Foundations — Non-homogeneous — method of undetermined coefficients (annihilator method)

Before you can run that method, you must be fluent with about a dozen symbols and ideas the parent note tosses around freely. Below, each one gets its plain-words meaning, its picture, and the reason the topic needs it — ordered so every item leans only on the ones above it.

1. The unknown function $y$ and its slope $\frac{dy}{dx}$

Picture. Draw the curve $y(x)$. At any point, lay a straight ruler flush against the curve (the tangent line). The derivative $y^{\prime }$ at that point is the ruler's slope: rise over run.

Why the topic needs it. Every differential equation is a sentence about slopes. You cannot even read $y^{\prime \prime }-3y^{\prime }+2y=g$ until "the slope of $y$" and "the slope of the slope" are concrete pictures, not squiggles.

2. Second derivative $y^{\prime \prime }$ — the slope of the slope

Picture. If $y^{\prime }$ is how fast you are climbing, $y^{\prime \prime }$ is whether that climb is speeding up (curve bends upward, like a bowl) or slowing down (bends downward, like a dome).

Why the topic needs it. The operators in the parent, like $D^2-3D+2$, contain a $D^2$. That $D^2$ is the instruction "take $y^{\prime \prime }$." Without it, an equation like $y^{\prime \prime }+y=\cos 2x$ is meaningless marks on paper.

3. The letter $D$ — turning "differentiate" into a number-like object

Picture. Think of $D$ as a box on a conveyor belt. A function rolls in on the left; its derivative rolls out on the right. Bolt several boxes in a row and you have $D^2$, or $D-3$, etc.

See Linear differential operators and the D-operator for the full algebra of these boxes.

4. A "linear operator $L$ with constant coefficients"

Linear means two fair-play rules hold: $L[y_1+y_2]=L[y_1]+L[y_2],L[cy]=cL[y].$ In words: $L$ respects addition and scaling — feed it a sum, get the sum of the outputs.

Why the topic needs it. Linearity is the entire licence for splitting the answer into $y_h+y_p$ and for handling sums $g_1+g_2$ piece by piece. It is formalised in Superposition principle for linear ODEs. "Constant coefficients" is what lets us factor $L$ and read off roots at all.

5. The equation $L[y]=g(x)$: homogeneous vs. non-homogeneous

Picture. Picture $L$ as a music amplifier. Homogeneous ($g=0$) is the amp humming on its own — its natural tones. Non-homogeneous is you singing $g(x)$ into it and asking what input $y$ produces exactly that sound out.

Why the topic needs it. The whole method is a bridge from the homogeneous problem (which we can solve) to the non-homogeneous one (which we want). See Homogeneous linear ODEs with constant coefficients.

6. Roots and the characteristic equation

Picture. Each root is a "seed"; the solution it grows is $e^{\alpha x}$ (a growing/decaying ramp) or a $\cos /\sin$ wave. Repeated seeds sprout extra $x$-multiplied copies. See Characteristic equation and repeated roots.

Why the topic needs it. The annihilator table is this dictionary read backwards: you look at $g(x)$, ask "which root would grow this?", and that root's factor $(D-\alpha )$ is the annihilator.

7. Repeated roots and the mysterious extra $x$

Picture. Imagine two overlapping seeds fighting for the same spot. One grows the plain ramp $e^{\alpha x}$; the second is forced to lean sideways, stretching with an extra factor of $x$.

Why the topic needs it. When $g$'s root coincides with a root of $L$, the annihilated equation $AL[y]=0$ has a higher-multiplicity root, which is exactly what produces the $x{e}^{2x}$ trial in Worked Example 2 of the parent.

8. Splitting the answer: $y=y_h+y_p$

Why the topic needs it. This is the skeleton of every worked example: find $y_h$, hunt $y_p$, add them. An alternative hunt for $y_p$ is Variation of parameters; the annihilator method is the shortcut when $g$ is "nice."

9. What "nice $g$" means

Picture. These functions are the "citizens" of the root-dictionary world. Anything born from a root can be killed by re-applying that root's factor. A function like $\tan x$ or $\ln x$ is an outsider — no polynomial in $D$ ever kills it, so the method refuses it (use Variation of parameters there).

Prerequisite map

Equipment checklist

Test yourself — reveal each only after you have an answer.