Members of the Multiphysics Materials Computation Research Group
Associate Professor, Department of Mechanical Engineering, Keio University (Third from left)
Researcher, Soft Material Simulation Team, National Institute of Advanced Industrial Science and Technology (AIST), (At the time of writing the paper, enrolled in the doctoral program at the Department of Mechanical Engineering, Keio University)(Second from right)
Prof. MuramatsuOur research lab focuses on understanding complex phenomena in metals, polymers, and ceramics, using solid mechanics as a foundation. We also work on developing and advancing new Computer Aided Engineering (CAE) techniques, and integrating computation with experimentation.
Recently, we released a paper*1 in collaboration with Mr. Matsuda from Fixstars Amplify, which utilizes Fixstars Amplify for simulating the equilibrium state of polymers.
*1 Endo, K., Matsuda, Y., Tanaka, S. et al. A phase-field model by an Ising machine and its application to the phase-separation structure of a diblock polymer. Sci Rep 12, 10794 (2022).
We converted the existing simulation method, the phase-field method, into a QUBO format and solved it using the Fixstars Amplify AE. As a result, we have gathered results that matched those obtained with traditional methods, but at a much faster speed (what used to take 100 minutes with conventional methods takes only 10 seconds with the Amplify AE). We believe that this method has the potential to be used in various fields, such as materials research in metals, as a new, efficient way to simulate physical phenomena.
Prof. Muramatsu In our field of research, we often deal with complex calculations that require a large amount of computing power. One of our main challenges has been to find ways to quickly perform these calculations. Using parallelized calculations. Nowadays parallelized calculations with supercomputers and GPUs are commonly used to faster computations.
While exploring solutions for large-scale computations, I heard Professor Yagawa, a well-known expert in numerical simulations, mention in a lecture that utilizing next-generation technologies like quantum computers will be crucial for future research in numerical simulations. My supervisor, Dr. Shizawa, also had a strong interest in quantum computing.
Inspired by the thoughts of senior professors, I read and learned about quantum computers. Then I started to ponder if we could apply these new technologies to our materials science field.
I shared my idea of "simulating physical phenomena using annealing" with Mr. Endo, who was a doctoral student in my lab at the time. He quickly made a prototype using a GPU machine and showed me the results. The outcome was so intriguing that I decided to seriously pursue this line of research.
Prof. Muramatsu I wanted to utilize an actual quantum annealing Ising machine in my research, so I reached out to Dr. Tanaka*2, Associate Professor at Keio University, who is also conducting research on quantum annealing Ising machines. During our consultation, Professor Tanaka introduced me to various companies and machines related to quantum computing, including Fixstars Amplify.
*2 Shu Tanaka, Department of Applied Physics and Physico-informatics Faculty of Science and Technology, Keio University
Dr. Endo I had previously worked on the IPA MITOU Program*3, where I became familiar with Fixstars Amplify as it was utilized in the project.
Prof. MuramatsuThere were three primary reasons.
Dr. Endo I was primarily responsible for implementing the program. I found that using Fixstars Amplify was simple to use, even with just a basic knowledge of quantum annealing and some programming experience. One of the features I found particularly helpful was the automatic conversion of polynomials to QUBO, it made the whole process much more user-friendly.
Based on my observation, Mr. Endo had a high level of programming skills and a good balance of creativity and ability to implement ideas, which made it easy for him to use Fixstars Amplify. Additionally, two students who recently began research using Amplify, while they had basic knowledge of Python, were able to effectively use the software with guidance and training from more experienced peers.
The tutorials and documentation available on the Amplify website were also very well-prepared and helpful.
Prof. Muramatsu I struggled with formulating the phase-field model in QUBO format. However, the calculation time was greatly reduced to around 10 seconds, which allowed me to quickly go through the implementation and verification cycle, and also enabled me to analyze various cases at once. I felt that this resulted in a significant achievement.
Dr. Endo As the Professor mentioned, I struggled with formulating the phase-field model into a QUBO format. To begin with, as far as I know, there were no precedents for using annealing for simulating physical phenomena, and it was an idea that started from nowhere. But when I was actually able to get the results that I imagined through the simulation, I was very happy. If this is possible, I feel that there is a great potential to use it in other types of physical simulations and to be able to simulate anything by writing the energy in a QUBO format.
Prof. Muramatsu In our research lab, we are not only utilizing Fixstars Amplify for simulating phase-field models, but also for analyzing the optimization of truss structures. We aim to solve larger problems faster and are currently using a paid plan. We are starting to see interesting results and are excited to see what the future holds.
In addition, we recently began collaborating with a group from RWTH Aachen University of Applied Sciences in Germany and a PhD student from Keio University in Japan on research related to fluid analysis. Although my expertise is in materials and solids, rather than fluids, I am using the insights gained from simulating with the Ising machine to explore the possibility of developing new analysis methods in the field of fluids.
I believe there is a lack of knowledge in the design of QUBO when it comes to mastering the use of Ising machines, including the representation of real numbers. In our research lab, we plan to continue conducting research utilizing the Ising machine in various fields and publishing papers to share our findings on how to design QUBOs to effectively utilize the Ising machine and the benefits that can be achieved in these fields.
Prof. Muramatsu In the field of material simulation, there is a significant interest not only in the final equilibrium state but also in the process leading to it (non-equilibrium state). Therefore, we are exploring the possibility of obtaining information that cannot be represented by linking the calculation process using the Ising machine (the trend of the decrease in the objective function) with the non-equilibrium state. We are also engaging in discussions with the people from Fixstars Amplify about this.
In addition, as it relates to the use of Fixstars Amplify and annealing in general, it is crucial to establish how to handle probabilistic results when incorporating probabilistic annealing into deterministic numerical simulations.
Prof. Muramatsu I frequently communicate with various researchers, and I sense that many of them are intrigued by the principle of the Ising machine, which is utilized to search for the ground state of energy without the concepts of time evolution or boundary conditions that are commonly present in general simulations.
Through this research, I realized that if we have the capability to think about how to design QUBO, there is a potential to apply Ising machines to a wide range of applications, including simulating physical phenomena. I hope to accumulate research results utilizing Ising machines in various fields.
* All information in this article is based on the information available at the time of the interview (November 2022).
I was greatly inspired by Prof. Muramatsu and everyone in her lab who developed a new method of using the Ising machine to simulate physical phenomena and consistently achieved results. As Prof. Muramatsu mentioned in the interview, I believe that the analytical method presented this time has the potential to be applied in various fields, such as materials research in metals. Our team will continue to work together to make the Fixstars Amplify useful for research.
Interviewer: Takahisa Todoroki (Senior Director, Partner Success Department, Fixstars Amplify Corporation)