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

Euler Characteristics of the Generalized Kloosterman sheaves for symplectic and orthogonal groups

Yu Fu Department of Mathematics
California institute of technology
Pasadena, CA, 91125
yufu@caltech.edu
 and  Miao (Pam) Gu Department of Mathematics
University of Michigan
Ann Arbor, MI 48109-1043
pmgu@umich.edu
Abstract.

We study the monodromy of certain \ell-adic local systems attached to the generalized Kloosterman sheaves constructed by Yun and calculate their Euler characteristics under standard representations in the cases of symplectic and split/quasi-split orthogonal groups. This provides evidence for the conjectural description of their Swan conductors at \infty which is predicted by Reeder-Yu on the Langlands parameters attached to the epipelagic representations.

Key words and phrases:
Kloosterman sheaves, Swan conductor, Langlands correspondence, epipelegic representations
2020 Mathematics Subject Classification:
Primary: 14D24 ; 22E57 Secondary: 11F70 ; 11L05

1. Introduction

Let k=𝔽qk=\mathbb{F}_{q} be a finite field of characteristic pp and F=k(t)F=k(t) be the rational function field over kk. Places of F are in natural bijections with closed points of X:=k1X:=\mathbb{P}_{k}^{1}. Let GG be a reductive quasi-split group over FF. Let K=FK=F_{\infty} be the local field of FF at \infty. In [ReederYu], Reeder and Yu constructed a family of supercuspidal representations of pp-adic groups called epipelagic representations, and later Yun [YunEpipe] constructed automorphic representations for quasi-split groups over FF, whose local components at \infty are epipelagic representation. They also constructed the associated Galois representations from an admissible parahoric subgroup 𝐏G(K)\mathbf{P}\subset G(K) under the geometric Langlands correspondence. The Galois representations give GL{}^{L}G-local systems which generalizes the Kloosterman sheaves constructed earlier by Heinloth, Ngo and Yun [HNY].

In this paper, we compute Euler characteristics of the \ell-adic local systems attached to the generalized Kloosterman sheaf and standard representation of G^\widehat{G} in the cases that GG is some symplectic group or orthogonal group. This is equivalent to calculating the Swan conductor at \infty and gives evidence for the prediction made by Reeder and Yu about the Langlands parameters of epipelagic representations.

1.1. The Kloosterman sheaves

In [HNY], generalizing Deligne’s and Katz’s constructions of the Kloosterman sheaves attached to families of Kloosterman sums, Heinloth, Ngo^\widehat{\mathrm{o}}, and Yun constructed Kloosterman sheaves KlG^,𝐈(𝒦,ϕ)\mathrm{Kl}_{\widehat{G},\mathbf{I}}(\mathcal{K},\phi) for reductive groups GG and 𝐈\mathbf{I} the Iwahoric subgroup of GG, which are G^\widehat{G}-local systems. The Kloosterman sheaves depend on 𝒦\mathcal{K} a character sheaf and ϕ\phi a stable linear functional, which corresponds to simple supercuspidal representations of Gross-Reeder [GrossReeder] on the automorphic side.

Generalizing the construction in [HNY], Yun constructed in [YunEpipe] a family of generalized Kloosterman sheaves KlG^,𝐏(𝒦,ϕ)\mathrm{Kl}_{\widehat{G},\mathbf{P}}(\mathcal{K},\phi) associated with parahoric subgroups 𝐏\mathbf{P}. We recall some details of the construction below.

We fix a prime \ell different from char(k)\mathrm{char}(k). We consider constructible ¯\overline{\mathbb{Q}}_{\ell}-complexes over various algebraic stacks over kk or k¯\overline{k}. Let (L𝐏,V𝐏)(L_{\mathbf{P}},V_{\mathbf{P}}) be θ\theta-groups studied by [Vinberg]. When char(k)\mathrm{char}(k) is large, let 𝐏:=𝐏m\mathbf{P}:=\mathbf{P}_{m}, where mm is a regular elliptic number uniquely associated with the admissible parahoric 𝐏\mathbf{P} (see Section 2.2).

Let π(χ,ϕ)\pi(\chi,\phi) be a cuspidal automorphic representation whose local component at \infty is an epipelegic supercuspidal representation as realized in [YunEpipe], where χ\chi is a character of L𝐏~\widetilde{L_{\mathbf{P}}} and ϕ\phi is a stable linear functional V𝐏kV_{\mathbf{P}}\to k.

By the global geometric Langlands correspondence, it is associated with the Kloosterman sheaf KlGL,𝐏(𝒦,ϕ)\mathrm{Kl}_{{}^{L}G,\mathbf{P}}(\mathcal{K},\phi) (which is a GL{}^{L}G-local system), where 𝒦\mathcal{K} is the character sheaf associated with χ\chi under sheaf-to-function dictionary. Let X=k1{0,}X^{\circ}=\mathbb{P}_{k}^{1}-\{0,\infty\}. The local system attached to the generalized Kloosterman sheaves and stardard representation VV of G^\widehat{G} can be interpreted using the following diagram

X×𝔊λ𝔸1Xfϕπ.\leavevmode\hbox to131.12pt{\vbox to52.06pt{\pgfpicture\makeatletter\hbox{\hskip 65.55966pt\lower-26.03014pt\hbox to0.0pt{\pgfsys@beginscope\pgfsys@invoke{ }\definecolor{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@rgb@stroke{0}{0}{0}\pgfsys@invoke{ }\pgfsys@color@rgb@fill{0}{0}{0}\pgfsys@invoke{ }\pgfsys@setlinewidth{0.4pt}\pgfsys@invoke{ }\nullfont\hbox to0.0pt{\pgfsys@beginscope\pgfsys@invoke{ }{}{}{}{{}}\hbox{\hbox{{\pgfsys@beginscope\pgfsys@invoke{ }{\offinterlineskip{}{}{{{}}{{}}{{}}}{{{}}}{{}{}{{ {}{}}}{ {}{}} {{}{{}}}{{}{}}{}{{}{}} { }{{{{}}\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@transformcm{1.0}{0.0}{0.0}{1.0}{-65.55966pt}{-26.03014pt}\pgfsys@invoke{ }\hbox{\vbox{\halign{\pgf@matrix@init@row\pgf@matrix@step@column{\pgf@matrix@startcell#\pgf@matrix@endcell}&#\pgf@matrix@padding&&\pgf@matrix@step@column{\pgf@matrix@startcell#\pgf@matrix@endcell}&#\pgf@matrix@padding\cr\hfil\hskip 0.0pt\hbox{{\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope{}}}&\hskip 0.0pt\hfil&\hfil\hskip 45.86386pt\hbox{{\pgfsys@beginscope\pgfsys@invoke{ }{{}}\hbox{\hbox{{\pgfsys@beginscope\pgfsys@invoke{ }{{}{}{{ {}{}}}{ {}{}} {{}{{}}}{{}{}}{}{{}{}} { }{{{{}}\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@transformcm{1.0}{0.0}{0.0}{1.0}{-17.55835pt}{0.0pt}\pgfsys@invoke{ }\hbox{{\definecolor{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@rgb@stroke{0}{0}{0}\pgfsys@invoke{ }\pgfsys@color@rgb@fill{0}{0}{0}\pgfsys@invoke{ }\hbox{${X^{\circ}\times\mathfrak{G}_{\leq\lambda}}$} }}\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}} \pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}} \pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}&\hskip 21.86389pt\hfil\cr\vskip 18.00005pt\cr\hfil\hskip 9.45555pt\hbox{{\pgfsys@beginscope\pgfsys@invoke{ }{{}}\hbox{\hbox{{\pgfsys@beginscope\pgfsys@invoke{ }{{}{}{{ {}{}}}{ {}{}} {{}{{}}}{{}{}}{}{{}{}} { }{{{{}}\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@transformcm{1.0}{0.0}{0.0}{1.0}{-5.15001pt}{0.0pt}\pgfsys@invoke{ }\hbox{{\definecolor{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@rgb@stroke{0}{0}{0}\pgfsys@invoke{ }\pgfsys@color@rgb@fill{0}{0}{0}\pgfsys@invoke{ }\hbox{${\mathbb{A}^{1}}$} }}\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}} \pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}} \pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}&\hskip 9.45555pt\hfil&\hfil\hskip 23.99997pt\hbox{{\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope{}}}&\hskip 0.0pt\hfil&\hfil\hskip 34.24022pt\hbox{{\pgfsys@beginscope\pgfsys@invoke{ }{{}}\hbox{\hbox{{\pgfsys@beginscope\pgfsys@invoke{ }{{}{}{{ {}{}}}{ {}{}} {{}{{}}}{{}{}}{}{{}{}} { }{{{{}}\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@transformcm{1.0}{0.0}{0.0}{1.0}{-5.93471pt}{0.0pt}\pgfsys@invoke{ }\hbox{{\definecolor{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@rgb@stroke{0}{0}{0}\pgfsys@invoke{ }\pgfsys@color@rgb@fill{0}{0}{0}\pgfsys@invoke{ }\hbox{${X^{\circ}}$} }}\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}} \pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}} \pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}&\hskip 10.24025pt\hfil\cr}}}\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}}{{{{}}}{{}}{{}}{{}}}} \pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}} {}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{}}{}{}{}{{{}{}}}\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@setlinewidth{0.39998pt}\pgfsys@invoke{ }{}{{ {\pgfsys@beginscope \pgfsys@setdash{}{0.0pt}\pgfsys@roundcap\pgfsys@roundjoin{} {}{}{} {}{}{} \pgfsys@moveto{-2.07988pt}{2.39986pt}\pgfsys@curveto{-1.69989pt}{0.95992pt}{-0.85313pt}{0.27998pt}{0.0pt}{0.0pt}\pgfsys@curveto{-0.85313pt}{-0.27998pt}{-1.69989pt}{-0.95992pt}{-2.07988pt}{-2.39986pt}\pgfsys@stroke\pgfsys@endscope}} }{}{}{{}}{}{}{{}}\pgfsys@moveto{-14.46025pt}{4.06738pt}\pgfsys@lineto{-46.11519pt}{-16.91365pt}\pgfsys@stroke\pgfsys@invoke{ }{{}{{}}{}{}{{}}{{{}}}}{{}{{}}{}{}{{}}{{{}}{{{}}{\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@transformcm{-0.83353}{-0.55247}{0.55247}{-0.83353}{-46.28188pt}{-17.02412pt}\pgfsys@invoke{ }\pgfsys@invoke{ \lxSVG@closescope }\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}{{}}}}\hbox{\hbox{{\pgfsys@beginscope\pgfsys@invoke{ }{{}{}{{ }}{ } {{}{{}}}{{}{}}{}{{}{}} { }{{{{}}\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@transformcm{1.0}{0.0}{0.0}{1.0}{-39.37108pt}{-2.01419pt}\pgfsys@invoke{ }\hbox{{\definecolor{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@rgb@stroke{0}{0}{0}\pgfsys@invoke{ }\pgfsys@color@rgb@fill{0}{0}{0}\pgfsys@invoke{ }\hbox{$\scriptstyle{f_{\phi}}$} }}\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}} \pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}} \pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope{}{ {}{}{}}{}{ {}{}{}} {{{{{}}{ {}{}}{}{}{{}{}}}}}{}{{{{{}}{ {}{}}{}{}{{}{}}}}}{{}}{}{}{}{}{}{{{}{}}}{}{{}}{}{}{}{{{}{}}}\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@setlinewidth{0.39998pt}\pgfsys@invoke{ }{}{}{}{}{{}}{}{}{{}}\pgfsys@moveto{13.0866pt}{4.06738pt}\pgfsys@lineto{44.54436pt}{-16.48958pt}\pgfsys@stroke\pgfsys@invoke{ }{{}{{}}{}{}{{}}{{{}}}}{{}{{}}{}{}{{}}{{{}}{{{}}{\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@transformcm{0.83711}{-0.54703}{0.54703}{0.83711}{44.71176pt}{-16.59897pt}\pgfsys@invoke{ }\pgfsys@invoke{ \lxSVG@closescope }\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}{{}}}}\hbox{\hbox{{\pgfsys@beginscope\pgfsys@invoke{ }{{}{}{{}}{} {{}{{}}}{{}{}}{}{{}{}} { }{{{{}}\pgfsys@beginscope\pgfsys@invoke{ }\pgfsys@transformcm{1.0}{0.0}{0.0}{1.0}{31.33565pt}{-3.96771pt}\pgfsys@invoke{ }\hbox{{\definecolor{pgfstrokecolor}{rgb}{0,0,0}\pgfsys@color@rgb@stroke{0}{0}{0}\pgfsys@invoke{ }\pgfsys@color@rgb@fill{0}{0}{0}\pgfsys@invoke{ }\hbox{$\scriptstyle{\pi}$} }}\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}} \pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope}}} \pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope \pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope{}{}{}\hss}\pgfsys@discardpath\pgfsys@invoke{\lxSVG@closescope }\pgfsys@endscope\hss}}\lxSVG@closescope\endpgfpicture}}.

Here 𝔊λ\mathfrak{G}_{\leq\lambda} is the variety corresponding to the dominant coweight λ\lambda associated with VV and π\pi is the projection map. When G=Sp2nG=\mathrm{Sp}_{2n}, 𝔊λ=Sym12(M)\mathfrak{G}_{\leq\lambda}=\mathrm{Sym}^{2}_{\leq 1}(M), where MM is 2n2n-dimensinal symplectic space. The ¯\overline{\mathbb{Q}}_{\ell}-local system KlG^,𝐏St(𝟏,ϕ)\mathrm{Kl}^{\mathrm{St}}_{\widehat{G},\mathbf{P}}(\mathbf{1},\phi) asssoicated to the Kloosterman sheaves KlG^,𝐏(𝟏,ϕ)\mathrm{Kl}_{\widehat{G},\mathbf{P}}(\mathbf{1},\phi) (the G^\widehat{G}-local system descended from KlGL,𝐏(𝒦,ϕ)\mathrm{Kl}_{{}^{L}G,\mathbf{P}}(\mathcal{K},\phi)) with the trivial character sheaf 𝒦=𝟏\mathcal{K}=\mathbf{1} and the standard representation of GG are

KlG^,𝐏St(𝟏,ϕ)π!fϕASψ[2n1](2n12).\mathrm{Kl}^{\mathrm{St}}_{\widehat{G},\mathbf{P}}(\mathbf{1},\phi)\cong\pi!f_{\phi}^{*}\mathrm{AS}_{\psi}[2n-1]\left(\frac{2n-1}{2}\right).

When GG is an orthogonal group, let (M,q)(M,q) be the associated quadratic spaces of dimension 2n2n or 2n+12n+1. The variety 𝔊λ(M)\mathfrak{G}_{\leq\lambda}\subset\mathbb{P}(M) is defined by Q(q)Q[i,mi]Q(q)-\cup Q_{[i,m-i]}, where Q(q)Q(q) is the quadric defined by q=0q=0 and Q[i,mi]Q_{[i,m-i]} is the quadric defined by q=0q=0 when restricted to MiMmiM_{i}\oplus\cdots\oplus M_{m-i}. We have

KlG^,𝐏St(𝟏,ϕ)π!fϕASψ[dimM2](dimM22).\mathrm{Kl}^{\mathrm{St}}_{\widehat{G},\mathbf{P}}(\mathbf{1},\phi)\cong\pi!f_{\phi}^{*}\mathrm{AS}_{\psi}[\mathrm{dim}M-2]\left(\frac{\mathrm{dim}M-2}{2}\right).

1.2. Main Theorem

In this paper we prove the following main theorems:

Theorem 1.1.

When G=Sp2nG=\mathrm{Sp}_{2n}, we have

χc(X~,KlG^,𝐏mst(χ,ϕ))=d,-\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi))=d,

where d=2nmd=\frac{2n}{m}.

Theorem 1.2.

When GG are split or quasi-split orthogonal groups whose root systems are of either type BnB_{n}, DnD_{n}, or Dn2{}^{2}D_{n}, we have

χc(X~,KlG^,𝐏mst(χ,ϕ))={2dBn,2dDn, ω1 is non-degenerate,2(d+1)Dn, ω1 is degenerate,2dDn2, ω1 is non-degenerate,2(d+1)Dn2, ω1 is degenerate..-\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi))=\begin{cases}2d&B_{n},\\ 2d&D_{n},\text{ }\omega_{1}\text{ is non-degenerate,}\\ 2(d+1)&D_{n},\text{ }\omega_{1}\text{ is degenerate,}\\ 2d&{}^{2}D_{n},\text{ }\omega_{1}\text{ is non-degenerate,}\\ 2(d+1)&{}^{2}D_{n},\text{ }\omega_{1}\text{ is degenerate.}\end{cases}.

Here d=2nmd=\frac{2n}{m} when GG is of type BnB_{n}, d|nd|n is even for type DnD_{n}, and d|nd|n is odd for type Dn2{}^{2}D_{n}; d=2(n1)dd=\frac{2(n-1)}{d} is odd when GG is of type DnD_{n} and even for type Dn2{}^{2}D_{n}.

Using the Grothendieck-Ogg-Shafarevich formula, χc(X~,KlG^,𝐏mst(χ,ϕ))-\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi)) is equivalent to the Swan conductor of KlG^,𝐏mst(χ,ϕ)\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi) at \infty. Our main theorems then provide evidence for Reeder-Yu’s conjecture [ReederYu] about the Swan conductor of the Langlands parameter of epipelagic representations of GG in the function field case.

Remark 1.3.

As mentioned in Section 1.1, when 𝐏=𝐈\mathbf{P}=\mathbf{I}, the local system matches with the local system in [HNY]. In that case, the corresponding regular elliptic number mm is the twisted Coxeter number (see Section 2.2). It was shown in [HNY, Corollary 5.1] that Swan(KlG^,𝐈Vθ(𝒦,ϕ))=rs(G^)\mathrm{Swan}_{\infty}(\mathrm{Kl}^{V_{\theta^{\vee}}}_{\widehat{G},\mathbf{I}}(\mathcal{K},\phi))=r_{s}(\widehat{G}), where VθV_{\theta^{\vee}} is the unique quasi-minuscule representation of G^\widehat{G} whose nonzero weights consist of short roots of 𝔤^\widehat{\mathfrak{g}} and rs(G^)r_{s}(\widehat{G}) is the number of short simple roots of G^\widehat{G}. When G=Sp2nG=\mathrm{Sp}_{2n}, the unique quasi-minuscule representation of G^\widehat{G} is the standard representation of G^\widehat{G} (see [Ses78] for reference). In 1.1, when d=1d=1, mm is the twisted Coxeter number, and 1.1 matches up with this result.

1.3. Outline of the paper

We set up the notations and review details about the epipelagic representations and the construction of the generalized Kloosterman sheaves in Section 2. We discuss the relation between the Euler characteristics of the local system and the Swan conductor of the conjectural Langlands parameter of epipelagic representations in Section 3. In Section 4 and Section 5, we prove the main theorems in the symplectic case and orthogonal case, respectively.

Acknowledgements

This project was suggested to us by Zhiwei Yun in Arizona Winter School 2022. We would like to thank Zhiwei Yun for his continued support during our work on the project. We would like to thank the organizers of Arizona Winter School 2022 for an excellent event, and for giving us the opportunity to work on this project. The first author would like to thank Yi Ni for helpful discussions. The second author would like to thank Robert Cass for answering many questions on derived categories, and Charlotte Chan, Tasho Kaletha, Jianqiao Xia, and Zeyu Wang for helpful discussions.

2. Preliminaries

In this section, we recall some details from [YunEpipe] about the realization of epipelagic representations as a local component of an automorphic representations of G(𝔸F)G(\mathbb{A}_{F}) and the construction of the Kloosterman sheaves.

2.1. The absolute group data and the Langlands dual group

Let 𝔾\mathbb{G} be a a split reductive group over kk whose derived group is almost simple. Fix a pinning =(𝔹,𝕋,)\dagger=(\mathbb{B},\mathbb{T},\dots) of 𝔾\mathbb{G} for 𝔹\mathbb{B} a Borel subgroup and 𝕋𝔹\mathbb{T}\subset\mathbb{B} a split torus. Let 𝕎=N𝔾(𝕋)/𝕋\mathbb{W}=N_{\mathbb{G}}(\mathbb{T})/\mathbb{T} be the Weyl group of 𝔾\mathbb{G} and let σAut(𝔾)\sigma\in\mathrm{Aut}^{\dagger}(\mathbb{G}) be the image of 11. Fix a cyclic subgroup /e\mathbb{Z}/e\mathbb{Z} of the pinned automorphism group of 𝔾\mathbb{G}. We assume char(k)\mathrm{char}(k) is prime to ee and k×k^{\times} contains eeth roots of unity μe\mu_{e}.

Fix a μe\mu_{e}-cover X~X\widetilde{X}\to X which is totally ramified over 0 and \infty. X~\widetilde{X} is isomorphic to k1\mathbb{P}_{k}^{1} with affine coordinate t1/et^{1/e}. We denote

X:=X{0,};X~:=X~{0,}.X^{\circ}:=X-\{0,\infty\};\widetilde{X}^{\circ}:=\widetilde{X}-\{0,\infty\}.

Let G^\widehat{G} be the reductive group over Q¯\overline{Q}_{\ell} whose root system is dual to that of 𝔾\mathbb{G}. We define the Langlands dual group GL{}^{L}G to be GL=G^μe{}^{L}G=\widehat{G}\rtimes\mu_{e}.

2.2. Admissible parahoric subgroups and regular elliptic numbers

We denote K=FK=F_{\infty} as the local field of FF at \infty. Let 𝐏G(K)\mathbf{P}\subset G(K) be a standard parahoric subgroup containing the standard Iwahori subgroup 𝐈\mathbf{I}. Let 𝔄\mathfrak{A} be the apartment in the building of G(K)G(K) corresponding to the maximal split torus of G(K)G(K). We denote Ψaff\Psi_{\mathrm{aff}} as the set of affine roots of G(K)G(K), which are certain affine functions on 𝔄\mathfrak{A}. 𝐏\mathbf{P} determines a facet 𝔉𝐏\mathfrak{F}_{\mathbf{P}} in 𝔄\mathfrak{A}, and let ξ𝔄\xi\in\mathfrak{A} denotes its barycenter. Let m=m(𝐏)m=m(\mathbf{P}) be the smallest positive integer such that α(ξ)1m\alpha(\xi)\in\frac{1}{m}\mathbb{Z} for all affine roots αΨaff\alpha\in\Psi_{\mathrm{aff}} (see [ReederYu, §3.3]).

Let 𝐏𝐏+𝐏++\mathbf{P}\supset\mathbf{P}^{+}\supset\mathbf{P}^{++} be the first three steps in the Moy-Prasad filtration of 𝐏\mathbf{P}. Here 𝐏+\mathbf{P}^{+} is the pro-unipotent radical of 𝐏\mathbf{P} and L𝐏:=𝐏/𝐏+L_{\mathbf{P}}:=\mathbf{P}/\mathbf{P}^{+} is the Levi factor of 𝐏\mathbf{P}. We denote V𝐏=𝐏+/𝐏++V_{\mathbf{P}}=\mathbf{P}^{+}/\mathbf{P}^{++}.

Let 𝕎=𝕎μe\mathbb{W}^{\prime}=\mathbb{W}\rtimes\mu_{e}. An element w𝕎w\in\mathbb{W}^{\prime} is \mathbb{Z}-regular if it permutes the roots freely and elliptic if 𝕏(𝕋ad)w=0\mathbb{X}^{\ast}(\mathbb{T}^{\mathrm{ad}})^{w}=0. The order of a \mathbb{Z}-regular elliptic element in 𝕎σ𝕎\mathbb{W}\sigma\subset\mathbb{W}^{\prime} is called a regular elliptic number of the pair (𝕎,σ)(\mathbb{W}^{\prime},\sigma).

𝐏\mathbf{P} is called admissible if there exists a closed orbit of L𝐏L_{\mathbf{P}} on the dual space V𝐏V_{\mathbf{P}}^{\ast} with finite stabilizers. When 𝐏\mathbf{P} is admissible and char(k)\mathrm{char}(k) is large, it was shown in [YunEpipe, §2.6] that there is a bijection between admissible parahoric subgroups 𝐏\mathbf{P} and regular elliptic numbers mm using [GLRY, Proposition 1],[Spr74, Proposition 6.4(iv)] and [ReederYu, Corollary 5.1].

Remark 2.1.

In particular, in the case of [HNY], the regular elliptic number corresponding to 𝐈\mathbf{I} is the twisted Coxeter number hσh_{\sigma}.

2.3. Epipelagic representations and automorphic representations

Let ψ:k(νp)×\psi:k\to\mathbb{Q}_{\ell}(\nu_{p})^{\times} be a nontrivial character and ϕ:V𝐏k\phi:V_{\mathbf{P}}\to k be a stable functional. As shown in [ReederYu, Proposition 2.4], the simple summands of the compact induction

cInd𝐏+G(K)(ψϕ)c-\mathrm{Ind}_{\mathbf{P^{+}}}^{G(K)}(\psi\circ\phi)

are called epipelagic representations of G(k)G(k) attached to the parahoric 𝐏\mathbf{P} and the stable functional ϕ\phi.

Let 𝐈0G(F0)\mathbf{I}_{0}\subset G(F_{0}) be the Iwahori subgroup corresponding to the opposite Borel 𝐁opp\mathbf{B}^{\mathrm{opp}} of GG, and let 𝐏0G(F0)\mathbf{P}_{0}\subset G(F_{0}) be the parahoric subgroup containing 𝐈0\mathbf{I}_{0} of the same type as 𝐏\mathbf{P}_{\infty}. Let 𝐏~0\widetilde{\mathbf{P}}_{0} be the normalizer of 𝐏0\mathbf{P}_{0} in G(F0)G(F_{0}) and L~𝐏=𝐏~0/𝐏0+\widetilde{L}_{\mathbf{P}}=\widetilde{\mathbf{P}}_{0}/\mathbf{P}_{0}^{+}. Let L𝐏scL_{\mathbf{P}}^{\mathrm{sc}} be the simply connected cover of the derived group of L𝐏L_{\mathbf{P}}, and let L𝐏(k)=Im(L𝐏sc(k)L𝐏(k))L_{\mathbf{P}}(k)^{\prime}=\mathrm{Im}(L_{\mathbf{P}}^{\mathrm{sc}}(k)\to L_{\mathbf{P}}(k)). We fix a character χ:L~𝐏(k)/L𝐏(k)¯×\chi:\widetilde{L}_{\mathbf{P}}(k)/L_{\mathbf{P}}(k)\to\overline{\mathbb{Q}}_{\ell}^{\times}.

Let π=x|X|πx\pi=\otimes^{\prime}_{x\in|X|}\pi_{x} of G(𝔸)G(\mathbb{A}) be an automorphic representation whose local components satisfying the following conditions: π\pi_{\infty} is an epipelagic representation attached to 𝐏+\mathbf{P}_{\infty}^{+} and ϕ\phi; π0\pi_{0} has an eigenvector under 𝐏~0\widetilde{\mathbf{P}}_{0} on which it acts through χ\chi via the quotient L~𝐏(k)\tilde{L}_{\mathbf{P}}(k); πx\pi_{x} are unramified for x0,x\neq 0,\infty.

It was shown in [YunEpipe, Proposition 2.11] that there is a unique cuspidal automorphic representation π=π(χ,ϕ)\pi=\pi(\chi,\phi) of G(𝔸F)G(\mathbb{A}_{F}) satisfying the above conditions.

2.4. The generalized Kloosterman sheaves.

Let Bun=Bun(𝐏0opp,𝐏++)\operatorname{Bun}=\operatorname{Bun}(\mathbf{P}_{0}^{opp},\mathbf{P}_{\infty}^{++}) be the moduli stack of bundles with 𝐏opp\mathbf{P}^{opp}-level structure at 0 and 𝐏++\mathbf{P}^{++}-level structure at \infty. The Hecke correspondence for Bun classifies (x,,,τ)(x,\mathcal{E},\mathcal{E}^{\prime},\tau) where xX~x\in\widetilde{X}^{\circ}, \mathcal{E}, Bun\mathcal{E}^{\prime}\in\operatorname{Bun} and τ:|X~x|X~x\tau:\mathcal{E}|_{\widetilde{X}-x}\to\mathcal{E}^{\prime}|_{\widetilde{X}-x} is an isomorphism of GG-torsors preserving the level structures at 0 and \infty. The moduli stack BunG(𝐏~0,𝐏+)\operatorname{Bun}_{G}(\widetilde{\mathbf{P}}_{0},\mathbf{P}^{+}_{\infty}) has a unique relevant point \mathcal{E} with trivial automorphism group. Let j:V𝐏Bunj:V_{\mathbf{P}}\hookrightarrow\operatorname{Bun} be the open embedding defined by \mathcal{E} and ASψ the pullback along the canonical pairing <,>:V𝐏×V𝐏,st𝔾a<-,->:V_{\mathbf{P}}\times V_{\mathbf{P}}^{*,st}\to\mathbb{G}_{a} of the Artin-Schreier sheaf on 𝔾a\mathbb{G}_{a} corresponding to a fixed character ψ:k¯\psi:k\to\bar{\mathbb{Q}}_{\ell}. According to [YunEpipe, Section 3], the Hecke eigensheaf can be described as

𝒜=(j×idV𝐏,st)!ASψ.\mathcal{A}=(j\times id_{V_{\mathbf{P}}^{*,st}})_{!}\operatorname{AS}_{\psi}.

The generalized Kloosterman sheaves. Let 𝔊\mathfrak{G} be the group of automorphisms of |X1\mathcal{E}|_{X-{1}} preserving the level structure at 0 and \infty. Let λ\lambda be a dominant coroot of GG and VλV_{\lambda} be the irreducible representation of G^\widehat{G} of highest weight λ\lambda. We have an affine Schubert variety Grλ\operatorname{Gr}_{\leq\lambda} in the affine Grassmannian Gr=L𝔾/L+𝔾\operatorname{Gr}=L\mathbb{G}/L^{+}\mathbb{G} defined by λ\lambda. Let 𝔊λ\mathfrak{G}_{\leq\lambda} be the preimage of L+𝔾\GrλL+𝔾\L𝔾/L+𝔾L^{+}\mathbb{G}\backslash\operatorname{Gr}_{\leq\lambda}\subset L^{+}\mathbb{G}\backslash L\mathbb{G}/L^{+}\mathbb{G} under the evaluation map ev1~\operatorname{ev}_{\widetilde{1}} at the preimage of 1X1\in X

ev1~:𝔊L1~+𝔾\L1~𝔾/L1~+𝔾.\operatorname{ev}_{\widetilde{1}}:\mathfrak{G}\to L^{+}_{\widetilde{1}}\mathbb{G}\backslash L_{\widetilde{1}}\mathbb{G}/L^{+}_{\widetilde{1}}\mathbb{G}.

The geometric Satake equivalence gives the intersection complex ICλIC_{\lambda} which corresponds to VλV_{\lambda}. Using the evaluation maps at 0 and \infty and compose with the projections 𝐏~0L~𝐏ab\widetilde{\mathbf{P}}_{0}\to\widetilde{L}_{\mathbf{P}}^{ab} and 𝐏+V𝐏\mathbf{P}^{+}_{\infty}\to V_{\mathbf{P}} we get

(f,f′′):𝔊λL~𝐏ab×V𝐏.(f^{\prime},f^{\prime\prime}):\mathfrak{G}_{\leq\lambda}\to\widetilde{L}_{\mathbf{P}}^{ab}\times V_{\mathbf{P}}.

By the existence of Hecke eigensheaves over the stable functionals V𝐏,stV_{\mathbf{P}}^{*,st} with eigenvalue KlG^,𝐏V(ϕ)\operatorname{Kl}_{\widehat{G},\mathbf{P}}^{V}(\phi), see [YunEpipe, Lemma 3.6, Corollary 3.7], applying the Fourier-Deligne transform we have [YunEpipe, Proposition 3.9]: Let Fourψ:Dcb(V𝐏)Dcb(V𝐏)\operatorname{Four}_{\psi}:D_{c}^{b}\left(V_{\mathbf{P}}\right)\rightarrow D_{c}^{b}\left(V_{\mathbf{P}}^{*}\right) be the Fourier-Deligne transform (without cohomological shift). We have

KlG^,𝐏Vλ(χ)Fourψ(f!′′(fχev1~ICλ))|V𝐏,st.\left.\mathrm{Kl}_{\widehat{G},\mathbf{P}}^{V_{\lambda}}(\chi)\cong\operatorname{Four}_{\psi}\left(f_{!}^{\prime\prime}\left(f^{\prime*}\mathcal{L}_{\chi}\otimes\operatorname{ev}_{\widetilde{1}}^{*}\mathrm{IC}_{\lambda}\right)\right)\right|_{V_{\mathbf{P}}^{*,st}}.

3. Euler characteristic of the local system and Swan conductor of epipelagic parameters

Let K=FK=F_{\infty} be the local field of FF at \infty. In this section, we relate the computation of the Euler characteristic of the local system corresponds to the generalized Kloosterman sheaves to Reeder-Yu’s prediction on the Swan conductor of the Langlands parameter of epipelegic representations.

3.1. Epipelegic representations and Kloosterman sheaves

Let π\pi_{\infty} be the epipelagic supercuspidal representation of G(k)G(k) as the local component at \infty of an automorphic representation π(χ,ϕ)\pi(\chi,\phi) as realized in section 2.3. As shown in [YunEpipe, Corollary 3.10], the GL{}^{L}G-local system KlGL,𝐏(𝒦,ϕ)\mathrm{Kl}_{{}^{L}G,\mathbf{P}}(\mathcal{K},\phi) is the global Langlands parameter attached to π(χ,ϕ)\pi(\chi,\phi). Here 𝒦\mathcal{K} is the character sheaf corresponding to χ\chi under sheaf-to-function correspondence.

3.2. Expected Langlands parameter

Let π\pi_{\infty} be as defined in the previous subsection. The local Langlands conjecture suggests that there should be a Galois representation ρπ:W(Ks/K)GL(¯)\rho_{\pi_{\infty}}:W(K^{s}/K)\to{}^{L}G(\overline{\mathbb{Q}}_{\ell}) attached to π\pi_{\infty} as the conjectural Langlands parameter. Based on the conjecture relating adjoint gamma factors and formal degrees in [GrossReeder], Reeder and Yu predicted in [ReederYu, Section 7.1] that

Swan(𝔤^)=#Φm.\mathrm{Swan}(\widehat{\mathfrak{g}})=\frac{\#\Phi}{m}.

3.3. Swan conductor and the Euler characteristic

We have the following lemma of the Swan conductor of the local system:

Lemma 3.1.

For X~k1{0,}\tilde{X}^{\circ}\cong\mathbb{P}_{k}^{1}-\{0,\infty\},

χc(X~,KlG^,𝐏mVλ(χ,ϕ))=Swan(KlG^,𝐏mVλ)\chi_{c}(\tilde{X}^{\circ},\mathrm{Kl}^{V_{\lambda}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi))=-\mathrm{Swan}_{\infty}(\mathrm{Kl}^{V_{\lambda}}_{\widehat{G},\mathbf{P}_{m}})
Proof.

By the Grothendieck-Ogg-Shafarevich formula, we have

χc(X~,KlG^,𝐏mVλ(χ,ϕ))=\displaystyle\chi_{c}(\tilde{X}^{\circ},\mathrm{Kl}^{V_{\lambda}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi))= χc(𝔾m)rk(KlG^,𝐏mVλ)x{0,}Swanx(KlG^,𝐏mVλ)\displaystyle\chi_{c}(\mathbb{G}_{m})\mathrm{rk}(\mathrm{Kl}^{V_{\lambda}}_{\widehat{G},\mathbf{P}_{m}})-\sum_{x\in\{0,\infty\}}\mathrm{Swan}_{x}(\mathrm{Kl}^{V_{\lambda}}_{\widehat{G},\mathbf{P}_{m}})
=\displaystyle= Swan0(KlG^,𝐏mVλ)Swan(KlG^,𝐏mVλ),\displaystyle-\mathrm{Swan}_{0}(\mathrm{Kl}^{V_{\lambda}}_{\widehat{G},\mathbf{P}_{m}})-\mathrm{Swan}_{\infty}(\mathrm{Kl}^{V_{\lambda}}_{\widehat{G},\mathbf{P}_{m}}),

Since 0 is the tame point as shown in [YunEpipe, Theorem 4.5], Swan0(KlG^,𝐏mVλ)=0\mathrm{Swan}_{0}(\mathrm{Kl}^{V_{\lambda}}_{\widehat{G},\mathbf{P}_{m}})=0. Then we deduce the lemma. ∎

Thus, computing the Euler characteristic of the local system corresponding to the Kloosterman sheaves is equivalent to computing the swan conductor at \infty. Therefore, combining this with Section 3.1 and Section 3.3, the computation of χc(X~,KlG^,𝐏mSt(χ,ϕ))\chi_{c}(\tilde{X}^{\circ},\mathrm{Kl}^{\mathrm{St}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi)) gives evidence to Reeder-Yu’s prediction.

4. Euler Characteristic for Symplectic Groups

The goal of this section is to prove Theorem 1.1. Throughout the section we work over k¯\bar{k} and ignore all the Tate twists. Let GG be a symplectic group. We first make a summary of [YunEpipe, Sec. 7] in subsection 4.1 and 4.2 where the local system attached to the Kloosterman sheaf KlG^,𝐏(𝒦,ϕ)\operatorname{Kl}_{\widehat{G},\mathbf{P}}(\mathcal{K},\phi) and the standard representation of G^\widehat{G} is calculated, then make an inductive argument to calculate the Euler characteristic of the local system.

4.1. The set-up

Let (M,ω)(M,\omega) be a symplectic vector space of dimension 2n2n over kk. One can extend ω\omega linearly to a symplectic form on MKM\otimes K and we denote 𝔾=Sp(M,ω)\mathbb{G}=\operatorname{Sp}(M,\omega) and G=Sp(MK,ω)G=\operatorname{Sp}(M\otimes K,\omega). Since regular elliptic numbers mm of 𝕎\mathbb{W} in this case are in bijection with divisors d|nd|n[Yun, 4.8], we have m=2n/dm=2n/d and let =n/d\ell=n/d. We fix a decomposition

M=M1M2MM+1MmM=M_{1}\oplus M_{2}\oplus\cdots\oplus M_{\ell}\oplus M_{\ell+1}\oplus\cdots\oplus M_{m}

such that dimMi=ddim_{M_{i}}=d and ω(Mi,Mj)0\omega(M_{i},M_{j})\neq 0 only if i+j=m+1i+j=m+1. We identify MjM_{j} with Mm+1jM^{*}_{m+1-j}.

Define the admissible parahoric subgroup 𝐏mG(K)\mathbf{P}_{m}\subset G(K) to be the stabilizer of the lattice chain

ΛmΛm1Λ1\Lambda_{m}\supset\Lambda_{m-1}\supset\cdots\Lambda_{1}

where

Λi=1jiMj𝒪K+i<jmMjω¯𝒪K\Lambda_{i}=\sum_{1\leq j\leq i}M_{j}\otimes\mathcal{O}_{K}+\sum_{i<j\leq m}M_{j}\otimes\bar{\omega}\mathcal{O}_{K}

and ω¯\bar{\omega} is a uniformizer of 𝒪F.\mathcal{O}_{F}. It has Levi quotient denoted by Lm=i=1GL(Mi)L_{m}=\prod_{i=1}^{\ell}GL(M_{i}) where the ii-th factor acts on MiM_{i} by the standard representation and on Mi=Mm+1iM^{*}_{i}=M_{m+1-i} by the dual of the standard representation. We have L~m=Lm\widetilde{L}_{m}=L_{m} and L~mabi=1𝔾m\widetilde{L}^{ab}_{m}\cong\prod_{i=1}^{\ell}\mathbb{G}_{m} given by the determinants of the GL-factors. The vector space Vm:=V𝐏mV_{m}:=V_{\mathbf{P}_{m}} can be described as

Vm=Sym2(M1)Hom(M2,M1)Hom(M,M1)Sym2(M).V_{m}=\operatorname{Sym}^{2}\left(M_{1}^{*}\right)\oplus\operatorname{Hom}\left(M_{2},M_{1}\right)\oplus\cdots\oplus\operatorname{Hom}\left(M_{\ell},M_{\ell-1}\right)\oplus\operatorname{Sym}^{2}\left(M_{\ell}\right).

We can arrange M1,,MmM_{1},\cdots,M_{m} into a cyclic quiver

M1{M_{1}}M2{M_{2}}{\cdots}M{M_{\ell}}Mm{M_{m}}Mm1{M_{m-1}}{\cdots}M+1{M_{\ell+1}}ϕ1\scriptstyle{\definecolor[named]{.}{rgb}{0,0,1}\color[rgb]{0,0,1}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,1}\phi_{1}}ψm\scriptstyle{\psi_{m}}ψ1\scriptstyle{\psi_{1}}ϕ1\scriptstyle{\definecolor[named]{.}{rgb}{0,0,1}\color[rgb]{0,0,1}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,1}\phi_{\ell-1}}ψ1\scriptstyle{\psi_{\ell-1}}ϕ\scriptstyle{\definecolor[named]{.}{rgb}{0,0,1}\color[rgb]{0,0,1}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,1}\phi_{\ell}}ψm1\scriptstyle{\psi_{m-1}}ϕm\scriptstyle{\definecolor[named]{.}{rgb}{0,0,1}\color[rgb]{0,0,1}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,1}\phi_{m}}ϕm1\scriptstyle{\definecolor[named]{.}{rgb}{0,0,1}\color[rgb]{0,0,1}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,1}\phi_{m-1}}ψ+1\scriptstyle{\psi_{\ell+1}}ψ\scriptstyle{\psi_{\ell}}ϕ+1\scriptstyle{\definecolor[named]{.}{rgb}{0,0,1}\color[rgb]{0,0,1}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,1}\phi_{\ell+1}}

so that the involution τ\tau sends {ψi:Mi+1Mi}\left\{\psi_{i}:M_{i+1}\rightarrow M_{i}\right\} to {ψmi:MmiMm+1i}\left\{-\psi_{m-i}^{*}:M_{m-i}^{*}\rightarrow M_{m+1-i}^{*}\right\}. Therefore, VmV_{m} is the set of τ\tau-invariant cyclic quivers of the above shape. The dual space VmV_{m}^{*} is the space of τ\tau-invariant cyclic quivers with all the arrows reversed, so we fit them within the same commutative diagram. Let ϕi:MiMi+1\phi_{i}:M_{i}\to M_{i+1} be the arrows and we view ϕm\phi_{m} (resp. ϕ\phi_{\ell}) as a quadratic form on MmM_{m} (resp. MM_{\ell}). Then ϕ=(ϕ1,,ϕm)Vm\phi=\left(\phi_{1},\ldots,\phi_{m}\right)\in V_{m}^{*} is stable if and only if

  • All the maps ϕi\phi_{i} are isomorphisms;

  • We have two quadratic forms on Mm:ϕmM_{m}:\phi_{m} and the transport of ϕ\phi_{\ell} to MmM_{m} using the isomorphism ϕ1ϕ1ϕm:MmM\phi_{\ell-1}\cdots\phi_{1}\phi_{m}:M_{m}\stackrel{{\scriptstyle\sim}}{{\rightarrow}}M_{\ell}. They are in general position in the same sense as explained in [YunEpipe, Section 6.2].

4.2. The local system.

Let =M𝒪X\mathcal{E}=M\otimes\mathcal{O}_{X} be the trivial vector bundle of rank 2n2n over XX with a symplectic form (into 𝒪X\mathcal{O}_{X} ) given by ω\omega. Define an increasing filtration of the fiber of \mathcal{E} at \infty by Fi=j=1iMjF_{\leq i}\mathcal{E}_{\infty}=\oplus_{j=1}^{i}M_{j} and a decreasing filtration of the fiber of \mathcal{E} at 0 by Fi0=j=imMjF^{\geq i}\mathcal{E}_{0}=\oplus_{j=i}^{m}M_{j}. The moduli stack BunG(𝐏0~,𝐏+)\operatorname{Bun}_{G}(\widetilde{\mathbf{P}_{0}},\mathbf{P}^{+}_{\infty}) then classifies triples (,Fi,Fi0)(\mathcal{E},F_{\leq i}\mathcal{E}_{\infty},F^{\geq i}\mathcal{E}_{0}). The group ind-scheme 𝔊\mathfrak{G} is the group of symplectic automorphisms of |X{1}\left.\mathcal{E}\right|_{X-\{1\}} preserving the filtrations F,FF_{*},F^{*} and acting by identity on the associated graded of FF_{*}. Let λ\lambda be the dominant short coroot. The subscheme 𝔊λ\mathfrak{G}_{\leq\lambda} consists of those g𝔊G(F)g\in\mathfrak{G}\subset G(F) whose entries have at most simple poles at t=1t=1, and Rest=1g\operatorname{Res}_{t=1}g has rank at most one. [YunEpipe, Lemma 7.4] shows that the subscheme 𝔊λ\mathfrak{G}_{\leq\lambda} can be embedded as an open subscheme of Sym12(M)\operatorname{Sym}^{2}_{\leq 1}(M), where Sym12(M)Sym2(M)\operatorname{Sym}^{2}_{\leq 1}(M)\subset\operatorname{Sym}^{2}(M) is the subscheme of symmetric pure 22-tensors. This is equivalent to say uu and vv are parallel vectors. We may write elements uvSym2(M)1u\cdot v\in\mathrm{Sym}^{2}(M)_{\leq 1} as u=(u1,,um)u=(u_{1},\dots,u_{m}) and v=(v1,,vm)v=(v_{1},\dots,v_{m}) with ui,viMiu_{i},v_{i}\in M_{i}. Define γi\gamma_{i} as follows

(4.2.1) γi(uv):=ω(vm+1i,ui).\displaystyle\gamma_{i}(u\cdot v):=\omega(v_{m+1-i},u_{i}).

The definition is independent of the choice of u,vu,v expressing uvu\cdot v therefore defines a regular function on Sym2(M)1.\mathrm{Sym}^{2}(M)_{\leq 1}. The following proposition gives an explicit description of KlG^,𝐏st(ϕ)\operatorname{Kl}_{\widehat{G},\mathbf{P}}^{st}(\phi) that will be used later in the computation of the Euler characteristic.

Proposition 4.1.

[YunEpipe, Cor. 7.6]

Let ϕ=(ϕ1,,ϕm)Vm,st(k)\phi=\left(\phi_{1},\ldots,\phi_{m}\right)\in V_{m}^{*,\mathrm{st}}(k) be a stable functional. Recall that 𝔊λ\mathfrak{G}_{\leq\lambda} in this case is Sym2(M)1i=1Γi\operatorname{Sym}^{2}(M)_{\leq 1}-\cup_{i=1}^{\ell}\Gamma_{i} where the divisor Γi\Gamma_{i} is defined by the equation γ1++γi=1\gamma_{1}+\cdots+\gamma_{i}=1 for functions γi\gamma_{i} (see [YunEpipe, Prop. 7.5]). Let fϕ:X×𝔊λ𝔸1f_{\phi}:X^{\circ}\times\mathfrak{G}_{\leq\lambda}\rightarrow\mathbb{A}^{1} be given by

fϕ(x,uv)=ω(ϕmvm,um)x+i=1ω(ϕivi,umi)1γ1(uv)γi(uv)f_{\phi}(x,u\cdot v)=\omega\left(\phi_{m}v_{m},u_{m}\right)x+\sum_{i=1}^{\ell}\frac{\omega\left(\phi_{i}v_{i},u_{m-i}\right)}{1-\gamma_{1}(u\cdot v)-\cdots-\gamma_{i}(u\cdot v)}

Let π:X×𝔊λX\pi:X^{\circ}\times\mathfrak{G}_{\leq\lambda}\rightarrow X^{\circ} be the projection. Then we have an isomorphism over XX^{\circ}

KlG^,𝐏mSt(𝟏,ϕ)π!fϕASψ[2n1](2n12).\mathrm{Kl}_{\widehat{G},\mathbf{P}_{m}}{\mathrm{St}}(\mathbf{1},\phi)\cong\pi!f_{\phi}^{*}\mathrm{AS}_{\psi}[2n-1]\left(\frac{2n-1}{2}\right).

4.3. Computation of Euler characteristic

Theorem 4.2.

We have

(4.3.1) χc(X~,KlG^,𝐏mst(χ,ϕ))=d\displaystyle-\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi))=d

By the same argument as in [Katz] the Swan conductor of KlG^,𝐏mst(χ,ϕ)\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi) at \infty does not depend on χ\chi, so does the Euler characteristic of KlG^,𝐏mst(χ,ϕ)\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi). We make an inductive argument as follows.

Proposition 4.3.

Let m3m\geq 3 be a regular elliptic number. Let ΓiSym12(M)\Gamma_{i}\subset\mathrm{Sym}^{2}_{\leq 1}(M) be the divisor defined by the equation γi=γ1++γi=1\gamma_{i}^{\prime}=\gamma_{1}+\cdots+\gamma_{i}=1 for functions γi\gamma_{i} (recall 4.2.1). Let Ui=Sym12(j=1iMjj=mi1mMj)j=1iΓiU_{i}=\mathrm{Sym}^{2}_{\leq 1}(\oplus_{j=1}^{i}M_{j}\oplus_{j^{\prime}=m-i-1}^{m}M_{j^{\prime}})-\cup_{j=1}^{i}\Gamma_{i} for 1i11\leq i\leq\ell-1, U=Sym12(M)j=1ΓiU_{\ell}=\mathrm{Sym}^{2}_{\leq 1}(M)-\cup_{j=1}^{\ell}\Gamma_{i}, and U0=Sym12(M1Mm)Γ1U_{0}=\mathrm{Sym}^{2}_{\leq 1}(M_{1}\oplus M_{m})-\Gamma_{1}. Let WiUiW_{i}\subset U_{i} be the divisor defined by ω(ϕm(vm),um)=0\omega(\phi_{m}(v_{m}),u_{m})=0 for 0i0\leq i\leq\ell. Assume χ=1\chi=1. We have

(4.3.2) (1)2n1χc(X~,KlG^,𝐏mst(χ,ϕ))=χc(Sym12(M1Mm))+χc(Γ1)+χc(Γ1)χc(Γ1Γ1).\displaystyle(-1)^{2n-1}\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi))=-\chi_{c}(\mathrm{Sym}_{\leq 1}^{2}(M_{1}\oplus M_{m}))+\chi_{c}(\Gamma_{1})+\chi_{c}(\Gamma_{1}^{\prime})-\chi_{c}(\Gamma_{1}\cap\Gamma_{1}^{\prime}).
Proof.

We assume 𝒦\mathcal{K} is the trivial character sheaf as in [YunEpipe, Prop 6.8]. By definition, U=𝔊λU_{\ell}=\mathfrak{G}_{\leq\lambda}. Let fi:U𝔸1f_{i}:U_{\ell}\to\mathbb{A}^{1} be the function [v]ω(ϕivi,umi)1γi(uv)[v]\mapsto\frac{\omega(\phi_{i}v_{i},u_{m-i})}{1-\gamma_{i}^{\prime}(u\cdot v)}. This function only depends on the coordinates u1,,ui,umi,,um,v1,,vi,vmi,,vmu_{1},\cdots,u_{i},u_{m-i},\cdots,u_{m},v_{1},\cdots,v_{i},v_{m-i},\cdots,v_{m}. Let fi=f1+fif_{\leq i}=f_{1}+\cdots f_{i}. Let fm=ω(ϕmvm,um)f_{m}=\omega(\phi_{m}v_{m},u_{m}).

Consider the projection π2:𝔾mrot×𝔊λ𝔊λ\pi_{2}:\mathbb{G}_{m}^{\mathrm{rot}}\times\mathfrak{G}_{\lambda}\to\mathfrak{G}_{\lambda}. The stalk of π2,!fϕASψ\pi_{2,!}f_{\phi}^{\ast}\mathrm{AS}_{\psi} over uvu\cdot v is

fASψHc(𝔾mrot,Tfm(uv)ASψ)f^{\ast}_{\leq\ell}\mathrm{AS}_{\psi}\otimes\mathrm{H}^{\ast}_{c}(\mathbb{G}_{m}^{\mathrm{rot}},T^{\ast}_{f_{m}(u\cdot v)}\mathrm{AS}_{\psi})

where Tfm(uv)T_{f_{m}(u\cdot v)} is the map 𝔾m𝔸1\mathbb{G}_{m}\to\mathbb{A}^{1} given by multiplication by fm(uv)f_{m}(u\cdot v).

When fm(uv)=0f_{m}(u\cdot v)=0 we have Hc(𝔾mrot,Tfm(uv)ASψ)=Hc(𝔾m,0ASψ)=Hc(𝔾m,¯)\mathrm{H}^{\ast}_{c}(\mathbb{G}_{m}^{\mathrm{rot}},T^{\ast}_{f_{m}(u\cdot v)}\mathrm{AS}_{\psi})=\mathrm{H}^{\ast}_{c}(\mathbb{G}_{m},0^{\ast}\mathrm{AS}_{\psi})=\mathrm{H}^{\ast}_{c}(\mathbb{G}_{m},\overline{\mathbb{Q}_{\ell}}). When fm(uv)0f_{m}(u\cdot v)\neq 0, since Hc(𝔸1,ASψ)=0H_{c}^{\ast}(\mathbb{A}^{1},\mathrm{AS}_{\psi})=0, we have Hc(𝔾mrot,Tfm(uv)ASψ)=¯[1]\mathrm{H}^{\ast}_{c}(\mathbb{G}_{m}^{\mathrm{rot}},T^{\ast}_{f_{m}(u\cdot v)}\mathrm{AS}_{\psi})=\overline{\mathbb{Q}_{\ell}}[-1]. Therefore we have

(4.3.3) (1)2n1χc(X~,KlG^,𝐏mst(1,ϕ))=χc(𝔊λ,π2!fϕASψ)=χc(U,fASψ)+χc(W,fASψ)\displaystyle\begin{split}(-1)^{2n-1}\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(1,\phi))&=\chi_{c}(\mathfrak{G}_{\leq\lambda},\pi_{2!}f_{\phi}^{\ast}\mathrm{AS}_{\psi})\\ &=-\chi_{c}(U_{\ell},f^{\ast}_{\leq\ell}\mathrm{AS}_{\psi})+\chi_{c}(W_{\ell},f^{\ast}_{\leq\ell}\mathrm{AS}_{\psi})\end{split}

We will show that the equalities

(4.3.4) χc(Ui+1,fi+1ASψ)=χc(Ui,fiASψ)\displaystyle\chi_{c}(U_{i+1},f^{\ast}_{\leq i+1}\mathrm{AS}_{\psi})=\chi_{c}(U_{i},f^{\ast}_{\leq i}\mathrm{AS}_{\psi})

and

(4.3.5) χc(Wi+1,fi+1ASψ)=χc(Wi,fiASψ)\displaystyle\chi_{c}(W_{i+1},f^{\ast}_{\leq i+1}\mathrm{AS}_{\psi})=\chi_{c}(W_{i},f^{\ast}_{\leq i}\mathrm{AS}_{\psi})

holds for 1i11\leq i\leq\ell-1. Consider the projection p:Ui+1Uip:U_{i+1}\to U_{i}^{\prime} by forgetting the Mmi2M_{m-i-2} component. We have

p!fi+1ASψ=fi+1ASψp!¯.p_{!}f_{\leq i+1}^{\ast}\mathrm{AS}_{\psi}=f^{\ast}_{\leq i+1}\mathrm{AS}_{\psi}\otimes p_{!}\overline{\mathbb{Q}_{\ell}}.

Hence

χc(Ui+1,fi+1ASψ)=χc(Ui,fi+1ASψp!¯)\chi_{c}(U_{i+1},f_{\leq i+1}^{\ast}\mathrm{AS}_{\psi})=\chi_{c}(U_{i}^{\prime},f_{\leq i+1}^{\ast}\mathrm{AS}_{\psi}\otimes p_{!}\overline{\mathbb{Q}_{\ell}})

Fix uvUiu^{\prime}\cdot v^{\prime}\in U_{i}^{\prime} such that

u=(u1,,ui+1,umi1,,um),u^{\prime}=(u_{1}^{\prime},\cdots,u_{i+1}^{\prime},u_{m-i-1}^{\prime},\cdots,u_{m}^{\prime}),
v=(v1,,vi+1,vmi1,,vm),v^{\prime}=(v_{1}^{\prime},\cdots,v_{i+1}^{\prime},v_{m-i-1}^{\prime},\cdots,v_{m}^{\prime}),

and let

γi=γ1(uv)++γi(uv)=ω(vm,u1)++ω(vm+1i,ui).\gamma_{i}^{\prime}=\gamma_{1}(u^{\prime}\cdot v^{\prime})+\cdots+\gamma_{i}(u^{\prime}\cdot v^{\prime})=\omega(v_{m}^{\prime},u_{1}^{\prime})+\cdots+\omega(v_{m+1-i}^{\prime},u_{i}).

The fiber of pp over uvu^{\prime}\cdot v^{\prime} is Mmi2M_{m-i-2}. Thus we have Hc(p1(uv))=¯[2d]H_{c}^{\ast}(p^{-1}(u^{\prime}\cdot v^{\prime}))=\overline{\mathbb{Q}_{\ell}}[-2d]. Moreover,

χc(Ui+1,fi+1ASψ)=χc(Ui,fi+1ASψ).\chi_{c}(U_{i+1},f_{\leq i+1}^{\ast}\mathrm{AS}_{\psi})=\chi_{c}(U_{i}^{\prime},f^{\ast}_{\leq i+1}\mathrm{AS}_{\psi}).

Since UiU_{i} can be identified with the subscheme of UiU_{i}^{\prime} where vi+1=0v_{i+1}=0, we have χc(Ui,fi+1ASψ)=χc(Ui,fiASψ)\chi_{c}(U_{i}^{\prime},f^{\ast}_{\leq i+1}\mathrm{AS}_{\psi})=\chi_{c}(U_{i},f^{\ast}_{\leq i}\mathrm{AS}_{\psi}).

Recall that U0=Sym12(M1Mm)Γ1U_{0}=\mathrm{Sym}^{2}_{\leq 1}(M_{1}\oplus M_{m})-\Gamma_{1}. Consider the projection p0:U1U0p_{0}:U_{1}\to U_{0} by forgetting the Mm2,Mm1M_{m-2},M_{m-1} component. We have

χc(U1,f1ASψ)=χc(U0,0ASψp0!f1ASψ)\chi_{c}(U_{1},f_{1}^{\ast}\mathrm{AS}_{\psi})=\chi_{c}(U_{0},0^{\ast}\mathrm{AS}_{\psi}\otimes p_{0!}f_{1}^{\ast}\mathrm{AS}_{\psi})

Fix uvU0u\cdot v\in U_{0} where u=(u1,um)u=(u_{1},u_{m}) and v=(v1,vm)v=(v_{1},v_{m}). By definition, γ1=ω(vm,u1)\gamma_{1}=\omega(v_{m},u_{1}). The fiber of p0p_{0} is Mm2Mm1M_{m-2}\otimes M_{m-1}, therefore Hc(p01(uv))=¯[4d]H_{c}^{\ast}(p_{0}^{-1}(u\cdot v))=\overline{\mathbb{Q}_{\ell}}[-4d]. One deduce that

(4.3.6) χc(U1,f1ASψ)=χc(U0).\displaystyle\chi_{c}(U_{1},f_{1}^{\ast}\mathrm{AS}_{\psi})=\chi_{c}(U_{0}).

Consider the projection p:Wi+1Wip:W_{i+1}\to W_{i}. We have

p!fi+1ASψ=fiASψp!fi+1ASψ.p_{!}f_{\leq i+1}^{\ast}\mathrm{AS}_{\psi}=f^{\ast}_{\leq i}\mathrm{AS}_{\psi}\otimes p_{!}f_{i+1}^{\ast}\mathrm{AS}_{\psi}.

Fix uvWiu^{\prime}\cdot v^{\prime}\in W_{i}. The fiber of p1(uv)p^{-1}(u^{\prime}\cdot v^{\prime}) is {(ui+1,umi2)Mi+1×Mmi2)|ω(vmi,ui+1)+Γi1}\{(u_{i+1},u_{m-i-2})\in M_{i+1}\times M_{m-i-2})|\omega(v_{m-i}^{\prime},u_{i+1})+\Gamma_{i}\neq 1\}. The function fi+1f_{i+1} along the fiber p1(uv)p^{-1}(u^{\prime}\cdot v^{\prime}) is a linear function in vi+1v_{i+1} (the vector parallel to ui+1u_{i+1}) given by fi+1=ω(ϕi+1vi+1,umi1)1ω(vmi,ui+1)Γif_{i+1}=\frac{\omega(\phi_{i+1}v_{i+1},u_{m-i-1})}{1-\omega(v^{\prime}_{m-i},u_{i+1})-\Gamma_{i}}. The stalk of p1,!fi+1ASψp_{1,!}f_{i+1}^{\ast}\mathrm{AS}_{\psi} at uvu^{\prime}\cdot v^{\prime} is Hc(p1(uv),fi+1ASψ)H_{c}^{\ast}(p^{-1}(u^{\prime}\cdot v^{\prime}),f_{i+1}^{\ast}\mathrm{AS}_{\psi}), which vanishes when umi10u^{\prime}_{m-i-1}\neq 0. We see that

Hc(p1(uv),fi+1ASψ)=Hc(p1(uv),0ASψ)=Hc(p1(uv),¯).H_{c}^{\ast}(p^{-1}(u^{\prime}\cdot v^{\prime}),f_{i+1}^{\ast}\mathrm{AS}_{\psi})=H_{c}^{\ast}(p^{-1}(u^{\prime}\cdot v^{\prime}),0^{\ast}\mathrm{AS}_{\psi})=H_{c}^{\ast}(p^{-1}(u^{\prime}\cdot v^{\prime}),\overline{\mathbb{Q}_{\ell}}).

When vmi=0v^{\prime}_{m-i}=0 this term equals ¯[4d]\overline{\mathbb{Q}_{\ell}}[-4d] and when vmi0v^{\prime}_{m-i}\neq 0, it is Hc(𝔾m)[4d+2]H_{c}^{\ast}(\mathbb{G}_{m})[-4d+2]. Therefore, p!fi+1ASψp_{!}f_{\leq i+1}^{\ast}\mathrm{AS}_{\psi} and the constant sheaf ¯\overline{\mathbb{Q}_{\ell}} are the same in the Grothendieck group of Dcb(Wi)D_{c}^{b}(W_{i}). Thus we have

χc(Wi+1,fi+1ASψ)=χc(Wi,fiASψp!fi+1ASψ)=χc(Wi,fiASψ)\chi_{c}(W_{i+1},f_{\leq i+1}^{\ast}\mathrm{AS}_{\psi})=\chi_{c}(W_{i},f^{\ast}_{\leq i}\mathrm{AS}_{\psi}\otimes p_{!}f_{i+1}^{\ast}\mathrm{AS}_{\psi})=\chi_{c}(W_{i},f^{\ast}_{\leq i}\mathrm{AS}_{\psi})

for 1i11\leq i\leq\ell-1. Similar as in the U0U_{0} case, we deduce that

(4.3.7) χc(W1,f1ASψ)=χc(W0).\displaystyle\chi_{c}(W_{1},f_{1}^{\ast}\mathrm{AS}_{\psi})=\chi_{c}(W_{0}).

Therefore by equation 4.3.3, equation 4.3.4, equation 4.3.6, and equation 4.3.5, equation 4.3.7, we have

(1)2n1χc(X~,KlG^,𝐏mst(χ,ϕ))=χc(U0)+χc(W0),(-1)^{2n-1}\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi))=-\chi_{c}(U_{0})+\chi_{c}(W_{0}),

where U0=Sym12(M1Mm)Γ1U_{0}=\mathrm{Sym}^{2}_{\leq 1}(M_{1}\oplus M_{m})-\Gamma_{1}, W0=Γ1Γ1W_{0}=\Gamma_{1}^{\prime}-\Gamma_{1} and Γ1\Gamma_{1}^{\prime} is the divisor defined by ω(ϕm(vm),um)=0\omega(\phi_{m}(v_{m}),u_{m})=0. Hence

(4.3.8) (1)2n1χc(X~,KlG^,𝐏mst(χ,ϕ))=χc(Sym12(M1Mm))+χc(Γ1)+χc(Γ1)χc(Γ1Γ1).\displaystyle(-1)^{2n-1}\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi))=-\chi_{c}(\mathrm{Sym}_{\leq 1}^{2}(M_{1}\oplus M_{m}))+\chi_{c}(\Gamma_{1})+\chi_{c}(\Gamma_{1}^{\prime})-\chi_{c}(\Gamma_{1}\cap\Gamma_{1}^{\prime}).

Now we consider the case when m=2m=2. Let U~0:=U0=Sym12(M1M2)Γ1\widetilde{U}_{0}:=U_{0}=\mathrm{Sym}_{\leq 1}^{2}(M_{1}\oplus M_{2})-\Gamma_{1} and let W~0U~0\widetilde{W}_{0}\subset\widetilde{U}_{0} be the divisor defined by ω(ϕ2v2,u2)=0\omega(\phi_{2}v_{2},u_{2})=0. We prove the following proposition.

Proposition 4.4.

Let U~0U~0\widetilde{U}_{0}^{\prime}\subset\widetilde{U}_{0} (resp. W0W0W_{0}^{\prime}\subset W_{0}) be the divisor defined by ω(ϕ1v1,u1)=0\omega(\phi_{1}v_{1},u_{1})=0 and let U~=U~0U~0\widetilde{U}=\widetilde{U}_{0}-\widetilde{U}_{0}^{\prime} (resp. W~=W~0W~0\widetilde{W}=\widetilde{W}_{0}-\widetilde{W}_{0}^{\prime}). Let (U~)(Sym12(M1M2))\mathbb{P}(\widetilde{U})\subset\mathbb{P}(\mathrm{Sym}_{\leq 1}^{2}(M_{1}\oplus M_{2})) (resp. (W~)\mathbb{P}(\widetilde{W})) be the projectivization of U~\widetilde{U} so that ukuu\cdot ku is equivalent to uuu\cdot u for ukuU~u\cdot ku\in\widetilde{U}. We have

χc(X~,KlG^,𝐏2st(1,ϕ))=χc((U~))χc((W~))χc(U~0)+χc(W~0).\displaystyle-\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{2}}(1,\phi))=\chi_{c}(\mathbb{P}(\widetilde{U}))-\chi_{c}(\mathbb{P}(\widetilde{W}))-\chi_{c}(\widetilde{U}_{0})+\chi_{c}(\widetilde{W}_{0}).
Proof.

we have

χc(X~,KlG^,𝐏2st(1,ϕ))\displaystyle-\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{2}}(1,\phi)) =χc(U~0,f1ASψ)+χc(W~0,f1ASψ)\displaystyle=-\chi_{c}(\widetilde{U}_{0},f_{1}^{\ast}\mathrm{AS}_{\psi})+\chi_{c}(\widetilde{W}_{0},f_{1}^{\ast}\mathrm{AS}_{\psi})
=χc(U~,f1ASψ)+χc(W~,f1ASψ)χc(U~0,f1ASψ)+χc(W~0,f1ASψ)\displaystyle=-\chi_{c}(\widetilde{U},f_{1}^{\ast}\mathrm{AS}_{\psi})+\chi_{c}(\widetilde{W},f_{1}^{\ast}\mathrm{AS}_{\psi})-\chi_{c}(\widetilde{U}_{0},f_{1}^{\ast}\mathrm{AS}_{\psi})+\chi_{c}(\widetilde{W}_{0},f_{1}^{\ast}\mathrm{AS}_{\psi})
=χc(U~,f1ASψ)+χc(W~,f1ASψ)χc(U~0,0ASψ)+χc(W~0,0ASψ)\displaystyle=-\chi_{c}(\widetilde{U},f_{1}^{\ast}\mathrm{AS}_{\psi})+\chi_{c}(\widetilde{W},f_{1}^{\ast}\mathrm{AS}_{\psi})-\chi_{c}(\widetilde{U}_{0},0^{\ast}\mathrm{AS}_{\psi})+\chi_{c}(\widetilde{W}_{0},0^{\ast}\mathrm{AS}_{\psi})
=χc(U~,f1ASψ)+χc(W~,f1ASψ)χc(U~0)+χc(W~0)\displaystyle=-\chi_{c}(\widetilde{U},f_{1}^{\ast}\mathrm{AS}_{\psi})+\chi_{c}(\widetilde{W},f_{1}^{\ast}\mathrm{AS}_{\psi})-\chi_{c}(\widetilde{U}_{0})+\chi_{c}(\widetilde{W}_{0})

By change of variable isomorphism and Ku¨\ddot{\mathrm{u}}nneth formula, we have

χc(U~,f1ASψ)+χc(W~,f1ASψ)\displaystyle-\chi_{c}(\widetilde{U},f_{1}^{\ast}\mathrm{AS}_{\psi})+\chi_{c}(\widetilde{W},f_{1}^{\ast}\mathrm{AS}_{\psi})
=\displaystyle= χc(𝔾m,ASψ)χc((U~))+χc(𝔾m,ASψ)χc((W~))\displaystyle-\chi_{c}(\mathbb{G}_{m},\mathrm{AS}_{\psi})\chi_{c}(\mathbb{P}(\widetilde{U}))+\chi_{c}(\mathbb{G}_{m},\mathrm{AS}_{\psi})\chi_{c}(\mathbb{P}(\widetilde{W}))
=\displaystyle= χc((U~))χc((W~))\displaystyle\chi_{c}(\mathbb{P}(\widetilde{U}))-\chi_{c}(\mathbb{P}(\widetilde{W}))

since χc(𝔾m,ASψ)=1\chi_{c}(\mathbb{G}_{m},\mathrm{AS}_{\psi})=-1. The proposition follows. ∎

The proof of Theorem 4.2 then reduces to the calculation of the compactly supported Euler characteristic of each pieces in Proposition 4.3 and 4.4, therefore boils down to the proof of a few lemmas listed below.

Proof of Theorem 4.2.

Let mm be a regular elliptic number.

Assume m \neq 2. We first notice that Sym12(M1Mm)\mathrm{Sym}_{\leq 1}^{2}(M_{1}\oplus M_{m}) is a cone over 𝐏(M1Mm)\mathbf{P}(M_{1}\oplus M_{m}), i.e. a line bundle \mathcal{L} over M1MmM_{1}\oplus M_{m} minus the zero section. Let EE be the total space of the line bundle and denote by E0E_{0} the zero section. Let EcE_{c} be the complement of the disk bundle D()D(\mathcal{L}).

Lemma 4.5.
χc(Sym12(M1Mm))=1.\chi_{c}(\mathrm{Sym}_{\leq 1}^{2}(M_{1}\oplus M_{m}))=1.
Proof.

We have

H(E/E0)=H(E/E0,Ec)=H(E/Ec,E0)\displaystyle H_{\star}(E/E_{0})=H_{\star}(E/E_{0},E_{c})=H_{\star}(E/E_{c},E_{0})

where E/EcE/E_{c} is the Thom space of \mathcal{L}. Then by Thom isomorphism

H+2(E/Ec,pt)H(M1Mm)H_{\star+2}(E/E_{c},pt)\cong H_{\star}(M_{1}\oplus M_{m})

The result follows from the Poincare duality. ∎

Lemma 4.6.
χc(Γ1)=0.\chi_{c}(\Gamma_{1})=0.
Proof.

Since Γ1\Gamma_{1} is defined by the symplectic form ω(u1,vm)=1\omega(u_{1},v_{m})=1, vmv_{m} is determined by the choice of u1u_{1} up to a non-zero scalar. Therefore we have the fibration

{vmω(u1,vm)=1}Γ1M1\{0}\{v_{m}\mid\omega(u_{1},v_{m})=1\}\to\Gamma_{1}\to M_{1}\backslash\{0\}

where the fiber {vmω(u1,vm)=1}\{v_{m}\mid\omega(u_{1},v_{m})=1\} contains a summand k.k^{*}. The fibration property of the Euler characteristic implies that

χc(Γ1)=χc(k)χc(M1\{0})=0.\chi_{c}(\Gamma_{1})=\chi_{c}(k^{*})\cdot\chi_{c}(M_{1}\backslash\{0\})=0.

Lemma 4.7.
χc(Γ1)=1.\chi_{c}(\Gamma_{1}^{\prime})=1.
Proof.

Let (Γ1)\mathbb{P}(\Gamma_{1}^{\prime}) be the projectivization of Γ1\Gamma_{1}^{\prime}, it is a projective quadric defined by the same homogenous equation as Γ1\Gamma_{1}^{\prime}. For Γ1{0}\Gamma_{1}^{\prime}-\{0\}, we have the following fibration

{(k1,u,vm)}Γ1{0}(Γ1)\{(k_{1},u,v_{m})\}\to\Gamma_{1}^{\prime}-\{0\}\to\mathbb{P}(\Gamma_{1}^{\prime})

over a fixed [v1](Γ1)(M1),[v_{1}]\in\mathbb{P}(\Gamma_{1}^{\prime})\subset\mathbb{P}(M_{1}), where k1v1k_{1}v_{1} is the preimage of [v1].[v_{1}]. But it is still a kk^{*}-bundle, therefore χc(Γ1{0})=0.\chi_{c}(\Gamma_{1}^{\prime}-\{0\})=0.

The fiber over v1=0v_{1}=0 is nothing but Sym12(Mm)\mathrm{Sym}_{\leq 1}^{2}(M_{m}). The lemma then follows from Lemma 4.5.

Lemma 4.8.
χc(Γ1Γ1)=d.\chi_{c}(\Gamma_{1}\cap\Gamma_{1}^{\prime})=-d.
Proof.

Let X=Γ1Γ1.X=\Gamma_{1}\cap\Gamma_{1}^{\prime}. Then

χc(X)\displaystyle\chi_{c}(X) =χc(Γ1Sym12)χc(Γ1ω(v1,um)=0))\displaystyle=\chi_{c}(\Gamma_{1}^{\prime}\cap\operatorname{Sym}^{2}_{\leq 1})-\chi_{c}(\Gamma_{1}^{\prime}\cap\omega(v_{1},u_{m})=0))
=χc(Γ1)χc(ω(ϕm(vm),um)=0(ω(v1,um)=0))\displaystyle=\chi_{c}(\Gamma_{1}^{\prime})-\chi_{c}(\omega(\phi_{m}(v_{m}),u_{m})=0\cap(\omega(v_{1},u_{m})=0))

Let Γ0:=ω(ϕm(vm),um)=0(ω(v1,um)=0)\Gamma_{0}:=\omega(\phi_{m}(v_{m}),u_{m})=0\cap(\omega(v_{1},u_{m})=0). Then Γ0\Gamma_{0} minus V(v1=0)V(v_{1}=0) is a bundle over (M1)\mathbb{P}(M_{1}). Consider its projection

ψ:Γ0V(v1=0)(M1)\psi:\Gamma_{0}-V(v_{1}=0)\to\mathbb{P}(M_{1})
(v,u)[v1].(v,u)\mapsto[v_{1}].

The fiber over each [v1][v_{1}] is a cone over (M)\mathbb{P}(M), i.e., the quadric ω(ϕm(vm),um)\omega(\phi_{m}(v_{m}),u_{m}) intersecting a hyperplane, therefore an affine quadric with one less variable. We have

χc(Γ0V(v1=0))=1χc((M1))=d.\chi_{c}(\Gamma_{0}-V(v_{1}=0))=1\cdot\chi_{c}(\mathbb{P}(M_{1}))=d.

Assume v1=0v_{1}=0. The intersection then becomes a single quadric ω(ϕm(vm),um)=0\omega(\phi_{m}(v_{m}),u_{m})=0 in Sym12(Mm)\operatorname{Sym}^{2}_{\leq 1}(M_{m}), whose compactly supported Euler characteristic is 11 by Lemma 4.7. Putting things together, one easily deduces the lemma.

Therefore 4.3.1 holds for m3m\geq 3.

Assume m=2. Throughout the proof, we make the following shorthand

  • U~0:=Q1Γ1\tilde{U}_{0}^{\prime}:=Q_{1}-\Gamma_{1}, where Q1Q_{1} is the quadric defined by ω(ϕ1v1,u1);\omega(\phi_{1}v_{1},u_{1});

  • W~0:=Q2Γ1\tilde{W}_{0}:=Q_{2}-\Gamma_{1}, where Q2Q_{2} is the quadric defined by ω(ϕ2v2,u2).\omega(\phi_{2}v_{2},u_{2}).

We deduce from Lemma 4.5-4.8 and (4.3.2) that

χc(U0)+χc(W0)=χc(Sym12(M1M2))+χc(Γ1)+χc(Q2)χc(Γ1Q2)=d.-\chi_{c}(U_{0})+\chi_{c}(W_{0})=-\chi_{c}(\mathrm{Sym}_{\leq 1}^{2}(M_{1}\oplus M_{2}))+\chi_{c}(\Gamma_{1})+\chi_{c}(Q_{2})-\chi_{c}(\Gamma_{1}\cap Q_{2})=d.

The theorem follows from the following proposition.

Proposition 4.9.
χc((U))χc((W))=0.\chi_{c}(\mathbb{P}(U))-\chi_{c}(\mathbb{P}(W))=0.
Proof.

We have

χc((U))χc((W))=χc((U0))χc((U0))χc((W0))+χc((W0U0))\chi_{c}(\mathbb{P}(U))-\chi_{c}(\mathbb{P}(W))=\chi_{c}(\mathbb{P}(U_{0}))-\chi_{c}(\mathbb{P}(U_{0}^{\prime}))-\chi_{c}(\mathbb{P}(W_{0}))+\chi_{c}(\mathbb{P}(W_{0}\cap U_{0}^{\prime}))

so it is sufficient to do the computation term by term. First,

(4.3.9) χc((U0))\displaystyle\chi_{c}(\mathbb{P}(U_{0})) =χc((M1M2))χc((Γ1))\displaystyle=\chi_{c}(\mathbb{P}(M_{1}\oplus M_{2}))-\chi_{c}(\mathbb{P}(\Gamma_{1}))
(4.3.10) =2d(2d2d)\displaystyle=2d-(2d-2d)
(4.3.11) =2d\displaystyle=2d

Then we notice that

(4.3.12) χc((U0))\displaystyle\chi_{c}(\mathbb{P}(U_{0}^{\prime})) =χc((Q1)χc((Q1Γ1)).\displaystyle=\chi_{c}(\mathbb{P}(Q_{1})-\chi_{c}(\mathbb{P}(Q_{1}\cap\Gamma_{1})).

By Lemma 5.4, χc((Q1))=2d\chi_{c}(\mathbb{P}(Q_{1}))=2d (when dd is even, 2d12d-1 when dd is odd). There are multiple ways to compute χc((Q1Γ1))\chi_{c}(\mathbb{P}(Q_{1}\cap\Gamma_{1})). For example one can take the projection onto the quadric (U0)\mathbb{P}(U_{0}^{\prime}) by sending [v1,v2][v_{1},v_{2}] to [v1][v_{1}]. The fiber Fv1F_{v_{1}} is M2M_{2} minus a hyperplane, therefore has Euler characteristic χc(Fv1)=11=0\chi_{c}(F_{v_{1}})=1-1=0. The fibration property implies χc((Q1Γ1)=0\chi_{c}(\mathbb{P}(Q_{1}\cap\Gamma_{1})=0, so χc((U0))=2d\chi_{c}(\mathbb{P}(U_{0}^{\prime}))=2d(when dd is even, 2d12d-1 when dd is odd).

We have χc((W0))=2d\chi_{c}(\mathbb{P}(W_{0}))=2d (when dd is even, 2d12d-1 when dd is odd) by the same reason.

The Euler characteristic of χc((W0U0))\chi_{c}(\mathbb{P}(W_{0}\cap U_{0}^{\prime})) can be computed as follows, combining with Lemma 5.4 and Lemma 5.5.

χc((W0U0))\displaystyle\chi_{c}(\mathbb{P}(W_{0}\cap U_{0}^{\prime})) =χc((Q1Q2))χc((Γ1Q1Q2))\displaystyle=\chi_{c}(\mathbb{P}(Q_{1}\cap Q_{2}))-\chi_{c}(\mathbb{P}(\Gamma_{1}\cap Q_{1}\cap Q_{2}))
=χc((Q1Q2))[χc((Q1Q2))χc(Γ1Q1Q2)]\displaystyle=\chi_{c}(\mathbb{P}(Q_{1}\cap Q_{2}))-[\chi_{c}(\mathbb{P}(Q_{1}\cap Q_{2}))-\chi_{c}(\Gamma^{\perp}_{1}\cap Q_{1}\cap Q_{2})]
=2d, if d is even\displaystyle=2d,\text{ if }d\text{ is even }
=2d2, if d is odd\displaystyle=2d-2,\text{ if }d\text{ is odd }

where Γ1\Gamma^{\perp}_{1} is the quadric in (M1M2)\mathbb{P}(M_{1}\oplus M_{2}) defined by ω(v1,v2)=0.\omega(v_{1},v_{2})=0. One then deduces the lemma by putting together. ∎

5. Euler Characteristic for Split and Quasi-split Orthogonal Groups

The goal of this section is to prove Theorem 1.2. Throughout the section we work over k¯\bar{k} and ignore all the Tate twists. Let GG be a split or quasi-split orthogonal group over KK.

5.1. The set-up

Assume char(k)2\mathrm{char}(k)\neq 2. Let (M,q)(M,q) be a quadratic space of dimension 2n2n or 2n+12n+1 over k. Let (,)(\cdot,\cdot): M×MkM\times M\to k be the associated symmetric bilinear form (x,y)=q(x+y)q(x)q(y)(x,y)=q(x+y)-q(x)-q(y). The regular elliptic numbers of mm of the root systems of type Bn,DnB_{n},D_{n} and Dn2{}^{2}D_{n} are in bijection with

  • Type Bn(dimM=2n+1)B_{n}(\operatorname{dim}M=2n+1) : divisors dnd\mid n (corresponding m=2n/d)\left.m=2n/d\right);

  • Type Dn(dimM=2n)D_{n}(\operatorname{dim}M=2n) : even divisors dnd\mid n (corresponding m=2n/d)\left.m=2n/d\right) or odd divisors dn1d\mid n-1 (corresponding m=2(n1)/dm=2(n-1)/d );

  • Type Dn2(dimM=2n){}^{2}D_{n}(\operatorname{dim}M=2n) : odd divisors dnd\mid n (corresponding m=2n/d)\left.m=2n/d\right) or even divisors dn1d\mid n-1 (corresponding m=2(n1)/dm=2(n-1)/d ).

Let m=2m=2\ell for some \ell. Fix a decomposition

(5.1.1) M=M0M1M1MM+1Mm1\displaystyle M=M_{0}\oplus M_{1}\oplus\cdots M_{\ell-1}\oplus M_{\ell}\oplus M_{\ell+1}\oplus\cdots\oplus M_{m-1}

where dimMi=d\operatorname{dim}M_{i}=d for i=1,,1,+1,,m1i=1,\cdots,\ell-1,\ell+1,\cdots,m-1, dimM0\operatorname{dim}M_{0} and dimM\operatorname{dim}M_{\ell} are either dd or d+1d+1, and we make sure that when dimM=2n+1\operatorname{dim}M=2n+1, dimM0\operatorname{dim}M_{0} is even. We see that

  • Type Bn:dimM0B_{n}:\operatorname{dim}M_{0} is even and dimM\operatorname{dim}M_{\ell} is odd;

  • Type Dn:dimM0=dimMD_{n}:\operatorname{dim}M_{0}=\operatorname{dim}M_{\ell} is even;

  • Type Dn2:dimM0=dimM{}^{2}D_{n}:\operatorname{dim}M_{0}=\operatorname{dim}M_{\ell} is odd.

The decomposition 5.1.1 satisfy (Mi,Mj)=0\left(M_{i},M_{j}\right)=0 unless i+j0modmi+j\equiv 0\bmod m. The restriction of qq to M0M_{0} and MM_{\ell} are denoted by q0q_{0} and qq_{\ell}. The pairing (,)(\cdot,\cdot) induce an isomorphism MiMmiM_{i}^{*}\cong M_{m-i}. Let M+=i>0MiM_{+}=\oplus_{i>0}M_{i}, then M=M0M+M=M_{0}\oplus M_{+} and correspondingly q=q0q+q=q_{0}\oplus q_{+}.

Let 𝐏~mG(K)\widetilde{\mathbf{P}}_{m}\subset G(K) be the stabilizer of the lattice chain

ΛmΛm1Λ0\Lambda_{m}\supset\Lambda_{m-1}\supset\cdots\supset\Lambda_{0}

where

Λi=0jiMj𝒪K+i<jm1Mjω¯𝒪K\Lambda_{i}=\sum_{0\leq j\leq i}M_{j}\otimes\mathcal{O}_{K}+\sum_{i<j\leq m-1}M_{j}\otimes\bar{\omega}\mathcal{O}_{K}

where ω¯\bar{\omega} is a uniformizer of 𝒪F.\mathcal{O}_{F}. Its reductive quotient L~m\widetilde{L}_{m} is the subgroup of O(M0,q0)×i=11GL(Mi)×O(M,q)\mathrm{O}\left(M_{0},q_{0}\right)\times\prod_{i=1}^{\ell-1}\mathrm{GL}(M_{i})\times\mathrm{O}\left(M_{\ell},q_{\ell}\right) of index two consisting of (g0,,g)\left(g_{0},\cdots,g_{\ell}\right) where det(g0)=det(g)\operatorname{det}\left(g_{0}\right)=\operatorname{det}\left(g_{\ell}\right).

The subgroup 𝐏m𝐏~m\mathbf{P}_{m}\subset\widetilde{\mathbf{P}}_{m}, defined as the kernel of 𝐏~mL~m{±1}\widetilde{\mathbf{P}}_{m}\rightarrow\widetilde{L}_{m}\rightarrow\{\pm 1\} by taking the determinant of the first factor, is an admissible parahoric subgroup of G(K)G(K) with m(𝐏m)=mm\left(\mathbf{P}_{m}\right)=m. The vector space Vm:=V𝐏mV_{m}:=V_{\mathbf{P}_{m}} is

Vm=Hom(M1,M0)Hom(M2,M1)Hom(M,M1).V_{m}=\operatorname{Hom}\left(M_{1},M_{0}\right)\oplus\operatorname{Hom}\left(M_{2},M_{1}\right)\oplus\cdots\oplus\operatorname{Hom}\left(M_{\ell},M_{\ell-1}\right).

And similarly, we can arrange VmV_{m} into a cyclic quiver

M1{M_{1}}M2{M_{2}}{\cdots}M1{M_{\ell-1}}M0{M_{0}}M{M_{\ell}}Mm1{M_{m-1}}Mm2{M_{m-2}}{\cdots}M+1{M_{\ell+1}}ϕ1\scriptstyle{\definecolor[named]{.}{rgb}{0,0,1}\color[rgb]{0,0,1}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,1}\phi_{1}}ψ0\scriptstyle{\psi_{0}}ψ1\scriptstyle{\psi_{1}}ϕ1\scriptstyle{\definecolor[named]{.}{rgb}{0,0,1}\color[rgb]{0,0,1}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,1}\phi_{\ell-1}}ψ2\scriptstyle{\psi_{\ell-2}}ϕ1\scriptstyle{\definecolor[named]{.}{rgb}{0,0,1}\color[rgb]{0,0,1}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,1}\phi_{\ell-1}}ψm1\scriptstyle{\psi_{m-1}}ϕ0\scriptstyle{\definecolor[named]{.}{rgb}{0,0,1}\color[rgb]{0,0,1}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,1}\phi_{0}}ψ1\scriptstyle{\psi_{\ell-1}}ϕ\scriptstyle{\definecolor[named]{.}{rgb}{0,0,1}\color[rgb]{0,0,1}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,1}\phi_{\ell}}ψm2\scriptstyle{\psi_{m-2}}ϕm1\scriptstyle{\definecolor[named]{.}{rgb}{0,0,1}\color[rgb]{0,0,1}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,1}\phi_{m-1}}ϕm2\scriptstyle{\definecolor[named]{.}{rgb}{0,0,1}\color[rgb]{0,0,1}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,1}\phi_{m-2}}ψ+1\scriptstyle{\psi_{\ell+1}}ψ\scriptstyle{\psi_{\ell}}ϕ+1\scriptstyle{\definecolor[named]{.}{rgb}{0,0,1}\color[rgb]{0,0,1}\definecolor[named]{pgfstrokecolor}{rgb}{0,0,1}\phi_{\ell+1}}

where the involution τ\tau sends {ψi:Mi+1Mi}\left\{\psi_{i}:M_{i+1}\rightarrow M_{i}\right\} to {ψm1i:Mm1iMm+1i}\left\{-\psi_{m-1-i}^{*}:M_{m-1-i}^{*}\rightarrow M_{m+1-i}^{*}\right\}. The dual space VmV_{m}^{*} is the space of τ\tau-invariant cyclic quivers with all the arrows reversed. Let ϕi:MiMi+1\phi_{i}:M_{i}\to M_{i+1} be the arrows. Then ϕ=(ϕ0,,ϕm1)Vm\phi=\left(\phi_{0},\ldots,\phi_{m-1}\right)\in V_{m}^{*} is stable if and only if

  • All the maps ϕi\phi_{i} have the maximal possible rank;

  • We have two quadratic forms on M0:q0M_{0}:q_{0} and the pullback of qq_{\ell} to M0M_{0} via the map ϕ1ϕ0:M0M\phi_{\ell-1}\cdots\phi_{0}:M_{0}\rightarrow M_{\ell}. They are in general position in the same sense as the symplectic and unitary cases.

5.2. The local system.

The moduli stack BunG(𝐏0~,𝐏+)\operatorname{Bun}_{G}(\widetilde{\mathbf{P}_{0}},\mathbf{P}^{+}_{\infty}) classifies 4-tuples

(,({}),({0}),δ),(\mathcal{E},\mathcal{E}(-\{\infty\}),\mathcal{E}(-\{0\}),\delta),

where the vector bundle \mathcal{E}, an increasing filtration ({})\mathcal{E}(-\{\infty\}) of the fiber of \mathcal{E} at \infty , a decreasing filtration ({0})\mathcal{E}(-\{0\}) at 0 and a trivialization δ\delta of 𝒪X\mathcal{O}_{X} are defined in [YunEpipe, Sec. 8.3]. The group ind-scheme 𝔊\mathfrak{G} is the group of orthogonal automorphisms of |X{1}\left.\mathcal{E}\right|_{X-\{1\}} preserving all the auxiliary data specified above. Let λ𝕏(T)\lambda\in\mathbb{X}_{*}(T) be the dominant minuscule coweight such that VλV_{\lambda} is the standard representation of the dual group G^=Sp2n\widehat{G}=\operatorname{Sp}_{2n} or SO2n\operatorname{SO}_{2n}. The subscheme 𝔊λ\mathfrak{G}_{\leq\lambda} consists of those g𝔊G(F)g\in\mathfrak{G}\subset G(F) whose entries have at most simple poles at t=1t=1, and Rest=1\operatorname{Res}_{t=1}g has rank one. [YunEpipe, Lem. 8.5 (3)] shows that the subscheme 𝔊λ\mathfrak{G}_{\leq\lambda} can be embedded as an open subscheme of the quadric Q(q)Q(q) in (M)\mathbb{P}(M) defined by q=0q=0. Let q[i,mi]q_{[i,m-i]} be the restriction of the quadratic form qq to MiMmiM_{i}\oplus\cdots M_{m-i} that extended to MM by zero on the rest of the direct summands. Similar to Proposition 4.1 in the case of symplectic groups, the following proposition gives an explicit description of KlG^,𝐏st(ϕ)\operatorname{Kl}_{\widehat{G},\mathbf{P}}^{st}(\phi) when GG is split or quasi-split orthogonal.

Proposition 5.1.

[YunEpipe, Cor. 8.7] Let ϕ=(ϕ0,ϕ1,,ϕm1)Vm st (k)\phi=\left(\phi_{0},\phi_{1},\ldots,\phi_{m-1}\right)\in V_{m}^{*\text{ st }}(k) be a stable functional. Recall that 𝔊λ\mathfrak{G}_{\lambda} in this case is Q(q)i=1Q(q[i,mi])Q(q)-\cup_{i=1}^{\ell}Q\left(q_{[i,m-i]}\right). Let fϕ:X~×𝔊λ𝔸1f_{\phi}:\widetilde{X}^{\circ}\times\mathfrak{G}_{\lambda}\rightarrow\mathbb{A}^{1} be given by

fϕ(x,[v])=(ϕ0v0,vm1)q[1,m1](v)xi=11(ϕivi,vmi1)q[i+1,mi1](v).f_{\phi}(x,[v])=-\frac{\left(\phi_{0}v_{0},v_{m-1}\right)}{q_{[1,m-1]}(v)}x-\sum_{i=1}^{\ell-1}\frac{\left(\phi_{i}v_{i},v_{m-i-1}\right)}{q_{[i+1,m-i-1]}(v)}.

Let π:X~×𝔊λX~\pi:\widetilde{X}^{\circ}\times\mathfrak{G}_{\lambda}\rightarrow\widetilde{X}^{\circ} be the projection. Then we have an isomorphism over X~\widetilde{X}^{\circ}

KlG^,mSt(𝟏,ϕ)π!fϕASψ[dimM2](dimM22).\mathrm{Kl}_{\widehat{G},m}^{\mathrm{St}}(\mathbf{1},\phi)\cong\pi_{!}f_{\phi}^{*}\mathrm{AS}_{\psi}[\operatorname{dim}M-2]\left(\frac{\operatorname{dim}M-2}{2}\right).

5.3. Computation of Euler characteristic

The goal of this subsection is to prove Theorem 5.2 for the regular elliptic numbers of mm of the root systems of type Bn,Dn,Dn2B_{n},D_{n},{}^{2}D_{n} and all parities of dimM\dim M and dd.

Theorem 5.2.

We have

χc(X~,KlG^,𝐏mst(χ,ϕ))={2dBn,2dDn, ω1 is non-degenerate,2(d+1)Dn, ω1 is degenerate,2dDn2, ω1 is non-degenerate,2(d+1)Dn2, ω1 is degenerate.-\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi))=\begin{cases}2d&B_{n},\\ 2d&D_{n},\text{ }\omega_{1}\text{ is non-degenerate,}\\ 2(d+1)&D_{n},\text{ }\omega_{1}\text{ is degenerate,}\\ 2d&{}^{2}D_{n},\text{ }\omega_{1}\text{ is non-degenerate,}\\ 2(d+1)&{}^{2}D_{n},\text{ }\omega_{1}\text{ is degenerate.}\end{cases}

By a similar argument as in [Katz] the Swan conductor of KlG^,𝐏mst(χ,ϕ)\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi) at \infty does not depend on χ\chi, so does the Euler characteristic of KlG^,𝐏mst(χ,ϕ)\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi). We make an inductive argument as follows.

Proposition 5.3.

Let Qi(M)Q_{i}\subset\mathbb{P}(M) be the quadric defined by q[i+1,m+i1]=0q_{[\ell-i+1,m-\ell+i-1]}=0. Let Ui(M0j=i+1m+i1Mj)U_{i}\subset\mathbb{P}(M_{0}\oplus_{j=\ell-i+1}^{m-\ell+i-1}M_{j}) be the projective variety defined by Q(q)j=1iQjQ(q)-\cup_{j=1}^{i}Q_{j} for 1i1\leq i\leq\ell. Let WiUiW_{i}\subset U_{i} be the quadric defined by (ϕiϕ0v0,vm+i1)=0(\phi_{\ell-i}\cdots\phi_{0}v_{0},v_{m-\ell+i-1})=0 for 1i1\leq i\leq\ell. Assume χ=1\chi=1. We have

(5.3.1) (1)dimM2χc(X~,KlG^,𝐏mst(1,ϕ))\displaystyle(-1)^{\mathrm{dim}M-2}\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(1,\phi)) =χc(U1)+χc(W1)\displaystyle=-\chi_{c}(U_{1})+\chi_{c}(W_{1})
Proof.

We assume 𝒦\mathcal{K} is the trivial character sheaf as in [YunEpipe, Prop 6.8]. We take the approach similar to the symplectic case.

Let fi:U𝔸1f_{i}:U_{\ell}\to\mathbb{A}^{1} be the function [v](ϕivi,vmi1)q[i+1,mi1][v]\mapsto\frac{(\phi_{i}v_{i},v_{m-i-1})}{q_{[i+1,m-i-1]}} for 1i11\leq i\leq\ell-1. This function only depends on the coordinates vi,,vmiv_{i},\cdots,v_{m-i}. Let fi=fi+f1f_{\geq\ell-i}=f_{\ell-i}+\cdots f_{\ell-1}.

Consider the projection π2:𝔾mrot×𝔊λ𝔊λ\pi_{2}:\mathbb{G}_{m}^{\mathrm{rot}}\times\mathfrak{G}_{\lambda}\to\mathfrak{G}_{\lambda}. The stalk of π2,!fϕASψ\pi_{2,!}f_{\phi}^{\ast}\mathrm{AS}_{\psi} over [v][v] is

fASψHc(𝔾mrot,Tf0([v])ASψ)f^{\ast}_{\leq\ell}\mathrm{AS}_{\psi}\otimes\mathrm{H}^{\ast}_{c}(\mathbb{G}_{m}^{\mathrm{rot}},T^{\ast}_{f_{0}([v])}\mathrm{AS}_{\psi})

where Tf0([v])T_{f_{0}([v])} is the map 𝔾m𝔸1\mathbb{G}_{m}\to\mathbb{A}^{1} given by multiplication by f0([v])f_{0}([v]).

When f0([v])=0f_{0}([v])=0 we have Hc(𝔾mrot,Tf0([v])ASψ)=Hc(𝔾m,0ASψ)=Hc(𝔾m,¯)\mathrm{H}^{\ast}_{c}(\mathbb{G}_{m}^{\mathrm{rot}},T^{\ast}_{f_{0}([v])}\mathrm{AS}_{\psi})=\mathrm{H}^{\ast}_{c}(\mathbb{G}_{m},0^{\ast}\mathrm{AS}_{\psi})=\mathrm{H}^{\ast}_{c}(\mathbb{G}_{m},\overline{\mathbb{Q}_{\ell}}). When f0([v])0f_{0}([v])\neq 0, since Hc(𝔸1,ASψ)=0H_{c}^{\ast}(\mathbb{A}^{1},\mathrm{AS}_{\psi})=0, we have Hc(𝔾mrot,Tf0([v])ASψ)=¯[1]\mathrm{H}^{\ast}_{c}(\mathbb{G}_{m}^{\mathrm{rot}},T^{\ast}_{f_{0}([v])}\mathrm{AS}_{\psi})=\overline{\mathbb{Q}_{\ell}}[-1]. Therefore we have

(5.3.2) (1)dimM2χc(X~,KlG^,𝐏mst(1,ϕ))=χc(𝔊λ,π2!fϕASψ)=χc(U,f1ASψ)+χc(W,f1ASψ)\displaystyle\begin{split}(-1)^{\mathrm{dim}M-2}\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(1,\phi))&=\chi_{c}(\mathfrak{G}_{\leq\lambda},\pi_{2!}f_{\phi}^{\ast}\mathrm{AS}_{\psi})\\ &=-\chi_{c}(U_{\ell},f^{\ast}_{\geq 1}\mathrm{AS}_{\psi})+\chi_{c}(W_{\ell},f^{\ast}_{\geq 1}\mathrm{AS}_{\psi})\end{split}

In the following we are going to show that for 1i11\leq i\leq\ell-1 we have

(5.3.3) χc(Ui+1,fiASψ)=χc(Ui,fi+1ASψ)\displaystyle\chi_{c}(U_{i+1},f^{\ast}_{\geq\ell-i}\mathrm{AS}_{\psi})=\chi_{c}(U_{i},f^{\ast}_{\geq\ell-i+1}\mathrm{AS}_{\psi})
(5.3.4) χc(Wi+1,fiASψ)=χc(Wi,fi+1ASψ)\displaystyle\chi_{c}(W_{i+1},f^{\ast}_{\geq\ell-i}\mathrm{AS}_{\psi})=\chi_{c}(W_{i},f^{\ast}_{\geq\ell-i+1}\mathrm{AS}_{\psi})

Let Ui(M0j=im+i1Mj)U_{i}^{\prime}\subset\mathbb{P}(M_{0}\oplus_{j=\ell-i}^{m-\ell+i-1}M_{j}) be defined by Q(q)j=1iQjQ(q)-\cup_{j=1}^{i}Q_{j}. Consider the projection p:Ui+1Uip:U_{i+1}\to U_{i}^{\prime} by forgetting the Mm+iM_{m-\ell+i} component. We have

p!fiASψ=fiASψp!¯.p_{!}f_{\geq\ell-i}^{\ast}\mathrm{AS}_{\psi}=f^{\ast}_{\geq\ell-i}\mathrm{AS}_{\psi}\otimes p_{!}\overline{\mathbb{Q}_{\ell}}.

Then we have

χc(Ui+1,fiASψ)=χc(Ui,fiASψp!¯).\chi_{c}(U_{i+1},f^{\ast}_{\geq\ell-i}\mathrm{AS}_{\psi})=\chi_{c}(U_{i}^{\prime},f^{\ast}_{\geq\ell-i}\mathrm{AS}_{\psi}\otimes p_{!}\overline{\mathbb{Q}_{\ell}}).

Fix [v]=[v0,vi,,vm+i1]Ui[v^{\prime}]=[v_{0},v_{\ell-i},\cdots,v_{m-\ell+i-1}]\in U_{i}^{\prime}, and we denote qi:=q[i+1,m+i1]q_{i}:=q_{[\ell-i+1,m-\ell+i-1]} . The fiber of pp over [v][v^{\prime}] is {vm+i|(vi,vm+i)+q(vm+i)=0,(vi,vm+i)+qi0}\{v_{m-\ell+i}|(v_{\ell-i},v_{m-\ell+i})+q(v_{m-\ell+i})=0,(v_{\ell-i},v_{m-\ell+i})+q_{i}\neq 0\}. When vi0v_{\ell-i}\neq 0, we have Hc(p1([v]))Hc(𝔾m)[2d+4]H^{\ast}_{c}(p^{-1}([v^{\prime}]))\cong H_{c}^{\ast}(\mathbb{G}_{m})[-2d+4]. When vi=0v_{\ell-i}=0, we have Hc(p1([v]))¯[2d2]H^{\ast}_{c}(p^{-1}([v^{\prime}]))\cong\overline{\mathbb{Q}}_{\ell}[-2d-2]. Since UiU_{i} can be identified with the subscheme of UiU_{i}^{\prime} with vi=0v_{\ell-i}=0, we have

χc(Ui+1,fiASψ)=χc(Ui,fiASψ)=χc(Ui,fi+1ASψ).\chi_{c}(U_{i+1},f^{\ast}_{\geq\ell-i}\mathrm{AS}_{\psi})=\chi_{c}(U_{i}^{\prime},f^{\ast}_{\geq\ell-i}\mathrm{AS}_{\psi})=\chi_{c}(U_{i},f^{\ast}_{\geq\ell-i+1}\mathrm{AS}_{\psi}).

Consider the projection p:Wi+1Wip^{\prime}:W_{i+1}\to W_{i}. We have

p!fiASψ=fi+1ASψp!fiASψ).p_{!}f_{\geq\ell-i}^{\ast}\mathrm{AS}_{\psi}=f^{\ast}_{\geq\ell-i+1}\mathrm{AS}_{\psi}\otimes p_{!}f_{\ell-i}^{\ast}\mathrm{AS}_{\psi}).

We decompose pp into two steps Wi+1p1Wip2WiW_{i+1}\xrightarrow{p_{1}}W_{i}^{\prime}\xrightarrow{p_{2}}W_{i}, where Wi(M0j=i+1m+iMj)W_{i}^{\prime}\subset\mathbb{P}(M_{0}\oplus_{j=\ell-i+1}^{m-\ell+i}M_{j}) is defined by (ϕi1ϕ0v0,vm+i)=0(\phi_{\ell-i-1}\cdots\phi_{0}v_{0},v_{m-\ell+i})=0. Fix [v′′]=[v0,vi+1,,vm+i][v^{\prime\prime}]=[v_{0},v_{\ell-i+1},\cdots,v_{m-\ell+i}], and let qi:=qi([v′′])q_{i}:=q_{i}([v^{\prime\prime}]). The fiber p1([v′′])={vi|(vi,vm+i)+q(vi)=0,(vi,vm+i)+qi0)}p^{\prime-1}([v^{\prime\prime}])=\{v_{\ell-i}|(v_{\ell-i},v_{m-\ell+i})+q(v_{\ell-i})=0,(v_{\ell-i},v_{m-\ell+i})+q_{i}\neq 0)\}. The function fi1f_{\ell-i-1} along the fiber is a linear function in viv_{\ell-i} by fi([v′′])=(ϕivi,vm+i)qif_{\ell-i}([v^{\prime\prime}])=\frac{(\phi_{\ell-i}v_{\ell-i},v_{m-\ell+i})}{q_{i}}. This the stalk of p1,!fiASψp_{1,!}f^{\ast}_{\geq\ell-i}\mathrm{AS}_{\psi} at [v′′][v^{\prime\prime}], which is Hc(p1([v′′]),fiASψ)H^{\ast}_{c}(p^{-1}([v^{\prime\prime}]),f^{\ast}_{\ell-i}\mathrm{AS}_{\psi}), vanishes when vm+i0v_{m-\ell+i}\neq 0. Thus the stalk of p2,!p1,!fiASψp_{2,!}p_{1,!}f_{i}^{\ast}\mathrm{AS}_{\psi} is isomorphic to ¯[2d2]\overline{\mathbb{Q}}_{\ell}[-2d-2]. Therefore, p!fi+1ASψp_{!}f_{\leq i+1}^{\ast}\mathrm{AS}_{\psi} and the constant sheaf ¯\overline{\mathbb{Q}_{\ell}} are the same in the Grothendieck group of Dcb(Wi)D_{c}^{b}(W_{i}). Thus we have

χc(Wi+1,fiASψ)=χc(Wi,fiASψp!¯ASψ)=χc(Wi,fiASψ).\chi_{c}(W_{i+1},f_{\geq\ell-i}^{\ast}\mathrm{AS}_{\psi})=\chi_{c}(W_{i},f^{\ast}_{\geq\ell-i}\mathrm{AS}_{\psi}p_{!}\overline{\mathbb{Q}_{\ell}}\mathrm{AS}_{\psi})=\chi_{c}(W_{i},f^{\ast}_{\geq\ell-i}\mathrm{AS}_{\psi}).

Combining equation 5.3.2, equation 5.3.3, and equation 5.3.4 we get

(5.3.5) (1)dimM2χc(X~,KlG^,𝐏mst(1,ϕ))\displaystyle(-1)^{\mathrm{dim}M-2}\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(1,\phi)) =χc(U1)+χc(W1)\displaystyle=-\chi_{c}(U_{1})+\chi_{c}(W_{1})

The proof of Theorem 5.2 then reduces to the calculation of the compactly supported Euler characteristic of U1U_{1} and W1W_{1}. We break these into a few lemmas listed below. Throughout the section,

  • QQ is the quadric on (M0+M)\mathbb{P}(M_{0}+M_{\ell}) defined by q0+q=0q_{0}+q_{\ell}=0;

  • Q1Q_{1} is the quadric on (M0+M)\mathbb{P}(M_{0}+M_{\ell}) defined by q0=0q_{0}=0, q=0q_{\ell}=0 (this is because the whole space here is Q(q)Q(q) and Q1Q(q)Q_{1}\subset Q(q));

  • W1W_{1} is defiend by ω1:=(ϕϕ0(v0),v)=0\omega_{1}:=(\phi_{\ell}\cdots\phi_{0}(v_{0}),v_{\ell})=0 and q0+q=0q_{0}+q_{\ell}=0;

  • The intersection W1,1:=W1Q1W_{1,1}:=W_{1}\cap Q_{1} defined by ω1,1:=ω1Q1\omega_{1,1}:=\omega_{1}\cap Q_{1}.

Lemma 5.4.

Suppose the root system is of type BnB_{n}. Then

χ(Q1)=2d.\chi(Q_{1})=2d.
Proof.

Without losing generality, we can assume dimM0=d\dim M_{0}=d, dimM=d+1\dim M_{\ell}=d+1 where dd is even. We break Q1Q_{1} into two parts, depending on whether the projection of a point (v0,v)Q1(v_{0},v_{\ell})\in Q_{1} to M0M_{0} is zero or not.

When v0=0v_{0}=0, we have v0.v_{\ell}\neq 0. Call this part Q1v0=0Q_{1}^{v_{0}=0}, it is just the quadric q=0q_{\ell}=0 in (M).\mathbb{P}(M_{\ell}). Therefore the Euler characteristic of Q1v0=0Q_{1}^{v_{0}=0} is dd. Here we do not have the primitive cohomology since the dimension of MM_{\ell} is odd, see [YunEpipe, Table 6.12].

When v00v_{0}\neq 0, we call this part Q1v00Q_{1}^{v_{0}\neq 0}. Similar to what we did previously, we can project Q1v00Q_{1}^{v_{0}\neq 0} onto the quadric q0=0q_{0}=0 in (M0)\mathbb{P}(M_{0}) by sending [(v0,v)][(v_{0},v_{\ell})] to [v0][v_{0}]. The fiber is a cone, so its Euler characteristic is 11. The Euler characteristic of Q1v00Q_{1}^{v_{0}\neq 0} is just the Euler characteristic of a quadric in P(M0)P(M_{0}) which is dd since dd is even. The total Euler characteristic is the sum of the Euler characteristics of the two parts. ∎

Lemma 5.5.

Suppose the root system is of type BnB_{n}. Then

χc(W1,1)=2d.\chi_{c}(W_{1,1})=2d.
Proof.

The proof is similar to 5.4. We split W1,1W_{1,1} into two (disjoint) parts

W1,1=W1,1v0=0+W1,1v00W_{1,1}=W_{1,1}^{v_{0}=0}+W_{1,1}^{v_{0}\neq 0}

so that the Euler characteristic is the sum of the Euler characteristics of the two parts.

When v0=0v_{0}=0, we have v0.v_{\ell}\neq 0. It is the quadric q=0q_{\ell}=0 in (M).\mathbb{P}(M_{\ell}). Therefore W1,1v0=0=dW_{1,1}^{v_{0}=0}=d.

When v00v_{0}\neq 0, project Q1v00Q_{1}^{v_{0}\neq 0} onto the quadric q0=0q_{0}=0 in (M0)\mathbb{P}(M_{0}) by sending [v0,v][v_{0},v_{\ell}] to [v0][v_{0}]. The fiber is again a cone, so its Euler characteristic is 11. The Euler characteristic of Q1v00Q_{1}^{v_{0}\neq 0} is just the Euler characteristic of a quadric in P(M0)P(M_{0}) which is dd. ∎

Remark 5.6.

Computations of the Euler characteristic of Q1Q_{1} and ω1,1\omega_{1,1} of type DnD_{n} and Dn2{}^{2}D_{n} follow the same path. The reader should be aware that the different parity of dimM0M\dim M_{0}\oplus M_{\ell}, dimM0\dim M_{0}, and dimM\dim M_{\ell} lead to differences in the Euler characteristics, see [YunEpipe, Table 6.12]. We will omit the proof of Type DnD_{n} and Dn2{}^{2}D_{n} and write down χc(Q1)\chi_{c}(Q_{1}) and χc(W1,1)\chi_{c}(W_{1,1}) directly from the table by Yun and the previous two lemmas. We make a summarization in the following table. In the last but two and three columns, we list the Euler Characteristic of W1W_{1} and W1,1W_{1,1}.

Table 1. Dimension, parity and Euler characteristics
Type dimM0M\dim M_{0}\oplus M_{\ell} Parity of dimM0M\dim M_{0}\oplus M_{\ell} dimM0\dim M_{0} Parity of dimM0\dim M_{0} dimM\dim M_{\ell} Parity of dimM\dim M_{\ell} χc(W1)\chi_{c}(W_{1}) χc(W1,1)\chi_{c}(W_{1,1}) χc(KlG^,𝐏mst)-\chi_{c}(\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}})
BnB_{n} 2d+1 odd d even d+1 odd 4d 2d 2d
2d+1 odd d+1 even d odd 4d 2d 2d
DnD_{n} 2d even d even d even 0 2d 2d
2d+2 even d+1 even d+1 even 0 2(d+1) 2(d+1)
Dn2{}^{2}D_{n} 2d even d odd d odd 0 2d-2 2d
2d+2 even d+1 odd d+1 odd 0 2d 2(d+1)
Proof of Theorem 5.2.

We compute the Euler characteristic χc(X~,KlG^,𝐏mst(χ,ϕ))-\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi)) when the root system is of type BnB_{n} with even d=dimM0d=\dim M_{0}, as an example to illustrate the usage of Table 1 in the proof of Theorem 5.2, and note that the computations of the Euler characteristic when the root system is of type DnD_{n} or Dn2{}^{2}D_{n} follows the same path.

When dimM0dimMell=2d+1\dim M_{0}\oplus\dim M_{e}ll=2d+1, QQ is a smooth quadric and W1W_{1} is either the intersection of two smooth quadrics or with singularities contains a factor isomorphic to kk^{*}, both are in (M0M)\mathbb{P}(M_{0}\oplus M_{\ell}). The table in [YunEpipe, Sec. 6] plus the fact that χc(k)=0\chi_{c}(k^{*})=0 implies

χc(Q)=2d, χc(W1)=4d.\chi_{c}(Q)=2d,\text{ }\chi_{c}(W_{1})=4d.

Therefore we have

χc(X~,KlG^,𝐏mst(χ,ϕ))=χc(Q)+χc(Q1)+χc(W1)χc(W1,1)=2d.-\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi))=-\chi_{c}(Q)+\chi_{c}(Q_{1})+\chi_{c}(W_{1})-\chi_{c}(W_{1,1})=2d.

In the case of type DnD_{n} and Dn2{}^{2}D_{n}, dimM\dim M (hence dimM2\dim M-2) is always even. So (5.3.1) implies that

χc(X~,KlG^,𝐏mst(χ,ϕ))=χc(U1)χc(W1)=χc(Q)χc(Q1)χc(W1)+χc(W1,1).-\chi_{c}(\widetilde{X}^{\circ},\mathrm{Kl}^{\operatorname{st}}_{\widehat{G},\mathbf{P}_{m}}(\chi,\phi))=\chi_{c}(U_{1})-\chi_{c}(W_{1})=\chi_{c}(Q)-\chi_{c}(Q_{1})-\chi_{c}(W_{1})+\chi_{c}(W_{1,1}).

Therefore one has to be careful with the signs of each term when calculating the Euler characteristic. ∎

References