Science with pulsar-timing arrays

Transcription

1 Science with pulsar-timing arrays Michele Vallisneri Jet Propulsion Laboratory California Institute of Technology

2 Summary Fitting that pulsars, after indirectly confirming the presence of GWs by loss GW detection with PTAs offers a very beautiful, of energy, yet should extremely offer a way to measure them directly. difficult challenge: building a detector the size of our galaxy, exploiting nature s most precise clocks, millisecond pulsars. Barring surprises (cosmic strings, nonstandard relic radiation, GW memory from early-universe events), PTAs will observe first the stochastic background from the cosmological population of supermassive black-hole binaries in Galactic nuclei. Improvements in sensitivity are limited by the increasing span of datasets and by the continued discovery of new pulsars. The most recent upper limits on the background are in tension with theoretical expectations, suggesting last-parsec physics, or faulty assumptions. See Kelley s talk! Nevertheless, if theoretical models are correct, detection is expected within 10 years. Establishing confident detection requires sophisticated statistical techniques and superior control of systematics. Unfortunately, recent hints of a signal seem to be subsiding.

3 The general theory of gravitational-wave detectors From formula: baseline helps. We only need two good clocks; we can make good clocks on Earth if only we could get a good one across the Galaxy! no gw L 12 = L Z 2 1 h( )d pulsar timing Doppler tracking, elisa, LIGO

4 [Joeri van Leeuwen] magnetars normal pulsars Double Pulsar double NSs msec binaries pulsars: Nature s precision clocks [Manchester 2015]

5 Pulsar-timing multiphysics [Stairs 2003, Manchester 2013, Manchester 2015, You et al. 2007, Weisberg et al. 2010]

6 J ASP/GASP ASP/GUPPI PUPPI/GUPPI Averaged Residual [µs] Whitened Residual [µs] DMX [10-3 pc cm -3 ] J Date [yr] B GASP GUPPI Averaged Residual [µs] Whitened Residual [µs] DMX [10-3 pc cm -3 ] Date [yr] [NANOGrav soon]

7 Pulsar-timing arrays [Foster and Backer 1990] [Nice 2016, NANOGrav soon] Date [yr] AO/1400 J AO/1400 J GBT/1400 J GBT/1400 J GBT/1400 J GBT/1400 J GBT/1400 J GBT/1400 J GBT/1400 J GBT/1400 J GBT/1400 J AO/1400 J GBT/1400 J GBT/1400 J GBT/1400 J AO/1400 J GBT/1400 J AO/1400 GBT/1400 AO/2100 J AO/1400 AO/2100 J AO/1400 J AO/2100 GBT/1400 J GBT/1400 J GBT/1400 J AO/1400 J AO/1400 B AO/1400 AO/2100 J GBT/1400 J AO/1400 AO/2100 J AO/1400 J GBT/1400 J AO/1400 J AO/1400 GBT/1400 AO/2100 B AO/1400 J AO/1400 B GBT/1400 J AO/1400 J AO/2100 AO/1400 J AO/1400 J GBT/1400 J AO/1400 AO/2100 J AO/1400 J AO/1400 J AO/1400 J AO/2100 GBT/1400 J AO/327 J AO/1400

8

9 Pulsar science: individual SMBH binaries [Graham et al. 2015] [Babak et al. 2016]

10 Pulsar science: bursts with memory [NANOGrav 2015] SMBH mergers:

11 Pulsar science: cosmic strings recombination vs string tension string tension vs loop size [NANOGrav 2016]

12 Pulsar science: relic radiation log tensor/scalar [Lasky et al. 2016] inflationary spectral index

13 Pulsar science: stochastic background from SMBH mergers [Phinney 2001, Sesana et al. 2008]

14 Stochastic background from SMBH mergers [Sesana et al. 2012, Ravi et al. 2014, Burke-Spolaor 2015]

15 Isotropic SMBH background: NANOGrav 9-year analysis [NANOGrav 2016]

16 Isotropic SMBH background: NANOGrav 9-year analysis [NANOGrav 2016, Sampson et al. 2015]

17 Detection probability given the PPTA limit [Taylor, Vallisneri, et al. 2015]

18 Time-to-detection scaling laws [Vigeland and Siemens 2016]

19 How to detect the SMBH background The detection of a GW background with PTAs relies on discerning a red-noise process correlated across the array with the tell-tale quadrupolar Hellings Down pattern. Each pulsar displays a host of noise processes, including GW-like red noise, which must be modeled probabilistically. Modern detection pipelines rely on the comparison of Bayesian evidence between a model that includes correlated GWs, and a null model that does not. Although time slides à la LIGO are not possible, evidence thresholds can be set using sky scrambles and red-noise phase shifts [Cornish and Sampson 2016, Taylor et al. 2017] For current datasets, the best few pulsars dominate all statistics, creating the danger that systematics simulate a red-noise process with correlation length comparable to the span of the data. Detection confidence will be helped by simulations, crossvalidation, more good pulsars, and increasing significance.

20 A PTA noise budget arxiv.org/

21 A PTA noise budget for J [NANOGrav 2016, internal]

22 A PTA noise model: everything is a Gaussian process timing residuals = radiometer noise (white) + timing-model errors van Haasteren & MV PRD 90, (2014) + jitter noise (white, epoch) + DM + timing noise (red) + GWs Basis picture Search over basis coefficients and hyperparameters y gp = Fa p(a) e at ( ) 1 a/2 Kernel picture Marginalize over basis coefficients, search over hyperparameters p(y gp ) e y T gpk( ) 1 y gp /2 K( )=F ( )F T

23 Stochastic GWs as correlated Gaussian process pulsar #1 pulsar #2 pulsar #3 pulsar #4 [Burke-Spolaor 2015] [Jenet et al. 2015]

24 GWB amplitude posteriors [NANOGrav 2017, PRELIMINARY]

25 Ephemeris systematics [NANOGrav 2017, PRELIMINARY] DE421 (2008): targets Mars DE430 (2014): ICRF 2.0, Moon++ DE435 (2016): targets Cassini DE436 (2016): targets Juno

26 J noise model [NANOGrav 2017, PRELIMINARY] green: DE430 blue: DE435

27 GWB amplitude posteriors [NANOGrav 2017, PRELIMINARY]