Edit/Keich
Understanding the properties of materials requires structural characterization on a large scale and at different scales to link microstructure to properties. The connection between microstructure and properties forms the basic paradigm of materials science. However, a key challenge is to characterize microstructure at appropriate scales (or multiple scales).
A general approach is to use scanning electron microscopy (SEM), a tool for material characterization that provides information about the surface or near-surface structure of a material, but still faces signal collection challenges in practical use.
The research team from Tsinghua University has developed a single-beam high-throughput SEM method that can be used to simultaneously collect secondary electron (SE) and backscattered electron (BSE) signals at nanoscale resolution, with a good signal-to-noise ratio and high imaging speed, combined with machine learning, to achieve efficient material research.
The study, titled "A macro-nano-atomic–scale high-throughput approach for material research," was published in Science Advances on December 1, 2021.
Secondary electron (SE) signals and backscattered electron (BSE) signals are two of the most important and widely used signals in SEM systems, where the image contrast mechanisms generated by these signals are called terrain contrast and component contrast, respectively.
However, in practice, large-scale scan-based investigations are not possible using traditional SEM because these tools are inherently too slow.
To solve this problem, multi-beam high-throughput SEM (mSEM) has been developed to increase throughput by increasing the number of main beams. However, in these mSEMs, the SE signal capability of each single beam is still similar to that in traditional SEM, and it is not possible to collect additional BSE signals at the same time. This creates significant restrictions on the use of mSEM in the field of materials science.
Here, the researchers developed a single-beam high-throughput SEM method that can be used to simultaneously collect SE and BSE signals at nanoscale resolution with good signal-to-noise ratio and high imaging speed (at 2100 megapixels per second, the minimum dwell time per pixel is 10 ns).
Schematic of a single-beam high-throughput SEM method.
The method is based on a single-beam high-throughput SEM equipped with a specially designed electron optics system and a detection system, utilizing a direct electron detector and an optimized deflection system. Combined with machine learning, the microscope can be used to identify and distinguish different phases, allowing for collected and post-analysis of experimental data on a wide range of length scales, thereby connecting experimental results at the centimeter level to the nanoscale or even the atomic level.
This is illustrated by a multiscale study of carbides in the second generation of nickel-based single crystal superalloys. Alloy samples are taken under five conditions: cast, heat treated, creep testing at 1038 °C/155 MPa to 22.2 and 131.6 hours, and after creep breakage (SA1, SA2, SA3, SA4 and SA5 below).
Each sample of the panorama was collected at nanoscale resolution, identifying primary (Ta, Hf)C carbides and secondary (Cr, Re)23C6 carbides.
Collect digital data from a panorama set
Panoramas produced by high-throughput data collection and automatic phase recognition through machine learning models can be easily converted into digital data for statistical analysis, based on data collected in sample-scale areas, from which the evolution of the size and volume (area) fraction of carbides can be quickly determined.
Statistical analysis of carbides based on sample-scale data.
The study found that the amount of M23C6 carbide in the sample decreased as the carbide grew during the creep process, and in addition, M23C6 carbide most likely played a key role in the creep process of this superalloy.
Additional microstructural characterization
The single-beam high-throughput SEM method enables rapid multiscale identification and quantification of nanoscale resolution over large sample areas (tens of square millimeters). However, determining the microstructure of the carbide and its relationship to the matrix (both of which are important for understanding the effects of carbides on creep properties) still requires some additional microstructure analysis.
A detailed examination of the dislocation structure around the carbide is followed, taking care to examine the carbide with the size, shape, and position representing the carbide identified in the high-throughput SEM observation.
Microstructure of M23C6 carbides and their relationship to surrounding substrates.
Combining the observed results with high-throughput SEM data shows that as the size of the M23C6 carbide increases, their shape becomes polyfaceted and connected to the matrix through the M23C6/interface. These interfaces are important throughout the creep period because they are barriers to dislocation movement.
Building on this understanding, the researchers also conducted STEM experiments to explain the relationship between M23C6 carbides and matrices by analyzing the M23C6/interface at the atomic scale.
Based on another set of high-throughput SEM observations of five deeply etched samples and STEM high-angle circular darkfield observations of sample SA5, the number of polyhedra M23C6 carbides was found in four typical types, defined by the coherent M23C6/interface of the three types.
It can be concluded that the carbide changes that occur during creep have a positive effect on the creep properties of the present superalloy.
"The comprehensive results we present here validate the feasibility and accuracy of the technique in identifying and distinguishing different phases in the micro or nanoscale, allowing for post-collection analysis of experimental data over a wide length range, linking macroscopic nanoatomic experimental results to performance," the researchers said.
As demonstrated in this study, the ability to combine SE/BSE high-throughput data collection is expected to be widely used to analyze the structure-performance relationships of heterogeneous materials. In addition, future applications may need to be combined with tomography techniques to build sample-scale three-dimensional (3D) models with complete microstructural details.
Thesis link: https://www.science.org/doi/10.1126/sciadv.abj8804
Artificial Intelligence × [ Biological Neuroscience Mathematics Physics Materials ]
"ScienceAI" focuses on the intersection and integration of artificial intelligence with other cutting-edge technologies and basic sciences.
Welcome to follow the stars and click Likes and Likes and Are Watching in the bottom right corner.