a-points of the Riemann zeta function Youness Lamzouri (York University) Joint work with Steve Lester (Tel Aviv University) and Maksym Radziwill (Institute for Advanced Study) Number Theory at Illinois A Conference in Memory of the Batemans and Heini Halberstam June 5, 2014 Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 1 / 18 History of the subject and previous results Definition Let a be a nonzero complex number. The roots of ζ(s) = a are called a-points of ζ(s). They are denoted by ρa = βa + iγa . Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 2 / 18 History of the subject and previous results Definition Let a be a nonzero complex number. The roots of ζ(s) = a are called a-points of ζ(s). They are denoted by ρa = βa + iγa . For every a 6= 0 there is n0 (a) such that for all n ≥ n0 (a) there is an a-point of ζ(s) quite close to s = −2n, and there are at most finitely many other a-points in Re(s) ≤ 0. These are called the trivial a-points of ζ(s). Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 2 / 18 History of the subject and previous results Definition Let a be a nonzero complex number. The roots of ζ(s) = a are called a-points of ζ(s). They are denoted by ρa = βa + iγa . For every a 6= 0 there is n0 (a) such that for all n ≥ n0 (a) there is an a-point of ζ(s) quite close to s = −2n, and there are at most finitely many other a-points in Re(s) ≤ 0. These are called the trivial a-points of ζ(s). The remaining a-points all lie in a strip 0 < Re(s) < A, where A depends on a, and we call these nontrivial a-points. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 2 / 18 History of the subject and previous results Definition Let a be a nonzero complex number. The roots of ζ(s) = a are called a-points of ζ(s). They are denoted by ρa = βa + iγa . For every a 6= 0 there is n0 (a) such that for all n ≥ n0 (a) there is an a-point of ζ(s) quite close to s = −2n, and there are at most finitely many other a-points in Re(s) ≤ 0. These are called the trivial a-points of ζ(s). The remaining a-points all lie in a strip 0 < Re(s) < A, where A depends on a, and we call these nontrivial a-points. Let Na (T ) = X 1. 1<γa <T βa >0 Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 2 / 18 Riemann-von Mangoldt type formula T Na (T ) = log 2π T 2π − T log 2 − δ1 (a) T + Oa (log T ), 2π 2π where δ1 (a) = 1 if a = 1 and equals 0 otherwise. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 3 / 18 Riemann-von Mangoldt type formula T Na (T ) = log 2π T 2π − T log 2 − δ1 (a) T + Oa (log T ), 2π 2π where δ1 (a) = 1 if a = 1 and equals 0 otherwise. The number a = 1 is special, since ζ(σ + it) = 1 + O(2−σ ) for σ large. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 3 / 18 Riemann-von Mangoldt type formula T Na (T ) = log 2π T 2π − T log 2 − δ1 (a) T + Oa (log T ), 2π 2π where δ1 (a) = 1 if a = 1 and equals 0 otherwise. The number a = 1 is special, since ζ(σ + it) = 1 + O(2−σ ) for σ large. Theorem (Levinson, 1975) Almost all of the a-points are close to the critical line Re(s) = 1/2. More precisely, for any δ > 0 X 1 1 = Na (T ) 1 + Oδ . log T 1/2−δ<βa <1/2+δ 1<γa <T Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 3 / 18 The distribution of a-points in Re(s) ≤ 1/2 Conjecture (Selberg) For any a 6= 0, three quarters of the a-points of ζ(s) are to the left of the critical line Re(s) = 1/2. More precisely, X 3 + o(1) Na (T ). 1= 4 0<βa <1/2 1<γa <T Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 4 / 18 The distribution of a-points in Re(s) ≤ 1/2 Conjecture (Selberg) For any a 6= 0, three quarters of the a-points of ζ(s) are to the left of the critical line Re(s) = 1/2. More precisely, X 3 + o(1) Na (T ). 1= 4 0<βa <1/2 1<γa <T For any a 6= 0, there are at most only finitely many a-points on the critical line Re(s) = 1/2. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 4 / 18 The distribution of a-points in Re(s) ≤ 1/2 Conjecture (Selberg) For any a 6= 0, three quarters of the a-points of ζ(s) are to the left of the critical line Re(s) = 1/2. More precisely, X 3 + o(1) Na (T ). 1= 4 0<βa <1/2 1<γa <T For any a 6= 0, there are at most only finitely many a-points on the critical line Re(s) = 1/2. Theorem (Selberg, unpublished) Assume the Riemann Hypothesis. Then at least (1/2 + o(1))Na (T ) of the a-points of ζ(s) with ordinates in (1, T ) lie to the left of the critical line. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 4 / 18 The distribution of a-points in Re(s) > 1/2 Let a be a nonzero complex number and 1/2 < σ1 < σ2 . Define X Na (σ1 , σ2 ; T ) := 1. σ1 <βa <σ2 1<γa <T Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 5 / 18 The distribution of a-points in Re(s) > 1/2 Let a be a nonzero complex number and 1/2 < σ1 < σ2 . Define X Na (σ1 , σ2 ; T ) := 1. σ1 <βa <σ2 1<γa <T Theorem (Borchsenius and Jessen, 1948) For any 1/2 < σ1 < σ2 , we have Na (σ1 , σ2 ; T ) ∼ ca (σ1 , σ2 )T . Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 5 / 18 The distribution of a-points in Re(s) > 1/2 Let a be a nonzero complex number and 1/2 < σ1 < σ2 . Define X Na (σ1 , σ2 ; T ) := 1. σ1 <βa <σ2 1<γa <T Theorem (Borchsenius and Jessen, 1948) For any 1/2 < σ1 < σ2 , we have Na (σ1 , σ2 ; T ) ∼ ca (σ1 , σ2 )T . Theorem (Matsumoto, 1988) For any 1 ≤ σ1 < σ2 , we have Na (σ1 , σ2 ; T ) = ca (σ1 , σ2 )T + OA Youness Lamzouri (York) T (log log T )A a-points of the Riemann zeta function . June 5, 2014 5 / 18 Our results Na (σ1 , σ2 ; T ) = X 1. σ1 <βa <σ2 1<γa <T Theorem 1 (L, Lester and Radziwill, 2014) Let 1/2 < σ1 < σ2 < 1. We have Na (σ1 , σ2 ; T ) = ca (σ1 , σ2 )T + O Youness Lamzouri (York) T log log T (log T )σ1 /2 a-points of the Riemann zeta function . June 5, 2014 6 / 18 Our results Na (σ1 , σ2 ; T ) = X 1. σ1 <βa <σ2 1<γa <T Theorem 1 (L, Lester and Radziwill, 2014) Let 1/2 < σ1 < σ2 < 1. We have Na (σ1 , σ2 ; T ) = ca (σ1 , σ2 )T + O T log log T (log T )σ1 /2 . Remark Our method also gives an improved version of Matsumoto’s result for σ1 ≥ 1. Namely, for 1 ≤ σ1 < σ2 we have T log log T √ Na (σ1 , σ2 ; T ) = ca (σ1 , σ2 )T + O . log T Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 6 / 18 The work of Borchsenius and Jessen Na (σ1 , σ2 ; T ) := X 1. σ1 <βa <σ2 1<γa <T Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 7 / 18 The work of Borchsenius and Jessen Na (σ1 , σ2 ; T ) := X 1. σ1 <βa <σ2 1<γa <T Borchsenius and Jessen studied the zeros of general almost periodic functions and applied their results to the function ζ(s) − a. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 7 / 18 The work of Borchsenius and Jessen Na (σ1 , σ2 ; T ) := X 1. σ1 <βa <σ2 1<γa <T Borchsenius and Jessen studied the zeros of general almost periodic functions and applied their results to the function ζ(s) − a. They showed that the Jessen function Z 1 T fa (σ) = lim log |ζ(σ + it) − a|dt T →∞ T 1 exists and is twice differentiable in σ, for 1/2 < σ < 1. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 7 / 18 The work of Borchsenius and Jessen Na (σ1 , σ2 ; T ) := X 1. σ1 <βa <σ2 1<γa <T Borchsenius and Jessen studied the zeros of general almost periodic functions and applied their results to the function ζ(s) − a. They showed that the Jessen function Z 1 T fa (σ) = lim log |ζ(σ + it) − a|dt T →∞ T 1 exists and is twice differentiable in σ, for 1/2 < σ < 1. They proved Na (σ1 , σ2 ; T ) ∼ ca (σ1 , σ2 )T , where ca (σ1 , σ2 ) = Youness Lamzouri (York) fa0 (σ2 ) − fa0 (σ1 ) . 2π a-points of the Riemann zeta function June 5, 2014 7 / 18 Strategy of the proof of Theorem 1 Theorem 1 (L, Lester and Radziwill, 2014) Let 1/2 < σ1 < σ2 < 1. We have Na (σ1 , σ2 ; T ) = ca (σ1 , σ2 )T + O Youness Lamzouri (York) T log log T (log T )σ1 /2 a-points of the Riemann zeta function . June 5, 2014 8 / 18 Strategy of the proof of Theorem 1 Theorem 1 (L, Lester and Radziwill, 2014) Let 1/2 < σ1 < σ2 < 1. We have Na (σ1 , σ2 ; T ) = ca (σ1 , σ2 )T + O T log log T (log T )σ1 /2 . Littlewood’s Lemma Let R be the rectangle with vertices σ1 + i, σ2 + i, σ2 + iT , σ1 + iT . Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 8 / 18 Strategy of the proof of Theorem 1 Theorem 1 (L, Lester and Radziwill, 2014) Let 1/2 < σ1 < σ2 < 1. We have Na (σ1 , σ2 ; T ) = ca (σ1 , σ2 )T + O T log log T (log T )σ1 /2 . Littlewood’s Lemma Let R be the rectangle with vertices σ1 + i, σ2 + i, σ2 + iT , σ1 + iT . Let φ(s) be holomorphic inside and on the boundary of R. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 8 / 18 Strategy of the proof of Theorem 1 Theorem 1 (L, Lester and Radziwill, 2014) Let 1/2 < σ1 < σ2 < 1. We have Na (σ1 , σ2 ; T ) = ca (σ1 , σ2 )T + O T log log T (log T )σ1 /2 . Littlewood’s Lemma Let R be the rectangle with vertices σ1 + i, σ2 + i, σ2 + iT , σ1 + iT . Let φ(s) be holomorphic inside and on the boundary of R. Let ν(σ, T ) be the number of zeros of φ(s) inside the rectangle R and such that Re(s) > σ. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 8 / 18 Strategy of the proof of Theorem 1 Theorem 1 (L, Lester and Radziwill, 2014) Let 1/2 < σ1 < σ2 < 1. We have Na (σ1 , σ2 ; T ) = ca (σ1 , σ2 )T + O T log log T (log T )σ1 /2 . Littlewood’s Lemma Let R be the rectangle with vertices σ1 + i, σ2 + i, σ2 + iT , σ1 + iT . Let φ(s) be holomorphic inside and on the boundary of R. Let ν(σ, T ) be the number of zeros of φ(s) inside the rectangle R and such that Re(s) > σ. Z σ2 Z 1 ν(σ, T )dσ = − log φ(s)ds. 2πi ∂R σ1 Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 8 / 18 Na (σ, T ) := X 1. βa >σ 1<γa <T Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 9 / 18 Na (σ, T ) := X 1. βa >σ 1<γa <T Littlewood’s Lemma σ2 Z Na (σ, T )dσ σ1 Z T Z T 1 1 log |ζ(σ1 + it) − a|dt − log |ζ(σ2 + it) − a|dt 2π 1 2π 1 Z σ2 Z σ2 1 1 − arg(ζ(σ + i) − a)dσ + arg(ζ(σ + iT ) − a)dσ 2π σ1 2π σ1 = Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 9 / 18 Na (σ, T ) := X 1. βa >σ 1<γa <T Littlewood’s Lemma Z σ2 Na (σ, T )dσ σ1 Z T Z T 1 1 log |ζ(σ1 + it) − a|dt − log |ζ(σ2 + it) − a|dt 2π 1 2π 1 Z σ2 Z σ2 1 1 − arg(ζ(σ + i) − a)dσ + arg(ζ(σ + iT ) − a)dσ 2π σ1 2π σ1 Z T Z T 1 1 log |ζ(σ1 + it) − a|dt − log |ζ(σ2 + it) − a|dt = 2π 1 2π 1 +Oa (log T ). = Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 9 / 18 An asymptotic formula for RT 1 log |ζ(σ + it) − a|dt Borchsenius and Jessen Z T log |ζ(σ + it) − a|dt ∼ T · fa (σ). 1 Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 10 / 18 An asymptotic formula for RT 1 log |ζ(σ + it) − a|dt Borchsenius and Jessen Z T log |ζ(σ + it) − a|dt ∼ T · fa (σ). 1 How can we obtain a quantitative asymptotic formula? Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 10 / 18 An asymptotic formula for RT 1 log |ζ(σ + it) − a|dt Borchsenius and Jessen Z T log |ζ(σ + it) − a|dt ∼ T · fa (σ). 1 How can we obtain a quantitative asymptotic formula? Unlike most classical mean value problems where Dirichlet polynomial approximations are used, there is no simple expression of log |ζ(σ + it) − a| as a Dirichlet series for σ > A. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 10 / 18 An asymptotic formula for RT 1 log |ζ(σ + it) − a|dt Borchsenius and Jessen Z T log |ζ(σ + it) − a|dt ∼ T · fa (σ). 1 How can we obtain a quantitative asymptotic formula? Unlike most classical mean value problems where Dirichlet polynomial approximations are used, there is no simple expression of log |ζ(σ + it) − a| as a Dirichlet series for σ > A. Instead, we exploit the fact that ζ(s) can be very well modeled by a random Euler product for Re(s) > 1/2. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 10 / 18 A random model for the values log ζ(σ + it) Let {X (p)}p prime be a sequence of independent random variables uniformly distributed on the Unit circle U = {z ∈ C : |z| = 1}. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 11 / 18 A random model for the values log ζ(σ + it) Let {X (p)}p prime be a sequence of independent random variables uniformly distributed on the Unit circle U = {z ∈ C : |z| = 1}. Define Y X (p) −1 ζ(σ, X ) := 1− σ . p p Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 11 / 18 A random model for the values log ζ(σ + it) Let {X (p)}p prime be a sequence of independent random variables uniformly distributed on the Unit circle U = {z ∈ C : |z| = 1}. Define Y X (p) −1 ζ(σ, X ) := 1− σ . p p This random product converges almost surely for any σ > 1/2. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 11 / 18 A random model for the values log ζ(σ + it) Let {X (p)}p prime be a sequence of independent random variables uniformly distributed on the Unit circle U = {z ∈ C : |z| = 1}. Define Y X (p) −1 ζ(σ, X ) := 1− σ . p p This random product converges almost surely for any σ > 1/2. Bohr and Jessen (1930) For any σ > 1/2, the distribution of log ζ(σ + it) converges, as t → ∞, to the distribution of the random variable log ζ(σ, X ). Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 11 / 18 A random model for the values log ζ(σ + it) Let {X (p)}p prime be a sequence of independent random variables uniformly distributed on the Unit circle U = {z ∈ C : |z| = 1}. Define Y X (p) −1 ζ(σ, X ) := 1− σ . p p This random product converges almost surely for any σ > 1/2. Bohr and Jessen (1930) For any σ > 1/2, the distribution of log ζ(σ + it) converges, as t → ∞, to the distribution of the random variable log ζ(σ, X ). More precisely, for any rectangle R in C with sides parallel to the coordinate axes we have lim T →∞ 1 meas{t ∈ [1, T ] : log ζ(σ + it) ∈ R} = P (log ζ(σ, X ) ∈ R) . T Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 11 / 18 A discrepancy bound for the distribution of log ζ(σ + it) Let Dσ (T ) = sup PT log ζ(σ + it) ∈ R − P log ζ(σ, X ) ∈ R R where the supremum is taken aver all rectangles R in C with sides parallel to the coordinate axes and where PT (f (t) ∈ R) := Youness Lamzouri (York) 1 meas{t ∈ [1, T ] : f (t) ∈ R}. T a-points of the Riemann zeta function June 5, 2014 12 / 18 A discrepancy bound for the distribution of log ζ(σ + it) Let Dσ (T ) = sup PT log ζ(σ + it) ∈ R − P log ζ(σ, X ) ∈ R R where the supremum is taken aver all rectangles R in C with sides parallel to the coordinate axes and where PT (f (t) ∈ R) := 1 meas{t ∈ [1, T ] : f (t) ∈ R}. T Theorem 2 (L, Lester and Radziwill, 2014) Let 1/2 < σ < 1 be fixed. Then Dσ (T ) σ Youness Lamzouri (York) 1 . (log T )σ a-points of the Riemann zeta function June 5, 2014 12 / 18 We use a“two-dimensional integration by part” technique to relate RT the integral 1 log |ζ(σ + it) − a|dt to the discrepancy Dσ (T ). Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 13 / 18 We use a“two-dimensional integration by part” technique to relate RT the integral 1 log |ζ(σ + it) − a|dt to the discrepancy Dσ (T ). However, in order for our method to work the we need to control contribution of the points t for which log |ζ(σ + it) − a| is large. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 13 / 18 We use a“two-dimensional integration by part” technique to relate RT the integral 1 log |ζ(σ + it) − a|dt to the discrepancy Dσ (T ). However, in order for our method to work the we need to control contribution of the points t for which log |ζ(σ + it) − a| is large. To bound the contribution of those t with |ζ(σ + it)| large, we use bounds for the distribution of extreme values of log |ζ(σ + it)| (L, 2011). Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 13 / 18 We use a“two-dimensional integration by part” technique to relate RT the integral 1 log |ζ(σ + it) − a|dt to the discrepancy Dσ (T ). However, in order for our method to work the we need to control contribution of the points t for which log |ζ(σ + it) − a| is large. To bound the contribution of those t with |ζ(σ + it)| large, we use bounds for the distribution of extreme values of log |ζ(σ + it)| (L, 2011). To bound the contribution of those t such that ζ(σ + it) is very close to a we use the following L2k bound. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 13 / 18 We use a“two-dimensional integration by part” technique to relate RT the integral 1 log |ζ(σ + it) − a|dt to the discrepancy Dσ (T ). However, in order for our method to work the we need to control contribution of the points t for which log |ζ(σ + it) − a| is large. To bound the contribution of those t with |ζ(σ + it)| large, we use bounds for the distribution of extreme values of log |ζ(σ + it)| (L, 2011). To bound the contribution of those t such that ζ(σ + it) is very close to a we use the following L2k bound. Proposition (L, Lester and Radziwill) Let 1/2 < σ < 1. Then, there exists a constant C = C (a, σ), such that for any integer k ≥ 1 Z T (log |ζ(σ + it) − a|)2k dt T (Ck)4k . 1 Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 13 / 18 An asymptotic formula for quantitative error term RT 1 log |ζ(σ + it) − a|dt with a Theorem 3 (L, Lester and Radziwill, 2014) Let 1/2 < σ < 1 be fixed. Then Z T log |ζ(σ + it) − a|dt = T · E (log |ζ(σ, X ) − a|) + O 1 T (log log T )2 (log T )σ . In particular, fa (σ) = E (log |ζ(σ, X ) − a|). Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 14 / 18 An asymptotic formula for quantitative error term RT 1 log |ζ(σ + it) − a|dt with a Theorem 3 (L, Lester and Radziwill, 2014) Let 1/2 < σ < 1 be fixed. Then Z T log |ζ(σ + it) − a|dt = T · E (log |ζ(σ, X ) − a|) + O 1 T (log log T )2 (log T )σ . In particular, fa (σ) = E (log |ζ(σ, X ) − a|). Remark A part from the power of log log T , the error term is the best that we can obtain using our discrepancy bound. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 14 / 18 Completing the proof of Theorem 1 Recall that Na (σ, T ) = |{βa > σ, 1 < γa < T }|. Littlewood’s Lemma Z σ2 σ1 1 Na (σ, T )dσ = 2π − Youness Lamzouri (York) 1 2π Z T log |ζ(σ1 + it) − a|dt 1 Z T log |ζ(σ2 + it) − a|dt + Oa (log T ) . 1 a-points of the Riemann zeta function June 5, 2014 15 / 18 Completing the proof of Theorem 1 Recall that Na (σ, T ) = |{βa > σ, 1 < γa < T }|. Littlewood’s Lemma Z σ2 σ1 1 Na (σ, T )dσ = 2π − 1 2π Z T log |ζ(σ1 + it) − a|dt 1 Z T log |ζ(σ2 + it) − a|dt + Oa (log T ) . 1 Theorem 3 (L, Lester and Radziwill) Z T log |ζ(σ + it) − a|dt = T · fa (σ) + O 1 Youness Lamzouri (York) a-points of the Riemann zeta function T (log log T )2 (log T )σ . June 5, 2014 15 / 18 Completing the proof of Theorem 1 Recall that Na (σ, T ) = |{βa > σ, 1 < γa < T }|. Littlewood’s Lemma Z σ2 σ1 1 Na (σ, T )dσ = 2π − 1 2π Z T log |ζ(σ1 + it) − a|dt 1 Z T log |ζ(σ2 + it) − a|dt + Oa (log T ) . 1 Theorem 3 (L, Lester and Radziwill) Z T log |ζ(σ + it) − a|dt = T · fa (σ) + O 1 T (log log T )2 (log T )σ . Borchsenius and Jessen fa (σ) = E (log |ζ(σ, X ) − a|) is twice differentiable in σ, for 1/2 < σ < 1. Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 15 / 18 Recall that Na (σ, T ) = |{βa > σ, 1 < γa < T }|. 1 h Z σ σ+h T Na (σ, T )dσ = 2π Youness Lamzouri (York) fa (σ) − fa (σ + h) h a-points of the Riemann zeta function +O T (log log T )2 h(log T )σ June 5, 2014 16 / 18 Recall that Na (σ, T ) = |{βa > σ, 1 < γa < T }|. 1 h Z σ σ+h T fa (σ) − fa (σ + h) T (log log T )2 Na (σ, T )dσ = +O 2π h h(log T )σ T 0 T (log log T )2 = − fa (σ) + O hT + . 2π h(log T )σ Since fa (σ + h) = fa (σ) + hfa0 (σ) + Oa (h2 ). Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 16 / 18 Recall that Na (σ, T ) = |{βa > σ, 1 < γa < T }|. 1 h Z σ σ+h T fa (σ) − fa (σ + h) T (log log T )2 Na (σ, T )dσ = +O 2π h h(log T )σ T 0 T (log log T )2 = − fa (σ) + O hT + . 2π h(log T )σ Since fa (σ + h) = fa (σ) + hfa0 (σ) + Oa (h2 ). Finally using that Na (σ, T ) is decreasing in σ, and choosing h optimally, leads to Na (σ1 , σ2 ; T ) = X σ1 <βa <σ2 1<γa <T Youness Lamzouri (York) f 0 (σ2 ) − fa0 (σ1 ) 1= a T +O 2π a-points of the Riemann zeta function T log log T (log T )σ1 /2 June 5, 2014 . 16 / 18 Last remarks Our result Na (σ1 , σ2 ; T ) = ca (σ1 , σ2 )T + O T log log T (log T )σ1 /2 , Relies on the discrepancy bound 1 sup meas{t ∈ [1, T ] : log ζ(σ + it) ∈ R} − P (log ζ(σ, X ) ∈ R) R T 1 σ , (log T )σ Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 17 / 18 Last remarks Our result Na (σ1 , σ2 ; T ) = ca (σ1 , σ2 )T + O T log log T (log T )σ1 /2 , Relies on the discrepancy bound 1 sup meas{t ∈ [1, T ] : log ζ(σ + it) ∈ R} − P (log ζ(σ, X ) ∈ R) R T 1 σ , (log T )σ Any improvement in the discrepancy bound for the distribution of log ζ(σ + it) will lead to an improvement in the error term of the number of a-points of ζ(s). Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 17 / 18 Thank you for your attention ! Youness Lamzouri (York) a-points of the Riemann zeta function June 5, 2014 18 / 18