PHYSICAL REVIEW B 97, 155428 (2018)

Nitrogen doping and CO2 adsorption on graphene: A thermodynamical study

Michele Re Fiorentin,1,* Roberto Gaspari,2,3 Marzia Quaglio,1 Gulia Massaglia,1,4 and Guido Saracco1
1
Center for Sustainable Future Technologies Istituto Italiano di Tecnologia, corso Trento 21, 10129 Torino, Italy
2
CompuNet, Istituto Italiano di Tecnologia, via Morego 30, 16163 Genova, Italy
3
Department of Nanochemistry Istituto Italiano di Tecnologia, via Morego 30, 16163 Genova, Italy
4
Department of Applied Science and Technology Politecnico di Torino, corso Duca degli Abruzzi 24, 10129 Torino, Italy

(Received 19 January 2018; revised manuscript received 9 April 2018; published 23 April 2018)

Nitrogen-doped graphene has raised considerable interest for its possible applications as carbon dioxide
adsorber and catalyst. In this paper, we provide a theoretical study of graphitic, pyridiniclike and pyrroliclike
nitrogen defects in a free-standing graphene layer, focusing on their formation and adsorption behavior. Using
density functional theory and thermodynamics, we analyze the various defects, highlighting the great stability
of graphitic nitrogen in a wide temperature and pressure range. CO2 adsorption proves to be moderately
thermodynamically disfavored around standard conditions for the most stable nitrogen defects and slightly favored
for the more energetic ones. The combination of the results on defect stability and CO2 adsorption may open
interesting possibilities in the design of carbon-based materials with promising adsorption performances.

DOI: 10.1103/PhysRevB.97.155428

I. INTRODUCTION We obtain the stability pattern of the considered defects, at

varying temperature and pressure, confirming the high stability
Beyond its unique electronic features, graphene has recently
of substitutional, graphitic nitrogen in a very wide range of
captured researchers’ attention for its possible employments
conditions. For each defect we also obtain the CO2 adsorption
in new environmentally-friendly technologies. In particular,
geometries and study the adsorption Gibbs free energy land-
graphene can be directly used for carbon dioxide capture
scape vs temperature and pressure. CO2 adsorption appears to
and electrocatalysis, or taken as a first theoretical setup for
be moderately disfavored for the most stable nitrogen defects,
approaching the study of various more complex metal-free,
while being slightly favored for the more energetic ones. Fi-
carbon-based materials, for the same purpose [1–6]. In this
nally, a simple adsorption model shows how the CO2 coverage
respect, graphene can be employed in its pristine form [7,8],
on the defects can be varied by changing the thermodynamical
decorated with nanoparticles [9,10] or after a doping treat-
variables. In addition to their intrinsic theoretical content, these
ment with other elements [11–16]. Among nonmetal dopants,
results can serve as the ground on which to develop a precise
nitrogen appears to play a very important role, both from the
phenomenology of nitrogen-doped carbon-based materials in
theoretical and applied point of view. Different configurations
their applications for CO2 adsorption.
of nitrogen atoms in the defected graphene layer have been
This work is organized as follows. In Sec. II we shall outline
taken into account in the literature and their formation energies
the theoretical framework and the computational details for
and electronic properties have been investigated [17–20].
the calculation of the formation and adsorption Gibbs free
While these results are of crucial importance, a key step for
energies, in Sec. III we will present and discuss the results
a thorough understanding of nitrogen defects in graphene and
obtained, while in Sec. IV we will comment on the possible
their behavior towards CO2 is represented by the study of the
sources of error and draw the conclusions.
thermodynamical stability of the defects and their adsorption
properties. In this paper, we aim at extending the already known
zero-temperature results by providing a complete study of the II. METHODS AND COMPUTATIONAL DETAILS
Gibbs free energy of formation for a large class of nitrogen de-
The aim of this paper is to compute the Gibbs free energies
fects and, for each of them, the Gibbs free energy of adsorption
of nitrogen defects formation in graphene and CO2 adsorption,
of a CO2 molecule in a wide temperature and pressure range.
by employing DFT and thermodynamics. The DFT compu-
The information obtained from density functional theory
tations are performed in the plane-wave pseudopotential ap-
(DFT) calculations are complemented with thermodynamics
proach with the QUANTUM ESPRESSO suite [23,24], employing
and statistical mechanics to predict the behavior of the N-doped
the PWSCF code for total energies calculations and the routines
graphene and its adsorption properties at any temperature and
in the PHONON package for phonon dispersions and normal
pressure condition. This will allow an exhaustive analysis of
modes calculations. The details of the DFT setup are reported
defect stability at different temperatures and pressures, and
below for each calculation.
a detailed investigation of carbon dioxide adsorption, going
beyond the pristine graphene studies [7,21,22].
A. Gibbs free energy of defects formation
We address the formation of nitrogen defects in graphene
*
[email protected] considering the six configurations in Fig. 1: graphitic-N (a),

2469-9950/2018/97(15)/155428(7) 155428-1 ©2018 American Physical Society

MICHELE RE FIORENTIN et al. PHYSICAL REVIEW B 97, 155428 (2018)

removed (negative) with respect to the pristine layer and μC

and μN are the carbon and nitrogen atom chemical potentials.
The volume term, pV f , can be safely neglected in this
calculation.
The Helmholtz free energy is given by the total electronic
energy, obtained by standard DFT calculations, and an entropic
term which, considering fixed the layer configuration, amounts
(a) (b)
to the total free energy of vibration. This in principle must be
computed for each layer and compared to pristine graphene.
Nevertheless, the difference in vibrational free energy between
the layers is expected to be small compared to the other terms in
Eq. (1). Indeed, an explicit calculation for graphitic-N shows
that this difference is smaller than 0.1 eV for T  2000 K.
Therefore, neglecting this contribution does not alter the
stability pattern (as will be clear in the following section) and
allows us to compute the Gibbs free energies of formation as
f
(c) (d) GX (T ,p) = EXtot − Eptot − nC μC (T ) − nN μN (T ,p), (2)
where EXtot and Eptot are the total electronic energies obtained
from DFT. The chemical potential of the carbon atom is
obtained from pristine, monolayer graphene, simulated in its
unit cell with Nat = 2 atoms as
1  tot 
μC (T ) = EC + FCvib (T ) , (3)
Nat

(e) (f) where ECtot is the total DFT energy and FCvib (T ) the free energy
of vibration of the pure graphene unit cell.
FIG. 1. Relaxed geometries of the nitrogen defects in graphene The chemical potential of the nitrogen atom
considered: (a) graphitic-N, (b) pyriN1, (c) pyriN2, (d) pyriN3, is obtained from the nitrogen gas molecule as
(e) pyriN4, and (f) pyrrolN1. Carbon atoms are shown in yellow, μN (T ,p) = 1/2[ENtot2 + ENZP2 + μN2 (T ,p)], where ENtot2
nitrogen atoms in blue. is the total electronic energy of the nitrogen molecule obtained
from a DFT calculation in a cubic cell with 20 Å side, at the 
pyriN1 (b), pyriN2 (c), pyriN3 (d), pyriN4 (e), and pyrrolN1 point. ENZP2 is the zero-point energy of the nitrogen molecule,
(f) [25–28]. In each case the simulated system is an 8 × 8 which can be obtained from the nitrogen molecule normal
supercell with lattice vectors a = b = 19.66 Å, in agreement mode [35], while μN2 encodes the temperature and pressure
dependence of the chemical potential as
with the experimental results [29], and 15 Å of vacuum  
in the c direction to ensure decoupling between periodic p
images. The DFT total energy calculations are performed with μN2 (T ,p) = H ◦ (T ) − H ◦ (0) − T S ◦ (T ) + kB T log ◦ ,
p
the Perdew-Burke-Ernzerhof (PBE) [30] exchange-correlation
(4)
functional and employing Rappe-Rabe-Kaxiras-Joannopoulos
ultrasoft pseudopotentials [31]. The wave function and charge where H ◦ and S ◦ are the nitrogen gas enthalpy and entropy
cutoffs are safely set to 50 Ry and 400 Ry, respectively, at standard pressure (p◦ = 1 bar), respectively, and can be
while the Brillouin zone sampling is performed on a 2 × 2 × 1 obtained from the NIST database [36].
Monkhorst-Pack mesh [32]. The final geometry for each defect,
reported in Fig. 1, is obtained with a full relaxation, after which B. Gibbs free energy of CO2 adsorption
the force on each atom is smaller than 0.01 eV/Å. Dispersion
interactions are taken into account by employing the DFT-D2 The Gibbs free energy of adsorption of CO2 is computed
method [33,34]. We perform spin-polarized calculations in for each defected layer introduced in the previous paragraph.
order to study the magnetic moments of the structures in exam. For each configuration, the adsorption geometry is obtained
f after fully relaxing the atoms in the supercell. In this study
The Gibbs free energy of formation GX (T ,p) at temper-
we focus on CO2 adsorption on the defect only and assume
ature T and pressure p, for each defect X can be obtained
the distance between defects is large enough to suppress any
as
adsorbate-adsorbate interaction. As a first approximation, the
f
GX = FXtot (T )−Fptot (T )−nC μC (T )−nN μN (T ,p)+pV f , Gibbs free energy of adsorption of CO2 on defect X can be
obtained as
(1)
GaX (T ,p) = EXtot∗ − EXtot + FCO
vib
∗ (T ) − μCO2 (T ,p), (5)
where FXtot and Fptot are the total Helmholtz free energies of 2

the defected and pristine layers, respectively. nC and nN are where EXtot∗ is the total electronic energy of the layer plus
the numbers of carbon and nitrogen atoms added (positive) or adsorbed CO2 , μCO2 is the chemical potential of CO2 in the gas

155428-2
NITROGEN DOPING AND CO2 ADSORPTION ON … PHYSICAL REVIEW B 97, 155428 (2018)

vib f
phase, and FCO ∗ is the vibrational free energy of the molecule TABLE I. Formation energy at 0 K, E0 , formation energy at 0 K
2 f
while adsorbed on the defected layer. This can be obtained from neglecting zero-point energies, Ê0 and nitrogen-carbon bond lengths
a normal modes calculation letting only the adsorbed molecule for each defect considered. Referring to Fig. 1, for pyriN2 the first
vibrate, while keeping fixed the atoms in the layer. We shall value of lN−C refers to N1 the second to N2 . For pyriN4, the two values
analyze the assumptions and implications of Eq. (5) in Sec. IV. of lN−C are the distances N-C1 and N-C2 , respectively.
The chemical potential of the CO2 molecule in the gas f f
phase is obtained through the same strategy employed for the Defect E0 (eV) Ê0 (eV) lN−C (Å)
nitrogen molecule. However, it has been shown that the PBE graphitic-N 0.932 0.885 1.407
functional introduces a systematic error in the calculation of pyriN1 5.692 5.444 1.321
the total energy of the gas-phase CO2 molecule, which can be pyriN2 4.892 4.555 1.319/1.377
compensated by adding 0.13 eV to the PBE result [37]. pyriN3 3.841 3.417 1.332
It is then possible to have a first insight on the process pyriN4 4.401 3.729 1.346/1.323
of adsorption on the various defects by employing a simple pyrrolN1 9.174 8.926 1.462
Langmuir adsorption model [38–40] to study the amount of
adsorbate at equilibrium as a function of the temperature and
CO2 partial pressure. As mentioned, we assume independence f
between defects and consider adsorption only on the nitrogen The values of Ê0 are in good agreement with previous stud-
defect (not on nearby graphene carbon atoms), thus neglecting ies [18,20,45,46]. On the other hand, we find large differences
coverage effects in the adsorption process [7]. We can obtain with the values reported in Ref. [17], as already noticed in
the adsorption rate per defect as [41–44] Ref. [18] (cf. note 41 therein), that can be traced back to the
different DFT treatment (LDA in Ref. [17], GGA in the present
  work) and to the different supercell size (4 × 4 in [17], 8 × 8 in
pA EXads
rX (T ,p) =
ads
 exp − , (6) this work). It is worth underlining the difference between E0
f
2π mCO2 kB T kB T
f
and Ê0 which shows the impact of the vibrational contribution
where p is the CO2 partial pressure, A the area of the active on the formation energetics at 0 K. Indeed, the zero-point
surface where adsorption on defect X takes place, and EXads energies of the carbon and nitrogen atoms amount to around
is the energy barrier the particle must overcome to adsorb on 0.15 eV.
the surface. As we shall see in the following section, dealing With Eq. (2) it is then possible to compute the Gibbs free
only with physisorption allows us to simplify Eq. (6) by taking energy of formation for each defect. In Fig. 2 we show the
EXads = 0. In our study we assume as active surface a 4 × 4 f
dependence of GX on μN ≡ 1/2μN2 , defined in Eq. (4),
subcell containing the defect, with side l = 7.4 Å. From Eq. (6) at T = 298.15 K. It can be noticed that up to μN  1 eV the
we can obtain the desorption rate as graphitic-N defect shows the lowest free energy of formation,
  while above this threshold the most stable configuration is
GaX (T ,p) given by pyriN4. This illustrates the possibility, in principle,
rX (T ,p) = rX (T ,p) exp
des ads
, (7)
kB T of modifying the nitrogen defect configuration by varying
the nitrogen chemical potential [20]. However, at standard
where GaX (T ,p) is computed with Eq. (5). By defining
temperature T = 298.15 K, μN = 1 eV is reached at p 
the coverage θ as the ratio between the number of adsorbed
1040 atm (see upper scale in Fig. 2), which makes it impossible
particles and the number of available sites, at equilibrium we
to achieve this result in practice. At lower values of μN
get the simple Langmuir isotherm relation
we can notice various crossovers between more energetic
rXads (T ,p) defects. In particular, moving around p  1 atm, the Gibbs
θ (T ,p) = . (8) free energies of pyriN3 and pyriN4 cross each other.
rX (T ,p) + rXdes (T ,p)
ads
The temperature dependence of the Gibbs free energies
This gives a rough estimation of the adsorption behavior of the of formation for each defect is shown in Fig. 3 at different
N defects in terms of the fraction of available sites occupied
by adsorbates.

III. RESULTS AND DISCUSSION

A. Defect stability
The relaxed geometries of the various defects addressed are
shown in Fig. 1. We found that all the considered structures
have zero magnetic moment except for pyriN3, with μ =
0.56 μB and pyrrolN1, with μ = 1.00 μB , where μB is the
Bohr magneton. For each defect, in Table I we report the
f
formation energy at 0 K, E0 , the formation energy at 0 K
neglecting the zero-point energies in the carbon and nitrogen FIG. 2. Variation of the Gibbs free energy of formation with
f
chemical potentials, Ê0 and the nitrogen-carbon bond lengths respect to μN (lower scale) at T = 298.15 K. In the upper scale
lN−C . we report the corresponding pressure.

155428-3
MICHELE RE FIORENTIN et al. PHYSICAL REVIEW B 97, 155428 (2018)

TABLE II. Adsorption energy of CO2 on various defects at 0 K

f
and neglecting vibrational contributions, Ê0 , distance of the CO2
carbon atom from the graphene plane d, angle of the oxygen-oxygen
segment with the vertical axis, θ, and CO2 bending angle α.

Defect Ê0a (eV) d (Å) θ (◦ ) α (◦ )

graphitic-N −0.245 3.139 83.56 179.57
pyriN1 −0.266 3.079 90.00 178.08
(a) (b) pyriN2 −0.337 3.069 86.81 178.56
pyriN3 −0.347 2.843 83.65 176.29
pyriN4 −0.376 2.816 89.64 177.09
pyrrolN1 −0.409 3.076 89.13 178.68

of the Gibbs free energy of formation, being μN (T ) the leading

term and the difference in vibrational free energy between
the pristine and defected slab still safely negligible. At higher
temperatures, the vibrational free energy difference is expected
(c) (d) to become more important, especially for the most defected
systems, though it shall not change the obtained stability
FIG. 3. (a) to (c): Gibbs free energies of formation for the various pattern.
defects at different pressures, (a) p = 10−3 atm, (b) p = 1 atm, and
(c) p = 103 atm. (d) Variation of the Gibbs free energy of formation
with respect to the temperature at p = 1 atm, showing the difference B. CO2 adsorption
between considering the temperature dependence of μC (T ) (solid In this paper we study the adsorption geometry and ener-
lines) and taking μC (T ) ≡ μC (T = 0) (dashed lines). getics of carbon dioxide on each of the considered defects.
In particular, we focus on the dominant process of physisorp-
tion mediated by dispersion interactions. In Fig. 4 we show
pressures. For pressures ranging from p = 10−3 atm Fig. 3(a),
the relaxed geometries of CO2 adsorption for each defect.
to p = 103 atm, Fig. 3(c), the free energy landscape is almost
On the left-hand side of each panel we report the top view of the
identical. The most relevant feature is the high stability of
simulated supercell, while the right-hand panel shows the side
graphitic-N, with respect to the other defects, which all lie at
view. As mentioned, in all six cases we have physisorption
higher energies. In this pressure range, at standard temperature,
of carbon dioxide over the defected graphene layer. This is
the pyriN3 defect appears to be relatively more stable than
also well displayed in Table II, where we report the adsorption
the other pyridinic and pyrrolic configurations, crossing with
energy at 0 K, Ê0a , neglecting zero-point energy, the distance d
pyriN1 at higher temperatures (T  800 K).
of the carbon atom in CO2 from the graphene plane, the angle
We can therefore conclude that in a wide temperature and −−→
pressure range, the most stable nitrogen defect in graphene is θ of the O1O2 direction (see Fig. 4) with the vertical axis and
the substitutional, graphitic one. This is energetically favored the CO2 molecule angle α ≡ O 1CO2 .
by more than 1 eV at temperatures up to 1600 K and pressures The distance d is computed with respect to the average
up to 1015 atm. This clearly reduces the possibility to easily graphene layer vertical position. It is found that the atoms
modify the nitrogen defect configurations in graphene by remain in the original plane, except for some small distortions
simply tuning pressure and temperature in an N2 gas envi- in pyriN1 and pyriN3, where displacements by more than
ronment. However, considering a less ordered material, such 0.1 Å in the vertical direction are observed (cf. Fig. 4). In
as graphene with native vacancies, it can be noticed that the the pyriN1 case, the nitrogen atoms move upwards, towards
formation energy of a graphitic nitrogen defect in the presence the adsorbed CO2 , of about 0.2 Å, while the nearby carbon
f
of a vacancy one hexagon away is E0 = 8.354 eV. This is atoms are displaced by 0.14 Å in the same direction. The
much higher than the formation energy of pyriN1, suggesting other atoms around the monovacancy remain undisplaced. For
that in disordered materials, graphitic-N and monovacancies pyriN3, the nitrogen atom in the lower hexagon lies below O1
tend to reorganize and give origin to more stable pyridinic in the CO2 molecule [see Fig. 4(d)] and is pushed downwards
configurations [18,47]. by 0.22 Å. Conversely, the other two facing nitrogen atoms are
Finally, it is interesting to observe the impact of the tem- pulled upwards by about 0.2 Å. The values of θ in Table II
f
perature dependence of the carbon chemical potential on GX differ from 90◦ by more than 1◦ in the graphitic-N and pyriN3
in the lower temperature range T  400 K, as can be seen in cases, reflecting the more marked difference between the layer
f
Fig. 3(d). In this picture, solid lines show GX (T ,p) computed regions directly below O1 and O2 . The angle α shows that
as in Eq. (2), while dashed lines show the Gibbs free energy of the carbon dioxide molecule undergoes a tiny bending due to
formation computed by approximating μC (T )  μC (T = 0) the interaction with the graphene layer. This effect is largest
in Eq. (2). In this case, the only temperature and pressure in the pyriN3 configuration, where CO2 is about 4◦ bent
dependence is given by μN . In this temperature range, this from the perfect linear geometry. This distortion entails the
is the next-to-leading contribution to temperature dependence generation of a non-null net electric dipole moment, which

155428-4
NITROGEN DOPING AND CO2 ADSORPTION ON … PHYSICAL REVIEW B 97, 155428 (2018)

(a) graphitic-N (b) pyriN1

(c) pyriN2 (d) pyriN3

(e) pyriN4 (f) pyrrolN1

FIG. 4. Relaxed geometries of CO2 adsorption on nitrogen defects in graphene. Oxygen atoms are shown in red. For each panel (a) to (f)
the left picture shows the simulated supercell from the top, while the right one from the side. The orientation is shown by the reference frame
in the bottom left of each picture.

can play a role, though small, in the interaction with the

graphene layer and with any applied external electric field [22].
Finally, the fundamental impact of the DFT-D2 correction is
checked against uncorrected, pure GGA calculations that give
positive adsorption energies, around 6.5 eV for each structure,
thus resulting in unbound states. The Gibbs free energy of
adsorption in Eq. (5) is computed for each defect and plotted
vs temperature and pressure in Fig. 5.
We can notice that at standard conditions (T = 298.15 K
and p = 1 atm) only for pyriN4 and pyrrolN1 we have GaX 
0, indicating a thermodynamically favored process. The other
defects show a moderately positive Gibbs free energy of
adsorption, 0 < GaX  0.12 eV, where we have GaX 
0.12 eV for graphitic-N. At room temperature, for lower
pressures p < 1 atm, the adsorption process is endergonic for
all defects, while for higher pressures it becomes exergonic
in all cases. From this analysis we can conclude that the
adsorption process becomes thermodynamically favored for
all defects only for high pressures or low temperatures, with
the upper limit given by ÊXa reached at T = 0 K. In particular,
for the most stable nitrogen defect, graphitic-N, adsorption
becomes favorable only for very low temperatures T  200 K
or high pressures T  102 atm. On the other hand, for the least
stable nitrogen defect here analyzed, pyrrolN1, CO2 adsorption FIG. 5. Gibbs free energies of formation vs temperature and
becomes unfavored for high temperatures T  300 K or very pressure for the various defects. Red areas have GaX > 0, while for
low pressures. In all cases, we have Gads X  0.4 eV in the blue areas GaX < 0. The GaX = 0 contour is indicated by the solid
whole temperature-pressure space considered, implying that, black line, while the standard conditions T = 298.15 K, p = 1 atm
though disfavored, adsorption cannot be considered a strongly are marked by the black cross.

155428-5
MICHELE RE FIORENTIN et al. PHYSICAL REVIEW B 97, 155428 (2018)

Our calculation of the Gibbs free energy of formation

neglected the difference in the vibrational properties between
pristine and defected graphene. A more precise calculation may
take into account the full phonon dispersion of the two systems,
though the results obtained here should not be affected by this
improvement.
In the second part of this paper we analyzed the adsorption
(a) (b) of carbon dioxide on the nitrogen defects. Our results indicate
that CO2 adsorption is not thermodynamically favored at stan-
FIG. 6. Langmuir isotherms for the various defects at dard conditions, except for the pyrrolic defect. Nevertheless,
T = 298.15 K (a) and at T = 400 K (b). θ is the coverage though disfavored, this process revealed to be moderately en-
defined as the ratio of occupied over available adsorption sites. dergonic at standard conditions, implying that at equilibrium,
the coverage is not totally negligible.
In the study of the Gibbs free energy of adsorption some
assumptions were made. Firstly, all the degrees of freedom
suppressed process in the (T ,p) range of our interest. Indeed,
of the adsorbate were treated as vibrational. A detailed study
the desorption rate in Eq. (7) is slightly enhanced by this
of the potential energy surface might point out some flat
positive values with respect to the adsorption one. This can
direction for which rotation and/or translation cannot be treated
be seen in Fig. 6 where the Langmuir isotherms are plotted for
as frustrated [48]. This, however, should occur at higher
all defects at T = 298.15 K (a) and at T = 400 K (b).
temperatures than those considered in this paper. We also
At room temperature, panel (a), the equilibrium coverage re-
assumed that the vibrational modes of the adsorbed carbon
flects the Gibbs free energy landscape, with pyrrolN1 showing
dioxide molecule can be considered decoupled from those of
the highest values and the graphitic-N the lowest. For graphitic-
the graphene layer. This can be supported by the fact that the
N, at standard conditions we have θg−N (T = 298.15 K,p =
CO2 molecule and the graphene layer are linked by dispersion
1 atm) = 0.7%, while increasing the pressure to p = 5 atm the
interactions only, which are much weaker than the covalent
coverage increases to θg−N (T = 298.15 K,p = 5 atm) = 4%.
bonds in the molecule and in the layer. Nevertheless, a complete
For pyrrolN1, already at standard conditions we have almost
calculation should consider the layer plus adsorbate as a whole
full coverage θpyrrolN1 (T = 298.15 K,p = 1 atm) = 82%. At
system and carry out a thorough vibrational study. Finally, in
T = 400 K, panel (b), at atmospheric pressure the coverage
studying the adsorption behavior we focused on the nitrogen
is below 1% for all defects, except for pyrrolN1, for which
defects only, assuming CO2 adsorption can take place only
θpyrrolN1 (T = 400 K,p = 1 atm) = 10%.
on the defected regions. A realistic study must consider CO2
adsorption also on nearby graphene sites, thus taking into
account adsorbate-adsorbate interactions and coverage effects.
IV. CONCLUSIONS In conclusion, our work extends the results on nitrogen
defect stability beyond the zero-temperature regime to a wide
In the first part of this paper we have carried out a theoretical range of temperatures and pressures and provides a detailed
study of the thermodynamics of the formation of a set of study of the geometry and thermodynamics of CO2 adsorption
nitrogen defects in free-standing graphene. The key quantity on the same defects. A combination of the results on defect sta-
analyzed is the Gibbs free energy of formation, which provides bility and CO2 adsorption can prove interesting for the design
insights on the stability of the different defects. The results of a carbon-based material with high adsorption performances,
obtained confirm the high stability of graphitic-N in a wide a crucial step towards carbon dioxide capture.
range of temperatures and pressures, as already suggested
by the formation energies alone [17,20]. The incorporation
ACKNOWLEDGMENTS
of independent vacancies in graphene alters the stability
scenario, favoring the rearrangement with graphitic-N atoms We acknowledge the CINECA award under the ISCRA
into pyridinic-like configurations [18]. This may suggest that initiative, for the availability of high performance computing
a more disordered material may accommodate more easily resources and support. M.R.F. would also like to acknowl-
N-defect configurations which in this study appear quite edge computational support provided by HPC@POLITO
disfavored. (http://www.hpc.polito.it).

[1] J. Wang, L. Huang, R. Yang, Z. Zhang, J. Wu, Y. Gao, Q. Wang, [4] W. Li, B. Herkt, M. Seredych, and T. J. Bandosz, Appl. Catal.,
D. O’Hare, and Z. Zhong, Energy Environ. Sci. 7, 3478 (2014). B 207, 195 (2017).
[2] X. Mao and T. A. Hatton, Ind. Eng. Chem. Res. 54, 4033 [5] W. Li, M. Seredych, E. Rodríguez-Castellón, and T. J. Bandosz,
(2015). ChemSusChem 9, 606 (2016).
[3] J. Wu, S. Ma, J. Sun, J. Gold, C. Tiwary, B. Kim, L. Zhu, N. [6] Y. Song, R. Peng, D. K. Hensley, P. V. Bonnesen, L. Liang, Z.
Chopra, I. Odeh, R. Vajtai, A. Yu, R. Luo, J. Lou, G. Ding, P. Wu, H. M. Meyer, M. Chi, C. Ma, B. G. Sumpter, and A. J.
Kenis, and P. Ajayan, Nat. Commun. 7, 13869 (2016). Rondinone, ChemistrySelect 1, 6055 (2016).

155428-6
NITROGEN DOPING AND CO2 ADSORPTION ON … PHYSICAL REVIEW B 97, 155428 (2018)

[7] K. Takeuchi, S. Yamamoto, Y. Hamamoto, Y. Shiozawa, K. [25] L. S. Panchakarla, K. S. Subrahmanyam, S. K. Saha, A. Govin-
Tashima, H. Fukidome, T. Koitaya, K. Mukai, S. Yoshimoto, daraj, H. R. Krishnamurthy, U. V. Waghmare, and C. N. R. Rao,
M. Suemitsu, Y. Morikawa, J. Yoshinobu, and I. Matsuda, Adv. Mater. 21, 4726 (2009).
J. Phys. Chem. C 121, 2807 (2017). [26] L.-S. Zhang, X.-Q. Liang, W.-G. Song, and Z.-Y. Wu, Phys.
[8] A. Ghosh, K. S. Subrahmanyam, K. S. Krishna, S. Datta, A. Chem. Chem. Phys. 12, 12055 (2010).
Govindaraj, S. K. Pati, and C. N. R. Rao, J. Phys. Chem. C 112, [27] L. Zhao, R. He, K. T. Rim, T. Schiros, K. S. Kim, H. Zhou,
15704 (2008). C. Gutiérrez, S. P. Chockalingam, C. J. Arguello, L. Pálová, D.
[9] J. Low, J. Yu, and W. Ho, J. Phys. Chem. Lett. 6, 4244 (2015). Nordlund, M. S. Hybertsen, D. R. Reichman, T. F. Heinz, P. Kim,
[10] D. C. B. Alves, R. Silva, D. Voiry, T. Asefa, and M. Chhowalla, A. Pinczuk, G. W. Flynn, and A. N. Pasupathy, Science 333, 999
Materials for Renewable and Sustainable Energy 4, 2 (2015). (2011).
[11] D. G. Hwang, E. Jeong, and S. G. Lee, Carbon letters 20, 81 [28] H. Gao, L. Song, W. Guo, L. Huang, D. Yang, F. Wang, Y. Zuo,
(2016). X. Fan, Z. Liu, W. Gao, R. Vajtai, K. Hackenberg, and P. M.
[12] J. Wu, M. Liu, P. P. Sharma, R. M. Yadav, L. Ma, Y. Yang, X. Ajayan, Carbon 50, 4476 (2012).
Zou, X.-D. Zhou, R. Vajtai, B. I. Yakobson, J. Lou, and P. M. [29] D. C. Elias, R. R. Nair, T. M. G. Mohiuddin, S. V. Morozov,
Ajayan, Nano Lett. 16, 466 (2016). P. Blake, M. P. Halsall, A. C. Ferrari, D. W. Boukhvalov, M. I.
[13] Y. Liu, J. Zhao, and Q. Cai, Phys. Chem. Chem. Phys. 18, 5491 Katsnelson, A. K. Geim, and K. S. Novoselov, Science 323, 610
(2016). (2009).
[14] G.-L. Chai and Z.-X. Guo, Chem. Sci. 7, 1268 (2016). [30] J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77,
[15] H. He and Y. Jagvaral, Phys. Chem. Chem. Phys. 19, 11436 3865 (1996).
(2017). [31] A. M. Rappe, K. M. Rabe, E. Kaxiras, and J. D. Joannopoulos,
[16] H. Wang, Y. Chen, X. Hou, C. Ma, and T. Tan, Green Chem. 18, Phys. Rev. B 41, 1227 (1990).
3250 (2016). [32] H. J. Monkhorst and J. D. Pack, Phys. Rev. B 13, 5188
[17] Y. Fujimoto and S. Saito, Phys. Rev. B 84, 245446 (2011). (1976).
[18] Z. Hou, X. Wang, T. Ikeda, K. Terakura, M. Oshima, M.-a. [33] S. Grimme, J. Comput. Chem. 27, 1787 (2006).
Kakimoto, and S. Miyata, Phys. Rev. B 85, 165439 (2012). [34] V. Barone, M. Casarin, D. Forrer, M. Pavone, M. Sambi, and A.
[19] Z. Hou, X. Wang, T. Ikeda, K. Terakura, M. Oshima, and M.-a. Vittadini, J. Comput. Chem. 30, 934 (2009).
Kakimoto, Phys. Rev. B 87, 165401 (2013). [35] JANAF THERMOCHEMICAL TABLES, Anal. Chem. 61,
[20] Y. Jing and Z. Zhou, ACS Catalysis 5, 4309 (2015). 1327A (1989).
[21] K.-J. Lee and S.-J. Kim, Bulletin of the Korean Chemical Society [36] D. R. Burgess, NIST Chemistry WebBook, NIST
34, 3022 (2013). Standard Reference Database 10.18434/T4D303,
[22] M. Muruganathan, J. Sun, T. Imamura, and H. Mizuta, Nano http://webbook.nist.gov/chemistry.
Lett. 15, 8176 (2015). [37] A. A. Peterson, F. Abild-Pedersen, F. Studt, J. Rossmeisl, and
[23] P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. J. K. Norskov, Energy Environ. Sci. 3, 1311 (2010).
Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococcioni, I. [38] I. Langmuir, J. Franklin Inst. 183, 102 (1917).
Dabo, A. D. Corso, S. de Gironcoli, S. Fabris, G. Fratesi, R. [39] I. Langmuir, Proc. Natl. Acad. Sci. USA 3, 251 (1917).
Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri, [40] I. Langmuir, J. Am. Chem. Soc. 40, 1361 (1918).
L. Martin-Samos, N. Marzari, F. Mauri, R. Mazzarello, S. [41] K. Reuter and M. Scheffler, Phys. Rev. B 73, 045433 (2006).
Paolini, A. Pasquarello, L. Paulatto, C. Sbraccia, S. Scandolo, [42] M. Karikorpi, S. Holloway, N. Henriksen, and J. Nørskov, Surf.
G. Sclauzero, A. P. Seitsonen, A. Smogunov, P. Umari, and R. Sci. 179, L41 (1987).
M. Wentzcovitch, J. Phys.: Condens. Matter 21, 395502 (2009). [43] G. R. Darling and S. Holloway, Rep. Prog. Phys. 58, 1595
[24] P. Giannozzi, O. Andreussi, T. Brumme, O. Bunau, M. B. (1995).
Nardelli, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, M. [44] A. Groß, Surf. Sci. Rep. 32, 291 (1998).
Cococcioni, N. Colonna, I. Carnimeo, A. D. Corso, S. de [45] Y.-X. Yu, Phys. Chem. Chem. Phys. 15, 16819 (2013).
Gironcoli, P. Delugas, R. A. D. Jr, A. Ferretti, A. Floris, G. [46] S. Kattel, P. Atanassov, and B. Kiefer, J. Phys. Chem. C 116,
Fratesi, G. Fugallo, R. Gebauer, U. Gerstmann, F. Giustino, T. 8161 (2012).
Gorni, J. Jia, M. Kawamura, H.-Y. Ko, A. Kokalj, E. Küçükbenli, [47] D. Usachov, A. Fedorov, O. Vilkov, B. Senkovskiy, V. K.
M. Lazzeri, M. Marsili, N. Marzari, F. Mauri, N. L. Nguyen, Adamchuk, L. V. Yashina, A. A. Volykhov, M. Farjam, N. I.
H.-V. Nguyen, A. O. de-la Roza, L. Paulatto, S. Poncé, D. Rocca, Verbitskiy, A. Grüneis, C. Laubschat, and D. V. Vyalikh, Nano
R. Sabatini, B. Santra, M. Schlipf, A. P. Seitsonen, A. Smogunov, Lett. 14, 4982 (2014).
I. Timrov, T. Thonhauser, P. Umari, N. Vast, X. Wu, and S. [48] D. W. Blaylock, T. Ogura, W. H. Green, and G. J. O. Beran,
Baroni, J. Phys.: Condens. Matter 29, 465901 (2017). J. Phys. Chem. C 113, 4898 (2009).

155428-7