Detection of Two New RRATs at 111 MHz

Abstract

At the LPA LPI radio telescope, a search was conducted for pulse signals in a specific area with declinations $+ 52^\circ < \delta < + 55^\circ$. The processing involved ten months of observations recorded in six frequency channels, each with a channel width of 415 kHz and a total bandwidth of 2.5 MHz. The analysis of the data revealed 22 000 events, exhibiting a pronounced dispersion delay of signals over the frequency channels, which are indicative of pulsar pulses. These pulses were found to belong to four known pulsars and two new rotating radio transients (RRAT). Additional pulse search was carried out in 32-channel data with a channel width of 78 kHz. This led to the discovery of 8 pulses for the transient J0249+52 and 7 pulses for the transient J0744+55. No periodic emission of transients was detected. The analysis of the observations indicates a high probability that the identified RRATs are pulsars with nullings, where the fraction of nullings is greater than 99.9%.

1 INTRODUCTION

In 2014, the Large Phased Array (LPA) of the Lebedev Physical Institute (LPI) of Russian Academy of Sciences underwent antenna upgrades, which increased the radio telescope’s fluctuation sensitivity by 2–3 times, and the number of simultaneously observed beams was expanded to 128. Following these improvements, regular monitoring observations commenced. These observations are used for studying interplanetary plasma [1], searching for pulsars and transients [2], and investigating the variability of active galactic nuclei [3]. The heightened sensitivity of the radio telescope has enabled the detection of individual pulses from more than 150 pulsars.¹ Some of these pulsars, which exhibit irregularly observed pulsed signals, were discovered the first time at the LPA LPI for the first time and are categorized as rotating radio transients (RRATs).

The optimal search for pulsed dispersed signals (pulses arriving first at a high and then at a low frequency), considering their scattering and scintillations in the interstellar medium, was considered by Cordes in 2003 [4]. The application of the proposed method to archival data obtained at the 64-m radio telescope (Parks, Australia) led to the discovery of 11 pulsars with distinctive properties in 2006 [5] and the discovery of extragalactic pulses in 2007 [6]. While ordinary pulsars emit pulses on every or almost every revolution, and they can be searched for in the standard way using power spectra or periodograms, RRATs may have long periods between successive pulses, sometimes not emitting a pulse for many rotations. Consequently, standard methods may not detect such pulsars.

The total number of known RRATs is small. The ATNF database² [7] and RRATalog³ list approximately 100–150 RRATs. The recent RRAT search using the 500-meter radio telescope FAST has reported the detection of another 76 transients [8]. That study pushed the number of known RRATs beyond 200. At the same time, estimates suggest that the expected number of RRATs may be twice as large as the number of conventional pulsars [9]. This means that only a small fraction of RRATs has been discovered thus far. Most of the RRATs have been detected using several radio telescopes with high instantaneous sensitivity: the 64-m Parkes Telescope (Australia) [5], the 300-m Arecibo telescope (Puerto Rico) [10], the 100-m Green Bank Telescope (GBT) (USA) [11], the LPA LPI telescope, which is an antenna array of 200 × 400 m (Russia) [2], the 500-m FAST telescope (China) [8], and the CHIME telescope with a size of 80 × 100 m (Canada) [12].

The occurrence of RRAT pulses is unpredictable. The typical time between successive pulses may range from minutes to hours [5], and in some cases, even tens of hours [13]. The study [14] notes two RRATs (J1132+25, J1336+33), in which no pulses were recorded in more than 250 observation sessions that took place once a day and had a duration of about 3.5 min. Thus, the search for RRAT requires not only an antenna with high fluctuation sensitivity, but also long series of observations.

Observations at the LPA LPI yield constant recordings of pulsed signals. A special examination of the quality of data obtained during monitoring revealed that approximately one-fifth of these signals are associated with pulsars, while the remaining signals are caused by interference [15]. A statistical approach was employed to assess the quality of monitoring data and enabled the detection of pulses from known pulsars (ranging from a few pulses to several thousands) and the discovery of four new RRATs in the data from the new recorder [15].

This study investigates short-duration signals found in raw data. We consider the statistics of detections of pulsed radiation sources in a new area that was included in the monitoring program of searching for pulsars in a test mode in the autumn of 2021. The aim is to identify these pulsed sources with interference or actual signals of extraterrestrial origin.

2 OBSERVATIONS

The LPA LPI radio telescope is a meridional instrument. It can observe any source in the sky once a day, with each observation session typically lasting 3–4 min. During the antenna’s upgrade, a scheme was implemented, enabling the operation of four independent radio telescopes based on one antenna field. Currently, two radio telescopes are in operation. One of them (LPA1) is used for standard pulsar observations. Its 512 beams cover declinations in the range $- 15^\circ < \delta < + 87^\circ$ with a beam overlap of 0.8. The second telescope (LPA3) has 128 beams, covering declinations $- 9^\circ < \delta < + 55^\circ$ with a beam overlap of 0.4. If the source coordinates coincide with the beam coordinates, the source is not observed in the beams above and below.

In the period of 2013–2014, the two recorders of the 128-beam radio telescope were connected to 96 beams covering declinations $- 9^\circ < \delta < + 42^\circ$. At the end of 2020, a new recorder was created, and 24 more beams were connected in a test mode, covering declinations $+ 42^\circ < \delta < + 52^\circ$. The quality of observations in these 24 beams was analyzed, and the results of the analysis were published in [15]. In addition to assessing the quality of data in the newly connected beams, four new RRATs were also detected. In 2021, the last 8 beams were put into operation, covering declinations $+ 52^\circ < \delta < + 55^\circ$. Thus, LPA3 uses all available beams (128 = 48 + 48 + (24 + 8)), to which three recorder units are connected. After the testing period of the equipment concluded on October 20, 2021, round-the-clock monitoring observations began. This paper presents the initial results of data analysis for the period from October 21, 2021 to August 31, 2022.

The characteristics of LPA3 are as follows: the center reception frequency is 110.25 MHz, the receiving bandwidth is 2.5 MHz, and the effective antenna area is approximately 45 000 m². Raw data are synchronously recorded in two time-frequency resolutions. Low frequency-time resolution data divides the receiving bandwidth into six 415-kHz-wide frequency channels with a point polling time of 100 ms. These data are used in the Space Weather project [1]. Noise (pulsed signals) is analyzed based on these data. Data with high time-frequency resolution are recorded in 32-channel mode at a channel width of 78 kHz. The point polling time is 12.5 ms. The 32-channel data is used when additional verification of detected pulses is required. Data recording is done at hourly intervals, with the next hour of observations starting immediately after the end of the previous recording.

To ensure equal gain in the frequency channels, a noise signal of known temperature is used, which is fed to a distributed amplification system. This is done in the OFF–ON–OFF mode (calibration step), where OFF represents the absence of a calibration signal with all intermediate amplifiers turned off. In this case, the noise in the antenna paths corresponding to the ambient temperature is recorded. The ON mode corresponds to turning on the calibration signal (temperature of 2400 K) with the dipole lines turned off. More details on working with the calibration step can be found in [3].

3 DATA PROCESSING

The pre-gain in the frequency channels is leveled through a calibration step, which is recorded six times a day. The recorded data is then divided into ten-second time segment. For each time interval and channel, several parameters are estimated, including the minimum and maximum intensity values in antenna degrees after step calibration, the median intensity value, and the noise standard deviations. Additionally, information about the date and hour of observations in Moscow time, as well as the beginning of the studied ten-second segment in sidereal time, is saved for each segment. The processed data are stored in a database, and its volume is ten times smaller than the volume of the original raw data. This database allows for the identification of interference levels at any selected time interval and enables the investigation of individual pulsed noise events.

The detailed description of the data processing is provided in [15, 16]. It should be noted that the coordinate of the observed maximum within each ten-second interval is stored, facilitating the alignment of maxima for each frequency channel in time. This aids in roughly determining the dispersion delay of the signal in frequency during the processing. Since the standard deviation of the noise inside each channel is known, the search for pulsed signals can be carried out at a given signal-to-noise (S/N) level. The S/N ratio is defined as S/N = $A{\text{/}}{{\sigma }_{{{\text{noise}}}}}$, where $A$ is the signal amplitude after subtracting the baseline (background signal), and ${{\sigma }_{{{\text{noise}}}}}$ is the standard deviation on a 10-s interval. If, for a particular segment, a pulse is found with an S/N greater than 5 in at least three of the six frequency channels, it is considered a candidate for transients. These potential RRAT candidates undergo additional verification against data with high time-frequency resolution.

4 RESULTS

During the observation interval from October 21, 2021 to August 31, 2022, a total of 7496 file-hours (312.33 sidereal days) were analyzed after subtracting data gaps. In the processing of 6-channel data, 2.5 million pulses were detected, and many of these pulses appeared simultaneously in several beams. These cases were automatically grouped into related events. In total, 160 504 related events were identified.

The distribution of the number of simultaneous events over different beams resembled the distribution observed in the earlier analysis of 24 beams [15]. Most often, pulses are either observed in only one beam or in all beams simultaneously. Actual pulsars’ pulses should be observed in one or two adjacent beams. However, this does not rule out interference, which can also fall into only one beam. Pulses observed in all beams are predominantly due to interference. Some very powerful pulses of pulsars can also be observed in multiple beams, appearing in the side lobes of the LPA PLI. Other sources of this type of pulses are also possible, but they will be the subject of a separate study.

After analyzing the pulses detected in one or two beams, we found that approximately 22 000 pulses, which is 13.9% of the total number, showed a pronounced dispersion delay, indicating similarity to pulsar pulses. The search for new pulsars based on their individual dispersed pulses was conducted for 8 out of the 32 beams ($+ 52^\circ < \delta < + 55^\circ$), since the search for 24 beams was already carried out in an earlier study [15].

Same as in [15], the area under study was divided into clusters, each 2 minutes long. For each cluster, various parameters were checked, including the number of detected pulses, right ascension and declination coordinates, the average dispersion measure of the pulses, beam number, Julian date (MJD), time (UT), sidereal time, S/N observed in frequency channels, and lists of beam numbers with similar pulses. Out of 5760 possible clusters within the study area, the processing program identified 45 clusters. Each cluster had a right ascension coordinate determined by the 2‑min interval and a declination coordinate determined by the LPA3 beam number. The identified 45 clusters contained at least one “pulsar” event. The coordinates of these clusters were determined, enabling an identification with the ATNF catalog.

Known strong pulsars can simultaneously occupy several neighboring clusters in both right ascension and declination. These pulsars were easily identified in the ATNF catalog and removed from further analysis. In total, four known pulsars were detected during the   analysis: B0329+54 ($P = 0.7145$ s; $DM =$ 26.7 pc/cm³); B0343+53 ($P = 1.9344$ s; $DM =$ 67.3 pc/cm³); B1508+55 ($P = 0.7396$ s; $DM =$ 19.6 pc/cm³) and B2021+51 ($P = 0.5291$ s; $DM = 22.5$ pc/cm³). These pulsars exhibited varying numbers of detected pulses, from one pulse for B0343+53 to more than 10 000 pulses for B0329+54. The pulses of the B0329+54 and B1508+55 pulsars were also observed in the side lobes.

In addition to the known pulsars, pulses belonging to two new RRATs (J0249+52 and J0744+55) were found in the records. Their profiles and dynamic spectra are shown in Fig. 1. In the case of J0744+55, a profile with a double peak is observed in some of the pulses, indicating the possibility of subpulses in this transient (see Fig. 2). The distance between the peaks in the profile is 25–35 ms.

Fig. 1.

Top panel: profiles of the strongest pulses of the found transients. The vertical axis is the flux density in janskys. Bottom panel: dynamic spectra of these pulses. The vertical axis represents the frequencies of several channels. The time intervals on the horizontal axis of the pulse profile and the corresponding dynamic spectrum coincide.

Fig. 2.

J0745+55 transient profile showing a double peak.

Table 1 provides information on the detected RRATs. Columns 1–3 contain the name of the transient and the coordinates of the source in right ascension and declination. The right ascension was determined as the median value of the detected pulses, and the error, due to the small number of detected pulses, was estimated as the size of the LPA pattern at half power. The pulses for J0249+52 were visible in one beam and not visible in the beams above and below, so the declination coordinate was determined as the beam declination, and the coordinate estimation accuracy was determined as half the declination distance between the beams. For J0744+55, the pulses were visible in two neighboring beams, and the coordinate was determined as the average of the declination coordinates of the beams. Columns 4–6 provide estimates of the dispersion measure $DM$, profile half-width ${{W}_{{0.5}}}$, and peak flux densities ${{S}_{{{\text{peak}}}}}$ of the weakest and the strongest pulse separated by a forward slash (“/”). Since the calibration step of a known temperature is written in all frequency channels 6 times a day, the data in the channels are calibrated using the step. Thus, the height of the found pulse is known in units of temperature. Recording is ongoing around the clock. In addition to the calibration steps, the record also contains discrete sources with a known flux density. Therefore, it is possible to recalculate the observed peak flux densities from temperature units to janskys. The given estimates are the lower estimates of the flux density. The exact coordinate of the transient is unknown both in right ascension and declination. Therefore, it is not possible to make corrections that take into account the possible impact of the pulse on the edge of the LPA radiation pattern and the possible discrepancy between the coordinates of the LPA beam and the declination coordinates of the transient. As a result, the peak flux density can be underestimated by up to 2 times. Column 7 indicates how many pulses were detected (${{N}_{1}}$) from high time-frequency data. A slash “/” indicates the number of days of pulse detection (${{N}_{2}}$). Column 8 shows the frequency of occurrence of pulses ($n$) with S/N $>$ 10 (${{S}_{{{\text{peak}}}}} > 3.5{-} 4$ Jy) in 32-channel data for one hour of observations. When obtaining the estimate of $n$, it was assumed that pulses are most likely to appear in the central part of the LPA3 radiation pattern, which is approximately equal to 3.5 min at half power. In total, (312 sessions × 3.5 min)/60 min = 18.2 h of observations are accumulated in the direction of each transient.

Table 1.   Characteristics of the found RRATs

Both RRATs had approximately the same number of pulses. However, while pulses for J0744+55 appeared almost evenly throughout the entire period of observations, all 8 pulses for J0249+52 were recorded in four consecutive months.

Since RRATs are pulsars, we made an attempt to find their periodic emission. The area with declinations $+ 52^\circ < \delta < + 55^\circ$ has not been studied before in the Pushchino Multibeams Pulsar Search (PUMPS) [17]. Assuming that the found RRATs could be ordinary second pulsars, we performed a standard search using Fourier power spectra. The days with the best quality of the noise track were selected for the search. In total, a third of the records were discarded out of 312 days of observations. For the remaining days, power spectra were obtained using the fast Fourier transform, which then were added. When adding the spectra, an increase in the S/N harmonics should be observed if there are periodic signals in the directions under study. In the summed power spectra, no features were found in the power spectra at the level ${\text{S/N}} > 5$ and at periods $P < 2$ s. It was shown in the early studies that the observed increase in S/N harmonics when spectra are summed is less than the root of the number of added sessions [17]. The increase in sensitivity when adding power spectra was estimated to be around 10 times, rather than 15 times, as could be expected. Considering the background temperature in the direction of the found transients and assuming the periods of the transients $P < 2$ s, we can give an upper bound on the expected peak flux density for regular pulsar emission: ${{S}_{{{\text{peak}}}}} < 0.3$ Jy (J0249+52), ${{S}_{{{\text{peak}}}}} < 0.2$ Jy (J0744+55). Since the periods of the found RRATs were not determined, it was not possible to obtain estimates of the integrated flux density.

In addition to using power spectra, an estimate of the RRAT period can be obtained based on the observed time interval between pulses by choosing the largest common divisor (the time interval that fits an integer number of times between the occurrences of any pulses), which will be the upper estimate of the transient period. The actual period can be an integer number of times smaller. Our search was carried out for obviously strong pulses (${\text{S/N}} > 10$ for the total pulse profile over 32 frequency channels). To estimate the period, we searched for weaker pulses (up to S/N = 5) in the vicinity of the found strong pulses. Weak pulses were not detected.

The nature of the discovered RRATs remains unclear. Previous studies [8, 18] have shown that RRATs are a mixture of known types of pulsars. Some of them are pulsars with very long nullings, while others are pulsars with a wide energy distribution of pulses, and for weak pulsars, strong pulses are observed from the tail of this distribution. Some RRATs are also pulsars with giant pulses. In the present study, two of the detected RRATs have only one pulse in 2.5 h of observations. If we assume that the found RRATs have a period $P = 1$ s, their nullings will be equal to 99.99%. This means that we see only one pulse out of 10 000. The upper peak flux densities in the average profile for J0249+52 and J0744+55 are 0.3 and 0.2 Jy, respectively. In this case (see Table 1), the observed pulse flux densities exceed the upper estimates of the peak flux density in the average profile by a factor of 40–100 or more. This difference in peak densities in an individual pulse and in the average profile can be inherent in both pulsars with giant pulses and pulsars with a long tail of the pulse energy distribution. The absence of regular emission supports the idea of the nulling nature of the discovered RRATs. However, for a definitive conclusion about the nature of J0249+52 and J0744+55, further observations with radio telescopes that are more sensitive than the LPA LPI are needed.

5 DISCUSSION OF RESULTS AND CONCLUSIONS

The program for monitoring the quality of ongoing observations based on data recorded with a low frequency-time resolution has shown high efficiency, making it possible to quickly respond to changes in both external conditions of observations and internal causes responsible for the deterioration of observations. In the course of observations, a pulsed signal is recorded that is visible in one channel or simultaneously in many beams on average every 3 min. In general, the quality of observations is high.

The main part of the detected pulses is associated with noise, and cluster analysis had to be used to isolate pulsar pulses from the sample. Tens of thousands of pulses with signs of “pulsation” were detected in the six-channel data over an almost one-year period of observations. For each of these events, a high frequency pulse arrives earlier than a low frequency pulse and is recorded in one or two neighboring beams. Checking these pulses shows that about half of them belong to pulsars, and the rest of the pulses are various kinds of processing artifacts.

A blind search led to the discovery of 4 known pulsars with 1 to more than 10 000 pulses in data with low frequency-time resolution over a total of 7496 observation hours (equivalent to over 312 days). In addition to the known pulsars, two new RRATs were also discovered. The occurrence of transient pulses on average every 2.5 hours corresponds to known cases [5]. The total number of RRATs detected at the LPA LPI reached 48 events.⁴

Based on the analysis, it is highly likely that the detected transients are pulsars with very long nullings. However, a definitive conclusion requires observations using radio telescopes with higher instantaneous sensitivity than that of the LPA LPI radio telescope.