서브메뉴
검색
Leveraging Machine Learning and Hydrodynamic Simulations of Galaxy Formation to Illuminate the Effects of Incident Radiation Fields on Atomic Gas Cooling and Heating Functions
Leveraging Machine Learning and Hydrodynamic Simulations of Galaxy Formation to Illuminate the Effects of Incident Radiation Fields on Atomic Gas Cooling and Heating Functions
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105225
- ISBN
- 9798291566633
- DDC
- 523
- 서명/저자
- Leveraging Machine Learning and Hydrodynamic Simulations of Galaxy Formation to Illuminate the Effects of Incident Radiation Fields on Atomic Gas Cooling and Heating Functions
- 발행사항
- [Sl] : University of Michigan, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 174 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-03, Section: B.
- 주기사항
- Advisor: Avestruz, Camille.
- 학위논문주기
- Thesis (Ph.D.)--University of Michigan, 2025.
- 초록/해제
- 요약Galaxies and stars form out of clouds of gas that collapse under the influence of gravity. This collapse stops when the gas heats up enough for its outward thermal pressure to counteract the gravitational compression. The cooling and heating functions of gas determine how efficiently gas can cool down and heat up by respectively emitting or absorbing photons. Gas thermodynamics drives the formation and evolution of galaxies. In this dissertation, we explore how locally varying radiation fields impact gas cooling and heating functions. We begin with a discussion of galaxy formation and evolution, and the atomic line emission and ionization processes that cool and heat interstellar gas. We also describe the main computational tools used in this dissertation: photoionization codes, hydrodynamic simulations, and machine learning. We then present results from galaxies simulated by the Cosmic Reionization on Computers (CROC) project, a hydrodynamic simulation of the Epoch of Reionization (EoR). The EoR is the era when ionizing radiation from early galaxies and stars caused the gas in the intergalactic medium to transition from neutral to ionized. We find that the cooling and heating functions of gas in these simulated galaxies cannot be well-described with a spatially constant radiation field, demonstrating the necessity of accurately accounting for spatial variations in the radiation field intensity. Next, we apply a machine learning approach to the problem of accurately approximating cooling and heating functions in a locally varying radiation field. We train models to predict cooling and heating functions at fixed metallicity using gas properties and four or more photoionization rates describing the radiation field. These models are able to compute cooling and heating functions more accurately than existing interpolation table methods. At arbitrary metallicity, we are able to reduce the frequency of the largest errors compared to an interpolation table approach. Interpolating between fixed metallicity models is the main bottleneck to further improvement. In a corollary study, we train machine learning models where the only radiation field parameters are its average intensity in various energy bins. We find that only three bins (one just below and one just above the hydrogen ionization energy, and one bin corresponding to high energy photons) are sufficient to accurately capture the cooling and heating behavior at fixed metallicity. Finally, to investigate the effects of these various modeling choices in a physical context, we implement one set of our machine learning cooling and heating models into a hydrodynamic simulation of an isolated galaxy. To assess the effects of improved accuracy in cooling and heating function approximations, we compare the gas temperature-density phase diagrams of this simulation to the same simulation with cooling and heating rates calculated using an interpolation table. This dissertation lays groundwork for more accurate modeling of baryonic processes in next-generation hydrodynamic simulations. This will improve our understanding of both systematic errors from baryonic uncertainties in the constraints on cosmological parameters from large-scale structure surveys and our understanding of gas physics in galaxy formation.
- 일반주제명
- Astrophysics
- 일반주제명
- Physics
- 일반주제명
- Astronomy
- 키워드
- Galaxy formation
- 키워드
- Machine learning
- 키워드
- Galaxies
- 키워드
- Hydrodynamics
- 키워드
- Radiation
- 키워드
- Gas cooling
- 기타저자
- University of Michigan Physics
- 기본자료저록
- Dissertations Abstracts International. 87-03B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017359853
■00520260202105225
■006m o d
■007cr#unu||||||||
■020 ▼a9798291566633
■035 ▼a(MiAaPQ)AAI32271845
■035 ▼a(MiAaPQ)umichrackham006262
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a523
■1001 ▼aRobinson, David B.
■24510▼aLeveraging Machine Learning and Hydrodynamic Simulations of Galaxy Formation to Illuminate the Effects of Incident Radiation Fields on Atomic Gas Cooling and Heating Functions
■260 ▼a[Sl]▼bUniversity of Michigan▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a174 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-03, Section: B.
■500 ▼aAdvisor: Avestruz, Camille.
■5021 ▼aThesis (Ph.D.)--University of Michigan, 2025.
■520 ▼aGalaxies and stars form out of clouds of gas that collapse under the influence of gravity. This collapse stops when the gas heats up enough for its outward thermal pressure to counteract the gravitational compression. The cooling and heating functions of gas determine how efficiently gas can cool down and heat up by respectively emitting or absorbing photons. Gas thermodynamics drives the formation and evolution of galaxies. In this dissertation, we explore how locally varying radiation fields impact gas cooling and heating functions. We begin with a discussion of galaxy formation and evolution, and the atomic line emission and ionization processes that cool and heat interstellar gas. We also describe the main computational tools used in this dissertation: photoionization codes, hydrodynamic simulations, and machine learning. We then present results from galaxies simulated by the Cosmic Reionization on Computers (CROC) project, a hydrodynamic simulation of the Epoch of Reionization (EoR). The EoR is the era when ionizing radiation from early galaxies and stars caused the gas in the intergalactic medium to transition from neutral to ionized. We find that the cooling and heating functions of gas in these simulated galaxies cannot be well-described with a spatially constant radiation field, demonstrating the necessity of accurately accounting for spatial variations in the radiation field intensity. Next, we apply a machine learning approach to the problem of accurately approximating cooling and heating functions in a locally varying radiation field. We train models to predict cooling and heating functions at fixed metallicity using gas properties and four or more photoionization rates describing the radiation field. These models are able to compute cooling and heating functions more accurately than existing interpolation table methods. At arbitrary metallicity, we are able to reduce the frequency of the largest errors compared to an interpolation table approach. Interpolating between fixed metallicity models is the main bottleneck to further improvement. In a corollary study, we train machine learning models where the only radiation field parameters are its average intensity in various energy bins. We find that only three bins (one just below and one just above the hydrogen ionization energy, and one bin corresponding to high energy photons) are sufficient to accurately capture the cooling and heating behavior at fixed metallicity. Finally, to investigate the effects of these various modeling choices in a physical context, we implement one set of our machine learning cooling and heating models into a hydrodynamic simulation of an isolated galaxy. To assess the effects of improved accuracy in cooling and heating function approximations, we compare the gas temperature-density phase diagrams of this simulation to the same simulation with cooling and heating rates calculated using an interpolation table. This dissertation lays groundwork for more accurate modeling of baryonic processes in next-generation hydrodynamic simulations. This will improve our understanding of both systematic errors from baryonic uncertainties in the constraints on cosmological parameters from large-scale structure surveys and our understanding of gas physics in galaxy formation.
■590 ▼aSchool code: 0127.
■650 4▼aAstrophysics
■650 4▼aPhysics
■650 4▼aAstronomy
■653 ▼aGalaxy formation
■653 ▼aMachine learning
■653 ▼aGalaxies
■653 ▼aHydrodynamics
■653 ▼aRadiation
■653 ▼aGas cooling
■690 ▼a0605
■690 ▼a0596
■690 ▼a0606
■71020▼aUniversity of Michigan▼bPhysics.
■7730 ▼tDissertations Abstracts International▼g87-03B.
■790 ▼a0127
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17359853▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


