Scanning electron microscope metrology: harder than you might think
Since its inception in the 1950s and its introduction as a commercial product in the 1960s, the scanning electron microscope (SEM) has become an indispensable tool for many and diverse applications. Moreover, over time the character and capabilities of these instruments have improved enormously. Electron sources have evolved from tungsten to lanthanum hexaboride to cold and thermal field emission, providing much higher instrument performance. Performance is also enhanced by the incorporation of new electromagnetic and electrostatic lens designs, as well as digital electronics. Today, additional operational enhancements such as automation, autofocusing, and autoastigmatism correction, as well as digital imaging, make the core microscopy functions nearly transparent to the user and the instrument itself deceptively easier to operate. But ease of operation can also foster operator complacency. In addition, user-friendliness has reduced the ‘apparent’ need for more thorough operator training. This overall attitude has fostered the concept that the SEM is just another expensive digital camera. Consequently, a person using the instrument may be lulled into thinking that all of the potential pitfalls have been eliminated and that everything they see on the micrograph is always correct. However, this may not be the case.
The SEM provides exciting imaging (see Figure 1) and a superb ability to measure very small structures. It is likely that one of the first questions asked even before the first scanning electron micrograph was ever recorded was ‘How big is that?’. The answer has improved substantially in recent years, thanks especially to research carried out to understand the fundamental mechanisms involved in imaging and making accurate measurements with these instruments. Today, SEMs are routinely used as the primary measuring tool on semiconductor processing lines to monitor manufacturing processes. At the National Institute of Standards and Technology (NIST), we have performed a great deal of work over 20 years to understand SEM instrument performance, instrument calibration, imaging, and metrology. In a paper published elsewhere,1 we describe some of these efforts and the potential issues related to signal generation in the SEM, instrument calibration, electron beam interactions, and the need for modeling2 to understand the actual image and its formation. For example, we present the NIST dimensional calibration standard RM 8820 along with earlier data indicating the degree of miscalibration found through an extensive interlaboratory study. In addition, we discuss the multiple sources of signal generation in the SEM and their contributions to measurement uncertainty. We have summed up all this research and standards development together in a discussion of how these issues effect every measurement made with the instrument.
An additional topic, impacting directly on the measurements, is the problem of beam-induced specimen contamination. In a second paper,3 we discuss the detrimental effect of specimen contamination and how it can be eliminated. Avoidance of contamination deposition (and its detrimental effects on imaging and measurements made with the SEM) is something that every user should know and understand before attempting any critical quantitative work. Sample contamination contributes to the total uncertainty of any measurement and must be considered in any uncertainty statement citing the accuracy of any measurement. Furthermore, especially for nanometer-scale imaging and measurements, electron-beam-induced contamination can seriously hamper or prevent work, as the few-nanometer-sized objects of interest are readily obscured under a layer of carbonaceous material. As a result of work carried out in cooperation by NIST and two small US companies, contamination-free imaging and measurements are now possible, and the paper describes the procedures.3
Proper instrument calibration, appropriate choice of operating parameters based on the knowledge about electron beam and specimen interaction, formation, modeling of the SEM image, and elimination of sample contamination are all important for accurate SEM imaging and measurement. This is especially true for semiconductor critical dimension control. Similarly, dimensional metrology is becoming more important in areas of nanotechnology and nanomanufacturing, where huge quantities of nanomaterials (such as carbon nanotubes or cellulose nanocrystals) need to be inspected, characterized, and measured to meet customer specifications. We are also researching additional components of measurement uncertainty, such as sample vibration, and will report on them at a later time.
Michael T. Postek, PhD is currently the senior scientist in the Semiconductor and Dimensional Metrology Division of the Physical Measurement Laboratory at NIST. He has been involved in SEM metrology for over 30 years and has written a book and over 200 publications in this area.
Andras E. Vladár, PhD is the leader of the Scanning Electron Microscope Metrology Project at NIST. He is an expert in SEM-based dimensional metrology and one the best-known research scientists and a technical leader of this field.