Artificial Intelligence In Microbiology by Dr. Prince C P
Seeing the beginning: the Cosmic Microwave Background and what it can tell us about our Universe
1. Cosmic Microwave Background Lensing: !
Fundamental Physics from Maps of the Invisible!
Blake D. Sherwin !
Department of Applied Mathematics and Theoretical Physics / Kavli Institute for Cosmology!
University of Cambridge!
2. But Galaxies Only Trace !
Even More Underlying Dark Matter!
2!
Galaxies! Dark Matter!
<20% of the mass!
Galaxies Trace !
Large Amounts of Underlying Dark Matter!
Galaxies, Gas: <20% of the mass!
3. 3!
Dark Matter: >80% of the mass – invisible!!
Galaxies Trace !
Large Amounts of Underlying Dark Matter!
Image: Virgo Consortium!
4. Dark Matter!
• > 80 percent of matter is invisible dark matter,
distributed in large structures across the universe. !
Image Credit:!
Millenium
Simulation
(Springel et al
2005) !
4!
5. Large Scale Dark Matter Structure!
• Want to probe distribution in detail, as contains much
information on open questions in cosmology and physics –!
!
!
• Dark matter / mass distribution is what theory robustly predicts.!
6. Time!
Many Open Questions in Cosmology
• We have a very successful theory – the standard (LCDM)
cosmological model with 6 free parameters, tested to sub-
percent precision.!
!
dark matter structure
forms through
gravitational collapse!
dark energy starts
accelerating universe apart !
inflation: fast
expansion in
early universe!
7. Time!
Many Open Questions in Cosmology
• We have a very successful theory – the standard (LCDM)
cosmological model with 6 free parameters, tested to sub-
percent precision. But: many parts not understood!!
!
dark matter structure
forms through
gravitational collapse!
dark energy starts
accelerating universe apart !
inflation: fast
expansion in
early universe!
8. Large Scale Dark Matter Structure!
• Want to probe distribution in detail, as contains much
information on open questions in cosmology and physics:!
What is the
physics of
inflation and the
early universe?!
9. Large Scale Dark Matter Structure!
• Want to probe distribution in detail, as contains much
information on open questions in cosmology and physics:!
What is dark
energy?!
!
What is the
physics of
inflation and the
early universe?!
10. Large Scale Dark Matter Structure!
• Want to probe distribution in detail, as contains much
information on open questions in cosmology and physics:!
What is dark
energy?!
!
What are the
properties /
masses of
neutrinos?!
What is the
physics of
inflation and the
early universe?!
11. Large Scale Dark Matter Structure!
• Want to probe distribution in detail, as contains much
information on open questions in cosmology and physics:!
What is dark
energy?!
!
What are the
properties /
masses of
neutrinos?!
Is standard
cosmology
correct?!
What is the
physics of
inflation and the
early universe?!
12. Outline!
• Introduction to CMB lensing: mapping cosmic dark matter!
• Lensing as a powerful signal for measuring neutrino masses
(+…)!
• Lensing as noise for studying inflation and the early universe!
12!
13. Observing Dark Matter: Gravitational Lensing!
13!
• Gravity from all matter, including dark matter, deflects radiation
that passes by!
real! apparent!
observer!
dark matter!
or mass!
14. Light Source for Studying Lensing: !
Cosmic Microwave Background (CMB)
time (+ distance light travels!)!
early universe is
dense and hot
plasma!
History of the universe!
15. time (+ distance light travels!)!
early universe is
dense and hot
plasma!
afterglow
radiation still
visible today:!
!
Cosmic!
Microwave!
Background!
Light Source for Studying Lensing: !
Cosmic Microwave Background (CMB)
History of the universe!
16. !
CMB temperature fluctuations (picture of the primordial plasma from the Planck Satellite)!
“Light” Source for Lensing: !
The Cosmic Microwave Background (CMB) Radiation!
hot/bright!
only 1 part
in 104!
!
cold/dim!
17. we
are
here
!
!
d
(=projected
matter
density)
Primordial CMB (most
distant and oldest
source of radiation)!
The observable universe:!
CMB: A Distant Backlight !
On the Matter Distribution !
dark
ma*er
19. “Light” Source for Lensing: !
The Cosmic Microwave Background (CMB)!
• CMB: leftover radiation from the hot primordial plasma – most
distant observable source of light!
• Distribution of dark matter deflects light that passes through!
19!
22. 22!
!
• Original, un-lensed, CMB fluctuations. Very well understood
statistical properties, e.g., isotropy.!
CMB Lensing: An Approximate Picture!
23. 23!
• Clump of dark matter in front…!
CMB Lensing: An Approximate Picture!
24. 24!
!
• Dark matter causes lensing magnification feature in the CMB!
CMB Lensing: An Approximate Picture!
25. 25!
!
• Dark matter causes lensing magnification feature in the CMB!
CMB Lensing: An Approximate Picture!
d!
d!
d!
described by
lensing
deflection
field: d!
!
(very small:
here
exaggerated
by x ~100,
actually a
few arcmins)!
26. CMB Lensing Measurement: !
An Approximate Picture!
d: lensing deflection
field!
CMB!CMB!
Temp.!
Local CMB power
spectrum!
27. CMB Lensing Measurement: !
An Approximate Picture!
d: lensing deflection
field!
CMB!
Temp.!
Infer magnification and lensing
from “stretching” of the local
CMB power spectrum!
shift to larger angular scales!
28. Lensing Reconstruction Details!
• Lensing changes known isotropic and Gaussian statistics:
introduces non-Gaussian correlations!
!
!
• So: measure lensing by looking for these non-Gaussian
correlations in the CMB two-point function!
[Zaldarriaga,
Seljak 1997,
Hu, Okamoto
2002]!
28!
hT(l)T⇤
(l L)i ⇠ d(L)
ˆd(L) ⇠
Z
d2
l T(l)T⇤
(l L)
T: temperature (Fourier mode)!
l: wavenumber!
29. CMB Polarization Lensing Measurement !
!
Characteristic
polarization!
indicates
overdensity!
CMB Polarization map!
NO background fluctuations – all signals from lensing!
No confusion, better measurement in polarization!!
better
lensing
map!
30. What Does Lensing Tell Us?!
• Lensing probes the projected total mass density in each
direction (of which most is dark matter), all the way out!
!
!
!
!
!
: fractional mass overdensity!
= (⇥ ¯⇥)/¯⇥
30!
lensing deflection!
d(ˆn) =
Z rCMB
0
drW(r) (ˆn, r)
radial!
distance!
geometric projection kernel, peak at z~2!
31. • Lensing probes the projected total mass density in each
direction (of which most is dark matter), all the way out!
• (I think) this is remarkable!!
!
!
!
!
: fractional mass overdensity!
= (⇥ ¯⇥)/¯⇥
geometric projection kernel, peak at z~2!
31!
lensing deflection!
d(ˆn) =
Z rCMB
0
drW(r) (ˆn, r)
What Does Lensing Tell Us?!
radial!
distance!
32. !
While Galaxy Surveys Are Extremely Powerful, !
Much of the Dark Universe Remains Unexplored!
Primordial CMB (most
distant source of
radiation)!
galaxy !
survey (z<1)!
The observable universe:!
33. !
!
Lensing Probes all the Mass !
Projected Back to the CMB Emission!
" "d !
=projected!
Matter!
density!
!
!
!
from z~0.5-5! Primordial CMB (most
distant source of
radiation)!
The observable universe:!
34. !Illustration: future !
full sky lensing
map of mass!
CMB Lensing Maps Can Gravitationally Probe all
the (Projected) Mass in the Observable Universe!
Lensing maps of mass: a powerful new probe of cosmology and physics!
35. Outline!
• Introduction to CMB lensing: mapping cosmic dark matter!
• Lensing as a powerful signal for measuring neutrino masses
(+…)!
• Lensing as noise for studying inflation and the early universe!
35!
36. Physics Lensing Can Tell Us:!
Neutrinos!!
36!
• Neutrinos: 3/12
elementary particles.
Have mass –
associated physics
not well understood!
• we don’t know:!
– absolute mass = new
scale!
!
!
!
!
37. Motivation: Measuring Neutrino Mass!
37!
normal!
ordering!
inverted
ordering!
mass!
• Neutrinos have mass
– associated physics
not well understood!
• we don’t know:!
– absolute mass = new
scale!
– order of neutrino
masses (2 light, 1
heavy? 2 heavy 1
light?)!
– their own antiparticle?!
– what gives neutrinos
mass? +…!
• Cosmology:
measurement of mass
sum " can give
insights!
!
!
X
m⌫
38. Physics Lensing Can Tell Us:!
Neutrinos!!
38!
normal!
ordering!
inverted
ordering!
mass!
X
m⌫
• If we measure mass
sum, can get
insight to some of the
questions (scale,
mass ordering, type)!
!
!
!
• Part of a big program
to understand this
new physics!!
N.B. +look for extra neutrinos!!
!
X
m⌫
Target: mass
at least
>60meV !
39. • Cosmic neutrino background affects how dark matter structure
grows: νs fast so don’t clump (effect size determined by mass)!
Neutrinos in Cosmology!
Background of
fast-moving
neutrinos! Cosmic
distribution
of dark
matter!
[Yu++2016]!
40. • The more massive neutrinos are, the more small scale dark
matter structure is suppressed. !
!
!
!
!
!
• Suppression also visible in lensing map – want to measure
(and compare with primordial CMB amplitude)!!!!
!
40!
Large-scale!
mass
distribution:
Neutrinos Affect How Cosmic Structure Grows
(via gravitational collapse)!
Image:!
Viel++!
2013!
Neutrino Mass Really Large !
(qualitative)!
Neutrino Mass Negligible!
41. 41!
Atacama Cosmology Telescope (ACT)!
• Arcminute resolution CMB telescope high in the Chilean Atacama desert,
with arrays of sensitive (TES bolometer) detectors!
42.
ACT CMB Temperature Map!
!
!
!
!
!
!
CMB Lensing Reconstruction Pipeline for ACT Data!
Statistical pipeline for CMB lensing measurement!
!
Challenges: !
"- Noise bias measurement!
"- Masks and missing data " " "-Bright radio galaxies!
"- Astrophysical signal from clusters"-Infrared galaxy emission, …
[Sherwin & Das 2010]!
43.
ACT CMB Lensing Dark Matter Map (12 degrees across)!
First CMB-only Lensing Measurement: !
Mapping Dark Matter!
brightness = density!
[Das, Sherwin++ 2011!
Sherwin, Dunkley, Das++2011]!
45.
Y axis: “How much!
lensing ….”!
!
!
" " " " "X axis: “for a lens of this angular scale?”!
Describe maps
statistically with
lensing power
spectrum.!
ACT CMB Lensing Dark Matter Map!
brightness = density!
Key Observable: CMB Lensing Power Spectrum ! Cdd
46. CMB lensing map d (e.g., ACT)!strength of
lensing!
angular
wavenumber!
Key Observable: CMB Lensing Power Spectrum ! Cdd
neutrino
mass!
(>blurring)!
Y axis: “How much!
lensing2 ….”!
!
!
" " " " " " " X axis: “for a lens of this angular scale?!
Cdd
L ⇠ hd2
i ⇠ hTTTTi
NG! NG!
Measurement: non-Gaussianity in four-
point function
47. 2011 (+2013), ACT!
ACT(Pol)
• Used to confirm: evidence for Dark Energy from CMB data alone!
• 2015: POLARBEAR, first detection of lens power in polarization!
[Das, Sherwin++ 2011!
Sherwin, Dunkley, Das++2011!
see also van Engelen++2013]!
Detection of the CMB Lensing Power Spectrum!
[Physical Review: listed as one of ≈30 landmark
papers published in GR in last 100yrs]!
48. 2011 (+2013), ACT!
ACT(Pol)
• Used to confirm: evidence for Dark Energy from CMB data alone!
• 2015: POLARBEAR, first detection of lens power in polarization!
[Das, Sherwin++ 2011!
Sherwin, Dunkley, Das++2011!
see also van Engelen++2013]!
Detection of the CMB Lensing Power Spectrum!
[Physical Review: listed as one of ≈30 landmark
papers published in GR in last 100yrs]!
49. 49!
Progress in CMB Lensing Power Spectrum
Measurements 2011-now!
(+Pol.)!
Planck, 2.5% precision!
Planck!
• Rapid progress – but only just beginning!!
52. ACTPol CMB Lensing Dark Matter Map!
Now: ACTPol/AdvACT – polarization sensitive ACT!
compare older ACT lensing map!
[Sherwin et al. 2016]!
!
• New, fast, polarization lensing trispectrum pipeline. Results in a
map of the dark matter (w. small-scale noise), 50 degs. across:!
ACTPol!
53. ACTPol/AdvACT – !
Early Lensing Power Spectrum Results!
[Sherwin et al. 2016, arXiv:1611.09753]!
Amplitude relative to
standard cosmology
expectation – our lensing is
consistent!!
!
!
• Lensing power is ~clean, few % level systematics (+can do more)!
54. ACTPol/AdvACT – !
Early Lensing Power Spectrum Results!
[Sherwin et al. 2016, arXiv:1611.09753]!
Amplitude relative to
standard cosmology
expectation – our lensing is
consistent!!
!
!
• Constraint: !
X
m⌫ < 396meV (95%C.L.)
55. !
!
!
• Previously analyzed map – less than 10% of our data!!
!
Now: Very Large ACTPol/AdvACT Lensing Maps!
N.B. multiplied by mask weight > stripe!
56. !
!
!
• Total observed through 2016: 15000 sq. deg. (1st release
500). >50 to <10uK-arcmin – working on lensing extraction
now!!
!
Now: Very Large ACTPol/AdvACT Lensing Maps!
N.B. multiplied by mask weight > stripe!
BOSS-N (~3000 sq. deg.) lens map
(noisy) - preliminary!!
Sky projected to cylinder!
58. 58!
Preliminary Data Run on Small Sub-field!
~16 sigma on small
part of sky!
caveat:!
many approximations!
59. ACTPol/AdvACT – Next Paper (2018):!
Expect around 3% precision (30-40 sigma)!
[Sherwin++ in prep.]!
c.f. lower limit, >60meV!
• 4000-12000 square
degrees, 170 million
pixels, temperature
and polarization –
lensing analysis is a
challenge!
• Potential to
independently test
Planck – e.g. Mv,
tensions (?) in high vs.
low z structure, H0…!
!
[van Engelen++ in prep.]!
(
X
m⌫) ⇠ 100 meV
60. ALMA!
ACT!
CLASS!
POLARBEAR!Existing!
Notional Pads for Simons Observatory Phase 2 and CMB S4!
Simons Observatory!
Notional Simons Observatory Phase 1 !
Power!
Control !Vehicle
s!
• Next generation, funded CMB experiment,
2020-!
• Combines ACT, POLARBEAR collaborations!
63. The Future: Simons Observatory and CMB Stage-IV !
Precision Lensing Power Spectra!
Errors inflated by 10x (for display)!!
(
X
m⌫) ⇠ 20meV• Will determine unknown neutrino
mass in any scenario!
c.f. limit, >60meV!
(Simons Obs.!
/ CMB-S4)!
Simons Observatory: <1% precise lensing, half sky !
(2020-2025); CMB-S4: ~0.2% precise!
64. • CMB lensing maps are unique probes of mass at high z>1!
!
!
!
!
!
!
!
!
!
• Can correlate maps of luminous objects (galaxies, quasars,
gas) with lensing/dark matter maps (z~0.5-5, peak z~2)!
Great Potential in Astrophysics for Correlating
Galaxies, Quasars… with CMB Lensing!
Contours:
CMB
lensing!
Color: Density of infrared galaxies!
[SPT,
Holder++
2013]!
65. • Correlation restores
distance / redshift
information. !
• Can probe structure
growth and probe DE, …
(w. sample variance
cancellation!)!
!
Structure
growth!
Redshift!
Great Potential in Cosmology for Correlating
Galaxies, Quasars… with CMB Lensing!
66. Rapid Progress in CMB Lensing Measurements!
• Working to solve
challenges with
students and S.O.
lensing group!
!
• Many challenges
will be solved with
novel theoretical
work!
!
*
*
*
*
-‐
B.
Sherwin
playing
a
leading
role
in
analysis
/
planning
Signal-to-noise ratio for temperature and polarization expts. !
*
*
*
now
• Improved estimators!
• Higher order corrections!
• Astrophysical and
instrumental systematics!
(…)!
[e.g., Sherwin, Boehm, Hill
in prep., Boehm, Schmittfull,
Sherwin 2016;]!
[Horowitz, Ferraro,
Sherwin 2017]!
67. Example Challenge I: !
Astrophysical Systematics!
• Multiplicative and additive
systematic biases from bi-/
trispectra of foregrounds!
!
• Typical picture: when nothing
is done, ~few percent effect.
Multifrequency data and
estimator “hardening” should
make these unobservable!
!
• …But work still required for
new high-precision regimes
and polarization!!
!
[Sherwin++ in prep. 2016b]!
[Liu, Hill, Sherwin++ 2016]!
[van Engelen++ 2013, Osborne+
+2013]!
N-‐body
simulated
CMB
lensing
68. • CMB lensing map assumed linear & Gaussian, but it is not!!
!
!
• Structure formation non-linear: neglect gives %-level bias!
68!
big cluster!
N-body simulated lensing! Gaussian lensing!
[Boehm, Schmittfull, Sherwin 2016]!
Example Challenge II: !
The Impact of Non-linear Structure !
[Sherwin, Boehm, Liu, Hill in prep.]!
N-body simulated CMB lensing! Gaussian CMB lensing!
Cdd
` ⇠ h ˜T ˜T ˜T0 ˜T0
i ⇠ hTTT0
T0
i + hT,ijdidjTT0
,kdkT0
i
69. time (+ distance light travels!)!
Lensing of CMB lets us map
dark matter to study neutrino
mass +… !
CMB light!
Recap!
CMB Lensing dark matter map: a signal to study neutrinos,
+…
CMB!
70. time (+ distance light travels!)!
But if we are interested in
the early universe! Now lensing is a problem!!
CMB light!
CMB Lensing as Noise for Early Universe Cosmology
71. Outline!
• Introduction to CMB lensing: mapping cosmic dark matter!
• Lensing as a powerful signal for measuring neutrino masses
(+…)!
• Lensing as noise for studying inflation and the early universe!
71!
72. Big Questions
• What happened at the
beginning?!
!
• What expanded and
flattened the universe?!
• What set up the first
fluctuations in density
(that turned into
galaxies)?!
!
74. Inflation
CMB!
• Inflation: initial
accelerated cosmic
expansion. !
!
• Good evidence for
idea – but we don’t
know for sure!
!
• Many (simple)
models make
inflationary gravity
waves*!
!
!
*N.B. Some other models also
produce GWs!
75. Inflation
CMB!
!
• Probe physics at
ultra-high energy (at
the doorstep of the
Planck scale)!
!
!
• The strength of the
waves - tensor-to-
scalar ratio r - tells
us about the energy
at which inflation
happened! !
!
N.B. Even improved "!
upper limits interesting. !
!
77. CMB Polarization Basics!
• Any polarization map can be decomposed into E and B
mode fields!
• B-mode: contains signals from inflation, if there!
[Image credit: CMBPol]!
77!
B mode map!
E mode map!
~known
physics!
new
inflation
physics!
78. 10°
78!
B
E
B-mode
polarization: no
background
from better
known physics!
*ignoring
lensing
for
now
CMB B-polarization with no inflationary signal!
79. 10°
79!
B
E
*ignoring
lensing
for
now
CMB B-polarization with no inflationary signal!
B-mode
polarization: no
background
from better
known physics!
80. 10°
80!
B
E
See signal
clearly as
there is no
background!
!
B-modes are a
“null channel”!
*ignoring
lensing
and
dust
for
now
CMB B-polarization* with small inflationary signal!
82. Lensed CMB B-Polarization!
10°
82!
B
E
Lensing-B noise obscures
the inflation signal in the
early universe!
!
!
!
!
This has already begun to
limit constraints!!
B ⇠ E ⇤ d
Gravitational
lensing
converts E-
to B-
polarization!
83. Future: Simons Observatory / CMB-S4 Error Budget for
Measuring Inflationary Grav. Wave Signal!
error from instrument noise only (1uK-arcmin)!
1-sigma!
Error!
On !
Strength!
Of !
Inflation!
Signal!
[i.e. σ(r)]!
!
!
84. error from instrument noise only (1uK-arcmin)!
error including lensing noise!
Removing lensing
is essential for
future
experiments!!!!!!
%
1-sigma!
Error!
On !
Strength!
Of !
Inflation!
Signal!
[i.e. σ(r)]!
!
!
Future: Simons Observatory / CMB-S4 Error Budget for
Measuring Inflationary Grav. Wave Signal!
85. error from instrument noise only (1uK-arcmin)!
error including lensing noise!
Removing lensing
is essential for
future
experiments!!!!!!
%
1-sigma!
Error!
On !
Strength!
Of !
Inflation!
Signal!
[i.e. σ(r)]!
!
!
Future: Simons Observatory / CMB-S4 Error Budget for
Measuring Inflationary Grav. Wave Signal!
• Reduction in error is essential to ruling out broad class of
large-field inflation models (with super-Planckian field range) !
"– or confirming one of them!!
86. Delensing the CMB: Lensing Removal!
10°
86!
B
E
Gravitational
lensing
converts E-
to B-
polarization!
87. Delensing the CMB: Lensing Removal!
10°
87!
B
E
Know lensing, so can reverse it (=delens)
and remove lensing from the B-modes!
delensing!
Gravitational
lensing
converts E-
to B-
polarization!
88. 10°
88!
B
E
Delensing the CMB: Lensing Removal!
89. New Method: Cosmic Infrared Background Delensing!
error from instrument noise only (1uK-arcmin)!
error with full lensing noise!
%!
CIB delensing!
[Sherwin, Schmittfull 2015]!
1-sigma!
Error!
On !
Strength!
Of !
Inflation!
Signal!
[i.e. σ(r)]!
!
!
90. New Method: Cosmic Infrared Background Delensing!
error from instrument noise only (1uK-arcmin)!
error with full lensing noise!
%!
CIB delensing!
- use maps of cosmic infrared background (CIB)
[=galaxy emission] to infer and remove the lensing
(used by SPT/BICEP now, main method for LiteBIRD
satellite)!
[Sherwin, Schmittfull 2015]!
1-sigma!
Error!
On !
Strength!
Of !
Inflation!
Signal!
[i.e. σ(r)]!
!
!
91. New Method: Cosmic Infrared Background Delensing!
error from instrument noise only (1uK-arcmin)!
error with full lensing noise!
%!
CIB delensing!
- use maps of cosmic infrared background (CIB)
[=galaxy emission] to infer and remove the lensing
(used by SPT/BICEP now, main method for LiteBIRD
satellite)!
- make multitracer delensing map!
[Sherwin, Schmittfull 2015]!
1-sigma!
Error!
On !
Strength!
Of !
Inflation!
Signal!
[i.e. σ(r)]!
!
!
[Yu, Sherwin, Hill 2017]!
92. New Method: Cosmic Infrared Background Delensing!
error from instrument noise only (1uK-arcmin)!
error with full lensing noise!
%!
CIB delensing!
- use maps of cosmic infrared background (CIB)
[=galaxy emission] to infer and remove the lensing
(used by SPT/BICEP now, main method for LiteBIRD
satellite)!
- make multitracer delensing map!
[Sherwin, Schmittfull 2015]!
1-sigma!
Error!
On !
Strength!
Of !
Inflation!
Signal!
[i.e. σ(r)]!
!
!
[Yu, Sherwin, Hill 2017]!
93. Demonstrating Delensing with (Temp.) Data!
Planck CMB Temperature Data !
- lensed, so blurred (cartoon)!
Delensed!
Planck CMB
Temperature Data!
(sharper) !
“Lensing Map” from Cosmic
Infrared Background!
[Larsen, Challinor, Sherwin, Mak 2016]!
[Sherwin, Schmittfull 2015]!
Reverse Lensing!
Deflections!
94. [Larsen, Challinor, Sherwin, Mak 2016]!
[Sherwin, Schmittfull 2015]!
Checking Success of Delensing – !
Lensing in CMB Temperature Power Spectrum!
CMB Temperature Power Spectrum [uK2]!
1 / angular scale!
95. [Larsen, Challinor, Sherwin, Mak 2016]!
[Sherwin, Schmittfull 2015]!
Checking Success of Delensing – !
Lensing in CMB Temperature Power Spectrum!
Lensing blurs out
features / wiggles in
the power. Use to
show de-lensing!!
!
!
!
!
(effect x10 shown)!
CMB Temperature Power Spectrum [uK2]!
1 / angular scale!
96. CMB Temperature Power [uK2]!
1 / angular scale!
Planck CMB Temp. Data !
Delensed!
Planck CMB Temp. Data !
Reverse Lensing!
Deflections!
“Lensing Map” from
Cosmic Infrared
Background!
take difference of
power spectrum
after delensing !
–!
before delensing!
Demonstrating Delensing with (Temp.) Data!
97. Demonstrating Delensing: !
Difference of Delensed - Lensed Temp. Spectra!
Red line shows expected
un-blurring of the CMB (De-
lensing sharpens features)!
!
!
!
!
!
!
!
!
!
Flat / zero if lensing removal
not working!
98. Demonstrating Delensing: !
Difference of Lensed and Delensed Temp. Spectra!
[Larsen, Challinor, Sherwin, Mak 2016]!
[Sherwin, Schmittfull 2015]!
Planck data result!
16 sigma detection!
99. [Larsen, Challinor, Sherwin, Mak 2016]!
[Sherwin, Schmittfull 2015]!
Planck data result!
16 sigma detection!
First Demonstration of Delensing in Data – it works!!
!
100. !
• SO/CMB-S4: Can
discover / constrain
much smaller inflation
(r~0.001) signals by
delensing!
!
!
• Excellent delensing is
crucial for future CMB
– more research
needed, but method
should work!!
Great Potential: !
Simons Observatory / CMB-S4 Delensing!
error, instrument noise only!
no delensing!
CMB Stage-IV delensing (r~0.001)!
%
1-sigma!
Error!
On !
Strength!
Of !
Inflation!
Signal!
[i.e. σ(r)]!
!
!
101. Future B Mode Map – !
Lensing-Dominated!
10°
101!
B
102. Delensed B Map – Inflation Signal?!
10°
102!
B
a signal from the
beginning of the
universe?!
103. Summary!
• With new lensing and delensing analyses, CMB lensing has rapidly
progressed. With new experiments (AdvACT / Simons Observatory /
CMB-S4) and advances in theory, it will soon be even more powerful!!
!
• With upcoming CMB lensing, we will!
– determine neutrino masses (+study dark energy / astrophysics…)!
– enable powerful probes of inflation and the early universe!
– map all the mass and dark matter in the observable universe!
104. Summary!
Thanks for your attention!!
• With new lensing and delensing analyses, CMB lensing has rapidly
progressed. With new experiments (AdvACT / Simons Observatory /
CMB-S4) and advances in theory, it will soon be even more powerful!!
!
• With upcoming CMB lensing, we will!
– determine neutrino masses (+study dark energy / astrophysics…)!
– enable powerful probes of inflation and the early universe!
– map all the mass and dark matter in the observable universe!
!
Also happy to talk about my work on other areas in cosmology theory / data analysis:!
Baryon Acoustic Oscillations!
Large scale structure theory!
New approaches to galactic dust removal !
New methods in cluster cosmology +…!
!
114. 0.0 0.4 0.8 1.2 1.6
⌦m
0.50
0.75
1.00
1.25
8
ACT+BAO
KIDS
ACTPol – Early Lensing Power Spectrum Results !
114!
• We are consistent with Planck – mild tension with KiDS?!
[Sherwin et al. 2016]!matter density!
sizeoffluctuations!
115. Example: Biases from Non-Linear Large Scale
Structure!
115!
!
• Higher order terms in lensing estimator (theory 8pt. function)!
!
[Boehm, Schmittfull, Sherwin 2016; Sherwin, Boehm++ in prep.]!
lensing signal!
neglected higher order correction!
(will soon matter!)!
theory!
116. Example: Biases from Non-Linear Large Scale
Structure!
116!
!
• Higher order terms in lensing estimator (ray-traced N-body sim.)!
!
[Boehm, Schmittfull, Sherwin 2016; Sherwin, Boehm++ in prep.]!
lensing signal!
neglected higher order correction!
(will soon matter!)!
simulation!
117. CMB Lensing Systematics!
• Multiplicative and additive
biases from bi/trispectra of
foregrounds (+lensing itself) !
!
• Typical picture: when
nothing is done, ~few
percent effect.!
!
• Multifrequency data and
estimator “hardening”
should make these
unobservable – but work still
required for new regimes!
[Sherwin++
in
prep.
2016b]
[Liu,
Hill,
Sherwin++
2016]
[van
Engelen++
2013,
Osborne+
+2013]
N-‐body
simulated
CMB
lensing
119. • Lensing maps probe matter density, projected over a
wide redshift range peaking at z~2:!
!
• Information clean: mainly on linear or quasi-linear
scales!
!
Properties of Future / S4 CMB Lensing Maps!
[S4
Science
Book]
120. • SO/CMB S4 (wide survey – half sky?) could target
overlap with lots of upcoming and future surveys:!
• Wide range of applications: Astrophysics at high-z,
modified gravity, non-Gaussianity, r via delensing… !
What science can we target with !
CMB lensing + LSS?!
121. • Astrophysics: Cross-correlations give a new window on
high-z, probing relation of dark matter to tracers to get:!
!
!
• Cosmology: CMB lensing can be used to calibrate
systematic / galaxy biases in lensing and galaxy surveys!
Why cross-correlate CMB lensing with matter
tracers?!
- Quasar halo properties, evolution, activity (lensing x quasars)!
- Bias of galaxies and evolution (lensing x galaxies)!
- Star-formation history (e.g. lensing x Cosmic Infrared Background)!
- Baryon-dark matter relation (lensing x Sz or Ly-alpha)!
[e.g. Sherwin et al. 2012; Ferrarro et al 2015; POLARBEAR
2014b; van Engelen, Sherwin et al. 2015, Hill+Spergel 2014]! 121!
⇥|d(l)| q⇤
(l0
)⇤ = (2⇥)2
Cdq
l (l0
l)
Cross-‐power
spectrum
tracer
map
lensing
map
122. 122!
Lots of Recent Work on CMB Lensing Correlations!
122!
ACTPol x Planck CIB!
[v.Engelen, Sherwin++ 2015]!
First lensing
from
individual
halos
(CMASS)!
[Madhavach
eril++ incl.
Sherwin
2015]!
Planck x WISE!
[Ferraro, Sherwin 2015]!
POLARBEAR
pol. lensing x!
Herschel!
[Sherwin
corr. auth.,
15]!
123. Example Astrophysical Cross-correlation: !
Quasars x CMB lensing!
• Lensing depends on
the matter distribution.
QSOs trace the matter
distribution, and their
redshift distribution
matches the lensing
kernel.!
• This implies both are
correlated!
123!
A
quasar
density
map
CMB Lensing
Kernel!
QSO redshift
distribution!
124. Example Astrophysical Cross-correlation: !
Quasars x CMB lensing!
• This implies both are
correlated!
124!
A
quasar
density
map
Figure
from
Geach
et
al.
2014
quasar
density
CMB
lensing
125.
Cross-power with SDSS
photometric quasars:!
4 sigma detection!
!
Constraint on bias: !
b(z~1.4)=2.5 ± 0.6 (or halo
mass)!
!
DiPompeo et al. 2015: extend
this to obscured quasars,
compare different populations.
(Obscured quasars seem
more biased; may live in
higher-mass halos?)!
!
[Sherwin, Das, Hajian et al. 2012
125!
First Measurement:!
Quasar-CMB lensing Cross-correlation!
Cdq
amplitude determined
by bias!
Cdq
126. • Astrophysics: Cross-correlations give a new window on
high-z, probing relation of dark matter to tracers to get:!
!
!
• Cosmology: CMB lensing can be used to calibrate
systematic / galaxy biases in lensing and galaxy surveys!
Why cross-correlate CMB lensing with matter
tracers?!
[e.g. Sherwin et al. 2012; Ferrarro et al 2015; POLARBEAR
2014b; van Engelen, Sherwin et al. 2015, Hill+Spergel 2014]! 126!
⇥|d(l)| q⇤
(l0
)⇤ = (2⇥)2
Cdq
l (l0
l)
Cross-‐power
spectrum
tracer
map
lensing
map
127. !
Optical Weak Lensing x CMB Lensing!
127!
CFHT (CS82) !
galaxy lensing!
"X !
ACT CMB Lensing!
detected at 4.2 sigma!
[Hand, Das, Leautheaud,
Sherwin et al 2015]!
• Can cross-correlate CMB lensing with another kind of
lensing measurement: galaxy weak lensing (also a
direct mass probe). First detection with ACT.!
= r · d/2
l3
Cdd
l/4
128. Near Future: !
CMB Lensing x Weak Lensing as a Calibrator!
128!
• Example:
AdvancedACT x
LSST forecast,
100 sigma…!
!
• Use CMB lensing
as a calibrator for
systematic
biases!!
• Dark energy
figure of merit
much improved!
l3
Cdd
l/4
129. CMB Lensing As a Weak Lensing Bias Calibrator!
• Future combinations for systematics calibration are very
promising for dark energy and neutrino mass!
!
!
!
• Dark energy FoM: joint analysis, factor of 2 improvement!
Core
x
Euclid
Weak
Lensing
[S4
similar]
[Kitching,
Heavens,
Das
2014]
130. Near Future, AdvancedACT and CMB Stage IV: !
Great Potential for Cross-Correlation Science!
• Extremely powerful lensing maps
over much of the sky, with lots of
overlap with other surveys!!
• Useful for astrophysics and weak
lensing calibration!
(near) full sky
lensing mass map!
136. Light Particles and Cosmology!
• Cosmic Neutrino Background: in radiation era, very
large part of the energy density - 41% of total!!
• Influences expansion rate H (as extra form of radiation): !
!
!
136!
137. Light Particles and Cosmology!
• Cosmic Neutrino Background: in radiation era, very
large part of the energy density - 41% of total!!
• Influences expansion rate H (as extra form of radiation): !
!
!
!
• Energy density parameterized via number of effective
neutrino species Neff. Measure via CMB!
– Measure this with Planck:!
– So what? We have Z-decay measurements.!
137!
138. Neff from Cosmology:!
A Universal Probe for Light Relics!
• Not just sensitive to particles with the SM couplings of
neutrinos!
• Gravity sees everything: cosmology probes all that is
neutrino like (radiation, free-streaming)!
• Can hunt for any new light (relativistic, weakly coupled)
particles!!
!
!
138!
139. History of an Extra Light Particle!
• At high energies, weakly
interacting particle is
produced in thermal
equilibrium. It freezes out at
Tfreeze-out. At first, ΔNeff~1.!
!
• Universe cools. Phase
transition! (e.g. muons/anti-
muons annihilate).!
!
!
139!
140. History of an Extra Light Particle!
• At high energies, weakly
interacting particle is
produced in thermal
equilibrium. It freezes out at
Tfreeze-out. At first, ΔNeff~1.!
• Universe cools. Phase
transition! (e.g. muons/anti-
muons annihilate).!
• Annihilating particles dump
energy into photons not
particle. ΔNeff goes DOWN
for every phase transition
after freeze out.! 140!
Impact on energy gets diluted !
141. A Particle Freezes Out. !
How Much Neff Does it Contribute?!
• Freeze out late / low energy:
" "ΔNeff ~1!
!
141!
142. !
• Here, some muons (e,g.)
will annihilate after freeze
out: ΔNeff ~0.5!
!
142!
[Baumann++ 2015]!
A Particle Freezes Out. !
How Much Neff Does it Contribute?!
143. !
• At high freeze-out
temperature, the particle
misses out on lots of energy
from the QCD phase
transition, so ΔNeff is very
small.!
143!
A Particle Freezes Out. !
How Much Neff Does it Contribute?!
144. Neff: What is the Target?!
!
!
!
!
!
!
!
!
!
• If we can measure this:
target sensitivity, can see
any new light particle that
was ever in equilibrium (out
to reheating temp.!)!
144!
145. Probing Neff in the CMB Power Spectra!
• Change in early expansion rate > change in diffusion time >
small change in damping seen in CMB power spectrum. !
• Also: small shift in phases (free-streaming). Delensing helps! !
145!
Cl
TT
multipole
l
0
2000
[Hou++ 2013]!
146. Challenges / Requirements: !
High Precision Small-scale CMB Spectra!
!
• Requirement to hit
targets: Very low
noise (~uK’) and
large sky areas
(fsky~1)!
• Good beam
systematics control!
• Or lower noise: 2
sigma, room for
improvement!
146!
148. ACT Data: Measuring Skewness (<T3>) due to
tSZ!
tSZ Decrements!
Observed CMB Temperature!
IR Sources
removed!
First detection (5 sigma):!
Wilson, Sherwin, Hill et al (2012)!
149. First Cosmological and Cluster Physics
Constraints from the Data!
• Can show: skewness proportional to σ8
10.5 (c.f. power spec.
σ8
7)!
• Obtain constraint on σ8 from this measurement!
• In future, constrain cluster physics with skewness/power
spec. combination!
Hill & Sherwin (2012)!
152. Best Fit Model and Constraints!
• New theory and data analysis framework developed to
model, measure, and extract cosmology from full PDF!
• Next steps: add lensing information!
156. JHU CAS Seminar! 156!B. D. Sherwin - CMB Lensing with ACT and ACTPOL!
157. CMB
lensing
is
a
probe
of
the
projected
mass
distribution
1)
Reconstruct
lensing
from
changes
in
background
CMB
Problem:
need
high
resolution,
low
noise,
polarized
data
for
delensing
To Delens, Need To Measure Lensing - How?!
157!
2)
Estimate
lensing
from
Large
Scale
Structure
tracers
of
lensing,
e.g.
CIB
=
Z
dzW(z) (z)
158. Cross-correlations have found !
excellent LSS tracers of lensing!
• CMB
lensing
κ
traces
same
mass
distribution
δ
as
large-‐scale
structure
tracers
• Best
lensing
proxy
found:
cosmic
infrared
background
(CIB).
• Diffuse
emission
from
dusty
star-‐forming
galaxies
[Planck
2015;
van
Engelen,
Sherwin
et
al.
2014]
high-‐significance
lensing-‐
CIB
cross-‐correlation
CI
l
=
Z
dzW(z) (z)
159. The Cosmic Infrared Background: !
An Excellent Lensing Tracer!
• ~80%
correlated
with
lensing!
• Due
to
similar
high-‐
redshift
origin
as
CMB
lensing,
from
z~2
Redshi]
origin
of
CIB
Redshi]
origin
of
CMB
Lensing
Planck
lens
deflection
data
stacked
on
CIB
bright
spot
160. Simple idea: use (Planck) CIB maps !
as lensing tracers to delens!
Planck
CIB
map
at
857
GHz
Planck
CIB
map
at
545
GHz
+
Planck
maps
at
217
GHz,
353
GHz
161. Delensing The CMB!
measured
B
map
–
tensors
+
lensing
• How
to
reduce
lensing
noise?
• Delensing:
make
Blensing
map
from
E+κ
and
subtract:
B
–
Blensing
• How
do
we
use
a
CIB
map,
I,
to
make
a
lensing
estimate?
es`mate
of
lensing
B
(from
k+E)
_
subtract
lensing
es`mate
from
CIB,
I
E
mode
map
deflect
ˆBlens
(r) ⇠ hCBB,lens
l + NBB,lens
l il<100
B
162. How to Delens Using the CIB - I!
• To
use
CIB
as
proxy
for
lensing:
normalize
with
function
f,
then
construct
lensing
B
estimate
• Choose
weight
f
such
that
residual
delensed
B-‐mode
is
minimized:
f
depends
only
on
observables
min[CB ˆBlens
l ] ! f =
CI
l
CII
l
ˆBlens
CIB =
Z
d2
l0
(2⇡)2
E(l0
)f(l0
, l)I(l l0
)
CIB
map
CIB-‐lensing
cross-‐power
CIB
power
163. !
• Scaled CIB : best current lensing map over most
scales. Example with Planck 545 GHz:!
• Use this map to estimate lensing B-mode, subtract from
data:!
!
How to Delens Using the CIB - II!
fl ⇥ Il
B ˆBlens
CIB
164. CIB-‐lensing
correla`ons
which
scales
ma*er
most?
Delensing The CMB with the CIB: !
Forecasting Performance!
• Error
reduction
depends
on
how
much
B-‐mode
power
is
removed
• Find
that
CIB
delensing
reduces
B-‐mode
power
by
a
factor
Where
[Sherwin,
Schmibull
2015]
⇢ =
CI
l
q
C
l CII
l
(1 ⇢2
)
165. Delensing with the CIB: Single Frequencies!
165!
• Good delensing performance!
original
B-‐mode
power
delensed
B-‐mode
power
166. • Error
is
reduced
as
lensing
B-‐mode
power
is
removed
• Define
delensing
improvement
factor
(how
much
is
any
detection’s
signal-‐to-‐noise
boosted?)
Quantifying Delensing Performance: How Much are
Constraints on r Improved by Delensing?!
↵ ⌘
0(r)
delensed(r)
=
hCBB,lens
l + NBB
l [ P ]il<100
hCBB,res
l [ P ] + NBB
l [ P ]il<100
Noise
Level
Noise
Power
Delensed
Residual
B-‐Power
(r) ⇠ hCBB,lens
l + NBB,lens
l il<100
CBB,res
l =
Z
d2
l0
(2⇡)2
W2
CEE
l C
l l0 (1 ⇢2
l l0 )
[Smith
et
al.
2012]
167. Expected CIB Delensing Improvement Factor!
167!
• Plot: factor by
which the signal-to-
noise is increased
by CIB delensing!
• Up to 2.2
improvement with
public, available
data!
• Corresponds to
2.22 larger survey.!
!
instrumental
noise
level
[Sherwin,
Schmibull
2015]
168. Expected CIB Delensing Improvement Factor!
168!
• Plot: factor by
which the signal-to-
noise is increased
by CIB delensing!
• Up to 2.2
improvement with
public, available
data!
• Corresponds to
2.22 larger survey.!
! Keck
Array
Now:
Equivalent
to
having
800
instead
of
400
square
degrees
169. 169!
• Example: SPT/
POLARBEAR +
CIB, delensing
Keck Array!
• Derive optimal
combination: large
improvements
when including
CIB.!
• Powerful
combination for
delensing with
current data!
add
CIB
Combined
Tracers
for
Delensing:
CIB
+
External
Lensing
Reconstruction
[Sherwin,
Schmibull
2015]
170. A Potential Problem: Uncertain CIB Spectra!
170!
• Say we have a measurement of CIB delensed B-mode
power. Now constrain r, using model for delensed B-
mode power:!
!
!
!
• Problem: residual lensing B-mode power depends on the
CIB spectra! If CIB spectra are poorly constrained, this
can be confused with r.!
• With Planck measurements of , do we know the CIB
spectra well enough?!
CI
l
171. Robust CIB Delensing, with !
CIB Bandpowers as Nuisance Parameters!
171!
!
• Don’t assume CIB spectra
known: parametrize CIB
generally, constrain with
Planck+BB!
• Constrain r marginalized
over CIB spectra, i.e., their
parameters ai!
• Fisher matrix formalism:
marginalizing over CIB
spectra does not degrade
delensing performance!!
!
CIB
bandpowers
ai
CI
l
l
a1
a2
a3
a4