Anodic coupling of carboxylic acids to electron-rich double bonds: A surprising non-Kolbe pathway to lactones

Article abstract of DOI:10.3762/bjoc.9.186

Carboxylic acids have been electro-oxidatively coupled to electron-rich olefins to form lactones. Kolbe decarboxylation does not appear to be a significant competing pathway. Experimental results indicate that oxidation occurs at the olefin and that the reaction proceeds through a radical cation intermediate.

Full text of DOI:10.3762/bjoc.9.186

Anodic coupling of carboxylic acids to electron-rich
double bonds: A surprising non-Kolbe
pathway to lactones
Robert J. Perkins, Hai-Chao Xu, John M. Campbell and Kevin D. Moeller*
Full Research Paper
Open Access
Address:
Beilstein J. Org. Chem. 2013, 9, 1630–1636.
Washington University in Saint Louis, Saint Louis, Missouri 63130,
United States
doi:10.3762/bjoc.9.186
Received: 12 June 2013
Accepted: 18 July 2013
Published: 09 August 2013
Email:
Kevin D. Moeller* - moeller@wustl.edu
* Corresponding author
This article is part of the Thematic Series "Organic free radical chemistry".
Guest Editor: C. Stephenson
Keywords:
carboxylic acid; cyclization; electrolysis; free radical; kolbe; radical
cation
© 2013 Perkins et al; licensee Beilstein-Institut.
License and terms: see end of document.
Abstract
Carboxylic acids have been electro-oxidatively coupled to electron-rich olefins to form lactones. Kolbe decarboxylation does not
appear to be a significant competing pathway. Experimental results indicate that oxidation occurs at the olefin and that the reaction
proceeds through a radical cation intermediate.
Introduction
Anodic cyclization reactions of the general type shown in these examples, the reactions can be viewed as arising from an
Scheme 1 can provide a powerful method for the synthesis of a oxidation that forms an olefinic radical cation that is then
variety of ring systems [1,2]. The reactions effectively reverse rapidly trapped by a nucleophile. This triggers a cascade of
the polarity of an electron-rich functional group and in so doing reactions that typically involve the loss of two protons (or a
open up entirely new opportunities for bond formation. In addi- silyl group), a second oxidation of a radical intermediate, and
tion, the oxidative reactions lead to products that either preserve solvent trapping. This reaction cascade leads to formation of the
or increase the level of functionality found in the initial sub- final product.
strate. This provides synthetic handles for further development
of the cyclic products generated.
A combination of computational studies and competition exper-
iments has begun to identify both cyclization reactions that
Such reactions can be used to build a variety of fused, bridged, proceed through different mechanistic pathways and cycliza-
and spirocyclic carbocyclic systems [3], cyclic amino acid tion reactions where the cascade of reactions downstream from
derivatives [4], cyclic ethers [5,6], and lactones [7,8]. In most of the cyclization plays an important role in product formation [9].
1630

Beilstein J. Org. Chem. 2013, 9, 1630–1636.
cyclization. Density functional theory (DFT) calculations
suggested that sulfonamide based cyclizations proceed by addi-
tion of the N-localized radical to the electron-rich double bond,
a reaction that led to product 8. In a competitive process, an
intramolecular electron-transfer reaction occurred that led to
formation of a radical cation from the olefin 7 and regeneration
of the sulfonamide anion. Formation of a radical cation from the
olefin led to the possibility of alcohol trapping and the forma-
tion of cyclic ether product 9. Experimental data suggested that
the alcohol trapping pathway provided the kinetic product but
that the cyclization was reversible. The sulfonamide radical
pathway afforded the thermodynamic product. In a very
interesting development, we found that the reaction could be
shifted toward the alcohol-trapping product by increasing the
current passed through the cell. The increased current appeared
to accelerate the removal of a second electron from an initial
Scheme 1: General scheme for anodic cyclization reactions.
For example, consider the competition experiment highlighted cyclic product like 3 (Scheme 1), a change that reduced the re-
in Scheme 2. In this experiment, an electron-rich olefin was versibility of the cyclization.
coupled to one of two competing nucleophiles, a sulfonamide
and an alcohol (5). When the oxidation was run with 2,6-luti- Clearly, the use of basic reaction conditions with the acidic sub-
dine as a base (not shown), the reaction led to the formation of a strate led to dramatic mechanistic changes to our early, simpli-
radical cation from the electron-rich olefin followed by fied view of the reactions. It is tempting to suggest that similar
trapping by the alcohol nucleophile to form product 9. No prod- situations would arise with other acidic substrates. For example,
uct from sulfonamide trapping (8) was observed. However, would the oxidation of a substrate having a carboxylic acid and
when the reaction was run under more basic conditions an electron-rich olefin lead to a carboxy radical pathway or a
(0.5 equivalents of lithium methoxide) cyclic voltammetry data radical cation type reaction? To date, we have been hesitant to
indicated that the oxidative cyclizations were initiated by the try such an experiment, and in fact we typically avoid substrates
oxidation of sulfonamide anion. This led to initial formation of with carboxylic acid functional groups [5]. This action was
N-localized radical 6. Depending on the reaction conditions, the taken because of the well-known Kolbe electrolysis reaction
experiment favoured either a sulfonamide or alcohol based (Scheme 3) [10,11]. In the Kolbe electrolysis (Scheme 3, reac-
Scheme 2: Anodic cyclization competition study.
1631

Beilstein J. Org. Chem. 2013, 9, 1630–1636.
minimize the risk of decarboxylation and lead to carboxylate
trapping of the radical cation to again form the cyclic radical 12.
A third mechanistic possibility is a direct oxidation of the
elctron-rich olefin to form radical cation 11 followed by
carboxylate trapping. All three mechanistic variations would
then require a successful single-electron oxidation of 12 to ulti-
mately lead to 13, the product of a direct oxidative coupling
between the carboxylic acid moiety and the electron-rich olefin.
Scheme 3: Kolbe electrolysis reactions.
We report here that this is the case and that the direct oxidative
tion 1), a carboxylic acid is oxidized. A decarboxylation reac- coupling of a carboxylic acid and an electron-rich olefin can be
tion then leads to the formation of a radical that is subsequently accomplished in good yield. In all cases, the reactions appear to
trapped by a second radical formed in solution. The reaction has proceed through an olefinic radical cation intermediate.
been used to form dimers [12], as well as in some cases cyclic
Results and Discussion
products (Scheme 3, reaction 2) [13-15].
Initial cyclization studies
However, the chemistry highlighted in Scheme 2 suggests that a The compatibility of the carboxylic acid with the anodic
Kolbe-type decarboxylation reaction might not interfere at all cyclization was initially studied by examining the anodic oxi-
with an oxidative coupling reaction between a carboxylic acid dation of substrates 14a–c (Table 1). The oxidation of 14a
and an electron-rich olefin. Examples in the literature of the nicely afforded the five-membered ring product using either
coupling of carboxylic acids to aromatic rings, though limited in lithium methoxide or 2,6-lutidine as a base. Clearly, decarboxy-
scope, were also encouraging [16-20].
lation of the carboxylate anion was not a problem. In fact,
cyclic voltammetry data suggest that the reaction originated
There are three mechanistic possibilities which would allow for from an oxidation of the ketene dithioacetal. The oxidation
a successful reaction. First, if the carboxy radical 10 is formed, potential (Ep/2) for 14a was measured to be +0.68 V versus
then it may simply add to the electron-rich olefin to give 12 Ag/AgCl. For comparison, the oxidation potential measured for
faster than it undergoes a decarboxylation reaction (Scheme 4). 10-undecenoic acid was +1.91 V versus Ag/AgCl in DMF
solvent and +1.85 V versus Ag/AgCl in acetonitrile. When
Second, it is possible for a carboxy radical like 10 to undergo an 0.5 equivalents of benzyltrimethylammonium hydroxide was
intramolecular electron-transfer reaction analogous to that added to the cyclic voltammetry solutions in order to generate
observed in the case of the sulfonamide-based radical. One the carboxylate and mimic the preparative oxidation conditions
might suspect the electron-transfer to form radical cation 11 to used for the reaction originating from 14a, the oxidation poten-
be even more favorable in the present case due to the stability of tial of the 10-undecenoic acid fell to 1.36–1.40 V versus
the resulting carboxylate anion. The electron-transfer would Ag/AgCl for the DMF solution and 1.38 V versus Ag/AgCl for
Scheme 4: Oxidative coupling between a carboxylic acid and electron-rich olefin.
1632

Beilstein J. Org. Chem. 2013, 9, 1630–1636.
Table 1: Anodic coupling of ketene dithioacetals and carboxylic acids.
Substrate
n
Ep/2a
Base (0.5 equiv)
Yield (15)
14a
14a
14b
14b
14c
14c
1
1
2
2
3
3
0.68
LiOMe
87%
74%
0%b
72%
0%c
0%
2,6-lutidine
LiOMe
0.71
1.06
2,6-lutidine
LiOMe
2,6-lutidine
aCyclic voltammetry data were obtained relative to a Ag/AgCl reference electrode with a sweep rate of 50 mV/s. b87% of the ring opened methyl ester
16b was obtained. c30% of the ring opened methyl ester 16c was obtained.
the acetonitrile solution. The carboxylate was only partially methoxide as the base, no cyclic product was observed when
soluble in acetonitrile. While the formation of the carboxylate 2,6-lutidine was employed. The reactions were quite messy and
significantly lowered the oxidation potential of the acid moiety, consistent with the formation of a radical cation followed by a
it did not lower it close to the oxidation potential measured for cyclization that was too slow to compete with decomposition.
substrate 14a or even the oxidation potential of a ketene dithio- The cyclic voltammetry data obtained were consistent with this
acetal (ca. +1.06 V versus Ag/AgCl) [21]. Hence, the oxidation observation. The +1.06 V versus Ag/AgCl oxidation potential
potential measured for 14a is most consistent with a fast measured for substrate 14c was the same as that measured for
cyclization originating from oxidation of the olefin. Fast the ketene dithioacetal in the absence of a trapping group [21].
cyclization reactions are known to significantly lower the oxi-
dation potential of a ketene dithioacetal. For example, the trap- The anodic coupling of a carboxylic acid group with a vinyl
ping of a ketene dithioacetal derived radical cation by an amine sulfide and an enol ether were also examined (Table 2). As with
was shown to lower the oxidation potential measured for the earlier alcohol and amine based cyclizations, reactions with the
ketene dithioacetal by 460 mV [21], a value consistent with the vinyl sulfide coupling partner proceeded much better than did
potential measured here for 14a.
their enol ether counterparts [21,22]. In the previous cases, the
argument was made that less polar radical cations underwent
The oxidation of 14b to afford the six-membered ring product better trapping reactions with heteroatomic nucleophiles [23],
also proceeded well. When 2,6-lutidine was used as the base for and a similar argument can be made here.
the reaction, a 72% isolated yield of the desired product was
obtained. In this case, the use of the more nucleophilic lithium
Table 2: Extension to other electron-rich olefins.
methoxide as the base led to cyclization followed by opening of
the lactone ring to form methyl ester product 16b. This product
did have the olefin functionalized with an alcohol and a
methoxy group indicating that the initial oxidative cyclization
had occurred. In this case, the slightly higher oxidation poten-
tial measured for the substrate (+0.71 V versus Ag/AgCl) is
consistent with a slightly slower cyclization reaction.
substrate
-X
Yield
Attempts to generate seven-membered ring products using the
oxidative cyclization were not successful. While 30% of a
methoxide opened product 16c could be obtained with lithium
17a
17b
-SMe
-OMe
74%
66%
1633

Beilstein J. Org. Chem. 2013, 9, 1630–1636.
The oxidation potential for the vinyl sulfide used in substrate (Ep/2 = +1.38 in acetonitrile) since the Ep/2 value measured for
17a is +1.08 V versus Ag/AgCl [9] and the oxidation potential substrate 19a under neutral conditions was +1.52 V versus
of the enol ether in substrate 17b is +1.18 versus Ag/AgCl [9]. Ag/AgCl. Several sets of conditions were attempted to achieve
Both oxidation potentials are lower than the potential measured the cyclization. However, in no case was the yield of cyclic
for the carboxylate suggesting that both reactions proceed product obtained more than 33%. Proton NMR data taken of the
through the olefinic radical cation. While an intramolecular crude reaction mixture showed no evidence for decarboxylation.
electron-transfer to form the carboxy radical can occur, the Instead, all products appeared to arise from a styrene derived
difference in oxidation potentials suggests that the equilibrium radical cation, an intermediate that might be formed by an
would lie on the side of the oxidized olefin.
intramolecular electron-transfer. Furthermore, the recovery of
starting material after 2 F/mol of current had been passed
through the reaction mixture indicated poor current efficiency.
Styrene substrates and additional insights
With the ketene dithioacetal, vinyl sulfide, and enol ether Why might an increase in the oxidation potential of the olefin
substrates, it appeared clear that the initial oxidation was occur- hurt the reaction?
ring at the olefin instead of at the carboxylate. We wondered
how the reaction might change if the olefin had a higher oxi- Past experience suggests that in such cases the low yield of
dation potential. With this in mind, a series of carboxylic acid product obtained might arise from either a slow initial cycliza-
substituted styrene substrates were examined (Table 3).
tion or the formation of a stable cyclic radical that then strug-
gles to undergo the second oxidation. Both pathways might lead
The first substrate studied was the simple styrene derivative 19a to the formation of a polymer or decomposition, and in so doing
(R = H). Cyclic voltammetry suggested that in the presence of lower the mass balance for the reaction as observed. Also, diffu-
base the initial oxidation should occur at the carboxylate sion of a stable radical or radical cation to the cathode would
Table 3: Extension to styrene derivatives.
Substrate
R
n
Ep/2a
Base/Temperature
F/mol
Yield
19a
19a
19a
19b
19c
19c
19d
19e
19e
19f
H
1
1
1
3
1
1
3
1
1
3
1
1
1
3
1
1
1
3
1.52
1.52
1.52
1.73
1.31
1.31
1.42
1.39c
1.39c
1.39c
1.42c
1.42c
1.42c
1.50
1.09
1.09
1.09
1.11
0.5 equiv LiOMe/rt
0.5 equiv LiOMe/rt
None/40 °C
NAb
2
15%
27%
33%
NA
H
10
10
NA
2
H
H
4-OMe
4-OMe
4-OMe
2-OMe
2-OMe
2-OMe
3-OMe
3-OMe
3-OMe
3-OMe
2,4-OMe
2,4-OMe
2,4-OMe
2,4-OMe
0.5 equiv LiOMe/rt
None/40 °C
NA
56%
76%
NA
2
NA
2
0.5 equiv LiOMe/rt
None/40 °C
NA
48%
59%
NA
2
NA
2
19g
19g
19g
19h
19i
0.5 equiv LiOMe/rt
0.5 equiv LiOMe/rt
None/40 °C
NA
4% (NMR)
35%
23%
NA
10
10
NA
2
0.5 equiv LiOMe/rt
None/40 °C
1.0 equiv LiOMe/rt
NA
48%
45%
74%
NA
19i
2
19i
2
19j
NA
aConditions: Substrates were dissolved in acetonitrile to a concentration of 0.025 M in a solution that contained 0.1 M tetraethylammonium tosylate.
Cyclic voltammetry was performed at a sweep rate of 50 mV/s using a platinum anode. Half-wave oxidation potentials were measured versus a
Ag/AgCl reference electrode. bNot applicable. cLithium perchlorate was used as the electrolyte.
1634

Beilstein J. Org. Chem. 2013, 9, 1630–1636.
result in a reduction to give the starting material, a scenario that
would lead to low current efficiency. Of the two mechanistic
explanations for these observations, we believe that a slow
second oxidation step is the problem.
The possibility of a slow cyclization was eliminated by consid-
eration of the observed half-wave oxidation potential measured
for 19a. The oxidation potential (Ep/2) of +1.52 V for 19a
(measured under neutral conditions) was significantly lower
than that measured for 19b (+1.73 V), a similar styrene with a
longer tether between the coupling partners. This observation is
consistent with the oxidation of 19a leading to a fast cycliza-
tion that rapidly removes the radical cation from the electrode
surface, resulting in a lower observed oxidation potential [21].
The seven-membered ring cyclization arising from the oxi-
dation of 19b would be significantly slower, giving rise to a
higher oxidation potential.
Scheme 5: Predicted relative rates of single-electron oxidation based
on resonance stabilization of the resulting cation.
Given the evidence for a fast cyclization reaction, we attributed membered ring counterpart 19h, indicating a fast cyclization.
the challenges associated with the styrene cyclizations to a Hence, the poor yield obtained from the oxidation of 19g was
problem with the second single-electron oxidation. This idea not due to a problem with the first step.
was supported by our next set of experiments. Methoxy-substi-
tuted styrenes 19c (R = 4-OMe), 19e (R = 2-OMe), and 19g Good yields and efficient consumption of starting material were
(R = 3-OMe) were prepared. Substrate 19c has a p-methoxy regained upon moving the methoxy group to the ortho position
substituent positioned perfectly for aiding in the removal of a (19e). Interestingly, 19e showed no oxidation potential drop
second electron from cyclic radical 21a (Scheme 5). Indeed, the from the analogous long-chain substrate 19f. It is tempting to
reactions of 19c proceeded better than the cyclizations origi- assign this to a slow cyclization, but we do not believe that this
nating from 19a, giving acceptable yields and complete is necessarily the case. The relatively low oxidation potentials
consumption of the starting material after 2 F/mol of current for 19e and 19f (as well as the potentials measured for 19i and
had been passed. In this case, the reaction benefited strongly 19j) may be due to an interaction between the o-methoxy
from using less basic reaction conditions and slightly elevated substituent and the radical cation (Figure 1). Such a stabilizing
temperatures. When more basic reaction conditions were used, interaction would be expected to lower the oxidation potential
an overoxidized product derived from the elimination of the of the olefin.
lactone ester to form a methoxy styrene from product 20c was
obtained. The elevated temperatures are known to assist the
intramolecular cyclization. The significantly lower oxidation
potential measured for 19c relative to that measured for its
seven-membered ring counterpart 19d was again consistent
with a fast initial cyclization.
We believe that the primary advantage of the methoxy group is
to facilitate the second electron oxidation, rather than merely
Figure 1: Radical cation stabilization by an o-methoxy substituent.
stabilizing an olefinic radical cation. To demonstrate this, the
position of the methoxy group on the ring was altered systemat-
ically (see Table 3). Cyclic radical 21b, arising from the reac- We also investigated the reactivity of the o,p-dimethoxy substi-
tion of m-methoxy substituted styrene 19g (Table 3), would not tuted styrene 19i towards the coupling reaction. This reaction
be expected to oxidize as readily as radical 21a. Indeed, the also led to a good yield of product and efficient consumption of
anodic oxidation of 19g gave low yields and poor current effi- the starting material. Surprisingly, the optimized conditions for
ciency, mirroring the results obtained from the oxidation of the oxidation of 19i were different from those optimized for the
unsubstituted styrene 19a. As with other substrates, the oxi- oxidation of either 19c or 19e, and the initial yields for the
dation potential of 19g was lower than that of its seven- cyclization of 19i were not good. The cyclization originating
1635

Beilstein J. Org. Chem. 2013, 9, 1630–1636.
7. Brandt, J. D.; Moeller, K. D. Org. Lett. 2005, 7, 3553–3556.
from 19i benefited from the use of more base. It appears that the
use of base reduced the formation of a quinone methide from
the product. For example, when a full equivalent of base was
used for the reaction, the desired product was obtained in 74%
isolated yield.
doi:10.1021/ol051296m
8. Xu, H.-C.; Brandt, J. D.; Moeller, K. D. Tetrahedron Lett. 2008, 49,
3868–3871. doi:10.1016/j.tetlet.2008.04.075
9. Campbell, J. M.; Xu, H.-C.; Moeller, K. D. J. Am. Chem. Soc. 2012,
134, 18338–18344. doi:10.1021/ja307046j
10.Torii, S.; Tanaka, H. Carboxylic Acids. In Organic Electrochemistry, 4th
ed.; Lund, H.; Hammerich, O., Eds.; Marcel Dekker: New York, 2001;
pp 499–543.
It is clear that the success of the styrene derived coupling reac-
tion depends strongly on the position of the methoxy group on
the aromatic ring. Ortho and para substitution leads to a
successful cyclization while meta substitution does not. This
observation is best explained by the reaction being dependent
upon the ease with which the cyclic benzyl radicals 21a–c are
oxidized.
11.Schäfer, H. J. Recent Contributions of Kolbe Electrolysis to Organic
Synthesis. In Electrochemistry IV; Steckhan, E., Ed.; Topics in Current
Chemistry, Vol. 152; Springer-Verlag: New York, 1990; pp 91–151.
12.Schäfer, H. J. Chem. Phys. Lipids 1979, 24, 321–333.
doi:10.1016/0009-3084(79)90117-8
13.Lebreux, F.; Buzzo, F.; Markó, I. E. Synlett 2008, 2815–2820.
doi:10.1055/s-0028-1083547
14.Huhtasaari, M.; Schäfer, H. J.; Becking, L.
Conclusion
Angew. Chem., Int. Ed. Engl. 1984, 23, 980–981.
doi:10.1002/anie.198409801
It was found that carboxylic acid nucleophiles can be coupled to
electron-rich olefins to form lactone products in good yield
without interference from competitive decarboxylation reac-
tions. The reactions are consistent with carboxylate trapping of
an olefin-derived radical cation intermediate and depend
strongly on the ease with which the second oxidation step in the
mechanism occurs. This is particularly true when styrene-based
substrates are used. Finally, optimization of reaction pH
required careful consideration of product stability.
15.Matzeit, A.; Schäfer, H. J.; Amatore, C. Synthesis 1995, 1432–1444.
doi:10.1055/s-1995-4112
16.Scott, A. I.; Dodson, P. A.; McCapra, F.; Meyers, M. B.
J. Am. Chem. Soc. 1963, 85, 3702–3704. doi:10.1021/ja00905a040
17.Coutts, I. G. C.; Edwards, M.; Musto, D. R.; Richards, D. J.
Tetrahedron Lett. 1980, 21, 5055–5056.
doi:10.1016/S0040-4039(00)71131-2
18.Coutts, I. G. C.; Culbert, N. J.; Edwards, M.; Hadfield, J. A.;
Musto, D. R.; Pavlidis, V. H.; Richards, D. J.
J. Chem. Soc., Perkin Trans. 1 1985, 1829–1836.
doi:10.1039/P19850001829
19.Thomas, H. G.; Schwager, H.-W. Tetrahedron Lett. 1984, 25,
4471–4474. doi:10.1016/S0040-4039(01)81469-6
20.Deffieux, D.; Fabre, I.; Courseille, C.; Quideau, S. J. Org. Chem. 2002,
67, 4458–4465. doi:10.1021/jo020023r
Supporting Information
Supporting Information File 1
Procedures for electrolysis and cyclic voltammetry
experiments, characterization of electrolysis products,
procedures for synthesis and characterization of electrolysis
starting materials.
21.Xu, H.-C.; Moeller, K. D. Angew. Chem., Int. Ed. 2010, 49, 8004–8007.
doi:10.1002/anie.201003924
22.Xu, G.; Moeller, K. D. Org. Lett. 2010, 12, 2590–2593.
doi:10.1021/ol100800u
23.Tang, F.; Moeller, K. D. Tetrahedron 2009, 65, 10863–10875.
doi:10.1016/j.tet.2009.09.028
[http://www.beilstein-journals.org/bjoc/content/
supplementary/1860-5397-9-186-S1.pdf]
License and Terms
Acknowledgements
We thank the National Science Foundation (CHE-1151121) for
This is an Open Access article under the terms of the
Creative Commons Attribution License
their generous support of this work.
(http://creativecommons.org/licenses/by/2.0), which
permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.
References
1. Sperry, J. B.; Wright, D. L. Chem. Soc. Rev. 2006, 35, 605–621.
doi:10.1039/b512308a
2. Yoshida, J.; Kataoka, K.; Horcajada, R.; Nagaki, A. Chem. Rev. 2008,
108, 2265–2299. doi:10.1021/cr0680843
The license is subject to the Beilstein Journal of Organic
Chemistry terms and conditions:
3. Moeller, K. D. Synlett 2009, 1208–1218. doi:10.1055/s-0028-1088126
4. Xu, H.-C.; Moeller, K. D. J. Am. Chem. Soc. 2010, 132, 2839–2844.
doi:10.1021/ja910586v
(http://www.beilstein-journals.org/bjoc)
5. Sutterer, A.; Moeller, K. D. J. Am. Chem. Soc. 2000, 122, 5636–5637.
doi:10.1021/ja001063k
The definitive version of this article is the electronic one
which can be found at:
6. Liu, B.; Duan, S.; Sutterer, A. C.; Moeller, K. D. J. Am. Chem. Soc.
2002, 124, 10101–10111. doi:10.1021/ja026739l
doi:10.3762/bjoc.9.186
1636