Forward Osmosis: How It Works, Where It Helps, and Where It Falls Short

Forward osmosis (FO) is a way of separating water from dissolved substances using a semi-permeable membrane and the natural pull of osmosis. In reverse osmosis (RO), the familiar technology inside most modern seawater desalination plants, pumps push salty water against a membrane hard enough to overcome its osmotic pressure. FO does something different. It puts a very concentrated solution, called the draw solution, on the other side of the membrane and lets osmosis do the work. Water moves on its own from the less concentrated feed (seawater, wastewater, fruit juice, muddy river water) into the more concentrated draw.

That simple idea has drawn decades of research, a handful of commercial products, several failed or shelved ventures, and a long-running argument about energy. This article explains the physics, the history, the membranes and draw solutions that make FO work, the applications that have actually been tried, and the limitations that have kept it from replacing reverse osmosis. Where the evidence is a company's claim and not an independent measurement, that is said plainly.

The basic idea in one paragraph

Put a membrane between two liquids. The membrane lets water through but blocks most dissolved salts and larger molecules. On one side is the feed, the water you want to clean or concentrate. On the other side is the draw solution, which is much saltier, sweeter, or otherwise more concentrated. Because water moves from regions of higher water chemical potential to lower water chemical potential, it crosses the membrane into the draw. The feed becomes more concentrated and the draw becomes more dilute. No high-pressure pump is needed for this step. The catch is that the water you have collected is now mixed into the draw solution. Unless you can drink or use the diluted draw directly, you need a second step to pull clean water back out of it, and that second step is where most of the cost and energy of FO desalination ends up.

Osmosis before "forward osmosis"

Osmosis was studied long before anyone tried to engineer it. The French cleric and physicist Jean-Antoine Nollet documented it in 1748. He sealed alcohol in a vessel with a wet animal bladder, submerged it in water, and found hours later that water had moved inward so strongly that the bladder bulged. When it was pricked, liquid shot out more than a foot. René Dutrochet (1776 to 1847) coined the terms "endosmose" and "exosmose," from which the modern word descends, and in 1867 Moritz Traube made highly selective precipitation membranes that made osmotic flow easier to measure.

People had used the principle much earlier without the theory. A 2006 review in the Journal of Membrane Science by Tzahi Cath, Amy Childress and Menachem Elimelech, which remains one of the most cited overviews of the field, opens by pointing out that salting food is an osmotic technology: in a salty environment, bacteria and fungi lose water and die or become inactive.

The same review notes that early membrane experimenters tried nearly every material they could find, including the bladders of pigs, cattle and fish, collodion (nitrocellulose), rubber, porcelain and goldbeater's skin. The modern era began around 1960, when the Loeb–Sourirajan process made it possible to cast defect-free, high-flux, asymmetric membranes. Those membranes were designed for reverse osmosis, and that origin would shape, and hamper, forward osmosis for decades.

The physics: osmotic pressure and flux

Osmotic pressure

Osmotic pressure (written π) is the pressure that, if applied to the more concentrated solution, would just stop water from flowing into it across an ideal membrane. For dilute solutions it can be estimated with the van 't Hoff relation:

π ≈ i · c · R · T

where i is the number of particles each dissolved unit splits into (about 2 for sodium chloride, which separates into Na⁺ and Cl⁻), c is the molar concentration, R is the gas constant (0.08314 L·bar per mol·K), and T is the absolute temperature.

Worked example. Take seawater as roughly 0.6 mol/L sodium chloride at 25 °C (298 K). The ideal estimate is 2 × 0.6 × 0.08314 × 298 ≈ 29.7 bar. Real seawater behaves less than ideally, and the Cath, Childress and Elimelech review gives its osmotic pressure as approximately 2.7 MPa, or about 27 bar. That figure is the number a reverse osmosis plant must beat just to get water to start moving, and it rises as the brine becomes more concentrated during processing. A forward osmosis draw solution must exceed it to pull water out of seawater at all.

One equation, three processes

The same equation describes water flux through the membrane in forward osmosis, reverse osmosis, and a third process called pressure-retarded osmosis (PRO):

J_w = A (σ Δπ − ΔP)

Here J_w is the water flux (volume per membrane area per time, usually reported in litres per square metre per hour, L/m²h), A is the membrane's water permeability, σ is the reflection coefficient (a measure of how well the membrane rejects solute), Δπ is the osmotic pressure difference, and ΔP is the applied hydraulic pressure difference.

The flux directions and driving forces of the three processes were characterized in the early 1980s by Lee, Baker and Lonsdale. If you plot flux against applied pressure, FO is a single point (zero applied pressure), PRO is a zone of positive pressure below the "flux reversal point," and RO is everything beyond it.

Solute moves too

Membranes are not perfect. The solute flux J_s across the membrane is commonly modelled with a Fick's-law form, J_s = B Δc, where B is the solute permeability of the membrane's selective layer and Δc is the concentration difference. In reverse osmosis, salt leakage goes one way: from the pressurized feed toward the product. In forward osmosis it can go both ways at once. Feed contaminants can leak into the draw, and draw solutes can leak backward into the feed. The second effect is called reverse solute flux, and it matters for three reasons:

1. The draw solute is lost and has to be replaced, which costs money.
2. The leaked draw solute ends up in the concentrated feed, which may be discharged or used for something else.
3. Draw solute building up on the feed side reduces the effective driving force.

A 2010 study in Environmental Science & Technology by William Phillip, Jui Shan Yong and Menachem Elimelech modelled reverse draw solute permeation in asymmetric FO membranes and tested it with a sodium chloride draw and a cellulose acetate FO membrane. They treated the ratio of forward water flux to reverse solute flux, the "reverse flux selectivity," as a key design parameter and concluded that its value is determined solely by the selectivity of the membrane's active layer, not by the support layer underneath. In practice this ratio, or its inverse, the "specific reverse solute flux," became a standard way to compare FO membranes and draw solutes.

Why real flux is much lower than theory: concentration polarization

If you put the bulk osmotic pressures of a typical feed and draw into the flux equation, you predict a large flux. Experiments consistently delivered far less. The main reason is concentration polarization: the concentrations right at the membrane's selective layer are not the same as the concentrations in the bulk liquids.

External concentration polarization

On the feed side, water leaves through the membrane and dissolved substances are left behind, so they pile up against the surface. This is concentrative external concentration polarization, and it also happens in reverse osmosis. On the draw side, incoming water dilutes the draw right at the membrane surface. That is dilutive external concentration polarization. Both shrink the effective driving force. Faster flow and more turbulence along the membrane can reduce them. The 2006 review points out that external polarization plays a relatively minor role in FO and is not the main cause of disappointing flux.

Internal concentration polarization

The bigger problem sits inside the membrane. Practical membranes are asymmetric. They have a very thin, dense selective layer on top of a thicker porous support that gives mechanical strength. In reverse osmosis that support just has to let product water drain away. In forward osmosis, solute has to diffuse into and out of the support's pores, and the support acts as an unstirred zone that cross-flow cannot reach.

• Dilutive internal concentration polarization happens in the usual desalination and water-treatment orientation, with the selective layer facing the feed and the porous support facing the draw. Water crossing the selective layer dilutes the draw solution trapped in the support pores, so the draw concentration right behind the selective layer is much lower than in the bulk.
• Concentrative internal concentration polarization happens in the opposite orientation, used in PRO, with the porous support facing the feed. Feed solutes are carried into the support and accumulate against the back of the selective layer.

The resistance of the support to solute diffusion is often summarized by a parameter K = tτ / (εD_s), where t is the support thickness, τ its tortuosity, ε its porosity and D_s the solute's diffusion coefficient. A thick, winding, low-porosity support gives a large K and severe internal polarization. Sidney Loeb and colleagues derived a simplified flux expression, J_w = (1/K) ln(π_high / π_low), which the 2006 review notes is valid only at very low fluxes. More general forms include the solute permeability B and the water permeability A.

How big is the penalty?

Very large. In bench tests of the ammonia–carbon dioxide draw solution described below, the cellulose triacetate FO membrane reached fluxes as high as 25 L/m²h with a calculated driving force of more than 200 bar. That sounds impressive, but an ideal membrane facing that driving force would have passed far more water. Analysis of the results found that the "performance ratio," experimental flux divided by theoretical flux, was at most 20 percent and on average between 5 and 10 percent. The authors attributed the gap to internal concentration polarization. The same review describes an experiment by Mehta and Loeb in which simply swapping which liquid faced which side of the membrane produced a sharp decline in flux.

This is why the history of FO membranes is largely a history of trying to make thinner, more porous, less tortuous support layers.

Membranes for forward osmosis

Borrowed RO membranes

In the 1970s, nearly all osmosis research, mostly aimed at PRO, used reverse osmosis membranes, and researchers always measured much lower flux than expected. Mehta and Loeb tested DuPont B-9 flat sheets and B-10 hollow fibres made of aromatic polyamide. The B-10 flux was about 8 L/m²h at a driving force of 80 bar. Early wastewater work from the University of Rhode Island, reported in 1974 and 1977, used commercial cellulose RO membranes to concentrate dilute industrial wastewater containing copper or chromium. The measured fluxes ran from zero to about 4.5 L/m²h, far below the 10 to 17 L/m²h calculated from manufacturer data. Salt from the simulated seawater draw also leaked through faster than expected: between 1 gram of sodium chloride for every 11.5 grams of water and 1 gram for every 688 grams of water crossed in the opposite direction. Planned pilot testing was cancelled, and the authors concluded that membranes would have to be specifically "tailored" for FO.

The first purpose-built FO membrane: cellulose triacetate

That tailoring arrived in the 1990s from a small Oregon company. Osmotek Inc. of Albany, Oregon, later renamed Hydration Technologies Inc. (HTI), developed a membrane specifically for forward osmosis. It is understood to be made of cellulose triacetate (CTA). Instead of a thick porous support, it has a polyester mesh embedded in the polymer for mechanical strength, and the whole membrane is less than 50 micrometres thick. Typical RO membranes, by comparison, combine an active layer under 1 micrometre thick with a thick porous support.

The difference showed up clearly in a test reported in the 2006 review, using a 100 g/L sodium chloride draw and deionized water feed:

The purpose-built membrane delivered roughly nine to thirty times the flux of the RO membranes tested.

Thin-film composite FO membranes

Researchers also turned to the polyamide thin-film composite (TFC) chemistry used in most reverse osmosis membranes and redesigned its support for forward osmosis. A widely cited 2010 paper by Ngai Yin Yip, Alberto Tiraferri, William Phillip, Jessica Schiffman and Menachem Elimelech in Environmental Science & Technology described a TFC forward osmosis membrane. It had a polyamide selective layer made by interfacial polymerization, on top of a polysulfone support made by phase separation onto a thin (40 micrometre) polyester nonwoven fabric. The authors reported that its high water flux was directly related to the thickness, porosity, tortuosity and pore structure of the polysulfone support, which confirms that the support, not the selective layer, was the bottleneck. The membrane's performance did not degrade after prolonged exposure to an ammonium bicarbonate draw solution.

What an ideal FO membrane needs

The 2006 review summarized the wish list:

• a dense selective layer for high solute rejection;
• a thin support with minimal resistance to diffusion, for low internal concentration polarization;
• hydrophilic surfaces, for higher flux and less fouling;
• enough mechanical strength to handle pressure if the membrane is to be used for PRO.

Membrane modules

A membrane is not useful until it is packed into a module. FO creates a design problem that RO does not, because liquid must be pumped along both sides of the membrane.

• Plate-and-frame modules are the simplest. They range from lab cells holding a single small coupon to full-scale systems holding more than 1,700 membranes. Their drawbacks are low packing density, which means a large footprint and higher capital and labour costs, plus sealing problems and limited operating pressures.
• Spiral-wound modules, the workhorse of the RO industry, cannot be used for FO in their standard form, because there is no way to force a draw solution to flow inside the envelope formed by two glued membrane sheets. A modified spiral-wound design by Mehta solved this by blocking the central tube halfway and adding a glue line so that feed flows into the envelope while draw flows between the rolled envelopes.
• Tubular and hollow-fibre modules are self-supporting, easy to pack densely, and let liquid flow freely on both sides. Hollow fibres (internal diameter under 1 mm) allow only laminar flow, which limits mixing. Larger tubes (internal diameter of 2 mm or more) can be run turbulently, which reduces polarization and fouling.
• Hydration bags are a double bag. The inner bag is made of FO membrane and holds the draw solution. The outer sealed bag holds the inner bag and the dirty water.

The 2006 review also distinguishes continuous operation, in which the draw is constantly reconcentrated and reused, from batch operation, in which the draw is diluted once and the device is often thrown away. Hydration bags and implanted osmotic drug pumps are both batch devices.

Draw solutions: the heart of the problem

The draw solution is the engine of forward osmosis. The literature calls it many things: osmotic agent, osmotic media, driving solution, osmotic engine or just brine. Its first job is to have a higher osmotic pressure than the feed. In most applications it also has to be separated from the product water afterward, cheaply, completely and without fouling the equipment.

A short history of draw solutes

The 2006 review traces a long line of proposals, many of them patents that never matured:

Natural brines have also been used or considered as draw solutions, including seawater, Dead Sea water and Salt Lake water, especially for PRO.

Sodium chloride: the default

Very often the draw is plain sodium chloride. It is cheap, highly soluble, produces a high osmotic pressure, and can be reconcentrated with reverse osmosis without much risk of scaling. Its weakness is that sodium chloride leaks back through the membrane relatively easily, and using RO to regenerate the draw means the overall system still contains a high-pressure RO step.

The ammonia–carbon dioxide "thermolytic" draw

The most influential draw solution research came out of Yale University. In a 2005 paper in Desalination, Jeffrey McCutcheon, Robert McGinnis and Menachem Elimelech described a forward osmosis desalination process that used an ammonium bicarbonate draw solution. Mixing ammonia and carbon dioxide in specific ratios produces a highly concentrated solution of ammonium salts (ammonium bicarbonate, ammonium carbonate and ammonium carbamate). According to the 2006 review, these draw solutions reached osmotic pressures above 250 atmospheres. Analysis showed driving forces as high as 238 bar for a feed of 0.05 M sodium chloride and 127 bar for a 2 M sodium chloride feed. A 2 M sodium chloride solution is roughly what remains as brine after seawater desalination reaches about 70 percent water recovery.

The trick is the recovery step. On moderate heating, near 60 °C, the ammonium salts decompose back into ammonia and carbon dioxide gas, which can be stripped out by distillation or membrane distillation and recycled, leaving fresh water behind. Because the separation runs on low-temperature heat instead of electricity-driven pressure, it was promoted as a way to use waste heat from power stations or industry.

The approach was commercialized by Oasys Water, founded by Robert McGinnis. According to the Wikipedia account of the field, the process was first used to treat oil and gas "produced water" in the Permian Basin of Texas and was later used at power and manufacturing plants in China. Brine concentration toward zero liquid discharge, where the goal is to shrink a wastewater stream as much as possible, played to the process's ability to handle very salty feeds that reverse osmosis cannot.

Draw solutions that never need separating

Some applications avoid the regeneration problem entirely by choosing a draw that is itself the useful product:

• Sugars for drinking. Hydration bags use an edible draw such as a sugar or flavoured beverage powder, so the diluted draw is the drink.
• Fertilizer for irrigation. In 2011, Sherub Phuntsho, Ho Kyong Shon and co-authors proposed fertilizer-drawn forward osmosis in the Journal of Membrane Science. Concentrated fertilizer pulls water from saline water, and the diluted fertilizer goes straight onto crops. They screened about nine common fertilizers and reported that 1 kg of fertilizer could extract between 11 and 29 litres of water from seawater. Later work by the same group found membrane scaling and flux decline when brackish groundwater was used as the feed.
• Seawater as the draw. In "osmotic dilution," seawater pulls water out of wastewater. The seawater gets slightly diluted before going into a conventional RO plant, which then needs less pressure, while the wastewater gets concentrated.

What a good draw solution needs

The 2006 review listed the most desirable properties: high osmotic pressure; low energy needed for regeneration or reconcentration; easy separation from the product water; low or no toxicity; and no chemical reaction with the polymer membrane. In practice, these goals conflict. Large molecules and nanoparticles leak back through the membrane less, but they tend to produce lower osmotic pressure and more viscous solutions, which worsens internal polarization. Small ions give high osmotic pressure but leak more. No draw solute has yet been found that satisfies every requirement at low cost.

The energy question

The popular claim

Forward osmosis is often described as "low energy" because the water-transfer step needs no high-pressure pump. In March 2010, National Geographic named FO among three technologies that could reduce the energy cost of desalination. Commercial developers made similar claims. Modern Water, a British company, said its process could save up to 30 percent of the specific energy consumption of a conventional RO process. That figure is a company claim, not an independent measurement.

The thermodynamic objection

The problem is that the first step of FO does not produce fresh water. It produces a diluted draw solution that is, by necessity, more concentrated than the feed it pulled water from. Otherwise water would not have moved. To get fresh water, you then have to separate water from that solution, and the minimum energy needed to separate water from a solution rises with its concentration.

In 2014, Ronan McGovern and John Lienhard of the Massachusetts Institute of Technology made this argument rigorously in the Journal of Membrane Science, in a paper titled "On the potential of forward osmosis to energetically outperform reverse osmosis desalination." According to MIT's summary of the study, McGovern found that even if the draw regeneration step were as efficient as reverse osmosis, the total energy use of FO desalination of seawater would always be higher than that of RO. The diluted draw solution is more concentrated than seawater, so regenerating it always takes more energy than desalinating the seawater directly. The authors concluded that FO effort would be better directed at other applications. MIT's coverage named hydration drinks as an example where FO has a real advantage, because only the first step is needed.

Where FO can still save energy

The objection is about seawater desalination, where the alternative, RO, is mature and efficient. It does not rule out energy benefits in other situations:

• When the second step is skipped. Hydration bags, fertilizer drawing, osmotic dilution before RO, and concentrating a liquid food all use only the first step.
• When the energy is low-grade heat. Thermolytic draws such as ammonia–carbon dioxide trade electricity for heat. If that heat would otherwise be wasted, the cost comparison changes even though the thermodynamic minimum does not.
• When the feed would wreck RO. Very salty or heavily fouling feeds such as landfill leachate, oilfield produced water and concentrated brines can exceed the pressure limits of RO membranes or foul them quickly. Here FO is competing with thermal evaporators, not with RO, and evaporators are energy-hungry.

The NASA wastewater project described below gives a sense of real numbers in a demanding application: specific power consumption almost always below 30 kWh per cubic metre of purified water. That is far higher than modern seawater RO, but it was for a lightweight, autonomous system treating hygiene wastewater, urine and humidity condensate, not seawater.

Applications that have actually been built or tested

The 2006 review summarized the status at the time: FO had been used at bench scale to treat industrial wastewater and liquid foods, at pilot and full scale to concentrate landfill leachate, at demonstration scale for potable reuse in life-support systems, and commercially in hydration bags and pharmaceutical osmotic pumps. It added that in most wastewater applications FO is not the final step but a high-level pretreatment before a final desalination process.

Landfill leachate in Oregon

The clearest full-scale success story in the early literature is at the Coffin Butte Landfill near Corvallis, Oregon. The site gets more than 1,400 mm of rain a year and generates roughly 20,000 to 40,000 cubic metres of leachate annually. Landfill leachate is notoriously difficult. It is a variable mixture of organic compounds, dissolved heavy metals, nitrogen compounds and dissolved salts. To meet discharge permits, the leachate had to be treated to below 100 mg/L total dissolved solids before land application.

In 1998 Osmotek built a pilot FO system using its CTA membrane and a sodium chloride draw. Over three months it achieved water recoveries of 94 to 96 percent with high contaminant rejection. Flux did not noticeably decline while treating raw leachate, but it fell by 30 to 50 percent when processing concentrated leachate, and cleaning restored it almost completely.

The full-scale plant that followed extracted water from pretreated leachate in six stages of FO cells. A three-pass RO system then produced purified water for land application and reconcentrated the draw to about 75 g/L sodium chloride. The concentrated leachate was solidified for disposal. Between June 1998 and March 1999 the plant treated more than 18,500 cubic metres of leachate, with an average water recovery of 91.9 percent and an average RO permeate conductivity of 35 µS/cm. Effluent concentrations of regulated contaminants were well below permit limits. For example, sodium dropped from about 1,620 mg/L in raw leachate to about 4 mg/L, and ammonia nitrogen from 1,110 mg/L to 1.6 mg/L.

Note what FO did here. It did not lower the energy bill compared with reverse osmosis. Its value was protecting the RO stage from a fouling, unpredictable feed. The Wikipedia summary of this application states the benefit directly: it is fouling resistance, not energy savings.

Wastewater recycling for spaceflight

Long missions to the Moon or Mars need to recycle water from hygiene wastewater, urine and humidity condensate with a system that is reliable, light, low-maintenance and low-power. NASA evaluated a system called direct osmotic concentration (DOC), which used FO as pretreatment ahead of a reverse osmosis stage. One variant also used osmotic distillation to catch small molecules such as urea that slip through semi-permeable membranes. NASA and Osmotek designed and tested the unit from 1994 to 1999. From 2002 to 2004 it was studied at the University of Nevada, Reno, and a new prototype with a separate urine-treatment process was being built in a third phase from 2004 to 2007.

Digested sludge liquids

At the Truckee Meadows Water Reclamation Facility in Reno, Nevada, researchers tested FO on centrate, the nutrient-rich liquid left after dewatering digested sludge. Using a CTA membrane and a sodium chloride draw that started at 70 g/L and diluted to about 25 g/L over roughly 12 hours, they observed flux decline from organic and suspended-solids fouling, which a short cleaning with sodium hydroxide largely or completely reversed. Phosphorus rejection exceeded 99 percent, ammonia rejection was about 87 percent, and total Kjeldahl nitrogen rejection about 92 percent. Colour and odour compounds were almost completely rejected. The lower ammonia rejection illustrates a general weakness of FO membranes: small, uncharged or weakly charged molecules pass more easily than salts.

Hydration bags

Hydration bags are among the few unambiguously commercial uses of FO. They were developed for military, recreational and emergency relief settings where safe drinking water is unavailable. A sealed pouch of FO membrane holds an edible draw such as a sugar or drink powder. Put it in water, even muddy water, and water slowly diffuses in, producing a sweet drink. Because the membrane rejects microorganisms, most macromolecules and most ions, the drink is safe in most situations. A small personal bag takes 3 to 4 hours to fully hydrate a 12-ounce drink. The bags need no power and foul very little.

The 2006 review also recorded a fair criticism: there is debate among experts over whether hydration bags provide water treatment at all, because the product is not water but a sugary drink suited only to particular uses. Hydration bags trade speed and versatility for simplicity and independence from power.

Seawater desalination

Several patents on FO desalination were awarded, but the 2006 review found that most had not matured or proven feasible and that very few peer-reviewed desalination studies existed. The ammonia–carbon dioxide process showed bench-scale salt rejections above 95 percent. The review summed up the two big limitations FO desalination had exposed: the lack of high-performance membranes and the need for an easily separable draw solution.

Commercial attempts followed. Modern Water, which called its process "manipulated osmosis," trialled its technology at a plant in Gibraltar and then built a plant at Al Khaluf in Oman for the Omani water ministry, announced in 2008. Trade coverage describes the Gibraltar trial plant at 18 m³ per day and Al Khaluf at about 100 m³ per day. A later Modern Water plant at Al Najdah, Oman, at 200 m³ per day, was described in coverage as the world's first commercial-scale FO facility, although the company had also used that phrase for Al Khaluf. Reports on when Gibraltar began supplying water differ: one says September 2008, another says public supply began in May 2009. These plants were small by desalination standards, where large RO plants produce hundreds of thousands of cubic metres per day, and trade press later reported that Modern Water's FO membrane revenues had dried up.

Thermal desalination pretreatment

Older thermal desalination plants, multi-stage flash (MSF) and multi-effect distillation (MED), are limited by scaling at high temperatures. One proposal uses FO to dilute the recirculating brine, drawing water from the plant's cooling water stream, to reduce scaling. A 2015 study by Darwish and colleagues, as summarized on Wikipedia, suggested this could raise the top brine temperature from 110 °C to 135 °C while keeping the same calcium sulfate scaling index. Wikipedia describes the idea as unexploited.

Cooling tower make-up water

Another proposed use treats evaporative cooling water as the draw. Cooling towers lose water to evaporation, which concentrates the circulating water. FO can use that concentrated water to pull fresh water from seawater, brackish water, treated sewage or industrial wastewater, replacing evaporation losses without a separate draw-recovery step, since the cooling loop needs the water anyway.

Food and beverage concentration

Concentrating fruit juice, coffee or other liquid foods by heating can damage flavour, aroma, colour and vitamins. FO concentrates at low temperature and low pressure. As of 2006 it had been studied only at laboratory scale. Highlights from the review:

• Popper and colleagues concentrated fruit juices with cellulose acetate membranes and saturated salt as the draw, reaching an average flux of 2.5 kg/m²h. Salt diffused into the grape concentrate, which shows the problem with salt draws in food.
• Beaudry and Lampi recommended a 72 °Brix sugar solution as the draw for juice concentration.
• A patent by Herron and colleagues reported laboratory fluxes as high as 4 L/m²h for orange juice and 6 L/m²h for coffee.
• Wrolstad and colleagues concentrated red raspberry juice by FO using high-fructose corn syrup as the draw. A trained sensory panel found no significant difference from juice concentrated by vacuum evaporation. Cost and energy were not compared.
• Petrotos and co-workers found that membrane thickness and draw viscosity were the main factors controlling flux when concentrating tomato juice.

Osmotic pumps in medicine

The most widespread use of osmotically driven membrane technology is probably inside the body. Osmotic pumps release drugs at a steady rate. The ALZET pump for animal research was commercialized in the mid-1970s, and later human systems such as DUROS and OROS could release drugs continuously for up to a year. In the typical design described in the 2006 review, a titanium cylinder 4 mm in diameter and 40 mm long contains a salt tablet "osmotic engine" behind a polyurethane semi-permeable membrane. As body water diffuses in, the engine swells and pushes a piston that drives the drug out through a small orifice. Because the salt stays saturated and the body's osmotic pressure is roughly constant, the delivery rate stays constant.

Osmotic power: pressure-retarded osmosis

Wherever fresh water meets salt water, as at a river mouth, there is energy available in the salinity difference. The idea of harvesting it was proposed by R. E. Pattle in 1954, and Sidney Loeb outlined a practical membrane method, pressure-retarded osmosis, in the 1970s. In PRO, fresh water crosses a membrane into pressurized seawater. The extra volume of diluted, pressurized seawater is partly sent through a turbine to make electricity and partly used, through a pressure exchanger, to pressurize incoming seawater.

Estimates of the global resource vary. Cath and colleagues cited a potential on the order of 2,000 TWh per year. The Wikipedia article cites a theoretical worldwide potential of 1,650 TWh per year. These are theoretical figures for how much energy is available in salinity gradients, not projections of what could be economically built.

The most prominent test was in Norway. The state-owned utility Statkraft, working with the research institute SINTEF, built a prototype at Tofte in Hurum on the grounds of a cellulose factory. It was officially opened by Crown Princess Mette-Marit and began test power production on 24 November 2009. Its expected output was only 2 to 4 kW, using 10 litres per second of fresh water and 20 litres per second of salt water. An earlier EU-funded research effort had set a goal of membranes producing at least 4 W per square metre. On 20 December 2013, Statkraft announced it was discontinuing its osmotic power work, saying the technology would not become competitive within the foreseeable future. The planned larger pilot at Sunndalsøra and the future commercial plant were never built.

The Norwegian experience mirrors FO's membrane problem. As the 2006 review put it, the future of PRO and of FO depend on the same kind of semi-permeable membrane, one that combines high flux, high rejection, low internal polarization and high strength, and that at the time did not commercially exist.

Fouling: genuinely lower, or just different?

A recurring claim is that FO fouls less than pressure-driven processes. The 2006 review was careful: FO "may have a lower membrane fouling propensity." The usual explanation is that without high hydraulic pressure pushing foulants against the membrane, the fouling layer is looser and easier to remove. The leachate and centrate experiments support this. In both, flux declined during operation and was substantially or fully restored by simple cleaning.

But the evidence is not one-sided, and the Wikipedia article itself states the contrast. One section says FO has "inherently low fouling" in desalination, while another says a main drawback in industrial use is high fouling when treating heavily loaded streams, which forces shutdowns for cleaning. Both can be true. Lower fouling in a given application does not mean no fouling, and the reverse solute flux that is unique to FO can itself worsen fouling by changing the chemistry near the membrane surface. Raw centrate in the Reno study, with high suspended solids, fouled more and did not fully recover after cleaning.

Strengths and limitations at a glance

Common misconceptions

"Forward osmosis desalination uses less energy than reverse osmosis." Not for seawater, when the full process including draw regeneration is counted. The MIT analysis shows that regenerating a draw solution more concentrated than seawater cannot beat direct RO on thermodynamic minimum energy. FO can win when the second step is unnecessary, when cheap waste heat is available, or when the feed is too difficult for RO.

"FO needs no energy." The water-transfer step needs little pumping energy, but liquids must still be circulated on both sides of the membrane, and in most cases the draw must be regenerated.

"FO is the opposite of reverse osmosis." It is better thought of as the natural process that RO reverses. FO, PRO and RO are three regimes of the same flux equation, distinguished by how much hydraulic pressure is applied.

"Any RO membrane can be used for FO." RO membranes can operate in FO mode, but their thick porous supports cause severe internal concentration polarization, which is why the first purpose-built FO membrane outperformed several RO membranes by factors of roughly nine to thirty in one comparison.

"Hydration bags make pure water." They make a sweet drink. The membrane keeps out microbes and most dissolved substances, but the product contains the draw sugar by design.

"The membrane blocks everything." No practical membrane is perfect. Small molecules such as ammonia slip through more easily, and draw solutes leak backward.

Frequently asked questions

What is the difference between forward osmosis and reverse osmosis? Reverse osmosis uses a pump to push water out of a salty solution through a membrane, against its osmotic pressure, producing fresh water directly. Forward osmosis uses a more concentrated draw solution to pull water out of the feed through a membrane without high pressure, producing a diluted draw solution that may then need further treatment.

Why not just drink the draw solution? Sometimes you can, which is the whole point of hydration bags. For most uses, such as municipal drinking water, the draw solute is something you do not want in your water (salt, ammonia, fertilizer), so it has to be removed.

What is a draw solution made of? Sodium chloride is the most common in research and in RO-regenerated systems. Others include sugars (glucose, fructose, sucrose), ammonium salts formed from ammonia and carbon dioxide, fertilizers, magnesium and other multivalent salts, polymers and experimental magnetic nanoparticles.

What is internal concentration polarization? It is the dilution or concentration of solute inside the membrane's porous support layer, where stirring cannot reach. It reduces the effective osmotic driving force, often drastically, and is the main reason FO fluxes are far below theoretical predictions.

Is forward osmosis used commercially? Yes, in niches: hydration bags, osmotic drug pumps, some landfill leachate and industrial wastewater treatment, brine concentration, and small desalination plants. It has not displaced reverse osmosis for large-scale seawater desalination.

What happened to osmotic power? Pressure-retarded osmosis works, as the Norwegian prototype demonstrated, but its output was tiny and the membranes were too expensive and not productive enough. Statkraft halted its programme in 2013.

Does forward osmosis remove bacteria and viruses? A good FO membrane is dense enough to reject microorganisms and most macromolecules. The 2006 review describes the permeate from hydration bags as free of microorganisms in most situations. As with any membrane, defects and seal failures can let contaminants through, so integrity matters.

Why is FO still researched if it loses to RO on energy for seawater? Because its strengths, such as handling difficult feeds, gentle treatment of sensitive products, fouling reversibility and the option to skip draw recovery, matter in applications where RO is weak. Research has shifted toward those niches and toward hybrids in which FO protects or feeds an RO stage.

Evidence quality: what to trust

The foundation of FO science, including the flux equation, the role of concentration polarization and the thermodynamic limits, is well established in peer-reviewed literature and textbooks. The bench-scale performance data quoted here come from peer-reviewed papers and the Cath, Childress and Elimelech review. The Coffin Butte leachate figures come from reports by the company that built the plant, as summarized in that review, so they are not fully independent. The energy-savings figures that companies have published, such as Modern Water's "up to 30 percent," are marketing claims and should be weighed against the MIT thermodynamic analysis. Global osmotic power estimates are theoretical upper bounds. A 2015 review by Devin Shaffer and colleagues in Desalination was pointedly titled "Forward osmosis: Where are we now?", which sums up the field's mood after the optimism of the 2000s: real progress on membranes, and a more sober view of where FO can actually beat the alternatives.

Summary

Forward osmosis lets osmosis move water across a membrane into a concentrated draw solution without high-pressure pumps. Its physics are simple. Its engineering is not. Real fluxes are cut sharply by concentration polarization inside the membrane's support, draw solutes leak backward, and, above all, the water ends up in a solution that usually has to be separated again. For seawater desalination, that second step makes FO less energy-efficient than reverse osmosis. Where the diluted draw is itself useful, where the feed is too dirty or salty for RO, where heat is cheap, or where a gentle process is needed, FO has proved its worth, in drink pouches, drug pumps, landfill leachate plants and brine concentrators. Its long-term future depends, as it has since the 1970s, on better membranes and better draw solutions.

Sources

• Forward osmosis (Wikipedia)
• Cath, Childress and Elimelech, "Forward osmosis: Principles, applications, and recent developments," Journal of Membrane Science 281 (2006) 70–87
• MIT News: "Study shows forward osmosis desalination not energy efficient" (2014), on McGovern and Lienhard, Journal of Membrane Science 469 (2014) 245–250
• McGovern and Lienhard paper record, DSpace@MIT
• McCutcheon, McGinnis and Elimelech, "A novel ammonia–carbon dioxide forward (direct) osmosis desalination process," Desalination 174 (2005) 1–11
• Yip, Tiraferri, Phillip, Schiffman and Elimelech, "High performance thin-film composite forward osmosis membrane," Environmental Science & Technology 44 (2010) 3812–3818
• Phillip, Yong and Elimelech, "Reverse draw solute permeation in forward osmosis: modeling and experiments," Environmental Science & Technology (2010)
• Phuntsho et al., "A novel low energy fertilizer driven forward osmosis desalination for direct fertigation," Journal of Membrane Science 375 (2011) 172–181
• Shaffer et al., "Forward osmosis: Where are we now?", Desalination 356 (2015) 271–284
• Statkraft osmotic power prototype in Hurum (Wikipedia)
• Statkraft: "Statkraft halts osmotic power investments" (2013)
• Modern Water (Wikipedia)
• WaterWorld: "Forward osmosis membrane revenues dry up for Modern Water"
• Osmosis (Wikipedia), for Nollet, Dutrochet and Traube