基于深层残差网络的脉冲星候选体分类方法研究

随着下一代射电天文望远镜的不断改进和发展,脉冲星巡天观测将发现数百万个脉冲星候选体,这给脉冲星的识别和新脉冲星的发现带来了巨大挑战,迅速发展的人工智能技术可用于脉冲星识别.使用Parkes望远镜的脉冲星数据集(The High Time Resolution Universe Survey,HTRUS),设计了一个14层深的残差网络(Residual Network,ResNet)进行脉冲星候选体分类.在HTRUS数据样本中,存在非脉冲星候选体(负样本)的数目远远大于脉冲星候选体(正样本)数目的样本非均衡问题,容易产生模型误判.通过使用过采样技术对训练集中的正样本进行数据增强,并调整正负样本的比例,解决了正负样本非均衡问题.训练过程中,使用5折交叉验证来调节超参数,最终构建出模型.测试结果表明,该模型能够取得较高的精确度(Precision)和召回率(Recall),分别为98%和100%,F1分数(F1-score)能够达到99%,每个样本检测完成只需要7 ms,为未来脉冲星大数据分析提供了一个可行的办法.     

The Parkes Multibeam Pulsar Survey – VI. Discovery and timing of 142 pulsars and a Galactic population analysis

We present the discovery and follow-up observations of 142 pulsars found in the Parkes 20-cm multibeam pulsar survey of the Galactic plane. These new discoveries bring the total number of pulsars found by the survey to 742. In addition to tabulating spin and astrometric parameters, along with pulse width and flux density information, we present orbital characteristics for 13 binary pulsars which form part of the new sample. Combining these results from another recent Parkes multibeam survey at high Galactic latitudes, we have a sample of 1008 normal pulsars which we use to carry out a determination of their Galactic distribution and birth rate. We infer a total Galactic population of  30 000 ± 1100  potentially detectable pulsars (i.e. those beaming towards us) having 1.4-GHz luminosities above 0.1 mJy kpc². Adopting the Tauris & Manchester beaming model, this translates to a total of  155 000 ± 6000  active radio pulsars in the Galaxy above this luminosity limit. Using a pulsar current analysis, we derive the birth rate of this population to be  1.4 ± 0.2  pulsars per century. An important conclusion from our work is that the inferred radial density function of pulsars depends strongly on the assumed distribution of free electrons in the Galaxy. As a result, any analyses using the most recent electron model of Cordes & Lazio predict a dearth of pulsars in the inner Galaxy. We show that this model can also bias the inferred pulsar scaleheight with respect to the Galactic plane. Combining our results with other Parkes multibeam surveys we find that the population is best described by an exponential distribution with a scaleheight of 330 pc. Surveys underway at Parkes and Arecibo are expected to improve the knowledge of the radial distribution outside the solar circle, and to discover several hundred new pulsars in the inner Galaxy.     

Pulsar candidate classification with deep convolutional neural networks

As the performance of dedicated facilities has continually improved, large numbers of pulsar candidates are being received, which makes selecting valuable pulsar signals from the candidates challenging. In this paper, we describe the design for a deep convolutional neural network(CNN) with 11 layers for classifying pulsar candidates. Compared to artificially designed features, the CNN chooses the subintegrations plot and sub-bands plot for each candidate as inputs without carrying biases. To address the imbalance problem, a data augmentation method based on synthetic minority samples is proposed according to the characteristics of pulsars. The maximum pulses of pulsar candidates were first translated to the same position, and then new samples were generated by adding up multiple subplots of pulsars. The data augmentation method is simple and effective for obtaining varied and representative samples which keep pulsar characteristics. In experiments on the HTRU 1 dataset, it is shown that this model can achieve recall of 0.962 and precision of 0.963.     

射电脉冲星周期跃变研究的进展

射电脉冲星周期跃变被认为是研究中子星内部结构和状态的极好探针。脉冲星高频巡天发现了一批年青脉冲星,脉冲星周期跃变的观测研究也有了飞快进展。至少发现了２５颗有跃变现象的脉冲星（简称跃变脉冲星）和７６次跃变事件。ＰＳＲＪ０８３５－４５１０是目前已有跃变脉冲星活动参数最高的,ＰＳＲＪ１３４１－６２２２０的跃变活动最频繁,而ＰＳＲＪ１６１４－５０４７在１９９５年发生的跃变是规模最大的,不同脉冲星的跃变事件     

Discovery of 28 pulsars using new techniques for sorting pulsar candidates

Modern pulsar surveys produce many millions of candidate pulsars, far more than can be individually inspected. Traditional methods for filtering these candidates, based upon the signal-to-noise ratio of the detection, cannot easily distinguish between interference signals and pulsars. We have developed a new method of scoring candidates using a series of heuristics which test for pulsar-like properties of the signal. This significantly increases the sensitivity to weak pulsars and pulsars with periods close to interference signals. By applying this and other techniques for ranking candidates from a previous processing of the Parkes Multi-beam Pulsar Survey, 28 previously unknown pulsars have been discovered. These include an eccentric binary system and a young pulsar which is spatially coincident with a known supernova remnant.     

On measuring the gravitational-wave background using Pulsar Timing Arrays

The long-term precise timing of Galactic millisecond pulsars holds great promise for measuring the long-period (months to years) astrophysical gravitational waves. Several gravitational-wave observational programs, called Pulsar Timing Arrays (PTA), are being pursued around the world.
Here, we develop a Bayesian algorithm for measuring the stochastic gravitational-wave background (GWB) from the PTA data. Our algorithm has several strengths: (i) it analyses the data without any loss of information; (ii) it trivially removes systematic errors of known functional form, including quadratic pulsar spin-down, annual modulations and jumps due to a change of equipment; (iii) it measures simultaneously both the amplitude and the slope of the GWB spectrum and (iv) it can deal with unevenly sampled data and coloured pulsar noise spectra. We sample the likelihood function using Markov Chain Monte Carlo simulations. We extensively test our approach on mock PTA data sets and find that the algorithm has significant benefits over currently proposed counterparts. We show the importance of characterizing all red noise components in pulsar timing noise by demonstrating that the presence of a red component would significantly hinder the detection of the GWB.
Lastly, we explore the dependence of the signal-to-noise ratio on the duration of the experiment, number of monitored pulsars and the magnitude of the pulsar timing noise. These parameter studies will help formulate observing strategies for the PTA experiments.     

The cross-correlation between the microwave and X-ray backgrounds: foregrounds and systematics

In this paper, we describe a novel experiment for the accurate estimation of pulsar dispersion measures (DMs) using the Giant Metrewave Radio Telescope. This experiment was carried out for a sample of 12 pulsars, over a period of more than one year (2001 January to 2002 May) with observations about once every fortnight. At each epoch, the pulsar DMs were obtained from simultaneous dual-frequency observations, without requiring any absolute timing information. The DM estimates were obtained from both the single-pulse data streams and from the average profiles. The accuracy of the DM estimates at each epoch is ∼1 part in 10⁴ or better, making the data set useful for many different kinds of studies.
The time-series of DMs shows significant variations on time-scales of weeks to months for most of the pulsars. An analysis of the mean DM values from these data shows significant deviations from catalogue values (as well as from other estimates in the literature) for some of the pulsars, with PSR B1642−03 showing the most notable differences. From our analysis results it appears that the constancy of pulsar DMs (at the level of 1 in 10³ or better) cannot be taken for granted. For PSR B2217+47, we see evidence of a large-scale DM gradient over a 1-yr period, which is modelled as being due to a blob of enhanced electron density sampled by the line of sight. For some pulsars, including pulsars with fairly simple profiles such as PSR B1642−03, we find evidence for small changes in DM values for different frequency pairs of measurement, a result that needs to be investigated in detail. Another interesting result is that we find significant differences in DM values obtained from average profiles and single-pulse data.     

基于双谱滤波的综合脉冲星时算法研究

毫秒脉冲星的自转频率非常稳定,提供了一种独立的基于遥远自然天体并能持续数百万乃至数十亿年的时间基准,具有稳定性强、运行时间长、服务范围广等特点.为了减弱毫秒脉冲星计时观测中各种高斯噪声对脉冲星时的影响,研究了一种基于双谱滤波的综合脉冲星时构建算法,处理分析了国际脉冲星计时阵(International Pul-sar Timing Array,IPTA)最新发布的4颗毫秒脉冲星(PSR J0437-4715、J0613-0200、J1713+0747和J1909-3744)的观测数据,分析了不同时间尺度综合脉冲星时的稳定性,并与构成国际原子时(International Atomic Time,TAI)的4家授时单位原子钟稳定性进行了比较.结果表明:双谱滤波算法能够较好地抑制观测噪声,提高综合脉冲星时的稳定性.相比于经典加权算法,综合脉冲星时1 yr、10 yr稳定度从7.77×10^-14、8.56×10^-16分别提高到1.50×10^-14、3.50×10^-16,单脉冲星时稳定性的提升也类似.同时发现,综合脉冲星时稳定性在5 yr及以上时间尺度上优于原子钟稳定性,可用于改善当前原子时的长期稳定性.     

基于卷积神经网络的快速射电暴候选体分类

针对目前从海量的快速射电暴(Fast Radio Burst, FRB)候选体中人工筛选FRB事件难以为继的现状,提出了一种基于卷积神经网络(Convolutional Neural Networks, CNN)的FRB候选体分类方法.首先,通过真实的观测数据和仿真FRB组成训练和测试样本集.其次,建立了二输入的深度卷积神经网络模型,并对其进行训练、测试和优化,获取FRB候选体分类器.最后,利用来自脉冲星的单脉冲数据对该分类器的有效性和性能进行了验证.实验结果表明,该方法可以快速从大量候选体中准确识别出单脉冲事件,极大地提高了FRB候选体的处理速率和效率.     

The Parkes Southern Pulsar Survey — II. Final results and population analysis

A survey of the entire southern sky for millisecond and low-luminosity pulsars using the ATNF Parkes radio telescope has now been completed. The survey detected 298 pulsars, of which 101 were previously unknown. The new pulsars include 17 millisecond pulsars. This is the largest sample of both normal and millisecond pulsars detected in any survey. Combining our sample with other recent surveys in the Northern Hemisphere, we present a statistical study of the populations of both normal and millisecond pulsars. We find that the improved statistics allow us to estimate the number and birth-rate of both types of pulsar down to a 400-MHz luminosity limit of 1 mJy kpc². The local surface densities of potentially observable normal pulsars and millisecond pulsars are both about 30 kpc⁻², corresponding to ∼ 30000 potentially observable pulsars of each type in the Galaxy. Once beaming effects are taken into consideration we estimate that the active population of normal pulsars is ∼ 160000. Although there is evidence for flattening of the luminosity function of normal pulsars, this is not evident for millisecond pulsars which probably have a substantial population with luminosities below 1 mJy kpc². After correcting for beaming effects, we estimate that a normal pulsar is born with a luminosity greater than 1 mJy kpc² between once every 60 and 330 yr in the Galaxy. The birth-rate of millisecond pulsars is at least 3 × 10⁻⁶ yr⁻¹ above the same luminosity limit. Modelling the observed transverse speeds of millisecond pulsars using a dynamical simulation, we find their mean birth velocity to be 130 ± 30 km s⁻¹, significantly lower than that of the normal pulsars.     

The Deutsch field gamma-ray pulsar – II. Application of the model

A new model for gamma-ray pulsars presented by Higgins & Henriksen is applied to the cases of the seven known gamma-ray pulsars. Those pulsars that are not presently observed in gamma-rays, but are candidates for observation by the next generation of gamma-ray telescopes, are discussed. The case of millisecond pulsars is discussed, and it is shown that these objects should radiate at detectable levels, in opposition to the predictions of other gamma-ray pulsar models.     

An interference removal technique for radio pulsar searches

Searches for radio pulsars are becoming increasingly difficult because of a rise in impulsive man-made terrestrial radio-frequency interference. Here, we present a new technique, zero dispersion measure filtering, which can significantly reduce the effects of such signals in pulsar search data. The technique has already been applied to a small portion of the data from the Parkes multi-beam pulsar survey, resulting in the discovery of four new pulsars, so illustrating its efficacy.     

Red supergiants in the LMC — II. Spectrophotometry and model atmospheres

We have analysed polarization data for a large number of isolated pulsars to investigate the evolution of pulsar radio beams. Assuming that a circular beam is directed along the axis of a dipolar magnetic field, we demonstrate that the distribution of magnetic inclination angles for the parent population of all pulsars is not flat but highly concentrated towards small inclination angles and that, consequently, the average beaming fraction is only ∼ 10 per cent. Furthermore, we find that there is a tendency for the beam axis to align with the rotational axis on a time-scale of ∼ 10⁷ yr. This has interesting consequences for statistical studies of the pulsar population. Finally, the luminosity of pulsars is shown to be independent of the impact parameter, which indicates that pulsar beams are sharp-edged and have a relatively flat integrated intensity distribution.     

The gamma-ray conversion efficiency of rotation-powered pulsars

We use a new self-consistent model to derive the conversion efficiency from rotation power to γ-ray power for pulsars (ε_th). Our result indicates that ε_th∝τ^6/7 P ², where τ and P are the characteristic age and period of the pulsar, which shows that although the efficiency increases with the characteristic age of the pulsar, it also depends on the pulsar period. We test our model results with the survey of high-energy γ-rays from pulsars by EGRET. Our model not only successfully explains the efficiency of the confirmed γ-ray pulsars but also explains why the γ-ray efficiency of millisecond pulsars is so low.     

The Parkes Southern Pulsar Survey — III. Timing of long-period pulsars

The Parkes survey of the entire southern sky for millisecond and other low-luminosity pulsars has now been completed. The survey system and initial results were described by Manchester et al. and the final results and population studies are described by Lyne et al. A total of 298 pulsars were detected, including 101 new discoveries of which 17 were millisecond pulsars. Here we report on timing observations at Parkes and Jodrell Bank of the 84 long-period (non-millisecond) pulsars discovered in the survey, including timing solutions for 78 of them. Pulse profiles and arrival times were obtained at several different frequencies over intervals of more than a year, yielding a position, period, period derivative and dispersion measure for each pulsar. Pulse profiles at frequencies near 400 MHz and 600 or 1400 MHz are presented for most of the observed pulsars. Significant timing noise was detected for five pulsars and a glitch was observed in the period of PSR J1123–6259.     

Analysis of the Precision of Pulsar Time Clock Modeltwo

Millisecond pulsars have a very high rotation stability, which can be applied to many research fields, such as the establishment of the pulsar time standard, the detection of gravitational wave, the spacecraft navigation by using X-ray pulsars and so on. In this paper, we employ two millisecond pulsars PSR J0437-4715 and J1713+0743, which are observed by the International Pulsar Timing Array (IPTA), to analyze the precision of pulsar clock parameter and the prediction accuracy of pulse time of arrival (TOA). It is found that the uncertainty of spin frequency is 10^?15 Hz, the uncertainty of the first derivative of spin frequency is 10^?23 s^?2, and the precision of measured rotational parameters increases by one order of magnitude with the accumulated observational data every 4～5 years. In addition, the errors of TOAs within 4.8 yr which are predicted by the clock model established by the 10 yr data of J0437-4715 are less than 1 μs. Therefore, one can use the pulsar time standard to calibrate the atomic clock, and make the atomic time deviate from the TT (Terrestrial Time) less than 1 μs within 4.8 yr.     

New pulsar rotation measures and the Galactic magnetic field

We measured a sample of 150 pulsar rotation measures (RMs) using the 20-cm receiver of the Parkes 64-m radio telescope. 46 of the pulsars in our sample have not had their RM values previously published, whereas 104 pulsar RMs have been revised. We used a novel quadratic fitting algorithm to obtain an accurate RM from the calibrated polarization profiles recorded across 256 MHz of receiver bandwidth. The new data are used in conjunction with previously known dispersion measures and the NE2001 electron-density model to study models of the direction and magnitude of the Galactic magnetic field.     

TeV neutrinos and gamma-rays from pulsars

Recent studies suggest that pulsars could be strong sources of TeV muon neutrinos provided positive ions are accelerated by pulsar polar caps to PeV energies. In such a situation, muon neutrinos are produced through the Δ-resonance in interactions of pulsar-accelerated ions with its thermal radiation field. High-energy gamma-rays should also be produced simultaneously in pulsar environment as both charged and neutral pions are generated in the interactions of energetic hadrons with the ambient photon fields. Here, we estimate TeV gamma-ray flux at the Earth from a few nearby young pulsars. When compared with the observations, we find that proper consideration of the effect of polar cap geometry in flux calculation is important. Incorporating such an effect, we obtain the (revised) event rates at the Earth due to a few potential nearby pulsars. The results suggest that pulsars are unlikely to be detected by the upcoming neutrino telescopes. We also estimate TeV gamma-ray and neutrino fluxes from pulsar nebulae for the adopted model of particle acceleration.     

脉冲星对氢原子钟的频率驾驭算法研究

氢原子钟具有较高的短期稳定度,将其作为主钟可在短期内产生高精度的本地时间信号.但氢钟存在频率漂移现象,导致其长期稳定度较差,从而影响本地时间的准确性.毫秒脉冲星自转高度稳定,借助于其长期稳定度高的特性,可定期实现对氢原子钟的频率驾驭,并对实时信号加以控制.首先分析了国际脉冲星计时阵(International Pulsar Timing Array,IPTA)第二批发布数据中四颗毫秒脉冲星的稳定度随时间的变化,同时采用哈达玛方差分析了中国科学院国家授时中心(National Time Service Center,NTSC)一台氢钟的频率稳定性能,最终给出了利用脉冲星驾驭氢原子钟频率的方法.     

High-energy γ-rays from globular clusters

It is expected that specific globular clusters (GCs) can contain up to a hundred of millisecond pulsars. These pulsars can accelerate leptons at the shock waves originated in collisions of the pulsar winds and/or inside the pulsar magnetospheres. Energetic leptons diffuse gradually through the GC Comptonizing stellar and microwave background radiation. We calculate the GeV–TeV γ-ray spectra for different models of injection of leptons and parameters of the GCs assuming reasonable, of the order of 1 per cent, efficiency of energy conversion from the pulsar winds into the relativistic leptons. It is concluded that leptons accelerated in the GC cores should produce well localized γ-ray sources which are concentric with these GCs. The results are shown for four specific GCs (47 Tuc, Ter 5, M13 and M15), in which significant population of millisecond pulsars have been already discovered. We argue that the best candidates, which might be potentially detected by the present Cherenkov telescopes and the planned satellite telescopes (AGILE, GLAST), are 47 Tuc on the Southern hemisphere, and M13 on the Northern hemisphere. We conclude that detection (or non-detection) of GeV–TeV γ-ray emission from GCs by these instruments put important constraints on the models of acceleration of leptons by millisecond pulsars.