path: root/noao/onedspec/doc/sys/rvidentify.ms

.RP
.TL
Radial Velocity Measurements with IDENTIFY
.AU
Francisco Valdes
.AI
IRAF Group - Central Computer Services
.K2
P.O. Box 26732, Tucson, Arizona 85726
.AB
The IRAF task \fBidentify\fR may be used to measure radial velocities.
This is done using the classical method of determining the doppler shifted
wavelengths of emission and absorption lines.  This paper covers many of
the features and techniques available through this powerful and versatile
task which are not immediately evident to a new user.
.AE
.NH
Introduction
.PP
The task \fBidentify\fR is very powerful and versatile.  It can be used
to measure wavelengths and wavelength shifts for doing radial velocity
measurements from emission and absorption lines.  When combined with
the CL's ability to redirect input and output both from the standard
text streams and the cursor and graphics streams virtually anything may
be accomplished either interactively or automatically.  This, of
course, requires quite a bit of expertise and experience with
\fBidentify\fR and with the CL which a new user is not expected to be
aware of initially.  This paper attempts to convey some of the
possibilities.  There are many variations on these methods which the
user will learn through experience.
.PP
I want to make a caveat about the suggestions made in this paper.  I wrote
the \fBidentify\fR task and so I am an expert in its use.  However, I am not
a spectroscopist, I have not been directly involved in the science of
measuring astronomical radial velocities, and I am not very familiar with
the literature.  Thus, the suggestions contained in this paper are based
on my understanding of the basic principles and the abilities of the
\fBidentify\fR task.
.PP
The task \fBidentify\fR is used to measure radial velocities by
determining the wavelengths of individual emission and absorption
lines.  The user must compute the radial velocities separately by
relating the observed wavelengths to the known rest wavelengths via the
Doppler formula.  This is a good method when the lines are strong, when
there are only one or two features, and when there are many, possibly,
weaker lines.  The accuracy of this method is determined by the
accuracy of the line centering algorithm.
.PP
The alternative method is to compare an observed
spectrum to a template spectrum of known radial velocity.  This is done
by correlation or fourier ratio methods.  These methods have the
advantage of using all of the spectrum and are good when there are many
very weak and possibly broad features.  Their disadvantages are
confusion with telluric lines, they don't work well with just a few
real features, and they require a fair amount of preliminary
manipulation of the spectrum to remove continuum and interpolate the
spectrum in logarithmic wavelength intervals.  IRAF tasks for
correlation and fourier ratio methods are under development at this
time.  Many people assume that these more abstract methods are inherently
better than the classical method.  This is not true, it depends on the
quality and type of data.
.PP
Wavelength measurements are best done on the original data rather than
after linearizing the wavelength intervals.  This is because 1) it is
not necessary as will be shown below and 2) the interpolation used to
linearize the wavelength scale can change the shape of the lines,
particularly strong, narrow emission lines which are the best ones for
determining radial velocities.  A second reason is that
\fBidentify\fR currently does not recognize the linear wavelength parameters
produced during linearization.  This will be fixed soon but
in the mean time the lines must be measured in pixels and converted
later by the user.  Alternatively one can determine a linear dispersion solution
with \fBidentify\fR but this is more work than needed.
.PP
This paper is specifically about \fBidentify\fR but one should be aware of the
task \fBsplot\fR which also may be used to measure radial velocities.  It
differs in several respects from \fBidentify\fR.  \fBSplot\fR works only on linearized
data; the wavelength and pixel coordinates are related by a zero point and
wavelength interval.  The line centering algorithms are different;
the line centering is generally less robust (tolerant
of error) and often less accurate.  It has many nice features but is
not designed for the specific purpose of measuring positions of lines
and, thus, is not as easy to use for this purpose.
.PP
There are a number of sources of additional information relating to the
use of the task \fBidentify\fR.  The primary source is the manual pages for
the task.  As with all manual pages it is available online with the
\fBhelp\fR command and in the \fIIRAF User Handbook\fR.  The NOAO
reduction guides or cookbooks for the echelle and IIDS/IRS include
additional examples and discussion.  The line centering algorithm
is the most critical factor in determining dispersion solutions and
radial velocities.  It is described in more detail under the help
topic \fBcenter1d\fR online or in the handbook.
.NH
Method 1
.PP
In this method arc calibration images are used to determine a wavelength
scale.  The dispersion solution is then transferred to the object spectrum
and the wavelengths of emission and absorption lines are measured and
recorded.  This is relatively straightforward but some tricks will make
this easier and more accurate.
.NH 2
Transferring Dispersion Solutions
.PP
There are several ways to transfer the dispersion solution from an arc
spectrum to an object spectrum differing in the order in which things are
done.
.IP (1)
One way is to determine the dispersion solution for all the arc images
first.  To do this interactively specify all the arc images as the
input to \fBidentify\fR.  After determining the dispersion solution for
the first arc and quitting (\fIq\fR key) the next arc will be displayed
with the previous dispersion solution and lines retained.  Then use the
cursor commands \fIa\fR and \fIc\fR (all center) to recenter and
recompute the dispersion solution, \fIs\fR to shift to the cursor
position, recenter, and recompute the dispersion solution, or \fIx\fR
to correlate features, shift, recenter, and recompute the dispersion
solution.  These commands are relatively fast and simple.
.IP
A important reason for doing all the arc images first is that this same
procedure can be done mostly noninteractively with the task
\fBreidentify\fR.  After determining a dispersion solution for one arc
image \fBreidentify\fR does the recenter (\fIa\fR and \fIc\fR), shift
and recenter (\fIs\fR), or correlation features, shift, and recenter
(\fIx\fR) to transfer the dispersion solutions between arcs.  This is
usually done as a background task.
.IP
To transfer the solution to the object spectra specify the list of
object spectra as input to \fBidentify\fR.  For each image begin by
entering the colon command \fI:read arc\fR where arc is the name of the
arc image whose dispersion solution is to be applied; normally the one
taken at the same time and telescope position as the object.  This will
read the dispersion solution and arc line positions.  Delete the arc
line positions with the \fIa\fR and \fId\fR (all delete) cursor keys.
You can now measure the wavelengths of lines in the spectrum.
.IP (2)
An alternative method is to interactively alternate between arc and
object spectra either in the input image list or with the \fI:image
name\fR colon command.
.NH 2
Measuring Wavelengths
.PP
.IP (1)
To record the feature positions at any time use the \fI:features file\fR
colon command where file is where the feature information will be written.
Repeating this with the same file appends to the file.  Writing to 
the database with the \fI:write\fR colon command also records this information.
Without an argument the results are put in a file with the same name as the
image and a prefix of "id".  You can use any name you like, however,
with \fI:write name\fR.  The \fI:features\fR command is probably preferable
because it only records the line information while the database format
includes the dispersion solution and other information not needed for
computing radial velocities.
.IP (2)
Remember that when shifting between emission and absorption lines the
parameter \fIftype\fR must be changed.  This may be done interactively with
the \fI:ftype emission\fR and \fI:ftype absorption\fR commands.  This parameter
does not need to be set except when changing between types of lines.
.IP (3)
Since the centering of the emission or absorption line is the most
critical factor one should experiment with the parameter \fIfwidth\fR.
To change this parameter type \fI:fwidth value\fR.  The positions of the
marked features are not changed until a center command (\fIc\fR) command
is given.  \fIWarning: The all center (\fIa\fR and \fIc') command automatically
refits the dispersion solution to the lines which will lose your
arc dispersion solution.\fR
.IP
A narrow \fIfwidth\fR is less influenced by blends and wings but has a larger
uncertainty.  A broad \fIfwidth\fR uses all of the line profile and is thus
stable but may be systematically influenced by blending and wings.  One
possible approach is to measure the positions at several values of
\fIfwidth\fR and decide which value to use or use some weighting of the
various measurements.  You can record each set of measurements with
the \fI:fe file\fR command.
.IP (4)
For calibration of systematic effects from the centering one should obtain
the spectrum of a similar object with a known radial velocity.  The systematic
effect is due to the fact that the centering algorithm is measuring a
weighted function of the line profile which may not be the true center of
the line as tabulated in the laboratory or in a velocity standard.
By using the same centering method on an object with the same line profiles
and known velocity this effect can be eliminated.
.IP (5)
Since the arcs are not obtained at precisely the same time as the object
exposures there may be a wavelength shift relative to the arc dispersion
solution.  This may be calibrated from night sky lines in the object
itself (the night sky lines are "good" in this case and should not be
subtracted away).  There are generally not enough night sky lines to act
as the primary dispersion calibrator but just one can determine a possible
wavelength zero point shift.  Measure the night sky line positions at the same
time the object lines are measured.  Determine a zero point shift from
the night sky to be taken out of the object lines.
.NH
Method 2
.PP
This method is similar to the correlation method in that a template
spectrum is used and the average shift relative to the template measures the
radial velocity.  This has the advantage of not requiring the user to
do a lot of calculations (the averaging of the line shifts is done by
\fRidentify\fR) but is otherwise no better than method 1.
The template spectrum must have the same features as the object spectrum.
.IP (1)
Determine a dispersion solution for the template spectrum either from 
the lines in the spectrum or from an arc calibration.
.IP (2)
Mark the features to be correlated in the template spectrum.
.IP (3)
Transfer the template dispersion solution and line positions to an object
spectrum using one of the methods described earlier.  Then for the
current feature point the cursor near the same feature in the object
spectrum and type \fIs\fR.  The mean shift in pixels, wavelength, and
fractional wavelength (like a radial velocity without the factor of
the speed of light) for the object is determined and printed.  A new
dispersion solution is determined but you may ignore this.
.IP (4)
When doing additional object spectra remember to start over again with
the template spectrum (using \fI:read template\fR) and not the solution
from the last object spectrum.
.IP (5)
This procedure assumes that the dispersion solution between the template
and object are the same.  Checks for zero point shifts with night sky
lines, as discussed earlier, should be made if possible.  The systematic
centering bias, however, is accounted for by using the same lines from
the template radial velocity standard.
.IP (6)
One possible source of error is attempting to use very weak lines.  The
recentering may find the wrong lines and affect the results.  The protections
against this are the \fIthreshold\fR parameter (in Version 2.4 IRAF) and
setting the centering error radius to be relatively small.
.NH
Method 3
.PP
This method uses only strong emission lines and works with linearized
data without an \fBidentify\fR dispersion solution.  \fBIdentify\fR has
a failing when used with linearized data; it does not know about the
wavelength parameters in the image header.  This will eventually be
fixed.  However, if you have already linearized your spectra and wish
to use them instead of the nonlinear spectra the following method will
work.  The recipe involves measuring the positions of emission lines in
pixels which must then be converted to wavelength using the header
information.  The strongest emission lines are found automatically
using the \fIy\fR cursor key.  The number of emission lines to be
identified is set by the \fImaxfeatures\fR parameter.  The emission
line positions are then written to a data file using the \fI:features
file\fR colon command.  This may be done interactively and takes only a
few moments per spectrum.  If done interactively the images may be
chained by specifying an image template.  The only trick required is
than when proceeding to the next spectrum the previous features are
deleted using the cursor key combination \fIa\fR and \fId\fR (all
delete).
.PP
For a large number of images, on the order of hundreds, this may be automated
as follows.  A file containing the cursor commands is prepared.
The cursor command format consists of the x and y positions, the window
(usually window 1), and the key stroke or colon command.  Because each new
image form an image template does not restart the cursor command file the
commands would have to be repeated for each image in the list.  Thus, a CL
loop calling the
task each time with only one image is preferable.  Besides redirecting
the cursor input from a command file we must also redirect the standard
input for the response to the database save query, the standard output
to discard the status line information, and, possibly, the graphics
to a metacode file which can then be reviewed later.  The following
steps indicate what is to be done.
.IP (1)
Prepare a file containing the images to be measured (one per line).
This can usually be done using the sections command to expand a template
and directing the output into a file.
.IP (2)
Prepare the a cursor command file (let's call it cmdfile) containing the
following two lines.
.nf
	1 1 1 y
	1 1 1 :fe positions.dat
.fi
.IP (3)
Enter the following commands.
.nf
	list="file"
	while (fscan (list, s1) != EOF) {
	print ("no") | identify (s1, maxfeatures=2, cursor="cmdfile",
		>"dev$null", >G "plotfile")
	}
.fi
.LP
Note that these commands could be put in a CL script and executed using the
command

	on> cl <script.cl

.PP
The commands do the following.  The first command initializes the image list
for the loop.  The second command is the loop to be run until the end of
the image file is reached.  The command in the loop directs the string
"no" to the standard input of identify which will be the response to the
database save query.  The identify command uses the image name obtained
from the list by the fscan procedure, sets the maximum number of features
to be found to be 2 (this can be set using \fBeparam\fR instead), the cursor
input is taken from the cursor command file, the standard output is
discarded to the null device, and the STDGRAPH output is redirected to
a plot file.  If the plot file redirection is not used then the graphs
will appear on the specified graphics device (usually the graphics terminal).
The plot file can then be disposed of using the \fBgkimosaic\fR task to either
the graphics terminal or a hardcopy device.