July 17, 2025
From seaweed to structural material: A seaweed called Ulva (righthand petri dish) is dried (center), powdered (left) and then mixed directly in with traditional cement (beaker). The darker cement cube (top center) contains 5% seaweed by weight.Mark Stone/University of Washington
The modern world is built with concrete: Humans use more concrete annually than any other material besides water. Yet cement, the key component of concrete, is the source of as much as 10% of all carbon dioxide emissions worldwide.
To address this problem, researchers at the University of Washington and Microsoft developed a new type of low-carbon concrete by mixing dried, powdered seaweed with cement. The seaweed-fortified cement has a 21% lower global warming potential while retaining its strength. And thanks to an assist from machine learning models, the team arrived at this new formulation in a fraction of the time that such work would ordinarily take.
The team published its findings July 8 in Matter.
“Cement is everywhere — it’s the backbone of modern infrastructure — but it comes with a huge climate cost,” said senior author Eleftheria Roumeli, a UW assistant professor of materials science and engineering. “What makes this work exciting is that we show how an abundant, photosynthetic material like green seaweed can be incorporated into cement to cut emissions, without the need for costly processing or sacrificing performance.”
UW doctoral student Meng-Yen Lin casts green cement samples into molds to cure and later test their structural properties.Mark Stone/University of Washington
Producing one kilogram of cement emits nearly a kilogram of CO2. Most of those emissions come from the fossil fuels used to heat raw materials and from a chemical reaction called calcination that occurs during the production process. Seaweed, in contrast, is a carbon sink: It pulls carbon out of the air and stores it while it grows. And, remarkably, it can directly replace some of the cement in concrete, giving the result a dramatically smaller carbon footprint.
Arriving at the ideal mixture of ingredients would have taken five years of trial and error, Roumeli estimated, because any concrete sample takes about a month to fully cure before its properties can be evaluated accurately.
To speed up the process, the team built a custom machine learning model and trained it on an initial set of 24 formulations of cement. They then used the model to predict ideal mixtures to test in the lab. By feeding the results of those tests back into the model, they were able to work in tandem with the model and move through formulations rapidly. The outcome was an optimal mixture of seaweed-enhanced cement with a reduced carbon footprint that passed compressive strength tests, discovered in just 28 days.
UW doctoral student Meng-Yen Lin tests the compressive strength of a cement cube to determine how the addition of seaweed is affecting its performance as a building material.Mark Stone/University of Washington
“Machine learning was integral in helping us dramatically shorten the process — especially important here, because we’re introducing a completely new material into cement,” Roumeli said.
From here, the team plans to deepen their understanding of how seaweed composition and structure affects cement performance. The larger goal is to generalize the work out to different kinds of algae (or even to food waste) so that producers can create local, sustainable cement alternatives around the world — and use machine learning to optimize them rapidly.
“By combining natural materials like algae with modern data tools, we can localize production, reduce emissions, and move faster toward greener infrastructure,” Roumeli said. “It’s an exciting step toward a new generation of sustainable building materials.”
Additional co-authors on this paper are Meng-Yen Lin, a UW doctoral student studying materials science and engineering; Paul Grandgeorge, a former UW postdoctoral researcher in the materials science and engineering department who is now an R&D engineer at the iPrint Institute; and Kristen Severson, a principal researcher at Microsoft Research.
This research was funded by Microsoft Research.
For more information, contact Roumeli at eroumeli@uw.edu.
Tag(s): College of Engineering • Department of Materials Science & Engineering • Eleftheria Roumeli