## Other

drummers-lowrise

Message boards : General discussion : b^n - b^m - 1 primes

Author Message
Bur
Volunteer tester

Joined: 25 Feb 20
Posts: 332
ID: 1241833
Credit: 22,611,276
RAC: 0

Message 145438 - Posted: 20 Nov 2020 | 18:09:38 UTC

I saw that LLR2 also allows numbers of form b^n - b^m - 1 as input. I never saw those before, what type is that?

Playing around with it a bit, I realized that they are actually just a special type of Riesel base b:

b^n - b^m - 1 = [b^(n-m) - 1] * b^m - 1

So why are they specifically mentioned?

____________
Primes: 1281979 & 12+8+1979 & 1+2+8+1+9+7+9 & 1^2+2^2+8^2+1^2+9^2+7^2+9^2 & 12*8+19*79 & 12^8-1979 & 1281979 + 4 (cousin prime)

JeppeSN

Joined: 5 Apr 14
Posts: 1378
ID: 306875
Credit: 21,623,822
RAC: 0

Message 145457 - Posted: 20 Nov 2020 | 20:32:44 UTC

How did you see it?

If I call LLR2 with -h, I see:

-q"expression" Test a single k*b^n+c or b^n-b^m+c number.

One special case is primes of form Phi(3, -(b^n)) where Phi(3, -x) is the third cyclotomic polynomial evaluated at negative x, which is x^2 - x + 1.

/JeppeSN

Bur
Volunteer tester

Joined: 25 Feb 20
Posts: 332
ID: 1241833
Credit: 22,611,276
RAC: 0

Message 145487 - Posted: 21 Nov 2020 | 5:58:17 UTC - in response to Message 145457.

b^n-b^m+c

For LLR2 c can be -1 or +1. I just looked at the -1 case, but +1 is the same. Then it's a special Proth number.

Of course it can also be any other number than -1/+1 but that it can still be expressed as k * b^n + c.

So why is it mentioned specifically?
____________
Primes: 1281979 & 12+8+1979 & 1+2+8+1+9+7+9 & 1^2+2^2+8^2+1^2+9^2+7^2+9^2 & 12*8+19*79 & 12^8-1979 & 1281979 + 4 (cousin prime)

Gelly
Volunteer tester

Joined: 13 Nov 16
Posts: 40
ID: 468732
Credit: 522,960,451
RAC: 0

Message 145537 - Posted: 22 Nov 2020 | 2:42:38 UTC - in response to Message 145487.

It actually would not be a Proth number. For a generalized Proth number k*b^n + 1 (or Thorp, k*b^n-1), the requirement is that k < b^n. For the rearrangement you provided, where k = b^(n-m)-1, it will not necessarily be true that k < b^m (such as when n-m > m)

Edit: Concerning your original question, programs usually have a limit on the size of k that are acceptable to test. With a different form, you don't have to fight with the k requirement.

Bur
Volunteer tester

Joined: 25 Feb 20
Posts: 332
ID: 1241833
Credit: 22,611,276
RAC: 0

Message 145575 - Posted: 23 Nov 2020 | 18:10:27 UTC

It might not be called Proth, but it will be accepted by most software anyway, I think.

But if a very large k can be circumvented by writing it like this, maybe that's the reason.

I still think it's strange that it's specifically mentioned but apparently not really used (at least by active users here). I might ask at mersenneforum.
____________
Primes: 1281979 & 12+8+1979 & 1+2+8+1+9+7+9 & 1^2+2^2+8^2+1^2+9^2+7^2+9^2 & 12*8+19*79 & 12^8-1979 & 1281979 + 4 (cousin prime)

Message boards : General discussion : b^n - b^m - 1 primes