Infrared molecular hydrogen lines in GRB host galaxies
K. Wiersema, A. Togi, D. Watson, L. Christensen, J. P. U. Fynbo, B. P. Gompertz, A. B. Higgins, A. J. Levan, S. R. Oates, S. Schulze, J. D. T. Smith, E. R. Stanway, R. L. C. Starling, D. Steeghs, N. R. Tanvir
aa r X i v : . [ a s t r o - ph . GA ] A ug MNRAS , 1–8 (2018) Preprint 29 August 2018 Compiled using MNRAS L A TEX style file v3.0
Infrared molecular hydrogen lines in GRB host galaxies
K. Wiersema , ⋆ , A. Togi , , D. Watson , L. Christensen , J. P. U. Fynbo , B.P. Gompertz , A. B. Higgins , A. J. Levan , S. R. Oates , S. Schulze , J. D. T.Smith , E. R. Stanway , R. L. C. Starling , D. Steeghs , N. R. Tanvir Department of Physics and Astronomy, University of Leicester, University Road, Leicester LE1 7RH, UK Department of Physics, University of Warwick, Coventry CV4 7AL, UK Department of Physics and Astronomy, The University of Texas at San Antonio, San Antonio, 1-UTSA Circle, TX 78249, USA Department of Physics and Earth Sciences, St. Mary’s University, One Camino Santa Maria, San Antonio, Texas 78228, USA The Cosmic Dawn Center, Niels Bohr Institute, University of Copenhagen, Juliane Maries Vej 30, 2100 Copenhagen Ø, Denmark Dark Cosmology Centre, Niels Bohr Institute, University of Copenhagen, Juliane Maries Vej 30, 2100 Copenhagen Ø, Denmark Department of Particle Physics and Astrophysics, Weizmann Institute of Science, Rehovot 7610001, Israel Ritter Astrophysical Research Center, University of Toledo, 2825 West Bancroft Street, M. S. 113, Toledo, OH 43606, USA
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
Molecular species, most frequently H , are present in a small, but growing, numberof gamma-ray burst (GRB) afterglow spectra at redshifts z ∼ − , detected throughtheir rest-frame UV absorption lines. In rare cases, lines of vibrationally excited statesof H can be detected in the same spectra. The connection between afterglow line-of-sight absorption properties of molecular (and atomic) gas, and the observed behaviourin emission of similar sources at low redshift, is an important test of the suitability ofGRB afterglows as general probes of conditions in star formation regions at high red-shift. Recently, emission lines of carbon monoxide have been detected in a small sampleof GRB host galaxies, at sub-mm wavelengths, but no searches for H in emission havebeen reported yet. In this paper we perform an exploratory search for rest-frame K band rotation-vibrational transitions of H in emission, observable only in the lowestredshift GRB hosts ( z . . ). Searching the data of four host galaxies, we detect asingle significant rotation-vibrational H line candidate, in the host of GRB 031203.Re-analysis of Spitzer mid-infrared spectra of the same GRB host gives a single lowsignificance rotational line candidate. The (limits on) line flux ratios are consistentwith those of blue compact dwarf galaxies in the literature. New instrumentation, inparticular on the
JWST and the
ELT , can facilitate a major increase in our under-standing of the H properties of nearby GRB hosts, and the relation to H absorptionin GRBs at higher redshift. Key words: gamma-rays: bursts, ISM:molecules
Gamma-ray burst (GRB) afterglow spectroscopy has showngreat promise as a probe of gas and dust properties within,and near, star forming regions in distant galaxies (see e.g.Schady 2015 for a review). The bright afterglows serve asbacklights with a simple (sometimes reddened) synchrotronspectrum, against which atomic and molecular absorptionlines are easily distinguished. In addition, the ultravio-let radiation of the rapidly fading afterglow excites meta-stable and fine structure atomic states. The variability of ⋆ E-mail: [email protected] absorption lines from these transitions allows precise dis-tance determination of the absorbing gas with respect tothe GRB in cases with high signal-to-noise spectra (e.g.,Vreeswijk et al. 2007). Long gamma-ray bursts (broadlyspeaking those with a duration longer than ∼ s) are as-sociated with the deaths of massive stars, and their rateis therefore thought to trace cosmic star formation, likelymoderated by a metallicity threshold for formation (e.g.Perley et al. 2016 and references therein). This means thatGRBs and their bright afterglows may form a valuable toolto locate, and study, actively star forming galaxies acrosscosmic time, unbiased by galaxy luminosity (except perhapsthrough a luminosity-metallicity dependence). A key advan- © K. Wiersema et al. tage offered by GRBs over quasars as backlights, is that thehost galaxies can be studied in emission once the afterglowshave faded, through spectral energy distributions (to ob-tain stellar population parameters, e.g. Perley et al. 2016)and emission lines (to obtain element abundances and starformation rates, e.g. Kr¨uhler et al. 2015). These studies inemission complement the line of sight studies of afterglows,connecting the afterglow sight lines through their hosts withhost galaxy integrated emission properties. Sample sizes ofhosts and afterglow spectra are growing, and statistical stud-ies of metal abundances, stellar populations and dust prop-erties are now possible, placing GRB host galaxies in thecontext of wider galaxy surveys (e.g. Vergani et al. 2007).An important component of the picture, though, the molec-ular content, is still poorly understood. Of particular interestis the H molecule, which plays a key role in the processesof star formation.An additional advantage of host studies is that it is notlimited to the subset of GRBs for which the optical after-glow is detected (provided the host can be reliably identified,see e.g. Perley et al. 2017). There is evidence that most ofthe GRB sightlines that pass by significant column densi-ties of molecules also will contain large dust column densi-ties and hence that such sightlines are underrepresented inthe subset of GRBs with well-detected optical afterglows(Prochaska et al. 2009; Kr¨uhler et al. 2013). The peculiarcase of GRB140506, for which CH+ molecules were detectedin absorption along with very steep UV extinction furthersupports this point (Fynbo et al. 2014; Heintz et al. 2017).The homonuclear H molecule does not have a dipolemoment, so electric dipole transitions between levels withdifferent vibrational quantum number ( ν ) or rotationalquantum number ( J ) in the ground state are forbidden.Quadrupole transitions, however, are allowed, and the purerotational lines (with ν = − ) are located at (mid-)infraredwavelengths, which makes them challenging to study, partic-ularly for faint sources like GRB hosts. Of more interest tous are the rotation-vibration (hereafter ro-vibration) tran-sitions in the ground state. In the following we adopt thestandard notation, where the difference in J is given by aletter (O, Q, S for ∆ J = + , , − , respectively), followed bythe final state J , and preceded by the vibrational transition(so 1–0 S(1) is ν = − , J = − ). The main ro-vibrationallines are located at near-infrared wavelengths. For example,the 1–0 S(0), 1–0 S(1) and 1–0 S(3) transitions are locatedat 2.22, 2.12 and 1.96 µ m, respectively, in the restframe. Atlow redshifts, these transitions can therefore be detected byground-based observatories.The first allowed transitions out of the H groundstate to an excited state are the Lyman and Wernerbands, which require ultraviolet (UV) photons. Thesetransitions have indeed been observed in a hand-ful of GRB afterglow spectra (Prochaska et al. 2009;Kr¨uhler et al. 2013; D’Elia et al. 2014; Friis et al. 2015, andpossibly in Fynbo et al. 2006), where the redshifts of theGRBs shift these transitions from the UV to the opticaldomain. Identification and analysis is difficult: these tran-sitions are located among the atomic hydrogen lines ofthe dense Lyman forest, and as such a fairly high signal-to-noise and spectral resolution is required to separatethem. In very rare cases, highly diagnostic absorption linesof vibrationally excited H are found (Scheffer et al. 2009; Kr¨uhler et al. 2013), that are located redwards of the Ly α line. The detection of Lyman-Werner lines, combined withfits to the atomic hydrogen Ly α line (in GRB sight linesoften found as a strong, highly damped line, a dampedLyman absorber [DLA], Jakobsson et al. 2006) has allowedestimates of the molecular gas fraction (integrated overthe line of sight), and helps to place the relatively smallGRB H absorption sample in the context of the muchlarger sample of quasar DLAs (e.g., Noterdaeme et al. 2008;Noterdaeme, Petitjean & Srianand 2015).The sightline selection function of long GRBsis arguably quite different from those of QSODLAs (e.g. Prochaska et al. 2008; Fynbo et al. 2008;Fynbo et al. 2009), which makes the long GRB afterglowH sample especially valuable as a probe of high-redshiftstar forming regions. Of particular interest is that longGRBs trace cosmic star formation (e.g. Greiner et al. 2015),and therefore the H absorption seen in afterglows mayprobe the conditions in star forming regions within (dwarf)star-forming galaxies at the peak of cosmic star formation( z ∼ − ).However, whilst afterglow sightlines likely probe regionsnear long GRBs in high mass star forming regions, whichshould be rich in H (e.g. Tumlinson et al. 2007), the lowdetection rate, the excitation state of the detected H , andoccasionally the association of the H absorber with excitedatomic metal fine structure lines, have shown that in sev-eral cases the H absorbers are likely located far from thestar forming region in which the GRB progenitor resided(e.g. D’Elia et al. 2014). The low detection rate of H in af-terglow spectra is puzzling. Several explanations have beenput forward, that likely all play a role: dissociation of the H molecules by a high UV radiation field from the intense starformation in the host galaxy (e.g. Hatsukade et al. 2014); ob-servational biases against sightlines with high dust columnsand against high-metallicity environments (e.g. Ledoux etal. 2009; Covino et al. 2013); and formation of stars fromatomic gas before H has a chance to form (e.g. Michalowskiet al. 2016).An alternative approach to detecting molecular speciesin GRB host galaxies is through emission line spectra. Thishas the added advantage of avoiding problems in interpret-ing line of sight measurements (e.g. the effects of ionisationand excitation by the GRB emission) and can help to placethe line of sight absorption in an integrated, or in low-zcases spatially resolved (e.g. Hatsukade et al. 2014), galaxycontext (Micha lowski et al. 2015). To date, a handful ofhost galaxies have been detected in molecular line emission,in all cases this is through emission lines of carbon monox-ide (CO) (Hatsukade et al. 2014; Stanway et al. 2015b;Micha lowski et al. 2016; Arabsalmani et al. 2018b;Micha lowski et al. 2018). The use of CO as a tracermolecule for H is a well established technique, thoughevidence that the CO to H conversion factor in GRB hostsightlines is comparable to Galactic translucent clouds,is limited to a single case (Prochaska et al. 2009): theonly afterglow spectrum so far with a detection of COabsorption lines. The metallicity dependence of the COto H conversion factor, and other environmental effects(e.g. Bolatto, Wolfire & Leroy 2013), make the CO toH conversion factor (and therefore a clear picture ofwhether GRB hosts are deficient in molecules or not) for MNRAS000
Gamma-ray burst (GRB) afterglow spectroscopy has showngreat promise as a probe of gas and dust properties within,and near, star forming regions in distant galaxies (see e.g.Schady 2015 for a review). The bright afterglows serve asbacklights with a simple (sometimes reddened) synchrotronspectrum, against which atomic and molecular absorptionlines are easily distinguished. In addition, the ultravio-let radiation of the rapidly fading afterglow excites meta-stable and fine structure atomic states. The variability of ⋆ E-mail: [email protected] absorption lines from these transitions allows precise dis-tance determination of the absorbing gas with respect tothe GRB in cases with high signal-to-noise spectra (e.g.,Vreeswijk et al. 2007). Long gamma-ray bursts (broadlyspeaking those with a duration longer than ∼ s) are as-sociated with the deaths of massive stars, and their rateis therefore thought to trace cosmic star formation, likelymoderated by a metallicity threshold for formation (e.g.Perley et al. 2016 and references therein). This means thatGRBs and their bright afterglows may form a valuable toolto locate, and study, actively star forming galaxies acrosscosmic time, unbiased by galaxy luminosity (except perhapsthrough a luminosity-metallicity dependence). A key advan- © K. Wiersema et al. tage offered by GRBs over quasars as backlights, is that thehost galaxies can be studied in emission once the afterglowshave faded, through spectral energy distributions (to ob-tain stellar population parameters, e.g. Perley et al. 2016)and emission lines (to obtain element abundances and starformation rates, e.g. Kr¨uhler et al. 2015). These studies inemission complement the line of sight studies of afterglows,connecting the afterglow sight lines through their hosts withhost galaxy integrated emission properties. Sample sizes ofhosts and afterglow spectra are growing, and statistical stud-ies of metal abundances, stellar populations and dust prop-erties are now possible, placing GRB host galaxies in thecontext of wider galaxy surveys (e.g. Vergani et al. 2007).An important component of the picture, though, the molec-ular content, is still poorly understood. Of particular interestis the H molecule, which plays a key role in the processesof star formation.An additional advantage of host studies is that it is notlimited to the subset of GRBs for which the optical after-glow is detected (provided the host can be reliably identified,see e.g. Perley et al. 2017). There is evidence that most ofthe GRB sightlines that pass by significant column densi-ties of molecules also will contain large dust column densi-ties and hence that such sightlines are underrepresented inthe subset of GRBs with well-detected optical afterglows(Prochaska et al. 2009; Kr¨uhler et al. 2013). The peculiarcase of GRB140506, for which CH+ molecules were detectedin absorption along with very steep UV extinction furthersupports this point (Fynbo et al. 2014; Heintz et al. 2017).The homonuclear H molecule does not have a dipolemoment, so electric dipole transitions between levels withdifferent vibrational quantum number ( ν ) or rotationalquantum number ( J ) in the ground state are forbidden.Quadrupole transitions, however, are allowed, and the purerotational lines (with ν = − ) are located at (mid-)infraredwavelengths, which makes them challenging to study, partic-ularly for faint sources like GRB hosts. Of more interest tous are the rotation-vibration (hereafter ro-vibration) tran-sitions in the ground state. In the following we adopt thestandard notation, where the difference in J is given by aletter (O, Q, S for ∆ J = + , , − , respectively), followed bythe final state J , and preceded by the vibrational transition(so 1–0 S(1) is ν = − , J = − ). The main ro-vibrationallines are located at near-infrared wavelengths. For example,the 1–0 S(0), 1–0 S(1) and 1–0 S(3) transitions are locatedat 2.22, 2.12 and 1.96 µ m, respectively, in the restframe. Atlow redshifts, these transitions can therefore be detected byground-based observatories.The first allowed transitions out of the H groundstate to an excited state are the Lyman and Wernerbands, which require ultraviolet (UV) photons. Thesetransitions have indeed been observed in a hand-ful of GRB afterglow spectra (Prochaska et al. 2009;Kr¨uhler et al. 2013; D’Elia et al. 2014; Friis et al. 2015, andpossibly in Fynbo et al. 2006), where the redshifts of theGRBs shift these transitions from the UV to the opticaldomain. Identification and analysis is difficult: these tran-sitions are located among the atomic hydrogen lines ofthe dense Lyman forest, and as such a fairly high signal-to-noise and spectral resolution is required to separatethem. In very rare cases, highly diagnostic absorption linesof vibrationally excited H are found (Scheffer et al. 2009; Kr¨uhler et al. 2013), that are located redwards of the Ly α line. The detection of Lyman-Werner lines, combined withfits to the atomic hydrogen Ly α line (in GRB sight linesoften found as a strong, highly damped line, a dampedLyman absorber [DLA], Jakobsson et al. 2006) has allowedestimates of the molecular gas fraction (integrated overthe line of sight), and helps to place the relatively smallGRB H absorption sample in the context of the muchlarger sample of quasar DLAs (e.g., Noterdaeme et al. 2008;Noterdaeme, Petitjean & Srianand 2015).The sightline selection function of long GRBsis arguably quite different from those of QSODLAs (e.g. Prochaska et al. 2008; Fynbo et al. 2008;Fynbo et al. 2009), which makes the long GRB afterglowH sample especially valuable as a probe of high-redshiftstar forming regions. Of particular interest is that longGRBs trace cosmic star formation (e.g. Greiner et al. 2015),and therefore the H absorption seen in afterglows mayprobe the conditions in star forming regions within (dwarf)star-forming galaxies at the peak of cosmic star formation( z ∼ − ).However, whilst afterglow sightlines likely probe regionsnear long GRBs in high mass star forming regions, whichshould be rich in H (e.g. Tumlinson et al. 2007), the lowdetection rate, the excitation state of the detected H , andoccasionally the association of the H absorber with excitedatomic metal fine structure lines, have shown that in sev-eral cases the H absorbers are likely located far from thestar forming region in which the GRB progenitor resided(e.g. D’Elia et al. 2014). The low detection rate of H in af-terglow spectra is puzzling. Several explanations have beenput forward, that likely all play a role: dissociation of the H molecules by a high UV radiation field from the intense starformation in the host galaxy (e.g. Hatsukade et al. 2014); ob-servational biases against sightlines with high dust columnsand against high-metallicity environments (e.g. Ledoux etal. 2009; Covino et al. 2013); and formation of stars fromatomic gas before H has a chance to form (e.g. Michalowskiet al. 2016).An alternative approach to detecting molecular speciesin GRB host galaxies is through emission line spectra. Thishas the added advantage of avoiding problems in interpret-ing line of sight measurements (e.g. the effects of ionisationand excitation by the GRB emission) and can help to placethe line of sight absorption in an integrated, or in low-zcases spatially resolved (e.g. Hatsukade et al. 2014), galaxycontext (Micha lowski et al. 2015). To date, a handful ofhost galaxies have been detected in molecular line emission,in all cases this is through emission lines of carbon monox-ide (CO) (Hatsukade et al. 2014; Stanway et al. 2015b;Micha lowski et al. 2016; Arabsalmani et al. 2018b;Micha lowski et al. 2018). The use of CO as a tracermolecule for H is a well established technique, thoughevidence that the CO to H conversion factor in GRB hostsightlines is comparable to Galactic translucent clouds,is limited to a single case (Prochaska et al. 2009): theonly afterglow spectrum so far with a detection of COabsorption lines. The metallicity dependence of the COto H conversion factor, and other environmental effects(e.g. Bolatto, Wolfire & Leroy 2013), make the CO toH conversion factor (and therefore a clear picture ofwhether GRB hosts are deficient in molecules or not) for MNRAS000 , 1–8 (2018) nfrared molecular hydrogen lines in GRB hosts GRB sightlines uncertain (e.g. Arabsalmani et al. 2018b;Micha lowski et al. 2018). In addition, most of the hostgalaxies with detected CO emission lines have a detectionof only a single transition. These reasons, together with thelow detection rate of CO absorption in optical afterglowspectra, makes a direct comparison between host galaxyCO emission and afterglow CO absorption difficult.No detections of H emission, through either pure ro-tational or ro-vibrational transitions, have been reported todate. In this paper we perform a first exploratory search forro-vibrational H lines in a sample of four, low redshift, longGRB host galaxies, to inform more sensitive searches withfuture observatories. In this paper we use a small sample of four low-redshiftGRB hosts as a pilot study. All four have z . . (Table1), to ensure a chance of detecting the 1-0 S(1) transitionin the usable range of near-infrared (NIR) spectrographs.Such low redshift GRBs are relatively rare (the mean red-shift for Swift -discovered GRBs is ∼ . , Selsing et al. 2018),and often hosts are too faint in the NIR range to have rea-sonable quality spectra at the H wavelengths. To give acrude guess at the required flux limits, we used the findingsof Pak et al. (2004), who observed a sample of vigorouslystar forming galaxies, and found that the ratio of the ro-vibrational 1–0 S(1) H line luminosity ( L H ) and the far-infrared (FIR) continuum luminosity ( L FIR ) are broadly con-stant at L H / L FIR ∼ − for a wide range in galaxy mass.The host of GRB 031203 has a well determined FIR lumi-nosity (Symeonidis et al. 2014), which gives an expected H flux of a few times − erg s − cm − .For three of the hosts in our sample, the spectra,their acquisition, reduction and calibration have been de-scribed in detail in previous papers: the hosts of GRBs060218 (Wiersema et al. 2007; Wiersema 2011), 031203(Watson et al. 2011) and 100316D (Starling et al. 2011;Wiersema 2011; Flores et al. in prep). The host ofGRB 100316D is large, and we use the spectrum of thebrightest star forming region in this host (a.k.a. source “A”,Starling et al. 2011).In addition to these sources, we observed the host ofnearby GRB 080517 (Stanway et al. 2015a). This last sourceis of particular interest because of a detection of an emis-sion line of carbon monoxide (Stanway et al. 2015b).We observed this host with the Long-slit Inter-mediate Resolution Infrared Spectrograph (LIRIS,Acosta-Pulido, Dominguez-Tagle & Manchado 2003) onthe 4.2m William Herschel Telescope, starting at 23:31UT on 3 March 2015. We used the low resolution HKgrism ( lr hk ) and a 1 arcsecond wide slit, which gave awavelength range . − . µ m. We obtained 4 noddingcycles, of 2 positions each, using an exposure time of 450s per exposure. Seeing conditions during the observationsvaried between 1.1 and 1.9 arc seconds, with the lattervalue measured on H and K s images taken directly afterthe science exposures. We reduced the data using version Figure 1.
The LIRIS spectrum of the host galaxy of GRB 080517.For plotting purposes the spectrum is smoothed with a 3 pixel me-dian filter. The wavelength is as observed. The area worst affectedby telluric absorption (i.e. with strongest residuals after telluricabsorption correction) is omitted from the plot. The dashed redlines indicate the wavelengths of Paschen α (left) and Brackett γ (right); the solid blue lines indicate the wavelengths of the H lirisdr package in IRAF . We observed thestar SAO 013747 (an A0 photometric standard star) toaid flux calibration: correction for telluric features, andflux calibration, was done using the SpeXtool softwarepackage (Cushing, Vacca & Rayner 2004), in particularthe xtellcor general task. To bring the spectrum onto anabsolute flux scale, we used LIRIS imaging observationsin J , H and K s filters. These imaging observations wereperformed on the night of 5 December 2014, starting at23:24 UT, and consisted of two cycles of 5 dither positionswith 30 seconds integration time for the H band, andthree cycles of 5 dither positions in the K s band. Theseeing was fair at 1.2 arc seconds FWHM. After sourceextraction using SExtractor (Bertin & Arnouts 1996),calibration onto 12 bright stars in the 2MASS survey givesthe following magnitudes (in the 2MASS Vega system) forthe host galaxy: H = . ± . and K s = . ± . .The response and telluric absorption calibrated host galaxyspectrum was corrected using these photometric values, andis shown in Figure 1.Generally speaking, the spectra we use in this paperwere obtained as part of GRB follow-up campaigns, andtherefore are a heterogeneous sample in depth, wavelengthrange and resolution. For each spectrum, we used a bright emission line in the datato fix the redshift as accurately as possible: this is generallyPaschen α , Brackett γ or He i λ lines). The lirisdr is supported by J. Acosta. IRAF is distributed by National Optical Astronomy Observa-tories, operated by the Association of Universities for Research inAstronomy, Inc., under contract with the National Science Foun-dation.MNRAS , 1–8 (2018)
K. Wiersema et al.
Table 1.
Observations used in this paper. ∗ : Note that the spectroscopic observations discussed in this paper concern the brighteststar forming region in this host galaxy, known as source A (Starling et al. 2011). The magnitudes given in this table are all integratedmagnitudes for the whole host galaxy. References for spectroscopy data: [1] Watson et al. (2011), [2] Wiersema (2011), [3] This work, [4]Starling et al. (2011). References for the host infrared magnitudes: [5] Prochaska et al. (2004), [6] Hjorth et al. (2012), [7] This work, [8]Micha lowski et al. (2015). References for abundance: [9] Guseva et al. (2011), [10] Wiersema et al. (2007); [11] Stanway et al. (2015a),[12] Starling et al. (2011).GRB host Instrument Obs date Redshift Host IR (Vega) magnitude + log ( O / H ) K ′ = . ± . [5] 8.20 [9]060218 VLT ISAAC 17 July + 10 Sep. 2008 [2] 0.033 K s = . ± . [6] 7.54 [10]080517 WHT LIRIS 3/4 March 2015 [3] 0.089 K s = . ± . [7] ∼ ∗ VLT X-Shooter 17 March 2010 [4] 0.059 K s = . ± . [8] 8.23 [12] Figure 2.
This plot shows postage stamp cut-outs of the spectraof the four host galaxies in this paper. Each spectrum covers therange − to + km/s around the position of four lines ofinterest, which are indicated with a vertical red dashed line. Theflux axis is identical for all lines and sources, with the exceptionof the Paschen α line panels. The different continuum brightnessbetween sources is readily apparent, as well as the varying con-tinuum noise as lines fall in or out of regions with strong telluricabsorption. A zoom-in of the top-right panel, with the only de-tection of a ro-vibrational H line (1–0 S(3)) in the sample, canbe seen in Figure 3. reason for doing this, is that in some cases small velocity off-sets may be present in the wavelength calibration comparedto redshift values in literature. In the mid-resolution (e.g. X-shooter) data of some hosts, in particular GRB 031203, theemission line profiles deviate somewhat from a Gaussian pro-file (previously noted by Guseva et al. 2011) for the brighterlines, a common property for GRB host galaxy spectra (seee.g. Wiersema et al. 2007). This can generally be attributedto kinematics within the host (e.g. Arabsalmani et al. 2018).In the flux measurements (and determination of upper lim-its) of the H lines in the X-shooter spectra, we make the Figure 3.
X-Shooter spectrum of the host galaxy of GRB 031203,around the location of the H µ mis the H i δ ) line, the one near 2.16 µ m is a He I line. assumption that the H line profile is the same as that ofthe Paschen and/or Balmer lines.After the redshift was found accurately, we used acatalogue of H ro-vibrational transition wavelengths (e.g.Black & van Dishoeck 1987) to search for emission lines. Incases of non-detection, we computed 3 σ upper limits on theline flux, using the emission line profile properties of Paschen α and/or Brackett γ . We note that the absolute flux calibra-tion of the spectra has a considerable uncertainty: the slitmay not encompass the entire extent of the galaxy, somedata were taken during somewhat non-spectrophotometricconditions, and telluric absorption often complicates fluxcalibration uncertainties. Line flux ratios are more reliable,especially of lines close together in wavelength (and similarlyaffected by telluric absorption), and we therefore primarilyexpress our limits and detections as a ratio, using the brightPa α line which is detected at high signal-to-noise in all foursources.Figure 2 shows some postage-stamp cutouts of the spec-tra at the positions of a selection of useful lines. It is clearfrom these that the signal-to-noise for a given source canvary a lot with wavelength: the different redshifts of the hostgalaxies shift lines nearer or further from telluric absorptionand emission features (the influence of these is stronger forthe low resolution spectra), or further into, or out of, wave-length regions where the spectrograph sensitivity is poorer(e.g. in X-shooter data the thermal noise in the K band is MNRAS000
X-Shooter spectrum of the host galaxy of GRB 031203,around the location of the H µ mis the H i δ ) line, the one near 2.16 µ m is a He I line. assumption that the H line profile is the same as that ofthe Paschen and/or Balmer lines.After the redshift was found accurately, we used acatalogue of H ro-vibrational transition wavelengths (e.g.Black & van Dishoeck 1987) to search for emission lines. Incases of non-detection, we computed 3 σ upper limits on theline flux, using the emission line profile properties of Paschen α and/or Brackett γ . We note that the absolute flux calibra-tion of the spectra has a considerable uncertainty: the slitmay not encompass the entire extent of the galaxy, somedata were taken during somewhat non-spectrophotometricconditions, and telluric absorption often complicates fluxcalibration uncertainties. Line flux ratios are more reliable,especially of lines close together in wavelength (and similarlyaffected by telluric absorption), and we therefore primarilyexpress our limits and detections as a ratio, using the brightPa α line which is detected at high signal-to-noise in all foursources.Figure 2 shows some postage-stamp cutouts of the spec-tra at the positions of a selection of useful lines. It is clearfrom these that the signal-to-noise for a given source canvary a lot with wavelength: the different redshifts of the hostgalaxies shift lines nearer or further from telluric absorptionand emission features (the influence of these is stronger forthe low resolution spectra), or further into, or out of, wave-length regions where the spectrograph sensitivity is poorer(e.g. in X-shooter data the thermal noise in the K band is MNRAS000 , 1–8 (2018) nfrared molecular hydrogen lines in GRB hosts high, Vernet et al. 2011). Generally speaking, we limit oursearch to a dozen lines bluewards of ∼ . µ m, the strong Qbranch lines at ∼ . µ m and the 2–1 S(1) line (useful diag-nostic for collisional excitation, Black & van Dishoeck 1987,see section 5) are unfortunately not in reach for these sourceswith the spectrographs we used.While long GRBs are (generally) accompanied by highlyenergetic type Ib/c supernovae (though not always detected,Fynbo et al. 2006b), the contribution of emission of suchSNe to our spectra is negligible: our spectra are either takenlong after the SN has faded away (Table 1), or, in thecase of 100316D, cover a region away from the GRB/SN(Starling et al. 2011).All spectra contain a high significance detection ofPaschen α . For three out of four hosts we do not detectany emission lines with significance > σ at the positions ofthe H transitions. In the Xshooter spectrum of the host ofGRB 031203, we detect a single H transition, 1–0 S(3), withflux significance of ∼ σ (Figure 3), and with the expectedcentre wavelength and line shape. We estimate its flux at ∼ . ± . × − erg s − cm − , with the caveats applyingto the line fluxes as listed above.The results of the flux ratio measurements for the linesthat are expected to be strongest (see below) are given inTable 2. The host of GRB 031203 is one of the very few hostgalaxies with good quality (mid-)infrared spectra taken by
Spitzer , using the Infrared Spectrograph (IRS) instrument(Watson et al. 2011). These spectra (shown in Figure 1 ofWatson et al. 2011) cover a range from ∼ − µ m in therest frame, and therefore cover strong pure rotational tran-sitions of H , in particular the S(0) to S(7) transitions. Mo-tivated by the possible detection of a ro-vibrational H line,we reanalyze these Spitzer spectra here.
Considering the diverse range of heating environments in thegalaxy interstellar medium (ISM), the traditional methodof fitting two or three discrete temperature molecule com-ponents to the excitation diagram is not realistic. Insteadwe assumed a continuous power law temperature distribu-tion for H to fit the excitation diagram and hence calcu-lated the total H gas mass in the ISM (for details referTogi & Smith 2016). We assumed that the column densityof H molecules is distributed as a power law function withrespect to temperature, dN ∝ T − n dT , where dN is the num-ber of molecules in the temperature range T to T + dT . Themodel consists of three parameters, upper and lower tem-perature with power law index, denoted by T u , T l , and n ,respectively.We attempted to fit the mid-infrared (MIR) spec-trum for the host of GRB 031203 using PAHFIT , a spec-tral decomposition tool to estimate the H line fluxes(Smith et al. 2007). No convincing H line flux was detected,except a tentative detection of the 0–0 S(7) line, with a fluxof ∼ . × − erg s − cm − . A section of the spectrum and the fitted components, including the 0–0 S(7) transi-tion, is shown in Figure 4. As is clear from Figure 4, the0–0 S(7) covers only a few datapoints with low flux errors,and we consider its detection tentative for that reason. The0–0 S(6) transition (the magenta component at 6.11 µ m inFig. 4) is very close to a strong polycyclic aromatic hydro-carbon (PAH) feature at 6.22 µ m, and the uncertainties inthe datapoints are larger than for 0–0 S(7). We thereforedon’t consider the 0–0 S(6) line reliably detected, and willin the following only consider the tentative detection of the0–0 S(7) line.The power law index could not be determined due tonon detection of any other rotational H lines. The aver-age power law index n is about 4.84 ± ∼ n = . , T l = 50 K, T u = 2000 K, with the S(7) line flux estimateabove, the calculated H gas mass is 1.7 × M ⊙ in the hostof GRB 031203 (2.3 × M ⊙ , including He and heavy ele-ment mass corrections). The variation in n from 4.0–5.0 canchange the gas mass in the range (0.37–7.5) × M ⊙ (and(0.5–10) × M ⊙ with He and heavy element correction).The upper dust mass limit estimated is 10 M ⊙ in this host(Micha lowski et al. 2015). Using a gas-to-dust ratio of 500,using the R´emy-Ruyer et al. (2014) scaling relation for themetallicity value of 8.2 for GRB 031203, the limit on the gasmass is < . × M ⊙ , consistent with our estimates. Ourstudy suggests the presence of a few × M ⊙ molecular gasin this host galaxy, but we caution that higher spectral reso-lution data, at higher signal-to-noise, is required to establishthe veracity and accurate flux of the S(7) line candidate. Wecan compare the H mass estimate with other GRB hosts,using Figure 3 of Stanway et al. (2015b), and using the starformation rate and stellar mass from the literature. Usingthe values from Savaglio, Glazebrook & Le Borgne (2009), log ( M ∗ / M ⊙ ) ∼ . , SFR ∼ . ⊙ / yr , we find that a H massaround or below M ⊙ is consistent with the behaviourof other GRB hosts in this diagram. However, we note thatestimates of stellar mass (and star formation rate) for thishost show a large scatter, ranging from log ( M ∗ / M ⊙ ) ∼ . to ∼ . (Guseva et al. 2011 and Micha lowski et al. 2015, re-spectively). The spectra of GRB host galaxies are dominated by strongemission lines from H ii regions, whereas H lines origi-nate in colder neutral molecular clouds. The H in ourhosts can be excited by two main processes: through col-lisional processes (e.g. involving shocks from stellar windsand supernovae, rational levels are populated by collisionsof H with neutral atoms or other H molecules); or throughfluorescence. In this latter process, the H molecules ab-sorb UV photons (the Lyman-Werner bands), and then de-cay into ro-vibrational states. Observations of extragalac-tic sources have shown both processes may occur (e.g.Izotov & Thuan 2016; Pak et al. 2004). The two processescan be differentiated by comparing fluxes of different H transitions with models (e.g. Black & van Dishoeck 1987). MNRAS , 1–8 (2018)
K. Wiersema et al.
Figure 4.
A small section of the
Spitzer spectrum with the PAH-FIT fit is shown (note that the model is fit over the full
Spitzer range, see Section 4.1). We use the standard PAHFIT conventionfor the components (see Smith et al. 2007): the black-body fitto the (dust) continuum (red), the stellar continuum (magenta),the combined continuum (thick grey), the PAH features (blue),the narrow atomic and molecular features (narrow magenta com-ponents), the composite fit (green) and the datapoints (black).The location of the 0-0 S(7) transition is indicated with a verti-cal arrow, the other magenta peak is the 0-0 S(6) transition (seeSection 4.1 for a discussion).
In particular, UV fluorescence should give rise to brighterlines from transitions of higher ν states.Recently, a large NIR spectroscopic survey of bluecompact dwarf galaxies and compact H ii regions in nearbygalaxies (Izotov & Thuan 2016; Izotov & Thuan 2011)showed that the majority of detected H ro-vibrationalline fluxes in that galaxy sample can be well explainedby fluorescence. The similarity between the sources inIzotov & Thuan 2016 and the physical properties of longGRB hosts in general (e.g. in metallicity, star formation rate,stellar age), suggests that fluorescence is likely to play a leadrole in GRB host emission too. Comparison of the measuredflux of 1–0 S(3) in the host of 031203 with the best limits onother ro-vibrational H lines in the same source, shows broadconsistency with expectations from fluorescence: we choosethe same models from Black & van Dishoeck (1987) asfavoured by Izotov & Thuan (2016), and use the measuredH flux ratios from Izotov & Thuan (2016) to compensatefor the absence of 1–0 S(3) from the prediction tables ofBlack & van Dishoeck (1987). In the host of GRB 031203,the ratio 1–0 S(1) / Br γ . . is comparable to thevalues found in the samples of Izotov & Thuan (2011);Izotov & Thuan (2016); Vanzi, Hunt & Thuan (2002).The host galaxy of GRB 080517 is the only onein our sample with a detection of a CO emission line(Stanway et al. 2015b). It is also a somewhat unusual sourcefor its bright NIR continuum, caused by a combination ofmass, redshift and the presence of a relatively bright olderstellar population in addition to the ongoing star formation,see Stanway et al. (2015a). While our WHT spectroscopy istoo shallow to give useful limits on the CO/H ratio, thissource will be a key target for higher resolution IR spectro-graphs on bigger telescopes.GRBs 060218 and 100316D gave rise to two of the beststudied GRB associated supernovae, and are therefore im-portant as keystones for GRB-host studies, as the super- novae may provide a direct link to stellar GRB progenitorproperties. In both cases the sources are faint (Table 1; notethat for 100316D the full integrated magnitude is given, notthe one of only source A), with very weak continuum andvery strong nebular lines. The host of 060218 has a very lowmass and metallicity (Table 1) and a high specific star for-mation, making it unlikely to host strong H lines, as pointedout in a first search by Wiersema (2011). The better reso-lution of X-shooter would allow a more sensitive limit onthe H fluxes to be set in the future, as the redshift of thissource places most lines in regions affected by strong telluricabsorption.In the coming era of highly sensitive instruments atNIR and MIR wavelengths, such as the Near Infrared Spec-trograph (NIRSpec) and Mid Infrared Instrument (MIRI)instruments onboard the James Webb Space Telescope(JWST), covering wavelengths 0.6–28 µ m, the ro-vibrationallines of H can easily be detected (the 1–0 S(1) line will bein the MIRI spectral range for the entire known GRB red-shift distribution), and the molecular gas properties in theGRB host galaxies can be studied using multiple transitions.Telluric absorption lines strongly limit the use of ground-based low resolution spectroscopy: the change to space-basedJWST data will allow access to a larger number of transi-tions in each spectrum. Using the flux estimate of the 1–0S(3) line in the host of GRB 031203, we find that NIRSpeccan provide a highly significant, spatially resolved, detectionof this line in this host using exposure times under an hour,as well as several other ro-vibrational transitions. The sameholds for MIRI spectroscopy targetting the rotational linesin low redshift GRB hosts. Detecting ro-vibrational linesin GRB host galaxies that showed Lyman-Werner lines intheir afterglow spectra, would require the presence of sub-stantially larger reservoirs of warm H than seen in the hostof GRB 031203. Motivated by the recent detections of CO molecule emissionin GRB host galaxies, we searched rest-frame infrared spec-tra of a sample of four low redshift GRB host galaxies forsignatures of H ro-vibrational emission lines. A single ro-vibrational H emission line candidate is detected at the po-sition of the 1–0 S(3) transition in the host of GRB 031203.The other GRB host spectra in our sample show no signifi-cant H line candidates, which is likely caused by signal-to-noise and resolution limitations, as well as the positions ofthe lines near telluric absorption features. We re-analysedlow resolution Spitzer mid-infrared spectra of the host ofGRB 031203 to search for H rotational lines. A single weakline candidate, at the position of the 0–0 S(7) transition,is seen, but the reality of this line is debatable, because ofthe low resolution of the Spitzer spectra. Observations withfuture facilities with better resolution and higher sensitiv-ity, particularly from space, will provide the means to de-tect the multiple lines required for proper comparison withlow-redshift galaxy samples and high redshift molecule de-tections in afterglow spectra.
MNRAS , 1–8 (2018) nfrared molecular hydrogen lines in GRB hosts Table 2.
Listed are the flux ratios of the molecular line fluxes (or limits) divided by the Paschen α line flux. Limits are σ . A dashindicates that a line measurement was not possible due to very strong telluric absorption, presence of residuals from sky emission lines,very poor signal-to-noise, or because the wavelength is not covered by the spectrum. ∗ : Note that the spectroscopic observations discussedin this paper concern the brightest star forming region in this host galaxy, known as source A (Starling et al. 2011).GRB host 3–1 S(1) 3–1 Q(1) 1–0 S(3) 1–0 S(2) 1–0 S(1) 1–0 S(0)(1.23 µ m) (1.31 µ m) (1.96 µ m) (2.03 µ m) (2.12 µ m) (2.22 µ m)031203 - - ( . ± . ) × − < . × − < . × − < . × − < . < . < . < . < . < . < . -100316D ∗ < . < . < . < . < . < . ACKNOWLEDGEMENTS
It is a pleasure to thank the staff at ING for their helpin obtaining the LIRIS observations in this paper, andJose Acosta Pulido for development of (and friendly as-sistance with) the lirisdr software. We thank the anony-mous referee for their constructive feedback. We acknowl-edge M. Micha lowski and M. Arabsalmani for useful dis-cussions. Based on observations collected at the EuropeanOrganisation for Astronomical Research in the SouthernHemisphere under ESO programmes 60.A-9022(C), 381.D-0723(C) and 084.A-0260(B). The WHT and its override pro-gramme (W/2015A/11 for observations in this paper) areoperated on the island of La Palma by the Isaac NewtonGroup in the Spanish Observatorio del Roque de los Mucha-chos of the Instituto de Astrof´ısica de Canarias. AT is grate-ful for support from the National Science Foundation undergrant no. 1616828. KW, NRT and RLCS acknowledge fund-ing from STFC. The Cosmic Dawn center is funded by theDNRF.
REFERENCES
Acosta-Pulido J., Dominguez-Tagle C., Manchado A., 2003, inIye M., Moorwood A. F. M., eds, Proc. SPIE Conf. Ser.Vol. 4841, Instrument Design and Performance for Opti-cal/Infrared Ground-based Telescopes. SPIE, Bellingham, p.437Appleton, P. N., Guillard, P., Togi, A., et al. 2017, ApJ, 836, 76Arabsalmani, M. et al. 2018, MNRAS 473, 3312Arabsalmani, M. et al. 2018b, MNRAS 476, 2332Bertin, E. & Arnouts, S. 1996, A&AS, 177, 393Black, J. H. & van Dishoeck, E. F. 1987, ApJ 322, 412Bolatto, A. D., Wolfire, M. & Leroy, A. K. 2013, ARA&A 51, 207Bragg, S. L., Smith, W. H., Brault, J. W. 1982 ApJ 263 999Covino, S. et al. 2013, MNRAS 432, 1231Cushing, M.C., Vacca, M.D., & Rayner, J.T. 2004, PASP 116,362D’Elia V. et al., 2014, A&A, 564, 38Draine, B. T. 2000, ApJ, 532, 273Draine, B. T. & Hao, L. 2002, ApJ, 569, 780Friis, M. 2015, MNRAS 451, 167Fynbo, J. P. U. et al. 2006, A&A, 451, L47Fynbo, J. P. U. et al. 2006b, Nature 444, 1047Fynbo, J. P. U. et al. 2008, ApJ, 683, 321Fynbo, J. P. U. et al. 2009, ApJS 185, 526Fynbo, J. P. U. et al. 2014, A&A 572, 12Greiner, J. et al. 2015, ApJ 809, 76Guillard, P., Boulanger, F., Lehnert, M. D., Appleton, P. N. &Pineau des Forˆets, G. 2015, SF2A-2015: Proceedings of the Annual meeting of the French Society of Astronomy and As-trophysics: Warm molecular Hydrogen at high redshift withthe James Webb Space Telescope, eds. Martins F., BoissierS., Buat V., Cambr´esy L., Petit P., pp81–85Guseva, N. G., Izotov, Y. I., Fricke, K. J., Henkel, C. 2011, A&A534, 84Hatsukade, B., Ohta, K., Endo, A., et al. 2014, Nature, 510, 247Heintz, K. E., et al. 2017, A&A, 601, A83Hjorth, J., et al. 2012, ApJ 756, 187Izotov, Y. I. & Thuan, T. X. 2011, ApJ, 734, 82Izotov, Y. I. & Thuan, T. X. 2016, MNRAS 457, 64Jakobsson, P. et al. 2006, A&A Letters, 460, L13Kr¨uhler T. et al. 2013, A&A, 557, 18Kr¨uhler T. et al. 2015, A&A 581, 125Ledoux, C. et al. 2009, A&A 506, 661Micha lowski, M. J. et al. 2015, A&A 582, 78Micha lowski M. J., et al., 2016, A&A, 595, A72Micha lowski M. J., et al., 2018, arXiv:1804.06492Noterdaeme, P., Ledoux, C., Petitjean, P., Srianand, R. 2008,A&A 481, 327Noterdaeme, P., Petitjean, P., Srianand, R. 2015, A&A 578, L5Pak, S., Jaffe, D. T., Stacey, G. J., Bradford, C. M., Klumpe, E.W., Keller, L. D. 2004, ApJ 609, 692Perley, D. A. et al. 2016, ApJ 817, 7Perley, D. A. et al. 2016, ApJ 817, 8Perley, D. A. et al. 2017, MNRAS 465, L89Prochaska, J.X. et al. 2004, ApJ 611, 200Prochaska, J. X., Chen, H.-W., Wolfe, A. M., Dessauges-Zavadsky, M., Bloom, J. S.Prochaska, J. X. et al., 2009, ApJ, 691, L27R´emy-Ruyer, A., Madden, S. C., Galliano, F., et al. 2014, A&A,563, A31Savaglio, S., Glazebrook, K. & Le Borgne, D. 2009, ApJ 691, 182Schady, P. 2015, JHEAp 7, 56Selsing, J. et al. 2018, arXiv:1802.07727Sheffer, Y., Prochaska, J. X., Draine, B. T., Perley, D. A., &Bloom, J. S. 2009, ApJ, 701, L63Smith, J. D. T., Draine, B. T., Dale, D. A., et al. 2007, ApJ, 656,770Stanway, E. R., Levan, A. J., Tanvir, N. R., Wiersema, K., vander Horst, A. J., Mundell, C. G., Guidorzi, C. 2015, MNRAS446, 3911Stanway, E. R., Levan, A. J., Tanvir, N. R., Wiersema, K., vander Laan, T. P. R. 2015, ApJ 798, 7Starling, R. L. C. et al. 2011, MNRAS 411, 2792Symeonidis, M. et al. 2014, MNRAS 443, 124Togi, A., & Smith, J. D. T. 2016, ApJ, 830, 18Tumlinson J., Prochaska J. X., Chen H. W., Dessauges-ZavadskyM., Bloom J. S., 2007, ApJ, 668, 667Vanzi, L., Hunt, L. K. & Thuan, T. X. 2002, A&A 390, 481Vergani, S. D. et al. 2017, A&A 599, 120Vernet, J. et al. 2011, A&A, 536, 105Vreeswijk, P. M. et al. 2007, A&A 468, 83MNRAS , 1–8 (2018)
K. Wiersema et al.
Watson, D., et al. 2011, ApJ 741, 58Wiersema, K., et al. 2007, A&A, 464, 529Wiersema, K. 2011, MNRAS 414, 2793This paper has been typeset from a TEX/L A TEX file prepared bythe author. MNRAS000