This repository is a guide to fitting spectral lines to identify potential activity-sensitive lines that could interfere with radial velocity detections. NEID files are made into templates, which are fed into the pipeline to pull out line parameters. A .csv file is saved from the pipeline so the resutls can be read and plotted in the appropriate notebook. This is where correlation coefficients are calculated, and k-means clustering is performed to identify groups of activity indicators.
Radial Velocity detections are one of the most robust ways of detecting exoplanets. When a planet orbits around its host star, the pull of the planet causes the star to "wiggle" about the center of gravity between the two objects. This creates a Doppler-shifting effect--the star appears blue-shifted as it moves toward us, and red-shifted when it moves away from us. We can observe this effect through spectroscopy, which is a method of separating light into its component parts by wavelength using a diffraction grating. We observe the Doppler shift in the star's spectrum, as seen below.
Unfortunately, the star has a very active atmosphere, which can distort spectral lines, and interfere with RV signals. In fact, in order to detect an Earth-like planet around a star similar in size to our Sun, we would need to be able to detect a velocity of ~9 cm/s, but stellar activity creates a noise floor around 1 m/s, nearly an order of magnitude above this signal. Thus, it is important to figure out which spectral lines are magnetically-sensitive, so our RV detections are less-prone to contamination from the stellar surface. We begin with a line list created for the Sun, which uses physics and radiative-transfer code to calculate theoretical wavelength, depth, and full-width half-maximum (FWHM) values for spectral lines we should observe in the Sun. This is the motivation for this project.
- spectra_pipeline.ipynb : Begin with this notebook. This notebook starts with reading in NEID spectra, which are in the format of .fits files with can be found at the Google Drive link below:
Link to neid_solar_data on Google Drive: https://drive.google.com/file/d/1sRNEKcAEOc93sTisQP-flMgLUUpcklcV/view?usp=drive_link Folder containing 3 NEID .fits solar files over the course of 20 days near solar noon. These are used to create the templates.
There are 3 .fits files containing spectra per day taken near solar noon over the course of 20 days. Using these files, we then create a template for each day. A template is a daily-averaged spectrum, and increases the SNR of our data while also filling in data gaps or nan values. We average these 3 files per day to create 20 total templates. An example template spectrum is shown below:
Once these templates are created, they are fed into the line-by-line pipeline. This goes through each individual line list wavelength, and finds where it should be located in the observed NEID spectrum (the template). Once a line is identified, a modified Gaussian is fit to it, and FWHM and Depth are recorded. Below shows an example of a single absorption line from the NEID spectra, along with the parameters that are measured.
These parameters are written to a .csv file in the user's directory, so they can be read in in the second notebook, plot_results.
- plot_results.ipynb : This notebook reads in the dataproducts from the .csv file created by the pipeline for analysis. Activity indicators trace stars' activity levels. We will be using two for our analysis. The first is Ca II H & K S-Index, which is a NEID data product that we pull from the .fits file. This index is calculated using the Ca II H and Ca II K absorption lines, which are two extremely magnetically-sensitive lines. The second is B_obs, technically called unsigned magnetic flux, which is a proxy for the Sun's magnetic field strength. We use Solaster, a package which calculates solar activity indicators using Solar Dynamics Observatory Data. Because there isn't data for a few of our days, we had to cut out two NEID templates. The solar activity indicators are taken daily, but are interpolated to the same time as our NEID observations. An example time series of a solar activity indicator is shown below.
We then calculate four Pearson Correlation Coefficients (PCC) for each line--Depth vs. B_obs, FWHM vs. B_obs, Depth vs. S-Index, and FWHM vs. S-Index. We then identify which lines make the PCC > 3.0 threshold for being activity-sensitive. We run a PCA on these data and use K-Means clustering with 4 clusters to see the separations between different correlation coefficients. We then create a dataframe with all the correlation coefficents, including those that do not meet the threshold for being activity-sensitive, and run a PCA. We perform K-means clustering on this data with 2 clusters, to identify activity-sensitive versus activity-insensitive lines in an alternative way. We see that the line parameter that is chosen for correlation has a stronger influence over grouping than the activity indicator that is chosen. This has ramifications for future RV studies, because multiple line parameters should be used to identify activity-sensitive lines which need to be removed from RV analysis.
-
tools.py : Contains functions used in spectra_pipeline.ipynb and plot_results.ipynb. Included in the import statements.
-
final_calcsold.csv : These are dataproducts from the Solaster pipeline, which calculates solar activity indicators using Solar Dynamics Observatory data.
-
new_gurtovenko_2015.csv : Theoretical line list pulled from Gurtovenko et al. 2015. Observed lines are matched to theoretical ones from this list based upon theoretical wavelength and theoretical line depth. If two observed lines are approximately equidistant from the theoretical line, the observed line with the depth closest to the theoretical depth is selected.
-
Link to neid_solar_data on Google Drive: https://drive.google.com/file/d/1sRNEKcAEOc93sTisQP-flMgLUUpcklcV/view?usp=drive_link Folder containing 3 NEID .fits solar files over the course of 20 days near solar noon. These are used to create the templates.