Time-Resolved Line Shapes of Single Quantum Emitters via Machine Learned Photon Correlations.

Area of Science:

• Quantum Optics
• Materials Science
• Machine Learning

Background:

• Solid-state single-photon emitters (SPEs) are crucial quantum light sources but suffer from spectral diffusion, limiting their characterization.
• Photon correlation Fourier spectroscopy (PCFS) is a technique for time-resolved line shape analysis, but is challenged by low signal-to-noise ratios at early times.

Purpose of the Study:

• To develop a computational framework for simulating PCFS correlation functions from diffusing spectra.
• To train a deep ensemble autoencoder machine learning model for accurate, noise-free reconstruction of PCFS data.
• To enable the study of single emitter line shapes on previously inaccessible timescales.

Main Methods:

• Simulated PCFS correlation functions using diffusing spectra models.
• Trained a deep ensemble autoencoder model on simulated data.
• Reconstructed time-resolved emission line shapes from noisy experimental PCFS data.

Main Results:

• The simulation framework accurately reproduces experimental PCFS data for colloidal quantum dots.
• The machine learning model successfully reconstructs noiseless correlation functions from noisy data.
• Time-resolved emission line shapes were obtained down to 10 ns, revealing spectral information on ultrafast timescales.

Conclusions:

• The developed machine learning approach enhances PCFS capabilities for characterizing SPEs.
• This method allows for optical coherence time extraction on timescales comparable to Hong-Ou-Mandel interference, with added spectral information.
• The approach is broadly applicable to various photon correlation spectroscopy techniques and SPE systems.