The four-dimensional Anderson model sits precisely at the upper critical dimension for the localization-delocalization transition, where the disorder is a random generalized function (white noise) and the associated elliptic SPDE requires renormalization just to define the operator. The precise open problem is: for the renormalized four-dimensional Anderson Hamiltonian constructed via singular SPDE techniques, prove or disprove the existence of a critical disorder strength lambda_c strictly between zero and infinity such that for lambda below lambda_c the spectrum is purely absolutely continuous (delocalized phase) while for lambda above lambda_c it is pure point (localized phase), and characterize the critical operator at lambda_c as a nontrivial renormalization group fixed point in the space of singular SPDEs.
This problem is open for several interlocking reasons. The singular SPDE framework, including regularity structures and paracontrolled calculus, provides existence and uniqueness of the renormalized operator but does not yet yield spectral-theoretic information because spectral analysis requires global properties of eigenfunctions while SPDE renormalization is inherently local. At dimension four the logarithmic divergences in the renormalization procedure introduce an additional free parameter (the renormalization scale), making the phase diagram two-dimensional rather than one-dimensional and obscuring where a critical point should lie. Furthermore, existing multiscale analysis techniques (Frohlich-Spencer, Bourgain-Kenig) are designed for operators with bounded or sub-Gaussian potentials, not for distributional ones, so the entire toolkit for proving localization at strong disorder must be rebuilt in this singular setting.
Solving this would accomplish several things simultaneously. It would provide the first rigorous example of an Anderson localization transition in a critical-dimension singular SPDE model, connecting the probabilistic theory of rough paths and regularity structures directly to spectral theory of random operators. It would also yield a concrete renormalization group flow in infinite-dimensional function space whose fixed points have direct physical meaning as critical quantum states, potentially informing the broader program of constructing quantum field theories via SPDE limits. The critical exponents governing how localization length diverges as lambda approaches lambda_c would give new universality predictions testable against numerical simulations.
The connection to the source papers is direct and multifaceted. The four-dimensional Anderson model paper explicitly identifies characterizing the spectral type of the renormalized operator as the central unresolved question after construction. The dynamical mean-field and SK model paper demonstrates that renormalization group and replica-symmetric free energy techniques from disordered systems can be made rigorous in high-dimensional limits, suggesting that similar averaging methods might be imported into the spectral analysis of the Anderson model. The Yau conjecture variance paper shows that nodal geometry of eigenfunctions in random settings can be studied statistically, hinting that even without full localization proofs one might characterize the critical eigenfunction geometry through variance asymptotics of spectral statistics analogous to the nodal length variance studied there.