Title: The SuperMACHO Project: Using Gravity to Find Dark Matter
1The SuperMACHO ProjectUsing Gravity to Find
Dark Matter
- Arti Garg
- November 1, 2007
- Harvard University
- Department of Physics and Harvard-Smithsonian
Center for Astrophysics
2Outline
- What is Dark Matter?
- How can we detect DM with a telescope?
- Gravitational Microlensing
- The SuperMACHO survey
- My work
- Image-Processing Software Verification
- Microlensing Event Selection
- Follow-up Observations
- Light curve Analysis
- Simulations
- Detection Efficiency
- Contamination Rate
3Outline
- What is Dark Matter?
- How can we detect DM with a telescope?
- Gravitational Microlensing
- The SuperMACHO survey
- My work
- Image-Processing Software Verification
- Microlensing Event Selection
- Follow-up Observations
- Light curve Analysis
- Simulations
- Detection Efficiency
- Contamination Rate
4Outline
- What is Dark Matter?
- How can we detect DM with a telescope?
- Gravitational Microlensing
- The SuperMACHO survey
- My work
- Image-Processing Software Verification
- Microlensing Event Selection
- Follow-up Observations
- Light curve Analysis
- Simulations
- Detection Efficiency
- Contamination Rate
5What is Dark Matter?
- Well, we dont really know
- What we do know
- Objects in the Universe behave as if they feel
stronger gravitational forces than what the
matter we see could generate - Most of the matter in the Universe is dark
- Places where dark matter might exist
Permeating the Universe
Galaxy Clusters
Galaxy Halos
Image Credit Jason Ware
Abel 2218 (http//spaceimages.northwestern.edu/p29
-abel.html)
http//zebu.uoregon.edu/1999/ph123/lec08.html
6Galactic Halo Dark Matter
- Rotation velocities are too fast
7Andromeda Galaxy
Image Credit Jason Ware
8Radial Profile of Rotation Velocity
From http//zebu.uoregon.edu/1999/ph123/lec08.html
9Galactic Halo Dark Matter
- Rotation velocities are too fast
- Radial profile of rotation velocities suggests
spherical distribution of dark matter the Halo
10NGC 4216 in a simulated halo
Visible Galaxy Disk
Dark Matter Halo
From http//chandra.as.utexas.edu/kormendy/dm-hal
o-pic.html
11Galactic Halo Dark Matter
- Rotation velocities are too fast
- Radial profile of rotation velocities suggests
spherical distribution of dark matter the Halo - One proposed candidate for the dark matter is in
the form of MAssive Compact Halo Objects
(MACHOs) - These can be detected through gravitational
microlensing
12What is Gravitational Lensing?
- Light from a star or galaxy is bent by a massive
object between it and the observer
Virtual Light Path
Light Path
Images
Source
Observer
Lens (e.g. galaxy)
13Infrared Image of a Gravitational Lens System
Image
Lens Galaxy
HE0435-1223
From CASTLES Survey http//cfa-www.harvard.edu/ca
stles/Individual/HE0435.html
14What is Gravitational Lensing?
- This can happen even in the case where the source
is not obscured by the lens
Source
Lens
Light Path
Observer
Image
Virtual Light Path
15What is microlensing?
- In microlensing, the separation between the
source and image is too small to be resolved - The lensed object just looks brighter
- Often the source, the lens, or both are moving so
the effect is temporal - For SuperMACHO, the time scale is 80 days
16What is microlensing?
- In microlensing, the separation between the
source and image is too small to be resolved - The lensed object just looks brighter
- Often the source, the lens, or both are moving so
the effect is temporal - For SuperMACHO, the time scale is 80 days
17Microlensing
Source
Lens Trajectory
Lens
Microlensing Light Curve
Observed Source Brightness
Time
18Microlensing to Detect Dark Matter
- In 1986, B. Paczynski suggested using
microlensing to detect MACHOs by their
gravitational effect on stars in nearby dwarf
galaxies such as the Magellanic Clouds
Milky Way Halo
Us
Large Magellanic Cloud
Light Path
From http//antwrp.gsfc.nasa.gov/apod/ap050104.htm
l Earth Image Apollo 17
MACHOs
19The MACHO project (1995-2000)
0.4
- Found t of 1.2 x 10-7 (Alcock et al 2000)
- Consistent with Milky Way Halo composed of 8-50
MACHOs - Event time scales 80 days
- Recent results from EROS-2 indicate some events
were not microlensing (Miltsztajn Tisserand
2005) - Revised MACHO fraction estimate 16 (Bennett
2005) - EROS-2 find a MACHO fraction of lt7 (Tisserand et
al. 2006)
- 0.3
Contamination
(Alcock et al. ApJ 542, 281 2000)
20The SuperMACHO Project
- SuperMACHO is a 5 year survey of the Large
Magellanic Cloud (LMC) to search for microlensing
events - Fifth season of observations completed in January
2006 - Observations conducted between Oct Jan on the
Cerro Tololo InterAmerican Observatory (CTIO) 4m
Blanco Telescope in Chile
21SuperMACHO Project
- More events
- CTIO 4m
- Mosaic imager big FOV
- 150 half nights over 5 years
- Completed Jan 2006
- blocks of 3 months per year
- Observe every other night in dark and gray time
- Single Filter custom VR-band
- Spatial coverage
- 68 fields, 23 sq deg.
- Difference Imaging
22CTIO Blanco 4m
23SuperMACHO fields
Primary field set Secondary field set
24SuperMACHO Team
- Harvard/CfA Arti Garg, Christopher W. Stubbs
(PI), - W. Michael Wood-Vasey, Peter Challis, Gautham
Narayan - CTIO/NOAO Armin Rest1, R. Chris Smith, Knut
Olsen2, Claudio Aguilera - LLNL Kem Cook, Mark E. Huber3, Sergei Nikolaev
- University of Washington Andrew Becker,
Antonino Miceli4 - FNAL Gajus Miknaitis
- P. Universidad Catolica Alejandro Clocchiatti,
Dante Minniti, Lorenzo Morelli5 - McMaster University Douglas L. Welch
- Ohio State University Jose Luis Prieto
- Texas AM University Nicholas B. Suntzeff
- Now Harvard University, Department of Physics
- Now NOAO North, Tucson
- Now Johns Hopkins University
- Now Argonne National Laboratory
- Now University of Padova
25Outline
- What is Dark Matter?
- How can we detect DM with a telescope?
- Gravitational Microlensing
- The SuperMACHO survey
- My work
- Image-Processing Software Verification
- Microlensing Event Selection
- Follow-up Observations
- Light curve Analysis
- Simulations
- Detection Efficiency
- Contamination Rate
26Image Reduction Pipeline
- Implemented in Perl, Python, and C
- Images processed morning after observing
- Stages of image processing
- Standard calibration (bias, flat field)
- Illumination correction
- Deprojection/Remapping (SWARP)
- Regular Photometry (DoPhot)
- Difference Imaging
- Photometry on Difference Images (Fixed PSF)
27Image Reduction Pipeline
- Implemented in Perl, Python, and C
- Images processed morning after observing
- Stages of image processing
- Standard calibration (bias, flat field)
- Illumination correction
- Deprojection/Remapping (SWARP)
- Regular Photometry (DoPhot)
- Difference Imaging
- Photometry on Difference Images (Fixed PSF)
28Empirical corrections to Difference flux errors
Error ratio of random positions on difference
image
sma5,amp 7 Jan 6, 2006
m -0.04 s 1.54
Should be 1.0, Errors are Underestimated!!
difference flux/flux err
29Empirical corrections to Difference flux errors
- Histograms of typical distributions of FDFSIG
- FDFSIG standard deviation of flux/flux err for
a grid of random positions in a difference image
(image keyword)
sm54 all amps
amp 1, all fields
FDFSIG for image
FDFSIG for image
m 1.47 s 0.2
m 1.43 s 0.2
30Outline
- What is Dark Matter?
- How can we detect DM with a telescope?
- Gravitational Microlensing
- The SuperMACHO survey
- My work
- Image-Processing Software Verification
- Microlensing Event Selection
- Follow-up Observations
- Light curve Analysis
- Simulations
- Detection Efficiency
- Contamination Rate
31Microlensing Event Selection
- Detecting microlensing
- We monitor tens of millions of stars in the Large
Magellanic Cloud - Tens of thousands of those appear to change
brightness - Need to determine whether those changes are
- Real, and not an artifact or cosmic ray
- Due to microlensing, or some other phenomenon
32Microlensing Event Selection
- Detecting microlensing
- We monitor tens of millions of stars in the Large
Magellanic Cloud - Tens of thousands of those appear to change
brightness - Need to determine whether those changes are
- Real, and not an artifact or cosmic ray
- Due to microlensing, or some other phenomenon
33Microlensing Event Selection
- Microlensing causes the brightness of a star to
change in a predictable way
Brightness
Time
34Microlensing Event Selection
- But many other things also change in brightness
such as supernovae - these turn out to be much more common
Brightness
Time
35Microlensing Event Selection
- And if your nights off from the telescope and the
weather conspire in the wrong way, its hard to
tell whats microlensing
36Challenges to Event Classification
- High volume of events
- Need sophisticated software tools (25 million
stars) - High rate of contamination
- Supernovae outnumber microlensing by up to 10
times - Gaps in sampling and low S/N
- No bright time (near full moon) observations
- Difficult to discriminate microlensing from other
phenomena
37Microlensing Event Selection
- So what do you do?
- You get a graduate student!
- Follow-up Observations
Magellan III 6.5m Telescopes
38Microlensing Event Selection
- So what do you do?
- You get a graduate student!
- Light Curve analysis tools
39Outline
- What is Dark Matter?
- How can we detect DM with a telescope?
- Gravitational Microlensing
- The SuperMACHO survey
- My Work
- Image-Processing Software Verification
- Microlensing Event Selection
- Follow-up Observations
- Light curve Analysis
- Simulations
- Detection Efficiency
- Contamination Rate
40Follow-up Program
- Developed computational tools and protocols for
analyzing many GBs of nightly CTIO observations
in almost real time to pick out interesting
events and prioritize them for follow-up
observation - Follow-up is time critical because events are
only active for a few weeks - Applied for many nights of Magellan telescope
time to follow interesting events as we
discovered them at CTIO
41Classifying events using follow-up
- Spectroscopic Observations
Intensity
Intensity
Wavelength
Wavelength
Source http//homepages.wmich.edu/korista/sun-im
ages/solar_spec.jpg
Spectrum of the Sun, a typical star (How
microlensing might look)
Spectrum of a supernova
42Magellan Spectroscopy
- Spectroscopy achieved to m21.5
- Positive identification of
- Supernovae (type Ia and type II), AGNs, CVs,
other Variable Stars - Many other objects with uncertain spectroscopic
identifications but definitely extragalactic
43SM-2004-LMC-821
VR21
Spectral classification Broad Absorption Line AGN
44Classifying events using follow-up
- Spectroscopy is an excellent way to classify an
event, but... - It is time-consuming and cant be done for faint
events - Obtaining a spectrum for every interesting event
is not feasible
45Classifying events using follow-up
- Multi-band observations - poor mans
spectroscopy
46(No Transcript)
47Classifying events using follow-up
- Multi-band observations - poor mans
spectroscopy - The ratio of brightness in different filters
gives a crude measure of the events wavelength
spectrum - The ratios for vanilla stars (i.e.
microlensing) differ from supernovae - This method is less precise but can be used for
faint events
48Stars have characteristic ratios of filter
intensities
49Outline
- What is Dark Matter?
- How can we detect DM with a telescope?
- Gravitational Microlensing
- The SuperMACHO survey
- My work
- Image-Processing Software Verification
- Microlensing Event Selection
- Follow-up Observations
- Light curve Analysis
- Simulations
- Detection Efficiency
- Contamination Rate
50A light curve describes an objects brightness as
a function of time
Brightness
Time
51Light Curve Analysis
- Why do we need it?
- Only have follow-up for 2 out of 5 years
- Follow-up is incomplete and sometimes
inconclusive - What is it?
- Software analysis tools that calculate 50
statistics describing the light curve - Unique?
- Significant and Well-sampled?
- Microlensing-like?
- Unlike other things?
52Unique?
-Frequent and periodic variability
-Year-to-Year change in baseline
Active Galactic Nucleus (AGN)
Variable Star
53Significant and well-sampled?
-Need more data after peak
54Microlensing-Like?
-This is a Supernova
55Unlike other phenomena?
-Fit well by microlensing and supernova models
56Passes all Criteria
57Outline
- What is Dark Matter?
- How can we detect DM with a telescope?
- Gravitational Microlensing
- The SuperMACHO survey
- My Work
- Image-Processing Software Verification
- Microlensing Event Selection
- Follow-up Observations
- Light curve Analysis
- Simulations
- Detection Efficiency
- Contamination Rate
58Light Curve Analysis
- Estimating how many microlensing event we can
detect (i.e. Detection Efficiency) - Simulate a large number of microlensing events of
all possible combinations of event parameters - source brightness, event duration, and
amplification - Determine which of these events survive selection
criteria - Estimating how many events we should expect
- Multiply by distribution of event parameters
consistent with various MACHO models to get
expected number of microlensing events
59Simulations
- Allows optimal tuning of selection criteria
- Allow the most microlensing events while
rejecting the most contaminants - Provides estimate of contaminant fraction
- Provides quantitative estimate of detection
efficiency - Fraction of simulated events that are recovered
- Differences between simulated population and
recovered population - Estimate how many events we should expect from
various models - Multiply by distribution of event parameters
consistent with various microlensing models to
get expected number of microlensing events (Rest
et al. 2005)
60Simulations
- Simulate a large number of events
- Microlensing all combinations of source star
brightness, event duration, and amplification - Determine which events survive selection criteria
- ? Detection Efficiency
- Supernovae all combinations of redshift,
extinction by dust, intrinsic shape - Determine which events survive selection criteria
- ? Contamination Rate
61Simulations
- Obtain light curves for a grid of positions
across our field-of-view - Add simulated event to each position
- Can add multiple events to the same light curve
- We simulated 57 million ML events and 4 million
SNe
62Simulations
Simulations of Microlensing events
Simulations of Supernovae
63Detection Efficiency Depends on Source Brightness
Simulated
Number of events
Recovered
Source Brightness (-2.5log(Intensity))
64Next Steps
- We are finalizing our selection criteria
- Final set of Candidates
- Final Detection Efficiencies
- Final Contamination Rate
- We will distinguish between microlensing models
by comparing the predicted rate of ML events with
our observed rate
65Summary
- Most of the matter in our Galaxy is dark
- We can detect Dark Matter with gravitational
lensing
66Summary
- SuperMACHO searches for Dark Matter in the form
of MACHOs in the Milky Way - Gravitational microlensing is easily confused
with other things
67Summary
- Additional observations and light curve analysis
improve event classification - Simulations allow for estimation of detection
efficiency and contamination rate
68Lens Equation
Source Blandford Narayan 1986
(Mollerach Roulet 2002)
69Microlensing
rE projection of qE at lens distance
u impact parameter
Lens Trajectories
Magnification Due to Lensing Event
source
Source Michael Richmond (RIT)
Source Paczynski 1991
70Microlensing Light Curve
fo x Amax a umin closest approach
to time of maximum brightness
t characteristic time ( )
Flux
fo baseline source flux
Time
71Observables for Event Ensemble
Ensemble of events has a uniform distribution of
umin
- t Optical depth toward source population
- likelihood that a source is within rE of a lens
at any time
(Mollerach Roulet 2002, Alcock et al. 2000)
G Distribution of
(Mollerach Roulet 2002)
72The MACHO project (1995-2000)
0.4
- Found t of 1.2 x 10-7 (Alcock et al 2000)
- Consistent with Milky Way Halo composed of 8-50
MACHOs - Event time scales 80 days
- Recent results from EROS-2 indicate some events
were not microlensing (Miltsztajn Tisserand
2005) - Revised MACHO fraction estimate 16 (Bennett
2005) - EROS-2 find a MACHO fraction of lt7 (Tisserand et
al. 2006)
- 0.3
Contamination
(Alcock et al. ApJ 542, 281 2000)
73SuperMACHO Project
- More events
- CTIO 4m
- Mosaic imager big FOV
- 150 half nights over 5 years
- Completed Jan 2006
- blocks of 3 months per year
- Observe every other night in dark and gray time
- Single Filter custom VR-band
- Spatial coverage
- 68 fields, 23 sq deg.
- Difference Imaging
74RR Lyrae from MACHO (black) and SuperMACHO (red)