Automated Critical Dimension Measurements (CD Measurements)

Critical dimension measurements in high-resolution microscopes are typically divided into two parts:

Image Acquisition
Image Evaluation

The image acquisition part consists of several elements, which may vary depending on the microscope type. Let's examine the typical workflow for a Scanning Electron Microscope:

CD Measurements with a SEM

Unlike light microscopes and atomic force microscopes, a Scanning Electron Microscope (SEM) covers a zoom range spanning several orders of magnitude -- for example, from 50x to 500,000x -- making it one of the most versatile types of microscopes. Additionally, a SEM can reveal quantitative material composition data using techniques such as EDX. However, here we will concentrate on the imaging aspect.

Image Quality

For SEM image acquisition, both image quality and precise positioning must be controlled. To obtain a good SEM image, both focus and stigmation need to be optimized. Furthermore, effects such as charging or sample degradation during imaging need to be controlled. This requires extensive control of the SEM via automation software to be able to select the best imaging conditions at each location and the same time achieve optimal working speed.

Autofocus/Auto Stigmation in SEM is challenging; under non-optimal stigmation, autofocus may converge on local minima, preventing optimal imaging conditions from being reached automatically. To address this issue, CorrMeas software uses a mathematical image analysis approach that can determine the optimal focus position even under suboptimal stigmation. This makes it possible to achieve high-quality imaging in a robust and automated way. Additional features, such as defining a tilted focal plane for the sample, can significantly improve acquisition speed.

To support scenarios involving varying imaging conditions -- such as different detectors, magnifications, or sensitivity to charging -- the CorrMeas state system allows users to interactively define complex acquisition sequences. These sequences can be taught step-by-step and later replayed as part of the automation process. This facilitates easy sequence testing and seamless integration into the overall automation workflow.

Positioning Accuracy

A target position can only be reached as accurately as the stage and other contributing uncertainties allow. In a typical SEM, stage precision may range from 2 to 5 micrometers for a stepper motor stage or as fine as around 100 nm for an encoder stage. Additional positioning errors arise from beam bending, structural deviations of the target, stage misalignment, and thermal expansion of the microscope and sample. In real-world conditions, positioning accuracy is often no better than about 500 nm with an encoder stage or 5 micrometers with a standard stage.

Still, in many cases, precise targeting is essential. The key to this is using local referencing. Reference data can be previously acquired images, vector graphics (e.g., GDS files), or other predefined structures. A high-precision positioning step generally involves:

Move to target location
Acquire an overview image
Correct the position via beam shift

Steps two and three may be repeated to achieve ultimate precision. Using this method, nanometer-scale positioning can be achieved even without a high-precision stage.

Having a high quality image from the exact target location is the optimal precondition for highly reproducible image evaluation results.

Image evaluation

Compared to a light microscope, a SEM can produce more artifacts. The grayscale in an SEM image depends on several factors: it is influenced by the material and it's work function, has topographic components, and is affected by local electric fields. Even when using the same detector, the same area can appear different under varying acceleration voltages and imaging currents. It is very difficult to predict how a structure will appear in an SEM image due to the many influencing parameters.

If a specific area needs to be analyzed, the simplest approach is to select appropriate imaging conditions and then optimize the parameters to ensure reliable image evaluation. The CorrMeas filter designer enables this process easily by allowing image analysis to be performed on the live image. This also makes it possible to test the robustness of the evaluation by checking how sensitive it is to small changes in imaging parameters.

Summary

One can see that hardware integrated automation for both image acquisition and analysis provides significant advantages over generalized image evaluation tools. Well-designed and easy-to-use built-in hardware control enables fast achievement of optimal working speed and robust evaluation results, thereby saving time and money.

Read further: CD Measurements with an AFM