Modeling Solar System Ephemeris Uncertainties & Their Impact On GW Constraints

Presentation given at European Pulsar Timing Array meeting in Amsterdam.

Transcript

1. Stephen Taylor EPTA Meeting, Amsterdam, 04-04-2017 © 2017 California Institute of Technology. Government sponsorship acknowledged Stephen R. Taylor (on behalf of NANOGrav Detection Working Group) Modeling Solar System Ephemeris Uncertainties & Impact On GW constraints JET PROPULSION LABORATORY, CALIFORNIA INSTITUTE OF TECHNOLOGY
2. Stephen Taylor EPTA Meeting, Amsterdam, 04-04-2017 Overview The Solar System ephemeris. JPL releases, and updates between versions. Modeling solar-system ephemeris uncertainties within our Bayesian pipelines. Impact on GW constraints, including preliminary NANOGrav 11yr results. Outlook and potential avenues for future study.
3. Stephen Taylor EPTA Meeting, Amsterdam, 04-04-2017 The Solar System Ephemeris All TOAs are referenced to the solar-system barycentre. Computing the location of the barycentre requires the masses and trajectories of all important dynamical objects in solar-system. JPL do not really care about the barycentre. They care about navigating probes to planets. The ephemeris time-series is not fit for in Tempo2. It is subtracted.
4. Stephen Taylor EPTA Meeting, Amsterdam, 04-04-2017 JPL releases and updates DE435 Created in Jan 2016 for the Cassini project, this is primarily an incremental improvement in the orbit of Saturn. Jupiter had already been updated in DE434 for the Juno mission. DE436 is a small improvement on DE435.
5. Stephen Taylor EPTA Meeting, Amsterdam, 04-04-2017 Quadratic and 1-yr sinusoid removed BIG DIFFERENCES! — Need to understand how this impacts GW searches Figure by J. Ellis JPL releases and updates
6. Stephen Taylor EPTA Meeting, Amsterdam, 04-04-2017 Modeling ephemeris uncertainties Ultimate goal is to introduce an ephemeris uncertainty term in our full model. This extra term is a perturbation to the ephemeris. We should be able to marginalize over this perturbation, regardless of which JPL version we started from. We want to “connect” the results of all the ephemerides. By marginalizing over ephemeris uncertainties, we want to isolate our GW constraints from the particular choice of ephemeris.
7. Stephen Taylor EPTA Meeting, Amsterdam, 04-04-2017 Ephemeris uncertainties as a Gaussian process Lentati, Taylor, Mingarelli et al. (2015), MNRAS, 453, 2576, arXiv:1504.03692 Just like red-noise or the GWB, we marginalize over “a” with a Gaussian prior. We parametrize the power in the error vector along each direction. Power spectrum model can be power-law, free-spectral, etc.
8. Stephen Taylor EPTA Meeting, Amsterdam, 04-04-2017 Simulate a PTA dataset. Inject an ephemeris uncertainty at Jupiter periodicity. Nothing else in dataset. Our model correctly isolates the ephemeris uncertainty, with no leakage into a common red process. Figure by J. Ellis
9. Stephen Taylor EPTA Meeting, Amsterdam, 04-04-2017 14.1 13.9 13.7 13.5 13.3 log10 Agwb 3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2 gwb 14.2 14.0 13.8 13.6 13.4 13.2 13.0 log10 Aclk 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 clk 9.0 8.8 8.6 8.4 8.2 log10 (| Mjup. |/M ) 0.0 0.2 0.4 0.6 0.8 1.0 sgn( Mjup. ) Taylor et al. (2017a), PRD, 95, 042002, arXiv:1606.09180 Deterministic planet-mass perturbations Model is a Roemer delay perturbation, as in Champion et al. (2010).
10. Stephen Taylor EPTA Meeting, Amsterdam, 04-04-2017 Modeling uncertainties in the true ephemerides Simulate 36 pulsars (IPTA MDC), with 11.4 yr baseline (same as NANOGrav 11yr dataset). 500 ns precision. No other processes in data. Create dataset with DE436. Analyze with others. Power-law red process for {x,y,z} ephemeris time-series
11. Stephen Taylor EPTA Meeting, Amsterdam, 04-04-2017 Perturbing the planet masses in Roemer delay correction. Modeling uncertainties in the true ephemerides
12. Stephen Taylor EPTA Meeting, Amsterdam, 04-04-2017 Preliminary NANOGrav 11yr Results 18 17 16 15 log10 AGWB 10 3 10 2 10 1 100 Deterministic object-mass perturbation model 9 objects (Mercury to Pluto) DE421 DE430 DE435 DE436 18 17 16 15 log10 AGWB 10 3 10 2 10 1 100 Power-law ephemeris model 30 linear-spaced frequencies (1/T to 30/T) DE421 DE430 DE435 DE436 Bayes factor for a common red process (i.e. leaving out H&D correlations) versus noise range from ~1 (DE435) to ~10 (DE430). It is crucial to marginalize over the difference in the ephemeris uncertainties for robust GW statistics. PRELIMINARY
13. Stephen Taylor EPTA Meeting, Amsterdam, 04-04-2017 Summary & Outlook PTAs are now at the sensitivity where the choice of ephemeris makes a huge difference! (This is good and bad!) The ephemeris effects upper limits, and will effect how we quote detection statistics. Our models perturb the Roemer delay to attempt to account for offsets in the ephemerides. These models are not yet good enough to “connect” the results of the different JPL ephemerides (DE421, DE430, DE435, …). Options being developed: As well as perturbing the planet mass, we can perturb the trajectory with a mixture model of the JPL ephemerides. Use dense low-frequency spectral coefficients to mitigate potential ephemeris spectral leakage from low frequencies. We are working with the ephemeris team as JPL to construct a search basis out of “perturbed" ephemerides, where they offset the initial conditions and masses used in their codes.