Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

What can one say about the set of all $n$-dimensional square matrices $A \in \text{GL}_n(\mathbb{C})$ that have an inverse with entries out of $\mathbb{C}$ with the properties:

  • unitary $:\Leftrightarrow A^*= A^{-1}$
  • hermitian $:\Leftrightarrow A^* = A$

where $A^*$ is the conjugate transpose of $A$.

What obviously follows is $$A^{-1} = A$$ The most simple matrix that is in this set is the identity matrix. Are there others? How do they look like?

share|cite|improve this question
What is a quadratic matrix? Is $A^H$ the conjugate transpose of $A$? – Mariano Suárez-Alvarez Jan 6 '11 at 20:32
Quadratic means $n \times n$: $n$ rows & $n$ columns; $A^H$ means the conjugate transpose. – a1337q Jan 6 '11 at 20:36
Strange. Most people would say square :) – Mariano Suárez-Alvarez Jan 6 '11 at 20:40
Sorry, I had this course in german ;) – a1337q Jan 6 '11 at 20:41
@a1337q: $A^*$ is more common than $A^H$ for conjugate transpose (though there seems to be little by way of very generally accepted notation). – Arturo Magidin Jan 6 '11 at 20:44
up vote 10 down vote accepted

Such a matrix is such that $A^2=I$, so it is diagonalizable, and its possible eigenvalues are $+1$ and $-1$. Since it is unitary, the eigenspaces corresponding to $1$ and to $-1$ are orthogonal.

Conversely, every diagonalizable matrix with eigenvalues contained in $\{+1,-1\}$ and orthogonal eigenspaces is of that form.

It follows that the set of your matrices is in bijection with the set of subspaces of $\mathbb C^n$. Explicitely: If $V$ is one such subspace, there is a unique linear transformation $f:\mathbb C^n\to\mathbb C^n$ such that $V$ and $V^\perp$ are eigenspaces for the eigenvalues $1$ and $-1$. The matrix $A_V$ of $f$ with respect to the standard basis of $\mathbb C^n$ satisfies your conditions. The bijection is $$V\in\{\mathrm{subspaces\;of\;}\mathbb C^n\}\leftrightarrow A_V.$$

As a consequence, considered as a whole, your set is a disjoint union of submanifolds homeomorphic to Grassmannian varieties. Literally books have been written about them.

share|cite|improve this answer
+1 Sounds creepy, but quite interesting. As often in maths. – a1337q Jan 6 '11 at 21:00

(in addition to Mariano's answer), these are sometimes called Householder transformations.

share|cite|improve this answer
+1 bc I know these but were actually unable to make a link (d'oh) – a1337q Jan 6 '11 at 20:54
Only when the multiplicity of $-1$ is exactly $1$. In any case, I never understood why people introduce a new name for what's already called an orthogonal reflection! – Mariano Suárez-Alvarez Jan 6 '11 at 20:55
@MAriano: Orthogonal reflection? How can that possibly become a Household name? ;-) – Aryabhata Jan 6 '11 at 21:20

A simple answer for $A \in \mathbb{R}^{2 \times 2}$:

Yes, there are more matrices. They are of the form

$A = \begin{pmatrix}\cos \alpha & \sin \alpha\\ \sin \alpha & -\cos \alpha \end{pmatrix}$.

It is obvious that these matrices are symmetric.

The inverse is $\frac{1}{(\cos \alpha) \cdot (-\cos \alpha) - sin^2 \alpha} \begin{pmatrix}-\cos \alpha & -\sin \alpha\\ -\sin \alpha & \cos \alpha \end{pmatrix} = \frac{-1}{\cos^2 \alpha + sin^2 \alpha} \begin{pmatrix}-\cos \alpha & -\sin \alpha\\ -\sin \alpha & \cos \alpha \end{pmatrix} = A$

Maybe you can find such forms for $\mathbb{C}^{n \times n}$, too.

I know this is not exactly an answer to your question. But I think this explicit, simple form is a nice answer for a part of the question.

share|cite|improve this answer
There's a small typo: you forgot the minus signs before the sines, while inverting $A$. – Andrea Orta Sep 12 '12 at 10:28
Thanks for mentioning it. – Martin Thoma Sep 12 '12 at 11:27
You're welcome! Oh, the signs are still missing in the previous step! :) – Andrea Orta Sep 12 '12 at 12:47

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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