Hybrid Monte Carlo estimators for multilayer transport problems.

Area of Science:

• Computational physics
• Numerical methods

Background:

• Monte Carlo methods are crucial for particle transport simulations.
• Existing transport-constrained unbiased radiance estimators (T-CURE) offer significant variance reduction but increase computational overhead.
• Efficiency, measured by figure of merit (FOM), balances variance and run time.

Purpose of the Study:

• To develop hybrid estimators that enhance computational efficiency for particle transport problems.
• To mitigate the increased overhead associated with T-CURE estimators.
• To maintain unbiased computations while improving the figure of merit.

Main Methods:

• A hybrid estimation strategy combining T-CURE and conventional terminal estimators.
• Treating refractive interface crossings with an extended next event estimator.
• Utilizing standard terminal estimators for other interactions.
• Comparing the hybrid strategy against conventional methods in multi-layer scenarios.

Main Results:

• The hybrid estimator effectively moderates computational efficiency.
• The strategy avoids introducing bias into the Monte Carlo computations.
• Significant efficiency gains were demonstrated in numerical examples with refractive interfaces.
• The hybrid approach circumvents the intractability of pure T-CURE for index-mismatched interfaces.

Conclusions:

• Hybrid estimators offer a practical solution for optimizing Monte Carlo efficiency in particle transport.
• This approach balances the variance reduction of T-CURE with the computational feasibility of terminal estimators.
• The proposed method is particularly beneficial for problems involving refractive interfaces.