Table Of Content

Graphene oxide and adsorption of chloroform: a density functional study Elena Kuisma, C. Fredrik Hansson, Th. Benjamin Lindberg, Christoffer A. Gillberg, Sebastian Idh, and Elsebeth Schrödera) Quantum Device Physics Laboratory, Microtechnology and Nanoscience (MC2), Chalmers University of Technology, SE-412 96 Go¨teborg, Sweden (Dated: 28 January 2016) Chlorinatedhydrocarboncompoundsareofenvironmentalconcerns,sincetheyaretoxictohumansandother mammals, are widespread, and exposure is hard to avoid. Understanding and improving methods to reduce the amount of the substances is important. We present an atomic-scale calculational study of the adsorption of chlorine-based substance chloroform (CHCl ) on graphene oxide, as a step in estimating the capacity of 3 6 graphene oxide for filtering out such substances, e.g., from drinking water. The calculations are based on 1 density functional theory (DFT), and the recently developed consistent-exchange functional for the van der 0 Waals density-functional method (vdW-DF-cx) is employed. We obtain values of the chloroform adsorption 2 energy varying from roughly 0.2 to 0.4 eV per molecule. This is comparable to previously found results for n chloroform adsorbed directly on clean graphene, using similar calculations. In a wet environment, like filters a fordrinkingwater,thegraphenewillnotstaycleanandwilllikelyoxidize,andthusadsorptionontographene J oxide, rather than clean graphene, is a more relevant process to study. 7 2 Keywords: graphene oxide, chloroform, vdW-DF, vdW-DF-cx, van der Waals, DFT, adsorption, water filtering, water cleaning ] h p - I. INTRODUCTION stances,bycalculatingthebindingenergyanditsdepen- m dence on the structure of GO. e Graphite oxide was first synthesized more than 150 GOhaspreviouslybeenstudied6–11inexperimentsand h yearsago1butcaughtgeneralinterest2–16onlyduringthe by use of calculational tools, including DFT. The GO it- .c past few decades when research in 2D materials, in par- self is expected to be reasonably well described12 by use s ticulargraphene,hasstartedtobloom. Graphiteoxideis of semilocal approximations of the exchange and corre- c i analternativepathtolarge-scaleproductionofgraphene, lation of DFT, such as in the PBE approximation,19 but s by liquid-phase exfoliation of graphite oxide into layers, for our subsequent studies of chloroform physisorption y h calledgrapheneoxide(GO),andsubsequentreductionto it is imperative that the dispersive nonlocal interactions p graphene,2–4 but already the GO sheets have intriguing be included in a consistent way. Therefore we here use [ andusefulfeatures. GOcanbeunderstoodasfunctional- thevanderWaals(vdW)densityfunctionalmethod20–24 izedgraphene,oxidizedwithhydroxyl,epoxideandsome (vdW-DF), in the vdW-DF-cx version,23,25 for all cal- 1 v carboxyl groups. Its properties depend on the details of culations except for a comparison with previous results, 9 the oxidation: the type of, the number of and the distri- where we use PBE for some calculations. 2 butionofthefunctionalgroups. GOhastuneableelectric Chloroform with graphitic or other carbon based ma- 3 properties, obtained by changing the functional groups, terialswaspreviouslystudiedinacoupleofexperimental 7 and with its thin size could be used for electronics.5 GO andcomputationalstudies.26–30Also,aDFTbasedstudy 0 is highly catalytic, highly solvable in water and other ofammoniaadsorptionwaspresentedearlier.31 However, . 1 solvents, and is proposed for use as a gas sensor.14 to our knowledge there are no previously DFT studies 0 GOhasbeensuggestedasamaterialforuseinfiltering of chloroform adsorption on GO using methods that in- 6 of toxic compounds,15,16 such as chlorinated hydrocar- clude the vdW interactions consistently, such as here. 1 bon compounds. These are some of today’s environmen- Certainly, physisorption of chloroform on GO with the : v tal concerns, since they are toxic to both humans and recent vdW-DF-cx has not previously been covered. Xi other mammals, and exposure is hard to avoid. Expo- This article is structured as follows: In Section II we sure to chlorine-based compounds arises, e.g., from con- describe the method of computation and the systems r a sumption of chlorinated drinking water or food supplies studied. In Section III the results are presented and dis- that have been contaminated by residues of industrial cussed, along with a discussion of the accuracy of our chemicals.17,18 calculations, and Section IV summarizes the study. We present a computational study of GO with state- of-the-art calculations, using a recent implementation of density functional theory (DFT). We study how GO II. PHYSICAL SYSTEM AND COMPUTATIONAL bindschloroform,oneofthecommonchlorine-basedsub- METHOD In the following we describe GO and its functional groupsandhowwemodeltherelevantpartsofGOinour a)Electronicmail: [email protected] adsorptionstudy. Wealsodescribethemethodsusedfor 2 the DFT calculations and define the binding energies for the functional groups in GO and for chloroform on GO, including a discussion of what constitutes the zero point of the binding energies in these calculations. This varies in the literature, making comparison difficult. A. GO and chloroform GO has a graphene carbon network with an almost amorphousdistributionoffunctionalgroups,anditisdif- ficulttodeterminethetypesoffunctionalgroupspresent and their positions. In Ref. 6 Lerf et al. used reactions withvariousreagentstosupplementtheirpreviousNMR measurements. They found that graphite oxide (and thus likely also GO) mainly has two types of functional FIG. 1. Graphene slab. The unit cell (area delimited by the groups: OboundintheC-Cbridgesite,whichisthe1,2- black solid line) used in this study consists of 3×6 conven- etherorepoxide,andtheC-OHorhydroxylgroup. They tionalrectangulargraphenecells,eachcontainingfourcarbon also found no support for O bound to two next-nearest atoms, double the size of the primitive unit cell (blue dashed line). Thelengthsofunitvectors(redsolidarrows)ofthecon- neighbor C atoms (the 1,3-ether) and very little for the √ ventional cell are a=2.456 ˚A and b=a 3=4.254 ˚A. Visu- carboxyl (the -COOH) group. Based on their measure- alization(hereandinFigures3,4and6)usingXCrySDen.33 ments they put forward a structural model for GO that has areas without functional groups (i.e., areas of clean graphene) and other areas with epoxide and hydroxyl sorbed molecule because these groups were supposedly groups randomly distributed but close together. In their the ones that influence the chloroform-GO binding the model the carbon grid of GO is almost flat, except for most. However, we see that also groups on the other the parts where C atoms are attached to hydroxyl. The side of the carbon grid affect the adsorption energy and GO has functional groups on both sides of the carbon ourstudyhasbeenenlargedtoenclosealsosystemswith gridandcarboxylgroupsareonlypresentattheedgesof functional groups on both sides. GO. In experiments, fully oxidized GO is found to have a C:O ratio approximately 2:1 or more.32 However, GO B. Unit cell is not always found in the fully oxidized state. Even thoughtheGOmodelofLerfetal.6 hasarelativelyhigh concentration of functional groups, the C:O ratio is only To study GO we start out with a graphene slab with about 5:1 in the areas that are not part of the GO sheet add√ed functional epoxide and hydroxyl groups. We use edges. a 3 3 × 6 orthorhombic unit cell with 72 graphene C InthepresentstudyofGOweconsiderstructureswith atoms, as illustrated in Figure 1, with a C-C distance clustersoffunctionalgroupsonotherwisecleangraphene, of 1.424 ˚A, and periodic boundary conditions. The unit the clusters being relatively small and disordered. Thus, cell height is varied such that the amount of vacuum be- we compute the structures and energies involved in the tween each copy of the system is approximately 10 ˚A, formation of epoxide and hydroxyl groups on graphene thus for clean GO calculations the unit cell height is forGOwithlowOconcentration(C:Oratiofrom72:1to 10.5 ˚A, and approximately 15.5 ˚A for physisorption of 15:1). We use GO with functional groups on either both chloroform on GO. In the calculations of adsorbed chlo- sides of the carbon grid (not symmetrically positioned) roform the molecule-molecule nearest-neighbor distance or one side with just a few clustered groups only. In our is12.8˚A(theunitcellwidth)andthesmallestlateraldis- study we use periodic boundary conditions in space and tance between any two atoms in neighboring chloroform thustheGOhasnoedges. Thismeansthataccordingto molecules is 10.4 ˚A, as illustrated in Figure 2. themodelbyLerfetal.thereshouldnotbeanycarboxyl groups included. Chloroform(CHCl )consistsofacentralcarbonatom C. Methods of computation 3 with three electronegative Cl atoms in a ‘tripod’ in one endandaHatomattheotherend. Thisgiveschloroform The formation of GO from graphene is a process a finite dipole moment, which affects its physisorption in which hydroxyl and epoxide groups chemisorb on properties. graphene. Such processes are expected to be well de- In our chloroform adsorption study we initially stud- scribed with a semilocal density functional like PBE.19 ied mainly (but not exclusively) GO structures that However, the adsorption of chloroform on GO is a physi- have all functional groups on the same side as the adsorption process and in such processes a robust de- 3 semilocal PBE functional. For the vdW-DF-cx results and most of the PBE re- sultsweuseQuantum Espresso35,36 (QE).Forfurther comparison we perform additional PBE calculations using the GPAW37,38 software. All calculations are carried out self-consistently. In the QE calculations we use ultrasoft pseudo- potentials39,40 with wavefunction and density cut-off en- ergies30and120Ry,respectively. Theforceconvergence thresholdvalueissetto2meV˚A−1,andthenumberofk- pointsis4×4×1(and1×1×1forsmallmolecules). With GPAWweusePAWsetups41,42intheGPAWstandard.43 The binding energy of the functional groups on graphene and of chloroform on GO, is calculated as E =−(E −E −E ), (1) b AB A B where E is the total energy of the full system and AB E and E are the total energies of its individual con- A B stituents (positive E for systems that bind). The zero b pointforthebindingenergydependsonthechoiceofin- dividualconstituentsinsubsystemAandB,e.g.,whether to use an isolated O atom or the O molecule as one of 2 the subsystems. These choices vary among authors of such studies. Our choice is discussed below. Spin polarization was not implemented in vdW-DF-cx atthetimeofourcalculations, andthusweapproximate the binding energy of an epoxide on graphene as44,45 Ecx =−(Ecx −Ecx−Ecx) (2) b GO G 3O ≈−(cid:0)Ecx −Ecx−(cid:0)EPBE−EPBE+Ecx(cid:1)(cid:1),(3) FIG.2. Sketchoftheunitcellusedforcalculationsofchloro- GO G 3O 1O 1O formadsorptiononGO.Thesizeoftheunitcell(whitebox)is where E and E are the total energies of GO and approximately12.8×14.8×15.5˚A.Colorlegend: Chloroform GO G graphene, and E and E are the energies of triplet has yellow C; lime Cl; blue H, and GO has gray C; red O. 3O 1O (3O) and singlet (1O) O atoms, respectively, and the Closest atom-atom distances between chloroform molecules superscript denotes the functional used, vdW-DF-cx or are indicated, d = 10.4 ˚A and d = 11.9 ˚A. Visualization 1 2 using VMD.34 PBE. Thus, the difference in O singlet and triplet total energy in vdW-DF-cx is approximated by the same dif- ferenceinPBE,Ecx −Ecx ≈EPBE−EPBE.Thebinding 3O 1O 3O 1O scription of the dispersive interactions is necessary. We energies of the hydroxyl groups are calculated in a sim- therefore perform the main DFT calculations using the ilar manner, whereas the chloroform-on-GO calculations vdW-DF method20–24 in the consistent exchange vdW- are carried out without considering spin, with a non- DF-cx version.23,25 The vdW-DF-cx functional has been spinpolarized calculation of chloroform and a graphene shown to work well for layered structures and aromatic calculation as the two individual constituents. molecules, and it accurately predicts the a and c lattice In the results section we report the height h of the constantsofsolidgraphite.23 Itthusprovidesabalanced chloroform molecule above GO in the adsorbed position. descriptionbetweenthechemicalsp2 bondingwithinthe The height is taken as the projection in the z-direction graphenesheetsandthevdWinteractionsbetweenlayers (i.e., perpendicular to graphene) of the distance between and in physisorption. the chloroform C atom and the nearest GO surface O atom. This means that an atom in chloroform may be To describe the GO used as a substrate it is impor- closer to the GO O-atom than the distance h. tantthatthemethodweusecanalsohandlethebalance between sp2 and sp3 binding, in the graphene patches and at the sites of the functional groups, respectively. III. RESULTS AND DISCUSSION The fact that vdW-DF-cx shows reasonable results for thephasetransistionbetweendiamondandgraphite,sp3 andsp2 materials,isencouraging.23 Tofurtherdocument A. Oxidized graphene the ability of vdW-DF-cx for this problem we compare formation energies for a number of unsaturated GO con- Wefocusonlightlyoxidizedgraphene,withonlyafew figurations, obtained with both the vdW-DF-cx and the functional groups per 72 C-atom unit cell. We use 10 4 different configurations with up to three epoxide groups TABLEI.BindingenergyE andbindingenergyperOatom and up to two hydroxyl groups per unit cell, as illus- b E /Oofepoxide(O)andhydroxyl(OH)groupsongraphene, trated in the left hand side of Figure 3. The functional b calculatedusingthevdW-DF-cxandPBEfunctionals. Ener- groups are placed such that they form clusters, as ex- gies in units of eV. See Figure 3 for the systems GO#. pectedfromexperiments.6Theresultingbindingenergies are presented in the graph on the right hand side and in GO# GO struct. Ecx Ecx/O EPBE EPBE/O b b b b Table I. Structures GO1–3 and GO6–9 have groups on GO1 O 2.341 2.341 2.166 2.166 one side of the graphene plane only, structures GO4–5 GO2 OH 0.996 0.996 0.769 0.769 have one epoxide on each side, and structure GO10 has GO3 2O 5.187 2.594 4.844 2.422 epoxide and hydroxyl groups on both sides. The C–O GO4 2O 5.301 2.652 4.937 2.469 lengths for the epoxides and hydroxyl groups are 1.47 ˚A and 1.52 ˚A with the vdW-DF-cx functional. GO5 2O 4.190 2.095 3.831 1.916 GO6 2OH 2.956 1.478 2.495 1.247 By studying the binding energy E for each structure b GO7 2O, OH 6.620 2.207 6.042 2.014 wefindthatE hasanalmostlineardependenceonnum- b berofCatomsinvolvedinbindingthefunctionalgroups: GO8 O, 2OH 4.835 1.612 4.189 1.396 the right hand side of Figure 3 shows that for every C GO9 3O, 2OH 10.583 2.117 9.592 1.918 atom involved (meaning there will be one less sp2-bound GO10 3O, 2OH 11.462 2.292 10.412 2.083 Catom)E increasesbyroughly1.5eV.Forcomparison, b one single hydroxyl group has a binding energy of about 1eVandtheepoxidegroup,withitstwoC-Obonds,has sorption). We find that the vdW-DF-cx formation ener- a binding energy of about 2.3 eV, so the gain in adding gies are systematically stronger than those of PBE, with afunctionalgrouptoaclusterissignificantlyhigher,per upto23%differenceinformationenergies. However,the C-Obond,thanjustaddingthegrouptoapatchofclean difference depends on the type of functional group(s) in- graphene. There is some spread in the numbers, and the volvedintheGOstructure: forepoxidesthedifferencein illustrations of the structures show that the clusters of formation energy of vdW-DF-cx compared to PBE is 7– functional groups are not all densely packed, leading to 9%,wherasthedifferenceis16–23%forhydroxylgroups, lessgaininenergyforsparseclustersthanformoredense and mixed structures in the range 9–13%. clusters. It is important to note that these differences in vdW- Thepreferenceforhavingfunctionalgroupsinclusters DF-cx and PBE formation energies include changes to isseenalreadyinaclusteroftwohydroxylgroups: wesee the positions of the atoms when changing functional be- a huge gain in energy (almost 1 eV) by pairing hydroxyl tweenvdW-DF-cxandPBEandsubsequentlystructually groups,insteadofhavingthemseparated. Thisissoeven relaxing the atomic positions. The hydroxyl functional whenthegroupsareonthesamesideofthegraphenegrid group has an H atom pointing away from graphene, this andgiverisetomoredistortionofthegraphenethanthe makeslong-rangeinteractionsmorerelevantforhydroxyl distortion created by one single hydroxyl group (struc- thatfortheepoxide. Thesingle-bondedHatomalsohas ture GO6 compared to GO2 times 2). a less stiff binding, and small changes in the forces on More generally, by visually comparing all of the struc- the atoms (from change of functional) can more easily tures in the left hand side of Figure 3 we find that struc- move the H atoms than the more stiffly bound O atoms. tures with groups on both sides (GO4, GO5, and GO10) Thesedifferencesbetweenthehydroxylandepoxidefunc- have a less distorted carbon structure than structures tional groups are likely at least part of the reason for with groups only on one side (all others), as expected.9 the larger PBE to vdW-DF-cx energy difference when Table II shows that for structures with two epoxides hydroxyl groups are involved: vdW-DF-cx may actu- (GO3, GO4, and GO5) an energy gain can be obtained ally turn out to describe those groups better than PBE! both by clustering, with gain per O atom 0.36 eV (from However, without experiments or high-quality quantum comparing GO1 and GO4), and by having the epoxide chemistry calculations to compare to we cannot claim groups on both sides of the graphene instead of one side that this is the case. (gain 0.11 eV) but also that the largest gain is obtained by having the epoxides as nearest neighbors as opposed to next-nearest neighbors (gain 1.21 eV). This is seen B. Adsorption of chloroform independentofDFTmethod(vdW-DF-cxandPBE)and computational code (QE and GPAW) and can also be The main goal is to examine how chloroform binds to seen in literature values.46 GO. Our focus is on studying the effect on the binding Returning to the full set of formation energy results, energy of the presence of and the positions of the epox- Table I, we can compare the results of PBE calculations ide and hydroxyl groups and the relative orientation of to vdW-DF-cx calculations. The PBE calculations are the chloroform molecule. Since GO is almost amorphous expected to get both the sp2 and sp3 binding of the C an exhaustive search is prohibitive. Instead only the few atoms reasonably correct (but not so for the long-range functional groups closest to the chloroform molecule are interactions, which are important in chloroform physi- examined, keeping the rest of the unit cell clean of func- 5 12 GO10 GO9 10 slope -1.5 eV/atom 8 V] GO7 [eb 6 GO4 E GO3 GO8 4 GO5 GO6 GO1 2 GO2 0 62 64 66 68 70 72 Number of sp2 C atoms FIG. 3. Left: Epoxide and hydroxyl groups on graphene. The top and side views of optimized geometries of systems GO1–10. Only a part of the unit cell is shown. Color legend for atoms: yellow C; red O; blue H. Right: The corresponding binding energies, as also presented in Table I, as a function of C atoms in the unit cell not bound to functional groups, i.e., number of remaining sp2 C atoms of the initial 72 C atoms in the clean graphene. The method vdW-DF-cx is used for the structures shown here. Overalltheadsorptionenergyliesapproximatelyinthe TABLE II. Binding energies E and energy differences for b range0.2–0.4eV.Acloserlookonthevariousadsorption GO structures with two epoxide groups, systems GO3, GO4 systems compared for similarities yields the following in- andGO5andasingleepoxidegroup,GO1. Calculationsper- sights: I compared to II shows that adsorption close to formed using QE and GPAW with the vdW-DF-cx and PBE an (isolated) epoxide is more favorable by about 0.08 eV functionals. The structures differ by having epoxide on one (GO3) or both sides (GO4) of the graphene plane, and by thatadsorptioncloseto(isolated)hydroxyl. Thesystems having the epoxides placed as nearest neighbors (GO4) or III, IV, and V compared to I show that adsorption close across the graphene carbon ring (GO5). The literature val- to a pair of epoxides is more favorable than adsorption ues (Ref. 46) are for similar but not identical systems. All on a single epoxide. The gain depends on the relative energies in units of eV. positions of the epoxides, whether they are on the same side of the graphene plane (III and V) or not (IV), and This work whether the epoxides are in nearest-neighbor positions QE QE GPAW Ref.46 (III) or sit across the C 6-ring (V). Placing the epoxides Structure vdW-DF PBE PBE PBE on both sides of the graphene plane shows to be more GO3 5.19 4.84 favorable by about 0.11 eV. GO4 5.30 4.94 5.15 4.76 Systems VI and VII are used to examine the depen- GO5 4.19 3.83 4.00 3.59 denceontherelativepositionsofhydroxylpairs: inthese GO4–GO3 0.11 0.10 calculations positioning the hydroxyl pair across a C 6- GO4–GO5 1.21 1.11 1.15 1.17 ringisfavored. Adsorptionontwohydroxylisalsoclearly GO4-2×GO1 0.72 0.61 preferable to adsorption on one hydroxyl only (II), by 0.20 eV. For more complex systems of functional groups in the tional groups (i.e., using graphene), even if that might patch on graphene used as a model of GO it is less result in a slightly worse binding energy than on fully clear which properties of the GO affect the adsorption oxidized GO. For the orientation of chloroform we con- the most. We can, however, examine whether having sider adsorption with the H atom pointing away from the chloroform H atom pointing towards or away from the GO (“H up”) or towards the GO (“H down”), be- GO is favorable. For clean graphene it has previously ing aware that in the end positional relaxation due to been found that an orientation with H pointing away the Hellmann-Feynman forces on the atoms moves the from graphene and the Cl-Cl-Cl tripod pointing towards groups and molecules to less well-defined orientations. graphene is most favorable, among the orientations con- Several configurations of chloroform on GO are stud- sidered, such as the Cl-Cl-H tripod pointing towards ied and data for 16 of these systems are presented in graphene, or the H atom pointing towards graphene.29 Figure 4 and Table III. The GO structures are mainly BecauseofthemoreunevenstructureoftheGOsurface, thosepresentedinFigure3andTableIplusafewother. compared to clean graphene with very little corrugation, 6 wecannotdistinguishtheorientationwiththechloroform TABLE III. Binding energies E of chloroform on GO and H atom pointing to GO and the Cl atoms all pointing b height h of chloroform above a GO O-atom (see Figure 4 for awaytotheorientationwiththeCl-Cl-Htripodpointing theGO-chloroformsystemnumbers). GO#referstotheGO to GO, we will therefore here only distinguish the situa- structures of Figure 3 and Table I. tionofchloroformHpointingmainlyawayfromGO(“H up”) from H pointing mainly towards GO (“H down”). System CCl H orient. GO struct. GO# E [meV] h [˚A] 3 b Systems VIII and IX differ in principle only by the I H up O GO1 257 3.34 orientation of chloroform (besides the thus induced po- II H up OH GO2 181 3.45 sitional relaxations of both the GO and the chloroform III H up 2O GO3 225 3.56 atoms). The adsorption energies of these systems indi- IV H up 2O GO4 382 2.72 cate that “H down” is preferable. However, results for V H up 2O - 268 3.40 the more complex systems X, XI and XII clearly show VI H up 2OH - 390 2.14 that there the “H up” orientation is more favorable, at least if there are functional groups on both sides of the VII H up 2OH GO6 199 3.69 grapheneplane,butalsothatitmattershowtheClatoms VIII H up 2O, OH GO7 219 3.81 are positioned relative to the atoms in epoxide. IX H down 2O, OH GO7 286 3.12 Systems XIV, XV and XVI all have the same number X H down O, 2OH - 319 2.99 of functional groups, placed either on one or both sides XI H down O, 2OH - 296 2.99 ofthegrapheneplane. Again,thedatashowthatplacing XII H up O, 2OH - 422 2.44 functionalgroupsonbothsidesoftheplaneispreferable, XIII H up 2O, 2OH - 243 3.45 evenwhenthenumberoffunctionalgroupsarerestricted XIV H up 3O, 2OH GO9 275 3.79 and the number of functional groups close to chloroform XV H up 3O, 2OH GO10 333 2.88 thus becomes less, compared to having all groups on the same side as the chloroform. XVI H up 3O, 2OH - 391 2.59 The measure h in Table III is an indication of the distance of chloroform from GO. Since GO is not flat, this measureismaybe bothshorteror longerthanforexam- going from left to right in the figure. The parameter ple the smallest atom-to-atom distance in chloroform to values chosen from this convergence test for production GO.Thehmeasuresthedistancebetweenthechloroform runs are more accurate than the default values of QE. C atom and the closest GO O-atom, projected onto the The convergence tests show that further improvements direction perpendicular to the underlying graphene net. of the parameters result in changes of adsorption ener- Thevaluesofhinourcalculationsfallintherange2.1to gies of a few meV or less. 3.8 ˚A. These are reasonable distances for physisorption. Ourcalculationsarecarriedoutwithoutcorrectionfor Theadsorptionofchloroformoncleangraphene,with- the dipole-dipole interaction between two copies of the out functional groups, was previously obtained29 as 0.36 system in the z-direction. The effect of a dipole correc- eV using a similar method. This value corresponds well tion on the systems is tested by applying a dipole cor- to the most favorable configurations for chloroform on rection along the z-axis, for a couple of our systems, in GO systems studied here. a manner described in Refs. 48 and 49. An example of Theinformationweobtainfromstudieslikethisofthe a dipole correction test is presented in Figure 6. The physisorption of small molecules, like the binding ener- electrostatic potential (ESP) curves with and without gies and the orientation of chloroform in this study, may dipole correction are almost overlapping in the figures, be useful both as direct results, but also as input for and hence, for clarity 100× the difference of these two modelling of larger systems.47 curves is also plotted. We see that the dipole correction only has minor effects on the (binding) energy values of the studied systems. This is not a trivial result, because C. Accuracy theGOslabandchloroformarebothpolarobjects. How- ever, the difference in binding energies with and without an applied dipole correction is only 0.5–1.0 meV in most We carefully check the convergence of parameters ofthestudiedsystems,withoneexceptionshowingaless used in our QE calculations. The parameter values are than 10 meV difference. changed,onebyone,toslightlybetterandslightlyworse values, and the binding energy for one of the structures with chloroform on GO is calculated. In Figure 5 we report the binding-energy dependence on unit cell size, IV. CONCLUSIONS wavefunction and density cut-off energies, the force convergence threshold value, the number of k-points, and Wepresentacomputationalstudyofchloroformphysi- thevacuumsize. ThesystemstudiedissystemIIinFig- sorption on GO using DFT calculations with the vdW- ure 4. The corresponding parameters are represented as DF-cx method. The binding energy values vary from a series of calculations, in which the accuracy increases approximately 0.20 to 0.40 eV, depending on the local 7 FIG. 4. Adsorption of chloroform on GO. Shown are top and side views of optimized geometries in systems I–XVI, obtained with the method vdW-DF-cx.. Only a part of the unit cell is shown. The corresponding binding energies are presented in Table III. Color legend for atoms: green Cl; yellow C; red O; blue H. rial for chlorinated water. 200 Graphene size Further,wedocumenttheabilityofvdW-DF-cxofbal- Vacuum [Å] ancing the sp2 and sp3 bindings of the GO C atoms. We k-points do this by comparing the formation energy of the un- ψ/ρ cut-offs [Ry] saturated GO structures, going from pure sp2 binding in F conv. thresh. [meV/Å] 190 clean graphene to a mixture of sp2 and sp3 in the for- V] 2√3x4 mation of GO from graphene. The GO structures are e m structurallyrelaxedwithuseofeitherthesemilocalPBE E [b 2x2x1 330√/31x260 30/284x08x1 50/200 70/280 or the vdW-DF-cx functionals. Besides a small offset in 180 7 2 10 1 4√3x8 the formation energies common to all the structures, we 4x4x1 14 find the same formation energies for PBE and vdW-DF- 20 cx calculations, underlining the ability of vdW-DF-cx of handling the change from sp2 to sp3 bindings. Our results for chloroform adsorption may be used as 170 -1 0 1 2 3 input data for modelling larger systems, with or without Test setup [#] thermodynamics, with or without additional molecules, such as water or ions. While our study is not exhaustive FIG. 5. Convergence tests of calculations for chloroform on in searching for all possible physisorption geometries we GO. For these tests the chloroform-GO system described as have included a number of structures such that we have number II of Figure 4 is used. In this test all calculations covered many of the relevant local environments of GO usethesameparametervaluesastheproductionruns(“Test functional groups. setup 0”) except for one parameter, the parameter indicated in the key of the figure. Lines are for ease of identification. ACKNOWLEDGMENTS environment on the GO, as well as the orientation of the chloroformmoleculerelativetographeneoxide. Thuswe WethankPerHyldgaardandKristianBerlandforpro- find that chloroform physisorbs rather strongly on GO viding access to a pre-release version of the vdW-DF-cx and that graphene-oxide has potential as filtering mate- code with Quantum Espresso. The computations were 8 11D. R. Dreyer, S. Park, C. Bielawski, and R. S. Ruoff, Chem. Soc.Rev.39,228(2009). 12L. Wang, Y. Y. Sun, K. Lee, D. West, Z. F. Chen, J. J. Zhao, andS.B.Zhang,Phys.Rev.B82,161406(2010). 13Y. Zhu, S. Murali, W. Cai, X. Li, J. W. Suk, J. R. Potts, and R.S.Ruoff,Adv.Mater.22,3906(2010). 14S. Basu and P. Bhattacharyya, Sensors Actuators B: Chemical 173,1(2012). 15S.M.Maliyekkal,T.S.Sreeprasad,D.Krishnan,S.Kouser,A.K. Mishra,U.V.Waghmare, andT.Pradeep,Small9,273(2012). 16R.K.Joshi,P.Carbone,F.C.Wang,V.G.Kravets,Y.Su,I.V. Grigorieva,H.A.Wu,A.K.Geim, andR.R.Nair,Science343, 752(2014). 17M. Rossberg, W. Lendle, G. Pfleiderer, A. Tögel, T. R. Torkel- son, andK.K.Beutel,“Chloromethanes,” (Wiley-VCHVerlag GmbH&Co.KGaA,Weinheim,2012). 18M. Rossberg, W. Lendle, G. Pfleiderer, A. Tögel, E.-L. Dreher, E. Langer, H. Rassaerts, P. Kleinschmidt, H. Strack, R. Cook, U.Beck,K.-A.Lipper,T.R.Torkelson,E.Löser, andK.K.Beu- tel (Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, 2006) Chap.ChlorinatedHydrocarbons. 19J.P.Perdew,K.Burke, andM.Ernzerhof,Phys.Rev.Lett.77, 3865(1996). 20M. Dion, H. Rydberg, E. Schröder, D. C. Langreth, and B. I. Lundqvist,Phys.Rev.Lett.92,246401(2004). 21M. Dion, H. Rydberg, E. Schröder, D. C. Langreth, and B. I. Lundqvist,Phys.Rev.Lett.95,109902(2005). 22T. Thonhauser, V. R. Cooper, S. Li, A. Puzder, P. Hyldgaard, andD.C.Langreth,Phys.Rev.B.76,125112(2007). FIG. 6. Effect of the dipole correction on the electrostatic 23K.Berland,C.A.Arter,V.R.Cooper,K.Lee,B.I.Lundqvist, potential in GO-chloroform systems. The plot shows a po- E.Schröder,T.Thonhauser, andP.Hyldgaard,J.Chem.Phys. 140,18A539(2014). tential line scan along the box z-axis through the chloroform 24K.Berland,V.R.Cooper,K.Lee,E.Schro¨der,T.Thonhauser, C atom (perpendicular to graphene) for the systems (a) II P.Hyldgaard, andB.I.Lundqvist,Rep.Prog.Phys.78,066501 and (b) IV shown in Figure 4. To illustrate the length scales (2015). and positions of the constituents a piece of the GO slab and 25K.BerlandandP.Hyldgaard,Phys.Rev.B89,035412(2014). the chloroform molecule are placed at their position along 26H.Ulbricht,R.Zacharia,N.Cindir, andT.Hertel,Carbon44, thez-axes. Theunitcelllengthsinthesetwocalculationsare 2931(2006). 16 ˚A (top) and 14 ˚A (bottom). Colors as in Figure 4. The 27A. Fujii, K. Shibasaki, T. Kazama, R. Itaya, N. Mikami, and solid blue line shows 100× the difference in potentials with S.Tsuzuki,Phys.Chem.Chem.Phys.10,2836(2008). and without the dipole correction. 28T.R.Rybolt,D.L.Logan,M.W.Milburn,H.E.Thomas, and A.B.Waters,J.ColloidInterfaceSci.220,148(1999). 29J. ˚Akesson, O. Sundborg, O. Wahlström, and E. Schröder, J. Chem.Phys.137,174702(2012). performed on resources at Chalmers Centre for Compu- 30J.Hooper,V.R.Cooper,T.Thonhauser,N.A.Romero,F.Zer- tationalScienceandEngineering(C3SE)providedbythe illi, andD.C.Langreth,ChemPhysChem9,891(2008). Swedish National Infrastructure for Computing (SNIC). 31Y.PengandJ.Li,Front.Environ.Sci.Eng.7,403(2013). 32T.NakajimaandY.Matsuo,Carbon26,357(1988). 33A.Kokalj,Comp.Mater.Sci.28,155(2003). 1B.C.Brodie,Phil.Trans.R.Soc.Lond.149,249(1859). 34W.Humphrey,A.Dalke, andK.Schulten,J.MolecularGraphics 2S.Stankovich,D.A.Dikin,G.H.B.Dommett,K.M.Kohlhaas, 14,33(1996). E. J. Zimney, E. A. Stach, R. D. Piner, S. Nguyen, and R. S. 35P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, Ruoff,Nature442,282(2006). C. Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococcioni, 3S.Stankovich,R.D.Piner,X.Q.Chen,N.Q.Wu,S.T.Nguyen, I. Dabo, A. D. Corso, S. Fabris, G. Fratesi, S. de Giron- andR.S.Ruoff,J.Mater.Chem.16,155(2006). coli, R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, 4H. C. Schniepp, J. L. Li, M. J. McAllister, H. Sai, M. Herrera- M. Lazzeri, L. Martin-Samos, N. Marzari, F. Mauri, R. Maz- Alonso,D.H.Adomson,R.K.Prud’homme,R.Car,D.A.Sav- zarello, S. Paolini, A. Pasquarello, L. Paulatto, C. Sbraccia, ille, andI.A.Aksay,J.Phys.Chem.B110,8535(2006). S. Scandolo, G. Sclauzero, A. P. Seitsonen, A. Smogunov, 5C. Gómez-Navarro, R. T. Weitz, A. M. Bittner, M. Scolari, P. Umari, and R. M. Wentzcovitch, J. Phys.: Condens. Mat- A.Mews,M.Burghard, andK.Kern,NanoLett.7,3499(2007). ter21,395502(2009). 6A.Lerf,H.He,M.Forster, andJ.Klinowski,J.Phys.Chem.B 36Open-source DFT code QE, http://www.quantum-espresso. 102,4477(1998). org/. 7T. Szabó, O. Berkesi, P. Forgo, K. Josepovits, Y. Sanakis, 37J.Enkovaara,C.Rostgaard,J.J.Mortensen,J.Chen,M.Du(cid:32)lak, D.Petridis, andI.Dekany,Chem.Mater.18,2740(2006). L.Ferrighi,J.Gavnholt,C.Glinsvad,V.Haikola,H.A.Hansen, 8J.T.Paci,T.Belytschko, andG.C.Schatz,J.Phys.Chem.C H. H. Kristoffersen, M. Kuisma, A. H. Larsen, L. Lehtovaara, 111,18099(2007). M. Ljungberg, O. Lopez-Acevedo, P. G. Moses, J. Ojanen, 9D. W. Boukhvalov and M. I. Katsnelson, JACS 130, 10697 T. Olsen, V. Petzold, N. A. Romero, J. Stausholm-Møller, (2008). M. Strange, G. A. Tritsaris, M. Vanin, M. Walter, B. Hammer, 10K. A. Mkhoyan, A. W. Contryman, J. Silcox, D. A. Stewart, H.Häkkinen,G.K.H.Madsen,R.M.Nieminen,J.K.Nørskov, G.Eda,C.Mattevi,S.Miller, andM.Chhowalla,NanoLett.9, M.Puska,T.T.Rantala,J.Schiøtz,K.S.Thygesen, andK.W. 1058(2009). 9 Jacobsen,J.Phys.Cond.Matt.22,253202(2010). setups/setups.html. 38Open-source,grid-basedPAW-methodDFTcodeGPAW,http: 44E. Ziambaras, J. Kleis, E. Schro¨der, and P. Hyldgaard, Phys. //wiki.fysik.dtu.dk/gpaw/. Rev.B76,155425(2007). 39D.Vanderbilt,Phys.Rev.B41,7892(1990). 45I.Hamada,Phys.Rev.B89,121103(2014). 40K. Garrity, J. Bennett, K. Rabe, and D. Vanderbilt, Comput. 46Zˇ.Sˇljivanˇcanin,A.S.Miloˇsević,Z.S.Popović, andF.R.Vuka- Mater.Sci.81,446(2014). jlović,Carbon54,482(2013). 41P.E.Blöchl,Phys.Rev.B50,17953(1994). 47V. R. Cooper, Y. Ihm, and J. R. Morris, Phys. Proc. 34, 34 42J.J.Mortensen,L.B.Hansen, andK.W.Jacobsen,Phys.Rev. (2012). B71,035109(2005). 48L.Bengtsson,Phys.Rev.B59,12301(1999). 43PAWsetupsasdescribedathttps://wiki.fysik.dtu.dk/gpaw/ 49B.MeyerandD.Vanderbilt,Phys.Rev.B63,205426(2001).