A Spherical Harmonic Analysis of the Ooty Wide Field Array (OWFA) Visibility Signal
Extracted source text · arXiv 1804.00493
\documentclass[referee,useAMS,usenatbib,fleqn]{mnras}
\usepackage{amsopn, natbib} \usepackage[utf8]{inputenc}
\usepackage{amsmath} \usepackage{caption} \usepackage{amsfonts}
\usepackage{amssymb} \usepackage{graphicx} \usepackage{hyperref}
\usepackage{subfigure} \usepackage{psfrag}
%\usepackage{auto-pst-pdf}
%\usepackage[left=2cm,right=2cm,top=2cm,bottom=2cm]{geometry}
%%%%%%%%%%%%%%%%%%%%%%
\def\visn{\mathcal{V}\left( \mathbfit{U}_n, \nu \right)}
\def\vis{\mathcal{V}\left( \mathbfit{U}, \nu \right)}
\def\visnonu{\mathcal{V}\left( \mathbfit{U}\right)}
\def\visnonup{\mathcal{V}^{'}\left( \mathbfit{U}\right)}
\def\v2{V_2\left(\mathbfit{U},\Delta \nu = 0\right)}
\def\viscor{V_2\left( \mathbfit{U}_a,\nu_i; \mathbfit{U}_b,\nu_j\right)}
\def\viscornonu{V_2\left( \mathbfit{U}_a; \mathbfit{U}_b\right)}
\def\viscoradjnonu{V_2\left( \mathbfit{U}_a; \mathbfit{U}_{a \pm 1}\right)}
\def\viscora{V_2\left( \mathbfit{U}_a,\nu_i; \mathbfit{U}_a,\nu_j\right)}
\def\viscorsame{V_2\left( \mathbfit{U}_a; \mathbfit{U}_a\right)}
\def\maps{\mathcal{C}_{\ell}\left( \nu_i, \nu_j \right)}
\newcommand{\norm}[1]{\left\lVert#1\right\rVert}
%\def\dBdTnu{\left(\frac{\partial B}{\partial T}\right)_\nu}
\def\dBdTnu{Q_{\nu}}
\def\dBdTnonu{Q}
\def\dBdT{\left(\frac{\partial B}{\partial T}\right)}
\def\Blm{B_{\ell}^m\left(\mathbfit{U}, \nu \right)}
\def\BlmR{ B_{\ell}^m\left(\U, \nu \right)\right}
%\def\BlmR{\left[ B_{\ell}^m\left(\U, \nu \right)\right]_{\R}}
\def\Blmnonu{B_{\ell}^m\left(\mathbfit{U} \right)}
%\def\BlmRnonu{\left[ B_{\ell}^m\left(\U \right)\right]_{\R}}
\def\BlmRnonu{B_{\ell}^m\left(\U \right)}
\def\Blma{B_{\ell}^m(\U_a )}
\def\alm{a_{\ell}^m}
\def\almR{\left[a_{\ell}^m\right]_R}
\def\almnu{a{\ell}^m \left( \nu \right)}
\def\Ylm{Y_{\ell}^m(\hat{\mathbfit{n}})}
\def\Ylms{Y_{\ell}^{*m}(\hat{\mathbfit{n}})}
\def\pb{A\left(\Delta \hat{\mathbfit{n}},\nu \right)}
\def\pbnonu{A\left(\Delta \hat{\mathbfit{n}}\right)}
\def\ap{\tilde{a}(\U - \U^{'}, \nu )}
\def\apnonu{\tilde{a}(\U - \U^{'})}
\def\Flm{F_{\ell}^m}
\def\Wl{\mathcal{W}_{\ell}\left(\U_a, \nu_i ;\U_b, \nu_j \right)}
\def\Wlsame{\mathcal{W}_{\ell}\left(\U_a, \nu_i ;\U_a, \nu_j \right)}
\def\Wladj{\mathcal{W}_{\ell}\left(\U_a, \nu_i ;\U_{a \pm 1}, \nu_j \right)}
\def\Wlsamec{\mathcal{W}_{\ell}\left(\U_a, \nu_c ;\U_a, \nu_c \right)}
\def\Wladjc{\mathcal{W}_{\ell}\left(\U_a, \nu_c ;\U_{a \pm 1}, \nu_c \right)}
\def\U{\mathbfit{U}}
\newcommand{\HI}{{H{\sc i}~}}
\def\n{\hat{\mathbfit{n}}}
\def\m{\hat{\mathbfit{p}}}
\def\dn{\Delta \mathbfit{n}}
\newcommand{\bk}{\mathbfit{k}}
\newcommand{\ppk}{\mathbfit{k}_{\perp}}
\newcommand{\kpar}{k_{\parallel}}
\def\SHE{SH}
\def\R{R}
%%%%%%%%%%%%%%%%%%%%%%%%%
%\newcommand{\tcm}{21-{\rm cm}}
%%%%%%%%%%%%%%%%%%%%%%%%%
\title{A Spherical Harmonic Analysis of the Ooty Wide Field Array (OWFA) Visibility Signal}
\author[Chatterjee et al.]{Suman Chatterjee$^{1,2}$\thanks{E-mail: suman05@phy.iitkgp.ernet.in},
Somnath Bharadwaj$^{1,2}$\thanks{E-mail: somnath@phy.iitkgp.ernet.in }\\ $^{1}$Department of Physics,
Indian Institute of Technology Kharagpur, Kharagpur - 721 302,
India.\\ $^{2}$Centre for Theoretical Studies, Indian Institute of
Technology Kharagpur, Kharagpur - 721 302, India.}
\begin{document}
\date{\today}
\pagerange{\pageref{firstpage}--\pageref{lastpage}} \pubyear{2018}
\maketitle
\label{firstpage}
\begin{abstract}
Considering redshifted $21$-cm intensity mapping with the upcoming
OWFA whose field of view subtends $\sim 57^{\circ}$ in the N-S
direction, we present a formalism which relates the
measured visibilities to the spherical harmonic coefficients of the
sky signal. We use this to calculate window functions
which relate the two-visibility correlations {\it i.e.} the correlation between
the visibilities measured at two baselines and two frequencies, to different multipoles
of the multi-frequency angular power spectrum
$C_{\ell}(\nu_1,\nu_2)$. The formalism here is validated using
simulations. We also present approximate
closed form analytical expressions which can be used to calculate the
window functions. Comparing the widely adopted
flat sky approximation, we find that its predictions match those
of our spherical harmonic formalism to within $16 \%$
across the entire OWFA baseline range. The match improves at large
baselines where we have $< 5 \%$ deviations.
\end{abstract}
\begin{keywords}
Interferometric; cosmology: observations, diffuse radiation,
large-scale structure of Universe.
\end{keywords}
\newpage
\section{Introduction} \label{sec:intro}
Intensity mapping with neutral Hydrogen \HI 21-cm radiation is a
promising tool to study the large scale structures in the
post-reionization Universe \citep{Bharadwaj2001b}. The redshifted \HI 21-cm
observations hold the potential of measuring the Baryon Acoustic
Oscillation (BAO) that is embedded in the power spectrum of \HI 21-cm
intensity fluctuations at all redshifts and the comoving scale of BAO
can be used as a standard ruler to constrain the evolution of the
equation of state for dark energy \citep{Wyithe2008, Chang2008,
Seo2010, Masui2010}. Further, a measurement of just the \HI 21-cm power
spectrum can also be used to constrain the cosmological parameters
\citep{Bharadwaj2009,Visbal2009}. The higher order statistics such as
the \HI 21-cm bispectrum holds the prospect of quantifying the
non-Gaussianities in the \HI 21-cm signal \citep{Ali2005, Hazra2012}.
Using the \HI signal in cross-correlation with the WiggleZ galaxy survey data,
the Green Bank Telescope (GBT) has made the first detection of the \HI
signal in emission at $z \approx 0.8$ \citep{Chang2010}.
A number of post-reionization experiments are either being planned or are ongoing at present.
The Giant Meterwave Radio telescope \citep[GMRT;][]{Swarup1991} is
sensitive to the cosmological \HI signal from a range of redshifts
in the post-reionization era \citep{Bharadwaj2003,
Bharadwaj2005}. The upgraded GMRT \citep[uGMRT;][]{Gupta2017} is expected
to have a larger bandwidth for which the prospects of a detection are
investigated in Chatterjee et. al. 2018(in preparation). The Canadian
Hydrogen Intensity Mapping Experiment \citep[CHIME;][]{Bandura2014} and
the Hydrogen Intensity and Real-time Analysis eXperiment \citep[HIRAX;]
[]{Newburgh2016} aims to measure the BAO in the redshift range
$0.8-2.5$. Future experiments, like Tianlai \citep{Chen2012,Chen2015} and SKA1-MID
\citep{Bull2015ApJ} also aim to measure the \HI 21-cm
signal from the post-reionization era.
The Ooty Wide Field Array (OWFA; \citealt{Subrahmanya2016a}) is an
upgraded version of the Ooty Radio Telescope
(ORT, \citealt{Swarup1971}). The upgrade will result in two
concurrently functioning modes named Phase I (PI) and Phase II
(PII). The primary science goals of OWFA have
been outlined in \citet{Subrahmanya2016b}, and the measurement of the
z = 3.35 post-reionization \HI 21-cm power spectrum is one of its
major objectives \citep{Ali2014}. It has been predicted
that a $5\sigma$ detection of the amplitude of
the \HI 21-cm power spectrum is possible with $\sim 150$ hrs of
observation \citep{Bharadwaj2015}. Further, \citet{Sarkar2016b} have
predicted that a $\sim 5\sigma$ measurement of the binned \HI 21-cm
power spectrum is possible in the $k$-range $0.05 \, {\rm Mpc}^{-1} \leq k \leq 0.3 \,
{\rm Mpc}^{-1}$ with $1,000$ hrs of observation.
The primary observable quantity for radio
telescopes are the visibilities, which are the correlation of the
measured voltages at each antenna. It is possible
to directly estimate the redshifted \HI 21-cm power spectrum from the visibilities
measured by a low frequency radio interferometric array
\citep{Bharadwaj2001a,Bharadwaj2005}. \citet{Ali2014} presents
theoretical predictions for the two-visibility correlations {\it i.e.} the correlation between
the visibilities measured at two baselines and two frequencies, expected
at OWFA considering both the \HI $21$-cm signal and also various
foreground components. \citet{Chatterjee2017}
and \citet{Marthi2016a} have used numerical simulations to
respectively predict the $21$-cm signal and the various
foreground contributions to the two-visibility correlations expected at
OWFA. Considering OWFA, \citet{Sarkar2016b} presents an analytic
technique to simulate the expected $21$-cm visibility signal.
The common assumption of all the earlier works mentioned here
is that the field of view (FoV) of the telescope is sufficiently
small so that the observed sky can be assumed to be
flat. In fact the flat sky approximation (FSA) is an underlying
assumption in a large fraction of the works related to
measuring the cosmological $21$-cm power spectrum
(e.g. \citealt{Morales2004}, \citealt{Ali2015}).
Considering OWFA (both PI and PII), the FoV covers $1.8^{\circ}$ in
the E-W direction while this is $4.8^{\circ}$ and $28.6^{\circ}$
in the N-S direction for PI and PII respectively. We expect the
FSA to be a reasonably good approximation for PI, however for PII the
large N-S extent of the FoV brings to question the validity of this
assumption. It is desirable to consider the spherical nature of the
sky in making predictions for PII.
\citet{Shaw2014} have introduced the ``m-mode'' formalism which
incorporates the spherical nature of the sky. This essentially deals
with drift scan observations, however their results can also be used
to predict the signal for observations where the telescope tracks a
fixed region of the sky. \citet{Liu2016} have introduced a spherical
Fourier-Bessel technique for analysing the 21-cm power spectrum,
that incorporates the spherical nature of the sky.
\citet{Zhang2016a} and \citet{Ghosh2017}
present sky map reconstruction methods based on the spherical
harmonics (SH) transformation.
In this work, we develop a spherical sky formalism which is
particularly suited for telescopes like OWFA, where the baselines are
all coplanar with the antenna aperture. The formalism relates the
visibilities to the SH coefficients of the sky signal
through the ``beam transfer function''. Section~\ref{sec:OWFA} of this
paper presents a brief overview of OWFA, whereas the formalism is
presented in Section~\ref{sec:formalism}. In addition to the numerical
evaluation of the beam transfer function, we also use the Limber
approximation (LA) to obtain an analytical expression for the same.
In Section~\ref{sec:vis_cor} we use the formalism to relate the two
visibility correlations to the multi-frequency angular power
spectrum (MAPS, \citealt{Datta2007}) which quantifies the statistics
of the 21-cm signal. In Section~\ref{sec:simulation} we validate our
formalism using all-sky simulations. We study the differences
between the FSA and the formalism of this paper in order to
quantify how important the spherical nature of the sky is for the
different phases of OWFA. The Results are presented in
Section~\ref{sec:results} and we present Summary and
Conclusion in Section~\ref{sec:discussion}.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5
\section{OWFA} \label{sec:OWFA}
\begin{figure*}
\psfrag{Equatorial}{blah} \psfrag{xx}{$x$} \psfrag{yy}{$y$}
\psfrag{zz}{$z$} \psfrag{bb}{$b$} \psfrag{dd}{$d$}
\psfrag{mm}{$\m$} \psfrag{nn}{$\hat{\mathbfit{n}}$}
\psfrag{rrp}{$\mathbfit{r}'$} \centering
%\begin{minipage}{190mm}
%\subfigure[OWFA aperture]{\label{fig:OWFA_aperture}
\includegraphics[scale=0.4]{OWFA_aperture.eps}%}
%\includegraphics[scale=0.35,angle=0]{figure_pbeam.eps}}
%\hskip 7mm
%\subfigure[Primary beam pattern for PII]{\label{fig:pb-PII}
%\includegraphics[scale=0.35,angle=0]{figure_pbeam.eps}}
%\end{minipage}
\caption{This shows the aperture of one of the OWFA antennas. The unit
vector $\m$ is the pointing direction of the antenna
beam, $\hat{\mathbfit{n}}$ points towards an arbitrary direction and
$\mathbfit{r}'$ refers to a displacement on the antenna aperture. The
$x$ and $y$ axes are along the N-S and E-W directions respectively,
while the $z$ axis, which is normal to the aperture, points towards
the celestial Equator ($\delta=0$).}
\label{fig:OWFA_aperture}
\end{figure*}
The Ooty Radio Telescope (ORT) is a $530 \, {\rm m}$ long (N-S) and $30
\, {\rm m}$ wide (E-W) offset-parabolic cylinder, operating at a nominal
frequency of $326.5 \, {\rm MHz}$. The telescope is equatorially
mounted, i.e. the long axis of the reflecting cylinder is in the
North-South direction, parallel to the Earth's rotation axis. The
telescope can be mechanically steered along the East-West direction
by a single rotation along the axis of the cylinder. The feed
consists of 1056 North-South dipoles of length $0.5 \lambda$ arranged
nearly end to end along the focal line of the reflector. The
telescope has a $530 \, {\rm m} \times 30 \, {\rm m}$ rectangular aperture.
Currently this telescope is being upgraded \citep{Subrahmanya2016a} to operate as an
interferometer the Ooty Wide Field Array (OWFA). The upgrade will
result in two concurrent modes namely OWFA PI and PII
respectively. In OWFA PI, the signals from $N_d=24$ adjacent dipoles
will be combined to form a single antenna. Each antenna has a
rectangular aperture of dimension $b \times d$, where $b = 30 \, {\rm
m}$ and $d = 11.5 \, {\rm m}$ respectively. The array will have $N_A =
40$ such antennas separated by $d = 11.5 \, {\rm m}$, arranged along the
North-South axis of the cylinder. In OWFA PII, the signals from $N_d=4$ adjacent dipoles will be
combined to form a single antenna. Each antenna has a rectangular
aperture of dimension $b \times d$, where $b = 30 \, {\rm m}$ and $d =
1.92 \, {\rm m}$ respectively. The array will have $N_A = 264$ such
antennas separated by $d = 1.92 \, {\rm m}$ arranged along the
North-South axis of the cylinder.
Figure~\ref{fig:OWFA_aperture}
shows a schematic diagram of the aperture for one of the OWFA
antennas. For both PI and PII this is a rectangle of dimensions $b
\times d$ with $b = 30 \, {\rm m}$, however as discussed earlier the
dimension $d$ is different for the two modes. In both the modes, the
digitised signals from $N_d$ successive dipoles are combined to form
the antenna beam. This allows the pointing direction
$\m$ of the antenna beam pattern (Figure
\ref{fig:OWFA_aperture}) to be steered electronically along the
North-South direction by introducing phases. The antenna primary beam
pattern $A\left(\Delta \hat{\mathbfit{n}},\nu \right)$ quantifies how
the individual antenna responds to the signal from different
directions $\hat{\mathbfit{n}}$ on the sky (Figure
\ref{fig:OWFA_aperture}), here $\Delta \n =
\n -\m$. It is possible to calculate
$A\left(\Delta \hat{\mathbfit{n}},\nu
\right)$ \citep[e.g][]{Chengalur2007} using,
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%Equation starts%%%%%%%%%%%%%%%%%%%%%%%%
\begin{equation}
A\left(\Delta \hat{\mathbfit{n}},\nu \right) = \int d^2 \U^{'} \;
e^{2 \pi i \U^{'}\cdot \Delta \hat{\mathbfit{n}} }\;
\tilde{a}\left(\U^{'}, \nu \right),
\label{eq:ft_aperture}
\end{equation}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%Equation ends%%%%%%%%%%%%%%%%%%%%%%%%
where, $\U^{'} = \mathbfit{r}'/\lambda$ refers to displacement
$\mathbfit{r}'$ on the antenna aperture and the aperture power pattern
$\tilde{a}\left(\U^{'}, \nu \right)$ is the auto-convolution of
the electric field pattern at the antenna aperture (Figure
\ref{fig:OWFA_aperture}). It is useful to note that the phase
factor $e^{2 \pi i \U^{'}\cdot \Delta \hat{\mathbfit{n}} }$ in
eq.~(\ref{eq:ft_aperture}) is invariant under a mirror reflection
with respect to the aperture plane. It follows that eq.~(\ref{eq:ft_aperture})
predicts a primary beam pattern which is exactly identical in the
upper hemisphere $(UH)$ and lower
hemisphere $(LH)$ of the sky. The telescope, however, only responds
to the $UH$ and the $LH$ is not accessible to
the telescope. Here we have exploited
the fact that eq.~(\ref{eq:ft_aperture}) predicts an
identical beam pattern in both the $UH$ and the $LH$ to simplify the
mathematical analysis in subsequent parts of this paper.
For the purpose of the present analysis we make the simplifying
assumption that the aperture is uniformly illuminated such that the
electric field is uniform everywhere on the $b\times d$ rectangular
aperture of the OWFA antenna. In this case the OWFA the aperture power
pattern can be expressed as \citep{Ali2014},
\begin{equation}
\tilde{a}\left(\U^{'}, \nu \right) = \frac{\lambda^2 }{bd}
\Lambda \left(\frac{U_x^{'}}{d} \lambda \right) \Lambda
\left(\frac{U_y^{'}}{b} \lambda \right) \, ,
\label{eq:OWFA_ap}
\end{equation}
%--------------------------
where $\U^{'} = (U_x^{'},U_y^{'})$, $\Lambda(\xi)$ is the triangular
function defined as, $\Lambda(\xi) = 1-|\xi|$ for $|\xi|<1$ and
$\Lambda(\xi) = 0$ for $|\xi|\geq 1 $ \citep[see Figure (2a)
of][]{Ali2014}.
We see that for OWFA, $\tilde{a}\left(\U^{'}, \nu \right)$ peaks
at $\U^{'} = 0$ and falls off as $\U^{'}$ is
increased. Further this has compact support and
$\tilde{a}\left(\U^{'}, \nu \right) = 0$ if $\U^{'}$ exceeds
the aperture dimensions. Note that these properties of
$\tilde{a}\left(\U^{'}, \nu \right)$ are not particular to OWFA
alone. In general, for any antenna, we expect
$\tilde{a}\left(\U^{'}, \nu \right)$ to peak at $\U^{'} =
0$, fall off as $\U^{'}$ is increased and to have compact support
whereby $\tilde{a}\left(\U^{'}, \nu \right) = 0$ if $\U^{'}$
exceeds the aperture dimensions.
\begin{figure*}
\psfrag{Equatorial}{blah} \centering
\begin{minipage}{190mm}
\subfigure[Primary beam pattern for PI]{\label{fig:pbeam-PI}
\includegraphics[scale=0.35,angle=0]{figure_phaseI.eps}}
\hskip 7mm \subfigure[Primary beam pattern for
PII]{\label{fig:pbeam-PII}
\includegraphics[scale=0.35,angle=0]{figure_phaseII.eps}}
\end{minipage}
\caption{The left and right panels show the primary beam pattern
on the visible upper hemisphere for OWFA PI and PII respectively. }
\label{fig:pbeam}
\end{figure*}
Using equation (\ref{eq:ft_aperture}) to calculate the OWFA primary
beam pattern $A(\dn,\nu)$, we get the product of two
sinc-squared functions
\begin{equation}
A(\dn,\nu)= {\rm sinc}^2 \left( \frac{\pi b \nu \,
\Delta n_y}{c} \right) {\rm sinc}^2 \left( \frac{\pi d \nu \,
\Delta n_x}{c} \right)
\label{eq:pb}
\end{equation}
where $\Delta n_y$ and $\Delta n_x$ are respectively the $y$ and $x$
components of $\dn$ \citep{Marthi2016a}.
Figure~\ref{fig:pbeam} shows the primary beam pattern for OWFA PI and
PII as a function of the celestial coordinates $(\alpha,\delta)$ on
the visible upper hemisphere of the sky. Here the telescope beam is assumed
to point towards $\m =(\alpha,\delta) = (0,0)$. The OWFA
beam is the diffraction pattern of a rectangular slit of dimension $b
\times d$. The beam pattern is normalized to $ A(\Delta
\mathbfit{n},\nu)=1$ at the maxima which occurs when $\Delta
\mathbfit{n}=0$ ({\it i.e.} $\mathbfit{n} = \m$), and $ A(\Delta
\mathbfit{n},\nu)$ falls off as $\dn$ is increased. The
angular extent of the main lobe is determined by the first null which
occurs at $\alpha_0 \approx \lambda / b= \pm 1.8^{\circ}$ along
$\alpha$ for both PI and PII, and at $\delta_0 \approx \pm4.8^{\circ}$
and $\pm 28.6^{\circ}$ along $\delta$ for PI and PII respectively.
After the first nulls, there are side lobes in the OWFA primary beam
pattern. The maxima of the $1^{st}$ and $2^{nd}$ side lobes are
respectively $5\%$ and $2\%$ of the maxima of the main lobe.
The Fourier relation (eq.~\ref{eq:ft_aperture}) implies that the
primary beam pattern gets wider if the antenna aperture dimensions are
reduced. Here, we see that for PI the angular extent of the main lobe,
which defines the antenna's FoV, is small and
restricted within a few degrees on the sky. For this small region
($\theta \ll 1$ in radians) it is a reasonably good approximation to
neglect the curvature of the sky, and it is adequate to carry out any
further analysis for OWFA PI using the FSA
\citep[e.g.][]{Marthi2016a}. For PII, however, we find that the FoV
extends across $57.2^{\circ}$ in the N-S direction. It is not possible
to ignore the curvature of the sky over such large angular extent, and
it is necessary to consider the spherical sky for any further analysis
for PII. Considering a radio-interferometer like OWFA, in the next
Section we develop a formalism to analyze the measured visibilities in
terms of a SH expansion which provide a natural basis for the
signal on a spherical sky.
\section{Visibilities and beam transfer functions} \label{sec:formalism}
\begin{figure*}
\centering \psfrag{nn}{$\hat{\mathbfit{n}}$}
\psfrag{mm}{$\m$} \psfrag{uu}{$\mathbfit{r}$}
\psfrag{dd}{$\mathbfit{r}'$} \psfrag{hh}{$b$} \psfrag{gg}{$d$}
\includegraphics[scale=0.4]{aperture1.eps}
\caption{This shows a pair of antennas in a radio
interferometer. Here $\mathbfit{r}$ is the antenna separation while
$\m$, $\hat{\mathbfit{n}}$ and $\mathbfit{r}'$ are as defined in Figure
\ref{fig:OWFA_aperture}.}
\label{fig:aperture}
\end{figure*}
The complex visibilities $\vis$ are the quantities measured in a radio
interferometric array. Here the baseline $\U$ refers to a pair of
antennas separated by $\mathbfit{r}$ (Figure \ref{fig:aperture}), and
$\U =\mathbfit{r}/\lambda$ is the antenna separation in units of
the observing wavelength. The baselines available at OWFA are
%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{equation}
\U_n = n \frac{d}{\lambda} \hat{\mathbfit{i}} \quad \quad \left(
1 \leq n \leq N_A - 1\right)
\label{eq:bline}
\end{equation}
where the values of $N_A$ and $d$ are different for PI and PII
respectively, and the $(x,y,z)$ coordinate system is as shown in
Figure~\ref{fig:OWFA_aperture}.
We then have the visibility given by \citep{Perley1989}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%Equation starts%%%%%%%%%%%%%%%%%%%%%%%%
\begin{equation}
\vis = \dBdTnu \int_{UH} d\Omega_{\n}
T\left(\n,\nu \right) A\left(\dn,\nu
\right) e^{-2 \pi i \U\cdot \dn },
\label{eq:vis_eq1}
\end{equation}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%Equation ends%%%%%%%%%%%%%%%%%%%%%%%%
where, $\dBdTnu = 2 k_B / \lambda^2$ is the conversion factor from
brightness temperature to specific intensity in the Raleigh - Jeans
limit, $T\left(\hat{\mathbfit{n}},\nu \right)$ is the brightness
temperature distribution on the sky and $d\Omega_{\n}$ is
the elemental solid angle in the direction $\n$.
For the $d\Omega_{\n}$ integral, we use $UH,LH$ and $S$ to
denote the upper hemisphere, the lower hemisphere and the entire
celestial sphere respectively. Note that the
integral here is over the visible upper hemisphere of the sky.
The entire analysis in the rest of
this sub-section is restricted to a single frequency, and in much
of the subsequent text
we do not show $\nu$ explicitly.
For a telescope with a sufficiently large aperture, the primary beam
pattern $\pbnonu$ falls off rapidly as $\dn$ is increased. This restricts
the FoV to a small
region of the sky. Considering the small FoV, we can adopt the FSA
whereby $\dn=\boldsymbol{\theta}$ is a 2D vector on
the plane of the sky with $ d\Omega_{\n} =
d^2\boldsymbol{\theta}$. It is now natural to interpret $\visnonu$
(eq.~\ref{eq:vis_eq1}) as the Fourier transform of the product
$T\left(\boldsymbol{\theta} \right) A\left(\boldsymbol{\theta}
\right)$. Here, the visibility can be expressed as a convolution \citep{Ali2014}
\begin{equation}
\visnonu = \dBdTnonu \int d^2 \U^{'} \, \apnonu \, \tilde{T}(\U^{'}) \, ,
\label{eq:vis_eq1a}
\end{equation}
%--------------------------------------------
where $\tilde{T}(\U^{'})$ is the Fourier transform of the
brightness temperature distribution on the sky. We may interpret
$\visnonu$ as the weighted sum of different Fourier modes
$\tilde{T}(\U^{'})$ with the weights being given by
the shifted aperture power pattern $\apnonu$. The contribution peaks at $\U^{'}
= \U$ where the value of $\apnonu$ is maximum. This tells us that we may
associate each visibility $\visnonu$ with a particular Fourier component
$\tilde{T}(\U)$ of the signal. To be more precise, the
sum extends over a width $ \vert \U -\U^{'} \vert$ which
is of the order of the dimensions of the antenna aperture in units of
the observing wavelength. As discussed earlier,
$\apnonu$ is zero beyond
the extent of the antenna aperture.
The FSA breaks down for the antennas with a small aperture which have
a large FoV. Here, one has to consider the spherical nature of the sky.
The SH function
$Y_{\ell}^m(\hat{\mathbfit{n}})$ provide a natural basis
for the signal on a spherical sky, and in the SH expansion we have
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%Equation starts%%%%%%%%%%%%%%%%%%%%%%%%
\begin{equation}
T\left(\n,\nu \right) = \sum _{\ell,m}
a_{\ell}^m\left(\nu\right) \; Y_{\ell}^m(\n),
\label{spi2sph}
\end{equation}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%Equation ends%%%%%%%%%%%%%%%%%%%%%%%%
where $\alm$ are the SH
coefficients of the brightness temperature distribution.
As discussed earlier by \cite{Shaw2014, Shaw2015} \citep[and also
by][]{Zheng2014, Zhang2016a, Zhang2016b, Liu2016}, the visibilities
can be expressed in terms of the SH coefficients
$\alm$ as,
\begin{equation}
\visnonu = \sum_{\ell,m} \, \alm \,\Blmnonu .
\label{eq:vis_sph_1}
\end{equation}
The visibility expressed in eq. (\ref{eq:vis_sph_1}) is a weighted sum
of the SH coefficients $\alm$ with the weights being given by
the beam transfer function
$\Blmnonu$. Note that this is in exact analogy with
eq. (\ref{eq:vis_eq1a}) with the difference that in the FSA we use Fourier
modes $\tilde{T}(\U)$ instead of the SH
coefficients $\alm$, and we have the shifted aperture
power pattern $\apnonu$ as the weights
instead of $\Blmnonu$. This analogy between
eqs.~(\ref{eq:vis_sph_1}) and (\ref{eq:vis_eq1a}) leads to a
picture where we interpret $\Blmnonu$ as the
spherical sky generalization of the shifted aperture power pattern
$\apnonu$.
The beam transfer function $\Blmnonu$ which quantifies
how a visibility $\visnonu$ responds to a particular SH
coefficient $\alm$ can be calculated \citep{Shaw2014}
using
\begin{equation}
\Blmnonu = \dBdTnonu \int_{UH} d\Omega_{\n} \Ylm \pbnonu e^{-2 \pi i \U\cdot \Delta
\n} \, .
\label{eq:blm1a}
\end{equation}
The integrand here contains two highly oscillatory
functions ($\Ylm$ and $ e^{-2 \pi i \U\cdot \Delta \n}$),
and the integral extends over the visible upper
hemisphere of the sky. It is computationally expensive and cumbersome to
evaluate such integrals, particularly for large values of $\ell$ and
$\U$. The major contribution to the integral comes
from the main lobe of the primary beam pattern, and one may consider
using this to restrict the integral. However the main lobe itself extends
over a large region of the sky for telescopes with a small aperture.
Further, the sidelobes also contribute to the integral, and it is
necessary to include the whole hemisphere to take this into account.
Finally, we note that eq.~(\ref{eq:blm1}) provides very little insight
into the behaviour of $\Blmnonu$ {\it i.e.} at which $\ell,m$ values
we have the maximum contribution for a particular baseline $\U$.
In this paper we develop a different formalism for calculating the
beam transfer function $\Blmnonu$. As noted earlier, it is
natural to interpret $\Blm$ as the spherical sky counterpart of the
shifted aperture power pattern $\apnonu$,
and in this paper we express $\Blmnonu$ as an integral of the aperture power
pattern $\apnonu$. The $d\Omega_{\n}$ integrals in
eqs.~(\ref{eq:vis_eq1}) and (\ref{eq:blm1}) are limited
to the visible upper hemisphere $(UH)$. The SH functions cease
to be orthonormal when the domain is restricted to a hemisphere. The
analysis is considerably simplified if we assume that the primary
beam pattern and also the sky signal
are both replicated on the lower hemisphere $(LH)$ by a reflection
with respect to the plane of the antenna aperture (the $xy$ plane in
Figure \ref{fig:OWFA_aperture}). The domain of
the $d\Omega_{\n}$ integrals are now extended to the
entire sphere $(S)$.
In this situation, the upper and lower hemispheres
both make an equal contribution to the visibility signal. We account
for this by introducing a factor of $1/2$ in the beam transfer
function which is now defined as
\begin{equation}
\Blmnonu = \frac{\dBdTnonu}{2} \int_{S} d\Omega_{\n} \Ylm \pbnonu
e^{-2 \pi i \U\cdot \Delta \n} \, .
\label{eq:blm1}
\end{equation}
The fact that $\Ylm \rightarrow (-1)^{\ell+m} \Ylm$ under
a mirror reflection with respect to the $xy$ plane implies that
$\Blmnonu=0$ when $\ell+m$ is odd, and $\Blmnonu$ is non-zero only
when $\ell+m$ is even.
It is necessary to note that $\alm$ in eq.~(\ref{eq:vis_sph_1}) now
refers to the SH expansion of the replicated
brightness temperature signal $T(\n)$ {\it i.e.} the signal in the
$LH$ is a mirror reflection of the signal in the visible $UH$, the
reflection here is with respect to the plane of the antenna aperture.
We proceed by expressing $\pbnonu$ in
eq.~(\ref{eq:blm1}) in terms of the aperture power pattern
$\apnonu$ using eq.~(\ref{eq:ft_aperture})
which gives us
\begin{equation}
\BlmRnonu = \frac{\dBdTnonu}{2} \int_S d\Omega_{\n} \Ylm e^{-2 \pi i \U\cdot \Delta
\n } \int d^2 \U^{'} \; e^{2 \pi i \U^{'}\cdot \Delta
\hat{\mathbfit{n}} }\; \apnonu \, .
\label{eq:vis_sph_2}
\end{equation}
Note that the $d\Omega_{\n}$ integral in eq.~(\ref{eq:vis_sph_2})
is over the entire sphere.
The subsequent analysis is considerably simplified if we assume that
the baselines $\U$ are coplanar with the antenna aperture. This
assumption is true for OWFA (Figure \ref{fig:aperture}). We also
expect this to be a good approximation for any compact array of fixed
antennas which point vertically overhead. Under this assumption,
$\U$ and $\U^{'}$ in eq.~(\ref{eq:vis_sph_2}) are co-planer vectors, and
we can write
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%Equation starts%%%%%%%%%%%%%%%%%%%%%%%%
\begin{equation}
\BlmRnonu =\frac{\dBdTnonu}{2} \int d^2 \U^{'} \; \apnonu \left[ \int_S
d\Omega_{\n} \;
\Ylm e^{2 \pi i \U^{'}\cdot \Delta \n }
\right].
\label{eq:blm1a}
\end{equation}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%Equation ends%%%%%%%%%%%%%%%%%%%%%%%%
We use the identity
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%Equation starts%%%%%%%%%%%%%%%%%%%%%%%%
\begin{equation}
e^{2 \pi i \U^{'}\cdot \n} = 4\pi \sum_{\ell,m}
i^{\ell} \; Y_{\ell}^m (\hat{\mathbfit{U}}^{'}) \; Y_{\ell}^{m \ast}
(\n) \; j_{\ell}( 2 \pi |\U^{'}| )
\label{eq:exp2sph}
\end{equation}
to evaluate the integral in the square brackets of
eq.~(\ref{eq:blm1a}), here $j_{\ell}( 2 \pi |\U^{'}| )$ is
the $\ell^{th}$ order
spherical Bessel function of first kind. We finally obtain
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%Equation starts%%%%%%%%%%%%%%%%%%%%%%%%
\begin{equation}
\BlmRnonu = 2 \pi i^{\ell} \, \dBdTnonu \int d^2 \U^{'} \; \apnonu
Y_{\ell}^m (\hat{\mathbfit{U}}^{'}) j_{\ell}( 2 \pi
|\U^{'}| )\; e^{- 2 \pi i \U^{'}\cdot \m} \, ,
\label{eq:blm2}
\end{equation}
which we use to compute the beam transfer function $\BlmRnonu$.
The integrand here contains three highly oscillatory
functions ($Y_{\ell}^m (\hat{\mathbfit{U}}^{'})$,
$j_{\ell}( 2 \pi |\U^{'}| )$ and $e^{- 2 \pi i \U^{'}\cdot
\m}$), however in contrast to eq.~(\ref{eq:blm1}), the domain of the
integral is restricted by $\apnonu$ which has compact
support. As mentioned earlier, $\apnonu$
is zero if $\U -\U^{'}$ exceeds the aperture dimensions in units of the
observing wavelength. This significantly reduces the computation
for calculating $\BlmRnonu$, particularly for small apertures where the
primary beam pattern covers a large region of the sky.
The subsequent analysis is restricted to the situation where the
pointing direction $\m$ of the antenna beam pattern is perpendicular
to the antenna aperture so that we have $e^{- 2 \pi i \U^{'}\cdot \m}=1$
for the phase factor in eq.~(\ref{eq:blm2}). We adopt the coordinate
system shown in Figure \ref{fig:OWFA_aperture} with the
baseline $\U$ (Figure \ref{fig:aperture}) aligned with $x$
axis. Expressing eq.~(\ref{eq:blm2}) in spherical polar coordinates,
the vector $\U^{'}$ is restricted to the $\theta = \frac{\pi}{2}$
plane and we then have $Y_{\ell}^m(\hat{\mathbfit{U}}^{'}) =
Y_{\ell}^m(\pi/2,0) \, e^{i m \phi}$.
So the beam transfer function $\BlmRnonu$ can be written as
\begin{equation}
\BlmRnonu = 2 \pi i^{\ell}\, Y_{\ell}^m (\frac{\pi}{2},0)
\, \dBdTnonu \int dU^{'} \, U^{'}\, j_{\ell}( 2 \pi
U^{'} ) \, \int d\phi \, \tilde{a}\left(U- U^{'} \cos \phi,
U^{'} \sin \phi \right) e^{i m \phi} \, ,
\label{eq:blm4}
\end{equation}
where the domain of integration is shown in Figure
\ref{fig:base_plane}.
The integral here contains two highly oscillatory
functions $j_{\ell}( 2 \pi U^{'} )$ and $e^{i m \phi}$.
Considering the complex conjugate $[\BlmRnonu]^{*}$, the fact that
$Y^{m*}_l(\theta,\phi)=(-1)^m \, Y^{-m}_l(\theta,\phi)$ implies that
$[\BlmRnonu]^{*}={B_{\ell}^{-m}\left(\mathbfit{U} \right)}$.
If we further assume that the aperture
power pattern is symmetric with respect to $U_y$ {\it i.e.}
$\tilde{a}(U_x,-U_y)=\tilde{a}(U_x,U_y)$ (which is true for OWFA) we
have $[\BlmRnonu]^{*}=(-1)^{\ell} \BlmRnonu$. This implies that
for even $\ell$ we have the non-zero elements
$B_{\ell}^{0}(\U),B_{\ell}^{2}(\U)=B_{\ell}^{-2}(\U),
B_{\ell}^{4}(\U)=B_{\ell}^{-4}(\U),...,B_{\ell}^{\ell}(\U)=B_{\ell}^{-\ell}(\U)$
which are all real.
For odd $\ell$ we have the non-zero elements
$B_{\ell}^{1}(\U)=-B_{\ell}^{-1}(\U),
B_{\ell}^{3}(\U)=-B_{\ell}^{-3}(\U),...,B_{\ell}^{\ell-1}(\U)=-B_{\ell}^{-\ell+1}(\U)$
which are all imaginary. The other elements of
$B_{\ell}^{m}(\U)$ are all zero.
\begin{figure*}
\centering
\psfrag{AA}{$\apnonu$}
\psfrag{phi}{$\phi$}
\psfrag{uxx}{$U_x$}
\psfrag{uyy}{$U_y$}
\psfrag{uuup}{$\mathbfit{U}'$}
\psfrag{uuu}{$\mathbfit{U}$}
\psfrag{bb}{$2\,b$}
\psfrag{dd}{$2\,d$}
\includegraphics[scale=0.4]{baseline-app.eps}
\caption{This shows the domain of integration in eq.~(\ref{eq:blm4}).
Here $\U$ corresponds to the baseline, $\U^{'}$ and $\phi$ can vary with
in the aperture of the antenna (shown by the rectangle). $\phi$ becomes
significantly small for large baseline $|\U|>>d/\lambda$. The $U_x - U_y$
plane here corresponds to the $\theta = \pi / 2$ for the previously adopted
spherical polar coordinate system.}
\label{fig:base_plane}
\end{figure*}
It is possible to obtain further insight into the behaviour of
$\BlmRnonu$ if we adopt the LA \citep{Limber1954}
\begin{equation}
j_{\ell} ( 2 \pi U^{'} ) \approx \sqrt{\frac{\pi}{2 \ell
+1}} \delta_D \left( \ell + 1/2 - 2 \pi U^{'} \right).
\label{eq:limber}
\end{equation}
which holds for large $\ell$. Here we also assume that the baseline is
large compared to the antenna aperture dimensions $U \gg b/\lambda$ so
that the $\phi$ range subtended by the aperture
(Figure~\ref{fig:base_plane}) is small whereby $\cos \phi \approx 1$
and $\sin \phi \approx \phi$. We then have
\begin{equation}
\BlmRnonu = i^{\ell}\, \sqrt{\frac{2 \ell+1}{16 \pi}}
Y_{\ell}^m (\frac{\pi}{2},0)
\, \dBdTnonu \int d\phi \, \tilde{a}\left(U- \frac{2 \ell+1}{4 \pi},
\frac{2 \ell+1}{4 \pi} \phi \right) e^{i m \phi} \, .
\label{eq:blm4b}
\end{equation}
For OWFA we can decompose the primary beam pattern (eq.~\ref{eq:pb}) as
$A\left(\Delta \hat{\mathbfit{n}}\right)=A_x(\Delta n_x) A_y(\Delta
n_y)$ and the aperture power pattern (eq.~\ref{eq:OWFA_ap}) as
$ \tilde{a}\left(\U^{'}\right)=\tilde{a}_x(U_x^{'})
\tilde{a}_y(U^{'}_y)$, where $A_x(\Delta n_x)$ and $A_y(\Delta
n_y)$ are respectively the Fourier transforms
(eq.~\ref{eq:ft_aperture}) of $\tilde{a}_x(U_x^{'})$ and
$\tilde{a}_y(U^{'}_y)$. It is then possible to analytically evaluate
the $\phi$ integral in eq.~(\ref{eq:blm4b}) and we have
\begin{equation}
\BlmRnonu = i^{\ell} \, \dBdTnonu\sqrt{\frac{\pi}{2 \ell +1}}
Y_{\ell}^m (\frac{\pi}{2},0) \tilde{a}_x\left(\U -\frac{2
\ell +1}{4 \pi}\right) A_y\left( \frac{m}{2 \ell + 1}\right) \, .
\label{eq:blm5}
\end{equation}
where $\tilde{a}_x(U_x^{'}) =(\lambda/d)
\Lambda(U_x^{'} \lambda / d)$ and $A_y(\Delta n_y)={\rm sinc}^2(\pi b
\Delta n_y/\lambda)$ for OWFA.
%%%%%%%%%%%%%%%%%%% New addition Ends %%%%%%%%%%%%%%%%%%%%%%
It is interesting to note that, in contrast to eq.~(\ref{eq:blm1}), the
alternate expression (eq.~\ref{eq:blm2}) obtained here provides useful
insight into the behaviour of the beam transfer function $\BlmRnonu$. We
see that the integrand in eq.~(\ref{eq:blm2}) is the product of
$\apnonu$, which is peaked at $\U=\U^{'}$,
and $j_{\ell}( 2 \pi |\U^{'}| )$, which is peaked at $\ell= 2
\pi|\U^{'}|$. Based on this we can infer that $\BlmRnonu$ has a maximum value
when $\ell= 2 \pi|\U|$, and the value of $\BlmRnonu$ falls off as the
difference between $\ell$ and $2 \pi|\U|$ is increased.
In other words, the visibility $\mathcal{V}(\U,\nu)$ receives maximum
contribution from the SH coefficients $ \, \alm$
with $\ell=2 \pi |\U|$. The spread $\Delta \ell$ around
this $\ell$ value depends on the aperture power pattern
with $\Delta \ell \sim 2 \pi D/\lambda$ where $D$ represents the
aperture size, {\it i.e.} we have a smaller spread in $\ell$ if the
aperture is small as compared to a large aperture. This behaviour
is explicit in the approximate analytical expression (eq. \ref{eq:blm5})
for $\Blmnonu$ .
%%%%%%%%%%%%%%%%%%% new addition for limber approximation %%%%%%%%%%%%%%%%%
\begin{figure*}
\psfrag{modflm}{$|\Flm|$}
\psfrag{mm}{$m$}
\psfrag{leq500}{$\ell = 500$}
\psfrag{leq1000}{$\ell = 1000$}
\psfrag{leq3000}{$\ell = 3000$}
\centering
\includegraphics[scale=0.6,angle=0]{flm_plot.eps}
\caption{This shows the variation of $|\Flm|$ as a function of $m$ for
the fixed $\ell$ values $500({\rm solid \, line}), 1000({\rm dased
\, line})$ and $3000({\rm dotted \, line})$. }
\label{fig:flm}
\end{figure*}
Eq.~(\ref{eq:blm5}) provides further insight in the behaviour of
$\Blmnonu$. For a fixed $\ell$, it is possible to analyze
the $m$ dependence of $\Blmnonu$ by considering the function
$\Flm = Y_{\ell}^m (\frac{\pi}{2},0) A_y\left(\frac{m}{2 \ell + 1}\right)$
which appears in eq.~(\ref{eq:blm5}).
Figure~\ref{fig:flm} shows the variation
of $|\Flm|$ as a function of $m$ for fixed $\ell$ values (mentioned in
the figure). We find that the $m$ dependence of $|\Flm|$ is very similar to
${\rm sinc}^2(\pi b m/(\lambda(2 \ell+1)))$. We expect the amplitude
of $\Blmnonu$ to be maximum for $m=0$ and decrease with
increasing $m$.
\begin{figure*}
\psfrag{sumwlmmm}{$\,\,\,\, |B_{\ell}^m|$}
\psfrag{l}{$\ell$}
\psfrag{mmmm0}{$m = 0$}
\psfrag{mmmm4}{$m = 4$}
\psfrag{mmmm14}{$m = 14$}
\psfrag{mmmm12}{$m = 12$}
\psfrag{bl5PI}{ $a = 5$}
\psfrag{bl21PI}{ $a = 21$}
\psfrag{bl30PII}{ $a = 30$}
\psfrag{bl60PII}{ $a = 60$}
\psfrag{OWFAPI}{PI}
\psfrag{OWFAPII}{PII}
\centering
\begin{minipage}{190mm}
\subfigure
{\includegraphics[scale=0.6,angle=0]{Blm-plot_pI-1.eps}}
\hskip 7mm \subfigure
{\includegraphics[scale=0.6,angle=0]{Blm-plot_pI-2.eps}}
\end{minipage}
\begin{minipage}{190mm}
\subfigure
{\includegraphics[scale=0.6,angle=0]{Blm-plot_pII-1.eps}}
\hskip 7mm
\subfigure
{\includegraphics[scale=0.6,angle=0]{Blm-plot_pII-2.eps}}
\end{minipage}
\caption{This shows the beam transfer function $\Blma$ as a function
of $\ell$ for fixed values of $\U_a$
with the $a$ and $m$ values as mentioned in the figure.
The top and bottom rows show results for
OWFA PI and PII respectively.
The solid and dashed lines represent $\Blma$
computed using eq.~(\ref{eq:blm2}) and the LA
(eq.~\ref{eq:blm5}) respectively.}
\label{fig:bt_func}
\end{figure*}
Figure~\ref{fig:bt_func} shows the OWFA beam transfer function
$\Blma$ as a function of $\ell$ for a few fixed values of $\U_a$
with the $a$ and $m$ values as mentioned in the figure.
The solid and dashed lines correspond to $\Blma$ computed using
eq.~(\ref{eq:blm2}) and the LA (eq.~\ref{eq:blm5})
respectively.
The top row shows the results for OWFA PI where we see that $|\Blma|$
peaks at $\ell_a =395$ and $1650$ for $a = 5$ and $21$
respectively. This is consistent with $\ell_a \approx 2 \pi a d /
\lambda$ where $d = 11.5 \, {\rm m}$ and $2 \pi d / \lambda \approx 79$.
We also see that, as expected, for all values of $a$ the beam
transfer function extends over a nearly fixed
interval of $\Delta \ell = \pm 79$ around the peak value.
We find that LA works very well for the larger
baseline whereas it is in reasonable agreement for the smaller baseline.
The bottom row of Figure~\ref{fig:bt_func}
shows the results for OWFA PII where we see that $|\Blma|$
peaks at $\ell_a =390$ and $782$ for $a = 30$ and $60$
respectively, which is consistent with $\ell_a \approx 2 \pi a d /
\lambda$ where $d = 1.92 \, {\rm m}$ and $2 \pi d / \lambda \approx 13$.
However, unlike PI, here we see that $|\Blma|$ shows several secondary
peaks (oscillations) after the
first zero to the left of $\ell_a$ and it shows a tail which extends
to $\Delta \ell \approx 40$ to the right of $\ell_a$. These features are
not present in the $\Blma$ computed using LA
and we find that the LA does not work well even
for the large baselines ({\it e.g.} $a = 60$). The oscillations here
are a manifestation of the oscillations in the
Bessel function $j_{\ell}(2 \pi U^{'})$ in eq.~(\ref{eq:blm4}).
For PI the aperture power pattern $\apnonu$ is wider than
the oscillation period of $j_{\ell}(2 \pi U^{'})$ and these
oscillations are averaged out in eq.~(\ref{eq:blm4}). However,
PII has a smaller aperture and consequently these oscillations persist
in $\Blma$.
\section{Visibility Correlation and Angular Power Spectrum} \label{sec:vis_cor}
The interest here is in a statistical detection of the redshifted
cosmological 21-cm signal using radio-interferometric observations.
In this section we consider a more general situation where we wish to
detect an arbitrary statistical sky signal using radio-interferometric
observations. We assume that the brightness temperature distribution
$T\left(\n,\nu \right)$
on the sky is the outcome of a statistically homogeneous and
isotropic random process. Note that, this assumption requires
$T\left(\n,\nu \right)$ to be defined on the entire celestial sphere
$S$. The two point statistics of this signal
can be completely quantified using
the multi-frequency angular power spectrum
(MAPS; \citealt{Datta2007}) which
jointly characterises the angular and the frequency dependence
of this sky signal. The MAPS of the brightness temperature distribution at two
different frequencies $\nu_i$ and $\nu_j$ is defined through
\begin{equation}
\maps =\left\langle a_{\ell}^{m} \left( \nu_i \right) \, a_{\ell}^{m \, *}
\left( \nu_j \right) \right\rangle .
\label{eq:maps_1}
\end{equation}
The measured visibilities $\vis$ arising from this random signal
are also random. We use two-visibility correlations
\citep{Bharadwaj2001a} to observationally quantify the statistical
properties of the input sky signal. The two-visibility correlation
\begin{equation}
\viscor \equiv \langle \mathcal{V}(\U_a,\nu_i) \, \mathcal{V}^*(\U_b,\nu_j) \rangle
\label{eq:viscor_1}
\end{equation}
refers to the correlation in the visibilities measured at two
baselines $\U_a$ and $\U_b$, and frequencies $\nu_i$ and $\nu_j$.
The angular brackets here denote an ensemble average over different random
realization of the sky signal. The two point statistics, or
equivalently the power spectrum (MAPS), of the sky signal is
contained in this two-visibility correlation.
The calculations are fairly simplified in the FSA \citep{Ali2014}
where we have
\begin{equation}
\viscor = \dBdT^2 \int d^2\U^{'} \tilde{a}\left(\U_a -\U^{'}, \nu_i \right)
\tilde{a}\left(\U_b -\U^{'}, \nu_j \right) \mathcal{C}_{2\pi U'}\left( \nu_i, \nu_j \right)\,,
\label{eq:viscor_2}
\end{equation}
which relates the visibility correlation to MAPS $\mathcal{C}_{2\pi U}\left( \nu_i, \nu_j \right)$. This can be obtained by using eq.~(\ref{eq:vis_eq1a})
in eq.~(\ref{eq:viscor_1}), and finally utlizing the
fact \citep{Datta2007} that
\begin{equation}
\langle \tilde{T}(\U, \nu_i) \tilde{T}(\U^{'}, \nu_j) \rangle = \delta_D^2 \left( \U - \U^{'}\right) \mathcal{C}_{2\pi U}\left( \nu_i, \nu_j \right)
\label{eq:maps_2}
\end{equation}
where $\delta_D^2 \left( \U - \U^{'}\right)$ is the $2$D
Dirac delta function. This expression for the visibility correlation
(eq.~\ref{eq:viscor_2}) is adequate for telescopes with small FoV
(e.g. OWFA PI).
It is necessary to consider the SH formalism (eq.~\ref{eq:vis_sph_1})
for telescopes with a large FoV (e.g. OWFA PII). To obtain an
expression for the visibility correlation in the SH formalism,
we revisit the visibilities as discussed earlier (in eq.~\ref{eq:vis_eq1}),
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%Equation starts%%%%%%%%%%%%%%%%%%%%%%%%
\begin{equation}
\visnonu = Q \int_{UH} d\Omega_{\n}
T\left(\n \right) A\left(\dn
\right) e^{-2 \pi i \U\cdot \dn } \, , \nonumber
\label{eq:vis_eq1a}
\end{equation}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%Equation ends%%%%%%%%%%%%%%%%%%%%%%%%
where the $d\Omega_{\n}$ integral is over the visible upper hemisphere $(UH)$.
For our convenience, we introduce the visibility
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%Equation starts%%%%%%%%%%%%%%%%%%%%%%%%
\begin{equation}
\visnonup = Q \int_{LH} d\Omega_{\n}
T\left(\n \right) A\left(\dn
\right) e^{-2 \pi i \U\cdot \dn } \, ,
\label{eq:vis_eq1b}
\end{equation}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%Equation ends%%%%%%%%%%%%%%%%%%%%%%%%
where the $d\Omega_{\n}$ integral is over the lower hemisphere $(LH)$ of the sky.
Here we can safely assume that the major contributions of the sky signal
to $\visnonu$ and $\visnonup$ are statistically uncorrelated which implies
that the $\visnonu$ and $\visnonup$ are statistically uncorrelated {\it i. e.}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{equation}
\left\langle (\visnonu + \visnonup) (\visnonu + \visnonup)^{*} \right\rangle = \left\langle |\visnonu|^2 \right\rangle + \left\langle |\visnonup|^2 \right\rangle \, .
\label{eq:viscor_sph1}
\end{equation}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
We also have,
%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{equation}
\left\langle |\visnonu|^2 \right\rangle = \left\langle |\visnonup|^2 \right\rangle \, .
\label{eq:viscor_sph1a}
\end{equation}
%%%%%%%%%%%%%%%%%%%%%%%%%
Following the formalism discussed in the previous section, the
total visibility can be expanded as,
%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{equation}
\visnonu + \visnonup = \sum_{\ell, m} b_{\ell}^m \,\Blmnonu \, .
\label{eq:vis_sph_3}
\end{equation}
%%%%%%%%%%%%%%%%%%%%%%%%%%%
Here $b_{\ell}^m$ are the SH coefficients of the sky signal
$T(\n)+T_R(\n)$, where $T_R(\n) $ is the $T(\n)$ replicated on the whole
sphere $(S)$ after a reflection with respect to the aperture plane of the
antenna. The statistical properties of $b_{\ell}^m$ can be quantified
using the MAPS of the brightness temperature distribution $T(\n)$
(discussed in eq.~\ref{eq:maps_1}) as,
%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{equation}
\left\langle b_{\ell}^m(\nu_i) \, b_{\ell^{'}}^{m^{'}*}(\nu_j) \right\rangle = 4 \, \mathcal{C}_{\ell}\left( \nu_i, \nu_j \right) \, \delta_{\ell \ell^{'}} \, \delta_{mm^{'}} \, .
\label{eq:cl_2}
\end{equation}
%%%%%%%%%%%%%%%%%%%%%%%%%%%
We then obtain the visibility correlation to be
\begin{equation}
\langle \mathcal{V}(\U_a,\nu_i) \,
\mathcal{V}^*(\U_b,\nu_j) \rangle = \sum_{\ell} \,
\mathcal{C}_{\ell}\left( \nu_i, \nu_j \right) \,
\mathcal{W}_{\ell}\left(\U_a, \nu_i ;\U_b, \nu_j \right) \, ,
\label{eq:viscor_sph}
\end{equation}
where $\mathcal{W}_{\ell}$ is a window function which is defined
as
\begin{equation}
\mathcal{W}_{\ell}\left(\U_a, \nu_i ;\U_b, \nu_j \right)
= 2\, \sum_m \, B_{\ell}^m\left(\U_a, \nu_i \right) \,
B_{\ell}^{m*}\left(\U_b, \nu_j \right).
\label{eq:wgt_1}
\end{equation}
Eqs.~(\ref{eq:viscor_2}) and (\ref{eq:viscor_sph}) both describe how the two-visibility correlations
$\viscor$, which can be measured directly from observations, is related to the statistics
of the sky signal namely MAPS $\maps$. We see that in both cases the two-visibility
correlation is a weighted sum of $\maps$. We have the weights
$\tilde{a}\left(\U_a -\U \right) \tilde{a}\left(\U_b -\U \right)$ in the FSA
(eq.~\ref{eq:viscor_2}) where $\ell= 2 \pi|\U|$. First considering a situation
where $\U_a = \U_b$, the weight $|\tilde{a}\left(\U_a -\U \right)|^2$ peaks
at $\ell= 2 \pi|\U_a|$ and we have a contribution from the range of
multipoles $\Delta \ell \sim \pm 2 \pi D/\lambda$, where $D$ is the dimension
of the antenna aperture, the weight falls to zero beyond this interval. Considering
two different baselines $\U_a \neq \U_b$, the weights are the
product of two functions, one which peaks at $\U = \U_a$ and another
peaks at $\U = \U_b$. The weights
have non-zero values only if there is some overlap between $\tilde{a}\left(\U_a -\U \right)$
and $\tilde{a}\left(\U_b -\U \right)$ {\it i.e.} $|\U_a - \U_b| < 2 \pi D/\lambda$. There is
no overlap and the visibilities are uncorrelated if $|\U_a - \U_b| > 2 \pi D/\lambda$. The
properties of the FSA visibility correlation have been analyzed in detail in several
earlier works \citep{Bharadwaj2001a, Bharadwaj2005}.
Considering the SH formalism (eq.~\ref{eq:viscor_sph}), we see that, here also the
weights are a product of two functions (eq.~\ref{eq:wgt_1}). Recollect that $\Blm$
peaks at $\ell= 2 \pi|\U|$ and has a width $\Delta \ell \sim \pm 2 \pi D / \lambda$
around this value. Considering the weight function for $\U_a = \U_b$ we see that this peaks
at $\ell= 2 \pi|\U_a|$, or in other words, the visibility correlation $\viscora$ essentially
responds to $\maps$ in a range of multipoles $\Delta \ell = \pm 2 \pi D/ \lambda$
peaked around $\ell= 2 \pi|\U_a|$. The fact that the index $m$ does not appear in
eq.~(\ref{eq:viscor_sph}) is a consequence of the assumption that the sky signal is
statistically homogeneous and isotropic on the sky. Next, considering two different
baseline $\U_a \neq \U_b$, here again we see that the expectations are qualitatively
similar to these for the FSA {\it i.e.} the visibilities are correlated only if
$|\U_a - \U_b| < 2 \pi D/\lambda$ and there is no correlation for larger separations.
Although the visibility correlations are expected to be qualitatively similar in both
the FSA and the \SHE, the predicted values are expected to differ. We have quantified
these differences in subsequent sections of this paper.
%%%%%%%%%%%%%%%%%%%%%%%% new addition %%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{The OWFA Window Function $(\mathcal{W}_\ell)$} \label{subsec:wl}
In this subsection we explicitly calculate the window function and discuss its behaviour
for the two modes PI and PII of OWFA. We have used the OWFA aperture power pattern
(eq. \ref{eq:OWFA_ap}) in eq. (\ref{eq:blm2}) to evaluate
$\Blm$. Finally, we have used eq. (\ref{eq:wgt_1}) to compute the
window function $\Wl$. For comparison, we have also used LA
(eq.~\ref{eq:limber}) to calculate $\Blm$ and used
this in eq. (\ref{eq:wgt_1}) to compute the corresponding window
function $\Wl$. Throughout the entire subsequent analysis we have
assumed that
the pointing direction $\m$ is towards the celestial equator {\it
i.e.} normal to the aperture (Figure~\ref{fig:OWFA_aperture})
which results in $e^{- 2 \pi i \U^{'}\cdot \m} = 1$.
\begin{figure*}
\psfrag{sumwlmmm}{$\qquad\qquad W_{\ell}$}
\psfrag{l}{$\qquad\qquad\qquad\qquad\qquad\ell$}
\psfrag{nol}{$\,$}
\psfrag{bl5PI}{{ $a = 5$}}
\psfrag{bl6PI}{{$a = 6$}}
\psfrag{bl21PI}{{ $a = 21$}}
\psfrag{bl22PI}{{$a = 22$}}
\psfrag{ztfz}{{$\bf 350$}}
\psfrag{zfzz}{{ $ \bf 400$}}
\psfrag{zffv}{{ $ \bf 450$}}
\psfrag{zfvzz}{{ $ \bf 500$}}
\psfrag{zfvfvz}{{ $ \bf 550$}}
\psfrag{oszz}{{$\bf 1600$}}
\psfrag{osfz}{{ $ \bf 1650$}}
\psfrag{osvzz}{{ $ \bf 1700$}}
\psfrag{osvfz}{{ $ \bf 1750$}}
\psfrag{oezz}{{ $ \bf 1800$}}
\psfrag{zr}{$0$}
\psfrag{5powm6}{$ \bf 5\times10^{-6}$}
\psfrag{1powm5}{$ \bf 1\times10^{-5}$}
\psfrag{15powm5}{$ \bf 1.5\times10^{-5}$}
\psfrag{2powm5}{$ \bf 2\times10^{-5}$}
\psfrag{bl30PI}{{$a = 30$}}
\psfrag{bl31PI}{{$a = 31$}}
\psfrag{bl60PI}{{$a = 60$}}
\psfrag{bl61PI}{{$a = 61$}}
\psfrag{ztsz}{{$\bf 360$}}
\psfrag{zfzz}{{ $ \bf 400$}}
\psfrag{zffz}{{ $ \bf 440$}}
\psfrag{zsfz}{{ $ \bf 750$}}
\psfrag{zsnz}{{ $ \bf 790$}}
\psfrag{zetz}{{ $ \bf 830$}}
\psfrag{zr}{$0$}
\psfrag{m1powm4}{$ \bf -1\times10^{-4}$}
\psfrag{1powm4}{$ \bf 1\times10^{-4}$}
\psfrag{2powm4}{$ \bf 2\times10^{-4}$}
\psfrag{3powm4}{$ \bf 3\times10^{-4}$}
\psfrag{4powm4}{$ \bf 4\times10^{-4}$}
%\psfrag{}{$ $}
\psfrag{OWFAPI}{PI}
\psfrag{OWFAPII}{PII}
\centering
\begin{minipage}{170mm}
\subfigure[]
{\label{fig:win_func-PI}
\includegraphics[scale=1.2,trim= 0 50 0 50,angle=0]{wlm-plot_pI-1.eps}}
\subfigure[]
{\label{fig:win_func-PII}
\includegraphics[scale=1.2,trim= 0 50 0 50,angle=0]{wlm-plot_pII-1.eps}}
\end{minipage}
\caption{This shows the window functions for OWFA PI and PII. Solid
lines in the top row show the window functions $\Wlsamec$ for the same
baselines of OWFA PI for $a = 5,6,21$ and $22$. The dot-dashed line shows the $\Wlsamec$
evaluated using LA for $a = 21$.
Solid lines in the bottom row show the window functions $\Wlsamec$ for the same
baselines of OWFA PII for $a = 30,31,60$
and $61$. The dashed lines in the figure represents the window functions
$\Wladjc$ for the corresponding adjacent baselines. }
\label{fig:win_func}
\end{figure*}
We first consider the OWFA window function $\Wl$ for a situation where
we have the same baseline $\U_a=\U_b=a d/\lambda$. Based on our
earlier discussion, the OWFA
window function $\Wlsame$ is expected to peak at $\ell_a = 2 \pi a d
/ \lambda$ and have a width of $\Delta \ell = \pm 2 \pi d / \lambda $
around $\ell_a$.
Figure~\ref{fig:win_func} shows the OWFA window function
$\Wl$ as a function of $\ell$ for a few fixed values of $\U_a$
(mentioned in the figure legend), and fixed frequency
$\nu_i = \nu_j = \nu_c = 326.5 \, {\rm MHz}$, which is
the nominal frequency for OWFA.
The top row shows the results for OWFA PI where we see that $\Wlsamec$
peaks at $\ell_a =395,471,1650$ and $1725$ for $a = 5,6,21$ and $22$
respectively which is consistent with $\ell_a \approx 2 \pi a d /
\lambda$ where $d = 11.5 \, {\rm m}$
and $2 \pi d / \lambda \approx 79$. We also see that, as expected,
for all values of $a$ the window function extends over a nearly fixed
interval of $\Delta \ell = \pm 79$ around the peak values.
The dot-dashed line in the right panel of the top row shows the window
function calculated using LA for
$a = 21$. We find that LA works well in the large
baseline regime.
The bottom row shows the results for OWFA PII where we see that $\Wlsamec$
peaks at $\ell_a =390,405,782$ and $797$ for $a = 30,31,60$ and $61$
respectively which is consistent with $\ell_a \approx 2 \pi a d /
\lambda$ where $d = 1.92 \, {\rm m}$ and $2 \pi d / \lambda \approx 13.13$.
Considering the width of the window function, we see that for all
values of $a$, $\Wlsamec$ goes to zero at $\Delta \ell \approx 13$ to
the left of $\ell_a$. However, unlike PI, here we see that
$\Wlsamec$ shows several secondary peaks (oscillations) after the
first zero to the left of $\ell_a$ and it shows a tail which extends
to $\Delta \ell \approx 40$ to the right of $\ell_a$. The oscillatory
features and the extended tail are both a consequence of the spherical
Bessel function $j_{\ell}( 2 \pi |\U^{'}| )$ in
eq.~(\ref{eq:blm2}). For PI these oscillations and the tail
are washed out because of the integral over the broader aperture power
pattern $ \tilde{a}\left(\U_{a} - \U^{'} \right)$.
We next consider the OWFA window function $\Wl$ for a situation where
we have the two different baselines $\U_a \neq \U_b$. The aperture
power patterns $\tilde{a}\left(\U_a - \U \right)$ and
$\tilde{a}\left(\U_{b} - \U \right)$ have an overlap only if
$b=a \pm 1$ {\it i.e.} the overlap is restricted to the adjacent
baselines and does not extend beyond (Figure 2 of \citealt{Ali2014}).
We further expect the window function $\Wladjc$ for the adjacent
baselines to peak at $\ell_a = \pi (2 a \pm 1) d / \lambda$
which corresponds to the average of the two baselines and have a
width $\Delta \ell = \pm \pi d / \lambda $ which is half the width
of $\Wlsamec$. For both PI and PII we find that the peaks of the
window function $\Wladjc$ are located at the expected $\ell$ values,
the width also is consistent with the expectations for PI. However,
for PII we find that $\Wladjc$ oscillates to the left of the peak and
has an extended tail to right. As mentioned earlier, these are
consequences of the spherical Bessel function $j_{\ell}( 2 \pi
|\U^{'}| )$ in eq.(\ref{eq:blm2}). We finally note that, for both
PI and PII there is an $\ell$ range where the window function
$\Wladjc$ become negative, while this is relatively prominent for PII
this feature is also present for PI.
\section{Simulation} \label{sec:simulation}
In the previous section we have developed a formalism which
relates the two-visibility correlations $\viscor$, which can be measured
from the observations, to the statistics of the sky signal namely MAPS $\maps$.
In this section we present simulations that we have carried out to
validate this theoretical formalism. The entire analysis is restricted
to a single frequency $\nu_c = 326.5 \, {\rm MHz}$ and do not show
this explicitly in much of the subsequent text.
We assume that the sky signal $T(\n)$ is a Gaussian random field with an
angular power spectrum $C_{\ell}$ which we have modelled as a power law
\begin{equation}
\mathcal{C}_{\ell} = A \, \left(\frac{\ell}{\ell_0}\right)^n \,,
\label{eq:cl_1}
\end{equation}
arbitrarily normalized to unity ({\it i.e.} $A = 1$) at $\ell =
\ell_0=1$. We have considered the power law index to have value
$n=-2$ in our analysis.
The simulations were carried out using the package HEALPix
\citep[Hierarchical Equal Area isoLatitude Pixelization of a
sphere;][]{Gorski2005}. For OWFA, the largest baseline corresponds to
an angular multipole
$\ell \approx 2 \pi \times 530 \, {\rm m}/\lambda_c=3624$. In the
simulations, we have set $\ell_{max}= 4096$ which corresponds to a
pixel size of $1.712^{'}$. We have used the SYNFAST routine of
HEALPix to generate different statistically independent realization of
the sky signal $T(\n)$ corresponding to the input $C_{\ell}$.
\begin{figure*}
\centering \includegraphics[scale=0.35]{l-2.eps}
\caption{This shows a single realization ofthe
brightness temperature fluctuations used in the simulations.}
\label{fig:l-2}
\end{figure*}
We have computed the OWFA visibilities from the simulated maps using
eq.~(\ref{eq:vis_eq1}). Here, it is convenient to adopt the coordinate
system defined in Figure 2 of \citet{Marthi2016a} whereby
\begin{equation}
\mathcal{V}(\U_a) = Q \, \Delta \Omega_{pix} \, \sum _{p=0}^{N_{pix} - 1} \, T(\alpha_p,
\delta_p) \, A(\alpha_p, \delta_p) e^{-2 \pi i U_a \, (\sin
\delta_p - \sin \delta_0)},
\label{eq:vis_sim1}
\end{equation}
where $\Delta \Omega_{pix}$ refers to the solid angle subtended by
each simulation pixel, $(\alpha_p, \delta_p)$ refers to the (RA, DEC) of
the $p$-th pixel, and $(\alpha_0, \delta_0)$ refers to the pointing
direction.
The sum here runs over all the pixels ($N_{pix}$ in number)
in the simulation.
The OWFA primary beam pattern
$A(\Delta \n)$ eq.~(\ref{eq:pb}) can be expressed as
\begin{equation}
A(\alpha_p, \delta_p) = {\rm sinc}^2 \left( \frac{\pi b \nu_c }{c} \,
{\rm cos} \delta_p \, {\rm sin} \left(\alpha_p - \alpha_0 \right)
\right) {\rm sinc}^2 \left( \frac{\pi d \nu_c}{c} \left({\rm sin}
\delta_p - {\rm sin} \delta_0\right) \right).
\label{eq:pb_sim}
\end{equation}
As mentioned earlier, we have used $(\alpha_0, \delta_0)=(0, 0)$
for the simulations presented here.
Figure \ref{fig:l-2} shows the brightness temperature fluctuations
corresponding to a particular realization of the simulations.
We have used an ensemble of $100$ independent realizations of the
simulated sky signal to estimate the visibility correlation
$\viscornonu$ and its variance.
\section{Results} \label{sec:results}
\begin{figure*}
\psfrag{V2V2UaUb}{$\viscornonu {\rm Jy}^2$}
\psfrag{U}{$\qquad \qquad \qquad \qquad \U$}
\psfrag{phaI}{PI}
\psfrag{SameBaseline}{ Same Baseline}
\psfrag{AdjacentBaseline}{Adjacent Baseline}
\psfrag{1theoretical}{ Theory}
\psfrag{1simulation}{ Simulation}
\psfrag{error}{$\, \, \, \Delta$}
\centering
\includegraphics[scale=1.0,angle=0]{sim_allbl_pI.eps}
\caption{The upper left and right panels show the two-visibility
correlation $\viscornonu$ for the same and the adjacent baselines respectively for
OWFA PI. The solid lines show the theoretical predictions (eq.~\ref{eq:viscor_sph})
while the points with error-bars show the mean and standard
deviation from the simulations considering a few selected baselines.
The points in the bottom panels show $\Delta$ (eq.~\ref{eq:Delta})
which quantifies the deviation between the theoretical predictions and
the simulations. }
\label{fig:PI_APS}
\end{figure*}
The upper panels of Figure~\ref{fig:PI_APS} show the two-visibility correlation
$\viscornonu$ for OWFA PI considering the fixed frequency $\nu_i = \nu_j = \nu_c$,
which has not been shown explicitly. The left panel considers the situation
where the two baselines are same $\U_a = \U_b$, whereas the right panel shows
the results for the adjacent baselines $\U_b = \U_{a+1}$. We have used the
$\mathcal{C}_{\ell}$ given by eq.~(\ref{eq:cl_1}) in eq.~(\ref{eq:viscor_sph})
to calculate the full SH theoretical predictions
for the two-visibility correlations
$\viscornonu$ shown by the solid lines in Figure~\ref{fig:PI_APS}. We find
that for both the same and the adjacent baselines, the theoretical predictions
are in good agreement with the simulation. Note that $N_r = 100$ independent
realizations were used to estimate the mean $\viscornonu$ (shown with points)
and the standard deviation $\sigma$ (shown with error bars). The lower panels
of Figure~\ref{fig:PI_APS} show
\begin{equation}
\Delta = \frac{\delta \viscornonu \, \sqrt{N_r}}{\sigma} \, ,
\label{eq:Delta}
\end{equation}
%--------------------------------------
where $\delta \viscornonu$ is the difference between the theoretical predictions
and the simulations. We expect, $\Delta$ to have a Gaussian distribution with zero
mean and unit variance. We find that the values of $\Delta$ in the
lower panel of
Figure~\ref{fig:PI_APS} are roughly centred around zero and distributed within
$\pm 3$, consistent with what one would expect from the Gaussian
distribution.
\begin{figure*}
\psfrag{V2V2UaUb}{$\viscornonu {\rm Jy}^2$}
\psfrag{U}{$\qquad \qquad \qquad \qquad \U$}
\psfrag{phaII}{PII}
\psfrag{SameBaseline}{ Same Baseline}
\psfrag{AdjacentBaseline}{ Adjacent Baseline}
\psfrag{1theoretical}{ Theory}
\psfrag{1simulation}{ Simulation}
\psfrag{error}{$\, \, \, \Delta $}
\centering
\includegraphics[scale=1.0,angle=0]{sim_allbl_pII.eps}
\caption{The same as Figure~\ref{fig:PI_APS} for OWFA PII.}
\label{fig:PII_APS}
\end{figure*}
Figure~\ref{fig:PII_APS} shows a comparison of the theoretical
predictions with the simulations for the two visibility correlations
for OWFA PII. Here also we find that the theoretical predictions
are in good agreement with the simulations. The fact that the
theoretical predictions are in good agreement with the simulations
for both OWFA PI and PII
validates our formalism for calculating the beam transfer function
(eq.~\ref{eq:blm2}), the window function (eq.~\ref{eq:wgt_1}) and the
visibility correlation (eq.~\ref{eq:viscor_sph}).
\begin{figure*}
\centering
\psfrag{U}{$\qquad \qquad \qquad \qquad \U$}
\psfrag{V2V2UaUb}{$\viscornonu {\rm Jy}^2$}
\psfrag{phaI}{PI}
\psfrag{covcovcovcovcvo}{{ \qquad \qquad \, FSA}}
\psfrag{sphsphsphsphsph}{{\qquad\qquad \quad \SHE}}
\psfrag{limberapproximation}{ \qquad \qquad \quad \, LA}
\psfrag{error}{$\%$-error}
\psfrag{SameBaseline}{\, \, Same Baseline}
\psfrag{AdjacentBaseline}{Adjacent Baseline}
\includegraphics[scale=1.0,angle=0]{stest_allbl_pI.eps}
\caption{The upper left and right panels show the two-visibility
correlation $\viscornonu$ for the same and the adjacent baselines respectively for
OWFA PI. The solid lines show the theoretical predictions obtained using the FSA
(eq.~\ref{eq:viscor_2}), the points correspond to the predictions
obtained from the full SH formalism (eqs.~\ref{eq:blm2}, \ref{eq:viscor_sph}, \ref{eq:wgt_1})and the
dot-dashed lines show predictions obtained using the SH formalism with LA for $\Blmnonu$
(eqs.~\ref{eq:blm5}, \ref{eq:viscor_sph}, \ref{eq:wgt_1}). In the lower panels,
we show the percentage deviation of the FSA (in solid lines) and
LA (in dot-dashed lines)predictions relative to the full SH analysis.}
\label{fig:PI_flat_APS}
\end{figure*}
Considering the two-visibility correlation $\viscornonu$ for OWFA PI,
the upper panels of Figure~\ref{fig:PI_flat_APS} show a
comparison between the FSA (eq.~\ref{eq:viscor_2}) (solid lines),
the full SH formalism
(eqs.~\ref{eq:blm2}, \ref{eq:viscor_sph}, \ref{eq:wgt_1}) (points),
and the SH formalism with LA for $\Blmnonu$
(eqs.~\ref{eq:blm5}, \ref{eq:viscor_sph}, \ref{eq:wgt_1})
(dot-dashed lines). The lower panels show the percentage deviation of
the FSA and LA predictions relative to the full SH analysis.
Considering the correlations at the same baseline (left panels), we find
that the FSA is in good agreement with the full SH predictions over
the entire $U$ range and the deviations are less than $5 \%$.
Considering the LA, we find that the results match the full SH
analysis at large baselines, the deviations are $\le 5 \%$ for $U \ge
60$. However, these deviations increase rapidly at smaller baselines
and we have $\sim 100 \%$ deviations at the smallest baseline$(U =
12.5)$. For the correlations at the adjacent baselines (right panel), we find
that the FSA is in good agreement with the full SH predictions for
$U \ge 40$ and the deviation goes upto $14 \%$ at the smallest baseline
$(U = 18.25)$. Considering the LA, we find that the results match the full SH
analysis at large baselines, the deviations are $\le 5 \%$ for $U \ge
60$. However, these deviations increase rapidly at smaller baselines
and we have $\sim 42 \%$ deviations for the correlation between the
two smallest baselines which corresponds to a mean value of $U =
18.25$.
Figure~\ref{fig:PII_flat_APS} shows a similar comparison for OWFA
PII. Considering the correlations at the same baseline (left panel),
we find that the FSA is in good agreement with the full SH
predictions over the entire $U$ range and the deviations are less
than $10 \%$.
Considering the LA, we find that the results match the full SH
analysis at large baselines, the deviations are $\le 5 \%$ for $U \ge
60$. However, these deviations increase rapidly at smaller baselines
and we have more than $ 200 \%$ deviations at the smallest baseline $(U =
2.1)$. For the correlations at the adjacent baselines (right panel), we find
that the FSA is in good agreement with the full SH predictions for
$U \ge 60$, and the deviation goes upto $16 \%$ at the smallest baseline pair $(U =
3.15)$. Considering the LA, we find that the results match the full SH
analysis at large baselines, the deviations are $\le 5 \%$ for $U \ge
60$. However, these deviations increase rapidly at smaller baselines
and we have $\sim 80 \%$ deviations for the smallest baseline pair $(U =
3.15)$.
\begin{figure*}
\centering
\psfrag{U}{$\qquad \qquad \qquad \qquad \U$}
\psfrag{V2V2UaUb}{$\viscornonu {\rm Jy}^2$}
\psfrag{phaII}{PII}
\psfrag{covcovcovcovcvo}{{\qquad\qquad \, FSA}}
\psfrag{sphsphsphsphsph}{{\qquad \qquad \quad \SHE}}
\psfrag{limberapproximation}{ \qquad \qquad \quad \, LA}
\psfrag{error}{$\%$-error}
\psfrag{SameBaseline}{ Same Baseline}
\psfrag{AdjacentBaseline}{ Adjacent Baseline}
\includegraphics[scale=1.0,angle=0]{stest_allbl_pII.eps}
\caption{The same as Figure~\ref{fig:PI_flat_APS} for OWFA PII.}
\label{fig:PII_flat_APS}
\end{figure*}
\section{Summary and Conclusion} \label{sec:discussion}
The upcoming Ooty Wide Field Array (OWFA;
\citealt{Subrahmanya2016a}) has the detection of the redshifted 21-cm
power spectrum as one of its primary science goals. Earlier works
({\it e.g.} \citealt{Ali2014}) dealing with the signal and foreground
predictions for the two modes (PI and PII) of OWFA have all assumed
the FSA. However, OWFA PII has a rather
large FoV ($\sim 57^{\circ}$) in the North-South direction
and it is important to incorporate the spherical nature of the sky. In
this paper we investigate the importance of this effect relative to
the earlier FSA. To this end we adopt the SH analysis which relates the measured
visibilities to the SH coefficients of the sky signal
through a beam transfer function \citep{Shaw2014}.
Considering a radio interferometer where the baselines are
all coplanar with the antenna aperture ({\it e.g.} OWFA, CHIME), we have
developed a new formalism to compute the beam transfer
function $\Blmnonu$ as an integral of the aperture power pattern
$\apnonu$ (eq.~\ref{eq:blm2}). In addition to computational
advantages, our formalism provides insight into the behaviour of
$\Blmnonu$, {\it i.e.} the $\ell$ and $m$ values at which we expect this to peak
for a particular baseline $\U$. For OWFA,
we find that $\Blmnonu$ is expected to
peak at $\ell \approx 2 \pi U$ and fall off with increasing
$|\Delta \ell |$ with a width $\Delta \ell \approx \pm 2
\pi d/\lambda$ around the peak value. We also provide a closed form
analytical expression for $\Blmnonu$ using the LA which is expected to hold at large $\ell$
or equivalently at large $U$. This analytic expression
(eq.~\ref{eq:blm5}) indicates that $\Blmnonu$ is expected to peak at
$m=0$ and fall off with increasing $m$ with a width $\Delta m \approx
\pm (2 \ell + 1) \lambda/b$ (Figure \ref{fig:flm}). It is worth noting
that while $m$ takes values in the range $-\ell \le m \le \ell$, the
sky signal is restricted to a range $-\ell \lambda/b \le m
\le \ell\lambda/b $ where $\lambda/b \approx 0.03$ for OWFA {\it i.e.}
OWFA only responds to $3 \%$ of the available $m$-modes because of
the relatively large width $b$ of the telescope aperture.
Figure \ref{fig:bt_func} shows $\Blmnonu$ for a few of the
baselines. For both PI and PII we find that the $\ell$ value where
$\Blmnonu$ peaks is roughly consistent with the expected values, for PI
the width also is consistent with this. However for PII $\Blmnonu$
oscillates to the left of the peak and has an extended tail to the
right of the peak. The LA provides a reasonable
match for PI, particularly at large $U$. For PII, LA correctly
predicts the peak, however, it does not provide a
good match to the behaviour of $\Blmnonu$ away from the peak even at
large $U$.
In the FSA, the two-visibility correlation
provides a direct estimate of the redshifted 21-cm power spectrum
\citep{Bharadwaj2001a,Bharadwaj2005}. In this paper we
incorporate the spherical sky and express the two-visibility
correlation (eq.~\ref{eq:viscor_sph}) as a weighted sum of
the multi-frequency angular power spectrum
(MAPS; \citealt{Datta2007}). The weights here are given by the window
functions $\mathcal{W}_{\ell}\left(\U_a, \nu_i ;\U_b, \nu_j \right)$
which are a sum (over $m$) of products of the beam transfer function
(eq.~\ref{eq:wgt_1}). The signal,
for OWFA, is present in the correlations at the same baseline and the
adjacent baselines only \citep{Ali2014}. Figure \ref{fig:win_func}
shows the window function for a few select baselines. We see that the
correlation at baseline $U_a$ pick up the signal corresponding to
$\ell \approx 2 \pi U_a$ whereas the correlation between the adjacent
baselines $U_a$ and $U_b$ corresponds to $\ell \approx \pi
(U_a+U_b)$. The LA provides a good match to the window function at
large baselines for PI, however the match (not shown in the Figure)
is not very good for PII.
We have carried out simulations to validate our spherical sky
formalism. We find (Figures \ref{fig:PI_APS} and \ref{fig:PII_APS})
that for both PI and PII our analytical predictions are consistent
with the results from the simulations, thereby validating our
formalism. Figures \ref{fig:PI_flat_APS} and \ref{fig:PII_flat_APS}
show a comparison between the predictions of our SH analysis with
those from the FSA. The situation where LA has been used to calculate
the beam transfer function is also considered for comparison. We find
that for both PI and PII, the FSA matches the SH predictions to within
$16 \%$ across the entire baseline range. As expected, the match is
better for PI as compared to PII. The match is also better for the
correlations at the same baseline as compared to the adjacent
baselines. In contrast, the LA provides a good match $< 5 \%$ only at
the large baselines $U \ge 60$. In conclusion, we note that the flat
sky approximation matches the full SH analysis to
within $15-16 \%$. It is necessary to use the latter if an accuracy
higher than this is required.
Finally, we note that the entire analysis here is restricted to
observations in a single FoV. This has advantage in
allowing detailed foreground modelling of the particular field which
may be useful in foreground removal. In contrast, the ``$m$-mode''
analysis proposed by \citet{Shaw2014} considers drift scan
observations which cover a large fraction of the sky. The latter
increases the available signal, this could however make detailed
foreground modelling more difficult. We plan to study $m$-mode
analysis in the context of OWFA in subsequent work.
\section*{Acknowledgement}
The authors would like to thank Jayaram N Chengalur and Visweshwar Ram Marthi for their
useful suggestions and discussions. SC acknowledges the University Grant Commission, India for
providing financial support through Senior Research Fellowship. SC would also
like to thank Sukannya Bhattacharya, Abinash Kumar Shaw, Anjan Kumar Sarkar and Debanjan Sarkar for
their help and useful discussions.
\bibliographystyle{mnras} \bibliography{sph_harmonics.bbl}
\end{document}
\begin{thebibliography}{42}
\expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
\bibitem[{Ali} \& {Bharadwaj}(2014)]{Ali2014}
{Ali} S.~S., {Bharadwaj} S., 2014, \japa, 35, 157
\bibitem[Ali et~al.(2005)Ali, Bharadwaj \& Pandey]{Ali2005}
Ali S.~S., Bharadwaj S., Pandey B., 2005, Monthly Notices of the Royal
Astronomical Society, 363, 1, 251
\bibitem[{Ali} et~al.(2015){Ali}, {Parsons}, {Zheng} et~al.]{Ali2015}
{Ali} Z.~S., {Parsons} A.~R., {Zheng} H., et~al., 2015, \apj, 809, 61
\bibitem[{Bandura} et~al.(2014){Bandura}, {Addison}, {Amiri}
et~al.]{Bandura2014}
{Bandura} K., {Addison} G.~E., {Amiri} M., et~al., 2014, in { Ground-based and
Airborne Telescopes V\/}, vol. 9145 of { \procspie\/}, 914522
\bibitem[{Bharadwaj} \& {Ali}(2005)]{Bharadwaj2005}
{Bharadwaj} S., {Ali} S.~S., 2005, \mnras, 356, 1519
\bibitem[{Bharadwaj} et~al.(2001){Bharadwaj}, {Nath} \&
{Sethi}]{Bharadwaj2001b}
{Bharadwaj} S., {Nath} B.~B., {Sethi} S.~K., 2001, \japa, 22, 21
\bibitem[{Bharadwaj} \& {Pandey}(2003)]{Bharadwaj2003}
{Bharadwaj} S., {Pandey} S.~K., 2003, Journal of Astrophysics and Astronomy,
24, 23
\bibitem[{Bharadwaj} et~al.(2015){Bharadwaj}, {Sarkar} \& {Ali}]{Bharadwaj2015}
{Bharadwaj} S., {Sarkar} A.~K., {Ali} S.~S., 2015, \japa, 36, 385
\bibitem[{Bharadwaj} \& {Sethi}(2001)]{Bharadwaj2001a}
{Bharadwaj} S., {Sethi} S.~K., 2001, Journal of Astrophysics and Astronomy, 22,
293
\bibitem[{Bharadwaj} et~al.(2009){Bharadwaj}, {Sethi} \&
{Saini}]{Bharadwaj2009}
{Bharadwaj} S., {Sethi} S.~K., {Saini} T.~D., 2009, \prd, 79, 8, 083538
\bibitem[{Bull} et~al.(2015){Bull}, {Ferreira}, {Patel} \&
{Santos}]{Bull2015ApJ}
{Bull} P., {Ferreira} P.~G., {Patel} P., {Santos} M.~G., 2015, \apj, 803, 21
\bibitem[{Chang} et~al.(2010){Chang}, {Pen}, {Bandura} \&
{Peterson}]{Chang2010}
{Chang} T.-C., {Pen} U.-L., {Bandura} K., {Peterson} J.~B., 2010, \nat, 466,
463
\bibitem[{Chang} et~al.(2008){Chang}, {Pen}, {Peterson} \&
{McDonald}]{Chang2008}
{Chang} T.-C., {Pen} U.-L., {Peterson} J.~B., {McDonald} P., 2008, Physical
Review Letters, 100, 9, 091303
\bibitem[Chatterjee et~al.(2017)Chatterjee, Bharadwaj \&
Marthi]{Chatterjee2017}
Chatterjee S., Bharadwaj S., Marthi V.~R., 2017, Journal of Astrophysics and
Astronomy, 38, 1, 15
\bibitem[{Chen}(2012)]{Chen2012}
{Chen} X., 2012, International Journal of Modern Physics Conference Series, 12,
256
\bibitem[{Chen}(2015)]{Chen2015}
{Chen} X., 2015, IAU General Assembly, 22, 2252187
\bibitem[Chengalur et~al.(2007)Chengalur, Gupta \& Dwarkanath]{Chengalur2007}
Chengalur J.~N., Gupta Y., Dwarkanath K., 2007, Low Frequency Radio Astronomy
3rd edition, NCRA-TIFR, India
\bibitem[{Datta} et~al.(2007){Datta}, {Choudhury} \& {Bharadwaj}]{Datta2007}
{Datta} K.~K., {Choudhury} T.~R., {Bharadwaj} S., 2007, \mnras, 378, 119
\bibitem[Ghosh et~al.(2018)Ghosh, Mertens \& Koopmans]{Ghosh2017}
Ghosh A., Mertens F.~G., Koopmans L. V.~E., 2018, Monthly Notices of the Royal
Astronomical Society, 474, 4, 4552
\bibitem[Gupta et~al.(2017)Gupta, Ajithkumar, Kale et~al.]{Gupta2017}
Gupta Y., Ajithkumar B., Kale H., et~al., 2017, CURRENT SCIENCE, 113, 4, 707
\bibitem[Górski et~al.(2005)Górski, Hivon, Banday et~al.]{Gorski2005}
Górski K.~M., Hivon E., Banday A.~J., et~al., 2005, The Astrophysical Journal,
622, 2, 759
\bibitem[Hazra \& Sarkar(2012)]{Hazra2012}
Hazra D.~K., Sarkar T.~G., 2012, Physical review letters, 109, 12, 121301
\bibitem[Limber(1954)]{Limber1954}
Limber D.~N., 1954, \apj, 119, 655
\bibitem[{Liu} et~al.(2016){Liu}, {Zhang} \& {Parsons}]{Liu2016}
{Liu} A., {Zhang} Y., {Parsons} A.~R., 2016, \apj, 833, 242
\bibitem[{Marthi} et~al.(2017){Marthi}, {Chatterjee}, {Chengalur} \&
{Bharadwaj}]{Marthi2016a}
{Marthi} V.~R., {Chatterjee} S., {Chengalur} J.~N., {Bharadwaj} S., 2017,
\mnras, 471, 3112
\bibitem[{Masui} et~al.(2010){Masui}, {McDonald} \& {Pen}]{Masui2010}
{Masui} K.~W., {McDonald} P., {Pen} U.-L., 2010, \prd, 81, 10, 103527
\bibitem[{Morales} \& {Hewitt}(2004)]{Morales2004}
{Morales} M.~F., {Hewitt} J., 2004, \apj, 615, 7
\bibitem[Newburgh et~al.(2016)Newburgh, Bandura, Bucher et~al.]{Newburgh2016}
Newburgh L., Bandura K., Bucher M., et~al., 2016, arXiv preprint
arXiv:1607.02059
\bibitem[Perley et~al.(1989)Perley, Schwab \& Bridle]{Perley1989}
Perley R.~A., Schwab F.~R., Bridle A.~H., 1989
\bibitem[Sarkar et~al.(2018)Sarkar, Bharadwaj \& Marthi]{Sarkar2016b}
Sarkar A.~K., Bharadwaj S., Marthi V.~R., 2018, Monthly Notices of the Royal
Astronomical Society, 473, 1, 261
\bibitem[{Seo} et~al.(2010){Seo}, {Dodelson}, {Marriner} et~al.]{Seo2010}
{Seo} H.-J., {Dodelson} S., {Marriner} J., et~al., 2010, \apj, 721, 164
\bibitem[{Shaw} et~al.(2014){Shaw}, {Sigurdson}, {Pen}, {Stebbins} \&
{Sitwell}]{Shaw2014}
{Shaw} J.~R., {Sigurdson} K., {Pen} U.-L., {Stebbins} A., {Sitwell} M., 2014,
\apj, 781, 57
\bibitem[{Shaw} et~al.(2015){Shaw}, {Sigurdson}, {Sitwell}, {Stebbins} \&
{Pen}]{Shaw2015}
{Shaw} J.~R., {Sigurdson} K., {Sitwell} M., {Stebbins} A., {Pen} U.-L., 2015,
\prd, 91, 8, 083514
\bibitem[Subrahmanya et~al.(2017{\natexlab{a}})Subrahmanya, Manoharan \&
Chengalur]{Subrahmanya2016a}
Subrahmanya C.~R., Manoharan P.~K., Chengalur J.~N., 2017{\natexlab{a}}, J.
Astrophys. Astron., 38, 10
\bibitem[Subrahmanya et~al.(2017{\natexlab{b}})Subrahmanya, Prasad, Girish,
Somashekar, Manoharan \& Mittal]{Subrahmanya2016b}
Subrahmanya C.~R., Prasad P., Girish B.~S., Somashekar R., Manoharan P.~K.,
Mittal A.~K., 2017{\natexlab{b}}, Journal of Astrophysics and Astronomy, 38,
1, 11
\bibitem[{Swarup} et~al.(1991){Swarup}, {Ananthakrishnan}, {Kapahi}, {Rao},
{Subrahmanya} \& {Kulkarni}]{Swarup1991}
{Swarup} G., {Ananthakrishnan} S., {Kapahi} V.~K., {Rao} A.~P., {Subrahmanya}
C.~R., {Kulkarni} V.~K., 1991, Current Science, Vol.~60, NO.2/JAN25, P.~95,
1991, 60, 95
\bibitem[{Swarup} et~al.(1971){Swarup}, {Sarma}, {Joshi} et~al.]{Swarup1971}
{Swarup} G., {Sarma} N.~V.~G., {Joshi} M.~N., et~al., 1971, Nature Physical
Science, 230, 185
\bibitem[{Visbal} et~al.(2009){Visbal}, {Loeb} \& {Wyithe}]{Visbal2009}
{Visbal} E., {Loeb} A., {Wyithe} S., 2009, \jcap, 10, 030
\bibitem[{Wyithe} et~al.(2008){Wyithe}, {Loeb} \& {Geil}]{Wyithe2008}
{Wyithe} J.~S.~B., {Loeb} A., {Geil} P.~M., 2008, \mnras, 383, 1195
\bibitem[Zhang et~al.(2016{\natexlab{a}})Zhang, Ansari, Chen, Campagne,
Magneville \& Wu]{Zhang2016a}
Zhang J., Ansari R., Chen X., Campagne J.-E., Magneville C., Wu F.,
2016{\natexlab{a}}, Mon. Not. Roy. Astron. Soc., 461, 2, 1950
\bibitem[Zhang et~al.(2016{\natexlab{b}})Zhang, Zuo, Ansari et~al.]{Zhang2016b}
Zhang J., Zuo S.-F., Ansari R., et~al., 2016{\natexlab{b}}, Research in
Astronomy and Astrophysics, 16, 10, 158
\bibitem[{Zheng} et~al.(2014){Zheng}, {Tegmark}, {Buza} et~al.]{Zheng2014}
{Zheng} H., {Tegmark} M., {Buza} V., et~al., 2014, \mnras, 445, 1084
\end{thebibliography}