Imaging Polarimetry with the Gemini Planet ImagerMarshall D. Perrin1, James R. Graham2, James E. Larkin1, Sloane Wictorowicz2, Jerome Maire4, Simon Thibault, Michael P. Fitzgerald3, Rene Doyon4, Bruce Macintosh3, Donald T. Gavel5, Ben R. Oppenheimer7, David W. Palmer3, Leslie Saddlemyer8, J. Kent Wallace6 and the Gemini Planet Imager team

1: UCLA, 2: UC Berkeley, 3: Lawrence Livermore National Lab, !4: !Universite de Montreal, 5: UC Santa Cruz, 6: JPL, 7: American Museum of Natural History, 8: NRC/Herzberg !Institute

Apai et al. 2004, A&A 415 p. 671Boccas et al. 2004 SPIE 5494 p. 239Graham 2006 (GPI OCDD, Gemini)Macintosh et al. 2006, SPIE 6272Macintosh et al. 2008, SPIE 7015

Maire et al. 2010, SPIEPerrin et al. 2004 , Science v303 p1345Perrin, et al., 2008, PASP p. 555Zuckerman et al. 2001 ARAA 39 p.549Zuckerman et al. 2004 ApJ 603 p. 738

References

28-04-2008 GPI_Volume_2A.doc Page 1 of 139

!

!"#$%$&'()%"*&+#),"-&

.!'+/!

!

!'+&012&34(5#"&67&

!

!

"##$%#&%#'

!!

!

()*+,-./!"0"!

122*-3)4!56/!! ! ! ! ! ! 789)/! ! !

::::5*;<)!=8<,.9-+>::::::::::::::!::::::::::::!!!:"##$%#&%#'!!!!!?! @A! ! B!

::::783)!@8CD)*::::::::::::::::::::::::::::::!!!:"##$%#&%#'!!!!!?!@=! ! B!

::::E)+C,)!F844C)D6)*:::::::::::::::::!!!:::::::!!!:"##$%#&%#':::!!FG! ! B!

!

Instrumental Polarization

The Gemini telescopes use protected silver coatings (Boccas et al. 2004) for all mirrors, including the 45º-incidence tertiary fold that likely dominates instrumental polarization. We have developed a high-fidelity model of the polarization properties of protected silver (Graham 2008), and find that a 45º reflection from protected silver induces a small linear polarization and phase shift: 0.99% and !164.5º at 1µm. This is almost as good as bare gold. This polarization changes slowly with wavelength and will be straightforward ot calibrate. Empirical measurements will be made during system integration and on the sky.

We have i nve s t i ga t ed t he systematic errors caused by instrumental polarization. By calibrating the instrumental polarization (of the Gemini telescope plus GPI AO system) with an accuracy of at least <0.3% in degree and <20º in phase, we can ensure that final uncertainties are not dominated by systematics.

– 37 –

Fig. 9.— Errors in recovered polarization signal as a function of calibration uncertaintiesas in Fig. 2, except the instrumental polarization due to a single 45! protected Al mirror

with 0.4% linear polarization and ! = !150! (see Figs. 8 and 7). The linear polarizationand retardance are calibrated to the accuracy indicated on the x- and y-axes respectively.RMS errors in Q/I and U/I are shown in the top two panels. Errors in the degree of linear

polarization and position angle are shown in the bottom panels.

Unc

erta

inty

in In

stru

men

tal

Ret

arda

nce

[deg

]

Uncertainty in Instrumental Linear Polarization [%]

RMS Error in Measured Linear Polarization [%]

– 37 –

Fig. 9.— Errors in recovered polarization signal as a function of calibration uncertaintiesas in Fig. 2, except the instrumental polarization due to a single 45! protected Al mirror

with 0.4% linear polarization and ! = !150! (see Figs. 8 and 7). The linear polarizationand retardance are calibrated to the accuracy indicated on the x- and y-axes respectively.RMS errors in Q/I and U/I are shown in the top two panels. Errors in the degree of linear

polarization and position angle are shown in the bottom panels.

RMS Error in Measured Position Angle [deg]

Above: Errors in recovered polarization signal as a function of calibration uncertainties, derived from Monte Carlo observation models.

10-7 10-6 10-5 10-4 10-3 10-2

Optical Depth !

0

5

10

15

20

Num

ber

Sola

r zod

i lev

el

Spitzer FEPS debris disks

IRAS-detected debris disks

debris disks seen inscattered light

!"#$%&'()*+,-.$/)012'$ 3&.,4)$56$7(1)'()$8&+19-+1&'$

!"#$%&'()%#*"+,"&(,"%,)-).)-/(-$(0$1(,2*3+#"(4*)56-%",,()%#*"+,",(-6"(&"-"#-+4)0)-/($3(3+#"7$%(&),8,9(:"-+(

;)#(),($%"($3(-6"(<$,-()%-"%,"0/()%.",-)5+-"&(,-+*,()%(-6"(,8/(="959'(>2#8"*<+%(?@@AB'(4"#+2,"($3()-,(

C*$<)%"%-():2);&'(&"4*),(&),89(!-*2#-2*"(1)-6)%(-6"(!(;)#(&),8(),(*",$0.+40"(+-(<)&7DE(1+."0"%5-6,($%(

-6"(0+*5",-(5*$2%&74+,"&(-"0",#$C",(=F+66+G("-(+09(?@@H'(F")%4"*5"*("-(+09(?@@HB9(I),8(,-*2#-2*"(#+%(4"(

C*$4"&(#0$,"*(-$(-6"(,-+*(1)-6(<)&7DE()<+5)%5(-6+%(1)-6(,#+--"*"&70)56-()<+5)%5'("."%(1)-6(J!K'(

4"#+2,"($3(-6"(<$*"(3+.$*+40"(&),87-$7,-+*(4*)56-%",,(*+-)$()%(-6"(<)&7DE9((L25"*"+2("-(+09(=?@@AB(<$&"0(

-6"(5*+)%(C*$C"*-)",(+%&(#$%#02&"(-6+-(C+*"%-(4$&)",(3$*(-6"(&2,-(+*"(0$#+-"&(4"-1""%(?@(+%&(AM@(LN(

1)-6(+(C"+8(+-(AA@(LN9(F")%4"*5"*("-(+09(=?@@HB($4-+)%"&(,C+-)+00/(*",$0."&(,C"#-*$,#$C/(-6+-(*"."+0"&(

".)&"%#"(3$*(+<$*C6$2,(+%&(#*/,-+00)%"(,)0)#+-",(1)-6)%(?@(LN($3(!(;)#9(J$1"."*'(4"#+2,"(-6"(!(;)#(&),8(

),("&5"($%'(1"(6+."(%$()%3$*<+-)$%($%(-6"(+O)<2-6+0(,-*2#-2*"(%""&"&(-$(C0+#"(-6"(,/,-"<(1)-6)%(-6"(

,#6"<"($3(P)52*"(AM9((((

(

!"#$%&'()*'+%&,-'./'-&0%"1'-"12'.34"567'-&348'9"48'6#&':6-634&-'/%.;'<$52&%;6,'=>>(?@'!"77&-'146%1'",-"564&'1A14&;1'

67%&6-A'";6#&-'0A'.34"567'.%',&6%BCD'5.%.,6#%6381@'E3&,'5"%57&1'%&3%&1&,4',&6%0A'57$14&%1'1$58'61'48&'F7&"6-&1@'+8&'

3.",4'/.%'48&'G$,'"1'061&-'.,'48&'H$"3&%'0&74'-$14';611'",/&%%&-'/%.;';.-&71@''C,5%&61",#'5.,4%614'6,-'5.%%&13.,-",#'

1&,1"4"I"4A'4.'1;677&%'6,-'1;677&%'.34"567'-&3481'&J4&,-1'48&'6#&'%6,#&'655&11"07&'4.'15644&%&-'7"#84'";6#",#@'+8&'-&0%"1'

-"121'1$%%.$,-",#'KL'(>)(MN':O%-"76'&4'67@'=>>M?'6,-'!.;6786$4':"'F1O?':H6761'&4'67@'=>>P?'9&%&'/"%14'1&&,'",'

15644&%&-'7"#84'",'=>>M'0&56$1&'./'48&'1&,1"4"I"4A'#6",1'3%.I"-&-'0A'48&'KG+'OQG'5.%.,6#%638@''Q$%%&,47AR'.,7A'6'

86,-/$7'./'48&'8"#8&14'.34"567'-&348R'A.$,#&14'6,-'&-#&B.,'1A14&;1'56,'0&'";6#&-'9"48'#%.$,-B061&-'1A14&;1@'+8$1R'

5$%%&,4'8"#8'5.,4%614'";6#",#'/65"7"4"&1'3%.0&'.,7A'.34"567'-&3481'./'6'/&9'(>BM@''

K6"($C-)#+0(C*$C"*-)",($3(&2,-(5*+)%,()%(&),8,(/)"0&(+&&)-)$%+0()%3$*<+-)$%(*"5+*&)%5(-6")*(,)O",(+%&(

#$<C$,)-)$%'(+%&(6"%#"(-6")*($*)5)%(="959'(I2#6Q%"("-(+09(?@@RB9(P$*("S+<C0"'(-6"(<$+%&(=(4+%&(#$0$*,($3(

-6"(LN(T)#(&"4*),(&),8'(-$5"-6"*(1)-6(-6"(,247<<(!UI(,6$1,(+(C*$%$2%#"&(0+#8($3(,<+00(=V(@9H(W<B(

5*+)%,()%(-6"()%%"*(&),8(=T"-#6".("-(+09(?@@RX(P)-O5"*+0&("-(+09(?@@YB9((!<+00(C+*-)#0",(<+/(4"(*"<$."&(4/(

#$+520+-)$%()%-$(0+*5"*(C+*-)#0",(=5*+)%(5*$1-6B'(&",-*2#-)$%(4/(;$/%-)%57E$4"*-,$%(+%&Z$*(#$*C2,#20+*(

&*+5'($*(*+&)+-)$%(C*",,2*"(40$1$2-9((J$1"."*'(-6"(LN(T)#(&),8(4"#$<",()%#*"+,)%50/(402"(+-(0+*5"*(

*+&))'(+-(0"+,-(4"-1""%(H@[\@(LN(=P)52*"(A]B9(K6"(,247<)#*$%(5*+)%,(,#+--"*)%5(.),)40"(0)56-(+-(1)&"(

,"C+*+-)$%,(=16"*"(-6"(#$00),)$%(+%&(;$/%-)%57E$4"*-,$%(-)<"(,#+0",(+*"(0$%5"*(-6+%(-6"(+5"($3(-6"(,-+*B(

<+/(4"(C*)<$*&)+0'(*+-6"*(-6+%(4")%5(40$1%($2-(3*$<(-6"()%%"*(&),89(K6","(5*+)%,(<+/(4"(+(*"<%+%-(3*$<(

8->$?@$ABBC$ "-2)$AD$&E$FD$

(

Disk Science with GPI

The majority of nearby disks have optical depths < 5e-4 (Zuckerman & Song 2004) yet only a handful of the highest optical depth systems have been imaged. Improved contrast is needed to enlarge the set of objects which can be resolved, allowing detections of fainter, older, and more face-on disks. GPI polarimetry will enable studies of protoplanetary and debris disks with optical depths as low as 4!10-5 (about 1/50 the optical depth of AU Mic) on scales of 0.15-2.0”. These data will provide insights into disk architectures, tidal interactions between disks and planets, and the formation and evolution of planetary systems.

Above left: Disk optical depths (as measured by IR excess) tend to decrease with stellar age. (Adapted from Zuckerman 2001). Center: Currently, only a handful of the highest-optical-depth systems have been imaged with adaptive optics or space coronagraphy. Right: GPI’s small inner working angle will enable observations of disk structure on scales comparable to the solar system for targets at ~100-150 pc distance. This figure shows simulated GPI observations of the inner regions of the transitional disk around HD 141569A. The inner working angle is ~ 3x better than HST NICMOS.

GPISensitivity

L54 IMAGING OF HD 141569 DISK Vol. 525

Fig. 1.—Left: Image of the reflected light disk around HD 141569 shown in a stretch that is linear with flux density. The eastern side of the disk is brighterbecause of the intrinsic scattering phase function of the dust. Right: Deprojected image of the disk using an inclination angle of 51!. Artifacts from small instabilitiesin the PSF subtraction process include the radial “streamers” and the dark “finger” in the inner disk at the 4 o’clock position. These residuals do not significantlyaffect the radial profile or the integrated flux density. The diffraction spikes narrow to higher radius as the overlap between the two sets of observations at differentorientations increases and allows pixels behind the spikes in one image to be replaced by pixels from the other image.

Fig. 2.—Azimuthally averaged surface brightness profile of the disk afterthe subtraction of three different PSF stars. The error bars represent only thestatistical uncertainty in the mean surface brightness at each radius. At radiigreater than !160 AU, the three profiles are consistent to within these un-certainties. The subtractions with t1 Eri and 49 Cet suggest that the disk fluxfalls off substantially inside of 160 AU.

applied to the final F110W images in which 1 ADU s =!1

Jy and 0 mag = 1775 Jy (M. Rieke 1999, private!62.031# 10communication).In the reduced and calibrated NICMOS images, scattered

light from the star still dominates the flux levels at radii of1!–5!, so in order to see the disk, a PSF star must be subtracted.The same observing strategy and the F110W filter were usedfor 14 other stars in this GTO program, and targets that do notshow extended disk emission form a library of possible PSF

stars. The detailed character of the NICMOS PSF changes withtime primarily because of thermal variations in the HST opticalassembly (Kulkarni et al. 1999); hence, some PSF star sub-tractions produce much lower residuals than others. Each PSFimage was registered using cubic convolution interpolation tothe images of HD 141569 at each spacecraft orientation. Onlystars brighter than HD 141569 were considered, and the PSFsubtraction residuals are dominated by systematic uncertainties.Scattered light from two nearby stars outside the field of

view (see § 5.1) pollutes the northwest corner of the image.For both roll angles, we used synthetic PSFs generated withTiny Tim (Krist & Hook 1997) and positioned them at thelocations of the two stars in order to subtract this excess scatter.A small amount of residual flux, however, can still be seen inFigure 1.Figure 2 compares the azimuthally averaged disk flux that

was observed after the subtraction of each of the three best-matched PSF stars. Two of these PSF stars, t1 Eridani and49 Ceti, were disk candidates in which no disk was detected,and the third, HR 4748, was observed specifically as a PSFstar. At radii greater than 1!.6 (160 AU), the morphology ofthe disk is consistent in all three subtractions. The subtractionof 49 Cet yielded the smallest amplitude residuals, and allfurther analysis was carried out on images using it as the PSF.The PSF-subtracted images from each telescope roll were

rotated to a common orientation and averaged. Regions of theimage that were obscured by the diffraction spikes in one rollwere replaced where possible by uncorrupted regions in theother roll.

4. RESULTS

The PSF-subtracted, roll-combined coronagraphic image ofHD 141569 is shown in the left panel of Figure 1. An elongateddisk is clearly evident at radii from !1!.6 to 4! (160–400 AU)from the star. Elliptical isophotes were fitted to the disk image,yielding a position angle of and a ratio of the minor356! " 5!to major axes of . If, as is likely for dust orbiting0.63" 0.04

HD 141569A

9” FOV

outer image:HST NICMOS F110W

Weinberger et al. 1999

GPI simH band3000 s

GPI is a high-contrast AO system being built for the Gemini Observatory (Macintosh et al. 2006, 2008). GPI combines

• high order AO wavefront control• coronagraphic diffraction suppression• precision IR wavefront calibration• a speckle-suppressing science camera

GPI is nearing completion of construction, and first light is planned for 2011.

GPI's " primary science goal is the direct imaging and characterization of young, Jovian-mass exoplanets in thermal emission. GPI also seeks to image scattered light from debris disks with an optical depths down to 4#10-5. To achieve this, GPI will include an imaging polarimetry mode that suppresses unpolarized speckles by a factor of >100.

!

!

!

!

"#! $%&! '()'&*(&#$! +,-$-./0! 1&'-2#! 3%/'&4! $%-'! /#/05'-'! 6/'! ,&30/.&7! 6-$%! /! '83%-'$-./$&7! 83$-.9)5983$-.! :,&'#&0!

'-;(0/$-8#<!=&'(0$'!>,8;!$%-'!/,&!3,&'&#$&7!)&086!?@&.$-8#!AB!/#7!7&'.,-)&7!&C$&#'-D&05!-#!/!.8;3/#-8#!3/3&,E<!

!"! #$%&'()*$&+,-*'-#*.+/$0+1#23&+4/&3+

FG"!.8#'-'$'!8>! '-C!;/H8,! '()'5'$&;'<!I%&!/7/3$-D&!83$-.'! ?JKB! '5'$&;!.8#$,80'! $%&!6/D&>,8#$! &,,8,'! -#7(.&7!)5! $%&!

/$;8'3%&,&4! $&0&'.83&4!/#7!FG"L'!86#!83$-.'<!I%&!.8,8#/2,/3%!;/'M'!.8#$,80! $%&!7->>,/.$-8#!8>!.8%&,&#$!'$/,0-2%$<!I%&!

./0-),/$-8#!-#$&,>&,8;&$&,!?+JNB!-'!8#&!8>!$%&!;8'$!(#-*(&!>&/$(,&'!O!/!38'$9.8,8#/2,/3%-.!6/D&>,8#$!'&#'8,4!-$!&C308-$'!

$%&!83$-./0!3,83&,$-&'!8>! $%&!.8,8#/2,/3%-.!3,8.&''! $8!'&#'&!$%&!6/D&>,8#$!/$!$%&!>8./0!30/#&!8..(0$8,4! $%&!M&5!08./$-8#<!

I%&! .,582&#-.! "#$&2,/0! :-&07! @3&.$,82,/3%! ?":@B! -'! $%&! 8#05! '.-&#.&! -#'$,(;&#$<! :-#/0054! /#! 83$89;&.%/#-./0!

'(3&,'$,(.$(,&! ?KP@@B! '(338,$'! /00! $%&! '()'5'$&;'!/#7! .8;38#&#$'<! I8390&D&0! '8>$6/,&Q! .88,7-#/$&'! $%&!83&,/$-8#!8>!

$%&!-#'$,(;&#$!/#7!-#$&,>/.&'!$8!$%&!8)'&,D/$8,5<!!

:-2(,&!R!'%86'!$%&!0-2%$!3/$%<!N-2%$!&#$&,'!$%&!-#'$,(;&#$!$%,8(2%!/!6-#786S!$%&!-#'$,(;&#$!&#.08'(,&!-'!8D&,3,&''(,-T&7!

6-$%!7,5!.0&/#!/-,!$8!3,8$&.$!83$-./0!'(,>/.&'<!J!7&3085/)0&!/$;8'3%&,-.!7-'3&,'-8#!.8,,&.$8,!/0-2#'!7->>&,&#$!6/D&0&#2$%'!

8>!0-2%$!-#!)8$%!$%&!.8,8#/2,/3%-.!>8./0!30/#&!/#7!?&*(/005!.,(.-/005B!8#!/00!$%&!FG"!-#$&,#/0!83$-.'4!'8!$%/$!$%&!0-2%$!'&&#!

)5!6/D&>,8#$!'&#'8,'!/.*(-,&'!$%&!'/;&!/)&,,/$-8#'!/'!$%&!0-2%$!-#!$%&!'.-&#.&!3/$%<!J!D/,-&$5!8>!./0-),/$-8#!'8(,.&'!./#!

)&!-#H&.$&7!-#$8!$%&!-#3($!>8.('<!I8!/.%-&D&!2887!.8,8#/2,/3%-.!-;/2&'4!$%&'&!0-2%$!'8(,.&'!;('$!3,87(.&!D&,5!(#->8,;!

-00(;-#/$-8#!8>!$%&!3(3-04!6%-.%!6-00!,&*(-,&!'&#7-#2!$%&!0-2%$!$%,8(2%!/!D&,5!';/00!3-#%80&!/#7!;/''-D&05!8D&,>-00-#2!$%&!

FG"!83$-.'<!J'!/!,&'(0$4!6&!,&*(-,&!/#!&C$,&;&05!),-2%$! 0-2%$!'(.%!/'!/!'(3&,.8#$-#((;!'8(,.&!6-$%!/!386&,!8>!'&D&,/0!

;6!-#!/!'-#20&9;87&!>-)&,<!!

!

:-2(,&!RU!FG"!@.%&;/$-.!7&'-2#<!

Above: A schematic layout of GPI showing the major subsystems.

Right: This Wollaston p o l a r i z i n g p r i s m mounted within the IR I F S e n a b l e s t h e polarization analysis mode, along with an achromatic half wave p la te and ro ta t ion mechanism mounted in the Cal WFS assembly. Prism Slide

Mechanism

WollastonPrism

Gemini Planet Imager Overview

Integrated End-to-end Modeling

PSF Sim: H band, 7th mag

IFS/Polarimeter Data Simulations

To verify the expected performance of GPI, we have carried out extensive modeling. Our detailed simulations of the AO control loop model multi-layer atmospheres, woofer/tweeter correction, and the full GPI AOC algorithm set to evaluate the effects of residual speckles on timescales of up to 30 seconds. Our static wave-optics model then uses a Fresnel optics code to propagate phase and intensity from arbitrary surfaces within the system, to evaluate the contribution of quasi-static components to the long-exposure PSF.

The derived PSF then is sent through our spectrograph/polarimeter data simulator, which propagates each individual lenslet’s light along the full optical path of the IFS through the dispersing elements and onto the detector, including effects such as sky rotation, instrumental polarization and modulation, and detector flat fielding, read noise, and photon noise. This simulator allows arbitrarily complicated scenes to be modeled, with any number of planetary companions, scattered light debris disks, and/or background stars.

Lastly, we reduce the data through our draft data reduction pipeline (see Maire et al., this meeting) to extract spectral and polarimetric datacubes, including speckle-suppression algorithms. These simulations verify that GPI will be achieve contrasts for planetary companions of < 10-7 in 1 hour (5 sigma) in spectral mode, or to resolve scattered light from debris disks with optical depths as faint as a few x 10-5 in polarimetric mode.

Above: Simulated H band GPI PSF. The dark hole is 1.8” across. Below: Simulated raw data from the IFS, showing individually dispersed polarization pairs from 40,000 lenslets.

Below: Simulated 1-hour H-band observation of an I=6 star at 20 pc. Differential polarimetry suppresses the stellar PSF, enabling the detection of the faint inner “asteroid belt” at r=6 AU.

Total Intensity Polarized Intensity

GPI remains stationary with respect to the telescope pupil—so the sky appears to constantly rotate. This allows angular differential imaging to reduce the residual speckle halo. However, it complicates polarimetry because each exposure has unique sky-projected polarization axes and rotation. The standard double-differencing data reduction will not suffice. Instead we adopt a forward-modeling approach to data reduction: Using the known instrumental response Mueller matrix for all exposure in a sequence, we derive the modulation matrix for the observation sequence. We then solve the resulting matrix equation via least-squares to derive the astronomical polarization for each pixel.

22.5 pixels

7 pixels

>41,000 lenslets, >82,000 spots

Lenslet Rows

Lens

let C

olum

ns

2 Polarizations for each lenslet

7 pixels

GPI’s science instrument is a lenslet-based near-infrared imaging spectrograph (shown at right). In spectral mode, the dispersed spectra have R~45 at H band, with a spatial sampling of 0.014”/pixel across a 2.8”x2.8” field of view. In dual channel polarimetry mode, the dispersing prism is replaced with a Wollaston prism, gaining sensitivity to orthogonal linear polarizations (though at reduced spectral resolution).

The polarizations are dispersed only after the lenslet array has pixellated the field of view, as shown below. This novel “integral field polarimetry” design minimizes differential wavefront error between the two polarizations, eliminates the need for precise image registration which has limited previous AO imaging polarimeters (e.g. Apai et al. 2004, Perrin et al. 2004, 2008), and does not require any reduction in field of view. It also prevents any distortion of the PSF due to lateral chromatism of the Wollaston prism, thus

removing any need for exotic materials with low birefringent chromaticity like YLF.

Linear Stokes polarimetry is achieved by modulation with a rotating half-wave plate. Because coronagraphic observations are more sensitive to wavefront errors prior to the occulting mask, the waveplate is mounted after the AO and coronagraphic optics. Hence the instrumental polarization is not modulated, and must instead be calibrated via observations of standards, and removed during data reduction. (see discussion at right.)

Above: The GPI IFS, shown with the dewar cover removed. The light path is shown in blue.

Left: Schematic of the polarimetry mode data format,showing individually dispersed polarizations for each lenslet.

IFS Dewar

H2RG

Prism Slide

Lenslet Array

Integral Field Polarimetry