Patterns that just seem random: A study shows that salts can be identified based on the stains they leave after drying. Based on subtle similarities in patterns between crystal structures, the AI system has so far been able to map 42 different types of salt spots with high accuracy. The researchers say this “fingerprinting” approach could lead to a simplified analysis process that could provide key information about materials when exploring celestial bodies, for example.
We know this from the kitchen: a drop of salted pasta cooking water leaves a white stain on the counter after it dries. If you look at it with a magnifying glass, you can see something that looks like an abstract artwork: filigree structures created by a sodium chloride crystal. Each drop forms individual patterns. This is because complex fluid movements and a number of environmental factors affect its sedimentation pattern. However, crystal growth is also affected by the special properties of the material in question. However, simple distinctions between the several types of salts are not possible: at least superficially, the residue of a sodium chloride solution is indistinguishable from a potassium chloride dye.
Can artificial intelligence develop the “expert’s eye”?
For this reason, researchers led by Bruno Battista of Florida State University in Tallahassee explored how well artificial intelligence can recognize subtle differences: Is it possible to identify salt deposits based solely on their appearance? To capture the underlying potential, the researchers took 7,500 detailed images of 42 different types of salt patches. They were “grown” from small droplets under uniform conditions on glass slides. Using a proprietary software concept, each image was then tagged using 16 parameters. These included features such as deposition area, compactness and texture. This in turn reflects the finer properties in the arrangement of small crystals into rings, needles and leaf-like structures.
The researchers then fed and trained the AI system with this image information: Machine learning methods allow the detection of hidden patterns in the data. Specifically, the system must learn to recognize typical features of salt patches of a given type based on subtle structural similarities. As the team reported, the approach was successful. This was demonstrated by their system’s ability to recognize the identity of the salt based on new spots. Despite a relatively modest training data set so far, the AI was able to identify the salt in question based on the spot image with a high level of accuracy. “We were surprised by how well it worked,” says senior author Oliver Steinbock of Florida State University. “Who would have thought that you could differentiate between sodium chloride and potassium chloride based on a picture? They look very similar in pictures, but the method recognizes the difference,” says the researcher.
Promising approach
As he and his colleagues emphasize, this is a testament to the feasibility of this so far: at least for pure aqueous brines, it has already been shown that it is possible to identify a substance based on the appearance of microgram-sized precipitates. Now this concept will be greatly expanded: the researchers plan to expand the training data set by analyzing hundreds of thousands of additional images. In addition, more compounds and mixtures of materials should be included, which will make the tool more accurate and versatile. However, this large number requires automation. Researchers are currently testing the use of an automated fall imaging device.
According to them, there is great application potential for this concept. Salts are of great importance in chemistry and nature, which makes it interesting to be able to identify them quickly and easily using a picture. Among other things, this process can be used in space travel. Because it is difficult and expensive to equip an alien celestial body exploration vehicle with a full-fledged chemistry laboratory. For the new concept, a high-resolution camera may be sufficient to obtain at least basic information about the composition of the sample material. According to the team, another advantage of this approach is the lower material requirements: just a few milligrams of salt deposits can reveal what it is.
Source: Florida State University, specialized article: Proceedings of the National Academy of Sciences, doi: 10.1073/pnas.2405963121
“Alcohol buff. Troublemaker. Introvert. Student. Social media lover. Web ninja. Bacon fan. Reader.”
More Stories
A woman explains the disease online
Discovery of “Trilobite Pompeii” – forschung.de
Why does Covid save some people? Researchers are baffled