Directional Derivatives and max rate of change

Join the discussion
Registration is free. Start your own thread to ask a follow-up.
3 replies · 5K views
Blkmage
Messages
11
Reaction score
0

Homework Statement



See attachment.

2. Homework Equations /solution attempt

Part (a)
Well, the gradient evaluated at (1,2-1) will give the rate of change. If we want the maximum rate of change then we need the directional direction such that the unit vector [tex]\mathbf{u}[/tex] is in the same direction as [tex]\nabla f(1,2,-1)[/tex]. So the unit vector would just be [tex]\frac{\nabla f(1,2,-1)}{\text{norm}(\nabla f(1,2,-1))}[/tex]? And the max rate of change is just the direction derivative at this point and direction?

Anyways, part (b) is what I'm having trouble with:

Since we want it to be in the direction of the vector (1,-1,-1), does that mean that we want [tex]\nabla g(x_{Q},y_{Q},z_{Q}) = (1,-1,-1)[/tex]? Since (1,-1,-1) isn't a unit vector, I didn't divide by the norm of the gradient...
 
Attachments
  • Screen shot 2011-04-26 at 7.52.34 PM.png
    Screen shot 2011-04-26 at 7.52.34 PM.png
    14.6 KB · Views: 572
Physics news on Phys.org
the gradient gives the direction of maximum rate of change, its magnitude is the rate of change moving along that direction

the magnitude of the gradient is equivalent to a dot product with a unit vector in the same direction, ie the value of the directional derivative in that direction
 
So for part (b), we have

[tex]\text{norm}(\nabla g(x_{Q},y_{Q},z_{Q})) = \nabla g(x_{Q},y_{Q},z_{Q}) \cdot \frac{1}{\sqrt{3}}\left(1,-1,-1\right)[/tex]

This seems so tedious to solve though. It's on a practice exam my teacher gave us and his tests are very reasonable given the time limits, so I'm guessing there is something I'm missing?

edit: not to mention that would give me a single equation of 3 variables, which I wouldn't be able to solve
 
[tex]g(x,y,z) = x^2yz[/tex]
[tex]\nabla g(x,y,z) = (2xyz, x^2z, x^2y)[/tex]

and you want to know when is in the direction (1,-1,-1), first clearly this will not be possible if any of the varibales are 0 as it will lead tio .

substituting in gives
2xyz = -x^2z _______(1)
2xyz = -x^2y _______(2)
x^2z = x^2y _______(3)

so you know y=z from (3), as x dne 0, which should simplify things