dick bond
DESCRIPTION
Dick Bond. Inflation & its Cosmic Probes, now & then. Cosmic Probes CMB, CMBpol (E,B modes of polarization) B from tensor: Bicep, Planck, Spider, Spud, Ebex, Quiet, Pappa, Clover, …, Bpol CFHTLS SN(192) , WL(Apr07) , JDEM/DUNE BAO, LSS,Ly a. - PowerPoint PPT PresentationTRANSCRIPT
Dick Bond
Inflation Then k=(1+q)(a) ~r/16 0<= multi-parameter expansion in (lnHa ~ lnk)
Dynamics ~ Resolution ~ 10 good e-folds (~10-4Mpc-1 to ~ 1 Mpc-1 LSS) ~10+ parameters? Bond, Contaldi, Kofman, Vaudrevange 07
r(kp) i.e. k is prior dependent now, not then. Large (uniform ), Small
(uniform ln). Tiny (roulette inflation of moduli; almost all string-inspired models)
KKLMMT etc, Quevedo etal, Bond, Kofman, Prokushkin, Vaudrevange 07, Kallosh and Linde 07
General argument: if the inflaton < the Planck mass, then r < .007 (Lyth96 bound)
Cosmic Probes CMB, CMBpol (E,B modes of polarization)
B from tensor: Bicep, Planck, Spider, Spud, Ebex, Quiet, Pappa, Clover, …, Bpol
CFHTLS SN(192),WL(Apr07), JDEM/DUNE BAO,LSS,Ly
Inflation & its Cosmic Probes, now & then
Inflation & its Cosmic Probes, now & then
Dick Bond
Inflation Now1+w(a)= sf(a/aeq;as/aeq;s) goes to ax3/2 = 3(1+q)/2 ~1
good e-fold. only ~2params
Cosmic Probes NowCFHTLS SN(192),WL(Apr07),CMB,BAO,LSS,Ly
Zhiqi Huang, Bond & Kofman 07 s=0.0+-0.25 now, late-inflaton (potential gradient)2
to +-0.07 then Planck1+JDEM SN+DUNE WL, weak as < 0.3 now <0.21 then
(late-inflaton mass is < Planck mass, but not by a lot)
Cosmic Probes Then JDEM-SN + DUNE-WL + Planck1
NOW
Inflation & its Cosmic Probes, now & then
–Dick Bond
Inflation now
Dynamical background late-inflaton-field trajectories imprint luminosity distance, angular diameter distance, volume growth, growth rate of density fluctuations
Prior late-inflaton primordial fluctuation information is largely lost because tiny mass
(field sound speed=c?)
late-inflaton may have an imprint on other fields?
New late-inflaton fluctuating field power is tiny
w-trajectories for V(): pNGB example e.g.sorbo et07
For a given quintessence potential V(), we set the “initial conditions” at z=0 and evolve backward in time.
w-trajectories for Ωm (z=0) = 0.27 and (V’/V)2/(16πG) (z=0) = 0.25, the 1-sigma limit, varying the initial
kinetic energy w0 = w(z=0)
Dashed lines are our first 2-param approximation using an a-averaged
s= (V’/V)2/(16πG) and 2 -fitted as.Wild rise solutions
Slow-to-medium-roll solutions
Complicated scenarios: roll-up then roll-down
Approximating Quintessence for Phenomenology
+ Friedmann Equations + DM+B
1+w=2sin2
Zhiqi Huang, Bond & Kofman 07
1+w=-2sinh2
Include a w<-1 phantom field, via a negative kinetic energy term
slow-to-moderate roll conditions
1+w< 0.2 (for 0<z<10) and gives a 1-parameter model (as<<1):
Early-Exit Scenario: scaling regime info is lost by Hubble damping, i.e.small as
1+w< 0.3 (for 0<z<10) gives a 2-parameter model (as and s):
CMB+SN+LSS+WL+Lya
w(a)=w0+wa(1-a)
effective constraint eq.
Some ModelsCosmological Constant (w=-1)
Quintessence
(-1≤w≤1)
Phantom field (w≤-1)
Tachyon fields (-1 ≤ w ≤ 0)
K-essence
(no prior on w)
Uses latest April’07
SNe, BAO, WL, LSS, CMB, Lya data
w(a)=w0+wa(1-a) models
45 low-z SN + ESSENCE SN + SNLS 1st year SN+ Riess high-z SN, 192 “gold”SN all fit with MLCS
illustrates the near-degeneracies of the contour plotillustrates the near-degeneracies of the contour plot
piecewise parameterization 4,9,40
z-modes of w(z)
1=0.12 2=0.32 3=0.63
Higher Chebyshev expansion is not useful:
data cannot determine >2 EOS
parameters 9 & 40 into Parameter
eigenmodes DETF Albrecht etal06, Crittenden etal06,
hbk07
Data used 07.04:
CMB+SN+WL
+LSS+Lya
49
40
Measuring constant w (SNe+CMB+WL+LSS)1+w = 0.02 +/- 0.05
Measuring s
(SNe+CMB+WL+LSS+Lya)Modified CosmoMC with Weak Lensing and time-varying w models
3-parameter parameterizationnext order corrections:
m (a) (depends on s redefines aeq)
vs (a) (adds new s parameter)
enforce asymptotic kinetic-dominance w=1 (add as power suppression)
perfect moderate-to-slow roll fits ... & even the baroque …
correction on w only ~ 0.01
fits the baroque wild-rising trajectories
Measuring the 3 parameters with current data• Use 3-parameter formula over 0<z<4 &
w(z>4)=wh (irrelevant parameter unless large).
as <0.3
Comparing 1-2-3-parameter results
Conclusion: for current data, the multi-parameter complications are largely irrelevant (as <0.3):
we cannot reconstruct the quintessence potential
we can only measure the slope s
CMB + SN + WL + LSS +Lyastrajectories are slowly
varying: why the fits are good
Thawing, freezing or non-monotonic?
• Thawing: 1+w monotonic up as z decreases
• Freezing: 1+w monotonic down to 0 as z decreases• ~15% thaw, 8% freeze, most non-monotonic with flat priors
With freezing prior:
With thawing prior:
the quintessence field is below the reduced Planck mass
Forecast: JDEM-SN (2500 hi-z + 500 low-z)
+ DUNE-WL (50% sky, gals @z = 0.1-1.1, 35/min2 ) +
Planck1yr
s=0.02+0.07-0.06
as<0.21 (95%CL)
Beyond Einstein panel: LISA+JDEM
ESA
• the data cannot determine more than 2 w-parameters (+ csound?). general higher order Chebyshev expansion in 1+w as for “inflation-then” =(1+q) is not that useful. Parameter eigenmodes show what is probed
• The w(a)=w0+wa(1-a) phenomenology requires baroque potentials• Philosophy of HBK07: backtrack from now (z=0) all w-trajectories arising from
quintessence (s >0) and the phantom equivalent (s <0); use a 3-parameter model to well-approximate even rather baroque w-trajectories.
• We ignore constraints on Q-density from photon-decoupling and BBN because further trajectory extrapolation is needed. Can include via a prior on Q (a) at z_dec and z_bbn
• For general slow-to-moderate rolling one needs 2 “dynamical parameters” (as, s) & Q to describe w to a few % for the not-too-baroque w-trajectories.
• as is < 0.3 current data (zs >2.3) to <0.21 (zs >3.7) in Planck1yr-CMB+JDEM-SN+DUNE-WL future
In the early-exit scenario, the information stored in as is erased by Hubble friction over the observable range & w can be described by a single parameter s.
• a 3rd param s, (~ds /dlna) is ill-determined now & in a Planck1yr-CMB+JDEM-SN+DUNE-WL future
• To use: given V, compute trajectories, do a-averaged s & test (or simpler s -estimate)• for each given Q-potential, velocity, amp, shape parameters are needed to define a w-trajectory
• current observations are well-centered around the cosmological constant s=0.0+-0.25 • in Planck1yr-CMB+JDEM-SN+DUNE-WL future s to +-0.07• but cannot reconstruct the quintessence potential, just the slope s & hubble drag info• late-inflaton mass is < Planck mass, but not by a lot
• Aside: detailed results depend upon the SN data set used. Best available used here (192 SN), soon CFHT SNLS ~300 SN + ~100 non-CFHTLS. will put all on the same analysis/calibration footing – very important.
• Newest CFHTLS Lensing data is important to narrow the range over just CMB and SN
Inflation now summary
THEN THEN
Inflation & its Cosmic Probes, now & then
Dick Bond
Inflation then
Amplitude As, average slope <ns>, slope fluctuations ns = ns -< ns > (running,
running of running, …)
for scalar (low L to high L CMB ACT/SPT; Epol Quad, SPTpol, Quiet2, Planck)
& tensor At, <nt>, nt = nt-< nt >
(low L <100 Planck, Bicep, EBEX, Spider, SPUD, Clover, Bpol)
& isocurvature Ais, <nis>, nis= nis-< nis> power spectra (subdominant)
Blind search for structure (not really blind because of prior probabilities/measures)
Fluctuation field power spectra related to dynamical background field trajectories
Defines a tensor/scalar functional relation between; both to Hubble & inflaton potential
Standard Parameters of Cosmic Structure Formation
Òk
What is the Background curvature of the universe?
Òk > 0Òk = 0Òk < 0
closed
flatopen
Òbh2 ÒË nsÒdmh2
Density of Baryonic Matter
Density of non-interacting Dark
Matter
Cosmological Constant
Spectral index of primordial scalar (compressional)
perturbations
PÐ(k) / knsà1
nt
Spectral index of primordial tensor (Gravity Waves)
perturbations
Ph(k) / knt
lnAs ø lnû8
Scalar Amplitude
r = A t=As
Tensor Amplitude
Period of inflationary expansion, quantum noise metric perturbations
üc
Optical Depth to Last Scattering
SurfaceWhen did stars
reionize the universe?
òø `à1s ; cf :ÒË r < 0.6 or < 0.28 95% CL
ns = .958 +- .015
.93 +- .03 @0.05/Mpc run&tensor
r=At / As < 0.28 95% CL
<.36 CMB+LSS run&tensor
dns /dln k = -.060 +- .022
-.038 +- .024 CMB+LSS run&tensor
As = 22 +- 2 x 10-10
The Parameters of Cosmic Structure FormationThe Parameters of Cosmic Structure FormationCosmic Numerology: aph/0611198 – our Acbar paper on the basic 7+; bckv07
WMAP3modified+B03+CBIcombined+Acbar06+LSS (SDSS+2dF) + DASI (incl polarization and CMB weak lensing and tSZ)
bh2 = .0226 +- .0006
ch2 = .114 +- .005
= .73 +.02 - .03
h = .707 +- .021
m= .27 + .03 -.02
zreh = 11.4 +- 2.5
CMBology
ForegroundsCBI, Planck
ForegroundsCBI, Planck
SecondaryAnisotropies (CBI,ACT)
(tSZ, kSZ, reion)
SecondaryAnisotropies (CBI,ACT)
(tSZ, kSZ, reion)
Non-Gaussianity(Boom, CBI, WMAP)
Non-Gaussianity(Boom, CBI, WMAP)
Polarization ofthe CMB, Gravity Waves
(CBI, Boom, Planck, Spider)
Polarization ofthe CMB, Gravity Waves
(CBI, Boom, Planck, Spider)
Dark Energy Histories(& CFHTLS-SN+WL)
Dark Energy Histories(& CFHTLS-SN+WL)
subdominant phenomena
(isocurvature, BSI)
subdominant phenomena
(isocurvature, BSI)
Inflation Histories(CMBall+LSS)
Inflation Histories(CMBall+LSS)
Probing the linear & nonlinear cosmic web
Probing the linear & nonlinear cosmic web
wide open braking
approach to preheating
Kahler modulus potential T=+i
2004
2005
2006
2007
2008
2009
Polarbear(300 bolometers)@Cal
SZA(Interferometer) @Cal
APEX(~400 bolometers) @Chile
SPT(1000 bolometers) @South Pole
ACT(3000 bolometers) @Chile
Planck08.8
(84 bolometers)
HEMTs @L2
Bpol@L2
ALMA(Interferometer) @Chile
(12000 bolometers)SCUBA2
Quiet1
Quiet2Bicep @SP
QUaD @SP
CBI pol to Apr’05 @Chile
Acbar to Jan’06, 07f @SP
WMAP @L2 to 2009-2013?
2017
(1000 HEMTs) @Chile
Spider
Clover @Chile
Boom03@LDB
DASI @SP
CAPMAP
AMI
GBT
2312 bolometer @LDB
JCMT @Hawaii
CBI2 to early’08
EBEX@LDB
LMT@Mexico
LHC
ACT@5170m
CBI2@5040m
why Atacama? driest desert in the world. thus: cbi, toco, apex,
asti, act, alma, quiet, clover
WMAP3 sees 3rd pk, B03 sees 4th
Current high L state
November 07
Current high L state
November 07
CBI sees 4th 5th pk
CBI excess 07
WMAP3 sees 3rd pk, B03 sees 4th
PRIMARY END @ 2012? PRIMARY END @ 2012?
CMB ~2009+ Planck1+WMAP8+SPT/ACT/Quiet+Bicep/QuAD/Quiet +Spider+Clover
http://www.astro.caltech.edu/~lgg/spider_front.htmhttp://www.astro.caltech.edu/~lgg/spider_front.htm
No Tensor
SPIDER Tensor Signal
Tensor
• Simulation of large scale polarization signal
forecast Planck2.5
100&143
Spider10d
95&150
Synchrotron pol’n
Dust pol’n
are higher in B
Foreground Template removals
from multi-frequency data
is crucial
Latham Boyle 07Latham Boyle 07
GW/scalar curvature: current from CMB+LSS: r < 0.6 or < 0.25 (.28) 95%; good shot at 0.02 95% CL with BB polarization (+- .02 PL2.5+Spider), .01 target
BUT foregrounds/systematics?? But r-spectrum. But low energy inflation
Very very difficult to get at with direct gravity wave detectors – even in our dreams (Big Bang Observer ~ 2030)
Inflation in the context of ever changing fundamental theory
1980
2000
1990
-inflation Old Inflation
New Inflation Chaotic inflation
Double InflationExtended inflation
DBI inflation
Super-natural Inflation
Hybrid inflation
SUGRA inflation
SUSY F-term inflation SUSY D-term
inflation
SUSY P-term inflation
Brane inflation
K-flationN-flation
Warped Brane inflation
inflation
Power-law inflation
Tachyon inflationRacetrack inflation
Assisted inflation
Roulette inflation Kahler moduli/axion
Natural inflation
Power law (chaotic) potentials
V/MP4 ~ 2 MP
-1/2
NI(k) +/3
ns-1+1NI(k) -/6
ntNI(k) -/6
=1, NI
rnsnt
=2, NI
rnsnt
Power law (chaotic) potentials
V/MP4 ~ 2 MP
-1/2
NI(k) +/3
ns-1+1NI(k) -/6
ntNI(k) -/6
=1, NI
rnsnt
=2, NI
rnsnt
MP-2= 8G
PNGB:V/MP4 ~red
4sin2fred -1/2
nsfred-2
1-nsexp[1-nsNI (k)] (1+1-ns -1
exponentially suppressed; higher r if lowerNI & 1-ns
to match ns.96, fred~ 5, r~0.032
to match ns.97, fred~ 5.8, r~0.048
cf. =1, rnsnt
PNGB:V/MP4 ~red
4sin2fred -1/2
nsfred-2
1-nsexp[1-nsNI (k)] (1+1-ns -1
exponentially suppressed; higher r if lowerNI & 1-ns
to match ns.96, fred~ 5, r~0.032
to match ns.97, fred~ 5.8, r~0.048
cf. =1, rnsnt
ABFFO93ABFFO93
Moduli/brane distance limitation in stringy inflation. Normalized canonical inflaton
over e.g. 2/nbrane1/2
BM06
= (dd ln a)2 so r < .007, <<?
ns.97, fred~ 5.8, r~0.048,
cf. =1, rns
cf. =2, rns
ns.97, fred~ 5.8, r~0.048,
cf. =1, rns
cf. =2, rns
roulette inflation examples r ~ 10-10 roulette inflation examples r ~ 10-10 possible way out with many fields assisting: N-flationpossible way out with many fields assisting: N-flation
energy scale of inflation & r
V/MP4 ~ Ps r (1-3) 3/2
V~ (1016 Gev)4 r/0.1 (1-3)
energy scale of inflation & r
V/MP4 ~ Ps r (1-3) 3/2
V~ (1016 Gev)4 r/0.1 (1-3)
roulette inflation examples V~ (few x1013
Gev)4
roulette inflation examples V~ (few x1013
Gev)4
H/MP ~ 10-5 (r/.1)1/2 H/MP ~ 10-5 (r/.1)1/2 inflation energy scale cf. the gravitino mass (Kallosh &
Linde 07) if a KKLT/largeVCY-like generation mechanism
1013 Gev (r/.01)1/2 ~ H < m3/2 cf. ~Tev
String Theory Landscape & Inflation++ Phenomenology for CMB+LSS
Hybrid D3/D7 Potential
KKLT, KKLMMT
f||
fperp
•D3/anti-D3 branes in a warped geometry
•D3/D7 branes
•axion/moduli fields ... shrinking holesBB04, CQ05, S05, BKPV06
large volume 6D cct Calabi Yau
endend
Roulette Inflation: Ensemble of Kahler Moduli/Axion InflationsBond, Kofman, Prokushkin & Vaudrevange 06
Roulette: which minimum
for the rolling ball depends upon the throw; but which roulette wheel we
play is chance too.
The ‘house’ does not just play dice with
the world. Very small r’s may arise from string-inspired low-energy inflation models
“quantum eternal
inflation” regime
stochastic kick >
classical drift
Ps (ln Ha) Kahler trajectoriesPs (ln Ha) Kahler trajectories
observable rangeobservable range
But it is much easier
to get models
which do not agree with
observationsi.e. data
selectsthe few that
‘work’ from the vast
cases that do not
But it is much easier
to get models
which do not agree with
observationsi.e. data
selectsthe few that
‘work’ from the vast
cases that do not
(ln a) (ln a)H (ln a)H (ln a)
10-1010-10
New Parameters of Cosmic Structure FormationÒk
Òbh2
lnP s(k)Òdmh2
scalar spectrumuse order N Chebyshev
expansion in ln k, N-1 parameters
amplitude(1), tilt(2), running(3), …
(or N-1 nodal point k-localized values)
òø `à1s ; cf :ÒË
tensor (GW) spectrumuse order M Chebyshev
expansion in ln k, M-1 parameters amplitude(1), tilt(2),
running(3),...Dual Chebyshev expansion in ln k:
Standard 6 is Cheb=2
Standard 7 is Cheb=2, Cheb=1
Run is Cheb=3
Run & tensor is Cheb=3, Cheb=1
Low order N,M power law but high order Chebyshev is Fourier-like
üc
lnP t(k)
lnPs Pt (nodal 2 and 1) + 4 params cf Ps Pt (nodal 5 and 5) + 4 params
reconstructed from CMB+LSS data using Chebyshev nodal point expansion & MCMC
lnPs Pt (nodal 2 and 1) + 4 params cf Ps Pt (nodal 5 and 5) + 4 params
reconstructed from CMB+LSS data using Chebyshev nodal point expansion & MCMC
no self consistency: order 5 in scalar and tensor power
r = .21+- .17 (<.53)
Power law scalar and constant tensor + 4 params
effective r-prior makes the limit stringent
r = .082+- .08 (<.22)
run+tensor r = .13+- .10 (<.33)
lnPs lnPt (nodal 5 and 5) + 4 params. Uniform in exp(nodal bandpowers) cf. uniform
in nodal bandpowers reconstructed from April07 CMB+LSS data using Chebyshev nodal point expansion & MCMC: shows prior dependence with current data
lnPs lnPt (nodal 5 and 5) + 4 params. Uniform in exp(nodal bandpowers) cf. uniform
in nodal bandpowers reconstructed from April07 CMB+LSS data using Chebyshev nodal point expansion & MCMC: shows prior dependence with current data
logarithmic prior
r < 0.075
uniform prior
r < 0.42
New Parameters of Cosmic Structure FormationÒk
Òbh2lnH(kp)
ï (k); k ù HaÒdmh2
=1+q, the deceleration parameter history
order N Chebyshev expansion, N-1 parameters
(e.g. nodal point values)
P s(k) / H 2=ï ;P t(k) / H 2
òø `à1s ; cf :ÒË
Hubble parameter at inflation at a pivot pt
Fluctuations are from stochastic kicks ~ H/2 superposed on the downward drift at lnk=1.
Potential trajectory from HJ (SB 90,91):
üc
à ï = d lnH =d lna
1à ïà ï = d lnk
d lnH
d lnkd inf = 1à ï
æ ïp
V / H 2(1à 3ï );
ï = (d lnH =d inf)2
H(kp)
lns (nodal 5) + 4 params. Uniform in exp(nodal bandpowers) cf. uniform in
nodal bandpowers reconstructed from April07 CMB+LSS data using Chebyshev nodal point expansion & MCMC: shows prior dependence with current data
lns (nodal 5) + 4 params. Uniform in exp(nodal bandpowers) cf. uniform in
nodal bandpowers reconstructed from April07 CMB+LSS data using Chebyshev nodal point expansion & MCMC: shows prior dependence with current data
log prior
r (<0.34; .<03 at 1-sigma!)
s self consistency: order 5 uniform prior
r = (<0.64)
1à ïà ï = d lnk
d lnH
V / H 2(1à 3ï ); d lnk
d inf = 1à ïæ ï
pP s(k) / H 2=ï ;P t(k) / H 2
ï = (d lnH =d inf)2
lns (nodal 5) + 4 params. Uniform in exp(nodal bandpowers) cf. uniform in nodal
bandpowers reconstructed from April07 CMB+LSS data using Chebyshev nodal point expansion & MCMC: shows prior dependence with current data
lns (nodal 5) + 4 params. Uniform in exp(nodal bandpowers) cf. uniform in nodal
bandpowers reconstructed from April07 CMB+LSS data using Chebyshev nodal point expansion & MCMC: shows prior dependence with current data
logarithmic prior
r < 0.33, but .03 1-sigma
uniform prior
r < 0.64
CL BB for lns (nodal 5) + 4 params inflation trajectories reconstructed from
CMB+LSS data using Chebyshev nodal point expansion & MCMC
CL BB for lns (nodal 5) + 4 params inflation trajectories reconstructed from
CMB+LSS data using Chebyshev nodal point expansion & MCMC
Planck satellite 2008.6 Spider
balloon 2009.9
Spider balloon 2009.9
uniform prior
log prior
Spider+Planck broad-band
error
CL BB for lnPs lnPt (nodal 5 and 5) + 4 params inflation trajectories reconstructed
from CMB+LSS data using Chebyshev nodal point expansion & MCMC
CL BB for lnPs lnPt (nodal 5 and 5) + 4 params inflation trajectories reconstructed
from CMB+LSS data using Chebyshev nodal point expansion & MCMC
Planck satellite 2008.6 Spider
balloon 2009.9
Spider balloon 2009.9
Spider+Planck broad-band
error
uniform prior
log prior
B-pol simulation: input LCDM (Acbar)+run+uniform tensor
r (.002 /Mpc) reconstructed cf. rin
s order 5 uniform prior s order 5 log prior
a very stringent test of the -trajectory methods: A+
Planck1yr simulation: input LCDM (Acbar)+run+uniform tensor
r (.002 /Mpc) reconstructed cf. rin
s order 5 uniform prior s order 5 log prior
Planck1 simulation: input LCDM (Acbar)+run+uniform tensor
Ps Pt reconstructed cf. input of LCDM with scalar running & r=0.1
s order 5 uniform prior
s order 5 log prior
lnPs lnPt (nodal 5 and 5)
Inflation then summarythe basic 6 parameter model with no GW allowed fits all of the data OK
Usual GW limits come from adding r with a fixed GW spectrum and no consistency criterion (7 params). Adding minimal consistency does not make that
much difference (7 params)
r (<.28 95%) limit comes from relating high k region of 8 to low k region of GW CL
Uniform priors in (k) ~ r(k): with current data, the scalar power downturns ((k) goes up) at low k if there is freedom in the mode expansion to do this. Adds GW
to compensate, breaks old r limit. T/S (k) can cross unity. But log prior in drives to low r. a B-pol could break this prior dependence, maybe Planck+Spider.
Complexity of trajectories arises in many-moduli string models. Roulette example: 4-cycle complex Kahler moduli in large compact volume Type IIB string theory
TINY r ~ 10-10 if the normalized inflaton < 1 over ~50 e-folds then r < .007
~10 for power law & PNGB inflaton potentials
Prior probabilities on the inflation trajectories are crucial and cannot be decided at this time. Philosophy: be as wide open and least prejudiced as possible
Even with low energy inflation, the prospects are good with Spider and even Planck to either detect the GW-induced B-mode of polarization or set a powerful
upper limit against nearly uniform acceleration. Both have strong Cdn roles. CMBpol
End End