Astronomical Science DOI: 10.18727/0722-6691/5066

The ESO Diffuse Interstellar Band Large Exploration ­Survey (EDIBLES)

Jan Cami1, 2 11 Department of Physics, The Catholic interstellar “mystery molecules”, and turn Nick L. J. Cox 3, 4 University of America, Washington DC, DIBs into powerful diagnostics of their Amin Farhang1, 5 USA environments in our Milky Way Galaxy Jonathan Smoker 6 12 Instituut voor Sterrenkunde, KU Leuven, and beyond. We present some prelimi- Meriem Elyajouri 7 Belgium nary results showing the unique capa- Rosine Lallement 7 13 George Washington University, bilities of the EDIBLES programme. Xavier Bacalla 8 ­Washington DC, USA Neil H. Bhatt1 14 UK Astronomy Technology Centre, Emeric Bron 9 Royal Observatory Edinburgh, UK The diffuse interstellar bands Martin A. Cordiner10, 11 15 ESTEC, ESA, Noordwijk, The Alex de Koter 3, 12 Netherlands One of the longest-standing problems in Pascale Ehrenfreund13 16 Université de Toulouse, UPS-OMP, modern astronomical spectroscopy is Chris Evans14 IRAP, Toulouse, France associated with the identification of the Bernard H. Foing15 17 CNRS, IRAP, Toulouse, France chemical species that produce the dif- Atefeh Javadi5 18 Sorbonne Université, Observatoire de fuse interstellar bands (DIBs; see Cami Christine Joblin16, 17 Paris, Université PSL, CNRS, LERMA, & Cox, 2012 for a recent review) — a Lex Kaper 3 Meudon, France problem that first surfaced almost a cen- Habib G. Khosroshahi 5 19 School of Chemistry, University of tury ago. The DIBs are a collection of Mike Laverick12 ­Nottingham, UK over 400 absorption features that appear Franck Le Petit18 20 Instituto de Astrofísica de Canarias in the spectra of reddened stars. Their Harold Linnartz 8 (IAC), La Laguna, Tenerife, Spain interstellar nature is clear, but their origin Charlotte C. M. Marshall19 21 Universidad de La Laguna, Tenerife, is unknown (although a few DIBs in the Ana Monreal-Ibero 20, 21 Spain near-infrared part of the spectrum are 22 22 + Giacomo Mulas INAF–Osservatorio Astronomico di attributed to C60; see below). Given their Evelyne Roueff18 Cagliari, Selargius, Italy strength and their widespread occur- Pierre Royer12 23 NASA Ames Research Center, Space rence in harsh interstellar environments, Farid Salama 23 Science & Astrobiology Division, the DIB carriers are most likely abundant, Peter J. Sarre19 ­Mountain View, USA stable, carbonaceous species such as Keith T. Smith 24 24 AAAS Science International, carbon chains, polycyclic aromatic hydro- Marco Spaans 25 ­Cambridge, UK carbons (PAHs), fullerenes or closely Jacco T. van Loon 26 25 Kapteyn Institute, University of related species. Despite their unknown Gregg Wade 27 ­Groningen, The Netherlands identity, the DIBs are being used increas- 26 Lennard-Jones Laboratories, Keele ingly often as tools, for example to map University, UK the interstellar medium in 3D (Bailey et al., 1 Department of Physics and Astronomy 27 Department of Physics, Royal Military 2016). The eventual identification of DIB and Centre for Planetary Science and College of Canada, Kingston, Canada carrier(s) will make DIBs very powerful Exploration (CPSX), The University of diagnostics in the interstellar medium. Western Ontario, London, Canada 2 SETI Institute, Mountain View, USA The ESO Diffuse Interstellar Band Large The definite identification of DIB carriers 3 Anton Pannekoek Institute for Astron- Exploration Survey (EDIBLES) is a Large must come from an accurate match omy, University of Amsterdam, The Programme that is collecting high-signal- between the observed spectroscopic Netherlands to-noise (S/N) spectra with UVES of a features in a low-temperature gas-phase 4 ACRI-ST, Sophia Antipolis, France large sample of O and B-type stars cov- laboratory experiment and the DIBs seen 5 School of Astronomy, Institute for ering a large spectral range. The goal in astronomical observations. Indeed, Research in Fundamental Sciences, of the programme is to extract a unique laboratory data not only provide accurate Tehran, Iran sample of high-quality interstellar spec- rest wavelengths and bandwidths (includ- 6 ESO tra from these data, representing differ- ing transitions beyond the origin band) of 7 GEPI, Observatoire de Paris, PSL ent physical and chemical environments, possible carriers, but the controlled con- Research University, CNRS, Université and to characterise these environments ditions in the laboratory also allow the Paris-Diderot, Sorbonne Paris Cité, in great detail. An important component derivation of oscillator strengths that are Meudon, France of interstellar spectra is the diffuse inter- needed to estimate column densities. 8 Sackler Laboratory for Astrophysics, stellar bands (DIBs), a set of hundreds However, the sheer number of possible Leiden Observatory, Leiden University, of unidentified interstellar absorption DIB carrier candidates to measure exper- The Netherlands lines. With the detailed line-of-sight imentally is so challenging — for exam- 9 ICCM, Madrid, Spain information and the high-quality spectra, ple, there are more than 1.2 million PAH 10 Astrochemistry Laboratory, NASA EDIBLES will derive strong constraints species with 100 or fewer C atoms — ­Goddard Space Flight Center, on the potential DIB carrier molecules. that targeted astronomical observations ­Greenbelt, USA EDIBLES will thus guide the laboratory are needed to guide the selection of the experiments necessary to identify these most promising candidates by providing

The Messenger 171 – March 2018 31 Astronomical Science Cami J. et al., EDIBLES

564 nm RedU HD 79186, 80 Flats S/N = 1400 1.02 1.02 EDIBLES S/N ~ 1000

x 1.00 1 flu

x flu

0.98 0.98 S/N ~ 555 Normalised 0.96 HD 79186, 5 Flats Normalised S/N = 880 0.96 ADP 0.94

0.94 6120 6125 6130 5794 5796 5798 5800 Wavelength (Å) Wavelength (Å) constraints on the carrier species. signal-to-noise ratio (median S/N ~ 500– Figure 1. Left: A comparison between the results High-resolution observations of DIB 1000 per target) over a large spectral of the ESO Archive Data Products (ADP) and the EDIBLES processing pipeline (bin size: 0.02 Å). line profiles can be used to estimate the range (3050–10420 Å), and with targets ­Figure from Cox et al. (2017), reproduced with per- size and geometry of the DIB carrier that represent very different physical mission from A&A. Right: The EDIBLES spectrum ­molecules as well as their excitation ­conditions along the lines of sight. DIB of HD 79186 around the 5797 Å DIB when using properties (see, for example, Marshall targets are typically bright, early-type (O- 80 flat fields (black) compared to the same obser­ vation using only five flat fields (red). et al., 2015). Correlation studies and and B-type) stars whose optical spectra investigations of how DIB strengths contain relatively few stellar lines. In these change in different environments yield spectra, the DIBs are more easily recog- A high S/N also requires a large number information about what drives variations nised and characterised. We selected of flat field exposures. We have devel- in the DIB properties, such as the ionisa- bright (V < 8 magnitudes) O- and B-type oped a custom data reduction procedure tion potential and chemical make up of stars, and constructed our sample so that to process flat field frames so that we can the DIB carriers (see, for example, Ensor we can probe a wide a range of inter­ attain the highest possible S/N (see et al., 2017). stellar environment parameters including Cox et al., 2017). Figure 1 illustrates how interstellar reddening E(B-V) ~ 0–2 mag- the addition of a large number of flat

While there has been steady progress nitudes, visual extinction AV ~ 0–4.5 mag- fields can greatly improve the quality of in the field over the years, most studies nitudes, total-to-selective extinction the resulting spectra. Using the same focus on the properties of a small num- ratio RV ~ 2–6, and a molecular hydrogen recipes and a higher number of flat field ber of DIBs, over particular wavelength fraction f(H2) range ~ 0.0–0.8. Our final frames — even when obtained on differ- ranges and in particular environments, target list contains 114 objects of which ent dates — can increase the S/N of or alternatively deal with large datasets 97 have been observed to date. Further other good-quality science observations and “average” properties of the DIBs details about the goals, objectives and in the UVES archive. In the red part of the (for example, Lan et al., 2015). Significant sample selection can be found in Cox spectrum, we could also improve and progress in the field can be expected et al. (2017). fine-tune the wavelength calibration using from a high-quality, sensitive survey of the large number of telluric lines available interstellar features (DIBs, but importantly EDIBLES is an approved Large Pro- in our high-resolution spectra. also known interstellar atoms and mole- gramme that started in September 2014 cules) over a large spectral range and (ESO period 94) under Programme ID EDIBLES is unique in its combination representing differing interstellar environ- 194.C-0883. The total allocated tele- of spectral resolution, wavelength cover- ments (Cami & Cox, 2014). The ESO scope time, excluding daytime calibra- age, sensitivity and sample size and this ­Diffuse Interstellar Band Large Explora- tions, is 284 hours. As a programme, promises great advances in the field. tion Survey (EDIBLES) is such a survey. EDIBLES has been optimised, in terms of As illustrated below, the high spectral selected targets and observing strategy, resolution allows us to study individual to obtain observations when weather DIB line profiles for a large number of EDIBLES conditions are typically too poor for regu- DIBs, yielding size estimates for their lar programmes (called “filler conditions”) ­carrier molecules when clear substruc- The aim of EDIBLES is to collect a large and thus helps to optimise the use of the tures are present. The high sensitivity sample of interstellar spectra with UVES telescope. This makes it a less efficient facilitates studying both strong and weak (Smoker et al., 2009) at high spectral process to reach a sufficiently high S/N, DIBs simultaneously. This is important ­resolution (R ~ 70 000 in the blue and and we therefore require a large number since laboratory spectra of typical DIB 100 000 in the red arm), with a very high of exposures. carriers often result in several transitions

32 The Messenger 171 – March 2018 of comparable strength in the optical Figure 2. Upper panel: range, in contrast with the fact that no ­Profiles of the 6614 Å DIB 1 towards HD 170740 (deeper, two DIBs have been found that undenia- blue trace) and HD 147165 bly originate from the same species (with (broader, red trace). The + spectra are normalised to a the possible exception of the C60 DIBs). We expect that EDIBLES will reveal the 0.96 common integrated inten- sity and the ordinate values first pair of perfectly correlating DIBs. are those for HD 170740. The spectra are shifted to The large spectral coverage combined 0.92 the interstellar restframe x using K I radial velocities.

with the high resolution — which allows flu Lower panel: The difference very good telluric corrections — will ve between the normalised explore wavelength ranges over which profiles of the DIB in few or even no sensitive DIB searches Relati 0.88 HD 170740 and HD 147165. have ever been carried out. Furthermore, this range includes a large number of known interstellar absorption lines — due −0.02 to atoms and small molecules such as

CN or C2 — which play an important role 0 in EDIBLES. They enable us to accurately define the physical conditions in a large 0.02 sample of lines of sight along which DIBs are measured, which is unprecedented 6612.0 6613.0 6614.0 6615.0 6616.0 on this scale. All these characteristics will Wavelength (air, Å) enormously constrain the number of pos- sible carrier molecules. interpreted as arising from a stronger hot ear molecule (Thorburn et al., 2003). Fur- band contribution (Marshall et al., 2015). ther constraints on the properties of the EDIBLES is also ideally suited for seren- carrier will come from detailed analyses dipitous studies that are not directly The details of the band profile depend of the line profile shape of the 2C -DIBs. related to the DIBs. The large number of on the rotational and vibrational tempera- Since many of the C2-DIBs are weak and observations of O and B stars will turn tures and EDIBLES is designed to make high spectral resolution is required to out to be very useful for studies of mas- progress in our understanding of pre- resolve the profiles, substructure in the sive stars and their stellar winds. Since cisely how these temperatures cause profiles of the 2C -DIBs was only reliably our survey also includes observations spectral changes. Indeed, the large num- established for three to four C2-DIBs, of the same lines of sight at different bers of EDIBLES sightlines sample the although asymmetries in observed pro- epochs, the survey can also be used and possible range of rotational and vibrational files suggested that more of the 2C -DIBs complemented by archival data, to study temperatures in interstellar clouds, and could exhibit substructure (Galazutdinov small-scale variations in interstellar lines information on these temperatures can et al., 2002). over time. be derived from other molecular absorp-

tion lines in the EDIBLES spectral range. In the EDIBLES dataset, we detect C2 A-X Resolved substructures in DIB line Modelling of these band profiles will thus rovibronic bands around 7700, 8800 and profiles provide the link between the observed 10 100 Å in 25 sightlines, representing The high-resolution, high-S/N spectra spectral variations and the changes in one of the largest samples to date for this of the EDIBLES survey reveal profile temperature; in turn, this will greatly con- important species (Cordiner et al., in ­substructures in a large number of DIBs, strain the properties (for example, size preparation). This is another demonstra- and subtle variations in these profiles and geometry) of the possible carrier tion of the unique, extremely high sensitiv- from one line of sight to another. A good molecules. ity of our EDIBLES survey for the detec- example is the well-known triplet struc- tion of weak spectral features. Excitation ture found in the 6614 Å DIB towards The C2-DIBs modelling of these C2 spectra permits HD 170740 and HD 147165 (σ Sco), shown A particularly intriguing subset of the new insights into the temperatures and in Figure 2. We compare the DIB profiles, DIBs are the so-called C2-DIBs (Thorburn densities of diffuse molecular gas. normalised to a common integrated et al., 2003); these are features that cor- intensity using an approach introduced relate especially well with the column From this sample, we selected several by Marshall et al. (2015). It is notable that density of C2 molecules along the line of EDIBLES targets that exhibit strong the absorption depths for the sub-peaks sight. This could imply that C2-DIB carri- C2 lines and that are “single clouds”, i.e., that are redwards and bluewards of the ers are chemically linked to C2. Moreover, there is only one dominant interstellar strongest central absorption differ for several C2-DIBs appear in pairs that are cloud along the direct line of sight to the the two sightlines, and that the redward separated by the same spacing of about target. For this sample of targets, we –1 tail is stronger and more extended for 20 cm , reminiscent of the spectroscopic find thatall C2-DIBs have resolved struc- HD 147165; this character­istic has been signature of spin-orbit interaction in a lin- ture in their profiles (Elyajouri et al., in

The Messenger 171 – March 2018 33 Astronomical Science Cami J. et al., EDIBLES

5418 Å 1.00 5512 Å 5769 Å 1.00 1.00 0.99 0.98 0.98 0.98 New 0.96 0.97 0.96

x 1.00 1.00 1.00 flu

0.99 0.98 0.99 Normalised

0.98 0.98 0.96

0.97 0.97

0.94 0.96 0.96 5417 5418 5419 5420 5421 5512 5513 5514 5768 5769 5770 Wavelength (Å) Wavelength (Å) Wavelength (Å)

Figure 3. A few examples of the C2-DIB profiles at son with interstellar spectra. The central uals) and a Gaussian fit to that depres- 5418.91, 5512.62 and 5769.08 Å. Coloured lines dis- wavelengths of the two strongest bands sion has an absorption depth that agrees play the individual sightlines. The black profile on top is the average normalised profile obtained by stack- in the laboratory experiments agree with with predictions, given the strength of the + ing the deepest bands of three single cloud targets. the central wavelengths of the two DIBs, other C60 bands. While this does not yet within uncertainties of ~ 0.1 Å. The labo- prove beyond any doubt that the 9428 Å ratory spectrum furthermore exhibits DIB is present, the EDIBLES data support + preparation; Figure 3). The high resolution three weaker bands and several authors the identification of these DIBs with C60 and S/N allow us to conclude with con­ have found evidence for these bands as (Lallement et al., 2018). fidence that these substructures are nei- well (for example, Walker et al., 2017; ther telluric nor stellar in origin, and that Cordiner et al., 2017). The high spectral resolution of the the single-cloud nature of the targets ­EDIBLES data furthermore allows the + eliminates multi-cloud confusion as well. Unequivocally confirming even the pres- study of some of the C60 DIBs profiles These substructures are thus intrinsic to ence of these weaker bands, however, is despite the telluric contamination. For the C2-DIBs and will reveal a lot about greatly complicated by the many strong example, in the lower panel of Figure 4 the properties of their carrier molecule(s). telluric spectral features in the wavelength we show how co-adding 79 individual + range over which the C60 bands occur observations of 43 different targets reveals + The DIBs attributed to C60 (Figure 4, top). Typically, the analysis is the first evidence for an enormous amount Foing & Ehrenfreund (1994) used labora- performed on a corrected spectrum, cre- of detail in the substructure of the strong + tory measurements to predict that inter- ated by dividing out the telluric lines by 9577 Å C60 DIB. This is possible because + stellar, gas-phase C60 molecules would means of a transmission model or the this DIB is only partially affected by telluric exhibit transitions near 9577 Å and spectrum of a telluric standard star. How- lines; the different interstellar cloud radial 9632 Å. A dedicated search did indeed ever, with such strong telluric lines — also velocities then cause different parts of result in the discovery of two DIBs near variable on short timescales — residuals the band to appear in the “telluric-free” those wavelengths with band characteris- are unavoidable and can result in severe window (Lallement et al., in preparation). tics that were expected for this species. artefacts. Figure 4 (middle) shows that, Small-scale variations in these substruc- However, the laboratory data that they even when dividing out a spectrum with a tures from one line of sight to another used were obtained in cryogenic matri- very similar telluric profile (in this case of can then be used to study the fullerene ces in which the band positions and pro- HD 54662), such residuals occur. molecular physics and the impact of the files may be affected. Obtaining a gas- local physical conditions. phase spectrum at low temperatures In the corrected spectrum, a clear feature has proven to be very challenging, but shows up at 9412 Å; this is a known DIB OH+ and the cosmic ray ionisation rate + Campbell et al. (2015) succeeded in that is unrelated to C60. At the same time, When Bhatt & Cami (2015) performed developing a technique to reliably meas- at the expected position of the 9428 Å a sensitive survey of interstellar lines + ure the electronic spectrum under con­ C60 band, there appears to be a depres- in the near-ultraviolet by stacking 185 ditions that are appropriate for compari- sion (somewhat masked by telluric resid- UVES archival observations of reddened

34 The Messenger 171 – March 2018 Figure 4. Upper: HD 169454 HD 54662 a time series since 1995 of the He I line The ­EDIBLES spectra of 2200 profile at 5876 Å in HD 148937, a mag- HD 169454 and HD 54662 in the range netic Of?p star that is at the centre of the 2000 of two weak DIBs: the bipolar emission nebula NGC 6164. This

x 9412 Å DIB (unrelated object has been monitored for a long Flu to C+ ) and a weak DIB 1800 60 time, and small variations in stellar line at 9428 Å which is due + profiles have been attributed to rotational to C60 (adapted from 1600 Lallement et al., 2018). modulation (Nazé et al., 2008). We noticed Middle: The HD 169454 9410 9420 9430 9440 9450 that the EDIBLES observations that were spectrum divided by carried out in 2015 revealed a very differ- the one of HD 54662. DIB ent line profile compared to all previously 9428 Å Lower: A com­posite 1.5 showing partially published profiles (c.f. Figure 14 in Cox resolved substructure et al., 2017). Many more stellar lines show in the strong 9577 Å 1.4 + clear variations in their shapes and their C60 DIB. All fluxes are in x arbitrary units. wavelengths. A careful analysis reveals

Flu that this object is a massive spectroscopic 1.3 DIB 9412 Å binary, with a preliminary period of at least 18 years. The ­EDIBLES observations 1.2 just happened to be acquired close to periastron passage, where effects on the 1.1 9410 9420 9430 9440 9450 spectrum would be most pronounced Wavelength (Å) (Wade et al., in preparation).

Outlook 1.00 With EDIBLES, we have currently ob­­

N served 97 early-type stars at high spectral umb resolution and at high S/N over a large

0.99 er

70 spectral range. The first studies, some of of of

x

cont which we have presented here, clearly Flu show the enormous potential for new dis- 60 ribu 0.98 coveries with this dataset. To ensure opti- ting ex ting mal exploitation and interpretation of this unique data set, the EDIBLES team is

50 posu composed of researchers from a diverse 0.97

re community, including observational s 40 astronomers as well as experts in molecu- lar astrophysics, interstellar physics and 9572 9574 9576 9578 9580 9582 9584 Wavelength (Å) chemistry, and laboratory astrophysics. In addition to the scientific analyses, we have targets, they discovered five new narrow The high S/N in the EDIBLES data and tested and demonstrated a recipe that is inter­stellar features in the stacked spec- the wavelength coverage in the near-­ available to the community, which can trum that were too weak to be discerned ultraviolet now show these very weak significantly increase the S/N of UVES in a single observation. Zhao et al. (2015) features almost routinely, and enable the observations. The EDIBLES team is fur- confirmed two of these features and calculation of the cosmic ray ionisation thermore committed to pro­viding the com- identified them with transitions due to rate for a much larger sample than before, munity with a catalogue detailing a large OH+; they also found several more OH+ thus expanding the available line-of-sight number of interstellar quantities for these lines in addition to the already well known information potentially also relevant for sightlines — including DIB measure- line at 3583.76 Å. With several lines the DIBs analysis (Bacalla et al., in prepa- ments. This catalogue is likely to serve as ­available that all arise from the ground ration; Figure 5). Indeed, the detection the DIB benchmark reference for many state level, it is possible to accurately and analysis of such diagnostic molecules years to come. While the EDIBLES team derive the corresponding­ population and allow the community to interpret the DIB focuses on the interstellar studies, the infer the cosmic ionisation rate in diffuse measurements within the context of their same data will also be fruitful ground for a inter­stellar clouds, something which physical surroundings. diverse range of other projects, including has only been possible using space tele- stellar spectroscopy analyses. It is clear scopes, and which yields comparable Variability in HD 148937 that ­EDIBLES will leave a significant legacy results. As an example of the potential for seren- to the astronomical community at large. dipitous discoveries, we show in Figure 6

The Messenger 171 – March 2018 35 Astronomical Science Cami J. et al., EDIBLES

635 HD 148937 630

625 x 1995

Flu 620

615

610 2002

1.000 2006 0.995 x flu

flux

0.990

0.985 2007 0.980 Normalised

Normalised 0.975 0.970 2008 0.965 3σ 0.0025 0.0000 2009 –0.0025 –3σ 3582.753583.25 3583.753584.25 3584.75 Wavelength (Å) 2010 Acknowledgements Figure 5 (above). The strongest OH+ line at 3584 Å in the EDIBLES Jan Cami and Amin Farhang acknowledge support spectrum of HD 80558, with a from a Natural Sciences and Engineering Research Voigt profile­ fit. 5860 5880 5900 Council (NSERC) Discovery Grant and a Science and Wavelength (Å) Engineering Research Board (SERB) Accelerator Award from Western University. Meriem Elyajouri acknowledges funding from the Region Île-de- HD 148937 France through the DIM-ACAV project. Peter J. Sarre thanks the Leverhulme Trust for award of a Lever- hulme Emeritus Fellowship. Charlotte C. M. Marshall thanks EPSRC and the University of Nottingham for financial support. OWN archive 26 May 2010

References

Bailey, M. et al. 2016, A&A, 585, A12 OWN archive 24 May 2013 Barba, R. H. et al. 2010, Revista Mexicana de Astronomía y Astrofísica, 38, 30 Bhatt, N. H. & Cami, J. 2015, ApJs, 216, 22 x flu

Cami, J. & Cox, N. L. J., eds. 2014, IAU Symposium, OWN archive 05 Jun 2014 Vol. 297, The Diffuse Interstellar Bands Campbell, E. K. et al. 2015, Nature, 523, 322 Cordiner, M. A. et al. 2017, ApJl, 843, L2

Cox, N. L. J. et al. 2017, A&A, 606, A76 Normalised Ensor, T. et al. 2017, ApJ, 836, 162 EDIBLES Discovery 15 Sep 2015 Foing, B. H. & Ehrenfreund, P. 1994, Nature, 369, 296 Galazutdinov, G. et al. 2002, A&A, 396, 987 Figure 6. Upper Right: Archival Lallement, R. et al. 2018, arxiv:1802.00369 observations showing the line pro- Lan, T.-W., Ménard, B. & Zhu, G. 2015, MNRAS, files of the He I line at 5876 Å in 452, 3629 EDIBLES Follow-up 23 Apr 2017 HD 148937 between 1995 and Marshall, C. C. M., Krełowski, J. & Sarre, P. J. 2015, 2010. Right: Archive observations MNRAS, 453, 3912 from the Southern Galactic O- and Nazé, Y. et al. 2008, AJ, 135, 1946 WN-type stars (OWN) campaign Smoker, J. et al. 2009, The Messenger, 138, 8 (Barba et al., 2010) and EDIBLES Thorburn, J. A. et al. 2003, ApJ, 584, 339 OWN Follow-up 15 Aug 2017 observations of the same line Walker, G. A. H. et al. 2017, ApJ, 843, 56 between 2010 and 2017. Note how Zhao, D. et al. 2015, ApJl, 805, L12 the character of the line profile changed markedly in 2014–2015 (Figure from Wade et al., in prepa- 5860 5880 5900 ration). Wavelength (Å)

36 The Messenger 171 – March 2018