Global description of action-angle duality for a Poisson-Lie deformation of the trigonometric \(\mathrm {BC}_n\) Sutherland system (Q1735504): Difference between revisions

The authors are concerned with integrable many-body systems of Ruijsenaars-Schneider-van Diejen type that display a curious action-angle duality. Basically, there are pairs of $2n$-dimensional symplectic manifolds $(M,\omega)$, $(\hat{M},\hat{\omega})$ with Darboux coordinates $\lambda_i,\theta_j$, $\hat{\lambda}_i,\hat{\theta}_j$, and Hamiltonians $H$, $\hat{H}$, that can be interpreted as many-body Hamiltonians describing the dynamics of $n$ interacting particles on a line, with positions $\lambda_i$, $\hat{\lambda}_i$, respectively. These two systems are said to be in \textit{action-angle duality} if there exists a global symplectomorphism $\mathcal{R}:(M,\omega)\to ({\hat M},\hat{\omega})$ such that under this correspondence the position coordinates $\hat{\lambda}_i$ are action-angle variables for $M$ and the position coordinates $\lambda_i$ are action-angle variables for $\hat{M}$. This means that $\mathcal{R}$ and $\mathcal{R}^{-1}$ can be interpreted as global action-angle maps for the Liouville integrable systems $(M, \omega,H)$ and $(\hat{M},\hat{\omega},\hat{H})$ and the solution of one system automatically leads to solution of the other system. In all known examples of this phenomena the Hamiltonian flows can be written down explicitly and globally. \par For some, but not all, of the examples it has been recognized that these dual pairs of integrable many-body systems can be derived by Hamiltonian reduction from a higher-dimensional `master phase space' $\mathcal{M}$ that admits a symmetry group $G$, and two distinguished independent abelian Poisson algebras $\mathcal{H}^1$ and $\mathcal{H}^2$ formed by $G$-invariant, smooth functions on $\mathcal{M}$. Then one can apply Hamiltonian reduction to $\mathcal{M}$ to obtain a reduced phase space $\mathcal{M}_{\mathrm{red}}$ equipped with two Abelian Poisson algebras $\mathcal{H}^1_{\mathrm{red}}$, $\mathcal{H}^2_{\mathrm{red}}$ that descend from $\mathcal{H}^1,\mathcal{H}^2$ respectively. Then one needs to construct two distinct models $M$ and $\hat{M}$ of $\mathcal{M}_{\mathrm{red}}$ yielding $(M,H,\lambda_i)$,$(\hat{M},\hat{H},\hat{\lambda}_i)$ in such a way that the reduction of $\mathcal{H}^1$ is represented by $\mathcal{H}$, the reduction of $\mathcal{H}^2$ is represented by $\hat{\mathcal{H}}^2$, and there are many-body Hamiltonians $H\in \mathcal{H},\hat{H}\in \hat{\mathcal{H}}^2$ with associated position variables $\{\lambda_i\},\{\hat{\lambda}_i\}$, respectively. If this is achieved, then there is a natural map $R: M \to \hat{M}$ that corresponds to the identity map on $\mathcal{M}_{\mathrm{red}}$ and relates the Abelian Poisson algebras on $M$ to those on $\hat{M}$ in the desired way. \par The authors refer to several papers in which this construction is successful. However, the standard application of Hamiltonian reduction does not provide globally valid descriptions of the reduced system. The authors study a single example and show that global validity can be achieved: reduction of the Heisenberg double of the standard Poisson Lie group $\mathrm{SU}(2n)$. Locally, the Hamiltonians $\hat H$, $H$ can be written as \[ H(\hat{\lambda}, \hat{\theta}) = U(\hat{\lambda}) - \sum_{j=1}^n \cos(\hat{\theta}_j)U_1(\hat{\lambda}_j)^{1/2}\prod_{k\ne j}^n\left[ 1-\frac{\sinh^2\mu}{\sinh^2(\hat{\lambda}_j-\hat{\lambda}_k)}\right]^{1/2}, \] with \[ U(\hat{\lambda}) =\frac{ e^{-2u} + e^{2v}}{2}\sum_{j=1}^n\exp(-2\hat{\lambda}_j), \] \[ U_1(\hat{\lambda}_j)=[1 - (1 + e^{2(v-u)}) \exp(-2\hat{\lambda}_j) + e^{2(v-u)} \exp(-4\hat{\lambda}_j). \] \[\begin{multlined} H(\lambda, \theta) = V (\lambda) + e^{v-u}\sum_{j=1}^n \frac{\cos(\theta_j)}{\cosh^2(\lambda_j)}[ 1-\frac{\sinh^2(v)}{\sinh^2(\lambda_j)}]^{1/2}[ 1-\frac{\sinh^2(u)}{\sinh^2(\lambda_j)}]^{1/2}\\ \times \prod_{k\ne j}^n [ 1 - \frac{\sinh^2 \mu}{\sinh^2(\lambda_j-\lambda_k)}]^{1/2} [ 1 - \frac{\sinh^2 \mu}{\sinh^2(\lambda_j+\lambda_k)}]^{1/2}, \end{multlined}\] \[\begin{multlined} V (\lambda) = e^{v-u}\left(\frac{\sinh(v)\sinh(u)}{\sinh^2 \mu}\prod_{j=1}^n[1-\frac{\sinh^2\mu}{\sinh^2\lambda_j}]\right.\\ -\left. \frac{\cos(v) \cosh(u)}{\sinh^2 \mu}\prod_{j=1}^n[1+\frac{\sinh^2 \mu}{\cosh^2\lambda_j}]+n e^{u-v} +\frac{\cosh(v - u)}{\sinh^2 \mu}\right). \end{multlined}\] Here $\mu,u,v$ are parameters. \par The master phase space is $\mathcal{M} = {\mathrm{SL}}(2n,\mathbb{C})$ viewed as a real Lie group. The construction uses its subgroups $K = {\mathrm{SU}}(2n)$, and $B= {\mathrm{SB}}(2n)$ (formed by upper triangular complex matrices with positive entries along the diagonal). Every $g \in \mathcal{M}$ admits the alternative Iwasawa decompositions $g = kb = b_Lk_R$, $k,k_R \in K$, $b, b_L \in B$. The Alekseev-Malkin symplectic form on $\mathcal{M}$ is \[ \omega_{\mathcal{M}}= \frac12 \ \text{Im tr}(db_Lb_L^{-1}\wedge dkk^{-1})+\frac12\ \text{Im tr}(b^{-1}db\wedge k_R^{-1}dk_R). \] The Poisson bracket on $\mathcal{M}$ determined by the symplectic form is known as the Heisenberg double of the standard Poisson Lie group ${\mathrm{SU}}(2n)$. Note that $\mathrm{sl}(2n,C) = \mathcal{K} + \mathcal{B}$ where $\mathcal K$ and $\mathcal{B}$ are the Lie algebras of $K$ and $B$. The linear mapping $R$ is defined as $R =\frac12 (\pi_\mathcal{K}- \pi_\mathcal{B})$ where $\pi_{\mathcal{K}},\pi_{\mathcal{B}}$ are the projectors on $\mathcal{K}$, $\mathcal{B}$. The space $\hat{\mathcal{H}}^2$ is spanned by the functions $\hat{\mathcal{H}}_j(g)=\frac12\ \text{tr}((bb^\dag)^j)$ for $j=1,\dots ,2n-1$. The space $\mathcal{H}^1$ is generated by the functions $\mathcal{H}_j(g)=\frac12\ \text{tr}((k^\dag IkI)^j)$ for $j=1,\dots,n$, where $I = \text{diag}(\mathbf{1}_n,-\mathbf{1}_n)$ and $\mathbf{I}_n$ is the $n\times n$ unit matrix. We introduce the subgroup $K_+ =\{ k \in K \mid k^\dag I k=I\}$. \par The symmetry group for the reduction is $K_+\times K_+$ acting by \[ K_+ \times K_+\times \mathcal{M}\to \mathcal{M}, \quad (\eta_L, \eta_R, g) \to \eta_L g\eta^{-1}_R. \] The momentum map generating this action sends $g$ to the pair of matrices given by block-diagonal parts of $b_L$ and $b$, and is accomplished in two steps, first by requiring $b$ and $b_L$ be of a certain explicit form and demanding that $\mathcal{M}_0$ be the subspace consisting of all $g\in \mathcal{M}$ such that $b$, $b_L$ take that form. This space is stable under the action of the gauge group $K_+(\sigma)\times K_+$ where $K_+(\sigma)$ is a subgroup of $K_+$ defined by an $n\times n$ upper triangular matrix $\sigma$ taken from the restriction of $b_L$ Then $\mathcal{M}_{\mathrm{red}}= \mathcal{M}_0/K_+(\sigma)\times K_+=K_+(\sigma)\backslash \mathcal{M}_0/K_+$. The model $\hat M$ for the reduction is adapted from [\textit{L. Fehér} and \textit{T. F. Görbe}, Nucl. Phys., B 901, 85--114 (2015; Zbl 1332.81082)]. It is obtained by explicitly constructing a global cross section of the gauge orbits in $\mathcal{M}_0$. There it is shown that $\mathcal{H}_j, \hat{\mathcal{H}}_j$ reduce to $\mathcal{H}_j^{\mathrm{red}},\hat{\mathcal{H}}^{\mathrm{red}}_j$ for $j=1,\dots n$, and $\omega_{\mathcal{M}}$ reduces to a symplectic form $\omega_{\mathrm{red}}$. In the model, $\hat{M}=\mathbb{C}^n$ in the variables $\mathbb{Z}_j$ with symplectic form $d\omega_{\mathrm{can}}=i\sum_{j=1}^n d\mathbb{Z}_j\wedge d\mathbb{Z}_j^*$. The ${\hat \lambda}_j$ generate the action of the torus $\mathbb{T}^n$ on $\hat M: (\mathbb{Z}_1, \dots ,\mathbb{Z}_n) \to (\mathbb{Z}_1 e^{it_1},\dots,\mathbb{Z}_ne^{it_n})$. The action has a single fixed point $\mathbb{Z}_1=\dots=\mathbb{Z}_n=0$ and the reduced Hamiltonian ${\hat H} ={\hat H}_1$ has the desired many-body form. \par The construction of the global model $M$ is much more complicated and constitutes the main new contribution of the authors. In [J. Phys. A, Math. Theor. 50, No. 31, Article ID 314004, 20 p. (2017; Zbl 1371.81158)], they have partial results for this description of $M$, but they must introduce new ideas and very lengthy technical computations for the global construction in this paper. The authors construct a space $\mathcal{N}$ acted upon by the $n$-torus $\mathbb{T}_n$ such that $\mathcal{M}_{\mathrm{red}}=\mathcal{N}/\mathbb{T}_n$. $M$ is constructed as a global cross section for this action. In the final construction the action-angle phase spaces $(M,\omega)$ and $(\hat{M}, \hat{\omega})$ are the same, both are equal to $(\mathbb{C}_n, \omega_{\mathrm{can}})$, although the coordinate identifications are, of course, different. The duality map $R$ is just the identity map on $M_{\mathrm{red}}$ written in terms of two distinct models. Another conclusion of the global models is that the particle positions evaluated along any fixed phase space trajectories of the Hamiltonians stay in a compact set and that both systems have a unique equilibrium point, \par In an appendix the authors show how the Hamiltonian $H$ is recovered as a scaling limit of van Diejen's 5-parametric integrable Hamiltonians.