Collaborative Research: RUI: New Insights from a Systematic Approach to Quasar Variability

天体物理学中一个长期存在的问题是了解星系在其一生中是如何形成和发展的。这样的理解对于揭示我们的宇宙是如何进化的，并深入了解我们银河系的起源是必要的。了解星系形成和演化的一个重要方面是研究类星体和其他活动星系核。它们本身也是有趣的天体物理现象，也是对相对论物理学的探索。它们与宿主星系共同演化，并追踪宇宙结构的演化。然而，大多数发现类星体的方法都是基于它们的宽带光谱能量分布的特性，几乎所有已知的类星体和候选类星体都来自使用某种类型的通量比或仅存在非热辐射的样本。变异性为类星体的发现提供了一种与光谱无关的方法。虽然变异性已经在单个或少数物体的基础上进行了大量研究，但基于变异性的类星体调查迄今为止仅限于天空的小区域，最多只有几千个物体和/或较差的时间分辨率。这项类星体变异性的研究采用了卡塔琳娜实时瞬变巡天（CRTS）数据集，该数据集覆盖了超过9年的基线时间内大约80%的天空。它的5亿个天体数据集目前包含了大约25万个已知类星体，50万个光度类星体候选者和大约100万个新的变异性选择类星体。这将形成迄今为止最大的类星体数据集。为了与CRTS开放数据政策保持一致，本项目中确定的所有类星体（和其他分类对象）将向社区发布。这将为类星体和更一般的变异性研究提供一个重要的新资源。所使用的统计方法也适用于任何不规则采样时间序列，并且将这些方法与机器学习技术相结合是数据密集型科学的案例研究。该项目与一所本科院校合作，直接加强了两名本科生的STEM教育。与加州理工学院数据驱动发现中心（CD3）的科学家合作，他们将接触到数据科学的前沿技术，包括数据挖掘的高水平使用和从这些大型数据集中提取有意义的结果。这个项目的数据产品也构成了美国和智利联合资助的拉塞雷纳数据科学学院的学生项目的基础，培养下一代处理大天文数据的应用工具。特别是，该项目将重点关注(i)类星体变异性特征，特别是特征时间尺度与物理参数（如亮度、黑洞质量和Eddington比率）的相关性；（ii）周期变异性作为超大质量黑洞双星的可能证据；（iii）用变率探测年轻尘埃覆盖的红类星体的暗度；（iv）量化变异性的波长依赖关系，以改善类星体选择和约束不同的物理过程模型。该研究将采用现代统计技术，可以自然地处理不规则采样的间隙时间序列，而无需重新投影或平滑。结合机器学习方法，这些将产生最佳的基于集合的结果，例如类星体选择的新变异性多色方法。该项目将成为进入LSST时代的光学类星体变异性研究的重点。在天空覆盖范围和类星体数量方面，它至少比以往的任何研究都要大两个数量级，在时间分辨率（观测次数/基线）方面则要好一个数量级。它还将大大增加已知的高可能性类星体候选者的数量，特别是在未被SDSS覆盖的天空区域。

A longstanding problem in astrophysics is to understand how galaxies form and develop throughout their lifetimes. Such understanding is necessary to uncover how our Universe evolved and to gain insight into the origin of our own Milky Way Galaxy. One important aspect of understanding galaxy formation and evolution is to study quasars and other active galactic nuclei. They are also interesting astrophysical phenomena in their own right and serve as a probe of relativistic physics. They co-evolve with their host galaxies and trace the evolution of cosmic structure. However, most of the methods for quasar discovery are based on the properties of their broadband spectral energy distributions, and nearly all known quasars and quasar candidates come from samples that use some type of flux ratios or just the presence of a non-thermal emission. Variability offers a spectrum-independent method for quasar discovery. Although variability has been much studied on the basis of individual or a few objects, variability-based quasar surveys have so far been limited to small dedicated regions of sky with at most a few thousand objects and/or poor time resolution. This study of quasar variability employs the Catalina Real-time Transient Survey (CRTS) data set, which covers about 80\% of the sky over a baseline greater than 9 years. Its 500 million object data set currently holds about 250,000 known quasars, 500,000 photometric quasar candidates and an estimated 1,000,000 new variability-selected quasars. This will form the largest quasar data set to date. In keeping with the CRTS Open Data policy, all quasars (and other classified objects) identified in this project will be released to the community. This will form a major new resource for both quasar and more general variability studies. The statistical methods to be used are also applicable to any irregular-sampled time series, and the combination of these with machine-learning techniques is a case study for data-intensive science.This project is a collaboration with a primarily undergraduate institution and directly enhances the STEM education of two undergraduates. Working with scientists at the Center for Data-Driven Discovery (CD3) at Caltech, they will be exposed to cutting-edge techniques in data science, including high level usage of data mining and extracting meaningful results from these large data sets. Data products from this project also form the basis for student projects at the joint US-Chile-funded La Serena School for Data Science, training the next generation in applied tools for handling big astronomical data.In particular, this project will focus on (i) the correlation of quasar variability features, particularly characteristic timescales, with physical parameters, such as luminosity, black hole mass, and the Eddington ratio; (ii) periodic variability as possible evidence for supermassive black hole binaries; (iii) variability as a probe of obscuration in young dust-enshrouded red quasars; and (iv) quantifying wavelength dependencies of variability to improve quasar selection and constrain different models of physical processes. The study will employ modern statistical techniques that can work naturally with irregularly-sampled gappy time series without the need for reprojection or smoothing. In combination with machine-learning methods, these will produce optimal ensemble-based results, such as new variability-polychromatic methods for quasar selection. This project will be a key study on optical quasar variability well into the LSST era. It is at least two orders of magnitude larger than any previous study in terms of sky coverage and number of quasars and an order of magnitude better in terms of time resolution (number of observations / baseline). It will also substantially increase the number of high likelihood quasar candidates known, particularly in the regions of the sky not covered by SDSS.