Experimental test of blind tip reconstruction for scanning probe microscopy

Abstract

Determination of the tip geometry is a prerequisite to converting the scanning probe microscope (SPM) from a simple imaging instrument to a tool that can perform width measurements accurately. Recently we developed blind reconstruction, a method to characterize the SPM tip shape. In principle this method allows estimation of the tip shape from an image of a tip characterizer sample that need not be known independently. In this work, we compare blind reconstruction results to those obtained by scanning electron microscopy for two diamond stylus profiler tips, one of which has a gentle shape and the other a more complicated profile. Of the two comparisons, the poorer agreement is still better than 30 nm for parts of the tip within a several micrometer neighborhood of the apex. In both cases the differences are comparable to the combined standard uncertainties of the measurements. We estimate uncertainties from five sources, the most significant of which is the repeatability of the stylus profiling instrument. In a separate measurement we determine the geometry of a silicon nitride SPM tip. The measured radius, 4-fold symmetry, included angle, and tilt are all consistent with expectations for such a tip.

Introduction

Originally developed to image surfaces with atomic resolution, scanning probe microscopies (SPM) such as scanning tunneling microscopy (STM) and atomic force microscopy (AFM) are now being used to perform width and roughness measurements on industrial surfaces like linewidth patterns of integrated circuits [1], optical surfaces [2], rough polysilicon [3], and data storage samples [4]. The advantages of using SPM over traditional methods (optical microscopy, electron microscopy) for such analyses include its unprecedented theoretical high vertical and lateral spatial resolution on both conducting and nonconducting devices, its versatility, and the fact that it requires very little sample preparation. As device dimensions continue to shrink, the requirements for accuracy, reproducibility, repeatability, mechanical and thermal stability, high-resolution positioning, and reliable calibration have become more stringent [5]. To fulfill these requirements, considerable improvement in SPM instrumentation and metrology has been achieved by manufacturer–vendors as well as by research and development laboratories.

A limitation not yet addressed by these developments is due to the nonvanishing size of the probe tip [6], [7], [8], [9], [10], [11]. In any SPM instrument, a probe tip is brought into close proximity to the sample surface. The topographic image can be described as the morphological dilation of the surface by the tip [12], [13], [14], [15], [16]. The dilation broadens peaks and bumps while holes and pits shrink (Fig. 1). Without an estimate of the tip shape, it is impossible to say what part of the image is due to the tip and what part is due to the sample. It has been found that for linewidth measurements the other sources of measurement uncertainty are usually significantly smaller than the uncertainty resulting from the tip size [17]. The situation is worse if the tip suffers wear between successive scans, because changes in tip geometry will generate inconsistencies in the collected data. Therefore it is important to establish a reliable method to characterize the tip geometry.

Previously proposed methods generally require the use of a special sample called a tip characterizer [11], [12], [13], [16], [18], [19], [20], [21], the geometry of which must be stable and independently measured with uncertainty small compared to the size of the tip. These requirements complicate tip determination by those methods.

We recently developed a method to determine the tip shape from the image alone, without independent knowledge of the characterizer geometry. This method relies on the basic tools of mathematical morphology, and in principle allows the derivation of an outer envelope that may closely approximate parts of the tip that were in contact with the sample surface during the scanning process. However there has not yet been a rigorous experimental confirmation of this technique. Such confirmation requires a comparison of the reconstructed tip shape to an actual shape given by a different reliable method. Finding an independent method for intercomparison is difficult at the size scales of interest for SPM (i.e., tip radii smaller than 50 nm) since other tip characterization methods also have important limitations at this size scale. For example, in the scanning electron microscope (SEM) the electron beam size, focus, charging effects, and beam–sample interaction make the specimen's edge positions uncertain at the nanometer level. Accordingly, to provide a quantitative comparison of blind reconstruction results to SEM images, we employed diamond stylus profiler tips. These tips are larger than AFM tips, and the SEM measurement uncertainties are therefore proportionally less of an issue. Two diamond tips were used in this experiment. The first has an approximately 500 nm radius with a smooth edge. The second had been damaged in previous work, causing it to be blunter (radius ≈2000 nm) but with a complicated profile that provides a stringent test of a tip measurement technique. A preliminary report of this work, including only results for the smooth tip and an abbreviated uncertainty analysis, was published in a conference proceedings [22].

In 2 Review of blind reconstruction, 3 Experimental procedure, 4 Results and intercomparisons, we will briefly present the blind reconstruction method, explain the experimental procedure conducted during this work, and present the intercomparisons with the associated uncertainty analysis. This quantitative comparison at a relatively large size scale is our main focus, but we will also present some preliminary results extending to size scales relevant for AFM in Section 5.