Simulation of ionizing radiation in cell phone camera image sensors

Runze Li,Alex Pizzuto,Justin Vandenbroucke,Brent Mode

2023-11-29

Abstract:The Distributed Electronic Cosmic-ray Observatory (DECO) is a cell phone app that uses a cell phone camera image sensor to detect cosmic-ray particles and particles from radioactive decay. Images recorded by DECO are classified by a convolutional neural network (CNN) according to their morphology. In this project, we develop a GEANT4-derived simulation of particle interactions inside the CMOS sensor using the Allpix$^2$ modular framework. We simulate muons, electrons, and photons with energy range 10 keV to 100 GeV, and their deposited energy agrees well with expectations. Simulated events are recorded and processed in a similar way as data images taken by DECO, and the result shows both similar image morphology with data events and good quantitative data-Monte Carlo agreement.

Instrumentation and Methods for Astrophysics,High Energy Astrophysical Phenomena,High Energy Physics - Experiment

What problem does this paper attempt to address?

### Problems Addressed by the Paper

The paper primarily explores how to detect cosmic ray particles and particles produced by radioactive decay using the image sensor of a mobile phone camera, and how to improve particle recognition through simulation techniques.

#### Core Issues
1. **Data Limitation**: The currently used Convolutional Neural Networks (CNN) rely on manually labeled data, which limits the amount of data available for training.
2. **Classification Limitation**: Existing CNNs can only classify events based on morphology and cannot determine the type or energy of particles.

#### Solutions
To overcome these issues, researchers have developed a simulation framework based on GEANT4 called Allpix2, which is used to simulate the interactions of particles in CMOS sensors. Through this method, simulated events containing real particle types and energies can be generated, thereby improving the CNN's ability to classify particle types.

#### Main Objectives
- **Simulation Technology**: Develop a detailed simulation framework to simulate the behavior of muons, electrons, and photons in CMOS sensors across different energy ranges.
- **Image Processing**: The simulated events undergo the same image processing workflow as actual data and are compared with real data.
- **Classification Performance**: Use simulated events to train the CNN to improve its ability to classify particle types and energies.

Through these methods, researchers hope to enhance the accuracy and efficiency of particle recognition without relying on a large amount of manually labeled data.