Procedures <Prev Next>

Calibrating Spectra

PGOPHER can be used to calibrate an experimental spectrum from anything displayed in the simulation window; a line list or a simulation of a known spectrum are likely to be the most useful sources. The following built in calibration sources are available (see File, New, Calibration):
  1. The visible B-X absorption spectra of I2. The line positions for this are calculated from the constants given in F. Martin, R. Bacis, S. Churassy and J. Verg J. Molec. Spectrosc. 116, 71 (1986) with the Franck-Condon factors estimated from RKR curves generated from these constants. Checks against high accuracy measurements in H. Knockel, B. Bodermann, and E. Tiemann, Eur. Phys. J. D 28 199 (2004) indicates a maximum error of 0.043 cm-1, but < 0.02 cm-1 for the v" < 11 transitions typically used for calibration. The experimental Doppler limited absorption spectrum of the same transition published on-line in electronic format by Salami and Ross (J. Molec. Spectrosc. 233, 157 (2005), doi:10.1016/j.jms.2005.06.002 and the Ohio State University Molecular Spectroscopy Archives, http://msa.lib.ohio-state.edu/jmsa_hp.htm) can also be loaded directly into PGOPHER and used for calibration, provided the overlay is inverted. (Right click on the spectrum after loading and select "invert").
  2. Ne and Fe atomic lines with positions and intensities taken from the NIST Atomic Spectra Database (version 3.0.3), Yu. Ralchenko, F.-C. Jou, D.E. Kelleher, A.E. Kramida, A. Musgrove, J. Reader, W.L. Wiese, and K. Olsen http://physics.nist.gov/asd3 [2006, June 18]. National Institute of Standards and Technology, Gaithersburg, MD.
  3. A set of line positions for use with Ne optogalvanic spectra, commonly used for calibrating pulsed dye lasers. The lines are taken from "An atlas of optogalvanic transitions in Neon" RAL report, RAL-91-069 (1991) by S.H. Ashworth and J.M. Brown, with Ne lines from the NIST database (see above) and a few points added from our own measurements.
If you have an etalon trace recorded with the data, then follow the procedure in Using Etalons to produce a linear frequency scale, and then use (if necessary) the procedure below to calibrate the resulting scale.

1. Load Experimental data and Calibration source

First load the experimental spectrum and set up the reference source for the calibration, so the simulation window looks something like the picture below; see Overlaying Files for how to do this. Note that both spectra should have upward pointing peaks; use Overlays,Invert if necessary to invert your spectra

In this case the spectrum is atomic lines, and the upper trace is an atomic line list.

2. The Calibration Window

Use Overlays, Calibrate (or right click on the experimental spectrum and select calibrate) to bring up the calibration window, and make sure the experimental spectrum to be calibrated is selected in the box at the top of the dialog:

Important: While calibrating, there are two horizontal scales that need to be considered:
As you start calibrating, the plot will always be of the new scale, but peak measurement and display changes for overlay channels as follows:
In addition, while the calibration window is open, Alt+drag will move the selected experimental spectrum rather than the simulation. Make sure the calibration window is closed when you have finished calibrating or you will be using an uncalibrated scale for peak measurements.

3. Initial Alignment

The traces need to be roughly aligned to start with, at least so the assignment to the reference spectrum is clear. If the experimental spectrum has an approximately correct scale, then an offset may be sufficient - alt + drag in the main plot window, provided the calibration window is open. In this mode, the Offset in the main window should be set to zero; do not use it to shift the spectra.

If you know the approximate limits of the experimental spectrum:
  1. Select "Other", "Set Range". This will add two measured "peaks" at the very start and end of the spectrum, and they will appear in the calibration window. (They are set with a large standard deviation of 1000, implying a large uncertainty in their values, so any subsequent assignment of calibration peaks with the default standard deviation of 1 will essentially ignore these points)
  2. Enter the corresponding frequencies in the "Actual" column for these two points.
  3. Ensure the polynomial value is 1, and press "Fit".
Alternatively, it is possible to manually edit the FrequencyOffset and FrequencyScale of the Experiment overlay. You will need to hit the simulate button in the main window to update the plot. These should be reset to their default values (0 and 1 respectively) if you use the steps below, or the calibration function will be applied twice.

4. Assigning Peaks

To use specific peaks in the spectrum for calibration:
  1. Right click and drag across an experimental peak; the peak position is measured and appears in the first column of the "Calibrating" form.
  2. The "Actual" box next to it will turn red to indicate the actual position to be filled in.
  3. Right click and drag across the corresponding peak in the calibration spectrum; the position of this will be filled in the square indicated above.
  4. Repeat as required - you will only need a few peaks in the initial stages.
To correct mistakes:

5. Fitting

Given some assignments, press the "fit" button. The residuals will be filled in and the experimental plot will be adjusted to reflect the newly fitted frequency scale. The residuals will be plotted in a separate window.
You will probably want to add more peaks - go back to step 4. If you manually set FrequencyOffset and/or  FrequencyScaleabove, you should now set them back to their default values (0 and 1 respectively).

6. Transferring the Calibration

Once you have a suitable calibration, you can transfer the calibration to another spectrum using one of the following entries on the "Other" menu:
In each case your are prompted to select the spectrum to apply the calibration to; select "all" the default to apply it to all overlays. Note that transferring a calibration of a spectrum onto itself (or selecting all) has the effect of removing the original frequency scale.