Detection of 107 glitches in 36 southern pulsars

Timing observations from the Parkes 64-m radio telescope for 165 pulsars between 1990 and 2011 have been searched for period glitches. Data spans for each pulsar ranged between 5.3 and 20.8 yr. From the total of 1911 yr of pulsar rotational history, 107 glitches were identified in 36 pulsars. Out of these glitches, 61 have previously been reported whereas 46 are new discoveries. Glitch parameters, both for the previously known and the new glitch detections, were measured by fitting the timing residual data. Observed relative glitch sizes Delta nu(g)/nu range between 10(-10) and 10(-5), where nu = 1/P is the pulse frequency. We confirm that the distribution of Delta nu(g)/nu is bimodal with peaks at approximately 10(-9) and 10(-6). Glitches are mostly observed in pulsars with characteristic ages between 10(3) and 10(5) yr, with large glitches mostly occurring in the younger pulsars. Exponential post-glitch recoveries were observed for 27 large glitches in 18 pulsars. The fraction Q of the glitch that recovers exponentially also has a bimodal distribution. Large glitches generally have low Q, typically just a few per cent, but large Q values are observed in both large and small glitches. Observed time constants for exponential recoveries ranged between 10 and 300 d with some tendency for longer time-scales in older pulsars. Shorter time-scale recoveries may exist but were not revealed by our data which typically have observation intervals of 2-4 weeks. For most of the 36 pulsars with observed glitches, there is a persistent linear increase in (nu) over dot (i.e. decrease in the slow-down rate vertical bar(nu) over dot vertical bar) in the interglitch interval. Where an exponential recovery is also observed, the effects of this are superimposed on the linear increase in (nu) over dot. In some but not all cases, the slope of the linear recovery changes at the time of a glitch. The double (nu) over dot values characterizing the linear changes in (nu) over dot are almost always positive and, after subtracting the magnetospheric component of the braking, are approximately proportional to the ratio of vertical bar(nu) over dot vertical bar and the interglitch interval, as predicted by vortex-creep models.