Cohomologically Calibrated Affine Connections

• Cohomologically calibrated affine connections are affine connections with totally antisymmetric torsion whose associated 3-form represents a prescribed nontrivial de Rham class in H³, encoding global topology.
• The framework uses Hodge decomposition to split the torsion into harmonic, exact, and co-exact parts, allowing precise control over curvature, Ricci tensor structure, and Einstein solvability.
• Applications on manifolds like S²×T² demonstrate how the choice of harmonic versus non-harmonic torsion representatives affects Einstein conditions and guarantees holonomic irreducibility.

Searching arXiv for the cited paper and closely related work on cohomologically calibrated affine connections. Cohomologically calibrated affine connections are affine connections whose torsion is constrained by a prescribed nontrivial class in $H^3(M;\mathbb{R})$, so that the harmonic component of the torsion encodes global topology directly in the connection. In the formulation developed for non-Riemannian Einstein geometry, one replaces the torsion-free Levi-Civita connection by a connection with totally antisymmetric torsion and then requires the associated 3-form $T^\flat$ to represent a fixed de Rham class. The resulting framework was applied in detail to $S^2\times T^2$, where the topology of the manifold, the Hodge decomposition of $T^\flat$, and the Einstein equation interact in a highly rigid way (Pigazzini et al., 4 Aug 2025).

1. Definition and basic geometric setup

The basic object is an affine connection $\nabla$ with totally antisymmetric torsion $T$, considered in place of the Levi-Civita connection. The associated 3-form is defined by

$T^\flat(X,Y,Z)=g(T(X,Y),Z).$

Cohomological calibration requires $T^\flat$ to represent a nontrivial cohomology class in $H^3(M;\mathbb{R})$. In this sense, the torsion is not merely a local deformation of the connection: its harmonic component is prescribed by global topology (Pigazzini et al., 4 Aug 2025).

On $S^2\times T^2$, the relevant topological fact is

$T^\flat$0

Accordingly, the calibration class is parametrized by two real numbers. The framework organizes all admissible torsions through the family

$T^\flat$1

where $T^\flat$2 is the unique harmonic representative of the de Rham class, while $T^\flat$3 and $T^\flat$4 are exact and co-exact 3-forms. This formulation separates the topological datum, carried by $T^\flat$5, from the non-harmonic freedom in the choice of torsion (Pigazzini et al., 4 Aug 2025).

A central feature of the theory is therefore the distinction between cohomological equivalence and geometric equivalence. All torsions in $T^\flat$6 encode the same de Rham class, but they do not generally yield the same Ricci tensor, curvature structure, or Einstein behavior. This dependence on the representative rather than only the class is one of the defining structural properties of the subject.

2. Hodge decomposition and the calibrated torsion family

Hodge theory gives the unique decomposition

$T^\flat$7

with $T^\flat$8 harmonic. Only $T^\flat$9 represents the nontrivial de Rham cohomology; the exact and co-exact parts are geometrically active but cohomologically trivial (Pigazzini et al., 4 Aug 2025).

For $S^2\times T^2$0, with orthonormal coframe $S^2\times T^2$1, the harmonic representative takes the explicit form

$S^2\times T^2$2

where $S^2\times T^2$3 correspond to the two generators of $S^2\times T^2$4. The associated harmonic torsion tensor $S^2\times T^2$5, defined by $S^2\times T^2$6, has the nonzero components

$S^2\times T^2$7

This explicit formula is the local algebraic realization of the calibration class (Pigazzini et al., 4 Aug 2025).

The framework treats $S^2\times T^2$8 as a family of admissible torsions rather than a single canonical object. A recurrent misconception is that the cohomological condition fixes the geometry uniquely. The opposite is true: the topology fixes only the harmonic part, while the exact and co-exact components remain available and can alter curvature and Einstein feasibility. The theory is therefore calibrated but not rigid in the naive sense.

3. Ricci curvature, torsion, and the Einstein condition

For a connection of the form $S^2\times T^2$9 with totally antisymmetric torsion, the Ricci tensor remains symmetric but acquires torsion-dependent contributions. In an orthonormal basis $T^\flat$0,

$T^\flat$1

where the sectional curvature depends on both the metric and the torsion contributions. The Einstein condition is

$T^\flat$2

for some constant $T^\flat$3 (Pigazzini et al., 4 Aug 2025).

The detailed computation on $T^\flat$4 shows that off-diagonal Ricci entries can be generated by the topology-driven torsion, especially $T^\flat$5. This mechanism creates a direct link between the calibration parameters $T^\flat$6 and the solvability of the Einstein equations. In particular, cohomological calibration is not merely compatible with curvature conditions; it actively reshapes the algebraic structure of the Ricci tensor.

The comparison between two torsion choices is especially instructive.

Torsion choice	Ricci structure	Einstein outcome
Canonical torsion with positive biorthogonal curvature	Non-diagonal; $T^\flat$7	No Einstein solution
Harmonic representative $T^\flat$8	Diagonal	Einstein for $T^\flat$9

For the canonical torsion previously used to obtain strictly positive biorthogonal curvature, the Ricci tensor has nonzero diagonal entries involving $\nabla$0 and the off-diagonal term

$\nabla$1

The Einstein condition forces $\nabla$2, so the cohomology class must be “pure,” but the remaining algebraic system is inconsistent. Consequently, no Einstein connection exists for this torsion choice, and this includes the Levi-Civita case (Pigazzini et al., 4 Aug 2025).

By contrast, for the harmonic representative $\nabla$3, all off-diagonal Ricci entries vanish, and the diagonal entries become

$\nabla$4

$\nabla$5

These equations admit a solution if and only if

$\nabla$6

in which case the Einstein constant is $\nabla$7, so

$\nabla$8

This gives an explicit non-Riemannian Einstein connection on $\nabla$9, with the calibration parameters determined by the topology-geometry interaction (Pigazzini et al., 4 Aug 2025).

4. Representative dependence, structural obstructions, and common misunderstandings

The $T$0 analysis establishes that the Einstein problem depends decisively on the representative of the cohomology class, not only on the class itself. The family $T$1 is large enough to contain geometrically incompatible torsions within the same topological calibration, and the difference between the positive-biorthogonal-curvature torsion and the harmonic torsion is decisive for Einstein existence (Pigazzini et al., 4 Aug 2025).

One misunderstanding is to identify cohomological calibration with the maximization of curvature positivity. The computations on $T$2 show that a torsion chosen to produce strictly positive biorthogonal curvature can obstruct the Einstein condition by generating unavoidable off-diagonal Ricci terms. The harmonic representative, rather than the curvature-positive choice, is the one that yields the Einstein solution. This suggests that within $T$3, curvature positivity and Einstein compatibility are distinct geometric objectives.

A second misunderstanding is to equate diagonal Ricci with geometric reducibility. Later work on product manifolds proves that a mixed calibration class can force irreducibility at the level of the full curvature tensor even when the Ricci tensor is diagonal, including the Einstein-calibrated case. The obstruction is therefore not exhausted by Ricci data; it resides in the off-diagonal structure of the full Riemann tensor (Pigazzini et al., 14 Aug 2025).

The main conceptual lesson is that harmonicity plays two roles at once. First, the harmonic part is the only part that represents the de Rham class. Second, in the $T$4 example, the harmonic representative is precisely the choice that removes off-diagonal Ricci terms and makes the Einstein equations solvable. The framework does not assert that harmonic representatives are always privileged in every problem, but the explicit computations identify them as decisive in this case.

5. Forced irreducibility on product manifolds

The framework was subsequently extended from the Einstein problem on $T$5 to a general irreducibility principle for product manifolds. For a product $T$6, if the calibration class $T$7 is mixed, then any cohomologically calibrated affine connection is holonomically irreducible with respect to the splitting $T$8 (Pigazzini et al., 14 Aug 2025).

The key mechanism is Hodge-theoretic. The harmonic part of the torsion induced by a mixed class has algebraic “mixing” properties across the product factors, and these generate nonzero off-diagonal components in the full curvature tensor. The non-harmonic pieces cannot globally cancel this effect. The formal non-cancellation is expressed by the integral orthogonality argument

$T$9

which rules out global elimination of the harmonic off-diagonal curvature contribution (Pigazzini et al., 14 Aug 2025).

The model example $T^\flat(X,Y,Z)=g(T(X,Y),Z).$0 is especially transparent because

$T^\flat(X,Y,Z)=g(T(X,Y),Z).$1

so every nontrivial class is mixed. In this case the off-diagonal block of the curvature tensor remains nonzero for every admissible connection, and this persists even when the Ricci tensor is diagonal. The result isolates a sharp distinction between Ricci diagonalization and holonomy reduction.

A further quantitative refinement gives a topological lower bound on the dimension of the off-diagonal holonomy subspace. For a metric connection with totally skew torsion calibrated by a mixed class on $T^\flat(X,Y,Z)=g(T(X,Y),Z).$2,

$T^\flat(X,Y,Z)=g(T(X,Y),Z).$3

where the right-hand side is the sum of the tensor ranks of the $T^\flat(X,Y,Z)=g(T(X,Y),Z).$4 and $T^\flat(X,Y,Z)=g(T(X,Y),Z).$5 Künneth components of the calibration class. This lower bound depends only on the mixed class and not on the product metric, thereby quantifying “forced irreducibility” in purely topological terms (Pigazzini et al., 15 Sep 2025).

6. Adjacent cohomological formulations and conceptual boundaries

The torsion-calibration framework belongs to a broader tendency in affine-connection theory to encode geometric data cohomologically, but it should not be conflated with other cohomological programs. In the theory of complex affine surfaces, the moduli space of affine surfaces is identified with the moduli space of regular meromorphic connections on Riemann surfaces, and the tangent space at $T^\flat(X,Y,Z)=g(T(X,Y),Z).$6 is identified with the hypercohomology group

$T^\flat(X,Y,Z)=g(T(X,Y),Z).$7

The derivative and coderivative of the holonomy map are also expressed by explicit morphisms of complexes. This is a cohomological control of deformation theory and holonomy rather than a torsion calibration by $T^\flat(X,Y,Z)=g(T(X,Y),Z).$8 (Apisa et al., 31 Jul 2025).

A different neighboring direction is the infinitesimal-algebraic encoding of affine connections by second-order infinitesimal groups. In that setting, symmetric affine connections correspond to abelian second-order i-group structures, while non-symmetric connections correspond to non-abelian ones, and torsion appears as the obstruction to symmetry. The truncated second-order Baker-Campbell-Hausdorff law

$T^\flat(X,Y,Z)=g(T(X,Y),Z).$9

encodes the Lie bracket at second order. This program is not presented in explicit cohomological language, but it provides an algebraic encoding of affine connections that is closely related to the role of torsion in calibrated theories (Bár, 2023).

Yet another distinct use of cohomology appears in Leibniz cohomology, where an affine connection is realized as a 1-cochain in the Leibniz cohomology complex of vector fields with adjoint coefficients. For the Levi-Civita connection, the coboundary decomposes into a Laplace-Beltrami term and a Ricci term. This yields a cohomological interpretation of curvature and spectral conditions, but again it is not the same as fixing a degree-three de Rham class for torsion (Lodder, 2020).

Canonical connection theory on almost complex manifolds supplies another adjacent framework. There one constructs canonical affine connections compatible with almost complex, conformal, or projective data by modifying a background affine connection with explicit tensors such as

$T^\flat$0

These constructions involve torsion characterizations and global invariant forms, including cohomological information carried by objects such as the Lee form, but they are organized by compatibility with $T^\flat$1, conformal structure, or projective structure rather than by calibration in $T^\flat$2 (Gover et al., 2012).

Taken together, these adjacent theories clarify the scope of cohomologically calibrated affine connections in the strict sense. The defining feature of the subject is not merely that cohomology appears somewhere in the theory of affine connections, but that a nontrivial class in $T^\flat$3 is imposed as a direct topological constraint on torsion and then allowed to govern curvature, Einstein solvability, and holonomy.