No. 73 - September 1993

A Message trom the New EditorIt is an honour and a pleasure to take on the responsibility of the Editorship of THE MESSENGER. The former editors, and

in particular Richard West who was editor for two terms (1976-1979 and 1986-1993), made the Messenger an informativeand friendly publication. It is therefore clear that no large changes should be brought to the Messenger.

However, because of the forthcoming deployment of the VLT- VLTI, the ESO observers' community has an increasing needfor technical information in all domains: site infrastructure, telescopes, instrumentation and detectors, calibration, dataanalysis and archiving. More emphasis will be given to these topics concerning both Paranal and La Silla. The Messenger willalso carry articles on the use of the VLT-VLTI for specific astronomical topics. These articles are written byastronomersinvited by the Editor to present their views. As before, observational results obtained at La Silla will also be presented. In thisarea, there will be no attempt at presenting a complete panorama of the observations carried out at ESO, and the number ofarticles in the observations section is likely to fluctuate from issue to issue. The mostapparent change in the look of the journalis a division into sections: first the telescopes and instrumentation including detectors. This will be followed byarticles onscience with the VLT. Then there will be a section on scientific results obtained at La Silla and a section on other astronomicalnews. The last section groups all the announcements of ESO conferences and workshops, some announcements forpositions (mostly in the Science Division), the tit/es of the accepted programmes and similar matters.

The technical editor of the Messenger remains Kurt Kjär and I look forward to our collaboration in producing theMessenger. Marie-Helene Ulrich

email Internet: mulrich eso.org

TELESCOPES AND INSTRUMENTATION

The 8.2-m Primary Mirrors of the VLTR. MUELLER, H. HOENESS, Schott Glaswerke, Mainz, Germany

J. ESPIARO, J. PASERI, REOSC, Longjumeau, France

P OIERlCKX, ESO

1. Introduction

The 8.2-m Zerodur primary mirrorsof the VLT telescopes are on their wayto the observatory. On June 25, the firstof the four mirror blanks (code nameJoe Dalton) was delivered by SCHOn,

the mirror blank manufacturer, toREOSC, who will polish it and mount itsinterfaces. Inspections performed be­fore delivery have not only demonstra­ted the compliance of the blank withits specifications, but also the perfor­mance of the blank manufacturer in

achieving superb quality.On July 19 at 14:20, Joe was trans­

ported to Mainz harbour, a few kilome­tres away from the SCHOn plant (Fig­ure 1). At 9 p. m. it was loaded onto thebarge ELDOR. The ELDOR went downthe Rhine to Rotterdam, travelled south

to Galais, then up the Seine to Evry aftercrossing Paris on Friday 23 overnight.On Monday 26 at 10 a. m. it was un­loaded and placed onto a trailer. Theroad transport to the REOSG plantstarted at 0:00 on Tuesday 27 and Joearrived at REOSG at 02:30 a. m. Thecontainer was opened the same day at14:00'.

One of the greatest difficulties in thedesign of telescopes larger than about 401' 5 metres is the manufacturing of theprimary mirror blank. The classical de­sign of a thick and solid mirror wouldimply enormous masses. First of a longseries of problems, the homogeneity ofa 200-ton casting is very unlikely tosatisfy the stringent requirements ap­plied to high accuracy optical compo­nents. Even under the unrealisticassumption that all manufacturing andassembling issues could be solved, an8-m mirror more than 1 m thick wouldstill deform too much under its ownweight. Remember the basic idea: get asurface the size of a small appartmentdown to sub-micron accuracy, andmaintain it.

There are three possible diets to getthe mass down: reduce its weight, makeit out of segments, 01' get it slim and useits support to control its shape. Thethree names behind these solutions are:Palomar, Keck, ND, respectively. How­ever, in the 8-m range, a Palomar-typeprimary mirror still needs some activecontrol as gravity and thermal loads mayaffect the surface accuracy. The effec­tiveness (with regard to performanceand safety) of very large honeycombstructures is still to be proven, while thesecond 10-m Keck telescope is on itsway, and the ND has already providedsuperb images. Each of the three solu­tions has its pros and cons, and thetrade-off is extremely difficult. There areno miracles, just tremendous efforts tobring the concepts into operation.

2. The Actively ControlledMeniscus Concept

The concept selected by ESO for theVLT is an extrapolation of the conceptdemonstrated first with an experimentalmirror 1 m in diameter and subsequentlywith the ND. It consists of a f/1.8, 8.2-mdiameter actively supported Zerodurmeniscus, 175 mm thick. The mirrorswill be supported axially by 150 hy­draulic actuators. These actuatorscan be seen as computer-controlledsprings: they do not constrain the spa-

I On ESO side the whole process has been exten­sively recorded by Messrs. Heyer, Madsen andZodet. They shall be acknowledged for their ex­ceptional dedication and efficiency.

2

Figure 1: Joe on its way to Mainz harbour.

tial position of the mirror at 150 loca­tions but aim at providing a force dis­tribution which results in a fully predict­able elastic deformation of the mirror.With an active mirror, flexibility is usedto compensate for manufacturing errorand for on-line, real-time optimization ofthe optical properties of the telescope.

The hydraulic support system is di­vided into three sectors, each of themproviding a virtual fixed point. Therefore,the axial position and tilt of the mirror isfixed. The actuators interface to themirror through axial pads, the geometryof which is a three-Ieg spider. Each legis connected to an invar pad about50 mm in diameter, which is glued ontothe convex surface. The distribution ofthe supports has been optimized to pro­vide the lowest possible print-throughon the mirror's surface. Additionally, thetripods introduce shear forces which re­duce local deformations. The residualerror is in the ,J40 rms range, and will bealmost totally polished off as the mirrorsare figured on their tripods. (See section4.3, Effectiveness of the FLIP method.)

Ouring operation the exact shape ofthe mirror will be monitored by Shack­Hartmann sensors. These sensors col­lect the light of an off-axis star to forman image of the pupil of the telescopeonto a grid of lenses. A GGO camerarecords the distribution of the point im­ages provided by the individual micro­Ienses. A phase error of the incomingwavefront would imply a variable incli­nation of the rays with respect to theaxis of each microlens thus resulting in adeviation from the nominal distribution

of spot images on the GGO. This devia­tion is measured and the wavefront re­constructed. The necessary correctionis translated into a force distributionwhich is applied to the mirror through itsactive support. With active optics thefrequency of correction is below about1Hz. With its six rings of support the VLTshould be able to compensate errorswith spatial frequencies up to about1 m-" within an accuracy in the range of50 to 100 nm rms, i.e. weil beyond theeffect of atmospheric turbulence.

On the user's side active opticsshould be fully automatic, i.e. the onlynoticeable effect should be the highquality of the data collected under per­manent computerized optimization. Ofcourse this situation is quite demandingin terms of tuning and reliability, and theconcept inevitably requires permanentmaintenance, as any large and complextelescope. On the other side the positiveaspects of the active optics conceptextend far beyond the potential excel­lence of the optical properties. Indeed itprovides far more design and manufac­turing perspectives than a passive con­cept, with positive consequences bothon performance and cost.

The fragility of the thin meniscus re­quires particular attention. All supportsand auxiliary equipment which will inter­face to the mirror at one stage 01' theother are designed to keep the probabil­ity of damage under about 10-5 undermaximum loads. Nevertheless, giventhe high flexibility of the mirrors and thehighly non-linear relation between dam­age probability and load parameters

The issue of matching is less criticalas weil, as the active support can handlea fairly large error of the conic constant.Although the option of a pentaprism teston the primary and secondary mirrorscombination can still be ordered, thecurrent strategy for the VLT is to haveseparate cross-checks performed onthe primary and secondary mirrors.These are direct tests, i.e. they do notinvolve relay lenses which could intro­duce systematic errors, and they aim atensuring that no significant matchingerror is left.

As for any optical component, opticaltesting is essential as it should ensurethat the manufacturing tolerances arefulfilled, and as a proper measurementof the surface data (radius of curvatureand conic constant) is requested for thecalibration of the field aberrations. In­deed, the latter have to be deductedfrom the off-axis measurement by thewavefront sensors, in order that the ac­tive system applies the appropriatecorrections.

3. Mirror Blanks

Large-diameter and medium-thick­ness menisci are economically manu­factured by using the spin casting tech­nique. This technique was specificallydevelopped by SCHOn in 1986-1987for the production of mirror blanks of the8-m class. Contrary to the conventionalcasting technique, spin casting is ableto produce a meniscus which ap­proaches the final shape of the mirrorwith only small oversizing, thus reducedmass. The meniscus shape is obtainedby rotating the mold (which has a spher­ical bottom) while the glass is not yetsolidified.

The spin casting technique hasproven successful in the production ofseveral raw blanks of the 8-m class.

._-.'

.......~..l!<%~///

~

"found another bubble?::J

specified in such a way that after con­volution with the atmospheric seeingtheir effect on the long-exposure imagewill be limited to a maximum decreaseof 5 % of the peak intensity in stellarimages, in the visible and with a seeingangle of 0.4 arcseconds full width at halfmaximum. It can be easily shown thatthe specification ensures simultaneous­Iy seeing-limited angular resolution andhigh signal throughput and signal-to­noise ratio.

A positive consequence of the activeoptics concept is that it allows themanufacturer to concentrate on highspatial frequency errors, which haveusually lower amplitude but much higherslopes and are therefore more detrimen­tal to performance than the low spatialfrequency ones. In other words, an ac­tive mirror is likely to be smoother than apassive one, and therefore to providebetter contrast and signal-to-noise ratio.

Table 1.

ParameterSpecifica- Achieve-

tions ment

Outer diameter 8200 ± 2 8201.52 mmInner diameter 1000 ± 0.5 999.81 mmConcentricity 51 0.01 mmThickness 1 176 + 2-0 177.9 mmRadius of curvature (concave surface) 28975 289682 mm

(convex surface) 28977 289793 mmProfile tolerance (concave surface) 2 0.124 mm

(convex surface) 2 0.055 mmResidual stresses (tensile) 50.10 max.0.07 MPa

(compressive) 50.60 max.0.08 MPaCoefficient of thermal expansion (mean value) o ± 0.15 -0.043 10-6o K- 1

(homogeneity) 50.05 0.009 1O-6O K- 1

I Normal to convex suriace, on a diameter of 8160 mm. - 2 Best fitting sphere. - 3 Best fitting sphere. -• Including the eHect of the 7mm error on the radius of curvature of the concave suriace. - 5 Including theeHect of the 2mm error on the radius of curvature of the concex suriace.

--

--'- I?'I'

~-------/"-/./ ~./"

./-----

(e.g. area of stress fields, magnitude ofstress, duration, surface finish), trans­port, handling and operation of anylarge mirror have to follow strict andrigorous rules.

The mirror itself is made out of ahighly homogeneous Zerodur sub­strate, with near-zero residual stresses.Table 1 summarizes the main specifica­tions for the mirror blanks, togetherwith the achievement by SCHOn forthe first mirror blank. In many respectsthe mirror blank is of exceptionalquality.

In addition to the geometrical andmaterial properties, the internal qualitywas specified in terms of maximumnumber of bubbles and inclusions.Again, the result is weil within thespecification. The convex surface hastwo holes (to be compared to 10allowed), maximum diameter 34.8 mmand maximum depth 40 mm, which re­sult from the removal of inclusions be­fore ceramization. Local acid etchinghas been applied and the effect of theholes on stress fields is such that theprobability of breakage is still beyond10-5 .

The optical specifications given to theoptical manufacturer are in line with theactive optics concept and follow theformalism developed at ESO to quantifyoptical performance of ground-basedtelescopes.

The mirrors will be polished on anactive support system which interfacesto the tripods. The manufacturer is per­mitted to physically remove low spatialfrequency errors, provided that he stayswithin the specified budget of ± 120Newton maximum active force. The op­erational performance of the mirrors isgiven by the residual errors after activecorrection. The latter have been

3

Figura 2: Lifting the overhead heating system.

After the resolution of start-up prob­lems, several castings could be suc­cessfully cooled down to room tempera­ture. The subsequent quality inspec­tions showed that the high requirementsset to the inner quality of ZERODUR arecompletely fulfilled.

All subsequent operations like han­dling, ceramization and machining werecarried out without any problems. Thefirst blank is already delivered, and thesecond blank is currently in the processof being transformed into glass ceram­ics in the ceramization furnace.

For the spin casting production ofmirror blanks of the 8-m class, allfacilities and the respective buildingswere built between June 1989 andMarch 1991. The facilities are entirelydedicated to the manufacture ofZERODUR blanks (mirror substrates ofthe 8-m class and smaller blanks of0.6 m up to 4.4 m in diameter). Produc­tion structures include new buildingswith about 50,000 m3 of enclosedspace, and modified and enlarged build­ings.

The main steps of themanufacturing process are:

• Melting of glass in 70-t melting tank(discontinuous)

4

• Pouring of molten glass into the cast­ing mold (which has a curved bot­tom). and generation of meniscus byspinning

• Coarse annealing of vitreous blank toambient temperature

• Machining (rough grinding of frontside, rear and edge)

• Transformation into glass ceramic(ceramization) by thermal retreatment

• Drilling of centre hole• Final grinding

The handling and quality controlswhich are necessary between the indi­vidual production steps are not men­tioned. In the following paragraphs theproduction process is described insome more detail.

• Me/tingFor this purpose, the batch is fed into

the tank and melted, refined, and subse­quently homogenized by stirring. Amelting cycle lasts 24 days.

• Pouring

The casting mold is preheated forabout 2 days, with the overhead heatingsystem being lowered. Following this,the casting mold is lifted to allow thefeeder (platinum tube for feeding the

molten glass from the tank into the cast­ing mold) to protrude downward fromthe bottom of the mold.

After heating the feeder nose with thehelp of proper devices, the molten glassflows from the melting tank onto a con­veyer.

Then the mold is progressively low­ered to keep the feeder nose at aspecified distance of the bottom of thecasting mold. The discharged glass flowis cut by an appropriate device whichallows the casting mold to be filled whileit is slowly being lowered. This fillingoperation lasts approximately 4 hours.

After completing the filling operation,the mold is moved from the casting areaonto the spinning area with the help ofthe transfer vehicle.

• Spinning

Once the transfer vehicle has beenfixed to the spinning unit along with thecasting mold, the overhead heating sys­tem is fastened to and slightly lifted bysome hoisting equipment from the cast­ing mold (Figure 2). This is followed byspinning the casting mold containingthe molten glass at a specified rotationrate. A cooling cap is placed above themold after the concave upper surface ofthe glass has been produced. The re­leased radiation energy (20 MW im­mediately after lifting the heating sys­tem) is dissipated through the evapora­tion of water. During the spinning phase,the molten glass cools down in such away that the glass surface retains itscontour. The spinning process lasts ap­proximately 1 hour.

• Annea/ing

After the spinning process, the cast­ing mold with the raw blank is enteredinto the annealing furnace. Subsequent­Iy, the transfer vehicle is moved out andthe furnace is closed. This is followed byannealing the vitreous-state raw blankto ambient temperature. During this pro­cess, the blank is supported by anadaptable support system which followsthe deformation and shrinking of theconvex surface through the annealingprocess. The annealing period iasts ap­proximately 3 months.

• Lifting and turn-over•whem annealing has been completed,

the annealing furnace is opened, afterwhich the raw blank is lifted from thecasting mold with a vacuum lifter andmoved to the turning position by thecrane equipment. There it is turned over(convex surface up) with the help of theturn-over installation (Figure 3).

Figure 3: Turning the mirrar blank.

During all handling operations, defor­mations of the handling equipmentshould not be transferred to the blank.Therefore the handling equipment musthave a flexible and adaptable couplingto the mirror blank. For this reason hy­draulic cylinders are used. All supportsystems are designed in such a way thattensile stresses induced in the blank arekept weil below 1 MPa.

• Rough machining

After turning over the raw blank, it isplaced on the support system of theCNC grinding machine. This supportsystem has a design similar to the de­sign of the handling equipment. Theblank convex surface and edge areroughly machined with diamond grind­ing wheels. This is followed by turning

over the blank again and machining thefront side.

• Transformation to a semi­crystalline state (ceramization)

After completing the rough machiningthe blank is submitted to a thermal after­treatment in the annealing furnace,which causes the vitreous state to betransformed into a semi-crystallinestate. This procedure causes the mate­rial to acquire its outstanding propertiesof thermal zero expansion.

This crystallization process is exo­thermic. During crystallization, the vol­ume shrinks by approximately 3 %.Hence, all efforts should be made tocause crystallization to proceed veryslowly, in order to prevent stresses fromcoming up and leading to the fracture ofthe disk due to non-controlled crystalli­zation and resultant non-uniform shrink­age. This is why transformation from thevitreous to the semi-crystalline statetakes approximately 8 months.

During the transformation process,the blank is supported by an adaptablesupport system within the annealing fur­nace.

• Drilling of the centre hole andfinal grinding

After transformation into a glass

Figure 4: The blank during quality inspection.

5

ceramic, the centre hole is drilled withthe CNC grinding machine.

The final grinding is also performed onthe machine, the accuracy of which is asfoliows:Rotating table at 7000 mm diameter:

radial (x) runout < 10 ~lm

axial (z) runout < 15 ~lm

angular positioning < 120 ~lm

Positioning in x and z: < 20 ~lm

Reproducibility in X and Z < 2 /olmTo preserve this accuracy the

machine raom has a temperature con­tral giving

spatial temperature homogeneity< ± 1.5°C

temporal temperature inertia< ± 1SC/h.

• Qua/ity contro/s

Quality controls and acceptance bycustomers are provided during and be­tween the individual production steps(Figure 4).

Material characteristics are deter­mined by the corporate Service RE­SEARCH & OEVELOPMENT. The pro­jecHied inspection means consist of:overall geometric testing system(theodolite with pracess contraI compu­ter), local geometric testing system, ul­trasonic thickness gauge system, stressmeasuring system, video system fortesting and documenting the inner quali­ty. For stress measurements the blank isput onto a special quality controlsupport system that generates extreme­Iy low stresses in the blank.

• Transfer to transportation system

After final acceptance, the blank isplaced onto the transportation systemwith the vacuum lifter and the crane. Thetransportation system (REOSC) isassembled beforehand in the produc­tion wing.

4. Mirror Figuring and Polishing

On July 25, 1989, REOSC obtainedfrom ESO the contract for the polishingof the four VLT mirrors and the followingtasks:• design and manufacture of the mirrar

handling tool and container,• transportation of each VLT mirror

blank fram SCHOn's plant in Mainz(Germany) to REOSC's plant,

• design, manufacture and assemblyon the mirror of special devices forthe mirror fixation in the cell,

• grinding, polishing and testing of thefour VLT mirrors.One of the governing design drivers in

this project was the very important flexi­bility of the VLT mirrors as they feature ahuge diameter of 8.2 m associated to avery low thickness of only 17.5 cm.

6

Figure 5: The REOSC optical shop.

4.1 THE OPTICAL SHOP

The new 8-m REOSC shop (Figure 5)was dedicated by the French Minister ofResearch and Space on April 24, 1992.

This plant is located close to theSeine, in order to optimize the mirrortransport. lts total surface is 1100 m2

and its dimensions are: length 70 m,width at the tower base 22 m, width ofthe shop 12.6 m, testing tower height32 m.

4.2 OVERVIEW ON THE TECHNICALPROCESS FOR FIGURING THE VLTMIRRORS

Oue to their huge size and high flexi­bility, the conventional technique of thefull sized flexible tool previously used forpolishing and figuring has been mod­ified in order to obtain a mirrar surfacefree of high frequency defects and com­pliant with the requested asphericalshape.

Thus, the VLT mirrors will be groundand polished by using tools of adequatesizes associated with the REOSC Com­puter Controlled Surfacing (CCS) Pra­cess, and axially sustained by a supportwhich is accurate enough not to blur themirrar high frequency defects with alarge amount of non-axisymmetric lowfrequency defects.

Grinding machine

The grinding machine is composed ofa ratating table, a removable millingbridge, a couple of two-arm machinesdesigned and manufactured accordingto the REOSC requirements and a sim­plified axial support composed of 150pneumatic actuators.

All this assembly is computer con­trolled through a unique console.

Po/ishing machine

The polishing machine (Figure 6) isidentical to the grinding one but it is notpermanently equipped with the millingbridge. Furthermore, this machine, 10­cated just under the testing tower, isequipped with an axial support of 150pneumatic actuators, each of them iscomputer controlled through a ShackHartmann CCO camera. The force accu­racy delivered by each actuator is betterthan 2N and this support will be able toremove physically a large amount of lowfrequency defects.

The suppport software enables theuser to adjust automatically the mirrorposition and its trim thanks to threelength sensors. This software featuresalso the mirror weighting, forces ex­certed by each actuator, mirrar position,etc....

4.3 OVERVIEW OF THE TESTINGMETHODS

The testing faci/ities: the tower

As the efficiency of the optical testsdepends on the vibration level, thermalstability and on the easiness to passfrom a polishing run to a testing one,REOSC has given particular care in de­signing the testing tower which featuresan external sun shield, a main wall madeof boarding with a thick lagging, an in­ternal tower totally independent framthe external one, made of a weldedmetallic structure which bears a doublewalled plastic tent. Thermally contralIedair is inflected between the two walls ofthis tent.

ManufacturingTest

Number Absolute Sensitive- Towerstage of points accuracy ness PIV requested

Rough grinding Spherometry 198 5000 nm 1000nm NoFine grinding IR tnterferometry 2000 500nm 300nm YesPolishing, final figuring FLIP method 250x250 5nm 5nm Yes

Figure 6: The 8.2-m dummy mirror on the polishing machine, as seen trom the test tower.

Furthermore, in the testing tower,REOSC has foreseen the possibility totest the matching of a convex mirrorassociated with a concave mirror by us­ing the Pentaprism method as alreadyperformed for the ESO 3.6-m telescope.

The testing plan

Ouring the whoie surfacing process ofthe VLT mirrors, the progress of thework is checked by the testing methodssummarized in Table 2.

The mirror aspherical shape will begenerated on the mirror at the roughgrinding stage and checked byspherometry. Then the mirror will bemoved to the polishing machine locatedunder the tower. Ouring the fine grind­ing, the mirror surface is tested by IRinterferometry. When the mirror surfaceis smooth enough and about a few mi­crons from the requested mirror shape,the polishing and the visible interfero­metric testing will start.

The spherometer

The spherometer body is made of analuminium honeycomb structure. Thereare 8 points of measure and about 200points of the mirror surface can beknown. The estimated time for perform­ing this measure and processing thedata is about half an hour. Thespherometer is carried by the millingbridge and its position and the positionof the rotating table are controlled on­line by a computerto which are input themirror sag measurements.

Visible interferometry and the FlowInterferogram Processing (FLIP)

The residual vibrations of the tower donot allow the use of the conventionalphase shift method. The phase methodbased on Fourier transforms and theworks of Claude and Franyois Roddierlead to a too long processing time foraveraging a large number of interfero­grams.

In order to overcome this difficulty,REOSC has developed a new algorithmwhich allows quick computing of thephase from only ONE interferogramwhich must present about 80 fringesinclined at 45° if a CCO camera of250 x 250 pixels is used for an 8-mmirror. Furthermore, as an image can be

Table2

caught in a thousandth of a secondthanks to the use of an electronic shut­ter, the disturbing effect of the vibrationsare eliminated.

FLIP is a special software which canbe used with any interferometer pro­vided it is equipped with a CCO cameraand an electronic shutter. REOSC thinksthat the FLIP method has the bestmoney value.

The interferograms are taken bybatches of 50 units and processed in1 minute for 250 x 250 points with ourpresent PC 386 equipped with a framegraber card, an accelerator card and acoprocessor. By adding two acceleratorcards, the processing time can be di­vided by two. The residual vibrations ofthe tower combined with the averagingof the results allow the elimination ofirregularities of illumination.

This software delivers the X and Yslope error map and the values of theRMS and PN errors, the wavefront errormap and the values of the RMS and PN,as weil as the mirror surface profilewhen the non-axisymmetric defectshave been removed.

The programme is user friendly and itis now currently used in REOSC's opti­cal shop.

Effectiveness ofthe FLIP method

In order to test the effectiveness ofthe VLT axial support when it is provided

with special tripods as interface be­tween the actuator shaft head and themirror, REOSC has figured a Zerodurspherical mirror of 1.7 m in diameter,17.5 cm thick and with a radius of curva­ture of 28.8 m, the same as the VLT one.This mirror is in fact apart of a wholeVLT mirror. For the experiment, this testmirror was resting on seven VLT ac­tuators, each of them was provided witha tripod. Furthermore, some of themwere equipped with a large dish in orderto compare the residual high frequencydefects between two kinds of supportfor polishing: on the tripod and on thedishes.

Since the testing tower of the newplant was not available at the date ofthis experiment, the 1.7-m mirror andits support were installed under theREOSC old testing tower of the 4-mmirror shop. The light beam was foldedtwice. The results obtained with this set­up and the FLIP method are thefollowing:

When the mirror is resting on 7 singlepoints, the measured and computedmirror deformations are as shown inTable 3.

There is a perfect correlation betweenthe experiment and the FE computation.

In order to evaluate the FLIP ability todetect minute defects among largerones, we have compared the resultsobtained by finite elements computationand FLIP measurements when themirror is resting on the tripods.

The mirror figure obtained by interfer­ometry is the averaging of a batch of 50interferograms taken by the FLIPmethod to which we have substractedthe 1.7-m mirror figure (see Table 4).

The conclusion is that the finite ele­ments results are very close to the ex­perimental ones. The difference in the

7

Interferometry Computation

PIV RMS PIV RMS

Mirror surface 148nm 24 nm 107nm 25nm

Interferometry Computation

PIV RMS PIV RMS

Mirror surface 59nm 7.3 nm 37nm 8nm

Table3

PN values comes from the presence inthe laser beam of some artefacts whichgenerate a few local sharp defects.

It is worth noting that the RMS valueof the high frequency defects generatedby the axial support is three times lowerwhen the mirror is resting on the tripods.This factor of three is in very goodagreement with the computation.

In order to evaluate the noise of themeasurements. we have computed thedifference between the average of three

batches of fifty interferograms of themirror resting on the tripods. The RMSerror on the mirror surface error is5.6 nm.

Table4.

5. Conclusion

Joe Dalton was delivered in time andweil within specifications. It is now Iyingon the grinding machine at REOSC,where everything is ready for the nextsteps: glueing of the axial interfaces (inAugust 1993). followed by the excitingtasks of grinding it aspherical to a fewmicrons accuracy. and then polishing tooptical accuracy. Looking forward tomeeting Bill, William and Averell!

Ground-Based Astronomy in the 10 and 20 !lmAtmospheric Windows at ESO - Scientific Potential atPresent and in the FutureH. u. KÄUFL, ESO

1. Introduction

This article is especially written forthose who have little or no experiencewith astronomical observations in the in­frared, especially longwards of 2.2 ~lm.

In recent issues of the Messenger(Käufl et al. 1992, Käufl 1993) therewere several reports on instrumentationat ESO for observations in the 10 and20 ~m atmospheric windows (N and Qband in Johnson's photometrie system).In this article a description of the per­spectives of ground-based astronomyin these windows will be given. A specialfocus is set on the astronomical applica­tions with respect to what is possible(and what not) with the present equip­ment; which improvements are antici­pated at the VLT and how these obser­vations compare to air-borne andspace-based infrared astronomy.

2. Targets tor Ground-basedIntrared Observations atWavelengths longer than 5 f-lm 1

2.1 Atmospheric constraints

The terrestrial atmosphere has a sub­stantial opacity in the infrared due to

1 Some of the argumentation holds also for therange of 3-5flm. This transition region betweenthe near infrared and the thermal infrared, how­ever. is outside the scope of this article.

8

rotational-vibrational transitions of tracemoleeules (e.g. CO2, H20, CH4 , 0 3), Insome spectral regions these moleculartransitions blend creating a practicallyopaque sky. The region between 8 and13 ~m, however, is reasonably free ofinterfering molecular transitions. Re­maining absorption lines in the windowitself and at the "red" edge are ratherstable and observations can be easilycorrected for these perturbations. Onthe "blue" edge H20 is a source of vari­able opacity following the rapid varia­tions of local humidity. Close to thisedge more careful observing proce­dures are required. Beyond = 13.3 ~lm

the atmosphere becomes opaque dueto the \/2-band of CO2 and opens againaround 16.5 ~m. A forest of lines of H20.however. prevents the sky from gettingclear so that even in the best part of the20 ~m window the average opacity moreor less corresponds to 50 % and is vari­able. Beyond 20 ~m the water vapourlines become a real problem. The term"Q-window" is misleading and shouldbe replaced by "Q-venetian-blind". De­pending on site and local weather, lim­ited astronomical observations are pos­sible up to 30-35 ~m. Beyond =35 ~mthe atmosphere remains opaque untilthe sub-millimetre region (~ 300~lm) isreached.

In conclusion, long wavelength in-

frared astronomical observations areconstrained to:• A=8-13 ~lm with very good condi­

tions• A= 16.5-30~m with reasonable to

poor conditions

2.2 Wien's Law, Kirchhoff's Lawand their consequences

Wien's law relates the maximum ofthe Planck curve for black body radia­tion with its temperature

A = 2898!lmmax T[K]

For a telescope at ambient tempera­ture )'max lies exactly in the centre of the10 ~lm atmospheric window. Translatedinto the words of optical astronomy thisis equivalent to observing stars fromwithin a furnace (Iike the one in whichthe VLT blanks have been casted) ratherthan from a dark astronomical site.

To those readers not familiar with in­frared astronomy it is also useful to re­call Kirchhoff's law

a(A, T) = e(}., T)

which states that for all wavelengths andtemperatures the emissivity equals ex­actly the absorptivity. For ground-basedinfrared astronomy this means that e.g. a