Revealing the Local Cosmic Web from Galaxies by Deep Learning (2008.01738v2)

Abstract: The 80% of the matter in the Universe is in the form of dark matter that comprises the skeleton of the large-scale structure called the Cosmic Web. As the Cosmic Web dictates the motion of all matter in galaxies and inter-galactic media through gravity, knowing the distribution of dark matter is essential for studying the large-scale structure. However, the Cosmic Web's detailed structure is unknown because it is dominated by dark matter and warm-hot inter-galactic media, both of which are hard to trace. Here we show that we can reconstruct the Cosmic Web from the galaxy distribution using the convolutional-neural-network-based deep-learning algorithm. We find the mapping between the position and velocity of galaxies and the Cosmic Web using the results of the state-of-the-art cosmological galaxy simulations, Illustris-TNG. We confirm the mapping by applying it to the EAGLE simulation. Finally, using the local galaxy sample from Cosmicflows-3, we find the dark-matter map in the local Universe. We anticipate that the local dark-matter map will illuminate the studies of the nature of dark matter and the formation and evolution of the Local Group. High-resolution simulations and precise distance measurements to local galaxies will improve the accuracy of the dark-matter map.

• The paper introduces a deep learning method that maps dark matter by analyzing galaxy positions and radial peculiar velocities using a CNN.
• The approach accurately replicates small-scale filamentary structures with average signal-to-noise ratios of at least 4.1 per pixel across simulations.
• This innovative method improves cosmic web reconstruction, offering enhanced simulations for galaxy formation and new tests for dark matter theories.

Revealing the Local Cosmic Web from Galaxies by Deep Learning

In the paper "Revealing the Local Cosmic Web from Galaxies by Deep Learning," Hong et al. present a method for mapping the dark-matter distribution in the local cosmic web using deep learning techniques. The research leverages galaxy data from Cosmicflows-3 and state-of-the-art cosmological simulations, namely Illustris-TNG and EAGLE, to train a convolutional neural network (CNN) that can infer the dark-matter structure from the spatial distribution and radial peculiar velocities of galaxies.

The paper addresses the challenge of mapping the cosmic web, a framework primarily composed of dark matter which, due to its non-luminous nature, is inherently difficult to observe. Conventionally, dark matter's presence has been surmised from its gravitational influence on baryonic matter. In this research, deep learning provides a novel approach to overcome these observational limitations by predicting the dark matter distribution with finer detail than traditional methods.

Core Methodology

The authors use a CNN-based architecture to analyze data from high-resolution simulations and observational catalogs. Their choice of input includes the spatial positions and radial peculiar velocities of galaxies, with the CNN trained to map these observations onto the dark-matter distribution. By employing different cosmological simulations, including the hydrodynamic models from Illustris-TNG and the dark-matter-only EAGLE simulation, the researchers ensure that the CNN can robustly generalize the mapping across varied simulation environments.

Key Findings and Numerical Results

Significantly, the paper finds that the CNN approach can successfully replicate small-scale filamentary structures within the cosmic web. The inclusion of radial peculiar velocity data is emphasized as crucial; models excluding this information fail to capture the intricate structure, essentially smoothing out the galaxy number distribution. The accuracy of this deep learning approach is validated by testing on independent simulation data not used in training, showing impressive alignment with the true dark-matter distributions.

The researchers report that their CNN model can detect dark matter structures with average signal-to-noise ratios of at least 4.1 per pixel at high Galactic latitudes. This level of precision marks a significant achievement over preceding endeavors in cosmic web reconstruction.

Implications and Future Directions

The reconstructed dark-matter map offers substantial potential for advancing cosmology. It can empower simulations of galaxy formation and evolution by providing accurate initial conditions. Furthermore, it opens avenues for testing dark matter theories by enabling cross-analysis with electromagnetic observations and gravitational wave data—potentially providing indirect evidence for dark matter compositions or interactions.

Future work would ideally expand on this framework by incorporating additional data sources and reducing systematic biases that arise from simulation assumptions. Greater resolution and accuracy in distance measurements, as well as integration of more extensive galaxy data, would refine the models further. Additionally, quantifying and reducing the theoretical uncertainties inherent in the simulation-to-observation mapping process will be crucial for optimizing this approach.

Overall, the paper highlights the power of combining cosmological simulations with advanced machine learning techniques to tackle longstanding challenges in astrophysics, marking a step forward in our quest to understand the large-scale structure of the universe and the role of dark matter within it.