1 of 70

FLASHy galaxies in cosmological simulations

PHISCC 2025

Emily Kerrison

with Samuel Ward & Stephanie Tonnesen

University of Sydney/CSIRO Space & Astronomy

CCA Predoctoral Fellow (Spring 2025)

Credit: NOIRLab/NSF/AURA/M.

2 of 70

First results 2: FLASH detections dominated by young, compact radio galaxies

The ASKAP-FLASH Survey @ PHISCC 2025 | Elizabeth Mahony + Hyein Yoon

Image: Emily Kerrison

z=0.55

z=0.73

z=0.58

z=0.81

z=0.67

Redshift

  • Fit using RadioSED (Kerrison et al. 2024) 
  • At least 65% of detections well characterised by a peaked spectrum model → young radio AGN (O’Dea & Saikia 2021)
  • FLASH is detecting radio galaxies at a specific evolutionary phase

OFFICIAL

OFFICIAL

3 of 70

First results 2: FLASH detections dominated by young, compact radio galaxies

The ASKAP-FLASH Survey @ PHISCC 2025 | Elizabeth Mahony + Hyein Yoon

Image: Emily Kerrison

z=0.55

z=0.73

z=0.58

z=0.81

z=0.67

Redshift

  • Fit using RadioSED (Kerrison et al. 2024) 
  • At least 65% of detections well characterised by a peaked spectrum model → young radio AGN (O’Dea & Saikia 2021)
  • FLASH is detecting radio galaxies at a specific evolutionary phase

OFFICIAL

OFFICIAL

4 of 70

PS sources with RadioSED (Kerrison+2024)

4

Data at low and high frequencies to capture a (sometimes) complex Spectral Energy Distribution (SED)

5 of 70

PS sources with RadioSED (Kerrison+2024)

Data at low and high frequencies to capture a (sometimes) complex Spectral Energy Distribution (SED)

5

RadioSED provides automated SEDs using 23 wide area radio surveys (76MHz - 200GHz)

Bayesian model inference and comparison to identify different SED classes

~300,000 sources across the southern sky!

6 of 70

PS sources with RadioSED (Kerrison+2024)

Thanks to well documented sensitivities (and Bayesian statistics) – we can incorporate upper limits to discover more PS sources!

6

7 of 70

Stripe 82 – a test case (Kerrison+ submitted)

Stripe 82 with RadioSED

~300 deg2 equatorial field (20h-4h)

2,760 fittable radio sources (S888MHz ≥ 10mJy)

7

Radio sources (2,760)

8 of 70

Stripe 82 – a test case (Kerrison+ submitted)

Stripe 82 with RadioSED

~300 deg2 equatorial field (20h-4h)

2,760 fittable radio sources (S888MHz ≥ 10mJy)

8

Radio sources (2,760)

Known PS sources (12)

9 of 70

Stripe 82 – a test case (Kerrison+ submitted)

Stripe 82 with RadioSED

~300 deg2 equatorial field (20h-4h)

2,760 fittable radio sources (S888MHz ≥ 10mJy)

9

Radio sources (2,760)

Known PS sources (12)

New PS sources (359)

10 of 70

Stripe 82 – a test case (Kerrison+ submitted)

Stripe 82 with RadioSED

~300 deg2 equatorial field (20h-4h)

2,760 fittable radio sources (S888MHz ≥ 10mJy)

10

Radio sources (2,760)

Known PS sources (12)

New PS sources (359)

~13% of an untargeted field are PS

11 of 70

Stripe 82 – a test case (Kerrison+ submitted)

Stripe 82 with RadioSED

~300 deg2 equatorial field (20h-4h)

2,760 fittable radio sources (S888MHz ≥ 10mJy)

11

Radio sources (2,760)

Known PS sources (12)

New PS sources (359)

~13% of an untargeted field are PS

What about FLASH?

12 of 70

A specific population

12

Frequency

Flux

Frequency

Flux

Compact

Breakout

13 of 70

A specific population

21/30 FLASH detections are towards compact radio jets (≲ 1kpc) (from broadband SED shape)

70% of radio galaxies are extended on these scales (O’Dea 1998, Callingham+17, Kerrison+subm.)

13

Frequency

Flux

Frequency

Flux

Compact

Breakout

14 of 70

A specific population

21/30 FLASH detections are towards compact radio jets (≲ 1kpc) (from broadband SED shape)

70% of radio galaxies are extended on these scales (O’Dea 1998, Callingham+17, Kerrison+subm.)

14

Frequency

Flux

Frequency

Flux

Compact

Breakout

Why are most detections in an untargeted survey towards compact sources?

jet-ISM interactions (Morganti+23) OR selection effects

15 of 70

Why are most detections in an untargeted survey towards compact sources?

jet-ISM interactions (Morganti+23) OR selection effects

16 of 70

A cool view of radio AGN - simulations

We know exactly:

  • Jet morphology
  • Radio luminosity
  • Gas mass
  • Gas distribution

Why are most detections in an untargeted survey towards compact sources?

jet-ISM interactions (Morganti+23) OR selection effects

17 of 70

A cool view of radio AGN - simulations

We know exactly:

  • Jet morphology
  • Radio luminosity
  • Gas mass
  • Gas distribution

Why are most detections in an untargeted survey towards compact sources?

jet-ISM interactions (Morganti+23) OR selection effects

So we can find out what matters amongst:

Galaxy orientation

Radio luminosity

Jet morphology

18 of 70

The project: A cool view of radio AGN in cosmological simulations

  1. Making simulated galaxies radio galaxies
    1. Cosmological simulations are the right tool (SIMBA)
    2. Finding the right galaxies
    3. ‘Observing’ cosmological simulations

  • Finding out what matters

Galaxy orientation

Radio luminosity

Jet morphology

19 of 70

The right tool for the job

Cosmological simulations

  • AGN-galaxy and galaxy-galaxy interactions
  • Large volume → statistical samples

CAMELS collaboration, Thomas+21

20 of 70

The right tool for the job

Cosmological simulations

  • AGN-galaxy and galaxy-galaxy interactions
  • Large volume → statistical samples

SIMBA (Dave+19)

  • Cold ISM is unresolved
  • BUT HI calculated on-the-fly and per-particle (Krumholz & Gnedin 2011)
  • Thomas+21 prescriptions for radio luminosity from SMBH properties

CAMELS collaboration, Thomas+21

– Observations

+ Simulations (low jet power)

– Observations

+ Simulations (high jet power)

21 of 70

Finding the right galaxies

Observationally-motivated snapshot choice (z=0.4)

HI-rich galaxies → focus on gas distribution rather than presence/absence

22 of 70

Finding the right galaxies

Observationally-motivated snapshot choice (z=0.4)

HI-rich galaxies → focus on gas distribution rather than presence/absence

AGN jet hosts (Thomas+21)

For experts:

Ṁ > 0

0 < fEdd < 0.02

23 of 70

The range of galaxies

100 kpc

More disky

More complex

24 of 70

The range of galaxies

100 kpc

More disky

More complex

Plus 250 more!

25 of 70

“Observing” SIMBA radio galaxies I - radio flux

We add to SIMBA:

26 of 70

“Observing” SIMBA radio galaxies I - radio flux

We add to SIMBA:

Static jets

27 of 70

“Observing” SIMBA radio galaxies I - radio flux

We add to SIMBA:

Static jets

Discretised flux

28 of 70

“Observing” SIMBA radio galaxies I - radio flux

We add to SIMBA:

Static jets

Discretised flux

3D ray tracing through gas particles

29 of 70

“Observing” SIMBA radio galaxies I - radio flux

With physical parameters:

  • Injected radio flux drawn from local radio luminosity function (Mauch & Sadler 07)
  • Absorption due to neutral hydrogen only :

We add to SIMBA:

Static jets

Discretised flux

3D ray tracing through gas particles

30 of 70

“Observing” SIMBA radio galaxies I - radio flux

With physical parameters:

  • Injected radio flux drawn from local radio luminosity function (Mauch & Sadler 07)
  • Absorption due to neutral hydrogen only :

We add to SIMBA:

Static jets

Discretised flux

3D ray tracing through gas particles

Radiative transfer!

31 of 70

“Observing” SIMBA radio galaxies I - radio flux

With physical parameters:

  • Injected radio flux drawn from local radio luminosity function (Mauch & Sadler 07)
  • Absorption due to neutral hydrogen only :

FR0

(confined)

FRI

FRII

(breakout)

We add to SIMBA:

Static jets

Discretised flux

3D ray tracing through gas particles

Radiative transfer!

32 of 70

“Observing” SIMBA radio galaxies I - radio flux

With physical parameters:

  • Injected radio flux drawn from local radio luminosity function (Mauch & Sadler 07)
  • Absorption due to neutral hydrogen only :

FR0

(confined)

FRI

FRII

(breakout)

We add to SIMBA:

Static jets

Discretised flux

3D ray tracing through gas particles

Radiative transfer!

33 of 70

“Observing” SIMBA radio galaxies I - radio flux

With physical parameters:

  • Injected radio flux drawn from local radio luminosity function (Mauch & Sadler 07)
  • Absorption due to neutral hydrogen only :

FR0

(confined)

FRI

FRII

(breakout)

We add to SIMBA:

Static jets

Discretised flux

3D ray tracing through gas particles

Why are most detections in an untargeted survey towards compact sources?

jet-ISM interactions (Morganti+23) OR selection effects

Radiative transfer!

34 of 70

“Observing” SIMBA radio galaxies II - instrument effects

Simulated Absorption of Neutral Gas for Radio Astronomy

(SANGRiA - Ward, Kerrison & Tonnesen in prep.)

Jet

FR0

FRII

35 of 70

“Observing” SIMBA radio galaxies II - instrument effects

Simulated Absorption of Neutral Gas for Radio Astronomy

(SANGRiA - Ward, Kerrison & Tonnesen in prep.)

Jet

Pixelate

FR0

FRII

36 of 70

“Observing” SIMBA radio galaxies II - instrument effects

Simulated Absorption of Neutral Gas for Radio Astronomy

(SANGRiA - Ward, Kerrison & Tonnesen in prep.)

Jet

Pixelate

Convolve

FR0

FRII

37 of 70

“Observing” SIMBA radio galaxies II - instrument effects

Simulated Absorption of Neutral Gas for Radio Astronomy

(SANGRiA - Ward, Kerrison & Tonnesen in prep.)

Jet

Pixelate

Convolve

FR0

FRII

38 of 70

“Observing” SIMBA radio galaxies II - instrument effects

Simulated Absorption of Neutral Gas for Radio Astronomy

(SANGRiA - Ward, Kerrison & Tonnesen in prep.)

Simulated!

FLASH

39 of 70

“Observing” SIMBA radio galaxies II - instrument effects

Simulated Absorption of Neutral Gas for Radio Astronomy

(SANGRiA - Ward, Kerrison & Tonnesen in prep.)

Simulated!

FLASH

Using the same Bayesian FLASHfinder to identify detections (Allison+2012)

where do we see HI?

40 of 70

Where would we see FLASH detections?

41 of 70

Galaxy orientation matters

Compact radio source (FR0) → significant changes to line profile from galaxy

orientation alone

42 of 70

Galaxy orientation matters

Compact radio source (FR0) → significant changes to line profile from galaxy

orientation alone

43 of 70

Galaxy orientation matters

Compact radio source (FR0) → significant changes to line profile from galaxy

orientation alone

Edge-on

deepest line (highest NHI)

Face-on

shallow line, blueshifted outflow (AGN feedback!)

44 of 70

Radio luminosity also matters

0o

60o

90o

45 of 70

Radio luminosity also matters

0o

60o

90o

L ~ 5 x 1026 W Hz-1

L ~ 325 W Hz-1

L ~ 3 x 1026 W Hz-1

46 of 70

Radio luminosity also matters … across one galaxy

47 of 70

Radio luminosity also matters … across one galaxy

48 of 70

100 kpc

More disky

More complex

Plus 250 more!

49 of 70

Radio luminosity also matters … across all galaxies

1e20

LAGN [W Hz-1]

NHI [cm-2]

Detection fraction

1e21

1e22

1e23

1e24

l

l

l

l

l

1e24

1e25

1e26

1e27

l

l

l

l

50 of 70

Radio luminosity also matters … across all galaxies

1e20

LAGN [W Hz-1]

NHI [cm-2]

Detection fraction

1e21

1e22

1e23

1e24

l

l

l

l

l

1e24

1e25

1e26

1e27

l

l

l

l

It’s easy to see HI with:

high NHI

high LAGN

51 of 70

Does radio morphology matter?

1e20

LAGN [W Hz-1]

NHI [cm-2]

Detection fraction

1e21

1e22

1e23

1e24

l

l

l

l

l

1e24

1e25

1e26

1e27

l

l

l

l

FR0 / confined

0.5 kpc

FRII / breakout

10 kpc

52 of 70

Does radio morphology matter?

1e20

LAGN [W Hz-1]

NHI [cm-2]

Detection fraction

1e21

1e22

1e23

1e24

l

l

l

l

l

1e24

1e25

1e26

1e27

l

l

l

l

FR0 / confined

0.5 kpc

FRII / breakout

10 kpc

53 of 70

Does radio morphology matter?

1e20

LAGN [W Hz-1]

NHI [cm-2]

Detection fraction

1e21

1e22

1e23

1e24

l

l

l

l

l

1e24

1e25

1e26

1e27

l

l

l

l

FR0 / confined

0.5 kpc

FRII / breakout

10 kpc

54 of 70

Does radio morphology matter?

1e20

LAGN [W Hz-1]

NHI [cm-2]

Detection fraction

1e21

1e22

1e23

1e24

l

l

l

l

l

1e24

1e25

1e26

1e27

l

l

l

l

FR0 / confined

0.5 kpc

FRII / breakout

10 kpc

Not detected!

55 of 70

Conclusions & Future work

FLASH is more likely to detect HI when:

  • Galaxies are closer to edge-on → higher NHI
  • Radio jets are brighter → higher S/N
  • Jets are compact → favoured geometry for backlighting HI

56 of 70

Conclusions & Future work

FLASH is more likely to detect HI when:

  • Galaxies are closer to edge-on → higher NHI
  • Radio jets are brighter → higher S/N
  • Jets are compact → favoured geometry for backlighting HI

Future work:

  • Importance resampling luminosities to match observed distribution(s)
  • Is FLASH seeing selection effects, or differences in the gas?
  • A FLASH-like lightcone for a full survey pointing

57 of 70

Conclusions & Future work

FLASH is more likely to detect HI when:

  • Galaxies are closer to edge-on → higher NHI
  • Radio jets are brighter → higher S/N
  • Jets are compact → favoured geometry for backlighting HI

Future work:

  • Importance resampling luminosities to match observed distribution(s)
  • Is FLASH seeing selection effects, or differences in the gas?
  • A FLASH-like lightcone for a full survey pointing

Questions? Get in touch emily.kerrison@sydney.edu.au

58 of 70

Conclusions & Future work

FLASH is more likely to detect HI when:

  • Galaxies are closer to edge-on → higher NHI
  • Radio jets are brighter → higher S/N
  • Jets are compact → favoured geometry for backlighting HI

Future work:

  • Importance resampling luminosities to match observed distribution(s)
  • Is FLASH seeing selection effects, or differences in the gas?
  • A FLASH-like lightcone for a full survey pointing

Questions? Get in touch emily.kerrison@sydney.edu.au

… or hire me! :)

59 of 70

More info: Mock observations

ASKAP beam : 30’’

pixel scale : 5’’

Noise: 5mJy/beam after conv.

Spectra : brightest pixel

(SIMBA spatial res. ~1kpc

<< FLASH angular res.)

Line identification: Bayesian FLASHfinder (Allison+12) using Multinest

lnB >= 30 for secure detections (FLASH)

60 of 70

More info: luminosity distributions

But FLASH detections are intrinsically very luminous

We can apply the Thomas+21 prescription to SIMBA galaxies to determine radio luminosities

And it is well matched to observations of the general radio population

61 of 70

What about realistic samples?

From our uniform grid

62 of 70

What about realistic samples?

From our uniform grid

With a suitable target distribution

63 of 70

What about realistic samples?

From our uniform grid

With a suitable target distribution

We can obtain a realistic distribution with importance resampling

64 of 70

What about realistic samples?

From uniform inputs

65 of 70

What about realistic samples?

From uniform inputs

To matched outputs!

66 of 70

Does radio morphology matter?

LAGN [W Hz-1]

NHI [cm-2]

FR0 / confined

0.5 kpc

FRII / breakout

10 kpc

67 of 70

Does radio morphology matter?

LAGN [W Hz-1]

NHI [cm-2]

FR0 / confined

0.5 kpc

FRII / breakout

10 kpc

68 of 70

Expert easter egg: confined vs. breakout statistics

Number of observations does suggest the difference is significant!

FR0 / confined

FRII / breakout

69 of 70

The ASKAP-FLASH survey

First Large Absorption Survey in HI (Allison+2022):

Image credit: Hyein Yoon

30,000 deg2, Dec < +15o

(excluding galactic plane)

70 of 70

The ASKAP-FLASH survey

First Large Absorption Survey in HI (Allison+2022):

~20’’ spatial, ~10km/s spectral resolution

0.4 < z < 1 HI absorption

in sources > 40mJy

Image credit: Hyein Yoon

20’’

30,000 deg2, Dec < +15o

(excluding galactic plane)