Catherine Watkinson - Academia.edu (original) (raw)

Papers by Catherine Watkinson

Monthly Notices of the Royal Astronomical Society, Jun 4, 2012

We demonstrate a methodology for optimizing the ability of future dark energy surveys to answer m... more We demonstrate a methodology for optimizing the ability of future dark energy surveys to answer model selection questions, such as 'Is acceleration due to a cosmological constant or a dynamical dark energy model?'. Model selection figures of merit (FoMs) are defined, exploiting the Bayes factor, and surveys optimized over their design parameter space via a Monte Carlo method. As a specific example, we apply our methods to generic multi-fibre baryon acoustic oscillation spectroscopic surveys, comparable to that proposed for Subaru Measurement of Images and Redshifts Prime Focus Spectrograph, and present implementations based on the Savage-Dickey Density Ratio that are both accurate and practical for use in optimization. It is shown that whilst the optimal surveys using model selection agree with those found using the Dark Energy Task Force (DETF) FoM, they provide better informed flexibility of survey configuration and an absolute scale for performance; for example, we find survey configurations with close-to-optimal model selection performance despite their corresponding DETF FoM being at only 50 per cent of its maximum. This Bayes factor approach allows us to interpret the survey configurations that will be good enough for the task at hand, vital especially when wanting to add extra science goals and in dealing with time restrictions or multiple probes within the same project.

Monthly Notices of the Royal Astronomical Society, Nov 28, 2020

Numerous studies have established the theoretical potential of the 21cm bispectrum to boost our u... more Numerous studies have established the theoretical potential of the 21cm bispectrum to boost our understanding of the Epoch of Reionization (EoR), and therefore early generation of stars and galaxies. In this paper we take a first look at the impact of foregrounds and instrumental effects on the 21cm bispectrum and our ability to measure it. Unlike the power spectrum for which (in the absence of instrumental effects) there is a window clear of smooth-spectrum foregrounds in which the 21cm power spectrum may be detectable, there is no such "EoR window" for the bispectrum. Instead, on smaller scales EoR structure modulates that of the foregrounds (FG) to alter the EoR+FG bispectrum from that of the foregrounds in a complex manner. On larger scales the EoR structures are completely swamped by those of the foregrounds, and the EoR+FG bispectrum is entirely dominated by that of the foregrounds. It is therefore unlikely that the bispectrum will be useful in cases where we are observationally restricted to using foreground avoidance techniques. We also find that there is potential for instrumental effects to seriously corrupt the bispectrum, possibly even rendering the bispectrum useless for parameter estimation. On larger scales (k perp 0.3 cMpc −1), foreground removal using GMCA is found to recover the EoR bispectrum to a reasonable level of accuracy (even better than the power spectrum for certain configurations). Further studies are necessary to understand the error and/or bias associated with foreground removal before the 21cm bispectrum can be practically applied in analysis of future data.

Monthly Notices of the Royal Astronomical Society, Jul 30, 2021

The bispectrum is a 3-point statistic with the potential to provide additional information beyond... more The bispectrum is a 3-point statistic with the potential to provide additional information beyond power spectra analyses of survey datasets. Radio telescopes which broadly survey the 21cm emission from neutral hydrogen (H) are a promising way to probe LSS and in this work we present an investigation into the H intensity mapping (IM) bispectrum using simulations. We present a model of the redshift space H IM bispectrum including observational effects from the radio telescope beam and 21cm foreground contamination. We validate our modelling prescriptions with measurements from robust IM simulations, inclusive of these observational effects. Our foreground simulations include polarisation leakage, on which we use a Principal Component Analysis cleaning method. We also investigate the effects from a non-Gaussian beam including side-lobes. For a MeerKAT-like single-dish IM survey at = 0.39, we find that foreground removal causes a 8% reduction in the equilateral bispectrum's signal-to-noise ratio / , whereas the beam reduces it by 62%. We find our models perform well, generally providing 2 dof ∼ 1, indicating a good fit to the data. Whilst our focus is on post-reionisation, single-dish IM, our modelling of observational effects, especially foreground removal, can also be relevant to interferometers and reionisation studies.

Monthly Notices of the Royal Astronomical Society, Apr 8, 2015

We investigate the impact of sinks of ionizing radiation on the reionization-era 21-cm signal, fo... more We investigate the impact of sinks of ionizing radiation on the reionization-era 21-cm signal, focusing on 1-point statistics. We consider sinks in both the intergalactic medium and inside galaxies. At a fixed filling factor of H II regions, sinks will have two main effects on the 21-cm morphology: (i) as inhomogeneous absorbers of ionizing photons they result in smaller and more widespread cosmic H II patches; and (ii) as reservoirs of neutral gas they contribute a non-zero 21-cm signal in otherwise ionized regions. Both effects damp the contrast between neutral and ionized patches during reionization, making detection of the epoch of reionization with 21-cm interferometry more challenging. Here we systematically investigate these effects using the latest semi-numerical simulations. We find that sinks dramatically suppress the peak in the redshift evolution of the variance, corresponding to the midpoint of reionization. As previously predicted, skewness changes sign at midpoint, but the fluctuations in the residual H I suppress a late-time rise. Furthermore, large levels of residual H I dramatically alter the evolution of the variance, skewness and power spectrum from that seen at lower levels. In general, the evolution of the large-scale modes provides a better, cleaner, higher signal-tonoise probe of reionization.

Monthly Notices of the Royal Astronomical Society, Nov 22, 2019

We compute the bispectra of the 21cm signal during the Epoch of Reionization for three different ... more We compute the bispectra of the 21cm signal during the Epoch of Reionization for three different reionization scenarios that are based on a dark matter N-body simulation combined with a self-consistent, semi-numerical model of galaxy evolution and reionization. Our reionization scenarios differ in their trends of ionizing escape fractions (f esc) with the underlying galaxy properties and cover the physically plausible range, i.e. f esc effectively decreasing, being constant, or increasing with halo mass. We find the 21cm bispectrum to be sensitive to the resulting ionization topologies that significantly differ in their size distribution of ionized and neutral regions throughout reionization. From squeezed to stretched triangles, the 21cm bispectra features a change of sign from negative to positive values, with ionized and neutral regions representing below-average and above-average concentrations contributing negatively and positively, respectively. The position of the change of sign provides a tracer of the size distribution of the ionized and neutral regions, and allows us to identify three major regimes that the 21cm bispectrum undergoes during reionization. In particular the regime during the early stages of reionization, where the 21cm bispectrum tracks the peak of the size distribution of the ionized regions, provides exciting prospects for pinning down reionization with the forthcoming Square Kilometre Array.

Journal of open source software, Oct 22, 2020

Monthly Notices of the Royal Astronomical Society, Aug 6, 2014

We explore the impact of reionization topology on 21-cm statistics. Four reionization models are ... more We explore the impact of reionization topology on 21-cm statistics. Four reionization models are presented which emulate large ionized bubbles around over-dense regions (21CMFAST/ global-inside-out), small ionized bubbles in over-dense regions (local-inside-out), large ionized bubbles around under-dense regions (global-outside-in) and small ionized bubbles around under-dense regions (local-outside-in). We show that first generation instruments might struggle to distinguish global models using the shape of the power spectrum alone. All instruments considered are capable of breaking this degeneracy with the variance, which is higher in outside-in models. Global models can also be distinguished at small scales from a boost in the power spectrum from a positive correlation between the density and neutral-fraction fields in outside-in models. Negative skewness is found to be unique to inside-out models and we find that pre-SKA instruments could detect this feature in maps smoothed to reduce noise errors. The early, mid and late phases of reionization imprint signatures in the brightnesstemperature moments, we examine their model dependence and find pre-SKA instruments capable of exploiting these timing constraints in smoothed maps. The dimensional skewness is introduced and is shown to have stronger signatures of the early and mid-phase timing if the inside-out scenario is correct.

Monthly Notices of the Royal Astronomical Society, Aug 29, 2020

Three-point and high-order clustering statistics of the high-redshift 21cm signal contain valuabl... more Three-point and high-order clustering statistics of the high-redshift 21cm signal contain valuable information about the Epoch of Reionization. We present 3PCF-Fast, an optimised code for estimating the three-point correlation function of 3D pixelised data such as the outputs from numerical and semi-numerical simulations. After testing 3PCF-Fast on data with known analytic three-point correlation function, we use machine learning techniques to recover the mean bubble size and global ionisation fraction from correlations in the outputs of the publicly available 21cmFAST code. We assume that foregrounds have been perfectly removed and negligible instrumental noise. Using ionisation fraction data, our best MLP model recovers the mean bubble size with a median prediction error of around 10%, or from the 21cm differential brightness temperature with median prediction error of around 14%. A further two MLP models recover the global ionisation fraction with median prediction errors of around 4% (using ionisation fraction data) or around 16% (using brightness temperature). Our results indicate that clustering in both the ionisation fraction field and the brightness temperature field encode useful information about the progress of the Epoch of Reionization in a complementary way to other summary statistics. Using clustering would be particularly useful in regimes where high signal-to-noise ratio prevents direct measurement of bubble size statistics. We compare the quality of MLP models using the power spectrum, and find that using the three-point correlation function outperforms the power spectrum at predicting both global ionisation fraction and mean bubble size.

Monthly Notices of the Royal Astronomical Society, Jan 5, 2022

We present the first application of the isosceles bispectrum to MCMC parameter inference from the... more We present the first application of the isosceles bispectrum to MCMC parameter inference from the cosmic 21-cm signal. We extend the MCMC sampler 21CMMC to use the fast bispectrum code, BIFFT , when computing the likelihood. We create mock 1000-h observations with SKA1-low, using PYOBS21 to account for uv-sampling and thermal noise. Assuming the spin temperature is much higher than that of the cosmic microwave background, we consider two different reionization histories for our mock observations: fiducial and late-reionization. For both models we find that bias on the inferred parameter means and 1 σ credible intervals can be substantially reduced by using the isosceles bispectrum (calculated for a wide range of scales and triangle shapes) together with the power spectrum (as opposed to just using one of the statistics). We find that making the simplifying assumption of a Gaussian likelihood with a diagonal covariance matrix does not notably bias parameter constraints for the three-parameter reionization model and basic instrumental effects considered here. This is true even if we use extreme (unlikely) initial conditions which would be expected to amplify biases. We also find that using the cosmic variance error calculated with Monte Carlo simulations using the fiducial model parameters while assuming the late-reionization model for the simulated data also does not strongly bias the inference. This implies we may be able to sparsely sample and interpolate the cosmic variance error o v er the parameter space, substantially reducing computational costs. All codes used in this work are publicly available.

Proceedings of the International Astronomical Union, May 1, 2014

We discuss a selection of semi-numerical simulations of reionization whose analysis investigates ... more We discuss a selection of semi-numerical simulations of reionization whose analysis investigates the effect of small absorbing systems, neutral hydrogen within galaxies and the efficiency with which galaxies produce X-rays. We focus on the consequences for both observing the 21-cm 1-point statistics and their interpretation.

PhDT, Aug 1, 2015

We live in an age where we can claim to have precisely characterised our Universe with the elegan... more We live in an age where we can claim to have precisely characterised our Universe with the elegant ΛCDM model. However, this model predicts that around 70% of the Universe's content is dominated by a 'dark energy' that opposes gravity, and about 30% of its content is in some form of non-baryonic 'dark' matter that interacts only via gravity. Even when we consider the less exotic contents of our Universe, such as stars and galaxies, we are again forced to acknowledge the huge gaps in our knowledge. Observations of quasar spectra inform us that our Universe is in a highly ionized state during its later stages. However, from the cosmic microwave background we know that it was neutral for much of its history. It is sensible to assume that high-energy radiation, produced in abundance once stars and galaxies formed, would have driven the process of reionizing the Universe. However, the exact nature of early generations of stars and galaxies, and this process of reionization, are poorly constrained. In this work, we describe efforts to constrain the epoch of reionization, concentrating on the potential of the high-redshift 21-cm line of the hydrogen atom. Observations of the 21-cm signal would provide 3D maps of neutral hydrogen, and could provide vital constraints on the nature of reionization. Observing the 21-cm signal is extremely challenging due to very strong radio foregrounds, we therefore concentrate on efforts to constrain the 21-cm signal statistically. In particular, we characterise the sensitivity of one-point statistics (or moments) of the 21-cm signal. We study the sensitivity of 21-cm moments to a wide range of morphological properties that might be exhibited by ionized bubbles during reionization. We then investigate how 21-cm moments are impacted by the details of inhomogeneous recombinations, where ionized hydrogen captures electrons to become neutral again. We then consider the sensitivity of 21-cm moments to the properties of X-ray production. This work is essential, as without fully understanding a statistic and how it depends on the underlying physical processes, we cannot hope to gain meaningful constraints from its observation. This work also exhibits how information rich the 21-cm moments are, providing strong motivation to better understand them, and for more effort to be put into constraining them from observations.

Monthly Notices of the Royal Astronomical Society, Nov 23, 2018

Upcoming experiments such as the SKA will provide huge quantities of data. Fast modelling of the ... more Upcoming experiments such as the SKA will provide huge quantities of data. Fast modelling of the high-redshift 21cm signal will be crucial for efficiently comparing these data sets with theory. The most detailed theoretical predictions currently come from numerical simulations and from faster but less accurate semi-numerical simulations. Recently, machine learning techniques have been proposed to emulate the behaviour of these semi-numerical simulations with drastically reduced time and computing cost. We compare the viability of five such machine learning techniques for emulating the 21cm power spectrum of the publicly-available code SimFast21. Our best emulator is a multilayer perceptron with three hidden layers, reproducing SimFast21 power spectra 10 8 times faster than the simulation with 4% mean squared error averaged across all redshifts and input parameters. The other techniques (interpolation, Gaussian processes regression, and support vector machine) have slower prediction times and worse prediction accuracy than the multilayer perceptron. All our emulators can make predictions at any redshift and scale, which gives more flexible predictions but results in significantly worse prediction accuracy at lower redshifts. We then present a proof-of-concept technique for mapping between two different simulations, exploiting our best emulator's fast prediction speed. We demonstrate this technique to find a mapping between SimFast21 and another publicly-available code 21cmFAST. We observe a noticeable offset between the simulations for some regions of the input space. Such techniques could potentially be used as a bridge between fast semi-numerical simulations and accurate numerical radiative transfer simulations.

Astrophysics Source Code Library, Jun 1, 2021

Monthly Notices of the Royal Astronomical Society, Oct 13, 2018

We present analysis of the normalized 21-cm bispectrum from fully-numerical simulations of interg... more We present analysis of the normalized 21-cm bispectrum from fully-numerical simulations of intergalactic-medium heating by stellar sources and high-mass X-ray binaries (HMXBs) during the cosmic dawn. Ly-α coupling is assumed to be saturated, we therefore probe the nature of non-Gaussianities produced by X-ray heating processes. We find the evolution of the normalized bispectrum to be very different from that of the power spectrum. It exhibits a turnover whose peak moves from large to small scales with decreasing redshift, and corresponds to the typical separation of emission regions. This characteristic scale reduces as more and more regions move into emission with time. Ultimately, small-scale fluctuations within heated regions come to dominate the normalized bispectrum, which at the end of the simulation is almost entirely driven by fluctuations in the density field. To establish how generic the qualitative evolution of the normalized bispectrum we see in the stellar + HMXB simulation is, we examine several other simulations-two fully numerical simulations that include quasistellar object (QSO) sources, and two with contrasting source properties produced with the semi-numerical simulation 21CMFAST. We find the qualitative evolution of the normalized bispectrum during X-ray heating to be generic, unless the sources of X-rays are, as with QSOs, less numerous and so exhibit more distinct isolated heated profiles. Assuming mitigation of foreground and instrumental effects are ultimately effective, we find that we should be sensitive to the normalized bispectrum during the epoch of heating, so long as the spin temperature has not saturated by z ≈ 19.

Astrophysics Source Code Library, Jun 1, 2021

We live in an age where we can claim to have precisely characterised our Universe with the elegan... more We live in an age where we can claim to have precisely characterised our Universe with the elegant ΛCDM model. However, this model predicts that around 70% of the Universe's content is dominated by a 'dark energy' that opposes gravity, and about 30% of its content is in some form of non-baryonic 'dark' matter that interacts only via gravity. Even when we consider the less exotic contents of our Universe, such as stars and galaxies, we are again forced to acknowledge the huge gaps in our knowledge. Observations of quasar spectra inform us that our Universe is in a highly ionized state during its later stages. However, from the cosmic microwave background we know that it was neutral for much of its history. It is sensible to assume that high-energy radiation, produced in abundance once stars and galaxies formed, would have driven the process of reionizing the Universe. However, the exact nature of early generations of stars and galaxies, and this process of reionization, are poorly constrained. In this work, we describe efforts to constrain the epoch of reionization, concentrating on the potential of the high-redshift 21-cm line of the hydrogen atom. Observations of the 21-cm signal would provide 3D maps of neutral hydrogen, and could provide vital constraints on the nature of reionization. Observing the 21-cm signal is extremely challenging due to very strong radio foregrounds, we therefore concentrate on efforts to constrain the 21-cm signal statistically. In particular, we characterise the sensitivity of one-point statistics (or moments) of the 21-cm signal. We study the sensitivity of 21-cm moments to a wide range of morphological properties that might be exhibited by ionized bubbles during reionization. We then investigate how 21-cm moments are impacted by the details of inhomogeneous recombinations, where ionized hydrogen captures electrons to become neutral again. We then consider the sensitivity of 21-cm moments to the properties of X-ray production. This work is essential, as without fully understanding a statistic and how it depends on the underlying physical processes, we cannot hope to gain meaningful constraints from its observation. This work also exhibits how information rich the 21-cm moments are, providing strong motivation to better understand them, and for more effort to be put into constraining them from observations.

Monthly Notices of the Royal Astronomical Society, 2021

The bispectrum is a three-point statistic with the potential to provide additional information be... more The bispectrum is a three-point statistic with the potential to provide additional information beyond power spectra analyses of survey data sets. Radio telescopes that broadly survey the 21-cm emission from neutral hydrogen (H i) are a promising way to probe LSS and in this work we present an investigation into the H i intensity mapping (IM) bispectrum using simulations. We present a model of the redshift space H i IM bispectrum including observational effects from the radio telescope beam and 21-cm foreground contamination. We validate our modelling prescriptions with measurements from robust IM simulations, inclusive of these observational effects. Our foreground simulations include polarization leakage, on which we use a principal component analysis cleaning method. We also investigate the effects from a non-Gaussian beam including side-lobes. For a MeerKAT-like single-dish IM survey at z = 0.39, we find that foreground removal causes an 8rmpercent8{{\ \rm per\ cent}}8rmpercent reduction in the eq...

Monthly Notices of the Royal Astronomical Society, 2021

We apply for the first time Gaussian Process Regression (GPR) as a foreground removal technique i... more We apply for the first time Gaussian Process Regression (GPR) as a foreground removal technique in the context of single-dish, low redshift H i intensity mapping, and present an open-source python toolkit for doing so. We use MeerKAT and SKA1-MID-like simulations of 21 cm foregrounds (including polarization leakage), H i cosmological signal, and instrumental noise. We find that it is possible to use GPR as a foreground removal technique in this context, and that it is better suited in some cases to recover the H i power spectrum than principal component analysis (PCA), especially on small scales. GPR is especially good at recovering the radial power spectrum, outperforming PCA when considering the full bandwidth of our data. Both methods are worse at recovering the transverse power spectrum, since they rely on frequency-only covariance information. When halving our data along frequency, we find that GPR performs better in the low-frequency range, where foregrounds are brighter. It p...

Monthly Notices of the Royal Astronomical Society, 2020

Three-point and high-order clustering statistics of the high-redshift 21 cm signal contain valuab... more Three-point and high-order clustering statistics of the high-redshift 21 cm signal contain valuable information about the Epoch of Reionization (EoR). We present 3PCF-fast, an optimized code for estimating the three-point correlation function (3PCF) of 3D pixelized data such as the outputs from numerical and seminumerical simulations. After testing 3PCF-fast on data with known analytical 3PCF, we use machine learning techniques to recover the mean bubble size and global ionization fraction from correlations in the outputs of the publicly available 21cmfast code. We assume that foregrounds have been perfectly removed and negligible instrumental noise. Using ionization fraction data, our best multilayer perceptron (MLP) model recovers the mean bubble size with a median prediction error of around 10rmpercent10 {{\ \rm per\ cent}}10rmpercent, or from the 21 cm differential brightness temperature with median prediction error of around 14rmpercent14 {{\ \rm per\ cent}}14rmpercent. A further two MLP models recover the global ion...

Monthly Notices of the Royal Astronomical Society, Jun 4, 2012

We demonstrate a methodology for optimizing the ability of future dark energy surveys to answer m... more We demonstrate a methodology for optimizing the ability of future dark energy surveys to answer model selection questions, such as 'Is acceleration due to a cosmological constant or a dynamical dark energy model?'. Model selection figures of merit (FoMs) are defined, exploiting the Bayes factor, and surveys optimized over their design parameter space via a Monte Carlo method. As a specific example, we apply our methods to generic multi-fibre baryon acoustic oscillation spectroscopic surveys, comparable to that proposed for Subaru Measurement of Images and Redshifts Prime Focus Spectrograph, and present implementations based on the Savage-Dickey Density Ratio that are both accurate and practical for use in optimization. It is shown that whilst the optimal surveys using model selection agree with those found using the Dark Energy Task Force (DETF) FoM, they provide better informed flexibility of survey configuration and an absolute scale for performance; for example, we find survey configurations with close-to-optimal model selection performance despite their corresponding DETF FoM being at only 50 per cent of its maximum. This Bayes factor approach allows us to interpret the survey configurations that will be good enough for the task at hand, vital especially when wanting to add extra science goals and in dealing with time restrictions or multiple probes within the same project.

Monthly Notices of the Royal Astronomical Society, Nov 28, 2020

Numerous studies have established the theoretical potential of the 21cm bispectrum to boost our u... more Numerous studies have established the theoretical potential of the 21cm bispectrum to boost our understanding of the Epoch of Reionization (EoR), and therefore early generation of stars and galaxies. In this paper we take a first look at the impact of foregrounds and instrumental effects on the 21cm bispectrum and our ability to measure it. Unlike the power spectrum for which (in the absence of instrumental effects) there is a window clear of smooth-spectrum foregrounds in which the 21cm power spectrum may be detectable, there is no such "EoR window" for the bispectrum. Instead, on smaller scales EoR structure modulates that of the foregrounds (FG) to alter the EoR+FG bispectrum from that of the foregrounds in a complex manner. On larger scales the EoR structures are completely swamped by those of the foregrounds, and the EoR+FG bispectrum is entirely dominated by that of the foregrounds. It is therefore unlikely that the bispectrum will be useful in cases where we are observationally restricted to using foreground avoidance techniques. We also find that there is potential for instrumental effects to seriously corrupt the bispectrum, possibly even rendering the bispectrum useless for parameter estimation. On larger scales (k perp 0.3 cMpc −1), foreground removal using GMCA is found to recover the EoR bispectrum to a reasonable level of accuracy (even better than the power spectrum for certain configurations). Further studies are necessary to understand the error and/or bias associated with foreground removal before the 21cm bispectrum can be practically applied in analysis of future data.

Monthly Notices of the Royal Astronomical Society, Jul 30, 2021

The bispectrum is a 3-point statistic with the potential to provide additional information beyond... more The bispectrum is a 3-point statistic with the potential to provide additional information beyond power spectra analyses of survey datasets. Radio telescopes which broadly survey the 21cm emission from neutral hydrogen (H) are a promising way to probe LSS and in this work we present an investigation into the H intensity mapping (IM) bispectrum using simulations. We present a model of the redshift space H IM bispectrum including observational effects from the radio telescope beam and 21cm foreground contamination. We validate our modelling prescriptions with measurements from robust IM simulations, inclusive of these observational effects. Our foreground simulations include polarisation leakage, on which we use a Principal Component Analysis cleaning method. We also investigate the effects from a non-Gaussian beam including side-lobes. For a MeerKAT-like single-dish IM survey at = 0.39, we find that foreground removal causes a 8% reduction in the equilateral bispectrum's signal-to-noise ratio / , whereas the beam reduces it by 62%. We find our models perform well, generally providing 2 dof ∼ 1, indicating a good fit to the data. Whilst our focus is on post-reionisation, single-dish IM, our modelling of observational effects, especially foreground removal, can also be relevant to interferometers and reionisation studies.

Monthly Notices of the Royal Astronomical Society, Apr 8, 2015

We investigate the impact of sinks of ionizing radiation on the reionization-era 21-cm signal, fo... more We investigate the impact of sinks of ionizing radiation on the reionization-era 21-cm signal, focusing on 1-point statistics. We consider sinks in both the intergalactic medium and inside galaxies. At a fixed filling factor of H II regions, sinks will have two main effects on the 21-cm morphology: (i) as inhomogeneous absorbers of ionizing photons they result in smaller and more widespread cosmic H II patches; and (ii) as reservoirs of neutral gas they contribute a non-zero 21-cm signal in otherwise ionized regions. Both effects damp the contrast between neutral and ionized patches during reionization, making detection of the epoch of reionization with 21-cm interferometry more challenging. Here we systematically investigate these effects using the latest semi-numerical simulations. We find that sinks dramatically suppress the peak in the redshift evolution of the variance, corresponding to the midpoint of reionization. As previously predicted, skewness changes sign at midpoint, but the fluctuations in the residual H I suppress a late-time rise. Furthermore, large levels of residual H I dramatically alter the evolution of the variance, skewness and power spectrum from that seen at lower levels. In general, the evolution of the large-scale modes provides a better, cleaner, higher signal-tonoise probe of reionization.

Monthly Notices of the Royal Astronomical Society, Nov 22, 2019

We compute the bispectra of the 21cm signal during the Epoch of Reionization for three different ... more We compute the bispectra of the 21cm signal during the Epoch of Reionization for three different reionization scenarios that are based on a dark matter N-body simulation combined with a self-consistent, semi-numerical model of galaxy evolution and reionization. Our reionization scenarios differ in their trends of ionizing escape fractions (f esc) with the underlying galaxy properties and cover the physically plausible range, i.e. f esc effectively decreasing, being constant, or increasing with halo mass. We find the 21cm bispectrum to be sensitive to the resulting ionization topologies that significantly differ in their size distribution of ionized and neutral regions throughout reionization. From squeezed to stretched triangles, the 21cm bispectra features a change of sign from negative to positive values, with ionized and neutral regions representing below-average and above-average concentrations contributing negatively and positively, respectively. The position of the change of sign provides a tracer of the size distribution of the ionized and neutral regions, and allows us to identify three major regimes that the 21cm bispectrum undergoes during reionization. In particular the regime during the early stages of reionization, where the 21cm bispectrum tracks the peak of the size distribution of the ionized regions, provides exciting prospects for pinning down reionization with the forthcoming Square Kilometre Array.

Journal of open source software, Oct 22, 2020

Monthly Notices of the Royal Astronomical Society, Aug 6, 2014

We explore the impact of reionization topology on 21-cm statistics. Four reionization models are ... more We explore the impact of reionization topology on 21-cm statistics. Four reionization models are presented which emulate large ionized bubbles around over-dense regions (21CMFAST/ global-inside-out), small ionized bubbles in over-dense regions (local-inside-out), large ionized bubbles around under-dense regions (global-outside-in) and small ionized bubbles around under-dense regions (local-outside-in). We show that first generation instruments might struggle to distinguish global models using the shape of the power spectrum alone. All instruments considered are capable of breaking this degeneracy with the variance, which is higher in outside-in models. Global models can also be distinguished at small scales from a boost in the power spectrum from a positive correlation between the density and neutral-fraction fields in outside-in models. Negative skewness is found to be unique to inside-out models and we find that pre-SKA instruments could detect this feature in maps smoothed to reduce noise errors. The early, mid and late phases of reionization imprint signatures in the brightnesstemperature moments, we examine their model dependence and find pre-SKA instruments capable of exploiting these timing constraints in smoothed maps. The dimensional skewness is introduced and is shown to have stronger signatures of the early and mid-phase timing if the inside-out scenario is correct.

Monthly Notices of the Royal Astronomical Society, Aug 29, 2020

Three-point and high-order clustering statistics of the high-redshift 21cm signal contain valuabl... more Three-point and high-order clustering statistics of the high-redshift 21cm signal contain valuable information about the Epoch of Reionization. We present 3PCF-Fast, an optimised code for estimating the three-point correlation function of 3D pixelised data such as the outputs from numerical and semi-numerical simulations. After testing 3PCF-Fast on data with known analytic three-point correlation function, we use machine learning techniques to recover the mean bubble size and global ionisation fraction from correlations in the outputs of the publicly available 21cmFAST code. We assume that foregrounds have been perfectly removed and negligible instrumental noise. Using ionisation fraction data, our best MLP model recovers the mean bubble size with a median prediction error of around 10%, or from the 21cm differential brightness temperature with median prediction error of around 14%. A further two MLP models recover the global ionisation fraction with median prediction errors of around 4% (using ionisation fraction data) or around 16% (using brightness temperature). Our results indicate that clustering in both the ionisation fraction field and the brightness temperature field encode useful information about the progress of the Epoch of Reionization in a complementary way to other summary statistics. Using clustering would be particularly useful in regimes where high signal-to-noise ratio prevents direct measurement of bubble size statistics. We compare the quality of MLP models using the power spectrum, and find that using the three-point correlation function outperforms the power spectrum at predicting both global ionisation fraction and mean bubble size.

Monthly Notices of the Royal Astronomical Society, Jan 5, 2022

We present the first application of the isosceles bispectrum to MCMC parameter inference from the... more We present the first application of the isosceles bispectrum to MCMC parameter inference from the cosmic 21-cm signal. We extend the MCMC sampler 21CMMC to use the fast bispectrum code, BIFFT , when computing the likelihood. We create mock 1000-h observations with SKA1-low, using PYOBS21 to account for uv-sampling and thermal noise. Assuming the spin temperature is much higher than that of the cosmic microwave background, we consider two different reionization histories for our mock observations: fiducial and late-reionization. For both models we find that bias on the inferred parameter means and 1 σ credible intervals can be substantially reduced by using the isosceles bispectrum (calculated for a wide range of scales and triangle shapes) together with the power spectrum (as opposed to just using one of the statistics). We find that making the simplifying assumption of a Gaussian likelihood with a diagonal covariance matrix does not notably bias parameter constraints for the three-parameter reionization model and basic instrumental effects considered here. This is true even if we use extreme (unlikely) initial conditions which would be expected to amplify biases. We also find that using the cosmic variance error calculated with Monte Carlo simulations using the fiducial model parameters while assuming the late-reionization model for the simulated data also does not strongly bias the inference. This implies we may be able to sparsely sample and interpolate the cosmic variance error o v er the parameter space, substantially reducing computational costs. All codes used in this work are publicly available.

Proceedings of the International Astronomical Union, May 1, 2014

We discuss a selection of semi-numerical simulations of reionization whose analysis investigates ... more We discuss a selection of semi-numerical simulations of reionization whose analysis investigates the effect of small absorbing systems, neutral hydrogen within galaxies and the efficiency with which galaxies produce X-rays. We focus on the consequences for both observing the 21-cm 1-point statistics and their interpretation.

PhDT, Aug 1, 2015

We live in an age where we can claim to have precisely characterised our Universe with the elegan... more We live in an age where we can claim to have precisely characterised our Universe with the elegant ΛCDM model. However, this model predicts that around 70% of the Universe's content is dominated by a 'dark energy' that opposes gravity, and about 30% of its content is in some form of non-baryonic 'dark' matter that interacts only via gravity. Even when we consider the less exotic contents of our Universe, such as stars and galaxies, we are again forced to acknowledge the huge gaps in our knowledge. Observations of quasar spectra inform us that our Universe is in a highly ionized state during its later stages. However, from the cosmic microwave background we know that it was neutral for much of its history. It is sensible to assume that high-energy radiation, produced in abundance once stars and galaxies formed, would have driven the process of reionizing the Universe. However, the exact nature of early generations of stars and galaxies, and this process of reionization, are poorly constrained. In this work, we describe efforts to constrain the epoch of reionization, concentrating on the potential of the high-redshift 21-cm line of the hydrogen atom. Observations of the 21-cm signal would provide 3D maps of neutral hydrogen, and could provide vital constraints on the nature of reionization. Observing the 21-cm signal is extremely challenging due to very strong radio foregrounds, we therefore concentrate on efforts to constrain the 21-cm signal statistically. In particular, we characterise the sensitivity of one-point statistics (or moments) of the 21-cm signal. We study the sensitivity of 21-cm moments to a wide range of morphological properties that might be exhibited by ionized bubbles during reionization. We then investigate how 21-cm moments are impacted by the details of inhomogeneous recombinations, where ionized hydrogen captures electrons to become neutral again. We then consider the sensitivity of 21-cm moments to the properties of X-ray production. This work is essential, as without fully understanding a statistic and how it depends on the underlying physical processes, we cannot hope to gain meaningful constraints from its observation. This work also exhibits how information rich the 21-cm moments are, providing strong motivation to better understand them, and for more effort to be put into constraining them from observations.

Monthly Notices of the Royal Astronomical Society, Nov 23, 2018

Upcoming experiments such as the SKA will provide huge quantities of data. Fast modelling of the ... more Upcoming experiments such as the SKA will provide huge quantities of data. Fast modelling of the high-redshift 21cm signal will be crucial for efficiently comparing these data sets with theory. The most detailed theoretical predictions currently come from numerical simulations and from faster but less accurate semi-numerical simulations. Recently, machine learning techniques have been proposed to emulate the behaviour of these semi-numerical simulations with drastically reduced time and computing cost. We compare the viability of five such machine learning techniques for emulating the 21cm power spectrum of the publicly-available code SimFast21. Our best emulator is a multilayer perceptron with three hidden layers, reproducing SimFast21 power spectra 10 8 times faster than the simulation with 4% mean squared error averaged across all redshifts and input parameters. The other techniques (interpolation, Gaussian processes regression, and support vector machine) have slower prediction times and worse prediction accuracy than the multilayer perceptron. All our emulators can make predictions at any redshift and scale, which gives more flexible predictions but results in significantly worse prediction accuracy at lower redshifts. We then present a proof-of-concept technique for mapping between two different simulations, exploiting our best emulator's fast prediction speed. We demonstrate this technique to find a mapping between SimFast21 and another publicly-available code 21cmFAST. We observe a noticeable offset between the simulations for some regions of the input space. Such techniques could potentially be used as a bridge between fast semi-numerical simulations and accurate numerical radiative transfer simulations.

Astrophysics Source Code Library, Jun 1, 2021

Monthly Notices of the Royal Astronomical Society, Oct 13, 2018

We present analysis of the normalized 21-cm bispectrum from fully-numerical simulations of interg... more We present analysis of the normalized 21-cm bispectrum from fully-numerical simulations of intergalactic-medium heating by stellar sources and high-mass X-ray binaries (HMXBs) during the cosmic dawn. Ly-α coupling is assumed to be saturated, we therefore probe the nature of non-Gaussianities produced by X-ray heating processes. We find the evolution of the normalized bispectrum to be very different from that of the power spectrum. It exhibits a turnover whose peak moves from large to small scales with decreasing redshift, and corresponds to the typical separation of emission regions. This characteristic scale reduces as more and more regions move into emission with time. Ultimately, small-scale fluctuations within heated regions come to dominate the normalized bispectrum, which at the end of the simulation is almost entirely driven by fluctuations in the density field. To establish how generic the qualitative evolution of the normalized bispectrum we see in the stellar + HMXB simulation is, we examine several other simulations-two fully numerical simulations that include quasistellar object (QSO) sources, and two with contrasting source properties produced with the semi-numerical simulation 21CMFAST. We find the qualitative evolution of the normalized bispectrum during X-ray heating to be generic, unless the sources of X-rays are, as with QSOs, less numerous and so exhibit more distinct isolated heated profiles. Assuming mitigation of foreground and instrumental effects are ultimately effective, we find that we should be sensitive to the normalized bispectrum during the epoch of heating, so long as the spin temperature has not saturated by z ≈ 19.

Astrophysics Source Code Library, Jun 1, 2021

We live in an age where we can claim to have precisely characterised our Universe with the elegan... more We live in an age where we can claim to have precisely characterised our Universe with the elegant ΛCDM model. However, this model predicts that around 70% of the Universe's content is dominated by a 'dark energy' that opposes gravity, and about 30% of its content is in some form of non-baryonic 'dark' matter that interacts only via gravity. Even when we consider the less exotic contents of our Universe, such as stars and galaxies, we are again forced to acknowledge the huge gaps in our knowledge. Observations of quasar spectra inform us that our Universe is in a highly ionized state during its later stages. However, from the cosmic microwave background we know that it was neutral for much of its history. It is sensible to assume that high-energy radiation, produced in abundance once stars and galaxies formed, would have driven the process of reionizing the Universe. However, the exact nature of early generations of stars and galaxies, and this process of reionization, are poorly constrained. In this work, we describe efforts to constrain the epoch of reionization, concentrating on the potential of the high-redshift 21-cm line of the hydrogen atom. Observations of the 21-cm signal would provide 3D maps of neutral hydrogen, and could provide vital constraints on the nature of reionization. Observing the 21-cm signal is extremely challenging due to very strong radio foregrounds, we therefore concentrate on efforts to constrain the 21-cm signal statistically. In particular, we characterise the sensitivity of one-point statistics (or moments) of the 21-cm signal. We study the sensitivity of 21-cm moments to a wide range of morphological properties that might be exhibited by ionized bubbles during reionization. We then investigate how 21-cm moments are impacted by the details of inhomogeneous recombinations, where ionized hydrogen captures electrons to become neutral again. We then consider the sensitivity of 21-cm moments to the properties of X-ray production. This work is essential, as without fully understanding a statistic and how it depends on the underlying physical processes, we cannot hope to gain meaningful constraints from its observation. This work also exhibits how information rich the 21-cm moments are, providing strong motivation to better understand them, and for more effort to be put into constraining them from observations.

Monthly Notices of the Royal Astronomical Society, 2021

The bispectrum is a three-point statistic with the potential to provide additional information be... more The bispectrum is a three-point statistic with the potential to provide additional information beyond power spectra analyses of survey data sets. Radio telescopes that broadly survey the 21-cm emission from neutral hydrogen (H i) are a promising way to probe LSS and in this work we present an investigation into the H i intensity mapping (IM) bispectrum using simulations. We present a model of the redshift space H i IM bispectrum including observational effects from the radio telescope beam and 21-cm foreground contamination. We validate our modelling prescriptions with measurements from robust IM simulations, inclusive of these observational effects. Our foreground simulations include polarization leakage, on which we use a principal component analysis cleaning method. We also investigate the effects from a non-Gaussian beam including side-lobes. For a MeerKAT-like single-dish IM survey at z = 0.39, we find that foreground removal causes an 8rmpercent8{{\ \rm per\ cent}}8rmpercent reduction in the eq...

Monthly Notices of the Royal Astronomical Society, 2021

We apply for the first time Gaussian Process Regression (GPR) as a foreground removal technique i... more We apply for the first time Gaussian Process Regression (GPR) as a foreground removal technique in the context of single-dish, low redshift H i intensity mapping, and present an open-source python toolkit for doing so. We use MeerKAT and SKA1-MID-like simulations of 21 cm foregrounds (including polarization leakage), H i cosmological signal, and instrumental noise. We find that it is possible to use GPR as a foreground removal technique in this context, and that it is better suited in some cases to recover the H i power spectrum than principal component analysis (PCA), especially on small scales. GPR is especially good at recovering the radial power spectrum, outperforming PCA when considering the full bandwidth of our data. Both methods are worse at recovering the transverse power spectrum, since they rely on frequency-only covariance information. When halving our data along frequency, we find that GPR performs better in the low-frequency range, where foregrounds are brighter. It p...

Monthly Notices of the Royal Astronomical Society, 2020

Three-point and high-order clustering statistics of the high-redshift 21 cm signal contain valuab... more Three-point and high-order clustering statistics of the high-redshift 21 cm signal contain valuable information about the Epoch of Reionization (EoR). We present 3PCF-fast, an optimized code for estimating the three-point correlation function (3PCF) of 3D pixelized data such as the outputs from numerical and seminumerical simulations. After testing 3PCF-fast on data with known analytical 3PCF, we use machine learning techniques to recover the mean bubble size and global ionization fraction from correlations in the outputs of the publicly available 21cmfast code. We assume that foregrounds have been perfectly removed and negligible instrumental noise. Using ionization fraction data, our best multilayer perceptron (MLP) model recovers the mean bubble size with a median prediction error of around 10rmpercent10 {{\ \rm per\ cent}}10rmpercent, or from the 21 cm differential brightness temperature with median prediction error of around 14rmpercent14 {{\ \rm per\ cent}}14rmpercent. A further two MLP models recover the global ion...