About Matrix Machine
A 2×2 matrix is a machine that moves every point of the flat plane. It keeps the origin where it is, keeps straight lines straight and keeps evenly spaced grid lines evenly spaced: the squares of the grid become a field of matching parallelograms. Here the plain squared paper on the light table is the "before", and the acetate sheet over it, drawn in chinagraph pencil, is the "after".
The trick is that the whole move is fixed by where two arrows land. The red arrow î starts as (1, 0) and lands on the matrix's first column; the blue arrow ĵ starts as (0, 1) and lands on the second column. Every other point (x, y) goes to x lots of the red arrow plus y lots of the blue. So you can type the four numbers, drag the two arrow tips, or press a preset (each one plays the move from the start).
The readouts: the determinant is what the move does to areas. The unit square becomes the shaded parallelogram, and its area is the size of the determinant; the determinant is negative, and the shading pink, when the plane has been flipped over (the F reads backwards). A determinant of 0 squashes everything onto a line (or, for the all-zero matrix, a single point), and that can't be undone, so there's no inverse. The eigenvectors are directions that stay on their own line, only stretched by a factor λ (the eigenvalue); they're drawn as green dashed lines. A rotation (other than a half turn or a whole turn) turns every direction, so it has none that are real. The trace (top-left plus bottom-right) is the two eigenvalues added up, and the determinant is them multiplied. INVERT swaps the matrix for its inverse, the move that puts everything back; press it again to come back.
COMBINE TWO does one matrix and then another. Doing A first and then B is the single matrix B·A (it's written right to left, like f(g(x))). Doing them the other way round usually gives a different picture: matrix multiplication cares about order.
THE STORY
This app grew from the Python script lib.py, a small linear algebra library with a Vector class and a Matrix class that can multiply a matrix by a vector, which is exactly what each point here goes through.
Arrays of numbers for solving sets of equations are very old: the Chinese text The Nine Chapters on the Mathematical Art (put together by around the 1st century AD) lays coefficients out in a grid on a counting board and eliminates them column by column, much as we do now. Determinants came long before matrices: Seki Takakazu in Japan (1683) and Gottfried Leibniz (1693) both worked with them, and the word "determinant" in its modern sense comes from Augustin-Louis Cauchy in 1812.
James Joseph Sylvester coined the word "matrix" in 1850 (Latin for womb: he saw it as the thing determinants are born from). Arthur Cayley's A Memoir on the Theory of Matrices of 1858 treated matrices as things in their own right that you can add, multiply and invert, noticed that the order of a product matters, and stated the Cayley–Hamilton theorem (he checked the 2×2 and 3×3 cases). The word "eigenvalue" is half German: David Hilbert was the first to use eigen ("own") for these ideas, in 1904.
HOW IT WORKS
For the matrix with rows (a, b) and (c, d), the point (x, y) goes to (ax + by, cx + dy). The determinant is ad − bc and the trace is a + d. The eigenvalues are the solutions of λ² − (trace)λ + determinant = 0; when (trace)² is less than 4 × determinant there are no real ones. The inverse is the matrix with rows (d, −b) and (−c, a), divided by the determinant. B·A is worked out row by column, and doing it the other way round gives A·B.
When a preset is pressed the picture slides from the start to the new matrix in under a second (rotations go round by the angle rather than sliding through a shrunken middle), then everything stops. Nothing runs by itself.
FURTHER WATCHING AND READING