Artificial Intelligence–Assisted Inversion (AIAI) of Synthetic Type Ia Supernova Spectra
Artificial Intelligence–Assisted Inversion (AIAI) of Synthetic Type Ia Supernova Spectra
复制标题
DOI：
10.3847/1538-4365/ab9a3b
复制
发表时间：
2019-11
期刊：
影响因子：
--
通讯作者：
Xingzhuo Chen;L. Hu;Lifan Wang
中科院分区：
文献类型：
--
作者：
Xingzhuo Chen;L. Hu;Lifan Wang
我们生成约 100,000 个 1a 型超新星（SNe Ia）的模型光谱，形成光谱库，用于构建理论模型的人工智能辅助反演（AIAI）算法。作为第一次尝试，我们将研究限制在 B 波段最大值附近的时间，并使用代码 TARDIS 计算具有 2000 至 10000 Å 宽光谱波长覆盖范围的理论光谱。基于理论计算光谱库，我们利用多残差卷积神经网络构建了 AIAI 算法，以检索不同离子种类对理论光谱的高度混合光谱剖面的贡献。人们发现 AIAI 在区分耦合原子跃迁引起的光谱模式方面非常强大，并且能够定量测量不同离子种类的贡献。通过将 AIAI 算法应用于一组观察良好的 SN Ia 光谱，我们证明该模型可以对这些 SNe Ia 的化学结构产生强大的约束。利用AIAI推导的化学结构，我们成功地重建了观测数据，从而证实了该方法的有效性。我们表明超新星 Ia 的光变曲线衰减率与喷射物中光球层上方 56Ni 的量相关。我们检测到 56Ni 质量随时间明显减少，这可归因于其放射性衰变。我们的代码和模型光谱可在网站 https://github.com/GeronimoChen/AIAI-Supernova 上获取。
We generate ∼100,000 model spectra of Type 1a supernovae (SNe Ia) to form a spectral library for the purpose of building an artificial intelligence–assisted inversion (AIAI) algorithm for theoretical models. As a first attempt, we restrict our studies to the time around B-band maximum and compute theoretical spectra with a broad spectral wavelength coverage from 2000 to 10000 Å using the code TARDIS. Based on the library of theoretically calculated spectra, we construct the AIAI algorithm with a multiresidual convolutional neural network to retrieve the contributions of different ionic species to the heavily blended spectral profiles of the theoretical spectra. The AIAI is found to be very powerful in distinguishing spectral patterns due to coupled atomic transitions and has the capacity to quantitatively measure the contributions from different ionic species. By applying the AIAI algorithm to a set of well-observed SN Ia spectra, we demonstrate that the model can yield powerful constraints on the chemical structures of these SNe Ia. Using the chemical structures deduced from AIAI, we successfully reconstructed the observed data, thus confirming the validity of the method. We show that the light-curve decline rate of SNe Ia is correlated with the amount of 56Ni above the photosphere in the ejecta. We detect a clear decrease of 56Ni mass with time that can be attributed to its radioactive decay. Our code and model spectra are available on the website https://github.com/GeronimoChen/AIAI-Supernova.