A short trajectory is all you need: A transformer-based model for long-time dissipative quantum dynamics

Area of Science:

• Quantum mechanics
• Artificial intelligence
• Computational physics

Background:

• Simulating quantum systems coupled to environments is computationally challenging.
• Predicting long-time dynamics of quantum dissipative systems is crucial for understanding quantum phenomena.
• Existing methods have limitations in accuracy and efficiency across different coupling regimes.

Purpose of the Study:

• To develop an efficient and accurate deep learning model for predicting long-time population dynamics of quantum systems.
• To leverage transformer neural networks with self-attention for quantum dynamics prediction.
• To assess the model's performance across various system-bath coupling strengths and non-Markovian regimes.

Main Methods:

• A deep artificial neural network utilizing a transformer architecture with self-attention layers was developed.
• The model was trained using short-time population dynamics data of a quantum system.
• The spin-boson model was used as a benchmark system to test the model's predictive capabilities.

Main Results:

• The transformer neural network model accurately predicts the long-time population dynamics of the spin-boson model.
• The model demonstrates high accuracy across weak to strong coupling, including non-Markovian regimes.
• Performance is superior to classical forecasting models like recurrent neural networks.

Conclusions:

• Deep artificial neural networks, specifically transformer architectures, offer a powerful tool for predicting quantum system dynamics.
• This approach provides an efficient and accurate alternative to traditional simulation methods for quantum dissipative systems.
• The model's ability to generalize across different coupling regimes highlights its potential for broader applications in quantum science.