1: \section{Search of VHE $\gamma$-rays from observations of Sagittarius Dwarf by H.E.S.S.}
2: \label{sec:analysis}
3: \subsection{The H.E.S.S. instrument}
4: H.E.S.S. (High Energy Stereoscopic System) is an array of four
5: imaging atmospheric Cherenkov telescopes located in the Khomas Highlands
6: of Namibia at an altitude of 1800 m above sea level. Each
7: telescope %of 13 m diameter
8: consists of an optical reflector of about 107 m$^2$ effective area
9: composed of 382 round mirrors arranged on a Davis-Cotton
10: mount~\cite{bernlohr}. The interaction of the primary $\gamma$-ray
11: in the Earth's upper atmosphere initiates an electromagnetic
12: shower. The reflector collects the Cherenkov light emitted by the
13: charged particles composing this shower, and focuses it onto a
14: camera comprising 960 fast photomultipliers (PMTs) of individual
15: field of view of 0.16$^{\circ}$ diameter~\cite{vincent}. Each tube
16: is equipped with Winston cones to limit the field of view of each
17: PMT and minimize the background light. The total field of view of
18: the H.E.S.S. instrument is 5$^{\circ}$ in diameter. The
19: stereoscopy technique used in the imaging atmospheric Cherenkov
20: telescopes allows for accurate reconstruction of the direction and
21: energy of the primary $\gamma$-rays as well as an efficient
22: rejection of the background induced by cosmic ray
23: interactions~\cite{trigger}. The energy threshold of H.E.S.S. at
24: zenith before selection cuts moved from 100 GeV at the
25: commissioning of the experiment in 2003 to 160 GeV due to the
26: degradation of the optical performance in 2006. The point source
27: sensitivity is better than $\rm 2\times 10^{-13} cm^{-2}s^{-1}$
28: above 1 TeV for a 5$\sigma$ detection in 25 hours~\cite{crabe}.
29:
30: \subsection{Dataset}
31: The observations of the Sgr dSph were taken in June 2006 with
32: zenith angles ranging from 7$^{\circ}$ to 43$^{\circ}$ around an
33: average value of 19$^{\circ}$. Data were taken in 28-minute
34: observation runs in the wobble mode method with pointing
35: directions offset by an angular distance of typically
36: $\pm$0.7$^{\circ}$ from the nominal target position. The dataset
37: suitable for analysis was selected using the standard run
38: selection procedure~\cite{crabe}, which in particular removes the
39: data taken under bad atmospheric conditions. A total of 25 runs
40: out of 26 are selected for the analysis. After calibration of the
41: raw shower images from PMT signals~\cite{aharonian0}, two
42: independent reconstruction techniques were combined to select
43: $\gamma$-ray events and reconstruct their direction and energy.
44: The first one uses the Hillas moment method~\cite{aharonian1}. The
45: second analysis referred hereafter as ``Model Analysis'', is based
46: on the pixel-per-pixel comparison of the shower image with a
47: template generated by a semi-analytical shower development model.
48: The event reconstruction relies on a maximum likelihood method
49: which uses available pixels in the camera, without requirement for
50: an image cleaning~\cite{denauroi0,denauroi1}. The reconstructed
51: shower parameter (energy, impact, direction and primary
52: interaction point) are obtained as a product of the minimization
53: procedure.
54: The separation between $\gamma$ candidates and hadrons is done
55: using a combination of the Model goodness-of-fit
56: parameter~\cite{denauroi1} and the Hillas mean scaled width and length
57: parameters, which results in an improved background
58: rejection~\cite{crabe}.
59: Standard cuts on the width and the length of Hillas ellipses
60: combined with the goodness-of-fit are used to suppress the
61: hadronic background~\cite{aharonian1}. An additional cut on the
62: primary interaction depth is used to improve background rejection.
63: Both methods yield a typical energy resolution of 15\% above
64: energy threshold. In the Model Analysis, the angular resolution at
65: the 68\% containment radius is found to be better than
66: 0.06$^{\circ}$ per $\gamma$-ray.
67: \subsection{Data analysis}
68: The on-source signal is defined by integrating all the events with
69: angular position $\theta$ in a circle around the target position
70: with a radius of $\theta_{cut}$. The target position is chosen
71: according to the photometric measurements of the Sgr dSph luminous
72: cusp showing that the position of the centre corresponds to the
73: centre of the globular cluster M54~\cite{monaco1}. The target
74: position is thus found to be $\rm (RA = 18^h55^m59.9s, Dec =
75: -30d28'59.9'')$ in equatorial coordinates (J2000.0) or $\rm (l =
76: 5^{\circ}41'12.9'',b = -14^{\circ}16'29.8'')$ in Galactic
77: coordinates. The signal coming from Sgr dSph is expected to come
78: from a region of 1.5 pc, about 30'', much smaller the H.E.S.S.
79: point spread function (PSF). A $\theta_{cut}$ value of
80: 0.14$^{\circ}$ suitable for a point-like source was therefore used
81: in the analysis. In case of a Navarro-Frenk-White (NFW) density
82: profile~\cite{nfw} for which $\rho$ follows r$^{-1}$ or a cored
83: profile~\cite{evans} folded with the point spread function (PSF)
84: of H.E.S.S., the integration region allows to retrieve a
85: significant fraction of the expected signal. See
86: Table~\ref{tab:table}. A cut on the image size of 60
87: photoelectrons is used to obtain a good sensitivity for weak
88: sources. In order to reduce systematic effects which affect images
89: close to the edges of the camera, only events reconstructed within
90: a maximum distance of 2.5$^{\circ}$ from the camera centre are
91: used for this analysis. The excess sky map is obtained by the
92: subtraction of a background model on the $\gamma$-ray candidate
93: sky distribution. The background level is estimated using the
94: ring-background method~\cite{puhlhofer} where the background rate
95: is calculated from the integration of $\gamma$-like events falling
96: in an annulus around the centre of the camera with identical
97: observation conditions and acceptances than that used for the
98: on-source region, which allows
99: an estimate of the background on every sky position.\\
100: The excess sky map of the $\gamma$-ray candidates is presented in
101: Fig.~\ref{fig:excessmap} in the RA/Dec J2000 coordinates centered
102: on the Sgr target position. For each bin, $\gamma$-ray like events
103: are summed within a radius of 0.14$^{\circ}$. No $\gamma$-ray
104: excess is found at the target position. Further, no significant
105: excess is observed anywhere else in the sky map. The $\theta^2$
106: distribution of the observed $\gamma$-ray events relative to the
107: target position, is presented in Fig.~\ref{fig:theta2} as well as
108: the background distribution. In order to check the robustness of
109: the results presented here, the dataset has been analyzed using
110: different analysis methods as the so-called Hillas and
111: model3D~\cite{lemoine} methods. No significant $\gamma$-ray excess
112: is detected
113: in the corresponding sky maps.\\
114: Since Fig.~\ref{fig:theta2} shows no $\gamma$-ray excess, we
115: derived the 95\% confidence level upper limit on the observed
116: number of $\gamma$-rays : $\rm N_{\gamma}^{95\%\, C.L.}$. The
117: limit is computed knowing the numbers of events above the energy
118: of 250 GeV in the signal region $\rm N_{ON} = 437$, in the
119: background region $\rm N_{OFF} = 4270$, and the ratio of the
120: off-source time to the on-source time $\rm\alpha = 10.1$. We use
121: the Feldman \& Cousins method~\cite{feldman} and obtain :
122: \begin{equation}
123: N_{\gamma}^{95\%\,C.L.} = 56\,.
124: \end{equation}
125: Given the acceptance of the detector for the observations of the
126: dSph Sgr, an 95\% confidence level upper limit on the $\gamma$-ray
127: flux is also derived :
128: \begin{center}
129: $\rm \Phi_{\gamma}(E_{\gamma} > 250\,GeV) < 3.6 \times 10^{-12}
130: \,cm^{-2}s^{-1}\,(95\%\,C.L.)$
131: \end{center}
132: