Galactic Radio Explorer: an all-sky monitor for bright radio bursts
Liam Connor, Kiran A. Shila, Shrinivas R. Kulkarni, Jonas Flygare, Gregg Hallinan, Dongzi Li, Wenbin Lu, Vikram Ravi, Sander Weinreb
DDraft version January 26, 2021
Typeset using L A TEX default style in AASTeX63
Galactic Radio Explorer: an all-sky monitor for bright radio bursts
Liam Connor , Kiran A. Shila , Shrinivas R. Kulkarni , Jonas Flygare , Gregg Hallinan,
1, 3
Dongzi Li, Wenbin Lu, Vikram Ravi, and Sander Weinreb Cahill Center for Astronomy and Astrophysics, MC 249-17, California Institute of Technology, Pasadena CA 91125, USA Onsala Space Observatory, Department of Space, Earth and Environment, Chalmers University of TechnologySE-41296 Gothenburg, Sweden Owens Valley Radio Observatory, MC 249-17, California Institute of Technology, Pasadena CA 91125, USA (Received January 26, 2021; Revised January 26, 2021; Accepted)
Submitted to Publications of the Astronomical Society of the PacificABSTRACTWe present the Galactic Radio Explorer (GReX), an all-sky monitor to probe the brightest burstsin the radio sky. Building on the success of STARE2, we will search for fast radio bursts (FRBs)emitted from Galactic magnetars as well as bursts from nearby galaxies. GReX will search down to ∼ ten microseconds time resolution, allowing us to find new super giant radio pulses from Milky Waypulsars and study their broadband emission. The proposed instrument will employ ultra-wide band(0.7–2 GHz) feeds coupled to a high performance (receiver temperature <
10 K) low noise amplifier(LNA) originally developed for the DSA-110 and DSA-2000 projects. In GReX Phase I (GReX-I), unitsystems will be deployed at Owens Valley Radio Observatory (OVRO), NASA’s Goldstone station, andat Telescope Array, Delta Utah. Phase II will expand the array, placing feeds in India, Australia, andelsewhere in order to build up to continuous coverage of nearly 4 π steradians and to increase ourexposure to the Galactic plane. We model the local magnetar population to forecast for GReX, findingthe improved sensitivity and increased exposure to the Galactic plane could lead to dozens of FRB-likebursts per year. Keywords: fast radio bursts, pulsars, instrumentation INTRODUCTIONThe advent of wide-field, broad band radio surveys combined with our ability to search data at high time resolutionhas led to a number of novel discoveries. The fast radio burst (FRB) phenomenon in particular has radically changedthe radio astronomy landscape (Cordes & Chatterjee 2019; Petroff et al. 2019). In response, many FRB surveys havebeen built or proposed.The detection of a Galactic FRB from SGR 1935+2154 by both STARE2 and CHIME was the most significant step todate in connecting extragalactic FRBs to a known phenomenon (The CHIME/FRB Collaboration et al. 2020; Bocheneket al. 2020b). The value of having such objects nearby is difficult to overstate, and detecting more ultra-bright burstsis essential to understanding the connection between magnetars and FRBs.We propose the Galactic Radio Explorer (GReX, rhymes with “T-Rex”) as a complement to deeper, high-spatialresolution surveys. This is similar to how X-ray all-sky monitors and more sensitive instruments work symbiotically;GReX will detect rare ultra-bright Galactic bursts that cannot be discovered without nearly continuous all-sky mon-itoring. We are therefore building on the STARE2 design (Bochenek et al. 2020a), but with greater sensitivity, afive-times larger bandwidth, and clusters of antennas dispersed around the world. The GReX design is meant tomaximize the detection rate of ∼ MJy bursts per unit cost.
Corresponding author: Liam [email protected] a r X i v : . [ a s t r o - ph . H E ] J a n Connor et al.
Figure 1.
The layout of antenna clusters for GReX-I (blue) and GReX-II (red) is shown in the top figure. The bottom figureshows the instantaneous primary beam coverage of GReX clusters around the world, plotted over the Planck 857 GHz map (as aproxy for molecular gas and thus star-formation and Galactic magnetars). The first Galactic FRB (SGR 1935+2154) is markedin orange color. A hypothetical configuration of GReX-II, with three stations in India (marked “GMRT”), Australia, and theNetherlands each, would nicely cover the inner Galaxy (where bulk of the magnetars are located). It would also increase thecoverage of the Northern sky.
FRBs have been detected over the frequency range 0.1 GHz to 8 GHz, though not simultaneously (Gajjar et al. 2018;Pastor-Marazuela et al. 2020; Pleunis et al. 2020). The angular distribution is isotropic and the daily all-sky rateis ∼ with fluence above a few Jy ms. The typical observed pulse widths are a few milliseconds, set by the ∼ msback-ends of most blind surveys; some FRBs are known to be tens to hundreds of microseconds in duration. A subsetof roughly a dozen FRBs is known to repeat, and two of those repeaters appear to do so periodically on weeks tomonths timescales (Chime/Frb Collaboration et al. 2020; Rajwade et al. 2020). One of the main debates in the FRBfield is whether all FRBs are repeaters but with a large variation in time between bursts, or if there is a population ofgenuine once-off FRBs.The events leading up to and during April 28, 2020 resulted in a dramatic link between magnetars and FRBs. On 27April 2020, the Swift Burst Alert Telescope reported multiple bursts from the soft γ -ray repeater (SGR) 1935+2154,signaling that the magnetar had entered a phase of heightened activity. The next day the CHIME/FRB collaborationreported a dispersed burst with F of few kJy ms (0.4–0.8 GHz) in a side-lobe, but this fluence value was later revisedby a factor of 10 . The daily inspection of recorded STARE2 triggers was then expedited, and ST 200428A was foundat approximately at the same time and dispersion measure (DM) as the CHIME event. However, ST 200428A had a fluence that was a one thousand times greater than that reported by CHIME (Figure 2). Because this event was foundin the complex sidelobes of CHIME, its fluence was simpler to measure in the STARE2 data. On 30 April 2020, the ReX Figure 2.
The dynamic spectrum of FRB 200428 from STARE2 (left), along with the combined localization region from itsand CHIME’s detection (right). The STARE2 burst is centered on the X-ray emission.
Five hundred metre Aperture Spherical Telescope (FAST) reported a weak (0.06 Jy ms) radio pulse and localized toSGR 1935+2154, with a DM consistent with the CHIME and STARE2 events.Shortly thereafter, a constellation of space-borne instruments reported a one-second-long X-ray (1–250 keV) burstfrom the direction of SGR 1935+2154 that occurred at precisely the same time as the CHIME bursts and ST 200428A.This partnership between radio facilities (The CHIME/FRB Collaboration et al. 2020; Bochenek et al. 2020b; Zhanget al. 2020) is summarized graphically in Figure 2. Following the mega burst of 28 April, SGR 1935+2154 emittedintermittently bursts with fluence of 100 Jy ms (Kirsten et al. 2020) and has now finally become an intermittent pulsarat the X-ray period of 3.25 s and fluence of 40 mJy ms (Zhu et al. 2020). Clearly, magnetars are capable of emitting megaJansky burst but not all magnetar X-ray flares are accompanied by intense radio bursts. We need more observationsof this phenomenon to understand the currently murky X-ray-radio connection.In this paper we first discuss the science that can be done with a global network of antennas searching the skycontinuously at tens of microseconds. We then describe the novel hardware and digital back-end that have beendeveloped for DSA-110 and DSA-2000, which will be modified for GReX. This includes an ultra-wide band feed,extremely low-noise amplifiers (LNAs), and a sub-band single pulse search strategy. In Sect. 4, we model the Galacticpulsar and magnetar distribution in order the forecast FRB and giant radio pulse science that will emerge from GReX. SCIENCE GOALSX-ray astronomy has shallow but very wide-field “all-sky monitors” or ASMs (e.g., Rossi X-ray Timing Explorer;MAXI) and also highly sensitive instruments (e.g., Rossi Proportional Counting Array; Chandra X-ray Observatory).The ASMs have historically played a major role in identifying rare but bright events that are missed by narrower fieldinstruments. In the same way, as demonstrated by STARE2, radio astronomy would benefit from having a powerfulradio ASM. GReX will fill this role.Now let us examine the science drivers for GReX in some detail. Connor (2019) has analyzed the performance ofFRB studies and argues that there likely exist many narrow bursts missed due to instrumental smearing. Giant radiopulses (GRP) from pulsars have typical widths of microseconds but a few at the nanoseconds timescale have beendetected (Soglasnov et al. 2004). These two points are well appreciated. The computational complexity is enormousfor major FRB surveys like those of ASKAP and CHIME, which search many radio beams and seek both to localizeFRBs and detect them at a high rate. As a result, these searches limit their blind detection temporal resolutionfor detection to ∼ √ B where B is the bandwidth is no longer optimal. However, by undertaking sub-bandsearches, the ultra-wideband of GReX acts as a frequency field of view. Galactic Magnetars
The mega burst ST 200428A from SGR 1935+2154 establishes a bridge between the enigmaticextragalactic FRB phenomenon and a known physical object. It also fills in an important chasm in the luminosityfunction of coherent radio pulses, as it was considerably more energetic than any known Galactic burst, but a couple of
Connor et al. orders of magnitude less energetic than the weakest FRBs. As demonstrated by the multi-telescope, multi-wavelengthcampaign on SGR 1935+2154, the value of having such a magnetar nearby cannot be overstated. Even the highly-activerepeating FRB 180916.J0158+65, at just ∼
150 Mpc, is much too far to reasonably expect a high-energy detection. Still,it is not yet known if Galactic FRBs are physically identical to extragalactic FRBs, rather than just phenomenologicallysimilar. Therefore, it is essential that we continue to capture Galactic FRBs and dig deeper into their luminosityfunction with the sensitivity improvements of GReX.The dynamic spectra of extragalactic FRBs show several distinct features. They are often band-limited, withdownward drifting subpulses (Hessels et al. 2019), known as the “sad trombone” effect. While CHIME’s detectionof FRB 200428 had significant time and frequency structure, it is not currently known if the mega bursts emitted byGalactic magnetars have similar dynamic spectra to extragalactic FRBs. The ultra-wideband receivers of GReX allowus to “catch” the narrowband bursts that STARE2 might have missed, but also to study their dynamic spectra overa 3:1 band—five times larger than the band of STARE2—and with almost 100 times better temporal resolution.If Galactic mega bursts such as ST 200428A are found to be the same physical phenomenon as extragalactic FRBs, wewill be able to answer major open questions in the FRB field. For example, is the coherent radio emission produced inthe neutron star’s magnetosphere, or is it produced in a relativistic shock well outside of the light cylinder? What is theorigin of periodic activity in repeating FRBs? If bright bursts from Galactic magnetars are found to be meaningfullydistinct from other FRBs, then that is proof that multiple mechanisms can produce ∼ erg Hz − radio pulses; thiswould be evidence for the multiple-class interpretation of the FRB population. Super-giant pulses from Galactic Pulsars
Young pulsars like the Crab and millisecond pulsars (MSPs) likePSR 1937+214 are known to emit giant radio pulses. Let us define “super-giant pulses” (SGP) as those with fluencehigher than 1 MJy µ s. So far, only the Crab pulsar is known to emit SGPs. The Crab GRP rate is a power-law functionof F ; from Bera & Chengalur (2019), we expect N ( > F ) = 10 − hr − ( F /
10 MJy µ s) − . . Assuming the typical pulse isbroad band in frequency and that known sources such as the Crab will be coherently dedispersed, GReX could detecta 1 µ s pulse at a fluence of a few MJy µ s. From the Crab we expect a rate of, R det ( > F ) ≈ yr − ( X/ . , (1)where X is the sensitivity improvement of GReX over STARE2. Thus GReX should detect some of the brightestsuper-giant pulses from Crab-like pulsars in our Galaxy and nearby. It is also salient that some GRP-emitting pulsarsexhibit a “kink” in their energy distribution, such that the powerlaw N ( > F th ) flattens in the high-energy tail, makingultra-bright bursts more common (Mahajan et al. 2018).GRPs have been detected from roughly a dozen sources, mostly either young pulsars or MSPs. They have all beendetected after the source or its supernova remnant were discovered, sometimes in targeted searches of pulsars withlarge light-cylinder magnetic field strengths (Knight et al. 2005). Pulsars are typically not discovered blindly via theirgiant pulses. The Crab is 10 years old, and we expect O (10) such young neutron stars in the Milky Way, giventhe Galactic core-collapse rate (Rozwadowska et al. 2021) as well as recent High-Altitude Water Cherenkov (HAWC)observations (Albert et al. 2020). Their normal radio emission may be beamed away from us, but giant pulses arenot from near the polar cap and may have different beaming properties (Philippov et al. 2020). The central compactobjects of supernova remnants whose radio emission has not been detected may fall into this category (De Luca 2017).It is likely that some SGPs from unknown pulsars may be detected by the unprecedented blind search of GReX. Apulse of 30 MJy in flux corresponds to νL ν = 1 . × erg s − at a distance of 2 kpc, so the (non-)detection of suchpulses will test whether giant pulse isotropic-equivalent luminosity can (temporarily) exceed the spin-down luminosityof the pulsar. Extragalactic FRBs
GReX’s wide, shallow survey strategy also allows us to probe the nearby, ultra-bright extra-galactic FRB population. While the primary science function of the instrument is as a Galactic explorer, we mightexpect O (1) FRBs from external galaxies over the instrument’s life time. Extrapolating from the ASKAP fly’s eyesurvey (Shannon et al. 2018), N ( > F ) (cid:39) × sky − yr − ( F /
100 Jy ms) − / (see Fig. 3 of Lu & Piro 2019), to thefluence threshold of GReX, one obtains the detection rate N det ( > F th ) = 0 .
45 yr − (cid:18)
50 kJy ms F th (cid:19) / . (2) ReX
Solar Astronomy
GReX will have continuous coverage of the sun with ∼ µ s sampling and high frequency resolutionover a large bandwidth. We will therefore have access to detailed dynamic spectra of fast solar phenomena such as“millisecond spike bursts” (Wang & Xie 1999). We will be able to study Type IV radio bursts, which are likelygenerated through coherent electron cyclotron maser (ECM) emission (Wang & Xie 1999; Liu et al. 2018).We remark on two questions at the end of this section. STARE2 could see roughly 25% of the northern sky, meaningat most Earth rotational phases it would not have seen the Galactic FRB. This explains the importance of coveringthe entire sky for rare but bright events. A world-wide GReX network (as described above) can be built for under amillion dollars. Brilliant bursts such as FRB 200428 can be detected via the side-lobes of powerful facilities; indeed,that is how CHIME discovered FRB 200428. However, without explicitly adapting their pipelines to preserve eventsthat arrive far off-axis and trigger multiple beams, ultrabright Galactic bursts will likely be mistaken for RFI. In thenorthern hemisphere, CHIME and GReX-I will cover a large portion of the visible sky from 400 MHz to 2 GHz, with100 MHz of overlap at 750 MHz. This will allow for future symbioses similar to the joint discovery of FRB 200428-likeevents.Low frequency facilities such as MWA and LWA naturally enjoy large field-of-view. However, computational costsand interstellar scattering increase as one attempts blind searches at lower frequencies. In the following section wecarry out detailed modelling of the magnetars whose bursts we hope to detect. Table 1.
GReX: Top-level specifications for searchSpecification ValueBand 0.7–2 GHzTsys 25 KPolarization dual/linearfield-of-view 1.5 steradianSampling time 32 µ s (initial) / 8 µ s (hardware)Channel width 150 kHz (initial) / 38 kHz (Phase II)Fluence 100 kJy for 1-ms burstTiming link to GPS ( ±
10 ns)3.
GALACTIC RADIO EXPLORER: IMPLEMENTATIONThe basic unit of GReX has a field-of-view of ∼ ∼ µ s. RFI rejection and crude localization (via timing) requires three units separated by at least onehundred kilometers. We call such a triplet as “cluster”. To cover the entire sky would require eight cluster (four in theNorth separated by 80 degrees in longitude and six in the South). Thus, a full-up GReX network would have 24 unitsystems. Reducing the unit cost is important and is an engineering requirements in the GReX pilot phase. DuringGReX Phase I we will build a cluster with a 3:1 radio band and, thanks to novel LNAs, sensitivity at least twice thatof STARE2. Combined, the increase in sensitivity relative to STARE2 could be as much as a factor of five. We areable to achieve such improvements by piggy-backing on the advances in electronics and antenna design provided byDSA-110 and DSA-2000. Leveraging this work, we will have the most sensitive single radiometer ever developed in ourproposed frequency range. While this is our starting point, there is enough flexibility to cater the feed and back-enddesign to GReX’s science goals, for example trading field of view for forward gain. Connor et al.
We have chosen the GReX configuration over, for example, a focal-plane phased-array without a reflector for thefollowing reasons: The added complexity and cost related to developing such a system would be non-trivial, as wouldthe increased cost of digitization, channelization, and beamforming compute hardware that follow from having morefeeds. The current feed design provides a frequency-independent beamwidth, and a single Stokes I beam requiresonly two digital backends per antenna, one for each polarization. Therefore, if our goal is to monitor the whole skycontinuously with high time and frequency resolution, we find that the most effective way is to deploy simple singleradiometer systems around the world. Another suggestion for an all-sky Galactic FRB survey relied on Citizens-Scienceand cellular communication devices to search for ∼
10 GJy bursts (Maoz & Loeb 2017). However, the detection ofST 200428A just above the STARE2 threshold established that the Galactic FRB brightness distribution cannot bevery flat and Giga bursts are likely exceedingly rare.With a world-wide GReX system that includes both more sky coverage and more exposure to the Galactic plane,we anticipate more than an order of magnitude increase in detection rate over STARE2. To achieve this, we aim tocreate an assembly kit that can be shipped at cost to interested parties around the world.3.0.1.
Wideband Antenna
Figure 3. ( A ) The current STARE2 low-cost “cake-pan” antenna which delivers uniform beam-width and excellent efficiencybut over a limited bandwidth. ( B ) Quad-ridge structure added to the cake-pan delivers wider bandwidth while maintaining alow-cost design. ( C ) The assembled quad-ridge choke horn structure concept to be used for GReX. The current “cake-pan” antenna of STARE2 is fabricated from a 6 (cid:48)(cid:48) -diameter aluminum pipe surrounded by two cakepans which reduce spillover and increase efficiency, providing a low-cost solution that delivers uniform beam-widthacross the 256 MHz band.Here, we propose to use a quad-ridge horn with a choke-ring structure (Figure 3) for wide-beam performance overa wide frequency range. The design is based on the quad-ridge flared horn (QRFH) technology, developed at Caltechby graduate student Ahmed Akgiray and his advisor Dr. Sandy Weinreb. To take advantage of the cost-effectivedesign of the cake-pan antenna, the quad-ridge structure will be integrated with the choke-rings (Figure 3). Thechoke-ring structure reduces side and back-lobes resulting in a near-symmetric beam pattern. In Figure 4, the beam isexemplified at 1.4 GHz, and the wide, near-constant, full-width-at-half maximum (FWHM) presented over frequency.The quad-ridge structure enable dual linear polarization within a compact footprint, and good impedance match tolow-cost 50 Ω coaxial connectors resulting in low input-reflection coefficient over the wide frequency band. To reducethe contribution of noise from ground pick-up to only a few K (Figure 7), a shield underneath the antenna will beused. 3.0.2. Low Noise Amplifiers (LNAs)
A break-through in low noise amplifier (LNA) technology has occurred in the past few years and enables a sensitivityimprovement of a factor of 3 compared to the monitors used for the previous STARE2 detection. This can be achieved https://old.astron.nl/r-d-laboratory/ska/embrace/embrace Akgriay’s 2013 Caltech PhD is a convenient reference: https://thesis.library.caltech.edu/7644/
ReX Figure 4. (left): Polar representation of the beam at 1.4 GHz. right): Simulation of beam full-width-at-half maximum (FWHM)in degrees and effective area, A e , as a function of frequency, ν (in GHz). by reducing the noise temperature of the LNA from 32 K to 10 K and contributions of feed spillover and other lossesfrom 28 K to 10 K to improve the system noise temperature from 60K to 20K. This can be accomplished without theutilization of cryogenic coolers which are costly, require much AC power, and require considerable maintenance. Figure 5.
The DSA-110 LNA showing exterior view (top left), interior view (top right), as mounted on the DSA-110 feed(bottom left), and noise temperature vs frequency at 3 temperatures (bottom right). The LNA requires no wires for bias orcontrol; the bias of +5V is supplied on the output coaxial line with control of an internal noise calibration source by a 32kHztone on this output line.
This LNA break-through has been demonstrated in the DSA110 array at the Caltech Owens Valley Radio Observatory(OVRO) where a system noise of 25 K has been measured on 25 dual-linear polarization 4.6 m paraboloidal reflectorantennas operating in the 1.28 to 1.53 GHz frequency range. The LNA for this project is summarized in Figure 5and fully described in a paper (Weinreb & Shi 2021). The key elements for this low noise are an extremely highperformance (Fmax of 550 GHz) high-electron-mobility-transistor (HEMT) on an indium-phosphide (InP) substrate
Connor et al.
Type pH-100 discrete InP HEMT and an extremely low loss ( < .Our goal for GReX is an LNA for 0.7 to 2 GHz with frequency-averaged noise temperature under 10K. This ischallenging due to the required bandwidth of the input matching network to transform 50 ohm impedance of thefeed to the optimum impedance driving the transistor which is known from previous studies. This network musthave extremely low loss with the realization that 0.1 dB of loss adds 7 K to the noise temperature. There are twoapproaches to this challenge illustrated in Figure 6: 1) a more complex, but very low loss, input matching network,and 2) cooling just the transistor chip to -40 ° C using a Peltier effect solid-state micro-cooler. Using computer-aided-design optimization with HFSS electromagnetic modeling software, a shunt stub Tee input network has been optimizedto give a frequency-averaged noise temperature of 10.1K with operation at 25 ° C or approximately 5 K less noise bycooling to -40 ° C. Analytic models of the loss and capacitance of the Tee connection were not available so a fieldanalysis utilizing finite elements to solve the Maxwell equations was required. A major challenge with cooling to -40 ° Cis the condensation of water and ice on the transistor. To prevent this over a long period of time, both high vacuumor pressurization with a low thermal conductivity gas such argon or xenon are being investigated. There is risk tothe cooling and our next step will be to construct a prototype LNA with the wideband input network, measure theLNA noise in a laboratory test setup, and then, measure the system noise with the feed including the ground radiationshield.
Figure 6. (left) Modeled LNA noise temperature in red without a microcooler, averaging 10.1 K, and with the shunt stubinput matching network shown below. (right) Measured physical temperature of a transistor chip mounted on the 6mm highmicrocooler shown below.
Following the LNAs additional amplification, filtering, and perhaps frequency conversion will be required to drivethe A/D converters and the digital spectrometer. These can be more conventional RF design modules similar to thoseused in DSA-110 but with wider bandwidth. 3.0.3.
Digital back-end
We propose to use SNAP-2 platform which allows direct sampling (5 Giga-samples per second, or 5 Gsps) of theentire 0.2–2 GHz RF signal to process the 0.7–2 GHz bandwidth. The ADCs sample at 10-bit precision which we believe Diramics AG, Zurich, https://diramics.com/products/ https://events.mpifr-bonn.mpg.de/indico/event/154/session/4/contribution/27 This includes a Kintex Ultrascale FPGA fed by dual FMC-mounted 5GAD ADCs. Both are supplied by the Institute for Automation inBeijing, China
ReX Figure 7.
Simulation of system temperature, separated as contributions from ground pick-up (T gp ), sky (T sky ; i.e. CMB,Galaxy, atmosphere), the low-noise amplifiers (T LNA ), and losses in the wideband antenna (T ant ). A ground radiation shield inthe form of a square with 4 m sides and 0.3 m, slightly angled, walls reduce the ground pick-up to a few K. A study for optimalshield-material and shape will be done to weight costs (mesh or solid material; curved or flat) and benefit (T gp ). is sufficient to excise strong RFI that is present in the band of interest to us. We will implement a standard four-tap polyphase filter-bank, beginning with a 16,384 channels across 2.5 GHz (153 kHz channel width) but preservingthe possibility of 64 k channelization. Following detection in each polarization we will integrate to the desired timeresolution, optimally re-quantized to 8-bit precision, and streamed across a 40 GbE direct connection to a computerserver (data rate of 8 Gbps). The use of a 40 GbE connection permits future upgrades towards streaming voltage datato the server for buffering for the entire usable bandwidth.3.0.4. Algorithm & Analysis Computer
GReX’s real-time detection pipeline and computing architecture will be tailored to the instrument’s broad radio bandand high time resolution; the traditional
Heimdall package, which was used for STARE2, is not well suited to the wideband of GReX, the top of whose frequency range is three times the lower frequency. We will employ a “sub-band”dedispersion algorithm, in which GReX’s large frequency range will be apportioned into uniform chunks in ν − thatwill be searched independently and later combined. To this end, we will customize the fast dedispersion algorithm FDMT (Zackay & Ofek 2017), which allows for sub-band searching and optimally detecting FRBs with significant frequencystructure using the Kalman filter . GReX will thereby serve as a proof-of-concept for future ultra-wideband surveyssuch as DSA-2000 and the funded Canadian project CHORD (Vanderlinde et al. 2019).The computer server will be equipped with sufficient processing power to reject impulsive RFI in real time, and tosearch for FRBs. Data will be piped through the analysis software using the psrdada framework. The RFI rejection,dedispersion, and pulse finding will be implemented on an Nvidia Quadro RTX4000 GPU. The hardware is well suitedto a final-stage machine learning classifier which will send out reliable, real-time triggers to other facilities Connor &van Leeuwen (2018).We plan to reserve roughly a dozen DM channels to be coherently dedispersed when known pulsars are in the beam,enabling us to search for super-giant radio pulses without the deleterious effects of instrumental smearing. The DMswill correspond to known millisecond pulsars and young pulsars, the sources most likely to emit giant pulses.Our design allows for upgrades to be explored by the group, in response to the evolving scientific landscape. Forexample, we can modify the digital firmware to stream voltage data for buffering on the server, such that when apulse is detected the data can be analyzed with better time and frequency resolution. Additionally, we have sufficientprocessing power on the FPGA to derive full-polarization data, if needed. https://bitbucket.org/bzackay/kalman detector/src/master/ Connor et al.
Future Stations
One of the goals of GReX is to expand University-based radio observatories in the United States. Specifically, weenvisage a future effort to place unit systems at sites that are managed or accessible to Universities, with emphasis onstate universities in the US. RFI studies need to be undertaken of potential future site. For clusters within ∼ thousandsof km of one another, we will offset the antennas in pointing to minimize overlap on sky.During GReX Phase II we plan expand beyond the United States, starting with Australia and India in order to buildup 24/7 coverage of the Galactic plane, where the majority of magnetars reside. Clusters of antennas in both WesternAustralia and Tasmania would provide 120 deg of coverage in right ascension, if they each point 7.5 deg off-zenith. Aset of three antennas in India could be located at GMRT. We also hope to deploy GReX instrumentation in WesternEurope, the Middle East, and elsewhere until we have an international network spanning the full sky’s 4 π steradians.We anticipate that this will require roughly 25 hardware kits. Hour angle (deg) M a g n e t a r f r a c t i o n UTC (hour) M a g n e t a r f r a c t i o n MWO, AustraliaGMRT, IndiaOVRO, USNetherlandsTotal exposure
Figure 8.
Modelling of the fraction of Galactic magnetars that are within the FoV of a GReX antenna for different potentialcluster locations around the world. The top panel shows this fraction as a function of hour angle (HA). The bottom panel showsthe same fraction vs. UTC, accounting for the different RAs to which each location is exposed at a given time. The dotted lineis the combined visible fraction of Galactic magnetars for the whole GReX-II array.4.
MODELLINGBy sending GReX clusters to southern latitudes we will gain exposure to the bulk of the Galactic plane, wheremagnetars reside. We also aim to increase our coverage in right ascension by deploying clusters at a range of longitudes.
ReX
Event rates
The system temperature of GReX is expected to be ∼ Event Rate (yr ) R e l a t i v e p r o b a b ili t y STARE2MWO, AustraliaGMRT, IndiaOVRO, USNetherlandsTotal exposure
Figure 9.
The event rate distributions based on one FRB-like event detected by STARE2 in 448 days on sky. The relativeprobabilities vs. detection rate are plotted for individual GReX clusters around the world (solid) as well as a hypotheticalGReX-II array (dotted) that has antennas in Australia, the Netherlands, India, and the United States. Even if FRB 200428were a rare event and STARE2 and CHIME/FRB were “lucky” to have detected it, the full array will likely find multiple suchevents per year. density on sky above the instrument’s detection threshold, N ( > s min ). We take the source brightness distribution tobe a power-law such that, N ( > s min ) ∝ s − α min , where s min is given by the radiometer equation. Assuming a baseline ratefor STARE2, R S2 , the detection rate of GReX-II will be a sum over all antenna clusters around the world, weightedby their average exposure to the Galaxy’s magnetars compared to STARE2. We call this weight for the i th GReXcluster location, w i . This gives,2 Connor et al. R GReX = R S2 n clust (cid:88) i w i Ω i Ω S2 (cid:32) SEFD S SEFD i (cid:114) B i B S (cid:33) α , (3)where SEFD refers to the system-equivalent flux density, or the ratio of system temperature to gain, T sys /G . Assumingeach GReX unit has the same FoV as STARE2 and that the pointings are mostly independent, we get R GReX = 1 /
448 days − (cid:16) . √ (cid:17) α (cid:88) i w i . (4)Equation 3 computes the maximum-likelihood value of the event rate based on STARE2’s detection ofST 200428A/FRB 200428, but we must include the uncertainty associated with just one burst. Using Bayes’ theo-rem we know, P ( R| N ) = P ( N |R ) P ( R ) P ( N ) , (5)where in this case N = 1. Taking a flat prior on R and assuming the detection of new bursts follow Poissonianstatistics, we can invert Equation 5 to get, P ( N = 1 | R ) = R e −R . (6)By scaling R for GReX using Equation 3, we can calculate the probability distribution in detection rate after theimproved sensitivity and exposure to Galactic magnetars. This is shown in Figure 9. SUMMARYWe have proposed GReX, a radio all-sky monitor that will detect the brightest bursts in the Galactic sky on sub-millisecond timescales. GReX Phase II will be an international network of ultra-wideband, high performance antennasthat will continuously search for Galactic FRBs and super giant pulses from radio pulsars. Each hardware componentwill be low-cost and replicable, such that clusters of at least three GReX antennas can be shipped as an assembly kitaround the world to span a wide range of latitude and longitude. The software and firmware back-end will also bestandardized such that each system will require little intervention once running. We expect to find multiple new FRB-like events from Galactic magnetars each year. As the first wide-field, blind single-pulse survey at the microsecondlevel, we should to find new super giant pulses from Galactic pulsars, as well as previously-unknown phenomena.ACKNOWLEDGEMENTSWe thank Dale Gary and Dan Werthimer for helpful discussions.REFERENCES
Albert, A., Alfaro, R., Alvarez, C., et al. 2020, ApJ, 905,76, doi: 10.3847/1538-4357/abc2d8Bera, A., & Chengalur, J. N. 2019, MNRAS, 490, L12,doi: 10.1093/mnrasl/slz140Bochenek, C. D., McKenna, D. L., Belov, K. V., et al.2020a, PASP, 132, 034202,doi: 10.1088/1538-3873/ab63b3Bochenek, C. D., Ravi, V., Belov, K. V., et al. 2020b,Nature, 587, 59, doi: 10.1038/s41586-020-2872-xChime/Frb Collaboration, Amiri, M., Andersen, B. C.,et al. 2020, Nature, 582, 351,doi: 10.1038/s41586-020-2398-2Connor, L. 2019, MNRAS, 487, 5753,doi: 10.1093/mnras/stz1666 Connor, L., & van Leeuwen, J. 2018, AJ, 156, 256,doi: 10.3847/1538-3881/aae649Cordes, J. M., & Chatterjee, S. 2019, ARA&A, 57, 417,doi: 10.1146/annurev-astro-091918-104501De Luca, A. 2017, in Journal of Physics Conference Series,Vol. 932, Journal of Physics Conference Series, 012006,doi: 10.1088/1742-6596/932/1/012006Gajjar, V., Siemion, A. P. V., Price, D. C., et al. 2018,ApJ, 863, 2, doi: 10.3847/1538-4357/aad005Hessels, J. W. T., Spitler, L. G., Seymour, A. D., et al.2019, ApJL, 876, L23, doi: 10.3847/2041-8213/ab13aeKirsten, F., Snelders, M., Jenkins, M., et al. 2020, arXive-prints, arXiv:2007.05101.https://arxiv.org/abs/2007.05101
ReX13