The Messenger 155

8/12/2019 The Messenger 155

1/68

The Messenger

No. 155 March 2014

LaserGuideStarFacilityup

grade

VUDS

V

IMOSUltraDeepSurvey

HUGSH

AWK-Iultradee

psurvey

Celeb

ratingESOs50years

inChile


2/68

2 The Messenger 155 March 2014

the instrument. The mass and size ofthe monster were daunting. Cooling theinstrument required the use of a pre-cooling circuit, prototyped on IRSPECand now common in most infrared instru-ments deployed at the VLT. In 1996,during the big-bang upgrade of the NTT,IRSPEC was retired and its cryostatwas used to test many of the mechanicalfunctions of ISAAC.

As the PI was fond of saying, ISAAC haddiamond-turned-nickel-coated-post-polished mirrors. The cold structurewhere the instrument optics weremounted was particularly complex andwould have been almost impossible tomachine from a single block. Two castaluminium structures were made at alocal foundry and the huge vacuum ves-sel came from France. The bearings usedto turn the massive gratings and lterwheels were off the shelf. They were dis-assembled by Jean-Louis down to theirsmallest components and then reassem-bled without the outgassing and freeze-prone lubricants and were equipped withnew rolling elements. The bearings forthe ISAAC lter wheels are comparable insize to an entire IRAC (the IR camera forthe 2.2-metre telescope commissioned in1994) lter wheel.

With the high number of functions inISAAC, the traditional warm motor witha feed through to the vacuum vesselwas not an option. A development pro-gramme inside ESO converted commer-cial stepper motors to function in cryo-genic conditions. For ESO, ISAAC wasalso the rst instrument equipped fromthe start with closed-cycle coolers. Theoriginal Leybold compressor outlastedthe instrument, with more than 100 000hours of continuous operation. The enor-mous platform which carried the pow-ered co-rotator and electronics (a featurethat became a common solution forinstruments at Paranal) arose as the onlyviable option as the instrument pro-ceeded through the construction phase.

To cool down such a massive load,a custom ESO cryogenic control systemwas constructed by Joar Brynnel, whoalso built the electronics that would drivethe instrument functions using earlyversions of the, then, ESO standardMACCON motor controllers. The pre-

Jason Spyromilio1

Jean-Gabriel Cuby2

Chris Lidman3

Rachel JohnsonAndreas O. Jaunsen4Elena Mason5

Valentin D. Ivanov1Linda Schmidtobreick1

(The ISAAC Instrument Scientists inchronological order)

1ESO2Laboratoire dastrophysique de

Marseille, Universit Aix-Marseille &CNRS, Marseille, France

3Australian Astronomical Observatory,Epping, Australia

4Institute of Astrophysics, University ofOslo, Norway

5INAFOsservatorio Astronomico deTrieste, Italy

ISAAC was switched off, almost cer-tainly for the nal time, on 12 December2013. The last observing block exe-cuted was OB1030962, the target,Supernova 2013ct, for a programmewhose principal investigator just hap-pened to be the rst instrument scien-tist for ISAAC. All constraints wererespected and spectra of the targetdetected are the public comments inthe log. A short history of ISAAC, fromthe instrument scientists viewpoint, ispresented.

Building ISAAC

The Infrared Spectrograph And ArrayCamera, ISAAC, was the second instru-ment mounted on the Very Large Tele-scope (VLT), following the FOcal Reducerand low dispersion Spectrograph (FORS1).

The Principal Investigator (PI) of ISAACwas Alan Moorwood. Although variousappreciations of Alan have already beenmade (e.g., de Zeeuw, 2011; Leibundgutet al., 2011) and it is cer tain that he madeprofound contributions to all aspectsof ESO, ISAAC was one of his proudestcreations. ISAAC was one of two rstgeneration instruments for the VLT builtin-house at ESO Garching by a teamof people working for Alan in a deeplypersonal and motivating atmosphere (the

other instrument was the Ultraviolet andVisible Echelle Spectrograph [UVES]).

Although much of the discussion on theinstrumentation suite for the VLT tookplace prior to the approval of the projectin 1987, the denitive instrumentationplan was authored by Sandro DOdoricoand Alan in June 1989. It appears pres-cient today, since almost all rst genera-tion instruments had their genesis inthat document. ISAAC in effect was theoutcome of option E4 for a combinedspectrometer/imager exploiting a singledetector. It should also be recalled thateven in 1989 the choice of site for the

VLT had not been settled. Vizchachasnext to La Silla with four Unit Telescopesin a linear conguration was still thedefault option for the VLT, with Paranalbeing considered as an alternate site.

At the time of ISAACs design design,and even in the early construction period,the instrument was base lined on 256 by256 pixel detectors. It should be recalledthat the rst infrared camera: IRCAMon the UK Infra Red Telescope (UKIRT)only had rst light in 1986 with a 62 58indium antimonide (InSb) detector(McLean et al., 1986). By the time theinstrument was deployed on Paranal in1998, the short wavelength arm had afancy new 1024 1024 HgCdTe detector,but the long wavelength detector was stilla 256 256 pixel InSb device.

The team assembled by Alan to buildISAAC relied very much on the instru-mentation experience at ESO, based onthe immediate previous generation ofinstruments (IRSPEC on the New Tech-nology Telescope [NTT] and the twoinfrared cameras on the MPG/ESO 2.2-metre telescope). Alan had alreadybrought an infrared heritage to ESO inthe Geneva days, from his work at ESAand ying balloons with infrared payloadsat University College London. Under hisleadership and a series of ever moresuccessful instruments, an effective teamhad been created.

The optical design, as with a lmost allinstruments at the VLT, came f rom thetrusted hands of Bernard Delabre.Gotthard Huster designed the mechanicalsystem with help from Ralf Gonzelmann.Jean-Louis Lizon and Armin Silver built

Telescopes and Instrumentation

ISAAC. An Appreciation


3/68

3The Messenger 155 March 2014

cooling accelerated the cool down, aprocess that came in very handy intesting, but also in operations when astuck function had to be unstuck!

Gert Finger was in charge of the detectorsystem, with Manfred Meyer buildingthe readout electronics. During the pro-gramme Jorg Stegmeier and LeanderMehrgan joined the detector group tomake the rst generation of IRACE (Infra-Red Advanced Controller Electronics)controllers. Giana Nicolini helped with thecharacterisation of detectors and pre-pared the upgrade of the Santa BarbaraResearch Corporation (SBRC) 256 256 pixel InSb array to the brand new32-channel Aladdin 1K 1K InSb array.

Transputers formed the basis of the read-out electronics, while the IRACE platform,in a shift from the VLT standards of

VxWorks on 68030 processors, moved tothe newfangled Sparc machines thatwere all the rage. In a minor panic duringthe early years of operation, the remain-ing worldwide supply of T2 and T8 trans-puter chips were procured by ESO so asto have sufcient spares for ISAAC and,of course, the other infrared instrumentsthat followed in its path.

Peter Biereichel who had written much, ifnot all, the software for the early infraredinstruments led the control software effortand along the way brought in ThomasHerlin and Jens Knudstrup to help. Mucheffort was needed to tame the earlyversions of the VLT common software.

Anton Van Dijsseldonk project-managedall this effort before the days when hewould have been called project manager.

Getting ISAAC to the sky

Following some design review (therewere not so many of them in those days),

Alan was convinced to hire an instrumentscientist to support the commissioningof the instrument and the development ofdata reduction tools. The rst author ofthis appreciation was hired in 1993 forthis post and was succeeded by the sec-ond author after less than 18 monthson the job, and long before the instru-ment ever went to sky. The third authortook over after commissioning, etc, etc.

The list of ISAAC instrument scientists ishopefully identical to the authorship ofthis article. Many large and small prob-lems needed to be resolved, such as howthe instrument interacted with the tele-scope, calibration challenges, observingsequences, etc, etc. For example, chop-ping via a synchronised timing signal,rather than a hardware connection, maybe a trivial issue today, but the ISAACcontroller, mentioned above, was notcompatible with the VLT standard hard-ware for timing. A simple solution of ahardware trigger from the TIM board(a VxWorks custom board built by ESOfor timing synchronisation across the VLT)was used.

The ISAAC software innovation stableprovided the three state buttons thatcontinue to plague the VLT control panelsand the Real Time Display (RTD). The

VLT standards forbade dedicated hard-

ware to connect the data acquisition sys-tem in the dome with the control room.

As a minimum for target acquisition therewas a need for a networked workstationdisplay that allowed astronomers to inter-act with the data live, rather than waitingfor a data reduction system to digest aFITS le. In the days before requirementsdocuments, a conversation between theinstrument scientist and Thomas Herlinto esh out the idea of a display thatwould stop astronomers looking at nd-ing charts through lamps to get the rightorientation, provide trivial zooming andscaling and allow the instantaneous dataas well as the integrated frame to appearin real time, resulted. Envisaged fromthe very start to have an application inter-face to allow plug-ins, the RTD was awild success and is not only used for alldetector systems on the VLT but in vari-ous incarnations also formed the basisfor theskycattool, the FORS InstrumentalMask Simulator (FIMS), and ended up inwide use by many observatories. Everyacquisition at ESOs optical telescopesrelies on the RTD to click the target or theguide star!

For infrared astronomers, prescriptionobserving per templates was a signicantpsychological challenge to overcome.However, ISAAC was in many waysresponsible for bringing the concepts oftemplates into VLT operations. Erik Allaertwas critical in making the toolkit to man-age scripted observations in a sensiblemanner: the resulting tool is now used

Figure 1. Jean-Louis Lizon, who powered ISAAC upin 1998, switches the instrument off for the last timein December 2013.

Figure 2.A younge r Jean-Louis Li zon than in Fig-ure 1 checks the ISAAC cold structure #1 during theconstruction phase in 1995.


4/68


motor cover on UT1 in February 1999, acomplete demolition of the control roomwhile instrument scientist number 2debugged software and a disconnectedtelescope cooling hose that washed thewhole Nasmyth platform down. ISAAC,together with FORS1, were incidentally therst instruments to be brought into theQC0 (Quality Control) and QC1 processes.

ISAAC observations

With over 850 papers to its name, ISAACis almost certainly amongst the mostproductive infrared instruments. It isespecially difcult to pick particular sci-ence results and our apologies are publicand up-front. We have picked a few thatentertained us.

Messier 16Mark McCaughrean was a regularChristmas visitor observer at UT1 in therst years of operation. His kind, beauti-fully handwritten appreciation letters were

ubiquitously at Paranal, as the Broker forObservation Blocks, BOB. As an aside,his inclusion of positively poor soundchoices also resulted in the silencing ofBOB on most consoles. Copying fromthe Infrared Space Observatory (ISO)operations, the idea of templates wasintroduced into the ISAAC draft usermanual in 1995 and was taken into theoperations scheme as the building blockfor OBs by the Data Flow Project Team,co-ordinated by Preben Grosbl, includ-ing, amongst others, Michle Peron,Dietrich Baade and Bruno Leibundgut.ISAAC was the rst VLT instrument witha full instrument observing simulatorwith panels and displays talking to codethat generated pseudo-realistic data andsimplistic pipelines.

The original proper ISAAC pipeline wasdeveloped by Nicolas Devillard and

Yves Jung under the tutelage of PascalBallester and instrument scientists 2, 3,etc. The xes, upgrades and updates,etc, are too numerous to mention. How-ever, at some point after much discus-sion, we did get the sky subtraction right.Pascal also provided the early exposuretime calculators.

Offspring and early days

Jean-Louis Lizon was expeditious in theassembly of the instrument in Garchingand, with Unit Telescope 1 (UT1) rst lightslipping to 1998, Alan decided to usethe expertise developed with ISAAC torapidly develop an instrument for thesoon to be upgraded NTT. Launched asa project in 1996, SofI, the Son of ISAAC,was in effect one imaging arm of ISAACwith a grism wheel providing the spectro-scopic capabilities.

While the correct name for SofI obviouslyshould have been JACOB (Jean-Gabrielsand Alans Camera for OBserving) orfor that matter ESAU (ESO Array Unit),pronunciation and modesty issues pre-vailed to leave us with SofI.

SofI went to the NTT in 1997 and wassuccessfully commissioned by the PI,the construction team and instrumentscientist number 2, while at the sametime bringing up to speed the futureinstrument scientist number 3. The com-

missioning went ahead with very fewproblems, thereby testing not only theinstrument, but also the bulk of theISAAC sof tware. By mid-1998, with UT1in full swing of commissioning, Jean-Louis arrived on Paranal to begin the re-integration of ISAAC.

It is worth recalling that the instrumentlaboratory in the control room at thattime had plastic sheets instead of win-dows and many facilities were sorelylacking. Alan oft would debate whetherthe occasional stuck lter wheel in ISAACcame from dust in the early days of inte-gration. It is true that Jean-Louis hasbeen to Paranal many times to keep theold dog going well past the design life-time of three years continuous operation.

Commissioning of ISAAC was successful,even if interrupted by a stuck altitude

Telescopes and Instrumentation Spyromilio J. et al., ISAAC. An Appreciation

Figure 3. ISAACJHKscolour combination mosaicimage of the Galactic HII region and star-forming com-plex M16 (NGC 6611). H2emission knots show in red.


5/68


posted on the whiteboard behind whatwas the UT1 console (now the VLT Inter-ferometer), and his spectacular imagesof various nebulae (one is shown in Fig-ure 3) have been icons of infrared astron-omy since the very rst images of Orionwith IRCAM at UKIRT.

Faint Infrared Extragalactic Survey(FIRES)Marijn Franx pestered the observatoryinto improving its efciency of operationsand calibration until we nally agreed thathe was right all along. A 26.5-hour totalintegration time in Ks, 25.9 hr inJsand24.4 hr in H, provided one of the mostimpressive deep elds from the ground.

The image quality of the data was trulyspectacular with the nal combinedimages below 0.52 arcseconds. An imagefrom the FIRES programme is shown inFigure 4.

Spectroscopy of high-redshift galaxiesIn the early days of the VLT quite a lot ofcatch-up with Keck was necessary. Oneof the best compliments to the instrumentwas the excellent dataset collected onhigh-redshift Lyman-break galaxies byMax Pettini, Alan, instrument scientistnumber 2 and Maxs California and EastCoast based collaborators. An exampleof a long-slit spectrum of a Lyman-breakgalaxy atz= 3.234 is shown in Figure 5.

Observing through the limb of JupiterSome observations simply dont t thetemplate package. Measuring, amongstothers, the temperature prole of theJovian atmosphere by taking spectra ofHIP 9369 during an occultation by the

Figure 5. ISAAC long-slit spectra of the quasarQ0347-C5 taken by Max Pettini in 1999 (fromPettini et al., 2001). The lef t panel shows the two-dimensional spectrum with the [O III] 4959 and5007 emission lines, the central panel anenlargement of the 5007 line kinematics and theright panel the tted radial velocity along the slit.

10

1

Figure 4.Three -colourcomposite image of theeld around the massivegalaxy cluster MS 1054-03 (z= 0.83) taken aspart of FIRES. ISAACJsand Ksand HST WFPC2F814W images werecombined as green, red,blue respectively. FromFrster Schreiber et al.(2006).

limb of Jupiter was certainly one of themost memorable observations made.Keeping the slit orientation parallel to thelimb in real time while everything is mov-ing and turning was an observationaltour de force. The results were publishedby Raynaud et a l. (2003).

Acknowledgements

While it is probable that we have missed some of thepeople who participated in building ISA AC, it is well-nigh impossible to count and name all the staf f atthe La Silla Paranal Observator y who have worked

(almost) every day for the past 15 years to keep theinstrument going. But it would be impossible to con-clude without attempting to thank some of them:Hans Gemperlein, Gustavo Rahmer, Markus Kissler-Patig, Vanesa Doublier, Steen Skole, AndreasKaufer, Gianni Marconi, Claire Moutou, Eline Tolstoy,Ueli Weilenmann, Roberto Castrillo, Pablo Barriga,Gordon Gillet, Pedro Mardones, MassimillianoMarchesi, Eduardo Bendek, Alfredo Leiva, JorgeJimenez, Nicolas Haddad and the rest of the cast,especially the instrument operation team membersover the past 15 years. Many scientists in the DataManagement Division in Garching, and observing atParanal, have had the pleasure of interacting withISAAC: Paola Amico, Mari o van den Ancker, TomBroadhurst, Fernando Comern, Danuta Dobrzycka,Lowell Tacconi-Garman, Christian Hummel,

Wolfgang Hummel, Sabine Mengel, Palle Mller,Monika Petr-Gotzens, Almudena Prieto, FrancescaPrimas, Martino Romaniello, David Silva, Elena

Valenti, Markus Wittkowski and Bodo Ziegler have allhad to live with the ips of the orientation of the slitand the rotations, the occasional glitch and the hun-dreds of users.

The ins trument science team of Rolf Ch ini, Geo rgeMiley, Tino Oliva and Jean-Loup Puget followedthe instrument through the trials and tribulations ofconstruction, picking lters and resolutions, andprovided constant support.

Our nal thanks must be reserved for Al an.

References

Andersen, M. et a l. 2004, A& A, 414, 969de Zeeuw, P. T. 2011, The Messenger, 145, 49Frster Schreiber, N. M. et al. 2006, AJ, 131, 1891Leibundgut, B. et al. 2011, The Messenger, 145, 50McLean, I. et al. 1986, SPIE, 627, 430Pettini, M. et al. 2001, ApJ, 554, 981Raynaud, E. et al. 2003, Icarus, 162, 344

Q0347C5 z= 3.234 PA = 4.5

2.100

4

2

0

A

rcsec

along

the

slit

2

4

2.110

Wavelength (m)

2.120 400

10

0

Distance

(kpc) 10

[OIII]5007

200 0

Rel velocity (km/s)

200 400

50

0

Distance

(kpc)

5

50 0

Rel velocity (km/s)

50


6/68


The launch optics, mounted behindthe secondary mirror of the main tele-scope, receive the visible-light outputfrom the relay bre and propagate anearly collimated 50 cm diameter beamonto the sky. They provide beam ex-pansion, focusing, tip-tilt beam stabilisa-tion and diagnostic functionality.

The scope of the upgrade reported herewas to install a new prototype lasersource. Other subsystems of the LGSFare essentially unchanged.

Sodium excitation

A sodium laser guide star (LGS) is formedby resonantly back-scattering laser lightfrom atomic sodium in the upper meso-sphere and lower thermosphere (abbrevi-ated MLT). Optimisation of the laser for-mat to maximise the return ux is beingactively studied (Holzlhner et al., 2010).

Optical radiation at 589 nm was rstobserved in the night-sky spectrum by

V. M. Slipher in 1929 (Sl ipher, 1929). Thisradiation was later attributed to emissionfrom neutral sodium atoms occurringmainly in the upper mesosphere (Bernard,1939), excited by solar radiation. Lyingabove the range of aircraft, but below thealtitude of orbiting spacecraft, the meso-sphere is one of the least accessibleand most poorly characterised regionsof the Earths atmosphere. Typically it isstudied with sounding rockets or ob-served remotely using laser light detec-

Steffan Lewis1

Domenico Bonaccini Calia1

Bernard Buzzoni1

Philippe Duhoux1

Gerhard Fischer1

Ivan Guidolin1

Andreas Haimerl1

Wolfgang Hackenberg1

Renate Hinterschuster1

Ronald Holzlhner1

Paul Jolley1

Thomas Pfrommer1

Dan Popovic1

Jose-Luis Alvarez1

Juan Beltran1

Julien Girard1

Laurent Pallanca1

Miguel Riquelme1

Frderic Gonte1

1ESO

The Laser Guide Star Facility is part ofVLT Unit Telescope 4 and provides asingle centre-launched sodium beaconfor the two adaptive optics instrumentsSINFONI and NACO. The original facil-ity, installed in 2006, employed a high-power dye laser source, PARSEC, pro-ducing an output beam that was deliv-ered via a single-mode optical breto launch optics located behind the tele-scope secondary mirror. We recentlyinstalled a new prototype laser source,PARLA, based on Raman optical bretechnology. Requirements for the newlaser include start-up times compatiblewith exible observing, an output beamappropriate for the existing bre-deliverysystem and an on-sky power of up to7 watts. This is the rst time that thistype of laser has been deployed at amajor observing facility, and it has apathnder role for future adaptive opticssystems. Reported here are the mainresults of the development, deploymentand early operation since the resumptionof science operation in February 2013.

The Laser Guide Star Facili ty

The main parts of the Laser Guide StarFacility (LGSF) installed on the VeryLarge Telescope Unit 4 (VLT UT4, Yepun)and described by Bonaccini Calia et al.(2006), are shown in Figure 1. These

include a laser cleanroom below theNasmyth platform, an optical bre beamrelay and a launch optical system behindthe telescope secondary mirror.

The laser cleanroom contains the bulkof the control and safety electronics, theoptical bre coupling system, and thelaser source. In the original facility, thelaser source was a dye laser system,PARSEC (Rabien et al., 2003), speciedto produce a single-frequency spectralline centred on the sodium D2aline atwavelength of 589.2 nm. The outputbeam from this laser was transmitted tothe input of the relay optical bre via afree-space beam relay on the laser bench.

The role of this beam relay is to formatthe beam spatially and spectrally, and toactively stabilise the beam position anddirection in order to maintain the opticalbre coupling efciency during operation.Spectral formatting consists of broad-ening the laser line using a phase modu-lator in order to increase the total powerthat can be transmitted through the relayoptical bre.

A large mode area solid-core photoniccrystal optical bre transports thehigh-power visible laser beam from thelaser cleanroom to the launch opticslocated behind the VLT UT4 secondarymirror. This optical bre has a lengthof 27.5 metres, a core mode eld diame-ter of approximately 14 m and anumerical aperture of 0.04. Transmissionlosses, excluding input coupling, areapproximately 10%.


Laser Guide Star Facility Upgrade

Figure 1. Sketch of VLTUT4 showing the mainparts of the Laser GuideStar Facility.


7/68


tion and ranging (LIDAR) systems, andfrom satellites.

The region where atomic sodium is pre-sent lies at an altitude range of around80120 kilometres with some seasonalvariation as well as latitude dependencies.

The MLT is one of the coldest regions onEarth, at a temperature of around 180 K.

This region is characterised by a numberof other features including strong zonalwinds, atmospheric tides and overturninggravity waves. Amongst the metals pre-sent there are Fe, K, Ca and Na, and it isgenerally held that this layer of metalatoms and ions is replenished by micro-meteors in the microgram range, thatis, interplanetary dust particles that areincident on the Earth and ablated in theupper atmosphere.

Although Fe is the most abundant metalatom in this region, Na is used for LGSapplications because it has the highestproduct of abundance and optical transi-tion cross-section at an accessible laserwavelength. Its mean column density is4 109cm2, which is a factor of 108ofthe number for air molecules present inthat region. At the lower boundary of thesodium layer, the concentration of oxygenradicals, which or iginate from ozone, ishigh enough to form products of sodiumhydroxyl and other molecules, and thusremove atomic sodium. Above a temper-ature-dependent threshold, this processappears to be very efcient and the loweredge of the sodium layer is more sharplydened than the upper edge, whereablated sodium ions recombine withelectrons from the ionosphere and fromSolar radiation to form atomic sodium.

The sodium D-line, which is split into ane structure doublet D1and D2atvacuum wavelengths of 589.76 nm and589.16 nm respectively, contributes tothe atmospheric air-glow and is responsi-ble for the familiar orangeyellow lightfrom high-pressure sodium street-lamps.

The D2-transition is approximately twiceas strong as the D1, and it is therefore themost efcient target for LGS generation.Despite this, the sodium layer is opticallythin and on average only 4% of the inci-dent laser light is actually absorbed, withmost of the remainder propagating intoouter space. As a further complication,the D2line is hyperne split into a closely

spaced doublet known as D2aand D2bwith a ratio of transition strengths of 5:3.

These two transitions are separated byabout 1.77 GHz, and under mesosphericconditions they partially overlap due toDoppler broadening of each line to a fullwidth half maximum of about 1.07 GHz.

There are many additional factors inu-encing the return ux, such as the lasertemporal, spectral and polarimetric for-mat as well as the Earths geomagneticeld. Calculations of sodium return uxtypically make extensive use of computersimulation. Nevertheless, most deployedlaser systems to date have concentratedon exciting the D2a-transition, simplybecause it has the highest peak trans-mission cross-section.

For the particular case of the LGSF, simu-lations also suggest that a single narrowline at the peak of the D2aline would bethe optimum format to maximise returnux per watt of launched optical power.In practice, the photonic crystal relay breis an important factor determining theoptical power and launched laser format.

This optical bre does not preserve thepolarisation state of the laser beam andstimulated Brillouin scattering, a nonlinearoptical effect, limits the maximum powerspectral density that it can transmit toapproximately 2.7 watts per spectral line,for lines that are sufciently well sepa-rated. Therefore, the single-frequency lineproduced by the PARLA laser was spec-trally broadened using a sinusoidal phasemodulator in order to launch a greaternumber of spectral lines into the existingrelay optical bre and hence more opticalpower onto the sky. The phase modula-tion is characterised by the modulationfrequency and the amplitude of the peakphase shift measured in radians. Theresulting broadened laser spectrum con-sists of multiple lines forming a frequencycomb inside an overall intensity envelope,where the power of each individual lineis required to remain below the limit forthe relay optical bre. The comb spacingis equal to the 110 MHz frequency of thephase modulator, and the width of theamplitude envelope and the number oflines in the nal laser spectrum increasesmonotonically with the peak phase shift.

The optimum phase shif t for the laserbeam is a trade-off between the totalpower transmitted through the bre and

the overlap between the broadened multi-line laser spectrum and the sodium D2aatomic transition. The overlap decreasesprogressively from 100% at zero peakphase shift to 67% at a peak phase shiftof 3.76 rad, which is close to the maxi-mum accessible with our equipment.

A peak phase shi ft of around 2.6 rad,corresponding to ve spectral lines, wasselected for the installed system for atotal power of 7 watts exiting the relaybre. During laboratory tests it was alsoveried that the spectrum did not changemeasurably after propagation throughthe relay bre and no signal was meas-ured at longer wavelengths where onewould expect to nd spectral compo-nents due to Raman scattering in theglass optical bre if they were present.

PARLA laser system

The PARLA laser source is based onsimilar technology to systems that areunder development for other ESO pro-

jects, which have been descr ibed indetail elsewhere (see Bonaccini Calia etal., 2010; Arsenault et al., 2006; andKaenders et al., 2010). Key elementsinclude a seed laser, a high-powerRaman optical bre amplier and an ef-cient frequency doubling scheme. Themain optical train is shown in Figure 2.Requirements for the system include ex-ible observing, high availability, excellentbeam quality and stability compatiblewith the existing optical bre beam relay.

The laser system includes an electronicscabinet and a laser head which emits afree-space TEM

00laser beam with a

maximum output power of 20.5 watts at589 nm. Although this is a 20-watt-classsystem, the nal operating point of thelaser source in the installed system isonly around 11 watts to achieve therequired 7 watts launched power on sky.

The laser line-width is specied to beless than 20 MHz, and the centre fre-quency is tuneable around the centre ofthe sodium D2atomic transition.

Downstream from the laser head, a peri-scope, phase modulator, and a beamexpander unit relay the optical beam tothe existing optical bre coupling system(see Figure 2). The laser includes a stand-alone computer-controlled system that is


8/68


connected via a point-to-point Ethernetlink to a VLT local control unit (LCU). Thisinterface makes it possible, in a very sim-ple way, to integrate the PARLA laser intothe standard VLT control system at theParanal Observatory.

Results from laboratory system tests,carried out in Europe, are shown inFigures 3 and 4. The laser was testedwith a spare relay optical bre in order toreplicate the optical interface to the tele-scope. The duration of this particulartest was 12 hours. Figure 3 shows theoptical power at the input and output ofthe relay bre. The output power wasmeasured with a bolometer, and the breinput power was measured using a cali-brated pick-off mirror at the bre input.During this test, the direction and lateralposition of the laser beam at the inputto the relay optical bre were actively sta-bilised using an automated closed loopsystem, as is the case at the telescope.

The average power at the input to therelay bre was 8.86 watts, the outputpower 7.16 watts, and the average opticalbre throughput was 80.8%. The through-put includes both the transmission lossesof the relay optical bre (approximately10%) and the input coupling losses,implying an input coupling efciency ofaround 90%.

The infrared laser wavelength was alsomeasured and logged during the testand Figure 4 shows the measured fre-quency error, in MHz, from the nominalset point during the test. The deviation iswithin a few MHz, which meets the goal


Figure 2. Block diagram showing the main opticalscheme for the PARLA laser. The parts of the newlaser system are shown with a blue outline.

Figure 4.Variation of the laser central f requencymeasured during the test in Europe is plotted for aperiod of nearly 12 hours.

Figure 3.The input and output opt ical power meas-ured through the relay bre during the test in the lab -oratory in Europe is plotted. The average throughputis 80.8%.

Lewis S. et al., Laser Guide Star Facility Upgrade


9/68


of 20 MHz. In addition to the selectedoptical characteristics described above,the laser system was subject to a com-prehensive set of performance and func-tional tests, including the control systemand interface to the VLT control software.

Installation and commissioning

The laser was installed on the telescopeat the start of 2013, and commissioningtook place in February 2013. The com-missioning process included standalonetests of the LGSF with the new laser sub-system, and commissioning and scienceverication (SV) with the completeobserving system and the two adaptiveoptics instruments, SINFONI and NACO.Figure 5 shows a photograph of UT4taken during the commissioning in Febru-ary 2013.

The LGS was rst commissioned in astandalone mode, before the start ofobservations and SV tests with theinstruments. To verify the correct outputwavelength, the centre frequency ofthe laser system was scanned in stepsof 250 MHz and 100 MHz around thenominal centre of the sodium D2atransi-tion and the relative return ux (bright-ness) of the LGS was measured using theUT4 guider camera. The guider output isin analogue-to-digital units (ADU) whichare nominally linear with the ux. Figure 6shows the plot of return ux versus laserwavelength taken during on-sky calibra-tion. The measured line shape is a con-volution of the laser spectrum and theDoppler-broadened sodium D2transition;the doublet is therefore not fully resolved.

The main peak corresponds to the D2atransition and the broad shoulder in thehigh frequency side corresponds tothe sodium D2btransition. The laser wastuned to the peak of the D2aline, deter-mined on sky in the nal conguration.

The angular size subtended by the LGSwas measured using the telescopeguide camera at different altitudes andis listed in Table 1. Images of a naturalguide star were taken before each LGSmeasurement and used to infer theseeing. The exposure time for the LGSimages was approximately 2 seconds.

This exposure duration averages themajority of the seeing effect, but formally

Figure 5. Image of VLTUT4 (Yepun) during thePARLA commissioning,taken on 14 February2013.

Figure 6.A wavelength scan through the sod ium D2doublet taken during commissioning.

ESO/G.H

depohl(atacamap

ho

to.c

om

)
http://www.atacamaphoto.com/http://www.atacamaphoto.com/http://www.atacamaphoto.com/http://www.atacamaphoto.com/http://www.atacamaphoto.com/http://www.atacamaphoto.com/http://www.atacamaphoto.com/http://www.atacamaphoto.com/http://www.atacamaphoto.com/http://www.atacamaphoto.com/http://www.atacamaphoto.com/http://www.atacamaphoto.com/http://www.atacamaphoto.com/http://www.atacamaphoto.com/http://www.atacamaphoto.com/http://www.atacamaphoto.com/http://www.atacamaphoto.com/


10/68


tem was handed back to science opera-tions for service observing.

Operation

At the time of writing, the LGSF has beenoperating with the PARLA laser for almostone year.

One of the goals of the laser upgradewas to enable exible observing. By thisit is meant that the laser can be availablefor science observation within a shortperiod of time, formally specied as30 minutes, should the need or opportu-nity arise for a laser-supported obser-vation. To achieve this aim, an idle statewas dened in which the laser waits in alow-power conguration. In this scheme,the lifetime of the high-power opticalcomponents is extended by reducingtheir usage outside of actual observingtime, while still ensuring that the systemis available at short notice. This is asignicant change from the original lasersystem which ran continuously at highpower when it was available. In the newimplementation, the telescope operatorcan wake up the laser using an auto-mated script from the control room andbe ready for operation in approximatelyten minutes. Figure 8 shows the availa-bility of the LGSF in hours per month forthe rst six months of operation after theupgrade.

speaking is too short to qualify as a truelong-exposure measurement of the spotsize.

The return ux, and apparent magnitude,of the LGS were not measured photo-metrically. However, the counts from theSINFONI wavefront sensor, which isbased on photon-counting avalanchephotodiodes, were accessible. The wave-front sensor counts and the system per-formance during SV were the two metricsused to verify that the return ux wassufcient to meet the operational require-ments. Measurements of the apparentLGS magnitude were made usingSINFONI at different altitudes and at dif-ferent times during the commissioningperiod. Typically the ux measured onthe wavefront sensor equated to around6 million counts per second. The opti-cal power launched onto the sky wasbetween 6 and 7 watts during these tests.

This level of photon return was found tobe sufcient to achieve good closed-loopadaptive optics performance during SV.

At the time of commissioning furthermeasurements needed to be donebefore a denitive comparison could bemade between the measured and theo-retical return ux for this system. Theprincipal uncertainty is the sodium abun-dance, which has been shown to varyby a factor of four seasonally (Simonichet al., 1979) and by a factor of up to twoduring a single night. Frequent measure-ments taken at different times of the yearare therefore necessary.

The observations selected for NACO weresuccessfully executed during the commis-sioning nights. Observations were mainlymade with the 7 7 wavefront sensor,and the adaptive optics loop remainedclosed without problems down to an alti-tude of 30 degrees during standard oper-ation tests. The instrument was operatedin a number of different congurations:tip-tilt correction only, high-order correc-tion only, and with full correction. On thenight of 17 February 2013 with full correc-tion, Strehl ratios in the range 15.637%in the Ks-band were measured in seeingranging from 0.76 to 0.99 arcseconds, asrecorded by the differential image motionmonitor (DIMM) at 500 nm. Figure 7shows an image of the cluster in OmegaCentauri taken on another commission-ing night. The e ld of view is 37 by 37arcseconds, the seeing was 0.75 arcsec-onds and the coherence time0= 3.7 ms.

The full width hal f maximum of the starsis between 0.094 and 0.11 arcsecondswith full adaptive optics correction.

The observations of Haumea andNGC 3621 with SINFONI were also com-pleted successfully. SINFONI was ableto work stably in closed loop with thelaser during standard operation tests.When observing the trans-Neptunianobject Haumea, it was possible toobserve up to the twilight limit withoutproblem. For the last two nights of theplanned commissioning period the sys-


Table 1.Measurements of the natural guide star(NGS) and laser guide star angular size.

Zenith angle

(deg)

19.419.424.636.241.754.9

LGS size

(arcsec)

1.161.281.341.591.791.62

NGS size

(arcsec)

0.550.680.660.920.880.90

Instrument

NACONACONACOSINFONISINFONI

Table 2.Science vericationtargets.

Description

Bulgeless galaxyActive galact ic nucleusGlobular cluster, Omega Centauri

Trans-Neptunian objectBulgeless galaxy

Object name

NGC 3621Centaurus ANGC 5139Haumea (TNO136108)NGC 36210.90

1000 2 000 3 000 4 000 5 000

Figure 7. Ks-band NACO image ofpart of the globular cluster OmegaCentauri. The eld of view is 37 arc-seconds square and the seeing was0.75 arcseconds. The full width halfmaximum of star images is between0.094 and 0.11 arcseconds.

Science verication

The observing targets selected for SVare given in Table 2. These were repeatsof existing and already published obser-vations.

Lewis S. et al., Laser Guide Star Facility Upgrade


11/68


After some months of operat ions withPARLA, the scientic use of the LGSFhas increased very signicantly to about35 hours per month on average (SINFONIonly). Remaining non-availability (seeFigure 8) has a < 15% impact on the totalLGSF scientic use. The reliability ofPARLA has made it possible to tackleother control issues, especially forSINFONI Multi-Applications Curvature

Adaptive Optics (MACAO). Observationswith the LGSF are now performed withsignicantly fewer overheads and ina more reliable way, providing that theatmospheric conditions are suitable.

The photon return ux, dened as thenumber of photons per unit area per sec-ond returning from the LGS, is also ofinterest over long timescales due to thelarge seasonal variability of this parame-ter (Holzlhner et al., 2010; Simonich etal., 1979). Automated logging of the num-

ber of counts on the SINFONI wavefrontsensor can be used as a relative measure,although instrumental effects precludea photometric analysis. Figure 9 showsthe mean and standard deviations of thewavefront sensor counts in a monthlytime series. These values include contri-butions from the complete observingsystem and there is not necessarily adirect correspondence with the atmos-pheric sodium column abundance.

Prospects

The LGSF at the Paranal Observatory hasbeen upgraded with a prototype lasersource based on Raman bre laser tech-nology. This is the rst time that this typeof laser has been operated as part of amajor astronomical observing facility.Results from almost one year of scienceoperation have shown that the LGSF canbe used more exibly and with signi-cantly higher availability after the upgrade.Experience with this system is providingvaluable feedback for the ESO AdaptiveOptics Facility, currently under develop-

ment, which will deploy four completeLGS units on the centrepiece of VLT UT4.

Acknowledgements

We would like to acknowledge the support ofRoberto Tamai for this project. We would also liketo thank Nadine Neumayer and Yazan Al Momanyfor processing and reducing the science images andJared ONeal for work on system performance moni-toring. We would like to thank the following compa-nies for their support: Toptica Photonics AG, MPBCommunications Inc. and Mitsubishi Cable.

References

Arsenault, R. e t al. 2006, The Messenger, 123, 6Bernard, R. 1939, ApJ, 89, 133Bonaccini Calia, D. et al. 2006, SPIE, 6272, 627207Bonaccini Calia, D. et al. 2010, SPIE, 7736, 77361U-1Holzlhner, R. et al. 2010, A&A, 510, A20Kaenders, W. G. et al. 2010, SPIE, 7736, 773621Rabien, S. et al. 2003, SPIE, 4839, 393Simonich, D., Clemesha, B. & Kirchhoff, V. 1979,

J. Geophys. Res.: Space Phys., 84, 1543Sliphe r, V. M. 1929, PASP, 41, 262

Figure 9. Monthly means and standard deviationsof the counts, recorded by the SINFONI wavefrontsensor, for the laser guide stars over the period ofthe upgraded LGSF.

Figure 8.Availab ilit yof the LGSF fo r the rs t sixmonths of science operation after the upgrade isplotted, in hours per month.

A colour image of the polar ring galaxy NGC 4650Ais shown. The image was formed by colour-coding

the MUSE spectral cube (range 48009300 )obtained during the instrument commissioning run inFebruary 2014. See Release eso1407 for moredetails.


12/68


Holographic imaging of crowded eldswas pioneered and presented by RainerSchdel (Schdel & Girard, 2012;Schdel et al., 2013). Here we demon-strate how the NACO non-AO mode,coupled with the aforementioned speckleimage processing technique, can beexploited to obtain near diffraction-limitedimages of objects of interest. We havedeveloped the required pipeline datareduction tool (Rengaswamy, Girard &Montagnier, 2010) for reconstructing animage from the speckle data1.

Imaging binary stars

The study of binary stars is an impor tantbranch of astrophysics. It is the mostcommon way by which stellar masses aredetermined. Multi-epoch images of abinary system enable the calculation ofthe orbital elements of the binary. A typicallong-exposure image does not resolvevery close binary stars, i.e., those at sub-arcsecond separation, as the resolvingpower is impaired by the atmosphericturbulence. Short-exposure images(freezing the atmosphere) can be post-processed to resolve the binary. The bot-tom left panels of Figure 1 show imagesof the binary star HIP24800, reconstructedfrom a series of 900 speckle imagesrecorded in NACO no-AO mode in Janu-ary 2010. The top panels show the bestand the worst images of the series. Thebest image is identied as the one forwhich the sum of the fourth power of themean subtracted pixel intensities is thehighest. The bottom right panel showsthe long-exposure image, a simple meanof all the short exposures. The separationbetween the components of the binary is161 milliarcseconds (mas) and their bright-ness ratio is about 2.3.

Imaging near-Earth objects

The no-AO mode of NACO also makesthe imaging of fast-moving near-Earthobjects at high angular resolution feasi-ble. The near-Earth passage of the aster-oid 2005 YU55in November 2011 pro-vided an excellent opportunity to testthis. As the adaptive optics could not lockonto this fast-moving object, data wererecorded in the no-AO mode in theKs-band. Figure 2 shows the resulting

Sridharan Rengaswamy1

Julien Girard1

Willem-Jan de Wit1

Henri Bofn1

1ESO

Long-exposure stellar images recordedwith large ground-based telescopesare blurred due to the turbulent natureof the atmosphere. The VLT employsactive and adaptive optics (AO) systemsto compensate for the deleteriouseffects of the atmosphere in real time.The speckle imaging technique providesan alternative way to achieve diffrac-tion-limited imaging by post-processinga series of short-exposure images. Theuse of speckle imaging with the no-AOmode of NACO at the VLT is demon-strated. Application of this technique isparticularly suited to theJ-band andit provides versatile high angular resolu-tion imaging under mediocre conditionsand/or in imaging extended objects.The implementation of this mode under-lines the continuing attractiveness ofNACO at the VLT.

Astronomical speckle imaging

The invention of speckle interferometry(Labeyrie, 1970) has revolutionised theeld of high-resolution imaging withground-based telescopes. In speckleinterferometry, a series of short-exposureimages of the object of interest (forexample, a spectroscopic binary star) isrecorded with a spectral lter whosebandwidth is smaller than the meanwavelength (quasi-monochromatic condi-tions). The modulus-squared Fouriertransform of the images averaged overthe ensemble (a quantity generally referredto as the energy spectrum) preservesthe information up to the diffraction limitof the telescope. Deconvolving this aver-age with a similarly obtained average ofan isolated single star, close in time andangular distance to the target, removesthe effects of the atmosphere and thetelescope on the energy spectrum, leavingbehind the energy spectrum of the objectof interest. A subsequent Fourier trans-form of the energy spectrum provides thediffraction-limited autocorrelation of the

object of interest. In the case of binarystars, the binary nature can be visuallyidentied from the peaks of the auto-correlation.

Speckle interferometry has made signi-cant contributions to the eld of binarystar research by creating several binarystar catalogues (e.g., Mason et al., 2013).Since the energy spectrum does notpreserve the Fourier phase information, itis not possible to obtain an image of theobject apart from the autocorrelation.Hence there is a 180-degree ambiguity inthe position angle of the b inary star. In1977, Gerd Weigelt showed, experimen-tally, that the phase of the complex tripleproduct of the Fourier transform of theimages at three spatial frequencies (thethird frequency being the sum of therst two) averaged over the ensemble,provides the sum of the Fourier phasesof the object at those three frequencies.

This complex product, called the bi-spectrum, is immune to atmospheric tur-bulence (as the random phase errorsintroduced by the atmosphere are com-pletely cancelled in the bi-spectrum)and provides a way to retrieve the Fourierphases of the object.

It should be noted that the phase of thebi-spectrum is the same as the closurephase, a term coined by Jennison (1958)in radio astronomy and used rst byRogstad (1968) for optical imaging throughturbulence. Combining the phase infor-mation with the energy spectrum obtainedfrom the speckle interferometry, animage of the object of interest can bereconstructed. This technique is knownas speckle imaging.

The no-AO mode of NACO

The Very Large Telescope (VLT) instru-ment NAOSCONICA (NACO; Lenzen etal., 2003) was built at the end of the1990s and its deformable mirror has atotal of 185 actuators (Rousset et al.,2003). The no-AO mode of NACO (Girardet al., 2010), essentially bypasses theadaptive optics module NAOS and usesCONICAs burst mode to record aseries of short-exposure images, exploit-ing the detectors windowing capability.

There are dif ferent ways in which theseshort-exposure images can be processed.


Speckle Imaging with VLT/NACO No-AO Mode


13/68


average reconstructed image of theasteroid. The nominal angular resolutionof 60 mas at the Ks-band corresponds toa linear resolution of 95 metres at the dis-tance (329 900 km) of the asteroid. Thedimensions of the asteroid were muchlarger than this resolution and thus thetechnique allowed the asteroid to be spa-tially resolved. The cross-sections of theimage, subjected to an edge enhance-ment operator (Prewitt edge enhancementoperator of the IDL software) to retrievethe asteroid dimensions, indicated a sizeof (261 20) (310 30) metres. Forcomparison, tri-axial diameters of 337 324 267 metres, with uncer tainties of15 metres in each dimension, were esti-mated for the size of the asteroid fromthe Keck adaptive optics system, sixhours before our observations. Account-ing for the time difference, and the18-hour rotation period of the asteroid,our results are in good agreement withthose of the Keck AO system, despitethe mediocre observing conditions (see-ing of 1.2 to 1.5 arcseconds at 550 nmand atmospheric coherence time of twomilliseconds). These observations dem-onstrate the robustness of the no-AOmode and the speckle imaging technique.

Imaging circumstellar envelopes aroundevolved stars

L2 Pup (HD 56096, HR 2748) is a semi-regular pulsating red giant star in theconstellation of Puppis with an angulardiameter of 17 milliarcseconds. Figure 3shows speckle-reconstructed imagesof L2 Pup at 2.27 m observed with the

Figure 1.Speckle reconstructed images based onNACO no-AO observations of the binar y HIP 24800inJ- (left) and Ks-(right) bands (low left panels). The

Figure 3.Speckle imagereconstruction of thecircumstellar envelopeof L2 Pup, observed inMarch and September2012. The theoretical(ideal) PSF is shown forcomparison.

Figure 2.Reconstructed image of asteroid 2005YU55(left) and its horizontal cross-secti on (right). Thered line indicates the cross-section before applyingan edge detection operator to the reconstructed

image. The green line indicates the cross-sectionafter applying the edge detection operator. The bluelines indicate the size.

top panels show the worst and best images of therecorded series. The lower right panel s show themean image.


14/68


were recorded nearly simultaneouslyalong with the speckle datacubes inthe case of the binary star shown in Fig-ure 1. Figure 5 shows good comparisonbetween the speckle-reconstructedimages obtained with the NACO no-AOmode and the corresponding AO-corrected images inJ- (left panel) andKs- (right panel) bands. The speckle-reconstructed images have been rescaled(between 0 and 255) and the contrasthas been reversed for the sake of display.Qualitatively, they match very well.

In order to have a quantitative compari-son, we estimated the Strehl ratios ofthe binary components from the recon-structed images. It should be notedthat speckle reconstruction yields theobject intensity distribution on the sky(of course sampled at the spatial sam-pling of the detector). This is quantita-tively different from the image intensitydistribution (i.e., the object convolvedwith the response of the telescope) or thepoint spread function (PSF) obtained innormal or adaptive optics imaging. Thus,

no-AO mode. The nominal resolution atthis intermediate band is about 60 mas.While the stellar continuum is unresolved,the extended circumstellar emissionfrom this star is visible in the recon-structed images. We have suppressedthe features with intensity values lessthan 5% of the maximum intensity ingenerating this gure. The circumstellarenvelope appears enlarged in the secondreconstruction, obtained six months afterthe rst one. This is perhaps due to the140-day pulsation period of the star ordue to its high mass-loss rate. The panelon the left is the ideal point spread func-tion at 2.27 m. The middle and rightpanels indicate speckle reconstructionsobtained from the March and September2012 data, respectively. These imagesserve rst and foremost as extremely val-uable information on the dynamicalevolution of the circumstellar envelope athigh angular resolution. Secondly, theycan serve as a good starting point foreven higher angular resolution investiga-tions with, for example, the Very Large

Telescope Interferometer.

Allen et al. (1972) alluded to the possiblepresence of dust shells around carbon-rich WolfRayet stars (WC stars) basedon the near-infrared excess inferred fromphotometric observations. This was ini-tially a surprising result, as the dust parti-cles would be expected to be destroyedby the strong stellar winds from thesestars. Therefore, these observations indi-cated the continuous formation of dustsomewhere within the environment ofthese stars. With the advent of aperturemasking interferometry, direct imagingof nearby WolfRayet stars became feasi-ble. A dusty pinwheel nebula structurewas discovered around two persistentlydust-forming WC stars, viz. WR 104(WC9d + B0.5V) and WR 98a (WC8-9vd)(Monnier et al., 1999; Tuthill et al., 1999).

Although these systems could be ex-plained if the dust can be formed at thedownstream regions of the shockfrontformed by the colliding winds of the WRstar and an associated secondary starof type OB, this has never been conclu-sively proven.

In an effort to clarify whether the dustformation in WC stars occurs at thecolliding wind region or in the self-shadowing regions of the clumpy winds

of an isolated star, we executed a pilotsurvey of six WC stars, consisting ofthree putative binaries and three isolatedsingle stars. Figure 4 shows the speckle-reconstructed images of WR 113 andWR 69 (supposedly binaries) and WR 15(a single star) in theJHKs-bands. Thecircumstellar envelope shows extendedemission conrming the presence ofthe dust. By resolving the dust emissionin several WC stars, at multiple epochsinJ-, H-, K- and L-bands, one can tracethe temperature of the dust distributionand thus distinguish between the twoscenarios of the dust formation.

Comparison between speckle-recon-structed and AO-corrected images

In an effort to assess the performance ofspeckle imaging, AO-corrected images

Telescopes and Instrumentation Rengaswamy S. et al., Speckle Imaging with VLT/NACO No-AO Mode

Figure 4.Reconstructed images of WR 15, WR 113and WR 69 inJ-, H- and Ks-bands. The top rowshows theoretical (ideal) point spread functions inJ-,H- and Ks-bands.


15/68


on fast-moving, extended, near-Earthobjects even under favourable conditions(since AO performs best on point sources)and speckle imaging is extremely valua-ble for imaging such objects (e.g., aster-oid 2005 YU55).

AO stands out gloriously when it isrequired to image faint targets with longexposures (deep imaging). It minimisesconfusion in crowded elds (e.g., theGalactic Centre) by separating closesources. Its use is inevitable for high-resolution spectroscopy and should bepreferred when the observing conditionsare excellent. However, there could besituations where AO cannot be usedefciently. For example, when an object isvery bright (e.g., L2 Pup with Ks= 2.3)long exposures will saturate the detectorand thus short exposures with the no-AOmode should be preferred.

Another advantage of speckle imaging isthe possibility of excluding bad imagesthat correspond to highly turbulent statesof the atmosphere. In other words, onecan select the best images from the re-corded series of images and then obtaina speckle-reconstruction from them. Thisis in line with the concept of lucky imag-ing, where typically 10% of the bestimages are used with a simple shift-and-add method to obtain a high-resolutionimage.

In our speckle processing, we use 90%of the frames when all images in therecorded sequence have similar contrast.However we discard more than 50% ofthe frames when the contrast of theframes shows signicant variations, whichoccurs particularly under mediocre con-ditions. This exibility facilitates speckleimaging under a wide variety of condi-tions. For example, the reconstructions ofL2 Pup were obtained from the datarecorded under 1.21.6 arcsecond seeingand one millisecond coherence time, afterdiscarding nearly 80% of the frames,thanks to the ability to record a few thou-sand frames. But the AO loop will notbe stable under highly varying conditions.Further, the quality of the AO correctionmay not be good under unstable atmos-pheric conditions, while it would still bepossible to obtain a speckle reconstruc-tion after excluding frames with degradedimage quality.

prior to estimating the Strehl ratios, weconvolved the speckle reconstructionswith an ideal synthesised PSF of thetelescope. We estimated the Strehl ratioas the ratio of volume under the transferfunction (Fourier transform of the PSF)of the reconstructed image (binary com-ponents in this case) to the volume underthe transfer function of an ideal (synthe-sised) PSF. Strehl ratios were obtainedwith a window size of 16 16 pixels.

The average Strehl ratios of ten consec-utive measurements inJ- and Ks-bandswere about 70% and 90 % respectivelyunder very good conditions (0.50.6 arc-seconds seeing) observing close to thezenith. We clearly see that speckle imag-ing tends to provide a higher Strehl ratio

than AO (typically 60% Strehl in Ks-bandand less at shorter wavelengths), undergood conditions. This could be interpretedas follows: as speckle image recon-struction corrects for the phase aberra-tions up to innite order (unlike adaptiveoptics imaging, which corrects the phaseaberrations up to a nite order of equiva-lent Zernike modes), we obtain higherStrehl ratios than adaptive optics imagingor normal imaging.

When to use speckle imaging with theno-AO mode

It is important to consider when to useNACO no-AO instead of AO. In general,speckle imaging with the no-AO modeis more benecial in theJ-band wherethe AO fails to perform well due to itslimited number of wavefront sensing/correcting elements. Further, it may notbe always possible to lock the AO loop

Figure 5.Comparison between speckle-recon-structed images and the images obtained with AOcorrections at the telescope inJ- (upper) and Ks-(lower) bands.


16/68


Monnier, J. D., Tuthill, P. G. & Danchi, W. C. 1999,ApJ, 525, L97

Rengaswamy, S., Girard, J. H. & Montagnier, G.2010, SPIE, 7734, 77341B

Rogstad, D. H. 1968, Appl. Opt., 7, 585Rousset, G. et al. 2003, SPIE, 4839, 140Schdel, R. & Girard, J. H. 2012, The Messenger,

150, 26Schdel, R. et al. 2013, MNRAS, 429, 1367

Tuthill, P. G., Monnier, J. D. & Danchi, W. C. 1999,Nature, 398, 487

Weigelt, G. P. 1977, Optics Communications, 21, 55

Links

1The speckle imaging data reduction tool can beobtained on request to: [email protected]

Addit ional benet of the No-AO mode

An additional use of this mode is theability to ascertain the proper tracking ofthe telescope (under wind shake, vibra-tions and occurrences of the primarymirror [M1] passive support motions). Aburst of 8000 images of a bright target,each of 20 mas exposure spans 160seconds. A software-based image regis-tration process (cross correlation usingFourier transform [FCC]) can generatea plot of image motion as a function oftime (as shown in Figure 6). Any slowlyvarying jitter with large amplitude (a fewarcseconds) could be attributed to theresidual tracking errors of the telescope.

The estimated root mean square (rms)jitter for the case shown in F igure 6 isabout 113 mas over 2.5 minutes withpeak-to-peak jitter of about 0.6 arcsec-onds. There were no passive supportmotions in M1 which could momentarilyincrease the amplitude of the jitter. Theestimated jitter is compatible with theexpected (as per design specications)rms tracking error for the VLT Unit Tele-scopes of 100 mas over 30 minutes.

Figure 6.Image jitter estimated by cross-correlatingthe speckle frames with a reference f rame. The rms

jitte r can be used to assess the qualit y of the track-ing. Wind shake and the M1 passive supportmotions can cause a momentary increase in themagnitude of the jit ter.

References

Allen, D. A., Harvey, P. M. & Swings, J. P. 1972,A&A, 20, 333

Girard, J. H. et al. 2010, SPIE, 7736, 77362NJennison, R. 1958, MNRAS, 118, 276Labeyrie, A. 1970, A&A, 6, 85Lenzen, R. et al. 2003, SPIE, 4841, 944Mason, B. D., Hartkopf, W. I. & Hurowitz, H. M. 2013,

AJ, 146, 56

Telescopes and Instrumentation Rengaswamy S. et al., Speckle Imaging with VLT/NACO No-AO Mode

Sunset over Paranal on5 July 2012, the date ofthe very low precipitablewater vapor event asdescribed in the follow-ing article. The photowas taken by ESO Photo

Ambassador GabrielBrammer, who foundthe scene to be extraor-dinarily clear and beauti-ful. Little did he knowthat the following nightwould be one of the

driest on record. SeePicture of the Week for3 February 2014 fordetails.
mailto:[email protected]:[email protected]


17/68


Florian Kerber1

Harald Kuntschner1

Richard R. Querel2,3

Mario van den Ancker1

1ESO2National Institute of Water & Atmos-

pheric Research (NIWA), Lauder,New Zealand

3Universidad de Chile, Santiago, Chile

Extremely low humidity (precipitablewater vapour [PWV] of ~ 0.1 mm) in theatmosphere above Paranal has beenmeasured by a water vapour radiometerover a period of about 12 hours. PWVvalues < 0.2 mm are usually only foundat very high altitude or in Antarctica. Infact a pocket of Antarctic air has beenshown to be responsible for this phe-nomenon and it may occur a few timesper year at Paranal. We highlight thescience opportunities created bynew atmospheric windows that arisein such conditions. The community isinvited to provide feedback on how tomake best use of low PWV with the VLT.

The dry episode over Paranal

The humidity in the atmosphere ismeasured in the form of precipitablewater vapour a measure of atmos-pheric water content. It is the amount (ordepth) of water vapour in a column ofthe atmosphere if it were all to condenseand fall as rain. The water vapour contentof the atmosphere can be measuredusing the strong water line at 183 GHz. Atthe Paranal Observatory these measure-ments are made by the Low Humidityand Temperature Proling radiometer(LHATPRO), manufactured by RadiometerPhysics GmbH1and described by Kerberet al. (2012).

Given Paranals sub-tropical location at24.5 degrees latitude south it soundsfar-fetched [pun intended] that Antarcticair would ever pass over Paranal, but thisis exactly what happened on 5 July 2012.During that night LHATPRO recordedan episode of extremely low (~ 0.1 mm)PWV that lasted for more than 12 hours.

While a beautiful panorama of Paranaltaken at sunset (Figure on p. 16) on 5 July2012 doesnt show anything unusualto the human eye, atmospheric condi-tions were in fact anything but ordinary.Located in the Atacama Desert at2635 metres above sea level, Paranalis a very dry place. Its median PWV isabout 2.4 mm with some seasonal varia-tions, but on this particular night it expe-rienced a dry episode that was trulyremarkable (Kerber et al., 2014). In Fig-ure 1 the temporal evolution of thehumidity is displayed as measured bythe LHATPRO radiometer and the twospectrographs CRIRES and X-shooterat the Very Large Telescope (VLT). Thespectroscopic observations were takenin support of routine science operationsand the PWV was deduced from ananalysis of their spectra using an atmos-pheric model (Querel et al., 2011). Suchdry conditions are more commonly ex-pected at sites at much higher altitude,such as the Atacama Large Millimeter/submillimeter Array (ALMA2) on theChajnantor Plateau (5050 metres abovesea level, median PWV 1.2 mm) or otherparticularly dry sites, such as locations in

Antarctica (the median PWV at Dome C,3233 metres above sea level, duringwinter is around 0.2 mm), which offerexcellent atmospheric transparency forinfrared and (sub-)millimetre astronomy.

Service mode and low PWV

This dry episode is the rst such event ata major observatory that has been fullydocumented in terms of atmospheric andmeteorological conditions. Other relevantambient conditions (seeing, temperature,wind, etc.) were around, or below, theirmedian values. Hence, the excellentatmospheric transparency at infraredwavelengths offered by the low humiditywould have offered ideal conditions forinfrared (IR) observations.

In this context it is worth noting that forservice-mode observations with severalIR instruments at the VLT (e.g., CRIRESand VISIR), PWV can be specied as auser-dened observing constraint. ThePWV measurements from the PWVmonitor are available in real-time in the

control room of the VLT, allowing deci-sions on which observations to conductin service mode to be taken with greatexibility. This schema allows potentiallyvery demanding observations requiringa very low (< 0.5 mm) column of watervapour to be scheduled when very dryand stable conditions occur, such as theepisode of extremely dry air described.

A careful analysis of meteoro logicalconditions involving numerical modelsshowed that on that date Antarctic airpassed over Paranal, driven far to thenorth by an unusual combination ofweather patterns; details are given inKerber et al. (2014). Is this a freak or aonce-in-ten-years event? In all likelihoodit isnt: based on archival data producedby VLT spectrographs, similarly dryconditions (PWV < 0.2 mm) occur veryrarely one or two nights per year,while PWV less than 0.5 mm is encoun-tered about 2% of the time, and lessthan 1 mm is found 15% of the time.

On account of the sparse nature of theseobservations the duration of events inthe past remains unknown. Are suchconditions also predictable in advance?Conditions for excursions of Antarcticair are certainly special, but weather fore-casting with a focus on astronomicalconditions is a eld of active research(Sarazin et al., 2013) and the predictivepower of atmospheric models is clearlyimproving.

What kind of science is enabled by lowPWV?

As already mentioned, atmospherictransparency in the infrared is dramati-cally enhanced in such dry conditionsbeyond the standard windows (J-, H-, K-,L-, M-, N-, and Q-bands) in the 125 mrange that are routinely used for ground-based astronomy. We have identied anumber of astrophysically importantlines (Kerber et al., 2014) that benet par-ticularly from such conditions.

A case in point is the hydrogen Paschen-(Pa-) line at 1875 nm, which is unobserv-able (transmission less than 2%) undermedian PWV conditions on Paranal. This


Antarctic Air Visits Paranal Opening New Science

Windows


18/68


contributions from ETH Zrich) andINAFArcetri Astrophysical Observatory.

The authors would like to encouragethe ESO community to send their feed-back on low-PWV science with the VLTand their thoughts about Pa-or otherlters in the context of instrument devel-opment or upgrades4.

Acknowledgements

Richard Querel acknowledges funding from Conicytthrough Fondecyt grant 3120150.

References

Kerber, F. et al. 2014, MNRAS, in pressKerber, F. et al. 2012, The Messenger, 148, 9McLean, I. S. et al. 2012, Proc. SPIE, 8446, 844619Motohara, K. et al. 2008, Proc. SPIE, 7014, 70142TQuere l, R. R., Naylor, D. A. & Kerber, F. 2011, PASP,

123, 222Sarazin, M., Kerber, F. & De Breuck, C. 2013,

The Messenger, 152, 17

Links

1Radiometer Physics GmbH, Germany:http://www.radiometer-physics.de

2ALMA Observatory:http://www.almaobservatory.org

3ERIS: http://www.eso.org/sci/facilities/develop/instruments/eris.html

4Comments for low PWV science wi th ESO instru-

mentation to: [email protected]

conditions on Paranal. The VLT, with itsuser-specied constraints on atmos-pheric conditions, is already prepared toaccept very demanding scientic pro-grammes in service mode that can onlybe performed during a small fractionof the time. Low-PWV science takes thisone step further, but for spectroscopicobservations. No additional investmentin hardware is required to conductlow-PWV science using existing spectro-graphs (in fact, a rst spectroscopicobservation of Pa-in a young stellarobject has already been secured undervery good but not excellent conditionsand will be the subject of a future publi-cation). This demonstrates that current

VLT spectrographs can take advantageof such conditions and deliver new andexciting science.

In terms of imaging, the VLT offers tele-scopes with an 8-metre primary mirrorand instruments with adaptive opticsproviding exceptional image quality.However, a small investment in hardwaremay be necessary to enable low-PWVscience with existing and planned imag-ers. Narrowband lters for Pa- andan off-band wavelength would have tobe procured. One imager that wouldcombine these qualities could be ERIS3,currently under design by ESO andexternal partners from the Max-PlanckInstitute for Extraterrestrial Physics (with

intrinsically strong line is prevalent inmany astrophysical sources, is lessaffected by dust extinction than H, and,due to its wavelength, benets signi-cantly from adaptive optics. At low PWVatmospheric transmission improvesrather dramatically (Figure 2), and reachesmore than 75% under the conditionsencountered on 5 July 2012. This is whysuch observations are currently eitherdone from space or at very high altitude.

The University of Tokyo operates a cam-era (Motohara et al., 2008) equippedwith Pa-lters at its 1-metre telescopeon Cerro Chajnantor (5640 metres) andrecently the First Light Infrared TestExperiment CAMera (FLITECAM) on theStratospheric Observatory for Infrared

Astronomy (SOFIA) received similar lters(McLean et al., 2012).

This of course begs the question of howsuch new science could be enabledby making use of these extraordinary


Figure 2.Transmiss ion of the Earth's atmosphere inthe region of the Pa-line at 1875.1 nm. At PWV =0.5 mm a transmission of 35% is reached. Note thatfor extragalactic sources even higher transmissionwill be achievable due to redshift and PWV as highas 1 mm may offer usable conditions (modied f romKerber et al., 2014).

Kerber F. et al., Antarctic Air Visits Paranal

Figure 1.Time series of the PWV measured aboveParanal from 56 July 2012. The data points in blackare from the LHATPRO radiometer, while PWV meas-urements derived from VLT instruments are overplot-ted. The dashed line indicates the Paranal overallnight-time median PWV (2.4 mm). Times of sunriseand sunset (see Figure on p. 16) are indicated; notethe larger variability of PWV during daytime. Insert:Enlarged view of the dry episode showing the highlevel of stability during the 12-hour period as well asthe extremely high precision of the radiometer data.

The larger number of VLT measurements taken on 6July (after MJD 56113.5) are in response to the verylow PWV recorded the previous night. Figure fromKerber et al., 2014.
http://www.radiometer-physics.de/http://www.almaobservatory.org/http://www.eso.org/sci/facilities/develop/instruments/eris.htmlhttp://www.eso.org/sci/facilities/develop/instruments/eris.htmlmailto:[email protected]:[email protected]://www.eso.org/sci/facilities/develop/instruments/eris.htmlhttp://www.eso.org/sci/facilities/develop/instruments/eris.htmlhttp://www.almaobservatory.org/http://www.radiometer-physics.de/


19/68


optimum coupling of the quasar calibratorand science target observations to pro-duce high delity images are discussedbelow.

The lling of the ALMA Quasar Catalogue

Quasar emission is produced near thenucleus of an associated massive galaxywhich may contain a black hole. Theemission is variable because of the inter-action of material with magnetic eldsnear the black hole, and often leads tothe ejection of material in narrow radio

jets. The changes in the emission fromthese objects have been widely studied(e.g., Antonucci, 1993), especia lly withvery long baseline interferometric tech-niques. But, their signicant variability inintensity is a complication in their use asamplitude (gain) calibrators. However,these quasars are sufciently far awaythat most of their emission is contained ina region that is less than 0.01 arcsecondsin projected size, so that their variabilityhas little effect on their use to monitorpath length changes.

In order to compile and store the infor-mation for hundreds of quasars, a sourcecatalogue for ALMA was implemented in2010. The initial content of the cata loguewas taken from several low frequencycatalogues from other observatorieswhich used them for similar calibrationpurposes1,2,3. However, the strength ofthese quasars at the ALMA frequencyrange of 85 to 900 GHz was unknownexcept for those monitored at these highfrequencies by the Submillimeter Array4,the Wilkinson Microwave AnisotropyProbe5and the Herschel Space Observa-tory6. On account of the great sensitivityof ALMA, a much larger calibrator data-base would be needed. With the ability of

ALMA to determine radio posit ions atthe milliarcsecond level, the quasar posi-tions were obtained from several VeryLong Baseline Catalogues7,8,9.

The special ALMA calibration observa-tions began in early 2011 when manyhundreds of sources were checked fortheir intensities. From these observationsand other observatory monitoring at orabove 85 GHz, about forty strong andrelatively stable sources, well-distributedover the sky, were chosen to be observed

Ed Fomalont1

Tim van Kempen2

Ruediger Kneissl3

Nuria Marcelino4

Denis Barkats3

Stuartt Corder1

Paulo Cortes1

Richard Hills5

Robert Lucas6

Alisdair Manning3

Alison Peck4

1National Radio Astronomy Observatory,Santiago, Chile

2Leiden University, Leiden,the Netherlands

3ESO4National Radio Astronomy Observatory,

Charlottesville, VA, USA5Cavendish Laboratory, Cambridge, UK6Institut dAstrophysique de Grenoble,

Grenoble, France

For ALMA to produce high qualityimages of astronomical objects withsub-arcsecond resolution at frequen-cies above 85 GHz, the radio signalsmust be combined from up to 66 anten-nas spread over 15 km with a maximumpath length delay difference of about0.025 mm. This accuracy requires pre-cise antenna structures, stable electron-ics, compensation for many temporalchanges in the system and the meas-urement of the path-changing watervapour emission in the line of sight. Thenal stage of path length calibration isprovided by frequent observations ofrelatively strong, point-like distant radiosources, quasars, that lie within a fewdegrees of the astronomical object. TheALMA Quasar Catalogue was imple-mented to provide a database that con-tains the essential parameters for hun-dreds of quasars and their brightnessvariations at several frequencies as afunction of time. This paper describesthe lling of the catalogue and the useof these quasar test signals to providethe path length accuracy needed for theimaging of radio sources.

The Atacama Large Mi llimeter/submil-limeter Array (ALMA) is an array of up to66 antennas, placed in congurationswith baselines (antenna-to-antenna vec-

tors) up to 16 kilometres, and located inthe Atacama Desert in Northern Chile.Each of the antennas receives the elec-tromagnetic waves from a celestialsource in the submillimetre (submm) tomillimetre (mm) wavelength range. Thesefaint signals (sensitivity of tens of Jy;1 Jansky = 1026W m2Hz1) are trans-ported to a central point where they arecombined at a virtual focus. The complexdigital processing device that focusesthe signals is the correlator. In order toproduce a high quality image, the signalsfrom each antenna must be combinedwith a phase difference that is less thanabout 0.5 radian. At a frequency of950 GHz with a wavelength of 0.3 mm,this phase difference corresponds a pathdelay accuracy of about 0.025 mm or0.08 picosecond in relative time delay.

There are several relevant time scalesassociated with the delay changes. Theshort-term path length noise (less thanone second) depends on the coherenceproperties of the ALMA electronics whichmeet the needed tolerance. The longer-term change of path delay, caused by theslowly changing properties of many

ALMA components, is in excess of thetolerance, although some aspects can bemonitored and their effects removed.Finally, the variable path delay associatedwith the propagation of the radio wavesin the atmosphere above each antennais one of the major contr ibutors to thedefocusing of the signal at the correlator,from time scales of one second to severalhours.

In order to monitor the changes in pathlength during an observation, test signalswith known properties can be propa-gated through the entire ALMA system from above each antenna to the correla-tor and the variable delay from eachantenna can be suitably adjusted (bothonline and ofine) before the image isformed.

Accurate and convenient test s ignalsare provided by point-like radio sourcesin the sky, most of them distant quasars,typically brighter than about 0.1 Jy, withaccurately known positions. A shortobservation of a quasar for a few minuteswith ALMA provides the test data neededto determine the path length adjust-ments to improve the array imaging. The


The Calibration of ALMA using Radio Sources


20/68


by ALMA at 90 GHz and 350 GHz peri-odically. This sample is called the gridsample because at least one of them isvisible at any time in the sky from ALMAand is usually within 30 degrees of anyscience target. These sources are suf-ciently strong that the calibration acrossall narrowband frequency channels canbe determined to a few percent accuracy.

Each ALMA grid session takes about40 minutes and consists of one scan ofabout eight sources. In order to deter-mine the ux density scale of the obser-vations, a scan of a Solar System objectis included. These objects have beenwell studied by many groups over theyears, so that their strength is accuratelyknown to about 5% in the ALMA fre-quency range. But, because of the highresolution of ALMA and the wanderingof these bodies around the sky, they arenot always available for use as uxdensity standards. The objects with suf-ciently accurate models, but not largerthan about 5 arcseconds in angular sizeare Mars, Uranus, Neptune, and themoons Titan, Ganymede and Callisto

(Europa and Io are often too close toJupiter), the asteroids Pallas, Juno and

Vesta and the dwarf planet Ceres.

Examples of the variability at 100 and350 GHz of four selected quasars areshown in Figure 1. These plots display therange of variability properties and werechosen because they have the mostextensive monitoring history so far. Thedata histories for these and others showthat their emission generally changessmoothly with a typical variation of 10%per month. Hence, a ux density meas-urement every two or three weeks, tied toa Solar System object, should providea ux density estimate accurate to < 10%at any time. This accuracy is sufcientfor the goals of most ALMA projects. Butoccasional outliers, which vary stronglyover less than one month (e.g., J0423-0120 and J1256-0547), do occur. Changesin a factor of two over yearly periods arecommon, and the content of the gridsource list will be modied accordingly.

The relat ively constant rat io between the100 and 350 GHz ux densities reects

the small range of quasar spectral slopearound 0.7, although ares occur atsomewhat different times at the two fre-quencies. Thus, the interpolation of aquasar ux density between 100 and350 GHz is accurate to about 10%, if thetwo ux density measurements weremade within two weeks. Extrapolation to500 and 900 GHz is somewhat moreuncertain, at the 15% level. Occasionalsimultaneous observations at threefrequencies of these quasars show thatmany have a spectral curvature between100 and 400 GHz that is sufc ientlysmall not to impact signicantly on thesimple linear spectral index extrapolationto higher frequencies. Since the goodobserving conditions necessary for> 600 GHz observations are at a pre-mium, calibrator measurements havelower priority than science observations.However, ux density estimates obtainedfrom science projects are entered intothe ALMA catalogue and provide somecheck of the extrapolation accuracy10.

Finally, catalogue data-lling observationsof weaker sources are also continuing in


Figure 1.The ux variat ions ofselected quasars are shown. The uxdensities for J0423-0120, J1256-0547(3C279), J1924-2914, J2258-2758 atBand 3 (100 GHz) in green and Band 7(350 GHz) in blue are shown in the fourpanels. The observations cover theperiod from January 2011 to January2014. The estimated uncertainties a reshown by the error bars from eachobservation. The gaps in monitoringoccur when the sources are near theSun and not monitored regularl y dur-ing the daytime.

Fomalont E. et al., The Calibration of ALMA using Radio Sources


21/68


time between the calibrator and target,generally from 20 seconds to 5 minutes.

The calibrator query for the most opti-mum phase calibrator has several criteriathat must be balanced. First, the closerthe quasar is to the science target in thesky, the more accurately it will removethe path length uctuations in the targetsource. On the other hand, the calibratormust have sufcient signal in order tomeasure these uctuations precisely andthis depends on the intensity of the cali-brator and its scan length. As shown inFigure 1, quasars often vary in ux den-sity by more than a factor of two over ayear. Since most of these fainter quasarcalibrators have limited observations,it is possible that one or more may be tooweak for use. Thus, a quick check of theux density of such quasar candidatesbefore the selection is recommended,especially for those quasars that are justabove the sensitivity limit needed for suf-ciently accurate measurements.

The switching time between calibratorand target also depends on the windspeed, the water vapour content, theobserving frequency and the size of thearray. Typically, a phase calibrator isobserved for a scan length 10 to 120 sec-onds with a repetition rate that is aboutve times the scan length. Thus, theadditional observations of calibratorsources can use up to 30% of the totalproject observing time.

The distribution of the nearly 600 quasarsmeasured with ALMA at Band 3 is givenin Figure 2. The probability of ndingone of these calibrators within a specieddistance from a random target locationis given in Figure 3. At the higher frequen-cies, the number of available calibratorsdecreases because of the poorer ALMAsensitivity and the lower ux density ofmost quasars. Although still at the experi-mental stage, it may be possible toobserve a quasar calibrator at, for exam-ple, 100 GHz, to calibrate a science tar-get at 350 or 650 GHz. This band-to-band calibration scheme will be success-ful if: (1) the phase change measured by

To accommodate the above scheme,just before the start of an observation asoftware-based calibrator query algo-rithm searches the ALMA Quasar Cata-logue in order to nd suitable quasartest sources that are needed to producegood quality images of the target. Thereare generally three kinds of test sources(calibrators) needed. If available, oneof the Solar System objects is observedfor about ten minutes in order to deter-mine the ux density scale of the obser-vation. Next, in order to determine thepath length and gain changes across therange of frequencies relevant to theastronomical observations, one of thebright grid sources is observed for up to15 minutes when narrowband channelsare used. Both of the above test obser-vations need only be done once per pro-

ject execution and that can last up toseveral hours. If the grid source ux den-sity has recently been measured anda Solar System object is not available,the grid source can be used for the uxdensity scale.

The nal calibration type, called the phasereferencing, removes path length varia-tions that occur in the atmosphere aboveeach of the antennas in the switching

order to nd quasars that are sufcientlystrong for use as phase calibrators. Atpresent there are 600 sources that aresatisfactory calibrators, but several thou-sand are needed for future ALMA ob-servations at baselines > 10 km, whentargetcalibrator separations of twodegrees are needed to remove the largedelay changes.

The ALMA quasar calibration methods

The observing schedule for an ALMAproject is produced dynamically in orderto accommodate optimally the require-ments of the project with the variableconditions. Thus, the choice of calibra-tors is often made at run time. The rele-vant parameters of the scientic targetare its position in the sky, the desiredresolution (required size of the ALMAarray) and the range of frequencies atwhich to observe. The latter depends onthe molecule(s) the astronomer wishesto study. The amount of observing timerequested depends on the image sensi-tivity that must be reached in order toachieve the proposed scientic results.

Figure 2.The distribution of ALMA Band 3 (100 GHz)calibrators are shown. Each circle represents a qua-sar with an ALMA-measured ux density at Band 3.

The largest po ints are in the 1020 Jy range; thesmallest are near 0.1 Jy.


22/68


References

Antonucci, R. 1993, ARA&A, 31, 473Nikolic, B. et al. 2008, The Messenger, 131, 14Nikolic, B. et al. 2013, A&A, 552, A104

Links

1

CRATES Flat-spectrum Catalogue: http://heasarc.gsfc.nasa.gov/W3Browse/all/crates.html2AT20G 20 GHz catalogue: http://heasarc.gsfc.

nasa.gov/W3Browse/all/at20g.html3VLA Calibrator Manual: http://www.vla.nrao.edu/

astro/calib/manual/csource.html4SMA calibrator list: http://sma1.sma.hawaii.edu/

callist/callist.html5WMAP catalogue: http://lambda.gsfc.nasa.gov/

product/map/dr3/ptsrc_catalog_get.cfm6Planck compact source catalogue:

http://www.sciops.esa.int/wikiSI/planckpla/index.php?title=Compact_Source_catalogues&instance=Planck_Public

7VLBA Sched Catalogue: http://www.aoc.nrao.edu/software/sched/Source_Catalog.html

8Petrov VLBI catalogue: http://astrogeo.org/vlbi/solutions/rfc_2013d/rfc_2013d_cat.txt

9ICRF2 catalogue: http://hpiers.obspm.fr/webiers/icrf2/icrf2

The Messenger 155

Documents