A note about Jordan’s bound on the size of finite linear groups
Peter Müller · 2026
Read A note about Jordan’s bound on the size of finite linear groups by Peter Müller (2026) 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-2603-15813.txt.
Introduction
Müller revisits Jordan's 1878 bound on finite linear groups. Jordan showed each such group holds an abelian normal subgroup. The quotient size is bounded by dimension alone, with no explicit number given. Blichfeldt later found an explicit bound by representation methods. Bieberbach gave a geometric proof, which Frobenius then simplified. Müller presents a cleaned, self-contained form of Frobenius's argument. The streamlined proof yields a quotient bound of 25 to the n squared. The note keeps the full chain in one place.
Catalog record
| CREATOR | Peter Müller |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 13606 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.15813
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A note about Jordan’s bound on the size of finite linear groups Peter Müller Abstract . In 1878 Camille Jordan showed that every finite subgroup G ≤ GL n ( ℂ ) G\leq\GL_{n}(\mathbb{C}) has an abelian normal subgroup A A such that | G / A | \lvert G/A\rvert is bounded in terms of n n , but he did not give an explicit bound. An explicit bound was obtained by Blichfeldt in a series of papers beginning in 1904, using representation-theoretic methods. In 1911 Bieberbach gave a geometric proof, which is quite different from the approaches of Jordan and Blichfeldt, together with an explicit bound. Frobenius simplified this proof in the same year, and the resulting argument is still the simplest known. We present a self-contained and streamlined variant of Frobenius’s argument, yielding the bound | G / A | ≤ 25 n 2 \lvert G/A\rvert\leq 25^{n^{2}} . 1. Introduction A classical topic in group theory is the study of finite subgroups of \GL_{n}(\mathbb{C}) . Fixing n , such groups can be arbitrarily large. For instance, for every m\geq 1 the group of scalar matrices \{\,\zeta I_{n}\,\mid\,\zeta\in\mathbb{C},\zeta^{m}=1\,\} has order m . However, large finite subgroups of \GL_{n}(\mathbb{C}) have large abelian normal subgroups by the following classical result of Camille Jordan: Theorem 1.1 (Jordan 1878) . Let G\leq\GL_{n}(\mathbb{C}) be a finite group. Then G has an abelian normal subgroup A such that \lvert G/A\rvert\leq f(n) , where f(n) depends only on n . The statement of this theorem appears on page 91, and its proof is given in Chapitre II of Jordan’s long paper [ 10 ] . Jordan’s proof, which is purely algebraic, does not give an explicit bound f(n) . See [ 2 ] for a modern exposition of Jordan’s original argument, and many details about the history of Jordan’s theorem. Beginning in 1904, several papers by Blichfeldt used representation theoretic methods to obtain a different proof and the explicit bound f(n)=n!6^{(n-1)(\pi(n+1)+1)} , where \pi(n+1) is the number of primes less than or equal to n+1 . See [ 7 , §30] for this bound, or [ 9 , Theorem (14.12)] and [ 11 , §70] for a slightly weaker bound, where 6 is replaced by 12 . Later in 1911 Bieberbach [ 1 ] found a completely different proof of Jordan’s theorem, along with the bound f(n)=(1+32^{4}n^{10})^{2n^{2}} . In the same year Frobenius [ 8 ] simplified Bieberbach’s proof and obtained the better bound f(n)=(\sqrt{8n}+1)^{2n^{2}}-(\sqrt{8n}-1)^{2n^{2}} . A modern account of Frobenius’s method is Dixon’s version [ 6 , Theorem 5.7] , yielding the bound f(n)=(49n)^{n^{2}} (which is somewhat weaker than the previous bounds). In [ 12 ] Tao sketched a variant of Frobenius’ method. Full details are given in [ 3 , Section 2] . The bound f(n)=n^{Cn^{3}} for some undetermined absolute constant C is worse than the previous bounds. These proofs yield no information regarding the optimal bound f(n) for a general n . To date, Blichfeldt’s bound is asymptotically the best known one which can be proven without invoking the classification of the finite simple groups. Note that if c>6 , then f(n)=c^{n^{2}/\log n} works for all sufficiently large n by the prime number theorem. In Jordan’s theorem we necessarily have f(n)\geq(n+1)! for n\geq 4 , as can be seen as follows: Let the symmetric group \Sym(n+1) act on \mathbb{C}^{n+1} by permuting the coordinates. The n -dimensional subspace which consists of those vectors whose coordinates sum up to 0 is invariant under \Sym(n+1) . This yields an embedding of \Sym(n+1) into \GL_{n}(\mathbb{C}) . Since \Sym(n+1) does not have a nontrivial abelian normal subgroup for n\geq 4 , it follows that f(n)\geq(n+1)! . Conversely, using the classification of the finite simple groups, Collins [ 4 ] showed that Jordan’s theorem in fact holds with f(n)=(n+1)! once n\geq 71 . He also determined the optimal cases for all n\leq 70 . While more than 100 years old, even today Frobenius’ proof is considered to be the simplest approach to Jordan’s theorem. His proof (as well as Bieberbach’s precursor) consists of three steps: Step 1: The unitary trick – known already to Bieberbach and Frobenius – allows us to assume that G is a subgroup of the unitary group \operatorname{U}_{n}(\mathbb{C}) . Step 2: Using a suitable matrix norm \lVert\cdot\rVert on \mathbb{C}^{n\times n} and a sufficiently small positive constant c , one defines the subgroup A which is generated by all g\in G with \lVert I_{n}-g\rVert<c and shows that it is an abelian normal subgroup of G . Step 3: A simple volume packing argument or alternatively the pigeon-hole principle shows that \lvert G/A\rvert is bounded from above in terms of n and c only. The crucial piece for Step 2 is Lemma 3.1 (or variants thereof) below. Its proof by Frobenius or the one in the standard references [ 6 , Proof of Th. 5.7, part (a)] and [ 5 , (36.17) Lemma] is somewhat involved. Below we provide a different proof which we believe is more natural. Furthermore, we replace Step 1 by a weaker result which does not require the spectral theorem, and which is sufficient for our purposes. We prove Jordan’s theorem with the bound f(n)=25^{n^{2}} . 2. Normed vector spaces and linear isometries Let V be a finite-dimensional normed vector space over \mathbb{C} . Let \lVert v\rVert denote the norm of v\in V . Thus \lVert\cdot\rVert is a function from V to \mathbb{R} such that for each v,w\in V and \lambda\in\mathbb{C} the following holds: \displaystyle\lVert v\rVert \displaystyle\geq 0, \displaystyle\lVert v\rVert \displaystyle=0\text{ if and only if }v=0, \displaystyle\lVert\lambda v\rVert \displaystyle=\lvert\lambda\rvert\cdot\lVert v\rVert, \displaystyle\lVert v+w\rVert \displaystyle\leq\lVert v\rVert+\lVert w\rVert. A linear isometry of V is a map x\in\GL(V) such that \lVert xv\rVert=\lVert v\rVert for all v\in V . Let \U(V) denote the group of linear isometries of V . Lemma 2.1 . Let V be a finite-dimensional complex vector space and G\leq\GL(V) be a finite group. Then V can be equipped with a norm such that G\leq\U(V) , that is, G acts by linear isometries on V . Proof. Let \lVert\cdot\rVert^{\prime} be an arbitrary norm on V . Set \lVert v\rVert=\sum_{h\in G}\lVert hv\rVert^{\prime} for v\in V . Then \lVert\cdot\rVert is a norm on V and \lVert gv\rVert=\sum_{h\in G}\lVert h(gv)\rVert^{\prime}=\sum_{h\in G}\lVert(hg)v\rVert^{\prime}=\lVert v\rVert, because if h runs through G , then so does hg . ∎ We recall the definition of the operator norm of a\in\End(V) . Since V is finite-dimensional, the set \{\,v\in V\,\mid\,\lVert v\rVert=1\,\} is compact. Hence the continuous function V\to\mathbb{R} , v\mapsto\lVert av\rVert assumes a maximum on this set. Accordingly, the operator norm of a is defined as \lVert a\rVert_{\text{op}}=\max\{\,\lVert av\rVert\,\mid\,\lVert v\rVert=1\,\} . Immediate consequences of this definition are \displaystyle\lVert av\rVert \displaystyle\leq\lVert a\rVert_{\text{op}}\cdot\lVert v\rVert, \displaystyle\lVert\lambda a\rVert_{\text{op}} \displaystyle=\lvert\lambda\rvert\cdot\lVert a\rVert_{\text{op}}, \displaystyle\lVert a+b\rVert_{\text{op}} \displaystyle\leq\lVert a\rVert_{\text{op}}+\lVert b\rVert_{\text{op}}, \displaystyle\lVert ab\rVert_{\text{op}} \displaystyle\leq\lVert a\rVert_{\text{op}}\cdot\lVert b\rVert_{\text{op}} for all a,b\in\End(V) , v\in V , and \lambda\in\mathbb{C} . We note three more immediate consequences of the definition of the operator norm: \displaystyle\lVert xay\rVert_{\text{op}} \displaystyle=\lVert a\rVert_{\text{op}}\text{ for all $a\in\End(V)$ and $x,y\in\U(V)$}, \displaystyle\lVert x\rVert_{\text{op}} \displaystyle=1\text{ for all $x\in\U(V)$}, \displaystyle\lvert\mu\rvert \displaystyle\leq\lVert a\rVert_{\text{op}}\text{ for each eigenvalue $\mu$ of $a\in\End(V)$}. 3. Commuting elements The key step in the proof of Jordan’s theorem is the following lemma, which shows that elements in a finite subgroup of \U(V) sufficiently close to the identity