Login

I keep hitting a wall trying to understand why the method of undetermined coefficients for solving non-homogeneous linear ODEs fails when the forcing function is part of the complementary solution. My textbook just states the rule about multiplying by x, but I can't intuitively grasp why the standard particular solution form becomes linearly dependent.

I remember hitting that wall too. I’d plug in a guess like P(x) times e^{rx} or a polynomial, and if that forcing term matched something in the homogeneous solutions, the left side just collapsed to zero. It felt like the guess was already in the solution space, so you can’t pin down coefficients.

The intuitive picture is that the operator acts like a filter that already wants to reproduce the patterns in the homogeneous solutions. When the forcing term lies in that same space, a naive guess is in the kernel too, so substituting just paints the same picture and you can’t solve for the coefficients. Multiplying by x nudges the guess into a direction the operator doesn’t annihilate, so you do get a resolvable equation. This rule is a practical guideline for the method of undetermined coefficients: if resonance happens, bump the guess up by x until you regain independence.

I once tried a concrete example: y'' - 3y' + 2y = e^x. The homogeneous equation has solutions e^x and e^{2x}. The right-hand side is e^x, which is in the homogeneous space. My first guess y_p = A e^x made the left side 0, so there was no A to solve for. Then I tried y_p = A x e^x, and that did give a value for A. It was a lightbulb moment, even though it felt a bit hacky.

Maybe the issue isn’t just resonance. Sometimes we overfit the forcing term to the homogeneous space and a tiny change in the forcing or the initial data breaks the problem loose. I’ve also seen people switch to variation of parameters for stubborn cases, which can feel slower but avoids chasing the right guess.

Login
Username:
Password:	Lost Password?
	Remember me