GMRT observation towards detecting the post-reionization 21-cm signal

Abstract

The redshifted 21-cm signal from neutral hydrogen (H i) is an important future probe of the high-redshift Universe. We have analysed 610 MHz Giant Metrewave Radio Telescope (GMRT) observations towards detecting this signal from z= 1.32. The multi-frequency angular power spectrum Cℓ(Δν) is used to characterize the statistical properties of the background radiation across angular scales ∼20 arcsec to 10 arcmin, and a frequency bandwidth of 7.5 MHz with resolution 125 kHz. The measured Cℓ(Δν) which ranges from 7 to 18 mK2 is dominated by foregrounds, the expected H i signal CHIℓ (Δν) ∼ 10−6 to 10−7 mK2 is several orders of magnitude smaller and detecting this is a big challenge. The foregrounds, believed to originate from continuum sources, is expected to vary smoothly with Δν whereas the H i signal decorrelates within ∼0.5 MHz, and this holds the promise of separating the two. For each ℓ, we use the interval 0.5 ≤Δν≤ 7.5 MHz to fit a fourth-order polynomial which is subtracted from the measured Cℓ(Δν) to remove any smoothly varying component across the entire bandwidth Δν≤ 7.5 MHz. The residual Cℓ(Δν), we find, has an oscillatory pattern with amplitude and period, respectively, ∼0.1 mK2 and Δν= 3 MHz at the smallest ℓ value of 1476, and the amplitude and period decreasing with increasing ℓ. Applying a suitably chosen high pass filter, we are able to remove the residual oscillatory pattern for ℓ= 1476 where the residual Cℓ(Δν) is now consistent with zero at the 3σ noise level. Based on this we conclude that we have successfully removed the foregrounds at ℓ= 1476 and the residuals are consistent with noise. We use this to place an upper limit on the H i signal whose amplitude is determined by ⁠, where and b are the H i neutral fraction and the H i bias, respectively. A value of greater than 7.95 would have been detected in our observation, and is therefore ruled out at the 3σ level. For comparison, studies of quasar absorption spectra indicate which is ∼330 times smaller than our upper limit. We have not succeeded in completely removing the residual oscillatory pattern, whose cause is presently unknown to us, for the larger ℓ values.