Engineering Math Series 14: Diagonalization and Dynamic Systems

The real reason to learn eigenvalues and eigenvectors is diagonalization. When diagonalization is possible, you can split a complex matrix problem into independent one-dimensional problems.

What Is Diagonalization

If matrix $A$ has a sufficient number of linearly independent eigenvectors, it can be written as

$$A = PDP^{-1}$$

where

- $P$ is the matrix with eigenvectors as column vectors

- $D$ is the diagonal matrix with eigenvalues on the diagonal

Diagonal matrices make exponentiation, matrix exponentials, and repeated application very easy. This is why they are powerful for dynamical system analysis.

Why Is It Important for Dynamic Systems

The system of differential equations

$$\mathbf{x}' = A\mathbf{x}$$

has a solution of the form

$$\mathbf{x}(t) = e^{At}\mathbf{x}(0)$$

If $A$ is diagonalizable, then

$$e^{At} = Pe^{Dt}P^{-1}$$

and $e^{Dt}$ simply applies the exponential function to each diagonal element.

This means you can directly read how fast the system grows or shrinks in each eigendirection.

Worked Example

Consider the matrix

$$

A =

\begin{pmatrix}

0 & 1 \\

-2 & -3

\end{pmatrix}

$$

The characteristic equation is

$$

\det

\begin{pmatrix}

-\lambda & 1 \\

-2 & -3-\lambda

\end{pmatrix}

= \lambda^2 + 3\lambda + 2 = 0

$$

so

$$\lambda = -1, \quad -2$$

Since the eigenvalues are distinct, diagonalization is possible. This fact alone tells us the system has two decaying modes.

The solution is approximately

$$\mathbf{x}(t) = c_1 e^{-t}\mathbf{v}_1 + c_2 e^{-2t}\mathbf{v}_2$$

As time passes, the $e^{-2t}$ term vanishes faster, and eventually the slower-decaying $e^{-t}$ mode dominates the long-term behavior.

This is exactly how you read the "dominant mode" in dynamics.

Engineering Applications

Stability in Control Engineering

If the real parts of all eigenvalues of the state matrix are negative, the linear system converges in a stable direction.

Networks and Diffusion

The spectrum of a matrix representing connection structure is connected to information propagation speed, diffusion modes, and convergence characteristics.

Mechanical and Structural Systems

Vibration systems with multiple degrees of freedom are fundamentally analyzed by decomposing into modes.

When Diagonalization Is Not Possible

Not every matrix is diagonalizable. But at the introductory level, understanding why "the diagonalizable case" is beneficial is more important first. Later, Jordan normal form or numerical decompositions handle more general situations.

Common Mistakes

Only memorizing the diagonalization formula

$A=PDP^{-1}$ is not just computational notation but the process of changing coordinates to the eigenvector directions.

Looking only at eigenvalues and ignoring initial conditions

How much each mode actually appears is determined by the initial conditions.

Not distinguishing fast-decaying and slow-decaying modes

In dynamic systems, practicing reading which mode dominates long-term behavior is very important.

One-Line Summary

Diagonalization is the core tool that decomposes coupled systems into independent modes so you can read dynamic behavior.

Next Post Preview

Starting with the next batch, we will move to **vectors and vector calculus**, dealing with change and flow in space.

References

- Erwin Kreyszig, _Advanced Engineering Mathematics_, 10th Edition

- Gene H. Golub, Charles F. Van Loan, _Matrix Computations_

- Hassan K. Khalil, _Nonlinear Systems_