**Eigenvector:**
An eigenvector of a matrix $A$ is a vector that *remains in the same direction* (or exactly opposite direction) after a linear [[Matrix Transformations|transformation]] represented by $A$. Mathematically, an eigenvector $x$ satisfies:
$ Ax= \lambda x$
**Eigenvalue:**
Each eigenvector has a corresponding eigenvalue $\lambda$, which represents the *scaling factor* applied to the eigenvector by the transformation $A$. This scaling changes the vector's magnitude but not its direction.
## Derivation
We can find an eigenvector $x$ when the transformation matrix $A$ is applied on the vector $x$ results in the same vector $x$ just scaled by some factor $\lambda$.
$Ax= \lambda x$
where:
- $A$: Transformation matrix with shape $(n \times n)$
- $x$: Vector of shape $(1 \times n)$
- $\lambda$: Scalar value
*Step 1: Setup the Characteristic Equation*
$ \begin{align}
Ax &= \lambda x \tag{1}\\[2pt]
Ax - \lambda x &= 0 \tag{2}\\[2pt]
x*(A- \lambda \mathbf I) &=0 \tag{3}
\end{align} $
(3) We need to add the identity matrix to $\lambda$ as subtracting a scalar from a matrix is not defined. The product of $\lambda \mathbf I$ will have $\lambda$ on all diagonal elements and zero elsewhere.
*Step 2: Solve for lambda*
For the equation to hold, either $x$ or $(A- \lambda \mathbf I)$ has to equal $0$. Since $x=0$ is a trivial solution, where all vectors have zero [[Vector Length|length]] and no direction, we focus on the scenario where $(A-\lambda \mathbf I)=0$.
$ \begin{align}
A-\lambda \mathbf I &=0 \tag{4}\\
\det(A- \lambda \mathbf I)&=0 \tag{5}\\
\end{align} $
(5) A matrix operation only results in $0$ if its [[Matrix Determinant]] is also $0$. This determinant equation is called the characteristic equation, and solving it gives the eigenvalues $\lambda$.
*Step 3: Solve for the eigenvector*
Once $\lambda$ is found substitute it back into the following to find the eigenvector $x$.
$ (A-\lambda \mathbf I) x=0 $
## Example
Computing the determinant for a high dimensional matrix is complicated. However we can show calculations for the case where $A$ has shape $(2 \times 2)$.
$
\begin{align}
\det\left( \begin{bmatrix} a&b\\c&d \end{bmatrix}
\begin{bmatrix} 1&0\\0&1 \end{bmatrix} *\lambda \right)&=0\\[10pt]
\det\left( \begin{bmatrix} a-\lambda &b\\c&d-\lambda \end{bmatrix} \right)&=0\\[12pt]
(a-\lambda)*(d-\lambda)-bc&=0\\[10pt] \lambda^2-(a+d)\lambda+(ad-bc)&=0
\end{align}
$
For a given matrix $A$ we can solve this equation w.r.t. $\lambda$ with the [[Quadratic Formula]]. Finally, we plug the solution for $\lambda$ into equation (3) to derive vector $x$.