Imagine drawing an arrow on a rubber sheet. Stretch the sheet, and most arrows change both length and direction. A few special arrows stay on the same line. Those are the directions an eigenvector describes.
A direction that survives
A matrix is a rule for moving vectors. An eigenvector is a nonzero vector for which the rule simplifies to multiplication by a single number:
The scalar λ is the eigenvalue. Positive values preserve orientation. Negative values reverse it. Zero collapses the vector to the origin. In every case, the image remains on the line spanned by the original vector.
A concrete example
1 2
Apply this matrix to (1, 1). The result is (3, 3): the same direction, three times as long. Apply it to (1, −1), and you get (1, −1) back. One direction expands; the other does not move.
A(1, −1) = 1(1, −1)
This symmetric matrix transforms the unit circle into an ellipse. The two eigendirections reveal its axes. A seemingly complicated change becomes two independent stretches when described in the right coordinates.
The useful caveat
Not every real matrix has real eigenvectors. A 90° rotation turns every nonzero vector away from its own line. There are no real eigendirections to draw, even though complex eigenvalues still exist.
And eigenvectors are not automatically the axes of every transformed circle. For a general matrix, singular vectors give that geometric description. The clean correspondence here comes from the symmetry of A.
Change the question
Instead of asking “Where did every point go?”, ask “Which directions make this transformation simple?” That shift—from tracking individual movements to finding structure—is what makes eigenvectors useful.
Try the transformation yourself ↗