---
title: Nakano Positivity in Hermitian Bundles
url: https://www.emergentmind.com/topics/nakano-positivity
type: topic
---

# Nakano Positivity in Hermitian Bundles

Nakano positivity is a curvature positivity notion for Hermitian holomorphic vector bundles that is stronger than Griffiths positivity and is naturally tied to the Bochner–Kodaira–Nakano identity, optimal $L^2$ solvability of $\bar\partial$, direct-image curvature, and vanishing theorems. In current work on $L^2$ extension of top-degree $E$-valued forms, it appears as the correct curvature condition: Nakano semipositivity is sufficient, whereas Griffiths positivity is not [2508.01753].

## 1. Smooth definition and operator formulations

Let $X$ be a complex manifold of dimension $n$, let $E \to X$ be a holomorphic vector bundle of rank $r$, and let $h$ be a smooth Hermitian metric. In local holomorphic coordinates $(z^1,\dots,z^n)$ and a local holomorphic frame $(e_\alpha)$, the Chern curvature is written as
$$
i\Theta_{E,h}
=
\sum_{i,j,\alpha,\beta}
\Theta_{i\bar j\,\alpha\bar\beta}\,
dz^i\wedge d\bar z^j\otimes e_\alpha\otimes e_\beta^\star.
$$
Nakano semipositivity means that the induced Hermitian form on $T^{1,0}X\otimes E$ is nonnegative:
$$
\sum_{i,j,\alpha,\beta}
\Theta_{i\bar j\,\alpha\bar\beta}\,
u_{i\alpha}\,\overline{u_{j\beta}}
\ge 0
\qquad
\text{for all }u=(u_{i\alpha})\in T^{1,0}X\otimes E.
$$
Strict Nakano positivity is the corresponding positive-definiteness condition [2004.05798].

A standard operator formulation uses the Lefschetz adjoint $\Lambda_\omega$ for a Kähler form $\omega$. On a Kähler manifold, smooth Nakano semipositivity is equivalent to
$$
[i\Theta_{E,h},\Lambda_\omega]\ge 0
$$
as a Hermitian endomorphism on $E$-valued forms, in particular on $A^{n,q}(X,E)$ for all $q\ge 1$ [2407.11412]. This equivalence is the analytic bridge between curvature and $L^2$ theory.

Dual Nakano positivity is defined by reversing the tensor indices in the quadratic form: $E$ is dual Nakano semipositive if
$$
\sum_{i,j,\alpha,\beta}
R_{i\bar j\,\alpha\bar\beta}\,
v^j_\alpha\,\overline{v^i_\beta}\ge 0
$$
for every tensor $v\in T^{1,0}X\otimes E$; equivalently, $E^\ast$ is Nakano seminegative [2301.03950]. For line bundles, Nakano, Griffiths, and dual-Nakano positivity coincide [1006.1465].

## 2. Relation to Griffiths positivity and other positivity notions

Griffiths semipositivity is the weaker condition obtained by testing curvature only on decomposable tensors:
$$
\sum_{i,j,\alpha,\beta}
\Theta_{i\bar j\,\alpha\bar\beta}\,
v^i\overline{v^j}\,\xi^\alpha\overline{\xi^\beta}\ge 0
\qquad
(v\in T^{1,0}_xX,\ \xi\in E_x).
$$
Nakano positivity implies Griffiths positivity, but the converse fails in general [2508.01753].

The failure of the converse is structurally important. For the dual metric $h^\ast$, Griffiths positivity is dual to Griffiths negativity, but there is no simple duality for $k$-positivity when $k\ge 2$ [2508.01753]. This asymmetry is one reason Nakano positivity is the curvature notion that enters directly into estimates on $E$-valued $(n,q)$-forms.

Several later positivity notions interpolate between decomposable positivity and Nakano-type conditions. “Positivity of Schur forms for strongly decomposably positive vector bundles” defines strongly decomposable positivity of type I and type II, both of which strictly strengthen decomposable positivity and contain Nakano or dual Nakano positivity as special cases. In that framework, Schur forms are weakly positive for type I and positive for type II; in particular, for Nakano-positive or dual Nakano-positive bundles all Schur forms are positive [2301.03950].

A distinct amplification phenomenon appears for determinant twists. If $E$ is ample on a compact Kähler manifold, then $S^kE\otimes \det E$ is both Nakano-positive and dual-Nakano-positive for every $k\ge 0$; more generally, if $(E,h)$ is Griffiths positive, then the induced metric on $S^kE\otimes \det E$ has both properties [1006.1465]. This shows that determinant twists can convert weaker positivity into Nakano-type positivity with stronger cohomological consequences.

## 3. Bochner–Kodaira–Nakano theory, optimal $L^2$ estimates, and extension

The analytic force of Nakano positivity comes from the Bochner–Kodaira–Nakano identity. For a weight $\phi$ and a Kähler metric $g$ with Kähler form $\omega$, the identity on $E$-valued $(n,q)$-forms $u$ has the form
$$
\|\bar\partial u\|^2+\|\bar\partial^\ast u\|^2
=
\big([\,i\Theta_h(E)+i\partial\bar\partial\phi,\Lambda\,]u,u\big)
+\text{boundary terms},
$$
with boundary terms vanishing or controllable under standard completeness or boundary hypotheses [2508.01753]. Combined with Demailly’s positivity calculus, this yields Hörmander–Skoda–Demailly $L^2$ estimates for $\bar\partial$.

The converse direction is equally important. “Positivity of holomorphic vector bundles in terms of $L^p$-conditions of $\bar\partial$” proves that optimal $L^2$ solvability characterizes Nakano semipositivity: if the optimal $L^2$ estimate for $\bar\partial$ holds with sharp constant $1$, then the bundle is Nakano semipositive; conversely, on complete Kähler manifolds, Nakano semipositivity gives that optimal estimate [2001.01762]. In the same paper, several weaker $L^p$ estimate and extension conditions are shown to imply Griffiths semipositivity rather than Nakano semipositivity [2001.01762].

The sharp extension theory of top-degree $E$-valued forms makes the distinction between Griffiths and Nakano positivity concrete. Let $X$ be essentially Stein with Kähler metric $g$, let $Z\subset X$ be a smooth hypersurface cut out by $T\in H^0(X,L_Z)$, and let $\lambda$ be a smooth metric on $L_Z$ such that $\sup_X |T|^2e^{-\lambda}\le 1$. If $(E,h)$ satisfies
$$
\Theta_h(E)\ge_{\mathrm{Nak}} t\,\delta\,\Theta_\lambda(L_Z)\otimes \mathrm{Id}_E
\qquad \text{for all } t\in[0,1],
$$
then every $f\in H^0(Z,K_Z\otimes E|_Z)$ with finite $L^2$ norm extends to $F\in H^0(X,K_X\otimes L_Z\otimes E)$ with $F|_Z=f\wedge dT$ and
$$
\int_X |F|_h^2 \le \pi\frac{1+\delta}{\delta}\int_Z |f|_h^2.
$$
The constant is sharp, and in the flat case it becomes $\pi$ [2508.01753].

The same paper gives an explicit counterexample showing that Griffiths positivity is not enough for such extension. On projective space, the universal quotient bundle $Q$ is Griffiths nonnegative but not Nakano-nonnegative; for $r=3$, the tangent bundle $T\mathbf P(V)$ is Griffiths positive but not Nakano positive, and the canonical restriction map needed for extension fails to be surjective [2508.01753]. This makes precise the statement that the commutator $[i\Theta_h(E),\Lambda]$ is controlled by Nakano positivity, not by Griffiths positivity.

## 4. Singular Hermitian metrics and $L^2$-based Nakano positivity

For singular Hermitian metrics on vector bundles, the smooth curvature tensor is often unavailable. A singular Hermitian metric on $E$ is a measurable assignment of positive semidefinite Hermitian forms on the fibers with $0<\det h<+\infty$ almost everywhere. In rank at least $2$, the Chern curvature need not exist as a current with measure coefficients, so positivity notions must avoid direct reliance on $\Theta_{E,h}$ [2004.05798].

A robust replacement is an $L^2$-solvability definition. Following Deng–Ning–Wang–Zhou and Inayama, singular Nakano semipositivity is defined by the requirement that on every Stein trivializing chart, for every Kähler form $\omega$, every smooth weight $\psi$ with $\theta+i\partial\bar\partial\psi\ge 0$, and every $\bar\partial$-closed $E$-valued $(n,q)$-form $g$ with finite curvature energy, there exists $u$ solving $\bar\partial u=g$ and satisfying the optimal estimate
$$
\int |u|^2_{he^{-\psi}}
\le
\int \big\langle (B_{\omega,\psi})^{-1}g,g\big\rangle_{he^{-\psi}},
\qquad
B_{\omega,\psi}=[(\,i\partial\bar\partial\psi+\theta\,)\otimes \mathrm{Id}_E,\Lambda_\omega].
$$
In the smooth case this is equivalent to classical Nakano semipositivity [2004.05798].

An approximation-theoretic version was developed in “Nakano positivity of singular Hermitian metrics: Approximations and applications.” There, a singular metric is called $\theta$-Nakano semipositive in the sense of approximations if it is the monotone limit of smooth metrics on Zariski open sets whose Nakano negativity is controlled by error terms tending to zero almost everywhere. The paper proves that this approximation property implies $\theta$-Nakano semipositivity in the global $L^2$ sense, and derives coherence of the sheaf of locally $L^2$ holomorphic sections and Nadel–Nakano type vanishing theorems [2402.06883].

Stability under monotone limits is another central feature. “Multiplier Submodule Sheaves and a problem of Lempert” proves that increasing limits of Nakano semipositive singular metrics remain Nakano semipositive, thereby answering affirmatively Lempert’s question on preservation of Nakano semipositivity under increasing limits. The same paper establishes strong openness and stability properties for the multiplier submodule sheaves associated with Nakano semipositive singular metrics [2111.13452].

The theory has also been extended from manifolds to complex spaces. “Griffiths and Nakano positivity and Nakano-Nadel vanishing theorems for singular Hermitian metrics on complex spaces” defines singular Nakano positivity on reduced complex spaces by the same $L^2$-estimate paradigm on the regular locus and proves $L^2$-Dolbeault fine resolutions, cohomological isomorphisms, and Nakano–Nadel vanishing theorems on weakly pseudoconvex complex spaces [2606.16275].

## 5. Direct images, Bergman spaces, and Hilbert bundles

Nakano positivity plays a distinguished role in direct-image geometry. In the smooth case, Berndtsson proved that for a holomorphic submersion and a smooth semipositive line bundle metric, the canonical $L^2$ metric on $f_\ast(K_{X/Y}+L)$ has Nakano-positive curvature; this result is cited as the starting point for later singular theories [2407.11412].

A higher-rank analogue appears in “Positivity and $L^2$ Extension.” Let $Y$ be Stein, let $X\subset Y$ be pseudoconvex, let $E$ be a holomorphic vector bundle over $Y$, and let $h(t)$ be a family of smooth Hermitian metrics. For the Hilbert bundle whose fiber at $t$ is the Bergman space
$$
H_t
=
\left\{
f\in H^0(X,K_X\otimes E)\,;\,
\int_X \langle f\wedge \bar f,h(t)\rangle<\infty
\right\},
$$
the paper proves: if $(E,h(t))$ is Nakano nonnegative on $X$ for each $t$ and $(E,h)$ is $k$-positive on $X\times B$, then the Hilbert bundle is $k$-positive in the sense that the Chern curvature of its $L^2$ metric is $k$-positive [2508.01753]. This result is a vector-bundle analogue of Berndtsson’s positivity theorem and is used in a Berndtsson–Lempert proof of sharp $L^2$ extension.

Twisted versions weaken the curvature hypotheses while retaining Nakano positivity of the direct image. “A Twisted Complex Brunn-Minkowski Theorem” introduces modified curvature operators $\Theta_\delta(h)$ and $\Xi_{\delta,\eta}(h)$ and proves Nakano semipositivity of the direct-image Hilbert bundle under twisted Griffiths and Nakano assumptions, allowing some curvature negativity in the fiber directions when compensated by a suitable auxiliary function $\eta$ [2111.03143]. “Twisted Nakano-Positivity of Fields of Hilbert Spaces” extends the same method to fields of Hilbert spaces over possibly unbounded Stein manifolds by an exhaustion argument [2111.05832].

In singular settings, the Narasimhan–Simha metric becomes the canonical object. For a proper Kähler fibration $f:X\to Y$ and a singular Hermitian line bundle $(L,h)$ with semipositive curvature current, the direct image
$$
E=f_\ast\big(\mathcal O_X(K_{X/Y}+L)\otimes \mathcal I(h)\big)
$$
carries the canonical $L^2$ metric
$$
\|s(y)\|_{NS}^2
=
\int_{X_y} c_n\, s\wedge \bar s\, e^{-\varphi_h}.
$$
“Singular Nakano positivity of direct image sheaves of adjoint bundles” proves that this metric is singular Nakano semipositive in the precise sense of optimal $L^2$ solvability, locally in general and globally when $X$ is Kähler [2407.11412].

Related direct-image results include Nakano semipositivity for adjoint line bundles with mild singularities over bounded pseudoconvex polydisks [2108.13715], and a dual statement in the smooth no-deformation case: “Dual Nakano positivity and singular Nakano positivity of direct image sheaves” proves that if the Kodaira–Spencer forms vanish, then the canonical $L^2$ metric on $f_\ast(K_{X/Y}\otimes L)$ is dual Nakano semipositive, while in the singular case the canonical metric is locally Nakano semipositive in the $L^2$ sense [2302.09398].

## 6. Vanishing theorems and further geometric consequences

The cohomological implications of Nakano positivity are classical and remain central. If $E$ is ample on a compact Kähler manifold, then for every $k\ge 0$ the bundle $S^kE\otimes \det E$ is both Nakano-positive and dual-Nakano-positive, and therefore
$$
H^{n,q}(X,S^kE\otimes \det E)=H^{q,n}(X,S^kE\otimes \det E)=0
\qquad
(q\ge 1).
$$
The same paper proves the corresponding statement for the naturally induced metric when $(E,h)$ is Griffiths positive [1006.1465].

For singular metrics, “Nakano positivity of singular Hermitian metrics and vanishing theorems of Demailly-Nadel-Nakano type” establishes $L^2$ estimates for Nakano positive singular Hermitian metrics and derives vanishing theorems generalizing both Nakano-type and Demailly–Nadel-type vanishing theorems [2004.05798]. On weakly pseudoconvex complex spaces, the complex-space theory yields
$$
H^q(X,GR(E,h))=0
\qquad
(q\ge 1)
$$
for Nakano-positive singular metrics in the $L^2$ sense [2606.16275].

Direct-image positivity also feeds vanishing. Under the assumptions of the singular direct-image theorem, if $\sqrt{-1}\Theta_{L,h}\ge f^\ast\omega_Y$ for a Kähler form $\omega_Y$ on $Y$, then
$$
H^q\big(Y,f_\ast(\mathcal O_X(K_X+L)\otimes \mathcal I(h))\big)=0
\qquad
(q>0),
$$
a Kollár–Ohsawa type vanishing theorem obtained from optimal $L^2$ estimates [2407.11412].

Finally, Nakano positivity has strong consequences beyond cohomology. For (dual) Nakano positive bundles, Schur forms are positive [2301.03950]. In direct-image geometry, strict pseudoconvexity can upgrade semipositivity to strict Nakano positivity: direct image bundles associated to strictly pseudoconvex families of bounded domains are shown to be strictly positive in the sense of Nakano [2403.19152], and for strictly pseudoconvex circular or Reinhardt families strict Nakano positivity persists even when the weight is only plurisubharmonic [2301.00160]. These results place Nakano positivity at the center of a common analytic framework for extension, convexity, Bergman-kernel variation, and vanishing.

Source: https://www.emergentmind.com/topics/nakano-positivity