Upgrade to Pro — share decks privately, control downloads, hide ads and more …

LOFAR Image Plane Transient Candidate #3

Ab44292d7d6f032baf342a98230a6654?s=47 transientskp
January 08, 2014

LOFAR Image Plane Transient Candidate #3

Adam Stewart

Ab44292d7d6f032baf342a98230a6654?s=128

transientskp

January 08, 2014
Tweet

Transcript

  1. Image Plane Transient Candidate #3 Found in the NCP Field

    by the TraP Adam Stewart, Tom Hassall, Rob Fender, Jess Broderick, Gosia Pietka LOFAR TKP Meeting - Amsterdam - 8-10 January 2014
  2. Brief Reminder of NCP Data • Recorded simultaneously, using a

    single sub band, with the initial MSSS-LBA observing run in 2011-2012 • Each observation: • Is 11 minutes long • At 60 MHz • 200 kHz of bandwidth • When in sequence observations are 4 minutes apart. • 175 deg2 searchable area. • MSSS calibrator is used to process the data. msss.astron.nl
  3. NCP Transient Simulations

  4. Transient Simulations • Wanted to test how the NCP field

    would react to a transient. • As we know, LOFAR reduction is quite dependent on the sky model used. • Nevertheless I wanted to confirm that a bright transient, not in the model, would appear or at least leave hints. • Eg. removing 3C 61.1 from the sky model, while causing bad artifacts, still leaves the source visible.
  5. Transient Simulations • Wanted to test how the NCP field

    would react to a transient. • As we know, LOFAR reduction is quite dependent on the sky model used. • Nevertheless I wanted to confirm that a bright transient, not in the model, would appear or at least leave hints. • Eg. removing 3C 61.1 from the sky model, while causing bad artifacts, still leaves the source visible.
  6. Transient Simulations • Wanted to test how the NCP field

    would react to a transient. • As we know, LOFAR reduction is quite dependent on the sky model used. • Nevertheless I wanted to confirm that a bright transient, not in the model, would appear or at least leave hints. • Eg. removing 3C 61.1 from the sky model, while causing bad artifacts, still leaves the source visible. 3C 61.1 removed from model
  7. Transient Simulations • Wanted to test how the NCP field

    would react to a transient. • As we know, LOFAR reduction is quite dependent on the sky model used. • Nevertheless I wanted to confirm that a bright transient, not in the model, would appear or at least leave hints. • Eg. removing 3C 61.1 from the sky model, while causing bad artifacts, still leaves the source visible. 3C 61.1 removed from model
  8. Inserting a Transient • Aim was to test whether a

    transient would be seen with the reduction method. • Any brightness of transient. • Transient inserted into pre-processed data before being reduced through MSSS pipeline. • Using calibrator gain solutions to insert - ideally want field phase solutions but proved difficult to merge and use table. Calibrate calibrator Transfer calibrator gain solutions to NCP Phase-only calibration using VLSS based model Image using AWimager Assuming transient survives demixing, flagging etc
  9. Inserting a Transient • Aim was to test whether a

    transient would be seen with the reduction method. • Any brightness of transient. • Transient inserted into pre-processed data before being reduced through MSSS pipeline. • Using calibrator gain solutions to insert - ideally want field phase solutions but proved difficult to merge and use table. Calibrate calibrator Transfer calibrator gain solutions to NCP Phase-only calibration using VLSS based model Image using AWimager Pre-processed Data Solutions + Transient BBS Assuming transient survives demixing, flagging etc
  10. Inserting a Transient 15 Jy 30 Jy It’s clear that

    at least something odd is in the dataset, and becomes clearer when brighter.
  11. Inserting a Transient 15 Jy 30 Jy It’s clear that

    at least something odd is in the dataset, and becomes clearer when brighter.
  12. Works Very Nicely in RSM 10 Jy transient source

  13. Works Very Nicely in RSM 10 Jy transient source

  14. Inserting a Very Bright Transient 150 Jy •Flux starts to

    be absorbed by 3C 61 •Now ~170 Jy •Image is poor with other sources missing.
  15. Ghosts... • A mirrored source appears when the transient was

    bright 50 - 80 Jy • Opposite 3C 61.1 - the brightest source in the field. • Ghost is brighter, roughly 60 - 20 Jy (in 80 Jy case) • Reminded me of something
  16. Ghosts... • A mirrored source appears when the transient was

    bright 50 - 80 Jy • Opposite 3C 61.1 - the brightest source in the field. • Ghost is brighter, roughly 60 - 20 Jy (in 80 Jy case) • Reminded me of something
  17. None
  18. Placing Transient in the Model At Ghost Location At True

    Location Ghost disappears when the true source is in the sky model
  19. Placing Transient in the Model At Ghost Location At True

    Location Ghost disappears when the true source is in the sky model
  20. Ghosts Not Limited to One Location • At first we

    were worried that there was a special distance such as Bell #1 that would scale with frequency. • Sampling the whole field reveals ghosts can be created in various locations. • In the NCP cases tested it seems to be concentrated to the right-hand side of 3C 61.1
  21. Ghosts Not Limited to One Location • At first we

    were worried that there was a special distance such as Bell #1 that would scale with frequency. • Sampling the whole field reveals ghosts can be created in various locations. • In the NCP cases tested it seems to be concentrated to the right-hand side of 3C 61.1
  22. Ghosts Not Limited to One Location • At first we

    were worried that there was a special distance such as Bell #1 that would scale with frequency. • Sampling the whole field reveals ghosts can be created in various locations. • In the NCP cases tested it seems to be concentrated to the right-hand side of 3C 61.1
  23. The Transient

  24. • An object only seen once and never again. •

    Snapshot taken on December 24th 2011 at 04:33. TraP Discovered a Similar Event
  25. TraP Discovered a Similar Event

  26. TraP Discovered a Similar Event

  27. TraP Discovered a Similar Event

  28. TraP Discovered a Similar Event

  29. TraP Discovered a Similar Event TraP extracted this source

  30. TraP Discovered a Similar Event 7 Jy Source? Ghost? 13

    Jy
  31. Can the Ghost be ? • With what was seen

    in the simulations the next step was to process the data again with suspected sources in the sky model. • Try a 20 Jy (13+7) point source at each location one at a time. 7 Jy Source Ghost? 13 Jy
  32. Transient in the Model 20 Jy at ghost Location Both

    sources still visible 20 Jy at Right/Source Location Ghost Source disappears! We came, we saw...
  33. Transient in Other Surrounding Snapshots Model • Process the surrounding

    4 observations with the transient in the sky model. • The source appears strongly only in the original snapshot where it was discovered.
  34. Attempts to Kill the Transient • Tried numerous methods to

    remove or at least greatly effect the transient. Subtract 3C 61.1 Different weighting scheme (here natural) • Also tried imaging using different baseline selections. • Different time compression before processing. • Checked other observations at the same LST - no hint of source. • It survived all these tests where somewhat similar candidates failed. • Though this is the only candidate which has a clear ghost source.
  35. There are similar objects found... • 8 more to be

    precise at similar fluxes of the initial source found 4 - 8 Jy 1 Jan 2012 2 Jan 2012 21 Jan 2012 11 Feb 2012 11 Feb 2012 10 Mar 2012 31 Mar 2012
  36. There are similar objects found... • 8 more to be

    precise at similar fluxes of the initial source found 4 - 8 Jy 1 Jan 2012 2 Jan 2012 21 Jan 2012 11 Feb 2012 11 Feb 2012 10 Mar 2012 31 Mar 2012 Need Testing! NCP Deep Image can also help
  37. There are similar objects found... • 8 more to be

    precise at similar fluxes of the initial source found 4 - 8 Jy 1 Jan 2012 2 Jan 2012 21 Jan 2012 11 Feb 2012 11 Feb 2012 10 Mar 2012 31 Mar 2012 Need Testing! NCP Deep Image can also help
  38. If real - what do we know? • Duration of

    11 mins. - When dataset split in half or thirds, the source is still present in each half/third. - Flux also relatively constant. - Processing effect? To test... Transient set split in half (in model) • Also not present in snapshots before or after. Combing next two also nothing. • Bright at ~ 25 +/- 5 Jy • Would suggest a rate of 1 / 2537 day-1 deg-2 with ∆t = 11 mins, 4 Jy limit (10σ TraP selection) • Not in EoR deep NCP map. • Localisation to ~120” (10 km baselines) - Will take a bit of care to incorporate longer baselines
  39. Optical Follow Up • r-band, slightly deeper than 21 magnitude.

    • With current localisation many sources are seen in the error box. • None show obvious variability on the timescales of minutes and one month. • Ongoing. Teo Muñoz-Darias
  40. What can we infer? (Reminder First) • Prospects of finding

    FRBs in image plane data detailed by Hassall, Keane & Fender 2013: How scattering effects the pulse Image plane can outperform beamformed in high scattering scenarios Low scattering the Beamformed > image plane FRBs/hour LBA
  41. • With 200 kHz of bandwidth - 11 minutes is

    due to scattering or intrinsic pulse width. - Because a DM of 6000 (very high! z~6) would mean a dispersive delay of 45s between 60 - 60.2 MHz. - Hence much more likely to be scattering or pulse width. What can we infer? T.Hassall
  42. • With 200 kHz of bandwidth - 11 minutes is

    due to scattering or intrinsic pulse width. - Because a DM of 6000 (very high! z~6) would mean a dispersive delay of 45s between 60 - 60.2 MHz. - Hence much more likely to be scattering or pulse width. • So lets take the upper bound flux of the first split 5 min image, 30 Jy, and the lower flux of the second 5 minute image, 20 Jy, and calculate the minimum DM. What can we infer? T.Hassall
  43. • With 200 kHz of bandwidth - 11 minutes is

    due to scattering or intrinsic pulse width. - Because a DM of 6000 (very high! z~6) would mean a dispersive delay of 45s between 60 - 60.2 MHz. - Hence much more likely to be scattering or pulse width. • So lets take the upper bound flux of the first split 5 min image, 30 Jy, and the lower flux of the second 5 minute image, 20 Jy, and calculate the minimum DM. • Using S = e(-t/tscatt) and the Bhat et al law: - t_scatt ~ 1600 hence a DM ~ 450 - Reasonable value if considering DMs of FRBs, which could last minutes at low frequencies. What can we infer? T.Hassall
  44. • Can attempt to say something about spectral index: -

    Take FRBs at 1.4 GHz, the NCP will be brighter by (1.4 / 0.06)α - FRB 10 Jy x 5 ms = 0.01 Jy/s - NCP 25 Jy x 660s = 13200 Jy/s - a 106 ratio - But could have a spectral broadening term, β - (1.4 / 0.06)α+β ~ 10-6 - Hence, α+β = -4 which is steep, but we should keep β in mind. What can we infer? T.Hassall
  45. Conclusions • Altering the model makes a definite impact to

    the data, removing the friend. • I cannot just create a source by inserting it into the sky model. • Doesn’t appear in other snapshots even when processing with it in model. • Follows transient simulation results. • Survives various tests thrown at it, when I have seen others not. For Against • Only seen in one snapshot. • If exponential decay then should be around ~10 Jy in next snapshot, cannot see anything. • No variation when splitting the dataset to image. • General uncertainty surrounding this LBA MSSS data - the data as a whole was never fully used for quality issues. • Do we understand artefacts in these short, limited UV coverage snapshots? A 25 Jy Transient at 60 MHz, ‘on’ for 11 minutes or less - Is it real?
  46. Conclusions • Lack of other data (more bw) is a

    blessing and a curse. • Perhaps will never be 100% sure because of this, but we have to take what we see. • Bell #1 situation of not being sure - Sky model tests on this data do not effect the transient. • Other candidates to check, may help in ways to prove/ disprove. • But this transient is tough to kill, it is yet to ever disappear. • RSM should be able to provide some answers...
  47. Finally thanks to... • The TraP team of John, Antonia,

    Gijs, Bart and Tim et al • Previously data had been combined to gather decent images for humans to work with. • The TraP in it’s current state made searching these images possible.
  48. DM Calculation • The flux of the pulse as a

    function of time is: • S = exp(-t/t_scatt) • so 30/20 = exp(-660/t_scatt) • t_scatt = -660/ln(20/30) ~ 1600 • This corresponds to a DM~450 from the Bhat et al. law. Current DMs of FRBs are in the range 550-1100, so this seems very reasonable.
  49. PSF Example

  50. UV Coverage Example For transient observation on 24th Dec