With Hua Zhou (XSD-SSM) & Yun Chi (Visiting Student) to discuss use of RSMap with High Energy Surface Diffraction being done on Sector 1. This is a technique in development. One if the ideas here is to scan the growth of layers being deposited on a surface.
This is very similar to what is being done at Sector 33. The main difference is in the energy being used. Sector 33 uses lower energy (~15keV) while Sector 1 is higher (listed between 50-`50keV on web page). At lower energies, the Ewald sphere is smaller and intersects fewer diffraction peaks. At higher energies, the Ewald sphere is larger and therefore intersects more peaks. In a single image of a scan, lower energies will typically intersect only one or two diffraction peaks at a time. At higher energies, it is much more common to see peaks and that there are in fact multiple peaks per image. In the lower energy case these peaks are the lower order peaks (100, 110, 200, etc.) while higher energies yield the higher order peaks (500, 311, 511, etc.). One problem in this range is that peak intensities can vary over several orders of magnitude. It would be better here to use a lower noise detector for collection so that lower intensity peaks can be distinguished along with high intensity peaks. Hua has had opportunity to collect a few data sets with a low noise Pilatus. The results here were promising and he is hoping on possibility that APS will procure a similar/better detector. For typical work that he is doing now, he is using Sector 1 GE detector. This has much higher noise levels so only stronger peaks may be distinguished.
At present, the setup at sector 1 is not using a standard 4C/6C type diffractometer like Sector 33 & others. The setup is more adhoc but seems that it may still be describable like a diffractometer so that xrayutilities may still be a possibility. (Hua had a picture of the setup try to get slide) There is only one active motor in a scan and it is essentially the phi motor spinning the sample on its axis. Data is collected in a fly scan of this 'phi' motor. The scan is driven by spec but Hua thinks that only the phi motor is captured. Scans will generally scan only 90-180 degrees depending on symmetry of the sample. Typically 250 images are collected for each 5 degrees of scan giving about 9000 images for a 180 degree scan. This gives about 64GB / hr of collection.
Images collected are in HDF format (Need to confirm this) some information on angles is in the spec file (phi angle positions). Need to see what this looks like compared to step scan though. Assume that this could be treated similar to the 4C data in xrayutilities. Need to make sure that proper angle information for theta, two-theta and chi get into the system. Can these simply be added to the spec file? What to do with existing data if it is missing? Is the energy stored?
The present setup of the GE detector does not use the full surface. The detector is mounted in such a way that half the detector is below the sample & does not see reflected signal, rather it sees transmission ( through the substrate as well). Typically only 1/4 of the detector is used (not sure what happens to the other 1/4). I believe that the processing should be able to handle masking of this so that we only deal with the active portion. The current use is limited by a fixed mount for the detector as used by other experiments. If the detector could be moved to a moving mount, then more of the active surface could be used.
Will have to deal with only extracting the proper part as the file is read to save on RAM. As with other cases will need to make sure that we can handle the images in a piecewise fashion to stay within memory constraints. Need to look at final grid requirements. At this time, the final grid is stored in memory at one time. This requires 4(words qx, qy, qz, I) * 4(bytes/word) * gridsize. Hua believes the final grid max size would be something like 500x500x500. This requires ~ 2GB of ram so this should be doable. Will have to look more at what this means when this data is moved into Paraview. How far can they push things on a desktop before they require a visualization cluster like Cooley?
Hua indicated that he has also already had some discussions with Haiden Wen about possibilities of bringing in a laser to do time resolved studies in conjunction with the High energy diffraction work.
I mentioned to Hua, that I already have a backlog of work on RSM. It may take some time to start working but it is not too early to start gathering preliminary data that we will use for the first round. Hua will start sending pointers to datasets. This will give a chance to review what is missing, and possibly feed back to collection to make sure we have all of the information that we need.