The \(\mathbf p\)-adic duality for the finite star-multiple polylogarithms (Q2418803)

For an index ${\mathbf{k}}=(k_1,\dots,k_m), k_i\geq 1$, the truncated multiple harmonic sums $\zeta_n(\mathbf{k})$ and $\zeta_n^{\star}(\mathbf{k})$ are defined by \[ \zeta_n({\mathbf{k}})=\sum_{n\geq n_1>\dots>n_m\geq 1}\frac{1}{n_1^{k_1}\dots n_m^{k_m}},\quad \zeta_n^\star({\mathbf{k}})=\sum_{n\geq n_1\geq\dots\geq n_m\geq 1}\frac{1}{n_1^{k_1}\dots n_m^{k_m}} \] respectively. Denote by $\mathcal{A}=\left(\prod_{p}\mathbb{Z}/p\mathbb{Z}\right)/\left( \bigoplus \mathbb{Z}/p\mathbb{Z}\right)$, where $p$ runs through all the prime numbers. The finite multiple zeta values and finite multiple zeta star values are defined by \[ \zeta_{\mathcal{A}}({\mathbf{k}})=( \zeta_{p-1}({\mathbf{k}}))_p, \zeta_{\mathcal{A}}^{\star}({\mathbf{k}})=( \zeta_{p-1}^{\star}({\mathbf{k}}))_p \] in the algebra $\mathcal{A}$ respectively. For an index ${\mathbf{k}}=(k_1,\dots,k_m), k_i\geq 1$, $wt({\mathbf{k}})=k_1+\dots+k_m=w$. Let $\mathbb{F}_2=\mathbb{Z}/2\mathbb{Z}$, then $\mathbf{k}$ corresponds to a vector $I_{\mathbf{k}}=(\{0\}^{k_1-1},1,\{0\}^{k_2-1}, 1,\dots, \{0\}^{k_m-1},1)$ in the $\mathbb{F}_2$-vector space $(\mathbb{F}_2)^w$. The Hoffman dual ${\mathbf{k}}^{{}^\vee}$ of $\mathbf{k}$ is the unique index which satisfies \[ I_{\mathbf{k}}+I_{\mathbf{k}^{{}^{\vee}}}=(\{1\}^{w-1},0). \] \textit{M. E. Hoffman} [Kyushu J. Math. 69, No. 2, 345--366 (2015; Zbl 1382.11066)] proved the following duality relation \[ \zeta_{\mathcal{A}}^{\star}(\mathbf{k})=-\zeta_{\mathcal{A}}^{\star}({\mathbf{k}}^{{}^{\vee}}) \] for any index ${\mathbf{k}}$. In this paper, the author extends the above result to more general setting. Denote by $\widehat{\mathcal{A}}={\mathrm{lim}}_{\leftarrow n}\left(\prod_{p}\mathbb{Z}/p^n\mathbb{Z}\right)/\left( \bigoplus \mathbb{Z}/p^n\mathbb{Z}\right)$, there is a natural projection $\pi: \prod_p \mathbb{Z}_p\rightarrow \widehat{\mathcal{A}}$. Define $\zeta_{\widehat{\mathcal{A}}}({\mathbf{k}})=\pi\left((\zeta_{p-1}({\mathbf{k}}))_p \right)$ and $\zeta^{\star}_{\widehat{\mathcal{A}}}({\mathbf{k}})=\pi\left((\zeta^{\star}_{p-1}({\mathbf{k}}))_p \right)$ . Furthermore, define \[ \mathcal{L}^{\mathrm{III},\star}_{n,{\mathbf{k}}}({\mathbf{t}})=\sum_{n\geq n_1\geq \dots \geq n_m\geq 1}\frac{t_1^{n_1-n_2}\dots t_{m-1}^{n_{m-1}-n_m}t_m^{n_m}}{n_1^{k_1}\dots n_m^{k_m}}. \] Let $\mathcal{L}^{\mathrm{III},\star}_{\widehat{\mathcal{A}},{\mathbf{k}}}({\mathbf{t}})=\pi\left( (\mathcal{L}^{\mathrm{III},\star}_{p-1,{\mathbf{k}}}({\mathbf{t}}))_p\right)$ (view $\mathbf t$ as formal variables) and ${\mathbf p}=\pi((p)_p)$, denote by \[ \widetilde{\mathcal{L}}^{\star}_{\widehat{\mathcal{A}},{\mathbf k}}(t)=\mathcal{L}^{\mathrm{III},\star}_{\widehat{\mathcal{A}},{\mathbf{k}}}(\{1\}^{m-1},t), \] then the main results of this paper are: Theorem 1. (The $\mathbf{p}$-adic duality theorem for FMZSVs) Let $\mathbf k$ be an index, then \[ \sum_{i=0}^{\infty}\zeta^{\star}_{\widehat{\mathcal{A}}}(\{1\}^i,{\mathbf k}){\mathbf p}^i=-\sum_{i=0}^{\infty}\zeta^{\star}_{\widehat{\mathcal{A}}}(\{1\}^i,{\mathbf k}^{{}^{\vee}}){\mathbf p}^i \] in the ring $\widehat{\mathcal{A}}$. Theorem 2. Let $\mathbf k$ be an index, then \[ \sum_{i=0}^{\infty}\left( \widetilde{\mathcal{L}}^{\star}_{\widehat{\mathcal{A}},(\{1\}^i,{\mathbf k})}(t)-\frac{1}{2}\zeta^{\star}_{\widehat{\mathcal{A}}}(\{1\}^i,{\mathbf k})\right){\mathbf p}^i=\sum_{i=0}^{\infty}\left(\widetilde{\mathcal{L}}^{\star}_{\widehat{\mathcal{A}},(\{1\}^i,{\mathbf k}^{{}^{\vee}})}(1-t) -\frac{1}{2}\zeta^{\star}_{\widehat{\mathcal{A}}}(\{1\}^i,{\mathbf k}^{{}^{\vee}})\right){\mathbf p}^i \] in the ring $\widehat{\mathcal{A}}_{\mathbb{Z}[t]}=\mathrm{lim}_{\leftarrow n}\left(\prod_p \mathbb{Z}/p^n\mathbb{Z}[t] \right)/\left(\bigoplus _p\mathbb{Z}/p^n\mathbb{Z}[t] \right)$. Theorem 3. Define a $\mathbf p$-adically convergent series $L_{\widehat{\mathcal{A}},{\mathbf k}}({\mathbf t})$ by \[ L^{\star}_{\widehat{\mathcal{A}},{\mathbf k}}({\mathbf t})=\sum_{i=0}^{\infty}\left( \mathcal{L}^{\mathrm{III},\star}_{\widehat{\mathcal{A}},(\{1\}^i,{\mathbf k})}({\{1\}}^i,{\mathbf t}\})-\frac{1}{2}\mathcal{L}^{\mathrm{III},\star}_{\widehat{\mathcal{A}},({\{1\}}^i,{\mathbf k})}({\{1\}}^i,{\mathbf t}_1)\right){\mathbf p}^i, \] where ${\mathbf t}_1=(t_1,\dots,t_{m-1},1)$. Let $r$ be a positive integer, ${\mathbf k}_1, \dots, {\mathbf k}_r$ indices and ${\mathbf t}=(t_1,\dots, t_r)$ a tuple of indeterminates. Define ${\mathbf k}=({\mathbf k}_1,\dots, {\mathbf k}_r)$ and ${\mathbf k}^\prime=({\mathbf k}_1^{{}^{\vee}},\dots, {\mathbf k}^{{}^{\vee}}_r)$. Furthermore, define $l_i=dep({\mathbf k}_i) $ and $l_i^{\prime}=dep({\mathbf k}_i^{{}^{\vee}})$ respectively for $i=1,\dots,r$. Then we have \[ L^{\star}_{\widehat{\mathcal{A}},{\mathbf k}}( \{1\}^{l_1-1},t_1,\dots,\{1\}^{l_r-1},t_r)= L^{\star}_{\widehat{\mathcal{A}},{\mathbf k}^\prime}( \{1\}^{l_1^\prime-1},1-t_1,\dots,\{1\}^{l_r^\prime-1},1-t_r) \] in the ring $\widehat{\mathcal{A}}_{\mathbb{Z}}[\mathbf t]$. Theorem 1 and Theorem 2 are special cases of Theorem 3. The following two formulas \[ (-1)^n \binom{p-1}{n}=(-1)^{p-1}(1-\frac{p}{n})\sum_{i=0}^{\infty}\sum_{p-1\geq m_1\geq \dots\geq m_i\geq n}\frac{p^i}{m_1\dots m_i}, \] \begin{multline*} \sum_{N\geq n_1\geq\dots \geq n_w\geq 1}(-1)^{n_1}\binom{N}{n_1}\frac{t_1^{n_1-n_2}\dots t_{w-1}^{n_{w-1}-n_w}(t_w^{n_w}-\frac{1}{2})}{n_1\dots n_w}\\ =\sum_{N\geq n_1\geq \dots \geq n_w\geq 1}\frac{(1-t_1)^{n_1-n-2}\dots (1-t_{w-1})^{n_{w-1}-n_w}\{(1-t_w)^{n_w}-\frac{1}{2} \}}{n_1\dots n_w} \end{multline*} play important roles to prove Theorem 3.