The 2HWC HAWC Observatory Gamma Ray Catalog

TL;DR

This paper presents the first catalog of TeV gamma-ray sources detected by the HAWC observatory, showcasing its sensitivity and continuous sky monitoring capabilities, resulting in 39 detected sources including many new discoveries.

Contribution

It provides the first comprehensive catalog of TeV gamma-ray sources from the completed HAWC observatory, with detailed source measurements and new source identifications.

Findings

39 sources detected with high significance

16 sources are new and not previously reported in TeV catalogs

HAWC's sensitivity allows for continuous sky monitoring at TeV energies

Abstract

We present the first catalog of TeV gamma-ray sources realized with the recently completed High Altitude Water Cherenkov Observatory (HAWC). It is the most sensitive wide field-of-view TeV telescope currently in operation, with a 1-year survey sensitivity of ~5-10% of the flux of the Crab Nebula. With an instantaneous field of view >1.5 sr and >90% duty cycle, it continuously surveys and monitors the sky for gamma ray energies between hundreds GeV and tens of TeV. HAWC is located in Mexico at a latitude of 19 degree North and was completed in March 2015. Here, we present the 2HWC catalog, which is the result of the first source search realized with the complete HAWC detector. Realized with 507 days of data and represents the most sensitive TeV survey to date for such a large fraction of the sky. A total of 39 sources were detected, with an expected contamination of 0.5 due to background fluctuation. Out of these sources, 16 are more than one degree away from any previously reported TeV source. The source list, including the position measurement, spectrum measurement, and uncertainties, is reported. Seven of the detected sources may be associated with pulsar wind nebulae, two with supernova remnants, two with blazars, and the remaining 23 have no firm identification yet.

Tables

Table 1. Table 1: Properties of the nine analysis bins: bin number ℬ ℬ \mathcal{B} , event size f hit subscript 𝑓 hit f_{\mathrm{hit}} , 68% PSF containment ψ 68 subscript 𝜓 68 \psi_{68} , cut selection efficiency for gammas ϵ γ MC superscript subscript italic-ϵ 𝛾 MC \epsilon_{\gamma}^{\mathrm{MC}} and cosmic rays ϵ CR data superscript subscript italic-ϵ CR data \epsilon_{\mathrm{CR}}^{\mathrm{data}} , and median energy for a reference source of spectral index − 2.63 2.63 -2.63 at a declination of 20 ∘ E ~ γ MC superscript subscript ~ 𝐸 𝛾 MC \tilde{E}_{\gamma}^{\mathrm{MC}} .

$ℬ$	$f_{\mathrm{hit}}$	${\psi }_{68}$	${ϵ}_{\gamma }^{\mathrm{MC}}$	${ϵ}_{\mathrm{CR}}^{\mathrm{data}}$	${\tilde{E}}_{\gamma }^{\mathrm{MC}}$
	(%)	(∘)	(%)	(%)	(TeV)
1	6.7 – 10.5	1.03	70	15	0.7
2	10.5 – 16.2	0.69	75	10	1.1
3	16.2 – 24.7	0.50	74	5.3	1.8
4	24.7 – 35.6	0.39	51	1.3	3.5
5	35.6 – 48.5	0.30	50	0.55	5.6
6	48.5 – 61.8	0.28	35	0.21	12
7	61.8 – 74.0	0.22	63	0.24	15
8	74.0 – 84.0	0.20	63	0.13	21
9	84.0 – 100.0	0.17	70	0.20	51

Table 2. Table 2: 2HWC source list and nearest TeVCat sources. The sources with a * symbol correspond to sources that are not separated from their neighbor by a large TS gap, as defined in section  3.5 .

								Nearest TeVCat source
Name	Search	TS	RA	Dec	l	b	1$\sigma$ stat. unc.	Dist.	Name
			[∘]	[∘]	[∘]	[∘]	[∘]	[∘]
2HWC J0534+220	PS	1.1E+4	83.63	22.02	184.55	-5.78	0.06	0.01	Crab
2HWC J0631+169	PS	29.6	98.00	17.00	195.61	3.51	0.11	0.39	Geminga
2HWC J0635+180	PS	27.4	98.83	18.05	195.04	4.70	0.13	0.97	Geminga
2HWC J0700+143	1.0∘	29	105.12	14.32	201.10	8.44	0.80	2.98	-
2HWC J0819+157	0.5∘	30.7	124.98	15.79	208.00	26.52	0.17	7.86	-
2HWC J1040+308	0.5∘	26.3	160.22	30.87	197.59	61.31	0.22	8.77	-
2HWC J1104+381	PS	1.15E+3	166.11	38.16	179.95	65.05	0.06	0.04	Markarian 421
2HWC J1309-054	PS	25.3	197.31	-5.49	311.11	57.10	0.22	3.27	-
2HWC J1653+397	PS	556	253.48	39.79	63.64	38.85	0.07	0.03	Markarian 501
2HWC J1809-190	PS	85.5	272.46	-19.04	11.33	0.18	0.17	0.31	HESS J1809-193
2HWC J1812-126	PS	26.8	273.21	-12.64	17.29	2.63	0.19	0.14	HESS J1813-126
2HWC J1814-173	PS	141	273.52	-17.31	13.33	0.13	0.18	0.54	HESS J1813-178
2HWC J1819-150*	PS	62.9	274.83	-15.06	15.91	0.09	0.16	0.51	SNR G015.4+00.1
2HWC J1825-134	PS	767	276.46	-13.40	18.12	-0.53	0.09	0.39	HESS J1826-130
2HWC J1829+070	PS	25.3	277.34	7.03	36.72	8.09	0.10	8.12	-
2HWC J1831-098	PS	107	277.87	-9.90	21.86	-0.12	0.17	0.01	HESS J1831-098
2HWC J1837-065	PS	549	279.36	-6.58	25.48	0.10	0.06	0.37	HESS J1837-069
2HWC J1844-032	PS	309	281.07	-3.25	29.23	0.11	0.10	0.18	HESS J1844-030
2HWC J1847-018	PS	132	281.95	-1.83	30.89	-0.03	0.11	0.17	HESS J1848-018
2HWC J1849+001	PS	134	282.39	0.11	32.82	0.47	0.10	0.16	IGR J18490-0000
2HWC J1852+013*	PS	71.4	283.01	1.38	34.23	0.50	0.13	1.37	-
2HWC J1857+027	PS	303	284.33	2.80	36.09	-0.03	0.06	0.14	HESS J1857+026
2HWC J1902+048*	PS	31.7	285.51	4.86	38.46	-0.14	0.18	2.03	-
2HWC J1907+084*	PS	33.1	286.79	8.50	42.28	0.41	0.27	1.15	-
2HWC J1908+063	PS	367	287.05	6.39	40.53	-0.80	0.06	0.14	MGRO J1908+06
2HWC J1912+099	PS	83.2	288.11	9.93	44.15	-0.08	0.10	0.24	HESS J1912+101
2HWC J1914+117*	PS	33	288.68	11.72	46.00	0.25	0.13	1.64	-
2HWC J1921+131	PS	30.1	290.30	13.13	47.99	-0.50	0.12	1.14	-
2HWC J1922+140	PS	49	290.70	14.09	49.01	-0.38	0.11	0.10	W 51
2HWC J1928+177	PS	65.7	292.15	17.78	52.92	0.14	0.07	1.18	-
2HWC J1930+188	PS	51.8	292.63	18.84	54.07	0.24	0.12	0.03	SNR G054.1+00.3
2HWC J1938+238	PS	30.5	294.74	23.81	59.37	0.94	0.13	2.75	-
2HWC J1949+244	1.0∘	34.9	297.42	24.46	61.16	-0.85	0.71	3.43	-
2HWC J1953+294	PS	30.1	298.26	29.48	65.86	1.07	0.24	8.44	-
2HWC J1955+285	PS	25.4	298.83	28.59	65.35	0.18	0.14	7.73	-
2HWC J2006+341	PS	36.9	301.55	34.18	71.33	1.16	0.13	3.61	-
2HWC J2019+367	PS	390	304.94	36.80	75.02	0.30	0.09	0.07	VER J2019+368
2HWC J2020+403	PS	59.7	305.16	40.37	78.07	2.19	0.11	0.40	VER J2019+407
2HWC J2024+417*	PS	28.4	306.04	41.76	79.59	2.43	0.20	0.97	MGRO J2031+41
2HWC J2031+415	PS	209	307.93	41.51	80.21	1.14	0.09	0.08	TeV J2032+4130

Equations

$$TS=2\text{ln}\frac{{\mathcal{L}}^{\max }(\text{Source Model})}{\mathcal{L}(\text{Null Model})},$$

$$dN/dE=F_0(E/E_0{)}^{\alpha },$$

Full text

HAWC is located in Mexico at a latitude of 19∘ North and was completed in March 2015. Here, we present the 2HWC catalog, which is the result of the first source search realized with the complete HAWC detector. Realized with 507 days of data and represents the most sensitive TeV survey to date for such a large fraction of the sky. A total of 39 sources were detected, with an expected contamination of 0.5 due to background fluctuation. Out of these sources, 16 are more than one degree away from any previously reported TeV source. The source list, including the position measurement, spectrum measurement, and uncertainties, is reported. Seven of the detected sources may be associated with pulsar wind nebulae, two with supernova remnants, two with blazars, and the remaining 23 have no firm identification yet.

\AuthorCollaborationLimit

=300

1 Introduction

The High Altitude Water Cherenkov Observatory (HAWC) is a newly completed very high energy (VHE; $>$100 GeV) gamma-ray observatory with a 1-year survey sensitivity of $\sim$5–10% of the flux of the Crab Nebula. The variation in sensitivity depends on the declination of the source under consideration over the observable sky, with declinations between $-20$∘ and 60∘ for the present study. Unlike imaging atmospheric Cherenkov telescopes (IACTs), such as H.E.S.S. (Aharonian et al., 2004), MAGIC (Aleksić et al., 2016), VERITAS (Holder et al., 2006), and FACT (Anderhub et al., 2011) which observe the Cherenkov light emitted by the extensive air showers as they develop in the atmosphere, HAWC detects particles of these air showers that reach ground level, allowing it to operate continuously and observe an instantaneous field of view of $>$1.5 sr. Prior to this work, unbiased VHE surveys were conducted by the Milagro (Atkins et al., 2003, 2004) and ARGO (Bacci et al., 2002) collaborations. Compared to these previous surface arrays, the sensitivity of HAWC is improved by more than an order of magnitude thanks to a combination of large size, high elevation, and unique background rejection capability. These features make HAWC ideally suited as a VHE survey instrument. High-sensitivity surveys of portions of the Galactic Plane have also been published by H.E.S.S. (Aharonian et al., 2006b), MAGIC (Albert et al., 2006) and VERITAS (Popkow et al., 2015). At lower energies, the Large Area Telescope on the space-based Fermi Observatory (Fermi-LAT) has detected many thousands of Galactic and extragalactic gamma-ray sources (Acero et al., 2015), but its small size limits its reach into the VHE band. The work presented here is the most sensitive comprehensive sky survey carried out above 1 TeV.

There are about 200 known VHE gamma-ray sources detected at high significance by a number of observatories (e.g. TeVCat catalog; Wakely & Horan, 2008).

Within the Galaxy, the VHE sources include pulsar wind nebula (PWNe), supernova remnants (SNRs), binary systems, and diffuse emission from the Galactic plane. The SNRs and PWNe represent the majority of the identified sources. Most Galactic gamma-ray sources have power-law spectra consistent with shock acceleration of electrons, though there is considerable evidence for gamma-ray production by hadronic cosmic rays interacting with matter. Most Galactic sources are observed as spatially extended by IACTs (Carrigan et al., 2013).

Beyond our galaxy, almost all known TeV sources are Active Galactic Nuclei (AGNs) and most of them are categorized as blazars. The TeV gamma-ray emission is generally observed to be variable and thought to originate from one or multiple regions of particle acceleration in the jet. While gamma-ray emission has been observed up to energies of about 10 TeV for some blazars (Acciari et al., 2011; Aharonian et al., 2001), the flux at and beyond such energies is strongly attenuated as a function of distance due to photon-photon interaction with the extragalactic background light (EBL). Since the sensitivity of HAWC peaks around 10 TeV (depending on the source spectrum and declination, see Section 4.1 for details), where absorption of TeV photons through the infrared component of the EBL becomes severe, the sensitivity of the HAWC survey to distant AGNs is relatively poor.

Many VHE sources are not unambiguously associated with objects identified at other wavelengths (a fifth of TeVCat sources are reported as unidentified). Further spectral and morphological studies are required to understand their origins and emission mechanisms.

In addition to a peak sensitivity at higher energies, the angular resolution of HAWC is larger than the IACT’s. Consequently, comparison of source significance and flux with IACT observations requires careful examination. For example, the HAWC instrument is relatively more sensitive to sources with harder energy spectra than softer ones, and to extended sources than pointlike sources. On the other hand, the surface detection method employed by HAWC permits continuous observation of the entire overhead sky, both during the day and night and under all weather conditions. For sources that transit through its field of view, HAWC typically accumulates 1500–2000 hours/yr of total exposure. Thus, above 10 TeV where photon statistics are poor, HAWC achieves better sensitivity than even long-duration observations by IACTs.

This paper presents a catalog of TeV gamma-ray sources resulting from a search for significantly enhanced point and extended emission detected in the gamma-ray sky maps of 17 months of HAWC data. More detailed morphology studies will be the subject of future papers.

In Section 2, we describe the HAWC detector. Section 3 describes the analysis of gamma-ray events and the construction of our source catalog. Results and discussion are provided in Sections 4, 5, and 6, and conclusions and outlook in Section 7.

2 HAWC Detector

The HAWC detector is located in central Mexico at 18∘59’41”N 97∘18’30.6”W and an elevation of 4100 m a.s.l. The instrument comprises 300 identical water Cherenkov detectors (WCDs) made from 5 m high, 7.32 m diameter commercial water storage tanks. Each tank contains a custom-made light-tight bladder to accommodate 190,000 liters of purified water. Four upward facing photomultiplier tubes (PMTs) are mounted at the bottom of each tank: a 10” Hamamatsu R7081-HQE PMT positioned at the center and three 8” Hamamatsu R5912 PMTs which are positioned halfway between the tank center and rim. The central PMT has roughly twice the sensitivity of the outer PMTs due to its superior quantum efficiency and its larger size. The WCDs are filled to a depth of 4.5 m, with 4.0 m (more than 10 radiation lengths) of water above the PMTs. This large depth guarantees that the electrons, positrons, and gammas in the air shower are fully absorbed by the HAWC detector well above the PMT level, so that the detector itself acts as an electromagnetic (EM) calorimeter providing an accurate measurement of EM energy deposition. High-energy electrons are detected via the Cherenkov light they produce in the water and gamma rays are converted to electrons through pair production and Compton scattering. Muons are also detected. They are more likely to be produced in air showers originating from hadronic cosmic-ray interactions with the atmosphere and tend to have higher transverse momentum producing large signals in the PMTs far from the air shower axis and thus serve as useful tags for rejecting hadronic backgrounds. The WCDs are arranged in a compact layout to maximize the density of the sensitive area, with about 60% of the 22,000 m2 detector area instrumented. See Figure 1 for a diagram of the HAWC detector.

Analog signals from the PMTs are transmitted by RG-59 coaxial cable to a central counting house. The signals are shaped and discriminated at two voltage thresholds roughly corresponding to 1/4 PE and 4 PEs and the threshold crossing times (both rising and falling) are recorded using CAEN V1190A time-to-digital converters. Individual signals that pass at least the low threshold are called hits. The time-over-threshold is used to estimate the charge. The response of this system is roughly logarithmic, so that the readout has reasonable charge resolution over a very wide dynamic range, from a fraction of 1 PE to 10,000 PEs. The timing resolution for large pulses is better than 1 ns. All channels are read out in real time with zero dead time and blocks of data are aggregated in a real-time computing farm. A trigger is generated when a sufficient number of PMTs record a hit within a 150 ns window (28 hits were required for most of the data used in this analysis, though other values were occasionally used earlier). This results in a $\sim$20 kHz trigger rate. Small events, with a number of hits close to the threshold value and which dominate the triggers, require a specific treatment and are removed from the analysis presented here. In the future their inclusion will significantly lower the energy threshold of HAWC. For sources with spectra that extend beyond 1 TeV, like the Crab Nebula, the sensitivity usually peaks above 5 TeV (depending on the source spectrum and declination) and excluding the near-threshold events does not significantly reduce the sensitivity. Details of the event selection for the present analysis are presented in the next section.

For each triggered event, the parameters of the air shower, like the direction, the size, and some gamma/hadron separation variables, are extracted from the recorded hit times and amplitudes, using a shower model developed through the study of Monte Carlo simulations and optimized using observations of the Crab Nebula (Abeysekara et al., 2017, submitted to ApJ). The angular resolution of the HAWC instrument varies with the event size (number of hit PMTs) and ranges from $\sim$0.2∘ (68% containment) for large events events hitting almost all the PMTs to $\sim$1.0∘ for events near the analysis threshold.

Gamma-ray induced showers are generally compact and have a smooth lateral distribution around the shower core (the position where the shower axis intersects the detector plane). In contrast, hadronic background events tend to be broader, contain multiple or poorly defined cores, and include highly localized large signals from muons and hadrons at significant distance from the shower axis. Selection cuts on shower morphology eliminate ${>}99\%$ of the hadronic background in the large event size samples and at least $85\%$ of the background near the analysis threshold, while usually retaining more than 50% of the gamma-ray induced signal events. Details of the data reconstruction, and analysis, and the verification of the sensitivity of the measurement will be presented in a future publication on the observation of the Crab Nebula with the HAWC Observatory (Abeysekara et al., 2017, submitted to ApJ).

3 Methodology

In this section we review the details of the dataset used in the analysis and describe the event selection and the construction of unbiased maps of the viewable sky, which include estimates of the cosmic-ray background rates. From the maps we compute a test statistic (TS) from the ratio of the likelihood that a source is present and the null hypotheses that the observed event population is due to background alone. We identify and localize sources from a list of local maxima in the TS maps with values greater than 25. The procedure is applied to the map to identify pointlike sources as well as sources with characteristic sizes 0.25∘, 0.5∘, 1∘, and 2∘. Many sources, particularly the bright ones, will likely be detected in both the point-source and extended-source maps. We find that there are some extended regions of gamma-ray emission that could either be interpreted as a single extended source or an ensemble of point sources. Below we describe the method employed to detect point and extended sources, to estimate their positions, extents, and spectra and finally discuss the principal sources of systematic uncertainty.

3.1 Dataset

The results presented here are obtained using data taken between 2014-11-26 and 2016-06-02. During this period, $8.8\times 10^{11}$ triggered events were recorded to disk. The full HAWC Observatory was inaugurated in 2015 March. During the construction phase prior to the inauguration, data were collected with a variable number of WCDs ranging from 250 to 300. Overall there was downtime of 40 days (7.2%) during this 553 day period, for the most part related to power issues or scheduled shutdowns for construction or maintenance. In addition, 7 days of data (1.3%) were removed based on requirements regarding the stability of the detector performance. The final livetime used for the analysis is 506.6 days, corresponding to 92% duty cycle.

The data were reconstructed and analyzed with Pass 4, which includes improved calibrations, improved event reconstruction, and improvements in the likelihood framework used for the map analysis. The new event reconstruction benefits from a directional fit using an improved shower model, a new algorithm to separate gamma-ray and hadronic events, and a better electronics model. For comparison, our previous search for sources in the inner Galactic Plane which defined the 1HWC source list (Abeysekara et al., 2016) was performed using 275 days of data taken with a detector consisting of about one third of the full HAWC array and using the Pass 1 analysis. This new pass, combined with the larger detector and longer exposure time, improves the sensitivity of the survey by about a factor of 5 with respect to the Pass 1 inner Galactic Plane search.

3.2 Event Selection

Events are classified by size in nine analysis bins $\mathcal{B}$, presented in Table 1, depending on the fraction $f_{\mathrm{hit}}$ of active PMTs in the detector that participate in the reconstruction of the air shower. We chose to define bins based on the fraction of the detector hit, rather than the absolute number of PMTs, in order to obtain more stable results for the various detector configurations of active WCDs over time.

The selection cuts on the gamma/hadron separation variables are optimized for each bin using observations of the Crab Nebula (Abeysekara et al., 2017, submitted to ApJ). The point spread function (PSF) of the reconstructed events depends on the event size. In Table 1, the $\psi_{68}$ column represents the 68% containment angle of the PSF, for a source similar to the Crab Nebula. Large events have a better PSF, a better hadronic background rejection, and correspond to higher energy primary particles. The efficiency of the gamma/hadron separation cuts is indicated in the $\epsilon_{\gamma}^{\mathrm{MC}}$ and $\epsilon_{\mathrm{CR}}^{\mathrm{data}}$ columns, where the gamma efficiency has been estimated using Monte Carlo simulation of the detector and the hadron efficiency has been measured directly using cosmic ray data. The $\tilde{E}_{\gamma}^{\mathrm{MC}}$ column represents the median energy of the simulated gamma-ray photons in this analysis bin for a source at a declination of 20∘ and for an energy spectrum $E^{-2.63}$ (Crab-Nebula-like source). Events in the same bin for a source with a harder spectrum or at larger declination will tend to have a larger energy on average.

3.3 Event and Background Maps

After reconstruction, event and background maps are generated. The event maps are simply histograms of the arrival direction of the reconstructed events, in the equatorial coordinate system. The background maps are computed using a method developed for the Milagro experiment known as direct integration (Atkins et al., 2003). It is used to fit the isotropic distribution of events that pass the gamma-ray event selection, while accounting for the asymmetric detector angular response and varying all-sky rate. As strong gamma-ray sources would bias the background estimate, some regions are excluded from the computation. These regions cover the Crab, the two Markarians, the Geminga region and, a region $\pm 3$∘ around the inner Galactic Plane. Nine event maps and nine background maps are generated, for the nine analysis bins.

The maps are produced using a HEALPix pixelization scheme (Górski et al., 2005), where the sphere is divided in 12 equal area base pixels, each of which is subdivided into a grid of $N_{\mathrm{side}}\times N_{\mathrm{side}}$. For the present analysis, maps were initially done using $N_{\mathrm{side}}=1024$ for a mean spacing between pixel centers of less than 0.06∘, which is small compared to the typical PSF of the reconstructed events as shown on Table 1.

3.4 Source Hypothesis Testing

The maximum likelihood analysis framework presented in Younk et al. (2016) is used to analyze the maps. The test statistic is defined using the likelihood ratio,

[TABLE]

to compare a source model hypothesis with a null hypothesis. The likelihood of a model $\mathcal{L}(\textrm{Model})$ is obtained by comparing the observed event counts with the expected counts, for all the pixels in a region of interest, and for all nine analysis bins.

For the null model, the expected counts are simply given by the background maps derived from data. For the source model, the expected counts correspond to the same background plus a signal contribution from the source derived from simulation. We assume a source model characterized by:

• •

a point source or a uniform disk of fixed radius and

• •

a power law energy spectrum.

The signal contribution is derived from the source characteristics and the detector response from simulation (expected counts for the spectrum and PSF, both functions of the analysis bin and the declination).

The TS is maximized with respect to the free parameters of the source model. This approach is used both to search for sources (with a TS threshold) and to measure the characteristics of said sources as a result of the maximization.

We make a TS map by moving the location of the hypothetical source across the possible locations in the sky. In the following searches the source flux is the only free parameter of the model while the extent and spectral index are fixed. The source and null model are nested; hence by Wilks’ Theorem the TS is distributed as $\chi^{2}$ with one degree of freedom if the statistics are sufficiently large. Consequently, the pre-trial significance, conventionally reported as standard deviations (sigmas), is obtained by taking the square root of the test statistic, $\sqrt{\mathrm{TS}}$ (here and after, what we denote $\sqrt{\mathrm{TS}}$ actually corresponds to $\textrm{sign}(TS)\,\sqrt{|TS|}$). Figure 2 shows the distribution of $\sqrt{\mathrm{TS}}$ across the sky for the point source search, as well as a standard normal distribution scaled by the number of pixels. For values lower than ${\sim}3$, the $\sqrt{\mathrm{TS}}$ is well reproduced by the normal distribution, whereas at greater values a large excess can be seen due to the presence of sources in the sky.

3.5 Catalog Construction

In order to take advantage of HAWC’s sensitivity to both pointlike and extended sources, multiple searches are conducted assuming either point or extended sources. The TS maps used for the search are computed using a source model consisting of a single test source with a fixed geometry (point source or uniform disk of fixed radius) and an energy spectrum consisting of a power law of fixed index,

[TABLE]

where $E_{0}$ is a reference energy, $F_{0}$ is the differential flux at $E_{0}$ and $\alpha$ is the spectral index.

For the known TeVCat sources that can be considered pointlike given the angular resolution of the HAWC instrument (i.e. the TeVCat extent is of the order of the PSF size or smaller), the spectral indices measured by HAWC vary around $-2.7$, from approximately $-3.1$ to $-2.5$, and are typically softer than the indices listed in TeVCat. This can be explained if the sources soften or cut off at the energies observed by HAWC. On the other hand, the Geminga PWN, which was first observed at TeV energies by the Milagro collaboration (Abdo et al., 2009), is detected by HAWC with an extent of about 2∘ and a hard spectral index around $-2$. To account for the range of source extents and spectra observed with HAWC, four different maps were used to build the catalog, testing various source hypotheses. In order to limit computing time, the resolutions of the maps are adapted to the characteristic dimension of the hypothetical source, without significantly affecting the results:

A point source map of index $-2.7$ (HEALPix map resolution $N_{\mathrm{side}}=1024$ or 0.06∘ per pixel). 2. 2.

An extended source map of radius 0.5∘ and index $-2.0$ ($N_{\mathrm{side}}=512$ or 0.1∘ per pixel). 3. 3.

An extended source map of radius 1.0∘ and index $-2.0$ ($N_{\mathrm{side}}=256$ or 0.2∘ per pixel). 4. 4.

An extended source map of radius 2.0∘ and index $-2.0$ ($N_{\mathrm{side}}=256$ or 0.2∘ per pixel).

When building the catalog, the priority is given to the point source search, then the extended searches ordered by increasing radius. This limits possible source contamination when multiple nearby sources are added together. However, a strong extended source may be found in the point source search, possibly multiple times (see e.g. Geminga below), as well as in the extended search. Hence, the exact search in which a source is first tagged is not a perfect indication of the source extent. More robust morphology studies will be performed in a future analysis and are beyond the scope of this catalog paper.

To select the sources in the maps, all local maxima with $\mathrm{TS}>25$ are flagged. In some regions, multiple local maxima are found very near each other. We define as primary sources all local maxima that are separated from neighboring local maxima of higher significance by a valley of $\Delta(\sqrt{\mathrm{TS}})>2$. We also define and include secondary sources when $1<\Delta(\sqrt{\mathrm{TS}})<2$. These sources are marked with an asterisk (*).

The final catalog comprises the sources of the point source search plus the sources of the extended searches, ordered by increasing radius, if their locations are more than 2∘ away from any hotspot with TS greater than 25 in the previous searches.

3.6 False Positive Expectation

When selecting the sources in the map, a background fluctuation can sometimes mimic a source and fulfill the selection criteria. To estimate this possible contamination, the search was run on randomized background maps. Events maps are generated for each of the nine analysis bins, and then the full search strategy as for the data map is employed, including point and extended source searches, as detailed on Section 3.5. This complete procedure was run with 20 sets of simulated maps. In 11 cases, no sources were flagged. In 9 cases, one source was flagged. In total, out of the 20 full searches performed over the entire sky, 9 sources were flagged, so the predicted number of background fluctuations passing the $\mathrm{TS}>25$ criterion is about $9/20=0.45$. Therefore, the predicted number of false positive in the catalog is about 0.5. These possible fluctuations are typically close to the threshold value $\mathrm{TS}=25$ and are usually out of the Galactic Plane, as it only represents a small fraction of the visible sky.

3.7 Source Position, Extent, and Energy Spectrum

The source positions reported in this catalog correspond to the first search in which they appear, as presented in Section 3.5. The statistical uncertainty of the position is defined as the maximum distance between the center and the 1-sigma contour obtained from the TS map.

After the search, a residual map is generated and halo-like structures are visible around several sources modeled as point sources. This halo is used to define a tentative source radius for the secondary source model when fitting the energy spectrum (results presented in Table 4.3 of the next section). This radius should not be regarded as a definite measurement of the source extent but can nonetheless provide useful information on how much the spectrum measurement depends on the source region definition. When this new source region definition is a good representation of the actual source, the newly fitted spectrum should better correspond to the source spectrum, however as it corresponds to a larger region it is more subject to contamination from other sources or possibly diffuse emission. Additionally, for some complex regions, or regions for which independent analyses are performed, the whole region is fit, explicitly including multiple sources, as an estimate of the total flux of the region. Such regions are discussed in Section 5.

Once the source location and size are defined, the source spectrum is fit using a power law (Equation 2). For the range of declinations considered, the reference energy of 7 TeV minimizes the correlation between the index and normalization, energy which corresponds to the region of maximum sensitivity (cf. Figure 3, right). We report the differential flux at 7 TeV ($F_{7}$), the index $\alpha$, and the statistical uncertainties on both parameters in Table 4.3.

3.8 Diffuse Galactic emission

At GeV energies, diffuse emission resulting from the interaction of cosmic rays with matter and photons is the dominant component of the gamma-ray sky. This diffuse emission has a steeper spectrum than galactic gamma-ray sources and as a result the TeV sky is source dominated. The Milagro and H.E.S.S. experiments measured the TeV diffuse emission in Abdo et al. (2008) and Abramowski et al. (2014). Both measured a higher flux than predicted – by the numerical cosmic-ray propagation code GALPROP (Strong et al., 2007) for Milagro111The conventional GALPROP version here, since the optimized version was derived to fit the EGRET excess which was latter refuted by Fermi-LAT., and a hadronic model for H.E.S.S.–, likely due to unresolved sources. A diffuse emission is not included in the likelihood model used in the present analysis. We are concerned that sources identified by this analysis may have a significant underlying diffuse component, or in extreme cases arise from background fluctuations in a continuous region of diffuse emission. To estimate the maximum possible contribution of the diffuse emission to the spectrum measurement, we simulate a uniform flux with a normalization corresponding to the peak flux value of the hadronic model reported by H.E.S.S. ($1\times 10^{-9}$ TeV*-1* cm*-2* s*-1* sr*-1* at 1 TeV) and a spectral index of $-2.7$. We estimate that, for the low latitude sources near the detection threshold (where the diffuse contribution will be the largest), the diffuse emission can contribute to $<$30% of the fluxes measured with the point source hypothesis.

As an alternative method of estimating the contribution from Galactic Diffuse emission, we can use a region of the Galactic Plane with no detected sources to derive a conservative upper limit on this contribution. As with the analyses by HESS and Milagro mentioned above, this approach will naturally overestimate the diffuse component since it includes unresolved sources. We use the region with longitude $l$ between 56∘ and 64∘ and latitude $|b|<0.5$∘, which does not contain detected sources. The median differential flux at 7 TeV measured in this region with the point source model is 2.1\times 10^{-15}$$\,\textrm{TeV}^{-1}\,\textrm{cm}^{-2}\,\textrm{s}^{-1}. This small excess over a large region indicates the presence of either the Galactic diffuse emission, some unresolved sources, or more likely a combination of both. We use it as an upper limit to estimate the impact of the diffuse on the flux of the sources measured in the plane near $l=60$∘. We extrapolate to lower latitudes using the shape of the longitudinal profile of the diffuse emission from GALPROP in Abdo et al. (2008). We find that in this approach the diffuse emission can contribute up to 60% of the flux measurement of the weak, low-latitude sources (TS close to 25), that have longitudes between 34∘ and 50∘. For $l>50$∘ the modeled diffuse emission is lower, and for $l<34$∘ all the detected sources have higher fluxes and they are not impacted significantly by the diffuse emission. The sources for which this conservative estimate is above 30% of the measured point source flux at 7 TeV are 2HWC J1852+013*, 2HWC J1902+048*, 2HWC J1907+084*, 2HWC J1914+117*, 2HWC J1921+131, and 2HWC J1922+140; as defined and discussed in Sections 4 and 5. In the likely case in which part or most of the flux measured in the $l=[56^{\circ},\,64^{\circ}]$ region indeed contains unresolved sources, the diffuse flux is lesser and so is its contribution of the flux reported on this catalog.

Future dedicated analysis of the HAWC data will allow to better constrain the Galactic diffuse emission.

3.9 Systematic Uncertainties

The absolute pointing of the HAWC Observatory is initially determined using a careful survey of the WCDs and PMTs and then refined using the observed position of the Crab Nebula. The positions of Markarian 421 and Markarian 501 are observed by HAWC within 0.05∘ of their known locations after the pointing calibration. Additional studies based on the observation of the Crab Nebula when it is farther from zenith showed that absolute pointing is still better than 0.1∘ up to a zenith angle of 45∘, which covers the full declination range considered in the present study. Therefore the systematic uncertainty on the absolute pointing of the catalog is quoted as 0.1∘.

For isolated point sources, the systematic uncertainties on the spectrum measurement are estimated to be $\pm 50\%$ for the overall flux and $\pm 0.2$ for the spectral index (Abeysekara et al., 2017, submitted to ApJ). In the present analysis, no detailed morphology study is performed. However, there is a correlation between the assumed source size and the measured spectrum. Simulation studies show that for isolated sources the unknown extent can induce an additional systematic uncertainty on the spectral index measurement of up to 0.3.

As we test the presence of a single source at a time without modeling the other sources, the likelihood computation may be impacted by events from a neighboring source. This is true in particular for the lower energy events where the PSF is wider. By adding events to the single hypothesized source, this contamination can increase the measured flux and make the spectral index softer. In the case of two identical point sources located 1∘ apart, the flux measurement, assuming a known spectral index, is increased by 20% to 30%, depending on the declination. When fitting the index as well, the index can change by up to 0.1 and the measured flux is changed by about 20% to 40%. This confusion is considered a systematic uncertainty of the present analysis and tends to be larger in the very populated regions of the sky with high source population.

4 Results

We present the result of the search, the 2HWC catalog. A total of 39 sources are found222Geminga is flagged twice but only counted as one here., 4 of which are detected with the extended search procedure only. As discussed in Section 3.6, the predicted number of background fluctuations passing the selection criteria is about 0.5. Out of these 39 sources, 16 are more than a degree away from known TeV sources listed in TeVCat.

4.1 HAWC Performance

Due to the development of air showers in the atmosphere, HAWC’s sensitivity as well as energy response varies with the source declination. The sensitivity of the point source search is represented in Figure 3, left. The curves correspond to the flux that gives a central expectation of a 5$\sigma$ signal for a point source with a power law flux of index $-2.0$, $-2.5$, and $-3.0$. The maximum sensitivity is obtained for sources transiting at the zenith of HAWC, i.e. whose declinations are close to 19∘. The sources found in the point source search are also represented here: the measured flux and statistical uncertainty are shown at the corresponding declination.

The energy range that contributes to most of the test statistic in the point source search, derived from simulation, is represented in Figure 3, right. More precisely, assuming a given spectral model, we show the energy range as the energy defining the central 75% of the contribution to the test statistic. Three spectral models are represented: power laws of index $-2.0$, $-2.5$, and $-3.0$. For a given spectral model, the energy range that contributes most of the test statistic shifts to lower values for sources transiting overhead than for sources whose declinations are far from 19∘.

4.2 Maps

The test-statistic map derived from the all-sky search for point sources with index $-2.7$ is presented in equatorial coordinates in Figure 4. The inner Galactic Plane is clearly visible. In the outer Galactic Plane, the Crab and Geminga are visible. Outside of the Galactic Plane, Markarian 421 and Markarian 501 stand out.

Figures 5 to 9 show detailed views of smaller regions of the sky. 2HWC sources are represented by white circles and labels below the circle. The source locations listed in TeVCat are also marked, with black squares and labels above the square symbol.

The maps of the regions around the Crab, Markarian 421, and Markarian 501 are shown in Figure 5. The region of the outer Galactic Plane around Geminga is mapped in Figure 6. The left map shows the result of the point source search; the right map that of the 2∘ extended search. The increased TS in the extended search supports the case of a significant extent of the two TeV sources detected by the HAWC Observatory in this region. Isolated sources found out of the Galactic Plane are shown on Figure 7. Finally, the inner Galactic Plane from the Cygnus region towards the center of the Galaxy is shown in Figures 8 and 9.

4.3 Catalog

Table 4.3 lists all sources found using the procedure described in Section 3.5, ordered by right ascension. The first column lists the HAWC catalog name. The second column specifies the search in which the source first appeared with a TS above the threshold value of 25. PS denotes the point source search, 0.5, 1, and 2∘ the radius of the disk in the extended search. The corresponding TS value is reported in the third column. The following columns compile the source positions in equatorial (J2000.0 epoch) and Galactic coordinates and the one-sigma uncertainty on the position of the maximum identified in the respective search. The second part of the table, after the vertical line, provides information on the nearest TeVCat source: the distance, then the corresponding name if this distance is less than 1∘.