Loading ...
Sorry, an error occurred while loading the content.
 

RE: [PrimeNumbers] Is phi(p^2-1)/(p^2-1) bounded?

Expand Messages
  • Paul Leyland
    ... I don t see why your belief is justified based on a small sample and no analysis. Just because a function is bounded over a small range doesn t mean it
    Message 1 of 20 , Jan 3, 2003
      > Briefly:
      >
      > forprime (p=1000000,1000500,
      > ep=eulerphi((p-1)*(p+1));print1(ep":"ep/(p^2-1)*1.0","))
      >
      > indicates 0.22~< phi(p^2-1)/(p^2-1) <~0.33
      >
      > leading one to the belief that it tends to a limit...

      I don't see why your belief is justified based on a small sample and no analysis.

      Just because a function is bounded over a small range doesn't mean it tends to a limit. Even if it's bounded over an infinite range doesn't mean it tends to a limit. For instance, sin(x) is bounded by +1 and -1 for all real x, but no limit exists as x tends to infinity.


      Paul
    • Jon Perry
      It was actually a fish for some voluntary labour. Pari chokes on primes 1000000, and I have no other means of performing such high powers tests that would be
      Message 2 of 20 , Jan 3, 2003
        It was actually a fish for some voluntary labour. Pari chokes on
        primes>1000000, and I have no other means of performing such high powers
        tests that would be required, let alone access to a modern machine.

        I did check it over a range of values, and my bounds were deduced from these
        tests. As phi(n)/n has no limit, I would assume this doesn't either, but I
        was surprised by the narrow region of results.

        Jon Perry
        perry@...
        http://www.users.globalnet.co.uk/~perry/maths/
        http://www.users.globalnet.co.uk/~perry/DIVMenu/
        BrainBench MVP for HTML and JavaScript
        http://www.brainbench.com
      • richard_heylen <richard_heylen@yahoo.co.
        Has anyone heard of any conjectures similar to the following ones I have come up with? Conjecture: For real x 1 the smallest n such that there s always a prime
        Message 3 of 20 , Jan 3, 2003
          Has anyone heard of any conjectures similar to the following ones I
          have come up with?

          Conjecture: For real x>1 the smallest n such that there's always a
          prime between x^n and (x+1)^n, is somewhat larger than 1.7632. n
          being the solution to x^n=113 and (x+1)^n=127.

          Conjecture: For all positive integers x the smallest n such that
          there's always a prime between x^n and (x+1)^n is roughly 1.5478
          n= ln 1151 / ln 95

          I would initially have said that there's always a prime between x^n
          and (x+1)^n for integers x for all n in the range ln 1151 / ln95 < n
          n < ln 523 / ln 57. That is 1.547777 < n < 1.548232.
          However, it turns out that theres no prime between
          593^1.54792=19609.5 and
          594^1.54792=19660.7
          so we have to exclude a tiny range of 6.3 millionths so that
          n > ln 19661 / ln 594
          or
          n < ln 19609 / ln 593

          Amazing!

          Richard Heylen
        • Phil Carmody
          ... Use calc instead. Or bc. Or the other calc . Or use gp and use p=nextprime(p+1) rather than forprime(p= . ... One of (p-1) and (p+1) is divisible by
          Message 4 of 20 , Jan 3, 2003
            --- Jon Perry <perry@...> wrote:
            > It was actually a fish for some voluntary labour. Pari chokes on
            > primes>1000000, and I have no other means of performing such high powers
            > tests that would be required, let alone access to a modern machine.

            Use 'calc' instead. Or bc. Or the other 'calc'. Or use gp and use
            'p=nextprime(p+1)' rather than 'forprime(p='.

            > I did check it over a range of values, and my bounds were deduced from these
            > tests. As phi(n)/n has no limit, I would assume this doesn't either, but I
            > was surprised by the narrow region of results.

            One of (p-1) and (p+1) is divisible by 2,
            the other is divisible by 4 or 2^i i>2

            One of (p-1) and (p+1) is divisible by 3.


            The 2s combine such that

            Phi((p-1)*(p+1)) = Phi(2.2^i.(p-1)/2.(p+1)/2^i)
            = 2^i.Phi((p-1)(p+1)/2^(i+1))

            The factor of three gives you a 2/3 factor.


            Therefore the highest value you will find will be
            1/2*2/3 = 1/3 from p=3,5,17
            and 1/3-eps from numbers with a few prime factors larger than 2 or 3 in p+1
            and p-1.

            e.g.
            499637 0.3333266618578673045764089881

            (23:01) gp > factor(499637-1)
            %3 =
            [2 2]
            [124909 1]
            (23:02) gp > factor(499637+1)
            %4 =
            [2 1]
            [3 1]
            [83273 1]

            Note that by HL it will reach 1/3-eps infinitely often

            Lower bound - anyone care for a stab? There should be some bound somewhere,
            I'm sure.

            Phil




            =====
            The answer to life's mystery is simple and direct:
            Sex and death. -- Ian 'Lemmy' Kilminster

            __________________________________________________
            Do you Yahoo!?
            Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
            http://mailplus.yahoo.com
          • David Broadhurst <d.broadhurst@open.ac.u
            Jon Perry mistakenly claimed that ... Invoke it with -p10000000 to precompute primes to 10M. Limit is about 430M, as I recall, but then you need to allocate
            Message 5 of 20 , Jan 3, 2003
              Jon Perry mistakenly claimed that

              > Pari chokes on primes>1000000

              Invoke it with -p10000000 to precompute primes to 10M.
              Limit is about 430M, as I recall, but then you
              need to allocate core with the -s<size> modifier.
            • David Broadhurst <d.broadhurst@open.ac.u
              ... .......................? No. p=3 gives 1/2 and p=2 gives 2/3 which is maximal, for prime p.
              Message 6 of 20 , Jan 3, 2003
                Phil:
                > Therefore the highest value you will find will be
                > 1/2*2/3 = 1/3 from p=3,5,17
                .......................?
                No. p=3 gives 1/2
                and p=2 gives 2/3 which is maximal, for prime p.
              • Phil Carmody
                ... Deliberate mistake, to see if Jon was paying attention! ;-) Phil (lying through his teeth!) ===== The answer to life s mystery is simple and direct: Sex
                Message 7 of 20 , Jan 3, 2003
                  --- "David Broadhurst <d.broadhurst@...>" <d.broadhurst@...> wrote:
                  > Phil:
                  > > Therefore the highest value you will find will be
                  > > 1/2*2/3 = 1/3 from p=3,5,17
                  > .......................?
                  > No. p=3 gives 1/2
                  > and p=2 gives 2/3 which is maximal, for prime p.

                  Deliberate mistake, to see if Jon was paying attention! ;-)

                  Phil
                  (lying through his teeth!)


                  =====
                  The answer to life's mystery is simple and direct:
                  Sex and death. -- Ian 'Lemmy' Kilminster

                  __________________________________________________
                  Do you Yahoo!?
                  Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
                  http://mailplus.yahoo.com
                • David Broadhurst <d.broadhurst@open.ac.u
                  ... 0. We believe (but cannot prove) that there are an infinite number of primes of the form primorial+1. That would be enough to make f(p)=phi(p^2-1)/(p^2-1)
                  Message 8 of 20 , Jan 3, 2003
                    Phil:
                    > Lower bound - anyone care for a stab?

                    0.

                    We believe (but cannot prove)
                    that there are an infinite number of primes of
                    the form primorial+1. That would be enough
                    to make f(p)=phi(p^2-1)/(p^2-1)
                    as close to zero as one likes.

                    At present we know that
                    p=392113#+1 is prime,
                    giving (a la Mertens)
                    f(p) < 0.0436

                    Can anyone get lower than that?

                    David
                  • David Broadhurst <d.broadhurst@open.ac.u
                    Let f(p)=phi(p^2-1)/(p^2-1). Say a prime p is lowest yet if there is no prime q
                    Message 9 of 20 , Jan 3, 2003
                      Let f(p)=phi(p^2-1)/(p^2-1).
                      Say a prime p is "lowest yet" if there is
                      no prime q<p with f(q)<f(p).
                      The "lowest yet" sequence begins
                      2, 3, 5, 11, 29, 131, 139, 181, 419, 1429, 17291, 23561,
                      23869, 188189, 315589, 483209, 614041, 1624349, 1729001,
                      8242961, 15431989, 22486309, 27033161, 36058021, 57762431,
                      61577671, 117048931, ...

                      (117048931^2-1)/4=
                      2*3*5*7*11*13*19*23*29*31*41*73*97

                      What comes next?
                    • Phil Carmody
                      ... It s what I would have guessed, but my brain has begun to stop working in the last few hours. (e.g. the p=3 - 0.5 line was on my screen when I typed p=3
                      Message 10 of 20 , Jan 3, 2003
                        --- "David Broadhurst <d.broadhurst@...>" <d.broadhurst@...> wrote:
                        > Phil:
                        > > Lower bound - anyone care for a stab?
                        >
                        > 0.

                        It's what I would have guessed, but my brain has begun to stop working in
                        the last few hours. (e.g. the p=3 -> 0.5 line was on my screen when I typed
                        p=3 -> 1/3, so I'm really not with it!)

                        > We believe (but cannot prove)
                        > that there are an infinite number of primes of
                        > the form primorial+1.

                        As many different prime factors as possible, such that
                        Product[(p-1)/p] could be over as many terms as possible.

                        > That would be enough
                        > to make f(p)=phi(p^2-1)/(p^2-1)
                        > as close to zero as one likes.

                        Of course.

                        > At present we know that
                        > p=392113#+1 is prime,
                        > giving (a la Mertens)
                        > f(p) < 0.0436
                        >
                        > Can anyone get lower than that?

                        Not without using a larger number, probably (it's possible, though, as you
                        can use fewer factors in p-1, and dope p+1 with them instead - all you
                        need's a few coincidences). However, finding primes of that size is not for
                        the faint hearted.

                        Phil


                        =====
                        The answer to life's mystery is simple and direct:
                        Sex and death. -- Ian 'Lemmy' Kilminster

                        __________________________________________________
                        Do you Yahoo!?
                        Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
                        http://mailplus.yahoo.com
                      • Phil Carmody
                        ... Nice concrete follow-on from Jon s original. I can t see a clever way to improve on brute-force search without leaving the possibility of missing some.
                        Message 11 of 20 , Jan 3, 2003
                          --- "David Broadhurst <d.broadhurst@...>" <d.broadhurst@...> wrote:
                          > Let f(p)=phi(p^2-1)/(p^2-1).
                          > Say a prime p is "lowest yet" if there is
                          > no prime q<p with f(q)<f(p).
                          > The "lowest yet" sequence begins
                          > 2, 3, 5, 11, 29, 131, 139, 181, 419, 1429, 17291, 23561,
                          > 23869, 188189, 315589, 483209, 614041, 1624349, 1729001,
                          > 8242961, 15431989, 22486309, 27033161, 36058021, 57762431,
                          > 61577671, 117048931, ...
                          >
                          > (117048931^2-1)/4=
                          > 2*3*5*7*11*13*19*23*29*31*41*73*97
                          >
                          > What comes next?

                          Nice concrete follow-on from Jon's original.

                          I can't see a clever way to improve on brute-force search without leaving
                          the possibility of missing some.
                          However, it might be possible to stick some markers in the ground for people
                          to aim at by finding squarefree-smooths that have isprime(sqrt(4s+1)).


                          Phil


                          =====
                          The answer to life's mystery is simple and direct:
                          Sex and death. -- Ian 'Lemmy' Kilminster

                          __________________________________________________
                          Do you Yahoo!?
                          Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
                          http://mailplus.yahoo.com
                        • richard_heylen <richard_heylen@yahoo.co.
                          ... ... I believe it continues as follows 181333151 267190769 331413809 376754951 636510601 1737265531 3019962791 One can obtain fairly low values of
                          Message 12 of 20 , Jan 3, 2003
                            --- In primenumbers@yahoogroups.com, "David Broadhurst
                            <d.broadhurst@o...>" <d.broadhurst@o...> wrote:
                            > Let f(p)=phi(p^2-1)/(p^2-1).
                            > Say a prime p is "lowest yet" if there is
                            > no prime q<p with f(q)<f(p).
                            > The "lowest yet" sequence begins

                            <snip>

                            > 61577671, 117048931, ...

                            I believe it continues as follows
                            181333151
                            267190769
                            331413809
                            376754951
                            636510601
                            1737265531
                            3019962791

                            One can obtain fairly low values of f(p) realtively easily. Consider
                            for example
                            p=58531393985146662592474024598667898081212671 prime
                            p-1=2.5.7.13.19.43.53.67.71.73.43520821168673.98287085283258329
                            p+1=2^8.3^3.11.17.23.29.31.37.41.47.59.61.79.83.89.97.101.103.107.109.
                            113.127.131.661

                            So the first 33 primes are factors of p^2-1 and I believe this gives
                            an f(p) around 0.113
                            This is significantly smaller than the f(p) around 0.148 for the best
                            of the minimal examples listed.
                            To break the 0.10 barrier you need primes up to 257.
                            To break the 0.05 barrier you need primes up to 75029
                            By this stage, the numbers are getting rather large.

                            Richard Heylen
                          • David Broadhurst <d.broadhurst@open.ac.u
                            ... I proved the first 4 of your addenda with the unsmart brute-force Pari-GP source m=1;mp=430*10^6; Jon please note
                            Message 13 of 20 , Jan 4, 2003
                              Richard:

                              > I believe it continues as follows
                              > 181333151
                              > 267190769
                              > 331413809
                              > 376754951
                              > 636510601
                              > 1737265531
                              > 3019962791

                              I proved the first 4 of your addenda with
                              the unsmart brute-force Pari-GP source

                              m=1;mp=430*10^6; \\ Jon please note
                              forprime(p=2,mp,n=p^2-1;s=eulerphi(n)/n;if(s<m,m=s;print(p)))

                              David
                            • Jon Perry
                              m=1;mp=430*10^6; Jon please note forprime(p=2,mp,n=p^2-1;s=eulerphi(n)/n;if(s
                              Message 14 of 20 , Jan 4, 2003
                                'm=1;mp=430*10^6; \\ Jon please note
                                forprime(p=2,mp,n=p^2-1;s=eulerphi(n)/n;if(s<m,m=s;print(p)))'

                                I'm looking...

                                'Use 'calc' instead. Or bc. Or the other 'calc'. Or use gp and use
                                'p=nextprime(p+1)' rather than 'forprime(p='.'

                                Is this the K.R. Matthews Number Theory calculator 'calc'?

                                Is there such a concept as the 'average value of f(p)'?

                                Jon Perry
                                perry@...
                                http://www.users.globalnet.co.uk/~perry/maths/
                                http://www.users.globalnet.co.uk/~perry/DIVMenu/
                                BrainBench MVP for HTML and JavaScript
                                http://www.brainbench.com
                              • David Broadhurst <d.broadhurst@open.ac.u
                                ... forprime is faster than nextprime if you can afford the memory up to p=430M
                                Message 15 of 20 , Jan 4, 2003
                                  > use 'p=nextprime(p+1)' rather than 'forprime(p='
                                  'forprime' is faster than 'nextprime'
                                  if you can afford the memory up to p=430M
                                • Phil Carmody
                                  ... Possibly. I m using Chongo s calc (Curt Landon Noll, record prime finder 2-3 decades ago), which is the standard GNU utility. The whereabouts of the
                                  Message 16 of 20 , Jan 4, 2003
                                    --- Jon Perry <perry@...> wrote:
                                    > 'm=1;mp=430*10^6; \\ Jon please note
                                    > forprime(p=2,mp,n=p^2-1;s=eulerphi(n)/n;if(s<m,m=s;print(p)))'
                                    >
                                    > I'm looking...
                                    >
                                    > 'Use 'calc' instead. Or bc. Or the other 'calc'. Or use gp and use
                                    > 'p=nextprime(p+1)' rather than 'forprime(p='.'
                                    >
                                    > Is this the K.R. Matthews Number Theory calculator 'calc'?

                                    Possibly. I'm using Chongo's calc (Curt Landon Noll, record prime finder 2-3
                                    decades ago), which is the standard 'GNU' utility. The whereabouts of the
                                    other calc is answered in the archives some time around a year back, maybe
                                    more.

                                    > Is there such a concept as the 'average value of f(p)'?

                                    I expect it to drift downards so it's not well-defined.
                                    (or maybe it is, maybe it's zero. On average numbers have 1/eps distinct
                                    divisors, i.e. a divergent number. That's got to take a toll on the phi
                                    value. Any sample up to 300000# is puny compared with the sizes of almost
                                    all integers...)

                                    Phil


                                    =====
                                    The answer to life's mystery is simple and direct:
                                    Sex and death. -- Ian 'Lemmy' Kilminster

                                    __________________________________________________
                                    Do you Yahoo!?
                                    Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
                                    http://mailplus.yahoo.com
                                  • Jon Perry
                                    I expect it to drift downards so it s not well-defined. (or maybe it is, maybe it s zero. On average numbers have 1/eps distinct divisors, i.e. a divergent
                                    Message 17 of 20 , Jan 4, 2003
                                      'I expect it to drift downards so it's not well-defined.
                                      (or maybe it is, maybe it's zero. On average numbers have 1/eps distinct
                                      divisors, i.e. a divergent number. That's got to take a toll on the phi
                                      value. Any sample up to 300000# is puny compared with the sizes of almost
                                      all integers...)'

                                      'Therefore the highest value you will find will be
                                      1/2*2/3 = 1/3 from p=3,5,17
                                      and 1/3-eps from numbers with a few prime factors larger than 2 or 3 in p+1
                                      and p-1.'

                                      Cough, cough. You make these up, or do they come naturally?

                                      Jon Perry
                                      perry@...
                                      http://www.users.globalnet.co.uk/~perry/maths/
                                      http://www.users.globalnet.co.uk/~perry/DIVMenu/
                                      BrainBench MVP for HTML and JavaScript
                                      http://www.brainbench.com
                                    • Phil Carmody
                                      ... Where the 3 was already pointed out as a typo. ... OK John. Which of the two statements do you think is wrong? And why? Come on, show us the flaws, I yearn
                                      Message 18 of 20 , Jan 4, 2003
                                        --- Jon Perry <perry@...> wrote:

                                        Quoting me:

                                        > 'I expect it to drift downards so it's not well-defined.
                                        > (or maybe it is, maybe it's zero. On average numbers have 1/eps distinct
                                        > divisors, i.e. a divergent number. That's got to take a toll on the phi
                                        > value. Any sample up to 300000# is puny compared with the sizes of almost
                                        > all integers...)'

                                        > 'Therefore the highest value you will find will be
                                        > 1/2*2/3 = 1/3 from p=3,5,17
                                        > and 1/3-eps from numbers with a few prime factors larger than 2 or 3 in p+1
                                        > and p-1.'

                                        Where the 3 was already pointed out as a typo.

                                        > Cough, cough. You make these up, or do they come naturally?

                                        OK John. Which of the two statements do you think is wrong?
                                        And why?

                                        Come on, show us the flaws, I yearn to be enlightened by your razer-sharp
                                        mathematical quill. I'll even fill in the ellipses, if you like, as have a
                                        feeling you're getting confused by my elision.

                                        �10 to Oxfam for each statement you persuade me to retract. I trust you'll
                                        reciprocate?

                                        Phil


                                        =====
                                        The answer to life's mystery is simple and direct:
                                        Sex and death. -- Ian 'Lemmy' Kilminster

                                        __________________________________________________
                                        Do you Yahoo!?
                                        Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
                                        http://mailplus.yahoo.com
                                      • Jon Perry
                                        £10 to Oxfam for each statement you persuade me to retract. I trust you ll reciprocate? What... Oxfam will pay me £10 to persuade you to retract them? I
                                        Message 19 of 20 , Jan 4, 2003
                                          '£10 to Oxfam for each statement you persuade me to retract. I trust you'll
                                          reciprocate?'

                                          What... Oxfam will pay me £10 to persuade you to retract them?

                                          I believe they are both correct, hence I will not allow Oxfam to waste their
                                          money on me, and this in turn leads me to believe that f(p) could have an
                                          average value.

                                          Jon Perry
                                          perry@...
                                          http://www.users.globalnet.co.uk/~perry/maths/
                                          http://www.users.globalnet.co.uk/~perry/DIVMenu/
                                          BrainBench MVP for HTML and JavaScript
                                          http://www.brainbench.com
                                        Your message has been successfully submitted and would be delivered to recipients shortly.