When a matrix acts on a vector, it usually transforms it, changing its direction, its length, or both. Think of a matrix as a function that takes an input vector and produces a new, output vector.

But for any given matrix, there are often special vectors that don't change their direction under the transformation. They might get stretched or shrunk, but their direction remains exactly the same. These special vectors are called eigenvectors.

The amount by which the eigenvector is scaled is called its eigenvalue. If the eigenvalue is 2, the eigenvector is doubled in length. If it's -0.5, it's halved and points in the opposite direction. If it's 1, it's completely unchanged.

This relationship is captured in a simple, powerful equation. If $A$ is a square matrix, $\vec{v}$ is an eigenvector, and $\lambda$ (the Greek letter lambda) is its corresponding eigenvalue, then:

This equation states that the action of the matrix $A$ on the vector $\vec{v}$ is the same as just scaling $\vec{v}$ by the number $\lambda$. For this to be meaningful, we're only interested in non-zero eigenvectors. The zero vector always satisfies this equation, but it doesn't tell us anything useful.

How do we find these special values? We start with the defining equation and rearrange it. To work with both sides as matrices, we can rewrite the right side by multiplying $\lambda$ by the identity matrix, $I$.

Now we can move everything to one side:

And factor out the vector $\vec{v}$:

This equation is the key. We are looking for a non-zero vector $\vec{v}$ that the matrix $(A - \lambda I)$ transforms into the zero vector. This only happens if the matrix $(A - \lambda I)$ squishes space into a lower dimension. A matrix that does this has a determinant of zero.

So, to find the eigenvalues, we must find the values of $\lambda$ that make this determinant zero.

Solving this equation for $\lambda$ will give us all the eigenvalues of the matrix $A$. The expression $\det(A - \lambda I)$ is a polynomial in $\lambda$, known as the characteristic polynomial.

Once we have the eigenvalues, finding the corresponding eigenvectors is straightforward. We take each eigenvalue one by one and plug it back into our rearranged equation:

For a given $\lambda$, this becomes a system of linear equations. The solutions for $\vec{v}$ form the set of eigenvectors for that eigenvalue. Let's walk through an example.

Suppose we have the matrix $A$:

First, we find the eigenvalues using the characteristic equation, $\det(A - \lambda I) = 0$.

Now, we calculate the determinant:

This simplifies to $(3-\lambda)(2-\lambda) = 0$, giving us two eigenvalues: $\lambda_1 = 3$ and $\lambda_2 = 2$.

Next, let's find the eigenvector for each eigenvalue.

For $\lambda_1 = 3$: We solve $(A - 3I)\vec{v} = \vec{0}$.

This matrix equation gives us the linear system $0x + 1y = 0$ and $0x - 1y = 0$. Both equations tell us that $y=0$. The value of $x$ can be anything. So, any vector of the form $\begin{pmatrix} x \\ 0 \end{pmatrix}$ where $x \neq 0$ is an eigenvector. A simple choice is:

For $\lambda_2 = 2$: Now we solve $(A - 2I)\vec{v} = \vec{0}$.

This gives us the equation $x+y=0$, or $x = -y$. Any vector where the x-component is the negative of the y-component is an eigenvector. If we choose $y=1$, then $x=-1$. A simple choice is:

So, we've found the eigenvalues $\lambda = 3$ and $\lambda = 2$, with their corresponding eigenvectors $\begin{pmatrix} 1 \\ 0 \end{pmatrix}$ and $\begin{pmatrix} -1 \\ 1 \end{pmatrix}$.

Eigenvectors represent the axes of a linear transformation. They are the directions that are preserved. Imagine stretching a sheet of rubber. The directions in which the material is only stretched, without any rotation or shear, are the directions of the eigenvectors.

The matrix $A = \begin{pmatrix} 3 & 1 \\ 0 & 2 \end{pmatrix}$ from our example performs a shear and stretch. Vectors along the x-axis, like our first eigenvector $\begin{pmatrix} 1 \\ 0 \end{pmatrix}$, are simply stretched by a factor of 3. Vectors along the line $y=-x$, like our second eigenvector $\begin{pmatrix} -1 \\ 1 \end{pmatrix}$, are stretched by a factor of 2. Any other vector will have its direction changed.

The diagram shows how the eigenvectors $\vec{v}_1$ and $\vec{v}_2$ stay on the same line after being transformed by matrix $A$, while another vector $\vec{u}$ is knocked off its original line. This geometric property is what makes eigenvectors and eigenvalues so fundamental to understanding the behavior of matrices.