A046644 From square root of Riemann zeta function: form Dirichlet series Sum b_n/n^s whose square is zeta function; sequence gives denominator of b_n.

1, 2, 2, 8, 2, 4, 2, 16, 8, 4, 2, 16, 2, 4, 4, 128, 2, 16, 2, 16, 4, 4, 2, 32, 8, 4, 16, 16, 2, 8, 2, 256, 4, 4, 4, 64, 2, 4, 4, 32, 2, 8, 2, 16, 16, 4, 2, 256, 8, 16, 4, 16, 2, 32, 4, 32, 4, 4, 2, 32, 2, 4, 16, 1024, 4, 8, 2, 16, 4, 8, 2, 128, 2, 4, 16, 16, 4, 8

Offset: 1

Author

N. J. A. Sloane

Keywords

• nonn
• easy
• frac
• nice
• mult

Comments

From Antti Karttunen, Aug 21 2018: (Start)

a(n) is the denominator of any rational-valued sequence f(n) which has been defined as f(n) = (1/2) * (b(n) - Sum_{d|n, d>1, d

Proof:

Proof is by induction. We assume as our induction hypothesis that the given multiplicative formula for A046644 (resp. additive formula for A046645) holds for all proper divisors d|n, dA046645(p) = 1. [Remark: for squares of primes, f(p^2) = (4*b(p^2) - 1)/8, thus a(p^2) = 8.]

First we note that A005187(x+y) <= A005187(x) + A005187(y), with equivalence attained only when A004198(x,y) = 0, that is, when x and y do not have any 1-bits in the shared positions. Let m = Sum_{e} A005187(e), with e ranging over the exponents in prime factorization of n.

For [case A] any n in A268388 it happens that only when d (and thus also n/d) are infinitary divisors of n will Sum_{e} A005187(e) [where e now ranges over the union of multisets of exponents in the prime factorizations of d and n/d] attain value m, which is the maximum possible for such sums computed for all divisor pairs d and n/d. For any n in A268388, A037445(n) = 2^k, k >= 2, thus A037445(n) - 2 = 2 mod 4 (excluding 1 and n from the count, thus -2). Thus, in the recursive formula above, the maximal denominator that occurs in the sum is 2^m which occurs k times, with k being an even number, but not a multiple of 4, thus the factor (1/2) in the front of the whole sum will ensure that the denominator of the whole expression is 2^m [which thus is equal to 2^A046645(n) = a(n)].

On the other hand [case B], for squares in A050376 (A082522, numbers of the form p^(2^k) with p prime and k>0), all the sums A005187(x)+A005187(y), where x+y = 2^k, 0 < x <= y < 2^k are less than A005187(2^k), thus it is the lonely "middle pair" f(p^(2^(k-1))) * f(p^(2^(k-1))) among all the pairs f(d)*f(n/d), 1 < d < n = p^(2^k) which yields the maximal denominator. Furthermore, as it occurs an odd number of times (only once), the common factor (1/2) for the whole sum will increase the exponent of 2 in denominator by one, which will be (2*A005187(2^(k-1))) + 1 = A005187(2^k) = A046645(p^(2^k)).

(End)

From Antti Karttunen, Aug 21 2018: (Start)

The following list gives a few such pairs num(n), b(n) for which b(n) is Dirichlet convolution of num(n)/a(n). Here ε stands for sequence A063524 (1, 0, 0, ...).

Numerators Dirichlet convolution of numerator(n)/a(n) yields

------- -----------

A046643 A000012

A257098 A008683

A317935 A003557

A318321 A003961

A317830 A175851

A317833 A078898

A317834 A078899

A317847 A303757

A317835 ε + A003415

A317845 ε + A001065

A317846 ε + A051953

A317936 ε + A100995

A317937 ε + A001221

A317938 ε + A001222

A317939 ε + A010051 (= A080339)

(End)

This sequence gives an upper bound for the denominators of any rational-valued sequence obtained as the "Dirichlet Square Root" of any integer-valued sequence. - Andrew Howroyd, Aug 23 2018

Links

• Antti Karttunen, Table of n, a(n) for n = 1..8192
• Wikipedia, Dirichlet convolution
• Index entries for sequences computed from exponents in factorization of n

Crossrefs

Cf. A000079, A005187, A028234, A037445, A050376, A067029, A082522, A268388.

See A046643 for more details. See also A046645, A317940.

Cf. A299150, A299152, A317832, A317926, A317932, A317934 (for denominator sequences of other similar constructions).

Programs

Mathematica

b[1] = 1; b[n_] := b[n] = (dn = Divisors[n]; c = 1;
Do[c = c - b[dn[[i]]]*b[n/dn[[i]]], {i, 2, Length[dn] - 1}]; c/2); a[n_] := Denominator[b[n]]; a /@ Range[78] (* Jean-François Alcover, Apr 04 2011, after Maple code in A046643 *)
a18804[n_] := Sum[n EulerPhi[d]/d, {d, Divisors[n]}];
f[1] = 1; f[n_] := f[n] = 1/2 (a18804[n] - Sum[f[d] f[n/d], {d, Divisors[ n][[2 ;; -2]]}]);
a[n_] := f[n] // Denominator;
Array[a, 78] (* Jean-François Alcover, Sep 13 2018, after A318443 *)

PARI

A046643perA046644(n) = { my(c=1); if(1==n,c,fordiv(n,d, if((d>1)&&(dA046643perA046644(d)*A046643perA046644(n/d)))); (c/2)); }
A046644(n) = denominator(A046643perA046644(n)); \\ After the Maple-program given in A046643, Antti Karttunen, Jul 08 2017

PARI

A005187(n) = { my(s=n); while(n>>=1, s+=n); s; };
A046644(n) = factorback(apply(e -> 2^A005187(e),factor(n)[,2])); \\ Antti Karttunen, Aug 12 2018

PARI

for(n=1, 100, print1(denominator(direuler(p=2, n, 1/(1-X)^(1/2))[n]), ", ")) \\ Vaclav Kotesovec, May 04 2025

Scheme

;; With memoization-macro definec.
(definec (A046644 n) (if (= 1 n) n (* (A000079 (A005187 (A067029 n))) (A046644 (A028234 n))))) ;; Antti Karttunen, Jul 08 2017

Formula

From Antti Karttunen, Jul 08 2017: (Start)

Multiplicative with a(p^n) = 2^A005187(n).

a(1) = 1; for n > 1, a(n) = A000079(A005187(A067029(n))) * a(A028234(n)).

a(n) = A000079(A046645(n)).

(End)

Sum_{j=1..n} A046643(j)/A046644(j) ~ n / sqrt(Pi*log(n)) * (1 + (1 - gamma/2)/(2*log(n))), where gamma is the Euler-Mascheroni constant A001620. - Vaclav Kotesovec, May 04 2025

Started in 1964 by Neil J. A. Sloane | Maintained by The OEIS Foundation Inc.

Content is available under The OEIS End-User License Agreement.