---
title: Symmetric Poisson Structures
url: https://www.emergentmind.com/topics/symmetric-poisson-structures
type: topic
---

# Symmetric Poisson Structures

Symmetric Poisson structures form a family of constructions that transpose characteristic features of Poisson geometry into settings where symmetry, rather than skew-symmetry, is primary. In current usage, the expression covers several non-identical notions: a symmetric bivector field coupled to a torsion-free connection and satisfying a connection-dependent integrability law; contravariant pseudo-Hessian and Koszul-Vinberg structures, where symmetric \(2\)-tensors play the role of Poisson tensors on affine or left-symmetric algebroid backgrounds; shifted Poisson structures whose low-shift components are invariant symmetric tensors; and Poisson structures obtained from symmetry reduction or Poisson symmetric spaces [2508.15890] [2001.03776] [2004.01774] [2412.12804].

## 1. Scope of the notion

The phrase “symmetric Poisson structure” does not designate a single universally fixed definition. In the differential-geometric literature, it can mean a pair \((\vartheta,\nabla)\) with \(\vartheta\in \Gamma(\Sym^2 TM)\) and \(\nabla\) torsion-free, subject to a symmetric Schouten-type condition. In affine and algebroid settings, it can mean a symmetric bivector \(h\in \Gamma(S^2TM)\) or \(H\in \Gamma(S^2A)\) satisfying a contravariant Codazzi identity or a Koszul-Vinberg self-commutation relation. In derived and representation-theoretic settings, the symmetric aspect appears through invariant symmetric tensors, symmetric Leibniz products, or Poisson symmetric-space structures.

| Setting | Symmetric datum | Integrability or interpretation |
|---|---|---|
| Manifold with torsion-free connection | \(\vartheta \in \Gamma(\Sym^2 TM)\) | \([\vartheta,\vartheta]_s=0\) |
| Affine manifold \((M,\nabla)\) | \(h\in \Gamma(S^2TM)\) | contravariant Codazzi equation |
| Left-symmetric algebroid \(A\) | \(H\in \Gamma(S^2A)\) | \([H,H]_A=0\) |
| Jacobi-left-symmetric algebroid \((A,\phi_0)\) | \(h\in \Gamma(S^2A)\) | \([h,h]_{2\phi_0}=0\) |
| \(2\)-shifted Poisson on \(CE(g)\) | \(\pi^{(2,0)}\in \operatorname{Sym}^2(g)\) | adjoint invariance |

These usages are linked by an explicit analogy repeatedly stated in the literature: ordinary Poisson geometry uses skew-symmetric tensors and Lie-type structures, whereas the symmetric theories replace them by symmetric tensors and left-symmetric, Hessian, or connection-dependent structures. At the same time, some papers use “symmetric” in the sense of Poisson symmetric spaces or symmetric Leibniz geometry rather than symmetric bivector fields, so the term is broader than a single tensorial definition [2412.20103] [1404.0988] [2003.12608].

## 2. Reduction, Poisson-Lie symmetry, and symmetric spaces

A classical source of Poisson structures with symmetry is the reduction of rigid-body phase space. For the symmetric top, the unreduced phase space is
\[
T^*SE(3)\simeq T^*(\mathbb{R}^3)\times SO(3)\times so(3)^*,
\]
with canonical nonzero brackets
\[
\{x_i,p_j\}=\delta_{ij},\qquad \{\pi_i,\pi_j\}=\varepsilon_{ijl}\pi_l,\qquad \{\pi_i,R_{jk}\}=\varepsilon_{ijl}R_{lk}.
\]
The canonical Poisson structure is invariant under the right action of \(SO(3)\), but the Hamiltonian of a symmetric top is invariant only under the subgroup
\[
S^1=\{Z\in SO(3): Z_{i3}=\delta_{i3}\},
\]
corresponding to rotations about the body symmetry axis. The invariant variable is the third column \(\nu_i=R_{i3}\), and the quotient
\[
T^*SO(3)/S^1 \simeq W_1 \simeq S^2\times so(3)^*
\]
inherits the reduced Lie-Poisson brackets
\[
\{\nu_i,\nu_k\}=0,\qquad \{\pi_i,\nu_j\}=\varepsilon_{ijl}\nu_l,\qquad \{\pi_i,\pi_j\}=\varepsilon_{ijl}\pi_l.
\]
The Casimirs are
\[
C_1(\nu,\pi)=\nu^2,\qquad C_2(\nu,\pi)=\nu\cdot\pi,
\]
and for \(\nu^2=1\) the relevant coadjoint orbit is diffeomorphic to \(T^*S^2\), with KKS form
\[
\Omega^{\mathcal O_{(\nu_0,\pi_0)}}_{KKS}
=
\Omega^{T^*S^2}_{can}-\rho^*\mathcal B.
\]
After adjoining the untouched translational variables, the reduced space becomes
\[
P_1=T^*SE(3)/S^1\simeq T^*\mathbb R^3\times W_1,
\]
with reduced Hamiltonian
\[
h((x,\nu),(p,\pi))=\frac{1}{2M}p^2+\frac{1}{2I_1}\pi^2+V(x,\nu).
\]
Here the term
\[
\left(\frac{1}{2I_3}-\frac{1}{2I_1}\right)\langle \nu,\pi\rangle^2
\]
is a Casimir contribution and can be dropped from the Hamiltonian dynamics [1403.2869].

A second, more explicitly group-theoretic use of symmetry appears in the Poisson geometry of bilinear forms. The space \(\mathcal A\) of bilinear forms on \(\mathbb C^N\), identified with \(\operatorname{Mat}_N(\mathbb C)\), carries the quadratic bracket
\[
\{A\overset{\otimes}{,}A\}=r_{12}(A\otimes A)-(A\otimes A)r_{12}+Ar_{12}A-Ar_{21}A.
\]
For the congruence action \(A\mapsto BAB^T\), the mixed bracket on \((B,A)\) is forced into exactly three cases:
\[
Q_{12}=0,\qquad Q_{12}=-r_{12},\qquad Q_{12}=r_{12}.
\]
Passing from pairs \((B,A)\) to triples \((B,C,A)\) and the action \(A\mapsto BACT\), one obtains an involutive anti-Poisson automorphism
\[
\mathcal O(B,C)=(C^{-T},B^{-T}),
\]
whose fixed-point set
\[
H=\{(H_1,H_2)\in GL_N\times GL_N\mid H_1=H_2^T\}
\]
gives a quotient \(GL_N\times GL_N/H\) immersed into \(\mathcal A\) by
\[
i(B,C)=BCT.
\]
The resulting identification makes \(\mathcal A\) into a Poisson symmetric space in the sense of Fernandes [1404.0988].

A related Poisson-Lie-symmetric picture appears on
\[
S(n,d)=\operatorname{Mat}_{n\times d}(\mathbb C)\times \operatorname{Mat}_{d\times n}(\mathbb C),
\]
where the holomorphic Poisson bracket contains quadratic terms and the constant coupling \(C_{n\times d}\). The natural left actions of \(GL(n,\mathbb C)\) and \(GL(d,\mathbb C)\) are Poisson for the standard multiplicative Poisson-Lie bracket
\[
\{g_1,g_2\}_G = K[g_1g_2,r_\ell].
\]
For the \(GL(n,\mathbb C)\)-action, the local moment map is built from
\[
T(A,B)=I_n+AB.
\]
The special case \(d=1\) provides a building block, and Theorem 1.1 constructs a local holomorphic Poisson diffeomorphism
\[
m:(S(n,1),\{\cdot,\cdot\}_K)^d \to (S(n,d),\{\cdot,\cdot\}_K),
\]
so the Poisson structure decouples locally into \(d\) commuting \(d=1\) blocks. The tensor is non-degenerate on a dense open subset, but not globally symplectic [2112.00381].

## 3. Connection-dependent symmetric bivectors

The most explicit modern definition of a symmetric Poisson structure replaces the skew bivector \(\pi\) of ordinary Poisson geometry by a symmetric bivector
\[
\vartheta \in \Gamma(\Sym^2 TM),
\]
which defines a symmetric bracket on functions,
\[
\{f,g\}:=\vartheta(df,dg),
\]
and a gradient map
\[
\grad := \vartheta\circ d,\qquad X_f:=\grad f.
\]
A basic obstruction is that the naive analogues of the usual Poisson integrability conditions fail: the ordinary skew Schouten bracket gives no content for a symmetric bivector, and both plausible Jacobiators force the symmetric bracket to vanish. To recover a nontrivial theory one fixes a torsion-free connection \(\nabla\), introduces the symmetric Lie bracket
\[
[X,Y]_s:=\nabla_XY+\nabla_YX,
\]
extends it to a symmetric Schouten bracket \([\, ,\, ]_s\), and defines a symmetric Poisson structure to be a pair
\[
(\vartheta,\nabla)
\]
satisfying
\[
[\vartheta,\vartheta]_s=0.
\]
The integrability condition is concretely
\[
\frac12[\vartheta,\vartheta]_s(\alpha,\beta,\eta)
=
(\nabla_{\vartheta(\alpha)}\vartheta)(\beta,\eta)
+
(\nabla_{\vartheta(\beta)}\vartheta)(\eta,\alpha)
+
(\nabla_{\vartheta(\eta)}\vartheta)(\alpha,\beta),
\]
hence equivalent to the cyclic identity
\[
(\nabla_{\vartheta(\alpha)}\vartheta)(\beta,\eta)+\text{cyclic}(\alpha,\beta,\eta)=0.
\]
In terms of the bracket on functions,
\[
\Jac(f,g,h) = dh\big([X_f,X_g]_s\big)+\text{cyclic}(f,g,h),
\]
where
\[
\Jac(f,g,h):=\{f,\{g,h\}\}+\{g,\{h,f\}\}+\{h,\{f,g\}\}.
\]
Accordingly, the theory is not governed by the Jacobi identity of the symmetric bracket alone; the Jacobiator is controlled by the symmetric Lie bracket of gradients [2508.15890].

The same paper distinguishes between symmetric and strong symmetric Poisson structures. A pair \((\vartheta,\nabla)\) is strong if
\[
\grad\{f,g\}=[\grad f,\grad g]_s,
\]
equivalently
\[
X_{\{f,g\}}=[X_f,X_g]_s.
\]
This is further equivalent to
\[
\nabla_{X_f}\vartheta=0\qquad \text{for all } f\in C^\infty(M),
\]
and therefore to
\[
\nabla_{\vartheta(\alpha)}\vartheta=0\qquad \forall\,\alpha\in\Omega^1(M).
\]
The resulting hierarchy is
\[
\{\text{parallel symmetric bivectors}\}\subseteq
\{\text{strong symmetric Poisson structures}\}\subseteq
\{\text{symmetric Poisson structures}\},
\]
and the inclusions are strict. This distinction is essential because equivalent formulations in ordinary Poisson geometry become inequivalent once skew-symmetry is replaced by symmetry.

## 4. Foliations, cotangent dynamics, and linear models

Given \(\vartheta\), the characteristic distribution and characteristic module are
\[
\im\vartheta:=\{\vartheta(\zeta)\mid \zeta\in T^*M\}\subseteq TM,\qquad
\mathcal{F}_\vartheta:=\{\vartheta(\alpha)\mid \alpha\in\Omega^1(M)\}\subseteq \mathfrak X(M).
\]
The condition \([\vartheta,\vartheta]_s=0\) implies
\[
[\mathcal{F}_\vartheta,\mathcal{F}_\vartheta]_s\subseteq \mathcal{F}_\vartheta.
\]
If \(\vartheta\) is regular, then \(\Gamma(\im\vartheta)=\mathcal F_\vartheta\), and Lewis’ theorem gives geodesic invariance of the characteristic distribution. More generally, the characteristic distribution of any symmetric Poisson structure is locally geodesically invariant. When the structure is involutive, the characteristic partition is totally geodesic; on each leaf \(N\), the restricted bivector \(\vartheta_N\) is non-degenerate and defines a leaf metric
\[
g_N:=\vartheta_N^{-1}.
\]
The ambient connection induces a torsion-free leaf connection \(\nabla^N\), and \((g_N^{-1},\nabla^N)\) is again a non-degenerate symmetric Poisson structure. In the strong case, \(\nabla^N\) is the Levi-Civita connection of \(g_N\), so strong symmetric Poisson structures correspond to totally geodesic foliations together with the induced metric and connection data [2508.15890].

The same framework equips each fiber of \(\im\vartheta\) with the characteristic metric
\[
g_{\vartheta_m}\big(\vartheta(\alpha),\vartheta(\beta)\big):=\vartheta(\alpha,\beta).
\]
For any \(\vartheta\)-admissible geodesic \(\gamma\), the squared norm
\[
\mathrm{Sq}_\vartheta(\dot\gamma):=g_\vartheta(\dot\gamma,\dot\gamma)
\]
is smooth, and if \((\vartheta,\nabla)\) is symmetric Poisson then \(\mathrm{Sq}_\vartheta(\dot\gamma)\) is constant.

A major dynamical construction uses the Patterson-Walker metric on \(T^*M\),
\[
g_\nabla = dp_j\odot dx^j - p_k\,\Gamma^k_{ij}\,dx^i\odot dx^j,
\]
whose inverse defines a symmetric bracket
\[
\{F,G\}_\nabla
=
\frac{\partial F}{\partial x^i}\frac{\partial G}{\partial p_i}
+
\frac{\partial F}{\partial p_i}\frac{\partial G}{\partial x^i}
+
2p_k\Gamma^k_{ij}\frac{\partial F}{\partial p_i}\frac{\partial G}{\partial p_j}.
\]
Vertical lift intertwines the symmetric Schouten bracket with the Patterson-Walker bracket:
\[
[\mathcal X,\mathcal Y]_s^v=\{\mathcal X^v,\mathcal Y^v\}_\nabla.
\]
The induced dynamics
\[
\grad_\nabla H:=g_\nabla^{-1}(dH)=\{H,\cdot\}_\nabla
\]
has integral-curve equations
\[
\dot\gamma_a^i=\frac{\partial H}{\partial p_i}\circ a,\qquad
\dot a_j=\frac{\partial H}{\partial x^j}\circ a + 2a_k\Gamma^k_{ij}\frac{\partial H}{\partial p_i}\circ a.
\]
This recovers the parallel transport equation, gradient extensions of dynamical systems, and the Newtonian equation
\[
\nabla_{\dot\gamma}\dot\gamma=-\grad_g f
\]
for conservative systems. If \(\vartheta\) is viewed as a quadratic polynomial \(\vartheta^v\) on \(T^*M\), then
\[
(\vartheta,\nabla)\text{ is symmetric Poisson}\iff \grad_\nabla(\vartheta^v)\ \text{is isotropic},
\]
and an integral curve \(a\) of \(\grad_\nabla(\vartheta^v)\) satisfies
\[
\dot\gamma_a=\vartheta(a),\qquad
\nabla_{\dot\gamma_a}a=\frac12(\nabla\vartheta)(a,a),\qquad
\nabla_{\dot\gamma_a}\dot\gamma_a=\frac14\,\iota_a\iota_a[\vartheta,\vartheta]_s.
\]
Thus \(\gamma_a\) is a geodesic for every such integral curve exactly when \([\vartheta,\vartheta]_s=0\).

In the linear case, with \(M=V^*\), a symmetric Poisson bracket is linear if it preserves linear functions, and the induced commutative product on \(V\) is defined by
\[
\iota_{u\cdot v}:=\{\iota_u,\iota_v\}.
\]
Linear symmetric Poisson structures correspond to Jacobi-Jordan algebras, that is commutative algebras satisfying
\[
\Jac(u,v,w):=u\cdot(v\cdot w)+v\cdot(w\cdot u)+w\cdot(u\cdot v)=0.
\]
Strong linear symmetric Poisson structures correspond to associative Jacobi-Jordan algebras. The theory contains flat pseudo-Riemannian examples, rank-one structures \(\vartheta=fX\otimes X\), Lie-group examples, and a \(5\)-dimensional linear example showing that
\[
\text{symmetric} \supsetneq \text{involutive symmetric} \supsetneq \text{strong symmetric}.
\]

## 5. Contravariant pseudo-Hessian and Koszul-Vinberg theories

A contravariant pseudo-Hessian manifold is a triple
\[
(M,\nabla,h),
\]
where \(\nabla\) is flat and \(h\in \Gamma(S^2TM)\) is a symmetric bivector satisfying the contravariant Codazzi equation
\[
(\nabla_{h^\sharp(\alpha)}h)(\beta,\gamma)
=
(\nabla_{h^\sharp(\beta)}h)(\alpha,\gamma).
\]
In affine coordinates, this becomes
\[
\sum_{l=1}^n h^{li}\,\partial_l h^{jk} - h^{lj}\,\partial_l h^{ik} =0.
\]
From \(h\) one canonically constructs a skew Poisson tensor \(\Pi\) on \(TM\) by
\[
\Pi(\alpha^h,\beta^h)=0,\qquad
\Pi(\alpha^v,\beta^v)=0,\qquad
\Pi(\alpha^h,\beta^v)=h(\alpha,\beta)\circ p,
\]
or locally
\[
\Pi=\sum_{i,j=1}^n h^{ij}(x)\,\frac{\partial}{\partial x^i}\wedge \frac{\partial}{\partial y^j}.
\]
The fundamental equivalence is
\[
(M,\nabla,h)\text{ is contravariant pseudo-Hessian}
\quad\Longleftrightarrow\quad
(TM,\Pi)\text{ is a Poisson manifold}.
\]
Moreover, \(\operatorname{Im} h^\sharp\) is integrable, its leaves carry pseudo-Hessian structures, and for every leaf \(L\), the tangent bundle \(TL\) is a symplectic leaf of \((TM,\Pi)\); with the induced symplectic form and the tangent-bundle complex structure, \(TL\) becomes a pseudo-Kähler manifold. The paper also notes that a full Darboux-Weinstein-type local normal form fails in general, although a corank-one normal form exists [2001.03776].

The Koszul-Vinberg approach shifts from affine manifolds to left-symmetric algebroids. For a left-symmetric algebroid \((A,\cdot_A,a_A)\), a symmetric tensor \(H\in \Gamma(S^2A)\) is a Koszul-Vinberg structure when
\[
[H,H]_A=0.
\]
This is the direct analogue of \([\pi,\pi]=0\) for a Poisson bivector. If \(H^\sharp\) is nondegenerate, then
\[
[H,H]_A=0 \quad\Longleftrightarrow\quad d_A(H^{-1})=0,
\]
so nondegenerate Koszul-Vinberg tensors correspond to pseudo-Hessian \(2\)-cocycles. The paper develops the full compatibility package parallel to Poisson-Nijenhuis theory: compatible KV structures, Koszul-Vinberg-Nijenhuis structures \((H,N)\), the hierarchy
\[
(H_{N^k})^\sharp:=N^k\circ H^\sharp,
\]
KV\(\Omega\)-structures \((H,B)\), complementary symmetric \(2\)-tensors \(B\), and pseudo-Hessian-Nijenhuis structures \((B,N)\). If \((H,N)\) is KVN, then every \(H_{N^k}\) is KV and
\[
[H_{N^k},H_{N^\ell}]_A=0
\]
for all \(k,\ell\ge 0\) [2004.01774].

The Jacobi extension replaces left-symmetric algebroids by Jacobi-left-symmetric algebroids \((A,\phi_0)\), where \(d_A\phi_0\) is symmetric, equivalently \(d_A\phi_0=0\). For \(h\in\Gamma(S^2A)\), the modified self-bracket is
\[
[h,h]_{2\phi_0}(a,B,y)
:=
[h,h]_A(a,B,y)+h(\phi_0,a)h(B,y)-h(\phi_0,B)h(a,y),
\]
and \(h\) is Jacobi-Koszul-Vinberg when
\[
[h,h]_{2\phi_0}=0.
\]
If this holds, then \(A^*\) acquires a Jacobi-left-symmetric algebroid structure. The theory also contains a Poissonization-like construction:
\[
\widetilde h:=e^{-t}h,
\]
and \(h\) is Jacobi-Koszul-Vinberg on \((A,\phi_0)\) if and only if \(\widetilde h\) is a Koszul-Vinberg structure on an extended left-symmetric algebroid over \(A\times \mathbb R\). On an affine manifold \((M,\nabla)\), a Jacobi-Koszul-Vinberg structure is a pair \((h,E)\) satisfying three explicit conditions, including
\[
(\nabla_a h)(B,y)-(a,E)h(B,y)
=
(\nabla_B h)(a,y)-(B,E)h(a,y),
\]
\[
(\nabla_E h)(B,y)-(y,\nabla_{h^\#B}E)+(B,E)(y,E)=0,
\qquad
\nabla_EE=0.
\]
In the nondegenerate case, the corresponding metric data produce a semi-Weyl manifold and, in fact, a locally conformally Hessian manifold [2412.20103].

## 6. Shifted, categorical, and algebraic variants

In derived algebraic geometry, shifted Poisson structures on a CDGA \(A\) are Maurer-Cartan elements
\[
\pi=\sum_{m\ge 2}\pi^{(m)}
\]
in the completed shifted polyvector dg Lie algebra, satisfying
\[
d\pi+\frac12[\pi,\pi]=0.
\]
For the Chevalley-Eilenberg algebra of an ordinary Lie algebra,
\[
CE(g)=\operatorname{Sym}(g^*[-1]),
\]
the low-shift classification is especially explicit. A \(2\)-shifted Poisson structure is exactly an invariant symmetric tensor
\[
\pi^{(2,0)}\in \operatorname{Sym}^2(g),
\qquad
(\operatorname{ad}_x\otimes \mathrm{id}+\mathrm{id}\otimes \operatorname{ad}_x)\,\pi^{(2,0)}=0,
\]
so the classification is
\[
(\operatorname{Sym}^2 g)^g.
\]
A \(1\)-shifted Poisson structure is equivalent to quasi-Lie bialgebra data. For Lie \(N\)-algebras, all \(n\)-shifted Poisson structures are trivial when
\[
n>2N.
\]
Hence only \(n=1,2\) can be nontrivial for ordinary Lie algebras, whereas Lie \(2\)-algebras admit nontrivial shifts \(n\in\{1,2,3,4\}\), with \(n=4\) again governed by invariant symmetric tensors and the lower shifts described by finite higher multilinear data except in the \(n=1\) case, where an infinite tower remains [2412.12804].

A higher-categorical refinement appears in the study of \(2\)- and \(3\)-shifted Poisson structures on semi-free CDGAs. A \(3\)-shifted Poisson structure has weight-\(2\) component
\[
\pi^{(2)}_{n=3}\in \operatorname{Sym}^2_A(T_A[-1])^{-1},
\]
and this produces a coexact homotopy yielding an infinitesimal syllepsis
\[
T_{MN}:(M\otimes_A N)\to (M\otimes_A N)[-1]
\]
on the symmetric strict monoidal \(2\)-category
\[
\big(\mathsf{Ho}_2({}_AdgMod^{sf}),\otimes_A,A,\gamma\big).
\]
The induced syllepsis is
\[
\Sigma:=\gamma T.
\]
By contrast, \(2\)-shifted Poisson structures produce infinitesimal \(2\)-braidings. Their second-order integration requires total symmetry,
\[
t_{\gamma_{UV}\boxtimes 1_W}=0,
\]
and coherency, which in the symmetric strict case can be written as
\[
L+L_{231}+L_{132}=\mathbf{0}.
\]
The obstruction is controlled by the third-weight Maurer-Cartan component,
\[
d_{\widehat{Pol}}(\pi^{(3)}) + \tfrac12[\pi^{(2)},\pi^{(2)}] = 0,
\]
so higher shifted Poisson data are read categorically as braidings and syllepses [2505.01949].

A different algebraic use of “symmetric Poisson” occurs for oscillator Lie algebras. There the Poisson structure is a commutative associative product \(\circ\) on the Lie algebra satisfying the Leibniz rule
\[
[u,v\circ w]=[u,v]\circ w + v\circ [u,w].
\]
For a generic oscillator Lie algebra \(\mathfrak g_\lambda\), the classification is rigid: there exists \(c\in \mathbb R\) such that the only nonzero product is
\[
e_{-1}\circ e_{-1}=c\,e_0,
\]
and all other products between basis elements vanish. The associated Lie-admissible product
\[
u\cdot v=[u,v]+u\circ v
\]
is then a symmetric Leibniz algebra. On the group side, these products correspond to bi-invariant torsion-free connections with the same curvature as the canonical bi-invariant connection
\[
V^0_XY=\tfrac12[X,Y],
\]
and if the Poisson structure is symmetric Leibniz the corresponding connection is locally symmetric and has the same holonomy as \(V^0\) [2003.12608].

Taken together, these developments suggest that “symmetric Poisson structure” is best understood as a family resemblance rather than a single axiom system. The recurring pattern is the replacement of skew-symmetric Hamiltonian data by symmetric \(2\)-tensors, symmetric multilinear operations, or symmetric-space symmetries, together with an auxiliary integrability mechanism—torsion-free connections, contravariant Codazzi equations, Koszul-Vinberg brackets, Maurer-Cartan towers, or Poisson-Lie reduction—that restores the structural role played by the Poisson condition in the skew-symmetric theory.

Source: https://www.emergentmind.com/topics/symmetric-poisson-structures