Formulas for the generalized Frobenius number of triangular numbers (Q7290490)

The purpose of this paper is to provide a confirmation of Komatsu's conjecture on triangular numbers $t_n=\frac{n(n+1)}{2}$ from [the reviewer, Ann. Comb. 26, No. 3, 757--779 (2022; Zbl 1498.11093)]. Let $A=(a_1,a_2,\dots,a_k)$ be a $k$-tuple of positive integers with $\gcd(a_1,a_2,\dots,a_k)=1$. For a non-negative integer $n$, let $d(n;A)=d(n;a_1,a_2,\dots,a_k)$ be the number of representations of $a_1 x_1+a_2 x_2+\cdots+a_k x_k=n$. For a given integer $s\ge 0$, define the generalized Frobenius number by \N\[\Ng(A;s)=g(a_1,a_2,\dots,a_k;s)=\max\{n\in\mathbb Z\mid d(n;A)\le s\}, \N\]\Nthat is the largest integer such that the number of representation by $a_1,a_2,\dots,a_k$ is at most $s$ ways. Notice that $g(a_1,a_2,\dots,a_k)=g(a_1,a_2,\dots,a_k;0)$ is the original Frobenius number. Komatsu formulated a conjecture for the explicit formula for $s\ge 0$ as follows. For some non-negative integer $s$, there exists an odd integer $q$ and integers $n_j$ ($j=1,2,3,4$) such that \N\[\Ng(t_n,t_{n+1},t_{n+2};s)+1=\begin{cases} \frac{(q n)(n+1)(n+2)}{4}&\text{ for even }n\ge n_2,\\\N\frac{(q n-3)(n+1)(n+2)}{4}&\text{ for odd }n\ge n_1,\end{cases}\N\]\Nand \N\[\Ng(t_n,t_{n+1},t_{n+2};s+1)+1=\begin{cases} \frac{(q n+6)(n+1)(n+2)}{4}&\text{ for even }n\ge n_4,\\\N\frac{(q n+3)(n+1)(n+2)}{4}&\text{ for odd }n\ge n_3, \end{cases}\N\]\N and so on. For some non-negative integer $s'$, there exists an even integer $q'$ and integers $n_j$ ($j=5,6$) such that \N\[\Ng(t_n,t_{n+1},t_{n+2};s')+1= \frac{(q' n)(n+1)(n+2)}{4}\quad(n\ge n_5) \N\]\Nand \N\[\Ng(t_n,t_{n+1},t_{n+2};s'+1)+1= \frac{(q' n+6)(n+1)(n+2)}{4}\quad(n\ge n_6) \N\]\Nand so on.\N\NThis paper can determine $g(t_n,t_{n+1},t_{n+2};s+1)$ from $g(t_n,t_{n+1},t_{n+2};s)$ for many integers $s\ge 0$ (or $s'\ge 0$) and the difference between those two is $6(n+1)(n+2)/4$. The first result of this paper presents an explicit formula for the generalized Frobenius number involving three consecutive triangular numbers for all $s\ge 0$. For even $n>6\lfloor{\sqrt{s+1}\rfloor}-6$, we have \N\[\Ng(t_n,t_{n+1},t_{n+2};s)=\frac{(n+1)(n+2)}{4}(q_s n+6 c_s)-1. \N\]\NFor odd $n>6\lfloor{(\sqrt{4 s+5}-1)/2\rfloor}-3$, we have \N\[\Ng(t_n,t_{n+1},t_{n+2};s)=\frac{(n+1)(n+2)}{4}(q_s n+6 c_s-3\delta_s)-1.\N\]\NHere, \N\[\Nq_s=2\lfloor{\sqrt{s}\rfloor}+2+\delta_s,\quad c_s=s-\lfloor{\sqrt{s}\rfloor}^2-\delta_s\lfloor{\sqrt{s}\rfloor},\quad \delta_s=\begin{cases} 1&\text{ if }s\ge\lfloor{\sqrt{s}\rfloor}^2+\lfloor{\sqrt{s}\rfloor},\\\N0&\text{otherwise}. \end{cases}\N\]\NThe second result confirms Komatsu's conjecture and provides a precise statement for the above-mentioned phenomena. For non-negative integers $s$, suppose that \N\[\Nn>\begin{cases} 6\lfloor{\sqrt{s+2}\rfloor}-6&\text{ if }n\text{ is even},\\\N6\lfloor{\frac{\sqrt{4s+9}-1}{2}\rfloor}-3&\text{ if }n\text{ is odd}. \end{cases} \N\]\NIf $s+1$ is in the form of neither $k^2$ nor $k(k+1)$, \N\[\Ng(t_n,t_{n+1},t_{n+2};s+1)-g(t_n,t_{n+1},t_{n+2};s)=\frac{6(n+1)(n+2)}{4}.\N\]\NIf $n$ is even and $s+1$ is in the form of either $k^2$ or $k(k+1)$, \N\[\Ng(t_n,t_{n+1},t_{n+2};s+1)-g(t_n,t_{n+1},t_{n+2};s)=\frac{(n-6k+6)(n+1)(n+2)}{4}.\N\]\NIf $n$ is odd and $s+1$ is in the form of either $k^2$ or $k(k+1)$, \N\[\Ng(t_n,t_{n+1},t_{n+2};s+1)-g(t_n,t_{n+1},t_{n+2};s)=\begin{cases} \frac{(n-6k+9)(n+1)(n+2)}{4}&\text{ if }s+1=k^2,\\\N\frac{(n-6k+3)(n+1)(n+2)}{4}&\text{ if }s+1=k(k+1). \end{cases}\N\]