Homogeneous with constant coefficients — characteristic equation: Visual walkthrough

We follow one concrete equation the whole way: $a y^{''} + b y^{'} + c y = 0, a \neq = 0.$

Let me name every symbol before using it, because that is the rule:

Step 1 — Read the equation as a balance, not a formula

WHAT. The equation $a y^{''} + b y^{'} + cy = 0$ says: take the curve's bendiness ( $y^{''}$ ), its slope ( $y^{'}$ ), and its height ( $y$ ); scale each by a number; the three must sum to zero at every point $x$ .

WHY. Before touching algebra we must see what we're hunting: a curve so self-consistent that its own height, slope and bend forever conspire to cancel. Most curves cannot do this — a random wiggle has no reason for its three quantities to stay in balance.

PICTURE. Look at the figure. A candidate curve is drawn in burnt orange. At one point we mark its height (the dot), its slope (the tangent arrow), and its bend (how the tangent is turning). The little scale-bars show $a \cdot y^{''}$ , $b \cdot y^{'}$ , $c \cdot y$ stacked — and they must add to the flat zero line.

Figure — Homogeneous with constant coefficients — characteristic equation

Step 2 — Only one family self-reproduces: the exponential

WHAT. We guess the curve is $y = e^{r x}$ . Here $e \approx 2.718$ is a fixed special number, and $r$ is an unknown number (possibly negative, possibly imaginary — we'll allow all of it later).

WHY THIS TOOL AND NOT ANOTHER? The balance in Step 1 needs $y^{''}$ , $y^{'}$ and $y$ to be the same shape so they can cancel. Ask: which curve has a derivative that is just a scaled copy of itself?

A parabola? Its derivative is a line — different shape. Rejected.
$sin x$ ? Its derivative is $cos x$ — a shifted shape. (Almost; it comes back around after two steps. Hold that thought — it re-enters in Step 8.)
$e^{r x}$ ? Its derivative is $r \cdot e^{r x}$ — exactly itself, only rescaled.

That last property is unique to the exponential, and it is precisely what lets three copies cancel.

$\frac{d}{d x} e^{r x} = scale factor r \cdot same shape back e^{r x} .$

PICTURE. The figure overlays $e^{r x}$ and its derivative $r e^{r x}$ (dashed). They are the same curve vertically stretched by $r$ — never a new shape. That "shape-invariance" is the whole reason we bet on it.

Step 3 — Differentiate the guess, twice

WHAT. Apply the shape-invariance rule from Step 2 repeatedly: $y = e^{r x}, y^{'} = r e^{r x}, y^{''} = r^{2} e^{r x} .$

WHY. We need $y^{'}$ and $y^{''}$ in the same language ( $e^{r x}$ ) so we can substitute them into the balance. Each differentiation just multiplies by another $r$ — that is the beauty of the choice.

Term by term:

$y = 1 \cdot e^{r x}$ — height.
$y^{'} = r \cdot e^{r x}$ — one factor of $r$ (one derivative).
$y^{''} = r^{2} \cdot e^{r x}$ — two factors of $r$ (two derivatives).

PICTURE. A little "ladder" figure: each rung is one derivative, and climbing a rung pins on one more $r$ . The exponential $e^{r x}$ rides along untouched at every rung.

Step 4 — Substitute and factor: calculus becomes algebra

WHAT. Put the three expressions into $a y^{''} + b y^{'} + cy = 0$ : $a r^{2} e^{r x} + b r e^{r x} + c e^{r x} = 0.$ Every term carries a common $e^{r x}$ . Pull it out front: $e^{r x} (a r^{2} + b r + c) = 0.$

WHY. Factoring exposes the structure. The messy calculus has collapsed into a product: an exponential times a plain polynomial in $r$ .

Annotating the factored form: $shape (never zero) e^{r x} \times a number — this is what we control via r (a r^{2} + b r + c) = 0.$

PICTURE. The figure shows the two factors as two blocks. The $e^{r x}$ block is shaded and stamped " $> 0$ always". The only way the product is zero is if the other block is zero.

Step 5 — Divide by $e^{r x}$ : the characteristic equation is born

WHAT. Because $e^{r x}$ is never zero (an exponential is always positive), we may divide it away. What survives is: $a r^{2} + b r + c = 0$ the characteristic equation (also called the auxiliary equation).

WHY. This is the payoff of the whole trick. We started with a differential equation about slopes and bends, and we have traded it for a plain quadratic — the same kind you factor in school. Solve for $r$ , and each $r$ hands you a solution $e^{r x}$ .

To solve it we use the quadratic formula — the tool that answers "for which $r$ does a parabola $a r^{2} + b r + c$ hit zero?": $r = \frac{- b \pm call this Δ b ^{2} - 4 a c}{2 a} .$

$- b$ centres the roots, $2 a$ scales them.
$\pm Δ$ is the spread of the two roots away from the centre.
$Δ = b^{2} - 4 a c$ is the discriminant — the number under the root that decides everything.

PICTURE. The figure plots the parabola $p (r) = a r^{2} + b r + c$ . Its crossings of the horizontal axis are the roots. The teal arrows show the two crossings; the plum bracket shows $Δ$ as their half-spread.

Step 6 — Case A: $Δ > 0$ , the parabola cuts the axis twice

WHAT. If $Δ > 0$ then $Δ$ is a real positive number, so we get two different real roots $r_{1} \neq = r_{2}$ . Two roots ⇒ two independent building-block solutions: $y = c_{1} e^{r_{1} x} + c_{2} e^{r_{2} x} .$

WHY the two constants $c_{1}, c_{2}$ ? A second-order equation loses two derivatives, so it needs two free knobs to fit any starting height and starting slope. (Independence is exactly what the Wronskian certifies.)

Each piece:

$e^{r_{1} x}$ , $e^{r_{2} x}$ — two genuinely different exponential shapes.
$c_{1}, c_{2}$ — free dials set later by initial conditions.

PICTURE. Parabola dips below the axis, crossing at two points $r_{1}, r_{2}$ . Beside it, the two resulting exponential curves (one steeper, one shallower).

Step 7 — Case B (degenerate): $Δ = 0$ , the parabola just kisses the axis

WHAT. If $Δ = 0$ then $Δ = 0$ , and both roots collapse to a single $r = - \frac{b}{2 a}$ . Now $e^{r x}$ is only one solution — but a second-order equation still demands two. The rescue: $y = (c_{1} + c_{2} x) e^{r x} .$

WHY the extra $x$ ? This is the subtle one, so draw it. The operator factors as $a (D - r)^{2} y = 0$ where $D = \frac{d}{d x}$ . Solving the inner $(D - r) u = 0$ gives $u = c_{2} e^{r x}$ ; solving $(D - r) y = u$ is a first-order ODE whose integrating factor cancels the exponential and leaves a bare constant to integrate, which manufactures the $x$ . (Full mechanism in Second Order Linear ODE — Reduction of Order.) So the $x$ is not a magic rule — it is the leftover of an integration.

Annotating:

$(c_{1}) e^{r x}$ — the ordinary exponential solution.
$(c_{2} x) e^{r x}$ — the reduction-of-order companion; the $x$ is the missing second dial.

PICTURE. The parabola is tangent to the axis at exactly one point (a "kiss"). Beside it, the two solution curves: $e^{r x}$ and $x e^{r x}$ — the second one starts at zero, rises, then the exponential takes over.

Step 8 — Case C: $Δ < 0$ , the parabola never touches the axis

WHAT. If $Δ < 0$ then $Δ$ is the square root of a negative number — no real $r$ works. We invent $i$ with $i^{2} = - 1$ , giving a complex conjugate pair: $r = α \pm i β, α = - \frac{b}{2 a}, β = \frac{4 a c - b ^{2}}{2 a} .$

$α$ — the real part: sets growth ( $α > 0$ ) or decay ( $α < 0$ ) or neither ( $α = 0$ ).
$β$ — the imaginary part: sets how fast it oscillates.

WHY sin and cos appear. The raw solutions $e^{(α \pm i β) x}$ are complex, but a real-world ODE wants real answers. Euler's formula converts: $e^{i β x} = cos β x + i sin β x .$ This is why the shifted-shape curve we rejected in Step 2 sneaks back in: an oscillation is what an exponential with an imaginary exponent is. Taking real and imaginary parts gives two real, independent solutions, so: $y = e^{α x} (c_{1} cos β x + c_{2} sin β x) .$

Term by term:

$e^{α x}$ — the envelope: shrinks the wave if $α < 0$ , grows it if $α > 0$ , flat if $α = 0$ .
$cos β x, sin β x$ — the wiggle at frequency $β$ .

PICTURE. The parabola floats entirely above the axis (no crossing). Beside it, an oscillation $cos β x$ hugged by a shrinking envelope $e^{α x}$ (a decaying wave) — the picture of a Damped Harmonic Oscillator.

Step 9 — All three cases are one parabola in disguise

WHAT. The only thing that changed across Steps 6–8 was where the parabola sits relative to the axis:

$Δ$	Parabola vs axis	Roots	Solution shape
$> 0$	cuts twice	two real	$c_{1} e^{r_{1} x} + c_{2} e^{r_{2} x}$
$= 0$	kisses once	one (double)	$(c_{1} + c_{2} x) e^{r x}$
$< 0$	misses	$α \pm i β$	$e^{α x} (c_{1} cos β x + c_{2} sin β x)$

WHY. Understanding the one picture — a parabola sliding up and down past the $r$ -axis — replaces memorising three unrelated rules. The height of the parabola is the discriminant.

PICTURE. Three copies of the same parabola shifted vertically, each with its solution sketch beneath. This is the flagship compression.

The one-picture summary

Everything collapses to: guess $e^{r x}$ → get quadratic → look where the parabola meets the axis.

Recall Feynman retelling of the whole walkthrough

We wanted a curve whose height, slope, and bend always cancel in fixed proportion. Only one curve in the world copies itself when you differentiate it — the exponential $e^{r x}$ — so we bet on it. Differentiating just sticks on extra $r$ 's, so the whole differential equation shrinks into a plain quadratic $a r^{2} + b r + c = 0$ . That quadratic is a parabola, and the whole story is where it meets the axis: cross twice → two exponentials; barely kiss → one exponential plus an $x$ -nudge partner; miss entirely → the roots go imaginary and the curve wiggles as sine and cosine, wrapped in a growing or fading envelope. Three answers, one parabola.

Connections

Homogeneous with constant coefficients — characteristic equation (parent)
Linear Differential Operators and Superposition
Euler's Formula and Complex Exponentials
Wronskian and Linear Independence of Solutions
Second Order Linear ODE — Reduction of Order
Damped Harmonic Oscillator
Method of Undetermined Coefficients