By using the cubic spline interpolation method , we reconstruct the shape of the primordial scalar and tensor power spectra from the recently released Planck temperature and BICEP2 polarization cosmic microwave background data .
We find that the vanishing scalar index running ( { d } n _ { s } / { d } \ln k ) model is strongly disfavored at more than 3 \sigma confidence level on the k = 0.0002 Mpc ^ { -1 } scale .
Furthermore , the power-law parameterization gives a blue-tilt tensor spectrum , no matter using only the first 5 bandpowers n _ { t } = 1.20 ^ { +0.56 } _ { -0.64 } ~ { } ( 95 \% { CL } ) or the full 9 bandpowers n _ { t } = 1.24 ^ { +0.51 } _ { -0.58 } ~ { } ( 95 \% { CL } ) of BICEP2 data sets .
Unlike the large tensor-to-scalar ratio value ( r \sim 0.20 ) under the scale-invariant tensor spectrum assumption , our interpolation approach gives r _ { 0.002 } < 0.060 ~ { } ( 95 \% { CL } ) by using the first 5 bandpowers of BICEP2 data .
After comparing the results with/without BICEP2 data , we find that Planck temperature with small tensor amplitude signals and BICEP2 polarization data with large tensor amplitude signals dominate the tensor spectrum reconstruction on the large and small scales , respectively .
Hence , the resulting blue tensor tilt actually reflects the tension between Planck and BICEP2 data .