[IPTA Clock] Progression of the clock project

[George.Hobbs at csiro.au]

Hi all,

 Nicolas has asked me how he can get involved more in the clock project.  I plan to write back specifically to him, but other people on this list may also be interested in a more general response.    A few dot points:

--
1. The PPTA team (mainly myself, Bill and Dick) have developed a technique for measuring the clock errors in a given PTA data set.  That technique has been applied to PPTA data and has been published by MNRAS.   

2. The PPTA data (particularly around the year 2000) show some interested features, but the data are not quite good enough for us to really be convinced that the pulsar data is showing a real clock error.  Therefore, it's clear that we should use the IPTA data set to improve our sensitivity.

3. My initial plan was simply to get the IPTA data set, run our algorithm on it and publish.  Done!  No real need for a large team.

4. Part of me still thinks in the same way.  This is not necessarily a major project to develop new algorithms etc.  We already have an algorithm that should work.  However, unless we get earlier NANOgrav data, I'm now of the opinion that the sensitivity of the IPTA data set will not be significantly better than just the PPTA data (note: earlier NANOGrav data would make a massive improvement).

5. If the sensitivity is not improved much then we need to do something else if we want to write a paper.  

6. KJ and Rutger both mentioned to me that they'd like to modify their GW detection algorithms to search for the clock signal and then compare that with the original PPTA algorithm.

7. Mike is very interested in applying a split-Cholesky algorithm to the clock signal and also to consider using his "common mode" time series.

8. Mike is interested in testing the algorithms via realistic simulations.

All this is good and  KJ, Rutger, Mike and I have some well-defined work to do.  However, there are quite a few more things that could be thought about:

a) making Mike's simulations more realistic.  They currently include DM variations, realistic sampling, white noise levels, red noise etc., but not RFI, strange points, EFACs/EQUADs, frequency dependent pulse shape variations etc.  The "toasim" structure allows new algorithms to be easily added and we'd be happy to help people learn how to use the toasim tools.

b) finishing the data sets.  Of course, we plan to use the "official" IPTA data sets, but we'll need to measure red noise spectra, model the DM variations etc.  

c) understanding the implications of the pulsar time scale.  This is a big sub-project and I'd certainly push for people to get involved in this.  We measure TT(IPTA).  What are we going to do with it?  What does it mean?  How can we combine it with TT(BIPM2012) to make an improved, combined data set?

d) simulating the future.  I'm guessing that all IPTA papers are going to have a section on "how well we'll do when we have the full IPTA data sets that extend another 10 years".  Somebody needs to make simulated data sets that we can then run our algorithms on and to consider how we'll present this work.

e) other algorithms could be developed.  We have the PPTA algorithm, KJ's and Rutger's, but I also know that some of the NANOGrav algorithms could easily be updated to search for the clock signal.  It's a shame that e.g. Sydney Chamberlin is not on this email list as, I think, she has an algorithm that could be applied to this problem.  My plan is to contact the NANOgrav people personally to try and get them more involved.

f) the literature contains suggestions of using the binary orbital parameters as the "pulsar clock" instead of the "pulsar spin".  I doubt that we'd discuss this (or implement any algorithms) for the IPTA paper, but somebody may wish to consider that problem in detail.

g) do we really believe our algorithms?  We'll have a few algorithms and they'll all give different values for the clock signal and its uncertainty.  Do we just use simulation to work out whether they are consistent or not?  We know that it's hard to present the uncertainties on the clock signal as, in the PPTA algorithm, the individual clock points can be highly covariant.  Detailed consideration of this would be very useful.

h) getting hold of the early NANOGrav data and combining that with our EPTA and PPTA data sets.  This would make a major increase in our sensitivity.

I'm sure that there are lots more things to do.  Please respond to this email if you have other ideas or wish to discuss a particular sub-project in more detail.

Cheers,
 George