the product-solution ansatz
/ ansatz: AHN-sahts /
An 'ansatz' is just a German word for an educated guess about the FORM of a solution — you write down a template with unknowns in it, plug it into the equation, and let the equation tell you whether the guess can work and what the unknowns must be. The product-solution ansatz is the specific guess that drives separation of variables: that a solution can be written as a product of single-variable functions, one for each independent variable.
In two variables it reads u(x,t) = X(x) T(t); in three it might be u(x,y,t) = X(x) Y(y) T(t), and in a disk it becomes u(r, theta) = R(r) Theta(theta). The whole point is that the factors do not mix variables — X depends on x only, T on t only. When you substitute such a product into a linear PDE with constant coefficients and divide through by the product itself, every term collapses into a piece that depends on one variable alone, and the equation cleanly falls apart into one ordinary differential equation per factor. You are not claiming every solution looks like this — most do not. You are claiming that ENOUGH product solutions exist to build the general one by adding them up.
Why allow such a restrictive guess? Because of superposition: the PDE is linear and homogeneous, so any sum of solutions is again a solution. Even though a single product cannot match a lumpy initial condition, an infinite combination of products — each a separated mode — can. The ansatz is the seed; the Fourier series is the harvest. If substituting the product does NOT separate the equation (the variables refuse to come apart), that is the equation telling you the method does not apply here.
For the wave equation u_tt = c^2 u_xx, the ansatz u = X(x)T(t) gives X T'' = c^2 X'' T, so T''/(c^2 T) = X''/X = -lambda. The spatial factor sin(n pi x / L) paired with the time factor cos(c n pi t / L) is a single 'normal mode' — a standing wave. The general vibration is a sum of these modes.
Each separated product is one mode; the real solution is the superposition of all of them.
The ansatz is a hypothesis, not a theorem: substituting it is a TEST. If the variables don't separate (e.g. mixed terms like u_xt, or coefficients that depend on both x and t), the product form fails and you need another method.