betaFeedback

WorksPhysics and Astronomy

HAWC, Fermi-LAT and Chandra Detections of High Energy Particles in the S106 Star Forming Region

Abstract

Local star-forming regions have recently been the subject of scrutiny as cosmic-ray accelerators and sources of ultra-high-energy gamma-ray emissions. The Cygnus-X region is one of the richest star forming regions in the galaxy and over the last few decades has been extensively studied. Here we report the detection of a star-forming region located in the southern part of Cygnus-X, Sharpless 106 (S106), using the HAWC Observatory. The S106 region is spatially located near the cataloged gamma-ray source 1LHAASO J2027+3657. In this work, we perform a multi-wavelength analysis of S106, by combining HAWC data with re-analysis of Fermi-LAT and archival Chandra data. We construct a fully leptonic model for the multi-wavelength emission from S106 from a single population of high-energy leptons, with the gamma-rays greater than 1 TeV being produced by inverse Compton scattering of the locally-produced IR photon field from S106 itself. Our models show that hadronic mechanisms are not always required to explain the observed flux from gamma-ray star-forming regions and that understanding the specific details of local photon fields is crucial to modeling the multi-wavelength emission from star-forming regions.

(HAWC Collaboration)

Corresponding author: R. Babu, S. DiKerby, T. Ergin, S. Zhang, M.U. Nisa

[email protected], [email protected], [email protected], [email protected], [email protected]

Introduction

Recent surveys by Very High Energy (VHE; 100 GeV–100 TeV) and Ultra High Energy (UHE; >100>100 TeV) instruments like the High-Altitude Water Cherenkov Observatory (HAWC) [10] and the Large High Altitude Air Shower Observatory (LHAASO) [24] have discovered a substantial population of TeV and PeV sources concentrated in the galactic plane, but the nature of these objects is still uncertain. Given that these gamma-ray sources emit particles and photons at energies far exceeding those in human-made particle accelerators, modeling the processes that produce their gamma-ray emission can contribute to the identification of new astrophysical laboratories to test the limits of our fundamental theories.

Several ongoing surveys are conducting low-energy investigations of promising galactic gamma-ray sources that lack confident association with known counterparts. Studies over the past decade have revealed that galactic gamma-ray sources can be attributed to a wide range of astronomical phenomena, including pulsars, pulsar wind nebulae (PWN), and supernova remnants (SNR). Despite wide-ranging interest, many gamma-ray sources remain “unassociated”, lacking confident low-energy counterparts. Identifying the low-energy counterparts of unassociated gamma-ray sources is a vital step towards completing the census of extreme systems within our own galaxy.

Whether high-energy systems produce observed gamma rays via leptonic – synchrotron, inverse Compton (IC), Bremsstrahlung – or hadronic – pion decay, proton synchrotron – processes is a particularly timely question, given the recent discovery of high-energy neutrino fluxes from flaring blazars [34], the galactic plane [25], and the nearby Seyfert galaxy NGC 1068 [35]. High-energy neutrino emission requires hadronic processes that produce or destroy leptons, creating neutrinos to conserve lepton number. The identification of galactic gamma-ray emitters as hadronic can help explain the observed galactic neutrino flux.

Giant molecular clouds are the birthplaces of stars and hosts to young stellar objects. These star-forming regions (SFRs) often display strong infrared emission, non-thermal radio signatures, and maser activity, which serve as key tracers of on-going star formation. The production of gamma-rays in such environments can occur due to interactions between accretion processes, collimated jets, stellar winds, and ambient material [47, 30]. In these ways, interactions between cluster-driven cosmic rays and surrounding dense gas can produce diffuse gamma-ray emission [5, 4, 7] (Saha et al. 2020).

Besides high-energy processes related to first-order Fermi acceleration, efficient particle acceleration can also produce non-thermal radiation from radio to X-ray energies. The power distributed into a SFR from even a single young stellar system is considerable; in massive star binaries, a substantial fraction of the wind kinetic power (exceeding ∼1033\sim10^{33} erg/s) can be transferred to high-energy particles and detectable gamma-ray emission (Tavani et al. [50]; Bednarek & Pabich [20]; Reitberger et al. [45]; Pshirkov [43]). In addition, protons might escape the system, interacting with nearby ambient matter. The collective activities of massive stars may form super-bubbles (e.g. Cygnus OB2 or the Cygnus Cocoon), where the acceleration of multiple shocks can raise the maximum energy of CR protons above 1 PeV (Aharonian et al. [7]; LHAASO Collaboration [38]).

The gamma-ray source 1LHAASO J2027+3657 (henceforth J2027) is an unassociated source (Cao et al. [24]) detected in the UHE KM2A component of LHAASO but not by the WCDA instrument. J2027’s position in the outer galactic plane, its moderate flux (N0=(5.0±0.6)×10−17N_0 = (5.0 \pm0.6) \times10^{-17} /s/cm2^2/TeV and E0=50E_0 = 50 TeV) but steep power-law slope (Γ=3.21±0.17\Gamma= 3.21 \pm0.17), and lack of counterparts make it a prime target for low-energy follow-up and analysis. In particular, determining whether the gamma-ray emission from J2027 is leptonic, hadronic, or a combination of both is a crucial step in evaluating its potential contribution to the galactic plane neutrino flux observed in IceCube (Collaboration*† et al. [25]).

A possible counterpart to J2027 is the nearby bipolar star-forming and HII region Sharpless 106 (henceforth S106). It is located at a distance of ∼1.4\sim1.4 kpc (Adams et al. [6]). Comerón et al. [27] revealed that the central object in the region, S106IR, is a close binary system, possibly responsible for the two-lobed structure of S106 as a whole. S106 also hosts a collection of young stars, many of which are strong X-ray emitters previously studied in Giardino et al. [31], and there may be diffuse X-ray flux from the collective interacting stellar winds of stars in S106.

The Fermi-LAT source catalog (4FGL-DR4, Ballet et al. [17]) includes the point-like gamma-ray source 4FGL J2026.5+3718c, which positionally coincides with S106, but is ∼20′\sim20' away from the J2027’s position as measured by the LHAASO-KM2A. Re-analysis of this archival Fermi-LAT data, along with the HAWC data analysis presented in this work, will more tightly constrain the gamma-ray landscape around S106 beyond the spectral point provided in the LHAASO catalog.

In this work we report analysis of a range of high-energy data to build a multiwavelength spectrum near J2027, testing what connection J2027 might have with S106 and other gamma-ray sources detected in Fermi-LAT and HAWC. In section 2 we describe the analysis of HAWC, Fermi-LAT, and Chandra data to construct a broadband SED for J2027. In section 3 we discuss several multiwavelength models linking the observed electromagnetic emission to underlying particle populations. Finally, in section 4 we present overall conclusions.

Observations and Data Analysis

HAWC Analysis

HAWC Data and Analysis Setup

To search for emission from J2027 in HAWC, we analyze ∼2900\sim2900 days of HAWC Pass 5 data via the neural network (NN) energy estimator above 1 TeV. This dataset includes on-array events, as well as off-array events with reconstructed shower cores falling off the main array up to 1.5 times its physical area, as discussed in Abeysekara et al. [2] and Albert et al. [9]. This dataset has stricter gamma-ray/hadron separation cuts based on a machine-learning algorithm, which further improves our sensitivity at lower energies (Alfaro et al. [12]).

We perform a binned maximum likelihood analysis of the data using the threeML (Vianello et al. [52]) with HAWC Accelerated Likelihood (HAL) plugin hawc-hal (Abeysekara et al. [3]). To quantify the goodness of fit of the sources in the region, we use a test statistic (TS), defined as:

TS=2×(ln⁡L1−ln⁡L0)(1)TS = 2 \times(\ln L_1 - \ln L_0) \tag*{(1)}

where L1L_1 and L0L_0 are the likelihoods of the presence (alternate hypothesis) and absence (null hypothesis) of a source respectively. Following Wilks’ theorem, we express the pre-trial significance as TS\sqrt{TS}. We use the source search pipeline outlined in Alfaro et al. [13]. In this work, we enhance the point source search stage by integrating the image processing driven source seed finder introduced in Babu et al. [16].

The HAWC pre-trial significance at the location of J2027 is ∼6.6σ\sim6.6\sigma in the NN energy estimator map as shown in Figure 1. The region of interest used for the modeling is from l∈[72.5∘,77.5∘]l \in[72.5^\circ, 77.5^\circ] and b∈[−4∘,4∘]b \in[-4^\circ, 4^\circ] in Galactic coordinates. This field of view includes one of the brightest sources in HAWC’s Field of View (FoV), MGRO J2019+368 [11].

Figure 1. HAWC pre-trial significance Map of the region created using the NN energy estimator map including off-array showers. This map is created assuming a point-like source morphology and a power-law type spectrum having a spectral index of 2.7. The galactic plane (b=0°) is shown in gray dashed lines. The orange label shows the LHAASO location of J2027.

Galactic Diffuse Emission Modeling

The Galactic plane is a sophisticated region of diffuse gamma rays, originating from the interaction of CRs with interstellar gas and radiation fields. The diffuse emission coupled with unresolved gamma-ray sources is also known as the Galactic Diffuse Emission (GDE). Since the significance of HAWC J2027 in the NN energy estimator map is ∼6.6σ\sim6.6\sigma, and the source is at a low galactic latitude, special attention must be paid to the GDE. This analysis employs the use of an energy-dependent GDE template generated with the HERMES software [29], incorporating emission from different radiative processes, Inverse Compton (IC) scattering, Bremsstrahlung, and neutral pion (π0\pi^{0}) decay from cosmic-ray interactions with atomic (HI) and molecular (H2_{2}) hydrogen. The HI and H2 gas maps used as spatial templates are derived from the HI4PI survey [33] and 12^{12}CO Dame Survey [28], respectively. For the (π0\pi^{0}) decay channel, we have adopted the cross-section calculations from Kelner et al. [37], as they are valid for the energy ranges seen by HAWC. A free scaling parameter is applied to the GDE template within the ROI, to allow adjusting its contribution to the local variations in the cosmic-ray density in the overall description of the region.

Modeling the GDE and S106 Emission in HAWC Data

The diffuse emission HERMES profile for the region is shown in the left panel of Figure 2. It is expected that the density of 12^{12}CO (for the molecular hydrogen) and atomic hydrogen is very high at and around the locations of the SFRs. TeV emission observed by HAWC appears as a diffuse morphology source in HAWC’s FoV. Modeling the HAWC J2027 source with a gaussian spatial morphology, along with the full radiative emission profile to account for the GDE over the full region, results in a post-trial detection of S106 at 4.8σ4.8\sigma, just below the 5σ5\sigma threshold for source detection in HAWC. From the fit results, we find that this is attributed to the contamination by the bright source MGRO J2019+368 to the overall diffuse emission in the region.

HERMES diffuse emission profile and Planck dust map

Figure 2. Left Panel: The HERMES diffuse emission profile templates used in this analysis. The HERMES template for the ICS+π0\pi^{0}(HI+H2_{2}) emission profile along with the location of 1LHAASO J2027+3657 (white) and the 0.5° mask at the centroid of the peak 12^{12}CO emission profile (red). Right Panel: Planck dust opacity map of the region surrounding the dense 12^{12}CO distribution spatially overlapping with S106’s position.

In order to separate the dense molecular gas emission that correlates with the position of S106 and its vicinity from the diffuse and less dense gas emission in the full ROI, we define a masked emission profile template that masks out a circular region around the location of the dense cloud region near S106 with a radius of 0.5∘0.5^\circ from the full HERMES diffuse emission profile. This is shown in the left panel of Figure 2 with a red circle.

Implementing this masked template captures the total gamma-ray emission carrying the fingerprints of the accelerated CRs within this SFR correlating with the molecular clouds in the S106 region.

S106 Gas Template Model based on Planck Dust Maps

Assuming that the gamma-ray excess that we observe near S106 is linked to the dense 12CO^{12}\mathrm{CO} clouds near the SFR, we fit a model for J2027 based on the 12CO^{12}\mathrm{CO} observations of the region within the FoV to check and confirm whether the TeV emission observed by HAWC is spatially extended. For the TeV extension model, we use a 2D spatial template based on GNILC Planck dust opacity map at 353 GHz shown on the right panel of Figure 2. The Planck dust opacity map is chosen as it is a more recent observation than the 12CO^{12}\mathrm{CO} Dame survey. This template also provides information on HI emission, as well as CO emission, because it shows the optically thick component of the total gas content. The top panel of Figure 3 shows the 2D spatial distribution of the 12CO^{12}\mathrm{CO} clouds in the vicinity of J2027.

The 3D spatial distribution of the $^{12}\mathrm{CO}$ molecular clouds in the ROI from the Dame survey

Figure 3. The 3D spatial distribution of the 12CO^{12}\mathrm{CO} molecular clouds in the ROI from the Dame survey [28]. The 12CO^{12}\mathrm{CO} region around S106 in black is not surrounded by excess gas in front and behind its location.

Since the Planck survey does not look into the 3D distribution of dust and clouds, we did a cross check to see if there are any 12CO^{12}\mathrm{CO} cloud distributions either in front or behind the S106 region using the 12CO^{12}\mathrm{CO} Dame survey data. The bottom panel of Figure 3 shows the spatial velocity distribution of the 12CO^{12}\mathrm{CO} clouds in the direction of S106. From this figure, we see that there is an excess of 12CO^{12}\mathrm{CO} clouds distribution within line of sight velocity range of −10<VLSR/(km/s)<5-10 < V_{\mathrm{LSR}}/(\mathrm{km}/\mathrm{s}) < 5, which corresponds to a near and far-kinematic distance of 0.0–5.7 kpc calculated using the tool described in Wenger et al. [53].

HAWC Modeling Results

Applying the source-finding algorithm described in Babu et al. [16] and using the masked HERMES diffuse emission profile template, our best fit model consists of four individual sources: in addition to J2027, we also detect one bright extended source (HAWC J2019+3646 (Albert et al. [11]) and two point-like sources (HAWC J2016+3711 and the newly discovered HAWC J2018+3610). Notably, HAWC J2018+3610 is a new unidentified TeV point source.

Initially, we used a point-like source model for J2027, which resulted in the detection of the source above 5σ5\sigma but for all the sources, we also tested alternative spectral and morphological models as described in Alfaro et al. (2025b). During the morphology test for J2027, the model with the spatial template derived from the Planck dust map (described in § 2.1.4) was preferred over the point-like morphology. We evaluated this preference using the Bayesian Information Criterion (Schwarz [48]) and the Akaike Information Criterion (Akaike [8]). The spatial gas template model is favored over the point-like morphology by a ΔBIC=BICgas−BICpoint=−31\Delta\mathrm{BIC} = \mathrm{BIC}_{\mathrm{gas}} - \mathrm{BIC}_{\mathrm{point}} = -31 and ΔAIC=AICgas−AICpoint=−6\Delta\mathrm{AIC} = \mathrm{AIC}_{\mathrm{gas}} - \mathrm{AIC}_{\mathrm{point}} = -6, where negative values indicate the preference for the spatial gas template because of its lower values.

The upper left panel of Figure 4 shows the sources found in the HAWC analysis and J2027 modeled with the Planck template shown in white contours. Meanwhile, the upper right panel of Figure 4 shows the residual map after the subtraction of model from the data map and the lower right panel of Figure 4 is the one dimensional binned version of the residual map, which demonstrates that the residual map is devoid of any significant excess, confirming the goodness-of-fit of our model.

Figure 4. Top Left Panel: The HAWC significance map is showing the location and morphology of the fitted sources in the ROI of J2027. The white contours plotted on HAWC J2027+3718 show the gas morphology represented by the Planck dust template, as described in § 2.1.4. Top Right Panel: The HAWC residual significance map of the J2027 ROI after modeling all the sources discussed in § 2.1.5, indicating no excess and no over-subtraction in the modeling. Bottom Left Panel: The zoomed-in significance map of HAWC J2027+3718, where its color-bar range goes up to 5.1σ\sigma. 12CO{}^{12}\mathrm{CO} emission contours of 8, 16, 24, 32 K km/s for [-10, 5] km/s taken from the 12CO{}^{12}\mathrm{CO} Dame survey are shown in cyan colors. In addition, the small black dashed circle shows the position and rough extension of S106 in IR, while the large white dashed circle shows the LHAASO positional error ellipse of J2027. Bottom Right Panel: The significance histogram (a 1D version of the significance map shown in the top right panel) showing no significant excess remains in the ROI after modeling all sources.

The best-fit parameters of HAWC J2027, the source of interest, are shown in Table 1, while the best-fit parameters of the other three sources are summarized in Table 2. All sources are detected with a significance above our detection threshold of TS>5σ\sqrt{\mathrm{TS}} > 5\sigma.

ParameterPoint SourcePlanck Template
RA [deg]306.98±0.04306.98 \pm0.04–
Dec [deg]37.30±0.0337.30 \pm0.03–
Φ0\Phi_0 [(TeV cm2^2 s)−1^{-1}]4.4−1.7+2.74.4^{+2.7}_{-1.7}4.8−0.9+1.1 −1.4+2.34.8^{+1.1}_{-0.9}\ ^{+2.3}_{-1.4}
α\alpha2.499±0.182.499 \pm0.182.61±0.11 −0.19+0.212.61 \pm0.11\ ^{+0.21}_{-0.19}
TS28.628.63131
BIC262102.53262102.53262072.0262072.0
AIC261847.48261847.48261841.2261841.2

Table 1. Best-fit values of HAWC J2027+3718

ParameterJ2019+3646J2016+3711J2018+3610GDE
RA [deg]304.89 ± 0.01 −0.017+0.018^{+0.018}_{-0.017}304.03 ± 0.02 ± 0.043304.57 ± 0.05 −0.057+0.071^{+0.071}_{-0.057}-
Dec [deg]36.766 −0.06+0.05^{+0.05}_{-0.06} −0.01+0.01^{+0.01}_{-0.01}37.19 ± 0.02 −0.033+0.035^{+0.035}_{-0.033}36.173 ± 0.03 −0.046+0.052^{+0.052}_{-0.046}-
σ\sigma [deg]0.361 ± 0.01 −0.014+0.019^{+0.019}_{-0.014}---
eccentricity0.835 ± 0.013 −0.023+0.024^{+0.024}_{-0.023}---
θ\theta15.6 ± 1.8 −2.8+3.3^{+3.3}_{-2.8}---
Φ0\Phi_0 (TeV cm2^2 s)−1^{-1}69 −6+9^{+9}_{-6}10.6 −2.2+2.8^{+2.8}_{-2.2} −3.3+11.9^{+11.9}_{-3.3}9.9 −2.2+2.8^{+2.8}_{-2.2} −3.4+11.9^{+11.9}_{-3.4}4.02 ± 0.24 −0.4+0.46^{+0.46}_{-0.4}
α\alpha1.57 ± 0.1 −0.079+0.431^{+0.431}_{-0.079}2.61 ± 0.1 −0.16+0.31^{+0.31}_{-0.16}2.67 ± 0.11 −0.18+0.3^{+0.3}_{-0.18}-
EcE_{\mathrm{c}}28.2 −3.4+4^{+4}_{-3.4}---
TS21537941309

Table 2. Best-fit values for the three other sources with HAWC Data in the ROI for the Planck Dust Template Model for HAWC J2027+3718. First set of errors is statistical; the second is systematic. Φ0\Phi_0 is the differential flux at E0=2E_0 = 2 TeV in units of ×10−14\times10^{-14} [(TeV cm2^2 s)−1^{-1}]. Φ0\Phi_0 for the GDE is the dimensionless scale factor.

NOTE - Φ0\Phi_0 is the differential flux at E0=7E_0=7 TeV and is in units of ×10−15\times10^{-15} [(TeV cm2^2 s)−1^{-1}]

NOTE - First set of errors is statistical, while the second is systematic. Systematic errors are reported only for the final best-fit model

Fermi-LAT Analysis

We re-analyzed archival Fermi-LAT data near J2027 to connect the >> TeV results discussed in § 2.1.5 with lower-energy processes. We use Fermi-LAT photon data between 4 August 2008 and 20 July 2024 with energies from 200 MeV up to 1 TeV and fermitools version 2.2.0 and fermipy version 1.1.6 [54]. The ROI for this analysis was circular with radius 20∘20^\circ centered at J2027’s LHAASO position (RA =306.88∘=306.88^\circ, Dec =36.95∘=36.95^\circ). The Fermi-LAT instrument response function version P8R3-SOURCE-V3 was used in the analysis while setting the zenith angle maximum at 90∘90^\circ.

We included all 4FGL-DR4 sources in the ROI in the gamma ray background model, plus the diffuse background model with the Galactic diffuse emission template (gll iem v07.fits) and the relevant isotropic emission template (iso P8R3 SOURCE V3 v1.txt) obtained at the Fermi science center [32].

With gtlike, we performed the binned maximum likelihood fitting on gamma-ray data. During the fit, we left free the normalization parameters of all the sources within 3∘3^\circ ROI with significance TS >20>20 as well as the diffuse emission components. The parameters of all the other sources were fixed to values given in the 4FGL-DR4 catalog [17].

Figure 5 shows the zoomed version of the TS map produced for the 10∘×10∘10^\circ\times10^\circ analysis region in the energy range of 1 GeV −- 1 TeV. We computed the best-fit position of 4FGL J2026.5+3718c assuming it to be a point source with log-parabola type spectrum. The best-fit position is found to be RA =306.82∘=306.82^\circ, Dec =37.48∘=37.48^\circ with a positional uncertainty of 0.04∘0.04^\circ. This new centroid is offset from the 4FGL position by 0.22∘0.22^\circ

Fermi-LAT TS maps in the ROI

Figure 5. Left Panel: The Fermi-LAT TS map in the ROI in the energy range of 1 GeV - 1 TeV, where the TS value range is fixed for both maps between 3 and ∼\sim32. The position and spatial extent of S106 are indicated by the green dashed circle, while the yellow dashed circle marks the position and positional uncertainty of the LHAASO source. The HAWC significance contours (2.75σ\sigma, 3.5σ\sigma, 4.5σ\sigma) shown in white solid color are taken from the bottom left panel of Figure 4. Right Panel: The TS map after adding the best-fit spatial model (RadialGaussian PL) to the background gamma-ray emission model, after the subtraction of 4FGL J2026.5+3718c (white cross for centroid position, white ellipse as error ellipse). 4FGL J2026.5+3718c with the best-fit extension is represented by the cyan solid circle.

To test whether the Fermi-LAT emission near J2027 is a point source, we employed the extension module in fermipy, which compares the point-source hypothesis with an extended spatial model using a likelihood-ratio test. For this source, we compared the RadialGaussian and RadialDisk spatial models and find a preference for an extended model in all combinations of spatial and spectral models, with TS >25>25 in all cases. Table 3 shows a summary of the extension measurements. The best-fit spatial model is RadialGaussian with a power law (PL) or broken power law (BPL) spectrum, where the Gaussian radius of the extended source is r=0.39∘±0.07∘r=0.39^\circ\pm0.07^\circ centered around the position of 4FGL J2026.5+3718c.

Spatial ModelSpectral ModelTSextTS_{\rm ext}
RadialDiskPower Law38.519
Broken Power Law38.519
Log-Parabola35.684
RadialGaussianPower Law39.456
Broken Power Law39.456
Log-Parabola35.516

Table 3. Fit results of Fermi-LAT data for 4FGL J2026.5+3718c produced in the energy range of 1 GeV − 1 TeV.

The resulting TS map after applying the RadialGaussian+PL model is shown in Figure 5. The right panel shows the result of fitting the RadialGaussian+PL model to the data after the subtraction of 4FGL J2026.5+3718c from the gamma-ray background model.

Using the best-fit spatial extension model described above (RadialGaussian+PL), we derived the energy spectrum of 4FGL J2026.5+3718c over the energy range from 200 MeV to 1 TeV. We found the total significance of the extended source to be ∼11σ\sim11\sigma with the spectral index fitted to Γ=2.18±0.06\Gamma=2.18\pm0.06. The total flux and energy flux of the extended gamma-ray emission was found to be (1.06±0.13)×10−8(1.06\pm0.13)\times10^{-8} cm−2^{-2} s−1^{-1} and (1.09±0.12)×10−5(1.09\pm0.12)\times10^{-5} MeV cm−2^{-2} s−1^{-1}, respectively. Using the standard Fermi-LAT energy bins, we produce the flux points between 200 MeV and 1 TeV shown in Figure 7 and Figure 8.

Figure 7. The E2dN/dEE^{2}dN/dE spectrum of high-energy emission near S106 (black circles for detections, grey arrows for upper limits, magenta circle for KM2A flux and slope). The fully leptonic model is plotted in blue: synchrotron emission from a ∼2.4 μG\sim2.4\ \mu\mathrm{G} magnetic field (blue and purple dotted), bremsstrahlung (green dashed) in a nH=250/cm3n_{\mathrm{H}}=250/\mathrm{cm}^{3} cloud, and IC emission (red dot-dash) dominated by IR photons from S106 itself all emerge from a single population of leptons. The sharp cutoff in bremsstrahlung at ≈1 GeV\approx1\ \mathrm{GeV} is an artifact from minimum particle energy in naima.

Figure 8. The E2dN/dEE^2dN/dE spectrum of high-energy emission near S106, but with a lepto-hadronic model. Instead of IC emission, pion decay (solid orange) produces the > TeV gamma-ray flux. In this case, the IC emission (red dash-dot) is subdominant at TeV energies.

Archival Chandra Data

There exists a single archival Chandra observation at S106, a ∼45\sim45 ks ACIS-I observation with observation ID 1893. This observation was the subject of a detailed analysis in Giardino et al. [31], studying the X-ray emission from numerous point sources including the central object S106IR. We revisit this archival observation with an eye towards connecting the physical diagnostics enabled with X-ray observations to the VHE emission observed with Fermi-LAT and HAWC. While Giardino et al. [31] focused on the diverse X-ray point sources in S106, we aimed to evaluate the diffuse X-ray emission near S106 and draw comparisons with the observed high-energy emission.

We downloaded observation 1893 from the Chandra Data Archive, and used the standard chandra repro task in ciao version 4.17 to reprocess the raw data files into a cleaned event file. Using fluximage, we produced an image of S106 in the “broad” ACIS spectral band (0.5 − 7.0 keV, shown in Figure 6 on the left).

The Chandra ACIS and UKIDSS J-band images of S106

Figure 6. The Chandra ACIS (left) and UKIDSS J-band (right) image of S106 obtained from SkyView, with the source ROI (2′ × 2′) in green highlighted in both views. We extract a spectrum by masking out the point sources studied in Giardino et al. [31]. The red dashed regions in the Chandra image are the borders of the ACIS chips.

As we are interested in diffuse, non-thermal emission throughout the S106 complex, we select a 2′ × 2′ source region enclosing the innermost X-ray point sources in the densest part of S106 (shown in green in Figure 6) in the north-east ACIS chip and then mask out the known point sources identified in [31] in a five pixel radius. Using a background region of the same

size in a source-free region of the same chip, we use specextract to extract a spectrum, response files, and background file for the diffuse emission in S106. We use sherpa version 4.17 to conduct X-ray analysis of this background-subtracted diffuse spectrum.

We adopt an absorbed power law model for the diffuse X-ray flux observed by Chandra; the point sources of X-rays are likely young stars ([31]) and are not related to the relativistic particles studied here. Leaving hydrogen column density NHN_H a free parameter to account for material close to S106 and in the intervening interstellar space, we obtain a best fit of NH=(1.4±0.5)×1022 /cm2N_{\rm H} = (1.4 \pm0.5) \times10^{22}\ /\mathrm{cm}^{2} and ΓX=2.0±0.4\Gamma_{\rm X} = 2.0 \pm0.4. The calculated 0.5 − 8.0 keV flux of the diffuse emission is (6±2)×10−13 erg/s/cm2(6 \pm2) \times10^{-13}\ \mathrm{erg}/\mathrm{s}/\mathrm{cm}^{2}, or at a distance of 1.4 kpc

Model NumberSpatial Model Named.o.f.Δ\DeltaAIC
11ES: RadialGaussian PL30
21ES: RadialGaussian BPL426

Table 4. The best-fit spatial model for a single extended source (1ES) in the 1 GeV − 1 TeV energy range is found by calculating the Δ\DeltaAIC value. Column (1) shows the spatial model name. (2)nd and (3)rd columns give the degrees of freedom (d.o.f.) of each model and the corresponding Δ\DeltaAIC value, respectively.

a diffuse X-ray luminosity of (2.96±0.5)×1031 erg/s(2.96 \pm0.5) \times10^{31}\ \mathrm{erg}/\mathrm{s}. This is about 10% of the total luminosity from the point sources identified in Giardino et al. [31].

We extract individual flux points from the Chandra spectrum by unfolding the spectrum around the measured Chandra sensitivity as given in the ARF and RMF files and binning the spectrum to S/N>5σS/N > 5\sigma in each bin, shown in Figure 7 and 8. We also calculate the synchrotron emission expected with no intervening absorption for comparison.

Multiwavelength Modeling

The Leptonic Model

Combining the diffuse X-ray flux from the archival Chandra X-ray data with the Fermi-LAT and HAWC gamma-ray emission, plus the reported LHAASO-KM2A flux from Cao et al. [24] and the upper limit implied by non-detection in the most recent Swift-BAT point source catalog (Lien et al. [39]), we produce the SED in the black points in Figure 7.

To test whether the Chandra, Fermi-LAT, and HAWC points could all be simultaneously fit by a single population of high-energy leptons, we develop a model with three discrete emission mechanisms, which we implement using the naima package (Zabalza [55]). The lepton population is an exponential cutoff power law of the form

dNdE=N0(ETeV)−αe−EEcut\frac{dN}{dE} = N_0 \left(\frac{E}{\mathrm{TeV}}\right)^{-\alpha} e^{-\frac{E}{E_{\rm cut}}}

These leptons produce emission via the following processes:

  • keV synchrotron radiation produced by leptons cycling around the ambient magnetic field lines in the core of S106, constrained by various previous observations ([36, 46, 51]) to be in the range of 10−500 μG10 - 500\,\mu\mathrm{G}. Below ≈1 keV\approx1\,\mathrm{keV} these X-rays are absorbed by intervening material.

  • Bremsstrahlung radiation as those same electrons interact with the ambient gas around S106 with a density of nH≈102−103/cm3n_\mathrm{H} \approx10^2 - 10^3/\mathrm{cm}^3.

  • IC up-scattering of ambient photons by the same leptons as they stream outwards into the galactic cosmic ray environment.

Notably, in addition to the standard photon fields relevant for inverse Compton (IC) emission, the cosmic microwave background (CMB), near- and far-infrared (NIR and FIR) radiation, and local synchrotron photons from synchrotron self-Compton (SSC), the intense infrared radiation field associated with S106 must also be included as a target photon field for up-scattering. We calculate the radiation field density associated with this S106 IR field, by adopting the dust luminosity of Adams et al. [6], Ldust≳(9.02±1.01)×104 L⊙L_\mathrm{dust} \gtrsim(9.02 \pm1.01) \times10^4\,L_\odot, derived from 12−350 μm12-350\,\mu\mathrm{m} photometry of S106 at an assumed distance of 1.4 kpc1.4\,\mathrm{kpc} and quoted as a lower limit because some stellar radiation may escape through the bipolar cavity. The radiation energy density at a distance rr from S106 is urad=Ldust/(4πr2c)u_\mathrm{rad} = L_\mathrm{dust}/(4\pi r^2 c), which gives 3.06×10−5 eV cm−33.06 \times10^{-5}\,\mathrm{eV\,cm}^{-3} at Earth and ≈60 eV cm−3\approx60\,\mathrm{eV\,cm}^{-3} at r=1 pcr = 1\,\mathrm{pc}. We evaluate this expression at the characteristic distance dICd_\mathrm{IC} from S106 at which the IC up-scattering by the lepton population occurs, treating dICd_\mathrm{IC} as a free parameter in the fit.

All the photon fields used in the IC model are described in Table 5, including their blackbody temperature TbbT_\mathrm{bb} and energy density uradu_\mathrm{rad}. The S106 field exceeds the combined CMB, FIR and NIR energy density (1.76 eV cm−31.76\,\mathrm{eV\,cm}^{-3}) inside ≈5.8 pc\approx5.8\,\mathrm{pc}, and falls below it at larger separations. Additionally, if the observed HAWC and LHAASO emission is produced by IC up-scattering on these spatial scales, the extension seen in TeV gamma-rays can be naturally reproduced.

FieldTbbT_\mathrm{bb} (K)uradu_\mathrm{rad} (eV cm−3\mathrm{eV\,cm}^{-3})
CMB2.70.261
FIR300.5
S10610060 (dIC/pc)−260\,(d_\mathrm{IC}/\mathrm{pc})^{-2}
NIR30001.0
SSC⋯\cdots(X-ray synchrotron field)

Table 5. Photon fields invoked for IC up-scattering. dICd_\mathrm{IC} is the characteristic distance from S106 at which the upscattering of IR photons takes place. SSC has a negligible energy density relative to the other photon fields.

Note— The S106 field is normalized from the dust luminosity Ldust≳(9.02±1.01)×104 L⊙L_\mathrm{dust} \gtrsim(9.02 \pm1.01) \times10^4\,L_\odot measured by Adams et al. [6] at d=1.4 kpcd = 1.4\,\mathrm{kpc}, via urad=Ldust/(4πr2c)u_\mathrm{rad} = L_\mathrm{dust}/(4\pi r^2 c).

It is worth noting that each of the emission mechanisms noted above have a single astrophysical parameter that can scale up or down the flux from that particular mechanism (BB for synchrotron, nHn_\mathrm{H} for bremsstrahlung, and the additional IR field from S106 for IC). Astrophysical modeling can constrain each of these parameters, but the extended emission could also occur in diverse environments around S106 with varying values for each. In our modeling, we fix the ambient gas density at nH=250/cm3n_\mathrm{H} = 250/\mathrm{cm}^3 and allow the magnetic field BB, the characteristic IC scattering distance dICd_\mathrm{IC}, and the index, cutoff energy and normalization of the lepton distribution to vary. The model parameters are fitted to the multi-wavelength data using the Markov Chain Monte Carlo (MCMC) implementation within naima’s framework, which samples the parent particle population parameters. We note that the parameters of theses models are correlated and are/or remain poorly constrained by the available data. Consequently, the best-fit parameter values should not be interpreted as independent measurements of the physical quantities, but rather as a plausible set of values that provides a reasonable description of the observed data. The best fit parameter values are summarized in Table 6 and the resulting spectrum is shown in Figure 7. The total energy in the electron population above 1 GeV1\,\mathrm{GeV}, WeW_\mathrm{e}, is derived from the best-fit values of N0N_0, EcutE_\mathrm{cut}, and α\alpha. The best-fit dIC=3.55 pcd_\mathrm{IC} = 3.55\,\mathrm{pc} therefore places the up-scattering region where photons from S106 itself supply the dominant seed population, exceeding the ambient fields by a factor of ≈2.7\approx2.7. Notably, this crossover radius is comparable to the observed extent of the TeV emission (0.25∘0.25^\circ, or ≈6.1 pc\approx6.1\,\mathrm{pc} at 1.4 kpc1.4\,\mathrm{kpc}), so the size of the IC-emitting region is set by the scale over which S106’s own radiation field dominates.

Parameterunitsvalue
N0N_{0}1/eV6.16×10316.16 \times10^{31}
EcutE_{\mathrm{cut}}TeV1753
α\alpha2.41
WeW_{\mathrm{e}}erg4.11×10454.11 \times10^{45}
BBμG\mu\mathrm{G}2.38
nHn_{\mathrm{H}}1/cm31/\mathrm{cm}^{3}250
dICd_{\mathrm{IC}}pc3.55

Table 6. Parameters for the leptonic model in Figure 7.

As shown in Figure 7, a single-population leptonic model can fit the entire SED for S106, from the Chandra diffuse emission to the HAWC and KM2A data points. If a population of high-energy hadrons is present, it would be sub-dominant at all relevant energies.

The Lepto-hadronic Model

In an alternative hybrid lepto-hadronic model, X-rays are still produced via diffuse synchrotron emission in S106, and GeV gamma-rays are still the product of lepton bremsstrahlung, but the gamma rays are produced by the decay of hadrons into gamma-rays via neutral pions. Figure 8 shows one possible lepto-hadronic model. The addition of a hadronic component introduces many new free parameters into the model, such as the energy density, slope, and cutoff of the hadron distribution while more weakly constraining the lepton distribution.

Figure 8 shows the broadband SED in this lepto-hadronic model, with gamma rays from pion decay explaining the TeV emission instead of IC scattering. This is achieved by increasing the effective distance at which the IC emission occurs with respect to S106, thereby decreasing the energy density of the IR field from S106 and greatly reducing the IC luminosity of the system. The >> TeV gamma-rays are then entirely the product of hadronic pion decay, with IC emission being subdominant at all relevant energies.

Table 7 gives the parameters of the lepto-hadronic model, once again calculating the energies in the leptons and hadrons WeW_{\mathrm{e}} and WpW_{\mathrm{p}} above 1 GeV. The hadrons in this model contain substantially more energy than the leptons, by dint of their higher cutoff energy and shallower spectrum.

Parameterunitsvalue
Leptons
N0N_01/eV1.878×10311.878 \times10^{31}
EcutE_{\mathrm{cut}}TeV99.5
α\alpha2.7
WeW_eerg5.3×10455.3 \times10^{45}
Hadrons
N0N_01/eV1.8×10331.8 \times10^{33}
EcutE_{\mathrm{cut}}TeV1000
α\alpha2.16
WpW_perg3.1×10463.1 \times10^{46}
BBμ\muG18.2
nHn_H1/cm3^3250
dICd_{\mathrm{IC}}pc15.1

Table 7. Parameters for the lepto-hadronic model in Figure 8.

Discussion and Conclusions

Discussion

Our fully leptonic model for the combined Chandra, Fermi-LAT, HAWC, and LHAASO data points in Figure 7 has a boosted ambient IR field close to S106 contributing most seed photons to >> TeV IC emission. Considering this model, it would seem that the gamma-ray emission from hadronic components is not strictly needed. This is not to say that emission above ≈\approx TeV cannot have hadronic components; In the lepto-hadronic model fit shown in Figure 8, lowering the leptonic contributions to the TeV flux while increasing the hadronic contribution results in a hybrid model.

The critical factors differentiating between the leptonic and hadronic models are the magnitude of the ambient photon fields for IC up-scattering and the total energy in the hadron population. If the ambient photon fields for IC up-scattering are high, perhaps because the scattering occurs very close the S106’s bright IR emission, then the gamma rays will be almost entirely leptonic and take place over an extended region. If the TeV gamma rays are hadronic in origin they would be correlated with the density of target matter throughout the region near S106.

Cooling Timescales

Is it feasible that a single population of leptons dominates the extended X-ray, GeV, and TeV emission observed near S106? The cooling timescales of each of these processes must be larger than the physical size of S106, such that a single population of leptons can cause all three emission mechanisms without depleting its reservoir of energy. S106 has a size on the sky of ≈2′\approx2', which at a distance of ≈1.4\approx1.4 kpc is a physical size of ≈3\approx3 ly. The cooling timescales of the emission mechanisms in our leptonic model must all be substantially greater than the light crossing time of S106 to ensure that a lepton population can participate in all three mechanisms.

For synchrotron emission, we compare the energy of a typical electron E=γ0mec2E = \gamma_0 m_e c^2 with the synchrotron power radiated by that electron Psyn=2γ2σTcUBsin⁡2(α)P_{\mathrm{syn}} = 2\gamma^{2}\sigma_{\mathrm{T}}cU_{B}\sin^{2}(\alpha). Adopting the pitch angle average ⟨sin⁡2(α)⟩=2/3\langle\sin^{2}(\alpha)\rangle= 2/3, together with B=2.38 μGB = 2.38\,\mu\mathrm{G} and conservatively, a Lorentz factor for the electrons of γ=109\gamma= 10^{9} (corresponding to a lepton energy E≈0.5 PeVE \approx0.5\,\mathrm{PeV}), we obtain the characteristic synchrotron cooling timescale of,

tsyn=EPsyn=mec243σTcγUB≈4329 yrt_{\mathrm{syn}} = \frac{E}{P_{\mathrm{syn}}} = \frac{m_{\mathrm{e}}c^{2}}{\frac{4}{3}\sigma_{\mathrm{T}}c\gamma U_{B}} \approx4329\ \mathrm{yr}

where mem_{\mathrm{e}} is the mass of the electron, c is the speed of light, σT\sigma_{\mathrm{T}} is the Thomson cross section, and UBU_{B} is the magnetic field energy density given by UB=B2/8πU_{B} = B^{2}/8\pi.

For bremsstrahlung, Abazajian et al. [1] and Collins et al. [26] give the cooling time as, with our assumed nH=250/cm3n_{\mathrm{H}} = 250/\mathrm{cm}^{3},

tbrem=4×107 yr(nHcm3)−1=160000 yrt_{\mathrm{brem}} = 4 \times10^{7}\ \mathrm{yr}\left(\frac{n_{\mathrm{H}}}{\mathrm{cm}^{3}}\right)^{-1} = 160000\ \mathrm{yr}

The cooling timescale for IC scattering in the Klein–Nishina (KN) regime, tICt_{\mathrm{IC}}, is given by (Moderski et al. [40]):

tIC(γ)=3mec4cσTγ∑i[Urad,i(1+4γε0,i)3/2]t_{\mathrm{IC}}(\gamma) = \frac{3m_{\mathrm{e}}c}{4c\sigma_{\mathrm{T}}\gamma\sum_{i}\left[\frac{U_{\mathrm{rad},i}}{(1+4\gamma\varepsilon_{0,i})^{3/2}}\right]}

Here, Urad,iU_{\mathrm{rad},i} and ε0,i\varepsilon_{0,i} are the energy density of the iith radiation field and the mean energy of its seed photons, respectively, where ε0,i\varepsilon_{0,i} is expressed in units of mec2m_{\mathrm{e}}c^{2}. Using the energy densities and mean photon energies listed in Table 5, with the latter calculated from the Planck blackbody relation ⟨E⟩=2.7kBT\langle E\rangle= 2.7k_{\mathrm{B}}T, we obtain an IC cooling timescale of tIC=28.6 kyrt_{\mathrm{IC}} = 28.6\ \mathrm{kyr}. We note that at γ=109\gamma= 10^{9}, the IR photons from S106 and NIR fields are strongly Klein–Nishina suppressed (4γε0≫14\gamma\varepsilon_{0} \gg1), so the IC cooling of the highest-energy electrons is dominated by the CMB even though the S106 field dominates the total radiation energy density at dICd_{\mathrm{IC}}.

All three of these timescales are large enough such that a single population of leptons streaming out at relativistic speeds will survive to produce the three emission mechanisms in the leptonic model before escaping into the wider galaxy as cosmic rays.

Total Energy Budget

Young stellar clusters are often invoked as sources of cosmic rays, the sites of dramatic particle acceleration. We have shown that it is feasible for X-ray, GeV, and TeV emission from systems like S106 to be attributed to fully leptonic processes. For the leptonic model we can also compare the total lepton energy budget We=4.1×1045 ergW_{\mathrm{e}} = 4.1 \times10^{45}\ \mathrm{erg} to the kinetic energy of the outflows from the central source of S106, thought to be a young O+B binary named S106IR (Comerón et al. [27]).

The total kinetic energy luminosity of the outflows from S106IR is given by

Lk=12M˙vflow2L_{\mathrm{k}} = \frac{1}{2}\dot{M}v_{\mathrm{flow}}^{2}

where M˙\dot{M} is the mass loss rate from S106IR and vflowv_{\mathrm{flow}} is the measured velocity of the winds. Simon & Fischer [49] and Bally et al. [19] find M˙≈10−6M⊙/yr\dot{M} \approx10^{-6}M_{\odot}/\mathrm{yr} and vflow≈150 km/sv_{\mathrm{flow}} \approx150\ \mathrm{km/s}, though admittedly both knots arrived at these values before the likely binary nature of S106IR was established. Using those values, we calculate Lk≈2×1041 erg/yearL_{\mathrm{k}} \approx2 \times10^{41}\ \mathrm{erg/year} in the outflows from S106IR.

For the duration over which this luminosity is sustained, several estimates can be applied. An upper limit is the age of the S106 complex as a whole, estimated in Oasa et al. [41] as roughly ∼Myr\sim\mathrm{Myr}. The lower limit might be the possible supersonic explosion 3500 years ago cited in Bally et al. [18]. Adopting the rather conservative IC cooling time noted above tIC=28.6 kyrt_{\mathrm{IC}} = 28.6\ \mathrm{kyr} we obtain a total reservoir of kinetic energy over that period of Ek=tIC×Lk≈5.72×1045 ergE_{\mathrm{k}} = t_{\mathrm{IC}} \times L_{\mathrm{k}} \approx5.72 \times10^{45}\ \mathrm{erg}. The order-of-magnitude similarity of this estimate for a single high-mass system to the energy stored in the lepton distribution supports the at-large shocked outflows from multiple young stars in S106 as the engine behind the high-energy fluxes analyzed here.

Particle Diffusion Coefficients

To evaluate whether particle transport through the S106 region is feasible, we parameterize the standard energy-dependent diffusion coefficient as

D(E)=χD0(EE0)δ,D(E) = \chi D_{0}\left(\frac{E}{E_{0}}\right)^{\delta},

where χ\chi accounts for suppression relative to the average Galactic diffusion coefficient, and δ\delta describes the turbulence regime ([44, 22]). We estimate the effective transport coefficient required for particles to reach an offset emission region by equating the observed propagation scale RR with the diffusion radius,

Rdiff(E,t)=4D(E)t.R_{\mathrm{diff}}(E,t) = \sqrt{4D(E)t}.

This gives

Dreq(R,t)∼R24t,D_{\mathrm{req}}(R,t) \sim\frac{R^{2}}{4t},

following the standard Green’s-function solution of the diffusion equation for particles escaping from a compact accelerator ([15, 42]). Therefore,

Dreq≃7.6×1027(R100 pc)2(t105 yr)−1 cm2 s−1.D_{\mathrm{req}} \simeq7.6 \times10^{27}\left(\frac{R}{100\ \mathrm{pc}}\right)^{2}\left(\frac{t}{10^{5}\ \mathrm{yr}}\right)^{-1}\ \mathrm{cm}^{2}\ \mathrm{s}^{-1}.

Here DreqD_{\mathrm{req}} is not an independently fitted quantity; rather, it is the effective diffusion coefficient required for particles to reach the observed distance RR within an activity time tt. We compare this value with the Bohm diffusion coefficient,

DB(E)=13rLc≃3.3×1025(E1 TeV)(1 μGB) cm2 s−1.D_{\mathrm{B}}(E)=\frac{1}{3}r_{\mathrm{L}}c\simeq3.3\times10^{25}\left(\frac{E}{1\ \mathrm{TeV}}\right)\left(\frac{1\ \mu\mathrm{G}}{B}\right)\ \mathrm{cm}^{2}\ \mathrm{s}^{-1}.

where rL≃E/(qB)r_{\mathrm{L}}\simeq E/(qB) is the Larmor radius for a relativistic particle. Bohm diffusion corresponds to the limiting case in which the particle mean free path is comparable to rLr_{\mathrm{L}}, and therefore DB(E)∝ED_{\mathrm{B}}(E)\propto E, i.e. δ=1\delta=1.

For the purely leptonic fit, the best-fit IC distance is

dIC=3.55 pc,d_{\mathrm{IC}}=3.55\ \mathrm{pc},

with

B=2.38 μG,Ecut,e=1753 TeV,B=2.38\ \mu\mathrm{G},\qquad E_{\mathrm{cut},e}=1753\ \mathrm{TeV},

and

We=4.11×1045 erg.W_e=4.11\times10^{45}\ \mathrm{erg}.

For this compact best-fit distance, the required diffusion coefficient is

Dreqlep≃9.6×1024(dIC3.55 pc)2(t105 yr)−1 cm2 s−1.D_{\mathrm{req}}^{\mathrm{lep}}\simeq9.6\times10^{24}\left(\frac{d_{\mathrm{IC}}}{3.55\ \mathrm{pc}}\right)^2\left(\frac{t}{10^5\ \mathrm{yr}}\right)^{-1}\ \mathrm{cm}^{2}\ \mathrm{s}^{-1}.

At the fitted electron cutoff energy, the Bohm coefficient is

DB,elep(1753 TeV)≃2.4×1028(2.38 μGB) cm2 s−1.D_{\mathrm{B},e}^{\mathrm{lep}}(1753\ \mathrm{TeV})\simeq2.4\times10^{28}\left(\frac{2.38\ \mu\mathrm{G}}{B}\right)\ \mathrm{cm}^{2}\ \mathrm{s}^{-1}.

The synchrotron cooling time is approximately

tsyn≃1.3×107(B1 μG)−2(Ee1 TeV)−1 yr.t_{\mathrm{syn}}\simeq1.3\times10^{7}\left(\frac{B}{1\ \mu\mathrm{G}}\right)^{-2}\left(\frac{E_e}{1\ \mathrm{TeV}}\right)^{-1}\ \mathrm{yr}.

For B=2.38 μGB=2.38\ \mu\mathrm{G} and Ee=1753 TeVE_e=1753\ \mathrm{TeV}, this gives tsyn∼1.3×103 yrt_{\mathrm{syn}}\sim1.3\times10^{3}\ \mathrm{yr}. The corresponding Bohm diffusion length is of order ∼20 pc\sim20\ \mathrm{pc}, larger than the best-fit dICd_{\mathrm{IC}}. Therefore, the updated purely leptonic solution is transport-compatible in the sense that the IC-emitting region is inferred to lie only a few parsecs from S106. This favors an interpretation in which the Fermi-LAT Bremsstrahlung component traces lower-energy electrons confined in the dense S106 environment, while the IC component is produced relatively close to the acceleration site.

We next consider the lepto-hadronic interpretation. In this model, the electron and proton spectra are described by power laws with exponential cutoffs, with

αe=2.70,Ecut,e=99.5 TeV,\alpha_e=2.70,\qquad E_{\mathrm{cut},e}=99.5\ \mathrm{TeV},
αp=2.16,Ecut,p=1000 TeV.\alpha_p=2.16,\qquad E_{\mathrm{cut},p}=1000\ \mathrm{TeV}.

The corresponding total particle energies are

We=5.300×1045 erg,Wp=3.081×1046 erg,W_e=5.300\times10^{45}\ \mathrm{erg},\qquad W_p=3.081\times10^{46}\ \mathrm{erg},

and the fitted magnetic field is

B=18.22 μG.B=18.22\ \mu\mathrm{G}.

As a compact lepto-hadronic transport estimate, we first evaluate the required diffusion coefficient for

dIC=3.55 pc,d_{\mathrm{IC}}=3.55\ \mathrm{pc},

matching the compact scale favored by the purely leptonic fit. This gives

DreqLH,3.55≃9.6×1024(dIC3.55 pc)2(t105 yr)−1 cm2 s−1.D_{\mathrm{req}}^{\mathrm{LH},3.55}\simeq9.6\times10^{24}\left(\frac{d_{\mathrm{IC}}}{3.55\ \mathrm{pc}}\right)^2\left(\frac{t}{10^5\ \mathrm{yr}}\right)^{-1}\ \mathrm{cm}^{2}\ \mathrm{s}^{-1}.

For the lepto-hadronic magnetic field, the Bohm coefficients at the fitted cutoff energies are

DB,eLH(99.5 TeV)≃1.8×1026(18.22 μGB) cm2 s−1,D_{\mathrm{B},e}^{\mathrm{LH}}(99.5\ \mathrm{TeV})\simeq1.8\times10^{26}\left(\frac{18.22\ \mu\mathrm{G}}{B}\right)\ \mathrm{cm}^{2}\ \mathrm{s}^{-1},

and

DB,pLH(1000 TeV)≃1.8×1027(18.22 μGB) cm2 s−1.D_{\mathrm{B},p}^{\mathrm{LH}}(1000\ \mathrm{TeV})\simeq1.8\times10^{27}\left(\frac{18.22\ \mu\mathrm{G}}{B}\right)\ \mathrm{cm}^{2}\ \mathrm{s}^{-1}.

Although DB,eLHD_{\mathrm{B},e}^{\mathrm{LH}} is larger than DreqLH,3.55D_{\mathrm{req}}^{\mathrm{LH},3.55}, this comparison alone does not include radiative cooling. For B≃18 μGB\simeq18\ \mu\mathrm{G} and Ee≃100 TeVE_e\simeq100\ \mathrm{TeV}, the synchrotron cooling time is only

tsyn∼4×102 yr.t_{\mathrm{syn}}\sim4\times10^{2}\ \mathrm{yr}.

The associated Bohm diffusion length is of order ∼1 pc\sim1\ \mathrm{pc}, smaller than dIC=3.55 pcd_{\mathrm{IC}}=3.55\ \mathrm{pc}. Thus, even in the compact lepto-hadronic case, the highest-energy electrons cannot easily propagate several parsecs in the Bohm limit without substantial cooling. Lower-energy electrons responsible for the GeV Bremsstrahlung component are less affected by these losses and can remain confined near the dense S106 gas.

We then evaluate the moderate-offset lepto-hadronic case,

dIC=15.1 pc≃15 pc.d_{\mathrm{IC}}=15.1\ \mathrm{pc}\simeq15\ \mathrm{pc}.

The required diffusion coefficient becomes

DreqLH,15≃1.7×1026(dIC15.1 pc)2(t105 yr)−1 cm2 s−1.D_{\mathrm{req}}^{\mathrm{LH},15}\simeq1.7\times10^{26}\left(\frac{d_{\mathrm{IC}}}{15.1\ \mathrm{pc}}\right)^2\left(\frac{t}{10^5\ \mathrm{yr}}\right)^{-1}\ \mathrm{cm}^{2}\ \mathrm{s}^{-1}.

This value is comparable to the Bohm coefficient of the ∼100 TeV\sim100\ \mathrm{TeV} electrons. However, because the synchrotron cooling time of these electrons is only ∼4×102\sim4 \times10^{2} yr, their cooling-limited diffusion length remains much smaller than 15.1 pc. Therefore, the offset TeV emission in the lepto-hadronic model cannot be explained by pure Bohm diffusion of the highest-energy electrons from S106 alone.

The proton component is less constrained by radiative cooling. In both lepto-hadronic transport estimates, the Bohm coefficient of the ∼1\sim1 PeV protons is larger than the required diffusion coefficient for dIC=3.55d_{\mathrm{IC}} = 3.55 pc and is also sufficient for transport over ∼15\sim15 pc on timescales shorter than 10510^{5} yr. Therefore, the updated lepto-hadronic solution favors a two-component interpretation: lower-energy electrons confined near the dense S106 environment produce the Fermi-LAT Bremsstrahlung emission, while the offset TeV emission requires either faster-than-Bohm diffusion in the outer region, a lower magnetic field outside the dense core, local re-acceleration, and/or a substantial hadronic contribution.

The hadronic interpretation is particularly important because the best-fit proton cutoff reaches

Ecut,p=1000 TeV,E_{\mathrm{cut,p}} = 1000\ \mathrm{TeV},

the upper boundary of the adopted proton energy range. Thus, the current lepto-hadronic fit supports a PeVatron-candidate interpretation. However, because the cutoff is boundary-limited, it should be treated as evidence for acceleration into the PeV regime rather than as a precisely measured spectral turnover.

Conclusions

In this work we have detected the LHAASO KM2A source 1LHAASO J2027+3657 with HAWC, characterizing its TeV gamma-ray emission in greater detail. We also re-analyzed Fermi-LAT data in this part of the sky, examining especially the cataloged source 4FGL J2026.5+3718c. Combining the Fermi and HAWC data allows for a more comprehensive understanding of the high-energy processes in this region.

We additionally re-analyzed archival Chandra data at the star-forming region S106. Instead of focusing on the point sources within S106, which are likely young stars Giardino et al. [31], we extracted a spectrum for the diffuse emission in the region by masking out the point sources. We find that diffuse emission is well-fit with a Γ=2.0±0.4\Gamma= 2.0 \pm0.4 power law accounting for ≈10%\approx10\% of the total X-ray luminosity of S106. This X-ray flux can be attributed to a non-thermal population of energetic particles produced in S106 by shocked stellar winds, though it may also contain unresolved point sources.

While it is often assumed that TeV emission from SFRs is hadronic in origin, our leptonic model demonstrates that the X-ray and gamma-ray emission from a star-forming region can be fully explained by leptons alone, and that gamma-rays produced by hadronic processes, do not necessarily dominate such systems. Key differences between fully leptonic and hadronic models include the morphology of the gamma-ray emission and the budget of energy in the outflows. Crucially, proper modeling of IC emission in systems like S106 must account for substantially boosted ambient photon fields that can be upscattered to TeV energies; neglecting this factor dramatically underestimates the flux of gamma-rays produced by IC scattering.

The diffusion estimates suggest a more nuanced transport picture. In the purely leptonic fit, the best-fit IC distance, dIC=3.55d_{\mathrm{IC}} = 3.55 pc, is sufficiently compact that electron transport from the S106 region remains plausible, even though the highest-energy electrons cool rapidly. In this case, the Fermi-LAT Bremsstrahlung component can be interpreted as tracing the lower-energy electron population confined in the dense S106 environment, while the IC component is produced relatively close to the acceleration site. In the lepto-hadronic interpretation, however, the higher magnetic field (B≃18 μGB \simeq18\ \mu\mathrm{G}) implies a short synchrotron cooling time for ∼100\sim100 TeV electrons. Therefore, pure Bohm diffusion of the highest-energy electrons from S106 is insufficient to explain the offset TeV emission on scales of dIC≃15d_{\mathrm{IC}} \simeq15 pc. The TeV component may instead require faster-than-Bohm diffusion in the outer region, a lower magnetic field outside the dense core, local re-acceleration, and/or a substantial hadronic contribution. The latter possibility is particularly relevant because the lepto-hadronic fit drives the proton cutoff to Ecut,p=1000E_{\mathrm{cut,p}} = 1000 TeV, supporting a PeVatron-candidate interpretation, while still leaving a degeneracy between the primary leptonic and hadronic contributions in the broadband gamma-ray SED.

The spatial coincidence of the GeV emission with the S106 SFR supports a dense-gas Bremsstrahlung origin, in which relativistic electrons accelerated by the S106IR’s jet/wind system interact with the surrounding molecular material. The Fermi-LAT emission traces the lower-energy electron population interacting with the dense gas in the S106 region. We assume these electrons to be primary particles accelerated by the S106IR’s jet/wind system. However, the GeV band is intrinsically degenerate: Bremsstrahlung from primary electrons, emission from secondary e±e^{\pm} produced in hadronic interactions, and the low-energy tail of pion decay emission can contribute with partially overlapping spectral shapes. The TeV emission detected by HAWC and LHAASO probes the highest-energy particles and is modeled as a combination of IC emission from high-energy electrons and pion decay emission from accelerated protons. The fitted hadronic cutoff suggests acceleration into the multi-100 TeV regime, supporting a PeVatron-candidate interpretation, although the present data do not uniquely separate the leptonic and hadronic components.

Revisiting S106 with further X-ray observations would be a fruitful path to monitor the X-ray resolved and unresolved point sources (which are expected to have some degree of variability) and better characterize the diffuse emission (which would not be expected to vary on short timescales). It is worth noting that the gamma-ray data analyzed from Fermi-LAT and HAWC above is comprehensive, stretching the entire durations of those observatories, while Chandra provides only a very brief snapshot from twenty years ago.

In the future, next-generation gamma-ray observatories like the Cherenkov Telescope Array Observatory (CTAO) will have substantially greater angular resolution compared to current telescopes. This will enable spatially-resolved studies of gamma-ray emission near SFRs like S106 [23] for direct discrimination between the IC scattering and pion decay scenarios for gamma-ray production. The detection of neutrinos from SFRs would also be a strong discriminator between hadronic and leptonic gamma-ray production, as neutrinos are only produced during hadronic particle interactions.

Acknowledgments

This research has made use of data and/or software provided by the High Energy Astrophysics Science Archive Research Center (HEASARC), which is a service of the Astrophysics Science Division at NASA/GSFC. This work was supported by NASA through the Fermi Guest Investigator (GI) Program under grant number 80NSSC26K0546.

The HAWC Collaboration acknowledges the support from: the US National Science Foundation (NSF); the US Department of Energy Office of High-Energy Physics; the Laboratory Directed Research and Development (LDRD) program of Los Alamos National Laboratory; Consejo Nacional de Ciencia y Tecnología (CONACyT), México, grants LNC-2023-117, 271051, 232656, 260378, 179588, 254964, 258865, 243290, 132197, A1-S-46288, A1-S-22784, CF-2023-I-645, CBF2023-2024-1630, cátedras 873, 1563, 341, 323, Red HAWC, México; DGAPA-UNAM grants IG101323, IG100726, IN111716-3, IN111419, IA102019, IN106521, IN114924, IN110521, IN102223; VIEP-BUAP; PIFI 2012, 2013, PROFOCIE 2014, 2015; the University of Wisconsin Alumni Research Foundation; the Institute of Geophysics, Planetary Physics, and Signatures at Los Alamos National Laboratory; Polish Science Centre grant, 2024/53/B/ST9/02671; Coordinación de la Investigación Científica de la Universidad Michoacana; Royal Society - Newton Advanced Fellowship 180385; Gobierno de España and European Union-NextGenerationEU, grant CNS2023-144099; The Program Management Unit for Human Resources & Institutional Development, Research and Innovation, NXPO (grant number B16F630069); Coordinación General Académica e Innovación (CGAI-UdeG), PRODEP-SEP UDG-CA-499; Institute of Cosmic Ray Research (ICRR), University of Tokyo. H.M. acknowledges support under grant number CBF2023-2024-1630. H.F. acknowledges support by NASA under award number 80GSFC21M0002. C.R. acknowledges support from National Research Foundation of Korea (RS-2023-00280210, RS-2026-25470489). We also acknowledge the significant contributions over many years of Stefan Westerhoff, Gaurang Yodh and Arnulfo Zepeda Domínguez, all deceased members of the HAWC collaboration. Thanks to Scott Delay, Luciano Díaz and Eduardo Murrieta for technical support.

Software: 3ML [52], hawc-hal [3], FTools [21], Xspec [14], FermiPy [54], Naima [55].

References

  1. [1]Abazajian, K. N., Canac, N., Horiuchi, S., & Kaplinghat, M. 2014, PhRvD, 90, 023526, doi: 10.1103/PhysRevD.90.023526DOI
  2. [2]Abeysekara, A. U., Albert, A., Alfaro, R., et al. 2019, ApJ, 881, 134, doi: 10.3847/1538-4357/ab2f7dDOI
  3. [3]Abeysekara, A. U., et al. 2021, PoS, ICRC2021, 828, doi: 10.22323/1.395.0828DOI
  4. [4]Abramowski, A., Acero, F., Aharonian, F., et al. 2012, A&A, 537, A114, doi: 10.1051/0004-6361/201117928DOI
  5. [5]Ackermann, M., Ajello, M., Allafort, A., et al. 2011, Science, 334, 1103, doi: 10.1126/science.1210311DOI
  6. [6]Adams, J., Herter, T., Hora, J., et al. 2015, The Astrophysical Journal, 814, 54
  7. [7]Aharonian, F., Yang, R., & de Oña Wilhelmi, E. 2019, Nature Astronomy, 3, 561, doi: 10.1038/s41550-019-0724-0DOI
  8. [8]Akaike, H. 1974, IEEE Transactions on Automatic Control, 19, 716, doi: 10.1109/TAC.1974.1100705DOI
  9. [9]Albert, A. ., et al. 2024, Astrophys. J., 972, 144, doi: 10.3847/1538-4357/adf2d
  10. [10]Albert, A., Alfaro, R., Alvarez, C., et al. 2020, ApJ, 905, 76, doi: 10.3847/1538-4357/abcd8
  11. [11]Albert, A., Alfaro, R., Alvarez, C., et al. 2021, The Astrophysical Journal, 911, 143, doi: 10.3847/1538-4357/abecdaDOI
  12. [12]Alfaro, R., Alvarez, C., Andrés, A., et al. 2025a, ApJ, 992, 156, doi: 10.3847/1538-4357/ae0186DOI
  13. [13]Alfaro, R., Alvarez, C., Araya, M., et al. 2025b, ApJ, 992, 22, doi: 10.3847/1538-4357/adf7aaDOI
  14. [14]Arnaud, K. A. 1996, in Astronomical Society of the Pacific Conference Series, Vol. 101, Astronomical Data Analysis Software and Systems V, ed. G. H. Jacoby & J. Barnes, 17
  15. [15]Atoyan, A. M., Aharonian, F. A., & Völk, H. J. 1995, Physical Review D, 52, 3265, doi: 10.1103/PhysRevD.52.3265DOI
  16. [16]Babu, R., Wentworth, P., Herzog, I., Salazar, D., & Nisa, M. U. 2025, arXiv e-prints, arXiv:2507.10307, doi: 10.48550/arXiv.2507.10307DOI
  17. [17]Ballet, J., Bruel, P., Burnett, T. H., Lott, B., & The Fermi-LAT collaboration. 2023, arXiv e-prints, arXiv:2307.12546, doi: 10.48550/arXiv.2307.12546
  18. [18]Bally, J., Chia, Z., Ginsburg, A., et al. 2022, The Astrophysical Journal, 924, 50, doi: 10.3847/1538-4357/ac30deDOI
  19. [19]Bally, J., Snell, R. L., & Predmore, R. 1983, ApJ, 272, 154, doi: 10.1086/161272DOI
  20. [20]Bednarek, W., & Pabich, J. 2011, A&A, 530, A49, doi: 10.1051/0004-6361/201116549DOI
  21. [21]Blackburn, J. K., Shaw, R. A., Payne, H. E., Hayes, J. J. E., & Heasarc. 1999, FTOOLS: A general package of software to manipulate FITS files, Astrophysics Source Code Library, record ascl:9912.002. http://ascl.net/9912.002
  22. [22]Blasi, P. 2013, Astronomy and Astrophysics Review, 21, 70, doi: 10.1007/s00159-013-0070-7DOI
  23. [23]Bonollo, A., Esposito, P., Giuliani, A., et al. 2026, Journal of High Energy Astrophysics, 50, 100465, doi: 10.1016/j.jheap.2025.100465DOI
  24. [24]Cao, Z., Aharonian, F., An, Q., et al. 2024, ApJS, 271, 25, doi: 10.3847/1538-4357/ae fd29
  25. [25]Collaboration†‡, I., Abbasi, R., Ackermann, M., et al. 2023, Science, 380, 1338, doi: 10.1126/science.adc9818
  26. [26]Collins, T., Rowell, G., Mitchell, A. M. W., et al. 2021, MNRAS, 504, 1840, doi: 10.1093/mnras/stab983DOI
  27. [27]Comerón, F., Schneider, N., Djupvik, A. A., & Schnugg, C. 2018, A&A, 615, A2, doi: 10.1051/0004-6361/201731979DOI
  28. [28]Dame, T. M., Hartmann, D., & Thaddeus, P. 2001, ApJ, 547, 792, doi: 10.1086/318388DOI
  29. [29]Dundovic, A., Evoli, C., Gaggero, D., & Grasso, D. 2021, A&A, 653, A18, doi: 10.1051/0004-6361/202140801DOI
  30. [30]Ergin, T., Saha, L., Bhattacharjee, P., et al. 2020, Monthly Notices of the Royal Astronomical Society, 501, 4226, doi: 10.1093/mnras/staa3817DOI
  31. [31]Giardino, G., Favata, F., & Micela, G. 2004, A&A, 424, 965, doi: 10.1051/0004-6361:20040363DOI
  32. [32]GSFC. 2024, Fermi-LAT Background Models, https://fermi.gsfc.nasa.gov/ssc/data/access/lat/BackgroundModels.html
  33. [33]HI4PI Collaboration, Ben Bekhti, N., Flöer, L., et al. 2016, A&A, 594, A116, doi: 10.1051/0004-6361/201629178
  34. [34]IceCube Collaboration, Aartsen, M. G., Ackermann, M., et al. 2018, Science, 361, 147, doi: 10.1126/science.aat2890DOI
  35. [35]IceCube Collaboration, Abbasi, R., Ackermann, M., et al. 2022, Science, 378, 538, doi: 10.1126/science.abg3395
  36. [36]Kazes, I., Troland, T. H., Crutcher, R. M., & Heiles, C. 1988, ApJ, 335, 263, doi: 10.1086/166924DOI
  37. [37]Kelner, S. R., Aharonian, F. A., & Bugayov, V. V. 2006, PhRvD, 74, 034018, doi: 10.1103/PhysRevD.74.034018DOI
  38. [38]LHAASO Collaboration. 2024, Science Bulletin, 69, 449, doi: https://doi.org/10.1016/j.scib.2023.12.040
  39. [39]Lien, A. Y., et al. 2025, Astrophys. J., 989, 161, doi: 10.3847/1538-4357/ade676DOI
  40. [40]Moderski, R., Sikora, M., Coppi, P. S., & Aharonian, F. 2005, Monthly Notices of the Royal Astronomical Society, 363, 954, doi: 10.1111/j.1365-2966.2005.09494.xDOI
  41. [41]Oasa, Y., Tamura, M., Nakajima, Y., et al. 2006, AJ, 131, 1608, doi: 10.1086/500086DOI
  42. [42]Ohira, Y., Murase, K., & Yamazaki, R. 2011, Monthly Notices of the Royal Astronomical Society, 410, 1577, doi: 10.1111/j.1365-2966.2010.17539.xDOI
  43. [43]Pshirkov, M. S. 2016, MNRAS, 457, L99, doi: 10.1093/mnrasl/slv205DOI
  44. [44]Ptuskin, V. S., Moskalenko, I. V., Jones, F. C., Strong, A. W., & Zirakashvili, V. N. 2006, The Astrophysical Journal, 642, 902, doi: 10.1086/501117
  45. [45]Reitberger, K., Reimer, A., Reimer, O., & Takahashi, H. 2015, A&A, 577, A100, doi: 10.1051/0004-6361/201525726DOI
  46. [46]Roberts, D. A., Crutcher, R. M., & Troland, T. H. 1995, ApJ, 442, 208, doi: 10.1086/175436DOI
  47. [47]Romero, G. E. 2010, Mem. Soc. Astron. Italiana, 81, 181 Saha, L., Domínguez, A., Tibaldo, L., et al. 2020, ApJ, 897, 131, doi: 10.3847/1538-4357/ab9ac2
  48. [48]Schwarz, G. 1978, The Annals of Statistics, 6, 461 , doi: 10.1214/aos/1176344136DOI
  49. [49]Simon, M., & Fischer, J. 1982, in Bulletin of the American Astronomical Society, Vol. 14, 925
  50. [50]Tavani, M., Sabatini, S., Pian, E., et al. 2009, ApJL, 698, L142, doi: 10.1088/0004-637X/698/2/L142DOI
  51. [51]Vallée, J. P., & Fiege, J. D. 2005, ApJ, 627, 263, doi: 10.1086/430341DOI
  52. [52]Vianello, G., Lauer, R. J., Younk, P., et al. 2015, arXiv e-prints, arXiv:1507.08343, doi: 10.48550/arXiv.1507.08343DOI
  53. [53]Wenger, T. V., Balser, D. S., Anderson, L. D., & Bania, T. M. 2018, ApJ, 856, 52, doi: 10.3847/1538-4357/aaaec8DOI
  54. [54]Wood, M., Caputo, R., Charles, E., et al. 2017, in International Cosmic Ray Conference, Vol. 301, 35th International Cosmic Ray Conference (ICRC2017), 824, doi: 10.22323/1.301.0824DOI
  55. [55]Zabalza, V. 2015, Proc. of International Cosmic Ray Conference 2015, 922

Paper details

vlnc:W_29sbhc4n
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 6

downloads · 2

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents