Analyzing sub-20 nm defects and ultrathin (~1 nm thick) residues in semiconductor processes

Background

With shrinking semiconductor device sizes and modern advanced packaging processes, it is crucial to eliminate sub-20 nm defects and surface contaminants. To do so effectively, one must be able to discern the molecular identity of a defect/contaminant. While the presence of defects as small as 20 nm can be detected with survey tools, the traditional suite of analytical tools—such as XPS, ToF-SIMS, or SEM/TEM EDX—have difficulty in clearly identifying the contaminating source of defects, especially if they are organic.

Organics are difficult to identify at this scale because traditional analytical tools either lack the spatial resolution required, or they only provide elemental information. Infrared photo-induced force microscopy (IR PiFM) can fill this gap by offering the ability to identify or name chemical compounds with nanometer-scale resolution. It does this by combining a non-contact AFM with IR spectroscopy to acquire topographical and chemical information concurrently at the nanoscale [1]. Since PiF-IR spectra match FTIR spectra for a given material, existing IR libraries can be used to identify defects analyzed with IR PiFM.

Given PiFM’s sub-5 nm spatial resolution, even a multi-component defect can be de-composed into pure components via multivariate data analysis of PiF-IR spectra from different regions of a defect.

Results and Discussion

Inorganic and organic nanoparticles

In Figure 1, a particle is found (within the red dotted circle) that is about 3 nm in height. The particle appears to be ~20 nm wide laterally. Since the AFM tip isn’t infinitely sharp, the particle is most likely a sphere with a diameter of ~3 nm whose width is dilated by the AFM tip’s radius of curvature which is about ~20 nm. The blue PiF-IR spectrum acquired on the particle matches the orange FTIR spectrum for silica from the Wiley IR library [2]. This demonstrates how easy it is to use PiF-IR spectra to identify unknown nanoparticles without having to make any inferences.

Another good example of how PiF-IR spectroscopy and PiFM imaging can be used to directly name materials at the nanoscale is shown in Figure 2. This sample is a 20 nm polystyrene (PS) particle deposited onto a mica surface.  The inset shows the AFM topography and the  PiFM images. PiFM images are essentially chemical absorption maps which show how strongly the material absorbs light of a given wavenumber, in this case 1492 cm⁻¹ that highlights the PS particle and at 1025 cm⁻¹ which highlights the mica surface. The lateral size is again dilated in the topography by the tip’s radius of curvature, so again, the height is going to be the best measurement of the diameter of the particle.

The red PiF-IR spectrum acquired on the nanoparticle and the blue spectrum on the surface match the FTIR spectra for PS (orange) and mica (green) well.  For PS, all the prominent FTIR peaks are clearly visible in the PiF-IR spectrum acquired on the particle. The additional peaks on the particle (between 1100 and 1350 cm⁻¹) may be due to the presence of a surfactant layer used to disperse the nanoparticles.  

Ultra-thin residues

Besides particulate defects, surface contamination can be another large issue in semiconductor processes. Dirty surfaces can interfere with bonding processes in modern advanced packaging processes. With monolayers, it is difficult to even detect its existence, let alone identify it. Many people currently rely on making inferences about the surface of a sample based on unreliable information like water contact angle measurements. However, PiF-IR spectra and PiFM images can detect and characterize monolayers directly.

In Figure 3a, five PiF-IR spectra acquired on an otherwise clean wafer from a storage container are shown at the bottom of the figure. The fact that they are quite repeatable even though they are acquired from 5 different locations suggests a homogeneous surface chemistry. In addition to the silicon oxide peak at 1130 cm⁻¹, they show a strong peak at ~ 1250 cm⁻¹.  When these spectra are averaged (blue spectrum) and a search performed in the Wiley IR library, one finds it matches the FTIR spectrum of perfluoro polyether (PFPE) which is shown in orange. Given the ultra-flat topography, the PFPE is expected to be a conformal monolayer that covers the surface quite uniformly. 

Note that even though the surfaces in Figure 3a and 3b are contaminated with different molecules, the micro-roughness of the silicon substrate remains the same, showing how thin the contaminant layers must be. Also, the PiFM image at 1130 cm⁻¹, which highlights the silicon oxide, shows much more contrast variations across the surface after the plasma cleaning as compared to the original one. This is consistent with the varying peak strengths observed in the spectra, suggestive of subtle differences in the oxide at different locations. This shows how useful nanoscale IR measurements are for looking at surface preparations. Without PiF-IR spectra, it would be impossible to know about surface contaminants like these, let alone track down where they might be coming from.

As another example, the cleanliness of copper surfaces as they are prepared for hybrid bonding can be examined. Hybrid bonding is gaining interest as replacement of thermal compression bonding for advanced packaging applications that require higher interconnect density. An important element to ensure successful hybrid bonding is an extremely flat, smooth, clean, and hydrophilic dielectric surface that is properly terminated with Si-OH bonds that can bond together instantaneously upon contact at 25 °C. Additionally, to form good metal-to-metal bonds, the Cu surfaces must be free of organic contaminants and carbides which prevent true Cu interfacial grain growth. It is known that benzotriazole (BTA), which is used as an inhibitor to protect recessed copper, reacts with copper ions to form a thin Cu(II)-BTA polymer on the copper surface that is difficult to remove [3].

Figure 4 shows averaged PiF-IR spectra of four copper surfaces: (a) as electroplated, (b) after the electroplated copper is treated by BTA, (c) after CMP process pf electroplated copper, and (d) after Ar ion cleaning following the CMP process. The as-electroplated copper (Figure 4a) shows a peak at around 730 cm⁻¹, which can be attributed to copper oxide. When the copper is treated with BTA (Figure 4b), two peaks at ~795 and 755 cm⁻¹ appear. These peaks are attributed to the C-H stretching vibration of the benzene ring in BTA, clearly indicating the formation of Cu-BTA polymer [3]. These two peaks are also present in the copper sample that undergoes a CMP process (Figure 3c) even though the peak at 795 cm⁻¹ appears as a shoulder of a new peak. The prominent new peak at ~615 cm⁻¹ is typically associated with Cu₂O⁴. When the same sample is cleaned via Ar ion beam (Figure 4d), the copper surface is mostly clean, except for a small peak at ~805 cm⁻¹, which may be due to the oxide peak associated with the AFM tip. Lastly, note the sharp feature at ~905 cm⁻¹ in all spectra is an artifact of laser power normalization, so it can be safely ignored.

Even though other surface sensitive techniques such as XPS can be used to check for the presence of the Cu-BTA complex, the advantage that PiFM offers is that its spatial resolution allows examination of individual copper pads, which are getting smaller than 1 µm in size.

Conclusion

IR PiFM can chemically identify sub-20 nm defects, regardless of whether they are organic or inorganic materials. It can also detect and characterize monolayer residues efficiently, without contaminating the sample. This is a huge advantage compared to something like SEM analysis which typically contaminates samples with carbon deposits.

IR PiFM also can be used to check post-CMP residues sensitively even on individual copper pads for Cu-Cu hybrid bonding applications; IR PiFM’s capability to evaluate the density of Si-OH bonds on SiO₂ surfaces after cleaning processes is covered in a separate application note. IR PiFM is available on instruments which support up to 12” wafers with many automation features, including automated defect analysis based on KLARF coordinates.

Contact us to discuss your defect and residue analysis needs and for a demo on your own samples.

References

1. D. Nowak, W. Morrison, H. K. Wickramasinghe, J. Jahng, E. Potma, L. Wan, R. Ruiz, T. R. Albrecht, K. Schmidt, J. Frommer, D. P. Sanders, and S. Park, “Nanoscale chemical imaging by photoinduced force microscopy,” Sci. Adv., 2:e1501571 (2016).
2. KnowItAll IR Spectral Database Collection – Wiley Science Solutions provides the library and the search algorithm.
3. Q. Wang, B. Tan, S. Tian, C. Han, L. Yang and B. Gao, “Study on infrared spectrum detection and analysis of BTA residual after copper CMP,” 2019 China Semiconductor Technology International Conference (CSTIC), Shanghai, China, pp. 1-3 (2019).
4. A. Chen, H. Long, X. Li, Y. Li, G. Yang, and P. Lu, “Controlled growth and characteristics of single-phase Cu2O and CuO films by pulsed laser deposition”, Vacuum 83, pp. 927-930 (2009)