OFF-DIAGONAL ESTIMATES FOR THE HELICAL MAXIMAL

FUNCTION
arXiv:2305.17133v1 [math.CA] 26 May 2023

DAVID BELTRAN, JENNIFER DUNCAN, AND JONATHAN HICKMAN

Abstract. The optimal Lp Ñ Lq mapping properties for the (local) helical

maximal function are obtained, except for endpoints. The proof relies on tools
from multilinear harmonic analysis and, in particular, a localised version of
the Bennett–Carbery–Tao restriction theorem.

1. Introduction
1.1. Main results. Let γ : I Ñ R3 be a smooth curve, where I :“ r´1, 1s, which
is non-degenerate in the sense that there is a constant c0 ą 0 such that
| detpγ p1q psq, γ p2q psq, γ p3q psqq| ě c0 for all s P I. (1.1)
This is equivalent to saying that γ has non-vanishing curvature and torsion. Pro-
totypical examples are the helix γpsq “ pcosp2πsq, sinp2πsq, sq or the moment curve
γpsq “ ps, s2 {2, s3 {6q. Given t ą 0, consider the averaging operator
ż
At f pxq :“ f px ´ tγpsqq χpsq ds,
R
defined initially for Schwartz functions f P SpR3 q, where χ P C 8 pRq is a bump
function supported on the interior of I. Furthermore, define the associated local
maximal function
Mγ f pxq :“ sup |At f pxq|.
1ďtď2
Here we are interested in determining the sharp range of Lp Ñ Lq estimates for
Mγ . To describe the results, let
T :“ convtp0, 0q, p1{3, 1{3q, p1{4, 1{6qu z tp1{3, 1{3qu,
so that T is a closed triangle (formed by the closed convex hull of three points)
with one vertex removed. We let intpT q denote the interior of T and L denote the
intersection of T with the diagonal: see Figure 1. Standard examples show that
Mγ fails to be Lp Ñ Lq bounded whenever p1{p, 1{qq R T : see §9. The following
theorem therefore characterises the type set of Mγ , up to endpoints.
Theorem 1.1. For all p1{p, 1{qq P intpT q Y L, there exists a constant Cγ,p,q ě 1
such that the a priori estimate
}Mγ f }Lq pR3 q ď Cγ,p,q }f }Lp pR3 q
3
holds for all f P SpR q.

2020 Mathematics Subject Classification. 42B25, 42B20.

Key words and phrases. Helical maximal function, Lp ´ Lq local smoothing estimates.
D.B. supported by the AEI grant RYC2020-029151-I.
J.D. partially supported by ERC grant 834728 and Severo Ochoa grant CEX2019-000904-S.
1
2 D. BELTRAN, J. DUNCAN, AND J. HICKMAN

1/q

1 1



2, 3
1 1
3, 3

T
1 1


L 4, 6

1 1


12 , 12

(0, 0) 1 1/p
2, 0

Figure 1. The known range of Lp Ñ Lq boundedness for the

helical maximal function. In [2] and [18], boundedness was shown
on the half-open line segment L connecting p0, 0q and p1{3, 1{3q.
By Theorem 1.1, the operator is bounded whenever p1{p, 1{qq P
intpT q, the interior of the triangle with vertices p0, 0q, p1{3, 1{3q
and p1{4, 1{6q. Frequency localised estimates are obtained at the
critical vertex p1{4, 1{6q by interpolating multilinear inequalities
at p1{2, 1{3q, p1{12, 1{12q and p1{2, 0q; see §3.5 below.

For the diagonal case (that is, p1{p, 1{qq P L), the sharp range of estimates was
established in [2] and [18], building on earlier work of [21]. Hence, our main result
is to push the range of boundedness to the region intpT q. As a consequence of
Theorem 1.1 (or, more precisely, Theorem 3.1 below) and [4, Theorem 1.4], pp, q 1 q-
sparse bounds for the full maximal operator Mγfull f pxq :“ suptą0 |At f pxq| follow
for p1{p, 1{qq P intpT q Y L. We omit the details and refer to [4] for the precise
statements.

1.2. Methodology. Here we provide a brief overview of the ingredients of the

proof of Theorem 1.1 and the novel features of the argument. For fixed t, the
averaging operators At correspond to convolution with an appropriate measure µt
on the t-dilate of γ. It is therefore natural to study these objects via the Fourier
transform, which leads us to consider the multiplier
ż
µ̂t pξq “ e´itxγpsq,ξy χpsq ds.
R
Stationary phase can be used to compute the decay rate of this function in dif-
ferent directions in the frequency space. This involves analysing the vanishing of
 ESTIMATES FOR THE HELICAL MAXIMAL FUNCTION 3

s-derivatives of the phase function ϕpξ, sq :“ xγpsq, ξy. Following earlier work on
the circular maximal function [20], it is also useful to study the Fourier transform of
At f pxq in both the x and the t variables. This leads us to consider the 4-dimensional
pξ, τ q frequency space.
Broadly speaking, this approach was taken in both works [18] and [2] to study
the Lp Ñ Lp mapping properties of Mγ . However, these papers focused on differ-
ent geometrical aspects of the problem. In very rough terms, the analysis of [18]
centres around a 3-dimensional cone Γ3 in pξ, τ q-space arising from the equations
Bs ϕpξ, sq “ 0, τ “ ϕpξ, sq. On the other hand, the analysis of [2] centres around
a 2-dimensional cone Γ2 in the pξ, τ q-space arising from the system of equations
Bs ϕpξ, sq “ Bs2 ϕpξ, sq “ 0, τ “ ϕpξ, sq.
It seems difficult to obtain almost optimal Lp Ñ Lq estimates using either the
approach of [18] or of [2] in isolation; rather, it appears necessary to incorporate
both geometries into the analysis. In order to do this, we apply a recent observation
of Bejenaru [1], which provides a localised variant of the Bennett–Carbery–Tao mul-
tilinear restriction theorem [7]. We describe the relevant setup in detail in §2 below;
moreover, in the appendix we relate the required localised estimates to the theory
of Kakeya–Brascamp–Lieb inequalities from [6]. The local multilinear restriction
estimate allows us to work simultaneously with the Γ2 and Γ3 geometries, by con-
sidering the embedded cone Γ2 as a localised portion of Γ3 . See Proposition 3.5
below.
On the other hand, the geometries of both Γ2 and Γ3 were previously exploited
in a non-trivial manner in [19] and [3] using the decoupling inequalities from [10].
This approach is inspired by earlier work of Pramanik–Seeger [21]. Decoupling
is effective for proving Lp Ñ Lp bounds for large p; here it is used to provide
a counterpoint for interpolation with the estimates obtained via local multilinear
restriction. See Proposition 3.6 below.

1.3. Notational conventions. Throughout the paper, I denotes the interval r´1, 1s.
We let Rp denote the frequency domain, which is the Pontryagin dual group of R
understood here as simply a copy of R. Given f P L1 pRd q and g P L1 pR p d q we define
the Fourier transform and inverse Fourier transform by
ż ż
1
fˆpξq :“ e´ixx,ξy f pxq dx and ǧpxq :“ eixx,ξy gpξq dξ,
Rd p2πqd Rp d
respectively. For m P L8 pRp d q we define the multiplier operator mpDq which acts
initially on Schwartz functions by
ż
1
mpDqf pxq :“ eixx,ξy mpξqfˆpξq dξ.
p2πqd Rp d
Given a list of objects L and real numbers A, B ě 0, we write A ÀL B or
B ÁL A to indicate A ď CL B for some constant CL which depends only items in
the list L and our choice of underlying non-degenerate curve γ. We write A „L B
to indicate A ÀL B and B ÀL A.

1.4. Organisation of the paper.

‚ In §2 we present the key localised trilinear restriction estimate.
‚ In §3 we describe a reduction of Theorem 1.1 to three local-smoothing-type
estimates: trilinear L2 Ñ L3 , linear L12 Ñ L12 and trivial L2 Ñ L8 .
4 D. BELTRAN, J. DUNCAN, AND J. HICKMAN

‚ In §4 we describe the basic properties of our operators and prove the trivial
L2 Ñ L8 estimate.
‚ In §5 we prove the trilinear L2 Ñ L3 estimate using the trilinear restriction
theorem from §2.
‚ In §6 we prove the linear L12 Ñ L12 estimate using decoupling.
‚ In §7 we bound a non-degenerate portion of the operator.
‚ In §8 we carry out the reduction described in §3 and thereby bound the
remaining degenerate portion of the operator.
‚ In §9 we demonstrate the sharpness of the range T .
‚ Finally, in Appendix A we present a proof of the localised trilinear restric-
tion theorem from §2.

Acknowledgements. The first and third authors would like to thank Shaoming
Guo and Andreas Seeger for discussions related to the topic of this paper over the
years.

2. Localised trilinear restriction

The key ingredient in the proof of Theorem 1.1 is a localised trilinear Fourier
restriction estimate. Here we describe the particular setup for our problem. As
in [18], it is necessary to work with functions with a limited degree of regularity.
Definition 2.1. Let 0 ă α ď 1 and U Ď Rd be an open set. We say a function
Q : U Ñ R is of class C 1,α pU q if it is continuously differentiable on U and, moreover,
the partial derivatives satisfy the α-Hölder condition
|∇Qpξ1 q ´ ∇Qpξ2 q|
sup ă 8.
ξ1 ,ξ2 PU |ξ1 ´ ξ2 |α
ξ1 ‰ξ2

Consider an ensemble Q “ pQ1 , Q2 , Q3 q of maps Qj : Uj Ñ R of class C 1,1{2 pUj q

p 3 is an open domain1 for 1 ď j ď 3. The graphs
where Uj Ď R
Σj :“ tpξ, Qj pξqq : ξ P Uj u
p4
are hypersurfaces in R , with some limited regularity. Each Σj has a Gauss map
given by
ˆ ˙
3 1 ´∇Qj pξq
νj : Uj Ñ S , νj pξq :“ for all ξ P Uj .
p1 ` |∇Qj pξq|2 q1{2 1
We further fix a smooth function u : U3 Ñ R satisfying |∇upξq| ą c0 ą 0 for all
ξ P U3 . This implicitly defines a smooth surface Z3 :“ tξ P R p 3 : upξq “ 0u, which
we lift to
Σ13 :“ tpξ, Q3 pξqq : ξ P Z3 u.
Thus, Σ13 is a codimension 1 submanifold of Σ3 , which is embedded in R p 4 . Defining
ˆ ˙
1 ∇upξq
ν31 : Z3 Ñ S 3 , ν31 pξq :“ for all ξ P Z3 ,
|∇upξq| 0
it follows that tν3 pξq, ν31 pξqu forms a basis of the normal space to Σ13 at pξ, Q3 pξqq
for all ξ P Z3 .

1Here an open domain in R

p d is an open, bounded, connected subset of R
p d.
 ESTIMATES FOR THE HELICAL MAXIMAL FUNCTION 5

We now fix aj P Cc pUj q with }aj }L8 pUj q ď 1 for 1 ď j ď 3 and we assume the
transversality hypothesis
ˇ ˆ ˙ˇ
ˇ det ∇Q1 pξ1 q ∇Q2 pξ2 q ∇Q3 pξ3 q ∇upξ3 q ˇ ą ctrans ą 0
ˇ ˇ
(2.1)
ˇ 1 1 1 0 ˇ
for all ξj P supp aj , 1 ď j ď 3. Furthermore, given 0 ă µ ă 1, we assume the
additional localisation hypothesis
|upξq| ă µ for all ξ P supp a3 . (2.2)
Finally, we define the extension operators
ż
Ej f px, tq :“ eipxx,ξy`tQj pξqq aj pξqf pξq dξ for f P L1 pUj q, 1 ď j ď 3.
p3
R
The key localised trilinear estimate is as follows.
Theorem 2.2 (Localised trilinear restriction). With the above setup, for all ε ą 0
and all R ě 1 we have
›ź3 › 3
1{3
ź
|Ej fj |1{3 › ÀQ,ε Rε µ1{6
› ›
› }fj }L2 pUj q
L3 pBp0,Rqq
j“1 j“1
1
for all fj P L pUj q, 1 ď j ď 3.
Here the implied constant depends on the choice of maps Qj and, in particular,
the lower bound in (2.1), but is (crucially) independent of the choice of parameter
µ in (2.2) and the choice of scale R.
If we consider smooth hypersurfaces rather than the C 1,1{2 class, then Theo-
rem 2.2 is a special case of [1, Theorem 1.3]. We expect that the arguments of [1]
can be generalised to treat C 1,α regularity for all α ą 0. However, in Appendix A we
observe that Theorem 2.2 is a rather direct consequence of the Kakeya–Brascamp–
Lieb inequalities from [6] (see also [24, 25]).

3. Initial reductions
3.1. Local smoothing estimates. The multipliers of interest are of the following
form. For I :“ r´1, 1s, let γ : I Ñ R3 be a smooth curve and fix ρ P Cc8 pRq
supported in the interior r1{2, 4s. Given a symbol a P C 8 pR p 3 zt0u ˆ R ˆ Rq, we
define ż
mγ raspξ; tq ” mraspξ; tq :“ e´itxγpsq,ξy apξ; t; sqψI psqρptq ds, (3.1)
R
for some ψI P C 8 pRq with support lying in I. Fix η P Cc8 pRq non-negative, even
and such that
ηprq “ 1 if r P I and supp η Ď r´2, 2s
k
and define β, β P Cc8 pRq by
βprq :“ ηprq ´ ηp2rq and β k prq :“ βp2´k rq for all k P Z. (3.2)
k k p 3.
By an abuse of notation, we also write ηpξq :“ ηp|ξ|q and β pξq :“ β p|ξ|q for ξ P R
p 3 zt0uˆRˆRq as above, we form a dyadic decomposition by writing
For a P C 8 pR
8 "
ÿ apξ; t; sq β k pξq for k ě 1
a“ ak where ak pξ; t; sq :“ . (3.3)
apξ; t; sq ηpξq for k “ 0
k“0
With the above definitions, our main result is as follows.
6 D. BELTRAN, J. DUNCAN, AND J. HICKMAN

Theorem 3.1 (Lp Ñ Lq local smoothing). Let γ : I Ñ R3 be a smooth curve and

p 3 zt0u ˆ R ˆ Rq satisfies the symbol condition
suppose a P C 8 pR
|Bξα Bti Bsj apξ; t; sq| Àα,i,j |ξ|´|α| for all α P N30 and i, j P N0 (3.4)
and that
3
ÿ
|xγ pjq psq, ξy| Á |ξ| for all pξ; sq P suppξ a ˆ I. (3.5)
j“1
Then for all p1{p, 1{qq P intpT q there exists some εpp, qq ą 0 such that
´ż 2 ¯1{q
}mrak spD; tqf }qLq pR3 q dt Àp,q 2´k{q´εpp,qqk }f }Lp pR3 q
1
holds for all k P N0 , where ak is defined as in (3.3).
The desired maximal bound follows from Theorem 3.1 using a standard Sobolev
embedding argument; we omit the details but refer the reader to [23, Chapter XI,
§3], [21, §6] or [2, §2] for similar arguments.
By results of [2], Theorem 3.1 is known to hold along the diagonal line L. By in-
terpolation, it therefore suffices to prove an estimate at the critical vertex p1{4, 1{6q
in the Riesz diagram (see Figure 1).
Proposition 3.2. Under the hypotheses of Theorem 3.1, for k P N0 and all ε ą 0,
we have
´ż 2 ¯1{6
}mrak spD; tqf }6L6 pR3 q dt Àε 2´k{6`εk }f }L4 pR3 q . (3.6)
1

By the preceding discussion, our main theorem follows from Proposition 3.2.
Henceforth, we focus on the proof of this critical estimate.
3.2. Trilinear reduction. If the hypothesis (3.5) is strengthened to
|xγ 1 psq, ξy| ` |xγ 2 psq, ξy| Á |ξ| for all pξ; sq P suppξ a ˆ I, (3.7)
then one can deduce the critical estimate (3.6) as a consequence of known local
smoothing inequalities from [21] (see Theorem 6.8 below) and the Stein–Tomas
Fourier restriction inequality. Given a small number 0 ă δ ă 1 and k P N, we
perform this analysis on the symbols
` ˘
ak,0 pξ; t; sq :“ ak pξ; t; sq 1 ´ ηp2´k δ ´20 G2 ps; ξqq (3.8)
ř2
where G2 ps; ξq :“ j“1 |xγ pjq psq, ξy|; note that ak,0 satisfies (3.7) with an implicit
constant depending on δ. We discuss this case in detail in §7.
The main difficulty is then to get to grips with the degenerate portion of the
multiplier. For the above choice of 0 ă δ ă 1, this corresponds to the condition
|xγ 1 psq, ξy| ` |xγ 2 psq, ξy| À δ 20 |ξ| for all pξ; sq P suppξ a ˆ I; (3.9)
note that this is satisfied on the support of ak :“ ak ´ak,0 . To control the degenerate
part, we work with a trilinear variant of Proposition 3.2, from which we deduce the
corresponding linear estimate (3.6) via a standard application of the broad-narrow
method from [11] (see also [16]).
To describe the trilinear setup, we introduce some notation. For 0 ă δ ă 1 as
above, let Jpδq denote a covering of I by essentially disjoint intervals of length δ.
Let J3,sep pδq denote the collection of all triples J “ tJ1 , J2 , J3 u Ă Jpδq which satisfy
the separation condition distpJi , Jj q ě 10δ for 1 ď i ă j ď 3. Given a bounded
 ESTIMATES FOR THE HELICAL MAXIMAL FUNCTION 7

interval J Ď R, we let ψJ P Cc8 pRq satisfy supp ψJ Ď J and |BsN ψJ psq| ÀN |J|´N
for all N P N. Similarly to (3.1), given a symbol a P C 8 pRp 3 zt0u ˆ R ˆ Rq, we define
the multipliers adapted to an interval J P Jpδq by
ż
J J
mγ raspξ; tq ” m raspξ; tq :“ e´itxγpsq,ξy apξ; t; sqψJ psqρptq ds.
R
With this setup, we prove the following estimate.
Proposition 3.3 (L4 Ñ L6 trilinear local smoothing). Let k P N0 , ε ą 0 and
δ ą 0. Under the hypotheses of Theorem 3.1 and further assuming (3.9), we have
´ż 2› ź ›6 ¯1{6
1{3
ź
|mJ rak spD; tqfJ |1{3 › dt Àε δ ´Op1q 2´k{6`εk
› ›
› }fJ }L4 pR3 q
6 3L pR q
1 JPJ JPJ
3,sep 3
whenever J P J pδq and fJ P SpR q for J P J.
Here we use the notation Op1q to denote an unspecified absolute constant. In
applications, we work with relatively large values of δ (namely, δ „ε 1) and accord-
ingly there is no need to precisely track the δ dependence. We will also assume
without loss of generality that 0 ă δ ă c where c ą 0 is a small absolute constant,
chosen to satisfy the forthcoming requirements of the argument, and k is sufficiently
large depending on δ ´1 .
As mentioned above, the (ostensibly weaker) trilinear estimate in Proposition 3.3
implies the linear estimate in Proposition 3.2 (under the additional assumption
(3.9)) using a variant of the procedure introduced in [11]. We postpone the details
of this reduction to §8 below.
3.3. Reduction to perturbations of the moment curve. At small scales, any
non-degenerate curve can be thought of as a perturbation of an affine image of the
moment curve γ˝ psq :“ ps, s2 {2, s3 {6q. We refer to [2, §4] for details (which involve
the affine rescalings described in §4.2 below), and just record here that it suffices
to consider curves in the class Gpδ0 q defined below for 0 ă δ0 ă 1 sufficiently small.
Definition 3.4. Given 0 ă δ0 ă 1 and M P N, let Gpδ0 , M q denote the class of all
smooth curves γ : I Ñ R3 that satisfy the following conditions:
i) γp0q “ 0 and γ pjq p0q “ ⃗ej for 1 ď j ď 3;
ii) }γ ´ γ˝ }C M pIq ď δ0 for all 0 ď j ď M .
Here ⃗ej denotes the jth standard Euclidean basis vector and
}γ}C M pIq :“ max sup |γ pjq psq| for all γ P C M pI; R3 q.
1ďjďM sPI

If M “ 4, then we will simply write Gpδ0 q for Gpδ0 , 4q

Henceforth we will always assume that γ P Gpδ0 q for δ0 :“ 10´10 .
3.4. Microlocal decomposition. Under the assumption (3.9), the non-degeneracy
condition (1.1) ensures that
|xγ 3 psq, ξy| Á |ξ| for all pξ; t; sq P supp a; (3.10)
indeed, for γ P Gpδ0 q this condition holds for all pξ; t; sq P suppξ a ˆ R ˆ I. We can
then assume that s ÞÑ xγ 3 psq, ξy has constant sign and henceforth we assume that
ξ3 ą 0. Following [2, §6], let θ2 : suppξ a Ñ I be the smooth mapping such that
xγ 2 ˝ θ2 pξq, ξy “ 0 for all ξ P suppξ a.
8 D. BELTRAN, J. DUNCAN, AND J. HICKMAN

It is clear that θ2 is homogeneous of degree 0. Let

upξq :“ xγ 1 ˝ θ2 pξq, ξy for all ξ P suppξ a. (3.11)
Since (3.10) is satisfied on the support of each
ak :“ ak ´ ak,0 , (3.12)
we decompose each of these pieces with respect to the size of |upξq|. Given ε ą 0
and 0 ă δ ă 1, we write
ÿ
ak “ ak,0 ` ak,ℓ ` ak,k{3 (3.13)
ℓPΛpkq

where ak,0 is as in (3.8) and

$ ` ˘
& ak pξ; t; sq β 2´k`2ℓ upξq if ℓ P Λpkq,
ak,ℓ pξ; t; sq :“ (3.14)
% a pξ; t; sq η `2´k`2tp1´2εqk{3u upξq˘ if ℓ “ k{3
k

for
Λpkq :“ t ℓ P N : rlog2 pδ ´8 qs ă ℓ ă tp1 ´ 2εqk{3u u. (3.15)
Here we assume that k is large enough so that the decomposition (3.13) makes
sense; note that Theorem 3.1 trivially holds for small values of k. In particular, we
concern ourselves with k P N satisfying k ě 4 log2 pδ ´8 q. In the definition (3.15), for
any x P R, we let txu denote the largest integer less or equal than x and rxs denote
the smallest integer greater or equal than x. It will also be useful to introduce the
notation
Λpkq :“ Λpkq Y tk{3u.
Note that the indexing set Λpkq depends on the chosen δ and ε, but we do not
record this dependence for notational convenience. We also note that here the
function β should be defined slightly differently compared with (3.2); in particular,
here βprq :“ ηp2´2 rq ´ ηprq (we ignore this minor change in the notation).
As mentioned in §3.2, for the extreme case ℓ “ 0 we have the non-degeneracy
condition (3.7) (with an implied constant depending on δ). This situation is easy to
handle using known estimates: see §7 below. On the other hand, for ℓ P Λpkq, the
multipliers mrak,ℓ s are degenerate in the sense that (3.9) now holds. A key aspect
of this decomposition is that for ℓ P Λpkq, Taylor series expansion shows that
|xγ 1 psq, ξy| ` 2´ℓ |xγ 2 psq, ξy| Á 2k´2ℓ for all pξ, sq P supp ak,ℓ p ¨ ; t; ¨ q; (3.16)
see, for example [3, (5.15)] for a detailed derivation. The weak non-degeneracy
condition (3.16) will allow for improved estimates depending on the value of ℓ.
Bounding these pieces, and the piece for ℓ “ k{3, is the difficult part of the argument
and is the focus of §§5–6 below.
3.5. Microlocalised estimates. Throughout this section we work under the hy-
potheses of Theorem 3.1 and, in addition, assume (3.9) holds for a specified value of
δ. That is, we let γ : I Ñ R3 be a smooth curve and suppose a P C 8 pR p 3 zt0uˆRˆRq
satisfies (3.4), (3.5) and (3.9). Furthermore, we define the symbols ak,ℓ as in (3.14).
The key ingredient in the proof of Proposition 3.3 is a trilinear estimate for
the multipliers associated to the localised symbols ak,ℓ . To describe this result,
3,sep
we work řwith a triple of integers ℓJ “ pℓJ qJPJ indexed by J P J pδq and write
|ℓJ | :“ JPJ ℓJ .
 ESTIMATES FOR THE HELICAL MAXIMAL FUNCTION 9

Proposition 3.5 (L2 Ñ L3 trilinear local smoothing). For k P N, ε ą 0 and

0 ă δ ă 1, we have
´ż 2 › ź ›3 ¯1{3 3
1{3
ź
|mJ rak,ℓJ spD; tqfJ |1{3 › 3 3 dt Àε δ ´Op1q 2´k{3`|ℓJ |{18`εk
› ›
› }fJ }L2 pR3 q
1 L pR q
JPJ j“1
3,sep 3
whenever J P J pδq, ℓJ “ pℓJ qJPJ with ℓJ P Λpkq and fJ P SpR q for J P J.
Proposition 3.5 is a fairly direct consequence of Theorem 2.2; we describe the
proof in §5 below. To deduce the critical L4 Ñ L6 estimate stated in Proposi-
tion 3.3, we interpolate Proposition 3.5 with the following linear inequalities.
Proposition 3.6 (L12 Ñ L12 local smoothing). For k P N, ε ą 0, 0 ă δ ă 1 and
ℓ P Λpkq, we have
´ż 2 ¯1{12
}mrak,ℓ spD; tqf }12
L12 pR3 q dt Àε 2´k{6´ℓ{12`εk }f }L12 pR3 q .
1

Lemma 3.7 (L2 Ñ L8 estimate). For k P N, ε ą 0, 0 ă δ ă 1 and ℓ P Λpkq, we

have
sup }mrak,ℓ spD; tqf }L8 pR3 q À 2k´ℓ{2 }f }L2 pR3 q .
1ďtď2

We remark that 0 ă δ ă 1 plays no significant rôle in the proofs of Proposition

3.6 and Lemma 3.7 and it is used only to set up the underlying decomposition in
the ak,ℓ . Similarly, ε ą 0 plays no significant rôle in Lemma 3.7.
Proposition 3.6 is a minor variant of estimates which have appeared in, for
instance, [19] and [3]. The result is highly non-trivial, and relies on the ℓp decoupling
inequality for the moment curve from [10]. We discuss the details in §6.
Lemma 3.7, on the other hand, is elementary. It follows from basic pointwise
estimates for the multipliers ak,ℓ , obtained via stationary phase. We discuss the
details in §4.1.
Given the preceding results, the key trilinear L4 Ñ L6 local smoothing estimate
is immediate.
Proof (of Proposition 3.3). By multilinear Hölder’s inequality, Propositions 3.6 and
3.7 imply their trilinear counterparts. Since
˜ ¸ ˜ ¸ ˜ ¸ ˜ ¸
1 1 1 1
4 3 12 7 2 1 2
1
“ ¨ 1
` ¨ 1
` ¨ ,
6
5 12
20 3
20 0
interpolation of the three estimates immediately gives
´ż 2› ź ›6 ¯1{6
|mJ rak,ℓJ spD; tqf |1{3 › 6 3 dt Àε δ ´Op1q 2´k{6´|ℓJ |{180`εk }f }L4 pR3 q
› ›
›
1 L pR q
JPJ
for ℓJ “ pℓJ qJPJ with 0 ď ℓJ ď k{3; see Figure 1. Here we carry out the interpola-
tion using a multilinear variant of the Riesz–Thorin theorem: see, for instance, [8,
§4.4].2 Using the geometric decay in 2´ℓJ for each J P J, we sum these bounds to
deduce the desired result. □
To prove Proposition 3.3, it therefore remains to establish Proposition 3.5,
Proposition 3.6 and Lemma 3.7. We carry this out in §§4–6 below.
2Alternatively, a suitable multilinear interpolation theorem can be proved by directly adapting
the argument used to prove the classical Riesz–Thorin theorem.
10 D. BELTRAN, J. DUNCAN, AND J. HICKMAN

4. Basic properties of the multipliers

4.1. Elementary estimates for the multiplier. Using stationary phase argu-
ments, we can immediately deduce Lemma 3.7.
Proof (of Lemma 3.7). By the Cauchy–Schwarz inequality, we have the elementary
inequality
ˇ ˇ1{2
}mpDqf }L8 pR3 q ď }m} 8 p 3 ˇsupp mˇ }f }L2 pR3 q .
L pR q
Fixing t P R, in view of the above it suffices to show
}mrak,ℓ sp ¨ , tq}8 À 2´pk´ℓq{2 , |supp mrak,ℓ sp ¨ ; tq| À 23k´2ℓ .
Since ∇upξq “ γ 1 ˝ θ2 pξq is bounded away from zero, the latter estimate is clear. On
the other hand, the former estimate is a consequence of a simple stationary phase
analysis. Indeed, for ℓ “ k{3 we apply van der Corput’s inequality with third order
derivatives. For ℓ P Λpkq, we apply van der Corput’s inequality with either first or
second order derivatives, using the lower bound (3.16). For further details see, for
example, [21, Lemma 3.3], [3, (5.15)] or Lemma 4.3 below. □
4.2. Scaling of the multiplier. Let σ P I, 0 ă λ ă 1 be such that rσ ´ λ, σ ` λs Ď
I. Denote by rγsσ the 3 ˆ 3 matrix
“ ‰
rγsσ :“ γ p1q pσq γ p2q pσq γ p3q pσq ,
where the vectors γ pjq pσq are understood to be column vectors. Note that this
matrix is invertible due to the non-degeneracy hypothesis (1.1). It is also convenient
to let rγsσ,λ denote the 3 ˆ 3 matrix
rγsσ,λ :“ rγsσ ¨ Dλ (4.1)
2 3 2 3
where Dλ :“ diagpλ, λ , λ q is the diagonal matrix with eigenvalues λ, λ , λ . With
this data, define the pσ, λq-rescaling of γ as the curve γσ,λ P C 8 pI; R3 q given by
` ˘
γσ,λ psq :“ rγs´1
σ,λ γpσ ` λsq ´ γpσq . (4.2)
A simple computation shows
` ˘
det rγsσ`λs
detrγσ,λ ss “ ` ˘ ,
det rγsσ
and therefore γσ,λ is also a non-degenerate curve. Furthermore, the rescaled curve
satisfies the relations
pjq
xγσ,λ psq, ξy “ λj xγ pjq pσ ` λsq, prγsσ,λ q´J ξy for all j ě 1. (4.3)
Combining this with the fact that γσ,λ is non-degenerate, we have
3
ÿ pjq
|ξ| À |xγσ,λ psq, ξy| À λ3 |prγsσ,λ q´J ξ| À |ξ|; (4.4)
j“1

here the last inequality is a simple consequence of the definition (4.1).

Defining the rescaled symbol
aσ,λ pξ; t; sq :“ apprγsσ,λ q´J ξ; t; σ ` λsq, (4.5)
a change of variables immediately yields
mγ raspD; tqf pxq “ λ mγσ,λ raσ,λ spD; tqfσ,λ pprγsσ,λ q´1 px ´ tγpσqqq (4.6)
 ESTIMATES FOR THE HELICAL MAXIMAL FUNCTION 11

where fσ,λ :“ f ˝ rγsσ,λ . In particular, by scaling,

1 1
}mγ raspD; ¨ qf }Lp pR3 qÑLq pR3`1 q À δ 1`6p q ´ p q }mγσ,λ raσ,λ spD; ¨q}Lp pR3 qÑLq pR3`1 q .
(4.7)
Finally, we observe that if the original symbol satisfies the condition
|Bξα apξ; t; sq| Àα |ξ|´|α| for all α P N30 , (4.8)
then so too does the rescaled symbol aσ,λ . Indeed, by the definitions (4.5) and
(4.1), we have
|pBξα aσ,λ qpξ; t; sq| Àα λ´3|α| |pBξα aqpprγsσ,λ q´J ξ; t; σ ` λsq|.
In view of the hypothesis (4.8) and (4.4), it follows that
ˇ ˇ´|α|
|pBξα aσ,λ qpξ; t; sq| Àα λ´3|α| ˇprγsσ,λ q´J ξ ˇ Àα |ξ|´|α| ,
as required.
4.3. Critical points. We next describe the critical points of the phase function
s ÞÑ xγpsq, ξy which, under the setup of §3.4, depend on the sign of the quantity
upξq introduced in (3.11).
Lemma 4.1 ([2, Lemma 6.2]). Let ξ P suppξ a and consider the equation
xγ 1 psq, ξy “ 0. (4.9)
i) If upξq ą 0, then the equation (4.9) has no solution on I.
ii) If upξq “ 0, then the equation (4.9) has only the solution s “ θ2 pξq on I.
iii) If upξq ă 0, then the equation (4.9) has precisely two solutions on I.
Following [2, §6], we can use Lemma 4.1 to construct a (unique) pair of smooth
mappings
θ1˘ : tξ P suppξ a : upξq ă 0u Ñ I
with θ1´ pξq ď θ1` pξq which satisfies
xγ 1 ˝ θ1˘ pξq, ξy “ 0 for all ξ P suppξ a with upξq ă 0.
Define the functions
v ˘ pξq :“ xγ 2 ˝ θ1˘ pξq, ξy for all ξ P suppξ a with upξq ă 0.
Taylor expansion yields the following.
Lemma 4.2 ([2, Lemma 6.3]). Let ξ P suppξ a with upξq ă 0. Then
ˇ „ |θ˘ pξq ´ θ2 pξq| „ |θ` pξq ´ θ´ pξq| „ ˇu ξ ˇ1{2 .
ˇ ˘ ` ξ ˘ˇ ˇ ` ˘ˇ
ˇv
|ξ| 1 1 1 |ξ|

4.4. Stationary phase. We next use the approach in [18] and apply stationary
phase to express the multipliers mJ rak,ℓ s as a product of a symbol and an oscillatory
term. In what follows, we let
q2 pξq :“ xγ ˝ θ2 pξq, ξy and q1˘ pξq :“ xγ ˝ θ1˘ pξq, ξy
for any value of ξ such that the expression is well-defined. Our analysis leads to
various rapidly decreasing error terms. Given R ě 1, we let RapDecpRq denote the
p 3 ˆ Rq which satisfy |epξ; tq| ÀN R´N for all N P N0 .
class of functions e P C 8 pR
Lemma 4.3. Let k P N and J P Jpδq.
12 D. BELTRAN, J. DUNCAN, AND J. HICKMAN

i) For some ek,k{3 P RapDecp2k q, we may write

mJ rak,k{3 spξ; tq “ 2´k{3 e´itq2 pξq bJk,k{3 p2´k ξ; tq ` ek,k{3 pξ; tq (4.10)
where bJk,k{3 P C 8 pR3`1 q is supported in Bp0, 10q, satisfies |upξq| À 2´2k{3`4εk{3
and distpθ2 pξq, Jq ă δ for ξ P suppξ bJk,k{3 and
|BtN bJk,k{3 p2´k ξ; tq| ÀN 23kεN `εk for all pξ; tq P R3`1 and all N P N0 . (4.11)
ii) For ℓ P Λpkq and some ek,ℓ P RapDecp2k q, we may write
ÿ ˘
mJ rak,ℓ spξ; tq “ 2´pk´ℓq{2 e´itq1 pξq bJ,˘
k,ℓ p2
´k
ξ; tq ` ek,ℓ pξ; tq (4.12)
˘

where the bJ,˘ 8

k,ℓ P C pR
3`1
q are supported in Bp0, 10q, satisfy |upξq| „ 2´2ℓ and
distpθ2 pξq, Jq ă δ for ξ P suppξ bJ,˘
k,ℓ and

|BtN bJ,˘
k,ℓ p2
´k
ξ; tq| ÀN 1 for all pξ; tq P R3`1 and all N P N0 . (4.13)
Remark 4.4. In part ii) of the lemma, upξq ă 0 always holds on the support of
the bJ,˘ ˘
k,ℓ and therefore, by Lemma 4.1, the functions θ1 pξq are well-defined. The
portion of the multiplier supported on the set where upξq ą 0 is incorporated into
the error term ek,ℓ .
Proof. i) By a change of variables,
ż
J
m rak,k{3 spξ; tq “ 2 ´k{3 ´itq2 pξq
e e´itΦk pξ;sq αkJ pξ; t; 2´k{3 sq ds
R
where:
p 3 zt0u ˆ R ˆ Rq satisfies
‚ The symbol αkJ P C 8 pR
|upξq| À 2k{3`4εk{3 , |ξ| ď 2k`1 , |t| ď 4 and θ2 pξq ` 2´k{3 s P J (4.14)
for pξ; t; 2´k{3 sq P supp αkJ and |BtN BsM αkJ pξ; t; sq| ÀM,N 1 for all M , N P N0 ;
‚ The phase Φk is given by
Φk pξ; sq :“ xγpθ2 pξq ` 2´k{3 sq, ξy ´ xγ ˝ θ2 pξq, ξy.
If |s| ě 2εk and k is sufficiently large then, by combining (4.14) with a simple
Taylor expansion argument, we see that |Bs Φk pξ; sq| Á 2kε . Therefore, by repeated
integration-by-parts, we obtain (4.10) for
ż
bk,k{3 p2 ξ; tq :“ e´itΦk pξ;sq αkJ pξ; t; 2´k{3 sqηp2´εk sqds
J ´k
R
k
and some ek,k{3 P RapDecp2 q.
By (4.14), we have distpθ2 pξq, Jq À 2´k{3`εk ă δ 2 for ξ P suppξ bJk,k{3 . On the
other hand, (4.11) now immediately follows for N “ 0, using the triangle inequality.
For larger values of N , the bounds follow from the estimate
|Φk pξ; sq| À 23kε for all pξ, t; sq P supp αkJ with |s| À 2εk .
which is again a consequence of (4.14) and Taylor expansion.
ii) If upξq ą 0, we know from Lemma 4.1 that the phase function s ÞÑ xγpsq, ξy has
no critical points, and one can therefore obtain rapid decay of the portion of the
multiplier where this condition holds; see [2, Lemma 8.1] for similar arguments. We
 ESTIMATES FOR THE HELICAL MAXIMAL FUNCTION 13

thus focus on the portion of the multiplier where upξq ă 0. Arguing analogously to
the proof of part i), for a given ρ ą 0 we may define
ż
k´3ℓ ˘
J,˘ ´k Φk,ℓ pξ;sq J,˘
bk,ℓ p2 ξ; tq :“ 2 pk´ℓq{2 ´ℓ
2 e´it2 αk,ℓ pξ; t; 2´ℓ sqηpρ´1 sq ds
R

where:
J,˘ p 3 zt0u ˆ R ˆ Rq satisfy
‚ The symbols αk,ℓ P C 8 pR

|upξq| „ 2k´2ℓ , |v ˘ pξq| „ 2k´ℓ , |ξ| ď 2k`1 , |t| ď 4, θ1˘ pξq ` 2´ℓ s P J
J,˘ J,˘
for pξ; t; 2´ℓ sq P supp αk,ℓ and |BtN BsM αk,ℓ pξ; t; sq| ÀM,N 1 for all M , N P
N0 ;
‚ The phase Φ˘ k,ℓ is given by

Φ˘
k,ℓ pξ; sq :“ 2
´pk´3ℓq
xγpθ1˘ pξq ` 2´ℓ sq ´ γ ˝ θ1˘ pξq, ξy.

Moreover, s “ 0 is the only critical point of s ÞÑ Φ˘

k,ℓ pξ; sq on the support
J,˘
of αk,ℓ if |s| ď ρ for ρ ą 0 sufficiently small.
An integration-by-parts argument similar to that in [2, Lemma 8.1] then shows
(4.12) holds for some ek,ℓ P RapDecp2k q.
J,˘
By the support properties of the αk,ℓ and Lemma 4.2, we have

distpθ2 pξq, Jq ď |θ2 pξq ´ θ1˘ pξq| ` distpθ1˘ pξq, Jq À 2´ℓ ă δ 2 for ξ P suppξ bJ,˘
k,ℓ

and ℓ P Λpkq. On the other hand, (4.13) for N “ 0 follows from van der Corput’s
2 J,˘
lemma, since |Bss Φ˘
k,ℓ pξ; 0q| “ 2
´k`ℓ ˘
|v pξq| „ 1 on supp αk,ℓ . For N ą 0, it suffices
to show that
ˇż k´3ℓ ˘ ˘N J,˘ ˇ
Φk,ℓ pξ;sq k´3ℓ ˘
`
ˇ e´it2 2 Φk,ℓ pξ; sq αk,ℓ pξ; t; 2´ℓ sq dsˇ À 2´pk´3ℓq{2 . (4.15)
ˇ ˇ
R

By Taylor expansion, we have

s2 “ ´k`ℓ ˘ ‰
Φ˘
k,ℓ pξ; sq “ 2 v pξq ` 2´k ωpξ; sqs (4.16)
2
J,˘
where |ωpξ; sq| À 2k for all pξ; t; sq P supp αk,ℓ ; in particular, |Φ˘
k,ℓ pξ; sq| À 1 in
J,˘
the support of αk,ℓ . In view of the factor s2 in (4.16), the bound (4.15) is again
a consequence of van der Corput’s lemma with second-order derivatives (in the
specific form of, for example, [22, Lemma 1.1.2]). □

5. Proof of the L2 Ñ L3 trilinear estimate

In this section we prove the key trilinear estimate from Proposition 3.5. After
massaging the operator into a suitable form, this is a consequence the localised
multilinear restriction inequality from Theorem 2.2.
14 D. BELTRAN, J. DUNCAN, AND J. HICKMAN

5.1. Reduction to multlinear restriction. Define the Fourier integral operators

ż
J
Uk,k{3 f px, tq :“ eipxx,ξy´tq2 pξqq bJk,k{3 p2´k ξ, tqfˆpξq dξ
p3
R
and
ÿż ˘
J
Uk,ℓ f px, tq :“ eipxx,ξy´tq1 pξqq bJ,˘
k,ℓ p2
´k
ξ, tqfˆpξq dξ
p3
R
˘
3,sep
for ℓ P Λpkq. Let J P J pδq and ℓJ “ pℓJ qJPJ with ℓJ P Λpkq. In light of
Lemma 4.3, to prove Proposition 3.5, it suffices to show
›ź ›
1{3
ź
|UJ fJ |1{3 › 3 3`1 Àε δ ´Op1q 2k{6´|ℓJ |{9`εk (5.1)
› ›
› }fJ }L2 pR3 q
L pB p0,1qq
JPJ JPJ
J 3
for UJ :“ and fJ P SpR q for J P J. This reduction follows from a standard
Uk,ℓ J
J J
localisation argument since the kernels Kk,ℓ associated to the propagators Uk,ℓ
satisfy the bounds
J
|Kk,ℓ px, tq| ÀN 2kpd´N q |x|´N for all |x| Á 1, N ą 0,
via an integration-by-parts argument.
We may remove the t-dependence from the symbols bJk,k{3 and bJ,˘ k,ℓ using a stan-
dard Fourier series expansion argument. Owing to the L2 -norm on the right-hand
side of (5.1), we may also freely exchange f and fˆ. Thus, after rescaling, we are
led to consider operators of the form
ż
J
Tk,k{3 gpx, tq :“ eipxx,ξy´tq2 pξqq bJk,k{3 pξqgpξq dξ
p3
R
and
ÿż ˘
J
Tk,ℓ gpx, tq :“ eipxx,ξy´tq1 pξqq bJ,˘
k,ℓ pξqgpξq dξ
p3
R
˘

for ℓ P Λpkq, where the symbols bJk , bJ,˘ 8 p3

k,ℓ P C pR q are bounded in absolute value
by 1 and further satisfy
supp bJk,k{3 Ď ξ P B 3 p0, 10q : distpθ2 pξq, Jq ă δ and |upξq| À 2´2k{3`4εk{3
␣ (
(5.2)
and
supp bJ,˘ 3
␣ ´2ℓ
(
k,ℓ Ď ξ P B p0, 10q : distpθ2 pξq, Jq ă δ and |upξq| „ 2 (5.3)
for ℓ P Λpkq. In particular, to prove the desired estimate (5.1), it suffices to show
›ź ›
1{3
ź
|TJ gJ |1{3 › 3 3`1 k Àε δ ´Op1q 2´|ℓJ |{9`εk (5.4)
› ›
› }gJ }L2 pR3 q
L pB p0,2 qq
JPJ JPJ
J 3
for TJ :“ Tk,ℓ J
and gJ P SpR q for J P J and ℓJ “ pℓJ qJPJ with ℓJ P Λpkq.
5.2. Verifying the hypotheses of Theorem 2.2. Enumerate the intervals in J
as J1 , J2 , J3 so that, writing ℓi :“ ℓJi and Ti :“ TJi , we have 0 ă ℓ1 ď ℓ2 ď ℓ3 ď
k{3, ℓi P Λpkq. The trilinear estimate (5.4) is a consequence of Theorem 2.2 where
we exploit the additional localisation of the symbol of T3 to the set |upξq| À 2´2ℓ3 .3
In order to apply Theorem 2.2, we must verify the regularity and transversality
hypotheses.

3Or the slightly larger set |upξq| À 2´2k{3`4εk{3 in the case ℓ “ k{3.
3
 ESTIMATES FOR THE HELICAL MAXIMAL FUNCTION 15

We begin by noting, as a consequence of the definition of the functions θ2 and

θ1˘ , that
∇q2 pξq “ γ ˝ θ2 pξq ` upξq∇θ2 pξq and ∇q1˘ pξq “ γ ˝ θ1˘ pξq. (5.5)

Regularity hypothesis. We first show that the functions q2 and q1˘ all satisfy (at
least) the C 1,1{2 condition. It is easy to see that the function q2 is C 1,1 in the sense
that
|∇q2 pξ1 q ´ ∇q2 pξ2 q| À |ξ1 ´ ξ2 | for all ξ1 , ξ2 P supp bJk,k{3 .
On the other hand, the functions q1˘ are less regular and only satisfy a C 1,1{2
condition, as first observed in [18, Lemma 3.5].
Lemma 5.1. For ℓ P Λpkq, we have

|∇q1˘ pξ1 q ´ ∇q1˘ pξ2 q| À min 2´ℓ , 2ℓ |ξ1 ´ ξ2 | ` |ξ1 ´ ξ2 | À |ξ1 ´ ξ2 |1{2
␣ (

for all ξ1 , ξ2 P supp bJ,˘

k,ℓ .

Proof. Fix ξ1 , ξ2 P supp bJ,˘

k,ℓ . In view of (5.5), it suffices to show

|θ1˘ pξ1 q ´ θ1˘ pξ2 q| À |ξ1 ´ ξ2 |1{2 . (5.6)

By differentiating the defining function for θ1˘ , we see that
|γ 1 ˝ θ1˘ pξq|
|∇θ1˘ pξq| “ „ 2ℓ for ξ P supp bJ,˘
k,ℓ .
|v ˘ pξq|
Thus, the fundamental theorem of calculus implies4
|θ1˘ pξ1 q ´ θ1˘ pξ2 q| À 2ℓ |ξ1 ´ ξ2 |. (5.7)
On the other hand, by Lemma 4.2,
|θ1˘ pξ1 q ´ θ1˘ pξ2 q| ď |θ1˘ pξ1 q ´ θ2 pξ1 q| ` |θ2 pξ1 q ´ θ2 pξ2 q| ` |θ2 pξ2 q ´ θ1˘ pξ2 q|
À |ξ1 ´ ξ2 | ` 2´ℓ . (5.8)

Combining (5.7) and (5.8), we deduce that

␣ (
|θ1˘ pξ1 q ´ θ1˘ pξ2 q| À min 2´ℓ , 2ℓ |ξ1 ´ ξ2 | ` |ξ1 ´ ξ2 |
which is precisely the first inequality in (5.6). Furthermore,

min 2´ℓ , 2ℓ |ξ1 ´ξ2 | “ min 2´ℓ |ξ1 ´ξ2 |´1{2 , 2ℓ |ξ1 ´ξ2 |1{2 |ξ1 ´ξ2 |1{2 ď |ξ1 ´ξ2 |1{2 ,
␣ ( ␣ (

and, since |ξj | À 1 for ξj P supp bJ,˘

k,ℓ , the second inequality in (5.6) immediately
follows. □

4Here we must be a little careful in applying the fundamental theorem of calculus because
the crucial condition |upξq| „ 2´2ℓ does not hold on a convex set. If |ξ1 ´ ξ2 | ! 2´2ℓ , then this
presents no problem. However, if |ξ1 ´ ξ2 | Á 2´2ℓ , then to apply the the fundamental theorem of
calculus we construct a continuous, piecewise smooth curve connecting ξ1 to ξ2 which consists of
two linear segments of length Op2´2ℓ q and a curve of length Op|ξ1 ´ ξ2 |q which traverses the level
set tξ P Bp0, 10q : upξq “ 2´2ℓ u.
16 D. BELTRAN, J. DUNCAN, AND J. HICKMAN

Transversality hypothesis. We now turn to verify the transversality hypothesis from

(2.1). This involves estimating expressions of the form
ˇ ˆ ˙ˇ
ˇ det ∇Q1 pξ1 q ∇Q2 pξ2 q ∇Q3 pξ3 q ∇upξ3 q ˇ
ˇ ˇ
ˇ 1 1 1 0 ˇ

where each Qj is either of the functions q2 or q1˘ . The columns where Qj “ q2 are
slightly more complicated since the formula for ∇q2 pξq in (5.5) involves multiple
terms. However, we can always treat the second term as an error and effectively
ignore it. Indeed, if Qj “ q2 , then we must have ℓj “ k{3 and so we consider
ξ P supp bJk,k{3 . In this case, |upξq| À 2´2k{3`4εk{3 and, by differentiating the
defining equation for θ2 , we also have |∇θ2 pξq| À 1. Since k is large, we can
therefore think of ∇q2 pξq as a tiny perturbation of γ ˝ θ2 pξq on supp bJk,k{3 . On the
other hand, for the final column we have

∇upξq “ γ 1 ˝ θ2 pξq. (5.9)

In view of the support conditions (5.2) and (5.3) and the derivative formulæ
(5.5) and (5.9), the transversality hypothesis would follow from the bound
ˇ ˆ 1
˙ˇ
ˇdet γps1 q γps2 q γps3 q γ ps3 q ˇ Á |V ps1 , s2 , s3 q||s1 ´ s3 ||s2 ´ s3 |
ˇ ˇ
(5.10)
ˇ 1 1 1 0 ˇ

for all s1 , s2 , s3 P I, where here and below

ź
V pt1 , . . . , tm q :“ pti ´ tj q
1ďiăjďm

denotes the Vandermonde determinant in the variables t1 , . . . , tm P R. In order to

make this reduction, we use the bound |θ2 pξq ´ θ1˘ pξq| À 2´ℓ ă δ 8 from Lemma 4.2.
For ℓ3 P Λpkq, we can think of θ2 pξq and θ1˘ pξq as approximately equal; this allows
us to reduce to a situation involving only three variables s1 , s2 , s3 . Here we use the
fact that the right-hand side of (5.10) is bounded below by (a constant multiple of)
δ 5 for si P I with distpsi , Ji q ă 2δ for 1 ď i ď 3 and J “ pJ1 , J2 , J3 q P J3,sep pδq.
By repeated application of the fundamental theorem of calculus, we may express
the left-hand determinant in (5.10) as
ż s2 ż s3 ż t3 ż s3 ż u4 ˆ ˙
γps1 q γ 1 pt2 q γ 2 pu3 q γ 3 pv4 q
det dv4 du4 du3 dt3 dt2 .
s1 s2 t2 t3 u3 1 0 0 0

By continuity and the reductions in §3.3, the inner determinant is single-signed and
bounded below in absolute value by some constant.
By the observations of the previous paragraph, the left-hand side of (5.10) is
comparable to the same expression but with γ replaced by the moment curve
γ˝ psq :“ ps, s2 {2, s3 {6q. Consequently, it suffices to prove (5.10) for this partic-
ular curve. However, in this case the left-hand side of (5.10) corresponds to (the
absolute value of a scalar multiple of) Bt V ps1 , s2 , s3 , s3 ` tq|t“0 . A simple calculus
exercise shows this agrees with the expression appearing in the right-hand side of
(5.10), as required. For similar arguments, see [14, 18].
 ESTIMATES FOR THE HELICAL MAXIMAL FUNCTION 17

6. L12 -local smoothing

In this section we upgrade the Lp -local smoothing estimates of [21] for p ě 12 by
exploiting the localisation of the spatio-temporal Fourier transform of mrak,ℓ spD; tqf
with respect to the 2-dimensional cone Γ2 in pξ, τ q-space from the introduction. The
arguments are similar to those of [19] and [3]. Crucially, we apply a decoupling in-
equality from [3], which is a conic variant of the celebrated decoupling inequality
for non-degenerate curves [10, 15]. After this step, the remainder of the argument
is similar to that of [21].5
8
6.1.
ř Decomposition along the curve. Fix ζ P C pRq with supp ζ Ď I such that
µPZ ζp ¨ ´ µq ” 1. For k P N and ℓ P Λpkq, we write
ÿ
ak,ℓ pξ; t; sq “ ak,ℓ pξ; t; sqζp2ℓ θ2 pξq ´ µq.
µPZ

Using stationary phase arguments, as in the proof of Lemma 4.3, we may localise
s to lie in a neighbourhood of θ2 pξq. Let ρ ą 0 be a small ‘fine tuning’ constant
chosen to satisfy the requirements of the forthcoming argument; for instance, we
may take ρ :“ 10´6 . We then define
aµk,ℓ pξ; t; sq :“ ak,ℓ pξ; t; sqζp2ℓ θ2 pξq ´ µqη ρ2ℓ ps ´ θ2 pξqq for ℓ P Λpkq,
` ˘

aµk,k{3 pξ; t; sq :“ ak,ℓ pξ; t; sqζp2k{3 θ2 pξq ´ µqη ρ2k{3´εk ps ´ θ2 pξqq

` ˘

for all µ P Z.
Lemma 6.1. Let 2 ď p ă 8. For all k P N and ℓ P Λpkq, we have
› ” ÿ µ ı ›
›m ak,ℓ ´ ak,ℓ pD; ¨ qf › ÀN,ε,p 2´kN }f }Lp pR3 q for all N P N.
› ›
Lp pR3`1 q
µPZ

By Lemma 4.2, if |s ´ θ2 pξq| " 2´ℓ , then s is far from any roots θ1˘ pξq of the
phase function. Hence Lemma 6.1 follows from non-stationary phase, as in the
analysis of the error term in Lemma 4.3 ii). Moreover, the proof is very similar to
that of [2, Lemma 8.1] and we therefore omit the precise details.6
The support properties of the symbols aµk,ℓ are best understood in terms of the
Frenet frame. Recall, given a smooth non-degenerate curve γ : I Ñ Rd , the Frenet
frame is the orthonormal basis te1 psq, . . . , ed psqu resulting from applying the Gram–
Schmidt process to the vectors tγ 1 psq, . . . , γ pdq psqu. With this setup, given 0 ă r ď 1
and s P I, recall the definition of the p1, rq-Frenet boxes π1 ps; rq introduced in [2];
namely,
p 3 : |xej psq, ξy| À r3´j for j “ 1, 2, |xe3 psq, ξy| „ 1 .
␣ (
π1 ps; rq :“ ξ P R
We have the following support property.
Lemma 6.2. Let k P N, ℓ P Λpkq and µ P Z. Then
suppξ aµk,ℓ Ď 2k ¨ π1 psµ ; 2´ℓ q.
The proof is similar to that of [2, Lemma 8.2, (a)], so we omit the details.
5Decoupling is also used in [21], but only with respect to a cone in ξ-space, leading to non-sharp
regularity estimates.
6The argument is in fact entirely the same as the proof of the case tk{3u ď ℓ ď tk{3u from [2,
ε
Lemma 8.1].
18 D. BELTRAN, J. DUNCAN, AND J. HICKMAN

6.2. Spatio-temporal localisation. The symbols are further localised with re-
spect to the Fourier transform of the t-variable. In particular, define the homoge-
neous phase function q2 pξq :“ xγ ˝ θ2 pξq, ξy as in Lemma 4.3 and let
` ˘
χk,ℓ pξ, τ q :“ η 2´pk´3ℓq´4εk pτ ` q2 pξqq for ℓ P Λpkq.
We introduce the localised multipliers mµk,ℓ , defined by
Ft mµk,ℓ pξ; ¨ q pτ q :“ χk,ℓ pξ, τ q Ft mraµk,ℓ spξ; ¨ q pτ q.
“ ‰ “ ‰

Here Ft denotes the Fourier transform acting in the t variable. A stationary phase
argument allows us to pass from mraµk,ℓ s to mµk,ℓ .
Lemma 6.3. Let 2 ď p ă 8. For all k P N, ℓ P Λpkq and µ P Z, we have
› mraµ s ´ mµ pD; ¨ qf › p 3`1 ÀN,ε 2´kN }f }Lp pR3 q
›` ˘ ›
k,ℓ k,ℓ L pR q
for all N P N.
The proof, which is based on a fairly straightforward integration-by-parts argu-
ment, is similar to that of [2, Lemma 8.3] and we omit the details.
To understand the support properties of the multipliers mµk,ℓ , we introduce the
primitive curve „şs ȷ
4 γ
γ̄ : I Ñ R , γ̄ : s ÞÑ 0 .
s
şs şs
Here 0 γ denotes the vector in R3 with ith component 0 γi for 1 ď i ď 3. Note
that γ̄ is a non-degenerate curve in R4 and, in particular,
` ˘ ` ˘
| det γ̄ p1q psq ¨ ¨ ¨ γ̄ p4q psq | “ | det γ p1q psq ¨ ¨ ¨ γ p3q psq | for all s P I.
Let pēj psqq4j“1 denote the Frenet frame associated to γ̄ and consider the p2, rq-Frenet
boxes for γ̄
p 3 ˆR
p : |xēj psq, Ξy| À r4´j for 1 ď j ď 3, |xē4 psq, Ξy| „ 1 ,
␣ (
π2,γ̄ ps; rq :“ Ξ “ pξ, τ q P R
as introduced in [2].
Lemma 6.4. For all k P N, ℓ P Λpkq with ℓ ě r4εks and µ P Z, we have
supp Ft mµk,ℓ Ď 2k ¨ π2,γ̄ psµ ; 24εk 2´ℓ q,
“ ‰

where sµ :“ 2´ℓ µ and Ft denotes the Fourier transform in the t-variable.

The proof, which is based on a fairly straightforward integration-by-parts argu-
ment, is similar to that of [2, Lemma 8.4] and we omit the details.
6.3. A decoupling inequality. With the above observations, we can immediately
apply the decoupling inequalities in [3, Theorem 4.4] associated to the primitive
curve γ̄ to isolate the contributions from the individual mµk,ℓ pD; ¨ q.
Proposition 6.5. Let p ě 12. For all k P N, ℓ P Λpkq, we have
›ÿ › 7
´ÿ ¯1{p
mµk,ℓ pD; ¨ qf › Àε 2Opεkq 2ℓp1´ p q }mµk,ℓ pD; ¨ qf }pLp pR3`1 q .
› ›
›
Lp pR3`1 q
µPZ µPZ

Proof. If ℓ P Λpkq satisfies ℓ ď r4εks, then we bound the left-hand side trivially
using the triangle and the Cauchy–Schwarz inequalities. For the case ℓ ą r4εks, we
partition the family of sets π2,γ̄ psµ ; 24εk 2´ℓ q for |µ| ď 2ℓ into Op24εk q subfamilies,
each forming a p2, 24εk 2´ℓ q-Frenet box decomposition in the language of [3, §4]. In
view of Lemma 6.4 and after a simple rescaling, the result now follows from [3,
Theorem 4.4] applied with d “ 3, n “ 4 to the primitive curve γ̄. □
 ESTIMATES FOR THE HELICAL MAXIMAL FUNCTION 19

6.4. Localising the input function. The Fourier multipliers mµk,ℓ pD; tq induce
a localisation on the input function f . We recall the setup from [2, §8.6]. Given
ℓ P Λpkq and m P Z define
p 3 : |ξ2 ´ ξ3 2´ℓ m| ď C2´ℓ ξ3 and C ´1 2k ď ξ3 ď C2k ,
␣ (
∆k,ℓ pmq :“ ξ P R
where C ě 1 is an absolute constant, chosen sufficiently large so that the following
lemma holds.
Lemma 6.6 ([2, Lemma 8.8]). If µ P Z, then there exists some mpµq P Z such that
` ˘
2k ¨ π1 psµ ; 2´ℓ q Ď ∆k,ℓ mpµq .
Furthermore, for each fixed k and ℓ, given m P Z there are only Op1q values of
µ P Z such that m “ mpµq.
For each µ P Z define the smooth cutoff function
χµk,ℓ pξ2 , ξ3 q :“ η C ´1 p2ℓ ξ2 {ξ3 ´ mpµqq ηpC ´1 2´k ξ3 q ´ ηp2C ´1 2´k ξ3 q .
` ˘` ˘

If pξ1 , ξ2 , ξ3 q P suppξ aµk,ℓ , then Lemmas 6.2 and 6.6 imply χµk,ℓ pξ2 , ξ3 q “ 1. Thus, if
we define the corresponding frequency projection
µ
fk,ℓ :“ χµk,ℓ pDqf,
it follows that
mraµk,ℓ spD; ¨ qf “ mraµk,ℓ spD; ¨ qfk,ℓ
µ
. (6.1)

Lemma 6.7. For all k P N and ℓ P Λpkq, we have

´ÿ ¯1{p
µ p
}fk,ℓ }Lp pR3 q À }f }Lp pR3 q .
µPZ

Proof. The case p “ 2 follows from Plancherel’s theorem via Lemma 6.6 and the
finite overlapping of the sectors ∆k,ℓ pmpµqq. For p “ 8, it is easy to see that
supµ }Fξ´1
2 ,ξ3
rχµk,ℓ s}L1 pR2 q À 1; indeed this is immediate for k “ ℓ “ 0 and the
general case follows since χµk,ℓ pξ2 , ξ3 q “ χµ0,0 p2ℓ´k ξ2 , 2´k ξ3 q. Interpolating these
two cases, using mixed-norm interpolation (see, for instance, [8, §5.6]) concludes
the proof. □
6.5. Local smoothing for the mraµk,ℓ spD, tq. We recall the following result, which
follows from [21, Theorem 4.1] when combined with the main result from [9].
Theorem 6.8 ([21, Theorem 4.1]). Let γ : I Ñ R3 be a smooth curve and suppose
p 3 zt0u ˆ R ˆ Rq satisfies the symbol conditions
that a P C 8 pR
|Bξα Bti Bsj apξ; t; sq| Àα,i,j |ξ|´|α| for all α P N30 and i, j P N0
and that
|xγ 1 psq, ξy| ` |xγ 2 psq, ξy| Á |ξ| for all pξ; sq P suppξ a ˆ I.
Let p ě 6, ε ą 0 and k ě 1. If ak is defined as in (3.3), then
´ż 2 ¯1{p
}mrak spD; tqf }pLp pR3 q dt Àε,p 2´2k{p`εk }f }Lp pR3 q .
1

By combining Theorem 6.8 with the rescaling from §4.2 we obtain the following
bound for our multipliers mraµk,ℓ s.
20 D. BELTRAN, J. DUNCAN, AND J. HICKMAN

Proposition 6.9. Let p ě 6. For all k P N, ℓ P Λpkq and µ P Z we have

´ż 2 ¯1{p
}mraµk,ℓ spD; tqf }pLp pR3 q dt Àε 2Opεkq 2´2pk´3ℓq{p´ℓ }f }Lp pR3 q . (6.2)
1

Proof. The argument is essentially the same proof as that of [21, Proposition 5.1].
We distinguish two cases:
Case: ℓ “ k{3. The result follows from interpolation between the elementary esti-
mates
sup }mraµk,k{3 sp ¨ ; tq}M 2 pR3 q À 2´k{3`Opεkq ,
tPR

sup }mraµk,k{3 sp ¨ ; tq}M 8 pR3 q À 2´k{3`Opεkq .

tPR

Both these inequalities are consequences of the size of the s support of aµk,k{3 . The
first is trivial. The second can be deduced, for instance, by adapting arguments
from [3, §5.6].
Case: ℓ P Λpkq. Fix ℓ P Λpkq, µ P Z and set σ :“ sµ and λ :“ 2´ℓ . With the
notation from §4.2, we define
γ̃ :“ γσ,λ , ã :“ paµk,ℓ qσ,λ , f˜ :“ fσ,λ , M̃ :“ prγsσ,λ q´J .
Thus, in view of (4.6), we have
mγ raµk,ℓ spD; tqf pxq “ λ mγ̃ rãspD; tqf˜ M̃ J x ´ tγpσq .
` ˘` ˘
(6.3)
We claim γ̃ and ã satisfy the hypotheses of Theorem 6.8. By (4.4), we have
|ξ| „ 2k´3ℓ for all ξ P suppξ ã. (6.4)
Combining this with (4.3) and (3.16), we have
|xγ̃ 1 psq, ξy| ` |xγ̃ 2 psq, ξy| Á |ξ| for all pξ; t; sq P supp ã. (6.5)
On the other hand, let θ̃2 and ũ be the functions defined in §3.4, but now with
respect to the curve γ̃. It follows that
σ ` λθ̃2 pξq “ θ2 ˝ M̃ pξq and ũpξq “ λu ˝ M̃ pξq
and so
` ˘ ` ˘
ãpξ; t; sq “ pak qσ,λ pξ; t; sqβ 2´k`3ℓ ũpξq ζpθ̃2 pξq ´ σ ´ µqη ρps ´ θ̃2 pξqq .
Using the fact that |xγ̃ p3q ˝ θ̃2 pξq, ξy| „ 2k´3ℓ „ |ξ| on the support of ã, it is a
straightforward exercise to show that
|Bξα θ̃2 pξq| Àα |ξ|´|α| and 2´k`3ℓ |Bξα ũpξq| Àα |ξ|´|α|
hold for all ξ P suppξ ã and α P N30 . The derivative bounds

|Bξα Bti Bsj ãpξ; t; sq| Àα,i,j |ξ|´|α| for all α P N30 and i, j P N0 (6.6)
then easily follow, noting that the derivatives of pak qσ,λ can be controlled following
the discussion at the end of §4.2. ř8
As a consequence of (6.4), we may write ã “ j“0 ãj where each ãj is a localised
symbol as defined in (3.3) and the only non-zero terms of the sum correspond to
 ESTIMATES FOR THE HELICAL MAXIMAL FUNCTION 21

values of j satisfying 2j „ 2k´3ℓ . In view of (6.5) and (6.6), for p ě 6 we can apply
Theorem 6.8 to obtain
´ż 2 ¯1{p
}mγ̃ rãspD; tqf˜}pLp pR3 q dt Àε 2Opεkq 2´2pk´3ℓq{p }f˜}Lp pR3 q .
1
This, together with (6.3) and an affine transformation in the spatial variables, gives
the desired inequality (6.2). □
6.6. Putting everything together. With the above ingredients, we can now
conclude the proof of the L12 local smoothing estimate.
Proof of Proposition 3.6. By successively applying Lemma 6.1, Lemma 6.3, Propo-
sition 6.5 and a second application of Lemma 6.3, we obtain
´ÿ ¯1{12
}mrak,ℓ spD; ¨qf }L12 pR3`1 q Àε,N 2εk{2 25ℓ{12 }mraµk,ℓ spD; ¨ qf }12
12
L pR3`1 q
µPZ

` 2´kN }f }L12 pR3 q

for any N ą 0. By the localisation (6.1) and Proposition 6.9 we have
´ÿ ¯1{12 ´ÿ ¯1{12
}mraµk,ℓ spD; ¨ qf }12
12
L pR3`1 q À 2εk{2 ´pk´3ℓq{6´ℓ
2 }f µ 12
} 12 3
k,ℓ L pR q .
µPZ µPZ

Combining the above observations with an application of Lemma 6.7 concludes the
proof. □

7. The non-degenerate case

In the non-degenerate case ℓ “ 0 we appeal to the classical (linear) Stein–Tomas
restriction estimate, rather than the trilinear theory from §5.
Proposition 7.1. Let γ : I Ñ R3 be a smooth curve and suppose that a P
p 3 zt0u ˆ R ˆ Rq satisfies the symbol conditions
C 8 pR
|Bξα Bti Bsj apξ; t; sq| Àα,i,j |ξ|´|α| for all α P N30 and i, j P N0
and that
|xγ 1 psq, ξy| ` |xγ 2 psq, ξy| Á |ξ| for all pξ; sq P suppξ a ˆ I. (7.1)
Let k ě 1. If ak is defined as in (3.3), then
´ż 2 ¯1{6
}mrak spD; tqf }6L6 pR3 q dt À 2k{3 }f }L2 pR3 q .
1

Proof. Decomposing the symbol a into sufficiently many pieces with small ξ and s
support, the non-degeneracy condition (7.1) can be strengthened to the following:
there exists B ą 1 such that
B ´1 |ξ| ď |xγ 2 psq, ξy| ď B|ξ| for all pξ; sq P suppξ a ˆ I (7.2)
and there exists s˚ P I such that
|xγ 1 ps˚ q, ξy| ď 10´10 B|ξ| for all ξ P suppξ a; (7.3)
see, for instance, [21, §4] or [12, Chapter 2] for details of this type of reduction,
which relies on the fact that the oscillatory integral mrak s is rapidly decreasing if
22 D. BELTRAN, J. DUNCAN, AND J. HICKMAN

the phase function s ÞÑ xγpsq, ξy has no critical points. Under conditions (7.2) and
(7.3), there exists a unique smooth mapping θ1 : suppξ a Ñ I such that
xγ 1 ˝ θ1 pξq, ξy “ 0 for all ξ P suppξ a. (7.4)
Let q1 pξq :“ xγ ˝ θ1 pξq, ξy. Arguing as in the proof of Lemma 4.3, we may use (7.2)
and van der Corput’s lemma with second-order derivatives to write
mrak spξ; tq “ 2´k{2 e´itq1 pξq bk p2´k ξ; tqpξ; tq
where bk P C 8 pR3`1 q is supported in Bp0, 10q and satisfies
|BtN bk p2´k ξ; tq| ÀN 1 for all pξ; tq P R3`1 and all N P N0 .
Following the reductions of §5.1, we consider an operator T of the form
ż
T gpx, tq :“ eipxx,ξy´tq1 pξqq bpξqgpξq dξ
B 3 p0,1q

p 3 q bounded in absolute value by 1 which, by (7.2), satisfies

for b P C 8 pR
supp b Ď tξ P B 3 p0, 1q : |vpξq| „ 1u where vpξq :“ xγ 2 ˝ θ1 pξq, ξy.
In particular, to prove the lemma, with the above setup it suffices to show
}T g}L6 pB 3`1 p0,2k qq À }g}L2 pR3 q . (7.5)
The inequality (7.5) follows from a generalisation of the classical Stein–Tomas
restriction theorem due to Greenleaf [13] (see also [23, Chapter VIII, §5 C.]). To
apply this result, we need to show that q1 is smooth over the support of b and
satisfies certain curvature conditions.
Arguing as in the proof of Lemma 5.1, we see that |∇θ1 pξq| „ 1 on supp b and,
furthermore, the function q1 is easily seen to be smooth with bounded derivatives
2
over this set. A simple computation shows that the hessian Bξξ q1 pξq is the rank
1 matrix formed by the outer product of the vectors ∇θ1 pξq and γ 1 ˝ θ1 pξq. By
2
elementary properties of rank 1 matrices, Bξξ q1 pξq therefore has a unique non-zero
eigenvalue given by
κpξq :“ xγ 1 ˝ θ1 pξq, ∇θ1 pξqy.
We claim that
|κpξq| „ 1 for all ξ P supp b; (7.6)
geometrically, this means that the surface formed by taking the graph of q1 over
some open neighbourhood of supp b has precisely one non-vanishing principal cur-
vature. This is precisely the geometric condition needed to apply the result of [13]
in order to deduce (7.5). To see (7.6) holds, we take the ξ-gradient of the defining
equation (7.4) for θ1 and then form the inner product with ∇θ1 pξq to deduce that
0 “ xγ 1 ˝ θ1 pξq, ∇θ1 pξqy ` xγ 2 ˝ θ1 pξq, ξy|∇θ1 pξq|2 “ κpξq ` vpξq|∇θ1 pξq|2 .
Since |vpξq| „ |∇θ1 pξq| „ 1 on supp b, the claim follows. □
6 6
One can interpolate Proposition 7.1 with the diagonal L Ñ L local smoothing
result of Pramanik–Seeger [21] (see Theorem 6.8 above) to directly deduce the
desired L4 Ñ L6 estimate for the non-degenerate piece ak,0 introduced in (3.8).
Lemma 7.2. For all k P N and ε ą 0, we have
´ż 2 ¯1{6
}mrak,0 spD, tqf }6L6 pR3 q dt Àε δ ´Op1q 2´k{6`εk }f }L4 pR3 q .
1
 ESTIMATES FOR THE HELICAL MAXIMAL FUNCTION 23

This lemma reduces the proof of Proposition 3.2 to establishing the L4 Ñ L6

bound in (3.6) with ak replaced with the localised symbol ak :“ ak ´ ak,0 .

8. Concluding the argument

Here we conclude the proof of Proposition 3.2 and, in particular, present the
details of the trilinear reduction discussed in §3.2.
Proof (of Proposition 3.2). Fix ε ą 0 and let 0 ă δ ă 1, Mε P N and Cε ě 1 be
constants, depending only on ε, and chosen to satisfy the forthcoming requirements
of the proof. We proceed by inducting on the parameter k. For 0 ď k ď δ ´100 , the
result is trivial and this serves as the base case. We fix k P N satisfying k ą δ ´100
and assume that for 0 ď n ď k ´ 1 the result holds in the following quantified sense.
p 3 zt0u ˆ R ˆ Rq
Induction hypothesis. Let γ P Gpδ0 , Mε q and suppose a P C 8 pR
satisfies the symbol condition
|Bξα Bti Bsj apξ; t; sq| ď |ξ|´|α| for all α P N30 and i, j P N0 with |α|, i, j ď Mε . (8.1)
For all 0 ď n ď k ´ 1, we have
´ż 2 ¯1{6
}mran spD; tqf }6L6 pR3 q dt ď Cε 2´n{6`εn }f }L4 pR3 q .
1
We remark that if Mε P N is chosen sufficiently large, then all the estimates
proved in this paper are uniform over all curves belonging to the class Gpδ0 , Mε q.
We now turn to the inductive step. Fix γ P Gpδ0 , Mε q and a satisfying (8.1) and
suppose ak,0 is defined as in (3.8). Provided Cε is chosen sufficiently large in terms
of δ, we may apply Lemma 7.2 to deduce a favourable bound for the corresponding
multiplier mrak,0 s. It remains to show
´ż 2 ¯1{6
}mrak spD; tqf }6L6 pR3 q dt ď pCε {2q2´k{6`εk }f }L4 pR3 q
1
for the ak symbols as defined in (3.12).
For convenience, write
Uk f px, tq :“ mrak spD, tqf pxq and UkJ f px, tq :“ mJ rak spD, tqf pxq for J P Jpδq.
By fixing an appropriate partition of unity,
ÿ
Uk f “ UkJ f.
JPJpδq

By an elementary argument (see, for instance, [18, Lemma 4.1]), we have a pointwise
bound ÿ ź
|Uk f pzq| À max |UkJ f pzq| ` δ ´1 |UkJ f pzq|1{3 . (8.2)
JPJpδq
JPJ3,sep pδq JPJ
6
Taking L -norms on both sides of (8.2), we deduce that
´ ÿ ¯1{6 ÿ ›ź ›
}Uk f }L6 pR3`1 q À }UkJ f }6L6 pR3`1 q ` δ ´1 |UkJ f |1{3 › .
› ›
›
L6 pR3`1 q
JPJpδq JPJ3,sep pδq JPJ
(8.3)
The first term on the right-hand side of (8.3) can be estimated using a combi-
p 3 q be
nation of rescaling and the induction hypothesis. To this end, let β̃ P Cc8 pR
a non-negative function satisfying β β̃ “ β and |ξ| „ 1 for ξ P supp β̃, and define
24 D. BELTRAN, J. DUNCAN, AND J. HICKMAN

β̃ k pξq :“ β̃p2´k ξq for k P N0 . For J P Jpδq fix ψ̃J P Cc8 pRq satisfying supp ψ̃J Ď 4¨J,
ψ̃J prq “ 1 for r P 2 ¨ J and |BrN ψ̃J prq| ÀN |J|´N for all N P N0 . We define the
Fourier projection fJ of f by
fˆJ pξq :“ χJ pξqfˆpξq where χJ pξq :“ β̃ k pξqψ̃J ˝ θ2 pξq.
By stationary phase arguments, similar to the proof of Lemma 6.1, we then have
}UkJ f }L6 pR3`1 q À }UkJ fJ }L6 pR3`1 q ` 2´10k }f }L4 pR3 q for each J P Jpδq.
Fix J P Jpδq with centre cJ . By the scaling relation (4.7), we have
}UkJ }L4 pR3 qÑL6 pR3`1 q À δ 1{2 }Ũk }L4 pR3 qÑL6 pR3`1 q
where Ũk is the rescaled operator
Ũk f px, tq :“ mγ̃ rãspD, tqf pxq for γ̃ :“ γcJ ,δ{2 , ã :“ pak ¨ ψJ qcJ ,δ{2 ,
with the rescalings as defined in (4.2) and (4.5). Note that γ̃ P Gpδ0 , Mε q and,
arguing as in the proof of Proposition 6.9, the symbol ã satisfies (8.1) (perhaps with
a slightly larger constant, but this can be factored out of the symbol). Furthermore,
in view of (4.4), the symbol ã satisfies
p 3 : |ξ| „ δ 3 2k u.
suppξ ã Ă tξ P R
ř8
In particular, we can write ã “ n“0 ãn where each ãn is a localised symbol as in
(3.3) and the only non-zero terms of this sum correspond to values of n satisfying
2n „ δ 3 2k . Thus, by the induction hypothesis,
}Ũk }L4 pR3 qÑL6 pR3`1 q À Cε pδ 3 2k q´1{6`ε “ Cε δ ´1{2`3ε 2´k{6`εk .
Combining these observations,
´ ÿ ¯1{6 ´ ÿ ¯1{4
}UkJ f }6L6 pR3`1 q À Cε δ 3ε 2´k{6`εk }fJ }4L4 pR3 q
JPJpδq JPJpδq
3ε ´k{6`εk
À Cε δ 2 }f }L4 pR3 q , (8.4)
where the final estimate follows from the orthogonality of the fJ via a standard
argument.7
On the other hand, each summand in the second term on the right-hand side
of (8.3) can be estimated using Proposition 3.3. In particular, for each fixed J P
J3,sep pδq we have
›ź ›
|UkJ f |1{3 › 6 3`1 Àε δ ´E 2´k{6`εk }f }L4 pR3 q (8.5)
› ›
›
L pR q
JPJ
for some absolute constant E ě 1.
Combining (8.3), (8.4) and (8.5), we deduce that
}Uk f }L6 pR3`1 q ď Cε Cε δ 3ε ` δ ´E´4 2´k{6`εk }f }L4 pR3 q ,
` ˘

7Indeed, by interpolation it suffices to show

ÿ
}fJ }2L2 pR3 q À }f }2L2 pR3 q and max }fJ }L8 pR3 q À }f }L8 pR3 q .
JPJpδq
JPJpδq

The former follows from Plancherel’s theorem and the finite overlap of the Fourier supports of the
fJ . For the latter, it suffices to show the kernel estimate supJPJpδq }F ´1 χJ }1 À 1. To see this,
we apply a rescaling as in the proof of Proposition 6.9, which transforms χJ into a function with
favourable derivative bounds.
 ESTIMATES FOR THE HELICAL MAXIMAL FUNCTION 25

where the constant Cε ě 1 is an amalgamation of the various implied constants

appearing in the preceding argument. Now suppose δ ą 0 and Cε have been
chosen from the outset so as to satisfy Cε δ 3ε ď 1{4 and Cε ě 4Cε δ ´E´4 . It then
follows that
}Uk f }L6 pR3`1 q ď pCε {2q2´k{6`εk }f }L4 pR3 q ,
which closes the induction and completes the proof. □

9. Necessary conditions
In this section we provide the examples that show that Mγ fails to be bounded
from Lp Ñ Lq whenever p1{p, 1{qq R T . By a classical result of Hörmander [17],
Mγ cannot map Lp Ñ Lq for any p ą q. Failure at the point p1{3, 1{3q was
already shown in [18] via a modification of the standard Stein-type example for the
circular maximal function. The line joining p1{3, 1{3q and p1{4, 1{6q is critical via
a Knapp-type example, whilst the line joining p0, 0q and p1{4, 1{6q is critical from
the standard example for fixed time averages.
9.1. The Knapp example. By an affine rescaling (as in §4.2), we may assume
γ pjq p0q “ ej for 1 ď j ď 3, where ej denotes the standard basis vector. Thus, if γ˝
denotes the moment curve as in §4.2, then γpsq “ γp0q ` γ˝ psq ` Ops4 q for s P I.
Furthermore, we may assume without loss of generality that a :“ γ3 p0q ą 0. Given
δ ą 0, let fδ :“ 1Rδ where
Rδ :“ ty P R3 : |yj | ă δ j , 1 ď j ď 3u.
Clearly, }fδ }Lp pR3 q À δ 6{p . Consider the domain
Eδ :“ x P R3 : |xj ´ x3 γj p0q{a| ă δ j {2, j “ 1, 2, a ď x3 ď 2a .
␣ (

By the moment curve approximation, there exists a constant cγ ą 0 such that if

|s| ă cγ δ, then the following holds. If x P Eδ and tpxq :“ x3 {a, then
|xj ´ tpxqγj psq| ď |xj ´ tpxqγj p0q| ` |tpxq||γj p0q ´ γj psq| ă δ j for j “ 1, 2
and
|x3 ´ tpxqγ3 psq| “ |tpxq||γ3 p0q ´ γ3 psq| ă δ 3 .
Thus, we conclude that x ´ tpxqγpsq P Rδ for all |s| ă cγ δ and therefore
}Mγ fδ }Lq pR3 q Á δ|Eδ |1{q Á δ 1`3{q .
The bound }Mγ fδ }Lq pR3 q À }fδ }Lp pR3 q therefore implies δ 1`3{q À δ 6{p ; letting δ Ñ 0,
this can only hold if 1 ` 3q ě p6 . This gives rise to the line joining p1{3, 1{3q and
p1{4, 1{6q in Figure 1.
9.2. Dimensional constraint. This is the standard example for Lp Ñ Lq bound-
edness for the fixed time averages. Given 0 ă δ ă 1, consider gδ “ 1Nδ pγq where
Nδ pγq :“ tx P R3 : |x ` γpsq| ď δ for some s P Iu.
Clearly, }gδ }Lp pR3 q À δ 2{p . Furthermore, x ´ γpsq P Nδ pγq for all |x| ď δ. This read-
ily implies Mγ gδ pxq ě A1 gδ pxq Á 1 for |x| ď δ, and consequently, }Mγ gδ }Lq pR3 q Á
δ 3{q . The bound }Mγ gδ }Lq pR3 q À }fδ }Lp pR3 q implies δ 3{q À δ 2{p ; letting δ Ñ 0, this
can only hold if 3q ě p2 . This gives rise to the line joining p0, 0q and p1{4, 1{6q in
Figure 1.
26 D. BELTRAN, J. DUNCAN, AND J. HICKMAN

Appendix A. Localised multilinear restriction estimates

Here we present the proof of Theorem 2.2. We use a simple Fubini argument
to essentially reduce the problem to particular cases of the multilinear restriction
inequalities from [6, Theorem 1.3] and [7, Theorem 5.1]. More precisely, we require
low-regularity versions of these results which apply to C 1,1{2 -hypersurfaces. How-
ever, the arguments of [6] and [7] extend to cover the C 1,α -class for any α ą 0 by
incorporating minor modifications to the induction-on-scale scheme as in the proof
of [18, Theorem 3.6]; we omit the details.
Proof (of Theorem 2.2). Let δ ą 0 be a small constant, which is independent of
µ and R and chosen to satisfy the forthcoming requirements of the proof. We
may assume without loss of generality that 0 ă µ ă δ, since otherwise the desired
estimate follows from the C 1,1{2 extension of the Bennett–Carbery–Tao multilinear
inequality [7, Theorem 5.1].
By localising the operators and applying a suitable rotation to the coordinate
domain, we may assume that there exists an open domain U31 Ď R p 2 and a smooth
1
map γ : U3 Ñ R such that
␣ (
ξ P supp a3 : upξq “ 0 “ tps, γpsqq : s P U31 u
and, moreover,
␣ (
supp a3 Ď ps, γpsq ` rq : s P U31 and |r| ă µ .
By differentiating the defining identity for γ, we observe that
pBξj uqps, γpsqq ` pBsj γqpsqpBξ3 uqps, γpsqq “ 0 for j “ 1, 2. (A.1)
By a change of variables, we may write
żµ
E3 f3 px, tq “ eirx3 E3,r f3,r px, tq dr
´µ

where f3,r psq :“ f3 ps, γpsq ` rq and

ż
E3,r gpx, tq :“ eipxΓpsq,xy`tQ3 ps,γpsq`rqq a3,r psqgpsq ds
p2
R
for Γpsq :“ ps, γpsqq and a3,r psq :“ a3 ps, γpsq ` rq. For each fixed |r| ă µ, the
operator E3,r is the extension operator associated to the 2-surface
␣ (
Σ13,r :“ ps, γpsq, Q3 ps, γpsq ` rqq : s P U31 .
When r “ 0, it follows from (A.1) that
!ˆ ˙ ˆ ˙)
´∇Q3 ˝ Γpsq ∇u ˝ Γpsq
span , “ Nξ Σ13,0 for ξ :“ pΓpsq, Q3 ˝ Γpsqq;
1 0
(A.2)
that is, the span of the two vectors is equal to the normal space to Σ13,0 at ξ.
After applying a simple Fubini–Tonelli argument, the problem is reduced to
showing
żµ ż ź2 3
ź
|Ej fj px, tq||E3,r f3,r px, tq| dxdtdr ÀQ,ε Rε µ1{2 }fj }L2 pUj q . (A.3)
´µ Bp0,Rq j“1 j“1

The key claim is that for each |r| ă µ, the trio of extension operators pE1 , E2 , E3,r q
satisfy the hypothesis of [6, Theorem 1.3]. In particular, provided δ ą 0 is chosen
 ESTIMATES FOR THE HELICAL MAXIMAL FUNCTION 27

sufficiently small, our transversality hypothesis (2.1) implies that the normal spaces
to the submanifolds Σ1 , Σ2 , Σ3,r factorise the space Rp 4 in the sense that
p4
Nξ1 Σ1 ‘ Nξ2 Σ2 ‘ Nξ3 Σ13,r “ R
for all |r| ă µ and all choices of ξ1 P Σ1 , ξ2 P Σ2 , ξ3 P Σ13,r . To see this, we first prove
the r “ 0 case by combining (A.2) and (2.1), and then extend to all |r| ă µ ă δ
using continuity. Consequently, we can use the formula proved in [5, Proposition
1.2] together with (2.1) to conclude that the Brascamp–Lieb constant associated
to the orthogonal projections onto the tangent spaces Tξ1 Σ1 , Tξ2 Σ2 , Tξ3 Σ3,r (with
Lebesgue exponents p1 “ p2 “ p3 “ 1{2) is uniformly bounded. We refer to [5, 6]
for the relevant definitions. This is precisely the hypothesis of [6, Theorem 1.3] and
invoking (a suitable C 1,1{2 generalisation of) this result we obtain
ż ź2 ź2
|Ej fj px, tq||E3,r f3,r px, tq| dxdt ÀQ,ε Rε }fj }L2 pUj q }f3,r }L2 pU31 q
Bp0,Rq j“1 j“1

uniformly over all |r| ă µ. We integrate both sides of this inequality with respect
to r and apply the Cauchy–Schwarz inequality to deduce that
żµ ż ź2
|Ej fj px, tq||E3,r f3,r px, tq| dxdtdr
´µ Bp0,Rq j“1

2
ź ´ż µ ¯1{2
ε 1{2
ÀQ,ε R µ }fj }L2 pUj q }f3,r }2L2 pU31 q dr .
j“1 ´µ

The desired estimate (A.3) now follows by reversing the original change of variables.
□

References
[1] Ioan Bejenaru, The multilinear restriction estimate: almost optimality and localization,
Math. Res. Lett. 29 (2022), no. 3, 599–630. MR 4516033
[2] David Beltran, Shaoming Guo, Jonathan Hickman, and Andreas Seeger, Sharp Lp bounds for
the helical maximal function, Am. J. Math., to appear. Preprint: arxiv.org/abs/2102.08272.
[3] , Sobolev improving for averages over curves in R4 , Adv. Math. 393 (2021), Paper
No. 108089, 85. MR 4340226
[4] David Beltran, Joris Roos, and Andreas Seeger, Multi-scale sparse domination, To appear in
Mem. Amer. Math. Soc., arXiv:2009.000227, 2020.
[5] Jonathan Bennett and Neal Bez, Some nonlinear brascamp–lieb inequalities and applications
to harmonic analysis, Journal of Functional Analysis 259 (2010), no. 10, 2520–2556.
[6] Jonathan Bennett, Neal Bez, Taryn C. Flock, and Sanghyuk Lee, Stability of the Brascamp-
Lieb constant and applications, Amer. J. Math. 140 (2018), no. 2, 543–569. MR 3783217
[7] Jonathan Bennett, Anthony Carbery, and Terence Tao, On the multilinear restriction and
Kakeya conjectures, Acta Math. 196 (2006), no. 2, 261–302. MR 2275834
[8] Jöran Bergh and Jörgen Löfström, Interpolation spaces. An introduction, Grundlehren
der Mathematischen Wissenschaften, No. 223, Springer-Verlag, Berlin-New York, 1976.
MR 0482275
[9] Jean Bourgain and Ciprian Demeter, The proof of the l2 decoupling conjecture, Ann. of Math.
(2) 182 (2015), no. 1, 351–389. MR 3374964
[10] Jean Bourgain, Ciprian Demeter, and Larry Guth, Proof of the main conjecture in Vino-
gradov’s mean value theorem for degrees higher than three, Ann. of Math. (2) 184 (2016),
no. 2, 633–682. MR 3548534
[11] Jean Bourgain and Larry Guth, Bounds on oscillatory integral operators based on multilinear
estimates, Geom. Funct. Anal. 21 (2011), no. 6, 1239–1295. MR 2860188
28 D. BELTRAN, J. DUNCAN, AND J. HICKMAN

[12] Aswin Govindan Sheri, On certain geometric maximal functions in harmonic analysis, PhD
Thesis, University of Edinburgh (2023).
[13] Allan Greenleaf, Principal curvature and harmonic analysis, Indiana Univ. Math. J. 30
(1981), no. 4, 519–537. MR 620265
[14] Philip T. Gressman, Shaoming Guo, Lillian B. Pierce, Joris Roos, and Po-Lam Yung, Re-
versing a philosophy: from counting to square functions and decoupling, J. Geom. Anal. 31
(2021), no. 7, 7075–7095. MR 4289255
[15] Shaoming Guo, Zane Kun Li, Po-Lam Yung, and Pavel Zorin-Kranich, A short proof of ℓ2
decoupling for the moment curve, Preprint: arXiv:1912.09798.
[16] Seheon Ham and Sanghyuk Lee, Restriction estimates for space curves with respect to general
measures, Adv. Math. 254 (2014), 251–279. MR 3161099
[17] Lars Hörmander, Estimates for translation invariant operators in Lp spaces, Acta Math. 104
(1960), 93–140. MR 121655
[18] Hyerim Ko, Sanghyuk Lee, and Sewook Oh, Maximal estimates for averages over space
curves, Invent. Math. 228 (2022), no. 2, 991–1035. MR 4411734
[19] , Sharp smoothing properties of averages over curves, Forum Math. Pi 11 (2023),
Paper No. e4, 33. MR 4549710
[20] Gerd Mockenhaupt, Andreas Seeger, and Christopher D. Sogge, Wave front sets, local
smoothing and Bourgain’s circular maximal theorem, Ann. of Math. (2) 136 (1992), no. 1,
207–218. MR 1173929
[21] Malabika Pramanik and Andreas Seeger, Lp regularity of averages over curves and bounds
for associated maximal operators, Amer. J. Math. 129 (2007), no. 1, 61–103. MR 2288738
[22] Christopher D. Sogge, Fourier integrals in classical analysis, second ed., Cambridge Tracts
in Mathematics, vol. 210, Cambridge University Press, Cambridge, 2017. MR 3645429
[23] Elias M. Stein, Harmonic analysis: real-variable methods, orthogonality, and oscillatory
integrals, Princeton Mathematical Series, vol. 43, Princeton University Press, Princeton, NJ,
1993, With the assistance of Timothy S. Murphy, Monographs in Harmonic Analysis, III.
MR 1232192
[24] Ruixiang Zhang, The endpoint perturbed Brascamp-Lieb inequalities with examples, Anal.
PDE 11 (2018), no. 3, 555–581. MR 3738255
[25] Pavel Zorin-Kranich, Kakeya-Brascamp-Lieb inequalities, Collect. Math. 71 (2020), no. 3,
471–492. MR 4129538

David Beltran: Departament d’Anàlisi Matemàtica, Universitat de València, Dr.

Moliner 50, 46100 Burjassot, Spain
Email address: david.beltran@uv.es

Jennifer Duncan: ICMAT, C. de Nicolas Cabrera 13-15, 28049 Madrid, Spain

Email address: jennifer.duncan@icmat.es

Jonathan Hickman: School of Mathematics, James Clerk Maxwell Building, The

King’s Buildings, Peter Guthrie Tait Road, Edinburgh, EH9 3FD, UK.
Email address: jonathan.hickman@ed.ac.uk