1: \begin{abstract}
2: A general framework for non-abelian symmetries is presented for
3: matrix-product and tensor-network states in the presence of
4: well-defined orthonormal local as well as effective basis sets.
5: The two crucial ingredients, the Clebsch-Gordan algebra for
6: multiplet spaces as well as the Wigner-Eckart theorem for
7: operators, are accounted for in a natural, well-organized, and
8: computationally straightforward way. The unifying
9: tensor-representation for quantum symmetry spaces, dubbed
10: \QSpace, is particularly suitable to deal with standard
11: renormalization group algorithms such as the numerical
12: renormalization group (NRG), the density matrix renormalization
13: group (DMRG), or also more general tensor networks such as the
14: multi-scale entanglement renormalization ansatz (MERA).
15: %
16: In this paper, the focus is on the application of the
17: non-abelian framework within the NRG. A detailed analysis is
18: presented for a fully screened spin-3/2 three-channel Anderson
19: impurity model in the presence of conservation of total spin,
20: particle-hole symmetry, and \SU{3} channel symmetry. The same
21: system is analyzed using several alternative symmetry scenarios.
22: This includes the more traditional symmetry setting \SSSS, the
23: larger symmetry \ASC, and their much larger enveloping
24: symplectic symmetry \SSP. These are compared in detail,
25: including their respective dramatic gain in numerical
26: efficiency. In the appendix, finally, an extensive introduction
27: to non-abelian symmetries is given for practical applications,
28: together with simple self-contained numerical procedures to
29: obtain Clebsch-Gordan coefficients and irreducible operators
30: sets. The resulting \QSpace tensors can deal with any set of
31: abelian symmetries together with arbitrary non-abelian
32: symmetries with compact, \ie finite-dimensional, semi-simple Lie
33: algebras.
34: \end{abstract}