THE FIRST DETECTION OF GeV EMISSION FROM AN ULTRALUMINOUS INFRARED GALAXY: Arp 220 AS SEEN WITH THE FERMI LARGE AREA TELESCOPE

Nearby star-forming and starburst galaxies have been identified to be GeV to TeV gamma-ray sources (Acero et al. 2009; VERITAS Collaboration et al. 2009; Abdo et al. 2010; Ackermann et al. 2012). Cosmic rays (CRs) accelerated by supernova remnants interact with the interstellar gas and produce neutral pions (schematically written as $p+p\to {\pi }^{0}+$ other products), which in turn decay into high-energy gamma-rays (${\pi }^{0}\to \gamma +\gamma$). High energy gamma-ray emission is thus a powerful tool to diagnose the non-thermal emission related with CRs in star-forming galaxies.

Arp 220, at a distance of d = 77 Mpc, is the nearest ultraluminous infrared galaxy (ULIRG), with a total infrared luminosity of (1–2) × 10¹² L_⊙ (Sanders et al. 2003; Gao & Solomon 2004; Rangwala et al. 2011). The system is a merger of two galaxies and contains two dense nuclei separated by 350 pc. Both nuclei have high star formation rates and dense molecular gas. Arp 220 has a molecular gas mass of 10¹⁰ M_⊙ (Scoville et al. 1997) and a total star formation rate of 240 ± 30 M_⊙ yr⁻¹, calculated with the far-infrared luminosity (Farrah et al. 2003). Radio detection of supernovae suggests a high rate of 4 ± 2 yr⁻¹ (Smith et al. 1998; Lonsdale et al. 2006; Parra et al. 2007). Although Arp 220 could contain active galactic nuclei (AGNs) (Paggi et al. 2013; Teng et al. 2015), the observed high supernova rate indicates that star formation provides a substantial fraction of the power radiated by the nuclei.

As a huge reservoir of molecular gas, the nuclei of Arp 220 should be complete proton calorimeters, i.e., the CR protons lose all of their energy to secondary pions via collisions with the gases before escaping (Torres 2004; Thompson et al. 2007; Lacki et al. 2011; Yoast-Hull et al. 2015). Given a high CR emissivity and a high efficiency of converting CR energy into gamma-rays, Arp 220 was predicted to emit GeV to TeV gamma-rays (Torres 2004; Lacki et al. 2011; Yoast-Hull et al. 2015). There have been intense efforts to search for GeV emission from Arp 220 (Lacki et al. 2011; Ackermann et al. 2012; Tang et al. 2014). Tang et al. (2014) obtained an upper limit of F_{0.1−100 GeV} < 3.5 × 10⁻⁹ ph cm⁻² s⁻¹ at the 95% confidence level using 68 months of Fermi/LAT data. Arp 220 was also observed by the VERITAS telescopes at TeV energies for more than 30 hr, but no significant excess over the background was found (Fleischhack et al. 2015). Here we report the detection of GeV emission from Arp 220 using 7.5 years of Fermi/LAT data reprocessed with Pass 8, the newest event-level analysis, which significantly improves the acceptance and angular resolution of the instrument.⁶

We present a description of the analysis and show the results in Section 2. Then we discuss the origin of the γ-ray emission in Section 3 and use it to constrain the CR acceleration efficiency in Section 4. Finally we give the conclusions in Section 5.

2.1. Data Selection

2.2. Spatial and Spectral Analysis

We start with a preliminary unbinned likelihood analysis on the ROI with the tool gtlike, only including "gll_iem_v06.fits," "iso_P8R2_SOURCE_V6_v06.txt" and point sources of 3FGL within 15° from the ROI center in the source model file. The residual test-statistic (TS) under this scenario is shown in the upper panel of Figure 1, which shows a gamma-ray excess around Arp 220. We add one point source with a power-law spectrum at the local maxima of the TS map in the source model file. We re-fit the data and find the best-fit γ-ray emission position with the tool gtfindsrc. We then change the position of the point source to the obtained best-fit location of γ-ray emission in the source model file. Repeating the standard analysis threads and computing the TS map, although we find a significantly improved fit, we note that there is still some residual emission to the northwest of the ROI center. In order to explore the high-energy emission more accurately near Arp 220 and reduce the contamination from nearby background sources, a second point source is added to the source model based on the residual γ-ray emission. By utilizing the tool gtfindsrc, we optimize the position of these two point sources, namely P1 (233677, 235163) ± 0054, which is near Arp 220 and P2 (233239, 238049) ± 0168, which is located at a position about 05 away from Arp 220, as marked in Figure 1. Next, we re-fit the data with point sources P1 and P2 in the source model file, and get a better result with Δ likelihood = 8 (i.e., we get a global improved fitting with significance TS = 16), in comparison with only one point source being added into the source model file. The significance of TS_P2 = 22 suggests that P2 should be a genuine additional γ-ray source. The bottom panel in Figure 1 illustrates that there is little residual γ-ray emission near the vicinity of Arp 220 region. The source model file including point sources P1 and P2 will be used in all the subsequent analysis.

Zoom In Zoom Out Reset image size

We now check possible alternative candidates for P1 and P2 in the following catalogs: the CRATES Flat-Spectrum Radio Source Catalog (Healey et al. 2007), the Veron Catalog of Quasars AGN, the 13th Edition (Véron-Cetty & Véron 2010), and the Candidate Gamma-Ray Blazar Survey Source Catalog (Healey et al. 2008). No other sources are found in the vicinity of r₉₅ of P1 except for Arp 220. With only a separation of 0058 between P1 and Arp 220, it is reasonable to ascribe the high-energy emission from P1 to Arp 220. In contrast, the separation between P2 and Arp 220 reaches 0547 and there are four candidates around r₉₅ of P2 in the above three catalogs: CRATES J153246+234400 (4902), an FSRQ with unknown redshift; SDSS J15323+2345 (9080), a QSO with redshift z = 1.465; SDSS J15337+2358 (14643), a Seyfert 1 galaxy with z = 0.067; and 1WGA J1533.8+2356 (15315), an AGN of unknown type located at z = 0.232. The values in the parentheses are the angular offsets of these sources from P2. We suspect that the gamma-ray emission from P2 is contributed by one or more of these AGNs.

We assume power-law spectra in the range of 0.2–100 GeV for both P1 and P2 in the aforementioned best source model files. The best-fit photon index for emissions of P1 is Γ₁ = 2.35 ± 0.16 with a total flux of F₁ = 1.76 ± 0.52 × 10⁻⁹ ph cm⁻² s⁻¹. Regarding P2, the photon index is Γ₂ = 2.45 ± 0.19 and the total flux is F₂ = 1.45 ± 0.52 × 10⁻⁹ ph cm⁻² s⁻¹. The corresponding energy fluxes in 0.2–100 GeV are 1.92 ± 0.43 × 10⁻¹² erg cm⁻² s⁻¹ and 1.39 ± 0.40 × 10⁻¹² erg cm⁻² s⁻¹ for P1 and P2, respectively. The spatial and spectral results of P1 and P2 are summarized in Table 1. The spectral energy distribution (SED) of P1 (Arp 220) is shown in Figure 2. We do not discuss P2 further in this paper since our focus is on the gamma-ray emission from Arp 220.

Zoom In Zoom Out Reset image size

Table 1.  The Best-fit Spectral Parameters of the Point Sources around the Arp 220 Region for Energy Band 0.2–100 GeV

Point	Position	r95	Separation	Photon Flux	Energy Flux	Γ	TS	Association
	(degree)	(degree)	(degree)	(10−9 ph cm−2 s−1)	(10−12 erg cm−2 s−1)
P1	(233.677, 23.5163)	0.090	0.058	1.76 ± 0.52	1.92 ± 0.43	2.35 ± 0.16	40	Arp 220
P2	(233.239, 23.8049)	0.279	0.547	1.45 ± 0.52	1.39 ± 0.40	2.45 ± 0.19	22	⋯

Note. The fourth column is the angle distance between the best-fit position of γ-ray excess and the position of Arp 220.

Download table as:  ASCIITypeset image

Gamma-ray emissions in star-forming galaxies are usually thought to originate from CR protons via collision with the gases inside the galaxies. This may especially be the case with Arp 220, as it has a high supernova rate and high gas density. Alternatively, Arp 220 could have hidden AGNs (Paggi et al. 2013; Teng et al. 2015), which in principle could give rise to gamma-ray emission as well. CR-induced gamma-ray emission in a star-forming galaxy is expected to be stable, while gamma-ray emission induced by AGNs should be temporally variable (Gasparrini et al. 2015). Thus, we examine the flux variability of Arp 220 (P1). We create a set of time bins for the light curve of photons with energy >400 MeV. In the first trial, the full observation period is divided linearly into five equal time bins. Each time bin is fitted by a separate maximum likelihood analysis, resulting in each bin having a detection significance TS > 4. We further check a finer light curve with 12 time bins. The χ² goodness-of-fit test is consistent with a constant flux with a reduced χ² of 0.83 for data points with a detection significance TS > 1. The results are shown in Figure 2. None of the two cases show significant changes in flux over the period of LAT observations.⁷ Generally speaking, the temporal behavior of GeV emission from Arp 220 resembles that of other starburst galaxies, such as M82 and NGC 253.

Meanwhile, there is a clear positive empirical relation between the γ-ray luminosity L_{0.1−100 GeV} and total infrared luminosity L_{8−1000 μm} in local group galaxies and nearby star-forming galaxies⁸ (Ackermann et al. 2012). With a pure sample of star-forming galaxies, Tang et al. (2014) have drawn a similar relation and extended the relation to a luminous infrared galaxy NGC 2146. The best simple linear fit for nearby star-forming galaxies excluding Arp 220 gives a relation of

$$\begin{eqnarray}&&{\rm{log}}\left(\displaystyle \frac{{L}_{0.1-100\mathrm{GeV}}}{\mathrm{erg}\;{{\rm{s}}}^{-1}}\right)=\alpha +\beta {\rm{log}}\left(\displaystyle \frac{{L}_{8-1000\mu {\rm{m}}}}{\mathrm{erg}\;{{\rm{s}}}^{-1}}\right).\end{eqnarray} \tag{ 1 }$$

(see Figure 3), where α = −14.98 ± 4.87 and β = 1.24 ± 0.11 are the intercept and slope, respectively. The Pearson's correlation coefficient is 0.98, with a chance probability of p ∼ 10⁻⁴. With a luminosity of L_{0.1−100 GeV} = 1.78 ± 0.30 × 10⁴² erg s⁻¹, Arp 220 is consistent with this quasi-linear relation. It is also consistent with the scaling relation between the gamma-ray luminosity and radio luminosity found for star-forming galaxies (Ackermann et al. 2012). The fact that the gamma-ray emission has no significant variability over the observation period and it is consistent with the above scaling relation suggests that the gamma-ray emission is likely to arise from CR interactions.

Zoom In Zoom Out Reset image size

As the ejecta of a SN encounter the ambient interstellar medium (ISM), a strong shock will be generated and a fraction of the kinetic energy of the ejecta will be transferred into CRs. The gamma-ray luminosity of a galaxy depends on both the efficiency of converting SN kinetic energy to CRs and the efficiency of converting CR energy to gamma-rays (i.e., the proton–proton collision efficiency). Because of the very dense ISM in Arp 220, it has been widely suggested that Arp 220 is an ideal proton calorimeter, since the collisional energy loss time of CR protons is much shorter than the time that CRs escape out of the galaxy (Lacki et al. 2011; Yoast-Hull et al. 2015). Then its gamma-ray luminosity provides a direct tool to measure the efficiency that SN explosion power goes into high-energy CRs.

Assuming that SNRs are the sole source of CR protons, and that a constant fraction (η) of SN kinetic energy is transferred into CR protons with energy above 1 GeV, the total injected CR power reads

$$\begin{eqnarray}&&{L}_{{\rm{CR}}}(\gt 1\;{\rm{GeV}})=1.3\times {10}^{44}\;{{\rm{erg}}{\rm{s}}}^{-1}{E}_{51}\eta \left(\displaystyle \frac{{{\rm{\Gamma }}}_{{\rm{SN}}}}{4\;{\mathrm{yr}}^{-1}}\right),\end{eqnarray} \tag{ 2 }$$

where E is the SN kinetic energy, which is typically 10⁵¹ erg, and Γ_SN = 4 ± 2 yr⁻¹ is the SN rate obtained by the radio observations (Smith et al. 1998; Lonsdale et al. 2006; Parra et al. 2007). On the other hand, by assuming that the system reaches a steady state, one can derive the injected CR power from the observed gamma-ray luminosity. As leptonic emission is expected to become increasingly important at lower energies, contaminating the estimate of the power in CR protons, we only consider the γ-ray emission above 1 GeV. Then we have

$$\begin{eqnarray}&&{L}_{{\rm{CR}}}(\gt 1\;{\rm{GeV}})=3{L}_{\gamma }(\gt 1\;\mathrm{GeV})({\rm{\Gamma }}-1){\beta }_{\pi }^{-1}.\end{eqnarray} \tag{ 3 }$$

Here the prefactor 3 is due to that only 1/3 of the lost CR power goes into gamma-rays, as inelastic proton–proton collisions produce both neutral and charged pions. The factor Γ − 1 arises from the fact that a fraction (Γ − 2)/(Γ − 1) of the energy of CRs above 1 GeV is transferred to lower energy CRs, where Γ is the CR spectral index, which is equal to the photon index of the gamma-ray emission in our case. Since not all the pionic gamma-rays produced by CRs will have energies above 1 GeV, we need another correction factor β_π ≃ 0.6 (for Γ = −2.35) to account for this effect (Lacki et al. 2011). Given that the luminosity of Arp 220 above 1 GeV is L_γ(>1 GeV) = (0.75 ± 0.28) × 10⁴² erg s⁻¹, we can obtain the CR acceleration efficiency from Equations (2) and (3), i.e.,

$$\begin{eqnarray}&&\eta \simeq (4.2\pm 2.6)\%{E}_{51}^{-1}{\left(\displaystyle \frac{{\beta }_{\pi }}{0.6}\right)}^{-1}{\left(\displaystyle \frac{{{\rm{\Gamma }}}_{{\rm{SN}}}}{4{\mathrm{yr}}^{-1}}\right)}^{-1}.\end{eqnarray} \tag{ 4 }$$

Note that here only the uncertainties in Γ_SN and L_γ(>1 GeV) are taken into consideration to estimate the uncertainty in η. To our knowledge, this is the first time that such an efficiency has been obtained directly from gamma-ray observations. The obtained efficiency in Arp 220 is consistent with estimates of a 3%–10% efficiency in the Milky Way (Strong et al. 2010) and a 6% efficiency in NGC 253 and M82 (Paglione & Abrahams 2012).

Through the analysis of 7.5 years of Fermi/LAT observations, we find high-energy gamma-ray emission above 200 MeV from Arp 220 with a detection significance of about 6.3σ. This is the first time that GeV emission has been detected from an ULIRG. The reconstructed energy spectrum is best modeled by a single power law with a photon index Γ = 2.35 ± 0.16 and the integrated flux in 0.2–100 GeV is F_{0.2−100 GeV} = (1.76 ± 0.52) × 10⁻⁹ ph cm⁻² s⁻¹. The fact that no evidence of significant variability is found and that the gamma-ray emission is consistent with some scaling relations for star-forming galaxies implies that the gamma-ray emission of Arp 220 is likely from CR-induced diffuse emission. As Arp 220 is an ideal proton calorimeter, the gamma-ray luminosity can be used to directly measure the CR acceleration efficiency by SNRs, which gives an efficiency of 4.2 ± 2.6% for CRs above 1 GeV.

We thank Anton Prosekin and Gwenael Giacinti for useful discussions and the anonymous referee for a helpful report. This work has made use of data and software provided by the Fermi Science Support Center. It has also made use of the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. This work is supported by the 973 program under grant 2014CB845800, the NSFC under grants 11273016 and 11033002, and the Excellent Youth Foundation of Jiangsu Province (BK2012011). J.W. acknowledges support from NSFC grant 11473021 and the Fundamental Research Funds for the Central Universities under grants 20720150168 and 20720160023.