the characteristic polynomial

For a constant-coefficient linear equation there is an almost magical shortcut: you do not really have to solve a differential equation at all. You solve an ordinary algebra problem — find the roots of a polynomial — and the solutions of the differential equation fall out of those roots. The characteristic polynomial is the bridge that turns calculus into algebra.

It comes from a single guess. For the homogeneous equation a_n y^(n) + a_(n-1) y^(n-1) + ... + a_1 y' + a_0 y = 0 with constant coefficients, try y = e^(rx). Each derivative just multiplies by r, so y^(k) = r^k e^(rx), and substituting turns the whole equation into e^(rx) times a_n r^n + a_(n-1) r^(n-1) + ... + a_1 r + a_0 = 0. Since e^(rx) is never zero, the bracket must vanish: that polynomial in r, of degree n, is the characteristic polynomial, and the equation setting it to zero is the characteristic equation. Its n roots (counted with multiplicity, by the fundamental theorem of algebra) are exactly the exponents that give solutions. Distinct real root r gives e^(rx); a complex pair a +/- bi gives e^(ax) cos(bx) and e^(ax) sin(bx); a repeated root contributes extra factors of x.

This single device solves every constant-coefficient linear ODE, of any order, by reducing it to root-finding. It also explains the deep link to linear algebra: the same polynomial is the characteristic polynomial of the companion matrix of the equation, tying scalar nth-order theory to eigenvalues. The honest limit: this shortcut needs CONSTANT coefficients — for variable coefficients there is no characteristic polynomial, and you turn to series methods or special equations like Cauchy-Euler instead. And finding the roots themselves can be hard: for degree five and up there is no general formula, so high-order cases may need numerical root-finding.