caf9f1eb9da6317b.tex
1: \begin{abstract}
2:   Using a recently developed perturbative renormalization group (RG)
3:   scheme, we calculate ana\-ly\-ti\-cal\-ly the spectral function of a
4:   Kondo impurity for either large frequencies $\w$ or large magnetic
5:   field $B$ and arbitrary frequencies.  For large $\w\gg \max[B,T_K]$
6:   the spectral function decays as $1/\ln^2[\w/T_K]$ with prefactors
7:   which depend on the magnetization.  The spin-resolved spectral
8:   function displays a pronounced peak at $\w \sim B$ with a
9:   characteristic asymmetry.  In a detailed comparison with results
10:   from numerical renormalization group (NRG) and bare perturbation
11:   theory in next-to-leading logarithmic order, we show that our
12:   perturbative RG scheme is controlled by the small parameter
13:   $1/\ln[\max(\w,B)/T_K]$.  Furthermore, we assess the ability of the
14:   NRG to resolve structures at finite frequencies.
15: \end{abstract}
16: