Estimates of Γ \Gamma -functions and their log \log{} -derivatives

Miron Bekker, Boris Mityagin · 2026

Read Estimates of Γ \Gamma -functions and their log \log{} -derivatives by Miron Bekker, Boris Mityagin (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-2605-30378.txt.

Introduction

Miron Bekker and Boris Mityagin give a lower estimate for a weighted ratio of Gamma functions and its log-derivative. They use this bound to bound the norm of the operator tied to the Cesáro operator in the Hardy space of the upper half-plane. This holds for p strictly between one and infinity. The work thus links the Gamma bound to the Cesáro norm.

Catalog record

CREATORMiron Bekker, Boris Mityagin
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE11719
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.30378

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Estimates of Γ \Gamma -functions and their log \log{} -derivatives Miron Bekker, Boris Mityagin Abstract We provide a lower bound for a weighted ratio of Γ − \Gamma- functions and its log − \log{}- derivative. Then we apply this result to estimate the norm of the operator associated with the Cesáro operator in the Hardy space H p ​ ( ℂ + ) H^{p}({\mathbb{C}}_{+}) , 1 < p < ∞ 1<p<\infty . MSC2020 : Primary 33B15. Secondary 30H10, 47B38 Key words and phrases : \Gamma -function and its log-derivative, Hermite-Hadamard Inequality, Hardy Spaces, Cesaro operator. The goal of this note is to prove a couple of inequalities related to \Gamma -functions and their ( \log{} ) derivatives and to apply them to analysis of the norm of the adjusted Cesáro operator in Hardy spaces. 1. Let us define a smooth function \Phi(x)=\begin{cases}2(1-\dfrac{1}{x})^{x-1}\dfrac{\Gamma(1/2+x/2)\Gamma(x)}{\Gamma(1+x/2)\Gamma(-1/2+x)}&x>1;\\ \dfrac{4}{\pi}&x=1,\end{cases} (1) on [1,\infty) . Proposition 1 . For s\geq-1 , s\neq 0 , the strict inequality \log{\Phi(s+2)}>R(s), (2) where R(s)=\frac{1}{24}\frac{s^{2}}{\kappa(\kappa+s)(\kappa+2s)},\text{and}\quad\kappa=1+1/\sqrt{12}, (3) holds. Moreover, \Phi(2)=1 . Proof. Applying the Legendre duplication formula to \Phi(x) and taking the logarithmnic derivative of both sides (see [ 1 ] , formulas 6.1.18, 6.3.1, and 6.3.16; also Appendix, B ), we arrive to the function \displaystyle F(s)\equiv\frac{d}{ds}\left(\log{\Phi(s+2)}\right)=\frac{\Phi^{\prime}(s+2)}{\Phi(s+2)}= (4) \displaystyle\log{\left(\frac{1+s}{1+s/2}\right)}-\sum\limits_{n=0}^{\infty}g(n;s), (5) where g(t;s)=\frac{1}{t+3/2+s/2}-\frac{1}{t+3/2+s},\quad-1<s<+\infty,\quad-\frac{1}{2}\leq t<+\infty. (6) Notice that the second derivative \frac{d^{2}}{dt^{2}}g(t;s)=2\left[\frac{1}{(t+3/2+s/2)^{3}}-\frac{1}{(t+3/2+s)^{3}}\right] (7) is well defined and is (a) strictly positive on [-3/2,+\infty) if s>0 , and (b) strictly negative for t\in[-1/2,+\infty) if -1<s<0. Therefore \displaystyle g(\cdot;s)\quad\text{is strictly convex on}\quad[-3/2,+\infty)\quad\text{if}\quad s>0; (8a) \displaystyle g(\cdot;s)\quad\text{is strictly concave on}\quad[-1/2,+\infty)\quad\text{if}\quad-1<s<0. (8b) If h(t) is a strictly convex (concave) function on the interval [a,b] then h(\frac{a+b}{2})<\frac{1}{b-a}\int\limits_{a}^{b}h(t)dt,\\ (9a) respectively h(\frac{a+b}{2})>\frac{1}{b-a}\int\limits_{a}^{b}h(t)dt,. (9b) This is the Hermite-Hadamard inequality (see [ 6 ] , Chapter 3, Statement 125, or [ 8 ] , Sec.1.10, formula (1.32)). Therefore, if s>0 , for any n\in{\mathbb{Z}}_{+} g(n;s)<\int\limits_{n-1/2}^{n+1/2}g(t;s)dt (10a) and \sum\limits_{n=0}^{\infty}g(n;s)<\sum\limits_{n=0}^{\infty}\int\limits_{n-1/2}^{n+1/2}g(t;s)dt=\int\limits_{-1/2}^{\infty}g(t;s)dt=\log{\left(\frac{1+s}{1+s/2}\right)}, and F(s)>0 if s>0 . If -1<s<0 we apply the Hermite-Hadamard inequality ( 9b ) and conclude by repeating an analogue of ( 10a ) that F(s)=\log{\left(\frac{1+s}{1+s/2}\right)}-\sum\limits_{n=0}^{\infty}g(n;s)<0. (10b) With F(0)=0 or \Phi^{\prime}(2)=0 , and \Phi(2)=1 , we can state as an intermidiate statement the following Claim 2. Under the notations ( 1 ) \log{\Phi(x)}\begin{cases}>0&\text{if}\quad+1\leq x<\infty,\quad x\neq 2,\\ =0&\text{if}\quad x=2.\end{cases} (11) 3. But we want to go farther and make inequality ( 11 ) quantitative, i.e. put as the right hand side some elementary functions of p or s=p-2 . Now we use Peano kernel formula of the midpoint integration rule. The gap E(f):=\frac{1}{b-a}\int\limits_{a}^{b}f(x)dx-f\left(\frac{a+b}{2}\right) (12) is a linear functional on C^{2}([a,b]) , and it has a representation E(f)=\int\limits_{a}^{b}K(t)f^{\prime\prime}(t)dt, where K(t)=\begin{cases}\dfrac{1}{2(b-a)}(t-a)^{2},&a\leq t\leq\mu:=\dfrac{a+b}{2},\\ \dfrac{1}{2(b-a)}(t-b)^{2},&\mu\leq t\leq b\end{cases} (13) (see [ 9 ] ; [ 10 ] , Sec. 13.2.2, formulas (13.49)-(13.51)). Peano kernel on the interval [-1/2,1/2] is k(t)=\frac{1}{2}\left(\frac{1}{2}-|t|\right)^{2}, it is strctly positive. Let us extend this function as a periodic one, with period 1 , i.e. k(t+m)=k(t),\quad|t|\leq 1/2,\quad m\in{\mathbb{Z}}_{+}, then F(s) as a total gap ( 5 ) can be represented in the following way: F(s)=\int\limits_{-1/2}^{\infty}k(t)g^{\prime\prime}(t;s)dt. Case s>0 . We have (-1)^{m}\frac{d^{m}}{dt^{m}}g(t;s)>0,\quad m=0,1,2,3, so g^{\prime\prime}(\cdot;s) is monotone, strictly decreasing. Choose \tau , 0<\tau<1/2 , in such a way that areas of sets \left\{0\leq t\leq\tau,k(\tau)\leq y\leq k(t)\right\} and \left\{\tau\leq t\leq 1/2,k(t)\leq y\leq k(\tau)\right\} are equal, i.e., with k being even and 1- periodic, \int\limits_{-\tau}^{\tau}[k(t)-k(\tau)]dt=\int\limits_{\tau}^{1-\tau}[k(\tau)-k(t)]dt (14) Then \frac{1}{2}\int\limits_{0}^{1/2}\left(\frac{1}{2}-t\right)^{2}dt=\frac{1}{2}\left(\frac{1}{2}-\tau\right)^{2}\frac{1}{2} (15) and \tau=\frac{1}{2}\left(1-\frac{1}{\sqrt{3}}\right),\qquad k(\tau)=\frac{1}{2}\left(\frac{1}{2\sqrt{3}}\right)^{2}=\frac{1}{24}. (16) Figure 1: Extended Peano kernel on [-1/2,1] . Horizontal line represents k(\tau)=1/24 . Points A,B,C correspond to values -\tau , \tau, and 1-\tau , respectively, \tau=1/2(1-1/\sqrt{3}) . Let us notice that if (X,\mu) , (Y,\nu) are spaces with positive measures such that \int_{X}d\mu=\int_{Y}d\nu, (17) and \exists c , p(x)>c>q(y) \forall x\in X , \forall y\in Y , then \int\limits_{X}p(x)d\mu>c\int\limits_{X}d\mu=c\int\limits_{Y}d\nu>\int\limits_{Y}q(y)d\nu. (18) This remark, with X=(-\tau,\tau) , \mu=[h(x)-h(\tau)]dx , Y=(\tau,1-\tau) , d\nu=[h(\tau)-h(y)]dy , c=g^{\prime\prime}(\tau) , and ( 14 ) being ( 17 ), where p and q coincide with g^{\prime\prime}(t) on the corresponding intervals, leads to \int\limits_{-\tau}^{\tau}[k(t)-k(\tau)]g^{\prime\prime}(t;s)dt>\int\limits_{\tau}^{1-\tau}[k(\tau)-k(t)]g^{\prime\prime}(t;s)dt or \int\limits_{-\tau}^{1-\tau}k(t)g^{\prime\prime}(t;s)dt>k(\tau)\int\limits_{-\tau}^{1-\tau}g^{\prime\prime}(t;s)dt The same holds for any shifted interval [-\tau,1-\tau]+m , m\in{\mathbb{Z}}_{+} , so \displaystyle F(s)=\int\limits_{-1/2}^{\infty}k(t)g^{\prime\prime}(t;s)dt> (19) \displaystyle\int\limits_{-\tau}^{\infty}k(t)g^{\prime\prime}(t;s)dt>\frac{1}{24}\int\limits_{-\tau}^{\infty}g^{\prime\prime}(t;s)dt=-\frac{1}{24}g^{\prime}(-\tau). Put \kappa=3/2-\tau=1+1/\sqrt{12} . The straightforward calculation shows that -g^{\prime}(-\tau)=\frac{1}{(\kappa+s/2)^{2}}-\frac{1}{(\kappa+s)^{2}} and \displaystyle\log{\Phi(s+2)}=\int\limits_{0}^{s}F(\sigma)d\sigma> \displaystyle\frac{1}{24}\int\limits_{0}^{s}\left[\frac{1}{(\kappa+\sigma/2)^{2}}-\frac{1}{(\kappa+\sigma)^{2}}\right]d\sigma=\frac{1}{24}\frac{s^{2}}{\kappa(\kappa+s)(2\kappa+s)}. Case s<0 . The function \gamma(t)=-g(t;s) is strictly convex on [-1/2,\infty) by ( 8b ), and by ( 19 ) -F(s)=\int\limits_{-1/2}^{\infty}k(t)\left[-\frac{d^{2}}{dt^{2}}g(t;s)\right]dt>\frac{1}{24}g^{\prime}(-\tau;s). (20) Therefore, \displaystyle\log{\Phi(s+2)}=\int\limits_{s}^{0}[-F(\sigma)]d\sigma> \displaystyle\frac{1}{24}\int\limits_{s}^{0}\left[\frac{1}{(\kappa+\sigma)^{2}}-\frac{1}{(\kappa+\sigma/2)^{2}}\right]d\sigma=\frac{1}{24}\frac{s^{2}}{\kappa(\kappa+s)(2\kappa+s)}. Q.E.D. It is interesting to compare functions \log{\Phi(s+2)} and R(s) given by ( 1 )-( 3 ) (see Figure 2). At infinity, by Stirling asympotic formula \lim\limits_{s\to\infty}\log{\Phi(s+2)}=\frac{3}{2}\log{2}-1=0.039720771\ldots and \lim\limits_{s\to\infty}R(s)=\frac{1}{24\kappa}=\frac{1}{12(2+1/\sqrt{3})}=0.032332948\ldots Figure 2: \Phi(s+2) ( solid line) and \exp{(R(s))} (dash line) for -0.8\leq s\leq 10 4. Applications. The function \Phi(x) , defined by the formula ( 1 ), appeared in [ 3 ] , where the authors made an attempt to show that the norm of the adjusted Cesáro operator V(p)=2\frac{p-1}{p}C-I,\quad 1<p<\infty,\quad p\neq 2 (21) in Hardy space H^{p}({\mathbb{C}}_{+}) , exceeds its spectral radius which is equal 1 . (To make our note self-contained we provide a few facts about Cesáro operator C:f\to\frac{1}{z}\int\limits_{0}^{z}f(\zeta)d\zeta (22) in Appendix). As a step to this direction, in [ 3 ] the test function h(z)=\frac{1}{(i+z)^{2}}. (23) was used. The staightforward calculations (we leave them to a reader or refer to formulas (5)-(7) in [ 3 ] ) lead to the following statement: Lemma 1 . On the Hardy space H^{p}(C_{+}) , 1<p<\infty , we have \|V(p)\|^{p}_{p}\geq\frac{\|V(p)h\|_{p}^{p}}{\|h\|^{p}_{p}}\geq\Phi(p) with equality only for p=2 , where \Phi is defined by ( 1 ). Now, with our results of Sections 1 and 2 above, we can claim the following. Proposition 2 . For the operator V(p) defined by ( 21 ) in Hardy space H^{p}({\mathbb{C}}_{+}) , 1<p<\infty , p\neq 2 , the norm \|V(p)\|_{p}>1 . Moreover \|V(p)\|_{p}^{p}>\exp{(R(p-2))} where R(s)=\frac{1}{24}\frac{s^{2}}{\kappa(\kappa+s)(2\kappa+s)},\quad\kappa=1+1/\sqrt{12}. It was proved in [ 2 ] that for p=2 the operator V(2) is unitary. 5. Appendix. In this Appendix we provide a brief information about Hardy spaces H^{p}({\mathbb{C}}_{+}) in the upper half-plabe {\mathbb{C}}_{+}=\{z:\im z>0\} and the Cesáro operator in these spaces. For detailed information about Hardy spaces we refer a reader to [ 5 ] (Chapter 11), [ 7 ] (in particular, Chapter VI, Sect. C, p. 112 ). We also recall a couple of identities related to \Gamma- function and its \log{}- deraivative that were used above. A. The Hardy space H^{p}({\mathbb{C}}_{+}) , 1\leq p<\infty consists of functions f(z) , which are analytic in the upper half-plane {\mathbb{C}}_{+} and staitsfy condition \left[\sup\left\{\frac{1}{2\pi}\int\limits_{-\infty}^{\infty}|f(x+iy)|^{p}dx:y>0\right\}\right]^{1/p}<\infty. (24) The space H^{\infty}({\mathbb{C}}_{+}) consists of functions which are analytic and bounded in {\mathbb{C}}_{+} . H^{p}({\mathbb{C}}_{+}) are Banach spaces. For 1\leq p<\infty the norm is defined by the left side of ( 24 ), and \|f\|_{\infty}=\sup\left\{|f|:z\in{\mathbb{C}}_{+}\right\} . For a function f\in H^{p}({\mathbb{C}}_{+}) the nontangential limit \lim\limits_{y\to 0}f(x+iy)=f^{*}(x) exists for almost all x , and the function f^{*}\in L^{p}({\mathbb{R}}) . The function f(z) can be restored from f^{*}(x) with the help of Cauchy formula. The Cesáro operator on the Hardy space H^{p}({\mathbb{C}}_{+}) in the upper half-plane was introduced in [ 4 ] . It is defined by the expression ( 22 ). In [ 4 ] it was proved that for 1<p\leq\infty the operator C is bounded on H^{p}({\mathbb{C}}_{+}) (another proof was given in [ 2 ] ) and for 1<p<\infty its spectrum \sigma(C) is given by \sigma(C)=\{w:|w-\frac{p}{2(p-1)}|=\frac{p}{2(p-1)}\}. (25) From the last formula it follows that the spectrum of the operator V(p) , defined by ( 21 ), is the unit circle. B. 1. Legendre duplication

Read the complete text