Exoplanet Microlensing Survey with the NEW Telescope (& WFIRST)

David BennettUniversity of Notre Dame

WFIRST

Unique Science from Space-based Survey• Exoplanet Survey Question #1: How do planetary systems

form and evolve?– complementary to Kepler– Exoplanet sensitivity down to sub-Earth masses at 0.5 AU - ∞– down to 0.1 Earth-masses over most of this range– free-floating planets down to 0.1 Earth-masses

• free-floating planet mass distribution is important for understanding planet formation.

• Exoplanet Survey Question #2: How common are potentially habitable worlds?– η = fraction of planetary systems with an earth-like planet in the outer

habitable zone– But what is earth-like? – Kepler results imply a wide variety of planetary systems– We need to answer question #1 to understand habitability

Microlensing Target Fields are in the Galactic Bulge

10s of millions of stars in the Galactic bulge in order to detect planetary companions to stars in the Galactic disk and bulge.

1-7 kpc from Sun

Galactic center Sun8 kpc

Light curve

Source starand images

Lens starand planet Telescope

Space vs. Ground Sensitivity

space

ground

Habitable Earths orbiting G & K stars accessible only from space

Expect ~190 free-floating Earths

How Low Can We Go?Limited by Source Size

θE ≈ μas

M p

M°Ú

⎛

⎝⎜

⎞

⎠⎟1/2

θ* ≈ μas

R*R°—

⎛

⎝⎜

⎞

⎠⎟

Mars-mass planets detectable

if solar-type sources can be monitored!

(Bennett & Rhie 1996)

angular Einstein radius

angular source star radius

For θE θ* :low-mass planet signals are rare and brief, but not weak

Ground-based confusion, space-based resolution

• Space-based imaging needed for high precision photometry of main sequence source stars (at low magnification) and lens star detection

• High Resolution + large field + 24hr duty cycle => Space-based Microlensing Survey

• Space observations needed for sensitivity at a range of separations and mass determinations

CTIO HST

High-magnification: Low-mass planets OGLE-2005-BLG-169Lb

• Detection of a ~17 M planet in a Amax= 800 event

• Caustic crossing signal is obvious when light curve is divided by a single lens curve.

• Detection efficiency for ~10 M planets is << than for Jupiter-mass planets

• Competing models with an Earth-mass planet had a signal of similar amplitude

• So, an Earth-mass planet could have been detected in this event, if it had a separation ~ RE !

FUN, OGLE, MOA & PLANET

Close Separation planets by Microlensing

• Faint main sequences sources needed to detecting low-mass planets• At separations < RE, planetary signals occur at low stellar magnification• Ground-based photometry seems to have systematic errors proportional to the

flux of blended stellar light.• For close-in (or HZ) planets, higher angular resolution & longer exposures help

s = 0.50 RE s = 0.25 RE

WFIRST vs. Kepler

Figures from B. MacIntosh of the ExoPlanet Task Force

WFIRST – w/ extended mission Kepler ~12 yr mission

Extraction of Exoplanet Light Curve Signal

Detailed fitting to the photometry yields the parameters of the detected planets.

Planets are revealed as short-duration deviations from the smooth, symmetric magnification of the source due to the primary star.

Time-series photometry is combined to uncover light curves of background source stars being lensed by foreground stars in the disk and bulge.

Mag

nific

atio

n

2.5

3

2

9.2 9.4 9.6 9.8Days

Deviation Due to Planet

Magnification by stellar lens

Offset from peak givesprojected separation

Mag

nific

atio

n

2.5

2

1.5

3

3.5

-20 -10 0 10 20Days

Twice EarthEarthHalf EarthNo Planet

~Neptune mass

€

tE =RE (M,D)v t

~ M /MJ dayM ： lens massMJ: Jupiter massD ： distancevt: velocity~ ２０ days for stars

WFIRST can detect Earth-mass FFPSumi et al. 2011As Many FFP as stars!

Free Floating Planet Eventshave tE < 2 days

WFIRST SDT Final Report is ObsoleteThere are issues with the exoplanet microlensing predictions

• Data Rate restrictions were ignored for DRM-2– With planned hardware, we can only observe 6 DRM-2

fields per 15 minute cycle (not 7)• Field locations and sampling was not optimized• Obsolete microlensing optical depth/rate numbers used

Microlensing Optical Depth & Rate

• MOA-II measures τ in vicinity of WFIRST field for 1st time (Sumi et al. 2012)– using 474 events with well measured microlensing parameters (215 in central fields)– luminosity function errors tend to cancel with all-star sample

• Selection of Red Clump Giant source stars is problematic– due to blending and possible luminosity function errors

• WFIRST SDT Final report on obsolete (on this point)

WFIRSTRed Clump Stars Only

GSFC NRO Design

These calculations use the GSFC design.

“only” 18H4RGs

WFIRST Field Selection & Data Rate

Preliminary optimization of DRM1 and DRM2 fields, with central MOA fields used for microlensing rate measurement. 9 and 6 fields, respectively, when optimum is 10.

Preliminary optimization of WFIRST-NRO fields, with central MOA fields used for microlensing rate measurement. L2 data rate limit is 5 fields. Optimum is 11 – assuming 15 minute sampling.

Assumes 40 sec slew & settle – probably conservative – we’d do more fields if we could

Comparison of Earth-mass Detection Numbers

DRM1-rep

DRM2-rep

DRM-1 DRM-2 NRO

L2-limited data rate 239 - 581 402 440

No data rate limit 239 176 610 483 821

Current L2 data rate limit makes L2 version of WFIRST inefficient. More pixels per star => fewer planet discoveries.

Likely trade: L2-NRO will require more microlensing observing time and less dark energy time. But, a factor-of-2 increase in L2 downlink would remove this issue. Maybe a larger high-gain antenna would be enough?

Photometry improvement from NRO is modest – but it may be more significant near the habitable zone.

The results assume Cassan et al. (2012) exoplanet MF and a detection threshold of Δχ2 > 300 in mission simulations using an IR version of the Bennett & Rhie (2002) code.

MPF in Geosynchronous Orbit

MPF fieldVernal Equinox Orbit Plane

Orbit

Ecliptic Plane

Equatorial Plane

W

23.528.7

Sun

MPF’s orbit allows continuous view of Galactic bulge planet search field and continuous data data downlink to a dedicated ground station in White Sands.

W

Detector Radiation Shield for Outer van Allen Belt

STScI Design

MPF study indicates “sealed” multilayer shield can remove trapped e- radiation

To get 100K focal plane may require multiple radiators protected by louvers

Lens System Properties

• Einstein radius : θE= θ*tE/t* and projected Einstein radius, – θ* = the angular radius of the star– from the microlensing parallax effect (due to Earth’s orbital motion).

%rE

%rE

RE =θEDL, so α =

%rEDL

=4GMc2θEDL

. Hence M =c2

4GθE%rE

• If only θE or is measured, then we have a mass-distance relation.

• Such a relation can be solved if we detect the lens star and use a mass-luminosity relation–This requires HST or ground-based

adaptive optics• With θE, , and lens star

brightness, we have more constraints than parameters

Finite Source Effects & Microlensing Parallax Yield Lens System Mass

M L =c2

4GθE

2 DSDL

DS −DL

ML =c2

4G%rE2 DS −DL

DSDL

ML =c2

4G%rEθE

%rE mass-distance relations:

%rE

HST Observations of OGLE-2003-BLG-235L/MOA-2003-BLG-53L

Lens and source perfectly aligned during event

Lens moves away form source after event, so centroid position of blended image is color dependent

Relative proper motion rel= 3.30.4 mas/yr from light curve analysis (rel= θ*/t*) implies 0.6 mas separation in B - I

HST Observation Predictions for OGLE-2003-BLG-235L/MOA-2003-BLG-53L

Fraction of total flux due to lens star.

Centroid Shift between HST-ACS/HRC passbands for follow-up images. (Units are 25 mas pixels.)

Relative proper motion rel= 3.30.4 mas/yr from light curve analysis (rel= θ*/t*)

Lens Star Identification from Space• Lens-source proper motion

gives θE = reltE

• rel= 8.40.6 mas/yr for OGLE-2005-BLG-169

• Simulated HST ACS/HRC F814W (I-band) single orbit image “stacks” taken 2.4 years after peak magnification – 2 native resolution– also detectable with HST

WFPC2/PC & NICMOS/NIC1• Stable HST PSF allows clear

detection of PSF elongation signal

• A main sequence lens of any mass is easily detected (for this event)

ML= 0.08 M

ML= 0.35 M

ML= 0.63 M

raw image PSF subtracted binned

Simulated HST images:

Stacked HST I-band Image of OGLE-2005-BLG-169 Source

HST images taken 6.5 yrs after event

Analysis by Jay Anderson

Subtracted Neighbor…

PSF ISGOOD!

Almost noresidualsWhen we Subtract aPSF from a(brighter)neighbor

Subtracted F814W Stack

This meansthat the residuals ofthe target-star subtractionare real.

Fit and Subtract Two Stars: Source & Lens

Very good subtractionresiduals when we fit for two sources

Two-source Solution:• Offset consistent in the

F814W, F555W, and F438W data:– Dx = 1.25 pixels = 50 mas– Dy = 0.25 pixel = 10 mas– FLUX: (left)

(right)• F814W 3392 e 3276 e-

• F555W 2158 e 3985 e-

• F438W 338 e

1029 e

• fI = 0.51• fV = 0.35• fB = 0.25

HST BVI observations imply M* = 0.63 M

Mp = 17 M

observed separation of 51 mas confirms planet model prediction of 54.3±3.7 mas

High-magnification: Low-mass planets OGLE-2005-BLG-169Lb

• Detection of a ~17 M planet in a Amax= 800 event

• Caustic crossing signal is obvious when light curve is divided by a single lens curve.

• Detection efficiency for ~10 M planets is << than for Jupiter-mass planets

• Competing models with an Earth-mass planet had a signal of similar amplitude

• So, an Earth-mass planet could have been detected in this event, if it had a separation ~ RE !

FUN, OGLE, MOA & PLANET

HST vs. WFIRST Astrometry• Our examples are all with HST data

– 2.4m diffraction limited optics– Imaging at λ = 0.8 μm– ~40 mas/pixel – 6-8 images per passband

• WFIRST– 1.1-2.4m diffraction limited optics– Imaging at λ = 1-2 μm– ~110-180 mas/pixel – 500-50,000 images per passband– Very wide filter 0.92-2.4 μm filter implies color dependent PSF– rel uncertainty scales as ~ FWHM2 so ~4× better w/ NRO

Relative Proper Motion

Gaudi’s relative proper motion from White Paper for J = 23 source+lens

Lens Detection Provides Complete Lens Solution

Estimates of the parameter uncertainties from these methods for the proposed Microlensing Planet Finder mission concept, which was one of 3 WFIRST precursors (Bennett, Anderson & Gaudi 2007). WFIRST should do at least as well as this.

WFIRST-NRO should do much better.

However, the Central Bulge is More Crowded in the IR CTIO HST/WFPC2

HST/WFC3/IR

Crowded fields give higher lensing rate, but complicate mass determination -> redundancy needed