4.6/5 - (35 votes)

Take a picture of the Mona Lisa and slide the top of the image sideways while the bottom stays fixed. Almost every point in the painting shifts to a new spot, except for the ones sitting on a single horizontal line running straight through her. Those unmoved points are not a coincidence. They are a textbook example of one of the most useful ideas in all of mathematics.

A painting that refuses to move

The sideways slide described above is called a shear transformation, one of the simplest ways mathematicians study how space bends and stretches. Most points in the image end up somewhere completely new after the shift.

But the points sitting exactly on that horizontal line do not rotate at all. They only get pushed left or right along the same direction they already pointed in. That surviving direction is what mathematicians call an eigenvector.

The stretch that got its own name

An eigenvector is a direction that a given transformation only stretches or shrinks, never turns. The exact amount of stretching is called the eigenvalue.

Mathematicians write the relationship as Av = λv, but in plain language it simply means this: apply the transformation A to the vector v, and you get the same direction back, only longer, shorter, or flipped, by a factor of λ.

In the Mona Lisa example, the points along that horizontal line barely move at all, which means their eigenvalue is exactly 1. No stretching, no shrinking, no flip, just a direction the transformation leaves completely alone.

A German word that simply means own

The idea did not appear all at once. In the 18th century, Leonhard Euler studied how rigid bodies rotate and noticed that certain axes behaved differently from every other direction in the object.

See also  Adding more equations to a math problem can make it impossible to solve

Joseph-Louis Lagrange later realized that those special axes Euler had found were exactly the eigenvectors of an object’s inertia matrix. Augustin-Louis Cauchy pushed the idea further, extending it to any number of dimensions and using it to classify quadric surfaces, coining the term characteristic root along the way.

It took until 1904 for the concept to get the name still used today. German mathematician David Hilbert, working on integral operators, borrowed the German word eigen, meaning own or characteristic, and eigenvalues and eigenvectors have carried that name ever since.

Why a real number matrix can still hide imaginary results

Two common assumptions trip people up almost immediately. The first is that the eigenvectors of a matrix must be perpendicular to each other. That is only true in special cases, such as symmetric matrices. In general, eigenvectors do not need to line up at right angles at all.

The second assumption is even stranger to accept. A matrix built entirely out of ordinary real numbers, no fractions, no imaginary components anywhere, can still produce eigenvalues that include an imaginary part. When that happens, those complex eigenvalues always arrive in matching conjugate pairs.

One guarantee does hold, though. Any real matrix with an odd number of rows and columns always has at least one fully real eigenvalue. That fact falls directly out of the intermediate value theorem applied to the matrix’s characteristic polynomial, a degree-n equation whose roots are precisely its eigenvalues.

From geology to the quantum world

Once the concept had a name, it spread far beyond rotating rigid bodies. Charles Hermite extended it to a new class of matrices in 1855, and by the early 20th century engineers had working numerical methods to actually calculate eigenvalues by hand, such as the power method published by Richard von Mises in 1929.

See also  The one straight line equation that breaks the formula you learned in school

Today the same math shows up in fields that seem to have nothing to do with each other on the surface, from reading fault lines in geology to describing the behavior of particles in quantum mechanics. A shear on a five-hundred-year-old portrait and the equations governing subatomic particles turn out to be speaking the exact same mathematical language.