Quantum impurity problems can be solved using the numerical renormalization
group (NRG), which involves discretizing the free conduction electron system
and mapping to a `Wilson chain'. It was shown recently that Wilson chains for
different electronic species can be interleaved by use of a modified
discretization, dramatically increasing the numerical efficiency of the RG
scheme [Phys. Rev. B 89, 121105(R) (2014)]. Here we systematically examine the
accuracy and efficiency of the `interleaved' NRG (iNRG) method in the context
of the single impurity Anderson model, the two-channel Kondo model, and a
three-channel Anderson-Hund model. The performance of iNRG is explicitly
compared with `standard' NRG (sNRG): when the average number of states kept per
iteration is the same in both calculations, the accuracy of iNRG is equivalent
to that of sNRG but the computational costs are signifficantly lower in iNRG
when the same symmetries are exploited. Although iNRG weakly breaks SU(N)
channel symmetry (if present), both accuracy and numerical cost are entirely
competitive with sNRG exploiting full symmetries. iNRG is therefore shown to be
a viable and technically simple alternative to sNRG for high-symmetry models.
Moreover, iNRG can be used to solve a range of lower-symmetry multiband
problems that are inaccessible to sNRG.
Quantum impurity problems can be solved using the numerical renormalization group (NRG), which involves discretizing the free conduction electron system and mapping to a `Wilson chain'. It was shown recently that Wilson chains for different electronic species can be interleaved by use of a modified discretization, dramatically increasing the numerical efficiency of the RG scheme [Phys. Rev. B 89, 121105(R) (2014)]. Here we systematically examine the accuracy and efficiency of the `interleaved' NRG (iNRG) method in the context of the single impurity Anderson model, the two-channel Kondo model, and a three-channel Anderson-Hund model. The performance of iNRG is explicitly compared with `standard' NRG (sNRG): when the average number of states kept per iteration is the same in both calculations, the accuracy of iNRG is equivalent to that of sNRG but the computational costs are signifficantly lower in iNRG when the same symmetries are exploited. Although iNRG weakly breaks SU(N) channel symmetry (if present), both accuracy and numerical cost are entirely competitive with sNRG exploiting full symmetries. iNRG is therefore shown to be a viable and technically simple alternative to sNRG for high-symmetry models. Moreover, iNRG can be used to solve a range of lower-symmetry multiband problems that are inaccessible to sNRG.
We show that the numerical renormalization group (NRG) is a viable multi-band impurity solver for Dynamical Mean Field Theory (DMFT), offering unprecedent real-frequency spectral resolution at arbitrarily low energies and temperatures. We use it to obtain a numerically exact DMFT solution to the Hund's metal problem for a three-orbital model with filling factor $n_d=2$. The ground state is a Fermi liquid. The one-particle spectral function has a strong particle-hole asymmetry, with a clear apparent power law for positive frequencies only. With increasing temperature it shows a coherence-incoherence crossover with spectral weight transfered from low to high energies and evolves qualitatively differently from a doped Mott insulator. The spin and orbital spectral functions show "spin-orbital separation": spin screening occurs at much lower energies than orbital screening. The renormalization group flows clearly reveal the relevant physics at all energy scales.