Legendre's equation and Legendre polynomials (intro): Visual walkthrough

We build everything. If a symbol appears, it was drawn first.

Step 1 — The equation, read like a sentence

WHAT. $y = y (x)$ is an unknown function we want to find. $y^{'}$ is its slope (how fast it rises), $y^{''}$ its curvature (how fast the slope itself changes). $n$ is a number we choose in advance — think of it as a dial. The whole line says: at every point $x$ , these three pieces must cancel to zero.

WHY read it this way. Notice the coefficient in front of $y^{''}$ is $(1 - x^{2})$ . At $x = 0$ that equals $1$ — perfectly healthy. At $x = \pm 1$ it equals $0$ — the curvature term vanishes, and the equation "loses control" there. Those two spots, $x = \pm 1$ , are the poles of the sphere in the physics. Everything dramatic in this topic happens because of that vanishing.

PICTURE. The curve of $(1 - x^{2})$ over the interval — a dome that touches zero exactly at the two ends.

Figure — Legendre's equation and Legendre polynomials (intro)

Which

x

-values make the

y^{''}

coefficient vanish?

x = \pm 1

, the two poles.

Step 2 — Guess a power series, and say why we're allowed to

WHAT. Each $a_{m}$ is just a number (a knob). Turn the knobs and you get different functions. Our job: find which knob-settings make the equation hold.

WHY a series, and why here. A smooth function near a "healthy" point can always be written as such a sum — that's the Power Series / Frobenius Method idea. We are allowed to use a plain Taylor series (no fractional powers) precisely because $x = 0$ is an ordinary point — the $y^{''}$ coefficient $(1 - x^{2})$ is nonzero there (Step 1). Had we expanded around $x = 1$ instead, we'd need the heavier Frobenius machinery.

WHY these tools — derivatives of a sum. To plug into the equation we need $y^{'}$ and $y^{''}$ . Differentiating a power term is the one rule we need: $\frac{d}{d x} x^{m} = m x^{m - 1}$ . So $y^{'} = \sum_{m} m a_{m} x^{m - 1}, y^{''} = \sum_{m} m (m - 1) a_{m} x^{m - 2} .$ Each derivative lowers the power and multiplies by the old power — remember that, it is the whole engine.

PICTURE. A stack of the basic bricks $1, x, x^{2}, x^{3}$ , each scaled by its knob $a_{m}$ , adding up to a wiggly curve.

Step 3 — Substitute, and line up every term at the same power

WHAT. Push the three pieces into the equation: $(1 - x^{2}) \sum m (m - 1) a_{m} x^{m - 2} - 2 x \sum m a_{m} x^{m - 1} + n (n + 1) \sum a_{m} x^{m} = 0.$

WHY line things up. For a sum of powers to be zero for every $x$ , the weight of each individual power $x^{m}$ must be zero on its own. (You can't cancel an $x^{2}$ against an $x^{5}$ — they're independent bricks.) So we must rewrite every sum so it reads "coefficient times $x^{m}$ ", the same power, then read off that coefficient.

WHAT we do to each piece — term by term.

$\sum m (m - 1) a_{m} x^{m - 2}$ : the power is $x^{m - 2}$ , two too low. Relabel by letting the dummy index jump up two ( $m \to m + 2$ ) so the power becomes $x^{m}$ : $\sum_{m} m (m - 1) a_{m} x^{m - 2} = \sum_{m} new front factor (m + 2) (m + 1) a_{m + 2} x^{m} .$
$- x^{2} \sum m (m - 1) a_{m} x^{m - 2} = - \sum m (m - 1) a_{m} x^{m}$ — already $x^{m}$ , leave it.
$- 2 x \sum m a_{m} x^{m - 1} = - \sum 2 m a_{m} x^{m}$ — already $x^{m}$ .
$n (n + 1) \sum a_{m} x^{m}$ — already $x^{m}$ .

PICTURE. Four ribbons of terms, three of them already parked on the " $x^{m}$ " shelf, the first one sliding two shelves over to join them.

Step 4 — Read off the coefficient of $x^{m}$ : the recurrence is born

WHAT. Collect everything multiplying $x^{m}$ and set it to zero: $from y^{''} (m + 2) (m + 1) a_{m + 2} from - x^{2} y^{''} - m (m - 1) a_{m} from - 2 x y^{'} - 2 m a_{m} from + n (n + 1) y + n (n + 1) a_{m} = 0.$

WHY. This is the "weight of $x^{m}$ is zero" rule from Step 3, made concrete. Now gather the three $a_{m}$ terms: $- m (m - 1) - 2 m + n (n + 1) = - [m (m - 1) + 2 m] + n (n + 1) = n (n + 1) - m (m + 1),$ because $m (m - 1) + 2 m = m^{2} + m = m (m + 1)$ .

PICTURE. A conveyor belt: feed in $a_{m}$ , the ratio-box multiplies it, out comes $a_{m + 2}$ — two steps down the line, never one.

Step 5 — Two independent chains: even and odd

WHAT. The general solution is $y = a_{0} built from a_{0} [even series] + a_{1} built from a_{1} [odd series] .$

WHY this matters. Because the chains are separate, a solution can be purely even (set $a_{1} = 0$ ) or purely odd (set $a_{0} = 0$ ). That is exactly the parity fact $P_{n} (- x) = (- 1)^{n} P_{n} (x)$ showing up at the source.

PICTURE. Two ladders side by side — a green even ladder ( $a_{0}, a_{2}, a_{4}, \dots$ ) and an orange odd ladder ( $a_{1}, a_{3}, a_{5}, \dots$ ) — each rung feeding the one two above it.

Step 6 — The miracle: for integer $n$ , one chain dies

WHAT happens. If $m = n$ makes the numerator zero, then $a_{n + 2} = 0$ . And since the chain multiplies forward, $a_{n + 4} = 0$ , $a_{n + 6} = 0$ , … — every coefficient beyond $a_{n}$ in that parity chain vanishes. The infinite series stops. It becomes a polynomial of degree exactly $n$ .

WHY this is the whole point. For an integer dial $n$ , one of the two chains truncates into a tidy degree- $n$ polynomial — that is the ==Legendre polynomial $P_{n} (x)$ ==. For a non-integer $n$ , the numerator never hits zero, the series runs forever, and (it can be shown) it blows up at the poles $x = \pm 1$ . Physics demands a finite temperature/potential at the poles, so it forces $n \in {0, 1, 2, \dots}$ . This is the selection rule.

PICTURE. The coefficient chain marching along, each bar shorter than the last, then — at $m = n$ — snapping to zero and staying flat forever after.

Why does the series terminate when

n

is a non-negative integer?

m = n

the numerator

m (m + 1) - n (n + 1) = 0

, so

a_{n + 2} = 0

and all higher same-parity coefficients die — leaving a degree-

n

polynomial.

Step 7 — Turning the crank: build $P_{2}$ from scratch

WHY normalize. One free knob $a_{0}$ remains — the ODE alone can't fix its size. We pin it with the universal convention $P_{n} (1) = 1$ : $a_{0} (1 - 3 \cdot 1^{2}) = a_{0} (- 2) = 1 \Rightarrow a_{0} = - \frac{1}{2} .$ Hence $P_{2} (x) = - \frac{1}{2} (1 - 3 x^{2}) = \frac{1}{2} (3 x^{2} - 1) . ✓$

PICTURE. The parabola $\frac{1}{2} (3 x^{2} - 1)$ over $[- 1, 1]$ : dips to $- \frac{1}{2}$ at the centre, rises to hit exactly $1$ at both ends $x = \pm 1$ (the normalization at work).

Step 8 — The edge case we must not forget: the other solution

WHAT $Q_{n}$ does. It does not terminate, and as $x \to \pm 1$ it grows without bound — a logarithmic blow-up at the poles. Concretely $Q_{0} (x) = \frac{1}{2} ln \frac{1 + x}{1 - x}$ , which shoots to $\pm \infty$ as $x \to \pm 1$ .

WHY we throw it away. Temperature, potential, and gravity must stay finite at the poles of the sphere. $Q_{n}$ violates that. So of the two mathematical solutions, physics keeps the bounded $P_{n}$ and discards the wild $Q_{n}$ . This is the same "bounded-at-poles" rule that forced $n$ to be an integer.

PICTURE. Two curves on $[- 1, 1]$ : the calm, finite $P_{0} = 1$ (flat) and the runaway $Q_{0}$ diving to $\pm \infty$ at the ends.

The one-picture summary

Everything on one canvas: the dome $(1 - x^{2})$ vanishing at the poles → series guess → the two-step recurrence → the even/odd chains → truncation at $m = n$ → the polynomial $P_{n}$ , with its wild twin $Q_{n}$ tossed aside.

Recall Feynman retelling — the whole walkthrough in plain words

We wanted a function that obeys one fussy rule everywhere on a line from $- 1$ to $+ 1$ . Instead of guessing the function outright, we said "it's some mix of the simplest building bricks — a flat bit, a slope bit, a bend bit, and so on" and left the amounts as knobs. Plugging that mix into the rule and demanding it hold for every point forced a tidy law: the knob two steps ahead is a fixed multiple of the current one. So the flat/bend/quartic knobs form one family, and the slope/cubic knobs form another. Now the magic: whenever our chosen dial $n$ is a whole number, that multiplier hits exactly zero right at step $n$ — so the chain of knobs suddenly stops. The endless mix collapses into a short, neat polynomial: a Legendre polynomial. And the other family of solutions? It never stops, and it explodes at the two ends of the line — the poles of a sphere. Since real temperatures and voltages can't be infinite at the poles, nature keeps the polite polynomial and throws the screaming one away. That single rule of "stay finite at the poles" is what secretly demands $n$ be a whole number in the first place.

Connections

Power Series / Frobenius Method — the machinery behind Steps 2–4.
Sturm-Liouville Theory — why the surviving $P_{n}$ come out mutually orthogonal.
Laplace's Equation in Spherical Coordinates — where the dial $n$ and $x = cos θ$ are born.
Fourier Series — same "expand in orthogonal building blocks" spirit.
Generating Functions — a slick alternate route to the same $P_{n}$ .
Associated Legendre Functions & Spherical Harmonics — the next chapter, when the sphere also varies around the equator.