When is a nanoparticle a cluster? An operando EXAFS study of amine borane dehydrocoupling by Rh4-6 clusters

Article abstract of DOI:10.1021/ja073331l

X-ray absorption fine structure (XAFS) is used to determine the structure of the rhodium cluster present during the catalyzed dehydrocoupling of amine boranes under operando conditions. We show how a variety of XAFS strategies can be used in combination with other analytical methods to differentiate homogeneous from heterogeneous systems. Analysis of the in situ XAFS spectra using a series of amine boranes (NH₃BH₃, R ₂NHBH₃, and RNH₂BH₃ where R = methyl, isopropyl, tert-butyl, and cyclohexyl) and rhodium catalyst precursor compounds (including chloro-(1,5-cyclooctadiene)rhodium (I) dimer, bis(1,5-cyclooctadiene)rhodium (I) trifluoromethanesulfonate, chlorodicarbonylrhodium (I) dimer, dichloro(pentamethylcylcopentadienyl)rhodium (III) dimer, hexarhodium hexadecacarbonyl, and tetrarhodium dodecacarbonyl) strongly suggest that the active catalyst species for this reaction is a homogeneous rhodium complex. Rhodium clusters containing four or six rhodium atoms (Rh_4-6) bound to amine boranes are observed as the major (>99%) rhodium containing species during and after the catalyzed anaerobic dehydrocoupling. During the later stages of the reaction a nonmetallic rhodium complex precipitates in which individual Rh_4-6 clusters likely form polymer chains ligated by the reaction products that have two or more ligating sites. The best fits of the XAFS data, using ab initio calculations of FEFF theory, show that the major rhodium species (80%) has each rhodium atom directly bound to three rhodium atoms with an observed bond distance of 2.73 A and to two boron atoms at 2.10 A. A minor (20%) rhodium species has each rhodium atom bound to four rhodium atoms with a bond distance of about 2.73 A and a single rhodium atom at a nonbonding distance of 3.88 A. No metallic rhodium was observed at any time during the anaerobic reaction.

Full text of DOI:10.1021/ja073331l

Published on Web 09/07/2007
When is a Nanoparticle a Cluster? An Operando EXAFS Study
of Amine Borane Dehydrocoupling by Rh4-6 Clusters
John L. Fulton,*,† John C. Linehan,*,† Tom Autrey,† Mahalingam Balasubramanian,‡
Yongsheng Chen,† and Nathaniel K. Szymczak§
Contribution from the Fundamental Science Directorate, Pacific Northwest National Laboratory,
Richland, Washington, 99352, AdVanced Photon Source, Argonne National Laboratory, Argonne,
Illinois 60439, and Chemistry Department, UniVersity of Oregon, Eugene, Oregon 97403
Received May 10, 2007; E-mail: john.linehan@pnl.gov; john.fulton@pnl.gov
Abstract: X-ray absorption fine structure (XAFS) is used to determine the structure of the rhodium cluster
present during the catalyzed dehydrocoupling of amine boranes under operando conditions. We show
how a variety of XAFS strategies can be used in combination with other analytical methods to differentiate
homogeneous from heterogeneous systems. Analysis of the in situ XAFS spectra using a series of amine
boranes (NH3BH3, R2NHBH3, and RNH2BH3 where R ) methyl, isopropyl, tert-butyl, and cyclohexyl) and
rhodium catalyst precursor compounds (including chloro-(1,5-cyclooctadiene)rhodium (I) dimer, bis(1,5-
cyclooctadiene)rhodium (I) trifluoromethanesulfonate, chlorodicarbonylrhodium (I) dimer, dichloro(penta-
methylcylcopentadienyl)rhodium (III) dimer, hexarhodium hexadecacarbonyl, and tetrarhodium dodecac-
arbonyl) strongly suggest that the active catalyst species for this reaction is a homogeneous rhodium
complex. Rhodium clusters containing four or six rhodium atoms (Rh4-6) bound to amine boranes are
observed as the major (>99%) rhodium containing species during and after the catalyzed anaerobic
dehydrocoupling. During the later stages of the reaction a nonmetallic rhodium complex precipitates in
which individual Rh4-6 clusters likely form polymer chains ligated by the reaction products that have two or
more ligating sites. The best fits of the XAFS data, using ab initio calculations of FEFF theory, show that
the major rhodium species (80%) has each rhodium atom directly bound to three rhodium atoms with an
observed bond distance of 2.73 Å and to two boron atoms at 2.10 Å. A minor (20%) rhodium species has
each rhodium atom bound to four rhodium atoms with a bond distance of about 2.73 Å and a single rhodium
atom at a nonbonding distance of 3.88 Å. No metallic rhodium was observed at any time during the anaerobic
reaction.
1. Introduction
amine boranes under ambient conditions to release H2 with
formation of borazines, and boron-nitrogen oligomers. In
addition, we have shown that there is an enhanced release of
H2 during the thermal dehydrocoupling of NH3BH3 at the
surface of mesoporous silica.11 This rhodium-catalyzed amine
borane chemistry has been the subject of several comprehensive
studies by Manners et al.12,13 Our initial in situ work14 in this
area suggested that small Rhn clusters (n ) 4-6) are the
catalytically active species for the dehydrocoupling of (CH3)2-
NHBH3 (DMAB) as shown in eq 1 although ex situ character-
ization suggested a much larger colloidal metallic rhodium
species.13 The goal of this work is to develop a detailed
understanding of the role of the rhodium-containing catalyst in
the dehydrocoupling of amine borane compounds.
Hydrogen-rich amine borane compounds such as NH3BH3
(AB) are considered potential materials for hydrogen storage
for fuel cell applications which has stimulated recent interest
in understanding their chemistry.1-10 The bonding of hydrogen
to the light atoms, nitrogen and boron, can meet the required
energy density goals for hydrogen storage. Rhodium-based
catalysts are known to efficiently catalyze dehydrocoupling of
† Pacific Northwest National Laboratory.
‡ Argonne National Laboratory.
§ University of Oregon.
(1) Baitalow, F.; Wolf, G.; Grolier, J. P. E.; Dan, F.; Randzio, S. L.
Thermochim. Acta 2006, 445, 121.
(2) Baumann, J.; Baitalow, E.; Wolf, G. Thermochim. Acta 2005, 430, 9.
(3) Bluhm, M. E.; Bradley, M. G.; Butterick, R.; Kusari, U.; Sneddon, L. G.
J. Am. Chem. Soc. 2006, 128, 7748.
It has become clear in recent years that the ex situ charac-
(4) Dixon, D. A.; Gutowski, M. J. Phys. Chem. A 2005, 109, 5129.
(5) Grant, D. J.; Dixon, D. A. J. Phys. Chem. A 2005, 109, 10138.
(6) Keaton, R. J.; Blacquiere, J. M.; Baker, R. T. J. Am. Chem. Soc. 2007,
129, 1844.
terization of colloidal or nanoparticle catalyst species can be
(11) Gutowska, A.; Li, L. Y.; Shin, Y. S.; Wang, C. M. M.; Li, X. H. S.; Linehan,
J. C.; Smith, R. S.; Kay, B. D.; Schmid, B.; Shaw, W.; Gutowski, M.;
Autrey, T. Angew. Chem., Int. Ed. 2005, 44, 3578.
(12) Jaska, C. A.; Temple, K.; Lough, A. J.; Manners, I. J. Am. Chem. Soc.
2003, 125, 9424.
(13) Jaska, C. A.; Manners, I. J. Am. Chem. Soc. 2004, 126, 9776.
(14) Chen, Y. S.; Fulton, J. L.; Linehan, J. C.; Autrey, T. J. Am. Chem. Soc.
2005, 127, 3254.
(7) Matus, M. H.; Nguyen, M. T.; Dixon, D. A. J. Phys. Chem. A 2007, 111,
1726.
(8) Nguyen, M. T.; Nguyen, V. S.; Matus, M. H.; Gopakumar, G.; Dixon, D.
A. J. Phys. Chem. A 2007, 111, 679.
(9) Stephens, F. H.; Baker, R. T.; Matus, M. H.; Grant, D. J.; Dixon, D. A.
Angew. Chem., Int. Ed. 2007, 46, 746.
(10) Yoon, C. W.; Sneddon, L. G. J. Am. Chem. Soc. 2006, 128, 13992.
9
11936
J. AM. CHEM. SOC. 2007, 129, 11936-11949
10.1021/ja073331l CCC: $37.00 © 2007 American Chemical Society

When is a Nanoparticle a Cluster?
A R T I C L E S
regeneration of the active catalyst for a subsequent catalytic
cycle. This concept of a recyclable homogeneous catalyst is
similar to that reported for a tungsten hydrosilylation catalyst
that forms an insoluble chlathrate species at the end of the
reaction.17
The ability of XAFS to derive the molecular structure of metal
clusters randomly dispersed in a solvent provides an alternative
to conventional single-crystal X-ray diffraction (XRD) when
the cluster cannot be recovered in its crystalline form. The
Rh4-6 clusters in this study have eight or more ligand sites,
and the reaction products can form bridges between rhodium
clusters thus making it difficult or impossible to generate a
crystalline form for XRD analysis. The disadvantage of XAFS
is that the ability to resolve the atom positions and atom types
is not as conclusive as full crystallographic techniques such as
XRD. Specifically, first shell ligands are often well-resolved,
but there is decreasing resolution for the second and higher
shells. However, in the present case, the rhodium-rhodium
backscattering functions will be significantly different than the
backscattering functions for boron, nitrogen, and carbon, and
hence rhodium nearest neighbors can be detected with high
sensitivity. This means that accurate determination of rhodium
cluster sizes is possible with XAFS. On the other hand the
backscattering functions for boron and nitrogen are quite similar,
and thus these atom types can only be discriminated for certain
highly optimum structures. Within these limitations the full
structure may not be resolved without a priori information of
the nature of the ligands about the metal atom. This is a situation
where the combination of the two operando methods, XAFS
and NMR, provides a powerful method to identify active catalyst
structures. Another tool in understanding the ligand structure
about the rhodium core is to conduct XAFS studies of a series
of related reactant and catalyst precursor compounds thereby
eliminating or allowing the possibilities of certain ligands. It is
this combination of methodologies that we employ here to
understand the nature of the active catalyst for dehydrocoupling
reactions of a series of amine borane compounds.
In this report, we use operando XAFS methods to show that
the active catalytic species is most likely a Rh4-6 cluster
(diameter of approximately 0.3 nm) rather than a 2 nm diameter
rhodium nanoparticle.12,13 We then use a combination of the
other methodologies described above to derive the most probable
ligand structure about the rhodium cluster. The reported structure
of the predominate rhodium species provides key insights into
the likely catalytic species and the mechanism of the formation
of hydrogen from amine borane compounds. Further we use
XAFS to determine the kinetics of formation for the new
rhodium clusters (presumably related to the catalyst species)
and show that the material that precipitates at the end of the
reaction contains the same rhodium cluster with only slightly
different amine borane ligands. We anticipate this class of Rh4-6
clusters may play an important role in both dehydrocoupling
and hydrogenation reactions.
compromised by the measurement technique. For instance it
has recently been shown13,15 that, under commonly used TEM
conditions, the electron beam can induce the formation of
rhodium nanoparticles from rhodium organometallic complexes
leading to ambiguous conclusions about the morphology of the
active species. Adding to this challenging analysis problem is
that the catalyst structure at the end of a reaction can be different
than the structure during the reaction. For instance, the starting
reactants or reaction intermediates may bind as ligands to the
rhodium cluster, helping to stabilize the soluble form. As these
initial reactants/ligands are consumed through the course of the
reaction, their ability to stabilize the rhodium cluster in solution
can be attenuated. It is therefore advantageous to use analytical
spectroscopic methods that examine the form of the catalyst in
the “operando” state, wherein the catalyst not only is in the
reaction solvent, at the reaction pressure and temperature (i.e.,
in situ), but also is undergoing chemical reactions with the reac-
tants in the system. Operando methods using NMR or FTIR
are especially useful for following the reaction pathways and
kinetics of the reactants, the products and with some limited
capability for catalyst characterization. On the other hand, X-ray
absorption fine structure (XAFS), employed as an operando
method, can selectively probe the predominate metal species
as the XAFS signal is primarily dependent on the backscattering
from the first two or three atomic shells about the X-ray absorb-
ing metal center. XAFS is also very useful for characterizing
clusters in which the highly fluxional nature of the ligands and/
or the ligand exchange rates are so high as to make identification
by NMR difficult or impossible since the time frame of the
complete XAFS scattering process (a few femtoseconds) is much
shorter than the time for any changes in the atom positions.
In our previous communication14 we proposed that, near the
end of the reaction shown in eq 1, the rhodium cluster binds
ligands from the product of the reaction, forming a precipitate
comprising large numbers of individual rhodium clusters to yield
an insoluble resting state of the rhodium catalyst species. In
this work we present further evidence supporting this hypothesis.
A related observation has been reported by Toshima16 for small
rhodium and platinum/rhodium clusters in a polymer-stabilized
colloidal dispersion. Their XAFS measurement showed a cluster
size of 13 rhodium atoms that was inconsistent with a TEM
analysis of 3.3 nm (∼500 atoms). They proposed a model of a
rhodium “microcluster” embedded in a much larger polymer
matrix. In this work we present a related rhodium structure in
which a Rh4-6 cluster forms its own polymer-like network from
products that are generated during a catalytic reaction. The solid
precipitate that forms at the end of the reaction can be
regenerated to the soluble active Rh4-6 cluster by simply adding
more of the amine borane reactant. This highlights an interesting
feature of this reaction, the ability to recover the active rhodium
catalyst with high efficiencies by self-precipitation and then
2. Experimental Section
2.1. Chemicals and Procedures. Dimethylamine borane (Callery),
tert-butylamine borane (tBAB, Callery), ammonia borane (Aviabor),
bis(1,5-cyclooctadiene)rhodium (I) trifluoromethanesulfonate (Strem),
chloro-(1,5-cyclooctadiene)rhodium (I) dimer (Strem), chlorodicarbo-
(15) Hagen, C. M.; Vieille-Petit, L.; Laurenczy, G.; Suss-Fink, G.; Finke, R.
G. Organometallics 2005, 24, 1819.
(16) Harada, M.; Asakura, K.; Toshima, N. J. Phys. Chem. 1994, 98, 2653.
(17) Dioumaev, V. K.; Bullock, R. M. Nature 2003, 424, 530.
9
J. AM. CHEM. SOC. VOL. 129, NO. 39, 2007 11937

A R T I C L E S
Fulton et al.
nylrhodium (I) dimer (Strem), dichloro(pentamethylcylcopentadi-
enyl)rhodium (III) dimer (Strem), hexarhodium hexadecacarbonyl
(Strem), and tetrarhodium dodecacarbonyl (Strem) were used as
received. Di-isopropylamine borane (DiPAB), cyclohexylamine borane
(CAB), and dicyclohexylamine borane (DCAB) were synthesized by
addition of the appropriate amine to an equimolar amount of a THF
solution of BH3-THF complex.18 The purity of all amine boranes used
in this study was greater than 99% as determined by 1H and 11B NMR.
Toluene and THF (Aldrich) were distilled from sodium and degassed
prior to use. All reactions were carried out under an inert atmosphere
unless otherwise stated.
The concentration of the precatalyst was varied between 3.5 and 35
mM, and that of the substrate, between 0.02 and 2 M. In a typical
procedure a solution of 20 mg of chloro-(1,5-cyclooctadiene)rhodium
(I) dimer ([Rh(1,5-COD)Cl]2) dissolved in 2 mL of toluene was added
to a solution containing 200 mg of DMAB in 2 mL of toluene under
an atmosphere of 4% H2 in He. The mixed solution inside of a 1.5 cm
diameter × 6 cm long sealed glass vial was taken from the glovebox
and placed in the X-ray beam. A new vial was used for each reaction.
Each vial was placed in a horizontal position to provide a 6 cm path
length for the X-ray beam. The glass sidewall of each vial was pierced
with a 1 mm diameter hole to allow for escape of the H2. To prevent
the ingress of air into the reaction vial the side hole was covered with
a 1 cm2 piece of Kapton film. The first XAFS spectrum was collected
within 3 min of mixing.
amplitude-reduction factor and is usually treated empirically. The fitted
parameters include Ni, the coordination number of the shell for each
type of neighboring atom, Ri, the shell distance, and σ2, the Debye-
i
Waller factor which represents the mean-square variation in Ri due to
both static and thermal disorder. The fitting of the FEFF8 theoretical
standards to the experimental data was accomplished using an analysis
program (FEFFIT).23
In addition to the structural parameters, a single
nonstructural parameter, ∆E0, is varied to correct for the simple estimate
of E0 made by FEFF8.
Various rhodium-containing standard compounds were used to
establish the value of the core-hole factor, S20, and to also better
understand the sensitivity of the XAFS method to the detection of longer
range rhodium atoms. In evaluating these compounds the ø(k) data were
fit to the FEFF8 theoretical standard using the atom positions that were
generated from the published crystallographic data for dichloro-
(pentamethylcylcopentadienyl)rhodium (III) dimer,24 bis(1,5-cyclooc-
tadiene)rhodium (I) trifluoromethanesulfonate, chlorodicarbonylrhodium
(I) dimer,25 Rh4(CO)12,26 Rh6(CO)16,27 chloro(1,5-cyclooctadiene)-
rhodium (I) dimer,28 and for rhodium metal.29 (Schematics of these
structures are given in the insets of Figures 9 and 10.) The complete
fits to these standards are provided in the Supporting Information. When
evaluating the new rhodium cluster, we used a core-hole factor of S02
) 0.89, i.e., the average from the five different solid standards. S20 has
an associated uncertainty of about 15%, and according to eq 2, we see
that this results in an approximate 15% uncertainty in the reported
coordination number.
2.2. XAFS Methods. The rhodium K-edge (23222 eV) XAFS spectra
were collected in transmission mode on the bending magnet beamline
(XOR/PNC, Sector 20) at the Advanced Photon Source, Argonne
National Laboratory. The bending magnet beamline was chosen over
the much higher flux insertion device line to minimize the potential
for beam damage to the rhodium complexes. No evidence of beam
damage was observed during exposure of the rhodium complexes to
the X-rays. Energy calibration was accomplished using the edge energy
of a Rh(0) foil (23222 eV). As a further measure, this rhodium foil
was placed far beyond the sample with an additional transmission
detector (I2) providing internal energy calibration for each sample
spectrum. We collected time series of X-ray absorption near edge
(XANES) spectra while the active catalyst was forming to study their
kinetics. Full extended X-ray absorption fine structure (EXAFS) spectra
were collected when the kinetics were slow or the conversion was
complete in order to capture the most molecular structural information
on the rhodium catalyst. Series of XANES spectra were normalized to
a common edge height.
The EXAFS ø(k) data were weighted by k2 and windowed in the
range 2.0 < k < 17.0 Å-1 using a Hanning window with dk ) 1.0
Å-1. The fits were to both the real and imaginary parts of ø˜(R) in the
region 0.8 < R < 5.0 Å. The quality of fits was evaluated using the
criteria defined in FEFFIT, the automated software for fitting the
structural parameters.
The structure of several Rh4 clusters have been characterized by
single-crystal XRD.26,30 One cluster of particular relevance for this study
is Rh4(CO)12 since the Rh coordination number is very close to that
measured for the rhodium catalyst in this study. Rh4(CO)12 has C3
symmetry with three equivalent rhodium atoms each binding two
terminal CO’s and sharing two bridging CO’s. The fourth apical
rhodium binds to only three terminal CO’s. The rhodium-rhodium
bond distances are only slightly different with rhodium bonds to the
apical rhodium of about 2.675 Å and slightly longer bonds between
the three equivalent rhodium atoms of 2.725 Å.26 The structure is highly
fluxional in solution, which makes it difficult to characterize by NMR.31
In the FEFF8 calculations used in this study, Rh4(CO)12 was used as
the base structure and modified by substituting B and N atoms at the
same terminal and bridging sites modified by realistic Rh-B and Rh-N
distances. Even with this more accurate representation of the rhodium
cluster much simpler FEFF8 structures generated from a single rhodium
atom coordinated with various B, N, or Rh ligands gave nearly identical
fit results showing the invariance of the fitted results with respect to
the starting model.
We used standard methods for the analysis of EXAFS data19 using por-
tions of the UWXAFS program.20 The EXAFS relationship is given by
Fi(k) S20Ni
2
2
i
ø(k) )
e-2k σi e-2R /λ(k) sin(2kRi + δi(k))
(2)
∑
kR2i
i
The EXAFS oscillations, ø(k), were extracted from the experimen-
tally measured absorption coefficient using an automated background
subtraction method (AUTOBK) developed by Newville et al.21 The
In several instances principal component analysis was used to fully
evaluate the reaction rate and to determine the number of chemical
states of the Rh species. In some systems, where only a limited number
wavenumber of the ejected photoelectron is given by
k )
2m (E - E )/p2 with E0 being the absorption edge energy. In eq 2,
x
e
0
(22) Zabinsky, S. I.; Rehr, J. J.; Ankudinov, A.; Albers, R. C.; Eller, M. J. Phys.
ReV. B 1995, 52, 2995.
(23) Newville, M.; Ravel, R.; Haskel, D.; Rehr, J. J.; Stern, E. A.; Yacoby, Y.
Physica B 1995, 208/209, 154.
(24) Chuchill, M. R.; Julis, S. A.; Rotella, F. J. Inorg. Chem. 1977, 16, 1137.
(25) Walz, L.; Scheer, P. Acta Crystallogr., Sect. C 1991, 47, 640.
(26) Farrugia, L. J. J. Cluster Sci. 2000, 11, 39.
(27) Farrar, D. H.; Grachova, E. V.; Lough, A.; Patirana, C.; Poe, A. J.; Tunik,
S. P. J. Chem. Soc., Dalton Trans. 2001, 2015.
(28) Deridder, D. J. A.; Imhoff, P. Acta Crystallogr., Sect. C 1994, 50, 1569.
(29) Singh, H. P. Acta Crystallogr. 1968, A24, 469.
(30) Ricci, J. S.; Koetzle, T. F.; Goodfellow, R. J.; Espinet, P.; Maitlis, P. M.
Inorg. Chem. 1984, 23, 1828.
(31) Cotton, F. A.; Kruczyns, L.; Shapiro, B. L. J. Am. Chem. Soc. 1972, 94,
6191.
Fi(k), δi(k), and λ(k) are the amplitude, phase, and mean-free-path factor,
respectively, that are derived from theoretical standards calculated by
FEFF8.22 The S20 term in the above equation is the core-hole or
(18) Brown, H. C.; Zaidlewicz, M.; Dalvi, P. V.; Narasimhan, S.; Mukho-
padhyay, A. Organometallics 1999, 18, 1305.
(19) Koningsberger, D. C.; Prins, R. X-ray Absorption: Principles, Applications,
Techniques of EXAFS, SEXAFS and XANES; John Wiley & Sons: New
York, 1988.
(20) Stern, E. A.; Newville, M.; Ravel, B.; Yacoby, Y.; Haskel, D. Physica B
1995, 208/209, 117.
(21) Newville, M.; Livins, P.; Yacoby, Y.; Rehr, J. J.; Stern, E. A. Phys. ReV.
B 1993, 47, 14126.
9
11938 J. AM. CHEM. SOC. VOL. 129, NO. 39, 2007

When is a Nanoparticle a Cluster?
A R T I C L E S
Table 1. List of Reactions between [Rh(1,5-COD)Cl]2 and Different Substrates in the Kinetic Study of the Active Catalyst Formation
cat.
precip a
t
0.05
reaction
no.
T
[Rh]
[substrate]
(mM)
t 1
/
2
precatalyst
substrate
solvent
(
°
C)
(mM)
[substrate]/[Rh]
(min)
(min)
[Rh(1,5-COD)Cl]2
DMAB
toluene
1
2
3
4
5
6
7
8
25
3.7
12
12
23
23
23
23
26
120
990
450
1900
970
490
240
850
32
85
37
84
42
21
10
33
7
2
17
8
5
10
11
15 (8)
-
-
36
80
45
19
20
24
9
10
40
25
23
23
970
240
42
10
3.5
6
14
9
toluene
THF
11
12
13
34
25
23
32
66
980
0.9
2.6
42
not converted
16 (13)
6
s
18
50
[Rh(1,5-COD)Cl]2
[Rh(1,5-COD)Cl]2
AB
THF
14
15
25
25
24
26
920
420
39
16
1.3
2
2
>6
CAB
toluene
16
17
18
19
20
21
23
23
23
23
21
24
20
41
460
460
450
770
0.9
1.7
20
20
22
n/a
3.5
<2
<2
<2
60
s
s
s
s
s
s
DCAB
tBAB
DiPAB
31
a Time 5% Rh precipitates out of solution; s, soluble.
of spectra are available, the full principal components analysis (PCA)
was not possible. In this case we used a simple empirical method that
gives almost identical results in order to determine the reaction rate.
The method entails finding the height difference between the maxima
at 23 245 eV and the minima at approximately 23 270 eV.
In many cases the complex starts to precipitate in the later stages of
the reaction. The XAFS edge height is proportional to the total amount
of soluble rhodium in the X-ray beam (for a constant beam path length).
Hence, the edge height provides an accurate measurement of the
concentration of the soluble rhodium species that includes both the
unreacted precatalyst and the soluble cluster. When rhodium-containing
species precipitate, the edge height decreases proportionally due to the
loss of the rhodium species.
3. Results
The formation of the catalyst must precede the amine borane
dehydrocoupling reaction described in eq 1. The conversion of
the rhodium-containing precursor compound, ([Rh(1,5-COD)-
Cl]2), to the rhodium cluster can be described by the sequence
in eq 3.
We have also hypothesized14 that, in the later stages of the
reaction, a rhodium-containing material precipitates and that
rhodium in the precipitate consists of small rhodium clusters
that form a polymer-like product bonded by amine borane dimer
in a reaction described in eq 4.
The goal of this work is to ascertain the details of these
formation kinetics and then the structure of the important catalyst
species and/or catalyst resting states. Using a variety of different
XAFS methods we determine many different aspects of this
reaction. First we use XAFS to follow the kinetics of formation
of the predominant rhodium species. Evaluation of time-resolved
XAFS spectra (1 min/spectrum) provides information about the
rate of cluster formation and also about the number of rhodium
species that are present during this reaction. We then use the
XAFS spectra to determine the solubility of the cluster during
the reaction from the measurement of the X-ray absorption edge
height. We show that the solubility of the complex greatly
diminishes in the later stages of the reaction for select amine
boranes but not for others. Finally we show how the same cluster
forms from (i) different rhodium precursor compounds and from
(ii) different amine borane compounds. This leads to the final
determination of the cluster structure by fitting the XAFS spectra
to theoretical standards.
3.1. Kinetics of the Rhodium Cluster Formation. 3.1.1.
Factors Affecting the Reactions Rates and Principal Com-
ponent Analysis of the Rhodium Cluster. A complete list of
the reaction conditions used in the study is presented in Table
1. We found that a range of different alkyl amine borane
compounds all convert the [Rh(1,5-COD)Cl]2 to a new species
with very similar structures. This series of alkyl amine borane
compounds include NH3BH3 (AB), Me2NHBH3 (DMAB), di-
iPrNHBH3 (DiPAB), tert-butylNH2BH3 (tBAB), CyNH2BH3
(CAB), and di-CyNHBH3 (DCAB). Also shown in Table 1 are
cat
0.5
precip
0.05
the t and t
values that represent the time to convert 50%
9
J. AM. CHEM. SOC. VOL. 129, NO. 39, 2007 11939

A R T I C L E S
Fulton et al.
Figure 2. Concentration of the rhodium clusters as a function of time for
DMAB in toluene at various concentrations as indicated. The complete
conditions are listed in Table 1 for reactions 4-7, 10 and 11.
Figure 1. Time series of XAFS spectra above the Rh K-edge XAFS from
3 to 32 min for the reaction of DMAB with [Rh(1,5-COD)Cl]2 in toluene
at 25 °C (Table 1, reaction 3).
Hence changes in the ligands in the later stages of the reaction
involving replacement of monomeric amine borane ligands with
dimer or trimer amine borane products would not necessarily
be detected.
of the precursor to the new species and the time at which 5%
of the rhodium complex have precipitated, respectively. The
different amine borane compounds convert [Rh(1,5-COD)Cl]2
to the new species at significantly different rates and to different
extents. When all of the amine borane starting material has been
consumed the rhodium complexes derived from AB and DMAB
precipitate, whereas the complexes in contact with DiPAB,
tBAB, CAB, and DCAB remain soluble. We also briefly
explored other factors that affect reaction rates such as
concentrations of the precatalyst and the amine borane substrate,
temperature, and solvent type. These conditions are indicated
in Table 1.
Figure 1 shows a time series of rhodium K-edge X-ray
absorption spectra of the reaction of DMAB with [Rh(1,5-
COD)Cl]2 in toluene at 25 °C (Table 1, reaction 3) spanning
the range from 23 230 to 23 380 eV that was typical of all the
amine borane reactions with this catalyst precursor. The
progression in the time series for the DMAB is very similar to
3.1.2. Rh Cluster Formation Kinetics for AB, DMAB,
CAB, DCAB, DiPAB. Figure 2 shows the rates of Rh cluster
formation for DMAB in toluene under different reaction
conditions described in Table 1 for reactions no. 4-7, 10, and
11. The tc0a.5t values for these reactions are also listed in Table 1.
In most cases the amount of the new Rh cluster increases linearly
with time. Most reactions had a significant excess of DMAB
(see the [substrate]/[Rh] ratios), which means that the DMAB
concentration does not change significantly over the course of
the catalyst forming reaction. Higher DMAB concentrations and
higher temperatures accelerate the formation of the Rh cluster.
The rate of formation of the Rh cluster for AB in THF is
almost 5 times faster than that for DMAB under the same
conditions (Figure 3 and reactions 13 and 14 in Table 1).
Likewise CAB has a Rh cluster formation rate that is over 5
times faster than that for DMAB with the same solvent and
temperature conditions (reactions 6 and 18). The rate data for
CAB and DiPAB (reactions 19 and 21) show a much different
behavior than those for the other systems. There appear to be
two different mechanisms involved: a rapid rate at times less
than about 10 min and then a very slow conversion over several
hours to the final state. This is consistent with the steric bulk
of these di-substituted amine boranes especially if two (or more)
of these are ligands on each rhodium atom in the cluster. The
first ligation will have less steric interaction compared to the
ligation of the second amine borane. The rate of cluster
formation increases as AB ≈ CAB ≈ tBAB> DMAB > DCAB
≈ DiPAB.
that obtained for the other amine borane compounds, although
cat
0.5
rates of reaction were different according to the t values
reported in Table 1. The primary feature of these spectra is the
presence of a unique set of isosbestic points indicated by arrows
in Figure 1. The presence of a unique set of isosbestic points32
reveals the presence of two chemical states of rhodium as
described by eq 3 in which the rhodium precursor is converted
to a single catalytic form33 (e.g., pure Rh4 clusters or to a fixed
ratio of mixed Rh4/Rh6 clusters as described later). We also
conducted a principal components analysis (PCA) on this time
series which showed that the spectra are a linear combination
of the precatalyst and the new rhodium cluster spectra. PCA
analysis also shows that no more than 2% of a third component
could exist in this reaction series. Any transient species are
present in negligible amounts. In addition, the XAFS analysis
is not sensitive to the third shell atoms about the rhodium core.
3.2. Solubility of the Rhodium Clusters. 3.2.1. Precipitation
during Reaction. In the later stages of the reaction the rhodium
clusters precipitate for AB and DMAB although to different
extents. Figures 4 and 5 show the percentage of soluble rhodium
with time under different reaction conditions for DMAB and
AB. The decrease in rhodium absorbance is the direct result of
the precipitation of the rhodium complex. To quantitatively
(32) Senior, W. A.; Verrall, R. E. J. Phys. Chem. 1969, 73, 4242.
(33) We realize that one cannot ever rule out very small amounts of unobservable
species being the actual catalyst. We also realize that what we are calling
the catalyst may in fact be a resting state of the actual catalyst. However
the structure of this resting state most likely has an impact on the actual
catalyst species.
9
11940 J. AM. CHEM. SOC. VOL. 129, NO. 39, 2007

When is a Nanoparticle a Cluster?
A R T I C L E S
Figure 3. Concentration of the rhodium clusters as a function of time for
DMAB, AB, and CAB at various concentrations as indicated. The complete
conditions are listed in Table 1 for reactions 11, 12, 14, 15, 16, and 17. All
reactions conducted in THF except for DMAB/Rh ) 0.9:1 that was
conducted in toluene.
Figure 5. Solubility of the rhodium clusters as a function of time for DMAB
and AB in THF at various concentrations as indicated. The complete
conditions are listed in Table 1 for reactions 12, 14, and 15. Concentration
of Rh is in mM.
group may be more inhibited toward polymer formation due to
steric effects. Alternatively the alkyl groups on the nitrogen may
increase the solubility of the rhodium catalyst complex over
the methyl or hydrogen found on DMAB and AB. These
arguments are consistent with the highest precipitation rate
observed for AB that ultimately forms a trimer (borazine) or a
polymer (polyborazalene) having three binding sites with
minimal steric hindrance.
3.2.2. Precipitation or Redispersion of Rhodium Clusters
by Adding DMAB Dimer or Monomer. The addition of
DMAB dimer to the catalytic system clearly hastens precipita-
tion of the rhodium complex, Figure 6a. In contrast the addition
of DMAB monomer to the precipitate at the very end of reaction
5 (Table 1), Figure 6b illustrates the redissolution of the
precipitate with the production of hydrogen gas and more
DMAB dimer. This latter result indicates that the precipitate
is not a metal or metal nanoparticle since it is unlikely that
simple replacement of the ligand could lead to breakup of the
metal-metal bonds. The observation of the precipitated rhodium
complex catalyzing the dehydrocoupling of amine boranes may
lead to the assumption of the operation of a heterogeneous
catalyst due to the dark opaque color of the solution.
Figure 4. Solubility of the rhodium clusters as a function of time for DMAB
in toluene at various concentrations as indicated. The complete conditions
are listed in Table 1 for reactions 4-7, 10, and 11.
precip
compare this effect, Table 1 lists the t
value that is the
3.3. XAFS Spectra of the Rhodium Clusters. 3.3.1.
Preliminary Examination of the Rhodium Coordination
Structure and the Effect of Solvent Type. The uncorrected,
k- and k2-weighted |ø˜(R)| plots for the predominant rhodium
species are shown in Figure 7a and 7b. The |ø˜(R)| functions
are generated by Fourier transformation of the XAFS spectrum
represented in the form of ø(k) as defined in eq 2. The |ø˜(R)|
plots represent the partial pair distribution (ppd) functions
convoluted with the photoelectron scattering functions shown
in eq 2. The plots approximately represent the probability of
finding a certain type of neighboring atom about the central
X-ray absorbing atom. The peak widths are artificially broadened
from the true ppd functions since the phase of the oscillations
for a certain scattering atom is dependent upon k which is not
the case for classical X-ray diffraction. Likewise the peak
positions in Figure 7 are before the phase correction and are
thus typically shorter by about 0.2-0.4 Å from the true position.
0.05
time at which 5% of the rhodium clusters precipitate. After the
initial rapid rate of precipitation, the soluble rhodium species
decreases until it reaches a saturation concentration. For higher
DMAB concentrations the saturation concentration of the Rh
cluster appears to be much higher, although this might reflect
the higher solubility of Rh clusters with appended DMAB
monomer vs dimer ligands. The system containing AB in THF
undergoes rapid catalyst precipitation at an early stage and
exhibits a much lower solubility of the active form.
The rhodium complex, generated from DiPAB, tBAB, CAB,
and DCAB, remains soluble throughout the reaction; hence the
nature of the ligand appended to the rhodium core plays an
important role in its solubility. The amine borane oligomeric
species formed have multiple sites of binding to the rhodium
cluster (see eq 4) that can lead to formation of polymeric
structures. The amine borane compounds with a bulky alkyl
9
J. AM. CHEM. SOC. VOL. 129, NO. 39, 2007 11941

A R T I C L E S
Fulton et al.
Figure 7. (a) k-weighted |ø˜(R)| plots for the rhodium clusters in both
toluene and tetrahydrofuran using the precursor, [Rh(1,5-COD)Cl]2. (b) k2-
weighted |ø˜(R)| (solid lines) and Im[ø˜(R)] (dashed lines) plots for the
rhodium clusters. A rhodium metal foil spectrum is also included that has
been scaled by 1/6 for comparison to the active catalyst spectra. In all cases
the k range for the Fourier transform is 2-17 Å-1. Distances are not
corrected for photoelectron phase shifts.
species that is identified by the intense new peak occurring at
2.4 Å (see Supporting Information for |ø˜(R)| of the precursor).
This peak can be uniquely assigned to rhodium-rhodium bonds
by examining the shape of the Im[ø˜(R)] portion of the transform
in Figure 7b between 2.2 and 2.8 Å. This “signature” of rhodium
scattering does not occur for boron or nitrogen atoms (see
Supporting Information). Also shown in Figure 7 is the spectrum
of the rhodium species produced in THF (Table 1, experiment
13) which is identical in every respect to that produced in
toluene. Thus the primary role of the solvent for this reaction
is to mediate the solubility of the reactants, products, and the
rhodium complexes.
Figure 6. (a) Precipitation of rhodium clusters upon addition of the cyclic
dimer product from the DMAB reaction. Reaction involving the precursor,
[Rh(1,5-COD)Cl]2, in toluene at 25 °C. (b) Dissolution of the spent catalyst
upon addition of DMAB in toluene at 25 °C.
However, the complex phase shift relationship for different
atoms is accurately predicted by FEFF so that the true peak
positions and widths are recovered later in the fitting process.
It is important to evaluate both the magnitude (|ø˜(R)|) and
imaginary (Im[ø˜(R)]) portions of the Fourier transforms since
both components are required to capture all the information
contained in the original XAFS spectra. It is also instructive to
evaluate both k- and k2-weighted |ø˜(R)| since the latter enhances
the peak heights of higher Z atoms such as rhodium. Examples
shown in the Supporting Information graphically illustrate the
ability of XAFS to clearly discriminate rhodium atoms in the
first shell from the second row elements such as B, C, or N
located at nearby positions.
Also shown in Figure 7 is the |ø˜(R)| data for metallic rhodium
in which the spectrum has been scaled by 1/6 to better compare
to the rhodium species produced with DMAB. The much greater
intensity of the peak at 2.4 Å is a result of 12 nearest rhodium
neighbors in the metallic form versus approximately 3 nearest
neighbors observed for the new rhodium species. (After phase
correction the rhodium-rhodium distance for metal is 2.68 Å,
and as shown in the Supporting Information, the distances for
the first through fourth rhodium shells are within 0.005 Å of
the XRD values.) The other significant features of the metallic
rhodium spectrum are the peaks at 3.6, 4.4, and 5.0 Å
corresponding to the second, third, and fourth rhodium shells,
respectively. In the Im[ø˜(R)] portion of the transform these
satellite peaks appear as an abrupt series of oscillations that are
unique to only rhodium-rhodium scattering paths. These
satellite peaks are completely absent in the rhodium species
Figure 7 provides the ø˜(R) plots for the predominant rhodium
species generated from [Rh(1,5-COD)Cl]2 with DMAB in
toluene at 25 °C corresponding to reaction 6 in Table 1. The
main features are the two peaks located near 1.6 and 2.4 Å,
corresponding to first-shell boron (most probable assignment,
described later) and rhodium atoms, respectively. [Rh(1,5-COD)-
Cl]2 rapidly reacts (tcat ) 10 min) to form the final rhodium
0.5
9
11942 J. AM. CHEM. SOC. VOL. 129, NO. 39, 2007

When is a Nanoparticle a Cluster?
A R T I C L E S
Figure 9. k2-weighted |ø˜(R)| and Im[ø˜(R)] plots for the rhodium clusters
from two different precursors, [Rh(1,5-COD)Cl]2 and from bis(1,5-
cyclooctadiene)rhodium (I) trifluoromethane sulfonate. The k range for the
Fourier transform is 2-15 Å-1. Distances are not corrected for photoelectron
phase shifts.
However since no significant differences in the second shell
structure are detected it appears that rhodium binds to boron.
Hydrogen cannot be detected using XAFS.
3.4. XAFS Spectra of the Rhodium Clusters from Various
Precursors. 3.4.1. Chloro-(1,5-cyclooctadiene)rhodium (I)
Dimer, Bis(1,5-cyclooctadiene)rhodium (I) Trifluoromethane-
sulfonate. Figure 9 shows the k2-weighted |ø˜(R)| plot of the
rhodium cluster species generated from two different starting
precursors. The first precursor has one cyclooctadiene and two
chloride ligands, while the latter compound has two cyclooc-
tadiene ligands and an outer shell triflate ion (CF3SO3-). The
XAFS spectra of these two precursor materials are significantly
different (see Supporting Information), and yet the rhodium
complexes produced from these two compounds are nearly
identical. The transformed spectra of the rhodium complex are
identical in all respects in both the |ø˜(R)| and Im[ø˜(R)] plots.
The 1,5-cyclo-octadiene ligands are displaced during the reaction
and hydrogenated under these conditions to produce cyclooctane
as determined by GC and NMR. The similarity of the rhodium
complexes produced from the two precursors conclusively shows
that Cl- no longer resides in the first coordination shell. As a
third row element, a Cl- ion provides a strong back scattering
signal that produces a distinct peak in the |ø˜(R)| plot at about
2 Å (e.g, see spectrum of [Rh(1,5-COD)Cl]2 in the Supporting
Information).
Figure 8. (a) k-weighted |ø˜(R)| plots for the rhodium clusters generated
from a series of alkyl amine borane compounds as shown starting from the
precursor, [Rh(1,5-COD)Cl]2. (b) k2-weighted |ø˜(R)| and (c) Im[ø˜(R)] plots
for the rhodium clusters. In all cases the k range for the Fourier transform
is 2-17 Å-1. Distances are not corrected for photoelectron phase shifts.
produced with amine boranes showing that the predominant
rhodium species does not contain higher rhodium shells.
3.3.2. Structure of the Rhodium Clusters Synthesized from
AB, DMAB, DiPAB, tBAB, CAB, and DCAB. The various
alkyl amine borane compounds all convert [Rh(1,5-COD)Cl]2
to nearly the same rhodium cluster with different conversion
3.4.2. XAFS Spectra of the Rhodium Complex from Other
Precursors. Four other precursor materials were studied includ-
ing three different carbonyl compounds and a pentamethyl
cyclopentadienyl complex. The spectrum of the rhodium clusters
from each of these different precursors is shown in Figure 10.
There is a remarkable transition for all six different precursors
in this study to a common form containing a rhodium cluster
core. Four of the precursor compounds shown in Figures 9 and
10 have no first shell rhodium-rhodium bonds resulting in very
diverse spectra (the Supporting Information compares the
precursor to its final form) due to the different ligand types and
coordination numbers. On the other hand, these four different
precursors undergo a transition to a structure with a nearly
cat
times according to the t values given in Table 1. Figure 8a
0.5
and 8b show the various ø˜(R) plots for this series of alkyl amine
boranes including AB and the primary and secondary amine
boranes. The first and second shell atoms of the amine borane
ligands that bond to rhodium are nearly identical for all the
alkyl amine boranes. This similarity suggests, but does not
prove, that the amine borane or oligomeric product bonds to
rhodium through the boron atom rather than through the amine
nitrogen. If the nitrogen was bound directly to the rhodium then
the differences in the number and type of alkyl groups would
contribute different intensities to the second-shell portion of the
ø˜(R) plots in the region from approximately 1.6 to 2.2 Å.
9
J. AM. CHEM. SOC. VOL. 129, NO. 39, 2007 11943

A R T I C L E S
Fulton et al.
Figure 10. (a) k-weighted |ø˜(R)| plots for the rhodium clusters derived from different precursor compounds including Rh6CO16 (T ) 40 °C in toluene, t )
20 h), Rh4(CO)12 (T ) 40 °C in THF, t ) 13 h), [Rh(CO)2Cl]2 (T ) 40 °C in toluene, t ) 13 h), a pentamethyl cyclopentadienyl complex (T ) 40 °C in
toluene, t ) 2 h), and [Rh(1,5-COD)Cl]2. (b) k2-weighted |ø˜(R)| and (c) Im[ø˜(R)] plots for the rhodium clusters from the various precursors. The k range for
the Fourier transform is 2-15 Å-1. Distances are not corrected for photoelectron phase shifts.
identical rhodium cluster but with some slight differences in
the ligand structure as shown in Figures 9 and 10. Rh6(CO)16
and Rh4(CO)12 already exist with the basic rhodium cluster core.
In these cases, the reaction with the amine borane is mostly a
ligand substitution reaction, displacing at least some of the
strongly binding CO ligands. As shown in Figure 10, these
carbonyl compounds have slightly different structures in the
range from 1 to 2 Å. Model fitting shows that the ligands are
likely a mixture of CO and amine borane compounds.
solution with the precatalyst solution. In the later stages of the
reaction, precipitation of a rhodium-containing species occurs
when the reactions are carried out with DMAB or AB (see Table
1). After several hours to days, depending on solvent and amine
borane-to-rhodium ratios, the supernatant is clear and the black
material settles to the bottom of the reaction vessel. For these
solutions there is no coating of the glass reactor with rhodium
complex, and these surfaces remain optically and spectroscopi-
cally clear. Hence all of the insoluble rhodium species remains
in the precipitated material. By conducting XAFS measurements
on this precipitate it is possible to evaluate the rhodium structure
3.4.3. XAFS Spectra of the Precipitate. The solution quickly
becomes opaque black shortly after mixing the amine borane
9
11944 J. AM. CHEM. SOC. VOL. 129, NO. 39, 2007

When is a Nanoparticle a Cluster?
A R T I C L E S
Figure 11. k2-weighted |ø˜(R)| plots for various forms of the rhodium
clusters derived from the precursor, [Rh(1,5-COD)Cl]2 with DMAB in
toluene at 25 °C. The inset shows the corresponding k2-weighted Im[ø˜(R)]
in the range from 3.3 to 6 Å. The k range for the Fourier transform is
2-17 Å-1. Distances are not corrected for photoelectron phase shifts.
in this material and also to explore the possibility of the presence
of a trace amount of a metal phase.
In Figure 11 we compare the k2-weighted |ø˜(R)| plots for the
rhodium complex produced with DMAB at t ) 0.8 h with the
precipitate that was anaerobically recovered after 72 h under
reaction conditions. The precipitate shows similar XAFS
characteristics to the soluble rhodium complex. For comparison,
the spectra of metallic rhodium is also shown and is scaled by
a factor of 1/6. The precipitate shows no evidence of the higher
rhodium-rhodium shells at 3.6, 4.4, and 5.0 Å. This is especially
clear in the expanded view of the Im[ø˜(R)] plots shown in the
inset of Figure 11. Thus, within the detection limits, we can
conclude that less than 1% of the rhodium precipitate could be
in the metallic form at this stage. This material has Virtually no
tendency, in a reducing environment (H2 saturated) or under a
nitrogen atmosphere, to convert to the metal over long time
periods. The ligands apparently stabilize the cluster and inhibit
reduction to the metal form.
Another dehydrocoupling reaction of DMAB was conducted
under aerobic conditions in a sample cell designed to readily allow
air exchange at the surface of the reacting mixture. The cell
was an open plastic trough of essentially the same dimensions
as the glass vials. In both the |ø˜(R)| and Im[ø˜(R)] plots in Figure
11 we see clear evidence that the precipitate from this reaction
contains metallic rhodium with higher rhodium-rhodium shells
that match the peaks at 3.6, 4.4, and 5.0 Å for bulk metal. From
the amplitudes of these peaks we can estimate that about 5%
of the total rhodium is in a metallic state. This counterintuitive
observation can be rationalized by the oxidation of the borane
ligands to borates (which is observed by 11B NMR). Presumably
some of the cluster reacts with O2 according to eq 5.
Figure 12. (a) k2-weighted |ø˜(R)| and Im[ø˜(R)] plots for the rhodium
clusters derived from the precursor, [Rh(1,5-COD)Cl]2 with DMAB in
toluene at 25 °C. The fitted lines are results from the FEFF8 standard and
parameters listed in Table 2. (b) Rh XAFS k2-weighted ø(k) plots for Rh
solution. The experimental data and the model fit are shown.
to air while a recently opened solution has virtually no odor.
Thus in the presence of air, the rhodium catalyst can be
converted to metal. We have also observed metallic rhodium
when an anaerobic precipitate was exposed to air. This leads to
an important point, that incorrect results could be obtained by
ex situ meausurements involving transporting material into the
analytical instruments, e.g., TEM or XPS, without rigorously
excluding O2.
3.5. Structural Parameters of the Rhodium Cluster from
XAFS Analysis. Quantitative information about the cluster
structure is obtained by fitting the experimental data to the
FEFF8 theoretical standards.20,23 Figure 12a and 12b show the
ø˜(R) and ø(k) plots of both the fits and the experimental data
for the rhodium clusters generated from DMAB in toluene at
25 °C. In all cases there is an excellent fit to the experimental
data. Table 2 provides the structural parameters that were
derived from these fits including the coordination numbers and
distance for atoms in the first and second shells about a central
rhodium atom. The rhodium-rhodium bond distances are 2.734
( 0.005 Å which are somewhat longer than the rhodium-
[Rh(1,5-COD)Cl]2 + DMAB (excess) f
[Rh(BH2NHMe2)2]4 + O2 f Rhmetal + nBxOy + 8NHMe2
(5)
Additional evidence for this reaction is the strong aroma of
dimethylamine from the reaction mixture upon long exposure
9
J. AM. CHEM. SOC. VOL. 129, NO. 39, 2007 11945

A R T I C L E S
Fulton et al.
Table 2. Results of the XAFS Analysis of the Rhodium Clusters at 25 °Ca,d
σ2SS
10-3 Å2)
b
solution
ID
bond
N
R (Å)
(×
E0
R
[Rh(1,5-COD)Cl]2
1
Rh-Rh
Rh-B
Rh-N
3.3 (0.3)
2.2 (0.4)
2.2c
2.734 (0.005)
2.095 (0.019)
2.67 (0.04)
8.2 (0.6)
6.3 (2.3)
16.1 (7.4)
-2.4
0.03
a For the fit, k2-weighting is applied, where σS2S, is the Debye-Waller factors for the single scattering path. Uncertainty is given in parentheses. In all
cases the k range used in the transform was from 2 to 17 Å-1 b Goodness of fit defined by a scaled sum of squares as described in FEFFIT.23 c Fixed to
.
the Rh-B value d Reported uncertainties for all parameters based upon the method described in FEFFIT.23
rhodium distance of 2.68 Å in bulk metal. The rhodium-
rhodium coordination number is 3.3 ( 0.5 which implies that
the average rhodium cluster contains approximately four
rhodium atoms (the coordination number uncertainty, (0.5, is
derived from the estimate of S20 and eq 2). Each rhodium atom
is ligated with approximately two boron atoms having a
rhodium-boron distance of 2.095 Å. More details of the fitting
procedure are given in the Discussion section.
measured independently from the main first shell peak at 2.4
Å. Alternatively the two clusters Rh4 and Rh6 are produced in
the same 4:1 ratio throughout the reaction.
4. Discussion
4.1. Structure of the Rhodium Core. 4.1.1. Rhodium
Cluster Coordination Number. For all rhodium precursors and
all the amine borane substrates examined in this study, the XAFS
results strongly suggest that the rhodium complex produced
consists of a small cluster composed of a few rhodium atoms
with a ligand shell. There are uniquely different features of the
XAFS spectra that place certain constraints on the maximum
size of this cluster. To demonstrate the sensitivity of XAFS to
determine the cluster size, it is necessary to evaluate the method
using known standards. Three of the rhodium precursor
compounds, [Rh(1,5-COD)Cl]2, [Rh(CO)2Cl]2, and [Cp*RhCl2]2,
are dimers that contain no rhodium in the first shell but that
have a single rhodium in the second shell. In each spectrum
there is an unambiguous backscattering signal from rhodium
in this second shell (see Figures S9, S10, and S11 in the
Supporting Information). For Rh6(CO)16, each rhodium atom
has four first-shell rhodium atoms and a single second-shell
atom. The XAFS spectrum of this compound also clearly shows
the existence of the single rhodium atom in the second shell
(see Figure S8). Rh(0) metal has a face centered cubic structure
in which there are 12 rhodium atoms in the first shell, 6 in the
second shell, 24 atoms in the third shell, and 12 atoms in the
fourth shell. As shown in Figure 7a, the ø˜(R) plots for the Rh-
(0) metal show that the peaks assigned to the third and fourth
shells at 4.4 and 5.0 Å are more intense than that of the second
shell at 3.6 Å due to the higher coordination numbers in these
shells. When the properties of these five standards are compared
to the new rhodium complex spectrum in Figure 7, several
conclusions are possible. The new complex shows no evidence
of backscattering from either the third or fourth rhodium shells
that would be indicative of metallic rhodium or rhodium
nanoparticles. On the other hand, in Figure 7 there is possibly
some scattering signal corresponding to the peak at 3.6 Å.
However the intensity of this peak is only about 20% of the
intensity for a Rh6 cluster such as Rh6(CO)16 that is shown in
Figure 10a. One possibility is that there is an equilibrium
between two relatively stable cluster sizes of Rh4 and Rh6 as
shown in eq 6. If we assume a simple equilibrium model of
80% Rh4 in equilibrium with 20% Rh6 based upon the intensity
of the peak at 3.6 Å, the calculated coordination number would
be 3.2 which is very close the coordination number of 3.3
4.1.2. Rhodium-Rhodium Bond Distance in the Rhodium
Cluster. As shown in Figure 13, small metal clusters, which
have no external ligands, undergo a bond contraction as cluster
size becomes smaller wherein the degree of bond contraction
follows an approximately n-1/3 relationship (n is the number of
atoms).34 This shrinking occurs as a result of minimizing the
surface free energy. At the two extremes, the metal-metal bond
distance for a metal dimer is typically about 0.2-0.3 Å shorter
than that for bulk metal for metals such as Pt, Cu, Au, and
Ir.34-37 In contrast, the Rh4-6 cluster in this study has a
rhodium-rhodium bond distance (2.73 Å) that is about 0.04 Å
longer than that for bulk metal (2.69 Å). This indicates that
the bonding with ligands leads to transfer/sharing of valence
electrons with the ligands that cause a weakening of the bonding
between the rhodium atoms, giving rise to a longer rhodium-
rhodium bond distance.
Other stable rhodium clusters have a longer bond distance
than that for bulk metal. As illustrated in Figure 13, Rh6(CO)16
has a rhodium-rhodium bond distance of 2.749 Å27 (or 2.754
Å from XAFS, Supporting Information) and Rh4(CO)12 has an
average bond distance of 2.71 Å.26 These ligands are capable
of stabilizing these clusters, and in so doing the rhodium-
rhodium bond distances are longer than that for bulk metal, an
indication that the ligands have overcome the high surface free
energies. Therefore the long rhodium-rhodium bond length for
the amine borane ligated Rh4-6 cluster strongly suggests that
the rhodium core is stabilized against agglomeration to metallic
particles or nanoparticles.
4.1.3. Are Rh Metal Nanoparticles Present during Reac-
tion? The evidence is quite strong that the vast majority of the
rhodium is converted to Rh4-6 clusters during the active part
of the catalytic dehydrocoupling reaction. There is also important
evidence that little, if any, nanoparticle or bulk metal phase is
present.
(34) Ankudinov, A. L.; Rehr, J. J.; Low, J. J.; Bare, S. R. J. Chem. Phys. 2002,
116, 1911.
(35) Montano, P. A.; Shenoy, G. K.; Alp, E. E.; Schulze, W.; Urban, J. Phys.
ReV. Lett. 1986, 56, 2076.
(36) Haberlen, O. D.; Chung, S. C.; Stener, M.; Rosch, N. J. Chem. Phys. 1997,
106, 5189.
(37) Ferrari, A. M.; Neyman, K. M.; Mayer, M.; Staufer, M.; Gates, B. C.;
Rosch, N. J. Phys. Chem. B 1999, 103, 5311.
9
11946 J. AM. CHEM. SOC. VOL. 129, NO. 39, 2007

When is a Nanoparticle a Cluster?
A R T I C L E S
Figure 13. Average rhodium-rhodium bond length for metal clusters having no ligands (n-1/3 relationship) and for various organometallic rhodium clusters
in this study.
The measured rhodium-rhodium coordination number is 3.3.
For low coordination number the XAFS uncertainties on the
measured cluster size are small. Frenkel38 reported a compre-
hensive study of platinum nanoparticles looking at the effect
of size and shape on the XAFS spectra. When the measured
first shell coordination number is less than about 6, accurate
measurements of the cluster size are possible. As the coordina-
tion number asymptotically approaches the bulk metal value of
12, the precision for the measured nanoparticle size is greatly
diminished. The smallest platinum nanoparticle in the Frenkel
study was 2.4 nm with a measured first shell coordination
number of 8.5 (XAFS). For this small cluster, scattering
contributions from the second, third, and fourth shells are still
clearly evident. In other words, for a 2.4 nm particle the ø˜(R)-
peaks in the range from 3 to 6 Å have amplitudes that remain
at about 20% of the amplitude of the first shell peak at 2.5 Å.
In a study of small 13 atom Cu clusters, the scattering from the
second and third shell metal was clearly detected.35 For the
rhodium cluster of this study there is no evidence of higher
shells in the 3 to 6 Å range. Hence this analysis is consistent
with the fact that the cluster size is about four atoms since the
higher shell scattering is not present. Further, for the active
catalyst system it is possible to evaluate both the magnitude
and the imaginary ø˜(R) peaks in the range from 3 to 6 Å to
establish an upper limit for the presence of trace metal
(nanoparticles or bulk) of <1%.
through the rhodium cluster stage. We can consider two different
classes of ligand types. One would be described as a weak ligand
wherein the free energy of the metal nanoparticle is lower than
that of the cluster intermediate. In this case, the rhodium
precursor is converted to Rh4-6 clusters, and then these clusters
agglomerate to form a metal nanoparticle. The rate of reaction,
for the conversion of the rhodium clusters into a nanoparticle
phase, would be limited by an activation barrier for cluster
agglomeration and growth. Another ligand class would include
a strong ligand in which the free energy of the cluster is lower
than that of the nanoparticle phase. In this case the reaction has
already proceeded to the minimum energy at the cluster phase
and cannot progress to the nanoparticle phase. We briefly
explore the hypothesis that the dehydrocoupling reaction
involves a weak ligand intermediate and that the dehydrocou-
pling reaction then occurs by the formation of a finite amount
of a metal nanoparticle phase. In this case, sufficient metal phase
must have formed within 1 h to be coincident with the maximum
rate of DMAB dehydrocoupling. Further, the agglomeration
process should continue at approximately the same rate in the
subsequent 72 h after the reaction is initiated. In other words,
for this weak ligand model, one would expect a nearly 100-
fold increase in the metal concentration at this last stage of the
reaction. In contrast, the XAFS results show that the rhodium
metal concentration is below the detection limit of 1% in this
end product (t ) 72 h) implying that the rhodium metal
concentration during the active DMAB dehydrocoupling (t )
1 h) would be less than 0.01%. It is thus most likely that the
amine borane ligands are strong ligands that stabilize the Rh4-6
cluster inhibiting agglomeration to a metal nanoparticle.
The spectrum of the precipitate after 72 h in an H2
environment provides another important test for the existence
of a trace metal nanoparticle species. The transition from the
precursor compound to a nanoparticle phase should progress
We have presented a range of XAFS results showing that
there is neither metal nor metal nanoparticles present during
(38) Frenkel, A. I.; Hills, C. W.; Nuzzo, R. G. J. Phys. Chem. B 2001, 105,
12689.
9
J. AM. CHEM. SOC. VOL. 129, NO. 39, 2007 11947

A R T I C L E S
Scheme 1
Fulton et al.
catalysis: (i) the observed rhodium-rhodium distance of 2.73
Å is longer than that for rhodium metal (2.68 Å) thus stabilized
against agglomeration; (ii) the rhodium-rhodium coordination
number is 3.3 which is much smaller than 12 for bulk metal;
(iii) there is no evidence of third and fourth rhodium-rhodium
shells that exist for nanoparticles or a metal; (iv) there is good
sensitivity for a second shell rhodium that is consistent with a
small amount of Rh6 cluster but not nanoparticles; (v) isosbestic
points show only two rhodium species, the precursor and the
Rh4-6 cluster; (vi) the maximum rate of amine borane dehy-
drocoupling is coincident with the maximum concentrations of
Rh4-6 clusters (from corresponding NMR kinetic data, not
shown); (vii) the precipitate after 72 h under reducing conditions
does not contain metal; (viii) added DMAB monomer redis-
solves the precipitate which is unlikely for rhodium metal; (ix)
added DMAB dimer precipitates the active catalyst which is
consistent with the cluster model; (x) the induction period
observed for the initiation of the dehydrocoupling reaction is
consistent with the delay to form the active Rh4-6 cluster from
the precursor.
the nitrogen atoms because the ligand binding region between
1 and 2 Å for a series of primary and secondary amine boranes
are all the same. If there was binding through the nitrogen, then
the carbon on the alkyl groups in the second shell would weakly
contribute to the scattering in the range around 2 Å, which is
not observed.
We also modeled a series of different ligand binding
geometries according to the structures shown in Scheme 1.
Hence in structures IV and V, we tested the hypothesis that the
amine borane would simply replace the CO terminal or bridging
sites in Rh4(CO)12 by fitting to the appropriate Rh-B and Rh-N
scattering paths generated through FEFF8. In both cases the
Rh-N distances were too long with respect to the measured
Rh-N distance of 2.67 ( 0.04 Å reported in Table 2. Likewise
structure III, in which there is a tetrahedral bond about the boron
atom, with a Rh-B-N bond angle of approximately 109°, can
be discounted on the same basis due to the fact that the Rh-N
distance would be too long. On the other hand, both structures
I and II are consistent with the measured distances and
coordination numbers for both the Rh-B and Rh-N bonds.
In this study we observe 99% of the rhodium containing
species present during the reaction. Even though our evidence
is extremely strong, we can never rule out the fraction of a
percent of the material that may be metallic in nature. However,
previous reports stating that the predominant species present
during catalysis is metallic rhodium are not consistent with our
observations.
4.2. Structure of Ligands on the Rh4-6 Catalyst. 4.2.1.
From Chloro-(1,5-cyclooctadiene)rhodium (I) Dimer. In
previous sections we have shown that the complex contains
Rh4-6 clusters. Although XAFS has a limited ability to resolve
the ligand structure about this cluster, through XAFS studies
of related chemical systems it is possible to generate a clear
picture of the mostly likely ligand structure. For instance we
have previously shown that chloride is not one of the first shell
ligands by conducting chemistry on the rhodium cyclooctadiene
analogue, bis(1,5-cyclooctadiene)rhodium (I) trifluoromethane-
sulfonate.
The Rh-N bond distance constrains the Rh-B-N bond
angle to slightly less than 90°. This constraint is satisfied by
structure II involving the cyclic DMAB dimer in a bridging
position and by structure I involving bridging of the linear dimer.
Overall, structure II is slightly favored because DiPAB and
DCAB do not readily form new intermolecular B-N bonds,
and yet the spectra of the observed rhodium complexes for both
DMAB and DiPAB/DCAB are nearly identical. The final DCAB
or DiPAB products can interact through the pi bonding with
another bound R2NBH2 product while bound to the rhodium
cluster to yield the same type of rhodium-boron and rhodium-
nitrogen interactions as the DMAB dimer.
We measure the rhodium-boron distance as 2.095 ( 0.02
Å. Crystallographic data exist for only a few rhodium-boron
complexes.39,40 Organoborane compounds can form rhodium-
boron bonds via a dihydridorhodium linkage having a single
(39) Westcott, S. A.; Marder, T. B.; Baker, R. T.; Harlow, R. L.; Calabrese, J.
C.; Lam, K. C.; Lin, Z. Y. Polyhedron 2004, 23, 2665.
(40) Hartwig, J. F.; Cook, K. S.; Hapke, M.; Incarvito, C. D.; Fan, Y. B.;
Webster, C. E.; Hall, M. B. J. Am. Chem. Soc. 2005, 127, 2538.
From the ø˜(R) plots in Figure 8 we have shown that the amine
borane ligands likely bind through the boron atom rather than
9
11948 J. AM. CHEM. SOC. VOL. 129, NO. 39, 2007

When is a Nanoparticle a Cluster?
A R T I C L E S
rhodium-boron bond with a length of 2.17 Å.39 Dialkoxy-
boranes (e.g., pinacolborane) can form single rhodium-boron
bond having lengths of about 2.07 Å. Dialkoxyboranes also form
semibridging rhodium-boron bonds between two rhodium sites
with nonequivalent bond lengths of 2.057 and 2.444 Å.39 Such
complexes also include both rhodium-rhodium bridging and
rhodium terminal hydride groups. Finally a dirhodium DiPPE
complex with B3H7 involves one bridging and two terminal
rhodium-boron bonds having distances of 2.16 and 2.25 Å,
respectively. This species contains both rhodium-rhodium and
rhodium-boron bridging hydrides and in addition to a terminal
hydride on the rhodium. None of these systems are completely
chemically equivalent to the Rh4-6 complex that contains Rh-
B-N bonds. Overall though the rhodium-boron bond distance
for the Rh4-6 cluster in this study is most consistent with the
rhodium-boron bond distances measured for the dialkoxyborane
compounds.
always pin down the exact structure of the active catalyst. In
this case the observed rhodium clusters most likely make up a
catalyst resting state; however the proposed structures could just
as easily be the active catalyst. However unlikely, we cannot
rule out that <1% of the unobserved rhodium might be a
metallic dehydrocoupling catalyst. In the same way, we cannot
rule out a monomeric rhodium complex being the active catalyst.
The latter is more amenable to the observed rhodium clusters,
especially if the Rh4 and Rh6 clusters are in equilibrium with
one another (as per eq 6).
We have shown the utility of operando XAFS for structural
determinations of catalytic, especially homogeneous catalytic,
systems. We are currently applying the methods of operando
XAFS described herein to help discriminate other homogeneous
vs heterogeneous cluster questions.
Acknowledgment. We acknowledge Drs. M. Bullock and D.
Dubois for helpful discussions. We also acknowledge the
assistance of Alan Willse in the PCA analysis. This work was
supported by the U.S. Department of Energy’s (DOE) Office
of Basic Energy Sciences, Chemical Sciences program. The
Pacific Northwest National Laboratory is operated by Battelle
for DOE. PNC/XOR facilities at the Advanced Photon Source,
and research at these facilities, are supported by the U.S.
Department of Energy-Basic Energy Sciences, a major facilities
access grant from NSERC, the University of Washington, Simon
Fraser University, the Pacific Northwest National Laboratory,
and the Advanced Photon Source. Use of the Advanced Photon
Source is also supported by the U.S. Department of Energy,
Office of Science, Office of Basic Energy Sciences, under
Contract DE-AC02-06CH11357.
Conclusions
We believe, in contrast to results derived from ex situ
measurements, we have demonstrated that there is little or no
metallic rhodium present during the catalyzed dehydrocoupling
of amine boranes. This use of operando XAFS sheds light onto
a particularly opaque analytical problem. XAFS does not allow
the discrimination of absolute structures and does not necessarily
yield quantitative information about the number of structures
in solution. However in this catalytic system we have shown
that the rhodium catalyst precursor compounds cleanly convert
to a new set of clusters containing four or six rhodium atoms
ligated most likely through the boron atom of the amine borane
substrates or amine borane products. All catalyst precursors and
all amine borane substrates convert to very similar rhodium
clusters. We have shown that the catalytic dehydrocoupling does
produce metallic rhodium in the presence of air and that the
rhodium clusters can be transformed into metallic rhodium by
air oxidation of the borane ligands. Unfortunately one cannot
Supporting Information Available: XAFS spectra and analy-
sis. This material is available free of charge via the Internet at
http://pubs.acs.org.
JA073331L
9
J. AM. CHEM. SOC. VOL. 129, NO. 39, 2007 11949