Title: LSST Target-of-Opportunity Observations of Gravitational Wave Events: Essential and Efficient

Abstract: We present simulated observations to assess the ability of LSST and the WFD survey to detect and characterize kilonovae - the optical emission associated with binary neutron star (and possibly black hole - neutron star) mergers. We expand on previous studies in several critical ways by exploring a range of kilonova models and several choices of cadence, as well as by evaluating the information content of the resulting light curves. We find that, depending on the precise choice of cadence, the WFD survey will achieve an average kilonova detection efficiency of $\approx 1.6-2.5\%$ and detect only $\approx 3-6$ kilonovae per year. The detected kilonovae will be within the detection volume of Advanced LIGO/Virgo (ALV). By refitting the best resulting LSST light curves with the same model used to generate them we find the model parameters are generally weakly constrained, and are accurate to at best a factor of $2-3$. Motivated by the finding that the WFD will yield a small number of kilonova detections, with poor light curves and marginal information content, and that the detections are in any case inside the ALV volume, we argue that target-of-opportunity follow-up of gravitational wave triggers is a much more effective approach for kilonova studies. We outline the qualitative foundation for such a program with the goal of minimizing the impact on LSST operations. We argue that observations in the $gz$-bands with a total time investment per event of $\approx 1.5$ hour per 10 deg$^2$ of search area is sufficient to rapidly detect and identify kilonovae with $\gtrsim 90\%$ efficiency. For an estimated event rate of $\sim20$ per year visible to LSST, this accounts for $\sim1.5\%$ of the total survey time. In this regime, LSST has the potential to be a powerful tool for kilonovae discovery, with detected events handed off to other narrow-field facilities for further monitoring.