Left-invariant geometries on \(\operatorname{SU}(2)\) are uniformly doubling (Q1794367)

The starting point of this paper is the uniform analysis of the family of all left-invariant Riemannian metrics on the Lie group \(\operatorname{SU}(2)\), but the authors intend to study a more general problem, formulating the following conjecture: ``For a connected real compact Lie group \(K\), and the family \(\mathcal{L}(K)\) of all left-invariant Riemannian metrics \(g\) on \(K\), many aspects of spectral analysis of the corresponding Laplace-Beltrami operator \(\Delta_g\) with \(g\in \mathcal{L}(K)\), as well as the analysis of the associated heat equation \(\partial_t-\Delta_g=0\), can be controlled uniformly over \(\mathcal{L}(K)\)''. If \(\lambda_g\) denotes the lowest non-zero eigenvalue of \(-\Delta_g\), then the authors concretize their conjecture through the inequality \(\lambda_g\leq \frac{C_K}{\operatorname{diam}_g^2}\), for all \(g\in \mathcal{L}(K)\), where the constant \(C_K\) may depend on \(K\) but not on \(g\), and \(\operatorname{diam}_g\) denotes the diameter of \(K\) with respect to \(d_g\). In terms of the heat equation, the conjecture states that there are constants \(c_i=c_i(K)\in (0,\infty)\), \(i=1,\dots 4\) such that the fundamental solution (heat kernel) \((t,x,y)\rightarrow p^g_t(x,y)\) of the heat equation on \((K,g)\) satisfies \[ \frac{c_1}{V_g(t)}\exp(-c_2d_g(x,y)^2/t)\leq p_t^g(x,y)\leq \frac{c_3}{V_g(t)}\exp(-c_4d_g(x,y)^2/t), \] where \(V_g(r)\) denotes the volume of the ball of radius \(r\) with respect to the Riemannian volume measure \(\mu_g\). One reason to believe that this conjecture might be true is that it can be reduced to a simpler question. Let \((X,d,\mu)\) be a metric measure space, that is, \((X,d)\) is a metric space and \(\mu\) is a Borel measure on \(X\). By \(B(x,r)\) we denote the ball centred at \(x\in X\) of radius \(r>0\) with respect to the distance \(d\). The metric measure space \((X,d,\mu)\) is volume doubling if \[ D(M,d,\mu):=\sup_{x\in X, r>0}\frac{\mu(B(x,2r))}{\mu(B(x,r))}<\infty. \] The paper is now focused on the particular case of \((X,d,\mu)=(K,d_g,\mu_g)\), with the volume doubling constant denoted by \(D_g:=D(K,d_g,\mu_k)\). Then, in the context of this particular case the following conjecture is stated: Conjecture 1.1. Let \(K\) be a connected real compact Lie group. There is a constant \(D(K)\) such that \(D_g\leq D(K)\) for all \(g\in \mathcal{L}(K)\), that is, \(K\) is uniformly doubling with constant \(D(K)\). The main result of the paper is the proof of Conjecture 1.1 for \(K=\operatorname{SU}(2)\). Theorem 1.2. There exists a constant \(D\) such that, for any left-invariant metric \(g\) on \(\operatorname{SU}(2)\), we have \(D_g\leq D\). Then, since the underlying manifold of \(\operatorname{SU}(2)\) is the 3-sphere \(S^3\), Theorem 1.2 insures uniform volume doubling for a large family of Riemannian metrics on \(S^3\). The authors mention that the only other case where Conjecture 1.1 is known to hold is \(K=\mathbb{T}^n\), the \(n\)-dimensional torus, for any fixed \(n\). The limit cases of Theorem 1.2 are also mentioned. In one case, the metric tends to infinity in one direction, and the manifold approaches a sub-Riemannian manifold, which itself is doubling. If the metric tends to zero in one direction, the 3-dimensional manifold \(\operatorname{SU}(2)\) collapses to a 2-dimensional quotient, which is also doubling. Then in some sense, the question becomes whether the doubling constant varies continuously with respect to these limits. One of the difficulties is that both cases must be considered simultaneously. An important part of the article (Section 8) is dedicated to some ``consequences of volume doubling'': 8.1 The Poincaré inequality on compact Lie groups, 8.2 Spectral gap, 8.3 Heat kernel estimates, 8.4 Harnack inequality, 8.5 Gradient inequality, 8.6 Weyl counting function, 8.7 Heat kernel estimates: ergodicity, 8.8 Infinite products. In the last section (Section 9) the authors apply their results in connection to so-called sub-Rimeannian geometries.