Desalination 312 (2013) 10–18

Contents lists available at SciVerse ScienceDirect

Desalination

j ourna l homepage: www.e lsev ie r .com/ locate /desa l

Using forward osmosis to teach mass transfer fundamentals to undergraduatechemical engineering students

Daniel Anastasio, Jeffrey R. McCutcheon ⁎Department of Chemical, Materials, and Biomolecular Engineering, Center for Environmental Sciences and Engineering, University of Connecticut, 191 Auditorium Rd. Unit 3222, Storrs,CT 06269-3222, USA

H I G H L I G H T S

► Forward osmosis is used to teach mass transfer to undergraduate engineering students.► A robust, user friendly test system was developed for use by undergraduate students.► Osmotic flux data was collected solely by undergraduate students during class.► Concentration polarization modeling allowed for prediction of osmotic water flux.► Student feedback on the laboratory experiment was largely positive.

⁎ Corresponding author.E-mail address: jeff@engr.uconn.edu (J. R. McCutche

0011-9164/$ – see front matter © 2012 Published by Elhttp://dx.doi.org/10.1016/j.desal.2012.10.037

a b s t r a c t

a r t i c l e i n f o

Article history:Received 27 August 2012Received in revised form 27 October 2012Accepted 29 October 2012Available online 12 December 2012

Keywords:Forward osmosisPressure retarded osmosisMembrane separationsConcentration polarizationMass transfer

A crossflow forward osmosis (FO) system was constructed to introduce senior-level chemical engineeringstudents to key facets of membrane separations and mass transport. Undergraduate students tested a commer-cial FOmembrane at various draw solution concentrations while monitoring mass flow rate into the draw solu-tion, feed solution concentration, and temperature. Ultimately, students used the data they collected to estimateeach membrane's hydraulic permeability coefficient (A) and the salt permeability (B). Mass transport boundarylayer theory was utilized to predict theoretical hydraulic flux at system operating conditions. These values werecompared to experimental results and discrepancies are discussed, drawing attention to potentially erroneoussimplifying assumptions. While an excellent illustration of mass transport on its own, coupling this systemwith a reverse osmosis (RO) experiment granted students a complete understanding of hydraulic membraneseparations and mass transport boundary layer phenomena.

© 2012 Published by Elsevier B.V.

1. Introduction

The increased need for clean water worldwide has prompted anincreased use of nontraditional sources that includewastewater, brackishwater, and seawater. Desalination technologies can be implemented totreat these waters, but the high operating and capital costs limit theirwidespread use to arid regions where few other freshwater sources areavailable.

The high costs of desalination have spurred efforts to develop desali-nation alternatives. One such technology is known as forward osmosis(FO), which utilizes an osmotic pressure gradient to drive water fluxthrough a membrane. Water flows naturally from the feed into a highlyconcentrated draw solution, which is designed such that the draw soluteis easier to extract from water than the feed solutes. Therein lies theprimary advantage of FO over a conventional membrane desalinationtechnique such as reverse osmosis (RO): water transport is enabledwithout requiring an applied pressure. The energy requirements are

on).

sevier B.V.

instead directed toward regeneration of the draw solute, where the sep-aration technique can be chosen and optimized based on the solutesavailable. This unique feature makes FO a cutting-edge separations tech-nology [1,2].

Aswater treatment, desalination, andmembrane technology becomemore commonplace in industrial processes and separations, employerswill demand that new engineering students have the knowledge andskills that prepare them to be an engineer in the 21st century. As such,curriculums must continually be tuned to incorporate new material,especially in the capstone laboratory course common tomany engineer-ing disciplines. In this study, a crossflow FO test systemwas constructedfor the Chemical Engineering (CHEG) Laboratory curriculum at theUniversity of Connecticut (UConn) as part of a newly implementedmembrane separations laboratory module. The first part of the moduleincludes a crossflow reverse osmosis (RO) and nanofiltration (NF) sys-tem that embodies a more conventional membrane separations ap-proach of pressure driven filtration. The second part of this module isthe FO component. In this lab, students are asked to analyze membranewater and salt flux behavior in forward osmosis conditions under awide variety of operating conditions. The first of its kind used to educate

11D. Anastasio, J. R. McCutcheon / Desalination 312 (2013) 10–18

undergraduate students on the basics of FO, this experimental system isdesigned to reinforce mass transfer fundamentals learned in the trans-port phenomena lecture course common to any CHEG curriculum.

FO serves as an excellent platform for experiential teaching of masstransport fundamentals using the context of novel membrane separa-tions. During the experiment, students gain experience evaluating keyFOmembrane performance characteristics such aswaterflux and soluteflux. The ratio of these two values, commonly known as the specificreverse solute flux, allows students to understand how effective amembrane when operating in FO. This experiment is also designed toreinforce basic mass transfer boundary layer theory through an examina-tion of concentration polarization (CP) and how performance variablesare impacted by the orientation of the membrane. In FO, CP impactsthe solute concentration at the membrane interface, which results indecreased driving force and membrane performance. Students will ob-serve how the degree of CP is impacted by various process parameters,such as cross flow velocity rate and membrane orientation. Ultimately,students must determine which membrane and/or operating conditionsare best-suited for FO, considering membrane permselectivity andreasonable goals for an FO process. Further discussion can be fosteredby having students consider other parameters that are vital to a completeFO desalination process, such as draw solution chemistry and recoverymethods, These samemethods have been used only in very recent inves-tigations throughout the FO community and students participating in thislab are among the first undergraduates in the world to learn these skills.

2. Theory

Forward osmosis theory is presented to students via a review paperwritten by Cath, Childress, and Elimelech [3]. The water flux, Jw, is com-monly measured in either gallons per square foot of membrane per day(GFD) or liters per square meter of membrane per hour (LMH). The ex-periment generates data in the form of amass flow rate of water, whichstudents can convert to a volumetric flow rate using density. Normaliz-ing this flow rate bymembrane area results in flux. The generalized fluxequation for FO is shown as Eq. (1).

Jw ¼ A Δπ−ΔPð Þ ð1Þ

where A is the hydraulic permeability constant, Δπ is the osmotic pres-sure difference across the membrane, and ΔP is the applied pressuregradient (which is 0 when feed and draw solutions are at equal hydrau-lic pressure). The osmotic pressure of a given solution, π, can be calcu-lated using the idealized form of the van't Hoff equation, below:

π ¼ iCRT ð2Þ

where i is the ionic dissociation constant, C is themolar concentration ofsolute, R is the gas constant, and T is the temperature. As i, R, and T areconstant during each experimental test, it can be said that C has a direct,linear relationship with osmotic pressure π. This is a reasonable as-sumption for the concentrations of the solutions considered here.

Concentration polarization has long been a topic of discussion andresearch in reverse osmosis. In FO, permeate gives rise to an additionalboundary layer at the draw-solution interface of the selective layerwhich results in a lower concentration of salt at the membrane inter-face as shown in Fig. 1, resulting in a lower net driving force. This phe-nomenon has been previously examined and is based on the CPmodeling work of McCutcheon and Elimelech [4]. After setting upand solving a simple shell balance around the boundary layer, stu-dents can determine that the CP equation for the draw solution in FOis as follows:

Cm;d−Cf

Cb;d−Cf¼ exp

−Jwk

� �ð3Þ

where Cm,d is the draw concentration at the membrane interface, Cb,d isthe draw concentration in the bulk solution, Cf is the bulk concentrationof the feed, and k is the mass transfer coefficient. The appropriateosmotic pressure terms (πm,d, πb,d, πf) can replace the concentrationterms in Eq. (3) as osmotic pressure is directly proportional to concen-tration. Eq. (3) can be used to calculate concentration of solute at themembrane interface, which can be used to determine the CP modulus(Cm,d/Cb,d), a quantity always less than one for FO. Low CP moduli indi-cate large boundary layers and are caused by low fluid crossflow rates(low k) or high water flux values.

On a fundamental level, the mass transfer coefficient, k, is defined asthe ratio of the molecular diffusivity constant D to the thickness of thesolute boundary layer δ (k=D/δ). Students will have difficulty usingthis definition, however.While there are numerous tables and equationsfor the determination of D, there is no way to measure the thickness ofthe boundary layer. Therefore, the most practical way to determine k isto evaluate the Sherwood number (Sh=kdh/D) using correlations pro-vided by Mulder for flow in a channel [6]. Now, Eqs. (1), (2) and (3)can be combined and iterated to estimate the water flux for the knownbulk concentrations, shown below:

Jw ¼ A Δπð Þ ¼ A πm;d−πf

� �ð4Þ

(from Eq. (1))

πm;d−πf ¼ πb;d−πf

� �exp

−Jwk

� �ð5Þ

(from Eqs. (2) and (3))

Jw ¼ A πb;d−πf

� �exp

−Jwk

� �ð6Þ

Note that this model assumes that the membranes are run in PROmode and that feed concentration is sufficiently low or the mem-brane is very selective such that internal CP is negligible. Thisassumption is not always true, but for undergraduate level masstransfer, it provides reasonably accurate results, much as it did forprevious studies [4]. For further background on CP for RO applica-tions, refer to Sablani et al. [5].

Finally, the solute flux, commonly referred to as Js, is calculatedsimilarly to water flux, as shown in Eq. (7) below:

Js ¼ B ΔCeff

� �ð7Þ

where B is the solute permeability coefficient and ΔCeff is the effec-tive concentration gradient, which is equal to Cm,d−Cf. Note thatEq. (7) accounts for external CP. It is typically desired to calculate Aand B values for membranes using a reverse osmosis test; thesevalues would be considered a more accurate A and B as concentra-tion polarization (CP) effects are more easily quantifiable in RO.The RO permeability values can be used as a point of comparison tostudent-generated FO data.

3. Experimental overview

This undergraduate experiment explores howmembrane properties(material and orientation) and operating conditions (draw concentra-tion, operating temperature, and flow rate) influence the osmotic fluxthrough amembrane. A systemdiagramof the cart-mounted FO systemis presented as Fig. 2. A FO membrane coupon that had been storedovernight in deionized (DI) water is sealed within the membrane cell.Feed and draw tanks are filled with DI water. The draw tank is placedon a balance and both solutions are circulated through the system topurge any air out of the lines which will cause errors in flux measure-ment. Once the balance has stabilized, concentrated saline stock

Fig. 1. Diagram showing concentration polarization CP phenomena for osmotic flow in PRO mode with no transmembrane hydraulic pressure gradient and deionized water feed.The dark gray vertical line is the active layer of the membrane, and the light gray vertical line is the support layer of the membrane. Note that salt concentration (indicated by theblack line) of the draw decreases with proximity to the membrane interface, lowering the effective driving force (πm,d−πf). This diagram assumes no salt crossover from the drawsolution.

12 D. Anastasio, J. R. McCutcheon / Desalination 312 (2013) 10–18

solution is added to the draw solution, and draw solution mass changeis recorded every minute. At regular time intervals, additional salinestock solution is added to the draw to increase the concentration, andthe conductivity and temperature of the feed are recorded. From theseobservations, thewaterflux and saltflux for each draw solution concen-tration can be calculated, permitting the estimation of the water and

Fig. 2. Schematic for cart-mounted cr

solute permeability constants for the membrane. Furthermore, thedraw solution concentration at the membrane surface can be predictedat a given set of experimental parameters (draw concentration,crossflow rate, etc.) using boundary layer film theory [6,7]. This predic-tion is then compared to the experimental data, and the quality of dataand validity of model assumptions are assessed.

ossflow forward osmosis system.

Fig. 3. Photographs of forward osmosis test cell; (a) exterior, (b) interior.

13D. Anastasio, J. R. McCutcheon / Desalination 312 (2013) 10–18

Individual FO tests require approximately 3.5 h for a complete exam-ination of one membrane coupon, including set-up, water and salt fluxmeasurements at four draw concentrations, and a system flush. Prior tothe experiment, students read an operation manual [8]. A two-day labo-ratory schedule provides students with the opportunity to test twodifferent membrane parameters (orientation, material, etc.). Due to thelarge number of independent variables related to non-membraneprocess parameters (feed and draw concentrations, feed and drawflow rates, temperature, etc.) the schedule can be expanded tomultiple days, or different student groups can be assigned differentexperimental variables to evaluate. At UConn, however, the experi-ment has been coupled with a reverse osmosis experiment for asix-day membrane separations experimental module. The RO exper-iment utilized in this module tasks students with characterizingmembranes for their hydraulic permeability and solute permeabilitywhile encouraging them to explore how process conditions impactmembrane performance and CP [9]. The two systems are run simul-taneously, allowing students to fully characterize a single FO mem-brane in both orientations with replicate data for error analysis, aswell as acquiring permeability and selectivity data (notably thetrue hydraulic and solute permeability of the FO membrane) fromthe RO tests. The additional data provided by the RO tests can en-hance FO data analysis, but is not required if the experiment is tobe performed as a demonstration of boundary layer effects.

4. Required equipment

4.1. Membrane

This experiment requires the use of commercially available forwardosmosis membranes. Previous investigations have indicated that com-mercial RO membranes exhibit poor flux performance in FO [10,11];therefore, commercial RO membranes are not recommended for thisexperiment as little meaningful result can be garnered. HydrationTechnologies Innovations (HTI) produces two types of FO membraneintended for use in freshwater purification with a sugar-electrolytebased draw solution [12].

For this experiment, the HTI Hydrowell cartridgemembrane is used.This membrane is the same used in previous investigations on FO[13–17]. The membrane is an integrated asymmetric cellulose acetatemembrane that is supported by a woven mesh. HTI makes anothertype of membrane used in its X-Pack and Sea Pack products. Thesecan also be used, though they tend to have lower flux and are supportedby a nonwoven fabric.

Other companies, such as Oasys Water (Boston, MA), are begin-ning to make new commercially available FO membranes using mate-rials other than cellulose acetate. HTI has also developed of a thin filmcomposite membrane [18]. These and other membranes are alsoworth considering for this experiment, though were unavailable atthe time of this study. Note that membranes made of materialsother than cellulose acetate may require additional preparation andstorage steps.

4.2. Cell design

The custom-made FO cell is composed of two identical halves fabri-cated from black delrin with stainless steel plates acting as additionalsupport. Each half has a crossflow channel with dimensions 3″ long by1″ wide by 1/8″ deep that is fed via threaded ports bored into thesides of the cell. Surrounding the channel on the bottom half are twoconcentric o-rings, one approximately 1 cm larger in radius than theother, that are seated in grooves bored into the delrin; the top halfonly contains the smaller o-ring. The small o-rings seal against themembrane, while the outer o-ring creates a watertight seal aroundthe membrane. The two halves are placed on threaded stainless steelrods that are attached to a stainless steel base plate, which can easily

be mounted to a surface. The cell halves are sealed and supportedusing washers and nuts. Fig. 3 shows photographs of the cell interiorand exterior.

Larger cells can be designed to accommodate larger membranecoupons, though we have found this size to be appropriate given theflux measurement method and the desire to preserve membrane. Thiswill give the user a greater ability to measure low fluxes and provide agreater sensitivity to flux changes as the system changes. The currentcell size was chosen for several reasons: 1) it matches the dimensionsof the RO cells used in the previously-mentioned RO experiment [9];2) uses smaller coupons to conserve membrane; and 3) fluxes throughcommercial FO membranes can be accurately measured using thismembrane area. If theflux is too low to be accuratelymeasured through

14 D. Anastasio, J. R. McCutcheon / Desalination 312 (2013) 10–18

this amount of membrane area, then the membrane performance islikely too poor to be considered for this experiment.

4.3. Key system components

To support the system, a cart (36″ tall×30″wide×20″ deep) wasfabricated with an angle-iron frame, painted to resist corrosion, andaluminum plates. Four locking, swiveling wheels were added for in-creased mobility. A 24″ tall aluminum backsplash was added to sup-port pressure gauges and flow meters as well as to provide supportfor stream return lines. Two 10″-wide shelves were added underthe top shelf to hold the draw and feed pumps. Fig. 4 shows a photo-graph of the completed cart. Note that all lines connected to vibrat-ing equipment (pumps, chiller, etc.) pass through a rigid surface,such as the countertop or backsplash to reduce the vibration aroundthe scale during the gravimetric data acquisition. Without this de-sign consideration, system stabilization will become too long to per-mit a thorough membrane examination within a laboratory period.Additional vibrational prevention measures, such as adding a foampad beneath the scale, adding Velcro strips to the scale and drawcontainer, or covering the scale with a wind guard, may also helpachieve faster stabilization.

Two 5-L polyethylene bottles from McMaster-Carr (Princeton, NJ)were selected to serve as the feed and draw tanks. Each tank is connectedthrough the tabletop using flexible color-coded PVC tubing to distinguishfeed anddraw lines. These lines connect toMicropumpvariable-flowgearconsole drive gear pump (fitted with A-Mount cavity style pump heads)from Cole-Parmer (Vernon Hills, IL). The draw solution reservoir sits ona PI-4002 Denver Instruments top-loading balance with 4000.00 g

Fig. 4. Photograph of the com

capacity (Fisher Scientific), which comes packaged with the appropriatedata-logging software, and the feed tank is placed on a magnetic stirplate (Fisher). To regulate the temperature of the streams, a 25′ coil ofwelded 316 stainless steel tubing was cut into thirds, which werefashioned into concentric heat exchange coils. The outermost and inner-most coils are connected to the feed anddraw lines, respectively. The cen-ter coil is connected to a Neslab RTE7 recirculating chiller (Fisher). Allthree coils are submerged in a 5-gal polypropylene tank containing DIwater that is agitated with a magnetic stir plate (Fisher). To conservespace, a shell and tube or concentric tube heat exchanger may besubstituted. Note that the chillermust not be in contact with the cart dur-ing operation to minimize vibrations. A pair of quick-disconnect fittings(without check valves) were placed in the tubing of the feed and drawlines between the pumps and the heat exchanger coils to allow air purg-ing of the lines between tests. After passing through the heat exchanger,the streams are fed through the backsplash and into the cell countercur-rent. Each line is connected to the cell with quick-disconnect fittings tofacilitate cell disassembly. Outlet streams from the cell are directedthrough the backsplash to two glycerin-filled, panel-mountable pressuregauges (0–30 psi, McMaster), followed by two panel-mountable flowmeterswith built-in needle valves (0–1 gpm,McMaster). The feed returnline is connected directly to the feed tank via quick-disconnect fitting.Draw returns to the draw tank via a stainless steel pipe that drains intoa funnel in the draw tank lid, shown in Fig. 5. A tube extends from the fun-nel stem into the draw solution to minimize vibrations from splashing.For best results, the tank should be centered on the scale and the returnshould drain into the center of the funnel. It is crucial that all metalcomponents in the system be 316 stainless steel to minimize the corro-sion high salt concentrations promote. All tubing, pipe, and pipe fittings

pleted FO system cart.

Fig. 5. Photograph of draw solution return line. The return is not directly connected tothe draw tank in order to minimize vibrations on the scale. This is a viable approach fornon-volatile draw solutes.

15D. Anastasio, J. R. McCutcheon / Desalination 312 (2013) 10–18

were purchased from McMaster, unless otherwise specified. A summaryof the approximate cost of each major system component is provided inTable 1.

4.4. Measurement devices

Toutilize the automatic data-logging feature of the scale, a computercapable of running Microsoft Excel with a USB port is needed. It isrecommended that the screen saver and automatic sleep mode bedisabled so data collection is not interrupted. An Oakton conductivityprobe with built-in temperature probe (Fisher) was used to monitor

Table 1Approximate cost of system components.

Component Supplier Approx. cost

Recirculating chiller Fisher Scientific $2000Two gear pump drives and heads Cole-Parmer $2000Denver instruments scale Fisher Scientific $2000Test cell n/a $500Custom cart n/a $500Tanks McMaster $100Meters and gauges McMaster $250Tubing, piping, and fittings McMaster $600Two magnetic stirrers and floating Stir bar Fisher Scientific $300Conductivity probe Fisher Scientific $600Total $8850

the condition of the feed solution. Holes should be drilled in the feedtank lid to accommodate these probes. The conductivity probe shouldbe calibrated to the selected draw solute prior to the test, which can eas-ily be accomplished bymeasuring the conductivity of a serial dilution of a2000 ppm stock solution.

5. Experimental procedure

At least 24 h prior to the experiment, membrane sheets must be cutinto small coupons for student use. The coupons should be stored inrefrigerated DI water to prevent membrane damage and biologicalgrowth. It is advised that the hydraulic and solute permeability ofthese membranes be determined, either through an RO test or bycontacting the manufacturer, as this information can be used to en-hance student analysis and understanding. Again, students couldperform such an RO test themselves as part of the experiment iftime permits [9]. Additionally, two liters of 5 M sodium chloridestock solution should be prepared. While other salts can be consid-ered, the high solubility limit of sodium chloride ensures this stockwill last for four complete membrane tests. This salt is also inexpen-sive when purchased in large quantity. Safety goggles should be wornas part of good laboratory practice.

At the start of each test, a membrane was loaded into the cell.Gloves were worn to minimize damage to the membrane, and themembrane was cut with scissors so it would fit between the smalland large o-rings in the cell. Great care was then taken to place themembrane's active layer against the draw side of the cell for initialtests. This orientation, known as PRO mode, allows for simplifiedmass transfer modeling appropriate for undergraduate students. Stu-dents may elect to run a test in FO mode, where the active layer facesthe feed solution, but the transport phenomena are complicated byinternal concentration polarization caused by the interaction of thedraw solution with the support layer. The membranes may curlslightly when loaded, so a spatula or DI water were used to gentlyflatten the membrane as the top was being placed on the cell to pre-vent folding.

As the feed and draw tanks were filled with 2 L of DI water each,the recirculating chiller was set to approximately 25 °C, adjustingthe setting depending on ambient air temperature to ensure that thefeed was 25 °C during the run. Both the feed and draw pumps wereturned on, and once no bubbles were observed in the lines, flowswere set to 1 LPM using the control knobs on the pumps. Feed anddraw pressure can range from 0 to 5 psi, but both streams were setto the same pressure. The data acquisition software was started torecord the mass of the draw tank. During this equilibration step, thescale registered losses in mass as air is purged from the system. Oncethe minute-to-minute change in draw mass was ±0.05 g, the systemwas considered stable; this process typically took under 10 min if thesystem was well-designed. If the scale does not stabilize in this timeperiod, center the tank on the scale andminimize all sources of air cur-rents, using cardboard or plastic to construct a wind shield if neces-sary. Once the system was stable, saline stock solution was added tothe draw, which increased the stream concentration to a known mo-larity, and the conductivity and temperature of the feed weremeasured. The data acquisition software collected 1–2 points of dataper minute for 30 min. After 30 min, the draw solution concentrationwas increased, feed conductivity and temperature were measured,and the process was repeated. Recommended draw solution con-centrations are 0.05 M, 0.1 M, 0.5 M, and 1 M. At the highest con-centration, students usually alter an independent variable, such astemperature or flow rate, and continue to take data as time permits.Table 2 summarizes typical testing conditions utilized by students inthe unit operations laboratory.

Once a membrane test was complete, the tanks were emptiedand refilled with DI water. The outlet lines were extended with flex-ible tubing placed in a bucket or drained directly into a sink. The

Table 2Typical operating conditions for student experiments.

Variable Typical draw-sidevalue

Typicalfeed-side value

Temperature (°C) 25 25Gauge pressure (psi) 0 0NaCl concentration (M) 0, 0.05, 0.1, 0.5, 1.0 0Initial DI water volume (L) 2 2Hydraulic flow rate (L/min) 1–2 1–2Frequency of massmeasurement(recordings/min)

1 n/a

Fig. 6.Water flux as a function of driving force (osmotic pressure for FO and applied pres-sure for RO) for HTI cartridge membrane (25 °C, feed and draw flow rate at 1 L/min). Thelower dashed line is a linear trend line for the FO data. Note that 1 gfd is approximately1.7 L/m2 h. The data points are an average of three student tests, with the error barsrepresenting standard deviation.

Fig. 7. Parity plot of experimentally observed experimental water flux versus waterflux predicted using boundary layer film theory model for draw concentrations of 0,0.05, 0.1, 0.5, and 1.0 M sodium chloride in PRO mode. Error bars indicate standarddeviation.

16 D. Anastasio, J. R. McCutcheon / Desalination 312 (2013) 10–18

water was pumped through the system into the sink or bucket, and thetanks were refilled with DI water as needed. Flushing continued untilthe conductivity of each outlet stream was below 10 microsiemens(μS). If DI water is in short supply, tap water may be used for an initialflush and about 1 gal of DI water can be used per tank as a polishingwash. Filtered compressed air was then used to purge the lines ofresidual water by connecting the air line to one of the air purge ports.Low-pressure air was applied until the lines had been evacuated.Great care should be taken during this step to avoid spiking the pressuregauges. Once both lines are purged, the membrane was removed fromthe cell, checked for defects, and discarded.

The versatility of this system permits many other tests using differ-ent independent variables. The independent variables mentioned hereare membrane type, membrane orientation, draw concentration, andtemperature. Additionally, feed and draw flow rate, draw solute type,and feed solution concentration can also be varied.

6. Typical student results and discussion

The primary results students will present when running this experi-ment are water flux (Jw) and salt flux (Js). To facilitate experimental anal-ysis, students frequently assume the feed and draw solutions are dilute;therefore, the solution is ideal, the properties ρ and μ can be evaluatedfor pure water at the appropriate temperature, and solute diffusivity Dcan be approximated using the Nernst-Haskell equation. While thisassumption is always appropriate for the feed, it becomes invalid fordraw solution concentrations greater than 0.1 M. Students find the as-sumption of ideal draw solution remains a tolerable approximation forthe purposes of this experiment. If other draw solutes are used, studentsmay choose to measure density and viscosity for more accurate masstransfer modeling.

Flux is plotted against the observed osmotic pressure difference,as shown in Fig. 6, which summarizes the results of several FO testsfor the HTI cartridge membrane in PRO mode with standard deviationerror bars. Note that the FO fluxes are lower than pure water flux inRO under equivalent pressure driving force (the upper dashed linein Fig. 5). The lower flux is a result of CP, which reduces the osmoticdriving force.

A parity plot of predicted water flux based on film theory and waterflux measured by a group of students is presented as Fig. 7. This datawas collected during pilot testing of the system during the laboratorycourse. The students accounted for external CP on the draw side of themembrane when analyzing their data. By performing this calculationusing Eq. (6), the students observed a decrease in external CP moduluswith increased flux, indicating a worsening of concentration polariza-tion effects. The plot shows that themodel is very accurate at predictingfluxes at low draw solution concentrations. As higher concentrationsare reached, the model tends to over-predict flux, likely caused by themodel's neglecting of internal CP effects caused by salt diffusion acrossthe membrane or the assumption of ideal draw solution. Membranesrun in FOmodewill exhibit lower flux due to significant internal CP act-ing directly on the draw solution. For further reading on internal CP andimpact of membrane orientation, consult McCutcheon et al. and Gray et

al. [19–21]. During the lab, students are encouraged to search thepeer-reviewed literature to learn about these more advanced concepts.

Salt flux is calculated by normalizing mass flow rate of salt into thefeed tank by membrane area. The summary of salt flux data at effectiveconcentrations (accounting for CP on the draw side) for several testsmade in the PRO mode using the HTI cartridge membrane is presentedas Fig. 8. Students will see that for high-concentration draw solutions,there is a high salt crossover into the feed solution in addition to a highwater flux. In order to help students understand how to optimize theFO process, they must calculate the specific reverse salt flux [22,23],which is the ratio of the solute flux to the water flux and typically hasunits of g/L. In other words, specific salt flux denotes the mass of solutethat permeates into the feed solution per volume of water that passesinto the draw. Ideally, this value should be as small as possible, denotingthat saltflux ismuch smaller thanwaterflux. For theHTI cartridgemem-branes, students have observed specific reverse salt fluxes between 0.85and 1.3 g/L during their experiments.

There are a number of other independent variables that can bechanged to demonstrate other fundamental mass transfer concepts.Feed and draw flow rates can be altered to see the impact of crossflow

Fig. 8. Observed salt flux for HTI cartridge membrane in PRO mode as a function ofeffective concentration gradient. Error bars indicate standard deviation.

17D. Anastasio, J. R. McCutcheon / Desalination 312 (2013) 10–18

velocity, and by extension Reynolds number, on water flux through themembrane. If flow velocity is been altered, it is expected that more tur-bulent flowswill diminish the thickness of the boundary layer, decreas-ing the severity of CP and increasing water flux. This study would helpstudents understand the interplay between the variables that contrib-ute to CP effects. Students could effectively mitigate the CP effects athigh draw concentrations by increasing fluid velocity to decrease thethickness of the boundary layer. Varying temperature can help studentslearn how water viscosity impacts water and salt permeability coeffi-cient. Different draw solutes, such as sugars, multivalent salts, andblended solutes, would provide students an opportunity to learn moreabout colligative properties and the properties of an ideal solution. Formore advanced students, both PRO and FOmode can be tested and stu-dents challenged with explaining why the fluxes for the same mem-brane in two orientations are different. The system is extremelyversatile, and if students can alter these variables independent of oneanother, they will be able to hypothesize if, how, and why water andsalt flux will change and justify their reasoning with mass transfer the-ory. When coupled with an RO experiment, students will gain an evenbetter understanding of membrane processes while gaining a compre-hensive understanding of concentration polarization phenomena.

All data presented in this manuscript has been generated by UConnchemical engineering undergraduate students using the experimental

Fig. 9. Student opinion of the reverse and forward osmosis experiments at UConn comparedscale of one to five, with 5 being the highest option. Error bars indicate standard deviation

apparatus and the HTI cartridge membrane oriented in PRO mode.The data was collected over the course of five separate runs of the sys-tem. Student feedback regarding this experiment has been positive. Atthe end of the Fall 2011 semester, students were asked to evaluatehow interested they were in the topic, how engaged they felt by the ex-periment as it was performed, and how much they felt they learnedfrom all experiments they performed were on a scale of 1 to 5, where5 was the highest mark. The results of this survey are shown in Fig. 9.Student opinion of the osmotic separations module (which containedboth a reverse and forward osmosis component) was higher than theaverage student opinion of all the experiments in general. Students be-come comfortable with the material if given an overview paper on thetechnology, such as the FO review prepared by Cath et al. [3] or thePRO review prepared by Achilli and Childress [24]. Upon completingonemembrane test, the students are able to test additional membraneswithout further assistance. Many students note a large amount of downtime during the experiment, as data acquisition takes 30 min per drawsolution concentration (less time is also acceptable), so they wereencouraged to review literature pertaining to the experiment duringthis time. Early tests were hampered by a failure of the data acquisitionsoftware mid-test; however, disabling sleep mode and screensaver onthe logging computer solved this problem. In their laboratory reports,students were able to determine water and solute fluxes accurately.Students were challenged by the flux modeling component of theexperiment, but many said this component helped them understandfactors that contribute to CP. In Spring 2011, students were given theopportunity to spend four laboratory periods using this system in con-junction with a crossflow RO system [9] to characterize various NF,RO, and FO membranes utilizing an array of independent variables,such as membrane orientation, solute type and concentration, flowrate, system backpressure, and temperature. This exercise allowedstudents to freely alter these multiple variables to grant them acomplete understanding of what impacts permeability, selectivity,and CP using both established and emerging water desalinationapproaches.

7. Conclusions

This paper describes the design and implementation of a forwardosmosis membrane testing system for a capstone-level chemicalengineering laboratory course. The system is novel in that it is the

to the average opinion of other experiments in the curriculum. Ratings were given on a.

18 D. Anastasio, J. R. McCutcheon / Desalination 312 (2013) 10–18

first FO system designed specifically to be robust and easy-to-use foreducational purposes. FO offers a unique learning opportunity for stu-dents in that they can explore a separations process that does not relyon temperature or applied pressure as a driving force. Students willlearn to characterize membrane permeability and selectivity will ob-serve how these properties change with membrane characteristics(structure, material, and orientation), hydrodynamic conditions, drawsolute concentration and/or type and temperature. The large numberof variables allow formanypotential laboratory scenarios.Mass transferboundary layer theory is reinforced by this experiment through a fluxmodeling component of the analysis. While the system functions verywell as a stand-alone experiment, coupling it with a reverse osmosisexperiment gives students a complete exposure to established andemerging membrane desalination technologies, highlighting similari-ties, differences, benefits, and drawbacks of each.

NomenclatureA Hydraulic permeability constant [L m−2 h−1 bar−1]B solute permeability constant [g m−2 h−1 M−1]C Solute concentration [mol/L (M)]D Molecular diffusivity of solute in water [m2/s]I Ionic dissociation constant of solute [mol ions/molmolecules]Js Solute flux [g m−2 h−1]Jw Volumetric water flux [L m−2 h−1 (LMH)]K Mass transfer coefficient [m/s, or LMH]P Pressure [bar]R Ideal gas constant [0.08314 bar L mol−1 K−1]Sh Sherwood numberT Temperature [K]

Subscriptsb Property of bulk solutiond Property of draw solutionf Property of bulk feed solutionm Property of solution at membrane interface

Greekδ Boundary layer thickness [m]μ Fluid viscosity [kg m−1 s−1]π Osmotic pressure [bar]ρ Fluid density [kg/L]Δ Difference evaluated between feed and draw interfaces

Acknowledgments

The authors would like to graciously acknowledge the Universityof Connecticut Chemical, Materials, and Biomolecular Engineering

Department. Additional funding for this project was provided by theRobert and Beatrice Mastracchio Endowed Scholarship. The authorswould also like to acknowledge Hydration Technology Innovationsfor their generous donation of membranes, and Jason Arena andEmily Cole, for their assistance with preliminary system testing andfine-tuning. Additional data for this manuscript was gathered byChristine Duvall, Nicholas Fluet, and Michael Pomykala as part oftheir laboratory experiment.

References

[1] McGinnis, McCutcheon, Elimelech, A novel ammonia-carbon dioxide osmoticheat engine for power generation, J. Membr. Sci. 305 (2007) 13–19.

[2] http://ngm.nationalgeographic.com/big-idea/09/desalination-pg2 (accessedDecember 26, 2010).

[3] Cath, Childress, Elimelech, Review: forward osmosis: principles, applications, andrecent developments, J. Membr. Sci. 281 (2006) 70–87.

[4] J. McCutcheon, M. Elimelech, Modeling water flux in forward osmosis: implica-tions for improved membrane design, AlChE J. 53 (2007) 1736–1744.

[5] S. Sablani, et al., Concentration polarization in ultrafiltration and reverse osmosis:a critical review, Desalination 114 (2001) 269–289.

[6] Mulder, Marcel, Basic Principles of Membrane Technology, 2nd Edition KluwerAcademic Publishers, 1996.

[7] Baker, W. Richard, Membrane Technology and Applications, 2nd Edition JohnWiley & Sons, Ltd, 2004.

[8] http://www.engr.uconn.edu/~daniel/www/FO2010.pdf (accessed December 26,2010).

[9] D. Anastasio, J. McCutcheon, Teaching mass transfer and filtration using crossflowreverse osmosis and nanofiltration: an experiment for the undergraduate unitops lab, CEE Chem. Eng. Educ. 46 (2012) 19–29.

[10] McCutcheon, McGinnis, Elimelech, A novel ammonia-carbon dioxide forward(direct) osmosis desalination process, Desalination 174 (2005) 1–11.

[11] Lee, Baker, Lonsdale, Membranes for power generation by pressure-retardedosmosis, J. Membr. Sci. 8 (1981) 141–171.

[12] http://www.htiwater.com(accessed April 20, 2012).[13] Martinetti, Childress, Cath, High recovery of concentrated RO brines using for-

ward osmosis membrane distillation, J. Membr. Sci. 331 (2009) 31–39.[14] Achilli, Cath, Childress, Power generation with pressure retarded osmosis: an ex-

perimental and theoretical investigation, J. Membr. Sci. 343 (2009) 42–52.[15] Esperanza, McCutcheon, Elimelech, Performance evaluation of sucrose concentra-

tion using forward osmosis, J. Membr. Sci. 338 (2009) 61–66.[16] J. McCutcheon, M. Elimelech, Desalination by ammonia-carbon dioxide forward

osmosis: influence of draw and feed solution concentrations on process perfor-mance, J. Membr. Sci. 278 (2006) 114–123.

[17] Wang, Ong, Chung, Double-skinned forward osmosis membranes for reducinginternal concentration polarization within the porous sublayer, Ind. Eng. Chem.Res. 49 (2010) 4824–4831.

[18] T. Pankratz, Water Desalin. Rep. EDS Barcelona Spec. 48 (17) (April 25, 2012).[19] J. McCutcheon, M. Elimelech, Influence of concentrative and dilutive internal concen-

tration polarization on flux behavior in forward osmosis, J. Membr. Sci. 284 (2006)237–247.

[20] J. McCutcheon, M. Elimelech, Influence of membrane support layer hydrophobic-ity on water flux in osmotically driven membrane processes, J. Membr. Sci. 318(2008) 458–466.

[21] Gray, McCutcheon, Elimelech, Internal concentration polarization in forwardosmosis: role of membrane orientation, Desalination 197 (2006) 1–8.

[22] N. Hancock, T. Cath, Solute coupled diffusion in osmotically driven membraneprocesses, Environ. Sci. Technol. 43 (2009) 6769–6775.

[23] Phillip, Yong, Elimelech, Reverse draw solute permeation in forward osmosis:modeling and experiments, Environ. Sci. Technol. 44 (2010) 5170–5176.

[24] A. Achilli, A. Childress, Pressure retarded osmosis: from the vision of Sidney Loebto the first prototype installation — review, Desalination 261 (2010) 205–211.