Abstract
In this paper, we formulate conjectural formulas for the arithmetic intersection numbers of special cycles on unitary Shimura varieties with minuscule parahoric level structure. Also, we prove that these conjectures are compatible with all known results.
Similar content being viewed by others
References
Cho, S.: The basic locus of the unitary Shimura variety with parahoric level structure, and special cycles (2018) preprint
Hironaka, Y.: Local zeta functions on Hermitian forms and its application to local densities. J. Number Theory 71, 40–64 (1998)
Hironaka, Y.: Classification of Hermitian forms by the Iwahori subgroup and their local densities. Comment. Math. Univ. St. Paul. 49, 105–142 (2000)
Howard, B.: Linear invariance of intersections on unitary Rapoport-Zink spaces. Forum Math. 31, 1265–1281 (2019)
Kudla, S.: Central derivatives of Eisenstein series and height pairings. Ann. Math. 146, 545–646 (1997)
Kudla, S., Rapoport, M.: Height pairings on Shimura curves and p-adic uniformization. Invent. Math. 142, 153–222 (2000)
Kudla, S., Rapoport, M.: Special cycles on unitary Shimura varieties I. Unramified local theory. Invent. Math. 184, 629–682 (2011)
Kudla, S., Rapoport, M.: Special cycles on unitary Shimura varieties II. Global theory. J. Reine Angew. Math. 697, 91–157 (2014)
Kudla, S., Rapoport, M., Yang, T.: Modular Forms and Special Cycles on Shimura Curves, Annals of Mathematics Studies, vol. 161. Princeton University Press, Princeton (2006)
Li, C., Zhang, W.: Kudla–Rapoport cycles and derivatives of local densities (2019) preprint
Liu, Y.: Arithmetic theta lifting and L-derivatives for unitary groups I. Algebra Number Theory 5, 849–921 (2011)
Liu, Y.: Arithmetic theta lifting and L-derivatives for unitary groups II. Algebra Number Theory 5, 923–1000 (2011)
Rapoport, M., Smithling, B., Zhang, W.: On the arithmetic transfer conjecture for exotic smooth formal moduli spaces. Duke Math. J. 166, 2183–2336 (2017)
Rapoport, M., Smithling, B., Zhang, W.: Regular formal moduli spaces and arithmetic transfer conjecture. Math. Ann. 370, 1079–1175 (2018)
Rapoport, M., Smithling, B., Zhang, W.: Arithmetic diagonal cycles on unitary Shimura varieties (2018) preprint
Rapoport, M., Zink, Th.: Period Spaces for p-divisible Groups, Annals of Mathematics Studies, vol. 141. Princeton University Press, Princeton (1996)
Sankaran, S.: Unitary cycles on Shimura curves and the Shimura lift I. Doc. Math. 18, 1403–1464 (2013)
Sankaran, S.: Improper intersections of Kudla–Rapoport divisors and Eisenstein series. J. Inst. Math. Jussieu 16, 899–945 (2017)
Vollaard, I., Wedhorn, T.: The supersingular locus of the Shimura variety for GU(1, s) II. Invent. Math. 184, 591–627 (2011)
Acknowledgements
I would like to thank Yifeng Liu for many helpful discussions. In particular, the discussions with him were very helpful to formulate our conjectures. I would like to thank Zhiyu Zhang for helpful comments. I would like to thank the anonymous referees for careful reading and helpful suggestions.
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Wei Zhang.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix A: Arithmetic intersection numbers of special cycles \({{\mathcal {Z}}}(x_1), \dots , {{\mathcal {Z}}}(x_{n+1})\) in \({{\mathcal {N}}}^1(1,n)\)
Appendix A: Arithmetic intersection numbers of special cycles \({{\mathcal {Z}}}(x_1), \dots , {{\mathcal {Z}}}(x_{n+1})\) in \({{\mathcal {N}}}^1(1,n)\)
In this appendix, we will prove that our conjectures are compatible with [10, Theorem 10.4.4]. This theorem relates the intersection number of \({{\mathcal {Z}}}(x_1), \dots , {{\mathcal {Z}}}(x_{n+1})\) in \({{\mathcal {N}}}^1(1,n)\) to representation densities. This intersection number can be regarded as the intersection number of
in \({{\mathcal {N}}}^n(1,2n-1)\), where
and \(B_1 \in X_{n+1}(E_v)\).
Therefore, the Conjecture 3.25 should be compatible with [10, Theorem 10.4.4]. To prove this compatibility, we will use the notation in Sect. 3.2.
Let L be an \(O_{E_v}\)-lattice with basis \(\lbrace u_1, \dots , u_{2n} \rbrace \) such that the matrix of inner product \((\langle u_i, u_j \rangle )\) is \(\pi B\). Let \(L_1\) be the \(O_{E_v}\)-lattice of rank \(n+1\) with basis \(\lbrace u_1, \dots , u_{n+1} \rbrace \), \(L_2\) be the \(O_{E_v}\)-lattice of rank \(n-1\) with basis \(\lbrace u_{n+1}, \dots , u_{2n} \rbrace \). Note that \(L_2\) is unimodular with respect to \(\langle \cdot , \cdot \rangle \).
Let M be an \(O_{E_v}\)-lattice with basis \(\lbrace v_1, \dots , v_m \rbrace \) (\(m=2n+2r\)) such that the matrix of inner product \((\langle v_i,v_j \rangle )\) is \(\pi A^{[r]}_{n}.\)
For \(0 \le i \le n-1\), let \(M_i\) be an \(O_{E_v}\)-lattice with basis \(\lbrace v_{1,i}, \dots , v_{2n,i} \rbrace \) such that the matrix of inner product \((\langle v_{k,i},v_{l,i} \rangle )\) is \(\pi A_{i}\).
We define the sets
and for \(0 \le i \le n-1\),
Then we have
and
for sufficiently large d.
Note that L has the rank \(n-1\) unimodular sublattice \(L_2\). Also, any unimodular sublattice in \(M_i\) has rank \(\le i\). Therefore, if \(i <n-1\), there is no \(\varphi \in {{\,\mathrm{Hom}\,}}_{O_{E_v}}(L,M_i/\pi ^dM_i)\) such that \(\langle \varphi (x), \varphi (y) \rangle \equiv \langle x,y \rangle {{\,\mathrm{mod}\,}}\pi ^d\) for sufficiently large d. This implies that \(J_d(L,M_i)\) is an empty set for \(i <n-1\). Therefore, \(W_{n-1,i}(B,0)=0\) for \(i <n-1\).
This implies that we only need to compute the constant \(\beta _{n-1}^{n-1}\) in Conjecture 3.25. We have the following proposition.
Proposition A.1
\(\beta _{n-1}^{n-1}=\dfrac{(1-(-q)^n)}{(-q)^{3n+1}(1-(-q))(1-(-q)^{-(n+1)})}\).
Proof
By the proof of theorem 3.16, we have
where \(0 \le h \le 2n\) (in this proposition, we only need \(h=n-1\), but we will use this general notation for later use). Therefore, we need to find the inverse matrix of \((-q)^{2n(2n-h)}{{\mathfrak {B}}}\). Note that
and
Therefore, we have
We define
and
Also, we define a \(2n+1\) \(\times \) \(2n+1\) matrix \({{\mathfrak {a}}}_h={{\,\mathrm{diag}\,}}(\alpha _{1,h}, \dots , \alpha _{2n+1,h})\) and a Vandermonde matrix
Then, we have \((-q)^{2n(2n-h)}{{\mathfrak {B}}}={{\mathfrak {X}}}{{\mathfrak {a}}}_h\)
Therefore, we have
Let \({{\mathfrak {X}}}^{-1}=(y_{ij})_{1 \le i \le 2n+1}\). It is well known that the (i, j)-th entry of the inverse matrix of a Vandermonde matrix is
We can think \(y_{ij}\) as the \(z^{2n+1-j}\)-coefficient of
Since we have
Note that \((1-x_{2n+1})=0\). Therefore we have,
Since \(x_n=(-q)\) we have
and
Combining these two and \(\alpha _{n,n-1}=(-q)^{n+1}\), we have
\(\square \)
Proposition A.2
(cf. [7, Proposition 9.9] and Proposition 3.20)
-
(1)
Let \(\lbrace O_j \rbrace \) be a set of representatives for the \(U(M_i)\)-orbits in the set of all sublattices O of \(M_i\) such that O is isometric to \(L_2\). Then we have
$$\begin{aligned} \begin{array}{ll} \vert J_d^1(L,M_i) \vert =&{}\sum _{j}\vert I_d(L_2,M_i;O_j) \vert \\ &{}\times \vert \lbrace \varphi _1 \in J_d^1(L_1,M_i) \vert \langle \varphi (L_1), O_j \rangle \equiv 0 ({{\,\mathrm{mod}\,}}\pi ^d) \rbrace . \end{array} \end{aligned}$$ -
(2)
Let \(\lbrace N_j \rbrace \) be a set of representatives for the U(M)-orbits in the set of all sublattices N of M such that N is isometric to \(L_2\). Then we have
$$\begin{aligned} \begin{array}{ll} \vert J_d^1(L,M) \vert =&{}\sum _{j}\vert I_d(L_2,M;N_j) \vert \\ &{}\times \vert \lbrace \varphi _1 \in J_d^1(L_1,M) \vert \langle \varphi (L_1), N_j \rangle \equiv 0 ({{\,\mathrm{mod}\,}}\pi ^d) \rbrace . \end{array} \end{aligned}$$
Proof
The proof of this proposition is identical to the proof of [7, Proposition 9.9] (and Proposition 3.20). \(\square \)
From now on, we fix \(i=n-1\). Note that \(L_2\) is unimodular. Therefore, by the arguments in [7, p.680], all sublattices \(O \subset M_{n-1}\) such that O is isometric to \(L_2\) are in the same \(U(M_{n-1})\)-orbit, and \(M_{n-1}=O \perp O^{\perp }\) where \(O^{\perp }:=(E_v O)^{\perp } \cap M_{n-1}\). Let us fix O the sublattice of \(M_{n-1}\) with the basis \(\lbrace v_{n+2,n-1}, \dots , v_{2n,n-1} \rbrace \) as a representative of the unique \(U(M_{n-1})\)-orbit of \(L_2\). Then, \(O^{\perp }\) is the lattice with basis \(\lbrace v_{1,n-1}, \dots , v_{n+1,n-1} \rbrace \). In particular, \(O^{\perp } \subset \pi M_{n-1}^{\vee }\).
Also, let us fix N the sublattice of M with the basis \(\lbrace v_{n+2}, \dots , v_{2n} \rbrace \). Then N is a representative of the unique U(M)-orbit of \(L_2\) and \(M=N \perp N^{\perp }\). We have the following analogue of [7, Lemma 9.10] (and Lemma 3.21).
Lemma A.3
-
(1)
For sufficiently large d, we have
$$\begin{aligned} \begin{array}{l} \vert \lbrace \varphi _1 \in J_d^1(L_1,M_{n-1}) \vert \langle \varphi _1(L_1),O \rangle \equiv 0 \text { }({{\,\mathrm{mod}\,}}\pi ^d) \rbrace \vert =\vert I_d(L_1,O^{\perp })\vert \end{array}. \end{aligned}$$ -
(2)
For sufficiently large d, we have
$$\begin{aligned}&\vert \lbrace \varphi _1 \in J_d^1(L_1,M) \vert \langle \varphi _1(L_1),N \rangle \equiv 0 \text { }({{\,\mathrm{mod}\,}}\pi ^d) \rbrace \vert \nonumber \\&\quad =q^{-2(n+1)}\vert I_d(L_1,N^{\perp } \cap \pi M^{\vee })\vert . \end{aligned}$$(A.0.2)
Proof
The proof of this lemma is almost identical to the proof of [7, Lemma 9.10] and Lemma 3.21. Therefore, we will only prove (2). First note that
for sufficiently large d such that \(\pi ^d N^{\vee } \subset N\).
Therefore, we have
Since \(\pi ^{d}N^{\vee } \subset \pi M^{\vee }\) for sufficiently large d, we can write (A.0.2) as
Now, fix a positive integer a such that \(\pi ^a N^{\vee } \subset N\).
Then we have
by replacing M by \(\pi ^a N^{\vee } \perp (\pi M^{\vee } \cap N^{\perp })\).
Now, we can write \(\varphi _1=\psi _1 + \psi _2\) where
Since we assume that d is sufficiently large such that \(\pi ^d N^{\vee } \subset N\), we have
This means that the condition \(\langle \varphi _1(x),\varphi _1(y) \rangle \equiv \langle x,y \rangle \text { }({{\,\mathrm{mod}\,}}\pi ^d)\) is the same as the condition \(\langle \psi _2(x),\psi _2(y) \rangle \equiv \langle x,y \rangle \text { } ({{\,\mathrm{mod}\,}}\pi ^d)\), and we do not need to impose any condition on \(\psi _1\).
Therefore, (A.0.2) is equal to
Here, we used the fact that \(\vert M:N \perp (\pi M^{\vee } \cap N^{\perp })\vert =q^2\) and \(N=N^{\vee }\) (since N is unimodular).
\(\square \)
Proposition A.2 and Lemma A.3 imply that
Note that
for sufficiently large d. Here \(\Delta \) is the lattice defined in the proof of the theorem 3.19.
Also, we have
for sufficiently large d.
Now, we need to compute these representation densities. We have the following proposition.
Proposition A.4
Let k, m, n be integers such that \(k,n \le m\) and let \(r=m-n\). Then we have
This formula is just a special case of [2, Theorem II]. In the proof of this proposition, we will use the notation in [2] (and [18, Page 921]).
Let \(\Lambda _n^+:=\lbrace \lambda =(\lambda _1, \dots , \lambda _n) \in {{\mathbb {Z}}}^n \vert \lambda _1 \ge \lambda _2 \ge \dots \ge \lambda _n \ge 0 \rbrace \).
For each \(\mu \in (\mu _1, \dots , \mu _n) \in \Lambda _n^+\), we define
Furthermore, for \(\mu \in \Lambda _n^+\), \(\xi \in \Lambda _m^+\), we define
For \(\mu , \lambda \in \Lambda _n^+\), we denote by \(\mu \le \lambda \) if \(\mu _i \le \lambda _i\) for all \(1 \le i \le n\).
For \(u \ge v \ge 0\), we define
Finally, for \(\mu , \lambda \in \Lambda _n^+\), we define
Now, we can state the following proposition.
Proposition A.5
([2, Theorem II]) For \(\lambda \in \Lambda _n^+\) and \(\xi \in \Lambda _m^+\) with \(m \ge n\), we have
Proof of Proposition A.4
In our case, \(\lambda =(0,0, \dots , 0) \in \Lambda _n^+\) and \(\xi =(1^k,0^{m-k}) \in \Lambda _m^+\). Therefore, \({\tilde{\lambda }}=(1^n) \in \Lambda _n^+\) and hence \(\mu \) runs over \((1^l,0^{n-l}) \in \Lambda _n^+\). We write \(^l\mu \) for \((1^l,0^{n-l})\).
First, we need to compute \(\prod _{j \ge 1} I_j(^l\mu ,\lambda )\).
For \(j=1\), we have
Since \(({\tilde{\lambda }})'_2=0\), i should be 0. Therefore, we have
For \(j>1\), it is easy to see that \(I_j(^l\mu ,\lambda )=1\).
Therefore,
Since \(\vert ^l\mu \vert =l\), \(n( ^l\mu )=\dfrac{l(l-1)}{2}\), and \(\langle \xi ', ^l\mu ' \rangle =kl\), we have
If \(k=0\), it is well-known that
where \(X=(-q)^{-r}\) (see [7, page 677]).
Therefore, by changing X to \((-q)^kX\), we have that (A.0.6) is equal to
\(\square \)
Now, we have the following proposition.
Proposition A.6
Proof
From (3.2.3), (A.0.3), (A.0.4), (A.0.5), we have
Note that
and hence \({{\,\mathrm{val}\,}}(\det (B_1))\equiv 0 ({{\,\mathrm{mod}\,}}2)\). Therefore \(\alpha (\left( \begin{array}{c@{\quad }c} \pi 1_n &{} \\ &{} \pi ^2 \end{array}\right) , \pi B_1)=0\).
This implies that
By Proposition A.4, we have
Therefore,
Here, we used the fact that \(\alpha (\pi C, \pi D)=q^{n^2}\alpha (C,D)\) where D is a \(n \times n\) hermitian matrix.
Similarly,
Here, we used Proposition A.1 and
\(\square \)
Note that \(\dfrac{\alpha \bigg (\left( \begin{array}{c@{\quad }c} 1_n &{} \\ &{} \pi \end{array}\right) ,B_1\bigg )}{\alpha (1_n,1_n)}\) is equal to \(\partial \text {Den}_{\Lambda }(L)\) and \(\dfrac{\alpha (1_{n+1},B_1)}{\alpha (1_{n+1},1_{n+1})}\) is equal to \(\text {Den}(L)\) in [10, Theorem 10.4.4] (but, our n is \(n-1\) in loc. cit.). Therefore, Conjecture 3.25 is compatible with their result.
Rights and permissions
About this article
Cite this article
Cho, S. Special cycles on unitary Shimura varieties with minuscule parahoric level structure. Math. Ann. 384, 1747–1813 (2022). https://doi.org/10.1007/s00208-021-02343-y
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00208-021-02343-y