The Atomic Secret Behind Your Ultrasound Machine Was Hiding in Plain Sight

MIT has finally mapped the atomic structure of relaxor ferroelectrics — a class of sensor materials that has powered sonar, ultrasound, and microphones for more than half a century. For the first time, engineers can see inside the material they have been working with blindly since the 1970s.

A team from MIT mapped the structure using multislice electron ptychography, according to MIT News. The results, published in Science, show that polarization nanoregions inside the material are significantly smaller than the leading simulations predicted. More importantly, the work connects those structures directly to molecular dynamics calculations for the first time — validating the models, or in this case, exposing where they failed.

"Previously, these models basically had random regions of polarization, but they didn't tell you how those regions correlate with each other," co-first author Michael Xu, a postdoc at MIT, told MIT News. "Now we can tell you that information."

The material is a lead magnesium niobate-lead titanate alloy — a cousin of the perovskite family but not itself a perovskite, and unrelated to quantum computing. It belongs to a class called relaxor ferroelectrics, which have powered microphones, ultrasound probes, and sonar since the 1970s, MIT News reported. Their appeal is extreme sensitivity to electric fields and the ability to convert mechanical stress into electrical charge and vice versa. That made them ideal for precision sensors. It also made them nearly impossible to study with conventional microscopy, because their defining feature is chemical disorder — atoms arranged in ways that resist the clean, periodic patterns that standard imaging techniques require.

The MIT team used multislice electron ptychography, a method that scans a nanoscale beam of high-energy electrons across a sample and reconstructs three-dimensional atomic positions from overlapping diffraction patterns. Applying it to a disordered relaxor ferroelectric at sufficient resolution was new, according to Interesting Engineering. What it revealed was a hierarchical structure spanning individual atoms up to mesoscopic features — and a surprise about scale.

"We realized the chemical disorder we observed in our experiments was not fully considered previously," co-first author Menglin Zhu, also a MIT postdoc, told MIT News.

The finding that polarization nanoregions are much smaller than predicted is more than a data point. It is a correction to the foundation on which an entire class of sensor and actuator designs has been built. LeBeau frames it plainly: "If our models aren't accurate enough and we have no way to validate them, it's garbage in, garbage out."

That is a significant statement from a researcher whose institution has spent decades developing those models. It is also, implicitly, an indictment of the prior half-century of materials science work that relied on them without verification.

The practical implication is direct. Engineers designing the next generation of ultrasound machines, naval sonar arrays, and piezoelectric actuators now have a benchmark against which to test their simulations. Whether they act on it depends on how urgent the performance gap is — and whether the manufacturers of PMN-PT-based devices consider their current models good enough. For defense systems, where material performance has direct consequences, the answer may come from the procurement side rather than the research side.

The technique itself is the more durable contribution. MEP can be applied to other disordered materials systems — and there are many, including other functional oxides, battery electrode interfaces, and solid-state electrolyte structures. This work serves as a proof-of-concept template for a broader class of problems where materials were deployed empirically long before their behavior was understood, MIT News noted. The MIT team used MIT.nano facilities for the electron microscopy work; the simulations were run with collaborators at Rice University, the University of Pennsylvania, the University of Alabama at Birmingham, and KAIST.

What the paper does not yet tell us is how large the performance gap actually is in practice. The Science paper's exact nanoregion size measurements — in angstroms, compared against prior simulation values — were behind the paywall and unavailable for this article. That is the specific number an engineer would need to quantify how wrong their existing models are. The preprint version, posted to arXiv in August 2024, may contain those details, but it predates the final published version and may differ in significant ways.

The broader question — how many other material classes are operating on validated models versus inherited assumptions — is harder to answer. Relaxor ferroelectrics have been studied for over 70 years. They were not chosen for this work because they were the most mysterious; they were chosen because they are the most commercially embedded. A material you have used for five decades in deployed systems without understanding why it works is not a curiosity. It is a liability that has been quietly working.

That is the story here. Not the imaging technique, not the specific nanoregion size, not even the Science paper. The story is that the engineers building your medical imaging equipment, your submarine sonar, and your smartphone microphones have been working from models that were wrong at the atomic scale — and they did not know it until now.