Ashley Villar astronomer

Detecting GW170817 with Spitzer

We recently submitted a new paper to Astrophysical Letters which presents Spitzer Space Telescope observations of GW170817 at late times.

We downloaded publicly available Spitzer infrared data taken 43 and 74 days after the merger at 3.6 and 4.5 microns. We reduced this data using standard methods, but we found that it was incredibly difficult to remove weird galactic artifacts. This is largely due to the Spitzer’s point spread function (PSF). Spitzer has a triangular PSF, and unfortunately the spacecraft was oriented in different directions between the different observations. This made image subtraction difficult for extended objects within the image.

Luckily, we know the position of GW170817 very well! So we were able to smooth over our subtracted image while masking out the position of GW170817. We then subtracted the smoothed version from the original version, revealing a source at the masked region. This procedure isn’t standard, so we did a lot of statistical testing to make sure we weren’t “detecting” noise or some weird pattern from the galaxy.

Spitzer images
Upper Left: Composite Image. Upper Right: 4.5 micron image after standard template subtraction. Bottom Row: Final images after subtracting a smooth galaxy.

In the end, we were confident of our results, shown in the light curve below. We found that the kilonova is dimmer than our model (originally presented here) predicts. This wasn’t super surprising, because we already knew that the late K-band data was also being over-predicted by our model.

Spitzer light curve

The discrepancy can be due to a number of factors, but we highlighted two. First we find that the optical depth is approximately one at 40 days post-merger, meaning that the ejecta is becoming optically thin. Our model assumes that the ejecta is optically thick so that the material can be thermalized. If the kilonova enters the nebular phase, our model assumptions break down. Second, it’s possible that the thermalization efficiency we assume is incorrect. The thermalization efficiency is a function of mass, velocity and time, and the uncertainty of this function is almost an order of magnitude at late times!

Finally, we discuss what GW170817 tells us about future missions. In (hopefully) early 2019, LIGO will turn on again for its third observing run. We expect to have improved sensitivity out to 120 Mpc for NS binary mergers. Based on GW170817, we can certainly observe future kilo novae out to similar distance with deeper observations. Looking further in the future, JWST and MIRI will be able to collect spectra and photometry of objects out to similar distances.