The Dark Energy Survey Supernova Program: Cosmological biases from supernova photometric classification
Cosmological analyses of samples of photometrically-identified Type Ia supernovae (SNe Ia) depend on understanding the effects of 'contamination' from core-collapse and peculiar SN Ia events. We employ a rigorous analysis on state-of-the-art simulations of photometrically identified SN Ia samples and determine cosmological biases due to such 'non-Ia' contamination in the Dark Energy Survey (DES) 5-year SN sample. As part of the analysis, we test on our DES simulations the performance of SuperNNova, a photometric SN classifier based on recurrent neural networks. Depending on the choice of non-Ia SN models in both the simulated data sample and training sample, contamination ranges from 0.8-3.5 %, with the efficiency of the classification from 97.7-99.5 %. Using the Bayesian Estimation Applied to Multiple Species (BEAMS) framework and its extension 'BEAMS with Bias Correction' (BBC), we produce a redshift-binned Hubble diagram marginalised over contamination and corrected for selection effects and we use it to constrain the dark energy equation-of-state, $w$. Assuming a flat universe with Gaussian $Ω_M$ prior of $0.311\pm0.010$, we show that biases on $w$ are $<0.008$ when using SuperNNova and accounting for a wide range of non-Ia SN models in the simulations. Systematic uncertainties associated with contamination are estimated to be at most $σ_{w, \mathrm{syst}}=0.004$. This compares to an expected statistical uncertainty of $σ_{w,\mathrm{stat}}=0.039$ for the DES-SN sample, thus showing that contamination is not a limiting uncertainty in our analysis. We also measure biases due to contamination on $w_0$ and $w_a$ (assuming a flat universe), and find these to be $<$0.009 in $w_0$ and $<$0.108 in $w_a$, hence 5 to 10 times smaller than the statistical uncertainties expected from the DES-SN sample.