MLx Home | Contents | MLx menu | MLx buttons | Widgets | Index | What's New | Multivariate Button

Superconductor Precursor Example

These sets of images are small (64x64) quantitated (JEOL) electron probe maps of ceramic superconductor precursor material. The material has several distinct phases which are seen in the maps, and more clearly seen in color overlays and scatter diagrams.


Load the images.

  1. MLx -> Image Files -> TIFF -> Read Multiple
    1. Unlike most other image processing programs, MacLispix must be told what type of file it is to read.
    2. The images can be loaded separately using MLx -> Image Files -> TIFF -> Read, but it is much easier to handle them as a group.
  2. Find Demo_Images / Super cond maps/ Set 2/ *.tiff
  3. Select one of the tiff files as a sample. A dialog showing all files in the folder with the same extension.
  4. In the next dialog, select all of the images except the total image by:
    1. Click the Select All button.
    2. Command-clicking the total.tiff file.
    3. Click the Load button.
  5. The postage stamp sized images will be stacked on top of each other and will look like this:
    1. Only one of the maps is visible
    2. The window is too small to read their names.
    3. Selecting the W menu with the option key held down will show all of the images in a menu. Selecting one of them will bring that image to the front, thereby selecting it. Lines will appear in the title bar.
    4. Information on the windows can be listed in the Monitor window. Get the Mlx Buttons window using the MLx -> Buttons menu. The Info... button in the Images collumn has two options. Select one, then click on the Monitor window to see the results.
      1. Dimensions

 The lines on top of the list are output of the image load ing function listing the images. The list of images has the

  1. name - This is the window title which can be changed. Second and third characters in the name denote the element, eg. "CPB110_2" denotes Lead. The initial C stands for Concentration (Wt.% X 100)
  2. type of image - These are integer images - unsigned, 16 bits for each pixel. Values can range from 0 - 65535.
  3. dimensions of the image. These are 64 pixels wide and 64 pixels high.
  4. memory - The image data arrays take 8 kilobytes each. Each image requires more memory than this for other things.
  5. window type - This class name describes the function of the window.
  6. file name - the entire file name for the image (if the image was loaded from a file, as in this case).

  (last part of image file names not shown)

 

 Selecting all of the images in the dialog (upon invoking the limits command) and then clicking on the Monitor Window (and perhaps scrolling it) gives this printout.

Note that the printout is in a text window, and could have been cut and pasted to this HTML file as text rather than as an image, but I wished to show exactly what the printout looks like.

Note, for later processing, that all of the images go to zero somewhere, and that all of them have values greater than 255, so that they must be scaled to be displayed.

The file names are printed in this list just as they are in the dimensions list above.

 


Viewing the Images

Here are a couple of convenient ways to view the images. The small version keeps the original size, the large one, for larger screens, zooms the images.

Small View

  1. Tile the images so that they can all be seen:
    1. MLx -> Buttons Pops the button window if it is not already visible.
    2. Arrange...-> Tile Arranges the images left to right at the top left of the screen.
    3. If the images are covered with the Monitor Window...
      1. Drag the monitor window out of the way .. or..
      2. Option-click the montor window title bar to send it to the rear (or bottom) ...or...
      3. Pop the image windows: Arrange...-> Pop
  2. The images are too small for their names to appear in the title bar. Make the titles appear as small annotations at the top of each window: Annotate...->Titles ->Notes, Select All.

 Immediately after tiling, none of the images is in front, so all of the title bars are white.

 
 If the images above are clicked on, left to right, in turn and then tiled again, their order is reversed. They are tiled in 'reverse' click order, ie. last clicked first, as this image is in front.  

Large View

  1. Zoom the images. Zoom... (in Images collumn), Select All images, Select 'Other' for the zoom factor, and enter 1.5 .
  2. Tile them, as above. Arrange...->Tile.

 Note that the images are not large enough to read the entire titles, but enough is visible to identify each image.
 


Color Overlay (also see Kowala example)

Although the regions of different phases (compositions) can be determined by examining the images separately, RGB color overlays make the regions stand out at a glance (Bright 1990, 1991)

  1. Other Windows -> Multivariate Makes a new button window for the following steps:
  2. Select the red, green and blue images
    1. X/Red Image -> CBI... Choose the Bismuth image to be shown in red.
    2. Y/Green Image -> CCA... Choose the Calcium image to be shown in green.
    3. X/Red Image -> CBI... Choose the Copper image to be shown in blue.

 The upper left part of the Multivariate Buttons window will then look like this. The image names replace the original button names.

The RGB overlay requires three gray level images of the same dimensions (but not necessarily of the same pixel type or intensity range. The overlay is made from the scaled images, which are always unsigned 8-bit images.) After selecting one image, only those images of the same size as the other two will be presented for selection.

 

  1. Make the RGB Overlay... RGB Overlay...->make
  2. Zoom the overlay by ...
    1. Using the zoom button as before
    2. Dragging the grow square.

 This overlay was zoomed by x1.5 using the zoom menu in order to make it the same size as the others.

There appear to be four phases present.

 


Scatter Diagrams

Another way to see how many phases there are, and quantify them as well, is to use scatter diagrams (see introduction). Using the three images already selected for the color overlay (also see Kowala example)

  1.  2d xy...->normal
  2. Click on the slider to 'thermalize' the image as desired.
  3. Axes makes the axes shown the concentration values (in this case) x100 for the clusters. The clusters correspond to the different phase.

To see what regions of the images correspond to the features of this scatter diagram, see traceback 1.

 

  1.  3D xyz...->normal
  2. Rotate the cube window to some orientation similar to this one by dragging the mouse across it.

  1. X, Y and Z are Bi, Ca and Cu respectively (see the ccch title bar).
  2. Axis labels are not available for the three - dimensional scatter diagram.
 


2-D Traceback #1

The traceback command makes a color labeled image (a traceback map) corresponding to the original images. The color for any pixel is the (automatically assigned) color of the parallelogram shaped region that encloses the corresponding point in the scatter diagram.

  1.  Select regions in the scatter diagram using the traceback...->normal button.
  2. A pink info line will appear across the top of the scatter diagram.
  3. For each parallelogram
    1. Drag one side.
    2. (mouse up)
    3. Drag that side parallel to itself to 'fatten' it into a parallelogram
  4. Click the small square 'go away' button on the pink line when done.

Note: when making the parallelograms, you may drag off the image. That is ok. Regions outside the image are ignored.

 

 After closing the pink info line, the traceback mask (zoomed here) will appear.

Note how the regions (but of course not the colors) roughly correspond to the pastel colored regions in the color overlay.

 
 This is a 'palette' window. It labels the colors of the parallelograms in the scatter diagram and the colored regions in the mask. Option-clicking a color or a label allows you to change it.  


Traceback - Three Dimensional

Traceback for the three-dimensional scatter diagrams is identical to that of the two dimensional diagrams. Regions are selected by 'mousing' parallelograms. The selected points appear inside the parallelogram -- physically, they are in a parallelopiped seen end-on. I have not implemented graphics to show side views of the selection region by rotating the scatter diagram. Once the diagram is rotated, the selection regions are not valid.

Try different views of the 3-D scatter diagram to make sure that unwanted bins are not selected because they are along the line of sight of desired bins.


About the Data

The data were taken by Ryna B. Marinenko, NIST, Gaithersburg, Md. and by Slavko Bernik, currently at the Jozef Stefan Institute, Ljubljana, Slovenia. (see references by Marinenko and Bernik)

The pixel values are weight percent x 100 (the images are 16 bit integers), except for the total map which is percent x 10.

Dwell time: 2 secs/pixel
Field of view: 50µm on a side, roughly. 
Magnification: 1700x mag. - expect some defocusing.
Beam: 70 nanoamps, 20 Kv.

Sets 1 and 2 - mixed phases .Set 3 - appears to have only one phase - the desired superconducting phase, The broadening of the ratio of (Bi+Pb)/(Sr+Ca) is probably due to self shielding in the e-probe.


References

D. S. Bright, R. B. Marinenko, S. Bernik "Identification of Multiphase Systems Using Compositional X-ray Maps", pp. 203-4 in MICROBEAM ANALYSIS, Proceedings of the 28th Annual MAS Meeting, John J. Friel ed., VCH NY, NY. 1994

Marinenko, R.B., Bright, D.S. and Bernik, S. (1996) "Multiphase Analysis of Bi-Sr-Ca-Cu-O High Tc Superconductors with X-Ray Compositional Mapping", SCANNING (18): 395-400

D.S.Bright (1990),"SOFTWARE TOOLS FOR EXAMINATION OF MICROANALYTICAL IMAGES", Microbeam Analysis 1990:73-78.

Bright, D.S. and Newbury, D.E. (1991) "Concentration Histogram Imaging", Analytical Chemistry 63(4):243A-250A, (Feb. 15) 1991