Surface Science Letters 259 (1991) L791-L796
North-Hoiiaod
surface science
letters
Surface Science Letters
Synthesis and decomposition of ammonia on transition metal surfaces:
bond-order-conse~ation-Morse-potential
analysis
Evgeny Shustorovich
Corporate Reseurch Laboratories, Emtman Kodak Company, Rochester, NY 14650-2001,
USA
and
Alexis T. Bell
Center for Advanced Materials, Chemical Sciences Division, Lawrence Berkeley Laboratory
and Department of Chemical Engineering, University of California, Berkeley, CA 94720, USA
Received 23 April 1991; accepted for publication 28 August 1991
The mechanism of ammonia synthesis and decomposition on transition metal surfaces has been analyzed using the BOGMP
(bond-order-~nse~ation-Mode-potentiai~
method. The analysis is based on calculations of the heats of chemisorption, &, for a&l
adsorbed species and activation barriers, A E *, for all elementary reactions believed to be involved in the reaction Nz + Hz @ NH,
over Ptfltl), RuK@1), Fe(llO), Re(OO1). The relevant experimental values of Q and AE * agree well with the BOC-MP estimates.
It is shown that along the periodic series Pt, Ru, Fe, Re, the dissociation activation barriers decrease but the recombination and
desorption barriers increase. In particular, we find that on all the surfaces the largest activation barrier corresponds to the
recombinative desorption 2N, -+ N,. This step is projected to be the rate-determining process for ammonia decomposition, and Pt
is projected to be the most efficient catalyst. For the dissociation N, + 2N,, we find that the activation barrier sharply increases in
the order Re I Fe Q: Pt, which makes Pt surfaces incapable of catalyzing ammonia synthesis. These and other BOC-MP
projections are in agreement with the results of mechanistic studies on Pt(lll), Ru(~l), and Fe(ll0).
1. Int~uction
Recently, we have refined our BOC-MP
(bond-order-conservation-Morse-potential)
method to allow one to calculate, in a uniform and
coherent way, the heats of chemisorption Q of
various species (diatomic and poIyatomic, monoand dicoordinated, with closed and open electronic shells) and activation barriers AE * for
dissociation, recombination, and disproportionation of adsorbates on metal surfaces Cl]. This
refined version has been used successfully to analyze the reaction pathways for Fischer-Tropsch
synthesis over Group VIII metals [2], methanol
0039-6028/91/$03.50
synthesis over Cu and Pd [31, and methane and
methanol on Ni and Pd [4]. The goal of the
present work is to apply the BOC-MP method to
the hydrogenation of N, on transition metal surfaces, and, more specifically, to determine the
energetics of the mechanism of ammonia synthesis and decomposition. To discern periodic reguIarities, we wilI examine the series Pt(lll),
Ru(OOl), Fe(llO), and Re(OOl). The results of our
theoretical projections are compared with mechanistic studies of ammonia decomposition on polycrystalline Pt [5], Ru@Ol~ 161, and single-crystal
Fe surfaces 171and ammonia synthesis on Fe(llO),
(loo), (Ill) surfaces [7a].
0 1991 - Elsevier Science Publishers B.V. All rights reserved
2.
Results and discussion
The formulas used to calculate Q and AE *
are summarized in the appendix. It is recalled
that the only parameters in the BOC-MP theory
that must be fitted to experimental data are the
atomic binding energies QA (A = H and N). The
experimental values of Qu and QN for Pt(lll),
Ru(OOl), Fe(llO), and Re(OO1) are given in table
1, which also lists the calculated values of QNH,,
x = l-3. These values of QNH, are compared
with the available experimental data [5-111 in
table 2. Given the differences in experimental
conditions and techniques used by different authors, the agreement between theory and experiment is very encouraging.
The mechanism of ammonia synthesis and decomposition N, + H, $ NH, has been proposed
to consist of the following elementary steps [7,121:
I&
e H,,s * 2% >
(1)
Nz,g + N,,s 3 2N,,
(2)
N, + H, @ NH,,
(3)
NH, + H, zi NH,,$,
(4)
NH,,s + H, @ NH,,s,
(5)
NH% +
+ NH,,.,
(6)
The calculated activation barriers for steps (l)-(6)
on Pt, Ru, Fe, and Re are given in table 3.
Comparison with experimental data can only be
made for the dissociation and recombination of
H, and N, (reactions (1) and (2)). As can be seen
from table 4, the agreement between theory and
experiment is remarkable.
Since mechanistic studies of ammonia synthesis on Fe, Ru, and Re [12] have demonstrated
that N, dissociation is the rate limiting step, it is
surprising that our BOC-MP calculations project
that the activation barrier for N, dissociation is
small (4-7 kcal/mol) compared to the barriers
for the subsequent hydrogenation of N, (see table
3). Since the calculated activation barriers for Fe
and Re are in good agreement with those observed e~erimentally
for thermal dissociation of
N,, the relative slowness of the process N2,g -+ 2N,
must be attributed to other factors. The most
probable explanation is that the sticking coefficient for dissociative adsorption is exceptionally
small. This interpretation
is consistent with the
experimental finding of sticking coefficients for
dissociative adsorption of N, on Fe(ll0) and
Re(OO1) in the range of 10-5-10-h [7,81.
Table 3 indicates that the activation barrier for
the process N, + H, -+ NH, for all four metals
lies between 36 and 42 kcal/moI, but that the
barriers for the subsequent hydrogenation steps
Table 1
Heats of chemisorption (Q) and total bond energies in the gas phase (D) and chemisorbed (D + QI states oflPt(lll),Ru(OOl),
Fe(llO), and Re@Ol) a)
Adsorbate
H
N
NH h,
NH2 hl
NH, it
Db’
75
169
279
‘) Ah energies in kcal/mol.
b, Ref. [7a], fig. 14.
‘) Ref. [5].
d, Ref. 161.
ef Ref. [7a].
n The extrapolated vatue.
gf Ref. 181.
h, Eq. (A.3).
i, Eq. (A.2).
RU
Pt
Re
Fe
Q
D+Q
Q
D+Q
Q
D+Q
Q
61 ‘)
116 ‘)
71
47
14
61
116
146
216
293
63 d,
135 d,
87
60
18
63
135
162
229
297
64 e,
139 e,
90
63
19
64
139
16.5
232
298
64 D
142 B)
93
6.5
20
See text for explanations.
D+Q
65
142
168
234
299
E. Shustorouich, A. T. Bell / BOC-MP analysis of synthesis and decomposition of NH3 on transition metal surfaces
Table 2
Comparison of calculated and e~erimentally
of chemisorption
System
Pt(lll)
Ru(OO1)
Fe(ll0)
Re(OO1)
14
18
19
20
NH,
Pt(ll1)
Fe0 10).
47
63
NH
Pt(lll)
FetllO)
71
90
Table 4
Comparison of calculated and experimentally observed activation barriers
Reaction
Q (kcal/mol)
Calc. ‘)
NH,
observed heats
Calc. a)
12
12-21
17
21
> 42
- 65
71-74
-100
[51
[6]
2% --* H2,g
Pt(ll1)
Ru(001)
Fe0101
21
23
24
19
22
24
Dal
[61
Val
N2,g
Pt(lll)
Fe0101
RefOOl)
21
5
4
16
7
3
[Sal
17al
Pt(lll)
Ru@Ol)
Fe(ll0)
Re@Ol)
27
50
57
62
22
44
51
60
Eal
[61
Bal
Val
[%I
Dal
--+ 2%
2% -+ N2,g
on the path to NH, are much smaller. This raises
the issue as to whether the first addition of hydrogen to N, occurs via reaction with H, or by
some other process. As an alternative, we have
considered the reaction
NS + H&S --+NH, + H,,
(7)
for which the activation barrier is smaller by
12-16 kcal/mol than that for reaction (3). This
AE,*
H 2.a
H 2-g
N 2.a
N 2.8
%+Hz.,
N,+H,
NH, + H,
NHa., + H,
NHzs + H,
NH,.,
@ Hz,,
ii 2H,
* N2,s
Ft 2N,
FINH,+HS
Ft NH,
* NH,,,
@ NH,,,
* NH3.s
* NH,.,
0
3
0
21
26
36
12
5
5
14
6 b,
21
9 =)
27
7
5
21
21
7
0
AE:
0
1
0
6
28
40
16
13
13
18
[63
Da3
@I
]7al
@I
suggests that if molecularly adsorbed H, is present in any significant concentration, it may serve
as the preferred source of hydrogen for forming
NH, species. One should also mention that the
surface of a working ammonia synthesis catalyst
Ru
AE:
-0
-0
a) From table 2.
Table 3
Activation barriers for forward AE: and reversed AE: reactions
decom~sition on Pt(lll), Ru(~l), Fe(llO), and Re(001) a)
Pt
1
[Sal
3
1
0
-+ 2%
[%I
191
Ref.
Exp.
Pt(lll)
Ru(001)
FetllO)
H2.g
a) From table 1.
Reaction
A E * (kcaI/mol)
Surface
Ref.
Exp.
comprising elementary
steps of ammonia synthesis and
Fe
Re
AE;
AE,*
AE:
6 b,
23
11 d,
50
8
4
20
18
0
‘0
0
0
0
5
29
41
17
15
17
19
6 b,
24
8 e,
57
,9
3
20
17
-2
0
AE:
AE:
0
-1
0
4
29
42
19
16
20
20
6 b,
25
6’-)
62
10
3
20,
16
-4
0
Calculated from eqs. (A.MA.7) with the relevant values of Q and D taken from table 1. The barriers for Na,s F, N2,, (Q,,>
were taken from experiment. Ail energies are in kcal/mol.
The assumed value (cf. ref. [l], table 2).
Ref. [lo].
Assumed the same as for N,/Ni(llO) [ll].
Ref. [7a].
Ref. 181.
E. Shustorouich, A. T. Bell / BOC-MF’ analysis of synthesis and decomposition ofNH3 on transition metal surfaces
may not be clean but, rather, contain a substantial amount of adsorbed atomic nitrogen. For the
sake of argument, let us consider a bee (100)
surface, say Fe(lO0). The BOC-MP theory predicts that when the nitrogen coverage exceeds
& = 0.25, the initial value of QN should decrease
and for the bee (lOO)-~(2 X 2)N surface, corresponding to ON= 0.5, the new value of QN will
become QG = 0,8OQ, (cf. eq. (17) in ref. [2b]).
Thus, instead of QN = 139 kcalfmol for Fe(100)
we project Q$ = 111 kcal/mol
for Fe(lOO)c(2 X 2)N, which is close to QN on Pt(ll1). This
change will contribute to a reduction in the activation barrier for the process N, + H, -+ NH,, as
well as al1 subsequent hydrogenation steps leading to NH,.
As seen from table 3, several periodic trends
relevant to ammonia synthesis and decomposition
can be discerned. The propensity to dissociate
chemical bonds and the propensity to recombine
chemisorbed species and desorb products are always opposite; nameIy, the former increases and
the latter decreases along the series Pt., Ru, Fe,
Re. In particular, the dissociation barrier AEt?jZL!
for N2,y -+ 2N, drops dramatically in the order
Pt z=-Ru > Fe > Re. The difference of more than
20 kcal/mol in the values of AE&, for Pt versus
Fe or Re makes for a difference of N lo9 in the
reaction rates at 500 K. Thus, we project that Ru,
Fe, and Re surfaces can readily dissociate N,, but
Pt surfaces cannot. Dissociative adsorption of H,
is practically unactivated on all four metals, and
hence the formation of H, will occur readily in all
cases. For a given surface, we find that the activation barriers for the process NH_ + H, -+
NH X+1s, x = O-2, decrease as x increases, and
that, for a given x, the activation barrier decreases in the order Re > Fe > Ru > Pt. The heat
of NH, adsorption, QNn,, decreases in the same
order.
From table 3 we immediately see that on ah
the surfaces the largest activation barrier is that
for nitrogen desorption (the reverse of reaction
(Z)), which we project to be the rate-determining
step for ammonia decomposition. Because this
barrier decreases along the series Re > Fe > Ru
> Pt, we further project that Pt surfaces should
be the best catalysts for decomposition of NH,.
Both of these projections are in full agreement
with the experimental findings on polycrystalline
Pt i.51,Ru(001) 161,and Fe(ll0) [7a].
The analysis given above for Fe has been restricted to Fe(llO), on which surface it is we11
known [13] that ~monia synthesis proceeds much
less rapidly than on the more open Fe(ll1) and
Fe(100) surfaces. It is worth noting, though, that
BOC-MP projections of the activation barrier for
N, dissociative adsorption indicate a decrease in
barrier height in the order Fe(ll0) > Fe(100) >
Fe(lll), consistent with e~erimental observation
E71.
Finally, we would like to comment on the
seeming contradiction in the activation barriers
for N, dissociation on Fe(ll1) obtained from
thermai dissociation studies, where no activation
barrier is found [7,14], and from molecular beam
studies revealing a distinct barrier which can be
overcome by translational activation [15]. In particular, molecular beam experiments show that
the initial probability for dissociative chemisorption of N,, S,, increases from N 10e6 for a
kinetic energy of 0.09 eV to over N 10-r for a
kinetic energy of 4.3 eV. Additional experimental
observations I:151 suggest that the translational
energy of N, serves to overcome a potential barrier to a precursor state (the weakly-bound a-N,
state) rather than promoting direct dissociation.
What is not revealed, though, is how S, depends
on kinetic energy for energies significantly below
0.09 eV, i.e., for thermal energies corresponding
to T< 1000 K. Analysis of thermal desorption
results [14] for Fe(lll)
suggests that at normal
thermal energies, there is no activation barrier
for adsorption into the o-N, precursor state. It is
significant to note, however, that both thermal
desorption [14] and molecular beam [15] studies
reveal that S, decreases with increasing surface
temperature for a constant kinetic energy of the
gas. In both cases, the phenomenological activation barrier is on the order of -0.5 kcal/moI,
suggesting that the barrier for dissociative
chemisorption from the precursor state is slightly
lower than the barrier for desorption from this
state. Thus, it appears that for temperatures (gas
and solid) below 1000 K thermal dissociation and
molecular beam studies may give similar results.
E. Shustorovich,A. T. Bell / BOC-MP analysisof synthesb and decompositionof NH, on transitionmetal surfaces
(i)
3. Concluding remarks
We have applied the BOC-MP theory to calculate the energetics of ammonia synthesis and
decomposition on transition metal surfaces along
the periodic series Pt(lll), Ru(OOl), Fe(llO), and
Re(001). The calculated values of Q and AE *
are in remarkable agreement with experiment. As
far as the general patterns of reactivity are concerned, we find that dissociation activation barriers decrease but recombination and desorption
barriers increase in the series Pt, Ru, Fe, Re.
More specifically, we project that the rate-determining step of ammonia decomposition on all
the surfaces is recombinative desorption of N,,
and that Pt surfaces are the most effective. For
dissociation of N,, we project that the activation
barrier dramatically increases in the order Re <
Fe K Pt, making Pt catalysts inefficient for ammonia synthesis. These and other projections of
our model are in agreement with experimental
studies of the mechanism of NH, synthesis and
decomposition.
Weak bonding (closed-shell molecules such as
(A.2)
(ii) Strong bonding (molecular
NH and NH,)
radicals such as
Qi
(A.3)
QAB=
QA+DAB'
II Activation
barriers
fur dkociation
+4
(8 From the gas phase
A E:B,g = + DAB +
QAQB
Q,+Q,
(A4
(ii) From a chemisorbed
A E&,,
Acknowledgment
state
QAQB
= $ DAB iQA+QB
i
(A.3
+QM-QA-QB
Alexis
T. Bell acknowledges support of this
work by the Division of Advanced Industrial Concepts, US Department of Energy, under Contract
DE-ACO3-76SFOOO98.
AB,,, +A,
ZZZ.Actiuation
barriers for recombination A, + B,
-)AqM
General the~od~amic
relationships
AEz-a,s = AE,*_,,g =Q~+Q,-~B+~&,
if AE&,g > 0.
Appendix
(A4
AEz_B,g = AEZ-,,s - AE&,$ = QA + QB - DAB
Here we will give the formulas used to calculate the values of Q and AE*. (For a derivation
of these relations, the reader is referred to ref.
if AE,*,, < 0.
(A.71
M.)
References
I. Heat of adsorption
For an adatom A in an n-fold site, M,-A,
Q/x=&0,42-
l/n),
(A4
where QoA is the two-center M-A bond energy.
For an admolecule AB with the bond energy DAB
in the M,-AB site with the A end down, there
are two cases:
[l] E. Shustoro~ch, Adv. Catal. 37 (1990) 101.
121 (a) E. Shustorovich, Catal. L&t. 7 (1990) 107;
(b) E. Shustorovich and A.T. Bell, Surf. Sci. 248 (1991)
359.
[3] E. Shustorovich and A.T. Bell, Surf. Sci. 253 (1991) 386.
[4] A.T. Bell and E. Shustorovich, J. Catal. 121 (1990) 1.
[S] (a) J.J. Vajo, W. Tsai and W.H. Weinberg, J. Phys.
Chem. 89 (198.5) 3243;
E. ~~~t#ro~i~h, A.T. Bell f XC-MP
~n~~ys~ of ~nthes~ and decomposition of NH3 om tra~~t~n me&alsurfaces
(b) W. Tsai, J.J. Vajo and W.H. Weinberg, J. Phys.
Chern. 89 (1985) 4926.
161 W. Tsai and W.H. Weinberg, J. Phys. Chem. 91 (1987)
5302.
f71 (a) G. Erti, Catal. Rev. Sci. Eng. 21 (1980) 201;
(b) G. Ertl and M. Huber, J. Catal. 61 (1980) 537.
@I G. Haase and M. Asscher, Surf. Sci. 191 (1987) 75.
[91 Z. Rosenzweig and M. Asscher, Surf. Sci. 225 (1990) 249.
[lOI H.A.C.M. Hendrickx, A. Hoek and B.E. Nieuwenhuys,
Surf. Sci. 135 (1983) 81.
[II] M. Golze, M. Grunze and W. Unertl, Progr. Surf. Sci. 22
(1986) 101, table 1, p. 177.
[12] A. Ozaki and K. Aika, in: Catalysis-Science and Technology, Eds. J.R. Anderson and M. Boudart, Vol. 1 (Springer,
New York, 1981) p. 87.
[ 131 N.D. Spencer and G.A. Somorjai, J. Catal. 78 (1982) 142.
[14] G. Ertl, S.B. Lee and M. Weiss, Surf. Sci. 114 (1982) 515.
[15] C.T. Rettner and H. Stein, Phys. Rev. L&t. 59 (1987)
2768.