Exploring temperature-dependent photoluminescence dynamics of colloidal CdSe nanoplatelets using machine learning...

Area of Science:

• Materials Science
• Quantum Dots
• Spectroscopy

Background:

• Understanding temperature-dependent photoluminescence (PL) is crucial for colloidal semiconductor nanoplatelets (NPLs).
• Predicting PL spectra across wide temperature ranges is challenging using traditional methods.

Purpose of the Study:

• To explore machine learning (ML) techniques for predicting temperature-dependent PL spectra in colloidal CdSe NPLs.
• To forecast PL spectra from 85 K to 270 K backward to 0 K and forward to 300 K.
• To compare ML predictions with theoretical models like Fan and Varshni equations.

Main Methods:

• Utilized polynomial regression models, including Tweedie, LassoLars, Linear Regression, Ridge, and Theil-Sen regressors.
• Trained models on experimental PL data from 85 K to 270 K.
• Employed a genetic algorithm (GA)-based approach for fitting experimental data to theoretical equations.

Main Results:

• Optimal 6th-degree polynomial models with Tweedie regression predicted band energy up to 300 K.
• 9th-degree models with LassoLars and Linear Regression were suitable for backward predictions to 0 K.
• Specific Lasso, Ridge, Tweedie, and Theil-Sen models showed effectiveness for exciton energy predictions across the temperature range.

Conclusions:

• Machine learning effectively predicts temperature-dependent photoluminescence spectra in CdSe NPLs.
• Developed models provide accurate spectral forecasts from 0 K to 300 K.
• ML predictions offer a valuable complement to theoretical models for materials characterization.