Grant: Alliance Grants
University of Victoria
$25,000.00
Deciphering the birth of planets using machine learning
EF 2022-23
Résumé
The idea that planets form in "protoplanetary disks" - gaseous disks surrounding newborn stars - dates back to Laplace. As planets form, they gravitationally interact with the host disk, and produce large scale structures, such as gaps and spiral arms. In the past decade, high-resolution imaging observations of protoplanetary disks have identified such structures in hundreds of disks. By carefully comparing observed disk structures with those produced in numerical simulations of gravitational disk-planet interactions, we can infer the presence of planets, and constrain their properties, such as masses and orbits. This information holds the key to deciphering how planets form, one of the four biggest questions in astronomy in Canada today as defined by Canadian Astronomy Long Range Plan 2020-2030.Currently, a key bottleneck in inferring the properties of forming planets from disk observations is that this process demands a huge amount of computational and human resources. This is because the expected observational signatures of disk-planet interactions are obtained using computationally intensive hydrodynamic simulations set up and tuned by human experts. We plan to develop a neural network that will greatly speed up this process, and potentially replace simulations all together. Once developed, our neural network will reduce the computational costs by orders of magnitude. Our program will enable astronomers to process the entire sample of disk imaging observations and obtain the statistics of planet formation for the first time. This will provide the most direct tests to planet formation theories to date.Our program will also result in significant economic and environmental benefits for Canada and the world. In simulating disk-planet interactions alone, astronomers consume 3,000,000 kwh of electricity per year, and generate 1,200 tons of CO2 emission. Our innovative technique will reduce these costs by orders of magnitude. In addition, the technique we plan to develop may be transferred to other fields in NSE, to help researchers reduce the computational costs in evolving physical systems in those fields.
Programme
Alliance Grants
Bénéficiaire
Emplacement
British Columbia, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
576488-2022
Name nom
Dong, Ruobing R
Department département
Physics and Astronomy
Organization id
7
Institution établissement
University of Victoria
Province en
British Columbia
Province fr
Colombie-Britannique
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2022
Competition year année de concours
2022
Award amount
25000
Program id
ALLRP
Program name en
Alliance Grants
Program name fr
Subventions Alliance
Group en
RESEARCH PARTNERSHIPS
Group fr
PARTENARIATS DE RECHERCHE
Committee code
96
Committee name en
RPP Internal Decision Cttee
Committee name fr
Cté de décision interne - PPR
Area of application code
9999
Area of application group en
Not available
Area of application group fr
Non disponible
Area of application en
Not available
Area of application fr
Non disponible
Research subject code
3200
Research subject group en
Astronomy and astrophysics
Research subject group fr
Astronomie et astrophysique
Research subject en
Astronomy and astrophysics
Research subject fr
Astronomie et astrophysique
Installment
inst:1
Partie
1
Nb partie
1
Application title
Deciphering the birth of planets using machine learning
Application summary
The idea that planets form in "protoplanetary disks" - gaseous disks surrounding newborn stars - dates back to Laplace. As planets form, they gravitationally interact with the host disk, and produce large scale structures, such as gaps and spiral arms. In the past decade, high-resolution imaging observations of protoplanetary disks have identified such structures in hundreds of disks. By carefully comparing observed disk structures with those produced in numerical simulations of gravitational disk-planet interactions, we can infer the presence of planets, and constrain their properties, such as masses and orbits. This information holds the key to deciphering how planets form, one of the four biggest questions in astronomy in Canada today as defined by Canadian Astronomy Long Range Plan 2020-2030.Currently, a key bottleneck in inferring the properties of forming planets from disk observations is that this process demands a huge amount of computational and human resources. This is because the expected observational signatures of disk-planet interactions are obtained using computationally intensive hydrodynamic simulations set up and tuned by human experts. We plan to develop a neural network that will greatly speed up this process, and potentially replace simulations all together. Once developed, our neural network will reduce the computational costs by orders of magnitude. Our program will enable astronomers to process the entire sample of disk imaging observations and obtain the statistics of planet formation for the first time. This will provide the most direct tests to planet formation theories to date.Our program will also result in significant economic and environmental benefits for Canada and the world. In simulating disk-planet interactions alone, astronomers consume 3,000,000 kwh of electricity per year, and generate 1,200 tons of CO2 emission. Our innovative technique will reduce these costs by orders of magnitude. In addition, the technique we plan to develop may be transferred to other fields in NSE, to help researchers reduce the computational costs in evolving physical systems in those fields.