Transcript talk at lensing and dark matter conference Ohio state 2004

Microlensing and Dark Matter
Jan 2005
Kim Griest, UCSD
•Surveys monitor millions of
stars for years to find
rare lensing events
• Bulge => stars, remnants,
planets, etc.
•LMC/SMC/M31 => DM
Microlensing of Dark Matter?
5 collaborations have returned dark matter results
• MACHO: strong evidence toward LMC, but
interpretation unclear
• EROS: evidence against toward LMC/SMC, but
not inconsistent with MACHO
• MEGA: moderate evidence in favor toward M31
• POINT/AGAPE: weak evidence against toward
M31, consistent with MACHO
• WeCapp, very weak evidence in favor (M31)
MACHO Collaboration (2000)
•
•
•
•
•
Monitored 11.9 million stars for 5.7 years
Found 13-17 events (depending on selection criteria)
Careful efficiency analysis including blending
Removed 8 Supernova behind LMC (contaminants)
Distribution in space, CMD, Amax, consistent with
microlensing interpretation
• Likelihood analysis to measure Macho DM, plus
events in disk, LMC, etc.
LMC in
neutral H
looks like
a face-on
disk.
Test of
systematic
error due to
contamination,
selection bias:
compare A
B criteria
Criteria A:
tighter cuts,
with less
contamination
Criteria B:
looser cuts,
with more
contamination
•Masses 0.1 1.0 Msun
preferred
•Halo fraction 8%
- 40% preferred
•Total mass in
Machos: 8-10
10^10 Msun
(MW disk=6
10^10 Msun, and
MW halo has 4-6
10^11 Msun)
•Optical depth =
1.2+0.4-0.3 10^-7
• Main conclusion: Macho’s as main
component of Dark Matter are ruled out
• But found significant extra microlensing
The number of non-Macho events is predicted to be much
smaller than the 13-17 events observed (using standard LMC
and Milky Way stellar populations.)
But these results need correcting
• Recently EROS (Glicenstein 2004) found that
event LMC-23 bumped again after 7 years =>
variable star, not lensing.
• LMC-23 contributed 8% of optical depth (and
halo fraction) (6% for set B), so all our optical
depths and halo fractions should be reduced by 8%
• => best f is 18.5%, and tau=1.1 10^-7
• More worrying: are there more events like this?
LMC-23
What does extra LMC
microlensing mean?
1. If events are in MW halo =>
- significant portion of DM
- problem exists: What are they?
-- stellar mass but can’t be stars (stars shine!)
-- stellar remnant (white dwarfs, black holes) would need
lots of early stars: no evidence for these (metal
enrichment, background light, etc.) WD observed?
-- primordial black holes? quark nuggets?
2. If events are LMC self lensing =>
- current LMC models wrong?
- lens stars should be seen?
3. Contamination in MACHO dataset?
Much written on LMC self lensing since Sahu/Wu/Gould 1994
•MACHO used Gyuk, Dalal, Griest review of LMC models, valid in
2000, to predict 1-2 LMC self-lensing microlensing events. At that
time no evidence of other stellar populations to do the self lensing.
•HOW ABOUT RECENT EVIDENCE?
•Zhao, Ibata, Lewis, & Irwin(2003) did 1300 2dF radial velocities:
no evidence for any extra population over expected LMC and Galaxy
Any new kinematically distinct population less than 1%.
(rules out Evans & Kerrins 2000 fluffy stellar halo model)
• Gallart, Stetson, Hardy, Pont, & Zinn (2004), search for a stellar
in a deep surface brightness CMD, and found no evidence for any
stellar halo
• However, Minniti, et al (2003), and Alves (2004) found RVs for 43
RR Lyaes and discovered an old and hot stellar halo! But they say
it is too small to account for all the extra microlensing
• But the structure of the LMC is being questioned: van der Marel,et
al (2002) says the LMC disk is not circular, but Nikolaev, et al.
(2004) disagree, saying it is warped. Both say it does not probably
affect self lensing much (e.g. Mancinit etal 2003 agree), but it does
show the LMC is still not well understood.
• Summary: no clear answer yet
Contamination?
•
•
Contamination was studied by MACHO; selection
criteria:
A: 13 events, tight cuts, less contamination., lower effs
B: 17 events, loose cuts, more contam., higher effs
tau(A) =1.1e-7, tau(B)=1.3e-7.
17% difference estimates contamination systematics
But Belokurov, Evans, & LeDu used neural net to
reanalyze MACHO LMC data. Say data set is badly
contaminated; find only 6 or 7 microlensing events =>
tau much smaller => no need for either Machos in dark
halo or extra LMC self lensing!
Wrong!
• Found events by running only on our selected events, but
calculated efficiencies without including effect of our
selection => badly miscalculated efficiencies.
• Analyzed only 22000 lightcurves out of 11.9 million
• Also used very weak statistics => much lower eff, and
many false positives (2 out of 22000) => probably would
not even work if applied to all 11.9 million lightcurves
• Rejected good microlensing, misidentified SN
Conclusion: BEL analysis is meaningless; neural nets may
be useful, but have yet to be applied correctly.
Contamination possible, but certainly not shown yet.
Results of MACHO LMC5.7 stand after
small correction for LMC-23.
What do to?
Other experiments!
EROS collaboration: 4 events in 50 LMC fields and 4 events in
10 SMC fields: Interpreted as limit on Halo dark matter
LMC
Events
Combined MACHO and EROS limits on short duration = small
mass objects
Limits vary
according to
Milky Way halo
model
Limits on Macho Dark Matter
• Objects with 10-7 < m < 10-3 Msun make up
less than 25% of DM. Objects with
3.5 10-7 < m < 4.5 10-5 make up less than
10% of DM
MEGA:
M31 Microlensing
Found 4 events:
Measure Macho
halo fraction
f=0.29 +0.30 -0.13
.01< m < 1 Msun
=> M31 halo
DM consistent
With LMC result!
BUT POINTAGAPE M31
3 events says
f<.25 (.6) for
.0001<m<.1
(.1<m<1 Msun)
WeCAPP
• (Wendelstein Calar Alto Pixellensing project)
• Found 2 events toward M31
• Say favor M31 halo lenses, but evidence very
weak (in my opinion)
What does it mean?
• Experimentally not clear: need more MEGA/POINTAGAPE M31 work, Supermacho on LMC. From Space
DIME can do parallax and (if approved) can answer
question of where lenses are; eventually SIM and do
astrometric microlensing. (Measure distance to 2 or 3
LMC lenses as 10 kpc to prove Macho DM. 3 or 4 at 50
kpc proves LMC self-lensing.)
• Theoretically fairly clear: Macho DM consistent with
Omega_baryon = 0.04, but causes problems with star and
galaxy formation, or requires very exotic objects.
BULGE
Microlensing:
three
collaborations
returned
results:
OGLE,
EROS,
MACHO
Microlensing towards bulge
• 50 million stars over 7 years
• >450 events, 60 on clump giants (less blended)
• ~40 binary events, parallax, extended source,
lensing of variable stars, etc.
• Optical depth = 2.18 +.45-.38 10-6, agrees with
models (e.g. Gould and Han 1.63 10-6)
• Also found optical depth as a function of (b,l)
and gradient in optical depth
Location of all
500 events.
(b,l)=(0,0) is
Galactic center
Many of these
Are blended.
Microlensing should be randomly distributed in Color-Magnitude
Select clump giants from color-magnitude diagram: 62 events
62 Clump giant
events.
Circle size is
proportional to
event duration.
Are events all
microlensing?
Microlensing is
uniformly
distributed in
impact
parameter,
umin ~1/Amax
K-S test shows
probability of
2.5% for these
258 events.
Deviation is
from blending.
For 60 clump
giant events
probability is
81%. So these
are unblended
microlensing
34 candidate
events probably
from the
recently
discovered
Sagitarious
dwarf galaxy
The first planet to be discovered by microlensing: OGLE 2003-BLG-233
MOA 2003-BLG-53; q=.0039. Likely star mass of 0.4 Msun, likely
Planet mass of 1.5 Mjupiter.
Microlensing Planet
Finder Mission:
(Bennett et al.)
4 year mission with 1 m
Telescope 290 M pixel
focal plane, in 2 bands
Conclusion
• The mystery of LMC microlensing is still
unsolved, and more work is needed
• If you want an inventory of all compact objects,
independent of luminosity microlensing is the way
to go, i.e. Microlensing has a bright future for
finding dark objects
Light bending => split and magnify image, move images
Around, and shear image shape
Are lenses DM in Galaxy or
LMC Self lensing?
If events are in MW halo =>
- significant portion of DM
- problem exists: What are they?
-- stellar mass but can’t be stars (stars shine!)
-- stellar remnant (white dwarfs, black holes) would need
lots of early stars: no evidence for these (metal
enrichment, background light, etc.)
If events are LMC self lensing =>
- current LMC models are wrong
- why are the lens stars not seen?
Lots of tests done: none conclusive yet
[Other lensing info?]
BULGE
Microlensing:
three
collaborations
returned
results:
OGLE,
EROS,
MACHO
Microlensing lightcurves
have well specified shapes
depending on 3 parameters:
Maximum magnification:
Amax, event duration t^hat,
and time of peak.
Blended lightcurves look
very similar, but have
different values for Amax
and t^hat