This example shows an advanced application of ProSEM Formulas, measuring field-to-field stitching error or layer-to-layer overlay error, an important metric of lithography processes. If you are not familiar with the basics of formulas, it may be useful to first read the prior examples demonstrating formulas. (links)
A typical use for this stitch formula is to measure a set of SEM images captured at different locations, with the output being one X and one Y stitch error measurement for each image / location, so this example will utilize ProSEM's Recipe and Summary Table features as well, to provide an automated measurement and data summary.
A package is available for this application example which contains sample images and necessary formulas scripts, as well as a ProSEM Recipe and a completed ProSEM Project. This package can be downloaded here:
Note: ProSEM 2.3.0 or greater is required for these formulas.
There are multiple different patterns and methods of measuring field stitch or layer overlay in lithography, such as verniers or box-in-box structures. There are even dedicated tools available designed specficially for this task. Here, a grating method is used to measure field-to-field stitching using ProSEM.
The stitch measurement is achieved by setting up 4 separate grating measurements, one for each quadrant of the pattern. The formulas will then compute the stitch error by analyzing the four measured gratings.
The scripts in this example require:
A detection fence will define the region of the image in which to search for other members of this grating. In this application using the detection fence causes the find operation to find only the elements in the upper left quadrant, but not those in the upper right quadrant.
By default, the fence positions are at the edges of the image.
Alternatively, the fence positions can be defined by entering the fence positions at the bottom of the Feature Detection panel. The positions are expressed as a percentage of the image, from the lower left corner of the image. In this application, the fences are defined to be somewhat larger than the actual quadrant, to allow for some position shift between SEM images of different stitch/overlay patterns. The detection fence definition is stored with each measured feature.
When all four quadrants have been measured, the image looks similar to:
The stitch error is evaluated for every feature in the image; this is the function of formulas within ProSEM. However, due to the definition of the formula, the same value will be calculated for all features in each image. These values can be seen for each meausred line in the Lines & Spaces data table. However, a Summary Table can be used to provide a more convenient table of just the measured stitch error for each image.
The Summary Table now shows the X and Y stitch errors for each image, although in this example only one image has been measured.
The Summary Table has been defined here, though, so that the Recipe to be created in the next section will already contain the summary. In that way, the same Recipe can be used on other image sets without the need to seperately add the summary.
If still viewing a Summary Table, switch back to ProSEM's SEM Image tab.
Select the image name in the Images Panel. The image name must be selected so that the recipe to be created will contain all four groups necessary for the stitch measurement.
Click the Set as Recipe button.
With a Recipe define, several actions are possible:
To use this recipe to measure other images in this project:
This Recipe can also be Exported for use on other projects of stitch images.
Once this recipe has been successfully run on other images, the resulting summary table looks like:
This is the goal of this example - a table of the measured stitch error for each image in the project. This table may be exported into a CSV file.
When using these Stitch/Overlay Formulas on multiple images, the degree of automatic success varies. The more the images vary from the template image in position, rotation, brightness, or contrast, the likelihood increases that some manual adjustment will be necessary.
Note the requirements listed above, specifically:
In case the recipe find does not find a feature match for one or more of the feature groups, it may be necessary to manaully adjust any failed measurements, and use Find Similar to find the rest of the grating. In these cases, it will be necessary to manually delete the "Lines & Spaces" group after the four gratings have been found.
In the case that one or more individual features within a grating are not found, it will be necessary to manually adjust the features and manually find and store the missing measurements. Those manual measurements should then be moved into the corresponding grating group in order for the Stitch formulas to function properly.