the anti field descent method
play

The Anti-Field-Descent Method Bernhard Schmidt Nanyang - PowerPoint PPT Presentation

The Anti-Field-Descent Method Bernhard Schmidt Nanyang Technological University Joint work with Ka Hin Leung Bordeaux, June 2017 Ryser (1963) Eigenvalues = = || 2


  1. The Anti-Field-Descent Method Bernhard Schmidt Nanyang Technological University Joint work with Ka Hin Leung Bordeaux, June 2017

  2. Ryser (1963)

  3. Eigenvalues 𝐼 π‘ˆ 𝐼 = 𝑀𝐽 𝑦 π‘ˆ 𝐼 π‘ˆ 𝐼𝑦 = |πœ‡| 2 𝑦 𝑦 π‘ˆ 𝐼𝑦 = πœ‡π‘¦ β‡’ 𝐼 π‘ˆ 𝐼 = 𝑀𝐽 β‡’ 𝑦 π‘ˆ 𝐼 π‘ˆ 𝐼𝑦 = 𝑀𝑦 𝑦 π‘ˆ πœ‡ ∈ β„‚ eigenvalue of Hadamard matrix of order 𝑀 ⇓ |πœ‡| 2 = 𝑀

  4. Circulants 𝑏 0 𝑏 1 β‹― 𝑏 π‘€βˆ’1 𝑏 π‘€βˆ’1 𝑏 0 β‹― 𝑏 π‘€βˆ’2 𝐼 = β‹― β‹― β‹― β‹― 𝑏 1 𝑏 2 β‹― 𝑏 0 Eigenvalues: π‘€βˆ’1 𝑏 𝑗 πœ‚ 𝑗𝑙 , 𝑙 = 0, … , 𝑀 βˆ’ 1, πœ‚ = exp( 2πœŒπ‘— 𝑗=0 𝑀 ) 2 π‘€βˆ’1 𝑏 𝑗 πœ‚ 𝑗𝑙 β‡’ = 𝑀 𝐼 Hadamard matrix 𝑗=0 Note: Implies 𝑀 = 4𝑣 2

  5. Basic Idea What does 2 π‘€βˆ’1 𝑏 𝑗 πœ‚ 𝑗𝑙 = 𝑀 𝑗=0 say about the 𝑏 𝑗 ’s?

  6. Divisibility Method (Turyn) 15 𝑏 𝑗 πœ‚ 𝑗 2 = 16, πœ‚ = exp( 2πœŒπ‘— Suppose 𝑗=0 16 ) 1 βˆ’ πœ‚ 3 β‹― (1 βˆ’ πœ‚ 15 ) Factorization: 2 = 1 βˆ’ πœ‚ 15 𝑏 𝑗 πœ‚ 𝑗 2 in β„€[πœ‚] Hence 1 βˆ’ πœ‚ 32 divides 𝑗=0 πœ‚ = 1 βˆ’ πœ‚ 15 = (1 + β‹― + πœ‚ 14 )(1 βˆ’ πœ‚) 1 βˆ’ 1 βˆ’ πœ‚ = (1 + β‹― + πœ‚ 14 )(1 βˆ’ πœ‚) 15 𝑏 𝑗 πœ‚ 𝑗 Thus 1 βˆ’ πœ‚ 16 divides 𝑗=0

  7. Divisibility Method (Turyn) 15 𝑏 𝑗 πœ‚ 𝑗 1 βˆ’ πœ‚ 16 divides 𝑗=0 15 𝑏 𝑗 πœ‚ 𝑗 4 divides 𝑗=0 15 𝑏 𝑗 πœ‚ 𝑗 = 𝑗=0 7 (𝑏 𝑗 βˆ’π‘ 𝑗+8 )πœ‚ 𝑗 Basis representation: 𝑗=0 4 divides 𝑏 𝑗 βˆ’ 𝑏 𝑗+8 for all 𝑗 𝑏 𝑗 ’s cannot be Β±1 No circulant Hadamard matrix of order 16

  8. Turyn (1965) Suppose a circulant Hadamard matrix of order 𝑀 = 4𝑣 2 , 𝑣 β‰₯ 2, exists. Then: β€’ 𝑣 is odd and 𝑣 β‰₯ 55 β€’ If π‘Ÿ is a prime power dividing 𝑣 , then π‘Ÿ 3 ≀ 2𝑣 2 β€’ β€œSelf - conjugacy” bound holds All results based on divisibility conditions for 𝑏 𝑗 ’s π‘€βˆ’1 𝑏 𝑗 πœ‚ 𝑗𝑙 2 = 𝑀 coming from 𝑗=0

  9. Further Known Results on Circulant Hadamard Conjecture Suppose a circulant Hadamard matrix of order 𝑀 = 4𝑣 2 , 𝑣 β‰₯ 2, exists. 2 π‘‘βˆ’1 𝐺(4𝑣 2 , 𝑣) β‰₯ 𝑣 2 ( 𝑑 = is number of distinct prime divisors of 𝑣 ) β€’ (S. 1999) β€’ Only finitely many possible 𝑣 if the prime divisors of 𝑣 are bounded by a constant (S. 1999) 𝐺 4𝑣 2 , 𝑣 2 /(4πœ’(𝐺 4𝑣 2 , 𝑣 ) β‰₯ 𝑣 2 (S. 2001) β€’ 𝐺 𝑣 2 , 𝑣 𝑣/πœ’(𝑣) β‰₯ 𝑣 2 (Leung, S. 2005) β€’ β€’ 𝑣 β‰₯ 11715 (Leung, S. 2005) β€’ Improved F-bounds (Leung, S. 2012)

  10. Parseval for Polynomials π‘€βˆ’1 𝑏 𝑗 𝑦 𝑗 and πœ‚ = exp( 2πœŒπ‘— Let 𝑔 𝑦 = 𝑗=0 𝑀 ) π‘€βˆ’1 π‘€βˆ’1 2 = 𝑔 πœ‚ 𝑙 𝑔 πœ‚ 𝑙 𝑔 πœ‚ 𝑙 𝑙=0 𝑙=0 π‘€βˆ’1 𝑏 𝑗 𝑏 π‘˜ πœ‚ 𝑙(π‘—βˆ’π‘˜) = 𝑙=0 𝑗,π‘˜ π‘€βˆ’1 πœ‚ 𝑙(π‘—βˆ’π‘˜) = 𝑏 𝑗 𝑏 π‘˜ 𝑗,π‘˜ 𝑙=0 = 𝑀 𝑏 𝑗2

  11. F-Bound (S.) Let’s try to prove 2 π‘€βˆ’1 𝑏 𝑗 πœ‚ 𝑗 < 𝑀 𝑗=0 2 = 𝑀 for all 𝑙 π‘€βˆ’1 𝑏 𝑗 𝑦 𝑗 . Then 𝑔 πœ‚ 𝑙 Write 𝑔 𝑦 = 𝑗=0 2 = 𝑀 2 π‘€βˆ’1 𝑔 πœ‚ 𝑙 Parseval β‡’ 𝑀 2 = 𝑀 𝑏 𝑗2 = 𝑙=0 2 = πœ’ 𝑀 𝑀 In particular, 𝑀 2 β‰₯ 𝑙:gcd 𝑙,𝑀 =1 𝑔 πœ‚ 𝑙

  12. F-Bound (S.) Suppose there is a divisor 𝐺 of 𝑀 such that π‘€βˆ’1 πΊβˆ’1 𝑏 𝑗 πœ‚ 𝑗 = 𝑏 𝑗𝑀/𝐺 πœ‚ 𝑗𝑀/𝐺 𝑗=0 𝑗=0 By the same argument, 2 = πœ’ 𝐺 𝑀 𝐺 2 β‰₯ 𝑔 πœ‚ 𝑙𝑀/𝐺 𝑙:gcd 𝑙,𝐺 =1 𝐺 2 Thus 𝑀 ≀ (F-bound) πœ’ 𝐺

  13. Field Descent (S. 1999) Goal: Find divisor 𝐺 of 𝑀 with π‘€βˆ’1 πΊβˆ’1 𝑏 𝑗 πœ‚ 𝑗 = 𝑏 𝑗𝑀/𝐺 πœ‚ 𝑗𝑀/𝐺 (βˆ—) 𝑗=0 𝑗=0 Let π‘ž is the largest prime divisor of 𝑀 = 4𝑣 2 If ord π‘ž 2 π‘Ÿ ≑ 0 (mod π‘ž) for all prime divisors π‘Ÿ β‰  π‘ž of 𝑣 , then (βˆ—) holds with 𝐺 = 𝑀/π‘ž 𝑀 π‘ž 2 𝑀 Hence 𝑀 ≀ β‡’ πœ’(𝑀) β‰₯ π‘ž πœ’ 𝑀 π‘ž

  14. What if Field Descent Fails? 𝑀 = 4𝑣 2 , 𝑣 odd, π‘ž largest prime divisor of 𝑣 If field descent fails, then ord π‘ž 2 π‘Ÿ β‰’ 0 (mod π‘ž) for some prime divisor π‘Ÿ β‰  π‘ž of 𝑣. In fact, π‘€βˆ’1 𝑀 π‘žβˆ’1 𝑏 𝑗 πœ‚ 𝑗 β‰  𝑏 π‘—π‘ž πœ‚ π‘—π‘ž (βˆ—βˆ—) 𝑗=0 𝑗=0 Idea: Use (βˆ—βˆ—) to construct another cyclotomic integer with β€œstrange” properties

  15. β€œTwisted” Cyclotomic Integer π‘€βˆ’1 𝑀 π‘žβˆ’1 𝑏 𝑗 πœ‚ 𝑗 β‰  𝑏 π‘—π‘ž πœ‚ π‘—π‘ž π‘Œ = (βˆ—βˆ—) 𝑗=0 𝑗=0 ord π‘ž 2 π‘Ÿ β‰’ 0 (mod π‘ž) ord π‘ž 2 𝑠 ≑ 0 (mod π‘ž) for 𝑠|𝑣 , 𝑠 β‰  π‘ž, π‘Ÿ β„š(ΞΆ) β„š) with Fix 𝜏 = β„š(πœ‚ π‘ž ) . Recall π‘Œ 2 = 𝑀 Let 𝜏 ∈ Gal( π‘Œπ‘Œ 𝜏 2 = 𝑀 2 and π‘Œπ‘Œ 𝜏 ≑ 0 (mod 𝑀/π‘Ÿ 2 ) Thus

  16. β€œTwisted” Cyclotomic Integer π‘Œπ‘Œ 𝜏 2 = 𝑀 2 π‘Œπ‘Œ 𝜏 ≑ 0 (mod 𝑀/π‘Ÿ 2 ) Set 𝑍 = π‘Œπ‘Œ 𝜏 /( 𝑀 π‘Ÿ 2 ) . Then 𝑍 is a cyclotomic integer with 𝑍 2 = π‘Ÿ 4 and 𝑍 βˆ‰ β„š(πœ‚ π‘ž ) 𝑐 ∈ β„€[πœ‚ π‘ž ] , where 𝐢 is an integral Now write 𝑍 = π‘βˆˆπΆ 𝑍 𝑐 𝑐 with 𝑍 basis of β„š(πœ‚)/β„š(πœ‚ π‘ž ) β„š), πœ‚ 𝜐 = πœ‚ π‘Ÿ . Then 𝑍 𝜐 = πœƒπ‘ for some root of Let 𝜐 ∈ Gal( β„š πœ‚ unity πœƒ

  17. β€œTwisted” Cyclotomic Integer 𝑍 2 = π‘Ÿ 4 and 𝑍 βˆ‰ β„š(πœ‚ π‘ž ) 𝑐 ∈ β„€[πœ‚ π‘ž ] 𝑍 = π‘βˆˆπΆ 𝑍 𝑐 𝑐 with 𝑍 𝑍 𝜐 = πœƒπ‘ 𝑍 𝑐 β‰  0 for some 𝑐 β‰  1 β‡’ 𝑍 𝑐 β‰  0 for at least ord π‘ž (π‘Ÿ) elements 𝑐 𝑍 𝛽 2 β‰₯ ord π‘ž (π‘Ÿ) Gal( β‡’ β„š πœ‚ β„š) π›½βˆˆGal( β„š πœ‚ β„š)

  18. β€œTwisted” Cyclotomic Integer 𝑍 𝛽 2 β‰₯ ord π‘ž (π‘Ÿ) Gal( β„š πœ‚ β„š) π›½βˆˆGal( β„š πœ‚ β„š) 𝑍 2 = π‘Ÿ 4 β‡’ ord π‘ž (π‘Ÿ) ≀ π‘Ÿ 4

  19. Result (Leung, S. 2016), Simplified Suppose a circulant Hadamard matrix of order 4𝑣 2 exists Write 𝑣 = π‘ž 𝑏 π‘₯ where π‘ž is the largest prime divisor of 𝑣 and gcd π‘ž, π‘₯ = 1 Let π‘Ÿ be the prime divisor of π‘₯ which β€œprevents” the field descent π‘ž 2𝑏 β†’ π‘ž 2π‘βˆ’π‘¦ 2 π‘‘βˆ’1 Then ord π‘ž π‘Ÿ ≀ π‘Ÿ 4𝑐 max π‘Ÿ 2 βˆͺ 𝑑 ) : 𝑑 ∈ 𝑇 𝑔 𝑑 (π‘‘βˆ’π‘”

  20. Circulant Hadamard Open Cases Order 4𝑣 2 β€’ Smallest open cases: 𝑣 = 11715 , 82005, 550605, 3854235 . β€’ 1371 open cases with 𝑣 ≀ 10 13 (computation by Borwein, Mossinghoff 2014) β€’ 423 of the 1371 cases ruled out by twisted cyclotomic integer result (Leung, S. 2016)

  21. Barker Sequences 𝑏 0 , … , 𝑏 π‘€βˆ’1 with 𝑏 𝑗 = Β±1 such that π‘œβˆ’π‘™βˆ’1 𝑏 𝑗 𝑏 𝑗+𝑙 ≀ 1, 𝑙 = 1, … , 𝑀 βˆ’ 1 𝑗=0 Exist for: 𝑀 = 1, 2, 3, 4, 5, 7, 11, 13 Conjecture: Barker sequences of length 𝑀 > 13 do not exist

  22. Known Results Suppose a Barker sequence of length 𝑀 exists. β€’ 𝑀 even β‡’ βˆƒ circulant Hadamard matrix of order 𝑀 β€’ 𝑀 odd β‡’ 𝑀 ≀ 13 (Storer, Turyn 1961) β€’ 𝑀 > 13 β‡’ 𝑀 β‰₯ 12,100 (Turyn 1965) β€’ 𝑀 > 13 β‡’ 𝑀 β‰₯ 1,898,884 and π‘ž ≑ 1 (mod 4) for all odd primes π‘ž dividing 𝑀 (Eliahou, Kervaire, Saffari 1990)

  23. Known Results (continued) Suppose a Barker sequence of length 𝑀 > 13 exists. β€’ 𝑀 > 4 βˆ™ 10 12 (β€œfield descent”, S . 1999) β€’ 𝑀 > 10 22 (Leung, S. 2005) β€’ 𝑀 > 2 βˆ™ 10 30 (Leung, S. 2012, Mossinghoff 2009)

  24. Open Cases with Length ≀ 10 50 𝒗 Factorization (computation by Borwein, Mossinghoff 2014)

  25. Result 2 𝑑 βˆ’ 1 ord π‘ž π‘Ÿ ≀ π‘Ÿ 4𝑐 max π‘Ÿ 2 βˆͺ 𝑑 ) : 𝑑 ∈ 𝑇 𝑔 𝑑 (𝑑 βˆ’ 𝑔

  26. Application to Length ≀ 10 50

  27. Consequences for Barker Sequences β€’ There is no Barker sequence of length 𝑀 with 13 < 𝑀 ≀ 4 βˆ™ 10 33 All known open cases with 𝑀 ≀ 10 50 ruled out β€’ 229,305 out of 237,807 open cases with 𝑀 ≀ 10 100 ruled out β€’ β€’ Smallest case known not to be ruled out: 𝑀 = 4 βˆ™ 30109 2 βˆ™ 1128713 2 βˆ™ 167849 2 βˆ™ 268813277 2 β‰ˆ 1.57 βˆ™ 10 51

  28. Thank you!

Recommend


More recommend