1. Tenth Summer Synthesis Imaging Workshop University of New Mexico, June 13-20, 2006 Low Frequency Interferometry Tracy Clarke (Naval Research Laboratory)

2. 2 History of Radio Astronomy: Low Frequencies Jansky  Radio astronomy was born in the 1930's with Karl Jansky's work at 20.5 MHz (14.5 m) at Bell labs  Reber continued radio astronomy work at 160 MHz (1.9 m) Reber

3. 3 History of Radio Astronomy  First radio telescopes operated at long wavelengths with low spatial resolution and very high system temperatures  Radio astronomy quickly moved to higher frequencies with better spatial resolution ( ) and lower system temperatures  ~ 1’, rms ~ 3 mJy/beam  ~ 10’, rms ~ 30 mJy/beam

4. 4 Science: Thermal vs. Synchrotron Emission Thermal Emission (Free-Free, Bremsstrahlung): Best observed at cm  GHz) Deflection of free electrons by positive ions in hot gas Depends on temperature of the gas Synchrotron Emission: Best observed at m   GHz) Relativistic electrons spiraling around magnetic field lines (high-energy astrophysics) Depends on the energy of the electrons and magnetic field strength Emission is polarized Can be either coherent or incoherent Thompson, Moran, & Swenson Synchrotron self absorption or free-free absorption Thermal: Rayleigh-Jeans Synchrotron

5. 5 Bursts From Jupiter & Extra-Solar Planets VLA SYSTEM CAN DETECT QUIESCENT EMISSION POSSIBLE TO DETECT BURST EMISSION FROM DISTANT “JUPITERS” Future instruments will resolve Jupiter and may detect extra-solar planets ➢ Jupiter's coherent cyclotron emission: complex interaction of Jupiter’s magnetosphere with Io torus VLA 74 MHz Jupiter images Bastian et al.

6. 6 Galactic Science Examples ➢ Galactic: - Galactic center black hole Sgr A* - non-thermal filaments: magnetic field orientation - transients - supernova remnant census - SNR acceleration - HII regions - diffuse nonthermal source (DNS): field strength

7. 7 Galactic Center Filaments Lang et al. (1999) ➢ Galactic Center: non-thermal filaments  Synchrotron filaments trace magnetic field lines and particle distribution.  Near the Galactic center filaments are perpendicular to the plane but the “Pelican” filament is parallel to the plane, allowing the magnetic field orientation to be further mapped

8. 8 (Hyman, et al., 2005, Nature) GCRT J1745-3009 ~10 minute bursts every 77 minutes – timescale implies coherent emission Coherent GC bursting source Transients ➢ Transients: sensitive, wide fields at low frequencies provide powerful opportunity to search for new transient sources ➢ candidate coherent emission transient recently discovered near Galactic center

9. 9 2 Color Image: Red: MSXat 8  m Blue: VLA 330 MHz Tripled (previously 17, 36 new) known SNRs in survey region! 330 MHz 8  m Brogan et al. (2006) Galactic Supernova Remnant Census ➢ Census: expect over 1000 SNR and know of ~230

10. 10 Comparison of GC 4 m and 6 cm Images Parkes 6 cm resolution 4’: Haynes et al. 1978, AuJPS, 45, 1 SNR: W28 Galactic Center HII Region: NGC 6357 SNR: Tornado HII Region: NGC 6334 VLA 4m resolution 2.1’ x 1.2’ A+B+C+D config. data Inner Galaxy HII Region; , Te T Gf T Gb

11. 11 Galactic Cosmic Ray 3-D Distribution  Galactic cosmic ray origin  Galactic magnetic field morphology CR energy ~ energy in starlight, gas pressure, and Galactic magnetic field Sun l b * Galactic Center Optically thick HII regions Typical T back ~5x10 4 K Typical T HII ~8x10 3 K T for + T HII T back  T for = T Gt + T obs_i – T HII  Emissivity= T for /D

12. 12 (LaRosa et al. 2005). Galactic Center: 330 MHz GBT Galactic Center: Revised Field Strength DNS VLA 74 MHz Diffuse nonthermal source (DNS) constrains the strength of the Galactic center magnetic field Energy density of GC field is ~4 orders of magnitude lower than previously thought

13. 13 SNRs: Shock Acceleration vs.Thermal Absorption Cas A A array A array + Pie Town 74/330 Spectral Index (T. Delaney – thesis with L. Rudnick)

14. 14 Pulsars Erickson 1983 Spectrum of 4C21.53: 1st msec pulsar Crab Nebula & pulsar @ 74 MHz Detecting fast (steep-spectrum) pulsars  – highly dispersed, distant PSRs – tight binaries – submsec? Probe PSR emission mechanism – explore faint end of luminosity function – spectral turnovers near 100 MHz New SNR/pulsars associations -- Deep, high surface brightness imaging of young pulsars

15. 15 Extragalactic Science Examples ➢ Extragalactic: - radio galaxy lifecycle - particle acceleration in radio galaxies - radio lobe particle content - energy feedback into the intracluster medium - particle acceleration in merger/accretions shocks - tracing Dark Matter - sample selection for Dark Energy studies - detection of high redshift radio galaxies - study of epoch of reionization

16. 16 Radio Galaxies: Outburst Lifecycle -12000 Hydra A at 4500 MHz (inset) shows an FR-I morphology on scales of <1.5' (100 kpc) ● New 74 and 330 MHz data show Hydra A is > 8' (530 kpc) in extent with large outer lobes surrounding the high frequency source Outer lobes have important implications for the radio source lifecycle and energy budget Lane et al. (2004)

17. 17 Nulsen et al. (2005) Driving Shocks into the ICM ● Chandra X-ray emission detects shock front surrounding low frequency radio contours ● Expanding radio lobes drive the shock over last ~1.4x10 8 yr ● Total energy input significantly exceeds requirements to offset X-ray cooling in cluster

18. 18 Galaxy Cluster Cores: AGN Feedback Chandra image with 8 GHz radio contours dashed areas show ghost holes in the thermal X-ray gas at radii larger than currently active central radio source 330 MHz radio data show an extension to the western X-ray hole holes are where the emission from a previous radio outburst has displaced the surrounding thermal gas McNamara et al. (2000) Clarke et al. (2005)

19. 19 Galaxy Cluster Cores: AGN Feedback 330 MHz 74 MHz Fabian et al. (2002)

20. 20 Cluster Mergers: Diffuse Synchrotron Emission Clarke & Ensslin (2006) Several clusters display large regions of diffuse synchrotron: 'halos' & 'relics' associated with merging clusters radio emission is generally steep spectrum location, morphology, spectral properties, etc... can be used to understand merger geometry Abell 2256

21. 21 Cosmology: Tracing Dark Energy Observations of cosmic acceleration have led to studies of Dark Energy: clusters should be representative samples of the matter density in the Universe study DE through various methods including the 'baryonic mass fraction' requires assumption of hydrostatic equilibrium merging cluster can be identified and removed using low frequency detections of halos and relics (Clarke et al. 2005) Allen et al. (2004)

22. 22 High Redshift Galaxies: Steep Spectrum -4-3-2012 -4 -2 0 2 INCREASING REDSHIFT log S log [ GHz ] THEORETICAL SYNCHROTRON AGING SPECTRA (KARDASHEV-PACHOLCZYK MODEL) Observations of cosmic acceleration have led to studies of Dark Energy: Synchrotron losses steepen the spectrum of radio galaxies at high z Inverse Compton losses act similarly to steepen the spectrum, especially at high z since IC losses scale as z 4. Spectrum is also red shifted to lower frequencies so that the entire observed spectrum is steep.

23. 23 Epoch of Reionization: z  6 (H I at 200 MHz) SDSS: Becker et al. (2001) Universe made rapid transition from largely neutral to largely ionized Appears as optical Gunn-Peterson trough in high-z quasars Also detectable by highly-red shifted 21 cm H I line in absorption against first quasars, GRB’s, SF galaxies … WMAP 3 yr : re-ionization epochs near z~11 (HI at 115 MHz)

24. 24 VLA Low Frequency Sky Survey: VLSS Survey Parameters – 74 MHz – Dec. > -30 degrees – 80” resolution – rms ~100 mJy/beam Deepest & largest LF survey – N ~ 10 5 sources in ~ 80% of sky – Statistically useful samples of rare sources => fast pulsars, distant radio galaxies, cluster radio halos and relics – Unbiased view of parent populations for unification models Important calibration grid for VLA, GMRT, & future LF instruments Data online at: http://lwa.nrl.navy.mil/VLSS Condon, Perley, Lane, Cohen, et al ~ 95 % complete

25. 25 ~20 o VLSS FIELD 1700+690  ~80”, rms ~50 mJy

26. 26 Low Frequency In Practice: Not Easy! Bandwidth smearing Distortion of sources with distance from phase center Finite Isoplanatic Patch Problem: Calibration changes as a function of position Interference: Severe at low frequencies Large Fields of View: Perley lecture Non-coplanar array (u,v, & w) Large number of sources requiring deconvolution Calibrators Phase coherence through ionosphere Corruption of coherence of phase on longer baselines

27. 27 Low Frequencies: Step 1

28. 28 Bandwidth Smearing ● Averaging visibilities over finite BW results in chromatic aberration worsens with distance from the phase center => radial smearing   )x(    synth  ~ 2 => I o /I = 0.5 => worse at higher resolutions Solution: spectral line mode (already essential for RFI excision) Rule of thumb for full primary beam targeted imaging in A config. with less than 10% degradation: 74 MHz channel width < 0.06 MHz 330 MHz channel width < 0.3 MHz 1420 MHz channel width < 1.5 MHz 151.4501420 11 41  MAX (‘) for 50% degradation 75 350 Radius of PB FWHM (‘) A-config.  synth (“) BW (MHz) Freq. (MHz) 66.0330 251.574 1.3

29. 29 Radio Frequency Interference: RFI As at cm wavelengths, natural and man-generated RFI are a nuisance – Getting “better” at low freq., relative BW for commercial use is low At VLA: different character at 330 and 74 MHz – 74 MHz: mainly VLA generated => the “comb” from 100 kHz oscillators – 330 MHz: mainly external – Solar effects – unpredictable Quiet sun a benign 2000 Jy disk at 74 MHz Solar bursts, geomagnetic storms are disruptive => 10 9 Jy! Ionospheric scintillations in the late night often the worst Powerful Solar bursts can occur even at Solar minimum! – Can be wideband (C & D configurations), mostly narrowband Requires you to take data in spectral line mode – RFI can usually be edited out – tedious but “doable”

30. 30 35 km Frequency Time RFI environment worse on short baselines Several 'types': narrow band, wandering, wideband,... Wideband interference hard for automated routines Example using AIPS tasks FLGIT, FLAGR Unfortunately, still best done by hand! afterbefore AIPS: SPFLG RFI Excision 12 km3 km

31. 31 RFI Excision in Practice ● Approach: averaging data in time and/or frequency makes it easier to isolate RFI, which averages coherently, from Gaussian noise, which does not ● Once identified, the affected times/baselines can be flagged in the un-averaged dataset ● Where to start? AIPS tasks: QUACK, SPFLG, TVFLG, UVPLT, UVFND, UVFLG, UVSUB, CLIP, FLGIT, FLAGR,... ● Stokes V can be helpful to identify interference signals before after

32. 32 Ionospheric Structure: ~ 50 km > 5 km <5 km Waves in the ionosphere introduce rapid phase variations (~1 °/s on 35 km BL) Phase coherence is preserved on BL < 5km BL > 5 km have limited coherence times Historically limited capabilities of low frequency instruments

33. 33 Ionospheric Effects Wedge Effects: Faraday rotation, refraction, absorption below ~ 5 MHz (atmospheric cutoff) Wave and Turbulence Effects: Rapid phase winding, differential refraction, source distortion, scintillations ~ 1000 km ~ 50 km Wedge Waves Wedge: characterized by TEC =  n e dl ~ 10 17 m -2 Extra path length adds extra phase  L  2  TEC  ~  L  ~ * TEC Waves: tiny (<1%) fluctuations superimposed on the wedge VLA  The wedge introduces thousands of turns of phase at 74 MHz  Interferometers are particularly sensitive to difference in phase (wave/turbulence component)

34. 34 Ionospheric Refraction & Distortion Both global and differential refraction seen. Time scales of 1 min. or less. Equivalent length scales in the ionosphere of 10 km or less. 1 minute sampling intervals Refractive wander from wedge

35. 35 Antenna Phase as a Function of Time Phase on three 8-km baselines A wide range of phenomena were observed over the 12-hour observation => MYTH: Low freq. observing is better at night. Often daytime (but not dawn) has the best conditions Scintillation Refractive wedge At dawn Quiesence ‘Midnight wedge’ TIDs

36. 36 'Dealing' with the Ionosphere ● Self-calibration models ionosphere as a time-variable antenna based phase:  i (t) ● Loop consisting of imaging and self-calibration - model improves and S/N for self-cal increases ● Typical approach is to use a priori sky-based model such as NVSS, WENSS, or higher frequency source model (AIPS: SETFC, FACES, CALIB, IMAGR) - freezes out time variable refraction - ties positions to known sky-model - DOES NOT ALWAYS WORK – e.g. fails due to thermal absorption ● This method assumes a single ionospheric solution applies to entire FOV - in reality the assumption is only valid over a smaller region but is probably ok if most of the flux is in the source of interest and only want small FOV

37. 37 Isoplanatic Patch Assumption Standard self-calibration sets single ionospheric solution across entire FOV:  i (t) – OK if brightest source is the only target of interest in the field – Problems: differential refraction, image distortion, reduced sensitivity – Solution: selfcal solutions with angular dependence  i (t)   i (t, ,  ) – Problem mainly for 74 MHz A and B arrays Zernike polynomial phase screen – Developed by Bill Cotton (NRAO) – Delivers astrometrically correct images – Fits phase delay screen rendered as a plane in 3-D viewed from different angles Key handicaps: – Need high S/N—significant data loss under poor ionospheric conditions – Total flux should be dominated by point sources New tools will be needed for next generation of instruments

38. 38 Breakdown of Finite Isoplanatic Assumption Self-calibration Zernike Model Average positional error decreased from ~45” to 17” AIPS: VLAFM

39. 39 Large Fields of View (FOV) I Noncoplanar baselines: (u,v, and w) Perley lecture Important if FOV is large compared to resolution => in AIPS multi-facet imaging, each facet with its own  synth E ssential for all observations below 1 GHz and for high resolution, high dynamic range even at 1.4 GHz Requires lots of computing power and disk space AIPS: IMAGR (DO3DIMAG=1, NFIELD=N, OVERLAP=2), CASA (aka AIPS++): w-projection Example: VLA B array 74 MHz: ~325 facets A array requires 10X more: ~ 3000 facets ~10 8 pixels

40. 40 Targeted Faceting Fly’s Eye Outliers ~ 4 degrees A array requires ~10,000 pixels! enormous processing required to image entire FOV reduce processing by targeting facets on selected sources (still large number!) overlap a fly's eye of the central region and add individual outliers AIPS: SETFC AIPS Tip: Experience suggests that cleaning progresses more accurately and efficiently if EVERY facet has a source in it. Best not to have extended sources spread over too many facets => often must compromise

41. 41 Large Fields of View (FOV) II Calibrators: Antenna gain (phase and amplitude) and to a lesser degree bandpass calibration depends on assumption that calibrator is a single POINT source Large FOV + low freq. = numerous sources everywhere At 330 MHz, calibrator should dominate flux in FOV: extent to which this is true affects absolute positions and flux scale => Phases (but not positions) can be improved by self-calibrating phase calibrator => Always check accuracy of positions Must use source with accurate model for bandpass and instrumental phase CygA, CasA, TauA, VirgoA 330 MHz phase calibrator: 1833-210 9 Jy 1 Jy

42. 42 Amplitude and bandpass calibration: Cygnus A (few x 2 min) –Blows through RFI! Phase calibration at 330 MHz: fairly easy –Sky is coherent across the array in C and D configurations Observe one strong unresolved source anywhere in sky –Traditional phase calibration in A and B arrays Now being superseded by NVSS Sky model – no phase calibration required! Phase calibration at 74 MHz: more challenging –Cygnus A (or anything bright) is suitable in the C and D arrays –A and B arrays: Cyg A works for initial calibration, because enough short spacings see flux to start self-cal process Selfcal can’t overcome breakdown of isoplanatic patch assumption Hourly scans on Cyg A => instrumental calibration for non-selfcal (Zernike polynomial) imaging –Calibration schemes continue to evolve rapidly with time! Avoid Sun particularly in compact configurations VLA LF Observing Strategy

43. 43  Two Receivers: 330 MHz = 90cm PB ~ 2.5 O (FOV ~ 5 O ) 74 MHz = 400cm PB ~ 12 O (FOV ~ 14 O )  Simultaneous observations  Max 330 MHz resolution 6 " (+PT resolution ~3 " )  Max 74 MHz resolution 25 " (+PT resolution ~12 " ) Current Low Frequency Interferometers: VLA

44. 44 74 MHz VLA: Significant Improvement in Sensitivity and Resolution 74 MHz VLA Many low frequency instruments in use!

45. 45  Five Receivers: 153 MHz = 190cm PB ~ 3.8 O (res ~ 20 " ) 235 MHz = 128cm PB ~ 2.5 O (res ~ 12 " ) 325 MHz = 90cm PB ~ 1.8 O (res ~ 9 " ) 610 MHz = 50cm PB ~ 0.9 O (res ~ 5 " ) Current Low Frequency Interferometers: GMRT 50 327 610 235 150 1420

46. 46 For more information: Further reading: White Book: Chapters 12.2, 15, 17, 18, 19, & 29 From Clark Lake to the Long Wavelength Array: Bill Erickson's Radio Science, ASP Conference Series 345 Data Reduction: http://www.vla.nrao.edu/astro/guides/p-band/ http://www.vla.nrao.edu/astro/guides/4-band/ Future Instruments: LWA, LOFAR, MWA, FASR http://lwa.nrl.navy.milhttp://lwa.nrl.navy.mil, http://lwa.unm.edu (Taylor lecture)http://lwa.unm.edu http://www.lofar.org/ Thanks to: C. Brogan (NRAO), N. Kassim, A. Cohen, W. Lane, J. Lazio (NRL), R. Perley, B. Cotton, E. Greisen (NRAO)

47. 47 Extra Slides

48. 48 Low Freqs and the EVLA The 74 and 330 MHz receiver systems are not slated for upgrade in the EVLA However, there will be benefits: New correlator will allow much wider bandwidths with sufficient channels to prevent bandwidth smearing at 1420 and 330 MHz 1420 MHz from 50 MHz to 1 GHz 330 MHz from 12 MHz to 40 MHz (limited by front-end filter) 74 MHz will still be limited by front end filter (and confusion) The 100 kHz oscillators that cause the “comb” will be eliminated Significant improvement requires a system designed for low frequencies => LWA (10-100 MHz) and LOFAR (100-300 MHz)

49. 49 For the future: the Long Wavelength Array (LWA) 74 MHz VLA demonstrates major breakthrough in sensitivity & angular resolution =>10 2 less collecting area than UTR-2, but 10 2 better sensitivity – Opens door for sub-mJy, arc-sec resolution LWA of greater size, collecting area, and frequency coverage Consortium of universities, the Naval Research Laboratory, and Los Alamos National Laboratory – Prototyping already underway LWA to explore the region of the EM spectrum below the FM bands – LWA intended to explore region of the spectrum below 100 MHz – 74 MHz VLA and past experience (e.g. Clark Lake) show that technology is in hand to do this at modest cost and with low technical or scientific risk

50. 50 LWA Concept 200 dipoles = 1 Station 1 LWA Antenna 1 meter 100 meters Central condensation ~30 km ~500 km Outliers  Fully electronic, broad-band antenna array - Frequency range:  90 MHz, no ionospheric limit on baseline length  Large collecting area:  1x10 6 m 2  Baselines  100 km  2-3 orders of magnitude improvement in resolution & sensitivity: - [4”, 1.6”] @ [30, 74] MHz;  1 mJy sensitivity  Low Cost: < $50M

51. 51 74 MHz VLA: Significant Improvement in Sensitivity and Resolution 74 MHz VLA

52. 52 History of Low Frequency Astronomy  Radio astronomy began at frequencies of ~ 20 MHz in the 30s with Karl Jansky  First all sky map ever is at 200 MHz (Droge & Priester 1956)  Low freq. receivers (dipoles) easy to make and cheap However: Resolutions poor – degrees => /D (wavelength / longest baseline length) => Ionosphere Sensitivity low – dominated by Galactic background – sky noise => T sys = T ant + T rec => synchrotron background due to several hundred MeV electrons spiraling in Galactic magnetic field

53. 53 All-sky Map – 408 MHz Best Tsys > 50 K resolution ~ 0.85 degrees T b ~ 500 K Haslam et al. (1982)

54. 54 All Sky Map – 150 MHz Best Tsys > 150 K resolution ~ 2.2 degrees T b ~ 3000 K Landecker & Wielebinski (1970)

55. 55 All-sky map, 45 MHz Best Tsys > 3000 K resolution ~ 4 degrees T b ~ 45,000 K Alvarez et al. (1997)

56. 56 Low Angular Resolution: Limits Sensitivity Due to Confusion  ~ 1’, rms ~ 3 mJy/beam  ~ 10’, rms ~ 30 mJy/beam

57. 57 Galactic Center ➢ Galactic Center: Galactic center black hole Sgr A*, non-thermal filaments, transients, supernova remnants, HII regions, diffuse nonthermal source (DNS) Spectrum of Sgr A* Sub-mm - Zhao, Bower, & Goss (2001) Radio - Zhao, Bower, & Goss (2001) 610 MHz – Roy et al. (2003) 330 MHz – Nord et al. 2004, ApJ, 601, L51. LaRosa et al. (2000) Sgr A*

58. 58 330 MHz Survey of Inner Galactic Plane  VLA 330 MHz mosaic composed of C+D configuration data  The resolution is 2.2’ x 1.4’ and the rms noise is ~15 mJy/beam  The mosaic is made up of 14 pointings, 3 from the VLA archive  Superior to any previous survey for  < 2 GHz. Brogan et al. (2004) 2695 MHz (11cm) Bonn Survey resolution 4.3’

59. 59 The VLA Galactic Plane Survey Area Greyscale Bonn 21cm (1465 MHz) Survey with 9.4’ resolution * 74 MHz (4 m) * A, B, C, & D configurations * final resolution ~45” * rms ~50 mJy/beam * 330 MHz (90cm) * B, C, & D configurations * final resolution ~20” * rms ~5 mJy/beam

60. 60 Ionospheric Structure: Limited Angular Resolution ~ 50 km Phase coherence preserved Phase coherence corrupted > 5 km <5 km Compared to shorter : Maximum antenna separation: 10 3 km) Angular resolution:  > 0.3  (vs. < 10 -3  ) Sensitivity confusion limited: rms  1–10 Jy (vs. < 1 mJy) => Over time push for higher resolution and sensitivity meant shorter Recent revolution due to advances in: (self) calibration, imaging, and overall computing power

61. 61 Interferometry Relies on Good Phase Stability: Dominated & “Corrupted” by the Ionosphere for  ~1 GHz 330 MHz A array74 MHz – 4 times worse

62. 62 Noise Characteristics  A+B array noise in 74 MHz maps decreases ~ t -1/2  Slower improvement with BW => confusion limited hours 1410 42 rms noise – mJy/beam 74 MHz: A+B array BW = 1.5 MHz t -1/2 rms noise vs. time rms noise vs. BW thermal 5001000 Bandwidth (kHz) 150 50 rms noise (mJy/beam) 74 MHz B array 1 hour

63. 63 Large Fields of View (FOV) III Calibrators There are no point-like calibrators below 100 MHz! => Must use source with accurate model for bandpass and instrumental phase CygA, CasA, TauA, VirgoA Then can try NVSS model, or other previous low freq. image (i.e. 330 MHz) of the field but be cautious! If field is dominated by thermal sources this will not work well and possibly not at all Positions can be off by significant amount (10s of arcseconds), especially if model is not a good representation of 74 MHz emission Cygnus A ~ 18,000 Jy

64. 64 VLA 4m resolution 2.1’ x 1.2’ using A+B+C+D config. Data rms ~ 0.1 Jy/beam Integrated Flux ~ 4000 Jy VLA 74 MHz (4 m) Image Brogan et al. (2004) LaRosa et al. (2000) => case where 330 MHz model didn’t work well!