|


Analysis and the DerivativeDate: 01/23/2001 at 14:19:58 From: Yvonne Subject: Analysis and the Derivative In my analysis class, we have come to the chapter on the derivative. The text suggests considering the following problem. I'm having a little trouble starting and I would greatly appreciate it if you could help me a little. Suppose that f:RtoR is differentiable at c and that f(c)=0. Show that g(x):=|f(x)| is differentiable at c if and only if f'(c)=0. Any suggestions would be terrific. Thanks.
Date: 01/24/2001 at 21:36:14
From: Doctor Fenton
Subject: Re: Analysis and the Derivative
Dear Yvonne,
Thanks for writing to Dr. Math. For this problem, you need to use the
definition of derivative:
a function F(x) is differentiable at x = a if
F(x)-F(a)
lim --------- exists. The value of this limit is F'(a).
x->a x - a
So, you know that (1) f(c)=0 ;
and (2) f is differentiable at x = c, or in other words,
f(x)-f(c)
lim --------- = f'(c).
x->c x - c
I think you'll need the epsilon-delta version of this: (I'll write e
for epsilon and d for delta)
Given e>0, there is a d>0 such that for all x such that
0<|x-c|<d, | f(x)-f(c) |
| ---------- - f'(c)| < e .
| x - c |
To show g(x) is differentiable at x = c, you must show that
g(x)-g(c)
lim --------- exists.
x->c x - c
That is, you must find a number L such that, given any e > 0, you can
find a d>0 such that if x satisfies
0 < |x-c| < d, then
| g(x)-g(c) |
| ---------- - L | < e .
| x - c |
If you look at a function near a point where it crosses the axis, and
if the function has a non-zero slope there (so f'(c) =/= 0, where
=/= means "is not equal to"), then the graph will look like
\ /
/ \ /
------/----- , so g(x) = |f(x)| looks like ---------- ,
/
because taking the absolute value reflects or folds the part of the
graph below the x-axis up into the part above the axis. That
intuitively explains why f'(c) must = 0 in order for |f(x)| to be
differentiable at c. You should be able to see that, in that case,
the derivative of g at c, g'(c), should also be 0, so we expect L=0.
Since f(c)=0, and if we assume that f'(c)=0, then we know that
(putting in 0 for f(c) and f'(c) in the e-d definition above,
Given e>0, there is a d>0 such that for all x such that
0<|x-c|<d, | f(x) |
| ---------- | < e .
| x - c |
We must consider
| g(x)-g(c) |
|---------- - (0) |
| x - c | , since as I indicated, we expect g'(c)=0 .
But
g(c)=|f(c)|
= 0 ,
so this is
| |f(x)| | | f(x) |
|-------- | = |-------|
| x - c | | x - c | ,
but this is exactly the expression above which we know is < e.
So the d you chose for f also works for g.
For the converse, it's easier to use contrapositive reasoning.
Assume that f'(c) =/= 0 .
Then the two-sided limit
f(x)-f(c)
lim --------- = f'(c) exists,
x->c x - c
so in particular, the two one-sided limits
f(x)-f(c) f(x)-f(c)
lim --------- = f'(c) , and lim --------- = f'(c) ,
x->c+ x - c x->c- x - c
or
f(x) f(x)
lim --------- = f'(c) , and lim --------- = f'(c) ,
x->c+ x - c x->c- x - c
since f(c) = 0. But then (remember that g(c)=0)
g(x) |f(x)|
lim --------- = lim --------- ,
x->c+ x - c x->c+ x - c
and
g(x) |f(x)|
lim --------- = lim --------- .
x->c- x - c x->c- x - c
But if f'(c) =/= 0, then it is either positive or negative. Assume it
is positive. Then
f(x)-f(c)
lim --------- = f'(c)
x->c+ x - c
> 0 ,
so all the quotients f(x)/(x-c) must be positive, for x close to c.
The denominator is clearly positive, so the numerator must also be
positive, and f(x)>0 for x>c (i.e. in some small interval c<x<c+d).
Similarly,
f(x)-f(c)
lim --------- = f'(c)
x->c- x - c
> 0 ,
and now the denominators are all negative, so the numerator must also
be negative, for x to the left of c.
But g(x)=|f(x)| is either f(x) (if f(x) is positive), or g(x) is -f(x)
(if f(x) is negative).
g(x) |f(x)| f(x)
lim --------- = lim --------- = lim ---------- = f'(c),
x->c+ x - c x->c+ x - c x->c+ x - c
and
g(x) |f(x)| - f(x)
lim --------- = lim --------- = lim --------- = - f'(c).
x->c- x - c x->c- x - c x->c- x - c
If f'(c) =/= 0, the two one-sided limits are unequal, so the two-sided
limit does not exist.
If you have further questions, please write again.
- Doctor Fenton, The Math Forum
http://mathforum.org/dr.math/
|
Search the Dr. Math Library: |
[Privacy Policy] [Terms of Use]


Ask Dr. MathTM
© 1994-2008 The Math Forum
http://mathforum.org/dr.math/