well" potential, which maintains the physical behavior of phase separation while significantly improving the stability of the solver.

Numerical Innovation: Block-Diagonal Preconditioning

When these continuous physical equations are discretized using the Finite Element Method (FEM), they result in massive, poorly conditioned linear systems—specifically saddle-point problems. Standard algorithms often fail to solve these systems efficiently as the resolution of the simulation increases.

The core innovation of this work is the design of a custom block-diagonal preconditioner used in conjunction with the Generalized Minimal Residual (GMRES) method. A preconditioner acts as a mathematical transformation that simplifies a complex matrix into a form that is easier to solve without changing the final solution.

The authors’ approach involves:

  1. Schur Complement Approximation: They approximate the most difficult parts of the system (the coupling between fluid pressure and velocity) using a matching approach.
  2. Algebraic Multigrid (AMG): The sub-blocks of the preconditioner are solved rapidly using AMG methods via the PyAMG library.

This strategy allows the solver to decouple the mathematical difficulty of the system from the grid resolution, ensuring that the computational cost remains manageable even for high-resolution 3D models.

Key Results and Performance

The proposed solver was tested across 1D, 2D, and 3D domains, demonstrating exceptional performance and reliability:

  • Linear Scaling: The computational runtime scaled linearly with the number of degrees of freedom. This "gold standard" for iterative solvers means that doubling the detail of a simulation only roughly doubles the time required, rather than increasing it exponentially.
  • Algorithmic Robustness: The preconditioned GMRES solver maintained a stable and low number of iterations (typically 15 to 35) per time step. Crucially, this performance remained consistent regardless of changes to the time-step size (ranging from $10^{-3}$ to $10^{-7}$) or the fineness of the spatial grid.
  • Morphology Reproduction: The model successfully captured the expected physical phenomena. Simulations showed the solvent layer shrinking due to evaporation while the polymer and NFA formed the characteristic interwoven domains necessary for efficient OPV active layers.

Conclusion and Future Work

This research provides a powerful tool for the virtual prototyping of organic solar cells. By proving that preconditioning is essential for scaling multiphysics simulations, the authors have paved the way for more detailed 3D studies of manufacturing processes.

While the current model uses simplifying assumptions—such as constant density and viscosity—future work will focus on introducing more physically specific parameters. As real-world materials involve viscosities and diffusion rates spanning several orders of magnitude, the next generation of this model will likely incorporate advanced multiscale numerical methods to maintain the stability and efficiency demonstrated here.

Summary updated

Abstract

We present a model for the morphology evolution of printed organic solar cells which occurs during the drying of a mixture of polymer, the non-fullerene acceptor and the solvent. Our model uses a phase field approach coupled to a Navier-Stokes equation describing the macroscopic movement of the fluid. Additionally, we incorporate the evaporation process of the solvent using an Allen-Cahn equation. The model is discretized using a finite-element approach with a semi-implicit discretization in time. The resulting (non)linear systems are coupled and of large dimensionality. We present a preconditioned iterative scheme to solve them robustly with respect to changes in the discretization parameters. We illustrate that the preconditioned solver shows parameter-robust iteration numbers and that the model qualitatively captures the behavior of the film morphology during drying.
BibTeX
@article{ciloglu2025preconditioning,
  author = {Pelin Çiloglu and Carmen Tretmans and Roland Herzog and Jan-Frederik Pietschmann and Martin Stoll},
  title = {{Preconditioning for a Cahn-Hilliard-Navier-Stokes model for morphology formation in organic solar cells}},
  journal = {Journal of Computational Physics},
  year = {2025},
}
Figure