Two solutions are not always two solutions
In the previous guide you met the central fact of this rung: the solutions of a homogeneous second-order equation a y'' + b y' + c y = 0 form a two-dimensional space. Two-dimensional means you need exactly two basis solutions, y1 and y2, and then every solution is a combination C1 y1 + C2 y2 — that is the superposition principle doing its work. But there is a catch hiding in the word *two*. If y2 is secretly just a multiple of y1, then C1 y1 + C2 y2 collapses to a single direction, and you have not spanned the plane at all — you have only covered a line through it.
Here is a concrete picture. For y'' - y = 0, both y1 = e^x and y2 = 2 e^x are honest solutions. But y2 = 2 y1 — same shape, just scaled. The combination C1 e^x + C2 (2 e^x) equals (C1 + 2 C2) e^x, which is still just *some* multiple of e^x. You can never build e^(-x) out of it, even though e^(-x) is also a solution. Two solutions that are scalar multiples of each other carry the information of only one. We call such a pair linearly dependent, and a pair that genuinely points in different directions linearly independent.
A determinant that watches the slopes
Checking by eye whether one function is a multiple of another works for e^x and 2 e^x, but fails the moment the functions are tangled — is sin(x) a multiple of cos(x)? Is e^x a multiple of x e^x? We want a mechanical test. The key insight: if y2 = k y1 for a constant k, then their slopes obey the *same* relation, y2' = k y1'. So at every point the value-and-slope pair (y2, y2') is just k times the pair (y1, y1') — the two pairs are parallel vectors. Two vectors in the plane are parallel exactly when the determinant they form is zero. That determinant is the Wronskian.
| y1 y2 |
W(x) = | | = y1 * y2' - y2 * y1'
| y1' y2' |
W = 0 at a point <--> (y1,y1') and (y2,y2') are parallel thereTry it on the dependent pair: W(e^x, 2 e^x) = e^x (2 e^x) - 2 e^x (e^x) = 0 everywhere — the test correctly flags them as dependent. Now the independent pair e^x and e^(-x): W = e^x (-e^(-x)) - e^(-x) (e^x) = -1 - 1 = -2, never zero. The Wronskian has done in one line what eyeballing could not promise: it has measured whether the two solutions truly open up a two-dimensional space.
Why the Wronskian actually fixes the constants
The Wronskian is not just a parallelism gadget; it is exactly the quantity that decides whether you can solve an initial value problem. Suppose you want a solution with y(x0) = A and y'(x0) = B. Writing y = C1 y1 + C2 y2 and imposing both conditions at x0 gives two linear equations in the two unknowns C1, C2. The coefficient matrix of that system is precisely the Wronskian matrix evaluated at x0. By the usual rule of linear algebra, you can solve for C1 and C2 for *any* targets A, B exactly when that matrix is invertible — that is, when W(x0) is not zero.
So the non-vanishing of the Wronskian is the green light for the whole theory: it guarantees that C1 y1 + C2 y2 can hit *every* allowed initial condition, which is what it means for those two solutions to deliver the full general solution. When W is non-zero, the pair {y1, y2} is called a fundamental set of solutions — a basis for the two-dimensional solution space — and the dimension-two promise from the last guide is finally cashed in concretely.
Abel's identity: all or nothing
Here is a remarkable fact that saves enormous effort. For solutions of the *same* second-order linear equation, the Wronskian cannot be zero at some points and non-zero at others. It is either zero everywhere (dependent) or non-zero everywhere (independent). This sharp dichotomy comes from Abel's identity, which says that W itself satisfies a tiny first-order equation, W' = -(b/a) W, whose solution is W(x) = W(x0) e^(-(b/a)(x - x0)) for the constant-coefficient case. Since an exponential is never zero, W flips between zero and non-zero only through its starting value W(x0).
Notice what this buys you. To certify a fundamental set, you only ever need to check the Wronskian at one single point — usually the easiest one. Abel's identity then propagates that verdict to the entire interval for free. The same identity has a second life later in this rung: in the reduction-of-order guide, that formula W(x) = W(x0) e^(-integral of p) is exactly the lever that pries a *second* solution out of a first one you already know.
Honest fine print: the Wronskian outside its home
The clean dichotomy above is true, but it has a precise hypothesis: y1 and y2 must both be solutions of the *same* linear equation, on an interval where its coefficients are continuous. For arbitrary functions plucked from nowhere, the Wronskian is less powerful. If W is non-zero at even one point, the functions are still genuinely independent — that direction always holds. But the converse fails: there exist independent functions whose Wronskian is identically zero, so W = 0 alone does not prove dependence in general.
The textbook counterexample is x^2 and x|x| on the whole real line: their Wronskian is zero everywhere, yet no single constant k makes x|x| = k x^2 hold on both sides of the origin, so they are independent. The escape hatch is exactly the hypothesis we kept stressing: x^2 and x|x| are not two solutions of one nice linear ODE across x = 0. Inside the protected world of solutions of a fixed linear equation, you are safe — W = 0 there really does mean dependent, and the green-light reading holds without asterisks.