The magic sign-smoothing conjecture for Dirichlet inverse functions

Let ff be a non-constant Dirichlet-invertible arithmetic function, with Dirichlet inverse f1f^{-1}. Define the partition sequences

p1(n)=[qn]m1(1+qm),p2(n)=[qn]m1(1+qm)1,p_1(n)=[q^n]\prod_{m\geq 1}(1+q^m),\qquad p_2(n)=[q^n]\prod_{m\geq 1}(1+q^m)^{-1},

and the transformations

s1[f](n):=j=1nf(j)p1(nj),s2[f](n):=(1)nj=1nf(j)p2(nj).s_1[f](n):=\sum_{j=1}^n f(j)p_1(n-j),\qquad s_2[f](n):=(-1)^n\sum_{j=1}^n f(j)p_2(n-j).

Say that ff has property P1\mathcal{P}_1 at NN if s1[f](n)s_1[f](n) has constant sign for all nNn\geq N, and property P2\mathcal{P}_2 at NN if s2[f](n)s_2[f](n) has constant sign for all nNn\geq N. Define

M1[f]:=sup{n1:f does not have property P1 at n},M_1[f]:=\sup\{n\geq 1:f\text{ does not have property }\mathcal{P}_1\text{ at }n\}, M2[f]:=sup{n1:f does not have property P2 at n}.M_2[f]:=\sup\{n\geq 1:f\text{ does not have property }\mathcal{P}_2\text{ at }n\}.

Magic sign-smoothing conjecture. If f(n)f(n) has constant sign for all n1n\geq 1, then M1[f1]M_1[f^{-1}] and M2[f1]M_2[f^{-1}] are bounded, with eventual signs respectively satisfying

sgn(f(M1[f1]))=sgn(f(1)),sgn(f(M2[f1]))=sgn(f(1)).\operatorname{sgn}(f(M_1[f^{-1}]))=-\operatorname{sgn}(f(1)),\qquad \operatorname{sgn}(f(M_2[f^{-1}]))=\operatorname{sgn}(f(1)).

If ff takes both signs, then M1[f1]M_1[f^{-1}] and M2[f1]M_2[f^{-1}] are bounded, with eventual signs respectively satisfying

sgn(f(M1[f1]))=sgn(f(1)),sgn(f(M2[f1]))=sgn(f(1)).\operatorname{sgn}(f(M_1[f^{-1}]))=\operatorname{sgn}(f(1)),\qquad \operatorname{sgn}(f(M_2[f^{-1}]))=-\operatorname{sgn}(f(1)).

The claim predicts eventual sign stabilization after two partition-convolution transformations of Dirichlet inverses; the source gives no proof or resolution.

Sources & referencesView supporting material

Primary source

Maxie Dion Schmidt, “Factorization theorems and canonical representations for generating functions of special sums”, arXiv:2209.12287 (2022).

Progress summary

Never refreshed

Nothing recorded yet. Refresh searches the literature and the public web for attempts on this problem, and writes the first summary here.

Solutions 0

No solutions have been posted yet.