Part I · Chapter 5 needs review

Analytic functions

A function is analytic if it equals its Taylor series near every point. This chapter gives examples and non-examples, shows that the decides where a Taylor series converges, and states the property that everything later depends on: an analytic function is determined by any small piece of it.

The definition

A function is analytic at a point cc if its Taylor series about cc converges to the function on some interval around cc. It is analytic on a domain if it is analytic at every point of the domain.

Examples of analytic functions:

The radius of convergence is decided in the complex plane

Take f(x)=11+x2f(x) = \dfrac{1}{1+x^{2}}. Its graph is a . The denominator is never 0 for real xx, so the function has no bad point anywhere on the real line.

Its Taylor series about 0 comes from the geometric series with −x2-x^{2} in place of xx:

11+x2  =  1−x2+x4−x6+⋯for ∣x∣<1\frac{1}{1+x^{2}} \;=\; 1 - x^{2} + x^{4} - x^{6} + \cdots \qquad \text{for } |x| \lt 1

The series stops converging at x=±1x = \pm 1. But at x=1x = 1 the function is 12\tfrac12, and nothing about its graph changes there. On the real line there is no reason for the series to fail.

The reason appears when xx is allowed to be complex. A power series only uses addition and multiplication, so it makes sense for a zz in place of xx. And for complex zz the function does have bad points:

1+z2=0whenz=i  or  z=−i1 + z^{2} = 0 \quad\text{when}\quad z = i \ \text{ or } \ z = -i Both points are at distance 1 from the center 0.

That distance is the radius of convergence. The rule is general:

The Taylor series of an analytic function about a center cc converges inside the largest disc around cc that contains no bad point, and diverges outside that disc. So the radius RR is the distance from cc to the nearest bad point, measured in the complex plane.

Test the rule by moving the center. The distance from a real number cc to ii is c2+1\sqrt{c^{2}+1}, so the series of 1/(1+x2)1/(1+x^{2}) about cc should have

R  =  c2+1R \;=\; \sqrt{c^{2}+1} About c=0c = 0 that is 1. About c=1c = 1 it is 2≈1.41\sqrt2 \approx 1.41, so the series converges for −0.41<x<2.41-0.41 \lt x \lt 2.41. About c=2c = 2 it is 5≈2.24\sqrt5 \approx 2.24.

Lab 5.1 the Taylor series of 1/(1 + x²) about a center c, and the disc that limits it
● the bad points i and −i● the center cthe largest disc around c with no bad point inside
1/(1 + x²) on the real lineits Taylor polynomial of degree N about cdashed lines: where the disc meets the real line
Top: the complex plane. The horizontal axis is the real line; the two white dots are the bad points. The circle passes through them. Bottom: the real line only. The polynomial matches the function between the dashed lines, which are exactly where the circle above crosses the real axis.

Three things follow.

Functions that are not analytic

Examples of functions that are not analytic:

Analytic functions are rigid

For functions in general, the values on one interval say nothing about the values on another. The function gg above agrees with the zero function on the whole left half of the line and differs from it on the right. For analytic functions this cannot happen:

If two analytic functions agree on an interval, however short, they agree everywhere that both are defined (on a connected domain). So a function given on a small region has at most one analytic extension to a larger one.

The reason: if two analytic functions agree near a point cc, they have the same derivatives at cc, so the same Taylor series about cc, so they agree on the whole interval where that series converges. Move the center toward the end of that interval and repeat.

A familiar special case: a polynomial of degree 2 is determined everywhere by its values at 3 points. An analytic function behaves like a polynomial of infinite degree.

Chapter 6 uses this to extend a function beyond the region where it was first defined.

Things to try

  • In Lab 5.1, leave c=0c = 0 and raise the degree. The polynomial improves between −1 and 1 and gets worse outside.
  • Slide the center to c=2c = 2. The circle grows until it touches the bad points again, and the interval on the real line widens to match.
  • Put the center at c=−3c = -3. The series now converges at x=−6x = -6, far outside the original interval, and still fails beyond the dashed lines.

Exercises

  1. Find the Taylor series of 1/(1+x2)1/(1+x^{2}) about 0 and its radius of convergence.
    Solution

    Put −x2-x^{2} in place of xx in the geometric series: 11+x2  =  1−x2+x4−x6+⋯\frac{1}{1+x^{2}} \;=\; 1 - x^{2} + x^{4} - x^{6} + \cdots valid when ∣x2∣<1|x^{2}| \lt 1, that is, ∣x∣<1|x| \lt 1. So R=1R = 1, the distance from 0 to the points ±i\pm i where the denominator vanishes.

  2. What is the radius of convergence of the Taylor series of 1/(1+x2)1/(1+x^{2}) about c=3c = 3, and for which real xx does it converge?
    Solution

    The nearest bad points are ±i\pm i, at distance 32+1=10≈3.16\sqrt{3^{2}+1} = \sqrt{10} \approx 3.16. The series converges for 3−3.16<x<3+3.163 - 3.16 \lt x \lt 3 + 3.16, that is, −0.16<x<6.16-0.16 \lt x \lt 6.16.

  3. The function ∣x∣|x| agrees with the function xx on the interval (0,1)(0, 1) but not at x=−1x = -1. Why does this not contradict the rigidity of analytic functions?
    Solution

    ∣x∣|x| is not analytic at x=0x = 0: it has a corner there and no Taylor series. Rigidity is a property of analytic functions only. On any domain where ∣x∣|x| is analytic, such as x>0x \gt 0, it does equal xx.

Check your understanding

  1. The function 1/(1+x2)1/(1+x^{2}) is smooth at every real xx. Why does its Taylor series about 0 stop converging at x=±1x = \pm 1?

    • Because the function is not analytic at x=1x = 1
    • The denominator is 0 at x=±ix = \pm i, and a Taylor series converges only out to the nearest bad point in the complex plane
    • Because the series was computed incorrectly

    The function is analytic at every real point, including ±1\pm 1. The obstruction is invisible on the real line. This is the main reason the natural setting for analytic functions is the complex plane.

  2. The function gg (0 for x≤0x \le 0, e−1/xe^{-1/x} for x>0x \gt 0) agrees with the zero function for all x≤0x \le 0 but is not zero. Why does this not contradict rigidity?

    • Because the two functions agree only at one point
    • It does contradict it
    • gg is not analytic at 0, and rigidity is a statement about analytic functions only

    Its Taylor series at 0 is identically 0 and does not equal gg to the right of 0. Smooth functions can be glued together like this; analytic ones cannot.

  3. Two analytic functions, both defined on the whole real line, agree for 0<x<0.0010 \lt x \lt 0.001. What follows?

    • They agree for every real xx
    • They agree only on that interval
    • Nothing; more information is needed

    Agreement near a point gives the same derivatives there, hence the same Taylor series, and re-centering carries the agreement along the whole line.