Real analytic functions
Let converge for . Define
has to be less than or equal to the radius of convergence. Functions which are locally (i.e, on some interval) defined by a convergent power series are called analytic functions. is a special kind of analytic function, since its power series at a single point converges on its entire domain, and hence is defined on its entire domain by the power series at that point, namely . In this lecture, we will only be working with such functions. In general, however, such a point may not exist. Here’s the general definition:
Definition 154.1.
d04d92A function is analytic if for each , there exists a power series
and a such that if , then the series converges and
We will soon show, in Thm 6, that functions like satisfy Def 1.
Let be as in (E1). Natural questions:
- Is continuous? Differentiable? Integrable?
- What functions are real analytic? In the above example, we started with a convergent series and defined a function. Can we go the other way around, i.e, start with a function, and express it as a convergent power series centered at any point in its domain?
More questions
At this point, we could just make the definitions
We have the tools to say that the above series converge for all , so and are defined for all . We can ask the above questions in particular for these functions.
Theorem 154.2(Rudin (1976) 8.1).
Suppose the series converges for . This allows us to define
which is the limit , where .
Then,
- converges uniformly on for all .
- is continuous and differentiable on , and
Since is also an analytic function, we can turn the crank again to get , which is also analytic, and so on. Thus, any analytic function is infinitely differentiable. This yields an explicit formula for the th derivative of :
Corollary 154.3.
523f3aIf
on , then exists for all , and is given by
The derivatives at are of particular interest:
Remark 154.4.
f331cbThis tells us that if a function is analytic, it must have a unique power series representation! The series
Is called the Taylor series for at . Furthermore, the series from Def 1 must be the Taylor series for at :
Additionally, we only need the derivatives at a single point to construct this power series about that point.
Remark 154.5.
3fdabeAlthough a function may have derivatives of all orders (i.e, be smooth), the series , where is computed by (E2), need not converge to for any . Since if had a power series representation, it would have to be , this tells us that is not analytic on any neighborhood of . For example, consider
for all . However, is not constant in any neighborhood of .
Having shown that an analytic function on is infinitely differentiable on , we can calculate all derivatives at any point . So, we can calculate the Taylor polynomial of at to any degree. But do these converge? If they do, do they converge to ? We will see that this is indeed the case.
Theorem 154.6(Taylor, Rudin (1976) 8.4).
34b9d2Suppose
If , then can be expressed as a power series about the point which converges in , and
Misc
Suppose converges for . If the series also converges at an endpoint, say , then is continuous not only in , but also at . We take for simplicity:
Proposition 154.7(Rudin (1976) 8.2).
Suppose converges. Put
Then,