Development of NIR Detectors and Science Requirements for SNAP Thesis Defense November 29th, 2006 Ma

1
Development of NIR Detectors and Science
Requirements for SNAPThesis Defense November
29th, 2006Matt BrownUniversity of
MichiganAdvisor Greg Tarlé
2
Development of NIR Detectors and Science
Requirements for SNAP Thesis Defense November
29th, 2006Matt BrownUniversity of
MichiganAdvisor Greg Tarlé

• Dark Energy and the Accelerating Universe
• The SNAP Satellite
• NIR Detector Characterization
• NIR Science Simulations
• Dark Energy Constraints

3
Fundamental Questions

• What is the nature of matter and energy at its
  most fundamental level?
• (What is the universe made of?)
• What is the evolution and destiny of the universe
  and how is it affected by the fundamental
  interactions of energy, matter, time and space?
• (Is the universe infinite? Will it last forever?)

4
A Startling Discovery

• Using type Ia supernovae, the Supernova Cosmology
  Project and the High-Z Supernova team constructed
  a Hubble diagram out to z 1.
• Both teams made the startling discovery that the
  expansion of the universe is accelerating.
• The thing causing the accelerated expansion is
  known as Dark Energy.

5
A Revolution in Cosmology
based on 42 SN ( Supernova Cosmology Project)
Constraints in the WM-WL plane as measured by the
Supernova Cosmology Project.
90
68
accelerating
decelerating
6
A Revolution in Cosmology
Perlmutter et al. 1999

• Weak lensing mass census
• Large scale structure measurements
• ?M 0.3

7
A Revolution in Cosmology
WMAP year 3
?total 1.02/-0.02

• Weak lensing mass census
• Large scale structure measurements
• ?M 0.3

Baryon Density ?B 0.044/-0.004
8
A Revolution in Cosmology
Flat universe ?total 1.02/-0.02
Baryon Density ?B 0.044/-0.004
WMAP

• Weak lensing mass census
• Large scale structure measurements
• ?M 0.3

9
Energy Budget of the Universe
Illustration creditAnn Field, STScI
10
What is the nature of dark energy?

• We now know that dark energy exists
• The dominant component of our universe
• Dark energy does not fit in current physics
  theory
• New theories propose a number of alternative
  physics explanations, each with different
  expansion history we can measure.
• Type Ia supernovae and weak gravitational
  lensing are two tools SNAP will use to measure
  the properties of dark energy

11
The Observational Tool SNe Ia

• C/O white dwarf accretes mass of a companion
  star leading to a thermonuclear explosion near
  the Chandrasekhar limit (1.4 MO)
• Explosion follows consistent patternwith nearly
  the same peak intensity
• Extremely bright event observable on
  cosmological distance scales
• Spectrum and brightness evolve with time
• Peak Magnitude is a standard candle to measure
  distance

.
12
Importance of High z Supernovae
Existing SNe data include 17 well observed type
Ia SNe discovered from space with HST. They
represent lt 10 of the total SNe sample, but
improve constraints on dark energy by a factor
of 2. Discovering these SNe requires space based
NIR observations.
No HST SNe
With HST SNe
13
The SNAP Satellite
14
Its a SNAP!

• A simple dedicated experiment to study dark
  energy

• Essentially nomoving parts
• 2 meter aperturetelescopesensitive to light
  fromdistant SN
• focal plane instrumented withgt 600 million
  pixels over 1 square degreeefficiently
  measures large number of supernovae
• Integral field optical and IR spectroscopy 350
  1700nmdetailed analysis of each SN

15
The SNAP Collaboration
16
Instrument Concept
Baffled Sun Shade
Solar Array, Sun Side
Instrument Suite
3-mirror anastigmat 2-meter Telescope
Solar Array, Dark Side
Instrument Radiator
Spacecraft Bus
17
Focal plane
D56.6 cm (13.0 mrad) 0.7 square degrees!
18
High-Resistivity CCDs for SNAP

• New kind of Charged Coupled Device (CCD)
  developed at LBNL.
• Better overall response than more costly
  thinned devices in use.
• High-purity radiation detector silicon has
  better radiation tolerance for space
  applications.
• The CCDs can be abutted on all four sides
  enabling very large mosaic arrays.

LBNL Red Hots NOAO September 2001 newsletter
19
Hybridized 1.7 mm cutoff HgCdTe Detectors

• Ongoing RD effort with Rockwell Scientific and
  Raytheon Vision Systems to produce high QE, low
  noise 2Kx2K detectors
• CMOS readout bump bonded to HgCdTe diode
• Non-destructive readout cosmic ray rejection,
  reduce read noise
• CdTe substrate will be removed proton induced
  luminescence

10mm
CdTe Substrate
800mm
20
UM NIR Laboratory
Dewar 1 Readout electronics Dewar
2 Power supply and temp. controller
Calibrated Flat-field Illuminator ESD
safe environment Spot-o-Matic
21
Characterizing NIR Detectors

• Characterizing NIR detectors is not as
  straightforward as it may seem!
• Each test can be affected by pixel self heating,
  capacitive coupling, bias/temperature drift, and
  measurement techniques
• Much of the effort in the last 4 years has
  focused refining measurement techniques to
  understand detector behavior
• Even standard tests face these issues
• Gain Capacitive coupling modifies the variance
  estimator
• Dark current Reset Anomaly is due to pixel
  self heating
• Fowler Read Noise Drifts and inter-pixel
  variations need to be removed before measuring
  the noise
• Quantum Efficiency Depends on detector gain and
  absolute photon calibration
• Intra-pixel variation Capacitive coupling
  modifies pixel response

22
Conversion Gain Measurement
Gain is measured with 3 techniques
variance estimatoraccounts for IPC
traditional variance estimator
standard gain measurement (Gaussian fit)
Ignoring correlated noise over-estimates the gain
by 20.(for this device)
Agreement between Gaussian and standard variance
methods confirms that outliers have been properly
masked.
23
Capacitive Coupling - Autocorrelation
? Cap. coupling occurs in mux and bump bond region
Average Correlation to neighboring pixels
4 (Nodal capacitance 32.2 fF 38.6 fF w/o IPC)
Average Correlation to neighboring pixels 1
(rows), 0.5 (columns) Nodal capacitance 75.1 fF
Average Correlation to neighboring pixels 2.5
(rows), 1 (columns) Nodal capacitance 77.7 fF
before after
epoxy underfill
trace topology in multiplexer
correlation increases by 2x
MGB, M. Schubnell, G. Tarle, Correlated Noise
and Gain in Unfilled and Epoxy Under-filled
Hybridized HgCdTe Detectors, Published in PASP,
Sept. 2006.
24
Dark Current
H2RG 40
RVS 09A
Internally funded
25
Dark Current Reset Anomaly
Local Heating
Reset Anomaly is due to local heating/cooling of
the individual pixels - Constant cadence readout
eliminates this effect
Local Cooling
26
Drift and Fowler Sampling
Noise e-
Fowler
When continuously clocking during the fowler
exposure the noise floor and total integrated
signal are reduced This was observed in our RVS
InGaAs detector with much gt amplitude Conclusion
Constant Cadence must be used to eliminate
reset anomaly
27
Quantum Efficiency
QE Measurement(5 absolute achieved 2 goal)
28
Intra-pixel variation Spot-o-matic
Micron-size NIR point projection system uncovers
sub-pixel structure
29
De-convolution Understanding Intra-pixel
Response
lets fit also the pixel width square
PRF (17.8 .1 mm) PSF (1.4
mm) charge diffusion (1.7 .02
mm) capacitive coupling (2.4 .1)
start with square PRF (18 mm) convolve with PSF
(1.4 mm) add charge diffusion (1.7.02 mm) add
capacitive coupling (2.2 .1) compare to data
published value 2.2 .1
Sub-pixel Response Measurements of Near-Infrared
Sensors, in preparation
30
NIR Science Simulations
31
Simulation Procedure
Use the SNAP simulation to connect detector
properties to the science output
32
Flux Variance at the Focal Plane
fzodi zodiacal flux (photons/m2/arcsec2/s) Atel
telescope collecting area (m2). 2m primary
assumed Apsf area of psf (arcsec2) ?(?d2
diffusion2), ?d 1.22?/D QE detector quantum
efficiency RN read noise (e/pix) DC dark
current (e/pix/s) Nexp Number of exposures (4
for visible, 8 for NIR) texp exposure time
(300s) Npix Number of pixels covered by psf
(3-6 depending on filter)
- Detector modeling at the aperture level
(2x2) pixels x (4 dithers) x (2 steps) 32
samples Gaussian noise models work well.
33
Zodiacal Background
Background for SNAP Imager assuming NIR QE 100
Zodi parameterization from Greg Aldering
34
Type Ia Spectrum
Ia spectrum at peak for restframe (left) and
z1.7 (right)
35
Simulated Detector Performance
z 1.7 supernova Ia
Detector parameters measured in the lab are used
to simulate light curves
36
Simulated Detector Performance
z 1.7 supernova Ia
Detector parameters measured in the lab are used
to simulate light curves Light curve Signal to
Noise is used to define detection threshold
S/N at peak, increasing QE
37
Simulated Detector Performance
z 1.7 supernova Ia
Detector parameters measured in the lab are used
to simulate light curves Light curve fits ?
parameter errors vs. detector noise
error on peak flux, QE 95
MGB et al., Development of NIR Detectors and
Science Driven Requirements for SNAP,
Proceedings of the SPIE, Volume 6265, May 2006.
38
Simulated Detector Performance
z 1.7 supernova Ia
Detector parameters measured in the lab are used
to simulate light curves Light curve fits ?
parameter errors vs. detector noise Multi-band
light curve fits ? error on SNe peak magnitude
Initial SNAP Spec
Magnitude error for z1.7 SNe (type Ia dispersion
0.12-0.15 mag)
MGB et al., Development of NIR Detectors and
Science Driven Requirements for SNAP,
Proceedings of the SPIE, Volume 6265, May 2006.
39
Distance Modulus vs. z
Intrinsic Dispersion 0.12 mag
Distance modulus error for 0, 1, 2, and 3 NIR
filters. NIR Detectors with QE 95, Total Noise
10e-
40
Simulation of 2000 SNe
41
Cosmological Constraints I (SNe Wm prior)
50 SNe per z0.05 redshift bin. Mag error Sqrt
0.152 stat2 (0.02z/1.7)2
Constraints on wa improve significantly with
prior knowledge of Wm This highlights the
complementarities between SNe and weak lensing
for constraining dark energy
No priors Wm .26 - 0.03 Wm .26 - 0.01
A Bayesian estimator with Wm 0.26 - 0.03
yields the contours shown in RED
42
Cosmological Constraints II (SNe only)
Brane World Dark Energy
Cosmological Constant
43
Cosmological Constraints III (SNe Weak Lensing)
sw0 0.04 swa 0.14 Combined constraints
consistent with DETF goals for stage IV experiment
44
Conclusions
Characterizing NIR detectors is a challenge that
has been met by the SNAP NIR team. A vigorous RD
program with Rockwell Scientific and Raytheon
Vision Systems has resulted in high quantum
efficiency, low noise NIR detectors. Simulations
show that QE leads to the largest gains for
supernova photometry, and that statistical errors
for a z1.7 supernova are at the level of the
intrinsic dispersion. SNAP will constrain the
dark energy equation of state and provide new
insights into the nature of dark energy.
45
Pixel Level Simulations

• So far all simulations use a parameterized
  description of the NIR detectors
• All pixels have the same mean RN, DC, and QE
• Understanding the distributions of these
  parameters is necessary to define operability
  specifications
• Pixel level simulations are also needed to
  understand secondary effects such as persistence

46
Persistence
Persistence is the release of charge following
illumination of HgCdTe arrays. - Appears to be
both flux and intensity dependent Currently
working to simulate persistence using VLT VMOS
galaxy data and USNO-B stars in the SNAP north
field. - Combine SNAP frames with measured data
to simulate persistence frames and develop
persistence specification
Grade B persistence from Rockwell 0.2
persistence in next frame
47
Simulated SNAP Fields
Add galaxies with pixsim using the luminosity
function in Zucca et al. (astro-ph/0506393) (Flux
normalization still needed)
All stars in USNO-B catalogue with RA272.5
DEC54.5
48
Astronomical Objects in SNAP Field

• SNAP pixel pitch is 0.17 arcsec ? 0.029 arcsec2
• Each 2k x 2k detector 0.09 deg2
• Point sources
• selected from USNO-B complete to V21 mag
• RA272.5 DEC54.5
• Mean number of objects per 0.09 deg2 field 70
• 400 pixels impacted by bright point sources
  (0.01)
• Galaxies
• Mean central surface brightness I 21.7
  mag/arcsec2 (Freemans law, see Binny and
  Merrifield, Galactic Astronomy pg. 221)
• Mag 0 is 2e10 photons/sec in NIR filters
  (Michael Richmond website)
• Mag 21.7 41 photons/sec/arcsec2
• Each pixel sees 1.2 photons/sec/pixel 360
  photons per 300s exposure for a typical galaxy
• There are a few very bright galaxies and/or AGN
  that do not obey Freemans law that need to be
  accounted for, but these will only impact a small
  number of pixels (similar to the point sources)
• Plan Use Hubble Deep Field to predict brightness
  distribution for SNAP NIR detectors

49
Persistence Distribution
Assume persistence is 0.1 (flux
independent) Simulate 500 x 500 pixel NIR
detector with galaxies USNO-B stars 1 (2400
pixels) have gt 1e- persistence 0.1 (286 pixels)
have gt 10e-
CAUTION The flux normalization in pixsim is in
progress I assume the brightest pixel in the
galaxy image is 100K photons - Equivalent to a
mag 18 galaxy spread over 6 pixels This is only
a demonstration of the method to estimate the
total number of pixels with persistence gt Read
Noise (10e-) The brightest objects are stars
these pixels can easily be masked during ground
processing
0.1 pixels gt 10e-
50
HgCdTe 141 Drift Corrections
Uncorrected

• For RVS detectors Reference pixels do not
  correlate or correct the scatter observed in dark
  current ramps
• - Rockwell ref. pixels do (see R. Smith)
• Using just 1 of the active pixels in a 100,000
  pixel region the scatter is greatly reduced

e-
After correction
Before correction
e-
Raw Data - Fit