--- title: Symplectic–Haantjes Structures in Integrability url: https://www.emergentmind.com/topics/symplectic-haantjes-structure type: topic --- # Symplectic–Haantjes Structures in Integrability Searching arXiv for papers on symplectic–Haantjes structures and related integrability results. arxiv_search(query="symplectic Haantjes structure integrability Darboux-Haantjes coordinates", max_results=10) arxiv_search(query="symplectic Haantjes structure", max_results=10) A symplectic–Haantjes structure is a geometric structure on a symplectic manifold that couples the symplectic form with an algebra of \((1,1)\)-tensor fields whose Haantjes torsion vanishes. In the formulation introduced by Tempesta and Tondo, the corresponding symplectic–Haantjes, or \(\omega\mathscr H\), manifolds provide a tensorial setting in which Liouville–Arnold integrability, separation of variables, and the construction of new integrable models can be stated in a unified way [1405.5118]. Subsequent work extended the framework to multiseparable and superintegrable systems, partial separability, magnetic Hamiltonians, and Jacobi-type generalizations [2012.09819] [2305.06844] [2507.11715]. ## 1. Definition and algebraic ingredients Let \(M\) be a smooth manifold and \(K:TM\to TM\) a \((1,1)\)-tensor field. The Nijenhuis torsion of \(K\) is \[ N_K(X,Y):=K^2[X,Y]+[KX,KY]-K\bigl([X,KY]+[KX,Y]\bigr), \] and the Haantjes torsion is \[ H_K(X,Y):=K^2N_K(X,Y)+N_K(KX,KY)-K\bigl(N_K(X,KY)+N_K(KX,Y)\bigr). \] A tensor \(K\) is called a Nijenhuis operator if \(N_K\equiv 0\), and a Haantjes operator if \(H_K\equiv 0\) [1405.5118]. A symplectic–Haantjes manifold of class \(m\) is a triple \((M,\omega,\mathscr H)\) in which \((M,\omega)\) is a \(2n\)-dimensional symplectic manifold and \(\mathscr H\) is a rank-\(m\) Haantjes algebra, namely an \(m\)-dimensional \(C^\infty(M)\)-module of Haantjes operators closed under composition and linear combinations [1405.5118]. In the later formulation used in the literature, \(\mathscr H\subset \Gamma(\operatorname{End}(TM))\) is an associative algebra of \((1,1)\)-tensor fields such that every \(H\in\mathscr H\) has vanishing Haantjes torsion, \(\mathscr H\) is closed under \(C^\infty(M)\)-linear combinations and composition, and the operators commute pairwise in the Abelian case [2012.09819]. If the identity \(I\) belongs to \(\mathscr H\), the manifold is said to have identity; if, in addition, all operators commute pairwise, it is an Abelian \(\omega\mathscr H\)-manifold [1405.5118]. A useful structural fact is that any polynomial in a Haantjes operator is again Haantjes, a point used repeatedly in the construction of cyclic Haantjes algebras [2305.06844]. ## 2. Symplectic compatibility and spectral geometry The defining compatibility condition with the symplectic form is \[ \omega(X,KY)=\omega(KX,Y), \] for every \(K\in\mathscr H\) and all vector fields \(X,Y\). Equivalently, if \(\Omega:TM\to T^*M\) denotes the bundle isomorphism \(X\mapsto \omega(X,\cdot)\), then \[ \Omega\,K=K^T\,\Omega. \] This means that each Haantjes operator in the algebra is \(\omega\)-self-adjoint [1405.5118] [2507.11715]. The spectral decomposition of a Haantjes operator is central to the geometry. For a general operator \(K\), one has pointwise generalized eigendistributions \[ \mathcal D_i=\ker\bigl(K-\lambda_i I\bigr)^{\rho_i}. \] Because \(K\) is \(\omega\)-self-adjoint, these generalized eigendistributions are symplectically orthogonal complements of one another, and the vanishing of the Haantjes torsion guarantees that each such distribution is integrable in Frobenius’ sense [2507.11715]. In the formulation of Reyes, Tempesta, and Tondo, every eigen-distribution has even rank, semisimplicity means that all Jordan blocks are one-dimensional, and maximal rank means that generically there are \(n\) distinct eigenvalues in the algebra [2305.06844]. This spectral picture explains the geometric role of \(\omega\mathscr H\)-structures. In the semisimple, maximal-rank case, the tangent bundle splits into mutually complementary, even-dimensional, integrable subbundles, and the associated foliations support full separability. When semisimplicity fails, or when the algebra has non-maximal rank, the same mechanism persists in block form and leads to partial separability rather than full additive separation [2305.06844]. ## 3. Haantjes chains and Liouville integrability The central theorem of the theory is the Liouville–Haantjes characterization of complete integrability. On a \(2n\)-dimensional Abelian \(\omega\mathscr H\)-manifold of class \(n\), suppose there exists a Haantjes chain of exact \(1\)-forms \[ dH_1=dH,\qquad dH_2=K_2^T\,dH,\ \dots,\ dH_n=K_n^T\,dH. \] Then the functions \(H_1,\dots,H_n\) are pairwise in involution and define a Lagrangian foliation; hence \(H\) is Liouville-integrable [1405.5118]. The converse statement is equally important. If \((M,\omega,H)\) is a non-degenerate integrable Hamiltonian system with \(n\) independent action–angle variables \((J_k,\phi_k)\), then one may define \(n\) commuting Haantjes operators \[ K_a=\sum_{k=1}^n \frac{\partial H_a/\partial J_k}{\partial H/\partial J_k} \left(\frac{\partial}{\partial J_k}\otimes dJ_k\right), \qquad a=1,\dots,n, \] which satisfy \[ K_a^T\,dH=dH_a. \] These operators have vanishing Haantjes torsion, commute, and generate an Abelian Haantjes algebra of rank \(n\) [1405.5118]. Later formulations sharpened the geometric content of these chains. In an \(\omega H\)-manifold, a function \(H\) generates a Haantjes chain of length \(m\) if \(dH_\alpha:=K_\alpha^T dH\) are closed. This is equivalent to the Frobenius integrability of the codistribution \[ \mathcal D_H^\circ=\mathrm{Span}\{K_1^T dH,\dots,K_m^T dH\}, \] and if the chain has length \(n\) in dimension \(2n\), then the resulting functions are pairwise in involution with respect to the Poisson bracket defined by \(\omega\) [2507.11715]. In the language of Kosmann-Schwarzbach, these are Lenard–Haantjes chains: the classical role of a single recursion operator and its powers is replaced by a family of commuting Haantjes operators [1712.08908]. A frequent source of confusion is the relation to Nijenhuis theory. A symplectic–Nijenhuis manifold is a special case of an \(\omega H\)-manifold in which every operator also has vanishing Nijenhuis torsion. The implication \(\tau_K=0\Rightarrow \mathcal H_K=0\) holds, but the converse does not hold in general [2507.11715]. This is precisely the point at which the Haantjes framework enlarges the admissible class of integrable systems. ## 4. Darboux–Haantjes coordinates and separation of variables A structural theorem guarantees the existence of local coordinates adapted simultaneously to the symplectic form and to the Haantjes algebra. On a semisimple Abelian \(\omega\mathscr H\)-manifold of class \(n\), every point has a neighborhood carrying Darboux–Haantjes coordinates \[ (q_1,\dots,q_n,p_1,\dots,p_n) \] such that \[ \omega=\sum_{i=1}^n dq_i\wedge dp_i, \] and every \(H\in\mathscr H\) takes diagonal form \[ H=\mathrm{diag}\bigl(\underbrace{\lambda_1,\dots,\lambda_1}_{2},\dots,\underbrace{\lambda_n,\dots,\lambda_n}_{2}\bigr), \] where each eigenvalue \(\lambda_i=\lambda_i(q_i,p_i)\) depends on exactly one pair \((q_i,p_i)\) [2012.09819]. In the more general Abelian case, Darboux–Haantjes coordinates still exist locally, but each operator is only block-diagonal with respect to the splitting into eigendistributions; in the semisimple case this reduces to full diagonalization [1405.5118] [2305.06844]. For a diagonal operator \[ K=\mathrm{diag}\bigl(\lambda^1,\dots,\lambda^n,\lambda^1,\dots,\lambda^n\bigr), \] the vanishing of the Haantjes torsion forces the eigenvalue functions \(\lambda^i\) to depend only on the pair \((q_i,p_i)\), with \[ \partial_{q_j}\lambda^i=\partial_{p_j}\lambda^i=0,\qquad j\neq i \] [1405.5118]. These coordinates are not merely normal forms; they are separation variables. If \(H\) generates a Haantjes chain of length \(n\), then the Darboux–Haantjes coordinates associated with \(\mathscr H\) are separation variables for each Hamiltonian in the chain [2012.09819]. Conversely, if a Hamiltonian is separable in Darboux coordinates, then one can construct diagonal Haantjes operators \[ K_a=\sum_{i=1}^n \bigl(\partial_{p_i}H_a\bigr)\, \bigl(\partial_{q_i}\otimes dp_i\bigr) +\bigl(-\partial_{q_i}H_a\bigr)\, \bigl(\partial_{p_i}\otimes dq_i\bigr), \] which span an Abelian semisimple Haantjes algebra compatible with \(\omega\), and the equations \(K_a^T dH=dH_a\) are solvable precisely because of the separability conditions [2012.09819]. The same logic extends to multiseparability and partial separation. An integrable Hamiltonian system admits as many inequivalent semisimple Abelian \(\omega\mathscr H\)-structures of class \(n\) as it has independent orthogonal separation schemes [2012.09819]. For partial separability, Reyes, Tempesta, and Tondo introduced generalized Stäckel matrices \[ S(q)=\bigl(S_{aj}(q^a)\bigr)_{a,j=1}^m, \] with the \(a\)-th row depending only on a block \(q^a\). Given separation data \(f_a(q^a,p^a)\), the Hamiltonians are defined by \[ \begin{pmatrix} H_1\\ \vdots\\ H_m \end{pmatrix} = S(q)^{-1} \begin{pmatrix} f_1(q^1,p^1)\\ \vdots\\ f_m(q^m,p^m) \end{pmatrix}, \] and the Hamilton–Jacobi equation splits into \(m\) first-order PDEs. When \(m=n\), one recovers full additive separation; when \(m