0
$\begingroup$

I read directional derivative few days ago.I also read total derivative of a function.Now I have totally confused with this two topics.

Suppose there is a function $f : \mathbb R^2 \longrightarrow \mathbb R$.Suppose directional derivative of $f$ at $(a,b)$ in the direction of $(u,v)$ exists.Then

$$\lim_{h \rightarrow 0} \frac {f(a + hu,b + hv) - f(a,b)} {h}$$ exists.

Now let us associate two functions $x : \mathbb R \longrightarrow \mathbb R$ and $y : \mathbb R \longrightarrow \mathbb R$ defined by $x (t) = a + (t - c)u$ and $y (t) = b + (t - c)v$ respectively where $c \in \mathbb R$.

Then we have by above limit

$$\lim_{h \rightarrow 0} \frac {f(x(c + h),y(c + h)) - f(x(c),y(c))} {h}$$

which is actually $\frac {df} {dt} \vert_(x(c),y(c))$ i.e. $\frac {df} {dt} \vert_(a,b)$.

I think there is some big error in my calculation.Otherwise directional derivative of $f$ at $(a,b)$ along any direction coincides with the total derivative.Which is not true.Because total derivative implies continuity where directional derivative doesn't.Please tell me what is my fault.This will help me a lot in understanding this concept.

Thank you in advance.

$\endgroup$
  • $\begingroup$ By 'total derivative' do you mean the usual (Fréchet) derivative? $\endgroup$ – copper.hat Dec 31 '16 at 5:28
  • $\begingroup$ I mean @copper.hat $\frac {df (x,y)} {dt} = \frac {\partial f} {\partial x} . \frac {dx} {dt} + \frac {\partial f} {\partial y} . \frac {dy} {dt}$ to be the total derivative of $f$ w.r.t. $t$. $\endgroup$ – Arnab Chatterjee. Dec 31 '16 at 5:36
  • $\begingroup$ But @copper.hat does the existence of the limit $$\lim_{h \rightarrow 0} \frac {f(x(c + h),y(c + h)) - f(x(c),y(c))} {h}$$ in my post not ensure that $f$ is continuous at $(x(c),y(c))$ i.e. at $(a,b)$.If the answer is 'no' then please explain. $\endgroup$ – Arnab Chatterjee. Dec 31 '16 at 7:38
  • $\begingroup$ Whatever you call the derivative above, it does not imply continuity, except along the direction of interest (where it it differentiable along that line). However, one can construct a simply example on the plane of a function $f$ that has a directional derivative everywhere, but is not continuous at the origin. $\endgroup$ – copper.hat Dec 31 '16 at 7:48
  • $\begingroup$ But @copper.hat if the above limit exists say this $l$ then for we can write it as $f(x(c + h),y (c + h)) - f(x(c),y(c)) = hl + {\epsilon} h$, where $\epsilon \rightarrow 0$ as $h \rightarrow 0$.It implies that $$\lim_{h \rightarrow 0} [f(x(h + c),y(h + c)) - f(x(c),y(c))] = 0$$.Does it not imply that $f$ is continuous at $(x(c),y(c))$?Please don't mind if my question is too silly.But I have to understand it clearly before going to further study. $\endgroup$ – Arnab Chatterjee. Dec 31 '16 at 8:27
0
$\begingroup$

There is a danger of slipping into a notational morass which I will try to avoid.

There is no error as far as I can see. I think you have confused yourself with notation. Hopefully this may help straighten things out.

The directional derivative of $f$ in the direction $d_0$ at the point $x_0$ is given by $df(x_0,d_0) = \lim_{t \to 0 } {f(x_0+td_0)-f(x_0) \over t}$. In my notation $x_0,d_0$ are in $\mathbb{R}^2$ (or $\mathbb{R}^n$, more generally).

Note that if we let $\phi(t) = f(x_0+td_0)$, then $\phi'(0) = df(x_0,d_0)$.

Returning to your notation, we have $x(s) = (a,b)+(s-c) (u,v)$, and we consider $\eta(t) = f(x(t+c))$ and note that $\eta(t) = f((a,b))+ t (u,v))$, hence $\eta'(0) = df((a,b),(u,v))$.

$\endgroup$

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.