Constraints on the emission region of 3C 279 during strong flares in 2014 and 2015 through VHE $\gamma$-ray observations with H.E.S.S.

https://doi.org/10.1051/0004-6361/201935704

© E S O 2019

Astronomy

&

Astrophysics

Constraints on the emission region of 3C279 during strong flares in 2014 and 2015 through VHE 7-ray observations with H.E.S.S.

H .E.S.S. C ollaboration * : H. A bdalla 1, R. A dam 26, F. A haronian 3,4,5, F. A it B enkhali3, E. O. A nguner 19, M. A rakaw a37, C. A rcaro 1, C. A rm and 22, H. Ashkar 17, M. B ackes8,1, V. B arbosa M artins33, M. Barnard 1, Y. B echerini10, D. B erge33, K. Bernlohr3, R. B lackw ell 13, M. B ottcher 1, C. Boisson 14, J. Bolm ont 15, S. B onnefoy33, J. B regeon 16, M . B reuhaus3, F. Brun 17, P. Brun 17,

M . Bryan 9, M . B uchele32, T. B ulik 18, T. B ylund 10, M. C apasso25, S. C aroff15, A. C arosi22, S. Casanova20,3, M. C erruti15,42, T. C hand 1, S. C handra 1, A. Chen 21, S. C olafrancesco21’**, M. C uryło 18, 1. D. D avids8, C. D eil3, J. D evin 24, P. deW ilt 13, L. D irson 2,

A. D jannati-Ata'127, A. D m ytriiev 14, A. D onath 3, V. D oroshenko25, L. O ’C. D rury4, J. D yks30, K. E gberts31, G. E m ery 15, J.-P. Ernenw ein 19, S. Eschbach 32, K. Feijen 13, S. Fegan26, A. Fiasson 22, G. F ontaine26, S. F unk32, M. FuBling33, S. G abici27, Y. A. G allant 16, F. G ate 22, G. G iavitto33, D. G law ion 23, J. F. G licenstein 17, D. G ottschall25, M .-H. G rondin 24, J. H ahn 3, M. H aupt33,

G. H einzelm ann 2, G. H enri28, G. H erm ann 3, J. A. H inton 3, W. H ofm ann 3, C. H oischen 31, T. L. H olch 7, M . H oller 12, D. H o rn s2, D. H uber 12, H. Iw asaki37, M. Jam rozy 34, D. Jankow sky32, F. Jankow sky23’***, A. Jardin-B licq3,1. Jung-R ichardt32, M. A. K astendieck2, K. K atarzyński35, M. K atsuragaw a38, U. K atz32, D. K hangulyan 37, B. K helifi27, J. K ing 23, S. K lepser33,

W. K luzniak30, Nu. K om in 21, K. K osack 17, D. Kostunin 33, M. K raus32, G. Lam anna22, J. L au 13, A. Lem iere27, M. Lem oine-G oum ard24, J.-P. Lenain 15, E. Leser 31,33, C. Levy 15, T. L ohse7, I. Lypova33, J. M ackey4, J. M ajum dar33, D. M alyshev25, V. M arandon 3, A. M arcow ith 16, A. M ares24, C. M ariaud 26, G. M arti-D evesa 12, R. M arx3, G. M aurin 22, P. J. M eintjes36, A. M. W. M itchell3,41, R. M oderski30, M. M oham ed 23, L. M ohrm ann 32, C. M oore29, E. M oulin 17, J. M uller 26,

T. M urach 33, S. N akashim a40, M. de N aurois26, H. N diyavala 1, F. N iederw anger 12, J. N iem iec20, L. O akes7, P. O ’B rien 29, H. O daka39, S. O hm 33, E. de O na W ilhelm i33, M . O strow ski34, I. O ya33, M. Panter 3, R. D. Parsons3, C. Perennes15, P.-O. P etrucci28, B. Peyaud 17, Q. Piel22, S. Pita27, V. P oireau22, A. Priyana N oel34, D. A. P rokhorov21, H. Prokoph 33, G. Puhlhofer 25, M. Punch 27,10, A. Q uirrenbach 23, S. Raab 32, R. R auth 12, A. R eim er 12, O. R eim er 12, Q. R em y 16, M. R enaud 16, F. Rieger 3, L. R inchiuso17, C. R om oli3,***, G. R ow ell 13, B. R udak 30, E. Ruiz-Velasco3, V. Sahakian 6, S. Saito37, D. A. Sanchez22,

A. S antangelo25, M . Sasaki32, R. Schlickeiser 11, F. Schussler 17, A. Schulz33, H. Schutte 1,U . Schw anke7, S. Schw em m er 23, M. Seglar-A rroyo 17, M. Senniappan 10, A. S. Seyffert 1, N. Shafi21, K. Shiningayam w e 8, R. S im oni9, A. Sinha27, H. Sol 14, A. Specovius32, M . Spir-Jacob 27, L. Staw arz34, R. Steenkam p 8, C. Stegm ann31,33, C. Steppa31, T. Takahashi38, T. Tavernier 17,

A. M . Taylor 33, R. Terrier27, D. T iziani32, M. Tluczykont2, C. T richard26, M. T sirou 16, N. T suji37, R. Tuffs3, Y. U chiyam a37, D. J. van der Walt 1, C. van E ldik32, C. van R ensburg 1, B. van Soelen 36, G. Vasileiadis 16, J. Veh 32, C. Venter 1, P. Vincent 15, J. Vink9,

F. Voisin 13, H. J. V olk3, T. V uillaum e22, Z. W adiasingh 1, S. J. W agner23, R. W hite3, A. W ierzcholska20,23,***, R. Yang3, H. Yoneda38, M. Z acharias^***, R. Z anin 3, A. A. Z dziarski30, A. Zech 14, A. Z iegler32, J. Zorn 3, N. Z yw ucka 1, a n d M . M eyer43

(Affiliations can be found after the references) Received 16 April 2019 / Accepted 12 June 2019

ABSTRACT

The flat spectrum radio quasar 3C 279 is known to exhibit pronounced variability in the high-energy (100 MeV < E < 100 GeV) y-ray band, which is continuously monitored with Fermi-LAT. During two periods of high activity in April 2014 and June 2015 target-of-opportunity observations were undertaken with the High Energy Stereoscopic System (H.E.S.S.) in the very-high-energy (VHE, E > 100 GeV) y-ray domain. While the observation in 2014 provides an upper limit, the observation in 2015 results in a signal with 8.7ix significance above an energy threshold of 66 GeV.

No VHE variability was detected during the 2015 observations. The VHE photon spectrum is soft and described by a power-law index of 4.2 ± 0.3.

The H.E.S.S. data along with a detailed and contemporaneous multiwavelength data set provide constraints on the physical parameters of the emission region. The minimum distance of the emission region from the central black hole was estimated using two plausible geometries of the broad-line region and three potential intrinsic spectra. The emission region is confidently placed at r > 1.7 x 1017 cm from the black hole, that is beyond the assumed distance of the broad-line region. Time-dependent leptonic and lepto-hadronic one-zone models were used to describe the evolution of the 2015 flare. Neither model can fully reproduce the observations, despite testing various parameter sets. Furthermore, the H.E.S.S.

data were used to derive constraints on Lorentz invariance violation given the large redshift of 3C 279.

Key words. radiation mechanisms: non-thermal - quasars: individual: 3C 279 - galaxies: active - relativistic processes

* e-mail: c o n ta c t.h e s s @ h e s s -e x p e r im e n t.e u

** Deceased.

*** Corresponding authors.

1. Introduction

3 C 2 7 9 (redshift z = 0.536, B urbidge & R osenberg 1965;

M arziani e t al. 1996, R A J2000 = 12h56m11.1s, D ecJ2000 = - 0 5 d4 7 m2 2 s) belongs to the class o f flat spectrum radio quasars (FSR Q s) th at are characterized by strong variability in all energy bands from rad io to y -rays, and b ro ad em ission lines (equivalent w idth > 5 A) in the optical spectrum signifying the existence o f a b road-line region (BLR). F SR Q s belong to the blazar class o f active galactic nuclei, and th eir je ts are closely aligned w ith the line o f sight (B landford & Rees 1974) resulting in strongly D oppler-boosted em ission. Spectral energy distributions (SED s) o f F SR Q s exhibit tw o broad, non-therm al com ponents. The low -energy com ponent peaks in th e infrared and is attributed to electron synchrotron em ission. In leptonic scenarios, the high-energy com ponent, w hich peaks below the G eV regim e, is attributed to inverse C om pton (IC) em ission o f the sam e electrons scattering off am bient, soft photon fields. Such soft p hoton fields can b e th e synchrotron em ission (synchrotron-self C om pton, o r SSC), photons from the accretion disk (IC /D isk), the broad-line region (IC /B LR ), o r th e infrared em ission o f the dusty torus (IC /D T). In lepto-hadronic m odels, the high-energy spectral com ponent is attributed to processes involving highly relativistic protons, such as proton synchrotron, o r secondary em ission from photo-m eson production. T he latter includes synchrotron em ission from charged pions, m uons, and the resulting secondary electrons an d positrons. F or a review o f these processes see, fo r exam ple, B ottcher (2007) .

W hile F SR Q s are b rig h t in the high-energy (HE, 100 M eV <

E < 100 G eV ) y -ra y dom ain, they are m uch fainter at very- high-energy (V H E, E > 100 G eV ) y-rays for a n um ber o f re a ­ sons. Firstly, the low p ea k energy around the low er end o f the H E y -ray dom ain m ig h t indicate a low m axim um p article L orentz fa c ­ tor, im plying em ission w ell b elow the V H E regim e. Secondly, if the y-rays are pro d u ced w ithin ~0.1 p c from th e central super- m assive b lack hole, any V H E em ission w ould b e strongly atten ­ uated by the B L R photon field. O bservations o f V H E em ission w ill therefore allow one to significantly constrain th e m inim um distance o f the em ission region from the b lack hole as th e in trin ­ sic absorption by the B L R cannot b e too severe. Thirdly, FSR Q s are found at rath e r large cosm ological redshifts, w ith th e closest V H E -detected F S R Q at z = 0.189 (PK S 0736+ 017, C erruti et al.

2017) . H ence, attenuation o f V H E y-rays by the extragalactic b ackground lig h t (EB L) w ill also red u ce the detectable y -ray flux.

3 C 2 7 9 w as detected a t V H E y-rays w ith M A G IC in 2006 (M A G IC C ollaboration 2008) and 2007 (A leksic et al.

2011) during b rig h t optical flares. H owever, it has n o t been detected a t V H E y -rays since then (H .E .S.S. C ollaboration 2 0 1 4 ; A le k s ic e ta l. 2 0 1 4 ; A rcham bault e t al. 2016) . In the H E y-ray regim e, 3 C 2 7 9 w as detected w ith both E G R E T (H artm an et al.

1999) and Ferm i-LA T (A cero et al. 2015) . D ue to the ongoing m onitoring o f Ferm i-LAT, several flares o f 3 C 2 7 9 have been observed in the last years, a few o f w hich have been subject to follow -up observations w ith C herenkov experim ents.

In A pril 2014 an d Ju n e 2015, 3 C 2 7 9 exhibited strong o u t­

bursts in the H E y -ra y b an d w ith integrated fluxes exceeding 10-5 p h cm -2 s-1 on tim escales o f a few hours (H ayashida et al.

2 0 1 5 ; P aliya 2015) . B oth flares w ere observed w ith Ferm i-LA T in pointing m ode, that is instead o f the usual survey m ode, the satellite was p ointed tow ards 3C 279 to increase the exposure. In the 2015 event, this resu lted in the detection o f very fast v ariabil­

ity on the o rder o f a few m inutes (A ckerm ann et al. 2016) on top o f th e longer-term (several hours) evolution o f the event. B oth o f these events have been follow ed up w ith the H igh E nergy

S tereoscopic S ystem (H .E .S.S.), and the results are reported here. W hile there is no detection in V H E y-rays in 2014, the 2015 observation has resu lted in a significant detection.

This p ap e r is organized as follow s: Sect. 2 describes the analysis o f the H .E .S.S. observations o f b oth flares. G iven the H .E.S.S. detection in 2015, th e analysis o f a m ultiw avelength data set o f that event is p resented in Sect. 3 . S ections 4 and 5 are devoted to a discussion and interpretation o f b o th events b ased on various m odels, w ith an em phasis placed on the 2015 event. L im its on L orentz invariance violations (LIV ) are derived in Sect. 6 . T he results are sum m arized in Sect. 7 .

T hroughout th e p ap e r a L am bda cold dark m atter co sm o l­

ogy is used w ith H 0 = 69.6 km s -1 M p c-1 , O M = 0.286, and

= 0.714 (e.g., B e n n e te ta l. 2014) . T he resulting lum inosity distance o f 3C 279 is dL = 3.11 Gpc.

2. H.E.S.S. data analysis

H .E.S.S. is located in the K hom as H ighland in N am ibia at about 1800 m above sea level. It is an array o f five Im aging A tm o ­ spheric C herenkov Telescopes, w ith four telescopes ( C T 1 - 4 ) w ith 107 m 2 m irror area arranged in a square o f 120 m side- length and one telescope (C T 5) w ith 614 m 2 m irror area in the center o f th e array. O bservations are carried out in individual runs o f typically 28 m in duration. F or p o int sources, such as 3C 279, th e array observes in w obble m ode, m eaning w ith alter­

nating offsets to the source in rig h t ascension an d declination betw een runs for im proved b ackground subtraction. W hile the array operates in stereo m ode - all telescopes poin t at the sam e sky coordinate - the analysis can b e perfo rm ed for different array layouts depending on the dem ands o f the observed source. A stereo analysis requires that C herenkov em ission b e d etected by at least tw o telescopes, w hile a m ono analysis considers p h o ­ tons detected b y C T 5. A m ono analysis w ith C T 5 typically p ro ­ vides a low er energy threshold com pared to analyses including C T 1 - 4 ow ing to the larger m irror area. T he m ain analysis is p er­

form ed using the Mo d e l analysis chain (de N aurois & R olland 2 0 0 9 ; H oller e t a l. 2015) . It is cross-checked w ith an in d e­

p en d en t calibration chain an d the analysis softw are ImPACT (P arsons & H inton 2 0 1 4 ; P arsons et al. 2015) .

In 2014, H .E .S.S. observed 3 C 2 7 9 w ith th e full array over three consecutive nights betw een A pril 2 and A pril 4 (M JD 56 7 4 9 -5 6 7 5 1 ). A m ono analysis has been conducted w ith v e r y l o o s e cu ts 1 (H .E.S.S. C ollaboration 2017) resulting in an energy threshold o f 66 GeV. Seven observation runs passed the quality selection (H .E .S.S. C ollaboration 2006) , resulting in 2.6 h o f acceptance-corrected observation tim e, an d yielding a 3 .6 ^ significance follow ing L i & M a ( 1983) . D ifferential upper lim its (99% confidence level) have been derived follow ­ ing F eldm an & C ousins ( 1998) assum ing a p h oton index o f 4.

T he index has been m otivated by the detection spectrum o f M A G IC C ollaboration (2008). T he u pper lim its are show n in Fig. 1.

O bservations in 2015 w ere conducted in five nights betw een June 15 and June 21 (M JD 5 7 1 8 8 -5 7 1 9 4 ) w ith changing array configurations. D uring the first night, Ju n e 15 (M JD 57188.7­

57188.9, “N ig h t 1” ), C T 5 was unavailable, and a stereo an al­

ysis w ith l o o s e cu ts2 (H .E .S.S. C ollaboration 2006) has been conducted on events rec o rd e d by C T 1 - 4 yielding an energy 1 The cuts refer to parameter settings for the air shower reconstruction.

2 Despite the different nomenclature, both mono and stereo analy­

sis cuts imply the lowest possible energy threshold for the respective analyses.

Fig. 1. Observed H.E.S.S. photon spectra for six data sets as labeled. Arrows mark upper limits (99% confidence level). The gray butterfly is the 1ix statistical uncertainty band of the 2015/Night 2 data set. Error bars are statistical only. The second label gives the telescope participation and the analysis used.

threshold o f 216 GeV. Q uality selection has resu lted in six obser­

vation runs fo r the analysis w ith 2.2 h o f acceptance corrected observation tim e an d a significance o f 1 .5 ^. As for 2014, d if­

ferential u pper lim its have been com puted w ith a photon index o f 4, cf. F ig. 1. A dditionally, an integrated upper lim it above 200 G eV has been com puted, w hich is show n in the lightcurve in F ig. 2 a .

D uring the second n ig h t o f observations, June 16 (M JD 57 1 8 9 .7 -5 7 1 8 9 .9 , “N ig h t 2 ”), C T 5 was available, and a m ono analysis has been conducted w ith v e r y l o o s e cuts and an energy threshold o f 66 GeV. Q uality selection has led to seven observation runs for th e analysis w ith 2.2 h o f acceptance cor­

rected observation tim e, resulting in a detection w ith 8.7<r signif­

icance. T he spectrum has been m odeled assum ing a pow er-law o f th e form

(1)

w ith norm alization N 0 = (2.5 ± 0.2stat ± 0.5sys) x 10 9 cm 2 s 1 TeV -1 , p h oton index r = 4.2 ± 0.3stat ± 0.2sys, and decorrelation energy E 0 = 98 GeV; see also Table 1. T he system atic errors have b een derived follow ing H .E .S.S. C ollaboration (2017) . The spectrum is show n as the gray butterfly ( 1 ^ statistical u ncer­

tainty band), points ( > 2 ^ significance level) and arrows (99%

confidence u pper lim its) in Fig. 1. T here is no indication for cur­

vature as the goodness-of-fit probability o f the pow er-law spec­

trum is p = 0.82. In the follow ing, H .E.S.S. d ata points that have been corrected for E B L absorption using the E B L m odel o f F ranceschini et al. (2008) , are used.

T he average flux above an energy threshold3 o f 200 G eV equals (7.6 ± 0.7stat ± 1.5sys) x 10-12 c m -2 s-1 , and is show n in Fig. 2 a . A zoom into N ig h t 2 is show n in Fig. 3a using run-w ise tim e bins. In order to b e com parable to the results o f M A G IC in 2006 and 2007 (M A G IC C ollaboration 2 0 0 8 ; A le k s ic e ta l.

2011), here the lightcurve is derived above an energy th resh ­ old o f 100 GeV. T he average flux is (6.5 ± 0.6stat ± 1.3sys) x 10-11 cm -2 s-1 , w hich is a factor ~ 1 0 less than the flux during the M A G IC detection in 2006 (M A G IC C ollaboration 2008) . T here is n o indication for statistically significant variations in this lightcurve, as a constant flux has a p robability o f p = 0.39 (X2/n d f = 7 .6 /6 ).

O bservations on June 17 (M JD 5 7 190.7344-57190.8569,

“N ig h t 3” ) w ere conducted using only C T 1, 3 and 4. Six runs passed th e quality selection, and a stereo analysis w ith l o o s e cuts resu lted in a significance o f - 0 . 6 ^ in 2.3 h o f acceptance corrected observation tim e. T he differential upper lim it spec­

trum (photon index 4) is show n in Fig. 1, w hile the integrated upper lim it above an energy threshold o f 200 G eV is show n in F ig. 2a .

O n June 18 (M JD 5 7 1 9 1 .7 819-57191.9193, “N ig h t 4 ”) all five telescopes p articipated in the observations. H owever, only tw o o f the five conducted runs p assed the C T 5 quality selection, w hich is w hy a stereo analysis w ith l o o s e cuts has been done on all five runs w ith only the sm all telescopes. T he analysis resulted in a significance o f - 2 . 0 ^ in 1.7 h o f acceptance corrected obser­

vation tim e. T he differential u pper lim it spectrum is show n in Fig. 1 an d w as com puted w ith a photon index o f 4, w hile the integrated u pper lim it above an energy threshold o f 200 G eV is given in F ig. 2a .

Two m o re runs w ere taken on Ju n e 20 (M JD 57193.8339­

57193.8740, “N ig h t 5”) w ith all five telescopes. However, as in N ig h t 4, the data reco rd ed w ith C T 5 did n o t pass the q u al­

ity selection. H ence again a stereo analysis w ith l o o s e cuts has been p erform ed on the d ata reco rd ed w ith the sm all telescopes.

D u e to m oon constraints th e observations started relatively late, resulting in elevations o f less than 52°. This explains the high energy threshold o f m o re than 40 0 G eV in this night. T he signifi­

cance is - 0 . 3 ^ in 0.7 h o f acceptance co rrected observation tim e.

As before, the differential upper lim it spectrum (photon index 4) is show n in F ig. 1, w hile the integrated upper lim it above an energy threshold4 o f 200 G eV is show n in F ig. 2 a ).

W hile th e lightcurve show n in F ig. 2 a m ay b e suggestive o f variability, th e u pper lim its and the flux p o in t have been achieved w ith different array configurations. A n analysis o f N ig h t 2 using only the d ata from C T 1 - 4 results in no detection w ith an integrated u pper lim it com parable to the o ther nights. As the m ultiw avelength flare subsided after N ig h t 2, and n o further detections w ere achieved w ith H .E.S.S. after that night, th e fo l­

low ing d iscussion w ill focus on N ights 1 and 2 only.

3. Multiwavelength observations of the 2015 flare In F igs. 2 and 3 lightcurves a t different w avelengths o f th e 2015 flare are shown. T he analyses are p resented below.

3 The threshold of 200 GeV has been chosen for comparison with the upper limits of the other nights.

4 This involves an extrapolation to this energy threshold, which is nec­

essary to be comparable with the other nights.

d N I E \ - r d E 4 E d )

5 h t t p s : / / f e r m i . g s f c . n a s a . g o v / s s c / d a t a / a c c e s s / l a t / B ackgroundM odels.htm l

Fig. 2. Observed multiwavelength lightcurves. (a) H.E.S.S. lightcurve derived above an energy threshold of 200 GeV in night-wise time bins with array configuration as indicated. Arrows mark upper limits (99%

confidence level). (b) Fermi-LAT lightcurve integrated above 100 MeV in 3 h bins. Gray arrows mark upper limits (95% confidence level).

(c) HE y-ray photon index measured with Fermi-LAT in 3 h bins.

(d) Swift-XRT lightcurve integrated between 2 and 10keV for individ­

ual pointings. (e) Optical R band lightcurve from ATOM and SMARTS for individual pointings. f ) Spectral index between the J and B band using SMARTS observations for individual pointings. In all panels, only statistical error bars are shown.

3.1. H E y--ray data

F or the H E band, d ata taken w ith the the L arge A rea Tele­

scope (A tw ood et al. 20 0 9 , LAT) on-board th e Ferm i satellite have been analyzed. T he F erm i-L A T analysis has been car­

ried out using th e Science Tool version 10.0.5 and In stru ­ m en t R esponse Functions (IR Fs) P8R2_SOURCE_V6. D ata have been analyzed first on a 28 day interval, from M JD 57174 to M JD 57202 using a B in n ed A nalysis m eth o d (M attox et al.

1996) on a square region o f interest o f 30° side length and an energy range going from 100 M eV to 300 GeV. N earby sources have been m odeled using the 3FG L catalog (A cero et al.

2015) up to a radial distance from the central source o f 25° . T he spectral param eters o f these back g ro u n d sources are k ep t free if they are w ithin a circle o f 5° from the p o si­

tion o f 3C 279. In th e annulus w ith angular distances betw een 5° and 15° only th e flux norm alization is left free to vary.

A ccording to th e recom m endations o f the Ferm i-LA T collab o ra­

tion, th e b ackground m odels iso_P 8 R 2 _ S O U R C E _ V 6 _ v 0 6 .tx t (isotropic) and g l l _ i e m _ v 0 6 . f i t (galactic)5 are used w ith their n orm alization fit to the data.

T he lightcurve and spectra for 3C 279 are obtained b y fix­

ing all the b ackground sources in the b est fit m o d el obtained from the 28-day tim e interval, leaving only the spectral p ara m ­ eters for 3C 279 free to vary. D u e to the very high level o f p h o ­ ton counts available w ith Ferm i-LA T for this event, it is possible firstly to perfo rm a d etailed 3 h b in n e d lightcurve o f the source near the peak o f the em ission show n in F igs. 2b and 3b along w ith the p hoton index in F igs. 2c an d 3c , and secondly to co m ­ pu te the H E y -ray spectrum in tim e intervals strictly sim ultane­

ous w ith the first and second nig h t o f th e H .E.S.S. observations.

In order to create a self-consistent m odel o f the evolution o f the flare (see Sect. 5.4) tw o m o re spectra are produced, nam ely for the “Preflare” tim e fram e an d the “M axim um ” o f the Ferm i-LAT lightcurve betw een N ig h t 1 and N ig h t 2. T he p recise integration tim es are given in Table 1. F or the calculation o f th e Ferm i-LAT SED points, a likelihood fit has been p erform ed in the d esig ­ nated energy range, w ith all free param eters fixed to the best pow er-law fit values except the n orm alization o f 3C 279. As for lightcurves, a flux p o in t has been com puted in case the signif­

icance in the bin is above 3 ^ , a 95% upper lim it has been cal­

culated otherw ise, assum ing the b est-fit pow er-law p h oton index over the entire energy range.

In the 3FG L catalogue the H E spectrum is b etter described by a log-parabola function o f th e form

d N ( E r ( r+3log % )

d E = ° ( Eq) ( )

w ith the curvature p aram eter )3. In th e short tim e intervals o f the observations considered here, only for the M axim um tim e fram e a curved spectrum is p referred on a 4 ^ significance level over a pow er-law . T he fit param eters are as follow s: N 0 = 31 ± 2 x 10- 5 p h c m - 2 s - 1 G eV - 1, r LAT = 1.96 ± 0.05, a n d j6 la t = 0.12 ± 0.03 at an energy scale E 0 = 0.342 GeV. T he b est fit spectral values using a pow er-law , E q. ( 1), are rep o rted in Table 1.

3.2. X -ray data

T he N e il G ehrels S w ift observatory (G ehrels et al. 2004) includes three instrum ents: the B u rst A lert Telescope (BAT, B arthelm y et al. 2005), the X -ray T elescope (XRT, B urrow s e t al. 2005) an d th e U ltraviolet/O ptical Telescope (UVOT, R om ing et al. 2005). T hese three instrum ents provide coverage o f the follow ing energy ranges: 5 - 1 5 0 k e V (BAT), 0 .3 -1 0 k e V (XRT), and in six optical and u ltraviolet filters in the 1 7 0 -6 0 0 n m w avelength ran g e (UVOT).

X R T data collected in 2015, w ith O bservation Ids 0 0035019171-00035019188, have been analyzed using version 6.21 o f th e H E A S O F T p ackage6. D ata calibration has been p erform ed using the x r t p i p e l i n e procedure and spectral fitting o f each single observation has been perfo rm ed w ith the XSPEC softw are (A rnaud 1996) . F or th e fitting, all observations have been b in n e d to have a t least 30 counts p er b in and each single observation has been fit w ith a single pow er-law m odel w ith a G alactic absorption value o f N H = 2.01 x 1020 cm -2 (K alberla et al. 2005) set as a frozen param eter.

T he only strictly sim ultaneous Sw ift observation was during the M axim um tim e fram e. F or N ig h t 1 and N ig h t 2, observations have been chosen th at w ere conducted close to the tim e fram es defined in Table 1. T he respective O bservation IDs, as w ell as observation tim es are sum m arized in Table 2 , w hile the spectral results are given in Table 1. T he lightcurve is show n in Fig. 2 d and zoom in on N ig h t 2 in Fig. 3d .

h t t p : / / h e a s a r c . g s f c . n a s a . g o v / d o c s / s o f t w a r e / l h e a s o f t 6

Table 1. Power-law fit of H.E.S.S. (E0 = 98 GeV), Fermi-LAT (E0 = 342 MeV), and Swift-XRT (E0 = 1 keV) observed spectra for the considered time frames.

Time frame H.E.S.S. Fermi-LAT Swift-XRT

MJD N0 [phcm-2 s-1 TeV-1] r H.E.S.S. N0 [phcm-2 s-1 GeV-1] Tlat N0 [phcm-2 s-1 keV-1] Txrt Preflare

Night 1 Maximum Night 2

57184.0-57187.0 57188.756-57188.880 57189.125-57189.250 57189.734-57189.888

Upper limit

(2.5 ± 0.2) x 10-9 4.2 ± 0.3

(1.1 ± 0.1) x 10-6 (9.2 ± 0.9) x 10-6 (27 ± 1) x 10-6 (7.7 ± 0.8) x 10-6

2.3 ± 0.1 2.2 ± 0.1 2.09 ± 0.04

2.1 ± 0.1

(5.9 ± 0.3) x 10-3 (8.3 ± 0.4) x 10-3 (3.8 ± 0.2) x 10-3

1.30 ± 0.05 1.16 ± 0.06 1.43 ± 0.07 Notes. The MJD values give the integration time for the Fermi-LAT spectra, and the other spectra are chosen to be as contemporaneous as possible.

Only statistical errors are given.

Time frame ObsID tstart [MJD] tdur [s] UVOT

Preflare _ — — —

Night 1 00035019176 57188.603 1996 U

Maximum 00035019180 57189.144 962 UVW2

Night 2 00035019181 57189.670 938 UVW2

Table 2. Swift-XRT observations of 3C 279 used for the time frames defined in Table 1.

Fig. 3. Observed multiwavelength lightcurves zoomed in on Night 2.

(a) H.E.S.S. lightcurve derived above an energy threshold of 100 GeV in run-wise time bins. (b) Fermi-LAT lightcurve integrated above 100 MeV in 3 h bins. (c) HE y-ray photon index measured with Fermi- LAT in 3 h bins. (d) Swift-XRT lightcurve integrated between 2 and 10keV for individual pointings. (e) Optical R band lightcurve from ATOM and SMARTS for individual pointings. In all panels, only statis­

tical error bars are shown, while horizontal bars mark the observation time.

3.3. U V/O ptical/IR data

Sim ultaneously w ith XRT, 3 C 2 7 9 w as m onitored in the ultraviolet and optical bands w ith the U V O T instrum ent.

O bservations w ere taken in six filters: U V W 2 (192.8 nm ), U V M 2 (224.6 nm ), U V W 1 (260.0 nm ), U (346.5 nm ), B (439.2 nm ), an d V (546.8 nm ) (P oole e t a l. 2008) . M a g ­ nitudes an d fluxes have been calculated using u v o t s o u r c e including all photons from a circular region w ith radius 5 ".

In o rder to determ ine the background, a circular region w ith a radius o f 10" located near the source area has been selected.

Notes. The columns give the time frame, the Observation ID, the start time and the duration of the observation. The last column gives the UVOT filter.

A ll d ata points are co rrected for dust absorption using the reddening E (B - V) = 0.0245 m ag (Schlafly & F inkbeiner 2011) and the ratios o f the extinction to reddening, Aa/E ( B - V) (G iom m i e t al. 2006) . U nfortunately, only one U V O T filter was used p er Sw ift pointing (see Table 2 ) during th e flare. H ence, w hile the resulting fluxes are used in the S E D in F ig. 4 , no lightcurve is show n in F ig. 2 .

T he A utom atic T elescope for O ptical M onitoring (ATOM, H auser e t a l. 2004) is a 75 cm optical telescope located at the H .E .S.S. site in N am ibia. S ince 2005, it has m o n ito re d around 300 y -ray em itters and provides optical d ata for H .E.S.S. obser­

vations. In 2015, 3 C 2 7 9 was m onitored w ith ATOM in the R -band from M arch until A ugust. F ollow ing a rise in flux in June and coinciding w ith th e H .E .S.S. T arget-of-O pportunity observations, coverage w as increased to up to 20 exposures per night, evenly spread during the tim e interval from 17h30 to 2 1 h 0 0 U T C . T he flux o f each observation has been derived using differential photom etry using six secondary standard stars from G onzalez-P erez et al. (2001) in the sam e field-of-view . T he data points have been extinction-corrected sim ilar to the U V O T data.

SM AR TS (Sm all an d M oderate A perture R esearch Tele­

scope S ystem ) is an optical an d infrared telescope d edicated for observations o f F erm i-LAT blazars, visible from th e SM ARTS site in C hile (B onning et al. 2012) . 3 C 2 7 9 has been m onitored w ith the instrum ent regularly since M ay 2008. In this paper, the observations collected for the b lazar in the season o f 2015 in the B , V , R , and J bands have been analyzed. SM AR TS data have been co rrected for extinction using the corresponding b and extinctions from th e G alactic D u st R eddening and E xtinction S ervice7.

T he R -band lightcurve is show n in Fig. 2 e , w hile the spectral index betw een th e B an d J band, calculated as

(3)

h t t p : / / i r s a . i p a c . c a l t e c h . e d u / a p p l i c a t i o n s / D U S T / 7

log vFj - log vFb a J - B = ^ ----,---,

log V j - log V b

Fig. 4. Observed multiwavelength SED for the considered time frames with black dots for the Preflare time frame, red filled squares for Night 1, green open squares for the Maximum, and blue diamonds for Night 2. The y-ray data have been corrected for EBL absorption using the model by Franceschini et al. (2008). The solid lines show a power-law interpola­

tion for the X-ray to y-ray spectrum, as described in the text.

is show n in F ig. 2 f . H ere, vFj and vFb are the energy fluxes in the J and B band, respectively, w hile vj and vb are the respective central frequencies o f th e filters. A zoom -in on th e R -band fluxes o f N ig h t 2 is show n in Fig. 3e .

3.4. D isc u ssio n

T he H E y -ray flux, cf. Fig. 2b , increases by roughly a factor 6 from th e P reflare p eriod to N ig h t 1, follow ed by another increase by a factor ~ 3. T he m axim um is, hence, a factor ~ 2 0 above the Preflare value. N ig h t 2 is a factor ~ 4 below the m axim um and about 30% below th e N ig h t 1 flux.

T he X -ray flux, cf. Fig. 2 d , increases by a factor ~ 2 from N ig h t 1 to the M axim um , and drops subsequently b y a factor

~ 3.5. T hese are sim ilar to the ratios o f the H E y -ra y lightcurve and indicate a ro ughly sim ultaneous variation o f th e tw o bands.

T he optical R -band flux rises b y about 40% from the P re ­ flare to N ig h t 1, and is at a sim ilar value in N ig h t 2, as is show n in F ig. 2 e . T he detailed lightcurves from ATOM , as given in Fig. 3e , indicate m in o r intranight fluctuations. H owever, the average value is a g o o d indicator o f the optical flux state across the observation window.

L ightcurves are typically exploited to derive a characteristic tim escale o f a flaring event. F or th e 2015 flare, A ckerm ann et al.

(2 016) derived a flux doubling tim escale o f less than 5 m in du r­

ing the M ax im u m tim e fram e. H ow ever, as the flare b racketed by N ights 1 an d 2 lasts for roughly a day, a tim escale on the order o f m inutes is n o t representative o f the w hole event. F rom the HE y -ray lightcurve in F ig. 2b , the rise tim e from th e low -point around N ig h t 1 to the M ax im u m is ab o u t 9 h. T he subsequent decay is w ell d escribed by an exponential function, if the sm all fluctuations on top o f the tren d are disregarded. A n exponen­

tial decay is expected from p article cooling, o r if th e particles

leave the em ission region on an energy independent tim escale.

Perform ing an exponential fit to the decaying lightcurve, one obtains a tim escale o f ~ 9 h. H ence, this value is considered as the characteristic tim escale o f the event.

T he observed m u ltiw avelength SED s are show n in Fig. 4 for the tim e fram es defined in Table 1. In cases w here m u lti­

p le observations are available w ithin a tim e fram e, th e d ata have been averaged. T he spectral param eters o f individual frequency ranges are im portant for m o deling purposes, since they reveal inform ation abo u t the underlying p article distribution.

T he high fluxes during th e flaring event allow a p recise d eter­

m ination o f th e spectral index in the H E y -ray b an d in th e 3 h tim e bins, as show n in F ig. 2c . D uring th e flaring event the index is ~ 2.2, and hardens significantly to ~ 2 .0 during the M axim um betw een N ig h t 1 and 2 (see also Fig. 3c ). A fterw ards the index softens w hile the flux returns to the quiescence level. A t this flux level, the error on the index becom es large fo r 3 h tim e bins, and n o further conclusions can be draw n as the evolution o f the index. T he specific param eters for the averaged spectra show n in Fig. 4 are listed in Table 1.

T he X -ray spectrum changes significantly during the flare, as given in Table 1. T he spectrum hardens from N ig h t 1 to the M axim um , and softens to N ig h t 2 w ith the spectrum o f N ight 2 being even softer than the one in N ig h t 1. E xtrapolating the X -ray spectra tow ards th e y -ray dom ain w ould overpredict the y -ray fluxes in all tim e fram es.

H ence, th e b ro ad ran g e o f frequencies betw een th e S w ift- X R T and F erm i-LAT spectrum (the explicit energy ranges are given in Table 3) has been interpolated. It is assum ed th at the frequency ran g e can b e fit by a pow er-law w ith spectral index a , th at is th e energy flux is described b y v F V ^ va w ith the spectral flux density F v. T he resulting indices are rep o rted in Table 3 and the interpolation is show n in Fig. 4 . T he index is positive and

Table 3. Spectral indices of the optical spectrum and interpolation between the X-ray and y-ray spectrum.

Time frame aj-B X-ray-y-ray index [Ex , Ey]

Preflare -0.47 ± 0.01 - -

Night 1 -0.55 ± 0.02 0.42 ± 0.02 [7.1 keV, 150 MeV]

Maximum - 0.45 ± 0.01 [5.5 keV, 150 MeV]

Night 2 -0.57 ± 0.01 0.44 ± 0.02 [5.2 keV, 150 MeV]

Notes. The fourth column gives the energy range of the X-ray to y-ray interpolation.

constant w ithin errors during the flare w ith a ~ 0.44. U n fo r­

tunately, there is no inform ation on the P reflare tim e fram e.

T he indices o f the interpolation are softer than the X -ray spec­

tral indices8. W hile the X -ray spectra them selves are com patible w ith sim ple pow er-law s, th eir spectral points and the interp o la­

tion lines in Fig. 4 are suggestive o f a b reak above a few keV.

T he indices in th e optical energy ran g e betw een the J and the B band, given in Table 3 an d show n in Fig. 2f , are derived from the SM A R TS observations as described in the previous section. T he spectrum softens significantly from th e Preflare tim e fram e to the flare, b u t is roughly co nstant during N ights 1 and 2. Sw ift-U V O T observations during the M axim um and N ight 2 tim e fram es utilized the U V W 2 filter. A s can b e seen in Fig. 4 , their fluxes are com patible, and the N ig h t 2 data poin t agrees w ell w ith an extrapolation o f th e o ther optical points. This in d i­

cates that the o ptical to U V flux m ay have been constant during the m axim um o f th e flare. A nother possibility could b e th at the flux in the optical b an d increased, b u t the spectrum softened in order to p reserve the U V flux.

4. The flare in April 2014

T he m ultiw avelength d ata o f th e flare in 2014 w ere a n a­

lyzed, m odeled an d discussed by P a liy a e ta l. (2015) and H ayashida e t al. (2015) . P aliya e t al. (2 015) provide a 3 h-binned H E lightcurve obtained w ith Ferm i-LAT. This allow s one to get the H E y -ray fluxes during the H .E .S.S. observation w in ­ dow. T hey are ~ 3 x 10-6 p h c m -2 s-1 , ~ 4 x 10-6 p h c m -2 s-1 , and

~ 4 x 10- 6 p h c m -2 s -1 , respectively. T hese fluxes coincide w ith low -points in th e lightcurve betw een separated peaks, sim ilar to N ig h t 1 and N ig h t 2 o f the 2015 cam paign (cf. Fig. 2 ) . T he HE fluxes in 2014 are a factor 2 to 3 low er than during N ig h t 1 and 2 o f 2015, w hich explains th e n on-detection a t V H E energies.

P aliya e t al. (2015) pro d u ced a H E spectrum integrated over 6 days since M JD 56749, w hich encom passes th e H .E.S.S.

observations. T he average spectrum is significantly curved w ith photon index r LAT = 2.05 ± 0.05 and curvature j6LAt = 0.13 ± 0.039. T hese param eters are com patible w ith the p ara m ­ eters obtained in Sect. 3.1 for th e M axim um tim e fram e o f 2015. T he n orm alization for the P a liy a e ta l. (2015) spectrum is N 0 = 5.0 x 10- 6 p h c m -2 s-1 G eV -1 , abo u t a factor 5 below the n orm alization o f the M axim um tim e fram e in 2015. E x trap ­ olating the P a liy a e ta l. (2015) spectrum to 1 0 0 G eV (using the corrected value for jdLAT) one obtains an energy flux o f 6.7 x 10-12 erg cm -2 s -1 , w hich is below th e H .E.S.S. u pper lim it at that energy (cf. Fig. 1) .

8 The index of the X-ray “vFv” spectrum is a XRT = 2 - r XRT.

9 One should note that a close inspection reveals that the given value for ySLAT is too small. Better compatibility with the spectral points in Fig. 4 of Paliya et al. (2015) is obtained withySLAT ~ 0.3.

H ayashida et al. (2015) derived a H E spectrum for a 6 h tim e p eriod around the m axim um flux (integration tim e:

M JD 5 6 7 5 0 .210-56750.477), w hich is betw een the first and sec­

ond n ig h t o f the H .E .S.S. observations in that year. T he derived H E spectrum is com patible w ith a pow er-law . T he param eters are r LAT = 2.16 ± 0.06, and No = 1.3 x 10-5 p h c m -2 s-1 G eV -1 , w hich are sim ilar to the param eters obtained for N ig h t 2 in 2015.

H ence, a detection a t V H E m ay have been possible during the p eak flux in 2014.

P aliya et al. (2015) and H ayashida et al. (2015) used leptonic one-zone m odels using different com binations o f SSC, IC /B LR and IC /D T em ission for the high-energy peak. T he H .E.S.S.

upper lim its cannot constrain th e m odels.

5. The flare in June 2015

T he significant detection o f the 3 C 2 7 9 flare w ith H .E .S.S. in 2015 gives im portant constraints on the p aram eter space. T hese constraints are discussed below, and tim e-dependent leptonic and lepto-hadronic one-zone m odels are tested to account for the variability. M o st notably, the com bined fit o f the Ferm i-LA T and H .E.S.S. spectra in N ig h t 2 provides strong constraints on the absorption o f y-rays, w hich can b e used to constrain th e m in i­

m u m distance o f the em ission region to th e b la ck hole. This is p resented first, follow ed b y a b rie f description o f the prevalent therm al p h oton fields surrounding the je t, w hich w ill be used for b oth m odeling attem pts.

5.1. M inim um d ista n c e o f th e e m is s io n region from th e b la ck h ole

T he contem poraneous d ata o f F erm i-L A T and H .E.S.S. enable the search for absorption features caused by pair production o f y -rays w ith photons o f the B LR . T he latter is derived follow ing the m odel o f F inke (2016) , w hich is m otivated b y reverberation m apping and assum es th at accretion disc radiation is absorbed by the B L R clouds and re-em itted as m onochrom atic lines at fixed distances from the b la ck hole. T he approach here closely follow s the m eth o d introduced by M eyer et al. (2019) , w ho used Ferm i- LAT d ata o f six b rig h t F SR Q s to search for absorption features.

Two geom etries o f the B L R are im plem ented in the m odel.

In the shell geom etry, B L R photons are em itted in infinitesim ally thin shells around the b lack hole, w hereas in the ring geo m ­ etry, the B L R photons originate from thin rings orthogonal to the je t axis. T he m o d el includes em ission lines from Lye to H a b ut neglects any contribution from the therm al continuum . E ach line has an associated lum inosity and is em itted in a shell o r a ring a t a fixed distance (see Table 5 in F inke 2016) . As input the m odel requires the black hole m ass, M •, and the lum in o s­

ity o f the H 0 line, L(H 0). F o r 3 C 2 7 9 , lo g 10( M J M 0 ) = 8.28 w ith the solar m ass M 0 , and L (Ę 6 ) = 1.7 x 1043e r g s -1 are adopted (L iu e t al. 2006) . U sing the relations sum m arized in F inke (2016) betw een L (Ę 6 ) an d L(5100 A), as w ell as betw een L (5 1 0 0 A ) and the radius o f the Ę 6 em itting shell together w ith Table 5 o f F inke (2016) , the radius o f the L y a em itting shell, R Lya ~ 7.6 x 1016 cm , is obtained. T he L y a lum in o s­

ity is th e highest in the m odel (a factor o f 12 higher than L(HjS)) and is therefore responsible for m o st o f the absorption.

T he values for R Lya and the L y a lum inosity are broadly co n ­ sistent w ith typical values obtained from reverberation m a p ­ ping (K a s p ie ta l. 2 0 0 7 ; B e n tz e ta l. 2 0 0 9 ; M eyer e ta l. 2019) T he resulting optical depths, r ry (r, E ), for both geom etries and different distances r o f the em ission region from the central black

Fig. 5. Optical depths for y-rays emitted along the jet axis at different distances r interacting with photons of the BLR emission lines. Left:

BLR modeled with the shell geometry. The crossing of lines at low energies is due to numerical inaccuracies. Right: BLR modeled with the ring geometry. The structure in the optical depth are caused by the contributions of different emission lines to the overall optical depth.

Fig. 6. Left: best-fit spectra for the BLR ring geometry to the combined Fermi-LAT and H.E.S.S. data. Both data sets are corrected for EBL absorption following Franceschini et al. (2008). The spectral shapes do not change significantly, if the shell geometry is assumed instead. Right:

likelihood profile as a function of the distance r for the different assumed intrinsic spectra and BLR geometries.

hole are show n as a function o f the y-ray energy in Fig. 5 . The shell geom etry generally results in hig h er values o f the o p ti­

cal depth (com pare also Fig. 14 in F inke 2016) . N evertheless, the optical depths are still low er com pared to predictions o f m ore sophisticated B L R m odels that include continuum em is­

sion (e.g., A bolm asov & P outanen 2017, see also the discussion in M eyer et al. 2019) . In that sense, constraints on the m inim um distance b etw een the y-ray em itting region and the central black hole can be regarded as conservative.

T he distance r is constrained by sim ultaneously fitting the Ferm i-LA T and H .E.S.S. data, bo th corrected for the E B L influence follow ing F ranceschini et al. (2008) , w ith an intrinsic spectrum F (E ) w hich is m odified by the absorption e x p ( - r r r ) (Fig. 6 , left). T he E B L m odel of F ranceschini et al. (2008) is in good agreem ent w ith o ther E B L m odels and w ith low er lim ­ its derived from galaxy num ber counts (see D w ek & K rennrich 2013, for a review ). S ince a spectral cut-off due to absorption is degenerate w ith a cut-off o f the intrinsic spectrum , different intrinsic spectral shapes, nam ely a log-parabola (LP), a pow er law w ith sub-exponential cut-off (SEPL) and a broken pow er law (BPL) are tested. F or each com bination o f intrinsic spec­

trum and assum ed B L R geom etry (ring o r shell), the p ara m e­

ters o f the intrinsic spectrum and r are optim ized. This is done using a m axim um likelihood optim ization, w here the likelihood o f each Ferm i-LA T and H .E.S.S. spectral flux poin t is approxi­

m ated w ith a G aussian centered on the m easured flux and w ith a w idth equal to the flux uncertainty in each bin. O ne-sided G aus­

sian distributions are u sed in case o f flux u p p er lim its.

T he resulting best-fit spectra for the ring geom etry are show n in the left p anel o f Fig. 6 . T he best-fit values for r are around

~11 RLya for the ring geom etry and around ~ 1 0 RLya in the shell geom etry regardless o f tested spectral shapes. T he figure includes the X values p er degrees o f freedom (d.o.f.). The reduced X values are all above unity and the fit qualities, m e a­

sured by the p -v alu e o f the X distribution w ith corresponding d.o.f., are 0.11, 0.12, 0.06 (0.01, 0.01, 0.003) for the LP, BPL, S EPL intrinsic spectra and the ring (shell) geom etry, resp ec­

tively. For the L P case, the dotted line additionally show s the case w hen r is fixed to 2 RLya. F or such sm all values o f r, the B L R absorption leads to a sharp cut-off o f the observed spec­

trum . W e note th at for the S E P L case, a sub-exponential cut-off is p referred by the data. A standard exponential cut-off could rep ro ­ duce the Ferm i-LA T d ata and the first tw o flux points obtained w ith H .E.S.S. but w ould under-predict the flux in the highest energy bin by an order o f m agnitude.

T he right panel o f Fig. 6 shows the profile likelihood o f the fit as a function o f r. It is evident from the figure that none o f the fits significantly prefers the p resence o f an absorption fea­

ture at these large distances over the no-absorption case (w hich corresponds to the m axim um tested distance, r ~ 30 RLya).

Therefore, the m axim um likelihood approach is used to derive 95% confidence low er lim its on r . T he low er lim its are found by decreasing r until th e likelihood increases b y A ln L = 2 .7 1 /2 . A ll assum ed intrinsic spectra resu lt in roughly the sam e value o f the lim it o f r > 5 .4 R Lya = 4.1 x 1017 cm . S ince th e optical depth is sm aller for the ring geom etry, the low er lim it in this case relaxes to r > 2.6 R Lya = 2.0 x 1017 cm for the L P and SEPL intrinsic spectra. T he low er lim it is slightly low er for the B PL spectrum , r > 2.2 R Lya = 1.7 x 1017 cm . It should b e n oted that if only the Ferm i-LA T d ata points are fit w ith a pow er law, w hich is then extrapolated to higher energies including B L R absorption, the flux for all H ESS data points is severely under-predicted for r < 7 x 1016 cm . This m o d el does n o t provide a satisfactory fit to the H .E .S.S. data an d is especially in tension w ith the highest energy H .E.S.S. data point, w hich it under-predicts b y m o re than an o rder o f m agnitude. In conclusion, th e em ission zone is confi­

dently placed beyond r ~ 1.7 x 1017 cm (or 3 x 103 S chw arzschild radii), outside the BLR.

5.2. T h e ex te rn a l p h o to n fields

In this section, the p h oton fields external to th e je t o f 3 C 2 7 9 are described. T he param eters are listed in Table 4 and are used for the leptonic and lepto-hadronic m odels described in th e next sections.

T he accretion d isk is m odeled as a S hakura-Sunyaev disk (S hakura & Sunyaev 1973) w ith a lum inosity L acc = 3.0 x 1045 erg s-1 , w hich is the average o f values given in th e litera­

ture (e.g., H ayashida e t al. 2 0 1 5 ; P aliya et al. 2015) . T he ac cre­

tion disk lum inosity is abo u t 8% o f the E ddington pow er Ledd = 3.78 x 1046 erg s-1 o f b lack hole w ith m ass M bh ~ 3 x 108 M 0 (and references therein H ayashida e t al. 2015) . T he inner radius o f the d isk is set to the innerm ost stable orbit o f a S chw arzschild b lack hole, nam ely R acc,in = 6 x R g w ith th e gravitational radius o f th e b lack hole Rg. T he o uter radius can b e estim ated follow ­ ing N etzer (2015) , and m arks the p o in t w here the self-gravity o f the d isk surpasses the gravity o f th e b lack hole leading to disk fragm entation further out. F or 3C 279 this corresponds to R acc,out ~ 430 x R g.

U nlike the lines, th e therm al B L R param eters are n o t w ell know n for 3C 279. U sing the num bers from the previous section, the radius o f the B L R is r BLR = R Lya, and the lum inosity is assum ed as Lblr = 2.3 x 1044 erg s-1 . This corresponds to about 8% o f the accretion disk lum inosity. T he given B L R lum inosity contains the sum o f the line lum inosities plus a therm al co n tri­

bution. T he B L R tem perature is set to TBLR = 1.0 x 104 K. N ote that for th e inverse C om pton process the B L R line em ission can be w ell approxim ated by a therm al continuum .

As th e d iscussion in Sect. 5.1 indicates th at the em ission region is located b eyond the BLR , its em ission m ay b e an in e f­

ficient target fo r the IC process. W hether the strong accretion disk radiation is a useful target field despite being strongly d e­

boosted, can n o t b e stated a priori. T herefore, w e also invoke the therm al field o f a dusty torus, despite the fact th at there is no evidence o f its p resence in 3C 279. U sing estim ates from H ayashida e t al. (2012) , the radius o f the D T becom es r DT = 4.23 x 1018 cm , w hile the lum inosity in this case is assum ed to be 10% o f the accretion disk. T he tem perature is assum ed to be Tdt = 500 K.

5.3. L ep to n ic o n e - z o n e m o d e l

T he leptonic one-zone m o d el is still the standard m odel for blazar physics, either in th e m o st fundam ental version w ith

Table 4. Parameter description of the external photon fields, symbol and value.

D efinition S ym bol Value

A ccretion disk lum inosity L acc 3.0 x 1045 e r g s -1 B L R lum inosity LBLR 2.3 x 1044e r g s -1

B L R radius r BLR 7.6 x 1016 cm

B L R tem perature Tblr 1.0 x 104 K

D T lum inosity _{l dt} 3.0 x 1044 e r g s -1

D T radius rDT 4.2 x 1018 cm

D T tem perature Tdt 5.0 x 102 K

Table 5. Leptonic model parameter description, symbol and value.

Definition Symbol Value

Emission region distance r 1.7 x 1017 cm

Emission region radius R 1.8 x 1016 cm

Doppler factor of emission region 8 30

Magnetic field of emission region B 0.65 G Electron injection luminosity Le .inj 8.0 x 1041 erg s-1 Minimum electron Lorentz factor Ymin 8.0 x 102 Maximum electron Lorentz factor Tmax 5.0 x 104

Electron spectral index se 2.94

Escape time scaling desc 5.0

Acceleration to escape time ratio dacc 1.0

Magnetic field variation AB1 -0.39 G

A B2 ^{-0.52 G}

e-injection luminosity variation A Le .,inj,1 6.0 x 1042 erg s-1 A L e

inj,2 3.6 x 1043 erg s-1 Min. e-Lorentz factor variation Ayemin. 8.0 x 102

e-spectral index variation Ase 0.18

Notes. Parameters listed below the horizontal line describe the variability.

synchrotron-self C om pton (SSC ) o r in the slightly extended v er­

sion w ith external fields, such as the accretion disk, th e broad- line reagion (B LR ) and the dusty torus (DT). Its advantage is the relatively low n um ber o f param eters, o f w hich a lo t can be constrained. F ro m now on param eters m arked w ith a prim e are considered in the h o st galaxy fram e, quantities w ith an asterisk are in th e ob serv er's fram e, and unm arked quantities are either in the com oving je t fram e o r invariant.

T he param eters used for the m odeling are listed in Tables 4 and 5 . P roper explanations o f th e param eters and the descrip ­ tion o f the code are given in A ppendix A . S om e o f the p ara m ­ eters can b e analytically constrained, w hich is also described in A ppendix A .

T he m odeling aim s to reproduce th e flare a t th e tim e around the H .E.S.S. observations. H ence, first the P reflare SED is rep ro ­ duced w ith the param eters listed above the horizontal line in Table 5, follow ed by N ig h t 1. T hen the M axim um is m odeled, after w hich the evolution is follow ed to N ig h t 2. T he variability is m odeled w ith th e follow ing p aram eter changes:

B(t) = B + AB1 (H [t, t*, t*] + H [t, tm, t*])

+ A B2^{H [t, t}

\,

L nj(0 = Lnj + ALnj,1 (H 4 t*, 4 + H 4 4 , t2])

+ ALnj,2 H 4 4 tm] (5)

Fig. 7. Multiwavelength spectra and models for the four time frames: Preflare (black dots), Night 1 (red filled squares), Maximum (green open squares), and Night 2 (blue diamonds). The y-ray data have been corrected for EBL absorption using the model by Franceschini et al.

(2008). The thick solid lines mark the leptonic models. The thin lines mark spectral components for the Preflare period as labeled.

rmin(t) = rmin + Ar!min H [u fs ’ f2\ (6)

se(t) = se + A se H [t’ t*’ f2\ ’ (7)

w here t* = M JD 57186.875 m arks the beginning o f the flaring event, t\ = M JD 57188.875 m arks N ig h t 1, fm = M JD 57189.125 the M axim um and t* = M JD 57189.875 N ig h t 2. W hile these tim e steps are defined in the o b serv er’s fram e, they are p ro p ­ erly transform ed to the com oving fram e in the code. T he step function H IX a ’ b] is 1 for a < x < b and 0 otherw ise. H ence, the variability is m odeled by 1 o r 2 step-function-like changes in the param eters. T he variability param eters are listed in Table 5 below the horizontal line. A reasoning fo r the adopted param eter changes is provided in A ppendix A .

T he resulting m odel S ED s are show n in Fig. 7 . T he optical regim e is dom inated by synchrotron photons, w hile th e X -ray regim e is m ostly SSC, and th e y -ra y regim e is d o m inated by the IC /B L R process. T he S E D s are reproduced w ell for the P re ­ flare, N ight 1 and M axim um tim e fram es except in th e X -ray dom ain. H ow ever, these tim e fram es can b e directly influenced by the changes in the param eters. Subsequently, the injection is returned to N ight 1 levels, so the continuing evolution is given by the cooling and escape o f the particles. As N ight 2 is not reproduced w ell in the X -ray and y -ray energy bands, the ch o ­ sen p aram eter set is not adequate to reproduce the decay from the M axim um to N ight 2.

In order to im prove th e fit in the X -ray dom ain, a h igher SSC flux is required. This could b e achieved b y a larger n um ber o f particles, w hich w ould how ever also increase the synchrotron and IC /B L R fluxes. This could be alleviated b y reducing the m agnetic field and the lum inosity o f the B L R . H owever, the latter is already close to the allow ed flux from the line m e a­

surem ents. Increasing th e m agnetic field, w hich w ould in turn increase the SSC flux, w ould req u ire a b righter B L R in order

to preserve the C om pton dom inance. A dditionally, this w ould require less particles in the em ission region. As th e SSC flux depends linearly on th e m agnetic field but quadratically on the particle density, the SSC flux w ould actually drop.

T he b ad fit o f N ight 2 is driven b y the slow p article escape due to the large em ission region. Instead o f leaving th e source, the particles are shifted to low er energies. This has no co n se­

quences for the o ptical dom ain, w hich is d o m inated b y particles that cool quickly, explaining the good fit. H owever, the X -ray and H E y -ra y dom ains are d o m inated b y th e inverse C om pton radiation o f less energetic particles. In this energy regim e p arti­

cles have p iled up, as the original ones have n o t yet cooled away, w hile further particles have reached this energy by cooling dow n from higher energies.

This could b e alleviated b y a faster escape o f the particles from th e em ission region. As the escape is controlled b y b oth the size o f the em ission region R, and the escape tim escale param eter nesc, either o f them could b e red u ced to accelerate th e escape.

H owever, ^esc is already set to only 5.0, im plying th at particles rem ain w ithin the em ission region fo r only five light crossing tim escales. This is already a very fast escape, as one expects som e diffusion w ithin the em ission region due to the m agnetic field.

H ence, reducing R is used to accelerate the escape o f particles, as the constraint from the characteristic variability tim escale only provides an u pper lim it. H owever, reducing R enhances the energy densities o f particles and p h oton fields w ithin the em ission region. W hile this can be accom m odated easily fo r th e synchrotron and external-C om pton com ponent by reducing the nu m b er o f injected particles, the SSC flux w ould drop, as outlined above, and therefore m ake the fit even w orse.

A nother p ossibility is to (additionally) increase the D oppler factor S. As this value has a d irect im p act on th e internal energy

Table 6. Poynting power, proton power, electron power, and radiative power in the observer’s frame for the leptonic model curves in Fig. 7.

Sym bol P reflare N ig h t 1 M axim um N ig h t 2

LB 1.8 x 1045 1.9 x 1044 5.3 x 1043 1.9 x 1044

Lp 5.3 x 1045 1.1 x 1046 2.3 x 1046 2.1 x 1046

L ep 4.9 x 1044 1.8 x 1045 6.9 x 1045 2.7 x 1045

LS 2.4 x 1045 1.6 x 1046 3.9 x 1046 3.0 x 1046

Notes. Powers in units of erg s 1.

densities o f the em ission region, the param eters have to be changed considerably. However, also in this case a p erfect fit is n o t possible under th e given constraints, w hich is show n in Fig. A .1 .

It should be n o te d th at despite the m e n tio n e d problem s, the H .E.S.S. spectrum is fit w ell. If th e escape p roblem could be solved, the fit w ould actually be p erfect in th e H .E .S.S. dom ain as th e F erm i-LAT spectra o f N ig h t 1 and 2 are sim ilar, and so w ould b e the m odels.

As m e n tio n e d above, a higher SSC flux could b e achieved w ith a larger nu m b er o f particles in the je t, w hile reducing the m ag n etic field and the external field. W hile reducing the B LR lum inosity is n o t possible, the em ission region could b e m o v ed to an even further distance from the b lack hole, w here the D T p hoton field dom inates th e external contribution. In fact, p ara m ­ eters can b e found th at allow for a go o d fit in large parts o f the spectrum , b u t n o t p erfectly at all energies, cf. F ig. A .2 . T he m ain issue is again the escape o f particles, b u t the delicate interplay o f the param eters does n o t allow to reduce the size o f th e em ission region in this case.

H ence, despite being able to fit the Preflare, N ig h t 1 and M axim um tim e fram es rath er w ell in som e cases, th e subsequent decay poses a severe p roblem for th e leptonic m odel. T he in ter­

play o f the param eters is delicate and requires incredible fine tuning, w hich could n o t b e achieved for all the details o f the spectrum .

N onetheless, it is interesting to study the resulting pow er output o f the m o d e l show n in Fig. 7 . Table 6 lists the Poynt- ing pow er, proton pow er, electron pow er, an d radiative power.

T he proton pow er is calculated assum ing one cold p roton per electron. T he pow ers have been derived u nder the assum ption that the b ulk L orentz factor is given by r = 8. C om pared to the E ddington pow er o f 3 C 2 7 9 ’s b la ck hole, L edd = 3.78 x 1046 erg s-1 , the total pow er is below the E ddinton lim it during the P reflare and N ig h t 1 tim e fram es. T he M axim um , and N ight 2 exceed the E ddington pow er. B y how m u c h depends on the actual value o f th e m ass o f the black hole, w hich has an u ncer­

tainty o f m ore than a factor 2 (e.g., H ayashida e t al. 2015) . The p ow er output o f th e je t is d om inated b y particles and radiation, w hile the P oynting pow er is com parable to th e o ther constituents only during the P reflare period. T he total pow er o f the je t o f N ig h t 2 could be reduced to below the E ddington lim it if the em ission region contains 90% pairs. S ince the radiative output o f the je t is already above the E ddington lum inosity for the M a x ­ im um (keeping the uncertainty in M bh in m ind), even a high pair content w ould n o t b e able to reduce the je t pow er below that threshold. It should also b e noted th at the m o d e l w ith a larger D oppler and bulk L orentz factor (show n in Fig. A .1) results in super-E ddington je t pow ers, how ever w ith a sm aller m argin, and a high fraction o f pairs m ay p ush the total je t pow er below the E ddington lim it in this case.

Table 7. Lepto-hadronic model parameter description, symbol and value.

Definition Symbol Value

Emission region distance r 1.7 x 1017 cm

Emission region radius R 1.8 x 1016 cm

Doppler factor of emission region 8 30

Magnetic field of emission region B 50.0 G Proton injection luminosity Lp .inj 7.0 x 1043ergs-1 Minimum proton Lorentz factor / . 5.0 x 105 Maximum proton Lorentz factor Tmax 3.0 x 108

Proton spectral index sp 2.11

Electron injection luminosity Leinj 3.3 x 1041 ergs-1 Minimum electron Lorentz factor T^in 5.0 x 101 Maximum electron Lorentz factor Tmax 2.0 x 103

Electron spectral index se 2.94

Escape time scaling nesc 5.0

Acceleration to escape time ratio ^acc 30.0 p-injection luminosity variation

Max. p-Lorentz factor variation e-injection luminosity variation e-spectral index variation

ALpnj,1 ALpoj,2 ATmax ALe .inj Ase

4.5 x 1044 erg s-1 1.17 x 1046ergs-1

3.0 x 108 3.0 x 1041 ergs-1

0.18

Notes. Parameters listed below the horizontal line describe the variability.

5.4. L e p to -h a d ro n ic o n e - z o n e m o d e l

To go b eyond th e sim ple one-zone leptonic m odel, the p o ssi­

b ility o f a one-zone lepto-hadronic m o d e l is explored, follow ­ ing a sim ilar set o f source assum ptions as m a d e for the lep- tonic m odel. Typically, lepto-hadronic m odels have difficulties in reproducing fast flares ow ing to the long cooling tim escales o f protons. H owever, it w as n o te d by P etropoulou et al. (2017) that sm all scale regions w ith kG m agnetic fields could account for the m inute-scale variability even in lepto-hadronic m odels. W hile the m in u te-scale variability is n o t a concern in the p rese n t study, it show s th e p rinciple p ossibility to use lepto-hadronic m o d els to account for the 2015 flare o f 3C 279.

T he param eters reproducing the P reflare p erio d are listed in Tables 4 and 7 above the horizontal line. T hey are explained along w ith a discussion o f the constraints and th e details o f the code in A ppendix B .

A gain, a self-consistent reproduction o f the 3C 279 spectra is attem pted b y changing inp u t param eters as follows:

Lp .(injw

0

S

0

0

0

Tmax(0 = Tmax + ATmax H \t, ts, (9)

L j ) = l% + a l % h \t, t*, o ] (10)

se( 0 = s e + A se H \t, t*, t* ] , (11) w here th e tim e steps are the sam e as in the leptonic case. The m ax im u m proton L orentz factor, th e electron injection lum in o s­

ity, and the electron spectral index are only varied once during the flare and rem ain at their levels until the en d o f the flare. The p roton injection, however, is varied until the beginning o f the M axim um , w ith a single injection on top o f that, after w hich it returns to Preflare levels. T he variability param eters are listed in Table 7 below the horizontal line. T he m ag n etic field is not varied, as there is no constraint on it in this case.

T he four derived spectra for th e lepto-hadronic m o d e l are show n in Fig. 8 . T he o ptical com ponent is w ell reproduced by