HR-MM 2.1 - Radial Velocity measuring

[Marco Leonardi]

Dear Collegaues,

Following the collaboration with Ernst Pollmann related to binary system monitoring I learned from him about an interesting tool usefull to compute the RV. So I have donwload and installed HR-MM 2.1 which is possible to download free at this web site:

http://www.astro.buecke.de/

This tool has been developed by Roland Bücke and it is based on mirroring method. The following is a pubblication very interesting to read in order to learn more about the different methods used to measure RV:

http://journals.cambridge.org/download. ... fec8f757c2

the following table (source: above website) summarize pro con of different methods used to compute the RV:

Table.jpg (70.4 KiB) Viewed 7079 times

basically the mirroring method is based on the bisectors method and in the following pubblication it is very well explained how the bisectors method works:

http://arxiv.org/pdf/astro-ph/0508096v1.pdf

I am testing the software on my last spectrum of pi Aqr, the last one approved by BeSS and dated December 29th , 2015.
I got a Radial Velocity of -4.8 km/sec but I wonder if the value is correct. To be honest I do not how to select correctly the spectrum zone. I have enclosed to this message two images of the different software windows. What do you think ? Is it correct my way to proceed ? What is the "Function of CC" window ?

Function_of_CC.jpg (124.93 KiB) Viewed 7079 times

is it correct to select spectrun zone in this way ?

mio_spectrum.jpg (122.24 KiB) Viewed 7079 times

may be Ernst can clarify some of my doubts. I think it could be useful not only for me but even for others to post my experience and questions. I found the RV matter very interesting.

May be some other of you is using this tool ? What do you think about the mirroring method vs cross-correlation method ?

Many Thanks and Regards,
Marco

[Ernst Pollmann]

Dear Marco,
I am very pleased of your attempts to calculate the RV of pi Aqr. In case of your last spectrum, I would prefer not to use selected sections of the profile, but to use the entire profile (see the following picture).

spectrum.jpg (38.82 KiB) Viewed 7068 times

In that picture you can recognize the corresponding parameter for the process "advanced parameters of spectral region": high level = 2.3, low level = 0. Corresponding the spectral parameters in your spectrum (dispersion and spectral resolution power) you have to select certain values for "correlated spectral region" (I selected 30 Angstr.) and for "velocity range of correlation" (I selected 110 pixel), laboratory wavelength = 6562.852 A.
With these pre-settings you can start the measurement "correlated line" (see the following picture)

correlated line.jpg (41.4 KiB) Viewed 7068 times

and the function of CC (see the following picture).

function of CC.jpg (45.61 KiB) Viewed 7068 times

Sometimes you have to "play" with this settings, depending on the lineprofile, in order to find out the best result of the CC.

The CC determines the size of the shift of the spectrum and/or a comparison (mirrored) spectrum (both leads to the same results, if one considers the direction of the shift), which is necessary, in order to obtain a maximum agreement of the correlated spectra (the measure of the agreement is defined by a coefficient of correlation).
An indicator for the "best result" is the CC curve itself. It has to be (as good as it possible) a symetrical distributed Gaussian function (curve). Unfortunately there is no reference of a correlation coefficient in that program, so the only possibility you have is to judge the CC curve.
I hope that this helps to understand the evaluation process and the CC a bit better.

Then you have to know that pi Aqr is a binary system with two kinds of radial velocities:
1) a RV of a traveling component within the Halpha emission, and
2) a RV of the underlying photospheric absorption profile.
Both with very different amplitudes, but with a common period of 84.1 days, corresponding the orbital period of the companion.
Bjorkman et al. (ApJ, 573, 812-824, 2002) published radial velocities for the Halpha emission of +/- 100 km/s.
With the ephemerides of that paper, you should be able to calculate the RV of your observed spectrum.

Finaly some words to the laboratory wavelength of Halpha:
The wavelength of Halpha is not so simple to specify, because of the fine structure and the hyperfine structure.
The National Institute for Standards and Technology (NIST; http://physics.nist.gov/cgi-bin/ASD/lines1.pl) gives for normal atmospheric pressure a wavelength of 6562.85175 A. This wavelength is used in VSpec among others. To get a reliable RV value, the correct normalizing of the spectrum is important.

Best wishes,
Ernst

[Robin Leadbeater]

Hi Marco,

For any typical line profile, particularly complex ones like this, there is no "right answer" for the absolute value of RV you get for the line as it will depend heavily on the method used to calculate it and the parameters used. With projects like this where you are trying to track changes, I have found that the most important thing is for all the measurements to be performed in exactly the same way using the same algorithm and parameters. (This is sometimes achieved for example by one person being responsible for making the measurements on all the submitted spectra.)

Cheers
Robin

[Marco Leonardi]

Thanks Ernst for your very clear and complete explanation and thanks Robin for your considerations. I agree about the point to keep all parameters consistent: same method, same algorithm, same parameters, same person for spectra and RV evaluation.

About the methods my opinion is that the mirroring method seems to be very effective for the RV evaluation of this type of line profile while the cross-correlation method needs a reliable and good template and I wonder if those templates are always available for binary systems or similar line profiles.

Considering another point highlighted by Ernst, which is accurate spectrum normalization, I believe that VSpec or ISIS can do the job in accurate way. Ernst, If you have any hint...thanks again for sharing !

Regards
Marco

[Robin Leadbeater]

Hi Marco,

I think the cross correlation method is more useful when you have a wide spectral range with many lines. The correlation maximum is then very clear (Often the star itself is used as the template and you measure the RV differentially at different dates) It does not work so well with just one line though.

For a symmetric line I usually use a Gaussian fit but if the line is significantly asymmetric or varying in shape I either try to separate it into different components representing different physical processes or just use the centroid. (The spectral analysis program SPLAT has some nice tools for measuring lines
http://star-www.dur.ac.uk/~pdraper/splat/splat.html )

As an example, for this paper (page 36, fig 8a),
http://arxiv.org/pdf/1409.5493v1.pdf
I measured the RV of the 7699A line for all the spectra from the three different observers using the centroid . (The line was very asymmetric and changed in shape dramatically during the measurement period). Originally the three observers had used different ways of measuring the rv and the agreement between us was poor. Only when I remeasured the other spectra using the same technique I had used, did the three sets of data come into good agreement.

Cheers
Robin

[Ernst Pollmann]

Dear Marco,
a very elegant way is to use the "Spline filter" function in VSpec.
I eliminated in your pi Aqr spectrum at first with the function "Supress Zone" the entire profile of the Halpha emission for that process.
Then I used in that "pseudo-continuum" the "Spline filter" function in order to find a "reasonable smoothed continuum" (see the following picture).

contfit.jpg (34.47 KiB) Viewed 7038 times

Finaly I divided the original spectrum by this "smoothed continuum" to get the normalized spectrum. That´s all.

Best wishes, Ernst

[Marco Leonardi]

Dear Ernst,
here we go ! I have followed your suggestions and I got an RV of -2.9 km/sec. Here below some images.

suppress-zone

supress_zone.jpg (83.16 KiB) Viewed 7030 times

spline-filter

HRV- MM show spectral region

HRV_MM_show_spectral_region.jpg (118.97 KiB) Viewed 7030 times

HRV MM Function of CC


As Laboratory wavelenght I used 6562.852 Amgstroms.

Thanks again !

Marco

[Ernst Pollmann]

Dear Marco,
-2.9 km/s is exactly what I measured too.
Now it could be interesting to see, what the ephemeries of the paper of Bjorkman et al. (ApJ, 824, 2000) says.
Related to these ephemeries, you took your spectrum at orbital phase 0.512.
The RV diagrams at page 820 shows that at this phase we should expect a RV between (approx.) 0 to 10 km/s (the Halpha absorption during the B phase of pi Aqr).
Since we are not able to observe the traveling emission component, whose RV is also shown in the diagram, we can compare our RV only with this Halpha absorption.
But Bjorkman et al. observed the star during its quasi-normal B phase, what means that we don´t have any possibilities at all, to compare the RV of our disk emission.
But I think, the RV of the Halpha absorption during the B phase of the star, can´t be far from the RV of the disk. Particularly because the disk is relatively small at the moment (as our EW monitoring shows).

Best wishes,
Ernst