On the parity of k-th powers modulo p

Similar documents
Pseudorandom Sequences II: Exponential Sums and Uniform Distribution

Research Statement. Enrique Treviño. M<n N+M

SUM-PRODUCT ESTIMATES APPLIED TO WARING S PROBLEM MOD P

Repeated Values of Euler s Function. a talk by Paul Kinlaw on joint work with Jonathan Bayless

The Least Inert Prime in a Real Quadratic Field

Prime Divisors of Palindromes

Roth s Theorem on Arithmetic Progressions

Degrees of Regularity of Colorings of the Integers

RESEARCH STATEMENT Jennifer Paulhus

DECIMATIONS OF L-SEQUENCES AND PERMUTATIONS OF EVEN RESIDUES MOD P

MATH 310: Homework 7

SUM-PRODUCT ESTIMATES APPLIED TO WARING S PROBLEM OVER FINITE FIELDS

Explicit Bounds for the Burgess Inequality for Character Sums

Roots of Polynomials in Subgroups of F p and Applications to Congruences

Expansions of quadratic maps in prime fields

M381 Number Theory 2004 Page 1

A numerically explicit Burgess inequality and an application to qua

The Riddle of Primes

Math 259: Introduction to Analytic Number Theory Exponential sums IV: The Davenport-Erdős and Burgess bounds on short character sums

198 VOLUME 46/47, NUMBER 3

EXPONENTIAL SUMS EQUIDISTRIBUTION

EXPLICIT BOUNDS ON MONOMIAL AND BINOMIAL EXPONENTIAL SUMS

Analytic Number Theory

LARGE PRIME NUMBERS (32, 42; 4) (32, 24; 2) (32, 20; 1) ( 105, 20; 0).

Summary Slides for MATH 342 June 25, 2018

Cullen Numbers in Binary Recurrent Sequences

WARING S NUMBER IN FINITE FIELDS JAMES ARTHUR CIPRA. B.S., Kansas State University, 2000 M.S., Kansas State University, 2004

Congruent Number Problem and Elliptic curves

Analytic. Number Theory. Exploring the Anatomy of Integers. Jean-Marie. De Koninck. Florian Luca. ffk li? Graduate Studies.

Erdős and arithmetic progressions

Decoupling Lecture 1

Fixed points for discrete logarithms

EXAMPLES OF MORDELL S EQUATION

Gauss periods as low complexity normal bases

On the Fractional Parts of a n /n

(Primes and) Squares modulo p

Maximal Class Numbers of CM Number Fields

A conjecture of Evans on Kloosterman sums

The distribution of consecutive prime biases

Proof of Infinite Number of Triplet Primes. Stephen Marshall. 28 May Abstract

Research Problems in Arithmetic Combinatorics

A FURTHER REFINEMENT OF MORDELL S BOUND ON EXPONENTIAL SUMS

Upper bounds for least witnesses and generating sets

On pseudosquares and pseudopowers

Zeros of Dirichlet L-Functions over the Rational Function Field

MATH 4400 SOLUTIONS TO SOME EXERCISES. 1. Chapter 1

Sum-Product Problem: New Generalisations and Applications

INTEGERS DIVISIBLE BY THE SUM OF THEIR PRIME FACTORS

DIOPHANTINE PROBLEMS, POLYNOMIAL PROBLEMS, & PRIME PROBLEMS. positive ) integer Q how close are you guaranteed to get to the circle with rationals

PRIMALITY TEST FOR FERMAT NUMBERS USING QUARTIC RECURRENCE EQUATION. Predrag Terzic Podgorica, Montenegro

Carmichael numbers and the sieve

#A11 INTEGERS 12 (2012) FIBONACCI VARIATIONS OF A CONJECTURE OF POLIGNAC

A NOTE ON CHARACTER SUMS IN FINITE FIELDS. 1. Introduction

Balanced subgroups of the multiplicative group

On The Weights of Binary Irreducible Cyclic Codes

On pseudosquares and pseudopowers

Character sums with Beatty sequences on Burgess-type intervals

THE LIND-LEHMER CONSTANT FOR 3-GROUPS. Stian Clem 1 Cornell University, Ithaca, New York

On Gelfond s conjecture on the sum-of-digits function

Solutions to Assignment 1

Math 101 Study Session Spring 2016 Test 4 Chapter 10, Chapter 11 Chapter 12 Section 1, and Chapter 12 Section 2

History of Mathematics

Proposed by Jean-Marie De Koninck, Université Laval, Québec, Canada. (a) Let φ denote the Euler φ function, and let γ(n) = p n

A Generating-Function Approach for Reciprocity Formulae of Dedekind-like Sums

Dense product-free sets of integers

MATH 433 Applied Algebra Lecture 4: Modular arithmetic (continued). Linear congruences.

NORMAL NUMBERS AND UNIFORM DISTRIBUTION (WEEKS 1-3) OPEN PROBLEMS IN NUMBER THEORY SPRING 2018, TEL AVIV UNIVERSITY

On the low-lying zeros of elliptic curve L-functions

k, then n = p2α 1 1 pα k

Sums of Squares. Bianca Homberg and Minna Liu

Some problems related to Singer sets

Sums and products. Carl Pomerance, Dartmouth College Hanover, New Hampshire, USA. Dartmouth Mathematics Society May 16, 2012

DIGITAL EXPANSION OF EXPONENTIAL SEQUENCES

MULTIPLE CHOICE. Choose the one alternative that best completes the statement or answers the question. C) x 8. C) y = x + 3 2

FINITE GROUPS AND EQUATIONS OVER FINITE FIELDS A PROBLEM SET FOR ARIZONA WINTER SCHOOL 2016

A talk given at Suzhou Univ. (June 19) and Nankai Univ. (June 25, 2008) Zhi-Wei Sun

Generalization of Hensel lemma: nding of roots of p-adic Lipschitz functions

EXPONENTIAL SUMS OVER THE SEQUENCES OF PRN S PRODUCED BY INVERSIVE GENERATORS

arxiv:math/ v1 [math.nt] 8 Apr 1998

On the arithmetic of modular forms

Math 110 HW 3 solutions

Old and new algorithms for computing Bernoulli numbers

Algebraic Number Theory and Representation Theory

Waring s problem, the declining exchange rate between small powers, and the story of 13,792

Winter Camp 2009 Number Theory Tips and Tricks

A Variation of a Congruence of Subbarao for n = 2 α 5 β, α 0, β 0

Bounds for the first several prime character nonresidues

THE HEIGHT OF ALGEBRAIC UNITS IN LOCAL FIELDS*

Discrete Math, Fourteenth Problem Set (July 18)

Characters (definition)

BURGESS INEQUALITY IN F p 2. Mei-Chu Chang

Quartic and D l Fields of Degree l with given Resolvent

NUMBERS OF SOLUTIONS OF EQUATIONS IN FINITE FIELDS. The equations to be considered here are those of the type

On the second smallest prime non-residue

ON DIRICHLET S CONJECTURE ON RELATIVE CLASS NUMBER ONE

Tomáš Madaras Congruence classes

Math 259: Introduction to Analytic Number Theory Exponential sums II: the Kuzmin and Montgomery-Vaughan estimates

Contents. Preface to the First Edition. Preface to the Second Edition. Preface to the Third Edition

MULTIPLICATIVE SEMIGROUPS RELATED TO THE 3x +1 PROBLEM. 1. Introduction The 3x +1iteration is given by the function on the integers for x even 3x+1

UNCC 2001 Algebra II

Transcription:

On the parity of k-th powers modulo p Jennifer Paulhus Kansas State University paulhus@math.ksu.edu www.math.ksu.edu/ paulhus

This is joint work with Todd Cochrane and Chris Pinner of Kansas State University and Jean Bourgain of the Institute for Advanced Study.

Notation p will denote an odd prime O ={1, 3, 5,..., p 2} Z/pZ E={2, 4, 6,..., p 1} Z/pZ e p ( ) = e 2πi /p the odd residues the even residues We consider N k = N k (A) = #{x E Ax k O} where k and A are any integers with p A. In previous work, we showed N k > 0 for p sufficiently large and (k, p 1) = 1, resolving a conjecture of Goresky and Klapper.

A Generalization Lehmer posed the problem of determining N 1 (1), the number of even residues with odd multiplicative inverse (mod p). The expectation is that N 1 (1) p/4, which was later proven by Zhang. For example when p = 13, N 1 (1) = 3. Residue Inverse Residue Inverse 2 7 8 5 4 10 10 4 6 11 12 12 In the general setting we no longer always have N k p/4.

A general estimate for N k Φ(k) = max a Z/pZ a 0 e p (ax k ), x 0 Φ (k) = max a Z/pZ a 0 (p 1)/2 x=1 e p (ax k ) Theorem For any integer k N k p < 1 ( ) 356p 4 π Φ (k) min{log Φ, log(5p)} (k) When k is even, Φ (k) = 1 2 Φ(k) and so N k p/4 when Φ(k) = o(p).

Ideas of Proof Let S = {x n } be a sequence defined by x n A2 k 1 n k mod p for 1 n p 1 2 and I = { 1, 2,..., p 1 2 }. x n I precisely when A(2n) k O so N k = p 1 2 n=1 χ I (x n ). Using the Fourier expansion of χ I and bounds on exponential sums we obtain the second estimate in the theorem: N k p/4 < 1 π Φ (k) log(5p). A better bound is obtained by using a smooth approximation to the characteristic function which leads to an Erdös-Turán type inequality.

Ideas of Proof, continued Theorem For any sequence S = (x 1,..., x N ) Z/pZ, interval I = {a + 1, a + 2,..., a + M} Z/pZ, and positive H < p N χ I (x n ) MN p N H + 1 N p + 2 ( log H + γ + π ) Φ π 2 S n=1 where γ is Euler s constant and N Φ S = max e p y p (yx n ). n=1 The left hand side corresponds to measuring the discrepancy of the sequence of points x n /p mod 1. To obtain the constants here we use a bound of Vaaler on the discrepancy of a sequence. The bound can be improved since our interval is of length 1 2.

Restating Φ(k) = max a Z/pZ a 0 e p (ax k ), x 0 Φ (k) = max a Z/pZ a 0 (p 1)/2 x=1 e p (ax k ) Theorem For any integer k N k p < 1 ( ) 356p 4 π Φ (k) min{log Φ, log(5p)} (k)

k odd Two parameters which help dictate the bias in the parity of kth powers are d = (k, p 1) and d 1 = (k 1, p 1). Similarly we have the values s = p 1 d and t = p 1 d 1. Theorem (a) If k is odd and t is even then N k p 0.35p 89/92 log 3/2 (5p). 4 (b) If k is odd and t is odd then N k p d 1 + p 4 log p. As long as d 1 = o(p) then N k p/4.

Small s = p 1 d An Example If k = p 1, x k = 1 identically so N k = 0 or p 1 2 depending on whether A is even or odd. If k = p 1 2, x k = ±1. A and A have opposite parity and roughly half even residues are quadratic residues so N k p 4. If k = p 1 3 then Ax k AC 1, AC 2, or AC 3 mod p where the C i are the cube roots of unity. Then N k = 0, p 1 6, p 1 p 1 3, or 2 depending on how many AC i are odd.

Small s = p 1 d, continued These examples suggest the following theorem. Theorem Let k, A be any integers with p A and (Z/pZ ) k = {C 1,..., C s }. (a) If k is even then N k = p 1 s 2s i=1 χ O(AC i ). In particular if k is even and s is even then N k = p 1 4. (b) If k is odd then N k p 1 4 < s 1 2π p log(5p).

Ideas of Proof Again, let {C 1,..., C s } be the kth powers in Z/pZ and define c i such that c k i C i mod p. N k = x χ E (x)χ O (Ax k ) = Main + Error where, after some manipulations, Error = 1 s Main = p 1 2s ψ s =ψ 0 ψ ψ 0 s s χ O (AC i ) i=1 i=1 χ O (AC i )ψ(c i ) x 0 ψ(x)χ E (x)

Ideas of Proof, continued If ψ( 1) = 1 then p 1 2 x=1 ψ(x) = 1 2 x ψ(x) = 0. This will happen to all ψ satisfying ψ s = ψ 0 if k is even. Hence Error = 0. If ψ( 1) = 1 we need to use a bound of Polya, Landau, Schur, and Vinogradov for incomplete character sums. (p 1)/2 ψ(x)χ E (x) = ψ(x) 1 p log(5p). π x x=1 When k is odd exactly half the ψ satisfying ψ s = ψ 0 will also satisfy ψ( 1) = 1. Error < s 1 p log(5p). 2π

Restating Theorem Let k, A be any integers with p A and (Z/pZ ) k = {C 1,..., C s }. (a) If k is even then N k = p 1 s 2s i=1 χ O(AC i ). In particular if k is even and s is even then N k = p 1 4. (b) If k is odd then N k p 1 4 < s 1 2π p log(5p).

Small t = p 1 d 1 An Example If k = p+1 2 then t = 2 and Ax k Ax or Ax mod p depending on whether x is a quadratic residue or not. So we expect about half the even residues to become odd. If k = p+2 3 then t = 3 and Ax k AC 1 x, AC 2 x or AC 3 x mod p To compute N k in this example we need to study the distribution of points on the lattices y AC i x mod p. When none of the lattices have a small nonzero point then even and odd values are equidistributed and N k p 4.

Bias If one of the lattices has a small point, there may be bias. An Example If k = p+2 3 then t = 3 and Ax k AC 1 x, AC 2 x or AC 3 x mod p. Our results give that, depending on the size of the smallest point in the lattices, N k is asymptotically between p 4 p 12 and p 4 + p 12. Another Example When t and A are both small odd numbers we get bias. In particular if A < (p/t) 1/2(t 1) and t log p then ( N k 1 1 ) p At 4

Open Problems How large does s need to be for Φ(k) = o(p)? Work of Bourgain says there is a constant c such that Φ(k) = o(p) if s > p c log log p. Montgomery, Vaughan, and Wooley conjecture Φ(k) dp log p which would give Φ(k) = o(p) if log p s 0 as p. We know from Shparlinski that Φ(k) = o(p) requires s log p as p. When is N k = 0? If p = 3s 1 2 and k = p 1 s then the k-th powers are just <3>= {1, 3, 3 2,..., 3 s 1 } which are all odd. Here s log p, are there examples with larger s?

The End