Since the installation of COS onto HST in 2009, the sensitivity of the FUV detector has decreased with time at a rate that has varied between 0 and 30% per year. Since November 2012, the COS FUV detector has shown little or no sensitivity loss with time. However, the reference file applied in the pipeline prior to September 18, 2013, did not reflect this change in rate of sensitivity loss, but rather included sensitivity loss rates as high as 9.4 %/yr. As a result, the absolute flux calibration of data taken after November 2012 yielded artificially high flux values, which increased with observation date.. By August 2013, fluxes provided for G140L observations were discrepant by 10% to 40% (FUVA and FUVB, respectively), G160M by 2-6%, and G130M by 2-5%. The G130M blue modes (1055, 1096, and 1222) were discrepant by 10-12 %.
On September 18, 2013, the COS team delivered a new time-dependent sensitivity (TDS) reference file (x9h1254fl_tds.fits) which more accurately characterizes the rate of sensitivity loss observed since November 2012. The relative flux accuracy with the new TDS files is 2%, with the exception of a discontinuity of up to 4% between data taken before and after March 23, 2012. This discrepancy, seen on Segment A, appears to be correlated with a change in the high voltage of that segment, needed to ameliorate gain sag effects.
The absolute flux accuracy is within 5% for all modes except for the G130M 1055, 1096, and 1222 (blue modes) CENWAVE settings and for G140L data on Segment B, (CENWAVE 1280 for wavelengths < 1195). The absolute flux accuracy for the 1055/1096/1222 settings is 20% while it is 40% for G140L/FUVB.
Users concerned with the absolute or relative flux calibration of data taken after November 2012 should re-retrieve their data from the Mikulski Archive for Space Telescopes.
Further improvements to the flux calibration, including new and improved sensitivity curves for the G130M blue modes and G140L/FUVB are expected to be implemented this winter. Stay tuned!