Lawrence Berkeley National Laboratory

To use type Ia supernovae as standard candles for cosmology we need accurate broadband magnitudes. In practice the observed magnitude may differ from the ideal magnitude-redshift relationship either through intrinsic inhomogeneities in the type Ia supernova population or through observational error. Here we investigate how we can choose filter bandpasses to reduce the error caused by both these effects. We find that bandpasses with large integral fluxes and sloping wings are best able to minimise several sources of observational error, and are also least sensitive to intrinsic differences in type Ia supernovae. The most important feature of a complete filter set for type Ia supernova cosmology is that each bandpass be a redshifted copy of the first. We design practical sets of redshifted bandpasses that are matched to typical high resistivity CCD and HgCdTe infra-red detector sensitivities. These are designed to minimise systematic error in well observed supernovae, final designs for specific missions should also consider signal-to-noise requirements and observing strategy. In addition we calculate how accurately filters need to be calibrated in order to achieve the required photometric accuracy of future supernova cosmology experiments such as the SuperNova-Acceleration-Probe (SNAP), which is one possible realisation of the Joint Dark-Energy mission (JDEM). We consider the effect of possible periodic miscalibrations that may arise from the construction of an interference filter.