Fluctuations of Biggins' martingales at complex parameters (Q2028944): Difference between revisions

Consider a branching random walk (BRW) on the real line. Let \(\mathcal Z_n\) be the point process encoding the positions of the particles at time \(n\). The Biggins martingale is defined by \[ Z_n(\lambda) := m(\lambda)^{-n} \int_{\mathbb R} e^{-\lambda x} \mathcal Z_n(dx), \] where \[ m(\lambda):= \mathbb E \int_{\mathbb R} e^{-\lambda x} \mathcal Z_1(dx). \] The main focus of the present paper is on the case when the parameter \(\lambda\) is complex. Under appropriate moment assumptions on the BRW, the Biggins martingale is known to converge to a non-degenerate limit \(Z(\lambda)\) for \(\lambda\) belonging to certain explicit domain \(\Lambda\) in the complex plane. The present paper studies the limit distributions of \(Z_n(\lambda)-Z(\lambda)\), as \(n\to\infty\), for \(\lambda\) belonging to this domain or to its boundary. There are three different regimes depending on the value of \(\lambda\). In the first regime, the limit laws are scale mixtures of the real or complex standard normal laws, the distribution of the parameter \(\sigma^2\) being essentially given by \(Z(2\mathrm{Re} \lambda)\). The boundary of this regime is also covered with the help of a Seneta-Heyde norming. In the second regime, the martingale fluctuations are determined by the extremal positions of particles in the branching random walk. Finally, there is a critical regime (typically on the boundary of \(\Lambda\)), where the limit laws are the laws of randomly stopped Lévy processes satisfying invariance properties similar to stability.