Matrices can rotate, reflect, shear, stretch, and squash. Amid all that motion, some directions are special: the matrix stretches or flips vectors on those lines but does not knock them onto a different line. Those directions are spanned by eigenvectors; the stretch factors are eigenvalues.

The defining equation

A nonzero vector $\mathbf{v}$ is an eigenvector of square matrix $A$ with eigenvalue $\lambda$ when $A\mathbf{v}=\lambda\mathbf{v}$. Applying $A$ has the same effect as simply multiplying by the scalar $\lambda$. Rearranged, $(A-\lambda I)\mathbf{v}=\mathbf{0}$, so $\lambda$ is an eigenvalue precisely when $A-\lambda I$ fails to be invertible — that is, when $\det(A-\lambda I)=0$.

Example Stretch the plane by $2$ in the $x$-direction and $3$ in the $y$-direction: $A=\begin{pmatrix}2&0\\0&3\end{pmatrix}$. Then $\mathbf{e}_1=(1,0)$ is an eigenvector with $\lambda=2$, and $\mathbf{e}_2=(0,1)$ with $\lambda=3$. Most other vectors get both stretched and realigned relative to the axes — their direction changes, so they are not eigenvectors.

Why anyone cares

  • Diagonalization. If you have a basis of eigenvectors, $A$ becomes a diagonal matrix in those coordinates — powers and exponentials of $A$ become easy.
  • Dynamics. For discrete systems $\mathbf{x}_{n+1}=A\mathbf{x}_n$, eigenvalues decide growth, decay, and oscillation.
  • Stability. In continuous linear systems $\mathbf{x}'=A\mathbf{x}$, signs of real parts of eigenvalues govern whether solutions blow up or settle.
Pitfall. Eigenvectors are defined only up to nonzero scaling. Also, real matrices can have complex eigenvalues (think rotations): there may be no real eigenvector even though complex ones exist.

Build the surrounding vocabulary in Linear algebra and follow the first-course pathway.

Citations & further reading