Integral Field Spectroscopy

Jeremy Allington-SmithUniversity of Durham

Contents

• Advantages of Integral Field Spectroscopy• Datacube "theorem"• Techniques of IFS• Lenslet-array• Fibres+lenslets• Image-slicing• Multiple IFS

What is IFS?

• Integral field spectroscopy produces a spectrum of each part of an image simultaneously

• This results in a datacube with axes (x, y,• This is sometimes called "3D imaging" or "2D

spectroscopy" or even "3D spectroscopy"!• 3D techniques which also produce a datacube

but not from a single observation (e.g Fabry-Perot or FTS) are not usually called IFS

Direct image Radial velocity Close up

SAURON: NGC 4365 (Lyon/Durham/Leiden/ESO)

Why use IFS?

"Boring" elliptical galaxy with odd kinematics!

Where do you put the slit?

• Slit gives only a 1D slice through object• Slit captures only part of the object's light• Only a 3D technique reveals the global velocity field

IFS – use info from adjacent slicesto correct velocity data

Slit spectroscopy – velocities in error since blobs not centred in slit

dispersion

Generic advantage of IFS

• Spectroscopy over full 2D field with high filling factor• No slit losses - all the light is used• Point and shoot target acquisition reduces operational

overheads• Can reconstruct white-light image to aid interpretation (and

target acquisition)• Almost immune to atmospheric dispersion • More accurate radial velocity determination:

– Obtain global velocity field - not just a 1-D section– Velocity field can be reconstructed accurately without errors due

to position of features within slit

Applications

• Galaxy kinematics: stars and gas (em & abs lines)• Distribution of ionising radiation (line ratios)• Distribution of stellar populations

(lines/continuum)• Studies of interacting galaxies (kinematic

resolution)• Unbiassed searches for primaeval line-emitting

galaxies (may be invisible in broadband image)• Searches for damped Ly aborbers near line of

sight to QSOs (with large impact parameter)• Outflows from young stellar objects

Dissecting active galaxies

NGC4151 observed with SMIRFS-IFU in J-band - Turner et al. MNRAS 331, 284 (2002)

Distribution of [FeII]

Velocity field (narrow Pa)

Datacube "theorem"

To first order… all 3D methods are equally efficient in generating the same datacube volume with the same number of pixels

x

y

Datacube with same equivalent volume Nnm

N observationseach withn x m pixels

Spectral and spatial informationencoded on detector in any way you like

Imaging spectroscopy E.g. Fabry-Perot interferometry & narrow-band imaging

Devote pixels entirely to imaging:

Datacube sliced into thin slices in wavelength.

Repeat observations with different wavelength range

Sensitive to changes in sky background

Each slice contains the fullfield imaged in one passband

x

y

Longslit spectroscopy

Longslit spectroscopy:

Each longslit pointing produces a x slice

Full datacube produced by stepping longslit in y

Each slice is one longslit spectrum

x

y

NB: No spatial information in y within each slice

Integral field spectroscopy

Devote pixels mostly to spectroscopy:

datacube sliced into narrow spatial fields - repeat observation with different pointings

Each piece contains all thespectra within a narrow field

x

y

... to second order?

• Which technique wins depends mostly on:– the dominant noise source

• detector read noise• detector dark current• photon noise from sky• photon noise from object• temporal variability in sky background

– how many pixels you can afford– details of the scientifc application, especially:

• the size of the total field required• the length of the total spectrum required

• A tradeoff between FTS and IFS for NGST/IFMOS indicated that IFS was preferrable

IFS "efficiency"

Aim is to maximise a figure of merit that is a function of: # spatial samples , # spectral samples , throughput

# spatial samples: pack spectra together tightly along slit. Overlaps will result between samples at the slit but this is okay if:– there is Nyquist sampling of the field at the IFU input– adjacent spectra come from adjacent elements on the sky– there is no wavelength offset between adjacent spectra

# spectral samples: maximise length of spectrum to fill complete detector length but, for a given detector, (#spatial #spectral) constant so can have multiple slits to increase #spatial by reducing #spectral

throughput: efficient design

Make the best possible use of the available detector pixels by minimising the dead space between spectra

Techniques of IFS

Lenslets

Fibres+lenslets

Imageslicer

Telescopefocus

Spectrographinput

Spectrographoutput

Pupilimagery

Fibres

Mirrors

slit

slit

1 2 3 4

1

2

3

4

x

y

Datacube

Both designs maximise the spectrum length and allows more efficient utilisation of detector surface.

Only the image slicer retains spatial information within each slice/sample high information density in datacube

Like SAURON and OASIS. Overlaps must be avoided low information density

in datacube

Lenslet IFU

• Example: SAURON* designed for wide-field galaxy kinematics• Short wavelength range for low-redshift MgB (517.4nm)• Spectra must not overlap otherwise information lost

Sauron built by CRAL (Lyon)

*Bacon et al. MNRAS 326, 23-35 (2001)

Lenslet+fibres: optical principle

Microlens array

Pickoff

mirror

Enlarger

fibre

slit

Spectrograph

grating

Fibre bundle

Slit (outof page)

Telescope focus

skyimage

pupil

image

fibre

GMOS-IFUAllington-Smith et alPASP 114, 892 (2002)

Input

x y

x

y

Original image

Alli

ngto

n-S

mit

h &

Conte

nt,

PA

SP 1

10

,12

16

(1

99

9)

Pseudo-slit

x y

Ensure critical sampling

here!

Fibre+lenslet detection process

x

Detector

y

x

monochromatic image of pseudo-slit

x

yreconstructedmonochromaticimage of sky

Computer

y’

Overlaps here don't

matter

GMOS• 0.07 arcsec/pixel image scale

• 5.5 x 5.5 arcmin field

• 0.4 - 1.1m wavelength coverage

• R = 10,000 with 0.25” slits

• Multiobject mode using slit masks

• Integral field spectroscopy mode

• Active control of flexure

GMOSwithout enclosure and electronics cabinets

fore opticsupportstructure

IFU/maskcassettes

Gemini instrumentsupportstructure

DewarCCD unitshuttermain optical

support structurecamera

grating turret& indexer unit

filter wheels

collimator

on-instrumentwavefrontsensor

Integral Field Unit

GMOS-IFU

Slit mask (containing two pseudoslits) interfaces with GMOS mask changer

Location of slits(covered)

The IFU

Requirements & solutions

• Exploit good images from GEMINI 0.2" sampling • Unit filling factor Fibres coupled to close-packed

lenslet array at input • Largest possible object field 7" x 5" (1000 fibres)• Provision to optimise accuracy of background

subtraction extra 5" x 3.5" field offset by 60" from object field for background estimation (500 fibres)

• Transparent change between modes IFU deployed by mask exchanger, input & output focus coplanar with masks

• High efficiency lenslet-coupled at output and input to convert F/16 beam to ~F/5 for efficient use with fibres

• Use of low risk construction technique (GEMINI request to reduce risk to schedule) fibre+lenslet not image slicer

4608 pixels

6144pixels

Optionally block off this slit to double spectrumlength but halvefield

1 arcmin

1 slit block containing 2 rows

Field to slit mapping

4608 pixels

6144pixels

Field to slit mappingOne slit blocked to give• Longer spectra• Half the field (can still beam-switch)

5.5'

Background subtraction

Field for Adaptive Optics

Objectfield

Backgroundfield

1 arcmin

• Various subtraction strategies • Beam switching supported• Optimised for AO (Altair in I)

Position of reference star during beam-switch

Typical/generousisoplanatic patch

Position of reference star during beam-switch

Typical/generousisoplanatic patch

One image at each velocity form the datacube (only 4% shown)

One spectrum for each element(only 4% shown)

The IFUrecords aspectrum for each element

Image taken by GMOS withoutusing the IFU

GMOS integral field unit observes NGC1068

[OIII]

Red Blue

Individualfibre

spectra

NGC1068 - raw data

• Composite plot of representative [OIII]4959+5007 spectra over the field

• The velocity structure is very complex.

NGC1068 - spectra

NGC1068 - datacube

•8 x 10" field (mosaiced from 5 pointings)

•Scan through [OIII]5007 line

Miller, Allington-Smith, Turner, Jorgensen

Jet

Galaxydisk

Nucleus

NE

SW

Observer

Bowshock

NE

SW

Advanced Image

Slicer (AIS)• Developed from MPE's 3D by the

University of Durham for highly-efficient spectroscopy over a two-dimensional field

• Optimum use of detector pixels since complete slices of sky are imaged (no dead space between spatial samples)

• Correct spectral sampling is obtained without degrading spatial resolution in dispersion direction

• Diffraction is only a 1-D issue

reduction in optics size/mass

• Optics may be diamond-turned from the same material as the mount to reduce thermal mismatch

good for space/cryo applications

• Adopted by GEMINI 8m Telescopes Project (GNIRS-IFU) and proposed by ESA for NGST

Field beforeslicing

Pseudo-slit

Slicing mirror (S1)

Spectrogram

Pupil mirrors(S2)

To spectrograph

Field optics (slit mirrors S3)

From telescopeand fore-optics

Focalplane

Gemini Near-IR Spectrograph

(0.2 x 0.1 x 0.1)m3

and 1Kg

• Cryogenic 1-5m spectrograph for GEMINI with IFU deployable via slit slide• GNIRS - NOAO, GNIRS-IFU - University of Durham

Field

Slit

Detector

3.2 "= 21 slices of 0.15"

4.4" = 29 px of 0.15"

2 pixels29 pixels

Detector: 1024 x 1024 pixelsSlit length (short camera)= 100" = 667 pixels

GNIRS-IFU summary • Wavelength range:

– Optimal: 1.0-2.5 m– Total: 1.0-5.0 m

• Field: 3.2”x 4.4” • Sampling: 0.15” • Spatial elements: 625• Spectrum length: 1024

px• Cryogenic environment• IFU fits in module in

GNIRS slit slide

Optical layout

S1, Slicing mirror

S3, slit mirrors

S2, pupil mirrors

F2, 1st reimagingmirror F1, pickoff

mirror

F3, 2nd reimagingmirror

From GNIRS fore-optics

To GNIRS collimatorSlice 1

Slice 2

Optical layout

S1

Monolithic S3Monolithic S2

Bi-lithic S1showing split

F2

Field 46x40" Sampling 0.19x0.19"

Fore-opticsFore-optics

Slicing unit Slicing unit

Blue+Red spectrograph(9 slits)

Fore-opticsFore-optics

Slicing unit Slicing unit

Blue+Red spectrograph(9 slits)

Fore-optics

Slicing unit

Spectrograph(1 slit)

4k x 4kdetector

1 slit

Field 3.8x2.6"Sampling 0.05x0.05"

MOS with IFS? - NGST/IFMOS HR LR

2kx2kdetector

9 slits

Fore-opticsFore-optics

Slicing unit Slicing unit

Blue+Red spectrograph(9 slits)

Work by NGST-IFMOS consortium sponsored by ESA

Did IFMOS get on NGST?

Work by NGST-IFMOS consortium sponsored by ESA. Picture from Astrium

No, but small-field IFU may be included in NIRSPEC alongside MOS mode

Multiple IFS

• IFS of multiple targets over wide field via deployable IFUs MOS with mapping to e.g. measure mass of many galaxies

• Total number of elements set by number of detector pixels:– This must be divided amongst the different IFUs– For example, 20 modules with 200 elements each could be

accommodated on a 4k x 4k detector small field/module

• Main focus is on near-infrared• Exploit "wide-field" AO on GEMINI and VLT • Existing small-field IFU system: VLT/Flames (NB: Falcon)

• Prototyping underway for image-slicing (e.g. VLT/KMOS)

Large-field multi-IFU prototype

•Complete deployable IFU module of 225 elements (Subaru F/2)

•Fishing rod deployment

Individual field15 x 15 (4.5" x 4.5")

Input

Output(slit for test only)

Probe arm + optics

30' primefocus field

Deqing Ren, PhD thesis, 2001. University of Durham

The enclosing circle is530mm diameterfor a 93mm diameterfield-of-view

UK-ATC

GIRMOS: gnomes around a pond

Feeds fixedimage-slicing IFUs

• stepping motor drive via worm gears

• for both ‘shoulder’ and ‘elbow’ actions

• two tubular arms in CFRP• the arms are not co-planar• four folds in each optical path• light re-imaged at x1.5

magnification

UK-ATC

light path

To fixed image slicer IFU From fore-optics

GIRMOS pickoff arm