This study explores the challenge of material degradation from radiation exposure, a phenomenon that significantly impacts fields ranging from materials science to nuclear engineering and space exploration. As of today, the primary solution of conventional simulation techniques are computationally expensive, making them impractical for widespread use. The difficulty of usage combined with complex methods to run simulations for many of these techniques limit accessibility to the research community, making it an important obstacle in solving radiation damage.

To address these challenges, we propose two solutions: First, we analyze meaningful data from Geant4 (a leading simulation tool), including solid properties, positioning, and materials, along with firing many different particles through the virtual world and collecting data through python scripts. Second, using these methods of data extraction, we will be able to predict the trajectories of any particle in a similar world state by building a convolutional neural network (CNN) which infers trajectories based on our data, bypassing the need for re-running computationally expensive simulations and directly providing trajectory outcomes.

Early results from this approach show promising signs of particle trajectory extractions and excellent scope for predictions through CNNs, streamlining the research process and facilitating faster, more efficient exploration of materials under radiation exposure. Through this method, we hope to mitigate radiation damage by providing more efficient and accessible ways to run simulations.

Vinay Gupta: Undergraduate Computer Science Student, Purdue University. Shrienidhi Gopalakrishnan: Undergraduate Engineering Student, Purdue University Brian Lee: Post-Doc at Purdue University, Alejandro Strachan: Professor of Materials Engineering. Purdue University First-Time-Researcher (FTR) Program

Passive-Components, passive-components.eu/wp-content/uploads/2017/10/1.1.-Passive-new-space-scenario.pdf. Accessed 3 Apr. 2024.

Tanya Faltens (2023), "Short Oral Research Communications: Elevator Pitch and 3 Minute Research Presentation," https://nanohub.org/resources/37505.

Researchers should cite this work as follows:

• Vinay Gupta, Shrienidhi Gopalakrishnan, Brian Hyun-jong Lee, Alejandro Strachan (2024), "Accelerating Radiation Damage Simulation Through Machine Learning," https://nanohub.org/resources/39025.

  BibTex | EndNote

1. deep learning
2. Geant4
3. machine learning
4. Monte Carlo
5. nanoHUB tools
6. physics-based machine learning
7. Poster
8. radiation damage
9. radiation hardening
10. SCALE@Purdue
11. SCALE (Scalable Asymmetric Lifecycle Engagement)
12. Transformer
13. undergraduate research