If the conformal invariance of electromagnetism is broken during inflation, then primordial magnetic fields may be produced. If this symmetry breaking is generated by the coupling between electromagnetism and a scalar field-e.g. the inflaton, curvaton, or the Ricci scalar-then these magnetic fields may be correlated with primordial density perturbations, opening a new window to the study of non-gaussianity in cosmology. In order to illustrate, we couple electromagnetism to an auxiliary scalar field in a de Sitter background. We calculate the power spectra for scalarfield perturbations and magnetic fields, showing how a scale-free magnetic field spectrum with rms amplitude of ∼ nG at Mpc scales may be achieved. We explore the Fourier-space dependence of the cross-correlation between the scalar field and magnetic fields, showing that the dimensionless amplitude, measured in units of the power spectra, can grow as large as ∼ 500HI /M , where HI is the inflationary Hubble constant and M is the effective mass scale of the coupling.

We show that a conformal-invariance violating coupling of the inflaton to electromagnetism produces a cross correlation between curvature fluctuations and a spectrum of primordial magnetic fields. According to this model, in the case of power-law inflation, a primordial magnetic field is generated with a nearly flat power spectrum and rms amplitude ranging from nG to pG. We study the cross correlation, a three-point function of the curvature perturbation and two powers of the magnetic field, in real and momentum space. The cross-correlation coefficient, a dimensionless ratio of the three-point function with the curvature perturbation and magnetic field power spectra, can be several orders of magnitude larger than expected as based on the amplitude of scalar metric perturbations from inflation. In momentum space, the cross-correlation peaks for flattened triangle configurations, and is three orders of magnitude larger than the squeezed triangle configuration. These results suggest likely methods for distinguishing the observational signatures of the model.

No abstract

No abstract