SCULPTing a new pulsar timing landscape

Abstract:

Pulsar timing arrays are only as good as the timing residuals they produce — and those residuals are only as clean as the method used to extract them. Traditional template-matching approaches measure pulse time-of-arrival (TOA) by fitting a single fixed or radio-frequency dependent template to observed profiles, but this leaves a critical vulnerability: any stochastic variation in pulse shape is silently absorbed into the TOA as a systematic bias, quietly corrupting the signals we are trying to extract.

SCULPT is a new technique built of old ideas that fixes this at its source. By constructing a basis directly from observed profiles via SVD, SCULPT learns the true structure of pulse shape variability from the data itself — no analytic assumptions, no oracle knowledge required. Stochastic shape changes are absorbed into nuisance amplitude coefficients within this basis, keeping them out of the timing solution. The result is a variational profile basis that is optimally adapted to the actual pulse morphology: expressive where it needs to be, simple everywhere else.

I will present the methodology behind this process and explore with the audience what it is we’re trying to achieve and why it’s an important goal.