| 研究課題/領域番号 |
20K03786
|
|---|
| 研究機関 | 京都大学 |
|---|
研究代表者 |
MOLINA JOHN 京都大学, 工学研究科, 助教 (20727581)
|
|---|
| 研究期間 (年度) |
2020-04-01 – 2023-03-31
|
| キーワード | Active Particles / Janus Particles / Electrophoresis / Quincke Rollers / Machine Learning / Multi-Scale Simulations / Optimal Control |
|---|
| 研究実績の概要 |
Working with other members of our lab, we have studied the collective motion of an active colloidal particle that is closely related to the ICEP Janus Particle: the Quincke rollers. Using a simple model for the propulsion, which assumed a fixed torque, but fully accounting for the hydrodynamics and the induced dipole-dipole interactions, we could better understand the rich behavior observed experimentally, and the crucial role played by the near-field hydrodynamic interactions.
We have also investigated the use of Machine Learning as (1) a probabilistic flow solver, (2) a way to learn constitutive relations, and (3) to tackle inverse optimal control problems. The first can help to interpret experiments on active particles, the second to develop continuum models, and the third to design optimal control strategies. For topic (1) we have developed a Bayesian Stokes flow solver that allows us to infer the velocity/pressure fields given only knowledge of the boundary values (in collaboration with an undergraduate student). Topic (2) was originally developed to accelerate Multi-scale simulations of polymer melts, and has now been extended to interacting/entangled polymer models (in collaboration with a master student). Topic (3) comes from a collaboration to understand the optimal government intervention strategy in an epidemic. For this, we have developed a physics informed neural network (Nash Neural Network) to infer the hidden utility of rational individuals from their behavior. We expect this will have broad applications in science, engineering, and biology.
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
3: やや遅れている
理由
Our study on the ICEP Janus particles, and the detailed mechanism behind the velocity reversal is still ongoing, as we have not yet obtained good quantitative agreement with experiments. Even though we can reproduce the velocity reversal, the dependence on the salt concentration is not consistent with experiments. To address these issues we need to further validate and improve our model.
However, we note that we have opened up new promising avenues of research, leveraging machine learning techniques, which can help in our goal of achieving designer active matter.
|
| 今後の研究の推進方策 |
We will continue to develop our Janus model, in order to achieve quantitative agreement with experiments. This includes finding a good enough model for the frequency dependent material properties. For this, we will leverage more detailed microscopic models that have been developed to describe these ICEP Janus particles. We will also finalize the Janus functionality into the latest Kapsel code, which now includes support for phase-separating fluids.
In addition, we will pursue the additional research avenues discussed above, in order to leverage the power of machine learning techniques to tackle such complex soft matter problems. This includes developing novel flow solvers (to help analyze experimental results), solve optimal control problems, and infer constitutive relations from microscopic/particle based data, in order to develop accurate coarse-grained models of active matter systems.
|
| 次年度使用額が生じた理由 |
The ongoing COVID-19 pandemic has continued to restrict our ability to travel, with most conferences we attended this year being held online. We thought that domestic travel would have returned to normal this past year, but this was also not the case. Going forward, we plan to use these funds to cover super-computing services, conference participation fees, publication fees, and if the situation allows, domestic and international travel costs.
|
