A sum-bracket theorem for simple Lie algebras (Q6169065)

Given an algebraic structure \(A\) (e.g.~a group, ring, field) and a finite subset \(X\subseteq A\), a natural question is: how many elements can be obtained from \(X\) using at most \(n\) applications of the structure operations (addition, multiplication, Lie brackets)? A flurry of fascinating and deep results in this vein have been established by Freiman, Green and Ruzsa (for abelian groups), by Erdős and Szemerédi (for the ring of integers), by Breuillard, Eberhard, Gill, Green, Helfgott, Murphy, Pyber, Szabó and Tao (for finite non-abelian groups), and many others. This paper studies sum-product phenomena for simple Lie algebras. Given a Lie algebra \(\mathfrak{g}\) and a symmetric subset \(X\subseteq \mathfrak{g}\) (that is, \(X=-X\) and \(0\in X\)), denote by \(X^k\) the set of all elements in \(\mathfrak{g}\) obtained by applying at most \(k-1\) operations (either addition or Lie brackets) on arbitrary elements from \(X\). For every simple Lie algebra of each one of the forms \(\mathfrak{sl}_n,\mathfrak{so}_n,\mathfrak{sp}_{2n},\mathfrak{e}_{6,7,8},\mathfrak{f}_4,\mathfrak{g}_2\) over a field \(K\), the author proves (Theorem 1.1) that symmetric generating subsets grow fast: if \(\text{char}(K)=0\) then \(|X^k|\geq |X|^{1+\varepsilon}\), for some absolute constant \(\varepsilon>0\) and for some \(k\) depending only on \(\dim_K \mathfrak{g}\); and if \(\text{char}(K)=p>0\) then \(|X^k|\geq \min \{|X|^{1+\varepsilon},p^{\dim_K\mathfrak{g}}\}\). Using this, the author proves a `diameter bound' for such Lie algebras. Consider a finite Lie algebra \(\mathfrak{g}\) over \(\mathbb{F}_p\). Its \textit{diameter} \(\text{diam}(\mathfrak{g})\) is defined to be the minimum number \(k\), for which \(X^k=\mathfrak{g}\) for \textit{all} symmetric generating subsets \(X\subseteq \mathfrak{g}\). This notion is a natural analog of the parallel term for (Cayley graphs of) finite groups. The author proves (Corollary 1.2) that \(\text{diam}(\mathfrak{g})\) is bounded by \(\log^C |\mathfrak{g}|\), where \(C\) depends (polynomially) only on \(\dim_{\mathbb{F}_p} \mathfrak{g}\), rather than \(p\). In order to prove Theorem 1.1, the author proves that symmetric generating subsets of finite-dimensional simple algebras (not necessarily Lie algebras) are `not too concentrated' in affine subspaces of \(\mathfrak{g}\). In detail, for every symmetric generating subset \(X\subseteq \mathfrak{g}\) and for every affine subspace \(V\subseteq \mathfrak{g}\), it is shown (Theorem 1.4) that \(|X\cap V|\leq |X^k|^{\frac{\dim_K V}{\dim_K \mathfrak{g}}}\), for some \(k\) depending on \(\dim_K \mathfrak{g}\) (at most exponentially for general simple algebras, but at most qudratically for Lie/associative/Mal'cev algebras).