Disorder and Localization Frontiers

Disorder research asks when spatial randomness changes transport, spectral structure, and phase organization. Localization is one outcome, but not every resistive or slowly relaxing sample is localized. A feasible project fixes the dimensionality, symmetry class, interactions, and observable before interpreting a size or temperature trend.

Helpful background. Anderson localization supplies the noninteracting mechanism; scaling theory explains the role of dimension and symmetry.

Localized eigenstates and insulating transport differ

For a normalized lattice eigenstate, the inverse participation ratio is

$$P_2=\sum_i|\psi_i|^4.$$

An extended state spread uniformly over $N$ sites has $P_2\sim N^{-1}$; a localized state has a nonvanishing large-system limit when its localization length stays finite. Critical multifractal states require a more refined scaling analysis.

Established: noninteracting localization and Anderson transitions depend on symmetry and dimension, as reviewed by Evers and Mirlin. Active: effects of interactions, correlated disorder, topology, and realistic environments. Not established by one finite sample: that a large $P_2$ or low conductance implies an asymptotic localized phase.

Transport also depends on contacts, inelastic processes, and temperature. A spectral calculation on an isolated Hamiltonian cannot automatically predict a measured resistance.

Competing approaches and controls

Transfer-matrix methods study localization lengths in long quasi-one-dimensional geometries. Exact diagonalization probes eigenfunctions and spectra at limited volume. Transport calculations add leads and boundary conditions. Field theory identifies universality classes but relies on correctly retaining the system’s symmetries.

Use the full distribution of a diagnostic when rare events matter. The mean and median can scale differently. Correlated or quasiperiodic potentials are distinct ensembles and should not be substituted for independent disorder merely because all produce irregular-looking landscapes.

Starter project: mean versus typical localization

Simulate a one-dimensional tight-binding chain with onsite disorder. Compute transmission or a localization-length proxy for many independent samples and several lengths. Compare arithmetic and geometric means and the distribution of $\log T$, with $T$ the transmission probability.

Deliver convergence with sample count, checks against the clean-chain limit, and an explicit treatment of contacts. A laptop or workstation is sufficient. Test whether rare resonant samples dominate the arithmetic mean. The failure test is a fitted localization length that changes strongly with sample count or lead mismatch.

Extend to research by introducing a controlled correlation length in the disorder, a symmetry-preserving hopping perturbation, or a topological band structure. Ask whether the relevant universality or observable distribution changes, rather than merely reporting stronger or weaker localization.

Open questions and boundaries

How do interactions and dephasing alter a single-particle localization diagnostic? Which rare-region effects survive in a finite device? Can disorder enhance one useful property while suppressing another, such as protecting a gap yet degrading coherent transport?

For interacting isolated systems, follow many-body localization; the noninteracting scaling results do not settle that debate. For materials, specify whether the bottleneck is intrinsic disorder physics or uncertainty in the microscopic disorder model.

Entry points and evidence horizon

Evers and Mirlin gives the foundational classification and numerical framework. The cited review is from 2008 and is used for established structure; links were checked on 2026-10-11. Active interacting questions are assessed separately rather than presented as conclusions of that review.

References

• Evers, F., and Mirlin, A. D., “Anderson transitions,” Reviews of Modern Physics 80, 1355–1417 (2008), arXiv:0707.4378.