Radio/Gamma-ray time lags: Large samples and statistics Walter Max-Moerbeck On behalf of the OVRO 40m blazar monitoring team AGN Monitoring Workshop MPIfR, Bonn, Germany March 14, 2011
Feb 22, 2016
Radio/Gamma-ray time lags:
Large samples and statistics
Walter Max-MoerbeckOn behalf of the OVRO 40m blazar monitoring team
AGN Monitoring WorkshopMPIfR, Bonn, Germany
March 14, 2011
Correlated radio/gamma-ray variability
The hypothesis of correlated variability in radio and gamma-ray is popular It would indicate a common spatial origin for
radio and gamma-ray emission But it needs to be proven!
Correlated radio/gamma-ray variability
Our approach: Large sample of objects Preselected as gamma-ray candidates Observed independently of gamma-ray state High cadence, observed twice per week Statistical tests for correlations
A first look at the radio/gamma-ray cross-correlation
Data Radio data published in Richards et al 2011 (ApJ
submitted) 2 year light curves for CGRaBS sources + a few
calibrators Gamma-ray data published in blazar variability
paper, Abdo et al. 2010 ApJ, 722, 520 106 sources 11-month light curves, weekly sampling
52/106 are in the CGRaBS sample
Radio lags Radio precedes
• Example cross-correlations. 3-month Fermi detections, using 11-months of Fermi data and 2 years of radio monitoring
β_radio = 2.5, β_gamma =
2.0• Significance evaluated using simulated data with a power-law PSD ~ 1/f^β
Radio/gamma-ray time lags and their significance
Radio lags Radio precedes
• Example cross-correlations. 3-month Fermi detections, using 11-months of Fermi data and 2 years of radio monitoring
β_radio = 2.5, β_gamma =
2.0• Significance evaluated using simulated data with a power-law PSD ~ 1/f^β
Radio/gamma-ray time lags and their significance
Statistical test for the cross-correlation:
Measuring the PSD
The significance level depends on the model used for the light curves
It is commonly assumed that it is red-noise with a simple power-law PSD
Uneven sampling complicates the model fitting We use the method of Uttley et al 2002 MNRAS 332, 231 With some modifications
Basic idea is to simulate data with a given PSD and process it as the data. The mean PSDs and deviations are used for model fitting
βradio = 2.5, βgamma-ray= 2.0 βradio = 2.0, βgamma-ray= 1.5
βradio = 0.0, βgamma-ray= 0.0
Significance versus PSD power-law exponent
Significance for longer time series
1 year of gamma-ray and 2 years of radio – dotted lines
5 years of gamma-ray and 6 years of radio – solid lines
Statistical test for the cross-correlation:
Measuring the PSD
J0017-0512
J0238+1636
Example light curves Goodness of fit –radio data
Some PSDs are hard to constrain, we need longer time series
A large fraction have well constrained PSDs slopes
β
β
n>/N
n>/N
PSD measurements first results
The distribution of PSD power-law indices is different for gamma-ray detected/non-detected sources This is consistent with
gamma-ray quiet objects looking like white noise, without flares
A peak near beta~2.0 can be used when measuring significance
Gamma-ray detected
Gamma-ray non detected
Cross-correlation the next step
Include all sources on 1LAC (Fermi first year catalog) with 2 years of data in gamma-ray and at least 2 years in radio, more for CGRaBS
Main problem is to extract all the gamma-ray light curves and deal with upper limits, sparse or adaptive sampling ~400 sources in our program
221 CGRaBS
Summary
Paper in preparation using published Fermi and OVRO data PSD is characterized for all radio sources Cross-correlation significance will incorporate this
new constraints on the variability behavior of blazars Will submit before Fermi Symposium
Next step is to extend this to a larger set of gamma-ray sources and longer light curves at both bands