0

In Amann's textbook Analysis II, the authors present below version of Taylor's theorem.


Let $E,F$ be Banach spaces. Suppose $X$ is open in $E, q \in \mathbb{N}^{\times}$, and $f$ belongs to $C^{q}(X, F)$. We use the following notation: $$ \partial^{k} f(x)[h]^{k}:=\left\{\begin{array}{lr} \partial^{k} f(x)[\underbrace{h, \ldots, h}_{k \text {-times }}], & 1 \leq k \leq q, \\ f(x), & k=0 \end{array}\right. $$ for $x \in X, h \in E$, and $f \in C^{q}(X, F)$. Then $$ f(x+h)=\sum_{k=0}^{q} \frac{1}{k !} \partial^{k} f(x)[h]^{k}+R_{q}(f, x ; h) $$ for $x \in X$ and $h \in E$ such that $[\![ x, x+h ]\!] \subset X$. Here $$ R_{q}(f, x ; h):=\int_{0}^{1} \frac{(1-t)^{q-1}}{(q-1) !}\left[\partial^{q} f(x+t h) \color{blue} {-\partial^{q} f(x)} \right][h]^{q} \mathrm d t \in F. $$


I have come across a version (at page 64 of this note) with below integral remainder, i.e., $$ R_{q}(f, x ; h):=\int_{0}^{1} \frac{(1-t)^{q-1}}{(q-1) !} \partial^{q} f(x+t h) [h]^{q} \mathrm d t. $$

Is it possible to deduce this remainder from the version in my textbook?

Dean Miller
  • 9,566
Analyst
  • 6,559

1 Answers1

1

It's a simple integration by parts: \begin{align} R_{q,\text{book}}(f,x,h)&:=\int_0^1\frac{(1-t)^{q-1}}{(q-1)!}[\partial^qf(x+th)-\partial^qf(x)][h^q]\,dt\\ &=\int_0^1-\left(\frac{d}{dt}\frac{(1-t)^q}{q!}\right)[\partial^qf(x+th)-\partial^qf(x)][h^q]\,dt\\ &=\bigg[-\frac{(1-t)^q}{q!}[\partial^qf(x+th)-\partial^qf(x)][h^q]\bigg]_0^1\\ &- \int_0^1\left(-\frac{(1-t)^q}{q!}\right)\frac{d}{dt}[\partial^qf(x+th)-\partial^qf(x)][h^q]\,dt\tag{IBP}\\\\ &= 0 +\int_0^1\frac{(1-t)^q}{q!}\partial^{q+1}f(x+th)[h^{q+1}]\,dt\\ &=R_{q,\text{notes}}(f,x,h). \end{align}

peek-a-boo
  • 69,172
  • Your IBP is valid if $\partial^{q+1} f$ exists and continuous. However, $f$ is only assumed to be $q$-times continuously differentiable. – Analyst Feb 17 '22 at 14:17
  • @Analyst in the notes, it is assumed $f$ is $n+1$ times continuously differentiable, and I assume $n=q$, so $f$ is $q+1$ times continuously differentiable by hypothesis. – peek-a-boo Feb 17 '22 at 14:18
  • Ah I see. We can not obtain such an integral remainder that only involves $\partial^q f$ if $f$ is only $q$-times continuously differentiable?? – Analyst Feb 17 '22 at 14:21
  • 2
    @Analyst well, the book gives you such a formula (which is actually interesting because I had always thought having an integral form of the remainder requires one extra degree of smoothness), so I'm not sure what else you're looking for. – peek-a-boo Feb 17 '22 at 14:22
  • Thank you so much for your help! – Analyst Feb 17 '22 at 14:22