Detection of Five New Pulsars with the BSA LPI Radio Telescope

During five years of daily observations, the search for new pulsars using the BSA LPI radio telescope in 96 spatial beams covering 17 000 square degrees was conducted. Five new pulsars were identified. Candidate pulsars were selected in summed power spectra. A noise generator was used to renormalize data and total the power spectra for individual directions correctly. As a result, the sensitivity increased by a factor of 10–20 relative to that in individual observational sessions. The sensitivity for pulsars with pulses longer than 100 ms at declinations +30° < δ < +40° was 1.2 and 0.4 mJy in and out of the galaxy plane, respectively.

The advent of wideband detectors, increased data processing rates, improvement of algorithms used to search for pulsars with subsequent reprocessing of archival records, development of new search techniques, and commissioning of new telescopes have contributed to the recent surge in pulsar studies. Numerous pulsar surveys covering the entire celestial sphere have already been conducted. New observational programs of this kind are worthwhile only if they utilize radio telescopes with high fluctuation sensitivity. This is the reason why such programs are underway at the 300-meter Arecibo telescope [[1]], 100-meter Effelsberg and Green Bank telescopes [[2]], the 64‑meter Parkes telescope [[4]], and the LOFAR [[5]] and GMRT [[7]] aperture-synthesis systems. The search for pulsars is also listed among the primary goals of the 500-meter FAST telescope [[9]].

Round-the-clock monitoring observations at 111 MHz with the completely reconstructed Big Scanning Antenna (BSA) of the Lebedev Physical Institute (BSA LPI) began in 2013 at the Pushchino Radio Astronomy Observatory (PRAO) as part of the "Space Weather" program [[10]]. Although these observations were initially limited to six frequency bands and a time resolution of 100 ms, monitoring data may be used to search for pulsars with periods on the order of a second. Seven new pulsars were found by an exhaustive search over periods and dispersion measures among individual records during 24 days of monitoring at declinations +21° < δ < +42° [[11]]. A total of 18 new pulsars were discovered in 4-year monitoring data at declinations –9° < δ < +42° with the summed power spectra [[12]]. This study concludes the search for pulsars in monitoring data with six frequency channels.

The performance capabilities of the reconstructed BSA antenna, observation modes, digital recorders, and data processing procedures were discussed in detail in [[10]–[12]]. Here, only a brief description is given.

Observations were carried out at the BSA LPI meridional radio telescope, showing an array of half-wave dipoles with an effective area of ~45 000 m2 in the zenith direction. The zenith corresponds to a declination of 55°. The center frequency in the reception band is 110.25 MHz, and its width is 2.5 MHz. The directional pattern of the antenna (beam) is approximately 0.5° × 1° in size, which allows one to track each source in the sky for 3–4 minutes per day. The monitoring part of the radio telescope currently features 96 beams covering the sky at –9° < δ < +42°. An area of ~17 000 square degrees is surveyed each day. Round-the-clock observations in six frequency channels began in 2013, but separate long periods of round-the-clock observations from 2012 are also available. Parallel observations in 32 frequency channels with a time resolution of 12.5 ms began in August 2014.

Monitoring observations are conducted in 1-h-long sessions. The start of each session is synchronized to an atomic frequency standard, and the time within the 1-hour interval is measured by a quartz oscillator. The time error accumulated within this interval may reach ±100 ms. Due to the probable large error in the determination of the signal arrival time and the low accuracy of period measurement in a single observational session (±0.0005s), timing based on monitoring data has not been implemented yet.

Candidate pulsars found in 6-channel data are verified using 32-channel observations. These candidates are then examined using a specialized pulsar detector with high time and frequency resolutions to identify them reliably as pulsars and determine their characteristics accurately. Subsequent studies will soon follow.

The direct search for pulsars in individual observation sessions showed that normalization is not needed [[11]]. Self-calibration against the noise track in each frequency channel was performed in each observational session in the search for pulsars in the total power spectra [[12]]. The procedure was as follows: interference was removed in each frequency channel within an interval of 204.8 s (2048 points), the baseline was subtracted, and the noise dispersion was calculated; the signal amplitude at each point was divided by the dispersion, thus making the dispersion over the set equal to unity; and power spectra were calculated for each frequency channel independently and summed. These operations were performed for each day of observations for a given direction. The normalization procedure was thought to suppress noise in those days when interference conditions were unfavorable or the telescope sensitivity decreased due to some other reasons, and the resulting reduction in quality of power spectra summed over all observation days was expected to be insignificant.

The subsequent analysis results of early observations revealed that days with low-quality data remain even after the exclusion of poor quality data. Therefore, if we total all the remaining power spectra within an observational period, the resulting signal-to-noise () ratio in the summed power spectrum will be lower than that in a spectrum with low-quality data removed. In the present study, another method for data calibration with a noise-generator signal was tested. The calibration signal was recorded in the form of OFF–ON–OFF, where each part is 5 s in length (see Fig. 1). The temperature of ON and OFF signals was 2400 and ~300 K (see details in [[10]]).

Graph: Fig. 1. Calibration signal (OFF–ON–OFF) and individual pulses from the known В1919+21 pulsar (PSR J1921+2153, s, pc/cm3) recorded in one of the six frequency channels. The channel width is 430 kHz.

Following calibration, one may arrange the signal dispersions in ascending order for each direction. The maximum dispersions will correspond to low-quality observations; the minimum ones, to the highest-quality observations. The observed area is split into ~40 000 directions (pixels). The dispersion is calculated each day for every pixel after normalization with the calibration signal. Approximately 108 estimates of the noise dispersion have already been obtained. The actual ratio may be estimated based on the known noise dispersion for each day and the minimum dispersion within the entire observation interval. It was assumed in [[12]] that the enhancement factor after the exclusion of days with low-quality observations is the square root of the number of remaining days. Now, it became possible to verify this assumption experimentally by considering actual estimates of the noise dispersion. Figure 2 presents the typical expected and actual enhancement patterns within the entire observation period for one direction.

Graph: Fig. 2. Time dependence of the signal-to-noise ratio. The upper curve represents the maximum possible theoretical S/N enhancement, while the lower curve illustrates the actual variation of in summed power spectra. The sharp drop seen in the lower curve is associated with the days having the highest noise dispersions.

The lower curve shows that the ratio first remains close to the theoretical curve, but then shifts further and further away from it. In five years of observations, 30–60 days with very poor quality of the noise track are recorded in different directions. If the noise dispersions are arranged in ascending order, the series end with these days; therefore, the ratio obtained by summing all power spectra drops sharply instead of increasing. The vertical arrow in Fig. 2 denotes the limit for the number of days used to obtain the summed power spectrum in a given pixel. Further summation beyond this boundary results in a negligible enhancement. It can be seen in Fig. 2 that the summed power spectrum was plotted for 1300 days and increased by a factor of 28.7, while the expected enhancement factor was 36. Testing other directions, it was discovered that the enhancement factor typically varies between 20 and 30. Since the time resolution in observations was 100 ms and the mean FWHM of a typical pulsar's pulse with a period on the order of 1 s is 20–30 ms, the resulting ratio will be 1.5–2 times lower (10–20 times higher than the theoretical ratio in a single observation).

A number of criteria were used in the previous BSA LPI search for pulsars in the summed power spectra [[12]] to verify the detection of new pulsars. These criteria were as follows:

(i) repetition of the signal in sidereal time;

(ii) presence of at least two harmonics in the power spectrum;

(iii) presence of a well-defined maximum in the ratio dependence in the mean profile on the dispersion measure;

(iv) existence of at least a single record in the 32‑channel frequency mode supporting the detection of a pulsar with > 6 in the mean profile;

(v) roughly equal heights of the mean profiles in the double-period record.

Regular observations of 21 out of the 26 pulsars detected earlier [[11]–[13]] have been performed starting from July 2017 at a pulsar facility with high frequency and time resolution (4.88 kHz × 460 channels and a time resolution of 2.46 or 5.12 ms). The radio emission detection from 18 sources was verified, and these sources are observed further to determine their coordinates and periods more accurately. Three pulsars, which apparently are weaker than the rest, are still being analyzed. The remaining five pulsars were identified as the weakest of the 18 pulsars detected earlier in the summed power spectra [[12]]. These objects are planned to be observed with a pulsar detector.

When observational data were subjected to another processing, BSA-Analytics[1] was used in the initial search for harmonics in the power spectrum, but all 40 000 directions were then inspected visually. Known pulsars (cross-checked against the ATNF[2] catalogue) were removed, and 87 new candidates were selected. Their verification revealed that a considerable fraction of the objects are known pulsars observed, in particular, in the side lobes of the BSA LPI directional pattern. These 23 pulsars have not been detected earlier and were not included into the monitored pulsar list. Eight candidates were found to be periodic interferences of various nature. A search through the initial data was conducted for the remaining 56 candidates, and five strongest sources with at least two harmonics (periods longer than 0.4 s) in the summed power spectra were selected.

The candidate verification procedure was the same as in the previous studies. The spectra with harmonics matching the frequency of harmonics in the summed power spectrum were selected out of all power spectra corresponding to the source direction. An exhaustive search over periods and dispersion measures was then performed. The observation date, the sidereal time at the checked interval's center, and the expected period were used as the input parameters. The code searches for pulsars within an interval of ±3 min from the set time with a step of 20 s. This allows one to more accurately determine the right ascension coordinates of detected pulsars. Thus, the overall length of the time interval for search in the initial data is 6 min. At each step, the code cycles through periods within ±10% of the set value. The dispersion measures are checked within 0–200 pc/cm3 at each interval, and the mean profiles at double periods with > 5 are recorded. The processing results for each chosen day are stored. In subsequent analysis, this provides an opportunity to load the data for several days and total the mean profiles or dependences of on the tested dispersion measure.

Since all the pulsars found have approximately the same ratios between the pulse duration and the period, all mean profiles are similar in shape. Thus, they are not presented here and were uploaded to the site of the observatory[3] together with the mean profiles of pulsars detected earlier.

The verification results of five candidates with their mean profiles plotted and the dispersion measures estimated based on 32-channel data are presented in Table 1. The pulsar's J2000 designation is given in the first column, and its right ascension and declination coordinates for the year 2000 are listed in columns nos. 2 and 3. The typical determination accuracy of right ascension and declination is ±30s and ±15′, respectively. The pulsar period determined with an accuracy of ±0.0005 s is given in the fourth column, and the dispersion measure and the observed FWHM of the mean profile are listed in columns nos. 5 and 6. The error of determination of the mean profile's FWHM may be rather large, since the probable pulse broadening due to the dispersion measure in the band of a single frequency channel was not considered. The actual FWHM of the mean profile may be smaller than the indicated estimate.

Characteristics of new pulsars

Individual observational days for verification in 32‑channel data could not be found for 51 candidates. The strongest of them is presented in Fig. 3. Source J1921+3357 has three harmonics in the spectrum and is observed in two neighboring antenna beams. The period corresponding to inverse of pulse frequency of the first harmonic is s. Pulsars to identify with this candidate could not be found in ATNF. Thus, it is highly likely that J1921+3357 is a pulsar, but its dispersion measure was not able to be determined. Objects J1743+1300 ( s, 20 pc/cm3) and J2022+2122 ( s, pc/cm3) are other candidate pulsar examples. They were not included into Table 1, since the ratio is lower than 4 at the maximum of the dependence on DM. During the preparation of this paper, it was discovered that these objects were identified as pulsars in [[6]]. Their characteristics turned out to be close to the ones determined here: s, pc/cm3 (J1745+12) and s, pc/cm3 (J2022+21).

Graph: Fig. 3. Power spectrum of J1921+3357. Harmonics multiple of 1 s and harmonics related to internal noise were removed.

In addition to new pulsars, more than 100 known pulsars with periods s, which were identified in ATNF and earlier studies, are found in the summed power spectra. The list of identified known pulsars grows constantly. These pulsars are not relevant to the present study. Their mean profiles and certain additional data may be found at the site of the observatory.[4]

As noted above, the identification of several objects in the summed power spectra as new pulsars could not be verified. Some of them may turn out to be interference-related or far side lobes of known pulsars. However, some of them can possibly be new pulsars. An example of such a candidate pulsar is shown in Fig. 3. Just as verified pulsars, candidate pulsars are observed in one beam or two neighboring beams, but the sensitivity of the BSA LPI radio telescope is not sufficient for a well-marked maximum to emerge in the dependence on in a single observational session. The minimum flux densities of pulsars observed in the power spectra have already been estimated at 0.2 (outside the Galaxy plane) and 0.6 mJy (in the Galaxy plane) at 111 MHz in the previous study [[12]]. With the plotted curves of actual growth of the ratio in accumulated spectra taken into account, these estimates may be corrected to 0.3 (outside the Galaxy plane) and 0.9 mJy (in the Galaxy plane) within a 4‑year observation interval. It is important to note that the sensitivity estimates in [[12]] are given for the zenith direction, i.e., these are the estimates of the maximum possible sensitivity. It was also noted in [[12]] that the difference between the maximum and minimum sensitivities may be as large as an order of magnitude due to the specifics of the BSA LPI antenna, which is a diffraction array. The maximum declination in observations was +42°, while the zenith direction corresponds to a declination of +55°. The source coordinates typically fall between the beams of the radiation pattern, thus reducing the sensitivity. The actual sensitivity may be roughly estimated due to the declination positioning of beams. It is approximately equal to 0.4 and 1.2 mJy at declinations +30° < δ < +40° for directions lying outside and within the Galaxy plane, respectively. At low declinations –9° < δ < +3°, the sensitivity decreases to 1.2 and 3.6 mJy for directions lying outside and within the Galaxy plane, respectively, both due to the correction for the zenith distance's cosine and to the effective reception band's narrowing.

The maximum sensitivity of the reconstructed BSA LPI antenna in a single session may be estimated roughly based on observational data (the measured flux density at 102 MHz [[14]] and the observed ratio in the mean profile) for known pulsars. These estimates suggest that pulsars with a flux density of ~5 mJy may be detected in a single observational session. Estimating the maximum sensitivity in a single observational session based on the effective area and other known parameters, one can find a ~4.4 mJy value for sources outside of the Galaxy plane in near-zenith directions [[11]].

New pulsar verification with flux densities below 5 mJy requires additional effort (e.g., observations with telescopes more sensitive than BSA LPI). Apparently, LOFAR is well-suited for such observations, since the lower effective area of this antenna is made up by its wide reception bands and tracking capability. The center reception frequency (140 MHz) is also close to the center frequency from observations. According to [[6]], the sensitivity of the LOFAR radio telescope in a 2-h-long observational session for pulsars with periods on the order of 1 s and dispersion measures below 100 pc/cm3 is up to 1.2 mJy, which is four times higher than the maximum sensitivity of BSA LPI in a single observational session. FAST observations will be even more efficient, since this radio telescope achieves a very high sensitivity on short observational intervals. Long-term BSA LPI observations are also an option. Since pulsars are variable objects, a possibility always remains that their flux density in certain sessions will be significantly higher than the mean density of <5 mJy.

Apart from power spectra with two or more harmonics, summed spectra also feature hundreds of single harmonics. Some of them are repeated in many beams and are apparently related to industrial and internal noise. However, some harmonics are observed only in individual beams, and the periods corresponding to these harmonics are not repeated.

The nature of these harmonics is important to note. The first harmonic in the pulsar's power spectra normally has the maximum height, while the height of other harmonics decreases. Therefore, if the height of the first observed harmonic is not that large, the next harmonic may be lost in the noise. Pulsars of this kind are weak objects for BSA LPI observations, and methods for processing the data on such pulsars need to be improved. A single harmonic is also produced by pure sine signals, which are characteristics of certain industrial noise types. The same single harmonic is produced by pulsars of the coaxial rotator type or pulsars with signal spreading due to the large dispersion measure and pulses occupying all or most of the period. The mean profiles of such pulsars appear unconvincing; it is hard to determine their ratio in the mean profile, and the errors of determination of the dispersion measure are large (see, e.g., pulsar J1844+4117 in [[12]]).

The problem of searching for pulsars in spectra with a single observed harmonic is solved partially by switching to 32-channel data. Figure 4 presents two summed power spectra plotted based on the data recorded in the full 2.5 MHz band in 6- and 32-channel modes for the known J1922+2110 pulsar with dispersion measure pc/cm3 and period s. Only the first harmonic may be distinguished in the power spectrum based on the 6-channel data, while four harmonics are visible in 32 channels. The harmonics unrelated to J1922+2110 were removed from the power spectra in Fig. 4. This increase in the number of observed harmonics is apparently related to the narrowing of pulses detected with a considerably higher time resolution (12.5 ms instead of 100 ms).

Graph: Fig. 4. Power spectra of pulsar J1922+2110 based on the 32-channel (upper panel) and 6-channel (lower panel) data. Low-frequency noise, which was partially removed, is seen at the start of records. The time resolution in observational sessions is indicated.

This example demonstrates that the search technique needs to be developed further to identify new weak pulsars and determine their parameters.

These pulsars have been compared to known pulsars with declinations –9° < δ < +42° (i.e., pulsars located within the observed area), dispersion measures pc/cm3, and rotation periods > 0.4 s selected from the ATNF catalogue. These constraints are related to the limitations of the survey. Figure 5 shows the distribution histograms of periods of ATNF pulsars and 30 pulsars detected by BSA LPI (see [[11]] and the present study).

Graph: Fig. 5. Distribution histograms of periods of ATNF pulsars (unshaded) and pulsars detected at 111 MHz (shaded).

Although the number of detected pulsars is small, it is evident that the distributions are different. The lower histogram has few pulsars with periods shorter than 0.9 s and a peak at 0.9–1.4 s. However, the upper histogram also features a slight excess in this range. The lack of short-period pulsars may be attributed to the reduced search frequency (111 MHz), but the more likely explanation is that it stems from the low time resolution of the search (100 ms).

The search for pulsars in 6-channel frequency data within a 5-year interval was conducted. Five new pulsars were found. Together with the results of [[11]–[13]], these new data bring the number of detected pulsars to 31. A set of 51 candidate pulsars with at least two harmonics in their summed power spectra was examined. Observational days for verification of their coordinates, periods, and dispersion measures could not be found in the dataset. It is expected that the flux density of candidate pulsars is below 5 mJy at 111 MHz. The identification of two objects (J1745+12, J2022+21) detected at 140 MHz [[6]] as pulsars was verified.

It was demonstrated that the ratio in a search with the use of summed power spectra within a 5-year interval increases by a factor of 10–20 depending on the direction.

The authors wish to thank A.I. Chernyshova for analyzing several power spectra, L.B. Potapova for her help in drafting the figures, and T.V. Smirnova for useful remarks.

Translated by D. Safin