Nova Measuring Instruments Inc.

ABSTRACT Thermo-compression bonding (TCB) is a key enabler for ultrafine-pitch interconnects in advanced semiconductor packaging, where flux-related contamination and cleaning limitations increasingly threaten yield and reliability. Flux-less TCB overcomes these limitations but requires controlled and residue-free oxide removal at solder interfaces, together with robust methods to verify surface chemistry on product-relevant structures. In this work, a hydrogen-radical-based in situ reduction process integrated into a chip-to-wafer TCB platform is evaluated, and oxide removal on 20 μm pitch solder-capped copper microbumps is quantitatively assessed using microspot X-ray photoelectron spectroscopy (XPS). A measurement methodology was developed to address challenges associated with small feature size and curved bump topography. High-resolution Sn 3d spectra enabled separation of metallic and oxidized tin contributions, allowing estimation of effective oxide thicknesses. Results show substantial reduction of native tin oxide following plasma treatment. The approach proved applicable to both cylindrical and dome-shaped bumps. Studying oxide thicknesses of treated samples after storage times of up to 24 hours showing reduced re-oxidation, promising results for extend Q-time processes and therefore enabling a larger process window.

Metrology in high-volume semiconductor manufacturing is increasingly constrained by the need for nanometer-level precision, extensive sampling, and the growing complexity of device architectures. As the global demand for semiconductors accelerates, there is an urgent need for scalable solutions that enhance measurement efficiency and throughput from manufacturing facilities without compromising on measurement integrity. In this study, we present results from an AI-driven in-line metrology framework developed as part of this work, which leverages big data analytics and machine-learning to significantly reduce measurement time while maintaining accuracy, precision, and sensitivity. By integrating AI with established film analysis algorithms, the system performs spectral reconstruction to compensate for signal degradation caused by reduced acquisition times. It is trained on large-scale historical datasets, enabling reliable performance in real-world production environments. Experimental evaluations conducted in a high-volume manufacturing setting at a GlobalFoundries facility demonstrate up to a twofold increase in metrology throughput, while remaining within bounds to ensure measurement integrity in terms of accuracy, precision, and sensitivity. This advancement offers a scalable and cost-effective approach to accelerate XPS metrology, with the potential to support next-generation fabrication demands.

A systematic study of the co-optimization of target design and metrology technique is presented to accurately measure the critical dimensions of backend of line (BEOL) metal line gratings. Rigorous coupled-wave analysis calculations and machine learning approaches are combined to evaluate various design scenarios with and without patterned underlayers in conjunction with either traditional scatterometry or vertical traveling scatterometry (VTS) using spectral interferometry. It was found that for traditional scatterometry techniques employing polarized reflectometry or ellipsometry, two levels of crossed metal lines buried below the level of interest are often sufficient to suppress most of the optical contributions from any underlayer stack beneath. Alternatively, VTS utilizing spectral interferometry and signal filtering can suppress all contributions from the underlayer stack independent of the design choice thus only the top layer of interest needs to be considered in the model analysis. Machine learning models trained on VTS data instead of traditional scatterometry data can improve the accuracy and ease of setup, for example, by utilizing simplified targets for training. Three relevant BEOL cases for measurements after an etch step, a polishing step, and dielectric layer deposition on patterned metal lines are addressed.

In this study, we introduce a machine learning approach designed to augment the conventional Rigorous Coupled-Wave Analysis (RCWA) method used in scatterometry measurements. The utility of this approach is illustrated through two practical examples. Initially, we applied it to a recess structure in trench MOSFET. Following the application of our machine learning method to the RCWA model, the recess depth measurement exhibited improved stability and uniformity across the wafer. In the second example, we measured a 2D line trench in silicon (with a depth of 22 µm); here, both the top and bottom widths are parameters of interest. We show that our machine learning based model is more robust compared to the conventional RCWA method. Our results were then cross-verified using atomic force microscopy results and cross-section Scanning Electron Microscopy data, respectively.

This paper demonstrates the successful lab-to-fab transition of dynamic secondary-ion mass spectrometry (SIMS). In comparison to traditional lab SIMS, the in-line version is optimized for automated wafer and measurement sequence handling and high throughput measurements in small areas. Key advantages are fast turn-around time, reduced scrap, increased yield, and the measured wafer can continue processing in the manufacturing line. The benefits of in-line SIMS in the production environment are demonstrated for several use cases: matching and monitoring the long-term stability of epitaxy tools on monitor wafers, process optimization and monitoring of epitaxial Si and SiGe layers on blanket and patterned wafers with blanket metrology targets, measurement of implant and dopant profiles on blanket and patterned wafers, and characterization of the Ge and B diffusion in multi-layer stacks stimulated by high-temperature annealing. Additionally, the characterization of the source/drain epitaxy in a fully integrated nanosheet gate-all-around transistor architecture is demonstrated and discussed. The results are compared to off-line lab SIMS and alternative methods where available.