![[Athena logo]](../../images/pallas_athene_thumb.jpg)
Sadly, most data that we measure is, frankly, kind of crappy. A single scan typically ranges from noisy to so dominated by noise that it's hard to recognize it as oscillatory. Fortunately, the central limit theorem works every single time. If you add up enough noisy scans, it eventually turns into real data!
All entries in the group list which are marked will be included in the merge. Make sure that the marked groups really are the ones you want to include in the merge!
Data can be merged as μ(E) , normalized μ(E) , or χ(k) . In most cases, it makes little difference which one you choose. χ(R) will be the same regardless. (This can be tested, in fact, I encourage you to do so.) The standard deviation in the merged data is also computed and stays with the data in project and certain output files, although ATHENA does nothing else with the standard deviation other than plotting it, which is shown below.
By default, each data set is weighted the same in the merge. You can specify the weight to use for each data set using the «importance» parameter, which is 1 by default for ever group. Alternately, you can let IFEFFIT determine the amount of noise in the data by looking at the signal between 15 and 25 in R-space and then use the noise to weight the data. In that way, cleaner data contributes more to the merge. The choice between using the noise or the «importance» is made using the check buttons in the menu.
The contents of the Merge menu, showing the various ways that data can merged.
Data merged as μ(E) plotted with its standard deviation.
The data merging feature is one of ATHENA's most important. My typical work pattern is to carefully align scans on a sample, merge them, then analyze the merged data in ARTEMIS.
It is essential that your data be well-aligned before merging. Merging poorly aligned data has the effect of significant attenuating the fine structure.
When merging as μ(E) or normalized μ(E) , if every data group included in the merge has a reference channel, the reference channels will also be merged and these two merged spectra will be tied together as a reference channels. Thus the merge of the reference channels can serve as the alignment standard for the merge of the data.
There is one situation where merging as μ(E) will be substantively different from the other two options. If something happens to change the values of the signals on the detectors in such a way that μ(E) for subsequent scans is of very different sizes but the normalized μ(E) data is consistent for all the data, then the merge in μ(E) will be dominated by the scans with larger absolute values while the merge in normalized μ(E) will treat all the data with similar weights. One way of having this happen is to change amplifier gains between scans. In that situation, you probably should choose to merge normalized or χ(k) data.
