PCCP
Dynamic Article Links
Cite this: Phys. Chem. Chem. Phys., 2011, 13, 7644–7648
PAPER
www.rsc.org/pccp
Crystal structure and dynamics of Mg(ND3)6Cl2w
Magnus H. Sørby,*a Ole Martin Løvvik,ab Masami Tsubota,c Takayuki Ichikawa,c
Yoshitsugu Kojimac and Bjørn C. Haubacka
Received 12th August 2010, Accepted 7th December 2010
DOI: 10.1039/c0cp01479f
The crystal structure and dynamics of Mg(ND3)6Cl2 have been investigated by powder neutron
diffraction and molecular dynamics. The powder diffraction data can be well described by 4
partly occupied deuterium sites in a square arrangement around the N atoms, which is seemingly
inconsistent with the 3-fold symmetry of the ND3 molecule. Molecular dynamics show highly
correlated rotational and translational motion of the ND3 molecules which explains the apparent
4-fold symmetry of the deuterium arrangement. A more disordered structure model based on the
molecular dynamics results gives a better fit to the experimental data and is in agreement with the
3-fold symmetry of ND3.
Introduction
Metal ammine halides M(NH3)nXm are produced by absorption
of ammonia into a metal halide MXm.1 Such solid/ammonia
systems have been investigated for ammonia separation purposes,2
as chemical heat pumps3 and for energy storage. Christensen
et al. suggested to use Mg(NH3)6Cl2 in combination with an
ammonia decomposition catalyst as a solid hydrogen storage
material4 with 9.1 mass% hydrogen. Other possibilities to use
Mg(NH3)6Cl2 in ammonia-mediated energy storage systems
are for example in combination with direct ammonia fuel
cells,5 or reaction of ammonia with metal hydrides to generate
hydrogen reversibly.6,7 The solid Mg(NH3)6Cl2 is easy to
handle, and can store ammonia safely with almost the same
volumetric density as liquid ammonia.8
The atomic positions of Mg, N and Cl in Mg(NH3)6Cl2 have
been determined from single crystal X-ray diffraction by
Hwang et al.9 No attempts were done to localize the H atoms.
The phase takes a cubic K2PtCl6-type structure (space group
Fm3m, a = 10.19 Å), which implies a face-centred cubic
packing of Mg with Cl in all tetrahedral interstices (Fig. 1a).
This can alternatively be described as a primitive cubic
packing of Cl with Mg in the body centre of every second
cube. Each Mg atom is octahedrally coordinated by six N
atoms with the Mg–N bonds parallel to the unit cell axes.
The N atoms are close to the face centres, but slightly
a
Institute for Energy Technology, Physics Department, P.O. Box 40,
2027 Kjeller, Norway. E-mail: magnuss@ife.no;
Fax: +47 6381 09 20; Tel: +47 6380 6000
b
SINTEF Materials and Chemistry, P.O. Box 124 Blindern,
0314 Oslo, Norway
c
Institute for Advanced Materials Research, Hiroshima University,
Higashi-Hiroshima 739-8530, Japan
w This article was submitted following the 1st workshop on Energy
Materials, organised by The Thomas Young Centre, and held on 7–9
September 2010 at University College London.
7644
Phys. Chem. Chem. Phys., 2011, 13, 7644–7648
out-of-plane, of the Cl cube that surrounds the MgN6 octahedron (Fig. 1b).
Full crystal structure determination and Fourier density
maps of several Ni hexammines, Ni(NH3)6X2 (X = Br , I ,
NO3 and PF6 ; H = natural hydrogen or deuterium), with
the K2PtCl6-type structure have been performed by single
crystal neutron diffraction.10–13 A common feature in the
density maps is four clear hydrogen density maxima in a
square planar configuration for each NH3 unit. The hydrogen
density maxima are between the N atom and the four X
anions in the closest face of the surrounding cube of X atoms.
Such a configuration is indicated in Fig. 1b. Calculations of
the crystal potential energy have shown that these positions
are favourable for hydrogen12,13 in agreement with the experimental results, which is not surprising considering the positive
charge of H in ammonia and the negative charge of the X
anions. The higher number of H density maxima than
H atoms means that the NH3 complexes are orientationally
disordered. Quasielastic neutron scattering data on
Ni(NH3)6Br2 have shown that the disorder is dynamic.14
Despite the apparent inconsistency between the 3-fold
symmetry of the ammonia molecule and the 4-fold symmetry
of maxima in the H density, several hexammine crystal
structures have been reported with four 75% occupied
hydrogen positions per ammonia molecule, in e.g. V(NH3)6I2,
Cr(NH3)6I2, Mn(NH3)6Cl2, Fe(NH3)6Cl2, Fe(NH3)6Br2,
Co(NH3)6Br2, Ni(NH3)6Cl2,15 Mn(NH3)6I2 and Fe(NH3)6I2.16
Other investigators have used a higher number of less
occupied sites to model the hydrogen distribution in e.g.
Co(NH3)6Cl2.17 However, only very few publications have
discussed the relationship between the local, instantaneous
orientations of NH3 and the four observed H density maxima.
Schiebel et al.13 proposed that the rotational motion of the
ammonia molecules in Ni(NH3)6I2 are strongly coupled with
translational motion. The N atom is shifted towards the centre
This journal is
c
the Owner Societies 2011
Fig. 1 (a) The K2PtCl6-type structure of Mg(NH3)6Cl2 without hydrogen atoms. (b) Cubic configuration of Cl atoms around the
MgN6-octahedra. The 4-fold hydrogen configuration often reported in isostructural hexammines is indicated. Small, dark spheres: magnesium;
large, light spheres: nitrogen; large, dark spheres: chlorine; small, white spheres: typical hydrogen sites, 75% occupied (see text for discussion).
of an edge in the H density square so that two of the square
corners can be occupied by H simultaneously. Thus, 2/3 of the
H will most of the time be close to the square corners, which
explain the clear H density maxima there.
In the present paper, powder neutron diffraction (PND)
and molecular dynamics (MD) are used to investigate the
deuterium distribution in Mg(ND3)6Cl2.
Methodology
Mg(ND3)6Cl2 was synthesized in a direct solid-state–gas reaction between MgCl2 (99.99%, Aldrich Co. Ltd.) and deuterated
ammonia, ND3 (99%, Euriso-top CEA group). Deuterated
ammonia was used due to the high incoherent neutron scattering
cross-section for natural hydrogen. 0.4 g MgCl2 was transferred
to a stainless steel reaction chamber with volume about 8 cm3
in a glove box with purified argon (o1 ppm O2, o1 ppm H2O).
The reaction chamber was evacuated and connected to a
60 cm3 reservoir volume where 5 bar ND3 was introduced at
ambient temperature. The temperature immediately increased
to about 50 1C due to the exothermic formation of the ammine
complex. The reaction was allowed to proceed for 2 hours but
appeared to be mostly completed after about 30 minutes as no
more heat was evolved. The final pressure was around 3 bar.
The reaction product was a snow white, fine powder.
Powder neutron diffraction (PND) data were collected at
the PUS instrument at the reactor JEEP II (Kjeller, Norway).18
The sample was contained in a rotating vanadium container
with 6 mm inner diameter. Neutrons with the wavelength
l = 1.555 Å were provided by a vertically focusing Ge
(5 1 1) monochromator. The instrument features two detector
banks, each with 7 vertically stacked 3He-filled position sensitive
detector tubes that cover a 201 range in scattering angle. The
2y range from 101 to 1301 was thus covered by moving each
detector bank to 3 different positions. The data were analyzed
with the Rietveld method implemented in the software
package GSAS19 with the graphical user interface EXPGUI.20
This journal is
c
the Owner Societies 2011
The Bragg profiles were described by the Thompson–Cox–
Hastings pseudo-Voigt function21 with 3 free parameters. The
background was fitted with a 12 term Chebyshev polynomial.
Different atoms of the same element were constraint to have
the same isotropic displacement factor.
First-principles molecular dynamics (MD) was performed
using the Vienna ab initio Simulation Package (VASP).22,23
The electronic structure was represented by plane-waves using
the projector augmented wave method24 and the generalized
gradient approximation.25 The plane-wave cut-off was 500 eV
for the MD runs, and 780 eV for the initial relaxation. The
k-space integration grid had a maximum distance of 0.2 Å 1
between the k-points. Smearing of the partial occupancies was
performed using the linear tetrahedron method, and the
convergence criterion for the electronic density relaxation
was 10 6 eV.
Since partial occupancy is not possible to implement in
atomistic studies like this one, the NH3 units were directly
represented, starting from the gas phase structure of NH3,
with the hydrogen atoms pointing in the same direction as in
the experimental structure with partial occupancy. The structure resulting from this construction was then relaxed, using
the quasi-Newton method with the RMM-DIIS algorithm.
The force relaxation criterion was 0.05 eV Å 1. This relaxed
structure was used as the input for the MD calculations. The
MD temperature was 300 K, with initial velocities randomly
generated by VASP. The time step was 1.3 fs, and the total
number of MD steps was 1400.
Results and discussion
The PND data could be fully indexed according to the cubic
unit cell suggested by Hwang et al. (a = 10.19 Å with
systematic absence according to F-centring).9 Rietveld refinement was first performed using the simple model described
above, with four 75% occupied D sites arranged in a square
for each N (space group Fm
3m). The refined model was in
Phys. Chem. Chem. Phys., 2011, 13, 7644–7648
7645
Fig. 2 Rietveld fit to PND data for Mg(ND3)6Cl2 using four 75%
occupied deuterium positions pr. ND3 unit (Model I). Open circles—
experimental data, solid line—calculated data, below—difference plot.
Bragg peak positions are marked with vertical ticks. Rwp = 5.10%.
good agreement with the data (Fig. 2) despite the inconsistency
with the 3-fold ND3 geometry (Rwp = 5.10%). The refined
structure data are shown in Table 1. The refined N–D
distances are 0.992(4) Å which is in good agreement with that
in gaseous ammonia (1.02 Å26). The D–N–D angles are on the
other hand too low as expected due to the 4-fold symmetry
(84.41 vs. 107.81 for gaseous ammonia). For comparison,
a model with 12 D sites (each 25% occupied) evenly distributed on a circle for each N-atom gave a much poorer fit
(Rwp = 7.49%). The latter would be a better description of the
D distribution if ND3 was rotating as a stiff unit between
four different positions with one D atom always pointing
toward a Cl.
To clarify this apparent contradiction, first-principles molecular
dynamics (MD) calculations were performed. The initial
structure model was generated from the refined model above,
but with the 0.75 occupied D4 groups replaced by fully
occupied D3 groups, giving molecular units ND3 resembling
the experimental structure of gaseous ND3. Fig. 3 shows the
trace of two ND3 units, viewed along their common 3-fold
symmetry axis, after 1000 MD steps. It is evident that the
average positions of the deuterium atoms are consistent with
the 4-fold symmetry demonstrated by diffraction. At the
same time, the local geometry of the ND3 units is quite stable
during the MD simulation (apart from various phonon
modes that should be expected at room temperature). This
means that the ammonia units may be regarded as bodies
with limited flexibility and molecular structure similar to that
of gaseous ND3.
Table 1 Results from Rietveld refinement of powder neutron diffraction data for Mg(ND3)6Cl2 at 298 K using Model I. Space group
Fm
3m, a = 10.199(1) Å. Calculated standard deviations are given in
parentheses
Atom Site
Mg
Cl
N
D
7646
x
y
z
4a (fixed) 0
0
0
1
1
8c (fixed) 14
4
4
24e (0 0 z) 0
0
0.2195(3)
96k (x x z) 0.0654(2) 0.0654(2) 0.2498(4)
Biso21/Å Occupancy
3.4(3) 1
3.4(1) 1
4.70(9) 1
6.4(1) 34
Phys. Chem. Chem. Phys., 2011, 13, 7644–7648
Fig. 3 Track of two ND3 units from the MD simulation. One unit is
directly below the other in the viewing direction. Each configuration is
represented by grey circles for the deuterium positions and black
circles for the nitrogen positions. The MD simulation was performed
at 300 K, with a time step of 1.3 fs. The first 1000 steps are shown in
this figure.
Furthermore, the movement of the ammonia molecules was
inspected from the MD movie. It was seen that the ND3 units
for most of the time stayed with D atoms in two of the four
density maxima seen in Fig. 3, with occasional rotations
between such structures. This implies that the nitrogen atoms
most of the time are shifted from their average position
towards one of the edges of the surrounding square. Opposite
ND3 molecules were seen to form 4-fold symmetry with four
of their six D atoms most of the time, except during the
rotations, which were clearly correlated.
This led us to propose an alternative structure model
(Model II) for the structural refinement. This is based on rigid
ND3 units with two of the D atoms placed in corners of the
square defined by the D positions of Model I. This leads to a
third D atom near the opposing square edge and a shift of the
N atom away from the centre of the square.
Rietveld refinement was used to check the consistency
between the PND data and Model II. The simple model
in Fig. 1b (Model I) was thus modified by reducing the
occupancy of the original D site from 75% to 50% (D1) and
introducing new, 25% occupied D sites (D2) between them. To
reduce the bias towards the suspected atomic arrangement, D2
was initially positioned so that the distances N–D1 and N–D2
were equal. This configuration yielded a poorer fit to the
experimental data (Rwp = 6.61%) as long as the atomic
positions were fixed. Refinement of the D positions did,
however, give a slightly better fit (Rwp = 5.01%) than for
Model I. The N–D1 distance was elongated while N–D2 was
shortened during the refinement in accordance with the
MD-model. In the last stage of the refinement, N was moved
slightly away from its 24e (x 0 0) position into a 96j position
(x dy 0) while reducing the occupancy to 25%. The N atoms
moved distinctly away from the 24e site on refinement. The fit
to the experimental data is improved with a final Rwp = 4.82
(Fig. 4). A Hamilton test27 shows that the decrease in Rwp is
significant within a 25% confidence level with the 3 additional
This journal is
c
the Owner Societies 2011
striking. The arrangement of partly occupied sites can easily be
regarded as a weighted superposition of the 4 different ND3
positions and orientations shown in Fig. 5. The N–D distances
within the outlined molecules are 0.966(6) Å (N–D1) and
1.06(1) Å (N–D2) which are in good agreement with the
reference value of 1.02 Å for free ammonia molecules. The
D–N–D angles are 102.8(3)1 (D1–N–D1) and 110.1(3)1
(D1–N–D2) which are in fair agreement with the experimentally determined gas phase value of 107.81. It should be noted
that the refinement was performed without any geometrical
constrains.
Conclusions
Fig. 4 Rietveld fit to PND data for Mg(ND3)6Cl2 using a superposition of four different ND3 orientations (Model II). Open circles—
experimental data, solid line—calculated data, below—difference plot.
Bragg peak positions are marked with vertical ticks. Rwp = 4.82%.
Table 2 Results from Rietveld refinement of powder neutron diffraction
data for Mg(ND3)6Cl2 at 298 K using Model II. Space group Fm3m,
a = 10.199(2) Å. Calculated standard deviations are given in parentheses.
Biso are constrained to have the same value for the same elements
Atom Site
x
y
z
Mg
Cl
N
D1
D2
0
0
0
1
4
1
4
1
4
4a (fixed)
8c (fixed)
96j (0 y z)
96k (x x z)
96j (0 y z)
Biso21/Å Occupancy
3.4(3) 1
3.0(1) 1
0
0.021(1) 0.2161(3) 4.70(9) 14
0.0739(3) 0.0739(3) 0.2427(5) 4.3(1) 12
0
0.0690(8) 0.2670(9) 4.3(1) 14
structural parameters (two for D2 and one for N). The refined
structural parameters are given in Table 2.
A ND3 unit from the converged structure (Model II) is
shown in Fig. 5 and the similarities with the MD model are
Fig. 5 Refined ND3 arrangement in Model II regarded as a weighted
superposition of four ND3 orientations. Small sphere: nitrogen; large
sphere: deuterium. The average nitrogen position from Model I is
found in the middle of the square. The deuterium atoms on the square
corners are denoted D1 and those slightly off the square edges are D2
in Table 2. The size of the N atoms is reduced for clarity.
This journal is
c
the Owner Societies 2011
Two crystal structure models for Mg(ND3)6Cl2 are proposed
and refined against PND data. In Model I, four partly
occupied deuterium sites arranged in a square are associated
with each nitrogen atom (Model I). The model is in good
agreement with the available PND data, but it is incompatible
with the 3-fold symmetry of ammonia. Model II was developed
based on MD simulations that showed that the nitrogen atom
is most of the time displaced from its ‘‘average’’ position, thus
allowing two deuterium atoms to be close to the D positions in
Model I simultaneously. It yields a better fit to the experimental data and is, more importantly, in good agreement with
the expected geometry of ammonia. Model II is thus preferred
over Model I.
Notes and references
1 W. Biltz and G. F. Hütting, Z. Anorg. Allg. Chem., 1921, 119,
115–131.
2 C. Y. Liu and K. Aika, Bull. Chem. Soc. Jpn., 2004, 77, 123–131.
3 W. Wongsuwan, S. Kumar, P. Neveu and F. Meunier, Appl.
Therm. Eng., 2001, 21, 1489–1519.
4 C. H. Christensen, R. Z. Sorensen, T. Johannessen, U. J. Quaade,
K. Honkala, T. D. Elmoe, R. Kohler and J. K. Norskov, J. Mater.
Chem., 2005, 15, 4106–4108.
5 J. C. Ganley, J. Power Sources, 2008, 178, 44–47.
6 Y. Kojima, S. Hino, K. Tange and T. Ichikawa, Mater. Res. Soc.
Symp. Proc., 2008, 1042, S06-01.
7 H. Yamamoto, H. Miyaoka, S. Hino, H. Nakanishi, T. Ichikawa
and Y. Kojima, Int. J. Hydrogen Energy, 2009, 34, 9760–9764.
8 C. H. Christensen, T. Johannessen, R. Z. Sorensen and
J. K. Norskov, Catal. Today, 2006, 111, 140–144.
9 I. C. Hwang, T. Drews and K. Seppelt, J. Am. Chem. Soc., 2000,
122, 8486–8489.
10 A. Hoser, W. Joswig, W. Prandl and K. Vogt, Mol. Phys., 1985, 56,
853–869.
11 A. Hoser, W. Prandl, P. Schiebel and G. Heger, Z. Phys. B:
Condens. Matter, 1990, 81, 259–263.
12 P. Schiebel, A. Hoser, W. Prandl, G. Heger, W. Paulus and
P. Schweiss, J. Phys.: Condens. Matter, 1994, 6, 10989–11005.
13 P. Schiebel, A. Hoser, W. Prandl, G. Heger and P. Schweiss,
J. Phys. I, 1993, 3, 987–1006.
14 J. A. Janik, J. M. Janik, A. Migdal-Mikuli and E. Mikuli,
Acta Phys. Pol., A, 1988, 74, 423–431.
15 R. Essmann, G. Kreiner, A. Niemann, D. Rechenbach,
A. Schmieding, T. Sichla, U. Zachwieja and H. Jacobs, Z. Anorg.
Allg. Chem., 1996, 622, 1161–1166.
16 H. Jacobs, J. Bock and C. Stuve, J. Less Common Met., 1987, 134,
207–214.
17 J. M. Newman, M. Binns, T. W. Hambley and H. C. Freeman,
Inorg. Chem., 1991, 30, 3499–3502.
18 B. C. Hauback, H. Fjellvåg, O. Steinsvoll, K. Johansson,
O. T. Buset and J. Jørgensen, J. Neutron Res., 2000, 8, 215.
Phys. Chem. Chem. Phys., 2011, 13, 7644–7648
7647
19 A. C. Larson and R. B. von Dreele, Los Alamos National
Laboratory Report, Los Alamos, 2004.
20 B. H. Toby, J. Appl. Crystallogr., 2001, 34, 210–221.
21 P. Thompson, E. D. Cox and J. B. Hastings, J. Appl. Crystallogr.,
1987, 20, 79–83.
22 G. Kresse and J. Furthmüller, Phys. Rev. B: Condens. Matter,
1996, 54, 11169–11186.
23 G. Kresse and J. Hafner, Phys. Rev. B: Condens. Matter, 1993, 47,
558–561.
7648
Phys. Chem. Chem. Phys., 2011, 13, 7644–7648
24 G. Kresse and D. Joubert, Phys. Rev. B: Condens. Matter Mater.
Phys., 1999, 59, 1758–1775.
25 J. P. Perdew, J. A. Chevary, S. H. Vosko, K. A. Jackson,
M. R. Pederson, D. J. Singh and C. Fiolhais, Phys. Rev. B:
Condens. Matter, 1992, 46, 6671.
26 J. C. Bailar, H. J. Emeléus, R. Nyholm and A. F. TrotmanDickenson, Comprehensive Inorganic Chemistry, Pergamon Press,
Oxford, 1973.
27 W. C. Hamilton, Acta Crystallogr., 1965, 18, 502–510.
This journal is
c
the Owner Societies 2011