Hermitian Geometric Formality Metrics

• Hermitian geometric formality metrics are defined by the wedge-stability of harmonic forms in Dolbeault, Bott–Chern, and Aeppli cohomologies, leading to refined torus-type numerical bounds.
• They enforce rigid constraints on Hodge numbers, Betti numbers, and Albanese maps, distinguishing Kähler from non-Kähler complex structures.
• Applications span Calabi–Eckmann manifolds, solvmanifolds, and complex surfaces, while phenomena like blow-ups and Massey products reveal obstructions to formality.

Hermitian geometrically formal metrics are Hermitian metrics on compact complex manifolds for which harmonic representatives in a complex Hodge theory are stable under wedge product in the appropriate sense. The systematic framework developed for Dolbeault, Bott–Chern, and Aeppli cohomologies shows that this notion is a genuinely bigraded refinement of Kotschick’s classical geometric formality, with strong consequences for Hodge numbers, Betti numbers, Albanese maps, and the existence of special metrics on surfaces, solvmanifolds, and Calabi–Eckmann manifolds (Sferruzza et al., 14 Jul 2025).

1. Definitions and basic framework

Let $(M,J)$ be a compact complex manifold of complex dimension $n$, and let $g$ be a Hermitian metric with fundamental form

$\omega(X,Y)=g(JX,Y).$

The complexified exterior algebra splits into $(p,q)$-forms $\mathcal A^{p,q}(M)$, with $d=\partial+\bar\partial$. The three cohomologies used in Hermitian geometric formality are

$$H^{p,q}_{\bar\partial}(M)=\frac{\ker \bar\partial}{\operatorname{im}\bar\partial},\qquad H^{p,q}_{BC}(M)=\frac{\ker\partial\cap\ker\bar\partial}{\operatorname{im}\partial\bar\partial},\qquad H^{p,q}_{A}(M)=\frac{\ker\partial\bar\partial}{\operatorname{im}\partial+\operatorname{im}\bar\partial}.$$

Given $g$, one considers the Dolbeault, Bott–Chern, and Aeppli Laplacians, denoted $\Delta_{\bar\partial}$, $n$0, and $n$1. These operators are elliptic and self-adjoint and yield Hodge decompositions, so their harmonic spaces canonically identify with the corresponding cohomologies (Sferruzza et al., 14 Jul 2025).

The Hermitian formal notions are organized as follows.

Notion	Harmonicity condition	Structural consequence
Geometrically Dolbeault formal	$n$2 is a bigraded algebra	Harmonic Dolbeault representatives multiply correctly
Geometrically Bott–Chern formal	$n$3 is a bigraded algebra	Bott–Chern harmonic representatives are wedge-stable
ABC-geometrically formal	$n$4 is closed under $n$5, $n$6, and $n$7	Equivalent to closure under wedge and Hodge star, and to giving both Bott–Chern and Aeppli cohomology isomorphisms
Geometrically Aeppli formal	Aeppli-harmonic forms wedge with Bott–Chern-harmonic forms into Aeppli-harmonic forms	Equivalent to ABC-formality plus the $n$8-lemma

A basic structural point is that geometrically Aeppli formal metrics are much more rigid than a naive “Aeppli wedge-stability” condition might suggest. The equivalence proved in Lemma 2.1 shows that geometric Aeppli formality is the same as ABC-geometric formality together with the $n$9-lemma, and in that case the harmonic spaces coincide: $g$0 Moreover, the complex-valued de Rham harmonic forms decompose as

$g$1

and are closed under wedge (Sferruzza et al., 14 Jul 2025).

This places Hermitian geometric formality strictly above ordinary cohomological formality. In particular, geometrically Aeppli formal metrics imply strong formality in the sense of Milivojević–Stelzig, and all de Rham, Dolbeault, Bott–Chern, and Aeppli Massey-type products vanish (Sferruzza et al., 14 Jul 2025).

2. Constant norms, dimension bounds, and Albanese rigidity

A central analytic feature of Hermitian geometric formality is the constant-norm phenomenon. On a geometrically Dolbeault formal manifold, every Dolbeault harmonic form has constant pointwise norm, and the pointwise inner product of any two Dolbeault-harmonic forms is constant. The same conclusion holds for Bott–Chern and Aeppli harmonic forms under the stronger Hermitian formal assumptions: on ABC-geometrically formal manifolds, Bott–Chern and Aeppli harmonic forms have constant norm and constant mutual inner products, and the analogous statement holds in the Aeppli formal case as well (Sferruzza et al., 14 Jul 2025).

These constant-norm results lead to sharp torus-type inequalities. If $g$2 is geometrically Dolbeault formal, then for all $g$3,

$g$4

where

$g$5

In particular,

$g$6

For ABC-geometrically formal metrics, and more generally for geometrically Bott–Chern $g$7 formal metrics, analogous inequalities hold for Bott–Chern and Aeppli numbers: $g$8 and

$g$9

again implying torus bounds for the Betti numbers (Sferruzza et al., 14 Jul 2025).

Another basic obstruction comes from holomorphic $\omega(X,Y)=g(JX,Y).$0-forms. On a compact complex manifold with a geometrically Bott–Chern formal metric, every holomorphic $\omega(X,Y)=g(JX,Y).$1-form is $\omega(X,Y)=g(JX,Y).$2-closed. Equivalently, the $\omega(X,Y)=g(JX,Y).$3-lemma holds for $\omega(X,Y)=g(JX,Y).$4-forms. As a consequence,

$\omega(X,Y)=g(JX,Y).$5

and, by duality,

$\omega(X,Y)=g(JX,Y).$6

(Sferruzza et al., 14 Jul 2025).

The Albanese map becomes correspondingly rigid. If $\omega(X,Y)=g(JX,Y).$7 carries a geometrically Bott–Chern formal metric whose Bott–Chern harmonic $\omega(X,Y)=g(JX,Y).$8-forms have constant norm, then the Albanese map

$\omega(X,Y)=g(JX,Y).$9

is a holomorphic submersion, and the induced pullback is injective on Bott–Chern cohomology as well as on $(p,q)$0 and $(p,q)$1. If $(p,q)$2, then $(p,q)$3 is Kähler; if $(p,q)$4, then $(p,q)$5 is biholomorphic to a complex torus (Sferruzza et al., 14 Jul 2025).

These theorems show that Hermitian geometric formality is not merely a property of harmonic representatives: it constrains the global complex geometry and forces torus-like numerical behavior.

3. Compact complex surfaces

The surface case exhibits both positive examples and stringent exclusions. Among rational and ruled surfaces, minimal rational surfaces of type $(p,q)$6 and Hirzebruch surfaces $(p,q)$7 with $(p,q)$8 even are Kähler geometrically formal. By contrast, Hirzebruch surfaces $(p,q)$9 with $\mathcal A^{p,q}(M)$0 odd admit no Hermitian metric that is geometrically Aeppli formal (Sferruzza et al., 14 Jul 2025).

A broader nonexistence theorem excludes several classical surface classes from all the principal Hermitian formal categories. Ruled surfaces of genus $\mathcal A^{p,q}(M)$1, minimal class VII surfaces with $\mathcal A^{p,q}(M)$2, classical Enriques surfaces, and K3 surfaces admit no geometrically formal metric, no geometrically Dolbeault formal metric, and no geometrically Bott–Chern $\mathcal A^{p,q}(M)$3 formal metric (Sferruzza et al., 14 Jul 2025).

The reasons differ by class but follow the same pattern. For ruled surfaces of genus $\mathcal A^{p,q}(M)$4, Dolbeault formality is obstructed by the inequality

$\mathcal A^{p,q}(M)$5

which contradicts the Dolbeault bounds above. For K3 and classical Enriques surfaces, the second Betti number already exceeds the torus bound $\mathcal A^{p,q}(M)$6, so both classical and Hermitian geometric formality fail. For class VII surfaces with $\mathcal A^{p,q}(M)$7, the same torus-bound obstruction applies (Sferruzza et al., 14 Jul 2025).

These results are important because they separate Hermitian geometric formality from broader Kähler-type phenomena. K3 and Enriques surfaces satisfy strong Hodge-theoretic conditions, but they still fail the torus-type constraints required by geometric formality. This directly illustrates that neither the existence of canonical metrics nor the presence of a rich Hodge theory is sufficient.

4. Solvmanifolds, complex parallelisable manifolds, and Calabi–Eckmann manifolds

The solvable and parallelisable settings provide a second major testing ground. For compact complex nilpotent manifolds, Dolbeault formality fails unless the manifold is a complex torus. Since geometric Dolbeault formality implies Dolbeault formality, complex nilpotent manifolds are not geometrically Dolbeault formal unless they are tori (Sferruzza et al., 14 Jul 2025).

The Bott–Chern theory is even more restrictive for complex parallelisable manifolds. If a compact complex parallelisable manifold is not a torus, then it cannot be geometrically Bott–Chern formal. The mechanism is simple and rigid: such manifolds admit a global frame of holomorphic $\mathcal A^{p,q}(M)$8-forms, while geometric Bott–Chern formality forces every holomorphic $\mathcal A^{p,q}(M)$9-form to be $d=\partial+\bar\partial$0-closed; this implies that the underlying complex Lie algebra is abelian, hence the manifold is a complex torus (Sferruzza et al., 14 Jul 2025).

For complex solvable manifolds in Nakamura’s low-dimensional classification, several families admit non-zero triple ABC-Massey products. This implies that they are not weakly formal and therefore cannot be ABC-geometrically formal or geometrically Aeppli formal. The paper raises the further question whether every complex solvable manifold admits non-vanishing triple ABC-Massey products; an affirmative answer would exclude all non-torus complex solvmanifolds from the Bott–Chern/Aeppli formal categories (Sferruzza et al., 14 Jul 2025).

Calabi–Eckmann manifolds furnish the main non-Kähler positive examples. For the standard Calabi–Eckmann manifold $d=\partial+\bar\partial$1, the Bott–Chern theory is controlled by an explicit finite-dimensional model generated by $d=\partial+\bar\partial$2, with

$d=\partial+\bar\partial$3

The classification is sharp: $d=\partial+\bar\partial$4 is geometrically Bott–Chern formal if and only if its underlying diffeomorphism type is one of

$d=\partial+\bar\partial$5

equivalently $d=\partial+\bar\partial$6. In those cases the standard metric is in fact ABC-geometrically formal. Moreover, the same manifolds are exactly the Calabi–Eckmann manifolds that carry Bismut-flat SKT metrics (Sferruzza et al., 14 Jul 2025).

The Dolbeault classification is different. The standard metric on $d=\partial+\bar\partial$7 is geometrically Dolbeault formal if and only if $d=\partial+\bar\partial$8, hence if and only if the manifold is diffeomorphic to $d=\partial+\bar\partial$9. These are precisely the locally conformally Kähler Calabi–Eckmann manifolds (Sferruzza et al., 14 Jul 2025).

Taken together, these examples show that the Dolbeault and Bott–Chern notions are genuinely distinct. They also indicate that Hermitian geometric formality interacts closely with special Hermitian connections and non-Kähler metric classes, but does so in a sharply selective way.

5. Blow-ups, deformation-theoretic contrast, and nearby Hermitian classes

Blow-up behavior is largely hostile to Hermitian geometric formality. If $$H^{p,q}_{\bar\partial}(M)=\frac{\ker \bar\partial}{\operatorname{im}\bar\partial},\qquad H^{p,q}_{BC}(M)=\frac{\ker\partial\cap\ker\bar\partial}{\operatorname{im}\partial\bar\partial},\qquad H^{p,q}_{A}(M)=\frac{\ker\partial\bar\partial}{\operatorname{im}\partial+\operatorname{im}\bar\partial}.$$0 is a complex torus and $$H^{p,q}_{\bar\partial}(M)=\frac{\ker \bar\partial}{\operatorname{im}\bar\partial},\qquad H^{p,q}_{BC}(M)=\frac{\ker\partial\cap\ker\bar\partial}{\operatorname{im}\partial\bar\partial},\qquad H^{p,q}_{A}(M)=\frac{\ker\partial\bar\partial}{\operatorname{im}\partial+\operatorname{im}\bar\partial}.$$1 is the blow-up along any compact complex submanifold, then $$H^{p,q}_{\bar\partial}(M)=\frac{\ker \bar\partial}{\operatorname{im}\bar\partial},\qquad H^{p,q}_{BC}(M)=\frac{\ker\partial\cap\ker\bar\partial}{\operatorname{im}\partial\bar\partial},\qquad H^{p,q}_{A}(M)=\frac{\ker\partial\bar\partial}{\operatorname{im}\partial+\operatorname{im}\bar\partial}.$$2 is not geometrically formal, not geometrically Dolbeault formal, and not geometrically Bott–Chern $$H^{p,q}_{\bar\partial}(M)=\frac{\ker \bar\partial}{\operatorname{im}\bar\partial},\qquad H^{p,q}_{BC}(M)=\frac{\ker\partial\cap\ker\bar\partial}{\operatorname{im}\partial\bar\partial},\qquad H^{p,q}_{A}(M)=\frac{\ker\partial\bar\partial}{\operatorname{im}\partial+\operatorname{im}\bar\partial}.$$3 formal. The cohomological reason is that the exceptional divisor increases Betti numbers beyond the torus bounds required by geometric formality (Sferruzza et al., 14 Jul 2025).

A stronger metric-level statement holds in the Kähler category. If $$H^{p,q}_{\bar\partial}(M)=\frac{\ker \bar\partial}{\operatorname{im}\bar\partial},\qquad H^{p,q}_{BC}(M)=\frac{\ker\partial\cap\ker\bar\partial}{\operatorname{im}\partial\bar\partial},\qquad H^{p,q}_{A}(M)=\frac{\ker\partial\bar\partial}{\operatorname{im}\partial+\operatorname{im}\bar\partial}.$$4 is the blow-up of a compact Kähler manifold along a compact complex submanifold, and

$$H^{p,q}_{\bar\partial}(M)=\frac{\ker \bar\partial}{\operatorname{im}\bar\partial},\qquad H^{p,q}_{BC}(M)=\frac{\ker\partial\cap\ker\bar\partial}{\operatorname{im}\partial\bar\partial},\qquad H^{p,q}_{A}(M)=\frac{\ker\partial\bar\partial}{\operatorname{im}\partial+\operatorname{im}\bar\partial}.$$5

is the standard blow-up metric, then $$H^{p,q}_{\bar\partial}(M)=\frac{\ker \bar\partial}{\operatorname{im}\bar\partial},\qquad H^{p,q}_{BC}(M)=\frac{\ker\partial\cap\ker\bar\partial}{\operatorname{im}\partial\bar\partial},\qquad H^{p,q}_{A}(M)=\frac{\ker\partial\bar\partial}{\operatorname{im}\partial+\operatorname{im}\bar\partial}.$$6 is not geometrically formal. This extends the classical projective blow-up obstruction to the standard Kähler blow-up metric on arbitrary compact Kähler manifolds (Sferruzza et al., 14 Jul 2025).

These nonexistence results contrast sharply with the behavior of the class of “special” Hermitian metrics introduced by the condition

$$H^{p,q}_{\bar\partial}(M)=\frac{\ker \bar\partial}{\operatorname{im}\bar\partial},\qquad H^{p,q}_{BC}(M)=\frac{\ker\partial\cap\ker\bar\partial}{\operatorname{im}\partial\bar\partial},\qquad H^{p,q}_{A}(M)=\frac{\ker\partial\bar\partial}{\operatorname{im}\partial+\operatorname{im}\bar\partial}.$$7

Every Kähler metric is special, and every special metric is both SKT/pluriclosed and astheno-Kähler. The blow-up stability theorem states that if a Hermitian metric $$H^{p,q}_{\bar\partial}(M)=\frac{\ker \bar\partial}{\operatorname{im}\bar\partial},\qquad H^{p,q}_{BC}(M)=\frac{\ker\partial\cap\ker\bar\partial}{\operatorname{im}\partial\bar\partial},\qquad H^{p,q}_{A}(M)=\frac{\ker\partial\bar\partial}{\operatorname{im}\partial+\operatorname{im}\bar\partial}.$$8 has fundamental form $$H^{p,q}_{\bar\partial}(M)=\frac{\ker \bar\partial}{\operatorname{im}\bar\partial},\qquad H^{p,q}_{BC}(M)=\frac{\ker\partial\cap\ker\bar\partial}{\operatorname{im}\partial\bar\partial},\qquad H^{p,q}_{A}(M)=\frac{\ker\partial\bar\partial}{\operatorname{im}\partial+\operatorname{im}\bar\partial}.$$9 satisfying

$g$0

then after blowing up along any compact complex submanifold one obtains a Hermitian metric $g$1 with

$g$2

In particular, the blow-up of a manifold carrying a special Hermitian metric again admits a special Hermitian metric (Ciulică, 2024).

This contrast is conceptually important. Special metrics provide a nearby class with strong $g$3-closedness properties for wedge powers of the fundamental form, but the relevant paper does not define geometric formality and does not study wedge products of harmonic forms. What it does show is that some Kähler-like $g$4-constraints are birationally stable even when Hermitian geometric formality is not (Ciulică, 2024).

The same paper proves a first-order obstruction to deformation stability. If $g$5 is a smooth family of Hermitian metrics on a deformation $g$6 with

$g$7

then necessarily

$g$8

This is not a theorem about Hermitian geometric formality itself, but it shows that neighboring $g$9-type structures already impose nontrivial cohomological constraints on Kodaira–Spencer classes (Ciulică, 2024).

6. Curvature, second Chern Ricci rigidity, and conceptual position

Curvature provides the main positive mechanism for classical geometric formality inside the Hermitian setting. If $\Delta_{\bar\partial}$0 is compact Kähler and the curvature operator is nonnegative, then every harmonic de Rham form is parallel by the Gallot–Meyer theorem. Consequently, $\Delta_{\bar\partial}$1 is geometrically formal in the de Rham sense, and $\Delta_{\bar\partial}$2 is strongly formal. Because the Kähler metric supplies the usual Hodge bigrading, all Dolbeault, Bott–Chern, and Aeppli cohomologies inherit the corresponding formal structures. Compact Kähler flat manifolds, and therefore Kähler solvmanifolds, fall into this class (Sferruzza et al., 14 Jul 2025).

The relationship with second Chern Ricci-flatness is more indirect. The paper on vanishing second Chern Ricci curvature does not define Hermitian geometric formality, but it identifies a rigid curvature regime in which Bochner-type parallelism becomes plausible. On compact manifolds carrying a background metric with nonpositive real bisectional curvature or nonnegative Schwarz bisectional curvature, any Hermitian metric with

$\Delta_{\bar\partial}$3

must share the same Chern connection as the background metric; if the background is Kähler, Kähler-like, or balanced, the second Chern Ricci-flat metric inherits the same structure. Moreover, on complex tori every second Chern Ricci-flat metric is flat (Broder et al., 2023).

In complex dimension $\Delta_{\bar\partial}$4, the same source states that second Chern Ricci-flat metrics on compact complex surfaces are Ricci-flat Kähler metrics and thus exist only on complex tori, K3 surfaces, and Enriques surfaces. Combined with the nonexistence of Hermitian geometrically formal metrics on K3 and classical Enriques surfaces, this suggests that, among compact complex surfaces, the overlap between second Chern Ricci-flatness and Hermitian geometric formality is essentially concentrated on complex tori (Broder et al., 2023).

A further conceptual boundary is that $\Delta_{\bar\partial}$5-lemma or Sullivan formality do not imply Hermitian geometric formality. An explicit non-Kähler threefold obtained by resolving a quotient of the Iwasawa manifold satisfies the $\Delta_{\bar\partial}$6-lemma and is Sullivan formal and Dolbeault formal, but it is neither geometrically formal, nor geometrically Dolbeault formal, nor geometrically Bott–Chern formal, because Bott–Chern numbers exceed the torus bounds and nontrivial ABC-Massey products survive (Sferruzza et al., 14 Jul 2025).

Hermitian geometrically formal metrics therefore occupy a narrow region within complex geometry. They are compatible with Kähler geometry under strong curvature conditions, with a few non-Kähler model spaces such as low-dimensional Calabi–Eckmann manifolds, and with torus-like numerical constraints. At the same time, they are typically destroyed by blow-up, obstructed by excessive cohomology, and incompatible with many classical complex surfaces and non-abelian solvable geometries.