Multiobject Spectroscopy

1
Multiobject Spectroscopy

• Jeremy Allington-Smith
• University of Durham

2
Contents

• Introduction to MOS
• Multislits and multifibres compared
• Multifibre systems
• Atmospheric effects
• Multislit systems
• Stability
• Optical performance
• Sky subtraction revisited
• Nod shuffle, microslits
• Alternatives to slit masks

3
Introduction to MOS
4
Basic principles
Non-contiguous sky spectrum
Object aperture
Detector
Sky
(S1)
(S1)
Sky apertures
A
A
B
B
(S2)
C
C
D
D
Spectrum of object only
Spectrum of object and contiguous sky background
(S2)
Non-contiguous sky spectrum
5
Top-level requirements

• Mandatory to obtain integrated spectrum of many
  objects
• One spectrum per object in defined aperture
• Estimate of spectrum of sky background
• preferably contiguous in same aperture
• or enough non-contiguous samples to build global
  model of sky
• Known mapping from sky to detector
• obtained simply by (wavelength calibration)
• mapping need not be simple!
• Optional to obtain spatially-resolved spectra
• Spatial resolution along slit/aperture
• Apertures can be tilted or curved
• to maximise throughput for extended source
• radial velocity distribution within aperture

6
Basic optical concepts
Collimator
Disperser
Camera
Multislit
Slit mask
Telescope focus
From telescope (or fore-optics)
Multifibre
Spectrograph optics
Long distance
Fibres
From telescope
Pseudo- slit
Fibre positioner
Telescope focus
(Dispersion shown rotated by 90? for simplicity)
7
Multislit vs multfibres

• Multislit
• Light goes directly from aperture into
  spectrograph
• ? distribution of spectra on detector is the same
  as that of apertures on the sky
• Overlaps between spectra are possible
• Difficult to observe objects which have same
  position perpendicular to the dispersion
  direction
• Multibre
• Light is conducted along flexible link (fibre)
• ? distribution of spectra on detector is
  independent of that of apertures on the sky
• Fibre outputs arranged as 'pseudo-slit' to avoid
  spectrum overlaps
• but fibre coupling may be lossy and destroys
  spatial info

8
Summary of pros and cons

• Multislit
• Efficient for faint sources
• fewer sources of light loss than fibres
• better sky subtraction - sky estimates in same
  slit
• limited field (?10') but fine resolution possible
  (0.1")
• Calibration straightforward
• Multifibre
• Very large fields possible (? 2?)
• Sky subtraction difficult - no adjacent sky
  estimates
• Good stability
• fibres immune to target position errors or
  guiding errors
• spectrograph can be gravity invariant eliminate
  flexure
• Calibration difficult

9
Sky subtraction

• Slits give adjacent sky estimates, contiguous
  with object
• Fibres do not, must build global sky model or
  beamswitch

A Object field B Background field
B
B
A
slit
A
Object
Fibres
Slit
10
Target position errors

• Slits retain image information perpendicular to
  dispersion direction
• Fibres scramble information on location of object
  within aperture

Centroid varies depending on position of object
within aperture of slit ? guiding/alignment
errors affect radial velocity measured
Slit
Input
Output
Centroid independent of position of object
within aperture of fibre ? guiding/alignment
errors have no effect on radial velocity measured
Fibre
11
Efficiency for surveys

• Multislit suffers from spectrum overlaps but
  target spacing can be small perpendicular to
  dispersion direction
• Multifibre does not suffer spectrum overlap, but
  limited by minimum closest approach of fibres

Spectra/fibres overlap
Max density for slits
Max density for fibres
Common objects (e.g normal galaxies)
Rare objects (active galaxies)
LogSurface density of targets
Min density for slits
Min density for fibres
Fibres
Sensitivity limit
Slits
Magnitude
Too few objects in field
12
Technology options
Multifibre
Multislit
Mini- bundles
Single fibres
Slit masks
Sliding slitlets
MEMS arrays
Normal
Micro-slits
mirrors
shutters
robots
spines
13
Multibre systems

• This is a review of the capabilities of current
  systems . Many of the technical issues which
  affect these systems also apply to multislit
  systems and will be discussed later

14
Two-degree Field (2dF, AAT)

• Field 2? diameter via corrector at f/3 prime
  focus
• 400 object fibres/field plate 4 guide fibre
  bundles,
• Fibre aperture 140mm (2 arcsec) diameter
• Fibre positioned by pick place robot
• Double-buffered observe with one plate while the
  other is configured
• Atmopheric dispersion compensator

15
2dF
16
Positioner performance

• Speed 6-7 seconds/fibre 1 hour/field ? double
  buffering
• Relibility one failure in every four fields
  configured
• Local positioning accuracy 15 mm (0.25 arcsec).
  
• Atmopheric refraction limits to Hour Angle /-
  2.5
• Active position control image back-illuminated
  fibres
• Fibre cross-overs must be dealt with carefully by
  s/w

17
2dF data Galaxy redshift survey
Each spectrograph handles 400 fibres (no overlaps)
18
Flames (ESO VLT)

• OzPoz (AAO)
• double-buffered
• fibre positioner at VLT Nasmyth
• 0.1" accuracy
• 10" minimum dist.

Gravity-stable Giraffe spectrograph
Fibre input (single fibres)
Pseudoslit
19
Flames fibre bundles

• Instead of 1 fibre use 20 to give image slicing
  or integral field capability? next lecture

Button deployed by positioner
20
"Image slicing"

• Gets both high spectral resolution and high
  throughput
• Can use fibres (for MOS) or little mirrors
• Don't confuse with integral field spectroscopy
• which retains spatial information (next lecture)

Slit is narrow to get high spectral resolution
but throughput is low
Reformatted slit (still narrow)
Slicing allows more light to be captured without
sacrificing spectral resolution
21
Issues for multifibre system

• Can't get fibres close together
• Limits on configuration flexibility due to
  cross-overs
• Reconfiguration time - longer for more fibres
• Atmospheric refraction ? update fibre positions
  but can't do this during observation
• Calibration of fibre throughput for each plate?
• Sky subtraction strategies global
  sky/beam-switch
• Stability
• fibres move but spectrograph stable (not 2DF)
• guiding error immunity for fibres

22
Alternative spines

• Mount fibres on spines, tilt to access small
  patrol field
• Natural match to studies of LSS (less good for
  clusters)
• Good for fast focii (PF of 8/10m) where
  inter-object distance is small (f/1.2, 8m
  50mm/arcsec) esp. ELTs

• Echidna (AAO) in progress for F/2 prime focus of
  8m Subaru as part of UK-Aus-Japan FMOS instrument
  
• 400 fibres/spines
• 7mm pitch (90")
• Possible for GSMT

23
Atmospheric effects
24
Effect of the atmosphere

• Parallel-plane model of atmosphere gives error in
  apparent position of object ("refraction")
• This also has a chromatic component since
  refractive index depends on wavelength
  ("dispersion")

25
Atmospheric refraction

• Significant for wide fields and fine image scales
• Motion of images at edges of a 2x2 field
  compared with 2? aperture for (1 HA) at dec
  50?
• Without correction, field is squashed and rotated

start
end
?sec2(ZD)
26
Atmospheric refraction

• Problem - field dependent!
• Errors in radial velocity since image moves in
  slit
• Errors in photometry since object not centred in
  slit
• Solutions
• If possible observe near transit
• Limit length of observation to reduce airmass
  range
• Use a different mask for each range of airmass,
  with slit positions allowing for refraction
• Make sure Telescope Control System correctly
  calculates refraction and has a correct
  atmospheric model (depends on pressure and
  temperature)
• If possible, use two guide probes to remove
  rotation empirically

27
Atmospheric dispersion(differential refraction)
To zenith
If slit is horizontal, light is lost at extreme
wavelength
Airmass sec(ZD) 2
For AAT (low altitude)
Error in position if guiding at 500nm
If slit is vertical, all light is in slit, but
spectrum will be curved
?tan(ZD)
28
Atmospheric dispersion

• Problem - wavelength dependent (but independent
  of field)
• Errors in photometry as light misses slit at
  extreme wavelengths
• Errors in radial velocity
• Solutions
• If possible, limit wavelength range to near
  reference wavelength
• Orient slit to always point to zenith -
  impossible for most observations of extended
  objects and usually for MOS
• Make sure Telescope Control System correctly
  calculates refraction and has a correct
  atmospheric model (depends on pressure and
  temperature)
• Use an Atmospheric Dispersion Compensator

29
ADC

• ADC contains a pair of prisms which rotate about
  optical axis
• The prisms produce no deviation at central
  wavelength
• Vector sum of dispersion arranged to cancel out
  atmospheric dispersion

Oiled gap between prisms
GMOS design
Result from 2DF (AAO)
Slit plane
30
Multislit spectrographs
31
GMOS multislit example
5.5 arcmin
A383 observed with GMOS
Note extra space required on detector to
accommodate spectra
32
Spectrum overlaps in MOS
Slit mask
Slit A
Slit B
Slit C
Slit D
33
Spectrum overlaps in MOS
Detector
1st order
2nd order
Zero order
Slit A
Assuming that only a clean 1st-order spectrum is
required
Slit B
Slit C
Slit D
34
Effect of anamorphism
Detector
1st order
2nd order
Zero order
Slit A
Slit B
Images of slit in direct image
Slit C
Slit D
35
Effect of distortion
Detector
Slit A
Slit B
Slit C
Slit D
36
Slit masks

• Using positions from a direct image
• eliminates astrometric errors (e.g. proper motion
  of fiducials)
• must account for optical distortion from mask to
  detector
• GMOS masks
• Minimum width 0.2 arcsec 0.12mm
• 5.5 x 5.5 arcmin mask 206 x 206mm
• 600 slits maximum/mask (200 slits/2 hours)
• Edge quality 1mm rms, position error 5mm rms
• Material 3-ply carbon fibre

37
GMOS mask handling system
Masks stored in one of several cassettes
From telescope
Mask position at focal plane
38
Errors in centroid of VRE

• VRE velocity resolution element,
• the monochromatic image of the slit as recorded
  by the detector

Target-slit error Centroid varies depending on
position of object with respect to slit due to
guiding error or movement between telescope and
slit
Slit-detector error Centroid varies due to
movement between slit and detector
39
Centroid errors

• Errors in slit position cause
• loss of throughput
• error in measured radial velocity
• Two nasty sources of astrophysical error
• plate scale error ? spurious radial dependence of
  RV or intensity and overestimate of velocity
  dispersion
• Mask rotated with respect to targets ? errors as
  above
• Some causes of error
• Errors in position of target (celestial or from
  image)
• Error in assumed plate scale (error depends on
  radius)
• Inaccuracy in mask maker (random or systematic)
• Error in guiding and aligning mask with sky
  during acquistion
• Atmospheric refraction varying through
  observation
• Instability in spectrograph between slit and
  detector

40
Atmospheric refraction again

• Significant for wide fields and fine image scales
• Motion of images at edges of a 5.5?x5.5? field
  compared with 0.3? aperture for (2 HA) at dec
  80?
• Image will move in slit causing velocity errors

0.3?
0.3?
Rotation corrected empirically
Rotation not corrected
41
Stability
42
GMOS hardware
43
GMOS gravity-induced flexure
Measured raw flexure as a function of tilt about
y and z axes

• Flexure between mask and detector
• Passive 12mm/hr
• mostly elastic
• Active 3mm/hr
• CCD translated using model derived from
  measurements of image of pinholes in mask

CCD translation stage
( at detector per 15-deg tilt 1 pixel
13.5 mm)
44
GMOS thermal-induced flexure
Enclosure acts as thermal buffer to reduce
temperature variations
Environment induces natural temperature
variations during the night and from night to
night

• Mauna Kea -5C lt T lt 5C 95 of the time
• Gradient dT/dt lt 0.8C/hr
• Aim to reduce temp fluctuations within GMOS
  enclosure to lt 0.24C/hr (0.48mm/hr at field
  edge)
• minimises image movement
• minimises focus changes (and plate scale changes)
• Image movement measured to be 0.62mm/hr

45
Guiding and wavefront sensing

• Purpose is to keep target accurately aligned with
  the slit
• Flexure and thermal effects between slit and
  guider must be small
• Guider can be upgraded to a wavefront sensor to
  provide telescope with e.g.
• tip/tilt/defocus signals

GMOS OIWFS Hardware Probe in GMOS focal plane 2x2
Shack-Hartmann Performance Tip/tilt, focus
signals gt M2
astigmatism corrn gt M1 R15 tip/tilt at
100Hz Max rate 200Hz Flexure 22mas/hr on sky
OIWFS image of star
46
Optical performance
47
Image scale

• Image scale at slit determined by collimator
• (dx/dc )slit fT F1DT in e.g mm/arcsec
• arrange for slitwidth to be physically
  manufacturable!
• GMOS example F1 FT 16, DT 8m
• ? (dx/dc )slit 621mm/arcsec,
• ? 0.2 arcsec slit 124mm
• Image scale on detector determined by focal
  reducer demagnification
• (dx/dc )det M(dx/dc )slit F2DT since M
  F2/F1
• arrange for slitwidth to project to gt 2 pixels
• GMOS example F2 4.8 so M 0.30, pixel size
  13.5mm
• ? (dx/dc )det 207mm/arcsec
• ? 0.2 arcsec 41.4/A mm gt 2.0 pixels for A lt 1.5

focal ratios of camera collimator
Anamorphic factor
48
Image quality

• IQ must be good enough to
• adequately oversample
• the slit (Nyquist 2 pixels/slitwidth)
  dispersion
• the seeing (Nyquist 2 pixels/FWHM) spatial
• not degrade spatial information from telescope
• be uniform over the field
• Example for GMOS
• Image scale 0.072"/pixel cf narrowest slit (0.2
  arcsec) with anamorphism
• Best images delivered by site 0.25 arcsec
  (10-ile)
• Telescope degrades by 15
• Instrument allowed to degrade by further 10
• Calculations assume "realistic" Kolmogorov PSF
  (not gaussian)

49
GMOS image quality

• Image quality defined by diameter encircling
  given of energy (EED)
• 50EED FWHM of 2-D gaussian (36 for PSF used
  for GMOS)
• 85EED defines wings (100EED ? ?)

50
GMOS throughput

• Throughput depends on
• main optics
• gratings
• filters
• detector
• GMOS example

51
Better sky subtraction? -Nod shuffle,
microslits
52
Sky subtraction with slit
B
A
Noise due to slit roughness
Corrected photon number
Signal to extract
distance along slit
estimated background signal uncertain slopes
due non-parallel sides
dispersion
Do this at every wavelength!
53
Sky subtraction near bright sky lines

• Poor cancellation of sky line due to
• Difference in line profile due to
• uneven slit width
• IQ varies over field
• Difference in line location due to
• tilt of slit
• poor wavelength calibration/ solution/

54
Nod shuffle (Va vient)

• Errors in sky subtraction
• Sky is spatially structured on scale of slit
  width
• Errors in slit fabrication lead to extra noise
• problems with flatfielding since calibration
  spectrum needs to match sky's spectrum
• fringing in CCDs
• Solution Use same detector pixels and optical
  path to alternately sample object and sky
  (beam-switch)?
• Advantages
• improved background subtraction
• can use shorter slits (microslits) to increase
  multiplex
• Potential drawbacks
• must alternate fast enough to cancel out temporal
  variations
• detector readnoise is increased due to multiple
  readouts

55
Nod shuffle in action
CCD

• Requirements
• ability to move telescope with good repeatability
• ability to move charge on CCD (controller upgrade)

Courtesy Karl Glazebrook
Glazebrook Bland-Hawthorn PASP 113, 197 (2001)
56
Nod shuffle on GMOS

• Example from engineering tests
• Shift object along normal slit
• 2 cycles of 60s in each position nod /- 1.5,
  shuffle 70 px

Slit length
After subtracting bottom half from top half
Anti-object
Object
57
Example object raw objectsky
I23.8
OH line forest
Courtesy Karl Glazebrook
58
Example object NS subtracted
OII3727at 770nm
I23.8 z1.07
Courtesy Karl Glazebrook
59
Microslits with NS

• Galaxy cluster AC114
• AAT/LDSS
• 586 microslitsnon-overlapping
• 40nm blockingfilter _at_ Ha
• I lt 22

Mask design software predicts layout of
spectra must have microslit landing on clean sky
after telescope nod
Couch et al. ApJ 549, 820 (2001)
60
AC114 Mask
61
Future challengesalternatives to slit masks?
62
MOS in space

• Key goal of NGST explore the epoch of initial
  galaxy formation
• The faintest galaxies are small and far apart.
• At AB29 half light diameter 0.2
• At AB30 galaxy density is 3 x 106 deg-2
• ? 17000 in 7.5 x 3.75 arcmin
• The multiobject capability of
• NIRSPEC will access most
• interesting galaxies in a large
• field simultaneously.
• 6000 galaxies at R40, 30 lt KAB lt 32 or zgt1.6
• 1600 galaxies at R1500, 28 lt KAB lt 29 or zgt2
• 600 galaxies at R5000, KAB lt 23.1
• ? Requirements
• Focal plane must be remotely configurable
• with no consumables and be reliable
• Address high surface density of targets

HDS-S image from STIS (to AB30)
63
MOS in cooled IR spectrographs

• Need to operate in temperatures depending on red
  cutoff and spectral resolution 240K?80K? 30K
• Slit masks must pre-cooled before installation in
  instrument cryostat equipped with gate valves
• Fibres can work in cold
• with attention to
• thermal mismatch but
• difficult with lenslets
• ? Requirements
• Focal plane must be
• remotely configurable

64
Alternative multislit systems

• What's wrong with slit masks?
• Have to make one for every new field (or
  airmass!) often requiring prior imaging campaign
  ? inefficient
• For cooled IR instruments the mask must
  thermalise ? cryostat must be opened and
  resealed, time delay
• not practical for space applications
• Reconfigurable focal planes
• can generate slits rapidly as a remote operation
• ? no operator intervention or consumables
• Current options have some drawbacks
• spatial quantisation limits flexibility in
  adressing targets
• spatial quantisation may introduce velocity
  flux errors (requires modelling or use of image
  data to correct)
• microshutter/mirror arrays - limited contrast
  ratios

65
Microshutter arrays and sliding slits
slide

• Each half of slitlet slides individually to give
  precise slit width and location in y
• Inflexibility in matching object locations in x
• Only 20-40 slits possible
• Multiple banks impossible
• Contrast ratio high

• Individual tiny elements can be swiched on or off
• Quantisation in both x and y
• Array gives fine quantisation (1k x 1k via
  mosaicing)
• Multiple banks OK
• Filling factor limited (support grid)
• Contrast ratio limited

66
Microshutter array
Baseline for NGST NIRSPEC 2kx1k (100x200mm) -
Moseley et al. NASA/GSFC
67
Sliding multislits
NB also VLT/FORS-1 has a 19-slit unit
Backup for NGST/NIRSPEC (Courtesy CSEM/Astrium)
68
Contrast ratio issues

• Problems if mask leaks light through
• bright sky leaks through enhancing background and
  reducing sensitivity
• bright objects leak through so their spectrum
  contaminates strips of the detector
• For NGST Near-Infrared Spectrograph (NIRSPEC)
• "bright spoilers" increase in density with lower
  galactic latitude so could wipe out Galactic
  science
• 2001 only barely adequate at Galactic poles
  (rejects Zodi)
• 20001 is okay for b gt 10? (probability of losing
  10 of field depending on detector noise and
  spectral resolution)
• Micromirror arrays rejected because of poor
  contrast
• Microshutter arrays likely to achieve 20001 but
  some science in Galactic will be impossible
• Only metal slits can help!