Lab 13. Symmetric and Hermitian Matrices: Independent Study

This lab accompanies Chapter 13: Symmetric and Hermitian Matrices.

The goal is to connect symmetric/Hermitian structure with computation, geometry, energy, and data:

  1. symmetric and skew-symmetric decompositions;
  2. Hermitian matrices and conjugate transpose;
  3. spectral decomposition using orthonormal eigenvectors;
  4. quadratic and Hermitian forms;
  5. positive definite and positive semidefinite tests;
  6. Gram matrices and \(A^*A\);
  7. Cholesky factorization.

This is an independent-study lab. Each main question includes a worked solution and a similar practice question.

Python practice notebook

You may use the Jupyter notebook version for longer Python practice:

Interactive lab

Study guide and worked questions

Question 1. Symmetric and skew-symmetric decomposition

Let

\[ M=\begin{bmatrix}1&4\\2&3\end{bmatrix}. \]

Find the symmetric part and the skew-symmetric part of \(M\).

Solution

Use

\[ S=\frac{M+M^T}{2}, \qquad K=\frac{M-M^T}{2}. \]

Then

\[ S=\begin{bmatrix}1&3\\3&3\end{bmatrix}, \qquad K=\begin{bmatrix}0&1\\-1&0\end{bmatrix}. \]

Thus

\[ M=S+K. \]

Similar practice

For

\[ M=\begin{bmatrix}2&-1\\5&0\end{bmatrix}, \]

find \(S\) and \(K\).

Answer

\[ S=\begin{bmatrix}2&2\\2&0\end{bmatrix}, \qquad K=\begin{bmatrix}0&-3\\3&0\end{bmatrix}. \]

Question 2. Hermitian check

Determine whether

\[ H=\begin{bmatrix}2&1+i\\1-i&4\end{bmatrix} \]

is Hermitian.

Solution

The diagonal entries are real. The off-diagonal entries are complex conjugates:

\[ \overline{1+i}=1-i. \]

Therefore

\[ H^*=H. \]

So \(H\) is Hermitian.

Similar practice

Determine whether

\[ B=\begin{bmatrix}1&2+i\\2+i&3\end{bmatrix} \]

is Hermitian.

Answer

No. The lower-left entry should be \(2-i\), not \(2+i\).

Question 3. Spectral decomposition

Let

\[ A=\begin{bmatrix}3&1\\1&3\end{bmatrix}. \]

Find an orthogonal diagonalization.

Solution

The eigenvalues are

\[ \lambda_1=4, \qquad \lambda_2=2. \]

Corresponding unit eigenvectors are

\[ \vec{u}_1=\frac{1}{\sqrt2}\begin{bmatrix}1\\1\end{bmatrix}, \qquad \vec{u}_2=\frac{1}{\sqrt2}\begin{bmatrix}1\\-1\end{bmatrix}. \]

Thus

\[ Q=\frac{1}{\sqrt2}\begin{bmatrix}1&1\\1&-1\end{bmatrix}, \qquad D=\begin{bmatrix}4&0\\0&2\end{bmatrix}, \]

and

\[ A=QDQ^T. \]

Similar practice

Find an orthogonal diagonalization of

\[ A=\begin{bmatrix}5&2\\2&5\end{bmatrix}. \]

Answer

Eigenvalues are \(7\) and \(3\), with the same orthonormal eigenvectors

\[ \frac{1}{\sqrt2}(1,1)^T, \qquad \frac{1}{\sqrt2}(1,-1)^T. \]

Question 4. Positive definite test by eigenvalues

Classify

\[ A=\begin{bmatrix}3&1\\1&3\end{bmatrix}. \]

Solution

The eigenvalues are \(4\) and \(2\), both positive. Therefore \(A\) is positive definite.

Similar practice

Classify

\[ B=\begin{bmatrix}1&2\\2&1\end{bmatrix}. \]

Answer

The eigenvalues are \(3\) and \(-1\). Since one eigenvalue is negative, \(B\) is indefinite.

Question 5. Sylvester’s criterion

Use Sylvester’s criterion to test whether

\[ A=\begin{bmatrix} 2&1&0\\ 1&3&1\\ 0&1&2 \end{bmatrix} \]

is positive definite.

Solution

The leading principal minors are

\[ \Delta_1=2>0, \]

\[ \Delta_2=\det\begin{bmatrix}2&1\\1&3\end{bmatrix}=5>0, \]

and

\[ \Delta_3=\det(A)=8>0. \]

All leading principal minors are positive. Therefore \(A\) is positive definite.

Similar practice

Test

\[ B=\begin{bmatrix}1&2\\2&1\end{bmatrix}. \]

Answer

The leading principal minors are \(1\) and \(-3\). Since the second is negative, \(B\) is not positive definite.

Question 6. Why \(A^TA\) is positive semidefinite

Let \(A\in\mathbb{R}^{m\times n}\). Prove that \(A^TA\) is positive semidefinite.

Solution

For any \(\vec{x}\in\mathbb{R}^n\),

\[ \vec{x}^{\,T}A^TA\vec{x}=(A\vec{x})^T(A\vec{x})=\|A\vec{x}\|^2\ge0. \]

Therefore \(A^TA\) is positive semidefinite.

Similar practice

When is \(A^TA\) positive definite?

Answer

It is positive definite when \(A\) has full column rank, because then \(A\vec{x}\ne0\) for every nonzero \(\vec{x}\).

Notebook submission suggestions

In your notebook, include:

  1. at least one symbolic computation by hand;
  2. at least one numerical computation in Python;
  3. at least one explanation of the geometry or energy interpretation;
  4. one example you create yourself;
  5. one AI companion reflection, checked by your own computation.