|
|
| Haga clic aquí para ver esta página en español. |
The Notorious Jumping Function of Continuum County
Submitted by Deb from Rochester, NY, 2/26/2000.
Original answer and this article by Allen Stenger.
Give an example of monotonic function on
[0,1]
which is discontinuous at all rationals.
Hint 1
Sneak up on it a few discontinuities at a time.
Hint 2
We can enumerate the rationals in the interval
[0,1]
in groups by their denominators as
\begin{array}{lccccccc}
1: & 0/1 &&&&&& 1/1 \\
2: & &&& 1/2 &&& \\
3: & && 1/3 && 2/3 && \\
4: & & 1/4 && 2/4 && 3/4 &
\end{array}
and so on. For each group, invent a monotonic function that is discontinuous
at each point of the group. Then what?
Hint 3
Step functions are the simplest monotonic functions with discontinuities;
let's keep it simple. We can make a function discontinuous on the
nth
group by
f_n(x) = \frac{k}{n} \mathrm{\ for \ } \frac{k}{n} < x \le \frac{k+1}{n}.
Here are graphs of
f_2,
f_3, and
f_4 :
Now how can we leverage these simple functions to get
one
function that is discontinuous at
all
rationals?
Hint 4
One way to combine functions is to add them together; we can't quite do that here
because the sum won't converge, but we can weight the functions and use
f(x) = \sum_{n=1}^\infty \frac{f_n(x)}{2^n}.
Prove this is well-defined, monotonic, and discontinuous at each rational.
Click here for rest of the solution.
The Rest of the Solution
The sum function is:
- Well-defined: We have
0 \le f_n(x) \le 1
and therefore the sum converges by comparison with the geometric series
\sum_{n=1}^\infty \frac{1}{2^n}
- Monotonic: if a < b then
f(a) = \sum_{n=1}^\infty \frac{f_n(a)}{2^n} \le \sum_{n=1}^\infty \frac{f_n(b)}{2^n} =f(b)
- Discontinuous at each rational:
if y > \frac{k}{m} then
f(y) - f\left( \frac{k}{m} \right)
= \sum_{n=1}^\infty \frac{f_n(y) - f_n(k/m)}{2^n}
\ge \frac{f_m(y) - f_m(k/m)}{2^m}
\ge \frac{1}{2^m} \; \frac{1}{m}
An Everywhere-Continuous, Nowhere-Differentiable Function
We know that if a function has a derivative at a point, then it is
continuous at that point.
The converse is not true:
The absolute value function is continuous at 0, but does not have
a derivative at 0, because it has a "corner" at 0.
Can there be a function that is continuous at all points but
differentiable at no point?
Yes, there can!
We can exploit the same idea we used in the original problem to define a
continuous function that is not differentiable anywhere.
The first such function was invented by Karl Weierstrass around 1861, and
the following example was invented by John McCarthy in 1953.
Define
g(x) =
\begin{cases}
1+x & \text{for $-2 \le x \le 0$} \\
1-x & \text{for $0 \le x \le 2$}
\end{cases}
and periodically outside this interval.

Then define
f(x) = \sum_{n=1}^\infty 2^{-n} g\left(2^{2^n} x\right).
See if you can discover how to prove it is nowhere differentiable.
References
- John McCarthy, "An Everywhere Continuous Nowhere Differentiable Function,"
American Mathematical Monthly, vol. 60 (1953), p. 709.
Reprinted in
A Century of Calculus,
Part I, ed. Tom M. Apostol et al., Mathematical Association of America, 1969, p.156.
Proof that the example given above is nowhere differentiable.
Also available on the web
here.
- E. C. Titchmarsh,
The Theory of Functions,
2nd edition, Oxford University Press, 1939.
Section 11.22 gives Weierstrass's original example of a nowhere differentiable
function, and Section 11.23 gives a third example due to van der Waerden.
- Click
here to view the original problem submitted by Deb.
|