On the q-factorization of power series

Robert Schneider, Andrew V. Sills and Hunter Waldron

arXiv: 2501.18744v2 · PDF · 11 June 2025
Categories: math.CO, math.NT
Journal: Ramanujan J 67, 92 (2025)
DOI: 10.1007/s11139-025-01140-4
Licence: CC BY 4.0 The site shows the text.

Abstract

Any power series with unit constant term can be factored into an infinite product of the form \prod_{n\geq 1} (1-q^n)^{-a_n}. We give direct formulas for the exponents a_n in terms of the coefficients of the power series, and vice versa, as sums over partitions. As examples, we prove identities for certain partition enumeration functions. Finally, we note q-analogues of our enumeration formulas.

Items 16

Proposition

Proposition 1.1

Q9NC

Let r(n) and a_n be defined as in (?). Then

r(n) = \sum_{\lambda \vdash n} \frac{ (a_1)_{m_1} (a_2)_{m_2} \cdots } {m_1!\ m_2! \cdots},

where (a)_k = (a)(a+1)(a+2)\cdots(a+k-1) is the usual rising factorial.

Theorem

Theorem 1.2

ZPU7

Let r(n) and a_n be defined as in (?). Then

a_n = \frac{1}{n} \sum_{d\mid n} \mu\left( \frac{n}{d} \right) d \sum_{\lambda\vdash d} \frac{ (-1)^{\ell - 1} (\ell-1)! \ r(1)^{m_1} r(2)^{m_2}\cdots } { m_1!\ m_2! \cdots } .

Used by Example 5.1, Example 5.2

Theorem

Theorem 2.1 (Fine)

6HDI

Let \psi_j(q) = \sum_{k\geq 0} C_j(k) q^k,\ (j = 1,2,\dots). Then

\prod_{j\geq 1} \psi_j(q^j) = \sum_{n\geq 0} q^n \sum_{\lambda\vdash n} C_1(m_1(\lambda) ) C_2(m_2(\lambda)) \cdots .

Example

Example 3.1

FU0L

A canonical special case of (?) is where all a_n = 1, and thus r(n) = p(n), the number of partitions of n. Then (1.1) reduces to the trivial statement

p(n) = \sum_{\lambda\vdash n} 1,

as (1)_m = m!. In other words, p(n) is obtained by literally counting all partitions of n, one by one.

Example

Example 3.2

T4MW

Let p_S(n) denote the number of partitions of n in which all parts are in some subset S\subseteq \mathbb Z^+ of the positive integers. Observe that

\sum_{n=0}^\infty p_S(n) q^n = \prod_{n\in S} \frac{1}{1-q^n}.

Upon noting (0)_0 = 1 (the empty product) and (0)_m = 1 when m is a positive integer, we see that

p_S(n) = \sum_{\lambda\vdash n} \chi( \mbox{ all parts of $\lambda$ are in $S$ } ),

where the characteristic function \chi(A) = 1 if A is true and 0 if A is false. So (3.2) is just a brute force count of allowable partitions of size n as one runs through the unrestricted partitions of n.

Example

Example 3.3

ZBS3

Recall that if \overline{p}(n) denotes the number of overpartitions [cl04] of n, then

\sum_{n=0}^\infty \overline{p}(n) q^n = \prod_{n=1}^\infty \frac{1+q^n}{1-q^n} = \prod_{n=1}^\infty \frac{1}{(1-q^n)^{a_n}}

where

a_n = \begin{cases} 1 &\mbox{if $n$ is even} \\ 2 &\mbox{if $n$ is odd} \end{cases} .

Thus we conclude that

\overline{p}(n) = \sum_{\lambda\vdash n} (m_1 + 1) (m_3 + 1) (m_5 + 1) \cdots .

Example

Example 3.4

KSPH

Fix a positive integer k. If p_k(n) denotes the number of k-color partitions of n, then

\sum_{n=0}^\infty p_k(n) q^n =\prod_{n=1}^\infty \frac{1}{(1-q^n)^{k}}.

Thus we conclude that

p_k(n) = \sum_{\lambda\vdash n} \frac{(k)_{m_1} (k)_{m_2} \cdots} {m_1!\ m_2! \cdots} = \sum_{\lambda\vdash n} \binom{k+m_1 - 1}{k-1} \binom{k+m_2 - 1}{k-1} \binom{k+m_3 - 1}{k-1} \cdots.

Example

Example 3.5

6G6U

If PL(n) denotes the number of plane partitions of n, then

\sum_{n=0}^\infty PL(n) q^n =\prod_{n=1}^\infty \frac{1}{(1-q^n)^{n}}.

Thus we conclude that

PL(n) = \sum_{\lambda\vdash n} \frac{(1)_{m_1} (2)_{m_2} (3)_{m_3} \cdots} {m_1!\ m_2! \ m_3! \cdots} = \sum_{\lambda\vdash n} \binom{m_2+1}{1} \binom{m_3+2}{2} \binom{m_4+3}{3} \cdots.

Example

Example 3.6

OSSK

The k-broken diamond partitions are defined by Andrews and P. Paule in [ap07]. If r_k(n) denotes the number of k-broken diamond partitions of n, then

\sum_{n=0}^\infty r_k(n) q^n = \prod_{n=1}^\infty \frac{1}{(1-q^n)^{a_n}},

where

a_n = \begin{cases} 2 &\mbox{if $n$ is even or $n\equiv (2k+1)\pmod{4k+2}$} \\ 3 &\mbox{otherwise} \end{cases} .

Thus

r_k(n) =\sum_{\lambda\vdash n} \prod_{j=0}^\infty \frac{ (2+m_{2j+1}) (1+m_{2j+2}) (1+m_{(4k+2)j+(2k+1)}) } { {(2+m_{(4k+2)j+(2k+1)})} } .

Remark.

Further results along similar lines are noted in Appendix D of [Schneider_PhD].

Remark

Remark Remark

HG0A

Further results along similar lines are noted in Appendix D of [Schneider_PhD].

Lemma

Lemma 4.1

X90J

We have that

\begin{gathered} n \sum_{\lambda\vdash n} \frac{ (-1)^{\ell(\lambda)-1} (\ell(\lambda)- 1)! \ r(1)^{m_1(\lambda)} r(2)^{m_2(\lambda)} \cdots } { m_1(\lambda)! m_2(\lambda)! \cdots } \\ = \sum_{j=1}^n j r(j) \sum_{\mu\vdash n-j} \frac{ (-1)^{\ell(\mu)} \ell(\mu)! \ r(1)^{ m_1(\mu)} r(2)^{ m_2(\mu)} \cdots } { m_1(\mu)! m_2(\mu)! \cdots }. \end{gathered}

Used by Example 5.1, Example 5.2

Proof of Lemma 4.1

Let \lambda be an arbitrary partition of size n. Thus we have

\begin{aligned} n &= \sum_{j=1}^n j m_j(\lambda)\\ & = \sum_{j=1}^n j \frac{m_j(\lambda)!}{(m_j(\lambda) - 1)!} \mbox{ (where we follow the convention that $\frac{0!}{(-1)!} := 0$)} \\ & = \sum_{j=1}^n j \frac {m_1(\lambda)! \cdots m_{j-1}(\lambda)! m_j(\lambda)! m_{j+1}(\lambda)! \cdots m_n(\lambda)! } { m_1(\lambda)! \cdots m_{j-1}(\lambda)! (m_j(\lambda)-1)! m_{j+1}(\lambda)! \cdots m_n(\lambda)! } , \end{aligned}

which implies

\frac{ n(-1)^{\ell(\lambda)-1} (\ell(\lambda) - 1)! }{ m_1(\lambda)! \cdots m_n(\lambda)!} \\ = \sum_{j=1}^n j \frac{ (-1)^{m_1(\lambda) + \cdots m_n(\lambda) - 1} (m_1(\lambda) + \cdots m_n(\lambda)-1)! } { m_1(\lambda)! \cdots m_{j-1}(\lambda)! (m_j(\lambda)-1)! m_{j+1}(\lambda)! \cdots m_n(\lambda)! }.

Upon multiplying both sides by r(1)^{m_1(\lambda)} r(2)^{m_2(\lambda)} \cdots and then summing over all partitions of n, the result follows.

Example

Example 5.1

KVCP

Let c(n) denote the number of compositions of n. It is well known that

c(n) = \begin{cases} 1 &\mbox{if $n=0$} \\ 2^{n-1} &\mbox{if $n>0$} \end{cases} ,

and that

\sum_{n=0}^\infty c(n) q^n = \frac{1-q}{1-2q},

but as a consequence of Theorem 1.2, we further conclude that

\sum_{n=0}^\infty c(n) q^n = \prod_{n=1}^\infty \frac{1}{(1-q^n)^{a_n}},

where

a_n = \frac{1}{n} \sum_{d\mid n} \mu\left( \frac{n}{d} \right) (2^d - 1),

i.e., the sequence \{ a_n\}_{n=1}^\infty = (1,1,2,3,6,9,18,30,56,99,186,335, 630, \dots), which is OEIS sequence A059966 [oeisA059966]. OEIS indicates that the a_n are the Lie analog of the sequence of partition numbers, which gives the dimensions of the homogeneous polynomials with one generator in each degree, and several other interpretations.

To see that (5.1) holds, apply Theorem 1.2 with r(n) = c(n) = 2^{n-1} if n>0 and r(0) = c(0) = 1. Thus we need to show that

d\sum_{\lambda\vdash d} \frac{(-1)^{\ell - 1} (\ell - 1)! (2^0)^{m_1} (2^1)^{m_2} (2^2)^{m_3} \cdots} {m_1! \ m_2! \ m_3! \cdots } = 2^d - 1.

Apply Lemma 4.1 with n = d to the left side of (?):

\begin{aligned} & \phantom{ = }d\sum_{\lambda\vdash d} \frac{(-1)^{\ell - 1} (\ell - 1)! (2^0)^{m_1} (2^1)^{m_2} (2^2)^{m_3} \cdots} {m_1! \ m_2! \ m_3! \cdots } \\ & = \sum_{j=1}^d j \ 2^{j-1} \sum_{\mu \vdash d-j} \frac{(-1)^\ell \ \ell! \ (2^0)^{m_1} (2^1)^{m_2} (2^2)^{m_3} \cdots}{m_1! \ m_2! \ m_3! \cdots } \\ & = \sum_{j=1}^d j \ 2^{j-1} \cdot \mbox{ coeff of } q^{d-j} \mbox{ in } \frac{1}{1+2q+4q^2+8q^3 +\cdots} \\ & = \sum_{j=1}^d j \ 2^{j-1} \cdot \mbox{ coeff of } q^{d-j} \mbox{ in } \frac{1-2q}{1-q} \\ & = \sum_{j=1}^d j \ 2^{j-1} \cdot \mbox{ coeff of } q^{d-j} \mbox{ in } 1 -q -q^2 -q^3 - q^4 - \cdots \\ & = d\ 2^{d-1} - \sum_{j=1}^{d-1} j\ 2^{j - 1} \\ & = 2^d - 1, \end{aligned}

where (?) follows from the preceding line by [ss22, p. 3, Lemma 1], and the last equality follows by induction on d.

Uses Theorem 1.2, Lemma 4.1

Example

Example 5.2

XGOR

Let F_n denote the nth Fibonacci number, where

F_1 = F_2 = 1, \mbox{ and } F_n = F_{n-1} + F_{n-2} \mbox{ for $n>2$ }.

It is well known that

\sum_{n=0}^\infty F_{n+1} q^n = \frac{1}{1-q-q^2},

but as a consequence of Theorem 1.2, we also see that

\sum_{n=0}^\infty F_{n+1} q^n = \prod_{n=1}^\infty \frac{1}{(1-q^n)^{a_n}},

where

a_n = \frac{1}{n} \sum_{d\mid n} \mu\left( \frac{n}{d} \right) L_{d},

where L_1 = 1, L_2 = 3, L_{n} = L_{n-1} + L_{n-2} for n>2 (the Lucas numbers). The sequence \{ a_n\}_{n=1}^\infty = (1, 1, 1, 1, 2, 2, 4, 5, 8, 11, 18, 25, 40, 58, 90, 135, 210, \dots) is OEIS sequence A006206 [oeisA006206].

To see that (5.2) holds, note that the d=1 and d=2 cases can be easily verified by direct computation. For d>2, we start by applying Theorem 1.2 with r(n) = F_{n+1}, and observe that we need to show that

d \sum_{\lambda\vdash d} \frac{(-1)^{\ell -1} (\ell - 1)! \ F_2^{m_1} F_3^{m_2} F_4^{m_3} \cdots } {m_1! \ m_2! \ m_3! \cdots } = L_d.

Apply Lemma 4.1 with n = d to the left side of (?):

\begin{aligned} & \phantom{ = }d\sum_{\lambda\vdash d} \frac{(-1)^{\ell - 1} (\ell - 1)! F_2^{m_1} F_3^{m_2} F_4^{m_3} \cdots} {m_1! \ m_2! \ m_3! \cdots } \\ & = \sum_{j=1}^d j \ F_{j+1} \sum_{\mu \vdash d-j} \frac{(-1)^\ell \ \ell! \ F_2^{m_1} F_3^{m_2} F_4^{m_3} \cdots}{m_1! \ m_2! \ m_3! \cdots } \\ & = \sum_{j=1}^d j \ F_{j+1} \cdot \mbox{ coeff of } q^{d-j} \mbox{ in }\frac{1} {F_1 + F_2 q + F_3 q^2 + F_4 q^3 + \cdots} \\ & = \sum_{j=1}^d j \ F_{j+1} \cdot \mbox{ coeff of } q^{d-j} \mbox{ in } (1-q-q^2) \\ & = \sum_{j=1}^d j \ F_{j+1} \cdot \begin{cases} 1 &\mbox{if $j = d$ }, \\ -1 &\mbox{if $j =d-1$ or $d-2$, }\\ 0 &\mbox{otherwise; } \end{cases} \\ & = -(d-2)F_{d-1} - (d-1)F_d + d F_{d+1} \\ & = -F_d - (d-2)(F_{d-1} + F_d) + 2 F_{d+1} + (d-2)F_{d+1} \\ & = -F_d - (d-2)F_{d+1} + 2F_{d+1} + (d-2)F_{d+1} \\ & = -F_d + F_{d+1} + F_{d+1} \\ & = F_{d-1} + F_{d+1} \\ & = L_d, \end{aligned}

where the equality at line (?) follows by [ss22, p. 3, Lemma 1].

Uses Theorem 1.2, Lemma 4.1

Theorem

Theorem 6.1

EZ67

If r(n) and a_n are defined as in equation (?) such that a_n\geq 0 is satisfied for all n\geq 0, then we have

\sum_{n =0}^\infty r_q(n) z^n = \prod_{n =1}^\infty \frac{1}{(z^n; q)_{a_n}}.

Proof of Theorem 6.1

Expanding the product

\prod_{n=1}^\infty \left(1 + \left[ \begin{matrix} a_n - 1 + 1 \\ 1 \end{matrix} \right]_{\displaystyle q} z^n + \left[ \begin{matrix} a_n - 1 + 2 \\ 2 \end{matrix} \right]_{\displaystyle q} z^{2n} + \left[ \begin{matrix} a_n - 1 + 3 \\ 3 \end{matrix} \right]_{\displaystyle q} z^{3n} + \cdots \right),

we obtain the right hand side of (?). Now, using the well-known q-series identity

\sum_{k=0}^\infty (qz)^k \left[ \begin{matrix} a - 1 + k \\ k \end{matrix} \right]_{\displaystyle q} = \frac{1}{(qz;q)_{a}}

with a\geq 0, we make the substitutions z \mapsto z^{a_n}q^{-1} and a \mapsto a_n\geq 0, take the product over n\geq 1, then compare with (Proof), to arrive at the result.

Source and licence

The text on this page is from arXiv:2501.18744v2 by Robert Schneider, Andrew V. Sills and Hunter Waldron, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2501/2501.18744.jsonl.