Optimized loops

• Restarting protocol for both U-driven and T-driven lines

• Hybrid Distributed/Restarting protocol for (U,T) scans

  Distributed loops require the Parallel Computing Toolbox, but if not installed they just fall back on regular loops, without throwing errors. So take it as an optional dependency, to enhance performance.

Generic bandwidth and frequency resolution

• Fixed an important bug affecting the D ≠ 1 runs.

• Fixed some subtleties arising whenever the frequency array contains the origin:

  • Modified the Fermi function to properly handle the beta = inf case, even when w = 0 (the exponential of inf*0 would give a NaN, which propagates dangerously downstream until breaking the dmft-loop when the norm of gloc_old-gloc is evaluated)
  • Modified the Luttinger integral so to carefully exclude the nonanalytic pole at w = 0, which would totally break the sum-rule.

Everything carefully tested, we could be approaching a stable release.

Structured namespaces + New markers

Changelog

• Significant refactoring of the code: definition of +plot and +phys namespaces
• Local inclusion of legacy code and reference material, for permanent reachability
• Reliable implementation of the Luttinger integral, as a sharp marker for the zero temperature MIT
• Introduction of a "strong correlation marker", for characterizing the phases on the whole U-T plane

Full rewrite of Nájera code

MATLAB port of Nájera python notebook.

Also featuring:

• Stable convergence handling through linear self-mixing

• Reliable extraction of the quasiparticle weight from the computed self-energies

• Pretty plots of spectral functions and other quantities