On the matrix formulation of time-delay interferometry

Abstract: Time-delay interferometry (TDI) is a processing step essential for the scientific exploitation of LISA, as it reduces the otherwise overwhelming laser noise in the interferometric measurements. The fundamental idea is to define new laser-noise-free observables by combining appropriately time-shifted measurements. First- and second-generation TDI combinations cancel laser noise under the assumption that the LISA armlengths are constant or evolve linearly with time, respectively. We recently extended TDI by solving for the laser-noise-free combinations implicitly, writing the likelihood of the data directly in terms of the basic measurements, and using a discretized representation of the delays that can accommodate any time dependence of the armlengths. We named the resulting formalism "TDI-infinity'' [PRD 103, 082001 (2021)]. According to Tinto, Dhurandhar, and Joshi [arXiv/2105.02054], our matrix-based approach is invalidated by the simplified start-up conditions assumed for the design matrix that connects the time series of laser-noise fluctuations to the time series of interferometric measurements along the LISA arms. Here we respond that, if those boundary conditions are indeed unrealistic, they do not invalidate the algorithm, since one can simply truncate the design matrix to exclude "incomplete'' measurements, or set them to zero. Our formalism then proceeds unmodified, except that the length of the laser-noise canceling time series is reduced by the number of excluded measurements. Tinto and colleagues further claim that the matrix formulation is merely a finite representation of the polynomial ring of delay operators introduced by Dhurandar, Nayak, and Vinet to formalize TDI. We show that this is only true if all interferometric delays are exact multiples of the sampling interval, which will not be possible in practical contexts such as LISA.