 "The goal of this study was to find the quantum phase transition at intermediate local disorder strengths on a Heisenberg chain. Exact diagonalization was used to find the reduced density matrices for a different number of consecutive spins for the lowest energy eigenstate of the Heisenberg model with an additional random field in z-direction at low and high disorder strengths. The resulting dataset representing extended and localized phases was used to train a neural network. Afterwards, the trained network was applied on intermediate disorder strengths to deduct the critical disorder strength for a phase transition. This phase transition was predicted for all system sizes to be around Wc = 2.5J for the system sizes L ∈ {8, 9, 10, 11, 12} and block sizes n ∈ [1, 6]. Low block and system sizes suffered from low accuracy and high losses in the machine learning model, whereas for n > 3 block sizes the Wc value showed smaller deviations from a previously published theoretical value Wc ≈ 3.6 calculated with entanglement entropy on systems up to L = 22. This deviation can be attributed to the effect of the smaller system sizes and the effect of open boundary conditions." 
0 Comments
Leave a Reply. |
|||