Logarithmic oscillatory multipliers and log-subdyadic square functions
Vicente Vergara · 2026
Read Logarithmic oscillatory multipliers and log-subdyadic square functions by Vicente Vergara (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-2605-27746.txt.
Introduction
Vicente Vergara develops square-function bounds for Fourier multipliers whose local oscillation scale sits between dyadic and power-subdyadic scales at high frequency. For symbols meeting a localized Sobolev test, he proves a pointwise square-function bound and a weighted L2 multiplier inequality. Adding a smooth low-frequency part yields unweighted Lp bounds. At p equals 2 the multiplier is bounded for all beta, with a sufficient beta test for other p and Lorentz endpoints at critical equality.
Catalog record
| CREATOR | Vicente Vergara |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 53877 |
| LICENSE | http://creativecommons.org/publicdomain/zero/1.0/ |
| RIGHTS | CC0 |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.27746
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Logarithmic oscillatory multipliers and log-subdyadic square functions Vicente Vergara Abstract We study Fourier multipliers with logarithmic oscillation at high frequency. The guiding example is the radial symbol m γ , β ( ξ ) = ( log ( e + | ξ | ) ) − β e i ( log ( e + | ξ | ) ) γ , γ > 1 , m_{\gamma,\beta}(\xi)=\bigl(\log(e+|\xi|)\bigr)^{-\beta}e^{i(\log(e+|\xi|))^{\gamma}},\qquad\gamma>1, whose natural frequency scale is smaller than dyadic but larger than every fixed power-subdyadic scale. We develop a square-function theory adapted to this logarithmic scale. The main square-function result is a pointwise estimate for Fourier multiplier operators whose symbols satisfy a localized logarithmic Miyachi condition. We prove the corresponding log-subdyadic frequency decomposition, the associated decoupling and recoupling estimates, and the local multiplier estimate needed to control the operator. We also establish a high-frequency weighted L 2 L^{2} multiplier estimate and derive unweighted L p L^{p} -boundedness for 1 < p < ∞ 1<p<\infty under the sufficient logarithmic decay condition β > d ( γ − 1 ) | 1 2 − 1 p | . \beta>d(\gamma-1)\left|\frac{1}{2}-\frac{1}{p}\right|. The logarithmic model multiplier above satisfies the localized hypothesis in the high-frequency region. 2020 Mathematics Subject Classification. Primary 42B15, 42B20; Secondary 42B25, 42B35 Keywords. Fourier multipliers, square functions, logarithmic oscillation, log-subdyadic decompositions, weighted inequalities, Littlewood–Paley theory 1 Introduction The starting point of this paper is the observation that the subdyadic geometry in the square-function theory of Beltran–Bennett [ 1 ] is dictated by a local oscillation scale. This viewpoint is rooted in the earlier work of Miyachi on oscillatory Fourier multipliers [ 5 , 6 ] and in the classical power-phase examples related to Wainger’s work [ 11 ] . For the Miyachi model multiplier m_{\alpha,\beta}(\xi)=|\xi|^{-\beta}e^{i|\xi|^{\alpha}}, the radial phase is \Phi(R)=R^{\alpha} , and the scale on which the phase changes by an amount comparable to one is |\Phi^{\prime}(R)|^{-1}\sim R^{1-\alpha}. This is precisely the radius of the \alpha -subdyadic balls used in the Beltran–Bennett decomposition. This places the argument in the broader Littlewood–Paley tradition of Stein’s square-function method [ 8 , 9 ] . It is also related in spirit to square functions associated with non-dyadic frequency decompositions, such as Rubio de Francia’s square function for arbitrary intervals [ 7 ] , and to square-function and maximal estimates for radial Fourier multipliers [ 4 ] . The logarithmic model considered here belongs to the broader family of oscillatory Fourier multipliers, but it has a different geometric scale from the power-phase examples above. Related contemporary work includes rapidly oscillating angular symbols [ 10 ] , bilinear oscillatory Fourier multipliers [ 3 ] , and oscillating spectral multipliers in non-Euclidean settings [ 2 ] . These results have different geometric hypotheses from those used below; they are cited to place the present logarithmic construction within the broader literature on oscillatory multiplier phenomena. The purpose of the present work is to develop the corresponding logarithmic theory. We fix \Phi(R)=(\log R)^{\gamma},\qquad\gamma>1, for R sufficiently large. The induced frequency scale is \rho(R)=\frac{R}{(\log R)^{\gamma-1}}. (1) The spatial aperture dual to this frequency scale is a_{\gamma}(t)=\rho(t^{-1})^{-1}=t(\log(1/t))^{\gamma-1}. (2) Thus the logarithmic approach region is wider than the classical cone by a logarithmic factor, while remaining much closer to dyadic geometry than the power-subdyadic regimes. The model multiplier is m_{\gamma,\beta}(\xi)=\bigl(\log(e+|\xi|)\bigr)^{-\beta}e^{i(\log(e+|\xi|))^{\gamma}}. (3) The parameter \beta measures logarithmic decay. Since frequencies |\xi|\sim t^{-1} correspond to the logarithmic size \log(1/t) , the output square function carries the compensating factor (\log(1/t))^{2\beta}. This is the logarithmic analogue of the Beltran–Bennett normalization in which g_{\alpha,\beta} carries the factor t^{-2\beta} for symbols with power decay |\xi|^{-\beta} . The logarithmic model considered here is motivated by, but not reducible to, the Miyachi–Beltran–Bennett framework for oscillatory multipliers with power phases, where the adapted frequency scale is dictated by the inverse derivative of the radial phase. Indeed, in the high-frequency region the first radial derivative has size |m^{\prime}_{\gamma,\beta}(R)|\simeq(\log R)^{-\beta+\gamma-1}R^{-1},\qquad R\gg 1, so the model lies outside the classical dyadic Mikhlin regime with logarithmic decay. At the same time, the oscillation scale forced by the phase is |\Phi^{\prime}(R)|^{-1}\sim\frac{R}{(\log R)^{\gamma-1}}, which lies strictly between the dyadic scale R and every fixed power-subdyadic scale R^{1-\alpha} , \alpha>0 , as R\to\infty . Thus the model is too oscillatory for a purely dyadic Littlewood–Paley analysis, while its natural scale is not one of the fixed power-subdyadic scales covered by Beltran–Bennett. The present decomposition is therefore adapted to a genuine logarithmic regime, rather than a formal reformulation of the power-subdyadic theory. The corresponding geometry and multiplier hypothesis are formalized in Definitions 2.2 , 2.3 and Hypothesis 2.6 ; the model verification is given in Proposition 6.1 . Equivalently, the logarithmic model can be compared with the formal propagators behind the power-phase Beltran–Bennett examples. In the power case one has U_{\alpha}(t)=e^{it|D|^{\alpha}}, and the corresponding time-one smoothed multiplier is |D|^{-\beta}U_{\alpha}(1)=|D|^{-\beta}e^{i|D|^{\alpha}}. In the present logarithmic regime the analogous formal propagator is U_{\log,\gamma}(t)=e^{it(\log(e+|D|))^{\gamma}}, and the model multiplier ( 3 ) is the logarithmically smoothed time-one operator (\log(e+|D|))^{-\beta}U_{\log,\gamma}(1). The respective inverse derivative scales are R^{1-\alpha} and R(\log R)^{1-\gamma} . This comparison is only at the level of Fourier multipliers. To the best of our knowledge, there is no systematic dispersive or Strichartz theory for this specific logarithmic propagator. No such estimates are claimed here. The corresponding logarithmic square function is g_{\log,\gamma,\beta}(f)(x)=\left(\int_{0}^{t_{0}}\int_{|y-x|\leq a_{\gamma}(t)}|f*\phi_{t}(y)|^{2}\frac{\mathrm{d}y}{a_{\gamma}(t)^{d}}(\log(1/t))^{2\beta}\frac{\mathrm{d}t}{t}\right)^{1/2}. We also use the robust square function g^{*}_{\log,\gamma,\beta,\lambda}(f)(x)=\left(\int_{0}^{t_{0}}\int_{\mathbb{R}^{d}}|f*\phi_{t}(y)|^{2}K_{t}^{\lambda}(x-y)(\log(1/t))^{2\beta}\frac{\mathrm{d}t}{t}\right)^{1/2}, where K_{t}^{\lambda}(z)=a_{\gamma}(t)^{-d}\left(1+\frac{|z|}{a_{\gamma}(t)}\right)^{-d\lambda}. These objects are defined precisely in Definition 2.5 . Our first main result is a logarithmic Beltran–Bennett pointwise estimate for the Fourier multiplier operator T_{m} . It says that, under the localized logarithmic Miyachi condition 2.6 , the logarithmic square function of T_{m}f is controlled pointwise by a robust logarithmic square function of f : g_{\log,\gamma,\beta}(T_{m}f)(x)\leq Cg^{*}_{\log,\gamma,0,\lambda}(f)(x). (4) More precisely, if the localized Sobolev order is \sigma>d/2 , then one may take \lambda=\frac{2\sigma}{d}>1. The constant in ( 4 ) may depend on d,\gamma,\beta,\sigma,\lambda , on the constants in the localized Miyachi condition, and on the fixed cutoffs, but not on f , x , the scale t , or the particular log-subdyadic ball. The full statement is Theorem 4.2 . The proof follows the Beltran–Bennett architecture [ 1 ] , but at the logarithmic scale \rho(R) . The log-subdyadic lattice decomposition is constructed in Section 3 . The two geometric estimates are the decoupling and recoupling inequalities, Propositions 3.2 and 3.4 . These are combined with the local multiplier estimate on a single adapted ball, Proposition 4.1 , to obtain Theorem 4.2 . Our second main result is a weighted L^{2} estimate for the high-frequency Fourier multiplier operator T_{m,\mathrm{hi}} . The logarithmic replacement for the Beltran–Bennett geometric maximal operator is M_{\log,\gamma,\beta}w(x)=\sup_{\begin{subarray}{c}0<t<t_{0}\\ |y-x|\leq a_{\gamma}(t)\end{subarray}}(\log(1/t))^{-2\beta}A_{t}w(y), where A_{t}w=\eta_{t}*w , \eta_{t}(x)=t^{-d}\eta(x/t) , and \eta is a fixed nonnegative Schwartz function with integral one. Thus A_{t}w is a smooth average of the weight at the physical Littlewood–Paley scale t , while the supremum is taken over centers y lying in the larger logarithmic tube |y-x|\leq a_{\gamma}(t) . The factor (\log(1/t))^{-2\beta} is not a normalization convention: it is the exact inverse of the logarithmic square-function weight appearing in g_{\log,\gamma,\beta} . The precise definition is given in Definition 5.1 . The weighted theorem states that \int_{\mathbb{R}^{d}}|T_{m,\mathrm{hi}}f(x)|^{2}w(x)\,\mathrm{d}x\lesssim\int_{\mathbb{R}^{d}}|f(x)|^{2}M^{2}\bigl(M_{\log,\gamma,\beta}M^{4}w\bigr)(x)\,\mathrm{d}x. (5) This is the logarithmic analogue of the weighted estimate in