Nanocluster formation synthetic, kinetic, and mechanistic studies. The detection of, and the methods to avoid, hydrogen mass-transfer limitations in the synthesis of polyoxoanion- and tetrabutylammonium-stabilized, near- monodisperse 40 ± 6 A? Rh(0) nanoclusters

Article abstract of DOI:10.1021/ja9719485

Previously we reported P₂W₁₅Nb₃O₆₂^9- polyoxoanion- and Bu₄N⁺- stabilized, 20 ± 3 A?, Ir(0)(~300) nanoclusters. These nanoscopic materials are synthesized from the reduction of [(C₄H₉)₄N]₅Na₃[(1,5- COD)M·P₂W₁₅Nb₃O₆₂] (M = Ir) by H₂ in acetone and isolable, highly catalytically active, with an unprecedented catalytic lifetime in solution. However, an initial attempt to synthesize Rh(0) nanoclusters from the analogous M = Rh precursor, and under conditions otherwise identical to the M = Ir synthesis, led to poydisperse 16 to 116 A? Rh(0) nanoclusters. The above results led, in turn, to the discovery that H₂ gas-to-solution mass-transfer limitations (MTL) were responsible for the failure of the initial synthesis, an important finding since H₂ is one of the most common reducing agents in syntheses of modern, near-monodisperse transition metal nanoclusters. The finding of a hydrogen MTL regime is fortified by stirring-rate-dependence data, kinetic data [demonstrating a catalyst nondependent (MTL) regime, and a catalyst-dependent (chemical-reaction rate-limiting regime, and transmission electron microscopy used as a mechanistic probe. The results provided evidence for a growth mechanism involving parallel autocatalytic surface growth (leading to near-monodisperse nanoclusters) in competition with diffusive agglomeration (leading to polydisperse nanoclusters). The above mechanistic insights were then used, in turn, to design conditions where only the autocatalytic surface-growth pathway occurs, conditions which led to the successful synthesis of the desired near-monodisperse, 40 ± 6 A? Rh(0) P₂W₁₅Nb₃O₆₂^9- polyoxoanion- and Bu₄N⁺-stabilized transition metal nanoclusters.

Full text of DOI:10.1021/ja9719485

J. Am. Chem. Soc. 1998, 120, 9545-9554
9545
Nanocluster Formation Synthetic, Kinetic, and Mechanistic Studies.†
The Detection of, and Then Methods To Avoid, Hydrogen
Mass-Transfer Limitations in the Synthesis of Polyoxoanion- and
Tetrabutylammonium-Stabilized, Near-Monodisperse 40 ( 6 Å Rh(0)
Nanoclusters
John D. Aiken III and Richard G. Finke*
Contribution from the Department of Chemistry, Colorado State UniVersity, Fort Collins, Colorado 80523
ReceiVed June 12, 1997. ReVised Manuscript ReceiVed July 17, 1998
Abstract: Previously we reported P2W15Nb3O629 polyoxoanion- and Bu4N -stabilized, 20 ( 3 Å, Ir(0)∼300
nanoclusters. These nanoscopic materials are synthesized from the reduction of [(C4H9)4N]5Na3[(1,5-
COD)M‚P2W15Nb3O62] (M ) Ir) by H2 in acetone and are isolable, highly catalytically active, with an
unprecedented catalytic lifetime in solution. However, an initial attempt to synthesize Rh(0) nanoclusters
from the analogous M ) Rh precursor, and under conditions otherwise identical to the M ) Ir synthesis, led
to polydisperse 16 to 116 Å Rh(0) nanoclusters. The above results led, in turn, to the discovery that H2
gas-to-solution mass-transfer limitations (MTL) were responsible for the failure of the initial synthesis, an
important finding since H2 is one of the most common reducing agents in syntheses of modern,
near-monodisperse transition metal nanoclusters. The finding of a hydrogen MTL regime is fortified by stirring-
rate-dependence data, kinetic data [demonstrating a catalyst nondependent (MTL) regime, and a catalyst-
dependent (chemical-reaction rate-limiting) regime], and transmission electron microscopy used as a mechanistic
probe. The results provided evidence for a growth mechanism involving parallel autocatalytic surface growth
-
+
(leading to near-monodisperse nanoclusters) in competition with diffusive agglomeration (leading to polydisperse
nanoclusters). The above mechanistic insights were then used, in turn, to design conditions where only the
autocatalytic surface-growth pathway occurs, conditions which led to the successful synthesis of the desired
9
-
+
near-monodisperse, 40 ( 6 Å Rh(0) P2W15Nb3O62 polyoxoanion- and Bu4N -stabilized nanoclusters. The
+
resultant Rh(0) nanoclusters are only the second example of polyoxoanion- and Bu4N -stabilized transition
metal nanoclusters.
Introduction
colloids (>10 nm or 100 Å) have been known for over 140
3
years, general methods for the designed synthesis of a precho-
Presently there is considerable interest in the synthesis of
near-monodisperse (e(15% in diameter) nanoclusters (<10
nm or 100 Å). In particular there is interest in controlling the
size, size-distribution, composition, and shape of such clusters
via designed syntheses. However, and despite the fact that metal
sen transition metal nanocluster remain unavailablesa statement
1
that is still true despite some notable recent synthetic advances.4
2
In fact, the empiricism in existing nanocluster synthetic routes
(
2) (a) See elsewhere for a review of nanocluster catalysis, which includes
2b
necessary key terms and definitions of: nanoclusters; traditional colloids;
monodisperse and near-monodisperse nanoparticles; “magic number” (i.e.,
full shell and thus enhanced stability) nanoclusters; Schwartz’s updated
definition of homogeneous vs heterogeneous catalysts; and inorganic
(“charge”) and organic (“steric”) stabilization mechanisms for colloids and
†
Part IV in a series.
(1) Reviews on nanoclusters: (a) Jena, P.; Rao B. K.; Khanna, S. N.
Physics and Chemistry of Small Clusters; Plenum: New York, 1987. (b)
Andres, R. P.; Averback, R. S.; Brown, W. L.; Brus, L. E.; Goddard, W.
A.; Kaldor, A.; Louie S. G.; Moscovits, M.; Peercy, P. S.; Riley, S. J.;
Siegel, R. W.; Spaepen, F.; Wang, Y. J. Mater. Res. 1989, 4, 704. This is
a Panel Report from the United States Department of Energy, Council on
Materials Science on “Research Opportunities on Clusters and Cluster-
assembled Materials”. (c) Thomas, J. M. Pure Appl. Chem. 1988, 60, 1517.
+
nanoparticles. A review of the Bu4N and polyoxoanion-stabilized Ir(0)∼300
nanoclusters is also provided in that review.2b (b) Aiken, J. D., III; Lin, Y.;
Finke, R. G. J. Mol. Catal. 1996, 114, 29-51.
(3) Faraday, M. Philos. Trans. R. Soc. 1857, 147, 145.
(4) (a) For instance, see Professor Reetz’s important electrochemical
synthesis4c (see elsewhere for a discussion of this interesting method), as
well as the papers cited in refs 15-19 of ref 8a. (b) Reetz, M. T.; Helbig,
W.; Quaiser, S. A.; Stimming, U.; Breuer, N.; Vogel, R. Science 1995,
267, 367. (c) Reetz, M. T.; Helbig, W. J. Am. Chem. Soc. 1994, 116, 7401.
(d) Reetz, M. T.; Quaiser, S. A. Angew. Chem., Int. Eng. Ed. 1995, 34,
2240. (e) Reetz, M. T.; Helbig, W.; Quaiser, S. A. Chem. Mater. 1995, 7,
2227. (f) Reetz, M. T.; Quaiser, S. A.; Breinbauer, R.; Tesche, B. Angew.
Chem., Int. Eng. Ed. 1995, 34, 2728. (g) Reetz, M. T.; Lohmer, G. Chem.
Commun. 1996, 1921. (h) Reetz, M. T.; Breinbauer, R.; Wanninger, K.
Tetrahedron Lett. 1996, 37, 4499. (i) Review: Reetz, M. T.; Helbig, W.;
Quaiser, S. A. In ActiVe Metals; F u¨ rstner, A., Ed.; VCH Publishers: New
York, 1996; Chapter 7, pp 279-297. (j) Kolb, U.; Quaiser, S. A.; Winter,
M. A.; Reetz, M. T. Chem. Mater. 1996, 8, 1889.
2b
(d) Henglein, A. Chem. ReV. 1989, 89, 1861. (e) A superb series of papers,
complete with a record of the insightful comments by the experts attending
the conference, is available in: Faraday Discuss. 1991, 92, 1-300. (f)
Bradley, J. S. In Clusters and Colloids. From Theory to Applications;
Schmid, G., Ed.; VCH: New York, 1994; pp 459-544. (g) Schmid, G. In
Aspects of Homogeneous Catalysis; Ugo, R., Ed.; Kluwer: Dordrecht, 1990;
Chapter 1. (h) B o¨ nnemann, H.; Braun, G.; Brijoux, W.; Brinkmann, R.;
Tilling, A. S.; Seevogel, K.; Siepen, K. J. Organomet. Chem. 1996, 520,
1
43-162 and the collection of “key publications” cited as refs 2-61 therein.
(
i) ActiVe Metals: Preparation, Characterization, Applications; F u¨ rstner,
A., Ed.; VCH: Weinheim, 1996. (j) Schmid, G. in Applied Homogeneous
Catalysis with Organometallic Compounds; Cornils, B., Herrmann, W.
A., Eds.; VCH: Weinheim, 1996; Vol. 2.
S0002-7863(97)01948-3 CCC: $15.00 © 1998 American Chemical Society
Published on Web 09/02/1998

9546 J. Am. Chem. Soc., Vol. 120, No. 37, 1998
Aiken and Finke
Scheme 1. The Minimum Mechanism Uncovered
Previously for the Formation of Ir(0) Nanoclustersa
8a
Figure 1. TEM image (100K magnification) of near-monodisperse
Ir(0)∼300 nanoclusters grown by the reduction of 1.2 mM 1 in acetone
under chemical-reaction-rate-limiting (i.e., non-MTL) conditions. This
image shows Ir(0)∼300 nanoclusters that are 20 ( 3 Å (i.e., (15%) in
diameter.
has led to comments in the literature expressing frustration over
the lack of clear, generally applicable synthetic guidelines and
mechanistic insights.1f,g,5
a
Previously we reported the synthesis of near-monodispersed,
polyoxoanion- and tetrabutylammonium-stabilized, 20 ( 3 Å
This mechanism consists of (a) slow, continuous, homogeneous
8a
nucleation (steps I-III), rate constant k
A f B, followed by (b) fast, autocatalytic surface growth (step IV),
rate constant k
for the pseudo-elementary step8a A + B f 2B.
Nucleation and growth are separated in time since k , k [B], which
in turn is a key to the observed formation of a near-monodisperse
1
for the pseudo-elementary step
6
Ir(0)∼300 nanoclusters, Figure 1. That synthesis is based on
our polyoxoanion-supported7 organometallic precursor [(n-
2
1
2
C4H9)4N]5Na3[(1,5-COD)Ir‚P2W15Nb3O62], 1, and involves re-
ducing this complex under hydrogen in acetone and in the
presence of cyclohexene as a catalytic substrate (vide infra), eq
(
e(15%) particle size distribution. In this example, diffusive ag-
glomeration (step V) does not compete kinetically with autocatalytic
surface growth, although literature precedent suggests that this step
(5) (a) Weller, H. Angew Chem., Int. Ed. Engl. 1993, 32, 41-53. (b)
18
always occurs at a finite rate.
Matijevic, E. Chem. Mater. 1993, 5, 412. (c) Steigerwald, M. L.; Brus, L.
Acc. Chem. Res. 1990, 23, 183. (d) Beattie, J. K. Pure Appl. Chem. 1989,
6
1, 937. (e) See also the quotes and references provided in footnotes 5 and
elsewhere. (f) LaMer’s classic work on the mechanism of formation of
1. The nanocluster formation reaction can be followed directly
by monitoring the production of cyclooctane throughout the
course of the reaction.
8
a
7
sulfur sols involving (i) “burst” nucleation from supersaturated solution,
and (ii) diffusive growth: LaMer, V. K.; Dinegar, R. H. J. Am. Chem. Soc.
1
950, 72, 4847. Lamer, V. K. Ind. Eng. Chem. 1952, 44, 1270.
6) (a) Lin, Y.; Finke, R. G. J. Am. Chem. Soc. 1994, 116, 8335-8353.
b) As described in refs 29 and 43a of ref 6a, we represent the 20 ( 3 Å
(
(
distribution of Ir(0)∼190 to Ir(0)∼450 nanoclusters present as their aVerage
composition of Ir(0)∼300 for convenience and clarity, only. We do not mean
to imply that a perfectly monodispersed, exactly Ir(0)∼300, cluster is the
only particle present. Similarly, we use Ir(0)∼900 to represent 30 ( 4 Å
Ir(0)∼640 to Ir(0)∼1460 nanoclusters. (c) Note that it is important to avoid
We have studied the kinetics and mechanism of this reaction
in detail for the iridium example; the results led to the finding
that nanocluster growth occurs by a non-LaMer mechanism
the misunderstanding possible from the use, in the literature, of “Pd561” to
x- 6d
represent what is really Pd570(30(phen)63(3(OAc)190(10]
,
or “Au55(PPh3)12-
5f
Cl6” to represent what is now realized to be “Au67(PPh3)14Cl8 with small
6e
involving a sequence of (i) slow, continuous nucleation and (ii)
clusters as impurities”. (d) Vargaftik, M. N.; Zagorodnikov, V. P.; Stolarov,
I. P.; Moiseev, I. I.; Kockubey, D. I.; Likholobov, V. A.; Chuvilin, A. L.;
Zamaraev, K. I. J. Mol. Catal. 1989, 53, 315. (e) Fackler, J. P.; McNeil, C.
J.; Winpenny, R. E. P.; Pignolet, L. H. J. Am. Chem. Soc. 1989, 111, 6434-
8
autocatalytic surface growth, Scheme 1. Significant features
of this new mechanism are that it predicts the formation of near-
6
435.
(
(8) (a) Part I: Watzky, M. A.; Finke, R. G. J. Am. Chem. Soc. 1997,
119, 10382-10400. (b) Part II: Watzky, M. A.; Finke, R. G. Chem. Mater.
1997, 9, 3083-3095. (c) Part III: Transition Metal Nanocluster Formation
Kinetics and Mechanism. A New Kinetic Method Based on Catalytic
Activity and the Concept of Pseudoelementary Steps. Watzky, M. A.;
Widegren, J. A.; Aiken, J. D., III; Finke, R. G. Submitted for publication.
7) P2W15Nb3O629 synthesis and characterization papers: (a) Weiner,
-
H.; Aiken, J. D., III; Finke, R. G. Inorg. Chem. 1996, 34, 7905 and
references therein. (b) Finke, R. G.; Lyon, D. K.; Nomiya, K.; Weakley, T.
J. Acta Crystallogr., Sec. C 1990, 46, 1592. (c) Edlund, D. J.; Saxton, R.
J.; Lyon, D. K.; Finke, R. G. Organometallics 1988, 7, 1692.

Nanocluster Formation Synthetic, Kinetic, and Mechanistic Studies
J. Am. Chem. Soc., Vol. 120, No. 37, 1998 9547
monodisperse nanoclusters, and that it can rationalize the
formation of so-called “magic number” clusters.8b Additional
synthetic insights and testable mechanistic predictions are
8b
available based on the realization that nanocluster autocatalytic
8
b
surface growth is a “living (metal) polymer” phenomenon.
The rate constants k1 and k2, Scheme 1, can be determined from
curve fitting the H2 pressure vs time to the appropriate analytic
kinetic equations. The rate constant k1 measures the overall
8
a
nucleation process while the k2 measures nanocluster growth.
Herein we report our discovery of, and subsequent mecha-
nistic data fortifying, an important guideline in nanocluster
syntheses: the need to aVoid hydrogen-mass-transfer limitations
when synthesizing near-monodisperse transition metal nano-
clusters under H2, especially when using rhodium and other,
more catalytically active second row metals. This finding was
discovered during our initial attempts to synthesize near-
monodisperse rhodium nanoclusters by the route we previously
6,9
synthesized polyoxoanion-stabilized Ir(0)∼300 nanoclusters. By
using several “fingerprint” tests from the well-established mass-
transfer literature, we show herein that nanoclusters grown under
hydrogen mass-transfer-limiting (MTL) conditions are polydis-
perse and irregularly shaped in comparison to the near-
monodisperse, spherical 40 ( 6 Å Rh(0) nanoclusters grown
under non-mass-transfer-limiting (non-MTL), chemical-reaction-
rate-limiting conditions. The results reveal the dramatic effects
on nanocluster shape and dispersity (see Figures 1-4) and show
that the solution hydrogen concentration provides one simple
but crucial parameter in the synthesis of near-monodisperse,
e15% size distribution, transition metal nanoclusters. The
present study is a continuation of our kinetic, mechanistic, and
Figure 2. TEM image (430K magnification) of polydisperse rhodium
nanoclusters grown by the reduction of 1.2 mM 2 in acetone under
hydrogen mass-transfer-limiting conditions and isolated following 100%
conversion of the precursor. The clusters have irregular shapes (see
the inset) and range in size from 16 to 116 Å with an average diameter
of 40 ( 14 Å (i.e., ( 35%). A particle size histogram of this sample
is provided as Figure C of the Supporting Information.
8
synthetic studies into nanocluster formation reactions.
research goals in nanocluster science.5 Hence, to test this
hypothesis, and to attempt to grow near-monodisperse rhodium
nanoclusters, we altered the synthetic reaction conditions in a
Results and Discussion
Transmission Electron Microscopy (TEM) Studies. Dis-
tinguishing between Polydisperse (.15% Size Distribution)
and Near-Monodisperse (e15% Size Distribution) Transi-
tion Metal Nanoclusters. The easiest way to illustrate the
dramatic effects observed by running nanocluster syntheses
under hydrogen mass-transfer-limited, vs chemical-reaction-rate-
limited, conditions is to view Figures 1-4. Figure 1 shows
the spherical, near-monodisperse, 20 ( 3 Å Ir(0)∼300 nano-
clusters obtained beginning from the Ir(I) precursor 1 under our
previously developed Standard Conditions6 (40 psig H2, 22 °C
and 1.2 ( 0.1 mM 1). Figure 2 shows dramatically different,
irregularly shaped, polydisperse-sized, 16 to 116 Å Rh∼520 to
Rh∼60000 nanoclusters obtained when we applied the identical
conditions to the rhodium precatalyst, 2. Note that the only
difference between the syntheses used for the results shown in
Figure 1 vs Figure 2 is the iridium vs rhodium metal in the
(10) (a) There are but two other systems that provide some initial hints,
and thus some precedent, for the hydrogen MTL effects seen herein. One
is our own initial study of nanoclusters syntheses, in particular Step V in
6
a
the mechanistic scheme in Figure 13 on pp 8346-8347 elsewhere. The
second is Whitesides’ work studying the reactions and stereochemistry of
Pt metal surface alkyls with H2 or D2. Note, however, that the metal products
were not a focus, and thus were not studied, in Whitesides’ series of valuable
b-f
papers.
(b) Lee, T. R.; Whitesides, G. M. Acc. Chem. Res. 1992, 25,
2
2
1
66-272. (c) Lee, T. R.; Whitesides, G. M. J. Am. Chem. Soc. 1991, 113,
568-2576. (d) Lee, T. R.; Whitesides, G. M. J. Am. Chem. Soc. 1991,
13, 368-369. (e) Lee, T. R.; Laibinis, P. E.; Folkers, J. P.; Whitesides,
a
G. M. Pure Appl. Chem. 1991, 63, 821-828. (f) Miller, T. M.; Whitesides,
G. M. J. Am. Chem. Soc. 1988, 110, 3164-3170.
(11) Some examples of nanoparticles grown with hydrogen as a reducing
agent include: (a) Yonezawa, T.; Tominaga, T.; Richard, D. J. Chem. Soc.,
Dalton Trans. 1996, 783. (b) A review: B o¨ nnemann, H.; Brijoux, W. In
ActiVe Metals: Preparation, Characterization, Applications; F u¨ stner, A.,
Ed.; VCH: Weinheim, 1996. See especially p 361 therein. (c) Schmid, G.;
Maihack, V.; Lantermann, F.; Peschel, S. J. Chem. Soc., Dalton Trans.
1996, 589. (d) Henglein, A.; Ershov, B. G.; Malow, M. J. Phys. Chem.
8-
precursor: [(1,5-COD)Ir‚P2W15Nb3O62] (1), or [(1,5-COD)-
1995, 99, 14129. (e) B o¨ nnemann, H.; Brijoux, W.; Brinkmann, R.; Fretzen,
R.; Joussen, T.; Koppler, R.; Korall, B.; Neiteler, P.; Richter, J. J. Mol.
Catal. 1994, 86, 129. (f) Schmid, G.; Harms, M.; Malm, J.-O.; Bovin, J.-
O.; van Ruitenbeck, J.; Zandbergen, H. W.; Fu, W. T. J. Am. Chem. Soc.
1993, 115, 2046. (g) Lee, T. R.; Whitesides, G. M. Acc. Chem. Res. 1992,
Rh‚P2W15Nb3O62]8- (2).
Having seen preliminary evidence for cluster agglomeration
10
in our earlier work, and noting that rhodium nanoclusters were
much more active cyclohexene hydrogenation catalysts than their
iridium counterparts, we postulated that the polydisperse
rhodium clusters shown in Figure 2 are the result of hydrogen
gas-to-solution, MTL reaction conditions. Furthermore, we
realized that this observation was an important one, given that
hydrogen is one of the most widely used reductants in transition
2
5, 266-272. (h) Toshima, N.; Takahashi, T. Bull. Chem. Soc. Jpn. 1992,
6
5, 400. (i) Lee, T. R.; Whitesides, G. M. Catal. Lett. 1991, 9, 461-472.
(
j) Lee, T. R.; Whitesides, G. M. J. Am. Chem. Soc. 1991, 113, 368-369.
(k) Lee, T. R.; Laibinis, P. E.; Folkers, J. P.; Whitesides, G. M. Pure Appl.
Chem. 1991, 63, 821-828. (l) Vargaftik, M. N.; Zargorodnikov, V. P.;
Stolarov, I. P.; Moiseev, I. I.; Kochubey, D. I.; Likholobov, V. A.; Chuvilin,
A. L.; Zamaraev, K. I. J. Mol. Catal. 1989, 53, 315-348. (m) Schmid, G.;
Lehnert, A. Angew. Chem., Int. Ed. Engl. 1989, 28, 780-781. (n) Schmid,
G.; Morun, B.; Malm, J.-O. Angew. Chem., Int. Ed. Engl. 1989, 28, 778.
11
metal nanocluster syntheses in the literature, and given that
size control and narrow particle size distributions are key, current
(
3
o) Meguro, K.; Torizuka, M.; Esumi, K. Bull. Chem. Soc. Jpn. 1988, 61,
41. (p) Harrison, J. B.; Berkheiser, V. E.; Erdos, G. W. J. Catal. 1988,
1
12, 126. (q) Boutonnet, M.; Kizling, J.; Stenius, P.; Maire, G. Colloids
(9) Lin, Y.; Finke, R. G. Inorg. Chem. 1994, 33, 4891-4910.
Surf. 1982, 5, 209.

9548 J. Am. Chem. Soc., Vol. 120, No. 37, 1998
Aiken and Finke
experiments, reported below, confirm that these pressure,
temperature, and precatalyst concentration conditions avoid any
problems with hydrogen mass transfer. In short, the two key
results here are (i) the strong evidence for MTL effects in these
particular nanocluster syntheses under H2 and (ii) the design of
conditions for the successful synthesis of near-monodisperse
Rh(0) nanoclusters.
Characterization of the Rh(0) nanoclusters was accomplished
by TEM and EDX analysis. TEM images (Figure 3) show
nonaggregated, near-monodisperse, 40 ( 6 Å spherical nano-
1
4
clusters. [TEM images at lower (100 K) and higher (400 K)
magnifications of these clusters are shown in Figure A of the
Supporting Information.] EDX analysis on this sample detects
O, Nb, and W, as well as Rh, confirming that the nanocluster
6a,8,9
9-
15
stabilizing
polyoxoanion, P2W15Nb3O62 , is also present.
As in the formation of near-monodisperse 20 ( 3 Å Ir(0)∼300
8a,b
nanoclusters, analytical curve-fitting of the hydrogen pressure
16
vs time data obtained during this reaction can be accomplished
8
if one incorporates our previously elucidated nucleation (A f
B), and then autocatalytic, surface-growth (A + B f 2B), steps,
Supporting Information, Figure B. This is strong evidence that
near-monodisperse Rh(0) nanoclusterssjust like their near-
monodisperse Ir(0)∼300 analoguessform via the previously
deduced mechanism involving these two steps, Scheme 1.8a
Formation of Polydisperse Ir Nanoclusters under Delib-
erately Induced Hydrogen MTL Conditions. An important
question to answer is the following: If we deliberately pick
conditions which disfaVor hydrogen mass transfer, specifically
the extreme conditions of a nonstirred solution, can we cause
the formation of amorphous, bulk Ir(0) metal? That is, can we
induce the opposite result compared to the synthesis of the near-
monodisperse Ir(0)∼300 nanoclusters shown in Figure 1, obtained
at stirring rates of g400 rpm? Figure 4 shows the results of a
deliberate and successful attempt to induce MTL conditions in
the case of the Ir(0) nanoclusters. This result provides good
evidence for the broader applicability of H2 MTL effects in
transition metal nanocluster syntheses.
Additional Evidence for a Hydrogen MTL Regime. (I)
Establishment of a Stirring Rate Dependence on the Reac-
(
12) The H2 concentration in neat acetone at 49 psig H2 and 3 °C is 1.5
-
2
×
10 M, unchanged (within experimental error) from its concentration
at 40 psig H2 and 22 °C. Thus, increasing the H2 pressure from 40 to 49
psig H2 compensates for the decrease in H2 solubility at lower temperatures,
and suggests that slower reaction chemistrysand not the dissolved H2
concentrationsis responsible for the change in particle dispersity in this
case. H2 solubilities were calculated by using both Ostwald coefficients
(The Matheson Unabridged Gas Book; Hydrogen; Matheson: East Ruth-
erford, NJ, 1974) and mole fraction data (Solubility Data Series: Hydrogen
and Deuterium; Pergamon Press: New York, 1981; p 224).
Figure 3. TEM image (300K magnification) and associated particle
size histogram (n ) 146 particles counted) of isolated, near-monodis-
perse 40 ( 6 Å Rh(0) nanoclusters grown by the reduction of 0.79
mM 2 in acetone under chemical-reaction-rate-limiting (i.e., non-MTL)
conditions (3 °C, 49 ( 1 psig H
conditions is independent of stirring rate and is linearly dependent upon
2] (see Figure L in the Supporting Information), evidence demonstrat-
ing that these conditions are non-MTL. The sample was harvested after
00% conversion of 2 into nanoclusters (i.e., after GLC analysis showed
2
). The reaction rate under these
(
13) The 40 ( 6 Å or 34 to 46 Å nanoclusters yield approximate metal
[
6a
content numbers of Rh∼1500 to Rh∼3700, calculated as before with the
23
equation N ) NorV/MW where No ) 6.02 × 10 atoms/mol, r ) 12.41
1
g/cm , V ) 4/3π(D/2) , and MW ) 102.9 g/mol. That is, 34 and 46 Å
rhodium particles contain ≈1500 and 3700 metal atoms, respectively, while
a perfect, spherical 40 Å rhodium particle contains ≈2430 metal atoms.
While the values of 1500 and 3700 metal atom particles are close to the
number of atoms predicted in so-called “magic number” clusters (at 1415
and 3871), the number of atoms in a 40 Å rhodium particle (≈2430) falls
between the two magic number values of 2057 and 2869.8a Hence, we have
avoided in this instance the approximate nomenclature “Rh(0)∼2400” as it
3
3
that 1.0 equiv of cyclooctane had evolved from 2).
manner that disfavors hydrogen MTL control, but favors
chemical-reaction-rate control, of the synthesis.
Formation and Characterization of Near-Monodisperse
40 ( 6 Å Rh(0) Nanoclusters under Non-MTL Conditions.
6c
is potentially misleading.
Figure 3 shows the results obtained following a change in three
key reaction variables in a way to help avoid hydrogen mass-
transfer limitations: higher H2 pressure (which increases H2
solubility), lower temperature (which slows subsequent reaction
chemistry that uses H2),12 and lower precatalyst concentration
(14) Interestingly, the 40 Å diameter of our rhodium nanoclusters is close
to those for Rh synthesized elsewhere with alcohol reductants, Os (∼<10
Å) < Ir (14-30 Å) < Pt (27 Å) < Rh (∼40 Å) < Pd (53 Å); see: Hirai,
H.; Toshima, N. J. Macromol. Sci. Chem. 1979, A13(6), 727.
(15) As expected, peaks for Na, Cr, and Cu are also seen. Na is present
as a countercation (from the precursor 2) whereas the Cr peak comes from
the specimen holder and the Cu peak comes from the copper grid. Being
below the limit of detection, peaks for C and N are not seen but must also
(
which also slows subsequent solution chemistry that uses H2).
These changes resulted in the formation of near-monodisperse,
0 ( 6 Å Rh(0)∼1500 to Rh(0)∼3700 nanoclusters.13 Additional
+
be present given that [(n-C4H9)4N] is present in the precursor 2.
4
(16) Lyon, D. K.; Finke, R. G. Inorg. Chem. 1990, 29, 1787.

Nanocluster Formation Synthetic, Kinetic, and Mechanistic Studies
J. Am. Chem. Soc., Vol. 120, No. 37, 1998 9549
Figure 4. TEM image (at lower, 50K magnification) of the isolated,
predominately bulk Ir(0) metal iridium product formed in an unstirred
solution. (A small number of irregular and aggregated clusters are also
seen at higher magnifications, Supporting Information, Figure K.)
tion Rate and on the Nanocluster Morphology. One classic
simple, but convincing, piece of evidence for the presence of a
gas-to-solution MTL regime is a demonstration of a stirring-
speed dependence to the observed reaction rate.17 Typically,
such a stirring-rate dependence can then be changed to stirring-
rate independence at higher stirring speeds, corresponding to a
change to a non-MTL, chemical-reaction-rate-limiting, regime.
This predicted stirring-rate dependence is exactly what is seen
in the present case, Figure 5, which shows the cyclohexene
hydrogenation rate as a function of stirring rate for both iridium
Figure 5. Hydrogenation rate of cyclohexene as a function of stirring
rate under chemical-reaction-rate-limiting, non-MTL conditions begin-
ning with 1 and showing saturation kinetics (A; 1.2 mM 1, 40 ( 1
psig H
A: 0 rpm, 0.4 mmol H
rpm, 3.3 mmol H /h. Rates in B: 0 rpm, 0.4 mmol H
rpm, 4.6 mmol H /h; 1300 ( 130 rpm, 4.7 mmol H /h. TEM images
2
, 22 °C) and 2 (B; 1.2 mM 2, 49 ( 1 psig H
2
, 3 °C). Rates in
/h; 1300 ( 130
/h; 450 ( 45
2
/h; 400 ( 40 rpm, 3.0 mmol H
2
2
2
2
2
(Figure 5A) and rhodium (Figure 5B) nanoclusters. The first
of the isolated reaction products at 400 and 1300 rpm (right-most two
points; chemical-reaction-rate-limiting regime) show mostly individual,
nonaggregated, near-monodisperse iridium nanoclusters (see Figure 1),
whereas TEM images of the major product isolated at 0 rpm stirring
two points in Figure 5A (iridium) correspond to the limits of
hydrogen MTL (point 1; 0 rpm) and non-MTL (point 2; vortex
stirring at 400 ( 40 rpm) conditions. Additionally, point 3 (far
right; 1300 ( 130 rpm) is well into the non-MTL regime. The
only additional, confirming data that is needed here is TEM
evidence for the shape and polydispersity or near-monodispersity
of the nanoclusters grown at each point on this curve. In fact,
point 1 in Figure 5A (0 rpm) represents the exact same sample
used for the TEM image (amorphous bulk Ir(0) metal) shown
in Figure 4. (A few nanoclusters are present, but they are mostly
aggregated.) In contrast to this, TEM images of the Ir
[
mass-transfer-limiting regime] exhibit amorphous, aggregated Ir(0)
metal (shown at 50K magnification in Figure 4 in the text and at 200K
magnification in Figure K of the Supporting Information).
nanoclusters grown and isolated at points 2 and 3 show well-
defined and unaggregated, near-monodisperse indiVidual clus-
ters analogous to the ones shown in Figure 1.
The stirring rate curves, Figure 5, plus the TEM evidence
for the drastic difference in cluster morphology of these two
examples provide additional evidence for hydrogen gas-to-
solution MTL effects in the present nanocluster syntheses under
hydrogen.
(
17) A few lead references to the more extensive literature of mass-
transfer-limited reactions: (a) Roberts, G. W. In Catalysis in Organic
Synthesis; Rylander, P. N., Ed.; Conference on Catalysis in Organic
Syntheses; Academic Press: New York, 1976; pp 1-48. (b) Hobbs, C. C.;
Drew, E. H.; Van’t Hof, H. A.; Mesich, F. G.; Onore, M. J. Ind. Eng. Chem.
Prod. Res. DeV. 1972, 11, 220. (c) Gogate, M.; Lee, S. Fuel Sci. Technol.
(II) Demonstration of Catalyst Dependent and Nonde-
pendent Kinetic Regimes. A second, classic piece of evidence
1994, 12, 35-50. (d) Mathieu, C.; Dietrich, E.; Delmas, H.; Jenck, J. Chem.
1
7
Eng. Sci. 1992, 47, 2289. (e) Notheisz, F.; Zsigmond, AÄ .; Bart o´ k, M.;
Szegletes, Z.; Smith, G. V. App. Catal. A: General 1994, 120, 105-114
and refs 1-9 therein. Note that the key test for gas-liquid MTL typically
involves showing whether the rate law contains a first-order, [catalyst]1
rate dependence. The equation cited therein of -d[H2]/dt ) krkm‚x/[kr‚x +
km] is useful in this regard (kr ) the chemical reaction-rate-limiting rate
constant; km ) the mass-transfer-limiting kinetic coefficient). (f) See also
references b-f herein. (g) H2 MTL effects on enantioselectivity in
asymmetric catalysis: Sun, Y.; Landau, R. N.; Wang, J.; LeBlond, C.;
Blackmond, D. G. J. Am. Chem. Soc. 1996, 118, 1348-1353.
for a mass-transfer-limiting kinetic regime is a demonstration
that one can go from a catalyst-dependent (reaction-rate limiting)
kinetic regime (e.g., typically a first-order dependence on
catalyst) to a regime zero order in catalyst (the MTL regime).
Figure 6, the rhodium nanocluster growth rate as a function of
[2] and hydrogen pressure (40 or 49 psig H2) at 22 °C, exhibits
just such a phenomenon. At 40 psig H2, this growth rate reaches
a saturation point above ca. 1.0 mM [2], indicating that the

9550 J. Am. Chem. Soc., Vol. 120, No. 37, 1998
Aiken and Finke
Scheme 2. Two Precedented, Parallel Nanocluster Growth
8
a
Pathways: Autocatalysis (top pathway) Leading to
Near-Monodisperse Nanoclusters, and Diffusive
1
8
Agglomerative Growth (bottom pathway) Leading to
Polydisperse Nanoclusters
Figure 6. Hydrogenation rate of cyclohexene as a function of [2] at
2 °C and stirring g750 rpm at both 40 ( 1 psig H (squares; dotted
line) and 49 ( 1 psig H (diamonds; solid line). The rate is initially
2
2
2
dependent upon [2] and hydrogen pressure, but then shows saturation
kinetics in both [2] and hydrogen. At saturation, the reaction is under
hydrogen MTL conditions. The rhodium nanoclusters isolated at the
point indicated by the asterisk (*) on the plot are polydisperse, as the
TEM image in Figure 2 demonstrates.
clusters are predicted throughout the whole reaction. TEM at
both short (e20%) and near-completion (g95%) reaction times
was used to rule out alternative mechanisms involving an
interconversion between poly- and mono-disperse nanoclusters.
This TEM experiment was carried out by first harvesting both
Ir(0) and Rh(0) nanoclusters (seven samples total) at times
corresponding to e20% M(0) formation and g95% M(0)
formation. The cyclooctane product yield accompanying the
conversion of the 1,5-COD in the precatalysts 1 and 2, eq 1 (a
direct monitor of the M(0) formation ), was also determined
by GLC (as detailed in the Experimental Section). The
experiment and the cyclooctane evolution results are shown in
Figure 7, and Table 1 summarizes the degrees of dispersity
determined by TEM for the seven samples.
reaction is operating under hydrogen MTL conditions at this
(and higher) precatalyst concentration. Indeed, the irregular,
polydisperse rhodium nanoclusters shown in Figure 2 were
grown and isolated under the conditions labeled with the asterisk
(
/) in Figure 6, conditions that are within the zero order in
catalystsand hence MTLsregime. Rather clearly, then, these
stirring rate, TEM, and catalyst dependent vs nondependent
kinetic data demonstrate that hydrogen gas-to-solution MTL
conditions can drastically affect nanocluster particle size and
morphology.
8
a
TEM Evidence Ruling Out Mechanisms That Involve a
Time-Dependent Interconversion between Near-Monodis-
perse and Polydisperse Nanoclusters. Three key pieces of
evidence presented so far are the following: (i) that hydrogen
MTL conditions can directly affect particle size and morphology,
The key results are that the resulting nanoclusters have the
same dispersity, regardless of whether they are isolated at
≈
100% or at e20% precatalyst conversions and regardless if
(ii) that iridium and rhodium nanoclusters grown under reaction-
they are Rh or Ir nanoclusters, so long as the conditions were
MTL or non-MTL throughout the synthesis. For example, when
Rh(0) nanoclusters grown under MTL conditions are isolated
at e20% conversion of the precatalyst 2, the resulting distribu-
tion of diameters is broad (ranging from ca. 20 to 80 Å), and
rate-limiting conditions are near-monodisperse, and (iii) that
iridium and rhodium nanoclusters grown under hydrogen MTL
conditions are polydisperse. This evidence, plus the literature
evidence for the generality of diffusive agglomeration of
nanoclusters,5 is fully consistent with two parallel, kinetically
competing reactions19 leading to near-monodisperse vs poly-
,18
2
1
bimodal (centering around 37 and 48 Å) and thus polydis-
perse, Figure C, Supporting Information. However, when the
Rh nanocluster synthesis is carried out under reaction-rate
20
disperse nanoclusters, Scheme 2.
Note that one prediction of this scheme is that the ratio of
polydisperse to monodisperse nanoclusters should increase with
time if the reaction happened to be in a regime where both
processes are kinetically competitive. Alternatively, a second
prediction is that if the reaction is in the non-MTL regime (i.e.,
where generally only autocatalytic growth and no diffusive
22
limiting, non-MTL conditions, TEM shows that the Rh(0)
nanoclusters are near-monodisperse, with an average diameter
(20) (a) We note that a Nature paper has appeared offering evidence
that aggregation is “universal” in the fractal sense. Also noteworthy here is
that we previously cited our one, preliminary piece of evidence for a parallel
step of diffusive agglomeration (Step V, Scheme 1) in competition with
autocatalytic nanocluster growth (Step IV, Scheme 1), specifically our initial
observation of nanoclusters that were “less spherical [in] shape” when grown
8a
agglomeration has been seen ), then only monodisperse nano-
6a
(18) Lin, M. Y.; Lindsay, H. M.; Weitz, D. A.; Ball R. C.; Klein, R.;
under lower H2 pressure conditions in iridium nanocluster syntheses. But,
as we noted previously, the proposed agglomeration step was, at that time,
“the only step in the minimum mechanism in [Scheme 2 of that report]
that is on less solid grounds”.
Meakin, P. Universality in Colloid Aggregation Nature 1989, 339, 360.
Note that by “universality” these authors are referring to the literature of
fractals and, hence, mean “independent of the details of the particular
system” or “independent of the specific details of how the particular particles
are formed”. (b) For a review titled Fractal Aggregates, see Meakin, P.
AdV. Colloid Interface Sci. 1988, 28, 249, which covers many of the
proposed mechanisms, computer simulations, and some of the experimental
evidence on colloid aggregation phenomenon.
(21) (a) A bimodal distribution under MTL conditions, and at e20%
conversion, means that nucleation is initiated at two, rather different times,
just as expected when H2 is not uniformly present throughout the solution
under MTL conditions. However, given that isolation of rhodium nano-
clusters at 100% conVersion under identical conditions yields smaller (40
( 14 Å), polydisperse nanoclusters (as compared to the bimodal distribution
centered about 37 and 48 Å obtained at e20% conversion), this explanation
requires that the nanoclusters decrease in size and become closer to unimodal
at some later time through, presumably, Ostwald ripening processes.
(19) (a) Moore, J. W.; Pearson, R. G. Kinetics and Mechanism, 3rd ed.;
John Wiley & Sons: New York, 1981; pp 299-304. (b) Espenson, J. H.
Chemical Kinetics and Reaction Mechanisms; McGraw-Hill: New York,
1
981; pp 57-59.

Nanocluster Formation Synthetic, Kinetic, and Mechanistic Studies
J. Am. Chem. Soc., Vol. 120, No. 37, 1998 9551
Table 1. Iridium and Rhodium Nanocluster Dispersities as a Function of Reaction Conditions (non-MTL vs MTL) and Nanocluster Isolation
a
Times
iridium
rhodium
e20% conversion
g95% conversion
e20% conversion
g95% conversion
non-MTL conditions
MTL conditions
near-monodisperse
near-monodisperse
(20 ( 3 Å)
bulk Ir(0) metal
near-monodisperse
(34 ( 5 Å)
polydisperse; bimodal
near-monodisperse
(40 ( 6 Å)
polydisperse (40 ( 14 Å)
(20 ( 3 Å)
a
The data show that, regardless of their individual isolation times, nanoclusters grown under non-MTL conditions are near-monodisperse while
nanoclusters grown under MTL conditions are polydisperse. A statistical analysis for bimodality was performed on particle size histograms which
appeared to be bimodal (i.e., the e20% conversion Rh example, Supporting Information, Figure C).
genation. To demonstrate that the Rh(0) nanoclusters exhibit
catalytic activity once isolated as a solid and redispersed in the
presence of fresh cyclohexene, a catalytic cyclohexene hydro-
genation reaction was performed beginning with ca. 2.5 mg of
isolated Rh(0) nanoclusters dissolved in 2.5 mL of acetone and
0
.5 mL of cyclohexene. The results show that the olefin
hydrogenation reaction proceeds, as expected, without an
2
3
induction period. Indeed, we have found, and will report
2
4
elsewhere, conditions under which polyoxoanion-stabilized
Rh(0) nanoclusters generated from 2 exhibit g150 000 total
turnovers of cyclohexene hydrogenation.
Summary and Conclusions
The highlights of the present work can be summarized as
follows:
Figure 7. Co-plot of hydrogen pressure (squares; left axis) and
cyclooctane production by GLC (diamonds, right axis: 0.09 equiv, 2.0
h; 0.10 equiv, 3.0 h; 0.23 equiv, 4.0 h; 0.50 equiv, 4.75 h; 0.63 equiv,
(1) Hydrogen gas-to-solution MTL has been recognized as a
5
.5 h; 0.98 equiv, 8.75 h; 1.0 equiv 9.5 h) versus time in the concomitant
factor that must be controlled in the synthesis of transition metal
nanoclusters under H2. This is especially true for the more
catalytically active second-row transition metals such as Rh, or
if the nanoclusters are formed in the presence of reducible
substrates such as olefins. Since H2 is one of the most common
reducing agents for the synthesis of nanoclusters, this finding
should be of broader significance and impact.
hydrogenation of cyclohexene and formation of near-monodisperse
Rh(0) nanoclusters. (Conditions: 1.2 mM 2, 1.65 M cyclohexene, 49
(
1 psig H
linear H pressure loss, due to cyclohexene hydrogenation. The
hydrogenation rate, -d[H ]/dt, may be reliably calculated from the slope
2
, 3 °C.) A ca. 2.5 h induction period is seen prior to the
2
2
8
a
of the linear portion of the line. Point 1 shows the first time at which
the nanoclusters were isolated (e0.2 equiv cyclooctane produced)
whereas point 2 shows the last time at which the nanoclusters were
isolated (g0.95 equiv cyclooctane produced).
(2) Several lines of evidence [stirring rate, catalyst dependent
or nondependent kinetic regimes, and TEM evidence, including
a statistical analysis for bimodal distributions] were obtained
for the existence of hydrogen MTL and non-MTL kinetic
regimes, depending upon the exact conditions. These lines of
evidence support a mechanism with two parallel, competing,
nanocluster growth steps: autocatalytic surface growth, leading
to near-monodisperse nanoclusters, and diffusive agglomeration,
leading to polydisperse nanoclusters. Hydrogen MTL effects
also negatively influence the nanocluster nucleation process,
leading to bimodal distributions of nanoclusters.
(3) The above mechanistic insights allowed, in turn, the design
of conditions leading to the synthesis of the desired, near-
monodisperse, 40 ( 6 Å Rh(0) nanoclusters. These Rh(0)
nanoclusters are of interest as only the second example of
polyoxoanion-stabilized transition metal nanoclusters.
of 34 ( 5 Å (Supporting Information, Figure D). These results
rule out mechanisms that involve a time-dependent intercon-
version between mono- and poly-disperse nanoclusters; the
results are, however, fully consistent with and supportive of
the anticipated parallel growth pathways mechanism shown in
Scheme 2.
Note that, under H2 MTL conditions, one might expect
nucleation to occur intermittently, thereby leading to polydis-
perse nanoclusters, if not to bimodal distributions. Hence, the
observation of a statistically significant bimodal distribution of
polydisperse nanoclusters (Figure C, Supporting Information)
is good evidence for H2 MTL effects on the nucleation step as
well as on the subsequent competing autocatalytic surface
growth, vs diffusive agglomeration, steps.
The availability, now, of both Rh(0) and Ir(0) polyoxoanion-
Control Experiment Demonstrating the Catalytic Activity
of the Isolated Rh(0) Nanoclusters for Cyclohexene Hydro-
+
and Bu4N -stabilized nanoclusters is by no means trivial, since
these nanoclusters are unusual among known transition metal
nanoclusters in that they undergo extensive catalytic turnovers
in solution, without the need for additional stabilization provided
by either polymer protectants or by affixing them to solid
(22) Sometimes TEM images of nanoclusters isolated at early, incom-
plete, or later reaction times show nanoclusters somewhat clumped together
rather than completely and individually dispersed, even though a closer
examination of the nanoclusters at higher magnification reveals that
individual nanoclusters themselves are not fused. This seems to be an artifact
of the deposition/TEM visualization process, so that one must be careful to
distinguish individual but “clumped” particles from fused, and thus truly
agglomerated, particles. This is perhaps why some investigators will sonicate
or otherwise redisperse their particles prior to their TEM visualization, but
we chose to not do this to avoid introducing any new artifacts that such a
process might include. We note here that fractal aggregate type “clumps”
are the expected result of agglomeration, for example Figure I of the
Supporting Information, which looks distinctly fractal-like.
2
b
supports. In future publications we will report on the scale-
(23) Aiken, J. D., III; Finke, R. G. Polyoxoanion- and Tetrabutyl-
ammonium-Stabilized Near-Monodisperse, 40 ( 6 Å Rh(0)∼1500 to Rh(0)∼3700
Nanoclusters: Synthesis, Characterization and Hydrogenation Catalysis,
manuscript in preparation.
(24) Aiken, J. D., III; Finke, R. G. Polyoxoanion- and Tetrabutyl-
ammonium-Stabilized Transition Metal Nanoclusters: Unprecedented Cata-
lytic Lifetime in Solution. Manuscript in preparation.

9552 J. Am. Chem. Soc., Vol. 120, No. 37, 1998
Aiken and Finke
up, isolation details, catalytic reactions, and the other properties
of the polyoxoanion- and Bu4N -stabilized Rh(0) nanoclusters.
control experiment, that is, using a solution of the rhodium-containing
nanocluster precursor complex 2, shows that it, too, does not form
nanoclusters under TEM conditions which were identical to those used
to examine the Rh(0) nanoclusters (Supporting Information, Figure E).
+
23
Experimental Section
Nanocluster samples were analyzed by energy-dispersive X-ray
analysis (EDX) on the JEOL TEM with a Kevex Super 8000, running
Materials. Cyclohexene (Aldrich, 99+%, stabilized with 0.01% 2,6-
di-tert-butyl-4-methylphenol) was distilled from Na under argon
immediately prior to use (g99% purity by GLC) and stored in a
Vacuum Atmospheres drybox. Acetone (Burdick and Jackson, 0.26%
H O) was transferred into the drybox and purged for g30 min with
2
nitrogen using a glass bubbler. The source, purity, and water content
Quantex Version 6.0, coupled to a Quantum thin window detector
2
(
mylar window; 30 mm detector area with a resolution of 116 eV at
1
.49 keV). Electron beam spot sizes (typically between instrument
setting of 2 and 5) were adjusted so that events occurred with a deadtime
of e50% at between 1000 and 2000 counts/s. The instrument was
calibrated to the K lines of Al and Cu; peak assignments are based on
emission energy and peak intensity.
of the acetone are known to be crucial to obtaining reproducible kinetic
9
rates. Toluene (anhydrous, 99.8%, packed under N
2
, Aldrich) was
degassed via 3 freeze-pump-thaw cycles and stored in the drybox.
Hydrogen gas (General Air, 99.5%) used on the hydrogenation
apparatus, vide infra, was prepurified by passage through a moisture
trap (Scott Specialty Gases, Baxter Scientific), an oxygen trap (R&D
Separations type OT3, Baxter Scientific), and an indicating oxygen trap
Gas-liquid chromatography (GLC) was performed on a Hewlett-
Packard 5890 Series II GC with a FID detector equipped with a 30 m
(
0.25 mm i.d.) Econo-cap Carbowax column (Alltech) and coupled to
a Hewlett-Packard 3395 integrator. Parameters were as follows: initial
temperature, 50 °C; initial time, 3 min; ramp, 10 °C/min; final
temperature, 160 °C; final time, 16 min; injector port temperature, 180
C; detector temperature, 200 °C; injection volume, 4 µL.
Hydrogenation Experiments (“Standard Conditions”) and Cal-
(
R&D Separations, Hewlett-Packard).
The polyoxoanion, [(n-C N]
of 1 and 2, was synthesized according to our more recent procedure.
The air-sensitive polyoxoanion-supported organometallic precatalysts
(n-C N] Na [(1,5-COD)Ir‚P 62], 1, and [(n-C N]
Na [(1,5-COD)Rh‚P 62], 2, were synthesized and characterized
H )
4 9 4
9 2 3
P W15Nb O62, used in the synthesis
°
7a
culation of Reaction Rates. Hydrogenation reactions were performed
[
4
9
H )
4
5
3
2
W
15Nb
3
O
4 9
H
)
4
5
-
9
using our previously described pressurized hydrogenation apparatus
3
2 3
W15Nb O
(
Supporting Information, Figure F), except that it was equipped with
25
as previously reported and stored in a drybox. The precatalyst 2 used
an improved Pyrex water-jacketed reaction flask (200 mL; supplier,
Andrews Glass Company). All reaction mixtures were prepared in a
nitrogen-filled drybox. In each experiment, the precursors 1 or 2 were
weighed into a 1 dram vial and dissolved in 2.5 mL of acetone (added
via a 10 mL gas-tight syringe) to yield a clear, bright-yellow solution.
Cyclohexene (0.5 mL, 1.65 M) was added and the resultant clear,
homogeneous solution was transferred via polyethylene pipet into a
fresh 22 × 175 mm borosilicate culture tube containing a 5/16 in. ×
for the data reported herein was a batch that exhibited the highest
catalytic activity observed, -d[H
2
2 2
]/dt ) 4.7 mmol H /h at 49 psig H
and 3 °C. This particular batch of 2 was found to contain 0.20%, or
-
0
.14 equiv, Br by elemental analysis. However, as detailed in the
Supporting Information, control experiments show that this trace amount
-
of Br has a negligible effect on the observed rates. (We do, however,
recommend that one thoroughly wash the polyoxoanion complexes as
7a
previously detailed, thereby removing any trace amounts of remaining
5/8 in Teflon-coated magnetic stir bar. The culture tube was then sealed
-
Br .)
inside of a Fischer-Porter pressure bottle that is part of the hydrogena-
tion apparatus (Figure F, Supporting Information); the Fischer-Porter
bottle was brought outside of the drybox, placed inside a constant
temperature circulating water bath (Neslab Escal, (0.1 °C), and attached
to the hydrogenation line via Swagelock TFE-sealed Quick-Connects.
Analytical Procedures. Unless otherwise reported all reaction
solutions were prepared under oxygen- and moisture-free conditions
with a Vacuum Atmospheres drybox (<1 ppm O
monitored by a Vacuum Atmospheres O -level monitor). Stirring rates
were determined in revolutions per minute (rpm) with a Pocket-Tach
2
as continuously
2
(Before attachment of the Fischer-Porter bottle to the hydrogenation
1
00 (Monarch Instrument).
line, the line itself was evacuated for at least 1 h to remove any trace
oxygen and water present, and then refilled with purified hydrogen.)
The Fischer-Porter bottle was then purged 15 times with hydrogen
Specimens were examined by transmission electron microscopy
(TEM) either at CSU, using a JEOL TEM 2000 EX-II, or at the
University of Oregon, using a Philips CM-12 with a 70 µm lens, the
(15 s per purge, with stirring) and stirred vigorously for 2 min, and
6
latter as previously described in detail. Both microscopes were
then t ) 0 was set. Hydrogen pressure in the Fischer-Porter bottle
was then monitored with an Omega Model PX621 pressure transducer
attached to an IBM PC via an Omega WB35 A/D converter. Pressure
vs time data were processed with Microsoft Excel.
operated at an accelerating voltage of 100 keV. Samples examined
by TEM at CSU were prepared with type A (300 mesh) Formvar and
carbon coated copper grids (supplier, Ted Pella). Grids were gently
suspended in chloroform for about 30 s immediately prior to use to
remove the Formvar coating and to expose a fresh carbon surface. One
drop of the nanocluster solution (at a concentration of ca. 1 mg/mL, in
either acetone or acetonitrile, vide infra) was placed on the carbon-
coated grid with a gas-tight syringe and allowed to air-dry. Samples
were examined at magnifications between 100 and 400K, and in at
least three different places on the sample grid to ensure that the images
seen were representative of the sample as a whole. Reported
magnifications are those of the original negative. Particle-size distribu-
tions were determined once the original negative had been digitally
scanned into Adobe Photoshop and expanded to >20 in. × 25 in. for
easier size measurement and resolution. Previously reported control
Reaction rates were calculated from the rate of hydrogen pressure
loss (-d [H
of the linear portion of the H
shown in Figure 7. As demonstrated elsewhere,8a this slope correlates
linearly with the bimolecular rate constant k , obtained by quantitative
curve-fitting of the hydrogen pressure vs time data to the appropriate
2
]/dt, expressed in mmol H
2
/h) as determined by the slope
2
uptake curve, an example of which is
2
8
a
nucleation plus autocatalytic growth kinetic expressions.
Nanocluster Sample Isolation and Preparation for TEM Analysis.
The Fischer-Porter bottle containing the reaction solution was discon-
nected from the hydrogenation line via the Quick-Connects and taken
back into the drybox, the hydrogen pressure was released, and the turbid
nanocluster solution was poured into a 1 dram vial which was capped
with a plastic snap-cap. The turbid solution was then allowed to stand
undisturbed in the drybox for g10 h, after which time the nanoclusters
had settled to form a black precipitate. The supernatant was carefully
removed via a polyethylene pipet and discarded. While still in a drybox,
the black nanocluster precipitate (now in e0.5 mL of solution) was
then dried under vacuum overnight at room temperature to yield a black
powder. A TEM sample was prepared by redissolving all of this black
powder in 3 mL of fresh acetone or acetonitrile to yield a clear amber,
homogeneous solution (no bulk metal was visible by the naked eye at
any time). A drop of this solution was then dispersed on a chloroform-
cleaned TEM grid.
6
a
experiments show (i) that the iridium nanocluster samples are not
perturbed by changing either the electron beam accelerating voltage
(40 or 100 keV) or by varying sample exposure times under the beam
(
seconds vs minutes) and (ii) that the iridium precursor 1 does not form
6
a
nanoclusters under the TEM beam. A repeat herein of this latter
(
25) (1,5-COD)M‚P2W15Nb3O628-: (a) Nomiya, K.; Pohl, M.; Mizuno,
N.; Lyon, D. K.; Finke, R. G. In Inorganic Syntheses; Cowley, A. H., Ed.;
John Wiley and Sons: New York, 1997; Vol. 31, pp 186-201. (b) Pohl,
M.; Lyon, D. K.; Mizuno, N.; Nomiya, K.; Finke, R. G. Inorg. Chem. 1995,
3
4, 1413. (c) Pohl, M.; Finke, R. G. Organometallics 1993, 12, 1453-
1457. (d) Finke, R. G.; Lyon, D. K.; Nomiya, K.; Sur, S.; Mizuno, N. Inorg.
Chem. 1990, 29, 1784-1787.

Nanocluster Formation Synthetic, Kinetic, and Mechanistic Studies
J. Am. Chem. Soc., Vol. 120, No. 37, 1998 9553
Quantitative Curve-Fitting of Hydrogen Pressure Ws Time Data.
(e0.1 mL) of the reaction solution were drawn with an 18 in. needle
attached to a gas-tight syringe that was inserted through the ball valve
at the top of the Fischer-Porter bottle, Figure F, Supporting Informa-
Curve-fitting of the hydrogen pressure (or, equivalently, cyclohexene
8
a
8a
concentration ) vs time data was performed as previously described.
Briefly, a nonlinear regression subroutine (RLIN), using a modified
Levenberg-Macquard algorithm and available in the IMSL Statistical
Library, was used (running on an IBM/AIX workstation). As a control,
the curve-fitting program was first tested on a previously published
tion, all while under a continuous flow of H
2
. Aliquots were removed
. Once
only after the needle and syringe were purged thoroughly with H
2
the aliquot was taken, the Fischer-Porter bottle was purged an
additional five times (15 s per purge) with hydrogen gas and resealed
8
a
data set and faithfully reproduced k
1
and k
2
values from that data set.
2
at 49 psig H . The production of cyclooctane vs time is summarized
Curve-fits of two rhodium nanocluster formation runs under non-MTL
conditions are supplied as Figure B of the Supporting Information.
Statistical Analysis for Bimodality in Particle Size vs Frequency
Histograms. A program was written in S-plus that utilizes a bootstrap
databased simulation method for determining the percent confidence
of bimodality in particle size vs frequency distributions which contain
more than one mode (i.e. Figure C of the Supporting Information).
This was applied to particle size histograms which appeared to contain
more than one mode. Full details are given in the Supporting
Information.
in Figure 7.
Hydrogenation of Cyclohexene Beginning with 1 and 2 as a
Function of the Stirring Rate. The effect of stirring rate on
nanocluster formation was measured in a series of otherwise Standard
Conditions experiments at a constant precursor concentration (20.0 (
2.0 mg, 3.6 ( 0.3 × 10 mmol, 1.2 ( 0.1 mM) by using both the
iridium-containing precursor 1 (at 40 psig H and 22 °C, non-MTL
conditions) and the rhodium-containing precursor 2 (at 49 psig H and
3 °C, non-MTL conditions). Hydrogenation rates vs stirring rate (rpm)
beginning with 1 are summarized in Figure 5A. Hydrogenation rates
vs stirring rate beginning with 2 are summarized in Figure 5B.
Mechanistic TEM Studies. Isolation and TEM Examination of
Iridium and Rhodium Nanoclusters at e20% Reaction Times. To
determine if the degree of polydispersity of the iridium and rhodium
nanoclusters changed throughout the reaction under both MTL and non-
MTL conditions, a series of nanoclusters were grown and isolated at
e20% of their growth cycle (see point 1 in Figure 7, the point at which
hydrogen consumption was complete) and examined by TEM for their
degree of dispersity. In all experiments the initial precatalyst concen-
tration was constant at 1.2 ( 0.1 mM. When cyclohexene hydrogena-
tion was complete (point 1, Figure 7), the reaction was quenched by
stopping the stirring, releasing the hydrogen pressure, and transferring
the Fischer-Porter bottle and its turbid-yellow reaction solution back
into the drybox. Next, the nanoclusters were isolated as detailed in
the Nanocluster Sample Isolation section, then redissolved in 3 mL of
2
6
-
3
2
2
Formation of Polydisperse Rh(0) Nanoclusters by Hydrogenation
of Cyclohexene under Hydrogen Mass-Transfer-Limiting Condi-
tions. A cyclohexene hydrogenation was performed under conditions
identical to those previously reported (i.e., the Standard Conditions6a,9
)
except with the rhodium precursor 2, [(n-C
62]. A clear, bright-yellow solution of 2 (20.5 mg, 3.7 ×
mmol) was made by dissolving the complex in 2.5 mL of acetone
4 9 4 5 3
H ) N] Na [(1,5-COD)Rh‚
2 3
P W15Nb O
-
3
10
and 0.5 mL of cyclohexene. A Standard Conditions hydrogenation
reaction was carried out as described above at 22 °C and with an initial
pressure of 40 psig H (3.7 atm). When the reaction was complete,
2
the Fischer-Porter bottle was disconnected from the hydrogenation
apparatus and transferred into the drybox and the hydrogen pressure
was released. The rhodium nanoclusters were isolated as described in
the Nanocluster Sample Isolation section and examined by TEM. The
results are presented in Figure 2. A histogram of the observed 16 to
1
16 Å frequency vs particle diameter is shown in Figure G of the
3
CH CN for TEM analysis. The degree of dispersity in each case is
Supporting Information.
With the goal of elucidating the hydrogenation rate vs [2] dependence
curve, eleven additional cyclohexene hydrogenation reactions at 22.0
shown in Table 1. The corresponding TEM images and particle size
histograms are shown in the Supporting Information: Figure C, Rh,
MTL conditions, 44 ( 11 Å, a bimodal distribution; Figure D, Rh,
non-MTL, 34 ( 5 Å; Figure I, Ir, non-MTL, 20 ( 3 Å.
(
0.1 °C were performed at either 40 ( 1 or 49 ( 1 psig H
of experiments); the concentration of 2 was varied between 0.3 and
.3 mM at each pressure. The results of this series of 12 experiments
2
(two sets
Investigation of the Origins of the 5-Fold Rh(0) Nanocluster
Precursor-Batch-Dependent Variability in the Observed Catalytic
Rates. Throughout the course of other work,23 we found that 2 of 7
of the batches of Rh precursor 2 showed catalytic hydrogenation rates
which were ca. 5-fold faster that the slowest batch under identical
reaction conditions. (The two fastest batches of Rh(0) nanoclusters
were used for the present studies of MTL effects, except for the control
experiment below, since these studies were done first, before our other
Rh(0) nanocluster work,23 and since the early two batches of Rh(0)
nanoclusters were the 5-fold faster ones.) However, within a batch of
Rh precursor 2, the rates were the reproducible 15-20% we have seen
2
are summarized in Figure 6.
Formation of Near-Monodisperse 40 ( 6 Å Rh(0) Nanoclusters
by Hydrogenation of Cyclohexene under Reaction-Rate-Limiting
(
6
Non-MTL) Conditions. The synthesis of near-monodisperse 40 (
Å Rh(0) nanoclusters was carried out in the exact manner described
above for the Standard Conditions except for the following changes
designed to avoid hydrogen mass-transfer limitations: (a) the amount
-
3
of the precatalyst 2 was lowered to 13.3 mg (2.4 × 10 mmol, 0.79
mM) to slow the reaction and to lower the demand for H ; (b) the
reaction temperature was reduced to 3.0 ( 0.1 °C to slow the reaction;
and (c) the initial hydrogen pressure was increased to 49 ( 1 psig H
After complete conversion of the precursor 2 into nanoclusters (g8 h;
confirmed by the release of 1.0 equiv of cyclooctane by GLC, vide
infra), the Fischer-Porter bottle containing the turbid, yellow-brown
reaction solution was taken into the drybox and the remaining hydrogen
pressure was released. The clusters were isolated as described above
and then examined by TEM (Figure 3) and EDX (Supporting Informa-
tion, Figure H). Two additional TEM images of this same sample at
higher (400K) and lower (100K) magnifications are shown in Figure
A of the Supporting Information.
2
6
a,8
previously, even inter-batch, for the Ir precursor.
As a source of
this variability we considered trace oxygen,27 trace Br- (see the
Supporting Information), or possibly trace water remaining in the
2
.
hydrogenation apparatus itself. However, data presented herein and
-
elsewhere demonstrated that this variability is not due to O
2
, or Br ,
or the apparatus; hence, we conclude it is intrinsic to the precatalyst 2
itself (which is known to exist as g2 isomers2 ), and thus most likely
variability in the nanocluster nucleation (and possibly also growth) steps.
Note, however, as the control experiment below demonstrates, this
variability does not affect the general results obtained herein under
MTL conditions. Hence, this variability as a function of the batch of
precatalyst 2 is addressed in more detail elsewhere.23
5b
GLC Determination of the Evolved Equivalents of Cyclooctane
Ws Time during Rh(0) Nanocluster Formation. Toluene (2 µL; as a
GLC internal standard) was added with a 5 µL gas-tight syringe to a
Control Experiment Demonstrating That H
2
MTL Effects Are
Found in the Slower Batch of Rh(0) Nanoclusters. As a control, an
-3
clear, bright-yellow solution of 2 (20.1 mg, 3.6 × 10 mmol, 1.2 mM)
(27) Structural studies or catalytic effects of oxygen in small metal
dissolved in 2.5 mL of acetone and 0.5 mL of cyclohexene.
A
particle systems: (a) Harada, M.; Asakura, K.; Ueki, Y.; Toshima, N. J.
Phys. Chem. 1992, 96, 9730-9738. (b) Reetz, M. T.; Quaiser, S. A.; Winter,
M.; Becker, J. A.; Sch a¨ fer, R.; Stimming, U.; Marmann, M.; Vogel, R.;
Konno, T. Angew. Chem., Int. Ed. Engl. 1996, 35, 2092. (c) Kolb, W.;
Quaiser, S.; Winter, M.; Reetz, M. T. Chem. Mater. 1996, 8, 1889-1894.
(d) Rothe, J.; Pollmann, J.; Franke, R.; Hormes, J.; B o¨ nnemann, H.; Brijoux,
W.; Siepen, K.; Richter, J. Fresenius J. Anal. Chem. 1996, 355, 372-274.
(e) See also p 371 of ref 1i.
hydrogenation reaction was then started by using an initial hydrogen
pressure of 49 psig and at a temperature of 3 °C (non-MTL conditions).
At prechosen times, the hydrogen pressure was released and aliquots
(26) Efron, B.; Tibshirani, R. J. An Introduction to the Bootstrap;
Monograph on Statistics and Applied Probability 57; Chapman & Hall:
New York, 1993; pp 227-234.

9554 J. Am. Chem. Soc., Vol. 120, No. 37, 1998
Aiken and Finke
experiment was done to show that Rh(0) nanoclusters, formed from a
batch of 2 that gave the slowest observed cyclohexene hydrogenation
at e20% precursor conversion under MTL conditions; Figure
D, TEM image and associated histogram of near-monodisperse
rhodium nanocluster diameters following isolation at e20%
precursor conversion under non-MTL conditions; Figure E,
TEM image (at 300K magnification) of a solution of 2 showing
that nanoclusters are not formed from 2 under TEM conditions;
Figure F, detailed schematic of the hydrogenation apparatus used
in the present study; Figure G, histogram of rhodium particle
diameters corresponding to Figure 2; Figure H, EDX spectros-
copy on the Rh(0) nanoclusters shown in Figure 2; Figure I,
TEM image and associated histogram of near-monodisperse
Ir(0)∼300 nanocluster diameters following isolation at e20%
precursor conversion under non-MTL conditions; Figure J, TEM
and associated histogram of irregular shaped, polydisperse 35
catalytic rate (-d[H
2 2 2
]/dt ) 1.0 mmol H /h at 49 psig H and 3 °C),
still show MTL effects. Specifically, under nonstirred (0 rpm) reaction
conditions, the results, Figure J, Supporting Information, show the
presence of both aggregated Rh(0) metal and polydisperse and irregular
shaped 35 ( 11 Å Rh(0) nanoclusters (ca. 31% polydispersity). This
control experiment confirms that the general conclusions of this studys
2
that is, the observation of H MTL effectssare not dependent upon
the particular batch of Rh(0) precursor 2 employed.
Acknowledgment. We wish to thank Dr. Eric Schabtach at
the University of Oregon for obtaining the TEM image shown
in Figure 2. We also thank Dr. John Chandler at the Colorado
State University Electron Microscopy and Digital Imaging
Center for helpful assistance with TEM images obtained at CSU
and Prof. Phil Chapman and Mr. Shannon Ritter in the Statistics
Department at CSU for performing the bimodal statistical
analysis cited in the text. Financial support was provided by
the Department of Energy, Chemical Sciences Division, Office
of Basic Research, via grant DOE FG06-089ER13998.
(
11 Å Rh(0) nanoclusters grown under MTL conditions (0
rpm stirring) for a slower batch of Rh(0) nanoclusters; Figure
K, amplified TEM image (200K magnification) of bulk Ir(0)
metal produced at 0 rpm stirring and shown at 50K magnifica-
tion in Figure 4; Figure L, observation of, and explanation for,
the non-hydrogen pressure dependence when under non-MTL
conditions; statistical analysis for bimodality in particle size vs
frequency histograms; control experiments describing the rate-
enhancing effects of added tetrabutylammonium bromide on the
observed catalytic hydrogenation activity beginning with 2 (15
pages, print/PDF). See any current masthead page for ordering
information and Web access instructions.
Supporting Information Available: Figure A, two ad-
ditional TEM images (at 100 and 400K magnifications) of the
near-monodisperse Rh(0) nanocluster sample shown in Figure
3
; Figure B, curve-fitting of the hydrogen pressure vs time data
for the synthesis of rhodium nanoclusters grown under non-
MTL conditions; Figure C, TEM image and associated histo-
gram of polydisperse rhodium nanoclusters following isolation
JA9719485