For decades, scientists have studied how frost spreads across surfaces, from airplane wings to heat exchangers. At the heart of this process lies a deceptively simple event: two neighboring droplets freezing and connecting through a tiny “ice bridge.” Once formed, this bridge allows freezing to rapidly propagate, turning isolated droplets into a continuous frost layer.
The mechanism seemed well understood. Ice bridges were believed to grow along the surface, crawling from one droplet to the next. This picture shaped how we model frost propagation and how we design surfaces to delay it.
But what if that picture was incomplete?
A question hiding in plain sight
Our journey began with a simple question: where exactly does an ice bridge grow in three-dimensional space?
Most previous studies relied on top-view imaging. From above, an ice bridge appears as a line connecting two droplets, but this view compresses a three-dimensional structure into two dimensions. It cannot tell whether the bridge is attached to the surface or suspended above it. In principle, side-view imaging could resolve this. In practice, it is nearly impossible. Ice bridges are only a few micrometers thick, they form unpredictably, and they exist for only a few seconds. Capturing such an event within the narrow focal depth of a high-magnification microscope is like trying to photograph a fleeting thread in midair.
For years, this limitation went largely unresolved. The assumption that ice bridges grow along the surface persisted.
Seeing what was previously invisible
To overcome this challenge, we turned to focal plane shift imaging (FPSI), a technique that allows us to infer vertical position by carefully scanning the focal plane.
What we observed was unexpected.
On superhydrophobic surfaces, the ice bridge did not lie on the substrate. When the bridge appeared sharp, the surface was out of focus. When the surface came into focus, the bridge blurred. The two were clearly separated in height. The bridge was not crawling along the surface. It was suspended in air. This “out-of-plane” growth mode stood in stark contrast to the conventional “in-plane” bridges observed on hydrophilic surfaces, where the bridge and substrate remain in the same focal plane, shown in the below picture.
A long-standing assumption had quietly broken.
From observation to mechanism
Once we realized that two distinct modes exist, the next question was obvious: what determines which mode occurs?
The answer lies in surface wettability. By systematically studying surfaces ranging from hydrophilic to superhydrophobic, we found that the ice bridge gradually lifts away from the surface as the contact angle increases. Around an advancing contact angle of approximately 105°, a clear transition occurs: suspended bridges begin to dominate.
But why?
The key is geometry. Surface wettability reshapes droplets, which in turn reshapes the shortest path for vapor diffusion between them. Ice grows along this path. On hydrophilic surfaces, droplets spread, and the shortest diffusion pathway lies near the substrate. The bridge forms along the surface. On superhydrophobic surfaces, droplets are more spherical, and the shortest path shifts upward, into the air between them. The bridge follows, forming a suspended structure.
What appears as a change in morphology is, at its core, a reconfiguration of vapor transport pathways.
Slower bridges, slower frost
This geometric shift has profound consequences.
Frost propagation is not controlled by how fast a droplet freezes, but by how quickly ice bridges grow. While droplet freezing occurs in milliseconds, bridge growth takes seconds. Suspended bridges grow significantly more slowly than their surface-attached counterparts. Unlike in-plane bridges, which remain thermally coupled to the cold substrate, suspended bridges are partially insulated by air. As they extend, their tip temperature rises, reducing the vapor pressure difference that drives growth.
In effect, the bridge slows down. This microscopic delay accumulates across many droplets, leading to a dramatic reduction in frost propagation speed. Our experiments show that frost spreading can be reduced by over 80% on superhydrophobic surfaces.
From microns to machines
One of the most rewarding aspects of this work was seeing the mechanism translate across scales. When applied to commercial finned-tube or microchannel heat exchangers, the suspended ice bridge delays frost formation, slows its spread, and prolongs efficient heat transfer operation. A microscopic shift in how ice connects droplets becomes a macroscopic improvement in system performance. This closes a loop that is often difficult to achieve: from fundamental physics to engineering relevance.
Rethinking frost formation
Looking back, what is striking is not just the discovery itself, but how long it remained hidden. The suspended ice bridge was not an exotic phenomenon requiring extreme conditions. It was present all along, simply overlooked because of how we chose to observe the system. This work suggests that frost propagation is not purely a two-dimensional process, as often assumed, but inherently three-dimensional, governed by the interplay of geometry, vapor transport, and thermal coupling. By revealing this hidden dimension, we hope to shift how the field thinks about freezing propagation, from surface-bound pathways to spatially distributed ones. And perhaps more broadly, it serves as a reminder that even in well-studied systems, there are still new perspectives waiting just beyond the focal plane.