arXiv Science⌕ Search

arXiv · 1607.08519

A self-consistent theory of localization in nonlinear random media

Abstract

The self-consistent theory of localization is generalized to account for a weak quadratic nonlinear potential in the wave equation. For spreading wave packets, the theory predicts the destruction of Anderson localization by the nonlinearity and its replacement by algebraic subdiffusion, while classical diffusion remains unaffected. In 3D, this leads to the emergence of a subdiffusion-diffusion transition in place of the Anderson transition. The accuracy and the limitations of the theory are discussed.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Nicolas Cherroret. 2016-10-27. A self-consistent theory of localization in nonlinear random media. https://doi.org/10.1088/0953-8984%2F29%2F2%2F024002

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

Numerical and theoretical studies of the Anderson transition in the chiral symmetry classes with weak topology

Combining lattice-model simulations with a renormalization group (RG) analysis of effective theories, we investigate the nature of the Anderson transition in chiral symmetry classes with one-dimensional (1D) weak topology. In the simulation study, we extend previous transfer matrix analyses to the chiral symplectic class, and study numerical Lyapunov exponents via a finite-size scaling (FSS) analysis that assumes spatially isotropic scaling. The analysis shows that, as in the other two chiral symmetry classes, the weak topology induces an intermediate quasi-localized (QL) phase between metal and Anderson insulator phases. In this QL phase, the localization length of wave functions diverges exclusively along the direction of the 1D weak topology. A comparison of the universal critical exponents across the three chiral symmetry classes indicates that the quantum criticality of the metal-to-QL transition varies, depending on the nature of the time-reversal symmetry. Motivated by this numerical observation, we revisit and extend our previous two-dimensional (2D) RG analysis to all three chiral classes, now newly incorporating a one-loop renormalization of the weak topological term in the analysis. The revised analysis reveals that a quasi-localized strong-coupling fixed point previously reported in the chiral unitary class is unstable under this new inclusion; instead, the strong-coupling phase is entirely governed by a stable fixed point with conventional localized character. Nevertheless, in the chiral unitary and chiral symplectic classes, the RG analysis still yields the hallmark of the 1D weak topology through the spatially anisotropic scaling of the Anderson transition criticality. These theoretical findings suggest that a conclusive numerical identification of the 2D QL phase requires a finite-size scaling approach that accommodates generic (anisotropic) spatial scaling.

cond-mat.dis-nn↗

Toward a Physics of Deep Learning and Brains

Deep neural networks and brains both learn and share superficial similarities: processing nodes are likened to neurons and adjustable weights are likened to modifiable synapses. But can a unified theoretical framework be found to underlie them both? Here we show that the equations used to describe neuronal avalanches in living brains can also be applied to cascades of activity in deep neural networks. These equations are derived from non-equilibrium statistical physics and show that deep neural networks learn best when poised between absorbing and active phases. Because these networks are strongly driven by inputs, however, they do not operate at a true critical point but within a quasi-critical regime -- one that still approximately satisfies crackling noise scaling relations. By training networks with different initializations, we show that maximal susceptibility is a more reliable predictor of learning than proximity to the critical point itself. This provides a blueprint for engineering improved network performance. Finally, using finite-size scaling we identify distinct universality classes, including Barkhausen noise and directed percolation. This theoretical framework demonstrates that universal features are shared by both biological and artificial neural networks.

cond-mat.dis-nn↗

Exact cost of the Euclidean Maximum Traveling Salesman Problem in one dimension

We study the random Euclidean Maximum Traveling Salesman Problem: finding a Hamiltonian cycle of maximal total cost through any set of $N$ cities on a line, where each edge cost is a function of its Euclidean length. We characterize the maximizing cycles for convex increasing, concave increasing, and convex decreasing cost functions. For points sampled independently and uniformly from the unit interval, we derive explicit expressions for the average maximal cost for arbitrary $N$ and asymptotic expansions with power-law edge costs.

cond-mat.dis-nn↗