This is a note written to address an issue we had in Tueday’s tutorial. Specifically the method I have shown you for doing one of the types of differentiability questions is somewhat flawed. There are a number of further assumptions that we need to make in order to make the analysis correct. I apologise for this oversight. However all is not lost – we can still ‘fix’ our (easier) method by taking these additional assumptions into account.
This question will now not be examinable in your summer exam.
The Question
Let be defined by:
Show that is differentiable but not twice differentiable.
The Flawed* Solution
Away from ,
is a polynomial and hence differentiable with derivative:
Now
.
and
.
Hence exists as the left and right derivatives are equal, hence
is differentiable.
Away from ,
is a polynomial and hence twice differentiable:
Now
and
Hence does not exist hence
is not twice differentiable
* this method leads to the correct conclusion, and it will do so for all cases we would encounter in MS 2001.
Essentially the error that was made by me was confusion in the definition of the left- and right-hand derivatives. Let and
.
Left- and right-hand Derivatives: Correct Definition
Right-hand Derivative at :
.
Left-hand Derivative at :
.
Left- and right-hand Derivatives: Incorrect Definition
Right-hand Derivative at :
.
Left-hand Derivative at :
.
Theorem
Let and
.
is differentiable at
if and only if the left- and right-hand derivatives exist and are equal.
Towards a Correct Analysis
The above flawed solution more-or-less looks at the function and says well it’s smooth there because the derivatives are matching up on both sides – i.e as a function, is continuous so
must be differentiable (click to sharpen):
A plot of .
We will see later that by this flawed analysis a discontinuous function can be differentiable with continuous derivative (!).
The Correct Solution
Away from ,
is a polynomial and hence differentiable with derivative:
Now
.
and
.
Hence exists as the left- and right-hand derivatives are equal, hence
is differentiable.
Away from ,
is a polynomial and hence twice differentiable:
Now
and
Hence does not exist hence
is not twice differentiable
What is wrong with the “flawed solution”?
Two counterexamples show that our original analysis is flawed (the second is not as important to us but is relevant in a subsequent proof):
Example 1:
Let
A plot of
Now geometrically it is clear that is does not have a well-defined tangent at
. Even more straightforwardly, we can see that
is not continuous at
and recalling the following theorem (and it’s contrapositive):
differentiable continuous
not continuous not differentiable,
It is clear that is not differentiable at
yet
thus,
and so our flawed analysis would suggest that is differentiable at
(and indeed infinitely differentiable). Our analysis is suggesting that the discontinuous function
is differentiable with continuous derivative (!).
Hence if the ‘flawed method’ is to work it will be necessary to show that the function is continuous.
Example 2
Let
Now away from ,
has derivative
but at
the tangent is vertical – hence the slope of the tangent is undefined (
undefined). Indeed plugging in
we get:
Uh-oh! However
which could lead to the nonsensical conclusion that is differentiable at
with derivative
.
Hence if the ‘flawed method’ is to work it will be necessary to show that the derivative is bounded (i.e. not infinite).
Theorem
Suppose that is continuous on
and differentiable except perhaps at
. Suppose further that
exists. Then
is differentiable at
with derivative
.
Proof: Let .
is continuous on
and differentiable on
. Then, by the Mean Value Theorem,
such that
Now taking the limit as on both sides:
Now as
hence the left hand side is equal to
:
Similarly taking and
we can show:
Hence is differentiable at
with derivative
Leave a comment
Comments feed for this article