Bacteria-bound sand: how ETH Zurich's biocement works and where it falls short

Researchers at ETH Zurich have turned loose sand into free-standing facade and partition elements at room temperature. They use a soil bacterium that grows limestone-like crystals between the grains. The work comes from architect Karen Antorveza Paez's doctoral research. It produced two digital manufacturing methods that bind sand into solid architectural elements. The headline claim is "building-grade", but the measured numbers say otherwise. By the team's own description the material behaves like mortar, not structural concrete, and its main by-product is still unsolved.

01How the bacteria cement sand

The chemistry is microbially induced calcite precipitation (MICP). Sporosarcina pasteurii breaks down urea using the enzyme urease. In a calcium-rich environment, the carbonate this releases precipitates as calcium carbonate. The resulting mineral crystals mechanically bind the grains together. Adding the crystals neither fuses nor melts the sand. They bridge contact points between grains, so strength depends on how much calcite forms and where it forms. Compressive strength of biomineralized specimens correlates positively with the amount of calcium carbonate precipitated.

None of this is new in principle. MICP has previously been used in self-healing concrete and to reduce soil erosion. The ETH contribution is to treat it as a fabrication step for a stand-alone material, rather than an improvement to existing ones.

02Two fabrication routes

3D biocement printing. The bio-ink is sand, living bacteria and a water-based gel. The gel contains alginate and xanthan. The gel gives the paste enough yield strength to hold its shape between nozzle and bath, and it carries the cells. After extrusion, the printed structure is immersed in a cementation bath of 0.5 M CaCl₂ and 0.5 M urea. Mineralisation takes two to five days, depending on the strength required. The part is then dried, and the bacteria are no longer active. Antorveza Paez likens the process to baking: the print is the soft dough, and the bacteria stand in for the oven.

Geometry matters because the reagents have to diffuse into the part. In a 2025 paper on digitally fabricated biocemented structures, the authors reported that the microorganisms produced more calcite in geometries with more surface exposure to the environment. They also noted that earlier 3D-printed MICP demonstrators were typically smaller than 10 cm, while their samples exceeded 20 cm in diameter. That favours thin, porous, lattice-like parts over solid blocks.

Robotic spraying. A robotic arm applies sand, bacteria and mineralisation substances in successive layers, with no immersion bath. The aim is surfaces rather than discrete objects. Antorveza Paez says printing suits architectural elements and spraying suits surfaces. Its selling point is automation: the process ran continuously for six months at the 2025 Venice Architecture Biennale. Spraying is the less characterised route. Printing is better understood as a material, and the spray process still needs work on its internal structure and mechanical performance.

The team's results are in three papers, including the printing study in npj Materials Sustainability. They appeared in npj Materials Sustainability, Architectural Intelligence and Biotechnology Design.

03What the strength numbers say

The printing paper gives the clearest data. For the bio-ink specimens, mean compressive strengths were 1.36 ± 0.46 MPa and 1.15 ± 0.38 MPa for the two biotic groups, against 1.06 ± 0.11 MPa for abiotic controls. The difference was not statistically significant. In other words, for that bio-ink the gel and sand packing were doing much of the work. In printed lattices, strengths ran from about 0.59–1.31 MPa in abiotic specimens to 0.52–2.18 MPa in biotic ones, depending on grid geometry and loading direction. Biocementation did raise strength and stiffness in sparse grids: compressive strength rose by 86% and 61% in two orientations. The authors concede that the strengths remain below those of structural cement-based materials.

Other MICP work shows what is possible with different protocols:

The last row is a different system. It reached unconfined strength above 50 MPa with 140 mm cementation depth by maximising aggregate packing density. The authors suggest prefabricated load-bearing components, where it could partially replace traditional concrete. That is a compacted, tuned specimen, not a printed, gel-supported lattice. The gap between ~2 MPa and 50 MPa shows that the binder chemistry can deliver far more. The limit lies in packing density, reagent transport and uniformity. ETH's open question on grain size fits this picture: sand particle size still affects final strength and porosity, so more testing is needed.

For context, one patent source states that high-strength concrete applications generally require at least 40 MPa. ETH's printed parts are one to two orders of magnitude below that. Hence the intended uses: facade elements, sun shading and partition walls.

04Carbon footprint: promising, not yet proven

The motivation is real. Cement production contributes around 7 to 8% of global CO₂ emissions. It needs kilns running above 1,000 °C. MICP avoids that heat, and one paper argues no process-related CO₂ is generated during biomineralisation. That claim comes with caveats. Whole-process results depend on reagents, curing and by-products:

• One life-cycle comparison for soil stabilisation estimated MICP at 4.3 kg CO₂ per tonne of stabilised soil, about 86% below cement stabilisation. The same paper found that in some applications MICP emissions exceed those of traditional materials, and raw-material preparation dominates. It also found nearly complete loss of compressive strength after 10 freeze–thaw cycles.
• A separate assessment of a related enzyme-based method found 3% lower global warming potential than Portland cement in soil stabilisation. That is a far smaller benefit.

These are soil-stabilisation studies, not assessments of printed architectural components. No life-cycle figure for the ETH process appears in the sources I could find.

The ammonium problem is the sharpest constraint. Chemically, the ureolytic reaction yields two moles of ammonium per mole of calcium carbonate precipitated. It is a pollutant, particularly for water. Antorveza Paez says that without handling the by-product, biocement is "unfortunately not sustainable." A baths-of-reagent process makes this worse, because each cubic metre of printed part leaves a volume of ammonium-laden liquor. ETH is now examining the wider environmental impact and ways to manage or avoid the ammonium, including treatment, recovery and alternative biological pathways. A third study on this is under way.

05What remains unproven

Four things stand between this and construction use:

1. Strength. Printed parts are around 1–2 MPa. Closing the gap to structural strengths means better packing, deeper and more uniform cementation, and probably different geometries.
2. Durability. Reported freeze–thaw sensitivity in related MICP soil work is a flag for exterior facades in cold climates. The ETH papers' long-term weathering performance was not available in the sources I reviewed.
3. By-product handling. Without a credible ammonium treatment or recovery route, the low-carbon case does not hold.
4. Scale and cost. Printed demonstrators now exceed 20 cm. Full-height panels, multi-day baths and reagent logistics are untested at building scale. One sand-stabilisation comparison lists high cost as a downside of MICP.

The result is best read as a credible, heat-free route to mortar-class, geometrically complex mineral parts such as shading, screens and partitions. It is not yet a cement replacement. The strength record from other labs suggests it could get stronger. The ammonium accounting will decide whether it is also cleaner.