| 其他摘要 | Tidal evolution research helps us understand the evolution of binary star systems and star-planet systems. Tidal forces play a crucial role in the formation and evolution of close-in giant planet systems. Parameters such as orbits, rotation speeds, and angular momentum within these systems are influenced by tidal forces, thereby affecting the evolution of close-in giant planets. Studies have found that during the orbital contraction of a planet, the transfer of angular momentum often results in stars within hot Jupiter systems rotating faster than stars without planets. With the increasing amount of observational data, scientists have also discovered very massive hot Jupiters and brown dwarfs near some stars. Generally, these close-in giant planets experience strong tidal forces from their host stars, leading to evolutionary timescales typically of just a few hundred million years. Further research indicates that their orbital periods often synchronize with the stellar rotation periods, a state known as double synchronization. In this scenario, tidal friction in the system may become extremely weak or even disappear entirely. However, in Sun-like stars, magnetic braking causes the star to continually lose angular momentum, potentially driving the system out of double synchronization. Despite this, some studies suggest that even considering the angular momentum loss due to stellar magnetic braking, double synchronization in such systems might be maintained for long periods. Quantitative studies on how long double synchronization can last and under what conditions are still relatively scarce. Additionally, in the era of rapid advancements in AI technology, the intersection of machine learning and astronomy is gaining increasing attention. The application of machine learning in the study of tidal interactions between stars and planets remains largely unexplored. This research focuses on the evolution of star-planet systems and aims to advance the application of machine learning in this field.In this paper, we first use the MESA code to model the tidal interactions between stars and planets in detail, considering a wide parameter space for both stars and planets and simulating and analyzing approximately 25,000 samples. The results show that in cases where the star's convective envelope is relatively thin, the initial rotation period of the star and the mass of the planet significantly impact the system's evolution. When a planet is engulfed, the weak stellar magnetic braking means that the angular momentum transferred from the planet can still have a long-term effect on the star's rotation. For stars with thicker convective envelopes, the stronger magnetic braking makes the initial stellar rotation speed an important factor. In systems with fast initial rotation speeds, the loss of angular momentum is more pronounced, making stellar magnetic braking more significant than the angular momentum transfer from the planet. Conversely, when the star's initial rotation speed is slow, the weaker magnetic braking makes the planet's influence on the star's rotation period more noticeable. Additionally, through the simulation analysis of the WASP-19 system, we observed that hot Jupiters located very close to their host stars can introduce significant errors in the age estimation of stars using gyrochronology.We also specifically focused on the phenomenon of double synchronization in star-planet systems, particularly in systems involving hot Jupiters and brown dwarfs. Our research indicates that in systems with thin stellar convective envelopes and massive planets, double synchronization is more likely to occur when the star's rotation rate is synchronized with the planet's orbital rate. By applying theoretical constraints to observational data, we can identify systems that are potentially in a long-term double synchronization state and predict which systems may exhibit this phenomenon in the future. As the precision in measuring the tidal quality factor improves, we anticipate discovering more systems in long-term double synchronization. This advancement will enhance our understanding of the evolutionary processes and dynamics within star-planet systems.Finally, we employed artificial neural networks (ANN) to perform regression analysis on the evolutionary curves generated by MESA simulations. The results show that with only six initial parameters, we can accurately predict the evolutionary curves of four key evolutionary quantities, with minimal error. This finding highlights the efficiency and accuracy of machine learning methods in generating evolutionary curves.Additionally, by applying the LightGBM model and oversampling techniques, we classified the star-planet system samples. Our model successfully distinguished between four main types of evolutionary curves and identified samples in long-term double synchronization with 80% accuracy. This demonstrates the potential of machine learning in classifying and predicting the evolutionary behavior of star-planet systems. |
修改评论