Skip to main content
Logo image

THE GEOMETRY OF MATHEMATICAL METHODS

Section 5.2 Definition of Hermitian Matrices

Definition 5.2. Hermitian Matrices.

On an \(n\times m\) matrix, \(N\text{,}\) you can take the Hermitian adjoint (usually denoted with a dagger, \(\dagger\)) which means take both the (complex) conjugate Section 2.3 and the transpose Section 3.4, in either order
\begin{equation} N^\dagger=N^*{}^T\text{.}\tag{5.2.1} \end{equation}
An \(n\times n\) (square) matrix \(M\) is Hermitian if it equals its conjugate transpose, that is, if
\begin{equation} M^\dagger = M\text{.}\tag{5.2.2} \end{equation}
If a matrix \(M\) is both Hermitian and real, then \(M\) is called a symmetric matrix. An anti-Hermitian matrix is one for which the Hermitian adjoint is the negative of the matrix:
\begin{equation} M^\dagger = -M\text{.}\tag{5.2.3} \end{equation}
A matrix which is both anti-Hermitian and real is called antisymmetric.

Identifying Hermitian Matrices.

For example, let \(M\) be a \(2\times2\) complex matrix, so that
\begin{equation} M = \begin{pmatrix} a\amp b\\ c\amp d \end{pmatrix}\text{,}\tag{5.2.4} \end{equation}
with \(a,b,c,d\in\CC\text{,}\) and its Hermitian adjoint is
\begin{equation} M^\dagger = \begin{pmatrix} a^*\amp c^*\\ b^*\amp d^* \end{pmatrix}\text{,}\tag{5.2.5} \end{equation}
If \(M\) is Hermitian, then \(M^\dagger=M\text{,}\) so, we must have
\begin{equation} a^* = a, b^* = c, d^* = d\text{,}\tag{5.2.6} \end{equation}
i.e. \(a\) and \(d\) are real and \(c\) is the complex conjugate of \(b\text{.}\)
In index notation, if the components of \(M\) are denoted \(m_{ij}\text{,}\) then \(M\) is Hermitian if and only if
\begin{equation} m_{ij} = m^*_{ji}\tag{5.2.7} \end{equation}
for all \(i\text{,}\) \(j\text{.}\) (See also, Section 3.5.) Thus, the diagonal elements of a Hermitian matrix must be real, and the off-diagonal elements come in complex conjugate pairs, paired symmetrically across the main diagonal.

An Important Special Case of Hermitian Matrices.

An important special case of a Hermitian matrix can be constructed from any column vector \(v\) by computing its outer square, which in traditional vector notation would be written \(vv^\dagger\) and in bra/ket notation would be written \(|v\rangle\langle v|\text{.}\) (See Section 5.6 for a description of how such operators are used to produce projections.)