This paper establishes a lower bound on the maximum distance among three points in an affine lattice that are near a helix with small curvature and torsion, advancing understanding of lattice-helix interactions.
Contribution
It provides a new lower bound for the maximum distance between three lattice points close to a small-curvature helix, a novel geometric insight.
Findings
01
Lower bound for maximum distance among three lattice points near a helix
02
Insights into lattice configurations close to curved structures
03
Advancement in geometric lattice analysis
Abstract
We obtain lower bound for the maximum distance between any three distinct points in an affine lattice which are close to a helix with small curvature and torsion.
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.
Taxonomy
TopicsFinite Group Theory Research · Graph theory and applications · Mathematical Approximation and Integration
We obtain lower bound for the maximum distance between any three distinct points in an affine lattice which are close to a helix with small curvature and torsion.
1 Introduction
The problem of estimating the number of lattice points on or close to a curve has a rich history. In 1926,
Jarník [7] proved that
the number of integer
points on a strictly convex closed curve of length L>3 does not exceed
3(2π)−1/3L2/3+O(L1/3) and the exponent and the
constant of the leading term are best possible. Assuming higher order smoothness conditions on the curve, a number of authors achieved sharper estimates, in particular Swinnerton-Dyer [9] and Bombieri and Pila [1].
Estimating the number of lattice points close to a curve is more recent topic. Of a survey of results on this topic and their applications one may see [3] and [4].
In 1972 Zygmund published a paper on spherical summability of Fourier series in two dimensions where an essential component was the following theorem.
Theorem 1**.**
(Schinzel) An arc of length 32R1/3 on a circle of radius R contains no more than two lattice points.
The original proof is due to Schinzel but the following short proof was provided by Pelczynski.
Proof.
Let A(x1,y1), B(x2,y2), and C(x3,y3) be distinct lattice points on a circle of radius R.
Denote by a, b, and c the lengths of the sides of △ABC, and its area by S.
Since the circle is strictly convex curve, we have
S=\frac{1}{2}|\left|\begin{array}[]{ccc}x_{1}&y_{1}&1\\
x_{2}&y_{2}&1\\
x_{3}&y_{3}&1\end{array}\right||\ \geq{1\over 2}.
On the other hand, by a formula attributed to Heron of Alexandria, we have S=4Rabc.
Thus, abc≥2R.
∎
In the paper [5] Howard and the second author generalized Schinzel’s result to the case of plane curve with bounded curvature and lattice points close to a plane affine lattice (where plane affine lattice is the set
{v0+mv1+nv2:m∈Z,n∈Z,} with
v0,v1,v2, and plane vectors in such that v1 and v2 are linearly independent).
There are only a few papers on the topic of lattice points close to a three-dimensional curve.
Huang [6] obtains estimates for the number of lattice points close dilations of a curve with parametrization (x,f1(x),f2(x)),
x∈[a,b]. Huang’s estimates depend on the dilation parameter q, the upper bound δ for the distance between the lattice points and the curve, and a constant depending on the functions f1 and f2 which is not explicitly computed.
In Chapter 4 of his PhD dissertation Letendre also obtains estimates for the number of lattice points close dilations of a curve with parametrization (x,f1(x),f2(x)),
x∈[a,b]. His results only assume bounds on certain quantities involving derivatives of f1 and f2.
We will consider the special case when the curve is a helix and obtain lower bound for the maximal distance between any three distinct points in a general lattice close to a helix.
We define a lattice in three-space as follows.
Definition 2**.**
Let v0,v1,v2,v3∈R2 be vectors in R3 with v1 , v2, and v3
linearly independent. Then, the affine lattice generated by v1, v2, and v3 with origin v0 is
[TABLE]
We will show in Section 2 that for each lattice L there exist positive constants DL and AL (depending on L) such that if A0 and A1 are two distinct lattice points in
L then ∣A0A1∣≥DL; furthermore, if A0, A1, and A2 are three non-collinear lattice points in L, then Area(△A0A1A2)≥AL.
Our main result is
Theorem 3**.**
Let H be a helix of curvature κ>0 and torsion τ>0, and let
[TABLE]
Let L be affine lattice, and let A0, A1, and A2 be three distinct points in L which are within δ of the helix H. Then,
the maximal distance between A0, A1, and A2 is at least
[TABLE]
The above theorem is a generalization of Schinzel’s result to the case of a helix.
Corollary 4**.**
Let H be a helix of curvature κ>0 and torsion τ>0 with τκ31≥0.4(κ2+τ2). Let A0, A1, and A2 be three distinct lattice points on H.
Then,
the maximal distance between A0, A1, and A2 is at least 1.1κ−31.
The paper is organized as follows. In Section 2 we prove some auxiliary lemmas. In Section 3 we prove the main result.
We expect that an analogue of our main result will hold for curves such that 0<c1K<κ(s)<c2K and 0<c3T<τ(s)<c4T where c1,c2,c3, and c4 are positive constants.
2 Some auxiliary results
In this section we prove several lemmas which will be needed in the proof of Theorem 3.
Lemma 5**.**
The area of a triangle in 3-space with non-collinear lattice point vertices is at least 21.
Proof.
Let A,B, and C be non-collinear lattice points in R3. Then, the area of △ABC equals 21AB×AC.
Since A, B, and C are lattice points, the components of the vectors AB, AC
are integers. Therefore, the components of the cross product AB×AC are integers, as well. Thus, Area(△ABC)=21n for some integer n. Since the points
A, B, and C are non-collinear, Area(△ABC)=0, so n=0. ∎
The next two lemmas where proved in [5] in the R2 case. The lemmas are easily extended to the n-dimensional case.
Lemma 6**.**
Let A, B, C, and C1 be points in Rn where n≥2. Let δ≥0 and ∣CC1∣≤δ. Denote by S the area of △ABC, and by S1 the area of △ABC1. Then,
[TABLE]
Proof.
Let h be the distance from the point C to the line AB, and let h1 be the distance from the point C1 to the line AB. Then, S=2∣AB∣h and S1=2∣AB∣h1.
By the triangle inequality,
[TABLE]
Therefore, ∣h−h1∣≤δ, so ∣S−S1∣=2∣AB∣∣h−h1∣≤2δ∣AB∣.
∎
Lemma 7**.**
Let △ABC and △A1B1C1 be triangles in Rn where n≥2, with areas S and S3, respectively.
Let δ≥0 and assume
[TABLE]
Then,
[TABLE]
Proof.
Let S1 be the area of △ABC1, and let S2 be the area of △AB1C1. Then
[TABLE]
By Lemma 6, ∣S−S1∣≤2δ∣AB∣. Applying Lemma 6 to the points A, C1, B, and B1, we get ∣S1−S2∣≤2δ∣AC1∣.
Finally, applying Lemma 6 to the points B1, C1, A, and A1, we get ∣S2−S3∣≤2δ∣B1C1∣.
Therefore,
[TABLE]
Note that the the triangle inequality gives us ∣AC1∣≤∣AC∣+∣CC1∣≤∣AC∣+δ and ∣B1C1∣≤∣B1B∣+∣BC∣+∣CC1∣≤∣BC∣+2δ.
Using the upper bounds for ∣AC1∣ and ∣B1C1∣ in (1) gives us the desired bound for ∣S−S3∣.
∎
We also need the following simple lemma.
Lemma 8**.**
For all x in the interval (0,π), we have 0<sinx−xcosx<3x3. Also, for all x in the interval (0,π/2], we have sinx≥π2x.
Proof.
Let f(x)=sinx−xcosx. Note f(0)=0. Consider f′(x)=xsinx. Clearly, f′(x)>0 on (0,π) and we have the first half of the inequality.
Now let g(x)=sinx−xcosx−3x3. Note g′(x)=x(sinx−x). If x>0, sinx−x<0 and thus g′(x)<0 on this interval also. Thus, because g(0)=0 and g is decreasing on (0,π), we have the second desired inequality.
To prove the last inequality of the lemma, consider the function h(x)=xsinx. We have h′(x)=x2xcosx−sinx. We proved above that sinx−xcosx>0 for x∈(0,π). Thus h(x) is decreasing in
(0,π/2] and its minimum is h(π/2)=2/π.
∎
Here, and throughout the end of the paper, when A is point and v is a vector, the equation A=v will mean OA=v, where O is the origin of the coordinate system.
A helix is a space curve with constant curvature and torsion.
Let A0<A1,A2 be three points on a helix H. In the next lemma we obtain formulas for the distances ∣A0A1, ∣A1A2∣, ∣A0A2∣, and for the area of triangle
△A0A1A2.
Lemma 9**.**
Let H be a helix with curvature κ>0 and torsion τ>0.
Let A0, A1, and A2 be three distinct points on the helix H whose arclengths from the origin of the helix are s0, s1, s2 respectively, with
s0<s1<s2. Define a=κ2+τ2κ, b=κ2+τ2τ, ti=a2+b2si for i=0,1,2, h1=t1−t0, and h2=t2−t1. Then,
(a) ∣A0A1∣2=4a2sin2(2h1)+b2h12, ∣A1A2∣2=4a2sin2(2h2)+b2h22, and
∣A0A2∣2=4a2sin2(2h1+h2)+b2(h1+h2)2;
(b) (Area(△A0A1A2))2=(T1+T2+T3)/4, where
T1:=16a4sin22h1sin22h2sin22h1+h2,
T2:=a2b2h12h22(2h2sin2h2−2h1sin2h1)2, and
T3:=16a2b2h1h2sin2h1sin2h2sin24h1+h2.
Proof.
First, we consider a helix H(a,b) with parametrization
[TABLE]
It is easy to check that we have κ=a2+b2a, τ=a2+b2b, and
κ2+τ2=a2+b21. From here, we get a=κ2+τ2κ, b=κ2+τ2τ. Also, for the natural parameter s we have
s=ta2+b2.
We have
[TABLE]
with t0<t1<t2.
Therefore,
[TABLE]
Therefore,
[TABLE]
One can obtain the formulas for ∣A1A2∣ and ∣A0A2∣ similarly.
For the area of △A0A1A2 we use
[TABLE]
where θ is the angle between A1A0 and A1A2. Next we calculate the dot product:
[TABLE]
Thus, we have
[TABLE]
We proved the lemma in the case of a helix H(a,b) with parametrization 2.
Now, consider a helix H′ in general position, with curvature κ>0 and torsion τ>0.
Let A0, A1, and A2 be three distinct points on the helix H′ whose arclengths from the origin of the helix are s0, s1, s2 respectively, with
0<s0<s1<s2.
Here we use the uniqueness part of the Fundamental Theorem of the local theory of curves (for example, see do Carmo [2] p.19). It states that if α,α′:I→R are regular parametrized curves with natural parameter s,
which have the same curvature κ(s) and the same torsion τ(s) for all s, then α′ differs from α by a rigid motion, that is there exist an orthogonal linear map ρ of R3, with positive determinant, and a vector c such that
α′=ρ∘α+c.
Since the helix H′ and the helix H(a,b) have the same curvature κ and the same torsion τ, there exists a rigid motion which maps the helix H′ onto the helix H(a,b);
and it maps the points A0,A1,A2 to some points A0′,A1′,A2′ respectively, which are on H(a,b) and have corresponding values of the natural parameter s0,s1,s2.
Note that the rigid motions do not change the distance between two points, and preserve the area of a triangle. Therefore, ∣A0A1∣=∣A0′A1′∣, ∣A1A2∣=∣A1′A2′∣, ∣A0A2∣=∣A0′A2′∣, Area(△A0A1A2)= Area
(△A0′A1′A2′). Moreover, a,b, h1, and h2 are uniquely determined by s0, s1, s2, κ and τ. Therefore, the lemma holds in the case of a helix H′ in general position.
∎
It is easy to estimate the terms T1 and T3. To estimate T2 we use the following lemma.
Lemma 10**.**
Let h1 and h2 be real numbers, both in the interval (0,2π). Then,
[TABLE]
Proof.
Applying the mean value theorem to the function xsinx, we have that there exists ζ between 2h1 and 2h2 such that
[TABLE]
By Lemma 8, we get that 0<x2sinx−xcosx<3x for every x∈(0,π). Therefore,
[TABLE]
∎
In the next two lemmas we prove the existence of the positive constants DL and AL mentioned in the introduction.
Lemma 11**.**
Let v1, v2, and v3 be linearly independent vectors in R3. Then, there exists a constant c>0 (depending on v1, v2, and v3) such that
∣∣m1v1+m2v2+m3v3∣∣≥c for any integers m1, m2, and m3 with (m1,m2,m3)=(0,0,0).
Proof.
We have ∣∣m1v1+m2v2+m3v3∣∣2=(m1v1+m2v2+m3v3)⋅(m1v1+m2v2+m3v3)=i=1∑3j=1∑3(vi⋅vj)mimj:=Q(m1,m2,m3).
The quadratic form Q(m1,m2,m3) is positive definite, since v1, v2, and v3 are linearly independent. Denote by M(Q) the matrix of the quadratic form Q(m1,m2,m3). Since M(Q) is symmetric and Q(m1,m2,m3) is positive definite, the eigenvalues of Q(M) are real positive numbers, say 0<λ1≤λ2≤λ3. Moreover, by the Spectral Theorem for real symmetric matrices, there exists an orthonormal basis of vectors w1,w2,w3 for
R3, where w1, w2, and w3 are eigenvectors corresponding to the eigenvalues λ1, λ2, and λ3. This implies that the minimum of Q(m1,m2,m3) on the unit sphere is λ1.
Therefore, the lemma holds with c=λ1.
∎
Lemma 12**.**
Consider a lattice L=L(v0,v1,v2)={v0+mv1+nv2+pv3:m,n,p∈Z}. There exist positive constants DL and AL (depending on L) such that
(i) if A0 and A1 are any two distinct lattice points in
L, then ∣A0A1∣≥DL;
and
(ii) if A0, A1, and A2 are any three non-collinear lattice points in L, then Area(△A0A1A2)≥AL.
Proof.
If A0 and A1 are two distinct lattice points in
L, then A0A1=m′v1+n′v2+p′v3 with m′, n′, p′ integers such that (m′,n′,p′)=(0,0,0). Now, (i) follows from Lemma 11.
Next, let A0, A1, and A2 be three non-collinear lattice points in L. We have that Area(△A0A1A2)=∣∣A0A1×A0A2∣∣/2.
Since A0, A1, and A2 are non-collinear, then Area(△A0A1A2)=0. Therefore, A0A1×A0A2=0.
Furthermore, since A0A1=m′v1+n′v2+p′v3 and A0A2=m′′v1+n′′v2+p′′v3 for some integers m′,m′′,n′,n′′,p′, and p′′, then
[TABLE]
with q=m′n′′−m′n′′, r=n′p′′−n′′p′, and s=m′p′′−m′′p′.
Since, A0A1×A0A2=0 we have (q,r,s)=(0,0,0).
Moreover, the vectors v1×v2, v2×v3, and v1×v3 are linearly independent.
Indeed, if c1(v1×v2)+c2(v2×v3)+c3(v1×v3)=0, then after taking the dot product of the last equation with v3, we get
c1((v1×v2)⋅v3)=0. However, the triple product (v1×v2)⋅v3=0, since the vectors v1, v2, and v3 are linearly independent.
We show similarly that c2=0 and c3=0.
Thus, the vectors v1×v2, v2×v3, and v1×v3 are linearly independent.
First, we consider the case when the lattice points are on the helix.
Theorem 13**.**
Let H be a helix with curvature κ>0 and torsion τ>0. Let L=L(v0,v1,v2,v3) be an affine lattice in R3, and
let A0, A1, and A2 be three distinct points in L which are also on the helix H and whose arclengths from the origin of the helix are s0, s1, s2 respectively, with
s0<s1<s2. Then ∣A0A2∣≥min(κ2+τ2πτ,1.5AL31κ−31) and the length of the arc \stackrel{{\scriptstyle\mbox{\large\frown}}}{{A_{0}A_{2}}} is at least
min(κ2+τ2π,2.4AL31κ−31).
Proof.
Define
a=κ2+τ2κ and b=κ2+τ2τ. Recall that
s=a2+b2t. Let ti=a2+b2si, for i=0,1,2. Then 0<t0<t1<t2 and
Ai=r(ti) for i=0,1,2. Let h1=t1−t0 and h2=t2−t1.
We consider two cases.
Case I:h1+h2≥π.
Then the arclength of the arc \stackrel{{\scriptstyle\mbox{\large\frown}}}{{A_{0}A_{2}}} is
Note that one cannot expect to get a nontrivial lower bound on ∣A0A2∣ depending only on κ. For example, consider the case when a is a large integer, b=2π1, t0=0, t1=2π, and t2=4π.
Case II:h1+h2<π.
Denote by S the area of △A0A1A2. From Lemma 9 (ii) we have 4S2=T1+T2+T3, where
T1=16a4sin22h1sin22h2sin22h1+h2,
T2=a2b2h12h22(2h2sin2h2−2h1sin2h1)2, and
T3=16a2b2h1h2sin2h1sin2h2sin24h1+h2.
Since 0<h1+h2<π, we have T1>0, T2≥0, and T3>0. Therefore, S>0, so the points A0, A1, and A2 are non-collinear. Now,
by Lemma 12 we have S≥AL>0. We obtain
[TABLE]
Next we get bounds for T1, T2, and T3.
Since 0<h1+h2 and sinx<x when x>0 we obtain
[TABLE]
(Here and below, we use the inequality ab≤(a+b)2/4.)
First, by Lemma 5 in the case of the standard lattice in R3, AL=21. Therefore, the condition τκ31≥0.4(κ2+τ2) implies
κ2+τ2πτ>1.5AL31κ−31 in the case of standard lattice L. The corollary follows by
noting that 321.5>1.1.
∎
Let A0′, A1′, and A2′ be three distinct lattice points
within δ of the helix H. Therefore, there exist points A0, A1, and A2 on the helix, such that
[TABLE]
Since A0′, A1′, and A2′ are in L, by Lemma 12 we have ∣A0′A1′∣≥DL, ∣A0′A2′∣≥DL, and ∣A1′A2′∣≥DL.
Since δ≤DL/4, by the triangle inequality, we get
[TABLE]
Similarly, ∣A0A1∣≥DL/2 and
∣A1A2∣≥DL/2
Thus, the points A0, A1, and A2 are distinct. Let the values of the natural parameter corresponding to A0, A1, and A2 be s0, s1, and s2, respectively.
Without loss of generality we can assume s0<s1<s2 (otherwise we relabel A0′, A1′, and A2′). As before, let ti=si(a2+b2)−21
for i=0,1,2, h1=t1−t0>0 and h2=t2−t1>0.
We consider two cases.
Case I: h1+h2≥π.
In this case, in exactly the same way as in the proof of Theorem 13, we obtain
[TABLE]
Now, by the triangle inequality, we get
[TABLE]
so
[TABLE]
Case II. h1+h2≤π.
Here, first we show that A0′, A1′ and A2′ are not on a straight line.
Recall that ∣A0A1∣≥DL/2. Therefore, s1−s0≥∣A0A1∣≥DL/2. Since s1−s0=h1(a2+b2)21, we get h1≥2DL(a2+b2)−21.
Similarly, s2−s1≥DL/2 and
h2≥2DL(a2+b2)−21.
Denote the area of △A0A1A2 by S.
By Lemma 9 (ii), we have 4S2=T1+T2+T3, where T1=16a4sin22h1sin22h2sin22h1+h2, T2≥0, and T3>0.
which contradicts the upper bound for δ, so our assumption is false. Hence, the points
A0′, A1′ and A2′are not on a straight line. By Lemma 12S3, the area of
△A0′A1′A2′ is at least AL.
Next, we get a lower bound for the area of △A0A1A2. Using (11) we obtain
[TABLE]
Now, ∣A0A1∣+∣A1A2∣+∣A0A2∣<2(s2−s0) and we showed above that δ≤(s2−s0)/4.
Therefore,
[TABLE]
We have
s2−s0=2(h1+h2)(a2+b2)21≤2π(a2+b2)21.
We obtain,
[TABLE]
Now, by the upper bound on δ, we have
[TABLE]
Therefore,
[TABLE]
Now, we get lower bound for h1+h2 the same way we did in the case of lattice points on the helix, in Case 2 of the proof of Theorem 13. The upper bounds of T1, T2, and T3 are the same as before, and the estimates
(4), (5), and (6) still hold.
The only difference is that the lower bound for the area of △A0A1A2 now is AL/2 rather than AL.
Since π3.8>1.2, and ∣A0′A2′∣≥∣A0A2∣−2δ we obtain the stated bound for ∣A0′A2′∣.
∎
Bibliography9
The reference list from the paper itself. Each links out to its DOI / PubMed record.
1[1] E. Bombieri and J. Pila, The number of integral points on arcs and ovals , Duke Math. J. 59 (1989), 337–-357.
2[2] Manfredo do Carmo, D ifferential Geometry of Curves and Surfaces, Prentice-Hall, Englewood Cliffs, New Jersey (1976).
3[3] M. Filaseta and O. Trifonov, The distribution of fractional parts with applications to gap results in number theory , Proc. London Math. Soc. 73 (1996), 241–278.
4[4] M. Filaseta, S. Graham, and O. Trifonov, Starting with gaps between k 𝑘 k -free numbers , Int. J. Number Theory 11 (2015), 1411–1435.
5[5] R. Howard and O. Trifonov, Bounding the number of lattice points close to a curve by curvature and arclength , Functiones et Approximatio (to apper), ar Xiv:2207.09532.
6[6] Jing-Jing Huang, Integral points close to a space curve , Mathematische Annalen 374 (2019), 1987–2003.
7[7] V. Jarník, Über bedingt konvergente Reihen , Math. Z. 24 (1926), 715–732.
8[8] P. Letendre, Topics in analytic number theory , Ph.D. thesis, Universite Laval, (2018).