Atacama Cosmology Telescope: Observations of supermassive black hole binary candidates. Strong sinusoidal variations at 95, 147 and 225 GHz in PKS 2131$-$021 and PKS J0805$-$0111

TL;DR

This paper uses Atacama Cosmology Telescope mm-wave data (95, 147, 225 GHz) to study strong sinusoidal variations in two blazars (PKS 2131–021 and PKS J0805–0111) previously identified as SMBHB candidates. By fitting synchronized sine waves across multiple bands and comparing with OVRO and ALMA data, the authors demonstrate frequency-dependent, monotonic phase shifts and achromatic sine amplitudes, consistent with a Doppler-beamed jet modulated by binary orbital motion. To explain the observations, they propose a Modified Kinetic Orbital (MKO) model in which a slower wind confines a relativistic jet, generating a jet-wind helix whose phase varies with emission radius; intermittency is attributed to disc-jet disturbances, supported by recent simulations. The results strengthen the SMBHB interpretation for these sources and underscore the potential of wide-field mm surveys (e.g., Simons Observatory) to uncover thousands of SMBHB candidates, with important implications for nanohertz gravitational-wave background studies and multi-messenger astrophysics.

Abstract

Large sinusoidal variations in the radio light curves of the blazars PKS J0805$-$0111 and PKS 2131$-$021 have recently been discovered with an 18-year monitoring programme at the Owens Valley Radio Observatory, making these systems strong supermassive black hole binary (SMBHB) candidates. The sinusoidal variations in PKS 2131$-$021 dominate its light curves from 2.7 GHz to optical frequencies. We report sinusoidal variations observed in both objects with the Atacama Cosmology Telescope (ACT) at 95, 147 and 225 GHz consistent with the radio light curves. The ACT 95 GHz light curve of PKS 2131$-$021 agrees well with the contemporaneous 91.5 GHz ALMA light curve and is comparable in quality, while the ACT light curves of PKS J0805$-$0111, for which there are no ALMA or other millimetre light curves, show that PKS 2131$-$021 is not an isolated case, and that this class of AGN exhibits the following properties: (a) the sinusoidal pattern dominates over a broad range of frequencies; (b) the amplitude of the sine wave compared to its mean value is monochromatic (i.e., nearly constant across frequencies); (c) the phase of the sinusoid phase changes monotonically as a function of frequency; (d) the sinusoidal variations are intermittent. We describe a physical model for SMBHB systems, the modified Kinetic Orbital model, that explains all four of these phenomena. Monitoring of ${\sim}8000$ blazars by the Simons Observatory over the next decade should provide a large number of SMBHB candidates that will shed light on the nature of the nanohertz gravitational-wave background.

Figures (11)

• Figure 1: PKS 2131$-$021 light curves from OVRO, ACT and ALMA. Heavy points are binned into 50-day intervals (100 for OVRO) to guide the eye, with individual measurements shown as lighter points. Sinusoidal fits, shown with the continuous curves, were done on the unbinned data. The sine fits for ALMA are from K25 and sine fits to ACT are done in this paper (see text). For the OVRO data, the solid curve is the best-fit for the ACT time range, while the dashed sine curve corresponds to the ALMA time range. A monotonic phase shift of the sine waves to earlier times with increasing frequency is visible by eye, and is reported relative to the OVRO phase as $\Delta\phi_0$ in Tables \ref{['tab:2131_fits']} and \ref{['tab:alma_2131_fits']}.
• Figure 2: Correlation of medium-term flux density variations in ALMA and ACT light curves of PKS 2131$-$021. Top: ALMA 91.5 GHz and ACT 95 GHz (PA5) light curves, which have been filtered by binning and subtracting a long timescale trend. The binning is in five day chunks, and only time chunks containing data from both ALMA and ACT are retained. The long timescale trend is removed by fitting a third degree B-spline with 120-day knot spacing to each binned light curve and subtracting; this effectively acts as a high pass filter of variations on timescales $\gtrsim 120$ days, removing the large sinusoidal pattern. Clear correlations between the light curves are seen on these time scales. Bottom: The Pearson correlation between the filtered ALMA and ACT light curves, for different spline knot spacings. Errors are estimated by calculating the correlations between the ALMA PKS 2131$-$021 light curve and each of the other 204 ACT light curves in our bright AGN library, which should be uncorrelated, and taking the standard deviation. The thick, coloured point at 120 days corresponds to the data shown in the upper panel.
• Figure 3: PKS J0805$-$0111 light curves at 15 GHz, 95 GHz, 147 GHz, and 225 GHz. The main panel shows the time range covered by the ACT observations while the inset shows the full range observed by OVRO. Heavy points are binned into 25-day intervals (100 days in the inset) to guide the eye, with individual measurements shown as lighter points; the sinusoidal, least-squares fits were done on the unbinned data. A monotonic phase shift of the sine waves to earlier times with increasing frequency is visible by eye, and is reported relative to the OVRO phase as $\Delta\phi_0$ in Table \ref{['tab:0805_fits']}. The dashed vertical line at MJD 59041 (2020 July 11) indicates the approximate date at which D25 found that the sinusoidal variation in the 15 GHz ceases. The ACT mm data also apparently flatten around the same time.
• Figure 4: PKS 2131$-$021 phase shifts measured by ACT at 95 GHz, 147 GHz, and 225 GHz (black points), compared to the phase shift results of K25 (blue points); see Table \ref{['tab:2131_phases']}. Note that the ACT 95 GHz and ALMA 91.5 GHz points overlap. The Haystack or OVRO 15 GHz light curves provided the phase reference in all cases. Uncertainties are determined by the MCMC sine-wave fits (Eq. \ref{['eq:mcmc_fit']}). The curved, blue line shows the quadratic fit determined by K25 (Eq. \ref{['eq:2131_quadratic_fit']}). The orange line is the best fit power law to the Haystack, ALMA and ACT data (Eq. \ref{['eq:powlaw']}, Table \ref{['tab:phase_comparison']}).
• Figure 5: Posterior probabilities of the power law MC fit to the sinusoid phase shifts (Eq. \ref{['eq:powlaw']}) for different data combinations. The contours show the 95% credible regions and the dots are the maximum likelihood values (c.f., Table \ref{['tab:phase_comparison']}). Flat priors on $a$ and $b$ in the ranges of $(0.05, 4)$ and ($-2, -0.05$) were used.