Showcasing research from the Group of Prof.

Maodu Chen at As featured in:

Dalian University of Technology, China.
Volume 21 Number 40 28 October 2019 Pages 22163–22660

Neural network potential energy surface and dynamical isotope

effects for the N+(3P) + H2 → NH+ + H reaction PCCP
Physical Chemistry Chemical Physics
rsc.li/pccp

This computational work constructs a global NH2+ potential

energy surface by combining numerous high-level ab initio
energy points with the permutation invariant polynomial neural
network method. Utilizing this newly constructed potential
energy surface, time-dependent wave packet calculations were
performed on the N+(3P) + H2 and N+(3P) + D2 reactions to study
ISSN 1463-9076

the microscopic reaction mechanism and dynamical isotope PAPER

Gauthier Eppe et al.
Understanding chemical interaction between
phosphonate-derivative molecules and a silver
surface cluster in SERS: a combined experimental and

effects. The present work can provide useful guidelines on the computational approach

experimental study of the two reactions. See Maodu Chen et al.,

Phys. Chem. Chem. Phys.,
2019, 21, 22203.

rsc.li/pccp
Registered charity number: 207890
 PCCP
View Article Online
PAPER View Journal | View Issue

Neural network potential energy surface

Published on 07 August 2019. Downloaded by Universidad de Salamanca on 2/12/2020 6:48:29 PM.

and dynamical isotope effects for the

N+(3P) + H2 - NH+ + H reaction
Cite this: Phys. Chem. Chem. Phys.,
2019, 21, 22203

Zijiang Yang,a Shufen Wang,a Jiuchuang Yuan b

and Maodu Chen *a

g ) - NH (X P) + H( S) reaction is important for initiating the chain reaction of

The N+(3P) + H2(X1S+ + 2 2

ammonia synthesis in the universe. To study the dynamics of this reaction, a global accurate potential
energy surface (PES) of the ground state NH+
2 was constructed by combining numerous high-level ab initio

energy points with the permutation invariant polynomial neural network method. Utilizing this newly
constructed PES, time-dependent wave packet calculations were performed on the state-to-state reactions
of N+(3P0) + H2 (v = 0, j = 0) and N+(3P0) + D2 (v = 0, j = 0) in order to study the microscopic reaction
Received 17th May 2019, mechanisms and dynamical isotope effects. Isotope effects have a significant influence on the rovibrational
Accepted 6th August 2019 state distributions and state resolved angle distributions of the product. The total differential cross sections
DOI: 10.1039/c9cp02798j (DCSs) present in the aforementioned reactions are dominated by both forward and backward scattering, as
expected from the observable deep well along the reaction path. Meanwhile, the rovibrational state-resolved
rsc.li/pccp DCSs show that both reactions are not entirely statistical at the state-to-state level.

1 Introduction energies, and only a limited number of states are included due to
its slight endothermicity. Thus, various statistical methods11,12,16,17
The nitrogen atom and nitrogenous molecules are important have been employed to treat the reactions of N+(3P) with H2 and
tracers when studying interstellar objects; nitrogen hydrides or its isotopic variants. Gerlich16 applied dynamically biased
their ions are also of significance in astrophysics due to their statistical theory to obtain the state-specific ICSs for the title
abundance in the universe. One of the simplest nitrogen reaction. The calculated results were very comparable with
chemical reactions of N+(3P) + H2(X1S+g) - NH+(X2P) + H(2S) known experiments; however, large discrepancies occurred
has been suggested as the first step in the synthesis of ammonia upon replacing H2 molecule with D2 molecule. Recently,
within dense interstellar clouds.1,2 In addition, as a prototype for Grozdanov et al.11 formulated a new statistical model to calculate
understanding ion–molecule reaction dynamics, it has highly the ICSs of reactions involving N+ with H2, D2 and HD molecules.
interesting characteristics for study because of its near thermo- The fine structure states of N+ ion were treated equally to other
neutrality. Especially, the dynamical behaviors are substantially internal motions, and the calculated results for the three reac-
different when the hydrogen undergoes isotopic substitution; tions could reproduce the experimental ICSs. In later work,12 the
numerous studies3–12 have focused on these isotope effects, authors concluded that the dynamics results were strongly
which have been consecutively explored over the past 30 years. dependent on the fine structure state of N+ ion and the initial
A variety of experimental results for the title reaction and its rotation state of the reactant H2 or HD.
isotopic counterparts, such as velocity-angle distributions13,14 However, theoretical treatment of statistical models is
and integral cross sections (ICSs),4,9 have been reported limited due to the addition of certain assumptions. A better
and provide abundant reference data for theoretical studies. method for accurately describing the characteristics of the title
Additionally, according to reported ab initio calculations,15 the reaction is based on the precise potential energy surface (PES)
title reaction processes a deep potential well (B6.5 eV) path, that covers the entire reactant and product regions. The first
resulting in the formation of long-lived complex NH+2 at low ground state PES of NH+2 was constructed by Gittins and Hirst.18
The energy points were calculated via the configuration inter-
action (CI) method, and a deep 3B1 well at C2v symmetry was
a
Key Laboratory of Materials Modification by Laser, Electron, and Ion Beams
represented in the PES. However, a crucial issue was that their
(Ministry of Education), School of Physics, Dalian University of Technology,
Dalian 116024, P. R. China. E-mail: mdchen@dlut.edu.cn
calculations of the product generated the wrong electronic state
b
State Key Laboratory of Molecular Reaction Dynamics, Dalian Institute of of NH+(a4S), which is actually the first excited state of NH+
Chemical Physics, Chinese Academy of Science, Dalian 116023, P. R. China molecule.19 Gonzalez et al.20 presented an ion-induced single

This journal is © the Owner Societies 2019 Phys. Chem. Chem. Phys., 2019, 21, 22203--22214 | 22203
 View Article Online

Paper PCCP

surface model for the title reaction and discovered that the 3A2 results. Furthermore, quantum dynamics simulations of the
well plays a predominant role in the energy range of 1 to 4 eV. The N+(3P) + D2 reaction have not been presented to date, and it is
most recent PES for the NH+2 system was reported by Wilhelmsson desirable to perform these simulations to study the influence of
et al.21 in 1992. In their work, a total of 752 configurations isotope effects on the title reaction. Herein, a new accurate
calculated by the complete active space self-consistent (CASSCF) ground state PES of the NH+2 system is constructed using a large
and multi-reference configuration interaction (MRCI) methods number of high level ab initio energy points and a neural
were used to fit the analytical form, and the average least-squares network (NN) model. Based on our new PES, rigorous quantum
deviation was 0.026 eV. Based on these PESs, dynamics simula- TDWP calculations for the N+(3P0) + H2 (v = 0, j = 0) - NH+ + H
tions of the N+ and H2 reaction and its isotopic variants using and N+(3P0) + D2 (v = 0, j = 0) - ND+ + D reactions are
Published on 07 August 2019. Downloaded by Universidad de Salamanca on 2/12/2020 6:48:29 PM.

classical trajectory calculations22 or the quasiclassical trajectory implemented to study the reaction mechanism and the influ-
(QCT) method6,7,20,23 have been extensively implemented. ence of isotope effects on the dynamics results.
Although the QCT method or statistical models can describe
the mechanism of the reactions of N+(3P) with H2 and its
isotopic variants to some extent, the significant quantum 2 Potential energy surface
effects are not included. The structure of the deep well and 2.1 Ab initio calculations
its nearly thermoneutral character imply that the quantum The electronic energies of the ground state (13A00 ) NH+2 within Cs
scattering resonances and the zero point energy (ZPE) effect symmetry were calculated by the internally contracted MRCI
are highly crucial to the reaction dynamics. Performing quantum method31,32 with the Davidson correction (icMRCI + Q), as
wave packet calculations on the title reaction is extremely implemented in the MOLPRO package.33 First, the Hartree–
expensive due to the numerous quasi-bound states supposed Fock wave functions were calculated as an initial guess,
by the deep well, which require a large number of grids and a followed by CASSCF34,35 calculations for the seven equally
long propagation time in order to achieve evolution of the wave weighted states (13A00 , 23A00 , 33A00 , 13A 0 , 11A 0 , 21A 0 , 11A00 ). The
packet. It is well known that the only quantum dynamics results icMRCI + Q calculations were carried out based on the opti-
for the N+(3P) + H2 reaction were calculated using the time- mized wave functions of the CASSCF process. In the CASSCF
dependent wave packet (TDWP) method proposed by Russell and icMRCI + Q calculations, six valence electrons are included
and Manolopoulos.24 They performed TDWP calculations for in seven active orbitals (6a 0 + 1a00 ). The augmented correlation-
the reactions of N+(3P) + H2 (v = 0 and 1, j = 0 and 1) on the consistent polarization valence quadruple-zeta (AVQZ) basis
Wilhelmsson PES. The ICS results of the ground vibrational set36 and the Dunning-weighted correlation consistent polarized
reactions showed different behaviors compared with QCT calcu- core-valence triple-zeta (WCVTZ)37 basis set were used for H and
lations due to the inability of the traditional QCT method to treat N atoms, respectively, which ensured accuracy of the calcula-
the vibrational ZPE effect of the product molecule. In the QCT tions and convergence of key energy points by testing numerous
calculations, only the ZPE of the reactant H2 molecule was geometries. A large number of ab initio energies over a great
corrected; thus, the title reaction became exothermic. range of configuration spaces were chosen in the Jacobi coordi-
According to previous studies, traditional QCT calculations nates, where the N+–H2 region was defined by 0.9 r rHH/a0 r 25,
are unable to correctly describe the dynamical behaviors of the 0.8 r RN+–HH/a0 r 25, and 0 r y/degree r 90 and the H–NH+
title reaction. The ZPE problem can be solved to some extent region was defined by 0.8 r r+NH/a0 r 28, 0 r RH–NH+/a0 r 25,
using various corrective methods; however, the Gaussian– and 0 r y/degree r 180.
Binning method displays excellent corrective abilities toward
this shortcoming of the QCT method. The Gaussian–Binning 2.2 Fitting the potential energy surface
procedure25,26 has been successfully incorporated into several
Representing global PESs from ab initio energies is challenging
barrierless reactions, such as H+ + H227,28 and C(1D) + H229
but can be achieved by combining electronic structure calcula-
reaction systems, and the calculated results were in good agree-
tions with advanced machine learning technologies. Artificial
ment with the quantum dynamics calculations. Another basic
NNs and Gaussian process regression are the most popular
problem of the QCT method is its poor capability to describe
machine learning methods to fit PESs by a large number of
quantum resonances, which are vital for deep well reactions.
ab initio data; these methods have been widely applied to
Therefore, using only quantum methods can give perfect
numerous molecular systems.38–46 In this work, the NN method
dynamics information for the title reaction. Although quantum
was used to obtain the analytical PES of the NH+2 system. The
dynamics calculations for the title reaction have been reported,
expression of the global surface consists of two-body and three-
the inaccuracy of the results is attributed to the lack of reliable
body terms and is written as follows:
PES. The error of the Wilhelmsson PES is relatively large due to
the limitations of early computational conditions. In addition, Vtotal ðR1 ; R2 ; R3 Þ ¼ VNHþ ðR1 Þ þ VH2 ðR2 Þ þ VNHþ ðR3 Þ
ab initio calculations30 and a recent experimental study10 have (1)
ð3Þ
proved that the endothermicity of the reaction (33 meV) on the þ VNHþ  f ðR1 ; R2 ; R3 Þ
2
Wilhelmsson PES is too large. Hence, performing quantum
dynamics calculations for the title reaction on a high-precision where R1, R2 and R3 are the internuclear distances of NH+a, HH
global PES is necessary to improve the accuracy of the dynamics and NH+b, respectively. For the ion–molecule reaction, the

22204 | Phys. Chem. Chem. Phys., 2019, 21, 22203--22214 This journal is © the Owner Societies 2019
 View Article Online

PCCP Paper
 
PNHþ ¼ exp 0:2RNHþ (4)
b b

PHH = exp(0.2RHH) (5)

Then, the following PIPs were adopted:

 .
G1 ¼ PNHþa þ PNHþ 2 (6)
b

G2 ¼ PNHþa PNHþ (7)

Published on 07 August 2019. Downloaded by Universidad de Salamanca on 2/12/2020 6:48:29 PM.

G3 = PHH (8)

Finally, G1, G2, and G3 were given normalized treatment as

 
2 Gi  Gi;min
Ii ¼    1; ði ¼ 1; 2; 3Þ (9)
Gi;max  Gi
Fig. 1 Comparison of ab initio data and fitting results for the potential
energy curves of H2(X1S+ + 2
g ) and NH (X P).
where Gi,min and Gi,max are the minimum and maximum values of
Gi, respectively. The NN included two hidden layers with 12
neurons, and the Levenberg–Marquardt algorithm51 was applied
accuracy of the long-range potential has substantial influence to train the NN. The final three-body NN expansion is presented as
on the dynamics results; thus, a switch function f (R1, R2, R3)
ð3Þ
was used to better describe the PES in the long-range region, VNHþ
2
and it is written as follows: ! ! !
X
12
ð3Þ
X
12
ð2Þ
X
3
ð1Þ ð1Þ ð2Þ
ð3Þ
Y    ¼j wi jð2Þ wj jð1Þ wk Ik þbj þbi þbð3Þ
;
1 Rn  Rd
f ðRÞ¼ 1  1 þ tanh ; (2) i¼1 j¼1 k¼1
n
2 Rw
(10)

where Rn represents the bond length of two atoms and Rd and where o and b represent the corresponding weights and bias,
Rw are the central position of f (R1, R2, R3) and the switch which were optimized by training the NN. The linear and
strength constant, respectively. In the process of two-body hyperbolic tangent functions were used as the transfer func-
term fitting, the NN structure included two hidden layers, tions f in the output layer and hidden layers, respectively. To
and 6 neurons were included in each hidden layer. 97 and 92 avoid overfitting, 30 519 energy points were divided randomly
ab initio points were generated to obtain the potential energy into three sets, namely the training set (90 percent), testing set
curves of HH and NH+, and the root mean square errors (5 percent), and validation set (5 percent). The overall RMSE of
(RMSE) are 0.10 and 0.12 meV, respectively. Fig. 1 shows the the PIP-NN PES is only 2.7 meV.
potential energy curves (PECs) of the H2(X1S+g) and NH+(X2P)
2.3 Features of the potential energy surface
molecules, where the fitting of the PECs is in good agreement
with the original ab initio data. The equilibrium bond length Fig. 2 shows the potential energies at four different insertion
and dissociation energy of H2(X1S+g) on the fitted PECs are angles in the reactant Jacobi coordinates, where rHH and RN+–HH
1.402 Bohr and 4.71 eV, respectively, which compare well with are the HH bond length and the distance between N+ and the
the corresponding experimental data47 of Re = 1.401 Bohr center of mass of the H2 molecule, respectively, and y is the
and De = 4.74 eV. In the case of NH+(X2P) molecule, the angle between rHH and RN+–HH. The energies are relative to
equilibrium bond length and dissociation energy are 2.047 Bohr the N+ + H2 dissociation limit, and the maximum energy was set
and 4.64 eV, respectively, showing good agreement with the as 5 eV in order to clearly display the structures of the wells.
experimental results48 of Re = 2.043 Bohr and De = 4.65  0.5 eV. There are two wells at every angle, which correspond to the 3B1
Thus, the fitting results for the two-body potential are sufficiently and 3A2 wells at the C2v structure, respectively. As the insertion
accurate. angle decreases, the deeper well corresponding to the 3B1 state
For the three-body term, permutation invariant polynomials at the C2v structure becomes less accessible and the depth of
(PIP)49,50 were employed to solve the problem of the adaptation the outer 3A2 well slowly decreases. The 3B1 well is located
of permutation symmetry caused by the two identical H atoms. at rHH = 3.76 Bohr at the T-shaped symmetry, indicating the
The geometries after PIP preprocessing were used in the input involvement of a bond stretching process prior to complex NH+2
layer of the NN. First, the bond length between two atoms was formation. Moreover, there is a saddle point between the two
transformed as wells, which can be attributed to prevention of the crossing of
the two lowest electronic states.20 In the case of C2v symmetry,
  the complex NH+2 processes form at the deep 3B1 state well
PNHþa ¼ exp 0:2RNHþa (3)
via the 3A2 - 3B1 intersection, and the two lowest states 3A2 and

This journal is © the Owner Societies 2019 Phys. Chem. Chem. Phys., 2019, 21, 22203--22214 | 22205
 View Article Online

Paper PCCP
Published on 07 August 2019. Downloaded by Universidad de Salamanca on 2/12/2020 6:48:29 PM.

Fig. 2 Potential energy surfaces of the ground state NH+

2 for insertion approaches at four y angles (01, 301, 601 and 901).

3
B1 return to the adiabatic ground state 3A00 in the Cs geometry. and previous ab initio calculations,15 and the Wilhelmsson PES
In the present work, the ground state PES is constructed in underestimates the depth of the well. However, there is a large
the adiabatic representation, and the conical intersection is deviation between our data and previous ab initio calculations52 for
hampered by the limitation of the Born–Oppenheimer approxi- the minimal energy of the collinear structure. This may be because
mation. Thus, the origin of the repulsive regions is the inter- the electron correlation energy cannot be treated well using the self-
action of the first excited state. consistent field method with rough CI corrections. The new PES can
The geometries and energies of the stationary points obtained on better reproduce the high-level ab initio calculations at the stationary
the new PES and original ab initio calculations are listed in Table 1 points, suggesting enhanced reliability in reaction dynamics studies.
and compared with reported results. The energy values are relative to The contour maps of the new PES at four different N+–H–H
the N+(3P) + H2 dissociation limit. For the C2v geometry, the angles (451, 901, 1351 and 1801) are presented in Fig. 3. RHH
equilibrium structure corresponds with the 3B1 well, which is a represents the bond length of the two H atoms, R+NH is the
highly crucial structure to the title reaction. As shown, the results distance between the N+ and one of the H atoms, and the
calculated on the new PES are in good agreement with the original N+–H–H angle is the included angle between RHH and R+NH.
The energy is set as zero at the N+ + H2 dissociation limit.
Two valleys exist at the bottom and left, corresponding to the
Table 1 Stationary points of the ground state NH+
2 N+(3P) + H2(X1S+g) channel and NH+(X2P) + H(2S) channel,
C2v rHH (Bohr) RN+–HH (Bohr) Energy (eV) respectively. The depth of the two valleys is comparable,
indicating that the title reaction is approximately thermoneutral.
NN PES 3.76 0.48 6.49
Wilhelmsson PES21 3.66 0.64 6.15 The deep well structure connects the two valleys at all four
icMRCI + Q 3.76 0.48 6.50 approaching angles, highlighting the important role of quantum
MRD-CI15 3.67 0.49 6.45 resonances in the reaction dynamics. At the beginning of the
title reaction, the N+ ion and H2 molecule convene in the bottom
CNv RHH (Bohr) R+NH (Bohr) Energy (eV)
valley, generating the long-lived NH+2 complex in the deep well
NN PES 1.68 2.24 2.20 and eventually dissociating to the NH+ molecule and H atom in
Wilhelmsson PES21 1.70 2.27 2.05
icMRCI + Q 1.67 2.24 2.22 the left valley. To more clearly describe the characteristics of
SCF-CI52 1.60 2.50 1.76 the title reaction on the new PES, the minimum energy paths

22206 | Phys. Chem. Chem. Phys., 2019, 21, 22203--22214 This journal is © the Owner Societies 2019
 View Article Online

PCCP Paper
Published on 07 August 2019. Downloaded by Universidad de Salamanca on 2/12/2020 6:48:29 PM.

Fig. 3 Potential energy surfaces of the ground state NH+ +

2 at four N –H–H angles (451, 901, 1351 and 1801).

(MEPs) of the N+(3P) + H2(X1S+g) - NH+(X2P) + H(2S) reaction at has a shorter lifetime when the reaction proceeds at a large
four N+–H–H angles (451, 901, 1351, 1801) are shown in Fig. 4. approaching angle. It can be seen from the minimum energy
The MEPs were determined by scanning the PES, as shown in paths that the title reaction is weekly endothermic, with an
Fig. 3, at different coordinates (RHH–R+NH) to obtain the minimum endothermicity of approximately 83 meV; this is consistent with
values of the potential energy. The depth of the well decreases the ab initio calculations. Table 2 lists the molecular constants
with increasing S–H–H angle, suggesting that the NH+2 complex of the reactants and products, which were obtained from the
analytical PES. When the ZPEs of the reactants and products
are considered, the endothermicities of the N+ + H2 and N+ + D2
reactions are approximately 8 and 23 meV, respectively, which
is in good agreement with the experimental measurements8
of 11  3 and 29  3 meV.
Fig. 5(a) displays the potential energy of N+ ion moving
around the H2 molecule with a fixed bond length of H2
(1.401 Bohr); the energy is set as zero when the N+ ion is far
from the H2 molecule. A 2.67 eV deep well is present at x = 0.0
Bohr and y = 2.3 Bohr. The depth of the well increases as the
HH bond stretches, which can promote insertion collisions at

Table 2 Molecular constants of the reactants and products

oe (cm1) oexe (cm1) D0 (eV) ZPEa (eV)

H2(X1S+g) 4398.2 121.3 4.445 0.269
D2(X1S+g) 3109.1 60.2 4.523 0.190
NH+(X2P) 2870.3 66.5 4.463 0.176
ND+(X2P) 2062.1 38.8 4.512 0.127
Fig. 4 Minimum energy paths for the N+(3P) + H2(X1S+ g ) - NH (X P) +
+ 2

a
H(2S) reaction at four N+–H–H angles (451, 901, 1351 and 1801). ZPE determined by oe/2  oexe/4.

This journal is © the Owner Societies 2019 Phys. Chem. Chem. Phys., 2019, 21, 22203--22214 | 22207
 View Article Online

Paper PCCP

where R is the distance from the N+ ion to the center of mass of

the H2 molecule and r is the bond length of the H2 molecule.
mR and mr are the corresponding reduced masses associated
with the R and r coordinates. Ĵ and ĵ are the total angular
momentum operator of the NH+2 system and the rotational
angular momentum operator of the H2 molecule, respectively.
V̂ is the potential energy of the NH+2 system which is obtained
on the NN PES. The state-to-state S-matrix was extracted by the
reactant coordinate-based method,55,56 and the second-order
Published on 07 August 2019. Downloaded by Universidad de Salamanca on 2/12/2020 6:48:29 PM.

split operator was adopted to evaluate the wave function. The

state-to-state reaction probability was calculated by
1 X  Je 

PJvj v0 j0 ¼ SvjK v0 j0 K0 2 (12)
2j0 þ 1 K;K
0

The state-resolved ICSs were obtained by

p XXX  2
 Je 
svj v0 j0 ¼ ð2J þ 1Þ SvjK v0 j0 K0  ;
ð2j0 þ 1Þkv0 j0 2 K K J
0

Fig. 5 (a) Contour plot of the PES for the N+ ion moving around the H2 (13)
molecule fixed at the equilibrium distance (1.401 Bohr). (b) Contour plot of
the PES for a H atom moving around the NH+ molecule fixed at the where kv0 j0 is the momenta in the entrance channel. The state-
equilibrium distance (2.043 Bohr). resolved differential cross sections (DCSs) can be calculated by
dsvj v0 j0 ðy; EÞ

low energies as the N+ ion slowly approaches the H2 molecule. dO

Similar contours to Fig. 5(a) for a H atom moving around the  2
1 X X  1 X 

J Je
NH+ molecule are presented in Fig. 5(b). The NH+ bond length ¼  ð2J þ 1ÞdKK ðyÞSvjK v0 j0 K0  ;
ð2j0 þ 1Þ K K 2ikv0 j0 J 0
is set to the equilibrium distance (2.043 Bohr), and the energy is 0

set as zero when the H atom is far from the NH+ molecule. (14)
It can be seen that a deeper well exists around the N+ ion, J
in which y is the scattering angle and dKK ðyÞ represents the
implying that the H atom can more readily escape from the side 0

of the H in the NH+ molecule in the title reaction. element of the reduced Wigner rotation matrix.
The initial rovibrational states of the reactant H2 molecule
were set as v0 = 0 and j0 = 0, and the TDWP calculations
included all Coriolis couplings. All partial waves were calcu-
3 Dynamics calculations
lated up to J = 75 for the N + H2 reaction and J = 81 for the N + D2
The TDWP method is a powerful tool for calculating atom– reaction, yielding converged ICSs and DCSs for collision
diatom reactive scattering and can give accurate information energies beyond 0.7 eV, which was set as the upper limit of
about the state-to-state reaction dynamics. Only an abbreviated the collision energy in the reaction dynamics. Large grids and
outline is introduced here; more detailed descriptions of the sufficient propagation times were used to ensure that the
method can be found in previous literature reports.53,54 In wave packet could reach the low energy region. The optimal
the body fixed representation, the Hamiltonian is expressed numerical parameters used in the calculations are listed in
as follows: Table 3 by numerous convergence tests.
The collision energy dependence of the total reaction prob-
2 2 2 2 ^ ^2 ^2 abilities for the N+ + H2 and N+ + D2 reactions with four
^ ¼  h @  h @ þ ðJ  j Þ þ j þ V^
H (11)
2mR @R 2 2mr @r 2 2mR R2 2mr r2 different partial waves are presented in Fig. 6. An extremely

Table 3 Numerical parameters used in the TDWP calculations

N+(3P) + H2 N+(3P) + D2
Grid ranges and sizes R (Bohr) A [0.1, 22.0], Ntot
R = 389 R (Bohr) A [0.1, 22.0], Ntot
R = 389
r (Bohr) A [0.4, 16.0], Ntot
r = 299 r (Bohr) A [0.4, 16.0], Ntot
r = 299
Nj = 100 Nj = 100
Initial wave packet Rc = 16.0 Bohr Rc = 16.0 Bohr
DR = 0.20 Bohr DR = 0.18 Bohr
k0 = (2E0mR)1/2 with E0 = 0.50 eV E0 = 0.50 eV
Propagation time 100 000 a.u. 100 000 a.u.
Highest J value 74 81

22208 | Phys. Chem. Chem. Phys., 2019, 21, 22203--22214 This journal is © the Owner Societies 2019
 View Article Online

PCCP Paper
Published on 07 August 2019. Downloaded by Universidad de Salamanca on 2/12/2020 6:48:29 PM.

Fig. 6 Total reaction probabilities for the N+(3P) + H2 and N+(3P) + D2 reactions as a function of collision energy with four different total angular
momentums (J = 0, 20, 40, 60).

dense oscillation feature was found in the reaction probability complexes in the deep well. The domination of long-lived
curves, which can be attributed to the formation of long-lived resonance is similar to the typical insertion reactions of
S(1D) + H257 and C(1D) + H2,58 and the title reaction has a
sharper oscillation peak because of the deeper well. For J = 0,
the threshold of the two reactions is consistent with the
corresponding endothermicity, implying a barrierless path at
low collision energies. With increasing J, the oscillation amplitude
becomes less pronounced due to the increased centrifugal barrier.
In the case of high J values, the threshold of the N+ + H2 reaction is
larger than that of the N+ + D2 reaction due to the heavier reduced
mass of ND+2. Because the increased collision energy diminishes
the constraint of the deep well, the resonance peak becomes
broader at high energy.
The total ICSs for the N+(3P) + H2 and N+(3P) + D2 reactions
obtained from the TDWP calculations on the NN PES together
with the experimental measurements9 using a guided ion bean
mass spectrometer and calculated by statistical theory11 are shown
in Fig. 7. Previous TDWP calculations24 for the N+(3P) + H2 reaction
on the Wilhelmsson PES are also presented. The ICS values for the
N+(3P) + H2 reaction calculated by the TDWP method on the NN
Fig. 7 Total ICSs for the N+(3P) + H2 and N+(3P) + D2 reactions as a
function of collision energy compared with previous TDWP calculations on
PES are larger than those on the Wilhelmsson PES due to the
the Wilhelmsson PES, the results calculated by statistical theory and smaller reaction threshold of the NN PES. As shown, the quantum
available experimental values. resonances remain strong on the TDWP ICS curves even though

This journal is © the Owner Societies 2019 Phys. Chem. Chem. Phys., 2019, 21, 22203--22214 | 22209
 View Article Online

Paper PCCP

various partial waves have been summed, and the oscillatory In general, the ICS values generated by the TDWP calcula-
structures calculated on the NN PES are narrower than the tions are close to the results calculated by the statistical
previous results due to the deeper well of the NN PES. The TDWP method and the experimental data. However, the collision
results show that the total ICS increases slightly as a function of energy dependence of the ICSs obtained by the TDWP calcula-
collision energy and then nearly reaches a plateau, which is tions is in contrast with the statistical and experimental results,
characteristic of an endothermic reaction. which present a monotonic decreasing dependence of collision
Published on 07 August 2019. Downloaded by Universidad de Salamanca on 2/12/2020 6:48:29 PM.

Fig. 8 Rovibrational state-resolved ICSs for the (a–d) N+(3P) + H2 and (e–h) N+(3P) + D2 reactions at four collision energies (0.1, 0.3, 0.5 and 0.7 eV).

22210 | Phys. Chem. Chem. Phys., 2019, 21, 22203--22214 This journal is © the Owner Societies 2019
 View Article Online

PCCP Paper

energy. This arises because the statistical and experimental

data include the excited spin–orbit states of N+ ion and the
higher rotational states of the H2 or D2 molecule. In the
previous experiment, the rotational temperature of H2 or D2
was 105 K, and the N+ ion was produced in a thermal distribu-
tion at 300 K. In this case, the rotational state populations of H2
at j = 0, 1, 2, and 3 were 0.243, 0.749, 0.008 and 0.0004,
respectively, and the rotational state populations of D2 at
j = 0, 1, 2, 3, and 4 were 0.472, 0.322, 0.193, 0.012 and 0.001,
Published on 07 August 2019. Downloaded by Universidad de Salamanca on 2/12/2020 6:48:29 PM.

respectively. The spin–orbit state distributions for N+ ion at 3P0,

3
P1 and 3P2 were 0.17, 0.39 and 0.44, respectively. In the
statistical theory results, the spin–orbit states of N+ ion were
treated equally to the vibrational and rotational states of H2 and
D2, and the endoergicities of the N+(3P) + H2 and N+(3P) + D2
reactions were set as 18.45 and 46.32 meV, respectively. The
electronic energy of N+ ion is 16.2 meV for the fine structure
state 3P2, and the rotational energy of H2 molecule at j = 1 is
14.4 meV, which are larger than the endoergicity of the title
reaction. The averaging effects of the excited spin–orbit states
and higher rotational states allow the two reactions to become
exothermic. It is clear that the statistical theory results success-
fully reproduce the experimental data of the N+(3P) + H2
reaction even when some assumptions are included at the
energy range. However, in the case of the N+(3P) + D2 reaction,
the statistical theory results and experimental data have relatively
large deviations. In addition, the statistical method cannot char-
acterize the quantum resonances, which is crucial for a deep well
reaction. Herein, only rigorous quantum dynamics calculations
on the ground adiabatic PES for the state-resolved reactions of
N+(3P0) + H2 and N+(3P0) + D2 were performed; thus, they displayed
different dynamics behavior compared with the statistical and
experimental results.
The rovibrational state-resolved ICSs of the product NH+ and
ND+ molecules at four collision energies (0.1 eV, 0.3 eV, 0.5 eV
and 0.7 eV) calculated by the TDWP method are shown in Fig. 8.
At lower collision energies, only the ground vibrational state is
populated, and the v 0 = 1 and v 0 = 2 channels are opened with
increasing collision energy. As shown, the products are difficult
to excite to higher vibrational states, implying that the energy
conversion efficiency is low. The main reason for this is the
Fig. 9 Three-dimensional plots of the total DCSs for the N+(3P) + H2 and
presence of a deep well on the reaction path: the N+ ion collides
N+(3P) + D2 reactions as a function of collision energy.
with the H2 or D2 molecule, generating the long-lived complex.
One H or D atom is ejected only when the rotational quantum
number of the continuously rotating complex is large enough to is because the vibrational frequency and rotational constant of
cleave the HH or DD bond. Thus, the majority of the collision ND+ molecule are smaller than those of NH+ molecule, and the
energies are transformed into translational energy of the pro- title reaction is very sensitive toward this subtle difference.
duct molecules, and the transformed internal energy of the Moreover, the N+(3P) + D2 reaction has smaller centrifugal
products is relatively low. In addition, a part of the collision barriers than the N+(3P) + H2 reaction at the special partial
energy is used to overcome the centrifugal barrier at the exit wave, which leads to the transformation of more collision
channel. As the collision energy increases, more rotational energy into internal energy. Therefore, the isotope effects have
states are populated, and the peak of the rotational state considerable influence on the product rovibrational state
distribution shifts to a higher j 0 value; this indicates that more distributions.
energy is transformed into the rotational energy of the NH+ or The DCSs can describe scattering features in greater detail,
ND+ molecule. Compared with the N+(3P) + H2 reaction, the whereas the angular distributions of the NH+ and ND+ mole-
N+(3P) + D2 reaction more easily excites the product to high cules formed by the two reactions have not been presented to
rovibrational states even though it has a larger threshold. This date. The three-dimensional plots of the total DCSs for the

This journal is © the Owner Societies 2019 Phys. Chem. Chem. Phys., 2019, 21, 22203--22214 | 22211
 View Article Online

Paper PCCP
Published on 07 August 2019. Downloaded by Universidad de Salamanca on 2/12/2020 6:48:29 PM.

Fig. 10 Total and vibrational state-resolved DCSs for the N+(3P) + H2 and N+(3P) + D2 reactions at 0.7 eV collision energy.

N+(3P) + H2 and N+(3P) + D2 reactions at collision energies below energy, and the corresponding vibrational state is v 0 = 0. It is
0.7 eV are given in Fig. 9. The angular distributions of NH+ and clear that the asymmetry of the forward–backward scattering is
ND+ molecules are nearly forward-backward symmetric in the more apparent. For example, at the rotational state j 0 = 10, the
whole selected collision energy range. The peaks appear at forward peak of the N+(3P) + H2 reaction is three times higher
two extreme angles, which is attributed to the long-lived than the backward peak. In the case of the N+(3P) + D2 reaction,
complex-forming mechanism. The collision energy dependence the peak at 01 disappears in the j 0 = 0 state, and the forward bias
of the DCSs shows sharp, dense oscillatory structures at each becomes dominant when the product is excited to high rota-
scattering angle; this stems from the long-lived quasi-bound tional states. This asymmetric behavior is also found in some
states supported by the deep well, which lead to densely long-lived complex-forming reactions, such as the H+ + D2,27
distributed quantum resonances. S(1D) + H259 and H + O260 reactions. The basic reason for the
Fig. 10 displays the total and vibrational state-resolved DCSs asymmetric angle distributions may be the strong scattering
of the two reactions at 0.7 eV collision energy. The total DCSs resonances even after summation over all the partial waves.
show that the forward and backward peaks are almost equal, Therefore, the quenching effect of the quantum interferences
and the angle distributions are symmetric with respect to 901. between various partial waves is less efficient than intuitively
However, the symmetries of the DCSs are broken at the vibra- expected at the state-resolved level. A comparison study59 of
tional state-resolved levels of the product molecules for the quantum and statistical methods for several typical complex-
two reactions. The forward scattering becomes more obvious forming reactions has proved that statistical theories are unable
when the product is excited to higher vibrational states, to accurately study product state-resolved reaction dynamics,
which suggests that both reactions demonstrate nonstatistical and the N+(3P) + H2 and N+(3P) + D2 reactions have similar
behavior at the vibrational state-resolved levels of the product nonstatistical behaviors. In addition, the DCSs of the two reac-
molecules despite their complex-forming natures. Fig. 11 shows tions have dramatic differences at specific rotation states; thus,
several rational states ( j 0 = 0, 5, 10, 15) of the resolved DCSs for the isotope effects have a significant impact on the state-resolved
the N+(3P) + H2 and N+(3P) + D2 reactions at 0.7 eV collision angle distributions of the product molecules.

22212 | Phys. Chem. Chem. Phys., 2019, 21, 22203--22214 This journal is © the Owner Societies 2019
 View Article Online

PCCP Paper
Published on 07 August 2019. Downloaded by Universidad de Salamanca on 2/12/2020 6:48:29 PM.

Fig. 11 Several rotational state-resolved ( j 0 = 0, 5, 10 and 15) DCSs for the N+(3P) + H2 and N+(3P) + D2 reactions at v 0 = 0 and a collision energy of 0.7 eV.

4 Conclusion comparison of the theoretical results with experimental data is

very difficult because it requires the inclusion of all accessible
In the present work, a new PES for the ground state NH+2 system spin–orbit states of N+ ion and rotational states of H2 or D2
was constructed over a large coordinate space; it can be used to molecule. In addition, a set of accurate diabatic PESs correlated
obtain accurate dynamics information for the significant inter- with the spin–orbit interaction should be used to calculate
stellar reaction of N+(3P) + H2(X1S+g) - NH+(X2P) + H(2S). The higher fine structure state reactions, which would be a very
ab initio energy points are calculated using the icMRCI + Q interesting but challenging task in future work.
method with high level basis sets (AVQZ for H atom and WCVQZ
for N atom). A total of 30 519 geometries were selected to fit the
PES using the PIP-NN model, and the fitting RMSE was only Conflicts of interest
2.7 meV. The structures and characteristics of the new PES are
There are no conflicts to declare.
described in detail and can give more accurate information than
previous PESs. TDWP calculations for the N+(3P0) + H2 (v = 0,
j = 0) and N+(3P0) + D2 (v = 0, j = 0) reactions were carried out on Acknowledgements
the NN PES to study the reaction mechanism and dynamical
isotope effects. Obvious quantum resonances are found in the This work was supported by the National Natural Science
reaction probabilities, ICSs, and DCS curves due to long-lived Foundation of China (No. 11774043) and China Postdoctoral
complex formation in the deep well (6.49 eV). The rovibrational Science Foundation (No. 2018M631828).
state-resolved ICSs reflect that the isotope effects greatly affect
the state distributions of the product molecules. The total DCSs References
have almost forward–backward symmetric angular distributions;
however, the state-resolved DCSs show nonstatistical behavior. 1 A. Dalgarno, E. Herbst, S. Novick and W. Klemperer,
Herein, accurate quantum dynamics calculations for the Astrophys. J., 1973, 183, L131–L133.
N+(3P0) + H2 (v = 0, j = 0) and N+(3P0) + D2 (v = 0, j = 0) reactions 2 A. Dalgarno and J. H. Black, Rep. Prog. Phys., 1976, 39,
were performed on the ground state adiabatic PES. A rigorous 573–612.

This journal is © the Owner Societies 2019 Phys. Chem. Chem. Phys., 2019, 21, 22203--22214 | 22213
 View Article Online

Paper PCCP

3 N. G. Adams and D. Smith, Chem. Phys. Lett., 1985, 117, 67–70. 32 P. J. Knowles and H. J. Werner, Chem. Phys. Lett., 1988, 145,
4 K. M. Ervin and P. B. Armentrout, J. Chem. Phys., 1987, 86, 514–522.
2659–2673. 33 H. J. Werner, P. J. Knowles, G. Knizia, F. R. Manby and
5 J. B. Marquette, C. Rebrion and B. R. Rowe, J. Chem. Phys., M. Schutz, Wiley Interdiscip. Rev.: Comput. Mol. Sci., 2012, 2,
1988, 89, 2041–2047. 242–253.
6 M. Gonzalez, A. Aguilar and R. Sayos, Chem. Phys., 1989, 34 H. J. Werner and P. J. Knowles, J. Chem. Phys., 1985, 82,
132, 137–151. 5053–5063.
7 G. Nyman and U. Wilhelmsson, J. Chem. Phys., 1992, 96, 35 P. J. Knowles and H. J. Werner, Chem. Phys. Lett., 1985, 115,
5198–5212. 259–267.
Published on 07 August 2019. Downloaded by Universidad de Salamanca on 2/12/2020 6:48:29 PM.

8 P. Tosi, O. Dmitriev, D. Bassi, O. Wick and D. Gerlich, 36 R. A. Kendall, T. H. Dunning and R. J. Harrison, J. Chem.
J. Chem. Phys., 1994, 100, 4300–4307. Phys., 1992, 96, 6796–6806.
9 L. S. Sunderlin and P. B. Armentrout, J. Chem. Phys., 1994, 37 D. E. Woon and T. H. Dunning, J. Chem. Phys., 1995, 103,
100, 5639–5645. 4572–4585.
10 I. Zymak, M. Hejduk, D. Mulin, R. Plasil, J. Glosik and 38 J. C. Yuan, D. He and M. D. Chen, Phys. Chem. Chem. Phys.,
D. Gerlich, Astrophys. J., 2013, 768, 86. 2015, 17, 11732–11739.
11 T. P. Grozdanov and R. McCarroll, J. Phys. Chem. A, 2015, 39 J. C. Yuan, D. He, S. F. Wang, M. D. Chen and K. L. Han,
119, 5988–5994. Phys. Chem. Chem. Phys., 2018, 20, 6638–6647.
12 T. P. Grozdanov, R. McCarroll and E. Roueff, Astron. Astro- 40 Z. J. Yang, J. C. Yuan, S. F. Wang and M. D. Chen, RSC Adv.,
phys., 2016, 589, 735–745. 2018, 8, 22823–22834.
13 S. G. Hansen, J. M. Farrar and B. H. Mahan, J. Chem. Phys., 41 J. Cui and R. V. Krems, Phys. Rev. Lett., 2015, 115, 073202.
1980, 73, 3750–3762. 42 J. Cui and R. V. Krems, J. Phys. B: At., Mol. Opt. Phys., 2016,
14 B. H. Mahan and W. E. W. Ruska, J. Chem. Phys., 1976, 65, 49, 224001.
5044–5051. 43 B. Kolb, P. Marshall, B. Zhao, B. Jiang and H. Guo, J. Phys.
15 S. D. Peyerimhoff and R. J. Buenker, Chem. Phys., 1979, 42, Chem. A, 2017, 121, 2552–2557.
167–176. 44 Y. F. Guan, S. Yang and D. H. Zhang, Mol. Phys., 2018, 116,
16 D. Gerlich, J. Chem. Phys., 1989, 90, 3574–3581. 823–834.
17 G. Nyman, J. Chem. Phys., 1992, 96, 3603–3612. 45 B. Jiang, J. Li and H. Guo, Int. Rev. Phys. Chem., 2016, 35,
18 M. A. Gittins and D. M. Hirst, Chem. Phys. Lett., 1975, 35, 479–506.
534–536. 46 S. F. Wang, J. C. Yuan, H. X. Li and M. D. Chen, Phys. Chem.
19 H. P. D. Liu and G. Verhaegen, J. Chem. Phys., 1970, 53, Chem. Phys., 2017, 19, 19873–19880.
735–745. 47 K. P. Huber and G. Herzberf, Constants of Diatomic Molecules,
20 M. Gonzalez, A. Aguilar and Y. Fernandez, Chem. Phys., Springer, 1979.
1986, 104, 57–66. 48 R. Colin and A. E. Douglas, Can. J. Phys., 1968, 46, 61.
21 U. Wilhelmsson, P. E. M. Siegbahn and R. Schinke, J. Chem. 49 B. J. Braams and J. M. Bowman, Int. Rev. Phys. Chem., 2009,
Phys., 1992, 96, 8202–8211. 28, 577–606.
22 M. A. Gittins and D. M. Hirst, Chem. Phys. Lett., 1979, 65, 50 B. Jiang and H. Guo, J. Chem. Phys., 2013, 139, 054112.
507–510. 51 M. T. Hagan and M. B. Menhaj, IEEE Trans. Neural Netw.
23 U. Wilhelmsson and G. Nyman, J. Chem. Phys., 1992, 96, Learn. Syst., 1994, 5, 989–993.
1886–1895. 52 D. M. Hirst, Mol. Phys., 1978, 35, 1559–1568.
24 C. L. Russell and D. E. Manolopoulos, J. Chem. Phys., 1999, 53 Z. G. Sun, S. Y. Lee, H. Guo and D. H. Zhang, J. Chem. Phys.,
110, 177–187. 2009, 130, 174102.
25 L. Bonnet and J. C. Rayez, Chem. Phys. Lett., 1997, 277, 54 Z. G. Sun, H. Guo and D. H. Zhang, J. Chem. Phys., 2010,
183–190. 132, 084112.
26 L. Bonnet and J. C. Rayez, Chem. Phys. Lett., 2004, 397, 106–109. 55 Z. G. Sun, X. Lin, S. Y. Lee and D. H. Zhang, J. Phys.
27 P. G. Jambrina, F. J. Aoiz, N. Bulut, S. C. Smith, G. G. Chem. A, 2009, 113, 4145–4154.
Balint-Kurti and M. Hankel, Phys. Chem. Chem. Phys., 56 S. Gomez-Carrasco and O. Roncero, J. Chem. Phys., 2006,
2010, 12, 1102–1115. 125, 054102.
28 P. G. Jambrina, J. M. Alvarino, F. J. Aoiz, V. J. Herrero and V. Saez- 57 J. C. Yuan, D. He and M. D. Chen, Sci. Rep., 2015, 5, 14594.
Rabanos, Phys. Chem. Chem. Phys., 2010, 12, 12591–12603. 58 Z. P. Sun, C. F. Zhang, S. Y. Lin, Y. J. Zheng, Q. T. Meng and
29 L. Banares, F. J. Aoiz, P. Honvault, B. Bussery-Honvault and W. S. Bian, J. Chem. Phys., 2013, 139, 014306.
J. M. Launay, J. Chem. Phys., 2003, 118, 565–568. 59 P. Larregaray and L. Bonnet, J. Chem. Phys., 2015,
30 R. Tarroni, P. Palmieri, A. Mitrushenkov, P. Tosi and 143, 144113.
D. Bassi, J. Chem. Phys., 1997, 106, 10265–10272. 60 Z. G. Sun, D. H. Zhang, C. X. Xu, S. L. Zhou, D. Q. Xie,
31 H. J. Werner and P. J. Knowles, J. Chem. Phys., 1988, 89, G. Lendvay, S. Y. Lee, Y. Lin and H. Guo, J. Am. Chem. Soc.,
5803–5814. 2008, 130, 14962.

22214 | Phys. Chem. Chem. Phys., 2019, 21, 22203--22214 This journal is © the Owner Societies 2019