Visual walkthrough — Parseval's theorem

We build one idea at a time. Read the figure, then the words.

---

Step 1 — What is a function's "energy"?

WHY square first? If we just integrated $f$, the parts above and below the axis would cancel and a big wobbly wave could score zero. Squaring makes every wiggle count as a positive contribution — exactly like how a loud sound wave carries power whether the pressure is above or below normal.

PICTURE: the red curve is $f$; the grey region is $[f(x){]}^2$. Energy is the grey area.

---

Step 2 — The function is secretly a sum of pure tones

WHY split it up? A messy curve is hard to reason about. Pure sines and cosines are simple and — crucially — they don't interfere with each other (Step 4). See Fourier Series for how the knobs $a_n,b_n$ are found.

PICTURE: the red target curve on top; below it, three pure tones that add up to it. Turn the knobs, stack the tones, recover the curve.

---

Step 3 — Energy of ONE pure tone

WHY is ${\cos }^2$ average $\frac{1}{2}$? Look at the figure: ${\cos }^2$ bobs between $0$ and $1$, symmetric about the line $y=\frac{1}{2}$. The area above and below that midline match — so the average height is exactly $\frac{1}{2}$. Same story for ${\sin }^2$.

PICTURE: red curve ${\cos }^2\frac{\pi x}{L}$; the dashed line at $\frac{1}{2}$ is its average; the grey area equals the area of the rectangle of height $\frac{1}{2}$.

---

Step 4 — Two different tones don't overlap (orthogonality)

WHY does this matter? When we square the whole series (next step), we get products of every pair of tones. This law says every "mixed" product silently deletes itself — only a tone times itself survives. That is the entire trick. This is exactly Orthogonality of functions — two waves are "perpendicular" if their product integrates to zero, just like two arrows are perpendicular when their dot product is zero.

PICTURE: red is the product $\cos \frac{\pi x}{L}\cos \frac{2\pi x}{L}$. The blue-shaded positive lobes and grey negative lobes have identical area — they cancel to nothing.

---

Step 5 — Square the whole series and integrate

WHY split into diagonal and cross? Because Step 4 already told us the fate of each family: every cross term integrates to $0$ (orthogonality), every diagonal term integrates to a clean $L$ (Step 3). We just sweep the table.

PICTURE: a grid. Rows and columns are the tones; each cell is a product. The red diagonal cells survive; every off-diagonal (cross) cell is crossed out — it vanishes on integration.

---

Step 6 — Add up the survivors

Divide both sides by $L$: $\boxed{\frac{1}{L}{\int }_{-L}^L[f(x){]}^{2}dx=\frac{a_0^2}{2}+\sum _{n=1}^{\infty }(a_n^2+b_n^2)}$

WHY divide by $L$? It turns the raw energy into an average (mean-square) and makes the DC term line up as $\frac{a_0^2}{2}$ — the same pattern as the others but wearing "half a coat." This is Parseval's theorem, the equality case of Bessel's Inequality.

PICTURE: left column is the single grey energy bar of $f$; right column is that same total height, sliced into stacked colored bins — one per harmonic. Same total, redistributed.

---

Step 7 — Degenerate & edge cases (so nothing surprises you)

WHY list these? A theorem you can't stress-test you don't trust. In every case the two sides match, and the "missing" coefficients simply mean empty energy bins, not a broken formula.

PICTURE: two bar charts side by side — the odd function fills only the sine bins (red), the even function fills only the cosine bins. Complementary, never contradictory.

---

The one-picture summary

Everything compressed: the wave, its energy area, its decomposition into tones, and the energy-bin bar chart that must sum to the same total. Read it left to right and you have re-derived Parseval.

Recall Feynman retelling — the whole walkthrough in plain words

Take a wiggly wave. Its "energy" is the shaded area you get after squaring it (squaring so nothing cancels). Now here's the secret: that wave is really a chord — a stack of pure tones (sines and cosines), each turned up by some volume knob. When you square the whole chord you get every tone times every other tone. But two different tones, multiplied and totaled, always cancel to nothing — they're perpendicular, like arrows at right angles. So all the messy mixed terms vanish. Only each tone-times-itself is left, and a single tone's energy is just its volume-knob squared (times a fixed factor). Add them up, divide by the interval width to make it an average, and you land on: total average energy = sum of squared volume knobs. Nothing leaks. That's why the same trick can add up $1+\frac{1}{4}+\frac{1}{9}+\cdots =\frac{{\pi }^2}{6}$ — see the Basel Problem — and why engineers use it for power in each harmonic.

Connections

• Parseval's theorem — the parent note (statement + worked sums).
• Fourier Series — where the tones and knobs $a_n,b_n$ come from.
• Orthogonality of functions — Step 4, the engine.
• Bessel's Inequality — Parseval is its equality case.
• Plancherel Theorem — the same idea for the continuous transform.