Wide-field, high sensitivity VLBI

1
Wide-field, high sensitivity VLBI

• surveying with mas resolution

• Adam Deller

• NRAO Postdoc Symposium
• April 2009

2
High resolution interferometry

• Traditionally, narrow fields for studying single
  compact objects (pulsars, AGN, masers)
• Astrometry is the current killer app
• Today something outside the square

A typical VLBI image
Very little actual work to report!
3
First a catalogue of surveys

• A number of incredibly useful surveys exist at
  radio wavelengths
• NVSS (1.4 GHz, 45 resolution)
• FIRST (1.4 GHz, 5 resolution)
• VLSS, WENSS, SUMSS, PMN, GB6
• And of course at non-radio wavelengths IRAS,
  2MASS, SDSS, ROSAT
• Future EVLA, ALMA, LSST, Pan-STARRS

3
4
The need for VLBI surveys

• Our knowledge of the high-resolution radio sky is
  poor! Previous surveys have typical sample
  sizes 100 largest (15 years VCS geodesy data)
  has 4000 bright sources
• SKA science driver Galaxy evolution, cosmology
  and dark energy looks to trace star formation
  over the history of the universe high resolution
  is vital

4
5
The need for VLBI surveys

• Limited surveys have been planned but not
  undertaken, usually involving preselection for
  expected detectability
• DEVOS (Deep Extragalactic VLBI detection Of SDSS
  quasars, Frey et al.) - preselected 9000 quasars
  from SDSS/FIRST (85 det.)
• VCS is also an example of this
• Of course, this biases any results

5
6
Why (until now) no real surveys?

• The problem is the twin demons of high resolution
  and low sensitivity plus data volume (later)
• VLBI sensitivity has traditionally lagged
  connected-element interferometers the data
  storage problem
• And most sources are already resolved out
• Hence source density is low - lucky to get one
  source in a pointing!!

6
7
The imminent changes for VLBI

• Data rates have been stagnant for the last 10
  years - time for an explosion
• This will mean gt10x increase in sensitivity over
  the next 10 years (4x in the next year or two)
• New, more flexible (software) correlator
  architecture will get away from single, narrow
  fields and allow the kind of surveying Im about
  to spruik

7
8
What the VLBA can (will) do
Data rate (Gbps) Snapshot sensitivity (?Jy/beam) 8hr track sensitivity (?Jy/beam)
Now (0.25) 380 40
Soon (4) 95 10
Future (32) 34 4

• 10 calibration o/head is only realistic for high
  sensitivity (every scan has in-beam)

3x 2 min tracks. Both sensitivities assume 10
calibration o/head
8
9
What the VLBA can (will) do

• Much improved uv coverage yields much improved
  imaging
• Thus getting close to the thermal noise
  prediction should be possible even for snapshots
• So, bear in mind that 6 minutes will be able to
  get you to down to 100 ?Jy

9
10
Imaging the full primary beam

• From hereon all calculations are for VLBA L band
  (1300-gt1800 MHz) at 4 Gbps
• For the 8000 km VLBA baselines, very fine
  time/frequency resolution is required to prevent
  smearing for wide fields
• http//www.aoc.nrao.edu/adeller/software/lba/
• http//astronomy.swin.edu.au/elenc/Calculators/

10
11
Imaging the full primary beam

• Reference choice 10 total smearing at beam edge
  (1800 MHz, 25m 27 FWHM)
• Need 4 kHz channels and 40 ms dump
• For an 8 hour track, 144 TB of baseband data is
  only reduced to 60 TB!!!
• And then, synthesized beam is
  5 mas, primary beam 27 need a
  130 Gpixel image (99.9999999 noise!)

sources
130 Gpixel !
primary beam
11
12
Imaging the full primary beam

• Thus the other reason VLBI surveys arent done
  - the datasets would be outrageous!!
• Could shift and average to known sources
  post-correlation, but consider the I/O on 60 TB
  This approach is impractical
• Is there another way?

12
13
A large N, small F alternative

• As already alluded, desired field around each
  target will be small - if the correlator
  uv-shifted to each target online, much more
  averaging could be done!
• Typical narrow-field parameters 1 MHz channels,
  4 second integrations (lt10 smearing for 3
  radius) yields just 2.5 GB per correlator
  pointing for 8 hours

13
14
A large N, small F alternative
primary beam

• So, even for 100s sources, the datasets would be
  lt 1 TB - not so scary in the EVLA era
• Plus the I/O problem is solved

• Time and frequency resolution internal
  to the correlator can be super high
  (1 kHz, 1 ms) and so the online uv-shifting is
  essentially penalty-free

uv-shifted pencil fields
14
15
Duplicating FIRST with VLBI

• FIRST is 9000 square degrees to 150 ?Jy
  sensitivity at 5 resolution (VLA B array),
  800,000 sources
• 0.2 square degrees per pointing, 45,000 pointings
  (20 sources/pointing)
• Need 2 mins to get to 150 ?Jy thermal Ive
  allowed 4 mins/pointing (original FIRST 3
  min/pointing)
• 125 days/3000 hours - several year project

15
16
Duplicating FIRST with VLBI

• 800,000 uv datasets and images 12 TB correlated
  data, 6.5 TB image data
• Expect many non-detections 30 hit rate (Porcas
  et al. 2004) still yields 240,000 VLBI images (I
  think thats optimistic)
• Hugely useful for understanding nature of a
  source in general studies
• Provides an excellent grid of reference sources
  for astrometry

16
17
Duplicating FIRST with VLBI

• Previously mentioned DEVOS study would be
  completed implicitly as part of such an
  undertaking
• But even better, the results would be unbiased -
  every FIRST source will have been checked
• Also useful for optical/radio reference frame ties

17
18
Narrower, deeper fields

• Already proposed for VLBA, using post-correlation
  shifting (Middelberg, CDFS).
• Will be made MUCH more efficient
• savings even greater than shallow surveys, more
  targets/pointing
• By going deep on fields with good multiwavelength
  data, can start SKA goal of understanding star
  formation/galaxy evolution (VLA SWIRE, COSMOS?)

18
19
Chandra Deep Field South (ATLAS)

• 4 sq. deg. using ATCA, 30 ?Jy, 800 sources, 10
  resolution
• Comparable VLBA would require 20 hours

19
20
VLA-COSMOS

• 2 sq deg. (large), 0.8 sq. deg (deep), 10 ?Jy,
  2 resolution, 3000 sources
• Could do to 10 ?Jy sensitivity with the VLBA in
  80 (30) hours

20
21
VLA SWIRE (104659)

• 0.8 sq. deg., 3 ?Jy, 1.6 resolution, 2000
  sources
• To get to same rms would take 300 VLBA hours -
  still possible

Part of the SWIRE field
21
22
Current activity

• Functionality for online phase shifting and
  correlation of multiple sources is presently
  being added to DiFX - the VLBA software
  correlator (this is a first)
• Expected to be testable form within a couple of
  months
• Widely available by the end of the year

22
23
Conclusions

• VLBI surveys are not a pipe-dream - with online
  uv-shifting, can achieve a speed-up of gt20 over
  traditional approaches
• Surveying the FIRST field to comparable
  sensitivity would take 3000 hours, and
  potentially yield 240,000 VLBI sources
• Very widely applicable results inbeam
  calibrators, reference frames, evolution

23
24
Addendum San Antonio

• Following the close of proceedings there is an
  excursion planned to San Antonio, famous for two
  things green chile cheeseburgers and birdlife
• Meet in the AOC car park at 510

25
Addendum San Antonio

• Stage 1 the famous Owl Bar for eats and drinks

26
Addendum San Antonio

• Stage 2 stay at the Owl Bar or head to Bosque
  del Apache to see some spectacular bird life