Face selection in addition and elimination in sterically unbiased systems

(1)

Face Selection in Addition and Elimination in Sterically Unbiased Systems

Mira Kaselj,

†,§

_{Wen-Sheng Chung,}

‡

_{and William J. le Noble*}

,†

Department of Chemistry, State University of New York, Stony Brook, New York 11794, and Department of Applied Chemistry, National Chiao Tung University, Hsinchu, Taiwan 30050, Republic of China

Received July 25, 1998 (Revised Manuscript Received January 20, 1999)

I. Introduction

For the purpose of this review, addition processes are considered to include all of the reactions in which there is an increase in the number of ligands bound to a central atom, and elimination covers all of the chemistry in which there is a decrease. With so wide a range of such processes, there can be little doubt that they are among the most important in all of chemistry, not only because of the multitude of reactions that fit the definition but also, and perhaps especially, because of the stereogenesis that charac-terizes additionsthat is where stereochemistry

be-gins. The question arises: when a new, additional ligand is going to be bound to the central atom, what will the preferred new configuration be? This ques-tion has created the need to know the factors that govern stereoselection.

Beyond any doubt, the best-documented and most powerful factor is the steric one, and for a long time, it was considered to be the only one. Even now, when the root mechanism of a successful new stereoselec-tive reagent or catalyst is discussed, it is virtually always taken for granted that the selectivity is the result of a difference in steric hindrance between the two approaches the reagent may choose. Yet it has been known since the 1950s that products sometimes form as the consequence of reactions in which it is not intuitively obvious how steric crowding could have led to the result. A variety of suggestions have been offered for these observations, and they may be captured under the general description of stereoelec-tronic effects.

The obvious way to study these effects is to employ probe molecules in which the two faces eligible for addition are sterically equivalent. The power of that approach has been dramatically demonstrated by Hammett, who used p-substituents in phenyl rings to influence rate and equilibrium constants in ways which are impervious to steric effects. The 2-nor-bornyl cation has provided us with an equally dra-matic example of the fact that failure to exclude possible steric effects where electronic influences are sought cannot lead to unassailable conclusions. The nonclassical ion controversy, in fact, was a study of face selection in an elimination reaction as well as of the stereochemistry of the final capture and addition step. In research of addition reactions involving neutral substrates, the most frequently used probes have been cyclohexanone and its deriva-tives, and from the beginning, there, too, arguments ensued about the degree and even direction of steric differences between the equatorial and axial ap-proaches (to say nothing about the conformational nonrigidity). Thus, important though this simple ring system may be in organic chemistry, it should be recognized that studies employing it will never convincingly reveal the reasons for the observed selectivities.

Unfortunately, it is not possible to design a probe containing a planar atom with two faces which are absolutely indistinguishable sterically, yet different electronically. As we shall see, this ideal can only be

†_{State University of New York} ‡_{National Chiao Tung University.}

§_{Present address: Geo-Centers, Inc., Building 3028, Picatinny} Arsenal, Mt. Arlington, NJ 07806.

1387 Chem. Rev. 1999, 99, 1387−1413

(2)

approached, and approached reasonably closely, but it cannot be reached. After all, electronic differences must be induced by means of one or more substitu-ents, and while one can place these substituents at positions remote from the site of reaction, one cannot guarantee that they are remote enough to avoid subtle steric differences, through distortion, polariza-tion, or other transmission. Indeed, a frequently raised point is whether the trigonal site to be added to is even planar, or whether the substitution has possibly induced pyramidality.

In this article, we review results that have been obtained by means of studies employing several probes that come quite close to the ideal of zero steric

bias. We did of course not attempt to incorporate a comprehensive listing of data which have been pro-duced by colleagues who are also authors of papers in this issue, but we did refer to those of their experiments that were needed to produce a seamless account. For investigators planning their own re-search in this area, it is important to consult earlier reviews and/or accounts by Wigfield,1 _Ashby,2 _and

Gung.3_{Reviews are also available in more specialized}

areas of narrower focus; they include papers by Zimmerman4_{and Pollack}5_{on keto-enol}

stereochem-istry, by Fallis6 _{on pericyclic stereochemistry, by}

Franck7 _{on the stereochemistry of simple}

cyclohex-anes, by Paddon-Row8 _{and by Bowden and Grubbs}9

on the through-bond and through-space transmission of substituent effects, by Juaristi10_{and Perrin}11 _on

the anomeric effect, and by Saunders12_{on carbocation}

chemistry. Each of these papers contains references to earlier summaries and seminal contributions to those topics. Excellent brief summaries may be found in the introductory paragraphs of recent research papers by Houk,13,14_Fraser,15_Fallis,16_Heathcock,17

Paddon-Row,18_Jorgensen,19_{and Tomoda.}20_{One of us}

has contributed reviews on the competing viewpoints of what lies behind nonsteric diastereoselection in addition21_{and in solvolysis.}22

While addition to trigonal centers lies at the heart of stereogenesis, elimination to produce such centers revolves about relative rates. These two processes are, of course, closely related, as each is essentially the microscopic reverse of the other; however, the addition reaction has received far more attention (with carbocation chemistry the main exception), presumably because of its importance in synthesis. Much of our own work in this field is based on the unique properties of the adamantane skeleton, such as its rigidity, a point which is elaborated further Mira Kaselj was born in Virovitica, Croatia, and attended the Faculty of

Science, University of Zagreb, where she received a B.A. degree in Chemistry in 1987. In 1990 she earned her M.Sc. degree in Chemistry, and in 1993 her Ph.D. degree in Organic Chemistry, both under the supervision of Dr. Kata Mlinaric-Majerski at Rudjer Boskovic Institute, Zagreb, Croatia. In 1994, she joined the research group of Professor William J. le Noble of SUNY at Stony Brook, New York. At present, she is a Senior Scientist at Geo-Centers, Inc., where she is involved in heterocyclic organic chemistry. Her extramural interests include traveling and hiking.

Wen-Sheng Chung, a native of Hualien, Taiwan, received his B.S. degree from National Tsing-Hua University, Hsinchu, in 1982 with distinguished honors. After two years of army service and one year as a teaching assistant at Tsing-Hua, he went to Columbia University to study organic photochemistry and high pressure chemistry under the guidance of Professors N. J. Turro and W. J. le Noble. He received his Ph.D. degree and a Pegram award in 1991 from Columbia University. He then worked with Professor J. A. Berson at Yale University for one year on a study of non-Kekule biradicals. He returned to Taiwan to begin his present position in 1991, first as an Associate Professor, and since 1977, as Professor. His current research interests in physical organic chemistry include host− guest chemistry and molecular sensors. He enjoys playing volleyball, tennis, and table tennis with his many friends.

Bill le Noble is a native of Rotterdam, The Netherlands. He studied chemical engineering as an undergraduate in Dordrecht and then moved to the United States in 1949. After a stint in the U.S. Army, which included service in Korea, he attended the University of Chicago and received a Ph.D. degree in organic chemistry with Professor G. W. Wheland. Postdoctoral work at Purdue University with Professor N. Kornblum was followed in 1959 by an appointment at the then new State University of New York at Stony Brook, where he has stayed ever since. He has held Visiting Professorships at the Free University of Amsterdam and the University of Groningen, received Humboldt Senior Scientist and Mom-busho Special Professorship Awards, and he has served as senior editor of The Journal of Organic Chemistry and Recueil as well as in the Chair’s position of his department. He is an avid swimmer and fisherman in his spare time.

(3)

below. Perhaps the first chemists to take advantage of these properties have been Schleyer23 _and

Mar-tin,24 _{who showed in 1969 that the allylic ester 1}

solvolyzes extremely slowly, thus demonstrating that when the leaving group is forced to depart in the plane perpendicular to the normally assisting π orbital, the vinyl group is an electron withdrawing one. Their work provided a dramatic example of the power of this probe to reveal insights that are hidden from view where nonrigid molecules are used.

In the sections that follow, we discuss the most important addition processes and explain the ob-served configurational preferences by means of tran-sition-state hyperconjugation, which we believe has proven to be the most successful theory to date for that purpose. In a final section, we mention the several alternative ideas to have emerged in the literature in recent years and their pros and cons. But first, we digress for a moment for two items of nomenclature. Throughout this paper and others already published, we have found it convenient to employ the numbering system shown in structure 2, even though it is sometimes incorrect; thus, we refer to 1,4,4-trichloroadamantane as the 2,2,5-species, and so on. We also prefer to use the E- and Z-stereochem-ical designations; E if the higher priority substituent among P and Q in 3 is anti to the higher priority substituent among R and S, and Z if it is syn. This is, of course, exactly the same convention as that proposed by Blackwood et al.25_{and now in common}

use for olefins; we see no reason why it should not also be used for achiral cyclic molecules such as 1,3-disubstituted cyclobutanes, etc. We refer to the two faces of C-2 in 2 as the en and zu faces, where en describes the distal face with respect to the substitu-ent at C-5, and zu repressubstitu-ents the proximal one.26_In

a few instances, we shall mention research done with chiral molecules; in all of those cases, the substances used were racemic.

II. Cation Formation and Capture

A. 5-Substituted 2-Adamantyl Cation

The stereochemistry of solvolysis went though a period of intense debate about the cause underlying often huge epimeric rate ratios.27_{The primary reason}

this dispute lasted for years without resolution was, as noted above, that the probes widely used to investigate the question, the exo-2-norbornyl esters 4, generate cations with two sterically inequivalent faces, and hence it is not possible to offer rigorous proof that the cause of the 103_{:1 exo/endo rate ratio}

is electronic in nature. In a few cases, unsaturated probes with sterically near-identical faces did find use, and theπ-participation postulated to cause their

very large solvolysis rate ratios was not controversial; perhaps the best known example is the 7-norbornenyl tosylate 5.28 _{The phenomenon of} _{σ-participation is}

now well-established as well, but it is interesting to speculate that it might have been studied, much earlier than it was, by means of 2-adamantyl esters or halides carrying a remote substituent to influence the ability of vicinalσ bonds to lend assistance in the heterolysis process. However, the parent 2-adamantyl ester was considered to provide the textbook case of unassisted solvolysis,29_{and thus, there was little or}

no incentive to study more elaborate specimens.

If one ignores this background, the 5-substituted adamantanes with a trigonal center at C-2 have much to recommend them as probes. The substituent, which is equatorial with respect to all three of the cyclohexyl rings converging at C-5, is remote enough from the trigonal center not to influence it sterically in any but the most extreme cases. There is virtually no conformational freedom in this rigid yet es-sentially strainless system, and Bredt’s rule forbids the possibility of elimination as a competing process. In early applications, the assignment of configuration was an obstacle that could be overcome only by means of X-ray studies;30_{however, such asignments}

are now routinely based on NMR spectra.31

The first chemist to study the solvolysis of a 2-adamantyl ester carrying a 5-substituent was Whiting,32_{who used a methyl group at that position}

to show that both secondary and tertiary esters 6 and 7 undergo acetolysis with predominant retention of configuration. The result was evidently surprising enough that the authors considered the possibility of an indirect steric effect, operating through the intervening axial hydrogen atoms, as the cause of the retention. Subsequent work in our own laboratory33

showed that predominant (92:8) retention also char-acterized the hydrolysis of 5-deuterio-2-adamantyl tosylate E-8 in 40% aqueous acetone (Scheme 1; note that Z-8 was not studied experimentally since its retention may be taken for granted. To suppose otherwise is to postulate aδ deuterium isotope effect). That observation rules out the steric possibility, Scheme 1

(4)

leaving assistance by the antiperiplanar vicinal bonds as the only reasonable explanation. One other pos-sibility is sometimes mentioned as an alternative, namely, that the solvent molecule trapped between the cation and the leaving group in a solvent-separated ion pair might be a stronger nucleophile than bulk solvent by virtue of its proximity to the anion and thus produce retention. This explanation cannot apply, however, to the tertiary 2-adamantyl cations, as noted in the following paragraphs.

The first insight that a 5-substituent can influence face selection in the capture of a free 2-adamantyl cation derived from a study34 _{of the}

5-phenyl-substituted propargyl chlorides shown in Scheme 2. Dissolution of both the E- and Z-isomers 9 in metha-nol caused partial methametha-nolysis and partial rear-rangement to the chloroallenes 10. The latter reac-tion is stereospecific, and hence there is evidently no crossover between or dissociation of the tight ion pair intermediates. In contrast, the methyl ethers 11 are produced in the same ratio from both epimers 9; evidently, the free cation is a common intermediate. The composition of this common mixture is of special interest: the Z-ether predominates by a ratio of 3:1 over the E-isomer. The remote phenyl group clearly has a directive effect, which we attribute to σ-delo-calization. Of the four C-C bonds vicinal to the cation center, the two proximal to the electron-withdrawing phenyl group are deactivated by it; the other two therefore are the bonds that stabilize the cation and control the stereochemistry. This clear evidence for delocalization is especially noteworthy because this cation is not just a tertiary species, but it is presum-ably stabilized by the triple bond. The conclusion that this stabilization is not enough to swamp the hyper-conjugative interaction with the neighboring bonds is inescapable.

These conclusions were supported by means of extensions to other systems. First, it was found that other electron-withdrawing substituents such as chloro-, fluoro-, trifluoromethyl in the 5-position lead to a similar or even larger preference of the

nucleo-phile for the zu face; thus, exposure of either of the two tertiary alcohols 12 to hydrogen chloride in dichloromethane leads to identical mixtures of 2-chlo-roadamantanes in which the Z-isomers are present in severalfold excess.33_{Neither the alcohols nor the}

products interconvert under these conditions (Scheme 3).

When the rates of generation of the cations rather than the ratios of the products are measured, far larger effects are encountered.35 _{Thus, a 5-fluoro}

substituent gives rise to an E/Z rate ratio of solvolysis of the secondary 2-adamantyl tosylates in aqueous acetone of less than 0.01. It was also found that if the 5-substituent is a donor such as trimethylstannyl, E/Z rate ratios far larger than unity result.36 _The

most powerful donor effect was found by Grob,37_who

replaced C-5 by a nitrogen atom. The assistance in this case derives from the unshared pair. The reac-tion of E-13 is some 105 _{times faster than that of}

2-bromoadamantane and produces the ring-opened species 14 as the initial product.

The remarkable persistence of the ability of anti-periplanar neighboring bonds to stabilize cations even if the positive center is bound to an ethynyl group led us to try to tune the product ratio in 2-phenyl-2-adamantyl cations by means of p-substi-tution. In one series of reactions, the cations were generated from mixtures of E- and Z-alcohols 15 by means of hydrogen chloride gas dissolved in meth-ylene chloride to give chlorides 16; in another, they were obtained by solvolysis of mixtures of E- and Z-tosylates 17 and captured with sodium borohydride to give reduction products 18. In all instances, the products resulting from attack at the zu face pre-dominated over their isomers by margins ranging narrowly and apparently randomly between 78:22 to 73:27. Evidently, the assistance is not even swamped by a p-anisyl group. It should be noted that these cations are essentially cumyl cations, the same spe-cies that provide the basis for theσ+ constants! One is led to see that the stability of cumyl cations is derived in part from delocalization by the methyl C-H bonds. It is evident that the σF-applications

(5)

which have been used so often27_{to call the existence}

of σ-delocalization into question provide at best a shaky basis on which to build such conclusions. When the solid salt 19 is treated with sodium borohydride, even then, the E-ether is formed in excess over the Z-isomer by a margin of 83:17.38 _{The conclusion is}

that σ-delocalization from vicinal bonds does not merely occur, but that it cannot be swamped by a substituent bearing unshared pairs.

We take note here of the presumed difference between σ-delocalization and hyperconjugation. As we have argued elsewhere,22_{when the participating}

or delocalizing bond is a vicinal one, there appears to be no difference; in fact, there is to our knowledge no clear definition of these two terms which ad-dresses the difference in the literature (however, see Kirmse et al.:39 _{“...there is no reason to mingle}

hyperconjugation withσ delocalization...”). The term σ-delocalization is perfectly descriptive and justified when the assistance is rendered through space by a more remote single bond; thus, McMurry’s cation 2040

and Sorensen’s cation 2141 _{are good examples of}

σ-participation by a remote C-H bond. Yet, the two-electron three-center bonds, even in these species, are essentially hyperconjugative in nature.

We also note a paper by Liu42_{in which he reports}

a decrease in the E/Z rate ratio of solvolysis of compounds 16 as Y changes from trifluoromethyl to methyl and claims that this shows our conclusion regarding the absence of swamping to be invalid. However, to “swamp” means to submerge completely, and Liu’s data do not indicate that his phenyl groups have stopped the assistance of the vicinal bonds; in fact, the product mixtures reported in Liu’s paper show that the capturing nucleophile preferentially attacked the zu face in every case, by margins between three and fifty to one.

The evidence that the 2-adamantyl cations are subject to stabilization by the vicinal bonds has been confirmed by Sorensen43_{in a dramatic way. He was}

able to study the 2,5-dimethyladamant-2-yl cation in superacid solution at low temperature by means of

13_{C NMR, and found it to be an equilibrating mixture}

of two structures, presumably 22 and 23. The equi-librium constant for the reaction as written is about 10. At least equally pertinent in this respect are the several studies by Laube, who has been able to carry out X-ray diffraction studies on solid carbocation salts at low temperatures.44_{The structure of the}

2-phenyl-2-adamantyl cation is of special interest.45_{The C} 1

-C2and C2-C3bonds are shortened by nearly 0.1 Å;

the C1C2C3bridge leans toward one side by 7.86°, and

the phenyl group by an additional 5.6°. The vicinal bonds on that side are elongated by about 0.05 Å; all of this is consistent with strong hyperconjugation. Bond shortening in the C-CH3bonds has also been

observed in the parent cumyl cation itself.46_While

one could also attribute the stereochemistry to the greater openness of the leaning bridge or the pyra-midality at C-2, because both of these distortions are themselves due to hyperconjugation, these are not different explanations.

The question arises how the substituent at C-5 makes its influence felt in the intervening anti-periplanar vicinal bonds. Carrying the argument of hyperconjugation one step further, we suppose that the effect is ascribable to delocalization of electrons in these bonds into theσ* orbital of the C-X bond, thus reducing their ability to assist at C-2. Strong support for efficient electron delocalization in this type of extended hyperconjugation has been reported by Grob,47 _Adcock,48-51 _Borden,52 _Michl,53

(6)

We summarize as follows. The physical properties of the 2-methyl- and 2-phenyl-2-adamantyl cations clearly show that they are subject to hyperconjuga-tion. We therefore feel that the stereochemistry of capture of 5-substituted 2-adamantyl cations as well as the epimeric rate ratios observed in their generation are not merely conveniently explained by hyperconjugation, but are plainly a consequence of it.

B. 4-Substituted 2-Adamantyl Cation

The preceding section gives rise to the intriguing question of how face selection in 2-adamantyl cations will be affected by a 4-substituent. If this substituent is axial (with respect to the cyclohexyl cation), it will obviously affect the stereochemistry in a steric way,57 _{and hence we will limit this discussion to}

cases in which it is equatorial. While an electron-withdrawing group in that position will inductively reduce the ability of the C3-C4 bond to assist at

C-2, it is not self-evident that this effect will be stronger than that exerted by the same substituent at C-5, the nature of the interaction being differ-ent.

Experimental data relevant to this question have been provided by Grob,58,59 _{who found the ability of}

substituents to affect solvolysis rates of 2-adamantyl esters weaker if they are in the equatorial 4-position than if they are located in the 5-position. He inter-preted this fact as evidence that conflicts with the proposal that hyperconjugation is the reason why 5-substituents have such drastic effects as noted in the preceding section.59_{However, this conclusion is}

certainly debatable; 5-substituents affect both inter-vening bonds vicinal to C-2, whereas equatorial 4-substituents affect only one. Note also that all four of the bonds vicinal to C-2 are related differently to C-4. We presume that the extended antiperiplanarity of 5-substituents is the factor that renders them so effective in this reaction.

III. Carbene Capture

A. 5-Substituted 2-Adamantylidene Carbene

Carbenes in which neither of the two groups bound to the carbenic center carries unshared electrons behave as strongly electron-deficient species, and hence they may be expected to mimic carbocations in their stereochemical properties. We see the ada-mantylidene carbene as a species in which C-2 has a filled sp2 _{orbital and a vacant p orbital. A report}

bearing out this expectation has been published by Majerski.60 _{He studied the formation of}

dehydro-adamantanes 24 and 25 in the Bamford-Stevens reaction shown below, and observed the product ratio to be 74:26 in favor of 24 when X is methyl, and 92:8 when it is chloro. These results signify that the secondary hydrogen atoms most remote from the substituent are the favored targets for carbene insertion, and that this preference becomes more pronounced the more electronegative the sub-stituent.

B. 5-Substituted 2-Vinylideneadamantane

We referred in section IIA to the neutral solvolysis of the propargyl chlorides E- and Z-9. When these same compounds are exposed to a basic medium, the resulting anions61 _{jettison chloride ion to give}

car-bene-anion pairs,62_{which in part undergo retentive}

isomerization to the allenes 10 and in part dissociate to the common free carbene 26. This carbene has a strongly dipolar character,63,64_{and the dipole moment}

should increase because of induction as a nucleophile approaches. Carbenes such as 26 should therefore mimic the 2-adamantyl cation in their behavior, and indeed, 26 captures methanol with a 3:1 preference for Z-11 as the product.65

IV. Nucleophilic Addition to Ketones

The stereochemistry of addition to carbonyl groups in open-chain compounds is obviously sensitive to steric and thermodynamic factors, and simple, well-known rules have been developed to predict both the conformation of the necessary nearby chiral center, the least hindered face for the approach of the reagent, and the stability of the products. It was only when cyclohexanone stereochemistry began to be studied that awareness gradually set in that these three influences alone were insufficient to explain the observed results.

The alternative and additional factors that can play a role have been much debated, and there is still much disagreement about them. Because most of the experimental background fueling these discussions is in the area of nucleophilic addition to ketones and because much of the discussion in the literature concerns the relevance of through-bond interactions between substituents and the carbonyl center, a brief digression is inserted here to describe the arguments favoring such interactions, and thus to clarify the purposes of the experiments. Readers wishing to see a fuller discussion of the main schools of thought may consult ref 21, papers quoted in that account, section XII below, and other papers in this issue; suffice it here to say that the similarity in stereochemical behavior between carbenes and cations naturally led to the question whether ketones might not also fit this mold.

(7)

Theπ and π* orbitals, which interact weakly with the vicinalσ and σ* orbitals, evolve into the new σ and σ* orbitals as the reaction coordinate is tra-versed. The electronic interaction of the incipient bond with vicinal bonds can take three forms: be-tweenσiandσv*, betweenσi* andσv, and betweenσi

andσv, where i stands for incipient and v for vicinal.

The first two of these, favored by Anh66_{and Cieplak,}67

respectively, are interactions between an empty and an antiperiplanar filled orbital, and, as such, they do in principle lower the transition-state energy; the third, favored by Felkin,68_{is a repulsive one between}

filled, synperiplanar orbitals and thus, it raises the barrier (Chart 1). The best Anh approach is clearly that which will donate electron density from the incipient bond into the most electron-deficient vicinal bonds; the best Cieplak approach is that which will donate electron density from the most electron-rich vicinal bonds. These two views therefore make clearly opposite predictions. The Felkin approach is least repulsive when the synperiplanar bond is electron deficient; hence, this interaction predicts the same stereochemistry as does the Cieplak model. All three have been successfully applied to cyclohexanone addition stereochemistry; however, the Anh and Felkin models also postulate various conformational distortions to account for the results. The Cieplak model postulates that C-H bonds are better donors than C-C bonds (in other words, that the Baker-Nathan order is correct) in order to explain the oft-observed axial approach of nucleophiles to cyclohex-anones.69

The 5-substituted adamantanone system is virtu-ally ideal to test these models. The skeleton has two carbonyl faces which would be identical but for the remote 5-substituent; its rigidity avoids the distor-tions apparently important in the Anh and Felkin models, and the propriety of the Baker-Nathan order is immaterial because all four vicinal bonds are C-C bonds which are differentiated only by the remote substituent. This same remoteness is of course also a disadvantage: the product (or rate) ratios are generally quite modest. But, one could say this also about secondary isotope effects. The point of the experiments is not to find large effects, but insightful ones, and, if that is achieved, they are well-chosen.

A. 5-Substituted Adamantanone

The first indication that the usefulness of the adamantanone probe was not limited to electron-deficient intermediates such as carbocations and carbenes came from the observation65_{that the}

ethy-nylation of 5-phenyladamantanone 27 did not give a 50/50 mixture as we had supposed; the E-isomer of 28 dominated by a margin of 2:1. In later work, we

learned that this deviation was general; the reduction of this ketone by means of the usual complex hydrides also produced E-alcohols in excess. Furthermore, other electron-withdrawing 5-substituents such as methoxycarbonyl,70 _{bromo, chloro, fluoro, hydroxy,}

and trifluoromethyl33_{caused the same behavior; the}

margins, ranging from 70:30 to 55:45, seemed to reflect the electronegativity of the substituents. In the case of 5-phenyl, in fact, it was possible to tune the ratio by means of variations in the para substitu-ent, although not as much as we at first reported.33,71,72

An electron-withdrawing substituent such as nitro tends to increase the tendency of the phenyl group to direct the nucleophile toward the zu face of the carbonyl carbon, whereas a p-amino group tends to diminish it. These effects can be most clearly seen in the 5,7-diphenyladamantan-2-ones 29-X,Y.73

Upon reduction with sodium borohydride in 2-pro-panol, ketones 29-H,NO2, 29-H,NH2, and 29-NH2

,-NO2 produce the E- and Z-alcohols in the ratios of

1.30, 0.78, and 1.64, respectively. It is evident that the nitro group which deactivates the proximal vicinal bonds directs the nucleophile to the zu face and that the amino function activates the same bonds and directs the reagent to the en face; in 29-NH2

,-NO2, both effects team up to give a ratio which is

the quotient of the other two. The effects of a donor substituent are also observable in the 5-trimethylsi-lyl- and -stannyl-2-adamantanones; thus, treatment of 30 with methyllithium gave the tertiary E- and Z-adamantanols in a ratio of 36:64.74

One might expect that the magnitude of the selectivity should be somewhat dependent on the nature of the nucleophile, but the variation is sur-prisingly small. The addition of 5-fluoroadamantan-2-one to six p-substituted phenylmagnesium bro-mides, with the p-substituent varying from trifluoro-methyl to ditrifluoro-methylamino, gave E/Z ratios between 68:32 and 76:24, with no obvious trend or correla-tion.75 _{In our search for such effects, we recalled a}

drastic case of nucleophile differentiation which had Chart 1

(8)

been reported by Gassman;76_{he had found a complete}

reversal of face selection in the two reactions in Scheme 4. When these two reagents were allowed to react with 5-fluoroadamantan-2-one, however, the E/Z ratios were found to be nearly identical (2.3 vs 3.0). Evidently, the double bond in norbornenone begins to play a role77_{when the ionic reagent}

pen-tafluoroethyllithium approaches the keto function. In any event, we now strongly suspected that transition-state hyperconjugation was driving the stereochemistry, and that the carbonyl group’s se-lectivity was a consequence of its carbocation char-acter. One possible test of this conjecture is to deactivate all four of its vicinal bonds completely. In pursuit of that notion, we studied the reaction of lithium borohydride dissolved in ether with 7-mono-hydrylperfluoroadamantan-2-one (Scheme 5), in which fluorine substitution abrogates the ability of the vicinal bonds to function as donors, and converts them into acceptors instead. As the scheme indicates, this ketone does indeed follow the prediction made on the basis of reverse electron flow.78

The apparently predictable selectivity of the ada-mantanones led us to search for ways to control it; i.e., to reverse it or to augment it at will. One obvious way to reverse it (by means other than the perflu-orination just mentioned) is by introducing a steric factor, for example, by the use of a bulky 5-substitu-ent, or reag5-substitu-ent, or both. Indeed, the data for 3179

show that en face can be favored in such cases:33

while a 50/50 product distribution was observed when Y in the reagent is H, alcohol Z-32 predominates by 58:42 when it is tert-butoxy. Even more effective approaches have been devised; they include the blocking of the normally preferred face by means of a catalyst (see section VA), and encapsulating the adamantanone in a cyclodextrin (an example is

discussed in section VIII).

There are two ways to enhance the selectivity. The first depends on rendering the substrate more “car-bocation-like”. This can be done by allowing the carbonyl oxygen atom to complex with a Lewis acid; we had already noted that with lithium aluminum hydride this selectivity was somewhat more pro-nounced than with sodium borohydride, and assumed that this was due to the complexation of the lithium ion as a Lewis acid. Indeed, the reduction of 5-phe-nyladamantan-2-one with sodium borohydride in the presence of a variety of Lewis acids such as alumi-num chloride, boron trifluoride, stannic chloride, and so on80 _{raised the E/Z ratio of alcohols by 15-30%.}

Antimony pentachloride was among the most effec-tive of these acids; this is of special interest because the complex has been isolated and studied by means of X-ray diffraction.81_{Laube found that the CdO bond}

and the C3-C4and C1-C9bonds in the complex are

lengthened by 0.045 and 0.017 Å, respectively, while the C1-C2and C2-C3bonds are shortened by 0.031

Å, compared to the free ketone. Thus, complexation has the effect of increasing the need for hyperconju-gative delocalization. The E/Z ratio (83:17) in the reduction of 5-fluoro-2-methoxy-2-adamantyl cation (19), as already noted, is about triple that of 5-fluo-roadamantan-2-one (62:38); this may be seen as an extreme case of complexation of the latter ketone by a Lewis acid, the methyl cation. Interestingly, no increase was seen at all when the methylimine 33 was compared with the iminium salt 34: under a variety of conditions, they showed identical E/Z ratios of about 2:1 in the reduction with sodium borohy-dride. We assume that the imine, the ratio for which seems abnormally large, is in fact in cationic form (protonated or complexed with lithium cation) during reduction.80

Similar effects can be brought about by the intro-duction of charge at the site of the 5-substituent. Thus, the sodium borohydride reduction of 5-dim-ethylaminoadamantan-2-one in D2O, which normally

has an E/Z ratio of 65:35, may be compared with the trimethylammonium salt which weighs in at 86:14, a more than threefold increase.80 _{The E/Z ratio for}

5-hydroxyadamantane (58:42) is diminished after treatment with sodium hydride to 49:51 and that for 5-methoxycarbonyladamantan-2-one (57:43) may be compared with 45:55 for the carboxylate salt.80_Even

Scheme 4

(9)

more drastic effects are observed if C-5 itself is replaced by positive nitrogen, as may be seen in Scheme 6. The E/Z product ratios in these two reactions are 60:40 and 96:4, respectively. This presents an increase in the ratio by a factor of 16, for two substrates which are isoelectronic! Similar results were obtained with iodide of cation 39 and betaine 40.82 _{The same high ratio was found in}

methanol, water, and saturated aqueous sodium chloride (for an alternative interpretation, see section XII).

The behavior of the neutral aminoketone itself is also of interest. We found the secondary alcohol E-42 to be in excess upon reduction in methanol (62:38), but a small excess of the tertiary alcohol Z-43 (55: 45) is obtained in the reaction with methyllithium in THF. We can readily account for this solvent effect; in the latter reaction, we assume that the syn vicinal bonds are activated by the nitrogen unshared pair and that this donation is prevented or thwarted in a hydrogen-bonding solvent (Chart 2). Senda83 _has

measured the ratio of sodium borohydride reduction in several alcohols and in THF; in fact, an excellent linear correlation of the E/Z ratio was found when it was plotted against the Swain solvent-acidity pa-rameter. The crossover in favor of Z-42 found by Senda in THF was included in the plot.

The effects of initial-state hyperconjugation can be seen very clearly in compounds 44 (R ) H or Me). Zefirov84 _{determined the X-ray diffraction}

pattern of these cations. The N5-C6bond was found

to be more than 0.1 Å longer than the N7-C6 bond

which clearly supports the contention that the alternative “no-bond structure” makes an important contribution. This type of hyperconjugation reduces the difference in donating power of the two sets of vicinal bonds, and the E/Z ratio of alcohols 46 in the sodium borohydride reduction of 45 in D2O is

ob-served85_{to be 7.3, compared to 24 for the monoaza}

analogue.

In discussions of face selection, questions about the approach angle are often raised. Indeed, if the rigidity of the adamantane skeleton and the location of the axial hydrogen atoms force an abnormal approach trajectory upon the nucleophile, one may argue that the adamantanone probe is not a good one. Con-versely, the near-symmetry of the molecule should allow the angle to be the same at both faces and thus cancel it out as a matter of concern. Be that as it may, we sought to gain an insight into this aspect by adjusting the length of the bridge represented by C-6 in the adamantanone probe. As the structures drawn below suggest, a more favorable approach is possible in the noradamantan-9-ones and an even more crowded one seems unavoidable in the homoada-mantan-9-ones. Studies of the reduction and alkyla-tion product ratios did not show much difference between these ketones, and the small variations observed were not systematic. The quaternary bridge-head aza analogues, with their substantially larger ratios, were more informative; they suggest that the adamantanone structure provides the largest ratios. Thus, the E/Z ratios for the alcohols obtained by sodium borohydride reduction of 47 and 39 are 87:13 and 96:4, respectively, and those of 48 and 49 are 88:12 and 83:17, respectively. We assume that this trend is due to the more nearly perfect antiperiplanarity of the substituent with the bonds vicinal to it.86 _{One additional point of interest is}

that both methylation and reduction of the azano-radamantanone show the same solvent dependence as does the azaadamantanone itself, and surely for the same reason: donation by the unshared nitrogen pair in THF, and electron withdrawal by the nitrogen in a hydrogen-bonding medium such as methanol.

Scheme 6

(10)

B. 4-Substituted Adamantanone

As noted in section IIB, a 4-substituent, to be useful at all, should be in the equatorial position because otherwise it will affect the stereochemistry sterically; but even then, it has the disadvantage of being poorly aligned and of affecting only one of the vicinal bonds. Nonetheless, its greater proximity to the trigonal center renders such probes interesting.

The steric effect of an axial substituent makes itself felt with even the smallest, namely fluoro: sodium borohydride attacks probe 50 exclusively at the en face to give the pure diaxial alcohol 51, whereas with the equatorial fluorine substitution in 52, the diequa-torial alcohol 53 is formed with a 67:33 preference over its isomer 54.87_{Similar data were found for the}

two bromoketones. In one instance, we were able to prepare a diequatorial species, namely, 4,9-dibro-moadamantan-2-one 55. Sodium borohydride reduc-tion in that case gave the two alcohols 56 and 57 in a ratio of 86:14.87_{This ratio is substantially but not}

spectacularly larger than that observed with 5-bro-moadamantan-2-one (59:41), which reinforces our impression that the effect of the bromine atom in the latter case is effectively transmitted via extended hyperconjugation.

Another interesting question is what might happen if C-4 and/or C-9 are replaced by nitrogen, but these compounds are not known. However, the oxa deriva-tive 58 has been prepared. Here we must deal with an uncertain difference in size between an axial hydrogen atom and an axial unshared pair and the ability of the oxygen to withdraw electrons vs its ability to participate with its unshared pair. In methylation, phenylation, and reduction, all of the reagents attack exclusively at the zu face88 _{to give}

alcohols such as 59, but whether this is due to reduced hindrance, oxygen electronegativity (or both), or even chelation is unsettled.

C. Other Carbonyl Probes

The face selection studies possible with 2,3-bis-endo-disubstituted 7-norbornanones 60 have been conducted principally by Mehta. Here also, the two faces are sterically equivalent; 60 differs from the 5-substituted adamantanones principally in that, the substituents being closer to the trigonal center, the product ratios are larger. A more minor consideration is that 60 is not as rigid, and the torsional motion possible in the C2-C3 and C5-C6 bonds was once

considered89 _{as a contributor to the high exo/endo}

solvolysis rate ratios of 2-norbornyl esters.

Mehta’s first report90_{revealed that the E-alcohol}

formed preferentially when X and Y are both meth-oxycarbonyl groups, both in reduction with complex hydrides (margins of 3:1 to 6:1) and in the reaction with methyllithium (>10:1). Rather surprisingly, however, when X and Y are both methoxymethyl, ethyl or vinyl, the Z-alcohols predominate, by mar-gins of 1.5 to 4:1 in the reduction and 5 in the methylation. Thus, if hyperconjugation in the Cieplak sense is expected, these substituents should be functioning as acceptors compared to the endo hy-drogen atoms at C5 and C6. Similar findings were

reported91_{in instances in which the C}

2-C3bond was

annelated with a five-membered ring on the endo side. In a later paper,92_{Mehta attempted to correlate}

the observed ratios with those calculated on the basis of the assumption that the selectivity has an elec-trostatic origin, with partial success. This explanation was also advanced for the product ratios in the reduction of a number 2-phenyl-substituted norbor-nanones in which the phenyl ring itself was also substituted.93_{However, as we pointed out above, a}

2-substituent alone affects all four of the vicinal bonds differently, making it difficult to predict what

(11)

will happen as a result of hyperconjugation. The tetracyclic bis-ether 61 undergoes reduction at the zu face to give the E-alcohol by a margin of more than 85:15; the result is reproduced by both MNDO and ab initio calculations, but these techniques suggest different reasons for the result (orbital and electro-static effects, respectively).94

An intriguing contribution by Gassman95_concerns

the tricyclic ketone 62. The methylation of this compound occurs with an E/Z ratio which is strongly dependent on the reagent: lithium dimethylcuprate, which operates via an initial electron transfer, re-verses to 8:92 the ratio attained with methyllithium (95:5) or methylmagnesium iodide (90:10). It will be of great interest to learn whether a similar reversal occurs in the 4,9-didehydroadamantan-2-one (the ethylene ketal of which is known.87_{) No reversal of}

stereochemistry was observed with 5-phenyladaman-tanone; the ratio was exactly the same as with sodium borohydride.96_{In the case of the more}

com-plex norbornanones 63-65, both sodium borohydride and methyllithium preferentially attack the zu face,97

by margins ranging from 57:43 to 68:32. In this case, at least, the norbornanones appear to behave as expected on the basis of transition-state hypercon-jugation, which points up the need to use rigid probes in these studies.

5,6-Bis-endo-disubstituted norbornenones have also been studied by Mehta. While the two faces differ sterically in these cases, one can look for trends as a function of the substituents. Thus, while the parent ketone 66-H reacts with methyllithium to give the tertiary E-alcohol by a margin of 74:26, two meth-oxycarbonyl groups cause a reversal98_{to 90:10 (note}

that the E- and Z-descriptors must be reversed too! See Scheme 7). In the case of 66-COOMe, the authors were able to obtain X-ray data, which allow several important conclusions to be drawn. One that is particularly pertinent is that there is no significant pyramidalization in this molecule: the three angles at the carbonyl carbon add up to 360°. On the other hand, the carbonyl bridge leans somewhat in the direction of the C5-C6 bond. Thus, whatever the

reason for this distortion, the reagent seeks out the

more hindered face.99_{Some of the 5-monosubstituted}

substrates 67-X also furnish interesting data; they seem in reasonable agreement with transition-state hyperconjugation, but they also vividly demonstrate the advantage of probes in which each face is anti-periplanar to only a single kind of vicinal bond. Thus, with 67-CN, -COOMe, and -COO-, sodium borohy-dride attacks the carbonyl face remote from the double bond by margins of 44:56, 68:32, and >90:10, respectively.

Mehta also found interesting changes in the reduc-tion product ratio ifβ-cyclodextrin is present:100_with

both 7-norbornenones and 7-norbornanones, the amount of Z-alcohol rises. Thus, the E/Z ratio for the reduction of 66-COOMe declines from 55:45 in metha-nol to 8:92 ifβ-cyclodextrin is added. Neither the R-norγ-cyclodextrins have much effect. The interpreta-tion is that the methoxycarbonyl groups become complexed in theβ-cyclodextrin cavity, thus leaving the en face more exposed. The R- andγ-cyclodextrins do not provide a good fit, and thus affect the ratio less or not at all. This experiment is similar to one reported earlier (see section VIII below).

Berg and co-workers have reported101_interesting

data on the sodium borohydride reduction of dione 68. Both carbonyl groups react at nearly equal rates, and both do so stereospecifically: only syn attack occurs in both branches. Both ketols 69 and 70 thereafter undergo a second reduction, the former mostly on the syn side, but the latter on the anti side. This latter result is unexpected in terms of hyper-conjugative donation to an electron-deficient incipient bond; a good explanation cannot be given at this time. Okada et al.102 _{have described their results with}

several 7-benzonorbornenones 71-X,Y. The carbonyl faces are not sterically equivalent, but the variation in product distribution as a function of the benzo substitution is significant. These authors report that in the reduction with lithium aluminum hydride, the percent E-alcohol equals 100 with 71-F,F; 92 with 71-Cl,Cl; 62 with 71-H,H; and 45 with 71-COOMe,H. Similar data were obtained with five other common reducing agents. While transition-state hyperconju-gation in the Cieplak sense would seem to account for these data, one could also maintain that the Scheme 7

(12)

nucleophile forms a complex with the benzo face and then attacks the carbonyl group on that side; such a complex would be expected to be strongest when the benzo substituents are the most electron-withdraw-ing.

Burnell has reported103_{the reactions of ketone 72}

with sodium borohydride and methyllithium, both of which take place at the carbonyl face closest to the double bond. The margins were about 5:1. This finding is readily interpreted in terms of the Cieplak model, and the authors did so.

A closely related experiment was carried out by Halterman.104 _{The reduction and methylation of}

ketones 73 gave rise to products resulting from addition of the nucleophile opposite the more electron-rich phenyl rings; a linear relationship was found between the logarithm of the ratio versus the σ constant of substituents varying from amino to nitro.

V. Electrophilic Addition to Olefins

A. 5-Substituted 2-Alkylideneadamantane

One of the interesting questions that may be raised for theories that purport to correctly interpret the

stereochemistry of nucleophilic addition to carbonyl groups is whether they can be used to predict the stereochemistry of electrophilic addition. If the σi*

orbital is indeed the recipient of electron donation by the antiperiplanar vicinal bond(s), the stereo-chemistry in the two types of reaction should be the same. In this view, it is immaterial whether the electrons that will populate the σi orbital are

con-tributed by the reagent nucleophile or by the sub-strate; the important feature is theσ* component and its acceptance of electrons from the richest available vicinal bond(s). The prediction is therefore that the stereochemistry should be the same. To put it in different words: hyperconjugative assistance will lower the transition-state energy in the same way whether the probe molecule is the substrate or the reagent (the nucleophile or the electrophile).105

Two early reports addressing this question ap-peared simultaneously in 1987. In one of them,106

5-fluoro-2-methyleneadamantane 74 was treated with several electrophiles, and indeed, syn attack was observed in every case. Thus, epoxidation by means of m-chloroperbenzoic acid gives the Z- and E-oxiranes in a ratio of 66:34, respectively; similar results were obtained in the capture of dihalocar-benes and in the anti Markovnikov hydroboration. Much larger ratios were seen in those cases in which the reagent becomes bound to the methylene carbon, leaving C-2 in cationic form to scavenge anions; in the reaction of 74 with hydrogen chloride in dichlo-romethane, for example, the E-dihalide product is almost undetectable (0.05%). The other paper just mentioned107_{describes a study of electrophilic}

addi-tions to monocyclic methylenecyclohexanes; the con-clusions were the same, although these are of course subject to the same objections which have so often been raised against cyclohexanones as probes.

Burgess108_{has published an interesting extension}

of our study with 5-substituted 2-methyleneadaman-tanes, using the hydroboration reaction to show not only that syn approach is favored if the substituent is fluoro or phenyl, but also that anti attack predomi-nates if it is trimethylsilyl. The E/Z product ratios were found to be 64:36, 53:47, and 47:53, respectively. Furthermore, all of the results are reversed when a rhodium catalyst is added; the ratios then are 46:54, 42:58, and 75:25, respectively. No solvent effects worth mentioning were encountered. The catalyst apparently affects the outcome of the reaction through prior complexation at the normally favored face.

(13)

A similar explanation may apply to the biohy-droxylations described by Bailey109_{and by Davis.}110

Bailey found that the resulting 2-hydroxy-5-substi-tuted adamantanes all had the E-configuration in a group of eight substrates, each with an electron withdrawing substituent; on the other hand, Davis reported that with the bridgehead substituted azi-doadamantane, the resulting 5-azidoadamantan-2-ol was a mixture with a 10:1 predominance of the Z-isomer. Because these oxygenations undoubtedly take place in an environment involving other complex biomolecules, no mechanistic conclusions can be drawn from these observations, interesting though they are.

As was the case in nucleophilic addition, replacing C-5 with positive nitrogen magnified the ratio con-siderably. To mention one example,111_{the epoxidation}

of amine oxide 75 gives the Z- and E-epoxides in a ratio of about 5:1. Additional instances may be found in the same paper.

B. 4-Substituted 2-Alkylideneadamantane

The only example appears to be Duddeck’s report112

which describes the catalytic hydrogenation of the 4-fluoro-2-methyleneadamantanes 76 and 77; how-ever, the presence of an axial fluorine in one case and that of a heterogeneous catalyst in both reactions rules out the possibility of a mechanistic interpreta-tion.

C. Other Olefinic Probes

An early paper by Hoffmann113_{describes his group’s}

results obtained with compounds 78 and 79 in the addition of dichlorocarbene and in hydroboration with 9-BBN. As would be expected, while the former compound directs these reagents to the syn face, a reversal occurs in the latter. It is possible that in the reaction of 79, electron donation from the cyclopropyl group’s bent C-C bond plays a significant role.

Mehta114-116_{and Ohwada}117 _{have made}

contribu-tions to our knowledge and understanding of elec-trophilic addition to 7-methylenenorbornanes as well; these are discussed elsewhere in this volume. Gan-dolfi118 _{has made good use of the 3,4-disubstituted}

cyclobutenes 80. The osmium tetroxide mediated dihydroxylation of these compounds is of course determined in part by steric interactions; neverthe-less, a trend can be seen as a function of the substituents which hints at an electronic contribu-tion. The syn products dominate with R ) OSO2Me

(8:1); the ratio is much smaller with R ) Cl (3:2), and only anti product is formed when R,R ) (CH2)4.

Malpass has been able to study119_{the chlorination}

of 7-azabenzonorbornenes 81-X and -norbornadienes 82-X, no small feat if one considers that the N-chloroamine products are configurationally stable only up to -50°. This instability has the advantage, however, that the thermodynamic ratios are also available. In this case also, the two faces differ sterically, but the effect of benzo substitution tells the story. Thus, while the parent compounds react with N-chlorosuccinimide to give the E-chloro com-pounds, the introduction of electron-withdrawing substituents on the benzo ring leads to a clear shift in favor of the Z-isomers. With tetrafluoro substitu-tion as in 82-F, the Z-isomer is the principal product. This is what the Cieplak model predicts; however, as Malpass pointed out, an increasing interaction between the benzo ring and the succinimide anion being released could also be responsible. Further-more, there is a fair correlation between the ratios under kinetically and under thermodynamically con-trolled conditions, so that product stability may also have influenced the transition-state barriers. In 83-X, where steric differences should be absent, syn chlorination predominates by a 3:1 margin.120_{But one}

other reservation must be expressed for these stud-ies: Ohwada121_{has found that even amide 84 has a}

distinctly pyramidal nitrogen atom, and this is surely also the case with Malpass’ amines.

Haltermen has used 3,3-diphenylcyclopentenes 85-X very effectively to study both the osmium-catalyzed cis-dihydroxylation122_{and the epoxidation}

mediated by peracetic acid.123_{In both cases, excellent}

correlations were obtained in plots of the logarithm of the ratio versus the σ constants, with p-nitro

(14)

promoting the largest amount of syn attack and p-amino or p-methoxy the largest amount of anti attack. These observations provide strong evidence for the notion that transition-state hyperconjugation is driving the face selection. The same reactions take place with bicyclo[2.2.2]octenes 86, and again, the preferred approach is syn in both of these cases.124

In the analogous compound 87-X, Ohwada125,126_found

a strong preference for syn attack if X is nitro; however, he attributed this toπ lobe unsymmetriza-tion.127

VI. Radical Capture

In all of the foregoing sections, the newly forming bond can easily be argued to be an electron-deficient one, at least in the transition state, as either the reagents or the substrate atoms undergoing addition are electron-deficient from the start. The intriguing question arises whether the rules will be different in radical abstraction reactions, which are more electroneutral. An important starting point128_{is that}

2-adamantyl has a planar C-2 radical center, while 2-methyl-2-adamantyl has a shallow pyramidal con-figuration with a barrier to inversion of perhaps 5 kcal/mol.

Two reactions have been reported129_{as favoring the}

capture of 5-substituted 2-adamantyl radicals at the face syn to the electron-withdrawing substituent. In the first, the radical was generated in a chain process involving an initiator, bromotrichloromethane solvent and 2-methylene-5-phenyladamantane 88. Bromide Z-89 is then the principal product (64:36). The other study concerned the Hunsdiecker decarboxylation of acid 90 and bromine abstraction by the secondary radical; here also, syn capture predominates, by a margin of 57:43. Thus, the radical behaves the same way as do the carbocation and the carbene, albeit with more modest ratios; we attributed these obser-vations to transition-state hyperconjugation.129

A subsequent study by Adcock130 _{poses a serious}

question for this interpretation: it was found that, while 5-fluoro-2-adamantyl does react with stannyllithium to give an excess of Z-92, the trimethyl-stannyl-2-adamantyl radical produces only 50/50 mixtures of products; in other words, the 5-trimeth-ylstannyl group fails to exhibit the strong donor pro-perties so obvious in the 2-adamantyl cation. For this reason, Adcock prefers electrostatic control as the cause of the modest directive effects by 5-phenyl- and -fluoro substituents, and by extension, to nucleophilic addition to ketones and electrophilic addition to olefins as well (see also section XII). One is reminded of Bartlett’s search131 _{for evidence bearing on the}

question whether the formation of 2-norbornyl radi-cals is subject to the σ-assistance then so strongly suspected in the corresponding cation.

In still more recent work, we defended our original proposition by means of a study of radical 93, which is isoelectronic with 5-phenyladamantyl but has a powerful remote donor instead of an electron-withdrawing group.132 _{This radical abstracts}

deute-rium from tri-n-butyltin deuteride at the zu face by a margin of approximately 3:2 (note that with the boron at the 5-position having a lower priority than the carbon at the 7-position, the en and zu faces are opposite in location to where they normally are).

A recent study by Pincock133_{has shown that radical}

cation 94 is captured by cyanide anion to give primarily the Z-nitrile (58:42).

VII. Carbanion Capture

Little has been done about this topic with sterically unbiased probes. Attempts to study the enolate of

(15)

ketone 96 were unsuccessful;96_{it resisted}

bromina-tion, and protonation and methylation gave only the enol and enol ether, respectively. However, the 5-phenyl-2-adamantyllithium 97 derived from either of the two bromides gave the Z-carboxylic acid and the Z-bromide in excess over the E-products upon carboxylation and bromination, respectively.96_Thus,

as noted earlier, newly forming bonds tend to be electron-deficient in the transition state, regardless of which of the two reactants is contributing the two electrons.

VIII. Cycloaddition

Cycloadditions are perhaps the most important examples of pericyclic reactions, in which doubly bound atoms of the first row undergo addition, and others, singly bound, undergo simultaneous elimi-nation. While the Woodward-Hoffmann rules in symmetry-allowed cases successfully specify whether these reactions occur suprafacially or antara-facially, they are silent about which faces will be presented for reaction. When it appeared that transi-tion-state hyperconjugation could account for face election in both nucleophilic and electrophilic additions, it occurred to us that pericyclic chemistry might also prove to be predictable by means of this device.

Several thermal cycloaddition processes have been studied with this objective.134_{It should be noted that}

the concerted or stepwise nature of these reactions is not an issue in these studies: if transition-state hyperconjugation is a valid predictor of configuration in pericyclic chemistry, the face selection will not be affected by the simultaneity or even the sequence of the bonding steps.

The [2+2] cycloaddition of dichloroketene and 2-methylene-5-phenyladamantane gives E- and Z-98 in a 44:56 ratio. The enolate of methyl isobutyrate adds to 5-fluoroadamantan-2-one to give lactone E-99 in a 60:40 excess over isomer Z-99, and the latter product eliminates carbon dioxide six times faster than the former. 5-Fluoro-adamantane-2-thione 100 undergoes the Diels-Alder reaction with 2,3-dimethylbuta-1,3-diene; the adduct E-101 is in excess by a margin of 67:33. Furthermore, the more elaborate but still mostly rigid substrate 102 reacts with tetracyanoethene to give E- and Z-103 in a 42:58 ratio.134_{Thus, all of these}

cycload-dition processes exhibit face selectivities in accord with the notion of transition-state hyperconjuga-tion;135_{in fact, the Diels-Alder reactions mentioned}

show this principle to apply to both diene and dienophile.

Similar findings136_{have been reported for several}

1,3-dipolar cycloadditions: benzonitrile oxide adds to

5-halo- and 5-phenyladamantanethiones and -2-me-thylene-adamantanes to give, respectively, the E-∆2

-1,4,2-oxathiazolines 104 and the Z-∆2_{-isoxazolines}

105 in excess, by margins ranging from 69:31 to 56: 44.

Several photocycloaddition studies have also been published. Adamantanones carrying a variety of electron-withdrawing 5-substituents were found137_to

undergo the Paterno-Bu¨ chi reaction with electron-poor olefins such as fumaro- and maleonitrile as well as with electron-rich alkenes such as Z-diethoxy-ethene to give moderate but clear excesses of the oxetanes 106-108, respectively.138 _{An especially}

interesting observation139 _{was that a trimethylsilyl}

group behaves in the same fashion (i.e., as a syn director) as do known electron-withdrawing groups, which poses the question whether this group in the excited ketone functions as an acceptor group.

(16)

As noted earlier, an interesting protocol has also been devised by one of us (W.-S.C.) for reversing face selections in the reactions of adamantanone deriva-tives: ifβ-cyclodextrin is present during the addition reaction, it complexes the substrate in such a way as to protect the favored face and leave the alterna-tive site open to attack. Models show that the cavity in R-cyclodextrin is too small to form such a complex and thatγ-cyclodextrin can accommodate both sub-strate and reagent; indeed, neither of these additives strongly affects the ratio of the isomers during the photocycloaddition of fumaronitrile to 5-substituted adamantan-2-ones, whereas β-cyclodextrin causes clear reversals from syn to anti attack. In a typical example, 5-phenyladamantan-2-one in aqueous solu-tion gives a syn/anti approach product ratio of 62: 38; this is reduced somewhat to 58:42 if either R- or γ-cyclodextrin is present, but with 1.5 equiv of β-cyclodextrin present, the ratio is reversed to 23:77. Complex 109 is clearly responsible.140_{In later work,}

even larger effects were reported;141_{in the presence}

of β-cyclodextrin, 5-trimethylsilyladamantan-2-one suffers as much as 98% attack at the face anti to the silicon. A molecular mechanics calculation by Jaime142

has confirmed our interpretation.

Another interesting extension that should be mentioned in this section is the use of substituted adamantylideneadamantanes 110-X. Hummelen143

reported an early instance of face selectivity in 110-NHCOC2H3: singlet oxygen addition furnished a 3:2

excess of the dioxetane resulting from syn attack. A much more drastic example was later found by Nelsen,144 _{who found that the preference for syn}

approach at -78° was as large as 25:1 in this reaction with 110-Cl. The active species in this chain reac-tion is the radical careac-tion; an interesting perspective on this process derives from the observation145

that the radical cation of the parent hydrocarbon 110-H has an ESR spectrum revealing strong cou-pling of the signal with the four remote bridgehead hydrogen atoms. In more recent work, Nelsen has measured the ratios with other halogen derivatives in epoxidation, dioxetane formation, and cycloaddi-tion to N-methyltriazolinedione, but although they found syn attack to be predominant in all cases,

they also argued that there were small variations that were difficult to square with the Cieplak hy-pothesis alone.146 _{The chemiluminescent dioxetane}

fission to give excited adamantanones should also be of interest, but although one example of a kinetic study has been reported,147 _{the authors made no}

mention of having separated the isomers of their substrate 111.

A large amount of work has been lavished on the Diels-Alder reactions of 5-substituted cyclopenta-dienes which are not free of steric bias but with which variations in product composition as a function of the substituent suggest that the steric factor alone can-not account for all of the data. An early example was reported by Woodward, who found that 5-acetoxycy-clopentadiene 112 adds ethylene on its syn side.28_The

same contrasteric prejudice has been observed with 1,2,3,4,5-pentachlorocyclopentadiene,148_but

5-iodocy-clopentadiene undergoes Diels-Alder cycloaddition at its en face;149 _{evidently, if the 5-substituent is}

bulky enough, the reaction becomes subject to steric control.

A systematic study by Fallis150 _{showed that the}

reaction of 5-substituted pentamethylcyclopenta-dienes 113-X with maleic anhydride gave only the syn adduct when X ) OH, OMe, NH2, or NHAc; there

is a small excess of syn adduct if X ) SH but a large excess of anti adduct if X ) SMe, and this becomes the only product with X ) SOMe or SO2Me. The

authors initially briefly summarized the several explanations that had been offered in the literature for these and similar observations; in their subse-quent view, the stereochemistry of these reactions can be understood in terms of transition-state hy-perconjugation tempered or even negated by steri-cally demanding substituents.151

(17)

Another remarkable observation by Fallis152

con-cerns the Diels-Alder reactions of 2,5-dimethylth-iophene oxide 114, generated in situ from the thiophene by peracid oxidation. Several dienophiles were used; in all instances, the cycloaddition occurred exclusively in syn fashion.

Other rather stark examples have been published by Ishida,153 _{who reported that N-phenylmaleimide}

attacks 5-cyanopentamethylcyclopentadiene exclu-sively at the zu face and the 5-hydroxymethyl ana-logue equally exclusively at the en face, and by Trost,154_{who found that the dimerization of 115 gives}

primarily compound 116; in this case, the same compound furnishes both the diene and the dieno-phile, and in both roles, the face syn to the oxygen is the reactive one. Burnell155_{and his group have also}

contributed much valuable information derived from 5-substituted cyclopentadienes; although their data tend to be in general harmony with those quoted above, these authors have stressed mostly the steric aspect of these reactions.

Gandolfi and co-workers have used cis-3,4-disub-stituted cyclobutenes to study the stereochemistry of 1,3-dipolar cycloadditions,156 _{and here also, a}

remarkable dichotomy was observed. With all of the eight dipoles, cycloaddition occurs exclusively trans to the trimethylene bridge of 117; however, in cyclobutene 118, the same dipoles add predominantly (and in a few cases, exclusively) to the zu face. Similar findings were reported by H.-D. Martin;157

thus, cis-3,4-dichlorocyclobutene reacts with diazo-ethane to give predominantly the adduct 119, but 2-diazopropane gives mostly product 120. Clearly, syn addition is induced by the chlorine atoms un-less the dipole makes this process sterically pro-hibitive. With dimethoxycyclobutene 121, the syn stereochemistry is predominant even with 2-diazo-propane.

Paquette has contributed158_{a study of the}

Diels-Alder reaction of N-methyl-triazolidinedione with dienes 122-X; while the approach to the diene is in favor of the face anti to the benzo ring (75:25) when X ) H, the ratio is reduced to 49:51 when X ) F.

The hexacyclic structure 123 and its reaction with more than twenty dienophiles has been studied by Coxon,159 _Pandey,160 _{and Marchand.}161

In most cases, the cycloaddition occurred exclusively at the “carbonyl face” to give adducts such as 124. In a few instances, mixtures were obtained, and in one, with diethyl azodicarboxylate, cycloaddi-tion took place exclusively syn to the cyclobutane ring to give product 125 (with the nitrogen atoms pyramidal and in s-trans conformation). Except for the azo derivatives, all of the dienophiles appear to add in accordance with expectations based on the Cieplak mode of transition-state hyperconju-gation. Coxon also reported162 _{sharply divergent}

results for ether 126, which reacts primarily at the face syn to oxygen with maleic anhydride but at the opposite face with dimethyl acetylenedicarboxylate and N-phenyltriazolidinedione, and attributed the latter results to repulsion between one of the oxygen unshared pairs and the π electrons of these dieno-philes.

An interesting finding by Mehta163 _{concerns the}

contrast between dione 123 and its derivatives 127. When R ) H, 128 is the principal product (78:22), but when R ) OMe or OAc, only 129 is formed. Mehta attributed the behavior of the disubstituted dienes to twisted conformations at the 1,4-positions due to these substituents.

Halterman164_{has used a version of his probe (130)}

Face selection in addition and elimination in sterically unbiased systems