The Law of the Iterated Logarithm for smooth functions
José G. Llorente and Artur Nicolau · 2026
Read The Law of the Iterated Logarithm for smooth functions by José G. Llorente and Artur Nicolau (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-2605-20018.txt.
Introduction
José Llorente and Artur Nicolau prove a version of a classical limit law for smooth functions on the upper half-space. Size conditions on the gradient, and on the gradient of the Laplacian, then yield self-improvement growth properties for such functions. The results apply even where harmonicity is absent, extending the classical law beyond harmonic settings. Those growth properties follow from the size conditions alone, without assuming the function is harmonic.
Catalog record
| CREATOR | José G. Llorente and Artur Nicolau |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 39120 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.20018
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
The Law of the Iterated Logarithm for smooth functions Thanks: Both authors are partially supported by the Spanish Ministerio de Ciencia e Innovación (grant PID2021-123151NB-I00). The second author is also supported by the Generalitat de Catalunya (grant 2021 SGR 00071) and the Spanish Research Agency through the María de Maeztu Program (CEX2020-001084-M) José G. Llorente and Artur Nicolau Abstract . A version of the Law of the Iterated Logarithm for smooth functions in the upper-half space is proved. As a consequence, we show that certain size conditions on the gradient and the gradient of the laplacian of a smooth function, lead to self-improvement growth properties. The results are applied in situations where harmonicity is not present. 1. Introduction In this paper, we study self-improving growth properties of smooth functions in the upper half-space \mathbb{R}^{d+1}_{+}=\{(x,y):x\in\mathbb{R}^{d}\,,y>0\} under size restrictions on their derivatives. A function u\in C^{1}(\mathbb{R}^{d+1}_{+}) is called a Bloch function if B=\sup\{y|\nabla u(x,y)|:(x,y)\in\mathbb{R}^{d+1}_{+}\}<\infty and we will refer to B as the Bloch constant of u . Analytic Bloch functions in a half-plane play a crucial role in several topics as conformal mappings, harmonic measure, Bergman spaces, trigonometric series and Zygmund functions among others (see [ 13 , 1 , 5 , 8 , 3 , 11 , 12 ] and references therein). Significant examples of harmonic Bloch functions are those of the form u=\log|f^{\prime}| where f:\mathbb{R}_{2}^{+}\to\mathbb{C} is a conformal mapping as well as harmonic extensions of lacunary trigonometric series with bounded coefficients. Observe that if u is a Bloch function in \mathbb{R}^{d+1}_{+} with Bloch constant B , then the following global estimate holds \limsup_{y\to 0}\frac{|u(x,y)|}{\log\big(1/y\big)}\leq B,\quad x\in\mathbb{R}^{d}. The celebrated Makarov’s Law of the Iterated Logarithm (LIL) for Bloch harmonic functions says that, in the presence of harmonicity, the global bound \log(1/y) can be substantially improved for almost every point x\in\mathbb{R}^{d} . The following theorem is the half-space version of Makarov’s original result in the unit disk ( [ 11 , 12 ] ). See also [ 2 , 10 ] for further results, still in the harmonic case. Theorem A . Let u be a harmonic Bloch function in \mathbb{R}^{d+1}_{+} with Bloch constant B . Then there exists a constant C>0 only depending on d and B such that (1.1) \limsup_{y\to 0}\frac{\displaystyle|u(x,y)|}{\sqrt{\log(1/y)\log\log\log(1/y)}}\leq C for almost every x\in\mathbb{R}^{d} . Moreover ( 1.1 ) is sharp in the sense that there exist harmonic Bloch functions u for which the \limsup in ( 1.1 ) is bounded below by a positive constant for almost every x\in\mathbb{R}^{d} . Our main purpose is to understand to what extent harmonicity can be relaxed in this result. This problem has already been considered in [ 7 , 4 ] but the results there do not provide the right analog of the LIL. Note that the function u(x,y)=\log(1/y) is Bloch and exhibits the maximal vertical growth allowed by the Bloch condition at every x\in\mathbb{R}^{d} so no self-improvement occurs. Since \Delta u(x,y)=y^{-2} , this example suggests that in order to obtain meaningful self-improvement results for non-harmonic Bloch functions, the behavior of their laplacian may play an important role. We introduce two classes of smooth functions in \mathbb{R}^{d+1}_{+} whose gradient and the gradient of the laplacian have controlled growth in terms of certain gauge functions. Let \psi,\varepsilon:(0,1]\to(0,+\infty) and define \displaystyle\mathcal{B}_{\psi} \displaystyle\equiv\{u\in C^{1}(\mathbb{R}^{d+1}_{+})\,:y|\nabla u(x,y)|\leq\psi(y),\,\,x\in\mathbb{R}^{d},\,0<y\leq 1\,\},\vskip 3.41418pt \displaystyle\mathcal{B}_{\psi,\varepsilon} \displaystyle\equiv\mathcal{B}_{\psi}\cap\{u\in C^{3}(\mathbb{R}^{d+1}_{+})\,:y^{3}|\nabla\Delta u(x,y)|\leq\varepsilon(y),\,\,x\in\mathbb{R}^{d},\,0<y\leq 1\,\}.\vskip 3.41418pt We will eventually require the following assumptions on \psi and \varepsilon : (1.2) \displaystyle\psi\,\text{is non-increasing, }\vskip 3.41418pt (1.3) \displaystyle\varepsilon\leq\psi,\vskip 3.41418pt (1.4) \displaystyle\displaystyle\frac{1}{y}\int_{0}^{y}\,\psi(t)dt\leq A\,\psi(y),\quad 0<y\leq 1,\vskip 3.41418pt for some positive constant A . Observe that ( 1.2 ) and ( 1.4 ) easily imply the following doubling property: (1.5) \psi(y/2)\leq 2A\psi(y),\quad 0<y\leq 1. Associated to \psi , define the square function \Psi as (1.6) \Psi(y)=\int_{y}^{1}\frac{\psi^{2}(t)}{t}dt\,\,,\qquad 0<y\leq 1. The choice of the constant function \psi\equiv B>0 gives \Psi(y)=B^{2}\log(1/y) and \mathcal{B}_{\psi} becomes then the class of Bloch functions with Bloch constant not greater than B . We now state the main result of this paper. Theorem 1.1 . Suppose that u\in\mathcal{B}_{\psi,\varepsilon} where \psi,\varepsilon satisfy ( 1.2 ), ( 1.3 ), ( 1.4 ) and let \Psi be the square function associated to \psi defined by ( 1.6 ). Then there exists a constant C>0 only depending on d , \psi , \varepsilon such that (1.7) \limsup_{y\to 0}\frac{\displaystyle\Big|u(x,y)-\int_{y}^{1}t\Delta u(x,t)dt\Big|}{\sqrt{\Psi(y)\log\log\Psi(y)}}\leq C, for almost every x\in\mathbb{R}^{d} . One of the remarkable ideas of the work of Makarov was that the boundary behavior of a harmonic Bloch function in a half space can be related to the asymptotic behavior of a dyadic martingale (see [ 11 , 12 ] ). Actually the mean value property of harmonic functions can be understood as an analog to the cancellation property of dyadic martingales. When the function u\in C^{3}(\mathbb{R}^{d+1}_{+}) is not harmonic, it turns out that instead of u , the following expression T(x,y)=u(x,y)-y\frac{\partial u}{\partial y}(x,y)-\int_{y}^{1}h\Delta u(x,h)dh,\quad 0<y<1,\,x\in\mathbb{R}^{d}, can be compared with a dyadic martingale. Since one can estimate the quadratic variation of this dyadic martingale by the functions \psi and \varepsilon , the proof of Theorem 1.1 can be transferred to the context of dyadic martingales. The assumptions on the function \varepsilon in the next two corollaries guarantee that the integral in the numerator of ( 1.7 ) can be absorbed by the denominator of ( 1.7 ). Corollary 1.1 . Let \psi , \varepsilon and u\in\mathcal{B}_{\psi,\varepsilon} be as in Theorem 1.1 . Suppose, in addition, that (1.8) \limsup_{y\to 0}\frac{\displaystyle\int_{y}^{1}\varepsilon(t)t^{-1}dt}{\sqrt{\Psi(y)\log\log\Psi(y)}}=M<\infty. Then there exists a constant C>0 only depending on d , \psi and M , such that \limsup_{y\to 0}\frac{\displaystyle|u(x,y)|}{\sqrt{\Psi(y)\log\log\Psi(y)}}\leq C, for almost every x\in\mathbb{R}^{d} . In the case of Bloch functions, one gets the following result. Corollary 1.2 . Let u\in C^{3}(\mathbb{R}^{d+1}_{+}) . Assume that there exists a constant B>0 such that y|\nabla u(x,y)|\leq B,\qquad y^{3}|\nabla\Delta u(x,y)|\leq\varepsilon(y)\leq B for 0<y\leq 1 , x\in\mathbb{R}^{d} . Suppose, in addition, that (1.9) \limsup_{y\to 0}\,\varepsilon(y)\sqrt{\frac{\log\big(1/y\big)}{\log\log\log\big(1/y\big)}}=D<\infty. Then there exists a constant C>0 only depending on d , B and D , such that ( 1.1 ) holds for almost every x\in\mathbb{R}^{d} . Hence a version of the LIL for smooth (non-harmonic) Bloch functions u holds under the restriction (1.10) y^{3}|\nabla\Delta(u)(x,y)|\leq C(\log(1/y))^{-1/2} if 0<y<1 and x\in\mathbb{R}^{d} . It is worth mentioning that the exponent 1/2 in ( 1.10 ) is sharp. Actually for 0<\alpha<1 , the function u(x,y)=(\log(1/y))^{\alpha} is in \mathcal{B}_{\psi,\varepsilon} where \psi(y)=\varepsilon(y)=(\log(1/y))^{\alpha-1} while the LIL ( 1.1 ) can only hold for \alpha\leq 1/2 . We know make two remarks. 1. Given two functions \psi,\eta:(0,1]\to(0,+\infty) consider the class \mathcal{K}_{\psi,\eta} of functions u\in C^{2}(\mathbb{R}^{d+1}_{+}) such that y|\nabla u(x,y)|\leq\psi(y) and y^{2}|\Delta u(x,y)|\leq\eta(y) for any x\in\mathbb{R}^{d} and 0<y\leq 1 . Assume that \psi satisfies ( 1.2 ) and ( 1.4 ) and in addition, that there exists a constant A^{\prime}>0 such that (1.11) \int_{y}^{1}\frac{\eta(t)}{t}dt\leq A^{\prime}\psi(y),\quad 0<y\leq 1. The proof of Theorem 1.1 applies with minor modifications and ( 1.7 ) holds for any function u\in\mathcal{K}_{\psi,\eta} . Since the assumption ( 1.11 ) implies that the integral in the numerator of ( 1.7 ) is bounded by the denominator of ( 1.7 ), it follows that ( 1.1 ) holds for any function u\in\mathcal{K}_{\psi,\eta} . If \psi is a constant function, condition ( 1.11 ) reduces to the integrability of \eta(t)/t . In this case, one can also prove ( 1.1 ) by decomposing u as the sum