Mycelium fills a microchannel, and freeze-drying turns it into a device

Grown rather than etched

Nanofluidics — steering ions through channels narrower than roughly 100 nanometres — normally begins in a cleanroom, with lithography and etching. A paper published online on 15 May 2026 in Nature Communications takes another route. The authors inoculated fungal mycelium into a microchannel, into a narrow gap and onto a metal mesh, let it grow, and then measured ion transport through whatever the fungus had built. Tian Li of Purdue University supervised the work; Qilong Cheng and Zhenyuan Niu contributed equally as first authors, with colleagues at Michigan State University, the University of California, Irvine, and Texas A&M University. The paper is open access, article 6461 in volume 17, and carries a version of record dated 17 July 2026.

One step in that sequence matters more than the headline. The hyphae grow alive, but every ion-conductivity measurement was made on dried material, freeze-dried at minus 57 degrees Celsius under a pressure below 20 Pa for more than 24 hours. Only then does the interior of a hypha — organelles and membranes suspended in cell fluid — become a nanoporous solid, and only then does the metabolic activity of living cells disappear. Live mycelium soaked in methylene blue, a cationic dye, did not stain at all; the same mycelium, freeze-dried, did. Nothing described here is a living fungus carrying an ionic current.

What was actually assembled

The microchannel was cast in PDMS silicone around a copper wire about 88 micrometres across, with an exposed run of roughly one centimetre serving as the channel. Liquid nutrient and 10 microlitres of culture went into a reservoir at one end, and the mycelium advanced through the channel along the oxygen gradient. Samples were incubated for more than ten days to give a dense hyphal network — a lower bound chosen for measurement consistency, not an optimum anyone determined.

Three cultivated species were used, bought as liquid cultures from a commercial supplier: Pholiota adiposa, Pleurotus ostreatus and Ganoderma sessile. No fruiting bodies and no wild collections appear anywhere in the study. Freeze-dried membranes of all three, immersed in potassium chloride at 1 × 10⁻² mol per litre and read on a streaming-potential analyser, gave zeta potentials between minus 12 and minus 18 millivolts at pH 7, corresponding to surface charges of minus 2.8 to minus 4.1 mC per square metre. That negative wall charge is what everything else rests on.

Each figure belongs to one device

Filling the 88-micrometre channel with P. adiposa raised ion conductivity by up to 467-fold against the same channel holding potassium chloride and no hyphae; the qualifier marks the peak of a concentration sweep rather than a fixed property. The empty channel behaved exactly like bulk solution, while the colonised one held a conductivity plateau near 0.1 mS per centimetre at potassium chloride concentrations below 1 × 10⁻³ mol per litre. Moving the bath from pH 5 to pH 9 increased conductance 3.0-fold. With one bath held at 1 × 10⁻⁴ mol per litre and the other raised to 1 × 10⁻¹, the current rectified only weakly, at 1.46 ± 0.06 — a ratio the authors themselves call modest and moderate, and attribute to the absence of concentration polarisation in a branching network. It is not an ionic diode to set beside a lithographically defined one.

The calcium experiment is the most striking result. Using the indicator Fluo-4FF, green-channel intensity in the colonised channel ran 55 ± 3 times higher for calcium at 10⁻⁶ mol per litre and 24 ± 2 times higher at 10⁻⁵ mol per litre than in the same channel without hyphae, lifting the signal-to-noise ratio from 2.3 to 99.7 and from 3.5 to 79.0 and improving the effective limit of detection by roughly an order of magnitude. It is enrichment, not capture or filtration: methylene-blue staining largely washed out again and the fluorescence faded after a rinse, which the authors read as reversible, transient accumulation in confined pathways.

Where measurement gives way to modelling

The three sizes the paper quotes — 49.7, 21.3 and 7.4 nanometres — are not widths anybody observed, and they are not a range for one device. They are equivalent channel diameters derived from a conductivity model, one for each species in its own geometry: P. adiposa in the microchannel, G. sessile across a gap of about 0.3 millimetres, and P. ostreatus through a stainless steel mesh with holes around 37 micrometres wide. For the gap and the mesh, the paper says the length and area of the conducting paths are neither well defined nor directly measurable, so a constant length-to-area ratio was assumed and a sensitivity analysis supplied; the microchannel needed no such assumption, because its length and cross-section could be measured directly. The authors’ own summary is the fair one: qualitative conclusions drawn from conductance trends are robust, while the quantitative values depend on the model.

Nor does a working sensor, chip or cell exist. Sensing, ion-based computing and energy conversion appear as prospects, with no power output or efficiency reported. Three of the authors have filed an invention disclosure, which is a filing rather than a granted patent or a product. And the abstract’s description of mycelium as the largest natural ion transport network in soil is the authors’ framing of why the idea is worth trying, not a quantity this study measured.

Something visible without instruments

The part worth watching needs no equipment. In the time-lapse that opens the paper, G. sessile hyphae travel about 0.67 millimetres along microchannels in four hours, and the discussion quotes a network fabrication rate of up to 0.167 millimetres per hour. The fungus enters the gap, climbs the wires of a steel mesh and pushes through by itself; anyone who has watched mycelium colonise a jar of substrate has seen the same capability, minus the electrodes at each end. The scale boundary is worth carrying away too: the gaps between hyphae run to tens of micrometres, far too wide for electric double layers to overlap, so the transport happens inside the hyphae. The full text and the four growth movies are mirrored in PubMed Central for anyone who would rather watch the colonisation itself.