Radical Rearrangements for the Chemical Vapor Deposition of Diamond

Article abstract of DOI:10.1021/jo971814+

A combination of chemical trapping and computations is used to determine the activation parameters for the interconversion of the 3-methylenebicyclo[3.3.1]nonan-7-yl (1), (3-noradamantyl)methyl (2), and 1-adamantyl (3) radicals. The three radicals model proposed intermediate surface radical structures in the chemical vapor deposition (CVD) of diamond on its 2 × 1 reconstructed [100] surface. The study finds that relatively low-level calculations previously applied to the problem of diamond growth are reliable, at least qualitatively.

Full text of DOI:10.1021/jo971814+

J . Org. Chem. 1998, 63, 4581-4586
4581
Ra d ica l Rea r r a n gem en ts for th e Ch em ica l Va p or Dep osition of
Dia m on d
Andreas M. Mueller and Peter Chen*
Laboratorium fu¨r Organische Chemie der Eidgeno¨ssischen, Technischen Hochschule,
CH-8092 Zu¨rich, Switzerland
Received October 1, 1997
A combination of chemical trapping and computations is used to determine the activation parameters
for the interconversion of the 3-methylenebicyclo[3.3.1]nonan-7-yl (1), (3-noradamantyl)methyl (2),
and 1-adamantyl (3) radicals. The three radicals model proposed intermediate surface radical
structures in the chemical vapor deposition (CVD) of diamond on its 2 × 1 reconstructed [100]
surface. The study finds that relatively low-level calculations previously applied to the problem of
diamond growth are reliable, at least qualitatively.
Sch em e 1. 5-Hexen yl Ra d ica l a n d th e Tw o Mod es
for Cycliza tion Usin g Ba ld w in ’s Design a tion s
In tr od u ction
The chemical vapor deposition (CVD) of diamond has
become an established technology for the production of
polycrystalline diamond for a variety of technical ap-
plications.1,2 Experimental studies on the effect of mac-
roscopic process variables, e.g., temperature, pressure,
added gases, flow rate, initiation method, etc., have
appeared in great number and contributed to the long,
incremental improvement in both the growth rate and
quality of CVD diamond. The chemical mechanism,
however, by which C1 and/or C2 units are incorporated
into a new layer of diamond lattice, has received less
scrutiny, with computational, and a few experimental,
studies appearing in recent years.3,4 Even those experi-
mental studies that have been done purport only to
identify the important gas-phase carbon-containing pre-
cursors to diamond on the surface.5 There are no
experiments pertaining to the structural transformation
on a microscopic level. One may suppose that a better-
than-incremental advance in the art will require at least
some mechanistic information on the surface reactions
of diamond.
CVD reactor by way of this rearrangement is shown in
Scheme 2, starting from a 2 × 1 reconstructed [100]
surface believed to be an important starting point for
growth.3 Notice that the carbon atom in the methyl
group external to the crystal is incorporated into the
framework.
The basic steps are as follows: (i) hydrogen abstraction
from a 2 × 1 reconstructed [100] surface by H•, (ii)
coupling of a methyl radical with the surface radical site,
(iii) abstraction of a hydrogen from the pendant methyl
group, (iv) ring-opening to generate the 5-hexenyl radical,
(v) ring-closure by radical addition to the other end of
the olefinic unit, and (vi) 1,5-hydrogen translocation. The
same basic rearrangement suffices to “shuffle” bridge
sites together to form the experimentally observed smooth
surfaces. Of these steps, only the 2 × 1 reconstruction
itself should require elevated temperatures. The remain-
ing steps are all well-known for simpler organic systems
in solution and should proceed rapidly at temperatures
substantially below these in typical CVD reactions.
Accordingly, while the gross kinetics of diamond growth
in CVD at 1200 K are probably not greatly affected by
differences in activation energies for the surface rear-
rangements, i.e., the surface rearrangements are not
rate-limiting,9 an accurate knowledge of the energetics
for the surface chemistry is nevertheless important as
one begins to design analogous processes that run at
much lower temperatures. Lower temperature operation
would be advantageous in the production of a technologi-
cally important material like diamond and provides the
motivation for the present work. Lastly, it should be
emphasized that, even if the proposed 5-hexenyl mech-
anism is operative in only some instances for CVD, the
simplicity of the chemistry makes this mechanism im-
portant to designed processes that will generate diamond
Of the several mechanisms proposed in the literature,
one based on the 5-hexenyl rearrangement appears most
plausible from the viewpoint of an organic chemist. The
5-hexenyl rearrangement, a “radical clock” reaction,6 has
been extensively investigated (Scheme 1). The proposed
mechanism8,9 for diamond growth from H• and CH3 in a
•
(1) For a review of properties and production of CVD diamond, see:
Yarbrough, W. A.; Messier, R. Science 1990, 247, 688.
(2) Derjaguin, B. V.; Fedoseev, D. V.; Lukyanovich, V. M.; Spitsyn,
B. V.; Ryabov, V. A.; Lavrentyev, A. V. J . Cryst. Growth 1968, 2, 380.
Eversole, W. G. U.S. Patent 3,030,188, 1961. Angus, J . C.; Will, H. A.;
Stanko, W. S. J . Appl. Phys. 1968, 39, 2915. Angus, J . C. U.S. Patent
3,630,677, 1971.
(3) For a review of terminology, surface morphology, and growth
mechanisms, see: Butler, J . E.; Woodin, R. L. Phil. Trans. R. Soc.
London A 1993, 342, 209.
(4) Harris, S. J .; Goodwin, D. G. J . Phys. Chem. 1993, 97, 23. Zhu,
M.; Hauge, R. H.; Margrave, J . L.; D’Evelyn, M. P. Proc. 3rd Int. Symp.
Diamond Mat. 1993, 138. Harris, S. J . Appl. Phys. Lett. 1990, 56, 2298.
(5) Lee, S. S.; Minsek, D. W.; Vestyck, D. J .; Chen, P. Science 1994,
263, 1596.
(6) Griller, D.; Ingold, K. U. Acc. Chem. Res. 1980, 13, 317.
(7) Baldwin, J . E. J . Chem. Soc., Chem. Commun. 1976, 734, 736,
738.
(8) Garrison, B. J .; Dawnkaski, E. J .; Srinivasta, D.; Brenner, D.
W. Science 1992, 255, 835.
(9) Shokov, S.; Weiner, B.; Frenklach, M. J . Phys. Chem. 1994, 98,
7073.
S0022-3263(97)01814-8 CCC: $15.00 © 1998 American Chemical Society
Published on Web 06/12/1998

4582 J . Org. Chem., Vol. 63, No. 14, 1998
Mueller and Chen
Sch em e 2. P r op osed Mech a n ism for Dia m on d Gr ow th on a 2 × 1 Recon str u cted [100] Su r fa ce w ith Atom ic
Hyd r ogen a n d Meth yl Ra d ica ls, Ad a p ted fr om Refs 8 a n d 9 (Descr ip tion s of Ea ch Step Ar e in th e Text)
Ta ble 1. Tr a p p in g P r od u ct Ra tios in th e Rea ction of 1
by non-CVD reaction sequences that mimic the high-
temperature chemistry in CVD.
w ith CCl4 a t Differ en t Tem p er a tu r es (Th e Ra d ica l
P r ecu r sor is Gen er a ted In Situ )
In the present report, we show a combination of
experimental and computational results on a set of
5-hexenyl radical rearrangements of cage hydrocarbons
that model the diamond [100] surface. By preparation
of the 3-methylenebicyclo[3.3.1]nonan-7-yl (1), (3-norada-
mantyl)methyl (2), and 1-adamantyl (3) radicals, com-
petitive trapping of the radicals as a function of temper-
ature, and molecular mechanics modeling of the corres-
ponding hydrocarbons, the relative enthalpies of the
three radicals and the transition states for their inter-
conversions are determined.
T (°C)
75
98
116
126
%4 ((0.5)
1.7a (1.7)
1.3 (1.5)
1.1 (1.4)
0.9 (1.4)
%5 ((1.0) 96.5c (96.5) 95.6 (96.2) 95.9 (95.9) 95.6 (95.6)
%6 ((0.9)
1.8b (1.8)
3.1 (2.3)
3.0 (2.7)
3.5 (3.0)
Average of 10 runs. bAverage of two runs. c Remainder of
trapped radical products. The numbers in parentheses are
predicted product yields from the kinetic simulation using the final
derived kinetic parameters (see Discussion).
a
The results are compared to published calculations of
the surface reactions on diamond, as well as ab initio
computations at a modest level, showing that, while the
rearrangements are eminently plausible as basic com-
ponents in diamond CVD chemistry, the level of theory
being currently brought to bear on the problem is
quantitatively in error, in particular overestimating
activation energies.
romethyl radicals are the only radical products observed
in the GC/MS traces. Similarly, from 1-adamantanol (the
direct precursor to 3), the only radical-derived products
were 6 and 7. The most important set of trapping studies
starts from methylenebicyclononanol (the direct precur-
sor to 1), from which all three desired products, 4-6, are
formed (at 75 °C), in a combined yield of 72%. Within
that total, the product distribution is shown in Table 1.
At temperatures much higher than the 126 °C run, e.g.,
135-170 °C, significant decomposition of the CCl4 solvent
occurred, which could perturb the product ratios. In
general, the measurement of reliable, quantitative prod-
uct ratios is complicated by a variety of experimental
difficulties. For example, the C10H15Cl compounds with
cage structures are surprisingly volatile, so care must be
exercised that the workup prior to analysis does not
differentially deplete any component by evaporation.
Also, one can conceive of a radical chain-induced isomer-
Resu lts
The trapping reactions of radicals 1-3 in CCl4, gener-
ated thermally from the “Barton” precursors prepared in
situ, produce 3-methylene-7-chlorobicyclo[3.3.1]nonane
(4), (3-noradamantyl)chloromethane (5), and 1-chloro-
adamantane (6). Two of the three sets of trapping results
are relatively uninformative, although they become
relevant in light of the third one: From noradamantyl-
methanol (the direct precursor to 2), the expected product
5 and the coupling product 7 of thiopyridinyl and trichlo-

Chemical Vapor Deposition of Diamond
J . Org. Chem., Vol. 63, No. 14, 1998 4583
Ta ble 2. P r od u ct Ra tios in th e Ra d ica l Ch a in
Cycliza tion of 3,7-Dim eth ylen ebicyclo[3.3.1]n on a n e w ith
Br om otr ich lor om eth a n e u p on γ Ir r a d ia tion (F r om Ref
12)
ization of 4 to either 5 or 6 in cases where solvent
decomposition produces Cl•. No other products assign-
able to the radicals 1-3 were found in the runs from
which kinetic data were extracted. The major side
product not derived from the radicals was monobasic acid
8, formed by hydrolysis of the remaining acid chloride
moiety in the monoester between alcohols and oxalyl
chloride by trace amounts of water. The total mass
balance of the desired products 4-6 plus 8 is 91% from
the oxalic acid monoester, with the remainder consisting
of solvolysis products of the alcohol or products from
incomplete decomposition of the oxalate ester. It was
very important to the acquisition of reliable product
ratios to account for as much of the mass as possible.
One should note that any side product coming from the
Barton precursor prior to double decarboxylation would
have no effect on the relative ratios of 4-6 and, thus, no
effect on the derived thermochemistry. As a check, we
also determined that the products themselves did not
interconvert after they were formed by subjecting a 5.3/
12.2/82.4 mixture (an arbitrary ratio) of 4, 5, and 6 to
the reaction conditions. No alteration in the initial ratio
was observed. A minor difficulty occurred when it was
discovered that small amounts of 6 appeared already
after the reaction of methylenebicyclononanol with oxalyl
chloride, presumably arising from traces of HCl in the
oxalyl chloride. Attempts to suppress the “premature”
formation of 6 met with only variable success. We judged
it more reliable to correct for existing 6. GC/MS analysis
of an aliquot removed from the reaction immediately
prior to addition to the thiopyridine N-oxide allowed us
to subtract off the spurious contribution to the total
chloroadamantane product, which amounted to only a
small correction. In the end, the product ratios for 1-3
at 75 °C were used to derive absolute rates for the 1 f
2 and 1 f 3 reactions. However, kinetic simulations for
higher temperatures, based on the absolute isomerization
rates of 1 at 75 °C, will be validated by comparing the
projected product ratios at higher temperatures to the
measured distributions. The success of that simulation
lends credence to the legitimacy of the procedure.
T (°C)
22
40
77
100
150
200
%5-exo
%6-endo
100.0
0.0
99.9
0.1
97.7
2.3
95.4
4.6
90.3
9.7
44.8
55.2
Ta ble 3. En th a lp ies, in k ca l/m ol Rela tive to th e
3-Meth ylen ebicyclo[3.3.1]n on a n -7-yl Ra d ica l (1), for th e
(3-Nor a d a m a n tyl)m eth yl (2) a n d 1-Ad a m a n tyl (3)
Ra d ica ls a n d th e Tr a n sition Sta tes for Th eir
In ter con ver sion
The closest model system for which product studies
give some indication of 5-exo vs 6-endo cyclization prefer-
ence is 3,7-dimethylenebicyclo[3.3.1]nonane, which in
BrCCl3 with radical chain initiation, gives two adducts
(Table 2).12,13 The product ratios at the higher temper-
atures are suspect because of the difficulties mentioned
earlier, but the general trend is similar. Direct compari-
son of these data to the computations for diamond CVD
is hindered by the extra substitution on the attacking
radical. Furthermore, the absence of any detectable
uncyclized products prevents extraction of absolute rate
data from the product ratios. Lastly, the necessary
auxiliary kinetic data are unavailable to convert the
product ratios into rates. All that can be concluded is
that, as the temperature is increased, the decrease in the
5-exo/6-endo ratio suggests a change from kinetic to
thermodynamic control.
Discu ssion
A considerable body of data exists for the 5-hexenyl
rearrangement, not only because of its preparative
importance, but also because of the role of the rearrange-
ment as a “radical-clock” reaction6 for competitive kinet-
ics in radical reactions. The wealth of experimental
results from which qualitative comparisons may be
drawn seems to have escaped notice in the CVD kinetics
community. While there is general agreement that the
5-exo cyclization of 5-hexenyl radicals is kinetically
preferred despite the thermodynamic preference for the
6-endo product, it has been noted that both the kinetic
and thermodynamic preferences are strongly dependent
on steric and electronic factors.10 In particular, alkyl
substituents at the 5-position in simple 5-hexenyl sys-
tems can reverse the kinetic preference from 5-exo to
6-endo cyclization.11 Given the presence of just such a
substitution pattern in the 3-methylenebicyclo[3.3.1]nonan-
7-yl structures in the proposed CVD mechanism, reliable
predictions ought to be difficult.
In a computational approach to the diamond CVD
process, Shokov et al.9 calculated, among other things,
the relative enthalpies for the three radicals 1-3 and the
transition states by which they interconvert (in their
“small” cluster calculations). The computed results for
the three radicals, listed in Table 3, provide the connec-
tion to experimentally accessible kinetics that can justify
(or not) the choice of computational method applied to
the larger clusters that were used in that study to model
the diamond [100] surface.
Because of the size of the computational problem, the
surface radical structures were treated with the PM3
semiempirical Hamiltonian at the unrestricted Hartree-
Fock (UHF) level. The primary motivation of the present
(12) Yurchenko, A. G. In Cage Hydrocarbons; Olah, G. A., Ed.;
Wiley: New York, 1990; Chapter 5.
(13) Yurchenko, A. G.; Zosim, L. A.; Dovgan, N. L.; Verpovsky, N.
S. Tetrahedron Lett. 1976, 4843. Yurchenko, A. G.; Zosim, L. A.;
Dovgan, N. L. Zh. Org. Khim. 1974, 10, 1996.
(10) March, J . Advanced Organic Chemistry, 4th ed.; Wiley
Interscience: New York, 1992; p 752.
(11) Chuang, C. P.; Gallucci, J . C.; Hart, D. J .; Hoffman, C. J . Org.
Chem. 1988, 53, 3218.

4584 J . Org. Chem., Vol. 63, No. 14, 1998
Mueller and Chen
study is to check these computational results against a
model system for which experimental data can be ob-
tained. As noted in ref 9, the kinetics of the surface
rearrangements are not rate-limiting in the existing CVD
process at 1200 K. Nevertheless, any attempt to run the
reactions for CVD, or an analogous set of reactions, at
lower temperatures would eventually run into a kinetic
bottleneck imposed by the radical reactions investigated
here. We therefore judged it important to obtain reliable,
quantitative experimental data on the reactions of 1-3.
The product ratios in Table 1 can be converted to
absolute rates with the aid of some auxiliary kinetic
information. As mentioned in the results section, gen-
eration of 2 gave only 5, while generation of 3 gave only
6. Only 1 gave all three trapping products. Moreover,
the temperature dependence of the trapping product
ratios from 1 over the 75-126 °C range, where we trust
the data, manifests itself primarily in the reduction in
the amount of 4 to a level below that which we can
reliably make an absolute determination. The ratio of 5
to 6 appears to change very slowly. An Arrhenius plot
with such data would not produce credible results.
Instead, the data from the runs at 75 °C, which after all,
were repeated many more times than the higher tem-
perature runs and were also subject to fewer possible
systematic errors, can be converted to rates for the
reactions 1 f 2 and 1 f 3 by using previously reported
kinetic data for the chlorine abstraction reactions by
primary, secondary, and tertiary radicals from CCl4. The
reactions of n-butyl, cyclopentyl, and tert-butyl radicals
with CCl4 have been determined by time-dependent ESR
spectroscopy.14 We use the Arrhenius parameters from
that report to compute the rates for chlorine abstraction
by 1-3, which must occur in kinetic competition with
isomerization of the radicals. Several other investiga-
tions15 report kinetic data for either only one of the three
classes of alkyl radicals or report rates only at a single
temperature. An examination of the different determi-
nations finds that no significant differences in the derived
activation energies for the 1 f 2 and 1 f 3 reactions
would occur if auxiliary data from one of the other reports
were to have been used. We used the kinetic parameters
from ref 14 primarily so that the final kinetic simulations
could be done with a consistent set of trapping rates.
Taking the density of CCl4 at 75 °C (F ) 1.4864),16 the
neat solvent is 9.66 M. The pseudo-first-order rate
constants for chlorine abstraction from CCl4 by primary,
secondary, and tertiary radicals are then, k1° ) 1.81 ×
106, k2° ) 1.62 × 105, and k3° ) 8.52 × 105 s-1. Using the
trapping rate, k2°, as a clock, the first-order rate constants
error boundaries due to propagation of the experimental
uncertainty in the product ratios are much smaller than
the reported error boundaries reported for the absolute
chlorine abstraction rates, so the latter are almost
entirely responsible for the (1.0 kcal/mol boundaries on
the two derived activation energies. A potential source
of a systematic error would be the underestimation of
the Arrhenius preexponential factor for the more rigid
system in the present study. For each factor of 10
increase in A, one would need to increase the derived Ea
by 1.6 kcal/mol.18
To complete the potential surface for the interconver-
sions of 1-3, we need the relative enthalpies for the three
radicals themselves. Combined with the activation ener-
gies derived above, the enthalpies would give the activa-
tion energies for the reverse processes. The enthalpy of
formation for the 1-adamantyl radical has been experi-
mentally determined19 to be ∆Hf,298[3] ) 14.8 ( 2.0 kcal/
mol, but for neither 1 nor 2 are there experimental data.
Nevertheless, values for ∆Hf,298[1] and ∆Hf,298[2] should
be obtainable with high reliability by combining a mo-
lecular mechanics calculation for the corresponding
hydrocarbon with canonical primary, secondary, and
tertiary bond dissociation energies. Polycyclic hydrocar-
bons like 3-methylenebicyclo[3.3.1]nonane and 3-meth-
ylnoradamantane are precisely the kind of structures for
which empirical force fields are parameterized. Test
calculations on 13 bicyclic, tricyclic, and tetracyclic
hydrocarbons20 with the MMX force field produces heats
of formation with a root mean square (rms) deviation
from experiment of less than 1.0 kcal/mol. The molecular
mechanics calculation gives ∆Hf,298[methylenebicyclo-
nonane]
) -11.7 kcal/mol and ∆Hf,298[methylnor-
adamantane] ) -23.2 kcal/mol. Combining these with
critically-evaluated C-H bond energies21 produces
∆Hf,298[1] ) 34.8 ( 1.1 and ∆Hf,298[2] ) 25.8 ( 1.1 kcal/
mol. As a control, the heat of formation of the 1-ada-
mantyl radical, computed in the same manner, is ∆Hf,298[3]
) 12.9 ( 1.1 kcal/mol, which compares well with the
number from experiment. The results for the three
radicals and the two transition states are listed in Table
3. A check of the final results can be made by performing
a kinetic simulation of the rearrangement and trapping
reaction sequence using PowerSim, a program for model-
ing nonlinear dynamics with feedback, like chemical
kinetics. The activation parameters from Table 3 are
input with those for chlorine abstraction and run at the
four temperatures listed in Table 1. The resulting
product ratios are listed in parentheses in Table 1. The
good agreement within the uncertainty boundaries in-
dicates that the derived activation parameters realisti-
cally model the observed chemistry.
for the reactions 1 f 2 and 1 f 3 are, therefore, k1f2
)
9.2 × 106 and k1f3 ) 1.7 × 105 s-1 at 75 °C. Using the
Arrhenius pre-exponential factor for the parent 5-hexenyl
rearrangement,17 log A ) 10.37 ( 0.32, the activation
energies for the radical isomerizations become Ea[1 f 2]
) 5.4 ( 1.0 and Ea[1 f 3] ) 8.2 ( 1.0 kcal/mol. The
Comparison of the calculated thermochemistry with
(18) In the final POWERSIM calculation of the product ratios, listed
in Table 1 as the numbers in parentheses, indistinguishable product
ratios can be obtained for a 100-fold higher A factor if the activation
energies are correspondingly reduced by 3.2 kcal/mol. Nevertheless,
given that log A < 13 is a safe assumption, the systematic error is not
(14) Hawari, J . A.; Davis, S.; Engel, P. S.; Gilbert, B. C.; Griller, D.
J . Am. Chem. Soc. 1985, 107, 4721.
too large.
A redetermination of absolute rate constants for Cl-
(15) Giese, B.; Hartung, J . Chem. Ber. 1992, 125, 1777. Katz, M.
G.; Horowitz, A.; Rajbenbach, L. A. Int. J . Chem. Kin. 1975, 7, 183.
Fritz, P. G.; McLauchlan, K. A. J . Chem. Soc., Faraday Trans. 2 1976,
72, 87. J ohnston, L. J .; Ingold, K. U. J . Am. Chem. Soc. 1986, 108,
2343.
(16) The density of carbon tetrachloride over a range of temperatures
is tabulated by: Timmermans, J . Physico-Chemical Constants of Pure
Organic Compounds; Elsevier: Amsterdam, 1950.
(17) Chatgilialoglu, C.; Ingold, K. U.; Scaiano, J . C. J . Am. Chem.
Soc. 1981, 103, 7739.
abstraction by adamantyl radical is also in progress to check the
assumptions.
(19) Kruppa, G. H.; Beauchamp, J . L. J . Am. Chem. Soc. 1986, 108,
2162.
(20) The 13 hydrocarbons are the test set for polycyclic systems in:
Engler, E. M.; Andose, J . D.; v. Schleyer, P. J . J . Am. Chem. Soc. 1973,
95, 8005. The prediction of reliable heats of formation by empirical
force fields is validated in this paper.
(21) Berkowitz, J .; Ellison, G. B.; Gutman, D. J . Phys. Chem. 1994,
98, 2744.

Chemical Vapor Deposition of Diamond
J . Org. Chem., Vol. 63, No. 14, 1998 4585
Chloroadamantane was obtained from Aldrich Chem. Co. FT-
IR spectra were recorded on NaCl plates with a Perkin-Elmer
Paragon 1000. Melting points are uncorrected.
the numbers reported here finds that the qualitative
behavior of the system is well-modeled by the relatively
modest level of theory used by Shokov et al.9 By
extension, the larger-scale computations for the diamond
[100] surface are likely to be qualitatively correct.
Nevertheless, the deviations are still large enough that
the actual rates and equilibrium constants would be
substantially different from the predicted values. For the
present high-temperature CVD process, e.g., in a fila-
ment-assisted CVD reactor, the deviations may in fact
be inconsequential because of the rate-limiting mass
transport to the surface, but for any potential low-
temperature analog of CVD, the rates become important.
The most significant finding in this regard is that the
activation barriers for the interconversions between
surface radical structures are even lower than predicted
1,3-Dibr om oa d a m a n ta n e. Bromination of adamantane
followed the procedure by Baugham.23 Br2 (216 mL, 4.12 mol),
BBr3 (7.8 mL, 82.0 mmol), and Al2Br6 (1.0 g) were placed in a
1 L three-necked flask fitted with a mechanical stirrer and
tubing leading to two gas washbottles (filled with concd
Na2S2O5 solution) and placed in an ice bath. Adamantane (50
g, 367 mmol, Aldrich Chemical Co.) was added slowly. After
addition was complete, the ice bath was removed. After 30
min, gas evolution ceased, and the reaction mixture was heated
to reflux for 3 h. The reaction mixture was then cooled again
in an ice bath and quenched with a concd solution of Na2S2O5,
resulting in vigorous evolution of both gas and heat. Further
Na2S2O5 was added as a solid until the reaction mixture
changed color. After neutralization with NaHCO3, the mixture
was filtered and extracted with chloroform. The combined
organic fractions were dried over sodium sulfate, redissolved
in ether, washed with further concd Na2S2O5, dried over
magnesium sulfate, and evaporated under reduced pressure,
yielding 87.5 g (81%) of crude 1,3-dibromoadamantane. An
aliquot for analysis was recrystallized from methanol: mp 112
°C; 1H NMR (200 MHz, CDCl3) δ 1.70 (br, 2H), 2.29 (br, 10H),
2.88 (s, 2H); 13C NMR (50 MHz, CDCl3) δ 33.5, 34.1, 46.4, 58.7,
62.0.
3-Meth ylen ebicyclo[3.3.1]n on a n -7-on e. The procedure
from Gaugneux was followed.24 Dibromoadamantane (30 g,
102 mmol) and 300 mL of 1 N aqueous NaOH were placed in
a steel autoclave filled with 270 mL of dioxane. The autoclave
was brought to 180 °C over 2 h and maintained at that
temperature for an additional 30 min. The reaction mixture
was washed with aqueous NH4Cl, extracted with chloroform,
and dried over sodium sulfate. Recrystallization from metha-
nol yielded 9.2 g (60%) of the product, which sublimes at 158-
161 °C (sealed tube): 1H NMR (200 MHz, CDCl3) δ 1.92 (br,
2H), 2.20-2.50 (br, 10H), 4.78 (s, 2H); 13C NMR (50 MHz,
CDCl3) δ 30, 32, 41, 47, 114, 141, 211.
•
by theory. If a functional substitute for H• and CH3 in
the CVD process could be designed, the only step that
would require even moderately high temperatures would
be the desorption of H2 in the 2 × 1 reconstruction of the
[100] surface. All of the other reactions would be efficient
at temperatures many hundreds of degrees below the
1200 K surface temperature in present CVD reactors.
Curiously, ab initio calculations at the MP2/6-31G* level
of theory, shown in Table 3sagain, a relatively low level
that could be used for larger clusters modeling a
surfacesdo a rather poor job of reproducing the energet-
ics for the isomerization. One concludes that the semiem-
pirical approach by Shokov et al.9 is a good compromise
between numerical accuracy and computational expense.
Con clu sion s
Model radical systems for the surface radicals in the
chemical vapor deposition of diamond are prepared and
trapped. Product ratios are used to derive absolute rates
for the isomerizations predicted to be important in the
incorporation of added carbon atoms into the diamond
lattice. The results from the small model systems agree
qualitatively with calculations at the relatively low level
of theory that are used for the computation of a large
section of the diamond surface. We find that the prin-
cipal difference from the calculated results is the over-
estimation of activation energies by the calculation.
3-Meth ylen ebicyclo[3.3.1]n on a n -7-ol.25 Reduction of me-
thylenebicyclononanone (1.76 g, 12 mmol) with LiAlH4 (0.22
g, 5.6 mmol) proceeded by slow, dropwise addition under dry
N2 of the ketone dissolved in 10 mL of THF into a stirred
suspension of the hydride suspended in 15 mL of THF. After
1 h at reflux and 4 h at rt, a concentrated solution of sodium
sulfate was added dropwise until a precipitate formed. The
reaction mixture was then diluted with THF, filtered, and
dried over sodium sulfate and the solvent evaporated to yield
1.5 g (86%) of a white solid. The 3:1 mixture of endo vs exo
alcohol was separated by flash chromatography over silica
using hexane/EtOAc in a 5:1 ratio. Endo isomer: mp 82-3
°C; 1H NMR (200 MHz, CDCl3) δ 1.53-1.68 (m, 4H), 2.10-
2.49 (br, 8H), 2.99 (d, J ) 12 Hz, 1H), 3.78 (m, 1H), 4.90 (m,
2H); 13C NMR (50 MHz, CDCl3) δ 28.16, 32.41, 39.83, 39.74,
62.93, 112.10, 148.35. Exo isomer: mp 86-7 °C; 1H NMR (200
MHz, CDCl3) δ 1.16 (s, 1H), 1.44 (td, 2H), 1.59 (m, 2H), 2.01
(br, 2H), 2.18 (t, 2H), 2.33 (br, 4H), 4.58 (m, 1H), 4.63 (m, 2H);
13C NMR (50 MHz, CDCl3) δ 28.34, 31.23, 37.64, 39.64, 61.83,
106.59, 146.45.
3-Nor a d a m a n tylm eth a n ol. BH3‚THF (1 M, 9.7 mL, 9.7
mmol) was added dropwise to a stirred solution of the
commercially available noradamantanecarboxylic acid (1.01 g,
6.1 mmol) in 10 mL of THF and stirred at rt overnight. The
solution was then quenched with water and extracted with
ether. The organic washes were dried and then evaporated
to yield 0.91 g of a white solid (98%): mp 146 °C (lit.26 142-4
°C); 1H NMR (200 MHz, CDCl3) δ 1.4 (br, 10H), 2.10 (br, 1H),
2.24 (br, 2H), 3.63 (s, 2H); 13C NMR (50 MHz, CDCl3) δ 35.23,
37.15, 40.28, 43.71, 45.62, 50.73, 68.83.
Exp er im en ta l Section
Radicals 1-3 were prepared by the pyrolysis of the Barton
precursors22 derived from the corresponding alcohols, oxalyl
chloride, and the sodium salt of thiopyridine N-oxide. This
approach was chosen from among several other prospective
routes because of the synthetic availability of the three
alcohols, convenience in radical generation, and clean subse-
quent radical chemistry. The latter condition is critical for
any quantitative determinations.
In the preparative work, 1H NMR spectra were recorded on
a Varian GEMINI 200 MHz spectrometer. Chemical shifts
are reported in ppm relative to TMS. Analysis by GC/MS was
done with a Fisons GC8060/MD800 instrument fitted with a
DB-5ms capillary column. In a typical run, the column
temperature was ramped from 50 to 108 °C over 4 min,
followed by a subsequent ramp up to 220 °C over 22 min.
Quantitative determinations by GC/MS were done relative to
anthracene as a standard. Response factors were determined
by independent synthesis of the trapping products and co-
injection of known amounts of the products with anthracene.
Ra d ica ls 1-3: Atm osp h er ic P r essu r e P r oced u r e. The
(23) Baugham, G. L. J . Org. Chem. 1964, 29, 238.
(24) Gaugneux, A. R.; Meier, R. Tetrahedron Lett. 1969, 1365.
(25) Stehelin, L.; Kanellias, L.; Ourisson, G. J . Org. Chem. 1973,
38, 851.
(22) Barton, D. H. R.; Crich, D. J . Chem. Soc., Perkin Trans. 1 1986,
1603.
(26) Vogt, B. R.; Hoover, J . R. E. Tetrahedron Lett. 1967, 30, 2841.

4586 J . Org. Chem., Vol. 63, No. 14, 1998
Mueller and Chen
3-methylenebicyclo[3.3.1]nonan-7-yl radical (1), (3-norada-
mantyl)methyl radical (2), and 1-adamantyl radical (3) were
prepared according to two procedures. For the runs at 75 °C,
i.e., refluxing CCl4 at atmospheric pressure, the procedure for
1 is given here as representative. The radicals were formed
fractions containing the desired product were combined and
dried (MgSO4), and the solvent was removed carefully to avoid
loss of the volatile compound. After sublimation, 0.038 g of 4
was obtained (3.4%): mp 61 °C; 1H NMR (200 MHz, CDCl3) δ
1.68 (br, 2H), 1.87 (td, 2H), 2.13-2.32 (br, 4H), 2.39 (br, 4H),
4.71 (m, 2H), 4.92 (m, 1H); 13C NMR (50 MHz, CDCl3) δ 29.39,
30.85, 37.29, 41.10, 52.88, 107.19, 146.14; GC/MS (DB-5 ms,
70 eV EI) 14.6 min, m/e ) 170 (M+, 10), 92 (100), 78 (95);
HRMS for C10H15Cl (M+) calcd 170.0857, found 170.0856.
3-Nor a d a m a n tylch lor om eth a n e (5). 3-Noradamantyl-
methanol (120 mg, 0.79 mmol) was dissolved in 5 mL of
benzene and added dropwise to a 2 M solution of oxalyl
chloride in dichloromethane (2.4 mL, 4.8 mmol) that had been
further diluted with 5 mL of benzene. After stirring, the
solvents and excess reagents were removed by a stream of dry
N2. The residue was redissolved in 5 mL of benzene and added
dropwise to a suspension of the thiopyridine N-oxide sodium
salt (302 mg, 2.0 mmol) in 200 mL of carbon tetrachloride over
a period of 3 h. The reaction mixture was then filtered, and
the majority of the solvent was distilled off. The concentrated
solution was eluted through a silica gel column with pentane.
The solvent was carefully removed by distillation to yield 23
mg (17.1%) of 5 as a colorless oil: 1H NMR (200 MHz, CDCl3)
δ 1.50-1.80 (br, 10H), 2.11-2.30 (br, 3H), 3.63 (s, 2H); 13C
NMR (50 MHz, CDCl3) δ 34.76, 37.20, 41.74, 43.46, 46.66,
50.34, 52.53; GC/MS (DB-5 ms, 70 eV EI) 13.1 min, m/e ) 170
(M+, 7), 135 (60), 79 (100); HRMS for C10H15Cl (M+) calcd
170.0857, found 170.0852.
Com p u ta tion a l Deta ils. Heats of formation for (nonradi-
cal) hydrocarbons that are unavailable by calorimetry were
determined by molecular mechanics using the MMX force field
derived from MM2 in the PCModel program (Serena Software)
on an IBM-compatible PC. The calculated heats of formation
for a number of cage and bicyclic compounds closely match
experimental values for those cases when the data exist for
the comparison. For example, the lowest energy conformer of
[3.3.1]bicyclononane is computed to have ∆Hf ) -30.5 kcal/
mol, compared to -30.49 kcal/mol by experiment.27 Further
computations for the isomerizations 1 f 2 and 1 f 3 were
done at the MP2/6-31G*//PM3 level of theory with the geom-
etries fully relaxed along the reaction coordinate using the
Gaussian 92 package of programs28 running on an IBM RS/
6000-590 UNIX workstation. Kinetic simulations were run
with PowerSim V2.01 (ModelData AS) on an IBM-compatible
PC.
from the alcohol by the Barton thiohydroxamate route.22
A
solution of 0.045 g (0.29 mmol) of the exo-3-methylene-
bicyclo[3.3.1]bicyclononan-7-ol in 10 mL of benzene was added
dropwise via cannula over 30 min to a solution of 0.88 mL (1.75
mmol) of oxalyl chloride in 10 mL of benzene under inert gas.
After the addition was complete, solvent and excess oxalyl
chloride were removed by evaporation under a stream of dry
N2. The mixture was redissolved in 5 mL of benzene. An
aliquot was removed, filtered through a plug of silica gel, and
analyzed by GC/MS to determine the amount of “premature”
chloroadamantane. The remainder of the solution was added
dropwise with vigorous stirring to 150 mL refluxing CCl4
containing an equimolar amount of suspended thiopyridine
N-oxide sodium salt. Within the first 2 min of the 10 min
addition, the solution turned yellow but became white again
after a further 15 min. An aliquot was then removed, filtered
through silica gel, and analyzed by GC/MS. The remaining
solution was carefully evaporated, redissolved in CDCl3, and
analyzed by NMR.
Ra d ica ls 1-3: High -P r essu r e P r oced u r e. Higher tem-
perature runs were performed in a sealed, thermostatically-
controlled high-pressure bomb reactor. Again, a representa-
tive procedure for 1 is described. Preparation of the oxalate
ester of exo-3-methylenebicyclo[3.3.1]nonan-7-ol proceeded as
noted above. An aliquot was removed prior to further reaction
to assay “premature” chloroadamantane. The bomb reactor
was fitted with a pressure-equalizing addition vessel with
stainless steel valves, pressure sensor, a temperature control-
ler with a thermistor sensor in contact with the solution, and
a belt-driven mechanical stirrer. The bomb was charged with
150 mL of CCl4 and an equimolar amount of the thiopyridine
N-oxide sodium salt and stirred vigorously for 3 h to achieve
uniform suspension of the solid. The bomb was further
pressurized and purged with 20 atm of N2 three times in
succession to check for leaks and remove traces of oxygen. After
the purging, the bomb was brought to the desired internal
temperature. The adduct of the alcohol and oxalyl chloride
in the addition vessel was added in small “squirts” every 60 s
over a period of 30 min with vigorous stirring in the bomb.
After the addition was complete, the addition vessel was briefly
pressurized and residual solution flushed into the reactor. The
bomb was held at the desired temperature for 15 min and then
quenched to room temperature. The reaction mixture was
then analyzed as described previously.
Su p p or tin g In for m a tion Ava ila ble: A typical GC/MS
trace from the radical trapping experiments as well as NMR
spectra of compounds 4 and 5 (20 pages). This material is
contained in libraries on microfiche, immediately follows this
article in the microfilm version of the journal, and can be
ordered from the ACS; see any current masthead page for
ordering information.
3-Meth ylen e-7-ch lor obicyclo[3.3.1]n on a n e (4). endo-3-
Methylenebicyclo[3.3.1]nonan-7-ol (1.0 g, 6.57 mmol) was
dissolved in pyridine to which 1.38 g (7.23 mmol) of p-
toluenesulfonic acid chloride was added. After the mixture
was stirred overnight, 0.56 g of lithium chloride (13.14 mmol)
was added, and after 24 h, the reaction mixture was extracted
with pentane, the extracts were washed with dilute acetic acid
and water and dried over MgSO4, and the solvents were
removed to yield 1.59 g of a colorless oil that was purified by
column chromatography on silica gel (hexane). A 2:5 mixture
of 1-chloroadamantane and 3-methylene-7-chlorobicyclo[3.3.1]-
nonane (4) was obtained and separated by HPLC (RP-18
column, MeOH, 30% H2O). The HPLC fractions were quickly
exracted with pentane to avoid product decomposition, the
J O971814+
(27) Parker, W.; Steele, W. V.; Watt, I. J . J . Chem. Thermodyn. 1977,
9, 307.
(28) Gaussian 92, Revision A, Frisch, M. J .; Trucks, G. W.; Head-
Gordon, M.; Gill, P. M. W.; Wong, M. W.; Foresman, J . B.; J ohnson, B.
G.; Schlegel, H. B.; Robb, M. A.; Resplogle, E. S.; Gomperst, R.; Andres,
J . L.; Raghavachari, K.; Binkley, J . S.; Gonzales, C.; Martin, R. L.;
Fox, D. J .; DeFrees, D. J .; Baker, J .; Stewart, J . J . P.; Pople, J . A.
Gaussian, Inc., Pittsburgh, PA, 1992.