Separation and Purification Technology 120 (2013) 373–377

Contents lists available at ScienceDirect

Separation and Purification Technology

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

Enhanced desalination using carboxylated carbon nanotube immobilizedmembranes

1383-5866/$ - see front matter � 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.seppur.2013.10.020

⇑ Corresponding author. Tel.: +1 (973) 5965611.E-mail address: mitra@njit.edu (S. Mitra).

Madhuleena Bhadra, Sagar Roy, Somenath Mitra ⇑Department of Chemistry and Environmental Science, New Jersey Institute of Technology, Newark, NJ 07102, USA

a r t i c l e i n f o

Article history:Received 24 April 2013Received in revised form 29 August 2013Accepted 9 October 2013Available online 21 October 2013

Keywords:DesalinationCarbon nanotubesMembranesMembrane distillation

a b s t r a c t

In carbon nanotube immobilized membrane (CNIM), the nanotubes serve as a sorbent that providesadditional pathways for solute transport. In this paper we present that carboxylated nanotubes whichare significantly more polar and can increase interactions with the water vapor in CNIM to improvedesalination efficiency in membrane distillation (MD). The encapsulation of the nanotubes in PVDFprevented the carboxylated nanotubes from making the overall membrane more hydrophilic and thusretain its performance. Overall, desalination was consistently better with carboxylated nanotubes thanwith unfunctionalized ones with flux reaching as high as 19.2 kg/m2 h in a sweep gas membranedistillation mode.

� 2013 Elsevier B.V. All rights reserved.

1. Introduction from forward osmosis [18] to nanofiltration [19,20]. The CNTs have

As the shortage of clean water looms in the horizon, there ismuch interest in developing novel, cost effective desalination tech-nology. Current methodologies include thermal, chemical and re-verse osmosis [1–6]. Membrane distillation (MD) has emerged asan alternative to address some issues related to the current tech-nologies [7–9]. Here a hot salt solution such as sea or brackishwater is passed through (or across) a hydrophobic membranewhich acts as a physical barrier separating the warm solution froma cooler permeate. The permeation is driven by a vapor pressuregradient resulting from the temperature difference and solutioncomposition gradients across the membrane. Typically, MD is car-ried out at 60–90 �C, which is significantly lower than conventionaldistillation. Therefore, it has the potential to generate high qualitydrinking water using only low temperature heat sources such aswaste heat from industrial processes and solar energy.

The main effort in optimal design involves the maximization of sol-ute rejection and flux, which would make MD commercially viable. Akey component in such a process is the membrane itself because itdetermines both flux and selectivity. Several membrane materialbased on polypropylene, Polyvinylidene fluoride (PVDF) and Teflonhave been used in MD [10,11]. Some recent developments include sur-faces made of zeolite [12], clay nanocomposites nanofiber [13], silanegrafting [14] and modification by hydrophobic porous alumina [15].

Carbon nanotube based membranes have been used in a varietyof separation applications in various formats [16,17], that range

been incorporated in the membranes by making polymer compos-ites [19] and via chemical vapor deposition [21]. Recently we havedemonstrated that immobilizing CNTs in different types of prefab-ricated membranes alter the solute-membrane interactions, whichis one of the major physicochemical factors affecting the perme-ability and selectivity. Referred to as carbon nanotube immobilizedmembrane (CNIM), here the CNTs serve as a sorbent and providean additional pathway for solute transport. These membranes havebeen used in nanofiltration, MD, solvent extraction and pervapora-tion, and have demonstrated superior performance [19,22,23].

An important consideration that is yet to be fully utilized inmembrane separation is that CNTs can be effectively functional-ized to alter its chemical properties, which could lead to specificinteractions with solutes, or just a change in hydrophilicity. Thishas been demonstrated in nanofiltration applications by our group[19]. In our previous efforts in MD using CNIM with plain CNTs, weattributed enhanced flux to the sorption of water vapor on thenanotube surface [22]. The objective of this research is to study iffunctionalization, in particular carboxylation will increase theinteraction of nanotubes with the polar water vapor and result inimproved desalination efficiency.

2. Experimental

2.1. Materials and methods

The membrane modules for MD were constructed in a shell andtube format using 1/4 in. polypropylene tubing. Ten, 16.6 cm longhollow fiber strands were used in the module. Each module

374 M. Bhadra et al. / Separation and Purification Technology 120 (2013) 373–377

contained approximately 12.50 cm2 of effective membrane contactarea (based on internal surface). The ends were then sealed withepoxy to prevent leakage into the shell side. Vacuum was appliedto one drain port to draw dry air through the other port, which cre-ated a higher pressure differential and provided a sweep air.

The synthesis of carboxylated CNTs (MWCNT–COOH) was car-ried out as follows. Pristine MWCNT was purchased from CheapTubes, Inc., Brattleboro, VT, USA. As previously reported by ourgroup [24], CNT carboxylation was carried out in a MicrowaveAccelerated Reaction system (CEM Mars) fitted with internal tem-perature and pressure controls. Three hundred milligram of origi-nal MWCNTs was added to the reaction chamber together with25 ml 1:1 conc. H2SO4 and HNO3. The reaction was carried out at120 �C for 40 min. After cooling, the product was vacuum filteredusing a Teflon membrane with pore size (0.45 lm), and the solidwas dried in a vacuum oven at 70 �C for 5 h. This led to the forma-tion of carboxylated MWCNTs (MWCNT–COOH) which was charac-terized by FTIR which confirmed the presence of carboxyl groups.The results are not presented here for brevity.

The CNIM with pure CNT (referred to as CNIM) and functional-ized CNT (referred to as CNIM-f) were prepared using Celgard typeX30-240 (Celgard, LLC, and Charlotte, NC, USA) hollow fiber withpore size (0.04 lm) as the starting material. For the preparationof CNIM and CNIM-f, each of 10 mg of MWCNT and MWCNT–COOHwere dispersed in a solution containing 0.1 mg of polyvinylidenefluoride in 15 ml of acetone by sonicating for 3 h. The Polyvinyli-dene fluoride (PVDF)-nanotube dispersion was forced under con-trolled vacuum into the bore of the polypropylene hollow fibermembrane. The PVDF served as glue that held the CNTs in placeand led to its encapsulation within the membrane, and this mayalso affect the membrane performance. The membrane wasflushed with acetone to remove excess nanotubes. The originalpolypropylene membrane was sonicated in PVDF solution in ace-tone without the CNTs, and this served as the control. The mor-phology of CNIM and CNIM-f were studied using scanningelectron microscopy (SEM, Model LEO 1530), and Thermogravi-metric analysis (TGA) was performed using a Perkin Elmer Pyris7 TGA instrument to study the thermal stability of the membrane.Differential scanning calorimetry (DSC) was also carried out usinga Universal V4.5A TA instrument to observe the alterations in ther-mal properties.

The schematic of experimental system is shown in Fig. 1. Thefeed used in these experiments contained 3.4 wt% NaCl solutions(Sigma Aldrich). This was pumped through the module using aMaster flex 7519-10 peristaltic pump. The preheated hot feed solu-tion travelled through a heat exchanger which was used to main-tain the desired temperature throughout the experiment. Dry airwas passed into the shell side and the permeate was collected ina trap. Air flow was maintained at 1 l min�1. The ionic strengthof the original solution, the permeate and the concentrate weremeasured using a Jenway Electrode Conductivity Meter 4310. Eachexperiment was repeated 3 times to check the reproducibility andrelative standard deviation was less than 1%.

Fig. 1. Schematic diagram of the experimental system

3. Results and discussions

3.1. Characterization of the prepared membranes

Scanning electron micrographs of the original membrane andCNIM-f are shown in Fig. 2a and b. The incorporation of the carbox-ylated CNTs is clearly evident in Fig. 2b. Additionally, Fig. 2c de-picts the intactness of CNT–COOH within the membranes after90 days of continuous usage. The TGA curve is shown in Fig. 3a.As observed, the thermal degradation of unmodified polypropylenemembrane started at around 260 �C. However, in line with previ-ous observations, the presence of CNT–COOH increased the degra-dation temperature by 40 �C. This implies that the CNT–COOH washighly stable and enhanced the thermal stability of the membrane.This is an important factor for MD, where the elevated tempera-tures can be used for desalination [22]. This data was also sup-ported by differential scanning calorimetry (DSC) and ispresented in Fig. 3b. No specific degradation or alterations wasobserved in the CNIM-f.

3.2. Desalination using CNIM-f

The CNIM and CNIM-f were tested for MD. The water vapor flux,Jw, across the membrane can be expressed as:

Jw ¼wp

t � A ð1Þ

where wp is the total mass of permeate, t is the permeate collectiontime and A is the membrane surface area. Also, Jw can be denoted as:

Jw ¼ kðCf � CpÞ ð2Þ

where k is the mass transfer coefficient, Cf and CP is the water vaporconcentration in feed and permeate side. Usually Cp is close to zero,since we utilize dry air as sweep gas. So overall mass transfer coef-ficient was calculated as:

k ¼ Jw

Cfð3Þ

As can be observed in Fig. 4a, increasing temperature increasedflux for all three-membrane types. Flux at 70 �C using theeither CNIM or CNIM-f was higher than what was obtainedby the unmodified membrane. Maximum flux reached up to19.2 kg/m2 h for CNIIM-f membrane and 15.6 kg/m2 h using CNIM.The enhanced performance on CNIM as compared to unmodifiedmembrane has already been studied previously in our group[22]. However, what is unique here is that, the permeate fluxwas the highest for CNIM-f membrane.

Desalination as a function of flow rate is shown in Fig. 4b. It canbe observed that increasing flow rate increased permeate flux. Asobserved, compared to unmodified membrane, CNIM and CNIM-fdemonstrated higher flux at all flow rates. At elevated flow rate,there was reduced boundary layer and adsorption–desorption pro-cesses were faster.

Fig. 4c depicts the effect of varying of feed concentration on per-meate flux. It is well known that concentration polarization is moreimportant at higher feed concentration. At higher feed concentra-tion, a more significant boundary layer develops next to the mem-brane interface and this reduces driving force of mass transfer. Thisleads to the decrease in permeate flux in case of unmodified mem-brane modules. On the other hand, in case of CNIM, and CNIM-f,the flux remained unchanged. The presence of CNTs increasedthe surface roughness that prevented the formation of stableboundary layers. As observed from Fig. 4c, for CNIM-F, the flux re-mained constant with increasing salt concentration, reaching up to19.2 kg/m2 h.

Fig. 2. (a) SEM images of the original membrane, (b) CNIM-f, and (c) CNIM-f after 90 days of operation.

Fig. 3. Thermo gravimetric analysis of unmodified membrane, CNIM, CNIM-f; (b)Differential Scanning Colorimetry of unmodified membrane, CNIM, CNIM-f.

M. Bhadra et al. / Separation and Purification Technology 120 (2013) 373–377 375

Incorporation of various weights of CNT loadings/cm2 of mem-brane area was also investigated. An optimum value of 0.005 mgper centimeter square loadings of MWCNT was required to en-hance the overall percent removal and flux. A further increase ofCNT loading (0.008 mg per centimeter square) did not showedany further enhancement. It was estimated that significantly high-er MWCNT amount would block the pores of the hydrophobicmembrane, thereby reducing flux and removal efficiency.

Additionally, as observed from Table 1, the mass transfer coef-ficients enhancements were found to be significantly higher forCNIM-f as compared to the unmodified membrane. Enhancementfor CNIM ranged between 50% and 77%. However, for CNIM-f,enhancement ranged from 95% to 116%. Table 2 indicates the effectof feed flow rate on mass transfer coefficients. As observed, theoverall mass transfer coefficient was enhanced by presence ofCNIM-f. Interestingly, the enhancement in mass transfer coefficientwas higher at a low flow rate. At a flow rate of 10 ml/min, the masstransfer coefficient of the CNIM-f was 145% higher than theunmodified membrane, whereas for CNIM enhancement was just56% but the corresponding values dropped to 27% and 59% wheninlet feed flow rate was 24 ml/min. In general, the presence ofthe CNT–COOH led to enhanced permeability through the mem-brane, and the CNIM-f showed a significantly higher overall masstransfer coefficient. An important observation from Fig. 5 is that,whereas an increase in feed concentration decreased k for theunmodified membrane, but remained almost constant and showednegligible decease for CNIM and CNIM-f. At 34,000 mg L�1, themass transfer coefficient was more than double for CNIM-f thanthe plain membrane, which was significantly higher than whatwas previously as reported [22].

3.3. Salt breakthrough and stability of the CNIM and CNIM-f

There was no observable salt breakthrough in any of theexperiments, and the permeate showed low conductivity of 1–2.5 ls/cm at 20 �C, implying that the water had over 99.9% purity.

unmodified

CNIM -f

CNIM

unmodified

CNIM -f

CNIM

(a)

(b)

(c)

Fig. 4. (a) Effect of temperature on permeate flux at a feed flow rate of 20 ml min�1;(b) effect of flow rate on permeate flux at 90 �C and (c) effect of feed concentrationon permeate flux at a feed flow rate of 20 ml min�1.

Table 1Mass transfer coefficient and enhancement% at various feed temperature at 20 ml/min.

Temp(�C) Mass transfer coefficient � 107 (kg/m2 s Pa) Enhancement (%)

Unmodified CNIM CNIM-f CNIM CNIM-f

70 0.499 0.856 1.07 72 11480 0.469 0.704 0.915 50 9590 0.349 0.618 0.753 77 116

Table 2Mass transfer coefficient and enhancement% at various feed flow rate at 90 �C.

Flow rate (ml/min) Mass transfer coefficient � 107 (kg/m2 s Pa)

Enhancement(%)

Unmodified CNIM CNIM-f CNIM CNIM-f

10 0.285 0.444 0.697 56 14520 0.349 0.618 0.753 77 11624 0.5 0.634 0.793 27 59

0.3

0.5

0.7

4000 14000 24000 34000Feed concentration (ppm)

Mas

s tr

ansf

er

coef

ficie

ntX1

07 (k

g/m

2 .s.P

a)

unmodified

CNIM

CNIM-f

Fig. 5. Effect of feed concentration on mass transfer coefficient at a feed flow rate of20 ml min�1, 90 �C.

Fig. 6. Operational period stability study of CNIM and CNIM-f membrane.

Feed solution

Membrane

Enhanced adsorption and fast transport by polar-

polar interaction by COOH on CNT surface

Activated diffusion via adsorption desorption on

CNT surface

Direct permeation

through membrane pores

Enhanced Hydrophobic

effect by PVDF

Sweep air

Water Vapor molecule

Water molecule

CNT with PVDF surface

cover

Fig. 7. Mechanism of action on CNIM-f.

376 M. Bhadra et al. / Separation and Purification Technology 120 (2013) 373–377

The stability of the membrane, especially the ability to retain theCNT coating on the surface was tested for long-term operation. Atest was carried out for 90 days and there was no observable

decrease in flux over this period of time using either CNIM orCNIM-f. This is shown in Fig. 6. The SEM images of CNIM-f after90 days of operation also did not show any visible signs of CNTerosion or damage.

M. Bhadra et al. / Separation and Purification Technology 120 (2013) 373–377 377

4. Mechanism

Due to a combination of factors mentioned above, significantlyhigher flux was observed for CNIM and CNIM-f as compared toconventional membrane. This was attributed to the fact that theCNTs serves as sorbent sites for vapor transport while rejectingthe liquid water [22]. The carboxylated CNTs are polar and theyprovided higher sorption for the water vapors than unfunctional-ized CNTs, thus enhancing flux (Fig. 7). Under normal circum-stances one would expect the hydrophilic CNT–COOH to decreasethe overall hydrophobicity of the membrane and also interact withthe sodium ions. Therefore, one would expect the performance ofCNIM-f to be lower than CNIM. However, since PVDF dispersionwas used to immobilize the CNT–COOH, the former encapsulatedthe latter, which prevented water as well as Na+ ions from reachingthe nanotubes. On the other hand, the water vapors that perme-ated through the PVDF surface was able to partition on the CNT-fand effectively permeate through the membrane.

5. Conclusions

Carboxylated CNTs were incorporated into CNIM to enhancepure water flux in membrane distillation. With the incorporationof CNTs, the desalination performance was consistently higherthan the conventional membrane. The carboxylated CNTs showedhigher performance than their unfunctionalized analogs. The per-meate flux achieved up to a maximum of 19.2 kg/m2 h and saltreduction higher than 99% in all cases. These results indicate thatthe incorporation of carboxylated CNTs favorably altered thewater-membrane interactions to enhance vapor permeabilitywhile preventing liquid penetration into the membrane pores.The membranes were stable over long periods of operation withoutany salt leakage.

Acknowledgement

The authors wish to acknowledge the support of Mr. AnthonyMancusi from Membrana Charlotte for their contribution of X30-240 membrane material.

References

[1] A.M. Delgado-Torres, L.G. Rodriguez, Status of solar thermal-driven reverseosmosis desalination, Desalination 216 (2007) 242–251.

[2] A.D. Khawaji, I.K. Kutubkhana, J.M. Wie, Advances in seawater desalinationtechnologies, Desalination 221 (2008) 47–69.

[3] B. Peñate, L.G. Rodríguez, Current trends and future prospects in the designof seawater reverse osmosis desalination technology, Desalination 284(2012) 1–8.

[4] N. Misdan, W.J. Lau, A.F. Ismail, Seawater reverse osmosis (SWRO) desalinationby thin-film composite membrane-Current development, challenges andfuture prospects, Desalination 287 (2012) (2012) 228–237.

[5] M. Li, Optimal plant operations of brackish water reverse osmosis (BWRO)desalination, Desalination 293 (2012) 61–68.

[6] S.S. Sablani, M.F.A. Goosen, R. Al-Belushi, M. Wilf, Concentration polarization inultrafiltration and reverse osmosis: a critical review, Desalination 141 (2001)269–289.

[7] P. Peng, A.G. Fane, X. Li, Desalination by membrane distillation adopting ahydrophilic membrane, Desalination 173 (2005) 45–54.

[8] B.R. Babu, N.K. Rastogi, K.S. Raghavarao, Concentration and temperaturepolarization effects during osmotic membrane distillation, J. Membr. Sci. 322(2008) 146–153.

[9] College of engineering, University of Texas at El Paso, Desalination WaterPurification Research and Development Program, Report No. 81, 2004.

[10] M.M. Teoh, T.S. Chung, Y.S. Yeo, Dual-layer PVDF/PTFE composite hollow fiberswith a thin macrovoid-free selective layer for water production via membranedistillation, Chem. Energy J. 171 (2011) 684–691.

[11] D. Hou, J. Wang, X. Sun, Z. Ji, Z. Luan, Preparation and properties of PVDFcomposite hollow fiber membranes for desalination through directcontact membrane distillation, J. Membr. Sci. 405 (2012) 185–200.

[12] C.H. Cho, K.Y. Oh, S.K. Kim, S.K. Yeo, P. Sharma, Pervaporative seawaterdesalination using NaA zeolite membrane: mechanisms of high water flux andhigh salt rejection, J. Membr. Sci. 371 (2011) (2011) 226–238.

[13] J.A. Prince, G. Singh, D. Rana, T. Matsuura, V. Anbharasi, T.S.Shanmugasundaram, Preparation and characterization of highly hydrophobicpoly(vinylidene fluoride) – Clay nanocomposite nanofiber membranes (PVDF–clay NNMs) for desalination using direct contact membrane distillation, J.Membr. Sci. 397 (2012) 80–86.

[14] S. Khemakhem, R.B. Amar, Modification of tunisian clay membrane surface bysilane grafting: application for desalination with air gap membrane distillationprocess, Colloids. Surf. A: Physicochem. Eng. Aspects 387 (2011) 79–85.

[15] H. Fang, J.F. Gao, H.T. Wang, C.S. Chen, Hydrophobic porous alumina hollowfiber for water desalination via membrane distillation process, J. Membr. Sci.403 (2012) 41–46.

[16] Y.T. Ong, A.L. Ahmad, S.H.S. Zein, K. Sudesh, S.H. Tan, Poly(3-hydroxybutyrate)-functionalised multi-walled carbon nanotubes/chitosan greennanocomposite membranes and their application in Pervaporation, Sep. Purif.Technol. 76 (2011) 419–427.

[17] F. Peng, C. Hu, Z. Jiang, Novel poly (vinyl alcohol)/carbonnanotube hybrid membranes for pervaporation separation of benzene/cyclohexane mixtures, J. Membr. Sci. 297 (2007) 236–242.

[18] Y.X. Jia, H.L. Li, M. Wang, L.Y. Wu, Y.D. Hu, Carbon nanotube: possiblecandidate for forward osmosis, Sep. Purif. Technol. 75 (2010) 55–60.

[19] S. Roy, S.A. Ntim, S. Mitra, K.K. Sirkar, Facile fabrication ofsuperior nanofiltration membranes from interfacially polymerized CNT-polymer composites, J. Membr. Sci. 375 (2011) 81–87.

[20] V. Vatanpour, S.S. Madaeni, R. Moradian, S. Zinadini, B. Astinchap, Fabricationand characterization of novel antifouling nanofiltration membrane preparedfrom oxidized multiwalled carbon nanotube/polyethersulfonenanocomposites, J. Membr. Sci. 375 (2011) 284–294.

[21] D. Manikandan, R.V. Mangalaraja, R. Siddheswaran, R.E. Avila, S.Ananthakumar, Fabrication of nanostructured clay–carbon nanotube hybridnanofiller by chemical vapour deposition, Appl. Surf. Sci. 258 (2012) 4460–4466.

[22] K. Gethard, O.S. Khow, S. Mitra, Water desalination using carbon-nanotube-enhanced membrane distillation, Appl. Mater. 3 (2011) 110–114.

[23] O.S. Khow, S. Mitra, Carbon nanotube immobilized composite hollow fibermembranes for pervaporative removal of volatile organics from water, J. Phys.Chem. C 114 (2010) 16351–16356.

[24] M. Bhadra, O.S. Khow, S. Mitra, Effect of carbon nanotube functionalization inmicro-solid-phase extraction (l-SPE) integrated into the needle of a syringe,Anal. Bioanal. Chem. 402 (2011) 1029–1039.