sign in sign up
<<
Home , Arithmetic function, Divisor function, Primorial

A Dedekind Psi Inequality N. A. Carella, December, 2011.

Abstract: This note shows that the !(N) / N = ! p | N (1+1/ p) , called the Dedekind psi

v1 v2 vk function, achieves its extreme values on the subset of N = 2 ⋅3  pk , where pi is the kth −2 γ prime, and vi ≥ 1. In particular, the inequality ψ (N)/ N > 6π e loglog N holds for all large squarefree primorial integers N = 2⋅3⋅5⋅⋅⋅pk unconditionally.

Mathematics Subject Classifications: 11A25, 11A41, 11Y35. Keywords: Psi function, function, Totient function, Squarefree Integers, Prime , .

1 Introduction

The psi function !(N) = N! p | N (1+1/ p) and its normalized counterpart !(N) / N = ! p | N (1+1/ p) arise in various mathematic, and physic problems. Moreover, this function is entangled with other arithmetic functions. The values of the normalized psi function coincide with the squarefree kernel

µ 2 (d) ! (1) d|N d

of the sum of ! (N) = !d | N d . In particular, !(N) / N = ! p | N (1+1/ p) = " (N) / N on the subset of square-free integers. This note proposes a new lower estimate of the Dedekind function.

2 Theorem 1. Let N ∈ ℕ be an , then ψ (N)/ N > 6π − eγ loglog N holds unconditionally for all sufficiently large primorial integer N = 2⋅3⋅5⋅⋅⋅pk, where pk is the kth prime.

An intuitive and clear-cut relationship between the Riemann hypothesis and the Dedekind psi function is established in Theorem 4 by means of the prime theorem. Recall that the Riemann hypothesis claims that the nontrivial zeros of the zeta function ζ(s) are located on the critical line { ℜe(s) = 1/2 }, see [ES]. A survey of various strikingly different reformulations of the Riemann hypothesis appears in [AM]. And a derivation of the reformulation of the Riemann hypothesis in terms of the psi function was recently resolved in [SP].

A Dedekind Psi Function Inequality

This estimate is quite similar to the lower estimate of the totient function, namely,

γ N k /ϕ(N k ) > e loglog N k (2) for all large squarefree primorial integers Nk = 2⋅3⋅5⋅⋅⋅pk. This inequality, called the Nicolas inequality, is also a reformulation of the Riemann hypothesis, see [NS].

The next section discusses background information, and supplies the proof of Theorem 1 in Subsection 2.2. The derivation of this result unfolds from a result on the oscillations theorem of finite prime product.

2 Elementary Materials And The Proof The basic concepts and results employed throughout this work are stated in this Section.

2.1 Sums and Products Over the Primes. The most basic finite sum over the prime numbers is the prime !1 harmonic sum #n"x p . The refined estimate of this finite sum, stated below, is a synthesis of various results !1 due to various authors. The earliest version #n"x p = loglog x + B1 +O(1/ log x) is due to Mertens.

Theorem 2. Let x ≥ 2 be a sufficiently large number. Then

1/2 %loglog x + B +O(e#c(loglog x) ), unconditionally, ' 1 1 ' #1/2 " = &loglog x + B1 +O((log x) ), conditional on the Riemann hypothesis, (3) p!x p ' loglog x B ((log x)#1/2 logloglog x / loglog x), unconditional oscillations, (' + 1 +$± where B1 = .2614972128… .

Proof: Use the integral representation of the finite sum

1 x d!(t) " = # , (4) p!x p c t where c > 1 is a small constant. Here, the prime counting function !(x) = #{ p ! x : p is prime } has the form

1/2 #li(x)+O(xe!c(log x) ), unconditionally, % !(x) = $li(x)+O(x1/2 log x), conditional on the Riemann hypothesis, (5) % li(x) (x1/2 logloglog x / log x), unconditional oscillations. &% +"±

The unconditional part of the prime counting formula arises from the delaVallee Poussin form of the prime 1/2 number theorem !(x) = li(x)+O(xe!c(log x) ), see [MV, p. 179], the conditional part arises from the Riemann form of the theorem !(x) = li(x)+O(x1/2 log x), and the unconditional oscillations part arises 1/2 from the Littlewood form of the prime number theorem !(x) = li(x)+!± (x logloglog x / log x), consult [IV, p.

2 A Dedekind Psi Function Inequality

x !1 51], [MV, p. 479] et cetera. Now proceed to replace the logarithm integral li(x) = "0 (t logt) dt , and the appropriate prime counting measure d!(t), and simplify the integral. ■

The proof of the unconditional part of this result is widely available in the literature, see [HW], [MV], [TN], et cetera. As an application of the last result, there are the following interesting products:

The omega notation f (x) = g(x) + Ω± (h(x)) means that both f (x) > g(x)+ c0h(x) and f (x) < g(x)! c0h(x) occur infinitely often as x → ∞, where c0 > 0 is a constant, see [MV, p. 5], [WK].

Theorem 3. Let x ∈ ℝ be a large real number, then

1/2 +e! log x +O(e!c(log x) log x), unconditionally, !1 - " 1 % - 1/2 )$1! ' = ,e! log x +O(x! log x), conditional on the Riemann hypothesis, (6) p(x # p& - e! log x (x!1/2 logloglog x / log x), unconditional oscillations, .- +*±

Proof: Consider the logarithm of the product

!1 1 1 1 log 1 1/ p B O(1/ x) , (7) #( ! ) = $ +$$ n = $ +! ! 1 + p"x p"x p p"x n%2 np p"x p where the Euler constant is defined by lim n#1 logn 0.577215665... , and the Mertens constant is ! = x!" %n$x ( # ) = defined by B1 = ! + #p"2 (log(1!1/ p)+1/ p) =.2614972128... , see [HW, p. 466]. The last equality in (7) stems n !1 from the expansion B1 = ! ! #p"2 #n"2 (np ) , which yields

1 1 B B O(1/ x) , (8) "" n = ! $ 1 $"" n = ! $ 1 + p#x n!2 np p>x n!2 np

The remaining steps follows from Theorem 2, and reversing the logarithm. ■

The third part above simplifies the proof given in [DP] of the following result:

The quantity

!1 x1/2 (1!1/ p) ! e! log x (9) ( # p"x ) attains arbitrary large positive and negative values as x → ∞.

3 A Dedekind Psi Function Inequality

Theorem 4. Let x ≥ x0 be a real number, then

1/2 +6! )2e" log x +O(e)c(log x) log x), unconditionally, - ! 1 $ - 2 1/2 (#1+ & = ,6! ) e" log x +O(x) log x), conditional on the Riemann hypothesis, (10) p'x " p% - 6 )2e" log x (x)1/2 logloglog x / log x), unconditional oscillations, .- ! +*±

Proof: For a large real number x ∈ ℝ, rewrite the product as

!1 !1 #(1+1/ p)(1!1/ p)(1!1/ p) = #(1!1/ p2 )#(1!1/ p) . (11) p"x p"x p"x

2 !2 !2 Replacing # p"2 (1!1/ p ) = $n"1µ(n)n = 6! in the first product on the right side, yields

2 !1 & 6 !2 ) !1 1!1/ p (1!1/ p) = ( 2 ! µ(n)n + (1!1/ p) #( )# '! %n$z *# p"x p"x p"x (12) 6 !1 log x 1 1/ p O( ), = 2 #( ! ) + c ! p"x x

c where z = O(x ), c ≥ 1 constant. Lastly, applying Theorem 3, to the last product above, yields the claim. ■

This result immediately gives an improved upper bound of the sum of function; currently, the best upper bound of the sum of divisors function is

! (N) / N ! e" loglog N +O(1/ loglog N) (13) for any integer N ≥ 1, see [RN].

Corollary 5. The sum of divisors function satisfies the followings inequalities unconditionally:

1/2 $e! loglog N +O(e"c(loglog N ) ), unconditionally, 1 & (i) ! = %e! loglog N +O((log N)"1/2 loglog N), conditional on the RH, d|N d & e! loglog N ((log N)"1/2 logloglog N / loglog N), unconditional oscillations, '& +#± the sum of divisors function σ(N)/N, (14)

1/2 %6! #1e" loglog N +O(e#c(loglog N ) ), unconditionally, 1 ' (ii) " = &6! #1e" loglog N +O((log N)#1/2 loglog N), conditional on the RH, d|N, µ(d )!0 d ' 6 #1e" loglog N ((log N)#1/2 logloglog N / loglog N), unconditional oscillations, (' ! +$±

4 A Dedekind Psi Function Inequality the squarefree kernel of σ(N)/N,

1/2 %(1# 6! #1)e" loglog N +O(e#c(loglog N ) ), unconditionally, 1 ' (ii) " = &(1# 6! #1e" )loglog N +O((log N)#1/2 loglog N), conditional on the RH, d|N, µ(d )!0 d ' (1 6 #1e" )loglog N ((log N)#1/2 logloglog N / loglog N), unconditional oscillations, (' # ! +$± the nonsquarefree kernel of σ(N)/N.

Proof: Utilize the decomposition of the sum of divisors function

1 1 1 ! = ! + ! , (15) d|N d d|N, µ(d )"0 d d|N, µ(d )=0 d and apply Theorem 4 to the identities

1 # 1 & 1 $! 2 ' $ 1 ' " = *%1+ (, and ! " & #1)*&1+ ), (16) d|N, µ(d )!0 d p)x $ p' d|N, µ(d )=0 d % 6 ( p"x % p(

where x = c1 log N , and c1 > 0 is a constant. ■

The oscillating nature of the sum of divisor function is readily visible in the Ramanujan expansion

! (N) " 2 ! cos" N 2cos2" N / 3 $ = #1+ + +!&. (17) N 6 " 22 32 %

2.2 The Extreme Values of Dedekind Psi Function. The proof of Theorem 1 presented below relies on the oscillation theorem of the finite prime product; a completely elementary proof, not based on the oscillation theorem is also available, but it is longer.

From the elementary inequality

"1 6 "1 1 1/ p 1 1/ p2 1 1/ p 1 1/ p , (18) !( + ) = !( " )!( " ) > 2 !( " ) p | N p |N p | N ! p | N

! "2 and Theorems 3 and 4, it is plausible to expect that ! p | N (1+1/ p) > 6e " loglog N infinitely often as N → ∞.

2 Theorem 1. Let N ∈ ℕ be a primorial integer, then ψ (N)/ N > 6π − eγ loglog N holds unconditionally for all sufficiently large N = 2⋅3⋅5⋅⋅⋅pk.

Proof : Theorem 4 implies that the product

5 A Dedekind Psi Function Inequality

6e! $ logloglog x ' 1 1/ p log x . (19) "( + ) = 2 +#± & 1/2 ) p!x " % x loglog x (

In particular, it follows that

6e! logloglog x 1 1/ p log x c (20) "( + ) > 2 + 0 1/2 p!x " x loglog x and 6e! logloglog x 1 1/ p log x c "( + ) < 2 # 0 1/2 p!x " x loglog x

occur infinitely often as x → ∞, where c0, c1, and c2 > 0 are constants. It shows that " p!x (1+1/ p) oscillates infinitely often, symmetrically about the line 6e" # $2 log x as x → ∞.

To rewrite the variable x ≥ 1 in terms of the integer N, recall that the Chebychev function ! (x) = "p!x log p , and ! , and . (21) log Nk = " log p = ! (pk ) ! (pk ) = pk + o(pk ) ! c1 log Nk p!pk

So it readily follows that pk ! x = c1 log Nk .

Moreover, since the maxima of the sum of divisor function σ(N) ≥ ψ(N) occur at the colossally abundant v1 v2 vk integers N = 2 ⋅3  pk , and v1 ≥ v2 ≥ ⋅⋅⋅ ≥ vk ≥ 1, see [AE], [BR], [LA], [RJ], it follows that the maxima of the Dedekind psi function ψ(N) occur at the squarefree primorial integers Nk = 2⋅3⋅5⋅⋅⋅pk. Therefore, expression (19) implies that

!(N ) k = (1+1/ p) N " k p!c0 log Nk (22) ! 6e loglogloglog Nk > 2 loglog Nk + c2 1/2 " (log Nk ) logloglog Nk as the primorial integer Nk = 2⋅3⋅5⋅⋅⋅pk tends to infinity. ∎

6 A Dedekind Psi Function Inequality

REFERENCES [AE] L. Alaoglu and P. Erdos, On highly composite and similar numbers, Trans. Amer: Math. Soc. 56 (1944), 448-469. [AM] American Institute of Mathematics, aimath.org [BR] Briggs, K. Abundant numbers and the Riemann hypothesis. Experiment. Math. 15 (2006), no. 2, 251-256. [ES] Edwards, H. M. Riemann's zeta function. Academic Press, 1974, New York; ISBN: 0-486-41740-9. [HW] Hardy, G. H.; Wright, E. M. An introduction to the theory of numbers. Sixth edition. Oxford University Press, Oxford, 2008. ISBN: 978-0-19-921986-5. [DP] Diamond, H. G.; Pintz, J. (2009). "Oscillation of Mertens' product formula". J. de Theorie des Nombres de Bordeaux 21: 523–533. [IV] Ivić, Aleksandar The Riemann zeta-function. The theory of the Riemann zeta-function with applications. A Wiley-Interscience Publication. John Wiley & Sons, Inc., New York, 1985. ISBN: 0-471-80634-X [LA] Jeffrey C. Lagarias, An Elementary Problem Equivalent to the Riemann Hypothesis, arXiv:math/0008177. [MV] Montgomery, Hugh L.; Vaughan, Robert C. Multiplicative . I. Classical theory. Cambridge Studies in Advanced Mathematics, 97. Cambridge University Press, Cambridge, 2007. [NS] J.-L. Nicolas, Petites valeurs de la fonction d’Euler, J. Number Theory 17 (1983) 375–388. [RJ] S. Ramanujan, Highly composite numbers, Proc. London Math. Soc. 14 (1915), 347407. [RN] Guy Robin, Grandes valeurs de la fonction somme des diviseurs et hypothe`se de Riemann. J. Math. Pures Appl. (9) 63 (1984), 187–213. [SP] Patrick Solé, Michel Planat, Extreme values of the Dedekind $\Psi$ function, arXiv:1011.1825. [TN] G. Tenenbaum, Introduction to analytic and probabilistic number theory, Cambridge Studies in Advanced Mathematics 46, (Cambridge University Press, Cambridge, 1995. [WK] Wikipedia.org.

7