Differentiation of y=xx


This is a merger of a couple of threads on the same subject.
By Sarah Shales on December 22, 1997 :

Can one differentiate xx (x to the power of x)

Thank you

Sarah Shales


By Stuart :

Hi:

You can. You'll need the chain rule (for differentiating a function of a function)

Sometimes things like ab look much easier if you take logs:

y = xx
=> log y = x log(x)


So, using the chain rule on the LHS:
(1/y) (dy/dx) = log(x) + x(1/x)

Thus
dy/dx = xx (log(x) + 1)


Hope this helps,
Stuart


By Anonymous on Friday, September 29, 2000 - 12:20 am :

I'm not too great with the binomial theorem, so the differential of

y=xx

has been puzzling me all day.

What is the differential, and how is it found?


By Dan Goodman (Dfmg2) on Friday, September 29, 2000 - 12:23 pm :

The way to do it is to write it in a way that you know the answer.

You know that the differential of ef(x) is f'(x)ef(x) (that's the chain rule).
Also, xx =(eln(x) )x =ex.ln(x) , so now you need to work out the differential of x.ln(x).

Write back again if you still can't get it.


By Anonymous on Saturday, September 30, 2000 - 08:21 pm :

I know that the differential of x.lnx is 1+lnx, but Im having trouble proving this.


By Dan Goodman (Dfmg2) on Sunday, October 1, 2000 - 12:57 am :

Do you know the product rule, that if f(x)=u(x)v(x), then f'(x)=u'(x)v(x)+u(x)v'(x). So, if you put u=x and v=ln(x), you should get the answer. Also, as you probably know, the differential of ln(x) is 1/x.

Does that answer your question? I can prove the results above if you want (e.g. chain rule, product rule, (lnx)'=1/x, etc.)


By Anonymous on Sunday, October 1, 2000 - 08:25 pm :

I know how to prove all the results you mentioned spare the product rule. How does the proof go?


By Dan Goodman (Dfmg2) on Monday, October 2, 2000 - 01:58 am :

Well, this is an almost proof; f+df=(u+du)(v+dv)=uv+v.du+u.dv+du.dv=f+v.du+u.dv+du.dv (as f=uv).
Subtract f from both sides to get df=u.dv+v.du+du.dv, divide by dx to get df/dx=u.dv/dx+v.du/dx+du.dv/dx, the last term is du.dv/dx which is 0, sort of.
Therefore df/dx=u.dv/dx+v.du/dx, or f'=uv'+vu'.
Tada!


[More on this here . - The Editor]


By Michael Doré P(904) on Monday, October 2, 2000 - 03:10 pm :

Alternatively, use logarithms:

y = uv

where y,u,v are functions of x.

ln y = ln(u) + ln(v)

Differentiate:

1/y dy/dx = 1/u du/dx + 1/v dv/dx

(Using the chain rule and assuming that ln(x) differentiates to 1/x. Both are easy to prove.)

dy/dx = y/u du/dx + y/v dv/dx

dy/dx = v du/dx + u dv/dx. So:

d(uv)/dx = v du/dx + u dv/dx.


By Dan Goodman (Dfmg2) on Monday, October 2, 2000 - 04:41 pm :

That's a better proof, thanks Michael.