Let be open, is an arbitrary point, and is a function. Recall that is called differentiable at if there is a linear map, denoted by , such that
The linear map is called derivative of at , denoted by . Notice that in the nominator of the above quotient, the symbol is simply the absolute value function. But this is no longer true for higher-order derivatives that we are going to define.
The following theorem is well-known.
Theorem 1 (1st order differentiability). The function is differentiable at if all partial derivatives exist in a neighborhood of and are continuous at .
When exists, we must have
In this note, we want to extend the above theorem for higher-order derivatives. To be more precise, we prove the following
Theorem 2 (2nd order differentiability). The function is twice differentiable at if all partial derivatives exist in a neighborhood of and are continuous at .
When exists, we must have
which is the Hessian matrix of at . The preceding formula suggests the relation between and for all .
Recursively, higher-order derivatives of , denoted by can be defined similarly. To be more precise, and for simplicity, let us treat the case . We say that the function is twice differentiable at if there is a bilinear map such that
Such a bilinear map is called the derivative at . Notice that if is a bilinear map, then is a linear map. To prove Theorem 2, we need the following lemma.
I. A CALCULUS LEMMA
The key ingredient of the proof of Theorem 2 is the following:
Lemma. exists if and only if all derivatives exist.
To prove the lemma, we first need some preparation. Obviously, the question is how to estimate . Recall that if is a linear map, then its norm is
Hence as a linear map, we know that
II. PROOF OF THE LEMMA: THE DIRECTION
Suppose that exists. By definition, this is a bilinear map. For clarity, let us denote by . In terms of we obtain
As a linear map, we have
with , thanks to . We now show that every exists. The idea is to show that
Indeed, take with
to get
Hence
which implies
Thus we must have
which immediately implies that exists for any .
III. PROOF OF THE LEMMA: THE DIRECTION
By definition, we must have
for each . Keep in mind that each is a vector, namely
Let us construct the following bilinear map
namely, with , we have
Making use of and with , we can verify
thanks to and the existence of for all . Thus, we have just shown that
proving the twice diffentiablity of .
IV. APPLICATION OF THE LEMMA: PROOF OF THEOREM 2
We are now in position to prove our main result. For arbitrary but fixed, because all second order partial derivatives with exist in a neighborhood of and are continuous at , we know that is differentiable at ; see Theorem 1. Now we apply the above lemma to conclude that is twice differentiable at .
Leave a comment