This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

The Spectral base and quotients of bounded symmetric domains by cocompact lattices

Siqi He Siqi He, Institute of Mathematics, Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing, 100190, China sqhe@amss.ac.cn Jie Liu Jie Liu, Institute of Mathematics, Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing, 100190, China jliu@amss.ac.cn  and  Ngaiming Mok Ngaiming Mok, Department of Mathematics, The University of Hong Kong, Pokfulam Road, Hong Kong nmok@hku.hk
Abstract.

In this article, we explore Higgs bundles on a projective manifold XX, focusing on their spectral bases —– a concept introduced by T. Chen and B. Ngô. The spectral base is a specific closed subscheme within the space of symmetric differentials. We observe that if the spectral base vanishes, then any reductive representation ρ:π1(X)GLr()\rho:\pi_{1}(X)\to\text{GL}_{r}(\mathbb{C}) is both rigid and integral. Additionally, we prove that for X=Ω/ΓX=\Omega/\Gamma, a quotient of a bounded symmetric domain Ω\Omega of rank at least 22 by a torsion-free cocompact irreducible lattice Γ\Gamma, the spectral base indeed vanishes, which generalizes a result of B. Klingler.

1. Introduction

Let XX be a projective manifold and denote by ΩX1\Omega_{X}^{1} its cotangent bundle. A symmetric differential is a holomorphic section of SymiΩX1\mathrm{Sym}^{i}\Omega_{X}^{1} for some positive integer ii. According to Hodge theory, the cotangent bundle ΩX1\Omega_{X}^{1} admits non-zero holomorphic sections if and only if the abelianization of π1(X)\pi_{1}(X) is infinite. However, the relationship between symmetric differentials and fundamental groups remains mysterious, and this question has been introduced by F. Severi [Sev59] and H. Esnault in various contexts.

Question 1.1 (Severi, Esnault).

What is the relationship between the symmetric differentials and the fundamental group?

The question mentioned above has been approached from different perspectives in [Kli13] and [BKT13], where the main tool is the non-abelian Hodge correspondence, which identifies the moduli space of Higgs bundles and character varieties. In this article, we attempt to gain insight into this question by adopting a novel approach proposed by T. Chen and B. Ngô in [CN20].

The moduli space of Higgs bundles has been extensively utilized to investigate the topology and geometry of character varieties [Sim91, Sim92, Sim94a, Sim97]. For a fixed rank rr, consider the moduli stack Higgsstack,r\mathcal{M}_{\mathrm{Higgs}}^{\textup{stack}{\color[rgb]{0,0,0}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@gray@stroke{0}\pgfsys@color@gray@fill{0},r}} of Higgs bundles of rank rr. The Hitchin morphism, introduced by Hitchin [Hit87a, Hit92], plays a crucial role in understanding the moduli space. By taking invariant polynomials, the Hitchin morphism is a map

𝗁X:Higgsstack,r𝒜Xr:=i=1rH0(X,SymiΩX1).\mathsf{h}_{X}:\mathcal{M}_{\mathrm{Higgs}}^{\textup{stack}{\color[rgb]{0,0,0}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@gray@stroke{0}\pgfsys@color@gray@fill{0},r}}\rightarrow\mathcal{A}_{X}^{r}:=\bigoplus_{i=1}^{r}H^{0}(X,\mathrm{Sym}^{i}\Omega_{X}^{1}).

Moreover, the affine space 𝒜Xr\mathcal{A}_{X}^{r} is called the Hitchin base.

When dimX2\dim\;X\geq 2, a Higgs bundle (,φ)(\mathscr{E},\varphi) must satisfy an extra integrability condition φφ=0\varphi\wedge\varphi=0, which makes the Hitchin morphism not surjective in general. T. Chen and B. Ngô introduced in [CN20] the spectral base 𝒮Xr\mathcal{S}_{X}^{r} as a closed subscheme of the Hitchin base 𝒜Xr\mathcal{A}_{X}^{r}. They demonstrated that the integrability condition leads to the Hitchin morphism 𝗁X\mathsf{h}_{X} factoring through the natural inclusion ιX:𝒮Xr𝒜Xr\iota_{X}:\mathcal{S}_{X}^{r}\rightarrow\mathcal{A}_{X}^{r}.

We denote by GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} the GLr()\mathrm{GL}_{r}(\mathbb{C}) character variety. Recall that a representation [ρ]GLr()[\rho]\in\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} is called rigid if it is an isolated point, and GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} is termed rigid if every representation [ρ]GLr()[\rho]\in\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} is isolated. In particular, since the character variety is a complex affine variety, the character variety is rigid if and only if it is zero-dimensional. Rigid representations are of particular interest. It has been shown in [Sim92] that rigid representations are \mathbb{C}-variations of Hodge structures (\mathbb{C}-VHS for short). Furthermore, it is conjectured that rigid representations (local systems) originate from geometric sources [Sim91].

A representation ρ:π1(X)GLr()\rho:\pi_{1}(X)\rightarrow\mathrm{GL}_{r}(\mathbb{C}) is called integral if it is conjugate to a representation π1(X)GLr(𝒪K)\pi_{1}(X)\rightarrow\mathrm{GL}_{r}(\mathcal{O}_{K}), where KK is a number field and 𝒪K\mathcal{O}_{K} is the ring of integers of KK. The motivation to consider rigid representations comes from Simpson’s integrality conjecture, which predicts that any rigid representation is integral. This conjecture has been confirmed by H. Esnault and M. Groechenig for cohomological rigid local systems (see also Remark 4.8).

Using the spectral base, we obtain the following result, which generalizes [Sim92, Ara02, Kli13], see also [Zuo96, Kat97, Eys04, CDY22] for various generalizations.

Theorem 1.2.

Let XX be a projective manifold such that 𝒮Xr=0\mathcal{S}^{r}_{X}=0 for some r1r\geq 1. Then the following statements hold:

  1. (1)

    Any reductive representation ρ:π1(X)GLr()\rho:\pi_{1}(X)\to\mathrm{GL}_{r}(\mathbb{C}) is rigid and integral. Moreover, it is a complex direct factor of a \mathbb{Z}-variation of Hodge structures.

  2. (2)

    Let FF be a non-Archimedean local field. Then any reductive representation ρ:π1(X)GLr(F)\rho:\pi_{1}(X)\to\mathrm{GL}_{r}(F) has bounded image.

Therefore, we will be particularly interested in the varieties with vanishing spectral bases. In view of Margulis’ superrigidity [Mar91], examples of varieties with rigid character varieties are Hermitian locally symmetric spaces of higher rank. Thus we would like to understand Simpson’s integrality conjecture from the perspective of Higgs bundles and spectral varieties in the case of Hermitian locally symmetric spaces with rank 2\geq 2, following the approach of Klingler [Kli13]. In particular, we obtain the following:

Theorem 1.3.

Let Ω=Ω1××Ωm\Omega=\Omega_{1}\times\cdots\times\Omega_{m} be a bounded symmetric domain of rank 2\geq 2 together with its decomposition into irreducible factors. Let ΓAut(Ω)\Gamma\subset{\rm Aut}(\Omega) be a torsion-free irreducible cocompact lattice, and write XΩ/ΓX\coloneqq\Omega/\Gamma. Then 𝒮Xr=0\mathcal{S}^{r}_{X}=0 for any r1r\geq 1.

By [Mok89, Appendix IV, Proposition 3], the cotangent bundle of a compact quotient XX of an irreducible bounded symmetric domain Ω\Omega of rank 2\geq 2 by a torsion-free lattice ΓAut(Ω)\Gamma\subset{\rm Aut}(\Omega) is big (aka almost ample in [Mok89]), and a similar proof yields the same when Ω\Omega is reducible and of rank 2\geq 2, and the lattice Γ\Gamma is irreducible — see also [BKT13, Theorem 1.1]. In particular, for kk sufficiently large and sufficiently divisible, we have

h0(X,SymkΩX1)O(k2dim(X)1).h^{0}(X,\mathrm{Sym}^{k}\Omega_{X}^{1})\sim O(k^{2\dim(X)-1}).

So Theorem 1.3 is somewhat surprising because the dimension of the Hitchin space 𝒜Xr\mathcal{A}^{r}_{X} is very large for r1r\gg 1, while the closed subset 𝒮Xr\mathcal{S}_{X}^{r} of 𝒜Xr\mathcal{A}_{X}^{r} is just a single point. This may also be seen as a strong piece of evidence that Theorem 1.2 may be applicable in other interesting cases.

On the other hand, we remark that it has been shown by B. Klingler in [Kli13, Theorem 1.6] the vanishing of the Hitchin base in some small ranges for compact quotients of certain irreducible bounded symmetric domains. Its proof is based on classical plethysm, a vanishing theorem of the last author [Mok89, p. 205 and p. 211] and a case-by-case argument depending on the types of Ω\Omega. However, one cannot expect to generalize its proof to large rr: the dimension of 𝒜Xr\mathcal{A}_{X}^{r} can be as large as possible for r1r\gg 1 as explained above. Our argument is purely geometric and provides a unified approach to all Ω\Omega and all rr. It relies on a Finsler metric rigidity theorem of the last author proved in [Mok04].

Finally we note that Theorem 1.3 cannot be strengthened to the rank-1 case, i.e., the case where Ω=𝐁n\Omega=\mathbf{B}_{\mathbb{C}}^{n}: there exists ball quotients satisfying H0(X,ΩX1)0H^{0}(X,\Omega_{X}^{1})\not=0 and then one can easily construct a non-zero element in 𝒮Xr\mathcal{S}_{X}^{r} for any r1r\geq 1 in this case.

Thanks to Theorem 1.2, we conclude from the vanishing of the spectral base the following result.

Corollary 1.4.

In the notation of Theorem 1.3 and under the assumption given there, any reductive representation ρ:ΓGLr()\rho:\Gamma\to\mathrm{GL}_{r}(\mathbb{C}) is rigid and integral for any r1r\geq 1. Moreover, it is a complex direct factor of a \mathbb{Z}-variation of Hodge structures.

The rigidity result allows us to recover Margulis’ rigidity result [Mar91] in the case of cocompact lattices. It can be seen as a strengthening of Klingler’s result [Kli13] for any rr and any Ω\Omega of rank 2\geq 2. Moreover, we note that Margulis’ rigidity only concerns representations, which are identified with topologically trivial Higgs bundles via the non-abelian Hodge correspondence, while we can also obtain a result on general Higgs bundles from the vanishing of the spectral base as in the corollary given below. This can be used to help us understand the analytic aspect of the Hermitian-Yang-Mills equation, for which we refer the reader to [He20] for a discussion.

Corollary 1.5.

In the notation of Theorem 1.3 and under the assumption given there, every Higgs bundle over XX is nilpotent.

Acknowledgements. The authors also wish to express their gratitude to a great many people for their interest and helpful comments. Among them are Shan-Tai Chan, Ya Deng, Ziyang Gao, Andriy Haydys, Thomas Walpuski, Pengyu Yang, Kang Zuo. S. He is supported by National Key R&D Program of China (No.2023YFA1010500) and NSFC grant (No.12288201). J. Liu is supported by the R&D Program of China (No.2021YFA1002300), the NSFC grant (No.12288201), the CAS Project for Young Scientists in Basic Research (No.YSBR-033) and the Youth Innovation Promotion Association CAS. N. Mok is supported by the GRF grant 17306523 of the Hong Kong Research Grants Council.

2. Higgs bundles and Hitchin morphism

In this section, we delve into the non-abelian Hodge correspondence and explore the Hitchin morphism over a projective variety. This subject has garnered extensive attention in various notable works, including [Hit87a, Sim88, Sim94a, Sim94b]. Readers interested in the topic may consult the surveys [GRR15, Wen16, Sch18].

2.1. Higgs bundle and non-abelian Hodge correspondence

Let XX be a projective manifold, and denote by ΩX1\Omega_{X}^{1} the holomorphic cotangent bundle of XX. We collect some basic definitions and facts about Higgs sheaves/bundles, which we refer [Sim92, BS94, BS09, LZZ17].

Definition 2.1.

A Higgs sheaf on XX is a pair (,φ)(\mathscr{E},\varphi), where \mathscr{E} is a torsion-free coherent sheaf on XX and φ:ΩX1\varphi\colon\mathscr{E}\rightarrow\mathscr{E}\otimes\Omega_{X}^{1}, called a Higgs field, such that the composed morphism

𝜑ΩX1φidΩX1ΩX1idΩX2\mathscr{E}\xrightarrow{\varphi}\mathscr{E}\otimes\Omega_{X}^{1}\xrightarrow{\varphi\otimes\mathrm{id}}\mathscr{E}\otimes\Omega_{X}^{1}\otimes\Omega_{X}^{1}\xrightarrow{\mathrm{id}\otimes\wedge}\mathscr{E}\otimes\Omega_{X}^{2}

vanishes. Following tradition, the composed morphism will be denoted by φφ\varphi\wedge\varphi and the equation φφ=0\varphi\wedge\varphi=0 is called the Higgs equation.

Given a Higgs sheaf (,φ)(\mathscr{E},\varphi) over a projective manifold XX, a coherent subsheaf \mathscr{F}\subset\mathscr{E} is said to be φ\varphi-invariant if and only if φ()ΩX1\varphi(\mathscr{F})\subset\mathscr{F}\otimes\Omega_{X}^{1}. Now we can introduce the concept of slope stability. Recall that, given a torsion-free coherent sheaf \mathscr{E} on an nn-dimensional projective manifold XX, the slope μ()\mu(\mathscr{E}) of \mathscr{E} with respect to ω\omega is defined to be

μ()=degω()rank=c1()ωn1rank.\mu(\mathscr{E})=\frac{\deg_{\omega}(\mathscr{E})}{\mathrm{rank}\;\mathscr{E}}=\frac{c_{1}(\mathscr{E})\cdot\omega^{n-1}}{\mathrm{rank}\;\mathscr{E}}.
Definition 2.2.

A Higgs sheaf (,φ)(\mathscr{E},\varphi) is called stable (resp. semistable) if and only if for any φ\varphi-invariant coherent subsheaf \mathscr{F}\subset\mathscr{E}, with 0<rank()<rank()0<\mathrm{rank}(\mathscr{F})<\mathrm{rank}(\mathscr{E}), we have

μ()<μ()(resp.μ()μ()).\mu(\mathscr{F})<\mu(\mathscr{E})\quad(\textup{resp.}\,\mu(\mathscr{F})\leq\mu(\mathscr{E})).

A Higgs sheaf (,φ)(\mathscr{E},\varphi) is called polystable if (,φ)(\mathscr{E},\varphi) is semistable and

(,φ)(1,φ1)(r,φr),(\mathscr{E},\varphi)\cong(\mathscr{E}_{1},\varphi_{1})\oplus\cdots\oplus(\mathscr{E}_{r},\varphi_{r}){\color[rgb]{0,0,0}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@gray@stroke{0}\pgfsys@color@gray@fill{0},}

where (i,φi)(\mathscr{E}_{i},\varphi_{i}) are stable Higgs sheaves with the same slope.

We will now focus on Higgs bundles, equivalently locally free Higgs sheaves, and will return to the general notion of Higgs sheaves in § 2.4, where such sheaves are constructed from spectral varieties defined by spectral data.

Let EE be a complex smooth vector bundle over XX. We write Ωp,q(E)\Omega^{p,q}(E) for the complex vector space of EE-valued (p,q)(p,q)-forms on XX. In the sequel of this paper, we will naturally identify the holomorphic structures on EE with the ¯\bar{\partial}-operators ¯E:Ωp,q(E)Ωp,q+1(E)\bar{\partial}_{E}\colon\Omega^{p,q}(E)\to\Omega^{p,q+1}(E) satisfying the integrability condition ¯E2=0\bar{\partial}_{E}^{2}=0. We denote by :=(E,¯E)\mathscr{E}:=(E,\bar{\partial}_{E}) the holomorphic vector bundle with the holomorphic structure defined by ¯E\bar{\partial}_{E} if there is no confusion.

Let gAut(E)g\in\mathrm{Aut}(E). Then gg acts on the Higgs bundles =((E,¯E),φ)\mathscr{E}=((E,\bar{\partial}_{E}),\varphi) by g(¯E,φ)=(g1¯Eg,g1Φg).g\cdot(\bar{\partial}_{E},\varphi)=(g^{-1}\circ\bar{\partial}_{E}\circ g,g^{-1}\circ\Phi\circ g). We define the moduli stack of polystable Higgs bundles of rank rr as

(1) Higgsstack,r{(,φ)|(,φ)polystable}/Aut(E).\begin{split}\mathcal{M}^{{\rm stack},r}_{\mathrm{Higgs}}\coloneqq\{(\mathscr{E},\varphi)|(\mathscr{E},\varphi)\;\mathrm{polystable}\}/\mathrm{Aut}(E).\end{split}

A complex smooth vector bundle EE is said to be topologically trivial if all the Chern classes of EE in H(X;)H^{*}(X;\mathbb{Q}) vanish. Let [(,φ)][(\mathscr{E},\varphi)] be the equivalence class of (,φ)(\mathscr{E},\varphi) in the orbit of Aut(E)\mathrm{Aut}(E). Under S-equivalence of Aut(E)\mathrm{Aut}(E) action, a semi-stable topologically trivial Higgs bundle [(,φ)][(\mathscr{E},\varphi)] is polystable. Following [Sim94b, Proposition 6.6], we define the Dolbeault moduli space Dolr\mathcal{M}_{\mathrm{Dol}}^{r} as the moduli space parametrizing topologically trivial polystable Higgs bundles on XX, which is a quasiprojective variety. It follows from [Sim88, Proposition 3.4] that a polystable Higgs bundle (,φ)(\mathscr{E},\varphi) is topologically trivial if and only if c1()ωn1=0c_{1}(\mathscr{E})\cdot\omega^{n-1}=0 and c2()ωn2=0c_{2}(\mathscr{E})\cdot\omega^{n-2}=0.

Let XX be a projective variety. The GLr()\mathrm{GL}_{r}(\mathbb{C}) character variety GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} of XX is defined to be the set of conjugacy classes of reductive representations of the fundamental group given by

(2) GLr(){ρ:π1(X)GLr()ρreductive}/.\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})}\coloneqq\{\rho\colon\pi_{1}(X)\to\mathrm{GL}_{r}(\mathbb{C})\mid\rho\;\mathrm{reductive}\}/\sim.
Theorem 2.3 ([Hit87a, Don87, Sim88, Cor88]).

There exists a bijective map

(3) ξ:DolrGLr(),\xi:\mathcal{M}_{\mathrm{Dol}}^{r}\to\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})},\;

which is real analytic over the smooth locus.

2.2. Hitchin morphism

The Hitchin morphism is a useful tool to study the moduli space of Higgs bundles. In this subsection, we will introduce the Hitchin morphism for projective manifolds, following [Hit87a, Hit87b, Sim94a]. Let XX be a projective manifold. Then the Hitchin base of XX with rank rr is defined to be

(4) 𝒜Xri=1rH0(X,SymiΩX1).\mathcal{A}^{r}_{X}\coloneqq\bigoplus_{i=1}^{r}H^{0}(X,\mathrm{Sym}^{i}\Omega_{X}^{1}).

The Hitchin morphism for the moduli stack of Higgs bundle is defined as follows:

(5) 𝗁X:Higgsstack,r𝒜Xr,[(,φ)](Tr(φ),Tr(φ2),Tr(φr)).\begin{split}\mathsf{h}_{X}:\mathcal{M}_{\mathrm{Higgs}}^{\textup{stack},r}\to\mathcal{A}^{r}_{X},\quad[(\mathscr{E},\varphi)]\mapsto(\mathrm{Tr}(\varphi),\mathrm{Tr}(\varphi^{2})\cdots,\mathrm{Tr}(\varphi^{r})).\end{split}
Theorem 2.4 ([Hit87a, Sim94a]).

The restriction 𝗁X|Dolr:Dolr𝒜Xr\mathsf{h}_{X}|_{\mathcal{M}_{\mathrm{Dol}}^{r}}:\mathcal{M}_{\mathrm{Dol}}^{r}\to\mathcal{A}^{r}_{X} is proper, and it is also surjective in the case where dim(X)=1\dim(X)=1.

2.3. Spectral base

We briefly recall the definition of the spectral base, which was introduced by T. Chen and B. Ngô in [CN20].

Definition 2.5.

The spectral base 𝒮Xr\mathcal{S}^{r}_{X} is the subset of 𝒜Xr\mathcal{A}^{r}_{X} consisting of the elements s=(s1,,sr)𝒜Xr\textbf{s}=(s_{1},\dots,s_{r})\in\mathcal{A}^{r}_{X} such that for any point xXx\in X, there exist rr elements ω1,,ωrΩX,x1\omega_{1},\dots,\omega_{r}\in\Omega_{X,x}^{1} satisfying si(x)=σi(ω1,,ωr)s_{i}(x)=\sigma_{i}(\omega_{1},\dots,\omega_{r}), where σi\sigma_{i} is the ii-th elementary symmetric polynomial in rr variables. Moreover, an element s𝒮Xr\textbf{s}\in\mathcal{S}^{r}_{X} is called a spectral datum.

Let VV be a complex vector space of dimension nn and let Chowr(V)\mathrm{Chow}^{r}(V) be the Chow variety of zero cycles of length rr on VV. By [CN20, Theorem 4.1], the following natural map

Chowr(V)V×Sym2V××SymrV,[v1,,vr](σ1,σ2,,σr),\mathrm{Chow}^{r}(V)\rightarrow V\times\mathrm{Sym}^{2}V\times\dots\times\mathrm{Sym}^{r}V,\quad[v_{1},\dots,v_{r}]\mapsto(\sigma_{1},\sigma_{2},\dots,\sigma_{r}),

is a closed embedding and thus it induces the following closed embedding

(6) Chowr(Tot(ΩX1)/X)Tot(ΩX1)×XTot(Sym2ΩX1)×X×XTot(SymrΩX1),\mathrm{Chow}^{r}(\mathrm{Tot}(\Omega_{X}^{1})/X)\hookrightarrow\mathrm{Tot}(\Omega_{X}^{1})\times_{X}\mathrm{Tot}(\mathrm{Sym}^{2}\Omega_{X}^{1})\times_{X}\dots\times_{X}\mathrm{Tot}(\mathrm{Sym}^{r}\Omega_{X}^{1}),

where Tot()\mathrm{Tot}(\bullet) denotes the total space of the corresponding vector bundle and the space Chowr(Tot(ΩX1)/X)\mathrm{Chow}^{r}(\mathrm{Tot}(\Omega_{X}^{1})/X) is the relative Chow space of zero cycles of length rr. In particular, under this closed embedding, the spectral base 𝒮Xr\mathcal{S}^{r}_{X} can be identified with the space of sections σ:XChowr(Tot(ΩX1)/X)\sigma:X\rightarrow\mathrm{Chow}^{r}(\mathrm{Tot}(\Omega_{X}^{1})/X) and so 𝒮Xr\mathcal{S}^{r}_{X} is a closed subset of 𝒜Xr\mathcal{A}^{r}_{X}.

The following observation shows that it suffices to check the condition in Definition 2.5 over general points to see whether an element s𝒜Xr\textbf{s}\in\mathcal{A}^{r}_{X} is a spectral datum.

Lemma 2.6.

Let s𝒜Xr\textbf{s}\in\mathcal{A}^{r}_{X} be an element. If there exists a dense Zariski open subset XX^{\circ} of XX such that s satisfies the condition in Definition 2.5 for any point xXx\in X^{\circ}, then s𝒮Xr\textbf{s}\in\mathcal{S}^{r}_{X}.

Proof.

Denote by

σ:XTot(ΩX1)×XTot(Sym2ΩX1)×X×XTot(SymrΩX1)\sigma:X\rightarrow\mathrm{Tot}(\Omega_{X}^{1})\times_{X}\mathrm{Tot}(\mathrm{Sym}^{2}\Omega_{X}^{1})\times_{X}\dots\times_{X}\mathrm{Tot}(\mathrm{Sym}^{r}\Omega_{X}^{1})

the section corresponding s. Then the image of σ\sigma is contained in Chowr(Tot(ΩX1)/X)\mathrm{Chow}^{r}(\mathrm{Tot}(\Omega_{X}^{1})/X) over XX^{\circ} by our assumption and (6). However, since (6) is a closed embedding and the image of σ\sigma is irreducible, the image of σ\sigma is contained in Chowr(Tot(ΩX1)/X)\mathrm{Chow}^{r}(\mathrm{Tot}(\Omega_{X}^{1})/X). Hence, σ\sigma is a section of Chowr(Tot(ΩX1)/X)X\mathrm{Chow}^{r}(\mathrm{Tot}(\Omega_{X}^{1})/X)\rightarrow X. ∎

Remark 2.7.
  1. (1)

    As an immediate consequence of Lemma 2.6, one can derive the birational invariance of the spectral base 𝒮Xr\mathcal{S}^{r}_{X} as proved by L. Song and H. Sun in [SS24, Theorem 5.3].

  2. (2)

    For any positive integers r<rr^{\prime}<r, there exists a natural inclusion 𝒮Xr𝒮Xr\mathcal{S}_{X}^{r^{\prime}}\subset\mathcal{S}_{X}^{r} defined as following:

    s=(s1,,sr)s:=(s1,,sr,0,,0)𝒜Xr.\textbf{s}^{\prime}=(s_{1},\dots,s_{r^{\prime}})\mapsto\textbf{s}:=(s^{\prime}_{1},\cdots,s^{\prime}_{r},0,\cdots,0)\in\mathcal{A}_{X}^{r}.

    We only need to show that s is contained in 𝒮Xr\mathcal{S}_{X}^{r}. Indeed, given an arbitrary point xXx\in X, let w1,,wrΩX,x1w_{1},\dots,w_{r^{\prime}}\in\Omega_{X,x}^{1} be the points such that si=σi(w1,,wr)s^{\prime}_{i}=\sigma_{i}(w_{1},\dots,w_{r^{\prime}}). Then one can easily conclude by considering the set {w1,,wr,0rr}\{w_{1},\dots,w_{r^{\prime}},0^{r-r^{\prime}}\}. In particular, if 𝒮Xr=0\mathcal{S}_{X}^{r}=0, then so is 𝒮Xr\mathcal{S}_{X}^{r^{\prime}} for any rrr^{\prime}\leq r.

Proposition 2.8 ([CN20, Proposition 5.1]).

The Hitchin morphism 𝗁X:Higgsstack,r𝒜Xr\mathsf{h}_{X}:\mathcal{M}_{\mathrm{Higgs}}^{\textup{stack},r}\to\mathcal{A}^{r}_{X} factors through the natural inclusion map ιX:𝒮Xr𝒜Xr\iota_{X}:\mathcal{S}^{r}_{X}\rightarrow\mathcal{A}^{r}_{X}. In other words, there exists a map 𝗌𝖽X:Higgsstack,r𝒮Xr\mathsf{sd}_{X}:\mathcal{M}_{\mathrm{Higgs}}^{\textup{stack},r}\to\mathcal{S}^{r}_{X} such that the following diagram commutes:

(7) Higgsstack,r{\mathcal{M}_{\mathrm{Higgs}}^{\textup{stack},r}}𝒮Xr{\mathcal{S}^{r}_{X}}𝒜Xr.{\mathcal{A}^{r}_{X}.}𝗌𝖽X\scriptstyle{\mathsf{sd}_{X}}𝗁X\scriptstyle{\mathsf{h}_{X}}ιX\scriptstyle{\iota_{X}}

The map 𝗌𝖽X\mathsf{sd}_{X} is called the spectral morphism

Proof.

Let (,φ)(\mathscr{E},\varphi) be a rank rr Higgs bundle with Tr(φi)=siH0(X,SymiΩX1)\mathrm{Tr}(\varphi^{i})=s_{i}\in H^{0}(X,\mathrm{Sym}^{i}\Omega_{X}^{1}). Given an arbitrary point xXx\in X, let dz1,dz2,,dzndz^{1},dz^{2},\cdots,dz^{n} be a frame of ΩX1\Omega^{1}_{X} at xXx\in X. If we write φ(x)=i=1nAidzi\varphi(x)=\sum_{i=1}^{n}A_{i}dz^{i}, then the condition φφ=0\varphi\wedge\varphi=0 implies that [Ai,Aj]=0[A_{i},A_{j}]=0 for any 1ij1\leq i\leq j. Thus AiA_{i}’s can be simultaneously upper-triangularized and so is φ(x)\varphi(x) as an r×rr\times r-matrix with values in one forms. In particular, after changing local coordinates, we may assume that φ(x)\varphi(x) is a upper triangular matrix and let ω1,ωrΩX,x1\omega_{1},\dots\omega_{r}\in\Omega^{1}_{X,x} be its diagonal elements. By the definition of the Hitchin morphism, we have si(x)=σi(ω1,,ωr)s_{i}(x)=\sigma_{i}(\omega_{1},\dots,\omega_{r}) and hence we are done. ∎

In [CN20], T. Chen and B. Ngô conjectured that 𝗌𝖽X\mathsf{sd}_{X} is surjective. This conjecture has been confirmed in [CN20] and [SS24] for smooth projective surfaces, in [HL23] for rank two case and studied in [BKU23] for abelian variety. However, in general the moduli stack Higgsstack,r\mathcal{M}_{\mathrm{Higgs}}^{{\rm stack},r} may be much larger than Dolr\mathcal{M}_{\mathrm{Dol}}^{r}. We are in particular interested in the image of the restriction of the spectral morphism to the Dolbeault moduli space, which leads to the following definition.

Definition 2.9.

The Dolbeault spectral base 𝒮X;Dolr\mathcal{S}^{r}_{X;\mathrm{Dol}} is defined to be the image 𝗌𝖽X(Dolr)\mathsf{sd}_{X}(\mathcal{M}_{\mathrm{Dol}}^{r}).

Clearly we have the natural inclusions 𝒮X;Dolr𝒮Xr𝒜Xr\mathcal{S}^{r}_{X;\mathrm{Dol}}\subset\mathcal{S}^{r}_{X}\subset\mathcal{A}^{r}_{X} and both of them are strict in the general case (see [HL23, Example 3.4]). Moreover, since by Theorem 2.4 the restriction 𝗁X|Dolr\mathsf{h}_{X}|_{\mathcal{M}_{\mathrm{Dol}}^{r}} is proper, the Dolbeault spectral base 𝒮X;Dolr\mathcal{S}^{r}_{X;\mathrm{Dol}} is a closed subset of 𝒮Xr\mathcal{S}^{r}_{X}.

2.4. Spectral variety and its decomposition

Let p:Tot(ΩX1)Xp:\mathrm{Tot}(\Omega_{X}^{1})\rightarrow X be the natural projection. Given a spectral datum s=(s1,,sr)𝒮Xr\textbf{s}=(s_{1},\dots,s_{r})\in\mathcal{S}^{r}_{X}, the spectral variety XsX_{\textbf{s}} corresponding to s is the closed subscheme of Tot(ΩX1)\mathrm{Tot}(\Omega_{X}^{1}) defined as follows

Xs{λrs1λr1++(1)r1sr1λ+(1)rsr=0},X_{\textbf{s}}\coloneqq\{\lambda^{r}-s_{1}\lambda^{r-1}+\dots+(-1)^{r-1}s_{r-1}\lambda+(-1)^{r}s_{r}=0\},

where λH0(Tot(ΩX1),pΩX1)\lambda\in H^{0}(\mathrm{Tot}(\Omega_{X}^{1}),p^{*}\Omega_{X}^{1}) is the Liouville form and for any 1ir1\leq i\leq r, the term siλris_{i}\lambda^{r-i} is regarded as an element of H0(Tot(ΩX1),SymrpΩX1)H^{0}(\mathrm{Tot}(\Omega_{X}^{1}),\mathrm{Sym}^{r}p^{*}\Omega_{X}^{1}). In particular, the subscheme XsX_{\textbf{s}} is locally defined by (n+r1r1)\binom{n+r-1}{r-1} equations. To understand the spectral variety, we introduce the notion of multivalued holomorphic 11-forms — see [CDY22, Definition 5.8].

Definition 2.10.

Let XX be a projective manifold, and let {Ui}\{U_{i}\} be an open covering of XX in the Euclidean topology. A multivalued holomorphic 11-form is a collection of multisets {ωi1,,ωir}\{\omega_{i1},\cdots,\omega_{ir}\} where ωilH0(Ui,ΩX1|Ui)\omega_{il}\in H^{0}(U_{i},\Omega_{X}^{1}|_{U_{i}}) and over UiUjU_{i}\cap U_{j}, we have {ωi1,,ωir}={ωj1,,ωjr}\{\omega_{i1},\cdots,\omega_{ir}\}=\{\omega_{j1},\cdots,\omega_{jr}\} counted with multiplicity. We write [(ω1,,ωr)][(\omega_{1},\cdots,\omega_{r})] to denote a multivalued holomorphic 11-form.

Let X^𝐬\widehat{X}_{\mathbf{s}} be the reduced scheme underlying X𝐬X_{\mathbf{s}}; that is, X^𝐬\widehat{X}_{\mathbf{s}} is the same topological space as X𝐬X_{\mathbf{s}}, but with the reduced structure sheaf. Then the natural morphism π:X^𝐬X\pi:\widehat{X}_{\mathbf{s}}\rightarrow X is surjective and finite. In particular, there exists a dense Zariski open subset XX^{\circ} of XX such that X^𝐬π1(X)X\widehat{X}^{\circ}_{\mathbf{s}}\coloneqq\pi^{-1}(X^{\circ})\rightarrow X_{\circ} is an unramified finite covering. Moreover, we can also write

X^𝐬=k=1mX^𝐬,k\widehat{X}^{\circ}_{\mathbf{s}}=\bigcup_{k=1}^{m}\widehat{X}^{\circ}_{\mathbf{s},k}

for the decomposition of X^𝐬\widehat{X}^{\circ}_{\mathbf{s}} into irreducible components. Since X^𝐬X\widehat{X}^{\circ}_{\mathbf{s}}\rightarrow X^{\circ} is unramified, the decomposition above is actually a disjoint union. On the other hand, one can easily see that each irreducible component X^𝐬,k\widehat{X}^{\circ}_{\mathbf{s},k} defines a multivalued holomorphic 11-form [(ω1k,,ωrkk)][(\omega_{1}^{k},\dots,\omega^{k}_{r_{k}})] over XX^{\circ} whose local representatives have no multiple elements.

Now we can define the multiplicity of X^𝐬,k\widehat{X}^{\circ}_{\mathbf{s},k} in the spectral variety X𝐬X_{\mathbf{s}}. For any kk and ll, we define the multiplicity m(k,l)m(k,l) of the section ωlk\omega^{k}_{l} to be its multiplicity as a root of the equation

λrs1λn1++(1)n1sn1λ+(1)nsn=0.\lambda^{r}-s_{1}\lambda^{n-1}+\dots+(-1)^{n-1}s_{n-1}\lambda+(-1)^{n}s_{n}=0.
Lemma 2.11.

The multiplicity m(k,l)m(k,l) is independent of ll.

Proof.

Define a function μ:X^𝐬o\mu:\widehat{X}_{\bf s}^{o}\to\mathbb{N} as follows. Any point wX^sow\in\widehat{X}_{s}^{o} is a tangent covector at x=π(w)x=\pi(w) of the form w=ωj(x)w=\omega_{j}(x) for some jj, 1jr1\leq j\leq r. Define now μ(w)\mu(w) to be the multiplicity of ωj\omega_{j} at xXox\in X^{o}. From the definition of the multiplicity it follows that μ\mu is locally constant on X^𝐬o\widehat{X}_{\bf s}^{o}, hence it must be constant on each connected component X^𝐬,ko\widehat{X}_{{\bf s},k}^{o}, 1km1\leq k\leq m. ∎

We shall denote m(k,l)m(k,l) by m(k)m(k). Then we have k=1mm(k)rk=r\sum_{k=1}^{m}m(k)r_{k}=r. Let X^𝐬,k\widehat{X}_{\mathbf{s},k} be the closure of X^𝐬,k\widehat{X}^{\circ}_{\mathbf{s},k} in X^𝐬\widehat{X}_{\mathbf{s}} and let π^k:X^𝐬,kX\hat{\pi}_{k}:\widehat{X}_{\mathbf{s},k}\rightarrow X be the natural morphism. We define

^kπ^k𝒪X^𝐬,k.\hat{\mathscr{F}}_{k}\coloneqq\hat{\pi}_{k*}\mathscr{O}_{\hat{X}_{\mathbf{s},k}}.

Since X^𝐬,k\hat{X}_{\mathbf{s},k} is integral, the structure sheaf 𝒪X^𝐬,k\mathscr{O}_{\hat{X}_{\mathbf{s},k}} is torsion-free. So ^k\hat{\mathscr{F}}_{k} is also torsion-free and it carries a natural Higgs field ψk\psi_{k} defined as follows:

(8) ψk:^k=π^k𝒪X^𝐬,k×λπ^k(𝒪X^𝐬,kπ^kΩX1)=^kΩX1.\psi_{k}:\hat{\mathscr{F}}_{k}=\hat{\pi}_{k*}\mathscr{O}_{\widehat{X}_{\mathbf{s},k}}\xrightarrow{\times\lambda}\hat{\pi}_{k*}\left(\mathscr{O}_{\widehat{X}_{\mathbf{s},k}}\otimes\hat{\pi}_{k}^{*}\Omega_{X}^{1}\right)=\hat{\mathscr{F}}_{k}\otimes\Omega_{X}^{1}.

Recall that a coherent sheaf \mathscr{F} over a complex manifold is said to be a normal sheaf if Hartogs’ extension across subvarieties of codimension 2\geq 2 holds, and \mathscr{F} is said to be a reflexive sheaf if =\mathscr{F}^{**}=\mathscr{F}. By [OSS11, Chapter 2, Lemma 1.1.12], a coherent sheaf on a complex manifold is reflexive if and only if it is normal and torsion-free.

Proposition 2.12.

Let k\mathscr{E}_{k} be the reflexive hull of ^k\hat{\mathscr{F}}_{k}, i.e., k=^k\mathscr{E}_{k}=\hat{\mathscr{F}}_{k}^{**}.

  1. (1)

    The Higgs field ψk\psi_{k} extends to a Higgs field φk\varphi_{k} on k\mathscr{E}_{k}.

  2. (2)

    The sheaf k\mathscr{E}_{k} carries a natural 𝒪X\mathscr{O}_{X}-algebra structure.

Proof.

There exists a dense Zariski open subset ι:UX\iota:U\hookrightarrow X such that XUX\setminus U is of codimension 2\geq 2 in XX and such that ^k\hat{\mathscr{F}}_{k} is locally free on UU, so that ^k|Uk|U\hat{\mathscr{F}}_{k}|_{U}\cong\mathscr{E}_{k}|_{U}. Hence, k|U\mathscr{E}_{k}|_{U} inherits from ^k\hat{\mathscr{F}}_{k} the structure of an 𝒪U\mathscr{O}_{U}-algebra. On the other hand, since k\mathscr{E}_{k} is reflexive, we have

kι(k|U)=ι(^k|U).\mathscr{E}_{k}\cong\iota_{*}(\mathscr{E}_{k}|_{U})=\iota_{*}(\hat{\mathscr{F}}_{k}|_{U}).

In particular, the restriction

k|U^k|Uψk|U^k|UΩU1k|UΩU1\mathscr{E}_{k}|_{U}\cong\hat{\mathscr{F}}_{k}|_{U}\xrightarrow{\psi_{k}|_{U}}\hat{\mathscr{F}}_{k}|_{U}\otimes\Omega_{U}^{1}\cong\mathscr{E}_{k}|_{U}\otimes\Omega_{U}^{1}

extends to a morphism φk:kkΩX1\varphi_{k}:\mathscr{E}_{k}\rightarrow\mathscr{E}_{k}\otimes\Omega_{X}^{1} and we have an extension of the structure of the 𝒪U\mathscr{O}_{U}-algebra structure on k|U\mathscr{E}_{k}|_{U} to an 𝒪X\mathscr{O}_{X}-algebra structure on k\mathscr{E}_{k}. ∎

Remark 2.13.

Let X~k\widetilde{X}_{k} be the variety defined as Spec𝒪Xk\textup{Spec}_{\mathscr{O}_{X}}\mathscr{E}_{k}. Then we have a natural finite birational morphism X~𝐬,kX^𝐬,k\widetilde{X}_{\mathbf{s},k}\rightarrow\widehat{X}_{\mathbf{s},k}, which is an isomorphism in codimension one. Moreover, since k\mathscr{E}_{k} is locally free in codimension two, the variety X~𝐬,k\widetilde{X}_{\mathbf{s},k} is Cohen-Macaulay in codimension two ([Ser65, IV, D, Corollaire 2]). In particular, the variety X~𝐬,k\widetilde{X}_{\mathbf{s},k} is Cohen-Macaulay if dim(X)=2\dim(X)=2.

Since XX is smooth and k\mathscr{E}_{k} is reflexive, there exists a dense Zariski open subset XX^{\circ\circ} of XX such that XXX\setminus X^{\circ\circ} is of codimension 3\geq 3 in XX and such that the restriction k|X\mathscr{E}_{k}|_{X^{\circ\circ}} is locally free ([OSS11, Chapter 2, Lemma 1.1.10]). Denote by φk\varphi_{k}^{\circ\circ} the restriction φk|X\varphi_{k}|_{X^{\circ\circ}} and let

(s1k,,srkk)𝗁X([k|X,φk])i=1rkH0(X,SymiΩX1).(s_{1k}^{\circ\circ},\cdots,s_{r_{k}k}^{\circ\circ})\coloneqq\mathsf{h}_{X^{\circ\circ}}\big{(}[\mathscr{E}_{k}|_{X^{\circ\circ}},\varphi_{k}^{\circ\circ}]\big{)}\in\bigoplus_{i=1}^{r_{k}}H^{0}(X^{\circ\circ},\mathrm{Sym}^{i}\Omega_{X^{\circ\circ}}^{1}).

Since codim(XX)3\mathrm{codim}(X\setminus X^{\circ\circ})\geq 3 and SymiΩX1\mathrm{Sym}^{i}\Omega_{X}^{1} is locally free, the sections siks_{ik}^{\circ\circ} extend to sections sikH0(X,SymiΩX1)s_{ik}\in H^{0}(X,\mathrm{Sym}^{i}\Omega_{X}^{1}). Let

Pk(T)Trks1kTrk1++(1)rksrkkP_{k}(T)\coloneqq T^{r_{k}}-s_{1k}T^{r_{k}-1}+\dots+(-1)^{r_{k}}s_{r_{k}k}

be the corresponding characteristic polynomial. Over the dense Zariski open subset XX^{\circ}, one can easily derive the following equality of polynomials

(9) P(T)Trs1Tr1++(1)ksk=k=1mPk(T)m(k).P(T)\coloneqq T^{r}-s_{1}T^{r-1}+\cdots+(-1)^{k}s_{k}=\prod_{k=1}^{m}P_{k}(T)^{m(k)}.

Then it follows that the equality above actually holds over the whole XX by comparing the coefficients. Now let X𝐬,kTot(ΩX1)X_{\mathbf{s},k}\subset\mathrm{Tot}(\Omega_{X}^{1}) be the subscheme defined by Pk(λ)=0P_{k}(\lambda)=0. Then clearly we have X𝐬,kp1(X)=X^𝐬,kX_{\mathbf{s},k}\cap p^{-1}(X^{\circ})=\widehat{X}^{\circ}_{\mathbf{s},k} and we may also write the equality (9) as an equation on cycles in the form

(10) [X𝐬]=k=1mm(k)[X𝐬,k],[X_{\mathbf{s}}]=\sum_{k=1}^{m}m(k)[X_{\mathbf{s},k}],

where for a pure-dimensional complex subspace AA of XX, [A][A] denotes the cycle in the Chow space Chow(X)(X) associated to AA.

Remark 2.14.

An analogue of the decomposition (10) was obtained by L. Song and H. Sun in [SS24, § 3] in the case where XX is a smooth projective surface and in higher dimension one can use (10) to reduce Chen–Ngô’s conjecture to the case where the spectral cover XsX_{\textbf{s}} is irreducible and generically reduced, i.e., m=1m=1 and m(1)=1m(1)=1.

The Hitchin morphism (5) and the spectral morphism (7) can be directly extended to Higgs sheaves and one can immediately derive the following result from our argument above.

Proposition 2.15.

Let XX be a projective manifold. Given a spectral datum 𝐬𝒮Xr\mathbf{s}\in\mathcal{S}_{X}^{r}, there exists a reflexive Higgs sheaf (,φ)(\mathscr{E},\varphi) of rank rr over XX such that 𝗌𝖽X([(,φ)])=𝐬\mathsf{sd}_{X}([(\mathscr{E},\varphi)])=\mathbf{s}.

Proof.

We conclude by letting (,φ)=k=1m(k,φk)m(k)(\mathscr{E},\varphi)=\oplus_{k=1}^{m}(\mathscr{E}_{k},\varphi_{k})^{\oplus m(k)}. ∎

Remark 2.16.
  1. (1)

    Since reflexive sheaves are locally free in codimension two ([OSS11, Chapter 2, Lemma 1.1.10]), one can use Proposition 2.15 to recover the surjectivity of the spectral morphism 𝗌𝖽X\mathsf{sd}_{X} proved in [CN20] and [SS24] in the surface case.

  2. (2)

    Assume that the spectral variety X𝐬X_{\mathbf{s}} is irreducible and generically reduced. As π:X𝐬X\pi\colon X_{\mathbf{s}}\rightarrow X is finite, the natural surjection 𝒪X𝐬𝒪X^𝐬\mathscr{O}_{X_{\mathbf{s}}}\rightarrow\mathscr{O}_{\hat{X}_{\mathbf{s}}} induces a surjection

    π𝒪X𝐬π^𝒪X^𝐬^,\mathscr{F}\coloneqq\pi_{*}\mathscr{O}_{X_{\mathbf{s}}}\rightarrow\hat{\pi}_{*}\mathscr{O}_{\hat{X}_{\mathbf{s}}}\eqqcolon\hat{\mathscr{F}},

    where π:X𝐬X\pi\colon X_{\mathbf{s}}\rightarrow X and π^:X^𝐬X\hat{\pi}\colon\hat{X}_{\mathbf{s}}\rightarrow X are the natural finite morphisms, respectively. Moreover, since ^\mathscr{F}\rightarrow\hat{\mathscr{F}} is an isomorphism over the generic point of XX and ^\hat{\mathscr{F}} is torsion-free, we must have

    ^/𝒯(),\hat{\mathscr{F}}\cong\mathscr{F}/\mathscr{T}(\mathscr{F}),

    where 𝒯()\mathscr{T}(\mathscr{F}) is the torsion subsheaf of \mathscr{F}. In particular, we have X^𝐬=Spec𝒪X^/𝒯()\hat{X}_{\mathbf{s}}=\textup{Spec}_{\mathscr{O}_{X}}\hat{\mathscr{F}}/\mathscr{T}(\mathscr{F}). This construction has already appeared in [CN20, Remark 7.1] and in some special case the sheaf ^\hat{\mathscr{F}} is already locally free, and hence reflexive — see [CN20, Example 8.1]. In particular, our construction of X~𝐬\widetilde{X}_{\mathbf{s}} can be viewed as a generalisation of that given in [CN20, Remark 7.1].

  3. (3)

    Assume that dim(X)=2\dim(X)=2, and X𝐬X_{\mathbf{s}} is irreducible and generically reduced. Then the reflexive hull ^\mathscr{E}\coloneqq\hat{\mathscr{F}}^{**} is locally free. So the variety X~𝐬Spec𝒪X\widetilde{X}_{\mathbf{s}}\coloneqq\textup{Spec}_{\mathscr{O}_{X}}\mathscr{E} is a finite Cohen-Macaulayfication of X𝐬X_{\mathbf{s}} (Remark 2.13 and [Ser65, IV, D, Corollaire 2]). On the other hand, T. Chen and B. Ngô has also constructed a Cohen-Macaulayfication XCMX^{\textup{CM}} of X𝐬X_{\mathbf{s}} in [CN20, Proposition 7.2] via the Hilbert scheme. Now we claim that X~𝐬\widetilde{X}_{\mathbf{s}} is actually isomorphic to XCMX^{\textup{CM}}. Indeed, let UU be the largest open subset of XX such that ^|U\hat{\mathscr{F}}|_{U} is locally free. Then XUX\setminus U has codimension 2\geq 2 in XX and we have

    |U|U|U\mathscr{F}|_{U}\cong\mathscr{E}|_{U}\cong\mathscr{E}^{\prime}|_{U}

    as 𝒪U\mathscr{O}_{U}-algebras ([CN20, Proposition 7.2 and Remark 7.1]), where π𝒪XCM\mathscr{E}^{\prime}\coloneqq\pi^{\prime}_{*}\mathscr{O}_{X^{\textup{CM}}} and π:XCMX\pi^{\prime}\colon X^{\textup{CM}}\rightarrow X is the natural finite morphism. Since both \mathscr{E} and \mathscr{E}^{\prime} are locally free, we get an isomorphism of \mathscr{E} and \mathscr{E}^{\prime} as 𝒪X\mathscr{O}_{X}-algebras. Hence, there exists an isomorphism between X~𝐬\widetilde{X}_{\mathbf{s}} and XCMX^{\textup{CM}} satisfying the following commutative diagram:

    X~𝐬{\widetilde{X}_{\mathbf{s}}}XCM{X^{\textup{CM}}}X^𝐬{\hat{X}_{\mathbf{s}}}\scriptstyle{\cong}

3. The spectral base for a quotient of a bounded symmetric domain with rank2\mathrm{rank}\geq 2.

Let XX be a quotient of a bounded symmetric domain by an irreducible torsion free cocompact lattice. In this section, we will explain the relationship between the spectral base and Finsler metrics. Moreover, we will use the Finsler metric rigidity theorem of the last author [Mok04] to prove the vanishing of the spectral base whenever rank(X)2{\rm rank}(X)\geq 2.

3.1. Finsler (pseudo-)metric

Let \mathscr{L} be a holomorphic line bundle over a complex manifold XX. We briefly recall the definition of a (singular) Hermitian metric on \mathscr{L}.

Definition 3.1.

A singular (Hermitian) metric hh on a line bundle FF is a metric which is given in any trivialization θ:L|UU×\theta:L|_{U}\xrightarrow{\cong}U\times\mathbb{C} by

ξh=|θ(ξ)|eφ(x),xU,ξx,\|\xi\|_{h}=|\theta(\xi)|e^{-\varphi(x)},x\in U,\xi\in\mathscr{L}_{x},

where φLloc1(U)\varphi\in L^{1}_{\textup{loc}}(U) is an arbitrary locally integrable function, called the weight of the metric with respect to the trivialization θ\theta.

The curvature current of \mathscr{L} is given formally by the closed (1,1)(1,1)-current 12πΘ,h=ddcφ\frac{\sqrt{-1}}{2\pi}\Theta_{\mathscr{L},h}=dd^{c}\varphi on UU. The assumption φLloc1\varphi\in L^{1}_{\textup{loc}} guarantees that Θ,h\Theta_{\mathscr{L},h} exists in the sense of distribution theory. Moreover, for the curvature current for is globally defined over XX and independent of the choice of trivialisations, and its de Rham cohomology class is the image of the first Chern class c1()H2(X,)c_{1}(\mathscr{L})\in H^{2}(X,\mathbb{Z}) in HdR2(X,)H^{2}_{\textup{dR}}(X,\mathbb{R}). If we assume in addition that φ𝒞(U,)\varphi\in\mathcal{C}^{\infty}(U,\mathbb{R}), then hh is the usual smooth Hermitian metric on \mathscr{L}.

Example 3.2.

Let DD be an effective divisor and let =𝒪X(D)\mathscr{L}=\mathscr{O}_{X}(-D) be the ideal sheaf of DD. Let 𝒪X\mathscr{L}\rightarrow\mathscr{O}_{X} be the natural non-zero map to the trivial line bundle 𝒪X\mathscr{O}_{X} over XX. Then the standard Hermitian metric over 𝒪X\mathscr{O}_{X} induces a singular Hermitian metric hh over \mathscr{L}. Indeed, let gg be the generator of 𝒪X(D)\mathscr{O}_{X}(-D) on an open subset UU of XX, then

θ(u)=ug\theta(u)=\frac{u}{g}

defines a trivialisation of 𝒪X(D)\mathscr{O}_{X}(-D) over UU, thus our singular metric is associated to the weight φ=log|g|\varphi=-\log|g|. By the Lelong–Poincaré equation, we find

i2πΘ=ddcφ=[D],\frac{i}{2\pi}\Theta_{\mathscr{L}}=dd^{c}\varphi=-[D],

where [D][D] denotes the current of integration over DD.

Let \mathscr{E} be a holomorphic vector bundle over a complex manifold XX. Let ()\mathbb{P}(\mathscr{E}) be the projectivisation in the geometric sense, i.e., ()\mathbb{P}(\mathscr{E}) parametrises the one-dimensional linear subspaces contained in the fibres of \mathscr{E}. Let 𝒪()(1)π\mathscr{O}_{\mathbb{P}(\mathscr{E})}(-1)\subset\pi^{*}\mathscr{E} be the dual tautological line bundle over ()\mathbb{P}(\mathscr{E}), where π:()X\pi:\mathbb{P}(\mathscr{E})\rightarrow X is the natural projection. Given a (singular) Hermitian metric hh over 𝒪()(1)\mathscr{O}_{\mathbb{P}(\mathscr{E})}(-1), we can define a pseudo-metric h¯\bar{h} over \mathscr{E} as in the following. For any vx{0}v\in\mathscr{E}_{x}\setminus\{0\}, we define

vh¯:=vh,\|v\|_{\bar{h}}:=\|v\|_{h},

where on the right-hand side we regard vv as the corresponding point in the fibre of the natural projection 𝒪()(1)()\mathscr{O}_{\mathbb{P}(\mathscr{E})}(-1)\rightarrow\mathbb{P}(\mathscr{E}) over [v][v]. Such a metric h¯\bar{h} is called a (complex) Finsler pseudo-metric and we call it a (complex) Finsler metric if the metric hh is a smooth Hermitian metric.

3.2. Bounded symmetric domain and Finsler metric rigidity

We collect some basic definitions and facts about bounded symmetric domains and we refer the interested reader to [Mok89] for more details. Let Ωn\Omega\Subset\mathbb{C}^{n} be a bounded domain in a complex Euclidean space. We say that Ω\Omega is a bounded symmetric domain if and only if at each xΩx\in\Omega, there exists a biholomorphism σx:ΩΩ\sigma_{x}:\Omega\rightarrow\Omega such that σx2=id\sigma_{x}^{2}=\mathrm{id} and xx is an isolated fixed point of σx\sigma_{x}. In this case, the Bergman metric dsΩ2ds_{\Omega}^{2} with Kähler form ω\omega on Ω\Omega is Kähler–Einstein and (Ω,ω)(\Omega,\omega) is a Hermitian symmetric space of the non-compact type. The rank of Ω\Omega is defined to be the rank of (Ω,dsΩ2)(\Omega,ds_{\Omega}^{2}) as a Riemannian symmetric manifold. We say that the bounded symmetric domain Ω\Omega is irreducible if and only if (Ω,dsΩ2)(\Omega,ds^{2}_{\Omega}) is an irreducible Riemannian symmetric manifold. Denote by Aut(Ω)\mathrm{Aut}(\Omega) the group of biholomorphic self-mappings on Ω\Omega. Write GG for the identity component Auto(Ω)\mathrm{Aut}_{o}(\Omega) and let KGK\subset G be the isotropy subgroup at a point oΩo\in\Omega, so that Ω=G/K\Omega=G/K as a homogeneous space.

Now we introduce the minimal characteristic bundle — see [Mok89, Chapter 6,§ 1] and [Mok02]. Let Ω\Omega be an irreducible bounded symmetric domain. Then we can identify Ω\Omega as a subdomain of its compact dual MM by the Borel embedding ([Mok89, Chapter 3, § 3]). Let GG^{\mathbb{C}} be the automorphism group of MM and PGP\subset G be the isotropy subgroup at oo. Then GGG^{\mathbb{C}}\supset G is a complexification of GG. Consider the action of PP on (ToM)\mathbb{P}(T_{o}M). There are exactly rr orbits 𝒪k(ToM)=(ToΩ)\mathcal{O}_{k}\subset\mathbb{P}(T_{o}M)=\mathbb{P}(T_{o}\Omega), 1kr1\leq k\leq r, such that the topological closures 𝒪¯k\bar{\mathcal{O}}_{k} form an ascending chain of subvarieties of (ToM)\mathbb{P}(T_{o}M) with 𝒪¯r=(ToM)\bar{\mathcal{O}}_{r}=\mathbb{P}(T_{o}M). In particular, the variety 𝒪1\mathcal{O}_{1} is the unique closed orbit, which is thus a homogeneous projective submanifold of (ToM)\mathbb{P}(T_{o}M). Moreover, the submanifold 𝒪1(ToM)\mathcal{O}_{1}\subset\mathbb{P}(T_{o}M) is nothing other than the variety of minimal rational tangents (VMRT) of MM at oo, i.e., the variety of tangent directions at oo of projective lines on MM passing through oo with respect to the first canonical projective embedding of MM. In particular, the subvariety 𝒪1(ToM)\mathcal{O}_{1}\subset\mathbb{P}(T_{o}M) is linearly non-degenerate, i.e., not contained in a hyperplane. The GG-orbit 𝒮(Ω)\mathcal{S}(\Omega) of a point 0[η]𝒪10\not=[\eta]\in\mathcal{O}_{1} is a holomorphic bundle of homogeneous projective manifolds over Ω\Omega, which is called the minimal characteristic bundle of Ω\Omega.

Let Ω\Omega be an arbitrary bounded symmetric domain, and write Ω=Ω1××Ωm\Omega=\Omega_{1}\times\cdots\times\Omega_{m} for its decomposition into the Cartesian product of its irreducible factors. Write TΩ=T1TmT\Omega=T_{1}\oplus\cdots\oplus T_{m} for the corresponding direct sum decomposition of the holomorphic tangent bundle. Let us denote by 𝒮i(Ω)(TΩ)\mathcal{S}^{i}(\Omega)\subset\mathbb{P}(T{\Omega}) the holomorphic bundle over Ω\Omega obtained from the natural embedding of 𝒮(Ωi)(TΩi)\mathcal{S}(\Omega_{i})\subset\mathbb{P}(T\Omega_{i}) into the projective subbundle (Ti)(TΩ)\mathbb{P}(T_{i})\subset\mathbb{P}(T\Omega). Let X=Ω/ΓX=\Omega/\Gamma be the quotient space by a torsion-free irreducible cocompact lattice. Then the Bergman metric dsΩ2ds^{2}_{\Omega} on Ω\Omega descends to a quotient metric gg on XX, which is again a Kähler-Einstein metric. We have the following Finsler metric rigidity theorem on XX proved by the last author in [Mok04].

Theorem 3.3 ([Mok04, Theorem and Remarks]).

Let Ω=Ω1××Ωm\Omega=\Omega_{1}\times\cdots\times\Omega_{m} be a bounded symmetric domain of rank 2\geq 2 together with its decomposition into irreducible factors. Let ΓAut(Ω)\Gamma\subset\mathrm{Aut}(\Omega) be a torsion-free irreducible cocompact lattice and set X:=Ω/ΓX{\color[rgb]{0,0,0}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@gray@stroke{0}\pgfsys@color@gray@fill{0}:}=\Omega/\Gamma. Let gg be the canonical Kähler–Einstein metric on XX, and let hh be a continuous complex Finsler pseudo-metric on XX such that the curvature current of the associated possibly singular continuous Hermitian metric on the line bundle 𝒪(TX)(1)\mathscr{O}_{\mathbb{P}(TX)}(-1) is non-positive. Denote by g\|\cdot\|_{g} (resp. h\|\cdot\|_{h}) lengths of vectors measured with respect to gg (resp. hh). Then there exist non-negative constants c1,,cmc_{1},\cdots,c_{m} such that for any vTXv\in TX that can be lifted to a vector vTiv^{\prime}\in T_{i} with [v]𝒮i(Ω)[v^{\prime}]\in\mathcal{S}^{i}(\Omega), 1im1\leq i\leq m, we have vh=civg\|v\|_{h}=c_{i}\|v\|_{g}.

3.3. Rigidity and Integrality

We will apply Finsler metric rigidity to study the rigidity and integrality of irreducible compact quotients of bounded symmetric domain of complex dimension 2\geq 2. We start with the following easy lemma from linear algebra.

Lemma 3.4.

Fix a positive integer nn and denote by σ1,,σn\sigma_{1},\cdots,\sigma_{n} the elementary symmetric polynomials in nn variables. Let VV be a complex vector space of dimension rr. Let L1,,LnL_{1},\cdots,L_{n} be nn ( possibly non-distinct) elements in the dual space VV^{*} and denote σk(L1,,Ln)\sigma_{k}(L_{1},\cdots,L_{n}) by PkP_{k}, 1kn1\leq k\leq n. Then we have 𝐁(L1,,Ln)=𝐁(P1,,Pk)\mathbf{B}(L_{1},\cdots,L_{n})=\mathbf{B}(P_{1},\cdots,P_{k}), where

𝐁(L1,,Ln){vV|Lk(v)=0, 1kn}\mathbf{B}(L_{1},\cdots,L_{n})\coloneqq\{v\in V\,|\,L_{k}(v)=0,\,1\leq k\leq n\}

and

𝐁(P1,,Pn){vV|Pk(v)=0, 1kn}.\mathbf{B}(P_{1},\cdots,P_{n})\coloneqq\{v\in V\,|\,P_{k}(v)=0,\,1\leq k\leq n\}.
Proof.

It is clear that 𝐁(L1,,Ln)\mathbf{B}(L_{1},\cdots,L_{n}) is contained in 𝐁(P1,,Pn)\mathbf{B}(P_{1},\cdots,P_{n}). So it remains to show the reverse inclusion.

First we claim that the set of common zeros of σk\sigma_{k}’s consists of only the origin. Indeed, set σ0=1\sigma_{0}=1. Then we have

i=1n(XXi)=i=0n(1)iσiXni.\prod_{i=1}^{n}(X-X_{i})=\sum_{i=0}^{n}(-1)^{i}\sigma_{i}X^{n-i}.

In particular, if x=(x1,,xn)\textbf{x}=(x_{1},\cdots,x_{n}) is a common zero of σk\sigma_{k}, the equality above implies

i=1n(Xxi)=Xn.\prod_{i=1}^{n}(X-x_{i})=X^{n}.

Then letting X=xiX=x_{i} shows that xin=0x_{i}^{n}=0 and hence xi=0x_{i}=0, 1in1\leq i\leq n. In other words, the point (0,,0)(0,\cdots,0) is the only common zero of σ1,,σn\sigma_{1},\cdots,\sigma_{n}.

Next we consider the natural linear map Φ:Vn\Phi:V\rightarrow\mathbb{C}^{n} defined as (L1,,Ln)(L_{1},\cdots,L_{n}). In particular, if we regard σk\sigma_{k} as a homogeneous polynomial of degree kk defined over n\mathbb{C}^{n}, then we get

𝐁(P1,,Pk)=𝐁(σ1Φ,,σnΦ).\mathbf{B}(P_{1},\cdots,P_{k})=\mathbf{B}(\sigma_{1}\circ\Phi,\cdots,\sigma_{n}\circ\Phi).

We have seen from above 𝐁(σ1,,σn)={(0,,0)}\mathbf{B}(\sigma_{1},\cdots,\sigma_{n})=\{(0,\cdots,0)\}. This yields

𝐁(P1,,Pn)=ker(Φ)=𝐁(L1,,Ln),\mathbf{B}(P_{1},\cdots,P_{n})=\ker(\Phi)=\mathbf{B}(L_{1},\cdots,L_{n}),

which finishes the proof. ∎

Now we are in the position to prove Theorem 1.3.

Proof of Theorem 1.3.

Assume to the contrary that 𝒮Xr0\mathcal{S}^{r}_{X}\not=0 and let s=(s1,,sr)𝒮Xr\textbf{s}=(s_{1},\cdots,s_{r})\in\mathcal{S}^{r}_{X} be a non-zero element. Let hkh_{k}, 1kr1\leq k\leq r, be the induced possibly singular Hermitian metric defined by sks_{k} on the dual tautological line bundle 𝒪(TX)(1)\mathscr{O}_{\mathbb{P}(TX)}(-1) over (TX)\mathbb{P}(TX). More precisely, for any v𝒪(TX)(1)v\in\mathcal{O}_{\mathbb{P}(TX)}(-1), we define the length of vv with respect to hh as following:

vhk|sk(vk)|1k,\|v\|_{h_{k}}\coloneqq|s_{k}(v^{k})|^{\frac{1}{k}},

where we regard skH0(X,SymkΩX1)s_{k}\in H^{0}(X,\mathrm{Sym}^{k}\Omega_{X}^{1}) as an element of H0((TX),𝒪(TX)(k))H^{0}(\mathbb{P}(TX),\mathcal{O}_{\mathbb{P}(TX)}(k)) with the canonical isomorphism

H0(X,SymkΩX1)H0((TX),𝒪(TX)(k))H^{0}(X,\mathrm{Sym}^{k}\Omega_{X}^{1})\cong H^{0}(\mathbb{P}(TX),\mathcal{O}_{\mathbb{P}(TX)}(k))

and vkv^{k} is viewed as a point contained in 𝒪(TX)(k)\mathcal{O}_{\mathbb{P}(TX)}(-k). Then hk=0h_{k}=0 if and only if sk=0s_{k}=0, and if sk0s_{k}\not=0, then the curvature current of hkh_{k} is non-positive as shown in Example 3.2.

Let gg be the canonical Kähler–Einstein metric on XX and denote by gg^{\prime} the induced Hermitian metric on 𝒪(TX)(1)\mathscr{O}_{\mathbb{P}(TX)}(-1). Then for any 1km1\leq k\leq m such that hk0h_{k}\not=0, by Theorem 3.3, there exist non-negative constants c1k,,cmkc_{1k},\cdots,c_{mk} such that for any vTXv\in TX that can be lifted to a vector vTiv^{\prime}\in T_{i} such that [v]𝒮i(Ω)(Ti)(TΩ)[v^{\prime}]\in\mathcal{S}^{i}(\Omega)\subset\mathbb{P}(T_{i})\subset\mathbb{P}(T\Omega), 1im1\leq i\leq m, we have

vhk,[v]=cikvg,[v].\|v\|_{h_{k},[v]}=c_{ik}\|v\|_{g^{\prime},[v]}.

The the result will follows directly from the following claim.

Claim 3.5.

There exist positive integers 1im1\leq i\leq m and 1kr1\leq k\leq r such that cik>0c_{ik}>0.

Proof of Claim 3.5.

We assume to the contrary that cik=0c_{ik}=0 for any ii and kk. Fix a point xXx\in X. By the definition of 𝒮X\mathcal{S}_{X}, there exist rr (maybe non-distinct) elements w1,,wrw_{1},\cdots,w_{r} contained in ΩX,x1=TxX\Omega^{1}_{X,x}=T^{*}_{x}X such that we have

sk(x)=(1)kσk(w1,,wr),s_{k}(x)=(-1)^{k}\sigma_{k}(w_{1},\cdots,w_{r}),

where σk\sigma_{k} is the kk-th elementary symmetric polynomial in nn variables. If cik=0c_{ik}=0, then for any element vTxXv\in T_{x}X that can be lifted to a vector vTiv^{\prime}\in T_{i} such that [v]𝒮i(Ω)[v^{\prime}]\in\mathcal{S}^{i}(\Omega), we always have sk(v)=0s_{k}(v)=0. In particular, for given ii, as cik=0c_{ik}=0 for any 1kr1\leq k\leq r, by Lemma 3.4, it follows that for any vTxXv\in T_{x}X that can be lifted to a vector vTiv^{\prime}\in T_{i} such that [v]𝒮i(ΩX)[v^{\prime}]\in\mathcal{S}^{i}(\Omega_{X}), we have

w1(v)==wr(v)=0.w_{1}(v)=\cdots=w_{r}(v)=0.

As a consequence, since the wkw_{k}’s are linear functionals over TxXT_{x}X, it follows that wkw_{k}’s vanish along the linear subspace of TxXT_{x}X spanned by the vectors vTxXv\in T_{x}X that can be lifted to vectors vTiv^{\prime}\in T_{i} such that [v]𝒮i(Ω)[v^{\prime}]\in\mathcal{S}^{i}(\Omega). On the other hand, since 𝒮i(Ω)(Ti)\mathcal{S}^{i}(\Omega)\subset\mathbb{P}(T_{i}) is pointwise linearly non-degenerate, it follows that the wkw_{k}’s vanish along the subspace of TxXT_{x}X generated by vectors vv which can be lifted to vTiv^{\prime}\in T_{i}. Thus the wkw_{k}’s vanish identically over TxXT_{x}X as ii is arbitrary. Then sk(x)=0s_{k}(x)=0 for all 1kn1\leq k\leq n and hence sk=0s_{k}=0 for any 1kr1\leq k\leq r, which is absurd. ∎

Now choose ii and kk such that cik0c_{ik}\not=0. Then for any vTxXv\in T_{x}X that can be lifted to vTiv^{\prime}\in T_{i} such that [v]𝒮i(Ω)[v^{\prime}]\in\mathcal{S}^{i}(\Omega), we have vhk=cikvg\|v\|_{h_{k}}=c_{ik}\|v\|_{g}. In particular, if v0v\not=0, then we have sk(v)0s_{k}(v)\not=0 and hence the curvature current of hkh_{k} at vv vanishes (see Example 3.2), which contradicts with Theorem 3.3. The proof of Theorem 1.3 is complete.

Remark 3.6.

As mentioned in §1, Theorem 1.3 does not hold for the rank-one case, i.e., compact quotient of complex balls. However, for the Kottwitz lattice ΓSU(n,1)\Gamma\subset\textup{SU}(n,1), B. Klingler proved in [Kli13, Theorem 1.11] that 𝒜Xr=0\mathcal{A}_{X}^{r}=0 for rn1r\leq n-1 if n+1n+1 is prime. So it should be interesting to ask whether the spectral base 𝒮Xr\mathcal{S}_{X}^{r} is trivial for any rr in this case.

4. Rigidity and integrality from spectral base perspective

The vanishing of the Hitchin base plays a significant role in understanding the rigidity and integrality of representations, as indicated in [GS92, Ara02, Kli13]. For further insights, see also [JZ97, CDY22] for generalizations to the quasi-projective situation.

Moreover, it is interesting to explore the relationship between symmetric differentials and the fundamental groups, as discussed in [BKT13]. As a consequence of the construction by Chen–Ngô, we observe that the Hitchin morphism factors through the spectral base, which allows us to strengthen many statements concerning the vanishing of the Hitchin base to apply to the vanishing of the spectral base as well.

4.1. Rigidity of the character variety

In this subsection, we will discuss the relationship between the rigidity of the character variety and the spectral base, as outlined in [Ara02].

Theorem 4.1 ([Ara02]).

If the Hitchin base 𝒜Xr=0\mathcal{A}^{r}_{X}=0, then GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} is rigid.

This method has been successfully used by B. Klingler [Kli13, Theorem 1.11] to study the rigidity problem of the Kottwitz lattice ΓSU(n,1)\Gamma\subset{\rm SU}(n,1) as mentioned in the introduction. However, on the one hand it is in general not an easy task to check the vanishing of 𝒜Xr\mathcal{A}_{X}^{r} in Theorem 4.1 on the other hand there are many examples of varieties with rigid character variety but having non-vanishing 𝒜Xr\mathcal{A}^{r}_{X} – see Examples 4.11 and 4.12 and [BDO11]. Arapura’s theorem above can be strengthened using the spectral base as follows.

Theorem 4.2.

The character variety GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} is rigid if and only if 𝒮X;Dolr=0\mathcal{S}^{r}_{X;\mathrm{Dol}}=0. In particular, if 𝒮Xr=0\mathcal{S}^{r}_{X}=0, then GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} is rigid.

Proof.

Note that GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} is an affine variety. So GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} is rigid if and only if it is compact. If 𝒮X;Dol=0\mathcal{S}_{X;\mathrm{Dol}}=0, then Dolr𝗌𝖽X1(0)\mathcal{M}_{\mathrm{Dol}}^{r}\subset\mathsf{sd}_{X}^{-1}(0) is compact by Theorem 2.4 and so is GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} by Theorem 2.3. Hence GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} is rigid.

Conversely, suppose that the character variety GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} is rigid. Then GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} only consists of a finite number of points. We assume to the contrary that there exists a Higgs bundle (,φ)Dolr(\mathscr{E},\varphi)\in\mathcal{M}_{\mathrm{Dol}}^{r} such that 𝗌𝖽X((,φ))0\mathsf{sd}_{X}((\mathscr{E},\varphi))\neq 0. Note that for any tt\in\mathbb{C}^{*}, the Higgs bundle (,tφ)(\mathscr{E},t\varphi) is again polystable and 𝗌𝖽X((,φ))𝗌𝖽X((,tφ))\mathsf{sd}_{X}((\mathscr{E},\varphi))\neq\mathsf{sd}_{X}((\mathscr{E},t\varphi)) for t1t\neq 1. Thus we obtain a non-trivial deformation family of Higgs bundles and then there exists a non-trivial deformation in GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} by the non-abelian Hodge correspondence (cf. Theorem 2.3), which contradicts the assumption that GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} is rigid. ∎

4.2. Harmonic Maps into Bruhat-Tits Buildings

In this subsection, we will briefly review the construction from harmonic maps into Bruhat-Tits buildings by Gromov–Schoen [GS92], as well as the construction of the spectral variety by means of harmonic maps. For more details, we refer to [Kli13, Appendix A], [DM21, CDY22, DM23].

Let FF be a non-Archimedean local field. For the group GLr(F)\mathrm{GL}_{r}(F), one can construct a Bruhat–Tits building, denoted as 𝒱\mathcal{V}. The Bruhat–Tits building is a contractible locally finite simplicial complex. Moreover, the group GLr(F)\mathrm{GL}_{r}(F) acts continuously on 𝒱\mathcal{V} by simplicial automorphisms, and the action is proper. The apartments of 𝒱\mathcal{V} are isomorphic to the Cartan subalgebra 𝔥\mathfrak{h}.

The Bruhat–Tits building is a metric space of non-positive curvature, and the theory of harmonic maps into metric spaces has been developed in [GS92, KS93]. Let FF be a non-Archimedean local field. A representation ρ:π1(X)GLr(F)\rho:\pi_{1}(X)\to\mathrm{GL}_{r}(F) is defined to be reductive if the Zariski closure of ρ(π1(X))\rho(\pi_{1}(X)) is a reductive subgroup of GLr(F)\mathrm{GL}_{r}(F).

Let X~\tilde{X} be the universal cover of XX. Given a reductive representation ρ:π1(X)GLr(F)\rho:\pi_{1}(X)\to\mathrm{GL}_{r}(F), by Gromov–Schoen [GS92], there exists a Lipschitz harmonic ρ\rho-equivariant map f:X~𝒱f:\tilde{X}\to\mathcal{V}. A point x~X~\tilde{x}\in\tilde{X} is called regular if there exists an apartment of 𝒱\mathcal{V} containing the image by ff of a neighborhood of x~\tilde{x}, and other points in X~\tilde{X} are called singular. For the covering map X~X\tilde{X}\to X, we denote by XregXX^{\mathrm{reg}}\subset X the image of the regular points and XsingX^{\mathrm{sing}} as the complement of XregX^{\mathrm{reg}} in XX. By [GS92, Theorem 6.4], the subspace XsingX^{\mathrm{sing}} has real Hausdorff codimension at least 22.

Let TT be the maximal FF-split torus of GLr(F)\mathrm{GL}_{r}(F) and let {h1,,hr}\{h_{1},\cdots,h_{r}\} be the root system of TT in GLr(F)\mathrm{GL}_{r}(F). For each apartment A𝒱A\subset\mathcal{V}, we can take the derivative of hih_{i}, obtaining rr real 1-forms {dh1,,dhr}\{dh_{1},\cdots,dh_{r}\} on the apartment 𝒱\mathcal{V}. Additionally, if UXU\subset X is an open set with f(U)Af(U)\subset A consisting of only regular points, we define ωi=f(dhi)1,0ΩX1\omega_{i}=f^{*}(dh_{i})^{1,0}\in\Omega_{X}^{1}. The harmonicity of ff implies that ωi\omega_{i} is holomorphic. Moreover, let A1A_{1} and A2A_{2} be two apartments. Then, over the intersection, the sets

{dh1,,dhr}|A1and{dh1,,dhr}|A2\{dh_{1},\cdots,dh_{r}\}|_{A_{1}}\quad\text{and}\quad\{dh_{1},\cdots,dh_{r}\}|_{A_{2}}

match up to permutation by the Weyl group action. Therefore, over the regular locus XregX^{\mathrm{reg}}, we obtain a multivalued holomorphic 1-form.

In analogy to the definition of the Hitchin morphism, we can define a map

(11) αρ:i=1r(Si𝔥)Wi=1rH0(Xreg,SymiΩXreg1),(h1,,hr)(σ1,,σr),\begin{split}\alpha_{\rho}&:\bigoplus_{i=1}^{r}(S^{i}\mathfrak{h}^{*})^{W}\to\bigoplus_{i=1}^{r}H^{0}(X^{\mathrm{reg}},\mathrm{Sym}^{i}\Omega_{X^{\mathrm{reg}}}^{1}),\\ &(h_{1},\cdots,h_{r})\mapsto(\sigma_{1},\cdots,\sigma_{r}),\end{split}

where σi\sigma_{i}’s are the symmetric polynomials taking values as (h1,,hr)(h_{1},\cdots,h_{r}). Based on the definition of the spectral base 𝒮Xreg\mathcal{S}_{X^{\mathrm{reg}}}, the image of αρ\alpha_{\rho} lies in the spectral base.

Moreover, since the singular set XsingX^{\mathrm{sing}} has Hausdorff codimension at least two and ff is a Lipschitz map, the sections σi\sigma_{i} uniquely extends to XX, and the extension of (σ1,,σr)(\sigma_{1},\cdots,\sigma_{r}) also lies in the spectral base 𝒮X\mathcal{S}_{X} by Lemma 2.6. Using the same notation, we write the extension map as

αρ:i=1r(Si𝔥)W𝒜Xr=i=1rH0(X,SymiΩX1).\alpha_{\rho}:\bigoplus_{i=1}^{r}(S^{i}\mathfrak{h}^{*})^{W}\to\mathcal{A}_{X}^{r}=\bigoplus_{i=1}^{r}H^{0}(X,\mathrm{Sym}^{i}\Omega_{X}^{1}).

Let GLr(F):={ρ|ρ:π1(X)GLr(F)}/\mathfrak{R}^{\mathrm{GL}_{r}(F)}:=\{\rho|\rho:\pi_{1}(X)\to\mathrm{GL}_{r}(F)\}/\sim be the character variety. The above construction allows us to define the Hitchin morphism 𝗌𝖽X;F\mathsf{sd}_{X;F} for non-Archimedean representations, which is defined as

(12) 𝗌𝖽X;F:GLr(F)𝒮X,𝗌𝖽X;F(ρ):=αρ.\begin{split}\mathsf{sd}_{X;F}:\mathfrak{R}^{\mathrm{GL}_{r}(F)}\to\mathcal{S}_{X},\;\mathsf{sd}_{X;F}(\rho):=\alpha_{\rho}.\end{split}

Recall that ρ:π1(X)GLr(F)\rho:\pi_{1}(X)\to\mathrm{GL}_{r}(F) is said to have a bounded image if ρ(π1(X))\rho(\pi_{1}(X)) is contained in a compact subgroup of GLr(F)\mathrm{GL}_{r}(F), where the topology of GLr(F)\mathrm{GL}_{r}(F) is defined by the topology of the local field FF.

Definition 4.3.

Let FF be a non-Archimedean local field. The non-Archimedean Dolbeault spectral base 𝒮X;DolNA;F\mathcal{S}_{X;\mathrm{Dol}}^{\mathrm{NA};F} is defined to be 𝒮X;DolNA;F:=𝗌𝖽X;F(GLr(F))𝒮X\mathcal{S}_{X;\mathrm{Dol}}^{\mathrm{NA};F}:=\mathsf{sd}_{X;F}(\mathfrak{R}^{\mathrm{GL}_{r}(F)})\subset\mathcal{S}_{X}.

For any reductive representation ρGLr(F)\rho\in\mathfrak{R}^{\mathrm{GL}_{r}(F)}, we write fρ:X~𝒱f_{\rho}:\tilde{X}\to\mathcal{V} for the corresponding harmonic map to the Bruhat-Tits building defined by GLr(F)\mathrm{GL}_{r}(F). We now state the following theorem, which is an analogue of Theorem 4.2.

Theorem 4.4.

For a non-Archimedean local field FF the non-Archimedean Dolbeault spectral base with respect to FF vanishes, 𝒮X;DolNA;F=0\mathcal{S}_{X;\mathrm{Dol}}^{\mathrm{NA};F}=0, if and only if every harmonic map fρf_{\rho} defined by ρGLr(F)\rho\in\mathfrak{R}^{\mathrm{GL}_{r}(F)} is a constant map. In particular, if 𝒮X=0\mathcal{S}_{X}=0, then every ρGLr(F)\rho\in\mathfrak{R}^{\mathrm{GL}_{r}(F)} has a bounded image.

Proof.

If 𝒮X;Dol=0\mathcal{S}_{X;\mathrm{Dol}}=0, then for αρ\alpha_{\rho} in (12), we have αρ=0\alpha_{\rho}=0, which implies that the harmonic map is a constant map. On the other side, if for every ρ\rho, fρf_{\rho} is a constant map, then based on the definition, αρ=0\alpha_{\rho}=0.

If 𝒮X=0\mathcal{S}_{X}=0, then fρf_{\rho} is a constant. As ρ(π1(X))\rho(\pi_{1}(X)) fixes the point f(X~)f(\tilde{X}), and GLr(F)\mathrm{GL}_{r}(F) acts properly on 𝒱\mathcal{V}, ρ(π1(X))\rho(\pi_{1}(X)) is bounded in GLr(F)\mathrm{GL}_{r}(F). ∎

It would be very interesting to know the relationship between the Dolbeault spectral base 𝒮X;Dol\mathcal{S}_{X;\mathrm{Dol}} and the non-Archimedean Dolbeault spectral base. Given an embedding σ:F\sigma:F\to\mathbb{C}, then σ\sigma induces a map σ:GLr(F)GLr()\sigma:\mathfrak{R}_{\mathrm{GL}^{r}(F)}\to\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})}. It will be very interesting to understand the following question:

Question 4.5.

Let FF be a non-Archimedean local field FF of characteristic zero and fix an embedding σ:F\sigma:F\to\mathbb{C}. For a representation ρGLr(F)\rho\in\mathfrak{R}^{\mathrm{GL}_{r}(F)}, do we have the following equality

𝗌𝖽X;F(ρ)=𝗌𝖽X(ξ1σρ)?\mathsf{sd}_{X;F}(\rho)=\mathsf{sd}_{X}(\xi^{-1}\circ\sigma\circ\rho)?

Here the map ξ1\xi^{-1} is the non-abelian Hodge correspondence map in (3) that maps the reductive representation σρ\sigma\circ\rho to a polystable Higgs bundle.

4.3. Applications

In this subsection, we will summarize previous results in [GS92, Ara02, Kli13] and rephrase them using the spectral base instead of the Hitchin base. The first one is a combination of Proposition 2.8 with [Ara02, Proposition 2.4] and [Kli13, Theorem 1.6].

Theorem 4.6.

Let XX be a projective manifold such that 𝒮Xr=0\mathcal{S}^{r}_{X}=0 for some r1r\geq 1. Then the following statements hold.

  1. (1)

    The character variety GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} is rigid.

  2. (2)

    Let FF be any non-Archimedean field. Then any reductive representation ρ:π1(X)GLr(F)\rho:\pi_{1}(X)\to\mathrm{GL}_{r}(F) has bounded image.

Proof.

(1) follows directly from Theorem 4.2 and (2) follows from Theorem 4.4. ∎

The second application is related to Simpson’s integrality conjecture.

Proposition 4.7 ([Sim92, Theorem 5] and [Kli13, Corollary 1.8]).

Let XX be a projective manifold such that 𝒮Xr=0\mathcal{S}^{r}_{X}=0 for some r1r\geq 1. Then any reductive representation ρ:π1(X)GLr()\rho:\pi_{1}(X)\to\mathrm{GL}_{r}(\mathbb{C}) is integral and it is a complex direct factor of a \mathbb{Z}-VHS.

Proof.

By Theorem 4.6, the character variety GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} is rigid and hence is zero-dimensional. In particular, as GLr()\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} is defined over \mathbb{Q}, there exists a number field KK such that the point [ρ]GLr()[\rho]\in\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})} is defined over KK. Then after replacing ρ\rho by some conjugation, we can assume that ρ\rho takes values in GLr(K)\mathrm{GL}_{r}(K). Let vv be an arbitrary finite place of KK. Then the induced representation ρv:π1(X)GLr(Kv)\rho_{v}:\pi_{1}(X)\rightarrow\mathrm{GL}_{r}(K_{v}), which is obtained from ρ\rho through the embedding KKvK\to K_{v}, is still reductive. So Theorem 4.6 implies that ρv\rho_{v} has bounded image in GLr(Kv)\mathrm{GL}_{r}(K_{v}). As vv is arbitrary, the image ρ(π1(X))\rho(\pi_{1}(X)) lies in GLr(𝒪K)\mathrm{GL}_{r}(\mathcal{O}_{K}).

Finally, since ρ(π1(X))\rho(\pi_{1}(X)) is contained in GLr(𝒪K)\mathrm{GL}_{r}(\mathcal{O}_{K}), the traces Tr(ρ(γ))\mathrm{Tr}(\rho(\gamma)) are algebraic integers for any γπ1(X)\gamma\in\pi_{1}(X). So it follows from [Sim92, Theorem 5] and the discussion in the paragraph before [Sim92, Corollary 4.9] that ρ\rho is a complex direct fact of a \mathbb{Z}-VHS. ∎

Remark 4.8.

We have learned from a talk by H. Esnault that the argument of [EG18] can be applied to show that if dim(GLr())=0\dim(\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})})=0, then any representation ρ:π1(X)GLr()\rho:\pi_{1}(X)\rightarrow\mathrm{GL}_{r}(\mathbb{C}) is integral.

Proof of Theorem 1.2.

It follows from Theorem 4.6 and Proposition 4.7. ∎

H. Esnault asked whether a projective manifold with infinite fundamental group must have a non-zero symmetric differential. This question was answered by Y. Brunebarbe, B. Klingler and B. Totaro in [BKT13] for its linear version. As the last application, we obtain the following variation of [BKT13] following the same argument there.

Theorem 4.9 (Variation of [BKT13, Theorem 6.1]).

Let XX be a projective manifold and let KK be a field of characteristic zero. If there is a linear representation ρ:π1(X)GLr(K)\rho:\pi_{1}(X)\to\mathrm{GL}_{r}(K) such that the image is infinite, then one of the following statements hold.

  1. (1)

    𝒮Xk0\mathcal{S}^{k}_{X}\neq 0 for some k1k\geq 1.

  2. (2)

    The semi-simplification ρs\rho^{\mathrm{s}} of ρ\rho is a complex direct factor of a \mathbb{Z}-VHS with infinite discrete monodromy group.

Proof.

Let ρs:π1(X)GLr(K)\rho^{\mathrm{s}}:\pi_{1}(X)\to\mathrm{GL}_{r}(K) be the semi-simplification of ρ\rho. Assume that the representation ρs\rho^{\mathrm{s}} has finite image. Then there exists a finite étale covering π:XX\pi:X^{\prime}\rightarrow X such that H0(X,ΩX1)0H^{0}(X^{\prime},\Omega_{X^{\prime}}^{1})\not=0 by the same argument of the proof of [BKT13, Theorem 6.1]. Choose a non-zero holomorphic 11-form αH0(X,ΩX1)\alpha\in H^{0}(X^{\prime},\Omega_{X^{\prime}}^{1}). Then we can define a Higgs bundle on XX as following:

φ:π𝒪X×απ(𝒪XΩX1)π𝒪XΩX1ΩX1,\varphi:\mathscr{E}\coloneqq\pi_{*}\mathcal{O}_{X^{\prime}}\xrightarrow{\times\alpha}\pi_{*}(\mathcal{O}_{X^{\prime}}\otimes\Omega_{X^{\prime}}^{1})\xrightarrow{\cong}\pi_{*}\mathcal{O}_{X^{\prime}}\otimes\Omega_{X}^{1}\eqqcolon\mathscr{E}\otimes\Omega_{X}^{1},

where the last isomorphism follows from the projection formula and the fact πΩX1ΩX1\pi^{*}\Omega_{X}^{1}\cong\Omega_{X^{\prime}}^{1} as π\pi is étale. One can easily see that the Higgs field φ\varphi is non-zero and so s:=𝗌𝖽X([,φ])\textbf{s}:=\mathsf{sd}_{X}([\mathscr{E},\varphi]) is a non-zero element in 𝒮Xk\mathcal{S}^{k}_{X}, where k=deg(π)k=\deg(\pi).

Now we assume that 𝒮Xk=0\mathcal{S}^{k}_{X}=0 for any k1k\geq 1 and the representation ρs\rho^{\mathrm{s}} has infinite image. Then, by Theorem 4.2, the representation ρs\rho^{\mathrm{s}} is rigid and then we can conclude by Proposition 4.7 that (2) in the statement of Theorem 4.9 holds.

Remark 4.10.
  1. (1)

    By the Grothendieck–Riemann–Roch theorem, since π:XX\pi:X^{\prime}\rightarrow X is a finite étale morphism, the Higgs bundle =π𝒪X\mathscr{E}=\pi_{*}\mathcal{O}_{X^{\prime}} is actually topologically trivial. So the non-zero spectral datum s is contained in 𝒮X;Dolr\mathcal{S}_{X;\mathrm{Dol}}^{r},

  2. (2)

    We briefly recall the argument of [BKT13] to show the existence of non-zero symmetric forms in the second case of Theorem 4.9. After replacing XX by a finite étale covering XX^{\prime}, we may assume that the monodromy group Γ\Gamma is torsion-free. Let YY be a resolution of the image of the period map Φ:XD/Γ\Phi:X^{\prime}\rightarrow D/\Gamma. Then YY is positive-dimensional as Γ\Gamma is infinite. Let ZXZ\rightarrow X be a resolution of the rational map XYX^{\prime}\dashrightarrow Y. Note that the cotangent bundle ΩY1\Omega_{Y}^{1} is big by [BKT13, Corollary 3.2]. This implies that in particular YY has non-zero symmetric forms and then it also induces non-zero symmetric forms on ZZ, which naturally descends to XX^{\prime}. As π:XX\pi:X^{\prime}\rightarrow X is étale, we have the following commutative diagram

    (TX){\mathbb{P}(T_{X^{\prime}})}(πTX){\mathbb{P}(\pi^{*}T_{X})}(TX){\mathbb{P}(T_{X})}X{X^{\prime}}X{X^{\prime}}X.{X.}\scriptstyle{\cong}π¯\scriptstyle{\bar{\pi}}=\scriptstyle{=}π\scriptstyle{\pi}

    Recall that π¯𝒪(TX)(1)𝒪(TX)(1)\bar{\pi}^{*}\mathcal{O}_{\mathbb{P}(T_{X})}(1)\cong\mathcal{O}_{\mathbb{P}(T_{X^{\prime}})}(1) and the Iitaka dimension is preserved under finite morphisms. Hence, using the following two natural identifications for any k0k\geq 0

    H0((TX),𝒪(TX)(k))H0(X,SymkΩX1)H^{0}(\mathbb{P}(T_{X^{\prime}}),\mathcal{O}_{\mathbb{P}(T_{X^{\prime}})}(k))\cong H^{0}(X^{\prime},\mathrm{Sym}^{k}\Omega_{X^{\prime}}^{1})

    and

    H0((TX),𝒪(TX)(k))H0(X,SymkΩX1),H^{0}(\mathbb{P}(T_{X}),\mathcal{O}_{\mathbb{P}(T_{X})}(k))\cong H^{0}(X,\mathrm{Sym}^{k}\Omega_{X}^{1}),

    one can derive that the existence of non-zero symmetric forms on XX^{\prime} also yields the existence of non-zero symmetric forms on XX.

  3. (3)

    One cannot expect to derive that 𝒮Xk0\mathcal{S}_{X}^{k}\not=0 for some k1k\geq 1 in the second case of Theorem 4.9 — see Theorem 1.3 and Example 4.12.

Proof of Corollary 1.4.

This follows directly from Theorem 1.3 and Theorem 4.6. ∎

Proof of Corollary 1.5.

This follows directly from the vanishing of the spectral base. ∎

4.4. Examples

In this subsection, we summarise the relations between the various vanishing of symmetric differentials and the rigidity/integrality/finiteness of representations in the following diagram.

𝒮Xr=0{\mathcal{S}_{X}^{r}=0}𝒮X;Dolr=0{\mathcal{S}_{X;\mathrm{Dol}}^{r}=0}𝒜Xr=0{\mathcal{A}_{X}^{r}=0}dim(GLr())=0{\dim(\mathfrak{R}^{\mathrm{GL}_{r}(\mathbb{C})})=0}Rigr(X){\textup{Rig}^{r}(X)}Intr(X){\textup{Int}^{r}(X)}Fin(X){\textup{Fin}(X)}FinLin(X){\textup{FinLin}(X)}𝒮X=0{\mathcal{S}_{X}=0}𝒜X=0{\mathcal{A}_{X}=0}Thm. 1.2[Ara02][Kli13]Thm. 4.2[EG18]Conj.?\scriptstyle{?}[BKT13]Conj.

Here Rigr(X)\textup{Rig}^{r}(X) means that every representation ρ:π1(X)GLr()\rho:\pi_{1}(X)\rightarrow\mathrm{GL}_{r}(\mathbb{C}) is rigid and Intr(X)\textup{Int}^{r}(X) means that every representation ρ:π1(X)GLr()\rho:\pi_{1}(X)\rightarrow\mathrm{GL}_{r}(\mathbb{C}) is integral. The notation Fin(X)\textup{Fin}(X) means that the fundamental group π1(X)\pi_{1}(X) is finite and FinLin(X)\textup{FinLin}(X) says that all the linear representations of π1(X)\pi_{1}(X) have finite image. Moreover, the condition 𝒮X=0\mathcal{S}_{X}=0 (resp. 𝒜X=0\mathcal{A}_{X}=0) just means that 𝒮Xr=0\mathcal{S}_{X}^{r}=0 (resp. 𝒜Xr=0\mathcal{A}_{X}^{r}=0) for all r1r\geq 1.

Example 4.11 ([BDO11, p. 1092, Example]).

There exists a simply connected smooth projective threefold XX such that 𝒮X20\mathcal{S}_{X}^{2}\not=0. In particular, the natural inclusion 𝒮X;Dol2𝒮X2\mathcal{S}_{X;\mathrm{Dol}}^{2}\subset\mathcal{S}_{X}^{2} is strict in this case.

Example 4.12.

Let X=Ω/ΓX=\Omega/\Gamma be a quotient of bounded symmetric domain of rank 2\geq 2 by an irreducible torsion-free cocompact lattice as in Theorem 1.3. Then we have 𝒮Xr=0\mathcal{S}_{X}^{r}=0 by Theorem 1.3 and dim(𝒜Xr)0\dim(\mathcal{A}_{X}^{r})\gg 0 for rr sufficiently large. In particular, the natural inclusion 𝒮Xr𝒜Xr\mathcal{S}_{X}^{r}\subset\mathcal{A}_{X}^{r} is strict for r0r\gg 0. Moreover, as Γ\Gamma is infinite, it follows that 𝒮X=0\mathcal{S}_{X}=0 does not imply FinLin(X)\textup{FinLin}(X); that is, the conclusion of Theorem 4.9 cannot be improved to 𝒮X0\mathcal{S}_{X}\not=0.

References

  • [Ara02] Donu Arapura. Higgs bundles, integrability, and holomorphic forms. Motives, polylogarithms and Hodge theory, Part II (Irvine, CA, 1998), 3:605–624, 2002.
  • [BDO11] Fedor Bogomolov and Bruno De Oliveira. Symmetric differentials of rank 1 and holomorphic maps. Pure Appl. Math. Q., 7(4, Special Issue: In memory of Eckart Viehweg):1085–1103, 2011.
  • [BKT13] Yohan Brunebarbe, Bruno Klingler, and Burt Totaro. Symmetric differentials and the fundamental group. Duke Math. J., 162(14):2797–2813, 2013.
  • [BKU23] Barbara Bolognese, Alex Küronya, and Martin Ulirsch. P=WP=W phenomena on abelian varieties. arXiv preprint arXiv:2303.03734, 2023.
  • [BS94] Shigetoshi Bando and Yum-Tong Siu. Stable sheaves and Einstein-Hermitian metrics. In Geometry and analysis on complex manifolds, pages 39–50. World Sci. Publ., River Edge, NJ, 1994.
  • [BS09] Indranil Biswas and Georg Schumacher. Yang-Mills equation for stable Higgs sheaves. Internat. J. Math., 20(5):541–556, 2009.
  • [CDY22] Benoit Cadorel, Ya Deng, and Katsutoshi Yamanoi. Hyperbolicity and fundamental groups of complex quasi-projective varieties. arXiv preprint arXiv:2212.12225, 2022.
  • [CN20] Tsao-Hsien Chen and Bao Châu Ngô. On the Hitchin morphism for higher-dimensional varieties. Duke Math. J., 169(10):1971–2004, 2020.
  • [Cor88] Kevin Corlette. Flat GG-bundles with canonical metrics. J. Differential Geom., 28(3):361–382, 1988.
  • [DM21] Georgios Daskalopoulos and Chikako Mese. Uniqueness of equivariant harmonic maps to symmetric spaces and buildings. arXiv preprint arXiv:2111.11422, 2021.
  • [DM23] Georgios Daskalopoulos and Chikako Mese. Notes on harmonic maps. arXiv preprint arXiv:2301.04190, 2023.
  • [Don87] Simon Kirwan Donaldson. Twisted harmonic maps and the self-duality equations. Proc. London Math. Soc. (3), 55(1):127–131, 1987.
  • [EG18] Hélène Esnault and Michael Groechenig. Cohomologically rigid local systems and integrality. Selecta Math. (N.S.), 24(5):4279–4292, 2018.
  • [Eys04] Philippe Eyssidieux. Sur la convexité holomorphe des revêtements linéaires réductifs d’une variété projective algébrique complexe. Invent. Math., 156(3):503–564, 2004.
  • [GRR15] Alberto García-Raboso and Steven Rayan. Introduction to nonabelian hodge theory: flat connections, higgs bundles and complex variations of hodge structure. Calabi-Yau Varieties: Arithmetic, Geometry and Physics: Lecture Notes on Concentrated Graduate Courses, pages 131–171, 2015.
  • [GS92] Mikhail Gromov and Richard M. Schoen. Harmonic maps into singular spaces and pp-adic superrigidity for lattices in groups of rank one. Inst. Hautes Études Sci. Publ. Math., (76):165–246, 1992.
  • [He20] Siqi He. The behavior of sequences of solutions to the Hitchin-Simpson equations. arXiv preprint arXiv:2002.08109, 2020.
  • [Hit87a] Nigel J. Hitchin. The self-duality equations on a Riemann surface. Proc. London Math. Soc. (3), 55(1):59–126, 1987.
  • [Hit87b] Nigel J. Hitchin. Stable bundles and integrable systems. Duke Math. J., 54(1):91–114, 1987.
  • [Hit92] Nigel J. Hitchin. Lie groups and Teichmüller space. Topology, 31(3):449–473, 1992.
  • [HL23] Siqi He and Jie Liu. On the spectral variety for rank two Higgs bundles. arXiv preprint arXiv:2310.18934, 2023.
  • [JZ97] Jürgen Jost and Kang Zuo. Harmonic maps of infinite energy and rigidity results for representations of fundamental groups of quasiprojective varieties. J. Differential Geom., 47(3):469–503, 1997.
  • [Kat97] Ludmil Katzarkov. On the Shafarevich maps. In Algebraic geometry—Santa Cruz 1995, volume 62, Part 2 of Proc. Sympos. Pure Math., pages 173–216. Amer. Math. Soc., Providence, RI, 1997.
  • [Kli13] Bruno Klingler. Symmetric differentials, Kähler groups and ball quotients. Invent. Math., 192(2):257–286, 2013.
  • [KS93] Nicholas J. Korevaar and Richard M. Schoen. Sobolev spaces and harmonic maps for metric space targets. Comm. Anal. Geom., 1(3-4):561–659, 1993.
  • [LZZ17] Jiayu Li, Chuanjing Zhang, and Xi Zhang. Semi-stable Higgs sheaves and Bogomolov type inequality. Calc. Var. Partial Differential Equations, 56(3):Paper No. 81, 33, 2017.
  • [Mar91] Grigorii Aleksandrovich Margulis. Discrete subgroups of semisimple Lie groups, volume 17 of Ergebnisse der Mathematik und ihrer Grenzgebiete (3). Springer-Verlag, Berlin, 1991.
  • [Mok89] Ngaiming Mok. Metric rigidity theorems on Hermitian locally symmetric manifolds, volume 6 of Series in Pure Mathematics. World Scientific Publishing Co., Inc., Teaneck, NJ, 1989.
  • [Mok02] Ngaiming Mok. Characterization of certain holomorphic geodesic cycles on quotients of bounded symmetric domains in terms of tangent subspaces. Compositio Math., 132(3):289–309, 2002.
  • [Mok04] Ngaiming Mok. Extremal bounded holomorphic functions and an embedding theorem for arithmetic varieties of rank 2\geq 2. Invent. Math., 158(1):1–31, 2004.
  • [OSS11] Christian Okonek, Michael Schneider, and Heinz Spindler. Vector bundles on complex projective spaces. Modern Birkhäuser Classics. Birkhäuser/Springer Basel AG, Basel, 2011. Corrected reprint of the 1988 edition, With an appendix by S. I. Gelfand.
  • [Sch18] L Schaposnik. An introduction to spectral data for higgs bundles. The geometry, topology and physics of moduli spaces of Higgs bundles, pages 65–101, 2018.
  • [Ser65] Jean-Pierre Serre. Algèbre locale. Multiplicités, volume 11 of Lecture Notes in Mathematics. Springer-Verlag, Berlin-New York, 1965. Cours au Collège de France, 1957–1958, rédigé par Pierre Gabriel, Seconde édition, 1965.
  • [Sev59] Francesco Severi. La géométrie algébrique italienne. sa rigueur, ses méthodes, ses problèmes. Matematika, 3(1):111–142, 1959.
  • [Sim88] Carlos T. Simpson. Constructing variations of Hodge structure using Yang-Mills theory and applications to uniformization. J. Amer. Math. Soc., 1(4):867–918, 1988.
  • [Sim91] Carlos T. Simpson. The ubiquity of variations of Hodge structure. In Complex geometry and Lie theory (Sundance, UT, 1989), volume 53 of Proc. Sympos. Pure Math., pages 329–348. Amer. Math. Soc., Providence, RI, 1991.
  • [Sim92] Carlos T. Simpson. Higgs bundles and local systems. Inst. Hautes Études Sci. Publ. Math., (75):5–95, 1992.
  • [Sim94a] Carlos T. Simpson. Moduli of representations of the fundamental group of a smooth projective variety. I. Inst. Hautes Études Sci. Publ. Math., (79):47–129, 1994.
  • [Sim94b] Carlos T. Simpson. Moduli of representations of the fundamental group of a smooth projective variety. II. Inst. Hautes Études Sci. Publ. Math., (80):5–79, 1994.
  • [Sim97] Carlos Simpson. The Hodge filtration on nonabelian cohomology. In Algebraic geometry—Santa Cruz 1995, volume 62 of Proc. Sympos. Pure Math., pages 217–281. Amer. Math. Soc., Providence, RI, 1997.
  • [SS24] Lei Song and Hao Sun. On the image of Hitchin morphism for algebraic surfaces: the case GLn\textup{GL}_{n}. Int. Math. Res. Not. IMRN, (1):492–514, 2024.
  • [Wen16] Richard A. Wentworth. Higgs bundles and local systems on Riemann surfaces. In Geometry and quantization of moduli spaces, Adv. Courses Math. CRM Barcelona, pages 165–219. Birkhäuser/Springer, Cham, 2016.
  • [Zuo96] Kang Zuo. Kodaira dimension and Chern hyperbolicity of the Shafarevich maps for representations of π1\pi_{1} of compact Kähler manifolds. J. Reine Angew. Math., 472:139–156, 1996.