# Do matrices with only elements along the main and anti-diagonals have a name?

To expand upon the title, I am wondering if there is a specific name for square matrices of the form: $$M = \begin{bmatrix} a_{11} & 0 & \cdots & 0 & \cdots & & 0 & b_{1n} \\ 0 & a_{22} & 0 & \cdots & & b_{2 \ i} & & 0 \\ 0 & \ddots & \ddots & & \cdot^{\textstyle \cdot^{\textstyle \cdot}} & 0 & & 0 \\ & & & a_{} & & \vdots & & \\ \vdots & 0 & \cdot^{\textstyle \cdot^{\textstyle \cdot}} & 0 & \ddots & 0 & & \vdots\\ & \cdots & b_{i \ n-(i+1)} & \cdots & & a_{i \ i} &\\ 0 & \cdot^{{\textstyle \cdot}^{\textstyle \cdot}} & & & & \ddots & \ddots & 0\\ b_{n 1} & 0 & \cdots & 0 & & \cdots & & a_{nn} \end{bmatrix} \\$$ where, $$M = A_{\text{diagonal}} + B_{\text{anti-diagonal}}$$ Essentially, M is a square matrix which is zero everywhere except the main and anti-diagonals. In other words, it is the sum of diagonal and anti-diagonal square matrices with the same dimension.

Apologies for the formatting, the normal LaTeX notation for up-right dots is not supported on this site.

• they have been called "X-matrices" in this post; conjugation with a permutation matrix brings them into a block-diagonal form (2x2 blocks along the main diagonal), so you could just work in that representation. Apr 3 '19 at 16:37
• @CarloBeenakker Is every real (complex) matrix orthonormal (unitary) equivalent to a x matrix? Apr 4 '19 at 19:25
• @AliTaghavi -- I don't think so, because not every matrix has a complete set of two-dimensional invariant subspaces, for example $\tiny\left( \begin{array}{cccc} 1 & 1 & 0 & 0 \\ 0 & 1 & 1 & 0 \\ 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 1 \\ \end{array} \right)$ Apr 5 '19 at 8:00
• +1 for the rising dots. Apr 7 '19 at 5:56
• @CarloBeenakker Thank you for your comment. I was not aware of the post referring to them as X matrices which is a very fitting name; however, I was hoping for a more standard terminology that is perhaps present in the literature. I will keep looking and report back at a later date. Apr 7 '19 at 14:42

The reason, I guess, why this name is not used more extensively, is that it seems more natural to reorder the basis vectors and work with $$2\times 2$$ block-diagonal matrices.