Fundamental properties of protoplanetary discs determined from simultaneous fits to thermal dust images and spectral energy distributions 2026 Fundamental properties of protoplanetary discs determined from simultaneous fits to thermal dust images and spectral energy distributions – B

Tim J. Harries · 2026

Read Fundamental properties of protoplanetary discs determined from simultaneous fits to thermal dust images and spectral energy distributions 2026 Fundamental properties of protoplanetary discs determined from simultaneous fits to thermal dust images and spectral energy distributions – B by Tim J. Harries (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-06230.txt.

Introduction

Harries presents a machine learning method producing dust continuum images and spectral energy distributions from radiative transfer training sets. With Bayesian optimization, he fits 1.3 millimetre ALMA ODISEA images of Ophiuchus protostellar discs and their spectral energy distributions to determine dust masses and radii. Good simultaneous fits emerge for Class II objects, though poorer for Class I and flat spectrum sources. Broader and shallower dust mass distributions than millimetre flux estimates predict raise high mass and low mass disc counts, through optical depth and temperature effects tied to size and inclination.

Catalog record

CREATORTim J. Harries
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE98181
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.06230

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Fundamental properties of protoplanetary discs determined from simultaneous fits to thermal dust images and spectral energy distributions 2026 Fundamental properties of protoplanetary discs determined from simultaneous fits to thermal dust images and spectral energy distributions – B Tim J. Harries Abstract We present a novel machine learning method that is capable of rapidly and accurately producing dust-continuum model images and spectral energy distributions from training sets created using a detailed radiative transfer code. We create a training set that encompasses the parameter space for protoplanetary discs, and then couple the trained machine learning method with a Bayesian optimisation algorithm. We then simultaneously fitted 1.3 mm ALMA ODISEA survey images of protostellar discs in ρ \rho Oph, and their spectral energy distributions, in order to determine fundamental discs parameters such as dust masses and radii. We find that good simultaneous fits may be found for the Class II objects in the survey, although the spectral fits are poorer for the Class I and flat spectrum sources. We find that the dust mass distributions of discs is broader and shallower than that predicted from 1.3 mm flux dust mass estimates, substantially increasing the numbers of objects with high-mass and low-mass discs. We show that this is due to a combination of optical depth and dust temperature effects, which are strongly related to the disc size and inclination constraints provided by the imaging fits. We show that there is a significant decrease in disc scale height and disc flaring when moving from the the Class I objects, to the flat spectrum sources, and the Class II discs. 1 Introduction Protostellar discs are the birthplace of new planetary systems, and understanding the conditions of dust and gas in these discs is key to making sense of the configurations of exoplanetary systems. One of the most basic quantities we wish to know is how much dust each protostellar disc contains, but unfortunately determining this parameter with certainty is proving challenging. As a first approximation it is often assumed that the emitting dust is optically thin and at a constant temperature ( T_{\rm dust} ). This enables the use of the sub-millimetre flux F_{\lambda} as a proxy for the dust mass: M_{\rm dust}=\frac{F_{\lambda}d^{2}}{\kappa_{\lambda}B_{\lambda}(T_{\rm dust})} (1) where d is the distance, \kappa_{\lambda} is the dust opacity, and B_{\lambda} is the Planck function ( 8 ) . This elegant formulation enables observers to simply estimate the dust mass, and it should be insensitive to the precise geometrical distribution of the dust. Broadly speaking this approach has remained unchanged for several decades, albeit with some small modifications, such as scaling T_{\rm dust} by the effective temperature of the protostar ( 3 ) . The advance in the observational capacity to observe in the sub-mm provided by ALMA has led to a host of sub-mm surveys of dust in nearby star forming regions, enabling observers to examine the evolution and environmental dependence of the dust mass of protoplanetary discs (see review by 31 ). Despite these successes, there is growing evidence that the main assumption the underpins sub-mm dust mass estimates (that the dust is optically thin to sub-mm radiation) isn’t necessarily true and that self-scattering by the emitting dust may be significant. This means that dust masses will be underestimated, perhaps by an order of magnitude in certain circumstances (e.g. 51 ; 7 ; 44 ; 28 ; 50 ). A more sophisticated approach is therefore necessary in order to accurately model the radiative transfer and thermodynamics of the disc. Despite the widespread use of sub-mm flux mass estimates, modern radiative transfer (RT) codes, predominantly based on Monte Carlo (MC) methods, are routinely used to model protoplanetary discs. These simulations self-consistently account for the self-scattering and should produce a much more reliable dust estimate. Until recently this required fitting the spectral energy distribution (SED), usually via a small grid of bespoke models created on an object-by-object basis e.g. 35 . This is a time-consuming business, as a typical RT model might take several minutes to run on a multi-core computer, and many thousands of models are needed to fit an SED and explore parameter degeneracies. However deep learning methods, using artificial neural networks (NN), may be used to speed up this process: Given a set of training data, and appropriately sampling disc parameter space, it is possible to take the disc parameters as inputs to a NN and the monochromatic fluxes of the model SED as outputs. The NN learns the complex non-linear, multi-dimensional relationship between the inputs and the outputs, and subsequently can produce an SED from an arbitrary set of disc parameters many orders of magnitude more quickly than the RT code. By coupling the trained NN with an optimiser it is possible to fit extant survey data of discs in nearby star-forming regions and use Markov chain Monte Carlo methods to explore parameter degeneracies. Studies that use this technique are starting to appear in the literature (e.g. 37 ; 22 ; 38 ). The ability to fit SED surveys is a powerful development, but disc parameters are not uniquely determined by fitting the SED alone, and a much tighter grip on parameter space may be exerted by employing other observables. ALMA surveys for example often provide size constraints on the disc thermal emission which helps constrain the disc radius. 7 demonstrated that by incorporating sub-mm disc-size constraints in tandem with SED fitting much more reliable dust mass estimates can be obtained. In this paper we extend the use of machine learning approaches to enable the generation of both SEDs and sub-mm imaging using a NN trained on models produced using a state-of-the-art radiative transfer code. By coupling the NNs with objective fitting methods were are able to simultaneously fit SEDs and imaging, thus producing reliable dust estimates for large surveys based on detailed RT modelling. In the following we first describe the neural network configurations that we use, and the production of our training data set. We subsequently apply the new technique to the ODISEA survey of protostellar discs in \rho Oph. We compare the resulting dust masses with those derived from 1.3 mm fluxes, and examine the role of optical depths on sub-mm dust mass estimates. Finally, we give our perspectives on the future applications of the method in enhancing not only our understanding of dust masses on a statistical basis but also our detailed knowledge of individual objects by simultaneous fitting of spectral, imaging and visibility data, from the dust continuum as well as spectral lines. 2 The machine learning algorithms Artificial neural networks rely on layers of neurons. In the simplest case of a fully connected, feed forward network an individual neuron in receives inputs from all neurons in the previous layer, and passes its output to all neurons in the subsequent layer. Consider an individual neuron i in layer \ell . It is associated with a bias value b_{i} and a set of weights w_{i,j} which are used to weight the outputs a_{j} of the neurons from the previous layer ( \ell-1 ) which contains N_{\ell-1} neurons. Thus the input to our individual neuron is given by x_{i}=\sum^{N_{\ell-1}}_{j=1}w_{i,j}a_{\ell-1,j}+b_{i} (2) If x_{i} is taken to be the output of the neuron i.e. a_{\ell,i}=x_{i}

Read the complete text