I want to consider the crossed module: $H \xrightarrow{t} Aut(H)$ for the case where $H = GL_n(\mathbb{Z}) = Aut(T^n)$ is the automorphism group of the $n$-torus. Any suggestions on how to understand this automorphism group, $Aut(GL_n(\mathbb{Z}))$, explicitly?

  • 4
    $\begingroup$ I'm not sure of the meaning of your first sentence, but for the second one, the automorphism group for $n\ge 3$ is reduced to the "obvious" one, namely $\mathrm{PGL}_n(\mathbf{Z})\rtimes\{\tau\}$ where $\tau$ is the inverse-transpose involutive automorphism. (Essentially, Mostow rigidity reduces to compute the normalizer in the real automorphism group, i.e., to show that $\mathrm{SL}_n(\mathbf{Z})$ equals its own normalizer in $\mathrm{SL}_n(\mathbf{R})$.) However it's very plausible that this was known earlier in this case by a direct algebraic approach. $\endgroup$
    – YCor
    Jan 6, 2019 at 17:31
  • $\begingroup$ @YCor Thank you! My first sentence was just a motivation in case it was helpful. Would you like to rewrite this as an answer so I can select it? It would be helpful if you added a reference or added a few more details about the semi-direct product with tau. $\endgroup$
    – cheyne
    Jan 6, 2019 at 18:24
  • $\begingroup$ I'm lazy at the moment to fill in the details, this is why a comment is maybe better to start with. $\endgroup$
    – YCor
    Jan 6, 2019 at 18:26
  • $\begingroup$ I completely understand. Thanks again! $\endgroup$
    – cheyne
    Jan 6, 2019 at 18:45

1 Answer 1


Hua and Reiner described this in their paper "Automorphisms of the unimodular group" (for them "unimodular" means determinant of absolute value 1, hence it's ${\rm GL}_n(\mathbf Z)$ rather than just ${\rm SL}_n(\mathbf Z)$) that appeared in 1951 here. The end result (their Theorem 4) is that the automorphisms of ${\rm GL}_n(\mathbf Z)$ are generated by inner automorphisms, the conjugate-transpose map $A \mapsto (A^{-1})^\top$, the map $A \mapsto (\det A)A$ (they include it "for even $n$ only," which must mean they have a way of getting it from the other generators for odd $n$ -- I haven't read the paper closely to see where that is indicated [Edit: see comment below]), and one additional automorphism when $n=2$.

  • 2
    $\begingroup$ For odd $n$ the endomorphism $A\mapsto \det(A)A$ is not injective (it maps $-I$ to $I$). $\endgroup$
    – YCor
    Jan 7, 2019 at 0:54
  • $\begingroup$ Ah, of course, good point. $\endgroup$
    – KConrad
    Jan 7, 2019 at 1:07
  • $\begingroup$ Thank you (both) very much. $\endgroup$
    – cheyne
    Jan 7, 2019 at 2:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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