Alcohol-induced drying of carbon nanotubes and its implications for alcohol/waterseparation: A molecular dynamics studyXingling Tian, Zaixing Yang, Bo Zhou, Peng Xiu, and Yusong Tu

Citation: The Journal of Chemical Physics 138, 204711 (2013); doi: 10.1063/1.4807484 View online: http://dx.doi.org/10.1063/1.4807484 View Table of Contents: http://scitation.aip.org/content/aip/journal/jcp/138/20?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Reactive adsorption of ammonia and ammonia/water on CuBTC metal-organic framework: A ReaxFF moleculardynamics simulation J. Chem. Phys. 138, 034102 (2013); 10.1063/1.4774332 Molecular dynamics study of free energy of transfer of alcohol and amine from water phase to the micelle bythermodynamic integration method J. Chem. Phys. 137, 094902 (2012); 10.1063/1.4747491 Highly selective adsorption of methanol in carbon nanotubes immersed in methanol-water solution J. Chem. Phys. 137, 034501 (2012); 10.1063/1.4732313 Hydrogen bond dynamics and microscopic structure of confined water inside carbon nanotubes J. Chem. Phys. 124, 174714 (2006); 10.1063/1.2194540 Diffusive dynamics of water in tert-butyl alcohol/water mixtures J. Chem. Phys. 120, 4759 (2004); 10.1063/1.1645782

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

130.39.62.90 On: Wed, 03 Sep 2014 02:47:15

THE JOURNAL OF CHEMICAL PHYSICS 138, 204711 (2013)

Alcohol-induced drying of carbon nanotubes and its implicationsfor alcohol/water separation: A molecular dynamics study

Xingling Tian,1 Zaixing Yang,2 Bo Zhou,1 Peng Xiu,2,a) and Yusong Tu3,a)

1Bio-X Lab, Department of Physics, Zhejiang University, Hangzhou 310027, China2Department of Engineering Mechanics and Soft Matter Research Center, Zhejiang University,Hangzhou 310027, China3Institute of Systems Biology, Shanghai University, Shanghai 200444, China

(Received 12 March 2013; accepted 7 May 2013; published online 30 May 2013)

Alcohols are important products in chemical industry, but separating them from their aqueous solu-tions is very difficult due to the hydrophilic nature of alcohols. Based on molecular dynamics simu-lations, we observe a striking nanoscale drying phenomenon and suggest an energy-saving and effi-cient approach toward alcohol/water separation by using single-walled carbon nanotubes (SWNTs).We use various common linear alcohols including C1-C6 1-alcohols and glycerol for demonstration(the phenol is also used as comparison). Our simulations show that when SWNTs are immersed inaqueous alcohols solutions, although the alcohols concentration is low (1 M), all kinds of alcoholscan induce dehydration (drying) of nanotubes and accumulate inside wide [(13, 13)] and narrow[(6, 6) or (7, 7)] SWNTs. In particular, most kinds of alcohols inside the narrow SWNTs form nearlyuniform 1D molecular wires. Detailed energetic analyses reveal that the preferential adsorption ofalcohols over water inside nanotubes is attributed to the stronger dispersion interactions of alcoholswith SWNTs than water. Interestingly, we find that for the wide SWNT, the selectivity for 1-alcoholsincreases with the number of alcohol’s carbon atoms (Ncarbon) and exhibits an exponential law withrespect to Ncarbon for C1-C5 1-alcohols; for narrow SWNTs, the selectivity for 1-alcohols is veryhigh for methanol, ethanol, and propanol, and reaches a maximum when Ncarbon = 3. The underlyingphysical mechanisms and the implications of these observations for alcohol/water separation are dis-cussed. Our findings provide the possibility for efficient dehydration of aqueous alcohols (and otherhydrophilic organic molecules) by using SWNT bundles/membranes. © 2013 AIP Publishing LLC.[http://dx.doi.org/10.1063/1.4807484]

I. INTRODUCTION

Alcohols are important products in chemical industry.However, separation of alcohols from their aqueous solutionis very difficult due to the hydrophilic nature of alcohols.Conventionally, the alcohol/water separation requires a dis-tillation step1 which is quite energy intensive and expensive.For example, distillation was estimated to consume more thanhalf of the total energy in the production of ethanol from fer-mentation broth.2 In recent years, several new technologies,such as pervaporation,3, 4 gas stripping,5 and adsorption ontozeolites,6, 7 have been proposed as alternative approaches fordehydration of aqueous alcohols. In this paper, on the basisof molecular dynamics (MD) simulations, we observe a strik-ing nanoscale drying phenomenon8–13 and suggest a new ap-proach toward efficient alcohol/water separation.

Herein, we have chosen single-walled carbon nanotubes(SWNTs) to realize alcohol/water separation. The carbonnanotubes possess well-defined hollow structures and, thus,serve as desirable materials for molecules adsorbing into theirinteriors.14–19 Recently, it has been proposed that the SWNTs(or SWNT bundles/membranes) can be used to separate wa-ter from salt solution,20 separate binary mixture of organic

a)Authors to whom correspondence should be addressed. Electronic ad-dresses: xiupeng2011@zju.edu.cn and ystu@shu.edu.cn.

(or gas) molecules,21, 22 and separate hydrophobic molecules,such as gases (H2, O2, and CO2)23 and methane,24 fromtheir aqueous solutions. In theory, the accumulation of hy-drophobic molecules inside SWNTs is not too unexpectedin view of the hydrophobic interactions (carbon nanotubesare hydrophobic as well). Interestingly, however, in previousstudies25–27 we found that even urea, a hydrophilic molecule,can spontaneously expel water out of SWNTs (induce dry-ing of the SWNTs) and accumulate inside SWNTs due tothe stronger dispersion interaction of urea than water withSWNTs. Inspired by these findings, in this paper, we in-vestigate whether SWNTs can be used to separate alcohols,which are also hydrophilic, from their aqueous solutions. Wenote here that, the selective adsorptions of methanol28, 29 andethanol30, 31 over water onto SWNTs’ inner surfaces had beenrecently observed by some researchers through computer sim-ulations. However, a systematic investigation of adsorptionof alcohols inside SWNTs is still lacking, and the physicalmechanism underlying the aforementioned selective adsorp-tion of alcohols inside nanotubes is still not very clear.

In current simulations, SWNTs are initially immersed invarious alcohols solutions (i.e., binary mixture of alcohol andwater) to investigate the adsorption behaviors of alcohols in-side the hydrophobic nanotubes. Here, we use wide [(13, 13)]and narrow [(6, 6) or (7, 7)] SWNTs, and various alcohols,

0021-9606/2013/138(20)/204711/8/$30.00 © 2013 AIP Publishing LLC138, 204711-1

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

130.39.62.90 On: Wed, 03 Sep 2014 02:47:15

204711-2 Tian et al. J. Chem. Phys. 138, 204711 (2013)

including glycerol and a series of 1-alcohols from methanolto hexanol, for demonstration; phenol is also used as com-parison. Although the concentration of alcohols/phenol is low(∼1 M), for both wide and narrow SWNTs, all kinds of alco-hols and phenol induce dehydration (drying) of nanotubes andaccumulate inside the SWNTs. In particular, most kinds of al-cohols and phenol inside the narrow nanotubes form nearlyuniform 1D molecular wires. Energetic analyses reveal thatalcohols have stronger dispersion interactions with SWNTsthan water which results in the selective adsorption of alco-hols over water inside SWNTs. Furthermore, we compare theselectivity of SWNTs to different alcohols, and compare theselectivity for wide and narrow SWNTs. Interestingly, the se-lectivity for C1-C5 1-alcohols inside the wide SWNT with re-spect to the number of alcohol’s carbon atoms (Ncarbon) can bewell fitted to an exponential function. The underlying physicalmechanisms and the implications for alcohol/water separationare discussed. Our findings suggest that we might use SWNTbundles/membranes to achieve efficient dehydration of aque-ous alcohols (and other hydrophilic organic molecules).

II. SYSTEM AND METHODS

The armchair (13, 13) SWNT, which is 1.74 nm in diam-eter and 3.03 nm in length, was used in current simulationsand referred as to the “wide SWNT.” The initial side lengthof cubic box for the wide SWNT system was 5.55 nm. Wealso used the “narrow SWNT” – (6, 6) or (7, 7) SWNT, whichis the narrowest armchair SWNT that the alcohols/phenol canpenetrate into. For methanol and ethanol, the “narrow SWNT”is the (6, 6) SWNT (0.80 nm in diameter), and the sizes ofinitial simulation box are 3.75, 3.75, and 5.69 nm; for otherkinds of alcohols and phenol, the “narrow SWNT” is the(7, 7) SWNT (0.94 nm in diameter), and the sizes of initialsimulation box are 3.88, 3.88, and 5.69 nm. The lengths of (6,6) and (7, 7) SWNTs are the same (3.27 nm).

Each SWNT was separately solvated in aqueous alco-hols/phenol solutions at a concentration of 1 M (see Fig. 1),and was aligned along the z-axis and positioned at the cen-tral region of the solvation box. The simulation systems ofthe ternary mixture (alcohol-water-SWNT) were prepared asfollows: First, the organic molecules were randomly placedin the solvated box. Here, the numbers of organic and wa-ter molecules in the given simulation boxes were calculatedaccording to the experimentally determined partial molar vol-umes of organic molecules (at a concentration of 1 M)32 andwater33 at 298 K. We used the TIP3P water model,34 andOPLSAA force field35 which was commonly used for simula-tions of organic molecules especially linear alcohols.16, 29, 36, 37

Second, the energies of the binary solution systems were min-imized with a steepest descent algorithm, followed by a 1-ns NPT equilibration at 298 K. Third, the SWNT was in-serted in the equilibrated binary solution, with water/organicmolecules removed if the distance between oxygen of waterand any carbon atom of SWNT was <2.7 Å, or if the distancebetween any heavy atom of alcohols/phenol and any carbonatom of SWNT was <2.4 Å.38 The resulting concentrationsof aqueous solutions are also roughly 1 M, with the number of

FIG. 1. Snap view of initial simulation system with a (13, 13) SWNT sol-vated in 1 M aqueous ethanol for demonstration. The SWNT is representedby gray bonds; ethanol is displayed in ball and stick representations; water isrepresented by red lines.

organic and water molecules for different production simula-tions summarized in Table I.

We modeled the carbon atoms of SWNT as unchargedLennard-Jones particles and chose the commonly used setof van der Waals (vdW) parameters for carbon atom ofSWNT/graphene, i.e., a cross section of σ cc = 0.34 nmand a depth of the potential well of εcc = 0.086 kcalmol−1.10, 17, 18, 26, 27, 39, 40 We note that a previous MD simu-lation of SWNT immersed in methanol-water solution indi-cated that a slight modification of εcc (change the current εcc

to 0.056 kcal mol−1) did not qualitatively change the results(i.e., selective adsorption of methanol over water in SWNT).29

Carbon-carbon bond lengths of 0.14 nm and bond angles of120◦ were maintained by harmonic potentials with springconstants of 93 800 kcal mol−1 nm−2 and 126 kcal mol−1

rad−2, respectively.18, 39 In addition, a weak dihedral angle po-tential was applied to bonded carbon atoms. The positions ofcarbon atoms at the SWNT inlet and outlet were constrained

TABLE I. Number of alcohols/phenol (Norganics) and water (Nwater)molecules for different simulation systems.a

Wide SWNT Narrow SWNTsystems systems

Alcohols/phenol Norganics Nwater Norganics Nwater

Methanol 96 5198 45 2429Ethanol 97 5078 45 23851-propanol 96 5002 48 25031-butanol 96 4922 48 24431-pentanol 95 4765 49 23971-hexanol 96 4757 49 2367Glycerol 96 5006 48 2503Phenol 96 4939 48 2466

aThe wide single-walled carbon nanotube (SWNT) is the (13, 13) SWNT and the narrowSWNT is the (6, 6) or (7, 7) SWNT.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

130.39.62.90 On: Wed, 03 Sep 2014 02:47:15

204711-3 Tian et al. J. Chem. Phys. 138, 204711 (2013)

FIG. 2. Number of alcohols/phenol (shown in black) and water molecules (shown in red) within the (13, 13) SWNT as a function of simulation time. Inset:Corresponding structures of the alcohols/phenol-water mixtures within the (13, 13) SWNT in equilibrium (from an axial view), with alcohols/phenol representedby licorice and water by vdW spheres.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

130.39.62.90 On: Wed, 03 Sep 2014 02:47:15

204711-4 Tian et al. J. Chem. Phys. 138, 204711 (2013)

by using the position restraint. Other carbon atoms were leftfree to vibrate.

All MD simulations were performed using GROMCAS

4.0.741 in an NPT (298 K, 1 atm) ensemble with periodicboundary conditions applied in all directions. The constanttemperature and pressure were maintained using the v-rescalethermostat42 (with a coupling coefficient of τT = 0.5 ps) andthe Parrinello-Rahman pressure coupling scheme43 (with acoupling coefficient of τ p = 5 ps), respectively. The particle-mesh Ewald method44 with a real space cutoff of 1.0 nm wasused to treat long-range electrostatic interactions, whereas thevdW interactions were treated with a cutoff distance of 1.2nm. LINCS was applied to constrain all bonds. A time step of2.0 fs was used, and the data were collected every 1 ps. Thesimulation lengths were 150 ns for methanol-(6, 6) SWNTsystem and 100 ns for other systems.

III. RESULTS AND DISCUSSION

A. Alcohol-induced drying for wide SWNT systems

In the following, we show that the alcohols win the com-petitive adsorption over water inside SWNTs and induce dry-ing of SWNTs. First, we have calculated the number of sol-vent molecules (water/alcohols/phenol) inside the (13, 13)SWNTs during the course of MD simulations, as shown inFig. 2. Here, a solvent molecule is defined as being insidethe SWNT once its center of mass enters the SWNT. Thecorresponding snapshots of the alcohol-water mixtures in-side the nanotubes, after the systems have reached equilib-rium, are also shown in the insets of Fig. 2. Strikingly, in allcases except methanol, most water molecules initially insidethe SWNTs (the concentrations of alcohols/phenol inside theSWNTs are also roughly 1 M from the initial solvation setup)are expelled out from the SWNTs. Eventually, the SWNT in-teriors are dominantly occupied by the alcohols/phenol. Thedrying effects for C4-C6 1-alcohols and phenol are so strongthat almost all water molecules initially inside the SWNTsare expelled out after the systems have reached equilibrium.It should be noted that, in the case of methanol, althoughthe number of water molecules remaining inside SWNT isstill large, the ratio of number of methanol to water insideSWNT is several times larger than that in bulk (more below).That is, it can be viewed that methanol also induces drying ofSWNT and accumulates inside SWNT. In addition, we havecalculated the average number of alcohols/phenol and watermolecules within (13, 13) SWNTs in equilibrium. The resultsare summarized in Table II.

To quantify the drying effect and the selectivity ofSWNTs to different alcohols, we have calculated the “selec-tivity” of a SWNT to alcohols (or phenol)22, 30, 45 (i.e., the“drying factor” in our previous paper26), fselec, defined as fol-lows:

fselec = RSWNT/Rbulk, (1)

where RSWNT and Rbulk are the ratio of the average numberof alcohols/phenol to water in SWNT and that in the bulkregion, respectively. A larger fselec means a higher selectiv-ity (i.e., stronger drying effect) for alcohols/phenol. The com-

TABLE II. Average number of alcohols/phenol (〈Norganics〉) and watermolecules (〈Nwater〉) inside the (13, 13) SWNT in equilibrium, as well asthe selectivity of the (13, 13) SWNT to alcohols/phenol (fselec, see text fordetails), for different simulation systems.

Alcohols/phenol 〈Norganics〉 〈Nwater〉 fselec

Methanol 16.85 104.9 8.696Ethanol 32.66 39.91 42.841-propanol 32.25 16.96 99.081-butanol 29.83 5.206 293.61-pentanol 26.55 1.523 876.11-hexanol 23.53 1.276 910.9Glycerol 25.95 35.01 38.65Phenol 30.75 3.247 487.2

puted fselec are presented in Table II, and the relationship be-tween fselec and Ncarbon is plotted in Fig. 3. It is clear that thefselec increase as Ncarbon increases for 1-alcohols. More inter-estingly, when Ncarbon ≤ 5, the fselec can be fitted to an expo-nential function with respect to Ncarbon very well,

fselec = 3.60 × exp(Ncarbon/0.91) + 3.25. (2)

When Ncarbon ≥ 5, the fselec seems to be saturated, becauseat those times almost no water can enter the nanotube (theaverage number of water inside SWNT is smaller than 2).Also, note that the fselec for glycerol and phenol are consider-ably large (39 and 487, respectively), but these values are stillsmaller than those of 1-propanol and 1-hexanol (they have thesame Ncarbon as glycerol and phenol, respectively). The under-lying mechanisms will be discussed in Sec. III B.

B. Physical mechanisms of alcohol-induced dryingof SWNTs

The observed phenomenon of alcohol-induced drying ofSWNTs can be explained from an energetic perspective. Fol-lowing the pioneering work by Hua et al.38 wherein the pref-erential binding of urea over water around proteins is inter-preted by analyzing the interaction energy distributions of sol-vents, we have computed the average nonbonded interactions,

FIG. 3. Selectivity of the (13, 13) SWNT to alcohols/phenol (fselec) withrespect to the number of carbon atoms of alcohols/phenol (Ncarbon). Black,green, and blue squares represent 1-alcohols, glycerol, and phenol, respec-tively. The red curve is fitted with an exponential function.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

130.39.62.90 On: Wed, 03 Sep 2014 02:47:15

204711-5 Tian et al. J. Chem. Phys. 138, 204711 (2013)

FIG. 4. Distribution of interaction energies of a solvent molecule, using the case of the ethanol interacting with the (13, 13) SWNT for demonstration. (a) and(b) The probability distribution of van der Waals (vdW) and electrostatic energies, respectively, for an ethanol molecule inside the SWNT and in bulk, with therest of system. (c) and (d) Distribution of interaction energies for water.

including both vdW and electrostatic interactions, for a sol-vent (alcohols/phenol/water) molecule inside the SWNT andin bulk (defined as 1.2 nm away from any atoms of SWNT forany atoms of the solvent) with the rest of the system. Figure 4shows the interaction energy distributions of a solvent in (13,13) SWNT and in bulk (the data are averaged over the first10 ns of simulations), using the ethanol case for demonstra-tion (the distributions for other cases are similar). Comparedto a bulk ethanol, the energy distribution of a confined ethanolshifts rightward (peak position shifting ∼2.2 kcal/mol towardthe high energy region) in the electrostatic term whereas shiftsleftward (peak position shifting ∼5.0 kcal/mol toward thelow energy region) in the vdW term, with the leftward shift-ing more noticeable. There are similar shifting tendencies forwater moving from bulk to SWNT, but the leftward shiftingin vdW energy is much less noticeable than that of ethanol.

Thus, when a solvent moves from bulk region into SWNT,ethanol’s gains in vdW energy can (over)compensate the lossin electrostatic energy, whereas water’s gains in vdW energycannot, resulting in the binding preference of ethanol over wa-ter inside the SWNT.

Table III summarizes the differences in the average in-teraction energies for a solvent in the SWNT and in bulk. Inall cases, as a bulk alcohol (or phenol) moves into SWNT, itloses electrostatic interaction energy; however, the confinedalcohol (or phenol) gains considerable vdW interaction en-ergy due to its strong dispersion interaction with the nan-otube, which overcompensates the loss in electrostatic in-teraction. On the other hand, in all cases, a water moleculemoving from bulk to SWNT gains a few vdW energies butloses more electrostatic energies. Overall, the changes in to-tal interaction energies are negative for alcohols/phenol but

TABLE III. Differences in average interaction energies (in kcal/mol) for a solvent molecule in the (13, 13) SWNT vs. in bulk.a

Alcohols/phenol �Eototal �Eo

elec �EovdW Eoc �Ew

total �Ewelec �Ew

vdW Ewc

Methanol −1.62 1.75 −3.37 −4.59 0.48 1.76 −1.28 −1.36Ethanol −2.95 1.99 −4.94 −6.51 0.63 1.97 −1.34 −1.251-propanol −4.28 2.62 −6.90 −8.54 0.68 2.05 −1.37 −1.161-butanol −5.97 3.17 −9.14 −10.78 0.75 2.17 −1.42 −1.181-pentanol −7.61 3.36 −10.97 −12.67 0.62 1.98 −1.36 −1.301-hexanol −9.27 2.95 −12.22 −15.48 0.64 2.01 −1.37 −1.29Glycerol −3.88 6.11 −9.99 −10.70 0.13 1.56 −1.43 −1.27Phenol −7.33 3.55 −10.88 −12.98 0.80 2.30 −1.50 −1.15

a�E?total , �E?

elec , and �E?vdW denote the differences in total, electrostatic, and vdW interaction energies, respectively, for a solvent molecule, where the superscript ? is o for organics

(alcohols/phenol) and w for water. Eoc and Ewc denote vdW interaction energies of an organic molecule and a water molecule, respectively, with carbon atoms of SWNT. These dataare averaged over the first 10 ns of simulations.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

130.39.62.90 On: Wed, 03 Sep 2014 02:47:15

204711-6 Tian et al. J. Chem. Phys. 138, 204711 (2013)

positive for water when penetrating into SWNTs, which isresponsible for the alcohols’ preferential binding over waterinto the SWNT. Thus, this interaction-energy-decompositionanalysis shows higher affinities of alcohols/phenol than waterto the SWNTs as a result of stronger dispersion interactionsof alcohols/phenol than water with hydrophobic nanotubes,consistent with previous MD studies10, 46 which also demon-strated that the attractive dispersion interaction is a very im-portant factor for the occurrence of drying transition in thenano-sized confined region. It should be noted that the re-placement of structurally confined water molecules by largeralcohols/phenol inside the SWNT is also favorable becauseof an overall solvent entropy gain. Indeed, our previous studyof urea-induced drying of SWNTs shows that the interaction-energy-decomposition analysis is consistent with the free-energy-decomposition analysis.26

Now we discuss the mechanisms behind the findings inFig. 3 according to the calculated energies in Table III. First,Fig. 3 shows that the selective adsorption of 1-alcohols intoSWNT becomes more profound as Ncarbon increases. This ten-dency is due to the fact that the dispersion interactions of alco-hols with SWNT, which dominate the total energetic changeswhen bulk alcohols penetrate into SWNT, increase with theNcarbon (see Table III). More interestingly, the absolute val-ues of �Eo

total (and Eoc) in Table III almost increase lin-early with respect to Ncarbon (see Fig. S1 in the supplementarymaterial52); this nearly linear tendency may correlate with theexponential function of fselec with Ncarbon when Ncarbon ≤ 5shown in Fig. 3 (see the related discussions in the supplemen-tary material52). Second, fselec of glycerol is smaller than thatof 1-propanol which has the same Ncarbon as the glycerol. Thereason is that glycerol has more hydroxyl functional groupsthan 1-propanol; when it moves from bulk into SWNT, its lossin electrostatic energy is much larger than that of 1-propanol

(although the additional hydroxyl groups gains more vdW en-ergy, this cannot compensate the loss in electrostatic energy).Third, somewhat unexpectedly, we find that fselec of 1-hexanolis larger than phenol which has the same Ncarbon as 1-hexanolbut enables π -π stacking interaction40, 47, 48 with the aromaticrings of nanotube. This may be due to the steric effect – phe-nol has a relatively larger width in the transverse directionthan 1-hexanol (as compared to the shorter length in the lon-gitudinal direction), so not all the inner phenol molecules canbenefit from the favorable π -π stacking interactions insidethe SWNT. In contrast, 1-hexanol possesses a linear, flexiblestructure, which can better fit the confined region of SWNT,thus enabling strong dispersion interactions with SWNT andwith themselves (the absolute values of Eoc, �Eo

vdW , and�Eo

total of 1-hexanol are all larger than those of phenol; seeTable III).

C. Alcohol-induced drying for narrow SWNT systems

In addition to wide SWNT systems, we have also in-vestigated the drying phenomenon for narrow [(6, 6) and(7, 7)] SWNTs. Here the “narrow SWNT” means the nar-rowest armchair SWNT that the alcohols/phenol can pene-trate into. Our simulations show that the “narrow SWNTs” are(6, 6) SWNT for methanol and ethanol and (7, 7) SWNT forother organic molecules used here. Similar to the situations ofwide SWNTs, almost all water molecules inside the SWNTsare replaced by alcohols/phenol within the first 50 ns (Fig. S2in the supplementary material52 displays the drying pro-cesses). After the systems have reached equilibrium, the inneralcohols/phenol exhibit single-file arrays, forming “perfect”molecular wires, or occasionally, forming “defective” molec-ular wires with a very small number of “water defect(s)”(see Fig. 5). In our previous paper,27 we systematically

FIG. 5. Snapshots of 1D molecular wires formed by alcohols/phenol (some with “water defect(s)”) inside narrow [(6, 6) or (7, 7)] SWNTs when the systemshave reached equilibrium.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

130.39.62.90 On: Wed, 03 Sep 2014 02:47:15

204711-7 Tian et al. J. Chem. Phys. 138, 204711 (2013)

TABLE IV. Average number of alcohols/phenol (〈Norganics〉) and watermolecules (〈Nwater〉) inside narrow [(6, 6) or (7, 7)] SWNTs in equilibrium,as well as the selectivity of narrow SWNTs to alcohols/phenol (fselec, see textfor details), for different systems.

Alcohols/phenol 〈Norganics〉 〈Nwater〉 fselec

Methanol 7.55 0.288 1416.5Ethanol 5.97 0.038 8382.51-propanol 6.58 0.003 1143731-butanol 4.84 2.003 123.061-pentanol 4.47 1.000 218.801-hexanol 3.99 1.000 192.72Glycerol 5.78 1.009 298.58Phenol 5.43 0.019 14757

investigated the structural and dynamical properties of 1Durea wire confined within the narrow SWNT. In that case,the urea wire is formed with a contiguous hydrogen-bondednetwork which is somewhat different from the currently pre-sented cases. For future studies, it is interesting to explorethe unique properties of these 1D alcohols/phenol wires andfind their potential applications such as the electronic de-vices for the signal transduction and multiplication at thenanoscale.17, 49, 50

Table IV lists the average number of alcohols/phenol(〈Norganics〉) and water molecules (〈Nwater〉) inside the narrow[(6, 6) or (7, 7)] SWNTs after the systems have reached equi-librium. The number of remaining water inside the narrowSWNTs is very small, resulting in very large fselec, particularlyfor phenol and 1-alcohols with Ncarbon ≤ 3. For 1-alcohols,as the Ncarbon increases, the selectivity of narrow SWNTs to1-alcohols increases and reaches a maximum when Ncarbon

= 3. It is noteworthy that for short-chain linear alcohols(Ncarbon ≤ 3), fselec for narrow SWNT are larger than thosefor wide [(13, 13)] SWNT; but for long-chain linear alco-hols (Ncarbon > 3), fselec for narrow SWNT are smaller thanthose for wide [(13, 13)] SWNT (see Tables II and IV).These interesting tendencies can be interpreted in terms ofthe competition between enthalpy (interaction energy) andentropy: for short-chain linear alcohols, the alcohols insidenarrow SWNTs can gain more vdW interaction energies withthe nanotube wall compared to those inside wide SWNT, sothe corresponding fselec are larger; when Ncarbon > 3, how-ever, the penalty in the entropy of linear alcohols inside nar-row SWNTs becomes significant and dominates the free en-ergy, resulting in relatively smaller fselec as compared to wideSWNT cases. This interpretation can be further validated bythe evidence that the fselec for phenol (it has the same Ncarbon

as 1-hexanol, but its penalty in entropy when penetrating intothe narrow SWNT is smaller than linear alcohols) is muchlarger than fselec for 1-hexanol inside narrow SWNT (see Ta-ble IV). These findings suggest that for short-chain linear al-cohols such as methanol, ethanol, 1-propanol, and glycerol,we can use the narrow SWNTs to achieve the highest selec-tivity; while for long-chain linear alcohols, it is better to usewide SWNTs to achieve efficient alcohol/water separation.

IV. CONCLUSIONS

In this study, we employ MD simulations to investigatethe competitive binding/adsorption between various alcohols(or phenol) and water inside SWNTs and unveil the underly-ing physical mechanisms. We observe that even at a low alco-hols/phenol concentration (∼1 M), all kinds of alcohols andphenol can induce drying of nanotubes and accumulate insideboth wide [(13, 13)] and narrow [(6, 6) or (7, 7)] SWNTs.In particular, for narrow SWNTs, most kinds of alcohols andphenol form nearly uniform 1D molecular wires under theextreme confinement of SWNTs. Our energetic analyses in-dicate that the preferential binding of alcohols over waterinside SWNTs results from the stronger dispersion interac-tions of alcohols with SWNTs than water. We find that forthe wide SWNT, the selectivity for 1-alcohols increases asNcarbon increases with an exponential function when Ncarbon

≤ 5. For narrow SWNTs, the favorable dispersion interactionsof alcohols/phenol with SWNTs lead to a very high selec-tivity for C1-C3 1-alcohols and phenol; the competition be-tween enthalpy and entropy results in a maximal selectivityfor 1-alcohols when Ncarbon = 3. By comparing the selectiv-ity of SWNTs to alcohols between wide and narrow SWNTs,we suggest that to achieve the highest efficiency of alco-hol/water separation, it is better to use narrow SWNTs to sep-arate short-chain linear alcohols, while using wide SWNTs toseparate long-chain linear alcohols. For practical applications,SWNT bundles15 or membranes51 can be used to realize thealcohol/water separation. In view of a previous Monte Carlosimulation30 which indicates that the separation performanceof SWNTs with appropriate diameters is better than zeolites,we expect that SWNT may be a more promising adsorbentthan the widely used zeolites for alcohol/water separation.Our findings hold promise for the potential use of SWNT bun-dles/membranes for efficiently extracting alcohols (and otherhydrophilic organic molecules) from their aqueous solutions.

ACKNOWLEDGMENTS

We thank Professor Xiaowei Tang and Dr. Xiuping Renfor their insightful suggestions. This work is supported by theNational Natural Science Foundation of China (NSFC) (GrantNos. 11204269 and 11105088), Zhejiang Provincial NaturalScience Foundation of China (Grant No. LY12A04007), theFundamental Research Funds for the Central Universities, andthe KYLIN-I Supercomputer in Institute for Fusion Theoryand Simulation, Zhejiang University.

1T. Remy, J. C. Saint Remi, R. Singh, P. A. Webley, G. V. Baron, and J. F.M. Denayer, J. Phys. Chem. C 115, 8117 (2011).

2S. I. Nakao, F. Saitoh, T. Asakura, K. Toda, and S. Kimura, J. Membr. Sci.30, 273 (1987).

3P. Kaewkannetra, N. Chutinate, S. Moonamart, T. Kamsan, and T. Y. Chiu,Desalination 271, 88 (2011).

4K. Sunitha, S. V. Satyanarayana, and S. Sridhar, Carbohydr. Polym. 87,1569 (2012).

5D. Nimcevic and J. R. Gapes, “The acetone-butanol fermentation in pilotplant and pre-industrial scale,” J. Mol. Microbiol. Biotechnol. 2, 15 (2000).

6N. Qureshi, S. Hughes, I. S. Maddox, and M. A. Cotta, Bioprocess Biosyst.Eng. 27, 215 (2005).

7K. Zhang, R. P. Lively, J. D. Noel, M. E. Dose, B. A. McCool, R. R.Chance, and W. J. Koros, Langmuir 28, 8664 (2012).

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

130.39.62.90 On: Wed, 03 Sep 2014 02:47:15

204711-8 Tian et al. J. Chem. Phys. 138, 204711 (2013)

8R. H. Zhou, X. H. Huang, C. J. Margulis, and B. J. Berne, Science 305,1605 (2004).

9P. Liu, X. H. Huang, R. H. Zhou, and B. J. Berne, Nature (London) 437,159 (2005).

10N. Choudhury and B. M. Pettitt, J. Am. Chem. Soc. 129, 4847 (2007).11Z. X. Yang, B. Y. Shi, H. J. Lu, P. Xiu, and R. H. Zhou, J. Phys. Chem. B

115, 11137 (2011).12X. Li, J. Y. Li, M. Eleftheriou, and R. H. Zhou, J. Am. Chem. Soc. 128,

12439 (2006).13M. G. Krone, L. Hua, P. Soto, R. H. Zhou, B. J. Berne, and J.-E. Shea,

J. Am. Chem. Soc. 130, 11066 (2008).14D. Tasis, N. Tagmatarchis, A. Bianco, and M. Prato, Chem. Rev. 106, 1105

(2006).15F. J. A. L. Cruz and J. P. B. Mota, Phys. Rev. B 79, 165426 (2009).16Q. Shao, L. L. Huang, J. Zhou, L. H. Lu, L. Z. Zhang, X. H. Lu, S. Jiang,

K. E. Gubbins, Y. D. Zhu, and W. F. Shen, J. Phys. Chem. C 111, 15677(2007).

17Y. S. Tu, P. Xiu, R. Z. Wan, J. Hu, R. H. Zhou, and H. P. Fang, Proc. Natl.Acad. Sci. U.S.A. 106, 18120 (2009).

18P. Xiu, B. Zhou, W. P. Qi, H. J. Lu, Y. S. Tu, and H. P. Fang, J. Am. Chem.Soc. 131, 2840 (2009).

19P. Das, Nanoscale 4, 2931 (2012).20B. Corry, J. Phys. Chem. B 112, 1427 (2008).21Z. G. Mao and S. B. Sinnott, J. Phys. Chem. B 105, 6916 (2001).22L. Huang, L. Zhang, Q. Shao, L. Lu, X. Lu, S. Jiang, and W. Shen, J. Phys.

Chem. C 111, 11912 (2007).23J. Lee and N. R. Aluru, Appl. Phys. Lett. 96, 133108 (2010).24A. Kalra, G. Hummer, and S. Garde, J. Phys. Chem. B 108, 544 (2004).25P. Das and R. H. Zhou, J. Phys. Chem. B 114, 5427 (2010).26P. Xiu, Z. X. Yang, B. Zhou, P. Das, H. P. Fang, and R. H. Zhou, J. Phys.

Chem. B 115, 2988 (2011).27P. Xiu, Y. S. Tu, X. L. Tian, H. P. Fang, and R. H. Zhou, Nanoscale 4, 652

(2012).28Y. Liu, S. Consta, and W. A. Goddard, J. Nanosci. Nanotechnol. 10, 3834

(2010).29W. H. Zhao, B. Shang, S. P. Du, L. F. Yuan, J. L. Yang, and X. C. Zeng,

J. Chem. Phys. 137, 034501 (2012).30L. H. Lu, Q. Shao, L. L. Huang, and X. H. Lu, Fluid Phase Equilib. 261,

191 (2007).31S. P. Du, W. H. Zhao, and L. F. Yuan, Chin. J. Chem. Phys. 25, 487 (2012).

32S. Cabani, P. Gianni, V. Mollica, and L. Lepori, J. Solution Chem. 10, 563(1981).

33D. P. Kharakoz, J. Solution Chem. 21, 569 (1992).34W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey, and M. L.

Klein, J. Chem. Phys. 79, 926 (1983).35W. L. Jorgensen, D. S. Maxwell, and J. TiradoRives, J. Am. Chem. Soc.

118, 11225 (1996).36J. Lehtola, M. Hakala, and K. Hamalainen, J. Phys. Chem. B 114, 6426

(2010).37 A. Vrhovšek, O. Gereben, A. Jamnik, and L. Pusztai, J. Phys. Chem. B

115, 13473 (2011).38L. Hua, R. H. Zhou, D. Thirumalai, and B. J. Berne, Proc. Natl. Acad. Sci.

U.S.A. 105, 16928 (2008).39G. Hummer, J. C. Rasaiah, and J. P. Noworyta, Nature (London) 414, 188

(2001).40Z. Yang, Z. Wang, X. Tian, P. Xiu, and R. Zhou, J. Chem. Phys. 136,

025103 (2012).41B. Hess, C. Kutzner, D. van der Spoel, and E. Lindahl, J. Chem. Theory

Comput. 4, 435 (2008).42G. Bussi, D. Donadio, and M. Parrinello, J. Chem. Phys. 126, 014101

(2007).43M. Parrinello and A. Rahman, J. Appl. Phys. 52, 7182 (1981).44T. Darden, D. York, and L. Pedersen, J. Chem. Phys. 98, 10089 (1993).45Z. Tan and K. E. Gubbins, J. Phys. Chem. 96, 845 (1992).46R. Zangi, R. Zhou, and B. J. Berne, J. Am. Chem. Soc. 131, 1535 (2009).47S. Tsuzuki, K. Honda, T. Uchimaru, M. Mikami, and K. Tanabe, J. Am.

Chem. Soc. 124, 104 (2002).48C. C. Ge, J. F. Du, L. N. Zhao, L. M. Wang, Y. Liu, D. H. Li, Y. L. Yang,

R. H. Zhou, Y. L. Zhao, Z. F. Chai, and C. Y. Chen, Proc. Natl. Acad. Sci.U.S.A. 108, 16968 (2011).

49Y. S. Tu, R. H. Zhou, and H. P. Fang, Nanoscale 2, 1976 (2010).50Y. S. Tu, H. J. Lu, Y. Z. Zhang, H. Tien, and R. H. Zhou, J. Chem. Phys.

138, 015104 (2013).51A. Kalra, S. Garde, and G. Hummer, Proc. Natl. Acad. Sci. U.S.A. 100,

10175 (2003).52See supplementary material at http://dx.doi.org/10.1063/1.4807484 for dis-

cussions on interaction energies with respect to Ncarbon and the physi-cal mechanism behind the exponential function of fselec with Ncarbon for1-alcohols, and some figures to show the drying processes for narrowSWNT systems.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

130.39.62.90 On: Wed, 03 Sep 2014 02:47:15