Hermitian Matrix - Definition, Properties, and Examples

The Matrix That Makes Quantum Measurements Come Out Real

Every physical quantity you can measure, energy, position, momentum, is represented in quantum mechanics by a matrix whose numbers are often complex. Yet every measurement you ever record is a plain real number. The bridge between complex bookkeeping and real results is one condition on the matrix: it must be Hermitian.

That single rule, matrix equals its conjugate transpose, forces the diagonal to be real and the eigenvalues to be real. Those two facts are exactly what physics needs, and they fall straight out of the definition.

What Is a Hermitian Matrix?

A Hermitian matrix is a square matrix $A$ with complex entries that is equal to its conjugate transpose:

$$A = A^{H}.$$

Here $A^{H}$ (also written $A^{*}$ or $A^{\dagger}$, and sometimes called the self-adjoint condition) is the matrix you get by transposing $A$ and then replacing every entry with its complex conjugate. Written entry by entry, $A$ is Hermitian when

$$a_{ij} = \overline{a_{ji}} \quad \text{for all } i, j.$$

The bar denotes the complex conjugate: the conjugate of $a + bi$ is $a - bi$. So the entry in row $i$, column $j$ must be the mirror-image entry's conjugate. A Hermitian matrix is the complex-number cousin of a real symmetric matrix.

What Is the Conjugate Transpose?

The conjugate transpose stacks two operations you already know, one after the other. Define it before testing anything, because skipping either step is the classic source of wrong answers.

Take

$$A = \begin{bmatrix} 3 & 2 + i \ 2 - i & 1 \end{bmatrix}.$$

Step 1, take the transpose, swapping rows and columns:

$$A^{T} = \begin{bmatrix} 3 & 2 - i \ 2 + i & 1 \end{bmatrix}.$$

Step 2, conjugate every entry (flip the sign of each imaginary part):

$$A^{H} = \overline{A^{T}} = \begin{bmatrix} 3 & 2 + i \ 2 - i & 1 \end{bmatrix}.$$

Since $A^{H} = A$, this matrix is Hermitian. The order of the two steps does not matter; conjugating then transposing gives the same result.

What Are the Properties of a Hermitian Matrix?

• The diagonal entries are always real. The condition $a_{ii} = \overline{a_{ii}}$ forces each diagonal entry to equal its own conjugate, which can only happen for a real number.

• Off-diagonal entries are conjugate mirror images. Whatever sits at position $(i, j)$, the conjugate of it sits at position $(j, i)$.

• The eigenvalues are always real, even when the matrix itself is packed with complex numbers. This is the property physics leans on, and the next section derives it.

• The trace and the determinant are real. The trace is a sum of real diagonal entries, and the determinant works out to a real number.

• Sums stay Hermitian, and real scalar multiples stay Hermitian. If $A$ and $B$ are Hermitian, so is $A + B$, and so is $kA$ for any real number $k$.

• The inverse is Hermitian whenever it exists, and every Hermitian matrix is normal, so it can be diagonalised.

Are the diagonal elements always real? Yes, without exception. That is the fastest first check: if any diagonal entry has a non-zero imaginary part, the matrix cannot be Hermitian, and you can stop there.

Why Are the Eigenvalues of a Hermitian Matrix Real?

Take an eigenvalue $\lambda$ of a Hermitian matrix $A$, with eigenvector $x \neq 0$, so that $Ax = \lambda x$. The argument runs one step at a time.

Multiply on the left by $x^{H}$:

$$x^{H} A x = \lambda, x^{H} x.$$

Now take the conjugate transpose of $Ax = \lambda x$, using $A^{H} = A$:

$$x^{H} A = \overline{\lambda}, x^{H}.$$

Multiply that on the right by $x$:

$$x^{H} A x = \overline{\lambda}, x^{H} x.$$

The left sides match, so the right sides are equal:

$$\lambda, x^{H} x = \overline{\lambda}, x^{H} x.$$

Because $x \neq 0$, the quantity $x^{H} x$ is a positive real number, so divide it out:

$$\lambda = \overline{\lambda}.$$

A number that equals its own conjugate is real. So every eigenvalue is real. The related idea of an eigenvalue is worth a full read if this is your first meeting with it.

How Is a Hermitian Matrix Different from a Symmetric Matrix?

A symmetric matrix satisfies $A = A^{T}$: it is its own transpose, with no conjugation involved. That test works cleanly for real matrices. A Hermitian matrix satisfies $A = A^{H}$: transpose and conjugate.

• Symmetric is the right idea for real entries. Every real symmetric matrix is automatically Hermitian, because conjugating a real number changes nothing.

• Hermitian is the right idea once complex entries appear. A matrix like $\begin{bmatrix} 3 & 2 + i \ 2 - i & 1 \end{bmatrix}$ is Hermitian but not symmetric, since $2 + i \neq 2 - i$.

• The close relative on the other side is the skew-Hermitian matrix, where $A^{H} = -A$; its diagonal entries are purely imaginary or zero.

So symmetric is what you test in a real-number world, and Hermitian is the correct generalisation the moment $i$ enters the entries.

Examples of Hermitian Matrix

The set runs from a direct 2x2 check, through the most common false-positive, to real eigenvalues and a full Hermitian-plus-skew decomposition. A first mistake many learners make is declaring a matrix Hermitian the instant its diagonal looks real, without ever testing the off-diagonal pairs.

Example 1

Is $A = \begin{bmatrix} 3 & 2 + i \ 2 - i & 1 \end{bmatrix}$ Hermitian?

Check the two conditions. Diagonal entries $3$ and $1$ are real. Off-diagonal: the $(1,2)$ entry is $2 + i$, and its mirror $(2,1)$ entry should be $\overline{2 + i} = 2 - i$, which it is.

Final answer: Yes, $A$ is Hermitian.

Example 2

Is $A = \begin{bmatrix} 2 & 3 + i \ 3 + i & 5 \end{bmatrix}$ Hermitian?

Wrong attempt. A student sees the real diagonal entries $2$ and $5$, notices both off-diagonal entries are the same, and concludes "symmetric-looking, so yes, Hermitian."

Test that against the rule. For Hermitian, the $(2,1)$ entry must be the conjugate of the $(1,2)$ entry. Here $(1,2) = 3 + i$, so $(2,1)$ would need to be $\overline{3 + i} = 3 - i$. But the actual $(2,1)$ entry is $3 + i$, not $3 - i$.

Correct conclusion. Build the conjugate transpose and compare:

$$A^{H} = \begin{bmatrix} 2 & 3 - i \ 3 - i & 5 \end{bmatrix} \neq A.$$

Final answer: No, $A$ is not Hermitian. Equal off-diagonal entries is the symmetric test, not the Hermitian one.

Example 3

Find the entry that makes $A = \begin{bmatrix} 4 & 5 - 2i \ z & 7 \end{bmatrix}$ Hermitian.

The $(2,1)$ entry $z$ must equal the conjugate of the $(1,2)$ entry:

$$z = \overline{5 - 2i} = 5 + 2i.$$

Final answer: $z = 5 + 2i$, giving $A = \begin{bmatrix} 4 & 5 - 2i \ 5 + 2i & 7 \end{bmatrix}$.

Example 4

Verify that $A = \begin{bmatrix} 1 & 2 + 3i & 4i \ 2 - 3i & 0 & 6 - 7i \ -4i & 6 + 7i & 3 \end{bmatrix}$ is Hermitian.

Diagonal entries $1$, $0$, $3$ are real. Now the three off-diagonal pairs:

$$\overline{2 + 3i} = 2 - 3i, \qquad \overline{4i} = -4i, \qquad \overline{6 - 7i} = 6 + 7i.$$

Each conjugate matches the mirrored entry across the diagonal.

Final answer: Yes, $A$ is Hermitian.

Example 5

Show that the eigenvalues of $A = \begin{bmatrix} 2 & 1 - i \ 1 + i & 3 \end{bmatrix}$ are real.

Form the characteristic equation $\det(A - \lambda I) = 0$:

$$(2 - \lambda)(3 - \lambda) - (1 - i)(1 + i) = 0.$$

Evaluate the product term:

$$(1 - i)(1 + i) = 1 - i^{2} = 1 + 1 = 2.$$

Expand and simplify:

$$\lambda^{2} - 5\lambda + 6 - 2 = 0,$$

$$\lambda^{2} - 5\lambda + 4 = 0,$$

$$(\lambda - 1)(\lambda - 4) = 0.$$

Final answer: The eigenvalues form the set ${1, 4}$, both real, exactly as the theory promised.

Example 6

Split $A = \begin{bmatrix} 1 & 2i \ 0 & 3 \end{bmatrix}$ into a Hermitian part and a skew-Hermitian part.

Any square matrix decomposes as $A = \tfrac{1}{2}(A + A^{H}) + \tfrac{1}{2}(A - A^{H})$. First build $A^{H}$:

$$A^{H} = \begin{bmatrix} 1 & 0 \ -2i & 3 \end{bmatrix}.$$

Hermitian part:

$$P = \tfrac{1}{2}(A + A^{H}) = \tfrac{1}{2}\begin{bmatrix} 2 & 2i \ -2i & 6 \end{bmatrix} = \begin{bmatrix} 1 & i \ -i & 3 \end{bmatrix}.$$

Skew-Hermitian part:

$$S = \tfrac{1}{2}(A - A^{H}) = \tfrac{1}{2}\begin{bmatrix} 0 & 2i \ 2i & 0 \end{bmatrix} = \begin{bmatrix} 0 & i \ i & 0 \end{bmatrix}.$$

Final answer: $A = P + S$, where $P$ is Hermitian and $S$ is skew-Hermitian. Check: $P + S = \begin{bmatrix} 1 & 2i \ 0 & 3 \end{bmatrix} = A$.

Where Are Hermitian Matrices Used?

"How do complex matrices produce real physical answers?"

In quantum mechanics, every observable, energy, spin, momentum, is a Hermitian operator. Because a Hermitian matrix has real eigenvalues, and measured values are exactly those eigenvalues, the theory can use complex numbers throughout while still predicting the real quantities an experiment records. Werner Heisenberg's 1925 matrix mechanics put this at the foundation of the subject.

Where Hermitian matrices do real work:

• Quantum mechanics. Observables are Hermitian precisely so that measurement outcomes are guaranteed real. The Hermitian matrix reference covers this link in depth.

• Signal processing and statistics. Covariance and correlation structures over complex signals are Hermitian, which keeps their eigenvalues (the variances) real and non-negative.

• Numerical linear algebra. Hermitian matrices are normal, so they diagonalise with an orthonormal set of eigenvectors, which makes them the best-behaved case for fast, stable eigenvalue algorithms.

That last point is the destination worth seeing early: the reality and diagonalisability you meet on a tiny 2x2 example are the same properties that make enormous Hermitian systems tractable in physics and engineering.

Common Mistakes with Hermitian Matrices

Mistake 1: Testing only the transpose and forgetting to conjugate

Where it slips in: Checking whether a complex matrix is Hermitian.

Don't do this: Compare $A$ with $A^{T}$ alone and call it Hermitian if they match.

The correct way: Compare $A$ with the conjugate transpose $A^{H}$, which needs both the transpose and the conjugation. The learner who runs only the symmetric test passes matrices that are not Hermitian and fails ones that are, because the sign of every imaginary part was never flipped.

Mistake 2: Assuming a real diagonal is enough

Where it slips in: A quick glance at the main diagonal.

Don't do this: Declare the matrix Hermitian because the diagonal entries are all real.

The correct way: A real diagonal is necessary but not sufficient. Every off-diagonal pair still has to satisfy $a_{ij} = \overline{a_{ji}}$, as Example 2 showed with a real diagonal that still failed the test.

Mistake 3: Confusing Hermitian with skew-Hermitian

Where it slips in: Working with $A^{H} = -A$ instead of $A^{H} = A$.

Don't do this: Expect real diagonal entries in a skew-Hermitian matrix.

The correct way: Hermitian means $A = A^{H}$ with a real diagonal; skew-Hermitian means $A^{H} = -A$ with a purely imaginary or zero diagonal. The second-guesser who swaps the sign condition ends up with the diagonal rule backwards.

Conclusion

• A Hermitian matrix is a square complex matrix with $A = A^{H}$, equal to its conjugate transpose.

• The diagonal entries are always real, and off-diagonal entries are conjugate mirror images.

• The eigenvalues are always real, which is why observables in quantum mechanics are Hermitian.

• A Hermitian matrix generalises a real symmetric matrix to complex entries; the sign-flipped version is skew-Hermitian.

• The common slips are skipping the conjugation, trusting a real diagonal alone, and mixing up the skew-Hermitian sign condition.

To go deeper into matrices with a teacher, explore Bhanzu's algebra tutor, work with a high school math tutor, or look into ongoing math tutoring.

Practice These to Solidify Your Understanding

1. Decide whether $A = \begin{bmatrix} 5 & 4 - i \ 4 + i & 2 \end{bmatrix}$ is Hermitian, and justify it in one line.

2. Find $z$ so that $\begin{bmatrix} 6 & z \ 1 + 3i & -2 \end{bmatrix}$ is Hermitian.

3. Show that the eigenvalues of $\begin{bmatrix} 0 & i \ -i & 0 \end{bmatrix}$ are real by solving the characteristic equation.

If Problem 1 comes out Hermitian without you conjugating the off-diagonal, return to Mistake 1. Want a live Bhanzu trainer to work through Hermitian and skew-Hermitian matrices with you? Book a free demo class.

Read More

• Skew symmetric matrix

• Unitary matrix

• Cofactor matrix

• Properties of matrices

• Adjoint of a matrix