17 Oct 2026 · Materials

Electrodes that fish lithium and salt out of seawater

A computer screen of 20,414 known compounds finds the few battery-style electrodes that should survive in seawater and soak up ions at safe voltages — some picky enough to grab only lithium, one broad enough to help make fresh water.

Sougat Purohit, Andreas Kuhlmann, Marten Huck, Hans-Georg Steinrück & Gopalakrishnan Sai Gautam

Why it matters

Seawater and salty brines hold both drinkable water and the lithium our batteries need, yet today’s main routes are slow, water-hungry evaporation ponds or membranes that cannot single out one element. By screening 20,414 known materials for ones that survive seawater and grab ions at safe voltages, this study gives experimenters a short list of 11 ion-selective electrodes — including LiFeP₂O₇, Fe₂(MoO₄)₃ and Li₃Fe₂(PO₄)₃ for lithium — plus olivine FePO₄ as the closest thing to an all-round desalination electrode.

The paper

Identifying intercalation electrodes for electrochemical ion recovery and desalination

Sougat Purohit, Andreas Kuhlmann, Marten Huck, Hans-Georg Steinrück & Gopalakrishnan Sai Gautam (Indian Institute of Science; Forschungszentrum Jülich; RWTH Aachen). arXiv:2610.20834 [cond-mat.mtrl-sci], submitted 14 Oct 2026. Read on arXiv · PDF

In plain language: an “intercalation” electrode works like the positive side of a lithium-ion battery. Push its voltage down and ions from the water slip into gaps in its crystal; raise it and they come back out into a small rinse stream. Pick an electrode that only lets lithium in and you are mining lithium; pick one that swallows every kind of ion and you are making fresh water. The catch is that water only tolerates a narrow voltage range before it splits into hydrogen or oxygen, and many good battery materials quietly dissolve in salty water.

The team pulled Li, Na, K, Mg and Ca compounds containing one to three redox-active metals and one to three anions from the Materials Project database, built “seawater Pourbaix diagrams” (maps of what is stable at each pH and voltage, with Cl, Na, Mg and S added to the water), and kept only electrode pairs that (1) decompose with at most 0.3 eV per atom of driving force into a protective solid that is not a pure element, (2) switch ions at an average voltage inside the seawater window across pH 7–8.5, and (3) stay metastable over at least 80 % of that window — the benchmark set by olivine LiFePO₄ (about 88 %). Graphs linking structures that share the same host then reveal which hosts take only one ion (for recovery) and which take several (for desalination). Their voltages match experiment where known — e.g. 0.46 V and 0.29 V vs SHE for Li and Na in olivine FePO₄, and 3.06 V vs Li predicted against a 2.92 V vs Li measured plateau for LiFeP₂O₇.

Try it

Voltages, metastability areas and screening counts below come from the paper (Table 1, Figs. 2 and 4). The water-window lines use the textbook Pourbaix equations with the paper’s −0.3 V / +0.5 V overpotential shifts. The ion-uptake curves are an illustrative ideal-solution (Nernst / competitive lattice-gas) model built on those voltages — not the authors’ code, and it ignores kinetics, capacity and structural change.

1 · Does the electrode work inside seawater’s safe voltage window?

Water pH

Hydrogen overpotential (V)

Oxygen overpotential (V)

Usable window at this pH (V vs SHE)

Window width at this pH

Paper pairs inside window, pH 7–8.5

Voltage vs pH: water stability and the paper’s shortlisted electrode voltages

LiNaKMgCaWater lines (dashed: ideal, solid: with overpotential)

2 · Can voltage control make one electrode pick one ion?

Host electrode (from Table 1)

Applied potential (V vs SHE)

Use typical seawater ion concentrations (Li 0.026 mM, Na 469 mM, Mg 53 mM, Ca 10 mM, K 10 mM — illustrative, not from the paper)

Ion filling most sites (toy)

Its share of captured ions (toy)

Voltage gap to next ion (paper)

Metastability area (paper)

Fraction of host sites filled by each ion vs applied potential (toy model)

3 · How hard the screen was (counts from the paper, Figs. 2b and 4)

Step	Li	Mg	Ca	Na	K
Structures queried from Materials Project	12,536	3,598	1,344	1,783	1,153
Metastable in seawater somewhere	10,131	2,437	780	1,185	559
Pairs metastable at their own voltage	754	195	95	68	21
…across the whole pH 7–8.5 range	573	140	65	49	15
…with ≥ 80 % metastability area	69	13	10	10	3
…with at least one experimentally known phase	8	4	4	7	0

The maths in this page

Water window. Water is thermodynamically stable between the hydrogen- and oxygen-evolution lines of a Pourbaix diagram. The paper widens it by overpotentials of 0.3 V (H₂) and 0.5 V (O₂), turning about 1.23 V into about 2 V, and takes seawater as pH 7–8.5:

E_H₂(pH) = −0.0592·pH − η_H E_O₂(pH) = 1.229 − 0.0592·pH + η_O (V vs SHE, 25 °C)

An electrode pair passes the voltage test only if its average intercalation voltage sits inside the narrowest window over pH 7–8.5, i.e. between E_H₂(7) and E_O₂(8.5) — about −0.71 V to +1.23 V with the paper’s overpotentials.

Average intercalation voltage (standard DFT recipe the paper uses with Materials Project energies; for ion A of charge z going from host composition A_x₁H to A_x₂H):

V̄ = − [ E(A_x₂H) − E(A_x₁H) − (x₂ − x₁)·μ_A ] / [ (x₂ − x₁)·z·e ]

Converting to the SHE scale shifts by the metal’s standard potential (e.g. 3.06 V vs Li ≈ 0.02 V vs SHE for LiFeP₂O₇).

Metastability rules. Decomposition energy ≤ 0.3 eV/atom into a non-elemental solid; intercalated and empty phases must overlap over ≥ 80 % of the seawater window. Hosts whose ions switch at voltages more than 0.1 V apart can be made selective by holding the potential (potentiostatic control); closer than 0.1 V, they tend to take both ions at once.

Toy uptake model (illustrative). Each ion i competes for the same sites with Nernst-weighted activity

a_i = (c_i/c°) · exp[ −z_iF(E − V̄_i) / RT ], θ_i = a_i / (1 + Σ_j a_j)

so an ion floods in once the applied potential E drops below its voltage, and a 0.1 V gap gives roughly a 50-fold preference for a +1 ion at 25 °C (RT/F ≈ 25.7 mV). Concentrations default to 1 M (c/c° = 1); the seawater toggle shows how scarce lithium (≈ 26 µM) is swamped by sodium unless the voltage gap is large. Real electrodes also depend on kinetics, site capacity and phase changes, which this toy ignores.

Illustrative teaching model built on published voltages — not the authors’ SPICES workflow and not a process design.

How to use it

In panel 1, slide the pH and the two overpotentials: the shaded band is where the electrode can work without splitting water, and the coloured ticks are the paper’s shortlisted electrode voltages over seawater pH. Set both overpotentials to zero to see how few electrodes would fit in the ideal 1.23 V window. In panel 2, pick a host — try olivine FePO₄ (Li 0.46 V vs Na 0.34 V) and β′-FePO₄ (Li 0.36 V vs Na 0.30 V) — then move the applied potential and watch which ion fills the sites first; a gap above 0.1 V lets voltage control pick one ion, a smaller gap makes the host take both, which is what you want for desalination. Tick the seawater toggle to see why lithium, about 18,000 times scarcer than sodium in the sea, needs a strongly lithium-selective host. Panel 3 shows how the 20,414 starting structures shrink to a handful of tested-in-the-lab candidates.