Stereoselection at the steady state. Group selective radical cyclizations of substrates containing two radical precursors and one radical acceptor

Article abstract of DOI:10.1021/ja9724102

The kinetic model for compound stereoselection presented in previous paper is verified experimentally by conducting a series of radical cyclizations of 1-(3-halo-2-(halomethyl)propyl)cycloalk-2-enecaboxylates with tributyltin hydride and measuring the ratios of products. Cyclization rate constants are abstracted from the data by an analysis that minimizes total error, and these rate constants compare favorably with rate constants that we measured directly on the corresponding monohalides. Transition state modeling was used to interpret the initial round of results and to design two new systems, one of which was predicted to cyclize with higher selectivity than the parent and the other of which was predicted to cyclize with reverse selectivity. These substrates bearing 4-tert-butyl groups were synthesized, and the experimental results verified the computational predictions.

Full text of DOI:10.1021/ja9724102

342
J. Am. Chem. Soc. 1998, 120, 342-351
Stereoselection at the Steady State. Group Selective Radical
Cyclizations of Substrates Containing Two Radical Precursors and
One Radical Acceptor
Dennis P. Curran,* Chien-Hsing Lin, Nicholas DeMello, and Jo1rg Junggebauer
Contribution from the Department of Chemistry, UniVersity of Pittsburgh, Pittsburgh, PennsylVania 15260
ReceiVed July 17, 1997
Abstract: The kinetic model for compound stereoselection presented in previous paper is verified experimentally
by conducting a series of radical cyclizations of 1-(3-halo-2-(halomethyl)propyl)cycloalk-2-enecarboxylates
with tributyltin hydride and measuring the ratios of products. Cyclization rate constants are abstracted from
the data by an analysis that minimizes total error, and these rate constants compare favorably with rate constants
that we measured directly on the corresponding monohalides. Transition state modeling was used to interpret
the initial round of results and to design two new systems, one of which was predicted to cyclize with higher
selectivity than the parent and the other of which was predicted to cyclize with reverse selectivity. These
substrates bearing 4-tert-butyl groups were synthesized, and the experimental results verified the computational
predictions.
Introduction
The previous paper1 outlined the conceptual basis for ste-
reoselection at the steady state within the framework of group
selective processes. At the steady state, the relative concentra-
tion of two stereomeric reactive intermediates remains constant.
And if a concentration gradient between these two intermediates
can be established, then they can be directed down different
pathways that lead to the same product by choosing appropriate
reaction conditions. The theory predicts, among other things,
that by combining the natural convergent topography of a group
selective process with the features of stereoselection at the steady
state, it is possible to execute a net “group selective” transfor-
mation solely by using reaction topography. Two stereotopic
reacting groups must be present in the precursor, but they need
not react selectively to result in selective formation of a final
product. This paper provides a series of experimental tests for
the theory in the field of diastereotopic group selective radical
cyclizations.
Radical cyclizations2 were the first among several types of
radical processes to receive attention from the standpoint of
stereochemistry.3 Most stereoselective radical cyclizations
involve face selection; a radical folds into a suitable conforma-
tion to attack one face or the other of an alkene or related
functional group. The Beckwith-Houk4 ,5 model is a powerful
tool to predict and rationalize the outcome of such cyclizations
Figure 1. The Beckwith-Houk model.
that can be applied qualitatively or quantitatively.6 In a
qualitative approach, the major product of a 5-exo cyclization
can often be predicted from the “chair-equatorial” model shown
in Figure 1. Trends in stereoselection (including exceptions to
the rule) can often be interpreted by comparing this chair with
the alternative ring-flipped chair and boatlike models.
Group selective radical cyclizations are much less common
than their face selective relatives.7 Two general types of group
selective radical cyclizations are possible: either there can be
two radical acceptors and one radical precursor (class I) or there
can be two radical precursors and one acceptor (class II).
Interpretation of the results from class I (two acceptors) is
generally straightforward. A key example that served as a
design element in this work is shown in Figure 2.8 Reductive
(1) Curran, D. P.; DeMello, N. J. Am. Chem. Soc. 1998, 120, 329.
(2) Review: Giese, B.; Kopping, B.; Gobel, T.; Dickhaut, J.; Thoma,
G.; Kulicke, K. J.; Trach, F. In Organic Reactions; John Wiley & Sons
Inc: New York, 1996; Vol. 48; p 301.
(3) (a) Curran, D. P.; Porter, N. A.; Giese, B. Stereochemistry of Radical
Reactions: Concepts, Guidelines, and Synthetic Applications; VCH: Wein-
heim, 1996; p 283. (b) Chapter 2 of this book is a review of substrate
controlled radical cyclizations.
(4) (a) Beckwith, A. L. J.; Easton, C. J.; Lawrence, T.; Serelis, A. K.
Aust. J. Chem. 1983, 36, 545. (b) Beckwith, A. L. J.; Easton, C. J.; Serelis,
A. K. J. Chem. Soc., Chem. Commun. 1980, 482. (c) Beckwith, A. L. J.;
Lawrence, T.; Serelis, A. K. J. Chem. Soc., Chem. Commun. 1980, 484.
(d) Beckwith, A. L. J.; Schiesser, C. H. Tetrahedron Lett. 1985, 26, 373.
(5) (a) Spellmeyer, D. C.; Houk, K. N. J. Org. Chem. 1987, 52, 959. (b)
Broeker, J. L.; Houk, K. N. J. Org. Chem. 1991, 56, 3651.
(6) Examples: (a) Belvisi, L.; Gennari, C.; Poli, G.; Scolastico, C.; Salom,
B. Tetrahedron-Asymmetry 1993, 4, 273. (b) Myers, A. G.; Condroski, K.
R. J. Am. Chem. Soc. 1995, 117, 3057. (c) Takahashi, T.; Tomida, S.;
Sakamoto, Y.; Yamada, H. J. Org. Chem. 1997, 62, 1912.
(7) a) Curran, D. P.; Geib, S. J.; Lin, C. H. Tetrahedron-Asymmetry 1994,
5, 199. (b) Curran, D. P.; Shen, W.; Zhang, J. C.; Geib, S. J.; Lin, C. H.
Heterocycles 1994, 37, 1773.
S0002-7863(97)02410-4 CCC: $15.00 © 1998 American Chemical Society
Published on Web 01/06/1998

Stereoselection at the Steady State
J. Am. Chem. Soc., Vol. 120, No. 2, 1998 343
Figure 2. A class I group selective radical process: 2 acceptors and
1 precursor.
cyclization of 1 with tributyltin hydride provides 2-exo and
2-endo in temperature-dependent selectivities that range from
30/1 (-78 °C) to 15/1 (80 °C). The selectivity is established
at the stage of the radical 3, which chooses between a chairlike
transition structure (3E) involving one olefin and an isomeric
chairlike structure (3A) involving the other. It follows from a
simple application of the Beckwith-Houk model that 3E should
be favored, and this prediction was borne out by the experi-
mental results.8
Although there are relatively few examples of this type of
radical process,9 it shows the usual features of other types of
group selective reactions. Radical 3 makes a simple choice
between one alkene and the other. Since both choices are
intramolecular, the stereoselectivity of this process cannot be
varied by changing the tin hydride concentration (although the
ratio of directly reduced to cyclized products is of course
affected by the tin hydride concentration). In this process, like
most other group selective processes, there is no topological
way to improve or erode the stereoselection; the only thing that
counts is the relative energy of the competing transition states
3A and 3E.
More conceptually interesting is the class of reactions that
has two radical precursors and one radical acceptor. According
to the prior paper,1 these reactions will exhibit a number of
unusual features, including dependence of stereoisomer ratios
on tin hydride concentration with the attendant ability to design
reaction conditions that provide final product selectivities that
exceed the level of selection in the group selective step.
There is at least one example of a group selective radical
cyclization in class II that appeared prior to our work, and this
is shown in Figure 3.10 Reduction of dibromide 4 with tin
hydride provides an 87/13 mixture of stereoisomers 5. This
process is completely group selective; the stereoisomer mixture
results from the accompanying face selective process. The
authors of this paper rationalized the result by positing the
intermediacy of diradical 6, which undergoes a group selective
cyclization as shown. This would be a standard type of group
selective process where two reactive entities directly compete
for a single functional group; however, we are unaware of any
group selective reactions of diradicals. In this mechanism, the
problem is not with the postulate that the diradical 6 will undergo
Figure 3. A class II group selective radical process: 2 precursors plus
1 acceptor.
stereoselective cyclization, it is with the postulate that diradical
6 will be generated in the first place. The reaction of a mono-
radical with a second trialkyltin tin radical is a rare event because
it is a radical-radical reaction, and when it does occur, it should
provide the standard products of radical-radical reactions
(coupling and disproportionation) rather than diradicals.
With 20/20 hindsight, we can now see that the reaction in
Figure 3 is a superbly selective example of a class II group
selective radical cyclization. If Br1 is abstracted (see 7a),
cyclization is possible because a normal “outside-outside”
bicycle is formed. But if Br2 is abstracted (see 7b), cyclization
is impossible because an “inside-outside” fused dioxabicyclo-
[2.2.1]heptane would result. The bromine abstraction surely
occurs with little or no selection, but the reaction topography
still dictates that the product is formed with complete group
selection: if Br1 is abstracted first, then cyclization precedes
reductive debromination, but if Br2 is abstracted first, then
reductive debromination precedes cyclization. While ideal from
a synthetic perspective, this process leaves something to be
desired when viewed as a “proof” of our analysis of stereose-
lection at the steady state. The problem is that the disfavored
cyclization is so slow that it could never occur under any
conditions. So the proposed variation of stereoisomer ratios as
a function of tin hydride concentration can never be tested. In
other words, the slower of the two cyclizations is outside of
the range when it can be affected by a third, independent process
(tin hydride trapping).
To experimentally test the kinetic model for this type of
topological group selective process requires the synthesis and
study of a series of double radical precursors whose derived
diastereomeric radicals cyclize over a range of different relative
rates. Ideally, neither the faster nor the slower of the processes
should be outside the range where it can be altered by changing
the tin hydride concentration. We previously described a highly
biased system with a kf/ks ratio of about 200, and showed that
the experimental data fit the kinetic model.11 Herein, we
describe in detail a series of structurally related substrates that
(8) (a) Curran, D. P.; Qi, H. Y.; DeMello, N. C.; Lin, C.-H. J. Am. Chem.
Soc. 1994, 116, 8430. (b) Qi, Hi Ph.D. Thesis, University of Pittsburgh,
1995.
(9) Additional cyclization examples can be found in the following: (a)
Reference 8b. (b) Lin, C.-H. Ph.D. Thesis, University of Pittsburgh, 1995.
For group selective hydrogen transfer, see: (c) Sugimura, T.; Goto, S.;
Koguro, K.; Futagawa, T.; Misaki, S.; Morimoto, Y.; Yasuoka, N.; Tai, A.
Tetrahedron Lett. 1993, 34, 505. (d) Sugimura, T.; Koguro, K.; Tai, A.
Tetrahedron Lett. 1993, 34, 509.
(10) Takano, S.; Ohashi, K.; Sugihara, T.; Ogasawara, K. Chem. Lett.
1991, 203.
(11) Curran, D. P.; Qi, H. Y. HelV. Chim. Acta 1996, 79, 21.

344 J. Am. Chem. Soc., Vol. 120, No. 2, 1998
Curran et al.
with tris(iodomethyl)methane (11a) provided 8a in 65% yield
after flash chromatography. The reaction is remarkable because
the triiodide 11a appears to be ripe for a competing E2
elimination process. The dibromides 8b and 14 were prepared
by alkylation of the enolate derived from 10 or 12 with the
iodide 11b followed by direct bromination of both (tert-
butyldimethylsilyl) ethers.
A preparative cyclization of the diiodide 8a was first
conducted with 2.5 equiv of tributyltin hydride (0.05 M). This
provided an inseparable 1/1 mixture of 9-exo and 9-endo in
high yield (eq 2). The structures were assigned by catalytic
Figure 4. Designing a stereoconvergent group selective substrate.
provide kf/ks ratios ranging from just over 1 to about 10. A
partial analysis of one of these substrates appeared in a
preliminary communication.8a The ability of the kinetic model
to fit the experimental data for a range of substrates with widely
differing rates of cyclization and over widely different reaction
conditions provides a strong validation of the kinetic model. In
addition, our results provide new insight into the factors
controlling stereoselection in radical cyclizations to cycloalk-
enes. Although these factors are presented in the context of
group selection, they should be translated back to face selective
processes that have similar features.
(2)
hydrogenation of a 15/1 mixture of 2-exo and 2-endo to provide
an authentic sample highly enriched in 9-exo (Figure 4). The
doubly reduced product 13 was available by reduction of 8a
with tin hydride at high concentration, but this product still
contained 9-exo and 9-endo. More conveniently, alkylation of
10 with isobutyl bromide gave pure 13.
With samples in hand and structures identified, cyclizations
of the diastereotopic diiodide 8a and dibromide 8b were studied
carefully. All the reactions were performed in benzene with
p-dimethoxybenzene as an internal standard, a catalytic amount
of AIBN, and an excess (10-15 equiv) of triphenyltin hydride
Results and Discussion
Design and Study of a Stereoconvergent Group Selective
Reaction. The Class II substrates 8 were designed as shown
in Figure 4 by the simple process of adding one radical precursor
and deleting one radical acceptor from the class I substrate 1.
This approach has three attractive features: (1) the starting
dihalides 8 should be readily available, (2) the configuration
assignment of the products from its cyclization required only a
simple one-step correlation to the products from 1, and (3) the
two diastereotopic radicals derived from 8 appeared to be so
similar to the diastereomeric radical transition structures 3E and
3A derived from 1 (Figure 2) that we could readily predict that
the rate constant ratio for the fast and slow cyclizations would
be about 15. This prediction turned out to be wrong, but the
substrate was nonetheless very informative.
1
in an oil bath at 75 °C. The H NMR spectrum of the crude
cyclized sample was obtained and it clearly showed three
characteristic resonances: a doublet (δ 0.96) for the methyl
group of the 9-exo, a doublet (δ 1.01) for the methyl group of
the 9-endo product, and a “triplet” (actually two overlapping
doublets centered at δ 0.84) for the methyl groups of the reduced
product 13. The yields of all the reactions were calibrated with
the internal standard by 1H NMR integration. In all cases, the
total yields were 100 ( 5%, and for the purposes of comparison,
the yields were calibrated to 100%.
Similar cyclizations of the cyclopentene dibromide 14 were
conducted with triphenyltin hydride (see eq 3). In this case,
To investigate the effect of different radical precursors and
ring sizes, we prepared both dibromides 8b and 14 and the
diiodide 8a. The diiodide was prepared directly in the remark-
able alkylation shown in eq 1. Reaction of the enolate of 10
(3)
product ratios were more easily obtained from the GC chro-
matograms, but an NMR internal standard was still used in each
case to ensure that the total yield of the three products was
quantitative (100 ( 5%). Again, data are normalized to a total
yield of 100%. Authentic samples of products were made by
reactions similar to those in eq 2.
(1)

Stereoselection at the Steady State
J. Am. Chem. Soc., Vol. 120, No. 2, 1998 345
Table 1. Reduction of Diiodide 8A with Ph3SnH
is greatly simplified by making the assumption that the pairs
of fast and slow rate constants are equal: kf1 ) kf2 * ks1 ) ks2.
This assumption seems intuitively reasonable. The only dif-
ference between the first and second cyclizations is that an
iodomethyl (or bromomethyl) group is changed to a methyl
group (compare 17x to 21x and 17n to 21n). It seems unlikely
that this change will have a large effect on the rate constant.
Furthermore, the satisfactory fitting of the data to the theoretical
model (see below) provides additional support for this assump-
tion.
entry
Ph3SnH/8a
[Ph3SnH]
9-exo
9-endo
13
1
2
3
4
5
6
7
8
2.6
15.5
12.1
12.9
10.2
13.6
11.3
15.2
0.02
0.10
0.24
0.50
0.74
1.01
1.49
2.01
53.0
54.1
56.6
59.6
60.2
57.5
54.3
52.1
47.0
45.5
41.1
28.8
26.0
20.2
19.7
15.1
0.4
2.3
11.6
13.8
22.3
26.0
32.8
After the two fast and slow rate constants are set as equal,
the product functions in eqs 4-6 can then be used for the exo,
endo, and reduced products, as derived in the prior paper.
The data for reduction of 8a are representative and are shown
in Table 1. Tables 2 and 3 in the Supporting Information
contain the data for 8b and 14. The data are plotted in Figure
6. These data show a number of unusual features for group
selective processes. The ratio of exo to endo products at low
tin hydride concentration is close to 50/50 in all three cases
(entry 1, Table 1). This confirms that there is no selectivity in
the abstraction of bromine or iodine by the tin radical, as
expected. This also shows that there is no equilibration of
radicals by bimolecular iodine transfer, a conclusion that can
likewise be reached by comparing relative rates (radical cy-
clization is much faster than bimolecular iodine transfer).12 As
the tin hydride concentration increases, the yield of exo products
begins to increase and the yield of endo products declines.
Concomitantly, the doubly reduced products begin to grow in.
After reaching a maximum in the vicinity of 60%, the yield of
the exo product begins to decline, but the yield of the endo
product declines more steeply, so the exo/endo ratio increases
toward infinity as the combined yield goes to zero.
kfast
kfast
1
2
[exo] )
+
‚
{
kfast + kH[SnH] kfast + kH[SnH]
kH[SnH]
(4)
(5)
(6)
(
)}
)}
kslow + kH[SnH]
kslow
kslow
1
2
[endo] )
+
‚
{
kslow + kH[SnH] kslow + kH[SnH]
kH[SnH]
(
kfast + kH[SnH]
kH[SnH]
kH[SnH]
1
2
[reduced] )
‚
+
{
(
)
kslow + kH[SnH] kfast + kH[SnH]
kH[SnH]
kH[SnH]
These data were analyzed within the mechanistic model
shown in Figure 5 for the cyclohexane substrates. This is
identical with the diastereoselective process presented in the
preceding paper, so it is not discussed extensively. Briefly, at
low tin hydride concentration, both the faster and the slower
cyclizations of the two initially generated radicals 17x and 17n
are faster than reduction by tin hydride. At this limit, the
diastereoselectivity is controlled by the halogen abstraction step;
since this is not selective, a 50/50 ratio of products forms. As
the tin hydride concentration increases, both the faster 17x and
slower 17n cyclizing radicals begin to be competitively trapped
by tin hydride, opening new pathways that converge to the
cyclized products or produce the doubly reduced product 13.
When the faster cyclizing radical 17x happens to be reduced
faster than cyclization, most of the resulting product 19
ultimately ends up at double reduction. But when the slower
radical 17 is reduced, the resulting product 20 mostly follows
the pathway to the major product 9-exo. In this scheme, the
yield of 9-exo is then being both supplemented and eroded. But
since the concentration of the slower cyclizing radical 17n
always exceeds that of the faster cyclizing radical, the yield of
9-exo is supplemented faster than it is eroded.
‚
(
)}
kfast + kH[SnH] kslow + kH[SnH]
A simple analysis of rate constant ratios with estimated values
can now be made visually by using the functions derived in the
prior paper as implemented by a program like MathCad. By
“guessing” values of rate constant ratios (plotted against an
arbitrary tin concentration), one finds (not shown) that the data
nicely fit the predicted behavior for a kf/ks ratio in the range of
3-5. This is considerably lower than predicted in the “design”
of the substrates outlined in the Introduction. Application of
actual tin hydride concentrations to the plots then gives estimated
rate constants. Although clearly supporting the analysis, this
trial and error process is not satisfactory from the standpoint of
accuracy and error estimation.
An independent computational approach was taken based on
this trial and error approach. We first constructed an error
function to calculate the agreement between the observed
product ratios and the calculated product ratios for arbitrarily
chosen values of kf and ks at a given tin hydride concentration.
We then constructed a second function which summed the
calculated error of all entries in Tables 1-3 for a given function
and reported the total error. This “total error function” defines
a surface whose range is all reasonable values of kf and ks and
whose minimum indicates the most accurate choice of these
values for the observed data. This type of fitting was compu-
tationally much more intensive and was accomplished on a
UNIX cluster composed of four DEC system 5000’s running a
Mathematica 2.2 kernel under Ultrix 4.2A. Full details are
provided in the Supporting Information along with a representa-
tive plot of this surface for the cyclization of 8a.
We tried a number of approaches to extract rate constants
from the data in these experiments. Newcomb’s recommended
rate constant for Ph3SnH was used for kH.13 Although some
efforts were made to analyze all four rate constants (kf1, kf2,
ks1, ks2) independently, this analysis is quite complicated, and
it gave results that were intuitively unsatisfactory. The analysis
(12) Rate constants for iodine transfer between primary alkyl radicals
are more than an order of magnitude lower than those from Ph3SnH, and
Ph3SnH is also used in large excess. (a) Newcomb, M.; Sanchez, R. M.;
Kaplan, J. J. Am. Chem. Soc. 1987, 109, 1195. (b) Newcomb, M.; Curran,
D. P. Acc. Chem. Res. 1988, 21, 206. (c) Drury, R. F.; Kaplan, L. J. Am.
Chem. Soc. 1972, 94, 3982.
The rate constants kf and ks derived from this analysis for all
three substrates are shown in Figure 6. These calculated values
can then be substituted into eqs 4-6 to generate theoretical lines
(13) Newcomb, M. Tetrahedron 1993, 49, 1151.

346 J. Am. Chem. Soc., Vol. 120, No. 2, 1998
Curran et al.
Figure 5. Mechanism of the topological group selective process.
for the yields of the corresponding products (exo, endo,
reduced), and these lines are plotted against the actual data points
in the graphs in Figure 6. It can be seen that the agreement
between calculation (lines) and experiment (points) is excellent.
The divergence of some points at very high tin hydride
concentrations could be caused by a number of sources (for
example, the pairs of rate constants may be similar, but not
equal), but it may simply be an experimental problem; at these
very high tin hydride concentrations (>50 vol % tin hydride),
the concentration values are not very accurate. During the
course of the research, this type of data analysis was actually
conducted on incomplete data sets, and the predictions for
missing values were confirmed by subsequent experiments.
observed. Even at this high concentration, the combined yield
of cyclic products 15 is still 79%.
It is also interesting to compare the results of the dibromide
8b and the diiodide 8a. Within our analysis, these should
provide identical rate constants even though the first pair of
radicals (17x,n) is different (X ) Br or I). This is because the
second pair of radicals (21x,n) is the same and because we
assumed that the first and second pair had identical rate constants
for cyclization. In our view, the differences between these pairs
of numbers are too small to attribute an origin to the apparent
error. The differences between these rate constants are not large
based on standard indirect kinetic methods, and our analysis is
even less direct than usual. While there could be small
differences between the rate constant for cyclization of iodom-
ethyl- and bromomethyl-substituted radicals 17x,n (X ) Br or
I), it is not safe to attribute the different rate constants obtained
in 8a and 8b to this cause as opposed to any of a number of
other experimental or analytical error factors.
There are several interesting features about the “best” rate
constants (Figure 6) that emerge from the total error analysis.
As deduced from the visual fitting, the ratios of rate constants
are about 3-5; not 15 as predicted. Within the big picture of
radical cyclizations, these rate constants are rather high. The
parent hexenyl radical cyclizes with kc ) 106 s-1 at 80 °C. Even
the slowest cyclizations in Figures 6 are more than 1 order of
magnitude faster than this, no doubt due largely to the decreased
rotational freedom of these radicals. Somewhat surprising is
the observation that the cyclization to make the more strained
bicyclooctane 15 (from 14) is faster than the cyclizations to
make the bicyclononanes 9 (from 8). This trend is outside of
experimental error, and can easily be seen by comparing actual
data points. The last experimental point for the cyclization of
14 is virtually in neat tin hydride, and this is just barely past
the concentration at which the maximum yield of 15-exo is
To provide an independent confirmation of the rate constant
analysis, we undertook the synthesis and kinetic analysis of the
“half reaction” substrates shown in eq 7. These substrates are
identical with those formed in the “second half” of the group
selective reaction whenever the initial radical is reduced. If
the first and second pair of cyclization rate constants are indeed
identical, then the rate constants measured for these substrates
should be identical with those in Figure 7. Ideally, the
diastereomeric radical precursors 19 and 20 (or 23 and 24) could
be studied individually to provide unambiguous measurements
of each of the slow and fast rate constants. In practice, it was

Stereoselection at the Steady State
J. Am. Chem. Soc., Vol. 120, No. 2, 1998 347
Figure 6. Fitting the data in Tables 1-3 with eqs 4-6. Data points were taken from Tables 1-3. Lines for exo, endo, and reduced products were
calculated from eqs 4-6 with the indicated rate constants (in s-1). Product ratios are plotted against tin hydride concentration [M].
not possible to separate the diastereomers, so kinetic experiments
had to be conducted on mixtures.
(8)
Knowing the ratio of the starting diastereomers (57/43), the
rate constants for the fast and slow cyclizations in each mixture
can then be extracted from the data by a relatively straightfor-
ward process that treats these as two independent reactions that
give one product which is different (the cyclic product) and
one product which is the same (eq 7). In this analysis, each
individual data point provides a pair of rate constants that were
then averaged over all the points to provide the “best estimate”
rate constants shown in Figure 7. The agreement of these rate
constants with those calculated in Figure 6 is only fair. In the
case of the cyclohexene substrate, the calculated rate constants
for the “half reaction” system are about 2-3 times lower than
those from the full doubly convergent system, although the rate
constant ratio of 4 is within the expected range. In the case of
the cyclopentene system, there is good agreement with kf, but
ks in the half reaction system is again low by about a factor of
2. Given the assumptions and the complexity of the systems
involved, we feel that this level of agreement is reasonable.
Computational Analysis of Transition States. The studies
with these substrates provided strong support for the model of
group selection detailed in the preceding paper and in Figure 5
of this paper. But they also firmly showed that the actual level
of selectivity in the group selective reactions of substrates 8
(7)
The substrates were prepared as shown in eq 8 as 57/43
diastereomeric mixtures. These results imply that there is a
small kinetic resolution in the alkylation step. These mixtures
were reduced in preparative experiments at low tin hydride
concentrations to give 57/43 mixtures of exo/endo products. No
doubly reduced products were formed. Kinetic experiments
were conducted as described above, and the data for these
experiments are shown in Tables 4 and 5 in the Supporting
Information.

348 J. Am. Chem. Soc., Vol. 120, No. 2, 1998
Curran et al.
Figure 7. Comparison of full-reaction with half-reaction rate constants (in s-1).
Figure 8. Models for transition state calculations.
with two precursors and one acceptor was significantly lower
(kf/ks ) 4) than that predicted by the model cyclization with
two acceptors and one precursor (kf/ks ) 15). We undertook a
series of calculations to gain insight into the reasons for the
reduced selectivity and with the ultimate goal of designing a
more selective substrate.
is not generally a good assumption, but in this specific case it
should be reasonable since the ground states only differ by the
interchange of a “CH3” and a “CH2 ” group substituted on a
•
common stereogenic carbon atom.
The calculations are described in detail and analyzed more
fully in the Supporting Information. Here we present only the
salient features for analysis of the group selection process along
with the predictions and conclusions that emerge. Briefly,
random geometries of the transition states were generated with
Monte Carlo simulation and optimized by using the Houk-
Spellmeyer molecular mechanics parameters for radical cy-
clization. Molecular orbital descriptions were created from the
resulting geometries by ab initio techniques, and the transition
state nature of the geometries was confirmed by quantum
mechanics frequency calculations of the molecular orbital
descriptions. Predicted isomer ratios were then generated by
applying Boltzmann distribution functions to all transitions
structures within 5 kcal/mol of the lowest energy one.
To decrease the number of possible transition states, we
exchanged the ester group (which has several possible rotational
isomers) for a methyl group (which has only one). This renders
the results of the calculations more qualitative; however, the
agreement of theory and calculation turned out to be remarkably
good. The substrates studied are shown in Figure 8. Dienes
25x and 25n are conformational isomers and are direct models
of 1 (Figure 4); however, they are drawn throughout the analysis
as enantiomers to facilitate a constant view of the cyclohexa-
diene ring. Radicals 26x and 26n model the cyclizations in
the group selective process. Therefore, these radicals are not
conformational isomers but diastereoisomers. The direct com-
parison of the transition state energies of diastereomers is usually
not especially meaningful because the transition states are not
in competition with each other. However, the presence of tin
hydride places these two transition states in competition in the
process in Figure 5, so we simply compare the calculated
diastereomeric transition state energies directly as if they were
conformational isomers. These calculations do not provide
direct isomer ratios, but they do provide rate ratios that can be
compared to the experimentally measured rate constants. Inher-
ent in this comparison is the assumption that the ground states
of the two diastereomers 26x and 26n are equal in energy. This
The two lowest energy transition states for 25 are shown in
Figure 9. The lowest energy transition structure leads to the
exo product, as expected from the results in Figure 2: the
forming cyclopentane ring is “chairlike” with an equatorial
methyl group (hereafter called “chair-Meeq”). Correspondingly,
the higher transition structure (TS) leads to the endo product
and is of the chair-Meax type. A corresponding “boat-Meeq”
TS leading to the endo product along with two other similar
transition states leading to the endo product were also located
(not shown). No other TSs leading to the exo product were

Stereoselection at the Steady State
J. Am. Chem. Soc., Vol. 120, No. 2, 1998 349
A straightforward prediction arises from these calculations:
destabilizing TS structure 26na should increase the stereose-
lectivity. This was done first in computations and then in
experiments by placing a tert-butyl group in the 4-position trans
to the side chain bearing the radical. This will destabilize the
lower energy endo TS (26na) and the higher energy exo TS
(26xb), which would now require axial tert-butyl groups.
This simple analysis was reflected in the calculations of 27x
and 27n. Although eight different transitions were located (four
for 27x and four for 27n), only two of these were close enough
in energy to contribute significantly to the product distribution.
These are shown in Figure 11. As anticipated from the analysis,
the calculated ratio of rates has now increased to 8/1.
Figure 9. Low energy transition structures of 25.
It was of special interest to calculate the results of the cis
stereoisomer 28, and these results are summarized in Figure
12. In the simple analysis, the favored transition states become
28xb and 28na, and this predicts that the selectivity in the
cyclization should be reversed. Although this may seem
surprising, it is actually expected. Analogy has been drawn
between related bicyclic transition states and cis-decalins.14 The
flipping of the six-membered ring (induced by the tert-butyl
group) also flips the forming five-membered ring and reverses
the product configuration. The calculations with 28 provided
four exo TSs and two endo TSs, but the lowest one of each
taken together (see Figure 12) accounted for more than 94% of
the products. The calculated product ratio is 5/1 and the endo
isomer is indeed predicted to be favored.
Experimental Testing of Computational Predictions. The
requisite samples to test the computational predictions were
prepared as shown in eq 9. Alkylation of the enolate derived
Figure 10. Low energy transition structures of 26.
located. The predicted Bolzmann distribution of exo/endo
products is 14/1, which is remarkably close (perhaps fortuitously
close, considering the structure change) to the observed ratio
with 1.
In contrast, analysis of the group selective system 26 provided
seven transition structures. Of these, the two (of three) lowest
energy structures of 26x are shown in Figure 11 along with the
two (of four) lowest energy structures from 26n. Working with
only these four structures (inclusion of the three higher energy
TSs has little effect), the predicted ratio of exo to endo products
is 1.6/1. The calculations clearly reflect the decreased selectivity
in moving from 25 to 26, although now the calculated level of
selectivity is lower than that observed (about 4/1).
(9)
Both TSs leading to the exo product are of the “chair-Meeq”
type; they differ in the direction of twist of the cyclohexene
ring. In the lower energy transition state 26xa, the side chain
bearing the radical is pseudoequatorial while the higher energy
TS 26xb has this group pseudoaxial. The TSs leading to the
endo product are both “boat-Meeq” with different half chair
conformations of the cyclohexene: 26na has a pseudoequatorial
side chain while 26na is pseudoaxial. The radical transition
states with the “chair-Meax” geometry are prohibitively high in
energy because they place the endo-Me group directly over the
cyclohexene.
from 34 with the various alkylating agents gave mixtures of
cis and trans isomers of the corresponding products 29-32. In
(14) Rajanbabu, T. V. Acc. Chem. Res. 1991, 24, 139.

350 J. Am. Chem. Soc., Vol. 120, No. 2, 1998
Curran et al.
Kinetic experiments with both 30-trans and 30-cis were
conducted as before, and the data for these experiments are
shown in Tables 6 and 7 in the Supporting Information and
plotted with the usual analysis in Figure 13. The data in Figure
13 clearly show that the predictions of the calculations have
been borne out in experiment. The plot of the 30-trans shows
increased selectivity compared to the parent 8. For example,
at 0.46 M concentration of tin hydride, the yield of 33-trans-
exo has increased from 50% to its maximum of 74% while the
yield of 33-trans-endo has declined from 50% to 19%; the
remaining 7% is the double reduction product 31. The rate
constant kf leading to the exo product is almost the same as
that in the parent, while the rate constant ks is lower than that
in the parent. The experimental ratio kf/ks of 11 is higher than
the parent and is close to the ratio of 8 predicted by the
calculations.
Figure 11. Low energy transition structures of 27.
Perhaps even more impressive are the results with 30-cis.
As predicted by the calculations, the endo isomer is indeed the
favored product; however, the predicted selectivity of 5 is not
borne out by experiment. The ratio of the two rate constants is
only about 2, and this is reflected in the gentle curvature of the
lines in Figure 13.
The rate constants of the corresponding half-reaction models
were again measured on mixtures of isomeric precursors 29-
cis and 29-trans. The process was similar to that described
above and the full details are provided in the Supporting
Information. Figure 14 shows the “best” rate constants that
emerge from this analysis. In the case of the 29-trans isomer,
the half-reaction rate constants agree reasonably well with the
rate constants in Figure 13. In the case of 29-cis, however, the
two rate constants are calculated to be about equal. We feel
that this is unlikely to be true and it is probably an error caused
by the fact that the two rate constants in the mixture being
measured are too close to resolve. Ideally, it would be better
to separate the two diastereomers of 29-cis and 29-trans and
measure the half reaction rate constants separately; however,
in no case was separation evident during analytical or preparative
chromatography of the precursors.
Figure 12. Low energy transition structures of 28.
all cases, these isomers were separable by chromatography.
Details of the synthesis, separation, and characterization of all
these products are contained in the Supporting Information.
Isomers 30-trans and 30-cis were first cyclized in preparative
experiments at low tin hydride concentrations to provide
authentic samples of the respective exo and endo products
(eq 10). Both reactions provided 50/50 mixtures of diastere-
Last, the two model substrates 32-cis were prepared to study
the effect of the methyl group on the rate constant for
cyclization. There is no issue of stereoselection in these
molecules (aside from the formation of a cis/trans ring fusion);
the analysis is made by comparing absolute rate constants. The
absolute rate constants are easily measured by standard competi-
tion kinetics against tin hydride (data in Supporting Information)
and are shown in Figure 15. Interestingly, both model
compounds cyclize significantly more slowly than the slower
of the two methyl-bearing diastereomers. Thus, in a formal
sense, the replacement of either hydrogen at C2 of the propyl
chain of 29 accelerates the radical cyclization; selectivity arises
because one cyclization is accelerated more than the other. This
trend presumably has its origins in the “Thorpe-Ingold” effect.
In addition, we learn from these models that there is a very
slight preference for cyclization of an equatorially oriented side
chain over an axially oriented one. It is this low preference
that is responsible for the relatively low selectivities obtained
with substrates 8.
(10a)
(10b)
omeric cyclic products 33 in excellent purified yields and free
from the doubly reduced product 31. The exo/endo configu-
ration of the cis isomers was determined by hydrolysis of these
compounds to the corresponding crystalline acids 37, whose
structures were solved by X-ray crystallography (see Supporting
Information). These crystal structures also proved that the
configuration in the alkylation was assigned correctly. The exo/
endo configuration of the trans isomer of 33 was assigned by a
series of qualitative spectroscopic correlations between the
known compounds (9-exo and 9-endo) and the corresponding
isomers of 33-cis.
Conclusions
The experimental results described in this paper strongly
support the model for compound stereoselection at the steady
state that was put forth in the preceding paper.1 The model
has been used both qualitatively to interpret the stereoselection

Stereoselection at the Steady State
J. Am. Chem. Soc., Vol. 120, No. 2, 1998 351
Figure 13. Fitting the data in Tables 6 and 7 with eqs 4-6. Data points were taken from Tables 6 and 7. Lines for exo, endo, and reduced
products were calculated from eqs 4-6 with the indicated rate constants (in s-1).
a
Figure 15. Model cyclizations lacking the methyl group.
reactions and/or catalytic reagents present significant challenges
Figure 14. Comparison of full-reaction rate constants with half-reaction
to experimental radical chemistry and other fields.
rate constants (in s-1).
Acknowledgment. We thank the National Science Founda-
tion for funding this work. Dr. Jo¨rg Junggebauer thanks the
Deutsche Forschungsgemeinschaft for a postdoctoral fellowship.
We thank Dr. Marcelo Preite for help in preparing the
manuscript.
as a function of tin hydride concentration and quantitatively to
extract relevant rate constants. The measured rate constants
provide insights into the factors that control both rate and
stereoselectivity in radical cyclizations to cycloalkenes. More
importantly, the success of the kinetic model in interpreting the
results leads us to conclude that stereoselection at the steady
state is a general process, and that the concepts and kinetic
equations outlined in part 1 can now be used to design substrates
and interpret results for other kinds of complex steady state
processes. The type of process outlined in this paper is by far
the simplest, and other processes involving only bimolecular
Supporting Information Available: Full experimental and
characterization data for all compounds, details of the error
analysis and calculations, and details of the two crystal structures
(55 pages). See any current masthead page for ordering and
Internet access instructions.
JA9724102