Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic...

40
P.Astier LSST@Belgrade (06/16) 1 Cosmology with supernovae in LSST Pierre Astier (LPNHE-Paris)

Transcript of Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic...

Page 1: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)1

Cosmology with supernovae in LSST

Pierre Astier (LPNHE-Paris)

Page 2: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)2

Foreword

● These slides were not blessed by the Supernova working group....

● … but they are not orthogonal to the concerns being followed there.

● … and are somehow twisted towards possible European contributions in the SN distance arena.

Page 3: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)3

The role of supernovae in constraining cosmology

● Distances are a “must do”– Constrain the average background/expansion

history and in particular Dark Energy.

– A successful approach so far.

JLA sample(Betoule et al2014)

Page 4: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)4

No more spectroscopic confirmation

● The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically identify all events in the Hubble-diagram

– As much 8-m time (spectroscopy) than CFHT time (imaging).

– Spectroscopy time was the bottleneck: some candidates could just not be observed.

– SDSS was in the same situation.

● DES-SN lives in this new paradigm:– Hosts redshifts on the AAT.

– “Live” SNe (~10%) on VLT and others .

Page 5: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)5

No more spectroscopic confirmation

● We need good (i.e low shot-noise) light curves:– For identification: we do not want the high-z end to

be more contaminated than the low-z end.

– At least 3 bands, for both light curve shapes and 2 colors or more.

– 3 bands improve the lever arm for color measurement.

● Good light curves come with other benefits:– Can effectively assemble a redshift-limited Hubble

diagram (no more Malmquist bias !)

– Efficient light curve model training.

Page 6: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)6

How well should light curves be measured (for distances)?

● Quality commensurate with natural spread...

– Brighter-bluer correction : c , with ~3

– color to 0.04 → distance modulus to ~ 0.12

Not very challenging:SNLS did it.(Guy et al 2010)

Page 7: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)7

How well should light curves be measured ?

● Precision of LC amplitudes

From Guy et al (2010)

Bottom-line:SNLS derived colorsessentially from gri

Page 8: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)8

Improving SN distances :light curve shot noise

Simulation for LSSTWith a 4-day cadence and ~10 mn per visit.

→ Significant improvement over SNLS and DES→ We might consider trading i and z depth in favor of y.

Page 9: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)9

Improving SN distances :rest-frame spectral coverage

● Rest frame UV of SNe is useless not very useful:– Only handful of nearby counterparts

– Large event to event scatter

– Suspicions of evolution

– Steep extinction curve

Page 10: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)10

Suspicions of UV evolution of SNeMaguire et al (2012) : - comparison of distant (z~0.5) and nearby (HST) spectra - empirical measurement of scatter

Marginal effectBut anyhow worrying

(Very) large scatter

Page 11: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)11

Suspicions of UV evolution of SNe

Marginal effectBut anyhow worrying

(Very) large scatter

i-band at z=0.95Maguire et al (2012)

Page 12: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)12

Redshift coverage for SNe with LSST

3 bands at >380nm→ z <= ~0.95

Page 13: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)13

The weight of high-z JLA events

JLA sample(Betoule et al2014)

(w) = 0.0575 (740 events)(w) = 0.0605 (z<0.8, 674 events)

Megacam has a low-sensitivity z band, and no y band.→ There is not much to gain at diving into deep UV

Page 14: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)14

What we would like for the LSST SN cosmology program:

● A deep enough “deep program” :– Means O(3) mn/day per band and per pointing.

– ~ Deep Drilling Fields observing scheme.

– We could think of exposing more in y and less in i and z.

● And a low-z SN program !– The average repetition in the wide is far too low to

measure distances.

– The often quoted number of SNe (O(105)) refers to detections, not distances.

Page 15: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)15

The LSST observing strategies

● Area● Time sampling

per band● Depth per visit

Wide Deep-Drilling Fields

~20 days ~5 days

23.5 → 24.5 ~20 to 50 timesmore

20,000 deg2 O(100) deg2

Page 16: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)16

Do we really care about low-z SNe ?

● Consider 5 DDF over 4 seasons (conservative)

→ yields ~ O(8000) SNe to z~0.95, very well measured.

● Add low-z events

~SDSS : z<0.35

– natural z distribution

For forecast hypotheses, seeArXiv:1409.8562.

Page 17: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)17

So,

● The high-z part of the LSST Hubble diagram is reasonably secured in the current observing plan

● Some part of the Wide survey should be “unevenly” scheduled to get a few thousand SNe in the SDDS redshift range (standard 30s visit).

– O(3000) deg2 for 6 months at a 4 day cadence.

– The DESC SN WG is indeed working on the cadence of the Wide LSST Survey.

– Without it, SN constraints are going to be poor.

Page 18: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)18

What are the identified limitations of LSST SN cosmology?

● Photometric calibration – dominates systematic uncertainties now

● Getting (precise) redshifts● The “host correction” (a.k.a “mass step”)● Identification (Ia vs others)● The redshift lever arm (the longer, the better)

Page 19: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)19

Photometric calibration

● Today, the zero points uncertainties are ~5 mmag and translate to (w) = 0.03 (Betoule et al, 2014)

● DESC (PCWG) is committed to transferring the NIST photometric physical scale to stellar standards.

DE FoM for a SN Hubble diagram, as a function of calibrationaccuracy.

~ 8k DDF SNe~ 8k Nearby SNeDetails in arXiv:1409.8562

Now

Goal

Page 20: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)20

Redshifts for the Hubble diagram

● The emerging standard way: collecting host redshifts after the fact.

● Requires a large multiplex because needed exposure times are O(20-50) ks on a 4m (z~1).

● For the high-z part , it requires dedicated time on a wide-field MOS instrument : 4most for south.

● Low-z : rely on the 4most survey footprint ?● Other suggestions ?

Page 21: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)21

The host correction (1)

● Since ~2010, all SN papers apply a host-dependent distance correction.

Residuals to the Hubble diagramare related tosome host galaxyproperty: e.g. the stellar mass

“The mass step”(Betoule+ 2014)

Page 22: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)22

The host correction (2)

M. Rigault & the SNfactory (2013) : the brightness of SNe Ia is related tothe local H emission.

M. Roman & co (2016, in prep) : the brightness of SNe is relatedto the local restframe U-V color (SFR indicator):

● 7 significance.● Makes more sense than some host mass ● Relies on imaging only

SNLS5 sample

Page 23: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)23

SN photometric identification

● Do we really need a very pure HD ?– Redshift-independent contamination is OK.

● Quality of light curves helps:– Low shot-noise

– 3 bands (to rely on color relations)

● A lot of work currently underway in DESC SN WG.

● Requires “live SN” spectroscopy to assess contamination: VLT !

Page 24: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)24

The redshift lever arm : z<0.1

Adding events to the base sample:● z<0.1 is the best deal● it is also the most difficult in practice

A few hundred of SNe @ z~0.05 are very useful for HD cosmology.● Detection by LSST is guaranteed● Follow up by LSST ? How ?● Follow-up with small (<2m) telescopes ? ●VST will soon be de commissioned● Will ZTF have done that by 202X ?

Page 25: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)25

The redshift lever arm : z>1

Proposal fo

r Eucli

d

left-o

ver tim

e

(if any)

LSST+ EuclidJoint SN survey

Page 26: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)26

What else can we do with SNe ?LSS constraints !

● Lensing (DDF SN sample).– Skewness of the distribution of residuals (or

evolution of the skewness with redshift).

– Angular correlation of residuals.

● Velocity correlations (nearby SN sample z<~0.1)– Requires a “sizable” time*volume at low z

– Constraints on and

e.g. Castro et al (2016)

Page 27: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)27

Velocity correlations from nearby SNe

1511.08695

3 realizations of HubbleResiduals (400 deg2), ignoring “intrinsic scatter”

Constraints from 1000 nearby SNe Uniform over 0.01<z<0.1 and 400 deg2

→ all events over ~ 50 years (!)

Page 28: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)28

Velocity correlations from low-z SNe

● The forecast from 1511.08695 seems unrealistic:– But we can aim at a few thousand events on a few

thousand deg2.

– Detection with LSST + follow-up elsewhere ?

– Worth the effort after ZTF ?

● But SN cosmology benefits a lot from nearby events:

– Effectively increases the redshift lever arm

– Light-curve fitter training.

Page 29: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)29

The LSST SN Hubble diagram

Wide : negociate adequatecadencingof some patches/seasons

Nearby : Considerpointed follow-upwith small telescopes

z>1 : remind Euclidheads that Euclidis about cosmology

Redshifts:4most, DESI,….

Distances:bridge laboratoryand stellar standards

DDF: monitorthe observing strategy. y bandis vital.

Page 30: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)30

The LSST SN Hubble diagram

DDF: monitorthe observing strategy. y bandis vital. Wide :

negociate adequatecadencingof some patches/seasons

Nearby : Considerpointed follow-upwith small telescopes

z>1 : remind Euclidheads that Euclidis about cosmology

Redshifts:4most, DESI,….

Distances:bridge laboratoryand stellar standards

Where EuropeCouldcontribute

Page 31: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)31

Summary/conclusions

● LSST is well equipped to assemble the best supernova Hubble Diagram of the next decade:

– Wide: SN statistics

– Fast: allows to collect low-z events

– Deep : actually reach z~1 with good light-curves.

● Collecting low-z light curves (z<0.35) requires some alterations to the base observing plan (“Rolling cadence”, within a constant allocation !)

● European facilities could play a leading role in follow-up

Page 32: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)32

More Slides

Page 33: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)33

Simulated surveys

Typically 20,000 SN events in total,with ~2,000 measured by Euclid NISP.

The assumed exposure timesdeliver precise distancesat the highest redshifts.

Euclid+LSST

Page 34: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)34

Lightcurve examples

The contribution ofshot noise to distanceuncertainty is at most 0.09,when the “natural” scatteraround the Hubble diagramis ~0.12

Page 35: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)35

Forecast approach

● Consider major sources of uncertainty :– SN statistics and photometric shot noise (including

finite depth of SN-free data).

– “natural” Hubble diagram scatter

– Photometric calibration uncertainties

– SN empirical “modeling” uncertainties● Finite training statistics● Irreducible residual scatter

– Ground distance error floor

● Conservative uncertainties: current state of the art.

Page 36: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)36

Forecast method

● Consider all parameters simultaneously and marginalise.

● Parameter sets:

– SNe lightcurve parameters (brightness, color, decline rate, date of maximum)

– SN model parameters (average SN restframe flux and color variations)

– Calibration offsets

– Distance estimator coefficients

– Cosmological parameters

● Typically 50,000 parameters

● Evaluate a sub-block of the inverse of the Fisher matrix

Page 37: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)37

The importance of calibration uncertainties

No calibration uncertaintyNo model uncertainty

The drop happens with both calibration and modeltraining turned on

With all uncertaintysources

FoM obtained using SNe+Planck (geometry only).

Page 38: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)38

Sensitivity to systematics

Irreducible distance error

NIRZPaccuracy

Page 39: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)39

Model-free forecast (1)

● We propose a covariance matrix of distances in redshift bins.

● If the tested model is not widely different from the vanilla cosmology, the obtained precision is accurate.

● We tested that we recover the direct fit precision for CDM, at an accuracy better than 1%.

● Of course, no external prior added: just SNe.

Page 40: Cosmology with supernovae in LSST · P.Astier LSST@Belgrade (06/16) 4 No more spectroscopic confirmation The SuperNova Legacy Survey (2003-08) was “able” to spectroscopically

P.Astier LSST@Belgrade (06/16)40

Model-free forecast (2)

● Paper : A&A 572 A80 (2014) http://arxiv.org/abs/1409.8562

● The appendix E describes the binned distance prior.

● We provide the covariance matrix, short note, and some code example at : http://supernovae.in2p3.fr/~astier/desire-paper/SN-prior/