2. What Do We Want to Know?
• What is the universe made of (the answer may surprise you)?
• How much stuff is there in the universe, anyways?
• When was the Big Bang?
• How quickly is it expanding?
• What’s going to happen to it?
• Can we say anything about physics in the very, very, very
young & hot universe? (when it was far hotter than anything
we can make with particle accelerators today)
3. The Plot Heard Round the Universe
• Edwin Hubble discovers in 1929 that
galaxies are moving away from us.
• Their speed is proportional to their
distance.
• Modern cosmology began with this
simple law.
Left: Edwin Hubble!
Above: First measurement of
distance-velocity relation.
4. Big Bang Prediction
• Big Bang would also have produced photons.
• Earlier in time, denser universe means hotter
temperature.
• Alpher & Gamov looked at helium in universe, saw there
was too much for stars to have made it all.
• Helium could have been made in Big Bang - if so, residual
photons should mean ~5 Kelvin temperature wherever
we look.
5. Discovery of the Cosmic Microwave Background
Penzias & Wilson (Bell Labs) discovered
excess radiation it while building & testing
antennas.
!
First thought it might have been a “white
dielectric substance.” Tried many things,
but couldn’t get 3K to go away. Showed
up everywhere they pointed in the sky.
!
Wandered up the road to Princeton to ask
if anyone there had any ideas. Dicke:
“Boys, we’ve been scooped.” Nobel prize in
1978
4.08 GHz (7.35 cm) ! “microwave”
6. Frequency Spectrum of the Microwave
Background
COBE Satellite measured temperature of sky in lots of direction.
Universe almost perfectly smooth. Biggest difference, at 0.1%
caused by our motion through universe.
This plot has error bars!
7. Nearly There…
• Basic picture of cosmology mostly in place. Not quite, though.
• Two classic problems in cosmology:
• Horizon - temp./density same in all directions (to 10 ppm). Light from
my right edge hasn’t had time to get to my left edge. Why so similar?
• Flatness - universe very closely balanced between expansion/
contraction. If I throw a ball up, either it comes back down, or leaves
earth orbit. 13.8 billion years later, still to close to call?
Matter curves space, bending light rays.
Critical expansion means light rays stay
parallel.
8. Inflation
• Phase changes in early universe generically give rise to huge
expansions.
• Expansion takes something the size of a proton to something
the size of the solar system in ~10-27 seconds.
• Expansion drives metric to flat, like blowing up a balloon.
• Because universe was tiny before inflation, plenty of time to
reach thermal equilibrium.
An ant on a balloon that inflates.
Before, the balloon looks curved.
Afterwards, it looks flat.
9. Inflation Predicts
• Heisenberg Uncertainty Principle: I can’t tell you where
something is and how fast it is going simultaneously.
• Alternative: I can’t tell you how much energy something had and
when it happened simultaneously.
• Energy equals mass (E=mc2), so energy uncertainty equals mass
uncertainty.
• Inflation freezes in mass uncertainties. Makes predictions about
what they should look like. In inflation, the largest things in the
universe come from the Heisenberg Uncertainty Principle.
11. The First 500,000 Years
• Inflation (or something similar) happens after big bang -
sets initial conditions, predicts “almost” scale-invariant
density spectrum.
• Universe expands, cools. ~25% of hydrogen burned to
helium in first several minutes. (lithium made)
• Perturbations from Heisenberg Uncertainty Principle
evolve under gravity/photon pressure.
• T~3000K (~400k years) electrons/protons combine to
form hydrogen. Universe becomes transparent.
• These photons come (almost) straight towards us. T now
2.72548±0.00057. Observe with mm telescopes.
• Temperature, density uniform to ~10 PPM → linear physics
(unlike stars, galaxies) Initial conditions+universe contents
+physics=quantitative statistical prediction.
12. What do We Expect to See?
• Sound! Matter wants to collapse under gravity.
• Because universe is opaque, the matter drags photons
with it. Photon pressure provides restoring force
• This leads to sound waves. All waves start as density
perturbations at inflation, so are in phase. Long waves
take longer to evolve than short waves.
• If we stop at any point in time, some waves will be at max
amplitude, some at zero.
13. What do we Expect to See?
Sound! Driven by gravity, photon
pressure. Normal modes sine waves,
like piano strings. (If strings were
longer than size of universe)
Sound speed=c/√3
Inflation: Everything starts at the same
time. Longer waves take longer to
oscillate.
Because we see a surface, we see a
snapshot in time.
WMAP and
the CMB sky.
14. Wave Amplitudes vs. Wavelength
(Power Spectrum)
Plot mean variance of
waves vs. k to get power
spectrum. Physics affects
the “sound quality” of the
universe.
!
Axes: Horizontal is l, full
moon is l~400. Vertical
noise in μK2. Typical
signals are a few to a few
dozen millionths of a
degree.
!
This is space in which data
and theory are compared.
“Cosmic Graphic Equalizer”
larger smaller
15. What This Looks Like: ns
!
Inflation sets the initial amplitude
of the fluctuations. Inflation
predicts almost scale-invariant
noise, but maybe with slightly
more amplitude on larger scales.
!
We call this parameter ns - if we
can measure it, we learn about
inflation!
!
!
!
ns=2
ns=1
ns=0
16. Initial Slope (ns) of Power Spectrum in PS
Change in power spectrum as
initial slope changes.
!
200 600 1000 1400 1800
l
17. Some Parameters: Regular Matter Density
Matter wants to collapse. It
drags light with it, but the light
doesn’t want to be squeezed.
The more matter there is, the
more it can compress the
photons. So, more baryons
means brighter patches.
!
Photons also spread out (“Silk
damping”). More electrons
means less spreading, and so
power on small scales falls off
less quickly.
!
(spectrum also falls off since
we’re really averaging over a
finite-thickness shell)
18. Some Parameters: Dark Matter Density
Dark matter doesn’t scatter light, so
it falls right through the photons.
So, no pressure means the dark
matter just collapses.
!
Dark matter tries to pull baryons
with it through gravity, so 1st, 3rd
etc. peaks, DM works with baryons,
2nd, 4th etc. peaks, DM works
against baryons.
!
Lots of DM + lots of baryons = big
1st, 3rd peak, smaller 2nd, 4th... Can
basically read off baryon/DM ratio
from relative even/odd peak
heights.
20. Pontificia Universidad Catόlica de Chile
University of Oxford
Stony Brook University
West Chester University of Pennsylvania
National Aeronautics and Space Administration
Goddard Space Flight Center (NASA GSFC)
University of British ColumbiaInstituto Nacional de
Astrofisica, Óptica y Electrónica (INAOE)
Carnegie Mellon University
University of Pennsylvania
Haverford College
Institute for Advanced Study (IAS)
National Institute of Standards and Technology
University of California, Berkeley
Canadian Institute for Theoretical Astrophysics
(CITA)
Princeton University
Cardiff University
University of Michigan
University of KwaZulu-Natal
University of Miami
University of Pittsburgh
Academia Sinica
Rutgers, The State University of New Jersey
Cornell University
The Johns Hopkins University
PhRvD 87, 3012 JCAP 10,60 JCAP 7,25 arXiv:1301.1037
23. ACT: Data Challenge
Requirements: Unbiased sky
estimate (need ~1% signal accuracy).
Optimal (data is precious). Ability
to handle complex noise. Fast - ACT
has a ton of data.
Go from
to
24. (Linear) Least Squares
2 = (d hdi)TN1(d hdi)
!
• Least squares has formed core of data analysis
for 200 years, going back to Gauss.
• First use was to rediscover Ceres. Been in
constant use ever since.
• ACT challenge:
• N is very complicated
• A is very complicated
• Have to invert huge matrices to solve for
~107-1010 parameters from 1012 data points.
25. 25
SciNet
@UofT:
GPC: 3780 nehalem
nodes=30240 cores 306
TFlops debut as #16 in
Top500
1 Rack: 692 cores, 7
TFlops, 1.3 TB RAM.
Mapping one ACT frequency from
one season takes ~13 CPU-years.
We’ve used 25 million CPU- hours.
Solve for 1010 params from 1012
data points.
28. Marginalised CMB-only
likelihood
Data from 4 totally
independent
experiments. There
is a model curve
under there - the
data are so precise
that you can’t see
the curve!
Restricting the range l 3500 where
the Cls are Gaussian – marginalize
over the secondary parameters!
29. Baryon Density
• Through the CMB, we
measure the density of
ordinary matter (baryons)
to an accuracty of 1%.
Ordinary matter makes up
4.82±0.06% of the
universe.
30. Dark Matter
30
• Through the CMB, we
measure the density of
dark matter to 25.8±0.4%
of the universe.
• What is the dark matter?
No idea! Would really like
to know…
• “But wait” I hear you say.
“25.8 + 4.8 only adds up
to 30%. What’s the rest?”
Also a good question.
Also would really like to
know… (cosmological
constant? Something that
evolves?)
31. ns
31
• Combination of CMB data measure ns to be
0.9614±0.0063, 6σ away from 1.
• Remember - inflation predicts ns to be a little bit
less than one. This is pretty strong evidence that
inflation happened!
•
r0.11
32. Matter Power Spectrum/Effective # Neutrinos Neff
Tegmark 03
!
Matter PS is a fundamental quantity. Observed CMB is matter PS +
snapshot effects + baryon physics. Others also measure matter PS,
nonlinear structures are biased tracers.
!
Perturbation stop growing when subhorizon during radiation-dominated
era, kink gets frozen in at transition to matter
dominated. Angular scale tells redshift of equality.
l
33. Effective relativistic species
Any extra light
particles would leave
a signature in the
CMB. We detect the
three neutrino
species we expect, but
don’t currently see
anything else.
neff=3.30+0.54-0.51
NB: highL=ACT+SPT(K11)
34. Particle Equilibrium in Big Bang
• In early universe, particles in thermal equilibrium.
• nx ~nγ if kTmxc2.
• As T drops, particles decouple from photons. For stable particles,
if Tdecouple mxc2, then final particle density set by temperature at
Tdecouple.
• If Tdecouple mxc2, density suppressed by exp(-mxc2/kT)
• Energy from these particles gets converted into photons.
• neutrinos decouple above mec2, so energy from e+e- goes into
photons, not neutrinos.
• Means photons hotter than neutrinos, Tν~(4/11)1/3Tγ
• Temp. of any light stable particle suppressed vs. γ by # of species
that annihilate after particle decoupling.
35. Gravitational lensing of the CMB
Intervening large-scale
potentials deflect CMB photons
and distort the CMB.
The RMS deflection is about 2.7
arcmins, but the deflections are
coherent on degree scales.
Lensing picks up intervening
matter power spectrum.
Jonathan Sievers, Columbia, March 7 2011
36. MASSIVE NEUTRINOS SUPPRESS
STRUCTURE FORMATION ON SMALL
SCALES
FS ⌧ FS
Jonathan Sievers, Columbia, March 7 2011
Graphics from Y. Wong
37. Effective relativistic species
Any extra light
particles would leave
a signature in the
CMB. We detect the
three neutrino
species we expect, but
don’t currently see
anything else.
neff=3.30+0.54-0.51
NB: highL=ACT+SPT(K11)
38. Fine structure constant
Claims of deviation from
unity at low redshifts z~2
using quasar absorption
spectra (Webb et al.)
!
!
!
!
!
α changes the physics of
recombination
!
WMAP7+ACT –
consistent with no
deviation at z~1100
Sievers, Hlozek et al. 2013 +Planck+BAO=0.9989±0.0037
39. Upgraded ACT to be polarization sensitive
!
Regular science observations started a year ago.
!
40. Polarization Results!
We put out first paper only 8 months
after beginning science observations!
Peaks in polarization spectrum are
very clear.
Left: UKZN postdoc Simon Muya
Kasanda and I measured phase of
polarization peaks. They are indeed
out of phase with the intensity peaks,
to an accuracy of a few degrees.
41. ACTpol Summary
• Mapping speed ~16 times ACT.
• Cover 4000 square degrees to 20 uK-arcmin,
150 to 5 uK-arcmin.
• Planck+ACTpol measures sum of neutrino
masses to ~0.06 eV - detection expected.
• Planck+ACTpol measures # of relativistic
species to 0.11
• Have first science data coming in now.
With ~2 months of data, 1/3 final # of
detectors, already close to ACT depth over
large patches.
Parameter forecasts from Galli et al.
Blue=Planck, Red=ACTpol,
Green=CMBpol (far future, unfunded).
42. Summary
• Leftover radiation from the Big Bang surrounds us, and at some
frequencies is the brightest thing in the sky.
• The universe appears to have been created in an explosion just
under 14 billion years ago we call the Big Bang. There are multiple
independent lines of very strong evidence for this.
• Also strong evidence the universe underwent inflation.
• Detailed measures of the CMB from when the universe was 400,000
years old gives us our best current handle on cosmology. We now
routinely make percent-level accurate measurements of the basic
parameters of the universe.
• Detailed measurements of the polarization of the CMB will open tell
us even more about the universe, particularly about inflation (and
particle physics).
• Advanced ACT (planned upgrade to ACTPol) got fully funded by NSF
a few months ago! Even better neutrino mass, will also constrain
formation of first generation of stars.