Description of need
Plasma simulation software is clunky and computationally expensive. This makes it difficult to use this software, and difficult to simulate the plasma conditions of interest for the design of fusion reactors.
Problem severity (1-10)
Who has this need
- Nuclear fusion companies
- Academic research labs
Total addressable market (TAM)
Solutions today, and their shortcomings
Potentially relevant capabilities
Convolutional Neural Network partial different equation solvers
Notes
According to Sam Wurzel, some characteristics of opportunities in this space are:
- Risks: Few customers, slow sales cycles
- Bring best in class fusion specific features
- Get tools in the hands of students and fusion energy companies ASAP for free, to get lock-in.
- Charge companies for support services. Have application engineers.
The simulation software that the fusion community uses is ancient, and not at all optimized for parallel computing, GPUs, cloud computing, etc. Nobody will trust a new code, but there may be an opportunity to completely refactor existing codes to optimize their operation. Not clear whether there is a business case for doing this, though.
References
2023-06-13 Amelia Cavallaro 2023-06-14 ARPA-e Fusion Meeting