---
title: Almost Born Structures and Integrability
url: https://www.emergentmind.com/topics/almost-born-structure
type: topic
---

# Almost Born Structures and Integrability

An almost Born structure is a geometric structure that packages para-quaternionic-type endomorphisms together with compatible metric and symplectic data. In the tangent-bundle formulation, it is a tuple
\[
(I,J,K,h,k,\omega),
\]
where \((I,J,K)\) is an almost para-quaternionic structure, \(h\) is a Riemannian metric, \(k\) is a pseudo-Riemannian metric of signature \((n,n)\), and \(\omega\) is a nondegenerate \(2\)-form, subject to the compatibility relations
\[
I=h^{-1}\circ \omega,\qquad J=k^{-1}\circ h,\qquad K=\omega^{-1}\circ k.
\]
In equivalent formulations, the same package is expressed by recursion operators or by a Künneth splitting. The qualifier “almost” indicates the absence of integrability assumptions; when appropriate closedness and Nijenhuis-type conditions are imposed, one obtains an integrable Born structure. Recent work has clarified that almost Born structures form a common framework for para-Hermitian geometry, pseudo-Kähler geometry, Hessian geometry, tangent-bundle constructions, and Lie-algebraic matched-pair constructions [2507.23264].

## 1. Algebraic package and defining relations

An almost para-quaternionic structure is a triple \((I,J,K)\) of \((1,1)\)-tensor fields satisfying
\[
-I^2=J^2=K^2=\mathrm{id},\qquad IJK=-\mathrm{id}.
\]
Equivalently,
\[
I=JK=-KJ,\qquad J=-KI=IK,\qquad K=-IJ=JI.
\]
An almost Born structure augments this triple by metric and \(2\)-form data so that the three endomorphisms and the tensors \((h,k,\omega)\) determine each other.

In the notation used for tangent-bundle Born structures, \(I\) is the almost complex-type endomorphism, while \(J\) and \(K\) are para-complex-type endomorphisms. In the recursion-operator formulation, one instead starts from a triple \((g,h,\omega)\), where \(g\) and \(h\) are pseudo-Riemannian metrics and \(\omega\) is a non-degenerate \(2\)-form, and defines recursion operators through the diagram
\[
g \xrightarrow{A} \omega \xrightarrow{-J} h,\qquad g \xrightarrow{B} h.
\]
Here the recursion operator from \(a\) to \(b\) is the unique endomorphism \(A\) such that
\[
a(A\cdot,\cdot)=b(\cdot,\cdot).
\]
This formulation yields the identities
\[
A^2=B^2=-J^2=\operatorname{Id},\qquad ABJ=\operatorname{Id},
\]
and the operators \(A,B,J\) pairwise anti-commute [2410.15402].

The metric signatures are constrained by the algebra. In the Künneth/recursion-operator formulation, the metric \(g\) is automatically neutral, while Proposition 16 gives the signature restriction for \(h\):
\[
(2p,2q),\qquad p+q=n=\frac12\dim M.
\]
Moreover,
\[
h(AX,AY)=h(BX,BY)=h(JX,JY)=h(X,Y),
\]
so \((h,J)\) is an almost pseudo-Hermitian structure with fundamental form \(\omega\) [2410.15402].

## 2. Equivalent formulations: para-quaternionic, Künneth, and generalized-geometric

A central development in the recent literature is the identification of equivalent descriptions of Born geometry. One may work with para-quaternionic tensors, with recursion operators, or with a Künneth structure; the corresponding formulations are stated to be equivalent reformulations of the same underlying package [2410.15402].

At the Künneth level, the basic object is an almost Künneth structure
\[
(\omega,F,G),
\]
where \(\omega\) is a non-degenerate \(2\)-form and \(F,G\subset TM\) are complementary \(\omega\)-isotropic subbundles. From such a splitting one obtains the associated almost product structure
\[
I|_F=+1,\qquad I|_G=-1,
\]
and the neutral metric
\[
g(X,Y)=\omega(IX,Y).
\]
If \(\omega\) is closed and \(F,G\) are integrable to Lagrangian foliations, the structure is Künneth.

The recursion-operator picture makes the decomposition particularly explicit. Writing \(L_\pm\) for the \(\pm1\)-eigenspaces of \(A\), one has
\[
TM=L_+\oplus L_-,
\]
and Proposition 12 shows that \(L_\pm\) are Lagrangian for \(\omega\). Corollary 13 states that \(L_\pm\) are null for the neutral metric \(g\). Section 3.3 then proves the converse construction: given any almost Künneth structure \((\omega,F,G)\), one may choose an isomorphism
\[
\widetilde J:F\to G,\qquad \omega(\widetilde JX,Y)=-\omega(X,\widetilde JY),
\]
extend it to an almost complex structure by
\[
J|_F=\widetilde J,\qquad J|_G=-\widetilde J^{-1},
\]
and define
\[
h(X,Y)=\omega(X,JY).
\]
Thus a Born structure is exactly an almost Künneth structure together with a compatible almost complex structure exchanging the two Lagrangian summands [2410.15402].

This equivalence also explains why Born geometry is stronger than an almost para-Hermitian or almost pseudo-Hermitian structure alone. The data simultaneously encode a neutral metric, a pseudo-Hermitian metric, a non-degenerate \(2\)-form, and a para-quaternionic-type family of endomorphisms.

## 3. Integrability and canonical connections

The distinction between almost and integrable Born geometry is sharp. In the Künneth formulation, “almost” means that \(\omega\) need not be closed and the recursion operators \(A,B,J\) need not be integrable. Integrability requires that \(\omega\) is closed and that at least two of the recursion operators are integrable; Proposition 21 then implies that all three are integrable [2410.15402].

Integrability is expressed through the Nijenhuis tensor
\[
N_T(X,Y)=[TX,TY]+T[X,Y]-T[TX,Y]-T[X,TY].
\]
For an almost product structure, vanishing \(N_T\) is equivalent to Frobenius integrability of the eigendistributions; for an almost complex structure, it is the Newlander–Nirenberg condition. In the integrable Born case, the compatible pair \((h,J)\) becomes pseudo-Kähler with Kähler form \(\omega\), and \((\omega,L_+,L_-)\) becomes a genuine Künneth or para-Kähler structure [2410.15402].

The tangent-bundle literature distinguishes two notions. An almost Born structure \((I,J,K,h,k,\omega)\) is **integrable** if \(I\), \(J\), and \(K\) are individually integrable and \(d\omega=0\). It is **strongly integrable** if the whole almost para-quaternionic triple \((I,J,K)\) is integrable in the simultaneous local normal form sense and \(d\omega=0\). In general, strong integrability is stricter than integrability, but for the tangent-bundle structures induced from \((\nabla,g)\) they coincide [2507.23264].

Connections organize these integrability conditions. For an almost Künneth structure there is a canonical Künneth connection \(\nabla^K\), the unique affine connection preserving \(F,G,\omega\) with vanishing mixed torsion; Theorem 4 says that it is torsion-free exactly when the Künneth structure is integrable, and Theorem 5 identifies it with the Levi-Civita connection of \(g\). The canonical connection associated to the almost Künneth structure is
\[
\nabla^c_XY=\frac12\left(\nabla^g_XY+A\,\nabla^g_X(AY)\right).
\]
Corollary 28 states that \(\nabla^c=\nabla^K\) iff
\[
d\omega^{1,2}=0=d\omega^{2,1}
\]
with respect to the \(L_+\oplus L_-\) bigrading; in particular, if \(\omega\) is closed, then \(\nabla^c=\nabla^K\). The averaged connection
\[
\nabla_XY=\frac12\left(\nabla^K_XY+B\,\nabla^K_X(BY)\right)
=\frac12\left(\nabla^K_XY-J\,\nabla^K_X(JY)\right)
\]
is compatible with the full Born structure, and in the integrable case agrees with the Born connection of Freidel–Rudolph–Svoboda [2410.15402].

## 4. Natural almost Born structures on tangent bundles

Given a manifold \(M\) endowed with an affine connection \(\nabla\) and a Riemannian metric \(g\), the tangent bundle \(TM\) carries a natural almost para-quaternionic structure induced by \(\nabla\), and if \(g\) is also given, then one gets a natural almost Born structure [2507.23264].

In local coordinates \((\theta^1,\dots,\theta^n)\) on \(M\), the induced coordinates on \(TM\) are
\[
x^i=\theta^i\circ\pi,\qquad y^i=d\theta^i,
\]
where \(\pi:TM\to M\) is the projection. Using the connection coefficients \(\Gamma^k_{ij}\) of \(\nabla\), one defines the adapted frame
\[
H_i=\frac{\partial}{\partial x^i}-(\Gamma^k_{ij}\circ\pi)y^j\frac{\partial}{\partial y^k},\qquad
V_i=\frac{\partial}{\partial y^i},
\]
with dual coframe
\[
H^{*i}=dx^i,\qquad
V^{*i}=(\Gamma^i_{jk}\circ \pi)y^k\,dx^j+dy^i.
\]

The induced endomorphisms are
\[
I = V_i\otimes H^{*i} - H_i\otimes V^{*i},\qquad
J = V_i\otimes H^{*i} + H_i\otimes V^{*i},\qquad
K = H_i\otimes H^{*i} - V_i\otimes V^{*i}.
\]
With respect to \((H_1,\dots,H_n,V_1,\dots,V_n)\), they take the block forms
\[
I=\begin{pmatrix}0&-1_n\\ 1_n&0\end{pmatrix},\qquad
J=\begin{pmatrix}0&1_n\\ 1_n&0\end{pmatrix},\qquad
K=\begin{pmatrix}1_n&0\\ 0&-1_n\end{pmatrix}.
\]

If \(g=(g_{ij})\) is a Riemannian metric on \(M\), one defines
\[
h=(g_{ij}\circ\pi)\bigl(H^{*i}\otimes H^{*j}+V^{*i}\otimes V^{*j}\bigr),
\]
\[
k=(g_{ij}\circ\pi)\bigl(H^{*i}\otimes V^{*j}+V^{*j}\otimes H^{*i}\bigr),
\]
\[
\omega=(g_{ij}\circ\pi)\bigl(H^{*i}\otimes V^{*j}-V^{*j}\otimes H^{*i}\bigr).
\]
Equivalently,
\[
h=\begin{pmatrix}G&0\\ 0&G\end{pmatrix},\qquad
k=\begin{pmatrix}0&G\\ G&0\end{pmatrix},\qquad
\omega=\begin{pmatrix}0&G\\ -G&0\end{pmatrix},
\]
where \(G=(g_{ij}\circ\pi)\). Thus \((I,J,K,h,k,\omega)\) is an almost Born structure on \(TM\) [2507.23264].

The local commutator relations are
\[
[H_i,H_j]=-(R_{ijk}^l\circ\pi)\,y^k V_l,\qquad [V_i,V_j]=0,\qquad [H_i,V_j]=-\Gamma_{ij}^k V_k.
\]
These formulas control the Nijenhuis tensors of the induced endomorphisms. For example,
\[
N_J(H_i,H_j)=T_{ij}^kH_k-R_{ijk}^l y^kV_l,
\]
\[
N_J(V_i,V_j)=T_{ij}^kH_k-R_{ijk}^l y^kV_l,
\]
\[
N_J(H_i,V_j)=R_{ijk}^l y^kH_l-T_{ij}^kV_k,
\]
where \(T_{ij}^k\) are the torsion coefficients of \(\nabla\). This shows that \(J\) is integrable exactly when \(\nabla\) is flat [2507.23264].

## 5. Hessian structures and the tangent-bundle integrability theorem

A Hessian structure on a manifold \(M\) is a pair \((\nabla,g)\) consisting of an affine connection and a Riemannian metric such that \(\nabla\) is flat,
\[
R^\nabla=0,\qquad T^\nabla=0,
\]
and the tensor \(\nabla g\) is totally symmetric. Equivalently, using the dual connection \(\nabla^*\) defined by
\[
Xg(Y,Z)=g(\nabla_XY,Z)+g(Y,\nabla^*_XZ),
\]
the following standard facts are recalled: if any two of
- \(\nabla\) torsion-free,
- \(\nabla^*\) torsion-free,
- \(\nabla g\) symmetric,
- \(\frac12(\nabla+\nabla^*)\) is the Levi-Civita connection of \(g\),

hold, then all four hold. In addition,
\[
R^\nabla=0\iff R^{\nabla^*}=0.
\]

The key proposition for the induced structure on \(TM\) gives the equivalences
\[
\text{(i)}\ \nabla \text{ is flat} \iff I \text{ is integrable} \iff J \text{ is integrable},
\]
\[
\text{(ii)}\ R^\nabla = 0 \iff K \text{ is integrable},
\]
\[
\text{(iii)}\ \nabla^* \text{ is torsion-free} \iff d\omega=0.
\]
From these, the main theorem follows:
\[
(\nabla,g)\text{ is a Hessian structure}
\Longleftrightarrow
(I,J,K,h,k,\omega)\text{ is integrable}
\Longleftrightarrow
(I,J,K,h,k,\omega)\text{ is strongly integrable}.
\]
More explicitly, if \((\nabla,g)\) is Hessian, then \(\nabla\) is flat and torsion-free, \(\nabla g\) is totally symmetric, and hence \(d\omega=0\) and all of \(I,J,K\) are integrable. Conversely, if the induced Born structure is integrable, then the integrability of \(I,J,K\) forces \(\nabla\) to be flat, and \(d\omega=0\) forces \(\nabla^*\) to be torsion-free; together with the standard Hessian identities, this implies that \((\nabla,g)\) is Hessian [2507.23264].

In an affine chart for a flat \(\nabla\), the formulas simplify to the constant block forms above. For the standard Hessian manifold \(\mathbb{R}^n\) with the Euclidean metric and standard flat connection, the induced structure on \(T\mathbb{R}^n\) identifies naturally with the standard strongly integrable Born structure on \(\mathbb{C}^n\) via
\[
T\mathbb{R}^n \to \mathbb{C}^n,\qquad (x,v)\mapsto x+i v.
\]

This theorem extends two earlier lines of work. Dombrowski showed that an affine connection induces natural almost complex and para-complex structures on the tangent bundle, with integrability tied to flatness. Satoh and later Marotta–Szabo showed that adding a metric yields an almost Hermitian or para-Hermitian-type structure whose Kähler condition is equivalent to the Hessian condition. The Born-geometric reformulation places these facts inside the broader framework introduced by Freidel–Leigh–Minic and developed mathematically by Freidel–Rudolph–Svoboda and Marotta–Szabo: Hessian geometry is the precise base-space geometry underlying integrable Born geometry on the tangent bundle [2507.23264].

## 6. Lie-algebraic models, bicross products, and classification

On a Lie algebra \(\mathfrak g\), a Born structure is again given by a triple
\[
(g,h,\omega),
\]
with recursion operators satisfying
\[
A^2=B^2=-J^2=\operatorname{Id},
\]
and it is called integrable if \(\omega\) is closed in the Chevalley–Eilenberg sense, \(J\) has vanishing Nijenhuis tensor, and the \(\pm1\)-eigenspaces of \(A\) are Lie subalgebras. A Lie algebra equipped with such an integrable structure is called a Born Lie algebra [2411.04856].

The eigenspace decomposition is structurally rigid. If \(\mathfrak g_\pm\) are the eigenspaces of \(A\), then
\[
\omega|_{\mathfrak g_\pm}=0,\qquad g|_{\mathfrak g_\pm}=0,\qquad \mathfrak g_+\perp_h \mathfrak g_-,
\]
and \(J\) interchanges them. The paper emphasizes the equivalence
\[
\text{integrable Born structure}
\Longleftrightarrow
\text{pseudo-K\"ahler structure }(h,J)\text{ plus complementary orthogonal subalgebras } \mathfrak g_\pm.
\]
Equivalently, a Born Lie algebra simultaneously encodes a pseudo-Kähler geometry \((h,J)\), a bi-Lagrangian or para-Kähler geometry \((\omega,\mathfrak g_+,\mathfrak g_-)\), and a complex product structure \((J,A)\) [2411.04856].

A major structural theorem states that every Born Lie algebra arises from a bicross product of two pseudo-Riemannian Lie algebras. Given Lie algebras \(\mathfrak g_+\) and \(\mathfrak g_-\) with representations
\[
\varphi:\mathfrak g_- \to \operatorname{End}(\mathfrak g_+),\qquad
\rho:\mathfrak g_+ \to \operatorname{End}(\mathfrak g_-),
\]
satisfying the matched-pair conditions, the bicross product
\[
\mathfrak g_\varphi^\rho=\mathfrak g_+\bowtie_\varphi^\rho \mathfrak g_-
\]
has bracket
\[
[X_++X_-,Y_++Y_-]
=
[X_+,Y_+]_+
+\varphi(X_-)Y_+-\varphi(Y_-)X_+
+[X_-,Y_-]_-
+\rho(X_+)Y_--\rho(Y_+)X_-.
\]
If \(h_\pm\) are pseudo-Riemannian metrics of the same signature and
\[
Q:(\mathfrak g_+,h_+)\to(\mathfrak g_-,h_-)
\]
is a linear isometry, then
\[
h=h_+\oplus h_-, \qquad
J=\begin{pmatrix} 0 & -Q^{-1}\\ Q & 0 \end{pmatrix}
\]
defines an almost Born structure. The integrability conditions are then expressed by the four relations
\[
\varphi(X_-)^a = Q^{-1}\nabla^-_{X_-}Q,
\qquad
\varphi(QX_+)^sY_+ = \varphi(QY_+)^sX_+,
\]
\[
\rho(X_+)^a = Q\nabla^+_{X_+}Q^{-1},
\qquad
\rho(X_+)^sQY_+ = \rho(Y_+)^sQX_+.
\]
Conversely, every Born Lie algebra is isomorphic to such a bicross product [2411.04856].

This bicross-product description is strong enough to support classification results. In dimension \(2\), every Lie algebra is Born, and any Born structure is automatically integrable. In dimension \(4\), a non-abelian Lie algebra admits an integrable Born structure iff it is isomorphic to one of
\[
\mathfrak{rh}_3,\ \mathfrak{rr}_{3,0},\ \mathfrak r_2\mathfrak r_2,\ \mathfrak r_2',\ \mathfrak r_{4,-1,-1},\ \mathfrak d_{4,1},\ \mathfrak d_{4,2},\ \mathfrak d_{4,1/2}.
\]
Moreover, a \(4\)-dimensional Lie algebra admits an integrable Born structure iff it admits both a pseudo-Kähler structure and a bi-Lagrangian structure. For non-abelian \(6\)-dimensional nilpotent Lie algebras, the complete list is
\[
\mathfrak h_4,\ \mathfrak h_7,\ \mathfrak h_8,\ \mathfrak h_9,\ \mathfrak h_{10},\ \mathfrak h_{11},\ \mathfrak h_{13}.
\]

The curvature behavior of the associated metrics is also constrained. In the \(4\)-dimensional classified cases, the neutral metric \(g\) is always flat. In the \(6\)-dimensional nilpotent case, the metrics \(h_\pm\) on the two subalgebras are always flat, the global metric \(h\) may be flat or non-flat, and \(\mathfrak h_{13}\) yields an example where the neutral metric \(g\) can also be non-flat for certain parameter values [2411.04856].

A recurrent misconception is that the almost and integrable levels differ only mildly. The available results indicate the opposite. On manifolds, almost Born data may be constructed from a symplectic splitting plus a compatible almost complex structure, but integrability requires closedness and Nijenhuis conditions. On tangent bundles, the induced almost Born structure is integrable exactly in the Hessian case. On Lie algebras, integrability is equivalent to a pseudo-Kähler structure together with complementary orthogonal subalgebras, and in examples on nilmanifolds, infra-nilmanifolds, and solvmanifolds, the extra complex and Kähler conditions can obstruct integrable Born geometry even when almost Künneth data exist [2410.15402].

Source: https://www.emergentmind.com/topics/almost-born-structure