The VLTI's Astronomical Multi-Beam Combiner

Telescopes and Instrumentation DOI: 10.18727/0722-6691/5106

The Life and Times of AMBER: The VLTI’s Astronomical Multi-BEam combineR

Willem-Jan de Wit1 instruments are scientifically inaugurated called the closure phase. The absolute Markus Wittkowski1 via their “first fringes”, which is the inter- phase of incoming light waves is scram- Frederik Rantakyrö 2 ferometric equivalent of “first light”. By bled by atmospheric turbulence, resulting Markus Schöller 1 the early 2000s, the integration of ESO’s in distortion over a pupil and global Antoine Mérand 1 interferometer into the VLT architecture phase shifts between the apertures in the Romain G. Petrov 3 was on track. array (called the piston). The degree and Gerd Weigelt 4 frequency of the scrambling increases Fabien Malbet 5 The first Paranal interference fringes were towards shorter wavelengths. As a result, Fabrizio Massi 6 produced by the VLT INterferometer the coherence time of the incoming Stefan Kraus7 Commissioning Instrument (VINCI) and wave ranges from a few milliseconds to Keiichi Ohnaka 8 MID-infrared Interferometric instrument (at best) some tens of milliseconds in the Florentin Millour 3 (MIDI), instruments that combined the optical regime. There is no way to beat Stéphane Lagarde 3 light from two telescopes. VINCI’s pur- the turbulence and recover the phase Xavier Haubois1 pose was to commission the interferome- without additional aids. When combining Pierre Bourget 1 ter’s infrastructure. MIDI, on the other three telescopes arranged in a closed Isabelle Percheron1 hand, was the first scientific instrument in triangle one can retrieve a new observa- Jean-Philippe Berger 5 operation using the VLTI in conjunction ble by adding the phases. This resulting Andrea Richichi 6 with the 8.2-metre Unit Telescopes (UTs). closure phase is invariant to atmospheric The second scientific VLTI instrument to perturbations, as the atmospheric phase arrive on Paranal was AMBER. It had noise terms from each individual tele- 1 ESO been conceived as a potential sea scope cancel out. The technique was first 2 Gemini Observatory change in optical interferometry, exploit- applied in radio interferometry. Physically, 3 Université Côte d’Azur, France ing the idea of spectro-interferometry — the closure phase quantity is a proxy 4 Max-Planck-Institut für Radioastrono- obtaining spatial information on milliarc- for the degree of asymmetry in the sci- mie, Bonn, Germany second scales at high spectral resolution. ence target. Closure phase information 5 Institut de Planétologie et d’Astrophy- It comprised three spectral settings, is a pre-requisite to reconstructing sique de Grenoble, France including a high spectral resolution of R = images from interferometric observables 6 INAF–Osservatorio Astrofisico di Arcetri, 12 000, and was foreseen to work at a (for example, Jennison, 1958; Baldwin Italy high sensitivity and with high visibility et al., 1996) and AMBER was the first 7 University of Exeter, UK accuracy in three infrared atmospheric instrument at the VLTI to deliver it. 8 Universidad Católica del Norte, Chile windows (J-, H-, and K-bands). Yet, argu- ably its most important asset was the AMBER produced clear first fringes of capacity to combine the light beams from the star θ Centauri on the night of 20 The sharpest images on Paranal are three separate telescopes at long base- March 2004 using two telescopes at a produced by the beam-combining lines, a novelty in long-baseline optical baseline of 64 metres, marking a mile- instruments of the Very Large Tele- interferometry which allowed millarcsec- stone after seven years of work. The scope Interferometer (VLTI). Currently, ond-resolution images to be synthesised instrument was offered to the community the VLTI is close to completing a tran at high spectral resolution. for the first time in observing period 76 sitional period, moving away from the (starting October 2005), fed by the large first generation of instruments (AMBER, The consortium of four institutes driving apertures of the UTs. MIDI) and offering new instruments the AMBER project consisted of the and subsystems to the community. In Observatoire de la Côte d’Azur (OCA: the this article, we report on the life and Principal Investigator institute) in Nice, the Optical principle and early years achievements of the recently decom- Laboratoire d’Astrophysique de l’Observ- missioned, near-infrared beam com- atoire de Grenoble (LAOG at the time, AMBER’s design corresponds broadly to biner instrument AMBER, the most pro- now called IPAG), the Max-Planck-Institut an optical configuration similar to the one lific optical interferometric instrument für Radioastronomie (MPIfR) in Bonn, and that creates fringe patterns in a Young’s to date. the Osservatorio Astrofisico di Arcetri in interference experiment, i.e., overlapping Florence. It built on the European exper- images coming from multiple telescopes tise of designing two-telescope combin- (or beams). Most importantly, before AMBER, a three-telescope combiner ers capable of exploiting spectro-inter the light is recombined, each light beam ferometry and the usage of single-mode is guided through a single-mode fibre. AMBER was one of three ambitious, fibres. Conceptually, to advance from A single-mode fibre acts as a spatial general-user, interferometric instruments two-telescope to three-telescope com- filter and rejects the distorted part of the proposed in 1997 for implementation on biners may seem a small step, but scien- wavefront, leading to a flattened exit the VLTI at Paranal (Paresce et al., 1996), tifically it constituted a leap forward. wavefront. The phase fluctuations are following the recommendations of the traded against fast intensity fluctuations Interferometry Science Advisory Commit- The crucial consideration is to provide (which are recorded) and a global piston tee to ESO. In optical interferometry, new access to the observational quantity (which is measured from the slope of

8 The Messenger 174 – December 2018 the dispersed fringes). Hence, AMBER AMBER consortium in 2007–2008, and (April 2009) standard AMBER operations implements three photometric channels through the continuous improvement of for medium and high spectral resolution for the simultaneous monitoring of the the VLTI infrastructure. were done in conjunction and simultane- beam intensity for each telescope beam. ously with FINITO. Since 2011, the FINITO Recombination, and with it the produc- A report analysing the accuracy in data have been delivered alongside the tion of fringe patterns, is done after absolute visibility, closure phase and dif- AMBER ones for optimised data reduc- forming three exit pupils. The exit pupils ferential phase identified critical software tion and post-processing purposes. With are physically placed in a non-redundant and hardware improvements required the advent of GRAVITY fringe tracking manner such that the set of three con- by AMBER (Malbet et al., 2008). The main has become an integral part of the sci- tained spatial frequencies in the final modifications in AMBER were the ence observations and the data from the interferometric image are fixed (i.e., non- replacement of its polarisation filters fringe tracker are used in the data reduc- homothetic) but different and identifiable. which were responsible for parasitic tion process. The four beams — three intensity moni- Fabry-Perot fringes in all the spectro-in- toring beams and the one interferometric terferometric measurements, and Continual enhancements of AMBER and beam containing all the information for improvements in its operation and main- the VLTI resulted in steady improvements the three baselines — are then spectrally tainability. On the VLTI, after a significant of the limiting magnitude and operational dispersed before detection (Petrov et al., improvement in the delay line models, a efficiency. AMBER’s self-coherencing 2007). continuous effort resulted in the progres- was introduced in April 2012 for the low- sive reduction of the vibrations in the resolution setup. This mode allowed auto- The integration of AMBER operations into coudé trains of the UTs. A decisive factor matic real-time fringe centring at a rela- the complex VLTI and telescope architec- was the implementation of a faster loop tively low cadence when fringe tracking ture was an iterative process (see Mérand to counter the flux dropouts seen in the with FINITO was not possible (for exam- et al., 2014). For example, operations instrument – a higher correction rate was ple, because of seeing conditions or low began with the UTs equipped with the made possible by offloading the meas- source flux). The instrument intervention Multiple Application Curvature Adaptive ured IRIS tip-tilt to the feeding optics of performed at the end of 2012 changed Optics (MACAO) guiding systems, before AMBER. AMBER’s performance could the spectrograph beamsplitter that the arrival of the versatile Auxiliary be further improved thanks to the arrival caused internal reflections; this resulted Telescopes (ATs). The ATs were commis- of the Fringe-tracking Instrument of NIce in almost a 30% improvement in through- sioned in the summer of 2006 and first and TOrino (FINITO) (Haguenauer et al., put in the interferometric channel. Better offered in April 2007 with a limited set 2008). polarisation control, by means of birefrin- of baselines. On the VLTI side, the injec- gent lithium niobate (LiNbO3) plates, was tion of the light into the instrument’s introduced in October 2014, following an single-mode fibres was optimised by Arrival of FINITO earlier implementation in the Precision controlling the tunnel tip-tilt inside the Integrated Optics Near-infrared Imaging VLTI laboratory using the InfraRed Image The art of fringe tracking was introduced ExpeRiment (PIONIER). Such plates allow Sensor (IRIS). This sensing sub-system into AMBER operations for the Period 80 the equalisation of the phase difference was operated from 2006 onwards but call for proposals in October 2007. The between the two polarisation stages and using the telescopic XY table as a correc- purpose of fringe tracking is to nullify the add them incoherently, improving the tive system. As a result, it operated at fringe movement caused by atmospheric sensitivity by a factor of nearly two. At the a sub-optimal slow rate of about 1 Hz, turbulence which blurs the contrast of same time the observing efficiency but it was nonetheless quickly seen as the fringes. With FINITO, longer detector improved dramatically by a factor of three a mandatory prerequisite for improved integration times could now be employed. since the start of operations, resulting in beam injection. Longer integration of fringe patterns much shortened execution times of allows the observation of targets at a 20 minutes per Observation Block (OB), After the first couple of years of operation higher spectral resolution, or of fainter down from one hour. it became clear that AMBER was not sources, or allows a lower intrinsic fringe fulfilling all of its potential on the VLTI. As contrast to be measured. Additionally, the first VLTI instrument to possess longer detector integration times also Science delivered high spectral resolution, and therefore to allow the full detector to be read out, require longer integration times, it increasing the spectral range covered by ESO’s AMBER and VLTI infrastructure demanded much more from the VLTI the observations. Fringe tracking was delivers observable quantities that reveal infrastructure than its predecessors. implemented in the VLTI by means of the the wavelength-dependent structure and Flux injection and phase stability had to separate instrument FINITO (Gai et al., geometry of astrophysical sources at a be significantly improved. The observa- 2004). Its purpose was to measure the very high angular resolution: indeed, the tion overheads were large, the quality of broad-band fringe jitter at kHz frequency best available from any of the ESO instru- the high-spectral-resolution data was in the H-band. The FINITO signal was ments. At the wavelength of operation, degraded, and the sensitivity was limited. processed and injected back into the AMBER can reach angular resolutions of These problems and others were tackled VLTI in real time via the Reflective Mem- the order of 1 milliarcsecond. The differen- thanks to increased efforts from the ory Network architecture. As of Period 83 tial phase accuracy allowed photocentre

The Messenger 174 – December 2018 9 Telescopes and Instrumentation de Wit W.-J. et al., AMBER: The VLTI’s Astronomical Multi-BEam combineR

displacements as small as 10 micro- Figure 1. Distribution (in percent) of science topics arcseconds to be measured, for example, of 153 peer-reviewed articles based on AMBER data. The large majority of papers (31%) are centred on in the alignment between the stellar young stars, in particular the structures that allow rotation axis and the orbits in the Fomal- the growth of the star (for example, accretion disc, haut debris disc system (Le Bouquin disc-wind, jet formation). et al., 2009). The superb angular resolution allows breakthrough science by delv- ing into spatially unexplored regions on stellar surfaces, in the circumstellar environment of young and evolved stars, and Young stellar objects: 30.7% around the active nuclei of galaxies. Evolved massive stars: 22.2% AGB: 13.7% ESO’s Telescope Bibliography telbib pro- Classical Be: 9.8% vides the following statistics for the sci- Multiples: 7.8% ence legacy of AMBER. Up to October Technical: 5.25% 2018, AMBER data contributed to 153 Other: 10.6% peer reviewed science papers. This number nearly equals the total number of totic giant branch [AGB] stars) and high- the first direct images of the innermost science papers produced with data from mass stars. Their fractional contribution part of the wind-wind collision zone, a MIDI, which was decommissioned in by sub-category is presented in Figure 1, key feature of the observed erratic behav- March 2015. The number of papers makes and we highlight some of these science iour of this object. A series of papers these two instruments the most produc- cases in the next section. reported investigations of the evolution of tive in terms of science papers produced novae and their environment (for exam- using data from a long-baseline optical The sub-topics cover a wide range of ple, Chesneau et al., 2007) and the interferometry facility. Over the period science cases, from the evolution of cool detection and characterisation of binaries 2015–2017, about ten papers per year evolved stars, concentrating on circum and higher order stellar multiples. AMBER were published with AMBER data; given binary discs of post-AGB stars and also observed the nucleus of the Seyfert that telbib publication analyses estimate supergiant B[e] stars, the inner wind galaxy NGC 3783, deriving a ring radius an average lag time between the execu- regions in neutral and ionised gas of for the toroidal dust distribution of tion of a programme and the publication post-red supergiants and unstable hyper- 0.74 milliarcseconds (Weigelt et al., 2012). of a paper of approximately 5.4 years for giants, and the nebulae of Wolf-Rayet 50% of the data, it is not unreasonable stars. The most cited AMBER paper to date to expect a few tens of peer-reviewed analyses the measured radii of seven AMBER papers to see the light of day in One example of exploiting aperture syn- low- and very-low-mass stars, finding the coming years. thesis imaging to better understand agreement between the observed radii evolved high-mass stars is the image of and the predictions of stellar evolutionary Regarding instrument modes, the relative the well-known luminous eruptive star models for magnetically active low-mass demand for the various AMBER spectral h Carinae, shown in Figure 2. It is one of stars (Demory et al., 2009; Figure 3). settings varied substantially, with the low-resolution and medium-resolution setups vying for dominance. The low- resolution time requested strongly dominated up to Period 88, with over 80% of the demand between Periods 79 and 82. After Period 88, the medium-resolution

settings (H- and K-band) became more BRDBNMC popular, leading to approximately 60%

of the time requested after Period 94. The LHKKH@Q high-spectral-resolution setting request fluctuated around 20% of the total time Figure 2. Aperture synthesis image of after Period 82, coinciding with the intro- NRHSHNM h Carinae in the Brackett g HI transi- duction of FINITO to VLTI and UT tion at a radial velocity of – 277 km s–1. UDO operations. l The image shows both the dense stellar wind surrounding the primary

1DK@SH star (red, yellow, and green regions) The topics of the science papers deal DX and the fan-shaped wind-wind colli- almost exclusively with stellar evolution, in l sion zone (blue). The image field of particular star formation and young stars, view is 50 × 50 milliarcseconds. Over- laid is a sketch of the orbit of the late evolutionary stages of intermediate- l l secondary star (adapted from Weigelt and low-mass stars (for example, asymp- 1DK@SHUDONRHSHNMLHKKH@QBRDBNMC et al., 2016).

10 The Messenger 174 – December 2018 We note that, after the first successful 1.0 Figure 3. The mass-radius relationship science observations, an issue of Astron- for M- and K-type dwarfs, for which radii have been obtained via direct omy & Astrophysics (Volume 464, No. 1, 0.8 measurements with AMBER and VINCI March II 2007) was dedicated to the first (filled circles). Other long-baseline )

results from AMBER, including the instru- ๬ observations are overplotted (solid, R

( 0.6 ment description and the first astrophysi- dashed and dotted lines) in open circles. Evolutionary predictions are for cal results. an age of 5 Gyr and different values for Radius 0.4 the convective overshoot parameter (adapted from Demory et al., 2009). The physics of young stars 0.2

AMBER’s scientific contributions in the Figure 4 (below). Left: Mid-infrared field of young stars are impressive. It is Spitzer composite image (3 × 3 arc- 0.05 minutes) of the surroundings of the clear that stars accrete mass from their ) 20 M⊙ young star IRAS+13481-6124, ๬ 0.00 R

environments, as revealed, for example, ( revealing the outflow from the star –0.05 by the spectroscopic and photometric C as indicated. Right: AMBER aperture activity of young objects. How the pro- O– –0.10 synthesis image zooming in on the accretion disc. Modelling shows that –0.15 cess of accretion actually manifests 0.10.2 0.30.4 0.5 0.6 0.70.8 0.9 the disc has a dust-free region inside remains less clear. AMBER contributed to Mass (M ) 9.5 astronomical units from the star. ๬ revealing the geometry of the accretion The structure is oriented perpendicu- environment in young stars. lar to the outflow direction (adapted from Kraus et al., 2010). Notably, an aperture synthesis image created by Kraus et al. (2010) showed a 3.6/4.5/8 µm VLTI/AMBER image 2.2 µm disc surrounding a young, 20 M star ⊙ Ou

at a spatial resolution of 2.4 milliarcsec- t ax ﬂo onds (see Figure 4). It demonstrates the Ou is w

tﬂ inevitability of disc formation for mass ax ow accretion to proceed, even in high-mass is luminous stars. How the disc is shaped and its structure closer to the stellar sur- Beam face are revealed in the 1500 visibility measurements reported by Benisty et al. (2010) where the inner few astronomical units dominate the emission in the H- and 3 0.006 K-bands. These hot disc regions give ಿ ೀ rise to large-scale ionised winds (for 600.000 au 20 au example, Malbet et al., 2007), or they diminish to very compact ionised regions possibly identifying the actual process of depositing material on the star’s surface Fast rotating stars Domiciano de Souza et al. (2012) infer or the jet launching environment (Kraus et the equatorial radius, the inclination angle al., 2008). The brightness, relative prox- The fastest rotating stars reaching critical and an equatorial rotational velocity of imity and complexity of various physical velocity (at which a star breaks up) are 298 +/– 9 km/s for the rapidly rotating

processes operating during the accre- the 6 to 15 M⊙stars. Indeed, the early B star Achernar using this technique (see tion process make bright young stars B-type stars, as a general group, display Figure 5). Owing to this fast rotation, the extremely suitable targets for spectro-in- the fastest rotation of all stars. Such classical Be stars are capable of launch- terferometry. These and other high-angu- rapid rotation can be assessed by either ing stellar material into a circumstellar lar-resolution findings formed the ration- measuring the shape of the star (via visi- disc. With the help of AMBER data, direct ale for the second conference dedicated bilities), or exploiting Doppler effects in evidence was obtained that these discs to intermediate-mass pre-main-sequence a spectral transition resulting from the are in clear Keplerian rotation, a sugges- stars organised in Vitacura in 2014 (de stellar rotation. In the latter case, one tion that dates back to their discovery Wit et al., 2014). exploits the fact that the photocentre of in 1866 (Meilland et al., 2007). Further the stellar surface in the approaching part investigation shows that this type of of the spectral line profile is different from disc is generally expanding, and forms a that in the receding part. AMBER allows one-armed spiral density pattern that this effect to be measured as it provides precesses with a period of a few years access to the phase changes relative to (Carciofi et al., 2009). each spectral bin.

The Messenger 174 – December 2018 11 Telescopes and Instrumentation de Wit W.-J. et al., AMBER: The VLTI’s Astronomical Multi-BEam combineR

Figure 5. A popular transition is the /G@RD "NMSQ@RS %KTW atmospheres, where the temperature Brackett g H I atomic line at 2.17 µm. is cool enough for dust to form. AMBER For the last AMBER fringes, the rapidly rotating star Achernar (aEridani) was observations of these stars have been targeted. Its rotation (~ 90% of the crit- shown to be largely consistent with ical velocity) causes its equatorial dynamic model atmospheres at individual diameter to be about 35% larger than phases, and have confirmed time varia its polar diameter. The blue back- ground image shows the interferomet- bility of molecular extensions on time ric beam with fringes (stretching from scales of weeks to months. x = –1.0 to 0.9), and overlayed are the extracted contrast for the three !Q For RSGs, it has been speculated that baselines and the closure phase. The a the same processes may explain their image also shows one photometric §L beam (x = 0.9 to 1.5), and overlaid is G mass loss as well. However, AMBER Achnernar’s flux spectrum. AMBER observations of RSGs showed extensions observations are sensitive to the disc KDMFS that are larger than expected based on rotating in the Brackett g HI transition, UD current dynamic model atmospheres. characterised by a decrease in con- 6@ trast and the opposite phase signature. Direct comparisons of AMBER data with 1D and 3D dynamic model atmospheres showed that current models of RSGs based on pulsation and convection alone cannot explain observed extensions of RSG atmospheres, and cannot explain how the atmosphere is extended to radii where dust can form (for example, Arroyo-Torres et al., 2015). This points to l l missing physical processes in current QAHSQ@QXTMHSR RSG dynamic models — an unsolved problem that is a heritage of AMBER and Cool evolved stars and their further or transition to hotter Wolf-Rayet stars, that is due to be investigated further by evolution depending on their mass. the next-generation interferometric instruments. AMBER has also provided obser- The AMBER instrument made significant Previous interferometric observations of vations of non-Mira red giants which are contributions to the study of cool evolved cool evolved stars were usually made via partly consistent with hydrostatic models stars and their further evolution through- broad filters or sequentially in a few and partly show discrepancies with out its operational lifetime. In total, about narrow bandpasses, the latter a time- models similar to RSGs. AMBER has pro- a quarter of the papers based on AMBER consuming technique. AMBER has been vided image reconstructions of both data fall into this scientific category. unique in providing detailed measure- the extended atmospheres of AGB stars ments of individual lines, in particular the (for example, Le Bouquin et al., 2009) Cool evolved stars comprise AGB and individual CO first overtone lines near and RSGs (for example, Ohnaka, Weigelt red supergiant (RSG) stars, which are 2.3 µm, with a high spectral resolution & Hofmann, 2017; see Figure 6). located on the Hertzsprung-Russell- of R ~ 12 000 (for example, Ohnaka et al., Diagram (HRD) at effective temperatures 2011), or measurements across the full between about 2500 and 4500 K. They K-band with the medium spectral resolu- Conclusion cover a large range of luminosities tion of R ~ 1500 (for example, Wittkowski depending on their initial mass, where et al., 2008). Cool evolved stars appear On the night of 3 September 2018, the AGB stars are low- to intermediate-mass to be extended in bandpasses that are interferometric instrument AMBER stars, and RSGs their massive and dominated by molecular layers, and observed its last fringes. After serving high-luminosity counterparts. Owing to much more compact in near-continuum the European astronomy community for the low temperatures of AGB and RSG bands. Observing spectral channels over thirteen years, the instrument was stars, molecules and dust can form across the K-band at once has been an decommissioned during the course of in their atmospheres, and they are sub essential tool to constrain dynamic model the interventions in the VLTI laboratory sequently expelled into the interstellar atmospheres. High-spectral-resolution that were necessitated by the arrival medium via a stellar wind with similar studies of CO first overtone lines showed of the adaptive optics system for the ATs mass-loss rates found in AGBs and extended CO layers in detail as well (NAOMI) and the Multi AperTure mid- RSGs. When AGB stars have lost a sig as their vigorous, inhomogeneous large- Infrared SpectroScopic Experiment nificant fraction of their mass, they evolve scale motions. (MATISSE). AMBER operations encour- again toward higher effective tempera- aged the development of the FINITO tures, and via a post-AGB phase they For Mira-variable AGB stars, it has been fringe-tracker to beat the atmosphere’s transition to planetary nebulae. RSG stars shown that pulsation and convection phase disturbance, which enabled longer explode as core-collapse supernovae can lead to strongly extended molecular detector integration times. With the

12 The Messenger 174 – December 2018 Figure 6. Velocity-resolved aperture synthesis imaging of the red supergiant Antares. This monochromatic image was obtained at the R centre of the CO transi- SX

L@ tion at 2.30665 µm. S AMBER’s high spatial DMRH RD

MS and spectral resolution EE

CH allowed the observa- MN tions to measure the “vigorous” motion above

@S HN the complex red supergiant photosphere -NQL@KHRD

#DBKHM (adapted from Ohnaka, l Weigelt & Hofmann, 2017).

l 1HFGS@RBDMRHNMNEERDSL@R decommissioning of both AMBER and (P2VM) method implemented in the Carciofi, A. et al. 2009, A&A, 504, 915 FINITO, the VLT Interferometer bids fare- instrument’s design. Originally invented Chesneau, O. et al. 2007, A&A, 464, 119 Demory, B. O. et al. 2009, A&A, 505, 205 well to the era of the first generation of for AMBER, this method has found de Wit, W.-J. et al. 2014, The Messenger, 157, 50 interferometric instruments at Paranal. its way into PIONIER and GRAVITY Domiciano de Souza, M. et al. 2012, A&A, 545, 130 The new era of VLTI operations is marked instruments. Gai, M. et al. 2004, SPIE, 5491, 528 by routinely making use of the four-tele- Haguenauer, P. et al. 2008, SPIE, 7013, 7013C Jennison, R. 1958, MNRAS, 118, 276 scope combiner instruments, GRAVITY, Furthermore, AMBER was the first Kraus, S. et al. 2008, A&A, 489, 1157 PIONIER, and the latest addition, instrument for which real time fringe- Kraus, S. et al. 2010, Nature, 466, 339 MATISSE. tracking data were used to enhance Le Bouquin, J. B. et al. 2009, A&A, 496, L1 the data reduction. This is routinely done Le Bouquin, J. B. et al. 2009, A&A, 498, 41 Malbet, F. et al. 2007, A&A, 464, 43 In many respects, AMBER represented a for GRAVITY, and will likely be done for Malbet, F. et al. 2008, arXiv:0808.1315 breakthrough in optical interferometry. MATISSE. The latter instrument also Meilland, A. et al. 2007, A&A, 464, 59 At Paranal, it was the first instrument to inherited the fringe combination scheme Mérand, A. et al. 2014, SPIE, 9146, 9146J combine the beams of three telescopes, employed by AMBER. Finally, the unique Ohnaka, K. et al. 2011, A&A, 529, 163 Ohnaka, K., Weigelt, G. & Hofmann, K.-H. 2017, providing access to the all-important aspect of the AMBER instrument was Nature, 548, 310 closure phase, without which it is not its spectral resolution, which initiated the Paresce, F. et al. 1996, The Messenger, 83, 14 possible to reconstruct images of celes- technique of spectro-interferometry. With Petrov, R. et al. 2007, A&A, 464, 1 tial objects from interferometric observa- a resolving power of 12 000, AMBER was Petrov, R. et al. 1998, The Messenger, 92, 11 Weigelt, G. et al. 2012, A&A, 541, L9 tions. As such, and in addition to the able to spectrally and spatially resolve Weigelt, G. et al. 2016, A&A, 594, 106 visibility and phase studies, AMBER has the dynamics of circumstellar phenom- Wittkowski, M. et al. 2008, A&A, 479, L21 delivered a great number of images at ena, and paved the way for GRAVITY and Woillez, J. et al. 2015, The Messenger, 162, 16 milliarcsecond scales, providing new MATISSE operations. The literature will insights into astrophysical areas that no doubt continue to see numerous Links could not be spatially resolved with single science papers originating from AMBER optical telescopes (see Figures 2, 4, 6). data in the coming years. 1 List of AMBER consortium members: The second generation of VLTI instru- http://amber.obs.ujf-grenoble.fr/spipe703.html ments has inherited and profited from the lessons learned, and the VLTI upgrade Acknowledgements started in 2015 is providing a further A large consortium of institutes, scientists and engi- enhanced facility (Woillez et al., 2015). neers contributed to AMBER. A list of the AMBER The performance demonstrated today by consortium members is included in this article1. We GRAVITY shows that the initial goals set thank Armando Domiciano de Souza and Thomas by AMBER were not unrealistic. Rivinius for useful discussions regarding classical Be stars. Part of AMBER’s legacy is the novel way in which interferometric observables References are extracted from the data, how fringe Arroyo-Torres, B. et al. 2015, A&A, 575, 50 patterns are initially recorded. Another Baldwin, J. et al. 1996, A&A, 306, 13 innovation is the pixel-to-visibility matrix Benisty, M. et al. 2010, A&A, 511, 75

The Messenger 174 – December 2018 13