Completing the localization theory of the disordered random lattice: the marginal dipolar regime, extended states and quantum diffusion at weak disorder, and coupling to external reservoirs

The Anderson model — a quantum particle or excitation on a lattice whose site energies are drawn at random from a distribution of width $W$ and whose sites are connected by hopping integrals $V$ — was introduced in 1958 with the result that sufficiently strong disorder localizes the eigenstates and forbids diffusion. The original argument, a locator-expansion convergence criterion, left four gaps that together define one problem: completing the localization and transport theory of the disordered random lattice. First, it excluded the marginal dipolar hopping law $V(r)\propto 1/r^3$ — and more generally long-range hopping $1/r^\alpha$ with $\alpha$ equal to the lattice dimension $d$ — as well as the strong-coupling regime $V\sim W$ . Second, its critical disorder ratio $(W/V)_0$ was an order-of-magnitude estimate built on a flat site-energy distribution, neglected self-energy corrections, and a connectivity-counting treatment of nonrepeating-path correlations. Third, it could not establish that perturbed eigenstates are labeled one-to-one by lattice sites, leaving possible rare delocalized exceptions. Fourth, it treated a closed, non-interacting lattice, omitting coupling to phonons or other reservoirs. Later literature resolved major parts — the scaling-theory picture, the numerically settled critical disorder $W_c=16.54$ (box distribution, cubic lattice) with exponent $\nu=1.57$ , delocalization of the three-dimensional dipolar case with anomalous Levy-flight transport, rigorous exponentially localized eigenfunctions in the localized regime, and variable-range hopping phenomenology for phonon-coupled insulators — but a precise open core survives: the rigorous extended-states and quantum-diffusion conjectures at weak disorder in three dimensions, the complete quantitative theory of the marginal $\alpha=d$ regime, and the microscopic theory of reservoir-coupled localization.

Background

In 1958 P. W. Anderson asked whether an excitation placed on one site of a disordered crystal would spread, and introduced what is now called the Anderson model: a tight-binding description in which the amplitude $a_j$ on lattice site $j$ evolves under a random on-site energy $E_j$ (drawn independently from a distribution of width $W$ , so $W$ measures the disorder strength) and hopping integrals $V_{jk}$ connecting sites $j$ and $k$ (Anderson 1958). The question came from two concrete physical settings recalled in his Nobel lecture (Anderson 1977): spin diffusion in dilute paramagnets, where spins interact through long-ranged dipole-dipole forces, and conduction in the impurity band of doped semiconductors.

His method was a locator expansion — a site-based perturbation series in the hopping, in which the unperturbed object is the single resonant site ('locator') $1/(E-E_j)$ rather than the free particle. Transport is decided by the convergence of this series for the self-energy, the energy shift of a site induced by virtual hops to other sites, whose terms are products of hopping integrals and energy denominators summed over nonrepeating paths (self-avoiding walks that never revisit a site). Convergence keeps the self-energy real and confines the eigenstate near one site: no diffusion. Divergence produces an imaginary self-energy and an excitation that escapes. The resulting localization theorem required the hopping to decay faster than $1/r^3$ with distance $r$ . The dipolar case $V(r)=A/r^3$ is exactly marginal: the single-jump sum over resonant neighbors diverges only logarithmically, so any transport is non-exponential and extremely slow, and the criterion of the theorem does not apply.

Four limitations of that analysis are treated here as one problem. (a) The marginal dipolar case, and more generally long-range hopping $V(r)\propto 1/r^\alpha$ with $\alpha$ equal to the lattice dimension $d$ , together with the strong-coupling (weak-disorder) regime $V\sim W$ , fall outside the convergence theorem. (b) The critical disorder ratio $(W/V)_0$ separating localized and conducting behavior was only an order-of-magnitude estimate, resting on a flat site-energy distribution $P(E)=1/W$ , neglect of the self-energy inside the energy denominators except for a lower cutoff, and a treatment of correlations among nonrepeating paths only through the connectivity $K$ (the number of independent directions a path can leave a site). (c) Within this Brillouin-Wigner-type framework — a perturbation scheme in which the unknown eigenenergy itself enters the denominators of the series and is determined self-consistently — the paper could not guarantee a one-to-one assignment between unperturbed site states and perturbed eigenstates, leaving open the existence of rare delocalized ('pathological') states. (d) The model is a closed, non-interacting lattice: coupling to phonons or other external reservoirs, ubiquitous in real materials, is absent and may enable transport even when the isolated model localizes.

The subsequent literature built the standard framework. Abou-Chacra, Thouless and Anderson (1973) reorganized the expansion into a self-consistent theory of localization that removes the nonrepeating-path restriction. The one-parameter scaling theory of Abrahams, Anderson, Licciardello and Ramakrishnan (1979) predicted that in one and two dimensions all states are localized for arbitrarily weak disorder, while in three dimensions a mobility edge — the boundary in the energy-disorder plane separating localized from extended states — moves with disorder and disappears at a critical disorder strength. For the three-dimensional cubic model with site energies drawn from a uniform ('box') distribution of width $W$ and nearest-neighbor hopping $t$ , the accepted reference values are the critical disorder $W_c=16.54\,t$ and the correlation-length exponent $\nu=1.57$ (Slevin and Ohtsuki 2018; Šuntajs, Prosen and Vidmar 2021); the correlation length — the length scale beyond which localized eigenfunctions decay — diverges at the transition as $\xi\propto|W-W_c|^{-\nu}$ . Evers and Mirlin (2008) review the resulting theory of Anderson transitions, including multifractality — anomalous scaling of eigenfunction intensities intermediate between fully localized and fully extended states. For long-range hopping, Levitov (1990) argued by real-space renormalization-group methods that dipole-like interactions ( $\alpha=d$ ) destroy localization, and Deng et al. (2016) found that the resulting three-dimensional extended states carry anomalous Levy-flight transport — spreading dominated by rare long jumps with a power-law step distribution — and are extended but not always ergodic, with multifractal (non-ergodic) states, which do not sample phase space uniformly, taking over as the lattice filling is reduced. In the opposite direction, the question of what a bath does to localization underlies many-body localization — the failure of thermalization in interacting disordered systems (Basko, Aleiner and Altshuler 2006) — while variable-range hopping, thermally assisted tunneling between sites whose distance and energy mismatch are traded off optimally, is the standard finite-temperature transport mechanism of Anderson insulators (Banerjee and Altman 2016; Feigel'man and Kravtsov 2019).

On the mathematical side, localization is rigorous at strong disorder, and in one dimension at any disorder: the spectrum is pure point (eigenvalues with square-summable eigenfunctions) with exponentially localized eigenfunctions, which settles the eigenstate-labeling question inside the localized regime. But for weak disorder in three dimensions the existence of extended states and of quantum diffusion — the spread of a wavepacket with a finite diffusion constant — remains an outstanding conjecture in mathematical physics (Black, Drogin and Hernández 2025). Rigorous delocalization is proven only for neighboring models: block Anderson models in dimensions $d\ge 7$ (Yang and Yin 2025), random band matrices in $d\ge 3$ (Dubova, Yang, Yau and Yin 2025), and time-averaged quantum diffusion for the lattice model in $d\ge 2$ (Black, Drogin and Hernández 2025). The first direct, energy-resolved experimental observation of the three-dimensional Anderson transition appeared only in 2026 (Yu et al. 2026).

Problem Statement

The problem is to complete the localization and transport theory of the disordered random lattice — the Anderson model — precisely in the regimes that the 1958 locator-expansion convergence argument did not cover. These four gaps form one problem with a single scientific objective, the full localization-delocalization phase diagram, spectral structure, and observable transport law of the model, and they support one another: the marginal and strong-coupling regimes (1) and the critical parameters (2) are the two halves of the phase boundary; the spectral labeling question (3) is the structure of the two phases that boundary separates; and the reservoir coupling (4) determines which of those single-particle predictions survive as observable transport in real solids. Concretely: (1) for the marginal dipolar law $V(r)\propto 1/r^3$ and the general long-range family $V(r)\propto 1/r^\alpha$ with $\alpha$ equal to the lattice dimension $d$ , and for the strong-coupling regime $V\sim W$ , determine whether localization survives at all, the character (ergodic, non-ergodic extended, multifractal) of the states, the anomalous transport laws, and the crossover structure between regimes; (2) determine the critical disorder ratio separating localized and delocalized behavior and the associated critical exponents with controlled quantitative accuracy, going beyond the 1958 estimate's flat site-energy distribution, neglected self-energy corrections, and connectivity-counting treatment of nonrepeating-path correlations; (3) establish the spectral structure of both phases: whether the localized regime carries a pure point spectrum whose eigenstates admit a one-to-one labeling by lattice sites, and whether delocalized exceptions exist — in modern terms, the existence of the mobility edge and of extended states with quantum diffusion at weak disorder in three dimensions; (4) determine how coupling to external reservoirs (phonons, baths) modifies observable transport: when isolated-lattice localization survives as slow reservoir-assisted hopping and what its microscopic description is. The critical numbers $W_c=16.54\,t$ and $\nu=1.57$ for the box-distribution cubic model are already established at the numerical/phenomenological level, and the localized-regime half of (3) is already established at the rigorous level. The open core that these gaps leave consists of three items. (i) A rigorous construction of extended states and quantum diffusion at weak disorder for the standard three-dimensional lattice model, which is the weak-disorder regime to which the strong-coupling regime $V\sim W$ of (1) belongs. (ii) The complete quantitative theory of the marginal $\alpha=d$ regime including its crossover structure and marginal phases. (iii) A microscopic theory of reservoir-coupled localization.

References

P. W. Anderson (1958). Absence of Diffusion in Certain Random Lattices. Physical Review 109, 1492-1505. DOI 10.1103/PhysRev.109.1492, https://doi.org/10.1103/PhysRev.109.1492
P. W. Anderson (1977). Local Moments and Localized States (Nobel Lecture, 8 December 1977). https://www.nobelprize.org/uploads/2018/06/anderson-lecture-1.pdf
R. Abou-Chacra, D. J. Thouless, P. W. Anderson (1973). A selfconsistent theory of localization. Journal of Physics C: Solid State Physics 6, 1734-1752. DOI 10.1088/0022-3719/6/10/009, https://doi.org/10.1088/0022-3719/6/10/009
E. Abrahams, P. W. Anderson, D. C. Licciardello, T. V. Ramakrishnan (1979). Scaling Theory of Localization: Absence of Quantum Diffusion in Two Dimensions. Physical Review Letters 42, 673-676. DOI 10.1103/PhysRevLett.42.673, https://doi.org/10.1103/PhysRevLett.42.673
L. S. Levitov (1990). Delocalization of vibrational modes caused by electric dipole interaction. Physical Review Letters 64, 547-550. DOI 10.1103/PhysRevLett.64.547, https://doi.org/10.1103/PhysRevLett.64.547
F. Evers, A. D. Mirlin (2008). Anderson transitions. Reviews of Modern Physics 80, 1355-1417. DOI 10.1103/RevModPhys.80.1355, arXiv:0707.4378, https://doi.org/10.1103/RevModPhys.80.1355
D. M. Basko, I. L. Aleiner, B. L. Altshuler (2006). Metal-insulator transition in a weakly interacting many-electron system with localized single-particle states. Annals of Physics 321, 1126-1205. DOI 10.1016/j.aop.2005.11.014, arXiv:cond-mat/0506617, https://doi.org/10.1016/j.aop.2005.11.014
F. Robicheaux, N. M. Gill (2014). Effect of random positions for coherent dipole transport. Physical Review A 89, 053429. DOI 10.1103/PhysRevA.89.053429, arXiv:1401.4776, https://doi.org/10.1103/PhysRevA.89.053429
T. Xu, R. V. Krems (2015). Quantum walk and Anderson localization of rotational excitations in disordered ensembles of polar molecules. New Journal of Physics 17, 065014. DOI 10.1088/1367-2630/17/6/065014, arXiv:1501.05063, https://doi.org/10.1088/1367-2630/17/6/065014
X. Deng, B. L. Altshuler, G. V. Shlyapnikov, L. Santos (2016). Quantum Levy Flights and Multifractality of Dipolar Excitations in a Random System. Physical Review Letters 117, 020401. DOI 10.1103/PhysRevLett.117.020401, arXiv:1604.03820, https://doi.org/10.1103/PhysRevLett.117.020401
S. Banerjee, E. Altman (2016). Variable-Range Hopping through Marginally Localized Phonons. Physical Review Letters 116, 116601. DOI 10.1103/PhysRevLett.116.116601, arXiv:1511.03676, https://doi.org/10.1103/PhysRevLett.116.116601
Z. Lenarčič, E. Altman, A. Rosch (2018). Activating Many-Body Localization in Solids by Driving with Light. Physical Review Letters 121, 267603. DOI 10.1103/PhysRevLett.121.267603, arXiv:1806.04772, https://doi.org/10.1103/PhysRevLett.121.267603
K. Slevin, T. Ohtsuki (2018). Critical Exponent of the Anderson Transition Using Massively Parallel Supercomputing. Journal of the Physical Society of Japan 87, 094703. DOI 10.7566/JPSJ.87.094703, https://doi.org/10.7566/JPSJ.87.094703
S. Whitlock, H. Wildhagen, H. Weimer, M. Weidemüller (2019). Diffusive to Nonergodic Dipolar Transport in a Dissipative Atomic Medium. Physical Review Letters 123, 213606. DOI 10.1103/PhysRevLett.123.213606, arXiv:1809.07532, https://doi.org/10.1103/PhysRevLett.123.213606
M. V. Feigel'man, V. E. Kravtsov (2019). Electron-phonon cooling power in Anderson insulators. Physical Review B 99, 125415. DOI 10.1103/PhysRevB.99.125415, arXiv:1809.09888, https://doi.org/10.1103/PhysRevB.99.125415
A. Kutlin, I. Khaymovich (2020). Renormalization to localization without a small parameter. SciPost Physics 8, 049. DOI 10.21468/SciPostPhys.8.4.049, arXiv:2001.06493, https://doi.org/10.21468/SciPostPhys.8.4.049
K. Kawa, P. Machnikowski (2020). Diffusion of excitations and power-law localization in strongly disordered systems with long-range coupling. Physical Review B 102, 174203. DOI 10.1103/PhysRevB.102.174203, arXiv:2009.05432, https://doi.org/10.1103/PhysRevB.102.174203
V. Ros, M. Müller (2021). Fluctuation-driven transitions in localized insulators: Intermittent metallicity and path chaos precede delocalization. Physical Review B 104, 094205. DOI 10.1103/PhysRevB.104.094205, https://doi.org/10.1103/PhysRevB.104.094205
J. Šuntajs, T. Prosen, L. Vidmar (2021). Spectral properties of three-dimensional Anderson model. Annals of Physics 435, 168469. DOI 10.1016/j.aop.2021.168469, arXiv:2103.05680, https://doi.org/10.1016/j.aop.2021.168469
X. Deng, I. M. Khaymovich, A. L. Burin (2024). Superdiffusion in a random two-dimensional system with time-reversal symmetry and long-range hopping. Physical Review B 109, 174208. DOI 10.1103/PhysRevB.109.174208, arXiv:2205.14715, https://doi.org/10.1103/PhysRevB.109.174208
V. Temkin, A. S. Ioselevich (2024). Coexistence of localized and extended states in the Anderson model with long-range hopping. Annals of Physics 462, 169620. DOI 10.1016/j.aop.2024.169620, arXiv:2309.06345, https://doi.org/10.1016/j.aop.2024.169620
F. Yang, J. Yin (2025). Delocalization of a general class of random block Schrödinger operators. arXiv:2501.08608, https://arxiv.org/abs/2501.08608
S. Dubova, F. Yang, H.-T. Yau, J. Yin (2025). Delocalization of Non-Mean-Field Random Matrices in Dimensions $d\ge 3$ . arXiv:2507.20274, https://arxiv.org/abs/2507.20274
A. Black, R. Drogin, F. Hernández (2025). Self-consistent equations and quantum diffusion for the Anderson model. arXiv:2506.06468, https://arxiv.org/abs/2506.06468
X. Yu, K. Xie, H. M. Quach, Y. Guo, M. Niranjan, S. Barré, J.-P. Banon, A. Aspect, N. Cherroret, V. Josse (2026). Direct Observation of the Three-Dimensional Anderson Transition with Ultracold Atoms in a Disordered Potential. arXiv:2602.07654, https://arxiv.org/abs/2602.07654
S. Liu, Y. Shi, Z. Zhang (2026). Localization and unique continuation for the Anderson-Bernoulli model with long-range hopping on $\mathbb{Z}$ . arXiv:2607.03472, https://arxiv.org/abs/2607.03472
E. Safonova, A. Lunkin, M. Feigel'man (2026). Analytical Theory for Anomalous Diffusion in the Anderson Model with Heavy Tails. arXiv:2608.10868, https://arxiv.org/abs/2608.10868