path: root/noao/onedspec/doc/sys/identify.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
August 1986
Revised August 1990
.AB
The IRAF task \fBidentify\fP 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
.sp 3
.NH
\fBIntroduction\fP
.PP
The task \fBidentify\fP 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\fP 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\fP 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\fP 
task.
.PP
The task \fBidentify\fP 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.
.PP
This paper is specifically about \fBidentify\fP but one 
should be aware of the task \fBsplot\fP which also may 
be used to measure radial velocities.  It differs in 
several respects from \fBidentify\fP.  \fBSplot\fP 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\fP.  The 
primary source is the manual pages for the task.  As 
with all manual pages it is available online with the 
\fBhelp\fP command and in the \fIIRAF User Handbook\fP.  
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\fP 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\fP.  After determining the dispersion solution for
the first arc and quitting (\fIq\fP key) the next arc will be displayed
with the previous dispersion solution and lines retained.  Then use the
cursor commands \fIa\fP and \fIc\fP (all center) to recenter and
\fIf\fP (fit) to recompute the dispersion solution.  If large shifts
are present use \fIs\fP (shift) or \fIx\fR (correlate peaks) to shift,
recenter, and compute a wavelength zero point shift to the dispersion
function.  A new dispersion function should then be fit with \fIf\fP.
These commands are relatively fast and simple.
.IP
An important reason for doing all the arc images first 
is that the same procedure can be done mostly noninteractively 
with the task \fBreidentify\fP.  After determining a 
dispersion solution for one arc image \fBreidentify\fP 
does the recenter (\fIa\fP and \fIc\fP), shift and 
recenter (\fIs\fP), or correlation features, shift, and 
recenter (\fIx\fP) 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\fP.  
For each image begin by entering the colon command 
\fI:read arc\fP 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\fP and \fId\fP (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\fP colon command.
.NH 2
Measuring Wavelengths
.IP (1)
To record the feature positions at any time use the \fI:features 
file\fP colon command where \fIfile\fP 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\fP 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\fP.  The \fI:features\fP 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\fP must be changed.  This may be done 
interactively with the \fI:ftype emission\fP and \fI:ftype 
absorption\fP 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\fP.  To change this parameter type \fI:fwidth value\fP.  
The positions of the marked features are not changed until a 
center command (\fIc\fP) command is given.
.IP
A narrow \fIfwidth\fP is less influenced by blends and wings but 
has a larger uncertainty.  A broad \fIfwidth\fP 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\fP 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\fP 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 identify) 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\fP.  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\fP) 
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\fP 
parameter 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\fP dispersion 
solution; though remember the caveats about rebinning the
spectra.  The recipe involves measuring 
the positions of emission lines.  The 
strongest emission lines may be found automatically using 
the \fIy\fP cursor key.  The number of emission lines to 
be identified is set by the \fImaxfeatures\fP parameter.  
The emission line positions are then written to a data file 
using the \fI:features file\fP 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 that when proceeding to the next spectrum the previous 
features are deleted using the cursor key combination \fIa\fP 
and \fId\fP (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 from 
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 a cursor command file (let's call it cmdfile) 
containing the following two lines.
.RS
.IP
.nf
.ft CW
1 1 1 y
1 1 1 :fe positions.dat
.ft P
.fi
.RE
.IP (3)
Enter the following commands.
.RS
.IP
.nf
.ft CW
list="file"
while (fscan (list,s1) !=EOF){
print ("no") \(or identify (sl,maxfeatures=2, cursor="cmdfile",
>"dev$null", >G "plotfile")
}
.ft P
.fi
.RE
.LP
Note that these commands could be put in a CL script and executed 
using the command
.sp
.IP
.ft CW
on> cl <script.cl
.ft P
.sp
.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\fP 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, the graphs will appear on the specified graphics 
device (usually the graphics terminal).  The plot file can then 
be disposed of using the \fBgkimosaic\fP task to either the  
graphics terminal or a hardcopy device.