An Introduction to the Automation of AFM

For many years, the dimensions of semiconductor devices have been moving towards 1X-nm nodes and below. This reduction in dimensions has been observed year-on-year and is the result of market demand for faster and more efficient designs.

Improvements in device manufacturing methods have occurred, ranging from 65 nm in 2006, down to the 1X node at 14 nm in 2014. The International Technology Roadmap for Semiconductors predicts that manufacturers will have the first sub-1X node devices, measuring 7nm, as soon as 2017 [1].

To continue at this rate, manufacturers must be able to meet metrology requirements that simultaneously call for enhancements in resolution, precision, and accuracy. Manufacturers will need tools that provide nanoscale imaging for critical dimension measurement. Repeatable and accurate results are also needed to enhance productivity in a mass production environment [2].

Park Systems has created a solution to these challenges that relates to nanometrology: atomic force microscopy (AFM). This tool makes the analysis of data simple, providing both customization and automation.

The Dubbed XEA [3] software is designed to enable a process control engineer to use AFM systems. The system is used to obtain precise and repeatable nanoscale images of target devices according to user-defined procedures in custom recipe files. The corresponding increase in productivity makes this combined software and hardware solution attractive wafer-level device fabrication plants everywhere.

Experimental

The experiment was conducted using a three-inch patterned silicon wafer sample. The target pattern on the wafer consists of rectangular pits which have pitches of 10μm and step heights of 120 nm. Analysis of the topography and roughness of the target pattern was also conducted on two pits. To perform the imaging, the sample was mounted onto the stage of a Park NX-HDM AFM system [4] using XEA software. A silicon-based cantilever allowed the AFM imaging in air in non-contact mode. Five measurement locations were selected inside the wafer as shown in Figure 1a:

The five measurement sites selected for imaging from the sample patterned silicon wafer.

Figure 1a. The five measurement sites selected for imaging from the sample patterned silicon wafer.

The reference measurement location (site #5 on Figure 1a) for creating the automated recipe was located at a NOR Gate device on the wafer. Four additional measurement sites of the same patterned device were selected at different XY coordinates inside the wafer to execute automated measurements of sample feature topography and roughness.

The first step in creating the measurement and analysis recipe is to put XEA software through a pair of teaching routines. This allows it to optically recognize where the cantilever and sample pattern are located. Figures 1b and 1c show these teaching routines, which consists of optical image and adjustments of the XY stage. To allow the hardware to bring the target pattern measurement site to the tip landing position, user-taught data is fed into the software.

Optical camera feed from the XEA software interface displaying a 95% confidence level in the software’s ability to recognize the sample pattern.

Figure 1b. Optical camera feed from the XEA software interface displaying a 95% confidence level in the software’s ability to recognize the sample pattern.

Optical camera feed from the XEA software interface displaying a 95% confidence level in the software’s ability to recognize the cantilever. The red crosshair is a user-defined estimate of the cantilever tip’s location. </h6

Figure 1c. Optical camera feed from the XEA software interface displaying a 95% confidence level in the software’s ability to recognize the cantilever. The red crosshair is a user-defined estimate of the cantilever tip’s location. </h6

Furthermore, the initial reference scan image of the sample pattern is used to adjust XY offsets in positioning when required.

For example, if a target pattern for the study is not consistently optically recognizable by the software, a user can define the pattern’s location relative from a nearby landmark that is. Then the recipe would be designed to automatically move the sample to the cantilever tip accordingly, using the landmark for orientation. This is performed with nanometer-level precision in order to accurately image the user-defined region.

When the teaching routines are completed and XY offsets are determined, the user then defines the scan parameters to obtain sample topography and roughness data.

Automated measurements on the sample are possible after the entire set of calibrations and instructions have been compiled into the recipe file. Immediately after each site is measured, the XEA recipe - which has been scripted to conduct an automatic analysis of the collected data right after measurement - completes. After all measurement sites have been investigated and the recipe has been exhausted, an aggregated report of the data acquired at each measurement site is made available to the user.

Results and Discussions

After scanning was finished, five user-defined measurement sites were accurately imaged with AFM. The topography and roughness was simultaneously measured according to the user-defined locations. The green rectangles in AFM images and the blue columns in the line profiles included in Figures 2a – 2e show those measurements.

To determine the step height and roughness of the device within the user-defined area of interest, two locations were measured in each area: VZ1 (located at the bottom of the first pit from the device’s left edge), and Reference (a site on the raised area between the first pit and the second pit to its right).

Fig. 2a – AFM topography image of measurement site #1 with green inset rectangle denoting a specific area of interest. This area’s line profile is also provided along with recorded step height (H) and roughness (R) values gathered by measuring sites VZ1 (in the first device pit) and Reference (in the raised area between the first and second device pits).

Fig. 2b – AFM topography image of measurement site #2, area of interest line profile, and collected step height (H) and roughness (R) values.

Fig. 2c – AFM topography image of measurement site #3, area of interest line profile, and collected step height (H) and roughness (R) values.

Fig. 2d – AFM topography image of measurement site #4, area of interest line profile, and collected step height (H) and roughness (R) values.

Fig. 2e – AFM topography image of measurement site #5, area of interest line profile, and collected step height (H) and roughness (R) values.

The difference between the average height calculated at a Reference location and the same calculated at the VZ1 location helped to identify the step height. The color map also shows that difference in height. The color map is used for the AFM images in Figures 2a – 2e.

On the map, deep and recess features (such as pits), are darker in color, whereas the raised areas are a lighter shade. Calculations of the roughness value, which was obtained by scanning the specific locations VZ1 and Reference, with the pit floor roughness (at VZ1), was considered to be of more interest.

The collected data shows that the height of 120 nm and pitches of 10 μm in length are present in the pits of all five sites. It is also worth noting that pitch length can be repeatedly visually confirmed by looking at each AFM topography image in Figures 2a – 2e.

Each topography image can also be used for visual confirmation. Even if the AFM did not manage to scan the exact same area of interest at each measurement site, the degree to which each rectangle was offset can still be measured within a single micron. Microns offer a distinctively small amount of variance, given the automation employed in the study. The final step height and roughness values collected are made available in tabular form by the software after the user-defined recipe finishes running (see Table 1).

Table 1. Step height (H) and roughness (R) values measured at each of the five selected sites with corresponding averages and standard deviations.

Step height (H) and roughness (R) values measured at each of the five selected sites with corresponding averages and standard deviations.

It has been made evident that the sample was exposed to ambient air for a significant amount of time prior to being scanned with the AFM. For this reason, contamination of the sample at several locations on the wafer occurred (especially at sites 3 and 4), but contamination was also present in all measurements. This correlates to the discrepancies depicted in the line profiles collected in Figures 2c and 2d and the roughness values shown at each of these two sites.

The usual deviation in roughness is larger than expected for a pristine sample. The deviation can be attributed to contaminants that collected on the surface of the sample over a prolonged period of time exposed to ambient air. The contamination had the unintended effect of further demonstrating the sensitivity of the AFM measurements and its ability to discern differences in features at nanoscale.

The successful imaging of Five identical NOR Gate devices on a three-inch patterned silicon wafer consisting of rectangular pits with 10um pitch and 120 nm step height was measured. This was made possible using the Park NX-HDM AFM system with XEA, which is automation software from Park Systems.

All five patterns exhibited accurate and repeatable nanoscale metrology with quantifiable data via automated data acquisition and analysis. The user-defined recipe enabled this by combining optical pattern recognition with the precision of AFM.

Given the ability to automate data acquisition and analysis, this process has viable applications in similar fields, such as in bare silicon wafer manufacturing (surface roughness) and in wafer design research (critical dimension measurement). Research is routinely carried out in the latter two industries, and these developments could drastically increase their throughput and efficacy by incorporating tools capable of fully-automated AFM.

References

[1] Moammer, K. (n.d.). TSMC Launching 10 nm FinFET Process in 2016, 7 nm in 2017 Read more: Http://wccftech.com/tsmc-promises-10nm-production-2016-7nm-2017/#ixzz4AXXp5M3v. Retrieved June 3, 2016, from http://wccftech.com/tsmc-promises-10nm-production-2016-7nm-2017/

[2] J. Foucher; R Therese; Y. Lee; S.-I.Park; S-J.8681, Metrology, Inspection, and Process Control for Microlithography XXVII, 868106 (April 18, 2013); doi:10.1117/12.2011463

[3] Park NX-HDM features – Automatic Measurement Control. (n.d.). Retrieved June 03, 2016, from http://www.parkafm.com/index.php/products/industrial-afm/park-hdm/applications

[4] Park HDM – Overview | Park Atomic Force Microscope. (n.d.). Retrieved June 3, 2016, from http://www.parkafm.com/index.php/products/industrial-afm/park-hdm/overview

This information has been sourced, reviewed and adapted from materials provided by Park Systems Inc.

For more information on this source, please visit Park Systems Inc.

Ask A Question

Do you have a question you'd like to ask regarding this article?

Leave your feedback
Submit