Introduction. The view from Earth:

TOMMASO TREU

(University of California Los Angeles)

What is the universe made of?

Outline

• 

Introduction. The view from Earth:

•  The standard model of particle physics

• 

The view from the Universe

•  Gravitational time delays and Dark energy

•  Strong lensing and dark matter

• 

A roadmap for the future

The view from Earth: standard

model of particle physics

The view from the universe

The Dark Universe

What is the universe made of?

(2013-2015)

Is this model correct? And, if so,

what is causing acceleration?

The current explanation is:

P=wρ

Cosmological constant? w=-1 Something else? w≠-1

Inflationary Big Bang predicts Universe is

“flat” (Euclidean geometry)

Cosmography with

gravitational lensing

What is Gravitational Lensing?

Matter curves space…

…and in rare circumstances create multiple images

Image separation is a direct measurement of mass, luminous or dark!

Cosmography from time delays:

how does it work?

Strong lensing in terms of Fermat’s principle

Excess time delay

Time delay distance

geometric time delay

Shapiro delay

Observables: flux, position, and arrival time of the multiple images

Time delay distance in practice

Steps:

•  Measure the time-delay between two images

•  Measure and model the potential

•  Infer the time-delay distance

•  Convert it into cosmlogical parameters

t D

(z

, z

) H

f (

, w, ...)

Low redshift measurements (like TD) are essential

Planck XVI

The power of time-delays (and other low-z probes)

Suyu, Treu et al. 2014

Cosmography from time delays:

A brief history

Ü  1964 Method proposed

Ü  70s First lenses discovered

Ü  80s First time delay measured

Ü  Controversy. Solution: improve sampling Ü  90s First Hubble Constant measured

Ü  Controversy. Solution: improve mass models Ü  2000s: modern monitoring (COSMOGRAIL,

Fassnacht & others); stellar kinematics (Treu &

Koopmans 2002); extended sources

Ü  2010s Putting it all together: precision measurements (6-7% from a single lens)

Ü  2014 first multiply imaged supernova discovered (50^th anniversary of Refsdal’s paper)

November 2014

Supernova ‘Refsdal’

1.1 1.2

1.3

MACS J1149.6+2223

10 arcsec

Lensing!

cluster!

member!

z=0.544

Host!

Galaxy!

z=1.49

CLASH/GLASS!

< Feb 2014

GLASS/Frontier Fields!

Nov 2014

Difference 1.4

Kelly, Rodney, Treu et al. 2014

Cosmography with strong lenses:

the 4 problems solved

Ü  Time delay – 2-3 %

Ü  Tenacious monitoring (e.g. Fassnacht et al.

2002); COSMOGRAIL (Meylan/Courbin) Ü  Astrometry – 10-20 mas

Ü  Hubble/VLA/(Adaptive Optics?) Ü  Lens potential (2-3%)

Ü  Stellar kinematics/Extended sources (Treu &

Koopmans 2002; Suyu et al. 2009)

Ü  Structure along the line of sight (2-3%)

Ü  Galaxy counts and numerical simulations (Suyu et al. 2010)

Ü  Stellar kinematics (Koopmans et al. 2003)

Cosmography with strong lenses:

measuring time delays

Vanderriest et al. 1989

COSMOGRAIL: better data & better techniques

C A

B

D

Cosmography with strong lenses:

measuring the lens potential

Host galaxy reconstruction; Suyu et al. 2012

Schechter et al. 1997

Cosmography with strong lenses:

measuring the lens potential

Stellar kinematics: Treu & Koopmans 2002

Kochanek & Schechter 2003

Cosmography with strong lenses:

Structure along the line of sight

Suyu et al. 2010

???

Pilot: B1608+656

B1608:

Constraints on Dark Energy

For curved wCDM

With WMAP7:

•  B1608+656 is comparable to

BAO [Percival et al. 2010]

•  B1608+656 and BAO both primarily constrain Ω_k

•  SN [Hicken et al. 2009]

primarily constrains w

Suyu et al 2010

Blind Analysis: 1131-1231

Blind Analysis

•  Prevents unconscious experimenter bias

•  allows us to test for the presence of residual

systematics, if any

•  PDF centroids of cosmological

parameters are hidden Blinded time-delay distance

Time delays of RXJ1131-1231

Time delay with 1.5%

accuracy!

[Tewes et al.

2013b]

Based on

state-of-the-art curve modeling techniques

[Tewes et al. 2013a]

Lens Model

Observed Image

Lens Model

mass distribution of lens

light distribution of extended source

Lens Model

mass distribution of lens

light distribution of extended source

light of lensed AGN

Lens Model

mass distribution of lens

light distribution of extended source

light of lens (Sersic) light of

lensed AGN

Lens Model

mass distribution of lens

light distribution of extended source

light of lens (Sersic) light of

lensed AGN +

time delays

Lens Model

Line-of-sight Effects

Keck LRIS

Velocity dispersion:

323 ± 20 km/s

Lens environment + Millennium Simulation

[Suyu et al. 2013]

In combination with WMAP7 in flat wCDM cosmology

Precision comparable to that of B1608+656 Accuracy?

After completing the blind analysis and agreeing we would publish the results without modification once unblinded…

Cosmological Results

Blinded

Constraints from Two Lenses

In combination with WMAP7 in wCDM cosmology:

(Suyu et al. 2013)

Cosmological Probe Comparison

WMAP7owCDM prior

•  contour orientations are different: complementarity b/w probes

•  contour sizes are similar: lensing is a competitive probe

(Suyu et al. 2013)

Immediate Prospects

•  time delays of lensed quasars from optical monitoring

•  expect to have delays with a few percent error for ~20 lenses HST

cycle 20 follow up

HST archival images for lens modeling

Immediate prospects

Future Prospects

• Currently ~10 lenses have precise time-

delays

• Future telescopes (e.g.

LSST) will discover and measure 100s of time delays (Oguri &

Marshall 2010; Treu 2010)

• A time delay survey could provide very

interesting constraints

on dark energy

What’s the (dark) matter?

Warm Dark Matter

Lovell et al. 2014 Free streaming ~kev scale thermal relic

Satellites as a probe of dark

matter “mass”

Dark Satellites in CDM vs WDM

Nierenberg, Treu, et al. 2013

Luminous Satellites in CDM vs WDM

Nierenberg, Treu et al. 2013

“Missing satellites” and lensing

•  Strong lensing can detect satellites based solely on mass!

•  Satellites are detected as “anomalies” in the gravitational potential ψ and its derivatives

– ψ’’ = Flux anomalies

– ψ’ = Astrometric anomalies – ψ = Time-delay anomalies

–  Natural scale is a few milliarcseconds. Astrometric perturbations of 10mas are expected

“Missing satellites” and lensing

Treu 2010

Flux Ratio Anomalies

T.Treu: Flux ratio anomalies and the substructure problem 3

Figure 1. The substructure problem. In simulations (top, from Kravtsov 2010), galaxies and clusters are self-similar and should have the same amount of satellites. In reality, this is not observed: galaxies have many fewer (luminous) satellites than expected based on dark matter substructure. Does this mean they are dark, or that they do not exist? Answering this question is the goal of this program.

Figure 2. HST-F160W images of the targets.

A smooth mass distribution would predict:

This to be 100x brighter These to be 2x brighter This to be 10% brighter

What causes this the anomaly?

1. Dark satellites?

2. Astrophysical noise (i.e. microlensing and dust)?

Anomalies detected in 7 radio lenses

– 15 –

0.0010 0.01 0.1 1

0.05 0.1 0.15 0.2 0.25

0.0010 0.01 0.1 1

0.05 0.1 0.15 0.2 0.25

Fig. 5.— Results for the observed lens sample with b = 0.^′′001. The heavy solid lines show the probability distributions assuming errors in the flux ratios of 5%, 10% and 20%. The points on the curves mark the median surface density (triangles) and the regions encompassing 68.3% (1σ, squares), and 95.4% (2σ, pentagons) of the probability. The dashed curves show the contributions from the individual lenses for the 10% case. The region between the vertical lines is the range of substructure mass fractions found in the Klypin et al. (1999) simulations. Normal satellite populations, with 10⁻⁴ <∼ f^sat<∼ 10⁻³, correspond to a region off the left edge of the figure.

Dalal and Kochanek 2002

Fraction of mass in satellites

Probability

How do we make progress?

1.  Larger samples

2.  High precision photometry and astrometry

3.  Avoid microlensing

4.  Direct detection a.k.a. "gravitational

imaging"

Dusty Torus and Narrow Line Region

Are not affected by microlensing

T.Treu: Flux ratio anomalies and the substructure problem 4

Figure 3. Signal-to-noise ratio maps for the proposed experiment: The top row shows the expected S/N maps obtained by rescaling the total line flux by the flux ratios as measured in the continuum from HST. The bottom row shows the expected S/N maps obtained by rescaling the total line flux by the flux ratios predicted by smooth models without substructure (see Table 1). The difference is apparent by eye. All simulations have been performed using the OSIRIS ETC developed by David Law assuming exposure times of 7200s (for 0924 and 1138) and 3600s (for 1422). The S/N ratio scale shown is 0-50 for 0924 and 1138 and 0-150 for 1422. The field of view shown is the OSIRIS field of view for 0.05 pixels in the appropriate narrow band filter.

Figure 4. Mid-IR Subaru image of 1422;

note how A and B are blended, while D is un- detected (Chiba et al. 2005). Our experiment will detect D and resolve all four images (see Figure 3).

References:

Benefits:

1. Confirm/

eliminate microlensing 2. High

resolution

spectroscopy rules out

wavelength- dependent suppression (e.g. dust) 3. Excellent

astrometry and photometry

T.Treu: Flux ratio anomalies and the substructure problem 3

Figure 1. The substructure problem. In simulations (top, from Kravtsov 2010), galaxies and clusters are self-similar and should have the same amount of satellites. In reality, this is not observed: galaxies have many fewer (luminous) satellites than expected based on dark matter substructure. Does this mean they are dark, or that they do not exist? Answering this question is the goal of this program.

Figure 2. HST-F160W images of the targets.

If the anomaly is from

microlensing…

If the anomaly is from

substructure…

Narrow line flux ratios of lensed AGN

OSIRIS detection of substructure

Nierenberg Treu et al 2014

A B

C D

G

OSIRIS detection of substructure

Nierenberg Treu et al 2014

Astrometric perturbations:

gravitational imaging

Mass substructure distorts extended lensed sources

Vegetti et al. 2010

Direct detection of a dark substructure

Vegetti et al 2010, 2012 HST/AO can detect down to 1e8 Msun

Statistics from gravitational imaging

Vegetti et al 2010, 2012, 2014 HST/AO can detect down to 3e8 Msun

Gravitational imaging:

Future Prospects

•  Gravitational imaging can now reach ~10⁸ solar mass sensitivity, limited by

resolution and S/N

(Vegetti et al. 2012, 2014)

•  With Next Generation Adaptive Optics and then ELTs we should reach 10⁷ solar masses, where the discrepancy with theory is strongest

• LARGE SAMPLES WITH SUFFICIENT SENSITIVITY WITHIN REACH

Flux ratio anomalies:

Future Prospects

• Narrow line flux ratio anomalies can currently be studied for 10 systems

• Future surveys will discover thousands of systems

• ELTs will provide spectroscopic follow-up and

emission line flux ratios

100 quasar lenses with Flux ratios and time-delays.

How do we do this in

practice?

Roadmap. I. Find Lenses

•  Carry out large imaging survey.

•  QSO forecasts by Oguri & Marshall (2010)

•  DES (~1000 lensed QSOs, including 150 quads)

•  LSST (~8000 lensed QSOs, including 1000 quads)

•  Euclid/WFIRST many more!

•  Find lenses:

•  Different strategies for lensed QSOs and galaxies (Marshall+, Gavazzi+,Kubo+,Belokurov+,Kochanek +,Faure+,Pawase+,Agnello+) and under

development (Marshall, Treu, LSST collaboration)

•  Successfully demonstrated

In large imaging surveys

HSC on Subaru

WFIRST

LSST Dark Energy Camera

Needle in a haystack!

Which ones are lenses? Agnello, Kelly Treu & Marshall 2015

We can find them using

machine learning techniques

Agnello et al. 2015a

And here they are!

Agnello et al. 2015b

Roadmap. II. Follow-up

• 

High resolution imaging: space or Adaptive Optics

• 

Time delays: dedicated monitoring in the optical or radio

• 

Deflector mass modeling: redshifts and

stellar velocity dispersions (Keck/VLT/ELTs)

High resolution information. Where

will it come from?

Imaging landscape after HST

Euclid/LSST will be great for

discovery but not for cosmography

Meng, TT et al. 2015 Euclid

590s LSST

10 yr 4500s

>10% >10%

Contribution of modeling error To time delay distance

WFIRST will be probably good enough for the brighter lenses

Meng, TT et al. 2015 WFIRST

900s

HST 9000s

2%

5%

Imaging landscape after 2015:

Adaptive Optics

2012: 0.3-0.4 strehl at 2micron; improvements under way: PSF/TT

Marshall et al. 2007; Fassnacht

Imaging landscape after 2015:

Next Generation Adaptive Optics

• For strong lensing at galaxy scales interested in high-

strehl small fov:

• Keck-NGAO: 90%

strehl at K, 60% at J (not funded yet)

• Gemini,VLT, Subaru etc are all developing AO+

• Resources spread between large fov and high strehl

Imaging landscape after 2018:

Extremely Large Telescopes

• With 30-40m apertures and advanced AO, in principle one can attain 10x resolution of HST

•  TMT should have strehl

>70% at K and >30% at Y

Imaging landscape after 2018:

JWST

Ü 

JWST is 6.5m,

diffraction limited beyond 2micron

Ü 

At best resolution equal to HST at

~0.7micron

Ü 

0.032”/pix

Ü 

Ok down to 1micron or so, 0.65 strehl.

Ü 

Resolution ~HST

The bill

•  100 gravitationally lensed AGN with deep images of host galaxies at 100mas resolution or better; ~200-300 orbits with HST; 4 nights with Keck NGAO; very fast with TMT/ELT

•  ALMA?

•  Time delays: some for free from LSST; will they be accurate enough? DES follow-up will require dedicated small

telescopes (a la COSMOGRAIL, or LCOGT)

•  Redshifts of source and deflector: ~2 weeks of Keck; a few days of TMT / ELT. Easy with ALMA.

Conclusions

•  Strong gravitational lensing is a cost-effective tool to study the composition of the universe:

•  A dedicated time-delay program can achieve sub- percent accuracy on H₀ and increase figure of merit of other dark energy experiments by x5 or more

•  Flux ratios and gravitational imaging can probe the subhalo mass function down to 1e7 solar masses

and thus help rule out (or confirm) WDM

• 

This is feasible in the next five years with a

concerted follow-up effort of quasar lenses

discovered in DES and other imaging surveys

The end

Roadmap. III. Modeling

•  Extended sources

•  At the moment each lens requires months of work by an expert modeler, and months of CPU (e.g. Suyu+, Vegetti+).

•  Need to get investigator time down to hours/lens

•  Massive parallelization is required (GPUs?) for efficient posterior exploration and analysis of systematics