Deep learning-based computational method for soft matter dynamics: Deep Onsager-Machlup method

Zhihao Li,Boyi Zou,Haiqin Wang,Jian Su,Dong Wang,Xinpeng Xu

2023-08-28

Abstract:A deep learning-based computational method has been proposed for soft matter dynamics -- the deep Onsager-Machlup method (DOMM), which combines the brute forces of deep neural networks (DNNs) with the fundamental physics principle -- Onsager-Machlup variational principle (OMVP). In the DOMM, the trial solution to the dynamics is constructed by DNNs that allow us to explore a rich and complex set of admissible functions. It outperforms the Ritz-type variational method where one has to impose carefully-chosen trial functions. This capability endows the DOMM with the potential to solve rather complex problems in soft matter dynamics that involve multiple physics with multiple slow variables, multiple scales, and multiple dissipative processes. Actually, the DOMM can be regarded as an extension of the deep Ritz method constructed using variational formulations of physics models to solve static problems in physics as discussed in our former work [Wang et al, Soft Matter, 2022, 18, 6015-6031]. In this work, as the first step, we focus on the validation of the DOMM as a useful computational method by using it to solve several typical soft matter dynamic problems: particle diffusion in dilute solutions, and two-phase dynamics with and without hydrodynamics. The predicted results agree very well with the analytical solution or numerical solution from traditional computational methods. These results show the accuracy and convergence of DOMM and justify it as an alternative computational method for solving soft matter dynamics.

Soft Condensed Matter

What problem does this paper attempt to address?

This paper attempts to address the computational methods for complex problems in soft - matter dynamics. Specifically, the author proposes a deep - learning - based computational method - the Deep Onsager - Machlup Method (DOMM) - to solve problems such as multiple physical processes, slow variables, multi - scale and dissipative processes in soft - matter dynamics.

### Main problems of the paper

1. **Limitations of traditional methods**:
- Traditional variational methods (such as Ritz - type methods) require careful selection of trial functions when dealing with soft - matter dynamics, which limits their scope of application and flexibility.
- For complex problems involving multiple physical processes, multiple slow variables, multi - scale and multiple dissipative processes, traditional methods are difficult to solve effectively.

2. **Advantages of DOMM**:
- **Flexibility**: DOMM uses deep neural networks (DNNs) to construct trial functions and can explore a richer and more complex set of acceptable functions.
- **Accuracy**: DOMM combines the Onsager - Machlup variational principle (OMVP) and can find the optimal solution without relying on specific trial functions.
- **Scalability**: DOMM can be regarded as an extension of the deep Ritz method (DRM) and is suitable for solving dynamic problems, not just static problems.

### Specific application scenarios

To verify the effectiveness of DOMM, the author applies it to several typical soft - matter dynamics problems:

1. **Diffusion of particles in dilute solutions**:
- DOMM is used to solve the diffusion problem of particles in one - dimensional space and is compared with the analytical solution. The results show that DOMM has high accuracy.
- By adjusting the initial and boundary conditions, the applicability and convergence of DOMM in different situations are verified.

2. **Two - phase dynamics**:
- The two - phase dynamics problems with and without hydrodynamic influence are considered, and the ability of DOMM to handle complex systems is verified.

### Conclusion

Through these applications, DOMM shows its potential as an alternative computational method, especially when solving complex problems in soft - matter dynamics. DOMM not only improves the computational accuracy and convergence but also provides a new tool for studying soft - matter dynamics.

### Summary of mathematical formulas

- **Onsager - Machlup integral**:
\[
L_{\text{om}}=\int_{t_0}^{t}dt'\int_{ - L}^{L}dx\frac{\zeta}{2n}(j - j^*)^2
\]
where \(j^* = - D\frac{\partial n}{\partial x}\) is the most probable flux and \(D=\frac{k_BT}{\zeta}\) is the diffusion coefficient.

- **Loss function**:
\[
L = w_{\text{om}}L_{\text{om}}+w_{\text{bc}}L_{\text{bc}}+w_{\text{ic}}L_{\text{ic}}+w_{\text{con}}L_{\text{con}}
\]
where \(L_{\text{om}}\) is the Onsager - Machlup integral, and \(L_{\text{bc}}\), \(L_{\text{ic}}\) and \(L_{\text{con}}\) represent the additional loss function terms for boundary conditions, initial conditions and physical or geometric constraints respectively.

Through these formulas and application examples, DOMM shows its broad application prospects in the field of soft - matter dynamics.