A continuum of Small-cap decouplings and Exponential Sums for the Moment Curve in ℝ 4 \mathbb{R}^{4}
Jacob Glidewell · 2026
Read A continuum of Small-cap decouplings and Exponential Sums for the Moment Curve in ℝ 4 \mathbb{R}^{4} by Jacob Glidewell (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-27065.txt.
Introduction
Glidewell uses the high-low method and wave-packet pruning to prove new small-cap split bounds for the moment curve in 4D. As a use, he proves a guess of Demeter on square-root cancellation of exponential sums tied to the moment curve in 4D. The result gives a range of square-root bounds. It links the Vinogradov mean value theorem in 3D with a result of Bourgain tied to beating the best-known estimate for the Lindelöf hypothesis.
Catalog record
| CREATOR | Jacob Glidewell |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 55150 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.27065
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A continuum of Small-cap decouplings and Exponential Sums for the Moment Curve in ℝ 4 \mathbb{R}^{4} Jacob Glidewell Abstract . We use the high-low method and wavepacket pruning to prove new small-cap decoupling estimates for the moment curve in ℝ 4 \mathbb{R}^{4} . As an application, we verify a conjecture of Demeter regarding the L 12 L^{12} square-root cancellation of exponential sums associated with the moment curve in ℝ 4 \mathbb{R}^{4} . This provides a continuum of square-root cancellation estimates that connects the Vinogradov MVT in ℝ 3 \mathbb{R}^{3} with a result of Bourgain, related to improving the best known estimate for the Lindelöf hypothesis. 1. Introduction We are motivated to attack the following conjecture of Demeter [ 6 , Conjecture 8.1] for the case of the moment curve \Phi(t)=(t,t^{2},t^{3},t^{4}) . The p=12 case is given in [ 6 , Conjecture 1.1] . We use the notation e(x)=e^{2\pi ix} for x\in\mathbb{R} . Conjecture 1.1 . Let 11\leq p\leq 12 . Assume a\geq b\geq 0 and a+b=\frac{p}{2}-3 . Let \Omega=[0,N]\times[0,N^{2}]\times[0,N^{a}]\times[0,N^{b}] . Then, (1) \int_{\Omega}\left|\sum_{n\sim N}e(x\cdot\Phi\left(\frac{n}{N}\right))\right|^{p}dx\lessapprox N^{\frac{p}{2}}|\Omega|. Some special cases of this are known. Notably, the case (p,a,b)=(12,3,0) corresponds to the Vinogradov Mean Value Theorem in \mathbb{R}^{3} ( [ 10 ] , [ 1 ] ). Indeed, using periodicity and VMVT, we get the estimate for each x_{4}\in[0,1] , \frac{1}{N^{3}}\int_{[0,N^{3}]^{3}}\left|\sum_{n\sim N}e(x\cdot\Phi\left(\frac{n}{N}\right))\right|^{12}\;dx_{1}dx_{2}dx_{3}\lessapprox N^{6}\cdot N^{6}. Integrating in x_{4} gives ( 1 ) in this case. The case (p,a,b)=(12,2,1) was shown by Bourgain [ 3 ] . In that paper, the following variant of ( 1 ) served as the main ingredient to his improvement on the Lindelöf hypothesis: (2) \int_{\Omega}\left|\sum_{n\sim N}e(x\cdot\tilde{\Phi}\left(\frac{n}{N}\right))\right|^{12}dx\lessapprox N^{6}|\Omega| with \tilde{\Phi}(t)=(t,t^{2},t^{3/2},t^{1/2}) . In particular, Bourgain showed ( 2 ) for any curve \tilde{\Phi}(t)=(t,t^{2},\phi_{3}(t),\phi_{4}(t)) satisfying some nondegeneracy and torsion conditions. In [ 6 ] , Demeter conjectured a continuum of estimates connecting these two endpoints, and proved the related cases where p=12 and a\in[\frac{3}{2},2] . All of these results are part of a vast program in Fourier analysis, initiated in [ 4 ] , called small-cap decouplings which have become useful for proving exponential sum estimates on spatial truncations of the torus. To the author’s knowledge, no prior estimates towards Conjecture 1.1 are known for the case when p<12 and b>0 . Bourgain’s original approach in [ 3 ] relies on a geometric observation that a bilinear version of ( 2 ) reduces matters to quadratic exponential sums in the plane. To show this bilinear version, he iterates L^{6} decouplings with a "magic" change of variables. One key feature of his argument uses that the domain [0,N] for x_{4} matches the period of x_{1} . His change of variables miraculously exploits this fact to move "volume" from x_{4} to x_{1} . See [ 6 , Section 3] for a discussion of this argument. The method in [ 6 ] generalizes Bourgain’s approach. By combining L^{6} decoupling for the parabola with Gauss sum estimates, Demeter was able to prove the sharp estimates in the regime a\in[\frac{3}{2},2] . However, in both versions of the argument, for the cases a>2 , this mechanism becomes inefficient due to the loss of genuine 4-dimensional behavior at the global scale. A direct manifestation of this is that the domain in x_{4} is now shorter than the period of x_{1} . Rather than clever manipulations of existing decouplings, we will progress Conjecture 1.1 by new small-cap decouplings. The new technology we use for this problem is the small-cap decoupling for the cone, developed by Guth and Maldague [ 7 ] , [ 8 ] . Let \beta\in[\frac{1}{3},\frac{1}{2}] and \alpha\in(0,\beta] . Define \mathcal{M}^{4}(R^{\beta},R)=\big\{(\xi_{1},\xi_{2},\xi_{3},\xi_{4})\;:\xi_{1}\in[\frac{1}{2},1],|\xi_{2}-\xi_{1}^{2}|\leq R^{-2\beta},\\ |\xi_{3}-3\xi_{1}\xi_{2}+2\xi_{1}^{3}|\leq R^{-1},|\xi_{4}+4\xi_{1}\xi_{3}-9\xi_{2}^{2}+4\xi_{1}^{4}|\leq R^{-1}\big\}. We define the small-caps \gamma as the partition \mathcal{M}^{4}(R^{\beta},R)\cap\{\ell R^{-\beta}\leq\xi_{1}\leq(\ell+1)R^{-\beta}\} for each integer \ell\in R^{\beta}[\frac{1}{2},1] . We will show the following generalized square-root cancellation phenomenon. Given F:\mathbb{R}^{4}\to\mathbb{C} , we define F_{\gamma} as the Fourier projection onto \gamma . For clarity, we call a S_{1}\times S_{2}\times S_{3}\times S_{4} box in \mathbb{R}^{4} regular if the side with length S_{i} is parallel to the x_{i} -axis. Theorem 1.2 . For any F:\mathbb{R}^{4}\to\mathbb{C} Schwartz with Fourier Transform supported in \mathcal{M}^{4}(R^{\beta},R)=\bigcup_{\gamma}\gamma and any R\times R\times R\times R^{\alpha} regular box \Omega_{R} , if \|F_{\gamma}\|_{\infty}\leq 1 , we have \|F\|_{L^{p}(\Omega_{R})}^{p}\lessapprox R^{\frac{p}{2}\beta}|\Omega_{R}|, with p=6+\frac{2(\alpha+1)}{\beta}, for \alpha and \beta in the range corresponding to p\geq 11 . \beta \alpha 1/2 2/5 1/3 1/3 1/2 1/4 0 Figure 1. Parameters \alpha and \beta . In the blue region, the critical exponent is p\in(12,14) . In the red, p\in(11,12) . In the orange, p<11 . We call this phenomenon "square-root cancellation" because, on average, F is comparable to R^{\beta/2} . This is seen easily with the L^{2} -moment: \|F\|_{L^{2}(\Omega_{R})}\lesssim R^{\beta/2}|\Omega_{R}|^{1/2} . We compare this to the triangle inequality, which gives |F(x)|\leq\sum_{\gamma}|F_{\gamma}(x)|\lesssim R^{\beta}. When |F(x)|\sim R^{\beta} , we call this "constructive interference." It is easy to see that for exponential sums, such as ( 1 ) or in Corollary 1.4 , constructive interference occurs in an O(1) -neighborhood of x=0 . The critical exponent p is the L^{p} -moment where the competing behavior of square-root cancellation and constructive interference match. For larger moments, constructive interference dominates while for smaller moments, square-root cancellation dominates. Observe that by Hölder’s inequality, Theorem 1.2 is true more generally for 2\leq p\leq 6+\frac{2(\alpha+1)}{\beta}, so for the rest of the paper we will only consider p=6+\frac{2(\alpha+1)}{\beta}. Notice p\in(10,14] for the range of parameters; see Figure 1 . When 10<p<11 , we expect an analogous 4 -linear estimate; however, our methods only show a 4 -broad version. A counterexample for the linear estimate when p<11 can be found in [ 6 , Theorem 8.2] . Theorem 1.3 . For any F:\mathbb{R}^{4}\to\mathbb{C} Schwartz with Fourier Transform supported in \mathcal{M}^{4}(R^{\beta},R)=\bigcup_{\gamma}\gamma , any R\times R\times R\times R^{\alpha} regular box \Omega_{R} , and any caps \tau_{1},\tau_{2},\tau_{3},\tau_{4} with |\tau_{i}|\sim 1 and \sim 1 -separated, if \|F_{\gamma}\|_{\infty}\leq 1 , we have \|\min_{1\leq i\leq 4}|F_{\tau_{i}}|\|_{L^{p}(\Omega_{R})}^{p}\lessapprox R^{\frac{p}{2}\beta}|\Omega_{R}|, with p=6+\frac{2(\alpha+1)}{\beta}, for the parameters \beta\in[\frac{1}{3},\frac{1}{2}] and 0<\alpha\leq\beta . As a corollary of Theorem 1.2 , we verify Conjecture 1.1 in the remaining range for p=12 , make progress for p\in[11,12) , and give new exponential sum estimates for p\in(12,14] . These estimates with c_{n}=1 and constructive interference on [0,\frac{1}{100}]^{4} show that the critical exponent p in Theorem 1.2 is sharp. Corollary 1.4 . Let 11\leq p\leq 14 . Assume a\in[2,3] , b\in(0,1] and a+b=\frac{p}{2}-3 . Let \Omega=[0,N]\times[0,N^{2}]\times[0,N^{a}]\times[0,N^{b}] . Then, for any |c_{n}|\leq 1 , \int_{\Omega}\left|\sum_{n\sim N}c_{n}e(x\cdot\Phi\left(\frac{n}{N}\right))\right|^{p}dx\lessapprox N^{\frac{p}{2}}|\Omega|. Therefore, the remaining open range in Conjecture 1.1 is p\in[11,12) with \frac{p-6}{4}\leq a<2 . This is analogous to the range for p=12 which Demeter proved in [ 6 ] , which our methods seem to not be able to address. These correspond to decouplings with parameters \alpha>\beta in the language of Theorem 1.2 . To establish Theorem 1.3 , we use a high-low stopping time argument with wavepacket pruning, combining the approaches of Guth and Maldague [ 8 ] and Maldague and Oh [ 9 ] . By a standard broad-narrow argument, we get the linear inequalities in Theorem 1.2 . The main novelity of this paper is the integration and simplification of the arguments from [ 8 ] and [ 9 ] . Further, we elaborate on the lack of linear inequalities for p<11 . One obstruction that we see explicitly is that the