Molecular chemisorption on open metal sites in Cu 3 ( benzenetricarboxylate ) 2 : A

spatially periodic density functional theory study

Taku Watanabe and David S. Sholl

Citation: The Journal of Chemical Physics 133, 094509 (2010); doi: 10.1063/1.3479041
View online: http://dx.doi.org/10.1063/1.3479041
View Table of Contents: http://scitation.aip.org/content/aip/journal/jcp/133/9?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in

Magnetic properties of single 3d transition metals adsorbed on graphene and benzene: A density functional
theory study
J. Appl. Phys. 110, 064303 (2011); 10.1063/1.3636112

Coadsorption of CO and NO on the Cu 2 O ( 111 ) surface: A periodic density functional theory study
J. Chem. Phys. 131, 174503 (2009); 10.1063/1.3251055

Adsorption of transition-metal atoms on boron nitride nanotube: A density-functional study

J. Chem. Phys. 125, 044711 (2006); 10.1063/1.2218841

Metal nanoparticles as models of single crystal surfaces and supported catalysts: Density functional study of size
effects for CO/Pd(111)
J. Chem. Phys. 117, 9887 (2002); 10.1063/1.1516798

Molecular and dissociative chemisorption of NO on palladium and rhodium (100) and (111) surfaces: A density-
functional periodic study
J. Chem. Phys. 108, 6447 (1998); 10.1063/1.476051

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: 147.143.2.5
On: Mon, 22 Dec 2014 04:35:03
 THE JOURNAL OF CHEMICAL PHYSICS 133, 094509 共2010兲

Molecular chemisorption on open metal sites in

Cu3„benzenetricarboxylate…2: A spatially periodic
density functional theory study
Taku Watanabe and David S. Sholla兲
School of Chemical & Biomolecular Engineering, Georgia Institute of Technology,
Atlanta, Georgia 30332-0100, USA
共Received 16 March 2010; accepted 22 July 2010; published online 7 September 2010兲

Chemical interactions of H2O, CO, NO, pyridine, C2H2, H2S, and NH3 with open metal sites in the
metal-organic framework 共MOF兲 Cu3共benzenetricarboxylate兲2 are examined using plane wave
periodic density functional theory 共DFT兲. In the case of single molecule adsorption on a Cu dimer,
NH3 and pyridine have the strongest binding, while NO binds weakly. Binding of pairs of molecules
on a Cu dimer shows significant interaction energies, that is, the binding energy of the pair of
molecules is not a simple summation of the binding energies of each molecule. The effect of
molecular adsorption on the magnetic moments of Cu atoms in the MOF is also examined. Using
the binding energies from DFT calculations, the effects of the pressure and temperature on the
chemisorbed species are investigated. Finally the effect of water adsorption on the elastic behavior
of Cu3共BTC兲2 is described. © 2010 American Institute of Physics. 关doi:10.1063/1.3479041兴

I. INTRODUCTION observed that the as-made sample showed continuous mass

Metal-organic frameworks 共MOFs兲 have been exten-
sively studied for a variety of applications including gas
separation, storage, and catalysis.1 Some MOFs are known to
have metal sites that can be activated by solvent removal and
these sites can play an important role in adsorption
behavior.2,3 These metal sites create the possibility of allow-
ing adsorption of molecules via chemisorption, as distinct
from the physisorption that controls that adsorption of mol-
ecules in many MOFs and other nanoporous materials.
Cu3关benzenetricarboxylate共BTC兲兴2, also known as
HKUST-1, is a well-known MOF with open metal sites.4–9
Cu3共BTC兲2 has Fm3̄m symmetry and a three dimensional
channel of pores, with adjacent square cavities of approxi-
mately 10⫻ 10 Å2 connected by a square window of ⬃7
⫻ 7 Å2.4,10,11 In this MOF, Cu dimers are bridged by BTC
linkers forming a paddlewheel structure. As synthesized,
these Cu sites are saturated by water molecules that can be
removed by heating the sample at moderate temperature
共⬃500 K兲.7,8 After this process, guest molecules can interact
with the open metal sites. Other examples of MOFs that
exhibit open metal sites include CPO-273,12 and MIL-101.13
Sorption properties of Cu3共BTC兲2 for a wide variety of
molecules have been investigated by a number of groups.
Wang et al.14 reported adsorption of N2, O2, CO, CO2, N2O,
CH4, ethylene, ethane, n-dodecane, and H2O. In their experi-
ments, CO uptake was found to be significantly lower
共⬃0.7 mmol/ g at 1 bar兲 than CO2 共⬃5 mmol/ g兲, C2H4
共⬃6.2 mmol/ g兲, and N2O 共⬃5.3 mmol/ g兲, and similar to
those of CH4, N2, and O2. Schlichte et al.7 examined the
thermal stability and catalytic properties of Cu3共BTC兲2. They
a兲
Author to whom correspondence should be addressed. Electronic mail:
david.sholl@chbe.gatech.edu.
loss of 32% up to 523 K. High temperature x-ray diffraction
共XRD兲 indicated that the reflection corresponding to the wa-
ter molecules bound to Cu sites diminished around 473 K.
They also reported that adsorption of gaseous ammonia re-
sulted in an irreversible structural change to another crystal,
but no information was given regarding the structure of the
final product. Xiao et al.15 performed gravimetric adsorption
and IR spectra measurements of NO and H2 on Cu3共BTC兲2.
The NO adsorption experiments were performed at 196 and
296 K, and a significant amount 共⬃2.2 mmol/ g兲 of NO was
found to remain adsorbed after the reduction of gas pressure.
Infrared 共IR兲 spectra at room temperature indicated the for-
mation of Cu2+ – NO adducts by the growth of a new com-
ponent at 1887 cm−1 upon contact with 30 Torr of NO. This
peak is consistent with the peak observed when NO is ad-
sorbed in Cu-ZSM-5 although the width is much narrower.16
Prestipino et al.8 described spectroscopic characterization of
Cu3共BTC兲2 upon activation and adsorption of guest mol-
ecules. Interaction of CO and H2 was investigated using
x-ray adsorption near edge structure 共XANES兲 and IR spec-
troscopy at 77 K. After the adsorption of CO in dehydrated
Cu3共BTC兲2, a substantial change in the XANES pre-edge
feature of Cu2+ species was observed, indicating the modifi-
cation of the first coordination sphere of Cu2+ atoms. The IR
spectra were collected by dosing a CO equilibrium pressure
of 1 Torr and by progressively reducing this pressure to
10−3 Torr. This IR study showed a prominent peak at
2178 cm−1 ascribed to Cu2+ – CO adducts that is consistent
with the same adducts in Cu-ZSM-5.17 As the CO pressure
was decreased, the peak shifted downward to 2174 cm−1 and
almost completely disappeared at 10−3 Torr. Bordiga et al.18
reported additional IR spectra measurements on Cu3共BTC兲2
including CO2 and N2 in addition to CO, NO, and H2.
Cu2+ – NO adducts were indicated by the peak at 1887 cm−1

0021-9606/2010/133共9兲/094509/12/$30.00 133, 094509-1 © 2010 American Institute of Physics

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: 147.143.2.5
On: Mon, 22 Dec 2014 04:35:03
 094509-2 T. Watanabe and D. S. Sholl
that increases proportionally with the coverage and without
substantial shifts. Clear shifts in the adsorbate vibrational
spectra were observed for CO 共+36 cm−1兲, CO2
共−19 cm−1兲, and NO 共+11 cm−1兲 but not for N2 and H2. We
are not aware of any similar studies that examined gas mix-
tures.
Quantum chemistry studies of adsorption of gas mol-
ecules in MOFs are limited.19 A large fraction of the earlier
studies focused on H2 adsorption in MOFs because of the
strong interest in hydrogen storage.20,21 However, this is one
of the most challenging systems to study with DFT because
the interaction is dominated by weak van der Waals forces
where standard DFT is inaccurate. Higher level quantum
chemistry approaches such as Møller–Plesset 共MP兲 theory
are necessary to describe H2-MOF interactions accurately.
Gao and Zeng22 used DFT and the second-order MP 共MP2兲
methods with correlation consistent triple-zeta basis sets 共cc-
pVTZ and aug-cc-pVTZ兲 to study the adsorption on ben-
zenelike linker fragments. They showed that the MP2 result
falls between DFT results with the local density approxima-
tion 共LDA兲 and generalized gradient approximation 共GGA兲,
and the binding energy can change 5 kJ/mol depending on
the H2 molecule orientation. Sagara et al.21,23 also performed
MP2 calculations with a split-valence quadruple-zeta plus
polarization 共QZVPP兲 basis set on IRMOF-1 fragments and
found binding energies of 4–9 kJ/mol on the benzyl linker
areas and 10 kJ/mol on the ZnO fragment.
Fewer studies have been done using quantum chemistry
calculations on adsorption of other gases. Xiang et al.10 per-
formed periodic DFT calculations to determine the binding
energies and geometries of adsorption of acetylene in
Cu3共BTC兲2. They found that acetylene binds to the frame-
work along the O–Cu–O bond with binding energies of
44.5 kJ/mol with LDA and 15.8 kJ/mol with a GGA func-
tional. Dubbeldam et al.24 investigated the interaction of
small molecules including CO2 and N2 in IRMOF-1 by DFT
with PBE/ 6-31Gⴱ and classical force fields. They examined
the effect of distances and angles on the adsorption energies
of the two molecular species and found that the linear mol-
ecules preferred to bind along the organic linker. Ramsahye
et al.25 characterized the CO2 adsorption sites in MIL-53 and
MIL-47 using the PW91-GGA functional in periodic simula-
tion cells. A number of possible adsorption geometries were
identified in the two polymorphs of MIL-53 and adsorption
energies were calculated. CPL-1 and CPL-2 have been stud-
ied by Nagaoka et al.26 using periodic LDA-DFT calculation
to characterize the bond length and angles of O2. They re-
ported a binding energy of ⬃40 kJ/ mol. Most of these stud-
ies focus on the adsorption of weakly binding molecules in
the MOFs. We have previously investigated the adsorption
and diffusion of CO2 / CH4 in Cu 共hfipbb兲0.5 · 共H2hfipbb兲0.5 by
combining molecular simulations and DFT.27 This material
has the interesting property that CO2 diffuses far faster than
CH4 through its pores. DFT is generally reliable for the for-
mation and breaking of chemical bonds. Plane wave DFT is
useful for extended solid materials such as MOFs and this
approach is far less computationally demanding than higher
level quantum chemistry approaches, so it is capable of mod-
eling the large unit cell structures that define many MOF
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: 147.143.2.5
On: Mon, 22 Dec 2014 04:35:03
J. Chem. Phys. 133, 094509 共2010兲
materials. Motivated by these observations, we examined
chemisorption of selected molecules on the open Cu site in
Cu3共BTC兲2 using plane wave DFT.
We will investigate the chemisorption behavior of H2O,
CO, NO, H2S, NH3, C2H2, and pyridine. H2O is a common
solvent used in the synthesis of many MOFs and separating
H2O from a gas mixture is also useful for producing dry
gases. CO and NH3 are important molecules for industrial
applications, but they are also common contaminants for the
metal catalyst on fuel cell anode materials.28 NO is an im-
portant molecule for biological systems and its delivery is
attractive for antibacterial, antithrombotic, and wound-
healing applications.29 H2S is a well-known contaminant in
natural gas and corrodes metallic parts used for transporta-
tion and processing.30 Pyridine was chosen since it is large
enough 共⬃4.9 Å from H to H across the ring兲 to signifi-
cantly modify the channel within Cu3共BTC兲2 but also small
enough to diffuse through the channels of the framework.
Qualitatively, Cu3共BTC兲2 is a Lewis acid,9 so good Lewis
bases such as NH3 and pyridine are expected to bind
strongly.
One possible effect of adsorption of molecules in MOFs
is the change in their mechanical response. A small number
of experiments have already studied the mechanical proper-
ties of Cu3共BTC兲2. Allendorf et al.31 used the adsorbate in-
duced mechanical response of Cu3共BTC兲2 to create a gas
sensing device. In their work, Cu3共BTC兲2 was deposited on a
piezoelectric cantilever and changes in the resistivity were
measured with varying amounts of moisture in the gas and
the presence of gas species including methanol, ethanol, and
CO2. It was found that the device responded more sensitively
to CO2 when Cu3共BTC兲2 is hydrated. Direct measurement of
the elastic property of Cu3共BTC兲2 at high pressure was per-
formed by Chapman et al.32 They measured the relative lat-
tice change of Cu3共BTC兲2 as a function of pressure at pres-
sures up to 10 GPa using a diamond anvil cell. Fluorinert,
isopropanol, and a methanol-ethanol-water 共MEW兲 mixture
were used as the fluid media in these experiments. A clear
change in the pressure dependence of the lattice volume was
observed when MEW or isopropanol was used, but when no
fluid or fluorinert was used, no such change in the mechani-
cal response was found. These studies involve mainly phys-
isorption of molecules, but little is known about the effect of
chemisorption on similar properties. Thus, we investigate the
effect of water adsorption on the elastic property of
Cu3共BTC兲2 using plane wave DFT.
II. METHODS
As discussed above, we performed calculations that used
the fully periodic structure of Cu3共BTC兲2. Our computational
supercell was one rhombohedral primitive cell of
Cu3共BTC兲2, which contains 156 framework atoms in its de-
hydrated form. All calculations with this approach were per-
formed with the Vienna ab initio simulation package33 ver-
sion 4.6 with the PW91-GGA 共Ref. 34兲 exchange-correlation
functional and an energy cutoff of 500 eV. This energy cutoff
was to give well converged total energies for test calculations
with the crystal structure reported by Chui et al.4 Trial cal-
 094509-3 Chemisorption on Cu sites in CuBTC
TABLE I. Crystal structure of dehydrated and hydrated Cu3共BTC兲2 from the
reported experimental values and our DFT optimized structures. All lengths
are in angstrom.
Dehydrated Hydrated
Expt. DFT Expt.
a 26.224a 26.51 26.343b
Cu–Cu 2.50a 2.48 2.682,b 2.64a
Cu–O 1.94a 1.97 1.952,b 1.96a
Cu– OH2 2.165,b 2.19a
a
Reference 8.
b
Reference 4.
culations were performed sampling k-space at the ⌫-point
and using 2 ⫻ 2 ⫻ 2 k-points. The energy difference between
these calculations was less than 3 ⫻ 10−4 eV/ Cu atom. Thus,
all further calculations were performed sampling reciprocal
space at the ⌫-point only. All geometry optimizations were
performed using a conjugate gradient method. The total en-
ergy and ionic force convergence criteria for energy minimi-
zation were 10−4 eV and 3 ⫻ 10−2 eV/ Å, respectively.
Cu3共BTC兲2 is known to show antiferromagnetic cou-
pling within each Cu dimer.35 Thus, spin polarization was
applied in calculations where the atomic magnetic moments
were averaged by the site projected electron density.36 The
system’s overall magnetic moments was relaxed without
constraints on the spin states. Total energy calculations were
performed for dehydrated Cu3共BTC兲2 in nonmagnetic, ferro-
magnetic, and antiferromagnetic states. The lowest energy
state was found to be antiferromagnetic. The total energy of
this spin state was lower than the nonmagnetic state by 0.10
eV/Cu atom.
Table I shows a comparison of the fully hydrated and
dehydrated crystal structures from experiments and our DFT
results in which both the unit cell size and shape and the
position of individual atoms were allowed to relax. The ini-
tial geometries of the hydrated and dehydrated structures
were the ones reported by the x-ray powder diffraction study
共XRPD兲 of Chui et al.4 The experimental values in the table
include the extended x-ray adsorption fine structure 共EX-
AFS兲 study by Prestipino et al.8 in addition to the XRPD
data of Chui et al. Only the lattice parameter and selected
bond lengths are shown. The calculated unit cell angles re-
main at 90.0° after geometry optimization. The unit cell pa-
rameters and bond lengths for the framework atoms are in
good agreement, with deviations of less than 1% between the
DFT and experimental structures. The bond length between
Cu and H2O is larger by 6%–7% in DFT than in the experi-
mental data. A careful XRD study of both single crystal and
powder Cu3共BTC兲2 samples was reported by Wu et al.37
Both samples showed negative thermal expansion of less
than 0.3% between 100 and 450 K. Thus we believe the
effect of thermal expansion relative to comparing our DFT
data to experimental data is negligible in Cu3共BTC兲2.
In our adsorption studies, either a single molecule or a
pair of molecules is placed adjacent to a single Cu dimer.
There are six dimers in a primitive cell, so these calculations
correspond to only adsorbing molecules on a small fraction
of the total metal sites. The atom nearest to the Cu dimer was
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: 147.143.2.5
On: Mon, 22 Dec 2014 04:35:03
J. Chem. Phys. 133, 094509 共2010兲
DFT
26.59
2.60
1.99
2.32
FIG. 1. Molecular structure of a hydrated Cu acetate molecule.
typically placed at the experimentally reported Cu– OH2O dis-
tance 共⬃2.2 Å兲. Geometry optimization was then performed
allowing all atoms to move while keeping the unit cell shape
and volume fixed. In these calculations, the DFT optimized
unit cell volume for the dehydrated material was used. To
test if keeping the unit cell fixed was appropriate, a test cal-
culation with no unit cell constraints was performed with a
pair of H2O molecules coordinated to a Cu dimer. The
change in the energy and volume of the primitive unit cell
from the DFT calculation that held the cell volume fixed was
only ⬃10−3 eV/ Cu and ⬃4 Å3. Therefore, the rest of the
adsorption calculations were performed with fixed unit cell
at the DFT optimized lattice parameter for computational
efficiency.
The total binding energy of a single molecule or a pair of
molecules was defined by
Eb = E共CuBTC兲 + ⌺iEi共mol兲 − E共CuBTC + mol兲,
where E共CuBTC+ mol兲, E共CuBTC兲, and Ei共mol兲 are the to-
tal energies of Cu3共BTC兲2 with adsorbate molecules,
Cu3共BTC兲2 only, and molecule of type i in gas phase, respec-
tively. i can be either 1 or 2, since our calculations were
performed with either a single molecules or a pair of mol-
ecules.
The reliability of plane wave DFT was further examined
by comparing the binding energy of two H2O molecules on a
single Cu dimer in Cu3共BTC兲2 and a localized basis set cal-
culation using a copper acetate 共CuAc兲 molecule. CuAc is
used as a precursor for the synthesis of Cu3共BTC兲2 and has
the same local coordination for the Cu dimer as in
Cu3共BTC兲2.4,38,39 The structure of a hydrated CuAc molecule
is shown in Fig. 1, while two views of the crystal structure of
Cu3共BTC兲2 is shown in Fig. 2. We used the PW91 and
B3LYP functionals with a 6-311+ +Gⴱⴱ basis set in GAUSS-
IAN03 to optimize this structure, using broken symmetry to
account for the antiferromagnetic coupling.40 The bond
length and angles of the optimized molecular structure are
compared in Table II with observations from neutron
diffraction.39 The bond length and angles of the covalent
bonds in the molecule are in excellent agreement between
the experiment and both localized basis set calculations. The
Cu– OH2 bond, however, is 7% 共8%兲 larger in the B3LYP
 094509-4 T. Watanabe and D. S. Sholl J. Chem. Phys. 133, 094509 共2010兲

FIG. 2. Crystal structure of dehydrated Cu3共BTC兲2 view along the 关100兴

direction 共left兲. Carbon, hydrogen, copper, and oxygen atoms are repre-
sented by the blue, white, brown, and red spheres. The paddlewheel struc-
ture of a Cu dimer in Cu3共BTC兲2 with water molecules coordinated to the
metal centers is shown on the right.

共PW91兲 calculation, an outcome very similar to our plane

wave DFT calculation for the crystalline MOF. Several of the FIG. 3. Binding energies of adsorbate molecules on a Cu dimer. Bars with
angles in the hydrated molecule also deviate from the neu- black, white, and gray are for the binding modes of single molecules, two
tron diffraction data. This may be associated with the fact identical molecules, and combinations, respectively.
that the experimental data are for a molecular crystal of hy-
drated CuAc rather than an isolated molecule. Finally, the
III. RESULTS
binding energy of the pair of water molecules was calculated
to be 0.84 共0.85 and 0.89兲 eV from plane wave DFT with the A. Binding energies
PW91 functional for the MOF 共localized basis set calcula- To illustrate our approach to examining the binding of
tions with the B3LYP and PW91 functionals for the CuAc H2O, CO, NO, H2S, C2H2, NH3, and pyridine, we first con-
molecule兲, showing good agreement between the calcula-
sider CO inside the pores of Cu3共BTC兲2. We first performed
tions.
calculations in which the center of a CO molecule was
The majority of our calculations examined molecules
placed in the center of the MOF main cavity with a random
binding to a single dimer in a Cu3共BTC兲2 unit cell. To probe
orientation and then the geometry of the entire system was
the importance of interactions among separate dimers on mo-
optimized as described above. This calculation gave a bind-
lecular adsorption energies, we calculated the adsorption en-
ing energy of only 0.003 eV and no obvious change in the
ergy for each pair of H2O molecules in fully hydrated
structure of Cu3共BTC兲2 or the bond length of CO was de-
Cu3共BTC兲2. This energy was 0.82 eV, which differs only
tected. This indicates, unsurprisingly, that configurations ex-
slightly from the 0.84 eV binding energy calculated when
ist for this small molecule inside the MOF where chemical
H2O bound only to a single dimer. This suggested that the
interactions with the MOF framework are negligible. We
latter calculation is sufficient to describe binding of small
molecules to the Cu sites in Cu3共BTC兲2. note the well-known fact that DFT calculations do not cap-
ture dispersion interactions accurately,41 so the contributions
TABLE II. Bond length and angles of a hydrated CuAc molecule from to the molecular adsorption energy in this and all of the
neutron diffraction 共Ref. 39兲 and localized basis set calculations. The labels configurations we consider do not include these interactions.
of atoms for the bond angles are shown in Fig. 1. X – Oi is the average of the
Figure 3 shows the binding energy of all the adsorbate
distances between atom X and O1, O2, O3, and O4.
molecules and their combinations we examined. When inter-
Expt. B3LYP PW91 action of a single molecule with a Cu dimer is considered,
NH3 and pyridine show the highest values of Eb
Bond length 共Å兲
共⬃0.8 eV兲. This is not surprising, since NH3 and pyridine
Cu–Cu 2.614 2.61 2.53
are strong Lewis bases and Cu in Cu3共BTC兲2 is known to be
Cu– OH2 2.151 2.30 2.32
Cu– Oi 1.992 1.97 1.99 a strong Lewis acid.9 Pyridine and bipyridine are important
C – Oi 1.263 1.27 1.27 linker molecules in MOFs and molecular crystals, so it is
C–C 1.500 1.51 1.52 natural that they show strong binding with the metal centers.
Bond angles 共deg兲 CO, H2O, and H2S have similar binding energies of 0.48,
O1-Cu-O2 89.21 89.47 89.69 0.49, and 0.53 eV, respectively. The weakest binding species
O2-Cu-O3 91.05 90.55 90.52 was NO with 0.30 eV. The binding energy of C2H2 was
Cu-O1-C 121.42 123.2 120.33 found to be 0.38 eV. An earlier theoretical study of acetylene
O1-Cu-Ow 93.14 84.21 81.68 binding on Cu sites by Xiang et al. gave binding energies of
O2-Cu-Ow 91.04 82.97 81.76
0.17 共0.46兲 eV with the PBE-GGA 共LDA兲 functional.10 Our
O3-Cu-Ow 98.12 108.45 105.53
O4-Cu-Ow 93.5 104.33 105.78
value lies in between their reported values, but it is not clear
Cu-Ow-Hw 115.01 97.17 94.23 whether these earlier calculations considered the antiferro-
magnetic nature of the Cu dimers.
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: 147.143.2.5
On: Mon, 22 Dec 2014 04:35:03
 094509-5 Chemisorption on Cu sites in CuBTC
TABLE III. Pair interaction energies from the DFT calculations.
⌬Eb,AA 共eV兲
2 C 2H 2 0.28
2 NO 0.31
2 CO 0.40
2 H 2S 0.32
2 H 2O 0.20
2 NH3 0.26
2 pyridine 0.31
⌬Eb,AB 共eV兲
CO+ NO 0.23
C 2H 2 + H 2O 0.19
NO+ H2O 0.07
CO+ H2O 0.22
H 2S + H 2O 0.19
NH3 + H2O 0.23
pyridine+ H2O 0.20
The binding energies of a single molecule are of limited
use for comparing with available experimental data since the
Cu dimers in the MOF are expected to be saturated in most
situations. Thus, we considered the cases where two mol-
ecules are bound to a Cu dimer. As in the case of single
molecule binding, calculations were initialized by placing
two molecules ⬃2.2 Å away from the Cu atoms in a dimer.
A principal result of these calculations is that the binding
energy of a pair of molecules on a Cu dimer is not simply
twice the binding energy of the individual molecule. This
effect can be quantified for molecules A and B binding to a
Cu dimer by defining the pair interaction energy
⌬Eb,AB = Eb,A + Eb,B + Eb,A+B .
Here, the terms on the right hand side correspond to the
binding energies of a single molecule of A, a single molecule
of B, and the two molecules together. With this definition, a
positive value of ⌬Eb,AB indicates having both molecules
present on the same Cu dimer gives a smaller net binding
energy than the two molecules would give individually. The
value of this pair interaction energy for each combination of
molecules we considered is summarized in Table III. All the
interaction energies lie between 0.19 and 0.40 eV, except for
the case of NO+ H2O, where the value is only 0.07 eV. Thus,
binding energies of a second molecule on a Cu dimer are
always lower than the first molecule.
Comparing the magnitudes of the binding energies, we
can conclude that if two CO, two H2S, two NO, or two C2H2
are adsorbed on a Cu dimer, they can be replaced by two
H2O molecules if a sufficient amount of water vapor is avail-
able. In order to make more careful comparisons with experi-
mental conditions, we consider the effect of temperature and
pressure in Sec. III D.
It is interesting to note that the binding energy of two
NO molecules is almost identical to that of a single NO
molecule. With our definition of the interaction energy, this
means that the interaction energy of two NO molecules is
equal to the binding energy of a single NO molecule. Exam-
ining the optimized geometry of two adsorbed NO molecules
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: 147.143.2.5
On: Mon, 22 Dec 2014 04:35:03
J. Chem. Phys. 133, 094509 共2010兲
showed that only a single NO was bound to the Cu dimer;
the other NO molecule was displaced to the pore of the cage.
This result agrees with the adsorption experiment by Xiao et
al.15 who observed that approximately 2.2 mmol/g of NO
remained strongly bound and was not released from the
sample after heating to 196 or 296 K. This amount is very
close to one NO molecule per Cu dimer.
B. Structure analysis
We now turn to the geometry of adsorbates on the metal
atoms. The bond lengths and angles between adsorbed mol-
ecules and Cu atoms and the separation of Cu atoms within
the Cu dimer are summarized in Table IV. The Cu to mol-
ecule bond length is defined relative to the atom in the mol-
ecule that is closest to Cu, which is O for H2O, C for CO and
C2H2, N for NO, pyridine, and NH3, and S for H2S. For
C2H2, the C–Cu distance is averaged between the two C–Cu
bonds. To characterize the adsorbate angle, we define a mo-
lecular axis for each molecule. For CO, NO, and C2H2, this
axis is simply the C–O, N–O, or C–C bond. For H2O, H2S,
and NH3, the vector is defined by the sum of each O–H,
S–H, or N–H bonds. For pyridine, the axis is from N to the C
on the opposite side of the aromatic ring. The adsorbate
angle is then defined by the angle between the Cu-adsorbate
bond and the adsorbate’s molecular axis. The Cu atoms in
the dimer are distinguished by labeling them as Cu1 and
Cu2.
It is clear from Table IV that the Cu-adsorbate bond
lengths simply reflect the size of the atoms in the adsorbate
molecule except for C2H2, where two atoms bind to the
metal center. H2S and H2O give the largest and second larg-
est values, respectively, as expected from the size of S and O
atoms. NH3 and pyridine are the same distance from the
metal center and are aligned almost perfectly along the axis
of the Cu dimer. The rest of the molecules bind to Cu atoms
at oblique angles. These angles decrease in the order of CO
共162°–165°兲, H2O 共123°–131°兲, NO 共123°–128°兲, and H2S
共96°–108°兲. H2S lies almost parallel to the plane of Cu–O
bonds of the framework. When H2O 共H2S兲 is present, one of
the two O–H 共S–H兲 bonds point toward the nearest O atom
bonded to the Cu atom as a part of the framework. This
orientation indicates formation of a very weak hydrogen
bond, although the distance between the framework O atom
and H in the adsorbate is large 共⬃2.8 Å兲 compared to typi-
cal hydrogen bonds 共⬍2.3 Å兲. In the case of C2H2, the mol-
ecule lies parallel to one of the O–Cu–O bonds along the
diagonal direction in the CuO4 plane. Thus unlike other mol-
ecules we considered, all the atoms in C2H2 contribute to the
binding to the framework. This geometry is also observed by
Xiang et al.10 although that work did not report bond lengths
or angles that would allow a quantitative comparison be-
tween our calculations and their results.
Finally, the Cu1-Cu2 distance is larger than the dehy-
drated MOF in all cases. This distance varies from 2.51 Å for
one NH3 to 2.67 Å for two CO. When CO or pyridine is
bound to the dimer, the Cu1-Cu2 distance is consistently
larger than the other cases. This result does not appear to be
correlated in a simple way to the molecular binding energies.
 094509-6 T. Watanabe and D. S. Sholl J. Chem. Phys. 133, 094509 共2010兲

TABLE IV. Bond lengths and angles between Cu atoms and adsorbate molecules, and Cu atom separation
within the dimer. For the mixed adsorbate, A + B, A and B correspond to mol1 and mol2, respectively. Cu1-Cu2
distance in the dehydrated state is 2.48 Å as shown in Table I.

Cu1-mol1 Cu2-mol2

Length Angle Length Angle Cu1-Cu2

Adsorbate 共Å兲 共deg兲 共Å兲 共deg兲 共Å兲

C 2H 2 2.475 90.0 ¯ ¯ 2.517

NO 1.925 127.2 ¯ ¯ 2.548
CO 2.061 165.2 ¯ ¯ 2.590
H 2O 2.304 131.4 ¯ ¯ 2.522
H 2S 2.636 107.7 ¯ ¯ 2.526
NH3 2.171 178.0 ¯ ¯ 2.505
pyridine 2.174 178.6 ¯ ¯ 2.547
2 C 2H 2 2.561 90.0 2.557 89.9 2.565
2 NO 1.972 129.9 ¯ ¯ 2.551
2 CO 2.189 161.9 2.195 165.1 2.666
2 H 2O 2.336 128.8 2.333 124.5 2.547
2 H 2S 2.685 106.3 2.650 95.6 2.584
2 NH3 2.207 177.6 2.209 174.1 2.547
2 pyridine 2.198 178.1 2.173 177.1 2.620
CO+ NO 2.108 165.3 2.079 123.2 2.602
C 2H 2 + H 2O 2.603 90.4 2.395 128.9 2.562
NO+ H2O 1.998 124.8 2.361 128.4 2.553
CO+ H2O 2.013 124.8 2.379 128.4 2.573
H 2S + H 2O 2.658 97.5 2.306 125.7 2.564
NH3 + H2O 2.219 177.4 2.302 122.6 2.547
Py+ H2O 2.163 178.8 2.320 123.1 2.578

C. Magnetization of Cu atoms The magnetization of the Cu atoms in a dimer and the

As mentioned in Sec. II, each Cu dimer in Cu3共BTC兲2
has antiferromagnetic coupling of the magnetic moments and
long range ferromagnetic order between the dimers.35 The
adsorption of CO or NO is known to quench the magnetic
moment of surrounding metal atoms on a metal surfaces such
as Ni共110兲.42 Our DFT calculations allow us to examine
whether similar effects occur in Cu3共BTC兲2.
The magnetization of Cu atoms and the simulation cell
when a single adsorbate molecule is coordinated with a Cu
dimer are shown in Table V. The magnetization with no ad-
sorbate molecules present is also included. All molecules in
these calculations are bound to Cu1. A range of outcomes
can be seen. When a single NO 共CO兲 is present, the magne-
tization of Cu1 and Cu2 are reduced by 54 共13兲% and 15
共12兲%, respectively. Adsorption of C2H2, H2O, or H2S
hardly alters the magnetization, but NH3 or pyridine in-
creases it by 6%–7%. There is no obvious correlation be-
tween the binding energies of molecules on the Cu sites and
the impact of the adsorbing molecules on the magnetization
of the Cu atoms. The total magnetization of the entire simu-
lation cell is nearly zero as expected from the collinear anti-
ferromagnetic order.
TABLE V. Magnetization of Cu atoms of the dimer and the simulation cell with no adsorbate and single
adsorbate molecules. All values are in Bohr.
simulation cell when two identical molecules are coordinated
with the dimer is shown in Table VI. Essentially the same
trend is observed as for NO, C2H2, H2O, and H2S adsorp-
tion. NO reduces the magnetization by 55% for Cu1 and 11%
for Cu2. C2H2, H2O, and H2S are again found to have no
effect on the magnetization. NH3 and pyridine increased the
magnetization by 6 共6兲% for Cu1 and 6 共5兲% for Cu2. The
only notable difference is that the pair of CO molecules com-
pletely quenched the magnetization of the dimer. The total
magnetization was mostly negligible although a small
change of 0.071 Bohr was observed when a pair of NO was
present.
We also examined the magnetization of Cu dimers and
the simulation cell when pairs of unlike molecules were co-
ordinated to the dimer. Results from these calculations are
summarized in Table VII. In this table, the combination A
+ B refers to a dimer where molecule A 共B兲 is bound to Cu1
共Cu2兲. The trend is, in general, consistent with the previous
results for the adsorption of single and identical pairs of mol-
ecules. One difference is that when NO 共CO兲 is paired with
H2O, the magnetization of the Cu atom bound to the H2O
also decreased by 7 共5兲%. Similarly to the case of the pair of

None C 2H 2 NO CO H 2S H 2O NH3 Pyridine

Cu1 0.488 0.497 0.224 0.426 0.497 0.509 0.523 0.523

Cu2 ⫺0.488 ⫺0.481 ⫺0.415 ⫺0.430 ⫺0.476 ⫺0.483 ⫺0.481 ⫺0.467
Total ⫺0.007 ⫺0.006 0.041 ⫺0.012 ⫺0.015 ⫺0.017 0.032 ⫺0.046

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: 147.143.2.5
On: Mon, 22 Dec 2014 04:35:03
 094509-7 Chemisorption on Cu sites in CuBTC
TABLE VI. Magnetization of Cu atoms of the dimer and the simulation cell with two identical adsorbate
molecules. All values are in the Bohr.
2 C 2H 2 2 NO 2 CO
Cu1 0.501 0.221 ⫺0.003
Cu2 ⫺0.498 ⫺0.420 0.005
Total 0.001 ⫺0.079 0.000
identical molecules, the total magnetization was generally
small and the highest value was observed when NO and H2O
molecules were bound to a Cu dimer.
D. Adsorption behavior of Cu3„BTC…2
Knowledge of binding energies of molecules on Cu sites
is a useful way to qualitatively order the binding strengths of
different molecules, but to make direct predictions about ex-
periments, the amount of adsorption for each species under
specific conditions must be determined. To this end, we per-
formed grand canonical Monte Carlo 共GCMC兲 simulations
of adsorption on the Cu sites in Cu3共BTC兲2. GCMC simula-
tion has been widely used to predict adsorption isotherms in
MOFs in situations where the process occurs via physisorp-
tion and can be modeled using a classical forcefield.6,19,43 In
our calculations, we take a somewhat different approach by
restricting our attention to molecules that bind to Cu sites in
the geometries characterized by the DFT calculations above.
In this case, Cu3共BTC兲2 defines a spatially regular lattice of
adsorption sites, and our DFT calculations provide well-
defined binding energies. We performed calculations using
4 ⫻ 4 ⫻ 4 unit cells, which contained 3072 Cu sites. When
two molecules are bound to a single Cu dimer, the pair in-
teraction energies from the DFT calculations are used to cor-
rect the binding energy of the second molecule. GCMC cal-
culations were performed with standard methods, treating
each gas phase species as an ideal gas.44 It would be possible
to extend these calculations to include bound and phys-
isorbed molecules simultaneously if a forcefield was used to
define the energies of physisorbed molecules, but we have
not pursued this direction in the current work.
Single component isotherms of NO, CO, H2S, H2O,
NH3, and pyridine at 300 and 500 K are shown in Fig. 4. The
partial pressures are set between 10−5 and 1 bar for NO, CO,
H2S, and NH3. The maximum pressure for H2O and pyridine
is limited to 10−2 bar since their saturation pressures at 300
K are roughly 0.05 bar for H2O and 0.03 bar for pyridine.45
Each isotherm is shown in terms of the fractional coverage of
each Cu site. At 300 K, NH3 and pyridine completely satu-
rate the metal sites over the entire pressure range because of
their large binding energies. H2O and H2S also cover a large
TABLE VII. Magnetization of Cu atoms of the dimer and the simulation cell with mixed adsorbate molecules.
All values are in Bohr.
CO+ NO C 2H 2 + H 2O NO+ H2O
Cu1 0.324 0.504 0.261
Cu2 0.389 ⫺0.510 ⫺0.454
Total 0.020 0.002 ⫺0.071
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: 147.143.2.5
On: Mon, 22 Dec 2014 04:35:03
J. Chem. Phys. 133, 094509 共2010兲
2 H 2S 2 H 2O 2 NH3 2 pyridine
0.494 0.500 0.518 0.518
⫺0.496 ⫺0.500 ⫺0.517 ⫺0.509
0.001 0.001 0.000 ⫺0.014
fraction 共0.5–1.0兲 of the adsorption sites, although H2O satu-
rates the sites at much lower pressure of ⬃5 ⫻ 10−3 bar. As
mentioned in the Introduction, it is known experimentally
that at room temperature Cu3共BTC兲2 is normally saturated
with H2O in air.4,7,8 Our calculations predict that at least
down to a water partial pressure of 10−3 bar, 90% of the
sites are occupied by water molecules. This means that when
the relative humidity of room temperature air is only ⬃2%,
most of the Cu sites are already saturated.
The coverage of CO remains constant at 0.5 for almost
the entire range of the pressure, but NO shows a significant
drop below 10−3 bar. If a sufficiently wide pressure range is
considered, all of the molecules except NO show a two step
adsorption profile during single component adsorption. For
NO, the maximum adsorbed amount within this description
is one molecule per dimer, as discussed above. When the
temperature is raised to 500 K, the isotherm for each species
is shifted towards higher pressures. No molecule can saturate
the adsorption sites below 5 ⫻ 10−3 bar. Our calculation
shows that even at this high temperature there is still a water
coverage of 0.2–0.5 for H2O partial pressures between 10−3
and 10−2 bar, and the value falls to nearly zero by 10−4 bar.
Dehydration of Cu3共BTC兲2 has been investigated by several
experimental groups and it is known to happen between 450
and 550 K.7,8 The EXAFS study by Prestipino et al.8 sug-
gests that a small fraction of Cu sites may still be coordi-
nated with H2O even after the dehydration at 453 K. Our
calculations are qualitatively consistent with these observa-
tions.
Our GCMC simulations make it straightforward to pre-
dict the composition of molecules bound to Cu sites in equi-
librium with multicomponent gas phase mixtures. We calcu-
lated the coverage of NO– H2O, CO– H2O, C2H2 – H2O,
H2S – H2O, NH3 – H2O, pyridine-H2O, and CO–NO mixtures
at 300 and 500 K using mixture GCMC simulations. The
coverage of NO in NO– H2O and NH3 in NH3 – H2O mix-
tures at 300 K are shown in Fig. 5; these two examples
illustrate the behavior that occurs for most of the mixtures
we considered. In the NO– H2O mixture, not all of the ad-
sorption sites are occupied by adsorbate molecules in the low
pressure region. When the partial pressures of both NO and
CO+ H2O H 2S + H 2O NH3 + H2O Py+ H2O
0.431 0.501 0.525 0.541
⫺0.463 ⫺0.495 ⫺0.493 ⫺0.483
0.012 0.018 ⫺0.037 ⫺0.029
 094509-8 T. Watanabe and D. S. Sholl
H2O are low, there are sufficient open metal sites to which
each adsorbate can bind so that the coverage of the NO is
nearly independent of the partial pressure of H2O. As the
partial pressure of each species increases, all the sites are
filled and there is competition between NO and H2O. How-
ever, because of the higher binding energy of H2O, H2O
dominates the coverage over NO molecules at the same par-
tial pressure. The same trend is observed in C2H2 – H2O,
CO– H2O, and H2S – H2O mixtures although the entire iso-
therm is shifted because of the difference in the binding en-
ergies. In the NH3 – H2O gas mixture, the coverage of NH3 is
independent of the partial pressure of NH3 in this pressure
range because of the large difference in the binding energies
of NH3 and H2O. Even though the binding energy of a pair
of NH3 is higher than that of either a pair of H2O or
NH3 – H2O pair, a significant amount of H2O molecules bind
at high water partial pressures. This indicates that the chemi-
cal potential of H2O is large enough in the range from 10−3
to 10−2 bar to partially displace the strongly bound NH3
molecules. The coverage of pyridine in the pyridine-H2O
mixture is almost indistinguishable from that of NH3 in the
NH3 – H2O mixture.
A qualitatively different type of behavior occurs for the
CO–NO mixture. The coverages of CO and NO in this mix-
ture at 300 K are shown in Fig. 6. In this case, not all of the
adsorption sites are filled by the adsorbate molecules. In the
low pressure region, there is competition between CO and
NO and only one molecule is bound to each dimer. As the
partial pressures increase, the competition becomes between
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: 147.143.2.5
On: Mon, 22 Dec 2014 04:35:03
J. Chem. Phys. 133, 094509 共2010兲
FIG. 4. Calculated single component
adsorption isotherms of NO, CO,
C2H2, H2S, H2O, NH3, and pyridine at
300 K 共left兲 and 550 K 共right兲.
CO–NO and CO–CO pairs, resulting in a change in the trend
of the coverage of the metal sites. Complete coverage of the
metal sites is reached only at the highest pressures we con-
sidered.
At 500 K, a large fraction of the sites is not filled at low
pressures in the mixture isotherms we examined. Figure 7
shows the coverages of H2S, H2O, and NH3 in the H2S – H2O
and NH3 – H2O mixtures. The coverages of NO– H2O,
CO– H2O, and C2H2 – H2O are qualitatively the same as that
of H2S – H2O. The maximum coverage of the adsorbate mol-
ecules in the H2S – H2O mixture is about 0.65, even at the
highest pressure we examined. The coverage in the
NH3 – H2O mixture is essentially the coverage of the single
component NH3, and H2O can occupy only a small fraction
of the coverage even at highest H2O partial pressure exam-
ined. The same qualitative trend as H2S – H2O was also ob-
served in the CO–NO gas mixture as shown in Fig. 8.
E. Effect of H2O adsorption on the elastic behavior
of Cu3„BTC…2
In the previous sections, we have focused on the inter-
action of small molecules with the open metal sites in
Cu3共BTC兲2 under equilibrium conditions. As introduced in
Sec. I, the mechanical behavior of Cu3共BTC兲2 is known to be
affected by the adsorption of molecules. Because the poten-
tial effect of chemically bound species on this behavior is not
FIG. 5. Calculated coverages of NO in
NO– H2O binary mixtures 共left兲 and
NH3 in NH3 – H2O 共right兲 binary mix-
tures at 300 K.
 094509-9 Chemisorption on Cu sites in CuBTC
well understood, we briefly consider how the presence of
chemisorbed H2O ligands affects the elastic properties of this
MOF.
Elements of an elastic constant tensor of a crystalline
solid are related to the element i of the stress, ␴i, and strain,
␧i, tensors by Hooke’s law,
␴i = Cij␧j ,
where i and j run from 1 to 6. With this notation, the elastic
stress ␴1 is the uniaxial stress in the x direction on the plane
perpendicular to the x axis, ␴xx. Similarly, index values of
共1, 2, 3, 4, 5, and 6兲 correspond to 共xx, yy, zz, xy, yz, and zx兲,
respectively. For the materials with cubic symmetry such as
Cu3共BTC兲2, there are only three independent nonzero elastic
constants, C11, C12, and C44.
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: 147.143.2.5
On: Mon, 22 Dec 2014 04:35:03
J. Chem. Phys. 133, 094509 共2010兲
FIG. 6. Calculated coverages of CO
共left兲 and NO 共right兲 in the CO–NO
binary mixtures at 300 K.
We examined the elastic constants of Cu3共BTC兲2 in both
its dehydrated and completely hydrated state using plane
wave DFT. For the hydrated sample, every Cu atom is coor-
dinated with a H2O molecule. Our rhombohedral primitive
cell is aligned so that the corresponding cubic unit cell edges
are along the x, y, and z axes to simplify the calculation. As
a result, C11 and C12 are determined by applying uniaxial
strain in the range of ⫺0.015 and 0.015 and calculating the
uniaxial stress in the x and y directions. C44 is obtained by
applying the shear strain in the x-y plane along the x direc-
tion and calculating the corresponding shear stress. Although
the simulation cell is fixed at each strain, all the atom posi-
tions are allowed to optimize to find the minimum energy
configuration.
FIG. 7. Calculated coverages of H2S
共top left兲 and H2O 共top right兲 in
H2S – H2O binary mixtures, and NH3
共bottom left兲 and H2O 共bottom right兲
in NH3 – H2O binary mixtures at
500 K.
 094509-10 T. Watanabe and D. S. Sholl J. Chem. Phys. 133, 094509 共2010兲

FIG. 8. Calculated coverages of CO

共left兲 and NO 共right兲 in CO–NO binary
mixtures at 500 K.

Figure 9 shows the stress tensors as functions of strains molecules on the Cu sites on Cu3共BTC兲2. Based on the DFT
for both dehydrated and hydrated Cu3共BTC兲2. The elastic results, we further investigated the effects of temperature,
constants are obtained from a linear fit of the data to Hooke’s pressure, and presence of other molecules using GCMC
law and the values are shown in Table VIII. C11, C12, and C44 simulations.
for dehydrated 共hydrated兲 Cu3共BTC兲2 are 31.8 共29.2兲, 21.2
共21.5兲, and 12.7 共14.7兲, respectively, with all values in GPa.
Thus, when the MOF is completely hydrated, C11 decreases
by 8.2%, but C12 and C44 increase by 1.4% and 15.7%. These
values of the elastic constants are similar to calculated values
for MOF-5 共C11 = 29.2 GPa, C12 = 13.1 GPa, and C44
= 1.4 GPa by Samanta et al.46 using LDA-DFT, C11
= 27.8 GPa, C12 = 10.6 GPa, and C44 = 3.6 GPa by Bahr
et al. using GGA-DFT, and E = 14.9 GPa by Greathouse and
Allendorf using a classical model兲.51 The reported shear
modulus for MOF-5 is unusually small, but it is not seen in
our calculations for Cu3共BTC兲2.
The bulk modulus of a cubic crystal is 共C11 + 2C12兲 / 3.
Our calculations give a bulk modulus for dehydrated 共hy-
drated兲 Cu3共BTC兲2 of 24.7 GPa 共24.0 GPa兲. The small de-
crease 共2.8%兲 in the bulk modulus upon hydrating the
sample indicates that the MOF becomes slightly softer under
isostatic strain. At the same time, it becomes harder to shear,
as indicated by the increase in C44. We separately calculated
the bulk modulus of dehydrated Cu3共BTC兲2 by applying iso-
static strain. The simulation cell was strained from +1 to
⫺1% in each direction along the unit cell parameters and the
total energies were calculated by DFT. The total energies
were fitted to the Birch–Murnaghan equation of state.47 This
approach gives B = 24.4 GPa, which is in good agreement
with the value obtained from the uniaxial strain. Our values
of the bulk modulus are lower than the experimental value
共⬃30.7 GPa兲 reported by Chapman et al.,32 but the bulk
modulus reported in these experiments showed quite large
variations as the contacting fluid was varied. For example, a
sample with isopropanol shows 25.9 GPa, while MEW can
give as high as 41.9 GPa. It is of course possible that these
diamond anvil cell experiments also measure effects due to
grain interactions or other defects that cannot be probed by
our calculations with a defect-free crystal.

IV. DISCUSSION

We performed DFT calculations to determine the trend FIG. 9. Calculated uniaxial and shear stress-strain from dehydrated 共filled
in the strength of bonds and geometry of the selected small circles兲 and hydrated 共open circles兲 Cu3共BTC兲2.
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: 147.143.2.5
On: Mon, 22 Dec 2014 04:35:03
 094509-11 Chemisorption on Cu sites in CuBTC
TABLE VIII. Elements of elastic constants, bulk modulus, and Young’s modulus of Cu3共BTC兲2 determined by
uniaxial and shear strain using plane wave DFT.
C11
共GPa兲
Dehydrated 31.8
Hydrated 29.2
One of our conclusions is that CO bonds with the Cu2+
metal center and it can occupy some fraction of Cu sites even
in the presence of some water. The IR study by Bordiga et al.
clearly shows binding of CO on unsaturated Cu sites.18 How-
ever, there is no other experimental confirmation of strong
binding of CO in Cu3共BTC兲2. Adsorption experiments by
Wang et al.14 with CO show a typical type I isotherm with
weak physisorption. Our results are inconsistent with this
experimental observation. One possible explanation for this
is that, in their adsorption experiment, H2O molecules were
not completely removed and kinetic limitations keep the CO
molecules from binding to the metal sites. Liu et al.48
pointed out that the pore volume from the sample of Wang
et al. is much smaller than theoretical values for an ideal
crystal and experimental values from carefully prepared
samples.
To the best of our knowledge, no one has explored the
possibility of mixed binding modes. In the case of NO ad-
sorption, we predict that a small fraction of NO was able to
adsorb on the metal sites even in the presence of water. Even
without the presence of the water, NO was able to cover only
half of all the adsorption sites. This result is consistent with
the experimental adsorption study by Xiao et al.15
Among all the molecules investigated, NH3 and pyridine
showed the highest binding energies. Unfortunately, there is
no experimental report on adsorption of NH3 in Cu3共BTC兲2
to which we can compare. Schlichte et al.7 briefly mentioned
in their report that they observed an irreversible transforma-
tion when NH3 is introduced in Cu3共BTC兲2, however, there
is no further information on the final structure. DFT can be
used to find a ground state geometry near the starting con-
figuration, but it is not capable of predicting complete struc-
tural transformations. We performed a preliminary calcula-
tion by covering all binding sites with NH3 molecules. All
the simulation cell parameters and atom positions are al-
lowed to relax using the same convergence criteria as the
completely hydrated structure. The final structure showed
isotropic contraction of 4.2% in volume, resulting from the
reduction of Cu–O bonds by 1.9%.
Besides the direct adsorption and separation of gas mol-
ecules, strongly binding species can be used to tune some
properties of MOFs.49 This functionalization approach has
been examined in a number of experiments. For example,
Bae et al. examined the effect of pyridine-CF3 functionaliza-
tion in Zn2共DMF兲2, where DMF= dimethylformamide.50
They found pyridine-CF3 does not improve the affinity to
CO2 compared to DMF functionalized MOF at room tem-
perature between 0 and 18 bar. This experiment demon-
strated that it is possible to functionalize MOFs with rela-
tively large molecules. The strong binding of pyridine in
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: 147.143.2.5
On: Mon, 22 Dec 2014 04:35:03
J. Chem. Phys. 133, 094509 共2010兲
C12 C44 B E
共GPa兲 共GPa兲 共GPa兲 共GPa兲
21.2 12.7 24.7 14.9
21.5 14.7 24.0 11.0
Cu3共BTC兲2 may be useful for functionalization of
Cu3共BTC兲2 since pyridine is relatively large compared to the
size of the cage of the MOF. By controlling the available
pore volume, it may be possible to control the adsorption or
separation behavior of specific gas mixtures.
The elastic constants of Cu3共BTC兲2 were found to be
similar to those of IRMOF-1 previously reported by Samanta
et al.46 Although there is no direct measurement of the elastic
constants for single crystal Cu3共BTC兲2 available, the small
change in the bulk modulus is in line with the experimentally
observed change in the unit cell volume.
Our approach of using DFT and GCMC is not specific to
Cu3共BTC兲2 and should be useful for describing chemisorp-
tion of molecules in MOFs in general. This type of DFT
calculation should be applied with the caveat that it is not
capable of accurately modeling the interaction of molecules
with MOFs through van der Waals forces. We have shown
how binding energies obtained from DFT can be used within
GCMC simulations to give information about the coverages
of chemically bound species as functions of temperature and
partial pressure.
ACKNOWLEDGMENTS
We thank Dr. Thomas Manz for his assistance with the
B3LYP calculations and Professor Krista Walton for helpful
discussion.
1
J. R. Li, R. J. Kuppler, and H. C. Zhou, Chem. Soc. Rev. 38, 1477
共2009兲.
2
C. Janiak, Dalton Trans. 14, 2781 共2003兲; S. Kitagawa, R. Kitaura, and
S. Noro, Angew. Chem., Int. Ed. 43, 2334 共2004兲.
3
J. Lee, O. K. Farha, J. Roberts, K. A. Scheidt, S. T. Nguyen, and J. T.
Hupp, Chem. Soc. Rev. 38, 1450 共2009兲.
4
S. S.-Y. Chui, S. M.-F. Lo, J. P. H. Charmant, A. G. Orpen, and I. D.
Williams, Inorg. Chem. Commun. 283, 1148 共1999兲.
5
J. L. C. Rowsell and O. M. Yaghi, J. Am. Chem. Soc. 128, 1304 共2006兲.
6
J. R. Karra and K. S. Walton, Langmuir 24, 8620 共2008兲.
7
K. Schlichte, T. Kratzke, and S. Kaskel, Microporous Mesoporous Mater.
73, 81 共2004兲.
8
C. Prestipino, L. Regli, J. G. Vitillo, F. Bonino, A. Damin, C. Lamberti,
A. Zecchina, P. L. Solari, K. O. Kongshaug, and S. Bordiga, Chem.
Mater. 18, 1337 共2006兲.
9
L. Alaerts, E. Seguin, H. Poelman, F. Thibault-Starzyk, P. A. Jacobs, and
D. E. De Vos, Chem.-Eur. J. 12, 7353 共2006兲.
10
S. C. Xiang, W. Zhou, J. M. Gallegos, Y. Liu, and B. L. Chen, J. Am.
Chem. Soc. 131, 12415 共2009兲.
11
E. Haldoupis, S. Nair, and D. S. Sholl, J. Am. Chem. Soc. 132, 7528
共2010兲.
12
A. C. McKinlay, B. Xiao, D. S. Wragg, P. S. Wheatley, I. L. Megson, and
R. E. Morris, J. Am. Chem. Soc. 130, 10440 共2008兲; F. Bonino, S. Cha-
van, J. G. Vitillo, E. Groppo, G. Agostini, C. Lamberti, P. D. C. Dietzel,
C. Prestipino, and S. Bordiga, Chem. Mater. 20, 4957 共2008兲.
13
S. Horike, M. Dinca, K. Tamaki, and J. R. Long, J. Am. Chem. Soc. 130,
5854 共2008兲.
14
Q. M. Wang, D. M. Shen, M. Bulow, M. L. Lau, S. G. Deng, F. R. Fitch,
 094509-12 T. Watanabe and D. S. Sholl J. Chem. Phys. 133, 094509 共2010兲

N. O. Lemcoff, and J. Semanscin, Microporous Mesoporous Mater. 55,
217 共2002兲.
15
B. Xiao, P. S. Wheatley, X. B. Zhao, A. J. Fletcher, S. Fox, A. G. Rossi,
I. L. Megson, S. Bordiga, L. Regli, K. M. Thomas, and R. E. Morris, J.
Am. Chem. Soc. 129, 1203 共2007兲.
16
C. Lamberti, M. Salvalaggio, S. Bordiga, F. Geobaldo, G. Spoto, A.
Zecchina, G. Vlaic, and M. Bellatreccia, J. Phys. IV 7, 905 共1997兲.
17
K. Hadjiivanov and H. Knozinger, J. Catal. 191, 480 共2000兲.
18
S. Bordiga, J. G. Vitillo, G. Ricchiardi, L. Regli, D. Cocina, A. Zecchina,
Bj. Arstad, M. Bjørgen, J. Hafizovic, and K. P. Lillerud, J. Phys. Chem. B
109, 18237 共2005兲.
19
S. Keskin, J. Liu, R. B. Rankin, J. K. Johnson, and D. S. Sholl, Ind. Eng.
Chem. Res. 48, 2355 共2009兲.
20
T. Mueller and G. Ceder, J. Phys. Chem. B 109, 17974 共2005兲; F. M.
Mulder, T. J. Dingemans, M. Wagemaker, and G. J. Kearley, Chem.
Phys. 317, 113 共2005兲; T. Sagara, J. Klassen, and E. Ganz, J. Chem.
Phys. 121, 12543 共2004兲; S. S. Han, J. L. Mendoza-Cortes, and W. A.
Goddard, Chem. Soc. Rev. 38, 1460 共2009兲.
21
T. Sagara, J. Klassen, J. Ortony, and E. Ganz, J. Chem. Phys. 123,
014701 共2005兲.
22
Y. Gao and X. C. Zeng, J. Phys.: Condens. Matter 19, 386220 共2007兲.
23
T. Sagara, J. Ortony, and E. Ganz, J. Chem. Phys. 123, 214707 共2005兲.
24
D. Dubbeldam, H. Frost, K. S. Walton, and R. Q. Snurr, Fluid Phase
Equilib. 261, 152 共2007兲.
25
N. A. Ramsahye, G. Maurin, S. Bourrelly, P. L. Llewellyn, C. Serre, T.
Loiseau, T. Devic, and G. Ferey, J. Phys. Chem. C 112, 514 共2008兲.
26
M. Nagaoka, Y. Ohta, and H. Hitomi, Coord. Chem. Rev. 251, 2522
共2007兲.
27
T. Watanabe, S. Keskin, S. Nair, and D. S. Sholl, Phys. Chem. Chem.
Phys. 11, 11389 共2009兲.
28
X. Cheng, Z. Shi, N. Glass, L. Zhang, J. J. Zhang, D. T. Song, Z. S. Liu,
H. J. Wang, and J. Shen, J. Power Sources 165, 739 共2007兲; J. J. Baschuk
and X. G. Li, Int. J. Energy Res. 27, 1095 共2003兲.
29
H. F. Zhu, B. Ka, and F. Murad, World J. Surg. 31, 624 共2007兲; M. R.
Miller and I. L. Megson, Br. J. Pharmacol. 151, 305 共2007兲.
30
A. J. Kidnay and W. R. Rarrish, Fundamentals of Natural Gas Processing
共CRC, Boca Raton, 2006兲.
31
M. D. Allendorf, R. J. T. Houk, L. Andruszkiewicz, A. A. Talin, J. Pikar-
sky, A. Choudhury, K. A. Gall, and P. J. Hesketh, J. Am. Chem. Soc.
130, 14404 共2008兲.
32
K. W. Chapman, G. J. Halder, and P. J. Chupas, J. Am. Chem. Soc. 130,
10524 共2008兲.
33
G. Kresse and J. Furthmuller, Phys. Rev. B 54, 11169 共1996兲.
34
J. P. Perdew and Y. Wang, Phys. Rev. B 45, 13244 共1992兲.
35
A. Poppl, S. Kunz, D. Himsl, and M. Hartmann, J. Phys. Chem. C 112,
2678 共2008兲; X. X. Zhang, S. S. Y. Chui, and I. D. Williams, J. Appl.
Phys. 87, 6007 共2000兲.
36
J. Hafner, J. Comput. Chem. 29, 2044 共2008兲.
37
Y. Wu, A. Kobayashi, G. J. Halder, V. K. Peterson, K. W. Chapman, N.
Lock, P. D. Southon, and C. J. Kepert, Angew. Chem., Int. Ed. 47, 8929
共2008兲.
38
J. N. van Niekerk and F. R. L. Schoening, Nature 共London兲 171, 36
共1953兲; O. Shekhah, H. Wang, D. Zacher, R. A. Fischer, and C. Woll,
Angew. Chem., Int. Ed. 48, 5038 共2009兲.
39
G. M. Brown and R. Chidamba, Acta Crystallogr., Sect. B: Struct. Crys-
tallogr. Cryst. Chem. B29, 2393 共1973兲.
40
M. J. Frisch, G. W. Trucks, H. B. Schlegel et al., GAUSSIAN 03, Revision
B.05, Gaussian Inc., Pittsburgh, PA, 2003; L. Noodleman, J. Chem. Phys.
74, 5737 共1981兲.
41
D. S. Sholl and J. A. Steckel, Density Functional Theory: A Practical
Introduction 共Wiley, Hoboken, 2009兲; R. M. Martin, Electronic Structure:
Basic Theory and Practical Methods 共Cambridge University Press, Cam-
bridge, 2004兲.
42
Q. Ge, S. J. Jenkins, and D. A. King, Chem. Phys. Lett. 327, 125 共2000兲;
S. J. Jenkins, Q. Ge, and D. A. King, Phys. Rev. B 64, 012413 共2001兲.
43
S. Keskin and D. S. Sholl, Ind. Eng. Chem. Res. 48, 914 共2009兲.
44
D. Frenkel and B. Smit, Understanding Molecular Simulation: From Al-
gorithms to Applications, 2nd ed. 共Academic, San Diego, 2002兲; S. Q.
Hao, M. Widom, and D. S. Sholl, J. Phys.: Condens. Matter 21, 11
共2009兲.
45
J. G. Speight, Lange’s Handbook of Chemistry, 16th ed. 共McGraw-Hill,
New York, 2005兲.
46
A. Samanta, T. Furuta, and J. Li, J. Chem. Phys. 125, 084714 共2006兲.
47
F. Birch, Phys. Rev. 71, 809 共1947兲; F. D. Murnaghan, Am. J. Math. 59,
235 共1937兲.
48
J. Liu, J. T. Culp, S. Natesakhawat, B. C. Bockrath, B. Zande, S. G.
Sankar, G. Garberoglio, and J. K. Johnson, J. Phys. Chem. C 111, 9305
共2007兲.
49
Z. Q. Wang and S. M. Cohen, Chem. Soc. Rev. 38, 1315 共2009兲; H. X.
Deng, C. J. Doonan, H. Furukawa, R. B. Ferreira, J. Towne, C. B. Kno-
bler, B. Wang, and O. M. Yaghi, Science 327, 846 共2010兲.
50
Y. S. Bae, O. K. Farha, J. T. Hupp, and R. Q. Snurr, J. Mater. Chem. 19,
2131 共2009兲.
51
J. A. Greathouse and M. D. Allendorf, J. Phys. Chem. C 112, 5795
共2008兲.

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: 147.143.2.5
On: Mon, 22 Dec 2014 04:35:03