This tutorial presents a sequence of self-contained numerical renormalization-group (NRG) calculations for the single-impurity Anderson model with NRG Ljubljana. Each numbered lesson adds one physical or numerical idea and can be carried out independently.
The supplied templates contain pre-generated, symmetry-resolved Hamiltonian and operator matrices. Mathematica is used only to regenerate these model matrices; it is not needed to run the calculations.
Install NRG Ljubljana and source its shell setup so that adapt, instantiate,
and nrg can be found on PATH. With a source installation this typically
means:
source "$HOME/nrgljubljana/share/nrgljubljanavars.sh"Then run all five lessons from the top-level directory:
makeEach calculation writes its numerical input, diagnostic output, and calculated
observables under the corresponding lesson's run/ directory. Remove all
calculated results with:
make cleanRun only the first lesson with make lesson1, or directly with
cd 01_minimal && ./run.sh.
| Lesson | Calculation |
|---|---|
01_minimal |
Map the bath to a Wilson chain and solve the symmetric SIAM |
02_thermodynamics |
Extract impurity entropy and susceptibility and follow Kondo screening |
03_rg_flow |
Follow the many-body flow from a local moment to the screened Fermi liquid |
04_spectral_function |
Resolve the Kondo resonance and Hubbard satellites with FDM-NRG |
05_z_averaging |
Average shifted logarithmic grids to reduce discretization artifacts |
Lessons 1 through 5 are tested with NRG Ljubljana 2026.09. To run them, install NRG Ljubljana and ensure the following are available:
adaptinstantiatemakenrg- Bash
Lesson 5 additionally requires broaden and a standard awk program.
Lesson 2 also uses awk to subtract matching thermodynamic tables. Python 3.9
or newer and Matplotlib are optional; they are needed only to recreate the
supplied SVG figures in Lessons 2 through 5.
Neither Mathematica nor nrginit is needed for the supplied calculations. They
are required only by the optional commands, documented in the lessons, that
regenerate the model matrices.
- NRG Ljubljana documentation
- NRG Ljubljana examples
- Uberlandia 2019 course
- Copenhagen 2021 course
- DMFT_NRG_KLM, which similarly uses pre-generated model matrices in a larger calculation
The tutorial material in this repository is released under CC0 1.0 Universal. NRG Ljubljana is a separate project distributed under its own license.