In this work, we map the phase diagrams of one-dimensional quasiperiodic models using artificial neural networks. We observe that the multiclass classifier precisely distinguishes the various phases, namely the delocalized, multifractal, and localized phases, when trained on the eigenstates of the long-range Aubry-Andr\'e Harper (LRH) model. Additionally, when this trained multilayer perceptron is fed with the eigenstates of the Aubry-Andr\'e Harper (AAH) model, it identifies various phases with reasonable accuracy. We examine the resulting phase diagrams produced using a single disorder realization and demonstrate that they are consistent with those obtained from the conventional method of fractal dimension analysis. Interestingly, when the neural network is trained using the eigenstates of the AAH model, the resulting phase diagrams for the LRH model are less exemplary than those previously obtained. Further, we study binary classification by training the neural network on the probability density corresponding to the delocalized and localized eigenstates of the AAH model. We are able to pinpoint the critical transition point by examining the metric ``accuracy'' for the central eigenstate. The effectiveness of the binary classifier in identifying a previously unknown multifractal phase is then evaluated by applying it to the LRH model.
In this work, we map the phase diagrams of one-dimensional quasiperiodic models using artificial neural networks. We observe that the multiclass classifier precisely distinguishes the various phases, namely the delocalized, multifractal, and localized phases, when trained on the eigenstates of the long-range Aubry-Andr'e Harper (LRH) model. Additionally, when this trained multilayer perceptron is fed with the eigenstates of the Aubry-Andr'e Harper (AAH) model, it identifies various phases with reasonable accuracy. We examine the resulting phase diagrams produced using a single disorder realization and demonstrate that they are consistent with those obtained from the conventional method of fractal dimension analysis. Interestingly, when the neural network is trained using the eigenstates of the AAH model, the resulting phase diagrams for the LRH model are less exemplary than those previously obtained. Further, we study binary classification by training the neural network on the probability density corresponding to the delocalized and localized eigenstates of the AAH model. We are able to pinpoint the critical transition point by examining the metric ``accuracy'' for the central eigenstate. The effectiveness of the binary classifier in identifying a previously unknown multifractal phase is then evaluated by applying it to the LRH model.
We uncover a systematic structure in the single-particle phase diagram of the quasiperiodic Aubry-Andr\'e-Harper (AAH) model with power-law hoppings ($\ensuremath{\sim}\frac{1}{{r}^{\ensuremath{\sigma}}}$) when the quasiperiodicity parameter is chosen to be a member of the metallic mean family of irrational Diophantine numbers. In addition to the fully delocalized and localized phases, we find a coexistence of multifractal (localized) states with the delocalized states for $\ensuremath{\sigma}<1$ ($\ensuremath{\sigma}>1$). The fraction of delocalized eigenstates in these phases can be obtained from a general sequence, which is a manifestation of a mathematical property of the metallic mean family. The entanglement entropy of the noninteracting many-body ground states respects the area law if the Fermi level belongs in the localized regime while logarithmically violating it if the Fermi level belongs in the delocalized or multifractal regimes. The prefactor of the logarithmically violating term is significantly larger in the delocalized phase in comparison to that in the multifractal phase. Entanglement entropy shows the area law even in the delocalized regime for special filling fractions, which are related to the metallic means.
Due to the rapid advancement of quantum information theory, some concepts such as fidelity and entanglement entropy have been introduced into the study of quantum phase transitions, which can be used not only to identify novel matter phases but also to detect the critical point and describe the critical behavior of the quantum phase transitions. From the point of view of the metric space, these physical quantities can be understood as the distance between the two functions in the metric space. In this work, we study a class of quasi-periodic system represented by the generalized Aubry-André-Harper (AAH) model, by using the distance between various wavefunctions or density distribution functions in real space. The generalized AAH model, an ideal platform to understand Anderson localization and other novel quantum phenomena, provides rich phase diagrams including extended, localized, even critical (multifractal) phases and can be realized in a variety of experimental platforms. In the standard AAH model, we find that the extended and localized phases can be identified. In addition, there exists a one-to-one correspondence between two distinct distances. We are able to precisely identify the critical point and compute the critical exponent by fitting the numerical results of different system sizes. In the off-diagonal AAH model, a complete phase diagram including extended phase, localized phase, and critical phase is obtained and the distance of critical phases is intermediate between the localized phase and extended phase. Meanwhile, we apply the metric space method to the wave packet propagation and discover that depending on the phase, the distance between wave functions or density functions exhibits varying dynamical evolution behavior, which is characterized by the exponent of the power-law relationship varying with time. Finally, the distance between the state density distribution functions is proposed, and it effectively identifies distinct matter phases and critical points. The critical phase which displays a multifractal structure, when compared with the other two phases, has the large state density distribution function distance. In a word, by defining the distances of a function under different parameters, we provide not only a physical quantity to identify familiar phase transitions but also an intuitive way to identify different matter phases of unknown systems, phase transition points, and their critical behaviors.
Classification of quantum phases is one of the most important areas of research in condensed matter physics. In this work, we obtain the phase diagram of one-dimensional quasiperiodic models via unsupervised learning. Firstly, we choose two advanced unsupervised learning algorithms, Density-Based Spatial Clustering of Applications with Noise (DBSCAN) and Ordering Points To Identify the Clustering Structure (OPTICS), to explore the distinct phases of Aubry-André-Harper model and quasiperiodic p-wave model. The unsupervised learning results match well with traditional numerical diagonalization. Finally, we compare the similarity of different algorithms and find that the highest similarity between the results of unsupervised learning algorithms and those of traditional algorithms has exceeded 98\%. Our work sheds light on applications of unsupervised learning for phase classification.
Abstract Classification of quantum phases is one of the most important areas of research in condensed matter physics. In this work, we obtain the phase diagram of one-dimensional quasiperiodic models via unsupervised learning. Firstly, we choose two advanced unsupervised learning algorithms, namely, density-based spatial clustering of applications with noise (DBSCAN) and ordering points to identify the clustering structure (OPTICS), to explore the distinct phases of the Aubry–André–Harper model and the quasiperiodic p-wave model. The unsupervised learning results match well with those obtained through traditional numerical diagonalization. Finally, we assess similarity across different algorithms and find that the highest degree of similarity between the results of unsupervised learning algorithms and those of traditional algorithms exceeds 98%. Our work sheds light on applications of unsupervised learning for phase classification.
Distinguishing the dynamics of an Anderson insulator from a many-body localized (MBL) phase is an experimentally challenging task. In this work we propose a method based on machine learning techniques to analyze experimental snapshot data to separate the two phases. We show how to train three-dimensional convolutional neural networks (CNNs) using space-time Fock-state snapshots, allowing us to obtain dynamic information about the system. We benchmark our method on a paradigmatic model showing MBL ($t\text{\ensuremath{-}}V$ model with quenched disorder), where we obtain a classification accuracy of $\ensuremath{\approx}80%$ between an Anderson insulator and an MBL phase. We underline the importance of providing temporal information to the CNNs and we show that CNNs learn the crucial difference between an Anderson localized and an MBL phase, namely the difference in the propagation of quantum correlations. Particularly, we show that the misclassified MBL samples are characterized by an unusually slow propagation of quantum correlations, and thus the CNNs label them wrongly as Anderson localized. Finally, we apply our method to the case with quasiperiodic potential, known as the Aubry-Andr\'e model (AA model). We find that the CNNs have more difficulties in separating the two phases. We show that these difficulties are due to the fact that the MBL phase of the AA model is characterized by a slower information propagation for numerically accessible system sizes.
We apply various unsupervised machine learning methods for phase classification to investigate the finite-temperature phase diagram of the spinless Falicov-Kimball model in two dimensions. Using only particle occupation snapshots from Monte Carlo simulations as input, each technique, including a straightforward classification based on principal component analysis (PCA), successfully identifies the phase boundary between ordered and disordered phases, independent of the type of phase transition. Remarkably, these techniques also distinguish between the weakly localized and Anderson-localized regimes within the disordered phase, accurately identifying their crossover, which is a challenging task for standard methods. Among the machine learning approaches used, PCA based analysis outperforms more complex methods, such as neural network predictors and autoencoders. These results underscore the effectiveness of simple unsupervised techniques in examining phase transitions and electron localization in complex correlated systems.