In a revolutionary advancement, researchers at the University of Illinois Urbana-Champaign have leveraged artificial intelligence (ai) to transform Atomic Force Microscopy (AFM), a technique long constrained by the limitations imposed by the size of its probe in accurately mapping three-dimensional material surfaces with high resolution. Led by Professor Yingjie Zhang from the materials science & engineering department, this team has developed a deep learning algorithm that outshines existing methods and enables microscopes to delineate material features smaller than the probe’s tip with unprecedented accuracy.
Revolutionizing Atomic Force Microscopy with ai
Atomic Force Microscopy (AFM) is a critical tool in the field of nanotechnology, renowned for delivering comprehensive topographical maps that showcase height profiles of surface features. However, when surface features approach the scale of the probe’s tip, which is approximately 10 nanometers in size, the microscope’s resolution falters. To tackle this challenge head-on, Zhang’s team has introduced an innovative ai solution that defies conventional limitations.
Deep learning for eliminating probe width effects
The researchers’ groundbreaking advancement revolves around an encoder-decoder framework meticulously trained to eliminate the probe’s width effects from AFM images. The study’s lead author, Lalith Bonagiri, a graduate student in Zhang’s group, underscored the significance of this ai-driven approach. By meticulously encoding raw AFM images, stripping away undesired effects, and decoding them into precise representations of material surfaces, this deep learning algorithm offers a deterministic solution to the challenge posed by the probe’s size constraints.
Traditional microscopy techniques have primarily been limited to providing two-dimensional snapshots of material surfaces. AFM, however, distinguishes itself by delivering three-dimensional topographical maps. Yet, the resolution falters when surface features approach the scale of the probe’s tip. Zhang’s team has boldly addressed this issue with their ai-driven solution, offering a promising alternative to conventional limitations.
Training the algorithm for transformative results
To train their deep learning algorithm, researchers generated artificial images of intricate three-dimensional structures and simulated AFM readouts. The algorithm was meticulously crafted to manipulate these simulated AFM images, extracting the underlying features obscured by the probe’s size effects. Bonagiri emphasized the unconventional approach taken, particularly the decision to retain absolute brightness and contrast in the images, enhancing the algorithm’s efficacy.
The team showcased their ai’s capabilities by synthesizing gold and palladium nanoparticles with precisely known dimensions on a silicon substrate. Remarkably, the algorithm successfully eradicated probe tip effects and accurately identified and characterized the nanoparticles’ intricate three-dimensional features. Zhang emphasized that this achievement marks a significant milestone but is just the beginning, as continued training on more extensive and diverse datasets promises even greater strides in unraveling nanoscale landscapes.
A new era for nanoscale imaging
The fusion of ai and AFM pioneered by the University of Illinois Urbana-Champaign signals a new era for nanoscale imaging. By transcending the limitations of conventional methodologies, this research offers unprecedented insights into material and biological systems and paves the way for transformative advancements in nanoelectronics development. This breakthrough demonstrates the immense potential of ai to revolutionize scientific imaging and unlock new discoveries at the nanoscale.
