FROM DIYLS TO YLIDES TO MY IDYLLNobel Lecture, 8 December 1979byGEORG WITTIGHeidelberg, Federal Republic of GermanyTranslation from the German textChemical research and mountaineering have much in common. If the goal orthe summit is to be reached, both initiative and determination as well asperseverance are required. But after the hard work it is a great joy to be at thegoal or the peak with its splendid panorama. However, especially in chemicalresearch - as far as new territory is concerned - the results may sometimesbe quite different: they may be disappointing or delightful. Looking back at mywork in scientific research, I will confine this talk to the positive results (1).Some 50 years ago I was fascinated by an idea which I investigated experi-mentally. The question was how ring strain acts on a ring if an accumulation ofphenyl groups at two neighboring carbon atoms weakens the C-C linkage andpredisposes to the formation of a diradical (for brevity called diyl) (Fig. 1).Among the many experimental results (2) I choose the synthesis of thehydrocarbons 1 and 4 (3), which we thought capable of diyl formation. Startingmaterials were appropriate dicarboxylic esters, which we transformed into thecorresponding glycols. While these were obtained under the influence ofphenylmagnesium halide only in modest yield, phenyllithium proved to besuperior and was readily accessible by the method of K. Ziegler, usingbromobenzene and lithium. The glycolates resulting from the reaction withpotassium phenylisopropylide formed - on heating with methyl iodide - thecorresponding dimethyl ethers, which supplied the equivalent hydrocarbons 1and 4 by alkali metal splitting and demetalation with tetramethylethylenedibromide.Fig. 1. Formation of a diradical (2).The resulting tetraphenylbenzocyclobutane(1),however,rearranged to triphenyldihydroanthracene (3) (Fig. 2). In contrast,G. Wittig369tetraphenyldihydrophenanthrene (4), which was prepared analogously, provedto be a stable hydrocarbon, even when substituents R were introduced thatforced the biphenyl system to twist. While 4 did not decompose at 340°C andwas stable in solution against oxygen, its aryl-weakened C-C bond could beobserved since it split with potassium into the ring-opened dipotassiumderivate. The results of these investigations on formation of radicals and ringstrain seem to indicate that ring closure is more likely to contribute tostabilization of the ethane bond.Fig. 2. Some reactions observed in the synthesis of 1 and 4 (3).This stabilizing influence is documented impressively by the behavior oftris(biphenylene)ethane (7), which was also synthesized (4) (Fig. 3). Thecarbinol6, which was formed by the reaction of the ketone 5 with o-lithiobiphenyl,transformed into the desired hydrocarbon 7 by an acid-catalyzed twofoldWagner-Meerwein rearrangement.56Fig. 3. Formation of tris(biphenylene)ethane (4).370Chemistry 1979This first aromatic propellane, which melted at 475”Cwithout decompositionand whose structure agreed with the nuclear magnetic resonance spectrum,proved to be insensitive to ethane linkage-breaking sodium-potassium alloy.Evidently the close aryl packing prevents penetration of the metal into theinterior of the molecule.Since the tendency to form diradicals was not evident with the hydrocarbonsmentioned, we intended to replace the phenyl groups by anisyls. Therefore,suitable dicarboxylic esters should be brought into reaction with p-lithioanisole(5). But, as we obtained unexpected smears, the functionally simplebenzophenone was used to treat the mixture that resulted from the reaction ofp-bromoanisolewith lithium. Instead of the expected p-anisyldiphenylcarbinol,the bromine-containing compound 9 was isolated, whose structure could beproved by conversion into the well-characterized derivative by zinc dustdistillation. Accordingly, p-lithioanisole, which was originally formed, metalatesthe p-bromoanisole that is still present into compound 8 which then reacts withbenzophenone to form the isolated compound 9 (Fig. 4).Fig. 4. Formation and characterization of compound 9 (5).When it was noted that phenyllithium too can modify pbromoanisole toform 9, we decided to look closer at the lability of the aromatic proton as afunction of the substituent. In the course of these studies we arrived at thesurprising result that aryl iodide, bromide, and even chloride can exchangewith the electropositive metal of phenyllithium (6). Later we called thisprinciple ofreaction umpolung, or reversal of polarity (7) (Fig. 5). Simultaneouslyand independently, H. Gilman found the same behavior when treating arylhalides with butyllithium.Fig. 5. Reversal of polarity (umpolung) (6).G. Wittig371Among the halogens of the various aromatic systems, fluorine proved to benot exchangeable with lithium (8). Here we found an unexpected reactionpath. First we observed that in the formation of biphenyl by the reaction ofmonohalobenzene with phenyllithium, fluorobenzene acted rapidly, formingapproximately 75 percent biphenyl, while the other halobenzenes producedonly 5 to 7 percent. We interpreted this result as indicating that biphenylformation was preceded by metalation of the halobenzene, which was stimulatedby the inductive effect of the strongly electronegative fluorine. This explanationwas supported by the finding that not biphenyl but o-lithiobiphenyl had beenproduced. In 1942 we further assumed that an elimination of metal andhalogen results that leads to the occurrence of dehydrobenzene (9), and thisis what changes phenyllithium into the o-lithiobiphenyl found experimentally(Fig. 6). Independent of our work, a proof for the intermediate occurrenceof dehydrobenzene was given by Roberts et al. (10), who reacted[ 1-14C] chlorobenzene with potassium amide in liquid ammonia and isolatedthe two expected anilines with approximately 50 percent yield (Fig. 7).Fig. 6. Mechanism of formation of o-lithiobiphenyl (8, 9)Fig. 7. Proof of the intermediate occurrence of dehydrobenzene in the reaction of 14C-labeledchlorobenzene with potassium amide in liquid ammonia (10, 24).372 Chemistry 1979Later we could prove the existence of the dehydrobenzene by expecting thatit would react as dienophile (11). For the diene and solvent we chose furan,which, being an ether, should favor organometallic exchange whilesimultaneously serving as a trapping agent. In an exciting experiment we hadc+fluorobromobenzene react with lithium amalgam in furan and isolated withgood yield the