Here $\delta=\frac{1}{2}\sum_{\alpha>0}\alpha$, where the sum is over all positive simple roots for some root system.

Lemma: Suppose $\mu\in\Lambda^+$ (dominant weight), and $\nu=\sigma^{-1}\mu$ for some $\sigma\in W$ (Weyl group). Then $(\nu+\delta,\nu+\delta)\leq(\mu+\delta,\mu+\delta)$, with equality holding iff $\sigma=1$ or $\mu=0$.

Proof: We have

\begin{aligned} (\nu+\delta,\nu+\delta) &= (\mu+\sigma\delta,\mu+\sigma\delta)\\ &= (\mu,\mu)+(\sigma\delta,\mu)+(\mu,\sigma\delta)+(\sigma\delta,\sigma\delta)\\ &= (\mu,\mu)+2(\mu,\sigma\delta)+(\delta,\delta)+(\delta,\mu)+(\mu,\delta)-2(\mu,\delta)\\ &= (\mu+\delta,\mu+\delta)-2(\mu,\delta-\sigma\delta). \end{aligned}

The claim follows if $(\mu,\delta-\sigma\delta)\geq 0$. Since $\delta$ is strongly dominant, $\sigma\delta\preceq\delta$. Thus $\delta-\sigma\delta\geq 0$, thus if $\delta-\sigma\delta=\sum_i c_i\alpha_i$, we have $c_i\geq 0$ for all $i$. Then $$ (\mu,\delta-\sigma\delta)=\sum_i c_i(\mu,\alpha_i)\geq 0 $$ since $(\mu,\alpha_i)\geq 0$ as $\mu$ is dominant. Then equality holds iff $(\mu,\delta-\sigma\delta)=0$. If $\mu\neq 0$, then $\delta-\sigma\delta=0$, or $\delta=\sigma\delta$. Since $\delta$ is strongly dominant, necessarily $\sigma=1$.

At the end, I don't follow why if $\mu\neq 0$, necessarily $\delta-\sigma\delta=0$. Couldn't $\mu$ be orthogonal to the $\alpha_i$ for which $c_i\neq 0$, but possibly not orthogonal to simple roots $\alpha_j$ where $c_j=0$? If feel like you could only make the conclusion if $\mu$ is assumed to be strongly dominant, not just dominant.


In my copy the proof continues as follows

... with equality only if $(\mu,\delta-\sigma\delta)=0$, i.e. $(\mu,\delta)=(\mu,\sigma\delta)=(\nu,\delta),$ or $(\mu-\nu,\delta)=0$. But $\mu-\nu$ is a sum of positive roots, and $\delta$ is strongly dominant, so $\mu=\nu$.

The point is that as $\mu\preceq\nu$ the inner product $(\nu-\mu,\delta)$ can vanish only if $\nu=\mu$. The inner product of $\delta$ and any positive root is, as you seem to know, strictly positive.

May be this is a version conflict? Mine is the 3rd printing, revised in 1980.

  • 1
    $\begingroup$ Because the key step is verbatim from Humphreys I am switching to CW. $\endgroup$ – Jyrki Lahtonen Mar 10 '17 at 8:11
  • $\begingroup$ Thanks, that's not present in mine, I'm using an old library copy from 1972. Perhaps I should get a newer version... $\endgroup$ – Nastassja Mar 10 '17 at 8:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.