About how large are algebraic Betti numbers?
Daniel Erman · 2025
Read About how large are algebraic Betti numbers? by Daniel Erman (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-2506-08133.txt.
Introduction
Daniel Erman asks how enormous certain numbers in algebra can grow, in a 2025 mathematics paper. The numbers, called algebraic Betti numbers, count the building blocks in a step-by-step breakdown of an algebraic shape's defining equations. For shapes placed in space by very strong embeddings, those counts can swell beyond easy reach. Erman works out rough bounds on their size rather than exact values.
His main tool is Boij-Söderberg theory, a framework that describes the possible shapes of these breakdowns. As a test case, he takes the flat plane embedded by degree-one-hundred maps into a space of more than five thousand dimensions. There, one middle Betti number turns out to have between 1,484 and 1,499 digits. The paper places such guesses on firm ground, giving the field order-of-magnitude answers where exact counts are out of reach.
Catalog record
| CREATOR | Daniel Erman |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 27 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-08133
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
About how large are algebraic Betti numbers? Daniel Erman Abstract . We use Boij-Söderberg theory to provide some order of magnitude bounds on algebraic Betti numbers. 1. Introduction Consider \mathbb{P}^{2}\subseteq\mathbb{P}^{5150} given by the degree 100 Veronese embedding. Let S be the coordinate ring of \mathbb{P}^{5150} and let I\subseteq S be the defining ideal of the image of \mathbb{P}^{2} . Write \beta_{i}(S/I) for the corresponding algebraic Betti number, i.e. \beta_{i}(S/I)=\dim\operatorname{Tor}_{i}(S/I,k) is the number of generators of the i ’th syzygy module of S/I . About how many digits does \beta_{i}(S/I) have? Our main results will give estimates for this type of question. For instance, we will show that in this example, \beta_{2000}(S/I) has between 1484 and 1499 digits. In [ 16 ] , Ein and Lazarsfeld coined the phrase asymptotic syzygies for the syzygies of algebraic varieties under increasingly positive embeddings. The first asymptotic syzygy results are perhaps the N_{p} theorems of Mark Green [ 32 ] , which showed that for smooth curves, as the positivity of the embedding increased, one obtained linear syzygies for more and more steps of the free resolution; see also [ 33 , 31 , 35 ] . Many later papers like [ 15 ] extended N_{p} theorems to higher dimensional varieties, but Ein and Lazarsfeld observed that, for very positive embeddings, these N_{p} results only describe a very small percentage of Betti diagram, and they began asking asymptotic questions about syzygies [ 16 , 17 ] . The qualitatitve asymptotic picture (which Betti numbers are nonzero?) is now largely complete [ 16 , 38 , 40 ] . We consider a small part of the quantitative story (how large are the Betti numbers?), which remains more open. We work throughout over an arbitrary field k . We will consider \mathbb{P}^{n} under the d -uple Veronese embedding. To simplify later notation, we write N=\binom{d+n}{n}-n-1 so we have \mathbb{P}^{n}\subseteq\mathbb{P}^{N+n} and N is the codimension. We write S for the coordinate ring of \mathbb{P}^{N+n} and I\subseteq S for the defining ideal of the image of \mathbb{P}^{n} . Write \beta_{i}(\mathbb{P}^{n};d):=\dim_{k}\operatorname{Tor}_{i}(S/I,k) for the i ’th total Betti number of S/I . Our main result about \mathbb{P}^{n} is the following: Theorem 1.1 . Fix n,d\geq 1 and let N=\binom{n+d}{d}-n-1 . For any i we have \binom{N}{i}\cdot N^{-n}\leq\beta_{i}(\mathbb{P}^{n};d)\leq\binom{N}{i}\cdot N^{n}. When d is small, the bounds are not terribly interesting. For instance, for the 5 -uple embedding \mathbb{P}^{2}\subseteq\mathbb{P}^{20} , let us use this to estimate \beta_{7}(\mathbb{P}^{2};5) . We have \binom{18}{7}\cdot 18^{-2}=98\tfrac{2}{9}\leq\beta_{7}(\mathbb{P}^{2};5)\leq 10310976=\binom{18}{7}\cdot 18^{2}. The actual Beti number is \beta_{7}(\mathbb{P}^{2};5)=417690 , and so the bounds are correct if unimpressive. However, as d grows, the binomial term \binom{N}{i} will totally overwhelm the error term of N^{\pm n} . To demonstrate this, let us consider an example where d is quite large. Example 1.2 . Consider the embedding of \mathbb{P}^{2} by the degree 10^{6} Veronese. In this case N\approx 5\times 10^{11} . Using some fairly basic estimation tricks (see § 5 ) our theorem implies that 10^{108661150967}\leq\beta_{10^{11}}(\mathbb{P}^{2};10^{6})\leq 10^{108661151026}. Thus \beta_{10^{11}}(\mathbb{P}^{2};10^{6}) has approximately 108661150996 digits, with an error of \pm 30 digits. 1 1 1 Since N^{\pm 2}\approx 10^{\pm 22} , our estimate could be tightened with a better approximation of the binomial term. We obtain similar results for other varieties. Let X be a variety and let L be a very ample line bundle on X . We let |L| be the projectived vector space H^{0}(X,L) so that \dim|L|=\dim H^{0}(X,L)-1 . Consider the embedding X\subseteq\mathbb{P}^{\dim|L|} . Let S be the coordinate ring of \mathbb{P}^{\dim|L|} and let I\subseteq S be the defining ideal of X . Write \beta_{i}(X;L):=\dim\operatorname{Tor}_{i}(S/I,k) . Theorem 1.3 . Let X be a variety and L\in\operatorname{Pic}(X) and continue with notation as in the previous paragraph. Let r be the Castelnuovo-Mumford regularity of S/I . For any i we have: \binom{\dim|L|-\dim X}{i}(\dim|L|)^{-r}\leq\beta_{i}(X;L)\leq\binom{\dim|L|}{i}(\dim|L|)^{r}. As above, this result is most dramatic when L is very positive. Example 1.4 . Let X\subseteq\mathbb{P}^{3} be a degree 13 hypersurface and let L be the pullback of {\mathcal{O}}_{\mathbb{P}^{3}}(1000) to X . We have \dim|L|=6441720\approx 6.4\times 10^{6} . We obtain the estimate 10^{1207673}\leq\beta_{10^{6}}(X;L)\leq 10^{1207707}. Thus, this Betti number has about 1207690 digits, with an error of \pm 17 digits. Again, see § 5 for details on this computation. The above results are both corollaries of our main result, which is a pair of more general algebraic bounds. Theorem 1.5 . Let M be a finitely generated, graded S=k[x_{1},\dots,x_{n}] -module, generated in degree 0 . We write \beta_{i}(M):=\dim_{k}\operatorname{Tor}_{i}(M,k) . We have 2 2 2 If \operatorname{codim}M=0 then the lower bound should be understood as 1 . \binom{\operatorname{codim}M}{i}(\operatorname{codim}M)^{-\operatorname{reg}M}\leq\frac{\beta_{i}(M)}{\beta_{0}(M)}\leq\binom{\operatorname{pdim}M}{i}(\operatorname{pdim}M)^{\operatorname{reg}M}. Here is the idea for the proof of Theorem 1.5 . The Betti table of M will have \operatorname{pdim}(M)+1 columns and \operatorname{reg}(M)+1 rows. Boij-Söderberg theory provides a decomposition of the Betti table of M as a sum of basic building blocks called pure diagrams , where each pure diagram is a particular Betti table with one nonzero entry in each column. For instance, here are potential shapes of several pure diagrams, each with 8 columns and 3 rows; each dot represents a nonzero entry. The main result of Boij-Söderberg theory (Theorem 2.1 below) decomposes \beta(M) as a sum of pure diagrams. For the pure diagrams appearing in this decomposition, the number of columns will range between \operatorname{codim}(M)+1 and \operatorname{pdim}(M)+1 , and the number of rows will be at most \operatorname{reg}(M)+1 . The theory was conjectured by Boij and Söderberg in [ 3 ] , and proven by Eisenbud and Schreyer in [ 21 ] , in combination with [ 20 , 2 ] and more. Properties of \beta(M) can be extracted by analyzing the corresponding pure diagrams. Imagine now that \operatorname{codim}(M)\gg\operatorname{reg}(M) ; each of the pure diagrams will then have many more columns than rows (see Figure 1 ). Figure 1. The heuristic behind our results is that the numerical properties of a pure diagram shaped like the bottom figure (300 columns and 3 rows) will be approximately the same as for the top figure (300 columns and 1 row). A pure diagram with only 1 row will correspond to a Koszul complex, and thus its Betti numbers precisely equal the binomial coefficients. Our key heuristic is: since any pure diagram with many more columns than rows will look approximately like a pure diagram with one row (e.g. the shapes in Figure 1 look similar, if you squint), any such diagram should behave approximately like a Koszul complex. This heuristic leads to a proof of the theorems above via an analysis of the numerical properties of such pure diagrams. We apply analytic techniques to pure diagrams, and then combine this with Boij–Söderberg theory to obtain our results. The main proofs involve little beyond first-year calculus. This basic idea has appeared before [ 27 , 37 , 4 ] , but our specific analyses are distinct from those. What should we make of these results? Are they surprising or trivial? How do they relate to other open questions about the quantitative behavior of (asymptotic) Betti numbers? We will attempt to provide partial answers. The lower bound in Theorem 1.5 is smaller than the bound predicted by the Buchsbaum-Eisenbud-Horrocks Conjecture (see [ 9 , p. 453] and [ 34 , Problem 24] ) due to the (\operatorname{codim}M)^{-\operatorname{reg}(M)}