FACTORS OF SUMS OF POWERS OF BINOMIAL COEFFICIENTS. Dedicated to the memory of Paul Erdős. 1. Introduction. n k. f n,a =

Similar documents
Bertrand s postulate Chapter 2

BINOMIAL COEFFICIENT HARMONIC SUM IDENTITIES ASSOCIATED TO SUPERCONGRUENCES

Binomial transform of products

Jacobi symbols. p 1. Note: The Jacobi symbol does not necessarily distinguish between quadratic residues and nonresidues. That is, we could have ( a

On the transcendence of infinite sums of values of rational functions

6.4 Binomial Coefficients

Automated Proofs for Some Stirling Number Identities

Bernoulli Polynomials Talks given at LSBU, October and November 2015 Tony Forbes

CERTAIN CONGRUENCES FOR HARMONIC NUMBERS Kotor, Montenegro

x !1! + 1!2!

A talk given at Institut Camille Jordan, Université Claude Bernard Lyon-I. (Jan. 13, 2005), and University of Wisconsin at Madison (April 4, 2006).

Problem. Consider the sequence a j for j N defined by the recurrence a j+1 = 2a j + j for j > 0

and each factor on the right is clearly greater than 1. which is a contradiction, so n must be prime.

arxiv: v1 [math.nt] 10 Dec 2014

THE GREATEST ORDER OF THE DIVISOR FUNCTION WITH INCREASING DIMENSION

1. By using truth tables prove that, for all statements P and Q, the statement

distinct distinct n k n k n! n n k k n 1 if k n, identical identical p j (k) p 0 if k > n n (k)

Modern Algebra 1 Section 1 Assignment 1. Solution: We have to show that if you knock down any one domino, then it knocks down the one behind it.

GENERALIZED LEGENDRE POLYNOMIALS AND RELATED SUPERCONGRUENCES

q-fibonacci polynomials and q-catalan numbers Johann Cigler [ ] (4) I don t know who has observed this well-known fact for the first time.

MATH 324 Summer 2006 Elementary Number Theory Solutions to Assignment 2 Due: Thursday July 27, 2006

18.S34 (FALL, 2007) GREATEST INTEGER PROBLEMS. n + n + 1 = 4n + 2.

Math 4707 Spring 2018 (Darij Grinberg): midterm 2 page 1. Math 4707 Spring 2018 (Darij Grinberg): midterm 2 with solutions [preliminary version]

Factors of sums and alternating sums involving binomial coefficients and powers of integers

Bertrand s Postulate

BINOMIAL PREDICTORS. + 2 j 1. Then n + 1 = The row of the binomial coefficients { ( n

MAT1026 Calculus II Basic Convergence Tests for Series

Proc. Amer. Math. Soc. 139(2011), no. 5, BINOMIAL COEFFICIENTS AND THE RING OF p-adic INTEGERS

#A18 INTEGERS 11 (2011) THE (EXPONENTIAL) BIPARTITIONAL POLYNOMIALS AND POLYNOMIAL SEQUENCES OF TRINOMIAL TYPE: PART I

SOME TRIBONACCI IDENTITIES

Bertrand s Postulate. Theorem (Bertrand s Postulate): For every positive integer n, there is a prime p satisfying n < p 2n.

A PROBABILITY PROBLEM

Some remarks on the paper Some elementary inequalities of G. Bennett

Zeros of Polynomials

n=1 a n is the sequence (s n ) n 1 n=1 a n converges to s. We write a n = s, n=1 n=1 a n

Problem 4: Evaluate ( k ) by negating (actually un-negating) its upper index. Binomial coefficient

42 Dependence and Bases

Stream Ciphers (contd.) Debdeep Mukhopadhyay

MT5821 Advanced Combinatorics

(s)h(s) = K( s + 8 ) = 5 and one finite zero is located at z 1

Ma/CS 6a Class 22: Power Series

Some results on the Apostol-Bernoulli and Apostol-Euler polynomials

Double Derangement Permutations

Math 2112 Solutions Assignment 5

Proof of two divisibility properties of binomial coefficients conjectured by Z.-W. Sun

A generalization of Morley s congruence

CSE 1400 Applied Discrete Mathematics Number Theory and Proofs

THE ASYMPTOTIC COMPLEXITY OF MATRIX REDUCTION OVER FINITE FIELDS

Lecture 10: Bounded Linear Operators and Orthogonality in Hilbert Spaces

Combinatorics and Newton s theorem

COMP 2804 Solutions Assignment 1

MATH 304: MIDTERM EXAM SOLUTIONS

Summer MA Lesson 13 Section 1.6, Section 1.7 (part 1)

Orthogonal Functions

Q-BINOMIALS AND THE GREATEST COMMON DIVISOR. Keith R. Slavin 8474 SW Chevy Place, Beaverton, Oregon 97008, USA.

THE ARTIN CARMICHAEL PRIMITIVE ROOT PROBLEM ON AVERAGE

The log-behavior of n p(n) and n p(n)/n

It is always the case that unions, intersections, complements, and set differences are preserved by the inverse image of a function.

Homework 1 Solutions. The exercises are from Foundations of Mathematical Analysis by Richard Johnsonbaugh and W.E. Pfaffenberger.

A Pair of Operator Summation Formulas and Their Applications

BOUNDS ON SOME VAN DER WAERDEN NUMBERS. Tom Brown Department of Mathematics, Simon Fraser University, Burnaby, BC V5A 1S6

Discrete Mathematics: Lectures 8 and 9 Principle of Inclusion and Exclusion Instructor: Arijit Bishnu Date: August 11 and 13, 2009

4 A Survey of Congruent Results 1

Week 5-6: The Binomial Coefficients

SOME FINITE SIMPLE GROUPS OF LIE TYPE C n ( q) ARE UNIQUELY DETERMINED BY THEIR ELEMENT ORDERS AND THEIR ORDER

Some p-adic congruences for p q -Catalan numbers

Lecture Outline. 2 Separating Hyperplanes. 3 Banach Mazur Distance An Algorithmist s Toolkit October 22, 2009

The minimum value and the L 1 norm of the Dirichlet kernel

Random Models. Tusheng Zhang. February 14, 2013

REGULARIZATION OF CERTAIN DIVERGENT SERIES OF POLYNOMIALS

DIVISIBILITY PROPERTIES OF GENERALIZED FIBONACCI POLYNOMIALS

ECE Spring Prof. David R. Jackson ECE Dept. Notes 20

Beurling Integers: Part 2

MDIV. Multiple divisor functions

Math 2784 (or 2794W) University of Connecticut

The Structure of Z p when p is Prime

Chapter 7 COMBINATIONS AND PERMUTATIONS. where we have the specific formula for the binomial coefficients:

The multiplicative structure of finite field and a construction of LRC

On Some Properties of Digital Roots

Lecture 1. January 8, 2018

arxiv: v1 [math.nt] 26 Feb 2014

X. Perturbation Theory

TR/46 OCTOBER THE ZEROS OF PARTIAL SUMS OF A MACLAURIN EXPANSION A. TALBOT

The Boolean Ring of Intervals

On Involutions which Preserve Natural Filtration

Chapter 2. Asymptotic Notation

The Gamma function Michael Taylor. Abstract. This material is excerpted from 18 and Appendix J of [T].

Hoggatt and King [lo] defined a complete sequence of natural numbers

On Divisibility concerning Binomial Coefficients

On Order of a Function of Several Complex Variables Analytic in the Unit Polydisc

6 Integers Modulo n. integer k can be written as k = qn + r, with q,r, 0 r b. So any integer.

Different kinds of Mathematical Induction

INMO-2018 problems and solutions

Harmonic Number Identities Via Euler s Transform

Commutativity in Permutation Groups

Enumerative & Asymptotic Combinatorics

Complete Solutions to Supplementary Exercises on Infinite Series

Accepted in Fibonacci Quarterly (2007) Archived in SEQUENCE BALANCING AND COBALANCING NUMBERS

Math 220A Fall 2007 Homework #2. Will Garner A

Math 155 (Lecture 3)

Transcription:

FACTORS OF SUMS OF POWERS OF BINOMIAL COEFFICIENTS NEIL J. CALKIN Abstract. We prove divisibility properties for sus of powers of bioial coefficiets ad of -bioial coefficiets. Dedicated to the eory of Paul Erdős It is well ow that if 1. Itroductio f,a = =0 ( ) a the f,0 = + 1, f,1 = 2, f,2 = ( ) 2, ad it is possible to show (Wilf, persoal couicatio, usig techiues i [8]) that for 3 a 9, there is o closed for for f,a as a su of a fixed uber of hypergeoetric ters. Siilarly, usig asyptotic techiues, de Bruij has show [2] that if a 4, the h 2,a has o closed for, where h,a = =0( 1) ( (clearly, h 2+1,a = 0.) I this paper we will prove that while o closed for ay exist, there are iterestig divisibility properties of f,2a ad h 2,a. We will illustrate soe of the techiues which ay be applied to prove these sorts of results. Our ai results are: Theore 1. for all positive ad a, ( ) 2 2 =0( 1) ( 2 Theore 2. For all positive itegers a,, j [ ] a t(, ) ( 1) j, =0 where the [ ] are the -bioial coefficiets, ad t(, ) is a iteger polyoial i with the property that t(, 1) is the odd part of ( ). ) a ) a. 1991 Matheatics Subject Classificatio. 05A10,05A30,11B65. 1

2. Bacgroud I atteptig to exted the results of previous wor [1], we were led to cosider factorizatios of sus of powers of bioial coefficiets. It uicly becae clear that for eve expoets, sall pries occurred as divisors i a regular fashio (Propositio 3), ad that this result could be exteded (Propositio 7) to odd expoets ad alteratig sus. Further ivestigatio revealed (Propositio 8) that for all alteratig sus, the pries dividig h 2,a coicided with those dividig ( ) 2. This led us to cojecture, ad subseuetly to prove, Theore 1: as part of our proof we obtai (Theore 2) a correspodig result for -bioial coefficiets. 3. No-alteratig Sus Propositio 3. For every iteger 1, if p is a prie i the iterval < p < 2a( + 1) 1 2a 1 = + 1 + + 1 (2a 1) the p f,2a. I particular, f,2a is divisible by all pries p for which < p < 2a( + 1) 1 2a 1 = + 1 + 2a 1. The followig lea will eable us to covert iforatio about divisors of f,a which are greater tha ito iforatio about divisors less tha. Lea 4. Let = s s 1... 2 1 0 be the expasio of i base p (ad siilarly for = s s 1... 2 1 0 ). The s f,a f i,a (od p). Proof: By Lucas Theore (see for exaple Graville [6]), ( ) ( ) s i (od p) i=0 i=0 i where as usual, ( ) i i 0 (od p) if i > i. Hece all the ters i the su over for which i > i for soe i disappear, givig ( ) a f,a = as claied. s s 1 s=0 s 1 =0 s i=0 i =0 s i i=0 =0 ( s i 0 0 =0 i=0 ( i i i ) a (od p) f i,a (od p) ) a (od p) Corollary 5. If l < p ad p f l,a the p f l+jp,a for all positive itegers j. 2

We are ow i a positio to prove Propositio 3: we proceed i two stages: first, the case whe < p. Lea 6. Let p be a prie i the iterval < p < 2a(+1) 1 2a 1. The p f,2a. Proof: Let p = + r where r > 0. The we have ( ) 2a p r ( ) 2a p r f,2a = (od p) =0 =0 p r ( ) 2a r + 1 ( 1) 2a (od p) =0 p r ( ) 2a r + 1 (od p) =0 p r ( ) 2a ( + 1)( + 2)... ( + r 1) (od p). (r 1)! =0 Writig x (0) = 1 ad x (r) for the polyoial x(x + 1)... (x + r 1) this last su becoes p r ( ) 2a ( + 1)(r 1). (r 1)! =0 We ow observe that the polyoials x (0), x (1),... x (d) for a iteger basis for the space of all iteger polyoials of degree at ost d. Hece there exist itegers c 0, c 1,... c (r 1)(2a 1) so that Thus f,2a (( + 1) (r 1) ) 2a 1 = p r 1 (r 1)! 2a =0 (r 1)(2a 1) j=0 (r 1)(2a 1) 1 (r 1)! 2a j=0 (r 1)(2a 1) j=0 p r =0 c j ( + r) (j). c j ( + 1) (r 1) ( + r) (j) c j ( + 1) (r+j 1) (r 1)(2a 1) 1 (p r + 1) (r+j) c (r 1)! 2a j. r + j j=0 Now, if r + (r 1)(2a 1) < p, the each of the ters i the su is divisible by p, ad (r 1)! is ot divisible by p: hece f,2a is divisible by p. But ad if ad oly if r + (r 1)(2a 1) = 2ra 2a + 1 = 2pa 2a 2a + 1 copletig the proof of the lea. 2pa 2a 2a + 1 < p p < 2a( + 1) 1 2a 1 3

Now, suppose that = ( 1)p + l with l > 0 ad l < p < 2a(l + 1) 1. 2a 1 The by lea 6, p divides f l,2a ad hece by corollary 5, p divides f,2a. But l < p if ad oly if < p, ad 2a(l + 1) 1 p < 2a 1 if ad oly if 2a( ( 1)p) 1 p < 2a 1 that is if 2a( + 1) 1 p <. 2a 1 Thus, if 2a( + 1) 1 < p < 2a 1 the p divides f,2a, copletig the proof of Propositio 3. 4. Alteratig Sus We ote that o siilar result holds for the case of odd powers of bioial coefficiets (with the trivial exceptio of a=1). Ideed, except for the power of 2 dividig f,2a+1 (which we discuss i Lea 12), the factorizatios of sus of odd powers see to exhibit o structure: for exaple, f 28,3 = 2 6.661.3671.5153.313527009031. However, for alteratig sus of odd powers, we have Propositio 7. p divides h 2,2a+1 for pries i the itervals < p < (2a + 1)( + 1) 1 (2a + 1) 1 = + 1 + + 1 (2a + 1) 1 Proof: ideed by exaiig the proof of Propositio 3, we see that if we defie ( ( )) a g,a = ( 1) =0 so that g,2a = f,2a ad g,2a+1 = h,2a+1, the g,a is divisible by all pries i each of the itervals ( + 1)a 1 < p < a 1 so Propositios 3 ad 7 are really the sae result. For all alteratig sus we have Propositio 8. If p ( 2 ) the p h2,a. 4

Proof: Clearly 2 divides h 2,a if ad oly if 2 divides the iddle ter, ( ) 2 a, as all of the other ters cacel (od 2). Sice 2 divides ( ) 2, 2 divides h2,a. Now let p be a odd prie dividig ( ) 2 : we will show that p divides h2,a. By Kuer s theore, at least oe of the digits of 2 writte i base p is odd (sice if all are eve, the there are o carries i coputig + = 2 i base p). Let the digits of 2 i base p be (2) s, (2) s 1,... (2) 1, (2) 0. The as i Lea 4 ) 2 a =0( 1) ( 2 (2) s i ( ) a ( 1) (2)i i i=0 i =0 i ), at least oe of the digits of 2 (sice p is odd, ( 1) = ( 1) 0+ 1 +... s ). Now, sice p ( 2 i base p is odd: but the the correspodig ter i the product is zero, ad so p h 2,a, copletig the proof of Propositio 8. After coputig soe exaples, it is atural to cojecture (ad the, of course, to prove!) Theore 1. 5. The Mai Theores We will prove Theore 1 by cosiderig bioial coefficets. Defiitios: Let be a positive iteger: throughout we will deote the uber of 1 s i the biary expasio of by l() (so that 2 l() ( ) 2 ): we further defie the followig polyoials i a ideteriate : (the -euivalet of ) (the th cyclotoic polyoial i ), (the -euivalet of!), ad (the -euivalet of ( ) ). Further, defie θ () = 1 1 = 1 + + 2 + + 1 φ () = (1 d ) µ( d ) d! = θ i () i=1 [ ] =!! ( )! r(x, ) = j x ( 1 j ) = (1 )! ad s(x, ) = ( ) 1 2j+1 t(, ) = 2j+1 x s(, ) s(, )s(, )s(, ).... 2 4 8 5

Note that the apparetly ifiite product i the deoiator is i fact fiite, sice s(x, ) = 1 if x < 1. We ow ae soe useful observatios about t(, ). First, so t(, ) = s(, ) = r(, ) r( 2, 2 ) s(, ) s(, ) s(, ) s(, ) s( 2, 2 )2 s( 4, 4 )2 s( 8, 8 )2 s(,... 16 )2 = r(, 2 )2 r(, 4 )2 r(, 8 )2 r(, ) r(, ) 2 r(, ) 4 r(, 2 2 ) r(, 4 2 ) r(, 8 2... r(, 4 2 ) 2 r(, 8 2 ) 2 r(, 16 2 ) 2 r(, ) = r(, 2 )2 r(, ) 2 r(, 4 )2 r(, 2 2 ) r(, 4 2 ) 2 r(, ) 4 r(, 8 )2 r(, 4 2 ) r(, 8 2 ) 2 where agai, the apparetly ifiite product is i fact fiite. Now, sice as 1, we see that r(x, ) r( x 2, )2 r(x, 2 ) r( x 2, 2 ) 2 li t(2, ) = 1 Further, t(2 + 1, ) has a factor ( 1), so { 1 x eve 1 x odd 2 ( 2 ) 2 l. li t(2 + 1, ) = 0. 1 r( 8, ) r( 16, )2 r( 8, 2 ) r( 16, 2 ) 2 I other words, sice 2 l ( ) 2, we ay regard t(2, ) as the -euivalet of the largest odd factor of ( ) 2. Lea 9. t(, ) = φ ()... where the product is over those odd for which is odd. Proof: Clearly, if is eve the φ () does t divide t(, ). Suppose is odd: the φ () divides s(, ) exactly /2 ties, ad hece φ () divides t(, ) /2 /2 /2 /2... 2 4 2 j ties. Now, by cosiderig the biary expasio of, it is iediate that this is 0 if is eve, ad 1 if is odd. 6

Lea 10. Let,, be o-egative itegers: ad write = + = + ( ) = ( ) + ( ) where, are the least o-egative residues of, (od ). The [ ] [ ] ( ) (od φ ()) where [ ] is tae to be 0 if <. Proof: We cosider polyoials odulo θ () ad φ (). First, observe that if ad oly if that is, if ad oly if (od ) 1 1 (od 1 ), θ () θ () θ () θ () (od θ ()), (od φ ()). Now, [ ] r(, ) = r(, )r(, ) ad reducig those ters which are coprie to φ () we obtai r( 1, ) ( ) r(, ) r(, ) r(, )r(( ), ) r(, )r(( ), ). Now, = + ( ) the ( ) = 0, ad [ ] [ ] r(, ) r(, )r(( ), ) (od φ ()) [ ] ( ) (od φ ()) sice θ j () j (od (1 )), ad hece (od φ ()). If + = + ( ), the ( ) = 1, ad [ ] 0 (od φ ()), ad sice >, we have as reuired. [ ] [ ] ( ) (od φ ()) 7

We ote that evaluatig [ ] at = 1 iediately iplies Kuer s theore, sice [ ] is a product of cyclotoic polyoials, ad sice φ (1) = p if = p i ad 1 otherwise, we have a factor of p correspodig to each positio for which there is a carry i = + ( ) base p. We are grateful to Professor Ira Gessel for iforig us that Lea 10 appears without proof as a property of -bioial coefficiets i [7], as Propositio 2.2 i [4] ad as Rear 2.4 i [3]. Proof of Theore 2: It is eough to show that if ad = are odd, the [ ] a φ () ( 1) j. But, fro Lea 10, [ ] a ( 1) j =0 [ = =0 =0 =0 ] a =0 [ ] a ( ) a ( 1) + j (od φ ()) ( 1) j =0 ( ) a ( 1) j ad sice ad are odd, the secod su is zero, ad we are doe. We observe ow that both sides of Theore 2 are iteger polyoials; thus whe we evaluate the at = 1, the left had side (if o-zero) will divide the right had side. But we have already observed that t(2, 1) = ( ) 2 /2 l, ad hece we have proved Corollary 11. ( ) ( ) 2 2 a 2 2 l() ( 1). =0 To prove Theore 1 it reais to show that 2 l() We prove a stroger result by iductio. 2 =0 Lea 12. For all positive itegers a ad, 2 l() ad 2 l() =0 ( ) a 2 ( 1). =0 ( ) a. ( ) a ( 1). Proof: The theore is clearly true whe = 1: assue ow that it it holds for all values less tha. For each 1 i l(), let = 2 i ad let,,,, ( ), ( ) be defied as i Lea 10: writig w i () = [ ] a ( ( ) a ) 8

we have [ ] a w i () (od φ 2 i()) =0 By our iductio hypothesis, sice l() = l( ) + l( ), 2 l() w i (1) for each i. We ow wish to cobie these euivaleces odulo θ 2 l()() = φ 2 ()φ 4 ()φ 8 ()... φ 2 l()() ad evaluate the at = 1. To do this, defie ad for i = 2, 3,..., l(). The, settig π 1 = 1 2 l 1 π i = 1 (1 ) 2l i+1 u i () = φ 2 ()φ 4 ()... φ 2 i 1()π i φ 2 i+1()... φ 2 l()() we have u 1 () = 1 2 (1 + l 1 2 )(1 + 4 )... (1 + 2l() 1 ) 1 (od (1 + )) ad for i 2, u i () 1 2 (1 l i+1 2 )(1 + 2 )(1 + 4 )... (1 + 2i 2 )(1 + 2i )... (1 + 2l() 1 ) Further, if i j, Hece, that is, 1 (1 2i 1 )(1 + 2i )(1 + 2i+1 )... (1 + 2l() 1 ) 2l i+1 1 (od (1 + 2i 1 )) =0 [ ] a u i () 0 (od φ 2 j()). l() i=1 w i ()u i () ( od θ 2 l()()) [ ] a l() = P ()θ =0 2 l()() + w i ()u i () i=1 where we wish to coclude that P () is a iteger polyoial. Observe that it is sufficiet to prove that each w i ()u i () is a iteger polyoial, sice θ 2 l() is oic. To do this, cosider w i (): first, observe that w i () is divisible by 2 l() by our iductive hypothesis, sice < : further, if is odd, so is, ad hece the -bioial su i w i () is syetric ad its coefficiets are eve: if is eve, the l( ) i 1, ad i each case, 2 l i w i () (that is, each coefficiet of w i () is divisible by 2 l i+1 ). Thus, for each i, w i ()u i () is a iteger polyoial. We have thus prove that =0 [ ] a = P ()θ 2 l()() + 9 l() i=1 w i ()u i ()

where P () has iteger coefficiets. Now, settig = 1 i both sides, we observe that u i (1) is a iteger for each i, 2 l() w i (1) for each i (ideed, u i (1) = 0 for i 2, ad u 1 (1) = 1), ad that θ 2 l()(1) = 2 l(). Hece each ter o the right is divisible by 2 l(), provig that ( ) a 2 l(). =0 To prove that ( ) a 2 l() ( 1) =0 we proceed siilarly, settig v i () = [ ] a ( ( ) ( 1) a ), with the oly ajor differece beig i the proof that l() i=1 v i ()u i () is a iteger polyoial: i this case, if is eve, thigs wor as above, ad if is odd, the we have is odd, ad v i () is idetically eual to 0. Note that we eed to have already prove the Lea for o-alteratig sus to prove the alteratig case. This copletes the proof of Lea 12 ad thus of Theore 1. We gratefully acowledge ay iforative discussios with Professors Joatha M. Borwei, Ira Gessel, Adrew J. Graville ad Herbert S. Wilf. Refereces [1] Neil J. Cali, A Curious Bioial Idetity, Discrete Math 131 (1994), 335-337. [2] N. G. DeBruij, Asyptotic Methods i Aalysis, (Dover, New Yor 1981). [3] Ma-Due Choi, George A. Elliott, ad Norio Yui, Gauss polyoials ad the rotatio algebra, Ivetioes Matheaticae 99 (1990), 225 246. [4] J. Désaréie, U aalogue des cogrueces de Kuer pour les -obres d Euler, Europ. J. Cobi. 3 (1982), 19 28. [5] A. J. Graville, Zaphod Beeblebrox s Brai ad the Fify-ith Row of Pascal s Triagle, Aerica Math. Mothly 99 (1992), 318 331. [6] A. J. Graville, The Arithetic Properties of Bioial Coefficiets Proceedigs of the Orgaic Matheatics Worshop (1996) http://www.cec.sfu.ca/orgaics/papers/graville/idex.htl (URL verified October 10, 1996). [7] Gloria Olive, Geeralized Powers, Aerica Math Mothly 72 (1965), 619 627. [8] Maro Petovše, Herbert S. Wilf ad Doro Zeilberger, A=B, (A. K. Peters, Wellesley, Massachusetts, 1996). School Of Matheatics, Georgia Istitute of Techology, Atlata, GA 30332 E-ail address: cali@ath.gatech.edu 10