Together we are stronger: synergies in astrophysics and AI to advance our understanding of the Universe

University of Delaware

Department of Physics and Astronomy

Biden School of Public Policy and Administration

Data  Science Institute

 

 

federica b. bianco

she/her

Site: Cerro Pachon, Chile

Funding: US NSF + DOE

 

Vera C. Rubin Observatory

Building an unprecedented catalog of Solar System Objects

LSST Science Drivers

LSST Science Drivers

Building an unprecedented catalog of Solar System Objects

LSST Science Drivers

Mapping the Milky Way and Local Volume

LSST Science Drivers

Building an unprecedented catalog of Solar System Objects

LSST Science Drivers

Mapping the Milky Way and Local Volume

LSST Science Drivers

Probing Dark Energy and Dark Matter

Building an unprecedented catalog of Solar System Objects

LSST Science Drivers

Mapping the Milky Way and Local Volume

LSST Science Drivers

Probing Dark Energy and Dark Matter

Exploring the Transient Optical Sky

 

 

 

 

To accomplish this, we need:

1) a large telescope mirror to be sensitive - 8m (6.7m)

2) a large field-of-view for sky-scanning speed - 10 deg2

3) high spatial resolution, high quality images - 0.2''/pixels

4) process images in realtime and offline to produce 10M nightly alerts and catalogs of all 37B objects

 

 

 

>=18000 sq degrees

~800 visits per field

2 visits per night (within ~30 min for asteroids)

+ 5x10sq deg Deep Drilling Fields with ~8000 visits

Objective: to provide a science-ready dataset to transform the 4 key science area

Are We There YET????!!!!

The DOE LSST Camera - 3.2 Gigapixel

3024 science raft amplifier channels

Camera and Cryostat integration completed at SLAC in May 2022,

Shutter and filter auto-changer integrated into camera body

LSSTCam undergoing final stages of testing at SLAC

Eye to the sky…on-sky engineering tests have begun at @nsfgov@energy Rubin Observatory using the world’s largest digital camera!🔭

 

 

April 17
 

June 23 - 16:00 CEST

Text

Text

LSST Science Verification June-September 2025

Dome work September-October 2025

LSST start >October 2025

Comcam commissioning  September-December 2024

LSSTcam commissioning  April-June 2025

Rubin LSST survey design

Rubin LSST survey design

Rubin has involved the community to an unprecedented level in survey design this is a uniquely "democratic" process!

2024

2024

80,000

Rubin LSST Transients by the numbers

 

 

17B stars (x10) Ivezic+19

~10k SuperLuminous Supernovae (from ~200)Villar+ 2018

~400 strongly lensed SN Ia (from 10) Ardense+24

~3.3M SN II 580k SN Ibc Hložek+20

~1M SNIa (z<0.5) 1M SNIa (z<2) Griz+24

~50 kilonovae (from 2) Setzer+19, Andreoni+19   (+ ToO)

> 10 Interstellar Objects fom 2....    ?)

 

 

 

 

 

Rubin LSST Transients by the numbers

 

 

17B stars (x10) Ivezic+19

20B Galaxies Ivezic+19

~3.3M SN II 580k SN Ibc Hložek+20

~1M SNIa (z<0.5) 1M SNIa (z<2) Griz+24

~50 kilonovae (from 2) Setzer+19, Andreoni+19   (+ ToO)

> 10 Interstellar Objects fom 2....    ?)

 

 

 

 

 

edge computing

Rubin LSST Transients by the numbers

 

 

17B stars (x10) Ivezic+19

20B Galaxies Ivezic+19

~35k-80k TDEs Brickman&Gombok20

~3.3M SN II 580k SN Ibc Hložek+20

~1M SNIa (z<0.5) 1M SNIa (z<2) Griz+24

~50 kilonovae (from 2) Setzer+19, Andreoni+19   (+ ToO)

> 10 Interstellar Objects fom 2....    ?)

 

 

 

 

 

Rubin LSST Transients by the numbers

 

 

17B stars (x10) Ivezic+19

20B Galaxies Ivezic+19

~35k-80k TDEs Brickman&Gombok20

103 galaxy-scale strongly lensed AGN systems O'Brian20

~1M SNIa (z<0.5) 1M SNIa (z<2) Griz+24

~50 kilonovae (from 2) Setzer+19, Andreoni+19   (+ ToO)

> 10 Interstellar Objects fom 2....    ?)

 

 

 

 

 

True Novelties!

Rubin LSST Transients by the numbers

 

 

17B stars (x10) Ivezic+19

20B Galaxies Ivezic+19

~35k-80k TDEs Brickman&Gombok20

103 galaxy-scale strongly lensed AGN systems O'Brian20

~400 strongly lensed SN Ia (from 10) Ardense+24

~50 kilonovae (from 2) Setzer+19, Andreoni+19   (+ ToO)

> 10 Interstellar Objects fom 2... (.    ?)

 

 

 

 

 

described in ls.st/LDM-612

world public!

10M alerts per night!! anything that changed by >5σ from "how the sky usually looks"

described in ls.st/LDM-612

proprietary

image stacks at expected depth

(ugrizy)=(23.80 24.50 24.03 23.41 22.74 21.96)

annual catalogs of 20B Galaxies and 17B stars 20PB

"BUT BIG DATA DOES NOT MEAN BIG SCIENCE"

 

Yang Huang,
University of Chinese Academy of Sciences

SpecCLIP talk

UniverseAI, Athens Greece

June 5, 2025

Synergies

- footprint coordination

- cadence coordination

- simulation coordination

- joint data processing

LSST Synergies

 

 

Roman+Rubin+Euclid: ~2,000 square degrees
- Roman High Latitude Wide Area Survey (HLWAS) southern extragalactic sky region
Roman+Euclid: ~2,000 square degrees
- HLWAS southern extragalactic sky region
Roman+Rubin: ~2,000+ square degrees
- HLWAS southern extragalactic sky region
- plus southern Galactic plane overlap

- Rubin+Euclid: ~7,000 square degrees
- southern extragalactic sky

LSST Synergies:

footprints

via Aaron Meisner - Science Synergies and Joint Survey Processing with Roman and Euclid - Rubin Community Workshop 2024

Euclid DFS (15k x2)

COSMOS (48k)

XMM-LSS  (22-24k)

CDF-S (22-24k)

ELAIS-S1 (22-24k)

LSST Synergies:

footprints

Brandt+12,18 AGN SC

LSST Synergies:

footprint

coordination

10k SN in Euclide DDF South

7% of LSST data

26.9

source http://astro-lsst-01.astro.washington.edu:8080/?runId=2

26.8

29.05

26.8

LSST DDFs

28.1

26.8

LSST DDFs

28.1

26.8

- footprint coordination

- cadence coordination

- simulation coordination

- joint data processing

LSST DDFs

28.1

26.8

  • V3.0 introduced near-sun twilight microsurvey (PSTN-055 recommendations 2.7.2); v3.2 updated this near-sun survey to be even nearer to the sun

  • We expect a call for microsurveys to be issued between April and June 2026 with a 2-3 month window

  • Criteria for selection:

Micorsurveys

  • Uses <3% of the LSST time

  • Unique use of the Rubin facility (e.g. need wide FoV)

  • Timely (scientific loss if delayed)

  • Compelling scientific case

- footprint coordination

- cadence coordination

- simulation coordination

- joint data processing

LSST DDFs

lssttooworkshop.github.io/

LSST ToO program

LSST ToO program

lssttooworkshop.github.io/

Joint data processing

Synergy with space photometric surveys

via Aaron Meisner - Science Synergies and Joint Survey Processing with Roman and Euclid - Rubin Community Workshop 2024

Synergy with space photometric surveys

The lower surface brightness limit of HSC allows fainter features in HST images to be modeled. Joint modeling is even more beneficial for the combination of LSST (comparable to HSC, i ~ 27) and the less sensitive Euclid (VIS < 25.2).
Figure courtesy Rémy Joseph (Princeton).

Synergy with space photometric surveys

"Comparison of these fits to SDSS spectra showed that emission line properties inferred from broad-band photometry were consistent with the results from spectroscopy for 91 per cent of objects."

 

 

 

 

Adam Marshall+2022

Synergy with space photometric surveys

Extending LSST coverage into the NIR (Roman/Euclide)

(e.g. UK Inkind HSC+VISTA, Banijer+Shirley) 1-1 pixel matching at the image level

=> source detection and measurement simultaneously in optical + NIR

Raphael Shirley

 

Synergy with space photometric surveys

https://www.youtube.com/watch?v=f9CYPKRsqeU&list=PLwQ-l1lnSF618jb5Lf5ESbKXACQ7yHKi-&index=7&t=3022s

We present an extension of the multi-band galaxy fitting method scarlet which allows the joint modeling of astronomical images
from different instruments, by performing simultaneous resampling and convolution

AI-aided data processing

AI-assisted superresolution cosmological simulations

Yin Li+2021

LOW RES SIM

HIGH RES SIM

AI-AIDED HIGH RES

AI-assisted superresolution cosmological simulations

Yin Li+2021

LOW RES SIM

HIGH RES SIM

AI-AIDED HIGH RES

INPUT

OUTPUT

TARGET

loss = D(OUTPUT-TARGET)

Generative AI

Generative AI

Generative AI

approximate q(xt−1∣xt)q(xt1xt) with a parameterized model pθ which (an ANN).

For small enough βttdiffusion steps,  p(t)~G(μ,σ)G(μ,σ)

 

Generative AI

approximate q(xt−1∣xt)q(xt1xt) with a parameterized model pθ which (an ANN).

For small enough βttdiffusion steps,  p(t)~G(μ,σ)G(μ,σ)

 

Generative AI

LSST follow-up

Swayamtrupta Panda's talk yesterday said all there was to say!

Discovery Engine

10M alerts/night

Community Brokers

target observation managers

When they go high, we go low... spectra classification at low resolution

Astrophysical spectra require the capture of enough photons at each wavelength: 

 

large telescopes

long exposure times

bright objects

Willow Fox Fortino

UDelaware

When they go high, we go low

Classification power vs spectral resolution for SNe subtypes

 

 

FASTlab Flash highlight

Willow Fox Fortino

UDelaware

When they go high, we go low

Classification power vs spectral resolution for SNe subtypes

 

 

FASTlab Flash highlight

Willow Fox Fortino

UDelaware

When they go high, we go low

Classification power vs spectral resolution for SNe subtypes

 

 

FASTlab Flash highlight

Kaggle PLAsTiCC challenge

AVOCADO classifier

https://arxiv.org/abs/1907.04690

Classification from sparse data: Lightcurves

Viswani 2017 Attention is all you need

AI was transformed in 2017 by this paper

Willow Fox Fortino

UDelaware

When they go high, we go low

Classification power vs spectral resolution for SNe subtypes

 

 

FASTlab Flash highlight

Willow Fox Fortino

UDelaware

When they go high, we go low

Classification power vs spectral resolution for SNe subtypes

 

 

FASTlab Flash highlight

Willow Fox Fortino

UDelaware

When they go high, we go low

Classification power vs spectral resolution for SNe subtypes

 

 

Willow Fox Fortino

UDelaware

FASTlab Flash highlight

Text

A new AI-based classifier for SN spectra at low resolution

Willow Fox Fortino

UDelaware

Optimal deep learning architectures for transients' spectral classification

Willow Fox Fortino

UDelaware

Optimal deep learning architectures for transients' spectral classification

Ask not what data I need to answer my questions, ask what questions can I ask of these beautiful data!

(paraphrased from Z. Ivezić)

atmosphere-aided flare studies with LSST

Riley Clarke, UDelware

dM                 Flare energy

dDCR               color                Flare temperature

NSF Award #2308016

 

Riley Clarke et al. 2024 ApJS

on the job market!

LSST ΔDCR detectability

Riley Clarke, UDelware

on the job market!

atmosphere-aided flare studies with LSST

Riley Clarke, UDelware

on the job market!

Riley Clarke, UDelware

on the job market!

Siddarth Chiaini, UDelaware

 

Most classifiers for variable stars use Random Forest (not distance based)

In distance based classification, optimal distances can be found for the class of interest: flexible, customizable, efficient

https://arxiv.org/pdf/2403.12120.pdf

Astronomy and computing

FASTlab Flash highlight

Shar Daniels is a NSF Graduate Research Fellow.

They use telescopes and cameras in innovative ways to show the stars in their time evolution at milliseconds rate

and uses cutting edge AI (transformers) to discover new physical phenomena

NSF Graduate Research

Fellowship Program

time: 1 pixel = 3.0 milliseconds

space: 1 pixel = 1 arcsecond

Faster transients

Any phenomenon that changes rapidly, in less than hours, is a technological challenge in astronomy 

e.g. blazars, stellar flares

FASTlab Flash highlight

Multi-city Urban Observatory Network

Studying the Universe cities as complex systems through imaging data

Studying cities as complex systems through imaging data

  • energy demand and consumption
  • ecology of flora and fauna
  • urban metabolism
  • circadiem rhythms

Studying the Universe cities as complex systems through imaging data

Multi-city Urban Observatory Network

Multi-city Urban Observatory Network

Multi-city Urban Observatory Network

Plumes and heat in NIR

Multi-city Urban Observatory Network

Plumes in hyperspectral imaging

Multi-city Urban Observatory Network

Us-topia:

the LSST Science Collaborations

An international community of practice built on principles of cooperation, equity, and solidarity

Rubin LSST Science Collaborations

8 SCs - 6 continents - 2000 people - 25 countries

is a word I am borrowing from Margaret Atwood to describe the fact that the future is us. 

However loathsome or loving we are, so will we be. 

Whereas utopias are the stuff of dream dystopias are the stuff of nightmares, ustopias are what we create together when we are wide awake

https://www.youtube.com/watch?v=QO3nY_u6hos

US-TOPIA

thank you!

 

University of Delaware

Department of Physics and Astronomy

 

Biden School of Public Policy and Administration

Data  Science Institute

federica bianco

fbianco@udel.edu

bit.ly/FBBscslsa25

Rubin Observatory

Site: Cerro Pachon, Chile

Funding: US NSF + DOE

 

 

 

 

 

To accomplish this, we need:

1) a large telescope mirror to be sensitive - 8m (6.7m)

2) a large field-of-view for sky-scanning speed - 10 deg2

3) high spatial resolution, high quality images - 0.2''/pixels

4) process images in realtime and offline to produce live alerts and catalogs of all 37B objects 

 

 

 

Objective: provide a science-ready dataset to transform the 4 key science area

2025

@fedhere

At this level of precision,everything is variable, everything is blended, everything is moving.

SDSS

LSST-like HSC composite

Field of View'
Image resolution'

DDFs'
Standard visit'
Photometric precision'
Photometric accuracy'
Astrometric precision'
Astrometric accuracy'
9.6 sq deg
0.2'' (seeing limited)

5 DDF
30 sec
5 mmag
10 mmag
10 mas
50 mas

' requirement: ls.st/srd

*simulation pstn-054.lsst.io

SDSS 2x4 arcmin sq griz

MYSUC (Gawiser 2014) 1 mag shallower than LSST coadds

u,g,r,i,z,y
Photometric filters'
saturation limit'
# visits*
mag single image*
mag coadd*
Nominal cadence
​u, g, r, i, z, y
~15, 16, 16, 16, 15, 14
53, 70, 185, 192, 168, 165
23.34, 23.2, 24.05, 23.55 22.03
25.4, 26.9, 27.0, 26.5, 25.8, 24.9
2-3 visits per night

At this level of precision,everything is variable, everything is blended, everything is moving.

' requirement: ls.st/srd

*simulation pstn-054.lsst.io

Rubin LSST Science Collaborations

8 SCs - 6 continents - 2000 people - 25 countries

number are quite a bit larger now! this plot is from ~2022

on the right is a connectivity network for the SCs

Rubin Observatory Status

5 / 2019

May 2022 - Telescope Mount Assembly

 

12/2022 TMA in action

weight 2e5 kg, max slew rate 0.2 rad/s

Most of the weight in a 10m disk
Angular momentum

5,000,000 ~\mathrm{kg~m^2 s^{-1}}

The DOE LSST Camera - 3.2 Gigapixel

3024 science raft amplifier channels

Camera and Cryostat integration completed at SLAC in May 2022,

Shutter and filter auto-changer integrated into camera body

LSSTCam undergoing final stages of testing at SLAC

July 2024 ComCam installed on  the telescope after M1M2 installation - Comcam is a 144Mpix version of LSSTCam

artist (me) impression of the first image taken by ComCam

https://community.lsst.org/c/news/7

 telescope's optical alignment close to optimal and the system delivering an image quality of around 1.5 arcseconds. 

The Data Management system effectively processed images, providing astrometric and photometric solutions.

AOS commissioning successfully running the system in a closed-loop configuration.

Continuous improvements in image quality were observed, achieving sub-arcsecond PSF FWHM

achieving PSF FWHM of 0.7" on several nights.

Observations for science pipeline commissioning continued, including filter exchanges and initial tests of difference image analysis and photometric calibration.

AOS closed loop was operating with over 90% of the optical degrees of freedom enabled.

 Full-color six-band (ugrizy) coverage of the Extended Chandra Deep Field South (ECDFS) LSST Deep Drilling Field.

Flux measurements demonstrated high repeatability across multiple nights and varying airmass.

The Data Management System successfully processed data with Difference Image Analysis, producing candidate sources and alerts.

Efforts were focused on increasing operational efficiency, including shutter open efficiency and telescope motion speed.

Repeated imaging in r and i bands allowed to build templates for six target fields. 

 First association of Solar System Object detections

A new single-night record of 99 in-focus visits taken with the automated scheduler.

LSST

data products

federica bianco - fbianco@udel.edu

Time

Domain

Science

 

Static

Science

Alerts based

 

Catalog based

Deep stack

based

Deep stack

based

Rubin Observatory LSST 

federica bianco - fbianco@udel.edu

Data Products

federica bianco - fbianco@udel.edu

data right holders only

federica bianco - fbianco@udel.edu

Rubin In-Kind Contribution Program

https://www.lsst.org/scientists/in-kind-program

federica bianco - fbianco@udel.edu

world public!

10Million alerts per night!

LSST survey strategy optimization

Exploring the Transient and Variable Optical Sky

Exploring the Transient and Variable Optical Sky

Exploring the Transient and Variable Optical Sky

Exploring the Transient and Variable Optical Sky

Exploring the Transient and Variable Optical Sky

Exploring the Transient and Variable Optical Sky

LSST Science Book (2009)

Operation Simulator (OpSim)

simulates the catalog of LSST observations + observation properties

 

Metric Analysi Framwork (MAF)

Python API to interact with OpSims specifying science performance on a science case with a metric

Lynne Jones

Peter Yoachim

~100s simulations

~1000s MAFs

Rubin LSST survey design

Rubin LSST survey design

Rubin has involved the community to an unprecedented level in survey design this is a uniquely "democratic" process!

Rubin LSST survey design

Rubin has involved the community to an unprecedented level in survey design this is a uniquely "democratic" process!

Rubin LSST survey design

Rubin has involved the community to an unprecedented level in survey design this is a uniquely "democratic" process!

85% submissins led by SC members

 

Rubin LSST survey design

Rubin has involved the community to an unprecedented level in survey design this is a uniquely "democratic" process!

Survey Cadence Optimization Committee

Rubin LSST survey design

Rubin has involved the community to an unprecedented level in survey design this is a uniquely "democratic" process!

Survey Cadence Optimization Committee

Rubin LSST survey design

Rubin has involved the community to an unprecedented level in survey design this is a uniquely "democratic" process!

Survey Cadence Optimization Committee

Rubin LSST survey design

2017

Rubin LSST survey design

Rubin has involved the community to an unprecedented level in survey design this is a uniquely "democratic" process!

2019

Rubin LSST survey design

Rubin has involved the community to an unprecedented level in survey design this is a uniquely "democratic" process!

2023

Rubin LSST survey design

Rubin has involved the community to an unprecedented level in survey design this is a uniquely "democratic" process!

2024

Rubin LSST survey design

Rubin has involved the community to an unprecedented level in survey design this is a uniquely "democratic" process!

2024

2024

LSST ToO program

The main Rubin assets are the 10 sqdeg FoV + rapid slew + depth

PSTN-055 (2022): The SCOC recommends a ToO program be enabled to respond to Gravitational Waves and MMA triggers with a fraction of ≤ 3% of dedicated survey time, with the possibility of extending it to additional types of targets in the future.

federica bianco - fbianco@udel.edu

Rubin ToO program

 

+80 authors!

Rubin ToO program

 

https://arxiv.org/pdf/2411.04793

Rubin ToO program

 

What about fast transients in the main survey?

cepheid

The minutes-second-subsecond Universe

 

  • GRB afterglow
  • Off-axis GRB
  • Kilonovae
  • SN-GRB breakout
  • SN-GRB pop studies
  • Fast Radio Bursts
  • relativistic TDEs
  • X-ray binary stars
  • cataclysmic variable stars
  • blazars
  • stellar flares
  • solar system occultations
  • technosignatures

Željko Ivezić et al 2019 ApJ 873 111

LSST: From Science Drivers to Reference Design and Anticipated Data Products

Marshall et al. 2017

Mortersen et al. 2019

Fast transients in LSST Wide Fast Deep

 

Smith +2019

Ragosta+ et al. 2023

Rubin LSST survey design

 

~800 per field

10 seasons, with each 6 months

2 visits per night (within ~30 min for Solar System Science)

revisit time => 4.5 nights

 

This will scatter significantly (weather, moon, ...)

 

The original survey plan didn't lead to good time domain astronomy (TDA) outcomes:

2 intranight obs in same filer +

2 intranight obs in another filter ~5 day later

Rubin LSST survey design

 

GRB =>

~800 per field

10 seasons, with each 6 months

2 visits per night (within ~30 min for Solar System Science)

revisit time => 4.5 nights

 

This will scatter significantly (weather, moon, ...)

 

The original survey plan didn't lead to good time domain astronomy (TDA) outcomes:

2 intranight obs in same filer +

2 intranight obs in another filter ~5 day later

Introducing Rolling Cadence

Current plan: rolling 8 out of the 10 years

Rubin LSST survey design up to 2018

 

# pairs of observations (1e5)

time gaps (days)

based on 2017 LSST simulations

2017 simulations: between 3 and 32 KN can be identified (~300 detected)

Text

Proposed 3 intranight obs

2 within 1 hour in different filters

1 at 4-8 hours separation w repeat filter

Intorducing Triplets

 

 

Intranight color (near instantaneous)

Intranight rate of change (~hour time scales)

Current plan: rolling 8 out of the 10 years

 Presto-Color, Bianco+ 2019

Proposed 3 intranight obs

2 within 1 hour in different filters

1 at 4-8 hours separation w repeat filter

Intorducing Triplets

 

Text

Intranight color (near instantaneous)

Intranight rate of change (~hour time scales)

Ofek+ 2024

Proposed 3 intranight obs

2 within 1 hour in different filters

1 at 4-8 hours separation w repeat filter

Intorducing Triplets

Current plan: 4% of the survey is currently conducted in triplets

Intranight color (near instantaneous)

Intranight rate of change (~hour time scales)

Ofek+ 2024

newer simulations ->

<-bad         good ->

2023 simulations: 62% improvement

newer simulations ->

4 – 24 hour gaps between epochs will enable kilonova parameter estimation

 

Kilonovae in LSST Wide Fast Deep

Andreoni+ 2022a

2023 simulations: 62% improvement

newer simulations ->

4 – 24 hour gaps between epochs will enable kilonova parameter estimation

 

Kilonovae in LSST Wide Fast Deep

Andreoni+ 2022a

GRB =>

2023 simulations: 62% improvement

newer simulations ->

4 – 24 hour gaps between epochs will enable kilonova parameter estimation

 

Kilonovae in LSST Wide Fast Deep

Andreoni+ 2022a

GRB =>

https://pstn-056.lsst.io/

Proposed reduction to 6 rolling years (3 2-year cycles) to improve intrasurvey uniformity

https://pstn-056.lsst.io/

8y rolling

no rolling

 6y rolling

~7% loss in KN characterization

Proposed reduction to 6 rolling years (3 2-year cycles) to improve intrasurvey uniformity

Shar Daniels

NSF Graduate Student Fellow

University of Delaware

TVS Science Collaboration

 Fast Transient Subgroup

join TVS! no fees no minimum req

Chair: Igor Andreoni

Shar Daniels

NSF Graduate Student Fellow

University of Delaware

TVS Science Collaboration

 Fast Transient Subgroup

Fast Transients MetricsCOordinations and White Paper

Let's work together!

join the overleaf paper

This is urgent! results must come through in the next ~ 2 months or the strategy may be set for year 1

SCSLSA Bianco - Serbia

By federica bianco

SCSLSA Bianco - Serbia

Rubin LSST building a legacy

  • 181