# §21.6(i) Riemann Identity

Let $\mathbf{T}=[T_{jk}]$ be an arbitrary $h\times h$ orthogonal matrix (that is, $\mathbf{T}\mathbf{T}^{\mathrm{T}}=\mathbf{I}$) with rational elements. Also, let $\mathbf{Z}$ be an arbitrary $g\times h$ matrix. Define

 21.6.1 $\mathcal{K}=\Integer^{g\times h}\mathbf{T}/(\Integer^{g\times h}\mathbf{T}\cap% \Integer^{g\times h}),$ Symbols: $\Integer$: set of all integers, $\cap$: intersection, $g$: positive integer, $h$: positive integer and $\mathcal{K}$: set of matrices Permalink: http://dlmf.nist.gov/21.6.E1 Encodings: TeX, pMML, png

that is, $\mathcal{K}$ is the set of all $g\times h$ matrices that are obtained by premultiplying $\mathbf{T}$ by any $g\times h$ matrix with integer elements; two such matrices in $\mathcal{K}$ are considered equivalent if their difference is a matrix with integer elements. Also, let

 21.6.2 $\mathcal{D}=|\mathbf{T}^{\mathrm{T}}\Integer^{h}/(\mathbf{T}^{\mathrm{T}}% \Integer^{h}\cap\Integer^{h})|,$ Symbols: $\Integer$: set of all integers, $\cap$: intersection, $h$: positive integer and $\mathcal{D}$: number of elements Permalink: http://dlmf.nist.gov/21.6.E2 Encodings: TeX, pMML, png

that is, $\mathcal{D}$ is the number of elements in the set containing all $h$-dimensional vectors obtained by multiplying $\mathbf{T}^{\mathrm{T}}$ on the right by a vector with integer elements. Two such vectors are considered equivalent if their difference is a vector with integer elements. Then

 21.6.3 $\prod_{j=1}^{h}\mathop{\theta\/}\nolimits\!\left(\sum_{k=1}^{h}T_{jk}\mathbf{z% }_{k}\middle|\boldsymbol{{\Omega}}\right)=\frac{1}{\mathcal{D}^{g}}\sum_{% \mathbf{A}\in\mathcal{K}}\sum_{\mathbf{B}\in\mathcal{K}}e^{2\pi i\trace\left[% \frac{1}{2}\mathbf{A}^{\mathrm{T}}\boldsymbol{{\Omega}}\mathbf{A}+\mathbf{A}^{% \mathrm{T}}[\mathbf{Z}+\mathbf{B}]\right]}\*\prod_{j=1}^{h}\mathop{\theta\/}% \nolimits\!\left(\mathbf{z}_{j}+\boldsymbol{{\Omega}}\mathbf{a}_{j}+\mathbf{b}% _{j}\middle|\boldsymbol{{\Omega}}\right),$

where $\mathbf{z}_{j}$, $\mathbf{a}_{j}$, $\mathbf{b}_{j}$ denote respectively the $j$th columns of $\mathbf{Z}$, $\mathbf{A}$, $\mathbf{B}$. This is the Riemann identity. On using theta functions with characteristics, it becomes

 21.6.4 $\prod_{j=1}^{h}\mathop{\theta\!\genfrac{[}{]}{0.0pt}{}{\sum_{k=1}^{h}T_{jk}% \mathbf{c}_{k}}{\sum_{k=1}^{h}T_{jk}\mathbf{d}_{k}}\/}\nolimits\!\left(\sum_{k% =1}^{h}T_{jk}\mathbf{z}_{k}\middle|\boldsymbol{{\Omega}}\right)=\frac{1}{% \mathcal{D}^{g}}\sum_{\mathbf{A}\in\mathcal{K}}\sum_{\mathbf{B}\in\mathcal{K}}% e^{-2\pi i\sum_{j=1}^{h}\mathbf{b}_{j}\cdot\mathbf{c}_{j}}\prod_{j=1}^{h}% \mathop{\theta\!\genfrac{[}{]}{0.0pt}{}{\mathbf{a}_{j}+\mathbf{c}_{j}}{\mathbf% {b}_{j}+\mathbf{d}_{j}}\/}\nolimits\!\left(\mathbf{z}_{j}\middle|\boldsymbol{{% \Omega}}\right),$

where $\mathbf{c}_{j}$ and $\mathbf{d}_{j}$ are arbitrary $h$-dimensional vectors. Many identities involving products of theta functions can be established using these formulas.

# ¶ Example

Let $h=4$ and

 21.6.5 $\mathbf{T}=\begin{bmatrix}1&1&1&1\\ 1&1&-1&-1\\ 1&-1&1&-1\\ 1&-1&-1&1\end{bmatrix}.$ Permalink: http://dlmf.nist.gov/21.6.E5 Encodings: TeX, pMML, png

Then

 21.6.6 $\mathop{\theta\/}\nolimits\!\left(\frac{\mathbf{x}+\mathbf{y}+\mathbf{u}+% \mathbf{v}}{2}\middle|\boldsymbol{{\Omega}}\right)\mathop{\theta\/}\nolimits\!% \left(\frac{\mathbf{x}+\mathbf{y}-\mathbf{u}-\mathbf{v}}{2}\middle|\boldsymbol% {{\Omega}}\right)\mathop{\theta\/}\nolimits\!\left(\frac{\mathbf{x}-\mathbf{y}% +\mathbf{u}-\mathbf{v}}{2}\middle|\boldsymbol{{\Omega}}\right)\mathop{\theta\/% }\nolimits\!\left(\frac{\mathbf{x}-\mathbf{y}-\mathbf{u}+\mathbf{v}}{2}\middle% |\boldsymbol{{\Omega}}\right)=\frac{1}{2^{g}}\sum_{\boldsymbol{{\alpha}}\in% \frac{1}{2}\Integer^{g}/\Integer^{g}}\,\sum_{\boldsymbol{{\beta}}\in\frac{1}{2% }\Integer^{g}/\Integer^{g}}e^{2\pi i\left(2\boldsymbol{{\alpha}}\cdot% \boldsymbol{{\Omega}}\cdot\boldsymbol{{\alpha}}+\boldsymbol{{\alpha}}\cdot[% \mathbf{x}+\mathbf{y}+\mathbf{u}+\mathbf{v}]\right)}\*\mathop{\theta\/}% \nolimits\!\left(\mathbf{x}+\boldsymbol{{\Omega}}\boldsymbol{{\alpha}}+% \boldsymbol{{\beta}}\middle|\boldsymbol{{\Omega}}\right)\mathop{\theta\/}% \nolimits\!\left(\mathbf{y}+\boldsymbol{{\Omega}}\boldsymbol{{\alpha}}+% \boldsymbol{{\beta}}\middle|\boldsymbol{{\Omega}}\right)\mathop{\theta\/}% \nolimits\!\left(\mathbf{u}+\boldsymbol{{\Omega}}\boldsymbol{{\alpha}}+% \boldsymbol{{\beta}}\middle|\boldsymbol{{\Omega}}\right)\mathop{\theta\/}% \nolimits\!\left(\mathbf{v}+\boldsymbol{{\Omega}}\boldsymbol{{\alpha}}+% \boldsymbol{{\beta}}\middle|\boldsymbol{{\Omega}}\right),$

and

 21.6.7 $\mathop{\theta\!\genfrac{[}{]}{0.0pt}{}{\tfrac{1}{2}[\mathbf{c}_{1}+\mathbf{c}% _{2}+\mathbf{c}_{3}+\mathbf{c}_{4}]}{\tfrac{1}{2}[\mathbf{d}_{1}+\mathbf{d}_{2% }+\mathbf{d}_{3}+\mathbf{d}_{4}]}\/}\nolimits\!\left(\frac{\mathbf{x}+\mathbf{% y}+\mathbf{u}+\mathbf{v}}{2}\middle|\boldsymbol{{\Omega}}\right)\mathop{\theta% \!\genfrac{[}{]}{0.0pt}{}{\tfrac{1}{2}[\mathbf{c}_{1}+\mathbf{c}_{2}-\mathbf{c% }_{3}-\mathbf{c}_{4}]}{\tfrac{1}{2}[\mathbf{d}_{1}+\mathbf{d}_{2}-\mathbf{d}_{% 3}-\mathbf{d}_{4}]}\/}\nolimits\!\left(\frac{\mathbf{x}+\mathbf{y}-\mathbf{u}-% \mathbf{v}}{2}\middle|\boldsymbol{{\Omega}}\right)\*\mathop{\theta\!\genfrac{[% }{]}{0.0pt}{}{\tfrac{1}{2}[\mathbf{c}_{1}-\mathbf{c}_{2}+\mathbf{c}_{3}-% \mathbf{c}_{4}]}{\tfrac{1}{2}[\mathbf{d}_{1}-\mathbf{d}_{2}+\mathbf{d}_{3}-% \mathbf{d}_{4}]}\/}\nolimits\!\left(\frac{\mathbf{x}-\mathbf{y}+\mathbf{u}-% \mathbf{v}}{2}\middle|\boldsymbol{{\Omega}}\right)\mathop{\theta\!\genfrac{[}{% ]}{0.0pt}{}{\tfrac{1}{2}[\mathbf{c}_{1}-\mathbf{c}_{2}-\mathbf{c}_{3}+\mathbf{% c}_{4}]}{\tfrac{1}{2}[\mathbf{d}_{1}-\mathbf{d}_{2}-\mathbf{d}_{3}+\mathbf{d}_% {4}]}\/}\nolimits\!\left(\frac{\mathbf{x}-\mathbf{y}-\mathbf{u}+\mathbf{v}}{2}% \middle|\boldsymbol{{\Omega}}\right)\\ =\frac{1}{2^{g}}\sum_{\boldsymbol{{\alpha}}\in\frac{1}{2}\Integer^{g}/\Integer% ^{g}}\,\sum_{\boldsymbol{{\beta}}\in\frac{1}{2}\Integer^{g}/\Integer^{g}}e^{-2% \pi i\boldsymbol{{\beta}}\cdot[\mathbf{c}_{1}+\mathbf{c}_{2}+\mathbf{c}_{3}+% \mathbf{c}_{4}]}\*\mathop{\theta\!\genfrac{[}{]}{0.0pt}{}{\mathbf{c}_{1}+% \boldsymbol{{\alpha}}}{\mathbf{d}_{1}+\boldsymbol{{\beta}}}\/}\nolimits\!\left% (\mathbf{x}\middle|\boldsymbol{{\Omega}}\right)\mathop{\theta\!\genfrac{[}{]}{% 0.0pt}{}{\mathbf{c}_{2}+\boldsymbol{{\alpha}}}{\mathbf{d}_{2}+\boldsymbol{{% \beta}}}\/}\nolimits\!\left(\mathbf{y}\middle|\boldsymbol{{\Omega}}\right)% \mathop{\theta\!\genfrac{[}{]}{0.0pt}{}{\mathbf{c}_{3}+\boldsymbol{{\alpha}}}{% \mathbf{d}_{3}+\boldsymbol{{\beta}}}\/}\nolimits\!\left(\mathbf{u}\middle|% \boldsymbol{{\Omega}}\right)\mathop{\theta\!\genfrac{[}{]}{0.0pt}{}{\mathbf{c}% _{4}+\boldsymbol{{\alpha}}}{\mathbf{d}_{4}+\boldsymbol{{\beta}}}\/}\nolimits\!% \left(\mathbf{v}\middle|\boldsymbol{{\Omega}}\right).$

# §21.6(ii) Addition Formulas

Let $\boldsymbol{{\alpha}}$, $\boldsymbol{{\beta}}$, $\boldsymbol{{\gamma}}$, $\boldsymbol{{\delta}}\in\Real^{g}$. Then

 21.6.8 $\mathop{\theta\!\genfrac{[}{]}{0.0pt}{}{\boldsymbol{{\alpha}}}{\boldsymbol{{% \gamma}}}\/}\nolimits\!\left(\mathbf{z}_{1}\middle|\boldsymbol{{\Omega}}\right% )\mathop{\theta\!\genfrac{[}{]}{0.0pt}{}{\boldsymbol{{\beta}}}{\boldsymbol{{% \delta}}}\/}\nolimits\!\left(\mathbf{z}_{2}\middle|\boldsymbol{{\Omega}}\right% )=\sum_{\boldsymbol{{\nu}}\in\Integer^{g}/\left(2\Integer^{g}\right)}\mathop{% \theta\!\genfrac{[}{]}{0.0pt}{}{\frac{1}{2}[\boldsymbol{{\alpha}}+\boldsymbol{% {\beta}}+\boldsymbol{{\nu}}]}{\boldsymbol{{\gamma}}+\boldsymbol{{\delta}}}\/}% \nolimits\!\left(\mathbf{z}_{1}+\mathbf{z}_{2}\middle|2\boldsymbol{{\Omega}}% \right)\*\mathop{\theta\!\genfrac{[}{]}{0.0pt}{}{\frac{1}{2}[\boldsymbol{{% \alpha}}-\boldsymbol{{\beta}}+\boldsymbol{{\nu}}]}{\boldsymbol{{\gamma}}-% \boldsymbol{{\delta}}}\/}\nolimits\!\left(\mathbf{z}_{1}-\mathbf{z}_{2}\middle% |2\boldsymbol{{\Omega}}\right).$

Thus $\boldsymbol{{\nu}}$ is a $g$-dimensional vector whose entries are either $0$ or $1$. For this result and a generalization see Koizumi (1976) and Belokolos et al. (1994, pp. 38–41). For addition formulas for classical theta functions see §20.7(ii).