Parseval's theorem: Foundations — Maths deep dive

This is a foundations page. If a symbol in the parent note made you pause, it is defined here — from zero, with a picture. Read top to bottom; each block leans on the one above it.

0. How to read the symbols (the alphabet first)

Before any formula, here is the raw notation the parent throws at you. We unpack each properly below — this is just so nothing on the page is a stranger.

Symbol	Said out loud	Rough meaning
$f (x)$	"eff of ex"	a function: a machine, feed in $x$ , get out a number
$L$	"ell"	a fixed positive number: the half-width of the interval
$π$	"pie"	the circle constant $\approx 3.14159$
$\int_{- L}^{L} d x$	"integral from minus L to L"	total area under a curve between two points
$\sum_{n = 1}^{\infty}$	"sum, n from 1 to infinity"	add up infinitely many terms
$sin, cos$	"sine, cosine"	the two basic wiggle-shapes
$\frac{nπ x}{L}$	inside the wiggle	controls how fast the wave wiggles
$a_{n}, b_{n}$	"a-sub-n, b-sub-n"	how much of each wave is in $f$
$[f (x)]^{2}$	"f squared"	the function multiplied by itself

1. What is a function $f (x)$ ? (the machine picture)

Why the topic needs it. Parseval's theorem is a statement about a function — specifically about how "big" a function is overall. Before we can measure a function we must agree it is a curve of heights.

The parent uses examples like $f (x) = x$ (a straight ramp) and $f (x) = x^{2}$ (a bowl). Both are just curves: at each $x$ , read off the height.

2. The interval: what is $L$ ? And what is $π$ ?

Together, $L$ and $π$ team up inside the waves as $\frac{nπ x}{L}$ (built in §3a) — $π$ supplies the circle, $L$ scales it to fit our window.

3. What is $\int_{- L}^{L} f (x) d x$ ? (area, and why we square)

The little $d x$ is a reminder that we slice the region into infinitely thin vertical strips of width $d x$ and add all their areas.

This is exactly why the left-hand side of Parseval is $\int f^{2}$ and not $\int f$ . Squaring is the "size" trick.

4. What are $sin$ and $cos$ ? (the two wiggle-shapes)

Why the topic needs them. Parseval is about breaking $f$ into building blocks. The chosen blocks are sine and cosine waves, because they are the natural "pure tones" — see Fourier Series.

4a. What does $\frac{nπ x}{L}$ do inside the wave?

The wave is written $cos \frac{nπ x}{L}$ . The fraction inside is the angle. Two knobs live there:

$\frac{π x}{L}$ stretches the wave so that one full cycle fits neatly across the interval $(- L, L)$ — it makes the wave periodic with the right period. (At $x = L$ the angle is $nπ$ , so the pattern lines up at the window edges.)
The whole-number $n = 1, 2, 3, \dots$ is the harmonic number. Bigger $n$ = more wiggles in the same interval = higher frequency.

$n = 1$ is the slow fundamental wave; $n = 2$ wiggles twice as fast; and so on.

5. What are $a_{n}$ and $b_{n}$ ? (recipe amounts — with their formulas)

But "how much" is not a feeling — it is computed by an exact integral. Each coefficient is found by multiplying $f$ by the wave you want to measure and integrating (this is orthogonality at work — §6):

Think of a smoothie: $a_{1}, b_{1}, a_{2}, b_{2}, \dots$ are the number of scoops of each pure ingredient. The parent's line $f (x) = \frac{a _{0}}{2} + \sum_{n = 1}^{\infty} (a_{n} cos \frac{nπ x}{L} + b_{n} sin \frac{nπ x}{L})$ just says: stack all these scaled waves and you rebuild $f$ .

6. Orthogonality — the reason cross-terms die

This is the single most important prerequisite. The parent calls it "the engine."

Different waves ( $m \neq = n$ ): product integrates to $0$ — they do not overlap.
The same wave times itself ( $m = n$ ): the product is always $\geq 0$ (a square), so its area is a positive number, namely $L$ .

Why the topic lives or dies on this. When you square the series and integrate, you get a storm of products. Orthogonality wipes out every "mixed" product (the cross-terms), leaving only each wave squared. That collapse is Parseval's theorem. Full treatment in Orthogonality of functions.

7. The vector analogy — why this is "Pythagoras for functions"

Vectors	Functions
axes $x, y, z$	waves $cos \frac{nπ x}{L}, sin \frac{nπ x}{L}$
perpendicular axes	orthogonal waves
parts $v_{x}, v_{y}, v_{z}$	coefficients $a_{n}, b_{n}$
length $^{2} = \sum v_{i}^{2}$	$\frac{1}{L} \int f^{2} = \frac{a _{0}^{2}}{2} + \sum (a_{n}^{2} + b_{n}^{2})$

This table is the whole page in one image. Everything above just earns the right to read it.

8. Sums to infinity — reading $\sum_{n = 1}^{\infty}$

Why the topic needs it. The right-hand side of Parseval is an infinite sum $\sum (a_{n}^{2} + b_{n}^{2})$ , because there are infinitely many harmonics $n = 1, 2, 3, \dots$ . Being comfortable that such a sum can equal a finite number is essential.

9. Prerequisite map

Equipment checklist

Cover the right side and test yourself.

I can read

f (x)

as a curve of heights

Yes — feed in

x

, the output is the height of the curve at that point.

I know what

L

A fixed positive number; the window runs from

- L

+ L

, total width

2 L

I know what

π

The circle constant

\approx 3.14159

; one full trip round a circle is angle

2 π

I know what

\int_{- L}^{L} f d x

measures

The signed area between the curve and the axis from

- L

L

I know why the parent squares

f

before integrating

Squaring makes every height positive, so the integral measures total size/energy instead of cancelling to zero.

I know why the averaging factor is

\frac{1}{L}

not

\frac{1}{2 L}

To cancel the

L

from each wave's self-integral, giving the tidy right side; it is twice the mean-square value.

I can say what

cos \frac{nπ x}{L}

A cosine wave whose speed of wiggling is set by the harmonic number

n

; one cycle fits the interval scaled by

π / L

I know the defining formulas for

a_{n}

and

b_{n}

a_{n} = \frac{1}{L} \int_{- L}^{L} f cos \frac{nπ x}{L} d x

b_{n} = \frac{1}{L} \int_{- L}^{L} f sin \frac{nπ x}{L} d x

a_{0} = \frac{1}{L} \int_{- L}^{L} f d x

I can state orthogonality in words

Multiply two different waves and integrate over

(- L, L)

; the result is

0

. Same wave squared gives

L

(but the constant wave gives

2 L

I know why the DC term needs

2 L

not

L

\int_{- L}^{L} 1 d x = 2 L

, twice every other wave, which is why

a_{0}

is stored as

a_{0} /2

I understand "Pythagoras for functions"

Coefficients are the function's parts along perpendicular wave-axes; length-squared = sum of squared parts.

I can read

\sum_{n = 1}^{\infty}

Add infinitely many terms and ask what finite number the total approaches.

Ready? Return to Parseval's theorem and the proof will read like plain English.

Connections

Parseval's theorem — the topic these foundations serve.
Fourier Series — where $a_{n}, b_{n}$ and the wave-stacking come from.
Orthogonality of functions — the engine unpacked further.
Basel Problem — the famous $\sum 1/ n^{2} = π^{2} /6$ payoff.
Bessel's Inequality — what happens with an incomplete set of waves.