Sufficient condition to apply integration by parts "infinitely many times"

76 Views Asked by At

My question is related to this one. Suppose we are trying to solve the following integral $$ \int f(x) g(x) dx,$$ where we know $f(x)$ is smooth, all of its derivatives are positive, and the sum of its derivatives $\sum_{n=0}^{\infty} f^{(n)}(x)$ converges. For the sake of simplicity, let's say $g(x)=e^x$ for now. Then if we naively apply integration by parts "infinitely times" with $u=f(x)$ and $v'=g(x)=e^x$ we get $$ \sum_{n=0}^{\infty} (-1)^n e^x f^{(n)}(x) = e^x \sum_{n=0}^{\infty} (-1)^n f^{(n)}(x). $$ Since the above series converges (by absolute value test and our assumption), were we justified in performing integration by parts infinitely times? In other words, is the condition that a series of a function's derivatives (which are all positive) converges, sufficient to performing integration by parts infinitely times (with $g(x)=e^x$)?

Now what about for a more general function $g(x)$? What conditions are needed on $g(x)$ to allow infinite applications of integration by parts (with the same assumptions on $f(x)$)? Thanks to all in advance.