A 4–8 GHz Galactic Center Search for Periodic Technosignatures

Radio searches for extraterrestrial intelligence have mainly targeted the discovery of narrowband continuous-wave beacons and artificially dispersed broadband bursts. Periodic pulse trains, in comparison to the above technosignature morphologies, offer an energetically efficient means of interstellar transmission. A rotating beacon at the Galactic Center (GC), in particular, would be highly advantageous for galaxy-wide communications. Here, we present blipss, a CPU-based open-source software that uses a fast folding algorithm (FFA) to uncover channel-wide periodic signals in radio dynamic spectra. Running blipss on 4.5 hr of 4–8 GHz data gathered with the Robert C. Byrd Green Bank Telescope, we searched the central 6′ of our galaxy for kHz-wide signals with periods between 11 and 100 s and duty cycles (δ) between 10% and 50%. Our searches, to our knowledge, constitute the first FFA exploration for periodic alien technosignatures. We report a nondetection of channel-wide periodic signals in our data. Thus, we constrain the abundance of 4–8 GHz extraterrestrial transmitters of kHz-wide periodic pulsed signals to fewer than one in about 600,000 stars at the GC above a 7σ equivalent isotropic radiated power of ≈2 × 1018 W at δ ≃ 10%. From an astrophysics standpoint, blipss, with its utilization of a per-channel FFA, can enable the discovery of signals with exotic radio frequency sweeps departing from the standard cold plasma dispersion law.

Broadband transient pulses enable signal detection agnostic to the choice of receiver center frequency (Clancy 1980).While such pulses experience dispersion and scattering from astrophysical plasma along the line of sight, ETI might introduce negative dispersion as a means of adding artificiality (Demorest et al. 2004;Siemion et al. 2013).For example, broadband radio pulses propagating through the interstellar medium undergo cold plasma dispersion, resulting in chromatic pulse arrival times, t a ∝ ν −2 .By introducing negative dispersion, ETI can intentionally overcorrect for astrophysical dispersion and transmit artificial bursts with delayed pulse arrival at high radio frequencies.Alternatively, artificiality can be incorporated through a repeating non-physical burst sequence such as a Fibonacci series (e.g., Clement et al. 2022).
Allowing for both natural and artificial dispersion, Gajjar et al. (2022) conducted extensive searches for broadband pulsed beacons from 1883 stars in the Milky Way.Their investigations reveal that 1 in 1000 stars transmit pulses repeating ≤ 500 s with equivalent isotropic radiated power (EIRP) 10 22 W at 4-8 GHz.Moreover, the power budget for transmission can be significantly lowered by incorporating periodicity in pulse emission.
Periodic pulsed beacons offer an inexpensive means of communication across vast interstellar distances.Sullivan (1991) proposed a set of plausible pan-Galactic SETI periods in the 10 −6 -10 5 s range, some of which are based on known pulsar periods in our galaxy.Further, Edmondson & Stevens (2003) suggested that ETI might potentially broadcast signals mimicking millisecond pulsar periodicities for interstellar communication.Transmission of such signals from their respective anti-pulsar directions relative to the Earth would then indicate their artificial origin.
On the Earth, periodic signals from human technology find extensive utility as radar for remote sensing, telecommunication, and aircraft navigation.Notable examples of air control radar include the Punta Salinas1 and TPS-75 2 radar operating in the United States between 1.3-1.7 and 2.9-3.1 GHz respectively.The former, with a radar rotation period of about 12 s, Though periodic signals are indispensable features of human technology, radio searches for periodic ETI transmissions have been scarce.Gray & Ellingsen (2002) looked for periodic transmissions in the locale of the "Wow!" signal, a strong narrowband emission detected near the 21 cm HI line in 1977 (Kraus 1979).Harp et al. (2018) conducted an autocorrelation search for periodic extraterrestrial signals in 1-10 GHz data gathered with the Allen Telescope Array in California, USA.
Here, we present the Breakthrough Listen Investigation for Periodic Spectral Signals (blipss), a novel software architecture utilizing a Fast Folding Algorithm (FFA: Staelin 1969;Morello et al. 2020) to enable routine searches for periodic technosignatures in radio data.Our target signals are channel-wide periodic pulse trains, an example of which is illustrated in Figure 1.Section 2 presents an energy budget comparison that evaluates the feasibility of ETI transmissions of our target signals against that of narrowband continuous-wave (CW) beacons and artificially dispersed broadband bursts.
Of all directions in the sky, the line of sight towards the Galactic Center (GC) offers the greatest propensity for the emergence of intelligent life (Gajjar et al. 2021).Further, radial beacons from the GC are viable given their advantageous positioning for galaxy-wide communications (Benford et al. 2010).Thus, the GC is an ideal first target for our periodic technosignature searches.Section 3 describes our 4-8 GHz observations undertaken as part of the Breakthrough Listen (BL) GC survey (Gajjar et al. 2021).We detail the methodology of blipss in Section 4. In Section 5, we present the findings obtained from running blipss on our GC data.Section 6 highlights the implications of our study with emphasis on its potential outstanding gains for SETI.Finally, we summarize and conclude our investigation in Section 7.

SIGNAL TRANSMISSION ENERGY BUDGETS
Consider an extraterrestrial antenna at the GC broadcasting three signal morphologies, i.e., steady Hz-wide CW beacons, broadband millisecond pulses, and channel-wide periodic pulse trains.For effective aperture η tx = 0.7 and telescope diameter D tx = 100 m similar to Robert C. Byrd Green Bank Telescope (GBT), the antenna gain is G tx ≈ η tx (πD tx c/ν) 2 ≈ 2.8×10 7 at ν = 6 GHz.Here, c is the vacuum speed of light.
Say that we receive extraterrestrial transmissions at the Earth using the GBT with systemequivalent flux density, S sys,GC ≈ 46.3 Jy (Suresh et al. 2022) towards the GC.For simplicity, let us assume that the transmitter and the receiver are perfectly aligned, have identical operating radio frequency bands, and have negligible signal digitization losses.
The minimum detectable EIRP of a narrowband CW beacon transmitted by the extraterrestrial antenna is (Enriquez et al. 2017 Assuming perfect CW beacon detection, we take β = 1 throughout Section 2. Equation 1 assumes a ∆ν ch tuned to B narrow for optimal detection.Using an integration time T int = 5 minutes and requiring a threshold signal-to-noise ratio (S/N) min = 10 for detection, EIRP min narrow ≈ 1.5 × 10 17 W.Over a transmitter operation lifetime L, the net energy cost associated with transmission of a Hz-wide CW signal from the GC is Similarly, for a single broadband burst (Gajjar et al. 2022) The blue and solid black curves both presume a pulse duty cycle, δ = 10 −5 .In contrast, the dashed and dotted black curves correspond to δ = 0.1 and 0.5 respectively.The red line assumes 100% signal dechirping efficiency.Finally, the red line and all black curves reflect numbers for an integration time of 5 minutes.
Here, we consider T int ≥ 3P for periodicity detection.The quantities E FFA ≈ 0.93 and ∆ν ch ≈ 2.86 kHz are, respectively, the FFA search efficiency and our channel bandwidth (see Section 3).Detection of periodic signals using a per-channel FFA requires that the channel bandwidth be matched to the signal bandwidth.
Transmission of kHz-wide periodic signals for detection with kHz-wide channels offers an advantageous arrangement for both the transmitter and the receiver.Where such a proposition allows the transmitter to avoid the elevated power costs of broadband signal transmission, the receiver benefits from negligible Doppler signal drift across channels.Assume that ETI correct for their own Doppler motion relative to the solar system barycenter.The largest contribution to any observed signal drift then comes from the Earth's daily rotation, which produces a maximum Doppler drift rate of ≈ 0.67 Hz/s (see Section 3 of Price et al. 2020) at 6 GHz.Over T int = 5 minutes, the resulting signal drift is ≈ 200 Hz ∆ν ch ∼ kHz.
While our target signals are kHz-wide periodic pulse trains, a per-channel FFA is also sensitive to periodically modulated CW signals.
For instance, consider a CW beacon turned on for 1 ms duration.The ensuing transmission has a sinc frequency response with a 1 kHz main-lobe bandwidth, i.e., comparable to our channel bandwidth.Periodic repetition of the modulation then produces a signal detectable via our per-channel FFA.
To facilitate consistent comparison between Equations 4 and 6, let us assume δ = 10 −5 , implying t pulse = 1 ms for P = 100 s.Then, EIRP min periodic ≈ 2.8 × 10 16 W. Applying Equation 5 to channel-wide periodic signals, Figure 2 compares transmission energy budgets at L ≥ 300 s for S/N = 10 detection of our three kinds of signals.Evidently, channel-wide periodic signals with δ = 10 −5 offer the least energy cost for the transmitter.Hence, searches for periodic spectral signals from the GC present a promising avenue of exploration for SETI.

OBSERVATIONS
The BL GC survey, conducted by the Breakthrough Listen Initiative (Worden et al. 2017;Gajjar et al. 2019), is an extensive 0.7-93 GHz campaign of the GC and neighbouring Galactic bulge fields for technosignatures, pulsars, bursts, spectral lines, and masers (Gajjar et al.  Figure 3 shows the 4-8 GHz BL GC survey footprint tiled by 19 distinct spatial pointings arranged in 3 concentric rings.From inner to outer, these rings are named A, B, and C respectively.All pointings made use of the singlebeam C-band receiver, resulting in a half-power beam width (HPBW), θ HPBW ≈ 2. 5 at 6 GHz.Table 1 presents an overview of our observing program comprised of alternating scans between pairs or triplets of pointings with centers separated by at least 2θ HPBW .We also observed a test pulsar at every observing epoch to verify our system integrity.
All observations utilized the Breakthrough Listen Digital Backend (MacMahon et al. 2018;Lebofsky et al. 2019) to channelize voltage data to filterbank products of various spectral and temporal resolutions.We refer readers to Table 4 of Gajjar et al. (2021) for details of data products generated by the BL GC sur-vey.Here, for our target fields, we utilized the mid-spectral resolution data products having ≈ 1.073 s time sampling.With a channel bandwidth of ≈ 2.86 kHz, these data contain 1,703,936 channels covering 3.56-8.44GHz, which overlaps the 3.9-8.0GHz operating range of the C-band receiver.Our test pulsar data come in the mid-time resolution format with ≈ 349.53 µs sampling time, ≈ 366.21 kHz channel bandwidth, and 13,312 channels spanning 3.56-8.44GHz.
Working with 4-8 GHz BL GC survey data, Gajjar et al. (2021) and Suresh et al. (2021Suresh et al. ( , 2022) ) reported the presence of significant radio frequency interference (RFI) between 4.24-4.39GHz.Clipping bandpass edges and masking RFI-affected channels, we restricted our radio frequency coverage to 4.39-7.81GHz, thus obtaining 3.42 GHz of usable bandwidth for our data analyses.

BLIPSS WORKFLOW
The FFA is a phase-coherent time-domain search for periodic signals (Staelin 1969;Morello et al. 2020).In comparison to traditional Fourier-domain periodicity searches involving incoherent harmonic summing, the FFA provides enhanced search sensitivity to short pulse duty cycles (δ 0.1).Combining the FFA with boxcar matched filtering of folded profiles, Morello et al. (2020) developed riptide, a software package for real-time radio pulsar searches.Here, we adapted utilities from riptide to build blipss 3 , a CPU-based Python package to uncover periodic pulsed extraterrestrial signals in radio data.
Figure 4 presents a flowchart of the blipss workflow split into three stages.
As in- 3 Code available at https://github.com/UCBerkeleySETI/blipss.put, our pipeline accepts one or more RFImasked filterbank files.Dynamic spectra (radio frequency-time data) from these files are processed channel-wise to enable periodic signal detection regardless of signal bandwidth and dispersion.
Consider a sequence of N filterbank files generated via position switching between two hypothetical targets X and Y. Say that odd and evennumbered files contain data of sources X and Y respectively.In the first stage of our workflow, all data are detrended channel-wise using a running median filter to remove slow baseline fluctuations.This red noise removal step is essential to enhance our sensitivity to long periods.Next, the data per channel are searched for periodic signals using the FFA.For every channel in each input data set, periodicity candidates are then grouped into sets of harmonically related periods.Within every such set, the candidate period with the highest detection S/N is labeled as the fundamental.Harmonics and subharmonics of the fundamental frequency are then defined in the Fourier sense as follows.
Here, 1/H m and 1/S n are, respectively, the m-th harmonic and the n-th subharmonic of the fundamental frequency, f 0 = 1/P 0 .The final output from our first processing stage is a list of periodicity candidates detected per channel for every input file.Key user-supplied parameters to the FFA include the minimum and the maximum trial periods, the shortest and the largest trial duty cycles, and the chosen duty cycle resolution of the search.Given a range of trial periods and a sampling time, spacings between consecutive trial periods are determined using Equation 5of Morello et al. (2020).In the second stage of our processing, we assign N -digit binary codes to periodicity candidates through comparisons between on-target and off-target scans.A "1" at the j-th element of a code (read from left to right) denotes candidate detection in the j-th input filterbank file.Similarly, zeroes indicate candidate non-detections in corresponding files.Say that N = 6 in our example of nodding observations between sources X and Y.A periodicity candidate with code 101010 would then be a SETI event of interest seen in all scans of target X and undetected in scans of target Y.
At our final processing step, we apply filters to shortlist candidates for visual inspection of their folded profiles.We define "filter n" (F n ) events as candidates that are detected in exactly n scans of the target field, and and are unseen in all off-target scans.In our example with N = 6, ON target X, and OFF target Y, candidates with binary code 101010 constitute the set of F 3 events.Figure 5 shows a sample F 3 candidate generated using simulated data and detected with blipss.A P = 30 s candidate is evident in all scans of source X, but unseen in scans of source Y.

Test Pulsar Detection
We verified the integrity of our pipeline by confirming the detection of the two test pulsars observed in our survey.First, we detrended our data in every channel using a running median window of width 2 s.For a pulsar with rotational period P 0 , we then searched for pulsations between 0.3P 0 and 2.5P 0 .Further, mandating a minimum N bins = 128 phase bins across a folding period, we obtained a duty cycle resolution of 1/N bins ≈ 0.78%.Allowing for boxcar widths extending up to 5 phase bins, we conducted  Figure 6.Detection of test pulsar B0355+54 using blipss.Left: Broadband signal detection at known pulsar rotational period, P 0 ≈ 156.37 ms.We obtain our highest matched filtering S/N values (color scale) at radio frequencies where interstellar scintillation best brightens the pulsar emission.The first subharmonic (P ≈ 312.75 ms) and the first two harmonics (P ≈ 78.19, 52.12 ms) of the pulsar rotational frequency are also detected.In addition, we detect the pulsar at rational multiples of P 0 near 5.2 GHz, where the pulsar emission is at its brightest in our band.These include detections at P ≈ 2P 0 /3 and P ≈ 3P 0 /2.Right: Phase-resolved pulsar spectrum revealing pulse dispersion from radio wave propagation through the interstellar medium.Since our search strategy does not assume a dispersion law, blipss can enable discovery of periodic signals with radio frequency drifts that differ from the often assumed cold plasma dispersion law (arrival times ∝ ν −2 ).
matched filtering searches for folded profiles with duty cycles between 0.78% and 3.91%.Setting a matched filtering S/N threshold of 25, we successfully detected both test pulsars B0355+54 and B1133+16 using blipss.
Figure 6 shows our detection of pulsar B0355+54 from epoch 1 of observations.Notably, our algorithm is agnostic to any radio frequency drifts exhibited in the dynamic spectrum.Therefore, blipss can enable discovery of periodic pulsed beacons with exotic swept frequency structures that are frequently missed by traditional broadband pulse search techniques assuming cold plasma dispersion (e.g., Lazarus et al. 2015;Gajjar et al. 2021Gajjar et al. , 2022)).fields.Visually noticing the presence of significant power-law red noise at f 0.1 Hz in power spectra of our channelized time series, we selected a running median filter of width, W med = 12 s, for time series detrending.Desiring at least N bins = 10 bins across a folding period, we adopted a duty cycle resolution of 10%  7).

blipss Searches of Target Fields
e Final two A00 scans were used as OFF for the two C07 scans.
for our FFA searches.Further, we set the shortest trial period of our searches at P min = 11 s to meet the requirement P min ≥ N bins t samp , where t samp ≈ 1.073 s is the sampling time of our data.Necessitating a minimum of three full periods within our data length, our scan duration of 5 minutes limits the maximum trial period of our FFA searches to P max = 100 s.Trialing boxcar filters spanning 1-5 phase bins, we explored pulse duty cycles, δ ∈ [10%, 50%] in our FFA searches.
As indicated in Table 2, we adopted matched filtering S/N thresholds of 7 and 6, respectively, for our ON and OFF pointings.Here, motivated by Enriquez et al. (2017), we chose a lower S/N threshold for our OFF scans to identify potential RFI candidates that may be weaker in our OFF observations.Such attenuation may occur for signals detected via telescope side lobes.Finally, we visually inspected folded profiles of all F n events, i.e., candidates found exclusively in ON pointings.

RESULTS
Table 3 summarizes the results of our FFA searches, yielding a total of 1,518,056 F 1 candidates and 1,213 F 2 events of interest across all pointings.Considering all possible binary codes, we notice significant variability in candidate counts across pointings.This finding likely suggests a dynamic RFI environment at the GBT.For instance, pointings A00, C01, and C07 that were jointly observed during epoch 1 (2019 Aug 07) have only about 200-400 candidates each.In contrast, pointings observed during epoch 2 have 1000 candidates each.Of note, all B-ring pointings with 10 6 can-didates were observed in the first 1.5 hours of epoch 2.
In Figure 7, we study the statistical distribution of F 1 and F 2 candidates in the ν-P and ν-S/N diagrams.Evidently, all F 2 events were detected in B04 scans at P ≈ 21.67 s. Figure 8 shows a sample F 2 candidate found in our study.Manually inspecting average pulse profiles of all such candidates, we attribute these events to coincident alignment between multiple instances of intermittent band-limited RFI.
Other notable features in Figure 7 include a narrowband line of events at 5.255 GHz and broadband streaks at multiple trial periods.(10) Here, n 0 ≈ 5.502 stars pc −3 is a normalization constant (Gowanlock et al. 2011), and h r = 2.25 kpc is the radial scale length.The constants z thin = 350 pc and z thick = 1 kpc are the vertical scale heights of the thin disk and the thick disk respectively.
At d GC = 8.18 kpc (GRAVITY Collaboration et al. 2019), our 6 survey footprint radius equates to a projected extent, ≈ 14.3 pc.Since z thin , z thick , we can integrate n(r, z) Consider an observing cone from the Earth pointing towards the GC along the Galactic mid-plane.Using Equation 11, we estimated the number of stars in our survey cone as a function of distance from the Sun.Our 4-8 GHz BL GC survey volume contains nearly 8 million stars within a distance of 8.24 kpc.Of these, about 600,000 stars reside between distances of 8.12-8.24kpc, i.e., at the GC.

Search Sensitivity
Equation 6 describes our theoretical survey sensitivity to periodic spectral signals, assuming an ideal Gaussian white noise background in time series data.However, real-world data frequently contain red noise that progressively hinders periodicity searches at large δ and long P .Following Section 2.2 of Suresh et al. (2021), we calibrated our A00 dynamic spectra to enable absolute survey sensitivity measurements.Injecting boxcar pulse trains into calibrated GBT time series data, we measured 7σ EIRP limits at several trial (P , δ).Performing 2D linear spline interpolation, Figure 9 illustrates our true search sensitivity EIRP min true in the P -δ plane.
Noticeably, at P ≥ 60 s and δ ≥ 25%, the presence of red noise in our GBT data significantly elevates EIRP min true above the theoretical sensitivity EIRP min periodic .Within our P -δ space in Figure 9, EIRP min true reaches a maximum of ≈ 2.5 × 10 21 W, i.e., 400 × EIRP min periodic at P = 100 s and δ = 50%.Nevertheless, at δ ≤ 15%, EIRP min true ≈ EIRP min periodic across all periods explored in our FFA searches.
Figure 10 places limits on the occurrence rate of transmitters in our survey as a function of EIRP min true for δ 10%.Our investigations constrain the number of extraterrestrial worlds transmitting periodic spectral signals to fewer than one in about 600,000 stars at the GC above a minimum detectable EIRP of ≈ 2 × 10 18 W at δ 10%.

Comparisons with Past Surveys
Our FFA searches bring novelty to radio SETI through their targeting of periodic spectral signals, a historically underexplored class of potential extraterrestrial transmissions.In contrast, traditional radio SETI programs have generally focused on the discovery of Doppler-drifting Hzwide CW beacons (Siemion et al. 2013;Harp et al. 2016;Tingay et al. 2016;Gray & Mooley 2017;Enriquez et al. 2017;Margot et al. 2018;Pinchuk et al. 2019;Price et al. 2020;Gajjar et al. 2021;Lacki et al. 2021;Margot et al. 2021;Traas et al. 2021;Franz et al. 2022;Garrett & Siemion 2023;Tusay et al. 2022) and artificially dispersed broadband bursts (Gajjar et al. 2021(Gajjar et al. , 2022)).Consequently, an equal comparison of our search sensitivity against that of past radio SETI campaigns is difficult.However, our searches are sensitive to sufficiently bright periodically modulated CW signals, rendering CW transmitter rate comparisons against previous surveys viable.Figure 11.Narrowband CW transmitter rate vs. minimum detectable EIRP for different surveys (various markers).Markers containing a white dot in their interior label surveys targeting exoplanets known to reside in the habitable zones around their host stars.The solid gray vertical line marks EIRP 10 13 W of an Arecibo-like transmitter.The dashed gray vertical line represents the the total solar power incident on the Earth.The diagonal gray band is a power-law fit derived from the best constraining limits (Gray & Mooley 2017;Price et al. 2020) on the CW transmitter rate at the low and high EIRP ends.Assuming a power-law distribution (N et ∝ EIRP α ) of transmitters in our Galaxy, the slope α ≈ −0.66 of our power-law fit constrains the luminosity function of extraterrestrial CW transmitters between EIRP limits of 2 × 10 12 W to 2 × 10 21 W. Plotted points for past surveys are taken from Figure 13 of Gajjar et al. (2021).
Lower CWTFM values correspond to more comprehensive searches for CW beacons.The product (N stars ν rel ) −1 is often termed the transmitter rate for narrowband CW beacons.Figure 11 compares the narrowband CW transmitter rate against EIRP min for various surveys.Past investigations marked in Figure 11 were dedicated searches for Doppler-drifting Hz-wide CW beacons.In contrast, our study, which targets channel-wide periodic signals, is sensitive to the periodicity at which potential CW beacons have been turned on or off.Therefore, while our search EIRP min overlaps with that of Gajjar et al. (2021), Figure 11 presents an unequal comparison of our search methodology against past uniform narrowband CW beacon search strategies.
Here, ξ AO = 10 −10 W −1 is a normalization constant such that PSSTFM = 1 when EIRP min = 10 13 W, N stars = 10 3 , ν rel = 0.5, P max /P min = 10 2 , and δ max /δ min = 10.For a given CWTFM, periodicity searches that cover larger ranges in P and δ have lower PSSTFM, and are therefore, said to be more complete.Our FFA searches achieve PSSTFM ∼ 10 3 -10 6 , where higher values reflect our increased EIRP min at longer P and larger δ.

Future Work
The coming decade promises to revolutionize radio SETI, with improved detection rates anticipated from the advent of ultra-wide bandwidth receivers, phased array feeds, and commensal backends for modern widefield sky surveys (Houston et al. 2021).The concomitant elevated data rates necessitate efficient automated pipelines for real-time event detection and follow-up.Enabling GPU acceleration of blipss will be essential to permit large-scale real-time searches for periodic spectral signals in future SETI campaigns.
The Breakthrough Listen Initiative endeavors to continuously expand the search space of technosignatures explored in SETI.Where blipss is sensitive to cyclic modulations of the total intensity, a natural extension to our SETI toolbox would be an autocorrelation search for signals that are cyclostationary in their complex voltages (see Section 2.10 of Morrison 2017).

SUMMARY AND CONCLUSION
Radio SETI has hitherto focused on the discovery of narrowband CW signals and artificially dispersed broadband pulses from extraterrestrial worlds.In contrast to these technosignature morphologies, periodic spectral signals offer an energetically efficient means of transmission across vast interstellar distances.A rotating beacon at the GC is, in particular, advantageously placed for galaxy-wide communications.
Here, we present blipss, a CPU-based software package to enable FFA searches for periodic spectral signals from alien worlds.Operating on radio dynamic spectra, blipss conducts FFA searches on a per-channel basis, thereby permitting periodicity detection regardless of signal bandwidth and dispersion.Consequently, blipss can uncover signals with exotic swept frequency structures frequently missed by traditional broadband pulse search techniques that assume cold plasma dispersion.
Running blipss on 4.5 hours of data from the 4-8 GHz BL GC survey, we report a non-detection of periodic pulsed signals with P ∈ [11 s, 100 s] and δ ∈ [10%, 50%].Thus, our investigations constrain the abundance of 4-8 GHz extraterrestrial transmitters of kHzwide periodic spectral signals to fewer than one in about 600,000 stars at the GC above a 7σ EIRP threshold of ≈ 2 × 10 18 W at δ 10%.
Future developments to our work include GPU acceleration of routines in blipss and incorporation of coherent voltage folding to expand our search space of periodic signals.
Progress on the former will enable integration of blipss into real-time search pipelines for largescale event discovery and follow-up.
Breakthrough Listen is managed by the Breakthrough Initiatives, sponsored by the Breakthrough Prize Foundation.P.N. was funded as a participant in the Berkeley SETI Research Center Research Experience for Undergraduates Site, supported by the National Science Foundation under Grant No. 1950897.S.Z.S. acknowledges that this material is based upon work supported by the National Science Foundation MPS-Ascend Postdoctoral Research Fellowship under Grant No. 2138147.The Green Bank Observatory is a facility of the National Science Foundation, operated under cooperative agreement by Associated Universities, Inc.

Figure 1 .
Figure 1.Detection of a target periodic spectral signal using blipss.Left: simulated radio frequency-time data containing 64 spectral channels of width 391 kHz each.A periodic pulsed signal has been inserted into the central channel.Right: radio frequency-period (P ) diagram showing the detection of a P 0 = 30 s signal in the central spectral channel.The first harmonic (cross) and the first subharmonic (plus) of the fundamental signal frequency (f 0 = 1/P 0 ) are also detected at P = 15 s and P = 60 s respectively.has often been picked up as periodic radio frequency interference (RFI) in historical transient searches (see Section 3.5 of Deneva et al. 2009) with the 300-meter William E. Gordon Arecibo radio telescope in Puerto Rico.

Figure 4 .
Figure 4.A schematic representation of the blipss workflow consisting of three stages.In Stage 1, a sequence of N input filterbank files is processed channel-wise to search for periodic signals.Operations on different input files are parallelizable across multiple cores of a machine.Stage 1 outputs a list of periodicity candidates detected per channel for every input file.In Stage 2, these candidates are assigned N -digit binary codes based on comparisons between on-target and off-target scans.A "1" at entry j of the code (read from left to right) indicates candidate detection in the j-th input filterbank file.Similarly, we use zeroes to label candidate non-detections in different files.Finally, in Stage 3, we define filters to select binary codes warranting visual candidate inspection.

Figure 5 .
Figure 5.A simulated candidate detected with 30 s period in all scans of source X, but unseen in scans of source Y. Left panels show periodograms generated from simulated single-channel time series data.Right panels illustrate phase-time diagrams (bottom subplots) and folded profiles (top subplots) of candidates at the periods highlighted by red vertical dashed lines in their respective left panels.Each row represents a different input data file.Rows {1, 3, 5} and {2, 4, 6} correspond to scans of sources X and Y respectively.

Figure 7 .
Figure 7. Statistical distribution of F 1 (blue) and F 2 (orange) events detected in our survey.Top left panel: Probability density function (PDF) of candidate periods.The period axis from 11-100 s is spanned by 39 bins of width 2.28 s each.Top right panel: PDF of candidate S/N values.The S/N axis from 7-9.81 is covered by 39 equal-sized bins.Bottom left panel: Scatter of candidates in the radio frequency-period plane.Bottom right panel: Scatter of events in the radio frequency vs. matched filtering S/N diagram.

Figure 8 .
Figure 8.An F 2 candidate with P ≈ 21.67 s (vertical red dashed line in left panels) detected exclusively in two B04 scans (rows 3 and 5).Phase-time diagrams (bottom subplots of right panels) on rows 3 and 5 trace the candidate origin to the coincident alignment between intermittent RFI instances in our data.In particular, some of the aforementioned RFI are observed close to the start times of different observing scans.
Number of Stars Surveyed The distribution of stars in the Milky Way constrains the number of potential transmitting worlds contained in the 4-8 GHz BL GC survey volume.Comparing different Galactic stellar density models, Gowanlock et al. (2011) argued that the Carroll & Ostlie (2006) estimation closely reproduces the observed stellar density profile in the solar neighborhood.Let r and z denote, respectively, the radial distance from the GC and the vertical distance from the Galactic midplane.Carroll & Ostlie (2006) modeled the stellar density in our Galaxy as n(r, z) = n 0 e −r/hr (e −|z|/z thin + 0.085e −|z|/z thick ).

Figure 9 .
Figure 9.Our true search sensitivity EIRP min true in the P -δ plane measured via injection of fake periodic signals into GBT time series data.The color scale quantifies EIRP min true relative to the theoretical sensitivity EIRP min periodic estimated from Equation6.Black circular markers label trial (P , δ) of fake signal injections.

Figure 10 .
Figure 10.Occurrence rate of 4-8 GHz extraterrestrial transmitters of kHz-wide periodic pulsed signals measured as a function of EIRP for a pulse duty cycle, δ 10%.The hatched region represents the parameter space excluded by our non-detection of channel-wide periodic signals in the 4-8 GHz BL GC survey.

A
standard metric used in literature to evaluate the CW search performance of a survey is the Continuous Waveform Transmitter Figureof-Merit (CWTFM) defined as (Enriquez et al. 2017) CWTFM = ζ AO EIRP min N stars ν rel .

6. 4 .
Periodic Spectral Signal Transmitter Figure-of-merit Motivated by CWTFM, we define a Periodic Spectral Signal Transmitter Figure-of-Merit (PSSTFM) to quantify our search completeness in relation to our processing parameters, our survey volume, the telescope and the instrumentation used.For periodicity searches spanning P ∈ [P min , P max ] and δ ∈ [δ min , δ max ], PSSTFM = ξ AO EIRP min N stars ν rel log Pmax P min log δmax δ min .

Table 1 .
Log of 4-8 GHz GBT observations processed in our study.Observing sequence of pointings grouped for position switching b Number of scans per pointing.Each scan lasted 5 minutes.
Table 2 lists various parameter values chosen for our blipss searches of paired target

Table 2 .
Parameter values for blipss searches of target fields

Table 3 .
Results from blipss searches of target fields ON pointing OFF pointing N scans a F 1 count b F 2 count b F 3 count b N cands c : Events detected in exactly n ON scans and zero OFF scans, n ∈ {1, 2, 3}.c Overall number of detected candidates, including all possible binary codes.d All candidates detected at P ≈ 21.67 s (see bottom left panel of Figure n Through visual candidate inspection, we identify the narrowband structure at 5.255 GHz with persistent RFI in our C02, C03, and C06 scans.We attribute the broadband features partly to noise and partly to bright transient RFI found near the edges of our observing scans.