The periodic Hankel transform of 1−x/r(x2)1-x/r(x^2)

At least 4 years old · documented by

Let r(x)r(x) be the generating function of the Rueppel sequence, and let the Hankel transform of a sequence be the sequence of determinants of its successive Hankel matrices. Periodic Hankel-transform conjecture. The Hankel transform of the sequence with generating function 1−x/r(x2)1-x/r(x^2) is the periodic sequence

1,−1,−1,0,1,−1,−1,0,1,−1,−1,0,….1,-1,-1,0,1,-1,-1,0,1,-1,-1,0,\ldots.

The claim is based on the displayed initial Hankel-transform values and is part of the paper's study of Rueppel analogs of Catalan-related sequences; no proof or resolution is given.

References

Primary source

Paul Barry, “Conjectures and results on some generalized Rueppel sequences”, arXiv:2107.00442 (2021).

Progress summary

Refreshed
Claimed solved

A 2020 paper claims to have proved the repeating-pattern conjecture, but neither that claim nor a new submitted proof has been independently checked.

Barry’s 2021 paper states the conjecture that the determinants associated with 1−x/r(x2)1-x/r(x^2) repeat as 1,−1,−1,01,-1,-1,0. It presents computed initial values but no proof.

2020 claimed proof

The paper On some conjectures of P. Barry says it proves Conjectures 66–1111 and 1616, which includes this conjecture. The retrieved record contains no independent verification or reported refutation.

Community submission (unverified), August 23, 2026

A submitted proof argues for the period-four formula using reciprocal Hankel identities, shifted Rueppel determinants, and formal power-series inversion. Its correctness is unverified.

Current status (as of August 2026): The conjecture has a claimed proof in a 2020 paper and an additional unverified submitted proof; neither has been independently verified.

Sources

Solutions 1

ProofThis solution needs a summarySee full solutionHide full solution

Proof of the period-four Rueppel Hankel conjecture

Let

r(x)=∑j≥0rjxj=∑k≥0x2k−1r(x)=\sum_{j\geq0}r_jx^j =\sum_{k\geq0}x^{2^k-1}

be the Rueppel generating function, and write

F(x)=1−xr(x2)=∑j≥0fjxj,Tn=det⁡(fi+j)0≤i,j≤n.F(x)=1-\frac{x}{r(x^2)}=\sum_{j\geq0}f_jx^j, \qquad T_n=\det(f_{i+j})_{0\leq i,j\leq n}.

We prove that

T4j=1,T4j+1=−1,T4j+2=−1,T4j+3=0(j≥0).(1)\boxed{ T_{4j}=1, \qquad T_{4j+1}=-1, \qquad T_{4j+2}=-1, \qquad T_{4j+3}=0 \quad(j\geq0). } \tag{1}

This is Conjecture 10 in Barry, Conjectures and results on some generalized Rueppel sequences.

Reciprocal Hankel identities

Let

A(x)=∑j≥0ajxj,a0=1,Q(x)=1A(x)=∑j≥0qjxj.A(x)=\sum_{j\geq0}a_jx^j, \qquad a_0=1, \qquad Q(x)=\frac1{A(x)}=\sum_{j\geq0}q_jx^j.

For m≥1m\geq1, formal-series inversion gives

det⁡(qi+j+1)0≤i,j<m=(−1)mdet⁡(ai+j+1)0≤i,j<m,det⁡(qi+j)0≤i,j<m=(−1)m−1det⁡(ai+j+2)0≤i,j<m−1.(2)\begin{aligned} \det(q_{i+j+1})_{0\leq i,j<m} &=(-1)^m\det(a_{i+j+1})_{0\leq i,j<m},\\ \det(q_{i+j})_{0\leq i,j<m} &=(-1)^{m-1}\det(a_{i+j+2})_{0\leq i,j<m-1}. \end{aligned} \tag{2}

Here the determinant of a zero-by-zero matrix is 11.

For completeness, let L=(ai−j)0≤j≤i<mL=(a_{i-j})_{0\leq j\leq i<m} be the unit lower triangular Toeplitz matrix, and let U=(qj−i)0≤i≤j<mU=(q_{j-i})_{0\leq i\leq j<m} be its unit upper triangular counterpart. The convolution identity

∑k=0dad−kqk=0(d≥1)(3)\sum_{k=0}^{d}a_{d-k}q_k=0 \qquad(d\geq1) \tag{3}

implies

L(qi+j+1)i,j<m=−(ai+j+1)i,j<mU,L(q_{i+j+1})_{i,j<m} =-(a_{i+j+1})_{i,j<m}U,

which proves the first equality in (2). Similarly, L(qi+j)i,j<mL(q_{i+j})_{i,j<m} has first column (1,0,…,0)T(1,0,\ldots,0)^{\mathsf T}. Deleting that column and the first row leaves

−(ai+j+2)0≤i,j<m−1U′,U′=(qj−i)0≤i≤j<m−1,-(a_{i+j+2})_{0\leq i,j<m-1}U', \qquad U'=(q_{j-i})_{0\leq i\leq j<m-1},

proving the second equality in (2).

Shifted Rueppel determinants

Set

Rm=det⁡(ri+j)0≤i,j<m,Jm=det⁡(ri+j+1)0≤i,j<m,Lm=det⁡(ri+j+2)0≤i,j<m,\begin{aligned} R_m&=\det(r_{i+j})_{0\leq i,j<m},\\ J_m&=\det(r_{i+j+1})_{0\leq i,j<m},\\ L_m&=\det(r_{i+j+2})_{0\leq i,j<m}, \end{aligned}

with R0=J0=L0=1R_0=J_0=L_0=1. The standard Rueppel determinant evaluation, recalled in Barry's introduction, is

Rm=(−1)(m2).(4)R_m=(-1)^{\binom m2}. \tag{4}

The Rueppel coefficients satisfy

r2j={1,j=0,0,j≥1,r2j+1=rj.(5)r_{2j}=\begin{cases}1,&j=0,\\0,&j\geq1,\end{cases} \qquad r_{2j+1}=r_j. \tag{5}

Reordering both rows and columns by parity in the matrix for JmJ_m, its off-diagonal parity blocks vanish. Its diagonal blocks are ordinary and once-shifted Rueppel Hankel matrices. Hence

J2t=RtJt,J2t+1=Rt+1Jt.(6)J_{2t}=R_tJ_t, \qquad J_{2t+1}=R_{t+1}J_t. \tag{6}

It follows inductively from (4) that

Jm∈{−1,1}(m≥0).(7)J_m\in\{-1,1\} \qquad(m\geq0). \tag{7}

For LmL_m, the diagonal parity blocks vanish. Thus an odd-size matrix has determinant zero, whereas an even-size matrix has two identical once-shifted blocks:

L2t+1=0,L2t=(−1)tJt2=(−1)t.(8)L_{2t+1}=0, \qquad L_{2t}=(-1)^tJ_t^2=(-1)^t. \tag{8}

Now specialize (2) to A(x)=r(x)A(x)=r(x). Equations (7) and (8) give

∣det⁡(qi+j+1)0≤i,j<m∣=1,(9)\left|\det(q_{i+j+1})_{0\leq i,j<m}\right|=1, \tag{9}

and

∣det⁡(qi+j)0≤i,j<m∣={1,m odd,0,m even.(10)\left|\det(q_{i+j})_{0\leq i,j<m}\right| = \begin{cases} 1,&m\text{ odd},\\ 0,&m\text{ even}. \end{cases} \tag{10}

The target Hankel determinants

Since

F(x)=1−xQ(x2),F(x)=1-xQ(x^2),

its coefficients satisfy

f0=1,f2j=0(j≥1),f2j+1=−qj.(11)f_0=1, \qquad f_{2j}=0\quad(j\geq1), \qquad f_{2j+1}=-q_j. \tag{11}

Reorder the rows and columns of its Hankel matrix by parity. For an odd-size matrix, expansion along the unique nonzero diagonal-parity entry gives

T2m=(−1)mdet⁡(qi+j+1)0≤i,j<m 2=(−1)m.(12)T_{2m}=(-1)^m \det(q_{i+j+1})_{0\leq i,j<m}^{\,2} =(-1)^m. \tag{12}

For an even-size matrix, the two off-diagonal blocks are square, giving

T2m+1=(−1)m+1det⁡(qi+j)0≤i,j≤m 2={−1,m even,0,m odd.(13)T_{2m+1}=(-1)^{m+1} \det(q_{i+j})_{0\leq i,j\leq m}^{\,2} = \begin{cases} -1,&m\text{ even},\\ 0,&m\text{ odd}. \end{cases} \tag{13}

Combining (12) and (13) proves the full period-four formula (1), namely

(Tn)n≥0=(1,−1,−1,0,1,−1,−1,0,…).(T_n)_{n\geq0} =(1,-1,-1,0,1,-1,-1,0,\ldots).