By induction, let's prove that (f∘g)(n)(a) is a sum of terms, each of which is a product of terms of the form
c(f(k)∘g)(a)(g(1)(a))d1(g(2)(a))d2⋯(g(n)(a))dn,with c, k, and d1,…,dn being nonnegative integers and k,d1,…,dn≤n.
This is so for the base case n=1 since
(f∘g)′(a)=(f′∘g)(a)g′(a),in which case the only term has c=1, k=1, and d1=1.
For the induction step, suppose the conjecture holds for positive integer n and that both f(n+1)(g(a)) and g(n+1)(a) exist. Then (f∘g)(n+1)(a) is the derivative of a sum of terms previously described. Hence, (f∘g)(n+1)(a) is a sum of terms of the form
c(f(k+1)∘g)(a)(g′)d1+1(g′′)d2⋯(g(n))dnor
c⋅di(f(k)∘g)(a)(g′)d1⋯(g(i))di−1(g(i+1))⋯(g(n))dn,with i≤n. Since (f(k+1)∘g)(a) and g(n+1)(a) are defined, all constituents of the forms above are defined. Hence, (f∘g)(n+1)(a) also is defined.