3D control of photomask etching using advanced CD AFM metrology
01/01/2008
EXECUTIVE OVERVIEW
A new generation of atomic force microscopes (AFM) has been designed for critical dimension (CD) metrology applications. Complex tip shapes, automated tip calibration capabilities, and a novel surface sampling protocol are all key factors enabling the latest CD AFMs to deliver high precision mask metrology data. Studies show that tip wear <2Å/site can be tolerated while maintaining proper CD accuracy and precision
By 2008, the ITRS roadmap predicts a CD precision of 0.86nm (3σ) for masks used in manufacturing 90nm node ICs. To achieve this level of precision, reticle makers will require advanced dimensional metrology to characterize and control processes involving novel materials, new structures, and shrinking enhancement features. A new generation of 3D atomic force microscopes (3D AFM) will meet the needs of the 65nm node and beyond. Specifically, these CD AFMs deliver both imaging and numeric data, including the 3D shape of mask features.
Mask metrology for defects and repair
Mask defects are of prime concern because they replicate. A single defect on a wafer might disable a single chip, but a killer defect on a mask can severely degrade the entire area of that flash field, up to a quarter of the wafer. Moreover, depending on the level, a mask defect might ruin the entire wafer. So while wafer defect reduction is essentially an analytical process where one may want to find root causes, on a mask all defects must be found and fully classified.
Mask defects must be characterized for size and volume, as well as for composition, to allow for mask repair techniques to be implemented. Where possible, repairing a mask is the most desirable solution to a defect problem, primarily because of the high cost of mask sets. Of course, repair is dependent on the availability of robust mask metrology. Specifically, it requires metrology tools that provide 3D imaging of defects for effective, accurate, and quantitative analysis of volume.
In terms of resolution requirements, 15-20nm dia. particles are already beginning to cause concern, so metrology tools must function at this level. With masks for immersion lithography, and eventually EUV, defect detection and characterization will become ever more challenging. An ideal metrology tool is thus one that meets the dimension requirements of today’s lithography technology and that can follow the roadmap through to EUV.
The AFM is the only technology with sufficient resolution and sensitivity to reliably detect 15-20nm diameter particles. Plus, it has sufficient overhead to support future EUV masks. And unlike other metrology techniques, the AFM is completely nondestructive. For example, it is known that SEMs (scanning electron microscopes) may cause a problem referred to as “staining” or “stamping” due to the effect often appearing like the image of a postal stamp. This secondary damage is caused by dissociating components within the vacuum chamber, which then deposit on the mask as a carbon film.
AFM technology overview
The principle of operation of an AFM is fairly straightforward. A fine tip at the end of a micro-fabricated cantilever is moved with a 3D piezoelectric actuator toward the surface under test. As the distance between the tip and the surface approaches zero, inter-atomic forces are at first attractive and then become strongly repulsive as contact is made. The AFM senses these forces by detecting the movement of the cantilever as the tip scans the sample surface. This is typically done by reflecting a laser beam off the top surface of the cantilever, and monitoring the reflected beam with a position-sensing photodetector.
For most surface morphology applications, the instrument can be operated in a continuous contact mode or in an intermittent contact “tapping” mode. Today, most AFM surface morphology metrology is carried out using tapping mode, where the cantilever is maintained in constant-amplitude oscillation, typically 10-50nm root-mean-square (rms). This enables the tip to be rapidly scanned across the test surface while reducing drag artifacts or damage to the tip or surface that arise from shear forces during constant contact scanning.
3D AFM advantages are high spatial resolution and nondestructive operation. In addition, they require no sample preparation, working well in both air as well as under liquids. In spite of these advantages, the AFM has traditionally offered only limited utility for photomask metrology. In particular, these instruments have provided accurate profiling only in the Z direction, i.e., measuring “horizontal” surfaces by sensing the up and down forces on the AFM tip. They have not provided the ability to perform sidewall metrology and also often have struggled to profile and measure the bottom surface of narrow, high aspect ratio (HAR) trenches.
However, every other established surface metrology tool has also proved to have significant limitations for mask metrology. For example, the scanning electron microscope (SEM) also provides high resolution “top down” images without detailed sidewall or trench-bottom information. Moreover, measurement accuracy can be compromised by surface charging effects. Cross-section techniques-such as SEM, FIB, and TEM-can provide high resolution cross-section information but are destructive, slow, and cumbersome. Optical scatterometry is a nondestructive tool, but requires lengthy testing to empirically develop robust models to derive accurate information from the raw data. Plus, each change in surface details necessitates the development of a new model, and the models can be altered by underlying surface layers and sample-to-sample variation.
CD AFM advances
In response to this situation, AFM tool builders have been working with mask technology experts to develop a new generation of AFMs specifically optimized for mask metrology. A primary focus of these efforts has been in improving tip shape and motion characteristics.
 Figure 1. a-b) Tip shape parameters for a CD-AFM tip used for CD and profile measurement, and c) the profile of a real tip reconstructed from TipQual calibration images. |
The conventional AFM probe tip has a conical or pyramidal profile terminating in a sharp point. One of the keys to successful 3D metrology with an AFM has been the development of specially shaped tips, designed for point contact profiling on both trench sides and lower surfaces. Figure 1 shows the theoretical and actual (measured) profile of one of these “boot-shaped” tips used in the new generation of CD AFM instruments. This two-sided boot profile with its lateral overhang can be used to profile sidewalls (Fig. 1b). Moreover, if the overhang is large enough, then this type of tip can be used to accurately profile re-entrant trenches with “reverse vertical surfaces,” trenches with sidewall angles >90°.
Reliable CD metrology requires that the precise shape and dimensions of the tip are known with high accuracy, including tip length, overhang, width, edge radius, and vertical edge height. The latest CD AFM instruments incorporate an on-board tip qualification and characterization routine, called TipQual [1], wherein the instrument measures the key parameters of the tip by scanning it across standard samples with highly calibrated CD values. For example, the tip may be first scanned across a standard vertical line with a known CD linewidth, and the tip width can then be determined from the AFM image width minus the known CD value. Next, the tip scans over a re-entrant trench having very sharp top corners; these corners act like tips and enable the software to reconstruct the tip shape from the measured AFM image.
 |
The table lists the typical limits to reject a tip for etched mask applications. For example, the edge overhang should not be <5nm or the tip may be unable to accurately image the sidewall of re-entrant trenches. Conversely, an overhang of >100nm implies that the tip is faulty and should be discarded. Similarly, the tip edge height should be <50nm in order to accurately image the middle CD of an etched mask feature. The optimum specifications for tip lengths vary with the application; a minimum tip length of 120nm will meet most mask CD metrology needs. Finally, when the tip scans across the vertical line during the qualification routine, the measured sidewall angle should be close to 90°.
Just as important as shape is the manner in which a tip moves across and contacts with mask surfaces. The problem with the conventional tapping mode is that it is designed only for accurate Z-axis metrology. Figure 2 shows tapping mode being used on a sample with vertical line features; the sidewall angles are inaccurately measured as <90°. Instead of this scanning mode, CD AFM instruments incorporate two different scanning algorithms: CD (critical dimension) and DT (deep trench) modes.
Unlike the usual tapping mode, CD and DT modes rely on adaptive scanning. Instrument software automatically slows down in the horizontal direction when the tip senses a sidewall, allowing the tip to slowly progress along the slope. In DT mode, the scanner servos in the vertical direction until it reaches the bottom of the trench, and then resumes horizontal scanning. As a result, the DT mode is optimized for depth measurement of HAR trench features. In the CD mode, the scanner slows when it senses a sidewall, and then servos at 45° to the sidewall surface, collecting 2D data on the sidewall profile. The CD mode thus traces and reproduces the full shape of both the flat and sidewall surfaces.
Calibration and wear studies
Repeated use of an AFM tip inevitably leads to wear. As a general rule for mask applications, a tip wear rate of <2Å/site is considered acceptable, with no significant impact on CD precision and accuracy. Since tip dimensions and shape are so critical for mask CD applications, this aspect of the CD AFM has to be rigorously addressed. Ongoing improvements focus on three areas: the development of wear-resistant tips with harder surfaces, reducing instrument-related causes for tip wear, and the use of robust tracking of tip parameters by frequent tip qualification and calibration.
 Figure 3. Summary of tip wear data on tested etched mask plates. |
We have recently evaluated two different tip materials-single crystal silicon, and single crystal silicon coated with silicon nitride-for use in mask metrology applications using three types of etched mask plates. Figure 3 summarizes the results of these studies, indicating that a two-stepped etch profile tends to consume the tip most quickly. Not surprisingly, the tough SiN coating reduces tip wear with all three mask plate types.
The latest CD AFMs reduce instrument-related tip wear by two different approaches. First, the software is carefully optimized regarding sample clearance to ensure proper surface-tip engagement. Another advance has been to reduce drift in the sample stage, which has been shown to be a major contributor to tip wear in the past.
Although some level of tip wear is an inevitable fact of life with CD AFM, our recent studies show that moderate levels of wear can be accommodated using robust calibration protocols based on tip qualification. Specifically, each subsequent measurement uses the data from the most recent tip qualification to calculate CD values from the raw data. Figure 4 shows two sets of bottom CD measurements made with normal tip wear and with deliberate exaggerated tip wear; running a qualification step after the ninth site can be seen in the return to proper measurement even after high tip wear. With normal tip wear, correct CD values are maintained between tip qualifications.
Etch CD and profile metrology data
Our studies have also evaluated the new CD AFM in several mask metrology applications. For instance, sidewall angle can have a critical impact on the resulting wafer CD, particularly when using off-axis lithography. In fact, a difference of just a few degrees can result in a change in the dose needed to expose the wafer.
 Figure 5. Comparison of a) CD SEM and b) CD AFM data for Cr-MoSi etch profiles. Notice the protuberance accurately profiled by the CD AFM. |
In another example, a change in process bias for dark field masks was observed using a CD SEM and initially attributed to charging problems. However, profiling with the CD AFM revealed that the profile change was actually a result of a process change. The CD SEM data were shown to contain “halo” artifacts (Fig. 5a) due to damage to the chrome masking layer. The halo effect can be seen as two white lines in each image running from top to bottom. The CD AFM showed that protuberances from the MoSi sidewall (Fig. 5b) prevented the CD SEM from obtaining accurate and repeatable linewidth CD values.
Etch depth is directly related to the phase shift of the final mask plate, so it is very important to control the etch depth during process development. As trenches become narrower, inverse RIE lag effects become more problematic in both wafer and mask etch processes, and manifest as dependence of the etch depth on trench width. This problem will become more severe at the 65nm node and beyond. Consequently, while large test structures can be used for phase targeting, they can result in phase errors near CD specifications.
Fortunately, this is an ideal task that can be rapidly accomplished using a CD AFM operating in DT mode. Moreover, because the DT mode only involves vertical scanning and can even be performed with a conventional conical tip, tip wear is very minor, and a tip can typically scan over 1000 sites before falling out of specification. Using FIB tips in DT mode, we have shown that the new CD AFMs can directly measure minute features at the pattern level with a very high level of precision, allowing accurate depth control and thus phase targeting.
Conclusion
In conclusion, advances in CD AFM technology have made this a powerful tool in the instrument arsenal for cutting edge mask process development at the 65nm node and beyond for CD, depth, and defect review applications.
Reference
1. G. Dahlen, et al, “Tip Characterization and Surface Reconstruction of Complex Structures with Critical Dimension Atomic Force Microscopy,” J. Vac. Sci. Technol. B, 23 (6), pp. 2297-2303, 2005.
Tianming Bao received his PhD in chemical engineering and his MS in statistics from Case Western Reserve U. He is an application scientist for Veeco Instruments. Veeco Metrology Inc., 112 Robin Hill Road, Santa Barbara, CA 93117; ph 805/967-1400, e-mail [email protected].
Dean Dawson received his MS in engineering management from Coventry Polytechnic, UK, and his BS in applied physics from De Montfort U., UK. He is senior director of marketing for Veeco Instruments’ Metrology Division.