Direction in which directional derivative is zero

In summary, the directional derivative of a function represents the rate of change of the function in a specified direction. When the directional derivative is zero, it indicates that there is no change in the function's value in that direction. This occurs at critical points or along level curves, where the gradient of the function is perpendicular to the direction being considered. Consequently, identifying directions with zero directional derivatives is essential for locating extrema and understanding the behavior of functions in multivariable calculus.
  • #1
songoku
2,384
351
Homework Statement
Please see below
Relevant Equations
Partial Derivative

Direction derivative in the direction of unit vector u = <a, b, c>:
Du f(x,y,z) = fx (x,y,z) a + fy (x,y,z) b + fz (x, y, z)
1697812128978.png

I want to ask about the direction in which ##D_v## is zero at point (1, 2, 1)

My attempt:
$$w_x=yz+\frac{1}{x}$$
$$w_y=xz+\frac{1}{y}$$
$$w_z=xy+\frac{1}{z}$$

At point (1, 2, 1), the ##\nabla w=<3, \frac{3}{2}, 3>##

$$D_v w=0$$
$$\nabla w \cdot v=0$$
$$
\begin{pmatrix}
3 \\
\frac{3}{2} \\
3
\end{pmatrix}
\cdot
\begin{pmatrix}
p \\
q \\
r
\end{pmatrix}
=0
$$
$$3p+\frac{3}{2} q+3r=0$$
$$2p+q+2r=0$$

Another equation is ##p^2+q^2+r^2=1##

But I can't find ##p, q## and ##r## from these equations. Is my working wrong?

Thanks
 
Last edited:
Physics news on Phys.org
  • #2
songoku said:
Homework Statement: Please see below
Relevant Equations: Partial Derivative

Direction derivative in the direction of unit vector u = <a, b, c>:
Du f(x,y,z) = fx (x,y,z) a + fy (x,y,z) b + fz (x, y, z)

View attachment 333900
I want to ask about the direction in which ##D_v## is zero at point (1, 2, 1)

My attempt:
$$w_x=yz+\frac{1}{x}$$
$$w_y=xz+\frac{1}{y}$$
$$w_z=xy+\frac{1}{z}$$

At point (1, 2, 1), the ##\nabla w=<3, \frac{3}{2}, 3>##

$$D_v w=0$$
$$\nabla w \cdot v=$$
$$
\begin{pmatrix}
3 \\
\frac{3}{2} \\
3
\end{pmatrix}
\cdot
\begin{pmatrix}
p \\
q \\
r
\end{pmatrix}
=0
$$
$$3p+\frac{3}{2} q+3r=0$$
$$2p+q+2r=0$$

Another equation is ##p^2+q^2+r^2=1##

But I can't find ##p, q## and ##r## from these equations. Is my working wrong?

Thanks
No. I can't find a mistake, except that I have no idea what ##\vec{u}=\vec{i}+\vec{j}## is since your coordinates seem to be ##x,y,z## and not ##\vec{i},\vec{j},\vec{k},## and why you did not calculate ##\nabla_{(1,2,1)}(\omega )\cdot \vec{u}## as requested.

The points in
$$
\left\{(p,q,r)\,|\,\nabla_{(1,2,1)}(\omega )\cdot (p,q,r)=\vec{0}\right\} =\left\{(p,q,r)\in \mathbb{R}^3\,|\,2p+q+2r=0\right\}=(2,1,2)^\perp
$$
form a hyperplane, i.e. a plane in the ##3##-dimensional Euclidean space. This makes sense because it is the kernel of a linear function of rank one, the kernel of the multiplication with the gradient, the kernel of the nabla operator.

Finally, you want to consider only unit vectors (why?). This means to intersect the plane with the unit sphere. The result will be an ellipse somewhere in space. Unfortunately, the emphasis is on "somewhere". It doesn't have a nice equation, but you can write ##q=-2p-2r,## plug it into ##p^2+q^2+r^2=1## and get the equation of the ellipse, which you can normalize or not.
 
Last edited:
  • Like
Likes songoku
  • #3
I am really sorry for late reply.

fresh_42 said:
No. I can't find a mistake, except that I have no idea what ##\vec{u}=\vec{i}+\vec{j}## is since your coordinates seem to be ##x,y,z## and not ##\vec{i},\vec{j},\vec{k},## and why you did not calculate ##\nabla_{(1,2,1)}(\omega )\cdot \vec{u}## as requested.
I have calculated ##\nabla_{(1,2,1)}(\omega )\cdot \vec{u}##

fresh_42 said:
Finally, you want to consider only unit vectors (why?).
Because the formula involves the dot product with unit vector. I am thinking the final answer would be something like ##k <p, q, r> ## (all vectors parallel to unit vector ##<p, q, r>## but I can't get that form.

fresh_42 said:
This means to intersect the plane with the unit sphere. The result will be an ellipse somewhere in space. Unfortunately, the emphasis is on "somewhere". It doesn't have a nice equation, but you can write ##q=-2p-2r,## plug it into ##p^2+q^2+r^2=1## and get the equation of the ellipse, which you can normalize or not.
$$p^2+q^2+r^2=1$$
$$p^2+(-2p-2r)^2+r^2=1$$
$$p^2+4p^2+8pr+4r^2=1$$
$$5p^2+8pr+4r^2=1$$

Then I don't know what to do.

Is it possible the answer in the form like:
$$v=
\begin{pmatrix}
p \\
q \\
r
\end{pmatrix}
$$
$$=
\begin{pmatrix}
p \\
-2p-2r \\
r
\end{pmatrix}
$$
$$=p
\begin{pmatrix}
1 \\
-2 \\
0
\end{pmatrix}
+ r
\begin{pmatrix}
0 \\
-2 \\
1
\end{pmatrix}
$$

Thanks
 
  • #4
songoku said:
I am really sorry for late reply.I have calculated ##\nabla_{(1,2,1)}(\omega )\cdot \vec{u}##Because the formula involves the dot product with unit vector. I am thinking the final answer would be something like ##k <p, q, r> ## (all vectors parallel to unit vector ##<p, q, r>## but I can't get that form.$$p^2+q^2+r^2=1$$
$$p^2+(-2p-2r)^2+r^2=1$$
$$p^2+4p^2+8pr+4r^2=1$$
$$5p^2+8pr+4r^2=1$$

Then I don't know what to do.

Is it possible the answer in the form like:
$$v=
\begin{pmatrix}
p \\
q \\
r
\end{pmatrix}
$$
$$=
\begin{pmatrix}
p \\
-2p-2r \\
r
\end{pmatrix}
$$
$$=p
\begin{pmatrix}
1 \\
-2 \\
0
\end{pmatrix}
+ r
\begin{pmatrix}
0 \\
-2 \\
1
\end{pmatrix}
$$

Thanks
I would say yes. But it might be a matter of convention. There are a few ways to describe a plane and I can't know whether you are expected to use any or a specific one. The same with an ellipse. It looks like
1698509228861.png


and you can calculate a rotation (coordinate transformation) so that it looks like a standrad ellipse. The new coordinates would be ##p'=\sqrt{5}p+4r/\sqrt{5}## and ##r'=2r/\sqrt{5}## if I made no mistake.
 
  • #5
fresh_42 said:
I would say yes. But it might be a matter of convention. There are a few ways to describe a plane and I can't know whether you are expected to use any or a specific one.
Now that you said it, I just realised that the answer is equation of plane passing through origin.

fresh_42 said:
and you can calculate a rotation (coordinate transformation) so that it looks like a standrad ellipse. The new coordinates would be ##p'=\sqrt{5}p+4r/\sqrt{5}## and ##r'=2r/\sqrt{5}## if I made no mistake.
Are there any resources to learn about rotation for ellipse and other geometrical shapes?

Thanks
 
  • #6
I would search for "conic sections" + pdf or look it up in Wikipedia, e.g. https://en.wikipedia.org/wiki/Parabola or more general https://en.wikipedia.org/wiki/Conic_section. Here is a good article https://arxiv.org/pdf/1205.5935.pdf but I'm not sure if it fits you. It is a bit "dry".

The equation of your plane through the origin was correct. It is a parametric representation, mine was a geometric representation by its normal vector. There is no better or worse, only different equations with the same solution space.
 
  • Like
Likes songoku
  • #7
Thank you very much for the help and explanation fresh_42
 
  • Like
Likes fresh_42

FAQ: Direction in which directional derivative is zero

What is a directional derivative?

A directional derivative is a measure of how a function changes as you move in a specific direction from a given point. It generalizes the concept of a partial derivative by considering the rate of change in any direction, not just along the coordinate axes.

When is the directional derivative zero?

The directional derivative of a function at a point is zero in a direction if the function does not change as you move in that direction from the point. This typically occurs when the direction is orthogonal (perpendicular) to the gradient vector of the function at that point.

How do you find the direction in which the directional derivative is zero?

To find the direction in which the directional derivative is zero, you need to determine the direction that is perpendicular to the gradient vector of the function at the given point. Mathematically, this involves solving for the direction vector that satisfies the condition of being orthogonal to the gradient vector.

What is the gradient vector and how is it related to the directional derivative?

The gradient vector of a function at a point is a vector that points in the direction of the steepest increase of the function and whose magnitude represents the rate of increase. The directional derivative in any direction is the dot product of the gradient vector and the unit vector in that direction. Hence, the directional derivative is zero when the direction vector is orthogonal to the gradient vector.

Can the directional derivative be zero in more than one direction?

Yes, the directional derivative can be zero in multiple directions. In fact, it will be zero in any direction that lies in the plane orthogonal to the gradient vector at the given point. For a function of two variables, this means any direction in the line perpendicular to the gradient vector. For higher dimensions, it means any direction in the hyperplane orthogonal to the gradient vector.

Similar threads

Replies
6
Views
832
Replies
3
Views
1K
Replies
4
Views
1K
Replies
3
Views
1K
Replies
9
Views
1K
Replies
6
Views
2K
Replies
4
Views
1K
Back
Top