Lower deviation for the maximum of two-speed branching Brownian motion (Q6152261)

The two-speed branching Brownian motion is a binary branching particle system, started from a particle at the origin, where particles independently: \begin{itemize} \item diffuse according to the SDE \(dX_s = (\sigma_1 \mathbf{1}_{s\leq bt} + \sigma_2 \mathbf{1}_{s> bt})\,dB_s\), where \(\sigma_1,\sigma_2,t>0,b\in (0,1)\) are fixed parameters, and \(B\) is a standard Brownian motion; \item branch, i.e.\ split into two particles at the same position, at rate \(1\). \end{itemize} Write \(M_t\) for the maximal position of a particle at time \(t\). It is known [\textit{A. Bovier} and \textit{L. B. Hartung}, Electron. J. Probab. 19, Paper No. 18, 28 p. (2014; Zbl 1288.60108)] that \(M_t-m_t\) converges in distribution as \(t\to \infty\) for an explicit \(m_t\), satisfying \(m_t\sim \sqrt{2}\beta t\) for some \(\beta>0\). This article focuses on the lower deviation for \(M_t\). More precisely, the author identifies the following large deviation rate \[ \lim_{t\to\infty} -\frac{1}{t} \log \mathbb{P}(M_t \leq \sqrt{2}\alpha t) \] for any \(\alpha<\beta\), similarly as in [\textit{B. Derrida} and \textit{Z. Shi}, Springer Proc. Math. Stat. 208, 303--312 (2017; Zbl 1386.60290)] where the ``single-speed'' case \(\sigma_1=\sigma_2\) was treated. This is done using the many-to-one lemma, and using precise estimates for the tail of Gaussian random variables. Depending on the value of \(\alpha\) with respect to the parameters (\(\sigma_1,\sigma_2,b,t\)) of the two-speed branching Brownian motion, the author is able to identify the event responsible for causing the deviation (time of the first branching event and position of the root particle).