A specimen-drift-free EDX mapping system in a STEM for observing two-dimensional profiles of low dose elements in fine semiconductor devices

Abstract

We developed a specimen-drift-free energy-dispersive X-ray (EDX) mapping system in a scanning transmission electron microscope (STEM) to improve the sensitivity and spatial resolution of EDX elemental mapping images. The amount of specimen drift was analysed from two STEM images before and after specimen drift by using the phase-correlation method, and was compensated for with an image-shift deflector of the STEM by the displacement of the scanning electron beam. We applied this system to observe the two-dimensional distribution of low dose arsenic in silicon semiconductor devices. The sensitivity of the elemental mapping was improved to several tenths atomic % for arsenic atoms while maintaining a spatial resolution of 2 nm.

Towards an adaptive minimum variance control scheme for specimen drift compensation in transmission electron microscopes

“…. , 59, it was found using ARMAsel that the de-trended y-direction drift can me modeled with an ARMA(2,1) model, with transfer function Figure 4 also presents the results of the standard minimum variance control algorithm, which assumes perfect knowledge of the transfer function in (8). Clearly, the drift reduction is significant with both techniques, even when strong measurement noise is present.…”

“…This is done by using image processing techniques such as phase-correlation or cross-correlation [8], [12] (different techniques are needed to measure drift in the zaxis direction [17], [20]) . Since the interest here is on d(k), the known value r b (k) = u(k) could be subtracted from the aforementioned displacement to produce the drift estimate, d(k), given byd…”

“…Although these techniques yield good results, they are not able to prevent the blurring of long-exposure images. Feedback control schemes without the necessary prediction capabilities (see, e.g., [8]), present a similar disadvantage.…”

Stability and performance analysis of dual-random-rate systems via Markov jump linear system theory

“…That is, its dynamics are slower than plant's time constants (see e.g., [2], [12]). On the other hand, a sporadic sensor is one that has fast dynamics but can be used only infrequently compared to the plant's actuation rate (e.g., those used for image drift control [6]). This leads to the following definition (see also [12], [13]).…”

“…These applications range from robot or industrial process control [1]- [3], to biology [4], medicine [5], and microscopy [6], [7]. A common property of these so-called image-based control systems is that, due to the need of acquiring and processing images for feedback purposes, their sensing (or measurement) rates are generally slower than their actuation rates.…”