![]() |
|
|
#1 |
|
Feb 2004
France
22·229 Posts |
In an old MersenneForum thread, I provided the proof of the following theorem:
q prime >5, Mq = A*B ==> there is a unique pair (x,y) such that : M[q] = (8x)^2 - (3qy)^2 (I) Here are some more information. 1) We know: (A prime or not) A | M[q] ==> A = 1 + 2qa . M[q] = A*B = (1+2qa)(1+2qb) = 1+2q(2qab+a+b) . Let say: S = a+b P = a*b D = b-a . We have: S^2 - 4P = D^2 M[q] = 1 + 2q(2Pq + S) = 1 + 4Pq^2 + 2Sq = 1 + (S^2 - D^2)q^2 + 2Sq = 1 + 2qS + (qS)^2 - (Dq)^2 M[q] = (1 + qS)^2 - (Dq)^2 (II) (I) and (II) ==> 8x = 1 + qS and 3y = D . That entails: D = 1 (mod 2) . Since: M[q-1] = 2^(q-1) - 1 is odd and M[q-1] = (M[q] - 1)/2 = q(2Pq + S) , then S = 1 (mod 2) . We know: q odd ==> q^2 = 1 (mod 8) . Since: 2Pq^2 = 2^(q-1) - (1+Sq) and 1+Sq = 8x then 2Pq^2 = 0 (mod 8). And thus: P = 0 (mod 4) . 2) We know: a and b are solutions of: x^2 - Sx + P = 0 . Let consider the curve [S,P] : y = x^2 - Sx + P . The minimum of [S,P] is the point: m ( S/2 , -(D/2)^2 ) . For a given q, all curves [Si,Pi] cross the point: M ( -1/(2q) , M[q]/(2q)^2 ) and all points m (from different curves [S,P] for a q) belong to the curve [Mq] : y = (M[q] - (1 + 2qx)^2) / (2q)^2 . M is the maximum of this curve. All points with coordinates (S,P) belong to the line: P = (M[q] - (1 + 2qS)) / (2q)^2 . The surface between [y = 0] and [S,P] is: - D^3 / 6 . The surface between [S,P] and [Mq] is: ((1 + qS)/(2q))^3 / 3 . It's useful to draw the curves with q = 11 and q = 29 ( 3 prime roots). |
|
|
|
|
|
#2 |
|
Feb 2004
France
22·229 Posts |
Here are 2 examples for the paragraph 1) of previous note.
a) q = 11 , M11 = 23*89 ======================= A=23 ; B=89 a= 1 ; b= 4 S= 5 ; P= 4; D= 3 x= 7 ; y= 1 M11 = (8*x)^2 - (3*q*y)^2 = (8*7)^2 - (3*11*1)^2 M11 = (1+29*S)^2 - (D*q)^2 = (1+5*11)^2 - (3*11)^2 b) q = 29 ; M29 = 233*1103*2089 =============================== -1) A1=233 ; B1=2304167 ----------------------- a1= 4 ; b1=39727 S1= 39731; P1=158908; D1=39723 x1= 144025 ; y1=13241 2*P1*q+S1 = M[q-1]/q = 9256395 8*x1 = 8*144025 = 1 + 29*39731 = 1 + q*S1 = 1152200 3*29*y1 = 3*29*13241 = 29*39723 = q*D1 = 1151967 M29 = (8*x1)^2 - (3*q*y1)^2 = (1+29*S1)^2 - (D1*q)^2 -2) A2=1103 ; B2=486737 ----------------------- a2= 19 ; b2= 8392 S2= 8411 ; P2= 159448; D2=8373 x2= 30490 ; y2= 2791 2*P2*q+S2 = M[q-1]/q = 9256395 8*x2 = 8*30490 = 1 + 29*8411 = 1 + q*S2 = 243920 3*29*y2 = 3*29*2791 = 29*8373 = q*D2 = 242817 M29 = (8*x2)^2 - (3*q*y2)^2 = (1+29*S2)^2 - (D2*q)^2 -3) A3=2089 ; B3=256999 a3= 36 ; b3= 4431 S3= 4467 ; P3= 159516; D3=4395 x3= 16193 ; y3= 1465 2*P3*q+S3 = M[q-1]/q = 9256395 8*x3 = 8*16193 = 1 + 29*4467 = 1 + q*S3 = 129544 3*29*y3 = 3*29*1465 = 29*4395 = q*D3 = 127455 M29 = (8*x3)^2 - (3*q*y3)^2 = (1+29*S3)^2 - (D3*q)^2 |
|
|
|
|
|
#3 |
|
Feb 2004
France
22·229 Posts |
On the figure are displayed the curves and points for q=11 :
M : (-1/2q , Mq/4q^2) = (-)1/22 , 2047/484) m : (S/2 , -(D/2)^2) = (5/2 , -9/4) (S,P) : (5,4) [S,P] is the green curve. [Mq] is the red curve. The yellow line contains all points such that: P = (M[q] - (1 + 2qS)) / (2q)^2 . The three curves (green, yellow, red) have a common point: M . |
|
|
|
|
|
#4 |
|
Feb 2004
France
22×229 Posts |
Here are the curves for q=23 .
The line containing the possible values for (S,P) cannot be drawn due to the scale. The green curve is the [S,P] curve for which a and b are solution of y=0 . The red curve is the [Mq] curve. |
|
|
|
|
|
#5 |
|
Feb 2004
France
22·229 Posts |
Here are the curves for q=29 .
The red curve is the [Mq] curve with maximum M. The green, yellow and blue curves are the [Si,Pi] curves. The green one deals with: a1=4 b1=39727 The yellow one deals with: a2=19, b2=8392 The blue one deals with: a3=36, b3=4431 The red curve [Mq] crosses the 3 other curves at m1, m2, m3, which are the minimum of each of these 3 curves. m1 = (S1/2 , -(D1/2)^2) = (39731/2 , -(39723/2)^2) . |
|
|
|
|
|
#6 |
|
Feb 2004
France
22×229 Posts |
Here is a graphical (2 parts) presentation of:
Mq = (8x)^2-(3qy)^2 = (1+Sq)^2 - (Dq)^2 . ![]() First part: |
|
|
|
|
|
#7 |
|
Feb 2004
France
39416 Posts |
Part 2:
|
|
|
|
|
|
#8 |
|
Feb 2004
France
22·229 Posts |
The same as previously plus squares q^2 .
|
|
|
|
|
|
#9 |
|
Feb 2004
France
11100101002 Posts |
Here is a nice visual proof of: Mq = 1 (mod 6) .
|
|
|
|
![]() |
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| mersenne numbers in pharmacy ? | science_man_88 | Miscellaneous Math | 0 | 2010-07-06 11:25 |
| conjecture about mersenne numbers | sascha77 | Math | 2 | 2010-01-07 08:06 |
| 6 digit numbers and the mersenne numbers | henryzz | Math | 2 | 2008-04-29 02:05 |
| LLT numbers, linkd with Mersenne and Fermat numbers | T.Rex | Math | 4 | 2005-05-07 08:25 |
| MATHS MUSINGS -I-Maths & Teaching Kung-Fu | devarajkandadai | Miscellaneous Math | 0 | 2004-09-16 04:27 |