The Hobby-Eberly Telescope D ark Energy Experiment (HETDE X)
Karl Gebhardt, Gary Hill, Phillip MacQueen Eiichiro Komatsu, Niv Drory, Povilas Palunas
McDonald Observatory & Department of Astronom y, University of Texas
Peter Schuecker, Ralf Bender, Uli Hopp, Claus Goe ssl, Ralf Koehler
MPE and Uni-Sternwarte Munich Martin Roth, Andreas Kelz
HET
Mt. Fowlkes west Texas
Hobby-Ebery Telescope (9.
2m)
Goals for HETDEX
• HETDEX measures redshifts for about 1 mill ion LAEs from 2<z<4
•Wavelength coverage: 340-550 nm at R~800
• Baryonic oscillations determine H(z) and Da (z) to 1% and 1.4% in 3 redshift bins
• Constraints on constant w to about one per cent
• Tightest constraints on evolving w at z=0.4 ( to a few percent)
Ly- emitters as tracers
Properties of LAEs have been investigated through NB imaging
Most work has focused on z ~ 3 – 4, little is known at z ~ 2
Limiting flux densities ~few e-17 erg/cm2/s
They are numerous
A few per sq. arcmin per z=1 at z~3
But significant cosmic variance between surveys
5000 – 10000 per sq. deg. Per z=1 at z~3
Largest volume MUSYC survey still shows significant variance in 0.25 sq. deg ree areas
Bias of 2 – 3 inferred
Basic properties of LAEs would make them a good tracer if they could be detected with a large area integral field spectrograph units (IFUs)
Has the advantage of avoiding targeting inefficiency
VIRUS
Visible IFU Replicable Unit Spectrograph
Prototype of the industrial replication concept
Massive replication of inexpensive unit spectrograph cuts costs and development time
Each unit spectrograph
Covers 0.22 sq. arcmin and 340-550 nm wavelength range, R=850
246 fibers each 1 sq. arcsec on the sky
145 VIRUS would cover
30 sq. arcminutes per observation
Detect 14 million independent resolution elements per exposure
This grasp will be sufficient to obtain survey in ~110 nights
Using Ly- emitting galaxies as tracers, will measure the galaxy po wer spectrum to 1%
Prototype is in construction
Delivery in April
Layout of 145 IFUs w/ 1/9 fill
(20’ dia field) New HET wide field corrector FoV
0.22 sq. arcmin
Layout with 1/9 fill factor is optimized for HETDEX
IFU separation is smaller than non-linear scale size
LAEs are very numerous so no need to fill-in – want to maximize area (HETD EX is sampling variance limited)
Well-defined window function
Dithering of pointing centers removes aliasing
VIRUS on HET
145 VIRUS units will be housed in two “saddle
bags” on the HET frame
Fiber feed allows offloading of the mass of the
instruments to this location
VIRUS on HET (detail)
HET will be upgraded with a new wide field corrector with 22 arc-minute field of view
Substantial upgrade: 3.5 arc-minutes 22 arc-minutes
New tracker and control system
Optical design of VIRUS module
Spherical collimator mirror
VPH Grating 115 mm beam
f/1.33 Camera 2kx2k CCD
Science driver requires coverage of 340-550 nm at R~800
Very few elements, simple to set up
Image quality easily meets spec
With dielectric mirror coatings (340-6 80 nm) expect 70% thorughput
Complexity of internal focus camera
Flat mirror
VIRUS Prototype Unit Spectrograph
Will be completed this s ummer
Tests the design and p erformance of the instr ument
Refines the cost estima te for replicating VIRUS Will be used for a 50 ni ght pilot survey of LAEs on the McDonald 2.7 m
Lyman-α Emitters
There are ever increasing num ber of observations on LAE
Compilation of the recent data and GALFORM modeling by D elliou et al. (2005)
Most recent data very consistent
“Theory” and data matching well
Not very reliable, but useful st arting point to design surveys
More accurate number counts will be obtained from VIRUS pr oto-type.
“Predicted” Number Counts
Sensitivity of VIRUS (5-)
2e-17 erg/cm2/s at z=2
1e-17 erg/cm2/s at z=3
0.8e-17 erg/cm2/s at z=4
Detected # LAEs approximately constant with redshift
sensitivity tracks distance modulus
predict ~5 / sq. arcmin = 18,000 / sq. deg. per z = 1
With z~1 and 1/9 fill factor, expect 3,000 LAEs/sq. degr ee
0.6 million in 200 sq. degrees
Sufficient to constrain the position of the BO peaks to <1%
HETDEX will require ~1100 hours exposure or ~110 goo d dark nights
Needs 3 Spring trimesters to complete (not a problem: HET is O UR telescope!)
Experimental Requirements
A LAE DE survey reaching <1% precision requires:
large volume to average over sample variance
200-500 sq. degrees and z ~ 2
this is 6-15 Gpc3 at z~2-4
surface density ~3000 per sq. degree per z=1; ~1 M galaxies
LAEs have 18,000 /sq. deg./z=1 at line flux ~1e-17 erg/cm2/s
only require a fill factor of ~1/9 to have sufficient statistics
so we can trade fill factor for total area
lowest possible minimum redshift (bluest wavelength coverage)
z = 1.8 at 3400 A is a practical limit
ties in well with high redshift limit of SNAP and other experiments
These science requirements determined the basic specifi cations of VIRUS
Status of HETDEX
• The prototype VIRUS unit is being built and will be on the McDonald 2.7m in Aug 2006, wit h 50 night observing campaign
• Pilot run on Calar Alto in Dec saw 4 hrs data i n 8 nights, but we will go back
• Full VIRUS is in design phase; with full fundi ng expect completion 2008-2009
• HETDEX will then take 3 years, finishing 2011 -2012
• $30M project (including operation cost and d ata analysis): $15M has been funded.
HETDEX Uncertainties
Current HETDEX design N/2
•HETDEX is sampling variance limited; thus, the exact number of objects does not matter too much.
H(z), Da(z), and w(z)
dz
z z z w
h z H
z X
m
0 3
1 ) ( 3 1
exp )
1 ( )
(
Point: The integral
dependence of H on w allows low-z constraints from high-z observations
HETDEX Measures w(z) at z~0.4
HETDEX
SNAP
Popular parameterization is:
) (
)
(a w0 w a a
w a p
It is important to choose the appr opriate pivot point to overcome de generacies.
Beyond w
a: Non-parametric Estima te of w(z)
dz
z z z w
h z H
z X
m
0 3
1 ) ( 3 1
exp )
1 ( )
(
2 2
2
Minimize
w
S d
S
L
From data to w(z)
H(z) more powerful than Da(z)
From data to w(z)
Non-linearity in BAO
E. Komatsu
Modeling Non-linearity:
3
rd-order Perturbation Theory
Suto & Sasaki (1991); Makino, Sasaki & Suto (1992)
PM code, 2563 particles
256/h Mpc box 512/h Mpc box
(70 sims averaged)
(22 sims averaged)
3PT prediction
Peacock&Dodds 96
Linear Theory
Error<1% at z>2!! 3PT is much better than PD96 even at z=1
Jeong & Komatsu (to be submitted)
Kaiser Effect + 3PT
Since we are measuring redshifts, the measured clustering length of galaxies in z-direction will be affected by peculiar velocity of galaxies.
Also known as the “redshift space distortion”.
Angular direction is not affected by this effect.
The clustering length in z-direction appears shorter than actually is.
z direction
In the linear regime, PPPwhich gives the original Kaiser formula in the linear regime. (=velocity divergence fiel d)
Work in progress (2): Non-linea r Bias
The largest systematic error is the effect of galaxy bi as on the shape of the power spectrum.
It is easy to correct if the bias is linear; however, it w on’t be linear when the underlying matter clustering i s non-linear.
How do we correct for it?
Non-linear Bias:
3
rd-order Perturbation Theo
ry
Powerful Test of Systematics
Work in progress (3):
Three-point Function
Status and Plans
VIRUS prototype is in construction
Will be used for pilot survey to establi sh properties of LAEs this fall
HET wide field upgrade is mostly f unded
Private fundraising for VIRUS is conti nuing
$30M total funding goal with $15M in hand
2009 start for survey with funding
3 years to complete
Why LAEs?
Target-selection for efficient spectroscopy is a challenge in measu ring DE with baryonic oscillations from ground-based observations
LRGs selected photometrically work well to z~0.8
High bias tracer already used to detect B.O. in SDSS
Higher redshifts require large area, deep IR photometry
Probably can’t press beyond z~2
Spectroscopic redshifts from absorption-line spectroscopy
[OII] and H emitters can work to z~2.5 with IR MOS
But difficult to select photometrically with any certainty
Lyman Break Galaxies work well for z>2.5
Photometric selection requires wide-field U-band photometry
Only ~25% show emission lines
Ly- emitters detectable for z>1.7
Numerous at achievable short-exposure detection limits
Properties poorly understood (N(z) and bias)