We calculate the cosmic microwave background (CMB) angular trispectrum, spherical harmonic transform of the four-point correlation function, from primordial non-Gaussianity in primordial curvature perturbations characterized by a constant non-linear coupling parameter, $f_{\rm NL}$. We fully take into account the effect of the radiation transfer function, and thus provide the most accurate estimate of the signal-to-noise ratio of the angular trispectrum of CMB temperature anisotropy. We find that the predicted signal-to-noise ratio of the trispectrum summed up to a given $l$ is approximately a power-law, $(S/N)(<l)\sim 2.2\times 10^{-9}f^2_{\rm NL}l^2$, up to the maximum multipole that we have reached in our numerical calculation, $l=1200$, assuming that the error is dominated by cosmic variance. Our results indicate that the signal-to-noise ratio of the temperature trispectrum exceeds that of the bispectrum at the critical multipole, $l_c \sim 1500~(50/|f_{\rm NL}|)$. Therefore, the trispectrum of the Planck data is more sensitive to primordial non-Gaussianity than the bispectrum for $|f_{\rm NL}|\gtrsim 50$. We also report the predicted constraints on the amplitude of trispectrum, which may be useful for other non-Gaussian models such as curvaton models.

We investigate the question of how tightly we can constrain the cosmological parameters by using the ``cosmic inversion'' method in which we directly reconstruct the power spectrum of primordial curvature perturbations, $P(k)$, from the temperature and polarization spectra of the cosmic microwave background (CMB). In a previous paper, we suggested that it may be possible to constrain the cosmological parameters using the fact that the reconstructed $P(k)$ does not depend on how many polarization data we incorporate in our inversion procedure if and only if the correct values of the cosmological parameters are used. The advantage of this approach is that we need no assumption regarding the functional form of $P(k)$. In this paper, we estimate typical errors in the determination of the cosmological parameters when our method is applied to the PLANCK observation. We investigate constraints on $h$, $\Omega_b$, $\Omega_m$, and $\Omega_\Lambda$ through Monte Carlo simulations.

We develop a new method to reconstruct the power spectrum of primordial curvature perturbations, $P(k)$, by using both the temperature and polarization spectra of the cosmic microwave background (CMB). We test this method using several mock primordial spectra having non-trivial features including the one with an oscillatory component, and find that the spectrum can be reconstructed with a few percent accuracy by an iterative procedure in an ideal situation in which there is no observational error in the CMB data. In particular, although the previous ``cosmic inversion'' method, which used only the temperature fluctuations, suffered from large numerical errors around some specific values of $k$ that correspond to nodes in a transfer function, these errors are found to disappear almost completely in the new method.

We reconstruct the primordial spectrum of the curvature perturbation, $P(k)$, from the observational data of the Wilkinson Microwave Anisotropy Probe (WMAP) by the cosmic inversion method developed recently. In contrast to conventional parameter-fitting methods, our method can potentially reproduce small features in $P(k)$ with good accuracy. As a result, we obtain a complicated oscillatory $P(k)$. We confirm that this reconstructed $P(k)$ recovers the WMAP angular power spectrum with resolution up to $\Delta \ell \simeq 5$. Similar oscillatory features are found, however, in simulations using artificial cosmic microwave background data generated from a scale-invariant $P(k)$ with random errors that mimic observation. In order to examine the statistical significance of the nontrivial features, including the oscillatory behaviors, therefore, we consider a method to quantify the deviation from scale-invariance and apply it to the $P(k)$ reconstructed from the WMAP data. We find that there are possible deviations from scale-invariance around $k \simeq 1.5\times10^{-2}$ and $2.6\times10^{-2}{\rm Mpc}^{-1}$.