---
title: Generalized Cunningham Chains
url: https://www.emergentmind.com/topics/generalized-cunningham-chains
type: topic
---

# Generalized Cunningham Chains

Searching arXiv for recent papers on generalized Cunningham chains and related formulations.
Generalized Cunningham chains extend the classical Cunningham-chain recursion from the special prime sequence \(p,2p+\epsilon,4p+3\epsilon,\dots\) to broader dynamical settings. In the most direct recent formulation, one studies the orbit \(\{f^n(z)\}_{n\ge 0}\) of a linear polynomial \(f(x)=ax+b\) with integer coefficients and asks for consecutive iterates that are prime; a further extension allows the initial value \(z\) itself to be composite and begins the chain at \(f(z)\) rather than at \(z\) [2508.18305]. A different but closely related generalization replaces integers by polynomials in \(\mathbb Z[x]\) and primality by irreducibility over \(\mathbb Q[x]\), leading to polynomial Cunningham chains governed by \(f_i(x)=x f_{i-1}(x)+\epsilon\) [1104.1579]. Across these settings, the subject combines affine dynamics, modular obstructions, irreducibility theory, and conjectural prime-distribution phenomena.

## 1. Classical pattern and modern generalizations

A classical Cunningham chain is a sequence of primes \(p_1,p_2,p_3,\dots\) satisfying
\[
p_i=2p_{i-1}+\epsilon,\qquad \epsilon\in\{-1,1\}.
\]
If \(\epsilon=1\), it is a chain of the first kind; if \(\epsilon=-1\), it is a chain of the second kind. The length is the largest \(k\) such that \(p_1,\dots,p_k\) are prime and \(p_{k+1}=2p_k+\epsilon\) is composite. In the integer setting, such chains are necessarily finite, and it is conjectured that for every positive integer \(k\), there are infinitely many Cunningham chains of length \(k\) [1104.1579].

Recent work generalizes this recurrence by replacing the map \(x\mapsto 2x+\epsilon\) with an arbitrary linear polynomial \(f(x)=ax+b\), with \(a,b\in\mathbb Z\), \(a>1\), and in the main results \(a,b>0\) and \(\gcd(a,b)=1\). The resulting orbit is
\[
\{f^n(z)\}_{n\ge 0},
\]
where
\[
f^n(z)=a^n z+b\,\frac{a^n-1}{a-1}.
\]
The classical first-kind case is recovered by \(f(x)=2x+1\) [2508.18305].

A second enlargement is the rooted formulation. For integers \(z,a,b\) with \(\gcd(a,b)=1\), a rooted Cunningham chain is
\[
\{f(z),f^2(z),\dots,f^{\ell(z)}(z)\},
\]
such that every listed term is prime, \(f^{\ell(z)+1}(z)\) is composite, and \(z\) itself is not required to be prime [2508.18305]. This removes a persistent restriction from the classical presentation: the root need not belong to the prime-valued segment of the orbit.

| Setting | Recurrence or orbit | Stopping condition |
|---|---|---|
| Classical integer chain | \(p_i=2p_{i-1}+\epsilon\) | first composite \(p_{k+1}\) |
| Linear-map generalized chain | \(f(x)=ax+b\), orbit \(\{f^n(z)\}\) | first composite iterate |
| Rooted chain | \(\{f(z),\dots,f^{\ell(z)}(z)\}\) | \(f^{\ell(z)+1}(z)\) composite |
| Polynomial chain | \(f_i(x)=x f_{i-1}(x)+\epsilon\) | first reducible \(f_{k+1}(x)\) |

## 2. Rooted chains for linear maps

The central quantitative object in the linear-map setting is the chain-length function \(\ell(z)\), defined as the maximal \(n\) such that the first \(n\) iterates after the root are prime:
\[
f(z),f^2(z),\dots,f^{\ell(z)}(z).
\]
Under the hypotheses \(a,b>0\), \(\gcd(a,b)=1\), and \(a>1\), two bounds are established. First, if \(z>1\) is coprime to \(a\), then
\[
\ell(z)<z.
\]
Second, there exists \(M\in\mathbb Z^+\) such that for all \(z>M\),
\[
\ell(z)<z.
\]
The second statement is the eventual bound that depends only on the size of \(z\), not on its prime factorization [2508.18305].

The proof of the coprime case is modular. Using
\[
f^n(z)=a^n z+b\,\frac{a^n-1}{a-1},
\]
one selects a prime divisor \(p\) and forces some iterate to vanish modulo \(p\). If \(p\mid(a-1)\), then for \(n=p\),
\[
f^p(z)\equiv 0 \pmod p,
\]
so \(f^p(z)\) is composite and \(\ell(z)<p\le z\). If \(p\nmid(a-1)\), Fermat’s little theorem yields
\[
a^{p-1}\equiv 1\pmod p
\]
and hence
\[
p\mid \frac{a^{p-1}-1}{a-1},
\]
so
\[
f^{p-1}(z)\equiv 0\pmod p,
\]
which gives \(\ell(z)<p-1<z\) [2508.18305]. An explicit remark in the paper weakens the formal hypothesis of the theorem: the argument needs only a prime divisor \(p\) of \(z\) such that \(p\nmid a\), not the full condition \(\gcd(z,a)=1\).

The eventual theorem addresses the remaining case, namely roots \(z\) all of whose prime factors divide \(a\). For this, the paper introduces the auxiliary sequence
\[
s_n=z-b\,\frac{a^n-1}{a-1},
\]
with recurrence
\[
s_1=z-b,\qquad s_{n+1}=s_n-a^n b.
\]
A prime divisor of some \(s_n\) then plays the role previously played by a prime divisor of \(z\). If \(a\) has \(k\) distinct prime factors and
\[
z>b+ab+\cdots+a^{k+1}b,
\]
then there exists a prime \(p\) dividing some \(s_i\), \(1\le i\le k+1\), such that \(p\nmid a\). From that point, the same modular strategy applies to an iterate of \(f\), and the paper gives the explicit threshold
\[
M=b+ab+\cdots+a^{k+1}b
\]
for which \(\ell(z)<z\) holds for all \(z>M\) [2508.18305].

## 3. Polynomial Cunningham chains

The polynomial analogue replaces integers by polynomials in \(\mathbb Z[x]\) and prime/composite by irreducible/reducible over \(\mathbb Q[x]\). A polynomial Cunningham chain is a sequence
\[
f_1(x),f_2(x),\dots
\]
such that \(f_i(x)\in\mathbb Z[x]\), \(f_1(x)\) has positive leading coefficient, each \(f_i(x)\) is irreducible in \(\mathbb Q[x]\) up to the stopping point, and
\[
f_i(x)=x f_{i-1}(x)+\epsilon,\qquad \epsilon\in\{-1,1\}.
\]
The chain is of the first kind for \(\epsilon=1\) and of the second kind for \(\epsilon=-1\). Its length is the least \(k\) such that \(f_{k+1}(x)\) is reducible over \(\mathbb Q\) [1104.1579].

The positive leading coefficient condition fixes a sign ambiguity. If it is dropped, then multiplying \(f_1\) by \(-1\) swaps first kind and second kind without changing reducibility behavior [1104.1579]. This is a structural distinction from the integer theory, where sign normalization is not part of the definition.

For the first kind, an explicit family is given. For integers \(m\ge 2\) and \(k\ge 1\),
\[
f_1(x)=m^2x^{k+3}+mx^{k+2}+mx^{k+1}+\cdots+mx+1,
\]
and the recurrence
\[
f_i(x)=x f_{i-1}(x)+1
\]
produces a sequence in which \(f_i\) is reducible if and only if \(i=k+1\). The proof derives
\[
f_n(x)=m^2x^{n+k+2}+m x^{n+k+1}+\cdots+m x^{n+1}+x^n+\cdots+x+1,
\]
and at the stopping time,
\[
f_{k+1}(x)=\bigl(mx^{k+1}+x^k+\cdots+x+1\bigr)\bigl(mx^{k+2}+1\bigr),
\]
which exhibits the unique reducible term [1104.1579].

For the second kind, if \(m,k\) are positive integers with
\[
m^2>k+1,
\]
then
\[
f_1(x)=m^2x-(m^2-k)
\]
and
\[
f_i(x)=x f_{i-1}(x)-1
\]
again yield a chain in which \(f_i\) is reducible if and only if \(i=k+1\). The explicit form is
\[
f_n(x)=m^2x^n-(m^2-k)x^{n-1}-x^{n-2}-\cdots-x-1,
\]
and
\[
f_{k+1}(1)=0,
\]
so \(x-1\) is a factor precisely at the stopping time [1104.1579].

Two corollaries are immediate. For every positive integer \(k\), there exist infinitely many polynomial Cunningham chains of length \(k\) of both kinds. Also, unlike the integer situation, there exist infinitely many polynomial Cunningham chains of infinite length of both kinds [1104.1579]. This is one of the sharpest contrasts between the classical and polynomial settings.

## 4. Irreducibility methods and structural differences

The polynomial theory is not merely formal imitation. Its irreducibility arguments use tools absent from the integer setting. One is the reciprocal polynomial
\[
\widetilde f(x)=x^d f(1/x)
\]
for a polynomial \(f(x)\) of degree \(d\). If \(f(0)\neq 0\), then \(f\) is irreducible over \(\mathbb Q\) if and only if its reciprocal \(\widetilde f\) is irreducible. This permits reductions to more convenient sparse forms [1104.1579].

A second ingredient is the Fried–Schinzel theorem on reducibility of quadrinomials. In the first-kind proof, after multiplying by \(x-1\), the transformed polynomial is
\[
F_n(x)=(x-1)f_n(x)=x^{n+k+3}+(m-1)x^{k+3}+(m^2-m)x-m^2.
\]
The argument shows that all roots of \(F_n\) lie outside the unit circle, that \(F_n\) is not of the exceptional Fried–Schinzel forms, and that \(F_n\) cannot be split into two pieces sharing a nonreciprocal factor. Therefore \(f_n\) is irreducible for all \(n\neq k+1\) [1104.1579].

In the second-kind case, the relevant transform is
\[
F_n(x)=-(x-1)f_n(x)=x^{n+1}+(m^2-k-1)x^2-(2m^2-k)x+m^2.
\]
The proof uses a contradiction arising from a hypothetical pair of reciprocal roots \(\alpha\) and \(1/\alpha\), together with Descartes’ rule of signs and the condition \(m^2>k+1\), to exclude reducibility except at the intended index [1104.1579].

These results show that generalized Cunningham behavior in \(\mathbb Z[x]\) is governed by factorization geometry rather than by modular obstructions alone. A plausible implication is that the polynomial analogue is substantially more flexible because irreducibility over \(\mathbb Q[x]\) admits techniques based on reciprocal structure, root location, and sparse factorization criteria that have no direct prime-number counterpart.

## 5. Related frameworks for estimating chain length

A separate line of work studies classical Cunningham-chain length through generalized Fibonacci sequences. For \(\alpha\ge 3\), the generalized Fibonacci sequence \(\mathcal F_\alpha=\{F_n\}_{n=0}^\infty\) is defined by
\[
F_0=0,\qquad F_1=1,\qquad F_{n+2}=\alpha F_{n+1}+F_n.
\]
The associated divisor function is
\[
{}_{\mathcal F_\alpha}\sigma(n)=\sum_{d\mid n,\ 0<d\in\mathcal F_\alpha}d.
\]
For odd primes \(p,q\), the paper proves that
\[
p=2q+1 \text{ or } 2q-1
\]
is equivalent to
\[
{}_{\mathcal F_\alpha}\sigma^2(F_p)={}_{\mathcal F_\alpha}\sigma(F_q)
\]
for some \(\alpha\ge 3\), and in fact for all \(\alpha\ge 3\) [2205.07650]. This gives an exact arithmetic encoding of the Cunningham step \(q\mapsto 2q\pm 1\).

The same paper defines
\[
\operatorname{ord}_\alpha(n):=\min\{k\ge 0 : {}_{\mathcal F_\alpha}\sigma^k(n)=1\},
\]
and introduces
\[
C_\alpha:=\limsup_{p\to\infty}\frac{\operatorname{ord}_\alpha(F_p)}{\log p}.
\]
If
\[
C_\alpha<\frac1{\log 2},
\]
then
\[
\limsup_{p\to\infty}\frac{l(p)}{\log p}
\le
\frac{C_\alpha}{1-C_\alpha\log 2}.
\]
A further reformulation uses
\[
D_\alpha:=\limsup_{n\to\infty}\frac{\operatorname{ord}_\alpha(n)}{\log\log n},
\]
and if
\[
D_\alpha<\frac1{\log 2},
\]
then the same type of upper bound follows for \(\limsup l(p)/\log p\) [2205.07650]. This reduces an upper-bound problem on prime chains to an iteration problem on natural numbers.

Another modular approach introduces rogueness. For \(i=1,2\) and an odd prime \(p\),
\[
A_i^p=(\mathbb Z/p\mathbb Z)^\times\setminus\{(-1)^i\bmod p\},
\]
and the affine recursion
\[
u_n^i=2u_{n-1}^i+(-1)^{i+1}
\]
on residues is used to define rogue sequences and rogue loops. The principal theorem is
\[
p\in \mathrm{Rog}_i \iff 2 \text{ is not a primitive root modulo } p.
\]
Thus rogueness is independent of the kind [2311.13375]. If
\[
Q_i^n=\{p\in \mathbb P\setminus \mathrm{Rog} : p<n,\ n\not\equiv (-1)^i \pmod p\},
\]
then
\[
|C_i[p]| \le \min_{q\in Q_i^p}\operatorname{ord}_{A_i^q}(p),
\]
and this minimum is always \(<p-1\) [2311.13375]. For \(p=89\), the method gives the exact value
\[
|C_1[89]|=6,
\]
where the older bound only yields \(|C_1[89]|\le 43\) [2311.13375]. A conjectural logarithmic estimate also appears:
\[
|C_i[p]|<\log_2(p+1)\qquad (p>5).
\]

## 6. Cunningham-chain products and applications beyond prime orbits

The Cunningham relation also appears in algebraic settings where the chain itself is not the main object but provides the arithmetic skeleton. One example is the study of separable field extensions of squarefree degree
\[
n=p_1\cdots p_\ell,
\qquad
p_i=2p_{i+1}+1 \ \ (1\le i\le \ell-1),
\]
so that the primes form a Cunningham chain. The paper calls such an \(n\) a Cunningham product [2508.03384].

In that context, the prime relation \(p_i-1=2p_{i+1}\) sharply constrains groups of order \(n\). The paper proves that there are
\[
F(\ell)
\]
groups of order \(n\), where \(F(\ell)\) is the \(\ell\)-th Fibonacci number with the convention \(F(2)=2\). These groups are assembled from cyclic factors \(C_{p_i}\) and possible semidirect products between consecutive primes, with at most one semidirect product in every triple of consecutive factors [2508.03384]. For \(\ell=3\), the three groups are
\[
C_{p_1}\times C_{p_2}\times C_{p_3},\qquad
(C_{p_1}\rtimes C_{p_2})\times C_{p_3},\qquad
C_{p_1}\times (C_{p_2}\rtimes C_{p_3}).
\]

Using the Greither–Pareigis theorem, Byott’s translation theorem, and Byott’s counting formula, the paper classifies transitive subgroups of \(\mathrm{Hol}(N)\) and counts Hopf–Galois structures. In the cyclic-type case \(N\cong C_n\), the transitive subgroups are shown to have the form
\[
J_{I,t}\rtimes A,
\]
where \(I\subseteq\{2,\dots,\ell\}\) contains no consecutive integers and \(A\) is a subgroup of a suitable automorphism group. These structures are all almost classically Galois [2508.03384].

This use of Cunningham chains is not a prime-generation problem. Rather, it shows that the same affine-doubling relation that defines classical and generalized Cunningham chains also governs structural classification in squarefree Hopf–Galois theory. A plausible implication is that Cunningham-chain relations function as a reusable arithmetic template across distinct domains: prime orbits, polynomial irreducibility, modular dynamics, and field-extension symmetry.

## 7. Conceptual significance

Several themes recur across the literature. First, generalized Cunningham chains are best understood as dynamical orbits rather than as isolated prime tuples. In the linear setting this is explicit through iteration of \(f(x)=ax+b\), and in the polynomial setting through \(f_i(x)=x f_{i-1}(x)+\epsilon\) [2508.18305] [1104.1579].

Second, the notion of length behaves differently in different categories. In the integer setting, classical chains are finite. In the linear generalized setting, one obtains the robust upper bound \(\ell(z)<z\) for all sufficiently large \(z\), and for every \(z>1\) with \(\gcd(z,a)=1\) [2508.18305]. In the polynomial setting, by contrast, infinite chains of both kinds exist [1104.1579].

Third, the recent literature separates two distinct generalizations that are sometimes conflated. One is the generalization in the map \(x\mapsto 2x+\epsilon\) to an arbitrary positive linear polynomial \(ax+b\). The other is the rooted generalization in which the root is not required to be prime and the chain begins at the first prime iterate [2508.18305]. A recurrent misconception is that a generalized Cunningham chain must still begin from a prime; the rooted definition explicitly removes that requirement.

Finally, current work places chain-length problems in broader frameworks: generalized Fibonacci divisor iterations produce conditional logarithmic bounds on \(l(p)\) [2205.07650], rogueness links upper bounds to primitive-root phenomena [2311.13375], and Cunningham-chain products organize squarefree Hopf–Galois classifications [2508.03384]. Taken together, these developments show that generalized Cunningham chains form a meeting point between affine prime dynamics, sparse factorization theory, modular order structures, and arithmetic group theory.

Source: https://www.emergentmind.com/topics/generalized-cunningham-chains