betaFeedback

WorksMathematics

Sharp L6L^6 Fourier moments and (3,δδ)-energy estimates for fractal measures on the truncated cone in R3\R^3

Abstract

For a δδ-separated (δ,d)(δ,d)-Katz–Tao set PP on the truncated cone in R3\R^3, 1<d≤21<d\le 2, we prove the sharp energy bound \E_{3,δ}(P)\lesssim_{\varepsilon} δ^{-\varepsilon} \max\{δ^{-(3d+2)},δ^{-(5d-1)}\}. The proof applies the Fu–Ren incidence theorem to triples with a fixed sum. As a corollary we also deduce a sharp L6L^6 moment estimate for the Fourier transform of Frostman measures supported on the cone.

Introduction

The main results for the cone

We consider the truncated cone

C={(rcos⁡θ,rsin⁡θ,r):12≤r≤1},\mathcal{C}=\{(r\cos\theta,r\sin\theta,r):\tfrac{1}{2}\leq r\leq1\},

and the key notation associated with it

Q(x,y,t)=t2−x2−y2,e(θ)=(cos⁡θ,sin⁡θ).Q(x,y,t)=t^2-x^2-y^2,\qquad e(\theta)=(\cos\theta,\sin\theta).

A probability measure μ\mu on C\mathcal{C} is called dd-Frostman if there is a constant CC such that μ(B(x,r))≤Crd\mu(B(x,r))\leq Cr^d for all 0<r≤20<r\leq2. We use μ^(ξ)=∫e−2πix⋅ξ dμ(x)\widehat{\mu}(\xi)=\int e^{-2\pi i x\cdot\xi}\,d\mu(x).

Theorem 1.1. Let 1<d≤21<d\leq2, and let μ\mu be a probability measure supported on C\mathcal{C} such that

μ(B(x,r))≤Crd,0<r≤2.(1)\mu(B(x,r))\leq Cr^d,\qquad0<r\leq2. \tag*{(1)}

For every ε>0\varepsilon>0 and R≥2R\geq2,

∫B(0,R)∣μ^(ξ)∣6 dξ≲ε,CRεmax⁡{R5−3d,R2−d}.(2)\int_{B(0,R)}|\widehat{\mu}(\xi)|^6\,d\xi\lesssim_{\varepsilon,C}R^\varepsilon\max\{R^{5-3d},R^{2-d}\}. \tag*{(2)}

When d≤1d\leq1 the supporting cone loses its relevance, as the measure could live entirely on a line segment.

The exponents 5−3d5-3d, 2−d2-d are both attainable, as explained in Remark 4.3.

In the last section we derive Theorem 1.1 from a discrete energy theorem, whose proof occupies most of this paper.

Key words and phrases. Additive energy, cone, Katz–Tao sets, planar incidences.

The author is partially supported by NSF grant DMS-2349828.

We call a subset PP of a metric space a (δ,d,C)(\delta,d,C)-KT set if (and we will call this a KT non-concentration condition; KT is short for Katz–Tao who first showed the relevance of such sets for continuous problems)

#(P∩B(x,ρ))≤C(ρ/δ)d,δ≤ρ≤2.(3)\#(P \cap B(x,\rho)) \le C(\rho/\delta)^d,\qquad\delta\le\rho\le2. \tag*{(3)}

Let P⊂CP \subset C be a finite δ\delta-separated set, N=#PN=\#P. Condition (3) gives N≲Cδ−dN\lesssim_C \delta^{-d}. We define the (3,δ)(3,\delta)-energy (in previous literature this was sometimes called (6,δ)(6,\delta)-energy) by

E3,δ(P)=#{(p1,…,p6)∈P6:∣p1+p2+p3−p4−p5−p6∣≤C∗δ}.\mathbb{E}_{3,\delta}(P)=\#\{(p_1,\ldots,p_6)\in P^6: |p_1+p_2+p_3-p_4-p_5-p_6|\le C_*\delta\}.

When 1<d≤21<d\le2, we will prove the sharp bound

E3,δ(P)≲ε,C,C∗δ−εmax⁡{δ−(3d+2),δ−(5d−1)}.(4)\mathbb{E}_{3,\delta}(P)\lesssim_{\varepsilon,C,C_*}\delta^{-\varepsilon}\max\{\delta^{-(3d+2)},\delta^{-(5d-1)}\}. \tag*{(4)}

Both exponents occur for sets with #P≳δ−d\#P\gtrsim\delta^{-d}, as we will prove below.

The case 0<d≤10<d\le1 does not have any substance. The trivial upper bound O(N5)O(N^5) for the energy is easily attainable by placing the points on a line inside the cone, in arithmetic progression. This is precisely possible because the Hausdorff dimension of the line is at least as big as the “dimension” dd of the points.

From here on we assume that 1<d≤21<d\le2.

The flat example. The δ\delta-neighborhood of the truncated cone is nearly planar on an angular sector of width ∼δ\sim\sqrt{\delta}. Fix θ\theta and let e=e(θ)e=e(\theta) and e⊥=e′(θ)e^\perp=e'(\theta). Points of the form

p(r,y)=(re+ye⊥,r2+y2),12≤r≤34,0≤y≤cδ(5)p(r,y)=\left(re+ye^\perp,\sqrt{r^2+y^2}\right),\qquad\frac{1}{2}\le r\le\frac{3}{4},\qquad0\le y\le c\sqrt{\delta} \tag*{(5)}

lie on a fixed such sector of CC. Note that

p(r,y)=(re+ye⊥,r)+O(c2δ).(6)p(r,y)=(re+ye^\perp,r)+O(c^2\delta). \tag*{(6)}

Take rr in a δ\delta-progression and yy in a δd/2\delta^{d/2}-progression. Each grid has

n≍δ−1δδd/2=δ−(d+1)/2n\asymp\delta^{-1}\frac{\sqrt{\delta}}{\delta^{d/2}}=\delta^{-(d+1)/2}

points. The planar grid in (6) has ≳n5\gtrsim n^5 additive 6-tuples: independently choose five points in the grid, and a positive proportion of the choices determine a sixth point within the grid. By (6), these are O(δ)O(\delta)-additive 6-tuples on the cone.

We select J≲δ−(d−1)/2J\lesssim\delta^{-(d-1)/2} angular sectors (of width ∼δ\sim\sqrt{\delta}) with separation ≲δ(d−1)/2\lesssim\delta^{(d-1)/2}. For each sector use the grid above, and we let PP be their union. Then

#P≍Jn≳δ−d,E3,δ(P)≳Jn5=δ−(3d+2).\#P\asymp Jn\gtrsim\delta^{-d},\qquad\mathbb{E}_{3,\delta}(P)\gtrsim Jn^5=\delta^{-(3d+2)}.

The Katz–Tao condition follows by counting grid points in a ball. For ρ≤δ\rho\le\sqrt{\delta} a ball meets O(1)O(1) grids and has O((ρ/δ)(1+ρ/δd/2))O((\rho/\delta)(1+\rho/\delta^{d/2})) points. For ρ≥δ\rho\ge\sqrt{\delta} it meets O(ρ/δ(d−1)/2)O(\rho/\delta^{(d-1)/2}) grids and has O(ρδ−(d+1)/2)O(\rho\delta^{-(d+1)/2}) points in each. Both bounds are O((ρ/δ)d)O((\rho/\delta)^d), as needed.

The radial arithmetic progression example. Let

h=δd−1,Rh={12+jh:0≤j≲h−1},h=\delta^{d-1},\qquad R_h=\left\{\frac{1}{2}+jh:0\le j\lesssim h^{-1}\right\},

and let Θ\Theta be a δ\delta-separated subset of [0,π2][0,\frac{\pi}{2}], with #Θ≍δ−1\#\Theta\asymp\delta^{-1}. Define

P={(rcos⁡θ,rsin⁡θ,r):r∈Rh, θ∈Θ}.P = \{(r\cos\theta, r\sin\theta, r) : r \in R_h,\ \theta\in\Theta\}.

Then #P≍h−1δ−1=δ−d\#P \asymp h^{-1}\delta^{-1}=\delta^{-d}. Moreover, for δ≤ρ≤1\delta\leq\rho\leq1,

#(P∩B(x,ρ))≲(1+ρ/h)ρ/δ≲(ρ/δ)d.\#(P\cap B(x,\rho))\lesssim(1+\rho/h)\rho/\delta\lesssim(\rho/\delta)^d.

The height coordinate of every triple sum belongs to Rh+Rh+RhR_h+R_h+R_h, which contains only O(h−1)O(h^{-1}) elements. At each such height, the two remaining sum coordinates occupy O(δ−2)O(\delta^{-2}) squares of side O(δ)O(\delta) in the plane. Thus the N3≳δ−3dN^3\gtrsim\delta^{-3d} triple sums occupy at most O(h−1δ−2)=O(δ−(d+1))O(h^{-1}\delta^{-2})=O(\delta^{-(d+1)}) δ\delta-cubes in R3\mathbb{R}^3. Cauchy–Schwarz gives

E3,δ(P)≳N6δ−(d+1)≳δ−(5d−1).(7)E_{3,\delta}(P)\gtrsim\frac{N^6}{\delta^{-(d+1)}}\gtrsim\delta^{-(5d-1)}. \tag*{(7)}

The two lower bounds cross at d=3/2d=3/2. For 1<d≤3/21<d\leq3/2 the flat term dominates, while for 3/2≤d≤23/2\leq d\leq2 the radial term dominates. Our main theorem gives the corresponding upper bound. In fact, as a by-product of our method, a more refined bound emerges, one that takes into account the size NN of the set.

Theorem 1.2. Let 1<d≤21<d\leq2 and let P⊂CP\subset C be a δ\delta-separated (δ,d,C)(\delta,d,C)-KT set, with N=#PN=\#P. For every ε>0\varepsilon>0,

E3,δ(P)≲ε,C,C∗Nδ−εmax⁡{δ−(2d+2),δ−(4d−1)}.(8)E_{3,\delta}(P)\lesssim_{\varepsilon,C,C_*}N\delta^{-\varepsilon}\max\{\delta^{-(2d+2)},\delta^{-(4d-1)}\}. \tag*{(8)}

In particular, (4) holds and is sharp up to δ−ε\delta^{-\varepsilon} for sets of maximal cardinality.

The conjectures for the parabolic cylinder and the parabola

Consider the truncated parabolic cylinder

Y={(x,x2,z):x∈[0,1], z∈[0,1]}⊂R3.\mathcal{Y}=\{(x,x^2,z):x\in[0,1],\ z\in[0,1]\}\subset\mathbb{R}^3.

The cylinder is often a good model for the cone, as both surfaces have zero Gaussian curvature. The analogous conjecture we make for a (δ,d)(\delta,d)-KT set S⊂YS\subset\mathcal{Y}, 1<d≤21<d\leq2, is

E3,δ(S)≲εδ−εmax⁡{δ−(3d+2),δ−(5d−1)}.(9)E_{3,\delta}(S)\lesssim_{\varepsilon}\delta^{-\varepsilon}\max\{\delta^{-(3d+2)},\delta^{-(5d-1)}\}. \tag*{(9)}

The two exponents are the same as for the cone, and their justification is also similar. The first obstruction comes from δ\sqrt{\delta}-wide strips of the cylinder, which are nearly planar at scale δ\delta. For the second, we use a set which projects to an arithmetic progression in the xx-coordinate. Both constructions yield the same exponents as above, via the same type of counting. The cylinder conjecture is at least as hard as the corresponding 3-energy conjecture for the parabola, introduced in [6].

Conjecture 1.3 (3-energy on the parabola). Let 0<s≤10<s\leq1, and let A⊂{(x,x2):0≤x≤1}A\subset\{(x,x^2):0\leq x\leq1\} be a δ\delta-separated (δ,s,C)(\delta,s,C)-KT set. Then for every ε>0\varepsilon>0,

E3,δ(A)≲ε,C,C∗δ−εmax⁡{δ−3s,δ−(5s−1)}.(10)E_{3,\delta}(A)\lesssim_{\varepsilon,C,C_*}\delta^{-\varepsilon}\max\{\delta^{-3s},\delta^{-(5s-1)}\}. \tag*{(10)}

Indeed, let A⊂{(x,x2):0≤x≤1}A\subset\{(x,x^2):0\leq x\leq1\} be a maximal (δ,s)(\delta,s)-KT set, where s=d−1s=d-1, and let ZZ be a δ\delta-arithmetic progression of length comparable to δ−1\delta^{-1}. Then S=A×ZS=A\times Z, with the coordinates ordered as above, is a (δ,s+1)(\delta,s+1)-KT set. Each approximate additive 6-tuple in AA can be paired with ≳δ−5\gtrsim\delta^{-5} exact additive 6-tuples in ZZ. Thus

E3,δ(S)≳δ−5E3,cδ(A)\mathrm{E}_{3,\delta}(S) \gtrsim\delta^{-5}\mathrm{E}_{3,c\delta}(A)

for a fixed c=O(1)c=O(1). Applying (9) with d=s+1d=s+1 would give

E3,δ(A)≲εδ−εmax⁡{δ−3s,δ−(5s−1)},0<s≤1,(11)\mathrm{E}_{3,\delta}(A) \lesssim_{\varepsilon} \delta^{-\varepsilon}\max\{\delta^{-3s},\delta^{-(5s-1)}\},\qquad0<s\leq1, \tag*{(11)}

which is precisely the conjectured parabola numerology. This gives a concrete sense in which the cylinder problem is harder. We do not know a comparable implication from the cone estimate to the parabola estimate.

Orponen [6] initiated the study of the δ\delta-energy problem for the parabola, deriving a power improvement over the (comparatively speaking, elementary) fourth-moment estimate, using Furstenberg-set theory. Subsequent progress is described in [2], where O’Regan and the author prove the bound E3,δ(A)≲εδ−7s/2−ε\mathrm{E}_{3,\delta}(A)\lesssim_{\varepsilon}\delta^{-7s/2-\varepsilon} for s≤2/3s\leq2/3. We refer to that paper for the intervening results and their references. More recently, Orponen and Shmerkin [7, Theorem 6.9] proved the sharp triple-sumset dimension formula dim⁡H(A+A+A)=3s\dim_{\mathrm{H}}(A+A+A)=3s for every ss-Ahlfors–David regular subset AA of the parabola with 0<s≤120<s\leq\frac{1}{2}. This Hausdorff-dimension result does not by itself imply the discretized energy estimate in Conjecture 1.3.

There is a contrast with familiar Fourier-analytic estimates. The cone and parabolic cylinder share similar critical exponents for a multitude of problems: they have the same critical ℓ2L6\ell^2 L^6 decoupling exponent, and comparable critical restriction and L4L^4 square-function phenomena. The cone is usually (both conceptually and technically) much harder, notably in reverse square-function estimates, see [4] and [1], because its curved and flat directions interact under rescaling. In our context however, the cone energy admits a full resolution, whereas the cylinder conjecture contains the unsolved parabola energy problem.

Remark 1.4 (Energy and the ring problem). For the parabola, Orponen’s improvement over the elementary energy exponent is connected to the discretized ring problem and to Furstenberg-set estimates. This is one motivation for the Furstenberg-theoretic methods used in [2]. We do not know an analogous implication from improved energy on the cone to a ring or sum–product theorem. The radial variable and the angular variable do not combine into the same one-dimensional additive–multiplicative structure. Our proof below uses planar incidences of lower complexity than the Furstenberg-set theorems known to have non-trivial sum-product consequences.

A warm-up: exact energy (=0) and Szemerédi–Trotter

We explain how the argument works in the technically simpler case when δ=0\delta=0. Let P⊂CP\subset C be finite, without any non-concentration assumptions. For each angle θ\theta, the segment

Lθ={(rcos⁡θ,rsin⁡θ,r):12≤r≤1}L_{\theta}=\{(r\cos\theta,r\sin\theta,r):\tfrac{1}{2}\leq r\leq1\}

is called a generator of the truncated cone. Let MM be the largest number of points of PP on any generator, and write

E3(P)=#{(p1,…,p6)∈P6:p1+p2+p3=p4+p5+p6}.\mathrm{E}_{3}(P)=\#\{(p_1,\ldots,p_6)\in P^6:p_1+p_2+p_3=p_4+p_5+p_6\}.

We prove the discrete analog of Theorem 1.2.

Proposition 2.1. For every finite P⊂CP \subset\mathbb{C},

E3(P)≲N13/3+NM4.E_3(P) \lesssim N^{13/3}+NM^4.

Remark 2.2. The estimate in Proposition 2.1 serves only as a warm-up for the transformation and -to some extent- for the incidence mechanism used below. Jing and Wu [5] recently proved that, for a finite set PP on a fixed irreducible algebraic surface of degree at least two (such as the cone), its fourth energy obeys

E2(P)≲εMN2+ε,E_2(P) \lesssim_{\varepsilon} MN^{2+\varepsilon},

where M≥1M \ge1 bounds the number of points on any line contained in the surface. Applying Young’s convolution inequality gives

E3(P)=∥1P∗1P∗1P∥22≤N2∥1P∗1P∥22≲εMN4+ε.E_3(P)=\lVert1_P * 1_P * 1_P\rVert_2^2 \le N^2\lVert1_P * 1_P\rVert_2^2 \lesssim_{\varepsilon} MN^{4+\varepsilon}.

This is stronger than the N13/3N^{13/3} term of Proposition 2.1 for bounded MM, though the bounds have different advantages when MM is large. We present the elementary Szemerédi–Trotter proof solely to introduce the nonlinear incidence transformation needed in the δ\delta-problem.

In preparation for the proof of Proposition 2.1, we distinguish the two possible cases for the triple sum of points on the cone. If pi=ri(e(θi),1)p_i=r_i(e(\theta_i),1) with 12≤ri≤1\frac{1}{2}\le r_i\le1 and q=p1+p2+p3q=p_1+p_2+p_3, then the definition of QQ gives

Q(q)=(r1+r2+r3)2−∣∑i=13rie(θi)∣2=2∑1≤i<j≤3rirj(1−cos⁡(θi−θj)).(12)Q(q)=(r_1+r_2+r_3)^2-\left|\sum_{i=1}^{3}r_i e(\theta_i)\right|^2 =2\sum_{1\le i<j\le3}r_i r_j\bigl(1-\cos(\theta_i-\theta_j)\bigr). \tag*{(12)}

To explain the second equal sign, expand the two squares separately:

(r1+r2+r3)2=∑i=13ri2+2∑i<jrirj,(r_1+r_2+r_3)^2=\sum_{i=1}^{3}r_i^2+2\sum_{i<j}r_i r_j,
∣∑i=13rie(θi)∣2=∑i=13ri2+2∑i<jrirje(θi)⋅e(θj)=∑i=13ri2+2∑i<jrirjcos⁡(θi−θj).\begin{aligned} \left|\sum_{i=1}^{3}r_i e(\theta_i)\right|^2 =\sum_{i=1}^{3}r_i^2+2\sum_{i<j}r_i r_j e(\theta_i)\cdot e(\theta_j) \\ &=\sum_{i=1}^{3}r_i^2+2\sum_{i<j}r_i r_j\cos(\theta_i-\theta_j). \end{aligned}

Here we used ∣e(θ)∣=1|e(\theta)|=1 and e(θi)⋅e(θj)=cos⁡(θi−θj)e(\theta_i)\cdot e(\theta_j)=\cos(\theta_i-\theta_j). Subtracting gives (12). Thus Q(q)≥0Q(q)\ge0.

We call a sum qq zero if Q(q)=0Q(q)=0 and nonzero if Q(q)>0Q(q)>0. Since every summand in (12) is nonnegative, a sum is zero exactly when all three points lie on one generator. Nonzero means that at least two of their angular directions differ. Only the nonzero case requires incidence geometry, while the zero case is trivial. We start with the nonzero case.

For q∈P+P+Pq\in P+P+P write

n(q)=#{(p1,p2,p3)∈P3:p1+p2+p3=q}.n(q)=\#\{(p_1,p_2,p_3)\in P^3:p_1+p_2+p_3=q\}.

Then E3(P)=∑qn(q)2E_3(P)=\sum_q n(q)^2 and ∑qn(q)=N3\sum_q n(q)=N^3. The distinguished feature of our argument is that we will derive uniform bounds for n(q)n(q), when Q(q)>0Q(q)>0. This allows us to fix qq. We will apply two transformations. The first one normalizes the geometry of points, to lay the ground for uniform estimates for n(q)n(q). The second one recasts triple sums contributing to n(q)n(q) as planar incidence estimates between points and lines.

A fixed sum and the linear Lorentz rescaling

Write q=(q1,q2,t)q=(q_1,q_2,t). Since the cone and its truncation are invariant under rotations in the horizontal (x,y)(x,y)-plane, we may rotate the coordinates so that q=(s,0,t)q=(s,0,t), where s=q12+q22≥0s=\sqrt{q_1^2+q_2^2}\ge0. This rotation preserves distances and additive relations. For a nonzero sum Q(q)>0Q(q)>0, we have t>st>s. Set h=t2−s2h=\sqrt{t^2-s^2} and consider the Lorentz rescaling

Bq(x,y,z)=(tx−szh,y,tz−sxh),Aq=h−1Bq.(13)B_q(x,y,z)=\left(\frac{tx-sz}{h},y,\frac{tz-sx}{h}\right),\qquad A_q=h^{-1}B_q. \tag*{(13)}

The map AqA_q is linear and satisfies Aqq=(0,0,1)A_qq=(0,0,1). Moreover, Q(Aqp)=h−2Q(p)Q(A_qp)=h^{-2}Q(p), so AqA_q maps the upper cone to itself. By linearity, if p1+p2+p3=qp_1+p_2+p_3=q, then

Aqp1+Aqp2+Aqp3=(0,0,1).A_qp_1+A_qp_2+A_qp_3=(0,0,1).

Apply AqA_q to all three points. Their images still lie on the upper cone, and their sum is now (0,0,1)(0,0,1). We may therefore write

Aqpi=ri′(e(θi′),1).A_qp_i=r'_i(e(\theta'_i),1).

The primed radii and angles generally differ from the original ones; in particular, the transformed radii need not belong to [1/2,1][1/2,1]. For points in a triple with sum qq, however, they satisfy absolute bounds. Indeed, writing the original points as pi=ri(e(θi),1)p_i=r_i(e(\theta_i),1), the formula for AqA_q gives

ri′=ri(t−scos⁡θi)h2=rit+s+ris(1−cos⁡θi)h2≥rit+s≥112.r'_i=\frac{r_i(t-s\cos\theta_i)}{h^2}=\frac{r_i}{t+s}+\frac{r_is(1-\cos\theta_i)}{h^2}\ge\frac{r_i}{t+s}\ge\frac{1}{12}.

Here we used ri≥1/2r_i\ge1/2, t=r1+r2+r3≤3t=r_1+r_2+r_3\le3, and 0≤s<t0\le s<t, hence t+s≤6t+s\le6. On the other hand, the normalized sum equation gives

r1′+r2′+r3′=1,∑i=13ri′e(θi′)=0.r'_1+r'_2+r'_3=1,\qquad\sum_{i=1}^{3}r'_ie(\theta'_i)=0.

Thus, for each ii, the triangle inequality implies

ri′=∣∑j≠irj′e(θj′)∣≤∑j≠irj′=1−ri′,r'_i=\left|\sum_{j\ne i}r'_je(\theta'_j)\right|\le\sum_{j\ne i}r'_j=1-r'_i,

so ri′≤1/2r'_i\le1/2. We have proved that every point in a contributing triple to qq has

112≤ri′≤12.\frac{1}{12}\le r'_i\le\frac{1}{2}.

Remark 2.3. We record a more refined observation. This is not needed in the discrete argument presented in this section, but will be needed when we deal with δ\delta-approximate incidences. If the transformed radii are ordered as r(1)′≥r(2)′≥r(3)′r'_{(1)}\ge r'_{(2)}\ge r'_{(3)}, then 2r(2)′+1/12≤12r'_{(2)}+1/12\le1. In particular, at least two of the three transformed radii belong to [1/12,11/24]⊂[1/24,1/2−1/24][1/12,11/24]\subset[1/24,1/2-1/24].

To simplify notation, from now on we omit the primes and write pi=ri(e(θi),1)p_i=r_i(e(\theta_i),1) for the transformed points.

The nonlinear change of coordinates

At this point, we could recast a triple contributing to qq as an incidence between points and ellipses having two degrees of freedom. The desired estimate would be a consequence of the generalized Szemerédi–Trotter theorem due to Pach and Sharir, [8]. We instead follow an alternative approach. The role of the nonlinear map VV is to straighten ellipses into lines. This perspective will be useful later when we estimate the δ\delta-energy, and apply an incidence theorem available for straight tubes.

For the transformed sum q=(0,0,1)q=(0,0,1), direct calculation gives

Q(q−p1−p2)=1−2(r1+r2)+2r1r2(1−cos⁡(θ1−θ2))=4r1r2[(12r1−1)(12r2−1)−cos⁡2θ1−θ22].(14)\begin{aligned} Q(q-p_1-p_2)&=1-2(r_1+r_2)+2r_1r_2(1-\cos(\theta_1-\theta_2))\\ &=4r_1r_2\left[\left(\frac{1}{2r_1}-1\right)\left(\frac{1}{2r_2}-1\right)-\cos^2\frac{\theta_1-\theta_2}{2}\right]. \tag*{(14)} \end{aligned}

For 0<r<1/20<r<1/2, choose a continuous angular coordinate θ\theta on a fixed interval of length less than 2π2\pi and define the non-linear

V(p)=V(r,θ)=2r1−2r e(θ/2),p=r(e(θ),1).(15)V(p)=V(r,\theta)=\sqrt{\frac{2r}{1-2r}}\,e(\theta/2),\qquad p=r(e(\theta),1). \tag*{(15)}

The notation V(p)V(p) means that the two parameters rr and θ\theta are the radial and angular coordinates of the cone point pp; thus V(pi)=V(ri,θi)V(p_i)=V(r_i,\theta_i). The angle is defined modulo 2π2\pi, but e(θ/2)e(\theta/2) changes sign when θ\theta is increased by 2π2\pi. We therefore use finitely many angular intervals on which θ\theta is chosen continuously, making VV single-valued.

Since r1+r2+r3=1r_1+r_2+r_3=1 and all three radii are positive, at most one of them can be greater than or equal to (or better said, equal to) 1/21/2. Thus every triple has at least two points whose transformed radii belong to (0,1/2)(0,1/2), where VV is well-defined. We apply VV only to these two points, which we choose to be p1,p2p_1,p_2, and split the count according to the three possible choices of pair. The third point is uniquely determined by the sum equation. Later, in § 3.3, we will take the two selected radii in the compact sub-interval [1/48,1/2−1/48][1/48,1/2-1/48], which is needed for the quantitative distortion estimates.

We next turn the equality

p1+p2+p3=q=(0,0,1)p_1+p_2+p_3=q=(0,0,1)

into an incidence between a point (namely V(p2)V(p_2)) and a line parametrized by V(p1)V(p_1). The role of p3p_3 will be solely coming from the fact that it belongs to the cone. The fact that it is in PP is irrelevant.

Set

ai=1−2ri2ri>0,ui=V(pi)=ai−1/2e(θi/2),i=1,2.a_i=\frac{1-2r_i}{2r_i}>0,\qquad u_i=V(p_i)=a_i^{-1/2}e(\theta_i/2),\qquad i=1,2.

Using e(α)⋅e(β)=cos⁡(α−β)e(\alpha)\cdot e(\beta)=\cos(\alpha-\beta), we have

u1⋅u2=cos⁡((θ1−θ2)/2)a1a2,cos⁡2θ1−θ22=a1a2(u1⋅u2)2.u_1\cdot u_2=\frac{\cos((\theta_1-\theta_2)/2)}{\sqrt{a_1a_2}},\qquad\cos^2\frac{\theta_1-\theta_2}{2}=a_1a_2(u_1\cdot u_2)^2.

Substituting into (14) gives

Q(q−p1−p2)=4r1r2a1a2(1−(u1⋅u2)2)=(1−2r1)(1−2r2)(1−(V(p1)⋅V(p2))2).\begin{aligned} Q(q-p_1-p_2)&=4r_1r_2a_1a_2(1-(u_1\cdot u_2)^2)\\ &=(1-2r_1)(1-2r_2)(1-(V(p_1)\cdot V(p_2))^2). \end{aligned}

Since 0<r1,r2<1/20<r_1,r_2<1/2, the prefactor is nonzero. Hence Q(q−p1−p2)=0Q(q-p_1-p_2)=0 if and only if

V(p1)⋅V(p2)=±1.(16)V(p_1)\cdot V(p_2)=\pm1. \tag*{(16)}

Fix p1p_1 and write V(p1)=(a,b)V(p_1)=(a,b) and V(p2)=(x,y)V(p_2)=(x,y). Since V(p1)≠0V(p_1)\ne0, equation (16) says

ax+by=1orax+by=−1ax+by=1 \qquad\text{or}\qquad ax+by=-1

Each equation describes a line in the (x,y)(x,y)-plane. Thus, for a fixed choice of sign, p1p_1 determines a line ℓp1\ell_{p_1}, and every compatible p2p_2 gives an incidence V(p2)∈ℓp1V(p_2)\in\ell_{p_1}.

We check that no multiplicities are introduced. First, the map p↦V(p)p\mapsto V(p) is injective. Indeed,

∣V(p)∣2=2r1−2r\lvert V(p)\rvert^2=\frac{2r}{1-2r}

and the right-hand side is strictly increasing for 0<r<1/20<r<1/2. Hence V(p)V(p) determines rr. Its direction determines θ/2\theta/2, and therefore θ\theta, once a continuous choice of angle has been fixed. Second, for a fixed sign the line ax+by=±1ax+by=\pm1 determines (a,b)(a,b) uniquely: the right-hand side is fixed and nonzero. Hence the map from p1p_1 to its line is injective as well. We count the two signs separately, so that writing u⋅v=1u\cdot v=1 and (−u)⋅v=−1(-u)\cdot v=-1 does not identify two lines within the same family.

Nonzero sums

Lemma 2.4. For every qq with Q(q)>0Q(q)>0, n(q)≲N4/3n(q)\lesssim N^{4/3}.

Proof. We recall the reduction from the previous two subsections. Since Q(q)>0Q(q)>0, after rotating the horizontal coordinates we may write q=(s,0,t)q=(s,0,t) with h=t2−s2>0h=\sqrt{t^2-s^2}>0. Thus the Lorentz map AqA_q is well-defined, maps the upper cone to itself, and satisfies Aqq=(0,0,1)A_q q=(0,0,1). By linearity, a triple with p1+p2+p3=qp_1+p_2+p_3=q satisfies Aqp1+Aqp2+Aqp3=(0,0,1)A_qp_1+A_qp_2+A_qp_3=(0,0,1). Writing the transformed points as ri(e(θi),1)r_i(e(\theta_i),1), comparison of the third coordinates gives r1+r2+r3=1r_1+r_2+r_3=1. In particular, at least two radii are strictly smaller than 1/21/2. This is needed in order to make the use of VV possible. Split according to the choice of the pair. For each such pair, (16) gives incidences between at most NN distinct points and at most NN distinct lines. Szemerédi–Trotter gives O(N4/3+N)O(N^{4/3}+N) incidences. Once the pair is fixed, the third point is uniquely determined. There are only three choices of pair and finitely many charts.

It follows immediately that

∑Q(q)>0n(q)2≤(sup⁡Q(q)>0n(q))∑qn(q)≲N13/3.(17)\sum_{Q(q)>0}n(q)^2\leq\left(\sup_{Q(q)>0}n(q)\right)\sum_q n(q)\lesssim N^{13/3}. \tag*{(17)}

:::

Zero sums

Lemma 2.5. The contribution to E3(P)E_3(P) from Q(q)=0Q(q)=0 is O(NM4)O(NM^4).

Proof. By (12), Q(q)=0Q(q)=0 forces all three points of a contributing triple to lie on the same generator. Two triples with the same zero sum lie on the same generator as each other. If a generator contains mm points, its energy is at most m5m^5 because five points determine the sixth. Summing over generators gives ∑m5≤M4∑m=NM4\sum m^5\leq M^4\sum m=NM^4.

Proof of Proposition 2.1. Combine (17) and Lemma 2.5.

Remark 2.6 (Exact and thickened energy). For a δ\delta-separated maximal-size KTKT set one always has M≲δ−1M\lesssim\delta^{-1}, and the exact theorem gives E3(P)≲δ−13/3+δ−(d+4)E_3(P)\lesssim\delta^{-13/3}+\delta^{-(d+4)}. This is not the sharp thickened exponent. In the latter problem a δ\sqrt{\delta}-wide plate, not merely a generator, behaves as a plane at tolerance O(δ)O(\delta). Ordinary Szemerédi–Trotter is therefore not the needed incidence input.

The δ\delta-energy estimate

This section is devoted to proving Theorem 1.2. All implicit constants may depend on CC and C∗C_\ast. Partition the bounded region containing P+P+PP+P+P into cubes of side cδc\delta, with c>0c>0 fixed. Let n(q)n(q) be the number of triples with sum in the cube qq. We may write

E3,δ(P)≲∑qn(q)2,∑qn(q)=N3.(18)E_{3,\delta}(P) \lesssim\sum_q n(q)^2,\qquad\sum_q n(q)=N^3. \tag*{(18)}

For a triple (p1,p2,p3)(p_1,p_2,p_3), its angular diameter is

diam⁡ang(p1,p2,p3):=max⁡1≤i<j≤3dist⁡S1(θi,θj),\operatorname{diam}_{\mathrm{ang}}(p_1,p_2,p_3):=\max_{1\leq i<j\leq3}\operatorname{dist}_{S^1}(\theta_i,\theta_j),

where dist⁡S1\operatorname{dist}_{S^1} is the shorter angular distance. We say that the triple has angular diameter σ\sigma when this number lies between σ/2\sigma/2 and σ\sigma. Identity (12), together with 1/2≤ri≤11/2\leq r_i\leq1, shows that Q(p1+p2+p3)≍σ2Q(p_1+p_2+p_3)\asymp\sigma^2: the largest angular separation contributes ∼σ2\sim\sigma^2, while the other two contribute O(σ2)O(\sigma^2).

Assume now σ2≫δ\sigma^2\gg\delta. For a sum cube qq and for a triple satisfying

p1+p2+p3=q+O(δ),diam⁡ang(p1,p2,p3)≍σp_1+p_2+p_3=q+O(\delta),\qquad\operatorname{diam}_{\mathrm{ang}}(p_1,p_2,p_3)\asymp\sigma

we may still conclude that

Q(q)≍σ2,Q(q)\asymp\sigma^2,

since Q(q)=Q(p1+p2+p3)+O(δ)Q(q)=Q(p_1+p_2+p_3)+O(\delta). The scales below C0δC_0\sqrt{\delta}, with C0C_0 a sufficiently large fixed constant, are included in the near-zero contribution of Lemma 3.3: their sums satisfy ∣Q(q)∣=O(δ)|Q(q)|=O(\delta). We apply Lorentz normalization only when σ≥C0δ\sigma\geq C_0\sqrt{\delta}. We work dyadically in this remaining range C0δ≤σ≲1C_0\sqrt{\delta}\leq\sigma\lesssim1.

Lemma 3.1 (The sector determined by a sum cube). Fix δ≪σ≤1\sqrt{\delta}\ll\sigma\leq1. There is a boundedly overlapping family Tσ\mathcal{T}_\sigma of angular arcs τ\tau of length ≲σ\lesssim\sigma and a rule assigning each sum cube qq with ∣Q(q)∣≲σ2|Q(q)|\lesssim\sigma^2 to one arc τ(q)∈Tσ\tau(q)\in\mathcal{T}_\sigma, with the following property. If

∣Q(q)∣≲σ2,p1+p2+p3=q+O(δ),|Q(q)|\lesssim\sigma^2,\qquad p_1+p_2+p_3=q+O(\delta),

then the directions of p1,p2,p3p_1,p_2,p_3 all lie in τ(q)\tau(q).

When ∣Q(q)∣=O(δ)|Q(q)|=O(\delta), we may use a family Tδ\mathcal{T}_{\sqrt{\delta}} consisting of arcs τ\tau of length ≲δ\lesssim\sqrt{\delta}.

Proof. We show the argument in the case δ≪σ≤1\sqrt{\delta}\ll\sigma\leq1. The other case follows by simple modifications. Write pi=ri(e(θi),1)p_i=r_i(e(\theta_i),1), R=∑iri≍1R=\sum_i r_i\asymp1, and x=∑irie(θi)x=\sum_i r_i e(\theta_i). Thus S:=p1+p2+p3=(x,R)S:=p_1+p_2+p_3=(x,R). Since S=q+O(δ)S=q+O(\delta) and δ≲σ2\delta\lesssim\sigma^2, Q(S)=Q(q)+O(δ)≲σ2Q(S)=Q(q)+O(\delta)\lesssim\sigma^2. The identity

∑i=13ri∣e(θi)−xR∣2=R−∣x∣2R=Q(S)R(19)\sum_{i=1}^3 r_i\left|e(\theta_i)-\frac{x}{R}\right|^2=R-\frac{|x|^2}{R}=\frac{Q(S)}{R} \tag*{(19)}

and ri≥1/2r_i\geq1/2 imply ∣e(θi)−x/R∣≲σ|e(\theta_i)-x/R|\lesssim\sigma for each ii. If σ\sigma is sufficiently small, then ∣x∣2=R2−Q(S)|x|^2=R^2-Q(S) gives ∣x∣≲1|x|\lesssim1, and hence

∣e(θi)−x∣x∣∣≲σ.\left|e(\theta_i)-\frac{x}{|x|}\right|\lesssim\sigma.

Writing v=(q1,q2)v=(q_1,q_2), we also have ∣v−x∣≲δ|v-x|\lesssim\delta, so ∣v∣≲1|v|\lesssim1 and

dist⁡S1(θi,arg⁡v)≲σ,i=1,2,3.(20)\operatorname{dist}_{S^1}(\theta_i,\arg v)\lesssim\sigma,\qquad i=1,2,3. \tag*{(20)}

It follows that there is L≲σL\lesssim\sigma so that each contributing triple θ1,θ2,θ3\theta_1,\theta_2,\theta_3 lies inside an LL-arc. Cover the circle with a boundedly overlapping collection Tσ\mathcal{T}_\sigma of 3L/23L/2-arcs, so that each LL-arc lies inside some τ∈Tσ\tau\in\mathcal{T}_\sigma.

The nearly zero contribution

For τ∈Tσ\tau\in\mathcal{T}_{\sigma} from Lemma 3.1, define

Cτ={r(e(θ),1):1/2≤r≤1, θ∈τ},Pτcone=P∩Cτ,Nτ=#Pτcone.C_{\tau}=\{r(e(\theta),1):1/2\le r\le1,\ \theta\in\tau\},\qquad P_{\tau}^{\mathrm{cone}}=P\cap C_{\tau},\qquad N_{\tau}=\#P_{\tau}^{\mathrm{cone}}.

Thus PτconeP_{\tau}^{\mathrm{cone}} consists of the original cone points whose angles belong to τ\tau.

Lemma 3.2. For δ≲σ≲1\sqrt{\delta}\lesssim\sigma\lesssim1,

max⁡τ∈TσNτ≲δ−dσd−1,∑τ∈TσNτ≲N.(21)\max_{\tau\in\mathcal{T}_{\sigma}}N_{\tau}\lesssim\delta^{-d}\sigma^{d-1},\qquad\sum_{\tau\in\mathcal{T}_{\sigma}}N_{\tau}\lesssim N. \tag*{(21)}

Proof. An angular sector of aperture O(σ)O(\sigma) on the truncated cone is covered by O(σ−1)O(\sigma^{-1}) balls of radius O(σ)O(\sigma). Apply (3) in each ball. The second assertion follows from bounded overlap.

Lemma 3.3 (Near-zero sums). Triples whose sum cube has ∣Q(q)∣=O(δ)|Q(q)|=O(\delta) contribute O(Nδ−2d−2)O(N\delta^{-2d-2}) to E3,δ(P)E_{3,\delta}(P).

Proof. Apply Lemma 3.1. For each output cube qq with ∣Q(q)∣=O(δ)|Q(q)|=O(\delta), the lemma places all triples counted by n(q)n(q) in the same sector τ(q)\tau(q). Within this sector, five points determine the sixth up to O(1)O(1) possibilities, since PP is δ\delta-separated. By bounded overlap the contribution is at most

∑τ∈TδNτ5≤(max⁡τNτ)4∑τNτ≲N(δ−d(δ)d−1)4=Nδ−2d−2.\sum_{\tau\in\mathcal{T}_{\sqrt{\delta}}}N_{\tau}^{5}\leq\left(\max_{\tau}N_{\tau}\right)^{4}\sum_{\tau}N_{\tau}\lesssim N\left(\delta^{-d}(\sqrt{\delta})^{d-1}\right)^{4}=N\delta^{-2d-2}.

We next deal with the far from zero-sums.

The normalization at angular scale σ\sigma for fixed qq

Consider a sum cube centered at qq such that Q(q)≍σ2≫δQ(q)\asymp\sigma^{2}\gg\delta. By Lemma 3.1, all triples contributing to n(q)n(q) use points in the single sector τ=τσ(q)\tau=\tau_{\sigma}(q). Let q=(q1,q2,t)q=(q_{1},q_{2},t). Rotate the horizontal coordinates so that (q1,q2)(q_{1},q_{2}) becomes (s,0)(s,0), where s=(q12+q22)1/2s=(q_{1}^{2}+q_{2}^{2})^{1/2}. This rotation preserves the value of Q(q)Q(q). It also preserves the angular diameter, in fact all angular parameters change by the same constant. Consequently, the comparison Q(q)≍σ2Q(q)\asymp\sigma^{2}, obtained before rotation, remains valid afterward. Since q=(s,0,t)q=(s,0,t) and t>0t>0, we have

t2−s2=Q(q)≍σ2>0,h:=(t2−s2)1/2=Q(q)≍σ.t^{2}-s^{2}=Q(q)\asymp\sigma^{2}>0,\qquad h:=(t^{2}-s^{2})^{1/2}=\sqrt{Q(q)}\asymp\sigma.

We now use AqA_{q} from (13). If

p1+p2+p3=q+O(δ)(22)p_{1}+p_{2}+p_{3}=q+O(\delta) \tag*{(22)}

then

Aqp1+Aqp2+Aqp3=(0,0,1)+O(δ/σ2).(23)A_{q}p_{1}+A_{q}p_{2}+A_{q}p_{3}=(0,0,1)+O(\delta/\sigma^{2}). \tag*{(23)}

Indeed, the symmetric matrix representing AqA_{q} has eigenvalues (t+s)/h2(t+s)/h^{2}, (t−s)/h2(t-s)/h^{2}, and 1/h1/h; the first two have eigenvectors (1,0,−1)(1,0,-1) and (1,0,1)(1,0,1). Hence its Euclidean operator norm is

∥Aq∥op=sup⁡∣v∣=1∣Aqv∣=t+sh2≲σ−2.\lVert A_{q}\rVert_{\mathrm{op}}=\sup_{|v|=1}|A_{q}v|=\frac{t+s}{h^{2}}\lesssim\sigma^{-2}.

Thus an additive error O(δ)O(\delta) becomes O(δ/σ2)O(\delta/\sigma^{2}) under AqA_{q}.

As explained in Remark 2.3, we may assume that the radii of at least two of the three points Aqp1,Aqp2,Aqp3A_qp_1,A_qp_2,A_qp_3 satisfy, say (assuming the error term O(δ/σ2)O(\delta/\sigma^2) is small enough)

1/48≤ri′,rj′≤1/2−1/48.1/48\leq r_i',r_j'\leq1/2-1/48.

Any three points on the circle belong to an arc of length at most 4π/34\pi/3, as follows by considering the three consecutive angular gaps. Fix a collection of 24 open arcs γ\gamma of length 3π/23\pi/2, with equally spaced starting points. Every arc of length 4π/34\pi/3 is contained in at least one of these γ\gamma. It follows that any triple Aq(p1),Aq(p2),Aq(p3)A_q(p_1),A_q(p_2),A_q(p_3) will have their polar angles θi′\theta_i' inside some γ\gamma. The only relevance is that the length of γ\gamma is <2π<2\pi, as used below.

Let τ\tau be the arc associated with qq via Lemma 3.1. We may and will assume it is centered at the origin. Indeed, after rotating the horizontal direction of qq, the original angles of all points in PτconeP_\tau^{\mathrm{cone}} satisfy ∣θ∣≲σ|\theta|\lesssim\sigma. This is the angular hypothesis used in Lemma 3.4 below. We now bound n(q)n(q) using this sector and the finite family of transformed-angle charts γ\gamma.

The planar map for a fixed $\gamma. We aim to do planar incidence geometry, so we shift the discussion to the plane. Recall our notation p=r(e(θ),1)p=r(e(\theta),1), Aqp=r′(e(θ′),1)A_qp=r'(e(\theta'),1). We define the planar map

T=Tq:(r,θ)⟼V(r′,θ′)=2r′1−2r′e(θ′/2).T=T_q:(r,\theta)\longmapsto V(r',\theta')=\sqrt{\frac{2r'}{1-2r'}}e(\theta'/2).

Here VV is the nonlinear map from Section 2. On the fixed chart we require 1/48≤r′≤1/2−1/481/48\leq r'\leq1/2-1/48 and the transformed angle θ′∈γ\theta'\in\gamma. Define the planar points

Pτ={T(r,θ):r(e(θ),1)∈Pτcone, 1/48≤r′≤1/2−1/48, θ′∈γ}.P_\tau=\{T(r,\theta):r(e(\theta),1)\in P_\tau^{\mathrm{cone}},\ 1/48\leq r'\leq1/2-1/48,\ \theta'\in\gamma\}.

The points of PτP_\tau lie in R2\mathbb{R}^2 and have cardinality O(Nτ)O(N_\tau). They arise from the points PτconeP_\tau^{\mathrm{cone}} on the cone-sector via two transformations, whose combined action is encoded by the useful planar map TT. They are first mapped via the linear map AqA_q, then via the nonlinear map VV (which has a linearizing effect, leading to lines). Since the domain of VV will need θ′\theta' in an interval shorter than 2π2\pi, we restrict PτP_\tau to an arc γ\gamma. This is a rather unimportant technicality.

The next lemma is instrumental in determining the KT non-concentration of the points PτP_\tau.

Lemma 3.4 (The map TT). Let q=(s,0,t)q=(s,0,t) and h=(t2−s2)1/2≍σh=(t^2-s^2)^{1/2}\asymp\sigma. The intermediate polar coordinates of AqpA_qp satisfy

r′=r(t−scos⁡θ)h2,tan⁡θ′2=t+shtan⁡θ2.(24)r'=\frac{r(t-s\cos\theta)}{h^2},\qquad\tan\frac{\theta'}{2}=\frac{t+s}{h}\tan\frac{\theta}{2}. \tag*{(24)}

Suppose ∣θ∣≲σ|\theta|\lesssim\sigma and 1/48≤r′≤1/2−1/481/48\leq r'\leq1/2-1/48. For a Euclidean ball of radius ww in the target VV-plane, the preimage under TT is contained in a region of radial width O(w)O(w) and angular width O(σw)O(\sigma w). All constants are uniform in σ\sigma.

Proof. The three coordinates of AqpA_qp are

x′=r(tcos⁡θ−s)h2,y′=rsin⁡θh,r′=r(t−scos⁡θ)h2.x'=\frac{r(t\cos\theta-s)}{h^2},\qquad y'=\frac{r\sin\theta}{h},\qquad r'=\frac{r(t-s\cos\theta)}{h^2}.

Since r′=x′2+y′2r'=\sqrt{x'^2+y'^2}, the half-angle identity gives

tan⁡θ′2=y′r′+x′=hsin⁡θ(t−s)(1+cos⁡θ)=t+shtan⁡θ2.\tan\frac{\theta'}{2}=\frac{y'}{r'+x'}=\frac{h\sin\theta}{(t-s)(1+\cos\theta)}=\frac{t+s}{h}\tan\frac{\theta}{2}.

Put k=(t+s)/h≍σ−1k=(t+s)/h\asymp\sigma^{-1}. For ∣θ∣≲σ|\theta|\lesssim\sigma,

dθ′dθ=k(1+tan⁡2(θ/2))1+k2tan⁡2(θ/2)≲σ−1,∣dθdθ′∣≲σ.\frac{d\theta'}{d\theta}=\frac{k(1+\tan^{2}(\theta/2))}{1+k^{2}\tan^{2}(\theta/2)}\lesssim\sigma^{-1},\qquad\left|\frac{d\theta}{d\theta'}\right|\lesssim\sigma.

Also t−scos⁡θ=(t−s)+s(1−cos⁡θ)≍σ2t-s\cos\theta=(t-s)+s(1-\cos\theta)\asymp\sigma^{2} and

r=h2r′t−scos⁡θ.r=\frac{h^{2}r'}{t-s\cos\theta}.

Differentiating shows ∣∂r/∂r′∣≲1|\partial r/\partial r'|\lesssim1 and ∣∂r/∂θ′∣≲1|\partial r/\partial\theta'|\lesssim1: indeed, the θ\theta-derivative of the coefficient is O(σ−1)O(\sigma^{-1}), which is multiplied by dθ/dθ′=O(σ)d\theta/d\theta'=O(\sigma).

It remains to take into account VV. On 1/48≤r′≤1/2−1/481/48\leq r'\leq1/2-1/48,

V(r′,θ′)=f(r′)e(θ′/2),f(r′)=2r′1−2r′.V(r',\theta')=f(r')e(\theta'/2),\qquad f(r')=\sqrt{\frac{2r'}{1-2r'}}.

Indeed,

f′(r′)=12r′(1−2r′)3/2,f'(r')=\frac{1}{\sqrt{2r'}(1-2r')^{3/2}},

so both ff and f′f' are bounded above and below by positive constants on this fixed radial interval. To see explicitly why VV has a uniformly bi-Lipschitz inverse on its image, take two points uj=V(rj′,θj′)=ρje(θj′/2)u_j=V(r'_j,\theta'_j)=\rho_j e(\theta'_j/2), where ρj=f(rj′)\rho_j=f(r'_j) and θj′∈γ\theta'_j\in\gamma. Then

∣u1−u2∣2=(ρ1−ρ2)2+ρ1ρ2∣e(θ1′/2)−e(θ2′/2)∣2=(ρ1−ρ2)2+4ρ1ρ2sin⁡2(θ1′−θ2′4)≍∣r1′−r2′∣2+∣θ1′−θ2′∣2.\begin{aligned} |u_1-u_2|^2 &=(\rho_1-\rho_2)^2+\rho_1\rho_2|e(\theta'_1/2)-e(\theta'_2/2)|^2\\ &=(\rho_1-\rho_2)^2+4\rho_1\rho_2\sin^2\left(\frac{\theta'_1-\theta'_2}{4}\right)\\ &\asymp|r'_1-r'_2|^2+|\theta'_1-\theta'_2|^2. \end{aligned}

For the last comparison, the mean value theorem controls ∣ρ1−ρ2∣|\rho_1-\rho_2| by ∣r1′−r2′∣|r'_1-r'_2| in both directions, while ∣θ1′−θ2′∣≤3π/2|\theta'_1-\theta'_2|\leq3\pi/2 ensures that ∣sin⁡((θ1′−θ2′)/4)∣≍∣θ1′−θ2′∣|\sin((\theta'_1-\theta'_2)/4)|\asymp|\theta'_1-\theta'_2|. The constants are independent of qq and σ\sigma. Equivalently, the inverse can be written as

r′=∣u∣22(1+∣u∣2),θ′=2Arg⁡γ/2(u),r'=\frac{|u|^2}{2(1+|u|^2)},\qquad\theta'=2\operatorname{Arg}_{\gamma/2}(u),

where Arg⁡γ/2\operatorname{Arg}_{\gamma/2} denotes the branch of the argument taking values in the lifted interval γ/2\gamma/2. This explains the role of restricting the transformed angle to an arc γ\gamma of length less than 2π2\pi: it allows a consistent single-valued angular branch (indeed, V(r′,θ′+2π)=−V(r′,θ′)V(r',\theta'+2\pi)=-V(r',\theta')) and gives the uniform chord–angle comparison above. The finite collection of such charts ensures that no contributing triple is lost. Consequently, the inverse image of a ww-ball under VV has r′r'-width and θ′\theta'-width O(w)O(w). The preceding derivative bounds for the inverse Lorentz coordinate map then imply radial width O(w)O(w) and angular width O(σw)O(\sigma w) in the original coordinates. □

The transformed incidence scale. Recall that

∥Aq∥op=sup⁡∣v∣=1∣Aqv∣=t+sh2≍σ−2.\|A_q\|_{\mathrm{op}}=\sup_{|v|=1}|A_qv|=\frac{t+s}{h^2}\asymp\sigma^{-2}.

Thus an additive error O(δ)O(\delta) becomes O(δ/σ2)O(\delta/\sigma^2). More precisely, Q(Aqv)=h−2Q(v)Q(A_qv)=h^{-2}Q(v) and Aqq=(0,0,1)A_qq=(0,0,1), so

Q((0,0,1)−Aqpi−Aqpj)=h−2Q(q−pi−pj)=O(δ/σ2).Q\bigl((0,0,1)-A_qp_i-A_qp_j\bigr)=h^{-2}Q(q-p_i-p_j)=O(\delta/\sigma^2).

For each a∈Pτa \in P_\tau define

La±={v∈R2:a⋅v=±1}\mathcal{L}_a^\pm= \{v \in\mathbb{R}^2 : a \cdot v = \pm1\}

and let

Lτ±={La±:a∈Pτ}.\mathcal{L}_\tau^\pm= \{\mathcal{L}_a^\pm: a \in P_\tau\}.

The identity (16) relating Q((0,0,1)−Aqpi−Aqpj)Q((0,0,1)-A_{q p_i}-A_{q p_j}) to V(Aqpi)⋅V(Aqpj)V(A_{q p_i})\cdot V(A_{q p_j}) shows that a compatible selected pair gives a point of PτP_\tau within O(δ/σ2)O(\delta/\sigma^2) of a line in Lτ+\mathcal{L}_\tau^+ or Lτ−\mathcal{L}_\tau^-. Set

η:=δ/σ2.\eta:= \delta/\sigma^2.

Lemma 3.5 (Rescaled Katz–Tao constant). Let τ∈Tσ\tau\in\mathcal{T}_\sigma, σ≫δ\sigma\gg\sqrt{\delta}. The planar point collection PτP_\tau and the line collections Lτ±\mathcal{L}_\tau^\pm are (η,d,Kσ)(\eta,d,K_\sigma)-KT sets, where

Kσ≲σ−d−1.(25)K_\sigma\lesssim\sigma^{-d-1}. \tag*{(25)}

Proof. By Lemma 3.4, the preimage under TT of a ww-ball with w≥ηw \geq\eta has radial width O(w)O(w) and angular width O(σw)O(\sigma w). It can be covered by O(σ−1)O(\sigma^{-1}) balls of radius O(σw)O(\sigma w). Since σw≥ση=δ/σ≥δ\sigma w \geq\sigma\eta= \delta/\sigma\geq\delta, the KT hypothesis for the original points on the cone (which translates into a similar condition in the (r,θ)(r,\theta) variables) gives

#{a∈Pτ:a∈B(v,w)}≲σ−1(σw/δ)d=σ−d−1(w/η)d.\#\{a \in P_\tau: a \in B(v,w)\} \lesssim\sigma^{-1}(\sigma w/\delta)^d = \sigma^{-d-1}(w/\eta)^d.

The map u↦{v:u⋅v=±1}u \mapsto\{v : u \cdot v = \pm1\} is bi-Lipschitz whenever ∣u∣|u| is bounded above and away from zero. Both requirements hold, since u=V(r′,θ′)u=V(r',\theta') and 1/48≤r′≤1/2−1/481/48 \leq r' \leq1/2-1/48. Therefore the same KT bound transfers from PτP_\tau to Lτ±\mathcal{L}_\tau^\pm. □\square

The incidence bound for one sum cube

We introduce the replacement for the Szemerédi–Trotter theorem. The Fu–Ren result takes advantage of the key features of our point-tube incidences derived in Lemma 3.5: non-concentration of both points and tubes.

Theorem 3.6 (Fu–Ren [3], Theorem 1.5). Let 1<d≤21<d\leq2 and 0<η<10<\eta<1. Let PP be a collection of points and L\mathcal{L} a collection of η\eta-neighborhoods of unit line segments in the plane. Assume that the point collection and the line-parameter collection satisfy the (η,d)(\eta,d) Katz–Tao bounds with constants KPK_P and KLK_L, respectively. Then, for every ϵ>0\epsilon>0,

I(P,L)≲ϵη−c−ϵ(KPKL)c(#P)1−c(#L)1−c,c=1max⁡{2d−1,2}.(26)I(P,\mathcal{L}) \lesssim_\epsilon\eta^{-c-\epsilon}(K_P K_L)^c(\#P)^{1-c}(\#\mathcal{L})^{1-c}, \qquad c=\frac{1}{\max\{2d-1,2\}}. \tag*{(26)}

Here II counts intersecting point-tube pairs.

We may now establish the upper bound for n(q)n(q).

Lemma 3.7 (Fixed-sum fibre). For a sum cube qq with Q(q)≲σ2≫δQ(q)\lesssim\sigma^2\gg\delta, let τ=τ(q)\tau=\tau(q) be its assigned sector from Lemma 3.1. Then, letting Nτ=#PτconeN_\tau=\#P_\tau^{\mathrm{cone}}

n(q)≲ϵδ−c−ϵσ−2cdNτ2−2c,c=1max⁡{2d−1,2}.(27)n(q)\lesssim_\epsilon\delta^{-c-\epsilon}\sigma^{-2cd}N_\tau^{2-2c}, \qquad c=\frac{1}{\max\{2d-1,2\}}. \tag*{(27)}

Proof. By Lemma 3.1, all triples counted by n(q)n(q) lie in PconeP^{\mathrm{cone}}, and the angular hypothesis of Lemma 3.4 holds on this sector after rotation. Normalize qq by AqA_q. Every normalized triple satisfies

r1′2+r2′2+r3′2=1+O(η).{r'_1}^{2}+{r'_2}^{2}+{r'_3}^{2}=1+O(\eta).

The radial observation preceding the definition of TT, based on Remark 2.3, gives a uniform positive lower bound on all three radii and places at least two in [1/48,1/2−1/48][1/48,1/2-1/48]. We split into the three possible choices of this pair.

For the selected pair, the thickened condition Q(q−pi−pj)=O(δ)Q(q-p_i-p_j)=O(\delta) becomes Q((0,0,1)−Aqpi−Aqpj)=O(η)Q((0,0,1)-A_qp_i-A_qp_j)=O(\eta). Equations (14)–(15) then turn it into an O(η)O(\eta)-thickened point–line incidence, between NτN_\tau points and NτN_\tau lines. Lemma 3.5 and (26) show that the number of admissible pairs is at most

η−c−εKσ2cNτ2−2c≲εδ−c−εσ−2cdNτ2−2c.\eta^{-c-\varepsilon}K_\sigma^{2c}N_\tau^{2-2c}\lesssim_\varepsilon\delta^{-c-\varepsilon}\sigma^{-2cd}N_\tau^{2-2c}.

Once the two original points are fixed, the third lies in an original O(δ)O(\delta)-ball, giving O(1)O(1) possibilities by separation. Notice that it is the original δ\delta, not the larger normalized thickness η\eta, that is used in this last step.

Summing over the scales

Recall the value of cc in (27).

Lemma 3.8 (One angular scale σ\sigma). For each dyadic δ≲σ≲1\sqrt{\delta}\lesssim\sigma\lesssim1,

Eσ:=∑qQ(q)≍σn(q)2≲εNδ−c−d(4−2c)−εσ(d−1)(4−2c)−2cd.(28)E_\sigma:=\sum_{\substack{q\\Q(q)\asymp\sigma}}n(q)^2 \lesssim_\varepsilon N\delta^{-c-d(4-2c)-\varepsilon}\sigma^{(d-1)(4-2c)-2cd}. \tag*{(28)}

Proof. Recall the assignment q↦τ(q)q\mapsto\tau(q) in Lemma 3.1, and let QτQ_\tau be those qq assigned to τ\tau. We start with the trivial first-moment estimate

∑q∈Qτn(q)≤Nτ3.\sum_{q\in Q_\tau}n(q)\le N_\tau^3.

Indeed, every triple in this sum belongs to (Pτcone)3(P_\tau^{\mathrm{cone}})^3, and its sum occupies exactly one output cube. We next apply Lemma 3.7

sup⁡q∈Qτn(q)≲εδ−c−εσ−2cdNτ2−2c.\sup_{q\in Q_\tau}n(q)\lesssim_\varepsilon\delta^{-c-\varepsilon}\sigma^{-2cd}N_\tau^{2-2c}.

Combining the two gives

∑q∈Qτn(q)2≲ε(sup⁡q∈Qτn(q))∑q∈Qτn(q)\sum_{q\in Q_\tau}n(q)^2\lesssim_\varepsilon\left(\sup_{q\in Q_\tau}n(q)\right)\sum_{q\in Q_\tau}n(q)
≲εδ−c−εσ−2cdNτ2−2cNτ3=δ−c−εσ−2cdNτ5−2c.\lesssim_\varepsilon\delta^{-c-\varepsilon}\sigma^{-2cd}N_\tau^{2-2c}N_\tau^3 =\delta^{-c-\varepsilon}\sigma^{-2cd}N_\tau^{5-2c}.

Summing over τ\tau and using the bounded overlap of τ\tau yields

Eσ≲δ−c−εσ−2cd∑τ∈TσNτ5−2c.E_\sigma\lesssim\delta^{-c-\varepsilon}\sigma^{-2cd}\sum_{\tau\in\mathcal{T}_\sigma}N_\tau^{5-2c}.

Using the two parts of (21) separately,

∑τNτ5−2c≤(max⁡τNτ)4−2c∑τNτ≲N(δ−dσd−1)4−2c.\sum_\tau N_\tau^{5-2c}\le\left(\max_\tau N_\tau\right)^{4-2c}\sum_\tau N_\tau\lesssim N\left(\delta^{-d}\sigma^{d-1}\right)^{4-2c}.

This proves (28).

Proof of Theorem 1.2. For 1<d≤3/21<d\le3/2, c=1/2c=1/2 and Lemma 3.8 reads

Eσ≲εNδ−3d−1/2−εσ2d−3.E_\sigma\lesssim_\varepsilon N^{\delta-3d-1/2-\varepsilon}\sigma^{2d-3}.

Since 2d−3≤02d-3\le0, the largest value is at σ≍δ\sigma\asymp\sqrt{\delta}, and equals Nδ−2d−2−εN^{\delta-2d-2-\varepsilon}.

For 3/2<d≤23/2<d\le2, c=(2d−1)−1c=(2d-1)^{-1} and the same lemma reads

Eσ≲εNδ−(4d−1)−εσ4d−6.E_\sigma\lesssim_\varepsilon N^{\delta-(4d-1)-\varepsilon}\sigma^{4d-6}.

Here 4d−6>04d-6>0, so the largest value is at σ≲1\sigma\lesssim1. The near-zero contribution in Lemma 3.3 is O(Nδ−2d−2)O(N^{\delta-2d-2}), which is bounded by Nδ−(4d−1)N^{\delta-(4d-1)} for d≥3/2d\ge3/2. There are O(log⁡(1/δ))O(\log(1/\delta)) dyadic scales σ\sigma, and we absorb this loss into δ−ε\delta^{-\varepsilon}.

L⁶ Fourier decay for fractal measures

We now prove Theorem 1.1, which we recall below. The derivation from the energy bound is rather standard. One decomposes μ\mu into 0,10,1-valued densities μn\mu_n that are locally constant at scale δ\delta. The different contributions are evaluated using the triangle inequality. The centers of the δ\delta-balls supporting μn\mu_n satisfy KT non-concentration (3), albeit with constants that are not O(1)O(1). Random sampling is typically used in the literature to go from large to O(1)O(1) constants. The even-ness of 6 allows for the sixth moment of μ^n\widehat{\mu}_n to be expressed in terms of the 3-energy of the centers. Rather than pursuing this somewhat technical approach, we follow a more elegant random sampling that could be of independent interest.

Theorem 4.1. Let 1<d≤21<d\le2, and let μ\mu be a probability measure supported on CC such that

μ(B(x,r))≤Crd,0<r≤2.(29)\mu(B(x,r))\le Cr^d,\qquad0<r\le2. \tag*{(29)}

For every ε>0\varepsilon>0 and R≥2R\ge2,

∫B(0,R)∣μ^(ξ)∣6 dξ≲ε,CRεmax⁡{R5−3d,R2−d}.(30)\int_{B(0,R)}|\widehat{\mu}(\xi)|^6\,d\xi\lesssim_{\varepsilon,C}R^\varepsilon\max\{R^{5-3d},R^{2-d}\}. \tag*{(30)}

We first pass from point sets to measures at scale δ\delta.

Lemma 4.2. Under the hypotheses of Theorem 4.1, for 0<δ≪10<\delta\ll1 and any fixed A>0A>0,

μ⊗6{(pi)i=16:∣p1+p2+p3−p4−p5−p6∣≤Aδ}≲ε,C,Aδ6d−εmax⁡{δ−(3d+2),δ−(5d−1)}.(31)\mu^{\otimes6}\left\{(p_i)_{i=1}^6: |p_1+p_2+p_3-p_4-p_5-p_6|\le A\delta\right\}\lesssim_{\varepsilon,C,A}\delta^{6d-\varepsilon}\max\{\delta^{-(3d+2)},\delta^{-(5d-1)}\}. \tag*{(31)}

Proof. Choose a maximal δ\delta-separated set {xj}\{x_j\} in supp⁡μ\operatorname{supp}\mu and partition supp⁡μ\operatorname{supp}\mu into measurable cells DjD_j associated with this net. Each DjD_j is contained in B(xj,δ)B(x_j,\delta). Write mj=μ(Dj)m_j=\mu(D_j). The points xjx_j lie on CC, are δ\delta-separated, and satisfy mj≤C1δdm_j\le C_1\delta^d by (29). Set aj=mj/(C1δd)≤1a_j=m_j/(C_1\delta^d)\le1.

We follow a probabilistic approach. Independently for each k=1,…,6k=1,\ldots,6 and each jj, select the center xjx_j with probability aja_j. Denote the resulting six random sets by SkS_k, and put S=⋃k=16SkS=\bigcup_{k=1}^6S_k.

We first prove the following: for every D>0D>0 there is a constant KDK_D such that, with probability at least 1−OD(δD)1-O_D(\delta^D),

#(S∩B(x,ρ))≤KDlog⁡(1/δ)(ρ/δ)d,δ≤ρ≤2.(32)\#(S\cap B(x,\rho))\le K_D\log(1/\delta)(\rho/\delta)^d,\qquad\delta\le\rho\le2. \tag*{(32)}

Here are the details. For each jj let Yj=1{xj∈S}Y_j=\mathbf{1}_{\{x_j\in S\}}. The variables YjY_j are independent as jj varies, and

P(Yj=1)=1−(1−aj)6≤6aj.\mathbb{P}(Y_j=1)=1-(1-a_j)^6\le6a_j.

Let rℓ=2ℓδr_\ell=2^\ell\delta, 0≤ℓ≤L0\le\ell\le L, where LL is the smallest integer such that rL≥2r_L\ge2. For each center xjx_j of the original δ\delta-separated net (whether or not xjx_j is selected into SS), put

Xj,ℓ:=#(S∩B(xj,2rℓ))=∑i:xi∈B(xj,2rℓ)Yi.X_{j,\ell}:=\#(S\cap B(x_j,2r_\ell))=\sum_{i:x_i\in B(x_j,2r_\ell)}Y_i.

Since the cells DiD_i are disjoint and Di⊂B(xi,δ)D_i\subset B(x_i,\delta), while δ≤rℓ\delta\le r_\ell, we have

λj,ℓ:=EXj,ℓ≤6C1δd∑xi∈B(xj,2rℓ)μ(Di)≤6μ(B(xj,3rℓ))C1δd≲C(rℓ/δ)d.\begin{aligned} \lambda_{j,\ell}:=\mathbb{E}X_{j,\ell}\le\frac{6}{C_1\delta^d}\sum_{x_i\in B(x_j,2r_\ell)}\mu(D_i) \\ &\le\frac{6\mu(B(x_j,3r_\ell))}{C_1\delta^d}\lesssim_C (r_\ell/\delta)^d. \end{aligned}

When 3rℓ>23r_\ell>2, the last inequality follows instead from μ(B)≤1\mu(B)\le1 and rℓ>2/3r_\ell>2/3. Put tℓ=Klog⁡(1/δ)(rℓ/δ)dt_\ell=K\log(1/\delta)(r_\ell/\delta)^d, with KK to be chosen. The Chernoff bound for sums of independent Bernoulli variables gives

P(Xj,ℓ≥tℓ)≤(eλj,ℓtℓ)tℓ≤2−tℓ≤δKlog⁡2,\mathbb{P}(X_{j,\ell}\ge t_\ell)\le\left(\frac{e\lambda_{j,\ell}}{t_\ell}\right)^{t_\ell}\le2^{-t_\ell}\le\delta^{K\log2},

provided KK is sufficiently large in terms of C,dC,d. There are O(log⁡(1/δ))O(\log(1/\delta)) choices of ℓ\ell and O(δ−3)O(\delta^{-3}) original centers xjx_j, by separation and the boundedness of supp⁡μ⊂C\operatorname{supp}\mu\subset C. Taking the union bound over these pairs (j,ℓ)(j,\ell), the probability that any estimate fails is at most

O(δ−3log⁡(1/δ) δKlog⁡2)=OD(δD),O(\delta^{-3}\log(1/\delta)\,\delta^{K\log2})=O_D(\delta^D),

if KK is large enough depending on DD.

Fix D>18D>18, and choose K=KDK=K_D sufficiently large above. Let GG denote the event that Xj,ℓ≤tℓX_{j,\ell}\le t_\ell simultaneously for every center xjx_j of the original net and every dyadic radius rℓr_\ell. The preceding union bound shows that

P(Gc)≲DδD.\mathbb{P}(G^c)\lesssim_D\delta^D.

For every outcome in GG, we first verify that the non-concentration estimate (32) holds for arbitrary balls. Take a ball B(x,ρ)B(x,\rho) with δ≤ρ≤2\delta\le\rho\le2. If S∩B(x,ρ)S\cap B(x,\rho) is empty, there is nothing to prove. Otherwise, choose one of its selected points xj∈S∩B(x,ρ)x_j\in S\cap B(x,\rho). For every z∈S∩B(x,ρ)z\in S\cap B(x,\rho), ∣z−xj∣≤∣z−x∣+∣x−xj∣<2ρ|z-x_j|\le|z-x|+|x-x_j|<2\rho, and hence

S∩B(x,ρ)⊂S∩B(xj,2ρ)⊂S∩B(xj,2rℓ),S\cap B(x,\rho)\subset S\cap B(x_j,2\rho)\subset S\cap B(x_j,2r_\ell),

where ℓ\ell is chosen so that ρ≤rℓ<2ρ\rho\le r_\ell<2\rho. The simultaneous estimates for the original centers therefore give

#(S∩B(x,ρ))≤Xj,ℓ≤tℓ≤2dKlog⁡(1/δ)(ρ/δ)d.\#(S\cap B(x,\rho))\le X_{j,\ell}\le t_\ell\le2^dK\log(1/\delta)(\rho/\delta)^d.

Absorbing 2d2^d into KDK_D proves (32) simultaneously for all centers and radii on GG.

Let Γ\Gamma consist of (deterministic) ordered six-tuples of cell indices whose representatives satisfy ∣xj1+xj2+xj3−xj4−xj5−xj6∣≤(A+6)δ|x_{j_1}+x_{j_2}+x_{j_3}-x_{j_4}-x_{j_5}-x_{j_6}|\le(A+6)\delta. We first relate this discrete sum to the measure on the left of (31). Write

FA,δ:={(p1,…,p6)∈C6:∣∑k=13pk−∑k=46pk∣≤Aδ}.F_{A,\delta}:=\left\{(p_1,\ldots,p_6)\in C^6:\left|\sum_{k=1}^{3}p_k-\sum_{k=4}^{6}p_k\right|\le A\delta\right\}.

If (p1,…,p6)∈FA,δ(p_1,\ldots,p_6)\in F_{A,\delta} and pk∈Djkp_k\in D_{j_k}, then ∣pk−xjk∣≤δ|p_k-x_{j_k}|\le\delta for each kk, and hence

∣∑k=13xjk−∑k=46xjk∣≤∣∑k=13pk−∑k=46pk∣+∑k=16∣pk−xjk∣≤(A+6)δ.\left|\sum_{k=1}^{3}x_{j_k}-\sum_{k=4}^{6}x_{j_k}\right| \le\left|\sum_{k=1}^{3}p_k-\sum_{k=4}^{6}p_k\right|+\sum_{k=1}^{6}|p_k-x_{j_k}| \le(A+6)\delta.

Thus (j1,…,j6)∈Γ(j_1,\ldots,j_6)\in\Gamma. Since the cells form a measurable partition and μ⊗6\mu^{\otimes6} is a product measure,

μ⊗6(FA,δ)=∑j1,…,j6μ⊗6(FA,δ∩(Dj1×⋯×Dj6))≤∑(j1,…,j6)∈Γμ(Dj1)⋯μ(Dj6)=∑Γmj1⋯mj6.(33)\mu^{\otimes6}(F_{A,\delta}) =\sum_{j_1,\ldots,j_6}\mu^{\otimes6}\left(F_{A,\delta}\cap(D_{j_1}\times\cdots\times D_{j_6})\right) \le\sum_{(j_1,\ldots,j_6)\in\Gamma}\mu(D_{j_1})\cdots\mu(D_{j_6}) =\sum_{\Gamma}m_{j_1}\cdots m_{j_6}. \tag*{(33)}

Define the random count

Z:=#{(j1,…,j6)∈Γ:xjk∈Sk (1≤k≤6)}.Z:=\#\{(j_1,\ldots,j_6)\in\Gamma:x_{j_k}\in S_k\ (1\le k\le6)\}.

Independence of the six selections, even when some indices coincide, gives

EZ=∑Γaj1⋯aj6.\mathbb{E}Z=\sum_{\Gamma}a_{j_1}\cdots a_{j_6}.

As mj=C1δdajm_j=C_1\delta^d a_j, (33) now yields

μ⊗6(FA,δ)≤∑Γmj1⋯mj6=(C1δd)6EZ.(34)\mu^{\otimes6}(F_{A,\delta})\le\sum_{\Gamma}m_{j_1}\cdots m_{j_6}=(C_1\delta^d)^6\mathbb{E}Z. \tag*{(34)}

For each outcome, ZZ is bounded by the energy E3,δ(S)E_{3,\delta}(S), where we take the fixed tolerance constant C∗=A+6C_* = A+6 in the definition of that energy; the underlying separation scale remains δ\delta. Split the expectation using the good event GG:

EZ=E(Z1G)+E(Z1Gc).\mathbb{E}Z=\mathbb{E}(Z1_G)+\mathbb{E}(Z1_{G^c}).

On GG, the Katz–Tao estimate (32) holds for SS. Consequently, for each individual outcome ω∈G\omega\in G, Theorem 1.2, applied to the δ\delta-separated set S(ω)⊂CS(\omega)\subset C with tolerance constant C∗=A+6C_*=A+6, gives the uniform pointwise upper bound

Z(ω)≤E3,δ(S(ω))≲ε,C,D,Aδ−εmax⁡{δ−(3d+2),δ−(5d−1)}.Z(\omega)\le E_{3,\delta}(S(\omega))\lesssim_{\varepsilon,C,D,A}\delta^{-\varepsilon}\max\{\delta^{-(3d+2)},\delta^{-(5d-1)}\}.

Thus

E(Z1G)≲ε,C,D,Aδ−εmax⁡{δ−(3d+2),δ−(5d−1)}P(G)≲ε,C,D,Aδ−εmax⁡{δ−(3d+2),δ−(5d−1)}.\mathbb{E}(Z1_G)\lesssim_{\varepsilon,C,D,A}\delta^{-\varepsilon}\max\{\delta^{-(3d+2)},\delta^{-(5d-1)}\}\mathbb{P}(G) \lesssim_{\varepsilon,C,D,A}\delta^{-\varepsilon}\max\{\delta^{-(3d+2)},\delta^{-(5d-1)}\}.

On GcG^c we use only the deterministic estimate Z≤#Γ≲δ−18Z\le\#\Gamma\lesssim\delta^{-18}, as there are O(δ−3)O(\delta^{-3}) original net points. Thus

E(Z1Gc)≲δ−18P(Gc)≲DδD−18.\mathbb{E}(Z1_{G^c})\lesssim\delta^{-18}\mathbb{P}(G^c)\lesssim_D\delta^{D-18}.

Our fixed choice D>18D>18 makes this term bounded (indeed vanishing) as δ→0\delta\to0. Together with the bound on GG, this yields

EZ≲ε,C,Aδ−εmax⁡{δ−(3d+2),δ−(5d−1)}.\mathbb{E}Z\lesssim_{\varepsilon,C,A}\delta^{-\varepsilon}\max\{\delta^{-(3d+2)},\delta^{-(5d-1)}\}.

Multiplying by (C1δd)6(C_1\delta^d)^6 in (34) proves (31).

Proof of Theorem 4.1. Choose a nonnegative, smooth, compactly supported function gg on R3\mathbb{R}^{3}, of sufficiently small support and with ∫g=1\int g=1. Put ϕ=g∗g~\phi=g*\widetilde{g}, where g~(x)=g(−x)\widetilde{g}(x)=g(-x). Then ϕ≥0\phi\ge0, ϕ^=∣g^∣2≥c>0\widehat{\phi}=|\widehat{g}|^{2}\ge c>0 on B(0,1)B(0,1), and ϕ\phi is supported in a fixed ball B(0,A)B(0,A). With ν=μ∗μ∗μ\nu=\mu*\mu*\mu and ν~(E)=ν(−E)\widetilde{\nu}(E)=\nu(-E), Fourier inversion gives

∫B(0,R)∣μ^(ξ)∣6 dξ≲∫R3∣μ^(ξ)∣6ϕ^(ξ/R) dξ=R3∫R3ϕ(Rx) d(ν∗ν~)(x)≲R3μ⊗6{(pi):∣∑i=13pi−∑i=46pi∣≤A/R}.\begin{aligned} \int_{B(0,R)}|\widehat{\mu}(\xi)|^{6}\,d\xi &\lesssim\int_{\mathbb{R}^{3}}|\widehat{\mu}(\xi)|^{6}\widehat{\phi}(\xi/R)\,d\xi\\ &=R^{3}\int_{\mathbb{R}^{3}}\phi(Rx)\,d(\nu*\widetilde{\nu})(x) \\ &\lesssim R^{3}\mu^{\otimes6}\left\{(p_i):\left|\sum_{i=1}^{3}p_i-\sum_{i=4}^{6}p_i\right|\le A/R\right\}. \end{aligned}

Apply Lemma 4.2 with δ=R−1\delta=R^{-1}. The last expression is

≲εR3−6d+εmax⁡{R3d+2,R5d−1}=Rεmax⁡{R5−3d,R2−d},\lesssim_{\varepsilon}R^{3-6d+\varepsilon}\max\{R^{3d+2},R^{5d-1}\}=R^{\varepsilon}\max\{R^{5-3d},R^{2-d}\},

as required.

Remark 4.3. We close by presenting the sharpness of the exponents in Theorem 4.1. For each R≥2R\ge2, set δ=R−1\delta=R^{-1} and choose a maximal-size (δ,d)(\delta,d)-KT set PP that maximizes the energy bound. Let μ\mu be normalized surface measure on the union of small δ\delta-neighborhoods of PP within the cone. The KT condition makes μ\mu a dd-Frostman measure with constant independent of RR. A standard positive-kernel argument then converts the discrete energy lower bound into the corresponding Fourier lower bound. Thus the exponents are sharp uniformly over dd-Frostman measures, allowing the measure to depend on RR.

Use of AI. The mathematical results in this paper are mostly due to the author. ChatGPT (GPT-6, OpenAI) was used to sharpen the exposition, to generate the map VV and the alternative proof to Theorem 4.1 presented here.

Department of Mathematics, Indiana University, 831 East 3rd St., Bloomington, IN 47405, USA.

Email address: [email protected]

References

  1. [1]A. Cushman, C. Demeter and S. Wu, Diameter-free reverse inequalities and superorthogonality, arXiv:2609.13105, 2026.arxiv.org/abs/2609.13105
  2. [2]C. Demeter and W. O’Regan, Incidence estimates for quasi-product sets and applications, arXiv:2511.15899, 2025.arxiv.org/abs/2511.15899
  3. [3]Y. Fu and K. Ren, Incidence estimates for α-dimensional tubes and β-dimensional balls in ℝ², Journal of Fractal Geometry, 11 (2024), no. 1-2, 1-30
  4. [4]L. Guth, H. Wang and R. Zhang, A sharp square function estimate for the cone in ℝ³, Annals of Mathematics 192 (2020), 551–581.arxiv.org/abs/1909.10693
  5. [5]Y. Jing and S. Wu, Near diagonal additive energy bound for points on algebraic surfaces, arXiv:2608.14467, 2026.arxiv.org/abs/2608.14467
  6. [6]T. Orponen, Additive properties of fractal sets on the parabola, arXiv:2205.02770.arxiv.org/abs/2205.02770
  7. [8]J. Pach and M. Sharir, On the number of incidences between points and curves, Combin. Probab. Comput. 7 (1998), 121–127DOI

Paper details

vlnc:W_qsrgtxfd
Version
v1 · latest
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