G54.1+0.3 and its environment
Laura N. Driessen
MSc Astronomy & Astrophysics
Astronomy & Astrophysics Track
Master Thesis
LOFAR observations of pulsar wind nebula G54.1+0.3
and its environment
by
Laura Nicole Driessen
1112 4660 July 2017 120 ECTS 2016 – 2017 Supervisor: Dr. Jason Hessels Associate Professor Co-supervisor: Dr. Jacco Vink Associate Professor
LOFAR observations of pulsar wind nebula G54.1+0.3 and its environment
Laura N. Driessen
Astronomy & Astrophysics Master’s Thesis The Anton Pannekoek Institute The University of Amsterdam
THE UNIVERSITY OF AMSTERDAM
Abstract
The Faculty of Science Physics & Astronomy Department
Research Master
LOFAR observations of pulsar wind nebula G54.1+0.3 and its
environment
byLaura N. Driessen
We present the calibration, imaging and analysis of LOw Frequency ARray (LOFAR) High Band Antenna (HBA) observations of the Galactic Plane. We investigate four calibration methods: prefactor, the Multi-frequency Snapshot Sky Survey (MSSS) imaging pipeline, direction-independent non-zero-phase cal-ibration, and direction-independent zero-phase calibration. We show that direction-independent zero-phase calibration produces the best results for this field and particular observation. We use WSClean to image the observations and pyBDSF to measure the flux density of sources in the field of view.
Due to the low observing frequency, wide-bandwidth, and large field-of-view of the LOFAR instrument it is excellent for investigating the Galactic Plane. This is because low-frequency objects with similar morphology, such as supernova remnants and HII regions, can be differentiated due to their behaviours at different frequencies. In particular, due to their spectral indices (α, where flux density S scales with observing frequency ν as S ∝ να) supernova remnants become brighter at lower frequencies (α ≈ −0.5) while HII regions become brighter at higher frequencies (α ' 0). Low-frequency observations are also useful for studying pulsar wind nebulae by investigating the lowest emission frequency of the electrons, which acts as a calorimeter for the total injected energy, and by measuring a range of frequencies to test pulsar wind nebula evolution models.
We present an investigation of pulsar wind nebula G54.1 + 0.3 and supernova remnants in the field of view, using LOFAR HBA and archival multi-frequency observations. We investigate and improve the multi-frequency spectral energy distribution of PWN G54.1 + 0.3. We also show that, contrary to previous suggestions, pulsar wind nebula G54.1 + 0.3 does not have a large scale supernova remnant shell. This confirms its classification as a filled-centre supernova remnant. We discuss pulsar wind nebula G54.1 + 0.3 in the context of supernova remnants with significant amounts of cold supernova ejecta dust.
Furthermore, we discuss other supernova remnant candidates in the field of view with LOFAR, showing that some candidates in the field are unlikely to be supernova remnants. We also present our discovery of a new supernova remnant, G53.41 + 0.03, the first supernova remnant discovered with LOFAR.
LOFAR observations of pulsar wind nebula G54.1+0.3 and its
environment
byLaura N. Driessen
Our own Milky Way Galaxy is home to a wide range of interesting objects visible to radio telescopes: pulsars, pulsar wind nebulae, supernova remnants, HII regions, and more. Supernova remnants are the shells of dust left after massive stars die in supernova explosions. We observe supernova remnants as bright bubbles in space. A supernova explosion can also leave a pulsar behind. A pulsar is a rapidly spinning, highly magnetised star with a radius of only 10 km but with a mass about that of the Sun. We observe pulsars as flashing “astronomical lighthouses”. Pulsars can accelerate particles in the highly magnetised region around them, called the magnetosphere. As these particles flow off the pulsar they produce a pulsar wind. This wind fills the bubble of the supernova remnant, and when this happens the wind-filled bubble is called a pulsar wind nebula. HII regions are a different kind of Galactic object. An HII region is an area of ionised hydrogen that can also look like a bubble in space, and these are typically regions where stars are being formed.
We can study these objects by observing them at low frequencies of light with arrays of antennas, called radio interferometers. In particular, both supernova remnants and HII regions can look like bright bubbles in the Galaxy. This makes it hard to tell them apart. Luckily, supernova remnants become brighter at lower frequencies, while HII regions become brighter at higher frequencies. So if we look at our Galaxy at a range of frequencies we can differentiate between objects that have a similar shape.
We can investigate pulsar wind nebulae using low-frequency observations as well. By observing a pulsar wind nebula at low frequency we can investigate how it produces and accelerates particles in its magne-tosphere, we can find out how old it is and how much energy it has overall, we can find out about the explosion that formed the pulsar and about the pulsar itself.
For this project we processed and analysed observations of a particular pulsar wind nebula, named G54.1 + 0.3, using low-frequency observations from the LOw Frequency ARray (LOFAR) radio interferometer. Pro-cessing low-frequency radio interferometric data is a challenging task, so we investigated various methods before finding the optimal strategy. We used our LOFAR observations, and archival observations at a range of frequencies, to investigate the properties of pulsar wind nebula G54.1 + 0.3 and show that it does not have a supernova remnant bubble around it, contrary to previous suggestions. We show that some objects that were classified as possible supernova remnants are not actually supernova remnants because we do not detect them in the LOFAR observations. We discovered a new supernova remnant, named G53.41 + 0.03, with the LOFAR observations and studied it using a range of observations at different frequencies.
Acknowledgements
I would like to thank my supervisors, Dr. Jason Hessels and Dr. Jacco Vink, for all of their help and support along the way. I would particularly like to thank Jason for helping me to apply for and choose a PhD position and for listening to my angst and whining while I was making the decision.
Thank you to Maria Arias de Saavedra who answered my silliest LOFAR questions and explained selfcal at least three times in a row until I understood it. And thank you to Vladim´ır Domˇcek for his expertise in X-ray astronomy.
I would like to thank Dr. Gemma Janssen for supporting my research journey since 2014. I really appreciate your advice and I could not ask for a better mentor.
Thank you to my sister, Dr. Brooke Driessen MD, for looking after my dog, Astro, and keeping up his instagram account while I have been overseas studying. I miss you and Astro every day.
Thank you to my sister, Chelsea Driessen, for always pushing me to be my best, for encouraging my ambition and especially for pushing me out of my comfort zone just when I need it.
My parents, Leo Driessen and Debra Driessen, have supported me every step of the way and have always encouraged me to just go for it. Not many parents would support their daughter moving all over the world to study astronomy. Thanks Dad for reading and editing all my essays, CVs and applications. Thanks Mum for listening to my whinging when I need it and for getting me to snap out of it when I need to. I would not be where I am today without parents like you.
Abstract i Popular abstract ii Acknowledgements iii List of Figures vi List of Tables ix Abbreviations x 1 Introduction 1 1.1 Supernova remnants. . . 2 1.2 Pulsars . . . 4
1.3 Pulsar wind nebulae . . . 4
2 LOFAR Observations and Processing 6 2.1 HBA observations . . . 6
2.2 Calibration methods . . . 7
2.2.1 Pre-Facet-Cal . . . 9
2.2.2 Direction-independent calibration . . . 10
2.2.3 Calibrating the G54.1+0.3 observation . . . 10
2.2.3.1 The MSSS imaging pipeline . . . 11
2.2.3.2 Pre-Facet-Cal . . . 12
2.2.3.3 Direction-independent calibration . . . 12
2.2.3.4 Chosen calibration method . . . 12
2.3 Imaging . . . 14 2.4 Flux determination . . . 16 3 PWN G54.1+0.3 19 3.1 Past work . . . 19 3.1.1 Discovery . . . 19 3.1.2 Observations. . . 20 3.1.2.1 Low-frequency observations . . . 20 iv
Contents v
3.1.2.2 X-ray observations . . . 22
3.1.2.3 Infrared observations . . . 23
3.1.2.4 Gamma-ray observations . . . 23
3.1.3 The distance to PWN G54.1+0.3 . . . 24
3.1.4 Magnetic Fields and Polarisation . . . 24
3.1.5 PSR J1930+1852: the central pulsar of PWN G54.1+0.3 . . . 24
3.2 Supernova remnant features of PWN G54.1+0.3 . . . 26
3.2.1 The infrared dust bubble around PWN G54.1+0.3. . . 27
3.2.1.1 IR-excess stellar objects. . . 27
3.2.1.2 A star-forming region . . . 27
3.2.1.3 Supernova ejecta dust . . . 29
3.2.2 The forward shock. . . 29
3.2.2.1 Detecting the forward shock at LOFAR frequencies . . . 29
3.3 Spectral energy distribution . . . 31
4 The G54.1+0.3 Field 36 4.1 Observations . . . 37
4.1.1 Radio observations . . . 37
4.1.2 Infrared observations . . . 37
4.2 Supernova Remnants and candidates . . . 37
4.2.1 G53.41+0.03 . . . 37
4.2.1.1 Radio observations . . . 39
4.2.1.2 Pulsar search . . . 40
4.2.1.3 Infrared observations . . . 42
4.2.1.4 X-ray observations . . . 42
4.2.1.5 High energy observations . . . 44
4.3 Discussion . . . 44
4.3.1 The distance to G53.41+0.03. . . 45
4.3.2 The age of G53.41+0.03 . . . 45
4.3.3 Other supernova remnants and candidates in the field of view . . . 46
5 Summary 50 A Calibration pipeline functions 53 B Measuring flux densities with pyBDSF 62 B.1 Script to convert VLSSR and WSRT catalogues to a usable format. . . 62
B.2 Running pyBDSF and aperture flux density measurements . . . 65
B.3 Extracting point source information from TGSS ADR and pyBDSF . . . 67
C Searching for a pulsar in G53.41+0.03 - a new supernova remnant detected with LOFAR 75
1.1 LOFAR low band antennas on the superterp located in Drenthe, the Netherlands. . . 2
1.2 A three colour image of shell-type SNR Cassiopeia A. Red is a high-resolution VLA ob-servation at 4.7 GHz. Green and blue are both Chandra X-ray obob-servations. Blue shows 4.2 − 6.2 keV X-rays showing the synchrotron continuum and green is the full Chandra band of 0.3 − 10.0 keV. The green X-ray emission at the edge of the SNR shows the location of the FS and the blue ring inside the SNR shows the RS. The data to produce this image were provided by Vladim´ır Domˇcek. . . 3
1.3 Basic structure of a PWN assuming spherical symmetry. This is an oversimplification as the outer edge of the PWN is confined by the ISM or the SNR it is surrounded by, and the ISM/SNR is not homogeneous. The overall structure of a PWN is typically much more complicated. . . 5
2.1 HBA SB160 with a central frequency of 149.9985 MHz and bandwidth of 0.1953 MHz. The top image has been averaged and demixed but not calibrated. The centre image has undergone direction independent calibration as part of the MSSS calibration pipeline The bottom image has been calibrated using the Pre-Facet-Cal (prefactor) pipeline. The synthesised beam is shown in the bottom left corner of each panel. In the bottom panel the synthesised beam is too small to see. . . 8
2.2 Clock-TEC solutions of the HBA observations produced when running prefactor. The colours indicate the different LOFAR stations. The y-axis indicates the ionospheric varia-tions. Time is in units of time steps in the MSs. . . 9
2.3 HBA SB000 after direction-independent MSSS pipeline calibration in the LTA. SB000 has a central frequency of 118.7485 MHz and bandwidth of 0.1953 MHz. The synthesised beam is shown in the bottom left corner. . . 11
2.4 HBA SB160 that has been calibrated using top: direction-independent, time-independent calibration with zero-phase=F and bottom: direction-independent, time-independent cali-bration with zero-phase=T. SB160 has a central frequency of 149.9985 MHz and bandwidth of 0.1953 MHz. The synthesised beams are shown in the bottom left corner of each panel. 13
2.5 Comparison of clean models without multiscale (top panel) and with multiscale (bot-tom panel). Both of these cleans were performed on the same HBA SB with a central frequency of 150.8 MHz. The image colour scale has been saturated at 50 mJy beam−1. The synthesised beams are shown in the bottom left of each panel.. . . 15
2.6 The PSFs of a multi-frequency clean of an HBA observation with a central frequency of 144 MHz. The panels have Briggs weightings from left to right of −1.0, 0.0, and 1.0. The synthesised beams of the cleans are shown in the bottom left corner of each panel. . . 15
2.7 Comparison of a clean without a uv-cut (top panel) and with a uv-cut of 100 λ (bottom panel). Both of these cleans were performed on the same HBA SB with a central frequency of 150.8 MHz. The synthesised beams are shown in the bottom left corner of each panel.. 16
List of Figures vii
2.8 Radio spectra of isolated point sources in the LOFAR HBA FoV using two different Briggs weightings and using averaged FITS files. Flux density measurements from VLSSR, TGSS ADR and WSRT are also plotted. These point sources have J2000 coordinates in degrees of (290.7, 18.8), (290.3, 17.0), and (293.4, 21.6) respectively. . . 18
3.1 Cambridge 5-km telescope observation of PWN G54.1 + 0.3 by Green (1985). This obser-vation has a resolution of 700× 2000 and an observing frequency of 2.7 GHz. . . 20
3.2 Low-frequency spectral energy distribution of PWN G54.1 + 0.3 produced by Velusamy & Becker (1988). The power law has a spectral index of α ∼ −0.13 confirming that PWN G54.1 + 0.3 has a spectrum similar to that of the Crab Nebula. . . 21
3.3 Chandra ACIS-S X-ray (0.5–10 keV) intensity map of PWN G54.1 + 0.3 (Lu et al., 2002). The intensity is plotted from 7.43 × 10−5 to 5.95 × 10−3 counts cm−2s−1arcmin−2. The inset image shows a closer view of the central region of the PWN containing the point source that is the central pulsar and the torus of the wind termination shock. . . 22
3.4 A schematic diagram of the basic structure of a shell-type SNR. . . 26
3.5 A three-colour image of PWN G54.1 + 0.3. Chandra X-ray observations are in blue, Merca-tor Telescope (MAIA) 890 nm optical observations are shown in green and Spitzer (MIPS-GAL) 24 µm IR observations are shown in red. We acquired the MAIA data while visiting the Mercator Observatory as part of the Observation Project course at the University of Amsterdam. . . 28
3.6 VLA 1.4 GHz observation of PWN G54.1 + 0.3 (Lang et al., 2010). The possible shell is circled in red and PWN G54.1 + 0.3 is the bright central source. . . 30
3.7 LOFAR HBA MFS image of PWN G54.1 + 0.3 and the surrounding environment. The white circles are known SNRs from Green’s SNR catalogue (Green, 2014) and the dashed white circles are SNR candidates from Anderson et al. (2017). The red, green and yellow circles are known, group and radio quiet HII regions respectively from the WISE HII region catalogue (Anderson et al., 2014). The synthesised beam is shown in the bottom right corner. 31
3.8 SED of PWN G54.1 + 0.3 from Gelfand et al. (2015). The orange points are VLA fluxes. The green line is the Chandra X-ray integrated flux. The blue points are VERITAS high-energy gamma-ray measurements. The black line is a fit based on their PWN evolution model. . . 32
3.9 SED of PWN G54.1 + 0.3 using archival observations discussed in Section 3.1.2. Note that the IR points measure the flux density of the SN ejecta dust around PWN G54.1 + 0.3, not the PWN itself. . . 32
3.10 Radio SED of PWN G54.1 + 0.3 using archival radio observations discussed in Section 3.1.2.1. The power laws are a result of fitting a power law or broken power law to the data, they are not a result of the PWN evolution model by Gelfand et al. (2015). The VLA power law is a simple power law fit to only the VLA observations. It has a spectral index of α = −0.29. The VLA+TGSS broken power law is a broken power law fit to the VLA and TGSS points. It has spectral indices of α1 = −0.03 and α2 = −0.42. The archival radio
power law is a simple power law fit to all the radio points (excluding the LOFAR points). It has a spectral index of α = −0.13, the same spectral index found by Velusamy & Becker (1988) (see Sec. 3.1.1). . . 33
3.11 Low-frequency radio SED of PWN G54.1 + 0.3 showing the TGSS ADR, VLSSR, and WSRT points and the LOFAR HBA points using Briggs 1.0, Briggs 0.6, and Briggs 1.0 averaged/combined images. The background brightness temperature is also plotted. In the two lowest frequency HBA SBs confusion with the HII region surrounding PWN G54.1+0.3 results in higher flux density measurements. . . 34
4.1 Observations of the Galactic Plane FoV where the LOFAR HBA observation, VGPS mosaic and WSRT mosaic coincide. Top: VGPS 1.4 GHz mosaic. Centre: WSRT 327 MHz mosaic. Bottom: Spitzer 24.0 µm MIPSGAL mosaic. . . 38
4.2 Observations of the Galactic Plane at 1.4 GHz (blue, VLA), 327 MHz (green, WSRT) and 144 MHz (red, LOFAR). The synthesized beam sizes are shown in the bottom left corner. Known SNRs from Green’s SNR catalogue (Green, 2014) are circled in solid white and candidate SNRs from Anderson et al. (2017) are circled in dashed white. The red, green, yellow, and cyan circles are known, group, radio quiet and candidate HII regions respectively (Anderson et al., 2014). . . 39
4.3 LOFAR HBA observations of G53.41 + 0.03, PWN G54.1 + 0.3, and SNR HC 40. Both panels have a frequency bandwidth of 1.95 MHz. The left panel has a central frequency of 131.3 MHz and the right panel has a central frequency of 154.8 MHz. The synthesised beam sizes are shown in the bottom right corner of both panels. . . 39
4.4 Observations of G53.41 + 0.03 at (a) 1.4 GHz using the VLA, (b) 24.0 µm using Spitzer and (c) X-rays using XMM-Newton. The dashed cyan contours are from LOFAR HBA 150.8 MHz data (contour levels: 0, 0.25, 0.5, 0.75, 1.0 Jy beam−1) and the solid yel-low contours are from VLA 1.4 GHz data (contour levels: 12, 14, 16, 18, 20, 22, 24 mJy beam−1) from the image in (a). . . 41
4.5 Standard Presto diagnostic plot of the test pulsar PSR J1928+1746 at 1450.168 MHz. This shows the pulse profile as a function of time and frequency during the course of this observation. . . 42
4.6 Left: X-ray spectrum showing the best-fit model. The Al Kα instrumental background line around 1.49 keV has been blanked out. Right: Contour plots of the ionization age and post-shock temperature. . . 44
4.7 SNR candidates from Anderson et al. (2017) in the FoV. In each row the left panel is the VGPS 1.4 GHz observation, the center panel is the WSRT 0.327 GHz observation, and the right panel is the LOFAR 0.144 GHz observation. From top to bottom the rows are the SNR candidates (circled in dashed white) from Anderson et al. (2017): G51.21 + 0.11, G52.37 − 0.70, and G53.07 + 0.49. The solid yellow circles are HII regions from the WISE HII catalog (Anderson et al., 2014). The synthesised beams are shown in the bottom left corner of each image. . . 47
4.8 SNR candidates from Anderson et al. (2017) in the FoV. In each row the left panel is the VGPS 1.4 GHz observation, the center panel is the WSRT 0.327 GHz observation, and the right panel is the LOFAR 0.144 GHz observation. From top to bottom the rows are the SNR candidates (circled in dashed white) from Anderson et al. (2017): G53.84 − 0.75 and G54.1+0.3. The solid yellow circles are HII regions from the WISE HII catalogue (Anderson et al., 2014). The synthesised beams are shown in the bottom left corner of each image. 48
5.1 Word Cloud summary of the most common words in this thesis. . . 51
C.1 A three-colour image of the Galactic Plane showing G53.41 + 0.03 (circled in dashed white), SNR HC 40 and PWN G54.1 + 0.3 (both circled in white). The three colours in this image are: LOFAR HBA 150 MHz in red, Westerbork Synthesis Radio Telescope (WSRT) 327 MHz in green and Very Large Array (VLA) Galactic Plane Survey (VGPS) 1400 MHz in blue. The red, yellow and green circles are known, radio quiet and group HII regions respectively from the WISE catalogue (Anderson et al., 2014). . . 77
List of Tables
2.1 Observation information for both of the calibrator scans and the target scan. All
observa-tions were taken on 2015/06/12. . . 6
3.1 Archival radio flux density measurements of PWN G54.1 + 0.3. . . 21
3.2 Archival X-ray observations of PWN G54.1 + 0.3. . . 23
3.3 Archival IR flux density measurements of PWN G54.1 + 0.3. . . 23
3.4 Archival gamma-ray observations of PWN G54.1 + 0.3. . . 24
3.5 The observed parameters of PSR J1930 + 1852. The Lu et al. (2007) values are from the 12th September 2002 Rossi X-Ray Timing Explorer observations. . . 25
3.6 The derived properties of PSR J1930 + 1852 by Camilo et al. (2002). . . 25
3.7 Observed epochs of J1930 + 1852 from Lu et al. (2007). Periods denoted by a * were taken from Camilo et al. (2002). . . 26
4.1 Details of the pointings for the Arecibo pulsar search of G53.41 + 0.03. The expected sensitivity is quoted for a minimum signal-to-noise ratio of 15 and pulse duty cycle of 10%. 40 4.2 The XMM-Newton best-fit model results. The abundances are provided in Solar units. . . 44
4.3 Flux densities of SNR candidates at 1.4 GHz from Anderson et al. (2017) and 327 MHz measured using WSRT observations (Taylor et al., 1996). The WSRT errors are 3σ statis-tical errors based on the RMS noise in the image; these errors do not take other sources of error, such as confusion, into account. . . 49
3FGL Third Fermi-LAT Catalogue of High-Energy Sources ACIS-S Advanced CCD for Imaging Spectrometer
ADR Alternative Data Release ALFA Arecibo L-band Feed Array
ASCA SIS Advanced Satellite for Cosmology and Astrophysics Solid-state Imaging Spectrometer ASTRON The Netherlands Institute for Radio Astronomy
Cas A Cassiopeia A
CCD Charge-Coupled Device CCSN(e) Core Collapse SuperNova(e) clock-TEC clock-Total Electron Count CSV Comma Separated Values DM Dispersion Measure
Einstein IPC Einstein Image Proportional Counter FC Filled-Centre
FITS file Flexible Image Transport System file FoV Field of View
FS Forward Shock
FWHM Full Width Half Maximum
GMRT Giant Metrewave Radio Telescope HBA High-Band Array
HEGRA High Energy Gamma-Ray Astronomy
HESSCAT High Energy Stereoscopic System CATalogue INTEGRAL INTErnational Gamma-Ray Astrophysics Laboratory IR InfraRed
IRAS The InfraRed Astronomical Satellite
Abbreviations xi
IRC Infra-Red Camera
IRIS Improved Reprocessing of the IRAS Survey ISM InterStellar Medium
Jy Jansky
LBA Low-Band Array
LOFAR The Low Frequency Array LTA Long Term Archive MFS Multi-Frequency
MIPSGAL Multiband Infrared Photometer for Spitzer GALactic Plane Survey MJD Modified Julian Date
MoM Management of Measurements MS Measurements Set
MSSS Multifrequency Snapshot Sky Survey
NE2001 Cordes-Lazio NE2001 Galactic Free Electron Density Model NIR Near Infra-Red
NIRC Near Infra-Red Camera
OSRT Ooty Synthesis Radio Telescope
PACS Photoconductor Array Camera and Spectrometer PALFA Arecibo Pulsar survey using ALFA
PANIC Persson’s Auxilliary Nasmyth IR Camera PSF Point Spread Function
PSR Pulsar
PWN(e) Pulsar Wind Nebula(e)
pyBDSF python Blob Detector and Source Finder RFI Radio Frequency Interference
RMS Root Mean Squared ROSAT R ¨Ontgensatellit RS Reverse Shock
RXTE Rossi X-ray Timing Explorer SB(s) SubBand(s)
SED Spectral Energy Distribution SN(e) Supernova(e)
SPIRE Spectral and Photometric Imaging Reciever TGSS TIFR GMRT Sky Survey
THOR The HI, OH, Recombination Line Survey of the Milky Way VERITAS Very Energetic Radiation Imaging Telescope Array System VLA Very Large Array
VGPS VLA Galactic Plane Survey
VLSSR VLA Low-Frequency Sky Survey Redux Source Catalogue WISE Wide-Field Infrared Survey Explorer satellite
WSRT Westerbork Synthesis Radio Telescope XMM X-ray Multi-Mirror Mission
Chapter 1
Introduction
The Galactic Plane of the Milky Way, as seen at low radio frequencies, is rich in point sources and extended emission including pulsars, pulsar wind nebulae (PWNe), supernova remnants (SNRs) and HII regions. Observing these objects at low frequencies with a wide frequency bandwidth and a large field of view (FoV) can be useful for investigating the characteristics of these objects and can be used to differentiate between objects with similar morphologies, like SNRs and HII regions.
LOFAR is the LOw Frequency ARray radio interferometer (van Haarlem et al., 2013). It has ‘Core’ and ‘Remote’ stations in the Netherlands and ‘International’ stations spread across Europe. Including the core and remote stations in the Netherlands means that LOFAR has baselines up to 100 km. Using the 100 km baselines results in an optimum resolution of 3.300. Because of its range of baselines, LOFAR is an excellent instrument for observing both extended emission and point sources with good resolution. LOFAR consists of two antenna arrays: the Low Band Antennas (LBA, shown in Fig.1.1) and High Band Antennas (HBA). The LBA observes between 10 and 90 MHz while the HBA observes between 110 and 250 MHz. The low observing frequencies, wide bandwidth, and large FoV of the LOFAR instrument make it ideal for observing and discovering steep-spectra objects and for differentiating between SNRs and HII regions.
Observing the Galactic Plane at low frequencies with a wide FoV presents unique interferometry challenges, particularly regarding ionospheric issues, calibration, and dealing with the extended emission that dominates the Galactic Plane. We present a detailed investigation of methods for processing and analysing LOFAR HBA imaging observations of the Galactic Plane in Chapter2. We discuss four methods of HBA calibration (Sec.2.2), producing HBA images using WSClean (Sec.2.3), and measuring point source flux densities with pyBDSF (Sec.2.4).
In this thesis we will present the analysis of LOFAR observations of the Galactic Plane. We will focus on investigating PWN G54.1 + 0.3 and on SNRs and SNR candidates in the FoV.
Figure 1.1: LOFAR low band antennas on the superterp located in Drenthe, the Netherlands.
1.1
Supernova remnants
While the extended emission in the Galactic Plane presents challenges for low-frequency interferometry, some low frequency extended emission is interesting and exciting to observe; for example SNRs. At the end of a massive (' 8 M ) star’s life, photo-disintegration occurs and the star can no longer support itself
through degeneracy pressure. This causes the star to collapse and subsequently explode as a core-collapse supernova (CCSN; Carroll & Ostlie, 2006). The explosion produces several solar masses of ejecta that expands into the surrounding medium at speeds of tens of thousands of kilometers a second (Slane,2017). At the interface of the expanding ejecta and the surrounding material a forward shock (FS) forms. As the FS decelerates a reverse shock (RS) forms that heats cold supernova ejecta dust that is also formed in the explosion. At the FS the material is compressed by a factor of ∼ 4. As the shock is expanding, the material inside is diluted. This compressed material with dilute material inside gives rise to the shell-morphology usually expected of SNRs (Vink,2012a). We further discuss CCSNe and the different types of SNRs - shell-type, filled-centre, and composite - in Section 3.2. A classic example of an SNR with a shell morphology and both a FS and RS is Cassiopeia A, shown in Figure 1.2.
SNRs have spectral indices of α ≈ −0.5, where Sν ∝ να for Sν flux density in Jy and ν frequency in Hz
(Oni´c, 2013). This negative spectral index is because SNRs emit synchrotron radiation. The FS of the SNR accelerates particles to relativistic energies and the particles emit synchrotron radiation as they spiral around magnetic field lines. Having a negative spectral index means that SNRs become brighter at lower frequencies and can be observed and detected using low-frequency observations with wide bandwidth.
Chapter 1: Introduction 3
Figure 1.2: A three colour image of shell-type SNR Cassiopeia A. Red is a high-resolution VLA observation at 4.7 GHz. Green and blue are both Chandra X-ray observations. Blue shows 4.2 − 6.2 keV X-rays showing the synchrotron continuum and green is the full Chandra band of 0.3 − 10.0 keV. The green X-ray emission at the edge of the SNR shows the location of the FS and the blue ring inside the SNR shows the RS. The
data to produce this image were provided by Vladim´ır Domˇcek.
There is another common source of extended emission in the Galactic Plane: HII regions. It can be difficult to differentiate between SNRs and HII regions as SNRs and HII regions can be morphologically similar. An HII region is a region of atomic hydrogen which is usually ionised due to heating by stars in the region. Typically, HII regions are star-forming regions and they can be compact or extended, morphologically irregular or bubble-like. The thermal emission from HII regions is caused by free-free radiation which results in a spectral index of α ' 0. This means that HII regions are brighter at higher frequencies. As HII regions become brighter at higher frequencies while SNRs become brighter at lower frequencies low-frequency, wide-bandwidth observations are excellent for differentiating between HII regions and SNRs. We use this concept in Section 4.3.3 to determine that two new Galactic SNR candidates are in fact much more likely to be HII regions while one is a good SNR candidate, and to show in Section 3.2.2.1that there
is no SNR shell around PWN G54.1 + 0.3.
1.2
Pulsars
CCSN explosions result in a shell of expanding ejecta, FS, RS and cold ejecta dust. The cold ejecta dust is heated by the RS. CCSNe also produce a central compact object: a neutron star or a black hole. Often what is left is a rapidly spinning neutron star with a strong magnetic field: a pulsar.
A pulsar is a rapidly rotating neutron star with a strong dipole magnetic field (Carroll & Ostlie, 2006). This makes pulsars like flashing “astronomical lighthouses”. By studying the pulsed light from a pulsar properties of the pulsar can be found. Dispersion measure (DM) is the column density of electrons between the pulsar and the observer. The DM of a pulsar in combination with a model of the distribution of free electrons in the Galaxy (eg.Cordes & Lazio 2002) can be used as a proxy for distance to the pulsar and SNR. The derivative of the spin period of a pulsar can be used to find an approximate age, called the characteristic age (τc), of the pulsar, and hence the SNR. The equation for the characteristic age of a
pulsar is given by:
τc≡
P
2 ˙P (1.1)
where P is the pulse period and ˙P is the period derivative or spin-down rate. This equation assumes that the initial period (P0) of the pulsar is much smaller than the current period (P ) and that the spin-down
is due to magnetic dipole radiation (Lorimer & Kramer,2004). It has been shown that the characteristic age can be inaccurate for pulsars with ages more accurately known through other methods. As such, the characteristic age should be used with care and should be considered an approximation only. We discuss some basic properties of the pulsar in PWN G54.1 + 0.3 in Section 3.1.5.
1.3
Pulsar wind nebulae
A pulsar wind is the highly magnetised, relativistic outflow of particles from a neutron star magnetosphere. This wind expands into the interstellar medium (ISM) or the SNR shell surrounding the pulsar to produce a PWN. As the wind expands into the PWN, which is now expanding slowly as it is confined by the ISM or SNR, this produces a termination shock. This is where the ram pressure of the unshocked wind balances with the pressure of the PWN (Slane, 2017). PWNe are structured as shown in Figure 1.3. The main focus of this Master’s research is a specific PWN, PWN G54.1 + 0.3. We present an in-depth investigation of PWN G54.1 + 0.3 in Chapter 3.
Observations and modelling of PWNe can help answer fundamental physics and astronomy questions (Gelfand et al.,2015). By investigating the spectral energy distribution (SED) of a PWN the multiplicity, or number of particles produced in the magnetosphere, of the neutron star and the acceleration of particles
Chapter 1: Introduction 5
Figure 1.3: Basic structure of a PWN assuming spherical symmetry. This is an oversimplification as the outer edge of the PWN is confined by the ISM or the SNR it is surrounded by, and the ISM/SNR is not
homogeneous. The overall structure of a PWN is typically much more complicated.
can be investigated. Modelling the evolution of a PWN can provide information about the CCSN that formed the pulsar such as the initial energy of the SN and the mass of the progenitor star. PWN evolution can be used to investigate the initial properties of the pulsar itself, such as the initial spin period of the pulsar. The low-frequency flux density of a PWN is important because the lowest frequency emitting electrons can act as a “calorimeter” for the total particle energy of the PWN. The low-frequency flux density can be used to test current evolution and SED models, for example by finding the cut-off energy (Gelfand et al.,2015). We investigate the SED of PWN G54.1 + 0.3 in Section3.3.
LOFAR Observations and Processing
The LOFAR observations analysed for this project were centred on PWN G54.1 + 0.3 and were taken in observing Cycle 4 as part of project LC4 011 on 2015/06/12. Both HBA (ObsID: 345918) and LBA (ObsID: 346360) observations were acquired, but this project is focused on the HBA observations. This is due to the extra calibration and processing required to perform LBA analysis.
2.1
HBA observations
Our HBA target and calibrator scans cover the frequency range from 118.7 MHz to 169.5 MHz. The observing bandwidth was split into 260, 195.3 kHz wide subbands (SBs). For these observations an 18 min calibrator scan of 3C380 was taken both before and after the 3 hr target scan. Details of the observations are shown in Table 2.1. Despite the information in the LOFAR Management of Measurements (MoM) tool it was difficult to ascertain what processing had been performed on the observations in the automatic pipeline. In this chapter we investigate and compare four different methods for calibrating LOFAR HBA observations (Sec.2.2), and we discuss imaging with WSClean (Sec.2.3), and measuring flux densities with pyBDSF (Sec.2.4).
The LOFAR observations were flagged to remove radio frequency interference (RFI), demixed, and averaged as part of standard LOFAR pre-processing. Demixing involves removing the effects of the very bright radio
Scan type Scan code
Central source
Right ascen-sion
Declination Start time (UTC) End time (UTC) Duration (seconds) Calibrator L345916 3C380 18h29m31.8s +48d44m46s 00:18:00.0 00:32:59.2 899.248 Target L345920 G54.1+0.3 19h30m31.0s +18d52m46s 00:34:00.0 03:04:00.5 9000.49 Calibrator L345926 3C380 18h29m31.8s +48d44m46s 03:05:00.0 03:19:59.2 899.248
Table 2.1: Observation information for both of the calibrator scans and the target scan. All observations were taken on 2015/06/12.
Chapter 2: LOFAR Observations and Processing 7
sources, Cassiopeia A and Cygnus A, that affect LOFAR images even when they are far from the phase centre of the FoV. The data were averaged from 64 channels per SB to 4 frequency channels per SB. The LOFAR synthesized beam size is 2.90× 2.30 at 150.8 MHz using a Briggs weight of 1.0. Different weighting schemes and their effect on the synthesised beam are discussed in Section 2.3.
The calibrator data were run through the calibrator pipeline, which performs flagging, demixing, and averaging as well as finding gain calibration solutions. The target data were run through the Multi-frequency Snapshot Sky Survey (MSSS, Heald et al.,2015) imaging pipeline (see Sec.2.2.3.1), where the data were calibrated using a direction-independent method. This is where the calibration solutions from the calibrator scan are directly applied to the target scan. A single SB uncalibrated clean image and the same SB after MSSS calibration are shown in Figure2.1.
Ionospheric variations during the observations were particularly pronounced. This is clear from the clock-TEC plots produced by prefactor1 (see Sec.2.2.1), where TEC is the total electron content of the ionosphere. The LOFAR core stations operate on the same clock, but the remote stations have their own clocks. This means that the relative drift between the core and remote station clocks needs to be corrected. The clock-TEC solutions of the HBA observations are shown in Figure2.2. The solutions should be smooth and clustered around 0. In contrast, one can see in Figure 2.2that the solutions are extremely noisy and have large variations. The ionospheric conditions during these observations thus likely caused some of the calibration problems discussed in Section 2.2.
2.2
Calibration methods
Calibrating radio interferometric data means calibrating the phase and amplitude of the signal as seen by the individual elements in the array (stations in the case of LOFAR). Phase errors occur due to e.g. the station clocks and the ionosphere and result in deconvolution artefacts and an overall increase in noise in the image (van Weeren et al.,2016). Ionospheric phase errors are caused by the column density of electrons in the ionosphere (TEC), which varies in time and is different along different lines of sight and for different stations. LOFAR is a geographically distributed array, which means that phase errors need to be corrected per station. Each station also has a different line of sight and hence experiences different ionospheric variations. This makes phase-calibrating LOFAR observations a challenge. Amplitude calibration is essential because the measured flux density depends on the telescope’s response, which varies with time and pointing direction. Correcting for amplitude means that flux densities measured in the final image are accurate.
1
Figure 2.1: HBA SB160 with a central frequency of 149.9985 MHz and bandwidth of 0.1953 MHz. The top image has been averaged and demixed but not calibrated. The centre image has undergone direction independent calibration as part of the MSSS calibration pipeline The bottom image has been calibrated using the Pre-Facet-Cal (prefactor) pipeline. The synthesised beam is shown in the bottom left corner
Chapter 2: LOFAR Observations and Processing 9
Figure 2.2: Clock-TEC solutions of the HBA observations produced when running prefactor. The colours indicate the different LOFAR stations. The y-axis indicates the ionospheric variations. Time is in
units of time steps in the MSs.
2.2.1 Pre-Facet-Cal
Pre-Facet-Calibration, or prefactor, is a calibration method specifically designed and developed to over-come the unique challenges of calibrating LOFAR HBA observations such as the ionospheric variations across the wide FoV, the clock drift between stations, and the different lines of sight (and hence different ionospheres) of the different stations. Here we will briefly discuss the steps of prefactor; more detail can be found in van Weeren et al.(2016).
The first steps in prefactor involve direction-independent correction. This means removal of RFI, re-moval of off-axis bright sources (demixing), averaging, solving for the calibrator complex gains, clock-TEC separation on the calibrator, transfer of amplitudes and clocks from the calibrator to the target and medium-resolution target field amplitude and phase calibration. RFI removal is performed by AOFlagger (Offringa,
2010). The final calibration on the target amplitude and phase is a self-calibration performed on groups of combined SBs (usually 10 per group).
The prefactor pipeline can be found on the LOFAR GitHub2. To run prefactor the user needs to find or produce the appropriate SkyModel files for their calibrator object and the target field. A SkyModel is a file that contains basic information (eg. right ascension, declination, semi-major and semi-minor axis, orientation, and flux density) about known sources in the field of interest. The user needs to edit the Pre-Facet-Cal parset and the “config” file to match their MSs and their computer or cluster specifications. Prefactor can then be run using a command similar to:
genericpipeline.py Pre-Facet-Cal.parset -c pipeline.cfg -d which was the command we used to run prefactor on the Dragnet3 cluster.
2.2.2 Direction-independent calibration
An alternative calibration method is a custom direction-independent calibration. The steps are:
• Flagging (both the calibrator scan and target scan): – Flagging the “ears”
– Auto flagging of RFI with AOFlagger (Offringa,2010) • Finding the calibrator gaincal solutions
• Applying the calibration solutions from the calibrator scan to the target scan • Flagging the target scan again using AOFlagger
It is important to flag the ears (HBA Core stations are split into HBA0 and HBA1 “ears”, each with 24 HBA tiles, and these have very short baselines between them) to remove the largest scale emission that causes a large-scale wave pattern in the image, an example of which is shown in figure 2.1(centre panel), and produces an incorrect clean model.
The gain calibration solution was found using a solution interval of 2. This means that a gaincal solution is found for every two time intervals. The calibration solutions are then exported using parmexportcal using zerophase=T or zerophase=F. Using parmexportcal automatically produces time-independent gain solutions. Setting zerophase=T sets the phase solution to zero. For most observations it is more appropriate to set zerophase=F. However, setting zerophase=T is useful to reduce the impact of poor ionospheric conditions. The calibration solutions are then applied to the flagged target measurement set.
2.2.3 Calibrating the G54.1+0.3 observation
As described above, we considered four possible calibration methods for our LOFAR observations:
2
https://github.com/lofar-astron/prefactor
3
Chapter 2: LOFAR Observations and Processing 11
Figure 2.3: HBA SB000 after direction-independent MSSS pipeline calibration in the LTA. SB000 has a central frequency of 118.7485 MHz and bandwidth of 0.1953 MHz. The synthesised beam is shown in the
bottom left corner.
• MSSS imaging pipeline • prefactor
• direction-independent, time-dependent, non-zero-phase calibration • direction-independent, time-dependent, zero-phase calibration
we will now compare each of these strategies in order to come to a conclusion on which methods deliver the best-possible calibration of our data.
2.2.3.1 The MSSS imaging pipeline
The data products available on the Long Term Archive (LTA) for this field were the averaged and demixed observations, as well as the averaged and demixed observations processed with the MSSS imaging pipeline. The MSSS imaging pipeline is the calibration pipeline used to calibrate observations for imaging as part of the MSSS survey. The MSSS imaging pipeline uses iterations of direction-independent calibration and self-calibration (Heald et al.,2015). Figure 2.1(centre panel) shows an example of a SB processed using the MSSS imaging pipeline. It appears that the pipeline does a good job of calibration. However, Figure
2.3shows that the pipeline does not achieve good results for all SBs. The MSSS imaging pipeline calibrates the most sensitive SBs well but does not produce reasonable results at the edges of the HBA frequency range.
2.2.3.2 Pre-Facet-Cal
As the ionospheric conditions were poor during these observations, prefactor attempted to calibrate using the clock-TEC solution shown in Figure 2.2. This resulted in prefactor producing poor calibration solutions. The result of the prefactor calibration pipeline on the PWN G54.1 + 0.3 FoV is shown in Figure 2.1(bottom panel).
2.2.3.3 Direction-independent calibration
An example of an SB in our FoV with a direction-independent, time-dependent and non-zero-phase calibra-tion compared to the same SB with a direccalibra-tion-independent, time-dependent and zero-phase calibracalibra-tion is shown in Figure2.4. It is clear that the zero-phase calibration greatly reduces the large-scale noise patterns visible in the non-zero-phase calibration.
2.2.3.4 Chosen calibration method
We found that the calibration method that produced the best results for all SBs was the direction-independent, time-direction-independent, zero-phase method. This method was selected for a combination of rea-sons. The sources in the field have realistic flux densities with maximum flux densities of point sources up to a few Janksy compared to hundreds of Janksy in some SBs from the MSSS imaging pipeline. The objects in the FoV have morphologies as expected from multi-frequency observations, such as with the Very Large Array and the Westerbork Synthesis Radio Telescope (Fig.4.1), unlike the images produced by the MSSS imaging pipeline (Fig.2.3) and the prefactor pipeline (Fig.2.1). The images do not suffer from large scale noise-patterns like the noise in the direction-independent, time-dependent, zero-phase method (Fig.2.4). Using a time-independent, zero-phase calibration is unusual but in this case using this method reduces the calibration problems caused by the poor ionospheric conditions. As such this was the method we used to process the LOFAR observations of the PWN G54.1 + 0.3 field.
A non-standard or custom calibration, such as a direction-independent, time-independent, zero-phase method, can be performed using the LOFAR specific software DPPP (Default Preprocessing Pipeline4) and LOFAR specific instruction files called “parsets”. These instruction files require parameters such as msin and msout to define the MS (or MSs) to perform the function on and the name of the output MS (or MSs). It then requires a steps parameter which is a list describing the functions to be performed in the order they should be performed. For example: steps=[aoflagger, gaincal] would first flag RFI using AOFlagger and then perform gain calibration. Additional requirements for the steps can then be added to the end of the parset. For example if one of the steps is flag then the user can define specific baselines to flag, such as the ears.
4
Chapter 2: LOFAR Observations and Processing 13
Figure 2.4: HBA SB160 that has been calibrated using top: direction-independent, time-independent calibration with phase=F and bottom: direction-independent, time-independent calibration with zero-phase=T. SB160 has a central frequency of 149.9985 MHz and bandwidth of 0.1953 MHz. The synthesised
beams are shown in the bottom left corner of each panel.
To perform the full calibration pipeline we used Python functions. Our pipeline first writes the required parsets and then executes them with DPPP in the correct order. At each step the pipeline writes new MSs instead of over-writing the original MSs. This is to prevent the calibration from altering the original data so that multiple calibration methods could be tested without re-downloading the data. This vastly increases the computer storage required to run the pipeline as each MS is typically on the order of Gigabytes and the pipeline runs on 260 target MSs and 260 calibrator MSs, but the extra MSs are deleted after the pipeline is completed. We produce basic cleans (using WSClean, see Sec.2.3) of the MSs produced at each step in order to make de-bugging simpler and to enable the user to check the products easily by eye at each step. The functions used to perform the calibration and the function to run the full pipeline are shown in Appendix A.
performed as the final calibration steps, the first round based on a model from the TGSS ADR5 catalogue (Intema et al.,2017).
2.3
Imaging
LOFAR imaging is performed using the WSClean tool (Offringa et al.,2014). WSClean uses a w-stacking clean algorithm to image interferometric radio data. WSClean has a wide range of parameters and options, here we will only discuss those used for the analysis in this work.
The basic WSClean parameters are scale, size, trim, niter and threshold. The scale parameter sets the size of the pixels in arcseconds or arcminutes. The size parameter sets the height and width of the image in pixels during cleaning which can then be trimmed to a height and width set by the trim parameter. The niter parameter stands for “number of iterations” and defines the number of major clean cycles. As the name suggests, the threshold parameter sets a threshold for how deep the clean is. If both niter and threshold are used then the clean will stop when the maximum number of major clean iterations is reached or when the threshold is reached, whichever is reached first.
The multiscale parameter is useful for observations containing extended emission. The difference between a clean without multiscale and a clean with multiscale is shown in Figure 2.5. A clean without multiscale uses simple delta functions to model the emission. Including the multiscale parameter means that WSClean uses delta functions and Gaussian functions to model the emission. This produces a much more accurate model.
In WSClean the weight parameter can be used to vary the weighting of different baselines. This is done using Briggs weighting, where a Briggs weight of −1.0 down weights the short baselines and up-weights the long baselines while a Briggs weight of 1.0 does the opposite. This means that a Briggs weight of −1.0 increases the resolution of an image, but increases the noise. Briggs −1.0 also resolves out extended emission. Briggs 1.0 decreases the resolution and noise of an image, and is better for observing extended emission. As we are interested in extended emission more than point source resolution, we opt for a Briggs weight of 1.0. The point spread functions (PSFs) of the same LOFAR image with three different Briggs weightings are shown in Figure2.6.
In an observation where extended emission dominates the short baselines can produce large scale noise. To prevent this it is possible to cut the shortest baselines during the clean using minuv-l. This can be used to remove the shortest baselines, and hence the largest scale emission, but also removes a large amount of noise and allows a deeper clean. The difference between a clean with a uv-cut and a clean without a uv-cut is shown in Figure 2.7. We use a uv-cut of 100 λ (where λ is wavelength) in all of our cleans to remove the large scale noise and achieve deeper cleans. Using this uv-cut the largest scale emission we can detect is ∼ 340.
5
Chapter 2: LOFAR Observations and Processing 15
Figure 2.5: Comparison of clean models without multiscale (top panel) and with multiscale (bottom panel). Both of these cleans were performed on the same HBA SB with a central frequency of 150.8 MHz. The image colour scale has been saturated at 50 mJy beam−1. The synthesised beams are shown in the
bottom left of each panel.
Figure 2.6: The PSFs of a multi-frequency clean of an HBA observation with a central frequency of 144 MHz. The panels have Briggs weightings from left to right of −1.0, 0.0, and 1.0. The synthesised
Figure 2.7: Comparison of a clean without a uv-cut (top panel) and with a uv-cut of 100 λ (bottom panel). Both of these cleans were performed on the same HBA SB with a central frequency of 150.8 MHz.
The synthesised beams are shown in the bottom left corner of each panel.
2.4
Flux determination
The point source flux densities were measured using pyBDSF6. pyBDSF is a LOFAR-specific program devel-oped at the Netherlands Institute for Radio Astronomy (ASTRON) that can be used to measure the flux densities of sources in Flexible Image Transport (FITS) files.
In order to check the flux densities in the FITS files we produced from the LOFAR HBA observations we used pyBDSF to measure the flux densities of point sources and compared the flux densities to those measured by the TGSS ADR, VLSSR7, and WSRT8 point source databases. We wrote a script to run pyBDSF on the FITS files we produced. The WSRT and VLSSR survey catalogues were in fixed-width column files with units and formatting inconsistent with TGSS and pyBDSF . We wrote a script to extract and convert the information and rewrite the appropriate data to new comma separated value (CSV) files.
6
http://www.astron.nl/citt/pybdsm/
7
VLA Low-Frequency Sky Survey Redux Source Catalogue, https://heasarc.gsfc.nasa.gov/W3Browse/ radio-catalog/vlssr.html
8
Chapter 2: LOFAR Observations and Processing 17
We then wrote a script to extract the TGSS ADR flux densities from the database. Finally, we wrote a script to extract the pyBDSF measured flux densities from the CSV files produced by pyBDSF. We could then compare the point source flux densities from pyBDSF to those from TGSS ADR, VLSSR and WSRT. The scripts discussed here can be found in Appendix B.
When comparing the point source flux densities we found that pyBDSF often measured higher flux densities than expected. Due to the larger beam size of the LOFAR images often pyBDSF would be measuring flux densities from multiple point sources where TGSS ADR would be measuring the flux density of one source. We edited our scripts to produce SAODS99 region files for the pyBDSF point sources and TGSS ADR point sources in order to check that the pyBDSF and TGSS ADR point sources matched. We also excluded sources in the pyBDSF files that had semi-major axes greater than 0.1◦ to restrict the results to point-like sources. We excluded any TGSS ADR sources that were within 0.1◦ of another TGSS ADR source to prevent confusion.
When pyBDSF finds a point source in different SBs it does not always define the point source as the same size in all of the SBs. If pyBDSF places a much larger ellipse around a source in one SB compared to another SB this can dramatically affect the measured flux density. As such, we altered our pyBDSF script to use the aperture parameter. Using the aperture parameter we set pyBDSF to measure the flux density of a point source only within an aperture of set size centred on the brightest pixel of the point source. We set the aperture radius to be the semi-major axis of the SB’s synthesised beam. This significantly reduced the scatter on the flux density measurements.
In Figure 2.8 the radio spectra of three points sources in the FoV are shown. The LOFAR HBA flux density measurements are not consistent with the flux density measurements from VLSSR, TGSS ADR and WSRT. We tested the effects of a different Briggs weightings in case this could affect the flux density measurements. Figure2.8shows that the Briggs weighting does not strongly affect the flux measurement. We also combined the FITS files into chunks of six images using Astropy10, but this did not significantly affect the measurements.
As the point sources are in the Galactic Plane, the steepening of the spectra may be due to the Galactic synchrotron background. We found the brightness temperature in the directions of the point sources using 408 MHz observations by Haslam (1983). The code to search the Haslam (1983) tables was written by David Gardenier11. We then converted the brightness temperature into flux density using:
Sν =
θ2
1.36 × 103· λ2 · TB (2.1)
where Sν is the flux density in Jy beam−1, θ is the semi-major axis of the synthesised beam in arcseconds,
λ is the wavelength in cm and TB is the brightness temperature in Kelvin. Figure 2.8 shows that the
9
http://ds9.si.edu/site/Home.html
10http://www.astropy.org/ 11
Figure 2.8: Radio spectra of isolated point sources in the LOFAR HBA FoV using two different Briggs weightings and using averaged FITS files. Flux density measurements from VLSSR, TGSS ADR and WSRT are also plotted. These point sources have J2000 coordinates in degrees of (290.7, 18.8), (290.3, 17.0),
and (293.4, 21.6) respectively.
background brightness temperature does not contribute significantly to the flux densities of the point sources.
It is therefore likely that the steepening effect in the LOFAR measurements is due to an instrumental or calibration effect. A possible cause that we have yet to test is the flux of the very large scale emission being added into the point sources. If the issue, this is likely to have occurred during the calibration. If there is large scale emission that is not accounted for in the sky model then the flux from that emission may be added into the point sources that are in the sky model. This could artificially increase the flux density of the point sources. To mitigate this problem in the calibration step a uv-cut could be applied to resolve out the large scale emission. A uv-cut could be applied that specifically cuts baselines such that the observable largest scale emission is exactly the size of the object of interest, in this case PWN G54.1 + 0.3. It would then be important at the clean stage to apply the same uv-cut and ensure that the correct Briggs weighting is applied to produce a synthesised beam approximately the size of PWN G54.1 + 0.3. This is future work for this project and could explain the unexpectedly steep spectra of the point sources.
Chapter 3
PWN G54.1+0.3
In this chapter we will discuss PWN G54.1 + 0.3. PWN G54.1 + 0.3 was the main source of interest that led to the proposal for and acquisition of the LOFAR observations analysed in this project.
3.1
Past work
3.1.1 Discovery
PWN G54.1 + 0.3, or the “Bull’s Eye” PWN, was first observed by Altenhoff et al.(1979), Downes et al.
(1980), and Reich et al. (1984) as a small diameter source in radio Galactic Plane surveys. Green(1985) identified it as a small, diffuse, apparently thermal source, shown in Figure 3.1, and noted that it may be an HII region or a filled-centre SNR. Reich et al.(1985) identified PWN G54.1 + 0.3 as a Crab-like, filled-centre SNR with a flat spectrum and significant polarisation. Velusamy et al.(1986) observed the region surrounding PWN G54.1 + 0.3 at 612 MHz using the Westerbork Synthesis Radio Telescope (WSRT) and they returned again to the idea that PWN G54.1 + 0.3 is an HII region associated with the large HII region centred at G053.935 + 0.228 (Anderson et al.,2014).
In order to confirm whether PWN G54.1 + 0.3 is an HII region or a SNR, Velusamy & Becker (1988) performed high resolution 1.4, 1.6, and 4.8 GHz Very Large Array (VLA) observations and measured the flux density at 327 MHz with the Ooty Synthesis Radio Telescope (OSRT). They also analysed Infrared Astronomical Satellite (IRAS) data. Velusamy & Becker (1988) used flux density measurements from
Green (1985), Velusamy et al. (1986), Caswell (1985) and Reich et al. (1985) to complement their flux densities and produce the SED in Figure3.2. PWN G54.1 + 0.3 was confirmed as a Crab-like, filled-centre SNR by Velusamy & Becker (1988) using high resolution VLA and OSRT observations to show that the spectrum is that of a non-thermal synchrotron emitting object.
Figure 3.1: Cambridge 5-km telescope observation of PWN G54.1 + 0.3 byGreen (1985). This obser-vation has a resolution of 700× 2000and an observing frequency of 2.7 GHz.
3.1.2 Observations
3.1.2.1 Low-frequency observations
PWN G54.1 + 0.3 was first discovered in the radio and has since been observed by various radio telescopes. Information regarding the observations and flux density measurements in the radio can be found in Table
Chapter 3: Pulsar Wind Nebula G54.1+0.3 21
Figure 3.2: Low-frequency spectral energy distribution of PWN G54.1 + 0.3 produced by Velusamy & Becker(1988). The power law has a spectral index of α ∼ −0.13 confirming that PWN G54.1 + 0.3 has
a spectrum similar to that of the Crab Nebula.
Source Telescope Frequency
(GHz)
Flux density (mJy)
Altenhoff et al. (1979) Effelsberg 4.875 400 Downes et al. (1980) Effelsberg 5 300 Reich et al. (1984) Effelsberg 2.7 580 Green et al. (1985) Cambridge Low-Frequency
Syn-thesis Telescope
0.151 0.55 ± 0.15
Caswell et al. (1985) Penticton Synthesis Telescope 1.4 364 Velusamy et al (1986) WSRT 0.612 430 ± 30 Velusamy et al (1988) VLA (C-configuration) 1.4 478 ± 30 Velusamy et al (1988) VLA(C-configuration) 1.6 417 ± 30 Velusamy et al (1988) VLA (C-configuration) 4.8 325 ± 20 Velusamy et al (1988) Ooty Synthesis Radio Telescope 0.327 495 ± 75 Taylor et al. (1996) WSRT 0.327 504 ± 17 Lang et al (2010) VLA 1.4 433.0 ± 30.0 Lang et al (2010) VLA 4.7 327.0 ± 25.0 Lang et al (2010) VLA 8.5 252.0 ± 20.0
Figure 3.3: Chandra ACIS-S X-ray (0.5–10 keV) intensity map of PWN G54.1 + 0.3 (Lu et al., 2002). The intensity is plotted from 7.43 × 10−5 to 5.95 × 10−3 counts cm−2s−1arcmin−2. The inset image shows a closer view of the central region of the PWN containing the point source that is the central pulsar
and the torus of the wind termination shock.
3.1.2.2 X-ray observations
Weak X-ray emission from PWN G54.1 + 0.3 was detected by Seward (1989) using the Einstein IPC instrument. Lu et al. (2001a) discovered the X-ray jet associated with PWN G54.1 + 0.3 using ROSAT and ASCA SIS observations. Lu et al. (2001b) also measured the X-ray flux density of PWN G54.1 + 0.3. Chandra observations of PWN G54.1 + 0.3 were taken in 2001 (Lu & Wang, 2001) with further details published in 2002 (Lu et al., 2002). One can see the X-ray morphology of PWN G54.1 + 0.3 in the intensity map in Figure3.3. PWN G54.1 + 0.3 was observed in the hard X-rays by the INTEGRAL space telescope (Krivonos et al.,2017). A summary of the X-ray observations can be found in Table 3.2.
Chapter 3: Pulsar Wind Nebula G54.1+0.3 23
Source Telescope Energy (keV)
Flux (erg cm−2s−1) Photon index (Γ)
Lu at al. (2001) ASCA SIS 0.7 − 10.0 1.14 × 10−11 2.9 ± 0.2 Lu et al. (2002) Chandra 2 − 10 (5.43 ± 0.035) × 10−12 2.09 ± 0.01 Krivonos et al (2017) INTEGRAL 17 − 60 (0.78 ± 0.10) × 10−11
Bocchino et al (2010) SUZAKU 2 − 10 (4.7 ± 0.7) × 10−12
Table 3.2: Archival X-ray observations of PWN G54.1 + 0.3.
Source Telescope Frequency (THz)
Flux density (Jy)
Extinction cor-rected flux den-sity (Jy) Taylor et al. (1996) IRAS 25 1.4
Taylor et al. (1996) IRAS 12 9.5 Taylor et al. (1996) IRAS 5 61.1 Taylor et al. (1996) IRAS 3 65.9
Temim et al. (2010) IRAC 51.7 0.2 0.3 Temim et al. (2010) IRAC 37.5 0.6 1
Temim et al. (2010) IRAC 12.5 3.8 ± 0.4 6.5 ± 0.7 Temim et al. (2010) IRAC 4.285 76 ± 15
Temim et al. (2017) Akari IRC 19.2 2.6 ± 0.26 Temim et al. (2017) SOFIA FORCAST 15.2 11.9 ± 2.4 Temim et al. (2017) Spitzer MIPS 12.5 23.1 ± 2.1 Temim et al. (2017) SOFIA FORCAST 11.9 12.7 ± 2.5 Temim et al. (2017) SOFIA FORCAST 9.52 42.5 ± 8.5 Temim et al. (2017) SOFIA FORCAST 8.62 41.7 ± 8.3 Temim et al. (2017) Herschel PACS 4.285 87.9 ± 11.4 Temim et al. (2017) Herschel PACS 3.0 68.8 ± 13.4 Temim et al. (2017) Herschel PACS 1.88 29.0 ± 14.9 Temim et al. (2017) Herschel SPIRE 1.2 6.9 ± 5.2 Temim et al. (2017) Herschel SPIRE 0.857 1.6 ± 2.8 Temim et al. (2017) Herschel SPIRE 0.60 0.4 ± 1.1
Table 3.3: Archival IR flux density measurements of PWN G54.1 + 0.3.
3.1.2.3 Infrared observations
As PWN G54.1 + 0.3 is in the Galactic Plane it has been observed as part of various IR Galactic Plane surveys. The observations have been summarised in Table 3.3. The IR emission will be further discussed in Section 3.2.1.
3.1.2.4 Gamma-ray observations
PWN G54.1+0.3 has been observed at high-energies by the High-Energy-Gamma-Ray Astronomy (HEGRA) telescope and Very Energetic Radiation Imaging Telescope Array System (VERITAS). These observations
Source Telescope Energy (GeV)
Photon density (photons cm−2s−1TeV−1) Acciari et al. (2010) VERITAS 311 (1.10 ± 0.56) × 10−11 Acciari et al. (2010) VERITAS 492 (4.2 ± 1.4) × 10−12 Acciari et al. (2010) VERITAS 780 (1.12 ± 0.45) × 10−12 Acciari et al. (2010) VERITAS 1.2 × 103 (6.2 ± 1.7) × 10−13 Acciari et al. (2010) VERITAS 3 × 103 (3.9 ± 2.1) × 10−14
Table 3.4: Archival gamma-ray observations of PWN G54.1 + 0.3.
are summarised in Table 3.4. The HEGRA measurement is an upper limit and can be found inAharonian et al. (2002). There is no associated Fermi source in the Fermi LAT 4-Year Point Source Catalogue.
3.1.3 The distance to PWN G54.1+0.3
The distance to PWN G54.1 + 0.3 has been approximated using a variety of methods. Lu et al. (2002) estimated a distance of ≈ 5 kpc using Galactic absorption. Camilo et al.(2002) used the DM of the central pulsar, PSR J1930 + 1852 (see Section 3.1.5), and NE2001 (Cordes & Lazio, 2002) to give a distance of ≤ 8 kpc. Weisberg et al. (2008) use Arecibo HI absorption measurements to put the pulsar between 3.2 kpc and 10 kpc.
Leahy et al.(2008) find a distance of 6.2 ± 0.1 kpc using a morphological association with a CO molecular cloud. LaterLee et al. (2012) discount this association and the distance to PWN G54.1 + 0.3 calculated using the CO cloud.
Using the photometric distance to the IR-excess objects (discussed in Section 3.2.1.1) Kim et al. (2013) find an approximate distance to PWN G54.1 + 0.3 of 6.0 ± 0.4 kpc.
3.1.4 Magnetic Fields and Polarisation
Lang et al. (2010) used the VLA to investigate the polarisation and magnetic field properties of PWN G54.1 + 0.3. They estimate the upper limit on the magnetic field strength to be ≈ 1250 µG.
3.1.5 PSR J1930+1852: the central pulsar of PWN G54.1+0.3
PSR J1930 + 1852, the pulsar associated with PWN G54.1 + 0.3, was discovered byCamilo et al.(2002). They found the pulsar using 1180 MHz Arecibo observations and subsequently also found the pulsar in archival ASCA X-ray observations. A summary of the pulsar parameters observed by Camilo et al.(2002) are shown in Table 3.5 while the derived parameters are shown in Table3.6.
Chapter 3: Pulsar Wind Nebula G54.1+0.3 25
Parameter Value - Camilo et al. (2002) Value - Lu et al. (2007) Right ascension (J2000) 19h30m30.13s Declination (J2000) +18◦52014.1” Period, P (ms) 136.855046957 (9) 136.871312 (4) Period derivative, ˙P (s s−1) 7.5057(1) × 10−13 7.5112(6) × 10−13 Epoch (MJD [TBD]) 52280.0 52530 Dispersion measure, DM (cm−3 pc) 308(4) Flux density at 1180 MHz (µJy) 60 ± 10
Flux at 2 − 10 keV (erg cm−2s−1) 1.7 × 10−12 1.2 × 10−12 (pulsed) Pulse FWHM at 1180 MHz (ms) 15 ± 2
Pulse FWHM at 2 − 10 keV (ms) ∼ 25
Table 3.5: The observed parameters of PSR J1930 + 1852. The Lu et al. (2007) values are from the 12th September 2002 Rossi X-Ray Timing Explorer observations.
Parameter Value
Distance, d (kpc) ≤ 8 Characteristic age, τc(yr) 2900
Spin-down luminosity, ˙E (erg s−1) 1.2 × 1037 Magnetic field strength, B (G) 1.0 × 1013 Pseudo luminosity at 1400 MHz,
(mJy kpc2)
∼ 1
Table 3.6: The derived properties of PSR J1930 + 1852 by Camilo et al.(2002).
Kaplan & Moon (2006) performed a search for near-infrared (NIR) counterparts to young pulsars using Persson’s Auxiliary Nasmyth Infrared Camera (PANIC) on the Magellan I (Baade) telescope and the Near-IR Camera (NIRC) on the 10 m Keck I telescope. They did not detect a NIR counterpart to PSR J1930+1852 but could define the upper limits:
J − band : < 0.17 mJy (3.1) K − band : < 0.03 MJy (3.2)
Ilardo (2006) searched for, but did not detect, giant pulses from PSR J1930 + 1852. Chevalier (2005) calculated the initial spin period of the pulsar to be 100 ms with a surface magnetic field of 1 × 1013G.
X-ray timing of PSR J1930+1852 using Rossi X-Ray Timing Explorer observations (RXTE) were performed by Lu et al. (2007). They found a best-fit constant spin-down rate of ˙P = 7.5112 (6) × 10−13s s−1. However, they note that there is evolution in the period, shown in Table 3.7, suggesting strong timing noise or glitch activity.
Date (UT) Epoch (MJD [TBD]) Period (s) 1997.04.27 50566 0.13674374 (5)* 2002.01.17 52280 0.136855046957 (9)* 2002.09.12 52530 0.136871313 (4) 2002.12.23 52632 0.136877919 (3) 2003.06.30 52820 0.136890130 (5)
Table 3.7: Observed epochs of J1930 + 1852 fromLu et al.(2007). Periods denoted by a * were taken fromCamilo et al.(2002).
Figure 3.4: A schematic diagram of the basic structure of a shell-type SNR.
3.2
Supernova remnant features of PWN G54.1+0.3
Theoretical predictions suggest that CCSNe can efficiently produce between 0.1 and 1.0 MJ of dust.This
is important because it is thought that CCSNe could be responsible for some of the dust in the Universe, particularly the large amounts of dust observed in some early galaxies (De Looze et al.,2017). However, there are only three known SNRs with a significant amount of supernova ejecta dust: Cassiopeia A (Cas A), SN 1987A, and the Crab.
The structure of a typical shell-type SNR is shown in Figure 3.4. Of the 294 known Galactic SNRs in the latest Green’s Catalogue (Green, 2014) 234 are considered to be shell-type. The basic features of SNRs were discussed in Section 1.1.
Filled-centre (FC) SNRs, or plerions, are SNRs similar to the Crab. FC SNRs have a flat, non-thermal radio spectral index and are centrally bright in X-rays. They lack a large-scale bright shell (the forward shock). It is usually assumed that these FC remnants are powered by a central pulsar (Wallace et al.,1997). Only
Chapter 3: Pulsar Wind Nebula G54.1+0.3 27
9 of the 294 known Galactic SNRs are classified as FC (Green,2014). A schematic of an FC remnant with a PWN is shown in Figure 1.3. If an FC remnant has a shell it is called a composite SNR.
Both Cas A and SN 1987A are shell-type SNRs. Cas A, shown in Figure1.2, is ≈ 330 years old and recent modelling suggests the presence of 0.3 − 0.5 MJ of SN ejecta dust. SN 1987A is, as the name suggests,
30 years old. Matsuura et al.(2015) suggest a dust mass of up to 0.8 MJ within SN 1987A. However, as
young shell-type SNRs both Cas A and SN 1987A have forward and reverse shocks. Simulations find that between 1 and 80% of the dust is likely to be destroyed by the reverse shock. For both of these SNRs at least 25% of the dust must survive to fit the theoretical model of CCSNe producing 0.1 − 1.0 MJ of dust.
The Crab, however, does not have a forward or reverse shock. It is an FC SNR and contains between 0.11 and 0.48 MJ of dust (Owen & Barlow,2015).
Currently, PWN G54.1 + 0.3 is classified as an FC SNR (Green, 2014). Due to the similar morphology of their PWNe, PWN G54.1 + 0.3 is considered a close-cousin of the Crab Nebula. As we will discuss below, PWN G54.1 + 0.3 fits in with Cas A, SN 1987A and the Crab as a dusty SNR as it is surrounded by a bubble of cold SN ejecta dust. However, it is still an open question as to whether PWN G54.1 + 0.3 possesses an SNR shell or not.
3.2.1 The infrared dust bubble around PWN G54.1+0.3
An IR bubble of dust surrounding PWN G54.1 + 0.3 was discovered by Koo et al. (2008) using the Akari (ASTRO-F) satellite. A Spitzer observation of the IR bubble is shown in Figure 3.5. The X-ray emission fills the IR cavity, strongly suggesting an association. Until recently there were two theories regarding the origin of this loop: a star-forming region containing young stellar objects (YSOs) and SN ejecta dust heated by late O- and B-type stars.
3.2.1.1 IR-excess stellar objects
The IR observations of PWN G54.1 + 0.3 reveal point-like sources in the IR observations. These point-like sources were first classified as YSOs by Koo et al. (2008) using Spitzer and Akari observations. Later, they were classified byTemim et al.(2010) as early-type stars. Kim et al.(2013) investigate the IR-excess objects further and find them to be late O- and early B-type main-sequence stars between O8 and B3.
3.2.1.2 A star-forming region
It was suggested byKoo et al.(2008) that the IR dust bubble is a star-forming region with embedded YSOs. They used Spitzer and Akari IR observations to identify at least 11 YSOs in the dust shell and conclude
