The Breakthrough Listen Search For Intelligent Life Near the Galactic Center. I.

There are two different ways of detecting evidence of ETI through radio technosignatures: detecting deliberate beacons, or eavesdropping on the leakage radiation that could be a byproduct of extraterrestrial technologies. However, it is much harder to speculate on the type of leakage radiation, and such putative leakage signals are also likely to be weaker. In this paper, we will focus on strong beacons deliberately transmitted by ETI. In the past, very few hours of observation on the GC have actually been performed to search for these ETI beacons. A narrowband Doppler-drifting beacon signal (Backus & Project Phoenix Team 2004; Harp et al. 2012) is a prime candidate for deliberately transmitted technosignatures. Shostak & Tarter (1985) carried out one of the few dedicated surveys of the GC for such narrowband signals in a 4 hr search with the Westerbork Synthesis Radio Telescope (WSRT) at a sensitivity threshold of ∼10¹⁸ W, reporting a nondetection.

Cole & Ekers (1979) suggest that, in addition to searching for signals concentrated in frequency space, searches for impulsive wideband signals should also be conducted. Powerful wideband radar systems are commonly used on Earth; they follow a dispersion relation similar to a pulse dispersed by the interstellar medium (ISM). An advanced society may know about pulsars and the techniques required to detect them. They could generate a pulsar-like broadband radar signal readily detected by another technologically capable society. Further, they could induce an artificial dispersion measure into the pulse to differentiate their transmitted signal from a pulsar. Such a signal serves as a second type of beacon readily identifiable as originating from nonastrophysical processes. Cole & Ekers (1979) further speculated that such broadband beacons will consist of extremely narrow (≤4 μs) in time pulses (nonphysical) and conducted searches for naturally dispersed pulses toward many targets including the GC. However, the Crab pulsar is known to produce μs structures at higher frequencies (Hankins & Eilek 2007). Thus, negative or artificial dispersion might be the only indicator of artificiality for such broadband signals.

No other SETI survey with moderate sensitivity has been performed for the GC region. Given the potential for discovery in the GC region, our outlined survey will contiguously span the large fraction of visible radio window from ground-based facilities, i.e., from 1 to 93 GHz. This constitutes the largest fractional bandwidth search for radio signals toward any source to date for SETI. In Section 2, we highlight that the GC is one of the most interesting locations to conduct a survey for technosignatures, with Section 2.4 outlining the ancillary science benefits. In Section 3, we outline our survey strategy for the Parkes telescope (Section 3.1) and GBT (Section 3.2), which will conduct observations across 0.7–3.8 GHz utilizing 352 hr and 3.9–93.0 GHz utilizing 280 hr, respectively. In Section 4, we outline observations that have already been carried out across 1–8 GHz from both these facilities. In Section 5, we detail our search for two beacon types. Section 5.1 highlights our search for the abovementioned narrowband Doppler-drifting signals, while Section 5.2 outlines our searches for four different types of broadband dispersed signals (originating from natural and nonastrophysical processes). Section 6 discusses the implications of our findings, and Section 7 lists our final conclusions.

The GC has the largest concentration and highest number density of stars within the Galaxy. The line of sight toward the GC offers the largest integrated Galactic star count of any direction in the sky. However, the emergence of intelligence near the GC depends on the probability of habitable planets along with their survival and stability due to nearby frequent supernovae (SNae) and other cataclysmic events such as stellar flybys and flares from magnetars near the GC. Lineweaver et al. (2004) also indicated that the inner region might be a less hospitable environment for habitability. In Jiménez-Torres et al. (2013), it is found that, within R = 0.8 kpc of the GC, stellar flybys are likely to disrupt Oort clouds, which would consequently result in the damage of planetary disks and systems. However, such flybys might deplete the Oort clouds on a fairly short timescale, thus making the planetary systems more habitable at longer times. Gowanlock et al. (2011) modeled the habitability of the Galaxy to host complex land-based life. Morrison & Gowanlock (2015) assessed the propensity of the Milky Way to host intelligent life by including an additional timescale to the prior work of Gowanlock et al. (2011) that accounts for the transition from complex to intelligent life, which required 0.6 Gyr on Earth. Both these models suggest the inner Galaxy as the most likely place for the emergence of habitable worlds and intelligent life. However, Gowanlock et al. (2011) elected to model the Galactic disk only at R ≥ 2.5 kpc, as the complicated formation history of the galactic bulge at R < 2.5 makes it difficult to model habitability in the region. To understand the propensity of intelligent life in the very inner region of the Milky Way for our BL survey, we update their model to account for the entire disk and include the R < 2.5 kpc region. As is clearly evident from Figure 1, the fraction of stars with a habitable planet is greatest in the inner disk of the Galaxy. This is because there is much earlier planet formation in the region than at the Solar neighborhood or the outskirts, due to the inside-out formation history of the Galaxy. Despite the high SNae rate in the inner Galaxy, there are more opportunities for land-based, subsurface, or oceanic complex life to emerge on planets in the region. The propensity for the emergence of intelligent life as introduced by Morrison & Gowanlock (2015) closely follows the density of habitable worlds and thus also suggests the inner region of the Galaxy as the most promising region. In summary, with findings from Gowanlock et al. (2011) and Morrison & Gowanlock (2015), as well as our extension to the inner disk (R < 2.5 kpc), if the emergence of life is commonplace within the galaxy, it follows that the inner Galaxy, especially regions around the GC, would provide an ideal location to search for technosignatures in our Galaxy.

Zoom In Zoom Out Reset image size

The higher density of stars and habitable planets at the GC will also be beneficial for the growth of more advanced societies. Close proximity among the likely inhabited worlds may accelerate development of interstellar communication and travel, which can give rise to advanced spacefaring societies. Newman & Sagan (1981) first discussed in detail diffusion of such advanced societies across the Milky Way. More recently, Carroll-Nellenback et al. (2019) modeled the expansion of spacefaring societies and suggested that the high density of stars provides settlement fronts to move much more quickly. Thus, if civilizations engage in settlements, planets in the inner region of the Galaxy are likely to be settled by advanced societies much earlier than other parts of the Milky Way. Similar arguments are used by Di Stefano & Ray (2016) to suggest that globular clusters are likely places to harbor spacefaring societies, due to their close proximity to each other. Such advanced societies are likely to produce technosignatures that can be detected across large interstellar distances.

In game theory, Schelling (1960) suggested an approach by which different members of a group can derive a mutually beneficial solution to a problem in the absence of direct communication. These arguments can be extended to propose a range of parameter spaces that ought to provide a maximal outcome in searches for technosignatures (Wright 2018). For example, suggestions of searching near the hydrogen-line frequency by Cocconi & Morrison (1959) and within the "cosmic water-hole" frequencies by Oliver & Billingham (1971) provide likely "Schelling points" for suitable frequencies to search for ETI beacons. In order to maximize the chances of detecting beacons across the entire sky, the optimal Schelling point for such a transmitter to exist would be the GC. The GC is a natural cynosure of the entire Milky Way, and also suggested by Project Cyclops as the only likely direction for sensitive technosignature searches (Oliver & Billingham 1971). ¹³

Benford et al. (2010a) argued various motivations for an advanced society to transmit powerful beacons and highlights the importance of an energy budget. They argued that transmitting strong pulses as beacons with a certain duty cycle would be the most cost-effective way to build a beacon. Benford et al. (2010a) suggested that advanced societies residing in our line of sight toward the GC might choose to transmit such powerful beacons toward the GC in order to get the maximum number of targets in the beam. Such transmitters might also direct back outward away from the GC because it is likely that societies like ours will be looking at the GC, due to the abovementioned Schelling point argument. They hence suggested a survey strategy to look at the GC on a roughly daily cadence over an annual period. Benford et al. (2010b) calculated the energy budget of such a strong beacon. As the required power for such a transmitter is proportional to the square of the distance to the target (i.e., P_transmitter ∝ D²), placing a transmitter at a location that can illuminate a large fraction of targets in the Milky Way with the least amount of energy would also favor the GC. Recently, Comisso & Asenjo (2021) suggested that it is possible to extract energy from the spinning black hole at the GC. An advanced ETI transmitter might be able to obtain enough energy this way to remain active for billions of years. Thus, we propose that by placing a powerful transmitter either at the GC or pointing toward and away from the GC are the most cost-effective ways, making it an ideal Schelling point for technosignature searches within the Milky Way.

The GC region is also an exciting observational target for a host of natural astrophysical phenomena, prominently including pulsars in close orbits around the central supermassive black hole, Sgr A* and other magnetars, or in new exotic systems such as a millisecond pulsar in a binary system with a black hole.

2.4.1. Pulsars

2.4.2. SGR J1745−2900

2.4.3. Fast Transients

We will use 352 on-source hours to observe the GC (0, 0) and bulge using Parkes. Parkes will perform the lower-frequency portion of the survey with the use of the newly installed UWL. This receiver covers 704 MHz to 4.032 GHz, with an average system temperature, T_sys = 22 K˙ (Hobbs et al. 2020). Because we will only be using one receiver and the HPBW of Parkes is larger than that of the Green Bank Telescope (GBT), the observing strategy will be to observe with approximately half the time spent on the GC (168 h) and the remaining time surveying the bulge.

At 700 MHz, the HPBW is approximately 30'. This defines the Galactic Center for the Parkes observations. At a distance of 8.2 kpc (distance to the GC; Gravity Collaboration et al. 2019), this corresponds to a circular region with a radius of ∼35 pc. We plan to sample this region by using seven beams at a distance of 92 from each other; this corresponds to the HPBW at the midpoint of the frequency band for the UWL (2.35 GHz). This is shown in Figure 2. We allocate approximately 24 h per pointing. Each pointing will be observed multiple times in 2–3 hr increments, at a cadence of every 1–2 weeks, to identify transient and highly scintillating signals. Our search parameters include a range of ±4 Hz s⁻¹, which is large enough to account for changes in the Doppler shifts due to Earth's spin, taking orbital velocity effects to be negligible, toward the GC on different days of observations.

Zoom In Zoom Out Reset image size

The Galactic bulge will be surveyed for 184 h. We define a region of 2° in Galactic latitude and 4° in Galactic longitude to cover the extent of the bulge. This corresponds to 375 pointings, including the abovementioned GC pointings. This will result in 30 minutes per pointing, with three ten-minute-long scans per pointing.

Table 1. Parkes Low-frequency and GBT High-frequency Receiver Bands Along with Other Necessary Details

Receiver	Obs. Freq.	Nbeams	Nbands	Δν/band	Total Δν	Beam Size	SEFD
	(GHz)			(GHz)	(GHz)	FWHM	(Jy)
UWL	0.704–4.032	1	1	4	4	92	40
C	3.9–8	1	1	4	4	25	10
X	8–11.6	1	1	2.4	2.4	14	15
Ku	12–15.4	2	1	3.5	3.5	54''	18
KFPA	18–27.5	2	2	8	9	32''	20
Ka	26–39.5	2	3	4	12	226	25
Q	38.2–49.8	2	2	4	8	16''	60
W	67–93	2	4	6	24	10''	100

Note. Columns list range of covered frequencies, number of beams, number of sub-bands, instantaneous bandwidth per sub-band, total bandwidth, full width at half maximum (FWHM) beam size computed for the central frequency, and system equivalent flux density (SEFD), respectively. KFPA is a seven-beam receiver; however, we are only planning to use two of these beams, with 4 GHz of band per beam, for our survey.

Download table as:  ASCIITypeset image

The GBT will be used for the high-frequency component of the survey starting from 4 GHz (C band) to 93 GHz (W band) (see Table 1) with a total observing time of around 280 h (excluding overhead). As mentioned earlier, a significant amount of observing time per receiver band will be dedicated for deep observations of the GC (0, 0). The remaining time will be used to observe the region around the GC. Many of the details in this section are taken from the GBT Proposer's Guide ¹⁵ and Observing Guide. ¹⁶

We would like to uniformly sample the GC region with the complete observable GBT band. However, the GBT beam size varies by a factor of 30 and the receiver SEFDs cover a factor of 10 in sensitivity from C band to W band. Also, the instantaneous bandwidth varies between receiver. As such, a compromise in survey area, bandwidth, and sensitivity must be made. The observation pointings and integration times are summarized in Figure 4 and Table 2. A deep integration, similar to the Parkes observing plan will be performed with each receiver of the GC (0, 0). The GC bulge around GC (0, 0), depending on the beam size and number of sub-bands to cover the full receiver bandwidth, will also be sampled at 15 minute integration times (three five-minute scans). Further details of the receiver specific observing strategies are presented below.

Table 2. Observing Time and Number of Pointings for Parkes UWL and Each GBT Receiver Planned to Be Used in This GC Survey

Receiver	Sampled	Obs Δν	Deep	Obs Time	Bulge	Obs Time	Total Obs
	Region	(GHz)	pointings,	per deep	pointings,	all bulge	(Hours)
			NInt,deep	pointing,	NInt,bulge	pointings,
				τobs,deep		τobs,bulge
				(hours)		(hours)
UWL	2° × 4°	4	7	24	368	184	352
C	$4^{\prime} \times 4^{\prime}$	4	1	10	18	4.5	14.5
X	$4^{\prime} \times 4^{\prime}$	2.4	1	10	36	9	19
Ku	$4^{\prime} \times 4^{\prime}$	3.5	1	10	126	31.5	41.5
KFPA	$2^{\prime} \times 2^{\prime}$	8	2	10	90	22.5	42.5
Ka	$2^{\prime} \times 2^{\prime}$	4	3	3	180	45	54
Q	$2^{\prime} \times 2^{\prime}$	4	2	3	180	45	51
W	$1^{\prime} \times 1^{\prime}$	6	4	2	144	36	44

Notes. The columns list the sampled region around the GC, available instantaneous bandwidth, number of deep pointings necessary to cover the full bandwidth at a given band and to fully sample GC (0, 0), total observation time per each deep pointing, total number of bulge pointings, total observing hours for the bulge, and total observing hours for the full survey at a given band, respectively. The total observing hours for each receiver are calculated as N_Int,deep×τ_obs,deep + τ_obs,bulge.

Download table as:  ASCIITypeset image

C Band (3.9–8 GHz) The C-band feed is a dual-polarization, single-beam receiver with 3.9 GHz of instantaneous bandwidth. This feed has been used in previous BL observations of the standard target list (Isaacson et al. 2017), stars located in the restricted Earth transit zone (Sheikh et al. 2020), and FRB 121102 (Gajjar et al. 2018). A 4' region around the GC can be fully sampled with 19 pointings, including a deep pointing of the GC (Figure 3). The BL digital backend (BLDB; MacMahon et al. 2018) will record baseband voltages across the entire 4 GHz of the receiver bandwidth. Combining the deep pointing of GC (0, 0) for 10 hr with the standard ON–OFF pointings of the 4' bulge, we spent around 14 hr surveying the GC region. Detailed analysis of these observations is presented in Sections 4.1 and 5.

Zoom In Zoom Out Reset image size

X Band (8.0–11.6 GHz) The X-band feed is a circularly polarized, single-beam receiver with 2.4 GHz of instantaneous bandwidth even though the total receiver bandwidth is 3.6 GHz. We aim to sample a 4' region around the GC with 37 pointings, including a deep 10 h pointing of the GC (0, 0) (Figure 4(a)) with a total observing time of around 19 h.

Zoom In Zoom Out Reset image size

Ku Band (12.0–15.4 GHz) The Ku band has two beams at a fixed separation ($5\buildrel{\,\prime}\over{.} 5$ in the cross-elevation direction), and each is circularly polarized. The instantaneous bandwidth is 3.5 GHz. Since the offset beam is likely to be outside the selected GC region for most of the GC pointings, it will be used to increase the sky coverage and to provide OFF sources for GC-region pointings. The BLDB is capable of recording 8 GHz of instantaneous bandwidth. This will allow us to capture full receiver coverage simultaneously from both beams, which will eliminate the need for ON-OFF cadence scans. A 4' region around the GC will be fully sampled with 127 pointings, including a deep pointing of the GC (0, 0) (Figure 4(b)) with a total observing time of 41.5 h.

K-Band Focal Plane Array (18.0–27.5 GHz) The K-band receiver is a focal plane array (KFPA) of seven dual-polarization (circular) beams at fixed separations on the sky in a hexagonal pattern. The 2' region around the GC can be fully sampled with 91 pointings as shown in Figure 4(c) considering an HPBW of 32''. There is a full seven-beam mode at 1.8 GHz instantaneous bandwidth per beam, and a dual-beam 8 GHz total instantaneous bandwidth experimental mode. Since the BL backend can capture 12 GHz of instantaneous bandwidth, we will use the single central beam with 8 GHz simultaneous bandwidth. For KFPA, we plan to use an ON–OFF observing strategy similar to that employed for the C band and X band. For the deep integration of the GC (0, 0) region, we plan to use the dual-beam mode of the KFPA with 4 GHz per beam. This doubles the number of independent pointings, to cover the entire bandwidth for the deep GC (0, 0) pointing. Such a dual-beam approach is especially necessary for transient signals, as dispersion delays at these frequencies are negligible, which renders them indistinguishable from radio-frequency interference (RFI). The 90 pointings of 15 min each from the GC bulge and two pointings (4 GHz per pointing) of 10 h each for GC (0, 0) gives a total observing time of 42.5 h.

Ka Band (26.0–39.5 GHz) The Ka-band receiver consists of two beams, each with a single polarization (linear) at a fixed separation (78'' in the cross-elevation direction). The beams are orthogonal in polarization. The band is split into three bands (26.0–31.0 GHz, 30.5–37.0 GHz, 36.0–39.5 GHz), only one of which can be used at a time, with an instantaneous bandwidth of 4 GHz. Thus, we will use both beams with 4 GHz per beam to survey the GC bulge and the GC (0, 0). Here, the second beam will provide OFF-source pointing for deep and standard pointings. Since three bands are required to sample the full receiver bandwidth, 60 pointings will be used per band for the bulge and a total of three deep observations of the GC (0, 0) (see Figure 4(d)). The 180 pointings of 15 min each from the GC bulge and three pointings (4 GHz per pointing) of 3 h each for GC (0, 0) give a total observing time of 54 h.

Q Band (38.2–49.8 GHz) Like the Ka band, the Q band also has a dual-beam receiver, each dual-polarization (circular) at a fixed separation (578 in the cross-elevation direction). The instantaneous bandwidth is 4 GHz. Thus, with the BLDB, we will record data from both beams. The offset beam will be used to increase the sky coverage and provide OFF-source pointings for the GC bulge and the GC (0, 0). The 2' region of the GC bulge is divided into 90 pointings (see Figure 4(e)). We plan to spend 15 min for bulge pointings and 3 h for the deep observations of the GC (0, 0) region. Since two bands are required to sample the full receiver bandwidth, 90 × 2 bulge pointings and 3 × 2 deep observations provide us a total of 51 h of observation.

W band (67–93 GHz) The W-band receiver also has two beams, each dual-polarization (linear) at a fixed separation (286'' in the cross-elevation direction). Similar to Ka and Q band, we will use the second beam to increase the sky coverage and obtain OFF-source pointings. We will sample a 1' region of the GC bulge with 37 pointings (see Figure 4(f)). We plan to spend the standard 15 min for bulge pointings and 2 h for the deep observations of the GC (0, 0) region. The band is split into four separate bands (67–74 GHz, 73–80 GHz, 79–86 GHz, and 85–93.3 GHz), only one of which can be used at a time. Thus, 36 × 4 pointings for the bulge along with 2 × 4 pointings for the GC (0, 0) will be completed in 44 h.

Here, we report observations on the GC carried out in five sessions between 2019 August 07 and 2019 September 11 using the C-band receiver at the GBT. As indicated in Section 3, these observations were conducted toward 19 pointings (see Figure 3) for three five-minute scans each. We also conducted deep observations of around 6.5 h toward the GC (0, 0), marked as pointing A00. We labeled the surrounding hexagon pointings as B and C (D, E, F, and G for higher bands), numbering the beams counterclockwise starting immediately to the right of A00. In order to eliminate false positives for the narrowband searches due to terrestrial RFI, we carried out observations in a sequence that allowed two HPBW separations between consecutive pointing centers. The shorter scans toward A00 were followed by two shorter scans of outer-ring pointings, i.e., C01 and C07. The other pointings were observed in the following pairs: B01–B04, B02–B05, B03–B06, C02–C04, C03–C05, C06–C08, C09–C11, and C10–C12 (see Figure 3). We also recorded noise-diode scans on the flux calibrators and a strong pulsar to measure our sensitivity and configuration during most sessions (see Table 3).

The observations, also listed in Table 3, were conducted using the BLDB. The BLDB provides a unique flexibility to capture and record baseband raw voltages across 12 GHz of bandwidth with 64 compute nodes at 32 bit resolution, making it one of the most powerful backends at the GBT. These compute nodes are clustered in eight banks, with each bank hosting eight compute nodes. We also have 10 storage nodes with a total recording capacity of 8.2 PB for long-term storage. For our observations, the BLDB repurposed the existing VEGAS backend (Versatile Greenbank Astronomical Spectrometer; Prestage et al. 2015) to tap into the GBT analog downconversion streams with four 1500 MHz wide tunable passbands covering the entire C-band receiver frequency response, spanning 3.9–8.0 GHz. This interface provided coarsely channelized raw digitized voltages into 512 polyphase channels—with 2.97 MHz of channel resolution. The recordings of these raw voltages are carried out across an array of 32 compute nodes in four banks, with each compute node configured to record 187.5 MHz of the bandwidth. We configured our backend with central on-sky frequencies of 4312.5, 5437.5, 6562.5, and 7687.5 MHz for the four overlapping bands. These frequencies were chosen to allow signal at the edges of each passband's intermediate frequency filter to overlap the adjacent passband by 2 × 187.5 MHz, allowing overlap of two compute nodes between banks. This resulted in a total 4875.0 MHz of bandwidth overlapping the entire C-band receiver instantaneous bandwidth of 4100 MHz.

We report observations of the GC carried out in four sessions between 2020 April 10 and 2020 April 25 using the UWL receiver at the Parkes telescope. We conducted observations toward 14 pointings, with three scans each for 10 minutes per scan (Figure 2). We observed the GC (0, 0), labeled as A1, and its immediate surrounding region (A2–A7), using OFF pointings one beamwidth away (B1–B7) to avoid detecting false positives as described above in Section 4.1. The pairs for our 14 pointings are as follows: [A1–B1], [A2–B2], ..., and [A7–B7]. The calibrators and test pulsars used to verify system integrity for most sessions can be found in Table 3.

Observations were recorded on the 26 BL compute nodes that comprise the BL data recorder system at Parkes. Price et al. (2018) provide a detailed summary of the BL data recorder and storage systems at Parkes. As mentioned previously, for these observations, we used the UWL receiver (Hobbs et al. 2020). The UWL digital systems digitize incoming voltages from the receiver at 16 bit resolution (for each pole), channelize the data streams into 26 coarse channels of width 128 MHz, and then output over high-speed Ethernet (total data rate ∼213 Gb s⁻¹). The BL data recorder (Price et al. 2018) captures these data (one 128 MHz sub-band per compute node) and produces spectral products as 32 bit floats (see Table 4). Copies of the data are also sent (via Ethernet multicast) to the primary telescope digital processor, Medusa (Hobbs et al. 2020).

The BL standard reduction pipeline at both the facilities reduces the collected raw-voltage data products into various temporal and spectral resolution products for offline processing. A detailed summary of this BL reduction pipeline and discussion on various products is provided by Lebofsky et al. (2019). These formats are shown in Table 4. As discussed in Section 5.2, the GC region has a large electron density, which is likely to induce large dispersion and scattering for any putative transient signals. At lower frequencies, i.e., lower than 4 GHz with Parkes, these effects are dominating (with also increased noise temperature from the GC) and thus do not allow sensitive searches for transient signals. Therefore, in this paper, we only searched for transient signals (naturally and artificially dispersed) above 4 GHz with the GBT. Moreover, at the GBT, we also produced further high-temporal-resolution products that will be useful for the community to search for pulsars and MSPs above 4 GHz. We plan to publicly release all of our data collected from these observations via the Breakthrough Listen Open Data Portal ¹⁷ .

Narrowband signals have been proposed to be prime candidates for a deliberately transmitted beacon across large interstellar distances (Cocconi & Morrison 1959). We will not discuss the merit of such signals any further here, and instead refer to the earlier literature (see Tarter 2003 for a review). Due to the relative velocity between the transmitter and observer, such narrowband signals exhibit Doppler drifting across the observed band. If the narrowband signal at a fixed frequency is originating from a rotating object, the nonlinear velocity difference will cause the observed frequency of this signal to change nonlinearly. However, for shorter observing lengths, we can approximate such Doppler drifting as a linear chirp. We used turboSETI (Enriquez et al. 2017) to search for such linearly chirped narrowband signals from the high-spectral resolution data products. Here, we used chunks of 2.9 MHz of zero-drift time-summed spectrum to determine the standard deviation (σ). We used a drift rate search in the range of ±4 Hz s⁻¹ with 860 and 1720 steps for the GBT and Parkes telescopes, respectively. For each trial drift rate step, we produced a time-averaged high-spectral resolution dechirp spectrum. Fine channels from this 2.9 MHz chunk that were found to be above 20σ were identified as narrowband drifting signals, or "hits".

Such a range is sufficient to detect drifting introduced by any transmitter located on typical Earth-sized planets, including highly eccentric near-Earth objects (Sheikh et al. 2019). Sheikh et al. (2019) recommend that ideally drift rates in a range of ±200 Hz s⁻¹ should be used to search for such narrowband signals. This range covers observed drift rates from transmitters possibly located on numerous exotic objects such as gaseous planets and planets in highly eccentric obits. However, because searching through such a wide range of drift rates is computationally expensive, here we limit our drift rate search in the range of ±4 Hz s⁻¹, with the caveat that this is by no means a full search of all possible narrowband drifting signals contained in all the observations analyzed in this paper. For example, ongoing efforts to improve our narrowband searching code, turboSETI, have led us to find an improvement in sensitivity that will be tackled in future papers. Due to our specific frequency and time resolution of high-spectral resolution products, signals with higher drift rates will spread across multiple nearby channels within a single spectra, incurring loss in sensitivity equivalent to 1/N, where N is the number of frequency channels over which a single signal is spread (see Margot et al. 2021 for a detailed discussion). However, it is possible to improve sensitivity by a factor of $\sqrt{N}$ for such high drifting signals by adding adjacent frequency bins or by using a varying-width moving boxcar (Price et al. 2020).

Using the search parameters listed in Table 5, we obtained a list of hits for each ON-OFF pair, as defined in Table 7, where each pair consists of a total of six observations—three ON and three OFF. We defined a "hit" as a single strong narrowband signal in an observation that is a fine channel above the set threshold. We found a total of 1,354,519 and 463,068 raw hits from Parkes and GBT observations, respectively. This hit density is relatively lower compared to the hit density obtained at L band from Price et al. (2020), as C band is relatively cleaner from interference and we have used a relatively higher threshold of 20σ. The distribution of these hits in frequency, S/N, and drift rate are shown in Figures 5(a), 5(b), and 6(a), respectively. Due to relatively lower observing frequencies with Parkes, we observed around double the number of total candidates with a roughly similar length of observations. We removed all the hits with a zero-drift rate, to eliminate as much RFI as possible, since any narrowband signals deliberately transmitted by ETIs would likely have some Doppler drift due to the relative velocity differences between the transmitter and the receiver. All the remaining hits, a total of 304,658 candidates combined from Parkes and GBT, are shown in Figures 5 and 6 as nonzero-drift-rate hits.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 5. Narrowband Search Parameters for the BL GC Survey

Parameter	Range
S/N${}_{\min }$ threshold	20
Drift rate search	±4 Hz s−1
GBT frequency range	3.9–8 GHz
GBT trial drift rates	860 steps
Parkes frequency range	1–4 GHz
Parkes trial drift rates	1720 steps
Filter 2	Candidates seen in at least one ON and no OFFs
Filter 3	Candidates seen in all three ONs and no OFFs

Download table as:  ASCIITypeset image

After collating hits from the three ON-OFF pairs, we remove any hits for which at least one of the OFF observations has a hit in the range

$$\begin{eqnarray}&&{\nu }_{\mathrm{bound},\mathrm{OFF}}=\nu \pm | \dot{\nu }| \times 2({\rm{\Delta }}T).\end{eqnarray} \tag{ 1 }$$

Here, ν is the detected frequency of the hit, $\dot{\nu }$ is the maximum drift rate, and ΔT is the length of observations. Any remaining hits were considered an event, where we defined an "event" as a strong narrowband signal that was associated with a hit in at least one ON observation and no OFF observations. We defined these as "filter 2" events, which totaled to 53,702 candidates combined from Parkes and GBT, and they are shown in Figures 5 and 6. Finally, we filtered out hits by removing all those that did not appear in all three ON observations and did appear in any OFF observation. We defined them as "filter 3" events, which totaled to only 249 candidates. These are also shown in Figures 5 and 6.

We plotted all of our event candidates for both filter 2 and 3, which totaled to around 5.4 × 10⁴ candidates, and vetted them by visual inspection. Any strong narrowband signals would appear in all three of the ON observations but none of the OFF observations. Table 7 shows the number of event candidates found for each ON source within each pair for both filter 2 and 3. We ran this process for each pair of observations: 14 Parkes UWL pointings and 19 GBT C-band pointings. Because the OFF observation for each ON observation is strategically placed inside the range of all our pointings, we are able to use this to our advantage by easily making double usage of each observation pair. For example, for the pair B01–B04 in C band, we can run turboSETI twice and look for narrowband drifting signals in two different areas: once using B01 as the ON source, and a second time using B04 as the ON source.

It is important to note that all of our pairs have a total of six observations; however, due to a split in our observation session for the GBT data, we only use four observations total (or two per pointing) for the pairs C07-A00 and C10-C12. Thus, an excess in the number of candidates for these pointings can be seen in Table 7. The rest of the pointings with high numbers of events are attributed to the large amount of RFI found in our data, as can be seen in 6(b), where many events were found to have drift rates close to zero. We did come across a few interesting Doppler-drifting candidates within this analysis; however, most of them can be attributed to RFI, since they appear in at least one of the OFF observations. Some examples of such interesting candidates are shown in Figure 7. Further details on these candidates are discussed in Section 6.1.

Zoom In Zoom Out Reset image size

As suggested in Section 1.1, ETI inhabiting the Galactic Center region might transmit powerful artificially dispersed broadband pulses from an isotropic and/or a rotating beam transmitting all across the Galaxy as a beacon. This would require a similar amount of energy as the isotropic narrowband transmitter we searched for in the previous section. Any broadband signal passing through the interstellar medium experiences dispersion delay Δt due to the free electrons, at the observed frequency of ν compared to the infinite frequency as

$$\begin{eqnarray}&&{\rm{\Delta }}t({DM},\nu )\,\propto \,\displaystyle \frac{{DM}}{{\nu }^{2}}.\end{eqnarray} \tag{ 2 }$$

Here, we can express the arrival time (Δt) and corresponding observed frequency (ν) in quadrants (±Δt, ±ν), also shown in Figure 8. It should be noted that the negative time and negative frequency only represent relative values for our given observation band in terms of respective channel and time sample indexes. For example, a signal with the noted arrival time of Δt = − 1 is only in reference to arrival at Δt = 0 for a signal without any (artificial) dispersion delay. Here, the quadrant 1 (i.e., + Δt and + ν) transient signals are the ones that we refer to as naturally dispersed pulses, which are known to originate due to astrophysical transients such as FRB-like magnetars near the GC. An advanced communicating ETI civilization might choose to artificially disperse a broadband pulse that is not known to occur in nature. These other three classes of transient signals, located in the remaining three quadrants, are nonphysical and can be considered prime candidates for artificially dispersed broadband pulses that are easily distinguishable from the standard naturally dispersed signals. Here, we refer to these artificially dispersed signals as nT ( − Δt, + ν), nF ( + Δt, − ν), and nTnF ( − Δt, − ν). Such signals have been speculated as negative DM signals (von Korff 2010; Siemion et al. 2010; Harp et al. 2018; Li et al. 2020), but to the best of our knowledge, no comprehensive searches have ever been performed. Thus, our analysis is going to be the first of its kind to target such artificially dispersed transient signals.

Zoom In Zoom Out Reset image size

In order to search these four types of broadband transient signals (shown in Figure 8), we have deployed a state-of-the-art GPU-accelerated and ML-assisted broadband signal detection pipeline named SPANDAK. This pipeline is an extension of the pipeline used for the detection of so far the highest known radio-frequency bursts from FRB (Gajjar et al. 2018). More details about the SPANDAK pipeline will be discussed in future publications (Gajjar et al. 2021 in preparation). As mentioned in Section 4.3, we only carried out these transient searches for our GBT data, since scattering and dispersion losses at lower frequencies render such searches less sensitive. Some of the search parameters for our GBT C-band transient search are listed in Table 6. For simplicity, natural DM candidates will be referred to as pDM (i.e., positive DM) candidates and three other types of transient signals—nT, nF, and nTnF—will collectively be referred to as artificial DM (i.e., aDM) candidates for the remainder of the discussion.

Table 6. Transient Search Parameters for the BL GC Survey

Parameter	Range
DM (pc-cm−3)	−5000 to +5000
Maximum sensitivity loss due to DM steps	≤15%
S/N${}_{\min }$ threshold for pDM	6
S/N${}_{\min }$ threshold for aDM	10
Width for pDM	0.38 to 97 ms —with steps 0.38 × 2n ; with n ∈ [0,8]
Width for aDM	0.7 to 194 ms —with steps 0.7 × 2n ; with n ∈ [0,8]

Note. Here, DM ranges refer to both natural and three types of artificial DMs.

Download table as:  ASCIITypeset image

5.2.1. Positive DM Signal Searches

The GC region is likely to harbor an FRB-like magnetar population as suggested in Section 2.4.3. We searched for such transients using SPANDAK with the parameters listed in Table 6 for the mid-temporal resolution data product mentioned in Table 4. We first collated two nearby frequency channels and reduced the frequency resolution by a factor of 2 to get 6656 channels, as SPANDAK has a limitation of processing only up to 8k channels. We searched up to a DM of 5000 pc cm⁻³. With our final temporal and frequency resolution products, the maximum channel smearing we will get for the highest DM is ∼0.38 ms. This is slightly higher than our sampling time for the data products (0.349 ms) used for this analysis. With this lower bound on the possible widths, we searched for bursts with widths up to 90 ms as mentioned in Table 6. We found a total of 4467 pDM candidates from all the pointings shown in Figure 3. A summary of the number of candidates observed from each of our C-band pointings are listed in Table 7. For each of the off-GC pointings (B and C rings) we observed roughly 80 to 100 candidates, which is consistent with our lower S/N${}_{\min }$ threshold and interference environment. As can also be seen from Table 7, around 70% of the total candidates were found toward the GC(0,0) region. Figure 9 shows an example output plot from the SPANDAK pipeline. We visually inspected such plots for 4467 candidates and separated 667 candidates for further inspection. Figures 10(a) and 10(b) show the distribution of all observed candidates from one of the 30 minute scans and the distribution of 667 candidates from our entire observations with a first visual inspection cut. Out of these candidates, we identified 603 candidates from the Galactic center magnetar, SGR J1745-2900, as is clearly evident from Figure 10(b), with a large fraction of these candidates around the DM of 1778 pc cm⁻³. The rest of the 63 candidates were viewed carefully by changing the number of sub-bands, number of time bins, and adjusting the detected DM to verify if the detected signal was real. Figure 11(a) shows one such example of an interesting candidate, which was later rejected. After this careful inspection, we were unable to verify the validity of these 63 remaining candidates and were unable to find any new pDM transient signals from our observations originating from astrophysical processes.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 7. Summary of BL Galactic Center Survey Conducted from the Parkes and GBT Telescopes Across 0.7 to 8 GHz

Parkes: 1.0 to 4.0 GHz
Pair	ON-source	R.A.	Decl.	l	b	f2	f3
	pointing	(h:m:s)	(d:m:s)	(deg)	(deg)	Events	Events
A1-B1	A1	17:45:40.04	−29:00:28.10	359.94	−0.05	291	0
	B1	17:42:34.15	−28:35:25.90	359.94	0.75	9,301	138
A2-B2	A2	17:45:18.09	−29:08:19.17	359.79	−0.05	1,424	0
	B2	17:42:12.06	−28:43:15.04	359.79	0.75	7,149	36
A3-B3	A3	17:46:01.93	−28:52:36.81	0.10	−0.05	1,012	3
	B3	17:42:56.19	−28:27:36.52	0.10	0.75	3,446	27
A4-B4	A4	17:46:00.21	−29:08:32.63	359.87	−0.18	1,116	0
	B4	17:42:54.00	−28:43:32.13	359.87	0.62	3,384	15
A5-B5	A5	17:46:22.11	−29:00:41.14	0.02	−0.18	1,036	0
	B5	17:43:16.04	−28:35:42.55	0.02	0.62	4,935	12
A6-B6	B6	17:44:57.98	−29:00:14.25	359.87	0.09	786	0
	B6	17:41:52.26	−28:35:08.43	359.87	0.88	2,436	9
A7-B7	A7	17:45:19.92	−28:52:23.38	0.02	0.09	802	0
	B7	17:42:14.35	−28:27:19.50	0.02	0.88	4,727	9
GBT: 3.9 to 8 GHz
Pair	ON-source	R.A.	Decl.	l	b	f2	f3	pDM	aDM
	pointing	(h:m:s)	(d:m:s)	(deg)	(deg)	Events	Events	Candidates	Candidates
A00-C01	A00	17:45:40.04	−29:00:28.10	359.944	−0.046	2	0	see below	see below
	C01	17:45:51.95	−28:56:11.99	0.028	−0.046	3,129	3	43	21
C07-A00	C07†	17:45:28.12	−29:04:44.14	359.861	−0.046	1,687	0	51	29
	A00†	17:45:40.04	−29:00:28.10	359.944	−0.046	1	0	see below	see below
B01-B04	B01	17:45:45.99	−28:58:20.05	359.986	−0.046	1	0	52	112
	B04	17:45:34.08	−29:02:36.13	359.906	−0.046	203	0	80	150
B02-B05	B02	17:45:34.57	−28:58:16.40	359.965	−0.010	75	0	113	240
	B05	17:45:45.52	−29:02:39.78	359.923	−0.082	53	0	84	140
B03-B06	B03	17:45:28.61	−29:00:24.42	359.923	−0.010	64	0	83	90
	B06	17:45:51.47	−29:00:31.72	359.965	−0.082	38	0	80	100
C02-C04	C02	17:45:40.52	−28:56:08.37	0.007	−0.010	271	0	73	55
	C04	17:45:23.14	−28:58:12.69	359.944	0.026	457	0	35	58
C03-C05	C03	17:45:29.10	−28:56:04.69	359.986	0.026	291	0	100	65
	C05	17:45:17.18	−29:00:20.67	359.903	0.026	2,038	0	86	27
C08-C06	C08	17:45:39.56	−29:04:47.83	359.882	−0.082	57	0	73	15
	C06	17:45:22.65	−29:02:32.42	359.882	−0.010	582	0	53	31
C11-C09	C11	17:46:02.90	−29:00:35.29	359.986	−0.118	1,605	0	51	23
	C09	17:45:51.00	−29:04:51.46	359.903	−0.118	118	0	46	81
C10-C12	C10†	17:45:56.95	−29:02:43.38	359.944	−0.118	282	16	51	20
	C12†	17:45:57.42	−28:58:23.65	0.007	−0.082	917	0	109	59
Deep	A00†	17:45:40.04	−29:00:28.10	359.944	−0.046	⋯	⋯	3190★	955★

Notes. The R.A., decl., and Galactic coordinates for each ON-source pointing are listed in columns 3–6. Columns 7 and 8 show the number of candidates detected by turboSETI for each ON source, for each filter as described in Section 5.1. The ON sources for which four observations were used instead of the standard six are labeled with a ^†; this only applies to our turboSETI analysis. The last two rows show the number of transient candidates found for both naturally (pDM) and artificially (aDM) dispersed signals. The transient analysis used all three five-minute-long observations per pointing, with the exception of A00, for which we included twelve thirty-minute-long searches as well. These values are labeled with a ^★, and are combined candidates from these deep and shorter observations toward the A00 field. The broadband analysis was not extended to the fields observed from the Parkes telescope, due to excess scattering and channel smearing losses

Download table as:  ASCIITypeset image

5.2.2. Artificial DM Signal Searches

We carried out searches for two different types of beacons that are likely to originate from: (1) a transmitter placed near the GC, illuminating the entire Galaxy; and/or (2) any star in the lines of sight of our pointings toward the GC.

6.1.1. Narrowband Drifting Candidates

6.1.2. Broadband Candidates

In Section 2, we highlighted that a line of sight toward the GC encounters the largest number of habitable stars compared to any other direction in the sky. In this section, we make an order-of-magnitude estimation of the number of stars surveyed based on the stellar number density (stars/pc³) in the radial direction of our survey. Gowanlock et al. (2011) outlined various stellar number density models and found that the Carroll & Ostlie (2007) estimation of the number density profile appears to match the observed stellar density in the Solar neighborhood. This number density is represented as a function of radial distance from the GC, R, and vertical height from the Galactic midplane, Z, as

$$\begin{eqnarray}&&n(Z,R)={n}_{0}\,({e}^{-Z/{Z}_{\mathrm{thin}}}\,+\,0.085{e}^{-Z/{Z}_{\mathrm{thick}}}){e}^{-{\rm{R}}/{{\rm{h}}}_{R}}\,\mathrm{stars}/{\mathrm{pc}}^{3}.\end{eqnarray} \tag{ 3 }$$

Here, n₀ is the normalization factor, given by Gowanlock et al. (2011) as 5.502 stars pc⁻³. The constants Z_thin = 350 pc, Z_thick = 1000 pc, and h_R = 2.25 kpc are thin-disk scale height, thick-disk scale height, and radial scale, respectively. As can be seen, although stellar density falls off very quickly in the Z direction, for smaller angular scales, it can be considered to have constant stellar density. For example, at the distance of 8.5 kpc with our survey region of $4^{\prime} \times 4^{\prime}$, the diameter of the region probed would be around 10 pc, for which the number density reduces by only 2.5% (from 5.96 to 5.81 stars pc⁻³). This is not a significant change; thus, we use the midplane density and take it as a constant for the Z scale height.

To calculate the total number of stars, we considered a cone in the Galactic midplane toward the GC with the Sun at the apex. The angle of the cone changes as a function of our observing frequencies and proposed target region. We considered stars up to a distance of 8.5 kpc from Earth—this includes the central Galactic bulge, which has the largest concentration of stars. For our current $4^{\prime} \times 4^{\prime}$ region, if we count the number of stars from Earth to the GC, we will encounter ∼6 × 10⁵ total stars. This number will decrease as we go to higher frequencies with the GBT, due to our reduced survey regions (see Table 2). For our already conducted survey of the Parkes region, which covers two $30^{\prime} \times 30^{\prime}$ fields (see Figure 2), we estimated that we surveyed around 60 million stars. The large concentration of these total surveyed stars are likely to reside close to the GC region. As indicated in Figure 1, a large fraction of these stars, especially the ones that are closest to the GC, may host habitable planets, and thus our survey provides some of the best constraints on the two types of beacons from the large number of habitable systems. We should highlight that these are only order-of-magnitude estimates and gauging the exact numbers, especially for Parkes (2° × 4°), is a nontrivial task due to the validity of our assumed stellar number density model and small angle approximations.

We have conducted searches for two different signal types, and limits can be placed on our detection sensitivity to both these signal types. For narrowband drifting signals, the minimum detectable flux can be given as

$$\begin{eqnarray}&&{S}_{\min ,\mathrm{narrow}}=\displaystyle \frac{S/{N}_{\min }}{\beta }\displaystyle \frac{{S}_{\mathrm{sys}}}{\delta {\nu }_{t}}\sqrt{\displaystyle \frac{\delta \nu }{{n}_{p}{\tau }_{\mathrm{obs}}}}.\end{eqnarray} \tag{ 4 }$$

Here, S_sys is the system equivalent flux density, τ_obs represents integration time, n_p is the number of polarizations, and δ ν_t and δ ν are bandwidths of transmitted and received narrowband signals, respectively. For our survey, we assumed a transmitted bandwidth of 1 Hz ¹⁸ . S/N${}_{\min }$ represents a desired threshold, and β is the dechirping efficiency. As mentioned in Section 5.1, due to our frequency (δ ν ∼ 3 Hz) and temporal (δ t ∼ 18 s) resolutions, high drift rate signals get spread across multiple channels, which reduces our sensitivity. This dechirping efficiency can also be expressed as follows for our survey:

$$\begin{eqnarray}\beta =\left\{\begin{array}{ll}1 & | \dot{\nu }| \leqslant 0.16\,\mathrm{Hz}\,{{\rm{s}}}^{-1}\\ \displaystyle \frac{3}{| \dot{\nu }| \times \,18} & | \dot{\nu }| \gt 0.16\,\mathrm{Hz}\,{{\rm{s}}}^{-1}.\end{array}\right.\end{eqnarray} \tag{ 5 }$$

Here, $\dot{\nu }$ is the trial drift rate under consideration. Our dechirping efficiency reaches its maximum for drift rates ≤0.16 Hz s⁻¹ and gradually declines beyond this limit. As mentioned in Section 5.1, future narrowband searches will improve this efficiency by collapsing nearby channels or by using a moving boxcar of corresponding width for the number of channels a given signal is likely to have spread across (Price et al. 2020).

For the broadband pDM and aDM transient signals, the minimum detectable flux can be expressed as

$$\begin{eqnarray}&&{S}_{\min ,\mathrm{transient}}=S{/N}_{\min }\displaystyle \frac{{S}_{\mathrm{sys}}\beta }{\sqrt{{n}_{{\rm{p}}}\ {\rm{\Delta }}{\tau }_{\mathrm{pulse}}\ {\rm{\Delta }}\nu }}.\end{eqnarray} \tag{ 6 }$$

Here, Δτ_pulse is the pulse width of pDM and aDM candidates, Δν is the total instantaneous bandwidth, and β is the loss due to digitization. For pDM candidates we used an S/N${}_{\min }$ threshold of 6, and for aDM candidates we used an S/N${}_{\min }$ threshold of 10. For simplicity, we considered scatter-independent pulse widths. A more accurate width estimation based on the scatter broadening and corresponding sensitivity limitation are discussed further in Section 6.6. For both the signal types (narrowband and broadband), the S_sys can also be represented as

$$\begin{eqnarray}&&{S}_{\mathrm{sys}}=\displaystyle \frac{2k({T}_{\mathrm{sys}}+{T}_{\mathrm{GC}})}{{A}_{\mathrm{eff}}}.\end{eqnarray} \tag{ 7 }$$

Here, T_sys is the system temperature at typical weather conditions ¹⁹ , k is the Boltzmann constant, and A_eff is the effective aperture of the receiving antenna. The GC region contains a significant amount of background flux, especially with the lower-frequency receivers (up to Ku band). The excess noise contribution, which is expressed as T_GC, will make a significant additional contribution to the S_sys (Johnston et al. 2006; Macquart et al. 2010), and needs to be included in the sensitivity calculations. Law et al. (2008) conducted a detailed continuum survey of the GC region using the GBT and computed a calibrated map of the region. Rajwade et al. (2017) used these measurements and derived the frequency dependence of this excess noise from the GC as

$$\begin{eqnarray}&&{T}_{\mathrm{GC}}=\displaystyle \frac{568}{{\nu }_{\mathrm{GHz}}^{1.13}}\,{\rm{K}}.\end{eqnarray} \tag{ 8 }$$

Here, ν_GHz is the observing frequency. As can be seen, this will add a huge contribution to the system temperature at lower frequencies. For example, at 0.7 GHz from the UWL at Parkes, the expected noise contribution from the GC would be on the order of 800 K. Considering this contribution, our survey sensitivity is listed in Figure 12 for both narrow and broadband signals. As it can be seen, Ku and KFPA bands provide maximum sensitivity for both types of signals, with contribution from the GC dominating at lower frequency and contribution from the Earth's atmosphere dominating the T_sys at higher frequencies. Based on our minimum ${{\rm{S}}}_{\min ,\mathrm{narrow}}$ and ${{\rm{S}}}_{\min ,\mathrm{broad}}$ at various frequencies, we can calculate the corresponding Equivalent Isotropic Radiated Power (EIRP) at all our observing bands as

$$\begin{eqnarray}&&{\mathrm{EIRP}}_{\mathrm{GC};(\mathrm{narrow},\mathrm{broad})}={S}_{\min ;(\mathrm{narrow},\mathrm{broad})}\times 4\pi {D}_{{GC}}^{2}.\end{eqnarray} \tag{ 9 }$$

Zoom In Zoom Out Reset image size

For transmitters at the distance of the GC, the minimum EIRP_GC are shown in Figure 12. It is evident that the minimum EIRP sensitivities of 3 × 10¹⁷ W and 4 × 10¹³ W/Hz are possible to achieve at the KFPA band for the narrowband and broadband transient signals, respectively. For our current survey with the GBT C band and Parkes-UWL, we estimate a minimum EIRP detection sensitivity of around 5 × 10¹⁷ W (≥10¹⁹ W for the signals with highest drift rates) and 4 × 10¹⁸ W (≥ 10²⁰ W for the signals with highest drift rates) for the narrowband drifting signals, respectively. Similarly, for the broadband artificially dispersed signals from the GBT C-band survey, we estimate a minimum detectable EIRP of the order of 1 × 10¹⁴ W/Hz. It is clearly evident that broadband signals allow much better sensitivity to provide better constraints on the weaker signals if they are band-limited. However, since such signals are assumed to last for only a fraction of a second, we rely upon the repeatability of such signals, which is further discussed in Section 6.5.

There have been several SETI surveys that have been conducted in the past six decades (Enriquez et al. 2017; Wright et al. 2018, and references therein). These surveys have primarily searched for narrowband drifting signals, and thus they can be compared with each other based on the total number of stars they surveyed (or total area of the sky covered), total bandwidth utilized, detection sensitivity, and other such parameters. For narrowband signals, the Drake figure of merit (DFM; Drake et al. 1984) has been computed by many previous SETI surveys. It can be expressed as

$$\begin{eqnarray}&&\mathrm{DFM}=\displaystyle \frac{{\rm{\Delta }}\nu \,{\rm{\Omega }}}{{S}_{{\min },{avg}}^{\tfrac{3}{2}}},\end{eqnarray} \tag{ 10 }$$

where Ω is the area of the sky surveyed, ${S}_{{\min },{avg}}$ is the minimum detectable flux average across all observing frequencies, and Δν is the total observing bandwidth. For our survey with the GBT across 3.9–8 GHz, we found the DFM to be ∼4 × 10²⁸ for the narrowband signals with drift rates ≤0.16 Hz s⁻¹ and ∼4 × 10²⁶ for the highest drift rate considered for our survey. It should be noted that the DFM has a number of limitations as outlined by Enriquez et al. (2017) and Margot et al. (2021). For example, the DFM does not consider a range of drift rates and dechirping efficiencies, which heavily impacts ${{\rm{S}}}_{{\min },{avg}}$ for a survey under consideration (Margot et al. 2021). Moreover, the DFM assumes a uniform distribution of ETI transmitters. For surveys like ours, where a large concentration of stars are likely to reside within our beam toward the GC, the DFM does not provide an accurate comparison. A better figure of merit is the Continuous Waveform Transmitter Rate (CWTFM), as introduced by Enriquez et al. (2017), which is also for narrowband signals and can be given as

$$\begin{eqnarray}&&{CWTFM}={\zeta }_{\mathrm{AO}}\displaystyle \frac{{EIRP}}{{N}_{\mathrm{stars}}{\nu }_{\mathrm{frac}}}.\end{eqnarray} \tag{ 11 }$$

Here, ν_frac is the fractional bandwidth Δν/ν, ζ_AO is the normalization factor such that CWTFM = 1 when the EIRP is equivalent to the EIRP of the Arecibo radio telescope's planetary radar (∼10¹³ W), ν_frac = 0.5, and N_stars = 1000. For our conducted survey, as discussed in Section 6.2, we can estimate the number of stars using the stellar density model and estimate the corresponding EIRP limits (see Section 6.3). We found that, for the UWL and C band, we surveyed around 60 million and 0.6 million stars, respectively. Figure 13 shows the measured CWTFM for our current C-band and UWL surveys compared with some of the prominent SETI surveys reported earlier. It is clearly evident that, with our current survey, we are already able to push the limit of transmitter-rate slope lower than what was obtained from previous SETI surveys. By extending this survey to other frequencies from the GBT and covering all the pointings from the Parkes telescopes, we are likely to push the CWFTM slope much lower and able to constrain the presence of technosignatures across a wide range of frequencies. Furthermore, Figure 12(a) shows a comparison of the sensitivity of narrowband signals of our survey with some of the prominent SETI surveys. As is clearly evident, no other survey has ever explored the frequency coverage we aim to search with our survey. Moreover, these comparisons are only carried out for narrowband drifting signals. Our survey is already the only survey to search for artificially dispersed signals that have never been explored, and thus it cannot be compared with any previous SETI surveys.

Zoom In Zoom Out Reset image size

We searched for three different types of artificially dispersed broadband transient bursts likely to originate from deliberately transmitted ETI beacons. In future work, we will provide a detailed discussion on harmonic searches for these broadband signals. In this section, we evaluate the repetition rate of these broadband bursts from our GC (0, 0) deep observations. Figure 14 shows one of the scenarios of repeating broadband signals originating from a rotating beam from an artificial beacon ²⁰ . We can also infer different mechanisms that do not necessarily require a rotating beam but can still generate repeated bursts of similar nature. Such rotating beams of an artificial nature have been proposed earlier. For example, Benford et al. (2010a) suggested such a beacon situated at the GC would be the most energy-efficient way to signal the entire Galaxy. Shostak & Tarter (1985) have also considered narrowband signals originating from such a rotating beacon. We did not find any strong aDM candidates; thus, considering length and separation between all our observations, we can reject repetition of all three types of broadband signals. Table 3 lists all the observing dates and lengths of each of these observations. We arranged them on the time axis and then simulated pulse trains with a range of periods. We also adjusted the phase (or offset) of these bursts within the corresponding period under consideration. For each period, we calculated the number of instances of offset phases that would have allowed us to see at least one pulse. Here, we assumed that we are likely to detect a burst at every period. Figure 14 shows the probability of rejecting periods for such bursts. We can reject any mechanism that can regularly produce pulses across 3.9 to 8 GHz with a repetition period of around 4.3 hr with high accuracy (considering imperfect recovery rate of our search pipeline). For larger repetition periods up to 10 hr, we can reject such a mechanism with ≥50% probability, while for further large periods up to 100 hr, the rejection probability is marginal (<10%) from our observations. In the future, we plan to extend this analysis for narrowband signals and will also schedule our observations such as to allow sufficient epoch sampling to reject larger repeating periodicity with better significance. These limits are nontrivial to apply to astrophysical sources such as magnetars, which are not known to produce strong transient pulses during every rotation and their beam of emission is not necessarily likely to reside in the Z = 0 plane as we assumed for this analysis.

Zoom In Zoom Out Reset image size

We detected 603 single pulses from SGR J1745–2900 across 3.9 to 8.0 GHz with a total of 6.6 h of observations. These single pulses exhibit a wide variety of pulse shapes with single and multiple components. Figure 15 shows examples of four different types of single pulses seen from SGR J1745–2900 from our observations: a narrow single pulse with a scattering tail, a pulse with unresolved multiple components, and two well-resolved dual-component pulses. This is consistent with single pulses seen around similar frequencies by Pearlman et al. (2018) and Wharton et al. (2019). A detailed spectro-temporal analysis of these bursts is beyond the scope of this paper and will be reported in future publications. These observations are useful to constrain the presence of other radio magnetars at the GC producing transient bursts.

Zoom In Zoom Out Reset image size

Scattering due to a relatively dense environment near the GC likely limits the detection of other transients. In order to estimate scatter-broadened pulse widths at our observed frequencies, we utilized a narrow single-peaked pulse from SGR J1745–2900 shown in Figure 15 for further analysis. Spitler et al. (2014) estimated that the scatter broadening (τ_d ) from SGR J1745–2900 is on the order of 1.3 ± 0.2 sec at 1 GHz with a spectral index of α = − 3.95 ± 0.2. Considering a similar spectral index, the expected τ_d will be ∼1.4 ms at 6 GHz. Measuring the observed τ_d is challenging, as suggested by Krishnakumar et al. (2017), due to the degeneracy of the unknown intrinsic pulse width and unresolved emission components. To model our underlying observed single pulse, we assumed a simple intrinsic single pulse with a Gaussian profile from the source, convolved with the $s(t)={e}^{-t/{\tau }_{d}}U(t)$ for transmission through the ISM, where U(t) is a step function with U(t) = 0 for t < 0 and U(t) = 1 for t ≥ 0. For the intrinsic pulse, we considered a pulse width of 1.8 ms reported from Pearlman et al. (2018) at 8.4 GHz, which is close to the highest end of our observed band. Similar techniques have been used by Krishnakumar et al. (2017) and Krishnakumar et al. (2019) to estimate τ_d . We found the τ_d ≈ 4 ms by fitting this model to the narrow pulse shown in Figure 15. This is slightly higher and not consistent with the spectral index of −3.8 reported earlier at lower frequencies by Spitler et al. (2014). Similar inconsistencies were also reported by Pearlman et al. (2018). It is also likely that the larger scatter broadening is due to an unresolved emission component at the trailing edge of the pulse.

The scatter broadening timescale found from the narrow SGR J1745–2900 burst can be utilized to place a limit on other FRB-like Galactic magnetars and millisecond pulsars (MSPs) located in a similar environment around the GC. We re-evaluated our minimum detection limit for the broadband signals calculated in Section 6.3. We find that, for our current survey across 3.9 to 8 GHz for transients, we were sensitive to bursts with peak flux density of around 68 mJy, which is equivalent to the peak luminosity of ∼10³¹ erg s⁻¹ across our band for a source located at the GC. Although average flux densities of folded profiles of radio-loud magnetars are relatively lower, single pulses from them do exhibit bright emission. For example, XTE J1810–197, a known Galactic radio magnetar, was seen to produce bright transient radio pulses up to 42 GHz (Camilo et al. 2006). XTE J1810–197 showed many single pulses with a peak flux of ≳10 Jy at 6.4 GHz, which corresponds to a peak luminosity of around 7 × 10³¹ erg s⁻¹ across their 500 MHz band. Similarly, another radio-loud magnetar, PSR J1622–4950, is known to show bursts with a peak flux of around 16 Jy (Levin et al. 2010b), which corresponds to a peak luminosity of 5 × 10³² erg s⁻¹. This is an order of magnitude larger than our limit. Moreover, the recently detected FRB-like Galactic magnetar, SGR 1935+2154, also exhibits a bright luminosity on the order of 10³⁶ erg s⁻¹ (The CHIME/FRB Collaboration et al. 2020), which is several orders of magnitude larger than our peak luminosity limit. Since our longest observing run was about 4.3 h (see Table 3), we can reject the presence of any transient signal with a typical burst-rate of ≥0.23 bursts h⁻¹ at our radio luminosity limit. A few of the radio magnetars are known to produce transient radio bursts with much larger burst rates. For example, SGR J1745–2900 produced ≳10 bursts hr⁻¹, as measured from our observations.

Our scatter-broadened limits are also important, as many different theories have been proposed to solve the missing pulsar problem. For example, the GC has been considered to favor MSP production (Bartels et al. 2016). As we found a scatter broadening timescale of ∼4 ms, any MSP signals with much smaller intrinsic pulse widths than normal pulsars are likely to get scattered and will therefore be undetectable up to our highest frequency of 8 GHz. Extension of our survey to higher frequencies would be ideal to search for other such MSP near the GC.