Conjecture on congruences for overcubic partition tuples

Less than 1 year old · traced to

Let n≥0n\geq 0, i≥3i\geq 3, and k≥0k\geq 0. The overcubic partition tuple function b‾m(r)\overline{b}_m(r) is considered at the indicated arithmetic progressions. Congruence conjecture. For all n≥0n\geq 0, i≥3i\geq 3, and k≥0k\geq 0,

b‾2ik+2i−1−1(8n+5)≡0(mod2i+2),\overline{b}_{2^i k + 2^{i-1}-1}(8n+5) \equiv 0 \pmod{2^{i+2}},

and

b‾2ik+2i−1−1(8n+7)≡0(mod2i+3).\overline{b}_{2^i k + 2^{i-1}-1}(8n+7) \equiv 0 \pmod{2^{i+3}}.

This conjecture proposes further infinite families of arithmetic congruences for overcubic partition tuples, extending the preceding divisibility results; the authors indicate that it may follow from techniques similar to those used in the paper.

References

Primary source

Saikat Maity and Manjil P. Saikia, “Extending Recent Arithmetic Properties of Overcubic Partition Tuples”, arXiv:2607.07270 (2026).

Progress summary

Refreshed
Claimed solved

The conjecture is recorded as open, but a complete proof has since been posted and has not been independently checked.

Maity and Saikia formulate two infinite divisibility families as Conjecture 1.5 in their July 2026 paper. The claims concern overcubic partition tuples at indices congruent to 55 and 77 modulo 88, with moduli depending on ii.

Known results

  • Sellers (2014) and Shivashankar–Naika (2018): related congruences for tuple size k=1k=1.
  • Lin (2014), Naika–Shivashankar (2017), and Ray–Barman (2020): related results for k=2k=2.
  • Nayaka (2026): related results for k=3k=3.
  • Buragohain–Saikia (2024) and Chaćon–Sellers (2026): further arithmetic families, but not these two conjectured families.

Posted attempt

A complete proof is claimed: it establishes base congruences for b‾3\overline b_3 and then lifts them to the stated indices using a generating-function expansion and 22-adic divisibility estimates. The argument has not been independently verified.

Current status (as of August 2026): the conjecture was published as open, while a complete proof has been posted but remains unverified, so the problem is not mathematically settled.

Sources

Solutions 1

ProofThis solution needs a summarySee full solutionHide full solution

Write

fj=∏m≥1(1−qjm),B(q)=f4f12f2,B(q)t=∑n≥0b‾t(n)qn.f_j=\prod_{m\ge1}(1-q^{jm}),\qquad B(q)=\frac{f_4}{f_1^2f_2},\qquad B(q)^t=\sum_{n\ge0}\overline b_t(n)q^n.

We prove, for every i≥3i\ge3, k,n≥0k,n\ge0, and t=2ik+2i−1−1t=2^ik+2^{i-1}-1, that

b‾t(8n+5)≡0(mod2i+2),b‾t(8n+7)≡0(mod2i+3).\overline b_t(8n+5)\equiv0\pmod{2^{i+2}},\qquad \overline b_t(8n+7)\equiv0\pmod{2^{i+3}}.

All power-series computations take place in Z[[q]]\mathbb Z[[q]].

First, Jacobi's identity gives

A(q):=B(q)−1=φ(−q)φ(−q2)=1+2C(q),A(q):=B(q)^{-1}=\varphi(-q)\varphi(-q^2)=1+2C(q),

where

U=∑j≥1(−1)jqj2,V=∑j≥1(−1)jq2j2,W=U+V,Z=UV,C=W+2Z.U=\sum_{j\ge1}(-1)^jq^{j^2},\quad V=\sum_{j\ge1}(-1)^jq^{2j^2},\quad W=U+V,\quad Z=UV,\quad C=W+2Z.

Since no integer congruent to 55 or 7(mod8)7\pmod8 is represented by x2+2y2x^2+2y^2,

[qN]A=[qN]C=0(N≡5,7(mod8)).(1)[q^N]A=[q^N]C=0 \qquad(N\equiv5,7\pmod8). \tag{1}

We establish the two base congruences

[q8n+5]B3≡0(mod32),[q8n+7]B3≡0(mod64).(2)[q^{8n+5}]B^3\equiv0\pmod{32},\qquad [q^{8n+7}]B^3\equiv0\pmod{64}. \tag{2}

Suppose first that N≡5(mod8)N\equiv5\pmod8, and let

R=#{(x,y)∈Z>02:x,y odd, x2+4y2=N}.R=\#\{(x,y)\in\mathbb Z_{>0}^2:x,y\text{ odd},\ x^2+4y^2=N\}.

Because C2≡W2(mod4)C^2\equiv W^2\pmod4, the UVUV term contributes nothing by (1), and V2V^2 has only even exponents,

[qN]C2≡[qN]U2=−2R(mod4).(3)[q^N]C^2\equiv[q^N]U^2=-2R\pmod4. \tag{3}

Indeed, each representation by two positive squares has one odd entry and one entry 2y2y with yy odd, and its two orders both contribute sign −1-1. Modulo 22, Frobenius gives

C3≡W3≡(U+V)(U(q2)+V(q2))(mod2).C^3\equiv W^3\equiv(U+V)(U(q^2)+V(q^2))\pmod2.

At an odd exponent, the first factor contributes an odd square from UU; the term U(q2)U(q^2) cannot contribute because it would represent NN as x2+2y2x^2+2y^2, whereas V(q2)V(q^2) gives exactly RR. Therefore

[qN]C3≡R(mod2),[qN]C4≡0(mod2).[q^N]C^3\equiv R\pmod2,\qquad [q^N]C^4\equiv0\pmod2.

Expanding B3=(1+2C)−3B^3=(1+2C)^{-3} modulo 3232 and using (1) yields

[qN]B3≡[qN](24C2−80C3+240C4)≡8(3(−2R)+2R)=−32R≡0(mod32).(4)[q^N]B^3\equiv[q^N](24C^2-80C^3+240C^4) \equiv8\bigl(3(-2R)+2R\bigr)=-32R\equiv0\pmod{32}. \tag{4}

Next suppose N≡7(mod8)N\equiv7\pmod8, and let

R=#{(x,y,z)∈Z>03:x,y,z odd, x2+4y2+2z2=N}.R=\#\{(x,y,z)\in\mathbb Z_{>0}^3: x,y,z\text{ odd},\ x^2+4y^2+2z^2=N\}.

The square residues modulo 88 give

[qN]W2=0,[qN]UV2=0,[qN]U2V=2R.(5)[q^N]W^2=0,\qquad[q^N]UV^2=0,\qquad[q^N]U^2V=2R. \tag{5}

In the last equality, the two UU-indices are an odd integer and 2y2y, with yy odd, while the VV-index is odd; both orders contribute sign +1+1. Also [qN]Z2[q^N]Z^2 is even because Z2≡Z(q2)(mod2)Z^2\equiv Z(q^2)\pmod2. Hence

[qN]C2=[qN](W2+4WZ+4Z2)≡0(mod8).(6)[q^N]C^2=[q^N](W^2+4WZ+4Z^2)\equiv0\pmod8. \tag{6}

The only term in W3W^3 contributing at residue 77 is 3U2V3U^2V, so [qN]W3=6R[q^N]W^3=6R. Moreover,

W2Z≡(U(q2)+V(q2))UV(mod2).W^2Z\equiv\bigl(U(q^2)+V(q^2)\bigr)UV\pmod2.

The term U(q2)UVU(q^2)UV cannot contribute at residue 77, since it would require two square residues to sum to 3(mod4)3\pmod4; V(q2)UVV(q^2)UV counts exactly the same triples as RR. Thus

[qN]C3≡[qN](W3+2W2Z)≡6R+2R≡0(mod4).(7)[q^N]C^3\equiv[q^N](W^3+2W^2Z)\equiv6R+2R\equiv0\pmod4. \tag{7}

Writing C2=C(q2)+2DC^2=C(q^2)+2D, one has C4≡C(q2)2(mod4)C^4\equiv C(q^2)^2\pmod4, so all odd-index coefficients of C4C^4 are divisible by 44. Finally,

C5≡W(q)W(q4)(mod2).C^5\equiv W(q)W(q^4)\pmod2.

Its odd exponents have the form x2+4y2x^2+4y^2 or x2+8y2x^2+8y^2, neither of which is 7(mod8)7\pmod8. Hence

[qN]C4≡0(mod4),[qN]C5≡0(mod2).(8)[q^N]C^4\equiv0\pmod4,\qquad[q^N]C^5\equiv0\pmod2. \tag{8}

The expansion modulo 6464 now gives

[qN]B3≡8[qN](3C2−10C3+30C4−84C5)≡0(mod64),[q^N]B^3 \equiv8[q^N](3C^2-10C^3+30C^4-84C^5) \equiv0\pmod{64},

proving both assertions of (2).

For the uniform lifting step, put

M=2ik+2i−1=2i−1(2k+1),L=M/4=2i−3(2k+1),e=v2(L)=i−3.M=2^ik+2^{i-1}=2^{i-1}(2k+1),\qquad L=M/4=2^{i-3}(2k+1),\qquad e=v_2(L)=i-3.

Since A≡B≡1(mod2)A\equiv B\equiv1\pmod2,

X:=B4−1∈8Z[[q]],X:=B^4-1\in8\mathbb Z[[q]],

and

BM−1=A(1+X)L=A+LAX+∑j=2L(Lj)AXj.(9)B^{M-1}=A(1+X)^L =A+LAX+\sum_{j=2}^{L}\binom Lj AX^j. \tag{9}

For every jj,

v2(Lj)≥e−v2(j),v_2\binom Lj\ge e-v_2(j),

by (Lj)=(L/j)(L−1j−1)\binom Lj=(L/j)\binom{L-1}{j-1}. Thus for odd j≥3j\ge3, every coefficient of (Lj)AXj\binom Lj AX^j is divisible by 2e+92^{e+9}. For even j≥2j\ge2, write X=8YX=8Y. Since A≡1(mod2)A\equiv1\pmod2 and YjY^j is a square modulo 22, its odd-index coefficients vanish modulo 22, and consequently

v2([qN](Lj)AXj)≥3j+1+e−v2(j)≥e+6(N odd).(10)v_2\left([q^N]\binom Lj AX^j\right) \ge3j+1+e-v_2(j)\ge e+6 \qquad(N\text{ odd}). \tag{10}

Thus every term with j≥2j\ge2 vanishes modulo 2e+6=2i+32^{e+6}=2^{i+3} at odd indices.

For N≡5,7(mod8)N\equiv5,7\pmod8, (1) and AX=B3−AAX=B^3-A reduce (9) to

b‾M−1(N)≡L b‾3(N)(mod2i+3).(11)\overline b_{M-1}(N) \equiv L\,\overline b_3(N)\pmod{2^{i+3}}. \tag{11}

Since v2(L)=i−3v_2(L)=i-3, the first base congruence in (2) implies divisibility by 2i+22^{i+2} at N=8n+5N=8n+5, and the second implies divisibility by 2i+32^{i+3} at N=8n+7N=8n+7. Both conjectured families therefore hold for all admissible i,k,ni,k,n.