Secants, Socles and Stability

Aaron Bertram · 2025

Read Secants, Socles and Stability by Aaron Bertram (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-14445.txt.

Introduction

Aaron Bertram and Brooke Ullery published a mathematics paper in 2025 linking three topics: secant varieties, socles, and stability. Symmetric tensors can be viewed in a second way, as a space of special graded rings whose top piece sits in a fixed degree. That swap lets tools from one field answer questions in the other.

The authors set up two stability tests, one for even degree and one for odd degree. These tests sort the whole space by the type of a standard filtration. They work the sorting out fully for low-degree cases in the plane. They then compare the result with the known tables that record how the rings are built from simpler pieces.

Catalog record

CREATORAaron Bertram
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE52 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-14445

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Secants, Socles and Stability Aaron Bertram Abstract 1. Introduction Let k be an algebraically closed field of characteristic zero and: S_{\bullet}=k[x_{0},...,x_{n}]_{\bullet} be the polynomial ring, graded by degree. The geometry of the secants to: {\mathbb{P}}(S_{1})=:{\mathbb{P}}^{n}\hookrightarrow{\mathbb{P}}^{n}_{d}:={\mathbb{P}}(S_{d}) (the “ d -uple embeddings”) is a polynomial version of the Waring problem for natural numbers [ War ] . In this context, one is interested in the secant loci in {\mathbb{P}}^{n}_{d} swept out by m -planes spanned by m+1 points of {\mathbb{P}}^{n} (possibly including limits), and one wants to know which points in such a secant locus “remember” the m+1 points. Alternatively, one can view {\mathbb{P}}(S_{d}) as the parameter space for degree d socles \sigma of Gorenstein graded S -modules R_{\bullet} . From that point of view, there are a pair of important numerical invariants associated to a point \sigma\in{\mathbb{P}}(S_{d}) , namely, (i) the palindromic Hilbert function of R_{\bullet} , and (ii) the (also palindromic) betti table of the minimal free resolution of R_{\bullet} . The second refines the first, and the minimal free resolution is also filtered by “linear strands.” Some guidance is needed, however, in finding the right coarsening of this overly fine filtration to explain, for example, why the classical variety of secant lines to the Veronese (two-uple) embedding of {\mathbb{P}}^{n} is defective (of smaller than the expected dimension) while the varieties of secant lines to the three-uple (and higher) embeddings are non-defective. In this paper, we appeal for such guidance to the canonical Harder-Narasimhan filtrations with respect to two natural stability conditions on the bounded derived category {\mathcal{D}}^{b}({\mathbb{P}}^{n}) of coherent sheaves on {\mathbb{P}}^{n} . This involves a third interpretation of the projective space {\mathbb{P}}^{n}_{d} via the vector space: \sigma\in\mbox{Hom}_{{\mathbb{P}}^{n}}({\mathcal{O}}_{{\mathbb{P}}^{n}},\omega_{{\mathbb{P}}^{n}}(-d)[n]) in the derived category, recovering the interior of the minimal free resolution of R_{\bullet} via the cone over the morphism. From this point of view, the uniqueness of the “spanner” of a secant reduces to the uniqueness of Harder-Narasimhan filtrations, and the challenge is to see how good the stability condition is at distinguishing betti tables from one another. This approach generalizes, in principle, to other varieties embedded by complete linear series. We will not discuss this here, but our treatment for {\mathbb{P}}^{n} mirrors the use of classical Mumford (in)stability to distinguish secant varieties to embedded curves, as explained to each of the authors by their advisor, Robert Lazarsfeld. [ 3 , 4 , 9 ] 2. Gorenstein Quotient Rings. Definition 2.1. A finite-dimensional graded quotient ring: S_{\bullet}\rightarrow R_{\bullet}=S_{\bullet}/I_{\bullet} is Gorenstein with socle in degree d if: (a) \dim_{k}R_{d}=1 (b) for all e\in{\mathbb{Z}} , the multiplication pairing: R_{e}\times R_{d-e}\rightarrow R_{d} is perfect, inducing vector space isomorphisms R_{d-e}\cong R_{e}^{\vee} . Note. The “socle” I_{d}=\ker(S_{d}\rightarrow R_{d})\in(S_{d})^{\vee} determines I_{e} for all e via: I_{e}=\ker(S_{e}\rightarrow S_{d-e}^{\vee})\ \mbox{for the map}\ (F,G)\mapsto F\cdot G+I_{d}\in R_{d} and I_{e}=S_{e} if and only if e is outside the range [0,d] . Example 2.1. For socles \sigma=I_{d} for which \dim(R_{1})=1 , I_{\bullet}=\langle I_{1}\cdot S_{\bullet},G\rangle\ \mbox{for any}\ G\in S_{d+1}-I_{1}\cdot S_{d} and \dim(R_{\bullet})=(1,1,....,1) . These socles therefore correspond to I_{1}\in(S_{1})^{\vee} . Passing to the dual, we may in general let: V=(S_{1})^{\vee} so that {\mathbb{P}}^{n} is the projective space of lines in V , and then (S_{d})^{\vee}\ \mbox{is the space of symmmetric $d$-tensors}\ \mbox{Sym}^{d}V\subset V^{\otimes d} and the d -uple embedding is the map {\mathbb{P}}^{n}\rightarrow{\mathbb{P}}^{n}_{d}\ \mbox{given by}\ v\mapsto v^{d} . Moreover: \bullet The image of the d -uple embedding is the locus of socles with \dim_{k}(R_{1})=1 . The Waring problem asks when a symmetric d tensor t has the form: t=\sum_{0}^{m}c_{i}v_{i}^{d} for distinct vectors v_{i}\in V . This is exactly the problem of determining whether a socle \sigma=t is in the locus of secant spaces to m+1 points of the d -uple embedding. Example 2.2. When d=2 , there is an identification: (S_{2})^{\vee}=\{q:S_{1}=V^{\vee}\rightarrow V=(S_{1})^{\vee}\ |\ q=q^{T}\} with the symmetric matrices, and the indeterminate middle dimension: \dim(R_{\bullet})=(1,\ r,\ 1) is the rank of the matrix q associated to \sigma since I_{1}=\ker(q) . The Waring problem in this context translates into the fact that a symmetric matrix of rank r is a always a sum of r distinct symmetric matrices of rank one, but of course the choice of the r matrices is not unique as, for example, v_{0}^{2}+v_{1}^{2}=\frac{1}{2}(v_{0}+v_{1})^{2}+\frac{1}{2}(v_{0}-v_{1})^{2} Example 2.3. When n=1 and d is arbitrary, the Waring problem has a very satisfactory answer. In that case, the Hilbert function of R_{\bullet} is a trapezoid: \dim(R_{e})=\left\{\begin{array}[]{l}e+1\ \mbox{for}\ 0\leq e\leq m-1\\ \\ m\ \mbox{for}\ m-1\leq e\leq d-(m-1)\\ \\ d-e+1\ \mbox{for}\ d-(m-1)\leq e\leq d\end{array}\right. and Waring’s problem (modified to allows for multiplicites) has the nicest possible result when 2m<d . Namely, each linear combination t=\sum_{i=0}^{m}c_{i}v_{i}^{d} of distinct rank one tensors uniquely determines the vectors v_{0},....,v_{m} (up to scalar multiples). Thus, for example, when n=1 and d=3 and m=1 , each sum: v_{0}^{3}+v_{1}^{3} has no other expression as a sum of two cubes. Note. The span of a non-reduced subscheme Z\subset{\mathbb{P}}^{1} of length m+1 may be absorbed into the Waring problem. When d=3 and m=1 , the span of double points is the union of the tangent lines to the twisted cubic in {\mathbb{P}}^{3} , and is disjoint from the variety of secant lines to pairs of points (minus the points themselves). The “right” way to describe the n=1 case is to say that when 2m\leq d , (1) each linear span of a subscheme Z\subset{\mathbb{P}}^{1} of length m+1 is a {\mathbb{P}}^{m}\subset{\mathbb{P}}^{d} . (2) the strata X_{m}-X_{m-1} of spans of length m+1 subschemes are locally closed. (3) when 2m<d , each \sigma\in X_{e}-X_{e-1} determines the subscheme that spans it. Next, consider the minimal free resolution of a Gorenstein quotient: \cdots\rightarrow F_{2}\rightarrow F_{1}\rightarrow S_{\bullet}\rightarrow R_{\bullet}\rightarrow 0 where F_{1} is the free module minimally generating I_{\bullet} , F_{2} is free on a minimal set of relations, etc. By the Hilbert Syzygy Theorem, this terminates at F_{n+1} . Example 2.4. When d=0 , the ideal is maximal: 0\rightarrow\langle x_{0},...,x_{n}\rangle\rightarrow S_{\bullet}\rightarrow k\rightarrow 0 and the minimal free resolution is the Koszul complex : 0\rightarrow\wedge^{n+1}V\otimes S(-n-1)\rightarrow\cdots\rightarrow\wedge^{i}V\otimes S(-i)\rightarrow\cdots\rightarrow V\otimes S(-1)\rightarrow S\rightarrow k Definition 2.2 . The array of natural numbers b_{i,j} such that: F_{i}=\bigoplus S(-j)^{b_{i,j}} is the betti table of R_{\bullet} , usually arranged (since j\geq i ) as: b_{0,0}=1 b_{1,1} b_{2,2} b_{3,3} \cdots b_{n+1,n+1} b_{1,2} b_{2,3} b_{3,4} \cdots b_{n+1,n+2} \vdots \vdots \vdots \vdots \vdots b_{1,d+1} b_{2,d+2} b_{3,d+3} \cdots b_{n+1,n+1+d} so that, for example, the betti table of the Koszul complex has a single row: b_{0,0}={{n+1}\choose 0} {{n+1}\choose 1} {{n+1}\choose 2} {{n+1}\choose 3} \cdots b_{n+1,n+1}={{n+1}\choose{n+1}} The left-right symmetry of this Koszul row generalizes via a well-known: Duality. [ 8 ] If R_{\bullet} is Gorenstein with socle in degree d , then (i) F_{n+1}=S(-n-1-d) , i.e. b_{n+1,n+1+d}=1 and b_{n+1,n+1+e}=0 for e<d . (ii) F_{n+1-m}\cong\mbox{Hom}_{S}(F_{m},S)(-n-1-d) , and when applied to the complex, (iii) \mbox{Hom}_{S}(F_{*},S)(-n-1-d)=-F_{n+1-*} (skew symmetry) Corollary. b_{n+1,n+1}=\cdots=b_{n+1,n+1+d-1}=0 . Example 2.5. When d=1 , there is a (smaller) Koszul complex for the first row: b_{0,0}={{n}\choose 0} {{n}\choose 1} {{n}\choose 2} {{n}\choose 3} \cdots b_{n,n}={{n}\choose{n}} 0 to which the second row is appended according to the Theorem: b_{0,0}={{n}\choose 0} {{n}\choose 1} {{n}\choose 2} {{n}\choose 3} \cdots {{n}\choose{n}} 0 0 {{n}\choose 0} {{n}\choose 1} {{n}\choose 2} \cdots {{n}\choose{n-1}} b_{n+1,n+2}={n\choose n} When d is arbitrary and \dim_{k}(R_{1})=1 the shape is analogous: b_{0,0}={{n}\choose 0} {{n}\choose 1} {{n}\choose 2} {{n}\choose 3} \cdots {{n}\choose{n}} 0 0 0 0 0 0 0 0

Read the complete text