Electrodialysis and EDR: How Ion Transport Demineralises Water

How does electrodialysis actually move salt out of water?

Electrodialysis is an electrically driven, ion-selective membrane process. A stack is built from many thin flow channels separated by an alternating series of two membrane types: cation-exchange membranes (CEM), which carry fixed negative charges and therefore admit only positively charged ions (Na⁺, Ca²⁺, Mg²⁺), and anion-exchange membranes (AEM), which carry fixed positive charges and pass only anions (Cl^-, SO₄^2-, NO₃^-). A DC voltage is imposed across the stack by an anode at one end and a cathode at the other.

Under that field, cations migrate toward the cathode and anions toward the anode. Consider one channel: a cation leaving it passes freely through the adjacent CEM (which is on the cathode side) but is then blocked by the next membrane it meets, an AEM. Simultaneously, anions in that same channel move the other way, pass an AEM, and are stopped by a CEM. The net effect is that both ionic species are swept out of every other channel and trapped in the channels in between. The depleted channels become the diluate (product water); the enriched channels become the concentrate (brine). A repeating CEM–diluate–AEM–concentrate unit is called a cell pair, and a commercial stack contains a few hundred of them clamped between a single electrode pair.

The distinction from pressure-driven membranes is fundamental. In a reverse osmosis or nanofiltration membrane the water is forced through the membrane and the salt is rejected; in electrodialysis the ions are forced through the membrane and the water stays put. That single inversion — move the minority component, not the majority component — is what makes the process energetics so different, and it is the reason ED occupies a distinct niche in the treatment map rather than simply competing head-to-head with RO everywhere.

Ion-exchange membranes achieve their selectivity through the Donnan exclusion principle: the high concentration of fixed charges inside the membrane electrostatically repels co-ions (ions of the same sign as the fixed charge) while readily hosting counter-ions. Practical membranes reach counter-ion transport numbers above 0.95, so the current is carried almost entirely by the intended ion. The residual leakage of co-ions and of water (electro-osmotic drag, roughly 4–8 water molecules dragged per ion) sets the ultimate limit on how concentrated a brine the stack can make and how pure the diluate can become.

How much current do you need? Faraday's law and salt removal

The single most important design relationship in electrodialysis is that the moles of salt removed are directly proportional to the electric charge passed. This is Faraday's law of electrolysis applied to ion transport, and it lets you translate a demineralisation target straight into a current requirement before you have chosen a single component.

The factor N in the denominator captures the elegance of the stack geometry: because the same current passes through all N cell pairs in series, each ampere removes salt from N channels at once. Doubling the number of cell pairs doubles the salt you can strip at a given current — you pay for it in stack voltage, not in current, as the next section shows.

The current efficiency η_cur is below unity because a small fraction of the charge is carried by co-ion leakage, by water splitting at high current, and by parasitic currents through the manifolds. In a well-designed brackish stack it sits comfortably above 0.9. Notice what the equation does not contain: the feed water volume in any absolute sense beyond the salt it carries. This is the mathematical heart of why ED energy tracks salt removed rather than water produced.

Worked example — current from a demineralisation target. A brackish groundwater at 2,500 mg/L as NaCl must be reduced to 500 mg/L at a product flow of 50 m³/h, using a stack of 300 cell pairs at 90% current efficiency.

• Molar mass of NaCl = 58.44 g/mol, so ΔC = (2,500 − 500) mg/L = 2,000 mg/L = 2.0 g/L ÷ 58.44 = 0.0342 mol/L = 34.2 mol/m³.
• Flow Q = 50 m³/h ÷ 3,600 = 0.01389 m³/s.
• Salt-removal rate = Q · ΔC = 0.01389 × 34.2 = 0.475 mol/s of NaCl.
• Charge demand (z = 1) = F · Q · ΔC = 96,485 × 0.475 = 45,800 C/s = 45,800 A across the full stack current path.
• Divide by N = 300 cell pairs and by η_cur = 0.90: I = 45,800 / (300 × 0.90) = 170 A.

So a 170 A rectifier moves this duty. The result is exact within the efficiency term; there is no empirical fudge factor, which is why Faraday's law is the anchor of every ED sizing spreadsheet.

Why can't you just push more current? Limiting current density

If salt removal is proportional to current, the naive instinct is to crank the current up and shrink the stack. Physics forbids it. As ions are pulled toward and through a membrane, they must first cross a thin, nearly stagnant diffusion boundary layer on the diluate side. Because the membrane transports the counter-ion faster than diffusion can resupply it, the ion concentration at the membrane face falls below the bulk value — a phenomenon called concentration polarisation.

Push harder and the surface concentration approaches zero. At that point the ions can no longer carry the demanded current, and the stack reaches the limiting current density (LCD). Beyond it, the only charge carriers left are H⁺ and OH^- from water splitting at the depleted membrane surface. That wastes energy, shifts local pH (precipitating carbonates and hydroxides on the membrane), and collapses current efficiency. Operating near but safely below the LCD is the central hydraulic constraint of ED design.

Two design levers follow directly. First, the LCD is proportional to bulk concentration, so it is lowest where the water is purest — at the diluate outlet. That is exactly why ED struggles to reach very high purity in a single pass and why deep polishing is left to a downstream ion-exchange or RO polishing step. Second, the LCD rises with turbulence because turbulence thins the boundary layer δ. Stacks therefore use mesh spacers that promote cross-flow mixing, and are run at channel velocities of roughly 5–12 cm/s to keep δ thin without excessive pressure drop.

Worked example — checking against the limit. Take a spacer with an established ratio i_lim/C_b = 0.6 (A/m² per mol/m³) at the design velocity. At the diluate outlet C_b has fallen to 500 mg/L = 8.6 mol/m³, so i_lim = 0.6 × 8.6 = 5.1 A/m². Designers typically operate at 70–80% of this, say 4 A/m². With the 170 A of the previous example, the required active membrane area per cell pair is 170 / 4 = 42.5 m² — which, if a single membrane sheet offers about 1 m² of active area, tells you the diluate must be staged through several passes in series to accumulate enough area at a falling current density. This staging is the real reason commercial units are built as multiple hydraulic stages rather than one long channel.

How much energy does electrodialysis use, and why does TDS decide it?

The stack energy is simply current times the voltage needed to drive it, integrated over time. Because current is fixed by the salt-removal duty (Faraday) and voltage rises with the number of cell pairs and the stack resistance, the energy per cubic metre is dominated by how much salt you remove, not how much water you make.

This proportionality to ΔC is the decisive contrast with reverse osmosis. In RO the energy floor is the osmotic pressure, which climbs steeply with feed salinity (roughly 0.78 bar per 1,000 mg/L of NaCl); desalting seawater costs 3–5 kWh/m³ largely because you must overcome ≈25–30 bar of osmotic pressure. Electrodialysis pays no osmotic penalty at all — it never pushes water through a membrane — so at low-to-moderate TDS its energy can undercut RO. But because ED energy grows linearly with the salt removed, it becomes uncompetitive at high salinity, exactly where RO's logarithmic osmotic term is comparatively kinder. The crossover for potable-quality product typically sits around 2,000–5,000 mg/L TDS.

Worked example — specific energy. Continue the 50 m³/h duty at I = 170 A. Suppose each cell pair needs V_cell = 1.0 V (typical for brackish water), so a 300-pair stack runs at V_stack = 300 V.

• Stack power = I × V_stack = 170 × 300 = 51,000 W = 51 kW.
• Per cubic metre: 51 kW ÷ 50 m³/h = 1.02 kWh/m³ for the stack.
• Add pumping (recirculation and feed, roughly 0.3–0.6 kWh/m³) and rectifier losses (≈5–8%), giving a plant figure of about 1.4–1.7 kWh/m³.

For this 2,500→500 mg/L duty that is broadly comparable to or better than brackish RO, and it improves further at still lower feed TDS. Run the same arithmetic on a 35,000 mg/L seawater and the ΔC — and hence current, voltage and energy — scales up roughly fourteen-fold, which is precisely why nobody desalinates seawater by electrodialysis when reverse osmosis is available.

What is EDR, and why reverse the polarity?

Electrodialysis reversal (EDR) is standard electrodialysis with one operational addition: the DC polarity is periodically reversed, typically every 15–30 minutes, and the diluate and concentrate flow paths are simultaneously switched by automatic valves. When polarity flips, every channel that was demineralising becomes a concentrating channel and vice versa, and the direction of ion migration reverses throughout the stack.

The payoff is self-cleaning scale and fouling control. In fixed-polarity ED, the concentrate channels and the membrane surfaces facing them steadily accumulate scale (CaCO₃, CaSO₄) and a fouling film of negatively charged colloids and organics attracted to the anion-membrane faces. Reversing polarity dissolves freshly deposited scale back into the now-diluting stream before it can harden, and it electrostatically repels the fouling layer that was migrating onto the membrane. The result is that EDR can run on waters with higher scaling and fouling potential, at higher recovery, and with far less chemical dosing (antiscalant, acid) than fixed-polarity ED or RO would tolerate. This robustness — tolerating chlorine, silica, and variable feed — is the main reason EDR, not plain ED, dominates the brackish-water market.

The cost of reversal is a short transition period after each flip during which the product briefly goes off-specification (the freshly reversed diluate channel still holds the previous cycle's concentrate) and must be diverted to waste or recycled. This trims net recovery by a few percent and demands robust automatic valving, but the operational resilience is almost always worth it. EDR plants routinely achieve 85–94% water recovery on brackish groundwater, higher than brackish RO can usually reach on the same scaling-prone feed without heavy chemical conditioning.

Where does electrodialysis win? Selective removal and reuse

Beyond bulk brackish desalination, electrodialysis has a growing role wherever the goal is to remove or recover specific ions rather than to strip everything. Because transport is charge-driven and membranes can be engineered for monovalent selectivity, ED variants can preferentially move one class of ion over another — something pressure-driven membranes, which reject by size and charge together, cannot do cleanly.

• Nitrate removal from drinking water. Selective anion-exchange membranes let a stack strip NO₃^- preferentially, meeting the potable limit without the full demineralisation (and remineralisation) that RO would impose.
• Monovalent-selective separation. Membranes that favour Na⁺ and Cl^- over Ca²⁺, Mg²⁺ and SO₄^2- allow desalting while retaining hardness or, conversely, softening a stream — useful upstream of other membrane systems or for partial desalination of process brines.
• Industrial water reuse and ZLD. EDR concentrates a waste stream to a smaller, more manageable brine at high recovery, cutting the volume that must go to evaporation in a zero-liquid-discharge scheme and reducing overall reuse cost. See our overview of industrial water reuse for where it fits.
• Food, dairy and bioprocessing. ED demineralises whey, deacidifies juices, and desalts amino-acid and organic-acid streams without the water loss and osmotic limits of RO.

The choice against neighbouring technologies is best made on the full water map. For hardness alone, ion-exchange water softening is simpler and cheaper; for high-salinity bulk desalting, reverse osmosis system design governs; and for genuine seawater duty, seawater desalination is an RO or thermal problem, not an ED one. Electrodialysis earns its place in the band between — low-to-moderate TDS, scaling-prone feeds, high-recovery targets, and selective-ion duties — where its Faradaic, osmosis-free energetics are a genuine structural advantage rather than a curiosity.

In short, the discipline of ED design is a chain of three physical anchors: Faraday's law fixes the current from the salt duty, the limiting current density fixes the safe area and staging, and the stack voltage fixes the energy. Get those three right, choose EDR when the feed is dirty or scaling, and the process delivers demineralised water at an energy cost that — for the right salinity band — nothing else matches.