Study of background from accidental coincidence signalsin the PandaX-II experiment*

Abdusalam Abdukerim, Wei Chen, Xun Chen, Yunhua Chen, C. H. Cheng, Xiangyi Cui, Yingjie 英杰 Fan 樊, Deqing Fang, Changbo Fu, Mengting Fu, Li‐Sheng Geng, Karl Giboni, Linhui Gu, Xuyuan Guo, K. Han, Changda He, Di Huang, Y. Huang, Zhou Huang, Xiangdong Ji, Yonglin Ju, Shuaijie Li, Huaxuan Liu, Jianglai Liu, Wenbo Ma, Y. G., Yajun Mao, Yue Meng, Kaixiang Ni, Jinhua Ning, Xuyang Ning, Xiangxiang Ren, Changsong Shang, Lin Si, Guofang Shen, Andi Tan, Anqing Wang, Hongwei Wang, Meng Wang, Qiuhong Wang, Siguang Wang, Wei Wang, Xiuli Wang, Zhou Wang, Mengmeng Wu, Shiyong Wu, Weihao Wu, Jingkai Xia, M. Xiao, Pengwei Xie, Binbin Yan, Jijun Yang, Yong Yang, Chunxu Yu, Jumin Yuan, Ying Yuan, Xinning Zeng, Dan Zhang, Tao Zhang, Li Zhao, Qibin Zheng, Jifang Zhou, N. Zhou, Xiaopeng ZhouChinese 2022 · DOI

TL;DR

This paper tackles a crucial background in dark matter detection: accidental coincidences of unrelated signals that mimic true dark matter interactions. The researchers meticulously studied these "accidental coincidences" in the PandaX-II experiment, identified their origins, and developed a machine learning approach to significantly reduce their impact, thereby improving the experiment's sensitivity.

What Problem Does This Paper Address?

The search for dark matter particles, particularly Weakly Interacting Massive Particles (WIMPs), relies on detecting rare interactions with ordinary matter. Experiments like PandaX-II use large liquid xenon detectors to achieve this. A significant challenge is distinguishing genuine WIMP-induced signals from background events. One such background arises from "accidental coincidences," where two independent, unrelated signals (a prompt scintillation, S1, and a delayed proportional scintillation, S2) occur within the detector's sensitive time window, mimicking the signature of a single WIMP interaction. This paper aims to understand the sources of these accidental coincidences, quantify their level, and develop effective methods to suppress them, which is critical for improving the sensitivity of dark matter searches.

Methods

The core of this paper involves a detailed study and suppression of accidental coincidence backgrounds. The authors first identified the potential physical and non-physical origins of isolated S1 and S2 signals, which are the building blocks of these accidental coincidences. They then developed several methods to accurately tag and estimate the rates of these isolated signals across different experimental runs.

The key technique for suppressing the accidental background is a Boosted Decision Tree (BDT) algorithm. This is a supervised machine learning method that combines multiple decision trees to make a classification. In this context, the BDT was trained to distinguish between simulated accidental background events and genuine nuclear recoil (NR) events (which are representative of WIMP signals). A comprehensive set of variables describing the properties of the S1 and S2 signals (e.g., charge, width, asymmetry) were used as input features for the BDT. The BDT was then applied to the experimental data to classify events and reject the accidental coincidences.

Datasets

The study primarily utilizes data from the PandaX-II experiment. This includes:

  • Dark matter search data: This is the primary dataset used for identifying and classifying events.
  • Calibration data: This includes data from AmBe (neutron source for nuclear recoils) and Tritium/Radon (for electron recoils) to understand the detector's response to different types of interactions.
  • Simulated accidental background events: These were generated by randomly pairing isolated S1 and S2 signals observed in the data, mimicking the accidental coincidence background.

No new benchmark datasets were introduced; the focus was on analyzing and processing existing experimental and simulated data.

Key Insights and Ideas

  • Accidental coincidences are a significant background: The paper clearly establishes that these events, formed by unrelated S1 and S2 signals, pose a real challenge to dark matter detection.
  • Multiple origins for isolated signals: Both physical processes (e.g., tiny sparks, scattering events in specific detector regions) and non-physical sources (e.g., PMT dark noise) contribute to the isolated S1 and S2 signals that form accidental coincidences.
  • PMT dark noise is a key contributor to isolated S1s: The higher rate of isolated S1s in Run 9, where PMTs operated at higher gains, strongly suggests that dark noise from PMTs plays a significant role. This highlights the importance of reducing dark noise for future experiments.
  • BDT is highly effective for suppression: The BDT algorithm demonstrates a remarkable ability to distinguish accidental coincidences from genuine WIMP-like (NR) signals, achieving substantial background reduction.
  • Specific signal properties are discriminative: Variables like S2 width (wS2), S2 asymmetry (S2SY2), and S1 asymmetry (S1TBA) were found to be particularly important for the BDT in identifying accidental events.

Key Results

  • The accidental coincidence background was reduced by 70% in the dark matter signal region using the BDT algorithm.
  • The BDT achieved a signal efficiency of about 90% for keeping genuine NR events below the NR median.
  • The estimated number of accidental background events below the NR median after BDT cuts was significantly reduced, for example, to 2.09 ± 0.25 for Run 9.
  • The study successfully suppressed the accidental background to a "trivial level," significantly improving the overall sensitivity of the PandaX-II experiment for dark matter searches.

Limitations and Future Work

  • Correlation in isolated S1s: Method 2 for tagging isolated S1s showed a decrease in events with increasing time difference, suggesting possible physical correlations (e.g., from 214Bi-214Po cascade decay). While this was accounted for, it indicates that perfect isolation isn't always guaranteed and requires careful correction.
  • Unknown origin of excess S1s in Run 11: An excess of isolated S1s was observed in Run 11 for short time differences, the origin of which remains unknown.
  • Future Work: The authors suggest that the developed analysis framework and suppression methods are directly applicable to future, larger experiments like PandaX-4T. They also anticipate that more advanced machine learning techniques, such as neural networks, could achieve even greater success in this area. The importance of reducing PMT dark noise is also emphasized for next-generation experiments.

Summary updated

Abstract

The PandaX-II experiment employed a 580kg liquid xenon detector to search for the interactions between dark matter particles and the target xenon atoms. The accidental coincidences of isolated signals result in a dangerous background which mimic the signature of the dark matter. We performed a detailed study on the accidental coincidence background in PandaX-II, including the possible origin of the isolated signals, the background level and corresponding background suppression method. With a boosted-decision-tree algorithm, the accidental coincidence background is reduced by 70% in the dark matter signal region, thus the sensitivity of dark matter search at PandaX-II is improved.
BibTeX
@inproceedings{abdukerim2022study,
  author = {Abdusalam Abdukerim and Wei Chen and Xun Chen and Yunhua Chen and C. H. Cheng and Xiangyi Cui and Yingjie 英杰 Fan 樊 and Deqing Fang and Changbo Fu and Mengting Fu and Li‐Sheng Geng and Karl Giboni and Linhui Gu and Xuyuan Guo and K. Han and Changda He and Di Huang and Y. Huang and Yanlin Huang and Zhou Huang and Xiangdong Ji and Yonglin Ju and Shuaijie Li and Huaxuan Liu and Jianglai Liu and Wenbo Ma and Y. G. and Yajun Mao and Yue Meng and Kaixiang Ni and Jinhua Ning and Xuyang Ning and Xiangxiang Ren and Changsong Shang and Lin Si and Guofang Shen and Andi Tan and Anqing Wang and Hongwei Wang and Meng Wang and Qiuhong Wang and Siguang Wang and Wei Wang and Xiuli Wang and Zhou Wang and Mengmeng Wu and Shiyong Wu and Weihao Wu and Jingkai Xia and M. Xiao and Pengwei Xie and Binbin Yan and Jijun Yang and Yong Yang and Chunxu Yu and Jumin Yuan and Ying Yuan and Xinning Zeng and Dan Zhang and Tao Zhang and Li Zhao and Qibin Zheng and Jifang Zhou and N. Zhou and Xiaopeng Zhou},
  title = {{Study of background from accidental coincidence signalsin the PandaX-II experiment*}},
  booktitle = {Chinese},
  year = {2022},
  doi = {10.1088/1674-1137/ac7cd8},
}
Figure