The Deep Imaging Multi-Object Spectrograph for Keck II by S. M. Faber and the DEIMOS Team

1
The Deep Imaging Multi-Object Spectrograph for
Keck IIbyS. M. Faber and the DEIMOS Team

• Supported by CARA, UCO/Lick Observatory, and the
  National Science Foundation

2
Structural Overview
3
During Assembly
4
Final Assembly Santa Cruz
5
At the Nasmyth Focus at Keck
6
Goals vs. Performance

• DEIMOS was conceived to be maximally efficient
  for faint-object spectroscopy of objects densely
  packed on sky
• Minimize the effect of sky background
• Get between OH lines ??1.25 A, R 6000 ,
  x4 speed gain
• Accurate flat-fielding 0.2 rms
  (photon-limited for 10-hr exposures)
• Stable image position (fringing) 0.6px rms
  (goal)
• High observing efficiency
• Long slit length on sky 16.7, gt130 slitlets
• Broad spectral coverage 2000 resolution
  elements
• High throughput 28 peak (with atm tel 50
  DEIMOS alone)
• Low readout noise 2.3 e
• Fast readout time 50 sec
• Rapid slitmask alignment 5 min (goal)
• Excellent image quality (3800 A to 10,500 A)
• Hoped for 0.6-0.8 px (1-d rms with 15?
  pixels)
• Actual 0.8-1.2 px gt 2.0-2.8 px
  FWHM

7
Detector Performance

• The detector is a mosaic of 8 2K x 4K CCDs from
  MIT/Lincoln Laboratories. The CCDs are
  high-resistivity, red-sensitive devices that are
  45 ? thick, with a peak QE of 85 and enhanced QE
  of 23 at 10,000 A.

8
Pretty pictures NGC 7331
9
DEIMOS Masks and Detector

• Slit masks are curved to match the focal plane
  and imaged onto an array of 2k ? 4k CCDs
• Readout time for full array (150 MB!) is 50
  seconds (8 amplifier mode)

10
Arc Spectrum 133 slitlets
11
First-light Spectrum
12
Sky-subtracted Sub-regions
13
Sky-subtracted Sub-regions
14
Sky-subtracted Sub-regions
15
Kinematic Information
16
Kinematic Information
17
Kinematic Information
18
Kinematic Information
19
Sky Subtraction is Key
Left Raw data from an unaligned DEIMOS slitmask,
with serendip (detail). Some slitlets are tilted
to allow rotation curve measurements this poses
unique challenges for automated sky subtraction.
Below test analysis of one tilted slitlet. From
top raw data, b-spline model of the night sky
lines, and rescaled residual. We already can
achieve sky subtraction at close to the Poisson
limit in cases like this.
20
Typical Extracted 1-d Spectrum
Unsmoothed 1-d spectrum with background sky (red)
offset and rescaled.
21
Poisson-Limited Sky Subtraction
Plot shows residual of flux from b-spline sky
model in region of sky emission lines, in units
of local RMS. Smooth curve is gaussian, width 1.
Work in progress to do non-local sky subtraction
using narrower, sky-only slitlets, for the
shortest slitlets where local sky subtraction is
impossible.
22
The UCB Automated Data Pipeline
A small group of galaxies with velocity
dispersion ? ? 250 km/s at z? 1. Note the clean
residuals of sky lines.
23
CCD Crosstalk

• The image from CCD 6 appears negatively on CCD5

• The amplitude saturates at about 2.5 e

• The main effect is to create negative sky
  lines. The widths depend on line brightness
  unpredictably

• Possibly due to open wire on CCD5 A amplifier

• Action is TBD

24
Optical Performance

• The camera was designed by Harland Epps. It has
  exceedingly wide field of view (11.4 radius),
  three steep aspherics, three large CaF2 elements,
  a passive thermal plate-scale compensator, and
  three fluid-coupled multiplets.

25
Camera/Dewar Layout
26
Images at First Assembly
Radial comatic tails, max 15 px
27
Causes of Radial Coma

• Inherent in optical design performance at room
  temperature differs from 0 C
• ? Accounts for about half of effect
• Element 8/9 spacing too short
• Detector too deep in dewar
• Multiplet 4 slightly too thick

28
Three Optical Adjustments
29
Sample Images Dome Lights
Detector center
Line profile
Image 0.5 pinholes
30
Line Profiles Top, Center, Bottom
Bottom
Center
Top
No coma
No coma
31
Far Corners vs. Center
32
Measured Image Sizes

• Estimated RMS image sizes, corrected for 0.5
  pinhole
• Actual
  Predicted
• Center Corners
  Center Corners
• 1-d ? 0.88 px 1.17 px 0.60
  px 0.82 px
• 13.2 ? 17.5 ?
  8.8 ? 12.0 ?
• FWHM 2.07 px 2.75 px 1.41 px
  1.93 px
• 31.7 ? 41.2 ?
  21.1 px 29.0 ?
• Extra source of broadening equivalent to 11.3?
  (1-d ?)
• Possibility refractive index inhomogeneities?
  CaF2?

33
Image Stability

• The original passive specification for image
  motion was 6 px peak-peak under 360? rotation in
  X and Y. This goal has not been met, but the
  final image stability specifications seem to be
  within reach nevertheless.

34
Image Stability/Flexure

• Reasons for wanting stable images
• Image quality
  X is along slit
• Needed during single exposure
  Y is along spectrum
• Affects both X and Y
• Specification lt 1 px rms
• Flat-fielding accuracy
• Needed between afternoon calibrations and evening
  observations
• Flat-fielding accuracy requirement 0.2 rms
• Affects Y only (along spectrum)
• Specification lt 0.6 px rms (originally 0.25 px
  rms)
• Use flat fields to delineate slitlet edges
• Needed between afternoon calibrations and evening
  observations
• Affects X only (across spectrum)
• Specification lt 1 px rms

35
Flexure Compensation System

• Closed feedback loop both centroid sensing and
  correcting
• Operates in both direct imaging and spectroscopy
  modes
• Sensing system
• Four optical fibers pipe CuAr light (or LED) into
  telescope focal plane at opposite ends of
  slitmask
• Two separate sensing CCDs are mounted on detector
  backplane flanking the science mosaic
• These FCS CCDs are read every 40 sec when shutter
  is open
• Feedback is achieved only when shutter is open
• Correcting system
• Steers image in X and Y no rotation
• X actuator motor in dewar moves detector along
  slit
• Y actuator piezo on tent mirror moves spectrum
  in ?

36
Flexure Compensation CCDs
37
FCS Actuators
38
Flexure History

• Initial image motion on first assembly
• X motion 40 px Correctable
  range 26 px
• Y motion 7 px
  13-23 px
• MUST FIX X MOTION!
• Year-long campaign discovered moving elements in
  camera and grating system
• Current image motion
• X motion 8 px
• Y motion 18-23 px (depends on
  grating or mirror)
• Lessening X increased Y to some degree
• Tilting grating is needed in Y in addition to
  tent mirror

39
Y Correction First Results

• Performance with closed-loop correction
• Total image motion through 360 rotation, in px
  slider 3 USING ONLY ONE FIBER ON ONE FCS
• Nature of motion sag in Y, larger with X (i.e.,
  a shear)
• Probable cause pitch of collimator
• Expectation final rms will be 0.4-0.5 px .
  meets goal

0.75 1.00 1.62
RMS 1.0 px
0.31 0.75 1.25
Y motions
RMS resid 0.4 px
0.50 1.25 1.19
Goal 0.6 px
Position on detector
40
X Correction First Results

• Performance with closed-loop correction
• TOTAL image motion through 360 rotation, in px
  slider 3 USING ONLY ONE FIBER ON ONE FCS
• Nature of motion shift in X, mainly bulk motion
• Probable cause flexure in the fiber mount
• Expectation final rms will be 0.6-0.7 px .
  meets goal

2.43 2.25 2.88
RMS 2.1 px
1.62 2.38 2.00
X motions
RMS resid 0.5 px
1.25 2.13 1.95
Goal 1.0 px
Position on detector
41
Lessons Learned

• Success-oriented does not work at this scale
• Expect that most mechanisms will NOT work as
  designed the first time. Hence
• Build prototypes and test extensively before
  putting into spectrograph
• The major source of flexure is not the main
  structure but rather mechanisms attached to the
  structure not easily analyzed using FEA hence
  the need for prototypes

42
Final Lesson Naming

• Phobos and Deimos were the horses that pulled
  the chariot of Aries, the god of war.
• Phobos means fear.
• Deimos means the awe one feels on the
  battlefield when in the presence of something
  greater than oneself.

MORAL be careful naming your instrument
names have a way of coming true
43
Comparison Between DEEP2 1HS and Local Surveys
SDSS
2dF
LCRS
DEEP2
z0
CFASSRS
PSCZ
z1
44
Masks Tiled on a 42x28 CFHT Pointing
45
Colors Pre-select Distant Galaxies

• Plotted at left are the colors of galaxies with
  known redshifts in our fields those at low
  redshift are plotted as blue, those at high
  redshift as red (diamonds are beyond the mag.
  limit of the survey).
• A simple color-cut defined by three line
  segments would yield a sample gt90 at zgt0.75 and
  missing lt3 of the high-z objects. Most of the
  failures are likely to be due to photometric
  errors.

46
Test of Photo-z Selection Procedure
Redshift distributions in early masks are
consistent with expectations
47
Simulated DEEP2 Spatial Sampling
Courtesy A. Coil
Targeted objects are included when our slitlet
assignment algorithm is performed on a mock DEEP2
survey created from an N-body simulation missed
objects are those not selected
48
Another Redshift Survey The VLT/VIRMOS Project

• 50,000 galaxies to IABlt 24 (1.2 sq. deg)
• 105 galaxies with IABlt 22.5 (9 sq. deg)
• 750 simultaneous slitlets (4 barreled instrument)
• Resolution R 180-2520 short spectra, multiple
  spectra per row
• 100 nights on VLT-3 Observations start
  November 2002

49
DEEP2 versus VLT/VIRMOS
HAS VIRMOS chosen quantity over quality?

• Only half their galaxies will be distant

• Most of their galaxies have resolution 200, not
  5000 no kinematic info inferior velocities?

• They cannot subtract sky accurately at R200
  will lose x2 overhead for nod and shuffle

50
Advantages of DEEP2 over VLT/VIRMOS

• Higher resolution
• Provides more precise redshifts and allows secure
  z measurements from the OII doublet alone
• Permits us to measure linewidths/rotation curves
• Reduces contamination by night skylines
• Necessary for many of our science goals e.g. T-F
  type relations, studies of bias (e.g. via
  redshift-space distortions), measurement of
  thermal motions, determining velocity dispersions
  of clusters, the dN/dz test None of these will
  be possible with low-resolution VLT/VIRMOS data.
• Photometric cut for zgt0.7 will eliminate 50 of
  all galaxies with IABlt 23.5 from target list,
  yielding denser sampling at z 1

51
Schedule of the DEEP2 Survey

• DEIMOS has been reassembled and tested at Mauna
  Kea
• Commissioning began June 2002 under clear skies
  and was extremely successful
• DEEP2 observing campaign began in July 2002. (so
  far we have had 49 science nights clear, and on
  34 of these, the TV camera was broken!)
• Observations complete late 2004 (we hope)
• Analysis complete late 2006