arXiv2004
This paper has been withdrawn from astro-ph and from the submission process to ApJ Letters since the variations "observed" are not real. As pointed out to me first by Willem van Straten and shortly thereafter by Matthew Bailes and Dick Manchester, the "Orbital Modulation of the Apparent Dispersion Measure Towards PSR J0737-3039A" is, in fact, an artifact of the data reduction. The problem occurs only for observations using very wide bandwidths of fast-moving and compact-orbit binary pulsars when reduced in a "traditional" manner (for folding of data, this means using a single set of polycos to fold the full bandwidth of data -- as was done for this paper). To quote Dick Manchester: "I believe I know the reason for the apparent DM variation in your data. We saw a similar effect as soon as we started taking the dual-freq [i.e. simultaneous 10cm/50cm data from Parkes] data. I showed that the effect was due to calling the polyco for folding the data with an incorrect RF frequency. The differential delay due to the dispersion moves the apparent orbital phase by a significant amount for these very short-period binaries. This leads to a sinusoidal modulation at the orbital period in the residuals, the amplitude (and sign) of which is a function of the difference between the prediction frequency and the observed frequency." and Willem Van Straten: "We were using a polyco created with a constant centre frequency [...] to fold data observed at all other frequencies. The canonical thinking behind this was to leave the relative dispersion delays in the data, so that tempo could later fit for DM. However, because the position of PSR J0737-3039A changes significantly during the dispersion delay time, the binary phase predicted at the [...] barycentric proper time [for the observing frequency used for the polycos] is significantly different to that at the proper time of other frequencies. For most pulsars, this doesn't matter much. But v/c for PSR J0737-3039A is about 10^-3. Another way to describe the problem is that, although the dispersion time delay does not evolve with binary phase, the corresponding pulsar _phase_ varies strongly as a function of dispersion delay (and therefore frequency)." Matthew Bailes pointed out that if this Doppler-like effect is not accounted for, than for any binary pulsar, an orbital variation in DM will occur of amplitude: dDM = V_orb/c * DM. This implies that for binaries where this effect is significant, an orbital phase dependent de-dispersion methodology is required. For pulsar folding, that can be accomplished by using a different set of polycos for _each_ frequency channel (or subband) in your data. I have confirmed that the DM variations do in fact go away when the data is processed correctly. Hopefully this mistake on my part did not cause too many people to waste significant amounts of time trying to explain the "effect"! Sincerely, Scott Ransom