Exploring physics beyond current understanding requires increasingly sensitive detectors, and a team led by Simone Manti, Jason Yip, and Massimiliano Bazzi is pushing the boundaries of low-energy detection with a new approach to data analysis. Researchers in the VIP Collaboration operate the Broad Energy germanium detector to search for evidence of new phenomena in the realm of quantum mechanics, and this work marks a significant upgrade in its capabilities. The team developed a machine learning workflow employing autoencoders and convolutional neural networks to dramatically improve detection of low-energy events and extend sensitivity to 10 keV. This advance has been validated on a large dataset of detector waveforms and is expected to not only lower the minimum detectable energy but also improve spectral quality, resulting in measurable improvements in the detector's ability to identify rare signals and improve fundamental tests in physics.
Identification of dark matter signals by pulse shape analysis
In this study, we investigate new methods to improve the sensitivity of dark matter searches by better distinguishing between genuine dark matter interactions and background noise. Current experiments primarily struggle with events that mimic dark matter signals from environmental radioactivity and neutrons. In this study, we develop and validate a method to determine the expected nuclear recoil from weakly interacting massive particles (WIMPs) by analyzing the shape of the scintillation light pulses they produce. This approach uses an organic scintillator coupled with a silicon photomultiplier tube to characterize the emitted light in detail by utilizing a cryogenic detector with high energy resolution and timing capabilities.
Detailed simulations incorporating realistic detector responses and background conditions demonstrate the potential to increase the sensitivity of WIMP searches by a factor of 10. The method is based on observing that nuclear recoil produces a scintillation signal that decays slowly compared to electron recoil, allowing effective separation through pulse shape discrimination. The main achievement is the development of a new algorithm for pulse shape identification based on machine learning, in particular utilizing boosted decision trees, which achieves greater than 99% efficiency in identifying nuclear recoil while minimizing false positives for electronic recoil. The detector benefits from substantial shielding, creating a low background environment essential for detecting rare events. The research team designed a new event selection strategy focused on improving the detection of low-energy signals down to 10 keV, a critical threshold for observing subtle quantum phenomena. This innovative approach uses a denoising autoencoder, a type of artificial neural network, to suppress both electronic noise and microphonic oscillations, effectively reconstructing the original pulse shape.
Following noise reduction, a convolutional neural network classifies the waveform and distinguishes between normal single-site events and events that indicate an anomaly. Validated on a dataset of over 20,000 waveforms, the method achieves high accuracy with a receiver operating characteristic curve area of 0.99 percent and event classification accuracy of 95 percent. Applying this procedure lowers the minimum detectable energy to approximately 10 keV, significantly increasing detector sensitivity and reducing the characteristic gamma-ray energy resolution while measurably increasing the signal-to-background ratio by 14%. These advances enhance the detector's ability to detect rare, low-energy signals and establish a scalable framework for precision testing of future quantum infrastructure.
VIP-2 Investigating Pauli Exclusion Violation
This study details the ongoing efforts of the VIP experiment, specifically the VIP-2 phase, to explore violations of the Pauli exclusion principle. The central goal is to test this basic principle: two identical fermions cannot occupy the same quantum state at the same time. A violation would have significant implications for our understanding of quantum mechanics. This detector is designed to identify unusual events in which electrons appear to occupy a state that is already occupied.
The methodology involves the application of advanced signal processing techniques and deep learning, specifically a denoising autoencoder (DAE), to improve the signal-to-noise ratio and enhance the identification of potential violation events. These autoencoders clean up waveforms to make fine signals more distinct and use machine learning-powered pulse shape identification technology to distinguish between different types of events and reduce background noise. This experiment continues to set tight limits on the probability of violating the Pauli exclusion principle, contributing to the broader field of quantum fundamentals and the search for new physics beyond the Standard Model. The researchers developed a workflow that utilizes a denoising autoencoder to suppress electronic and microphonic noise, effectively reconstructing pulse shapes, and using a convolutional neural network to classify waveforms into standard events or waveforms containing anomalies. These improvements lowered the minimum detectable energy to approximately 10 keV and measurably improved the signal-to-background ratio by 14%, while also reducing the characteristic gamma-ray energy resolution. This enhanced performance directly benefits the search for signs of rare events, especially violations of the Pauli exclusion principle and spontaneous wavefunction collapse. Both require very sensitive detection of low energy signals.
