Spiral Delone sets and three distance theorem
Shigeki Akiyama

TL;DR
This paper characterizes when a constant angle progression on the Fermat spiral forms a Delone set, linking it to the property of the angle being badly approximable.
Contribution
It establishes a precise condition relating the angle's Diophantine approximation properties to the Delone set formation on the Fermat spiral.
Findings
Constant angle progression forms a Delone set iff the angle is badly approximable.
Provides a characterization connecting number theory and geometric point sets.
Enhances understanding of spatial distributions on spirals.
Abstract
We show that a constant angle progression on the Fermat spiral forms a Delone set if and only if its angle is badly approximable.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Spiral Delone sets and three distance theorem
Shigeki Akiyama
Abstract.
We show that a constant angle progression on the Fermat spiral forms a Delone set if and only if its angle is badly approximable.
In botany, phyllotaxis stands for angular patterns of plant stems or leaves, which tends to be correlated with Fibonacci numbers and the golden mean. In this short note, we study a related fundamental problem on the spiral patterns, which is not addressed before. Let be a subset of which is identified with the complex plane . Denote by the open ball of radius centered at . is -relatively dense if there exists that for any , holds. is -uniformly discrete if there exists that for any we have . is a -Delone set if it is both -relatively dense and -uniformly discrete. It is clear that -relatively dense implies -relatively dense if , and -uniformly discrete implies -uniformly discrete if , and holds for a -Delone set. We omit the prefixes -, - and - if we are only interested in the existence of and/or .
Denote by . Fix an angle and a strictly increasing function from to itself. We study a point set
[TABLE]
on a spiral curve . Clearly is not relatively dense if the angle is rational, since is contained in a union of finite number of lines passing through the origin.
Lemma 1**.**
If is -relatively dense, then . If is -uniformly discrete, then .
Proof.
Assume that is -relatively dense. Then
[TABLE]
Since has cardinality and
[TABLE]
for , we see which implies .
If is -uniformly discrete, then are disjoint disks for . From , we obtain
[TABLE]
which leads to , i.e., for . ∎
Our target is to obtain a condition that is a Delone set. It is of interest to discuss general increasing functions , however, in this paper we specify and is irrational and study the set
[TABLE]
From now on, we assume
[TABLE]
in light of Lemma 1.
In other words, we are interested in the sequence of points on the Fermat spiral that progresses by a constant angle (see Figure 1). A real number is badly approximable if there exists a positive constant so that
[TABLE]
holds for all . It is known that is badly approximable if and only if partial quotients of continued fraction expansion of is bounded by a positive integer (see [5, Theorem 23]). Indeed, and are roughly inverse-proportional;
[TABLE]
c.f. [3, Theorem 1.9]. We use this notation and throughout this paper. In particular, if is a real quadratic irrational, then is badly approximable, due to Lagrange Theorem. In this note we will prove
Theorem 2**.**
The following four statements are equivalent.
- a)
* is relatively dense,* 2. b)
* is uniformly discrete,* 3. c)
* is a Delone set,* 4. d)
the angle is badly approximable.
By definition of a Delone set, the proof is finished when we prove the equivalence between a) and d) in §2, and then the one between b) and d) in §3. We prove them in a quantitative form.
Theorem 3**.**
Let be the above bounds of when is badly approximable.
- a)
If is badly approximable, then is -relatively dense. 2. b)
If is -relatively dense, then is badly approximable with . 3. c)
If is badly approximable, then is -uniformly discrete. 4. d)
If is -uniformly discrete, then is badly approximable with .
Roughly speaking, the bound corresponds to -Delone set where is proportional to and is inverse-proportional to as in (2). As -Delone set is more uniform when is closer to , the case , i.e., is the best choice. This shows an important role of the golden angle observed in many instances in phyllotaxis.
1. Three Distance Theorem
The torus is divided into subintervals by
[TABLE]
Steinhaus conjectured that there are at most three different lengths appear in this partition for any and . This is first proved by Sós [9, 10] and then Świerczkowski [12], Surányi [11], Halton [4] and Slater [8]. We follow a formulation in Alessandri-Berthé [1], see also [2, Chapter 2.6] and [6, Chapter 6.4, p.518]. Let , and
[TABLE]
for with a unique . When is irrational, the sequence is positive, strictly decreasing and given by Euclidean division process of intervals. Concatenating
[TABLE]
the continued fraction expansion of
[TABLE]
is naturally associated with this algorithm. Define and
[TABLE]
for . Then we observe that the convergent is equal to and by induction. We later use an important basic inequality
[TABLE]
see e.g. [7, Chapter 2] ,[5, Theorem 13], [3, Corollary 1.4]. For a positive integer there exists a unique expression:
[TABLE]
with and . Three distance theorem reads
Theorem 4**.**
The torus is subdivided by into intervals of at most three lengths:
- •
* intervals have length ,*
- •
* intervals have length ,*
- •
* intervals have length .*
Rotating all points by in with a fixed , the same statement is valid for . See [9, 10, 12, 11, 4, 8] for the proof of Theorem 4. We just comment that it makes easier to understand them after we comprehend the recursive (dynamical) induction structure of irrational rotation . Inducing an irrational rotation on the interval of length to an interval of length located at the end of in the sign direction, the first return map gives another irrational rotation . As the number of points increases, we inductively observe the induced irrational rotations. Three distance theorem can be easily understood as a consequence of this structure.
Corollary 5**.**
* is badly approximable if and only if there is a positive constant , that for any , the ratios of three distances generated by the points are bounded by .*
Proof.
Clearly is the sum of the other two. We obtain
[TABLE]
and
[TABLE]
On the other hand, for there are exactly two lengths and which appear in the partition. From we see
[TABLE]
Summing up, the bound of partial quotients and satisfy
[TABLE]
∎
2. Diophantine condition for relative denseness
First we give a Diophantine characterization of relatively denseness. The distance of from is designated by . Hereafter we often use
[TABLE]
which follows from .
Lemma 6**.**
If is -relatively dense then for any and , the system of inequalities
[TABLE]
has a solution . Conversely the system of inequalities (5) is solvable, then is -relatively dense.
Proof.
Assume that is -relatively dense. Then for with , we find that . Then
[TABLE]
gives
[TABLE]
On the other hand,
[TABLE]
shows
[TABLE]
Conversely, take a complex number with and assume the system (5) has a solution . Then
[TABLE]
Here we used which follows from . This shows for any point with , the intersection is non empty. For , we have . By (1), is -relatively dense. ∎
Proof of a) and b) of Theorem 3. We show that if is badly approximable with the constant in Corollary 5, then the system of inequalities (5) is solvable. Take in Corollary 5 and fix . For , the closed interval contains integers. By Theorem 4 the points
[TABLE]
gives a partition of into intervals of at most three lengths where
[TABLE]
Every subdivided interval has length not larger than , because otherwise all lengths are larger than which would cause overlapping. Therefore any and there exists such that
[TABLE]
which shows that (5) is solvable, recalling (1). In view of Lemma 6 and (4), to show -relatively denseness of , we need
[TABLE]
and therefore suffices.
Let us assume that is -relatively dense. By Lemma 6, there exists that for any and any , there exists satisfying (5). Our goal is to find an upper bound of . We may assume and select that
[TABLE]
has cardinality with for 111The inequality ensures the existence of not less than . . This is possible since the function
[TABLE]
from to satisfies
[TABLE]
By this choice of , we have
[TABLE]
By Theorem 4, the torus is partitioned into subintervals having three lengths , and by
[TABLE]
From the choice of , Theorem 4 and (3), a subinterval of length
[TABLE]
exists. Letting be the mid point of , we have . Summing up, we obtain
[TABLE]
by (6). This shows under the assumption . Clearly implies . Therefore we obtained . ∎
3. Uniform discreteness of
Finally we prove c) and d) of Theorem 3. If is not -uniformly discrete, then we can find with such that
[TABLE]
Similar computation with Lemma 6 leads to
[TABLE]
Putting , we have . Therefore is not badly approximable by the constant .
Conversely assume is not badly approximable by the constant , i.e., has an integer solution . Then
[TABLE]
Putting and using (1), the right side is estimated by
[TABLE]
This shows that is not -uniformly discrete.
4. Open Questions
It is of interest to generalize this result to a strictly increasing function asymptotically equal to , or to substitute by a uniformly distributed sequence on .
Is there a higher dimensional analogy of this result ? One possibility is to find a transitive action on , so that
[TABLE]
is a Delone set in .
Acknowledgments
The author came across this problem from an inspiring lecture delivered by J. F. Sadoc at RIMS Kyoto in Oct. 2018. He wishes to show his heartfelt gratitude to Y. Yamagishi at Ryukoku Univ, who organized the meeting, for advices, discussion and critical reading of the original manuscript. He is also indebted to J. Lagarias, Y. Bugeaud and A. Haynes for crucial remarks and discussion. This research is partially supported by JSPS grants (17K05159, 17H02849, BBD30028).
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] P. Alessandri and V. Berthé, Three distance theorems and combinatorics on words , Enseign. Math. (2) 44 (1998), no. 1-2, 103–132.
- 2[2] J.P. Allouche and J. Shallit, Automatic sequences , Cambridge University Press, Cambridge, 2003, Theory, applications, generalizations.
- 3[3] Y. Bugeaud, Approximation by algebraic numbers , Cambridge Tracts in Mathematics, vol. 160, Cambridge University Press, Cambridge, 2004.
- 4[4] J. H. Halton, The distribution of the sequence { n ξ } ( n = 0 , 1 , 2 , ⋯ ) 𝑛 𝜉 𝑛 0 1 2 ⋯ \{n\xi\}\,(n=0,\,1,\,2,\,\cdots) , Proc. Cambridge Philos. Soc. 61 (1965), 665–670.
- 5[5] A. Ya. Khinchin, Continued fractions , The University of Chicago Press, Chicago, Ill.-London, 1964.
- 6[6] D. E. Knuth, The art of computer programming, vol 3: Sorting and searching , Addison Wesley, London, 1973.
- 7[7] A. M. Rockett and P. Szüsz, Continued fractions , World Scientific Publishing Co., Inc., River Edge, NJ, 1992.
- 8[8] N. B. Slater, Gaps and steps for the sequence n θ mod 1 𝑛 𝜃 mod 1 n\theta\ {\rm mod}\ 1 , Proc. Cambridge Philos. Soc. 63 (1967), 1115–1123.
