Absolute discriminant of the Fermat linear-cycle lattice

edit · history · discussion · files · long url · integer number theory
Numbers
$n$
$d$ 
$D_{n,d}$
2
3:
1
2
4:
64
2
5:
244140625
2
6:
3454279995636458717184
2
7:
36703368217294125441230211032033660188801
2
8:
431359146674410236714672241392314090778194310760649159697657763987456
2
9:
11433811272836884826665874049685357613602127080237382571153151471568389249023393608050222706416077770721
2
10:
15625000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000
2
11:
88592213009575201230959050581819203496673983850936311247456617209331422390098646799274067240224511892912343600256063311645277101292135934503249769447807774230858090529333267215281318750801671448715521
2
12:
470334193977228364182039705989729385036680927844521642182127731523820636777177081180879381361576312916392641625312929088402114797948502850853289537287836438668555534641141148861968727206245304867509557997217777335462781990494735273361293665074907616804447774048256
2
13:
15240727288217481097169688317555229380061886407509143835918755533810868338102949921365781625200196777115677757738931891442055818688224083255364665144842754930572721401780637025735352407911082739515663622750278129367018845556889930590117427957681737928048742606808680064227380113254754424438292390026122481050902313200443614663022782001
2
14:
4749863008467037658161426810949450102916838010285729719906280552459419822052504800566714372980091454911086800492333473302166498475530964733056865121577946373569842708216083039490184252890202436319138290851235611373802032659619283559739274517289539585412493652087568097173450404116454361009290373577433005588011382712660082322284590749419179669465808391564747736206934856535496193651711046487247236647968822960716775424
4
3:
27
4
4:
85070591730234615865843651857942052864
4
5:
61363668316221581913697149437328445537190193906243119563894363964838815684021448205751223550466649403464794644564148481414598982630308986285093935301362043153163818305748274106736772637304966337978839874267578125
Definition
For an integral basis $(e_i)$ of the Fermat linear-cycle lattice $L_{n,d}$ in (5), $D_{n,d}=|\det(\langle e_i,e_j\rangle)|$.
Parameters
$n$
—   complex dimension ($n\geq2$ is even)
$d$
—   defining degree ($d\geq3$)
Formulas
(1)
The number of generating planes is $(n+1)!!\,d^{n/2+1}$, where $(n+1)!!=1\cdot3\cdots(n+1)$ ([1], equation (2)).
(2)
For two generating planes $P,Q$ with projective intersection dimension $r$, $\displaystyle P\cdot Q=\frac{1-(1-d)^{r+1}}{d}$, including $r=-1$ for an empty intersection and $r=n/2$ for $P=Q$ ([1], equation (6)).
(3)
If $s_1,\ldots,s_\rho$ are the nonzero Smith invariant factors of the full generating Gram matrix, then $D_{n,d}=\prod_{j=1}^{\rho}s_j$. Equivalently, if $G=CKC^{\mathsf T}$ with $C$ integral and its rows spanning $\mathbb Z^\rho$, then $D_{n,d}=|\det K|$ ([1], Section 2.2).
Comments
(4)
Let $X_{n,d}:X_1^d+\cdots+X_{n+1}^d=X_0^d\subset\mathbb P^{n+1}_{\mathbb Q}$, with $n\geq2$ even and $d\geq3$. The pairing is the topological intersection form on $H_n(X_{n,d}(\mathbb C),\mathbb Z)$, as in [1], Sections 2.2 and 3.1; every plane has its complex orientation.
(5)
On $X_{n,d}$ from (4), put $\xi_d=\exp(\pi i/d)$ in the standard complex embedding, and use the all-plus model $\sum_{j=0}^{n+1}Y_j^d=0$. For each perfect matching of $0,\ldots,n+1$, orient each pair increasingly and order the pairs by their first element. For every tuple $c_0,\ldots,c_{n/2}\in\{0,\ldots,d-1\}$ take the plane with equations $Y_{b_{2j}}=\xi_d^{1+2c_j}Y_{b_{2j+1}}$, $0\leq j\leq n/2$. Transport its complex-oriented cycle to $X_{n,d}$ by $Y_0=\xi_dX_0$, $Y_j=X_j$ for $j>0$. This is the coordinate form of [1], equation (2).
(6)
The lattice is the integral span of these planes, including the hyperplane-section class, which is a sum of $d$ planes. It is neither the lattice generated only by primitive differences nor a saturation in middle homology or in the full integral Hodge lattice. The matrix on all plane generators is redundant; its zero determinant is not $D_{n,d}$.
(7)
The absolute-discriminant convention is positive. [1], Table 1, lists elementary divisors with leading signs; only the product of its positive elementary divisors is used here. Several printed leading signs differ from the signed determinants of the intersection lattices, for example $(n,d)=(2,6)$. Exact signed determinants and signatures are retained in the attached certificates; the elementary divisors and their positive products agree.
Programs
(P1)
Sage
# Put attached lattice.py and modular.py beside this script; use SageMath with numberdb.
import numberdb.sage as numberdb
from lattice import gram, quotient

def absolute_discriminant(n, d):
    if n < 2 or n % 2 or d < 3:
        raise ValueError("need even n>=2 and integer d>=3")
    planes, G = gram(n, d)
    J, H, C, K = quotient(G)
    return abs(K.det())

absolute_discriminant(2, 4)  # 64
# Default stored selection: every (2,d) with 3<=d<=14,
# and every (4,d) with 3<=d<=5. Larger inputs can be expensive.
References
[1]
E. Aljovin, H. Movasati and R. Villaflor Loyola, Integral Hodge conjecture for Fermat varieties, arXiv version 2 (27 March 2019), Theorem 1, Table 1, Sections 2.2, 3.1 and 3.3. (arXiv)
Links
Similar tables
Pham periods of Fermat varieties —   gives integrals on explicitly marked Pham cycles of the same surface and fourfold models; the present invariant uses the integral intersection lattice generated by complex-oriented linear algebraic cycles
Data properties
Entries are of type: integer
Table is complete: no (it holds every surface degree $3\leq d\leq14$ and every fourfold degree $3\leq d\leq5$, 15 pairs from the published reference range in [1], Table 1. These noncanonical bounds include composite surface degrees and broaden to fourfolds; the sextic fourfold is deferred because of the substantially higher exact-lattice cost documented in Section 4. Higher dimensions and degrees remain uncomputed)
How they were obtained:

The attached generator multiplies the exact elementary divisors transcribed from [1], Table 1. All 15 lattices were also recomputed from their complete plane intersections with exact integer and rational arithmetic. Integral quotient certificates record $G=CKC^{\mathsf T}$, with the rows of $C$ spanning $\mathbb Z^\rho$; this excludes a proper finite-index sublattice from a merely rational selection of planes.

more

A separate checker uses cyclotomic equation-matrix ranks for every relative matching and phase configuration, verifies the complete Gram matrices and integral quotient certificates, and compares every invariant factor with the source. Large integer Smith reductions are replaced by exact Schur complements modulo prime powers, with checked unit-block inverses and valuation sums matching the exact determinant; this method is controlled against native Smith results. Exact characteristic polynomials check the signatures. Known-answer and deliberately invalid controls are retained. The certificates concern the plane span; no independent integral embedding into the full Hodge lattice or complete plane-period vectors in Pham coordinates are asserted.