Visual walkthrough — Binomial theorem for rational indices (approximate values)

Everything below assumes only: you can multiply, you know what "a bit more than 1" means, and you have met a slope (how steep a curve is) once before. Nothing else. We build the rest.

---

Step 1 — Start with the honest picture: a curve near $x=0$

WHAT. Draw the function $f(x)=(1+x{)}^n$. This just means: take the number $1+x$ (a number close to $1$ when $x$ is small) and raise it to the power $n$. Here $n$ can be any number — a whole number, a half ($\frac{1}{2}$ means square-root), or a negative number ($-1$ means "one over").

WHY. Before we can approximate the curve, we must look at it. Everything we do lives near the single point $x=0$, because there $1+x=1$ and $(1{)}^n=1$ — a value we know for free. We will slowly walk away from that anchor.

PICTURE. The blue curve is $(1+x{)}^{1/2}=\sqrt{1+x}$. Notice it passes through the yellow dot at $x=0,y=1$. That dot is our launch pad.

---

Step 2 — Term zero: the flat guess

WHAT. The crudest possible approximation: pretend the curve is a flat horizontal line at height $1$. In symbols, $f(x)\approx 1$.

WHY. At $x=0$ the true height is exactly $1$. If we refuse to move, height $1$ is the best single number we can name. This is the "start" term of the series — the constant $1$.

PICTURE. The pale-yellow horizontal line sits at height $1$. Near the dot it hugs the blue curve; far away it drifts off. The gap between line and curve (pink) is the error we must fix in the next steps.

$\underset{\text{height at }x=0}{\underbrace{1}}=f(0)=(1+0{)}^n$

Each symbol: $f(0)$ is "the curve's height when $x=0$"; $(1+0{)}^n=1^n=1$ always, no matter what $n$ is.

---

Step 3 — Term one: tilt the line to match the slope → the $nx$ term

WHAT. A flat line ignores that the curve is climbing (or falling). Fix it: use a tilted straight line through the dot whose steepness equals the curve's steepness at $x=0$. Steepness of a curve is its slope, written $f^{\prime }(x)$ (read "f-prime"). The slope tells you how much height you gain per step sideways.

WHY use slope here? Because a straight line is completely decided by (a) a point it passes through and (b) its steepness. We already have the point (the yellow dot). The one extra number that makes the line kiss the curve rather than cross it randomly is exactly the curve's slope at that point. No other tool gives "best straight-line match" — that is precisely what slope means.

To get the slope of $(1+x{)}^n$: the rule for powers is that differentiating $(1+x{)}^n$ pulls the power down in front and lowers the power by one: $f^{\prime }(x)=n(1+x{)}^{n-1}.$ At the dot, $x=0$, so $(1+0{)}^{n-1}=1$ and the slope is just $n$.

$f^{\prime }(0)=n\cdot 1=n\Rightarrow \text{line}=\underset{\text{start height}}{\underbrace{1}}+\underset{\text{slope}}{\underbrace{n}}\cdot \underset{\text{sideways step}}{\underbrace{x}}$

So the tilted line is $1+nx$. Term by term: $n$ is how steep, $x$ is how far sideways, their product $nx$ is how much height we gained.

PICTURE. The pink tilted line now leaves the dot at the correct angle and clings to the blue curve far longer than the flat line did.

Recall Why "slope of

$(1+x{)}^n$ is $n(1+x{)}^{n-1}$"? Question ::: What are the two things the power rule does to the exponent? Answer ::: It (1) drops the old power $n$ down as a multiplier out front, and (2) reduces the exponent by one to $n-1$. At $x=0$ the $(1+0{)}^{n-1}$ part is $1$, leaving just $n$.

---

Step 4 — The line still bends away: introduce curvature → the $x^2$ term

WHAT. Look again at Step 3's picture: even the tilted line eventually peels off the curve, because the curve bends and a straight line cannot. We measure bending with the slope of the slope, written $f^{\prime \prime }(x)$ ("f-double-prime") — how fast the steepness itself is changing.

WHY a second derivative? A straight line captures direction. To also capture bending we need one more number: how the direction changes. That is exactly $f^{\prime \prime }$. Add a gently-curved piece (a parabola, shape $x^2$) whose bending matches, and the fit tightens dramatically.

Differentiate the slope again — same rule, pull down and lower: $f^{\prime \prime }(x)=n(n-1)(1+x{)}^{n-2}\Rightarrow f^{\prime \prime }(0)=n(n-1).$

Now, why divide by $2$? The bit of height a parabola $c{x}^2$ contributes has slope-of-slope equal to $2c$ (differentiate $c{x}^2$ twice: first $2cx$, then $2c$). We want that to equal $f^{\prime \prime }(0)=n(n-1)$, so $2c=n(n-1)\Rightarrow c=\frac{n(n-1)}{2}.$ That division by $2$ (the $2!$ in the formula) is not decoration — it is the price of matching curvature honestly.

$\text{new term}=\underset{\text{half the bending}}{\underbrace{\frac{n(n-1)}{2}}}\underset{\text{parabola shape}}{\underbrace{x^2}}$

PICTURE. The green curve $1+nx+\frac{n(n-1)}{2}{x}^2$ now bends with the blue curve and stays glued much further out.

---

Step 5 — See the pattern: each new term $=\frac{f^{(r)}(0)}{r!}{x}^r$

WHAT. We keep going. The $r$-th correction fixes the $r$-th layer of bending. Its recipe is always the same: take the $r$-th derivative at $0$, divide by $r!$ (read "$r$ factorial" $=1\cdot 2\cdot 3\cdots r$), multiply by $x^r$.

WHY the factorial $r!$? Same reason as the $2$ in Step 4. Differentiating $x^r$ a total of $r$ times spits out $r\cdot (r-1)\cdots 2\cdot 1=r!$. To undo that pile-up and keep the coefficient honest, we divide by $r!$.

Each derivative of $(1+x{)}^n$ at $0$ pulls one more factor off the front: $f^{(r)}(0)=n(n-1)(n-2)\cdots (n-r+1).$

So the $r$-th coefficient is $a_r=\frac{f^{(r)}(0)}{r!}=\frac{\stackrel{r\text{ falling factors}}{\overbrace{n(n-1)\cdots (n-r+1)}}}{\underset{\text{undo the pile-up}}{\underbrace{r!}}}.$

PICTURE. A staircase: term $0$ is flat, term $1$ tilts, term $2$ bends, term $3$ bends the bending — each rung a shorter correction than the last.

Assembling the whole staircase: $(1+x{)}^n=1+nx+\frac{n(n-1)}{2!}{x}^2+\frac{n(n-1)(n-2)}{3!}{x}^3+\cdots$

---

Step 6 — Why it must stop making sense at $|x|=1$ (the convergence edge)

WHAT. The staircase is infinite for non-integer $n$ — no factor $(n-r)$ ever hits zero to end it. So we must ask: do the corrections keep shrinking? Only then does the infinite pile add up to a finite answer.

WHY does size of $x$ decide this? Term $r$ carries an $x^r$. If $|x|<1$ (a fraction), then $x^2,x^3,\dots$ get tinier and tinier — the tower converges. If $|x|>1$, then $x^r$ grows, each correction bigger than the last, and the sum flies to infinity — meaningless. The knife-edge is $|x|=1$.

PICTURE. Two shrinking-vs-growing bar towers. Left ($x=0.5$): bars collapse to nothing — safe. Right ($x=1.5$): bars explode — the series is nonsense there.

---

Step 7 — Degenerate check: does it still work when $n$ is a whole number?

WHAT. Feed in $n=3$ and watch the infinite staircase collapse to the finite ordinary binomial.

WHY check this? A good general formula must contain the special case it generalises. If our rational-index machine didn't reproduce plain $(1+x{)}^3$, it would be wrong.

$$\begin{array}{c}{a}_0=1,a_1=3,a_2=\frac{3\cdot 2}{2}=3,a_3=\frac{3\cdot 2\cdot 1}{6}=1,\\ a_4=\frac{3\cdot 2\cdot 1\cdot \stackrel{n-3=0}{\overbrace{0}}}{24}=0.\end{array}$$

The factor $(n-3)=0$ kills $a_4$ and every term after it. So $(1+x{)}^3=1+3x+3x^2+x^3,$ the familiar terminating result. The infinite tail vanishes exactly when $n$ is a non-negative integer — connecting straight back to Binomial Theorem for Positive Integral Index.

PICTURE. The infinite staircase for $n=\frac{1}{2}$ beside the truncated 4-rung staircase for $n=3$; rung 4 onward is greyed out (zero).

---

The one-picture summary

Everything at once: the flat guess ($1$), the tilt ($+nx$), the bend ($+\frac{n(n-1)}{2}{x}^2$) — each new colour hugging the blue truth curve $\sqrt{1+x}$ a little longer, all anchored at the yellow launch dot, valid only inside the shaded band $|x|<1$.

Recall Feynman retelling — the whole walk in plain words

Stand on a curve at the spot where $x=0$; there the height is exactly $1$ — that's our free starting number. First guess: "stay flat at $1$." Bad far away, so we tilt our line to match how steeply the curve rises — the steepness is $n$, so we add $nx$ ("$n$ times how far I stepped"). The curve still bends away from any straight line, so we bolt on a gentle parabola whose bending matches the curve's; matching curvature forces the $\frac{n(n-1)}{2}$ out front (the $2$ pays for double-differentiating $x^2$). Keep matching higher and higher bends — each fix divides by a bigger factorial and rides a higher power of $x$, so each is smaller than the last. That shrinking is guaranteed only when $x$ is a fraction ($|x|<1$); make $x$ bigger than $1$ and the fixes grow instead — chaos. Finally, if $n$ happens to be a whole number, one falling factor becomes zero and the endless staircase snaps shut into the plain school-book binomial. That's the entire story: start at 1, add $n$ times the small bit, polish with the bend, stop when the bits are tiny.

Connections

• Maclaurin & Taylor Series — Steps 2–5 are the Maclaurin recipe $a_r=f^{(r)}(0)/r!$.
• Binomial Theorem for Positive Integral Index — Step 7's collapse.
• Convergence of Series & Infinite Geometric Series — Step 6's $|x|<1$ edge.
• Error & Approximation in Calculus — the leftover gap after truncating is ~ the first dropped term.