---
title: Projectively Flat Bundles
url: https://www.emergentmind.com/topics/projectively-flat-bundles
type: topic
---

# Projectively Flat Bundles

Searching arXiv for recent and foundational papers on projectively flat bundles and closely related correspondences.
{"query":"all:projectively flat bundles Higgs bundles 1911.03593 OR projectively flat vector bundles torus noncommutative torus 2205.11235 OR curved Riemann-Hilbert projectively flat 2411.19595", "max_results": 10}
I found several directly relevant arXiv papers, including the 2019 correspondence paper on projectively flat bundles and semistable Higgs bundles, the 2022 work relating projectively flat bundles on tori to noncommutative tori and generalized theta functions, and the 2024 curved Riemann–Hilbert correspondence for projectively flat connections.
Searching for additional arXiv papers on classification, torus geometry, and projectively flat structures beyond the Higgs-bundle correspondence.
{"query":"all:\"projectively flat\" bundle torus factor automorphy classification foliation log smooth", "max_results": 10}
Additional relevant arXiv work includes projectively flat bundles on higher-dimensional complex tori and their interpretation via factors of automorphy, as well as later classification-oriented papers on numerically projectively flat tangent or logarithmic tangent bundles and foliations.
A projectively flat bundle is the natural scalar-curvature analogue of a flat bundle: for a complex vector bundle \(E\) of rank \(r\) over a complex manifold \(X\), projective flatness means that \(E\) admits a connection whose induced connection on the associated \(PGL(r,\mathbb C)\)-bundle is flat, equivalently that its curvature is everywhere scalar, \(\sqrt{-1}F_D=\alpha\,\mathrm{Id}_E\) for some real \((1,1)\)-form \(\alpha\) [1911.03593]. In this form, projective flatness isolates the trace part of curvature and forces the trace-free curvature to vanish. The notion sits at the intersection of complex differential geometry, Higgs-bundle theory, torus geometry, representation theory, and twisted or logarithmic variants of flatness.

## 1. Definition and basic curvature model

For a connection \(D\) on a rank-\(r\) complex vector bundle \(E\), the defining condition of projective flatness is
\[
\sqrt{-1}F_D=\alpha\otimes \mathrm{Id}_E,
\]
or, in the smooth formulation,
\[
\nabla^2=h\otimes \mathbbm 1_E
\]
for some \(2\)-form \(h\) [1911.03593] [2411.19595]. The two formulations express the same geometric principle: the induced projective connection is flat, while the remaining curvature is central. When \(c_1(E)\) has a real \((1,1)\)-representative, \(\alpha\) may be chosen in that class [1911.03593].

This condition is weaker than ordinary flatness and stronger than a general Hermitian–Einstein-type condition. In the terminology of the correspondence paper, the only curvature left is the trace part; the trace-free curvature vanishes [1911.03593]. On complex tori and elliptic curves this often appears as constant central curvature for the Chern–Bott connection, while in smooth topology it appears as projective monodromy rather than honest linear monodromy [2205.11235] [2411.19595].

The definition has several equivalent incarnations. On a complex torus, projectively flat Hermitian bundles admit factors of automorphy of the standard form
\[
j(\gamma,z)=U(\gamma)\exp\!\left(\frac1{2r}R(z,\gamma)+\frac1{4r}R(\gamma,\gamma)\right),
\]
with \(U:\Gamma\to U(r)\) a semirepresentation and \(R\) a Hermitian form satisfying \(\operatorname{Im}R(\Gamma,\Gamma)\subset \pi\mathbb Z\) [1705.04007]. On a manifold with basepoint, projectively flat connections lead not to honest representations of \(\pi_1\), but to projective representations whose multiplier records the scalar curvature defect [2411.19595].

## 2. Correspondence with Higgs bundles

The principal modern structural result is an extension of the Corlette–Donaldson–Hitchin–Simpson correspondence from flat bundles to projectively flat bundles. On a compact Kähler manifold \((X,\omega)\), the category of \(\omega\)-semi-stable (resp. \(\omega\)-poly-stable) Higgs bundles \((E,\bar\partial_E,\phi)\) satisfying
\[
\bar\partial_E\phi=0,\qquad \phi\wedge\phi=0,
\]
and the vanishing discriminant condition
\[
\bigl(2r\,c_2(E)-(r-1)c_1^2(E)\bigr)\cdot[\omega]^{n-2}=0
\]
is equivalent to the category of semisimple (resp. simple) projectively flat bundles \((E,D)\) with
\[
\sqrt{-1}F_D=\alpha\,\mathrm{Id}_E
\]
[1911.03593].

The discriminant equality is the Bogomolov-type equality case that forces the Hermitian–Einstein or Hitchin–Simpson curvature to be as special as possible. In the projectively flat situation, it is exactly the bridge between stability on the Higgs side and scalar curvature on the connection side [1911.03593]. Analytically, a simple projectively flat bundle admits a unique Hermitian metric \(H\) solving the projectively Hermitian–Einstein equation
\[
\sqrt{-1}\,\Lambda_\omega G_H=\lambda\,\mathrm{Id}_E,
\]
and on an astheno-Kähler manifold the trace-free part actually vanishes, producing a harmonic metric and hence a Higgs structure [1911.03593].

The correspondence is not confined to the Kähler case. It extends to compact Hermitian manifolds satisfying
\[
\partial\bar\partial\omega^{n-1}=0,\qquad \partial\bar\partial\omega^{n-2}=0,
\]
together with a global \(\partial\bar\partial\)-lemma-type condition [1911.03593]. Under these hypotheses the paper proves a one-to-one correspondence between moduli spaces,
\[
\mathcal{CDol}(E)\cong \mathcal{CDR}(E,\alpha),
\]
and establishes the filtration statement needed to pass from stable and simple objects to semistable and semisimple ones [1911.03593].

An important application is a vanishing theorem for characteristic classes. For a projectively flat bundle \((E,D)\), the odd real classes
\[
v_{2j+1}(E,D)\in H^{2j+1}_{\mathrm{dR}}(X,\mathbb R)
\]
constructed in the style of Kamber–Tondeur and Bismut–Lott vanish on astheno-Kähler manifolds:
\[
v_{2j+1}(E,D)=0,\qquad j\ge 1
\]
[1911.03593].

## 3. Complex tori, factors of automorphy, and theta-function realizations

Higher-dimensional complex tori provide an explicit model theory. For a complex torus
\[
T^{2n}_{J=T}=\mathbb C^n/2\pi(\mathbb Z^n+T\mathbb Z^n),
\]
a holomorphic bundle \(E_{(r,A,\mu,U)}\) is built from data \((r,A,\mu,U)\), and its mirror object is the affine Lagrangian multisection
\[
L(r,A,p)=\left\{(\check x,\check y)\in \mathbb R^{2n}\mid \check y=\frac1r A\check x+\frac1r p\right\}
\]
with unitary local system \(L(r,A,p,q)\) [1705.04007]. The holomorphicity condition is
\[
AT=(AT)^t,
\]
and the curvature is computed as
\[
\Omega_{(r,A,\mu,U)}=\frac{i}{2\pi r}\,dx\,A\,dy\cdot I_r,
\]
so the bundle is projectively flat because the curvature is scalar-valued [1705.04007].

The same paper identifies these bundles with classical projectively flat bundles defined by factors of automorphy. The associated Hermitian form is
\[
R=\frac{i r'}{2\pi r}(T-\overline T)^{-1}A,
\]
and Theorem 3.8 gives an explicit exponential gauge transformation identifying the two constructions [1705.04007]. This makes projectively flat bundles the holomorphic-side avatars of affine Lagrangian submanifolds with unitary local systems.

The torus picture also supports triangulated and mirror-symmetric phenomena. If an exact triangle has cone again isomorphic to a simple projectively flat bundle, then for
\[
\alpha:=\frac{A}{r}-\frac{B}{s}
\]
one must have
\[
\operatorname{rank}\alpha=1,
\]
which on the mirror side becomes
\[
\operatorname{codim}\bigl(L(r,A,\check p)\cap L(s,B,\check p)\bigr)=1
\]
[1705.04007]. Thus exact triangles are controlled by codimension-one intersections of affine Lagrangian graphs.

On a complex \(2\)-torus, projectively flat bundles also appear as Hermitian–Einstein bundles encoded by rational noncommutative tori. For coprime positive integers \(r,q\), the rank-\(r\), degree-\(q\) bundle \(E_{r,q}\to \mathbb C/\Lambda\) has canonical connection of constant curvature, and for \(\theta=q/r\) the construction from an irreducible finite-dimensional representation of \(A_\theta\) yields
\[
\nabla^2=\theta\,\omega\,1_H
\]
up to normalization [2205.11235]. The commutator relation
\[
2\pi i\,[Q,P]=\theta\,1
\]
is the operator-theoretic form of the central-curvature condition [2205.11235].

In Matsushima’s framework, holomorphic sections of \(E_{r,q}\) become vector-valued theta functions, and the section space carries commuting actions of \(A_\theta\) and the dual torus algebra \(A_{1/\theta}\). The space \(\Gamma(E_{r,q})\) is an \(A_\theta\)-\(A_{1/\theta}\) bimodule, and the geometry is exchanged by the Fourier–Mukai–Nahm duality
\[
E_{r,q}\longleftrightarrow E_{q,r}
\]
[2205.11235]. In this setting projectively flat bundles are the geometric carrier of Morita equivalence.

## 4. Cohomological classification and projective monodromy

On a pro-torus \(\mathbb T\), projectively flat bundles admit a cohomological classification. Isomorphism classes of projectively flat rank-\(q\) complex vector bundles are classified by the first Chern class
\[
[\mathcal E]\longmapsto c_1(\mathcal E)\in H^2(\mathbb T,\mathbb Z),
\]
while flat \(q\times q\) matrix bundles are classified by the obstruction class
\[
[\mathcal A]\longmapsto \beta_q(\mathcal A)\in H^2(\mathbb T,\mu_q)
\]
coming from the connecting map
\[
H^1(\mathbb T,PGL(q,\mathcal C_{\mathbb T})) \to H^2(\mathbb T,\mu_q)
\]
[2509.10812]. The vector-bundle and matrix-bundle pictures are related by
\[
\mathcal A\cong \operatorname{End}(\mathcal E)\cong \mathcal E\otimes \mathcal E^*.
\]

A key point of this classification is the contrast with ordinary flat complex bundles on tori. Because \(U(q)\) is connected, flat complex vector bundles on a torus are trivial as bundles; projectively flat bundles are the next nontrivial level, and the first Chern class becomes the complete invariant [2509.10812]. The same mechanism recovers the bundle-theoretic description of rational noncommutative tori.

The representation-theoretic analogue is a projectively flat Riemann–Hilbert correspondence. For a connected smooth manifold \(M\), the category \(\mathrm{PF}_0(M)\) of projectively flat vector bundles is equivalent not to all projective representations of \(\pi_1(M;x_0)\), but to the subcategory
\[
\mathrm{PRep}^0_{H^2_B(\pi_1(M),\mathbb C^*)}(\pi_1(M))
\]
whose projective multiplier class lies in the subgroup \(H^2_B(\pi_1(M),\mathbb C^*)\) [2411.19595]. The obstruction criterion is the composite
\[
H^2(\pi_1(M),\mathbb C^*) \longrightarrow H^2(M,\mathbb C^*) \longrightarrow H^3(M,\mathbb Z),
\]
and a projective representation comes from a projectively flat bundle if and only if its class maps to \(0\) [2411.19595].

The torus example in that paper is especially concrete: if \(A,B\in GL_n(\mathbb C)\) satisfy
\[
AB=-BA,
\]
then the resulting projective representation of \(\pi_1(\mathbb T^2)\cong \mathbb Z^2\) is realized by a projectively flat connection
\[
\nabla=d-i\pi y\,dx,\qquad h=i\pi\,dx\wedge dy
\]
[2411.19595]. This exhibits projective monodromy as the holonomy shadow of central curvature.

## 5. Numerical, logarithmic, and foliation-theoretic extensions

Projectively flat bundle techniques extend to tangent bundles, logarithmic tangent bundles, and foliation tangent bundles. For a regular codimension-\(1\) foliation \(\mathcal F\subset T_X\) on a smooth complex projective manifold of dimension at least \(4\), the condition that the normalized bundle
\[
S^{n-1}\mathcal F\otimes \det(\mathcal F)
\]
is numerically flat forces a rigid classification: either \(X\) carries a \(\mathbb P\)-bundle structure over a finite étale quotient of an abelian variety and \(\mathcal F\) induces a flat Ehresmann connection; or, after a finite étale cover, the foliation is linear on an abelian variety; or it is induced by an abelian scheme over a smooth complete curve of genus at least \(2\); or \(K_X\) is ample and \(\kappa(K_{\mathcal F})=\dim X\) [2401.03311]. The same paper proves the dichotomy that for any regular codimension-\(1\) foliation, either the normal bundle \(N\) is pseudo-effective or the conormal bundle \(N^*\) is nef [2401.03311].

In the logarithmic setting, for a reduced log smooth pair \((X,D)\), the normalized logarithmic tangent bundle
\[
S^nT_X(-\log D)\otimes \mathcal O_X\big(-(K_X+D)\big)
\]
being numerically flat implies, after a finite cover and a birational morphism given by blowing up finitely many points away from the boundary, that one reaches either a log smooth pair with numerically flat logarithmic tangent bundle or the model
\[
(Z,B)\cong (\mathbb P^n,\mathbb P^{n-1})
\]
[2112.05449]. In the first case the structure is already understood: such pairs are toric fiber bundles over finite étale quotients of abelian varieties [2112.05449].

These results use projective-flatness identities as numerical constraints on Chern classes and as input to semistability and flatness arguments. They also show that projectively flat behavior persists well beyond the setting of ordinary vector bundles, governing integrable subbundles and logarithmic geometries with controlled singularities.

## 6. Analytic variants and terminological distinctions

On non-compact Gauduchon manifolds, projectively flat bundles are studied through Hermitian-Poisson metrics. For a projectively flat bundle \((E,D)\) with
\[
\sqrt{-1}F_D=\alpha\otimes \mathrm{Id}_E,
\]
the basic Hermitian decomposition
\[
D=D_H+\psi_H
\]
leads to the curvature-like term
\[
G_H=(D_H'')^2,
\]
and a Hermitian metric \(H\) is Hermitian-Poisson if
\[
\sqrt{-1}\Lambda_\omega G_H=\lambda\cdot \mathrm{Id}_E
\]
[2507.10973]. On a non-compact Gauduchon manifold satisfying the stated finite-volume, exhaustion, \(L^1\)-to-\(L^\infty\), and \(L^2\)-boundary assumptions, a simple projectively flat bundle admits such a metric, while conversely, in the balanced case, existence of a Hermitian-Poisson metric with \(\psi_H\in L^2\) forces semi-simplicity [2507.10973]. The proof uses a perturbed equation and a heat flow for Hermitian metrics.

A persistent source of confusion is the distinction between projectively flat bundles and projectively induced Kähler metrics. A Kähler metric is projectively induced if it admits a holomorphic and isometric immersion into \((\mathbb{CP}^N,g_{FS})\), finite or infinite dimensional [1912.05223]. This is a different notion. The non-projective-inducedness of Calabi’s Ricci-flat metrics on holomorphic line bundles over compact Kähler–Einstein manifolds, and of every positive multiple of the Eguchi–Hanson metric, concerns projective embeddings of metrics rather than projective flatness of connections [1912.05223]. The two notions meet historically through curvature and moduli questions, but they are not interchangeable.

Across these settings, projectively flat bundles recur as objects whose curvature is central, whose projective monodromy is flat, and whose residual geometry can be read through Higgs fields, factors of automorphy, noncommutative torus actions, obstruction classes, or numerical-flatness conditions. That combination of analytic rigidity and categorical flexibility is what makes them a unifying class in complex and differential geometry.

Source: https://www.emergentmind.com/topics/projectively-flat-bundles