international journal of hydrogen energy 34 (2009) 1922–1930

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/he

Hydrogen dissociation and diffusion on transition metal

([ Ti, Zr, V, Fe, Ru, Co, Rh, Ni, Pd, Cu, Ag)-doped
Mg(0001) surfaces

M. Pozzoa,b, D. Alfèa,b,c,d,*
a
Material Simulation Laboratory, University College London, Gower Street, London WC1E 6BT, United Kingdom
b
Department of Earth Sciences, University College London, Gower Street, London WC1E 6BT, United Kingdom
c
Department of Physics and Astronomy, University College London, Gower Street, London WC1E 6BT, United Kingdom
d
London Centre for Nanotechnology, University College London, 17–19 Gordon Street, London WC1H 0AH, United Kingdom

article info abstract

Article history: The kinetics of hydrogen absorption by magnesium bulk is affected by two main activated
Received 6 March 2008 processes: the dissociation of the H2 molecule and the diffusion of atomic H into the bulk.
Received in revised form In order to have fast absorption kinetics both activated processed need to have a low
9 October 2008 barrier. Here we report a systematic ab initio density functional theory investigation of H2
Accepted 14 November 2008 dissociation and subsequent atomic H diffusion on TM (¼ Ti, V, Zr, Fe, Ru, Co, Rh, Ni, Pd, Cu,
Available online 19 January 2009 Ag)-doped Mg(0001) surfaces. The calculations show that doping the surface with TMs on
the left of the periodic table eliminates the barrier for the dissociation of the molecule, but
Keywords: the H atoms bind very strongly to the TM, therefore hindering diffusion. Conversely, TMs
Metal hydrides on the right of the periodic table do not bind H, however, they do not reduce the barrier to
Hydrogen absorption dissociate H2 significantly. Our results show that Fe, Ni and Rh, and to some extent Co and
First principles calculations Pd, are all exceptions, combining low activation barriers for both processes, with Ni being
the best possible choice.
ª 2008 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights
reserved.

1. Introduction (i) they need to be capable of storing hydrogen in excess of

6.5% in weight, (ii) the kinetics of absorption has to be fast, i.e.,
Hydrogen is regarded by many as a possible energy vehicle (or with time scales of minutes, and (iii) the temperature at which
fuel) for future mobile applications, and has been targeted to they release hydrogen (the decomposition temperature) needs
replace the current use of liquid hydrocarbons in the next few to be ideally in the range 20–100  C. Cyclability of the material
decades [1]. Unlike fossil fuels, it is an environmentally is also a desired property.
friendly, non-polluting fuel simply because its combustion Metal hydrides are natural hydrogen storing materials, and
product is water, provided it is produced by renewable energy the relatively strong bonds between hydrogen and the host
resources, obviously. metal satisfy the safety requirement. Unfortunately, however,
One of the main challenges faced by the development of no material which has all the properties mentioned above
the so-called hydrogen economy is the capability of storing exists today (see for example the review by Sakintuna et al. [2]).
hydrogen safely and efficiently. For mobile applications the Magnesium hydride, MgH2, satisfies some of the above
storage materials need to satisfy a number of requirements: requirements. It has high storage capacity (7.6 wt%), good

* Corresponding author. London Centre for Nanotechnology, University College London, 17–19 Gordon Street, London WC1H 0AH,
United Kingdom.
E-mail address: [email protected] (D. Alfè).
0360-3199/$ – see front matter ª 2008 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijhydene.2008.11.109
 international journal of hydrogen energy 34 (2009) 1922–1930 1923

cyclability and it is relatively inexpensive. However, its and also that the strength of the hydrogen-metal bond is
enthalpy of formation is too high (76 kJ/mol), requiring similar to that on the pure metal surface. The strength of the
temperatures in excess of 300  C to decompose it into H2 and H-TM (TM ¼ Ni, Ti) bonding was found to be correlated to the
Mg bulk. The formation of the hydride also has slow kinetics height of the diffusion barrier. We therefore might expect to
[3,4], making this material not good enough. However, MgH2 see an analogous trend in the dissociation and diffusion
represents a good test material to study how various treat- barriers by doping the Mg(0001) surface with various transi-
ments can affect its properties. In particular, it has been found tion metals. In fact, we will show that the elements on the left
that by doping the material with transition metals can weaken of the periodic table make the H2 dissociation barrier to vanish
the Mg–H bond and reduce the stability of the hydride (see for but are responsible for high diffusion barriers, while those on
example [5–10], and references therein). Ball-milling further the right cannot catalyse the dissociation of the molecule.
enhances the sorption processes by increasing the number of Among the elements studied here, we found that Fe, Ni and
possible paths for the diffusion of H (see for example [5,8,11] Rh, and to some extent Co and Pd offer a good compromise
and references therein). A new method of chemical fluid between the promotion of dissociation and the hindering of
deposition in supercritical fluids (SCF) has also recently been diffusion, and qualify as good catalysts for accelerating the
used on metal hydrides [12]. This method offers the same kinetics of hydrogen absorption.
sorption properties of ball-milled samples, but with a hugely
improved cyclability (the catalytic effect of the metal being
almost constant for SCF samples, while decreasing for ball- 2. Computational method
milled samples after about 100 cycles) and in particular shows
that the catalytic effect of Ni on hydrogen sorption processes All the density functional theory (DFT) calculations were per-
is higher than that of Pd. formed with the ab initio simulation package VASP [21] using
Liang et al. [13] and Schulz et al. [14] found that V and Ti are the projector augmented wave (PAW) method [22,23] and the
better catalysts than Ni for hydrogen absorption and desorp- PBE exchange-correlation functional [24]. An efficient charge
tion from MgH2–metal composites, showing faster absorption density extrapolation was used to speed up the calculations
kinetics at T w 300  C and faster desorption kinetics above [25]. We used a plane-wave basis set to expand the electronic
w250  C than other 3d-elements investigated. They also found wave-functions with the same plane-wave energy cutoff of
an enthalpy of hydride formation for different catalysts 270 eV as in Ref. [20], which guarantees convergence of
similar to that of MgH2. By contrast, theoretical calculations adsorption energies within 1 meV. Surfaces were modeled
and experimental results of Song et al. [7] and Shang et al. [8] using periodic slabs, with five atomic layers and a vacuum
led to different conclusions. They found that the stability of thickness of 10 Å. The topmost three atomic layers were
MgH2–Ni is reduced when compared to that of MgH2–Ti. allowed to relax, while the bottom two were held fixed to the
Moreover (and contrary to the experimental findings of Liang positions of bulk Mg. Calculations were performed using 2  2
et al. [13]), the heat of formation of the metal-doped MgH2 surface unit cells, with 9  9  1 k-point grids and replacing
hydrides is smaller than that of MgH2. In particular, MgNiH2 one of the four surface Mg atoms by one TM (¼ Ti, V, Zr, Fe, Ru,
shows a smaller enthalpy of formation than MgTiH2. Co, Rh, Ni, Pd, Cu, Ag) atom. These settings were extensively
Zaluska et al. [5] used Li, Al, V, Mn, Zr and Y as catalysts for tested and guarantee convergence of activation energies to
the hydrogenation/dehydrogenation of Mg alloy samples. better than 0.02 eV. Activation energies have been calculated
According to their results, V remarkably improves the H with the nudged elastic band (NEB) method [26] using 17
absorption kinetics, but Zr is better for lower temperature H. replicas, which proved to be sufficient to reach convergence of
However, the best kinetic results are achieved with mixtures, activation energies to better than 0.01 eV, and display all the
i.e., V þ Zr or Mn þ Zr Mg alloys. Bobet et al. [15] have shown main features of the minimum energy path. The initial state of
that the hydrogen storage properties are enhanced when the NEB calculations for the dissociation of H2 is represented
using reactive mechanical alloying of Mg þ 10 wt% Co, Ni and by the hydrogen molecule sitting on top the TM at a distance of
Fe mixtures. Co, unlike Ni, is found to significantly increase 5 Å, and the final state is the most energetically favourable
the quantity of MgH2 formed. However, Bobet et al. [6] later among four possible adsorption sites for the two dissociated
reported that the hydrogen sorption properties of Mg–Co hydrogen atoms (see Ref. [20] for details). For the diffusion
mixtures are less effective than those reported for MgH2– process, the initial state is represented by the final state of the
metal mixtures. Gutfleish et al. [16] have recently presented dissociation process, and the final state by a configuration
results achieved with an Mg sample alloyed with Ni (1 wt%) where one of the two hydrogen atoms has been displaced into
and Pd (0.2 wt%). Their sample shows excellent hydrogen a nearby hollow site (see details in Section 3.2).
absorption/desorption kinetics and cyclic stability, exhibiting Figs. 2 and 3 have been made using the XCRYSDEN soft-
an overall reversible H2 storage capacity of 6.3 wt%. ware [27].
Previous theoretical and experimental investigations over
pure surfaces of transition metals belonging to the left of the
periodic table have shown that H2 dissociation is promoted, but 3. Results
also that the bonding between the hydrogen atoms and the
metal is strong (see [17] and references therein; see also [18,19]). In the following section, we report calculations for the bulk
In our previous paper [20], we have shown that H2 disso- structural parameters of the various elements investigated, as
ciation on the metal (Ni,Ti)-doped Mg surface has a barrier a test of the quality of the PAW and the PBE exchange-corre-
similar to that on the corresponding pure metal (111) surface, lation functionals. In Section 3.2, we report results for the H2
1924 international journal of hydrogen energy 34 (2009) 1922–1930

dissociation and diffusion barriers, which we also analyse in

Table 1 – Bulk properties of pure transition metals
terms of the position of the centre of the d-band of the various (TM [ Ti, V, Fe, Co, Ni, Cu, Zr, Ru, Rh, Pd, Ag). For each
transition metals employed as dopants on the Mg(0001) element we report the bulk lattice constant a (together
surface. with c/a for hcp metals), the bulk modulus k0, the
electrons treated as valence (VE) and the core radius rcore
3.1. Bulk parameters of the PAW potentials. References for values previously
reported in literature follow in the last column.
We obtained bulk structural properties of the pure transition a (Å), c/a k0 (GPa) VE, rcore (Å) References
metals by calculating energy versus volume curves, and fitting Ti 2.92, 1.583 120 3d24s2, 1.5 Ref. [20]
them to a Birch–Murnaghan equation of state [28]. The 2.92, 1.583 118 Ref. [29]
elements investigated here were Zr, V, Fe, Ru, Co, Rh, Pd, Cu [Expt.] [2.95, 1.59] [105] Ref. [31]
and Ag, together with Ti and Ni already presented in Ref. [20]. V 3.00 179 3s23p63d34s2, 1.2 This worka
The bulk parameters were derived using 13  13  13 and 3.00 182 Ref. [30]
18  18  12 k-point grids for those metals with the cubic and 2.99 185 Ref. [29]
the hexagonal structure, respectively. The corresponding [Expt.] [3.00] [162] Ref. [31]
standard version of the PAW potential was used for all of Fe 2.84 169 3d64s2, 1.2 This worka
them, with the exception of V and Zr for which we used the 2.71 281 Ref. [32],c
version of the PAW treating, respectively, the 3s23p63d34s2 and [Expt.] [2.87] [168] Ref. [31]
4s24p64d25s2 electrons in valence (which give results closer to Co 2.49, 1.617 212 3d74s2, 1.2 This worka
the experimental values than the standard versions of the 2.40, 1.62 384 Ref. [32],c
PAW functionals). Calculated bulk parameters values are [Expt.] [2.51, 1.62] [191] Ref. [31]
reported in Table 1, together with the details of the PAW Ni 3.52 194 3d84s2, 1.2 Ref. [20]
potentials, which include the electrons treated in valence and 3.52 194 Ref. [33]
the core radii. Overall, the lattice parameter a is always 3.52 201 Ref. [34]
overestimated, and the bulk modulus is underestimated with [Expt.] [3.52] [186] Ref. [31]
respect to the experimental values. This is in agreement with Cu 3.64 136 3d104s1, 1.2 This worka
the findings from previous theoretical calculations, and is 3.63 142 Ref. [35]
typical of the PBE functional. Among the elements investi- [Expt.] [3.61] [137] Ref. [31]
gated, only Ni, Co and Fe are magnetic. We find a magnetic Zr 3.24, 1.602 93 4s24p64d25s2, 1.3 This worka
moment of 0.63 [20], 1.70 and 2.15 mB/atom for Ni, Co and Fe, 2.99, 1.86 108 Ref. [32],b
respectively, which are in agreement with the corresponding 3.23, 1.600 101 Ref. [36]
experimental values of 0.61, 1.71 and 2.22 mB/atom [31]. [Expt.] [3.23, 1.59] [83] Ref. [31]
However, as discussed in Section 3.2, Fe is the only element [3.23, 1.59] [92  3] Ref. [37]

which required spin-polarised calculations. Ru 2.72, 1.578 310 4d75s1, 1.4 This worka
2.68, 1.59 360 Ref. [32],c
3.2. H2 dissociation and diffusion 2.69, 1.606 322 Ref. [38]
[Expt.] [2.71, 1.58] [321] Ref. [31]

The activation barriers for H2 dissociation over the various Rh 3.84 251 4d85s1, 1.3 This worka
metal-doped Mg surfaces are reported in Fig. 1 and Table 2, 3.83 259 Ref. [39]
where we also report two experimental values for the H2 3.86 258 Ref. [34]
[Expt.] [3.80] [270] Ref. [31]
dissociation/recombination on the Mg(0001) surface. The value
reported in Ref. [48] (w1.0 eV) refers to the recombination Pd 3.95 163 4d10, 1.3 This worka
barrier (which, in this particular case, is similar to the dissoci- 3.95 163 Ref. [35]
[Expt.] [3.89] [181] Ref. [31]
ation barrier) identified with the barrier for desorption from the
surface. This value was not directly measured in the thermal Ag 4.17 88 4d105s1, 1.3 This worka
programmed desorption (TPD) experiments of Ref. [48] because 4.20 87 Ref. [40]
[Expt.] [4.09] [101] Ref. [31]
complete desorption spectra as function of temperature could
not be taken, due to the onset of Mg sublimation at w450 K a Reported values do not include room temperature thermal
which overlaps with the temperature at which H2 desorbs. expansion.
However, it was noted that the onset of H2 desorption appears b From tight-binding calculations.
c From tight-binding non-magnetic calculations.
at 425 K, which is similar to that of the H/Be(0001) system that
has a determined desorption energy of w1 eV [50], and so, by
analogy, it was suggested that the activation energy for 1.05 eV, respectively. It should be noted, however, as pointed
desorption might be the same on the H/Mg(0001) system too. out in Ref. [49], that the experimental situation may not be the
The value reported in Ref. [49] (0.75  0.15 eV) has been same as the theoretical ones, due to the possible presence of
obtained by the interpretation of TPD experiments performed steps on the surface which might be more reactive sites and
on a 400-Å thick magnesium film. This value is in good agree- lower the H2 dissociation barrier. Moreover, the inferred
ment with the calculated PBE dissociation energy, but is dissociation energy of 0.75  0.15 eV is based on the use of the
significantly lower than the dissociation energies calculated Arrhenius relations with assumed pre-factors of w1012 Hz.
with RPBE by Vegge [45] and Du et al. [46] which are 1.15 and As shown in Refs. [51] and [52], these values could be
 international journal of hydrogen energy 34 (2009) 1922–1930 1925

experiments and other DFT (GGA and PBE) calculations [19,53–

Table 2 – Activation energies for H2 dissociation (Ediss) on
the pure Mg and metal-doped Mg surfaces (ordered by 57] over the pure Cu(111) surface. In addition, we find that the
increasing atomic number). energy difference between the final state and the initial state
is 0.19 eV, which is in line with the findings of Kratzer et al.
Metal surface Ediss (eV)
[58], who found that H2 dissociation on the pure Cu(111)
Pure Mg 0.87a, 0.4b,c, 0.5d,e, 1.15f, 1.05g, 0.95h, surface is exothermic, with a gain of 0.2 eV. On the Pd-doped
[Expt.] 1.0i, 0.75  0.15j
surface we calculate a dissociation barrier of 0.39 eV, which
Ti-doped Mg Nulla, negligibleg
reduces to 0.30 eV when a smaller 5  5  1 k-points grid is
V-doped Mg Nullk
Fe-doped Mg 0.03k used (instead of the 9  9  1 grid). The reason for mentioning
Co-doped Mg 0.03k the result obtained with the coarser grid is because we want to
Ni-doped Mg 0.06a compare with the findings of Dong et al. [59], who performed
Cu-doped Mg 0.56k calculations with a similar grid for the dissociation of H2 on
Zr-doped Mg Nullk the Pd(111) surface, and found a barrier of 0.29 eV when the
Ru-doped Mg Nullk
molecule dissociates on top a Pd atom, therefore a value very
Rh-doped Mg 0.04k
close to our value of 0.30 eV on the Pd-doped Mg surface (note,
Pd-doped Mg 0.39k
Ag-doped Mg 1.18k however, that they report the bridge–bridge as the preferred
dissociation path, with a barrier of only 70 meV, which is
a Ref. [20].
consistent with the theoretical value found by Nobuhara et al.
b Ref. [41] for a jellium system.
[60] and in good agreement with the experimental value of
c Ref. [42] from DFT LDA calculations and PES. This lower value as
compared to other calculations is explained as due to the well- 50 meV [61]). Finally, in the case of Ag, our activation barrier of
known LDA over-binding. 1.18 eV is in agreement with the experimental results which
d Ref. [43] for a jellium system and PES. predict a dissociation barrier on Ag(111) larger than that on Cu
e Ref. [44] for a jellium system and PES. and somewhat larger than 0.8 eV [62], and we also agree with
f Ref. [45] from DFT RPBE. previous theoretical calculations [63] also obtained on the
g Ref. [46] from DFT PAW RPBE calculations.
pure Ag(111) surface, which reported an activation energy of
h Ref. [47] from PES calculations.
i Ref. [48] (see comments in the main text).
1.11 eV. Fig. 1 shows the MEPs for hydrogen dissociation on
j Ref. [49] (see comments in the main text). a TM ¼ V, Fe, Co, Cu, Zr, Ru, Rh, Pd, Ag-doped Mg surface
k This work. investigated here, together with those on a TM ¼ Ni, Ti-doped
Mg surface that we have reported previously [20].
Fig. 2 shows the dissociation of H2 on the Ag-doped Mg
underestimated by more than two orders of magnitudes surface, which is found to have the largest barrier value
because the classical pre-factors do not include the enhance- among all the dopants investigated here (since this barrier is
ment due to the large entropy increase as the molecules leave larger than that on pure Mg, obviously H2 will not dissociate
the surface, in which case the activation energy could be up to onto this site, but will rather choose some regions of the Mg
w0.25 eV higher. surface free of Ag). For Fe, Co, Cu, Rh and Pd-doped Mg
The geometry of adsorption of the H atoms on the metal- surfaces, the images at the IS, TS and FS of the NEB are similar,
doped Mg surfaces listed in Table 2 appears to be somewhat with the H2 molecule at the TS sitting closer (i.e., Cu and Pd) or
correlated to the height of the dissociation barrier. We find further away from the surface (i.e., Fe, Co and Rh) compared to
that when the barrier is large (i.e., for Cu, Pd, Ag) the H atoms
Ag−doped Mg
fall into filled hollow sites, and when the barrier is null (i.e., for pure Mg
1 Cu−doped Mg
Ti, Zr, V and Ru) they fall into empty hollow sites. In between Pd−doped Mg
Ni−doped Mg
there are elements showing a small energy barrier for which Rh−doped Mg
Fe−doped Mg
the preference towards filled hollow sites (i.e., for Ni) instead 0.5 Co−doped Mg
Relative Energy (eV)

Ru−doped Mg
of empty hollow sites (i.e., for Fe, Co and Rh) is weaker (less V−doped Mg
Zr−doped Mg
than about 30 meV). Ti−doped Mg

Fe is the only dopant for which magnetic calculations are 0

really required, with the total magnetic moment on the Fe
atom being of 2.8 and 2.5 mB in the initial and final state of the
−0.5
dissociation process, respectively (Co is magnetic too, but
when used as dopant of the Mg surface our calculations show
that it can be treated as non-magnetic). In particular, non- −1
magnetic calculations for the Fe-doped surface would give
significantly different results, reducing the dissociation
barrier to almost zero and increasing by 60% the energy −1.5
difference between the initial and final states.
Reaction Coordinates
As we noted before [20], the activation barrier for the
dissociation of H2 over a metal-doped Mg surface is similar to Fig. 1 – Minimum energy paths for the dissociation of the
that on the corresponding pure metal surface. For example, in H2 molecule, and subsequent diffusion of one of the two H
the case of Cu our calculated barrier is 0.56 eV, which is close atoms, on a pure Mg surface and on (Ti, V, Fe, Co, Ni, Cu, Zr,
to the dissociation barrier of about 0.5 eV suggested by Ru, Rh, Pd, Ag)-doped Mg surfaces.
1926 international journal of hydrogen energy 34 (2009) 1922–1930

Fig. 2 – H2 (dark red) dissociation over the Ag-doped Mg surface as viewed from side (top figures) and top (bottom figures)
positions respectively at IS (left-hand panel), TS (central panel) and FS (right-hand panel). The Mg, Ag and H atoms are
represented respectively by light grey, dark grey and black colours. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.)

the behavior shown in our previous paper [20] for the Ni- is the dopant. In other words, it appears that H2 dissociates on
doped Mg surface (see dHsurf values reported in Table 3). For top of the dopant atom for those doped Mg surfaces which
Ag we note that the hydrogen molecule at the TS dissociates show a very small barrier (i.e., Fe, Co, Ni and Rh), slightly
closer to the surface than on the Ni-doped Mg surface, and shifted to the side of the dopant atom for the Pd and Cu-doped
that it does so on a side of the dopant atom (see Fig. 2). Mg surfaces having a non-negligible barrier, and completely
A closer look at the geometry of the dissociation process on the side of the dopant atom on the Ag-doped Mg surface
shows an interesting correlation between the height of the which shows a very large dissociation barrier.
barrier and the geometry of the transition state. The dissoci- The dissociation of the H2 molecule is only the first step for
ation of molecular hydrogen into two hydrogen atoms the absorption of hydrogen. A second fundamental step is the
happens on top of the dopant atom when Fe, Co, Ni or Rh are diffusion of the products away from the catalytic site. To study
used as dopants, slightly shifted to the side when Pd and Cu this, we performed NEB calculations in which the initial state
are the metals used as dopants, and fully on the side when Ag was the final state of the dissociation process, and the final

Fig. 3 – H (dark red) diffusion on the Fe-doped Mg surface as viewed from top. Figures show positions at the final state of the
dissociation which is the initial state for the diffusion process (left), at the transition state (centre) and final state (right) of
the diffusion process. The Mg, Fe and H atoms are represented respectively by light grey, dark grey and black colours. (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
 international journal of hydrogen energy 34 (2009) 1922–1930 1927

Table 3 – The d-band centre position with respect to the

Fermi energy (Ed), the activation energy barrier for the 1.1
dissociation of H2 (Ediss), the energy difference between

Dissociation Barrier (eV)

Ag

Diffusion Barrier (eV)

the final and initial state (EFS–IS
diss
) of the dissociation, the 0.9 Cu

activation energy barrier for the diffusion of atomic H Pd

Ni
(Ediff) and the corresponding energy difference (EFS–ISdiff ) on 0.7 Rh
the pure Mg surface as opposed to the metal-doped Mg Fe
surfaces (these have been ordered so as to highlight the 0.5 Co
overall dependence along each column of the periodic Ru
table, as we go from right to left across the periodic table). V
0.3
Also reported in the last column is the average distance of Ti
Zr
molecular hydrogen from the surface as measured at the
transition state, dH–surf . 0.1

Surface Ed Ediss EFS–IS

diss Ediff EFS–IS
diff dHsurf
-0.1
(eV) (eV) ðeVÞ (eV) ðeVÞ (Å) -5 -4 -3 -2 -1 0 1 2

Mg pure a
– 0.87 0.04 0.11 0.02 0.9
d-band center (eV)
Ag-doped Mgb 4.14 1.18 0.15 null 0.18 0.7
Fig. 4 – Activation energy barrier for hydrogen dissociation
Cu-doped Mgb 2.27 0.56 0.19 0.10 0.04 0.7
(black) and diffusion (red) of hydrogen on pure Mg and
Pd-doped Mgb 1.84 0.39 0.18 0.08 0.07 1.0
Ni-doped Mga 0.79 0.06 0.66 0.27 0.09 1.5 metal-doped Mg surfaces as a function of the d-band centre
Rh-doped Mgb 0.75 0.04 0.72 0.31 0.05 1.8 positions. The dashed lines have been drawn for eye
Co-doped Mgb 0.16 0.03 1.03 0.41 0.07 2.0 guidance only.
Ru-doped Mgb 0.14 null 1.20 0.54 0.26 –
Fe-doped Mgb 0.72 0.03 0.76 0.30 0.17 2.0
V-doped Mgb þ0.82 null 1.39 0.73 0.68 –
Zr-doped Mgb þ1.32 null 1.46 0.94 0.94 – The step-limiting process in the hydrogen absorption is the
Ti-doped Mga þ1.08 null 1.34 0.75 0.74 – one with the largest energy barrier between dissociation and
a Ref. [20]. diffusion, so the best dopant is the one which minimizes the
b This work. largest energy barrier. It is customary to define the activity of
a catalyst in terms of the rate of the reaction which is being
catalysed. This can often be accurately approximated by an
state was obtained by displacing one H into a nearby hollow Arrhenius relation, and therefore the natural logarithm of the
site. Fig. 3 shows the diffusion path of one of the two hydrogen rate is proportional to the negative of the activation energy
atoms on the Fe-doped Mg surface as an example. The MEPs barrier. We can then interpret the maximum of the two
for the diffusion processes are also shown in Fig. 1. We barriers shown in Fig. 4 as indicating the activity of the cata-
observe that the height of the diffusion barrier Ediff is strongly lyst. If we draw a line across these points, we see that the
anti-correlated to the height of the dissociation barrier Ediss various transition metals investigated here fit on an inverse
(see Table 3). In fact, Ti, V, Zr and Ru have zero dissociation volcano plot, with Ni, Fe and Rh sitting near the top of the
barriers, but they bind the products very strongly, which volcano, and therefore being the most active catalysts.
results in high values of Ediff. By contrast, Ag, Cu and Pd
produce large dissociation barriers, but they have low diffu-
sion barriers (in fact, no barrier at all for Ag). In between there 1.25
are Fe, Ni and Rh, which represent the best compromise in
combining low activation barriers for both processes. Ni is the
0.75
best possible choice overall.
Ag
We note in passing that the catalytic effect of Ni dopant on
E(FS) – E(IS) (eV)

E(FS) – E(IS) (eV)

Cu
0.25
MgH2 for the dehydrogenation process (not studied here) has Pd
Ni
been experimentally demonstrated by Jensen et al. [64], Rh
-0.25 Fe
showing an activation energy reduced by 0.5 eV with respect Co
Ru
to that obtained with pure MgH2.
-0.75 V
It is interesting to correlate the height of the barriers with Ti
the position of the d-band of the transition metal dopant with Zr

respect to the Fermi energy EF (here we define the d-band, pd (E ), 1.25

as the projection of the electronic density of states onto d type
spherical harmonics). In particular, it is useful to consider the -1.75
-5 -4 -3 -2 -1 0 1 2 3
first energy moment of the d-band, or d-band centre, defined as
R E0 d-band center (eV)
Ed ¼ N dEðE  EF Þpd ðEÞ, where E0 is some cutoff energy which
we chose to be 7 eV above the Fermi energy. In Fig. 4 we plot the Fig. 5 – The energy difference between the final and initial
dissociation and the diffusion energy barriers, Ediss and Ediff, state, E(FS-IS), for hydrogen dissociation (black) and
against Ed for all the systems explored (see Table 3). It is obvious diffusion (red) on pure Mg and metal-doped Mg surfaces as
that the heights of the barriers are strongly correlated with the a function of the d-band centre positions. The dashed lines
position of the d-band centre. have been drawn for eye guidance only.
1928 international journal of hydrogen energy 34 (2009) 1922–1930

As a matter of interest, in Fig. 5 we plot the energy differ- references

ence between the final and the initial state EFS–IS (both for
the dissociation and for the diffusion process) as a function
of the d-band centre. We can observe some correlation [1] Schlapbach L, Züttel A. Hydrogen storage materials for
between the two quantities, although this is less strong than mobile applications. Nature (London) 2001;414(6861):
that observed in Fig. 4 for the height of the two energy barriers. 353–8.
If follows that the correlation between the energy barriers and [2] Sakintuna B, Lamari-Darkrim F, Hirscher M. Metal hydride
materials for solid hydroge storage: a review. Int J Hydrogen
EFS–IS is also weaker than that between the energy barriers
Energy 2007;32(9):1121–40.
and the d-band centre, leaving the latter a better parameter to [3] Huot J, Liang G, Boily S, Van Neste A, Schulz R. Structural
characterize the catalyst. study and hydrogen sorption kinetics of ball-milled
From an inspection of Figs. 4 and 5 the d-band centre magnesium hydride. J Alloys Compd 1999;293–295:495–500.
correlation is evident, and points to an ideal d-band centre [4] Zaluska A, Zaluski L, Ström-Olsen JO. Structure, catalysis and
value of about 1.29 eV. This value cannot be obtained with atomic reactions on the nano-scale: a systematic approach
to metal hydrides for hydrogen storage. Appl Phys A 2001;
any of the TM-doped Mg surfaces investigated here.
72(2):157–65.
Recently, Vegge et al. [9] have investigated magnesium 3d
[5] Zaluska A, Zaluski L, Ström-Olsen JO. Nanocrystalline
TM alloys. They showed that the d-band centre values of magnesium for hydrogen storage. J Alloys Compd 1999;
the expanded alloys obtained with TM belonging to the first 288(1–2):217–25.
raw of the Periodic Table range from þ0.93 to 6.88 eV going [6] Bobet JL, Chevalier B, Darriet B. Effect of reactive mechanical
from MgSc to MgZn. In particular, MgCu gives a value of grinding on chemical and hydrogen sorption properties of
2.37 eV while the neighbor MgNi 0.82 eV. It would be the Mg þ 10 wt.% Co mixture. J Alloys Compd 2001;330–332:
738–42.
interesting to broaden their investigation to 4d and 5d TMs
[7] Song Y, Guo ZX, Yang R. Influence of selected alloying
to see if the optimal d-band centre value of about 1.3 eV elements on the stability of magnesium dihydride for
that we have extrapolated here could be obtained with hydrogen storage applications: a first-principle investigation.
some alloys, but this is beyond the purpose of the present Phys Rev B 2004;69(9):094205–15.
investigation. [8] Shang CX, Bououdina M, Song Y, Guo ZX. Mechanical
alloying and electronic simulations of (MgH2 þ M) systems
(M ¼ Al, Ti, Fe, Ni, Cu and Nb) for hydrogen storage. Int J
Hydrogen Energy 2004;29(1):73–80.
4. Conclusions [9] Vegge T, Hedegaard-Jensen LS, Bonde J, Munter TR,
Nørskov JK. Trends in hydride formation energies for
We have performed here a systematic DFT/PBE study of magnesium 3d transition metal alloys. J Alloys Compd 2005;
hydrogen dissociation and subsequent diffusion over Mg 386(1–2):1–7.
surfaces doped with different transition metals. The dopants [10] Bhat V, Rougier A, Aymard L, Nazri GA, Tarascon JM.
Enhanced hydrogen storage property of magnesium hydride
investigated were Ti, Zr, V, Fe, Ru, Co, Rh, Ni, Pd, Cu and Ag.
by high surface area Raney nickel. Int J Hydrogen Energy
We have observed that the transition metals on the left of 2007;32(18):4900–6.
the periodic table (Ti, V, Zr), together with Ru, eliminate the [11] Zaluski L, Zaluska A, Tessier P, Ström-Olsen JO, Schulz R.
dissociation barrier altogether, however, the products stick Effects of relaxation on hydrogen absorption in Fe–Ti
too strongly to the metal dopant, therefore hindering diffu- produced by ball-milling. J Alloys Compd 1995;227(1):53–7.
sion away from the catalytic site. This would result in [12] Bobet JL, Aymonier C, Mesguich D, Cansell F, Asano K,
Akiba E. Particle decoration in super critical fluid to improve
a quick deactivation of the catalyst and therefore a slow
the hydrogen sorption cyclability of magnesium. J Alloys
absorption process. On the contrary, the transition metals on
Compd 2007;429(1–2):250–4.
the right of the periodic table do not bind too strongly the H [13] Liang G, Huot J, Boily S, Van Neste A, Schulz R. Catalytic
atoms (in fact, Ag does not bind them at all), allowing easy effect of transition metals on hydrogen sorption in
diffusion, however, their effect on the dissociation barrier is nanocrystalline ball milled MgH2–Tm (Tm ¼ Ti, V, Mn, Fe and
small. We have shown that these two opposite catalytic Ni) systems. J Alloys Compd 1999;292(1–2):247–52.
properties are well correlated to the position of the d-band [14] Schulz R, Liang G, Huot J. Hydrogen sorption in mechanically
alloyed nanocrystalline and disordered materials. In:
centre, according to the Hammer and Nørskov [65] model. In
Dinesen, editor. Proceedings of the 22nd Risø international
fact, we have shown that the catalytic activity for the H
symposium on material science: Science of metastable and
absorption process can be described well by a volcano plot, nanocrystalline alloys: Structure, properties and modelling;
with the most active catalysts Ni, Fe and Rh sitting near the 2001. p. 141–53. Risø National Laboratory, Roskilde, Denmark.
top of the volcano. [15] Bobet J-L, Even C, Nakamura Y, Akiba E, Darriet B. Synthesis
of magnesium and titanium hydride via reactive mechanical
alloying Âdinfluence of 3d-metal addition on MgH2
synthesize. J Alloys Compd 2000;298(1–2):279–84.
Acknowledgments [16] Gutfleisch O, Dal Toè S, Herrich M, Handstein A, Pratt A.
Hydrogen sorption properties of Mg–1 wt.% Ni–0.2 wt.% Pd
prepared by reactive milling. J Alloys Compd 2005;404–406:
This work was conducted as part of a EURYI scheme award as
413–6.
provided by EPSRC (see www.esf.org.euryi). Calculations have
[17] Ward JW. Electronic structure, bonding and chemisorption in
been performed on the LCN cluster at University College metallic hydrides. J Less-Common Met 1980;73(1):183–92.
London. We thank Sam French and Alvaro Amieiro for very [18] Christmann K. Interaction of hydrogen with solid surfaces.
useful discussions. Surf Sci Rep 1988;9(1–3):1–163.
 international journal of hydrogen energy 34 (2009) 1922–1930 1929

[19] Hammer B, Nørskov JK. Why gold is the noblest of all the [40] Li WX, Stampfl C, Scheffler M. Oxygen adsorption on Ag(111):
metals. Nature 1995;376(6537):238–40. a density-functional theory investigation. Phys Rev B 2002;
[20] Pozzo M, Alfè D, Amieiro A, French S, Pratt A. Hydrogen 65(7):75407–25.
dissociation and diffusion on Ni- and Ti-doped Mg(0001) [41] Hjelmberg H. Hydrogen chemisorption on Al, Mg and Na
surfaces. J Chem Phys 2008;128(9):094703–13. surfaces Âdcalculation of adsorption sites and binding
[21] Kresse G, Furthmüller J. Efficient iterative schemes for ab energies. Surf Sci 1979;81(2):539–61.
initio total-energy calculations using a plane-wave basis set. [42] Bird DM, Clarke LJ, Payne MC, Stich I. Dissociation of H2 on
Phys Rev B 1996;54(16):11169–86. Mg(0001). Chem Phys Lett 1993;212(5):518–24.
[22] Blöchl PE. Projector augmented-wave method. Phys Rev B [43] Johansson PK. Chemisorption of molecular hydrogen on
1994;50(24):17953–79. simple metal surfaces. Surf Sci 1981;104(2–3):510–26.
[23] Kresse G, Joubert D. From ultrasoft pseudopotentials to the [44] Nørskov JK, Høumoller A, Johansson PK, Lundqvist BI.
projector augmented-wave method. Phys Rev B 1999;59(3): Adsorption and dissociation of H2 on Mg surfaces. Phys Rev
758–1775. Lett 1981;46(4):257–60.
[24] Perdew JP, Burke K, Ernzerhof M. Generalized gradient [45] Vegge T. Locating the rate-limiting step for the interaction
approximation made simple. Phys Rev Lett 1996;77(18): of hydrogen with Mg(0001) using density-functional theory
3865–8. calculations and rate theory. Phys Rev B 2004;70(3):
[25] Alfè D. Ab initio molecular dynamics, a simple algorithm 035412–8.
for charge extrapolation. Comp Phys Comm 1999;118(1): [46] Du AJ, Smith SC, Yao XD, Lu GQ. First-principle study of
31–3. adsorption of hydrogen on Ti-doped Mg(0001) surface. J Phys
[26] Mills G, Jonsson H, Schenter GK. Reversible work transition Chem B 2006;110(43):21747–50.
state theory –Application to dissociative adsorption of [47] Arboleda Jr NB, Kasai H, Nobuhara K, Dino WA, Nakanishi H.
hydrogen. Surf Sci 1995;324(2–3):305–337; Henkelman G, Dissociation and sticking of H2 on Mg(0001), Ti(0001) and
Johnsson H. Improved tangent estimate in the nudged La(0001) surfaces. J Phys Soc Jpn 2004;73(3):745–8.
elastic band method for finding minimum energy paths [48] Sprunger PT, Plummer RW. An experimental study of the
and saddle points. J Chem Phys 2000;113(22):9978–9985; interaction of hydrogen with the Mg(0001) surface. Chem
Henkelman G, Uberuaga BP, Johnsson H. A climbing image Phys Lett 1991;187(6):559–64.
nudged elastic band method for finding saddle points [49] Johansson M, Ostenfeld CW, Chorkendorff I. Adsorption of
and minimum energy paths. J Chem Phys 2000;113(22): hydrogen on clean and modified magnesium films. Phys Rev
9901–9904. B 2006;74(19):193408–11.
[27] Kokalj A. Computer graphics and graphical user interfaces as [50] Ray KB, Hannon JB, Plummer EW. An experimental study of
tools in simulations of matter at the atomic scale. Comp hydrogen adsorption on beryllium. Chem Phys Lett 1990;
Mater Sci(2):155–68. Code available from: http://www. 171(5–6):469–74.
xcrysden.org/, 2003;28. [51] Alfè D, Gillan M. Absolute rate of thermal desorption from
[28] Birch F. Finite elastic strain of cubic crystals. Phys Rev 1947; first-principles simulation. J Phys: Condens Matter 2006;
71(11):809–24. 18(37):L451–7.
[29] Jahnátek M, Krajèı́ M, Hafner J. Interatomic bonding, [52] Alfè D, Gillan MJ. Ab initio statistical mechanics of surface
elastic properties, and ideal strength of transition metal adsorption and desorption. I. H2O on MgO(001) ay low
aluminides: a case study for Al3(V,Ti). Phys Rev B 2005;71(2): coverage. J Chem Phys 2007;127(11):114709–20.
024101–16. [53] Hammer B, Scheffler M, Jacobsen KW, Nørskov JK.
[30] Verma AK, Modak P. Structural phase transitions in Multidimensional potential energy surface for H2
vanadium under high pressure. Europhys Lett 2008;81(3). dissociation over Cu(111). Phys Rev Lett 1994;73(10):1400–3.
37003(1–5). [54] Gross A, Hammer B, Scheffler M, Brenig W. High dimensional
[31] Kittel C. Introduction to solid state physics. 7th ed. New quantum dynamics of adsorption and desorption of H2 at
York: Wiley; 1996. Cu(111). Phys Rev Lett 1994;73(23):3121–4.
[32] Mehl MJ, Papaconstantopoulos DA. Applications of a tight- [55] Murphy MJ, Hogdson A. Adsorption and desorption
binding total-energy method for transition and noble metals: dynamics of H2 and D2 on Cu(111): the role of surface
elastic constants, vacancies, and surfaces of monatomic temperature and evidence for corrugation of the dissociation
metals. Phys Rev B 1996;54(7):4519–30. barrier. J Chem Phys 1998;108(10):4199–211.
[33] Kresse G, Hafner J. First-principles study of the adsorption of [56] Šljivanèanin Ž, Hammer B. H2 dissociation at defected Cu:
atomic H on Ni(111), (100) and (110). Surf Sci 2000;459(3): preference at vacancy and kink sites. Phys Rev B 2002;65(8):
287–302. 085414–7.
[34] Pozzo M, Carlini G, Rosei R, Alfè D. Comparative study of [57] Sakong S, Groß A. Dissociative adsorption of hydrogen on
water dissociation on Rh(111) and Ni(111) studied with strained Cu surfaces. Surf Sci 2003;525(1–3):107–18.
first principles calculations. J Chem Phys 2007;126(16): [58] Kratzer P, Hammer B, Nørskov JK. Geometric and electronic
164706–17. factors determining the differences in reactivity of H2 on
[35] Da Silva JLF, Stampfl C, Scheffler M. Converged properties of Cu(100) and Cu(111). Surf Sci 1996;359(1–3):45–53.
clean metal surfaces by all-electron first-principles [59] Dong W, Kresse G, Hafner J. Dissociative adsorption of H2 on
calculations. Surf Sci 2006;600(3):703–15. the Pd(111) surface. J Mol Catal A: Chem 1997;119(1–3):69–76.
[36] Nie Y, Xie Y. Ab initio thermodynamics of the hcp metals Mg, [60] Nobuhara K, Kasai H, Dinõ WA, Nakanishi H. H2 dissociative
Ti, and Zr. Phys Rev B 2007;75:174117–23. adsorption on Mg, Ti, Ni, Pd and La surfaces. Surf Sci 2004;
[37] Zhao Y, Zhang J, Pantea C, Qian J, Daemen LL, Rigg PA, et al. 566:703–7.
Thermal equations of state of the a, b, and u phases of [61] Resch C, Berger HF, Rendulic KD, Bertel E. Adsorption
zirconium. Phys Rev B 2005;71:184119–24. dynamics for the system hydrogen/palladium and its
[38] Nie Y, Xie Y, Peng H, Xiaobo L. Ab initio thermodynamics of relation to the surface electronic structure. Surf Sci 1994;
metals: Pt and Ru. Physica B 2007;395(1–2):121–5. 316(3):L1105–9.
[39] Ganduglia-Pirovano MV, Scheffler M. Structural and [62] Healey F, Carter RN, Worthy G, Hodgson A. Endothermic
electronic properties of chemisorbed oxygen on Rh(111). dissociative chemisorption of molecular D2 on Ag(111). Chem
Phys Rev B 1999;59(23):15533–43. Phys Lett 1995;243(1–2):133–9.
1930 international journal of hydrogen energy 34 (2009) 1922–1930

[63] Xu Y, Greeley J, Mavrikakis M. Effect of subsurface oxygen on nickel-doped magnesium hydride investigated by in situ
the reactivity of the Ag(111) surface. J Am Chem Soc 2005; time-resolved powder X-ray diffraction. Int J Hydrogen
127(37):12823–7. Energy 2006;31(14):2052–62.
[64] Jensen TR, Andresen A, Vegge T, Andreasen JW, Stahl K, [65] Hammer B, Nørskov JK. Electronic factors determining the
Pedersen AS, et al. Dehydrogenation kinetics of pure and reactivity of metal surfaces. Surf Sci 1995;343(3):211–20.