To appear in ApJ Letters First Spectroscopic Imaging Observations of the Sun at Low Radio Frequencies with the Murchison Widefield Array Prototype 1 1 Divya Oberoi1, Lynn D. Matthews1, Iver H. Cairns2, David Emrich3, Vasili Lobzin2, Colin 0 1 4 5 5 3 2 J. Lonsdale , Edward H. Morgan , T. Prabu , Harish Vedantham , Randall B. Wayth , 6 4 7 8 3 n Andrew Williams , Christopher Williams , Stephen M. White G. Allen , Wayne Arcus , a 9 1 10 11 David Barnes , Leonid Benkevitch , Gianni Bernardi , Judd D. Bowman , Frank H. J 12 8 13 1 14 3 Briggs , John D. Bunton , Steve Burns , Roger C. Cappallo , M. A. Clark , Brian E. Corey1, M. Dawson12, David DeBoer8,15, A. De Gans12, Ludi deSouza8, Mark Derome1, R. ] R G. Edgar14,16, T. Elton8, Robert Goeke4, M. R. Gopalakrishna5, Lincoln J. Greenhill10, S 17 3 4 5 18 Bryna Hazelton , David Herne , Jacqueline N. Hewitt , P. A. Kamini , David L. Kaplan , . h p Justin C. Kasper10, Rachel Kennedy1,15, Barton B. Kincaid1, Jonathan Kocz12, R. Koeing8, - Errol Kowald12, Mervyn J. Lynch3, S. Madhavi5, Stephen R. McWhirter1, Daniel A. o r Mitchell10, Miguel F. Morales17, A. Ng8, Stephen M. Ord10, Joseph Pathikulangara8, Alan t as E. E. Rogers1, Anish Roshi,5,19, Joseph E. Salah1, Robert J. Sault20, Antony Schinckel8, N. [ 5 5 8 5 2 Udaya Shankar , K. S. Srivani , Jamie Stevens , Ravi Subrahmanyan , D. Thakkar , 1 Steven J. Tingay3, J. Tuthill8, Annino Vaccarella12, Mark Waterson3,12, Rachel L. v 0 Webster20 and Alan R. Whitney1 2 6 0 . 1 0 1 1 : v i X r a – 2 – ABSTRACT We present the first spectroscopic images of solar radio transients from the prototype for the Murchison Widefield Array (MWA), observed on 2010 March 27. Our observations span the instantaneous frequency band 170.9– 201.6 MHz. Though our observing period is characterized as a period of ‘low’ to ‘medium’ activity, one broadband emission feature and numerous short-lived, narrowband, non-thermal emission features are evident. Our data represent a significant advance in low radio frequency solar imaging, enabling us to follow the spatial, spectral, and temporal evolution of events simultaneously and in unprecedented detail. The rich variety of features seen here reaffirms the coro- nal diagnostic capability of low radio frequency emission and provides an early 1MIT Haystack Observatory, Westford, MA USA 2University of Sydney, Sydney, Australia 3Curtin University, Perth, Australia 4MIT Kavli Institute for Astrophysics and Space Research, Cambridge, MA USA 5 Raman Research Institute, Bangalore,India 6The University of Western Australia, Perth, Australia 7Air Force Research Laboratory,Kirtland, NM USA 8CSIRO Astronomy and Space Science, Australia 9Swinburne University of Technology, Melbourne, Australia 10Harvard-Smithsonian Center for Astrophysics, Cambridge, MA USA 11School of Earth and Space Exploration, Arizona State University, Tempe, AZ USA 12The Australian National University, Canberra, Australia 13Burns Industries, Inc. Nashua, NH USA 14 Harvard University, Cambridge, MA USA 15University of California, Berkeley, CA USA 16Massachusetts General Hospital, Boston, MA USA 17University of Washington, Seattle, WA USA 18University of Wisconsin - Milwaukee, Milwaukee, WI USA 19National Radio Astronomy Observatory,Green Bank, WV USA 20The University of Melbourne, Melbourne, Australia – 3 – glimpse of the nature of radio observations that will become available as the next generation of low frequency radio interferometers come on-line over the next few years. Subject headings: Sun: corona — Sun: radio radiation — radiation mechanisms: non-thermal — instrumentation: interferometers 1. Introduction Low radio frequency (ν . 300 MHz) emission provides powerful diagnostics of the so- lar corona. However, high fidelity solar imaging at low frequencies is challenging. Coronal emission features are complex and dynamic, evolving rapidly in space, time, and frequency. Consequently, the limited instantaneous spatial and frequency coverage provided by cur- rent radio interferometers have been inadequate to simultaneously resolve transient solar phenomena spatially, temporally, and spectrally. This situation should change dramatically in the next few years as a new generation of low-frequency radio arrays becomes available, leveraging recent advances in digital signal processing hardware and computational capacity. The Murchison Widefield Array (MWA) (Lonsdale et al. 2009) will be one such array, with most of its 512 elements spread over 1.5 km and a few outliers out to 3 km. The resulting dense instantaneous monochromatic uv coverage will provide a radio imaging capability with unprecedented fidelity and flexibility. The MWA iscurrently under construction at theMurchison RadioastronomyObservatory, in the remote and radio-quiet Western Australian outback, and a prototype system comprising 32 interferometer elements (tiles) is operational on site. This system, hereafter referred to as the “32T”, serves as an engineering testbed and provides early science opportunities, in advance of the full MWA. We present here solar imaging observations obtained with the 32T. The quiescent solar emission in the MWA frequency range of 80–300 MHz is dominated by coronal emission from heights of ∼1–10 R above the photosphere. These are among the ⊙ first high-fidelity, high dynamic-range, spectroscopic images of the Sun with a goodtemporal and spectral resolution at meter wavelengths. 2. Data The 32T tiles are arranged along the arms of a randomized Reuleaux triangle (see Cohanim et al. 2004), providing a fairly uniformly sampled uv plane with baseline lengths up to 350 m (Fig. 1). For solar observations, the correlated flux is dominated by the Sun, – 4 – reducing the effective field-of-view (FoV) to ∼1◦, for which the uv sampling provided by the 1 32T exceeds the Nyquist criterion . This, together with a high signal-to-noise ratio, permits robust, high fidelity imaging (Bracewell & Roberts 1954). Initsusualinterferometricmodeofoperation,the32Tprovidesauto-andcross-correlations (visibilities) for all 64 input signals (32 tiles×2 linear polarizations). The entire RF band is directly sampled and filtered into 1.28 MHz wide coarse channels. Twenty four of these coarse channels are further processed by the correlator, which can currently provide a time resolution of 50 ms with a 50% duty cycle over a 30.72 MHz band and a spectral resolution of 40 kHz. The 32T design closely follows the MWA architecture described in Lonsdale et al. (2009). The data presented here were obtained on 2010 March 27 from 04:24:53 to 04:34:03 UT and span the range 170.8–201.6 MHz. They were smoothed to a time resolution of 1 sec- ond before further processing. Data editing, calibration, and imaging were performed using the Astronomical Image Processing System. Self-calibration, to solve for frequency inde- pendent complex gains, was performed individually for each of the coarse channels using a 10-second interval where the Sun was in a relatively quiescent state. These gain solutions were then applied to the full time interval. No absolute flux calibration was possible, owing to the lack of observations of a suitable calibrator. The amplitude part of the bandpass was calibrated using the total power spectra from the same quiescent time interval as used for self-calibration. As will become evident later, the choice of a spectral index only im- pacts the underlying spectral slope. We assumed a spectral index α=2.6 for the quiet Sun (S ∝ να where S is the flux density and ν is the observing frequency); this choice is based ν ν on 32T observations during the recent deep solar minimum (2008 November), which yielded α = 2.6±0.4, a value consistent with Erickson et al. (1977). All the data presented here cor- respond to the east-west (XX) polarization. The wide FoV MWA tiles are expected to have significant (but stable) cross-polarization leakage, and an absolute polarization calibration was not attempted. Imaging and deconvolution were performed using the standard CLEAN algorithm with robust (R=0) weighting, resulting in a synthesized beam at the band center of 961′′ ×796′′ (FWHM). For the images presented here, the data were averaged over a single coarse channel during the gridding process and restored with a circular beam with FWHM 800′′. The edges of the coarse channels could not be calibrated satisfactorily, hence the first six and the last four spectral channels of each coarse channel were flagged during imaging. 1δu ∼ 1/2θ0, where δu is length scale in the uv plane beyond which visibilities are no longer correlated, and θ0 is the size of the FoV in radians – 5 – 200 150 100 h) 50 gt n wavele 0 v ( -50 -100 -150 Nyquist cell size corresponding to a FoV of 1 deg -200 200 150 100 50 0 -50 -100 -150 -200 u (wavelength) Fig. 1.— The instantaneous uv coverage for a single coarse channel centered at 186.2 MHz. The circle indicates the size of a uv-cell, corresponding to the Nyquist sampling criterion for a 1 degree FoV. – 6 – Fig. 2.— Observed visibilities as a function of frequency (XX polarization). The left panels show amplitude (arbitrary units), and the right panels phase (in degrees). The top and bottom rows correspond to baselines with projected lengths of ∼40λ and ∼196λ, respectively, at the band center (186.26 MHz). The frequency (x-axis) ranges from 170.8 to 201.6 MHz; the time (y-axis) spans 550 s starting at 04:24:53 UT on 2010 March 27. The horizontal bar below each panel shows the color coded scale and the dark vertical streaks arise from flagging of the coarse channel edges (Sec. 2). – 7 – 3. Analysis and Results Amplitudes and phases observed on a representative short (∼64 m) and long (∼315 m) projectedbaselineareshowninFig.2. Baselinesofdifferentlengthsandorientationsmeasure independent Fourier components of the source structure. The examples shown here illustrate several key features of the data. Only one polarization is shown as both polarizations show very similar behavior. The most prominent feature in Fig. 2, visually obvious on all baselines in both amplitude and phase, spans the entire observing band around 04:30:10 UT. We hereafter refer to this as the “broadband” feature. Its properties are similar to a weak type III burst (Sec. 4). In addition, a large number of shorter-lived, narrowband features are evident in the visibility amplitudes. Many of these amplitude variations are accompanied by corresponding variations in phase, implying a change in the brightness temperature (T ) B distribution in the corona. These features become more numerous and prominent visually with increasing baseline length and are referred to as “narrowband” features hereafter. A dearth of truly “quiescent” periods is also evident, and the ubiquitous modest variations seen in the spectral structure are solar, not instrumental, in origin. To further illustrate the diversity in the types of emission seen in the short span of data presented here, as well as its spectrally complex and highly time-variable behavior, Fig. 3 shows several series of autocorrelation (total power) spectra. The bottom panel shows one of the few quiescent intervals; the baseline spectral slope seen here is due to the spectral index of the quiet Sun, dominated by thermal emission. An amplitude variation of ∼5% over a 10 s interval is seen even in “featureless” parts of the data. The middle panel shows a series of spectra illustrating one of the numerous “narrowband” features seen in Fig. 2. These features show remarkably rapid evolution in spectral shape and intensity. They typically span ∼5–10 MHz in bandwidth, outside of which the spectral flux density returns to that of the quiet Sun. Their peak flux density approaches ∼1.5 times the quiescent solar flux density at that frequency. Finally, the top panel of Fig. 3 shows the dynamic behavior of the “broadband” feature around 04:30:10 UT. Its bandwidth exceeds the ∼30 MHz observing bandwidth, and at its peak, its flux density is ∼2.5 times the quiescent solar flux density. The abrupt changes from one second to the next imply that the physical changes leading to the production of this emission are temporally undersampled. The datapresented here permit a highfidelity imageof theSunforevery individual time and frequency slice (i.e. every pixel on the dynamic spectra shown in Fig. 2). Fig. 4 shows a collage of images that illustrates the rapid evolution of the appearance of the solar corona during our observations. Although the spatial resolution of the 32T is limited, the fidelity and dynamic range of these images is unprecedented at these frequencies; the dynamic range of ∼2500 exceeds that of earlier images by about an order of magnitude (cf. Mercier et al. – 8 – Fig. 3.— Sample autocorrelation spectra from tile 6, X polarization. Top panel: a 14 s period bracketing the most prominent “broadband” feature seen in Fig. 2. Middle panel: variations seen across 10 s for one of the numerous shorter-lived “narrowband” intensity enhancements. Bottom panel: a 10 second interval exhibiting little intensity variation. The inset in the bottom panel zooms in on a part of the band. As with Fig. 2, the data gaps reflect flagged data. – 9 – Fig. 4.— Top and middle rows: sets of images, 1 s apart, from the “broadband” and one of the numerous “narrowband” features, respectively, highlighting the rapid temporal variation in the emission morphology. Animations including these panels are available in the on-line journal. Bottom row: A 32T radio image of the quiescent Sun (left) and a 304˚A image from SOHO/EIT (right), taken a few hours earlier (01:19 UT). The bright region in the northeastern quadrant is the region 11057. The red circle on the 32T image represents the size of the optical solar disc and has been registered by aligning the centroid of radio emission at the location of the active region 11057 with its counterpart on the 304˚A image, ignoring the time offset. All radio images are at 193.3 MHz (XX polarization) and have the same intensity scale (0–2500, arbitrary units). All images have celestial north on top. – 10 – 2006; Mercier & Chambe 2009). The higher fidelity and dynamic range enables us not only to image the quiescent solar emission in the presence of much brighter features, unlike most earlier observations (e.g., Kai 1970; Kundu et al. 1986; Krucker et al. 1995; Vilmer et al. 2002; Ramesh et al. 2010), but also to track low level variations in the coronal emission. The bottom panels in Fig. 4 illustrate the morphological relationship between our 32T radio imagesandtheextremeultraviolet(EUV)imagestakenonthesameday. Apersistent feature coinciding with the location of the NOAA Space Weather Prediction Center (SWPC) region 11057 (hereafter region 11057), is seen in the solar radio images, even at quiescent times. Rapid and significant changes in the radio morphology of the Sun on timescales of seconds are clearly visible in the images presented in Fig. 4 and are further highlighted in the two “movies” available in the on-line journal. A significant fraction of this variability maps back to the location of the region 11057. Subtle changes in T as a function of time B are also discernible at the locations of the active regions on the Eastern limb and close to the center of the disc. Fig. 5 highlights the spectroscopic capability of the 32T by showing the spectra from representative pixels across the entire emitting region. Various interesting features can be seen here. Even at quiescent times, there is a significant difference in the spectral slope of regions close to and away from region 11057. For both the “narrowband” and “broad- band” features, the changes in spectral structure are confined to the vicinity of region 11057, and to a smaller extent to the locations of other active regions visible in the EUV image (Fig. 4). The “narrowband” features are consistent with spectral features superposed on the underlying quiescent spectrum. In contrast, the underlying quiescent spectrum is no longer visible in spectra of the “broadband” feature, and the changes in spectral characteristics are significantly larger, more dynamic, and complex. 4. Discussion and Conclusions At low radio frequencies, a featureless continuum due to thermal bremsstrahlung from a corona with optical depth approaching unity constitutes thebulk ofthe “quiet” Sunemission and is expected to vary only slowly in time (e.g., Sheridan & McLean 1985). In this regime, emission that exhibits rapid spectral variations is dominated by plasma emission, a resonant emission mechanism in which electrostatic Langmuir waves are converted to electromagnetic radiation at the local plasma frequency, f , and its harmonic, 2f (e.g., Robinson & Cairns P P 2000). This is the accepted emission mechanism for type II (e.g., Nelson & Melrose 1985) and type III solar bursts (e.g., Suzuki & Dulk 1985).