Sustainable operation of membrane distillation for enhancementof mineral recovery from hypersaline solutions

Kerri L. Hickenbottom, Tzahi Y. Cath n

Colorado School of Mines, Golden, CO, USA

a r t i c l e i n f o

Article history:Received 31 July 2013Received in revised form15 November 2013Accepted 17 December 2013Available online 25 December 2013

Keywords:Membrane distillationDesalinationHypersaline brineScalingMembrane cleaningEvaporation pond

a b s t r a c t

Membrane distillation (MD) is an emerging desalination technology that has the ability to desalinatehypersaline brines, including those used in mineral production. MD can potentially replace evaporationponds in conventional mineral production processes because of its small footprint and ability to utilizeindustrial low-grade heat. In the current study MD was investigated for sustained water recovery andconcentration of hypersaline brines. Direct contact MD (DCMD) experiments were performed with waterfrom the Great Salt Lake (4150,000 mg/L total dissolved solids) as the feed stream and deionized wateras the distillate stream. DCMD was able to concentrate the feed solution to twice its originalconcentration, achieving close to complete inorganic salt rejection. During experiments water fluxdeclined to 80% of its initial value (from 11 to 2 L m�2 h�1). Real-time microscopy revealed thatprecipitation of salts on the membrane surface was the main contributor to the decline in water flux. Theapplication of novel scale-mitigation techniques was highly effective in preventing scale formation onmembrane surfaces, sustaining high water flux and salt rejection, and eliminating chemical consumptionused for membrane cleaning. MD was compared to natural evaporation and was found to potentiallyreplace 4047 m2 (1 acre) of evaporation ponds with approximately 24 m2 (259 ft2) of membrane area andto be nearly 170 times faster in concentrating hypersaline brines.

& 2013 Elsevier B.V. All rights reserved.

1. Introduction

As population grows, an increased stress is placed on naturalresources [1]; thus, there is a need for more holistic approachesto process intensification, in which process waste is considereda resource. For example, in desalination, brine is considereda byproduct and treated as a waste stream, whereas in mineralproduction water is considered a byproduct and, as commonpractice, is evaporated to the atmosphere. Ironically, the wastestream of one process is the product of the other. Efficientutilization of brines and on-site energy resources could result inproduction of both water and high-value minerals for beneficialuse, including potable water for urban areas and minerals forfertilizers or road deicing.

In mineral production, evaporation ponds are traditionally utilizedfor concentration of saline water and precipitation of minerals, whichare then further processed in chemical plants. Evaporation pondscommonly use large areas, they are time and energy intensive, andwhen used, large volumes of valuable water are lost to the atmosphere[2]. In order to improve the efficiency of mineral recovery, replacement

of evaporation ponds with desalination processes could minimize landuse and increase water recovery from hypersaline streams.

Current engineered processes for desalination of brackish waterand seawater include thermal distillation or membrane processessuch as reverse osmosis (RO), nanofiltration (NF), and electrodia-lysis (ED). Conventional thermal distillation processes are capableof achieving high water recovery, but they are limited by high-energy consumption needed to heat the feed stream [3]. While RO,NF, and ED are commonly utilized membrane processes fordesalination [2], when feed solutions are highly concentrated orapproach saturation, these processes are limited by operatingpressures (RO and NF) or applied voltage (ED), and in many casesmembrane scaling [4].

Alternatively, membrane distillation (MD) is a novel and uniquemembrane process that can synergistically assist in mineralrecovery and simultaneously produce pure water. MD is a ther-mally driven membrane process in which the driving force formass transfer of water is the partial vapor pressure differenceacross a microporous hydrophobic membrane. Thus, compared tohydraulic pressure and electric field driven membrane processes,MD is minimally affected by increased salt concentrations [5]. Indirect contact MD (DCMD), a warm feed stream (e.g., brine) and acooler fresh water stream (e.g., deionized water) are in directcontact with the active and support sides of the membrane,respectively [5]. In DCMD, water evaporates from the feed solution

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/memsci

Journal of Membrane Science

0376-7388/$ - see front matter & 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.memsci.2013.12.043

n Corresponding author. Tel.: þ1 303 273 3402; fax: þ1 303 273 3413.E-mail address: tcath@mines.edu (T.Y. Cath).

Journal of Membrane Science 454 (2014) 426–435

at the feed-pore interface on the feed side of the membrane, watervapors then diffuse through the membrane pores, and ultimatelythey condense into the distillate stream at the distillate-poreinterface on the support side of the membrane.

1.1. Factors affecting DCMD process performance

While MD is a unique process that can be utilized for desalina-tion of hypersaline streams, several transport phenomena maylimit water flux through the membrane. These include decrease inpartial vapor pressure, heat and mass transfer resistance across themembrane, and concentration and temperature polarization (orloss in heat transfer) across the membrane [5,6].

Heat and mass transfer dominate the vapor pressure drivingforce in MD. Because several phase changes occur during MD, theheat transfer resistances across the boundary layers of the mem-brane surface are often the rate-limiting step [5]. Although DCMDis considered one of the simpler configurations of MD, theconductive heat transfer across the membrane is greater than inother MD configurations [6,7].

Heat transfer across the MD membrane gives rise to tempera-ture polarization (TP) in which the temperature of the feedsolution at the feed–membrane interface declines and the tem-perature of the distillate at the distillate–membrane interfaceincreases. Mass transfer across the MD membrane gives rise toconcentration polarization (CP) in which the vaporization of waterfrom the feed stream through the MD membrane results in anincreased solute concentration and thus a lower partial vaporpressure of water at the feed–membrane interface. Similar to othermembrane processes, CP can also induce membrane scaling, whichfurther reduces process performance [8,9].

Membrane scaling occurs when inorganic salts precipitate andaccumulate on the membrane surface, thus blocking the pores forvapor to diffuse across the membrane and subsequently loweringwater flux. Scaling of sparingly soluble salts such as CaCO3, CaSO4,and silicates has been identified as a cause of flux decline whenrecovering water from natural streams, including brines fromdesalination processes [10–17]. Two types of membrane scalingcan occur: homogeneous and heterogeneous scaling. Homoge-neous scaling occurs when crystals that form in the bulk solutionprecipitate on the membrane surface, and heterogeneous scalingoccurs when salts crystallize directly on the membrane surface[18–20]. Scaling can alter membrane surface properties (i.e.,hydrophobicity), change membrane pore structure, and ultimatelydecrease process efficiency and potentially lead to wetting of themembrane pores [12,13]. In MD it is essential that the porousmembrane maintains its hydrophobicity to prevent membranewetting, which will allow passage of water in a liquid phasethrough the membrane pores [5]. Pore wetting of the membranehinders water flux, lowers salt rejection, and further impairsmembrane integrity (i.e., loss of hydrophobicity).

1.2. DCMD for concentration of supersaturated solutions in mineralproduction

Replacing traditional concentration methods with MD couldproduce high quality minerals and water, reduce land footprint ofevaporation ponds, and eliminate the required pumping of waterfrom pond to pond in mineral production sites. Recent studieshave shown that MD consumes less energy than traditionalthermal distillation such as multi-stage flash and multi-effectdistillation, and can further concentrate brines from desalinationprocesses such as RO, NF, and ED [10,13,15,16,21–23]. Furthermore,utilization of low-grade heat sources such as industrial heatemissions and solar energy can offset the overall energy consump-tion needed for MD [24–27].

Recent studies have coupled membrane processes with crystal-lizers to concentrate and recover minerals in hypersaline solutions[14,22,28–35]; however, none of these studies have effectivelymitigated membrane scaling. While membrane scaling has beeninvestigated [10–13,23,34,36–40], effective scale mitigation tech-niques for maintaining and restoring water flux and salt rejectionwhen desalinating saturated solutions are still lacking. In thecurrent study, DCMD was applied to concentrate Great Salt Lake(GSL) water. The main objectives of the study were to evaluate theperformance of DCMD in concentrating hypersaline brines fromthe GSL, and in doing so, optimize operating conditions tomaximize water recovery and mitigate membrane scaling. Severalunique methods were developed and tested to identify andmitigate membrane scaling. Finally, the replacement of evapora-tion ponds with DCMD was assessed as a means to intensify themineral production process.

2. Materials and methods

2.1. Membranes

Two hydrophobic microporous membranes were acquired fromGE Water (Minnetonka, MN). The first membrane (TS22) is acomposite membrane consisting of a thin polytetrafluoroethylene(PTFE) active layer and a polypropylene woven support layer. Theoverall thickness of the TS22 membrane is 175 μm, with an activelayer thickness of 5–10 μm. The second membrane (PP22) is anisotropic membrane made of polypropylene (PP), and is approxi-mately 150 μm thick. Both membranes have a nominal pore size of0.22 μm and a porosity of approximately 70% [41]. After experi-ments, the membranes were rinsed with deionized water andstored in a desiccator until analysis. A new membrane coupon wasused for each set of experiments.

2.2. Membrane cells

Experiments were performed with acrylic membrane cells fittedto test flat sheet membranes. The cells were fabricated withsymmetric flow channels on either side of the membrane, allowingfor parallel flow of feed and distillate streams on the opposite sidesof the membrane. Nitrile rubber gaskets were used to form flowchannels, approximately 2 mm deep, on each side of the mem-brane. Turbulent enhancing spacers were placed in the flowchannels to reduce temperature polarization effects, increase waterflux, and ensure that the membrane lay flat and centered in the cell[41]. Experiments were performed using a modified SEPA-CF cellwith an effective membrane surface area of 139 cm2. To preventprecipitation of salts on the membrane surface, the membrane cellwas positioned horizontally with the feed side (active side) facingdown. To observe real-time membrane scaling, an additional set ofexperiments was performed with a stereomicroscope (Stemi 2000,Carl Zeiss Microscope, Thornwood, NY) and a direct observationmembrane cell that has a glass observation port (12.7 cm�2.54 cm)and an effective membrane surface area of 89 cm2. During theseexperiments, the feed side of the membrane was facing up.

2.3. Bench-scale system

Bench-scale experiments were performed to investigate waterflux, salt rejection, and membrane scaling. A supervisory control anddata acquisition (SCADA) system (LabVIEW, National Instruments,Austin, TX; and a LabJack UE-9 Pro, Lakewood, CO) was utilized tocontrol the temperatures of the feed and distillate streams andcollect data to calculate water flux and batch recovery.

K.L. Hickenbottom, T.Y. Cath / Journal of Membrane Science 454 (2014) 426–435 427

A flow schematic of the test unit is illustrated in Fig. 1. Thethermally insulated feed and distillate reservoirs were connected togear pumps (Micropump, Cole Parmer, Vernon Hills, IL) thatcirculate the feed and distillate streams co-currently on the oppo-site sides of the membrane. It is important to note that due to thesmall size of the membrane, operating in co-current mode hadnegligible effects on process performance in this study. Thermo-couples were installed at the inlets of the feed and distillatechannels and connected to the SCADA system. The flow rate ofthe two streams was 1.6 L min�1 when using the modified SEPA-CFcell, and 0.6 L min�1 when using the direct observation cell. As feedwater evaporates through the membrane and absorbed into thedistillate stream, water overflowed from the distillate reservoir intoa beaker positioned on an analytical balance (Model S-8001, DenverInstruments, Bohemia, NY), which was connected to the SCADAsystem. The overflow rate was used to calculate water flux throughthe membrane. The conductivity of the distillate reservoir wascontinuously measured (Waterproof pH/CON 300 m, Oakton Instru-ments, Vernon Hills, IL) and changes in conductivity were used tocalculate salt rejection and detect membrane wetting.

2.4. Preliminary pure water permeability experiments

A set of preliminary experiments was performed to evaluate thepure water permeability of the TS22 and PP22 membranes and toselect experimental operating temperatures. In these experiments,deionized water was used as the feed and distillate streams. Thetemperature of the feed ranged from 30 to 70 1C in increments of10 1C and the temperature of the distillate was 20 or 30 1C.

2.5. Direct contact membrane distillation batch experiments

Experiments were conducted in a batch mode, simultaneouslyconcentrating the feed solution and producing purified water. The feedsolution was representative water from the Bear River Bay (Great SaltLake, UT) on the eastern shore of the GSL and the distillate solutionwas deionized water. The raw GSL water was pre-filtered through a0.5-μm cartridge filter to remove any suspended solids.

2.5.1. Successive water flux and salt rejection experimentsThe performance of DCMD after multiple successive batch

experiments was evaluated. These experiments utilized the PP22and TS22 membranes. A high (40 1C) and low (20 1C) temperaturedifference across the membrane was chosen for these experi-ments. The temperature of the distillate stream was 30 1C and thetemperature of the feed stream was either 50 or 70 1C. Theexperiments were performed until the water flux decreased tonearly 80% of its initial value, then the collected distillate was

returned to the feed reservoir (diluting the feed solution to itsoriginal concentration), and the experiment resumed.

2.5.2. Scale identification experiments2.5.2.1. Optical microscope. The onset of membrane scaling wasobserved with a stereo-microscope and the direct observationmembrane cell. These experiments were performed with the PP22membrane and 3 L of filtered GSL water as the feed. The feed anddistillate temperatures were 50 and 30 1C, respectively. Images ofmembrane scaling were captured throughout the experiments(EOS Rebel XTi, Canon, Lake Success, NY).

2.5.2.2. Isolation of sparingly soluble salts. Experiments werecarried out to evaluate the effect of sparingly soluble salts on thedecline in water flux. To isolate the potential scalants, experimentswere performed with a feed solution of 150 g/L NaCl (ACS grade).The modified SEPA-CF cell, PP22 membrane, and 3 L of feedsolution were used. The feed and distillate temperatures were 50and 30 1C, respectively.

2.5.3. Extended scaling experimentsA set of experiments was performed to evaluate the onset of

membrane scaling over a prolonged period of time. Each experimentwas initiated with 10 L of filtered GSL water recirculating on the feedside, and process performance was evaluated over 125 h. In theseexperiments, both the PP22 and TS22 membranes were tested andthe feed and distillate temperatures were 50 and 30 1C, respectively.Each experiment was divided into 5–6 cycles. In each cycle, 1 L ofdistillate overflow was collected and removed from the system, thenthe distillate overflow (permeate) was continuously returned to thefeed reservoir for approximately 12 h. After the 12 hours, a new cyclestarted and another 1 L of distillate overflow was removed from thesystem. The cycles were repeated until the feed water was concen-trated approximately 2.5 times relative to its original concentration.

2.6. Scale mitigation experiments

Three scaling mitigation techniques were investigated to pre-vent or remove membrane scaling and maintain membraneintegrity. Batch experiments were performed with the modifiedSEPA-CF cell, PP22 membrane, and 8.5 L of filtered GSL feed water.The PP22 membrane was chosen because of its isentropic struc-ture. The temperatures of the feed and distillate were 60 and 30 1C,respectively.

2.6.1. Mitigating rapid flux declineIn the first scale mitigation procedure, DCMD batch experi-

ments were terminated before the water flux started to rapidlydecline. After 12 h of operation, or when 35–40% of the feed waterwas recovered, the collected distillate was returned to the feedreservoir, diluting the feed to its original concentration. Subse-quently, the next batch experiment began. Eight successive batchexperiments were performed without membrane cleaningbetween cycles.

2.6.2. Flow reversalSimilar to the first technique, batch experiments were termi-

nated before scaling occurred on the membrane surface, or after35–40% of the feed water was recovered. Then, the feed side of themembrane was rinsed with approximately 200 mL of deionizedwater, and the feed and distillate channels of the membrane cellwere exchanged: the feed side became the distillate side andthe distillate side became the feed side. Subsequently, the nextsuccessive batch experiment began. This procedure was repeatedsix times.

Fig. 1. Flow schematic of the DCMD bench scale system.

K.L. Hickenbottom, T.Y. Cath / Journal of Membrane Science 454 (2014) 426–435428

2.6.3. Temperature reversalSimilar to the previous scale mitigation techniques, batch

experiments were terminated before scaling occurred on themembrane surface, or after 35–40% of the feed water wasrecovered. Then, a colder GSL feed stream (15 1C) was circulatedon the feed side of the membrane while the warmer distillate(30 1C) continued to be circulated on the distillate side of themembrane, thus reversing the direction of the driving force acrossthe membrane. After 20 minutes, another batch concentrationcycle started with a warmer feed (60 1C at 150,000 mg/L) andcooler distillate (30 1C) on their respective sides of the membrane.This procedure was repeated eight times. It is important to notethat the colder feed stream could be taken directly from a coldersource water; therefore, eliminating feed water cooling. In thiscase, the colder feed stream could be taken directly from the GSL,which has an average temperature of 15 1C [42].

2.7. Solution chemistry and analytical methods

The GSL water was characterized using standard analyticalmethods. Water samples were prepared and analyzed for dis-solved solids according to Standard Methods (APHA, 2005).Samples were diluted and filtered through a 0.45-μm filter andanalyzed for anions with an ion chromatograph (Model ICS-90,Dionex, Sunnyvale, CA) and for cations with an inductively coupledplasma atomic emission spectrometer (ICP-AES) (Optima 5300 DV,PerkinElmer Inc., Waltham, MA). The average total dissolved solids(TDS) concentration of the GSL water from the Bear River Bay wasapproximately 150,000 mg/L, most of which is sodium chloride(�84% w/w). A detailed ionic composition of the GSL water usedin this study is summarized in Table 1. The feed solution chemistrywas simulated with OLI Stream Analyzer™ (OLI Systems, Inc., MorrisPlains, NJ) to determine the scaling tendencies at increasing tem-peratures (50–70 1C) and concentrations (150,000–500,000 mg/L).

2.8. Economic implications of DCMD utilization

The economic implication of using DCMD for concentration of GSLwater was evaluated and compared to the use of evaporation ponds.The evaporative water loss for the Bear River Bay (approximately 1.4 mper year [42]) and the average DCMD water flux obtained from the

temperature reversal experiments were standardized and used forcomparison of the different concentration methods.

3. Results and discussion

3.1. Pure water permeability experiments

Water flux as a function of feed temperature for the twomembranes is shown in Fig. 2. Water flux increased exponentiallywith increasing temperature difference (or vapor pressure drivingforce) across the membrane. The water flux through the TS22 wasconsistently higher than the water flux through the PP22. This isbecause the PP22 is a thicker membrane with more tortuous poresthat increase the resistance to vapor diffusion through the mem-brane pores, thus resulting in a lower permeability [41,43]. Basedon these results, a high temperature differential (ΔT¼40 1C) andlow temperature differential (ΔT¼20 1C) were chosen for thesuccessive batch concentration experiments; the temperature ofthe distillate stream was kept at 30 1C and the temperature of thefeed stream was either 50 or 70 1C.

3.2. Direct contact membrane distillation batch experiments

3.2.1. Successive batch experiments: water flux and salt rejectionWater flux as a function of GSL water total solids concentration

for experiments performed with the two membranes operated at aΔT of 40 and 20 1C is shown in Fig. 3. In all experiments the waterflux gradually declined as the feed solution concentrationincreased, and thus the partial vapor pressure of water in the feedsolution decreased. Thereafter, a sharp decline in water flux wasobserved in all the experiments.

Compared to the pure water permeability experiments, a lowerinitial water flux was observed during batch experiments per-formed with the GSL water. For example, in experiments performedwith the PP22 and GSL feed solution, the initial water flux was 11%(ΔT of 20 1C) and 20% (ΔT of 40 1C) lower than that of the purewater permeability experiments performed with the same mem-brane and the same temperature differences. The experimentsperformed with the TS22 and GSL feed solution resulted in aninitial water flux of 28% (ΔT of 20 1C) and 38% (ΔT of 40 1C) lowerthan in the pure water permeability experiments performed withthe same membrane and temperature differences. The percentdecrease in initial water flux for the experiments performed withthe GSL feed water was lower for temperature difference of 20 1Cand for the PP22. The lower partial vapor pressure of water in thehighly concentrated feed solution (150,000 mg/L total solids) wasthe main reason for a lower driving force across the membrane andthe lower initial water flux. For example, Song et al. reported thatsolutions with salt concentrations of 10% and 20% reduced thepartial vapor pressure by 6% and 20%, respectively [44].

At increased temperature differentials (ΔT of 40 1C), a higherterminal GSL water concentration was achieved with the PP22.Although experiments conducted with the TS22 at a higher tempera-ture difference (ΔT of 40 1C) resulted in greater initial water fluxescompared to those with the PP22, the initial water flux decreasedafter each successive batch experiment. Several studies have reportedthat polarization effects are more severe at increased operatingtemperatures [8,16,23,43,45]; however, these effects are reducedwhen using membranes with low thermal conductivities [7,25,45].Thus, it is likely that the lower thermal conductivity of the thickerPP22 mitigated the effects of temperature polarizationwhen tested atincreased temperature differences across the membrane. Tempera-ture polarization effects were more severe during experimentsperformed with the TS22 at increased temperature differences, whichresulted in increased membrane scaling. Higher flux also results in

Table 1Ionic composition of the GSL water. All values arefor the cartridge filtered GSL water.

Analyte Concentration (mg/L)

B3þ 40.0Ba2þ 1.08Ca2þ 360Fe2þ 0.96Kþ 2800Liþ 24.0Mg2þ 5200Mn2þ 0.04Naþ 43,200Ni2þ 0.21S2�n 270Si4þ 9.16Ti4þ 1.85Zn2þ 1.38Sr2þ 2.97Cl� 81,600Br� 72NO3

� 13SO4

2� 10,300TOC 56.8

n Sulfur not associated with SO42� .

K.L. Hickenbottom, T.Y. Cath / Journal of Membrane Science 454 (2014) 426–435 429

higher heat fluxes, which subsequently decreases the temperaturedifference across the membrane and increases temperature polariza-tion effects [8,9,16,43].

Temperature polarization also affects the mechanisms of mem-brane scaling. Because the temperature is lower at the feed–membrane interface than in the bulk solution, the calcium speciessolubility is expected to increase, whereas NaCl solubility isexpected to decrease. Additionally, increased water flux resultsin increased rates of scaling. Therefore, it is expected that the

potential for scaling of NaCl will be higher for the TS22 membrane(higher flux) than the PP22 membrane.

Water flux and distillate conductivity as a function of time areshown in Fig. 4 from data presented in Fig. 3. The distillateconductivity for both sets of experiments performed at a ΔT of20 1C continuously decreased, indicating that the membranerejected nearly 100% of all inorganic and non-volatile constituents.However, the distillate conductivity increased in both experimentsoperated at a higher temperature gradient. The distillate

0

10

20

30

40

50

60

70

Wat

er fl

ux, L

m-2

hr-1

Feed/Distillate temperature, C

TS22 - 20 C PP22 - 20 C

0

10

20

30

40

50

60

70

30/20 40/20 50/20 60/20 40/30 50/30 60/30 70/30

Wat

er fl

ux, L

m-2

hr-1

Feed/Distillate temperature, C

TS22 - 30 C PP22 - 30 C

Fig. 2. Water flux as a function of feed temperature for experiments performed with the TS22 and PP22 MD membrane. The distillate temperature (Td) was either (a) 20 or(b) 30 1C, the feed temperature was between 30 and 70 1C, and the flow rates were kept constant at 1.6 L min�1.

0

5

10

15

20

25

30

35

40

45

50

Wat

er fl

ux, L

m-2

hr -

1

Total solids concentration, mg/L

50/30 - Run 1 70/30 - Run 1

50/30 - Run 2 70/30 - Run 2

50/30 - Run 3 70/30 - Run 3

0

5

10

15

20

25

30

35

40

45

50

150,000 250,000 350,000 150,000 250,000 350,000

Wat

er fl

ux, L

m-2

hr -

1

Total solids concentration, mg/L

50/30 - Run 1 70/30 - Run 1

50/30 - Run 2 70/30 - Run 2

50/30 - Run 3 70/30 - Run 3

Fig. 3. Water flux as a function of total solids concentration for the successive batch experiments (runs) performed with (a) the TS22 membrane and (b) the PP22 membrane.Experiments were performed with filtered GSL water as feed and deionized water as distillate. The distillate temperature was 30 1C and the feed temperature was either50 (50/30) or 70 1C (70/30). The flow rate was kept constant at 1.6 L min�1.

K.L. Hickenbottom, T.Y. Cath / Journal of Membrane Science 454 (2014) 426–435430

conductivity increased throughout the experiment performedwith the PP22. In experiments performed with the TS22, thedistillate conductivity started to increase only towards the end ofthe experiment, indicating that the TS22 is less susceptible towetting at higher operating temperatures. A previous study byGryta et al. reported that operating at temperatures greater than68 1C may reduce the hydrophobicity of polypropylene mem-branes and lead to membrane wetting [36]. Saffarini et al. reportedthat PTFE membrane support layers showed no signs of degrada-tion when exposed to high temperatures (o350 1C); however, thepotential for wetting of PTFE membranes does increase withincreasing feed temperatures and salinities [46].

When operated at a higher ΔT of 40 1C, the water flux had morethan tripled, increasing the overall water flux from 12.8 to47 L m�2 h�1 during experiments with the TS22 and from 12.3to 40 L m�2 h�1 during experiments with the PP22. Yet, while theprocess was accelerated at high temperature differences across themembrane, both salt rejection and initial water flux decreasedafter each successive batch experiment. When experiments wereconducted at a ΔT of 20 1C, the salt rejection of the membrane washigh, the water flux after each successive batch run was restored,and the GSL water was concentrated to more than 300,000 mg/Ltotal solids. Therefore, operating temperatures lower than 70 1C(50 and 60 1C) for the feed and 30 1C for the distillate were chosenfor the following experiments.

3.2.2. Membrane scaling investigationTo further investigate the onset of rapid flux decline, a set of

experiments was performed in conjunction with stereomicroscopeobservation. Water flux and total solids concentration as a functionof time are shown in Fig. 5. The dashed lines in Fig. 5 indicatewhen pictures of scaling on the membrane surface were taken.Images of the PP22 membrane surface were captured before andafter the onset of rapid flux decline. Similar to results shown inFig. 3, the water flux begins to rapidly decline at a total solidsconcentration of approximately 300,000 mg/L.

Images of the different stages of scale formation on the feedside of the membrane are also shown in Fig. 5. The onset ofmembrane scaling is first visualized after 7.5 h of operating time,or when the feed solution concentration approached 250,000 mg/Ltotal solids. Thereafter, crystals continued to precipitate on themembrane surface and the water flux continued to decline. Afterapproximately 9.5 h of operation, or at a bulk feed solution concen-tration of 300,000 mg/L total solids, the membrane surface wasmostlycovered with salt resembling NaCl crystals [22,47].

To further evaluate the effect of sparingly soluble salts on fluxdecline, results from experiments performed with pure NaCl as thefeed solution were super-imposed on results from experimentsperformed with GSL water (Fig. 5). Interestingly, similar to resultsfrom experiments with GSL water, the same sharp decline in waterflux occurred during the experiments performed with the NaClwater, indicating that the onset of homogeneous precipitation ofsalts, mainly NaCl, correlates to the onset of rapid water flux decline.

Water flux was higher during experiments with NaCl feed thanduring experiments with GSL water. Also, compared to the experi-ments with the GSL water, the rapid decline inwater flux is delayed inthe experiments with NaCl feed. This can be explained by furtherevaluating the complexity of the solution chemistry for the GSL water.Sparingly soluble salts and organic matter are present in the GSL water(Table 1). OLI modeling results revealed that at a bulk GSL water feedsolution temperature of 50 1C, calcium species are the first to reachsaturation (at 340,000 mg/L TDS), followed by NaCl (at 400,000 mg/LTDS). Curcio et al., found that divalent calcium ions in the presence ofhumic acid form complexes with the carboxyl functional groups andcause membrane scaling [17]. The calcium scaling then serves asnucleation sites for other species, such as NaCl [16]. Therefore, scalingof sparingly soluble salts and NaCl were both the cause of rapid waterflux decline during the experiments with GSL feed water.

3.2.3. Extended scaling experimentsWater flux and distillate conductivity as a function of time and

feed total solids concentration are shown in Fig. 6a and b. Long-

0

50

100

150

200

250

0

5

10

15

20

25

30

35

40

45

50

Dis

tilla

te c

ondu

ctiv

ity,

S cm

-1

Wat

er fl

ux, L

m-2

hr -1

Elapsed time, hr

Water flux - 50/30 Water flux - 70/30 Distillate conductivity - 50/30 Distillate conductivity - 70/30

0

50

100

150

200

250

0

5

10

15

20

25

30

35

40

45

50

0 10 20 30 0 10 20 30

Dis

tilla

te c

ondu

ctuv

uty,

S

cm-1

Wat

er fl

ux, L

m-2

hr -

1

Elapsed time, hr

Water flux - 50/30 Water flux - 70/30 Distillate conductivity - 50/30 Disitllate conductivity - 70/30

Fig. 4. Water flux and distillate conductivity as a function of total solids concentration for (a) the TS22 membrane and (b) the PP22 membrane. Successive batch experimentswere performed with filtered GSL feed water and deionized water as the distillate stream. The distillate temperature was kept at 30 1C and the feed temperature was either50 (50/30) or 70 1C (70/30). The flow rate was kept constant at 1.6 L min�1.

K.L. Hickenbottom, T.Y. Cath / Journal of Membrane Science 454 (2014) 426–435 431

term batch experiments were performed in a unique operatingmode to evaluate the effect of membrane scaling over time. Thefirst part of the cycle was performed to evaluate how water fluxdecreases as concentration increases. Similar to results obtainedfor the successive batch experiments (Section 3.2.1), in both sets ofexperiments, the water flux gradually declined until the feedsolution reached approximately 300,000 mg/L total solids. Thesecond part of the cycle (recirculation step) was performed underconstant conditions to evaluate membrane scaling over time andits effects on water flux. During the recirculation step, water fluxcontinued to declined, further indicating that in addition to areducing partial vapor pressure of the feed solution, nucleation of

sparingly soluble salts on the membrane also contribute to thegradual decline in water flux.

3.3. Scaling mitigation techniques

Three unique operating techniques were investigated to mitigatemembrane scaling. These include reduced operating time interval,reduced operating time interval with flow reversal, and reducedoperating time interval with temperature reversal. The PP22 mem-brane was chosen for these experiments because of its isotropicstructure. The operating temperatures were chosen because anaccelerated operating time and increased feed concentration can be

0

2

4

6

8

10

12

14

16

Wat

er fl

ux, L

m-2

hr-1

Total solids concentration, mg/L

GSL

NaCl

(a) (b)(c)(d)

Fig. 5. Water flux and total solids concentration as a function of time for experiment performed with GSL and NaCl feed solutions. Labeled dashed lines correspond tostereomicroscope images of the feed side of the membrane surface during the experiment. Feed and distillate temperatures were 50 and 30 1C, respectively, feed volume was1.5 L, flow velocities were 0.8 L min�1, and the PP22 membrane surface area was 89 cm2.

0

10

20

30

40

50

60

70

80

90

0

2

4

6

8

10

12

14

16

18

Dis

tilla

te c

ondu

ctiv

ity,

S cm

-1

Wat

er fl

ux, L

m-2

hr -

1

Elapsed time, hr

Water flux - TS22 Water flux - PP22 Distillate conductivity - TS22 Distillate conductivity - PP22

0

10

20

30

40

50

60

70

80

90

0

2

4

6

8

10

12

14

16

18

0 50 100 150 150,000 250,000 350,000

Dis

tilla

te c

ondu

ctiv

ity,

S cm

-1

Wat

er fl

ux, L

m-2

hr -

1

Total solids concentration, mg/L

Water flux - TS22 Water flux - PP22 Distillate conductivity - TS22 Distillate conductivity - PP22

Fig. 6. Water flux and distillate conductivity as a function of (a) elapsed time and (b) total solids concentration in the feed for experiments with the PP22 and TS22membranes using cartridge filtered GSL feed water. The temperature for the feed and distillate streams were 50 and 30 1C, respectively, and the stream flow rates were1.6 L min�1.

K.L. Hickenbottom, T.Y. Cath / Journal of Membrane Science 454 (2014) 426–435432

achieved without wetting the membrane and compromising itsperformance.

3.3.1. Mitigating rapid flux declineThe first technique to prevent scale formation during succes-

sive batch experiments was to terminate the experiment before arapid flux decline occurred. Water volume recovered from the feedand distillate conductivity as a function of elapsed time are shownin Fig. 7. Each line represents a successive batch experiment, andthe slope of each line divided by the membrane area (0.0139 m2) isthe average water flux (in L m�2 h�1) during each batch experi-ment (labeled above each line). During the first five successivebatch experiments the water recovery only minimally changed.However, during the 6th batch, scaling on the membrane andwetting of some pores caused a rapid decline in water flux and asharp increase in distillate conductivity. Following these results,two new operating techniques were tested to mitigate and reversescaling of minerals on the membrane.

3.3.2. Flow reversalWater volume recovered (i.e., distillate collected) and distillate

conductivity as a function of elapsed time are shown in Fig. 8 and eachline represents a successive batch experiment. In this scale mitigationtechnique, the feed and distillate flow channels were exchanged aftereach successive batch experiment. The average water flux (inL m�2 h�1) during each batch experiment is labeled above each line.The average water flux during all experiments was 19.5 L m�2 h�1

with a standard deviation of 1.44 L m�2 h�1, indicating that mem-brane scaling was minimal. Following the first experiment, thedistillate conductivity increased. This increase in distillate conductivitywas mostly due to residual salts in the distillate hydraulic loop fromthe previous cycle/experiment and/or dissolution of scalants thatdeposited on the membrane surface and in the membrane pores.The cause for the different trends in distillate conductivity on eitherside of the membrane is not well understood; however, it is likely thatthe slight difference in surface characteristics on the opposite sides ofthe membrane resulted in dissimilar scaling and wetting patterns on

the membrane. Additional research on membrane characteristics,nucleation kinetics, and scale formation could provide further insightto this trend.

3.3.3. Temperature reversalWater volume recovered and distillate conductivity as a

function of elapsed time for the third scale mitigation techniqueare shown in Fig. 9. The average water flux (in L m�2 h�1) duringeach batch experiment is labeled above each line. In thistechnique, the temperature difference across the membranewas reversed for a period of time before a new batch experimentwas performed.

The average water flux during these experiments was20.6 L m�2 h�1 with a standard deviation of 0.95 L m�2 h�1. Thewater flux slightly declined during the sixth experiment and thedistillate conductivity slowly increased. Overall, the water flux andsalt rejection were higher during this operating technique than theprevious scale mitigation techniques, and the use of freshwater toflush the feed channel was eliminated.

Flow and temperature reversal techniques proved to be veryeffective in maintaining water flux and mitigating membranescaling. Compared to previous experiments performed withoutscale mitigation techniques, experiments performed with the flowand temperature reversal techniques resulted in sustained, highwater fluxes throughout batch concentration experiments. Also,these operating techniques were performed without the use ofchemicals (i.e., antiscalants, acids, and bases) to remove scalants,and without additional energy to cool the feed water. Scalemitigation via temperature reversal achieved the greatest averagewater fluxes, rejected nearly 100% of non-volatiles for the first sixbatch concentration cycles, and was performed without additionalwater and energy inputs. Therefore, this scale mitigation techniquecould prove to be very impactful in MD. A recent study by Kesieme

21.7 20.3

7.2

21.3

19.6

4.1

20.7

16.8

0

30

60

90

120

150

180

210

240

0

500

1000

1500

2000

2500

3000

3500

4000

0 20 40 60 80 100

Dis

tilla

te c

ondu

ctiv

ity,

S cm

-1

Wat

er re

cove

red,

mL

Elapsed time, hr

Run 1 Run 4 Run 7 Run 2 Run 5 Run 8 Run 3 Run 6 Dist. cond.

Fig. 7. Water recovered and distillate conductivity as a function of elapsed time forthe successive batch experiments (runs) performed with the PP22 membrane andfiltered GSL feed water. The numbers at the top of each line represent the averagewater flux (L m�2 h�1) for each batch run. The experiments were conducted withfeed and distillate temperatures of 60 and 30 1C, respectively.

19.0 21.2 21.1 19.3 19.3

17.8

0

150

300

450

600

750

900

1050

1200

0

500

1000

1500

2000

2500

3000

3500

4000

0 50 100

Dis

tilla

te c

ondu

ctiv

ity,

S cm

-1

Wat

er re

cove

red,

mL

Elapsed time, hr

S1_1 S2_1 S1_2 S2_2 S1_3 S2_3 Dist. cond.

Fig. 8. Water recovered and distillate conductivity as a function of elapsed time forthe successive batch experiments with alternating feed and distillate channels. Thefeed and distillate channels were alternated three times each. S1 and S2 denote theinitial feed and distillate sides, respectively, whereas 1, 2, and 3 denote the first,second, and third alternations of the feed and distillate sides. The numbers at thetop of each line represent the average water flux (L m�2 h�1) for each batchexperiment. The experiments were performed with the PP22 membrane andfiltered GSL feed water, and were conducted with feed and distillate streamtemperatures of 60 and 30 1C, respectively, at 1.6 L min�1.

K.L. Hickenbottom, T.Y. Cath / Journal of Membrane Science 454 (2014) 426–435 433

et al. [48] reported that addition of a 0.5 μm filter before the feedchannel inlet was effective in capturing precipitated salts thatwould have otherwise deposited on the membrane. Hybridizationof the proposed scale mitigation techniques with a 0.5 μm filtra-tion of the feed could further reduce membrane scaling andwetting in MD.

3.4. Efficiency of MD over natural evaporation

When considering replacement of evaporation ponds withDCMD, two central considerations are the time and costs involvedin concentrating brines. Production of high-value minerals willcontinue to increase with growing demands, and acquisition ofland for additional ponds can be costly or in some cases impos-sible. Therefore, the efficiency of natural evaporation of the BearRiver Bay was compared to MD.

The net annual evaporation rate at the Bear River Bay is1040 mm (41 in.) per year [42]. From a simple unit conversion, onaverage 2.85 mm of water is evaporated from the bay each day.From results obtained in this research, DCMD can concentrate GSLwater at an average rate of 20 L m�2 h�1 (Fig. 9), and from a simpleunit conversion, approximately 480 mm of high-quality distillatewater can be recovered from GSL water each day using a DCMDmembrane of the same area. Therefore, applying DCMD to mineralproduction not only recovers high quality water, but also acceleratesthe natural evaporation process in concentrating hypersaline solu-tions by approximately 170 times.

In terms of land use, one acre (4047 m2) of evaporation pondscould be replaced with approximately 259 ft2 (24 m2) of flat sheetDCMD membrane. Several studies have shown that DCMD is aneconomically and environmentally competitive water treatmentprocess to RO when low-grade heat is utilized [25,48,49]. A studyby Al-Obaidani et al. estimated that operating DCMD with a heatrecovery system could reduce water cost to $0.64 m�3 of waterproduced (�40 kWh/m3), making DCMD a competitive membraneprocess to RO ($0.50 m�3 of water produced) [25]. A more recentstudy estimated that water production costs with DCMD could be

further reduced to $0.57 m�3 when low-grade heat is utilized andcarbon tax is applied [48]. Therefore, in addition to concentratingGSL water for mineral recovery, the high-quality water producedcan be sold to further offset operating costs.

4. Conclusions

DCMD was effective in concentrating GSL water to greater than350,000 mg L�1. Operating DCMD at high ΔT of 40 1C was notsustainable; the membrane performance was compromisedbecause of membrane scaling and pore wetting. Consequently,operating DCMD in successive batch mode without the use of scalemitigation techniques resulted in decreased membrane perfor-mance (i.e., lowered salt rejection and water fluxes).

Flow reversal and temperature reversal are new operatingtechniques that proved very effective in sustaining high waterfluxes and membrane performance [50]. The scale mitigationtechniques were effective in inhibiting homogeneous precipitationof salts and disrupting nucleation of sparingly soluble salts on themembrane surface. Of the three scale mitigation techniques, thetemperature reversal technique was most effective in maintaininghigh water fluxes (420 L m�2 h�1) and high salt rejection. Thenew techniques were simple to operate and very impactful inmitigating scaling. Furthermore, the need for antiscalants andother chemicals used for membrane cleaning was avoided.

Replacing natural evaporation ponds with DCMD can result inenhanced operations and reduced environmental footprints. Oper-ating DCMD with low-grade heat recovered from the on-sitechemical processing plant can drastically reduce MD operatingcosts, and high-quality water recovered from the GSL water canaid in offsetting operating costs.

Acknowledgments

The authors would like to thank Compass Minerals, the NSF-BDprogram, NSF-CBET (Grant #1236846), and the EPA-STAR fellow-ship grant for the support of this research. Special thanks to DanielPannell, Jerry Poe, and Mark Reynolds with Compass Minerals.

References

[1] World Water Assessment Programme, The United Nations World WaterDevelopment report 4: Managing Water under Uncertainty and Risk, UNESCO,Paris, 2012.

[2] A. Perez-Gonzalez, A.M. Urtiaga, R. Ibanez, I. Ortiz, State of the art and reviewon the treatment technologies of water reverse osmosis concentrates, WaterRes. 46 (2012) 267–283.

[3] G. Raluy, L. Serra, J. Uche, Life cycle assessment of MSF, MED and ROdesalination technologies, Energy 31 (2006) 2361–2372.

[4] M. Mulder, Basic Principles of Membrane Technology, Kluver AcademicPublishers, Dordrecht, The Netherlands, 1991.

[5] K.W. Lawson, D.R. Lloyd, Membrane distillation, J. Membr. Sci. 124 (1997)1–25.

[6] M. Khayet, Membranes and theoretical modeling of membrane distillation: areview, Adv. Colloid Interface Sci. 164 (2011) 56–88.

[7] A.M. Alklaibi, N. Loir, Heat and mass transfer resistance analysis of membranedistillation, J. Membr. Sci. 282 (2006) 362–369.

[8] L. Martinez-Diez, M.I. Vazquez-Gonzalez, Temperature and concentrationpolarization in membrane distillation of aqueous salt solutions, J. Membr.Sci. 156 (1999) 265–273.

[9] R.W. Schofield, A.G. Fane, C.J.D. Fell, Heat and mass transfer in membranedisitillation, J. Membr. Sci. 33 (1987) 299–313.

[10] D. Qu, J. Wang, B. Fan, Z. Luan, D. Hou, Study on concentrating primary reverseosmosis retentate by direct contact membrane distillation, Desalination 247(2009) 540–550.

[11] M. Gryta, Long-term performance of membrane distillation process, J. Membr.Sci. 265 (2005) 153–159.

[12] M. Gryta, Fouling in direct contact membrane distillation, J. Membr. Sci. 325(2008) 383–394.

21.7 21.0 19.6

21.8 20.4

21.4 20.0

19.3

0

25

50

75

100

125

150

175

200

0

500

1000

1500

2000

2500

3000

3500

4000

0 20 40 60 80 100

Dis

tilla

te c

ondu

ctiv

ity,

S cm

-1

Wat

er re

cove

red,

mL

Elapsed time, hr

Run 1 Run 4 Run 7 Run 2 Run 5 Run 8 Run 3 Run 6 Dist. cond.

Fig. 9. Water recovered and distillate conductivity as a function of elapsed time forthe successive batch experiments (runs) performed with the temperature reversaltechnique. The numbers at the top of each line represent the average water flux(L m�2 h�1) for each batch run. The experiments were performed with the PP22membrane and filtered GSL feed water, and were conducted with feed and distillatestream temperatures of 60 and 30 1C, respectively, at 1.6 L min�1.

K.L. Hickenbottom, T.Y. Cath / Journal of Membrane Science 454 (2014) 426–435434

[13] F. He, K.K. Sirkar, J. Gilron, Studies on scaling of membranes in desalination bydirect contact membrane distillation: CaCO3 and mixed CaCO3/CaSO4 systems,Chem. Eng. Sci. 64 (2009) 1844–1859.

[14] E. Curcio, A. Criscouli, E. Drioli, Membrane Crystallizers, Ind. Eng. Chem. Res.40 (2001) 2679–2684.

[15] E. Drioli, E. Curcio, A. Criscuoli, G.D. Profio, Integrated system for recovery ofCaCO3, NaCl and MgSO4 �7H2O from nanofiltration retentate, J. Membr. Sci.239 (2004) 27–38.

[16] J.-P. Mericq, S. Laborie, C. Cabassud, Vacuum membrane distillation of sea-water reverse osmosis brines, Water Res. 44 (2010) 5260–5273.

[17] E. Curcio, X. Ji, G. Di Profio, S. Al Obaidani, E. Fontananova, E. Drioli, Membranedistillation operated at high seawater concentration factors: role of themembrane on CaCO3 scaling in presence of humic acid, J. Membr. Sci. 346(2010) 263–269.

[18] S. Lee, C.-H. Lee, Effect of operating conditions on CaSO4 scale formationmechanism in nanofiltration for water softening, Water Res. 34 (2000)3854–3866.

[19] S. Shirazi, C.-J. Lin, S. Doshi, S. Agarwal, P. Rao, Comparison of foulingmechanism by CaSO4 and CaHPO4 on nanofiltration membranes, Sep. Sci.Technol. 41 (2007) 2861–2882.

[20] B. Mi, M. Elimelech, Gypsum scaling and cleaning in forward osmosis:measurements and mechanisms, Environ. Sci. Technol. 44 (2009) 2022–2028.

[21] J.H. Hanemaaijer, J. van Medevoort, A.E. Jansen, C. Dotremont, E. van Sonsbeek,T. Yuan, L. De Ryck, Memstill membrane distillation – a future desalinationtechnology, Desalination 199 (2006) 175–176.

[22] X. Ji, E. Curcio, S. Al Obaidani, G. Di Profio, E. Fontananova, E. Drioli, Membranedistillation-crystallization of seawater reverse osmosis brines, Sep. Purif.Technol. 71 (2010) 76–82.

[23] C.R. Martinetti, A.E. Childress, T.Y. Cath, High recovery of concentrated RObrines using forward osmosis and membrane distillation, J. Membr. Sci. 331(2009) 31–39.

[24] F. Suarez, S.W. Tyler, A.E. Childress, A theoretical study of a direct contactmembrane distillation system coupled to a salt-gradient solar pond forterminal lakes reclamation, Water Res. 44 (2010) 4601–4615.

[25] S. Al-Obaidani, E. Curcio, F. Macedonio, G. Di Profio, H. Al-Hinai, E. Drioli,Potential of membrane distillation in seawater desalination: thermal effi-ciency sensitivity study and cost estimation, J. Membr. Sci. 323 (2008) 85–98.

[26] G. Zuo, R. Wang, R. Field, A.G. Fane, Energy efficiency evaluation and economicanalyses of direct contact membrane distillation system using Aspen Plus,Desalination 283 (2011) 237–244.

[27] A.M. Alklaibi, The potential of membrane distillation as a stand-alonedesalination process, Desalination 223 (2008) 375–385.

[28] G.D. Profio, E. Curcio, E. Drioli, Supersaturation control and heterogeneousnucleation in membrane crystallizers: facts and perspectives, Ind. Eng. Chem.Res. 49 (2010) 11878–11889.

[29] J.T.M. Sluys, D. Verdoes, J.H. Hanemaaijer, Water treatment in a membrane-assisted crystallizer (MAC), Desalination 104 (1996) 135–139.

[30] C.Y. Tai, Crystal growth kinetics of two-step growth process in liquid fluidized-bed crystallizers, J. Cryst. Growth 206 (1999) 109–118.

[31] E. Curcio, X. Ji, A.M. Quazi, S. Barghi, G. Di Profio, E. Fontananova, T. Macleod,E. Drioli, Hybrid nanofiltration-membrane crystallization system for thetreatment of sulfate wastes, J. Membr. Sci. (2012)493–498.

[32] E. Drioli, G.D. Profio, E. Curcio, Progress in membrane crystallization, Curr.Opin. Chem. Eng. 1 (2012) 178–182.

[33] M. Gryta, Concentration of NaCl solution by membrane distillation integratedwith crystallization, Sep. Purif. Technol. 37 (2002) 3535–3558.

[34] M. Gryta, Direct contact membrane distillation with crystallization applied toNaCl solutions, Chem. Pap. 56 (2001) 14–19.

[35] F. Macedonio, L. Katzir, N. Geisma, S. Simone, E. Drioli, J. Gilron, Wind-AidedIntensified eVaporation (WAIV) and membrane crystallizer (MCr) integratedbrackish water desalination process: advantages and drawbacks, Desalination273 (2011) 127–139.

[36] M. Gryta, J. Grzechulska-Damszel, A. Markowska, K. Karakulski, The influenceof polypropylene degradation on the membrane wettability during membranedistillation, J. Membr. Sci. 326 (2009) 493–502.

[37] F. He, J. Gilron, H. Lee, L. Song, K.K. Sirkar, Potential for scaling by sparinglysoluble salts in crossflow DCMD, J. Membr. Sci. 311 (2008) 68–80.

[38] F. He, K.K. Sirkar, J. Gilron, Effects of antiscalants to mitigate membrane scalingby direct contact membrane distillation, J. Membr. Sci. 345 (2009).

[39] K. He, H.J. Hwang, M.W. Woo, I.S. Moon, Production of drinking water fromsaline water by direct contact membrane distillation (DCMD), J. Ind. Eng.Chem. 17 (2011) 41–48.

[40] L.D. Nghiem, T.Y. Cath, A scaling mitigation approach during direct contactmembrane distillation, Sep. Purif. Technol. 80 (2011) 315–322.

[41] T.Y. Cath, V.D. Adams, A.E. Childress, Experimental study of desalination usingdirect contact membrane distillation: a new approach to flux enhancement,J. Membr. Sci. 228 (2004).

[42] J.W. Gwynn, P.A. Sturm, Great Salt Lake: a scientific, historical, and economicoverview, Utah Geol. Miner. Surv., 1980.

[43] R.W. Schofield, A.G. Fane, C.J.D. Fell, R. Macoun, Factors affecting flux inmembrane distillation, Desalination 77 (1990) 279–294.

[44] L. Song, B. Li, K.K. Sirkar, J.L. Gilron, Direct contact membrane distillation-based desalination: novel membranes, devices, larger-scale studies and amodel, Ind. Eng. Chem. Res. 46 (2007) 2307–2323.

[45] F. Lagana, G. Barbieri, E. Drioli, Direct contact membrane distillation: model-ling and concentration experiments, J. Membr. Sci. 166 (2000) 1–11.

[46] R.B. Saffarini, B. Mansoor, R. Thomas, H.A. Arafat, Effect of temperature-dependent microstructure evolution on pore wetting in PTFE membranesunder membrane distillation conditions, J. Membr. Sci. 429 (2013) 282–294.

[47] C.M. Tun, A.G. Fane, J.T. Matheickal, R. Sheikholeslami, Membrane distillationcrystallization of concentrated salts – flux and crystal formation, J. Membr. Sci.257 (2005) 144–155.

[48] U.K. Kesieme, N. Milne, H. Aral, C.Y. Cheng, M. Duke, Economic analysis ofdesalination technologies in the context of carbon pricing, and opportunitiesfor membrane distillation, Desalination 323 (2013) 66–74.

[49] K. Tarnacki, M. Meneses, T. Melin, J.v. Medevoort, A. Jansen, Environmentalassessment of desalination processes: reverse osmosis and Memstills,Desalination 296 (2012) 69–80.

[50] K.L. Hickenbottom, T.Y. Cath, D.K. Pannell, J. Poe, Methods for sustainablemembrane distillation concentration of hypersaline streams, US 20130327711A1, Filing date: June 11, 2013.

K.L. Hickenbottom, T.Y. Cath / Journal of Membrane Science 454 (2014) 426–435 435