This package contains the code I developed for my master thesis in cooperation with dr. Andrew Tkachenko. It consists of two major parts. The first computes the least-squares deconvolution profiles of the input spectrum, using the masks given in the UserInput.dat file. The second uses the same masks and the computed LSD profiles to build a reconstruction of the observed spectrum, but with a higher SNR. Both routines are written in Fortran 90/95, although a python script is included, which provides the commands to compile & run the code, and plot the results. More information on the used algorithms can be found in: https://ui.adsabs.harvard.edu/abs/1997MNRAS.291..658D/abstract https://ui.adsabs.harvard.edu/abs/2010A%26A...524A...5K/abstract https://ui.adsabs.harvard.edu/abs/2013A%26A...560A..37T/abstract
In addition, the "Data" folder contains files for a tutorial.
ACKNOWLEDGEMENTS If the code is used for scientific publications, please cite: Tkachenko A., Van Reeth T., Tsymbal V., Aerts C., Kochukhov O., Debosscher J., 2013, "Denoising spectroscopic data by means of the improved least-squares deconvolution method", Astronomy & Astrophysics, Volume 560, id.A37, 14 pp. (https://ui.adsabs.harvard.edu/abs/2013A%26A...560A..37T/abstract)
WHEN USING THE CODE FOR THE FIRST TIME:
If you want to use the overarching python script, you have to create a link to the file /Source/LSDProject.py in the main "LSDProject" folder. In GNU/Linux, this can be done by going to the LSDProject folder in the terminal and entering the command:
$ ln -s ./Source/LSDProject.py
The code can then be compiled by adding the option "-c" when running the python script:
$ python LSDProject.py -c
This has to be done twice. Error messages shown after the first compilation can be ignored. (This is due to the fact that the underlying modules have to be compiled before the executable can be compiled properly.) By default, the code uses the ifort compiler and its mkl library. If you want to use another compiler, this can be done by compiling the code "manually" or adapting the python script. The code should work for other compilers as well, though in this case, you have to (install and) use liblapack and libblas (i.e., include the options "-lblas -llapack" when compiling) instead of the mkl library. Additional information on the use of the python script can be found by entering the command
$ python LSDProject.py -h
in the terminal.
FOR GENERAL USE OF THE CODE:
The algorithm makes use of a large section of a normalised spectrum. In this range, there should be a large number of metal lines, but no hydrogen, helium or metal lines with damping. A good example is the section between 5000 and 5800 angstrom in the spectrum of an F-type star. The used "masks" on the other hand, are in fact line lists containing the central wavelengths, the residual fluxes at the center of the lines, and the names of the ions corresponding to the lines. (This information can for instance be obtained from the VALD website.) To allow for a more robust error estimation, there is also an option to include a file containing the variance of the observed spectrum. However, this is not necessary for the code to run. The files have to be located in the "Spectra" and "Models" subdirectories of the folder "Data". These locations then have to be filled in into the file UserInput.dat. For the tutorial this gives us:
Spectrum = /Tutorial/synthetic_binary_070-030.dat Variance = / Model = /Tutorial/lp0000_08500_0340_0020_on_new.lin Model = /Tutorial/lm0003_07300_0350_0020_on_new.lin
Here we have a "/" for the variance, as we do not provide it to the code in this case. Similarly, we also have to provide the location in the "/Data/Results" folder where we want to save the results.
Result = /Tutorial/
Other related entries include:
wavel_beg & wavel_end: marking the edges of the studied spectral range vel_beg & vel_end: marking the edges of the computed profile in velocity space Regular: the Tikhonov regularisation parameter, which allows us to "smooth out" the enhanced noise resulting from the matrix inversion in the algorithm. The value used in the tutorial (0.001) can be considered to be "safe" in most cases. In addition, this code does not smooth the LSD profile itself, but the difference between the LSD profile and the CCF profile of the star. While this results in a slightly higher noise level in the result, it does prevent the smoothing of intrinsic characteristics of the profile. Nr of LSDs: As the algorithm in this code is based on the improved least-squares deconvolution (iLSD) proposed by Kochukhov et al. (2010), we have the possibility to compute LSD profiles for groups of spectral lines with different line strengths. The "separations" between the groups are provided in the next line ("Line limits"). When computing a single LSD profile for our mask, the "separation values" are ignored.
There is also the possibility to make use of multiple models, allowing us to compute the LSD profiles for multiple contributing stellar components, or to take into account "anomalies", such as groups of spectral lines for a pulsating star, which respond differently to the influence of pulsations.
The computation of the LSD profiles then results in the following files being created in the chosen "results" folder:
comp1.lsd: the computed LSD profile(s). The first column gives us the velocity in km/s. If N groups of spectral lines were defined, based on the line strength, then the following N columns give us the values of the N LSD profiles in the velocity domain. The final N columns give us the error margins on these profiles. Naturally, if multiple models were used in the computation, multiple files comp.lsd are created. log: Contains the numerical integrals (computed with the trapezium rule) of the computed LSD profiles, and their sum for each individual model which was used. The ratio of these integrals can be used to roughly estimate the light factor of a binary, as discussed in my thesis text.
Finally, the computed LSD profiles and (if included) the reconstructed spectrum can be plotted by adding the option "-p" or "--plot" when running the python script.
$ python LSDProject.py -p
Questions, remarks and information on encountered bugs can always be send to timothy.vanreeth@kuleuven.be.
Good luck with your analyses!
Timothy Van Reeth