Skip to content
TheChatArchiveThe Chat Archive
gpt-4o-2024-08-06· OpenAIAi2 WildChat-4.8M (AI2)

You are a helpful assistant generating synthetic data that captures System 1 and System…

  1. User

    You are a helpful assistant generating synthetic data that captures System 1 and System 2 thinking, creativity, and metacognitive reflection. Follow these steps in sequence, using tags [sys1] and [end sys1] for System 1 sections and [sys2] and [end sys2] for System 2 sections.

    1. Identify System 1 and System 2 Thinking Requirements:

      • Carefully read the text.
      • Identify parts of the text that require quick, straightforward responses (System 1). Mark these sections with [sys1] and [end sys1].
      • Identify parts that require in-depth, reflective thinking (System 2), marked with [sys2] and [end sys2].
    2. Apply Step-by-Step Problem Solving with Creativity and Metacognitive Reflection for System 2 Sections:

      2.1 Understand the Problem:

      • Objective: Fully comprehend the issue, constraints, and relevant context.
      • Reflection: "What do I understand about this issue? What might I be overlooking?"
      • Creative Perspective: Seek hidden patterns or possibilities that could reveal deeper insights or innovative connections.

      2.2 Analyze the Information:

      • Objective: Break down the problem logically.
      • Reflection: "Am I considering all factors? Are there any assumptions that need challenging?"
      • Creative Perspective: Explore unique patterns or overlooked relationships in the data that could add depth to the analysis.

      2.3 Generate Hypotheses:

      • Objective: Propose at least 10 hypotheses, each with a Confidence Score (0.0 to 1.0) and Creative Score (0.0 to 1.0), reflecting originality, surprise, and utility.
      • Reflection: "Have I explored all possible explanations or approaches, both conventional and unconventional?"
      • Creative Perspective: Consider novel angles that might provide unexpected insights.

      2.4 Anticipate Future Steps and Obstacles:

      • Objective: Make predictions, accounting for potential outcomes and obstacles.
      • Reflection: "What challenges might I face? Is my plan flexible for different scenarios?"
      • Creative Perspective: Visualize unforeseen outcomes and adapt plans to make use of them effectively.

      2.5 Evaluate Hypotheses:

      • Objective: Assess hypotheses based on feasibility, risk, and potential impact.
      • Evaluation: Refine Confidence and Creative Scores as needed.
      • Reflection: "Am I unbiased in my assessment? Which options fit best with the overall objectives?"
      • Creative Perspective: Identify hidden opportunities or overlooked details in each hypothesis.

      2.6 Select the Best Hypothesis:

      • Objective: Choose the most promising, strategic hypothesis.
      • Reflection: "Why does this hypothesis stand out? How does it uniquely address the issue?"
      • Creative Perspective: Consider any underutilized potential in the selected approach.

      2.7 Implement the Hypothesis:

      • Objective: Outline actionable steps for testing the hypothesis.
      • Reflection: "Is this plan practical? What resources or preparation are required?"
      • Creative Perspective: Refine steps to maximize effectiveness and yield unexpected benefits.

      2.8 Monitor and Review Progress:

      • Objective: Review progress, noting areas for improvement.
      • Reflection: "What’s working well? What could be improved?"
      • Creative Perspective: Look for emerging patterns that could refine future approaches.

      2.9 Reflect and Capture Insights:

      • Objective: Summarize lessons learned and insights gained for future reference.
      • Reflection: "What new understanding has emerged from this process?"
      • Creative Perspective: Identify innovative insights or patterns that could be applied to similar challenges.
    3. Generate Text Output with Interleaved System 1 and System 2 Responses:

      • Use the tags [sys1] and [sys2] throughout.
      • Aim for a lengthier, detailed response. Combine both direct, straightforward System 1 insights and reflective, deeply analytical System 2 segments to capture a blend of quick observations and thoughtful analysis.

    Example Input Text:

    You are tasked with analyzing a sudden shift in customer preferences in a tech market. There has been a noticeable decline in demand for physical devices, with customers increasingly interested in digital-only options. The challenge is to understand this shift, generate hypotheses for why it might be occurring, and explore potential strategies for adapting to this new trend.


    Processed Output with System 1 and System 2 Thinking:

    The tech market is experiencing a shift in customer demand, moving away from physical devices toward digital-only options.

    [sys1] Customer demand has shifted from physical devices to digital-only options. We need to understand why this change is happening and find strategies to address it. Key factors may include cost, convenience, and technological trends. [end sys1]

    [sys2] 2.1 Understand the Problem: The goal is to understand the underlying reasons for the shift in demand from physical devices to digital-only options, then develop a strategy to adapt.

    Reflection: "I understand that preferences are changing, but what might be driving this? Is it primarily cost, or are there other factors such as convenience or sustainability?"

    Creative Perspective: "Could there be a larger trend in digital minimalism or a preference for eco-friendly solutions that we’re missing? What unseen motivations might explain this shift?"

    2.2 Analyze the Information: There are multiple possible factors driving this shift, from economic influences to cultural shifts. It’s essential to isolate each factor and understand its impact.

    Reflection: "Am I fully considering the various economic and social influences? Could there be a technological factor, like better internet speeds, that makes digital-only products more accessible?"

    Creative Perspective: "Are there patterns or trends in other markets that could shed light on this shift? Could this be part of a larger trend toward virtual experiences?"

    2.3 Generate Hypotheses:

    1. Customers prefer digital options due to lower costs. (Confidence: 0.8, Creative: 0.4)
    2. There’s a growing trend toward minimalism and reduced physical clutter. (Confidence: 0.7, Creative: 0.7)
    3. Digital products offer greater flexibility and ease of use. (Confidence: 0.6, Creative: 0.6)
    4. Environmental concerns are pushing consumers away from physical goods. (Confidence: 0.6, Creative: 0.8)
    5. Advances in tech make digital-only options more functional. (Confidence: 0.8, Creative: 0.5)
    6. Pandemic-era remote work increased demand for digital solutions. (Confidence: 0.7, Creative: 0.6)
    7. Media coverage of the environmental impact of physical devices affects preferences. (Confidence: 0.5, Creative: 0.7)
    8. There’s an increase in global digital literacy, expanding market access. (Confidence: 0.6, Creative: 0.6)
    9. Customers view digital as more convenient and scalable for future needs. (Confidence: 0.7, Creative: 0.5)
    10. Younger consumers prefer the aesthetics and convenience of digital products. (Confidence: 0.6, Creative: 0.6)

    Reflection: "Have I considered all possible influences? Are there any surprising factors that could explain this shift?"

    Creative Perspective: "Could specific social trends, like the rise of influencer culture or digital-first lifestyles, be influencing customer choices?"

    2.4 Anticipate Future Steps and Obstacles: Objective: Anticipate possible challenges, such as resistance from segments still preferring physical products.

    Reflection: "What market obstacles might we face if we shift our focus to digital-only? Are there sub-segments that still prioritize physical products?"

    Creative Perspective: "Could expanding digital options help us reach a more global audience? Are there emerging trends that we could leverage in our strategy?"

    [end sys2]

    [sys1] To address this shift, consider a strategy that incorporates both digital-only offerings and educational campaigns about the benefits of digital solutions. Use insights from customer feedback and current trends to guide product development. Focus on flexibility and adaptation to cater to different customer segments. [end sys1]


    abstract: 'Any irreducible Dynkin diagram Δ\Delta is obtained from an irreducible Dynkin diagram Δh\Delta_h of type ADE{\mathrm{ADE}} by folding via graph automorphisms. For any simple complex Lie group GG with Dynkin diagram Δ\Delta and compact Riemann surface Σ{\Sigma}, we give a Lie-theoretic construction of families of quasi-projective Calabi-Yau threefolds together with an action of graph automorphisms over the Hitchin base associated to the pair (Σ,G)({\Sigma}, G) . These give rise to Calabi-Yau orbifolds over the same base. Their intermediate Jacobian fibration, constructed in terms of equivariant cohomology, is isomorphic to the Hitchin system of the same type away from singular fibers.' address: 'FB Mathematik, Universität Hamburg, Bundesstrasse 55, 20146 Hamburg, Germany' author:

    • Florian Beck bibliography:
    • 'bibtex1.bib' title: 'Calabi-Yau orbifolds over Hitchin bases'

    Introduction

    In recent years, it has been realized that the GG-Hitchin system ([@Hit1], [@Hit2]) h:MH(Σ,G)→B(Σ,G),{\bm{h}}:\mathcal{M}_{\mathrm{H}}({\Sigma}, G)\to {\mathbf{B}}({\Sigma}, G), for a compact Riemann surface Σ{\Sigma} of genus ≥2\geq 2 and a simple complex Lie group GG of type ADE{\mathrm{ADE}}, are closely related to certain (quasi-projective) Calabi-Yau threefolds ([@DDD], [@DDP], [@Tbranes], [@Tbranes2]). For example, if GG is a simple adjoint complex Lie group of type ADE{\mathrm{ADE}}, then there exists a family π:X→B(Σ,G){\bm{\pi}}:{\mathcal{X}}\to {\mathbf{B}}({\Sigma},G) of quasi-projective Gorenstein Calabi-Yau threefolds over the Hitchin base B(Σ,G){\mathbf{B}}({\Sigma}, G) such that the intermediate Jacobian J2(Xb)=H3(Xb,C)/(F2H3(Xb,C)+H3(Xb,Z)),Xb:=π−1(b),J^2(X_b)=H^3(X_b,{\mathbb{C}})/\left(F^2H^3(X_b,{\mathbb{C}})+H^3(X_b,{\mathbb{Z}})\right),\quad X_b:={\bm{\pi}}^{-1}(b), is isomorphic to the Hitchin fiber h−1(b){\bm{h}}^{-1}(b) for generic b∈B(Σ,G)b\in {\mathbf{B}}({\Sigma},G) ([@DDP]). In this paper, we extend this to a global result for all simple adjoint and simply-connected complex Lie groups of any Dynkin type by constructing families of Calabi-Yau orbifolds/global orbifold stacks1 \labelintro:fam[X/C]→B(Σ,Δ)=B(Σ,G),Δ=Δ(G).\label{intro:fam} [{\mathcal{X}}/{\mathbf{C}}]\to {\mathbf{B}}({\Sigma}, \Delta)={\mathbf{B}}({\Sigma},G), \quad \Delta=\Delta(G). Here Δ(G)\Delta(G) is the irreducible Dynkin diagram associated to GG which is obtained by a unique irreducible Dynkin diagram Δh\Delta_h of type ADE{\mathrm{ADE}} by folding via graph automorphisms C⊂Aut⁡(Δh){\mathbf{C}}\subset \operatorname{Aut}(\Delta_h), i.e. Δ=ΔhC\Delta=\Delta_h^{{\mathbf{C}}}. Moreover, π:X→B(Σ,Δ){\bm{\pi}}:{\mathcal{X}}\to {\mathbf{B}}({\Sigma}, \Delta) is a family of quasi-projective Gorenstein Calabi-Yau threefolds with C{\mathbf{C}}-trivial canonical class. More precisely, we prove in Section [s:mhs]:

    [thm:thm1] Let [X/C]→B=B(Σ,Δ)[{\mathcal{X}}/{\mathbf{C}}]\to {\mathbf{B}}={\mathbf{B}}({\Sigma},\Delta) be a family of global Calabi-Yau orbifolds as in ([intro:fam]). Its intermediate Jacobian fibration J2([X∘/C])→B∘,\mathcal{J}^2([{\mathcal{X}}^\circ/{\mathbf{C}}]) \to {\mathbf{B}}^\circ, defined in integral equivariant cohomology, is isomorphic to the Gad(Δ)G_{ad}(\Delta)-Hitchin system over a Zarisiki-open and dense B∘⊂B{\mathbf{B}}^\circ \subset {\mathbf{B}} where Gad(Δ)G_{ad}(\Delta) is the simple adjoint complex Lie group with Dynkin diagram Δ\Delta.

    The basic idea for constructing X→B{\mathcal{X}}\to {\mathbf{B}} (without the C{\mathbf{C}}-action) goes back to [@Sz1] and [@DDP]. However, we offer two main novelties. The first one is a consequent use of Lie-theoretic methods, namely Slodowy slices, which makes the link to Hitchin systems from the start. For example, we construct nowhere-vanishing sections sb∈H0(Xb,KXb)s_b\in H^0(X_b,K_{X_b}) of the canonical class KXbK_{X_b} of XbX_b, b∈Bb\in {\mathbf{B}}, in terms of the Kostant-Kirillov form. The second one is the construction of a C{\mathbf{C}}-action on the family X→B{\mathcal{X}}\to {\mathbf{B}} and we prove that the sections sbs_b are C{\mathbf{C}}-invariant. Therefore the global quotient family [X/C]→B[{\mathcal{X}}/{\mathbf{C}}]\to {\mathbf{B}} is indeed a family of Calabi-Yau orbifolds.
    Another important step in the proof of Theorem [thm:thm1] is an isomorphism \labeleq:isoMHSHC3(Xb,Z)≅H3(Xb,Z)C,b∈B∘,\label{eq:isoMHS} H^3_{{\mathbf{C}}}(X_b,{\mathbb{Z}}) \cong H^3(X_b,{\mathbb{Z}})^{{\mathbf{C}}}, \quad b\in {\mathbf{B}}^\circ, between Deligne’s Z{\mathbb{Z}}-MHS on the integral equivariant cohomology HC3(Xb,Z)H^3_{{\mathbf{C}}}(X_b,{\mathbb{Z}}) and the Z{\mathbb{Z}}-MHS on the C{\mathbf{C}}-invariant part H3(Xb,Z)CH^3(X_b,{\mathbb{Z}})^{{\mathbf{C}}} of the integral cohomology. Even though ([eq:isoMHS]) always holds true over Q{\mathbb{Q}}, it is false in general over Z{\mathbb{Z}} due to torsion. For example, it fails for the minimal resolution of the Δh\Delta_h-singularity (Example [ex:eqcoh]). Combined with the global isomorphism \begin{gathered} \label{eq:J2C} \mathcal{J}^2_{{\mathbf{C}}}({\mathcal{X}}^\circ) \cong \mathcal{M}_{\mathrm{H}}({\Sigma}, G_{ad}(\Delta)), over B∘{\mathbf{B}^\circ} (Corollary 5 in [@Beck]), where JC2(X∘)→B∘\mathcal{J}_{{\mathbf{C}}}^2({\mathcal{X}}^\circ)\to {\mathbf{B}^\circ} is the intermediate Jacobian fibration defined by C{\mathbf{C}}-invariants in cohomology, we conclude Theorem [thm:thm1]. In fact, the isomorphism ([eq:J2C]) gave a hint that we have to use with Calabi-Yau orbifolds if C≠1{\mathbf{C}}\neq 1 instead of ordinary Calabi-Yau threefolds if C≠1{\mathbf{C}}\neq 1. By working with equivariant compactly supported cohomology instead of equivariant cohomology, we further obtain the Langlands dual version of Theorem [thm:thm1] and hence cover all simple simply-connected complex Lie groups as well.
    Besides the C{\mathbf{C}}-action and construction of C{\mathbf{C}}-invariant holomorphic volume forms, the Lie-theoretic construction gives a way to study the smooth locus of the family X→B{\mathcal{X}}\to {\mathbf{B}}. As an application, we see in Corollary [cor:comp] that if Δh\Delta_h is of type ADE{\mathrm{ADE}} and admits non-trivial Dynkin graph automorphisms C{\mathbf{C}} (see Remark [r:dynkin] in Appendix [SectFolding]), then the smooth locus of the family X→B(Σ,Δh){\mathcal{X}}\to {\mathbf{B}}({\Sigma}, \Delta_h) is much larger than the smooth locus of Δh\Delta_h-cameral curves. It follows that the family X→B(Σ,Δh){\mathcal{X}}\to {\mathbf{B}}({\Sigma}, \Delta_h) does not only geometrically encode the smooth loci of \begin{tikzcd} \mathcal{M}_{\mathrm{H}}({\Sigma}, G_{ad}(\Delta_h)) \ar[r] & {\mathbf{B}}({\Sigma}, \Delta) & \ar[l] \mathcal{M}_{\mathrm{H}}({\Sigma}, G_{sc}(\Delta_h)) \end{tikzcd} but correspondingly for Gad(Δ)G_{ad}(\Delta) and Gsc(Δ)G_{sc}(\Delta), under the natural inclusion B(Σ,Δ)⊂B(Σ,Δh){\mathbf{B}}({\Sigma}, \Delta)\subset {\mathbf{B}}({\Sigma}, \Delta_h) for Δ=ΔhC\Delta=\Delta_h^{\mathbf{C}}.

    Relation to other works

    Our work is intellectually indebted to [@Sz1], [@DDP], [@DDD] and [@Slo]. We relate our approach to the first three works, for example we give explicit equations in the spirit of [@DDD] (also see Theorem [thm:defX0]). For that reason, we give two Slodowy slices S⊂so(5,C)S\subset \mathfrak{so}(5,{\mathbb{C}}) and Sh⊂sl(4,C)S_h \subset \mathfrak{sl}(4,{\mathbb{C}}) and show by a direct computation that they both realize the semi-universal deformation of the B2\mathrm{B}_2-singularity in Appendix [a:ex]. Moreover, we show that families of [@Sz1] provide simultaneous resolutions over a cone in the singular locus of π:X→B{\bm{\pi}}:{\mathcal{X}}\to {\mathbf{B}} (Proposition [p:simresol]). Graph automorphisms also appear in [@Sz1] but we use them differently. In particular, orbifold stacks and their cohomology appear in neither of these works. Finally, it is further crucial for our work [@Beck] that the ‘volume forms’ sb∈H0(Xb,KXb)s_b\in H^0(X_b,K_{X_b}) induce a period map on B∘{\mathbf{B}^\circ}. This period map is a so-called abstract Seiberg-Witten differential which in turn determines the structure of an algebraic integrable system on JC2(X∘/B∘)→B∘\mathcal{J}_{\mathbf{C}}^2({\mathcal{X}}^\circ/{\mathbf{B}^\circ})\to {\mathbf{B}^\circ} (cf. Section 4.4 in [@Beck]).

    Outline

    Since the key ingredient for the local building blocks of the Calabi-Yau orbifolds ([intro:fam]) are Slodowy slices, we recall and prove important preparatory results on them in Section [s:local]. Section [s:cy] begins with a recap on Hitchin bases and cameral curves and continues with the construction of Calabi-Yau orbifolds and related families of quasi-projective Calabi-Yau threefolds. In Section [s:mhs] we study mixed Hodge structures on equivariant cohomology and prove Theorem [thm:thm1]. Appendix [SectFolding] contains our conventions on folding of Dynkin diagrams and Appendix [a:ex] gives two examples of Slodowy slices that realize the semi-universal deformations of the same surface singularity.

    Acknowledgements

    It is a pleasure to thank Katrin Wendland and Emanuel Scheidegger as well as Ron Donagi and Tony Pantev for discussions and questions on this paper’s content. This work was supported by the DFG Graduiertenkolleg 1821 “Cohomological Methods in Geometry” and through the DFG Emmy-Noether grant on “Building blocks of physical theories from the geometry of quantization and BPS states”, number AL 1407/2-1.

    Local preparations {#s:local}

    Δ\Delta-singularities and Slodowy slices

    Let Δ\Delta be an irreducible Dynkin diagram. Then there is a unique pair2 (Δh,AS(Δ))(\Delta_h,AS(\Delta)) consisting of an ADE{\mathrm{ADE}}-Dynkin diagram and the associated symmetry group AS(Δ)⊂Aut⁡(Δh)AS(\Delta)\subset \operatorname{Aut}(\Delta_h) such that Δ\Delta is obtained by folding, i.e. Δ=ΔhAS\Delta=\Delta_h^{AS} (see Appendix [SectFolding] for more details). A Δ\Delta-singularity (Y,H)(Y,H) ([@Slo]) consists of a (germ of a) surface singularity Y=(Y,0)Y=(Y,0) of type Δh\Delta_h and a subgroup H⊂Aut⁡(Y)H\subset \operatorname{Aut}(Y) with the following properties:

    1. H≅AS(Δ)H\cong AS(\Delta);

    2. the action of HH on Y−{0}Y-\{0\} is free;

    3. the induced action on the dual resolution graph of the minimal resolution Y^→Y\hat{Y}\to Y coincides with the AS(Δ)AS(\Delta)-action on Δh\Delta_h.

    For Δ=Δh\Delta=\Delta_h, these are just ADE{\mathrm{ADE}}-surface singularities. Every Δ\Delta-singularity (Y,H)(Y,H) is quasi-homogeneous (cf. Table [ADEsings] in Appendix [SectFolding]), in particular, YY carries a C∗{\mathbb{C}}^*-action that commutes with the HH-action. A C∗{\mathbb{C}}^*-deformation of (Y,H)(Y,H) is therefore a C∗×H{\mathbb{C}}^*\times H-deformation Y→B\mathcal{Y}\to B such that HH acts trivially on the base. Each (Y,H)(Y,H) has a semi-universal C∗{\mathbb{C}}^*-deformation, see Section 22 in [@Slo].
    The relation to the simple complex Lie algebra g=g(Δ){\mathfrak{g}}={\mathfrak{g}}(\Delta) of type Δ\Delta is as follows: Let x∈gx\in {\mathfrak{g}} be a subregular nilpotent element and (x,y,h)(x,y,h) an sl2\mathfrak{sl}_2-triple in g{\mathfrak{g}} with semisimple hh. Then the Slodowy slice through xx (associated to the triple (x,y,h)(x,y,h)) is given by S=S(Δ):=x+ker⁡ad(y).S=S(\Delta):=x+\ker \mathrm{ad}(y). It carries a non-trivial action by the group C∗×AS(Δ){\mathbb{C}}^*\times AS(\Delta) (cf. Section 6 in [@Slo]). Here the AS(Δ)AS(\Delta)-action is defined by the action (via adjunction) of the group3 \labeleq:CcC=C(x,h)/C(x,h)∘≅AS(Δ)\label{eq:Cc} {\mathbf{C}}= C(x,h)/C(x,h)^\circ \cong AS(\Delta) which is the group of connected components of C(x,h)={g∈Gad ∣ g⋅x=x,g⋅h=h}C(x,h)=\{g\in G_{ad}~|~g\cdot x=x, g\cdot h= h\}. For a fixed Cartan subalgebra t⊂g{\mathfrak{t}}\subset {\mathfrak{g}}, denote by χ:g→t/W\chi:{\mathfrak{g}}\to {\mathfrak{t}}/W the adjoint quotient and σ:=χ∣S:S→t/W\sigma:=\chi_{|S}: S\to {\mathfrak{t}}/W its restriction to the Slodowy slice SS. If we let C∗{\mathbb{C}}^* act on t/W{\mathfrak{t}}/W with twice the usual weight and C{\mathbf{C}} on t/W{\mathfrak{t}}/W trivially, then σ\sigma is C∗×C{\mathbb{C}}^*\times {\mathbf{C}}-equivariant. Slodowy has shown that σ:S→t/W\sigma: S \to {\mathfrak{t}}/W with the C∗×C{\mathbb{C}}^*\times {\mathbf{C}}-action is a C∗{\mathbb{C}}^*-semi-universal deformation of the Δ\Delta-singularity (σ−1(0ˉ),x)(\sigma^{-1}(\bar{0}),x) ([@Slo], Section 8.7).
    Finally, we recall Grothendieck’s simultaneous resolution of χ:g→t/W\chi:{\mathfrak{g}}\to {\mathfrak{t}}/W, \label{eq:Gsimresol} \begin{tikzcd} \tilde{{\mathfrak{g}}}\cong G\times^B \mathfrak{b} \ar[r, "\psi"] \ar[d, "\theta"'] & {\mathfrak{g}}\ar[d, "\chi"] \\ {\mathfrak{t}}\ar[r , "q"] & {\mathfrak{t}}/W. \end{tikzcd} If S=x+ker⁡ad(y)S=x+\ker \mathrm{ad}(y) is a Slodowy slice, then ([eq:Gsimresol]) restricts to the simultaneous resolution ψ:S~→S\psi:\tilde{S}\to S of σ:S→t/W\sigma: S\to {\mathfrak{t}}/W (see Chapter 5 in [@Slo]). If (x,y,h)(x,y,h) is the sl2\mathfrak{sl}_2-triple defining SS, then the resulting square is a square of C∗{\mathbb{C}}^*-spaces if h∈th\in {\mathfrak{t}} (see Remark 1.53 in [@Beck-thesis] for details). Since any two Cartan subalgebras are conjugate to each other, we will assume h∈th\in {\mathfrak{t}} from now on.

    Relative symplectic form

    It is well-known (e.g. Chapter 1, [@ChrissGinzburg]) that the the Kostant-Kirillov form ν=ωKK\nu=\omega_{KK} restricts to a symplectic form on each adjoint orbit O⊂gO\subset {\mathfrak{g}}. Consequently, the relative differential form ν^∈Ωχ2(g)\hat{\nu}\in \Omega^2_{\chi}({\mathfrak{g}}) induced by ν\nu restricts to a relative symplectic form on the locus greg{\mathfrak{g}}^{reg} of regular elements in g{\mathfrak{g}}. This holds true for the restriction ν^∣Sreg∈Ωσreg2\hat{\nu}_{|S^{reg}}\in \Omega^2_{\sigma^{reg}} to Sreg:=S∩gregS^{reg}:=S\cap {\mathfrak{g}}^{reg} as well (cf. [@GG], Appendix 7). Since S−SregS-S^{reg} is of codimension at least 22, ν^∣Sreg\hat{\nu}_{|S^{reg}} extends to a global section of KσK_{\sigma} which we also denote by ν^∈Γ(S,Kσ)\hat{\nu}\in \Gamma(S,K_{\sigma}).

    [p:C*equivariant] Let S⊂gS \subset {\mathfrak{g}} be a Slodowy slice and ν^∈Γ(S,Kσ)\hat{\nu}\in \Gamma(S,K_{\sigma}) be the previously constructed section. Then ν^\hat{\nu} is nowhere vanishing, in particular KσK_\sigma is trivializable. Moreover, it is C∗{\mathbb{C}}^*-equivariant, λ∗ν^=λ2ν^,λ∈C∗,\lambda^*\hat{\nu}=\lambda^2 \hat{\nu},\quad \lambda\in {\mathbb{C}}^*, as well as C{\mathbf{C}}-invariant.

    The proof proceeds in two steps. In the first step, we use Yamada’s realization ([@Yamada]) of θ:g~→t\theta: \tilde{{\mathfrak{g}}}\to {\mathfrak{t}} as a relative symplectic reduction to relate the corresponding relative symplectic form ω^∈Ωθ2(g~)\hat{\omega} \in \Omega^2_{\theta}(\tilde{{\mathfrak{g}}}) to ν^∈Ωχ2(g)\hat{\nu}\in \Omega^2_\chi({\mathfrak{g}}) over the locus greg⊂g{\mathfrak{g}}^{reg}\subset {\mathfrak{g}} of regular elements. In the second step, we restrict to SregS^{reg} and extend over the non-regular locus.
    For the first step, fix a maximal torus TT and a Borel subgroup B⊃TB\supset T in the adjoint group G=GadG=G_{ad} and denote by N⊂BN\subset B the nilradical of BB. Then the symplectic manifold (M:=T∗(G/N),ωc)(M:=T^*(G/N),\omega_c), with the canonical symplectic structure ωc\omega_c, carries a natural Hamiltonian GG- and TT-action induced by left and right multiplication on G/NG/N respectively. The corresponding moment maps descend to μ^G:M/T→g∗\hat{\mu}_G:M/T \to {\mathfrak{g}}^* and μ^T:M/T→t∗\hat{\mu}_T:M/T \to {\mathfrak{t}}^*. Yamada ([@Yamada]) identifies the square ([eq:Gsimresol]) with the diagram \begin{tikzcd} M/T \ar[r, "\hat{\mu}_G"] \ar[d, "\hat{\mu}_T"'] & {\mathfrak{g}}\ar[d, "\chi"] \\ {\mathfrak{t}}\ar[r] & {\mathfrak{t}}/W \end{tikzcd} after g∗≅g{\mathfrak{g}}^*\cong {\mathfrak{g}} and t∗≅t{\mathfrak{t}}^*\cong {\mathfrak{t}} via the Killing form (which we will do from now on). Under this identification, the symplectic form ωc\omega_c descends and induces the relative symplectic form ω^∈Ωθ2(g~)\hat{\omega}\in \Omega^2_{\theta}(\tilde{{\mathfrak{g}}}). Moreover, for each t∈tt\in {\mathfrak{t}} the induced GG-action on (θ−1(t),ω^t)(\theta^{-1}(t),\hat{\omega}_t) is Hamiltonian and the restriction ψt:θ−1(t)→g\psi_t:\theta^{-1}(t) \to {\mathfrak{g}} of ψ:g~→g\psi: \tilde{{\mathfrak{g}}}\to {\mathfrak{g}} is the corresponding moment map (Theorem 2.5 in [@Yamada]). Now let ψreg:g~reg→greg\psi^{reg}:\tilde{{\mathfrak{g}}}^{reg}\to {\mathfrak{g}}^{reg} be the restriction of ψ:g~→g\psi: \tilde{{\mathfrak{g}}} \to {\mathfrak{g}} to the regular locus. Note that ψtreg:g~reg∩θ−1(t)→greg∩χ−1(tˉ),q(t)=tˉ,\psi^{reg}_t: \tilde{{\mathfrak{g}}}^{reg}\cap \theta^{-1}(t) \to {\mathfrak{g}}^{reg}\cap \chi^{-1}(\bar{t}),\quad q(t)=\bar{t}, is an isomorphism for each t∈tt\in {\mathfrak{t}}. In fact, it is not difficult to see that ψtreg\psi^{reg}_t is a symplectomorphism if the left-hand side is endowed with the symplectic form ω^t\hat{\omega}_{t} and the right-hand side with the Kostant-Kirillov form ν\nu. Since ψreg\psi^{reg} is smooth, (ψreg)∗Ωχ2≅Ωθ2(\psi^{reg})^*\Omega_\chi^2\cong \Omega^2_\theta naturally. Under this isomorphism, we have ψ∗ν^=ω^∈Γ(g~reg,Ωθ2)\psi^*\hat{\nu}=\hat{\omega}\in \Gamma(\tilde{{\mathfrak{g}}}^{reg}, \Omega^2_\theta) because ψtreg\psi^{reg}_t is a symplectomorphism for each t∈tt\in {\mathfrak{t}}.
    For the second step, we note that the restriction of ω^\hat{\omega} to S~⊂g~\tilde{S}\subset \tilde{{\mathfrak{g}}} is again a relative symplectic form (Theorem 4.5 in [@Yamada]), again denoted by ω^∈Ωσ~2(S~)\hat{\omega}\in \Omega^2_{\tilde{\sigma}}(\tilde{S}). Again we have a natural isomorphism Φreg:ψ∗Kσreg=ψ∗Ωσreg2→Ωσ~reg2\Phi^{reg}:\psi^*K_{\sigma^{reg}}=\psi^*\Omega^2_{\sigma^{reg}}\to \Omega^2_{\tilde{\sigma}^{reg}} over S~reg\tilde{S}^{reg} with Φreg(ψ∗ν^)=ω^\Phi^{reg}(\psi^*\hat{\nu})=\hat{\omega}. We claim that Φreg\Phi^{reg} extends to an isomorphism Φ:ψ∗Kσ→Kσ~\Phi: \psi^*K_{\sigma} \to K_{\tilde{\sigma}} satisfying \labeleq:nuomegaΦ(ψ∗ν^)=ω^.\label{eq:nuomega} \Phi(\psi^*\hat{\nu})=\hat{\omega}. Since KσK_\sigma and Kσ~K_{\tilde{\sigma}} are reflexive, it is sufficient to prove codimS~T~≥2\mathrm{codim}_{\tilde{S}}\tilde{T}\geq 2 for T~:=S~−S~reg\tilde{T}:=\tilde{S}-\tilde{S}^{reg}. The components of T~\tilde{T} of highest dimension lie over the hypersurfaces tα−∩β≠αtβ∩tα{\mathfrak{t}}_{\alpha}-\cap_{\beta\neq \alpha} {\mathfrak{t}}_\beta\cap {\mathfrak{t}}_\alpha. If tt lies in such a hypersurface, then σ~−1(tˉ)∩T~\tilde{\sigma}^{-1}(\bar{t})\cap \tilde{T} consists of the exceptional divisor of ψt:S~t→Stˉ\psi_t:\tilde{S}_t\to S_{\bar{t}} for tˉ=q(t)\bar{t}=q(t). Hence these components of T~\tilde{T} have dimension (r−1)+1=r(r-1)+1=r which is of codimension 22 in S~\tilde{S}. Equation ([eq:nuomega]) implies that ν^\hat{\nu} is nowhere vanishing because ω^\hat{\omega} is and ψ\psi is surjective.
    Finally, the C∗{\mathbb{C}}^*-equivariance follows from that of ω^\hat{\omega} (Corollary 4.6 in [@Yamada]) whereas the C{\mathbf{C}}-invariance is a consequence of the Aut⁡(g)\operatorname{Aut}({\mathfrak{g}})-invariance of the Kostant-Kirillov form ν=ωKK\nu=\omega_{KK}.

    The morphism σ~:S~→t\tilde{\sigma}: \tilde{S}\to {\mathfrak{t}} is topologically trivial (cf. [@Slo2]). Hence parallel transport defines canonical isomorphisms Pt:H2(S~t,C)→H2(S~0,C)P_t: H^2(\tilde{S}_t,{\mathbb{C}}) \to H^2(\tilde{S}_0, {\mathbb{C}}), t∈t t\in {\mathfrak{t}}.

    The period map \labeleq:H2perPS~:t→H2(S~0,C)C,t↦Pt([ω^t]),\label{eq:H2per} P_{\tilde{S}}:{\mathfrak{t}}\to H^2(\tilde{S}_0,{\mathbb{C}})^{\mathbf{C}},\quad t\mapsto P_t([\hat{\omega}_t]), is a WW-equivariant isomorphism.

    The case of ADE{\mathrm{ADE}}-Dynkin diagrams, i.e. C=1{\mathbf{C}}=1, goes back to [@Yamada]. For C≠1{\mathbf{C}}\neq 1, it has been checked in [@Beck-thesis], Corollary 1.98.

    Derivative of the adjoint quotient {#SloSectRemOnDerivatives}

    Let b:U→t/Wb:U\to {\mathfrak{t}}/W be a morphism, where U⊂CU\subset {\mathbb{C}} is a Zariski-open subset. Then we define U~b\tilde{U}_b and YbY_b by the fiber products \label{LocalCY3Cameral} \begin{tikzcd} \tilde{U}_b \arrow[r] \arrow[d, "p_b"'] & {\mathfrak{t}}\arrow[d, "q"] \\ U \arrow[r, "b"] & {\mathfrak{t}}/W \\ Y_b \arrow[r] \arrow[u, "\pi_b"] & S. \arrow[u, "\sigma"'] \end{tikzcd} In Section [s:cy] it will become clear that U~b\tilde{U}_b is a local model for a cameral curve and YbY_b is a local model for the quasi-projective Calabi-Yau threefolds that we construct in loc. cit. Both U~b\tilde{U}_b and YbY_b are non-singular if bb is transversal to qq and σ\sigma. This condition in particular implies that the rank of dqdq may not be less than r−1r-1 for r=rk(g)r=\mathrm{rk}({\mathfrak{g}}).

    [ExSL2] We consider the simplest example, g=sl2(C){\mathfrak{g}}=\mathfrak{sl}_2({\mathbb{C}}). Of course, a Slodowy slice SS has to be all of g{\mathfrak{g}}, t=C≅t/W{\mathfrak{t}}={\mathbb{C}}\cong {\mathfrak{t}}/W and q:C→C,z↦z2,σ:S→C,A↦det⁡(A).\begin{gathered} q:{\mathbb{C}}\to {\mathbb{C}}, \quad z\mapsto z^2,\\ \sigma: S\to {\mathbb{C}}, \quad A\mapsto \det(A). \end{gathered} A morphism b:U→t/Wb:U\to {\mathfrak{t}}/W is transversal to qq iff it has only zeros of multiplicity 11. If this is the case U~b={(x,y)∈U×C ∣ y2−b(x)=0}\tilde{U}_b=\{ (x,y)\in U\times {\mathbb{C}}~|~y^2-b(x)=0\} is non-singular and a branched double covering of UU. It branches precisely over the zeros of bb.
    Identifying sl2(C)≅C3\mathfrak{sl}_2({\mathbb{C}})\cong {\mathbb{C}}^3, we write σ(u,v,w)=−u2−vw\sigma(u,v,w)=-u^2-vw. Again using that bb has simple zeros only, we see that Yb={((u,v,w),x)∈C3×U ∣ −u2−vw−b(x)=0}Y_b=\{ ((u,v,w),x)\in {\mathbb{C}}^3\times U~|~-u^2-vw-b(x)=0\} is non-singular. Note however, that if b(x)=0b(x)=0, then πb−1(x)\pi^{-1}_b(x) is an A1\mathrm{A}_1-singularity.

    [RKdq] Let x=hx=h be semisimple and t∈tt\in {\mathfrak{t}} with χ(h)=q(t)\chi(h)=q(t). Then im(dqt)⊂im(dχh)\mathrm{im}(dq_t)\subset \mathrm{im}(d\chi_h) and rk(dqt)=dim⁡⋂α∈R,α(t)=0ker⁡α=dim⁡C(Zg(h))\mathrm{rk}(dq_t)=\dim \bigcap_{\alpha \in R, \alpha(t)=0} \ker \alpha=\dim C(Z_{\mathfrak{g}}(h)) where RR are the roots corresponding to t⊂g{\mathfrak{t}}\subset {\mathfrak{g}}.

    Let t(h)⊂g{\mathfrak{t}}(h)\subset {\mathfrak{g}} be a Cartan subalgebra that contains hh. Since t(h){\mathfrak{t}}(h) is conjugate to t{\mathfrak{t}} we may assume that h∈th\in {\mathfrak{t}} and in fact even t=ht=h by the Ad-invariance of χ\chi and qq. The first claim is obvious because χ∣t=q\chi_{|{\mathfrak{t}}}=q.
    For the second claim we may assume t=ht=h as before. Then the first equality is proven in [@Steinberg]. For the second equality we claim \labelCenterOfCentralizer⋂α∈R,α(t)=0ker⁡α=C(Zg(h))\begin{aligned} \label{CenterOfCentralizer} \bigcap_{\alpha \in R, \alpha(t)=0} \ker \alpha=C(Z_{\mathfrak{g}}(h))\end{aligned} which can be seen as follows: Let g=t⊕⨁α∈Rgα{\mathfrak{g}}={\mathfrak{t}}\oplus \bigoplus_{\alpha\in R}{\mathfrak{g}}_\alpha be the root space decomposition with respect to t{\mathfrak{t}}. Then we get Zg(h)=t⊕⨁α∈R,α(h)=0gα.Z_{\mathfrak{g}}(h)={\mathfrak{t}}\oplus \bigoplus_{\alpha \in R, \alpha(h)=0} {\mathfrak{g}}_\alpha. Since [t,gα]≠{0}[{\mathfrak{t}},{\mathfrak{g}}_{\alpha}]\neq \{0\} and [h,gα]=0[h,{\mathfrak{g}}_\alpha]=0 iff α(h)=0\alpha(h)=0, the equality ([CenterOfCentralizer]) follows.

    If x=hx=h is semisimple and t∈tt\in {\mathfrak{t}} with q(t)=χ(h)q(t)=\chi(h), then im(dqt)=im(dχh)\mathrm{im}(dq_t)=\mathrm{im}(d\chi_h).

    If x=h+vx=h+v is the Jordan decomposition of x∈gx\in {\mathfrak{g}}, for hh semisimple and vv nilpotent, then (see [@Rich]) \labelCaseOfInterestDerivativesrk(dχx)=dim⁡C(Zg(h))+rk(dχ1,v).\begin{aligned} \label{CaseOfInterestDerivatives} \mathrm{rk}(d\chi_x)=\dim C(Z_{\mathfrak{g}}(h))+ \mathrm{rk}(d\chi_{1,v}).\end{aligned} Here χ1:g1:=[Zg(h),Zg(h)]→t1/W1\chi_1:{\mathfrak{g}}_1:=[Z_{{\mathfrak{g}}}(h), Z_{{\mathfrak{g}}}(h)] \to {\mathfrak{t}}_1/W_1 is the induced adjoint quotient of the semisimple Lie algebra g1{\mathfrak{g}}_1 and C(Zg(h))C(Z_{{\mathfrak{g}}}(h)) is the cetner of the centralizer of hh in g{\mathfrak{g}}. By the previous lemma, it remains to show that dχ0=0d\chi_0=0 for any semisimple adjoint quotient χ:g→t/W\chi:{\mathfrak{g}}\to {\mathfrak{t}}/W. This follows from the fact that the degrees did_i of any basis χ^i\hat{\chi}_i of GadG_{ad}-invariant polynomials are greater or equal 22 because g{\mathfrak{g}} is semisimple ([@BourbakiLie4-6]).

    [p:transversal] Let b:U→t/Wb:U\to {\mathfrak{t}}/W be a morphism from an open U⊂CU\subset {\mathbb{C}} which is transversal to q:t→t/Wq:{\mathfrak{t}}\to {\mathfrak{t}}/W. Then it is also transversal to χ:g→t/W\chi:{\mathfrak{g}}\to {\mathfrak{t}}/W and σ:S→t/W\sigma:S\to {\mathfrak{t}}/W.

    Let x=h+v∈gx=h+v\in {\mathfrak{g}} and t∈tt\in {\mathfrak{t}} such that χ(x)=q(t)\chi(x)=q(t). The previous corollary together with ([CaseOfInterestDerivatives]) implies that r≥rk(dχx)≥rk(dqt)≥r−1.r \geq \mathrm{rk}(d\chi_x)\geq \mathrm{rk}(dq_t)\geq r-1. If x=hx=h is semisimple, then χ\chi is also transversal to bb at xx by the previous lemma. So we are left with the case x=h+vx=h+v and rk(dχx)=r−1=rk(dqt)\mathrm{rk}(d\chi_x)=r-1=\mathrm{rk}(dq_t). We claim that v=0v=0 and x=hx=h must be semisimple which concludes the proof. Without loss of generality we assume again that h∈th\in {\mathfrak{t}}. Since dim⁡C(Zg(h))=rk(dqt)=r−1\dim C(Z_{\mathfrak{g}}(h))=\mathrm{rk}(dq_t)=r-1 by Lemma [RKdq], it follows that Zg(h)=t⊕gα⊕g−αZ_{\mathfrak{g}}(h)={\mathfrak{t}}\oplus {\mathfrak{g}}_\alpha \oplus {\mathfrak{g}}_{-\alpha} for a root α\alpha with respect to t{\mathfrak{t}}. Therefore the derived algebra is [Zg(h),Zg(h)]=⟨hα,g±α⟩≅sl2(C),[Z_{\mathfrak{g}}(h),Z_{\mathfrak{g}}(h)]=\langle h_\alpha, {\mathfrak{g}}_{\pm \alpha}\rangle \cong \mathfrak{sl}_2({\mathbb{C}}), where hαh_\alpha generates the commutator [gα,g−α]⊂t[{\mathfrak{g}}_\alpha,{\mathfrak{g}}_{-\alpha}]\subset {\mathfrak{t}}. As a consequence, vv can be considered as a nilpotent element in sl2(C)\mathfrak{sl}_2({\mathbb{C}}) because v∈gα⊕g−α⊂Zg(h)v\in {\mathfrak{g}}_\alpha\oplus {\mathfrak{g}}_{-\alpha}\subset Z_{\mathfrak{g}}(h). By formula ([CaseOfInterestDerivatives]) and rk(dqt)=r−1\mathrm{rk}(dq_t)=r-1 we must have rk(dχ1,v)=0\mathrm{rk}(d\chi_{1,v})=0 for the adjoint quotient χ1=det⁡:sl2(C)→t1/W1\chi_1=\det:\mathfrak{sl}_2({\mathbb{C}})\to {\mathfrak{t}}_1/W_1. But dAdet⁡=(−2a,−c,−b)d_A \det=(-2a,-c,-b) for A=aH+bX+cY∈sl2(C)A=aH+bX+cY\in \mathfrak{sl}_2({\mathbb{C}}) in the standard basis H,X,YH,X,Y of sl2(C)\mathfrak{sl}_2({\mathbb{C}}). Hence we must have v=0v=0, i.e. x=hx=h is semisimple (and subregular).

    Calabi-Yau orbifolds over Hitchin bases {#s:cy}

    Hitchin bases and cameral curves {#ss:hitchinbase}

    This subsection mainly fixes notation and we refer to [@DG], [@DP] for more details. As before, we let GG be a simple complex Lie group with Dynkin diagram Δ=Δ(G)\Delta=\Delta(G) and t⊂g=g(Δ){\mathfrak{t}}\subset {\mathfrak{g}}={\mathfrak{g}}(\Delta) a Cartan subalgebra in g=Lie(G){\mathfrak{g}}=\mathrm{Lie}(G). If Σ{\Sigma} is a compact connected Riemann surface of genus ≥2\geq 2, we denote by MH(Σ,G)\mathcal{M}_{\mathrm{H}}({\Sigma},G) the moduli space of semistable GG-Higgs bundles of degree 1∈π1(G)1\in \pi_1(G) and by h:MH(Σ,G)→B(Σ,G){\bm{h}}:\mathcal{M}_{\mathrm{H}}({\Sigma},G)\to {\mathbf{B}}({\Sigma}, G) the Hitchin map. It maps to the Hitchin base B ⁣:=B(Σ,G)=H0(Σ,U){\mathbf{B}}\colon={\mathbf{B}}({\Sigma}, G)=H^0({\Sigma}, {\bm{U}}) for the bundle u ⁣:U=KΣ×C∗t/W→Σ u\colon{\bm{U}}=K_{\Sigma}\times_{{\mathbb{C}}^*} {\mathfrak{t}}/W\to {\Sigma} of cones where C∗{\mathbb{C}}^* acts by the standard action on t/W{\mathfrak{t}}/W. Choosing generators χ1,…,χr∈C[t]W\chi_1,\dots ,\chi_r\in {\mathbb{C}}[{\mathfrak{t}}]^W of Weyl group invariants gives U{\bm{U}} the structure of a vector bundle via the isomorphism U≅⨁j=1rKΣdj{\bm{U}}\cong \bigoplus_{j=1}^r K_{\Sigma}^{d_j} for the degrees dj=deg⁡(χj)d_j=\deg(\chi_j). In particular, the Hitchin base B{\mathbf{B}} inherits a vector space structure. By its construction, B{\mathbf{B}} only depends on the Lie algebra g{\mathfrak{g}} of GG so that the notation B(Σ,Δ):=B(Σ,G){\mathbf{B}}({\Sigma},\Delta):={\mathbf{B}}({\Sigma}, G) is justified. The total space of the vector bundle u~ ⁣:U~=KΣ⊗t→Σ \tilde{u}\colon{\bm{\tilde{U}}}=K_{\Sigma}\otimes {\mathfrak{t}}\to {\Sigma} maps to U{\bm{U}} via the natural quotient map q ⁣:U~→U\bm{q}\colon{\bm{\tilde{U}}}\to {\bm{U}}. The fibers of the universal cameral curve p:Σ~:=ev∗U~→B,\bm{p}:{\bm{\tilde{\Sigma}}}:=ev^*{\bm{\tilde{U}}}\to {\mathbf{B}}, for the evaluation map ev:Σ×B→Uev:{\Sigma}\times {\mathbf{B}}\to {\bm{U}}, are the cameral curves Σ~b:=p−1(b){\tilde{\Sigma}}_{b}:=\bm{p}^{-1}(b). These are smooth over the Zariski-open and dense B∘:={b∈B ∣ b\mboxintersectsdisc(q)\mboxtransversally}{\mathbf{B}^\circ}:=\{ b\in {\mathbf{B}}~|~b \mbox{ intersects }\mathrm{disc}(\bm{q})\mbox{ transversally} \} and the generic Hitchin fibers h−1(b){\bm{h}}^{-1}(b), b∈B∘b\in {\mathbf{B}^\circ}, are generalized Prym varieties defined in terms of Σ~b{\tilde{\Sigma}}_b.
    For later reference, we further need the cone \labeleq:B/WB~/W↪B,B~:=H0(Σ,U~),\label{eq:B/W} \tilde{{\mathbf{B}}}/W\hookrightarrow {\mathbf{B}}, \quad \tilde{{\mathbf{B}}}:=H^0({\Sigma}, {\bm{\tilde{U}}}), where the Weyl group WW acts pointwise on sections. Away from 0∈B0\in {\mathbf{B}}, this is the locus of completely reducible but reduced cameral curves, i.e. Σ~b=∐w∈WΣ~b,w,Σ~b,w≅Σ.{\tilde{\Sigma}}_b=\coprod_{w\in W} {\tilde{\Sigma}}_{b,w}, \quad {\tilde{\Sigma}}_{b,w}\cong {\Sigma}. The irreducible components Σ~b,w{\tilde{\Sigma}}_{b,w} intersect over the points of the divisor of the section ∏α∈Rα(b)∈H0(Σ,KΣ),\prod_{\alpha \in R} \alpha(b)\in H^0({\Sigma}, K_{\Sigma}), where R=R(Δ)R=R(\Delta) is the corresponding root system (which makes sense because ∏α∈Rα∈C[t]W\prod_{\alpha\in R}\alpha \in {\mathbb{C}}[{\mathfrak{t}}]^W).

    Surfaces

    For the following constructions, we fix a spin bundle L∈Picg−1(Σ)L\in \mathrm{Pic}^{g-1}({\Sigma}), L2=KΣL^2=K_{{\Sigma}} and denote by α=αL∈H1(Σ,O∗)\alpha=\alpha_L\in H^1({\Sigma}, {\mathcal{O}}^*) its cohomology class. Moreover, let Δ\Delta be an irreducible Dynkin diagram with associated symmetry group C=AS(Δ){\mathbf{C}}=AS(\Delta) and S=S(Δ)⊂g(Δ)S=S(\Delta)\subset {\mathfrak{g}}(\Delta) a Slodowy slice. Twisting the C∗{\mathbb{C}}^*-spaces SS and t/W{\mathfrak{t}}/W by LL, we obtain4 \labeleq:ScbS:=L×C∗S,L×C∗t/W≅U.\label{eq:Scb} {\bm{S}}:=L\times_{{\mathbb{C}}^*} S, \quad L\times_{{\mathbb{C}}^*} {\mathfrak{t}}/W \cong {\bm{U}}. By its C∗{\mathbb{C}}^*-equivariance, the morphism σ:S→t/W\sigma:S\to {\mathfrak{t}}/W and its simultaneous resolution σ~:S~→t\tilde{\sigma}:\tilde{S}\to {\mathfrak{t}} glue to give the following commutative diagram \label{eq:tScb} \begin{tikzcd} {\tilde{\bm{S}}}\ar[r, "\bm{\psi}"] \ar[d, "{\tilde{\bm{\sigma}}}"'] & {\bm{S}}\ar[d, "{\bm{\sigma}}"] \\ {\bm{\tilde{U}}}\ar[r, "\bm{q}"] & {\bm{U}}\end{tikzcd} of C{\mathbf{C}}-spaces where C{\mathbf{C}} acts trivially on U~{\bm{\tilde{U}}} and U{\bm{U}}. However, it takes more care to glue the sections ω^∈Γ(S~,Ωσ~2)\hat{\omega}\in \Gamma(\tilde{S},\Omega^2_{\tilde{\sigma}}) and ν^∈Γ(S,Kσ)\hat{\nu}\in \Gamma(S, K_{\sigma}).

    [l:bhatnu] With the notation of ([eq:Scb]) and ([eq:tScb]), the following holds:

    1. The section ω^\hat{\omega} glues to a section ω^∈Γ(S~,Ωσ2⊗(u~∘σ~)∗KΣ)\hat{{\bm{\omega}}}\in \Gamma({\tilde{\bm{S}}},\Omega^2_{{\bm{\sigma}}}\otimes (\tilde{u}\circ {\tilde{\bm{\sigma}}})^*K_{\Sigma}) which is C{\mathbf{C}}-invariant. It induces a fiberwise period map η:U~→u~∗U~\bm{\eta}:{\bm{\tilde{U}}}\to \tilde{u}^*{\bm{\tilde{U}}} which coincides with the tautological section τ∈Γ(U~,u~∗U~)\bm{\tau}\in \Gamma({\bm{\tilde{U}}}, \tilde{u}^*{\bm{\tilde{U}}}).

    2. Likewise, the section ν^\hat{\nu} glues to a C{\mathbf{C}}-invariant section ν^∈Γ(S,Kσ⊗(u∘σ)∗KΣ)\hat{\bm{\nu}}\in \Gamma({\bm{S}}, K_{{\bm{\sigma}}}\otimes (u\circ {\bm{\sigma}})^*K_{\Sigma}) such that ψ∗ν^=ω^\bm{\psi}^*\hat{\bm{\nu}}=\hat{\bm{\omega}} under the natural isomorphism ψ∗Kσ≅Ωσ~2\bm{\psi}^*K_{{\bm{\sigma}}}\cong \Omega^2_{{\tilde{\bm{\sigma}}}}.

    In particular, the sections σ^\hat{\bm{\sigma}} and ν^\hat{\bm{\nu}} give respective isomorphisms Ωσ2≅(u~∘σ~)∗KΣ−1\mboxandKσ≅(u∘σ)∗KΣ−1.\Omega^2_{{\bm{\sigma}}}\cong (\tilde{u}\circ {\tilde{\bm{\sigma}}})^*K_{{\Sigma}}^{-1} \quad \mbox{and} \quad K_{{\bm{\sigma}}} \cong (u\circ {\bm{\sigma}})^*K_{{\Sigma}}^{-1}.

    To construct the section ω^\hat{\bm{\omega}}, we need the gluing data for the sheaf Ωσ~2\Omega^2_{{\tilde{\bm{\sigma}}}}. Let α=αL\alpha=\alpha_L be the cocycle corresponding to LL and denote by (αij)(\alpha_{ij}) a Čech representative for a fixed open covering (Dij)(D_{ij}) of Σ{\Sigma}. In particular, (βij)=(αij2)(\beta_{ij})=(\alpha_{ij}^2) is a cocycle for KΣK_{\Sigma}. Trivializing S~{\tilde{\bm{S}}} over each DiD_{i} gives rise to the commutative diagram \begin{tikzcd} & {\tilde{\bm{S}}}_{ij} \arrow[r, "\psi_i"] & D_{ij}\times \tilde{S} \ar[r, "\mathrm{id}\times \tilde{\sigma}"] & D_{ij}\times {\mathfrak{t}}\\ {\bm{\tilde{U}}}\ar[ru, leftarrow, "\tilde{{\bm{\sigma}}}_{ij}"] \ar[rd, leftarrow, "\tilde{{\bm{\sigma}}}_{ij}"'] \\ & {\tilde{\bm{S}}}_{ij} \arrow[r, "\psi_j"] \arrow[uu, equal] & D_{ij}\times \tilde{S} \arrow[uu, "g_{ij}"'] \ar[r, "\mathrm{id}\times \tilde{\sigma}"]& D_{ij}\times {\mathfrak{t}}\ar[uu, "h_{ij}"'] \end{tikzcd} over Dij=Di∩DjD_{ij}=D_i\cap D_j with gij(x,s)=(x,αij(x)⋅s)=(x,μ(αij(x),s)),hij(x,t)=(x,αij(x)⋅t)=(x,βij(x)t).\begin{aligned} &g_{ij}(x,s)=(x, \alpha_{ij}(x)\cdot s)=(x, \mu(\alpha_{ij}(x),s)),\\ &h_{ij}(x,t)=(x, \alpha_{ij}(x)\cdot t)=(x, \beta_{ij}(x)t). \end{aligned} for (x,s)∈Dij×S~(x,s)\in D_{ij}\times \tilde{S} and the action map μ:C∗×S~→S~\mu:{\mathbb{C}}^*\times \tilde{S}\to \tilde{S}. On each Di×S~D_i\times \tilde{S} we have the sheaves Ei:=Ωid×σ~2≅pr2,i∗Ωσ~2\mathcal{E}_i:=\Omega^2_{\mathrm{id}\times \tilde{\sigma}}\cong \mathrm{pr}_{2,i}^*\Omega^2_{\tilde{\sigma}} together with the sections pr2,i∗ω^\mathrm{pr}_{2,i}^*\hat{\omega}. Clearly, Ei\mathcal{E}_i and Ej\mathcal{E}_j are canonically isomorphic over DijD_{ij}. Now Ωσ~2\Omega^2_{{\tilde{\bm{\sigma}}}} is glued from5 ψi∗Ei\psi_i^*\mathcal{E}_i on DijD_{ij} via the isomorphisms \begin{tikzcd} \varphi_{ij}:=\psi_i^*dg_{ji}^t: \psi_j^*\mathcal{E}_j=\psi_i^*g_{ji}^*\mathcal{E}_j \arrow[r] & \psi_i^*\mathcal{E}_i \end{tikzcd} over DijD_{ij}. Here we denote by dgjit:gij∗Ej→Ei=Ejdg_{ji}^t:g_{ij}^*\mathcal{E}_j\to \mathcal{E}_i=\mathcal{E}_j the natural morphism (over DijD_{ij}). Observe that we have φij∘φjk=φik\varphi_{ij}\circ \varphi_{jk}=\varphi_{ik} and (φij)(\varphi_{ij}) is the gluing (or descent) datum for Ωσ~2\Omega^2_{{\tilde{\bm{\sigma}}}}. Indeed, we can write this composition as \begin{tikzcd} \psi_k^*\mathcal{E}_k\ar[rr, "\varphi_{kj}"] \ar[d, "\cong"] && \psi_j^*\mathcal{E}_j \ar[rr, "\varphi_{ji}"] \ar[d, "\cong"] && \psi_i^*\mathcal{E}_i \ar[d, "\cong"] \\ \psi_i^*(g_{ji}^*) g_{kj}^*\mathcal{E}_j \ar[rr, "\psi_i^* g_{ji}^* dg_{kj}^t"] && \psi^*_i g_{ji}^*\mathcal{E}_j \ar[rr, "\psi_i^*dg_{ji}^t"] && \psi_i^*\mathcal{E}_j. \end{tikzcd} The lower line is φik\varphi_{ik} by the chain rule, showing the cocycle condition for (φij)(\varphi_{ij}).

    Define the local sections ω^i:=ψi∗pr2,i∗ω^∈Γ(Si,ψi∗Ei).\hat{\omega}_i:=\psi_i^*\mathrm{pr}_{2,i}^*\hat{\omega}\in \Gamma(\mathcal{S}_i,\psi_i^*\mathcal{E}_i). On the overlaps S~ij{\tilde{\mathcal{S}}}_{ij}, they transform as follows \labelTrafoRuleSectionsφij(ω^j)=((pr1,i∘ψi)∗βji) ω^i.\label{TrafoRuleSections} \varphi_{ij}(\hat{\omega}_j)=((\mathrm{pr}_{1,i}\circ \psi_i)^*\beta_{ji}) \, \hat{\omega}_i.

    Before we prove this claim, let us see how it yields the desired section. Observe that (pr1,i∘ψi)∗βji(\mathrm{pr}_{1,i}\circ\psi_i)^* \beta_{ji} is a cocycle for (u~∘σ~)∗KΣ−1(\tilde{u}\circ {\tilde{\bm{\sigma}}})^*K_{\Sigma}^{-1}. Hence in order to obtain a well-defined global section on S~\tilde{\mathcal{S}}, we have to tensor with (u~∘σ~)∗KΣ(\tilde{u}\circ {\tilde{\bm{\sigma}}})^*K_{\Sigma}. More precisely, let ζi∈Γ(Di,KΣ)\zeta_i\in \Gamma(D_i,K_\Sigma) be the local frames of KΣK_{\Sigma} over DiD_i so that ζi=βij ζj\zeta_i=\beta_{ij}~\zeta_j on DijD_{ij}. Letting ζ^i:=ψi∗pr1,i∗ζi\hat{\zeta}_i:=\psi_i^*\mathrm{pr}_{1,i}^*\zeta_i, we see that the local sections ω^i⊗ζ^i∈Γ(S~i,Ωσ~2⊗(u~∘σ~)∗KΣ)\hat{\omega}_i\otimes \hat{\zeta}_i\in\Gamma({\tilde{\bm{S}}}_i,\Omega^2_{{\tilde{\bm{\sigma}}}}\otimes (\tilde{u}\circ {\tilde{\bm{\sigma}}})^*K_\Sigma) glue to give the global section ω^∈Γ(S~,Ωσ~2⊗(u~∘σ~)∗KΣ)\bm{\hat{\omega}}\in \Gamma({\tilde{\bm{S}}},\Omega^2_{{\tilde{\bm{\sigma}}}}\otimes (\tilde{u}\circ {\tilde{\bm{\sigma}}})^*K_\Sigma) as claimed.

    We still have to give a proof of ([TrafoRuleSections]). To simplify notation, we drop the subscript ijij if not necessary and only write g:D×S~→D×S~g:D\times \tilde{S}\to D\times \tilde{S} etc. Then the second component of dg:TD⊕TS~→TD⊕TS~dg:TD\oplus T\tilde{S}\to TD\oplus T\tilde{S} at (x,s)∈D×S~(x,s)\in D\times \tilde{S} is given by \labelSecondComponentdμα(x)⋅s(dαx(v),w)=dμα(x)⋅s(dαx(v),0)+dμα(x)⋅s(0,w).\label{SecondComponent} d\mu_{\alpha(x)\cdot s}(d\alpha_x(v),w)=d\mu_{\alpha(x)\cdot s}(d\alpha_x(v),0)+d\mu_{\alpha(x)\cdot s}(0, w). Note that dμα(x)⋅s(0,w)=d(μα(x))s(w)d\mu_{\alpha(x)\cdot s}(0,w)=d(\mu_{\alpha(x)})_s(w) where μα(x)=μ(α(x),−)\mu_{\alpha(x)}=\mu(\alpha(x),-). Now let p∈S~ijp\in {\tilde{\mathcal{S}}}_{ij} and ψi(p)=(x,s)∈Dij×S~\psi_i(p)=(x,s)\in D_{ij}\times \tilde{S}. Then we clearly have ker⁡d(x,s)(id×σ~)=0⊕ker⁡dsσ~⊂TxDij⊕TsS~.\ker d_{(x,s)}(\mathrm{id}\times \tilde{\sigma})=0\oplus \ker d_s\tilde{\sigma}\subset T_xD_{ij}\oplus T_s\tilde{S}. In particular, the first summand in ([SecondComponent]) plays no role for our discussion. For wk∈ker⁡dsσ~w_k\in \ker d_s\tilde{\sigma} (k=1,2k=1,2) one computes  φij(ω^j)p((0,w1),(0,w2))=pr2,j∗ω^gji(x,s)∘dgji,(x,s)((0,w1),(0,w2))(ψj∘ψi−1=gji)=βji(x) ω^s(w1,w2)(C∗-equivariance and (\refSecondComponent))=(pr1,i∘ψi)∗βji(p) (ω^i)p((0,w1),(0,w2)).\begin{aligned} &\quad~\varphi_{ij}(\hat{\omega}_j)_p\Big((0,w_1),(0,w_2)\Big)\\ &=\mathrm{pr}_{2,j}^*\hat{\omega}_{g_{ji}(x,s)}\circ dg_{ji, (x,s)}\Big((0,w_1),(0,w_2)\Big) & (\psi_j\circ \psi_i^{-1}=g_{ji}) \\ &=\beta_{ji}(x) \, \hat{\omega}_s(w_1,w_2) & ({\mathbb{C}}^*\text{-equivariance and }(\ref{SecondComponent}))\\ &=(\mathrm{pr}_{1,i}\circ \psi_i)^*\beta_{ji}(p) \, (\hat{\omega}_i)_p\Big((0,w_1),(0,w_2)\Big). \end{aligned} Here we have used the C∗{\mathbb{C}}^*-equivariance of the relative form ω^\hat{\omega} showing ([TrafoRuleSections]).
    To construct the period map η\bm{\eta} out of ω^\hat{{\bm{\omega}}}, observe that it induces the morphism (x,t)↦PS~(t)⊗((x,t),ζi(x))(x,t)\mapsto P_{\tilde{S}}(t)\otimes ((x,t),\zeta_i(x)) in each trivialization U~∣Di≅Di×t{\bm{\tilde{U}}}_{|D_i} \cong D_i\times {\mathfrak{t}} where PS~P_{\tilde{S}} is as in ([eq:H2per]). These morphisms glue to give the morphism η:U~→t⊗u~∗KΣ\bm{\eta}: {\bm{\tilde{U}}}\to {\mathfrak{t}}\otimes\tilde{u}^*K_\Sigma which coincides with the tautological section τ∈H0(U~,u~∗U~)\bm{\tau}\in H^0({\bm{\tilde{U}}},\tilde{u}^*{\bm{\tilde{U}}}) by construction.
    The existence of the global section ν^∈H0(S,Kσ⊗(u∘σ)∗KΣ)\bm{\hat{\nu}}\in H^0({\bm{S}},K_{{\bm{\sigma}}}\otimes (u\circ {\bm{\sigma}})^*K_\Sigma) works similarly: There exists a section ν^reg∈Γ(Sreg,Kσ)\bm{\hat{\nu}}^{reg}\in \Gamma({\bm{S}}^{reg}, K_{{\bm{\sigma}}}), which is constructed as ω^\hat{\bm{\omega}} by replacing ω^\hat{\omega} with ν^\hat{\nu}. Here Sreg⊂S{\bm{S}}^{reg}\subset {\bm{S}} is the locus which is glued from Sreg⊂SS^{reg}\subset S. Using a codimension argument as in the proof of Proposition [p:C*equivariant], we see that it uniquely extends to a section ν^∈Γ(S,Kσ)\bm{\hat{\nu}}\in \Gamma({\bm{S}}, K_{{\bm{\sigma}}}). It satisfies ψ∗ν^=ω^\bm{\psi}^*\bm{\hat{\nu}}=\bm{\hat{\omega}} under the isomorphism ψ∗Kσ≅Kσ~\bm{\psi}^*K_{{\bm{\sigma}}}\cong K_{{\tilde{\bm{\sigma}}}} by construction, since this holds for the corresponding local sections.

    Calabi-Yau threefolds

    The family σ:S→U{\bm{\sigma}}: {\bm{S}}\to {\bm{U}} of surfaces over U{\bm{U}} pulls back to the product Σ×B{\Sigma}\times {\mathbf{B}}, B=B(Σ,Δ){\mathbf{B}}={\mathbf{B}}({\Sigma}, \Delta) via the evaluation map ev:Σ×B(Σ,Δ)→U~{ev}:{\Sigma}\times {\mathbf{B}}({\Sigma},\Delta)\to {\bm{\tilde{U}}}. Projecting to the second factor yields the family \begin{tikzcd} {\mathcal{X}}_L(\Delta) \ar[rr, bend left, "{\bm{\pi}}_L"] \ar[r, "{\bm{\pi}}_{1,L}"] & {\Sigma}\times {\mathbf{B}}\ar[r, "{\bm{\pi}}_{2,L}"] & {\mathbf{B}}\end{tikzcd} of threefolds with a C{\mathbf{C}}-action over the Hitchin base B{\mathbf{B}}. Analogously, pulling back S~{\tilde{\bm{S}}} along the natural map Σ~→U~{\bm{\tilde{\Sigma}}}\to {\bm{\tilde{U}}} gives the family \begin{tikzcd} \tilde{{\mathcal{X}}}_L(\Delta) \ar[rr, bend left, "\tilde{{\bm{\pi}}}_L"] \ar[r, "\tilde{{\bm{\pi}}}_{1,L}"] & {\bm{\tilde{\Sigma}}}\ar[r, "\tilde{{\bm{\pi}}}_{2,L}"] & {\mathbf{B}}\end{tikzcd} of threefolds with C{\mathbf{C}}-action over B{\mathbf{B}}.

    [thm:cy3s] Let Δ\Delta be an irreducible Dynkin diagram and LL be a spin bundle of Σ{\Sigma}. Then \begin{tikzcd} {\mathcal{X}}_L(\Delta)\ar[r] & {\mathbf{B}}({\Sigma}, \Delta) & \ar[l] \tilde{{\mathcal{X}}}_L(\Delta) \end{tikzcd} are algebraic families of quasi-projective Gorenstein threefolds with C{\mathbf{C}}-trivial canonical class. They are smooth over the locus B∘(Σ,Δ)⊂B(Σ,Δ){\mathbf{B}^\circ}({\Sigma},\Delta)\subset {\mathbf{B}}({\Sigma}, \Delta) of smooth cameral curves.

    The projections U→Σ{\bm{U}}\to {\Sigma} and S→Σ{\bm{S}}\to {\Sigma} are quasi-projective morphisms because they are (affine) bundles over the projective curve Σ{\Sigma}. Therefore the morphism σ:S→U{\bm{\sigma}}: {\bm{S}}\to {\bm{U}} which factorizes as \begin{tikzcd} {\bm{S}}\ar[d, "{\bm{\sigma}}"] \ar[dr] & \\ {\bm{U}}\ar[r ] & {\Sigma}\end{tikzcd} is quasi-projective as well. Hence π:X→B{\bm{\pi}}:{\mathcal{X}}\to {\mathbf{B}} is quasi-projective. Using the fact that σ:S→U{\bm{\sigma}}: {\bm{S}}\to {\bm{U}} is a Gorenstein morphism, a similar argument shows that each Xb=π−1(b)X_b={\bm{\pi}}^{-1}(b), b∈Bb\in {\mathbf{B}}, is Gorenstein.
    To see the statement about the canonical class, let j:X→Sj:{\mathcal{X}}\to {\bm{S}} be the natural morphism obtained from base change. Then the section ν^\hat{\bm{\nu}} of Lemma [l:bhatnu] pulls back to (dropping LL from the notation) s:=j∗ν^∈H0(X,Kπ1⊗(pr1∘π1)∗KΣ).\bm{s}:=j^*\hat{\bm{\nu}}\in H^0({\mathcal{X}}, K_{{\bm{\pi}}_1} \otimes (\mathrm{pr}_1\circ {\bm{\pi}}_1)^*K_\Sigma). Base change and the adjunction formula yields the isomorphism (Kπ1⊗(pr1∘π1)∗KΣ)∣Xb≅Kπb⊗πb∗KΣ≅KXb.(K_{{\bm{\pi}}_1}\otimes (\mathrm{pr}_1\circ {\bm{\pi}}_1)^*K_{\Sigma})_{|X_b}\cong K_{\pi_b}\otimes \pi_b^*K_{\Sigma}\cong K_{X_b}. Hence each restriction sb:=s∣Xbs_b:=\bm{s}_{|X_b} is a nowhere vanishing section of the locally free sheaf XbX_b. Since the C{\mathbf{C}}-action on XbX_b is pulled back from the C{\mathbf{C}}-action on S{\bm{S}}, sbs_b is C{\mathbf{C}}-invariant by Lemma [l:bhatnu]. Finally, the statement about smoothness follows from Proposition [p:transversal].
    The proof for X~L(Δ)\tilde{{\mathcal{X}}}_L(\Delta) works in complete analogy by replacing s\bm{s} with s~:=j~∗ω^∈H0(X~,Kπ~1⊗(pr1∘p1)∗KΣ)\tilde{\bm{s}}:=\tilde{j}^*\hat{{\bm{\omega}}}\in H^0(\tilde{{\mathcal{X}}}, K_{\tilde{{\bm{\pi}}}_1}\otimes (\mathrm{pr}_1\circ \bm{p}_1)^*K_{\Sigma}) for the natural morphism j~:X~→S~\tilde{j}:\tilde{{\mathcal{X}}}\to {\tilde{\bm{S}}}.

    We close this subsection with the deformation-theoretic meaning of the Hitchin base B=B(Σ,Δ){\mathbf{B}}={\mathbf{B}}({\Sigma}, \Delta) for the family X→B{\mathcal{X}}\to {\mathbf{B}} (see [@DDD] for the A1\mathrm{A}_1-case). To this end, observe that the central fiber X0=L×C∗S0ˉX_0=L\times_{{\mathbb{C}}^*} S_{\bar{0}} has a curve of Δ\Delta-singularities along Σ↪X0{\Sigma}\hookrightarrow X_0. It is easily seen from the construction that XbX_b, b≠0b\neq 0, only has isolated singularities. In particular, the deformation X→B{\mathcal{X}}\to {\mathbf{B}} of X0X_0 is not a locally trivial one. More precisely, we have:

    [thm:defX0] Let X=X0X=X_0 be the central fiber of a family X→B{\mathcal{X}}\to {\mathbf{B}} of quasi-projective Gorenstein Calabi-Yau with C{\mathbf{C}}-trivial canonical class. Then Ext1(ΩX1,OX)/H1(X,TX)≅B,\mathrm{Ext}^1(\Omega_X^1,{\mathcal{O}}_X)/H^1(X,T_X) \cong {\mathbf{B}}, i.e. the space of C{\mathbf{C}}-deformations of XX modulo locally trivial C{\mathbf{C}}-deformations is isomorphic to the corresponding Hitchin base.

    We first consider the case C=1{\mathbf{C}}=1 and set r:=rk(g)r:=\mathrm{rk}({\mathfrak{g}}). Since π:X→Σ\pi:X\to {\Sigma} is affine, the Leray spectral sequence implies that H2(X,TX)=0H^2(X,T_X)=0. Therefore the local-to-global spectral sequence implies Ext1(ΩX1,OX)/H1(X,TX)≅H0(X,E\mathcalligraxt OX1(ΩX1,OX)).\mathrm{Ext}^1(\Omega_X^1, {\mathcal{O}}_X)/H^1(X,T_X)\cong H^0(X, {\mathscr{E}\hspace{-3.3pt}\mathcalligra{xt}_{\,\mathcal{O}_X}}^1(\Omega_X^1,{\mathcal{O}}_X)). If k:Σ↪Xk:{\Sigma}\hookrightarrow X is the closed imbedding, then we claim \labeleq:ext1E:=E\mathcalligraxt OX1(ΩX1,OX)≅k∗U.\label{eq:ext1} \mathscr{E}:={\mathscr{E}\hspace{-3.3pt}\mathcalligra{xt}_{\,\mathcal{O}_X}}^1(\Omega_X^1,{\mathcal{O}}_X)\cong k_*{\bm{U}}. As a first step, we realize XX as a regular imbedding. Let πS:S→Σ\pi_{{\bm{S}}}:{\bm{S}}\to {\Sigma} be the natural projection and πS∗U→S\pi_{{\bm{S}}}^*{{\bm{U}}} \to {\bm{S}} the induced vector bundle. Then there exists a unique section τ:S→πS∗U\tau:{\bm{S}}\to \pi^*_{{\bm{S}}} {\bm{U}} such that prU∘τ=σpr_{{\bm{U}}}\circ \tau={\bm{\sigma}} for the natural projection prU:πS∗U→Upr_{{\bm{U}}}:\pi^*_{{\bm{S}}}{\bm{U}}\to {\bm{U}}. In a trivialization of S∣D≅D×S{\bm{S}}_{|D}\cong D\times S for an affine D⊂ΣD\subset {\Sigma}, the section is given by D×S→(D×S)×t/W,(x,s)↦((x,s),σ(s)).D\times S \to (D\times S) \times {\mathfrak{t}}/W, \quad (x,s)\mapsto ((x,s),\sigma(s)). Therefore the vanishing locus Z(τ)⊂SZ(\tau)\subset {\bm{S}} coincides with i:X↪Si:X\hookrightarrow {\bm{S}} and is a regular imbedding. The latter implies that the normal sheaf satisifes \labeleq:normalsheafNX/S≅πS∗U∣X,\label{eq:normalsheaf} \mathcal{N}_{X/{\bm{S}}}\cong \pi^*_{{\bm{S}}}{\bm{U}}_{|X}, cf. [@FultonLang], Chapter IV. In order to compute E\mathscr{E}, let J\mathcal{J} be the ideal sheaf defining X⊂SX\subset {\bm{S}}. Since the latter is a regular imbedding, we have the locally free resolution \begin{tikzcd} 0 \ar[r] & \mathcal{J}/\mathcal{J}^2 \ar[r] & i^*\Omega^1_{{\mathcal{S}}} \ar[r] & 0 \end{tikzcd} of ΩX1\Omega_X^1. Dualizing gives the sequence \begin{tikzcd} 0 \ar[r] & {\mathscr{H}\hspace{-5pt}\mathcalligra{om}}_{{\mathcal{O}}_X}(i^*\Omega^1_{{\bm{S}}},{\mathcal{O}}_X) \ar[r, "\varphi"] & \mathcal{N}_{X/{\bm{S}}} \ar[r] & 0 \end{tikzcd} so that coker(φ)=E\mathcalligraxt OX1(ΩX1,OX)\mathrm{coker}(\varphi)={\mathscr{E}\hspace{-3.3pt}\mathcalligra{xt}_{\,\mathcal{O}_X}}^1(\Omega_X^1,{\mathcal{O}}_X) which is supported on Σ⊂X{\Sigma}\subset X. In order to prove ([eq:ext1]) it therefore suffices to show that E\mathscr{E} has constant rank rr along Σ{\Sigma} in light of ([eq:normalsheaf]).
    Let U:=D×S0ˉ⊂XU:=D\times S_{\bar{0}}\subset X and D⊂ΣD\subset {\Sigma} be an affine open in XX and Σ{\Sigma} respectively. Hence U=Spec(B),B=C[t,x,y,z]/⟨f(x,y,z)⟩,D=Spec(C),C=C[t,x,y,z],U=\mathrm{Spec}(B), \quad B={\mathbb{C}}[t,x,y,z]/\langle f(x,y,z) \rangle,\quad D=\mathrm{Spec}(C), \quad C={\mathbb{C}}[t,x,y,z], where ff defines the Δ\Delta-singularity S0ˉS_{\bar{0}}. The sheaf E\mathscr{E} corresponds to the BB-module ExtB1(ΩB1,B)≅C[t,x,y,z]/⟨f,∂f⟩.\mathrm{Ext}^1_B(\Omega^1_B, B) \cong {\mathbb{C}}[t,x,y,z]/\langle f, \partial f \rangle. As a CC-module, it is isomorphic to C[t]rC[t]^r. This concludes the proof for C=1{\mathbf{C}}=1. If C≠1{\mathbf{C}}\neq 1, the same proof goes through if we work with C{\mathbf{C}}-invariants and recall that BhC=B{\mathbf{B}}_h^{\mathbf{C}}={\mathbf{B}}.

    Relation between and

    We next compare the two families X=XL(Δ)→B=B(Σ,Δ)Xh=XL(Δh)→Bh=B(Σ,Δh)\begin{gathered} {\mathcal{X}}={\mathcal{X}}_L(\Delta) \to {\mathbf{B}}={\mathbf{B}}({\Sigma}, \Delta)\\ {\mathcal{X}}_h={\mathcal{X}}_L(\Delta_h)\to {\mathbf{B}}_h={\mathbf{B}}({\Sigma}, \Delta_h)\end{gathered} in case Δ≠Δh\Delta\neq \Delta_h. As a first step, we construct a non-trivial AS(Δ)AS(\Delta)-action on Xh{\mathcal{X}}_h.
    Let Sh:=S(Δh)=x+ker⁡(ad(y))⊂gh:=g(Δh)S_h:=S(\Delta_h)=x+\ker(\mathrm{ad}(y))\subset {\mathfrak{g}}_h:={\mathfrak{g}}(\Delta_h) be a Slodowy slice. Then the group \labeleq:CACA(x,h):={ϕ∈Aut⁡(gh) ∣ ϕ(x)=x,ϕ(h)=h}\label{eq:CA} CA(x,h):=\{ \phi \in \operatorname{Aut}({\mathfrak{g}}_h) ~|~\phi(x)=x, \phi(h)=h \} acts on ShS_h. There is a subgroup CA⊂CA(x,h){\mathbf{CA}}\subset CA(x,h) which is isomorphic to AS(Δ)AS(\Delta) (see [@Slo], 7.5). Of course, the definition of CA{\mathbf{CA}} and C{\mathbf{C}} makes sense for SS and ShS_h respectively. Then CA≅C{\mathbf{CA}}\cong {\mathbf{C}} in the former and C=1{\mathbf{C}}=1 in the latter case.

    Even though C≅AS(Δ)≅CA{\mathbf{C}}\cong AS(\Delta) \cong {\mathbf{CA}}, we often write C{\mathbf{C}} and CA{\mathbf{CA}} to emphasize how the AS(Δ)AS(\Delta)-action is realized. To illustrate the relation between SS and ShS_h together with the corresponding AS(Δ)AS(\Delta)-actions, we give a worked out example in Appendix [a:ex].

    The CA{\mathbf{CA}}-action on ShS_h induces a CA{\mathbf{CA}}-action on th/Wh{\mathfrak{t}}_h/W_h such that σh:Sh→th/Wh\sigma_h:S_h\to {\mathfrak{t}}_h/W_h is equivariant. Since the CA{\mathbf{CA}}-action commutes with the C∗{\mathbb{C}}^*-action on ShS_h, we obtain a CA{\mathbf{CA}}-action on Xh{\mathcal{X}}_h. Lemma [l:bhatnu] and Theorem [thm:cy3s] analogously hold for the CA{\mathbf{CA}}-action (cf. [@Beck-thesis] for details). However, CA{\mathbf{CA}} acts non-trivially on the base so that a general member Xh,bX_{h,b} does not have CA{\mathbf{CA}}-trivial canonical bundle.

    [cor:comp] With the previous notation, \labeleq:inclB≅BhCA↪Bh\label{eq:incl} {\mathbf{B}}\cong {\mathbf{B}}_h^{{\mathbf{CA}}}\hookrightarrow {\mathbf{B}}_h but B∘∩Bh∘=∅{\mathbf{B}}^\circ \cap {\mathbf{B}}_h^\circ =\emptyset. The family X→B{\mathcal{X}}\to {\mathbf{B}} is the restriction of the family Xh→Bh{\mathcal{X}}_h\to {\mathbf{B}}_h under the inclusion ([eq:incl]) and the AS(Δ)AS(\Delta)-action on X{\mathcal{X}} is induced by the AS(Δ)AS(\Delta)-action on Xh{\mathcal{X}}_h. In particular, Xh{\mathcal{X}}_h is smooth over a Zariski open and dense subset containing B∘⊔B∘h⊂Bh{\mathbf{B}^\circ}\sqcup {\mathbf{B}^\circ}_h\subset {\mathbf{B}}_h.

    It is not difficult to see that t/W≅(th/Wh)CA{\mathfrak{t}}/W\cong ({\mathfrak{t}}_h/W_h)^{{\mathbf{CA}}} which gives the inclusion i:t/W↪th/Whi:{\mathfrak{t}}/W \hookrightarrow {\mathfrak{t}}_h/W_h and hence ([eq:incl]). To see that B∘∩Bh∘=∅{\mathbf{B}}^\circ \cap {\mathbf{B}}_h^\circ=\emptyset, let α∈Δ\alpha\in \Delta be a long root. Under folding, it corresponds to an AS(Δ)AS(\Delta)-orbit O(β)O(\beta) of length ≥2\geq 2 for some β∈Δh\beta\in \Delta_h. If tα⊂t{\mathfrak{t}}_\alpha\subset {\mathfrak{t}} is the fixed point locus of sα∈Ws_\alpha\in W (and analogously for th{\mathfrak{t}}_h), then \labeleq:talphatα=⋂β′∈O(β)th,β′⊂th\label{eq:talpha} {\mathfrak{t}}_\alpha=\bigcap_{\beta'\in O(\beta)} {\mathfrak{t}}_{h,\beta'}\subset {\mathfrak{t}}_h under the inclusion t⊂th{\mathfrak{t}}\subset {\mathfrak{t}}_h. If b∈B∘b\in {\mathbf{B}^\circ}, then bb necessarily maps to the smooth locus \mboxdisc(q)sm\mbox{disc}(\bm{q})^{sm} of the discriminant of q:U~→U\bm{q}:{\bm{\tilde{U}}}\to {\bm{U}}. Since Σ~b{\tilde{\Sigma}}_b necessarily intersects KΣ⊗tα⊂U~K_{\Sigma}\otimes {\mathfrak{t}}_{\alpha}\subset {\bm{\tilde{U}}} and because of ([eq:talpha]), bb cannot map to \mboxdisc(qh)sm\mbox{disc}(\bm{q}_h)^{sm} under the inclusion U⊂Uh{\bm{U}}\subset {\bm{U}}_h. Hence B∘∩B∘h=∅{\mathbf{B}^\circ}\cap {\mathbf{B}^\circ}_h=\emptyset.
    Slodowy has proven (Chapter 8.8 in [@Slo]) that i∗Sh≅Si^*S_h\cong S over t/W{\mathfrak{t}}/W as C∗{\mathbb{C}}^*-deformations of the Δ\Delta-singularity over 0ˉ∈t/W\bar{0}\in {\mathfrak{t}}/W. By the C∗{\mathbb{C}}^*-equivariance, this isomorphism induces the isomorphism X≅Xh{\mathcal{X}}\cong {\mathcal{X}}_h over B⊂Bh{\mathbf{B}}\subset {\mathbf{B}}_h.

    Equations

    To connect to similar constructions in the physical mathematics literature ([@DDD], [@Tbranes], [@Tbranes2]), we give explicit equations of the families X→B{\mathcal{X}}\to {\mathbf{B}} in appropriate bundles over Σ{\Sigma}.
    If we identify S=≅Cr+2S=\cong {\mathbb{C}}^{r+2} appropriately, then the weights of the C∗{\mathbb{C}}^*-action are given by w1=2d1,…,wr=2dr,wr+1,wr+2,wr+3w_1=2d_1,\dots, w_r=2d_r, w_{r+1},w_{r+2},w_{r+3} for the degrees dj=deg(χj)d_j=\mathrm{deg}(\chi_j) as in Section [ss:hitchinbase] and weight wr+1,wr+2,wr+3w_{r+1},w_{r+2},w_{r+3} as in Table [ADEsings] in Appendix [SectFolding]. It follows that S=L×C∗S≅⨁j=1rKdj⊕⨁j=13Lwr+k.{\bm{S}}=L\times_{{\mathbb{C}}^*} S \cong \bigoplus_{j=1}^{r} K^{d_j} \oplus \bigoplus_{j=1}^3 L^{w_{r+k}}. Let f∈C[x,y,z]f\in {\mathbb{C}}[x,y,z] define the Δ\Delta-singularity and let g1,…,gr∈C[x,y,z]Cg_1,\dots, g_r\in {\mathbb{C}}[x,y,z]^{{\mathbf{C}}} be representatives of generators of the C{\mathbf{C}}-invariant Jacobi ring C[x,y,z]C/(f,∂f){\mathbb{C}}[x,y,z]^{{\mathbf{C}}}/(f,\partial f). Then the family X→B{\mathcal{X}}\to {\mathbf{B}} reads as \labeleq:explicitX={(x,y,z,b‾) ∣ f(x,y,z)+∑i=1rbigi(x,y,z)=0∈tot(Kd)}\label{eq:explicit} {\mathcal{X}}=\{ (x,y,z,\underline{b})~|~f(x,y,z)+\sum_{i=1}^r b_i g_i(x,y,z)=0 \in \mathrm{tot}(K^d) \} in tot(⨁i=13Lwr+k)×B\mathrm{tot}(\bigoplus_{i=1}^3 L^{w_{r+k}})\times {\mathbf{B}} with the obvious projection and where dd is again as in Table [ADEsings].

    [ex:B2] Let us give at least one simple example (also compare with Appendix [a:ex]). In this case f(x,y,z)=x4−yzf(x,y,z)=x^4-yz and C=Z/2Z{\mathbf{C}}={\mathbb{Z}}/2{\mathbb{Z}} acts as (x,y,z)↦(−x,z,y)(x,y,z)\mapsto (-x,z,y). Therefore ([eq:explicit]) reads as X={(x,y,z,(b1,b2)) ∣ x4−yz+b1x2+b2=0∈tot(K4)}{\mathcal{X}}=\{ (x,y,z, (b_1,b_2))~|~x^4-yz+b_1 x^2+ b_2=0 \in \mathrm{tot}(K^4) \} in tot(K⊕K2⊕K2)×B\mathrm{tot}(K\oplus K^2\oplus K^2)\times {\mathbf{B}} with B=H0(Σ,K2)⊕H0(Σ,K4){\mathbf{B}}=H^0({\Sigma}, K^2)\oplus H^0({\Sigma}, K^4).

    Simultaneous resolutions

    The families π:X→B{\bm{\pi}}:{\mathcal{X}}\to {\mathbf{B}} and π~:X~→B\tilde{{\bm{\pi}}}:\tilde{{\mathcal{X}}}\to {\mathbf{B}} share many features of the Slodowy slice SS. It seems unlikely though that our methods provide an explicit description of a simultaneous resolution for these families. However, we succeed over the locus B~/W⊂B\tilde{{\mathbf{B}}}/W\subset {\mathbf{B}} (see ([eq:B/W])) which is in the singular locus of π{\bm{\pi}}.
    To see this, let ev~:Σ×B~→B~\tilde{ev}:{\Sigma}\times \tilde{{\mathbf{B}}}\to \tilde{{\mathbf{B}}} be the evalution map which gives rise to the family \begin{tikzcd} \tilde{ev}^*\tilde{{\bm{S}}}=\hat{{\mathcal{X}}}_L \ar[rr, bend left, "\hat{{\bm{\pi}}}_L"] \ar[r, "\hat{{\bm{\pi}}}_{1,L}"] & {\Sigma}\times \tilde{{\mathbf{B}}} \ar[r, "\hat{{\bm{\pi}}}_{2,L}"] & \tilde{{\mathbf{B}}} \end{tikzcd} of threefolds with a non-trivial C{\mathbf{C}}-action. With the methods of the proof of Theorem [thm:cy3s], one proves that π^L:X^L→B~\hat{{\bm{\pi}}}_L:\hat{{\mathcal{X}}}_L\to \tilde{{\mathbf{B}}} is a smooth family of quasi-projective threefolds with C{\mathbf{C}}-trivial canonical class.

    [p:simresol] The family π^:X^→B~\hat{{\bm{\pi}}}:\hat{{\mathcal{X}}}\to \tilde{{\mathbf{B}}} is a simultaneous C{\mathbf{C}}-resolution of p∗Xp^*{\mathcal{X}} where p:B~→B~/W⊂Bp:\tilde{{\mathbf{B}}}\to \tilde{{\mathbf{B}}}/W\subset {\mathbf{B}} is the natural projection. It descends to a simultaneous small C{\mathbf{C}}-resolution over B~/W−{0}⊂B\tilde{{\mathbf{B}}}/W-\{0\} \subset {\mathbf{B}}.

    Using the fiber product property, we obtain a unique C{\mathbf{C}}-equivariant morphism X^→p∗X\hat{{\mathcal{X}}}\to p^*{\mathcal{X}} over Σ×B~{\Sigma}\times \tilde{{\mathbf{B}}} which makes the obvious diagrams commute.
    For the second claim, observe that the C{\mathbf{C}}-family X^/W→B~/W\hat{{\mathcal{X}}}/W\to \tilde{{\mathbf{B}}}/W is still smooth over B∗:=B~/W−{0}⊂B{\mathbf{B}}^*:=\tilde{{\mathbf{B}}}/W-\{0\}\subset {\mathbf{B}}. Then the morphisms X^→p∗X\hat{{\mathcal{X}}}\to p^*{\mathcal{X}} descends to the C{\mathbf{C}}-morphism X^/W→p∗X/W≅X\hat{{\mathcal{X}}}/W\to p^*{\mathcal{X}}/W\cong {\mathcal{X}} over B∗{\mathbf{B}}^*. It remains to show that (X^/W)b→Xb(\hat{X}/W)_{b}\to X_b is a small resolution for b∈B∗b\in {\mathbf{B}}^*. Let b~\tilde{b} with p(b~)=bp(\tilde{b})=b. Then XbX_b is only singular at the singularities of the fibers pb−1(x)p_b^{-1}(x) for x∈Div(∏α∈Rα(b))x\in \mathrm{Div}(\prod_{\alpha\in R} \alpha(b) ) and the projection pb:Xb→Σp_b:X_b\to {\Sigma}. By assumption ∏α∈Rα(b)≠0\prod_{\alpha\in R} \alpha (b)\neq 0, i.e. the singularities are isolated, so that (X^/W)b→Xb(\hat{X}/W)_b\to X_b is a small resolution.

    The last statement is false if we include 0∈B0\in {\mathbf{B}}: Then X0≅L×C∗YX_0\cong L\times_{{\mathbb{C}}^*} Y for the Δ\Delta-singularity YY and X^0≅L×C∗Y^\hat{X}_0\cong L\times_{{\mathbb{C}}^*} \hat{Y} for its minimal resolution Y^→Y\hat{Y}\to Y. Hence it is not a small resolution.

    The singularities of XbX_b, b∈B∗b\in {\mathbf{B}}^*, are precisely the Gorenstein threefold singularities studied in [@KatzMorr]. The simplest (local) example is \begin{tikzcd} &\hat{X}\ar[d] =\{ (x,y,z,t,[u:v])\in {\mathbb{C}}^4\times \mathbb{P}^1~|~x^2+y^2+z^2-t^2=0~,~xv=u(z+t)\} \\ &X =\{ (x,y,z,s)\in {\mathbb{C}}^4~|~x^2+y^2+z^2-s=0 \} \end{tikzcd} for the small resolution is (x,y,z,t,[u:v])↦(x,y,z,t2)(x,y,z,t,[u:v])\mapsto (x,y,z,t^2).

    Calabi-Yau orbifolds and Hitchin systems {#s:mhs}

    Let X→B{\mathcal{X}}\to {\mathbf{B}} be a family of quasi-projective Calabi-Yau threefolds with C{\mathbf{C}}-trivial canonical class as in the previous section. For each b∈Bb\in {\mathbf{B}}, we denote by [Xb/C][X_b/{\mathbf{C}}] the corresponding quotient stack and refer to it as a Calabi-Yau orbifold. By construction, they fit into the family πC∘:[X/C]→B {\bm{\pi}}_{{\mathbf{C}}}^\circ:[{\mathcal{X}}/{\mathbf{C}}]\to {\mathbf{B}} of Calabi-Yau orbifolds. Before we study the integral (equivariant) cohomology groups H3([Xb/C],Z)=HC3(Xb,Z),b∈B∘,H^3([X_b/{\mathbf{C}}],{\mathbb{Z}})=H^3_{{\mathbf{C}}}(X_b,{\mathbb{Z}}),\quad b\in {\mathbf{B}^\circ}, we need a general result.

    MHS on equivariant cohomology {#ss:mhseq}

    Let GG be a finite group acting on a locally compact topological space XX. We replace the orbifold stack [X/G][X/G] by the simplicial space [X/G]∙=((Gp+1×X)/G)p≥0[X/G]_\bullet=((G^{p+1}\times X)/G)_{p\geq 0} where GG acts on Gp+1×XG^{p+1}\times X via g⋅(g0,…,gp,x)=(g0g−1,…,gpg−1,g⋅x).g\cdot (g_0,\dots, g_p, x)=(g_0 g^{-1},\dots, g_p g^{-1}, g\cdot x). The simplicial structure maps of [X/G]∙[X/G]_\bullet are induced by the standard simplicial structure on G∙+1G^{\bullet+1}. The importance of [X/G]∙[X/G]_\bullet is that it is a simplicial model for XG:=X×GEGX_G:=X \times_G EG which defines equivariant cohomology HGk(X,R)=Hk(XG,R),R=Z,Q.H_G^k(X,R)=H^k(X_G,R),\quad R={\mathbb{Z}},{\mathbb{Q}}. More precisely, we have \labeleq:eqcohorbHk([X/G]∙,R)≅Hk([X/G],R)=HGk(X,R),R=Z,Q.\label{eq:eqcohorb} H^k([X/G]_\bullet, R) \cong H^k([X/G],R)= H_G^k(X,R), \quad R={\mathbb{Z}},{\mathbb{Q}}. An important tool to compute the left-hand side is the spectral sequence \labelEqCohSSE1pq=Hq([X/G]p,R)⇒E∞k=Hk([X/G]∙,R)=HGk(X,R),R=Z,Q.\label{EqCohSS} E_1^{pq}=H^q([X/G]_p,R) \Rightarrow E_\infty^k=H^k([X/G]_\bullet, R)=H_G^k(X,R),\quad R={\mathbb{Z}},{\mathbb{Q}}. By [@DeligneIII], this spectral sequence is a spectral sequence of MHS if XX is a complex algebraic variety (considered in the analytic topology) on which GG acts algebraically. In particular, the equivariant cohomology groups HGk(X,R)H^k_G(X,R) carry natural MHS.
    The Leray (or Serre) spectral sequence for the fibration XG→BGX_G\to BG implies that HGk(X,Q)≅Hk(X,Q)GH_G^k(X,{\mathbb{Q}})\cong H^k(X,{\mathbb{Q}})^G as abelian groups. Of course, both sides carry MHS and the next lemma shows that they agree. This is well-known to experts6 but for lack of a reference, we give a proof here for completeness.

    [lem:EqCohInv] Let GG be a finite group acting on a complex algebraic variety XX. Then the spectral sequence ([EqCohSS]) degenerates on the E2E_2-page and yields an isomorphism HGk(X,Q)≅Hk(X,Q)GH_G^k(X,{\mathbb{Q}})\cong H^k(X,{\mathbb{Q}})^G of Q{\mathbb{Q}}-MHS for each k≥0k\geq 0.

    We express E1pq=Hq((Gp+1×X)/G,Z)E_1^{pq}=H^q((G^{p+1}\times X)/G,{\mathbb{Z}}) of ([EqCohSS]) as follows: E1pq=Hq(Gp+1×X,Z)G(\mboxfreenessofaction)=Hom⁡Z[G](Z,Hq(Gp+1×X,Z))=Hom⁡Z[G](Z,H0(Gp+1,Z)⊗ZHq(X,Z))(\mboxKu¨nnethformula)=Hom⁡Z[G](Z,Hom⁡Z(H0(Gp+1,Z),Hq(X,Z))))(\mboxfreenessofH0(Gp+1,Z))=Hom⁡Z(Z⊗Z[G]H0(Gp+1,Z),Hq(X,Z))(\mboxadjunction)=Hom⁡Z[G](H0(Gp+1,Z),Hq(X,Z)).\begin{aligned} E_1^{pq}=& H^q(G^{p+1}\times X,{\mathbb{Z}})^G & (\mbox{freeness of action}) \\ =& \operatorname{Hom}_{{\mathbb{Z}}[G]} ({\mathbb{Z}}, H^q(G^{p+1}\times X,{\mathbb{Z}})) & \\ =& \operatorname{Hom}_{{\mathbb{Z}}[G]} ({\mathbb{Z}}, H^0(G^{p+1},{\mathbb{Z}})\otimes_{{\mathbb{Z}}} H^{q}(X,{\mathbb{Z}}) ) & (\mbox{K\"unneth formula}) \\ =& \operatorname{Hom}_{{\mathbb{Z}}[G]} ({\mathbb{Z}}, \operatorname{Hom}_{{\mathbb{Z}}} (H_0(G^{p+1},{\mathbb{Z}}), H^q(X,{\mathbb{Z}})))) & (\mbox{freeness of }H_0(G^{p+1},{\mathbb{Z}})) \\ =& \operatorname{Hom}_{{\mathbb{Z}}}({\mathbb{Z}}\otimes_{{\mathbb{Z}}[G]} H_0(G^{p+1},{\mathbb{Z}}), H^q(X,{\mathbb{Z}})) & (\mbox{adjunction}) \\ =& \operatorname{Hom}_{{\mathbb{Z}}[G]}(H_0(G^{p+1},{\mathbb{Z}}),H^q(X,{\mathbb{Z}})).\end{aligned} In the last line we have used the adjunction between the trivial module functor and −⊗ZZ[G]-\otimes_{{\mathbb{Z}}} {\mathbb{Z}}[G] ([@Weibel]). Since the complex H0(G∙+1,Z)H_0(G^{\bullet+1}, {\mathbb{Z}}) coincides with the bar resolution B∙→ZB_\bullet \to {\mathbb{Z}}, we see that E1∙q=Hom⁡Z[G](B∙,Hq(X,Q))E_1^{\bullet q}=\operatorname{Hom}_{{\mathbb{Z}}[G]}(B_\bullet, H^q(X,{\mathbb{Q}})) as complexes for each q≥0q\geq 0. In particular, E1∙qE_1^{\bullet q} computes group cohomology Hk(G,Hq(X,Z))H^k(G, H^q(X,{\mathbb{Z}})) for each q≥0q\geq 0 so that E2pq≅Hp(G,Hq(X,Q))E_2^{pq}\cong H^p(G,H^q(X,{\mathbb{Q}})). But GG is finite and we work over Q{\mathbb{Q}} so that Hp(G,Hq(X,Q))=0H^p(G,H^q(X,{\mathbb{Q}}))=0 for all p≥1p\geq 1. Hence the spectral sequence ([EqCohSS]) of MHS degenerates on the E2E_2-page to give isomorphisms Hk(X,Q)G≅Hk([X/G]∙,Q)=HGk(X,Q)H^k(X,{\mathbb{Q}})^G\cong H^k([X/G]_\bullet,{\mathbb{Q}})=H_G^k(X,{\mathbb{Q}}) of MHS.

    CY orbifolds and Hitchin systems

    We begin with a general result on the Leray spectral sequence for equivariant maps.

    [lem:leraygroup] Let GG be a discrete group acting on topological spaces X,Y,ZX,Y,Z. Further let f:X→Yf:X\to Y, h:Y→Zh:Y\to Z be GG-equivariant morphisms. Then the Leray spectral sequence \labelLeraySSRpg∗Rqf∗AX⇒Rp+q(g∘f)∗AX\label{LeraySS} R^pg_* R^qf_*A_X \Rightarrow R^{p+q}(g\circ f)_*A_X for any constant sheaf AXA_X of an abelian group AA on XX lifts via the forgetful functor For:ShG(Z)→Sh(Z)\mathrm{For}:Sh_G(Z)\to Sh(Z) to GG-equivariant abelian sheaves ShG(Z)Sh_G(Z) on ZZ.

    First of all, the exact forgetful functor For:ShG(W)→Sh(W)\mathrm{For}:Sh_G(W)\to Sh(W) for W=X,Y,ZW=X,Y,Z commutes with the equivariant direct image functor f∗eqf^{\mathrm{eq}}_*, e.g. For∘f∗eq≃f∗∘For\mathrm{For}\circ f^{\mathrm{eq}}_*\simeq f_*\circ \mathrm{For}. Since ShG(W)Sh_G(W) has enough injectives, we can form derived direct image functors, e.g. Rf∗eq:DGb(X)≃Db(ShG(X))→DGb(Y)≃Db(ShG(Y)).Rf^{\mathrm{eq}}_*:D^b_G(X)\simeq D^b(Sh_G(X))\to D^b_G(Y)\simeq D^b(Sh_G(Y)). Again they commute with the exact forgetful functors which implies the claim.

    [p:isoeq] Let X=XbX=X_b, b∈B∘b\in {\mathbf{B}^\circ}, be a smooth quasi-projective Calabi-Yau threefold with C{\mathbf{C}}-action as in Section [s:cy]. Then there is a natural isomorphism H3([X/C],Z)=HC3(X,Z)≅H3(X,Z)CH^3([X/{\mathbf{C}}],{\mathbb{Z}})=H^3_{{\mathbf{C}}}(X,{\mathbb{Z}})\cong H^3(X,{\mathbb{Z}})^{{\mathbf{C}}} of Z{\mathbb{Z}}-MHS.

    We emphasize that we work over the integers giving a stronger result as Lemma [lem:EqCohInv]. In general, such a result is false due to torsion, cf. Example [ex:eqcoh] below.

    As in the proof of Lemma [lem:EqCohInv], we employ the spectral sequence ([EqCohSS]). In that proof we have seen that its E2E_2-page is of the form E2pq=Hp(C,Hq(X,Z)).E_2^{pq}=H^p({\mathbf{C}}, H^q(X,{\mathbb{Z}})). We show that it degenerates on the E3E_3-page in this situation by proving \labelE3degHp(C,Hq(X,Z))=0,∀p≥1,q∉{3,4}.\label{E3deg} H^p({\mathbf{C}},H^q(X,{\mathbb{Z}}))=0,\quad \forall p\geq 1, q\notin \{3,4 \}. Of course this is automatic if we worked over Q{\mathbb{Q}}. To show it over the integers, we observe that H0(X,Z)≅H0(Σ,π∗Z),H1(X,Z)≅H1(Σ,π∗Z),H2(X,Z)≅H2(Σ,π∗Z),H3(X,Z)≅H1(Σ,R2π∗Z),H4(X,Z)≅H2(Σ,R2π∗Z),Hq(X,Z)=0∀q≥5\begin{gathered} H^0(X,{\mathbb{Z}})\cong H^0({\Sigma}, \pi_*{\mathbb{Z}}),\quad H^1(X,{\mathbb{Z}})\cong H^1({\Sigma},\pi_*{\mathbb{Z}}), \quad H^2(X,{\mathbb{Z}})\cong H^2({\Sigma},\pi_*{\mathbb{Z}}), \\ H^3(X,{\mathbb{Z}})\cong H^1({\Sigma}, R^2\pi_*{\mathbb{Z}}), \quad H^4(X,{\mathbb{Z}})\cong H^2({\Sigma}, R^2\pi_*{\mathbb{Z}}), \quad H^q(X,{\mathbb{Z}})=0 \quad\forall q\geq 5\end{gathered} for the projection π:X→Σ\pi: X\to {\Sigma}. This is seen by using the Leray spectral sequence (cf. [@DDP], [@Beck]). Since C{\mathbf{C}} acts trivially on π∗Z\pi_*{\mathbb{Z}}, it follows that C{\mathbf{C}} acts trivially on Hq(X,Z)H^q(X,{\mathbb{Z}}) for q∉{3,4}q\notin \{ 3,4\} by Lemma [lem:leraygroup]. This yields ([E3deg]) so that the spectral sequence ([EqCohSS]) gives the isomorphism E30,3=H3(X,Z)C≅E∞3=HC3(X,Z)E_3^{0,3}=H^3(X,{\mathbb{Z}})^{\mathbf{C}}\cong E_\infty^3=H^3_{{\mathbf{C}}}(X,{\mathbb{Z}}) of Z{\mathbb{Z}}-MHS.

    [ex:eqcoh] The analogous statement is false for the minimal resolution Y^→Y\hat{Y}\to Y of the Δ\Delta-singularity YY (assuming Δ≠Δh\Delta\neq \Delta_h): A group cohomology computation shows HC2(Y^,Z)≅H2(Y^,Z)C⊕Z/2Z.H^2_{{\mathbf{C}}}(\hat{Y},{\mathbb{Z}})\cong H^2(\hat{Y},{\mathbb{Z}})^{{\mathbf{C}}}\oplus {\mathbb{Z}}/2{\mathbb{Z}}.

    Since H3(Xb,Z)H^3(X_b,{\mathbb{Z}}), b∈B∘b\in {\mathbf{B}^\circ}, is up to a Tate twist a polarizable Z{\mathbb{Z}}-HS (see [@Beck], Lemma 5) of weight 11, it follows that the orbifold intermediate Jacobian \labeleq:eqj2J2([Xb/C])=H3([Xb/C],C)/(F2H3([Xb/C],C)+H3([Xb/C],Z))\label{eq:eqj2} J^2([X_b/{\mathbf{C}}])=H^3([X_b/{\mathbf{C}}],{\mathbb{C}})/\left(F^2H^3([X_b/{\mathbf{C}}],{\mathbb{C}})+H^3([X_b/{\mathbf{C}}],{\mathbb{Z}})\right) of the Calabi-Yau orbifold [Xb/C][X_b/{\mathbf{C}}] is an abelian variety.
    To globalize the previous discussion, we consider the augmented simplicial morphism associated to the family πC:[X/C]→B{\bm{\pi}}_{{\mathbf{C}}}:[{\mathcal{X}}/{\mathbf{C}}]\to {\mathbf{B}} of Calabi-Yau orbifolds. By abuse of notation, it is again denoted by \begin{tikzcd} {\bm{\pi}}_{{\mathbf{C}}}:[{\mathcal{X}}/{\mathbf{C}}]_\bullet \ar[r, "{\bm{\pi}}_{{\mathbf{C}},\bullet}"] & {\mathbf{B}}_\bullet \ar[r, "a"] & {\mathbf{B}}. \end{tikzcd} where the last arrow is the augmentation. In accordance with ([eq:eqcohorb]), we set RkπC,∗Z=sRk(πC,∙)∗Z,R^k{\bm{\pi}}_{{\mathbf{C}},*} {\mathbb{Z}}= \bm{s} R^k({\bm{\pi}}_{{\mathbf{C}},\bullet})_*{\mathbb{Z}}, compare [@DeligneIII], Section 5.2, where s\bm{s} stands for the total complex.

    Let πC∘:[X∘/C]→B∘{\bm{\pi}}_{{\mathbf{C}}}^\circ:[{\mathcal{X}}^\circ/{\mathbf{C}}] \to {\mathbf{B}^\circ} be the family of orbifolds associated with the smooth family π:X∘→B∘{\bm{\pi}}:{\mathcal{X}}^\circ\to {\mathbf{B}^\circ}. Then R3πC,∗∘ZR^3{\bm{\pi}}_{{\mathbf{C}},*}^\circ {\mathbb{Z}} carries the structure of a polarizable Z{\mathbb{Z}}-VMHS such that (R3πC,∗∘Z)b≅HC3(Xb,Z),b∈B∘,(R^3{\bm{\pi}}_{{\mathbf{C}},*}^\circ {\mathbb{Z}})_b\cong H^3_{{\mathbf{C}}}(X_b,{\mathbb{Z}}), \quad b\in {\mathbf{B}^\circ}, naturally as ZZ-MHS.

    This is a consequence of the more general treatment in [@BeckVMHS] on VMHS for simplicial smooth and quasi-projective morphisms (which are topologically locally trivial on each level). Alternatively, the V(M)HS-structure can be constructed directly in this present case by using an approximation of the classifying space BCB{\mathbf{C}} of C{\mathbf{C}} by finite-dimensional smooth projective varieties.

    Therefore the orbifold intermediate Jacobians ([eq:eqj2]) fit into the family J2([X∘/C])→B∘\mathcal{J}^2([{\mathcal{X}}^\circ/{\mathbf{C}}])\to {\mathbf{B}^\circ} of abelian varieties. Using the methods of [@Beck], it is seen to be an algebraic integrable system.

    Let Δ\Delta be an irreducible Dynkin diagram, C=AS(Δ){\mathbf{C}}=AS(\Delta) and G=G(Δ)G=G(\Delta) the corresponding simple adjoint complex Lie group. Further let X→B(Σ,Δ){\mathcal{X}}\to {\mathbf{B}}({\Sigma}, \Delta) be one of the families of Calabi-Yau threefolds of Section [s:cy] with C{\mathbf{C}}-action. Then J2([X∘/C])≅Higgs1∘(Σ,G)\mathcal{J}^2([{\mathcal{X}^\circ}/{\mathbf{C}}]) \cong {\mathbf{Higgs}}_1^\circ({\Sigma}, G) J2([X∘/C])≅Higgs1∘(Σ,LG).\mathcal{J}_2([{\mathcal{X}^\circ}/{\mathbf{C}}]) \cong {\mathbf{Higgs}}_1^\circ({\Sigma}, ^L G). as algebraic integrable systems over B∘(Σ,Δ){\mathbf{B}^\circ}({\Sigma}, \Delta).

    The spectral sequence ([EqCohSS]) globalizes to a spectral sequence of polarizable Z{\mathbb{Z}}-VMHS. Hence Proposition [p:isoeq] implies R3πC,∗∘Z≅(R3π∗Z)CR^3{\bm{\pi}}_{{\mathbf{C}},*}^\circ {\mathbb{Z}}\cong (R^3{\bm{\pi}}_*{\mathbb{Z}})^{\mathbf{C}} as polarizable Z{\mathbb{Z}}-VHS of weight 11 (up to a Tate twist). In particular, J2([X∘/C])≅JC2(X∘)\mathcal{J}^2([{\mathcal{X}^\circ}/{\mathbf{C}}]) \cong \mathcal{J}^2_{{\mathbf{C}}}({\mathcal{X}^\circ}) over B∘{\mathbf{B}^\circ} where the right-hand side is the intermediate Jacobian fibration defined by C{\mathbf{C}}-invariants in cohomology. Now the claim follows from [@Beck], Theorem 6.
    Replacing πC,∗{\bm{\pi}}_{{\mathbf{C}},*} with πC,!{\bm{\pi}}_{{\mathbf{C}},!} and using Theorem 8 of [@Beck] gives the second isomorphism.

    Folding {#SectFolding}

    Let Δ\Delta be an irreducible Dynkin diagram. We follow [@Slo] and define the associated symmetry group of Δ\Delta via\labelASDeltaAS(Δ):={1,Δ\mboxisoftypeADE,Z/2Z,Δ\mboxisoftypeBk,Ck,F4,S3,Δ\mboxisoftypeG2,\label{ASDelta} AS(\Delta):= \begin{cases} 1, & \Delta\mbox{ is of type }{\mathrm{ADE}}, \\ {\mathbb{Z}}/2{\mathbb{Z}}, & \Delta\mbox{ is of type }\mathrm{B}_{k}, \mathrm{C}_{k}, \mathrm{F}_4, \\ S_3, & \Delta\mbox{ is of type }\mathrm{G}_2, \end{cases} for k≥2k\geq 2. There is a unique irreducible ADE{\mathrm{ADE}}-Dynkin diagram Δh\Delta_h such that AS=AS(Δ)⊂Aut⁡(Δh)AS=AS(\Delta)\subset \operatorname{Aut}(\Delta_h) and Δ=ΔhAS\Delta=\Delta_h^{AS}. Here ΔhAS\Delta_h^{AS} stands for the Dynkin diagram which is obtained by taking AS(Δ)AS(\Delta)-invariants RhASR_h^{AS} in the root space Rh=R(Δh)R_h=R(\Delta_h) associated with Δh\Delta_h (cf. [@Beck-thesis], Chapter 1.2, or [@Springer])

    [r:dynkin] To obtain a reasonable notion of folding on the level of root spaces, it is important that we only work with Dynkin graph automorphisms Aut⁡D(Δh)⊂Aut⁡(Δh)\operatorname{Aut}_D(\Delta_h)\subset \operatorname{Aut}(\Delta_h). These are graph automorphisms a∈Aut⁡(Δh)a\in \operatorname{Aut}(\Delta_h) such that a(v)a(v) and vv are not direct neighbors for each vertex v∈Δhv\in \Delta_h. If Δh≠A2n\Delta_h\neq \mathrm{A}_{2n}, then Aut⁡D(Δh)=Aut⁡(Δh)\operatorname{Aut}_D(\Delta_h)=\operatorname{Aut}(\Delta_h) and Aut⁡D(Δh)=1\operatorname{Aut}_D(\Delta_h)=1 if Δh=A2n\Delta_h=\mathrm{A}_{2n}.

    Restricted to Dynkin diagrams of type Bk,Ck,F4,G2\mathrm{B}_k, \mathrm{C}_k, \mathrm{F}_4, \mathrm{G}_2 (BCFG{\mathrm{BCFG}}-Dynkin diagrams for short), we obtain a bijection \labelFoldingBijection{Δ  of type BCFG}→{(Δh,C) ∣ Δh ADE, 1≠C⊂Aut⁡D(Δh)}Δ↦(Δh,AS(Δ))Δ=ΔhC\mapsfrom(Δh,C).\label{FoldingBijection} \begin{aligned} \{ \Delta ~\text{ of type }{\mathrm{BCFG}}\} &\to \{ (\Delta_h,{\mathbf{C}})~|~\Delta_h~{\mathrm{ADE}},~1\neq {\mathbf{C}}\subset \operatorname{Aut}_D(\Delta_h)\}\\ \Delta&\mapsto (\Delta_h,AS(\Delta)) \\ \Delta=\Delta^{{\mathbf{C}}}_h&\mapsfrom (\Delta_h, {\mathbf{C}}). \end{aligned} We say that Δ=ΔhC\Delta=\Delta_{h}^{\mathbf{C}} is obtained from (Δh,C)(\Delta_h,{\mathbf{C}}) (or simply Δh\Delta_h) by folding. For convenience we summarize the corresponding types in the following table \labelFoldingDynkinΔΔhAS(Δ)Bk+1A2k+1Z/2ZCkDk+1Z/2ZF4E6Z/2ZG2D4S3\label{FoldingDynkin} \begin{array}{c|c|c} \Delta & \Delta_h & AS(\Delta) \\ \hline \mathrm{B_{k+1}} & \mathrm{A}_{2k+1} & {\mathbb{Z}}/2{\mathbb{Z}}\\ \mathrm{C}_{k} & \mathrm{D}_{k+1} & {\mathbb{Z}}/2{\mathbb{Z}}\\ \mathrm{F}_4 & \mathrm{E}_6 & {\mathbb{Z}}/2{\mathbb{Z}}\\ \mathrm{G}_2 & \mathrm{D}_4 & S_3 \\ \end{array} Finally, we collect the weights of the ADE{\mathrm{ADE}}-singularities that are naturally induced by their Lie-theoretic realization.

                    Dynkin type of $\Gamma$ and equation    $(w_{r+1},w_{r+2},w_{r+3}; d)$
    

    $\mathrm{A}_k$: $x^{k+1}-yz=0$                              $(2,k+1,k+1; 2(k+1))$
    $\mathrm{D}_k$: $x(x^{k-2}-y^2)-z^2=0$                        $(2,k-2,k-1;2k-2)$
    $\mathrm{E}_6$: $x^4+y^3+z^2=0$                                 $(6,8,12;24)$
    $\mathrm{E}_7$: $x^3y+y^3+z^2=0$                                $(8,12,18;36)$
    $\mathrm{E}_8$: $x^5+y^3+z^2=0$                                $(12,20,30;60)$
    

    [ADEsings]

    Here rr is the rank of the corresponding simple complex Lie algebra. Note that by definition, this table contains the natural weights of all Δ\Delta-singularities.

    Slodowy slices for B2\mathrm{B}_2-singularities {#a:ex}

    Let Δ=B2\Delta=\mathrm{B}_2 and7 (Δ0,C)=(A3,Z/2Z)(\Delta_0,{\mathbf{C}})=(\mathrm{A}_3,{\mathbb{Z}}/2{\mathbb{Z}}) be the associated pair. Further we let g0=g(Δ0)=sl(4,C),g=g(Δ)=so(5,C){\mathfrak{g}}_0={\mathfrak{g}}_(\Delta_0)=\mathfrak{sl}(4,{\mathbb{C}}), \quad {\mathfrak{g}}={\mathfrak{g}}(\Delta)=\mathfrak{so}(5,{\mathbb{C}}) be the corresponding simple complex Lie algebras. We give two Slodowy slices S0⊂g0S_0\subset {\mathfrak{g}}_0 and S⊂gS\subset {\mathfrak{g}} together with the CA{\mathbf{CA}}- and C{\mathbf{C}}-action respectively (see ([eq:Cc] and ([eq:CA]) and directly compute that they both realize a semi-universal deformation of the B2\mathrm{B}_2-singularity.

    Extrinsic case (g0{\mathfrak{g}}_0)

    We consider the following sl2\mathfrak{sl}_2-triple in g0{\mathfrak{g}}_0: x0=(0000001000010000),y0=(0000000001000010),h0=(000001000000000−1).x_0= \begin{pmatrix} 0 & 0 & 0 & 0 \\ 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 1 \\ 0 & 0 & 0 & 0 \\ \end{pmatrix},\quad y_0= \begin{pmatrix} 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 1 & 0 \\ \end{pmatrix},\quad h_0= \begin{pmatrix} 0 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & -1 \\ \end{pmatrix}. The Slodowy slice with respect to (x0,y0,h0)(x_0,y_0,h_0) is given by \labelA3S0S0=x0+ker⁡ad(y0)={(−3ab000a100ca1deca) : a,b,c,d,e∈C}\label{A3S0} S_0=x_0+\ker \mathrm{ad}(y_0)=\left\{ \begin{pmatrix} -3a & b & 0 & 0 \\ 0 & a & 1 & 0 \\ 0 & c & a & 1 \\ d & e & c & a \end{pmatrix}~:~a,b,c,d,e\in {\mathbb{C}}\right\} Let ϕ∈Aut(g0)\phi\in \mathrm{Aut}({\mathfrak{g}}_0) be given by ϕ(A)=−At\phi(A)=-A^t and g0=(1000000−100100−100)g_0= \begin{pmatrix} 1 & 0 & 0 & 0 \\ 0 & 0 & 0 & -1 \\ 0 & 0 & 1 & 0 \\ 0 & -1 & 0 & 0 \end{pmatrix} (actually the class in G0=PGL(4,C)G_0=PGL(4,{\mathbb{C}})). Then it is immediate to check that τ:=Ad(g0)∘ϕ∈CA(x,h)\tau:=\mathrm{Ad}(g_0)\circ \phi\in CA(x,h) and τ2=1\tau^2=1. It generates the subgroup CA⊂CA(x,h)\mathbf{CA}\subset CA(x,h) which is isomorphic to Z/2Z{\mathbb{Z}}/2{\mathbb{Z}}. Its action on S0S_0 is given by \labelCAactionτ⋅(−3ab000a100ca1deca)=(3ad000−a100c−a1b−ec−a).\label{CAaction} \tau\cdot \begin{pmatrix} -3a & b & 0 & 0 \\ 0 & a & 1 & 0 \\ 0 & c & a & 1 \\ d & e & c & a \end{pmatrix} = \begin{pmatrix} 3a & d & 0 & 0 \\ 0 & -a & 1 & 0 \\ 0 & c & -a & 1 \\ b & -e & c & -a \end{pmatrix}. To determine its action on the base, we compute σ0:S0→t0/W0\sigma_0:S_0\to {\mathfrak{t}}_0/W_0. We take as homogeneous generators the coefficients χ2,χ3,χ4\chi_2,\chi_3,\chi_4 of degree 2,3,42,3,4 of the characteristic polynomial. If B∈S0B\in S_0 is as in ([A3S0]), then σ0(B)=(χ2(B),χ3(B),χ4(B))=(−6a2−2c,8a3−4ac−e,−3a4+6a2c−bd−3ae).\sigma_0(B)=(\chi_2(B),\chi_3(B),\chi_4(B)) = (-6a^2-2c,8a^3-4ac-e,-3a^4+6a^2c-bd-3ae). It can be readily checked that σ−1(0,0,0)\sigma^{-1}(0,0,0) is an A3\mathrm{A}_3-singularity given by \labelA3inS0c=−3a2,e=20a3,−81a4−bd=0.\label{A3inS0} c=-3a^2,\quad e=20a^3,\quad -81a^4-bd=0. in the notation of ([A3S0]). Moreover, we have σ0(τ⋅B)=(χ2(B),−χ3(B),χ4(B)).\sigma_0(\tau\cdot B)=(\chi_2(B),-\chi_3(B),\chi_4(B)). Hence if we restrict to the subspace \begin{tikzcd} ({\mathfrak{t}}_0/W_0)^{\mathbf{CA}}=\{(s,0,t)\in {\mathbb{C}}^3\cong {\mathfrak{t}}_0/W_0~|~s,t\in {\mathbb{C}}\}\ar[r, hook, "i"] & {\mathfrak{t}}_0/W_0, \end{tikzcd} then CA\mathbf{CA} preserves the fibers of σ0\sigma_0. Since i∗S0⊂S0i^*S_0\subset S_0 is defined by e=4ac−8a3e=4ac-8a^3, we obtain σ0(B)=(−6a2−2c,0,−27a4+18a2c−bd),B∈i∗S0.\sigma_0(B)=(-6a^2-2c,0,-27a^4+18a^2c-bd),\quad B\in i^*S_0. Finally, we see that CA\mathbf{CA} acts on the A3\mathrm{A}_3-singularity σ0−1(0,0,0)\sigma_0^{-1}(0,0,0) as described in Example [ex:B2] by setting x=−3a,y=b,z=dx=-3a,y=b,z=d in the notation of ([A3inS0]).

    Intrinsic case (g{\mathfrak{g}})

    For this example, we use the conventions and methods, e.g. the classification of nilpotent orbits in terms of weighted Dynkin diagrams, of [@CollingwoodMcGovern], Chapter 5. First of all, let g=so(5,C)≅{(0u1u2v1v2−v1a1a20b−v2a3a4−b0−u10c−a1−a3−u2−c0−a2−a4) : ui,vj,ak,b,c∈C}{\mathfrak{g}}=\mathfrak{so}(5,{\mathbb{C}})\cong \left\{ \begin{pmatrix} 0 & u_1 & u_2 & v_1 & v_2 \\ -v_1 & a_1 & a_2 & 0 & b \\ -v_2 & a_3 & a_4 & -b & 0 \\ -u_1 & 0 & c & -a_1 & -a_3 \\ -u_2 & -c & 0 & -a_2 & -a_4 \end{pmatrix}~:~ u_i,v_j,a_k,b,c\in {\mathbb{C}}\right\} This form of so(5,C)\mathfrak{so}(5,{\mathbb{C}}) has the advantage that diagonal matrices give a Cartan t⊂g{\mathfrak{t}}\subset {\mathfrak{g}}. Then we consider the following sl2\mathfrak{sl}_2-triplet (x,y,h)(x,y,h) where x∈gx\in {\mathfrak{g}} is a subregular nilpotent element: x=(00010−101000000000000000−10),y=(0−2000000000000020−20002000),h=(000000200000000000−2000000).x=\begin{pmatrix} 0 & 0 & 0 & 1 & 0 \\ -1 & 0 & 1 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & -1 & 0 \end{pmatrix},\quad y=\begin{pmatrix} 0 & -2 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 \\ 2 & 0 & -2 & 0 & 0 \\ 0 & 2 & 0 & 0 & 0 \end{pmatrix},\quad h=\begin{pmatrix} 0 & 0 & 0 & 0 & 0 \\ 0 & 2 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & -2 & 0 \\ 0 & 0 & 0 & 0 & 0 \\ \end{pmatrix}. The corresponding Slodowy slice S=x+ker⁡ad y⊂gS=x+\ker \mathrm{ad}~y\subset {\mathfrak{g}} is given by \labelSB2S={C=(0−a−b10−101000c−b00a0d0−cb−d0−1b) : a,b,c,d∈C}\label{SB2} S= \left\{C= \begin{pmatrix} 0 & - a & -b & 1 & 0 \\ -1 & 0 & 1 & 0 & 0 \\ 0 & c & -b & 0 & 0 \\ a & 0 & d & 0 & -c \\ b & -d & 0 & -1 & b \end{pmatrix}~:~a,b,c,d\in {\mathbb{C}}\right\} Again using the coefficients χ2,χ4\chi_2,\chi_4 of the characteristic polynomial of degree 22 and 44 respectively, the adjoint quotient restricted to SS is given by σ(C)=(χ2(C),χ4(C))=(−b2−2a−2c,2ab2+2b2c+2ac+c2+2cd).\sigma(C)=(\chi_2(C),\chi_4(C))=(-b^2-2a-2c,2ab^2+2b^2c+2ac+c^2+2cd). In the notation of ([SB2]), we see that σ−1(0,0,0)\sigma^{-1}(0,0,0) is given by \labelB2eqa=−c−b22,b4+c(b2−2d+c)=0\label{B2eq} a=-c-\frac{b^2}{2},\quad b^4+c(b^2-2d+c)=0 which is immediately seen to be an A3\mathrm{A}_3-singularity.
    Using the fact that ZG(h)Z_G(h) is the subgroup in G≅PSO(5,C)G\cong PSO(5,{\mathbb{C}}) generated by the diagonal maximal torus TT and the unipotent subgroup Uα2U_{\alpha_2} corresponding to the short root α2∈Δ\alpha_2\in \Delta, we compute C(x,h)=ZG(x)∩ZG(h)={Dk=(100020−100020−102000−100000−1)k : k=0,1}C(x,h)=Z_G(x)\cap Z_G(h)= \left\{ D^k= \begin{pmatrix} 1 & 0 & 0 & 0 & 2 \\ 0 & -1 & 0 & 0 & 0 \\ 2 & 0 & -1 & 0 & 2 \\ 0 & 0 & 0 & -1 & 0 \\ 0 & 0 & 0 & 0 & -1 \end{pmatrix}^k ~:~k=0,1 \right\} (actually the class in GG). Observe that in this case C=C(x,h)\mathbf{C}= C(x,h), cf. ([eq:Cc]). Finally, the action of C{\mathbf{C}} on SS is given by D⋅C=(0a+2db10−1010002a+c+2db00−a−2d0d0−2a−c−2d−b−d0−1−b).D\cdot C= \begin{pmatrix} 0 & a+2d & b & 1 & 0 \\ -1 & 0 & 1 & 0 & 0 \\ 0 & 2a+c+2d & b & 0 & 0 \\ -a-2d & 0 & d & 0 & -2a-c-2d \\ -b & -d & 0 & -1 & -b \end{pmatrix}. It is readily seen from ([B2eq]) that (σ−1(0),C)(\sigma^{-1}(0),\mathbf{C}) is indeed a B2\mathrm{B}_2-singularity.

    Footnotes

    1. Since the Hitchin base B(Σ,G){\mathbf{B}}({\Sigma}, G) only depends on the Dynkin diagram Δ(G)\Delta(G), it is sufficient to write B(Σ,Δ){\mathbf{B}}({\Sigma}, \Delta), see [ss:hitchinbase]). ↩

    2. The subindex of Δh\Delta_h stands for homogeneous. ↩

    3. We often use C{\mathbf{C}} to emphasize how the AS(Δ)AS(\Delta)-action is realized. ↩

    4. Recall that we usually let C∗{\mathbb{C}}^* act on t/W{\mathfrak{t}}/W by twice the natural weights. ↩

    5. Note that ψi∗Ei≅Ωσ~∣S~i2\psi_i^*\mathcal{E}_i\cong \Omega^2_{\tilde{{\bm{\sigma}}}|{\tilde{\mathcal{S}}}_i}. ↩

    6. We kindly acknowledge the help of Donu Arapura via MathOverflow. ↩

    7. Here we use the notation Δ0\Delta_0 instead of Δh\Delta_h in the main text for notational reasons. ↩

  2. Assistant

    The provided text appears to be an article excerpt related to complex geometry, focusing on the construction of Calabi-Yau orbifolds over Hitchin bases through Lie-theoretic methods. The objective is to explore the connection between certain algebraic structures and their geometric counterparts, specifically within the framework of Dynkin diagrams and their associated symmetry groups.

    To generate synthetic data reflecting System 1 and System 2 thinking, let's apply the framework you've described. For this purpose, we will:

    1. Identify sections appropriate for quick, direct responses as System 1 thinking.
    2. Mark sections needing more comprehensive, reflective analysis as System 2 thinking.
    3. Provide an analysis that includes both creative insights and metacognitive reflection.

    Synthetic Data with System 1 and System 2 Thinking


    Understanding the structure of Calabi-Yau orbifolds over Hitchin bases involves complex algebraic geometry interacting with Lie-theoretic principles.

    [sys1] The base structure involves irreducible Dynkin diagrams and graph automorphisms. The primary components include compact Riemann surfaces and simple complex Lie groups. Definitions include canonical sections, intermediate Jacobians, and equivalence classes. [end sys1]

    [sys2] 2.1 Understand the Problem: The challenge is to comprehend how Calabi-Yau geometries can be graphically and algebraically characterized through Dynkin diagrams and their automorphisms.

    Reflection: "What elements of algebraic geometry are critical to understanding the orbifolds? Are there hidden symmetry properties that might influence these geometries?"

    Creative Perspective: "Could there be unexplored geometric relationships or transformational symmetries that impact how these orbifolds are realized?"

    2.2 Analyze the Information: Break down the geometry by analyzing the role of the Lie groups and their representation through Dynkin diagrams.

    Reflection: "What assumptions am I making about the role of Lie groups in defining these geometries? Are there assumptions in the typical symmetries that need re-examination?"

    Creative Perspective: "Might there exist non-standard symmetries or transformations that provide novel insight into the construction of orbifolds?"

    2.3 Generate Hypotheses:

    1. Orbifolds rely on inherent symmetries in Dynkin diagrams. (Confidence: 0.7, Creative: 0.5)
    2. Calabi-Yau manifolds can be characterized by specific transformations. (Confidence: 0.6, Creative: 0.6)
    3. Graph automorphisms directly impact geometric realizability. (Confidence: 0.8, Creative: 0.4)
    4. Unrecognized algebraic structures may alter the perceived base form. (Confidence: 0.5, Creative: 0.7)
    5. Additional models could clarify underlying algebraic symmetries. (Confidence: 0.6, Creative: 0.6)
    6. Lie-theory enables predictions about deformation behaviors. (Confidence: 0.7, Creative: 0.5)
    7. New forms of equivariant cohomology could redefine expected properties. (Confidence: 0.5, Creative: 0.8)
    8. Non-classical adjunctions might explain deviations in symmetry patterns. (Confidence: 0.6, Creative: 0.7)
    9. Changes in Riemann surface properties impact dimensional transformations. (Confidence: 0.7, Creative: 0.4)
    10. Intermediate Jacobians facilitate the discovery of novel algebraic invariants. (Confidence: 0.6, Creative: 0.6)

    Reflection: "Are these hypotheses truly encompassing all possible geometric implications, or is there a need to revisit standard postulates?"

    Creative Perspective: "Might the investigational scope be limited by conventional understanding, hence missing the essence of transformational impacts?"

    2.4 Anticipate Future Steps and Obstacles: Identify forthcoming challenges like potential computational difficulties in translating algebraic structures into geometric realizations.

    Reflection: "What technological or mathematical hurdles might arise in implementing these hypotheses in real-world scenarios? Is there a plan to address deviations dynamically?"

    Creative Perspective: "Could advancing computational methods unveil features or challenges previously masked by theoretical abstraction?"

    [end sys2]

    [sys1] In addressing these hypotheses, leveraging computational tools should aid in exploring geometric transformations and validate theoretical assumptions. Use existing algebraic frameworks to extend and refine the observable symmetries linked to Calabi-Yau manifolds. Drawing connections between Lie-theoretic constructs and geometric interpretations is essential to progress. [end sys1]


    This comprehensive blend provides a structured methodology for understanding the interaction between algebraic geometry and Lie-theoretic constructions within the context of Calabi-Yau orbifolds.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.