The science of an aqueous magnesium-ion battery

The premise

A battery is a controlled, reversible chemical reaction. Two electrodes (one chemically hungry for electrons, one full of them) are connected by an external wire and separated by an electrolyte — a liquid or gel that lets ions move between the electrodes but blocks electrons. When the cell discharges, electrons flow out through the wire (doing useful work along the way: lighting your house, running a server rack, powering a forklift) while ions move through the electrolyte to balance the charge. When the cell charges, an external voltage forces both flows to run in reverse, restoring the original state.

Three things make a stationary-storage battery good. Energy density — how many kWh you can pack into a given volume or weight. Power density — how fast you can deliver those kWh on demand. And cycle life — how many full charge-discharge cycles the cell survives before its capacity decays past usefulness. For an electric vehicle, energy density dominates (every kg matters). For grid storage in a warehouse next to a substation, weight is essentially free; the dominant metric is cycle life multiplied by safety. Aqueous magnesium-ion is positioned for the second case.

Lithium-ion in one breath

In a lithium-ion cell, the working ion is Li⁺ — a single positive charge on the smallest metal atom. The cathode is a layered metal oxide (LiCoO₂, LiFePO₄, NMC). The anode is graphite. The electrolyte is a lithium salt dissolved in an organic solvent — typically LiPF₆ in ethylene carbonate / dimethyl carbonate. That solvent is flammable. The cell runs at ~3.7 V — well above water’s 1.23 V electrochemical window. At that potential water would split into H₂ and O₂, so the electrolyte must use a non-aqueous (and flammable) organic solvent instead.

Lithium-ion is extraordinary at energy density (~250 Wh/kg in modern NMC packs) and that’s why it won the EV market. The downsides: thermal runaway (when a cell is punctured, the flammable solvent ignites and the energy density that made the pack useful now makes it dangerous), cycle life ceiling (~3,000–6,000 cycles for the best LFP stationary chemistry), and supply chain (lithium concentrated in Chile, Argentina, Australia, China; cobalt in DRC; the geopolitics is well-rehearsed).

For stationary grid storage — where energy density is nearly irrelevant and cycle life is everything — lithium-ion is overqualified on weight and underqualified on safety and longevity. It is winning the 2020s market because nothing else was ready. The 2030s will look different if a longer-lived, non-flammable, commodity-fed chemistry can reach manufacturable scale.

What changes when water is the electrolyte

Water has an electrochemical stability window of about 1.23 V. That means: if you push a cell above ~1.23 V (or below the corresponding negative potential), water itself begins to split into hydrogen and oxygen gas at the electrodes. Either gas is a problem. Hydrogen is flammable; oxygen accelerates corrosion. The first century of aqueous batteries (lead-acid, NiCd, NiMH) all live inside this window or work around the gassing with vented cells.

To get a useful battery from water, you either: (1) accept the low voltage and design around it (lead-acid runs at 2.05 V per cell despite gassing, because the gassing is slow and the cell is vented), or (2) use a chemistry where the working ion has a redox potential far enough from water’s splitting potentials that you can run higher voltage without electrolyzing the solvent. The CityU paper’s 2.2 V full-cell window is achieved by combining a cathode and anode whose redox potentials sit just inside — not exceed — the practical kinetic limits of water splitting in concentrated MgCl₂. Concentrated salts narrow the activity of free water, which broadens the practical electrochemical window beyond the thermodynamic 1.23 V.

The payoff for staying inside water’s window: the electrolyte cannot burn. No organic solvent, no LiPF₆, no thermal runaway. Inherent safety is not a feature added by engineering; it’s a property of the chemistry choice itself.

Why magnesium

Magnesium is the eighth most abundant element in Earth’s crust and the third most abundant dissolved cation in seawater (~1.3 g/L). It costs ~$300/ton for industrial-grade MgCl₂; lithium carbonate is $9,000–17,000/ton. There are no rare-earth supply chains involved, no Congolese cobalt, no Chilean salar evaporation. The U.S. has domestic Mg production (Intrepid Potash, NM), and Europe has its own at Nedmag in the Netherlands. The supply chain doesn’t bend to a single country’s policy.

The other reason magnesium matters: it’s a divalent cation — Mg²⁺, two positive charges per ion, versus lithium’s single charge. Each Mg²⁺ that intercalates into a cathode lattice carries twice the charge of a Li⁺. In principle this means higher theoretical capacity per ion moved. In practice it also means stronger coulombic interactions with the host lattice, which is harder on the lattice and is exactly why most cathode materials studied since the 1990s have failed to host Mg²⁺ reversibly. The CuFe-PBA cathode (chapter 6) is one of a small number that does.

Magnesium also doesn’t dendrite the way lithium does. Lithium plating during fast charge grows whisker-like dendrites that can short across the cell separator and trigger thermal runaway. Mg²⁺ intercalation chemistry that doesn’t plate metallic Mg sidesteps that failure mode entirely.

The cell, in three pieces

An aqueous Mg-ion cell has the same three parts every battery has: anode (where electrons leave the cell during discharge), cathode (where they re-enter), and electrolyte (the ionic-conduction medium between them). What’s specific to this chemistry is the choice of material at each position.

The cathode is a Prussian-blue analogue — copper hexacyanoferrate, Cu₃[Fe(CN)₆]₂, abbreviated CuFe-PBA. It’s an open framework of iron-cyano octahedra bridged by copper, with large open channels that magnesium ions can slip into and out of without breaking the framework. (Chapter 6 unpacks this in detail.)

The anode is a Covalent Organic Polymer (COP) — a class of materials with conjugated π-electron backbones that accept and release electrons reversibly, with Mg²⁺ coordinating to the polymer’s functional groups during the redox cycle. Polymer anodes are unusual in aqueous batteries; they’re chosen here because they’re flexible (mechanically forgiving of volume changes during cycling) and tunable in voltage.

The electrolyte is a mixed magnesium/calcium chloride brine in water — dominated by MgCl₂ (~5.8 mol/L at 25°C) with a substantial CaCl₂ fraction — held at pH 7.0±0.5 throughout cycling. The concentration is high enough that there are roughly more Mg²⁺ and Ca²⁺ ions than free water molecules in the solution. (Chapter 7.)

Current collectors on each electrode are thin metal foils that conduct the electrons to and from the external wire: titanium foil at the cathode (Ti is electrochemically inert in the brine) and SS 316L foil (25 µm) at the anode. The cell housing is sealed; the only thing leaving the cell during operation is electrons through the terminals.

The Prussian blue analogue

Prussian blue itself — ferric ferrocyanide, Fe₄[Fe(CN)₆]₃ — is the pigment first synthesized accidentally in 1704 in a Berlin paint shop. It’s a three-dimensional cubic lattice in which iron atoms sit at the corners and cyanide ions (CN⁻) bridge them. The cube has open channels through its center, roughly 3–5 Å across — large enough for small alkali and alkaline-earth cations to enter, sit, and leave without disturbing the cage.

A Prussian-blue analogue (PBA) is the same architecture with one or both of the metals swapped — copper replacing some of the iron, manganese replacing others, and so on. CuFe-PBA is the specific variant chosen by the CityU group: copper at the high-spin metal site, iron at the cyano-bridged site. The crystal structure looks like a microscopic warehouse of cubic rooms, each room sized to fit exactly one Mg²⁺ ion plus a few solvating water molecules.

Why does this matter for cycle life? Most cathode materials suffer structural collapse after a few thousand cycles — the host lattice can’t accommodate the volume change that comes with repeated ion insertion/extraction, so it cracks, isolates pockets of active material, and loses capacity. PBA’s open framework absorbs the volume change without cracking: the cubes flex but don’t shatter. The lattice doesn’t care that the ion came back; the room was already waiting.

This is the structural answer to why aqueous Mg-ion can target 120,000 cycles — about twenty times more than the best stationary LFP. The chemistry doesn’t fight itself.

The dual-cation brine, neutral pH

Conventional aqueous electrolytes — like the 1 M sulfuric acid in a lead-acid battery — are dilute. Most of what’s in solution is free water: water molecules not bonded to ions, available to participate in electrolysis at the electrode surfaces. That’s why the practical electrochemical window for such dilute electrolytes is close to the thermodynamic 1.23 V limit of water.

At saturation — a concentrated brine dominated by MgCl₂ (~5.8 mol/L) with a substantial CaCl₂ fraction at room temperature — most water molecules are coordinated to magnesium, calcium, or chloride ions. Each Mg²⁺ or Ca²⁺ typically organizes 6–8 water molecules into its hydration shell; each Cl⁻ coordinates 2–3. Free, electrochemically-available water is rare. This is sometimes called a "water-in-salt" electrolyte, a term coined in the lithium aqueous-battery literature (Suo et al., 2015) for highly concentrated systems. The practical voltage window opens up by roughly 1 V — enough to run the 2.2 V CuFe-PBA / COP couple without ripping water apart.

The pH 7.0±0.5 operating range is the other story. Acidic electrolytes corrode the metal current collectors; alkaline electrolytes degrade the PBA framework. Neutral pH is the chemistry’s sweet spot, and it’s also why the electrolyte is environmentally non-hazardous: a leak is salt water, food-grade. The Phase 1 protocol monitors pH continuously across the 12-month cycle campaign with Rosemount 372 pH probes; the kill threshold is pH outside 4.0–8.0 (operational target 4.91–7.02).

Which ions get in

Here is the question that reframes everything about purification. A real brine — even a carefully made one — is never a single salt. Alongside the magnesium and calcium there are traces of sodium, potassium, sometimes rubidium, barium, and lithium: every one a cation looking for a home. The CuFe-PBA cathode is an open framework of cubic rooms (chapter 6). Which of these cations does the framework actually let in, and in what order? The property that governs the answer is called selectivity.

Measured against the CuFe-PBA lattice, the affinity runs in a definite sequence — and it contains a surprise:

Read that carefully. The framework likes the monovalents best. Rubidium, potassium, and sodium are the right size for the channel and slip in with almost no penalty. The divalents — barium, calcium, magnesium — engage the lattice less eagerly. Lithium, too small and too tightly wrapped in its hydration shell, barely engages at all. So if the framework prefers the monovalents, why are they the poisons of this cell?

Because affinity is not the same as usefulness. Each monovalent that lodges in a room carries only a single charge, and then holds that room against the divalents that should be doing the work. A sodium ion parked in a channel is a spent seat: half the charge of a Mg²⁺ or Ca²⁺, and slow to leave. Enough of them and the cathode’s usable capacity quietly bleeds away, cycle after cycle. The monovalents don’t rupture the cell — they occupy it.

This is the flip that makes the whole purification story make sense. Conventional brine chemistry casts calcium as the villain — the “hardness” you spend real money to strip out, and separating calcium from magnesium is one of the classic expensive problems in inorganic processing. In this cell, calcium is a co-worker, not a contaminant. It is a divalent; it carries two charges exactly like magnesium; and it actually inserts more readily than magnesium does (Ca sits just above Mg in the sequence). A cell running mixed Mg/Ca brine has two productive charge carriers instead of one. The proven CityU cell is precisely this dual-cation system — not pure MgCl₂ — and the mix is not a compromise. It is the design.

Which is why the purification target in chapter 13 is narrower, and cheaper, than it first appears. The job is not the classic, costly separation of calcium from magnesium. The job is to knock out the monovalents — the sodium and potassium — along with the transition-metal traces that attack the framework, while keeping both divalents. Reject the monovalents; keep the Mg and the Ca. That single reframing is what makes the frontier ambition — running the proven cell on lightly processed, unseparated brine — look less like a gamble and more like an economy.

How a cycle works

Start with the cell at 0% state of charge — the COP anode is in its reduced form (electrons sitting on the polymer backbone), the CuFe-PBA cathode is in its oxidized form (its iron sites are Fe³⁺), and the Mg²⁺ ions are mostly resting inside the cathode lattice.

Charge. Apply an external voltage of about 2.4 V across the terminals (positive on the cathode side). Electrons are pulled out of the cathode — the Fe sites oxidize further (or stay at Fe³⁺) and Mg²⁺ ions are expelled from the lattice, swimming out through the channels into the electrolyte. At the anode, electrons are pushed onto the COP polymer, reducing its functional groups; Mg²⁺ ions from the electrolyte coordinate to the polymer’s reduced sites. The cell voltage climbs as the redox states shift. When the cell reaches ~2.2 V open-circuit, it’s at 100% SOC.

Discharge. Connect the terminals through a load (a grid inverter, a building’s electrical panel). Electrons leave the anode through the wire, do useful work in the load, and return to the cathode. Inside the cell, the chemistry reverses: Mg²⁺ ions leave the polymer and travel back through the electrolyte to re-enter the PBA cathode lattice. The cell voltage drops as the redox states relax. The discharge curve is roughly flat in the middle — a desirable property for inverter compatibility — with a sharp dropoff near 0% SOC.

Why 120,000 cycles

Lithium-iron-phosphate (LFP) stationary cells — today’s grid-storage benchmark — advertise 6,000–10,000 cycles to 80% capacity retention. The CityU paper reports 120,000 cycles at 20 A/g charge rate in coin cells — running the mixed magnesium/calcium (dual-cation) electrolyte, not pure MgCl₂ — with capacity retention above 80% throughout. Twenty times longer.

Three structural reasons underlie that gap:

(1) Lattice stability. Chapter 6: the PBA cubic framework absorbs volume change without cracking. The lithium-ion equivalent (a layered metal oxide) suffers cumulative mechanical fatigue from repeated ion insertion. After ~10,000 cycles, an LFP cathode’s particles have visibly fragmented under SEM. PBA particles after 120,000 cycles look very nearly the same as they did at cycle 100.

(2) Electrolyte stability. Lithium-ion electrolyte degrades chemically during use — the LiPF₆ salt hydrolyzes to HF in any trace moisture, attacking the electrodes and SEI (solid-electrolyte interphase) layer. Aqueous Mg/Ca chloride brine at pH 7 has no analogous degradation pathway. The electrolyte at cycle 100,000 is the same electrolyte at cycle 1.

(3) No SEI to maintain. Lithium-ion cells form a protective SEI layer on the anode during the first few cycles. The SEI must remain stable for the life of the cell — any cracking or dissolution accelerates capacity loss. In aqueous Mg-ion, there is no SEI; the COP polymer interacts with the electrolyte directly. One fewer thing to maintain over a 30-year deployment.

The 120,000-cycle number translates to roughly 30 years at one full cycle per day with several deep-cycle excursions during peak-shaving events. That’s the lifetime stationary storage needs to claim if it’s going to compete with the 30–40 year service life of a natural-gas peaker plant on amortized economics. This is why the chemistry matters.

The kill modes

Nothing in chemistry is free. Aqueous Mg-ion has known failure modes that Phase 1 is designed to detect, characterize, and either accept, design around, or use to kill the program.

Copper leach. The CuFe-PBA cathode contains copper in the lattice corners. Over many cycles, a fraction of that copper can dissolve into the electrolyte, slowly depleting the cathode’s active material and reducing capacity. The Phase 1 kill threshold is >15 ppm Cu per 100 cycles; target is <5 ppm/100 cycles. Rosemount continuous monitoring catches this in real time; the proposed Cu recovery loop (electrowinning) closes the loop by re-precipitating dissolved Cu back as solid-phase electrolyte cleanup.

pH drift. The electrolyte is engineered to stay at pH 7±0.5. Drift outside 4–8 indicates a parasitic reaction is occurring — possibly water electrolysis at the electrode surfaces, possibly a contamination problem. Rosemount 372 pH probes monitor continuously; kill threshold is sustained excursion below 4 or above 8.

Voltage hysteresis growth. The voltage difference between charge and discharge at a given SOC — the hysteresis — reflects internal resistance. A growing hysteresis means the cell is becoming less efficient over time. Kill threshold: hysteresis at C/4 mid-SOC exceeding 400 mV. Target: stay under 250 mV.

EIS impedance growth. Electrochemical impedance spectroscopy measures the cell’s frequency-dependent internal resistance. A >50% growth in impedance over 1,000 cycles flags a problem the simpler metrics may miss — passivation layers forming on the electrodes, channel blockage in the PBA framework, or unwelcome interface chemistry.

Capacity fade beyond the projection. The 120,000-cycle stability is reported at the paper’s specific test conditions (20 A/g, partial depth of discharge). Phase 1 runs at multiple C-rates and 80% DoD — the more aggressive conditions stationary storage actually faces. If capacity drops below 90% retention by cycle 1,000 at C/4 with 80% DoD, the commercial extrapolation breaks. This is the load-bearing kill gate.

Cell → skid → grid

A single coin cell is a 0.5 cm², ~1 mAh disc — the scale at which the CityU paper made its measurements. A grid-storage installation needs megawatt-hours. The path from cell to grid runs through three engineering layers, each of which Emerson’s existing product portfolio can address.

Cell to module. Cells are stacked or packed into modules at the kilowatt-hour scale. Each module needs uniform electrolyte fill, balanced cell voltages, thermal management (aqueous chemistry generates much less heat than lithium-ion, but heat removal is still a design consideration), and a cell management system that monitors each cell’s SOC and balances them during charging. This is the layer where Zitara’s battery management software — an Emerson Ventures portfolio company — runs.

Module to skid. Modules are arrayed into skids — transportable, weatherproof enclosures typically at the 50–500 kWh scale, sized to fit a standard shipping container. The skid handles AC/DC power conversion, fire detection (although the aqueous chemistry removes most of the fire-detection requirement), and the physical connection to the building or substation. Ovation Green BESS (Emerson’s Power & Water Solutions division, owner of the prior Westinghouse battery line) is the skid-level integration platform.

Skid to grid. Multiple skids interconnect at the substation level, with grid-tie inverters, transformer integration, and SCADA control. DeltaV handles the process-control plumbing; Plantweb Insight Aqueous (the new SaaS layer the proposal builds) handles the predictive analytics: aggregating telemetry from every deployed installation, learning the degradation patterns across the fleet, predicting maintenance needs before they become outages. This is the moat — the chemistry is published, but the data layer accrues to whoever ships the first 1,000 installations.

What Phase 1 actually tests

Phase 1 is the 12-month, $1.4–1.8M experimental program that answers the open chemistry questions before any commercial capital commits. It runs as two parallel tracks.

Track A: cell-chemistry validation. 520 pouch cells (≥1 Ah each) built in the Phase 1 lab using CuFe-PBA cathode, COP anode, and mixed Mg/Ca chloride (dual-cation) electrolyte. Cells are cycled on the NI HPS-17000 battery cycler at four C-rates (C/10, C/4, C/1, and 2C) for 12 months. The deliverable is a measured capacity-fade curve at each rate, extrapolated to commercial cycle life, against the kill thresholds in chapter 11 and the lab manual.

Track B: purification process development. Run in parallel: develop a repeatable, instrumented DeltaV-controlled process to purify industrial-grade MgCl₂ from Intrepid Potash (Carlsbad, NM) to battery-grade (≥99.9% MgCl₂, <5 ppm transition-metal impurities). Validate via inductively-coupled plasma (ICP) analysis against Nedmag pre-purified bischofite as the reference standard. The deliverable is a Phase-2-ready purification skid spec — an Emerson product line in itself, beyond its role in this program.

The two tracks intersect at month 9: Track B’s output (purified MgCl₂) becomes Track A’s electrolyte for the final 25% of the cell campaign, confirming that domestically-purified electrolyte performs identically to vendor-supplied battery-grade. If both tracks land their kill-gate criteria, Phase 1.5 builds the first integration prototype skid. If either track misses, the program kills at <$1.8M — failure-cheap by design.

What this textbook covered is the scientific case. The proposal covers the business case. The lab manual covers the experimental case. Each document anchors to the others. The chemistry is published. The instruments are in the pantry. Phase 1 is what closes the loop.