Microplastics – the tiny particles of plastic shed when litter breaks down – are everywhere, from the deep sea to Mount Everest, and many researchers worry that they could harm human health.
I am a machine learning researcher. With a team of scientists, I have developed a tool to make identification of microplastics using their unique chemical fingerprint more reliable. We hope that this work will help us learn about the types of microplastics floating through the air in our study area, Michigan.
Microplastics – a global problem
The term plastic refers to a wide variety of artificially created polymers. Polyethylene, or PET, is used for making bottles; polypropylene, or PP, is used in food containers; and polyvinyl chloride, or PVC, is used in pipes and tubes.
Microplastics are small plastic particles that range in size from 1 micrometer to 5 millimeters. The width of a human hair, for comparison, ranges from 20 to 200 micrometers.
Most scientific studies focus on microplastics in water. However, microplastics are also found in the air. Scientists know much less about microplastics in the atmosphere.
When scientists collect samples from the environment to study microplastics, they usually want to know more about the chemical identities of the microplastic particles found in the samples.
Plastic bottles are often made of polyethylene, while food containers usually containe polypropylene.
Anton Petrus/Moment via Getty Images
Fingerprinting microplastics
Just as fingerprinting uniquely identifies a person, scientists use spectroscopy to determine the chemical identity of microplastics. In spectroscopy, a substance either absorbs or scatters light, depending on how its molecules vibrate. The absorbed or scattered light creates a unique pattern called the spectrum, which is effectively the substance’s fingerprint.
Spectroscopy can match a substance with its unique fingerprint.
VectorMine/iStock via Getty Images Plus
Just like a forensic analyst can match an unknown fingerprint against a fingerprint database to identify the person, researchers can match the spectrum of an unknown microplastic particle against a database of known spectra.
However, forensic analysts can get false matches in fingerprint matching. Similarly, spectral matching against a database isn’t foolproof. Many plastic polymers have similar structures, so two different polymers can have similar spectra. This overlap can lead to ambiguity in the identification process.
So, an identification method for polymers should provide a measure of uncertainty in its output. That way, the user can know how much to trust the polymer fingerprint match. Unfortunately, current methods don’t usually provide an uncertainty measure.
Data from microplastic analyses can inform health recommendations and policy decisions, so it’s important for the people making those calls to know how…
