xi Astronomer Maurice Wilson

Finding X-ray Coronal Cycles

During the summer of 2014, I worked on research within the field of stellar astrophysics at the Harvard-Smithsonian Center for Astrophysics (CfA) in Cambridge, MA. I am grateful to have earned the NSF REU internship that they offer. This was my second internship but my first REU, and it was a phenomenal experience. I am eager to share a small part of that experience here. I will briefly discuss the concepts and methodology of my research project. I will supplement the nonubiquitous astronomy jargon and relatively complex concepts you may see here with a few comprehensible and easy to read blog posts, which is what I currently do with my exoplanetary research project. This stellar research is still ongoing but I am happy to explain my progress thus far. I will continue to update this web site as I learn more about magnetic activity in stellar coronae. I hope you enjoy!

Objective: I seek to verify whether or not 10 selected stellar sources within the Chandra Deep Field South exhibit magnetic activity cycles via X-ray emission from coronae.

Magnetic Activity Cycle

In order to find stars that have X-ray magnetic activity cycles, I must, of course, first know what an X-ray coronal cycle looks like. Our Sun is the most convenient source of information regarding what a magnetic activity cycle is.

Figure 1: Solar X-ray Cycle (NASA, ISAS Yohkoh Mission)

Figure 1 gives a nice look at how the Sun's magnetic activity fluctuates over the course of about 11 years. The foremost image with Figure 1 shows a somewhat strong magnetic activity. (The brightness of areas represent strength of magnetic activity.) As you look from there in a counter-clockwise direction you'll notice that the Sun gets dimmer and thus less magnetic activity is seen. At the top, where the furthest image of the Sun in this mosaic lies, the Sun has very weak magnetic activity. This point in time is known as solar minimum. Continuing counter-clockwise from there, the magnetic activity increases significantly until you reach the bottom left image of the Sun, which is the time when the strongest magnetic activity is seen. This indicates the time of solar maximum. Solar maximum marks the time when the Sun's north and south magnetic poles flip! Needless to say, this is a pretty big deal. I find it pretty cool that the astrophysics community, with as many geniuses it has to , has yet to find out exactly why this happens and why this happens every ~11 years. I also think it's cool that humans are oblivious to this phenomenon--mostly because none of the effects are physically felt by humans on Earth. This ~11-year magnetic pole flip is known as the sunspot cycle. It can also be called half of the 22-year magnetic activity cycle.

Figure 2: The Solar Chromosphere
Figure 3: The Solar Corona (composite image)
(ESA/Proba-2/SWAP, S. Koutchmy/J.Mouette)

Different events can be seen in different layers of the Sun's atmosphere. The chromosphere is a thin layer immediately above the photosphere. In the photosphere, we can see sunspots. In the chromosphere, we may see spicules or prominences. Temperatures in the chromosphere can range from 6,000 to 30,000 Kelvin (K). The thickness of the chromosphere is about 20,000 km.
The corona is above the chromosphere and is the outermost atmospheric layer of the Sun. Figure 3 shows how far the coronal structures can extend from the Sun. This atmospheric layer is immensely thicker than the chromosphere and any of the Sun's other layers. The blue and white regions are show an image of a solar eclipse, but the moon is replaced by an image of the Sun (when looking at its corona face-on) in this composite image. Within the corona, we can see active regions and flares. Temperatures range from about 500,000 K to 5 million K depending on location within the corona. The depth of the corona is not rigidly defined because material and large structures from its corona extend far away from the Sun.

The light, or electromagnetic waves, emanating from these atmospheric layers differ substantially in energy. The feasibility of detecting periodicity in the magnetic activity of stars differs depending on which energy band is observed. Thus far, stellar ultraviolet (UV) emission has proven to be the most convenient medium for detecting magnetic flux variability over long time scales. Simply put, ground based telescopes can easily detect the UV band from stars and this allows us to observe the Ca II H and K emission lines that have provided conclusive evidence for the existence of stellar magnetic cycles in stars other than our Sun. (This is a pretty big deal. Feel free to check out the "Chromospheric Variations in Main-Sequence Stars" paper.) Unfortunately, X-ray observatories must operate above the Earth's atmosphere since X-rays are thoroughly absorbed before reaching the Earth's surface. Because of the substantial expenses and immense competition for observing time that must come with such an observatory, there have been more observation projects monitoring stars in the UV bands. However, X-ray observations have been important when searching for activity cycles because this energy range conveys coronal activity information from stars. In main-sequence stars, coronae contain the active regions where charged plasma follow magnetic field lines. These active regions are typically above a pair of sunspots in the photosphere. Because active regions from distant stars provide magnetic cycle information, instruments used for detecting magnetic activity periodicities should be capable of detecting variability without the aid of the photosphere's sunspots.

Because of the difficulties involved with X-ray observations, there are only four stars that have been confirmed to have an X-ray coronal cycle along with their chromospheric magnetic activity cycle. Does that not sound strange?? After all these years of "advanced" technology, we've only confirmed that FOUR stars in our entire Universe have a magnetic activity cycle in their coronae! Of course, there are probably billions more but we just have not confirmed them yet. Nonetheless, our obliviousness to this process occurring on other stars actually fascinates me. I like mysteries, I guess. But, I digress.
One of the four stars includes our Sun. Another of these stars is HD 81809, which is where the data plotted in Figure 4 comes from.

Figure 4: An Ideal Light Curve

The space-based XMM-Newton Telescope observed the star HD 81809 in the X-ray bands over the course of ~9 years. The measurements from two of its detectors (MOS1 and PN) are plotted in Figure 4. (XMM-Newton's third detector is called MOS2 and its data is not shown here only for the sake of clarity. Its data agrees with MOS1 within 1 sigma anyway.) Data in Figure 4 exhibits a sinusoidal curve, which is ideally what one would like to see when looking for a cycle/periodicity in a light curve. That one lonely datum at the top is most likely due to a flare that occurred during that epoch of observations. Such flaring is not considered when searching the light curve for periodicity, therefore that lonely data point is ignored. Such a sinusoidal pattern is what I want to see in the light curves of the stars I investigate.

Chandra Deep Field South

For my project, ten stars, that have been observed repeatedly by Chandra and XMM-Newton, are analyzed as potential candidates for detecting cyclic variability in their magnetic activities. These ten were chosen because they all are in the Chandra Deep Field South (CDFS). Chandra and XMM-Newton have observed them frequently for over a decade. This field is primarily observed for studying galaxies. However, there are stellar point sources seen near the edge of this field. Although these point sources are faint, both X-ray observatories have gathered much data over the years since their missions commenced. Because of the long length of time these stars were observed, it is likely that I may discern flux variability over a long time period similar to our Sun's magnetic activity cycle. Table 1 lists the positions and identications of these sources. Table 2 shows their magnitudes in various wavelengths. Figure 5 illustrates each source's position in the observed field.

Figure 5: Chandra Deep Field South

Figure 5 shows an exposure corrected mosaic image of 56 Chandra Deep Field South observations. This image covers an energy range of 0.5 - 6.0 keV and only the ACIS-I (CCD ID = 0-3) detector of Chandra. The total exposure time of this image is 3.8 million seconds taken over the course of ~11 years. Fortuitously, XMM-Newton observed this field over a ~9 year timescale as well (see Table 4). The magenta circles correspond to the sources listed in Tables 1-4 that I analyze. The size of the circles are exaggerated just to allow the reader to see the brightness (or faintness) of the stars. Within some circles, there seems to be very bright sources. However, this is misleading because those sources are not stars. This is why those bright sources within a few circles (e.g. circle #8) are off-centered. Most of the bright sources seen in this region of the sky are not stars, hence why it is considered to be a galactic field. To determine which sources were extended or pointlike (stellar) sources, I used the "wavdetect" function in software known as Chandra's Interactive Analysis of Observations (CIAO). After I used CIAO to find the stellar sources within the CDFS, search for the ten most bright stars within this galactic field. Early in the analysis, we decided that the 5th source was too faint to obtain a good quality spectral fit. Thus, you will not see information for this star in the tables of the appendix.

Figure 5 pretty much sums up the major problems that I have to deal with in this project. I'm dealing with faint stars that do not even remain in the field of view over the 11 year period. As you can see in Figure 5, the 56 images stacked on each other do not have the same orientation. This is due to the rotation of the ACIS-I detector as it observes the CDFS over 11 years. This means that some (if not all) of my nine stars will be missing from several of the 56 observations! Source #3 practically has no chance of being observed for a long period of time. It's too close to the edge of the detector. Notice how extremely faint it is too. Like most of my stars, you cannot see it within its circle. Although, there might be a different color scale that I could have used when making the mosiac that perhaps would bring out the stars' light enough so that we could see my dim pointlike objects. However, I am not sure if that this is true. I only tried a few different color scales and did not have such luck. Nonetheless, this blue color scale on the image is beautiful! And of course, at the end of the day that's all that really matters. (Blue is my favorite color, by the way.)

After reading the previous two paragraphs, you're probably wondering why I am bothering to analyze faint stars within a galactic field when I am running into several big issues anyway. Well, this field is actually the most promising field left for one on the hunt for X-ray coronal cycles. This is true primarily because this field has been observed over the course of ~11 years in the X-rays. Such a field is far from common, due to the costs and competition surrounding X-ray observatories as mentioned previously. Furthermore, there are some other fields that have bright X-ray, stellar sources observed for a long period of time but those stars have already been researched. Among the very few--probably less than a handful of--chronically observed fields left up for grabs, the CDFS is my best bet for finding an X-ray coronal cycle.

Optimal Spectral Fit

After finding the brightest ten stars, I grouped the observations into epochs (shown in Tables 3 and 4). I then extracted the spectra of each star for each observation.

Figure 6: Spectra for one epoch of observations from XMM-Newton and Chandra. This data has the background still included. A model was fit to the flux due to both the source and background.

In Figure 5, it is worth noting that the Chandra flux data have much larger error bars than the flux for XMM-Newton. Also, if you look at the y-axis you'll notice that the flux is much weaker for the Chandra data than the XMM-Newton data. The weaker flux is due to the significantly less number of counts detected by the Chandra satellite in comparison to XMM-Newton. These relations for the error bars and counts between both observatories are seen for most of the observations. Such relations conspicuously affect the final results, as you will see.

For fitting these spectra (1 per epoch), I used the APEC model for the stellar source's counts. This model is well-suited for my needs because it accounts for the optically thin coronal plasma of my stars. Determining what to do about the background noise was quite difficult however. Because of the very few stellar source counts (i.e. few X-ray photons from faint stars), it was statistically problematic to simply subtract the background. (This situation is opposite of the aperture photometry/background subtraction I performed for the bright stars of my exoplanet research.) Instead, I sought for a model that would fit the background well. After weeks of searching and testing, I finally found that the background model that fits the background the best is actually two models. For a complicated reason, that I will not discuss here, we determined that the best background fit came from a Power Law added to a Constant Function for the XMM-Newton data and a mere Constant function was needed for the Chandra data. These models may have proven to be the best, but the light curves will show that the fit for the Chandra data was still far from satisfactory.

Light Curves

Figure 7: Four Resultant Light Curves

Amongst the nine stars scrutinized, I present light curves for the four brightest sources. The effect of analyzing such faint sources is illustrated in the error bars. Some of the large errors are due to few observations (with exposure times of ~10ks) within certain epochs. In all four light curves, the XMM-Newton fluxes have smaller error bars than the Chandra fluxes. As stated previously, the XMM-Newton data provided much more counts than Chandra, and thus the spectral fits for the XMM-Newton data provided more stringent constraints than for the Chandra data. Nonetheless, valuable information can be extracted from the light curves in regards to the sources' long term variability.

None of my sources seem to be exhibit flux variability of a factor ~10 like the Sun. The flux of source #4 seems to remain constant throughout the observations. There is a lack of data between 2002 and 2008. Moreover, the one data point within that time period is associated with a large error value. Therefore, we are not certain if the corona truly remained constant throughout these years. However, we are confident that the flux variability did not exceed a factor of 3. Source #6 certainly expresses flux variability, although this does not seem to be periodic. We are not sure if this is a sign of long-term variability. This is the brightest source out of the nine selected (although still faint). Its brightness allowed for a good spectral fit, which resulted in the relatively small errors associated with its flux values. There is an excessive time span (near 2002 to 2008) in which there were no observations of this source. The flux variability does seem to reach a factor of 2 if we assume its quiescent flux is ~\( 7 \times 10^{-15} \frac{erg}{s\cdot cm^2} \). Source #7 unfortunately has relatively large errors for every point on the light curve. Its flux seems to remain constant throughout all epochs. The flux of source #9 also seems to remain quiescent throughout the epochs. However, like the other light curves, we cannot confidently conclude that it is not an active star. Among other obvious reasons, this lack of confidence is due to a 6 year gap in our data.


For this analysis, I utilized CIAO and SAS to download and reprocess the data for nine point sources observed within the CDFS. Although faint, these sources were useful to analyze due to their frequent appearance within the Chandra and XMM-Newton observations across a period of ~11 years. The lack of photons detected (for the point sources) significantly decreased the feasibility of obtaining good quality spectral fits. We find that an APEC model comprised of merely one function known as "xsapec" (in Sherpa) sufficed in describing our stellar sources. The Chandra background spectra were low in counts primarily due to the small background regions used to extract the spectra. The background regions needed to be small because of the many dispersed sources (point and extended) within the field observed. Consequently, the Chandra spectral fits were not satisfactory. However, the XMM-Newton spectral fits were satisfactory but the small amount of epochs used for our light curves hindered us from confidently concluding whether or not our four brightest sources exhibited long-term flux variability similar to the Sun. Source #6 seems to be the best candidate within our nine star sample for possibly having a long-term coronal cycle. However, we do conclude that none of the four sources express a long-term variability above a factor of 3. This analysis attests to the difficulty of conclusively discovering long-term X-ray coronal cycles without the initial aid of Ca II H and K emission information from the stellar chromospheres.


Table 1: Identities

* Because these sources are faint, few people have observed and/or analyzed them. Ellipses (...) are seen because these sources have not been classified in the literature. It has also been difficult to find some of their magnitudes, as Table 2 shows. See the footnote at the below Table 2 for the references corresponding to the information (assigned with numerical superscripts) in this table.

Table 2: Magnitudes

These measurements are in the AB astronomical magnitude system. This system is based on flux measurements that were calibrated in absolute units.

  • \( ^{1} \)(Silverman et al. 2010)
  • \( ^{2} \)(Luo eta al. 2010)
  • \( ^{3} \)(Moy et al. 2003)
  • \( ^{4} \)(Groenewegen et al. 2012)
  • \( ^{5} \)(Virani et al. 2006)
  • \( ^{6} \)(Cutri et al. 2003)
  • \( ^{7} \)(Lehmer et al. 2005)
  • \( ^{8} \)(Giavalisco et al. 2004)
  • \( ^{9} \)(Giacconi et al. 2002)

Table 3: Epochs - Chandra

Observation identifications are listed in epochs comprised of observations that occurred within relatively short time periods of each other.

Table 4: Epochs - XMM-Newton

Observation IDs are shown for the observations that we used within these epochs.

    Table References

  • Cutri, R. M., Skrutskie, M. F., van Dyk, S., et al. 2003, VizieR Online Data Catalog, 2246, 0
  • Giacconi, R., Zirm, A., Wang, J., et al. 2002, ApJS, 139, 369
  • Giavalisco, M., Ferguson, H. C., Koekemoer, A. M., et al. 2004, ApJ, 600, L93
  • Groenewegen, M. A. T., Girardi, L., Hatziminaoglou, E., et al. 2002, A\&A, 392, 741
  • Lehmer, B. D., Brandt, W. N., Alexander, D. M., et al. 2005, ApJS, 161, 21
  • Luo, B., Brandt, W. N., Xue, Y. Q., et al. 2010, ApJS, 187, 560
  • Moy, E., Barmby, P., Rigopoulou, D., et al. 2003, A\&A, 403, 493
  • Silverman, J. D., Mainieri, V., Salvato, M., et al. 2010, ApJS, 191, 124
  • Virani, S. N., Treister, E., Urry, C. M., & Gawiser, E. 2006, AJ, 131, 2373
Acknowledgements: Harvard-Smithsonian Center for Astrophysics, National Science Foundation REU