betaFeedback

WorksMathematics

Two-Sided large oscillations of the argument of the Riemann zeta function and its iterates

Abstract

Assuming the Riemann hypothesis, we establish lower bounds for large oscillations of the iterated argument functions Sn(t)S_n(t) of the Riemann zeta function. As a corollary, we prove that, for every fixed integer n≥0n \geq 0, Sn(t)=Ω±((log⁡t)1/2(log⁡3t)1/2(log⁡2t)n+1/2).S_n(t)=Ω_{\pm}\Big(\frac{(\log t)^{1/2}(\log_3 t)^{1/2}}{(\log_2 t)^{n+1/2}} \Big). For even nn, this recovers two results of Chirre and Mahatab in 2021. For odd nn, our result strengthens theirs by establishing the missing direction of oscillation with the same order of magnitude.

Introduction

Throughout this paper, we denote the Riemann zeta function by ζ(s)\zeta(s), where s=σ+its = \sigma+ it is a complex variable. For 1/2≤σ≤11/2 \le\sigma\le1 and t>0t > 0, define

S(σ,t):=1πarg⁡ζ(σ+it),S(\sigma,t) := \frac{1}{\pi}\arg\zeta(\sigma+ it),

where the argument is obtained by continuous variation along straight line segments joining the points 22, 2+it2 + it and σ+it\sigma+ it, assuming that the segment from σ+it\sigma+ it to 2+it2 + it has no zeros of ζ(s)\zeta(s), and with the convention that arg⁡ζ(2)=0\arg\zeta(2) = 0. If this path has zeros of ζ(s)\zeta(s) (including the endpoint σ+it\sigma+ it), define

S(σ,t)=S(σ,t+0)+S(σ,t−0)2.S(\sigma,t) = \frac{S(\sigma,t+0) + S(\sigma,t-0)}{2}.

Set S0(σ,t)=S(σ,t)S_0(\sigma,t) = S(\sigma,t). Then, for integer n≥1n \ge1, define

Sn(σ,t)=∫0tSn−1(σ,τ) dτ+δn,σ,S_n(\sigma,t) = \int_0^t S_{n-1}(\sigma,\tau)\,\mathrm{d}\tau+ \delta_{n,\sigma},

where the constant δn,σ\delta_{n,\sigma} is given by Fujii [11]. More precisely, for n=2kn = 2k with k≥1k \ge1,

δ2k,σ=(−1)k−1∫σ1∫u2k1⋯∫u31∫u21du1du2⋯du2k=(−1)k−1(1−σ)2k(2k)!.\delta_{2k,\sigma} = (-1)^{k-1}\int_{\sigma}^{1}\int_{u_{2k}}^{1}\cdots\int_{u_3}^{1}\int_{u_2}^{1}\mathrm{d}u_1\mathrm{d}u_2\cdots\mathrm{d}u_{2k} = \frac{(-1)^{k-1}(1-\sigma)^{2k}}{(2k)!}.

and for n=2k−1n = 2k - 1 with k≥1k \ge1,

δ2k−1,σ=(−1)k−1π∫σ∞∫u2k−1∞⋯∫u3∞∫u2∞log⁡∣ζ(u1)∣ du1du2⋯du2k−1.\delta_{2k-1,\sigma} = \frac{(-1)^{k-1}}{\pi} \int_{\sigma}^{\infty} \int_{u_{2k-1}}^{\infty} \cdots\int_{u_3}^{\infty} \int_{u_2}^{\infty} \log\lvert\zeta(u_1) \rvert\,\mathrm{d}u_1\mathrm{d}u_2 \cdots\mathrm{d}u_{2k-1}.

The function S(t)S(t) is closely related to the distribution of the non-trivial zeros of the Riemann zeta function. The classical Riemann–von Mangoldt formula states that

N(t)=t2πlog⁡(t2πe)+78+S(t)+O(t−1).N(t) = \frac{t}{2\pi} \log\left(\frac{t}{2\pi e}\right) + \frac{7}{8} + S(t) + O(t^{-1}).

where N(t)N(t) denotes the number of zeros ρ=β+iγ\rho= \beta+ i\gamma of ζ(s)\zeta(s) with 0<γ≤t0 < \gamma\le t, counted with multiplicity, where zeros with ordinate tt are counted with half weight. In particular, for 0<h≤10 < h \le1, we have

N(t+h)−N(t)=h2πlog⁡t+S(t+h)−S(t)+O(h+t−1).N(t+h) - N(t) = \frac{h}{2\pi} \log t + S(t+h) - S(t) + O(h+t^{-1}).

This relation shows that the oscillations of S(t)S(t) are intimately connected with the local distribution of zeros of ζ(s)\zeta(s). More generally, the iterated functions Sn(t)S_n(t) describe successive integrals of these fluctuations. Understanding their large values and oscillations is therefore a natural problem in the study of the Riemann zeta function.

The study of large values of S(t)S(t) has a long history. In 1977, Montgomery [15] established, assuming the Riemann hypothesis (RH), the following Ω±\Omega_{\pm}-result:

S(t)=Ω±((log⁡t)1/2(log⁡2t)1/2).S(t) = \Omega_{\pm}\left(\frac{(\log t)^{1/2}}{(\log_2 t)^{1/2}}\right).

Here and throughout this paper, we write log⁡j\log_j for the jj-th iterated logarithm. A heuristic prediction for the true order of magnitude of S(t)S(t) was proposed by Farmer, Gonek and Hughes [10]. Based on random matrix theory, they conjectured that

lim sup⁡t→∞S(t)(log⁡t)1/2(log⁡2t)1/2=1π2.\limsup_{t\to\infty} \frac{S(t)}{(\log t)^{1/2}(\log_2 t)^{1/2}} = \frac{1}{\pi\sqrt{2}}.

This conjecture suggests that large values of S(t)S(t) may be substantially larger than the currently known lower bounds.

Tsang [16] subsequently obtained several refinements of the classical Ω\Omega-results. More precisely, he proved that

S(t)=Ω±((log⁡t)1/3(log⁡2t)1/3),unconditionally,S1(t)=Ω±((log⁡t)1/2(log⁡2t)3/2),under RH.S(t) = \Omega_{\pm}\left(\frac{(\log t)^{1/3}}{(\log_2 t)^{1/3}}\right), \qquad\text{unconditionally}, \qquad S_1(t) = \Omega_{\pm}\left(\frac{(\log t)^{1/2}}{(\log_2 t)^{3/2}}\right), \qquad\text{under RH}.

In addition, Tsang obtained unconditional estimates for S1(t)S_1(t), with different orders in the two directions. In a subsequent paper, Tsang [17] further improved Ω+\Omega_+-result for S1(t)S_1(t).

A significant development came from the resonance method. In 2018, Bondarenko and Seip [2 Theorem 2] proved, under RH, that

S(t)=Ω((log⁡t)1/2(log⁡3t)1/2(log⁡2t)1/2),S1(t)=Ω+((log⁡t)1/2(log⁡3t)1/2(log⁡2t)3/2).S(t) = \Omega\left(\frac{(\log t)^{1/2}(\log_{3} t)^{1/2}}{(\log_{2} t)^{1/2}}\right), \qquad S_{1}(t) = \Omega_{+}\left(\frac{(\log t)^{1/2}(\log_{3} t)^{1/2}}{(\log_{2} t)^{3/2}}\right).

These results introduced an additional factor of (log⁡3t)1/2(\log_{3} t)^{1/2} compared with the earlier conditional work, although the bound for S(t)S(t) did not distinguish between its positive and negative values.

Using a modified version of the resonance method in [2], Chirre [7] investigated large values of Sn(σ,t)S_{n}(\sigma,t) near the critical line. Assuming RH, when σ∈[1/2,1/2+1/(log⁡2T)]\sigma\in[1/2,1/2+1/(\log_{2} T)], he obtained lower bounds of order, for sufficiently large TT,

(log⁡T)1−σ(log⁡3T)σ(log⁡2T)σ+n.\frac{(\log T)^{1-\sigma}(\log_{3} T)^{\sigma}}{(\log_{2} T)^{\sigma+n}}.

When n≡1(mod4)n \equiv1 \pmod{4}, he established an Ω+\Omega_{+}-result for Sn(σ,t)S_{n}(\sigma,t), whereas in the other cases, a corresponding Ω\Omega-result is available. In 2021, Chirre and Mahatab subsequently made two important contributions. Specifically, in [8] they obtained that

S(t)=Ω±((log⁡t)1/2(log⁡3t)1/2(log⁡2t)1/2).S(t) = \Omega_{\pm}\left(\frac{(\log t)^{1/2}(\log_{3} t)^{1/2}}{(\log_{2} t)^{1/2}}\right).

This result complements the missing negative direction in [2]. On the other hand, Chirre and Mahatab [9] extended this approach to the iterated functions Sn(σ,t)S_{n}(\sigma,t) near the critical line. Assuming RH, they established large oscillations of Sn(σ,t+h)−Sn(σ,t)S_{n}(\sigma,t+h)-S_{n}(\sigma,t) for every fixed integer n≥1n \ge1. Their result gives both directions when nn is odd, but only one prescribed direction when nn is even. Consequently, their results for Sn(t)S_{n}(t) are two-sided for even nn, whereas for odd nn only one direction is obtained at this order of magnitude. They also derived certain unconditional one-sided consequences for higher iterates.

More recently, Inoue, Kobayashi and Toma [13] further studied explicit extreme values of S(t+h)−S(t)S(t+h)-S(t) in short intervals under RH. Their results also have applications to large and small gaps between consecutive zeros of the Riemann zeta function. For more developments and details about the argument of ζ(s)\zeta(s) and its iterates, we recommend [3, 4, 5, 6, 11, 12] and the references therein.

Motivated by these developments, we investigate whether large oscillation estimates of Chirre and Mahatab [9] can be strengthened to both directions for every fixed integer n≥1n \ge1. Our main result gives an affirmative answer under RH, without imposing any restriction on the parity of nn.

Theorem 1.1. Assume RH. Let n≥1n \ge1 be a fixed integer and let 0<β<10 < \beta< 1 be fixed. Then there exists a constant cn,β>0c_{n,\beta} > 0 such that, for all sufficiently large TT and

h∈[0,(log⁡2T)−1]h \in\left[0,(\log_{2} T)^{-1}\right], we have

max⁡Tβ≤t≤Tε(Sn(t+h)−Sn(t))≥cn,βh(log⁡T)1/2(log⁡3T)1/2(log⁡2T)n−1/2,ε∈{−1,1}.\max_{T^{\beta}\le t\le T}\varepsilon\left(S_{n}(t+h)-S_{n}(t)\right)\ge c_{n,\beta}h\frac{(\log T)^{1/2}(\log_{3}T)^{1/2}}{(\log_{2}T)^{n-1/2}},\qquad\varepsilon\in\{-1,1\}.

Corollary 1.2. Assume RH. Let n≥0n\ge0 be a fixed integer and let 0<β<10<\beta<1 be fixed. Then there exists a constant cn,β>0c_{n,\beta}>0 such that, for all sufficiently large TT, we have

max⁡Tβ≤t≤TεSn(t)≥cn,β(log⁡T)1/2(log⁡3T)1/2(log⁡2T)n+1/2,ε∈{−1,1}.\max_{T^{\beta}\le t\le T}\varepsilon S_{n}(t)\ge c_{n,\beta}\frac{(\log T)^{1/2}(\log_{3}T)^{1/2}}{(\log_{2}T)^{n+1/2}},\qquad\varepsilon\in\{-1,1\}.

In particular,

Sn(t)=Ω±((log⁡t)1/2(log⁡3t)1/2(log⁡2t)n+1/2).S_{n}(t)=\Omega_{\pm}\left(\frac{(\log t)^{1/2}(\log_{3}t)^{1/2}}{(\log_{2}t)^{n+1/2}}\right).

We now compare our results with the work of Chirre and Mahatab. In fact, [9 Theorem 1] establishes two-sided large oscillations of Sn(t+h)−Sn(t)S_{n}(t+h)-S_{n}(t) when nn is odd, but only one prescribed direction when nn is even. Theorem 1.1 removes this restriction and establishes both directions for every fixed integer n≥1n\ge1, with the same order of magnitude. On the other hand, for Sn(t)S_{n}(t), Corollary 1.2 recovers [9 Corollary 2] when nn is even and σ=1/2\sigma=1/2. When nn is odd, it strengthens their result by establishing the missing direction. Furthermore, when n=0n=0, Corollary 1.2 recovers [8 Theorem 1].

Although our results are stated on the critical line for simplicity, the same method yields the corresponding estimates uniformly for 1/2≤σ≤1/2+(log⁡2T)−11/2\le\sigma\le1/2+(\log_{2}T)^{-1}, with Sn(t)S_{n}(t) replaced by Sn(σ,t)S_{n}(\sigma,t).

The choice of the kernel K(u)K(u), defined in (1), is inspired by Jaskari [14]. The resulting convolution formula, established in Proposition 2.3, plays a crucial role in the proof of Theorem 1.1. In contrast to [9 Proposition 8], our approach avoids determining which of two finite sums yields the main term based on the parity of nn. This is achieved by suitably choosing the phase of the sine function in K(u)K(u). Since in(−i)n−1=ii^{n}(-i)^{n-1}=i, this choice guarantees that, after taking the imaginary part, the second sum always appears with coefficient ±1/2\pm1/2, independent of nn.

Preliminary tools

A convolution formula

For h∈Rh\in\mathbb{R} and a function FF, we write

ΔhF(t)=F(t+h)−F(t−h).\Delta_{h}F(t)=F(t+h)-F(t-h).

The following two lemmas, due to Chirre and Mahatab [9], play a crucial role in establishing the convolution formula we require.

Lemma 2.1 (Chirre-Mahatab [9], Lemma 6). Assume RH. Let 0<β<10<\beta<1 be a fixed number. Let α>0\alpha>0, H∈RH\in\mathbb{R} and TT be sufficiently large. Then for 1/2<ν≤21/2<\nu\le2, 0≤h≤10\le h\le1

and Tβ≤t≤Tlog⁡TT^\beta\le t \le T\log T, we have

∫−(log⁡T)3(log⁡T)3Δhlog⁡ζ(v+i(t+u))(sin⁡αuu)2eiHu du=−πi∑m≥2Λ(m)ωm(α,H)sin⁡(hlog⁡m)(log⁡m)mv+it+O(h(1+∣log⁡(2v−1)∣)e2α+∣H∣(log⁡T)3),\int_{-(\log T)^3}^{(\log T)^3}\Delta_h\log\zeta(v+i(t+u))\left(\frac{\sin\alpha u}{u}\right)^2e^{iHu}\,du = -\pi i\sum_{m\ge2}\frac{\Lambda(m)\omega_m(\alpha,H)\sin(h\log m)}{(\log m)m^{v+it}}+O\left(\frac{h(1+|\log(2v-1)|)e^{2\alpha+|H|}}{(\log T)^3}\right),

where ωm(α,H)=max⁡{0,2α−∣H−log⁡m∣}\omega_m(\alpha,H)=\max\{0,2\alpha-|H-\log m|\}.

Lemma 2.2 (Chirre-Mahatab [9], Lemma 7). Assume RH. Let n≥1n\ge1 be a fixed integer, and t,h∈Rt,h\in\mathbb{R} such that t≠±ht\ne\pm h. Then, we have

ΔhSn(t)=1πIm⁡(in(n−1)!∫1/22(v−12)n−1Δhlog⁡ζ(v+it) dv)+On(h).\Delta_hS_n(t)=\frac{1}{\pi}\operatorname{Im}\left(\frac{i^n}{(n-1)!}\int_{1/2}^{2}\left(v-\frac{1}{2}\right)^{n-1}\Delta_h\log\zeta(v+it)\,dv\right)+O_n(h).

We now construct the convolution formula we need. To this end, for large TT, n≥1n\ge1 and ε∈{−1,1}\varepsilon\in\{-1,1\}, define

K(u)=2+εsin⁡(n−12π−ulog⁡2T)=2e0+εin2(−i)n−1eiulog⁡2T−εin2e−iulog⁡2T.(1)\begin{aligned} K(u)=2+\varepsilon\sin\left(\frac{n-1}{2}\pi-u\log_2 T\right) \\ &=2e^0+\frac{\varepsilon i^n}{2}(-i)^{n-1}e^{iu\log_2 T}-\frac{\varepsilon i^n}{2}e^{-iu\log_2 T}. \tag*{(1)} \end{aligned}

Note that K(u)K(u) is always positive. More precisely, 1≤K(u)≤31\le K(u)\le3.

Proposition 2.3. Assume RH. Let n≥1n\ge1 be a fixed integer and let 0<β<10<\beta<1 and 0<γ≤1/20<\gamma\le1/2 be fixed. Let TT be sufficiently large. Then for 0≤h≤10\le h\le1 and Tβ≤t≤Tlog⁡TT^\beta\le t\le T\log T, we have

∫−(log⁡T)3(log⁡T)3ΔhSn(t+u)(sin⁡(γulog⁡2T)u)2K(u) du=Im⁡(2in+3A1(n,t))+ε2Re⁡A2(n,t)+On(hlog⁡2T),\int_{-(\log T)^3}^{(\log T)^3}\Delta_hS_n(t+u)\left(\frac{\sin(\gamma u\log_2 T)}{u}\right)^2K(u)\,du =\operatorname{Im}\left(2i^{n+3}A_1(n,t)\right)+\frac{\varepsilon}{2}\operatorname{Re}A_2(n,t)+O_n(h\log_2 T),

where

Aj(n,t)=∑m≥2Λ(m)aj(m)(log⁡m)n+1m1/2+it,j=1,2.A_j(n,t)=\sum_{m\ge2}\frac{\Lambda(m)a_j(m)}{(\log m)^{n+1}m^{1/2+it}},\qquad j=1,2.

Moreover,

a1(m):=a1(m,T,h)=ωm(γlog⁡2T,0)sin⁡(hlog⁡m),a_1(m):=a_1(m,T,h)=\omega_m(\gamma\log_2 T,0)\sin(h\log m),
a2(m):=a2(m,T,h)=ωm(γlog⁡2T,log⁡2T)sin⁡(hlog⁡m).a_2(m):=a_2(m,T,h)=\omega_m(\gamma\log_2 T,\log_2 T)\sin(h\log m).

Proof. Set α=γlog⁡2T\alpha=\gamma\log_2 T in Lemma 2.1, and set H=−log⁡2T,0,log⁡2TH=-\log_2 T,0,\log_2 T, respectively. Then, we have

∫−(log⁡T)3(log⁡T)3Δhlog⁡ζ(v+i(t+u))(sin⁡(γulog⁡2T)u)2K(u) du\int_{-(\log T)^3}^{(\log T)^3}\Delta_h\log\zeta(v+i(t+u))\left(\frac{\sin(\gamma u\log_2 T)}{u}\right)^2K(u)\,du
=−2πi∑m≥2Λ(m)a1(m)(log⁡m)mv+it+πε2(−i)n−1∑m≥2Λ(m)a2(m)(log⁡m)mv+it+O(h(1+∣log⁡(2v−1)∣)(log⁡T)2−2γ).(2)= -2\pi i\sum_{m\ge2}\frac{\Lambda(m)a_1(m)}{(\log m)m^{v+it}}+\frac{\pi\varepsilon}{2}(-i)^{n-1}\sum_{m\ge2}\frac{\Lambda(m)a_2(m)}{(\log m)m^{v+it}}+O\left(\frac{h(1+\lvert\log(2v-1)\rvert)}{(\log T)^{2-2\gamma}}\right). \tag*{(2)}

Here, we use the linear combination (1). Furthermore, when 0<γ≤1/20<\gamma\le1/2, we have ωm(γlog⁡2T,−log⁡2T)=0\omega_m(\gamma\log_2 T,-\log_2 T)=0 for all m≥2m\ge2. Multiplying both sides of (2) by (v−1/2)n−1(v-1/2)^{n-1} and integrating over v∈[1/2,2]v\in[1/2,2] gives

∫−(log⁡T)3(log⁡T)3(∫1/22(v−12)n−1Δhlog⁡ζ(v+i(t+u)) dv)(sin⁡(γulog⁡2T)u)2K(u) du=−∑m≥2Λ(m)(log⁡m)mit(2πia1(m)−πε2(−i)n−1a2(m))∫1/22(v−1/2)n−1mv dv+O(h(log⁡T)2−2γ∫1/22(v−12)n−1(1+∣log⁡(2v−1)∣) dv)=−∑m≥2Λ(m)(log⁡m)mit(2πia1(m)−πε2(−i)n−1a2(m))∫1/22(v−1/2)n−1mv dv+O(h(log⁡T)2−2γ).(3)\begin{aligned} &\int_{-(\log T)^3}^{(\log T)^3}\left(\int_{1/2}^{2}\left(v-\frac{1}{2}\right)^{n-1}\Delta_h\log\zeta(v+i(t+u))\,dv\right)\left(\frac{\sin(\gamma u\log_2 T)}{u}\right)^2K(u)\,du \\ &=-\sum_{m\ge2}\frac{\Lambda(m)}{(\log m)m^{it}}\left(2\pi i a_1(m)-\frac{\pi\varepsilon}{2}(-i)^{n-1}a_2(m)\right)\int_{1/2}^{2}\frac{(v-1/2)^{n-1}}{m^v}\,dv \\ &\quad+O\left(\frac{h}{(\log T)^{2-2\gamma}}\int_{1/2}^{2}\left(v-\frac{1}{2}\right)^{n-1}(1+\lvert\log(2v-1)\rvert)\,dv\right) \\ &=-\sum_{m\ge2}\frac{\Lambda(m)}{(\log m)m^{it}}\left(2\pi i a_1(m)-\frac{\pi\varepsilon}{2}(-i)^{n-1}a_2(m)\right)\int_{1/2}^{2}\frac{(v-1/2)^{n-1}}{m^v}\,dv \\ &\quad+O\left(\frac{h}{(\log T)^{2-2\gamma}}\right). \tag*{(3)} \end{aligned}

From the argument presented in [9 p. 250], the inner integral contained in the main term on the right-hand side of (3) satisfies, for each m≥2m\ge2,

∫1/22(v−1/2)n−1mv dv=(n−1)!(log⁡m)nm+O(1m2log⁡m).\int_{1/2}^{2}\frac{(v-1/2)^{n-1}}{m^v}\,dv=\frac{(n-1)!}{(\log m)^n\sqrt{m}}+O\left(\frac{1}{m^2\log m}\right).

Combining this estimate with (3), we get

∫−(log⁡T)3(log⁡T)3(∫1/22(v−12)n−1Δhlog⁡ζ(v+i(t+u)) dv)(sin⁡(γulog⁡2T)u)2K(u) du=−2π(n−1)!iA1(n,t)+πε2(n−1)!(−i)n−1A2(n,t)+On(hlog⁡2T).\begin{aligned} &\int_{-(\log T)^3}^{(\log T)^3}\left(\int_{1/2}^{2}\left(v-\frac{1}{2}\right)^{n-1}\Delta_h\log\zeta(v+i(t+u))\,dv\right)\left(\frac{\sin(\gamma u\log_2 T)}{u}\right)^2K(u)\,du \\ &=-2\pi(n-1)!iA_1(n,t)+\frac{\pi\varepsilon}{2}(n-1)!(-i)^{n-1}A_2(n,t)+O_n(h\log_2 T). \end{aligned}

Here we also use the following trivial upper bound that, for j=1,2j=1,2,

∣aj(m)∣≪hlog⁡mlog⁡2T.(4)\lvert a_j(m)\rvert\ll h\log m\log_2 T. \tag*{(4)}

Multiplying by the factor in/(π(n−1)!)i^n/(\pi(n-1)!) and taking the imaginary part of both sides, we complete the proof of Proposition 2.3. □

The resonator and some estimates

We now construct the required resonator. Indeed, the resonator we use here coincides with the one chosen in [8], both following the idea of [1, 2, 7]. For completeness, we recall its construction and the estimates needed in our proof.

Fix β∈(0,1)\beta\in(0,1) and set κ=(1−β)/2\kappa=(1-\beta)/2, so that κ+β<1\kappa+\beta<1. For sufficiently large TT, set

N=⌊Tκ⌋.N=\lfloor T^\kappa\rfloor.

Let P\mathcal{P} denote the set of primes pp satisfying

elog⁡Nlog⁡2N<p≤exp⁡((log⁡2N)1/8)log⁡Nlog⁡2N.e\log N\log_2 N<p\leq\exp\left((\log_2 N)^{1/8}\right)\log N\log_2 N.

Next, we define the multiplicative function f(n)f(n) to be supported on the set of square-free numbers, with values for p∈Pp\in\mathcal{P} as

f(p)=log⁡Nlog⁡2Nlog⁡3N1p(log⁡p−log⁡2N−log⁡3N).f(p)=\sqrt{\frac{\log N\log_2 N}{\log_3 N}}\frac{1}{\sqrt{p}(\log p-\log_2 N-\log_3 N)}.

Furthermore, if p∉Pp\notin\mathcal{P}, set f(p)=0f(p)=0.

For k∈{1,…,⌊(log⁡2N)1/8⌋}k\in\{1,\ldots,\lfloor(\log_2 N)^{1/8}\rfloor\}, we define the set Pk\mathcal{P}_k of prime numbers as follows

Pk:={p:eklog⁡Nlog⁡2N<p≤ek+1log⁡Nlog⁡2N}\mathcal{P}_k:=\{p:e^k\log N\log_2 N<p\leq e^{k+1}\log N\log_2 N\}

and

Mk:={n∈supp⁡(f):n has at least 3log⁡Nk2log⁡3N prime divisors in Pk}.\mathcal{M}_k:=\left\{n\in\operatorname{supp}(f):n\text{ has at least }\frac{3\log N}{k^2\log_3 N}\text{ prime divisors in }\mathcal{P}_k\right\}.

Furthermore, define

M:=supp⁡(f)∖⋃k=1⌊(log⁡2N)1/8⌋Mk.\mathcal{M}:=\operatorname{supp}(f)\setminus\bigcup_{k=1}^{\lfloor(\log_2 N)^{1/8}\rfloor}\mathcal{M}_k.

Let L\mathcal{L} be the set of integers ℓ\ell such that

[(1+1T)ℓ,(1+1T)ℓ+1]∩M≠∅.\left[\left(1+\frac{1}{T}\right)^\ell,\left(1+\frac{1}{T}\right)^{\ell+1}\right]\cap\mathcal{M}\ne\varnothing.

Then, we let mℓm_\ell be the minimum of [(1+T−1)ℓ,(1+T−1)ℓ+1]∩M\left[(1+T^{-1})^\ell,(1+T^{-1})^{\ell+1}\right]\cap\mathcal{M} for ℓ∈L\ell\in\mathcal{L}. Define the set

M′:={mℓ:ℓ∈L}.\mathcal{M}':=\{m_\ell:\ell\in\mathcal{L}\}.

and define

r(mℓ):=(∑(1−T−1)ℓ−1≤n≤(1+T−1)ℓ+2n∈Mf(n)2)1/2r(m_\ell):=\left(\sum_{\substack{(1-T^{-1})^{\ell-1}\leq n\leq(1+T^{-1})^{\ell+2}\\n\in\mathcal{M}}}f(n)^2\right)^{1/2}

for every mℓ∈M′m_\ell\in\mathcal{M}'. Moreover, the resonator is defined by

R(t):=∑m∈M′r(m)m−it.R(t):=\sum_{m\in\mathcal{M}'}r(m)m^{-it}.

We conclude this section with some estimates. The Cauchy-Schwarz inequality implies that

∣R(t)∣2≤R(0)2≪Tκ∑k∈Mf(k)2.(5)|R(t)|^2\leq R(0)^2\ll T^\kappa\sum_{k\in\mathcal{M}}f(k)^2. \tag*{(5)}

Furthermore, set Φ(y):=e−y2/2\Phi(y) := e^{-y^2/2} as in [1]. It follows from [2 Lemma 5] (also see [8 Proposition 4]) that

∫−∞∞∣R(t)∣2Φ(tT) dt≪T∑k∈Mf(k)2.(6)\int_{-\infty}^{\infty} |R(t)|^2\Phi\left(\frac{t}{T}\right)\,\mathrm{d}t \ll T\sum_{k\in\mathcal{M}} f(k)^2. \tag*{(6)}

We will make frequent use of (5) and (6) in the subsequent argument. Finally, owing to the work of Bondarenko and Seip [2], we state the following lemma:

Lemma 2.4 (Bondarenko-Seip [2], Lemma 7). If

G(t)=∑m≥2Λ(m)um(log⁡m)m1/2+itG(t)=\sum_{m\ge2}\frac{\Lambda(m)u_m}{(\log m)m^{1/2+it}}

is absolutely convergent and um≥0u_m\ge0 for each m≥2m\ge2, then we have

∫−∞∞G(t)∣R(t)∣2Φ(tT) dt≫T(log⁡T)1/2(log⁡3T)1/2(log⁡2T)1/2(min⁡p∈Pup)∑k∈Mf(k)2.\int_{-\infty}^{\infty}G(t)|R(t)|^2\Phi\left(\frac{t}{T}\right)\,\mathrm{d}t \gg T\frac{(\log T)^{1/2}(\log_3 T)^{1/2}}{(\log_2 T)^{1/2}}\left(\min_{p\in\mathcal{P}}u_p\right)\sum_{k\in\mathcal{M}}f(k)^2.

Proof of Theorem 1.1

In this section, we combine the resonance method with Proposition 2.3 to prove Theorem 1.1. The case h=0h=0 is trivial, so we assume that h∈(0,(2log⁡2T)−1)h\in(0,(2\log_2 T)^{-1}) throughout the proof. Set γ=1/8\gamma=1/8. First, by Proposition 2.3, we have

∫TβTlog⁡T(∫−(log⁡T)3(log⁡T)3ΔhSn(t+u)(sin⁡(γulog⁡2T)u)2K(u) du)∣R(t)∣2Φ(tT) dt=2Im⁡(in+3∑m≥2Λ(m)a1(m)(log⁡m)n+1m(∫TβTlog⁡Tm−it∣R(t)∣2Φ(tT) dt))+ε2Re⁡(∑m≥2Λ(m)a2(m)(log⁡m)n+1m(∫TβTlog⁡Tm−it∣R(t)∣2Φ(tT) dt))+O(hTlog⁡2T∑k∈Mf(k)2).(7)\begin{aligned} &\int_{T^\beta}^{T\log T}\left(\int_{-(\log T)^3}^{(\log T)^3}\Delta_hS_n(t+u)\left(\frac{\sin(\gamma u\log_2 T)}{u}\right)^2K(u)\,\mathrm{d}u\right)|R(t)|^2\Phi\left(\frac{t}{T}\right)\,\mathrm{d}t\\ &=2\operatorname{Im}\left(i^{n+3}\sum_{m\ge2}\frac{\Lambda(m)a_1(m)}{(\log m)^{n+1}\sqrt{m}}\left(\int_{T^\beta}^{T\log T}m^{-it}|R(t)|^2\Phi\left(\frac{t}{T}\right)\,\mathrm{d}t\right)\right)\\ &\quad+\frac{\varepsilon}{2}\operatorname{Re}\left(\sum_{m\ge2}\frac{\Lambda(m)a_2(m)}{(\log m)^{n+1}\sqrt{m}}\left(\int_{T^\beta}^{T\log T}m^{-it}|R(t)|^2\Phi\left(\frac{t}{T}\right)\,\mathrm{d}t\right)\right)\\ &\quad+O\left(hT\log_2T\sum_{k\in\mathcal{M}}f(k)^2\right). \tag*{(7)} \end{aligned}

Here, in the last step, we make use of (6) and the fact that both ∣R(t)∣2|R(t)|^2 and Φ(t/T)\Phi(t/T) are non-negative.

We first consider the second term on the right-hand side of (7), namely the one associated with a2(m)a_2(m), as this contributes the main term. To this end, we wish to expand the integral over R\mathbb{R}, so that Lemma 2.4 can be applied. By the definition of ωm\omega_m, a2(m)=0a_2(m)=0 unless (log⁡T)3/4≤m≤(log⁡T)5/4(\log T)^{3/4}\le m\le(\log T)^{5/4}. Thus, combining this with (4) and (5), we obtain

∣∑m≥2Λ(m)a2(m)(log⁡m)n+1m(∫0Tβm−it∣R(t)∣2Φ(tT) dt)∣≪hTκ+βlog⁡2T∑k∈Mf(k)2∑2≤m≤(log⁡T)5/4Λ(m)(log⁡m)n+1m.(8)\begin{aligned} \left|\sum_{m\ge2}\frac{\Lambda(m)a_2(m)}{(\log m)^{n+1}\sqrt{m}}\left(\int_0^{T^\beta}m^{-it}|R(t)|^2\Phi\left(\frac{t}{T}\right)\,\mathrm{d}t\right)\right| \\ &\ll hT^{\kappa+\beta}\log_{2} T\sum_{k\in\mathcal{M}} f(k)^2\sum_{2\leq m\leq(\log T)^{5/4}}\frac{\Lambda(m)}{(\log m)^{n+1}\sqrt{m}}. \tag*{(8)} \end{aligned}

By using [6 Lemma B.1], we have

∑2≤m≤(log⁡T)5/4Λ(m)(log⁡m)n+1m≪(log⁡T)5/8(log⁡2T)n.\sum_{2\leq m\leq(\log T)^{5/4}}\frac{\Lambda(m)}{(\log m)^{n+1}\sqrt{m}}\ll\frac{(\log T)^{5/8}}{(\log_{2}T)^n}.

Since κ+β<1\kappa+\beta<1, substituting this upper bound into (8) implies that

∣∑m≥2Λ(m)a2(m)(log⁡m)n+1m(∫0Tβm−it∣R(t)∣2Φ(tT)dt)∣≪hTκ+β(log⁡T)5/8(log⁡2T)n−1∑k∈Mf(k)2≪hT∑k∈Mf(k)2.(9)\left|\sum_{m\geq2}\frac{\Lambda(m)a_{2}(m)}{(\log m)^{n+1}\sqrt{m}}\left(\int_{0}^{T^{\beta}}m^{-it}|R(t)|^{2}\Phi\left(\frac{t}{T}\right)dt\right)\right| \ll hT^{\kappa+\beta}\frac{(\log T)^{5/8}}{(\log_{2}T)^{n-1}}\sum_{k\in\mathcal{M}}f(k)^2\ll hT\sum_{k\in\mathcal{M}}f(k)^2. \tag*{(9)}

Moreover, the exponential decay of Φ\Phi yields the following crude upper bound:

∣∑m≥2Λ(m)a2(m)(log⁡m)n+1m(∫Tlog⁡T∞m−it∣R(t)∣2Φ(tT)dt)∣≪hT∑k∈Mf(k)2.(10)\left|\sum_{m\geq2}\frac{\Lambda(m)a_{2}(m)}{(\log m)^{n+1}\sqrt{m}}\left(\int_{T\log T}^{\infty}m^{-it}|R(t)|^{2}\Phi\left(\frac{t}{T}\right)dt\right)\right|\ll hT\sum_{k\in\mathcal{M}}f(k)^2. \tag*{(10)}

On the other hand, we seek an upper bound for the first term on the right-hand side of (7). By the definition of ωm\omega_m, we have a1(m)=0a_1(m)=0 whenever m>(log⁡T)1/4m>(\log T)^{1/4}. Hence, the sum runs over 2≤m≤(log⁡T)1/42\leq m\leq(\log T)^{1/4}. Together with (4) and (6), this gives

∣∑m≥2Λ(m)a1(m)(log⁡m)n+1m(∫TβTlog⁡Tm−it∣R(t)∣2Φ(tT)dt)∣≤∣hlog⁡2T∑2≤m≤(log⁡T)1/4Λ(m)(log⁡m)nm∫−∞∞∣R(t)∣2Φ(tT)dt∣≪hTlog⁡2T∑2≤m≤(log⁡T)1/4Λ(m)(log⁡m)nm∑k∈Mf(k)2.(11)\begin{aligned} \left|\sum_{m\geq2}\frac{\Lambda(m)a_1(m)}{(\log m)^{n+1}\sqrt{m}}\left(\int_{T^\beta}^{T\log T}m^{-it}|R(t)|^{2}\Phi\left(\frac{t}{T}\right)dt\right)\right| &\leq\left|h\log_{2}T\sum_{2\leq m\leq(\log T)^{1/4}}\frac{\Lambda(m)}{(\log m)^n\sqrt{m}}\int_{-\infty}^{\infty}|R(t)|^{2}\Phi\left(\frac{t}{T}\right)dt\right|\\ &\ll hT\log_{2}T\sum_{2\leq m\leq(\log T)^{1/4}}\frac{\Lambda(m)}{(\log m)^n\sqrt{m}}\sum_{k\in\mathcal{M}}f(k)^2. \tag*{(11)} \end{aligned}

By using [6 Lemma B.1] again, we have

∑2≤m≤(log⁡T)1/4Λ(m)(log⁡m)nm≪(log⁡T)1/8(log⁡2T)n.\sum_{2\leq m\leq(\log T)^{1/4}}\frac{\Lambda(m)}{(\log m)^n\sqrt{m}}\ll\frac{(\log T)^{1/8}}{(\log_{2}T)^n}.

Thus, inserting this into (11) yields that

∣∑m≥2Λ(m)a1(m)(log⁡m)n+1m(∫TβTlog⁡Tm−it∣R(t)∣2Φ(tT)dt)∣≪hT(log⁡T)1/8(log⁡2T)n−1∑k∈Mf(k)2.(12)\left|\sum_{m\geq2}\frac{\Lambda(m)a_1(m)}{(\log m)^{n+1}\sqrt{m}}\left(\int_{T^\beta}^{T\log T}m^{-it}|R(t)|^{2}\Phi\left(\frac{t}{T}\right)dt\right)\right|\ll\frac{hT(\log T)^{1/8}}{(\log_{2}T)^{n-1}}\sum_{k\in\mathcal{M}}f(k)^2. \tag*{(12)}

To simplify the notation, we write

Z=(max⁡Tβ/2≤t≤2Tlog⁡TεΔhSn(t)),Z+=max⁡{0,Z}.\mathcal{Z}=\left(\max_{T^\beta/2\leq t\leq2T\log T}\varepsilon\Delta_hS_n(t)\right),\qquad\mathcal{Z}_{+}=\max\{0,\mathcal{Z}\}.

We shall prove that Z+>0\mathcal{Z}_{+}>0 for sufficiently large TT, which will imply that Z=Z+\mathcal{Z}=\mathcal{Z}_{+}. Since

Tβ−(log⁡T)3≥Tβ2,Tlog⁡T+(log⁡T)3≤2Tlog⁡TT^{\beta}-(\log T)^3\ge\frac{T^{\beta}}{2},\qquad T\log T+(\log T)^3\le2T\log T

for sufficiently large TT, we have

εΔhSn(t+u)≤Z+\varepsilon\Delta_hS_n(t+u)\le\mathcal{Z}_{+}

whenever Tβ≤t≤Tlog⁡TT^{\beta}\le t\le T\log T and ∣u∣≤(log⁡T)3|u|\le(\log T)^3. Using 1≤K(u)≤31\le K(u)\le3, together with the classical identity

∫−∞∞(sin⁡(au)u)2du=πa,\int_{-\infty}^{\infty}\left(\frac{\sin(au)}{u}\right)^2du=\pi a,

we obtain

ε∫TβTlog⁡T(∫−(log⁡T)3(log⁡T)3ΔhSn(t+u)(sin⁡(ulog⁡2T/8)u)2K(u) du)∣R(t)∣2Φ(tT) dt≤Z+∫−(log⁡T)3(log⁡T)3(sin⁡(ulog⁡2T/8)u)2K(u) du∫TβTlog⁡T∣R(t)∣2Φ(tT) dt≤3π8Z+log⁡2T∫−∞∞∣R(t)∣2Φ(tT) dt≪Z+Tlog⁡2T∑k∈Mf(k)2.(13)\begin{aligned} \varepsilon\int_{T^\beta}^{T\log T}\left(\int_{-(\log T)^3}^{(\log T)^3}\Delta_hS_n(t+u)\left(\frac{\sin(u\log_2T/8)}{u}\right)^2K(u)\,du\right)|R(t)|^2\Phi\left(\frac{t}{T}\right)\,dt\\ \le\mathcal{Z}_{+}\int_{-(\log T)^3}^{(\log T)^3}\left(\frac{\sin(u\log_2T/8)}{u}\right)^2K(u)\,du\int_{T^\beta}^{T\log T}|R(t)|^2\Phi\left(\frac{t}{T}\right)\,dt\\ \le\frac{3\pi}{8}\mathcal{Z}_{+}\log_2T\int_{-\infty}^{\infty}|R(t)|^2\Phi\left(\frac{t}{T}\right)\,dt\ll\mathcal{Z}_{+}T\log_2T\sum_{k\in\mathcal{M}}f(k)^2. \tag*{(13)} \end{aligned}

Here, in the last step, we apply (6). Noting that both ∣R(t)∣2|R(t)|^2 and Φ(t/T)\Phi(t/T) are real and even, we have

Re⁡(∫0∞m−it∣R(t)∣2Φ(tT) dt)=12∫−∞∞m−it∣R(t)∣2Φ(tT) dt.(14)\operatorname{Re}\left(\int_0^\infty m^{-it}|R(t)|^2\Phi\left(\frac{t}{T}\right)\,dt\right)=\frac{1}{2}\int_{-\infty}^{\infty}m^{-it}|R(t)|^2\Phi\left(\frac{t}{T}\right)\,dt. \tag*{(14)}

Multiplying (7) by ε\varepsilon, and then using (9), (10), (12), (13) and (14), we obtain

Z+Tlog⁡2T∑k∈Mf(k)2≫14∑m≥2Λ(m)a2(m)(log⁡m)n+1m∫−∞∞m−it∣R(t)∣2Φ(tT) dt+O(hT(log⁡T)1/8log⁡2T∑k∈Mf(k)2).(15)\begin{aligned} \mathcal{Z}_{+}T\log_2T\sum_{k\in\mathcal{M}}f(k)^2 \gg\frac{1}{4}\sum_{m\ge2}\frac{\Lambda(m)a_2(m)}{(\log m)^{n+1}\sqrt{m}}\int_{-\infty}^{\infty}m^{-it}|R(t)|^2\Phi\left(\frac{t}{T}\right)\,dt\\ +O\left(hT(\log T)^{1/8}\log_2T\sum_{k\in\mathcal{M}}f(k)^2\right). \tag*{(15)} \end{aligned}

We next show that the main term on the right-hand side of (15) is positive. Since we assume h≤(log⁡2T)−1h\le(\log_2T)^{-1}, it follows that

sin⁡(hlog⁡m)≫hlog⁡m\sin(h\log m)\gg h\log m

whenever m≤(log⁡T)2m\le(\log T)^2. Moreover, by the definition of ωm\omega_m,

ωm(18log⁡2T,log⁡2T)≫log⁡m\omega_m\left(\frac{1}{8}\log_2T,\log_2T\right)\gg\log m

holds for all m∈[(log⁡T)7/8,(log⁡T)9/8]m\in[(\log T)^{7/8},(\log T)^{9/8}]. Hence, when m∈[(log⁡T)7/8,(log⁡T)9/8]m\in[(\log T)^{7/8},(\log T)^{9/8}],

a2(m)≫h(log⁡m)2.a_2(m)\gg h(\log m)^2.

Since N=⌊Tκ⌋N=\lfloor T^\kappa\rfloor, we deduce that

(log⁡T)7/8≤p≤(log⁡T)9/8(\log T)^{7/8}\le p\le(\log T)^{9/8}

for each p∈Pp\in\mathcal{P} when TT is sufficiently large. Thus, we get a2(p)≫h(log⁡p)2a_2(p)\gg h(\log p)^2 and

min⁡p∈Pa2(p)(log⁡p)n≫min⁡p∈Ph(log⁡p)n−2≫h(log⁡2T)n−2.\min_{p\in\mathcal{P}}\frac{a_2(p)}{(\log p)^n}\gg\min_{p\in\mathcal{P}}\frac{h}{(\log p)^{n-2}}\gg\frac{h}{(\log_2 T)^{n-2}}.

Then, setting um=a2(m)/(log⁡m)nu_m=a_2(m)/(\log m)^n in Lemma 2.4 yields

∑m≥2Λ(m)a2(m)(log⁡m)n+1m∫−∞∞m−it∣R(t)∣2Φ(tT) dt≫hT(log⁡T)1/2(log⁡3T)1/2(log⁡2T)n−3/2∑k∈Mf(k)2.(16)\begin{aligned} \sum_{m\ge2}\frac{\Lambda(m)a_2(m)}{(\log m)^{n+1}\sqrt{m}}\int_{-\infty}^{\infty}m^{-it}|R(t)|^2\Phi\left(\frac{t}{T}\right)\,\mathrm{d}t \\ &\gg hT\frac{(\log T)^{1/2}(\log_3 T)^{1/2}}{(\log_2 T)^{n-3/2}}\sum_{k\in\mathcal{M}}f(k)^2. \tag*{(16)} \end{aligned}

Combining (15) with (16), we observe that

hT(log⁡T)1/8log⁡2ThT(log⁡T)1/2(log⁡3T)1/2(log⁡2T)3/2−n=(log⁡2T)n−1/2(log⁡T)3/8(log⁡3T)1/2=o(1),\frac{hT(\log T)^{1/8}\log_2 T}{hT(\log T)^{1/2}(\log_3 T)^{1/2}(\log_2 T)^{3/2-n}}=\frac{(\log_2 T)^{n-1/2}}{(\log T)^{3/8}(\log_3 T)^{1/2}}=o(1),

since nn is fixed. Thus, for all sufficiently large TT, (15) and (16) imply

Z+≫n,βh(log⁡T)1/2(log⁡3T)1/2(log⁡2T)n−1/2>0.\mathcal{Z}_+\gg_{n,\beta}h\frac{(\log T)^{1/2}(\log_3 T)^{1/2}}{(\log_2 T)^{n-1/2}}>0.

Consequently, Z+=Z\mathcal{Z}_+=\mathcal{Z}, and hence

max⁡Tβ/2≤t≤2Tlog⁡TεΔhSn(t)≫n,βh(log⁡T)1/2(log⁡3T)1/2(log⁡2T)n−1/2.\max_{T^\beta/2\le t\le2T\log T}\varepsilon\Delta_hS_n(t)\gg_{n,\beta}h\frac{(\log T)^{1/2}(\log_3 T)^{1/2}}{(\log_2 T)^{n-1/2}}.

Finally, we adjust the parameters to obtain the desired result. Recall that

Δh/2Sn(v)=Sn(v+h/2)−Sn(v−h/2).\Delta_{h/2}S_n(v)=S_n(v+h/2)-S_n(v-h/2).

Putting t=v−h/2t=v-h/2, we have

Δh/2Sn(v)=Sn(t+h)−Sn(t).\Delta_{h/2}S_n(v)=S_n(t+h)-S_n(t).

Now, fix 0<β<β1<10<\beta<\beta_1<1 and put T1=T/(4log⁡T)T_1=T/(4\log T). We apply the preceding estimate with TT replaced by T1T_1, β\beta replaced by β1\beta_1, and hh replaced by h/2h/2. Since 0<h≤(log⁡2T)−10<h\le(\log_2 T)^{-1}, we have

0<h2≤12log⁡2T10<\frac{h}{2}\le\frac{1}{2\log_2 T_1}

for sufficiently large TT. Thus,

max⁡T1β1/2≤v≤2T1log⁡T1εΔh/2Sn(v)≫n,β1h(log⁡T1)1/2(log⁡3T1)1/2(log⁡2T1)n−1/2.\max_{T_1^{\beta_1}/2\le v\le2T_1\log T_1}\varepsilon\Delta_{h/2}S_n(v)\gg_{n,\beta_1}h\frac{(\log T_1)^{1/2}(\log_3 T_1)^{1/2}}{(\log_2 T_1)^{n-1/2}}.

Moreover, since β1>β\beta_1>\beta, we have, for all sufficiently large TT,

[T1β12−h2,2T1log⁡T1−h2]⊆[Tβ,T].\left[\frac{T_1^{\beta_1}}{2}-\frac{h}{2},2T_1\log T_1-\frac{h}{2}\right]\subseteq[T^\beta,T].

Consequently,

max⁡Tβ≤t≤Tε(Sn(t+h)−Sn(t))≫n,βh(log⁡T)1/2(log⁡3T)1/2(log⁡2T)n−1/2.\max_{T^\beta\le t\le T}\varepsilon\left(S_n(t+h)-S_n(t)\right)\gg_{n,\beta}h\frac{(\log T)^{1/2}(\log_3 T)^{1/2}}{(\log_2 T)^{n-1/2}}.

where we used log⁡jT1∼log⁡jT\log_j T_1\sim\log_j T for j=1,2,3j=1,2,3. Since the estimate holds for each ε∈{−1,1}\varepsilon\in\{-1,1\}, the proof of Theorem 1.1 is complete.

Proof of Corollary 1.2

Set β′=(1+β)/2\beta'=(1+\beta)/2. Applying Theorem 1.1 with nn replaced by n+1n+1, TT replaced by T/2T/2, we choose h=(log⁡2(T/2))−1h=(\log_2(T/2))^{-1}, which lies in the admissible range of the theorem. Then, we obtain

max⁡(T/2)β′≤t≤T/2±(Sn+1(t+h)−Sn+1(t))≫n,βh(log⁡T)1/2(log⁡3T)1/2(log⁡2T)n+1/2.\max_{(T/2)^{\beta'}\le t\le T/2}\pm\left(S_{n+1}(t+h)-S_{n+1}(t)\right)\gg_{n,\beta}h\frac{(\log T)^{1/2}(\log_3 T)^{1/2}}{(\log_2 T)^{n+1/2}}.

Since

Sn+1(t+h)−Sn+1(t)=∫tt+hSn(τ) dτ,S_{n+1}(t+h)-S_{n+1}(t)=\int_t^{t+h}S_n(\tau)\,\mathrm{d}\tau,

it follows that

max⁡t≤τ≤t+h±Sn(τ)≥±1h(Sn+1(t+h)−Sn+1(t)).\max_{t\le\tau\le t+h}\pm S_n(\tau)\ge\pm\frac{1}{h}\left(S_{n+1}(t+h)-S_{n+1}(t)\right).

Since β′>β\beta'>\beta, we have (T/2)β′≥Tβ(T/2)^{\beta'}\ge T^\beta and T/2+h≤TT/2+h\le T for all sufficiently large TT. Hence, [t,t+h]⊆[Tβ,T][t,t+h]\subseteq[T^\beta,T], which completes the proof of Corollary 1.2.

AI Declaration

The authors used ChatGPT solely for language revision in this work and assume full responsibility for the mathematical arguments, results, and final text.

References

  1. [1]A. Bondarenko and K. Seip, Large greatest common divisor sums and extreme values of the Riemann zeta function, Duke Math. J., 166 (2017), pp. 1685–1701.
  2. [2]A. Bondarenko and K. Seip, Extreme values of the Riemann zeta function and its argument, Math. Ann., 372 (2018), pp. 999–1015.DOI
  3. [3]H. M. Bui, S. J. Lester, and M. B. Milinovich, On Balazard, Saias, and Yor’s equivalence to the Riemann Hypothesis, J. Math. Anal. Appl., 409 (2014), pp. 244–253.DOI
  4. [4]E. Carneiro, V. Chandee, and M. B. Milinovich, Bounding S(t) and S₁(t) on the Riemann hypothesis, Math. Ann., 356 (2013), pp. 939–968.
  5. [5]E. Carneiro and A. Chirre, Bounding Sₙ(t) on the Riemann hypothesis, Math. Proc. Camb. Philos. Soc., 164 (2018), pp. 259–283.DOI
  6. [6]E. Carneiro, A. Chirre, and M. B. Milinovich, Bandlimited approximations and estimates for the Riemann zeta-function, Publ. Mat., 63 (2019), pp. 601–661.DOI
  7. [7]A. Chirre, Extreme values for Sₙ(σ,t) near the critical line, J. Number Theory, 200 (2019), pp. 329–352.
  8. [8]A. Chirre and K. Mahatab, Large oscillations of the argument of the Riemann zeta-function, Bull. Lond. Math. Soc., 53 (2021), pp. 1776–1785.arxiv.org/abs/1904.11051
  9. [9]A. Chirre and K. Mahatab, Large values of the argument of the Riemann zeta-function and its iterates, J. Number Theory, 225 (2021), pp. 240–259.arxiv.org/abs/2006.04288
  10. [10]D. W. Farmer, S. M. Gonek, and C. P. Hughes, The maximum size of L-functions, J. reine angew. Math., 609 (2007), pp. 215–236.DOI
  11. [11]A. Fujii, On the zeros of the Riemann zeta function, Comment. Math. Univ. St. Pauli, 51 (2002), pp. 1–17.DOI
  12. [12]S. Inoue, Extreme values for iterated integrals of the logarithm of the Riemann zeta-function, Acta Arith., 205 (2022), pp. 97–119.arxiv.org/abs/2009.04099
  13. [13]S. Inoue, H. Kobayashi, and Y. Toma, Explicit extreme values of the argument of the Riemann zeta-function, Math. Proc. Camb. Philos. Soc., 181 (2026), pp. 1149–1163.
  14. [14]M. Jaskari, Applying the resonance method to Re(e^{-iθ} log ζ(σ + it)), Funct. Approx. Comment. Math., 70 (2024), pp. 263–275.DOI
  15. [15]H. L. Montgomery, Extreme values of the Riemann zeta function, Comment. Math. Helv., 52 (1977), pp. 511–518.DOI
  16. [16]K.-M. Tsang, Some Ω-theorems for the Riemann zeta-function, Acta Arith., 46 (1986), pp. 369–395.DOI
  17. [17]K.-M. Tsang, An improved Ω₊ theorem for S₁(t), Q. J. Math., 42 (1991), pp. 501–513.DOI

Paper details

vlnc:W_ngma56fn
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC0-1.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents