An official website of the United States government
ABSTRACT Forward-modeling observables from galaxy simulations enables direct comparisons between theory and observations. To generate synthetic spectral energy distributions (SEDs) that include dust absorption, re-emission, and scattering, Monte Carlo radiative transfer is often used in post-processing on a galaxy-by-galaxy basis. However, this is computationally expensive, especially if one wants to make predictions for suites of many cosmological simulations. To alleviate this computational burden, we have developed a radiative transfer emulator using an artificial neural network (ANN), ANNgelina, that can reliably predict SEDs of simulated galaxies using a small number of integrated properties of the simulated galaxies: star formation rate, stellar and dust masses, and mass-weighted metallicities of all star particles and of only star particles with age <10 Myr. Here, we present the methodology and quantify the accuracy of the predictions. We train the ANN on SEDs computed for galaxies from the IllustrisTNG project’s TNG50 cosmological magnetohydrodynamical simulation. ANNgelina is able to predict the SEDs of TNG50 galaxies in the ultraviolet (UV) to millimetre regime with a typical median absolute error of ∼7 per cent. The prediction error is the greatest in the UV, possibly due to the viewing-angle dependence being greatest in this wavelength regime. Our results demonstrate that our ANN-based emulator is a promising computationally inexpensive alternative for forward-modeling galaxy SEDs from cosmological simulations.
more »
« less
Accelerated Prediction of Photon Transport in Nanoparticle Media Using Machine Learning Trained With Monte Carlo Simulations
Abstract Monte Carlo simulations for photon transport are commonly used to predict the spectral response, including reflectance, absorptance, and transmittance in nanoparticle laden media, while the computational cost could be high. In this study, we demonstrate a general purpose fully connected neural network approach, trained with Monte Carlo simulations, to accurately predict the spectral response while dramatically accelerating the computational speed. Monte Carlo simulations are first used to generate a training set with a wide range of optical properties covering dielectrics, semiconductors, and metals. Each input is normalized, with the scattering and absorption coefficients normalized on a logarithmic scale to accelerate the training process and reduce error. A deep neural network with ReLU activation is trained on this dataset with the optical properties and medium thickness as the inputs, and diffuse reflectance, absorptance, and transmittance as the outputs. The neural network is validated on a validation set with randomized optical properties, as well as nanoparticle medium examples including barium sulfate, aluminum, and silicon. The error in the spectral response predictions is within 1% which is sufficient for many applications, while the speedup is 1–3 orders of magnitude. This machine learning accelerated approach can allow for high throughput screening, optimization, or real-time monitoring of nanoparticle media's spectral response.
more »
« less
- Award ID(s):
- 2102645
- PAR ID:
- 10553718
- Publisher / Repository:
- ASME
- Date Published:
- Journal Name:
- ASME Journal of Heat and Mass Transfer
- Volume:
- 145
- Issue:
- 5
- ISSN:
- 2832-8450
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Solar cells containing complex geometric structures such as texturing, photonic crystals, and plasmonics are becoming increasingly popular, but this complexity also creates increased computational demand when designing these devices through costly full-wave simulations. Treating these complex geometries by modeling them as homogeneous slabs can greatly speed up these computations. To this end, we introduce a simple and robust method to solve the branching problem in the homogenization of metamaterials. We start from the branch of the complex logarithm in the Nicolson-Ross-Weir method with the minimum absolute mean derivative in the low frequency range and enforce continuity. This is followed by comparing the reflectance, transmittance, and absorptance of the original and homogenized slabs. We use our method to demonstrate accurate and fast optical simulations of patterned PbS colloidal quantum dot solar cell films. We also compare patterned solar cells homogenized via equivalent models (wavelength-scale features) and effective models (sub-wavelength-scale features), finding that for the latter, agreement is almost exact, whereas the former contains small errors due to the unphysical nature of the homogeneity assumption for that size regime. This method can greatly reduce computational cost and thus facilitate the design of optical structures for solar cell applications.more » « less
-
Measurements of volcano deformation are increasingly routine, but constraining complex magma reservoir geometries via inversions of surface deformation measurements remains challenging. This is partly due to deformation modeling being limited to one of two approaches: computationally efficient semi-analytical elastic solutions for simple magma reservoir geometries (point sources, spheroids, and cracks) and computationally expensive numerical solutions for complex 3D geometries. Here, we introduce a pair of Graph Neural Network (GNN) based elasto-static emulators capable of making fast and reasonably accurate predictions (error upper bound: 15 %) of surface deformation associated with 3D reservoir geometries: a spheroid emulator and a general shape emulator, the latter parameterized with spherical harmonics. The emulators are trained on, and benchmarked against, boundary element (BEM) simulations, providing up to three orders of magnitude speed up compared to BEM methods. Once trained, the emulators can generalize to new reservoir geometries statistically similar to those in the training data set, thus avoiding the need for re-training, a common limitation for existing neural network emulators. We demonstrate the utility of the emulators via Bayesian Markov Chain Monte Carlo inversions of synthetic surface deformation data, showcasing scenarios in which the emulators can, and can not, resolve complex magma reservoir geometries from surface deformation. Our work demonstrates that GNN based emulators have the potential to significantly reduce the computational cost of inverse analyses related to volcano deformation, thereby bringing new insights into the complex geometries of magmatic systems.more » « less
-
The focus of this work is on the measurement and analysis of the radiative properties of polycrystalline SiO2 particle beds with various layer thicknesses. The particles are polydispersed with average diameters of 222, 150, and 40 μm. The spectral, directional–hemispherical reflectance and transmittance of the particle bed are measured at wavelengths from 0.4 to 1.8 μmusing a monochromator, and the reflectance measurement is extended to 15 μmusing a Fourier-transform infrared spectrometer. Particles are closely packed between two transparent windows for measuring the radiative properties. In the visible and near-infrared region up to 1.8 μm, the inverse adding–doubling method yields the effective absorption and scattering coefficients. The results suggest that short wavelength absorption needs to be included in modeling the behavior of particle beds due to multiple scattering. A discrete-scale Monte Carlo ray-tracing method is developed to model the radiative properties by assuming monodispersed spherical particles, and the simulated results compare well with measurements. The effective absorption and scattering coefficients of the particle beds obtained from the independent scattering theory are compared to those from the inverse method. The impact of dependent scattering on the packed beds is observed for smaller-sized particles.more » « less
-
Monte Carlo Simulations (MCSs) allow for the estimation of photon propagation through media given knowledge of the geometry and optical properties. Previous research has demonstrated that the inverse of this problem may be solved as well, where neural networks trained on photon distributions can be used to estimate refractive index, scattering and absorption coefficients. To extend this work, time-dependent MCSs are used to generate data sets of photon propagation through various media. These simulations were treated as stacks of 2D images in time and used to train convolutional networks to estimate tissue parameters. To find potential features that drive network performance on this task, networks were randomly generated. Generated networks were then trained. The networks were validated using 4-fold cross validation. The consistently performing top 10 networks typically had an emphasis on convolutional chains and convolutional chains ending in max pooling.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1115/1.4062188
