Purification of endohedral trimetallic nitride fullerenes in a single, facile step

Article abstract of DOI:10.1021/ja055089t

A major hurdle hampering the development of fullerenes, endohedral metallofullerenes, and nanotubes has been the difficulty of obtaining high purity samples. Soots prepared in the usual manner via a Kraetschmer-Huffman electric-arc generator consist of mixtures of insoluble carbonaceous materials and soluble fullerenes: C₆₀, C₇₀, C₇₆, C ₇₈, C₈₄, etc. When metals are introduced as endohedral species the complexity of the resultant soot is even greater because of the presence of multiple isomers of both the empty fullerenes and the endohedral metallofullerenes. Here, for the first time, we report that lanthanide trimetallic nitride endohedral metallofullerenes, A₃N@C₈₀ (A = lanthanide atom, e.g., Er, Gd, Ho, Lu, Sc, Tb, Tm, Y), can be obtained in pure form directly from as-prepared soots in a single facile step by taking advantage of their extraordinary kinetic chemical stability with respect to the other fullerenes in Diels-Alder reactions with a cyclopentadiene-functionalized resin. We show that careful control of conditions (stoichiometry, time, temperature) allows separation of fullerenes with different cage sizes, as well as isomeric species. Furthermore, the Diels-Alder reaction is thermally reversible, and we demonstrated that the bound empty-cage fullerenes and classical endohedral metallofullerenes can be recovered by displacement with maleic anhydride.

Full text of DOI:10.1021/ja055089t

Published on Web 10/28/2005
Purification of Endohedral Trimetallic Nitride Fullerenes in a
Single, Facile Step
Zhongxin Ge,† James C. Duchamp,‡ Ting Cai,† Harry W. Gibson,*,† and
Harry C. Dorn*,†
Department of Chemistry, Virginia Polytechnic Institute and State UniVersity,
Blacksburg, Virginia 24060, and Department of Chemistry, Emory and Henry College,
Emory, Virginia 24327
Received July 27, 2005; E-mail: hwgibson@vt.edu; hdorn@vt.edu
Abstract: A major hurdle hampering the development of fullerenes, endohedral metallofullerenes, and
nanotubes has been the difficulty of obtaining high purity samples. Soots prepared in the usual manner via
a Kra¨tschmer-Huffman electric-arc generator consist of mixtures of insoluble carbonaceous materials and
soluble fullerenes: C60, C70, C76, C78, C84, etc. When metals are introduced as endohedral species the
complexity of the resultant soot is even greater because of the presence of multiple isomers of both the
empty fullerenes and the endohedral metallofullerenes. Here, for the first time, we report that lanthanide
trimetallic nitride endohedral metallofullerenes, A3N@C80 (A ) lanthanide atom, e.g., Er, Gd, Ho, Lu, Sc,
Tb, Tm, Y), can be obtained in pure form directly from as-prepared soots in a single facile step by taking
advantage of their extraordinary kinetic chemical stability with respect to the other fullerenes in Diels-
Alder reactions with a cyclopentadiene-functionalized resin. We show that careful control of conditions
(stoichiometry, time, temperature) allows separation of fullerenes with different cage sizes, as well as isomeric
species. Furthermore, the Diels-Alder reaction is thermally reversible, and we demonstrated that the bound
empty-cage fullerenes and classical endohedral metallofullerenes can be recovered by displacement with
maleic anhydride.
Introduction
selective chemical reactivity. These TNT EMFs are currently
being seriously investigated for a number of diagnostic (MRI
The scarcity of purified, homogeneous samples has hampered
wider study and application of carbonaceous nanomaterials
(fullerenes, endohedral metallofullerenes, and nanotubes). Pro-
duction of single-walled nanotubes usually leads to distributions
in terms of diameter, chirality, and length, and formation of
multiwalled nanotubes.1 The usual Kra¨tschmer-Huffman (K-
H) electric-arc generator produces a mixture of fullerenes
including C60, C70, C76, C78, C84, etc.2 The complexity is even
greater for endohedral metallofullerenes because of the presence
of multiple isomers of both the fullerenes and metallofullerenes.3
Until now, the only reliable methods for obtaining purified
fullerene and endohedral metallofullerene (EMF) samples
involved extensive, repetitive, time-consuming chromatographic
procedures that required large volumes of solvent.
and X-ray contrast agents) and therapeutic medical applica-
tions.7,8
Our purification protocol is based on the kinetic stability of
the TNT EMFs relative to empty-cage fullerenes and classical
EMFs, such as Ax@C2y (x ) 1-3, y ) 30-50), which are
concomitantly produced and also of interest in medical
imaging.8-10 Computational and experimental results demon-
strate significant charge transfer (6 electrons) to the icosahedral
(Ih symmetrical) cage of TNT EMFs, e.g., [Sc3N]+6@[C80]-6.11,12
(5) Stevenson, S.; Fowler, P. W.; Heine, T.; Duchamp, J. C.; Rice, G.; Glass,
T.; Harich, K.; Hajdu, E.; Bible, R.; Dorn, H. C. Nature 2000, 408, 427-
428.
(6) Olmstead, M. M.; de Bettencourt-Dias, A.; Duchamp, J. C.; Stevenson, S.;
Marciu, D.; Dorn, H. C.; Balch, A. L. Angew. Chem., Int. Ed. 2001, 40,
1223-1225.
(7) (a) Iezzi, E. B.; Duchamp, J. C.; Fletcher, K. R.; Glass, T. E.; Dorn, H. C.
Nano Lett. 2002, 2, 1187-1190. (b) Stevenson S.; Lee H. M.; Olmstead
M. M.; Kozikowski C.; Stevenson P.; Balch A. L. Chem. Eur. J. 2002, 8,
4528-4535. (c) Fatouros, P. P.; Corwin, F. D.; Chen, Z.-J.; Broaddus, W.
C.; Tatum, J. L.; Ge, Z.; Gibson, H. W.; Kile, J. L.; Leonard, A. P.;
Duchamp, J. C.; Dorn, H. C. Radiology 2005, 236, in press.
(8) Gorman, J. Sci. News 2002, 162, 26.
(9) Kato, H.; Kanazawa, Y.; Okumura, M.; Taninaka, A.; Yokawa, T.;
Shinohara, H. J. Am. Chem. Soc. 2003, 125, 4391-4397.
(10) (a) Bolskar, R. D.; Benedetto, A. F.; Husebo, L. O.; Price, R. E.; Jackson,
E. F.; Wallace, S.; Wilson, L. J.; Alford, J. M. J. Am. Chem. Soc. 2003,
125, 5471-5478. (b) Toth, E.; Bolskar, R. D.; Borel, A.; Gonzalez, G.;
Helm, L.; Merbach, A. E.; Sitharaman, B.; Wilson, L. J. J. Am. Chem.
Soc. 2005, 127, 799-805.
(11) (a) Kobayashi, K.; Nagase, S. Chem. Phys. Lett. 1996, 262, 227-232. (b)
Kobayashi, K.; Nagase, S.; Akasaka, T. Chem. Phys. Lett. 1996, 261, 502-
506. (c) Kobayashi, K.; Sano, Y.; Nagase, S. J. Comput. Chem. 2001, 22,
1353-1358.
In this paper, we report that pure trimetallic nitride template
(TNT) EMFs, A3N@C80 (A ) lanthanide atom),4-6 can be
obtained from crude soots in a single, facile step based on
† Virginia Polytechnic Institute and State University.
‡ Emory and Henry College.
(1) See Acc. Chem. Res. 2002, 35, Special Issue on Nanotubes. Haddon, R.
C., Ed.; Niyogi, S.; Hamon, M. A.; Hu, H.; Zhao, B.; Bhowmik, P.; Sen,
R.; Itkis, M. E.; Haddon, R. C. Acc. Chem. Res. 2002, 35, 1105-1113.
(2) Kra¨tschmer, W.; Lamb, L. D.; Fostiropoulos, K.; Huffman, D. R. Nature
1990, 347, 354-358.
(3) Shinohara, H. In Fullerenes: Chemistry, Physics and Technology; Kadish,
K. M., Ruoff, R. S., Eds.; Wiley: New York, 2000, pp 357-393.
(4) Stevenson, S.; Rice, G.; Glass, T.; Harich, K.; Cromer, F.; Jordan, M. R.;
Craft, J.; Hadju, E.; Bible, R.; Olmstead, M. M.; Maltra, K.; Fisher, A. J.;
Balch, A. L.; Dorn, H. C. Nature 1999, 401, 55-57.
9
16292
J. AM. CHEM. SOC. 2005, 127, 16292-16298
10.1021/ja055089t CCC: $30.25 © 2005 American Chemical Society

Endohedral Trimetallic Nitride Fullerenes
A R T I C L E S
Scheme 1. Synthesis of Cyclopentadiene (Cp)-Functionalized
Table 1. Minimum Bond Resonance Energies of Reported
Fullerenes (from ref 12)
Resin 2
predicted
species
symmetry
BRE/
â
reactivity ranka
C60
C70
C76
C78
C78
C78
C82
C82
C82
C82
C84
C84
C84
C84
C84
C84
C84
C84
C84
C84
Ih
+0.0822
+0.0519
+0.0074
-0.0098
-0.0440
-0.0887
+0.0096
-0.0314
-0.0405
-0.0607
+0.0819
+0.0771
+0.0351
+0.0108
-0.0702
-0.0254
-0.0316
-0.0501
-0.0564
-0.0574
+0.155
+0.1620
+0.1931
+0.1652
+0.0743
+0.0296
-0.0212
-0.0736
+0.0839
+0.0430
-0.0514
26
22
16
15
9
D5h
D2
reported and rationalized in terms of redox potentials.19 Obvi-
ously, the nature of the endohedral species plays a role in the
reactivity and even regiochemistry, as recently demonstrated
for A ) Y vs A ) Sc of A3N@C8020 and the recent production
of a diadduct from reaction of Gd3N@C80 with the o-quin-
odimethane derived from 6,7-dimethoxyisochromanaone in less
time (1 h @ 214 °C)21 than for production of the monoadduct
(predominantly) of the Sc3N@C80 analogue (24 h @ 214 °C).15
Nonetheless, although not an absolute predictor of reactivities,
the min BREs provide a useful framework for predicting and
understanding the relative reactivities of fullerenes, EMFs and
TNT EMFs.
D3
C2V-1
C2V-2
Cs
1
17
12
10
4
25
24
20
18
3
13
11
8
6
5
28
29
31
30
23
19
14
2
Cs-2
C2-1
C2-2
D2-1
Td
†
†
D6h
D2d
D3d
D2-2
Cs
C2
D2d-2
D2-3
D3
-6
C68
Results and Discussion
-6
C78
D3h
Ih
D5h
-6
C80
A. Preparation of Reactive Resin for Purification of TNT
EMFs and Other Fullerenes. On the basis of the observed
low reactivity of the TNT EMFs, we reasoned that a suitably
designed insoluble resin system could remove more reactive
fullerenes from mixtures produced in the K-H process and
allow the purification of the TNT EMFs in a facile manner.
Fullerenes act as dienophiles in Diels-Alder cycloaddition
reactions.15,16,21-23 We, therefore, prepared a cyclopentadiene-
functionalized styrene-divinylbenzene resin (2) from the chlo-
romethylated (Merrifield) resin 1 in a fashion similar to the
procedure reported by Rotello et al.24 (Scheme 1). However,
we used a lower level of cross-linking (1% vs 2%) and a higher
loading of chloromethylated sites (3.5-4.5 vs 1.04 meq of Cl/
g) for synthesis of resin 2.
Figure 1 illustrates the concept; as the solution of the fullerene
mixture passes through a column packed with the cyclopenta-
dienyl resin 2 the species are bound to the resin by Diels-Alder
cycloaddition in order of their reactivities, empty cages (green
balls) first, then classical EMFs (pink balls) and under appropri-
ate time/temperature conditions the TNT EMF (blue ball) is
left unbound and can be isolated in a pure state in the eluent
solution.
-6
C80
-4
b
C82
Cs-1
-4
b
C82
C2-1
-4
b
C82
C2-2
-4
b
C82
Cs-2
-4
C84
D2d
Cs
27
21
7
-4
C84
-4
C84
C2V-4
a Highest reactivity ) 1. b Not yet an isolated/identified isomer.
Aihara has calculated the minimum bond resonance energies
(min BRE) for numerous fullerenes and metallofullerenes;
negative BREs represent kinetic instability and higher chemi-
cally reactivity. This leads to the prediction of exceptional
kinetic and thermodynamic stability of the icosahedral [C80]-6
cage of A3N@C80 that is significantly greater than neutral
empty-cages C60 and C70, which are the major products of the
K-H process (Table 1).12 Note that the min BRE value for the
icosahedral [C80]-6 cage is the highest on the list, meaning it
should be the least reactive in general. Consistent with this
prediction, Sc3N@C80 is stable in air to temperatures of 600-
650 K13 and in nanotube “peapods” in vacuo to temperatures
in excess of 1300 K.14 An isochromanone derivative of
Sc3N@C80 can be prepared only at elevated temperatures,15 in
comparison with the analogous lower temperature reaction with
the empty-cage fullerene C60.16 Similarly, the Prato reaction of
Sc3N@C80 produces a monoadduct17 under conditions which
yield multi-adducts with empty cage fullerenes18a and classical
endohedral metallofullerenes.18b Recently, the low reactivity of
Sc3N@C80 in comparison to La2@C80 (a singlet ground state
in which the cage has been assigned11,12 as the Ih hexaanion,
C80-6) in photochemical disilarane cycloadditions has been
B. Purification of Synthetic Binary Mixtures of C60 with
A3N@C80. As a first test of the concept, 1:1 (mass:mass, 1.5-
2.0:1.0 mol:mol) binary mixtures of C60, the predominant
(17) (a) Cai, T.; Ge, Z.; Iezzi, E. B.; Glass, T. E.; Harich, K.; Gibson, H. W.;
Dorn, H. C. Chem. Commun. 2005, 3594-3596. (b) Cardona, C. M.;
Kitaygorodkiy, A.; Ortiz, A.; Herranz, M. A.; Echegoyen, L. J. Org. Chem.
2005, 70, 5092-5097.
(18) (a) Tagmatarchis, N.; Prato, M. Synlett 2003, 6, 768-779. Kordatos, K.;
Bosi, S.; Da Ros, T.; Zambon, A.; Lucchini, V.; Prato, M. J. Org. Chem.
2001, 66, 2802-2808. (b) Lu, X.; He, X.; Feng, L.; Shi, Z.; Gu, Z.
Tetrahedron 2004, 60, 3713-3716.
(19) Iiduka, Y.; Ikenaga, O.; Sakkuraba, A.; Wakahara, T.; Tsuchiya, T.; Maeda,
Y.; Nakahodo, T.; Akasaka, T.; Kako, M.; Mizorogi, N.; Nagase, S. J.
Am. Chem. Soc. 2005, 127, 9956-9957.
(12) (a) Aihara, J. Phys. Chem., Chem. Phys. 2001, 3, 1427-1431. Aihara, J.
Chem. Phys. Lett. 2001, 343, 465-469. (b) Aihara, J. J. Phys. Chem. A
2002, 106, 11371-11374.
(13) Dorn, H. C. et al. In Electronic Properties of NoVel Materials-Molecular
Nanostructures; Kuzmany, H., et al., Eds.; Am. Inst. of Phys. 2000, pp
135-141.
(14) Russo, R.; Smith, B. W.; Satishkumar, B. C.; Luzzi, D. E.; Dorn, H. C.
Mater. Res. Soc. Symp. Proc. 2001, 675, W1.3.1.-W1.3.7.35.
(15) (a) Iezzi, E. B.; Duchamp, J. C.; Harich, K.; Glass, T. E.; Lee, H. M.;
Olmstead, M. M.; Balch, A. L.; Dorn, H. C. J. Am. Chem. Soc. 2002, 124,
524-525. (b) Lee, H. M.; Olmstead, M. M.; Iezzi, E.; Duchamp, J. C.;
Dorn, H. C.; Balch, A. L. J. Am. Chem. Soc. 2002, 124, 3494-3495.
(16) Belik, P.; Guegel, A.; Kraus, A.; Spickermann, J.; Enkelmann, V.; Frank,
G.; Mu¨llen, K. AdV. Mater. 1993, 5, 854-856.
(20) Cardona, C. M.; Kitaygorodskiy, A.; Echegoyen, L. J. Am. Chem. Soc.
2005, 127, 10448-10453.
(21) Stevenson, S.; Stephen, R. R.; Amos, T. M.; Cadorette, V. R.; Reid, J. E.;
Phillips, J. P. J. Am. Chem. Soc. 2005, 127, 12776-12777.
(22) Maeda, Y.; Miyashita, J.; Hasegawa, T.; Wakahara, T.; Tsuchiya, T.;
Nakahodo, T.; Akasaka, T.; Mizorogi, N.; Kobayashi, K.; Nagase, S.; Kato,
T.; Ban, N.; Nakajima, H.; Watanabe, Y. J. Am. Chem. Soc. 2005, 127,
12190-12191.
(23) Sliwa, W. Fullerene Sci. Techn. 1997, 5, 1133-1175. Nakamura, Y.;
Okawa, K.; Nishimura, J. Bull. Chem. Soc. Jpn. 2003, 76, 865-882.
(24) Guhr, K. I.; Greaves, M. D.; Rotello, V. M. J. Am. Chem. Soc. 1994, 116,
5997-5998.
9
J. AM. CHEM. SOC. VOL. 127, NO. 46, 2005 16293

A R T I C L E S
Ge et al.
Figure 3. HPLC (PYE column, toluene, 3mL/min, detection at 390 nm)
traces of (a) the toluene extract of “scandium soot”, (b) the room-temperature
toluene eluent from application of the extract to a column packed with
cyclopentadiene-functionalized resin; the peak at 27.9 min corresponds to
Sc3N@C80; there are minor peaks due to Sc3N@C68 (16 min) and Sc3N@C78
(27 min) and at retention times of 5-10 min due to unreactive polycyclic
aromatic hydrocarbons from the soot; see further discussion below and (c)
the eluent from the column after elution (b) upon heating at 85 °C in the
presence of maleic anhydride.
Figure 1. Separation of TNT EMFs from empty cage fullerenes and
classical EMFs by selective cycloadditions with resin 2.
Ho3N@C80; see the Supporting Information for equivalently
good results obtained for the other two mixtures.
C. Purification of Sc3N@C80 Soot Extract. Next we applied
the purification protocol to the toluene extract of the soot
produced from a K-H preparation of Sc3N@C80. The soot from
the arc process with Sc2O3-loaded graphite rods was subjected
to Soxhlet extraction with toluene. Figure 3a shows the HPLC
trace of the toluene extract containing both empty cage fullerenes
and Sc3N@C80. Figure 4 contains the CI mass spectrum of the
extract, demonstrating the presence of empty cage fullerenes
C60, C70, C76, C78, C82, C84, C86 (m/z 1032), C88 (m/z 1056), C90
(m/z 1080), C92 (m/z 1104), C94 (m/z 1128) and C96 (m/z 1152),
along with TNT EMFs Sc3N@C68,25 Sc3N@C78,6 and
Sc3N@C80.4,5 The concentrated extract was applied to a column
packed with resin 2 and the column was flushed slowly with
toluene. As demonstrated above with the simple binary mixtures
and in accord with Figure 1, the empty-cage fullerenes present
in the extract react with the cyclopentadiene-functionalized resin
by Diels-Alder cycloaddition, allowing the less reactive TNT
EMF, Sc3N@C80 (see Table 1), to pass through (essentially
completely unreacted as demonstrated below), thus affording
relatively pure Sc3N@C80 (Figure 3b); see also the discussion
of purification of the raw “scandium soot” below.
Figure 2. HPLC traces (Buckyclutcher column, toluene, 1 mL/min,
detection at 390 nm) of (a) a synthetic mixture of C60 and Ho3N@C80 and
(b) pure Ho3N@C80 obtained by application of the mixture to a column of
resin 2 and flushing with toluene for 4 h at room temperature. (c) MALDI
TOF mass spectrum (9-nitroanthracene matrix, negative ion) of the pure
Ho3N@C80.
The Diels-Alder reaction is thermally reversible and Rotello
et al. investigated recovery of C60 from similar cyclopentadiene-
functionalized supports upon heating.24,26 And Saunders et al.
utilized the reversibility for isomer release after anthracene
Diels-Alder adducts had been purified.27
material formed in the K-H process in all cases, and the TNT
EMFs were examined. Toluene solutions of the mixtures were
applied to columns packed with resin 2 and toluene; the column
was flushed with toluene and the resultant solution was analyzed
by HPLC. The C60 was cleanly removed in every case, yielding
pure lanthanide TNT EMFs, Sc3N@C80, Gd3N@C80, and
Ho3N@C80. Figure 2 shows the results for purification of
Here, we demonstrated recovery of the empty-cage fullerenes
trapped by resin 2; Figure 3c shows the HPLC trace of the eluent
(25) Olmstead, M. M.; Lee, H. M.; Duchamp, J. C.; Stevenson, S.; Marciu, D.;
Dorn, H. C.; Balch, A. L. Angew. Chem., Int. Ed. 2003, 42, 900-903.
(26) Nie, B.; Rotello, V. M. J. Org. Chem. 1996, 61, 1870-1871.
9
16294 J. AM. CHEM. SOC. VOL. 127, NO. 46, 2005

Endohedral Trimetallic Nitride Fullerenes
A R T I C L E S
Figure 4. Chemical ionization mass spectrum (negative ion) of the toluene extract of soot made from graphite rods packed with Sc2O3; as indicated, empty
cage fullerenes and TNT EMFs are present.
from the column upon heating at 85 °C in the presence of maleic
anhydride. The overall recovery of empty cage fullerenes was
60% relative to the extract initially applied to the resin. Note
that no or very little Sc3N@C80 had been bound to the resin, in
accord with its greatly diminished reactivity relative to the other
species in the complex mixture; however, see the discussion
below. It is noteworthy that exposure of the extract to resin 2
and the recovery process resulted in a change in the ratio of
C78 isomers as shown by comparison of Figure 3, parts a and
c, in accord with the differing reactivities of the three known28,29
isomers predicted by the calculations shown in Table 1.
No attempt was made in the present work to optimize the
recovery process, but it should be noted that Rotello et al.
achieved 48% recovery of bound C60 using their cyclopenta-
diene-functionalized resin24 and 94% of bound C60 and 41% of
bound C70 using an analogous cyclopentadiene-functionalized
silica.26
D. Purification of Raw Soots. We then proceeded to study
the raw soots themselves. As found with the extracts described
above, the expected high degree of selectivity was observed in
reactions with resin 2.
1. Lu3N@C80. The raw soot from K-H arc treatment of
graphite rods packed with Lu2O3 was applied as a powder to
the top of a column of resin 2 and toluene; the column was
flushed with toluene. As shown in Figure 5, mass spectral and
HPLC analyses confirmed the identity and high purity of the
effluent, the previously reported Lu3N@C80.7a,7b The high degree
of selectivity is illustrated by the wide range of empty-cage
fullerenes (C60, C70, C76, C78, C84, etc.) and classical EMFs
[Lu2@C82 and Lu2@C84, presumed by analogy with scandium
Figure 5. Purification of the raw soot to produce pure Lu3N@C80. (a)
HPLC (Buckyclutcher column, toluene, 1 mL/min, detection at 390 nm)
trace of the toluene extract of the soot produced in the presence of Lu2O3;
the peaks correspond in approximate order of elution left to right: C60,
C70, C76, C78, C84 (the latter three with g2 identified isomers), classical
EMFs Lu2@C82, Lu3N@C80, and Lu2@C84 (g2 isomers). (b) chemical
ionization mass spectrum (negative ion) and (c) HPLC trace of eluted, pure
Lu3N@C80.
-4
analogues to possess tetraanionic cages, C82
(unknown
symmetry)11b,30 and C84 (D2d, Cs, and C2V);31 see Table 1]
present in the soot that readily react with, and are trapped by,
the cyclopentadiene-functionalized resin.
-4
2. Sc3N@C80. The collected raw soot from arc-vaporization
of graphite/Sc2O3 was placed (as a powder) on top of a column
packed with the cyclopentadiene-functionalized resin 2/toluene
and toluene was passed through the column. As was the case
with the “scandium soot” extract described above (Figures 3
and 4) Sc3N@C80 was isolated in a relatively high state of purity,
(27) Wang, G.-W.; Saunders: M.; Khong, A.; Cross, R. J. J. Am. Chem. Soc.
2000, 122, 3216-3217.
(28) Ettl, R.; Chao, I.; Diederich, F.; Whetten, R. Nature 1991, 353, 149-153.
(29) Diederich, F.; Whetten, R.; Thilgen, C.; Ettl, R.; Chao, I.; Alvarez, M. M.
Science 1991, 254, 1768-1770.
(30) Wang, C.-R.; Inakuma, M.; Shinohara, H. Chem. Phys. Lett. 1999, 300,
379-384.
9
J. AM. CHEM. SOC. VOL. 127, NO. 46, 2005 16295

A R T I C L E S
Ge et al.
Figure 7. HPLC traces (Buckyclutcher column, toluene, 1 mL/min,
detection at 390 nm) trace of (a) the toluene extract (Soxhlet, 4 h) of a soot
mixture prepared from graphite rods packed with Gd2O3 and (b) purified
Gd3N@C80 from toluene elution of the soot through a column packed with
cyclopentadiene-functionalized resin 2 over a 20 h period; (c) the chemical
ionization mass spectrum (negative ion) of the eluent noted in (b), identifying
the pure TNT EMF and (d) the theoretical spectrum for Gd3N@C80 predicted
on the basis of the natural isotopes of Gd, namely Gd152 (0.2%), Gd154
(2.1%), Gd155 (14.8%), Gd156 (20.6%), Gd157 (15.7%), Gd158 (24.8%) and
Gd160 (21.8%).
Figure 6. (a) HPLC (Buckyclutcher column, toluene, 1 mL/min, detection
at 390 nm) trace of the toluene extract of “scandium soot”; (b) HPLC trace
of the room-temperature toluene eluent from application of the raw soot to
a column packed with cyclopentadiene-functionalized resin 2; the three peaks
noted correspond to TNT EMFs with different cage sizes and the peak with
the asterisk is contamination introduced by the injector. (c) the chemical
ionization mass spectrum (negative ion) of the eluent noted in (b) showing
the presence of the three TNT EMFs.
For the three known31 classical EMF Sc2@C84 isomers (Cs,
C2V, and D2d) produced in this process the min BREs for the
resultant [C84]-4 cages (Table 1) are indicative of higher
reactivities than the TNT EMFs, in agreement with their higher
experimentally observed reactivities with, and resultant trapping
by, the cyclopentadiene-functionalized resin. Furthermore, it is
noteworthy that the D5h isomer of Sc3N@C80, which is observed
as a longer retention time shoulder in HPLC analyses of samples
purified chromatographically,32 is not observed in these experi-
ments (Figure 6). Furthermore, careful examination of Figure
3c indicates that this isomer reacted with the resin and was
released upon treatment with maleic anhydride at 85 °C. This
process comprises a separation of these two isomers.33
3. Gd3N@C80. The raw soot from K-H treatment of graphite
rods packed with Gd2O3 was analogously treated by elution
through a column of resin 2 with toluene at room temperature.
Figure 7a displays the HPLC trace of the toluene extract of the
soot, indicating the presence of the usual mixture of empty cage
fullerenes, classical EMFs and TNT EMFs. The sample resulting
containing the homologous TNT EMFs with smaller cage sizes,
namely Sc3N@C68 and Sc3N@C78, in minor proportions (3.0
and 3.7%, respectively, by area from the chromatogram), as
demonstrated by HPLC and mass spectrometric analyses (see
Figure 6).
Assuming that, as in Sc3N@C80, six electrons are transferred
to the cages, corresponding to [Sc3N]+6@[C68]-6 (D3) and
[Sc3N]+6@[C78]-6 (D3h), the low chemical reactivity of these
smaller TNT EMFs observed in the present work is consistent
with the high kinetic stability deduced theoretically (Table
1).11,12 Note, however, that these TNT EMF homologues are
expected to be more reactive than Sc3N@C80 (Table 1).12
Therefore, the extent of removal of these homologues is
dependent on the details of the kinetic purification process. [Note
that these homologues are not observed for the Lu3N analogues,
because the larger trimetallic nitride cluster is too large to fit
the smaller C68 and C78 cages; hence, Lu3N@C80 is isolated as
a pure single isomer (Figure 5).]
(32) (a) Duchamp, J. C.; Demortier, A.; Fletcher, K. R.; Dorn, D.; Iezzi, E. B.;
Glass, T.; Dorn, H. C. Chem. Phys. Lett. 2003, 375, 655-659. (b) Krause,
M.; Dunsch, L. ChemPhysChem 2004, 5, 1445-1449.
(33) Echegoyen and co-workers have recently reported an electrochemical
method for separation of the Ih and D5h isomers of Sc3N@C80: Elliott, B.;
Yu, L.; Echegoyen, L. J. Am. Chem. Soc. 2005, 127, 10885-10888. The
present protocol offers an alternative and complementary means of preparing
the two pure isomers from the raw soot.
(31) Inakuma, M.; Yamamoto, E.; Kai, T.; Wang, C.-R.; Tomiyama, T.;
Shinohara, H.; Dennis, T. J. S.; Hulman, M.; Krause, M.; Kuzmany, H. J.
Phys. Chem. B. 2000, 104, 5072-5077.
9
16296 J. AM. CHEM. SOC. VOL. 127, NO. 46, 2005

Endohedral Trimetallic Nitride Fullerenes
A R T I C L E S
similar low min BREs, and are quite reactive; however, two
other isomers have been isolated but not identified. Notably,
however, a few isomers possess higher kinetic stability (e.g.,
D2-1 and Td, Table 1), thereby potentially allowing them to be
isolated in this manner; the unbound C84 noted in Figure 8, parts
g and h, is presumed to be a mixture of these less reactive
isomers.
Conclusions
High purity lanthanide trimetallic nitride endohedral metal-
lofullerenes, A3N@C80 (A ) lanthanide atom, e.g., Gd, Ho, Lu,
Sc), can be obtained directly from as-prepared soots in a single
facile step by taking advantage of their extraordinary kinetic
chemical stability with respect to the Diels-Alder reaction with
cyclopentadiene-functionalized resin. The present results indicate
that it is possible to separate isomers of these TNT EMFs as
well as various empty cage fullerenes. The thermal reversibility
of the Diels-Alder reaction can, in principle, provide selective
elution of fullerenes and endohedral metallofullerenes from the
resin column as a function of temperature, flow rate, stoichi-
ometry, and addition of competitive dienophiles. Moreover,
complete removal of the bonded fullerenes would allow multiple
cycles with the same resin, providing a purification approach
that is scalable and consistent with Green Chemistry practices.
We are currently pursuing these refinements of this protocol.
This new purification protocol has two advantages relative
to the use of the HPLC method. First, the yields are higher by
about 2-fold. This is due to the use of a lower temperature
extraction solvent with reduced exposure to oxygen, hence
decreased oxidative losses. In addition, repeated losses in the
chromatographic process from multiple injections are avoided.
A second major advantage is the time for obtention of pure
endohedral trimetallic nitride fullerene samples, which is reduced
from ca. 30 days to one (1) day.
Figure 8. Kinetic stability of empty cage fullerenes with resin 2 at room
temperature in toluene. (a) HPLC (PBB column, toluene, 2 mL/min,
detection at 390 nm) trace of the initial empty cage fullerenes. HPLC traces
of the solution 0.5 h after addition of (b) the first portion (4 mg), (c) the
second portion (4 mg), (d) the third portion (10 mg), (e) the fourth portion
(4 mg), (f) the fifth portion (4 mg), (g) the sixth portion (4 mg) and (h) the
seventh portion (2 mg) of resin 2.
from flushing the soot through the column was shown by HPLC
(Figure 7b) and mass spectrometry (Figure 7c) to be pure
Gd3N@C80. The experimental mass spectrum agrees well with
the theoretically predicted one (Figure 7d).
E. Purification of Empty Fullerenes. The kinetic stability
approach outlined here is not limited to purification of the TNT
class of endohedral metallofullerenes, but can be extended to
other fullerenes and presumably functionalized nanotubes. To
demonstrate this, a small amount of resin 2 was added every
30 min to a stirred toluene solution of empty-cage fullerenes
(C60, C70, C76, C78 and C84) at room temperature. The composi-
tion of the solution was monitored by HPLC. The results are
shown in Figure 8. The removal of C76 and C78 is faster than
removal of C60, C70 and C84. C76 and C78 disappear first and
then C60 and C70. When the other four empty cage fullerenes
have been removed, some C84 remains. These results are
consistent with the very low min BREs (high reactivities) for
the known isomers of C76 (D2),28 C78 (C2V),29,34 C78 (C2V)29,34
and C78 (D3)29,34 relative to C60 (Ih) and C70 (D5h); see Table 1.
Most of the eight known34-40 neutral C84 isomers also have
Experimental Section
Materials and Methods. Graphite rods (99.9995% C, 6.15 mm ×
152 mm) and graphite powder (99.9995% C) were obtained from Alfa
Aesar. The graphite rods were drilled longitudinally to provide a 5/32”
(4 mm) hole, which was packed with the metal oxide, FexN and graphite
powder (2-15 µm); the latter three components were mixed with a
mortar and pestle and the graphite rod was packed with the resultant
mixture using a cotton tip applicator, followed by the blunt end of the
5/32” drill bit. Scandium (III) oxide (Sc2O3, 99.999%), lutetium (III)
oxide (Lu2O3, 99.995%), holmium (III) oxide (Ho2O3, 99.999%) and
gadolinium (III) oxide (Gd2O3, 99.995%) were obtained from Stanford
Materials Corporation. Iron nitride (99.9%, FexN, x ) 2-4) was
obtained from Cerac Specialty Inorganic Chemicals. Merrifield’s resin
(chloromethylated styrene-1% divinylbenzene copolymer, 3.5-4.5 meq
of Cl/g) and sodium cyclopentadienylide (2.0 M solution in tetrahy-
drofuran) were obtained from Sigma-Aldrich. C60 was obtained from
Mer Corporation and used as received. A semipreparative Buckyclutcher
column (10 x 250 mm, Regis Chemical Company) was used in high-
pressure liquid chromatography (HPLC) for both analysis and purifica-
tion. HPLC was also carried out on a â-(1-pyrenyl)ethyl silica (PYE)
column (10 × 250 mm, Alltech Associates) and a pentabromobenzy-
loxypropyl silica (PBB) column (4.6 × 250 mm, Alltech Associates).
Cyclopentadiene-Functionalized Resin 2. A suspension of chloro-
methylated styrene-divinylbenzene copolymer (1 g) in toluene (200 mL)
was cooled to -20 °C. To the stirred suspension, sodium cyclopenta-
dienylide (16 mL of a 2.0 M solution in tetrahydrofuran, 32 mmol, ∼8
equiv) was added dropwise. Then the mixture was stirred for 2 h at 20
°C and filtered; the beads were washed with toluene (600 mL) to give
the dark brown cyclopentadiene-functionalized resin 2.
(34) Kikuchi, K.; Nakahara, N.; Wakabayashi, T.; Suzuki, S.; Shiromaru, H.;
Miyake, Y.; Saito, K.; Ikemoto, I.; Kainosho, M.; Achiba, Y. Nature 1992,
357, 142-145.
(35) Manolopoulos, D. E.; Fowler, P. W.; Taylor, R.; Kroto, H. W.; Walton, D.
R. M. Faraday Trans. 1992, 88, 3117-3118.
(36) Dennis, T. J. S.; Kai, T.; Tomiyama, T., Shinohara, H. Chem. Commun.
1998, 619-620.
(37) Tagmatarchis, N.; Avent, A. G.; Prassides, K.; Dennis, T. J. S.; Shinohara,
H. Chem. Commun. 1999, 1023-1024.
(38) Crassous, J.; Rivera, J.; Fender, N. S.; Shu, L.; Echegoyen, L.; Thilgen,
C.; Herrmann, A.; Diederich, F. Angew. Chem., Int. Ed. 1999, 38, 1613-
1617.
(39) Dennis, T. J. S.; Kai, T.; Asato, K.; Tomiyama, T.; Shinohara, H.; Yoshida,
T.; Kobayashi, Y.; Ishiwatari, H.; Miyake, Y.; Kikuchi, K.; Achiba, Y. J.
Phys. Chem. A 1999, 103, 8747-8752.
(40) Azamar-Barrios, J. A.; Dennis, T. J. S.; Sadhukan, S.; Shinohara, H.;
Scuseria, G. E.; Pe´nicaud, A. J. Phys. Chem. A 2001, 105, 4627-4632.
9
J. AM. CHEM. SOC. VOL. 127, NO. 46, 2005 16297

A R T I C L E S
Ge et al.
Sc3N@C80. Sc3N@C80was prepared by arc-vaporization of graphite
rods packed with Sc2O3 (0.73 g, 5.3 mmol), FexN (0.40 g) and graphite
powder (1.00 g, total carbon ∼2.5 g, 0.21 g at) in a K-H generator
under a dynamic N2 (20 mL/min)/He (∼900 mL/min) atmosphere (at
300 Torr total pressure). The graphite rods were typically baked at
∼1100 °C under N2 for ∼18 h just prior to arc-vaporization. A potential
difference of 31 V was applied between the ends of the rods and
maintained via electronic control. After each rod had been consumed
(∼45 min), the resulting soot was collected.
To study the reversibility of the covalent binding process, the column
was wrapped with heating tape and a digital thermometer was inserted
into the resin. At room temperature, two column volumes of toluene
saturated with maleic anhydride were flushed through the column; no
fullerenes were eluted. The column was then heated to 85 °C and kept
at this temperature overnight. The solution in the column turned red,
which indicated that the empty cage fullerenes had been released and
replaced by the more the reactive dienophile, maleic anhydride. The
column was flushed with toluene saturated with maleic anhydride at
85 °C over a 2 h period. The red solution was subsequently passed
through a silica gel column to remove a small amount of maleic
anhydride; removal of the solvent yielded 22.4 mg of recovered
fullerenes; total recovery from extract: 23.4 mg, 61% (Figure 3c).
Lu3N@C80. Lu3N@C80 was prepared from graphite rods packed with
Lu2O3 (0.85 g, 2.1 mmol) and FexN (0.33 g) as outlined above for the
Sc analogue.
Gd3N@C80. Gd3N@C80 was prepared from graphite rods packed with
Gd2O3 (0.80 g, 2.2 mmol) and FexN (0.33 g) as outlined above for the
Sc analogue.
Ho3N@C80. Ho3N@C80 was prepared from graphite rods packed with
Ho2O3 (0.80 g, 2.1 mmol) and FexN (0.33 g) as outlined above for the
Sc analogue.
Purification of Raw Lu3N@C80 Soot Using Resin 2. The recovered
soot (7.18 g) from two graphite rods packed with Lu2O3 was placed
(as a powdery solid) on a glass column (38 × 25 mm, d × h) packed
with the cyclopentadiene-functionalized resin 2 (40 g, ∼160 mmol of
reactive sites based on the starting chloromethylated resin capacity) in
toluene and toluene was passed through the column by gravity feed
over a 20 h period. Pure Lu3N@C80 (9.0 mg, 0.21% yield based on
Lu2O3 as the limiting reagent, 0.13% by mass of the soot) was isolated
(Figure 5).
Empty Cage Fullerenes. Empty cage fullerenes were prepared by
arc-vaporization of graphite rods in a K-H generator under a dynamic
He atmosphere (200 Torr). A potential difference of 31 V was applied
between the ends of the rods and maintained via electronic control.
After the rod had been consumed, the resulting soot was collected and
extracted with toluene. Purification by HPLC (PBB column, toluene)
afforded a mixture with similar amounts of C60, C70, C76, C78, and C84
(Figure 8a).
Purification of Sc3N@C80 Present in a Binary Mixture with C60.
A binary mixture of C60 (1.0 mg, 1.4 µmol) and Sc3N@C80 (1.0 mg,
0.90 µmol, purified as previously reported4,5 ) was applied as a toluene
solution to the top of a glass column [13 × 25 mm, diameter (d) ×
height (h)] packed with 4 g of resin 2 (∼16 mmol of reactive sites
based on the starting chloromethylated resin capacity) in toluene. The
column was flushed with toluene by gravity feed for 4 h. The resultant
solution was concentrated and analyzed by HPLC (see SI).
Purification of Gd3N@C80 Present in a Binary Mixture with C60.
A binary mixture of C60 (1.0 mg, 1.4 µmol) and Gd3N@C80 (1.0 mg,
0.69 µmol purified as previously reported)41 was applied as a toluene
solution to the top of a glass column (13 × 20 mm, d × h) packed
with 4 g of resin 2 (∼16 mmol of reactive sites based on the starting
chloromethylated resin capacity) in toluene. The column was flushed
with toluene by gravity feed for 4 h. The resultant solution was
concentrated and analyzed by HPLC (see SI).
Purification of Raw Sc3N@C80 Soot Using Resin 2. The recovered
soot (8.0 g) from two graphite rods packed with Sc2O3 was placed (as
a powdery solid) on a glass column (38 × 25 mm, d × h) packed with
the cyclopentadiene-functionalized resin 2 (40 g, ∼160 mmol of reactive
sites based on the starting chloromethylated resin capacity) in toluene
and toluene was passed through the column by gravity feed over a 20
h period. High purity Sc3N@C80 (11 mg, 0.19% yield based on C as
the limiting reagent, 0.14% by mass of the soot) was isolated (Figure
6).
Purification of Raw Gd3N@C80 Soot Using Resin 2. The recovered
soot (3.0 g) from one graphite rod packed with Gd2O3 was placed (as
a powdery solid) on a glass column (28 × 22 mm, d × h) packed with
the cyclopentadiene-functionalized resin 2 (20 g, ∼80 mmol of reactive
sites based on the starting chloromethylated resin capacity) in toluene
and toluene was passed through the column by gravity feed over a 20
h period. Pure Gd3N@C80 (0.30 mg, 0.014% yield based on Gd2O3 as
the limiting reagent, 0.01% by mass of the soot) was isolated (Figure
7).
Purification of Empty Cage Fullerenes Using Resin 2. To a stirred
solution of empty-cage fullerenes (C60, C70, C76, C78, and C84) (1.0 mg,
∼1.2 µmol total assuming equal masses) in 2 mL of toluene at room
temperature, resin 2 was incrementally added every 30 min in the
following sequence. The first and the second portions were 4 mg (∼16
µmol of reactive sites based on the starting Merrifield resin capacity
of 4.0 meq Cl/g) the third portion was 10 mg (∼40 µmol of reactive
sites based on the starting Merrifield resin capacity); the fourth, fifth,
and sixth portions were also 4 mg each; the seventh portion was 2 mg.
Total resin added: 32 mg (∼128 µmol of reactive sites based on the
starting Merrifield resin capacity of ∼4.0 meq Cl/g). Before every
addition of resin 2, the solution was monitored by HPLC (Figure 8).
Purification of Ho3N@C80 Present in a Binary Mixture with C60.
A binary mixture of C60 (1.0 mg, 1.4 µmol) and Ho3N@C80 (1.0 mg,
0.68 µmol purified as previously reported)42 was applied as a toluene
solution to the top of a glass column (13 × 20 mm d × h) packed with
4 g of resin 2 (∼16 mmol of reactive sites based on the starting
chloromethylated resin capacity) in toluene. The column was flushed
with toluene by gravity feed for 4 h. The resultant solution was
concentrated and analyzed by HPLC as shown in Figure 2.
Purification of Sc3N@C80 Soot Extract Using Resin 2 and
Recovery of other Fullerenes. The raw soot from three graphite rods
packed with Sc2O3 was collected and extracted in a Soxhlet device
with toluene for 12 h, yielding 38.3 mg of mixed fullerenes and
endohedral metallofullerenes after removing the solvent in vacuo; see
Figure 4 for the mass spectrum of this material. This extract was
redissolved in toluene and applied to a glass column (28 × 22 mm, d
× h) packed with ∼20 g of cyclopentadiene-functionalized resin 2
(∼180 mmol of reactive sites based on the starting chloromethylated
resin capacity) in toluene. Toluene was flushed through by gravity feed
(∼6 mL/h) during 48 h. Sc3N@C80 was obtained after removing the
solvent from the eluent (Figure 3b,c).
Acknowledgment. We thank NSF for support through awards
CHE-043700 (H.C.D.) and DMR-0507083 (H.C.D., H.W.G.).
Supporting Information Available: HPLC traces of binary
mixtures of C60 with Gd3N@C80 and Sc3N@C80 and the
resultant purified A3N@C80 species; ultraviolet-visible absorp-
tion spectra of Gd3N@C80, Ho3N@C80, Lu3N@C80 and
Sc3N@C80 purified using resin 2; complete ref 13. This material
is available free of charge via the Internet at http://pubs.acs.org.
(41) Stevenson, S.; Stevenson, J. P.; Reid, J. E.; Olmstead, M. M.; Rath, S. P.;
Balch, A. Chem. Commun. 2004, 2814-2815.
(42) Dunsch, L.; Krause, M.; Noack, J.; Georgi, P. J. Phys. Chem. Solids 2004,
65, 309-315.
JA055089T
9
16298 J. AM. CHEM. SOC. VOL. 127, NO. 46, 2005