University of Groningen Large-scale 21-cm Cosmology with LOFAR and AARTFAAC Gehlot, Bharat Kumar

Document Version

Publisher's PDF, also known as Version of record

Publication date: 2019

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):

Gehlot, B. K. (2019). Large-scale 21-cm Cosmology with LOFAR and AARTFAAC. University of Groningen.

Copyright

Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policy

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

6

6

6

in order to achieve a reliable detection of the 21-cm signal. This thesis has addressed various critical challenges faced by 21-cm signal CD and EoR ex-periments and discussed methods and strategies to either mitigate or bypass these challenges. This chapter summarises the main results of the thesis and provides an outlook for future work and prospects.

6.1 Main results

This section summarises the key results and conclusion of this thesis. The work presented in the thesis can be divided into two parts: (1) identifying the challenges in 21-cm signal CD experiments using the LOFAR-LBA system; (2) understanding and quantifying similar challenges on large angular scales using the LOFAR-AARTFAAC LBA and HBA systems.

6.1.1 A wide-field LOFAR-LBA power-spectrum analysis

Contamination or corruption of the 21-cm signal due to foregrounds (Galac-tic and Extra-galac(Galac-tic), calibration errors and ionospheric effects, pose major challenges in various Cosmic Dawn and Epoch of Reionization experiments. Chapter 2 has quantified a range of wide-field and calibration effects such as gain errors, polarised foregrounds, and ionospheric effects using the results of a pilot study of a field centred on the radio source 3C196 using LOFAR-LBA observations. The main conclusions of this chapter are:

1. Signal-corruption effects, such as polarisation leakage and ionospheric distortions as well as systematic biases in the signal power spectra called “excess-noise”, are greatly amplified at lower frequencies in CD experi-ments (50-80 MHz) when compared to the EoR experiexperi-ments (110-200 MHz).

6

power (or “excess-noise”) in Stokes I compared to Stokes V . This ex-cess remains more or less constant as a function of baseline length, and possibly a function of frequency. However, using a baseline cut in cal-ibration affects this excess, and it increases on baselines excluded from the calibration. The “excess-noise” is caused by calibration errors due to an incomplete sky-model, imperfect source subtraction and ionospheric effects. Using all baselines in calibration lowers the power on shortest baselines due to suppression of foregrounds which are not included in the calibration and can cause a suppression of the desired 21-cm signal. 3. The bright source Cas A, outside the observed field, causes a “pitchfork” structure in the power spectrum of Stokes I and the total polarised in-tensity in the baseline-delay space. A comparison of the power levels in this pitchfork structure between Stokes I and the total polarised inten-sity suggests this is caused by strong instrumental polarisation leakage, of the order of 30% in LOFAR-LBA, of two bright sources at lower el-evations. Subtraction of Cas A fro the data reduces the power on and around the ‘pitchfork but leaves power residuals of the order of 10%. 4. The residuals after Cas A subtraction decorrelate over a few minutes

time-scales. The ionospheric scintillation due to strong ionospheric tur-bulence can cause such strong residuals. The power spectrum towards Cas A takes the form of a compact source corrupted by Kolmogorov-type turbulence. This result suggests strong ionospheric activity towards Cas A, which is at a lower elevation, during the observations. A small diffractive scale is found towards Cas A (∼ 400 m at 60 MHz). To our knowledge, this is the smallest diffractive scale ever measured at 60 MHz.

It is crucial to mitigate or circumvent the effects observed in this study, e.g. excess noise, foreground suppression, polarisation leakage and ionospheric ef-fects. Excluding the short baselines from the instrument gain calibration avoids the suppression of the foregrounds and 21-cm signal but at the same time enhances excess noise on these baselines. This increase in noise can be mitigated somewhat by enforcing frequency smoothness as a constraint on the gain solutions. Because the instrument polarisation leakage in LOFAR-LBA is significant, a careful modelling and subtraction of the foregrounds outside the field of view are required to avoid the foregrounds leaking from Stokes I to Q and U and vice versa. Moreover, it is crucial to remove bright sources using calibration solution intervals below the coherence time-scale of iono-spheric scintillation (which is typical of the order of 10-30 seconds depending on ionospheric activity) to avoid significant residuals after source subtraction.

6

to calibrate at very low frequencies.

2. The calibration strategy used in this study enforces frequency smooth-ness in the gain calibration and uses lower calibration solution intervals for direction-independent calibration and to subtract Cas A and Cyg A. As a result of this, the “excess-noise” in the differential power spectrum of Stokes I is significantly reduced compared to the previous study; how-ever, it is still is higher than Stokes V .

3. There remains an unexplained excess in the ratio of noise estimates from Stokes I compared to the estimates Stokes V , determined by differencing the visibilities on a 12 s time scale. Still, this excess reduced when the data differencing is done on a 2 s time scale. This excess appears uncor-related along the frequency direction and with baseline length. Further-more, it does not seem to depend on the observed field, nor on calibra-tion, and even appears in the raw auto-correlations. Surprisingly, it only affects the Stokes I emission but leaves Stokes Q and U unaffected. So far we were unable to determine the exact cause of this excess noise, but we suspect it is due to some, still unknown, physical emission process outside the instrument since it only affects Stokes I. Thus, we refer to this unexplained excess as the “the physical excess noise” and treat it as an additional white noise for all practical purposes in the analysis. 4. A new Gaussian Process Regression (GPR) foreground removal

tech-nique and its application were illustrated to subtract the foregrounds that remain after the subtraction of calibration sky model. The remain-ing foregrounds consist of unmodelled sources, imperfectly subtracted sources, sources at or below the confusion noise, and diffuse mainly Galactic emission on the shortest baselines. GPR removes the smooth

6

resid-uals after foreground removal are largely noise-like and used for power spectrum estimation.

5. A modified LOFAR-EoR power spectrum extraction pipeline was used on the residuals, after foreground removal, to determine the brightness-temperature power spectra for both fields. The residual Stokes I power spectra for both fields are higher than the corresponding Stokes V power spectra. The ratio observed is somewhat higher than the “physical excess noise”, due to a combination of additional calibration errors and the “physical excess noise” in the data. It is necessary to remove the noise bias from the Stokes I power spectrum in order to determine the 21-cm power spectrum. I accounted for the “physical excess noise” while removing the noise bias from Stokes I and obtained a 95% confidence upper limit on the 21-cm power spectrum of ∆2

21 < (14561 mK)2 and

∆2

21 < (14886 mK)2 at k ∼ 0.038 h cMpc−1 in the redshift range z =

19.8− 25.2 for the two fields, respectively.

6.1.3 The AARTFAAC Cosmic Explorer

Chapter 4 provided an overview of the AARTFAAC Cosmic Explorer pro-gramme, which is a long-term project that we have begun in order to probe the 21-cm signal brightness temperature fluctuation at the location of the al-leged 21-cm absorption feature reported by the EDGES collaboration. We use a unique “semi drift-scan” mode of observations for the ACE observations, where every 15 minutes observing blocks are phase-tracked to the direction which crosses the zenith halfway through the corresponding observation block. The “physical excess noise” is also present in the ratio of Stokes I noise esti-mates compared to Stokes V in the ACE data and is twice as high compared to the LOFAR-LBA observations. The origin of this excess remains unknown. For longer integrations, the noise estimates on short baselines average down as expected but the “physical excess noise” remains at more or less the same level suggesting that it is coherent in time and does not average down. Whereas, on longer baselines the Stokes I noise estimate doesn’t average compared to the Stokes V leading to a much higher “physical excess noise” on longer base-lines for longer integration. Besides, solar and ionospheric activities cause higher power in Stokes V before sunset. This observation implies that night-time observations are more suitable of the ACE project compared to daynight-time observations. Finally, a first exploratory power spectrum results determined from 6 h of the ACE data is presented, processed with the LOFAR-EoR pro-cessing pipeline somewhat adjusted and refined to support the propro-cessing of AARTFAAC data. Two strategies were presented to remove the noise bias in 21-cm power spectrum limit estimation. A 2σ upper limit of ∆2

6

1 degree scale, the diffuse emission has a brightness temperature variance of about (13 K)2 _{and its angular power spectrum scales as ℓ}−2_{. Not including}

this emission in the sky-model can have a severe impact on the calibration of any instrument which probes these scales. Two methods to model this diffuse emission were demonstrated, viz. multiscale CLEAN and shapelet decomposi-tion. Each method has its pros and cons, e.g. the multiscale CLEAN method models the emission very well on scales smaller than a few degrees, but leaves the larger scales mostly unmodeled. On the contrary, the shapelet decompo-sition method can model the emission on scales larger than several degrees but is incapable of modelling the small-scale structures. A hybrid approach that combines both methods is required in order to produce a model which captures a broader range of spatial scales of the diffuse emission.

6.2 Future outlook and concluding remarks

Observing the redshifted 21-cm signal from the Cosmic Dawn and the Epoch of Reionization remains a daunting task because of many significant chal-lenges. All ongoing, as well as several upcoming, observational projects that seek to measure the cosmological 21-cm signal using current and planned in-terferometers, have made significant strides in recent years in improving their observational strategies and developing novel methods for data processing and analysis.

In this thesis, I have addressed an array of challenges faced by experiments probing the Cosmic Dawn at low-frequencies of 40_{− 80 MHz, which thus far} has remained largely unexplored and faces considerably more difficult chal-lenges than EoR experiments at higher frequencies. I have discussed various methods to mitigate or bypass these challenges, which can be applied to other

6

“loose ends”. For example, the presence of “physical excess noise” in the data is an intriguing, unexplained, and unexpected phenomenon. It is crucial to understand this effect in order to remove this bias from the 21-cm signal power spectrum. A part of my future effort will constitute understanding the ori-gin of this “physical excess noise” and developing methodologies to mitigate or remove this bias in 21-cm signal power spectrum analyses. I also plan to improve the calibration of the LOFAR-LBA data further and utilise novel imaging methods in the processing of CD and EoR data in order to achieve additional improvements in the power spectrum sensitivity.

In the thesis, I have also provided a sketch of the ACE project which aims to probe the Cosmic Dawn with the LOFAR-AARTFAAC system, which is a unique instrument in many aspects. Apart from a 21-cm power spectrum measurement, ACE data will also provide deep wide-field images of the sky on large angular scales in both Stokes I and polarised intensity. The LOFAR-EoR processing pipeline, due to its flexibility, has proven its ability to process ACE data. However, the current pipeline can only perform incoherent aver-aging of long-duration ACE observations. I plan to develop new methods that allow a coherent combination of longer-duration observations, using spherical harmonics imaging. This approach allows one to achieve much higher power spectrum sensitivity. Besides this, I also plan to improve the sky-model for calibration which will help to refine the calibration of ACE data. The sky maps obtained from calibrated ACE data will provide a catalogue of radio sources at low frequencies similar to the VLSS catalogue, as well as the most com-prehensive image-cube yet of the diffuse emission in the northern sky at low radio frequencies. Similarly, the AARTFAAC-HBA observations will help in modelling the diffuse emission in the LOFAR-EoR primary observing window (i.e. the NCP) and support the gain calibration of LOFAR-EoR observations. The work I have presented in my thesis has already proven to be informative and valuable in the process of mitigating many challenges in the ACE project and other experiments probing the Cosmic Dawn and the Epoch of Reioniza-tion. It will surely be of importance as well for the HERA instrument, under construction, that observes in a very similar drift-scan mode to AARTFAAC. I am confident that the combined efforts of radio astronomers, engineers and computer scientists, who are jointly searching for this feeble signal, will lead to breakthroughs and a first confirmed detection 21-cm signal in the next several years. The near future will undoubtedly be inspiring, as unexpected discoveries are made, and new astrophysical phenomena are unveiled!