Visual walkthrough — Parametric differentiation — dy - dx, d²y - dx²

Before anything, two dots of vocabulary we will lean on constantly:

Step 1 — The bug and its two speeds

WHAT. We place a moving point (the "bug") on a curve. Its position at clock-time $t$ is $(x(t),y(t))$. In one tick $dt$ it slides a little.

WHY. Because the curve has no equation $y=f(x)$ — it may loop, cross itself, or stand vertical. The only handle we have is the clock $t$. So we describe motion first, and squeeze geometry out of it afterward.

PICTURE. The little step splits into a horizontal piece $dx=\dot{x}dt$ (the run) and a vertical piece $dy=\dot{y}dt$ (the rise). In the figure, follow the teal east arrow and the orange north arrow; together they build the dashed plum step along the curve.

• $dx=\dot{x}dt$ — how far east the bug drifts in one tick: east-speed $\dot{x}$ times the tick $dt$. If $\dot{x}>0$ it goes right, if $\dot{x}<0$ it backs up left.
• $dy=\dot{y}dt$ — how far north it drifts in the same tick: north-speed times the same $dt$.

Step 2 — Slope is the ratio of the two speeds

WHAT. The slope of the path is rise-over-run. Divide the vertical step by the horizontal step: $\frac{dy}{dx}=\frac{\dot{y}dt}{\dot{x}dt}=\frac{\dot{y}}{\dot{x}}$

WHY. Slope must not depend on how fast the bug walks — only on the shape of the road. The clock-tick $dt$ appears in both rise and run, so it cancels, leaving a pure ratio of speeds.

• $\dot{y}$ (top) — north speed, the "rise per tick".
• $\dot{x}$ (bottom) — east speed, the "run per tick".
• $dt$ — the shared tick that divides out; this is why walking-speed is irrelevant.

PICTURE. Two bugs, fast and slow, on the same hill. Different arrow lengths, identical arrow angle — same slope.

This first result is the whole first derivative. Call it $y^{\prime }$: $y^{\prime }\equiv \frac{dy}{dx}=\frac{\dot{y}}{\dot{x}}(\dot{x}\ne 0)$

Step 3 — The trap: "just do it again"

WHAT. The second derivative feels like it should be $\ddot{y}/\ddot{x}$ — copy the pattern from Step 2 but with accelerations on top and bottom.

WHY it's wrong. Look at the notation carefully. The symbol $\frac{d^{2}y}{d{x}^2}$ is shorthand for "apply the slope-taking operation $\frac{d}{dx}$ to the slope $\frac{dy}{dx}$": $\frac{d^{2}y}{d{x}^2}\equiv \frac{d}{dx}\left(\frac{dy}{dx}\right)$

• The outer $\frac{d}{dx}$ says "differentiate with respect to $\mathbf{x}$" — not with respect to $t$. It measures change per step east.
• $\ddot{y}/\ddot{x}$ would be a ratio of time-accelerations. That answers a different question: "how does the bug's velocity vector turn per tick?", not "how does the map-slope change as I move east."

PICTURE. Two labelled boxes. The teal box is the question we actually want (slope changing per unit of eastward distance). The plum box is the seductive impostor (accelerations per tick). They are simply not the same measurement.

Step 4 — Reuse the Step-2 trick on the slope itself

WHAT. We need $\frac{d}{dx}(y^{\prime })$, but we only know how to differentiate against $t$. So apply the exact same speed-ratio idea from Step 2 (the chain-rule engine), now to the quantity $y^{\prime }$ instead of $y$: $\frac{d}{dx}(y^{\prime })=\frac{d{y}^{\prime }/dt}{dx/dt}=\frac{\frac{d}{dt}\left(y^{\prime }\right)}{\dot{x}}$

WHY. Anything that varies with $t$ obeys "rate-along-x = (rate-along-t) ÷ (x-speed)". The slope $y^{\prime }$ is just another passenger riding the clock. So it gets the same treatment $y$ got.

• Numerator $\frac{d}{dt}(y^{\prime })$ — how the slope changes per tick.
• Denominator $\dot{x}$ — the east speed we must divide by to convert "per tick" into "per unit east". This is the future source of the extra power.

PICTURE. A relabelled version of the Step-2 triangle: the vertical axis is now "slope value $y^{\prime }$", the horizontal axis is still $x$. Same divide-by-$\dot{x}$ move.

Step 5 — Differentiate the slope with the quotient rule

WHAT. Now compute $\frac{d}{dt}(y^{\prime })$ where $y^{\prime }=\frac{\dot{y}}{\dot{x}}$. This is a fraction whose top and bottom both depend on $t$, so the Quotient rule is the right tool: $\frac{d}{dt}\left(\frac{\dot{y}}{\dot{x}}\right)=\frac{\stackrel{{\text{(top)}}^{\prime }\cdot \text{bottom}}{\overbrace{\ddot{y}\dot{x}}}-\stackrel{\text{top}\cdot (\text{bottom}{)}^{\prime }}{\overbrace{\dot{y}\ddot{x}}}}{{\dot{x}}^2}$

WHY the quotient rule and not something simpler? Because $y^{\prime }$ is a ratio of two changing things. If we only differentiated the top we would pretend the bottom ($\dot{x}$) is frozen — but $\dot{x}$ is itself changing at rate $\ddot{x}$. The quotient rule is precisely the bookkeeping that credits both changes.

Term by term:

• $\ddot{y}\dot{x}$ — "differentiate the top, keep the bottom." The north-acceleration scaled by east-speed.
• $-\dot{y}\ddot{x}$ — "keep the top, differentiate the bottom." This is the term that vanishes only when $\ddot{x}=0$; drop it and the cycloid example breaks.
• ${\dot{x}}^2$ — the quotient rule always squares the denominator. First appearance of a power of $\dot{x}$.

PICTURE — read it like this. In the figure the two ingredients sit at the corners: $\dot{y}$ (orange, top) and $\dot{x}$ (teal, bottom). Each diagonal arrow differentiates the corner it leaves from:

• the arrow from $\dot{y}$ down to the $\dot{x}$-corner builds the product $\ddot{y}\dot{x}$ (top differentiated, times bottom);
• the arrow from $\dot{x}$ up to the $\dot{y}$-corner builds $\dot{y}\ddot{x}$ (bottom differentiated, times top). Then — crucially — the order of subtraction is "top-diagonal minus bottom-diagonal": $\ddot{y}\dot{x}-\dot{y}\ddot{x}$. Reverse them and every concavity sign flips, so mind the minus.

Step 6 — Divide once more: the cube is born

WHAT. Put Step 4 and Step 5 together. Step 4 said "divide $\frac{d}{dt}(y^{\prime })$ by $\dot{x}$." Step 5 gave that numerator. So: $\frac{d^{2}y}{d{x}^2}=\frac{1}{\dot{x}}\cdot \frac{\ddot{y}\dot{x}-\dot{y}\ddot{x}}{{\dot{x}}^2}=\boxed{\frac{\ddot{y}\dot{x}-\dot{y}\ddot{x}}{{\dot{x}}^3}}$

WHY the cube. Two divisions stack:

• ${\dot{x}}^2$ from the quotient rule (Step 5),
• $\times \frac{1}{\dot{x}}$ from converting "per tick" to "per unit east" (Step 4).

${\dot{x}}^2\cdot \dot{x}={\dot{x}}^3$. That is the entire origin of the fingerprint cube. No cube ⇒ you forgot one of the two divisions.

PICTURE — read it bottom-up. The figure is a stack. The lowest tile is the quotient-rule denominator ${\dot{x}}^2$. Directly above it sits the second tile, the extra $÷\dot{x}$ from Step 4. Follow the downward arrow that merges the two tiles: ${\dot{x}}^2\cdot \dot{x}$ collapses into the single label ${\dot{x}}^3$ in the basement, and the finished formula is printed beneath. The point of the picture is that there are two separate floors of division, not one.

Step 7 — The degenerate case: $\dot{x}=0$

WHAT. Every step above quietly divided by $\dot{x}$. What if the bug's east-speed is zero at some instant?

WHY it matters. $\dot{x}=0$ means the bug is moving straight up or down for that tick — the tangent is vertical. Then:

• $\frac{dy}{dx}=\frac{\dot{y}}{0}$ is undefined (slope $\to \infty$),
• $\frac{d^{2}y}{d{x}^2}$ has ${\dot{x}}^3=0$ underneath, also undefined.

This is geometry telling the truth, not a bug in the formula: on a vertical tangent there is genuinely no finite "rise per run". See Tangents and normals for how to handle such points (swap roles: use $\frac{dx}{dy}$ instead).

PICTURE. A unit circle $x=\cos t,y=\sin t$. At $t=0$ the point is $(1,0)$, the arrow points straight up, $\dot{x}=-\sin 0=0$ — vertical tangent, formula undefined. At $t=\pi /2$ the point is $(0,1)$, $\dot{x}=-1\ne 0$, slope $=0$ (flat top). Both cases shown.

The one-picture summary

This single diagram compresses all seven steps: start with four raw numbers ($\dot{x},\dot{y},\ddot{x},\ddot{y}$), form the slope by one division, differentiate that slope with the quotient rule (giving ${\dot{x}}^2$), then divide once more (giving the total ${\dot{x}}^3$).

Active recall

Connections

• Parametric differentiation — dy - dx, d²y - dx² — the parent result this page derives visually.
• Chain rule — converts "per tick" rates into "per unit $x$" rates (Steps 2 & 4).
• Quotient rule — differentiates the slope $\dot{y}/\dot{x}$ (Step 5).
• Concavity and second derivative — reads off concave up/down from the sign (Step 7).
• Tangents and normals — handles the vertical-tangent case where $\dot{x}=0$.
• Cycloid, Parametric curves — curves where the $\dot{y}\ddot{x}$ term truly matters.