My Summer as a NASA Intern

• Hello! My name is Corin Marasco, I’m a physics major at Georgia Tech, and this summer I worked with Dr. Kristin Madsen in the X-ray astrophysics lab at NASA Goddard on cross-calibrations of x-ray satellites using the quasar 3C 273!

• Here are the four satellites we cross-calibrated. They all observed the quasar 3C 273 from 2015-2021, and we have analyzed the data from all seven years for a more accurate cross-calibration analysis.
• The satellites we used are NuSTAR, Swift, Chandra, and XMM-Newton.
• NuSTAR observes in by far the hardest X-ray band out of these four observatories, from 3-79 keV. However, unlike the other three, it does not operate in the soft X-ray band below 3 keV. It has two theoretically identical instruments that it uses for observations, FPMA and FPMB, which were analyzed separately in this study.
• Swift is a multiwavelength observatory which houses an X-ray telescope called XRT that operates in about the 0.2-10 keV band. Swift XRT collects data through two modes, photon counting and windowed timing. For our analysis, we used the windowed timing data because it was available for every observation.
• Chandra observes in the 0.08-10 keV band using two grating spectrometers: LETGS and HETGS. They observe in different X-ray bands for better overall resolution. In order to match the bands from other observatories, just the data from HETGS was used.
• Our last satellite, XMM-Newton, observes in the 0.15-15 keV band. It houses three instruments that it uses for X-ray observations, MOS1 and MOS2, which are theoretically identical, and pn, which uses different gratings for collection. These instruments were analyzed separately in this study.
• Lastly, these satellites have different orbits, influencing their collection data. NuSTAR and Swift have a low Earth orbit, while Chandra and XMM have highly elliptical orbits.

• Good time intervals or GTIs are periods of time when the observatory was collecting usable data.
• To compare the observational data between observatories, we first found good time intervals where the observations overlapped.
• Like I mentioned in the last slide, NuSTAR and Swift have low earth orbits, which causes them to have these short GTI spans as the Earth gets in the way of the target. Chandra and XMM are uninterrupted because their orbit allows them to swing out far from the Earth and collect data for longer periods of time.
• For better accuracy, the satellites were compared and analyzed as pairs instead of limiting our study to the short periods of time where all four GTIs overlap.
• The lightcurve on the right shows the counts per second collected by each satellite in 2018.
• As you can see in the lightcurves, there is a lot of data that occurs outside the short spans where all four observations overlap, and some of that might be worth including, like the dip in the blue Swift data.

• This table details the GTIs shown in the plot on the left.
• Because we analyzed the satellites in pairs, we identified six GTI intervals for every year. The six for 2018 are listed here.
• You can see how much more of the exposure time we get to use when we compare the satellites in pairs. Instead of using just 14,000 seconds of NuSTAR data because it’s limited by Chandra, we can use 35,000 seconds when it’s compared to Swift and 30,000 when compared with XMM-Newton.
• So comparing the satellites in pairs allows us to use more of the collected data and hopefully obtain a more accurate cross-calibration.

• Before we look at the collected data, let's talk about the source we used.
• 3C 273 is a Seyfert 1 quasar dominated by its coronal emission.
• A quasar is a very bright active galactic nuclei, and, similar to the sun, it has a corona and its corona is the primary source of its x-ray emission.
• 3C 273 the closest high-luminosity quasar, and was the first quasar ever detected.
• It’s so bright that if it was as far as 10 parsecs away, it would be as bright as the sun!
• Its spectrum is dominated by a power-law model in the X-ray band that we examined.
• We use 3C 273 for cross-calibration because it is very bright across the X-ray band, it has been heavily studied, and it’s spectrum is very featureless, so it’s easier to fit a model to.

• For each of the six satellite pairs, we fit a power-law model to the data.
• Because of NuSTARs harder energy band, the three pairings that involved NuSTAR were analyzed in the 3-7 keV band, and the other three pairings without NuSTAR were analyzed in the 1-5 keV band.
• We used a power-law model because, as shown in the ratio plot in the figure on the left, the power-law fit describes 3C 273 very well. Because we only analyzed data in the 1-10 keV range, the power law accurately fits the energy range our study focuses on, even though there’s a dip in the end of the plot.
• For each satellite pairing, we first fit the model to the spectrum from each instrument separately to determine the photon index and flux within that GTI.
• Then, as shown in the picture below with Swift and NuSTAR, we fit the data from all instruments in the observatory pair together to get the ratio of the measured flux to the model flux.

• Here are the estimates for the flux and photon index values for each satellite pair that was fitted in 2018. These confidence contours show the range we expect the flux and photon index to be in.
• The actual photon index and flux value is most likely (68%) to be in the centermost oval, it is probably within the 2nd oval (95%), and almost certainly within the outermost.
• We created these confidence contours because they are a useful way to visualize how much agreement there is between observatories.
• If they overlap, like NuSTAR and Chandra in the center of the top row, those satellites are more likely to agree with each other.
• If the contours don’t touch each other, like Swift and XMM or Swift and Chandra in the bottom row, we can deduce that the satellites do not agree with each other.

• Now let's look at the pairings involving nustar over time.
• These are the NuSTAR and Swift confidence contours for all seven years of data.
• As you can see, there was some clear agreement between Swift and NuSTAR in 2015, 2018 and 2021, but there is none in 2016, 2017, 2019, and 2020.
• This drastic change indicates that there is an issue with our Swift data extractions, we are working on extracting the data a second time and consulting with others about the issue.

• These flux ratio plots show the ratio of the measured flux to our model flux when an observatory pair is fitted together. These give us an idea of how good our fit was. Ideally, this ratio would be one at all energy levels for all instruments.
• There is a clear slope in the 2016 and 2020 plots that is also visible in the 2017 and 2019 plots which should not be there. Also, Swift’s flux ratio typically increases with energy, similarly to the flux ratio in 2015, but this pattern is not repeated.
• The low flux ratio of the Swift data is likely what introduces this slope. Again, this suggests that we need to extract and fit the Swift data again. We plan on introducing a cross-normalization constant to the fit for Swift to offset this.

• Here are the confidence contours for NuSTAR and Chandra.
• These are closer to what we expect the confidence contours to look like over time.
• They are all similar, but the years with the best agreement are 2018, 2019, and 2021.

• These are the flux ratios for chandra and NuSTAR, which also look more like what we’d expect.
• Here, the NuSTAR flux ratios are more scattered around a flux ratio of 1.
• The Chandra flux ratio gets higher as energy increases, but this is a documented trend for Chandra, so it’s to be expected.

• Lastly, these are the confidence contours for NuSTAR and XMM.
• The 2021 XMM data is not available yet, so it was not included.
• Looking at these, there is typically good agreement between NuSTAR and the XMM MOS instruments, but XMM’s pn (cyan) instrument ranges from agreeing with the other instruments, like in 2015 or 2019, to measuring a much lower flux value, like in 2016 and 2017.

• Here are the flux ratio plots for NuSTAR and XMM.
• The NuSTAR data has been offset by -0.2 to make the plots easier to read.
• These plots consistently show off a characteristic V in the XMM flux ratio. This has been well-documented and is expected.
• The NuSTAR flux ratios again are scattered around a flux ratio of 1, which is also what we would expect.

• Finally, here are the cross-normalization constants for every year.
• These were calculated by comparing the flux of each instrument listed in the columns to the flux of the NuSTAR instruments on the left.
• Here, we notice trends that can be expected, such as FPMB being slightly higher than FPMA or the NuSTAR instruments being very similar to the XMM MOS instruments.
• However, because the Swift data hasn’t been revisited yet, there are wide variations in the Swift cross-normalization constants.

• Here are the average cross-normalization constants and their standard deviations.
• Again, FPMB is higher than FPMA, the MOS instruments have good agreement with the NuSTAR instruments, and pn is lower than the NuSTAR instruments, which is all expected. The main outlier is still Swift’s low cross-normalization constant and high standard deviation.

• In the future, we plan to double check our spectral data for Swift and Chandra.
• These extractions were done by myself, so we plan to check them with experts and to clean up the mistakes in the Swift extractions and fit.
• We also plan to include the 2021 XMM data once that’s available.
• Lastly, we plan to publish a paper with these results!
• The main reason this work is so essential is because cross-calibrations of satellites ensures accurate results can be obtained from these observatories, which impacts many areas of X-ray astrophysics. So we hope that this work can serve as a reference for those using data collected from different satellites at different times to ensure that their measurements are as accurate as possible.
• Thank you!