Most of the limestone on Earth was never fired. Living things grew it, one shell and one crust at a time. Microbial carbonates still cement beachrock on warm coasts, and stromatolites laid down layers for billions of years, although it is still debated whether microbes or chemistry alone built the oldest candidates. At ETH Zurich, the architect Karen Antorveza Paez has now automated rock that microbes build.

ETH News reported on 14 October 2026, in a piece by Corinne Landolt, that Antorveza Paez completed three studies during her doctorate at D-ARCH. She worked with Benjamin Dillenburger’s Digital Building Technologies chair and André R. Studart’s Complex Materials group. Her subject is microbiologically induced calcite precipitation (MICP), “or simply biocementation”. In her words: “We’re using bacteria to produce a sort of mineral glue.”

The chemistry, held under arithmetic

The bacterium is Sporosarcina pasteurii. Its enzyme, urease, splits urea and releases carbonate. The carbonate meets calcium from a calcium chloride bath and precipitates as calcite between the sand grains. Stoichiometry sets the cost: every mole of CaCO₃ comes with 2 mol NH₄⁺. The comparison is the kiln. ETH puts cement firing at over 1,000 °C, CEMBUREAU gives 1,450 °C for clinker, and Carbon Brief puts cement at about 8% of global CO₂ emissions. Antorveza Paez’s parts dry at room temperature, with “No heat sources,” as ETH writes.

What was actually measured

The first paper appeared in npj Materials Sustainability on 7 May 2026. A UR10e robot prints a bio-ink of sand, alginate–xanthan gel and live bacteria through a 4 mm nozzle. The part then mineralises in a 0.5 M CaCl₂ + 0.5 M urea bath that is renewed every day, for 2 to 5 days depending on the target strength, according to ETH. Printed grid cubes reached up to 2.18 MPa in compression. In bending, beams made with bacteria held 2.74 ± 1.15 MPa, against 0.13 ± 0.08 MPa for the abiotic control. The paper calls this an “~21-fold increase”. The wide error bar shows a living process that has not yet settled into a narrow band. The demonstrator is a double-curved TPMS facade element, 30 × 20 × 20 cm, printed in 40 minutes. It is the largest part made with this method so far. The authors place their values “within the range of other porous mineral systems, such as clay foams and rammed earth.” So the realistic uses are a screen, a sunshade or a partition. Not a column.

The second paper, in Architectural Intelligence (21 May 2026), drops the bath. A robot arm sprays bacteria, xanthan and sand in cycles modelled on how stromatolites grow. At the 2025 Venice Biennale, the installation “Geological Microbial Formations” in the Arsenale ran for six months on a 1.2 × 0.5 m demonstrator. It grew about 4 cm a month, to a peak of 24 cm, and the material was 77–86% CaCO₃. The toolpaths came from a differential-growth algorithm in Grasshopper and Daniel Piker’s Kangaroo Physics. The paper reports no MPa values and calls its relevance non-structural.

The credit goes well beyond one name. Robert O. Kindler co-authored both papers. Medusoil SA of Etoy, Vaud, led by co-author Dimitrios Terzis, supplied about 100 litres of S. pasteurii culture. Few architects know this Swiss microbiology supply chain exists. Alessandro Dutto, Tazio Pleij, Gustavo Zorzetto, CheWei Lin and Fergal B. Coulter did much of the lab and robot work. The microbe was already used to stabilise soil. Turning it into a process you can print and spray is the contribution, and probably the harder half.

The by-product is part of the material

The trade-off is plain. You swap a kiln running above 1,000 °C for a bath that leaves two ammonium ions behind for every carbonate it locks in, and you get strengths in the rammed-earth range. Antorveza Paez names this limit herself: “If we don’t factor the disposal of this chemical by-product into our designs, then unfortunately using biocement for construction won’t be such a sustainable practice.” Her third study, on environmental impact, is still in progress and unpublished. Building has seen this pattern before: the materials that hurt us were scaled up before anyone understood how they failed. Here the builder named the weak point first. Keep the ammonium line in every specification conversation.

←TODAY: A 30 × 20 × 20 cm facade element, grown in a daily-renewed bath, up to 2.18 MPa, dried at room temperature. →3012: Zurich walls specified by cure days and a by-product ledger, not by kiln temperature. Fulcrum: the bath water belongs to the material.

Atelier: For a Zurich studio working on a facade or fit-out brief, biocement is not something to tender yet. It is a non-load-bearing option to understand before a client asks about it. For offices already working in Grasshopper, the toolpath now decides how a part cures, not only its shape. Monday move: add a process-ledger row to your material options sheet (cure temperature, cure days, by-product per kg) and fill it in for one bio-based and one fired option on a current project.

Hack: Price the ammonium before you price the panel. Breaking down urea fixes the ratio at 2 mol NH₄⁺ per mol CaCO₃, so four lines of GhPython or plain Python give the theoretical upper limit per kilogram of calcite.

M_CACO3, M_NH4 = 100.09, 18.04  # g/mol
def nh4_kg(caco3_kg):
    return caco3_kg * 2 * M_NH4 / M_CACO3  # 2 mol NH4+ per mol CaCO3
print(round(nh4_kg(1.0), 2))  # 0.36 kg NH4+ per kg calcite, upper bound

Multiply a part’s mass by its CaCO₃ fraction and feed in the result. This is stoichiometry, not a measurement: the real amount depends on how much urea the bath supplies and how much carbonate actually precipitates.

Read the methods section of the npj paper before your next facade workshop, and bring the 2:1 ratio along with the rammed-earth strength range.

Sources & Further Reading

• Primary: ETH News — When bacteria become masons
• Reinforcing: npj Materials Sustainability — 3D biocement printing: scaling up living mineral structures