Differentiablilty of composition functions

79 Views Asked by At

Two questions I suppose. One comes from a test I recently took that I didn't quite get/understand the method I should be using (or even how I should proceed)

Let $f:R^2 \rightarrow R$ s.t $f$ is an element of $C^2$(i.e. continuous first/second partials). Let $h(s,t)=(s+t,s-t)$.

a) Show that $F=f∘h$ is differentiable and its first partials are also differentiable. [Note - no problem here. I got this problem correct. Just putting it here for context]

b) Prove that $D_{1,2}(f∘h)=D_{1,1}(f∘h)-D_{2,2}(f∘h)$

On b I guess I'm just not sure how to proceed.

What I tried to do via playing around with it(though I think I'm wrong about my attempts to think about it):

$f(h(s,t)=f(s+t,s-t)$

$D_1(f∘h)= D_1f*Dh_1+D_2f*D_1h_2 = D_1f+D_2f$

$D_{1,1}(f∘h)=...$

I wrote something for this but it was pointed out that it had to be wrong...I'm not sure if I can evaluate $D_1f+D_2f$ further but someone made me think so. So atm, I'm just confused on how to proceed. [I decided to make something separate for the other question so nevermind on the second question].

EDIT: With a prompt elsewhere, I may have found a way forward that I wasn't considering. The test question asked me for $D_{1,2}(f∘h)=D_{1,1}(f∘h)-D_{2,2}(f∘h)$

My initial thought was to calculate $D_{1,1}(f∘h)-D_{2,2}(f∘h)$ and set it equal to the LHS to tease the proof out by some difference. Instead I went ahead with a direct route:

$D_{2}F=D_{1}fD_{2}h+D_{2}fD_{2}=D_{1}f-D_{2}f$

Then:

$D_{1,2}F=(D_{1,1}f-D_{2,1}h)+(D_{2,1}f-D_{2,2}f$)

Since F is an element of C^2(R^2) => $D_{1,2}f=D_{2,1}f$

Then $D_{1,2}F=D_{1,1}-D_{2,2}f$

which implies:

$D_{1,2}(f∘h)=D_{1,1}(f∘h)-D_{2,2}(f∘h)$

So, I may or may not have it figured out... depending if there is a mistake in my logic.