Small-span characteristic polynomials of integer symmetric matrices - - PowerPoint PPT Presentation

small span characteristic polynomials of integer
SMART_READER_LITE
LIVE PREVIEW

Small-span characteristic polynomials of integer symmetric matrices - - PowerPoint PPT Presentation

Small-span characteristic polynomials of integer symmetric matrices James McKee (RHUL) ANTS 9, July 20, 2010 PLAN ISMs, characteristic polynomials, minimal polynomials 1/1729 PLAN ISMs, characteristic polynomials, minimal polynomials


slide-1
SLIDE 1

Small-span characteristic polynomials

  • f integer symmetric matrices

James McKee (RHUL) ANTS 9, July 20, 2010

slide-2
SLIDE 2

PLAN

  • ISMs, characteristic polynomials, minimal polynomials

1/1729

slide-3
SLIDE 3

PLAN

  • ISMs, characteristic polynomials, minimal polynomials
  • Small span polynomials

1/1729

slide-4
SLIDE 4

PLAN

  • ISMs, characteristic polynomials, minimal polynomials
  • Small span polynomials
  • Intersecting the two problems

1/1729

slide-5
SLIDE 5

PLAN

  • ISMs, characteristic polynomials, minimal polynomials
  • Small span polynomials
  • Intersecting the two problems
  • Computational Results

1/1729

slide-6
SLIDE 6

PLAN

  • Small span polynomials
  • Intersecting the two problems
  • Computational Results
  • Theorem

1/1729

slide-7
SLIDE 7

PLAN

  • Intersecting the two problems
  • Computational Results
  • Theorem
  • Application

1/1729

slide-8
SLIDE 8

PLAN

  • Computational Results
  • Theorem
  • Application
  • Future work

1/1729

slide-9
SLIDE 9

Integer symmetric matrices (ISMs) These are things like:

  

1 −2 3 −2 3 7

  

(symmetric square matrix, integer entries) 2/1729

slide-10
SLIDE 10

Properties of ISMs Their characteristic polynomials

  • are monic
slide-11
SLIDE 11

Properties of ISMs Their characteristic polynomials

  • are monic
  • have integer coefficients
slide-12
SLIDE 12

Properties of ISMs Their characteristic polynomials

  • are monic
  • have integer coefficients
  • have all roots real
slide-13
SLIDE 13

Properties of ISMs Their characteristic polynomials

  • are monic
  • have integer coefficients
  • have all roots real

To what extent is the converse true?

slide-14
SLIDE 14

Example 1 The polynomial x2 − 2 is monic, has integer coefficients, and all roots real.

slide-15
SLIDE 15

Example 1 The polynomial x2 − 2 is monic, has integer coefficients, and all roots real. It is the characteristic polynomial of

  • 1

1 1 −1

slide-16
SLIDE 16

Example 2 Consider the polynomial x2 − 3. Can this be the characteristic polynomial of an ISM?

slide-17
SLIDE 17

Example 2 Consider the polynomial x2 − 3. Can this be the characteristic polynomial of an ISM? No!

slide-18
SLIDE 18

Example 2 Consider the polynomial x2 − 3. Can this be the characteristic polynomial of an ISM? det

  • x − a

−b −b x − c

  • =

x2 − (a + c)x + ac − b2 det

  • x − a

−b −b x + a

  • =

x2 − a2 − b2

slide-19
SLIDE 19

Example 2 Consider the polynomial x2 − 3. Can this be the characteristic polynomial of an ISM? det

  • x − a

−b −b x − c

  • =

x2 − (a + c)x + ac − b2 det

  • x − a

−b −b x + a

  • =

x2 − a2 − b2

slide-20
SLIDE 20

Example 2 Consider the polynomial x2 − 3. Can this be the characteristic polynomial of an ISM? det

  • x − a

−b −b x − c

  • =

x2 − (a + c)x + ac − b2 det

  • x − a

−b −b x + a

  • =

x2 − a2 − b2 We need a2 + b2 = 3.

slide-21
SLIDE 21

Example 2 (continued) Consider the polynomial x2 − 3. Can this be the min. poly. of an ISM?

slide-22
SLIDE 22

Example 2 (continued) Consider the polynomial x2 − 3. Can this be the min. poly. of an ISM? Yes!

slide-23
SLIDE 23

Example 2 (continued) Consider the polynomial x2 − 3. Can this be the min. poly. of an ISM? Yes!

    

−1 1 1 1 1 1 1 1 −1 1 −1 −1

    

slide-24
SLIDE 24

Example 3 Consider the polynomial x3 − 4x − 1.

slide-25
SLIDE 25

Example 3 Consider the polynomial x3 − 4x − 1. This is not the characteristic polynomial of an ISM. (Why?)

slide-26
SLIDE 26

Example 3 Consider the polynomial x3 − 4x − 1. This is not the characteristic polynomial of an ISM. (Why?) But it is the min. poly. of the following 6 × 6 ISM:

         

1 1 1 1 1 1 −1 1 1 1 −1 1 −1 −1 1 1

         

slide-27
SLIDE 27

Theorem of Estes and Guralnick (1993) Let f(x) be a monic, separable polynomial with integer coeffi- cients, degree n, and with all roots real. If n ≤ 4, then f is the min. poly. of a 2n × 2n ISM.

slide-28
SLIDE 28

Question 1 of Estes and Guralnick (1993) Let f(x) be a monic, separable polynomial with integer coeffi- cients, degree n, and with all roots real. Is f always the min. poly. of an ISM?

slide-29
SLIDE 29

Question 1 of Estes and Guralnick (1993) Let f(x) be a monic, separable polynomial with integer coeffi- cients, degree n, and with all roots real. Is f always the min. poly. of an ISM? They conjectured that the answer is ‘yes’.

slide-30
SLIDE 30

Answer: Dobrowolski (2008)

  • No!
slide-31
SLIDE 31

Answer: Dobrowolski (2008)

  • No!
  • Indeed there exist infinitely many f (monic, separable, integer

coefficients, and with all roots real) for which f is not the

  • min. poly. of any ISM.
slide-32
SLIDE 32

Answer: Dobrowolski (2008)

  • No!
  • Indeed there exist infinitely many f (monic, separable, integer

coefficients, and with all roots real) for which f is not the

  • min. poly. of any ISM.
  • He shows that if f (degree n) is the min. poly. of an ISM,

then the discriminant of f is at least nn. For large, highly composite m, the discriminant of the min. poly. of 2 cos(π/m) is too small.

slide-33
SLIDE 33

Let’s change the question What is the smallest n such that there is a monic, separable polynomial f(x) of degree n, with integer coefficients and with all roots real, and with f not the min. poly. of any integer symmetric matrix?

slide-34
SLIDE 34

Let’s change the question What is the smallest n such that there is a monic, separable polynomial f(x) of degree n, with integer coefficients and with all roots real, and with f not the min. poly. of any integer symmetric matrix?

  • Dobrowolski: 5 ≤ n ≤ 2880
slide-35
SLIDE 35

Let’s change the question What is the smallest n such that there is a monic, separable polynomial f(x) of degree n, with integer coefficients and with all roots real, and with f not the min. poly. of any integer symmetric matrix?

  • Dobrowolski: 5 ≤ n ≤ 2880
  • More precise answer: n ∈ {5, 6}
slide-36
SLIDE 36

Some degree-6 examples I claim that the following polyomials are monic, separable, with all roots real, but do not arise as the min. poly. of any ISMs:

  • x6 − x5 − 6x4 + 6x3 + 8x2 − 8x + 1
  • x6 − 7x4 + 14x2 − 7
  • x6 − 6x4 + 9x2 − 3
slide-37
SLIDE 37

Summary to this point We don’t fully understand which polynomials arise as character- istic polynomials of integer symmetric matrices. We don’t fully understand which polynomials arise as min. polys. of integer symmetric matrices.

slide-38
SLIDE 38

SMALL-SPAN POLYNOMIALS

  • Definition
slide-39
SLIDE 39

SMALL-SPAN POLYNOMIALS

  • Definition
  • Equivalence
slide-40
SLIDE 40

SMALL-SPAN POLYNOMIALS

  • Definition
  • Equivalence
  • History
slide-41
SLIDE 41

SMALL-SPAN: DEFINITION A totally real, monic polynomial with integer coefficients, f(x) = xd + ad−1xd−1 + · · · + x0 , with roots α1 ≤ · · · ≤ αd, has span αd − α1.

slide-42
SLIDE 42

SMALL-SPAN: DEFINITION A totally real, monic polynomial with integer coefficients, f(x) = xd + ad−1xd−1 + · · · + x0 , with roots α1 ≤ · · · ≤ αd, has span αd − α1. The span is small if it is strictly less than 4.

slide-43
SLIDE 43

SMALL-SPAN: DEFINITION A totally real, monic polynomial with integer coefficients, f(x) = xd + ad−1xd−1 + · · · + x0 , with roots α1 ≤ · · · ≤ αd, has span αd − α1. The span is small if it is strictly less than 4. WHY 4?

slide-44
SLIDE 44

SMALL-SPAN: EQUIVALENCE

  • For any integer c, and any ε = ±1, the polynomials f(x) and

εdf(εx + c) will be called equivalent.

slide-45
SLIDE 45

SMALL-SPAN: EQUIVALENCE

  • For any integer c, and any ε = ±1, the polynomials f(x) and

εdf(εx + c) will be called equivalent.

  • Equivalent polynomials have the same span.
slide-46
SLIDE 46

SMALL-SPAN: EQUIVALENCE

  • For any integer c, and any ε = ±1, the polynomials f(x) and

εdf(εx + c) will be called equivalent.

  • Equivalent polynomials have the same span.
  • Any small-span polynomial is equivalent to one with all its

roots in the interval [−2, 2.5).

slide-47
SLIDE 47

SMALL-SPAN: WHY 4?

  • Suppose that f(x) (monic, integer coefficients, all roots real)

has all its roots in the interval [−2, 2]. Then the roots of f(x) are all of the form 2 cos(2π/m), where m is a natural number.

slide-48
SLIDE 48

SMALL-SPAN: WHY 4?

  • Suppose that f(x) (monic, integer coefficients, all roots real)

has all its roots in the interval [−2, 2]. Then the roots of f(x) are all of the form 2 cos(2π/m), where m is a natural number.

  • I’ll call such a polynomial a cosine polynomial.
slide-49
SLIDE 49

SMALL-SPAN: WHY 4?

  • Suppose that f(x) (monic, integer coefficients, all roots real)

has all its roots in the interval [−2, 2]. Then the roots of f(x) are all of the form 2 cos(2π/m), where m is a natural number.

  • I’ll call such a polynomial a cosine polynomial.
  • Any small-span polynomial that is not equivalent to a cosine

polynomial is especially interesting.

slide-50
SLIDE 50

SMALL-SPAN: HISTORY

  • For fixed degree, the number of equivalence classes of small-

span polynomials is finite.

slide-51
SLIDE 51

SMALL-SPAN: HISTORY

  • For fixed degree, the number of equivalence classes of small-

span polynomials is finite.

  • Robinson (1964): up to degree 6; conjectured lists for de-

grees 7 and 8.

slide-52
SLIDE 52

SMALL-SPAN: HISTORY

  • For fixed degree, the number of equivalence classes of small-

span polynomials is finite.

  • Robinson (1964): up to degree 6; conjectured lists for de-

grees 7 and 8.

  • Flamming, Rhin, Wu (2009, unpublished): up to degree 13.
slide-53
SLIDE 53

SMALL-SPAN: HISTORY

  • For fixed degree, the number of equivalence classes of small-

span polynomials is finite.

  • Robinson (1964): up to degree 6; conjectured lists for de-

grees 7 and 8.

  • Flamming, Rhin, Wu (2009, unpublished): up to degree 13.
  • Capparelli, Del Fra, Sci`
  • (2010): up to degree 14, conjec-

tured lists for degrees 15 and 16.

slide-54
SLIDE 54

SMALL-SPAN: HISTORY

  • Robinson (1964): up to degree 6; conjectured lists for de-

grees 7 and 8.

  • Flamming, Rhin, Wu (2009, unpublished): up to degree 13.
  • Capparelli, Del Fra, Sci`
  • (2010): up to degree 14, conjec-

tured lists for degrees 15 and 16.

  • Stop press (June 2010): Rhin et al have verified the degree

15 list.

slide-55
SLIDE 55

SMALL-SPAN: SUMMARY degree # classes # non-cosine degree # classes # non-cosine 1 1 9 21 19 2 4 1 10 28 15 3 5 3 11 11 10 4 14 10 12 16 9 5 15 14 13 4 4 6 17 13 14 10 9 7 15 15 15 7 6 8 26 21 16 ≥ 9 ≥ 3

slide-56
SLIDE 56

SMALL-SPAN CHARACTERISTIC POLYNOMIALS We can intersect the previous two (unsolved) problems, and get an easier problem: Which small-span polynomials arise as characteristic polynomials (or minimal polynomials) of ISMs?

slide-57
SLIDE 57

SMALL-SPAN CHARACTERISTIC POLYNOMIALS We can intersect the previous two (unsolved) problems, and get an easier problem: Which small-span polynomials arise as characteristic polynomials (or minimal polynomials) of ISMs? There is a natural notion of equivalence.

slide-58
SLIDE 58

SMALL-SPAN CHARACTERISTIC POLYNOMIALS: AN EXAMPLE

② ② ② ② ② ② ② ② ② ② ② ②

1 1 1 1 1 1

s s s s s s s s s s

slide-59
SLIDE 59

SMALL-SPAN CHARACTERISTIC POLYNOMIALS: AN EXAMPLE

② ② ② ② ② ② ② ② ② ② ② ②

1 1 1 1 1 1

s s s s s s s s s s

Eigenvalues: −1.4955. . . , −1.4955. . . , −1, −1, −0.2196. . . , −0.2196. . . , 1.2196. . . , 1.2196. . . , 2, 2, 2.4955. . . , 2.4955. . . .

slide-60
SLIDE 60

GROWING

  • FACT: For n > 1, any small-span n-by-n ISM can be ‘grown’

from an (n − 1)-by-(n − 1) small-span ISM.

slide-61
SLIDE 61

GROWING

  • FACT: For n > 1, any small-span n-by-n ISM can be ‘grown’

from an (n − 1)-by-(n − 1) small-span ISM.

  • GROWING ALGORITHM: find all 1-by-1 examples (up to

equivalence), grow to 2-by-2, 3-by-3, etc..

slide-62
SLIDE 62

RESULTS: MAXIMAL SMALL-SPAN ISMs UP TO EQUIVALENCE n # #′ n # #′ n # #′ 1 1 6 48 11 15 2 1 7 36 12 17 3 2 8 59 13 15 4 21 9 25 14 16 5 22 10 27 15 17

slide-63
SLIDE 63

RESULTS: REMOVING MEMBERS OF 10 FAMILIES n # #′ n # #′ n # #′ 1 1 1 6 48 43 11 15 2 2 1 1 7 36 28 12 17 3 3 2 1 8 59 50 13 15 4 21 19 9 25 15 14 16 5 22 19 10 27 15 15 17

slide-64
SLIDE 64

CLASSIFICATION THEOREM n > 12 # #′ n n + 2

slide-65
SLIDE 65

APPLICATION: A QUESTION OF ESTES AND GURALNICK Computations + a small argument produce lots of small-degree counterexamples to the conjecture of Estes and Guralnick con- cerning minimal polynomials of ISMs.

slide-66
SLIDE 66

FUTURE WORK?

  • Degree 5?
slide-67
SLIDE 67

FUTURE WORK?

  • Degree 5?
  • Entries from OK for various number fields K? (Hermitian)
slide-68
SLIDE 68

FUTURE WORK?

  • Degree 5?
  • Entries from OK for various number fields K? (Hermitian)
  • Gary Greaves has completed [K : Q] = 2.
slide-69
SLIDE 69

FUTURE WORK?

  • Degree 5?
  • Entries from OK for various number fields K? (Hermitian)
  • Gary Greaves has completed [K : Q] = 2.
  • Two of the three degree-6 polynomials now appear as mini-

mal polynomials.

slide-70
SLIDE 70

THANK YOU FOR LISTENING ω = (1 + √−3)/2

② ② ② ② ② ② ② ② ② ② ② ②

x6 − 7x4 + 14x2 − 7 x6 − 6x4 + 9x2 − 3

✁❆ω s s s s s s s s s s s s s s s s s s s s ✁❆ω

slide-71
SLIDE 71

QUESTIONS? QUESTIONS?