7
19644 Phys. Chem. Chem. Phys., 2013, 15, 19644--19650 This journal is c the Owner Societies 2013 Cite this: Phys. Chem. Chem. Phys., 2013, 15, 19644 The binding nature of light hydrocarbons on Fe/MOF-74 for gas separationHeejin Kim, a Joonho Park a and Yousung Jung* ab The application of a metal–organic framework (MOF) has expanded into the area of heterogeneous catalysis, gas storage and separation, drug delivery, and lightweight magnets. Herein, we investigate the nature of olefin and paraffin binding on Fe/MOF-74 and identify several factors that determine separa- tion efficiency using the first-principles calculations. The calculated binding energies and magnetic orderings are in excellent agreement with those observed in experiments. While the olefin strongly interacts with Fe atoms through a well-known p-complexation, the HOMO 1(2) of the paraffin weakly interacts with Fe atoms without back-donation, facilitating the olefin–paraffin separation primarily. However, the mutual gas–gas interactions and magnetic transitions of the MOF host also contribute significantly to the total binding energy of each gas molecule as much as 2–28% and 6–8%, respectively, emphasizing the necessity that these subtle effects must be handled carefully when considering selective binding with small energy differences. In particular, Fe/MOF-74 is shown to be a unique system where the guest-dependent magnetic transition observed only for the olefin adsorption is a secondary reason for the high olefin–paraffin adsorption selectivity measured. The understanding of the hydrocarbon binding energetics can provide a way to modify MOFs for enhanced separation/sorption properties that can be complemented by principles of kinetic separation. Introduction Light hydrocarbons such as methane, ethane, propane, acetylene, ethylene and propylene are widely used as the energy resources and chemical feedstocks in diverse industries. While the applica- tion areas are different for each hydrocarbon species, they are produced all at the same time by thermal cracking of large hydrocarbons and thus separation processes are required to obtain high purity ingredients. Commercially, the separation of binary gas mixtures such as ethane–ethylene and propane– propylene is the most important but difficult process due to the similar physicochemical properties of the mixed gases. 1 Although the cryogenic distillation has been successfully used for these separations, its operating cost is exceedingly high since low temperatures and high pressures are required, demanding the development of alternative technologies. 2,3 Separation methods based on reversible chemical absorption have been investigated as a promising approach. 1,3,4 In earlier attempts on the adsorp- tive separation, however, the low surface area (capacity) and selectivity were two major limitations. 1,4 To overcome the latter problems simultaneously, a number of research studies have focused on the Ag(I) or Cu(I) cation impregnated mesoporous adsorbents and indeed, they exhibited good separation selec- tivities due to the p-complexation 5 between the metal cations and olefins. 1,4,6,7 As an extension, metal–organic frameworks (MOFs) have also been explored for olefin–paraffin separation. 8–15 MOFs are composed of metal-containing clusters and organic linkers 16 with extremely high surface area that enables MOFs to be used in gas manipulations such as carbon dioxide, 17–25 hydrogen, 26–33 and hydrocarbons. 34–39 Especially MOFs with coordinatively unsaturated metal sites can strongly bind with the guest olefins. To the best of our knowledge, open Cu sites in HKUST-1 13 and ZJU-5, 40 open Fe sites in MIL-100(Fe), 14 and open M sites in M/MOF-74 (M = Mg, Mn, Fe, Co, Ni, Zn) 8–12 were reported to undergo p-complexation with olefins. Among them, the M/MOF-74 (or CPO27-M) have been studied most extensively due to their outstanding separation performance. For instance, Fe/MOF-74 shows 2–3 times higher selectivity and capacity for hydrocarbon separation compared to the zeolite-based solid adsorbents without capacity loss and hysteresis for dozens of cycles. 10 While they exhibit overall good olefin–paraffin selectivity a Graduate School of EEWS, Korea Advanced Institute of Science and Technology (KAIST), Yuseong-gu, Daejeon 305-701, Republic of Korea. E-mail: [email protected] b KAIST Institute NanoCentury, Korea Advanced Institute of Science and Technology (KAIST), Yuseong-gu, Daejeon 305-701, Republic of Korea † Electronic supplementary information (ESI) available: PDOS of Fe at different Fe–Fe distances, more calculation details, comparison of DFT methods, geo- metric changes, and comparison between different magnetic orderings. See DOI: 10.1039/c3cp52980k Received 16th July 2013, Accepted 3rd October 2013 DOI: 10.1039/c3cp52980k www.rsc.org/pccp PCCP PAPER Published on 04 October 2013. Downloaded by Korea Advanced Institute of Science & Technology / KAIST on 19/01/2014 06:11:23. View Article Online View Journal | View Issue

RSC CP C3CP52980K 3. - dspace.kaist.ac.kr€¦ · The application of a metal–organic framework (MOF) has expanded into the area of heterogeneous catalysis, gas storage and separation,

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: RSC CP C3CP52980K 3. - dspace.kaist.ac.kr€¦ · The application of a metal–organic framework (MOF) has expanded into the area of heterogeneous catalysis, gas storage and separation,

19644 Phys. Chem. Chem. Phys., 2013, 15, 19644--19650 This journal is c the Owner Societies 2013

Cite this: Phys. Chem.Chem.Phys.,2013,15, 19644

The binding nature of light hydrocarbons onFe/MOF-74 for gas separation†

Heejin Kim,a Joonho Parka and Yousung Jung*ab

The application of a metal–organic framework (MOF) has expanded into the area of heterogeneous

catalysis, gas storage and separation, drug delivery, and lightweight magnets. Herein, we investigate the

nature of olefin and paraffin binding on Fe/MOF-74 and identify several factors that determine separa-

tion efficiency using the first-principles calculations. The calculated binding energies and magnetic

orderings are in excellent agreement with those observed in experiments. While the olefin strongly

interacts with Fe atoms through a well-known p-complexation, the HOMO � 1(2) of the paraffin

weakly interacts with Fe atoms without back-donation, facilitating the olefin–paraffin separation

primarily. However, the mutual gas–gas interactions and magnetic transitions of the MOF host also

contribute significantly to the total binding energy of each gas molecule as much as 2–28% and 6–8%,

respectively, emphasizing the necessity that these subtle effects must be handled carefully when

considering selective binding with small energy differences. In particular, Fe/MOF-74 is shown to be a

unique system where the guest-dependent magnetic transition observed only for the olefin adsorption

is a secondary reason for the high olefin–paraffin adsorption selectivity measured. The understanding of

the hydrocarbon binding energetics can provide a way to modify MOFs for enhanced separation/sorption

properties that can be complemented by principles of kinetic separation.

Introduction

Light hydrocarbons such as methane, ethane, propane, acetylene,ethylene and propylene are widely used as the energy resourcesand chemical feedstocks in diverse industries. While the applica-tion areas are different for each hydrocarbon species, they areproduced all at the same time by thermal cracking of largehydrocarbons and thus separation processes are required toobtain high purity ingredients. Commercially, the separation ofbinary gas mixtures such as ethane–ethylene and propane–propylene is the most important but difficult process due tothe similar physicochemical properties of the mixed gases.1

Although the cryogenic distillation has been successfully usedfor these separations, its operating cost is exceedingly highsince low temperatures and high pressures are required, demandingthe development of alternative technologies.2,3 Separation methodsbased on reversible chemical absorption have been investigated

as a promising approach.1,3,4 In earlier attempts on the adsorp-tive separation, however, the low surface area (capacity) andselectivity were two major limitations.1,4 To overcome the latterproblems simultaneously, a number of research studies havefocused on the Ag(I) or Cu(I) cation impregnated mesoporousadsorbents and indeed, they exhibited good separation selec-tivities due to the p-complexation5 between the metal cationsand olefins.1,4,6,7

As an extension, metal–organic frameworks (MOFs) havealso been explored for olefin–paraffin separation.8–15 MOFs arecomposed of metal-containing clusters and organic linkers16

with extremely high surface area that enables MOFs to be used ingas manipulations such as carbon dioxide,17–25 hydrogen,26–33

and hydrocarbons.34–39 Especially MOFs with coordinativelyunsaturated metal sites can strongly bind with the guest olefins.To the best of our knowledge, open Cu sites in HKUST-113 andZJU-5,40 open Fe sites in MIL-100(Fe),14 and open M sites inM/MOF-74 (M = Mg, Mn, Fe, Co, Ni, Zn)8–12 were reported toundergo p-complexation with olefins.

Among them, the M/MOF-74 (or CPO27-M) have been studiedmost extensively due to their outstanding separation performance.For instance, Fe/MOF-74 shows 2–3 times higher selectivity andcapacity for hydrocarbon separation compared to the zeolite-basedsolid adsorbents without capacity loss and hysteresis for dozens ofcycles.10 While they exhibit overall good olefin–paraffin selectivity

a Graduate School of EEWS, Korea Advanced Institute of Science and Technology

(KAIST), Yuseong-gu, Daejeon 305-701, Republic of Korea. E-mail: [email protected] KAIST Institute NanoCentury, Korea Advanced Institute of Science and Technology

(KAIST), Yuseong-gu, Daejeon 305-701, Republic of Korea

† Electronic supplementary information (ESI) available: PDOS of Fe at differentFe–Fe distances, more calculation details, comparison of DFT methods, geo-metric changes, and comparison between different magnetic orderings. See DOI:10.1039/c3cp52980k

Received 16th July 2013,Accepted 3rd October 2013

DOI: 10.1039/c3cp52980k

www.rsc.org/pccp

PCCP

PAPER

Publ

ishe

d on

04

Oct

ober

201

3. D

ownl

oade

d by

Kor

ea A

dvan

ced

Inst

itute

of

Scie

nce

& T

echn

olog

y / K

AIS

T o

n 19

/01/

2014

06:

11:2

3.

View Article OnlineView Journal | View Issue

Page 2: RSC CP C3CP52980K 3. - dspace.kaist.ac.kr€¦ · The application of a metal–organic framework (MOF) has expanded into the area of heterogeneous catalysis, gas storage and separation,

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 19644--19650 19645

regardless of the particular metal species, the best separatingmaterials are different for each case of gas mixtures. For instance,Fe/MOF-74 and Mn/MOF-74 are optimal for the separation ofethylene–ethane and propylene–propane mixtures, respectively,11

and Co/MOF-74 is the best for the separation of methane froma mixture of C2 hydrocarbons.12 To resolve such mixture-dependent adsorption performance as well as to modulatemetal centers and functional groups for obtaining suitablecharacteristics, an atomic scale understanding is necessary.

Recently, several theoretical studies have been performed onthe hydrocarbon binding/separation in MOF-74.41–43 Hou et al.analysed the ethylene and methane adsorptions on Mg/ andZn/MOF-74 by employing cluster models.41 In their calculations,only a forward donation was identified for ethylene adsorption,while both forward and backward donations are observed inmethane adsorption, conflicting with the common p-complexationphenomena. Sillar et al. investigated the methane adsorption onMg/MOF-74 focusing on the lateral interactions where they showedthat the mutual gas–gas interaction contributes to the totalbinding energy by 10–15%.42 Verma et al. examined the adsorp-tion of C1–C3 hydrocarbons particularly on Fe/MOF-74.43 Theyobserved notable forward and backward donations for theethylene adsorption, but not for the ethane adsorption. Whiletheir calculated energies are qualitatively consistent with experi-ments, explanations for distinctive adsorption of acetylene, thebinding nature of paraffin, mutual gas interactions, magnetictransitions, and their effects on the hydrocarbon separation werenot addressed and still remain unclear.

In Fe/MOF-74, the intrachain magnetic ordering of Fe atomsalong the c-direction (Fig. 1) is switched from ferromagnetic(FM, Fig. 1b) to antiferromagnetic (AFM, Fig. 1c) after olefinadsorption, while it remains FM for the paraffin adsorption.10

This selective magnetic transition can affect the adsorptionstrength by a different amount for each gas molecule, therebyaffecting the separation selectivity.

In this paper, we investigate the binding nature of olefins andparaffins on Fe/MOF-74 using the periodic DFT calculations thatcan fully describe the magnetic orderings, mutual gas interactions,and their effects on the hydrocarbon separations. In particular, thepresent study addresses the following aspects: (1) what is thenature of a distinctive adsorption of acetylene with sp hybridiza-tion compared to olefins with sp2 hybridization? (2) what is thebinding nature of paraffin, is it just dispersion interactions, orare there distinct orbital interactions different from the olefincase? (3) can mutual gas–gas interactions affect the separationefficiency, and if so, to what extent? and (4) how and to whatextent did the magnetic transition from ferro- to anti-ferromag-netism upon olefin adsorption observed experimentally, butnot paraffin adsorption, affect the separation efficiency?

Calculation details

The periodic density functional theory (DFT) calculationswere performed under the generalized gradient approxi-mation (GGA) with dispersion corrections (D2)44 as imple-mented in the Vienna Ab initio Simulation Package (VASP).45

Both Perdew–Burke–Ernzerhof (PBE)46 and revised PBE (RPBE)47

exchange–correlation functionals give a qualitatively identicaltrend in binding energies of hydrocarbons, but RPBE + D2provides quantitatively superior results. While we also examinedHubbard U-corrections48 with Ueff = 4.0 eV for Fe,49 the bindingenergies were much worse than the uncorrected results. SeeTable S1 in the ESI† for the comparison of calculation methods.Therefore, we adopted RPBE + D2 in all calculations. The initialstructures were taken from the experiments and geometries10

were fully optimized until the force is less than 0.025 eV �1. Weused a plane wave basis set and the projector-augmented wave(PAW) method50 with an energy cutoff of 500 eV. The Brilloulinzone integration was performed at the gamma point. All crystalstructures were visualized using the VESTA program.51

To address the ground state magnetic orderings for each gasadsorbed structure, we adopted supercells containing 36 adsorptionsites to describe AFM ordering along the 1D chain of Fe/MOF-74 asshown in Fig. 1. We assumed that gas molecules occupy allunsaturated Fe sites. For each gas adsorption, we examinedthree different magnetic orderings that are observed in experi-ments: (i) intrachain FM and interchain antiparallel (Fig. 1b),(ii) intrachain AFM and interchain antiparallel (Fig. 1c), and(iii) intrachain FM and interchain parallel (magnetized).

ResultsElectronic structure of bare Fe/MOF-74

The electronic structure of Fe/MOF-74 and its effects on thebinding of hydrocarbons can be understood from the conventional

Fig. 1 (a) Crystal structure of Fe/MOF-74 used in calculations. The blue, red, gray,and white are Fe, O, C, and H, respectively. The ground state magnetic ordering (b) forthe bare and paraffin adsorbed Fe/MOF-74, and (c) for the olefin adsorbed Fe/MOF-74.The dark and light polyhedrons represent up and down spin states, respectively.

Paper PCCP

Publ

ishe

d on

04

Oct

ober

201

3. D

ownl

oade

d by

Kor

ea A

dvan

ced

Inst

itute

of

Scie

nce

& T

echn

olog

y / K

AIS

T o

n 19

/01/

2014

06:

11:2

3.

View Article Online

Page 3: RSC CP C3CP52980K 3. - dspace.kaist.ac.kr€¦ · The application of a metal–organic framework (MOF) has expanded into the area of heterogeneous catalysis, gas storage and separation,

19646 Phys. Chem. Chem. Phys., 2013, 15, 19644--19650 This journal is c the Owner Societies 2013

ligand field theory. Fig. 2a shows the projected density of states(DOS) for each d-orbital of Fe and that for oxygen, where thetop and bottom sides correspond to the up- and down-spinelectrons, respectively. From the geometric viewpoint, eachFe ion has a square pyramidal coordination of oxygen atoms(Fig. 1 and 2b) and one coordination site is empty toward thetunnel to afford one more ligand. The highest and secondhighest energy levels of dx2�y2 and dz2 orbitals, respectively, arelargely in agreement with the energy splitting by the squarepyramidal coordination of ligands.52 However, the dxy, dyz, anddzx orbitals, namely dt2g

orbitals, are more similar in energythan typical five-coordinate systems as noted with the splittingshown in (i) in Fig. 1, since the neighboring Fe atoms arelocated close together at 2.79 Å along the diagonal direction(Fig. 2b), disturbing the orbital symmetry.

One down-spin electron in d6 of Fe(II) occupies the afore-mentioned dt2g

orbitals that are located just below the Fermilevel (Fig. 2a–(i)), so it is the highest occupied molecular orbital(HOMO) of bare Fe/MOF-74. Fig. 2c shows an isosurface ofelectron density at the energy level of HOMO that illustrates theorbital shape and interaction between electrons. As shown inthe DOS (Fig. 2a–(i)) and electron density (Fig. 2c), the HOMO iscomposed mainly of Fe-dt2g

electrons and each orbital splitsinto multiple bands, indicating the existence of a direct inter-action between Fe atoms along the 1D chain (c-direction)through electrons in the HOMO. While the equilibrium dis-tance between Fe atoms in our calculation regime (2.79 Å) isshorter than that in experiments (2.99 Å), we note that thesedirect interaction phenomena also hold at the experimentalFe–Fe distance (See Fig. S1 in the ESI†).

Olefin adsorption

The binding nature of olefin–metal complexes has beenexplained well in organometallic chemistry by employing thes(p)-bonding and p-backbonding concepts, namely p-complexation,where the outermost shell of a metal and the HOMO of anolefin are involved.5 The HOMO of a free olefin gas is thep-bonding molecular orbital which is placed at 2–3 eV belowthe Fermi level of bare Fe/MOF-74 as shown in Fig. 2a. Insteadof an isotropic 4s orbital in the olefin–Ag(I) or –Cu(I) complexa-tion, the 3d orbitals participate in the olefin–Fe/MOF-74 com-plexation as the outermost shell, owing to the open geometry ofMOF-74. Therefore, the spatial direction and phase of Fe-3dorbitals should match well with the interacting orbitals of theolefin to form a complex.

For the acetylene (Ay) adsorption, for example, the projectedDOS and corresponding electron density isosurfaces at severalrelevant energy levels are presented in Fig. 3a and b, respectively.The most significant orbital overlap appears at the energy levelof (i), which is attributed to the bonding between the dz2 orbitalof Fe and the HOMO of Ay, which locates in the z-direction,namely pz-HOMO. We denote this s-type orbital interaction asthe pz–dz2 bonding (Fig. 3b–(i)).

In contrast to ethylene (Ey) and propylene (Py) that belong tothe sp2 hybridization group, the sp hybridized Ay has twodegenerate HOMOs, pz and px, rather than one. This additionalpx-HOMO can interact with the Fe–dxz orbital, namely px–dxz

bonding, as shown in Fig. 3a–(ii) and b–(ii). Although thisp-type overlap of px–dxz bonding is much smaller than that ofpz–dz2 bonding, it increases the binding strength between Ayand Fe/MOF-74 compared to the other olefins to some extent.

Meanwhile, the lowest unoccupied molecular orbital (LUMO)of the olefin gas locates at 3–4 eV above the Fermi level of bare

Fig. 2 (a) The decomposed DOS for Fe and oxygen ions in bare Fe/MOF-74. Thedotted and solid lines in the lower part of the figure indicate the relative positionsof HOMO and LUMO levels for isolated paraffin (Ma = methane, Ea = ethane,Pa = propane) and olefin (Ay = acetylene, Ey = ethylene, Py = propylene) gasmolecules. (b) Local geometries with Fe–O distances, Fe–Fe distance, Fe–O–Feangle. (c) The electron density isosurface at the HOMO level as noted with(i) in Fig. 2a.

Fig. 3 (a) The decomposed DOS for the acetylene (Ay) adsorbed Fe/MOF-74.(b) (i) The bonding between the dz2 orbital of Fe and the pz-HOMO of Ay;(ii) bonding between the dxz orbital of Fe and the px-HOMO of Ay; (iii) thep-backbonding between the dyz orbital of Fe and LUMO of Ay. (c) The Fe–Odistances, Fe–O–Fe angle, and Fe–Fe distance are increased by Ay adsorption.

PCCP Paper

Publ

ishe

d on

04

Oct

ober

201

3. D

ownl

oade

d by

Kor

ea A

dvan

ced

Inst

itute

of

Scie

nce

& T

echn

olog

y / K

AIS

T o

n 19

/01/

2014

06:

11:2

3.

View Article Online

Page 4: RSC CP C3CP52980K 3. - dspace.kaist.ac.kr€¦ · The application of a metal–organic framework (MOF) has expanded into the area of heterogeneous catalysis, gas storage and separation,

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 19644--19650 19647

Fe/MOF-74 and it accepts electrons from the Fe-dyz orbital byforming the p-backbonding as shown in Fig. 3a–(iii) and b–(iii).These bondings and backbonding synergistically contribute tothe strong adsorption of olefins on Fe/MOF-74, not only by wayof typical electron donation–backdonation, but also from ageometric viewpoint as follows.

Upon the Ay adsorption, an Fe atom is pulled out toward thechannel direction, with an increase of the coordination numberfrom five to six by forming a bond with Ay. While the strongpz–dz2 bonding mainly induces the pulling and bond formation,the p-backbonding additionally helps the binding strength,where it diminishes Fe–Fe attractions by drawing electronsfrom the HOMO that are responsible for direct Fe–Fe inter-actions. Indeed, the slightly delocalized HOMO bands (Fig. 2a–(i))in bare MOF disappear after the Ay adsorption, indicating thedisappearance of the direct interaction between Fe atoms.Instead, all the down-spin electrons are localized in the dyz

orbital and participate in the p-backbonding with Ay molecule(Fig. 3a–(iii)). This p-backbonding orbital is the HOMO of Ayadsorbed Fe/MOF-74.

As a result, Fe–O bond distances increase by 0.08 Å onaverage, and the distance between neighboring Fe atomsincreases from 2.79 to 3.31 Å. The Fe–O–Fe bond angle alsoincreases up to 107.11. These geometric evolutions can be theorigin of the observed magnetic transition along the 1D chainof Fe atoms for the olefin adsorption.53–55 The C–C distance ofgas molecules, the Fe–Fe distance, and bond angles before/afterthe adsorption that can be a measure of complexation, aresummarized in the ESI† (Table S2).

Paraffin adsorption

Hereafter, the notation [mol]geom indicates that the geometrywas taken from the complex denoted as ‘‘geom’’, and the part(or whole) of the complex under consideration is denoted as‘‘mol’’. For example, [MOF + Pa]MOF+Pa denotes the fully opti-mized structure of the propane (Pa) adsorbed MOF (both hostand guest included), [MOF]MOF+Pa denotes the MOF host only,whose geometry is extracted from the fully optimized MOF + Pa,and [MOF]MOF denotes the optimized structure of the bareFe/MOF-74. Similarly, [Pa]MOF+Pa denotes the Pa guest moleculesonly, whose geometries were taken from the fully optimizedMOF + Pa. [1Pa]MOF+Pa denotes a single gas molecule taken fromMOF + Pa. These notations are illustrated in Fig. S3 in the ESI†for clarity.

Fig. 4a shows the projected DOS of Fe atoms for [MOF +Pa]MOF+Pa, [MOF]MOF+Pa, and [MOF]MOF. The overall DOS changeof Fe atoms due to the Pa adsorption is negligible compared tothat due to the olefin adsorption (Fig. 4b). To be more specific,the change in electronic structure due to the structural change ofthe MOF host upon Pa adsorption (from green [MOF]MOF toblue [MOF]MOF+Pa) is exceedingly small, consistent with smallgeometry changes for paraffin adsorptions (Table S2, ESI†).Moreover, the additional changes by adding Pa molecules to thedistorted MOF (from blue [MOF]MOF+Pa to black [MOF + Pa]MOF+Pa)are also small, indicating that the electronic interactions betweenPa and Fe/MOF-74 are insignificant compared to the olefin cases

where the gas adsorption also induces a noticeable alteration inthe electronic structure of the host MOF as shown in Fig. 4b. Themost significant changes in electronic structure for Pa adsorptionare marked with (i)–(iv) and they are discussed below.

The key interactions in the paraffin adsorbed Fe/MOF-74 canbe categorized as follows, for example, in the case of Pa.

(1) Paraffin–Fe interaction: since the HOMO of Pa gasresembles a p*-antibonding orbital (composed of C–H s-bonds),it can interact with the in-phase dt2g

orbitals (dxy, dyz, and dxz) ofFe atoms instead of the dz2 orbital. To do that, the Pa gas shouldwithdraw a sufficient amount of electrons from dt2g

that formsthe Fe–Fe direct interaction. However, the latter interactioncannot be so strong since the HOMO of Pa is a combination ofthe localized s-orbitals that are far below the Fermi level ofFe/MOF-74 (Fig. 2a). Indeed, the Fe–Fe distance increases onlyby 0.02 Å after the Pa adsorption (see the ESI† for detailedgeometries), contrast to the 0.52 Å increment for the Ayadsorption. Instead, the HOMO � 2 of Pa is a s-resembledorbital close to the unsaturated Fe atoms, and hence instead

Fig. 4 (a) Solid lines: the projected DOS of Fe atoms for Pa adsorbed Fe/MOF-74(black [MOF + Pa]MOF+Pa), for the same framework without Pa molecules (blue[MOF]MOF+Pa), and for the bare MOF (green [MOF]MOF). Shaded gray from top tobottom: the DOS for the adsorbed Pa guest molecules in [MOF + Pa]MOF+Pa, forthe 36 Pa molecules in MOF + Pa framework without MOF ([Pa]MOF+Pa), and forthe single Pa molecule in MOF + Pa framework without MOF ([1Pa]MOF+Pa).(b) The same notation as Fig. 4a for propylene adsorbed Fe/MOF-74. (c) Theelectron density isosurfaces at the energy levels of (i), (iii), and (iv) of Fig. 4a. The(i) presents the gas–MOF interaction and the (iii) and (iv) show the in-phase andout-of-phase gas–gas interactions.

Paper PCCP

Publ

ishe

d on

04

Oct

ober

201

3. D

ownl

oade

d by

Kor

ea A

dvan

ced

Inst

itute

of

Scie

nce

& T

echn

olog

y / K

AIS

T o

n 19

/01/

2014

06:

11:2

3.

View Article Online

Page 5: RSC CP C3CP52980K 3. - dspace.kaist.ac.kr€¦ · The application of a metal–organic framework (MOF) has expanded into the area of heterogeneous catalysis, gas storage and separation,

19648 Phys. Chem. Chem. Phys., 2013, 15, 19644--19650 This journal is c the Owner Societies 2013

this HOMO � 2 interacts with the dz2 orbital of Fe atomsas shown in Fig. 4a–(i) and c–(i). In the same manner, theHOMO � 1 of ethane (Ea) can interact with Fe atoms. Note thatthe p-backbonding to the paraffin is not allowed, and the aboveoverlap between HOMO � 2(1) of paraffin and Fe/MOF-74 ismuch smaller than that for the olefins, both leading to a weakadsorption of the paraffin in Fe/MOF-74.

(2) Paraffin–ligand interaction: the short distances betweenH in Pa and O in Fe/MOF-74 (2.59 Å in calculation, 2.48 Å inexperiment10) allow the interaction between Pa and the ligandsof Fe/MOF-74 at the energy level of (ii), for example, in Fig. 4a.This type of interaction is also available for the olefin cases.

(3) Mutual gas interaction: in the DOS of Pa (shaded inFig. 4a), the band splitting in the multiple gas molecules ([Pa]MOF+Pa)compared to a single isolated gas molecule ([1Pa]MOF+Pa) implies theinteraction between gas molecules. Since the band split isattributed to the in-phase interaction (iii) and out-of-phaseinteraction (iv), it depends on the spatial shape of the mole-cular orbitals. Note that the extent of DOS splitting, or mutualgas interaction, decreases as the molecular size decreases(Pa > Ea > Ma). Consistently, the calculated gas–gas interactionenergies for Ay, Ey, Py, Ma, Ea, and Pa in the Fe/MOF-74framework are 0.9, 2.1, 5.5, 1.8, 4.3, and 8.0 kJ mol�1, respec-tively, whose magnitudes are in accord with the Ma–Ma inter-action in Mg/MOF-74 (2.4 kJ mol�1).42 This dispersion-dominantinteraction is stronger for larger molecules since their polariz-abilities are higher and intermolecular distances are closerwithin the confined volume. In addition, the mutual gas inter-action is smaller in olefins since the gas–gas intermoleculardistances are further apart, the molecular sizes and polarizabi-lities are smaller, and the polarizable HOMO electrons are inlarge part delocalized through the bonding with open metal sitesin Fe/MOF-74 (shaded in Fig. 4b)

Binding energies

The binding energy is calculated between the fully optimizedgeometries/energies under the ground state magnetic orderings foreach structure. Consistent with experiments,10 Fe/MOF-74 shows aclear olefin–paraffin selectivity in binding energies: it bindsolefin gases (45–50 kJ mol�1) and paraffin gases (15–30 kJ mol�1)(Fig. 5).

For olefins, the exact order of calculated binding energy(Py > Ay > Ey) is different from the Bloch’s experiments (Ay >Ey > Py) but the same as another measurement using the sameexperimental method,12 indicating that the energy differences(within 5 kJ mol�1) are perhaps in the error range of measure-ments. For comparison, the binding energies in the absence ofmutual gas interactions are considered. As mentioned above,Ay has an additional px-bonding that contributes weakly to theadsorption and hence, Ay shows a stronger gas–MOF inter-action (white bar) than Ey and Py by B3 kJ mol�1 as shown inFig. 5. The mutual gas interactions in olefins, however, arelargest in the Py, perhaps as expected.

For paraffin, the order of calculated binding energy isconsistent with two independent experiments, Pa > Ea > Ma.10,12

This order is completely consistent for all aforementioned

binding components: (1) the molecular orbitals of paraffinsthat are able to interact with Fe are placed higher in energy inthe order of Pa > Ea > Ma, (2) the paraffin–ligand dispersioninteraction is higher for larger and more polarizable molecules,and (3) the strength of the gas–gas interaction is in the orderof Pa > Ea > Ma. A consistent slight underestimation of thebinding energy for paraffins is probably due to the overestimationof the distance between gas molecule and Fe/MOF-74 (2.59 Å incalculation, 2.48 Å in experiment for Pa10) and the consequentunderestimation of gas–MOF interactions, as well as the limita-tions of the present density functionals used. Also, the partialoccupancy that depends on the loading can alter the bindingstrength of gas molecules.

The binding energy can be further decomposed into contribu-tions from the geometric distortion, magnetic transition, mutualgas interaction, and electronic effects. Fig. 6 shows relative energiesfor each gas adsorption state. By using the same [mol]geom

notations as in the previous section, state (i) is [MOF]MOF withintrachain FM ordering and 36 free gas molecules (isolated MOFand gas molecules in each ground state, i.e. a reference energystate); state (ii) is [MOF]MOF+gas with intrachain FM ordering and

Fig. 5 Experimental and calculated (with and without gas–gas interaction) bindingenergies for olefin and paraffin adsorption in kJ mol�1. The numbers are thecontribution of the gas–gas interaction.

Fig. 6 The contributions of each component on total binding energy. Thenumbers correspond to the component denoted below the x-axis: distortion,magnetic transition by MOF host (Mag.h), mutual gas interaction, and totalelectronic effects, from left to right.

PCCP Paper

Publ

ishe

d on

04

Oct

ober

201

3. D

ownl

oade

d by

Kor

ea A

dvan

ced

Inst

itute

of

Scie

nce

& T

echn

olog

y / K

AIS

T o

n 19

/01/

2014

06:

11:2

3.

View Article Online

Page 6: RSC CP C3CP52980K 3. - dspace.kaist.ac.kr€¦ · The application of a metal–organic framework (MOF) has expanded into the area of heterogeneous catalysis, gas storage and separation,

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 19644--19650 19649

36 [1gas]MOF+gas (distorted MOF and gas molecules withoutmagnetic transition); state (iii) is [MOF]MOF+gas in ground statemagnetic ordering and 36 [1gas]MOF+gas (same distorted geo-metries as state (ii) but in ground state magnetic ordering);state (iv) is [MOF]MOF+gas with ground state magnetic orderingand [gas]MOF+gas (MOF host is the same as state (iii) but mutualgas interactions are included); and state (v) is [MOF + gas]MOF+gas

(fully electronically relaxed gas adsorbed MOF, i.e. the finaladsorption state). Therefore, the numbers between each stateare, from left to right, the binding energy contributions of thegeometric distortion, magnetic transition by MOF host (Mag.h),mutual gas interaction (gas–gas), and total electronic effects. Thetotal electronic effects include all interactions between MOF andgas molecules such as charge transfer, polarization and disper-sion, as well as the additional magnetic transition effect byelectronic interaction (Mag.e). Note that the total contributionof magnetic transition is the sum of Mag.h and Mag.e.

The olefin adsorption (solid lines) induces distortionenergy penalties for both the MOF host and gas molecules of40–50 kJ mol�1 and 6–17 kJ mol�1, respectively. This distortiveenergy penalty is most significant for Ay adsorption as expectedfrom the extent of geometry changes (Table S2 in the ESI†). Thislarge distortion is stabilized by the magnetic transition (total2.8–3.3 kJ mol�1), mutual gas interaction (0.9–5.5 kJ mol�1),and mostly by electronic interactions between MOF and gas(90–110 kJ mol�1). The strength of electronic interactions is notablylarger for Ay adsorption, consistent with the distinctive orbitalinteraction picture of Ay as described in the previous section.

In contrast, the paraffin adsorption (dashed lines) does notalter the geometry of the MOF host and gas molecules signifi-cantly; thereby distortive energy penalties are below 1 kJ mol�1.The magnetic transition does not occur for the paraffin adsorp-tion and thus its contribution is zero. While the adsorbedparaffins are mostly stabilized by electronic effects as in the olefins(12–21 kJ mol�1), the mutual gas interactions (1.8–8.0 kJ mol�1)become a more important factor in paraffins as compared to theolefin cases. Also, we identified in the previous section that thecharge transfer between paraffin and host MOF is insignificant,consistent with Verma’s charge analysis results,43 implying thatthe electronic effects in paraffin adsorption is largely attributed tothe long-range interactions.

Table 1 shows the total binding energies with and without aconsideration of experimentally observed magnetic transition.

Note that the intrachain magnetic ordering of the paraffinadsorbed Fe/MOF-74 is identical to that of bare Fe/MOF-74(Fig. 1b), while the olefin adsorption overturns it from FM toAFM (Fig. 1b and c).10 By including this magnetic transition,the binding energies of Ay, Ey and Py increase by 3.3, 3.2 and2.8 kJ mol�1, respectively, where the increments are not negli-gible (6–8%). More importantly, since the magnetic transitionappears only for the olefin adsorption, i.e., guest-dependent,it can be another reason for the high olefin–paraffin selectivityin Fe/MOF-74.

Conclusion

We fully investigated the nature of hydrocarbon binding onFe/MOF-74 with resultant magnetic transitions observed in experi-ments, and identified the factors that determine the separationefficiency. The following findings are noteworthy:

(1) The s(p)-bonding and p-backbonding between the olefinand Fe/MOF-74 synergistically enhance the olefin adsorption inboth electronic and geometric ways. Especially, two degeneratedHOMOs of acetylene increase its interaction with Fe/MOF-74,compared to the other olefins.

(2) For the paraffins, the HOMO does not interact, but insteadthe HOMO � 2 (or HOMO � 1) weakly interacts with the MOFframework. This different binding nature mainly enablesFe/MOF-74 to separate olefins from paraffins, and vice versa.

(3) We show that the mutual gas interactions (1–8 kJ mol�1)must also be considered carefully when evaluating the total bindingenergies of gas molecules in Fe/MOF-74, in particular when con-sidering the selective binding within small energy differences.

(4) The magnetic transition of Fe/MOF-74 appears onlyfor the olefin adsorption, increasing the binding strength ofolefins by B3 kJ mol�1, but not for the paraffins. This selectivemagnetic transition and associated energy scale are not negli-gible but contribute sizably to the olefin separation perfor-mance of Fe/MOF-74.

The present understanding and characterization of thedeciding factors for hydrocarbon separation using MOF-74 asa prototype example give us hints to the way of tuning MOFsfor improved hydrocarbon separation/sorption properties. Forexample, more electronegative functional groups in the organiclinker can selectively increase binding strength for olefins,especially acetylene, due to the increased forward donation tothe more electron deficient Fe atoms owing to the electrondrawing from the linkers. Modifying framework morphology orsteric groups may also be utilized to further aid kinetic separa-tion in conjunction with the present thermodynamic guideline.

Acknowledgements

H. Kim and J. Park contributed equally to this work. Weacknowledge the support from the Korea CCS R&D Center,KAIST EEWS Initiative, and Basic Science Research (2010-0023018) funded by the Ministry of Education, Science andTechnology of Korean government. A generous computing timefrom KISTI is gratefully acknowledged.

Table 1 The calculated binding energies with/without the consideration ofmagnetic transition for each gas adsorption

Adsorbate

Binding energy (kJ mol�1)

Gound state w/o Mag. Tr. Experimentsa

Acetylene 46.4 43.2 47Ethylene 44.4 41.1 45Propylene 48.1 45.3 44

Methane 14.1 14.1 20Ethane 22.5 22.5 25Propane 28.1 28.1 33

a Ref. 10.

Paper PCCP

Publ

ishe

d on

04

Oct

ober

201

3. D

ownl

oade

d by

Kor

ea A

dvan

ced

Inst

itute

of

Scie

nce

& T

echn

olog

y / K

AIS

T o

n 19

/01/

2014

06:

11:2

3.

View Article Online

Page 7: RSC CP C3CP52980K 3. - dspace.kaist.ac.kr€¦ · The application of a metal–organic framework (MOF) has expanded into the area of heterogeneous catalysis, gas storage and separation,

19650 Phys. Chem. Chem. Phys., 2013, 15, 19644--19650 This journal is c the Owner Societies 2013

Notes and references

1 J. Padin and R. T. Yang, Chem. Eng. Sci., 2000, 55, 2607.2 R. B. Eldridge, Ind. Eng. Chem. Res., 1993, 32, 2208.3 D. J. Safarik and R. B. Eldridge, Ind. Eng. Chem. Res., 1998,

37, 2571.4 S. U. Rege, J. Padin and R. T. Yang, AlChE J., 1998, 44, 799.5 G. L. Miessler and D. A. Tarr, Inorganic Chemistry, Prentice

Hall College Div, 1998.6 F. Iucolano, P. Aprea, D. Caputo, C. Colella, M. Eic and

Q. Huang, Adsorption, 2008, 14, 241.7 S. U. Rege and R. T. Yang, Chem. Eng. Sci., 2002, 57, 1139.8 Y.-S. Bae, C. Y. Lee, K. C. Kim, O. K. Farha, P. Nickias,

J. T. Hupp, S. T. Nguyen and R. Q. Snurr, Angew. Chem., Int.Ed., 2012, 51, 1857.

9 Z. Bao, S. Alnemrat, L. Yu, I. Vasiliev, Q. Ren, X. Lu andS. Deng, Langmuir, 2011, 27, 13554.

10 E. D. Bloch, W. L. Queen, R. Krishna, J. M. Zadrozny,C. M. Brown and J. R. Long, Science, 2012, 335, 1606.

11 S. Geier, J. A. Mason, E. Bloch, W. Queen, M. Hudson,C. M. Brown and J. R. Long, Chem. Sci., 2013, 4, 2054.

12 Y. He, R. Krishna and B. Chen, Energy Environ. Sci., 2012,5, 9107.

13 N. Lamia, M. Jorge, M. A. Granato, F. A. Almeida Paz,H. Chevreau and A. E. Rodrigues, Chem. Eng. Sci., 2009,64, 3246.

14 J. W. Yoon, Y.-K. Seo, Y. K. Hwang, J.-S. Chang, H. Leclerc,S. Wuttke, P. Bazin, A. Vimont, M. Daturi, E. Bloch, P. L.Llewellyn, C. Serre, P. Horcajada, J.-M. Greneche, A. E. Rodriguesand G. Ferey, Angew. Chem., Int. Ed., 2010, 122, 6085.

15 Y. He, S. Xiang, Z. Zhang, S. Xiong, F. R. Fronczek, R. Krishna,M. O’Keeffe and B. Chen, Chem. Commun., 2012, 48, 10856.

16 N. L. Rosi, J. Kim, M. Eddaoudi, B. Chen, M. O’Keeffe andO. M. Yaghi, J. Am. Chem. Soc., 2005, 127, 1504.

17 S. R. Caskey, A. G. Wong-Foy and A. J. Matzger, J. Am. Chem.Soc., 2008, 130, 10870.

18 S. Choi, T. Watanabe, T.-h. Bae, D. S. Sholl and C. W. Jones,J. Phys. Chem. Lett., 2012, 3, 1136.

19 D. D’alessandro, B. Smit and J. R. Long, Angew. Chem., Int. Ed.,2010, 49, 6058.

20 E. Haldoupis, S. Nair and D. S. Sholl, J. Am. Chem. Soc., 2012,134, 4313.

21 S. S. Han, D.-H. Jung and J. Heo, J. Phys. Chem. C, 2013,117, 71.

22 K. Sumida, D. L. Rogow, J. A. Mason, T. M. McDonald,E. D. Bloch, Z. R. Herm, T.-H. Bae and J. R. Long, Chem. Rev.,2012, 112, 724.

23 C. E. Wilmer and R. Q. Snurr, Chem. Eng. J., 2010, 171, 775.24 H. Yang, Z. Xu, M. Fan, R. Gupta, R. Slimane, A. Bland and

I. Wright, J. Environ. Sci., 2008, 20, 14.25 A. O. Yazaydin, R. Q. Snurr, T.-H. Park, K. Koh, J. Liu, M. D.

Levan, A. I. Benin, P. Jakubczak, M. Lanuza, D. B. Galloway,J. J. Low and R. R. Willis, J. Am. Chem. Soc., 2009, 131, 18198.

26 B. Chen, N. W. Ockwig, A. R. Millward, D. S. Contreras andO. M. Yaghi, Angew. Chem., Int. Ed., 2005, 44, 4745.

27 M. Dinca and J. R. Long, Angew. Chem., Int. Ed., 2008,47, 6766.

28 S. S. Han, J. L. Mendoza-Cortes and W. a. Goddard,Chem. Soc. Rev., 2009, 38, 1460.

29 Y.-H. Kim, Y. Y. Sun, W. I. Choi, J. Kang and S. B. Zhang,Phys. Chem. Chem. Phys., 2009, 11, 11400.

30 G. J. Kubas, Chem. Rev., 2007, 107, 4152.31 N. L. Rosi, J. Eckert, M. Eddaoudi, D. T. Vodak, J. Kim,

M. O’Keeffe and O. M. Yaghi, Science, 2003, 300, 1127.32 J. L. C. Rowsell and O. M. Yaghi, J. Am. Chem. Soc., 2006,

128, 1304.33 W. Zhou, H. Wu and T. Yildirim, J. Am. Chem. Soc., 2008,

130, 15268.34 M. Eddaoudi, J. Kim, N. Rosi, D. Vodak, J. Wachter,

M. O’Keeffe and O. M. Yaghi, Science, 2002, 295, 469.35 Z. Guo, H. Wu, G. Srinivas, Y. Zhou, S. Xiang, Z. Chen,

Y. Yang, W. Zhou, M. O’Keeffe and B. Chen, Angew. Chem.,Int. Ed., 2011, 50, 3178.

36 S. Noro, S. Kitagawa, M. Kondo and K. Seki, Angew. Chem.,Int. Ed., 2000, 39, 2081.

37 H. Wu, W. Zhou and T. Yildirim, J. Am. Chem. Soc., 2009,131, 4995.

38 S. Xiang, W. Zhou, J. M. Gallegos, Y. Liu and B. Chen, J. Am.Chem. Soc., 2009, 131, 12415.

39 S. Xiang, W. Zhou, Z. Zhang, M. A. Green, Y. Liu andB. Chen, Angew. Chem., Int. Ed., 2010, 49, 4615.

40 X. Rao, J. Cai, J. Yu, Y. He, C. Wu, W. Zhou, T. Yildirim,B. Chen and G. Qian, Chem. Commun., 2013, 49, 6719.

41 X.-j. Hou, P. He, H. Li and X. Wang, J. Phys. Chem. C, 2013,117, 2824.

42 K. Sillar and J. Sauer, J. Am. Chem. Soc., 2012, 134, 18354.43 P. Verma, X. Xu and D. Truhlar, J. Phys. Chem. C, 2013,

117, 12648.44 S. Grimme, J. Comput. Chem., 2006, 16, 1787.45 G. Kresse and J. Furthmuller, Comput. Mater. Sci., 1996,

6, 15.46 J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett.,

1996, 77, 3865.47 B. Hammer, L. Hansen and J. Norskov, Phys. Rev. B: Condens.

Matter Mater. Phys., 1999, 59, 7413.48 S. L. Dudarev, G. A. Botton, S. Y. Savrasov, C. J. Humphreys

and A. P. Sutton, Phys. Rev. B: Condens. Matter Mater. Phys.,1998, 57, 1505.

49 L. Wang, T. Maxisch and G. Ceder, Phys. Rev. B: Condens.Matter Mater. Phys., 2006, 73, 195107.

50 P. E. Blochl, Phys. Rev. B: Condens. Matter Mater. Phys., 1994,50, 17953.

51 K. Momma and F. Izumi, J. Appl. Crystallogr., 2008, 41, 653.52 R. G. Burns, Mineralogical applications of crystal field theory,

Cambridge University Press, 1993.53 J. Goodenough, Phys. Rev., 1960, 117, 1442.54 J. B. Goodenough, Magnetism and the Chemical Bond,

John Wiley & Sons, Inc., New York, 1963.55 J. Park, H. Kim and Y. Jung, J. Phys. Chem. Lett., 2013,

4, 2530.

PCCP Paper

Publ

ishe

d on

04

Oct

ober

201

3. D

ownl

oade

d by

Kor

ea A

dvan

ced

Inst

itute

of

Scie

nce

& T

echn

olog

y / K

AIS

T o

n 19

/01/

2014

06:

11:2

3.

View Article Online