CMB-S4 will regularly survey a large fraction of the sky with high sensitivity. This pattern of repeated observations of a single (or few) patches opens the possibility of doing a high-quality time-domain survey of emissive objects. The millimeter band, for reasons of field of view and observing priorities on non-CMB-focused instruments, has had almost no such survey conducted (SPTpol has done one; there have been no others), with the result that we know extremely little about the time-domain sky in these wavelengths (and it is a broad desert -- there are essentially no surveys between optical and 1 GHz radio, and the 1 GHz surveys are currently not that sensitive). At the same time, our observing band is one of significant scientific interest for understanding a wide variety of physics: Where mm observations have been used for follow-up of extragalactic transients or variable sources, they have greatly increased our understanding, in particular of GRBs.
As a general rule, sensitivity to these sources is optimized by good angular resolution (point-source sensitivity scales as the square of primary mirror diameter), high-frequency coverage (to reduce the diffraction limit of beams), and a steady re-observation cadence of a single sky patch. The slopes of the dN/dS distributions for the target sources are such that trading depth for sky area is nearly a wash, with a slight preference for area over depth. The survey requirements are thus quite similar to those for Neff. The key timescale for most of these objects is days to weeks, so we need to reobserve every part of the field every day for maximum sensitivity.
There is almost no point in doing single-object follow-up with any CMB instrument as other telescopes (ALMA, ATCA, etc.) have much better point-source sensitivity. Our strength, much like for CMB, is statistics and wide-area surveys.
As a further general consideration, observing time reduces noise as 1/sqrt(t) for periods less than the expected duration of the source and looks like sky area for periods longer than the duration of the source.
Gamma-ray bursts are the most promising target for CMB-S4. GRB afterglows have a self-absorbed synchrotron spectrum, with peaks at around 200 GHz (the center of a basically flat region from ~ 50 GHz to 400 GHz) a day or two after the initial explosion and gamma-ray emission. This gives us uniquely good sensitivity to so-called orphan afterglows, in which the afterglow is seen but not the primary gamma-ray emission, which are a hard prediction of our current understanding of GRBs. No one has seen such an object despite over 20 years of effort (there's some twiddle here), which is a situation we are in a position to change. Finding them is a major science goal of next-generation radio surveys like SKA and, sooner, ASKAP/VAST. CMB-S4 will have more than an order of magnitude better sensitivity than SKA and should see more than 1000 afterglows.
Orphan afterglows may be seen for one of four reasons:
- For whatever reason, the gamma-ray satellites were not looking in the right place at the right time to see the emission. This is a calculable scaling factor (they monitor about 2/3 of the sky at any time).
- The mm emission is less tightly beamed than the gamma-ray emission, so we are sensitive to more off-axis bursts. The ratio tells you about the beaming angles and evolution of the sources.
- The gamma-ray emission was suppressed by attenuation on gas. Such bursts are a possible source of the diffuse TeV+ neutrino background.
- The gamma-ray emission came from very high red shift (z > 10) and was too time-dilated and red-shifted to see.
There are a few overarching science goals:
- Demonstrate that the rate of off-axis GRBs matches our understanding of jet angles from jet break measurements.
- Constrain any excess population of "hidden" GRBs or GRB-like objects with large energy budgets but invisible in gamma-rays.
- Search for ultra-high-redshift (z > 20) GRBs from Pop-III stars. These are very hard to trigger on with gamma-ray satellites (a purpose-built mission, THESEUS, has been proposed to look for them), but we should also have good sensitivity here. There is very limited theoretical work on expected rates, but the one paper I have been able to extract a number from suggested a few/year.
S4 should see hundreds to over one thousand of the first (guaranteed) class of object (Ghirlanda et al. 2014) in an f_sky=0.4, Ndet=400000 configuration. That number scales weakly with most survey choices except concentration of sensitivity at low frequencies (< 90 GHz), which will hurt sensitivity. It is not fully clear what the theoretical uncertainties are on that prediction.
- Daily map noise < ~ 5 mJy/sqrt(day). In principle, there may be very rare, very bright afterglows, but none have been seen with fluxes > a few tens of mJy and we need to be sensitive to those.
- Field reobservation daily or better. The shortest expected source length is ~ 3 days and we need multiple data points.
A very large fraction (~half) of the steady point sources seen by CMB-S4 will be AGN, with several thousand sources detected at high S/N. These can have variability of factors of several in flux and polarization variation on timescales of days to months. Aside from one small project at OVRO, there is no consistent long-term monitoring of these sources at wavelengths longer than X-rays. Our band is also convenient for peering into AGN central engines.
There seem to be three main ways we can contribute to studies of AGN variability:
- Because there are very, very few unbiased surveys of AGN variability at long wavelengths, the statistical interpretation of follow-up in radio/millimeter is at best unclear. The value of data from ALMA/ATCA et al. following up high-energy triggers would be quite substantially increased by an understanding of the PDFs. For example, the IceCube measurement of neutrino emission from TXS 0506+056 had long-wavelength data, but no interpretation was given because the realm of normal behavior was unknown.
- Any high-energy-budget process occurring in the AGN that does not emit gamma-rays would probably escape the notice of the entire community at present. Having even a single long-wavelength survey would substantially improve the ability to trigger on -- or constrain -- such things. To use the IceCube TXS example again, there was apparently some emission (one neutrino) during a 2017 gamma-ray flare observed by Fermi-LAT. But, in December 2014, the neutrino luminosity of the source seems to have been at least an order of magnitude larger and there is no evidence of any gamma-ray emission at that time. And no one has any idea if there was any other emission, because no one was looking.
- Flares evolve from short to long-wavelengths. The cross-correlation of the Fermi-LAT and long-wavelength brightnesses with time provides a high-quality probe of the dynamics of the evolution of these flares, the presence of shock acceleration, etc. which is at present poorly constrained and limited to a few (~10) marquee objects (this is the goal of the OVRO project "Blazars in the Fermi era").
We need the following:
- Revisit times of < 1 week (flares can be as short as days, but are usually more like weeks to months)
- The number of sources we can see goes as (map noise)^-1.5, but there are many very bright sources, so there does not seem to be a cliff here.
Last year, the source AT2018cow was found with incredibly high millimeter emission. It remains unclear what it is and it was found by chance in an optical survey. CMB-S4, in any configuration, would have see this object at very high signal-to-noise (daily S/N > 10 for 3 weeks) without a trigger anywhere in the survey area.