Divisor-count conjecture for classes of upper k-gap balancing numbers

About 8 years old · traced to

Let k≥0k\geq 0 be an integer. An upper kk-gap balancing number is a balancing number associated with a gap of kk, and solutions are grouped into classes as in the paper. Divisor-count conjecture. The number of classes of upper kk-gap balancing numbers is equal to the number of positive divisors of 2k2−12k^2-1. There appears to be no uniform upper bound on the number of such classes as kk varies; the conjecture is supported by numerical evidence for the values of kk examined in the paper, but no proof is given.

References

Primary source

Jeremiah Bartz, Bruce Dearden and Joel Iiams, “Classes of Gap Balancing Numbers”, arXiv:1810.07895 (2018).

Progress summary

Refreshed
Claimed solved

The original paper left the conjecture open, but a subsequently posted argument claims a complete proof; nobody has independently checked it.

Bartz, Dearden, and Iiams proposed in 2018 that the number of upper kk-gap classes equals the number of positive divisors of 2k2−12k^2-1, based on numerical evidence.

Known results

  • For k>1k>1, there are at least two classes and at most max⁡{1,k}\max\{1,k\}.
  • The formula is known for odd k≥2k\geq 2 when 2k2−12k^2-1 is prime, by Tekcan, Tayat, and Özbek.
  • The original paper gives no general proof and anticipates no uniform upper bound as kk varies.

Posted attempt

A reader-posted argument claims a complete proof: it identifies classes with bounded solutions of a Pell-type equation, counts norm orbits in Z[2]\mathbb{Z}[\sqrt{2}], and claims each orbit has exactly one admissible seed. The argument has not been independently verified.

Current status (as of August 2026): The conjecture remains unverified in general; the posted proof claim is the only reported new development, while the prime-valued special case and numerical evidence are established.

Sources

Solutions 1

ProofThis solution needs a summarySee full solutionHide full solution

Let k≥1k\geq1 and put N=2k2−1N=2k^2-1. By the seed characterization following Proposition 4.1, the number of classes equals the number of pairs (x,y)(x,y) satisfying

0≤x<k,y>0,y2=8x2+8(1−k)x+(2k−1)2.0\leq x<k,\qquad y>0,\qquad y^2=8x^2+8(1-k)x+(2k-1)^2.

With z=2x+1−kz=2x+1-k, this becomes

y2−2z2=N.(1)y^2-2z^2=N. \tag{1}

Conversely, every integer solution of (1) with y>0y>0 and ∣z∣<k|z|<k yields a unique admissible seed. Reduction modulo 88 shows that zz is odd when kk is even and even when kk is odd. Hence z≡1−k(mod2)z\equiv1-k\pmod2, so

x=z+k−12∈Z,x=\frac{z+k-1}{2}\in\mathbb Z,

and ∣z∣<k|z|<k gives 0≤x<k0\leq x<k.

Consider the Euclidean domain R=Z[2]R=\mathbb Z[\sqrt2], with norm

Nm⁡(u+v2)=u2−2v2.\operatorname{Nm}(u+v\sqrt2)=u^2-2v^2.

Every prime p∣Np\mid N is odd and satisfies 2k2≡1(modp)2k^2\equiv1\pmod p. Therefore 22 is a quadratic residue modulo pp, and pp splits in RR as

p=πpπp‾.p=\pi_p\overline{\pi_p}.

Because RR is a principal ideal domain and 1+21+\sqrt2 has norm −1-1, the factors can be chosen with

Nm⁡(πp)=Nm⁡(πp‾)=p.\operatorname{Nm}(\pi_p)=\operatorname{Nm}(\overline{\pi_p})=p.

Writing N=∏ppepN=\prod_p p^{e_p}, unique factorization shows that every element of norm NN has the form

α=u∏p∣Nπpjpπp‾ ep−jp,0≤jp≤ep,\alpha=u\prod_{p\mid N}\pi_p^{j_p} \overline{\pi_p}^{\,e_p-j_p}, \qquad 0\leq j_p\leq e_p,

where uu has norm 11. Distinct exponent vectors give distinct unit orbits. The norm-one units are

±(3+22)m,m∈Z.\pm(3+2\sqrt2)^m,\qquad m\in\mathbb Z.

Restricting to positive elements removes the sign. Thus the positive elements of norm NN split into exactly

∏p∣N(ep+1)=τ(N)\prod_{p\mid N}(e_p+1)=\tau(N)

orbits under multiplication by ε=3+22\varepsilon=3+2\sqrt2.

Each orbit contains exactly one admissible seed. Indeed, since ε=(1+2)2\varepsilon=(1+\sqrt2)^2, there is a unique integer mm such that

N1+2<αεm<N(1+2).(2)\frac{\sqrt N}{1+\sqrt2} < \alpha\varepsilon^m < \sqrt N(1+\sqrt2). \tag{2}

Neither endpoint can occur, since equality would imply z2=N/2z^2=N/2, impossible for odd NN. Writing αεm=y+z2\alpha\varepsilon^m=y+z\sqrt2, positivity of this element and its conjugate N/(αεm)N/(\alpha\varepsilon^m) gives y>0y>0, and

z=αεm−N/(αεm)22.z=\frac{\alpha\varepsilon^m-N/(\alpha\varepsilon^m)}{2\sqrt2}.

Therefore (2) is equivalent to

∣z∣<N/2=k2−12,|z|<\sqrt{N/2}=\sqrt{k^2-\tfrac12},

which for integral zz is exactly ∣z∣<k|z|<k. Hence each positive norm-NN orbit produces exactly one class seed.

The number of classes is consequently

τ(2k2−1).\boxed{\tau(2k^2-1)}.

Finally, k=0k=0 is the ordinary cobalancing case with one class, agreeing with the single positive divisor of −1-1. Thus the formula holds for every k≥0k\geq0.