astr633 astrophysical techniques spectroscopy · 2018. 11. 1. · 2 typical slit spectrometer d 1 =...

13
1 ASTR633 Astrophysical Techniques SPECTROSCOPY (original notes by Pat Henry, edited by Mike Liu and Jonathan Williams) "If you're not doing spectroscopy, you're not doing science." Spectrometer/spectrograph = instrument with some kind of dispersing element that breaks light into its component wavelengths. Spectral analysis of celestial objects is probably the most important way of learning about them. A large fraction of telescope (and your) time is devoted to this analysis. 1. Definitions Angular dispersion (“A”) = [radians/Angstrom] Linear dispersion: linear counterpart of “A”. measure of linear separation of wavelengths, usually in a focal plane (e.g. as seen by the CCD). linear dispersion: = = [mm/Angstrom] reciprocal linear dispersion (a.k.a. the plate factor): = 1 [Angstrom/mm]

Upload: others

Post on 19-Feb-2021

1 views

Category:

Documents


0 download

TRANSCRIPT

  • 1

    ASTR633AstrophysicalTechniques

    SPECTROSCOPY(originalnotesbyPatHenry,editedbyMikeLiuandJonathanWilliams)

    "Ifyou'renotdoingspectroscopy,you'renotdoingscience."Spectrometer/spectrograph=instrumentwithsomekindofdispersingelementthatbreakslightintoitscomponentwavelengths.Spectralanalysisofcelestialobjectsisprobablythemostimportantwayoflearningaboutthem.Alargefractionoftelescope(andyour)timeisdevotedtothisanalysis.

    1. Definitions Angulardispersion(“A”)

    𝑨 = 𝒅𝜷 𝒅𝝀⁄ [radians/Angstrom]Lineardispersion:linearcounterpartof“A”.measureoflinearseparationofwavelengths,usuallyinafocalplane(e.g.asseenbytheCCD).

    lineardispersion:𝑑𝑙 𝑑𝜆 = 𝑓 𝑑𝛽 𝑑𝜆 = 𝑓𝐴⁄⁄ [mm/Angstrom] reciprocallineardispersion(a.k.a.theplatefactor):𝑃 = 1 𝑓𝐴⁄ [Angstrom/mm]

  • 2

    Typicalslitspectrometer

    d1=sizeofbeamatthecollimator(notthesizeofthecollimatoritself!)d2=sizeofbeamatthecamera.Maydifferentthand1duetodisperser.DisperserhasangulardispersionA,withdirectionofdispersionparalleltotheslitwidth(i.e.intheplaneofthediagram).Theanglesubtendedbytheslitasseenfrom3ofthesystemelements:- Asseenfromobjective:φ=w/f=angleofslitprojectedonthesky)

    where“w”isthephysicalwidthoftheslit.

    - Asseenfromthecollimator:da=w/f1

    - Asseenfromcamera:db=w’/f2 wherew’isthewidthoftheslitimageatthebackend(e.g.onthedetector).

    AnamorphicMagnification=magnificationinthedispersiondirectionIfadispersingelementisplacedbetweenthelenses,rotationalsymmetryabouttheopticalaxisislost,andtheequalityofthesubtendedanglesisnotnecessarilythesameforobjectsseenatdifferentorientations.

    Inthedirectionperpendiculartothedispersion,thebeampassingthoughthedisperserisunchanged.Inthedirectionalongthedispersionaxis,therearepossiblemagnificationeffectsduetothedisperser.Theanamorphicmagnificationis:

    MeganAnsdell

  • 3

    BytheLagrangeInvariant(H=htanuisconserved):

    Sosubstituteintoouraboveexpressionfor“r”

    i.e.ratioofthebeamatthecollimatortobeamatthecamera PhysicalSlitWidthRelatedtoSystemParameters

    fromourpreviousequation,theslitsizeasseenfromtheobjectiveis

    (whereF1isf/number,a.k.a.focalratio)ImageofSlitWidthRelatedtoSystemParametersfromourequationfortheslitsizeasseenfromthecamera:

    r � ⇤⇥⇤�

    =w�

    w

    f1f2

  • 4

    Example:Keck10-metertelescope,φ=1”Wanttobeabletomachine“w”easily,somakeit1mm. F1=w/D/φ=1e-3m/(10m)/(1”/206265”/rad)=20.6àveryslow,soeasyWant“w’“tocover2CCDpixels=44microns F2=w’/D/φ=44e-6m/(10m)/(1”/206265”/rad)=0.91àhard,butdo-ableàspectrographsonlargetelescopesneedfastcamerasi.e.forfixed(w’/φ),needfastercameraforlargertelescopes(e.g.,DEIMOSonKeck2)SpectralPurity=wavelengthchangeacrosstheimageoftheslitwidth.Bydefinitionoflineardispersion

    or

    δ𝜆 ≡ 𝑤3

    𝑓4𝐴= 𝐷𝜙𝑓4/𝑑4𝑓4𝐴

    = 𝐷𝜙𝐴𝑑4

    SpectralResolution=dimensionlessmeasureofspectralpurityThisisthemorecommonquantityusedtodescribespectrographs

    Thushighspectralresolutionrequires

    1. Smallφ(i.e.slitwidth)–butthiscanleadtolossoflight2. Largecamera–expensive(morespecifically,cameralargeenoughto

    accommodatelargebeamcomingfromthedisperser)3. Largeangulardispersion–dispersersthatdothishavelowerefficiencies

    Notethatthefundamentallimitofthisissetbydiffraction:φ=1.22λ/D

    RD=Ad2/1.22

  • 5

    2. Dispersers 2.1 Prisms

    Prism=wedge-shapedpiece,composedofmaterialwithindexofrefractionn(λ).Becauseofwavelength-dependenceofdispersion,thereissomewavelengthwheretheraysintheprismgoparalleltothebase(“minimumdeviationcondition”,i.e.anglein=angleout).Fortheserays,thediagramissymmetricabouttheverticalbisectoroftheprism,asshownabove.Thissetupisaspecificpedagogicalchoicetoshowhowtheprismworks.Ifthewavelengthoflightisdifferent,youhavetochangeθtogetthissetup.Getconstructiveinterferencebetweenthe2lightpathsabovewhen Takederivativew.r.t.lambda

    Fromthefigure,weseethat

    Sothisgives

    n(⇥)t = 2L cos �

    L sin� = a⇥ + 2� + ⇤ = ⌅, or � =

    12(⌅ � ⇤ � ⇥) ⇥ d�

    d⇥= �1

    2

  • 6

    where(t/a)istheratiooftheprismbaselengthtothebeamwidth.Indexofrefractionofmostopticalglassescanbeexpressedas Sothen

    àndecreaseswithλ,i.e.,dn/dλA<0i.e.,angulardispersiondecreaseswithlongerwavelengths.Bluelightisdeviatedmorethanredlight.àSinceA=dβ/dλisnumericallylargeratbluerwavelengths,aprismdispersesbluelightmorestronglythanredlight. Notethatthereisno(anamorphic)magnification(i.e.,r=1)becausetheincoming&outgoingbeamarethesamesize.Spectralresolutioninminimumdeviationconditionisgivenby(pluggingin“A”aboveintoourpreviousequationfor“R”):

    𝑅 = 𝜆𝛿𝜆 =

    𝜆𝐴𝑑4𝜙𝐷 =

    𝜆𝐴𝑎𝜙𝐷 =

    𝜆𝜙𝑡𝐷𝑑𝑛𝑑𝜆

    Example:ObjectiveprismonaSchmidttelescope a=D=1meter t=150mm φ=1”(typicalseeing,sincedataarecollectedslitless) [email protected]:dn/dλ=0.066µm-1 àA=0.57degs/µm àR=103i.e.notveryhighbutstilluseful(e.g.high-zQSOs,low-Zstars) Weneedmuchhigherresolution,R~104,toresolvelinesinstars.

    tdn

    d⇥= (�2a)

    ��1

    2

    ⇥d�

    d⇥⇤ A ⇥ d�

    d⇥=

    t

    a

    dn

    d⇥

  • 7

    2.2 Diffraction Gratings Grating=planar(orcurved)opticalsurfacewithaseriesoffinelyspacedgrooves(orslits).Primarydisperserinmostastronomicalapplications. Pros:SignificantlylargerRthanprismofcomparablesize. Cons:Notasefficientasprism.Limitedλcoverage(typically<2x)Gratingsworkbothintransmissionandreflection.TransmissiongratingMulti-slitversionofYoung’stwo-slitexperiment(RiekeFigure6.2):

    TheincomingbeamemergesfromtheslitsasapatternofHuygenswavefronts,spreadoverarangeofanglesduetodiffraction.Lightraysthatemergeperpendiculartotheslitsaddtogethertoproducethewhitelightpeakatm=0.Atdifferentangles,interferenceprovidesawavelengthdependenceandpeaksatm=1(firstorder),m=2(secondorder),etc.Thepathdifference,p,forlightpassingthroughneighboringslitsis

    𝑝 = 𝑑 sin 𝑖 + 𝑑 sin 𝜃Fori=0shownabove,thephasedifferenceis

    𝛿 = 2𝜋𝑝𝜆 =

    2𝜋𝑑 sin 𝜃𝜆

  • 8

    Themagnitudeoftheconstructivewavefrontdecreaseswithincreasingorderbecausetheintensityofthelightdecreaseswithdiffractionangle.Thisworkstodisperselightbutisinefficientbecausemostofthelightisinthewhitelightpeak.Bettertoreplaceslitswithmirrors.ReflectiongratingNotjustmoreefficientbuteasiertocutlinesinsolidthantocuttinyslitsthroughit(thediagramhereisfromen.wikipedia.org/wiki/Blazed_gratingbutalsoseeRiekeFigure6.3):

    Themultiplereflectionsfromthesetoftiltedmirrorsproducesaninterferencepatternthatisafunctionofλ.Thepathlengthrequirement(similartoabove)forwavestoaddinphasegivesthegratingequation:

    𝑚𝜆 = 𝑑(sin 𝛼 + sin 𝛽)Theblazeangle,θB,setsthetiltofthemirrorandischosentomaximizetheefficiencyofthegratingataparticularblazewavelength,λB.Practicallythismeansthatthediffractionandreflectionangleslineup.ThiscanbeachievedfordifferentcombinationsofanglesandwavelengthsbuttheLittrowconfigurationiswhentheincidenceangleisthesameasthereflection(i.e.,theincidentbeamisperpendiculartothetilt).Thatisα=β=θBandtherefore

    𝜃J = sinKL𝑚𝜆J2𝑑

  • 9

    AngulardispersionDifferentiatethegratingequationw.r.t.lambda,atfixedα(angleofincidence)

    Thisequationshowsthattheangulardispersioninagivenordermisafunctionofdandβ.i.e.choosedifferentgroovespacingorusegratingatdifferentangleofdiffractiontogetdifferentA.

    Canalsore-writethisas:

    SoavalueofAatagivenwavelengthissetentirelybyanglesαandβ,independentofmandd.ThusagivenAcanbeobtainedwithmanypossiblecombinationsofmandd,providedtheanglesatthegratingsareunchangedandm/d=constant.

    Thereforeabletogethighangulardispersionbymakingαandβlarge(~60degs)whenusingcoarselyruledgratings(“echelles”=Frenchfor“ladder”,lowgroovedensitybutoptimizedforuseinhighorder).

    § Ordinarygratingspectrometer:m=1or2,d=1/300–1/1200mm§ Echellegratingspectrometer:m=10-100,d=1/30mm

    Notelargeαàhighlytiltedgratingàneedalargegratingtofillthebeam.

    m d⇥ = d cos � d�

    A =d�

    d⇥=

    m

    d cos �

    A =d⇥

    d⇤=

    m

    d cos ⇥=

    sin⇥ + sin�⇤ cos ⇥

  • 10

    SpectralresolutionStartingfromourgeneralequationabove

    whereW=d2/cos(β)=physicalsizeofthegrating

    Astelescopegetslarger,somustgratingsinordertopreservespectralresolution,e.g.Keck/HIRESgratingis1.2m(actuallymosaicof3echellegratingsegments).

    Thelimitingresolutionoccursinthediffraction-limitedcase,φ=1.22λ/D:

    (middlesubstitutionbasedonouraboveequationfor“A”)

    whereN=W/D=total#ofgroovesinthegrating:moregroovesmeanmoreinterferenceandfinerwavelengthdiscrimination.Example:considersamesizegratingasourobjectiveprismexample,W=1meter. SamesizetelescopeD=1m Gratinggroovespacing:d=1/1200mm φ=1” λ=0.5um α=β=17.5deg

    àmuchlargerthanprismFreeSpectralRangeForagivenpairof{α,β},thegratingequationissatisfiedforallλforwhich“m”isaninteger.Overlapwhen𝑚𝜆 = (𝑚 + 1)𝜆′

    ð Freespectralrangeis∆𝜆 = 𝜆3 − 𝜆 = 𝜆/𝑚

    e.g.m=1,λ’=0.8µm,thenλ=0.4μmandΔλ=0.4μmThereare2waystosortoutoverlappingorders:

    (1) Ordinarygratingm=1andΔλislargeàuseafilterEchellegratingm=10andΔλissmallàaddacross-disperser.i.e.aprismorgratingtoseparatethepiecesintheperpendiculardirection.e.g.IRTF/Spexcross-disperseddata

  • 11

    3. Practical Considerations for Actually Obtaining Useful Data Slitsize§ Thisisoneoftwothingstheastronomerhascontrolover,theotherbeingthegrating.§ Narrowslitsyieldimprovedspectralresolution.Thathelpssolongasthespectral

    featuresareresolved.§ However,don’tmaketheslitwidth~2”)sowantslitsoforderthatsizetogetallthelightinthegalaxy.Besides,thelinesingalaxiesarenotnarrowanyways,duetostellarmotions,sonoadvantageinfinalspectralresolutionbyusinganarrowslit.

    Gratings§ Makesuretopickappropriateblazewavelengthforobjects.§ Trytogetefficiencycurve,sincesomegratingsarebetterthanothers.§ Caneasilyloseafactorof2inlightinthegrating.Multi-slitspectroscopy§ Canget~10-100’sofobjectsinasingleexposure,bymillinglittleslitsalloverthefocal

    plane.(~1000sifusingfiberspectrograph)§ Makesureslitlengthis>~10”,inordertogetenoughblankskydatatomakeagood

    determinationofitslevel.WavelengthCalibrationLamps§ Takearcspectratogetwavelengthcalibration,i.e.matchobservedarcemissionlines

    withlistofknownwavelengths.§ Frequencyoftakingarcimagesdependsonflexurepropertiesofspectrograph(e.g.at

    CassorNasymthfocus).Forbestpractices,ReadtheManual(RTM).Flats§ Notasimportantaswithimages,sinceonlytoflattentheregionnearthespectrum,not

    overthewholeimage§ Noneedtomatchthespectrumofthetargetandtheflat,becausethespectrographputs

    thesameλonthesamepixelinbothcases.Justneedsomefluxatallλ.§ Makesurenottosaturatethedetector.§ Formulti-slits,needaflatforeveryindividualmask,b/cmappingofλtopixelsis

    differentforeachmask.§ Forsingleslits,frequencyofflats(e.g.everyimage,everytarget,oreverynight)

    dependsonphysicalstabilityoftheinstrument.RTFM.Spectrophotometricstandards§ Thesearestarsofknownspectraltype,socanusethisknowledgetocorrectforallthe

    wavelength-dependenteffects(atmosphere,telescope,optics,grating,detector)andrecoverthetruespectrumofthesciencetarget.

  • 12

    § Trytousestandardswithoutstrongabsorptionlinesinthem.e.g.A0stars.Bestspectraltypesdependonwavelength.

    § Intheoptical,makesureyoucantakelong(>~15sec)exposures.Someshuttersarenon-linear.(Notrelevantforthenear-IR,sincenoshuttersareinvolved–arraysarecontinuallyreadout.)

    4. Reduction of Data Reducingspectroscopydataisbasicallythesameasphotometry,justdoingitinonespatialdimensionandhundredsoftimesinthedispersiondirection.The“images”areofteninthebackground-limitedregime,soaccurateskysubtractionisveryimportant.

    1.Dobiasremovalandflat-fieldingintheusualway.2a.Tomeasurethesciencetarget,foreverycolumn:

    § Collapsetheobjectspectrumtoonenumberbyaveraging,sincestellarprofileisonly3-4pixelswide

    § Measuretheskyspectrumbymedian-averagingovertheremainingpixels.Want(#ofskypixels)>>(#ofobjectpixels)sothatPoissonnoiseinthemeasurementoftheaverageskyleveldonotdominate.

    § Subtractthe2numberstoyieldthenetDNperpixel=DNnet(x)2b.Dothesameforstandardstardata:DNstd(x)3.Todeterminethewavelengthcalibration:

    § Identifypairsof(pixel,λ)inthearcspectrum.

  • 13

    § Usethesetoestablishthemappingfrompixel->λusingasimplefunction(e.g.linearorlow-orderpolynomialfit,b/cspectrographsarereasonablylinear).

    4.Applywavelengthcalibrationtogetnetfluxa.f.o.λ:DNnet(λ),DNstd(λ)5.Thenthefinalfullycalibratedmeasurementforthesciencetargetis

    Sinceweknowthetruephysicalspectrumofthestandardstar,thisallowsustoremoveallthewavelengthdependenteffects(e.g.atmosphere,telescope,grating/blaze,etc.)Bewaryofdetectedspectral“features”nearstrongnightskylines.