Sunday Function

A tricky one:

i-d4bb9fe09d6b78eb832985821d8a307f-1.png

At x = 2, f(x) = 1 + 2 + 4 + 8 + ...

Which, uh, is clearly not going to end anywhere finite. The series fails the single most basic necessary (but not sufficient) condition for convergence of infinite sums - that the terms of the series approach zero as you go farther and farther up the chain of addition.

On the other hand, f(x) is equally clearly not divergent for all x.

For instance, at x = 0, f(x) = 1 + 0 + 0 + ... = 1

So that's a start. Maybe there's other numbers x can equal where the series has a finite value. What are they? This is one of those problems where it's easier if you already know something extra about the question. Take a look at this function:

i-88516ca505d84f969b7ab15a965c3fb3-2.png

If you find the series expansion of that function around x = 0, you'll get precisely the f(x) we posed earlier. But there's a complication. While our neat fraction is perfectly well defined at x = 2, we already saw that our series entirely failed to converge there. What gives?

The answer lies in the one place our fraction isn't defined - where x = 1, the denominator is zero and it's impossible to divide by zero. And here's the catch. If you create a series expansion of a function around a particular point (here we chose x = 0) that expansion will converge as it should right up until you hit the place where the function isn't analytic. Ours has such a point at x = 1, and so the series will fail to converge there or anywhere past it. Further, that distance between our series start and the bad point serves to define a radius of convergence. Not only will our series not work for x equal to or greater than 1, it won't work for x equal to or less than -1. In the complex plane, it won't work outside the circle centered at the origin.

We can if we want create another series expansion on the other side of the discontinuity, but of course the same problem would still exist. It wouldn't work outside the radius defined by the distance of our expansion point from the point x = 1.

Here, for comparison, are the graphs of the fractional version of our function and the series version of our function graphed on a domain running from -1.5 to 1.5. The fractional version of our function is in purple, the series (the first eleven terms) is in blue.

i-7b7142f1efa09b47c3b8242f89bcb9f4-graph.png

This particular kind of series is called the geometric series, and find application not just in physics but also in finance, engineering, and pure mathematics.

More like this

If you're a regular reader of this site, you might remember a post about this fascinating specimen from the collection of unusual functions. I'm only showing it on the interval [-1,1] for reasons that will become apparent, but outside that region the growth tapers off rapidly and the function…
Note to the reader: this post is relatively stiff mathematically. For those not mathematically inclined, I think you might enjoy reading it anyway and enjoying it as you would a tour of a widget factory; even if you're not worried about the details of the nuts and bolts, it's fun to see it done.…
Before her career took an unfortunate wrong turn, a young and talented Lindsay Lohan gave us a charming and popular comedy called Mean Girls. Time has been good to the careers of some of the others involved, Tiny Fey and Rachel McAdams perhaps most notably. But the film did something that very…
"So," Herr Schrodinger says to us, "I'm looking for a function of x. It needs to be equal to the negative of its second derivative, up to a constant factor. When x is zero, the function itself should be zero. And when x equals L for some constant L greater than zero, the function also needs to…

f(0)=1+0+0+...=1

By Anonymous (not verified) on 01 Mar 2009 #permalink

This is also a very good example to give people an idea of what happens with analytic continuation. In essence, we have a function f(x)=1+x+x^2+x^3... that only makes sense for |x|<1. But we can find a function that agrees with f everywhere that f is defined (that is the function 1/(1-x)). And this new function is defined everywhere but a few pesky points (in this case the singularity at x=1). This is at some level the same idea of what we are doing when we find the analytic continuations of things like the Riemann zeta function and the gamma function.

Ray Kurzweil, Vernor Vinge, et al. wax eloquet and voluminous about the Singularity! - social and technological convergence beyond conceptualization by past civilizations (including this one),

http://en.wikipedia.org/wiki/File:ParadigmShiftsFrr15Events.svg

First derivatives sell books, second derivatives sell guns. Know whether extrema are peaks (Manhattan Project) or troughs (NINJA loans). As your plot shows, an epsilon skew makes all the difference in the world.

ok, don't know why the rest of my post got cut off. Trying that again. Our function only makes sense |x| less than 1 (this is true even if we look at complex valued x). But we have this other function 1/(1-x) that agrees with f everywhere they are both defined. This second function is defined and infinitely differentiable everywhere by x=1. Thus, we have a continuation of x to the entire complex plane. This is very similar to the process that allows us to get the analytic continuations of the Riemann zeta function even though its series is only defined for x with the real part of x greater than 1.

Just to be difficult, 1+2+...=-1 is no problem if you use the right norm. Specfically, if you define a "size" of rational numbers so that |2^a b| is 2^-a when b is prime to 2, you get a perfectly reasonable metric on the rational numbers. In this metric that series converges nicely, and to -1. Of course, when you complete the number system, you get something different from the real numbers - in fact its topology looks rather like a Cantor set. But this object, the set of 2-adic numbers, turns out to be a very useful object in number theory - for example, "closeness" means congruence modulo a high power of two.

Neat post. I always enjoy the Sunday function. This graph is hard to read for us color blind folks, though. Those 2 colors are indistinguishable. Different weights, or dashed and solid lines would be much easier to read.

Why no proof that the series converges to 1/(1-x) when |x| < 1 and diverges when |x| >= 1?

There is one more (mildly) interesting thing about the function.

y = 1 + x + x^2, is a parabola which intersects the point (0, 1).

y = 1 + x + x^2 + x^3, is a curve with a point of inflection at (0, 1).

If you continue to add the next term and then plot, each curve ending with an even power of x resembles the first curve as a distorted parabola intersecting (0, 1) and y has no negative values. Each curve ending with an odd power of x resembles the second curve and has the same point of inflection and y has both positive and negative values.

By Joe Doakes (not verified) on 07 Mar 2009 #permalink