Lp-Lq estimates for the circular maximal operator on Heisenberg radial functions
L p boundedness of the circular maximal function M H 1 on the Heisenberg group H 1 has received considerable attentions. While the problem still remains open, L p boundedness of M H 1 on Heisenberg radial functions was recently shown for p > 2 by Beltran et al. (Ann Sc Norm Super Pisa Cl Sci. https://doi.org/10.2422/2036-2145.202001-006, 2021). In this paper we extend their result considering the local maximal operator M H 1 which is defined by taking supremum over 1 < t < 2 . We prove L p – L q estimates for M H 1 on Heisenberg radial functions on the optimal range of p, q modulo the borderline cases. Our argument also provides a simpler proof of the aforementioned result due to Beltran et al.
Keywords: 42B25; 22E25; 35S30
introduction
For the spherical maximal function is given by
Graph
where is the -dimensional sphere centered at the origin and is the surface measure on . When , it was shown by Stein [[21]] that is bounded on if and only if . The case was later settled by Bourgain [[5]]. An alternative proof of Bourgain's result was subsequently found by Mockenhaupt, Seeger, Sogge [[11]], who used a local smoothing estimate for the wave operator. We now consider the local maximal operator
Graph
As is easy to see, the maximal operator can not be bounded from to unless . However, is bounded from to for some thanks to the supremum taken over the restricted range [1, 2]. This phenomenon is called improving. Almost complete characterization of improving property of was obtained by Schlag [[17]] except for the endpoint cases. A different proof which is based on – smoothing estimate for the wave operator was also found by Schlag and Sogge [[18]]. They also proved – boundedness of for which is optimal up to the borderline cases. Most of the left open endpoint cases were settled by the second author [[8]] but there are some endpoint cases where – estimate remains unknown though restricted weak type bounds are available for such cases. There are results which extend the aforementioned results to variable coefficient settings, see [[18]]. Also, see [[1], [4], [14]] and references therein for recent extensions of the earlier results.
The analogous spherical maximal operators on the Heisenberg group also have attracted considerable interests. The Heisenberg group can be identified with under the noncommutative multiplication law
Graph
where and A is the matrix given by
Graph
The natural dilation structure on is . Abusing the notation, since there is no ambiguity, we denote by the usual surface measure of . Then, the dilation of the measure is defined by . Thus, the average over the sphere is now given by
Graph
We consider the associated local spherical maximal operator
Graph
Similarly, the global maximal operator is defined by taking supremum over . As in the Euclidean case, boundedness of is essentially equivalent to that of (for example, see [[2]] or Section 2.5). The spherical maximal operator on was first studied by Nevo and Thangavelu [[13]]. It is easy to see that is bounded on only if by using Stein's example ([[21]]):
Graph
where is a cutoff function supported near the origin. For , boundedness of on the optimal range was independently proved by Müller and Seeger [[10]], and by Narayanan and Thangavelu [[12]]. Furthermore, for , Roos, Seeger and Srivastava[[15]] recently obtained the complete – estimate for except for some endpoint cases. Also see [[7]] for related results.
However, the problem still remains open when .
Definition
We say a function is Heisenberg radial if for all .
Beltran, Guo, Hickman and Seeger [[2]] obtained boundedness of on the Heisenberg radial functions for . In the perspective of the results concerning the local maximal operators [[8], [15], [17]], it is natural to consider – estimate for . The main result of this paper is the following which completely characterizes improving property of on Heisenberg radial function except for some borderline cases.
Theorem 1.1
Let and , and let be the closed region bounded by the triangle . Suppose . Then, the estimate
1.1
Graph
holds for all Heisenberg radial function f. Conversely, if , then the estimate fails.
Though the Heisenberg radial assumption significantly simplifies the structure of the averaging operator, the associated defining function of the averaging operator is still lacking of curvature properties. In fact, the defining function has vanishing rotational and cinematic curvatures at some points, see [[2]] for a detailed discussion. This increases the complexity of the problem. To overcome the issue of vanishing curvatures, Beltran et al. [[2]] used the oscillatory integral operators with two-sided fold singularities and the variable coefficient version of local smoothing estimate [[3]] combined with additional localization.
The approach in this paper is quite different from that in [[2]]. Capitalizing on the Heisenberg radial assumption, we make a change of variables so that the averaging operator on the Heisenberg radial function takes a form close to the circular average, see (2.1) below. While the defining function of the consequent operator still does not have nonvanishing rotational and cinematic curvatures, via a further change of variables we can apply the – local smoothing estimate (see, Theorem 3.1 below) in a more straightforward manner by exploiting the apparent connection to the wave operator (see (2.2) and (2.3)). Consequently, our approach also provides a simplified proof of the recent result due to Beltran et al. [[2]]. See Sect. 2.5.
Even though we utilize the local smoothing estimate, we do not need to use the full strength of the local smoothing estimate in since we only need the sharp – local smoothing estimates for (p, q) near (7/3, 7/2). Such estimates can also be obtained by interpolation and scaling argument if one uses the trilinear restriction estimates for the cone and the sharp local smoothing estimate for some large p (for example, see [[9]]).
The estimate (1.1) remains open when . However, we expect that those borderline cases should be true. Most of the corresponding endpoint estimates for the circular maximal function (in ) are known to be true [[8]], but to implement the approach in [[8]] we need the local smoothing estimate without -loss regularity, which we are not able to establish yet even under the Heisenberg radial assumption.
We close the introduction showing the necessity part of Theorem 1.1.
Optimality of
p, q range. We show (1.1) implies , that is to say,
Graph
To see , let be the characteristic function of a ball of radius , centered at 0. Then, is also supported in a ball B of radius and on B. Thus, is finite only if . For let be the characteristic function of a ball of radius centered at 0. Then, when is contained in a neighborhood of for a small constant . Thus, (1.1) implies , which gives if we let . Finally, to show we consider which is the characteristic function of an neighborhood of with . Then, when is in an ball centered at 0. Thus, (1.1) gives , which yields .
The maximal estimate (1.1) for general functions has a smaller range of p, q. Let be a characteristic function of the set for a sufficiently small . Then if for a small constant independent of r. Thus, (1.1) implies . It seems to be plausible to conjecture that (1.1) holds for general f modulo some endpoint cases as long as , , and . The range of p, q is properly contained in .
Proof of Theorem 1.1
In this section we prove Theorem 1.1 while assuming Proposition 2.1 and Proposition 2.2 (see below), which we show in the next section.
Heisenberg radial function
Since f is a Heisenberg radial function, we have for some . Let us set
Graph
Then, it follows . Since , we have
2.1
Graph
Let us define an operator by
2.2
Graph
Using Fourier inversion, we have
2.3
Graph
Since is also Heisenberg radial,[1] A computation shows . Therefore, we see that the estimate (1.1) is equivalent to
2.4
Graph
In what follows we show (2.4) holds for p, q satisfying
2.5
Graph
Then, interpolation with the trivial estimate proves Theorem 1.1.
Decomposition
Let denote a positive smooth function on supported in such that for . We set . To show (2.4) we decompose as follows:
Graph
We break g via the Littlewood–Paley decomposition and try to obtain estimates for each decomposed pieces. For the purpose we denote and and define the projection operators
Graph
Our proof of (2.4) mainly relies on the following two propositions, which we prove in Sect. 3.
Proposition 2.1
Let and . Suppose
2.6
Graph
Then, for we have
2.7
Graph
The estimate (2.7) continues to be valid for the case . However, the range of p, q for which (2.7) holds gets smaller.
Proposition 2.2
Let and . Suppose , and . Then, for we have
Graph
We frequently use the following elementary lemma (for example, see [[8]]) which plays the role of the Sobolev imbedding.
Lemma 2.3
Let I be an interval and let F be a smooth function defined on . Then, for ,
Graph
Proof of (2.4)
We prove (2.4) handling the three cases and , separately. We first consider a change of variables
2.8
Graph
which plays an important role in what follows. Note that
2.9
Graph
In order to show (2.4), we shall use the change of variables (2.8) to apply the local smoothing estimate to the averaging operator (see Sect. 3.1). Since , for . Thus, the cases can be handled directly by using local smoothing estimates for the half wave propagator. However, the determinant of the Jacobian may vanish when . This requires further decomposition away from the set . See Sect. 3.3. This is why we need to consider the three cases separately.
Let us set and so that . Then, we break
2.10
Graph
We use Propositions 2.1 and 2.2 to obtain the estimate for , whereas we show the estimate for by elementary means using (2.2).
Case k≤-2
We claim that
2.11
Graph
holds provided that p, q satisfy , , and (2.6). Thus (2.11) holds for p, q satisfying (2.5).
We first consider . We shall show that
2.12
Graph
holds for . We recall (2.2) and note that is uniformly bounded because . Since and , we have by the Mikhlin multiplier theorem. Applying Lemma 2.3 to , we see that (2.12) follows if we show
2.13
Graph
We now make use of the change of variables (2.8). Since and , we have . Thus the left hand side of (2.13) is bounded by
Graph
Changing variables and gives
Graph
where . Since and is a smooth function supported in the set , is a smooth multiplier whose derivatives are uniformly bounded. So, the multiplier operator given by is uniformly bounded from to for . Thus, via scaling we obtain (2.13) and, hence, (2.12).
Using the triangle inequality and (2.12), we have
Graph
because . We now consider for which we use Proposition 2.1. Since
Graph
and since p, q satisfy , , and (2.6), using the estimate (2.7), we get
Graph
Combining this with the above estimate for gives (2.11) and this proves the claim.
Case k≥2
In this case we show
2.14
Graph
if , , and (2.6) holds. So, we have (2.14) if (2.5) holds.
In order to prove (2.14) we first prove the following.
Lemma 2.4
Let . If and , then
2.15
Graph
where
Proof
We note that
Graph
where
Graph
We note since . Thus, changing variables , by integration by parts we have for any . Since and , we see . Therefore, we get (2.15).
Proof of 2.14
We begin by observing a localization property of the operator . From (2.1) we note that
Graph
for if k is large enough, i.e., . Thus, from (2.1) and (2.3) we see that
2.16
Graph
where . Clearly, the intervals are finitely overlapping and so are the supports of . Since , by a standard localization argument it is sufficient for (2.14) to show
2.17
Graph
for .
Using the decomposition (2.10), we first consider . Changing variables , we have
Graph
Since , , and , by Lemma 2.4 . Hence,
Graph
The second inequality follows by Young's convolution inequality and the third is clear because and . We now handle . Since
2.18
Graph
and since , , and (2.6) holds, using the estimate (2.7), we get
Graph
Therefore, we get (2.17).
Case |k|≤1
To complete the proof of (2.4), the matter is now reduced to obtaining
Graph
if p, q satisfy (2.5). In order to show this we use Proposition 2.2. Using the decomposition (2.10), we first consider . Since and , by Lemma 2.4 we have . Hence, it follows that
Graph
for .
We now consider . Note that (2.6) is satisfied if (2.5) holds. Since , by (2.18) and Proposition 2.2 we see
Graph
taking a small enough . Therefore we get the desired estimate.
Global maximal estimate
Using the estimates in this section, one can provide a simpler proof of the result due to Beltran et al. [[2]], i.e.,
2.19
Graph
for . In order to show this we use the following lemma which is a consequence of Propositions 2.1 and 2.2.
Lemma 2.5
Let . Then, for some we have
2.20
Graph
Proof
We briefly explain how one can show (2.20). In fact, similarly as before, we decompose
Graph
where
Graph
and Then, the estimate (2.20) follows if we show , for some . The estimate for follows from (2.12) and summation over . Using the estimate of the second case in (2.7), one can easily get the estimate for . The estimate for is obvious from Proposition 2.2. By Proposition 2.1 combined with the localization property (2.16) we can obtain the estimate for . However, due to the projection operator we need to modify the previous argument slightly.
From (2.1) and (2.3) we see
2.21
Graph
where . Note that for any N and . If , , and k is large enough, then we have
Graph
for any N since and . Hence it follows that
Graph
for any N. We break . Using the last inequality and then Proposition 2.1, we obtain
Graph
for some by taking an N large enough.
Once we have (2.20), using a standard argument which relies on the Littlewood–Paley decomposition and rescaling (for example, see [[2], [5], [16]]) one can easily show (2.19). Indeed, we break the maximal function into high and lower frequency parts:
Graph
where
Graph
For we claim
2.22
Graph
This gives . Since is bounded on for , for we get
Graph
We now proceed to prove (2.22). Note that and is a smooth function supported on . Thus, similarly as in (2.21) we note that where . Since for any N, for we see
2.23
Graph
because and Hence, taking an N large enough, we note that
2.24
Graph
provided that . Indeed, to show this we only have to consider the case since the other case is trivial. By scaling we may assume that . Thus, it is enough to show for with an N large enough. However, this is easy to see since and .
Therefore, combining (2.23) and (2.24), one can see
Graph
Here denotes the Hardy-Littlewood maximal function on . This proves the claim (2.22) since .
So we are reduced to showing for . For the purpose it is sufficient to show
2.25
Graph
because and By scaling, using (2.2), we can easily see the inequality (2.25) is equivalent to (2.20) while j replaced by k. So, we have (2.25) and this completes the proof of (2.19).
Proof of Propositions 2.1 and 2.2
In order to prove Propositions 2.1 and 2.2, we are led by (2.2) to consider for which we use the following well known asymptotic expansion (see, for example, [[20]]):
3.1
Graph
where is a smooth function satisfying
3.2
Graph
for if . The expansion (3.1) relates the operator to the wave propagator. After changing variables, to prove Propositions 2.1 and 2.2 we can use the local smoothing estimate for the wave operator (see Proposition 3.1 below).
Local smoothing estimate
Let us denote
Graph
We make use of – local smoothing estimate for the wave equation in .
Theorem 3.1
Let . Suppose (2.6) holds. Then, for we have
3.3
Graph
This follows by interpolating the estimates (3.3) with , , and (4, 4). The estimate (3.3) with is a straightforward consequence of Plancherel's theorem and (3.3) with can be shown by the stationary phase method (for example, see [[8]]). The case is due to Guth et al. [[6]].
From Theorem 3.1 we can deduce the following estimate via simple rescaling argument.
Corollary 3.2
Let . Suppose (2.6) holds. Then, for we have
Graph
Proof
Changing variables , we see
Graph
Thus, using (3.3) we have
Graph
So, rescaling gives the desired inequality.
Proof of Proposition 2.1
We now recall (2.2) and (3.1). To show Proposition 2.1 we first deal with the contribution from the error part . Let us set
Graph
Lemma 3.3
Let . Suppose (2.6) holds. Then, we have
3.4
Graph
Proof
We first consider the case . Using Lemma 2.3, we need to estimate and in . For simplicity we denote . We first consider . Changing variables , we note that
Graph
where
Graph
Since , using (3.2), we have for via integration by parts. Thus, we have for with a positive constant C. Young's convolution inequality gives . Thus, reversing , after a simple manipulation we get
3.5
Graph
for Indeed, we need only note that because and .
We now consider . Note that
3.6
Graph
Using (3.2), we can handle similarly as before. In fact, since and , we see
Graph
Hence, combining this and (3.5) with Lemma 2.3, we get (3.4) for .
We now consider the case . We first claim that
3.7
Graph
We use the transformation (2.8). By (2.9) we have . Therefore,
Graph
where
Graph
Note that . Changing and , using (3.2) and integration by parts, we have for and . Young's convolution inequality gives
Graph
Thus, we get (3.7). As for , we use (3.6) and repeat the same argument to see since , , and . Thus, we get
Graph
Putting (3.7) and this together, by Lemma 2.3 we obtain (3.4) for .
By (3.1) and Lemma 3.3, to prove Propositions 2.1 and 2.2 we only have to consider contributions from the remaining . To this end, it is sufficient to consider the major term since the other terms can be handled similarly. Furthermore, by reflection it is enough to deal with since the estimate (3.3) clearly holds with the interval [1, 2] replaced by .
Let us set
3.8
Graph
To complete the proof of Proposition 2.1, we need to show
3.9
Graph
Using Lemma 2.3, the matter is reduced to obtaining estimates for and in . Note that
3.10
Graph
By the Mikhlin multiplier theorem one can easily see
Graph
where denotes . Therefore, by Lemma 2.3 it is sufficient for (3.9) to prove that
Graph
We first consider the case . As before, we use the change of variables (2.8). Since from (2.9) and since and , we have
Graph
since . Thus, Corollary 3.2 gives the desired estimate (3.9) for . The case can be handled in the exactly same manner. The only difference is that . Thus, the desired estimate (3.9) immediately follows from Corollary 3.2.
Proof of Proposition 2.2
As mentioned already, the determinant of the Jacobian may vanish when . So, we need additional decomposition depending on . We also make decomposition in depending on to control the size of the multiplier in a more accurate manner (for example, see (3.22)).
For let us set
Graph
so that . We additionally define
Graph
So it follows that
3.11
Graph
Proposition 3.4
Let us set Let and . Suppose (2.6) holds. Then, for we have
3.12
Graph
In order to prove Proposition 3.4, we make the change of variables (2.8). Since , we need only to consider (r, t) contained in the set . Set
Graph
By (2.8) . From (2.9) we note if . Thus, changing variables we obtain
3.13
Graph
Therefore, for (3.12) it is sufficient to show
3.14
Graph
for p, q satisfying (2.6). For the purpose we need the following lemma, which gives an improved estimate thanks to restriction of the integral over . Indeed, one can remove the localization .
Lemma 3.5
Let . Then, we have
3.15
Graph
Proof
We write Then, changing variables and we see
Graph
where and By Plancherel's theorem in the variable and integrating in t, we have
Graph
A computation shows , so on the support of . Thus, by changing variables and Plancherel's theorem we get (3.15).
We also use the following elementary lemma.
Lemma 3.6
For any , j, and m, we have
Graph
Proof
Since , it suffices to prove the second inequality only. By Young's inequality we need only to show By scaling it is clear that Note that is supported in a rectangular box with dimensions . So, is supported in a cube of side length and it is easy to see is uniformly bounded for any . This gives . Therefore, after scaling we get
Proof of 3.14
In view of interpolation the estimate (3.14) follows for p, q satisfying (2.6) if we show the next three estimates:
3.16
Graph
3.17
Graph
The first estimate follows from Lemma 3.5. Corollary 3.2 and Lemma 3.6 give the other two estimates.
It is possible to improve the estimate (3.12) when .
Proposition 3.7
Let and . Suppose , , and , then
Graph
Proof
By (3.13) it is sufficient to show
Graph
for p, q satisfying , . In fact, by interpolation with the estimates (3.16) and (3.17) we only have to show
3.18
Graph
Let us set
Graph
Then Therefore, (3.18) follows if we show
3.19
Graph
when . Note that if . So, is contained in a conic sector with angle . Let be a sector centered at the origin in with angle and be a cut-off function adapted to . Then, by integration by parts it follows that
Graph
if . (See, for example, [[8]]). Now (3.19) is clear since the support of can be decomposed into as many as such sectors.
Finally, we prove Proposition 2.2 making use of Propositions 3.4 and 3.7. We recall (2.2) and (3.1). As mentioned before, by Lemma 3.3 we need only to consider (see (3.8)) and it is sufficient to show
3.20
Graph
for p, q satisfying , and .
Proof of 3.20
Let us set and . Then, we decompose
Graph
Combining this with (3.11) and using , by the triangle inequality we have
Graph
where
Graph
The proof of (3.20) is now reduced to showing
3.21
Graph
for p, q satisfying , and .
Before we start the proof of (3.21), we briefly comment on the decomposition , . As for and , which are easier to handle, the sizes of and are sufficiently small on the supports of the associated multipliers, so we can remove the dependence of t by an elementary argument. For and , we use Lemma 2.3 combined with (3.10) to control the maximal operators. Different magnitudes of contribution come from and , so we need to compare them. Writing , we note
3.22
Graph
The decompositions in and are made according to comparative sizes of and in terms of l, m, and j.
We first consider . Using Lemma 2.3, we need to estimate and in . Note that and . Thus, recalling (3.10), we apply Lemma 2.3 and the Mikhlin multiplier theorem to get
Graph
Thus, by Proposition 3.4 it follows that
Graph
Since and , we obtain (3.21) with .
We can show the estimate (3.21) with in the same manner. As before, since and , using (3.22), Lemma 2.3, and the Mikhlin multiplier theorem, we have
Graph
Thus, by (3.13) and Theorem 3.1, we have which gives (3.21) with .
We now consider , which we handle as before. Since , if . Similarly, and if . Using (3.22) and (3.10), we see
Graph
Since , using Proposition 3.7, we get (3.21) for .
We handle and in an elementary way without relying on Lemma 2.3. Instead, we can control and more directly. Concerning we claim that
3.23
Graph
if and . This clearly gives (3.21) with for p, q satisfying , and . We note that
Graph
where
Graph
and is a smooth function supported in such that . If , then . Thus, for any . We remove the dependence of t by using a bound on the coefficient of Fourier series, not the Sobolev embedding. Expanding into Fourier series on we have while . Since , the estimate (3.23) follows after scaling if we obtain
Graph
where
Graph
When , changing variables and following the argument in the proof of Lemma 3.5 we have On the other hand, (3.18) gives Interpolation between these two estimates gives
Graph
for . Since the support is contained in a rectangular region of dimensions , by Bernstein's inequality we have
Graph
for . Since , this proves the claimed estimate (3.23).
Finally, we show (3.21) with . Changing variables , we observe
Graph
where
Graph
Note that . Since for any , expanding into Fourier series on we have while . Hence, similarly as before, changing variables , to show (3.21) with it is sufficient to obtain
3.24
Graph
for . Clearly, the left hand side is bounded by . The Fourier transform of is supported on the rectangle . Thus, using Bernstein's inequality in , we get
Graph
for . Another use of Bernstein's inequality gives (3.24) for . This completes the proof of (3.20).
Acknowledgements
Juyoung Lee was supported by the National Research Foundation of Korea (NRF) grant no. 2017H1A2A1043158 and Sanghyuk Lee was supported by NRF grant no. 2021R1A2B5B02001786.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
[
References
1
Anderson TC, Hughes K, Roos J, Seeger A. - bounds for spherical maximal operators. Math. Z. 2021; 297: 1057-1074. 4229592. 10.1007/s00209-020-02546-0. 1461.42012
2
Beltran, D, Guo, S, Hickman, J, Seeger, A: The circular maximal operator on Heisenberg radial functions. Ann. Sc. Norm. Super. Pisa Cl. Sci. (2021). https://doi.org/10.2422/2036-2145.202001-006
3
Beltran D, Hickman J, Sogge CD. Variable coefficient Wolff-type inequalities and sharp local smoothing estimates for wave equations on manifolds. Anal. PDE. 2020; 13: 403-433. 4078231. 10.2140/apde.2020.13.403. 1436.35340
4
Beltran, D, Oberlin, R, Roncal, L, Seeger, A, Stovall, B: Variation bounds for spherical averages. Math. Ann. (2021). https://doi.org/10.1007/s00208-021-02218-2
5
Bourgain J. Averages in the plane over convex curves and maximal operators. J. Anal. Math. 1986; 47: 69-85. 874045. 10.1007/BF02792533. 0626.42012
6
Guth L, Wang H, Zhang R. A sharp square function estimate for the cone in. Ann. Math. 2020; 192: 551-581. 4151084. 10.4007/annals.2020.192.2.6. 1450.35156
7
Kim, J: Annulus maximal averages on variable hyperplanes. arXiv:1906.03797
8
Lee S. Endpoint estimates for the circular maximal function. Proc. Am. Math. Soc. 2003; 131: 1433-1442. 1949873. 10.1090/S0002-9939-02-06781-3. 1042.42007
9
Lee S, Vargas A. On the cone multiplier in. J. Funct. Anal. 2012; 263: 925-940. 2927399. 10.1016/j.jfa.2012.05.010. 1252.42014
Müller D, Seeger A. Singular spherical maximal operators on a class of two step nilpotent Lie groups. Isr. J. Math. 2004; 141: 315-340. 2063040. 10.1007/BF02772226. 1054.22007
Mockenhaupt G, Seeger A, Sogge C. Wave front sets, local smoothing and Bourgain's circular maximal theorem. Ann. Math. 1992; 136: 207-218. 1173929. 10.2307/2946549. 0759.42016
Narayanan E, Thangavelu S. An optimal theorem for the spherical maximal operator on the Heisenberg group. Isr. J. Math. 2004; 144: 211-219. 2121541. 10.1007/BF02916713. 1062.43016
Nevo A, Thangavelu S. Pointwise ergodic theorems for radial averages on the Heisenberg group. Adv. Math. 1997; 127: 307-334. 1448717. 10.1006/aima.1997.1641. 0888.22002
Roos, J, Seeger, A: Spherical maximal functions and fractal dimensions of dilation sets. arXiv:2004.00984
Roos, J, Seeger, A, Srivastava, R: Lebesgue space estimates for spherical maximal functions on Heisenberg groups. Int. Math. Res. Not. (2021). https://doi.org/10.1093/imrn/rnab246
Schlag, W: estimates for the circular maximal function. Ph.D. Thesis, California Institute of Technology (1996)
Schlag W. A generalization of Bourgain's circular maximal theorem. J. Am. Math. Soc. 1997; 10: 103-122. 1388870. 10.1090/S0894-0347-97-00217-8. 0867.42010
Schlag W, Sogge CD. Local smoothing estimates related to the circular maximal theorem. Math. Res. Lett. 1997; 4: 1-15. 1432805. 10.4310/MRL.1997.v4.n1.a1. 0877.42006
Sogge CD. Propagation of singularities and maximal functions in the plane. Invent. Math. 1991; 104: 349-376. 1098614. 10.1007/BF01245080. 0754.35004
Stein EM. Harmonic analysis: real variable methods, orthogonality and oscillatory integrals. 1993: Princeton; Princeton Univ. Press. 0821.42001
Stein EM. Maximal functions: spherical means. Proc. Natl. Acad. Sci. USA. 1976; 73: 2174-2175. 420116. 10.1073/pnas.73.7.2174. 0332.42018
]
[
Footnotes
This is true because is an abelian group. However, is not commutative in general, so the property is not valid in higher dimensions.
]
By Juyoung Lee and Sanghyuk Lee
Reported by Author; Author