Hyperspectral imaging combines spectroscopy, which provides material and topological information, with imaging for shape and structure recognition. This convergence of technology yields a comprehensive data set for those in the manufacturing sector. Hyperspectral imaging has been leveraged in a number of applications from medical imaging and agriculture to defense and surveillance. But the technology can also be a much-needed boon in the factory as an aid for materials and quality inspection.
Silicon wafers require precise inspection during production. Courtesy of iStock.com/Aaron Hawkins.
For instance, a layer stack data set can reveal layer thickness distribution, layer composition homogeneity, defect presence and classification, and pore detection and quantification. It can also provide quality classification and downstream production step quality predictions. Depending on the industry, this can be invaluable to lowering production costs, as it may reduce manufacturing times and identify defective products early in the process.
In semiconductor manufacturing, hyperspectral imaging is used to evaluate the spatial distribution of thin-film thickness of oxides, resists, or surface parameters before or after every processing step in lithography, chemical vapor deposition, physical vapor deposition, and atomic layer deposition. The combination of detailed spectral data with high-resolution imaging enables a thorough analysis of wafer properties, ensuring quality and consistency.
Specim’s FX17 hyperspectral camera. Courtesy of Specim.
DIVE imaging systems builds hyperspectral vision systems, integrating hardware, software, and comprehensive solutions for industrial inspection tasks. DIVE addresses the need for meticulous inspection and quality control in thin-layer application processes. Any deviation from specifications in these layers can lead to malfunctions in highly specialized applications, such as microchips, making accurate assessment crucial. DIVE’s technology offers a comprehensive evaluation of surface characteristics with strong spatial resolution, which is particularly beneficial in semiconductor manufacturing.
To help in this endeavor, DIVE integrated its VEpioneer, a one-button benchtop system for at-line usage, with Specim’s FX10 and FX17 hyperspectral cameras to create a hyperspectral vision-based solution that can make these assessments quickly and with minimal error.
DIVE imaging’s VEsolve Pro benchtop inspection system. Courtesy of DIVE imaging systems.
Creating hyperspectral vision
For DIVE, the term hyperspectral vision refers to the combination of its VEpioneer hardware with its VEsolve Pro all-in-one software suite. VEsolve Pro uses AI to process all gathered hyperspectral data, which can be viewed as both a stack of images and a bundle of spectra, resulting in millions of available data points for various forms of timely analysis.
“In order to achieve nearly real-time processing of the data, we use our own machine learning solution,” said Philipp Wollmann, CEO of DIVE imaging systems. “Since we are working on industrial applications, our algorithms must be fast, reliable, and repeatable. Over the past ten years, DIVE has gained a wealth of experience for industrial application of AI algorithms in hyperspectral vision solutions.”
Using hyperspectral vision, the success of processing steps in semiconductor manufacturing can be reapplied to direct tests on production wafers, significantly reducing the need for standard test wafers. This, of course, also relies on the hyperspectral cameras inside the system.
Camera integration
“We are combining classical hyperspectral imaging cameras with manual or automated sample positioning, an in-house-developed excitation and an all-in-one software suite for a holistic hyperspectral vision solution,” Wollmann said. “Our technology is designed to inspect thin layers and performance surfaces. Thus, DIVE’s instrumentation differs immensely from other hyperspectral imaging solutions in its application area.”
Since the company was packaging so many functions into one solution, it needed to find a camera that would be able to capture enough data so that the system’s processing capabilities would not be wasted. The camera also needed to be small enough to be integrated easily into the system.
“Specim’s FX series is superior to other hyperspectral cameras when it comes to industrial applications due to their compact design and robust housing without any enclosure openings,” Wollmann said. “From our perspective, the small plug connectors and the GigE interface enable an easy, straightforward integration without any compromises regarding cable lengths, flexibility, and overall setup size.”
A close look at a Specim camera analyzing a wafer integrated into the VEsolve Pro. Courtesy of Specim.
The company also identified other areas where the FX10 and FX17 excelled at their intended application. These include noninvasive measurement, 100% area coverage of the inspection site, and rapid scan times of 300-mm wafers in 30 s. Given these advantages, DIVE found that the VEpioneer is faster and more efficient with the integration of the hyperspectral cameras, moving closer to reaching the zero-defect goal for the system.
“The integration of Specim’s hyperspectral cameras has revolutionized our fully integrated hyperspectral imaging solution for industrial inspection. We have seen remarkable improvements in quality and efficiency. Looking ahead, we are excited to leverage this technology further to meet the evolving demands of the semiconductor industry,” Wollmann said.
Inspecting more than wafers
By resolving critical inspection challenges, Specim’s hyperspectral imaging technology has enabled DIVE to establish efficient and reliable wafer inspection, improving its partners’ production processes.
But beyond wafer inspection, Wollmann believes hyperspectral vision technology will revolutionize machine vision in other aspects of semiconductor manufacturing, due to its ability to capture unique, detailed data from hundreds of images. From bipolar plates for fuel cells and glass to printed circuit boards and steel, the integration of machine learning and AI algorithms with chemical-analytical methods allows for innovative approaches in the assessment of product quality.
“Looking ahead, we are excited to leverage this technology further to meet the evolving demands of the semiconductor industry,” Wollmann said.