DRAFT VERSION NOVEMBER 15, 2018 Preprint typeset using LATEX style emulateapj v. 01/23/15 THE SERENDIP III 70 CM SEARCH FOR EXTRATERRESTRIAL INTELLIGENCE STUART BOWYER,MICHAEL LAMPTON,ERIC KORPELA,JEFF COBB,MATT LEBOFSKY, AND DAN WERTHIMER Space Sciences Laboratory, University of California, Berkeley, CA 94720 Draft version November 15, 2018 ABSTRACT We employed the SERENDIP III system with the Arecibo radio telescope to search for possible artificial ex- traterrestrial signals. Over the four years of this search we covered 93% of the sky observable at Arecibo at least once and 44% of the sky five times or more with a sensitivity of ∼ 3×10−25Wm−2. The data were sent to a 4×106 channel spectrum analyzer. Information was obtained from over 1014 independent data points and the results were then analyzed via a suite of pattern detection algorithms to identify narrow band spectral power peaks that were not readily identifiable as the product of human activity. We separately selected data coincident with interesting nearby G dwarf stars that were encountered by chance in our sky survey for suggestions of ex- cess power peaks. The peak power distributions in both these data sets were consistent with random noise. We report upper limits on possible signals from the stars investigated and provide examples of the most interesting candidates identified in the sky survey. This paper was intended for publication in 2000 and is presented here without change from the version submitted to ApJS in 2000. Keywords: 1. INTRODUCTION pared, for example, with the SETI Institute targeted search), Early radio searches for extraterrestrial intelligence used our sensitivity is still substantial because of the large col- dedicated telescope time to search for emission from nearby lecting area of the Arecibo telescope and the outstanding re- stars (see Tarter 1991 for a partial listing, and Tarter 2001 for ceivers that are available for use with this instrument. a full listing of these searches). This type of search became We report the results of our sky survey and provide upper increasingly difficult to carry out at major facilities because limits on possible signals from stars in this paper. of a general reluctance to devote dedicated telescope time to 2. OBSERVATIONS AND ANALYSIS such projects, which though interesting, are acknowledged to The SERENDIP III data were obtained with the National have a low probability of success. In addition, sky surveys Astronomy Ionosphere Center’s radio observatory in Arecibo, were carried out which scanned substantial portions of the Puerto Rico with a 430 MHz receiver. Data collection began sky. The Berkeley SERENDIP project (Search for Extrater- in April, 1992 and ended in October, 1998. restrial Radio Emissions from Nearby Developed Intelligent The feed used for SERENDIP III was located at the oppo- Populations) solved the dedicated telescope time problem by site carriage house from the primary observer’s feed. This using data obtained simultaneously with ongoing astronom- resulted in three observation modes. In the first, represent- ical research. This program began over twenty years ago ing about 45% of the observing time, the primary observer’s (Bowyer et al. 1983) and has continued to the present day feed was tracking a point on the sky. This resulted in the with ever increasing sensitivity and an ever-widening set of SERENDIP III beam moving across the sky at roughly twice search parameters. the sidereal rate. In the second, representing about 40% of the Other sky surveys using dedicated telescopes have been, observing time the feeds were stationary, resulting in motion and are continuing to be, carried out. The Ohio State program at the sidereal rate. In the third, representing about 15% of the (Dixon 1985) was the earliest sky survey; this project is now observing time, the SERENDIP III feed was tracking a point terminated. The Harvard search (Leigh & Horowitz 1997) on the sky. has also been terminated. The Argentinian search (Lemarc- Power spectra were generated by the SERENDIP III four hand et al. 1997), and the Australian search (Stootman et al. million channel spectrum analyzer which had a 1.7 second in- 1999) continue. Targeted searches of nearby stars have been arXiv:1607.00440v1 [astro-ph.IM] 2 Jul 2016 tegration period and a 0.6 Hz frequency resolution (2.5 MHz initiated by the SETI Institute (Tarter 1997) using substan- instantaneous band coverage). The sensitivity achieved in a tial amounts of dedicated telescope time which were obtained 1.7 second integration at our typical system temperature of in return for a substantial financial contribution to the tele- −25 −2 scope upgrade which was carried out after the conclusion of 45 K was 3 × 10 W m . This bin size is wide enough the SERENDIP III observations. to encompass Doppler frequency drifts caused by the Earth’s We discuss our sky survey search for artificial extraterres- motions plus reasonable accelerations of a hypothetical trans- trial signals with the SERENDIP III system (Bowyer et al. mitter’s reference frame. The receiver’s entire 12 MHz band 1997) and the Arecibo telescope. Although the search was a was processed in 2.4 MHz steps taking about 8.5 seconds to sky survey, nearby solar-type stars inevitably fell within the complete a single sweep. The SERENDIP III fast Fourier beam pattern of the telescope in the course of these observa- transform based hardware is described in detail by Werthimer, tions. As part of the analysis of the SERENDIP III data, we et al. (1997). have separately investigated the data from observations of the Adaptive thresholding was achieved by baseline smoothing sky coincident with these nearby stars. Although our integra- the raw power spectra with a sliding eight thousand channel tion times for individual targets are relatively short (as com- local-mean boxcar and searching for channels exceeding 16 times the mean spectral power. These signals were recorded 2 SERENDIPIII (Real-Time at Arecibo) Also performed at Arecibo * Baseline smoothing and normalization * Course-resolution spectra * Telescope interface 4 Million Pt. FFT * Data logging * Instrument control Event detection (thresholding) (Off-Line at UCB) Candidate Also performed at UCB Database * Health monitoring * Reference frame correction Dead-time Non-drifting Drifting RFI rejection Pattern detection Candidate merging and ranking rejection RFI rejection Figure 1. An illustration for the SERENDIP III data flow showing the steps taken to detect candidate signals. Real-time processing consists of power spectra generation and application of an adaptive threshold to the resulting spectra. Off-line data reduction included RFI rejection techniques, pattern detection, and candidate extraction. along with time, pointing coordinates, detection frequency, nals were rejected if they (1) were detected over broad ar- and signal power. eas of the spectrum in one or more integration periods (broad Following the real-time data analysis by the SERENDIP III spectrum interference), (2) persisted at the same receiver fre- instrument, the reduced data was shipped to Berkeley of off- quency through multiple telescope beams, or (3) persisted in line data reduction and candidate generation. Off-line data the same channel of the spectrum analyzer. reduction included radio frequency interference (RFI) rejec- Broad spectrum interference was identified by the rejection tion techniques, pattern detection, and candidate extraction. test: The overall data flow is shown in Figure 1. ∑d2 2.1. Off-line Data Reduction x > 50 and > S (1) x Data were transferred from Arecibo to Berkeley across the Internet where off-line data analysis activities began. Data re- where d is the number of frequency bins (0.6 Hz/bin) between duction consisted of removing data taken during periods when simultaneous events above the threshold and x is the number the telescope was slewing too fast or too slow for our analy- of events above threshold in the spectrum. For SERENDIP sis procedures, followed by the application of a suite of RFI III, the threshold value was set at S = 108. filters. RFI rejection algorithm (2) uses a statistical method to de- Excessively rapid telescope slew rates precluded acquisi- termine if several detections at the same observing frequency tion of accurate positioning information. In addition, during could be ruled out. If these detections occur with a signifi- times when the receiver tracks a point on the sky, it is not pos- cantly above-average hit rate, and they continue to occur when sible for our analysis programs to differentiate between con- our observing beam has moved beyond one beam width (0.17 tinuous RFI and a potential signal. Therefore, our first filter degrees), we reject the hypothesis of random Poisson events was to censor data acquired during periods of rapid telescope being the cause. Instead we mark these detections as being movement and periods when the telescope was tracking sky due to external RFI, and reject them. objects. Roughly 15% of the data were removed for this rea- RFI rejection algorithm (3) uses the same statistical test son. as (2) but applies it to hit sequences that have the same The next step in data reduction was non-drifting RFI rejec- intermediate-frequency bin number. In this way it rejects in- tion. SERENDIP’s non-drifting RFI rejection algorithms in- terference generated within the observatory. corporated dynamically adaptive statistical analysis routines Data surviving the first three rejection criteria were further that detect spurious signals from terrestrial and near-space analyzed for RFI that drifts rapidly in frequency and were sources. Three cluster analysis tests were conducted on each therefore not rejected by algorithms (2) and (3) above. Figure input data file spanning several hours of observation. Sig- 2 illustrates SERENDIP’s drifting frequency RFI detection al- BOWYER ET AL. 3 9000 8000 7000 6000 Time 5000 4000 3000 Time in seconds 2000 1000 0 Frequency 422 424 426 428 430 432 434 436 Figure 2.