Axions play a central role in many realizations of large field models of inflation and in recent alternative mechanisms for generating primordial tensor modes in small field models. If these axions couple to gauge fields, the coupling produces a tachyonic instability that leads to an exponential enhancement of the gauge fields, which in turn can decay into observable scalar or tensor curvature perturbations. Thus, a fully self-consistent treatment of axions during inflation is important, and in this work we discuss the perturbative constraints on axions coupled to gauge fields. We show how the recent proposal of generating tensor modes through these alternative mechanisms is in tension with perturbation theory in the in-in formalism. Interestingly, we point out that the constraints are parametrically weaker than one would estimate based on naive power counting of propagators of the gauge field. In the case of non-Abelian gauge fields, we derive new constraints on the size of the gauge coupling, which apply also in certain models of natural large field inflation, such as alignment mechanisms.

We investigate the inflationary dynamics of two kinetically-coupled massless $U(1)$ gauge fields with time-varying kinetic-term coefficients. Ensuring that the system does not have strongly coupled regimes shrinks the parameter space. Also, we further restrict ourselves to systems that can be quantized using the standard creation, annihilation operator algebra. This second constraint limits us to scenarios where the system can be diagonalized into the sum of two decoupled, massless, vector fields with a varying kinetic-term coefficient. Such a system might be interesting for magnetogenesis because of how the strong coupling problem generalizes. We explore this idea by assuming that one of the gauge fields is the Standard Model $U(1)$ field and that the other dark gauge field has no particles charged under its gauge group. We consider whether it would be possible to transfer a magnetic field from the dark sector, generated perhaps before the coupling was turned on, to the visible sector. We also investigate whether the simple existence of the mixing provides more opportunities to generate magnetic fields. We find that neither possibility works efficiently, consistent with the well-known difficulties in inflationary magnetogenesis.

The damping of a non-uniform magnetic field between the redshifts of about $10^4$ and $10^6$ injects energy into the photon-baryon plasma and causes the CMB to deviate from a perfect blackbody spectrum, producing a so-called $\mu$-distortion. We can calculate the correlation $\langle\mu T\rangle$ of this distortion with the temperature anisotropy $T$ of the CMB to search for a correlation $\langle B^2\zeta\rangle$ between the magnetic field $B$ and the curvature perturbation $\zeta$; knowing the $\langle B^2\zeta\rangle$ correlation would help us distinguish between different models of magnetogenesis. Since the perturbations which produce the $\mu$-distortion will be much smaller scale than the relevant density perturbations, the observation of this correlation is sensitive to the squeezed limit of $\langle B^2\zeta\rangle$, which is naturally parameterized by $b_{\text{NL}}$ (a parameter defined analogously to $f_{\text{NL}}$). We find that a PIXIE-like CMB experiments has a signal to noise $S/N\approx 1.0 \times b_{\text{NL}} (\tilde B_\mu/10\text{ nG})^2$, where $\tilde B_\mu$ is the magnetic field's strength on $\mu$-distortion scales normalized to today's redshift; thus, a 10 nG field would be detectable with $b_{\text{NL}}=\mathcal{O}(1)$. However, if the field is of inflationary origin, we generically expect it to be accompanied by a curvature bispectrum $\langle\zeta^3\rangle$ induced by the magnetic field. For sufficiently small magnetic fields, the signal $\langle B^2 \zeta\rangle$ will dominate, but for $\tilde B_\mu\gtrsim 1$ nG, one would have to consider the specifics of the inflationary magnetogenesis model. We also discuss the potential post-magnetogenesis sources of a $\langle B^2\zeta\rangle$ correlation and explain why there will be no contribution from the evolution of the magnetic field in response to the curvature perturbation.

We investigate the phenomenological consequences of a modification of the initial state of a single inflationary field. While single-field inflation with the standard Bunch-Davies initial vacuum state does not generally produce a measurable three-point function (bispectrum) in the squeezed configuration, allowing for a non-standard initial state produces an exception. Here, we calculate the signature of an initial state modification in single-field slow-roll inflation in both the scale-dependent bias of the large-scale structure (LSS) and mu-type distortion in the black-body spectrum of the cosmic microwave background (CMB). We parametrize the initial state modifications and identify certain choices of parameters as natural, though we also note some fine-tuned choices that can yield a larger bispectrum. In both cases, we observe a distinctive k^-3 signature in LSS (as opposed to k^-2 for the local-form). As a non-zero bispectrum in the squeezed configuration correlates a long-wavelength mode with two short-wavelength modes, it induces a correlation between the CMB temperature anisotropy on large scales with the temperature-anisotropy-squared on very small scales; this correlation persists as the small-scale anisotropy-squared is processed into mu-type distortions. While the local-form mu-distortion turns out to be too small to detect in the near future, a modified initial vacuum state enhances the signal by a large factor owing to an extra factor of k_1/k. For example, a proposed absolutely-calibrated experiment, PIXIE, is expected to detect this correlation with a signal-to-noise ratio greater than 10, for an occupation number of about 0.5 in the observable modes. Relatively calibrated experiments such as Planck and LiteBIRD should also be able to measure this effect, provided that the relative calibration between different frequencies meets the required precision. (Abridged)

Single-field slow-roll inflation with a non-vacuum initial state has an enhanced bispectrum in the local limit. We numerically calculate the local-type fNL signal in the CMB that would be measured for such models (including the full transfer function and 2D projection). The nature of the result depends on several parameters, including the occupation number N_k, the phase angle \theta_k between the Bogoliubov parameters, and the slow-roll parameter \epsilon. In the most conservative case, where one takes \theta_k \approx \eta_0 k (justified by physical reasons discussed within) and \epsilon\lesssim 0.01, we find that 0 < fNL < 1.52 (\epsilon/0.01), which is likely too small to be detected in the CMB. However, if one is willing to allow a constant value for the phase angle \theta_k and N_k=O(1), fNL can be much larger and/or negative (depending on the choice of \theta_k), e.g. fNL \approx 28 (\epsilon/0.01) or -6.4 (\epsilon/0.01); depending on \epsilon, these scenarios could be detected by Planck or a future satellite. While we show that these results are not actually a violation of the single-field consistency relation, they do produce a value for fNL that is considerably larger than that usually predicted from single-field inflation.

In 2004, Creminelli and Zaldarriaga proposed a consistency relation for the primordial curvature perturbation of all single-field inflation models; it related the bispectrum in the squeezed limit to the spectral tilt. We have developed a technique, based in part on the Creminelli and Zaldarriaga argument, that can greatly simplify the calculation of the squeezed-limit bispectrum using the in-in formalism; we were able to arrive at a generic formula that does not rely on a slow-roll approximation. Using our formula, we explicitly tested the consistency relation for power-law inflation and for an exactly scale-invariant model by Starobinsky; for the latter model, Creminelli and Zaldarriaga's argument predicts a vanishing bispectrum whereas our quantum calculation shows a non-zero bispectrum that approaches zero in the long-wavelength limit and for inflation with a large number of e-folds.