Single photon emitters are core building blocks of quantum technologies, with established and emerging applications ranging from quantum computing and communication to metrology and sensing. Regardless of their nature, quantum emitters universally display fluorescence intermittency or photoblinking: interaction with the environment can cause the emitters to undergo quantum jumps between on and off states that correlate with higher and lower photoemission events, respectively. Understanding and quantifying the mechanism and dynamics of photoblinking is important for both fundamental and practical reasons. However, the analysis of blinking time traces is often afflicted by data scarcity. Blinking emitters can photo-bleach and cease to fluoresce over time scales that are too short for their photodynamics to be captured by traditional statistical methods. Here, we demonstrate two approaches based on machine learning that directly address this problem. We present a multi-feature regression algorithm and a genetic algorithm that allow for the extraction of blinking on/off switching rates with ⩾85% accuracy, and with ⩾10× less data and ⩾20× higher precision than traditional methods based on statistical inference. Our algorithms effectively extend the range of surveyable blinking systems and trapping dynamics to those that would otherwise be considered too short-lived to be investigated. They are therefore a powerful tool to help gain a better understanding of the physical mechanism of photoblinking, with practical benefits for applications based on quantum emitters that rely on either mitigating or harnessing the phenomenon.
Read full abstract