betaFeedback

Reading v1View latest version

WorksMathematics

Geometric progressions in distance sets in odd dimensions over prime fields

Abstract

For E⊂FpdE\subset\mathbb{F}_p^d, let Δ(E)={Qd(x−y):x,y∈E}Δ(E)=\{Q_d(x-y):x,y\in E\}, where Qd(z)=z12+⋯+zd2Q_d(z)=z_1^2+\cdots+z_d^2. We introduce and study the problem of the existence of nondegenerate three-term geometric progressions in the distance set Δ(E)Δ(E) when dd is odd. If MGP,d(p)M_{GP,d}(p) denotes the largest size of a set whose distance set contains no such progression, then, for every prime p≥7p\geq7 and every integer m≥0m\geq0, pmMGP,1(p)≤MGP,2m+1(p)≤4pmMGP,1(p). p^mM_{GP,1}(p)\leq M_{GP,2m+1}(p) \leq4p^mM_{GP,1}(p). For all sufficiently large primes pp, we also prove pm+726≤MGP,2m+1(p)≤Cpm+1625, p^{m+\frac{7}{26}}\leq M_{GP,2m+1}(p) \leq Cp^{m+\frac{16}{25}}, where CC is an absolute positive constant.

Introduction

Throughout the paper, pp is an odd prime and Fp\mathbb{F}_p is the field with pp elements. For d≥1d \ge1, let

Qd(z)=z12+⋯+zd2,Δ(E)={Qd(x−y):x,y∈E}(E⊂Fpd).Q_d(z) = z_1^2 + \cdots+ z_d^2, \qquad\Delta(E) = \{Q_d(x-y) : x,y \in E\} \quad(E \subset\mathbb{F}_p^d).

The finite-field Erdős–Falconer problem asks how large a set must be to determine a positive proportion of all possible distances. Bourgain, Katz, and Tao [4] connected planar distance estimates with the sum–product problem. Iosevich and Rudnev [10] developed a Fourier-analytic approach and proved that, for d≥2d \ge2, the condition ∣E∣≥Cpd+12|E| \ge Cp^{\frac{d+1}{2}} implies ∣Δ(E)∣≥cp|\Delta(E)| \ge cp. Hart, Iosevich, Koh, and Rudnev [7] showed that the exponent d+12\frac{d+1}{2} cannot in general be lowered in odd dimensions. In even dimensions, it is conjectured that the sharp exponent is d/2d/2. This was confirmed recently, up to a log factor, in [16].

In this paper, we ask whether the structure of distance sets allows us to obtain a prescribed configuration, such as an arithmetic or a geometric progression, under weaker size assumptions on EE than those in the distance estimates above. Previous work has considered square and nonsquare distances [12], and the values that can be represented as ratios of nonzero distances [8, 9, 15].

In particular, we take a first step in this direction by studying three-term geometric progressions in distance sets.

We start with a definition.

Definition 1.1. Let D⊂FpD \subset\mathbb{F}_p. A nondegenerate three-term geometric progression in DD is a triple

x, xr, xr2∈D,x,r∈Fp×,r2≠1.x,\ xr,\ xr^2 \in D,\qquad x,r \in\mathbb{F}_p^\times,\qquad r^2 \ne1.

The three terms are nonzero and pairwise distinct. Equivalently, they are distinct nonzero values t1,t2,t3t_1,t_2,t_3 satisfying t22=t1t3t_2^2 = t_1t_3. Multiplication by a nonzero scalar preserves the existence of such a progression. Unless otherwise stated, geometric progressions have three terms.

The main question is formulated formally as follows.

Question 1.2. How large must E⊂FpdE \subset\mathbb{F}_p^d be in order that Δ(E)\Delta(E) contain a nondegenerate geometric progression?

Positive-density results already give partial answers for the existence of such progressions. Szemerédi’s theorem [21] states that, for each fixed k≥3k \ge3 and δ>0\delta> 0, every subset of {1,…,N}\{1,\ldots,N\} of size at least δN\delta N contains a kk-term arithmetic progression when NN is sufficiently large. Applying this theorem to a short interval of exponents in the cyclic group Fp×\mathbb{F}_p^\times gives the corresponding assertion for geometric progressions of every fixed length in sets of fixed positive density. For three terms, Roth’s theorem [19] is sufficient. The quantitative bounds of Bloom and Sisask [3] and Raghavan [18] give more: for every fixed cc, K>0K > 0, the condition

∣D∣≥cp(log⁡p)K|D| \ge\frac{cp}{(\log p)^K}

forces such a progression in D⊂FpD \subset\mathbb{F}_p when pp is sufficiently large in terms of cc, KK.4 Therefore, regarding Question 1.2, by [16], the condition ∣E∣≫pd/2|E| \gg p^{d/2} is sufficient when dd is even. This condition is also necessary, except in one case involving nonsplit forms, which will be explained in § 5.

It is worth noting that the arithmetic and geometric properties of the distance set play an important role in the study of Question 1.2. For arbitrary sets of field elements, a lower bound with a fixed exponent below one does not force such a progression. More precisely, for each fixed ε>0\varepsilon> 0, a size condition ∣D∣≥p1−ε|D| \ge p^{1-\varepsilon} does not force a nondegenerate three-term geometric progression in D⊂FpD \subset\mathbb{F}_p. Indeed, applying Behrend’s construction [2] to a short interval of exponents gives sets of size at least pexp⁡(−Clog⁡p)p\exp(-C\sqrt{\log p}) with no such progression; see § 5.

We now turn to odd dimensions. For each d≥1d \ge1, define

MGP,d(p)=max⁡{∣E∣:E⊂Fpd, Δ(E) contains no nondegenerate geometric progression}.M_{\mathrm{GP},d}(p) = \max\{|E| : E \subset\mathbb{F}_p^d,\ \Delta(E)\text{ contains no nondegenerate geometric progression}\}.

Our main result gives upper and lower bounds for this extremal quantity.

Theorem 1.3. There are absolute constants C≥1C \ge1 and p0≥7p_0 \ge7 such that, for every prime p≥p0p \ge p_0 and every integer m≥0m \ge0,

pm+726≤MGP,2m+1(p)≤Cpm+1625.(1)p^{m+\frac{7}{26}} \le M_{\mathrm{GP},2m+1}(p) \le Cp^{m+\frac{16}{25}}. \tag*{(1)}

The constants are independent of the dimension. For m≥1m \ge1, the sufficient exponent m+1625m + \frac{16}{25} is strictly smaller than the exponent m+1=d+12m + 1 = \frac{d+1}{2} for a positive proportion of distances in odd dimensions.

The main ingredient in the proof is the following one-dimensional result.

Proposition 1.4. The following statements hold.

(i) There is an absolute constant CC such that, for every prime p≥7p \ge7 and every A⊂FpA \subset\mathbb{F}_p, the condition

∣A∣≥Cp1625|A| \ge Cp^{\frac{16}{25}}

implies that (A−A)2(A-A)^2 contains a nondegenerate geometric progression.

(ii) For every sufficiently large prime pp, there exists a set A⊂FpA \subset\mathbb{F}_p such that

∣A∣≥p726|A| \ge p^{\frac{7}{26}}

and κ(A−A)2\kappa(A-A)^2 contains no nondegenerate geometric progression for every κ∈Fp×\kappa\in\mathbb{F}_p^\times.

In dimension one, Δ(A)=(A−A)2\Delta(A) = (A-A)^2, where (A−A)2={(a−b)2:a,b∈A}(A-A)^2 = \{(a-b)^2 : a,b \in A\}.

Geometric progressions in sumsets provide another relevant comparison. Ahmadi and Shparlinski [1] proved that A+BA+B contains three distinct nonzero terms in geometric progression whenever A,B⊆FpA,B \subseteq\mathbb{F}_p and ∣A∣∣B∣≥20p43|A||B| \ge20p^{\frac{4}{3}}. Taking B=−AB=-A gives such a progression in A−AA-A when ∣A∣≥20p32|A| \ge\sqrt{20}p^{\frac{3}{2}}. Shparlinski [20] strengthened the vector-space result and obtained progressions with a separate ratio in each coordinate.

The upper bound of Theorem 1.3 gives the following improvement for the case A=BA=B.

Corollary 1.5. There is an absolute constant CC such that, for every prime p≥7p \ge7 and every A⊆FpA \subseteq\mathbb{F}_p with ∣A∣≥Cp1625|A| \ge Cp^{\frac{16}{25}}, there exist x,r∈Fp×x,r \in\mathbb{F}_p^\times such that

x, xr, xr2∈A−A,r4≠1.x,\ xr,\ xr^2 \in A-A,\qquad r^4 \ne1.

Sketch of main ideas of the proofs. We first reduce the problem in odd dimensions to dimension one, using a one-dimensional version of the quotient argument of Pham, Shen, and Xue [17]. We choose an mm-dimensional totally isotropic subspace WW. More precisely, averaging over its orthogonal-group orbit controls the number of pairs of points of EE in common affine cosets of WW. In a suitable affine coset of W⊥W^\perp, projection to the one-dimensional space W⊥/WW^\perp/W then produces a large set A⊆FpA \subseteq\mathbb{F}_p with κ(A−A)2⊆Δ(E)\kappa(A-A)^2 \subseteq\Delta(E) for some κ≠0\kappa\ne0. For the lower bound, replacing each point of AA by a translate of WW multiplies the cardinality by pmp^m and changes the distance set only by a nonzero scalar.

To prove Proposition 1.4(i), suppose that (A−A)2(A-A)^2 contains no nondegenerate geometric progression. For t≠0t \ne0 with t4≠1t^4 \ne1, the map

(a,b,c)⟼(b−ta,c−tb)(a,b,c) \longmapsto(b-ta,c-tb)

is injective on A3A^3. This gives ∣A∣3≤p2|A|^3 \le p^2 and bounds the sizes of intersections of translates and dilates of AA. Plünnecke’s inequality also gives ∣A−A∣≥∣A∣54|A-A| \ge|A|^{\frac{5}{4}}. We apply these estimates to the variance of the number of representations as a product of two nonzero differences. At every nonzero squared difference this number is small, whereas its mean is large when ∣A∣|A| is large. Comparing the resulting lower and upper bounds for the variance gives ∣A∣≪p1625|A| \ll p^{\frac{16}{25}}.

For part (ii), we select vectors in {0,1,2,3}L\{0,1,2,3\}^{L} with prescribed digit counts and project them linearly to Fp\mathbb{F}_p. We delete points to remove relations that would produce nondegenerate geometric progressions among squared differences. The choice of four digits ensures that polynomial identities give only degenerate progressions. Suitable digit weights bound the number of deleted points and leave a set of size at least p726p^{\frac{7}{26}}. The cylinder construction then gives the examples in every odd dimension.

Structure. Section 2 contains the reduction to dimension one and deduces Theorem 1.3 from Proposition 1.4. Sections 3 and 4 prove parts (i) and (ii) of that proposition, respectively. Section 5 discusses even dimensions and the remaining extremal questions.

We write X≪YX \ll Y if X≤CYX \le CY for an absolute constant C>0C>0, X≫YX \gg Y if Y≪XY \ll X, and X≍YX \asymp Y if both inequalities hold. Constants may vary from one occurrence to the next.

Reduction: from 2m+12m+1 to 11

We compare the extremal problem in odd dimensions with the one-dimensional problem. The comparison does not depend on the exponents in Theorem 1.3.

Proposition 2.1. For every prime p≥7p \ge7 and every integer m≥0m \ge0,

pmMGP,1(p)≤MGP,2m+1(p)≤4pmMGP,1(p).(2)p^m M_{\mathrm{GP},1}(p) \le M_{\mathrm{GP},2m+1}(p) \le4p^m M_{\mathrm{GP},1}(p). \tag*{(2)}

The constants are independent of the dimension. Thus, every one-dimensional upper bound or construction transfers to all odd dimensions, with the factor pmp^m. For the upper bound, we extract a large set in an isotropic quotient. For the lower bound, we replace each point of a one-dimensional set by a translate of a totally isotropic subspace. We first record the quadratic-form and Fourier identities used in these two arguments.

Quadratic forms and isotropic quotients

Let QQ be a nondegenerate quadratic form on Fpd\mathbb{F}_p^d. Its associated bilinear form is

BQ(u,v)=Q(u+v)−Q(u)−Q(v)2.B_Q(u,v) = \frac{Q(u+v)-Q(u)-Q(v)}{2}.

For a subspace W⊂FpdW \subset\mathbb{F}_p^d, we write

W⊥={u∈Fpd:BQ(u,w)=0 for every w∈W}.W^\perp= \{u \in\mathbb{F}_p^d : B_Q(u,w)=0\text{ for every }w \in W\}.

For the standard form QdQ_d, we have BQd(u,v)=u⋅vB_{Q_d}(u,v)=u\cdot v. All orthogonal complements below are taken with respect to the form under consideration.

The subspace WW is totally isotropic if Q(w)=0Q(w)=0 for every w∈Ww \in W. Since pp is odd, this implies BQ(w,w′)=0B_Q(w,w')=0 for all w,w′∈Ww,w' \in W. Hence, W⊂W⊥W \subset W^\perp. If u∈W⊥u \in W^\perp and w∈Ww \in W, then

Q(u+w)=Q(u)+2BQ(u,w)+Q(w)=Q(u).Q(u+w)=Q(u)+2B_Q(u,w)+Q(w)=Q(u).

Thus, the value of Q(u)Q(u) depends only on the coset u+Wu+W, so QQ defines a quadratic form on W⊥/WW^\perp/W. In particular, passing from an affine coset of W⊥W^\perp to this quotient preserves the quadratic values of all differences.

The induced form is nondegenerate. Since dim⁡W⊥=d−dim⁡W\dim W^\perp=d-\dim W, we also have

dim⁡(W⊥/W)=d−2dim⁡W.(3)\dim(W^\perp/W)=d-2\dim W. \tag*{(3)}

Witt decomposition shows that every nondegenerate quadratic form on Fp2m+1\mathbb{F}_p^{2m+1} has a totally isotropic subspace of dimension mm; see [9 Lemma 2.1]. For such a subspace WW, the quotient W⊥/WW^\perp/W is one-dimensional and nondegenerate. Therefore, the induced quadratic form is κt2\kappa t^2 with κ∈Fp×\kappa\in\mathbb{F}_p^\times.

Extraction of a one-dimensional set

The next lemma extracts a set A⊂FpA \subset\mathbb{F}_p whose squared differences, after multiplication by a nonzero scalar, belong to Δ(E)\Delta(E).

Lemma 2.2. Let p≥7p \ge7, let m≥1m \ge1, and let ∅≠E⊂Fp2m+1\varnothing\ne E \subset\mathbb{F}_p^{2m+1}. There are A⊂FpA \subset\mathbb{F}_p and κ∈Fp×\kappa\in\mathbb{F}_p^\times such that

∣A∣≥∣E∣pm(2+∣E∣pm+1),κ(A−A)2⊂Δ(E).(4)|A| \ge\frac{|E|}{p^m\left(2+\frac{|E|}{p^{m+1}}\right)}, \qquad\kappa(A-A)^2 \subset\Delta(E). \tag*{(4)}

When ∣E∣≤pm+1|E| \le p^{m+1}, the denominator in parentheses in (4) is at most 3. We retain the full bound so that Proposition 2.1 can be proved without an initial restriction on ∣E∣|E|.

Proof. We seek a totally isotropic mm-dimensional subspace WW and an affine coset of W⊥W^\perp in which many cosets of WW meet EE. After translation, their images in W⊥/WW^\perp/W will form the required set AA. To find WW, we estimate the number of pairs of points of EE lying in the same coset of WW.

We first bound the number of pairs at distance zero. Write S0={x∈Fp2m+1:Q2m+1(x)=0}S_0=\{x\in\mathbb{F}_p^{2m+1}:Q_{2m+1}(x)=0\}. We have ∣S0∣=p2m|S_0|=p^{2m} and

∣{(x,y)∈E2:Q2m+1(x−y)=0}∣≤∣E∣2p+pm∣E∣.(5)\left|\{(x,y)\in E^2:Q_{2m+1}(x-y)=0\}\right|\le\frac{|E|^2}{p}+p^m|E|. \tag*{(5)}

a proof can be found in [11].

We now average over isotropic subspaces. Choose one such subspace of dimension mm, and choose WW uniformly from its orbit under the orthogonal group. To compute the probability that a fixed vector lies in WW, we use transitivity on the nonzero isotropic vectors. This can be checked directly. For two such vectors v,v′v,v', reflection across the hyperplane orthogonal to v−v′v-v' sends vv to v′v' if v⋅v′≠0v\cdot v'\ne0. If v⋅v′=0v\cdot v'=0, choose uu with Q2m+1(u)(v⋅u)(v′⋅u)≠0Q_{2m+1}(u)(v\cdot u)(v'\cdot u)\ne0 and first reflect vv across the hyperplane orthogonal to uu. Its inner product with v′v' is then nonzero. Such a uu exists because the displayed nonzero polynomial has degree four, less than pp, and so cannot vanish at every point of Fp2m+1\mathbb{F}_p^{2m+1}. It follows by double counting that, for every nonzero isotropic vector vv,

Pr⁡(v∈W)=pm−1p2m−1=1pm+1<p−m.(6)\Pr(v\in W)=\frac{p^m-1}{p^{2m}-1}=\frac{1}{p^m+1}<p^{-m}. \tag*{(6)}

The sum ∑x+W∣E∩(x+W)∣2\sum_{x+W}|E\cap(x+W)|^2 counts ordered pairs of points whose difference lies in WW. By (5) and (6), its expectation is at most

∣E∣+p−m(∣E∣2p+pm∣E∣)=2∣E∣+∣E∣2pm+1.|E|+p^{-m}\left(\frac{|E|^2}{p}+p^m|E|\right)=2|E|+\frac{|E|^2}{p^{m+1}}.

Fix WW for which this bound holds. The Cauchy–Schwarz inequality then shows that at least

∣E∣2+∣E∣pm+1\frac{|E|}{2+\frac{|E|}{p^{m+1}}}

affine cosets of WW meet EE. Each lies in one affine coset of W⊥W^\perp. Since there are pmp^m cosets of W⊥W^\perp, some affine subspace z+W⊥z+W^\perp contains at least

∣E∣pm(2+∣E∣pm+1)\frac{|E|}{p^m\left(2+\frac{|E|}{p^{m+1}}\right)}

cosets of WW that meet EE.

Translate E∩(z+W⊥)E\cap(z+W^\perp) by −z-z and take its image in W⊥/WW^\perp/W. By (3), the quotient is one-dimensional and nondegenerate, so its form can be written as κt2\kappa t^2 with κ≠0\kappa\ne0. Identify the image with a set A⊂FpA\subset\mathbb{F}_p, and choose one representative xa∈E∩(z+W⊥)x_a\in E\cap(z+W^\perp) for each a∈Aa\in A. The quotient form gives

κ(a−b)2=Q2m+1(xa−xb)(a,b∈A).\kappa(a-b)^2=Q_{2m+1}(x_a-x_b)\qquad(a,b\in A).

This proves both assertions and completes the proof. □

The cylinder construction

The reverse construction replaces each element of AA by a translate of an mm-dimensional totally isotropic subspace. This multiplies the cardinality by pmp^m, while the distances depend only on the corresponding elements of AA. The following lemma is the coordinate-free form of the construction in [9 Lemma 2.1(ii) and Section 4.2].

Lemma 2.3. Let m≥0m\ge0 be an integer and let d=2m+1d=2m+1. There exist a totally isotropic mm-dimensional subspace W⊂FpdW\subset\mathbb{F}_p^d and a vector e∈W⊥∖We\in W^\perp\setminus W such that κ:=Qd(e)≠0\kappa:=Q_d(e)\ne0 and, for every A⊂FpA\subset\mathbb{F}_p, the set

EA:={w+ae:w∈W, a∈A}E_A:=\{w+ae:w\in W,\ a\in A\}

satisfies

∣EA∣=pm∣A∣,Δ(EA)=κ(A−A)2.(7)|E_A|=p^m|A|,\qquad\Delta(E_A)=\kappa(A-A)^2. \tag*{(7)}

Proof. If m=0m=0, then the conclusion follows by taking W={0}W=\{0\} and e=1e=1, so that κ=Q1(e)=1\kappa=Q_1(e)=1. Therefore, it remains to prove the result in the case m≥1m\ge1.

Choose a totally isotropic subspace W⊂FpdW\subset\mathbb{F}_p^d of dimension mm, as in Section 2.1. The quadratic form induced by QdQ_d on W⊥/WW^\perp/W is nondegenerate and one-dimensional. Hence, for any e∈W⊥∖We\in W^\perp\setminus W,

κ:=Qd(e)≠0.\kappa:=Q_d(e)\ne0.

Since e∉We\notin W, the representation w+aew+ae, with w∈Ww\in W and a∈Aa\in A, is unique. Therefore,

∣EA∣=∣W∣∣A∣=pm∣A∣.|E_A|=|W||A|=p^m|A|.

Moreover, for w,w′∈Ww,w'\in W and a,b∈Aa,b\in A, the total isotropy of WW and the fact that e∈W⊥e\in W^\perp give

Qd(w−w′)=0,(w−w′)⋅e=0.Q_d(w-w')=0,\qquad(w-w')\cdot e=0.

Consequently,

Qd((w+ae)−(w′+be))=Qd(w−w′)+2(a−b)(w−w′)⋅e+(a−b)2Qd(e)=κ(a−b)2.Q_d\left((w+ae)-(w'+be)\right)=Q_d(w-w')+2(a-b)(w-w')\cdot e +\left(a-b\right)^2Q_d(e) =\kappa\left(a-b\right)^2.

Thus, Δ(EA)⊆κ(A−A)2\Delta(E_A)\subseteq\kappa(A-A)^2. The reverse inclusion follows by taking w=w′=0w=w'=0. Hence, Δ(EA)=κ(A−A)2\Delta(E_A)=\kappa(A-A)^2. This completes the proof.

Proof of Proposition 2.1, Theorem 1.3, and Corollary 1.5

We now combine the two lemmas. The estimate MGP,1(p)≤(p−1)/2M_{\mathrm{GP},1}(p)\leq(p-1)/2 allows us to absorb the dependence on ∣E∣|E| in the denominator of (4).

Proof of Proposition 2.1. If m=0m=0, then (2) follows directly from the definition of MGP,1(p)M_{\mathrm{GP},1}(p). We henceforth assume that m≥1m\geq1. We first note that

MGP,1(p)≤p−12.M_{\mathrm{GP},1}(p)\leq\frac{p-1}{2}.

Indeed, if A⊂FpA\subset\mathbb{F}_p has more than p/2p/2 elements, then AA and u+Au+A intersect for every u∈Fpu\in\mathbb{F}_p, so A−A=FpA-A=\mathbb{F}_p. Hence, (A−A)2(A-A)^2 contains all nonzero squares. The set of nonzero squares is a cyclic multiplicative subgroup of order (p−1)/2≥3(p-1)/2\geq3. Therefore, it contains a nondegenerate geometric progression.

Let E⊂Fp2m+1E\subset\mathbb{F}_p^{2m+1} be nonempty and suppose that its distance set contains no nondegenerate geometric progression. By Lemma 2.2 and the invariance under nonzero scaling noted after Definition 1.1,

MGP,1(p)≥∣E∣pm(2+∣E∣pm+1).M_{\mathrm{GP},1}(p)\geq\frac{|E|}{p^m\left(2+\frac{|E|}{p^{m+1}}\right)}.

Multiplying through by the denominator and collecting the terms containing ∣E∣|E|, we obtain

∣E∣(1−MGP,1(p)p)≤2pmMGP,1(p).|E|\left(1-\frac{M_{\mathrm{GP},1}(p)}{p}\right)\leq2p^mM_{\mathrm{GP},1}(p).

Since MGP,1(p)≤p−12M_{\mathrm{GP},1}(p)\leq\frac{p-1}{2}, the factor on the left is greater than 12\frac{1}{2}. Hence, ∣E∣≤4pmMGP,1(p)|E|\leq4p^mM_{\mathrm{GP},1}(p). This proves the upper bound in (2).

For the lower bound, choose A⊂FpA\subset\mathbb{F}_p with ∣A∣=MGP,1(p)|A|=M_{\mathrm{GP},1}(p) such that (A−A)2(A-A)^2 contains no nondegenerate geometric progression. Lemma 2.3 constructs a set EA⊂Fp2m+1E_A\subset\mathbb{F}_p^{2m+1} satisfying ∣EA∣=pmMGP,1(p)|E_A|=p^mM_{\mathrm{GP},1}(p) and Δ(EA)=κ(A−A)2\Delta(E_A)=\kappa(A-A)^2 for some κ≠0\kappa\ne0. Since multiplication by κ\kappa preserves nondegeneracy, Δ(EA)\Delta(E_A) also contains no nondegenerate geometric progression. Consequently, MGP,2m+1(p)≥pmMGP,1(p)M_{\mathrm{GP},2m+1}(p)\geq p^mM_{\mathrm{GP},1}(p). This proves (2) and completes the proof.

The main theorem follows by applying the comparison to the two one-dimensional bounds.

Proof of Theorem 1.3. Proposition 1.4(i) gives an absolute constant C1C_1 such that MGP,1(p)≤C1p1625M_{\mathrm{GP},1}(p) \le C_1p^{\frac{16}{25}} for every prime p≥7p \ge7. Part (ii) gives an absolute p0≥7p_0 \ge7 such that MGP,1(p)≥p726M_{\mathrm{GP},1}(p) \ge p^{\frac{7}{26}} for every prime p≥p0p \ge p_0. Hence, Proposition 2.1 yields

pm+726≤pmMGP,1(p)≤MGP,2m+1(p)≤4pmMGP,1(p)≤4C1pm+1625.p^{m+\frac{7}{26}} \le p^m M_{\mathrm{GP},1}(p) \le M_{\mathrm{GP},2m+1}(p) \le4p^m M_{\mathrm{GP},1}(p) \le4C_1p^{m+\frac{16}{25}}.

Taking C=max⁡{1,4C1}C = \max\{1,4C_1\} proves the theorem. Both CC and p0p_0 are independent of mm. This completes the proof.

Proof of Corollary 1.5. By Proposition 1.4(i), there exist u,v,w∈(A−A)∖{0}u,v,w \in(A-A) \setminus\{0\} such that u2,v2,w2u^2,v^2,w^2 are pairwise distinct and form a geometric progression. Thus,

v4=u2w2,v^4 = u^2w^2,

and hence

(v2−uw)(v2+uw)=0.(v^2-uw)(v^2+uw)=0.

Therefore, v2=uwv^2=uw or v2=−uwv^2=-uw. Since A−AA-A is symmetric, replacing ww by −w-w if necessary, we may assume that v2=uwv^2=uw.

Set x=ux=u and r=v/ur=v/u. Then x,r∈Fp×x,r \in\mathbb{F}_p^\times and

x=u,xr=v,xr2=v2u=w.x=u,\qquad xr=v,\qquad xr^2=\frac{v^2}{u}=w.

Thus, x,xr,xr2∈A−Ax,xr,xr^2 \in A-A. Moreover,

r4=v4u4=w2u2≠1,r^4=\frac{v^4}{u^4}=\frac{w^2}{u^2}\ne1,

since u2≠w2u^2 \ne w^2. This completes the proof.

We prove the two parts of Proposition 1.4 in the next two sections.

Proof of Proposition 1.4(i)

The proof compares two bounds for the variance of the number of representations of a nonzero field element as a product of two differences from AA. The absence of a progression makes this number small at every nonzero squared difference. We first show that there are many such squared differences and establish the injectivity estimate needed for the upper bound on the variance.

Consequences of the absence of a progression

The three conclusions of the next lemma have different roles. The bound for ∣A∣|A| controls the exceptional slopes in the variance estimate, the bound for ∣A−A∣|A-A| supplies many squared differences, and the bound for the number of solutions to b−ta=sb-ta=s controls the remaining slopes. All three follow from the injectivity of (a,b,c)↦(b−ta,c−tb)(a,b,c) \mapsto(b-ta,c-tb) when t≠0t \ne0 and $t^4 \ne 1.

Lemma 3.1. Let p≥7p \ge7 be a prime and let ∅≠A⊂Fp\varnothing\ne A \subset\mathbb{F}_p. If (A−A)2(A-A)^2 contains no nondegenerate geometric progression, then

∣A∣3≤p2,∣A−A∣≥∣A∣54.(8)|A|^3 \le p^2,\qquad|A-A| \ge|A|^{\frac{5}{4}}. \tag*{(8)}

Moreover, for t≠0t \ne0 with t4≠1t^4 \ne1 and every s∈Fps \in\mathbb{F}_p,

∣{(a,b)∈A2:b−ta=s}∣≤p∣A∣.(9)\left|\{(a,b)\in A^2:b-ta=s\}\right| \le\frac{p}{|A|}. \tag*{(9)}

Proof. Fix t∈Fp×t \in\mathbb{F}_p^\times with t4≠1t^4 \ne1 and consider the map

Φt:A3⟶Fp2,Φt(a,b,c)=(b−ta,c−tb).\Phi_t:A^3\longrightarrow\mathbb{F}_p^2,\qquad\Phi_t(a,b,c)=(b-ta,c-tb).

If two triples (a,b,c)(a,b,c) and (a′,b′,c′)(a',b',c') have the same image, then

b−b′=t(a−a′),c−c′=t(b−b′).b-b'=t(a-a'),\qquad c-c'=t(b-b').

If the triples are distinct, then a−a′≠0a-a' \ne0, and these three differences are nonzero and have common ratio tt. Their squares would form a nondegenerate geometric progression in (A−A)2(A-A)^2, since t4≠1t^4 \ne1. Thus, Φt\Phi_t is injective. Since 24≠12^4 \ne1 for p≥7p \ge7, taking t=2t=2 yields ∣A∣3≤p2|A|^3 \le p^2.

To obtain the lower bound for ∣A−A∣|A-A|, apply Plünnecke’s inequality [14 Theorem 1.1] to AA and −A-A. This gives a nonempty set X⊂AX \subset A such that

∣X−A−A∣≤(∣A−A∣∣A∣)2∣X∣.|X-A-A| \le\left(\frac{|A-A|}{|A|}\right)^2|X|.

For every (a,b,c)∈A×X×X(a,b,c) \in A \times X \times X, we have

b−2a=b−a−a∈X−A−A,c−2b=c−b−b∈X−A−A.b-2a=b-a-a\in X-A-A,\qquad c-2b=c-b-b\in X-A-A.

Since Φ2\Phi_2 is injective,

∣A∣∣X∣2≤∣X−A−A∣2≤(∣A−A∣∣A∣)4∣X∣2.|A||X|^2 \le|X-A-A|^2 \le\left(\frac{|A-A|}{|A|}\right)^4|X|^2.

Since X≠∅X \ne\varnothing, it follows that ∣A−A∣≥∣A∣54|A-A| \ge|A|^{\frac{5}{4}}.

Finally, for each s∈Fps \in\mathbb{F}_p, the map Φt\Phi_t sends {(a,b,c)∈A3:b−ta=s}\{(a,b,c)\in A^3:b-ta=s\} injectively into {s}×Fp\{s\}\times\mathbb{F}_p. Comparing cardinalities gives

∣A∣∣{(a,b)∈A2:b−ta=s}∣≤p,|A|\left|\{(a,b)\in A^2:b-ta=s\}\right| \le p,

which proves (3.2). This completes the proof.

The variance argument

For a nonempty set A⊂FpA \subset\mathbb{F}_p, let rA−A(h)=∣{(a,b)∈A2:a−b=h}∣r_{A-A}(h)=|\{(a,b)\in A^2:a-b=h\}| for h∈Fph\in\mathbb{F}_p. Thus, rA−A(h)>0r_{A-A}(h)>0 if and only if h∈A−Ah\in A-A. To count products of nonzero differences with their multiplicities, define

F(z):=∑x,y∈Fp×xy=zrA−A(x)rA−A(y)(z∈Fp×).F(z):=\sum_{\substack{x,y\in\mathbb{F}_p^\times\\xy=z}}r_{A-A}(x)r_{A-A}(y)\qquad(z\in\mathbb{F}_p^\times).

Its mean is

F‾:=1p−1∑z∈Fp×F(z)=∣A∣2(∣A∣−1)2p−1.\overline{F}:=\frac{1}{p-1}\sum_{z\in\mathbb{F}_p^\times}F(z)=\frac{|A|^2(|A|-1)^2}{p-1}.

The formula for the mean follows from ∑h≠0rA−A(h)=∣A∣(∣A∣−1)\sum_{h\ne0}r_{A-A}(h)=|A|(|A|-1).

Let h∈(A−A)∖{0}h\in(A-A)\setminus\{0\}. If x,y∈Fp×x,y\in\mathbb{F}_p^\times satisfy xy=h2xy=h^2 and rA−A(x)rA−A(y)>0r_{A-A}(x)r_{A-A}(y)>0, then x,y∈A−Ax,y\in A-A. Consequently, x2,h2,y2∈(A−A)2x^2,h^2,y^2\in(A-A)^2 and

h2x2=yx=y2h2.\frac{h^2}{x^2}=\frac{y}{x}=\frac{y^2}{h^2}.

Thus, these three values form a geometric progression, possibly degenerate. When no nondegenerate progression is present, only a bounded number of pairs (x,y)(x,y) are possible. Thus, F(h2)F(h^2) is small. For ∣A∣>6p|A|>6\sqrt{p}, these values lie below half the mean and give the lower bound in the following lemma. The upper bound will follow from Lemma 3.1.

Lemma 3.2. Let p≥7p\ge7 be a prime and let ∅≠A⊂Fp\varnothing\ne A\subset\mathbb{F}_p. Suppose that (A−A)2(A-A)^2 contains no nondegenerate geometric progression. Then

∑z∈Fp×(F(z)−F‾)2≤5p2∣A∣3.\sum_{z\in\mathbb{F}_p^\times}(F(z)-\overline{F})^2\le5p^2|A|^3.

If ∣A∣>6p|A|>6\sqrt{p}, then also

∑z∈Fp×(F(z)−F‾)2≫(∣A−A∣−1)∣A∣8p2,\sum_{z\in\mathbb{F}_p^\times}(F(z)-\overline{F})^2\gg\frac{(|A-A|-1)|A|^8}{p^2},

where the implied constant is absolute.

Proof of Lemma 3.2. We first bound the variance from above by counting pairs on lines of a fixed slope. We then obtain the lower bound from the small values of FF at nonzero squared differences.

The equation xy=uv≠0xy=uv\ne0 can be written as u=txu=tx and y=tvy=tv for some t≠0t\ne0. This relates the second moment of FF to the numbers of solutions of b−b′=t(a−a′)b-b'=t(a-a'). We count these solutions through the fibers of (a,b)↦b−ta(a,b)\mapsto b-ta. For s,t∈Fps,t\in\mathbb{F}_p, let

it(s)=∣{(a,b)∈A2:b−ta=s}∣,V(t)=∑s∈Fp(it(s)−∣A∣2p)2.i_t(s)=|\{(a,b)\in A^2:b-ta=s\}|,\qquad V(t)=\sum_{s\in\mathbb{F}_p}\left(i_t(s)-\frac{|A|^2}{p}\right)^2.

Since ∑s∈Fpit(s)=∣A∣2\sum_{s\in\mathbb{F}_p}i_t(s)=|A|^2, expanding the square in V(t)V(t) reduces the calculation to counting solutions of b−b′=t(a−a′)b-b'=t(a-a'). When a≠a′a\ne a', each choice of a,a′,b,b′a,a',b,b' gives one value of tt; when a=a′a=a', a solution requires b=b′b=b' and then all pp values of tt are allowed. Hence,

∑t∈Fp∑s∈Fpit(s)2=(∣A∣2−∣A∣)∣A∣2+p∣A∣2.\sum_{t\in\mathbb{F}_p}\sum_{s\in\mathbb{F}_p} i_t(s)^2=(|A|^2-|A|)|A|^2+p|A|^2.

Subtracting ∣A∣4|A|^4 gives

V(t)=∑s∈Fpit(s)2−∣A∣4p≥0,∑t∈FpV(t)=p∣A∣2−∣A∣3≤p∣A∣2.(10)V(t)=\sum_{s\in\mathbb{F}_p}i_t(s)^2-\frac{|A|^4}{p}\ge0,\qquad\sum_{t\in\mathbb{F}_p}V(t)=p|A|^2-|A|^3\le p|A|^2. \tag*{(10)}

For t≠0t\ne0 with t4≠1t^4\ne1, (3.2) implies V(t)≤∑s∈Fpit(s)2≤p∣A∣V(t)\le\sum_{s\in\mathbb{F}_p}i_t(s)^2\le p|A|. For the at most four values of tt satisfying t4=1t^4=1, we use V(t)≤∑s∈Fpit(s)2≤∣A∣3V(t)\le\sum_{s\in\mathbb{F}_p}i_t(s)^2\le|A|^3. For t≠0t\ne0 with t4≠1t^4\ne1, we have V(t)2≤p∣A∣V(t)V(t)^2\le p|A|V(t). The values satisfying t4=1t^4=1 contribute at most 4∣A∣64|A|^6. Thus, (3.3) and the bound ∣A∣3≤p2|A|^3\le p^2 from Lemma 3.1 yield

∑t∈Fp×V(t)2≤p2∣A∣3+4∣A∣6≤5p2∣A∣3.(11)\sum_{t\in\mathbb{F}_p^\times}V(t)^2\le p^2|A|^3+4|A|^6\le5p^2|A|^3. \tag*{(11)}

We now transfer this estimate to FF. Grouping the weighted solutions of xy=uv≠0xy=uv\ne0 by t=ux=yvt=\frac{u}{x}=\frac{y}{v} gives

∑z∈Fp×F(z)2=∑t∈Fp×(∑h∈Fp×rA−A(h)rA−A(th))2.(12)\sum_{z\in\mathbb{F}_p^\times}F(z)^2=\sum_{t\in\mathbb{F}_p^\times}\left(\sum_{h\in\mathbb{F}_p^\times}r_{A-A}(h)r_{A-A}(th)\right)^2. \tag*{(12)}

Moreover, since multiplication by any h≠0h\ne0 permutes Fp×\mathbb{F}_p^\times,

∑t∈Fp×∑h∈Fp×rA−A(h)rA−A(th)=(∑h∈Fp×rA−A(h))2=∣A∣2(∣A∣−1)2=∑z∈Fp×F(z).\sum_{t\in\mathbb{F}_p^\times}\sum_{h\in\mathbb{F}_p^\times}r_{A-A}(h)r_{A-A}(th)=\left(\sum_{h\in\mathbb{F}_p^\times}r_{A-A}(h)\right)^2=|A|^2(|A|-1)^2=\sum_{z\in\mathbb{F}_p^\times}F(z).

Thus, FF and the function t↦∑h∈Fp×rA−A(h)rA−A(th)t\mapsto\sum_{h\in\mathbb{F}_p^\times}r_{A-A}(h)r_{A-A}(th) have the same mean F‾\overline{F} and second moment.

For t∈Fp×t\in\mathbb{F}_p^\times, the sum ∑s∈Fpit(s)2\sum_{s\in\mathbb{F}_p}i_t(s)^2 counts solutions of b−b′=t(a−a′)b-b'=t(a-a') in A4A^4. Removing the ∣A∣2|A|^2 solutions with a=a′a=a' gives

∑h∈Fp×rA−A(h)rA−A(th)=∑s∈Fpit(s)2−∣A∣2=V(t)+∣A∣4p−∣A∣2.\sum_{h\in\mathbb{F}_p^\times}r_{A-A}(h)r_{A-A}(th)=\sum_{s\in\mathbb{F}_p}i_t(s)^2-|A|^2=V(t)+\frac{|A|^4}{p}-|A|^2.

Combining this identity with (3.5) and the equality of the means, we obtain

∑z∈Fp×(F(z)−F‾)2=∑t∈Fp×(V(t)−1p−1∑u∈Fp×V(u))2≤∑t∈Fp×V(t)2≤5p2∣A∣3,(13)\sum_{z\in\mathbb{F}_p^\times}(F(z)-\overline{F})^2=\sum_{t\in\mathbb{F}_p^\times}\left(V(t)-\frac{1}{p-1}\sum_{u\in\mathbb{F}_p^\times}V(u)\right)^2 \le\sum_{t\in\mathbb{F}_p^\times}V(t)^2\le5p^2|A|^3, \tag*{(13)}

where the last inequality follows from (3.4).

We now prove the lower bound. For h∈(A−A)∖{0}h \in(A-A) \setminus\{0\}, each nonzero summand in F(h2)F(h^2) corresponds to nonzero x,y∈A−Ax,y \in A-A with xy=h2xy=h^2. The terms x2,h2,y2x^2,h^2,y^2 form a geometric progression with ratio h2x2\frac{h^2}{x^2}. This progression must be degenerate, so (h2x2)2=1\left(\frac{h^2}{x^2}\right)^2=1, or equivalently x4=h4x^4=h^4. There are at most four choices for xx, each determining yy. Since rA−A≤∣A∣r_{A-A}\leq|A|, we obtain

F(h2)≤4∣A∣2.(14)F(h^2)\leq4|A|^2. \tag*{(14)}

Assume now that ∣A∣>6p|A|>6\sqrt{p}. Then

F‾=∣A∣2(∣A∣−1)2p−1≥∣A∣44p.\overline{F}=\frac{|A|^2(|A|-1)^2}{p-1}\geq\frac{|A|^4}{4p}.

Since ∣A∣>6p|A|>6\sqrt{p}, we also have F‾>8∣A∣2\overline{F}>8|A|^2.

Since A−AA-A is symmetric and pp is odd, the squaring map is two-to-one on (A−A)∖{0}(A-A)\setminus\{0\}, so the set D=(A−A)2∖{0}D=(A-A)^2\setminus\{0\} has cardinality ∣D∣=∣A−A∣−12|D|=\frac{|A-A|-1}{2}. Moreover, (3.7) and the bound F‾≥8∣A∣2\overline{F}\geq8|A|^2 give F(z)≤4∣A∣2≤F‾2F(z)\leq4|A|^2\leq\frac{\overline{F}}{2} for every z∈Dz\in D. Therefore,

∑z∈Fp×(F(z)−F‾)2≥∑z∈D(F(z)−F‾)2≥∣D∣4F‾2=∣A−A∣−18F‾2≫(∣A−A∣−1)∣A∣8p2.(15)\begin{aligned} \sum_{z\in\mathbb{F}_p^\times}(F(z)-\overline{F})^2 &\geq\sum_{z\in D}(F(z)-\overline{F})^2\\ &\geq\frac{|D|}{4}\overline{F}^2 =\frac{|A-A|-1}{8}\overline{F}^2\\ &\gg\frac{(|A-A|-1)|A|^8}{p^2}. \tag*{(15)} \end{aligned}

This completes the proof.

We can now prove the upper bound. The next calculation explains the exponent 1625\frac{16}{25}.

Proof of Proposition 1.4(i). Assume that (A−A)2(A-A)^2 contains no nondegenerate geometric progression. If ∣A∣≤6p|A|\leq6\sqrt{p}, then the required bound ∣A∣≪p16/25|A|\ll p^{16/25} is immediate. Hence, we may assume ∣A∣>6p|A|>6\sqrt{p}. Comparing the two estimates in Lemma 3.2, we obtain

(∣A−A∣−1)∣A∣5≪p4.(|A-A|-1)|A|^5\ll p^4.

Since (3.1) and ∣A∣>6p|A|>6\sqrt{p} imply ∣A−A∣−1≫∣A∣5/4|A-A|-1\gg|A|^{5/4}, it follows that

∣A∣25/4≪p4,so∣A∣≪p16/25.|A|^{25/4}\ll p^4,\qquad\text{so}\qquad|A|\ll p^{16/25}.

Choosing the constant in Proposition 1.4(i) sufficiently large proves the assertion. This completes the proof.

Proof of Proposition 1.4(ii)

We construct a large family of integer vectors, project it to Fp\mathbb{F}_p, and delete points to remove the relations that would produce a nondegenerate geometric progression.

Random projection and deletion

To construct the set in Proposition 1.4(ii), we must control relations among all differences of AA. The following criterion allows us to work before squaring those differences.

For κ∈Fp×\kappa\in\mathbb{F}_p^\times, the set κ(A−A)2\kappa(A-A)^2 contains a nondegenerate geometric progression if and only if there exist u,v,w∈(A−A)∖{0}u,v,w \in(A-A)\setminus\{0\} with

v2=uw,u4≠v4.(16)v^2=uw,\qquad u^4\ne v^4. \tag*{(16)}

Indeed, v2=uwv^2=uw gives the progression κu2,κv2,κw2\kappa u^2,\kappa v^2,\kappa w^2. The inequality u4≠v4u^4\ne v^4 is precisely its nondegeneracy condition. Conversely, a progression of squared differences satisfies v4=u2w2v^4=u^2w^2. Thus, v2=±uwv^2=\pm uw. Since A−AA-A is symmetric, we may replace ww with −w-w to remove the minus sign.

We begin with a family of integer vectors in {0,1,2,3}L\{0,1,2,3\}^L and project it linearly to Fp\mathbb{F}_p. A collision in the projection or an unwanted quadratic relation can be removed by deleting points so that one of its differences no longer occurs. There are two kinds of quadratic relations. If the defining polynomial is nonzero, its probability of vanishing is at most 2/p2/p. If it is identically zero, the choice of four digits will imply that the progression is degenerate.

In the next lemma, rC−C(u)r_{C-C}(\mathbf{u}) counts the ordered pairs with difference u\mathbf{u}. For a relation involving three differences, we delete the occurrences of a least frequent difference. The cost is at most the geometric mean of the three multiplicities. This is why the cube-root sum J(C)J(C) appears in the estimate.

Lemma 4.1. Let L≥1L\ge1 and C⊆{0,1,2,3}LC\subseteq\{0,1,2,3\}^L. With differences and intersections taken in ZL\mathbb{Z}^L, let

D:=C−C,rC−C(u):=∣C∩(C+u)∣,J(C):=∑u∈DrC−C(u)1/3.\mathcal{D}:=C-C,\qquad r_{C-C}(\mathbf{u}):=|C\cap(C+\mathbf{u})|,\qquad J(C):=\sum_{\mathbf{u}\in\mathcal{D}}r_{C-C}(\mathbf{u})^{1/3}.

For every prime p>18p>18, there exists A⊂FpA\subset\mathbb{F}_p such that (A−A)2(A-A)^2 contains no nondegenerate geometric progression and

∣A∣≥∣C∣−∣C∣2p−2J(C)3p.(17)|A|\ge|C|-\frac{|C|^2}{p}-\frac{2J(C)^3}{p}. \tag*{(17)}

Proof. We will choose a map from ZL\mathbb{Z}^L to Fp\mathbb{F}_p and delete points from CC. The set AA will be the image of the remaining subset C′C'. For α=(α1,…,αL)∈FpL\boldsymbol{\alpha}=(\alpha_1,\ldots,\alpha_L)\in\mathbb{F}_p^L, define

ϕα:ZL⟶Fp,\phi_{\boldsymbol{\alpha}}:\mathbb{Z}^L\longrightarrow\mathbb{F}_p,
c=(c1,…,cL)⟼∑i=1Lci‾αi,\mathbf{c}=(c_1,\ldots,c_L)\longmapsto\sum_{i=1}^{L}\overline{c_i}\alpha_i,

where ci‾∈Fp\overline{c_i}\in\mathbb{F}_p is the residue class of the integer cic_i modulo pp. Thus, we first reduce the coordinates modulo pp and then take their dot product with α\boldsymbol{\alpha} in FpL\mathbb{F}_p^L. We use this map on CC and on its difference set D\mathcal{D}.

We need to satisfy two requirements. First, ϕα\phi_\alpha must be injective on C′C'. Second, with A=ϕα(C′)A=\phi_\alpha(C'), every δ1,δ2,δ3∈(A−A)∖{0}\delta_1,\delta_2,\delta_3\in(A-A)\setminus\{0\} must satisfy

δ22=δ1δ3⟹δ14=δ24.\delta_2^2=\delta_1\delta_3 \quad\Longrightarrow\quad\delta_1^4=\delta_2^4.

By (16), the second requirement excludes nondegenerate geometric progressions in (A−A)2(A-A)^2. The first gives ∣A∣=∣C′∣|A|=|C'|, so the required cardinality bound follows if

∣C∖C′∣≤∣C∣2p+2J(C)3p.|C\setminus C'|\le\frac{|C|^2}{p}+\frac{2J(C)^3}{p}.

We choose α∈FpL\boldsymbol{\alpha}\in\mathbb{F}_p^L uniformly. A collision occurs when a nonzero difference vector projects to zero. A relation among three projected differences gives a quadratic equation in α\boldsymbol{\alpha}. We will delete points to remove all collisions and all such relations defined by nonzero polynomials. Each deletion will make one of the relevant difference vectors disappear. For a fixed u∈D∖{0}\mathbf{u}\in\mathcal{D}\setminus\{0\}, the event ϕα(u)=0\phi_\alpha(\mathbf{u})=0 has probability 1/p1/p. A fixed nonzero quadratic polynomial vanishes with probability at most 2/p2/p. These estimates will bound the expected number of deleted points. Polynomial identities require a separate argument: we will show at the end that the four-digit assumption makes every remaining relation degenerate.

We now define the deletions. For each u∈D\mathbf{u}\in\mathcal{D}, define

Ru=C∩(C+u).R_{\mathbf{u}}=C\cap(C+\mathbf{u}).

We have ∣Ru∣=rC−C(u)|R_{\mathbf{u}}|=r_{C-C}(\mathbf{u}). Whenever x,y∈C\mathbf{x},\mathbf{y}\in C satisfy x−y=u\mathbf{x}-\mathbf{y}=\mathbf{u}, we have x∈Ru\mathbf{x}\in R_{\mathbf{u}}, so deleting RuR_{\mathbf{u}} leaves no pair with difference u\mathbf{u}.

For each ordered triple τ=(u,v,w)∈(D∖{0})3\tau=(\mathbf{u},\mathbf{v},\mathbf{w})\in(\mathcal{D}\setminus\{0\})^3, define

Pτ(X)=(v⋅X)2−(u⋅X)(w⋅X)in Fp[X1,…,XL].P_\tau(X)=(\mathbf{v}\cdot X)^2-(\mathbf{u}\cdot X)(\mathbf{w}\cdot X)\quad\text{in }\mathbb{F}_p[X_1,\ldots,X_L].

Choose d(τ)\mathbf{d}(\tau) among u,v,w\mathbf{u},\mathbf{v},\mathbf{w} to minimize rC−C(d(τ))r_{C-C}(\mathbf{d}(\tau)), resolving ties by a fixed rule independent of α\boldsymbol{\alpha}. Let C′C' be obtained from CC by deleting RuR_{\mathbf{u}} whenever u∈D∖{0}\mathbf{u}\in\mathcal{D}\setminus\{0\} and ϕα(u)=0\phi_\alpha(\mathbf{u})=0, and deleting Rd(τ)R_{\mathbf{d}(\tau)} whenever Pτ≢0P_\tau\not\equiv0 and Pτ(α)=0P_\tau(\boldsymbol{\alpha})=0.

We next estimate the expected number of deleted points. Every nonzero u∈D\mathbf{u}\in\mathcal{D} remains nonzero modulo pp, since its coordinates have absolute value at most 33 and p>18p>18. Hence, ϕα(u)=0\phi_\alpha(\mathbf{u})=0 with probability 1p\frac{1}{p}. Similarly, if Pτ≢0P_\tau\not\equiv0, then PτP_\tau has total degree two and at most 2pL−12p^{L-1} zeros in FpL\mathbb{F}_p^L, so Pτ(α)=0P_\tau(\boldsymbol{\alpha})=0 with probability at most 2p\frac{2}{p}. Since the number of deleted points is bounded by the sum of the cardinalities of the deleted sets, its expectation is at most

1p∑u∈D∖{0}rC−C(u)+2p∑τ∈(D∖{0})3Pτ≢0rC−C(d(τ)).\frac{1}{p}\sum_{\mathbf{u}\in\mathcal{D}\setminus\{0\}}r_{C-C}(\mathbf{u})+\frac{2}{p}\sum_{\substack{\tau\in(\mathcal{D}\setminus\{0\})^3\\P_\tau\not\equiv0}}r_{C-C}(\mathbf{d}(\tau)).

For τ=(u,v,w)\tau=(\mathbf{u},\mathbf{v},\mathbf{w}), we have

rC−C(d(τ))=min⁡{rC−C(u),rC−C(v),rC−C(w)}≤(rC−C(u)rC−C(v)rC−C(w))13.r_{C-C}(\mathbf{d}(\tau))=\min\{r_{C-C}(\mathbf{u}),r_{C-C}(\mathbf{v}),r_{C-C}(\mathbf{w})\} \leq\left(r_{C-C}(\mathbf{u})r_{C-C}(\mathbf{v})r_{C-C}(\mathbf{w})\right)^{\frac{1}{3}}.

Enlarging the triple sum to D3\mathcal{D}^3 bounds the expected number of deleted points by

∣C∣2−∣C∣p+2p(∑u∈DrC−C(u)13)3=∣C∣2−∣C∣+2J(C)3p.\frac{|C|^2-|C|}{p}+\frac{2}{p}\left(\sum_{\mathbf{u}\in\mathcal{D}}r_{C-C}(\mathbf{u})^{\frac{1}{3}}\right)^3 =\frac{|C|^2-|C|+2J(C)^3}{p}.

Fix a choice of α\alpha for which the number of deleted points does not exceed this bound, and set A=ϕα(C′)A=\phi_\alpha(C'). The restriction of ϕα\phi_\alpha to C′C' is injective: if distinct x,y∈C′\mathbf{x},\mathbf{y}\in C' had the same image, then u=x−y\mathbf{u}=\mathbf{x}-\mathbf{y} would satisfy u≠0\mathbf{u}\ne\mathbf{0} and ϕα(u)=0\phi_\alpha(\mathbf{u})=0, so x∈Ru\mathbf{x}\in R_{\mathbf{u}} would have been deleted. Consequently, ∣A∣=∣C′∣|A|=|C'|, which gives (17).

It remains to show that (A−A)2(A-A)^2 contains no nondegenerate geometric progression. Suppose that nonzero differences δ1,δ2,δ3∈A−A\delta_1,\delta_2,\delta_3\in A-A satisfy δ22=δ1δ3\delta_2^2=\delta_1\delta_3. Choose nonzero vectors u,v,w∈C′−C′\mathbf{u},\mathbf{v},\mathbf{w}\in C'-C' whose images under ϕα\phi_\alpha are δ1,δ2,δ3\delta_1,\delta_2,\delta_3, respectively. For τ=(u,v,w)\tau=(\mathbf{u},\mathbf{v},\mathbf{w}), we then have Pτ(α)=0P_\tau(\alpha)=0. If Pτ≢0P_\tau\not\equiv0, the set Rd(τ)R_{\mathbf{d}(\tau)} would have been deleted, contradicting d(τ)∈C′−C′\mathbf{d}(\tau)\in C'-C'. Therefore, Pτ≡0P_\tau\equiv0.

Unique factorization in Fp[X1,…,XL]\mathbb{F}_p[X_1,\ldots,X_L] implies that the three nonzero linear forms u⋅X\mathbf{u}\cdot X, v⋅X\mathbf{v}\cdot X, and w⋅X\mathbf{w}\cdot X are proportional. Every 2×22\times2 minor formed from their coefficient vectors vanishes modulo pp. Since each such minor is an integer of absolute value at most 18<p18<p, it also vanishes over Z\mathbb{Z}. Thus, the vectors are proportional over Q\mathbb{Q}, so we may write

u=λ1h,v=λ2h,w=λ3h,\mathbf{u}=\lambda_1\mathbf{h},\qquad\mathbf{v}=\lambda_2\mathbf{h},\qquad\mathbf{w}=\lambda_3\mathbf{h},

where h\mathbf{h} is a primitive integer vector and λ1,λ2,λ3\lambda_1,\lambda_2,\lambda_3 are nonzero integers. The coordinate bounds give ∣λj∣≤3|\lambda_j|\leq3. Since h\mathbf{h} is primitive, it is nonzero modulo pp, and substituting these expressions into Pτ≡0P_\tau\equiv0 yields

λ22≡λ1λ3(modp).\lambda_2^2\equiv\lambda_1\lambda_3\pmod p.

Together with ∣λ22−λ1λ3∣≤18<p|\lambda_2^2-\lambda_1\lambda_3|\leq18<p, this gives the integer equality λ22=λ1λ3\lambda_2^2=\lambda_1\lambda_3. The only factorizations of 11, 44, or 99 into two nonzero integers of absolute value at most 33 have factors of equal absolute value. It follows that

∣λ1∣=∣λ2∣=∣λ3∣,δ12=δ22=δ32.|\lambda_1|=|\lambda_2|=|\lambda_3|,\qquad\delta_1^2=\delta_2^2=\delta_3^2.

In particular, δ14=δ24\delta_1^4=\delta_2^4, so (16) shows that (A−A)2(A-A)^2 contains no nondegenerate geometric progression. This completes the proof.

The choice of the four digits 0,1,2,30,1,2,3 ensures that the polynomial identities left untreated by our deletion rule produce only degenerate progressions. This property fails for five consecutive digits. Indeed, the difference set of {0,1,2,3,4}L\{0,1,2,3,4\}^L contains h\mathbf{h}, 2h2\mathbf{h}, 4h4\mathbf{h}, where h\mathbf{h} is a coordinate vector, and

(2h⋅X)2−(h⋅X)(4h⋅X)≡0.(2\mathbf{h}\cdot X)^2-(\mathbf{h}\cdot X)(4\mathbf{h}\cdot X)\equiv0.

For any projection with t=ϕα(h)≠0t=\phi_\alpha(\mathbf{h})\ne0, these differences give the squared values t2t^2, 4t24t^2, 16t216t^2, which form a nondegenerate geometric progression when p>18p > 18. Since the displayed polynomial is identically zero, it vanishes with probability one, and the 2/p2/p bound for nonzero polynomials does not apply. Thus, extending this construction to five consecutive digits would require an additional argument to handle such identities.

A family with prescribed digit counts

To apply Lemma 4.1, we need a large family C\mathcal{C} for which J(C)3J(\mathcal{C})^3 is small compared with p∣C∣p|\mathcal{C}|. We prescribe the digit counts so that all vectors in the family have the same weight. The weighted pair count in the full cube then factors over the coordinates and gives the required estimate for J(C)J(\mathcal{C}).

Proof of Proposition 1.4(ii). We choose a family in which each digit has a prescribed number of occurrences. First we estimate its size and the quantity J(CL)J(\mathcal{C}_L). We then choose LL so that the deletion bound in Lemma 4.1 leaves at least half the family.

Let LL be a positive multiple of 26, to be chosen below. For x∈{0,1,2,3}L\mathbf{x} \in\{0,1,2,3\}^L, write Nj(x)=∣{i:xi=j}∣N_j(\mathbf{x}) = |\{i : x_i=j\}| for j∈{0,1,2,3}j \in\{0,1,2,3\}, and set

CL={x∈{0,1,2,3}L:(N0(x),N1(x),N2(x),N3(x))=L26(5,8,8,5)}.(18)\mathcal{C}_L = \left\{ \mathbf{x} \in\{0,1,2,3\}^L : (N_0(\mathbf{x}),N_1(\mathbf{x}),N_2(\mathbf{x}),N_3(\mathbf{x})) = \frac{L}{26}(5,8,8,5) \right\}. \tag*{(18)}

With

h0=log⁡26−513log⁡5−813log⁡8,h_0 = \log26-\frac{5}{13}\log5-\frac{8}{13}\log8,

Stirling’s formula gives absolute constants c1,c2>0c_1,c_2>0 such that

c1L−32eh0L≤∣CL∣=L!((5L26)!)2((8L26)!)2≤c2L−32eh0L.(19)c_1L^{-\frac{3}{2}}e^{h_0L}\leq|\mathcal{C}_L|=\frac{L!}{\left(\left(\frac{5L}{26}\right)!\right)^2\left(\left(\frac{8L}{26}\right)!\right)^2}\leq c_2L^{-\frac{3}{2}}e^{h_0L}. \tag*{(19)}

To estimate J(CL)J(\mathcal{C}_L), set (w0,w1,w2,w3)=(1,3,3,1)(w_0,w_1,w_2,w_3)=(1,3,3,1) and define

ω(x)=∏i=1Lwxi,x∈{0,1,2,3}L.\omega(\mathbf{x})=\prod_{i=1}^{L}w_{x_i},\qquad\mathbf{x}\in\{0,1,2,3\}^L.

For every x∈CL\mathbf{x}\in\mathcal{C}_L, we have

ω(x)=3N1(x)+N2(x)=38L13.\omega(\mathbf{x})=3^{N_1(\mathbf{x})+N_2(\mathbf{x})}=3^{\frac{8L}{13}}.

For j∈{−3,…,3}j\in\{-3,\ldots,3\}, let

ρ(j)=∑a,b∈{0,1,2,3}a−b=jwawb,\rho(j)=\sum_{\substack{a,b\in\{0,1,2,3\}\\a-b=j}}w_aw_b,

so that

ρ(0)=20,ρ(±1)=15,ρ(±2)=6,ρ(±3)=1.(20)\rho(0)=20,\qquad\rho(\pm1)=15,\qquad\rho(\pm2)=6,\qquad\rho(\pm3)=1. \tag*{(20)}

Write DL=CL−CLD_L=\mathcal{C}_L-\mathcal{C}_L and use the difference multiplicity rCL−CLr_{\mathcal{C}_L-\mathcal{C}_L} defined in Lemma 4.1. For each u∈DLu \in\mathcal{D}_L, we obtain

316L13rCL−CL(u)=∑x,y∈CLx−y=uω(x)ω(y)≤∑x,y∈{0,1,2,3}Lx−y=uω(x)ω(y)=∏i=1Lρ(ui).(21)3^{\frac{16L}{13}}r_{\mathcal{C}_L-\mathcal{C}_L}(\mathbf{u}) = \sum_{\substack{\mathbf{x},\mathbf{y}\in\mathcal{C}_L\\ \mathbf{x}-\mathbf{y}=\mathbf{u}}} \omega(\mathbf{x})\omega(\mathbf{y}) \le \sum_{\substack{\mathbf{x},\mathbf{y}\in\{0,1,2,3\}^{L}\\ \mathbf{x}-\mathbf{y}=\mathbf{u}}} \omega(\mathbf{x})\omega(\mathbf{y}) =\prod_{i=1}^{L}\rho(u_i). \tag*{(21)}

Taking cube roots and enlarging the sum to {−3,…,3}L\{-3,\ldots,3\}^{L} gives

J(CL)=∑u∈DLrCL−CL(u)13≤3−16L39∑u∈{−3,…,3}L∏i=1Lρ(ui)13=3−16L39s0L,J(\mathcal{C}_L)=\sum_{\mathbf{u}\in\mathcal{D}_L}r_{\mathcal{C}_L-\mathcal{C}_L}(\mathbf{u})^{\frac{1}{3}} \le3^{-\frac{16L}{39}}\sum_{\mathbf{u}\in\{-3,\ldots,3\}^{L}}\prod_{i=1}^{L}\rho(u_i)^{\frac{1}{3}} =3^{-\frac{16L}{39}}s_0^{L},

where

s0=203+2153+263+2,b0=3log⁡s0−1613log⁡3.(22)s_0=\sqrt[3]{20}+2\sqrt[3]{15}+2\sqrt[3]{6}+2,\qquad b_0=3\log s_0-\frac{16}{13}\log3. \tag*{(22)}

In particular,

J(CL)3≤eb0L.(23)J(\mathcal{C}_L)^3\le e^{b_0L}. \tag*{(23)}

We now choose LL so that Lemma 4.1 gives a set of size at least ∣CL∣2\frac{|\mathcal{C}_L|}{2}. Since 0<h0≤log⁡40<h_0\le\log4 and s0>12s_0>12, we have

b0>3log⁡12−1613log⁡3>2log⁡4≥2h0>0.b_0>3\log12-\frac{16}{13}\log3>2\log4\ge2h_0>0.

Hence,

β0:=h0b0−h0∈(0,1).\beta_0:=\frac{h_0}{b_0-h_0}\in(0,1).

The logarithmic correction in the following choice of LL compensates for the factor L−32L^{-\frac{3}{2}} in (19). For a fixed constant K>0K>0, to be specified below, take

L=26⌈log⁡p−32log⁡log⁡p−K26(b0−h0)⌉.(24)L=26\left\lceil\frac{\log p-\frac{3}{2}\log\log p-K}{26(b_0-h_0)}\right\rceil. \tag*{(24)}

For all sufficiently large pp, this is a positive multiple of 2626 and satisfies L≍log⁡pL\asymp\log p. Applying Lemma 4.1 to CL\mathcal{C}_L and using (23), we obtain a set A⊂FpA\subset\mathbb{F}_p such that (A−A)2(A-A)^2 contains no nondegenerate geometric progression and

∣A∣≥∣CL∣−∣CL∣2p−2eb0Lp.|A|\ge|\mathcal{C}_L|-\frac{|\mathcal{C}_L|^2}{p}-\frac{2e^{b_0L}}{p}.

By (19) and (24),

2eb0Lp∣CL∣≤2c1L32e(b0−h0)Lp≤2e−Kc1(b0−h0)32.\frac{2e^{b_0L}}{p|\mathcal{C}_L|} \le\frac{2}{c_1}L^{\frac{3}{2}}\frac{e^{(b_0-h_0)L}}{p} \le\frac{2e^{-K}}{c_1(b_0-h_0)^{\frac{3}{2}}}.

Fix KK so that the last expression is at most 14\frac{1}{4}. Moreover, since eh0L≤pβ0e^{h_0L}\le p^{\beta_0} for all sufficiently large pp, we have

∣CL∣p≤c2pβ0−1=o(1).\frac{|\mathcal{C}_L|}{p}\le c_2p^{\beta_0-1}=o(1).

It follows that ∣A∣≥∣CL∣2|A|\ge\frac{|\mathcal{C}_L|}{2} for all sufficiently large pp.

It remains to estimate the size of AA in terms of pp. The definition of LL also gives

L≥log⁡p−32log⁡log⁡p−Kb0−h0−26.L \ge\frac{\log p-\frac{3}{2}\log\log p-K}{b_0-h_0}-26.

Therefore,

eh0L≥e−β0K−26h0pβ0(log⁡p)−32β0.e^{h_0L}\ge e^{-\beta_0K-26h_0}p^{\beta_0}(\log p)^{-\frac{3}{2}\beta_0}.

Combining this with (19), ∣A∣≥∣CL∣2|A|\ge\frac{|C_L|}{2}, and L≍log⁡pL\asymp\log p, we obtain

∣A∣≫pβ0(log⁡p)32(1+β0),β0=h0b0−h0.(25)|A|\gg\frac{p^{\beta_0}}{(\log p)^{\frac{3}{2}(1+\beta_0)}},\qquad\beta_0=\frac{h_0}{b_0-h_0}. \tag*{(25)}

To verify that β0>726\beta_0>\frac{7}{26}, it suffices to show 33h0>7b033h_0>7b_0, since b0−h0>0b_0-h_0>0. Substituting the definitions yields

13(33h0−7b0)=log⁡(26429⋅31125165⋅8264⋅s0273).(26)13(33h_0-7b_0)=\log\left(\frac{26^{429}\cdot3^{112}}{5^{165}\cdot8^{264}\cdot s_0^{273}}\right). \tag*{(26)}

The integer inequalities

27153>20⋅10003,24673>15⋅10003,18183>6⋅100032715^3>20\cdot1000^3,\qquad2467^3>15\cdot1000^3,\qquad1818^3>6\cdot1000^3

imply s0<132851000s_0<\frac{13285}{1000}, while direct integer arithmetic gives

26429⋅3112⋅1000273>5165⋅8264⋅13285273.26^{429}\cdot3^{112}\cdot1000^{273}>5^{165}\cdot8^{264}\cdot13285^{273}.

Thus, the ratio in (26) is greater than 1, proving β0>726\beta_0>\frac{7}{26}. Since

pβ0−726(log⁡p)32(1+β0)⟶∞,\frac{p^{\beta_0-\frac{7}{26}}}{(\log p)^{\frac{3}{2}(1+\beta_0)}}\longrightarrow\infty,

the bound (25) implies ∣A∣≥p726|A|\ge p^{\frac{7}{26}} for all sufficiently large pp. Finally, multiplication by any κ∈Fp×\kappa\in\mathbb{F}_p^\times preserves the absence of a nondegenerate geometric progression, as noted after Definition 1.1. This completes the proof.

Numerically, β0≈0.2693287509\beta_0\approx0.2693287509, so 726\frac{7}{26} is a convenient rational exponent below the exponent obtained in (25). We do not claim that the digit set, the prescribed counts, or the weights are optimal, or that β0\beta_0 is the optimal lower exponent.

Further remarks

The preceding sections reduce the odd-dimensional extremal problem to dimension one. We now explain what is known in even dimensions, where the split and nonsplit forms behave differently. We then show why the condition ∣D∣≥p1−ε|D|\ge p^{1-\varepsilon}, for fixed ε>0\varepsilon>0, does not force a progression in an arbitrary set D⊂FpD\subset\mathbb{F}_p, and state the remaining questions.

Even dimensions

The distance estimate of Pham, Shen, Tran, and Xue [16], combined with a quantitative form of Roth’s theorem, gives a uniform sufficient threshold in even dimensions. This threshold is sharp for split forms. For the standard nonsplit form, the remaining extremal problem reduces to the plane. We first bound the cardinality of an arbitrary set containing no nondegenerate geometric progression. This will be applied to distance sets in both cases.

Lemma 5.1. There is an absolute constant a>0a > 0 such that, for every sufficiently large prime pp, every set D⊂FpD \subset\mathbb{F}_p containing no nondegenerate geometric progression satisfies

∣D∣≤pexp⁡[−a(log⁡plog⁡log⁡p)16].(27)|D| \le p \exp\left[-a\left(\frac{\log p}{\log\log p}\right)^{\frac{1}{6}}\right]. \tag*{(27)}

Consequently, for every c,K>0c, K > 0, there exists p0=p0(c,K)p_0 = p_0(c, K) such that

∣D∣≥cp(log⁡p)K⟹D contains a nondegenerate geometric progression|D| \ge\frac{cp}{(\log p)^K} \quad\Longrightarrow\quad D\text{ contains a nondegenerate geometric progression}

whenever p≥p0p \ge p_0 is prime and D⊂FpD \subset\mathbb{F}_p.

Proof. Let r3(N)r_3(N) be the largest cardinality of a subset of {0,…,N−1}\{0,\ldots,N-1\} containing no three distinct terms in arithmetic progression. By Raghavan [18 Theorem 1.4], there is an absolute constant b>0b > 0 such that

r3(N)≤Nexp⁡[−b(log⁡Nlog⁡log⁡N)16]r_3(N) \le N \exp\left[-b\left(\frac{\log N}{\log\log N}\right)^{\frac{1}{6}}\right]

for all sufficiently large NN.

Suppose that DD contains no nondegenerate geometric progression. Choose a primitive root γ∈Fp×\gamma\in\mathbb{F}_p^\times and let

S={j∈{0,…,p−2}:γj∈D}.S = \{j \in\{0,\ldots,p-2\} : \gamma^j \in D\}.

If SS contained integers u,u+v,u+2vu,u+v,u+2v with v>0v > 0, then γu,γu+v,γu+2v\gamma^u,\gamma^{u+v},\gamma^{u+2v} would be three distinct nonzero elements of DD in geometric progression. Their distinctness follows because their exponents belong to {0,…,p−2}\{0,\ldots,p-2\}. Thus, ∣S∣≤r3(p−1)|S| \le r_3(p-1) and

∣D∣≤1+r3(p−1)≤pexp⁡[−a(log⁡plog⁡log⁡p)16]|D| \le1+r_3(p-1) \le p\exp\left[-a\left(\frac{\log p}{\log\log p}\right)^{\frac{1}{6}}\right]

for all sufficiently large pp, with a suitable absolute constant a>0a > 0. The final assertion follows since

exp⁡[−a(log⁡plog⁡log⁡p)16]=o((log⁡p)−K)\exp\left[-a\left(\frac{\log p}{\log\log p}\right)^{\frac{1}{6}}\right] = o((\log p)^{-K})

for every fixed K>0K > 0. This completes the proof. □

A uniform sufficient threshold. We apply the preceding lemma to the recent distance estimate. For a quadratic form QQ on Fp2m\mathbb{F}_p^{2m}, let

ΔQ(E)={Q(x−y):x,y∈E}.\Delta_Q(E) = \{Q(x-y) : x,y \in E\}.

A nondegenerate form on Fp2m\mathbb{F}_p^{2m} is called split if it has an mm-dimensional totally isotropic subspace, and nonsplit otherwise. In the nonsplit case, the largest dimension of a totally isotropic subspace is m−1m-1. Pham, Shen, Tran, and Xue [16 Theorem 1.1 and Remark 1.2] proved that, for every nondegenerate QQ and every E⊂Fp2mE \subset\mathbb{F}_p^{2m} with ∣E∣>pm|E| > p^m,

∣ΔQ(E)∣≥c0plog⁡2(2p),(28)|\Delta_Q(E)| \ge\frac{c_0p}{\log^2(2p)}, \tag*{(28)}

where c0>0c_0 > 0 is absolute. If QQ is nonsplit, the same conclusion holds when ∣E∣=pm|E| = p^m; in fact, their estimate gives ∣ΔQ(E)∣≥c0plog⁡(2p)|\Delta_Q(E)| \ge\frac{c_0p}{\log(2p)} in this case, after decreasing c0c_0 if necessary. The constants are independent of mm.

Corollary 5.2. There is an absolute constant p0p_0 such that the following holds for every odd prime p≥p0p \ge p_0 and every integer m≥1m \ge1. Let QQ be a nondegenerate quadratic form on Fp2m\mathbb{F}_p^{2m} and let E⊂Fp2mE \subset\mathbb{F}_p^{2m}. If ∣E∣>pm|E| > p^m, then ΔQ(E)\Delta_Q(E) contains a nondegenerate geometric progression. The same conclusion holds if ∣E∣=pm|E| = p^m and QQ is nonsplit.

Proof. For p≥2p \ge2, we have log⁡(2p)≤2log⁡p\log(2p) \le2\log p. Thus, (28) gives

∣ΔQ(E)∣≥c0p4(log⁡p)2.|\Delta_Q(E)| \ge\frac{c_0p}{4(\log p)^2}.

Lemma 5.1, with c=14c0c = \frac{1}{4}c_0 and K=2K = 2, proves the assertion when ∣E∣>pm|E| > p^m. The quoted estimate for nonsplit forms gives the conclusion at equality as well. Since c0c_0 is absolute, the lower bound on pp is independent of mm and QQ. This completes the proof.

Split forms. The strict inequality ∣E∣>pm|E| > p^m is sharp when QQ is split. Indeed, a split form has a totally isotropic subspace WW of dimension mm. Taking E=WE = W gives ∣E∣=pm|E| = p^m and ΔQ(E)={0}\Delta_Q(E) = \{0\}. For the standard form Q2mQ_{2m}, the split condition is that (−1)m(-1)^m be a square in Fp\mathbb{F}_p; see [17 Lemma 2.7]. Consequently,

MGP,2m(p)=pm(29)M_{\mathrm{GP},2m}(p) = p^m \tag*{(29)}

for all sufficiently large primes pp whenever mm is even or p≡1(mod4)p \equiv1 \pmod{4}.

Nonsplit forms. For the standard form, the remaining case is p≡3(mod4)p \equiv3 \pmod{4} and mm odd. A totally isotropic subspace then has dimension at most m−1m-1, so the preceding example does not establish sharpness of the exponent mm. A cylinder over a planar set gives a lower comparison with MGP,2(p)M_{\mathrm{GP},2}(p), and the extraction theorem gives the upper comparison. We record both inequalities together with the available planar bounds.

Proposition 5.3. There are absolute constants c>0c > 0, C≥1C \ge1, and p0p_0 such that, for every prime p≥p0p \ge p_0 with p≡3(mod4)p \equiv3 \pmod{4} and every odd integer m≥1m \ge1,

pm−1MGP,2(p)≤MGP,2m(p)≤Cpm−1MGP,2(p).(30)p^{m-1}M_{\mathrm{GP},2}(p) \le M_{\mathrm{GP},2m}(p) \le Cp^{m-1}M_{\mathrm{GP},2}(p). \tag*{(30)}

Moreover,

p726≤MGP,2(p)≤Cpexp⁡[−c(log⁡plog⁡log⁡p)16].(31)p^{\frac{7}{26}} \le M_{\mathrm{GP},2}(p) \le Cp \exp\left[-c\left(\frac{\log p}{\log\log p}\right)^{\frac{1}{6}}\right]. \tag*{(31)}

Proof. We first prove (30) and then establish the planar bounds in (31).

By the classification of quadratic forms over Fp\mathbb{F}_p, a linear change of coordinates identifies Q2mQ_{2m} with

u⋅v+Q2(z),u,v∈Fpm−1,z∈Fp2;u\cdot v+Q_2(z), \qquad u,v\in\mathbb{F}_p^{m-1}, \quad z\in\mathbb{F}_p^2;

see [17 Lemma 2.7]. If P⊂Fp2P\subset\mathbb{F}_p^2 has no nondegenerate geometric progression in its distance set, then, in these coordinates, the set

Fpm−1×{0}m−1×P\mathbb{F}_p^{m-1}\times\{0\}^{m-1}\times P

has pm−1∣P∣p^{m-1}|P| elements and distance set Δ(P)\Delta(P). This proves the lower bound in (30).

For the upper bound, let E⊂Fp2mE\subset\mathbb{F}_p^{2m} have no nondegenerate geometric progression in Δ(E)\Delta(E). Corollary 5.2 gives ∣E∣<pm|E|<p^m. The extraction theorem of Pham, Shen, and Xue [17 Theorem 1.1], with the residual binary form identified with Q2Q_2, gives a set P⊂Fp2P\subset\mathbb{F}_p^2 such that Δ(P)⊆Δ(E)\Delta(P)\subseteq\Delta(E) and

∣P∣≥1C0∣E∣/pm−11+∣E∣/pm+1≥∣E∣2C0pm−1,|P|\ge\frac{1}{C_0}\frac{|E|/p^{m-1}}{1+|E|/p^{m+1}}\ge\frac{|E|}{2C_0p^{m-1}},

where C0C_0 is absolute. Since ∣P∣≤MGP,2(p)|P|\le M_{\mathrm{GP},2}(p), the upper bound follows.

To prove (31), let A⊂FpA\subset\mathbb{F}_p be the set from Proposition 1.4(ii). Then Δ(A×{0})=(A−A)2\Delta(A\times\{0\})=(A-A)^2, giving MGP,2(p)≥p726M_{\mathrm{GP},2}(p)\ge p^{\frac{7}{26}}. Next, let P⊂Fp2P\subset\mathbb{F}_p^2 be nonempty and have no nondegenerate geometric progression in Δ(P)\Delta(P). Again, Corollary 5.2 gives ∣P∣<p|P|<p. Since Q2Q_2 is anisotropic, the planar estimate of Pham, Shen, Tran, and Xue [16 Theorem 1.3] yields

∣P∣log⁡(2∣P∣)≪∣Δ(P)∣≤pexp⁡[−a(log⁡plog⁡log⁡p)16],\frac{|P|}{\log(2|P|)}\ll|\Delta(P)|\le p\exp\left[-a\left(\frac{\log p}{\log\log p}\right)^{\frac{1}{6}}\right],

where the last inequality is Lemma 5.1. Since ∣P∣<p|P|<p, we obtain

∣P∣≪plog⁡(2p)exp⁡[−a(log⁡plog⁡log⁡p)16]≪pexp⁡[−c(log⁡plog⁡log⁡p)16]|P|\ll p\log(2p)\exp\left[-a\left(\frac{\log p}{\log\log p}\right)^{\frac{1}{6}}\right]\ll p\exp\left[-c\left(\frac{\log p}{\log\log p}\right)^{\frac{1}{6}}\right]

for a suitable absolute constant c>0c>0. This completes the proof.

In particular, for p≡3(mod4)p\equiv3\pmod{4} and odd mm,

pm−1+726≤MGP,2m(p)≪pmexp⁡[−c(log⁡plog⁡log⁡p)16].(32)p^{m-1+\frac{7}{26}}\le M_{\mathrm{GP},2m}(p)\ll p^m\exp\left[-c\left(\frac{\log p}{\log\log p}\right)^{\frac{1}{6}}\right]. \tag*{(32)}

Thus, MGP,2m(p)=o(pm)M_{\mathrm{GP},2m}(p)=o(p^m) as pp tends to infinity through these primes. The upper bound does not give a fixed power improvement below pmp^m, and the lower bound does not establish sharpness of the exponent mm. Determining the optimal exponent in this nonsplit case reduces, by (30), to determining it for MGP,2(p)M_{\mathrm{GP},2}(p) when p≡3(mod4)p \equiv3 \pmod{4}.

Cardinality alone is not enough

Lemma 5.1 forces a progression at every density c/(log⁡p)Kc/(\log p)^K. A fixed power bound is different: for every ε>0\varepsilon> 0, the size condition ∣D∣≥p1−ε|D| \ge p^{1-\varepsilon} does not force a nondegenerate geometric progression in an arbitrary set D⊂FpD \subset\mathbb{F}_p. Indeed, let ℓ0=⌊p−13⌋\ell_0 = \left\lfloor\frac{p-1}{3} \right\rfloor. Behrend’s construction [2] gives a set B⊂{0,…,ℓ0−1}B \subset\{0,\ldots,\ell_0-1\} with ∣B∣≥ℓ0exp⁡(−Clog⁡ℓ0)|B| \ge\ell_0 \exp(-C\sqrt{\log\ell_0}) containing no arithmetic progression of three distinct terms. Choose a generator γ\gamma of Fp×\mathbb{F}_p^\times and set D={γb:b∈B}D = \{\gamma^b : b \in B\}. A nondegenerate geometric progression in DD would give distinct b1,b2,b3∈Bb_1,b_2,b_3 \in B satisfying b1+b3≡2b2(modp−1)b_1+b_3 \equiv2b_2 \pmod{p-1}. Since ∣b1+b3−2b2∣≤2(ℓ0−1)<p−1|b_1+b_3-2b_2| \le2(\ell_0-1) < p-1, this congruence would imply b1+b3=2b2b_1+b_3=2b_2 over the integers, contrary to the choice of BB.

After increasing the absolute constant CC if necessary, we obtain

∣D∣=∣B∣≥pexp⁡(−Clog⁡p)|D| = |B| \ge p\exp(-C\sqrt{\log p})

for all sufficiently large pp. For every fixed ε>0\varepsilon> 0, this lower bound exceeds p1−εp^{1-\varepsilon} once pp is sufficiently large, showing that such a size condition alone does not force a nondegenerate geometric progression. The argument in Section 3 uses the additional structure of squared difference sets.

Remarks over arbitrary finite fields

Let qq be a power of a prime p>18p > 18. Define Δ(E)\Delta(E) and MGP,d(q)M_{\mathrm{GP},d}(q) over Fq\mathbb{F}_q by the same formulas as over Fp\mathbb{F}_p. The arguments in Sections 2–4 extend to this setting. For every integer m≥0m \ge0,

qmMGP,1(q)≤MGP,2m+1(q)≤4qmMGP,1(q).q^m M_{\mathrm{GP},1}(q) \le M_{\mathrm{GP},2m+1}(q) \le4q^m M_{\mathrm{GP},1}(q).

Moreover, there are absolute constants C≥1C \ge1 and q0q_0 such that

qm+726≤MGP,2m+1(q)≤Cqm+1625q^{m+\frac{7}{26}} \le M_{\mathrm{GP},2m+1}(q) \le Cq^{m+\frac{16}{25}}

whenever q≥q0q \ge q_0. These constants are independent of mm, pp, and the degree of the field extension. Questions of sharpness and the construction of extremal examples are more involved over general finite fields. To keep the presentation focused and consistent with recent progress on the distance problem over prime fields, we state and prove the main results only over prime fields.

Open questions

The one-dimensional extremal problem. Proposition 1.4 gives, for all sufficiently large primes pp,

p726≤MGP,1(p)≪p1625.p^{\frac{7}{26}} \le M_{\mathrm{GP},1}(p) \ll p^{\frac{16}{25}}.

The gap between these bounds leads to the following question.

Question 5.4. What is the order of magnitude of MGP,1(p)M_{\mathrm{GP},1}(p) as pp tends to infinity through the primes?

By Proposition 2.1,

MGP,2m+1(p)≍pmMGP,1(p),M_{\mathrm{GP},2m+1}(p) \asymp p^{m}M_{\mathrm{GP},1}(p),

with absolute implied constants, so an answer to Question 5.4 would also determine the extremal order in every odd dimension.

Other arithmetic patterns. A natural next question is how large EE must be for Δ(E)\Delta(E) to contain other patterns, such as a nontrivial three-term arithmetic progression or a geometric progression of length at least four.

On (u,s)(u,s)–Salem sets. Fraser introduced (u,s)(u,s)–Salem sets and studied their distance sets in [6]; see also [5] for subsequent results. It would be interesting to determine whether suitable (u,s)(u,s)–Salem hypotheses lower the size thresholds for geometric progressions proved in this paper.

Acknowledgement. D. T. Tran would like to thank the Vietnam Institute for Advanced Study in Mathematics (VIASM) for its warm hospitality and excellent research environment.

References

  1. [1]O. Ahmadi and I. E. Shparlinski, Geometric progressions in sumsets over finite fields, Monatshefte für Mathematik, 152(3) (2007), 177–185.DOI
  2. [2]F. A. Behrend, On sets of integers which contain no three terms in arithmetical progression, Proceedings of the National Academy of Sciences of the United States of America, 32(12) (1946), 331–332.pubmed.ncbi.nlm.nih.gov/16578230
  3. [3]T. F. Bloom and O. Sisask, An improvement to the Kelley–Meka bounds on three-term arithmetic progressions, arXiv:2309.02353, 2023.arxiv.org/abs/2309.02353
  4. [4]J. Bourgain, N. H. Katz, and T. Tao, A sum-product estimate in finite fields, and applications, Geometric and Functional Analysis, 14(1) (2004), 27–57.
  5. [5]D. Cheong, G. Ge, D. Koh, T. Pham, D. T. Tran, and T. Zhang, Additive structures imply more distances in F_q^d, arXiv:2510.26364, 2025.arxiv.org/abs/2510.26364
  6. [6]J. M. Fraser, L^p averages of the Fourier transform in finite fields, Indiana University Mathematics Journal, to appear. arXiv:2407.08589.arxiv.org/abs/2407.08589
  7. [7]D. Hart, A. Iosevich, D. Koh, and M. Rudnev, Averages over hyperplanes, sum-product theory in vector spaces over finite fields and the Erdős–Falconer distance conjecture, Transactions of the American Mathematical Society, 363(6) (2011), 3255–3275.arxiv.org/abs/0707.3473
  8. [8]A. Iosevich, D. Koh, and H. Parshall, On the quotient set of the distance set, Moscow Journal of Combinatorics and Number Theory, 8(2) (2019), 103–115.DOI
  9. [9]A. Losevich, D. Koh, and F. Rakhmonov, The quotient set of the quadratic distance set over finite fields, Forum Mathematicum, 36(5) (2024), 1341–1358.
  10. [10]A. Losevich and M. Rudnev, Erdős distance problem in vector spaces over finite fields, Transactions of the American Mathematical Society, 359(12) (2007), 6127–6142.
  11. [11]A. Losevich, D. Koh, and M. Lewko, Finite field restriction estimates for the paraboloid in high even dimensions, Journal of Functional Analysis, 278(11) (2020), 108450.
  12. [12]D. Koh, M. Q. Pham, and T. Pham, Structural theorems on the distance sets over finite fields, Forum Mathematicum, 35(4) (2023), 925–938.arxiv.org/abs/2111.14076
  13. [14]G. Petridis, New proofs of Plünnecke-type estimates for product sets in groups, Combinatorica, 32(6) (2012), 721–733.
  14. [15]T. Pham, Group action and L²-norm estimates of geometric problems, Journal of Mathematical Analysis and Applications, 531(1) (2024), 127748.DOI
  15. [16]T. Pham, C.-Y. Shen, D. T. Tran, and B. Xue, Quadratic distances in even dimensions over prime fields, arXiv:2610.01795v2, 2026.arxiv.org/abs/2610.01795
  16. [17]T. Pham, C.-Y. Shen, and B. Xue, On Erdős–Falconer distance problem in even dimensions, arXiv:arXiv:2607.17324v1, 2026.arxiv.org/abs/2607.17324
  17. [18]R. Raghavan, Improved bounds for 3-progressions, arXiv:2603.27045v3, 2026.arxiv.org/abs/2603.27045
  18. [19]K. F. Roth, On certain sets of integers, Journal of the London Mathematical Society, 28(1) (1953), 104–109.DOI
  19. [20]I. E. Shparlinski, Geometric progressions in vector sumsets over finite fields, Finite Fields and Their Applications, 68 (2020), 101747.DOI
  20. [21]E. Szemerédi, On sets of integers containing no k elements in arithmetic progression, Acta Arithmetica, 27(1) (1975), 199–245.

Paper details

vlnc:W_ce5c37pc
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents