Why Bigger Telescopes Alone Cannot Perfect Pulsar Timing

From stellar rotation to measured arrival

Pulsar timing compares measured pulse arrival times with predictions accounting for stellar rotation, orbital motion and observing geometry. The differences, called timing residuals, contain both signals of interest and unwanted fluctuations. Pulsar timing arrays combine measurements from multiple millisecond pulsars to search for extremely low-frequency gravitational waves.

The preprint follows the measurement chain from the neutron star and its radio-emitting magnetosphere, through intervening plasma, to telescope electronics and data processing. It distinguishes uncertainty in measuring an arrival time from genuine changes in when a pulse arrives. A precisely measured pulse can still be a poor clock reading if the star’s rotation fluctuates or propagation delays change.

An inventory, not a new gravitational-wave detection

Cordes and colleagues combine established research with analytical derivations, numerical simulations and analyses using existing observations, extensively drawing on NANOGrav datasets. Their focus is understanding and reducing physical timing errors, rather than developing a complete statistical inference pipeline or announcing another gravitational-wave detection.

They classify effects by their dependence on radio frequency and their variability over time. Chromatic effects change with observing frequency; achromatic effects do not. Receiver noise, pulse-to-pulse variability and rapidly changing scattering affect short observations. Spin fluctuations and slower plasma changes influence measurements across days, months or longer.

The work also extends analytical treatment of frequency-dependent dispersion measures to broad observing bands. Dispersion measure describes the electron column encountered by radio waves. Scattering makes different frequencies sample different bundles of paths, complicating the assumption that one electron-column value describes an entire observation.

Sensitivity encounters astrophysical limits

Larger telescopes reduce the contribution from receiver and background noise. They do not automatically suppress pulse jitter: random changes in individual pulse shapes and phases. Averaging more pulses generally reduces jitter, making longer observations useful even after additional collecting area offers diminishing returns.

Interstellar scintillation, the changing interference pattern produced by scattered radio waves, adds another limitation. A finite observation samples only a finite number of bright and faint interference features, called scintles. Their changing contributions perturb the measured pulse shape and arrival time.

The authors argue that propagation effects will become increasingly important as instrumental sensitivity improves. Observing at higher radio frequencies usually reduces plasma effects, but pulsars generally become fainter there. The best observing band therefore depends on the pulsar, its intervening plasma and the telescope.

Why scattering corrections can introduce bias

A central finding concerns the pulse broadening function, which describes how scattering spreads an initially sharp signal over time. A commonly used model assumes a simple exponential tail. The paper shows that broader distributions of plasma structure can instead produce extended, heavy tails whose shapes also change with frequency.

The characteristic scattering time is consequently not interchangeable with the timing correction. The measured offset depends on the intrinsic pulse shape, the actual broadening function and the template used to estimate arrival time.

In one numerical example, a Gaussian pulse with a width of 50 microseconds was scattered using a heavy-tailed function with a characteristic time of 5 microseconds. An unscattered Gaussian template produced an arrival-time error of positive 6.53 microseconds. A fitted Gaussian-plus-exponential template produced negative 2.16 microseconds. These are simulation results under specified assumptions, not measured errors shared by all pulsars.

Incorrect propagation models can also leave contamination in arrival times intended to be frequency independent. Distinct events, including plasma-lensing episodes or sudden intrinsic profile changes, can produce departures from the bell-shaped statistics assumed in Gaussian models.

Choosing pulsars and correcting observations

The authors develop quality measures for receiver-noise errors, jitter, scintillation, spin stability and several propagation effects, then compare 63 millisecond pulsars in NANOGrav’s timing program. Brightness alone does not determine usefulness: B1937+21 performs exceptionally well for basic pulse measurement but suffers strong long-term, frequency-independent noise.

The combined rankings use illustrative, explicitly non-optimal weights. Changing the priorities changes the rankings; they are decision aids, not a universal league table.

The main recommendation is to estimate dispersion and scattering corrections using simultaneous broad-frequency data from each observing epoch. Scattering corrections should be customized to individual pulsars, supported where possible by measurements of intensity versus time and frequency.

Limits and unresolved questions

The forecasts inherit substantial scatter in empirical spin-noise and scattering relations. Intrinsic pulse shapes and the spatial distribution of scattering material are imperfectly known. The paper’s assignments of dominant long-term noise sources to individual pulsars are partly subjective.

Cyclic spectroscopy can resolve fine scintillation structure and potentially recover propagation information, but the authors note that coherent scattering removal has not yet delivered better arrival-time precision than standard methods. Their proposed improvements therefore require further observational demonstration. The manuscript is a preprint submitted to Living Reviews in Relativity; submission does not establish completed peer review.