Sang-Joon Cho, vice president and director of R&D Centre, Park Systems Corp, and Ilka M. Hermes, principle scientist, Park Systems Europe
Advances in lithographic processing are allowing for the production of ever smaller semiconductor devices, but nanometre-sized defects on wafer substrates can limit the performance of such devices. The detection and classification of these defects requires characterisation methods that deliver a resolution in the nanometre-range. Conventional automatic optical inspection (AOI) cannot achieve a sufficient resolution in the nanometre-range due to the diffraction limit of visible light, meaning that quantitative imaging and consequent defect classification is impaired. Automatic defect review with atomic force microscopy (ADR-AFM), however, visualises defects in three dimensions at the nanometre-resolution customary for AFM, therefore reducing uncertainty in defect classification and qualifying it as a more favourable technique for defect review.
Defect inspection and defect review
As semiconductor devices are becoming smaller and smaller in accordance with Moore’s law, defects of interest (DOI) are also decreasing in size. DOI are defects that potentially reduce the performance of semiconductor devices and are therefore of interest to process yield management. The decreasing size of DOI is a challenge for defect analysis; suitable characterisation methods must be able to image defects non-invasively at high lateral and vertical resolution in a two-digit or single-digit nanometre-range.
Traditionally, defect analysis in the semiconductor industry comprises two steps. In step 1 (called defect inspection), imaging methods that afford a high throughput but low resolution—such as automated optical inspection (AOI) or scanning surface inspection (SSI)— provide maps with the coordinates of defect positions on the wafer surface. However, due to their low resolutions, AOI and SSI provide insufficient information when characterising nanometre-sized DOI, thus necessitating the second step.
In step 2 (called defect review), high-resolution microscopy methods—such as transmission electron microscopy (TEM), scanning electron microscopy (SEM) or AFM—image smaller areas of the wafer surface to resolve the DOI using the defect coordinate maps from defect inspection. Utilising the coordinate maps from AOI or SSI minimises the scan area of interest and thus the measurement time of the defect review.
The electron beams of SEM and TEM can potentially damage the wafers, but AFM is able to scan surfaces non-invasively if a non-contact mode is employed. Furthermore, only AFM is capable of imaging defects at a high vertical resolution as well as a high lateral resolution. Therefore, AFM provides quantitative 3D information on defects required for reliable defect classification.
AFM
AFM achieves the highest vertical resolution among conventional imaging methods by mechanically scanning surfaces using a nanometre-sized tip on the end of a cantilever. As well as contact mode, AFM can be operated in non-contact mode, meaning the cantilever oscillates above the surface; here, changes in the oscillation amplitude or frequency provide information on the sample topography. Non-contact mode ensures non-invasive imaging of surfaces at high lateral and vertical resolutions.
Recent developments in automated AFM have meant that its application has spread from academic research to industry sectors such as hard disk manufacturing and semiconductor manufacturing. Industry has started to focus on the versatility of AFM and its ability to non-invasively characterise nanostructures in three dimensions. Hence, AFM is evolving into a next-generation inline measurement solution for defect analysis.
ADR-AFM
One of the biggest challenges of AFM-based defect review is the transfer of defect coordinates from AOI to AFM. Originally, the user manually marked defect coordinates using an optical microscope in an additional step between AOI and AFM, then searched for these coordinates in the AFM. However, this additional step proved time-consuming and lowered throughput significantly.
ADR-AFM, on the other hand, imports the defect coordinates from the AOI data. This requires an accurate alignment of the wafer as well as the compensation of stage errors between AOI and AFM. An optical surface analysis tool possessing higher position accuracy than AOI can reduce the stage error in a quick intermediate calibration step. The ensuing ADR-AFM measurement comprises a large-scale survey scan at the given defect coordinates, a high-resolution image of the defect and the defect classification. Since the measurement is automated, the user does not have to be present and throughput increases by up to an order of magnitude. ADR-AFM is used in non-contact mode to maintain the nanometre-range tip radius and high resolution for multiple subsequent scans. This prevents tip wear and ensures a quantitative defect review.
Comparing AOI and ADR-AFM
The results of a defect review using AOI and ADR-AFM on the same nanometre-sized defects are compared in figure 1. AOI estimates the size of the defects based on the intensity of scattered light, and ADR-AFM directly images defects by mechanically scanning the surface.
Figure 1
Left: A direct comparison table of six defect sizes determined using AOI and ADR-AFM (protruding defects are referred to as bumps and indenting defects as pits). Right: The corresponding AFM topography scans for all six defects in the table.
ADR-AFM measures the height or depth of the defects as well as their lateral size, and thus allows differentiation between protruding bump and indenting pit defects. The visualisation of the 3D shape of the defects ensures a reliable defect classification, which cannot be achieved via AOI.
If one compares the defect sizes determined via AOI and ADR-AFM, those estimated by AOI strongly differ from the ones measured via ADR-AFM. For bump defects, AOI consistently underestimates the defect size by more than half. This underestimation is especially evident for defect 4; AOI gave a size of 28 nm, roughly one third of the size determined by ADR-AFM at 91 nm. However, the largest deviations between AOI and ADR-AFM can be observed for measurements on pit defects 5 and 6; AOI underestimated defects at sizes in the micrometre-range by more than two orders of magnitude. The comparison of the defect sizes determined using AOI and ADR-AFM clearly shows that AOI alone is insufficient for the imaging and classification of defects.
Comparing ADR-SEM and ADR-AFM
As well as ADR-AFM, it is also possible to use ADR-SEM for high-resolution defect review. ADR-SEM conducts the automatic defect review based on DOI coordinates from the AOI data via an SEM measurement, during which a high-energy electron beam scans the wafer surface. Although SEM offers a high lateral resolution, it generally cannot provide quantitative height information on defects.
A number of wafer surfaces that have been imaged by ADR-SEM and then ADR-AFM are compared in figure 2. The ADR-AFM images exhibit a changed surface at, or near to, the ADR-SEM scan areas, these being visible as rectangles.
Figure 2
A comparison of wafer surface imaging using ADR-SEM and ADR-AFM (the ADR-SEM scan areas are visible as rectangles on the ADR-AFM images). a) An ADR-AFM image of a bump defect previously missed by ADR-SEM. b) An ADR-AFM image of a defect at a height of 0.5 nm (left), which is not visible in the ADR-SEM image (right). c) Examples of surfaces damaged by the electron beam during ADR-SEM measurements.
Figure 2a shows the improved visibility of the ADR-SEM scan area in ADR-AFM, zooming in on a bump defect missed by ADR-SEM, located just above the ADR-SEM scan area.
Figure 2b shows that ADR-AFM, due to its high vertical resolution, afforded sufficient sensitivity to detect surface defects at a height of 0.5 nm. These defects could not be imaged by ADR-SEM because of its lack of vertical resolution.
Lastly, figure 2c shows damage caused to the wafer surface by the high-energy electron beam in ADR-SEM. The ADR-SEM scan areas can be recognised in the ADR-AFM images as rectangles surrounding the defects. In contrast, non-invasive imaging and high vertical resolution make ADR-AFM ideally suited as a characterisation technique for defect review.
Conclusion
The manufacture of ever-smaller semiconductor devices has led to AFM gaining importance as a high-resolution and non-invasive analysis method for defect review. The automation of AFM measurements has simplified and sped up the previously time-consuming workflow for AFM in defect characterisation. The progress in AFM automation was the basis for the introduction of ADR-AFM, in which the defect coordinates can be imported from prior AOI measurements and the subsequent AFM-based characterisation does not require the presence of a user. ADR-AFM therefore qualifies as an inline methodology for defect review.
ADR-AFM complements conventional AOI, especially for defect sizes in the one- or two-digit nanometre range, since the high vertical resolution of AFM facilitates a reliable and 3D defect classification. Non-contact mode ensures non-invasive surface characterisation and prevents wear of the AFM tip, thus guaranteeing that a high resolution is maintained in numerous consecutive measurements.
Park Systems Europe