Tipos de estructuras organizacionales.

Since the discovery of the original 11 sources, followup timing observations have been performed primarily at Parkes, as part of observing proposal P511, with a few additional observations undertaken at Arecibo. Numerous observations of J1819−1458 and J1913+1330 were performed at Jodrell Bank, as have been described in Chapter 5. In addition to the PMSingle sources, Table 6.1 lists the timing solutions which have been obtained for 7 of the 11 sources.

6.1.4

New Discoveries

Using the Parkes Telescope, two surveys have been performed off the Galactic plane, i.e. outside the region where the PMPS observations were. These were performed at intermediate and high Galactic latitudes of 5◦ < |b| < 30◦ (Edwards et al., 2001; Jacoby et al., 2009). The surveys used the same specifications as the PMPS, except with faster time sampling of 125 µs and shorter pointings of 4.4 minutes. Recently, Burke-Spolaor & Bailes (2010) have analysed these surveys in search of isolated bursts and presented 14 new sources, 7 of which were candidates which had never been confirmed. One of these unconfirmed sources was in fact re-detected by the authors soon after their publication (Burke- Spolaor, private communication), but 6 sources remained unconfirmed. As part of the P661 observing project, these 6 sources were observed in search of single pulses and 3 of these sources have been confirmed. Two of these sources have been regularly observed since January 2010 and both have provisional timing solutions, coherent over a timescale of & 200 days, at the time of writing.

J0735−62

J0735−62 is not detected in two followup observations, each of ten minutes dura- tion. Recently we have made a third, 30-minute observation where it was easily detected, and thus confirmed for the first time, with 20 strong single pulses. Analysing the TOA differences, as described in § 4.4, we determine a topocentric period of P = 4.865(1) s, consistent with the initial estimate of P = 4.862 s period published by Burke-Spolaor & Bailes (2010). Additionally the two non- detections support their claim that the source suffers from severe scintillation. For this reason we do not yet know if this source is solvable using a reasonable amount of observing time, but a single lengthy observation is planned for P661 observing sessions in the October 2010 – March 2011 semester, to investigate this very question.

J1226−32

The original detection of J1226−32 contained only 3 pulses but this was sufficient for Burke-Spolaor & Bailes (2010) to predict a period of P = 6.193 s. We have confirmed this candidate and have observed 45 pulses in almost 3 hours of fol- lowups, although in one third of the observations it is not detectable. We confirm

the published period. Our provisional timing solution is coherent since January 2010 and regular observations as part of the P661 project are planned. These should reveal a full timing solution once the data span surpasses the one year mark.

J1654−23

The original detection of J1654−23 also consisted of just 3 pulses. We have confirmed this source and have determined a period of 545 ms, which differs from the published estimate of Burke-Spolaor & Bailes (2010). This is not very surprising given their small number of detected pulses. Interestingly, the period we determine, from 106 pulses detected in 2.3 hours, is not at a different harmonic. This suggests that perhaps one of the 3 pulses initially identified was terrestrial in origin. As for J1226−32 we have a provisional timing solution, coherent since January 2010 and regular observations as part of the P661 project are planned. A full timing solution is expected for this source, once a year of monitoring has been made.

Unconfirmed Sources

In addition to the above three sources, we have attempted to confirm three other sources. We have observed J0923−31 and J1610−17 for 1.0 and 1.2 hours re- spectively but have not been able to make a confirmation. We have detected 5 weak pulses from J1753−12 at the correct DM, during 1.3 hours of observation, although we hope a more significant confirmation will come with time. We have not yet followed up these 3 sources for as long as the 3 new confirmations. This is, in some sense, by design, as these sources showed just 1, 1 and 3 pulses re- spectively in their discovery observations, so we decided to initially focus on the higher burst rate source (which were subsequently confirmed).

6.1.5

An Aside: The Perils of EFAC

If we have some data, and a model for that data, it is common to determine the χ2/nfree value for the model, to test its validity. Suppose the data are as shown in

Figure 6.3. In this example, if your model is a straight line of slope 1 you will get a large χ2/nfree value, suggestive of a ‘bad fit’. This can be interpreted in two ways:

-35 -30 -25 -20 -15 -10 -5 0 5 10 15 0 2 4 6 8 10 Data x data model data + EFAC

Figure 6.3: This plot illustrates the dangers of using EFAC in pulsar timing. The

χ2/nfree for the top straight-line fit to the data is large. If we blindly scale our errors to

force χ2/nfree = 1 then we can obtain a ‘good fit’ to a straight line but we have washed

out the true underlying sinusoidal features.

underestimated the errors in your data points. If you have χ2/nfree= EFAC2then

scaling up your errors by a factor of EFAC will give you χ2/nfree = 1, a good fit!

Such an operation is very bad practise, completely unjustified and should in no way be endorsed. Furthermore, forcing χ2_/n

free to equal 1 completely defeats

the purpose of the test for checking the validity of the model, i.e. you are simply accepting the model without testing it and setting the ‘best fit’ solution with this model to be the true solution. Unfortunately, the use of EFAC has actually been practised in pulsar timing studies in the past. We do not recommend this. For instance, if we had taken the χ2_/n

free of our timing model for J1707−4417,

ignoring the systematic signature of the two well-separated bands, we might apply an EFAC of 30. With an average measured TOA error of 2.8 ms, this means we would scale this up to ∼ 84 ms. This essentially suggests that we can determine the peak of each pulse only to an accuracy of several times the pulse width (!) when in fact we expect a relationship of the form σTOA ≈ W/(S/N ). Even in the

cases where all systematic trends have been removed there is still no justification for an EFAC, as our model cannot be perfect, e.g. the timing solutions presented above implicitly assume a stable pulse profile but this is clearly not the case for some of the sources with extra scatter in Figure 6.2, above and beyond the size

of the residuals. Another similar quantity sometimes used in pulsar timing is EQUAD, an error added in quadrature to all TOAs. However, just as there is no reason to assume all our measurements are incorrect by a constant factor, there is no reason to assume that all of our measurements are subject to an identical additive contribution from some unknown systematic. Use of EFAC and EQUAD is not recommended.