MIT Study Finds 1-μm Wrinkle Cuts GaN Thermal Conductivity by 4–5×

A microscopic wrinkle measuring just one micrometer across slashes thermal conductivity in gallium nitride by a factor of four to five, according to new findings from MIT scientists. The defect also lowers thermal boundary conductance at the GaN-silicon interface by approximately 25 percent, to roughly 2.12 × 10⁷ W/m²·K.

The study, published in Nature Communications, shows that such imperfections do not merely slow heat transfer—they distort it directionally. Heat approaching the wrinkle from one side behaves differently than heat arriving from the opposite direction, transforming an otherwise uniform thermal pathway into a localized, asymmetric bottleneck.

Mingda Li, associate professor of nuclear science and engineering at MIT, stated: “I think overheating has become the real bottleneck in device performance.” He emphasized that conventional modeling assumes perfect crystals, yet large wrinkles are common in two-dimensional materials—and their thermal impact had remained unquantified until now.

Researchers mapped this disruption using a novel combination of laser pulses and ultrafast X-ray diffraction. A laser rapidly heated the material, and time-resolved X-ray measurements tracked atomic lattice shifts caused by thermal expansion—changes that directly correlate with heat propagation.

This method overcomes key limitations of existing thermal imaging tools. Time-domain thermoreflectance, for example, delivers only surface-weighted optical signals and cannot resolve buried layers. Infrared cameras lack the spatial and temporal resolution needed for nanoscale, sub-nanosecond thermal events.

Seeing heat through layered structures

As transistor density rises, managing heat across multilayered devices grows more complex. Real-world chips often contain five or more stacked layers, each contributing to overall thermal resistance—but conventional techniques struggle to isolate behavior within individual strata.

Jeehwan Kim, MIT professor and co-author, explained: “Because that technique uses optics, it doesn’t allow you to study different layers. Real devices have five or more layers. It also only provides an overall signal, and that makes it hard to see thermal transport happening in layers buried under the surface.”

In contrast, X-rays penetrate deeply and interact with atomic lattice spacing. By recording diffraction patterns at multiple positions and time delays after laser excitation, the team reconstructed how heat propagated both laterally and vertically—specifically within a 500-nanometer-thick GaN film and across its interface with silicon.

The wrinkle’s asymmetric thermal effect

The experiment used a GaN-on-silicon sample containing a naturally occurring wrinkle approximately one micrometer wide. Scanning across the defect revealed a local thermal conductivity of 21.2 ± 1.2 W/m·K—four to five times lower than in adjacent, unwrinkled regions.

Crucially, the reduction was not isotropic. Heat flow exhibited marked asymmetry: thermal transport differed depending on whether heat arrived from the left or right side of the wrinkle. This directional distortion confirms that isolated defects can impose vector-specific constraints on phonon pathways.

“People never even knew how much heat is blocked by these wrinkles. Those are things we can now directly observe with this technique,” Li said. The method captures localized effects without averaging them into bulk material behavior—a critical distinction for diagnosing hotspot origins.

Toward defect-aware thermal engineering

The researchers envision the technique supporting next-generation thermal design, particularly for AI hardware, high-power electronics, and flexible devices where compactness intensifies heat concentration risks.

While currently confined to synchrotron-based X-ray facilities and demonstrated on GaN-on-silicon—not full commercial chips—the approach is adaptable to other material systems. A semiconductor industry consortium has already expressed interest in applying it to diverse chip architectures.

Ultimately, the work shifts focus from treating electronic materials as homogeneous blocks toward identifying and characterizing the specific microscopic features—wrinkles, interfaces, grain boundaries—that govern where heat accumulates. For increasingly miniaturized electronics, locating those hidden bottlenecks may prove essential to sustaining performance and reliability.