In an informal exam tonight, my professor asked me to demonstrate that for $f(x)=\sin(x),\, f'(x)=\cos(x)$ using the definition of the derivative, $f'(x)=\lim_{h\rightarrow0}\frac{f(x+h)-f(x)}{h}$.

Well, plugging things in, we get $$\lim_{h\rightarrow0}\frac{\sin(x+h)-\sin(x)}{h}$$ $$=\lim_{h\rightarrow0}\frac{\sin(x)\cos(h)+\cos(x)\sin(h)-\sin(x)}{h}$$$$=\lim_{h\rightarrow0}\frac{\sin(x)\left(\cos(h)-1\right)+\cos(x)\sin(h)}{h}$$

And here I managed to stump him. In order to prove that this equals $\cos(x)$, we need to demonstrate that $\lim_{h\rightarrow0}\frac{\cos(h)-1}{h}=0$ and that $\lim_{h\rightarrow0}\frac{\sin(h)}{h}=1$. You can't simply plug in $h=0$ because that would lead to an indeterminate form. L'Hôpital's Rule does indeed get you the required limits, but that requires the very derivative we're trying to prove.

Is there a way, using the definition of the derivative, to actually prove that the derivative of sine is cosine, without a proof by construction or a proof by contradiction?

  • $\begingroup$ In case anyone's wondering, I managed to satisfy my professor by using Taylor Series, glossing over the fact that those equalities can only be proven to hold using the derivatives we're trying to find... $\endgroup$ – DonielF Jan 28 '20 at 2:54
  • 8
    $\begingroup$ These two identities are often proved using a geometric argument and the squeeze theorem. $\endgroup$ – Michael Burr Jan 28 '20 at 2:55
  • 1
    $\begingroup$ The $sinh/h=1$ can be shown through inscribed and circumscribed triangles as well as an accompanied circle sector all spanned by angle $h$. This is a non calculus approach and I am sure it can be found on this site somewhere. I am glad though that you realize that both Taylor series and Hospital Rule is NOT the correct way to prove these limits $\endgroup$ – imranfat Jan 28 '20 at 2:56
  • $\begingroup$ I mean, if you don't want to use Taylor series, then what definition of $\sin(x)$ and $\cos(x)$ are you using. I personally treat those as the definition but there are other equivalent definitions $\endgroup$ – QC_QAOA Jan 28 '20 at 2:56
  • 2
    $\begingroup$ The mentioned proof using a geometric argument and the squeeze theorem is here. $\endgroup$ – dxdydz Jan 28 '20 at 2:57

Without a proper definition, you can't prove anything. For instance, the limit

$$\lim_{t\to0}\frac{\sin t}t$$ has no meaning as long as the function $\sin t$ has not been defined, or at least some properties given. And it is not so easy to have a definition that avoids a circular argument.

The analytical definition relies on the complex exponential, via Euler's formula $$\sin t=\Im e^{it}$$ where the exponential is the entire function $$e^z=\sum_{n=0}^\infty\dfrac{z^n}{n!}$$ or equivalently a solution of the ODE $$f'(z)=f(z).$$

This obviously implies the derivative of the sine "by definition".

A slightly more geometric approach is by analytical geometry, from the equation of the unit circle, $$x^2+y^2=1,$$ giving by differentiation, $$y+x\frac{dx}{dy}=0.$$

Now if we accept the formula for the element of arc, $$ds^2=dx^2+dy^2,$$ we have

$$s=\int_0^s ds=\int_0^y\frac{dy}{\sqrt{1-y^2}}=f(y)$$ which defines a functional relation between $s$ and $y$. And by the derivative of the inverse function, let $g(s):=f^{(-1)}(s)$, we obtain


A purely geometric argument is more difficult, as it requires to prove facts about lengths/areas of triangles and circular segments, and relate the axioms of geometry to calculus.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.