Melting temperatures are often measured after carefully calibrating crystal structures or drawing thermodynamic free energy curves when a material melts, creating a phase change from a solid to a liquid. But when high-temperature materials exceed 2,000 or 3,000 degrees, finding an experimental chamber to make the measurements can be a challenge. And sometimes rocks have complex mixtures of minerals not much larger than a grain of sand, so getting enough samples of a single mineral can also be a challenge. This is where AI comes in.
“We use machine learning methods to fill this gap by creating a fast and accurate map from the chemical formula to the melting temperature,” said Qi-Jun Hong, one of the scientists involved in this research. in a press release. “The model we have developed will facilitate the analysis of large-scale data involving melting temperature in a wide range of areas. These include the discovery of new high-temperature materials, the design of new metallurgical processes extractive, the modeling of mineral formation, the evolution of the earth throughout geological time and the prediction of the structure of exoplanets.
Hong’s approach allows melting temperatures to be calculated in milliseconds for any chemical or compound formula input. To do this, the research team built a model based on a neural network architecture and trained its machine learning program on a custom database that includes 9,375 materials, of which 982 compounds have melting temperatures above 3,100 degrees Fahrenheit. Materials at this temperature glow as white.
Bidirectional search
The methodology is being used to explore two lines of research: 1) predicting the melting temperatures of nearly 5,000 minerals and 2) finding new materials that have extremely high melting temperatures above 5,000 degrees Fahrenheit.
For the mineral project, Hong’s team was able to predict melting temperatures and correlate them with the major known geologic epochs in Earth’s history.
These melting temperatures obtained by AI were applied to minerals made since the formation of the earth about 4.5 billion years ago. The oldest minerals originate directly from stars or from condensates of interstellar and solar nebulae that predate the formation of the Earth 4.5 billion years ago. These are the most refractory, with melting temperatures around 2600 F.
For the most part, there was a gradual decrease in the calculated melting temperatures of minerals identified on earth in more recent times, with two major exceptions.
“The gradual global decrease in the melting temperature of minerals formed during Earth’s history is interrupted by two anomalies, which are clearly manifested in the mean and median melting temperatures using binning 250 or 500 million years ago” , Alexandra Navrotsky, co-author of the study. that presents these findings, he said.
The first anomaly in Earth’s early history came from a dramatic spike in temperature caused by a moment of major meteorite impacts, possibly leading to the formation of the moon.
“The increase at 3.75 billion years ago correlates with the proposed timing of the late bombardment, hypothesized exclusively from the dating of lunar samples and currently debated,” Navrotsky said.
Formation of minerals that contain water
The team also noted a large temperature drop in the melting temperatures of minerals about 1.75 billion years ago, which is related to the first known occurrences of a large number of hydrated (water-containing) minerals and correlates with the Huronian glaciation, the longest ice. It was thought to be the first time our planet was completely covered in ice.
With their machine learning program trained to successfully replicate the melting of minerals in Earth’s early history, the team next turned their attention to finding new materials that have extremely high melting temperatures.
Dozens of new materials were identified and computationally predicted to have extremely high melting temperatures above 5,000 degrees Fahrenheit, more than half the temperature of the sun’s surface.
The team made the model simple and reliable enough that any user can obtain the melting temperature in seconds for any compound based only on its chemical formula.
“To use the model, a user must visit the web page and enter the chemical compositions of the material of interest,” Hong said. “The model will respond with a predicted melting temperature in seconds, as well as the actual melting temperatures of the nearest neighbors (i.e., the most similar materials) in the database. Therefore, this model not only serves as predictive model but also as a melting temperature manual”.