Very fractional geometric progressions with an integer ratio

The geometric progression 1/3, 2/3, 4/3, 8/3, 16/3,… is notable for being dyadic (ratio 2) and staying away from integers as much as possible (distance 1/3 between this progression and the set of integers; no other dyadic progression stays further away from integers). This property is occasionally useful: by taking the union of the dyadic partition of an interval with its shift by 1/3, one gets a system of intervals that comfortably covers every point: for every point x and every (small) radius r there is an interval of size r, in which x is near the middle.

Endpoints of the intervals in one group are away from the endpoints of the other

It’s easy to see that for any real number x, the distance between the progression {x, 2x, 4x, 8x, …} and the set of integers cannot be greater than 1/3. Indeed, since the integer part of x does not matter, it suffices to consider x between 0 and 1. The values between 0 and 1/3 lose immediately; the values between 1/3 and 1/2 lose after being multiplied by 2. And since x and 1-x yield the same distance, we are done.

Let’s find the most fractional progressions with other integer ratios r. When r is odd, the solution is obvious: starting with 1/2 keeps all the terms at half-integers, so the distance 1/2 is achieved. When r is even, say r = 2k, the best starting value is x = k/(2k+1), which achieves the distance x since rx = k-x. The values between 0 and k/(2k+1) are obviously worse, and those between k/(2k+1) and 1/2 become worse after being multiplied by r: they are mapped to the interval between k-x and k.

The problem is solved! But what if… x is irrational?

Returning to ratio r=2, it is clear that 1/3 is no longer attainable. The base-2 expansion of x cannot be 010101… as that would be periodic. So it must contain either 00 or 11 somewhere. Either of those will bring a dyadic multiple of x within distance less than 0.001111… (base 2) of an integer, that is distance 1/4.

The goal is to construct x so that its binary expansion is as balanced between 0 and 1 as possible, but without being periodic. The Thue-Morse constant does exactly this. It’s constructed by starting with 0 and then adding the complement of the sequence constructed so far: x = .0 1 10 1001 10010110 … which is approximately 0.412. The closest the dyadic geometric progression starting with x comes to an integer is 2x, which has distance about 0.175. The Wikipedia article links to the survey The ubiquitous Prouhet-Thue-Morse sequence by Allouche and Shallit, in which Corollary 2 implies that no other irrational number has a dyadic progression with a greater distance from integers, provided that this distance is attained. I have not been able to sort out the case in which the distance from a progression to the integers is not attained, but it seems very likely that Thue-Morse remains on top.

What about other ratios? When the ratio r is even, the situation is essentially the same as for r=2, for the following reason. In base r there are two digits nearest (r-1)/2, for example 4 and 5 in base 10. Using these digits in the Thue-Morse sequence, we get a strong candidate for the most fractional progression with ratio r: for example, 0.455454455445… in base 10, with the distance of about 0.445. Using any other digit loses the game at once: for example, having 3 in the decimal expansion implies that some multiple of 10 is within less than 0.39999… =  0.4 of an integer.

When the ratio is odd, there are three digits that could conceivably be used in the extremal x: namely, (r-1)/2 and its two neighbors. If the central digit (r-1)/2 is never used, we are back to the Thue-Morse pattern, such as  x = 0.0220200220020220… in base 3 (an element of the standard Cantor set, by the way). But this is an unspectacular achievement, with the distance of about 0.0852. One can do better by starting with 1/2 = 0.1111111… and sprinkling this ternary expansions with 0s or 2s in some aperiodic way, doing so very infrequently. By making the runs of 1s extremely long, we get the distance arbitrarily close to 1 – 0.2111111… base 3, which is simply 1/2 – 1/3 = 1/6.

So it seems that for irrational geometric progressions with an odd ratio r, the distance to integers can be arbitrarily close to the number 1/2 – 1/r, but there is no progression achieving this value.

The Kolakoski-Cantor set

A 0-1 sequence can be interpreted as a point in the interval [0,1]. But this makes the long-term behavior of the sequence practically invisible due to limited resolution of our screens (and eyes). To make it visible, we can also plot the points obtained by shifting the binary sequence to the left (Bernoulli shift, which also goes by many other names). The resulting orbit  is often dense in the interval, which doesn’t really help us visualize any patterns. But sometimes we get an interesting complex structure.

The Kolakoski-Cantor set, KC

The vertical axis here is the time parameter, the number of dyadic shifts. The 0-1 sequence being visualized is the Kolakoski sequence in its binary form, with 0 and 1 instead of 1 and 2. By definition, the n-th run of equal digits in this sequence has length {x_n+1}. In particular, 000 and 111 never occur, which contributes to the blank spots near 0 and 1.

Although the sequence is not periodic, the set is quite stable in time; it does not make a visible difference whether one plots the first 10,000 shifts, or 10,000,000. The apparent symmetry about 1/2 is related to the open problem of whether the Kolakoski sequence is mirror invariant, meaning that together with any finite word (such as 0010) it also contains its complement (that would be 1101).

There are infinitely many forbidden words apart from 000 and 111 (and the words containing those). For example, 01010 cannot occur because it has 3 consecutive runs of length 1, which implies having 000 elsewhere in the sequence. For the same reason, 001100 is forbidden. This goes on forever: 00100100 is forbidden because it implies having 10101, etc.

The number of distinct words of length n in the Kolakoski sequence is bounded by a power of n (see F. M. Dekking, What is the long range order in the Kolakoski sequence?). Hence, the set pictured above is covered by {O(n^p)} intervals of length {2^{-n}}, which implies it (and even its closure) is zero-dimensional in any fractal sense (has Minkowski dimension 0).

The set KC apparently does not have any isolated points; this is also an open problem, of recurrence (whether every word that appears in the sequence has to appear infinitely many times). Assuming this is so, the closure of the orbit is a totally disconnected compact set without isolated points, i.e., a Cantor set. It is not self-similar (not surprising, given it’s zero-dimensional), but its relation to the Bernoulli shift implies a structure resembling self-similarity:

KC is covered by two copies scaled by 1/2

Applying the transformations {x\mapsto x/2} and {x\mapsto (1+x)/2} yields two disjoint smaller copies that cover the original set, but with some spare parts left. The leftover bits exist because not every word in the sequence can be preceded by both 0 and 1.

KC covered by two copies scaled by 2

Applying the transformations {x\mapsto 2x} and {x\mapsto 2x-1} yields two larger copies that cover the original set. There are no extra parts within the interval [0,1] but there is an overlap between the two copies.

The number {c = \inf KC\approx 0.146778684766479} appears several times in the structure of the set: for instance, the central gap is {((1-c)/2, (1+c)/2)}, the second-largest gap on the left has the left endpoint {(1-c)/4}, etc. The Inverse Symbolic Calculator has not found anything about this number. Its binary expansion begins with 0.001 001 011 001 001 101 001 001 101 100… which one can recognize as the smallest binary number that can be written without doing anything three times in a row. (Can’t have 000; also can’t have 001 three times in a row; and 001 010 is not allowed because it contains 01010, three runs of length 1. Hence, the number begins with 001 001 011.) This number is obviously irrational, but other than that…

In conclusion, the Python code used to plot KC.

import numpy as np
import matplotlib.pyplot as plt
n = 1000000
a = np.zeros(n, dtype=int)
j = 0                  
same = False  
for i in range(1, n):
    if same:
        a[i] = a[i-1]    
        same = False
        a[i] = 1 - a[i-1]
        j += 1            
        same = bool(a[j])
v = np.array([1/2**k for k in range(60, 0, -1)])
b = np.convolve(a, v, mode='valid')
plt.plot(b, np.arange(np.size(b)), '.', ms=2)