The Anti-Field-Descent Method Bernhard Schmidt Nanyang Technological University Joint work with Ka Hin Leung Bordeaux, June 2017
Ryser (1963)
Eigenvalues πΌ π πΌ = π€π½ π¦ π πΌ π πΌπ¦ = |π| 2 π¦ π¦ π πΌπ¦ = ππ¦ β πΌ π πΌ = π€π½ β π¦ π πΌ π πΌπ¦ = π€π¦ π¦ π π β β eigenvalue of Hadamard matrix of order π€ β |π| 2 = π€
Circulants π 0 π 1 β― π π€β1 π π€β1 π 0 β― π π€β2 πΌ = β― β― β― β― π 1 π 2 β― π 0 Eigenvalues: π€β1 π π π ππ , π = 0, β¦ , π€ β 1, π = exp( 2ππ π=0 π€ ) 2 π€β1 π π π ππ β = π€ πΌ Hadamard matrix π=0 Note: Implies π€ = 4π£ 2
Basic Idea What does 2 π€β1 π π π ππ = π€ π=0 say about the π π βs?
Divisibility Method (Turyn) 15 π π π π 2 = 16, π = exp( 2ππ Suppose π=0 16 ) 1 β π 3 β― (1 β π 15 ) Factorization: 2 = 1 β π 15 π π π π 2 in β€[π] Hence 1 β π 32 divides π=0 π = 1 β π 15 = (1 + β― + π 14 )(1 β π) 1 β 1 β π = (1 + β― + π 14 )(1 β π) 15 π π π π Thus 1 β π 16 divides π=0
Divisibility Method (Turyn) 15 π π π π 1 β π 16 divides π=0 15 π π π π 4 divides π=0 15 π π π π = π=0 7 (π π βπ π+8 )π π Basis representation: π=0 4 divides π π β π π+8 for all π π π βs cannot be Β±1 No circulant Hadamard matrix of order 16
Turyn (1965) Suppose a circulant Hadamard matrix of order π€ = 4π£ 2 , π£ β₯ 2, exists. Then: β’ π£ is odd and π£ β₯ 55 β’ If π is a prime power dividing π£ , then π 3 β€ 2π£ 2 β’ βSelf - conjugacyβ bound holds All results based on divisibility conditions for π π βs π€β1 π π π ππ 2 = π€ coming from π=0
Further Known Results on Circulant Hadamard Conjecture Suppose a circulant Hadamard matrix of order π€ = 4π£ 2 , π£ β₯ 2, exists. 2 π‘β1 πΊ(4π£ 2 , π£) β₯ π£ 2 ( π‘ = is number of distinct prime divisors of π£ ) β’ (S. 1999) β’ Only finitely many possible π£ if the prime divisors of π£ are bounded by a constant (S. 1999) πΊ 4π£ 2 , π£ 2 /(4π(πΊ 4π£ 2 , π£ ) β₯ π£ 2 (S. 2001) β’ πΊ π£ 2 , π£ π£/π(π£) β₯ π£ 2 (Leung, S. 2005) β’ β’ π£ β₯ 11715 (Leung, S. 2005) β’ Improved F-bounds (Leung, S. 2012)
Parseval for Polynomials π€β1 π π π¦ π and π = exp( 2ππ Let π π¦ = π=0 π€ ) π€β1 π€β1 2 = π π π π π π π π π π=0 π=0 π€β1 π π π π π π(πβπ) = π=0 π,π π€β1 π π(πβπ) = π π π π π,π π=0 = π€ π π2
F-Bound (S.) Letβs try to prove 2 π€β1 π π π π < π€ π=0 2 = π€ for all π π€β1 π π π¦ π . Then π π π Write π π¦ = π=0 2 = π€ 2 π€β1 π π π Parseval β π€ 2 = π€ π π2 = π=0 2 = π π€ π€ In particular, π€ 2 β₯ π:gcd π,π€ =1 π π π
F-Bound (S.) Suppose there is a divisor πΊ of π€ such that π€β1 πΊβ1 π π π π = π ππ€/πΊ π ππ€/πΊ π=0 π=0 By the same argument, 2 = π πΊ π€ πΊ 2 β₯ π π ππ€/πΊ π:gcd π,πΊ =1 πΊ 2 Thus π€ β€ (F-bound) π πΊ
Field Descent (S. 1999) Goal: Find divisor πΊ of π€ with π€β1 πΊβ1 π π π π = π ππ€/πΊ π ππ€/πΊ (β) π=0 π=0 Let π is the largest prime divisor of π€ = 4π£ 2 If ord π 2 π β‘ 0 (mod π) for all prime divisors π β π of π£ , then (β) holds with πΊ = π€/π π€ π 2 π€ Hence π€ β€ β π(π€) β₯ π π π€ π
What if Field Descent Fails? π€ = 4π£ 2 , π£ odd, π largest prime divisor of π£ If field descent fails, then ord π 2 π β’ 0 (mod π) for some prime divisor π β π of π£. In fact, π€β1 π€ πβ1 π π π π β π ππ π ππ (ββ) π=0 π=0 Idea: Use (ββ) to construct another cyclotomic integer with βstrangeβ properties
βTwistedβ Cyclotomic Integer π€β1 π€ πβ1 π π π π β π ππ π ππ π = (ββ) π=0 π=0 ord π 2 π β’ 0 (mod π) ord π 2 π β‘ 0 (mod π) for π |π£ , π β π, π β(ΞΆ) β) with Fix π = β(π π ) . Recall π 2 = π€ Let π β Gal( ππ π 2 = π€ 2 and ππ π β‘ 0 (mod π€/π 2 ) Thus
βTwistedβ Cyclotomic Integer ππ π 2 = π€ 2 ππ π β‘ 0 (mod π€/π 2 ) Set π = ππ π /( π€ π 2 ) . Then π is a cyclotomic integer with π 2 = π 4 and π β β(π π ) π β β€[π π ] , where πΆ is an integral Now write π = πβπΆ π π π with π basis of β(π)/β(π π ) β), π π = π π . Then π π = ππ for some root of Let π β Gal( β π unity π
βTwistedβ Cyclotomic Integer π 2 = π 4 and π β β(π π ) π β β€[π π ] π = πβπΆ π π π with π π π = ππ π π β 0 for some π β 1 β π π β 0 for at least ord π (π) elements π π π½ 2 β₯ ord π (π) Gal( β β π β) π½βGal( β π β)
βTwistedβ Cyclotomic Integer π π½ 2 β₯ ord π (π) Gal( β π β) π½βGal( β π β) π 2 = π 4 β ord π (π) β€ π 4
Result (Leung, S. 2016), Simplified Suppose a circulant Hadamard matrix of order 4π£ 2 exists Write π£ = π π π₯ where π is the largest prime divisor of π£ and gcd π, π₯ = 1 Let π be the prime divisor of π₯ which βpreventsβ the field descent π 2π β π 2πβπ¦ 2 π‘β1 Then ord π π β€ π 4π max π 2 βͺ π‘ ) : π‘ β π π π‘ (π‘βπ
Circulant Hadamard Open Cases Order 4π£ 2 β’ Smallest open cases: π£ = 11715 , 82005, 550605, 3854235 . β’ 1371 open cases with π£ β€ 10 13 (computation by Borwein, Mossinghoff 2014) β’ 423 of the 1371 cases ruled out by twisted cyclotomic integer result (Leung, S. 2016)
Barker Sequences π 0 , β¦ , π π€β1 with π π = Β±1 such that πβπβ1 π π π π+π β€ 1, π = 1, β¦ , π€ β 1 π=0 Exist for: π€ = 1, 2, 3, 4, 5, 7, 11, 13 Conjecture: Barker sequences of length π€ > 13 do not exist
Known Results Suppose a Barker sequence of length π€ exists. β’ π€ even β β circulant Hadamard matrix of order π€ β’ π€ odd β π€ β€ 13 (Storer, Turyn 1961) β’ π€ > 13 β π€ β₯ 12,100 (Turyn 1965) β’ π€ > 13 β π€ β₯ 1,898,884 and π β‘ 1 (mod 4) for all odd primes π dividing π€ (Eliahou, Kervaire, Saffari 1990)
Known Results (continued) Suppose a Barker sequence of length π€ > 13 exists. β’ π€ > 4 β 10 12 (βfield descentβ, S . 1999) β’ π€ > 10 22 (Leung, S. 2005) β’ π€ > 2 β 10 30 (Leung, S. 2012, Mossinghoff 2009)
Open Cases with Length β€ 10 50 π Factorization (computation by Borwein, Mossinghoff 2014)
Result 2 π‘ β 1 ord π π β€ π 4π max π 2 βͺ π‘ ) : π‘ β π π π‘ (π‘ β π
Application to Length β€ 10 50
Consequences for Barker Sequences β’ There is no Barker sequence of length π€ with 13 < π€ β€ 4 β 10 33 All known open cases with π€ β€ 10 50 ruled out β’ 229,305 out of 237,807 open cases with π€ β€ 10 100 ruled out β’ β’ Smallest case known not to be ruled out: π€ = 4 β 30109 2 β 1128713 2 β 167849 2 β 268813277 2 β 1.57 β 10 51
Thank you!
Recommend
More recommend