学术文献库

This thesis introduces a set of methods for testing models of modified gravity using galaxy clusters. In particular, a technique for constraining models with a chameleon screening is introduced. In addition, the outlined technique is expanded to test a wider class of models, such as the theory of emergent gravity. Finally, the first part of the thesis is concluded by adapting the mentioned tests for model independent constraints. The obtained results indicate that galaxy clusters can be used to obtain some of the most powerful constraints on cosmological scales. The second part of the thesis is dedicated to the topic of cosmological emulators. More specifically, a technique of emulating cosmological N-body simulation output data based on machine learning is introduced. Generative adversarial networks (GANs) are used to emulate dark matter-only as well as hydrodynamical simulation data. In addition, N-body modified gravity simulations are explored as well. The presented investigation of the GAN algorithm shows that such emulators offer a powerful, fast and efficient way of producing simulation output data with different cosmological parameters. The power spectrum analysis indicates a 1-20% difference between the training and the generated data depending on the dataset used and whether Gaussian smoothing is applied or not.