I wanted to derive centripetal acceleration from scratch and tried using differential equations. But no matter what I did I hit a snag as follows:
$\alpha=$ centripetal acceleration
$\omega=$ angular velocity
$v =$ linear velocity
$s =$ distance traveled (i.e. arc length)
$r =$ radius
$t =$ time
As $\omega = \frac{\mathrm d\theta}{\mathrm dt}$, we have $\alpha =\frac{\mathrm d^2\theta}{\mathrm dt^2}$.
From $\mathrm d\theta = \frac{\mathrm ds}{r}$ we get $\mathrm d^2\theta=\frac{\mathrm d^2s}{r}$.
Also, from $\mathrm dt=\frac{\mathrm ds}{v}$ we get $\mathrm dt^2=\frac{\mathrm ds}{v}\mathrm dt$.
Putting this together, $$ \begin{align} \alpha&=\frac{\mathrm d^2s/r}{\mathrm ds\mathrm dt/v}=\frac{\mathrm d^2s v}{r\mathrm ds\mathrm dt}=\\ &=\frac{\mathrm ds v}{rs\mathrm dt}=\frac{v\mathrm dv}{rs\mathrm dt}=\\ &=\frac{v^2}{rs} \end{align} $$
So you can see the "snag" in the final result - I have a random $1/s$ getting in the way of the actual formula $\alpha=v^2/r$.
I have had a look at the calculus proof (using vectors) of this formula and am aware that there are many others. Obviously something has gone wrong here, and it may be as simple as "you just can't do that". But this has my head tied in knots so I was wondering whether anyone could actually explain why this doesn't work and why $1/s$ pops up in the answer?