1. May 6, 2017

### vishnu 73

<Moderator's note: moved from a technical forum, so homework template missing.>

so this is the question.
i want to know if there is a solution without using calculus maybe trig substitution maybe other methods?
i tried trig substitition
i let u = √2 cosx
and v be sinx
am i on the right track

Last edited by a moderator: May 8, 2017
2. May 6, 2017

### Staff: Mentor

You need two variables - with only x you cannot cover the whole plane you have to search.

3. May 8, 2017

### vishnu 73

ok fine let
v be siny
but this only makes the problem complicated how am i supposed to minimum value of function with different variables

4. May 8, 2017

### Staff: Mentor

It has to be a minimum with respect to both variables. While that is technically not sufficient to have a global minimum, it will do the job here.

5. May 8, 2017

### vishnu 73

so terms containing x should be minimum themselves and terms containing y must be local minimum too?

6. May 8, 2017

### Staff: Mentor

Yes, or the same with u and v without substitution. If a point is not a minimum with respect to these variables, there is a point nearby which has a lower value.

7. May 8, 2017

### Staff: Mentor

Are you allowed to use a calculator??

8. May 9, 2017

no

9. May 9, 2017

### vishnu 73

@mfb
i am just realizing i got back to using calculus and partial derivative i know that method already is there any other

10. May 9, 2017

### Staff: Mentor

Sometimes there are other options. You have the sum of two squares. Both squares are not negative, so if there is a point where both squares are zero, or at least at a global minimum within the allowed parameter range, you found a global minimum. Then you just have to check if it is within the given range of the coordinates.
There is no such point in this problem (unless you allow imaginary arguments).

11. May 10, 2017

### vishnu 73

so is the calculus the only method? without imaginary numbers only real because i have it done the calculus way already

12. May 10, 2017

### Buffu

Did you try AM-GM ?

13. May 12, 2017

### haruspex

One thing that makes it hard is having both variables in both terms. If you were to make a substitution to two new variables, x and y, so that one term only contains x then you can concentrate first on minimising the other term wrt y.
The downside is that checking the given bounds gets awkward.

Another possible start is to consider what happens on the boundaries.

Last edited: May 12, 2017
14. May 13, 2017

### vishnu 73

@Buffu
how to apply am-gm here

@haruspex
yup i am finally seeing that calculus is the most straight forward approach here

15. May 13, 2017

### haruspex

Standard differential calculus approach will lead to quartics to solve.
I think I have the answer, using my first suggestion. Write x+u for v and find the min wrt u. That should give you an expression for v in terms of u. It is rather nasty, but see if you can spot a fairly simple pair of values satisfying it.
Showing that produces the minimum is easier than finding it from calculus.

16. May 13, 2017

### vishnu 73

17. May 13, 2017

### Buffu

$\displaystyle (u - v)^2 + \left(\sqrt{2- u^2} - {9\over v}\right)^2 \ge 2(u-v)\left(\sqrt{2- u^2} - {9\over v}\right)$

Maybe do $v = ku$ now.

I don't think it simplifies the equation much but better than nothing, right ?

18. May 13, 2017

### vishnu 73

okay i will give that a try too give me some time i am now doing another problem

19. May 13, 2017

### haruspex

Very neat, and yes, that is the answer I got.
The method at that link does produce quartics - see the expression with both y and 1/y3 - but it turns out to factorise nicely.

20. May 13, 2017

### vishnu 73

oh if thats what you meant by quartics then yes but anyways i will give your method a try too