Cosmic Reionization May Still Have Started Early and Ended Late: Confronting Early Onset with CMB Anisotropy and 21 cm Global Signals

The global history of reionization was shaped by the relative amounts of starlight released by three halo mass groups: atomic-cooling halos (ACHs) with virial temperatures Tvir > 10^4 K, either (1) massive enough to form stars even after reionization (HMACHs, >~ 10^9 Msun) or (2) less-massive (LMACHs), subject to star formation suppression when overtaken by reionization, and (3) H2-cooling minihalos (MHs) with Tvir < 10^4 K, whose star formation is predominantly suppressed by the H2-dissociating Lyman-Werner (LW) background. Our previous work showed that including MHs caused two-stage reionization - early rise to x ~ 0.1, driven by MHs, followed by a rapid rise, late, to x ~ 1, driven by ACHs - with a signature in CMB polarization anisotropy predicted to be detectable by the Planck satellite. Motivated by this prediction, we model global reionization semi-analytically for comparison with Planck CMB data and the EDGES global 21cm absorption feature, for models with: (1) ACHs, no feedback; (2) ACHs, self-regulated; and (3) ACHs and MHs, self-regulated. Model (3) agrees well with Planck E-mode polarization data, even with a substantial tail of high-redshift ionization, beyond the limit proposed by the Planck Collaboration (2018). No model reproduces the EDGES feature. For model (3), |dTb| <~ 60 mK across the EDGES trough, an order of magnitude too shallow, and absorption starts at higher z but is spectrally featureless. Early onset reionization by Population III stars in MHs is compatible with current constraints, but only if the EDGES interpretation is discounted or else other processes we did not include account for it.