SOME OPEN PROBLEMS Math118, O. Knill WAVE FRONTS AND CAUSTICS ON THE TORUS.

Math118, O. Knill WAVE FRONTS AND CAUSTICS ON THE TORUS. Is it true that for a general metric on
the two dimensional torus (a bumpy doughnut), the wave front of a point becomes dense on
ABSTRACT. We summarize some open problems in dynamics. Most have been mentioned the torus? Is it true that for a nonflat metric, the caustic of a point is dense on the torus?
during this course. One possibility for a project is to write down a short essay about such a
problem. Discussion. This seems completely unexplored and might be not too difficult. It would
be enough to verify that the wavefront Kt contains longer and longer pieces with shrinking
STABILITY OF EXTERIOR BILLIARDS. Does there exist a billiard table, for which the curvature.
exterior billiard features an unbounded orbit?
HILBERTS 16th PROBLEM. Find upper bounds on the number of limit cycles for a differential
Discussion. A semicircle had been mentioned as a candidate for unstability. One doesn’t equation on the plane. A concrete problem: verify that a differential equation
know the answer also for general polygonal billiards. Smooth strictly convex tables produce
d
bounded orbits (KAM). Some ”rational polygons” produce bounded orbits too. x = p(x, y)
dt
d
INTEGRABLE EXTERIOR BILLIARDS. Does there exist an integrable smooth convex y = q(x, y)
exterior billiard map different from the ellipse? Integrability means that there should exist a dt
smooth function F for which each set {F = c} outside the table is a curve. with polynomials p and q of degree n has only finitely many limit cycles. Find a bound for
their number H(n).
Discussion. The problem could be changed by allowing F to be continuous only or requiring
F to be realanalytic. For some polygons like the sqare, the billiard is integrable. Discussion: this is an old and probably very difficult problem.

MEASURE OF PERIODIC POINTS OF BILLIARDS. Can a smooth billiard table have a set CELLULAR AUTOMATA. Find closed formulas for the topological entropy of each of the 256
of periodic points which have positive area? elementary CA in one dimension.

Discussion. It is known that periodic points of period 2 or 3 have zero area. See Rychliks Discussion: there are some automata, for which we know the answer, like for rule 90, the
paper of 1998. shift. For subshifts of finite type, the entropy is log(κ), where κ is the largest eigenvalue of an
integer matrix.
MEASURE OF PERIODIC POINTS OF EXTERIOR BILLIARDS. Can a smooth strictly
convex exterior billiard table have a set of periodic points which have positive area? SPEED SPECTRUM OF CA. Given a CA in the plane like life, one can define the speed
spectrum as the set of possible speeds, gliders can have. If T n (x) = σ m (x), then the velocity
Discussion. I have not seen that question asked and the answer could be easier to find than is v = m/n and the speed |m|/v. The speed spectrum of a cellular automaton is a discrete
in the billiard case. Some polygons like the square have a full set of periodic orbits. subset of [0, d], where d is the diameter of the CA rule. Find a higher dimensional automaton,
where the speed spectrum can be proven to be dense in [0, d].
MEASURE OF NONCOLLISION SINGULARITIES. what is the measure of noncollision
singularities of the Newtonian n-body problem. Discussion: I don’t think, this is known in the specific example of ”Life”. There are many
gliders known. ”Life” should be rich enough to have a desne speed spectrum.
Discussion. This is an old problem appearing for example in Simons list of unsolved prob-
lems for the 21’st century: The difficulty to figure out all possible configurations leading to ENTROPY OF STANDARD MAP. Does the Standard map have a positive average Lyapunov
noncollision singularities suggest that a completely different approach is necessary. exponent for large λ. More specifically, is it true that for all n,
INTEGRABIILITY AND DYNAMICAL LOGARITHM PROBLEM. Does nonintegrability (in 1
Z 2π Z 2π
log ||A(T n−1 (x, y) · · · A(x, y)|| dxdy ≥ log(c/2)
the sense that no smooth invariant function F exists) imply that the dynamical logarithm n 0 0
problem can not be solved efficiently? Specifically, for the nonintegrable map
" # " #
where T (x, y) = (2x + c sin(x) − y, x) and A(x, y) = dT (x, y).
x x + 2α
T = . Discussion. The problem had been posed in the sixties by Sinai. One knows that the Lyapunov
y x+y
exponent is positive on horse-shoes which form Cantor sets, but these sets have zero measure.
on the torus. Assume α = π, find n such that T n (0.5, 0.5) is within distance 10−1000 of (0, 0).
BOUNDED ORBITS IN SITNIKOV PROBLEM. What is the measure of the set of initial
LATTICE POINTS NEAR PARABOLA. For every 0 ≤ δ < 1, there exists a positive constant conditions for which the Sitnikov planet stays bounded?
C such that the number M (n, δ) of 1/n-lattice points in a 1/n1+δ neighborhood of a parabola
satisfies M (n, δ)/n1−δ → C for n → ∞. Discussion: there is a continuum of bounded solutions constructed by the horse-shoe con-
struction. The problem is already difficult for maps in the plane, which are explicitely given
Discussion: we know that the result is true for δ < 1/3. We suspect it is true for δ < 1. like the area preserving Henon map.
STABILITY OF SOLAR SYSTEM An ancient problem. Is the solar system stable? More
(
x/2, x even
precisely: is the n-body problem with the initial condition given by the position leading to a 3N+1 PROBLEM. Is 1 → 4 → 2 the only attractor of the Collatz map T (x) = .
3x + 1, x odd
solution in which orbits stay in a bounded region for all times? How big is the set of initial
conditions, for which one has stability? Discussion. The mathematical tools seem not to catch. The best hope for success is probably
to find a periodic orbit different from the trivial one. Heuristic arguments show however that
Discussion. One knows that the measure of initial conditions leading to stability is positive, the ”chance” for success is small. As higher periodic orbits you look for, they are more and
but one knows no estimates for the size of the stability region. more ”unlikely”.

NONCOLLISION SINGULARITIES WITH 4 BODIES. Does Gervers construction lead to A LATTICE POINT PROBLEM. The map T (x) = (3/2)x mod 1 is related to a lattice point
noncollision singularity with four bodies? problem for a function on the real line. It is conjectured that the orbits of T are uniformly
distributed modulo 1.
Discussion. Gervers construction works in the plane. Xias construction needed a lot of
mathematical resources. It is likely that also this problem is very difficult. Discussion. For T (x) = 2x mod 1, the distribution is uniform for most initial points x.

CONSTRUCTING LATTICE POINTS CLOSE TO PARABOLA. Can one solve the dynami- HOMOCLINIC POINTS AND INTEGRABILITY. Assume that a smooth map T on the
cal logarithm problem efficiently for for maps obtained from polynomials. plane has a hyperbolic fixed point with a transverse homoclinic point. Prove that there is no
smooth function F which is invariant for which {F = c} is either a curve or a finite set.
Discussion: if it were, one could factor integers fast.
Discussion. This might be relatively easy to settle. The existence of a transverse homoclinic
point produces a horse shoe as we have seen. The function is constant on that horse shoe as
LYAPUNOV EXPONENT FOR DOUBLE PENDULUM. Is the Lyapunov exponent positive
for the double pendulum (equal mass and equal length of the legs). The same question can be well as on the union of the invariant stable and unstable manifolds. While the horse shoe has
dimension bigger than 0, it could be part of a complicated level set which is a union of arcs.
asked for other Hamiltonian systems for which the flow leaves a bounded region invariant like
But unlike realanalytic functions, smooth function can have complicated level sets.
the Henon-Heils system.

Discussion. There are many Hamiltonian systems of two degree of freedom, where chaos is BLANCHARD PROBLEM. Does every transitive automaton have a dense set of periodic
observed. While it is often possible to construct a Cantor set on which the dynamics is chaotic, points?
the question is whether one can establish this kind of motion on a set of positive measure.
Discussion: An automaton T is called transitive, if it has a dense orbit in X. We have seen
that the shift is transitive. We also have seen that the shift has a dense set of periodic points.
LOCAL CONNECTIVITY OF MANDELBROT SET. is the Mandelbrot set locally connected?
Francois Blanchard writes: ”The answer, positive or negative, is a necessary step before on
Discussion. This is the holy grail in complex dynamics. understands the meaning of chaos in the field. ” Source: This problem can be found in Michael
Misiurewicz list of open problems in dynamical systems (http://www.math.iupui.edu/ mmisi-
ure/open)
CHAOTIC SMOOTH CONVEX BILLIARD. Find a smooth convex billiard table for which
the Lyapunov exponent is positive on a set of positive area.
COHEN MAP. Does the hyperbolic point of period 14 found by Hubbard in the map
Discussion. One could try to smoothen out the stadium but any of these attempts produces √
T (x, y) = ( x2 + 1 − y, x)
stable periodic orbits which distroy ergodicity. Also invariant curves near the boundary will
prevent ergodicity. One has to prove coexistence of stable and unstable behavior. have stable and unstable manifolds which intersect transversly?

CHAOTIC THREE BODY PROBLEM. Is the Lyapunov exponent positive on a set of positive Discussion: I had been asked as an undergraduate to make experiments with this map to
area for some restricted three body problem. see whether it is integrable or not - and did not find any signs of non-integrability. Marek
Rychlik told me in 1998, that numerical experiments by John Hubbard revealed a hyperbolic
Discussion. By constructing horse shoes, one can get Cantor sets of zero area, for which the periodic orbit of period 14: (x, y) = (u, u) with u = 1.54871181145059. The largest eigenvalue
Lypapunov exponent is positive. of dT 14 (x, y) is λ = 1.012275907. The existence of a hyperbolic point of such a period makes
integrability unlikely since homoclinic points might exist, but it is not impossible. It is difficult
NORMALITY OF PI. Are the digits of Pi uniformly distributed?
√ In other words, is π normal. to find other hyperbolic periodic points. An other indication for non-integrability is a result
The same question can be answered for other numbers like 2. of Rychlik and Torgenson who have shown that this map has no integral given by algebraic
functions.
Discussion. Some doubts have been raised that π produces good random numbers:
http://news.uns.purdue.edu/UNS/html4ever/2005/050426.Fischbach.pi.html