lable at ScienceDirect

Journal of Power Sources 326 (2016) 466e475

Contents lists avai

Journal of Power Sources

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

Preparations of an inorganic-framework proton exchangenanochannel membrane

X.H. Yan 1, H.R. Jiang 1, G. Zhao, L. Zeng, T.S. Zhao*

Department of Mechanical and Aerospace Engineering, The Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong, China

h i g h l i g h t s

* Corresponding author.E-mail address: metzhao@ust.hk (T.S. Zhao).

1 These authors contributed equally to this work.

http://dx.doi.org/10.1016/j.jpowsour.2016.07.0220378-7753/© 2016 Elsevier B.V. All rights reserved.

g r a p h i c a l a b s t r a c t

� A PEM composed of straight andaligned proton conducting nano-channels is prepared.

� The proton conductivity of thenanochannel membrane reaches11.3 mS cm�1 at 70 �C.

� The activation energy for protontransfer is 0.06 eV assisted with wa-ter molecules.

a r t i c l e i n f o

Article history:Received 1 April 2016Received in revised form17 June 2016Accepted 7 July 2016

Keywords:Proton exchange membraneInorganic materialProton conductivityFirst-principles study

a b s t r a c t

In this work, a proton exchange membrane composed of straight and aligned proton conductingnanochannels is developed. Preparation of the membrane involves the surface sol-gel method assistedwith a through-hole anodic aluminum oxide (AAO) template to form the framework of the PEM nano-channels. A monomolecular layer (SO3He(CH2)3eSie(OCH3)3) is subsequently added onto the innersurfaces of the nanochannels to shape a proton-conducting pathway. Straight nanochannels exhibit longrange order morphology, contributing to a substantial improvement in the proton mobility and subse-quently proton conductivity. In addition, the nanochannel size can be altered by changing the surface sol-gel condition, allowing control of the active species/charge carrier selectivity via pore size exclusion. Theproton conductivity of the nanochannel membrane is reported as high as 11.3 mS cm�1 at 70 �C with alow activation energy of 0.21 eV (20.4 kJ mol�1). First-principle calculations reveal that the activationenergy for proton transfer is impressively low (0.06 eV and 0.07 eV) with the assistance of watermolecules.

© 2016 Elsevier B.V. All rights reserved.

1. Introduction

Proton exchange membranes (PEMs) have attracted muchattention due to their widespread utilization in electrochemicalenergy systems such as fuel cells [1e4], batteries [5,6] and

electrolyzers [7,8]. Themost commonly used PEMs are composed ofpolymer-based materials, such as perfluorosulfonic acid [9], withwhich themost famous example is commercialized under the tradename Nafion, sulfonated polyether ether ketone [10] and acryl-amide-tert-butyl sulfonic acid [11]. With significant headway beingmade in nanofluidic science, nanostructured polymer membranes,which contain self-organized ion nanochannels driven by the in-compatibility of hydrophilic and hydrophobic polymer moieties,have seen substantial development in recent years [12e16]. For

X.H. Yan et al. / Journal of Power Sources 326 (2016) 466e475 467

example, Watanabe et al. [17] synthesized quaternized aromaticmultiblock copolymers with well-developed ion nanochannels;Ran et al. [18] developed a rod-coil graft copolymer comprising ofhydrophobic rigid main chains and hydrophilic flexible graftchains; Titvinidze et al. [19] prepared a series of multiblock co-polymers consisting of highly sulfonated poly (phenylene sulfone)and poly (phenylene ether sulfone) segments as the PEM, in whichthe spontaneous phase separation results in ordered and contin-uous ion nanochannels; Bae et al. [20,21] reported poly (aryleneether sulfone ketone) multiblock copolymer membranes exhibitinghighly sulfonated hydrophilic blocks, where distinct hydrophilic/hydrophobic phase separations were observed to form distinct ionnanochannels. The nanostructured membrane displays remarkableadvantages: the well-defined nanochannels exhibiting long-rangeorder morphology can significantly enhance the ion mobility byreducing the morphology barrier, and in turn improve the protonconductivity [22,23]. In addition, the issue of fuel permeation,which plagues many electrochemical energy systems, can be solvedvia size exclusion effect in such a confined microenvironment[24e26]. While promising, limitations for this type of nano-structured polymer membranes need to be addressed. First,nanoscale morphology of polymer-based PEMs proves difficult tobe controlled with precision. Second, swelling is observed from dryto hydrated state during operation, causing a volumetric discrep-ancy and disturbing the nanostructure. Membrane properties, suchas permeability and stability, are thus affected. Meanwhile, itshould be mentioned that the “nanochannels” of this polymermembrane are separated hydrophilic phases with ionic clustersrather than true paralleled and straight channels.

It is currently feasible to obtain inorganic materials withdesignable architecture and tunable pore structures, where theinorganic material itself is free of swelling. Hence, one method toovercome the limitations of the polymer membrane is to adopt theuse of inorganic material based PEM. In this work, a silica nanotube(SNT) membrane, the framework of the PEM, is prepared by asurface-sol-gel (SSG) method assisted with a through-hole anodicaluminium oxide (AAO) template. The size of the nanochannel canbe controlled by the SSG condition. The nanochannels shape theproton-conducting pathway and the assembly of the inner wallswith a monomolecular layer (SO3He(CH2)3eSie(OCH3)3) enhancesthe ion exchange capacity, allowing the Grotthuss transport mech-anism to be achieved. As a result, a PEM composed of straight,aligned and size-controllable proton conducting nanochannels isdeveloped. The basic fabricationprocess of the inorganic frameworkproton exchange nanochannel membrane is illustrated in Fig. 1.

2. Experimental methods

2.1. Fabrication of the silica based nanochannel membrane

Commercially available through-hole AAO template with porediameters of 30 nm and a thickness of 30 mmwas used. The surfacesol-gel (SSG) synthetic method was used to fabricate the SiO2nanotube (denoted as SNT) membrane. The AAO template was firstimmersed in a 99% SiCl4 solution (Sigma-Aldrich) for 4 min andsubsequently rinsed with hexane and immersed in fresh hexane for30 min to remove any unbound molecules. The template was thenplaced inmethanol/hexane (1:1) and ethanol for 5 min respectivelyto replace the hexane, and dried under a stream of N2. The exper-imental procedures were completed in an argon-filled glovebox toavoid undesired hydrolysis of SiCl4. Finally, the templatewas placedinto the deionized water for 5 min to make the adsorbed SiCl4molecules hydrolyze to form SiO2. The aforementioned methoddescribes one cycle of SSG. In total, four cycles were carried out forthe synthesis of the silica nanotube membrane.

2.2. Functionalization of the nanochannel membrane

The fabricated nanochannel membrane (evacuated at 393 K)was placed in dry toluene solution with 0.5 vol% 3-mercaptopropyltrimethoxysilane (MPTMS) at ambient tempera-ture for 100 h to allow diffusion of the MPTMS molecules into thenanochannel to react with the nanopatterned SieOH on the innerwalls of the SNTs. The membrane was subsequently rinsed withtoluene to remove residual MPTMS and dried under a stream of N2.The eSH end groups assembled on the nanochannel were thenconverted to eSO3H groups by mild oxidation with 35 wt% H2O2 atambient temperature for 48 h. The membrane was then washedwith water and ethanol, respectively, and acidified with0.1 M H2SO4, followed by thorough washed with water and dried.

2.3. Characterization

The morphologies of the nanochannel membrane wereobserved by a high resolution scanning electron microscope (SEM).The SNTs, which act as the membrane framework, were observedby a transmission electron microscopy (TEM) fitted with energy-dispersive X-ray spectroscopy (EDS). The functionalization of theinner wall was confirmed by X-ray photoelectron spectroscopy(XPS), Fourier transform infrared spectroscopy (FTIR) and time offlight secondary ion mass spectrometry (ToF-SIMS). The membraneproton conductivity was measured with a potentiostat (EG&GPrinceton, model M2273). The prepared nanochannel membranewas sandwiched by a pair of gold-coated stainless steel electrode.This setup was then clamped by a home-made module andimmersed in deionized water. The proton conductivity wasmeasured by the potentiostat through an AC impedance method.The spectra were recorded at a frequency range from 1000 kHz to1 Hz with wave amplitude of 10 mV. The proton conductivity s canbe obtained from:

s ¼ LRU � A

(1)

where L represents the thickness of the membrane, A is the surfacearea of the membrane and RU is the membrane resistance whichwas obtained by calculating the intercept of the high frequencyregion.

2.4. Computational methods

All density functional theory (DFT) based first-principles studieswere performed using ABINIT code [27,28]. The electron-ion in-teractions were described by the projector augmented wave (PAW)method [29], and the electronic exchange correlation effect wasimplemented within the generalized gradient approximation(GGA) of the Perdew-Burke-Ernzerhof (PBE) type [30]. The valenceelectrons were expended on a planewave basis with a 22 Ha energycutoff. The Brillouin zone integrationwas sampled using a 4� 4� 4and 2 � 2 � 1 Monkhorst-Pack [31] k-point mesh for bulk andsurfaces, respectively. A 20 Å vacuum along z-direction was used toavoid the interaction between adjacent supercell in the slab model.The Self-Consistent-Field (SCF) cycles were continued until thetolerance for differences of forces reached 4.0 � 10�5 Ha/Bohr ontwo successive cycles, and the maximal absolute force tolerance forstructural optimization was set to be 4 � 10�4 Ha/Bohr.

First-principles studies of amorphous silica are challenging dueto lack of periodicity in the system. Previous investigations used avariety of crystalline structures to model the structure of amor-phous silica and demonstrated high feasibility [32e36]. In thiswork, a-quartz SiO2 with hexagonal crystal structure is used to

Fig. 1. Schematic illustration for the fabrication of the proton exchange nanochannel membrane.

X.H. Yan et al. / Journal of Power Sources 326 (2016) 466e475468

represent silica. A 2 � 2 � 1 SiO2 (001) supercell containing sixlayers alternating with two layers of oxygen atoms and one layer ofsilicon atoms was chosen to model the fully hydroxylated SiO2surface, which had been shown in previous works as the mostfavorable orientation of amorphous silica. The hydroxylated surfacewas then modified by 3-mercaptopropyl trimethoxysilane(MPTMS). A ladderlike-grafting mode was formed by self-condensation, as described in previous works [37]. And the mer-capto group was substituted by sulfonic acid groups in furthermodification. All atoms were fully relaxed except for the bottomthree layers fixed at the bulk position.

To calculate the energy barriers, a minimum energy path (MEP)for proton-hopping was determined by a simplified string method[38,39], which had been validated to find structures at the transi-tion state.

3. Results and discussion

3.1. Fabrication and characterization of the inorganic frameworknanochannel membrane

The surface sol-gel (SSG) method involves repeating a two-stepadsorption/hydrolysis deposition to fabricate the nanochannelmembrane (denoted as NM hereafter). The molecular precursor,SiCl4, is first adsorbed onto the hydrated surface of AAO, and thenrinsed to desorb any unbounded adsorbate molecules from thepores. In the second step of this process, the adsorbed SiCl4 is hy-drolyzed to form SNT. The layer-by-layer growth process creates isadvantageous in creating a uniform and smooth SNT preparation.Fig. 2a depicts the SEM image of the top view of the AAO templatewhich displays through-holes measured at an average diameter of30 nm. At the conclusion of four cycles of SSG, pore diametersdecrease due to the growth of silica nanotubes. A layer of silica filmis also coated upon the top and bottom surfaces of the AAO, asshown in Fig. 2b. It was observed that the majority of the SNT poresare exposed, while only a very small number of pores are blockedby the silica film formed during the SSG process. More SSG cycleswere also performed, but the close-ends SNTs would be predomi-nant. TEM images are shown in Fig. 2cef, which demonstrate

robust SNTs with smooth and uniformwalls. It can be seen that theshape of the SNT replicates the pore structure of the AAO template;its outer diameter is analogous with the pore diameter of AAO as30 nm. The SNT also exhibits uniform inner diameter of 20 nmwhich is determined by the SSG cycle times. The SNT has an averagewall thickness of 5 nm. An EDS analysis was performed during theTEM characterization. No chlorine was detected during this char-acterization, suggesting that silica conversion from SiCl4 wascomplete. Thus, the chemical composition of the oxide nanotubecan be described as (SiO2)x(SiOH)y. The SNTs comprise the frame-work of the nanochannel membrane and shape the pathway forproton transfer.

Fig. 3a and b depict the top and cross-sectional views of thenanochannel membrane, respectively. Fig. 3a indicates that thenanochannel membrane exhibits parallel channels and highporosity (6.8%, 2.0� 1010 pores cm�2). Only a few silica particles areobserved on the top surface, which were formed during the SSGprocess. The thickness of the membrane is 30 mm as shown inFig. 3b and the tortuosity of the nanochannel in the SNT membraneis 1. The photograph of the final membrane in Fig. 3c shows hightransparency and a large area. The membrane is highly hydrophilicwhich allows further modification in the liquid phase, which can beseen in Fig. 3d.

3.2. Functionalization of the nanochannel membrane

During hydrolysis, the silica surface is covered by silicon hy-droxyl (SieOH). De Boer et al. [40] and Peri et al. [41] demon-strated that the silicon hydroxyl number could be 4.55 ^SieOHgroups nm�2 on silica surface, offering abundant sites for surfacemodification. Immersing the nanochannel membrane into toluenesolution in addition of a small amount of 3-mercaptopropyl-trimethoxysilane (MPTMS) leads to co-condensation of themethoxy groups of the MPTMS with the hydroxyl groups on thewall of the nanochannel as a process of self-assembly. The eSHend group of the MPTMSmolecule was then oxidized to eSO3H byH2O2 and H2SO4. To confirm the uniform functionalization, a longreaction time is required for these processes. The thickness of theself-assembled HSO3e(CH2)3eSieOeSiemonolayer is ~1 nm, and

Fig. 2. (a) The SEM image of the pristine AAO template, scale bar is 100 nm; (b) the SEM image of the silica nanotubes based nanotube membrane after 4 cycles of SSG, scale bar is100 nm; (cee) TEM image of the silica nanotubes, scale bar is 50 nm; (f) TEM image of silica nanotube with higher magnification, scale bar is 20 nm.

Fig. 3. (a) Overview of the top-surface of the nanochannel membrane, scale bar is 1 mm; (b) SEM image of the cross-sectional view of the nanochannel membrane, scale bar is10 mm; (c) Photograph of nanochannel membrane with high transparency, scale bar is 5 mm; (d) The contact angle of the nanochannel membrane, scale bar is 0.4 mm.

X.H. Yan et al. / Journal of Power Sources 326 (2016) 466e475 469

X.H. Yan et al. / Journal of Power Sources 326 (2016) 466e475470

the channel size of the NM shrinks to 18 nm after self-assembly.FTIR was used to clarify the functionalization of the nano-channel membrane, in comparison with the pristine AAO tem-plate. As the FTIR spectrum of the membranes in Fig. 4a suggests,the functionalized nanochannel membrane displays severalnewly-presented characteristic peaks typically from 1000 cm�1 to1200 cm�1. The peaks appearing at 1080 cm�1, 1092 cm�1,1124 cm�1 and 1160 cm�1 are assigned to SieOeSi due to theformation of SNTs. A typical characteristic peak around 1134 cm�1

is identified which belongs to eSO3 asymmetric stretching vi-bration. A around 1058 cm�1 peak belongs to eSO3 symmetricstretching vibration. The peak at 1112 cm�1 is assigned to CH2chain wag and the peak around 1143 cm�1 corresponds to CeCstretching [42e44]. These results indicate that silane-based self-assembly onto the nanochannels was successful.

Fig. 4b and c depict the XPS spectra of the functionalizednanochannel membrane. Peaks at around 285.0 eV and 169.4 eV

Fig. 4. (a) FTIR spectra of the pristine AAO template and the nanochannel membraneafter functionalization; (bec) XPS spectra of C 1s and S 2p of the nanochannel mem-brane after functionalization; (d) ToFeSIMS results of phase depth profile for theproton exchange nanochannel membrane.

were detected, which are assigned to C 1s and S 2p, respectively. Itshould be clarified that the electrolyte used for AAO fabrication issulfuric acid, which may result in the detection of residual sulfur.Nevertheless, the presence of both C and S can confirm the surfacefunctionalization by MPTMS. A proton exchange membrane mustconsist of proton transfer pathways, which is realized by the inor-ganic nanochannels created herein. The other important require-ment in our proposed inorganic framework proton exchangenanochannel membrane is the formation of a continuous self-assembled HSO3e(CH2)3eSieOeSie layer, where continuous pro-ton hopping can occur. To further guarantee that the assembly ofthe monomolecular layer was continuous over the entire nano-channel surface, the ToF-SIMS was also performed with a depth of6.7 mm‚ a valuewhich is approximately one fourth of themembranethickness. The ToF-SIMS profile, shown in Fig. 4d, indicates the co-existence of both carbon and sulfur which confirms the assembly ofthe HSO3e(CH2)3eSieOeSie monomolecular layer. This profilealso shows that the carbon and sulfur content do not decrease withdepth, and their contents are relatively uniform at the depth di-rection. The aforementioned experimental results prove that thenanochannels have been successfully functionalized and the silane-based self-assembly within the nanochannels is continuous ratherthan intermittent, allowing proton hopping between sulfonic acidgroups.

3.3. Proton conductivity

Fig. 5 presents the measured proton conductivities of the inor-ganic framework proton exchange nanochannel membrane atdifferent temperatures. It is shown that the conductivity increaseswith an increase in temperature. The literature has reported thatthe proton conductivity of silica itself is very low at only5� 10�3 mS cm�1 at 25 �C, suggesting that proton transfer throughpure silica via surface hydroxyl groups would be very difficult [45].In comparison with pure silica, the proton conductivity of ourproposed SNT based membrane reaches 4.46 mS cm�1 at 25 �C,indicating that the surface sulfonic acid groups on the nanochannelwalls play a critical role in proton transfer which can be furtherillustrated by the density functional theory (DFT) calculation. Withan increase in temperature to 70 �C, the proton conductivity in-creases to 11.3 mS cm�1. The high conductivity can be ascribed tothe following: (1) a straight nanochannel structure introduced by

Fig. 5. Proton conductivity of the proton exchange nanochannel membrane in DIwater at different temperatures.

X.H. Yan et al. / Journal of Power Sources 326 (2016) 466e475 471

SNT results in a lowmorphology barrier for protonmobility and theshortest path for proton transportation through the membrane; (2)the continuous self-assembly of silane with sulfonic acid groupsonto the nanochannel walls allows continuous proton hoppingthrough the entire nanochannel. The conductivity of this inorganicframework nanochannel membrane shows Arrhenius-type tem-perature dependence and the activation energy is recorded to be20.4 kJ mol�1 (0.21 eV). In contrast, the activation energy of puresilica is as high as 46.0 kJ mol�1, reported by previous work [46].

3.4. Proton transfer mechanism

The proton transfer process in the SNT based nanochannel issimulated by the first-principles study to clarify the reasoningbehind the high conductivity. In our proposed membrane, theGrotthuss transport mechanism should be predominant, as can beseen from the large discrepancy between the activation energiesbefore and after surface modification as above mentioned. Theoptimized lattice constants of a-quartz SiO2 are a¼ b¼ 5.025 Å andc ¼ 5.526 Å, which are consistent with previous experimental(~2.2% larger) and theoretical (~0.5% smaller) investigations [47,48].The side-view and top-view of the fully hydroxylated SiO2 (001)surface are shown in Fig. 6a and b, respectively. It is found thatthere are two types of hydroxyl groups present on the hydroxylatedsurface, labeled as O1 and O2. The O1 is the more exposed oxygenatom with protruded hydrogen atom, and O2 is the moreembedded one with retracted hydrogen atom. The electron densitydistribution in the middle of O1 and O2 atoms along z-direction ispresented in Fig. 6c. A higher electron density is seen on the, O1atom as opposed to the O2 atom, suggesting that the O1 atom ex-hibits higher activity, which is congruent with previous reports[49e51]. As a result, it is reasonable to propose that the self-assembly with MPTMS occurs on O1 atom sites rather than on O2atoms. The fully modified structure of the nanochannel is shown inFig. 6d, with half of the O1 atoms bonded to MPTMS. In addition,self-condensation is assumed to occur on unilateral SieOH sites to

Fig. 6. (a) Side view and (b) top view of the hydroxylated SiO2 (001) surface; (c) the electrostructure of nanochannel.

model the proton-hopping in the bonded case (SieOeSi) andseparated case (SieOH HOeSi).

The motion of protons inside the nanochannel is proposed to berealized by a series of hopping between oxygen atoms of the sul-fonic acid groups. Based on the surface structure shown in Fig. 6d,there are two steps for proton-hopping e in the bonded andseparated cases. In each case, we calculated the minimum energypath (MEP) for proton hopping to obtain the activation energy.Moreover, in order to reveal the effect of watermolecules on protonmotion, the activation energy obtained from one water moleculeassisted pathway is compared with that of the anhydrous protontransfer pathway. With water adsorbed, proton-hopping shows a“rocking” mechanism instead of a “jumping” mechanism [46,52].

3.4.1. Proton-hopping in bonded caseFirst, we investigated the proton-hopping in the bonded case

(Step 1 in Fig. 6d); the energy profiles of the anhydrous and water-assisted proton-hopping pathways are shown in Fig. 7a. It is foundthat the activation energy for anhydrous pathway is 0.18 eV.However, the involvement of water molecules in the transportationprocess decreases the activation energy barrier substantially, to avalue of 0.06 eV, indicating that the dominant pathway in thebonded case is water-assisted.

To gain a better understanding into the mechanisms of proton-hopping in anhydrous and water-assisted pathways, the calculatedstructures of initial state (IS), transition state (TS) and final state(FS) are shown in Fig. 7b with the simulated interatomic distancesand angles listed in Table 1. The proton-hopping pathway consid-ered here is the proton from the initial site O1 oxygen atom to thefinal site O2 oxygen atom. In the transition state of anhydrouspathway, the distance between O1 and O2 atoms is shortened to2.552 Å to promote proton transfer. On the contrary, in the case of awater-assisted pathway, the distance is elongated to 3.897 Å toincorporate the water molecule. Although the distance increases,the activation energy for proton-hopping in water-assisted path-ways decreases which is mainly attributed to two facts: i) the

n density in the middle of O1 atom and O2 atom along z-direction; (d) the optimized

Fig. 7. (a) The energy profile of anhydrous and water-assisted proton-hopping pathways in the bonded case; (b) the calculated structures of initial state (IS), transition state (TS) andfinal state (FS) for anhydrous pathway and water-assisted pathway in bonded case.

X.H. Yan et al. / Journal of Power Sources 326 (2016) 466e475472

hydrogen-bond angles (:O1eH1eO2) of initial and final states inanhydrous pathway are 166.01� and 167.01�, which is far from theideal 180� for proton-hopping in water system. For the water-assisted pathway, however, both the hydrogen-bond angle in theinitial state (:O1eH2eO5) and the final state (:O5eH3eO2) in-crease to the value of 175.96� and 175.63� respectively, which aremuch closer to the ideal 180�, leading to a lower activation energy;and ii) the oxygen atoms in the nanochannel form much tighterOeH bonds (in other word, shorter bond length) with proton in theanhydrous pathway compared to that of the water-assistedpathway, resulting in a much higher energy requirement for sep-aration. For example, it is shown that the O1eH1 bond (1.027 Å,anhydrous pathway) is shorter than the O1eH2 bond (1.135 Å,water-assisted pathway) in the initial state, and the H1eO2 bond(1.031 Å, anhydrous pathway) is shorter than the H3eO2 bond(1.135 Å, water-assisted pathway) in the final state. It is also shownthat the O1eS1 bond (1.575, 1.531 and 1.491 Å for initial, transitionand final states) in the anhydrous pathway is longer than that(1.547, 1.515 and 1.489 Å for initial, transition and final state) in thewater-assisted pathway. Meanwhile the O2eS2 bond (1.491, 1.530and 1.574 Å for initial, transition and final states) in anhydrouspathway is longer than that (1.488, 1.513 and 1.545 Å for initial,transition and final states) in water-assisted pathway. To accountfor the weaker binding energy between channel oxygen atoms (O1,O2) and protons in the water-assisted pathway, the bond length inthe water molecule is also given. In the transition state, the O1eH2bond (1.395 Å) is longer than the H2eO5 bond (1.136 Å), and theH3eO2 bond (1.387 Å) is longer than the O5eH3 bond (1.130 Å),suggesting that the oxygen atom in the water molecule gives a drag

force to the proton and will decrease the interaction between thenanochannel and protons.

3.4.2. Proton-hopping in separated caseThe energy profiles of anhydrous and water-assisted proton-

hopping pathways in the separated case (Step 2 in Fig. 6d) arepresented in Fig. 8a, showing that the activation energies foranhydrous and water-assisted proton-hopping pathways are0.26 eV and 0.07 eV, respectively. A much lower activation energyin water-assisted pathway indicates that it is the preferred proton-hopping pathway even in the separated case.

Fig. 8b shows the calculated structures of initial state (IS),transition state (TS) and final state (FS), and Table 2 lists thesimulated interatomic distances and angles. In this case, theproton-hopping pathway considered here is the proton from initialsite O6 to the final site O7 atom. Similar to the bonded case, thedistance between O6 and O7 oxygen atoms in water-assistedpathway (3.700, 3.758 and 3.827 Å for initial, transition and finalstates) is still larger than that in anhydrous pathway (2.639, 2.571and 2.624 Å for initial, transition and final states) due to the exis-tence of the water molecule. To account for the lower activationenergy in the water-assisted pathway, the hydrogen-bond anglesand bond lengths are also given. The hydrogen-bond angles in theinitial (:O6eH2eO5) and final states (:O5eH3eO7) are 176.15�

and 172.45� in water-assisted pathway, respectively, which arelarger than that (:O6eH1eO7) in anhydrous pathway (173.75�

and169.29� for initial and final states), indicating that in the water-assisted pathway, the hydrogen-bond angle is closer to the ideal180� and it favors proton-hopping. Moreover, the O6eH1 bond

Table 1The simulated interatomic distances and angles of the initial state (IS), transitionstate (TS) and final state (FS) for proton-hopping on anhydrous and water-assistedpathways in bonded case (angle in degrees, distance in Å).

Anhydrous Water-assisted

IS TS FS IS TS FS

:S1eO1eH1 114.53 121.98 127.59:S1eO1eH2 114.99 119.33 122.55:O1eH1eO2 166.01 167.50 167.01:H1eO2eS2 130.52 123.51 114.93:O1eS1eO3 110.46 111.35 112.13 110.85 111.49 112.34:O2eS2eO4 111.87 111.24 110.37 112.02 111.44 110.69:O1eH2eO5 175.96 173.44 170.10:H2eO5eH3 107.94 108.13 108.09:O5eH3eO2 169.04 173.58 175.63:H3eO2eS2 134.10 126.75 117.31:H2eO5eH4 114.75 111.29 107.40:H3eO5eH4 108.10 110.32 113.12O1eS1 1.575 1.531 1.491 1.547 1.515 1.489O1eH1 1.027 1.279 1.538O1eH2 1.135 1.395 1.649H1eO2 1.553 1.288 1.031O2eS2 1.491 1.530 1.574 1.488 1.513 1.545O3eS1 1.452 1.460 1.471 1.456 1.464 1.468O4eS2 1.473 1.460 1.455 1.475 1.469 1.463O1/O2 2.562 2.552 2.553 3.895 3.897 3.935O3/O4 2.986 2.921 2.860 3.272 3.228 3.220O1/O3 2.487 2.470 2.458 2.472 2.463 2.456O2/O4 2.455 2.468 2.488 2.456 2.464 2.475H2eO5 1.309 1.136 1.007O5eH4 0.972 0.983 0.973O5eH3 1.007 1.130 1.309H3eO2 1.642 1.387 1.135O1/O5 2.442 2.526 2.647O5/O2 2.637 2.513 2.442H2/H3 1.881 1.834 1.884

Fig. 8. (a) The energy profile of anhydrous and water-assisted proton-hopping pathways inand final state (FS) for anhydrous pathway and water-assisted pathway in separated case.

X.H. Yan et al. / Journal of Power Sources 326 (2016) 466e475 473

(1.014 Å, anhydrous pathway) is shorter than the O6eH2 bond(1.096 Å, water-assisted pathway) in the initial state and H1eO7bond (1.019 Å, anhydrous pathway) is shorter than the H3eO7 bond(1.085 Å, water-assisted pathway) in the final state. And the S2eO6bond (1.578, 1.536 and 1.487 Å for initial, transition and final states)in the anhydrous pathway is longer than that (1.552, 1.506 and1.482 Å for initial, transition and final state) in the water-assistedpathway. Meanwhile, the O7eS1 bond (1.485, 1.530 and 1.576 Åfor initial, transition and final states) in the anhydrous pathway isalso longer than that (1.482, 1.504 and 1.560 Å for initial, transitionand final states) in the water-assisted pathway. As a result, themechanisms found in the separated case are in agreement withthose found in the bonded case.

To sum up, in this nanochannel, the water-assisted proton-hopping pathway is the dominant one when compared with theanhydrous pathway. The significant lower activation energies in thewater-assisted pathway is mainly due to two factors: i) thehydrogen-bond angle in the water-assisted pathway is much closerto the ideal angle of 180� in comparison with the anhydrouspathway, which is favorable to proton-hopping, and ii) the dragforce from oxygen atoms on the water molecule to protons de-creases the binding energy between the nanochannel and theprotons, allowing much more proton mobility when water mole-cules are involved. To further confirm the validity of the proton-hopping mechanism depicted by first-principle calculations, wecompare the experimental and computational activation energyvalues and discuss the consistency between them. The measuredactivation energy is 0.21 eV, larger than the calculated results(0.06 eV for bonded case and 0.07 eV for separated case). It shouldbe mentioned that the computational results are obtained fromideal model with sulfonic acid groups periodically existing.

the separated case; (b) the calculated structures of initial state (IS), transition state (TS)

Table 2The simulated interatomic distances and angles of the initial state (IS), transitionstate (TS) and final state (FS) for proton-hopping on anhydrous and water-assistedpathways in separated case (angle in degrees, distance in Å).

Anhydrous Water-assisted

IS TS FS IS TS FS

:S2eO6eH1 112.34 121.26 128.51:S2eO6eH2 112.63 127.68 137.14:O6eH1eO7 173.75 174.74 169.29:O6eH2eO5 176.15 172.58 166.81:H2eO5eH3 103.57 108.87 106.31:O5eH3eO7 160.65 168.45 172.45:H3eO7eS1 114.93 133.55 116.06:H2eO5eH4 106.33 109.03 107.16:H3eO5eH4 105.83 103.47 113.42:H1eO7eS1 140.32 128.89 114.96:O6eS2eO2 111.26 111.39 112.39 110.01 111.72 112.43:O6eS2eO4 106.35 109.70 111.29 109.65 110.90 111.72:O7eS1eO1 111.05 109.02 105.84 111.56 110.94 109.09:O7eS1eO3 112.64 111.70 111.38 112.22 111.69 110.73S2eO6 1.578 1.536 1.487 1.552 1.506 1.482O6eH1 1.014 1.280 1.616O6eH2 1.096 1.381 1.710H1eO7 1.629 1.294 1.019O7eS1 1.485 1.530 1.576 1.482 1.504 1.560O6/O7 2.639 2.571 2.624 3.700 3.758 3.827H2eO5 1.386 1.120 0.997O5eH4 0.975 0.983 0.972O5eH3 0.994 1.102 1.401H3eO7 1.728 1.377 1.085O6/O5 2.481 2.496 2.690O5/O7 2.686 2.467 2.480H2/H3 1.886 1.808 1.934

X.H. Yan et al. / Journal of Power Sources 326 (2016) 466e475474

However, in real experiments, it is extremely difficult to make allthe SieOH groups on the nanochannel surface assemble with sul-fonic acid groups; thus, the distance between adjacent sulfonic acidgroups at some sites is larger than that in theoretical model. Fromour results, it is found that the distance between adjacent sulfonicacid groups can greatly influence the calculated activation energy.For example, in the theoretical model of Figs. 7b and 8b, the dis-tance between O1 and O2 is 2.92 Å and the distance between O6 andO7 is 3.36 Å. The larger OeO distance in separated case than bondedcase leads to the larger activation energies for both anhydrous(0.26 eV > 0.18 eV) and water-assisted (0.07 eV > 0.06 eV) path-ways. Thus, due to the larger distance between adjacent sulfonicacid groups in experiments than theoretical model in some parts ofthe nanochannel, the measured activation energy value is largerthan the calculated results. More importantly, the calculated acti-vation energy in our nanochannel is only 0.06 eV, which is evenlower than that in Nafionwith the value of 0.09 eV [53], suggestingthe membrane synthesized in this work can theoretically achievemuch higher conductivities.

4. Conclusions

In summary, we propose an inorganic framework nanochannelmembrane as a novel PEM, where the inorganic framework itselfcan achieve designable architecture and tunable pore structures.The experimental studies show that the proton conductivity of thisnanochannel membrane could reach 4.46 mS cm�1 at 25 �C and11.3 mS cm�1 at an elevated temperature of 70 �C. These results areattributed to the low morphology barrier from straight nano-channels and the presence of a continuous proton hopping envi-ronment from the self-assembled proton conductingmonomolecular layer. A two-step proton transfer mechanism isestablished, with each step demonstrating a low activation energy(0.06 eV and 0.07 eV) with the assistance of water molecules. A

continuous stream of proton hopping along the monomolecularlayer is therefore created, resulting in high proton conductivity. Thepresent study provides an attractive alternative for developingnanostructured PEMS and guidance in designing inorganic-material based membranes.

Acknowledgements

The work described in this paper was fully supported by a grantfrom the Research Grants Council of the Hong Kong SpecialAdministrative Region, China (Project No. 16213414).

References

[1] B.C. Steele, A. Heinzel, Nature 414 (2001) 345e352.[2] X. Zhao, W. Yuan, Q. Wu, H. Sun, Z. Luo, H. Fu, J. Power Sources 273 (2015)

517e521.[3] X. Yan, T. Zhao, L. An, G. Zhao, L. Zeng, Appl. Energy 138 (2015) 331e336.[4] Q. Wu, H. Li, W. Yuan, Z. Luo, F. Wang, H. Sun, X. Zhao, H. Fu, Appl. Energy 160

(2015) 146e152.[5] W. Wang, Q. Luo, B. Li, X. Wei, L. Li, Z. Yang, Adv. Funct. Mater. 23 (2013)

970e986.[6] J. Huang, Q. Zhang, H. Peng, X. Liu, W. Qian, F. Wei, Energy & Environ. Sci. 7

(2014) 347e353.[7] F. Barbir, Sol. Energy 78 (2005) 661e669.[8] M. Carmo, D.L. Fritz, J. Mergel, D. Stolten, Int. J. Hydrogen Energy 38 (2013)

4901e4934.[9] K. Kreuer, Solid State Ion. 97 (1997) 1e15.

[10] K. Kreuer, J. Membr. Sci. 185 (2001) 29e39.[11] T. Yamaguchi, F. Miyata, S. Nakao, Adv. Mater 15 (2003) 1198e1201.[12] N. Li, C. Wang, S.Y. Lee, C.H. Park, Y.M. Lee, M.D. Guiver, Angew. Chem. 123

(2011) 9324e9327.[13] G. He, Z. Li, J. Zhao, S. Wang, H. Wu, M.D. Guiver, Z. Jiang, Adv. Mater 27 (2015)

5280e5295.[14] J. Zhang, J. Li, H. Tang, M. Pan, S.P. Jiang, J. Mater. Chem. a 2 (2014)

20578e20587.[15] E.M. Tsang, Z. Zhang, A.C. Yang, Z. Shi, T.J. Peckham, R. Narimani, B.J. Frisken,

S. Holdcroft, Macromolecules 42 (2009) 9467e9480.[16] Q. Li, L. Liu, Q. Miao, B. Jin, R. Bai, Polym. Chem. 5 (2014) 2208e2213.[17] M. Tanaka, K. Fukasawa, E. Nishino, S. Yamaguchi, K. Yamada, H. Tanaka,

B. Bae, K. Miyatake, M. Watanabe, J. Am. Chem. Soc. 133 (2011) 10646e10654.[18] J. Ran, L. Wu, B. Wei, Y. Chen, T. Xu, Sci. Rep. 4 (2014).[19] G. Titvinidze, K. Kreuer, M. Schuster, C.C. de Araujo, J.P. Melchior, W.H. Meyer,

Adv. Funct. Mater. 22 (2012) 4456e4470.[20] B. Bae, K. Miyatake, M. Watanabe, Macromolecules 43 (2010) 2684e2691.[21] B. Bae, K. Miyatake, M. Uchida, H. Uchida, Y. Sakiyama, T. Okanishi,

M. Watanabe, ACS Appl. Mater. Interfaces 3 (2011) 2786e2793.[22] H. Chinen, K. Mawatari, Y. Pihosh, K. Morikawa, Y. Kazoe, T. Tsukahara,

T. Kitamori, Angew. Chem. Int. Ed. 51 (2012) 3573e3577.[23] C. Dellago, M.M. Naor, G. Hummer, Phys. Rev. Lett. 90 (2003) 105902.[24] D. Huh, K. Mills, X. Zhu, M.A. Burns, M. Thouless, S. Takayama, Nat. Mater. 6

(2007) 424e428.[25] Z. Cao, Y. Peng, T. Yan, S. Li, A. Li, G.A. Voth, J. Am. Chem. Soc. 132 (2010)

11395e11397.[26] M. Thomas, B. Corry, Philos. Trans. A. Math. Phys. Eng. Sci. 374 (2016), http://

dx.doi.org/10.1098/rsta.2015.0020.[27] X. Gonze, J. Beuken, R. Caracas, F. Detraux, M. Fuchs, G. Rignanese, L. Sindic,

M. Verstraete, G. Zerah, F. Jollet, Comput. Mater. Sci. 25 (2002) 478e492.[28] X. Gonze, Z. für Kristallogr. Cryst. Mater. 220 (2005) 558e562.[29] G. Kresse, D. Joubert, Phys. Rev. B 59 (1999) 1758.[30] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865.[31] H.J. Monkhorst, J.D. Pack, Phys. Rev. B 13 (1976) 5188.[32] Y. Kang, J. Kang, K. Chang, Phys. Rev. B 78 (2008) 115404.[33] M.Z. Hossain, Appl. Phys. Lett. 95 (2009) 143125.[34] P. Shemella, S.K. Nayak, Appl. Phys. Lett. 94 (2009) 2101.[35] T.C. Nguyen, M. Otani, S. Okada, Phys. Rev. Lett. 106 (2011) 106801.[36] A. Rudenko, F. Keil, M. Katsnelson, A. Lichtenstein, Phys. Rev. B 84 (2011)

085438.[37] J. Wu, L. Ling, J. Xie, G. Ma, B. Wang, Chem. Phys. Lett. 591 (2014) 227e232.[38] E. Weinan, W. Ren, E. Vanden-Eijnden, Phys. Rev. B 66 (2002) 052301.[39] E. Weinan, W. Ren, E. Vanden-Eijnden, J. Chem. Phys. 126 (2007) 164103.[40] J. De Boer, M. Hermans, J. Vleeskens 60 (1957) 234.[41] J. Peri, A. Hensley Jr., J. Phys. Chem. 72 (1968) 2926e2933.[42] P. Larkin, Infrared and Raman Spectroscopy; Principles and Spectral Inter-

pretation, Elsevier, 2011.[43] J. Fan, P.K. Chu, Silicon Carbide Nanostructures: Fabrication, Structure, and

Properties, Springer, 2014.[44] T. Remyamol, P. Gopinath, H. John, RSC Adv. 4 (2014) 29901e29908.[45] S. Lu, D. Wang, S.P. Jiang, Y. Xiang, J. Lu, J. Zeng, Adv. Mater 22 (2010)

971e976.

X.H. Yan et al. / Journal of Power Sources 326 (2016) 466e475 475

[46] Y. Zhou, J. Yang, H. Su, J. Zeng, S.P. Jiang, W.A. Goddard, J. Am. Chem. Soc. 136(2014) 4954e4964.

[47] L. Levien, C.T. Prewitt, D.J. Weidner, Am. Mineral. 65 (1980) 920e930.[48] T. Goumans, A. Wander, W.A. Brown, C.R.A. Catlow, Phys. Chem. Chem. Phys. 9

(2007) 2146e2152.[49] J. Li, J. Wu, C. Zhou, B. Han, E.J. Karwacki, M. Xiao, X. Lei, H. Cheng, J. Phys.

Chem. C 113 (2009) 9731e9736.

[50] B. Han, Q. Zhang, J. Wu, B. Han, E.J. Karwacki, A. Derecskei, M. Xiao, X. Lei,M.L. O’Neill, H. Cheng, J. Phys. Chem. C 116 (2011) 947e952.

[51] G. Ni, B. Han, H. Cheng, J. Phys. Chem. C 117 (2013) 22705e22713.[52] M.J. Janik, R.J. Davis, M. Neurock, J. Am. Chem. Soc. 127 (2005) 5238e5245.[53] S. Tanimura, T. Matsuoka, J. Polym. Sci. Part B Polym. Phys. 42 (2004)

1905e1914.