摘要:

QUARTERLY REVIEWS THE CHEMISTRY OF NATURALLY OCCURRING 1,2-EPOXIDES By A. D. CROSS (IMPERIAL COLLEGE OF SCIENCE AND TECHNOLOGY SOUTH KENSINGTON LONDON S.W.7) INCREASING attention has been paid to the chemistry of epoxidesl during the past two decades much of the stimulus being derived from their importance as synthetic intermediates. Several recent reviews have dealt with sugar epoxides2 and with mechanisms of epoxide reactions3 A later development has been the recognition of the epoxide group in a number of natural product^.^ Elucidation of their structures which vary greatly in com- plexity has frequently been complicated by intramolecular nucleophilic attack upon the reactive epoxide with novel skeletal rearrangements. Inter- pretations of these reactions have often involved subtle application of mechanistic and stereochemical principles.Indeed solutions ofthe structural problems presented by limonin and picrotoxinin must surely rank among the finest achievements of organic chemists. Physical methods have found extensive application and nuclear magnetic resonance spectroscopy is likely to prove especially useful for the detection and study of the epoxide group. It is the purpose of this Review to survey the chemistry of naturally occurring epoxides with special reference to interesting reactions of the epoxide group. No significance attaches to the order of presentation since no general classification is possible though many compounds do fall into the terpene class. Biogenetic relationships with other natural products are considered and the r61e of epoxides as biogenetic precursors is commented upon.An outline of possible biogenetic pathways is annexed. \ / / ‘ / 0 W I ,c=c -rc >c-c - -c-y- Ly t HQ QH 8 YH P - - -C-$- -c-c- Fi’ C- I P = Cyclisation products e.g. furanoid and pyranoid rings. Throughout this Review the term epoxide refers to the 1 ,Zepoxide (oxiran) group. Newth Quart. Rev. 1959 13 30. Parker and Isaacs Chem. Reviews 1959,58 737. As late as July 1956 Conroy5 when proposing his structure for picrotoxinin wrote “The oxiran group occurs in natural products so exceedingly rarely that its proposal merits exceptional concern.” Conroy J. Amer. Chem. Soc. 1957,79 1726. 1 317 318 QUARTERLY REVIEWS Generally sources of the various natural products are not enumerated since comprehensive lists of references are cited in footnotes.Monoterpene Epoxides.-Linalool epoxide (I) was identified by Naves and Bachmann,g and may be formed by aerial oxidation of linalool within the tree.’ Penfold and Simonsen* considered a constituent of the essential oil of Zieria Smithii to be “car-3-ene 5,6-epoxide” (11). However (v) (VI) (VI I) (VI I I) their findings have been re-interpreted by Blan~hard,~ who after further experiment concluded that the substance is ChrysanthenonelO (111). Piperitone oxide (IV) and piperitenone oxide (lippione) (V) occur in several Mentha species together with biogenetically related alcohols ketones olefins and the respective diosphenolsll (VI and VII); the diosphenols are readily obtained by acid treatment of the epoxides. Rotundifolone has been identified with piperitone oxide12 (IV) and alka- line rearrangement of this epoxide to the cyclopentane acid (VIII) has been rationalised as an intramolecular displacement through rear attack by the enolate anion upon the epoxide followed by cyclopropanone ring fission.13 Menthofuran (IX) is also isolable from Mentha extracts its formation from the olefinic ketone isopulegone by biochemical or acid- catalysed rearrangement of the intermediate epoxide14 (X) is considered to duplicate a biosynthetic pathway.Levisalles has briefly discussed this mode of furan bi0gene~is.l~ Sesquiterpene Epoxides.-Pyrethrosin was shown to be a 10-membered ring sesquiterpenoid lactone in which the epoxide is singularly susceptible Naves and Bachmanxi Helv. Chim. Acta 1945,243 1227 1231. Simonsen and Owen “The Terpenes” University Press Cambridge 1947,Vol.I Blanchard jun. Chem. and Ind. 1958 293. 2nd Edn. p. 66. * Penfold and Simonsen J. 1939 1496. lo Cf. Kotake and Nonaka Annalen 1957,607 153. l1 Reitsema J. Amer. Pharm. ASSOC. 1958 47 265. l2 Shimizu Bull. Agric. Chent. Sac. Japan 1957 21 107. Nelson and Mortimer J. Org. Chem. 1957 22 1149. l4 Fritel and Fetizon J. Org. Chem. 1958 23 481. lri Levisalles Perfumery Essent. Oil Record 1958 49 627. CROSS NATURALLY OCCURRING 1 ,ZEPOXIDES 319 to transannular nucleophilic attack.16 The earlier structure16 (XI) has been revised1' to a lactone (XII) with the ring closed at the 8- rather than the 6-position. This change does not affect the original interpretation of the cyclisation mechanism as for example with chromic acid (XII+XIIJ + XIV. U (xv) (xv 1) Parthenolide (XV) like pyrethrosin is found in the Chrysanthemum genus.ls Ozonolysis of dihydroparthenolide gave formic acid and a methyl ketone aldehyde by rearrangement of the ap-epoxy-ozonide (XVI) probably as outlined.18 Caryophyllene a 9-membered ring sesquiterpene of novel structure also occurs as its epoxide19 (XVII) the chemistry of which has been reviewed.16 Features of interest are the acid- and base-catalysed transannular reaction products (XVIII and XX) of caryophyllene oxide (XVII) and a ketonic oxidation product (XIX) respectively.Daucol the per-acid oxidation product of carotol,20 is a 1,5-epoxide not a l,2-epoxide2l The Prague group have reported the isolation of isolaserpjtin22 (XXI) also ~ynthesised~~ from its known natural olefinic precursor deoxodehydro- laserpitin22 by the action of per-acid.Both the epoxide and the olefin may l6 Barton and de Mayo J. 1957 150; Quart. Rev. 1957 11 189. l7 Barton Bockman and de Mayo J. 1960 2263. l8 Herout SouEek and Sorm Chem. and Znd. 1959 1069. l9Treibs Chem. Ber. 1947 80 56. 2o Asahina and Tsukamoto J. Phurm. SOC. Japan. 1925,525,961. 21 Sqkora Novotny and Sorm Tetrahedron Letters 1959 No. 14 24. a2 Holub Herout and sorm Coll. Czech. Chem. Comm. 1959,24 3926. 320 QUARTERLY REVIEWS be regarded as intermediates in the biosynthesis of la~erpitin~~ (XXII). Picrotoxinin and picrotin are the two components of the bitter principle “picrotoxin”. The constitution of picrotoxinin (XXIII) was brilliantly deduced by C ~ n r o y ~ ~ who interpreted and augmented previous extensive (XVI I> (XVI I I) R = COCMe CHMe investigation^.^^ No simple proof of the presence of an epoxide in picro- toxjnin has been adduced and its existence is accepted primarily because the structure (XXIII) accounts rationally for the many reactions of this remarkable cornpo~nd.~ jS-Brornopicrotoxinin (XXIV) is converted by alkali into p-bromopicrotoxinic acid (XXV) which yields a-picrotoxininic acid (XXVII) on zinc debromination.Reactions involving the epoxide of picrotoxinin demonstrate the supreme importance of three-dimensional geometry in the chemistry of rigid cage systems. Neither picrotoxinin nor its derivatives containing the intact epoxide ring are attacked by hot dilute acid in no case is the related 1,2-glycol obtainable directly. Consideration of the structures (XXIII- XXV) reveals that the epoxide is strongly shielded by a lactone ring against nucleophilic attack from the rear by external anions.Indeed the epoxide is only opened when an anion is generated internally from either the shielding lactone or another oxygen atom. Interesting illustrations are the reduction of /3-bromopicrotoxininic acid (XXV) by borohydride with an internal nucleophilic displacement as indicated (XXV+XXVI) and of /3-bromopicrotoxinin (XXIV) by lithium aluminium hydride to the trio12* (XXVIII). A similar explanation has been offered for the reduction of picr~toxinin.~~ 28 8s Sorm personal communication. 24 Conroy J. Amer. Chem. SOC. 1951,73,1889; 1957,79,1726,5550; Chem. and Ind. 1957,604. 25 The early literature (to mid-1949) has been collected,26 and more recently reviewed together with the later publications (to mid-1957) which led to the structure 27(XXIIT).For this Review primarily concerned with the epoxide chemistry the post-1956 publica- tions are particularly relevant. 26 Sutter and Schlittler Helv. Chim. Acta 1949 32 1855. 27 F. Korte Barkemeyer and I. Korte Fortschr. Chern. org. Naturstqfe 1959,17 155. 28 J. S. E. Holker K. U. Holker McGookin Robertson Sargeant and Hathway J. 1957 3746. CROSS NATURALLY OCCURRING I ,2-EPOXIDES 32 1 Thermal decarboxylation of di hydro-a-picrotoxinink acid (XXVII isopropenyl reduced) furnishes picrotoxinide an ap-unsaturated cyclo- pentenone. Though glycidic acids are usually decarboxylated to the a- ketone abnormal opening with hydride-ion migration is postulated by Conroy for this reaction as outlined.24 Alkali attacks p-bromopicrotoxin- inic acid (XXV) at the &lactone carbonyl to give an oxanion (cf.action of BH,-) which displaces the epoxide to yield an ortho-acid itself susceptible to attack by base Formation of picrotoxic acid (XXIX) involves internal nucleophilic attack on the epoxide by the C(3) hydroxyl liberated on opening of the lactone ring in picrotoxinin by base.24*29*30 (XXlll) (XXV) (XXVI) 0 &?JocH2Br oc-0 (XXIV) (XXVI I) CO H (XXVI I I) In these and intramolecular rearrangements conformational changes to less strained structures are apparent and probably provide a driving force €or many of the reactions. Conformational factors dictate which oxygen atom attains the sterically favourable position for epoxide displacement. A biogenetic relation between picrotoxinin and steroids has been suggested,32 but it appears more attractive to consider picrotoxinin as a sesquiterpenoid whose carbon skeleton is arrived at by cyclisation of a farnesyl chain followed by two 1,2 C-methyl migrations (XXX-tXXXI) with subsequent oxidation and lactonisation.Tutin and coriamyrtione occur in several Coriaria species. Both show an extraordinarily similar physiological activity to picrotoxinin and BB Burkhill J. S. E. Holker Robertson and Taylor J. 1957 4945. 8o Hathway J. 1957 4953. s1 J. S. E. Holker Robertson Taylor K. U. Holker and Williamson J. 1958,2987; Carman Hassan and Johns J. 1959,130. Conroy J. Amer. Chem. Soc. 1952,74 3046. 322 QUARTERLY REVIEWS mainly on the basis of analogous chemical reactions Kariyone and Okuda advanced constitution (XXXII) for c~riamyrtione.~~ (XXIX) f OH pox (XXX) A (XXXII) (X = e.g.,enzyme surface) (XXXI) Fumagillin is a potent antibiotic isolable from the mould Aspergillus fumigatus.Early reports established fumagillin as the hemiester of deca- 2,4,6,8-tetraene-l 10-dicarboxylic acid and alcohol-I the monomethyl ether of a sesquiterpenoid di01.~~ First results underlined the complexity and practical difficulties of the problem and confirmed the presence of one methoxyl a secondary hydroxyl and a 4-methylpent-3-enyl side chain oxygenated in the 1-position. After the existence of an epoxide ring in alcohol-I was p r o ~ e d ~ ~ ~ ~ a range of reduction products was prepared which required that alcohol-I contain a second reducible ether ring and only one carb~cycle.~~ Tarbell and his collaborators thereafter solved the structural problem finally assigning non-isoprenoid structures to fumagil- lin (XXXIII) and alcoh01-1~~ (XXXIV).Nuclear magnetic resonance spectral data provided valuable confirmatory evidence in the later work. The two epoxides dominate the chemistry of alcohol-I and are respons- ible for the profusion of and base-~atalysed~~ i somerisation and hydration products. Much interest remains in the stereochemistry and mode of formation of these derivatives. Of the two epoxides that on the side chain is less reactive than the spiro-epoxide. Elimination of the tertiary hydroxyl from tetrahydro-alcohol-Iab monoacetatet (XXXV ; R = OH R’ = OAc) took place without skeletal rearrangement to give a mixture of olefins and epoxidation of the exomethylene isomer afforded 33 Kariyone and Okuda Bull. Inst.Chem. Res. Kyoto Univ. 1953 31 387; cf. Chem. Abs. 1954,48 9971. 34 Schenk Hargie Tarbell and Hoffman J. Amer. Chem. SOC. 1953 75 2274; Brown and Landquist Chem. and Ind. 1953,973. 35 Landquist J. 1956 4237. 36 Ross Tarbell Lovett and Cross J. Amer. Chem. Sac. 1956 78 4675. as Cross and Tarbell J. Arner. Chem. SOC. 1958,80,3682. 7 Nomenclature for the reduction products is given in ref. 36. Tarbell Carman Chapman Huffman and McCorkindale J. Arner. Chem. SOC. Chapman and Tarbell J. Arner. Chern. Soc. 1958,80,3679. 1960,82,1005. CROSS NATURALLY OCCURRING 1 ,ZEPOXIDES 323 dihydro-alcohol-Ia acetate (XXXIV; R = Ac side chain reduced). The isomeric cyclic olefin gave isohexanaldehyde on treatment with periodate. Before the existence of a second epoxide ring was established this last result had been interpreted by assigning the terminal methylene epoxide function to the side chain ar-po~ition.~~ @-q-Jwp OMe OMe OMe OH (XXXV) (XXXVD OR R' (XXXIII ; R=CO.[CH:CH];C02H) (XXXIV ; R = H) Hydrogenation of the olefin mixture and reduction of the derived tosylate (XXXV; R = H R' = OTos) led to a dideoxy-derivative (XXXV; R = R = H) the epoxide of which is isomerised by acid to furnish an allylic alcohol (XXXVI). Attempted tosylation of the deoxy-disecondary alcohol (XXXV; R = H R' = OH epoxide reduced) gave the perhydro- benzofuran (XXXVII; R = OTos) by cyclisation with loss of methanol analogous cyclisations are known.42 Dehydrogenation of the perhydro- benzofuran (XXXVII ; R = H) furnished the corresponding benzofuran identical with a synthetic specimen and 6-rnethyl-2-0-tolylheptane.~~ (XXxVl I) (XXXVl I I) (XXXlX) When the /I-hydroxy-epoxide (XXXV; R = R' = OH) was treated with lithium aluminium hydride simple reduction of the epoxide occurred and also a base-catalysed isomerisation followed by reduction to the allylic alcohol (XXXVIII) presumably by the mechanism which operates when moradiol diacetate oxide (XXXTX) is treated with this reagent.41 Olefin formation from a /3-hydroxy-epoxide may be considered41a a further example of 'fragrnentatio~i'.~l~ Steroid and Triterpene Epoxides Microbiological Epoxidation.-Apart from a small group of toad venoms and limonin and its congeners epoxides are rare among steroids and triterpenes.This is surprising since 40 Carman Chapman McCorkindale Tarbell Varino West. and Wilson J. Amr. Chem. SOC.1959 81 3151. 41 Barton and Brooks J. 1951 257. 41a Tarbell personal communication 41b Grob Experientia 1957 13 126. 4aChapman Cremer Carman Kunstmann McNally Rosowsky and Tarbell J. Amer. Chern. Soc. 1960,82 1009. 324 QUARTERLY REVIEWS Bloom and Shull have established that several micro-organisms capable of introducing an axial hydroxyl at C of a saturated steroid can also in- troduce an epoxide group “axial” at C in the corresponding unsaturated Moulds which cause equatorial h ydroxylation do not effect ep~xidation.~~ Hydroxylation and epoxidation may represent two aspects of the same or similar mechanisms involving initial attack by peroxy- radical (or the equivalent in the enzymatic system) followed by loss of h y d r ~ x y l . ~ ~ By these means Bloom and Shull prepared 9B,l lp- and 14a,15a- epoxides.Examples of hydroxylated steroids are legion the corresponding epoxides do not therefore appear to be intermediates in sterol formation in vivo except possibly where 1,2-glycol systems result. Microbiological transformation of epoxides are known as exemplified by the conversion of the pregnanone (XL) by fermenting yeast into the diol (XLII),45 a C-methyl migrating in the initially-formed carbonium ion (XLI). Tschesche’s structure (XLIIIa) for scymn01,~~ a constituent of shark oil was altered to (XLIIIb) by Asikari4’ on the strength of its conversion into a known cholic acid. Fieser and F i e ~ e r ~ ~ noted that warm chromic acid would attack such an epoxide and proposed a trimethylene oxide ring (XLIIIc) as a likely alternative. Glycol-cleavage experiments on products obtainable from opening of the oxide ring by reduction or hydration and nuclear magnetic resonance spectral studies should permit a more definite assignment of structure.HO VR” @ A H (XLI I I) From their investigations R R‘ R” 0”H O;} - & o ~ c ~ ~ ~ i ~ -CH-&H-CtiiOH C #-OH H f)-CH of the toad poisons Meyer and his collabora- 43 Bloom and Shull ibid. 1955 77 5767; U.S. Patent 2,830,935; cf. Chem. A h . 44 Bloom Hayano Saito Stone and Dorfman Federation Proc. 1956 15 222. 45 Camerino and Vercellone Gazzetta 1956 86 260. 46 Tschesche 2. physiol. Chern. 1931 203 263. 47 Asikari J. Biochem. Japan 1939 29 319. 48 Fieser and Fieser “Steroids” Chapman and Hall London 1959 p. 432. 1958,52 14723h CROSS NATURALLY OCCURRING 1,2-EPOXIDES 325 50 arrived at structures for marinobufagin (XLIVa) and resibufogenin (XLIVb).The 148,15/3-epoxide group a possibility also envisaged by Thies~en,~~ is demonstrated by the reactions with anhydrous and with aqueous acid as depicted. Meyer’s group52 also developed structures for the closely related compounds cino bufagin (XLIVc) and cinobufotalinin probably (XLIVLI),~~ while jamaicobufagin appears to contain an e p ~ x i d e ~ ~ (infrared absorption data) presumably 148,15p by analogy. Bufotalinin has the constitution (XLIVe).55 All five compounds (XLIV a-e) have been inter-related or converted into other toad poisons of established structure. R R’ R ” g OH Me H b H M e H 5 H Me OAc 4 OH Me OAc 5 OH CHOH (XLIV) {fi=={fi HO Cl 0 -&OH H -{fro A In view of Bloom and Shull’s observation^^^ (see above) it is of interest that all toad poisons of known constitution contain a 14/3-hydro~yl~~ or a 14/3,15p-epoxide with a cis-fused C/D ring junction in which the 14/3-0 bond has an equatorial (with respect to ring c) orientation.Formation of both a- and /I-epoxides in the cholestane series suggests that enzymatic oxidation is not seriously ,subject to steric hindrance. Nevertheless it would be of interest to know whether enzymes capable of oxidation at the a-face in the cholestane series can also effect oxidation of the same carbon atom from the /I-face in the coprostane series. Derivation of the 14/3,158- epoxide in the toad venoms by direct enzymatic oxidation as proposed here differs from the mode of formation suggested by Thie~sen.~~ The s t r ~ c t u r e ~ ~ . ~ ~ and stereochemi~try~~ of limonin (XLV) the bitter principle of citrus fruits have been elucidated more than 100 years after 40 Linde and Meyer Helv.Chim. Acta 1959 42 807. 5 0 Schroter Rees and Meyer Helv. Chim. Acta 1959 42 1385. 51 Thiessen Chem. and Ind. 1958 440. 52 Hofer Linde and Meyer Experietitia 1959 15 297. 53 Prof. Meyer personal communication. 54 Barbier Schroter Meyer Schindler and Reichstein Helv. Chim. Acta 1959 42 65 Schroter Tamm and Reichstein Helv. Chim. Acta 1958,41 720. 56 Tamm Fortschr. Chem. org. Natrirstofe 1956 13 188. 57 Arigoni Barton Corey; and Jeger in collaboration with Cagliotti Dev Ferini Glazier Melera Pradhan Schaffner Sternhell Templeton and Tobinaga Experientia 1960 16 41. 5* Arnott Davie Robertson Sim and Watson Experientia 1960,16,49. 2486. 326 QUARTERLY REVIEWS its isolation was announced.59 Three teams led by Arigoni and Jeger Barton and Corey deduced two alternative structures one (XLV) being independently arrived at through X-ray crystallographic study by Monteath Robertson and his co-w~rkers.~~ The X-ray group examined the iodoacetate of epilimonol (XLVI) the borohydride reduction product of limonin and were solely responsible for the determination of the stereo- chemistry.Evidence for the a/3-epoxy-6-lactone came from the conversion of limonin into desoxylimoninso (XLVII) by hydriodic acid or chromous chloride. Tetrahydrolimonin (XLV ; furan ring reduced) was transformed by acid with epoxide ring opening and a cis-hydrogen shift into a keto- lactone isolated as the enolic form (XLVIII) and which on ozonolysis followed by hydrolysis furnished oxalic acid similar acid treatment of deoxytetrahydrolimonin caused no reaction as expected.A novel re- arrangement of the glycidic acid moiety of hexahydrolimoninic acid (XLIX) with acid led to a neutral y-lactone (L). Oxidation of the latter gave a non-enolisable a-keto-y-lactone. Another noteworthy reaction involving the epoxide was the treatment of limonol (LI) with base leading 6D Bernays Anmlen 1841 40 317. CROSS NATURALLY OCCURRING 1,2-EPOXIDES 327 to merolimonol (LIII) and f~ran-3-aldehyde,~O apparently via the inter- mediate trimethylene oxide (LII) since neither epilimonol (equatorial 7/3-hydroxyl) nor limonin gives the same reaction. Limonin" (XLV) may be considered a degraded tetracyclic triterpenoid of the euphol type a biogenetic scheme has been outlined involving cleavage and rearrangement reactions of established pre~edent.~' Ring D is postulated to arise by epoxidation of a cyclopentenone followed by a Baeyer-Villiger ring expansion to the a#?-epoxy-8-lactone.This necessitates epoxidation from the hindered side of the molecule (cf. the toad venoms p. 325) as a consideration of limonin (LIV) and molecular models of 1 3a-methyl-7-oxo-14,15-unsaturated triterpenoids illustrates. B-Epoxida- tion locks ring c in its boat conformation. Plausible structural formulae have been suggested5' for nomilin61* 62 and obacunone (ca~imirolide),~~-~~ (LV) and (LVI) respectively. No stereo- chemical correlations have as yet been established though recent experi- mental work supports these structures and C/D trans-ring fusion.64 By analogy with limonin the epoxide is expected to be 148,15b.Carotenoid Epoxides.-A neat picture of biogenetic relations within a class of natural products is presented by the carotenoids with epoxides as intermediates in the formation of furanoid oxides from olefins (e.g. LVII+LIX). Epoxides of one or both terminal cyclic ethylenic bonds of the polyene chromophore (LVII or LXII) are found and the isomerisa- tion (enzymatic or acid-catalysed) to 2,5-dihydrofurans (furanoid oxides) is of diagnostic value since it results in a characteristic change of the visible spectrum. The known natural epoxides all Cqo compounds (LX; R and R = CI1 units) are listed in Table 1. Discovery of more epoxides is to be expected and an early review by KarreF5 already requires revision. * Numbering and lettering in limonin (XLV) follow biogenesis from a euphol skeleton.6o Melera Schaffner Arigoni and Jeger Helv. Chim. Acta 1957 40 1420 and the 61 Emerson J. Amer. Chem. SOC. 1948,70 545; 1951 73 2621. Dean and Geissman J. Org. Chem. 1958 23 596. 83 Sondheimer Meisels and Kincl J. Org. Chem. 1959 24 870. 84 Barton Sternhell and Templeton personal communication. 65 Karrer Fortsch. Chem. org. Natwstofe 1948 5 1. earlier literature summarised therein. TABLE 1 Related Ca carotenoid olejins epoxides and furanoid oxides** (LX) w N 00 Mono-epoxide 7 t Mono-furanoid Di-epoxide t Di- furanoid oxide t * Ethylenic precursor a- Car0 tene R = LVII R‘ = LXI P-Carotene oxide Flavochrome 66b R = LIX R’ = LM Citroxant hin (mutatochrome) Rubichronie 66c R = LIX R’ = LVII 66d R = LXIV R’ = LXV Mutatoxan thin 66e R = LXIV R’ = LXII Flavoxan thin maxanthin) 66e (chrysanthe- R = LXIV R’ = LXVI Trollichrome R = LXIV ? R’ = LXVII Cryp tofla vin - a-Carotene epoxide R = LVIII 66a R’ = LXI 8- Car0 tene monoapoxide R = LVIII 67 R’ = LVII Rubixanthin epoxide R = LXIII - R’ = LXV Antheraxanthin R = LXIII 66e R’ = LXII Eloxanthin 66e f * 8- Carotene - di-epoxide R = R’ = LVIII 69 Aurochrome R = R’ = LVII R = R’ = LIX * Rubixanthin R = LXII R’ = LXV Zeaxanthin R = R’ = LXII * Violaxanthin R = R’ = LXIII 66e Auroxanthin R = R’ = LXIV 66e Xanthophyll * R = LXII R’ = LXVI Unknown R = LXII ? R’ = LXVII Cryptoxanthin R = LXIII R’ = LXVI Trollixanthin R = LXIII 66g ? R = LXVII Cryp toxan th in monoepoxide R = LVIII - R’ = LXII * * Cryptoxanthin diepoxide R = LVIII - R’ = LXIII Cryptochrome R = LVII R’ = LXII 66h R = LIX R’ = LXII 66h R = LIX R’ - LXIV f Reference to elucidation of structure of epoxide.$ Reference to synthesis of epoxide. * Epoxides of unconjugated or acyclic terminal olefin groups are not found. ** Italics denote synthetic only. CROSS NATURALLY OCCURRING 1,2-EPOXIDES 329 Structures66Q for trollixanthin and trollichrome (see Table) may require revision since the allylic tertiary alcohol functions proposed should prove highly labile especially in acid leading to rearrangements. Taraxanthin and tarachrome appear to be stereoisomers of trollixanthin and trolli- chrome.66i Revised structures for capsanthin (LX; R = LXII R' = LXVIII) and capsorubin (LX; R = R' = LXVIII) are consistent with R / (y- GLr OH (LXV) (LXVI I 0 (LXI ; R=R'=H) (LXVI;R=OH,R=H) (UVI I; R=R'=OH) biosynthesis from antheraxanthin and violaxanthin respectively by pina- colic rearrangement of the epoxides or related oxygenated intermediates.Nuclear magnetic resonance spectral studies of carotenoids indicate that this technique may aid detection of the epoxide group through proton signals of the methyl attached to epoxidic carbon.69 Carotenoid epoxides seem to fulfil an important r81e in Nature as oxygen donors. 70 Both 01- and p-carotene epoxides display provitamin A effects in mammals,66i but the epoxide group is not the cause of the physiological activity. Epoxides of Coumarins etc.-A number of oxygen heterocycles is known in which an isoprenoid side chain exhibits the sequence olefin+epoxide +glycol+cyclisation products the related compounds sometimes occur- ring in a single plant.Cyclisation products may conceivably arise from one of several oxidation states of the side chain. Seshadri and his colleagues71 have reviewed such compounds and since the epoxide chemistry is un- exceptional only brief mention is made here. 66 Karrer et al. (a) Helv. Chim. Acta 1945,28 1146; (b) ibid. p. 471 ; (c) ibid. p. 427; ( d ) ibid. 1947 30 531 ; ( e ) ibid. 1945 28 300; ( f ) ibid. p. 1526; (6) ibid. 1955 38 638; (h) ibid. 1946 29 229; ( i ) ibid. 1957 40 69; ( j ) ibid. 1950 33 1481. 67 Savinov and Protsenko Ukrain. khirri. ZJwr. 1954 20 399. 68 Barber Jackman Warren and Weedon Proc. Chem. SOC. 1960 19. 69 Barber Davis Jackman and Weedon J. 1960 2870. 'O Cholnoky Gyorgyfy Nagy and Panczel Nature 1956 178 410. 71 Aneja Mukerjee and Seshadri Tetrahedron 1958 4 256.330 QUARTERLY REVIEWS O~ypeucedanin~~ (LXIX) isoimperatorin the corresponding olefin and ostruthol a monoester of the related 1,2-glyco! are all found in Impera- toria ostruthium. Phellopterin and byakangelicin are the biogenetically related olefin and glycol respectively of byakangelicol 73 (LXX). A ~ r o p t e n ~ ~ (maranzin) (LXXI) is the epoxide of osthol and a c ~ l e a t i n ~ ~ (LXXII) is related to the glycol toddalolactone. Fukugetin (garcinin) is an e p ~ x i d e ~ ~ (LXXIII). pJ4J-y / / ;,:&y R' R"' /o\ 0 (U(IX;R=H,R:.0.CYCH.CMe2) (LXXI ; R=Cl-L#H-kMe2,&~~ R' H) (LXX ; R=OCHiCHkM%.R'=Ce) 10 (MI I ; R=H ,R'= RzOMe d:CH;CH.CMe,) P U (WXI I I) Fatty Acid Epoxides.-Since G~nstone'~ showed vernolic acid to be cis-12,13-epoxyoctadec-9-enoic acid (LXXIV) not 1 l-hydroxyoctadec-9- enoic acid as suggested earlier,78 three more cis-epoxy-acids have been discovered and their structures determined all are C I S compounds.0 / \ 0 / \ Me[CH2]4CH.CH.CHiCH CH+J-~&~H Me.[CH,],CH -CH.[CH&CO,H (LXX I v) (LX xv) (LXXVI) (UXVI I) 0 / \ 0 / \ Me-[C%]iCH :CH CH;CHCH. [CH2]jCO2H MeCHiCHCH.[CH,CH :CHI i[CH&02H 15,16-Epoxylinoleic acid (LXXVII) is a constituent of cameline oil 7 9 and coronaric acid is very probably ci~-9,10,epoxyoctadec- 12-enoic acid (LXXVI).80 Very recently cis-9,lO-epoxystearic acid (LXXV) was isolated from tragopogon and from the spores of numerous parasitic fungiS2 72 Spath and Klager Ber. 1933 66 914. 73 Noguchi and Kawanami Ber. 1939,72,483. 74 Bohme and Schneider Ber. 1939 72 780. Dutta J.Indian Chem. Soc. 1942 19 425. 76 Murakimi and Irie Doc. Imp. Acad. Tokyo 1934 10 568. 77 Gunstone J. 1954 1611. 78 Vidgarthi Patna Univ. J. 1945 1 51. 79 Gunstone and Morris J. 1959 2127. 79a Osbond Proc. Chem. SOC. 1960 221. 8o Smith jun. Koch and Wolff Chem. andInd. 1959 259. Chisholm and Hopkins Chem. and Ind. 1959 p. 1154. 82 Tulloch Craig and Ledingham Canad. J. Microbiol. 1959 5 485; Tulloch personal communication and Canad. J. Chem. 1960,38 204. CROSS NATURALLY OCCURRING 1,2-EPOXIDES 33 1 Thus the series 9,10-epoxy-stearic oleic and linoleic acid exists as pre- dicted by Gunstone and Morris. 79( &t)-Vernolic acid has been synthesi~ed.~~~ Fatty acid epoxides undergo standard epoxide addition hydrogenation and olefin and 1,2-glycol oxidative cleavage reactions. Hopkins and Bernsteins3 have described a nuclear magnetic resonance spectroscopic method for detection of epoxides in natural oils where the amount exceeds 5 %.In support of the biosynthetic scheme the corresponding 1,2-dihy- droxy- and keto-acids occur naturally. Mould Metabolite Epoxides.-Moulds are the source of a rich variety of chemical structures. Among the simpler products is ethylene oxide-tram- 1 ,2-dicarboxylic another metabolite of Aspergillus fumigatus (see p. 322). Three of the seven metabolites isolated from Coprinus quadrijidus are believed to constitute the sequence olefin-tepoxide (LXXVIII)+1,2- glycol.85 A provisional structure incorporating an epoxide ring for sclerotiorina6" has been abandoned by the same authors.s6b Alkali isomerisation of the epoxide in terreic acid (LXXIX)87 led to a mixture in which the presence of 3,6-dihydroxytoluquinone (LXXX) was verified by reductive acetylation to the tetra-acetoxytoluene (LXXXI ; R = H).Opening of the epoxide by treatment with boron trifluoride- acetic anhydride furnished the penta-acetate (LXXXI; R = OAc). Proton resonance absorption for the epoxidic protons of terreic acid occurs at the same field strength as those h in 2,3-epoxynap ht haquinone. n HCC-CCCCCH-CH /o CH~OH Me@ Hob IoH AcO\ MC@C R HO (UXVI I I) U (UXIX) Magnamycins8 (LXXXII) the first of the macrolide epoxides to yield to structural investigation has been amply reviewed.89 Another macrolide 83 Hopkins and Bernstein Canad. J. Chem. 1959,37,775. 84 Birkinshaw Bracken and Raistrick Biochem. J. 1945 39 70. 85 Jones and Stephenson J. 1959 2197.86 (a) Robertson Whalley et. al. Chem. SOC. Special Publ. 1956 No. 5 p. 27; 88 Woodward Angew. Chem. 1957 69 50; "Festschrift Arthur Stoll" Birkhauser 8 9 van Tamelan Fortschr. Chem. org. Naturstofe 1958 16 90; Brink and Harman (h) idem J. 1958 1814. Basel 1957 p. 524. Quart. Reviews 1958 12 93. Sheehan Lawson and Gaul J. Amer. Chem. SOC. 1958 80 5536. 332 QUARTERLY REVIEWS antibiotic pimaricingo (LXXXIII) contains an ap-epoxy-carbonyl func- tion manifest from the liberation of iodine on addition of iodide ion. Hydrogenation saturated the tetraene chain and reduced the epoxide to an acetylatable hydroxyl. Cyclodehydration of N-acetyldodecahydro- pimaricin took place in acidic media with the formation of a fury1 ketone (LxxxIv+Lxxxv). CH,*OH 0 CHmOH CHiOH ++ OH OH I ( L X M I I i) (LXXXIV) (LXXXV] H2z/i“ 0’ (LXXXVa) Oleandomycin a third macrolide readily forms a chlorohydrin from which the antibiotic is re-formed by treatment with base properties indicative of an epoxide.91 The complete structure of oleandomycin (LXXXVa) is now known.91u Alkaloid Epoxides.-A few Senecio alkaloids some tropane alkaloids undulatine and annotinine contain an epoxide group. An early structure for quinamine was based on its conversion into cinchonamine by lithium aluminium hydride. 92 Witkop rejected an epoxidic structure after effecting this transformation in reverse using peracetic acid which does not oxidise indoles to their 2,3-epoxy-derivatives. 93 Jacobineg4 and tomentosineg5 (otosenineg6) two Senecio a1 kaloids contain the same epoxidic acid moiety. Alkaline hydrolysis of jacobine yields retronicine (an amino-alcohol) jaconecic acid and isojaconecic acid.Jaconecic acid and an unknown amino-alcohol have been obtained (a) Patrick Williams Wolf and Webb J . Amer. Chem. SOC. 1958 SO 6688; (b) Patrick Williams and Webb J . Amer. Chern. Sac. 1958 80 6689. 91 Els Celmer and Murai J . Amer. Chem. SOC. 1958 80 3777. 91a Wood ward personal communication. s2 Gnutarel Janot Prelog and Taylor Helv. Chim. Acta 1950 33 150. O3 Witkop J . Amer. Chem. SOC. 1950 72 231 1. 94 Adams and Gianturco in “Festschrift Arthur Stoll” Birkhauser Basel 1957 p. 72; g5 Adams Gianturco and van Duiiren J . Amer. Chem. SOC. 1956 78 3513. Q6 Santavy Planta Med. 1958 6 78; Chem. Abs. 1958 52 14,971. Bradbury and Culvenor Austral. J. Chenz. 1954 7 378. CROSS NATURALLY OCCURRING 1 ,&EPOXIDES 333 from similar treatment of tomentosine though isojaconecic acid should also be formed.97 Several structures were advanced for jaconecic acid (LXXXVI 98 LXXXVII,99 and LXXXVIIP5) and isojaconecic acidg8 (LXXXIX) but none accounted satisfactorily for the known facts,s4 since the tacit assump- tion was made that the epoxide ring present in the alkaloid remained intact during alkaline hydrolysis.Acceptable structures have now been formulated by Geissmanlo0 and Bradbury and Masamuneg7 for jaconecic acid (XC) isojaconecic acid (XCI) and jacobine (XCII). The cyclisation mechanism (XCIII; R and R' = H or amino-alcohol) could operate before or after hydrolysis of one or both of the ester linkages. Related alkaloids found with jacobine include jacoline jaconine and senecionine which are respectively the 1,2-glyc01,~~~ chlorohydrin,loO and olefin corre- sponding to the epoxide group in jacobine.Laboratory interconversions have been realised.loO With hydrochloric acid jacobine is converted into jaconine and a chloro-dilactone long thought to be a di-y-lactone (vcE0 1781 cm.-l) but now a s ~ i g n e d ~ ~ * ~ ~ ~ a &lactone structure (XCIV) formed by abnormal epoxide opening ester hydrolysis and lactonisation. Infra- red and nuclear magnetic resonance spectral studies on analogous com- pounds supported the new structures. Me / \ 0 h;le Ye /O Ft /O\ ?H H O i i H *F-FCH ,CH20H H0,CCH *~-~CHiCH(OH) .Me H0,CCH *CH-F-$-CO,H Me C02H Me COP Me Me (LXXXVI) (LXXXVI I) ( U X X V I I I) Ye Y byle (?CH.FH2 Me H0,C.C H C- 7 C02H I 02c Me OH Warnhoff and Wildmanlol have clarified the structure of undulatine (XCV).A useful method of epoxide detection consisted of successive reduc- tion with lithium aluminium deuteride oxidation and base equilibration 87 Bradbury and Masamune J. Amer. Chem. Soc. 1959 81 5201. Bradbury and Willis Austral. J. Chem. 1956 9 258. 99 Bradbury Tetrahedron 1958 2 363. loo Geissman Aristral. J. Chem. 1959 12 247. lol Warnhoff and Wildman Chem. and Ind. 1958 1293. 334 QUARTERLY REVIEWS of the resultant monodeuterated ketone (XCVJ) to obtain the deuterium- free epimer (XCVII).lol Reduction leads to hydroxyl and deuterium sub- stituents at the carbon atoms originally attached to the cyclic ethereal oxygen atom and since loss of deuterium during base equilibration can occur only if the deuterium is on carbon 01 with respect to the carbonyl (XCVI) the cyclic ether must be an epoxide.(XCVI ; R=’H,OM~ axial) ’ ‘ “‘.L (XCVlI ; R= H,OMe equatorial) ( X c v ) (XCVI 1 r) Conflicting opinion1o2Jo3 on the structure of annotinine an alkaloidal constituent of the club moss Lycupodium annotinurn has been resolved by X-ray crystallography which established the structure of annotinine bromo- hydrin (XCVIII; R‘ = Br R = H2),lo4 and hence vindicated the structure for annotinine (XCIX; R = H,) developed by Weisner Valenta and their collaborators.lo2 Maclean and Prime had earlier demonstrated the presence of an epoxide by conversion of annotinine lactam (XCIX; R = 0) into a chlorohydrin (XCVIII; R = 0 R’ = Cl).lo5 Whereas acid-catalysed hydration of annotinine leads to a 1,2-glyc01,~~~ alkali yields first an epimeric carboxylic acid which under more vigorous conditions gives the diol (C; R = 0).lo2 Formation of the latter is made possible by 01- orientation of the carboxylate anion in the epirnerised acid.lo2 A similar argument has been advanced for the formation of diphenylannotinine (C; R = Ph2) from the alkaloid and phenyl-lithium.lo6 The tropane alkaloid scopolamine (C1 ; R = -CO-CHPh-CH,.OH) on very mild hydrolysis furnishes scopine (CI; R = H).More vigorous treat- lo2 Weisner Valenta et al. Chem. arid Znd. 1957 564; .?. Amer. Chem. Sac. 1956,78 2867; Tetrahedron 1958 4 87. Io3 Martin-Smith Greenhalgh and Marion Canad. J. Chem. 1957 35 409. lo* Przybylska and Marion Canad. J. Chem. 1957,35 1075. Io6 MacLean and Prime Canad. J . Chem. 1953,31 543. lo6 Perry MacLean and Manske Canad. J . Ckem. 1958 36 1146. CROSS NATURALLY OCCURRING 1,2-EPOXIDES 335 ment with base or acid leads to scopoline (CII) by the now familiar rear- ward displacement of epoxide oxygen. As expected pseudoscopine (3- epimer of CI) does not undergo this transformation. Fodor and his colleagues1o7 have recently achieved a total synthesis of scopolamine. Fodor et al. J. 1959 3461 and references summarised therein.

ISSN:0009-2681    

DOI:10.1039/QR9601400317

出版商:RSC    

年代:1960

数据来源: RSC

摘要:

THE INTERACTIQN OF FREE RADICALS WITH SATURATED ALIPHATIC COMPOUNDS By J. M. TEDDER (DEPARTMENT OF CHEMISTRY UNIVERSITY OF SHEFFIELD) THE chief characteristic of aliphatic chemistry is the inertness of the hydro- carbon chains. The so-called functional groups in an aliphatic compound can undergo a vast series of transformations while the hydrocarbon chain remains unaltered. There is however an important exception to this general rule and this is the behaviour of aliphatic compounds when attacked by free radicals. The position is then reversed and the “functional group” is often unaltered while the hydrocarbon chain is attacked. The most common example is the halogenation particularly the chlorination of aliphatic compounds. This has received extensive study for technical reasons1 but only recently has it attracted the serious attention of the organic chemist.2 At the other end of the scale is the interaction of methyl radicals with aliphatic compound~.~ This subject has received much attention from the kineticists but little serious attempt has been made to correlate these results either with types of organic structure or with the data from halogenation.New techniques of which gas-phase chromatography and infrared and nuclear magnetic resonance spectroscopy are the most important have in the last few years made it more profitable to study the complex mixtures which are so often obtained from the interaction of free radicals with aliphatic compounds. The present situation is that the preliminary surveys have been made and the main features of these reactions recognised.It is the purpose of this Review to summarise these features. The next quin- quennium is likely to see a more detailed study of free-radical reactions and it is hoped that this article may help those chemists who are interested spectators rather than participants to follow the progress of the investiga- tions. The formation of free radicals will not be discussed as this has been adequately covered elsewhere.* The Reactions of Free Radicals.-The reactions of free radicals can be divided into six principal types (a) addition (b) radical transfer (c) dissociation (d) radical recombination (e) disproportionation and (f) insertion (e.g. of CH,). The addition reactions of free radicals have received the greatest attention. They include the simple radical additions to olefinic double Hass McBee and Weber Znd.Eng. Chern. 1935,27 1190; 1936 28 333. Ash and Brown Rec. Chern. Progr.,.1948 9 81. Trotman-Dickenson Quart. Rev. 1953 7 198. (a) Waters “The Chemistry of Free Radicals,” Oxford Univ. Press London 1948; (b) Walling “Free Radicals in Solution,” John Wiley & Sons New York 1957; (c) Trotman-Dickenson “Free Radicals,” Methuen London 1959. 336 TEDDER FREE RADICALS WITH SATURATED ALIPHATIC COMPOUNDS 337 bonds,6 radical polymerisation,6 and the so-called homolytic substitution of aromatic compounds in which the first step is without doubt one of addition.7 None of these processes is to be discussed here. The present Review is concerned with (b) the transfer reactions of free radicals though inevitably (c) radical dissociation and ( d ) radical re- combination will have to receive some consideration.In radical-transfer reactions the free radical abstracts an atom from another molecule thereby creating a new free radical A:B + X- -+ A* + X:B In saturated aliphatic compounds the atom abstracted is nearly always a hydrogen atom R-H + X. -+ R* + H-X The fate of the alkyl radical so produced depends on the reaction system and the nature of the radical. It may dissociate into an olefin and another radical (with ordinary alkyl radicals this process usually only occurs extensively at fairly elevated temperatures) CSH,* -+ C,H + CH,. it may combine with another radical R* + R* (X*) -+ R-R (R-X) or it may in its turn abstract an atom from another molecule. The last pro- cess is particularly common in aliphatic chemistry and leads to a chain reaction of which chlorination is the most studied CI -+ 2 CI* RH + CI* -+ R* + HCI R* + CI + RCI + CIS To the organic chemist the three most important questions are Which hydrogen atom will be most readily extracted? What effect will substituents in the aliphatic compound have? And finally will different radicals abstract different hydrogen atoms ? Before considering these questions we shall very briefly consider the variation in reactivity of free radicals.The Reactivity of Free Radicals.-The reaction we are principally concerned with is hydrogen abstraction. Not all radicals are sufficiently reactive to remove hydrogen atoms from aliphatic compounds. Thus triphenylmethyl radicals and iodine atoms show no tendency to attack hydrocarbon solvents at normal temperatures. If the abstraction process were controlled entirely by thermodynamic considerations we should expect the bond formed between the free radical and the abstracted hydrogen atom to be a measure of the radical’s reactivity.Table 1 lists Cadogan and Hey Quart. Rev. 1954 8 308. Mayo and Walling Chem. Rev. 1950,46 191. Augood and Williams Chem. Rev. 1957,57 123. 338 QUARTERLY REVIEWS the bond strengths of some radical-hydrogen bonds together with the activation energies for the abstraction of a hydrogen atom from ethane by the same radicals. It will be seen that although the bond dissociation energy of H-X falls as the activation energy rises no direct relation is observed. Thus although the strength of the bond formed is very important there must be other factors which have considerable control over these processes.TABLE 1. The strength of the radical-hydrogen bond formed and the activation energies for the abstraction reaction with ethane X- +- C2H6 +XH + C,H,*. X D(H-X) E (kcal./mole) Ref. F. 136 e l . 103 CF3* 103 H- 103 Br- 87 CH3* 101 0.2 1.0 7.5 8.7 11.2 13.3 8 9 10 11 3 12 Hydrogen Abstraction from Aliphatic Compounds.-(a) The experimental results. It has long been known that the ease of abstraction of hydrogen atoms from aliphatic hydrocarbons follows the order tertiary> secondary >primary. Thus the more branched an alkane the more susceptible it is to free-radical attack. In his pioneering work on chlorination Hass introduced a very useful term “relative selection” (RS) which is defined as the ratio of the rates of attack at the different types of hydrogen atom in a molecule multiplied by the inverse ratio of the number of hydrogen atoms of each type.Thus in n-butane the “relative selection” of secondary to primary hydrogen atoms (written RS,”) is given by Assuming steady state conditions we can equate the ratio of the two butyl halides formed with the rates of attack at the primary and secondary hydrogen atoms. Hass performed his chlorinations over a range of temperatures and subsequently his results were used to calculate the Arr- henius parameters. This method which can give ratios of A factors and differences of activation energies has been much used for halogenation recently. For other free radicals however although precise kinetic data were often obtained they only referred to the rate of attack on the whole * Fettis Knox and Trotman-Dickenson J.1960 1064. Knox and Nelson Trans. Furuday Soc. 1959,55 937. lo Ayscough and Steacie Canad. J. Chem. 1956 34 103. l1 Darwent and Roberts Discuss. Furuduy SOC. 1953 14 55. la Andersen and Van Artsdalen J. Chem. Phys. 1944 12 478. TEDDER FREE RADICALS WITH SATURATED ALIPHATIC COMPOUNDS 339 molecule and did not distinguish the rates at the different types of hydrogen atom. This is particularly striking in the case of methyl radicals where in spite of a large amount of accurate data the rates of attack at different types of hydrogen atom could only be obtained by approximate inter- polation. During the last three years this position has been partially rectified and quantitative estimations of the relative rates of attack by various radicals on the primary secondary and tertiary hydrogeil atoms in butane and isobutane have been reported.These results are tabulated below; the relative selectivities are all referred to the primary hydrogen atoms. Table 2 gives a very concordant picture. The selectivity of a particular radical varies with its reactivity in a fairly regular manner. It appears from these results that for chlorination and bromination the ratio (Ep - Et)/(Ep - Es) is larger than for methyl radicals i.e. the halogen atoms are relatively more affected by increased branching. However there are really too few results to be certain about this. TABLE 2. X. RS at 300"~ EP-ES EP-Et EP Ref. The selectivity of direrent radicals X* for primary secondary and tertiary atoms:I3 X* + RH -+- R. + HX. CHS- CH; CHr (cal./mole) F. 1 1.2 1-4 90 - 220 13 8 c1.1 3.9 5- 1 480 650 lo00 13 9 CD,. I 35 - 2100 - 11,OOO 14 CH3* 1 - - 2300 2900 11,400 15 16 Br- 1 82 1600 (3170) (5900) 14,000 13 17(18) (There is some discrepancy between the three sets of bromination results.) Until recently there have been no data of the absolute rate of radical attack on a specific hydrogen atom and it has been common practice to treat all primary or all secondary hydrogen atoms as equivalent. There is ample evidence that this is likely to be a poor approximation and for chlorination it is no longer necessary. Using a competitive technique Knox and Nelson have been able to estimate the absolute rates of chlorina- tion of the primary hydrogen atoms in some of the simple alkanes. There are two generalisations which can be made about these results (Table 3).First the activation energies in general fall with increasing molecular weight of the alkane (a similar trend was noted with secondary hydrogen atoms) and secondly branching adjacent to the methyl group l3 Anson Fredericks and Tedder J. 1959 918. l4 McNesby and Gordon J. Arner. Chenz. SOC. 1956 78 3570. l5 Rice and Vanderslice J . Arner. Chem. SOC. 1958 80 292. l6 Trotman-Dickenson Bichard and Steacie J. Chem. Phys. 1951 19 161 l7 Eckstein Scheraga and Van Artsdalen J. Chern. Phys. 1954 22 28. l8 Fettis and Trotman-Dickenson J. Arner. Chern. SOC. 1959 81 5260. 340 QUARTERLY REVIEWS containing the primary hydrogen atoms causes a slight rise in the activation energy. TABLE 3. Absolute rate factors for the chlorination of primary hydrogen^:^ Cl. + RH + R. + HC1.atom atom at 2 9 3 " ~ RH 10-13A per H E (cal./mole) 10-13k per H (mole-l cm. sec.-l) (mole-l ~ m . ~ sec.-l) C2H6 1-5 & 0.2 1040 & 75 0-25 C3H8 1.7 5 0.2 lo00 75 0-3 1 n-C4H1 0 1.4 rfr 0.2 790 & 80 0.37 iso-C,H, 1.3 5 0-2 820 Ifr 80 0-33 neo-C,H, 1.4 & 0.2 920 & 80 0.30 So far only unsubstituted alkanes have been considered; the effect of substituents can be very pronounced but the situation is complicated by the fact that different radicals are influenced in different ways by any particular substituent. This is clearly demonstrated by the different rates of attack on methane ethane and the halogenated methanes by methyl radicals and chlorine atoms. Table 4 shows that the introduction of substituent chlorine atoms into methane causes an almost uniform decrease TABLE 4.Hydrogen abstraction from methane ethane and the chlorinated methanes by methyl radicals and by chlorine a t o m ~ . ~ * ~ J ~ CH c1* log A E log A E CH4 11-6 12-8 10.4 3.9 C2H6 11.3 10.4 11.1 1.0 CH3C1 11.8 9.4 10-7 3 4 CH 2C12 11.3 7.2 11.6 5.5 CHCl 10-8 5.8 11.6 6.5 in the activation energy to attack b y methyl radicals while it increases the activation energy for chlorination once the first chlorine atom has been introduced. Substituents quite often affect hydrogen abstraction b y alkyl radicals and halogen atoms in different ways. Thus an alkyl radical will attack a carboxylic acid (or acid chloride) in the a-position adjacent TABLE 5. Hydrogen abstraction from propionic acid relative selections.2Q*21 Radical CH3- CH2- C02H (mole-l 1. sec.-l) (kcal. mole-l) (mole-l 1.sec.-l) (kcal. mole-l) CH3* 1 7.8 Cl- 1 0-03 l9 Goldfinger Jeunehomme and Martens J. Chem. Phys. 1958 29 456. 2o Price and Morita J . Amer. Chem. Soc. 1953 75 3686. 21 Magritte and Bruylants Znd clzim. belge 1957 22 547. TEDDER FREE RADICALS WITH SATURATED ALIPHATIC COMPOUNDS 341 to the carboxyl group,20*22 whereas attack by a halogen atom ( c h l ~ r i n e ~ ” ~ ~ or fluorinez4) occurs preferentially at sites remote from the carboxyl group (see Table 5). Substituent halogen atoms and carbonyl groups also facilitate attack by methyl radicals but impede attack by chlorine atoms. The property of these substituents which impedes chlorination at adjacent carbon atoms has been shown to be largely polar. K ~ o y m a n ~ ~ and Wallingz6 and their collaborators have shown that the relative rates of chlorination of substituted toluenes obey the Hammett op relation.Walling and Miller’s results are shown in Fig. 1. Van Helden and Kooyman -0.2 0-0 0.2 0 . 4 0.6 Hammctt CY FIG. 1. Hammett ap plot for chlorination of substituted toluenes. (Reproduced 4181.) by permission from Walling and Miller J. Amer. Chem. Soc.. 1957. 79. also found that the rate of chlorination of substituted t-butyl compounds gave a linear relationship with the acidities of the corresponding acids (see Table 6). Other instances where the deactivating influence of the substituent must be largely polar have been observed in the chlorination of l,l,l-trifluoro- 23 Bruylants et al. Bull. SOC. chim. belges 1949 58 210; 1952 61 266; Den Hertog 24 Bockemuller Annalen 1953 506 20. 25 Van Helden and Kooyman Rec.Trav. chim. 1954 73 269. 26 Walling and Miller J. Amer. Chem. SOC. 1957 79 4181. Kharasch and Gladstone J. Amer. Chem. Sac. 1943 65 15 de Vries and van Bragt Rec. Trav. chim. 1955 74 1561. 342 QUARTERLY REVIEWS propane,27 l,l,l-trifluorobutane,28 and 1,1 l-trichlor~butane.~~ All these results show that a polar group on an adjacent site can have a marked effect on chlorination. However when the substituent is on the same carbon atom as the hydrogen atom being abstracted the polar effect is not always the predominating influence. Table 4 showed that methyl chloride TABLE 6 . The relative rates of chlorination of t-butyl-X.25 CH3 CH3 I I I I CH3 CH3 CH3-C-X + C1. -+ .CH,-C-X + HCI x Relative rate -log K (CH3)3C 1.2 & 0.1 0.7 p-NO 2 C g H 4 0.5 -+ 0.1 -0.4 C,H5C0 0-4 -+ 0.1 -0.8 Cl-co 0.19 rl 0.03 -1.5 CN 0.17 & 0.03 - 1.7 Relative rate = (Rate t-butyl-)()/(Rate t-butylbenzene) log Ka == log ( K of X.CH,*COzH) - log ( K of CGH5CO2H) is more readily chlorinated than methane and Table 7 compares the rates of chlorination at the various positions in butane and 1- and 2-chloro- butane.The most easily chlorinated site in 2-chlorobutane is the 2-posi- tion although it is less reactive than the methylene groups in unsubstituted butane. Similarly although the least easily chlorinated site in l-chloro- butane is the 1-position it is far less deactivated than would be expected on polar TABLE 7 . The chlorination of butane and the butyl chlorides at 78°:31*32 relative select ions. CH3 CH2 CH2 CH3 1 3.6 3.6 1 CH2C1 CH2 CH,---CH3 0.8 2.1 3.7 1 0.2 3.0 2.9 0.8 CH3 CHCl CH2 CH3 As would be expected attack by the other halogen atoms fluorine and bromine is also influenced by polar substituents but the extent of this Heme and Whaley J.Amer. Chem. SOC. 1942,64 1157. 28 Heme and Hinkamp J. Amer. Chem. SOC. 1945,67 1197. Brown and Ash J. Amer. Chem. SOC. 1955,77,4019. 30 Fredricks and Tedder Chem. and Znd. 1959,490. 31 Fredricks and Tedder J. 1960 144. 32 Fredricks and Tedder unpublished work. TEDDER FREE RADICALS WITH SATURATED ALIPHATIC COMPOUNDS 343 influence is very different. Table 8 compares the selectivities of the three halogen atoms when reacting with 1 -fluorobutane. Fluorination which is TABLE 8. The halogenation of 1-jluorobutane relative ~elections.~~ CH 2F CH2 CH2 CH3 Fluorination at 20" < 0.3 0.8 1.0 1 Chlorination at 78" 0.9 1-7 3.7 1 Bromination at 146" 10 9 88 1 by far the least selective halogenation process (see Table 2) is clearly retarded by the substituent polar group and this retardation slightly apparent at carbon atom 3 is most pronounced at the carbon atom carrying the substituent.Bromination the most selective halogenation process is not detectably affected at carbon atom 3 is strongly retarded at carbon atom 2 but proceeds ten times more rapidly at the substituted carbon atom 1 than at the other terminal carbon atom 4. Polar influences in bromination have also been reported by Kooyman et u Z . ~ ~ who found that the bromination of substituted toluenes obeyed the Hammett equation Halogenation is the most studied free-radical process in which polar effects have been observed.However polar effects have also been found to be important in hydrogen abstraction by a variety of other free radicals. Qualitative evidence has been obtained for polar effects with ethoxy- radicals and the attack of arylperoxy-radicals on substituted cumenes has been to obey the Hammett relation (p = - 0.43). The Hammett equation has also been found to correlate the rates of hydrogen abstrac- tion from substituted thiophenols by 1 -cyano- 1 -cyclohexyl radicals (p = - 0-4),35 from substituted toluenes by trichloromethyl radicals (p = - 1.46),36 and from substituted benzaldehydes by p-chlorobenzoyl- peroxy-radicals (p = - 0~8).~' It should also be added that polar effects have been found to be important in other free-radical processes besides hydrogen abstraction. The results cited above represent only a small fraction of the work that has been done on the interaction between free radicals and aliphatic compounds.However the inclusion of more results in which other free radicals or other substituents were employed would not reveal any new principles. The cardinal observations can be summarised. Rydrogen abstraction by free radicals from unsubstituted alkanes invariably follows the order tertiary> secondary > primary though the selectivity of different radicals varies very considerably. The introduction of a substituent into 3s Kooyman Van Helden and Bickel Proc. k. ned. Akad. Wetenschap. 1953 B 56 75. 84 Russell J. Amer. Chem. SOC. 1956 78 1047. s6 Schaafsma Bickel and Kooyman Rec. Truv. chim. 1957 76 180. s6 Huyser J. Amer. Chem. SOC. 1960 82 394.(p = - 1.05). Walling and McElhill J. Amer. Chem. Soc. 1951 73 2927. 344 QUARTERLY REVIEWS the alkane can have a very pronounced effect which depends on the attacking radical and may accelerate the attack of one radical while retard- ing another. Even such similar species as the three halogen atoms fluorine chlorine and bromine are affected differently by any particular substituent. Finally one of the influences affecting hydrogen abstraction by the halogen atoms and certain other radicals has been shown to be of a polar nature. In electrophilic aromatic substitution the most reactive site in a molecule toward one reagent will also be the most reactive to all the others although the extent to which different reagents discriminate will vary. The same applies to nucleophilic substitution in aliphatic compounds provided there is no change in mechanism.However the attack on aliphatic compounds by free radicals shows no such regularity and the most reactive site on a molecule towards one radical may be the least reactive towards another. Hydrogen Abstraction from Aliphatic Compounds.-(b) The factors controlling the reaction. In the previous section the rates of abstraction of hydrogen atoms from different molecules and different sites in the same molecule have been compared. Tables 2 and 4 show that the differences in rate for these particular cases are primarily due to differences in the activa- tion energy of the process and that the pre-exponential factors are of secondary importance. The chlorination of the butyl halides (cf. Tables 7 and 8) has also been carried out over a range of temperatures and again the results indicate that the different rates of attack at the various carbon atoms are principally due to differences in activation energy.It seems reasonable therefore to assume that the differences in rates discussed previously are in general due to differences in activation energy. A large number of these reactions are carried out in the gas phase and the rate- determining process is a particularly simple one involving an atom or radical X reacting with a molecule RH to yield a new molecule XH and a new radical R X * + H - R - + X * * * H * - * R + X H + R* The factors controlling such a process are very conveniently discussed in terms of potential-energy surfaces in a manner originally developed by Evans and P~lanyi.~* Such a surface is pictured in Fig.2. The section MN (shown in Fig. 3) is the curve for the molecule R-H when X is far away. As X approaches H-R there will be a series of similar curves produced by cutting sections parallel to MN. At the transition state 0 the section is POQ and if there were no reaction it would continue along the dotted line PO’P’ (Fig. 3). The section RS represents the repulsion between the new molecule X-H and the new radical R when the distance XH is kept con- stant at its equilibrium value. The curve Q Q’ represents this same repulsion when X-H has been partly stretched and the difference in potential between S and Q represents the energy required to stretch X-H from its 38 Evans and Polanyi Trans. Faradczy Soc. 1938 34 11. TEDDER I U c 0 X C W Y 3 .c.W 9 pr 0 c 0 .c. cn .- n Z RADICALS WITH SATURATED ALIPHATIC COMPOUNDS 345 Distance between R and H FIG. 2. Potential-energy surfaces for the reaction X + HR + XH + R. \ \ \ \ QI'\ x 0 C W 0 c W 0 P - .- ). Y a \ \ \ \ \ \ \ \ \ P' 4- -- - 4 0 / / - / / / Distance between R and H F'IG. 3. Potential-energyprofiles for the reaction X + HR += XH + R. 346 QUARTERLY REVIEWS equilibrium distance to that in the transition state. The broken parts of the lines PP’ and QQ‘ intersect at 0’ but the curve POQ is rounded off owing to resonance between the states X* + RH and R- + XH. The activation energy E is represented by the vertical distance between the bottom of the curve MN and the top of the POQ curve at 0. Consideration of Fig. 3 shows that the activation energy E is controlled by four factors An increase in the strength of this bond would correspond to a lowering of the point S which in turn would lower curve QQ’ so that its intersection with PP’ would be lower i.e.the activation energy would be less. (2) The strength of the bond H-R which is broken. An increase in the strength of this bond would correspond to the raising of the point P’ so that the right-hand side of the curve PP’ would be steeper and its intersection with QQ‘ would also be raised i.e. the activation energy would be greater. (3) The repulsion between the new radical R and the new molecule XH. An increase in this repulsion corresponds to a steepening of the curve QQ‘ so that its intersection with PP’ will be higher i.e. the activation energy is increased. (4) The repulsion between the incoming radical X and the molecule RH.An increase in this repulsion would raise the height of the curve PP‘ above MN and thus its intersection with QQ’ would also be higher i.e. the activation energy would be greater. The first factor the strength of the bond formed is exactly what was considered on pp. 337 and 338 when the reactivity of free radicals was being discussed. Table 1 showed that the new bond formed between the free radical and the abstracted hydrogen atom was a major factor in determin- ing the activation energy of the process. Thus the reaction between a fluorine atom and ethane yields the very strongly bound hydrogen fluoride D(H-F) = 136 kcal. mole-l and the activation energy is very small ( E = 2>0 cal. mole-’); the reaction of bromine atoms with the same alkane yields the relatively weakly bound hydrogen bromide D(H-Br) = 87 kcal.mole-l and the activation energy is very much greater ( E = 13,300 cal. mole-l). The importance of the second factor the strength of the bond broken is illustrated by the relative ease of abstraction of primary secondary and tertiary hydrogen atoms. Indeed attention has been drawn to the fact that there is a linear relation between the activation energy and the strength of the C-H bond broken in the reaction of methyl radicals with the alkane^.^ The existence of such relations which in a more general sense imply a relation between the activation energy and the heat of reaction was predicted by Evans and PolanyL3* This relation takes the form E = aH (O<a<l) and can be expected to hold when X is kept constant and R is varied (e.g.X = CH,; R = Alkyl) or when X is varied and R is (1) The strength of the bond X-H which is formed. TEDDER FREE RADICALS WITH SATURATED ALIPHATIC COMPOUNDS 347 kept constant (e.g. X = Na K Cs extracting C1 from CH,Cl). However Fig. 3 makes it quite clear that no relation can be expected when the repulsive forces (factors 3 and 4) vary from one compound to another (cf. Table 4 the chlorination of the chloromethanes). In order to predict the points of attack by a free radical in a particular molecule it is necessary 10 - - 0 x W lu 5 - € = 0 - 5 D ( C - H ) - 3 8 - 5 k cat. 95 100 D (C-H)(kcal.) RG. 4. Activation energy plotted against bond-dissociation energy for abstraction of hydrogen by methyl radicals. to have some qualitative way of estimating the relative strengths of the different carbon-hydrogen bonds.One method of attempting this is to consider the possible resonance stabilisation of the incipient alkyl radical. Thus the experimentally determined order of carbon-hydrogen bond strengths primary> secondary> tertiary can be accounted for in terms of the resonance stabilisation of the incipient radical by hyperconjugation CH? /,/ CH CH3,[CH2 H- C. i--t / CH / CH This argument would suggest that the reactivity of primary hydrogen atoms themselves would vary from molecule to molecule depending on the extent of substitution at the adjacent carbon atom. Such a variation has been observed in the rates of chlorination of the primary hydrogen atoms in 348 QUARTERLY REVIEWS n-butane isobutane and neopentane which decrease in that order (see Table 3).The effects of substituents now being considered it becomes clear that any group containing .rr-electrons or any atoms with non-bonded p-electrons will favour hydrogen abstraction from the carbon atom to which it is attached by resonance stabilisation of the incipient radical. These predictions are well borne out by the behaviour of methyl radicals + - CI-CH,* ++ *CI-CHS CH,-CH*C-OEt t+ CH,*CH*C-OEt I +O* II 0 (Tables 4 and 5). However it must be remembered that there are other factors which affect bond strength and this simple picture is likely to break down in some cases. We come now to the third and the fourth factor controlling the activation energy of the hydrogen-abstraction process. These are the repulsive forces between the reaction products HX and R- (factor 3) and the repulsion between the reactants X.and RH (factor 4). Before these can be considered in detail it is necessary to review the nature of these repulsions. The major part of these forces are non-classical and in simple cases can be estimated by using London’s theory. However the organic chemist is not concerned with the actual magnitude of these forces but with variations due to sub- stituents in the reacting molecule. The electronic theory of organic chem- istry describes the effect of substituents in electrostatic terms so we are concerned with extra Coulombic repulsions due to electronic displacements in the reacting molecule. The nature and relative magnitude of these electronic displacements can readily be predicted by elementary organic theory. Let us now consider in detail the repulsion between the new molecule H-X and the new radical R.In hydrogen abstractions by methyl radicals the new molecule is methane which is essentially non-polar so there will be no extra Coulombic forces and it would be expected that methyl radicals would preferentially attack the weakest carbon-hydrogen bond and be almost indifferent to any polar properties of the molecule. This is in good accord with observation-(cf. Tables 4 and 5). However it will be shown later that methyl radicals may be slightly susceptible to polar effects between the reactants (factor 4). Other radicals in which the HX molecule formed is non-polar are other alkyl radicals and hydrogen atoms. With atomic halogenation on the other hand the new molecule (HX) will be highly polar and we may expect polar substituents in the aliphatic com- pound to exert a considerable effect.The chlorination of the substituted toluenes (Fig. l) and of the substituted t-butyl compounds (Table 6) are good examples of this. The changes of reactivity within the two series can be attributed almost entirely to changes in electron density at the reaction TEDDER FREE RADICALS WITH SATURATED ALIPHATIC COMPOUNDS 349 site. In Table 4 it is shown that the introduction of a substituent chlorine atom into methane reduces the activation energy for the abstraction of a hydrogen atom by a chlorine atom. A similar effect is observed with methyl radicals and we have already attributed this to a weakening of the C-H bond. However the drop in activation energy is proportionally much less for chlorination and for the chlorination of methylene dichloride and chloroform the activation energy rises again.This type of effect is very familiar to the organic chemist and occurs when the inductive effect ( - I ) is in opposition to the mesomeric effect (+M). The two opposing effects are even more clearly illustrated by the halogenation of the butyl halides (Tables 7 and 8). Table 7 shows that a substituent chlorine atom retards chlorination at the adjacent carbon atom owing to the polar effect but at the carbon atom to which it is attached the deactivating influence of the polar effect is almost completely counteracted by the resonance stabilisa- tion of the incipient radical. The hydrogen halides become less polar and TABLE 9. Dipole moments and dissociation energies of the hydrogen halides.X F c1 Br PHX (D) D(H-X) (kcal. mole-l) 1.19 136 1 -08 103 0.79 87 less strongly bound as the molecular weight increases (Table 9). We should expect therefore an increase in selectivity together with a decrease in the relative importance of the polar effect in changing from fluorination to chlorination or from chlorination to bromination. How these opposing effects combine is shown very clearly in the halogenation of n-butyl fluoride (Table S) and the at first sight puzzling changes in the reactivity of carbon atom 1 are readily explained. In all the examples cited so far the polar effect has been such that the activation energy has been increased. However an electron-repelling substituent (+ l ) would be expected to lower the activation energy for halogenations by reducing the repulsion.Evidence for this type of effect is to be found in Table 1 where the activa- tion energy for hydrogen abstraction from ethane is very much less for chlorine atoms than for hydrogen atoms although the heats of reaction are the same. Similarly it was pointed out on p. 339 that in Table 2 chlorination and bromination are more affected by branching in the alkane than is attack by methyl radicals. Factor 3 will be important whenever the bond formed between the radical and the abstracted hydrogen atom is polar so that besides the halogen atoms hydroxy- alkoxy- cyano- and peroxy- radicals will all be affected in a similar way. Finally we must consider the fourth factor i.e. the repulsion between the attacking radical (X) and the initial molecule (RH). For there to be additional Coulombic forces the attacking radical itself must have polar 2 350 QUARTERLY REVIEWS properties.Factor 4 will therefore be unimportant in reactions involving atoms. At present the results which can be definitely attributed to this factor are not numerous. The rate of attack by trichloromethyl radicals on substituted toluenes has been found by H ~ y s e r ~ ~ to follow the Hammett ap relationship. These results are in direct contradiction to an earlier report by Kooyman et aZ.33 who found trichloromethyl radicals to be unaffected by polar substituents. However Huyser’s work is more extensive and appears reliable. If it is accepted that polar effects are present then these must be largely due to polar repulsion between the reaction sites of the trichloromethyl radical and the toluene molecule since the carbon- hydrogen bond in chloroform is considered to be almost non-polar and repulsions between the reaction products will be unimportant.Table 1 shows that the activation energy of the abstraction of hydrogen atoms from ethane by trifluoromethyl radicals hydrogen atoms and methyl radicals are respectively 7.5 8.7 and 11.2 kcal. mole-l although the heats of reaction are almost the same. In reactions with hydrogen atoms no polar effects are possible while any Coulombic forces between the reactants would be attractive with trifluoromethyl radicals and repulsive with methyl radicals. Factor 4 therefore appears to be the best explanation of these observations. The present section can now be summarised. The variations in the rate of hydrogen abstraction from different sites in a molecule by an atom or radical have been attributed to changes in the activation energy of the process.By using the usual transition-state picture of the process varia- tions in the activation energy have been attributed to four factors (i) the strength of the H-X bond formed (ii) the strength of the C-H bond broken (iii) polar forces between the products of the reaction the radical Re and the molecule H-X and (iv) polar forces between the reactants i.e. the radical X* and the molecule R-H. Combinations of these factors have been adequate to explain the wide variety of experimental results. Steric Effects in Free-radical Reactions.-When a hydrogen atom is abstracted from an asymmetric carbon atom the asymmetry is lost and the chlorination of (+)- 1 -chloro-2-methylbutane yielded inactive 1,2- dichlor0-2-methylbutane.~~ The tervalent carbon therefore either assumes a planar sp2-configuration or it is pyramidal but there is rapid interconver- sion of the two pyramidal forms (cf.the NH3 molecule). Although the tervalent carbon atom does not retain the configuration of the molecule from which it is derived it can in certain cases be stereospecific in its reac- tions. The halogenation of the 2-halogenobutanes yielded both the expected 2,3-dihalogenobutanes threo and erythro ; however the amounts of these two isomers were not equal (see Table 10).32940 It seems probable that the two 2-halogeno- 1 -methylpropyl radicals (CH,CHCI.CH.CH,) have 30 Brown Kharasch and Chao J . Amer. Chem. SOC. 1940 62 3435. 40 Fredricks and Tedder Proc.Chem. SOC. 1959 9. TEDDER FREE RADICALS WITH SATURATED ALIPHATIC COMPOUNDS 351 preferred conformations both of which favour substitution from one direction. If the bonds about the tervalent carbon atom are nearly planar TABLE 10. Proportions of the isomeric 2,3-dihalogenobutanes formed during the halogenation of the 2-halogenob~tanes:~~*~~ CH,-CHXCH2.CH3 Y - CH3*CHX*CHYCH3 X c1 F Cl F Y erythro (yo) threo (yo) c1 71-4 28.6 c1 59.4 40.6 F 66-6 33.4 F 60.2 39.8 then the other three bonds attached to the tetrahedral carbon atom can rotate relative to this plane (see Fig. 5). It has been suggested that they will have preferred conformations in which the methyl group of the tervalent carbon atom will lie close above the hydrogen atom of the FIG. 5. tetrahedral carbon.The incoming molecule must approach approximately perpendicularly to the free-radical plane and this approach will be easier on the side away from the substituent halogen so that in both cases it will favour the erythro-compound.40 In agreement with this explanation the stereospecificity is greater when chlorine atoms than when fluorine atoms are the substituents. A similar stereospecificity has been reported by Kooyman and Vegter41 from the studies of the halogenation of bicyclo- (2,2,l)heptane (norbornane). This halogenation gives mainly the two 2-halides. Halogenation with molecular chlorine or bromine yields about 70% of the em-isomer and 20-25y0 of the endo-halide. Halogenation with sulphuryl chloride or carbon tetrachloride initiated by benzoyl peroxide yielded 95% of the exo-chloride.In this molecule the conformation is fixed and approach to the tervalent carbon atom is shielded on the endo-side (cf. Fig. 6). As would be expected this shielding becomes increasingly Kooyman and Vegter Tetrahedron 1958 4 382. 352 QUARTERLY REVIEWS important as the size of the molecule with which the norbornyl radical reacts increases. The reactions of free radicals will of course be affected by steric inhibi- tion of resonance. The importance of resonance-stabilisation of the incipient radical was emphasised in the previous section. Kooyman and Strang42 found that trichloromethyl radicals attacked m- or p-cymene about five times faster than o-cymene; similarly the attack on hexaethyl- benzene was only about half as fast as that on ethylbenzene although ando -Approach FIG.6. (Reproduced by permission from Kooyman and Vegter Tetrahedron 1958 4 322.) hexamethylbenzene was attacked eight times faster than toluene. Reso- nance-stabilisation of the incipient benzyl radical can occur only if the radical is planar and this is not possible with either o-cymene or hexaethyl- benzene. Solvent Effects.-It is often considered that solvents have little effect on free-radical reactions. While it is true that there are reactions which have been shown to have identical or very similar kinetics in both the gas and the solvent phase,43 examples have long been known in which a change from the gas phase to the liquid phase has a pronounced effect. In his pioneering studies on chlorination Hass reported that the selectivity of the reaction was much less in the liquid phase than in the gas phase at the same temperature.l Table 11 presents Hass’s results as recalculated by Walling.44 This work is not quite in line with more recent results but both gas-13 and liquid-p h a ~ e ~ ~ results have been repeated separately within the last three years and the difference between the two has been confirmed.It is to be hoped that a careful study of the chlorination in the gas and the liquid phase will soon be attempted by the same operator using the more refined modern techniques so that a quantitative measure of this effect 43 Kooyman and Strang Rec. Trav. chim. 1953,72 329. 43 Bell Ann. Reports 1939 36 84. 44 Ref. 4(b) p. 358. 45 Walling and Mayahi J. Amer. Chem. SOC. 1959 81 1485. TEDDER FREE RADICALS WITH SATURATED ALIPHATIC COMPOUNDS 353 can be obtained.The effect has been attributed to a "solvent cage." The rate of the reaction in the gas phase is of the order 1013 mole-l-~m.~ sec.-l at room temperature which is comparable with the rate of diffusion of chlorine atoms in the liquid phase. Thus a chlorine atom caged in by the TABLE 11. Relative selections for chlorination in the gas and the liquid phase. Temp. -CH3 >CH SCH -CH3 >CH X H - 1 3.7 5.4 200" - - 100 1 2.0 3.0 1 4.3 7.0 50 1 2.9 4.5 1 4.8 - 0 1 4.5 7.0 - - - Liquid Gas solvent will be held at a particular site long enough for there to be a high probability of reaction even though there are more reactive sites elsewhere. reported an even more profound solvent effect. He found that certain solvents particularly aromatic compounds greatly increased the selectivity of the liquid-phase chlorination.Table 12 presents In 1957 TABLE 12. Solvent effects in the chlorination of n - b ~ t a n e . ~ ~ Temp. Gas Liquid Solvent phase phase phase C6H6 (9M) cs2 (11M) cs2 ( 1 3 ~ ) 0" 4.13 3.09 - 13.2 62.5 - 7.8 39.6 34 3-89 - 68 3.60 2.69 5.1 6.6 8-0 the later results of Walling and Mayahi because these experiments were performed at more than one temperature. The Table shows that the increase in selectivity is reflected in an increase in activation-energy differences. In fact the value (Ep-E8 I= 5.6 kcal. mole-l) calculated from the results in 13~-carbon disulphide is very much greater than the actual activation energy for the abstraction of a primary hydrogen atom in butane by chlorine atoms in the gas phase (Ep = 0-78 kcal. mole-l). There is little doubt that the chlorine atoms form a complex with the solvent.The solvents which form these complexes include aromatic compounds carbon disulphide and NN-dimethylformamide but not aliphatic hydrocarbons olefins or esters. Russell was able to show that the increase in the selectivity of chlorination in aromatic solvents was related to the basicity of the solvent as measured by the equilibrium constant for the interaction of the aromatic compound with hydrogen chloride at -78". The complex is 46 Russell J. Amer. Chem. SOC. 1957 79 2977; 1958 80 4987 4997 5002; Tetra- hedron 1960 8 101. 354 QUARTERLY REVIEWS probably a charge-transfer complex and from Walling's data its heat of formation can be greater than 8 kcal. mole-l. Russell's results suggest that the electronegativity of the chlorine atoms plays an important part in the stability of the complex and complexes from alkyl radicals are probably only weak if formed at all.This means that abstraction of hydrogen from an alkane by a chlorine atom will change from an exothermic process in the gas phase to an endothermic one in a complex-forming solvent. The increased activation energies are thus readily explained. The Vicinal Effect.-In all the work described so far it has been assumed that the radicals formed by the abstraction of hydrogen atoms from aliphatic compounds do not decompose or rearrange at normal tempera- tures. However when there is a substituent halogen on the carbon atom adjacent to the tervalent carbon there is a strong tendency to form an olefin by the elimination of a halogen atom R* -CHCI-CH,- -+ -CHCI-CH- -+ -CH=CH- f CI* This effect was first observed by Rust and Vaughan when chlorinating n-propyl and n-butyl ~ h l o r i d e s .~ ~ The apparent rate of substitution at the Bearbon atom decreased rapidly at temperatures above 200" until at 380" and 320" respectively no 1 ,Zdichloropropane or 1 ,Zdichlorobutane was formed. More recent work31 indicates that there is some decomposition of the 1-chloro-2-butyl (1-chloromethyl-n-propyl) radical even at 150". The corresponding bromobutyl radical loses bromine at room temperature and the chlorination of n-butyl bromide at 35" yielded less than 0.3% of 1 -bromo-2-chlorobutane but some 1 ,Zdichlorobutane was formed as a result of the addition of chlorine to the b ~ t - l - e n e . ~ ~ Allylic Halogenation.-The chlorination of an olefin such as propene can yield two products either allyl chloride or 1 ,Zdichloropropane CH,.CH=CH + CI* k d f k - 1 v 2 C H,-C H *C H,C I *CH,*CH= CH k I CI .1 k I CI J.CH,CHCI*CH,CI CICH2*CH= CH Rust and Vaughan found that at low temperatures addition predominated but that as the temperature rose so the extent of substitution increased at the expense of addition.48 Earlier it had been shown that the ratio of substitution to addition could be increased by lowering the concentration of the chlorine.49 Although the preparation of allyl chloride by the high- O7 Rust and Vaughan J. Org. Chem. 1941 6 479. 48 Rust and Vaughan J. Org. Chem. 1940,5,472. 49 Stewart Dod and Stenmark J. Arner. Chem. Soc. 1937 59 1765. TEDDER FREE RADICALS WITH SATURATED ALIPHATIC COMPOUNDS 355 temperature chlorination of propene is in most of the textbooks the explanation of this reaction was not given until 1956.50 In the above dis- cussion on the “vicinal effect” attention was directed to the instability of alkyl radicals in which a halogen atom is attached to a carbon atom adjacent to the tervalent carbon.This structure is produced when a chlorine atom adds to an olefinic double bond in the first step of the addition process. In considering the proportions of addition and sub- stitution there are three rate constants we need to consider (k, kPl and k2). The reactions between the radicals and the chlorine molecules (k3 and k4) will be very fast in both cases. The activation energies are all likely to be very small but that for the abstraction of a hydrogen atom (Ed is likely to be larger than that for addition (El).Similarly the reverse reaction will have a higher activation energy (E-,>E,). This means that at room temperature kl will be larger than k- or k2 but that as the temperature rises so k- and k2 will increase more rapidly than k,; also the increasing temperature will decrease the concentration of chlorine molecules. It is easy to see how these three factors can account for the change from addition to substitution at higher temperatures. We must now consider allylic substitution by special reagents of which N-bromosuccinimide is the most i r n p ~ r t a n t . ~ ~ It is generally accepted that this is a free-radical process and the mechanism usually proposed assumes that the succinimidyl radicals abstract hydrogen atoms.This raises the question why do the succinimidyl radicals not add to the double bond? Curiously this key question is ignored by the majority of writers on the S-Br -+ S* + Bra CH,-CO So + -CH,*CH=CH- -+ -CHCH=CH- + SH -CH-CH=CH- + SBr -+ -CHBr-CH=CH- + S* S = I ‘N CHZ-CO / subject. However Walling52 has suggested that it can be explained in terms of the strengths of the bonds formed [i.e. B(S-H) > D(C-S)]. On the basis of the allylic chlorination described above Goldfinger proposes an entirely different mechanism. He suggests that the function of the N-bromosuccinimide is to furnish a low concentration of molecular bromine. The reaction is thus a simple bromination with the chain pro- pagated by hydrogen bromide molecules which react with N-bromo- succinimide to yield more molecular bromine S-Br + H-Br -f S-H + Brz Bra + -CH2*CH=CH- + -CH,*CH*CHBr- Br* 4- -CH2*CH=CH- -f -CH-CH=CH- + H-Br Br -t -CHCH=CH- -f -CHBr*CH=CH- + Br.50 Adam Gosselain and Goldfinger Nature 1953 171 704; Bull. Sac. chim. belges 51 Ziegler Spaeta Schaaf Schumann and Winkelmann Annulen 1942 551 80; 52 Ref. 4(b) p. 383. 1956 65 523 533. Djerassi Chem. Rev. 1948 43 271. 356 QUARTERLY REVIEWS Thus the bromine concentration is maintained at a very low level such that allylic substitution is highly favoured. This mechanism has much to commend it. It readily explains the various initiators used with N-bromo- succinimide and unlike the other mechanism it requires no ad hoc explanations. If it is correct it should be possible to achieve allylic bromin- ation simply by adding bromine in very low concentrations to the olefin in refluxing carbon tetrachloride (the conditions normally used for N- bromosuccinimide). There is some evidence that this can be done.53 However no physical device for adding low concentrations of bromine can hope to maintain constant the extremely low concentrations that N-bromosuccinimide may be expected to nor will the hydrogen bromide be removed regularly from the system. 53 McGrath and Tedder unpublished work.

ISSN:0009-2681    

DOI:10.1039/QR9601400336

出版商:RSC    

年代:1960

数据来源: RSC

摘要:

THE PINACOL REARRANGEMENT By CLAIR J. COLLINS (CHEMISTRY DIVISION OAK RIDGE NATIONAL LABORATORY OAK RIDGE TENNESSEE U.S.A.) 1. Historical (a) Scope of the Pinacol and Related Rearrangements.-In 1860 Fittigl reported that when pinacol (1) was treated with cold concentrated Y HO OH HO OH I I ‘Me.COCMe3 RR,&-,LR2R3 -H,O Me&-CMe - R R ’C- C0R3 sulphuric acid pinacolin (methyl t-butyl ketone) (2) was produced. Evidently a molecule of water has been lost during the reaction and a methyl group had shifted from one of the central carbons to another. Since Fittig’s original observation many other examples of the acid- catalysed rearrangement of a-glycols to ketones or aldehydes have been reported and several reviews on this topic have appeared. In the reacting glycol each hydroxyl-bearing carbon may be primary secondary or tertiary and the substituent groups aryl or alkyl in nature-that is R R1 R2 and R3 of structure (3) may be hydrogen aryl or alkyl-further the two adjacent carbon atoms bearing the functional groups may be part of a cyclic sy~tem.~ Although concentrated or dilute sulphuric acid has been most commonly used many other acids (such as phosphoric per- chloric formic and oxalic) will also catalyse the rearrangement.Moreover in some cases the presence of an acid catalyst is apparently not required. For example triphenylethylene glycol (3; R = H R1 = R2 = R3 = Ph) when heated above its melting point has been reported4 to yield diphenyl- methyl phenyl ketone (4; R = H; R1 = R2 = R3 = Ph). a-Amino-alcohols ( 5 ; X = NH,) and a-halogeno-alcohols ( 5 ; X = Cl Br or I) also undergo rearrangement to produce ketones (6 and 7) usually with one of the two possible products formed in much greater yield.5 The rearrangement of amino-alcohols is brought about through the action of sodium or potassium nitrite in acid solution whereas that of the a-halogeno-alcohols is effected by the action of salts of the heavy metak5 Oxides (8) are sometimes formed during rearrangement of the Fittig Annalen 1859 110 17 23; 1860 114 54; Stadeler ibid.1859 111 277. (a) Bennett and Chapman Ann. Reports 1930 27 114-120; 1923 20 115; 1925 22 1 16; 1928,25 134; (6) Wheland “Advanced Organic Chemistry,” Wiley New York 1949 Chap. 12; (c) Ingold “Structure and Mechanism in Organic Chemistry,” Cornell Univ. Press Ithaca N.Y. 1953 Chap. 124. Meerwein and Unkel Annalen 1910 376 152; Meerwein ibid.1913 396 200; Nametkin and Delektorski Ber. 1924,57 583. Danilov J. Rum. Phys. Chem. Soc. 1929,61,723. McKenzie and Richardson J. 1923 123 79; Luce Compt. rend. 1925 180 145; Tiffeneau Ann. Chim. Phys. 1907 10 322. 357 358 QUARTERLY REVIEWS more highly sterically hindered glycols (3) and are often convertible into the same products (4) as are the glycols (3) themselves. In the present 0 (5) (6) (7) (8) Review the rearrangements of a-glycols alone will be considered; the related rearrangements will be discussed only insofar as they bear upon the a-glycol rearrangement itself. No serious attempt has been made to classify the pinacol rearrangement except incompletely according to the number of substituents borne by the reacting a-glycol. By this principle there should be six kinds of a-glycol since each methyl-carbon atom can be primary secondary or tertiary in the nature of its substituents.Further every unsymmetrically substituted glycol can rearrange in two conceivable ways depending upon which of two hydroxyl groups is lost during reaction. It is thus possible to recognise at least seventeen separate sub-classes of rearrangements based solely upon the structures of the reactant a-glycols. In practice however several of these types are unimportant and there seems little reason to discuss the sub-classification except for the historical interest involved. Thus in 1930 Bennett and Chapman2” grouped the rearrangements of pinacols under three main headings (i) the “pinacolinic change proper” of tetrasubsti- HO X RRt&-&Ra - RCYX~HRR~ + R ~ C H R R ~ RRIC/-&R2R3 HY PH F RR’C-CRzR3 -H?o RR’C-COR’ ( 3) (4) tuted (tertiary-tertiary) glycols (3)+(4) (ii) the “semihydrobenzoinic change”6 of trisubstituted (tertiary-secondary) glycols “in which the tertiary hydroxyl is lost” and (iii) the “semipinacolinic change”6 of f HO OH I I RR‘C-CHR~ - RR’C-CHO or RR’CI-I-COR~ (9) ( 10) (I I) trisubstituted (tertiary-secondary) glycols “in which the secondary hydroxyl is lost” (9)+(6) + (7).(b) Problems in understanding the Mechanism.-The pinacol rearrange- ment can be considered as a complicated Wagner-Meerwein reactionzb* in which either one or both of two adjacent hydroxyl groups [a or /3 of Originally named by Tiffeneau and Levy Bull. Soc. chim. 1923,33,758. COLLINS THE PINACOL REARRANGEMENT 359 structure (3)] can be lost during rearrangement and in which any one or all of four substituents [R R1 Rz R3 of structure (3)] can undergo 1,Zshifts.If one of the substituents is hydrogen (e.g. R = H) and an aldehyde is produced the aldehyde itself may suffer further rearrangement. These uncertainties are the factors which differentiate the pinacol from the Wagner-Meerwein rearrangement ?nd which have made it difficult to study.’ In the following paragraphs of this section we shall illustrate these factors considering the tetrasubstituted and trisubstituted glycols separately. In a subsequent section (p. 364) we shall give a mechanistic rationalisation of these factors. (i) Tetrasubstituted glycols. In order that the rearrangements of the tetrasubstituted glycols be understood the following problems have required study (1) the direction of rearrangement as determined by which of two hydroxyl groups is initially removed; (2) the “migratory aptitudes” of the substituent groups; (3) the effect of steric properties of the glycol upon the course of the reaction; (4) the role of oxides as intermediates during rearrangement ; ( 5 ) the stability of the products themselves under reaction conditions; and (6) the effect of reaction media.Bennett and Chapmanzu pointed out that the general direction of rearrangement is dependent upon which of two hydroxyl groups is removed and that this in turn is decided by which pair of groups is the more effective in electron release. Thus Meerweinsa referred to the “affinity demands” exerted by adjacent substituents upon a hydroxyl in tetra-alkyl-substituted glycols such as (12).He concluded that alkyl groups possessing odd numbers of carbon atoms exerted greater “affinity demands” than do alkyl groups possessing even numbers of carbon atoms and also that the “affinity demands” of the ethyl isopropyl and t-butyl groups diminished in that order. Nyberghsb studied the rearrangements in cold concentrated sulphuric acid of glycols (12) and (13) to ketones (14) and (15) and found that glycol (13) (whether meso or dl was unspecified) was converted into ketones (14) and (15) in the ratio of 4:1 and that glycol (12) under identical conditions was converted into these same ketones in the ratio 1 20. The conclusion was that under the specified conditions of reaction whereas the ethyl group has a greater “migratory aptitude” than methyl two methyl groups are better able to release an adjacent hydroxyl group 69 OH P T PH M%C-kEt - kC-CMuE\+ Me,CEt.CD.Et - EtC-CEt (12) (1 4) (1 5) he he (13) The formal similarity between the pinacol and Wagner-Meerwein rearrangements has been well established (see ref.2b and 24. That the migrating group in specifk cases proceeds to the migration terminus with inversion a3 in the Wagner-Meerwein reaction has been established by Bartlett and Brown J. Amer. Chem Soc. 1940 62 2927. See also Brown ibid. 1952,74,428; Brown Nardmann and Madoft ibid. 1952 74 432; Brown,ibid. 1954,76 1279. * (a) Meerwein Annulen 1919 419 121-175; (b) Nybergh Ber. 1922 55 1960; (c) Stiles and Meyer J. Amer. Chem. SOC. 1959,81 1497. 360 QUARTERLY REVIEWS than are two ethyl groups. The effect of changing the acidic medium is illustrated by the observationg that glycol (16) in cold concentrated sulphuric acid yielded the ketone (18) although the same glycol when Ph*COCMe,Ph HQ QH (17) HP QH PhMeC-C Me Ph Ph&Me-COMe (19) \ P h&-CMe2 (I 6) (1 8) subjected to the influence of a trace of sulphuric acid in acetic anhydride yielded ketone (17).These same investigatorsg report that under the influence of either of the foregoing catalysts the symmetrical glycol (19) is converted exclusively into ketone (1 8). Initial attempts2a to assign relative “migratory aptitudes” or a mobility sequence representative of the ease with which various groups move to an adjacent carbon during the “pinacolinic change proper” were reasonably successful although the probability was early recognised1° that such apti- tudes might depend upon the relative positions of the groups in a given molecule; that is whether such glycols as (13) and (19) for example were meso or DL in configuration.Other difficulties attending both the assign- ment of migratory aptitudes2a and the order in which certain pairs of groups “facilitate the fission of the adjacent hydroxyl group”2a had to do with the possibility that often the compound isolable as a result of re- arrangement was simply the least soluble and not the most abundant product.ll Bailar and Bachmann,12 as a result of their studies of the rearrangements of symmetrical pinacols of general structure (20) to the ketones (21) and HO OH I I PhRC-CRPh -c PYC-COR + Ph-CDCPhR (201 (2 1) (22) (22) (in which R is substituted phenyl) were able to assign specific numbers to a series of substituted phenyl groups representing the “migratory aptitudes” of those groups with respect to phenyl.For example Bachmann Ramart-Lucas and Salmon-Legagneur Compf. rend. 1928 188 1301 ; Montagne Rec. Trav. chim. 1901,21,6. Such must be the explanation fdr the observation by Roger and McKay J. 1931 2229 that the “8-farm” (erythvo) of 1,2-diphenyl-l-m-t~lylethylene glycd when treated with cold concentrated sulphuric acid yields a mixture cf m-methyldiphenyl- methyl phenyl ketone and diphenylmethyl rn-tclyl ketone with the latter “preponderat- ing” whereas Collins and Bowman J . Amer. Chem. Soc. 1959,81,3614 showed by the isot ipe-dilution technique that rn-methylcfiphenylmethy1 phenyl ketone was the pre- dominant product actually produced in 70 % yisld.l3 Bailar J. Amer. Cfzem. Soc. 1930 52 3596; Bachmann and Ferguson ibid. 1934 56 2081 ; Bachmann and Sternberger ibid. 1934 56 170; Bachmann ibid. 1932 54 2112. lo Montagne Rec. Trav. chim. 1906 26 253. COLLINS THE PINACOL REARRANGEMENT 361 and Moser13 have shown that when R in glycol (20) is p-tolyl rearrange- ment of the glycol in a mixture of acetyl chloride glacial acetic acid and benzene (4 2 8 by vol.) produces ketone (22) in 94 % yield and ketone (2 1) in 6% yield. From this experiment they concluded that the migratory aptitude of p-tolyl with respect to phenyl is 94:6 or 15.7. Some other migratory aptitudes obtained in a similar fashion are p-anisyl (500) ; p-phenetyl (500); p-biphenylyl (1 1.5) ; p-isopropylphenyl (9); rn-tolyl (2); m-methoxyphenyl(1-6); phenyl (1.0) ; p-bromobiphenyl (0.7) ; o-methoxy- phenyl (0.3); o-chlorophenyl (0).Most of the foregoing data were obtained however without any knowledge of whether the configurations of the glycols (20) were meso or racemic. As predicted by Montagne,lo this configuration can influence the course of the rearrangement at least in the caie of the two 1,2-di-a- naphthyl- 1,2-diphenyIethylene glycols (20 ; R = a-naphthyl). One form of m.p. 220° has been shown14 to yield only ketone (22; R = a-naphthyl) when treated with glacial acetic acid and iodine or acetyl chloride. The other form m.p. 159" produces only ketone (21) upon similar treat- ment.14J5 The influence of configuration upon the course of rearrangement has been studied rather exhaustively by Curtin and his co-workers16 in the closely related semipinacolinic deamination.For example erythro-2- amino-1 -p-methoxyphenyl- 1 ,Zdiphenylethanol(23) yields upon treatment with potassium nitrite in acid solution approximately 90 % of diphenyl- methyl p-methoxyphenyl ketone (25) through migration of a phenyl group whereas the threo-isomer (26) when similarly treated yields approximately 90% of ketone (28) through migration of the rnethoxy- phenyl group. Curtin explained these results by postulating that the trans- transition states (24) and (27) are more easily formed than the correspond- ing cis-transition states (29) and (30). It was shown later1' that in the deamination-rearrangement of stereospecifically labelled [the carbon- 14 label is denoted by an asterisk] 2-amino-1,l-diphenylpropanol [(+)-(3 I)] both of the products (-)-(34) and (+)-(35) are formed nearly ex- clusively through the trans-transition states expected by migration of the labelled phenyl group in ion (32) and of the unlabelled phenyl group in ion (33).Oxidative degradation of the products followed by radioactivity assay demonstrated that the ketone (-)-(34) possessed essentially all of its carbon- 14 in the phenyl group attached to the asymmetric carbon atom whereas the ketone (+)-(35) possessed essentially all of its carbon-14 in the phenyl adjacent to the carbonyl group. l3 Bachmann and Moser J. Amer. Chern. Soc. 1932,54 1124. 14 Bachmann and Shankland J. Arner. Chern. SOC. 1929,51 306. 15 Bergmann and Schuchardt Annalen 1931 487 234. 16 Pollak and Curtin J. Amer. Chem. Soc. 1950 72 961 ; Curtin and Pollak ibid.1951 73 992; Curtin Harris and Pollak ibid. 1951 73 3453; Curtin and Crew ibid. 1955 77 355. 17 Benjamin Schaeffer and Collins J. Amer. Chem. SOC. 1957 79 6160. 362 QUARTERLY REVIEWS Concerning the questions of product stability and of the role of oxides as intermediates Fry18 and his co-workers have shown that benzpina- colone labelled with carbon-14 in the carbonyl group (36a) undergoes isotope position isomerisation to (36b) when treated with strong acids high temperatures. Gebhart and Adamsl studied the rearrangement at of benzpinacol (37) to benzpinacolone (36) in mixtures of perchloric and acetic acid (0.12~ in perchloric acid) and demonstrated that the reaction takes place through two routes route (a) being direct rearrangement of the pinacol and route (b) involving the intermediate formation of the oxide (ii) Trisubstituted glycols.The rearrangements of trisubstituted glycols present in addition to difficulties discussed in the preceding sections some very special problems of their own. These special problems are related to (33).* l8 Fry Carrick and Adams J. Amer. Chem. SOC. 1958 80 4743. * At 75" approximately 80 Gebhart and Adams J . Amer. Chem. Soc. 1954,76 3925. of the ketone is formed through the oxide. The authors (ref. 19) suggest that oxide intermediates are important where "serious steric retardation" exists [such as in the rearrangement of 371 and unimportant in the rearrangement of for example triphenylethylene glycol. COLLINS THE PINACOL REARRANGEMENT 363 the multiplicity of paths through which trisubstituted glycols can rearrange to the anomalous aldehyde-ketone rearrangement and to the role of hydrogen during the reaction.Consider for example the structure (39) which could conceivably rearrange through paths (a) and (b) correspond- ing to the “semihydrobenzoinic change”,2a*b and paths (c) and (d) corres- ponding to the “semipinacolinic Ketone (40) could be formed through paths (a) and (4 indistinguishable without isotopic tracer studies or through path (b) by further rearrangement of the aldehyde (41). Ketone (42) can conceivably arise through paths (b) or (c). The rearrangements of trisubstituted aldehydes (4 1) have long been considered anomalous.2 For example dimethylphenylacetaldehyde (43) in cold concentrated sulphuric acid yields 3-phenylbutan-2-one (44) and not isopropyl phenyl ketone (45).20 This result is in contrast to the well-establishedg conversion of (19) into (18) with exclusive migration of phenyl rather than of methyl under conditions of both strong and mildly acid medium.21*22 R’COCHR (42) L RCOCHRR’ (a) (a) loss of tertiary hydroxyl group; hydrogen migrates (b) loss of tertiary hydroxyl group; R migrates (c) loss of secondary hydroxyl group; R migrates (d) loss of secondary hydroxyl group; R‘ migrates M%YCHO - MeCHCOMe not Me,CH-COPh (44) (45) 6h (43) Ph In the rearrangements of a-glycols containing less than four substituents hydrogen itself often undergoes a 1,2-shift from one carbon atom to another2b [path (a) (3940)].This shift can conceivably be either an intermolecular or an intramolecular process. One intermolecular mechan- ism which has received much attention is the so-called process of “vinyl dehydrati~n”,~~-~~ through which triphenylethylene glycol (46) for See also Tiffeneau and Cahnmann Bull.SOC. chim. 1935 1876 and Curtin and Schmukler J. Amer. Chem. SOC. 1955 77 1107 who state that even in dea?fination reactions “10 is a conservative estimate for the phenyl/alkyl migration ratio. . . . 22 The apparent reversal of migratory aptitudes in the aldehyde-ketone rearrangement is quite general2 For a discussion of this phenomenon see also Collins and Benjamin J. Amer. Chem. SOC. 1956,78,4329. a3 Tiffeneau Bull. SOC. chim. 1923,33 759. McKenzie and Roger J. 1924 125 844. 25 As recently as 1959 the theory of “vinyl dehydration” was seriously considered as a possibility for the rearrangement of (46). See Gould “Mechanism and Structure in Organic Chemistry” Henry Holt and Company New York 1959 p.602. 2o OrCkhov and Tiffeneau Compt. rend. 1926,182,67. 364 QUARTERLY REVIEWS example would supposedly form diphenylmethyl phenyl ketone (47) through the intermediates (48) (49) and (50). 2. Recent Progress with the Mechanism (a) Enumeration of Problems.-In the preceding sections were outlined the main problems which must be resolved before our understanding of the mechanism of the pinacol rearrangement can be on the same basis as that of the Wagner-Meerwein rearrangement. These problems are (i) Which of two hydroxyl groups is initially removed during reaction? (ii) How is the course of the rearrangement altered by changing the acid catalyst? (iii) How does the stereochemistry of the glycol affect the pro- ducts of rearrangement? (iv) Under what conditions do the products themselves rearrange? (v) To what extent are oxides involved in the re- arrangements? (vi) To what extent are aldehydes involved in the re- arrangement of other than tetrasubstituted glycols ? (vii) What is the reason for the anomalous migratory aptitudes in the aldehyde-ketone rearrangement ? (viii) Does hydrogen migrate intermolecularly or intra- molecularly (i.e.does “vinyl d e h y d r a t i ~ n ” ~ ~ - ~ ~ occur) ? During the past five years considerable progress has been made towards answering these questions chiefly through the application of isotopic- tracer techniques. In the present section a review and evaluation of that recent work is presented. (b) The Rearrangement of Triphenylethylene Glycol (46).-(i) Multiplicity of paths of rearrangement.The rearrangement of triphenylet hylene glycol (46) under the influence of various acid catalysts to produce in quantitative yields diphenylmethyl phenyl ketone (47) or a mixture of (47) and triphenylacetaldehyde (5 1) has received much attention.26 An interpre- tation of the mechanism of this rearrangement however was frustrated by the difficulties mentioned in Section 1. By labelling triphenylethylene glycol in one of the methyl carbon atoms (46a) and alternatively in the phenyl group (462) with carbon-14 and then subjecting these isotope- position isomers to rearrangement it has been possible to determine the fate of each label under the influence of a specific cataly~t.~’ The possible paths for rearrangement of (46) are outlined in Chart 1.The phenyl- labelled isomer (46c) alone is illustrated although similar charts could be drawn for two different chain-labelled isomers and the alternative ring- 26 Gardeur Bull. Acad. roy. Belg. 1897 34 67; Chem. Zentr. 1897 68 TI 622; Tiffeneau Compt. rend. 1908 146 29; Legrave Ann. Chim. 1927 8 416; Danilov J. Russ. Phys. Chem. SOC. 1917 49 282; DaniIov and Venus-Danilova Ber. 1926 59 377. 27 Collins J. Arneu. Chem. Soc. 1955 77 5517. COLLINS THE PINACOL REARRANGEMENT 365 labelled isomer since each was synthesised and s t ~ d i e d . ~ ~ ~ ~ ~ The scheme outlined in Chart 1 represents all possibilities mentioned by previous i n ~ e s t i g a t o r s ~ ~ * ~ ~ * ~ ~ with the hypothetical intermediates expressed as carbonium ions.27 Hydrogen has been as having undergone CHART 1 H e YH HP + Phk- CHPh - Ph)fC-CHPh- P;hCOCHFh,.... Path I (474 PhCCHPh LL Ph-CHCOPh ........... Path 2 (53:) 4 A (474 * + w Ph&CH-OH PhGCHO ............ Path 3 (54) 4 t (5 1) Y +QH* * * PhGCHPh - Ph,CH.COPh .......... Path 4 (53 4 (47 c d ) intramolecular migration rather than “vinyl dehydration” ;23-25 the evidence29 for such intramolecularity is discussed later. As explained by Collins,27 a determination of the yields of the products (47) and (51) through the isotope-dilution method plus a knowledge (gained through appropriate degradation techniques) of the fraction of rearrangement undergone by the carbon-14 labels of (46a) and (46c) are sufficient to allow calculation of the contribution of each path (1,2,3,4) to the overall reaction under the influence of a given acid catalyst.Table 1 gives the results of such experiments for five different catalysts and we can draw these important conclusions (1) the contributions of paths 2 3 and 4 vary considerably depending upon the acid catalyst employed (2) the phenyl/ hydrogen “migratory aptitude” (k,/kH) varies by a factor of approximately 180 1 within the range of catalysts employed,* and ( 3 ) with the exception of the rearrangement in dioxan-hydrochloric acid-water the interconver- sion of the intermediates ( 5 3 3 4 ) is reversible. (ii) Conditions for secondary hydroxyl loss. In cold concentrated sulphuric acid the tertiary hydroxyl group of triphenylethylene glycol (46) is removed in preference to the secondary hydroxyl group by a factor of at least 30 1. Thus it can be concluded that two phenyl groups are better able to stabilise the carbonium ion (53) than one phenyl is able to stabilise carbonium ion (52).It is interesting however that if the 2-phenyl McKenzie Mills and Myles Ber. 1930 63 904. McKenzie Roger and McKay J. 1932,2597; Roger and McKay J. 1933,332. 29 Collins Rainey Smith and Kaye J. Amer. Chem. Soc. 1959 81 460. * The evaluation of kb/kH depends upon (a) the fact that the two chemically identical carbonium ions (53c) and (53cd) can be distinguished isotopically and (b) the assump- tion that the rate of free rotation about the C-C bonds of the intermediates of Chart 1 is much faster than the ratcs of phenyl or hydrogen migration during the interconversion of these same intermediates. 366 QUARTERLY REVIEWS TABLE 1. Ratios of k,/kH in the rearrangement of trQhenylethylene glycoZ (46 Chart 1)" Catalyst 1 2 3 4 Ratio k4JkH Formic acid 4.7 39.0 0 56.3 1 -44-1 *70b Oxalic acid 2.7 45.3 12.6 39.4 0.MC Dilute H2S04 3.2 67.4 16.5 12-9 0-44 Dioxan-H,O-HCl 0 96.1 3.9 0 0.04 1 (a) Taken from ref.27; (6) the value 1-70 is from ref. 29 and was recalculated on the basis of 6.9% rearrangement of the chain label of (51) when converted in boiling formic acid into (47); (c) taken from data of ref. 29-the value for kb/kH of 1.13 given therein is in error. % Contribution of path Conc. H2S0 2.5 11.7 0 85.8 7-33 group of (46) is replaced by p-tolyl as in 1 1-diphenyl-2-p-tolylethylene glycol ( 5 9 the ratio of tertiary to secondary hydroxyl group removal with cold concentrated sulphuric acid as the catalyst is 3.3 1 whereas the same glycol (55) when treated with boiling formic acid exhibits no more than 5 % removal of its secondary hydroxyl group.22 A similar large varia- tion in percentage removal of the two hydroxyl groups of (56) was noted.3o H? QH Unfortunately the action of cold concentrated sulphuric acid on (56) yielded no isolable products; the action of boiling formic acid however caused preferential removal of secondary hydroxyl groups in the ratio of 1.3 1 whereas the action of perchloric-acetic acid mixtures (0.005-0*82~ in perchloric acid) at 50" favoured secondary hydroxyl group removal in the ratio of approximately 6 1.(iii) The question of "vinyl dehydration". In 1953 Mislow and Siege131 obtained the first evidence for the intramolecular nature of hydrogen migration through the conversion in boiling dilute sulphuric acid of HO OH OH i I + I Ph,C-CDPh - Phs-CDPh - PhfD-COPh (46e) (534 (47e?) (+)- 1 -phenyl- 1 -0-tolylethylene glycol into optically active phenyl-o- tolylacetaldehyde.More recently the rearrangement of the deuterium- containing 1 172-triphenyl [2-2H,]ethylene glycol (46e) to dipheny1C2H1]- methyl phenyl ketone (47e) was studied.29 In cold concentrated sulphuric acid the ketone (47e) was formed from (46e) without loss of deuterium. When (46e) was heated with dilute sulphuric acid (47e) was produced 8o Collins and Krauth unpublished work. a1 Mislow and Siege] J. Amer. Chem. SOC. 1952 74 1060. COLLINS THE PINACOL REARRANGEMENT 367 with partial loss of deuterium whereas boiling formic acid or boiling oxalic acid solution caused the formation of ketone (47) completely devoid of deuterium.It should not necessarily be inferred from the latter experiments however that deuterium does not under these conditions migrate intramolecularly for the deuterium-containing ketone (47e) when boiled with formic acid suffers complete exchange of hydrogen for deuterium. The internal nature of deuterium migration has also been demonstrated by Smith32 and his co-workers during the rearrangement of 2-methyl [3-2H,]butane-2,3-diol (57) in aqueous perchloric acid solutions. HP S>H &? tvk$-CDMe Me,C-CHiOMe Me2CH-CH (OMe)* (57) (58) (59 1 Earlier Ley and Vernon33 had obtained presumptive evidence for intra- molecular hydrogen migration in the rearrangement of 1,2-dimethoxy-2- methylpropane (58) in deuterated methanolic hydrochloric acid for they determined that the product 1,l -dimethoxy-2-methylpropane (59) possessed less than the statistical amount of deuterium.“Vinyl dehydra- ti0n”23-25 has thus been r e f ~ t e d ~ ~ * ~ ~ * ~ ~ as a mechanism for several pinacol- type rearrangements. Further under those conditions in which compound (46e) leads to (47) devoid of deuterium it has been that diphenyl- [2H,]methyl phenyl ketone (47e) itself suffers complete loss of deuterium. It thus seems unlikely that “vinyl dehydration” can be of any great general importance. Cc) Anomalous Migratory Aptitudes in the Rearrangements of Aide- hydes.-The apparent reversal of the usual migratory aptitudes in the rearrangements of aldehydes to ketones was discussed in section 1. This anomaly has now been cleared up22*34-36 through a study of the rearrange- ment of diphenyl-p-tolylacetaldehyde (60) and closely related compounds.Benjamin and Collins demonstrated22 through the isotopic-dilution tech- p-Me-CbHqCP$CHO - p-MeC6H4CHPhCOPh + p -Me.C6H4CO-CH Ph (60) (6 0 (62) nique that in cold concentrated sulphuric acid the aldehyde (60) yielded that ketone (61) formed through phenyl migration in greater yield than the ketone (62) formed by p-tolyl migration in the ratio of 4.7 1. When the statistical factor of two phenyls and only one p-tolyl was allowed for it thus appeared that the p-tolyllphenyl migration ratio was 0.43 or less than unity. Never before had ap-tolyllphenyl migration ratio or “migratory 32 Smith Bowman and Kmet J. Amer. Chem. Soc. 1959 81 997. s3 Ley and Vernon J. 1957 2987. Raaen and Collins J. Amer.Chern. SOC. 1958 80 1409. 35 Kendrick Benjamin and Collins J. Amer. Chern. SOC. 1958,80,4057. 36 Collins and Bowman J. Amer. Chem. SOC. 1959 81 3614. 368 QUARTERLY REVIEWS aptitude’’ of less than unity been o b ~ e r v e d ~ ~ * ~ ~ so it was concluded that the ratio of the two ketones was not a proper measure of the “migratory aptitude” . The reversibility of the carbonium-ion transformation (53 S 4 ) (Chart 1) indicated that the rearrangement of diphenyl-p-tolylacetaldehyde (60) to the two ketones (61) and (62) could best be interpreted through the scheme outlined in Chart 2 which shows the three ions (X,) (X,,) and (X,) in a dynamic equilibrium. If the mechanism shown in Chart 2 is correct then the quantity 2 k ~ l k p [and not the value 0.43 calculated from the product ratio of compounds (61) and (62)] is the true value of the p-tolyllphenyl migration ratio.An exact to the mechanism shown in Chart 2 is given by equation (l) in which m62 and m61 are the mole fractions of the [y+-:;;] . . . (1) kT . k~ k ~ ~ i m62. - - - . - . - - kP k k‘H m61 two ketones formed during reaction. After experimental determinations of the two ratios k ~ / k and kTol/k’H7 by use of equation (1) it was possible to calculate that the quantity 2k~/kp is about 3 a normal p-tolyllphenyl migration ratio.37 CHART 2 p-Me-C6H4CPhiCH0 -p-MeC6HiCPh;EH*OH & p-MeC6H42Ph.CHPhp Oqp” 9 (60) o(,> $1 pm (62) AH (X3) (6 0 Phg HCOC6Hi Me - p 5 P h2;C HC6H4.M e- p p - Me -C6H4C H P h CO Ph The two quantities k H / k + and k ~ ~ l / k ’ ~ can be determined in principle by subjecting the glycols (63) and (55) to the same conditions of rearrange- ment in cold concentrated sulphuric acid as was the aldehyde (601 for both of these glycols should generate the same equilibrating ions (Xl)7 (X 2) and (X3).Through the same type of isotope-dilution double-labelling experiments as those performed27 with triphenylethylene glycol (46) (see Chart l) it was possible to determine22 the contributions of each of five possible paths to the over-all reaction of the glycol (63) (both threo- and erythro-forms were studied). Since the glycol (63) labelled with carbon-14 in the 1-phenyl position proceeds to the two ketonic products as shown* p - MeC6HiCPh-CH P h P h&-CHC6H4-Me -p (63) (55) 37 Burr J. Amer. Chcm. SOC. 1953 75 5008; Burr and Ciererzko ibid. 1952 74 5426; Tietz and McEwen ibid.1955 77 401 1 ; McEwen Gilliland and Sparr ibid. 1950 72 3212; Ege and Sherk ibid. 1953 75 354; Curtin and Crew ibid. 1954 76 3719; Benjamin and Collins ibid. 1956 78 4952; Bachmann and Moser ibid. 1932 54 1124. * The unimportant paths owing to removal of secondary hydroxyl groups have been neglected in Chart 3 to simplify the discussion. COLLINS THE PINACOL REARRANGEMENT 369 in Chart 3 it was necessary only to divide the contribution of path 1 by the sum of the contributions of paths 2 and 3 to evaluate the ratio Ic~/k,. Evaluation of k ~ ~ l / k ’ ~ was also possible through a less direct method.22 Additional verification for the general validity of the mechanistic schemes CHART 3 p-MeC6H4CHFhCOPh ...... Path I +* PH (63c) -tp-MeC6HiCPh-CHPh o$J 3% (6! c) * * * + + * * p-Me.C,H,.CPhiCH-OH & p-Me.CbH4€Ph-CHPh sp-MeC6HiCHPhCOPh ov kJ [%,I *a (x,) bH (61cd) .....Path 2 OH of Charts 2 and 3 can be seen from the fact that the ratio of the contribu- tion of path 2 to that of path 3 (Chart 3) is 4.8 in excellent agreement with the observed ratio (4.7) of the two ketone products during the rearrange- ment of the aldehyde (60). In order that supporting evidence could be gained for the mechanism of the aldehyde-ketone rearrangement outlined in Charts 2 and 3 experi- ments similar to those performed with (60) (63) and (55) were carried upon the analogous compounds (64) (65) and (66) each of which undergoes rearrangement in cold concentrated sulphuric acid to mixtures. M e ~ h - H P h HO OH (64) Me Me (65) (66) Me of the ketones (67) and (68).The aldehyde (64) under these conditions is converted into the ketones (67) and (68) essentially in equal quantities. Once again a “reversed” p-tolyllphenyl migration ratio of 1 2 can be calculated from the product ratio. Application of the same mechanistic p-MeC6H4CHPh.COC6HiMe- p (p-MeC6H4)2CH .COPh (67) (68) concepts (Charts 2 and 3)22 to the experimental data obtained35 during the rearrangements of (65) (66) and (64) when coupled with a mathematical solution similar to equation (l) led to the conclusion that again the true p-tolyllphenyl migration was greater than unity. 370 QUARTERLY REMEWS Analogous tracer experiments have been carried out with diphenyl-o- tolylacetaldehyde and with the three diphenyl-o-tolylethylene as well as with the corresponding m-tolyl derivative^.^^ In all cases the mechanism outlined in Chart 2 for the aldehyde-ketone rearrangement was supported.The so-called anomaly of “reversal” of the normal “migratory aptitudes” in the rearrangement of aldehydes to ketones has therefore been solved. (d) Effect of Steric Properties upon Course of Rearrangement of Glycol.- (i) Tetrasubstituted glycols. Very little progress has been made in under- standing how the steric properties of tetrasubstituted diastereoisomeric gly- cols [e.g. the meso and racemic forms of (1 3) (19) and (20)] affect the course of rearrangement. It has been pointed out [Section (l)] that the meso and racemic forms of (20; R= a-naphthyl) undergo rearrangement with the same catalyst to yield different product~.~*J~ Since the a-naphthyl group possesses an ortho-substituent however it is not clear that meso and racemic glycols of general structure (20) in which the R groups are para-substituted phenyl radicals should necessarily always rearrange with different con- sequences.* threo- and erythro-l-Phenyl-l,2-di-p-tolyl- ethylene glycol (65) when subjected to rearrangement in cold concentrated (ii) Trisubstituted glycols. OH OH Ph p c T 7 t h o - (65) rrythro- (6 5) sulphuric acid produce different yields of the two ketones (67) and (68). This has been shown to be a consequence of different fractions of secondary hydroxyl Whereas the glycol erythro-(65) undergoes reaction with 18 % of secondary hydroxyl loss accompanied by p-tolyl migration the threo-glycol threo-(65) exhibits a maximum secondary hydroxyl loss of 3%. It is interesting that those portions of the reactions of both threo- and erythro-isomers of (65) which take place with tertiary hydroxyl removal do so with identical consequences.This fact must be a result of the similar effective bulks of the phenyl and p-tolyl groups38 occupying the tertiary position whereas the large variation in secondary hydroxyl removal is most probably caused by the greater anchimeric assistance39 provided by the tolyl group on the tertiary carbon atom of erythro465) 38 It is well established that in certain Grignard additions and lithium aluminium hydride reductions the “effective bulks” of phenyl and para-substituted phenyl groups are identical. See for example Stocker Sidisunthorn Benjamin and Collins J . Amer. Chem. Suc. 1960,82 3913; Cram and Kopecky ibid. 1959,81,2748. 39 Winstein Lindegren Marshall and Ingraham J.Arner. Chem. Soc. 1953,75 147. COLLINS THE PINACOL REARRANGEMENT 37 1 than by the phenyl group on the tertiary carbon atom of threo-(65) as they undergo migration through their respective trans-transition states? (e) Stability of Ketonic Products and Role of Oxides as Intermediates.- The observations by Fry and his co-workersls on the carbon-14 rearrange- ment which takes place when carbonyl-labelled benzpinacolone (36a) is subjected to catalysis by strong acids under conditions often used to bring about the pinacol rearrangement have already been mentioned. -This example is remarkable in that the product (36a,b) has actually been formed with a shift of the carbonyl-oxygen from one carbon atom to another. Another type of acid-catalysed ketonic rearrangement is exempli- fied by the interconversion of [l4C1]methyl t-butyl ketone (2a) with its isotope-position isomer (2b).40 This interconversion was explained*O through the rapid reversible protonation of the carbonyl group followed by reversible 1,2-shift of methyl to the carbonium centre.A similar mechanism must obtain in the conversion catalysed by concentrated sulphuric acid of 2,2,4,4-tetramethyl [3J4 Clpentan-2-one (69) into 3,3,4,4- tetramethyl [2J4C]pentan-2-one (70).41 The foregoing rearrangements are in contrast to the lack of rearrange- ment of the isotope-position isomers (47a) and (47c) of 14C-labelled diphenylmethyl phenyl ketone in the presence of cold concentrated sulphuric acid or of boiling formic for neither Fry’s “oxygen function rearrangement”lS nor the rearrangement40*41 accompanied by aryl shift was exhibited.The ketones (61) (62) (67) and (68) were similarly resistant to rearrangement in cold concentrated sulphuric Even more remarkably diphenyl [2H,]methyl phenyl ketone (47) did not lose PhCH &P h Ph2CHCO$h Ph CO-CHPh- C6H4 Me - p (47s) (47s) (+) - (6 I) deuterium,29 and optically active p-tolyldeoxybenzoin [(+)-(61)] was not racemised by treatment for 15 minutes with concentrated sulphuric acid at 0°.42 Only partial racemisation occurred even under the action of cold concentrated sulphuric acid for 16 hr. At higher temperatures the ketone (47e) does lose deuterium however in formic acid.29 It thus appears that although ketones such as (36) (2) and (69) which can be formed by rearrangement of tetrasubstituted glycols are themselves often prone to 40 Rothrock and Fry J.Arner. Chem. Soc. 1958 80 4349. Pertinent references to other examples of ketonic rearrangements are listed therein. *l Barton and Porter J. 1956 2483. 42 Raaen and Collins unpublished work. 372 QUARTERLY REVIEWS further rearrangement under conditions of the pinacol reaction such is not generally the case for those ketones which are obtainable from unhindered trisubstituted glycols. There is no proof that oxides are intermediates in the acid-catalysed ketone interconver~ion.~~ Tetrasubstituted glycols on treatment with acid often form oxides however in addition to the usual ketonic products.lg Also some hindered trisubstituted glycols yield oxides on treatment with dilute acid. For example erythro-1 -a-naphthyl-l,2-diphenylethylene glycol (71) upon treatment with dilute sulphuric acid is reported43 to yield an oxide (72) which when dissolved in cold concentrated sulphuric acid is converted into diphenylmethyl a-naphthyl ketone (73).In studies of the rearrangements in concentrated or dilute sulphuric acid of threo- and erythro-l,2-diphenyl-l-o-tolylethylene g l y ~ o l s ~ ~ t ~ ~ however no mention is made of the isolation of an oxide. Also Brown4 failed to find evidence of oxide formation in his studies of the highly hindered cis- and trans- 1,2-di-o- tolylacenaphthene- 1,2-diols. Oxide formation does not seem to be there- fore a general reaction of hindered glycols. In the rearrangement of triphenylethylene glycol (46a) I4C-labelling HO OH 1 I + (46a) Ph2c-CHPh\4 ph,CH.tOPh + Ph,?H*COPh (5la) Ph,C?HO i7 (47a) (47Q experiments have ~ h o ~ n ~ ~ g ~ ~ a measurable amount of rearrangement of the chain label in the product [about 5% of (47b) is formed] when the reaction is catalysed by boiling formic acid.This rearrangement of the chain label is undoubtedly not a consequence of secondary hydroxyl removal followed by phenyl migration (path 1 of Chart l) for later experiment^^^ have shown that carbonyl-labelled triphenylacetaldehyde (5 1 a) on similar treatment yields 8% of (47b). It is possible that triphenylethylene oxide could be an intermediate in the transformation ( 5 la)A47b) although the rapid reversible formation from (5 la) of triphenylethylene diformate followed by migration of one of the tertiary phenyl groups has also been suggested22 as a mechanism. It is clear therefore that much more information will be reqxired before the role of oxides in the pinacol rearrangement and the interconversions of ketones will be understood.43 McKenzie and Roger J. 1924 125 853; McKenzie and Dennler ibid. 2105 44 Roger and McKay J. 1933 332. Tiffeneau and OrCkhoff Compt. rend. 1924 178 1619. COLLINS THE PINACOL REARRANGEMENT 373 (f) Kinetic Studies; Deuterium Isotope Effect.-Several kinetic studies pertain to the mechanism of the pinacol rearrangement. 7,19,29.32,33145.46,47 Bunton and his c o - ~ o r k e r s ~ ~ showed that pinacol when subjected to rearrangement in aqueous acid enriched with oxygen- 18 undergoes oxygen exchange at a measurable rate. Collins et al.29 showed that both threo- and erythro-l,2-diphenyl-l-p-tolylethylene glycol (63) were converted into the same (1 1) mixture of threo- and erythro-glycol (63) in aqueous-ethanolic sulphuric acid at a rate faster than the rearrangement itself took place.The mechanism suggested by Bunton et aZ.45 applied32 to the rearrangement of 2-methyl [3-2H,]butane-2,3-diol (57) (Chart 4) leads to the kinetic equation (2) in which kexp is the experimental rate constant and h is the kexp = Kkzk,hO/(k-2 + k3) . . . . . . . (2) antilog of the Hammett acidity function48 H,. The experimental rate con- stant (kexp) is therefore not a simple function of the slow or “rate- determining” step. Both Smith,32 who studied the kinetics of rearrange- ment of normal and deuterated 2-methylbutane-2,3-diol(57) and Duncan and Lynn,46 working with pinacol (l) found plots of log kexp versus H, and of (log kexp)/[H+] versus [H+] to be linear.The isotope effect (kH/kD) in the rearrangement of (57) however was independent of the acidity of the medium and fairly constant at about 1.6-1-8. Smith3 cited reasonable evidence to show that the term [ k 3 / ( k 2 + k3)] in equation (2) is constant for the rearrangement of deuterated and non-deuterated diols (57) and concluded that the observed isotope effect arose primarily in the loss of water from (74) to produce (75) [k,]. Because the observed isotope effect was so much larger than any yet observed attributable to hyperconjuga- tion Smith49 postulated participation of hydrogen during the slow step. CHART 4 HY PH HP pH He PH HP PH Ph&-CHPh Ph,C-CDPh Ph2C-CHC6H4Me-p Ph2C- CDC6H4&-p With the aid of multiple carbon-14 labels Collins et al.29 studied the kinetics of the rearrangements of the glycols (46) (46e) (55) and (55e) in aqueous ethanolic sulphuric acid at 43.3 ’.All four compounds rearranged under these conditions with almost complete loss of tertiary hydroxyl (46) (46e) (55) (554 45 Bunton Hardwick Llewellyn and Pocker J. 1958 403. 46 Duncan and Lynn J. 1956 3512. 47 Den0 and Perizzolo J . Org. Chem. 1957 22 836. 48 Hammett “Physical Organic Chemistry” McGraw-Hill New York 1940 p. 267. Winstein and Takahaski Tetruhedrun 1958 2 316. 374 QUARTERLY REVIEWS group. The relative rates of reaction were respectively 1-00 0-588 3-31 and 2-75. It was further shown that (a) removal of tertiary hydroxyl group is rapid and reversible; (b) hydrogen migration takes place at a nearly constant rate despite the substitution of p-tolyl for phenyl at position 2 of the glycol; (c) the rate of aryl migration is essentially independent of the substitution of deuterium for hydrogen; (d) the decreased rate of rearrange- ment of the deuterated glycols is due only to that portion of the reaction proceeding with deuterium migration ; (e) the mechanism of the rearrange- ment in dilute aqueous-ethanolic sulphuric acid involves a rapid and reversible removal of hydroxyl ion followed by a rate-determining 1,2- shift of hydrogen deuterium or aryl; and (f) themost reasonable explana- tion for the large variation of k @ H with catalyst is control of these phenyl/ hydrogen migration ratios by the relative populations of particular ionic conformations.Although Collins et aZ.29 suggested the same general mechanism as B ~ n t o n ~ ~ Duncan and Lynn,4s and Smith et aZ.,32 they believed that in the reactions they studied the two equilibrium steps (corresponding to K and k2/k- of Chart 4) are extremely fast and that the isotope effect is due almost completely to a rate-determining shift of deuterium or of hydrogen (corresponding to k4 of Chart 4).also through a study of product ratios (or carbon-14 distributions) during the rearrangements of l*C-labelled isotope-position isomers of (46) and (46e) that the k ~ / k ~ isotope effect was approximately constant at a value of about 3 and independent of the acid catalyst used to effect the rearrangements. Since it had been shown2' that the ratio of phenyl to hydrogen migration ( k # / k ~ Chart 1 and Table 1) could vary by nearly 200-fold depending upon the catalyst employed it was that under all the conditions studied the highly stable tertiary carbonium ion (53) was formed without assistance (participation) by neighbouring It was OH + I Ph,C-CHPh (53) phenyl or hydrogen and then destroyed by phenyl or by hydrogen migration followed by loss of a proton as illustrated in Chart 1.(g) Relation Between Migratory Aptitude and Loss of Adjacent Hydroxyl Groups.-It appears that there is some conflict amongst the published data concerning the relation between the migratory aptitudes exhibited by aryl and alkyl groups and the abilities of these same groups to facilitate the loss of an adjacent hydroxyl group. NyberghEb and MeerweinEa showed that in the rearrangement of the glycol (1 2) in cold concentrated sulphuric acid the hydroxyl group adjacent to the two methyl groups was removed in preference to that adjacent to the two ethyl groups by a factor of 20 1.In the rearrangement of glycol (1 3) under identical conditions however migration of the ethyl group predominated over migration of the methyl COLLINS THE PINACOL REARRANGEMENT 375 group by a factor of 4 1. A quite different situation exists in the rearrange- m e n t ~ ~ in cold concentrated sulphuric acid of glycols (16) and (19). The unsymmetrical dimethyldiphenylethylene glycol (1 6) rearranges exclusively with methyl migration meaning that the two phenyl groups facilitate exclusive loss of their adjacent hydroxyl group whereas the symmetrical glycol (1 9) undergoes exclusively phenyl migration. With glycols (12) and (13) although two methyl groups are better able to facilitate loss of hydroxyl than are two ethyl groups the ethyl group possesses the greater migratory aptitude whereas in the examples (16) and (19) the same groups -phenyl-possess the greater migratory aptitude and the greater ability to facilitate loss of adjacent hydroxyl group.That this apparent anomaly is not necessarily a consequence of different steric properties of glycols (13) and (19) (that is whether meso or racemic) follows from the work of Stiles and MeyerSC on 14C-labelled compounds of the general structure (78; H OH RMeC-CMe2 Q I* 08) R = Me Et But). These authorssC demonstrated that when R = ethyl the hydroxyl group adjacent to two methyl groups was removed 74% of the time (to 26% for the other) whereas the migratory aptitude of ethyl with respect to methyl was 3935 1 ; when R was t-butyl the product was formed with almost exclusive t-butyl migration signifying that there was no loss of hydroxyl adjacent to the t-butyl group.The apparent anomaly in certain of these examples is a consequence of the presumption that the same property.of a group-namely the electron-releasing ability-which facilitates migration should also facilitate removal of an adjacent hydroxyl group. There is ample evidence for this presumption in work already with respect to the abilities of p - tolyl and p-methoxyphenyl to compete with two phenyl groups in hydroxyl release [structures ( 5 5 ) and (56)] in the observation of OrCkhof and TSeneauso that 1 -p-methoxyphenyl-2-phenylethylene glycol (79) is con- verted exclusively into p-methoxybenzyl phenyl ketone (80),51 and in the HO OH observation52 that both threo- and erythro- 1-phenyl-2-p-tolylethylene glycol undergo rearrangement in perchloric-acetic acid mixtures of varying molarity with preferential removal of the hydroxyl group adjacent to p-tolyl in the ratio of about 9 1.6o Orekhof and Tiffeneau Bull Soc. chim. 1925,37 1410. 61 See the discussion by Curtin and Crew ref. 16 concerning the greater migratory 52 Benjamin and Collins unpublished data. aptitudes of p-anisyl than of phenyl. 376 QUARTERLY REVIEWS A rationalisation of the anomaly posed in this section is to be found in the work of B~nton*~ and ~ t h e r ~ ~ ~ * ~ ~ * ~ ~ who demonstrated rapid and reversible exchange of the tertiary hydroxyl group under conditions of the pinacol rearrangement and in the work of Collins et aZ.,29 who showed the rates of rearrangement and racemisation in aqueous-ethanolic sulphuric acid of glycol (46) to be identical.If tertiary-hydroxyl group exchange in (46) is rapid and re~ersible,~~ the identity of the rearrangement and race- misation rates indicates that secondary-hydroxyl exchange if it takes place at all must be very much slower than tertiary-hydroxyl exchange. It is therefore reasonable to postulate that the rearrangementsC of the tetra- substituted glycol (78; R = But) takes place through a mechanism such as outlined in Chart 5 . If such be the case then althoughk > k2 if kgu k M e the yield of ketone (84) could predominate over that of ketone (83). CHART 5 + ?,H kM H3? * R - y - 7 - q - R-7-COMe (81) H3C CH3 H3C (83) 3. Summary During the past five years much progress has been made in understand- ing the mechanism of the hitherto mysterious pinacol rearrangement.Through the use of 14C-labelling it has been possible to determine for several cases the relative fractions of removal of the two hydroxyl groups and to conclude from these data that the groups with the greater electron- donating abilities also possess the greater abilities to facilitate ionisation of adjacent hydroxyl-bearing carbon atoms. Whenever such ionisation is rapidly reversible the migratory aptitudes of the moving groups can then become important in determining the course of reaction. The effect of change of acid catalyst on the course of reaction has been demonstrated in the rearrangement of triphenylethylene glycol and has been explained as being a consequence of changes in the ionic conformations of the inter- mediates with changes in the surrounding reaction media.The importance of aldehydes or of their conjugate acids as intermediates in the rearrange- ments of trisubstituted glycols also has been clarified. Within the range of catalysts and reactants employed it has been shown that in cold con- centrated sulphuric acid the aldehyde intermediate reaches a maximum importance becoming less important under the influence of the weaker acid catalysts. In this connection the question of the anomalous migratory COLLINS THE PINACOL REARRANGEMENT 377 aptitudes in the aldehyde-ketone rearrangement has been solved for it has been shown that these migratory aptitudes cannot be calculated directly from the product ratios but are obtainable only through a consideration of the several ionic intermediates which are present in a dynamic equili- brium.Glycols labelled with deuterium have been subjected to the con- ditions of the pinacol rearrangement and it has thus been established that the ketonic products can be formed with complete retention of deuterium ; in such cases it has been concluded that the theory of “vinyl dehydration” is invalid. Although some slight progress has been made with respect to the influence of steric properties of reactants upon the course of the rearrange- ment this problem particularly in the case of tetrasubstituted glycols is still not well understood. The role of oxides in the pinacol and related aldehyde-ketone rearrangements also remains unsolved. The impact of the isotopic method upon the pinacol rearrangement has thus provided a solution to the primary mysteries of the pinacol rearrange- ment such that our state of knowledge of this interesting and historical reaction is now equal or nearly so to our knowledge of the Wagner- Meerwein rearrangement. The Reviewer acknowledges the support both of the United States Atomic Energy Commission and of Union Carbide Corporation by whom Oak Ridge National Laboratory is operated.

ISSN:0009-2681    

DOI:10.1039/QR9601400357

出版商:RSC    

年代:1960

数据来源: RSC

摘要:

INFRARED SPECTRA OF ADSORBED GASES By V. CRAW FORD^ (THE UNIVERSITY MEDFORD MASS. U.S.A.) 1. Introduction THE infrared spectrometer has become firmly established as a useful and important tool for investigating the adsorption of gases on solids. Work published during the last ten years has described the results obtained in studies of both physical adsorption and chemisorption. It is the object of this Review to illustrate how spectroscopic methods may be used to con- tribute towards elucidating certain general problems of adsorption and for this purpose examples will be chosen from some of the solid-gas systems which have been studied. The traditional distinction between physical adsorption and chemisorption will be observed although it is becoming increasingly evident that in certain instances there are no clear-cut criteria to enable such a distinction to be made with certainty.2. Physical Adsorption (A) Spectroscopic Results to be expected from Perturbing Effects of Surface Forces on Adsorbed Molecules and vice versa.-When a gas molecuIe approaches the surface of a solid interaction between the mole- cule and the surface occurs the interaction energy U(r) being a function of the distance r from the surface. Irrespective of its functional form U(P) satisfies the following conditions U=O a t r = o o and U=cx>atr=O If the adsorption minimum in the potential curve U(r) plotted against r is located at values of r where there is practically no overlap between the wave functions of the adsorbed molecules and the lattice of the solid then physical adsorption has occurred.In this case the perturbing effects of the surface might be manifested in the following ways. (i) The surface forces distort one “side” of the adsorbed molecule more than the other and this induced asymmetry might be expected to result in the occurrence of new bands in the spectrum of the adsorbate i.e. bands which for reasons of symmetry are forbidden in the spectrum of the isolated unperturbed adsorbate molecules. Further this effect should be particularly marked in the case of highly symmetrical molecules and depending on the degree of distortion the degeneracy of vibrations may either be reduced or completely destroyed. (ii) It is well known that in passing from gaseous to condensed phases the frequencies of the corresponding bands are diminished. Since in physical t Present address Physical Chemistry Laboratories Imperial College of Science and Technology S.Kensington London S.W.7. 378 CRAWFORD INFRARED SPECTRA OF ADSORBED GASES 379 adsorption the perturbing effects of the surface forces on the adsorbate molecule are comparable to those of the surrounding molecules in the liquid state the frequency shifts would be expected to be of a similar order of magnitude to that observed in passing from the gas phase to the liquid state. Indeed the magnitude of the frequency shift might perhaps prove useful in distinguishing between physical adsorption and chemisorption in doubtful cases. (iii) In carrying out a statistical thermodynamic study of the adsorbed phase a model is chosen and various thermodynamic quantities of interest are computed for certain assumed degrees of freedom of the adsorbed molecule.Comparison of the computed with the experimental quantities then permits deductions to be made concerning the degrees of freedom lost by the adsorbate in passing from the gaseous to the adsorbed phase. It might be expected that study of band shapes and in particular high- resolution studies would provide direct information about the degrees of freedom of molecules in the adsorbed state. (iv) Condonl has studied the infrared spectrum induced by electrical fields and as might have been expected found the selection rules for the induced spectrum to be the same as those for Raman spectra. It was also shown that the intensity of bands induced by electrostatic forces should depend on the square of the field strength. Study of the infrared spectra of adsorbed molecules on ionic adsorbents should then permit determination of the field strength prevailing at the adsorption equilibrium distance.(v) It has been suggested2 that in a previous study of the adsorption of ammonia on barium fluoride hydrogen bonding between adsorbate molecules on the surface was neglected. Since hydrogen bonding has been very successfully studied by infrared spectroscopy it might be expected that the same technique would provide direct information on the occurr- ence of hydrogen bonding in the adsorbate on the surface. (vi) Statistical and thermodynamic studies of the adsorbed phase almost invariably neglect the perturbation of the adsorbate on the adsorbent. Nevertheless the presence of adsorbate molecules would be expected to perturb the adsorbent surface.This effect should be easily detectable spectroscopically if the surface possesses chemical functional groups which terminate the bulk structure of the solid e.g. OH on a silica surface. Thus when a molecule is adsorbed near the OH group the motions of the atoms of the group are perturbed and consequently a shift in the group vibration frequency might be expected. (B) Experimental Requirements.-Having indicated the type of informa- tion which might be obtained from a study of the infrared spectra of adsorbed molecules we must consider whether such spectra are in fact physically realisable. The intention here is not to discuss details of the types of cell and methods of sample preparation that have been used but E. U. Condon Phys. Rev. 1932,41,759. R. M. Dell and R.A. Beebe J. Phys. Chem. 1955,59 754. 380 QUARTERLY REVIEWS rather to direct attention to the special combination of circumstances which make it possible to obtain well-defined spectra. Since a layer of liquid 1-100 p thick is sufficient to give measurable absorption bands a monolayer 10 A thick would require 1,000 traversals of infrared radiation to give measurable spectra. Owing to surface scatter- ing however radiation is lost on each traversal. This loss due to scattering by individual particles can be greatly diminished by using particles of a size less than the wavelength of the incident radiation; but decreasing the particle size increases the specific surface of the adsorbent and so the means used to reduce scattering has the simultaneous beneficial effect of ensuring that a correspondingly greater amount of gas is adsorbed at a given pressure.It is this fortunate combination of factors which makes possible the spectroscopic study of adsorbed molecules. In practice however it is found that a layer of powder has a marked residual scattering even though the particles of which it composed satisfy the size criterion for minimum scattering. This is undoubtedly due to the formation of agglomerates having an effective size greater than the wavelength of the incident radiation. On the other hand adsorbents in the form of gels or porous glasses seem to scatter much less radiation than a powder of equivalent path length and specific surface. The reason for this it has been ~uggested,~ is that the individual surfaces are separated from each other more often than in the case of a layer of powder.In addition to losses due to scattering possible additional losses may occur because of absorption by the adsorbent. However since many solids do not exhibit strong absorption in certain regions of the spectrum this leaves certain frequency ranges available for study. The major requirements then for obtaining well-defined spectra are these (i) That there be a sufficiently high concentration of adsorbed gas in the path of the incident radiation; this can be secured by using an adsorbent of sufficiently high specific surface. (ii) Scattering by particles of the adsorbent should be eliminated or at least diminished as much as possible and this can be achieved by using particles of a size smaller than the wavelength of the incident radiation.(iii) The adsorbent should be transparent to infrared raaiarion. The remainder of this section will now be devoted to experimental evidence illustrating and confirming the expectations enumerated in section 2(A). (C) Perturbation of the Surface by the Adsorbate.-As indicated pre- viously perturbation of the surface by the adsorbate should be detected very easily if the surface possesses functional groups. Indeed the very first infrared absorption study of a solid-gas system4 was with one demonstrat- ing just this effect. a N. Sheppard Spectrochim. Acta 1959 14 249. N. G. Yaroslavsky and A. N. Terenin Doklady Akad. Nauk S.S.S.R. 1949 66 885. CRAWFORD INFRARED SPECTRA OF ADSORBED GASES 38 1 In their pioneering studies on the infrared spectra of physically adsorbed molecules Yaroslavsky and Terenin4 used as adsorbent a porous glass-like silica.Sirice it was several mm. thick and absorbed the long-wavelength radiation their work was confined to the near-infrared region (5000- 10,000 cm.-l). A band was found at 7325 cm.-l and assigned to the first overtone of the fundamental frequency of the surface OH groups. The location of this band was subsequently confirmed by Yaroslavsky5 and by Yaroslavsky and Karyakh6 On the adsorption of gases and vapours this band is broadened reduced in intensity (sometimes disappearing) and displaced to lower frequencies. The effect on the surface OH groups of gases which are physically adsorbed is demonstrated nicely by the work of Yaroslavsky and Karyakin6 on the adsorption of nitrogen on porous glass. After degassing at SO" the sharp peak at 7325 cm.-l was obtained.The adsorbent at - 180" was then exposed to 1 atm. pressure of nitrogen and the spectra were recorded at 1 10 20 and 120 min. respectively after the admission of the nitrogen. After 20 min. the 7325 cm.-l band disappeared and was replaced by one at 7257 cm.-l the latter band appearing 1 min. after the admission of nitrogen. The 7325 cm.-l band was undetectable after 20 min. but the 7257 cm.-l band increased in intensity between 20 and 120 min. Since the original band at 7325 was obtained on raising the temperature to 20" this suggests that the shift of the OH frequency is caused by physical adsorption of nitrogen. Similar results were obtained on adsorption of oxygen except that in this case the admission of oxygen at -180" resulted in (i) the instan- taneous disappearance of the 7325 cm.-l and appearance of the 7257 cm.-l band and (ii) the finding that the 7325 cm.-l band is only restored on heating to 200".Sidorov using transparent plates of glass obtained the spectra not only of surface groups but also of several different types of molecule in the adsorbed phase. Carbonyl-containing adsorbates such as acetone and benzaldehyde cause the OH overtone band to be widened increased in intensity and displaced to lower frequency the displacements being 370 and 290 cm.-l respectively for acetone and benzaldehyde. Similar effects on the first overtone band of OH have been reported by Filimono~.~*~ For example exposure of silica gel to the vapour of chlorobenzene and nitromethane produced shifts of 140 and 160 cm.-l respectively.Some very comprehensive studies of the silica surface have been carried out by McDonald.lO The adsorbent used was Aerosil 2491 or Cabosil which is a pure fumed silica. Before degassing the adsorbent showed a N. G. Yaroslavsky Zhur. Jiz. Khim. 1950,24 68. N. G. Yaroslavsky and A. V. Karyakin Doklady Akad. Nauk S.S.S.R. 1952 85 A. N. Sidorov Doklady Akad. Nauk S.S.S.R. 1954 95 1235. V. N. Filimonov Optika i Spektroskopiya 1956 1 450. V. N. Filimonov and A. N. Terenin Doklady Akad. Nauk S.S.S.R. 1956,109,982. 1103. lo R. S. McDonald J. Amer. Chem. Soc. 1957 79 850. 3 382 QUARTERLY REVIEWS sharp band at 3749 and a broader one at 3400 cm.-l. On out-gassing the former increased and the latter decreased in intensity until after prolonged evacuation at 300-350” only the 3749 cm.-l band remained.This was assigned as the 0-1 stretching frequency of OH groups oriented in such a way as to be incapable of interacting with their surroundings. The effect on this band of physically adsorbed rare-gas atoms and non-polar molecules was then investigated and in every case the OH frequency diminished as shown in Table 1. TABLE 1. Perturbation of free OH frequency of Aerosil by various adsorbates. Adsorbate None Argon Krypton Xenon Nitrogen Oxygen Methane Polarisabili ty (A”> - 1-65 2.54 4.13 1 *76 1 *60 2.60 OH Frequency (cm.-l) 3749 3741 3733 3730 3725 3737 3717 Frequency shift (cm.-l) 8 16 19 24 12 32 - McDonald sought a correlation between the frequency shifts produced by the various adsorbates and their polarisabilities. In the case of the rare gases there appeared to be such a correlation for argon krypton and xenon which have polarisabilitiesll of 1.65 2.54 and 4.1 3 A3 produced frequency shifts of 8 16 and 19 cm.-l respectively.However oxygen and nitrogen show no such correlation. These molecules have approximately the same polarisabilityll (1 -60 and 1 -76 pi3 respectively) yet produce frequency shifts of approximately 12 and 24 cm.-l respectively. Frohnsdorff and Kingtonf2 have suggested that thareason for this is to be found in the difference in the quadrupole moments of these molecules (0,<0*1 A2;13 N2 < 0.5 A2 14). Very approximate calculations based on various assumed positions and orientations of the quadrupole show that interaction between proton and quadrupole moment can satisfactorily account for the fre- quency difference provided the effective protonic charge is not less than -0.2 x e.s.u.The suggestion that the quadrupole moment of nitrogen cannot be ignored is reinforced by a study of the intensities of the OH bands. Thus McDonald’s high-pressure measurements show that the free OH band disappears at higher relative pressures for oxygen than for nitrogen indicating that the latter molecule perturbs the OH group more strongly than oxygen does. This is in accordance with the fact that l1 Landolt-Bornstein “Zahlenwerte und Funktionen” Vol. I Part 3 1951 p. 510. la G. J. C. Frohnsdorff and G. L. Kington Trans. Faraday SOC. 1959,55 1173. Is W. V. Smith and R. Howard Phys. Rev. 1950,79 132. l4 C. H. Tomes and A. L. Schawlow “Microwave Spectroscopy” McGraw-Hill New York 1955 p. 365. CRAWFORD INFRARED SPECTRA OF ADSORBED GASES 383 the quadrupole of nitrogen interacts with the field gradient of the OH group giving a higher heat of adsorption.0.0 0 0 g 0.1 0' 0-2 n $ 0.3 >" 0-4 ,o 0.5 1-0 1.5 v) .- CI 0 m 3700 3700 3700 3700 F r aq u enc y ( c m:' ) FIG. 1 . Infrared absorption due to SiOH of Cabosil (12.5 mg.lcm2 pressed at 12,000 Ib./in.2). (a) Before degassing; peak at 3747 cm.-'. (b) After degassing for 30 min. at 500" in vacuo; peak at 3748 cm?. (c) After degassing for 15 min. at 940" in vacuo; peak at 3749 cm.-'. (d) After degassing for 8.5 hr. at 940" in vacuo; peak at 3750 cm.-'. 3700 3700 3700 F r Q q u Q n c y (c m:' ) FIG. 2. Infrared absorption due to SiOH of Mallinckrodt Special Bulky Silicic Acid (12.5 mg./cm.2 pressed at 12,000 1b./in.2). (a) Degassed for 30 min.at 500" in vacuo; peak at 3740 cm.-'. (b) Degassed for 15 min. at 940" in vacuo; peak at 3748 cm.-'. (c) Degassed for 8.5 hr. at 940" in vacuo; peak at 3748 cm.-l. (Figs. 1 and 2 are reproduced with permission from R. S. McDonald J. Phys. Chern. 1958 62 1168.) More recently McDonald15 described an infrared study of silanol groups on the surface of two varieties of pure amorphous silica Mallin- ckrodt Special Bulky Silicic Acid (MSBS) which is a pure precipitated silica and Cabosil. The results indicate that OH on the surface of silica exists in different states. Thus the shift and narrowing of the residual SiOH band of MSBS is much more pronounced than for Cabosil as in- dicated in Figs. 1 and 2. l5 R. S. McDonald J. Phys. Chern. 1958 62 1168. 384 QUARTERLY REVIEWS The breadth of the band after degassing at 500" (Fig.2a) shows that silanol groups of MSBS interact with each other much more than those of Cabosil which has not been degassed at all. However degassing at 940" in a vacuum for eight hours destroyed hydrogen bonding and left ap- preciable amounts of isolated silanol groups on the surface of both silicas. As a result of the degassing the absorption band of surface silanol groups was narrowed and displaced to higher frequency. This would indicate that the groups destroyed during outgassing had an environment markedly different from those which remained after outgassing. Very recently Folman and Yates16 studied effects due to hydrogen bonding between physically adsorbed molecules and the OH groups present on the surface of porous silica glass.In this case the surface OH groups show a very strong absorption even after evacuation for eleven hours at 450". The band is not only much wider than the normal OH band but it is also highly asymmetric. This implies that the band is composite consisting of a narrow band at approximately 3740 crn.-I due to isolated OH groups and another broader band at a lower frequency. Since hydrogen bonding is known to produce broad bands it can be inferred that the broadening in the observed band is probably due to OH groups hydrogen-bonded to adjacent groups. On adsorption of sulphur dioxide chloroform acetone and ammonia a new broad band appeared in each case and at a frequency lower than that attributed to OH on the free surface. The results obtained are shown in Table 2. TABLE 2.Shifts in OH frequencies produced by various adsorbates on the surface of porous silica glass. Adsorbate Temp. so2 24 O CH3Cl 24 (CM3)2CO 25 75 135 NH3 25 75 100 150 Average displacement (cm.-l) 115 110 330 305 270 820 750 710 640 In each case the new band appearing at lower wave-number is attributed to perturbation of the surface OH groups by the adsorbate as a result of hydrogen-bond formation. At room temperature the magnitude of the perturbation of the surface OH groups increases in the order CH,Cl < SO < (CH,)&O < NH, the last compound forming a particularly M. Folman and D. J. C. Yates Proc. Roy. SOC. 1958 A 246 32. CRAWFORD INFRARED SPECTRA OF ADSORBED GASES 385 strong hydrogen bond. Later work17 showed that within experimental error dv is independent of coverage and that vt the half-width of the perturbed OH band increases with increasing temperature.In the case of acetone and ammonia where measurements were made at several temperatures the magnitude of the shift decreases with rise in temperature. This behaviour is similar to that displayed by solutions in which hydrogen bonding is present. (D) Perturbation of the Adsorbate by the Surface.-The infrared absorp- tion spectra of methane ethylene acetylene and hydrogen adsorbed on porous glass were investigated by Sheppard and Yates.18 In addition to obtaining the spectra they simultaneously determined the coverage. This work provided the first definite infrared spectroscopic evidence which clearly showed the perturbing effect of surface forces on adsorbed mole- cules and for this reason will be considered in some detail.Their results are shown in Table 3 which lists not only the frequencies for the adsorbed state but also those for the corresponding liquids and gases. It will be noted that (i) Where data are available the frequencies in the adsorbed phase are either lower than or equal to those of the corresponding liquids; thus the perturbing effects of the surface forces are greater than those of the surrounding molecules in the liquid. (ii) All the shifts in frequency are small and approximately 2 % of the gaseous value indicating that no change in chemical species could have occurred as a result of adsorption. TABLE 3. Frequencies (in cm.-l) of bands of adsorbed molecules. Molecule Gas Liquid Adsorbed vgas- vadsorbed CH v3 3018.8” 301 8 C2H4 vQ 3105*5c 3105 vll 2989.5c 2980 v1 2916~5~ (R) - vl 3019*3c (R) - C2H2 v3 3287c - 4160*2d (R) - H2 Vl 3006 12.8 2899 17.5 3 100 5.5 2980 9.5 3010 9.3 3240 47 4131 29.2 a D.R. J. Boyd H. W. Thompson and R. L. Williams Proc. Roy. SOC. 1952 A b B. P. Stoicheff C. Cumming C. E. St. John and H. L. Welsh J. Chern. Phys. 1952 c G. Herzberg “Infrared and Raman Spectra of Polyatomic Molecules,” New York d G. Herzberg Canad. J. Res. 1950 28 A 144. 213 42. 20 498. Van Nostrand 1945. As indicated previously when a molecule is adsorbed it is necessarily distorted to some extent. If the adsorbate is a highly symmetric poly- atomic molecule this reduction in its symmetry on transference from the l7 M. Folman and D. J. C. Yates J. Phys. Chzm. 1959,63,183. Is N. Sheppard and D. J. C. Yates Proc. Roy. SOC.1956 A 238 69. 386 QUARTERLY REVIEWS gaseous to the adsorbed phase allows vibrations to appear in the spectra which previously were forbidden. Also vibrations which are degenerate in the isolated gas molecule may have the degeneracy lifted or completely destroyed depending on the extent to which the symmetry is reduced. Thus in the case of methane the totally symmetric vibration vl which occurs at 2916 crn.-l in the Raman spectrum is as expected unobserved in the infrared spectra of the gaseous and the liquid phases. In the adsorbed phase however a band occurs in the infrared spectrum at 2899 cm.-l and is assigned to vl. The v3 band is a triply degenerate C-H stretch and if on adsorption the symmetry of the molecule were reduced from Td (free gas) to CSv in the adsorbed phase v3 would be split into two components of symmetry species E (doubly degenerate) and A (non-degenerate) whose intensities would be approximately in the ratio 2:l.If however transference of methane from gas to adsorbed phase should reduce its symmetry from Td to CZv the degeneracy of v3 would be completely destroyed yielding three bands of approximately equal intensity. Sheppard and Yates in fact obtained a single broad but fairly symmetrical band. (i) Rotational degrees of freedom. It is well known that the bands of a spherical-top molecule consist of P Q and R branches each branch consisting of lines which correspond to transitions between quantised rotational levels of the molecule. However a sharp fine structure will be seen only as long as the molecules can absorb radiation without interrup- tion.This can be accomplished for instance by increasing the pressure of the absorbing gas in which case the individual lines become broadened. When this becomes sufficiently great the individual rotational lines merge but the merging may have very little effect on the overall shape of the band. Sheppard and Yates then attempted to obtain some information about rotational degrees of freedom from the effects of rotational motions on the shape of the v3 band of adsorbed methane. Three possibilities which in principle can be distinguished spectroscopically are conceivable namely no free rotation in the adsorbed state; free rotation about one axis probably perpendicular to the surface; and three degrees of rotational freedom. No fine structure was observed in the spectrum of adsorbed methane at W O K whereas the spectrum of the gas phase exhibited clear fine structure.This however does not necessarily rule out the possibility of free rotation as the individual rotational energy levels may have merged. Consequently detailed calculations as follows were made of the shapes of the bands to be expected for methane adsorbed in the three ways suggested. Case I. The molecule may be so strongly adsorbed that the rotational degrees of freedom present in the gas phase become torsional oscillations on adsorption. Since rotational fine structure is absent most of the intensity occurs in a single peak and so the overall shape of the band should be well represented by a Lorentz-type curve viz., CRAWFORD INFRARED SPECTRA OF ADSORBED GASES 387 In &/I) = a / [ ( v - vo)2 + b2] where vo is the centre of the band 2b is its half-width and a/b* = Case II.Two degrees of rotational freedom are lost on adsorption the remaining one being about an axis perpendicular to the plane of the surface. In this case the rotational energy levels are given by In (IO/I)v()- E,(K) = K2h2/8n21 where K = 0 1,2 - the selection rules being AK = & 1 for perpendicular bands and AK = 0 for parallel bands. As usual the relative intensities depend largely on the factor exp [-E,(K)/kET] the statistical weights being 1 if K = 0 and 2 if K>O. Case ZII. No rotational degrees of freedom are lost on adsorption. In this case the energy levels are given by the familiar expression E,(J) = J(J + l)h2/8n21 where J = 0 1 2 the relative intensities of the rotational lines being roughly proportional to exp [-E,(J”)/kT) times the mean value of the statistical weights for the upper and lower levels [(W + 1)2].The results of the calculation showed that Case 111 can very definitely be ruled out for it allocates too much intensity to the P and R branches. So Cases I and I1 remain for consideration. Unfortunately the experi- mental evidence available does not permit a clear-cut and decisive distinc- tion to be made between these two alternatives. Since the magnitude of the induced dipole moment is proportional to the polarisability for a fixed field strength and since the polarisability ellipsoid of methane is a sphere the induced dipole moment will be independent of the orientation of the molecule. Hence the band shape is independent of rotational motions and the spectrum should consist of a single peak corresponding to a Q branch.On the basis of the free rotational model the width of the vl band should be less than that of v3 for the latter consists of P Q and R branches. Thus a simple com- parison of the relative widths of the two bands might seem to favour the free-rotational model. On the other hand the greater width of the vs band could possibly also be due to an unresolved splitting of the three-fold degenerate vibration due to surface forces. (iii) Hydrogen. The absorption band of adsorbed hydrogen was found to be symmetrical with a half-width of about 21 cm.-l. Unfortunately the shape of the band is not decisively helpful in distinguishing between models I (no free rotation on the surface) and I1 (free rotation about an axis perpendicular to the surface).The Raman spectrum of hydrogen has been obtained several timesfe and consists of a very strong Q branch together with widely spaced but lD F. Rasetti Phys. Rev. 1929 34 367. (ii) vl Band. 388 QUARTERLY REVIEWS weaker rotational wings. These rotational branches do not appear in the infrared spectrum of adsorbed hydrogen but it would be invalid to con- clude from this that the hydrogen molecule in the adsorbed state has completely lost its rotational degrees of freedom. The absence of the rotational structure could be due to insufficient hydrogen on the surface of the adsorbent. Crawford and Dagg20 had obtained the spectrum of hydrogen at high pressure under the influence of electrostatic fields and confirmed Condon’s prediction that the intensity of bands induced by electrostatic fields varies as the square of the field strength.Sheppard and Yates combined their intensity measurements with the data of Crawford and Dagg and com- puted the field strength to be about 7 x lo6 v/cm. at the equilibrium adsorption distance. This figure is of the right order of magnitude and clearly shows that the calculation of field strengths outside the adsorbent surface from intensity measurements is along the right lines. 3. Chemisorption If the minimum in the plot of U(r) against distance occurs at sufficiently small values of r so that overlap of wave functions cannot be neglected chemisorption has occurred. In this case the forces which hold the mole- cules to the surface are of an exchange nature and result in the formation of new bonds.Here several problems are of interest and it is worth inquiring what contributions might be expected from the application of infrared techniques. In the case of chemisorption these techniques were pioneered by Eischens and his colleagues.21*22 The objective is to obtain a metal adsorbent which (i) will cause as little radiation as possible to be lost by scattering and (ii) have a specific surface such as to ensure that sufficient gas is chemisorbed to give a well-defined spectrum. Since metals are very good absorbers of infrared radiation this is an additional source of loss and might seem to make impossible the application of infrared techniques to a study of chemisorption on metals. The absorption coefficient however is dependent on the size of the metal particle and so infrared study of chemisorption is rendered possible because of the happy circumstance that the means used to diminish scattering of radiation and increase the specific surface simultaneously make the adsorbent transparent to infrared radia- tion.Thus Eischens et al. 21,22 have shown that metal particles of size 3 x 10-2p are opaque but that satisfactory spectra can be obtained pro- vided the gas is chemisorbed on metal particles ,(10-2,u. In practice the the metal particles are dispersed in a non-porous silica support the particles of which are in the range 1.5-2-0 x 10-2p. (A) Surface Heterogeneity.-Probably the most significant single pro- perty of adsorption is the energetics of the process. The differential heat of 2o M. F. Crawford and I. R. Dagg Phys. Rev. 1953 91 1569.21 R. P. Eischens W. A. Pliskin and S. A. Francis J. Chem. Phys. 1954,22 1786. 22 R. P. Eischens W. A. Pliskin and S. A. Francis J . Phys. Chem. 1956,60 194. CRAWFORD INFRARED SPECTRA OF ADSORBED GASES 389 chemisorption falls markedly with coverage this variation of heat with coverage being quite complex in some cases. The fall in heat has been attributed to surface heter~geneity~~ and repulsive interactions between adsorbed species.24 Since interactions are never large it seems that in those cases where there is a marked fall in initial heat this is to be at- tributed to the heterogeneity of the surface. In particular the chemisorption of carbon monoxide on metals has been abundantly studied and by analogy with the structure of metallic carbonyls carbon monoxide chemisorption might be expected to take place by a single-site mechanism.At the surfaces of nickel25 and tungsten some evidence for this had been advanced but recently alternative possi- bilities have been suggested.28 Thus on molybdenum and rhodihm films the chemisorption of carbon monoxide was found to be almost exactly equal to that of hydrogen and on tantalum it was approximately equal to that of oxygen suggesting that on these metals the carbon monoxide molecule covers two sites. On iron and tungsten the chemisorption was respectively 1.2 and 1.4 times that of hydrogen thus implying mixed one- and two-site chemisorption. Observation of the infrared spectrum of chemisorbed carbon monoxide and variation in the spectrum with cover- age might be expected to provide some information on the heterogeneity of metal surfaces towards carbon monoxide.A good example of the study of surface heterogeneity by observing variations in the infrared spectra of a chemisorbed species with coverage is provided by the work of Eischens Pliskin and Francis,22 on the spectra of carbon monoxide chemisorbed on silica-supported palladium. The results are shown in Fig. 3 spectra A-E being recorded at progressively higher coverages. In interpreting the spectra the following considerations are pertinent. (a) With increasing coverage all bands might grow at the same relative rate. The implication of this would be that the position of the band was not a function of the bond strength and that the multiple band spectrum is not due to surface heterogeneity. (b) The bands might grow at different rates and this would be good evidence that the surface was heterogeneous.In this case the bands appearing first would be associated with the most strongly bonded carbon monoxide. (c) It is conceivable that as the coverage increases new bands might appear while simultaneously bands formed at low coverages might dis- appear. This would indicate that the structure of the chemisorbed carbon monoxide was a function of coverage. 23 H. S. Taylor J. Phys. Chem. 1926 30 145. 24 J. K. Roberts Proc. Roy. SOC. 1935 A 152 445. 25 0. Beeck A. E. Smith and A. Wheeler Proc. Roy. SOC. 1940 A 177,62. 26 B. M. W. Trapnell Proc. Roy. SOC. 1951 A 206,39. 27 Sir Eric Rideal and B. M. W. Trapnell Proc. Roy. SOC. 1951 A 205,409. 28 M. A. H. Lanyan and B. M. W. Trapnell Proc. Roy. SOC. 1955 A 227,387. 390 QUARTERLY REVIEWS Referring again to Fig.3 it will be noticed that as coverage increases new bands appear at 1835 1887,2062 and 1923 crn.-I. The last two bands increase in intensity with increasing coverage the 2062 cm.-l band growing proportionately more than the 1923 cm.-l band. On desorption the bands disappear in the reverse order of their appearance indicating that the species responsible for the 2062 cm.-l band is the least tightly bound. 100 n s u 9 0 t 0 Y) ul .- .- 5 C 8 0 t- 7 0 4.8 5 . 0 5.2 5.4 5.6 5.8 Wavelengt h ( p ) FIG. 3. Efect of increasing surface coverage on the spectrum of carbon monoxide chemi- (Reproduced by permission from R. P. Eischens S. A. Francis and W. A. Pliskin sorbed on palladium. (Coverage increases from A to E.) J. Phys. Chem. 1956 60 194.) The bands formed are assigned by analogy with the spectra of metal carbonyls and their spectra are of two types.Carbonyls of nickel,29 cobalt,32 manganese,33 and rhenium,33 in which the carbon monoxide is bound to the metal atom via the carbon have bands in the 2083-2000 cm.-l region. On the other hand in the case of dicobalt ~ctacarbonyl,~~ iron nonaca~bonyl,~~ and iron tetra~arbonyl~~ bands occur in the 1852-1 8 18 cm.-l region and these are attributed to carbon monoxide bridging two metal atoms. On this basis then the bands found for carbon monoxide chemisorbed on palladium indicate that the gas occurs on the surface bound in two ways (i) linearly to a single metal atom as Pd-CEO and (ii) bridged between two metal atoms i.e. thecarbonis bonded to two metal atoms as in (A). The spectroscopic evidence then is that the surface B.L. Crawford and P. C. Cross J. Chem. Phys. 1938 6 525. R. K. Sheline and K. S. Pitzer J. Amer. Chem. SOC. 1950 72 1107. R. K. Sheline J . Amer. Chem. Soc. 1951,73 1615. J. W. Cable R. S. Nyholm and R. K. Sheline J. Amer. Chem. SOC. 1954,76 3373. 89 G. 0. Brumm M. A. Lynch and W. Sesny J. Amer. Chem. SOC. 1954,76 3831. CRAWFORD INFRARED SPECTRA OF ADSORBED GASES 39 1 is heterogeneous and in particular indicates the presence of two types of site. 0 In the case of carbon monoxide on platinum the spectrum shows only a single intense band which occurs in the short-wavelength region and is attributed to Pt -C_O. With increasing coverage no additional bands appear but the one intense band does shift its position. Since in this case increasing coverage does not result in a multiplicity of bands it is im- possible to say whether the observed shift is due to heterogeneity or inter- action between molecules.As previously mentioned linearly chemisorbed carbon monoxide on palladium produced a band at 2070 ern? which was readily removed on desorption. In contrast the linearly chemisorbed gas on platinum is strongly bonded and so the band position for linear carbon monoxide on different metals is no indication of the chemisorption bond strength. Further spectroscopic evidence relating to the heterogeneity of the adsorbent surface is provided by the work of Yang and GarlandM on the infrared spectra in the region 1700-4000 cm.-l of carbon monoxide on rhodium surfaces. The metal in this instance was supported on a high-area alumina and spectra were taken of carbon monoxide chemisorbed on both sintered and unsintered surfaces.Not only were simpler spectra obtained in the case of the unsintered surface but in addition the behaviour of the bands on desorption was different for the two types of surface. Here then is additional spectroscopic corroboration of a fact which has long been known viz. that the character of an adsorbent surface varies with the conditions under which it has been prepared. In this particular case the difference was shown to be due to the presence of adsorbed water on the surface of the unsintered sample. Assignment of the spectra obtained and their variation with coverage would indicate that on the adsorbents used there were present at least three types of site. The type of site and the band obtained from carbon monoxide chemisorbed on it are indicated in Table 4.More recent spectroscopic evidence for the heterogeneity of an adsorbent surface is provided by the work of Terenin and roe^^^ on the spectra of nitric oxide adsorbed on transition metals and their oxides. (i) Metals. Iron chromium and nickel were obtained in the disperse state by adsorbing the corresponding metal carbonyl on alumina gel and then decomposing it in vacuo at 150". The spectra obtained for nitric oxide chemisorbed on the metals dispersed in alumina are summarised in Table 5. 34 A. C. Yang and C. W. Garland J. Phys. Chern. 1957,61,1504. 35 A. Terenin and L. Roev Spectrochim. Acta 1959 946. 392 QUARTERLY REVIEWS In interpreting these results the following facts are pertinent.In the presence of an electron acceptor nitric oxide is easily converted into the NO+ ion in which case the vibration frequency of 1876 cm.-l for the nitric oxide molecule in the gaseous state is displaced to 2000-2400 ~ r n . - l . ~ ~ TABLE 4. Location of bands (in cm.-l) in spectra of Coon rhodium. 0 0 111 II C C oc co \/ Sample -Rh- I /\ preparation Coverage -Rh- >Rh-Rh< unsintered All 2027 2095 - - 2% Rh 8% or 16% Rh Low - - 2040 unsintered High 2027 2095 2055 1905 8% or 16% Rh Low - - 2045 - sintered High 2040 2108 2062 1925 TABLE 5. Metal Frequency Desorption behaviour Iron 2008vs Disappears on desorption at 20" Frequencies (in cm.-l) of bands of NO adsorbed on Fe Cr Ni 1735w 1805w } Disappear after 1 hour at 150" Disappear on desorption at 350" 1698w 1660w 1625w Nickel 1850vs 1735w 1698w 1660w 1625w Chromium 2010 after long desorption at room Disappear after heating 1735 1698 1660 lrn Alternatively in the presence of an electron donor nitric oxide is con- verted into the NO- anion which has a vibration frequency located at 1000-1 100 ~rn.-l.~' In addition with transition-metal atoms and ions 36 W.R. Angus and A. H. Leckie Proc. Roy. Soc. 1935 A 149 327. 37 L. N. Short Rev. Pure Appl. Chem. (Austruliu) 1954 4 41. CRAWFORD INFRARED SPECTRA OF ADSORBED GASES 393 nitric oxide forms different types of covalent and co-ordinate bonds which are characterised by definite vibration freq~encies.~~ For example the characteristic frequency of N=O is known39 to be located in the range 1700-1 870 cm.-l and so the occurrence of a band in this region would be indicative of covalent bonding.With this background then it is evident that the very strong band at 2008 cm.-l in the case of iron indicates an ionised state of the chemisorbed molecule which might perhaps be attached to the surface as *NO+ the asterisk representing a bond to the surface. Judged by the ready removal of this band on desorption this kind of chemisorption is weak. Nevertheless the nitric oxide molecule undergoes a pronounced change in its electronic structure which is reflected in the large shift of its vibration frequency on chemisorption. In addition to this type of binding the bands at 1805 and 1735 cm.-l are characteristic of a double bond between nitrogen and oxygen and the nitric ,oxide molecule might perhaps be bound to the surface as ,N=O. The spectra in the case of chromium like that of iron indicate the occurrence of more than one type of binding.(ii) Oxides. Here Fe,O and Cr,O were used as gels but NiO was prepared by the thermal decomposition of the nitrate. The spectrum of nitric oxide chemisorbed on Fe203 gel is summarised in Table 6. TABLE 6. Frequencies (in cm.-l) of bands of NO adsorbed on Fe,O gel. Treatment 1927 1865 1806 1770 1738 On standing in contact . with NO (12 hr.) Increase in intensity Desorption at 20" Dis- Relative Dis- Relative appears decrease appears decrease in intensity in intensity On admission of Disappears Disappears oxygen Desorption on heating and a broad intense band appears at 1620 cm.-l Disappear at 150" v L- I Bands at 1698 1660 and 1625 cm.-l disappear at 350" The bands here may be classified into two types according to their behaviour on desorption.Thus the group of bands above 1700 cm-l are removed at 150" whereas the group below 1700 cm.-l disappear only at 350". That this classification corresponds to two different types of adsorp- tion centre is reinforced by the fact that oxygen acts only on the first 38 W. Hieber and A Jahn 2. Naturforsch. 1958 13b 195. 39 W. G. Burns and H. J. Bernstein J . Chem. Phys. 1952 20 380. 394 QUARTERLY REVIEWS group of bands. Tarte's results40 indicate that the absorption bands of 0-N=O are found in the range 1610-1690 cm.-l and so the bands 1700 1665 and 1625 cm.-l are presumably to be ascribed to a covalent bonding of nitric oxide with oxygen atoms. Nitric oxide on silica and alumina gives bands below 1700 cm.-l but only in the case of the transition-metal oxides are bands found above 1700 cm.-l.Hence the bands above 1700 cm.-l are attributed to adsorption on metal cations of the oxides. In the case of transition-metal oxides then adsorption can occur on both cations and anions and binding to the former may be either ionic or covalent. It is noteworthy that in no case is there any spectroscopic evidence for the formation of NO-. In this respect the behaviour of nitric oxide is very similar to that of carbon monoxide which is always chemisorbed with electron transfer from the gas to the solid. (B) Mechanism of Chemisorption.-During chemisorption the adsorbate undergoes chemical change which usually results in its dissociation into independent fragments. Consequently any discussion of the mechanism of chemisorption should consider not only the nature of the surface bonding but also the nature of the new species formed during chemi- sorption and it is with regard to the latter that infrared studies might prove most useful.For example the chemisorption and hydrogenation of ethylene has proved particularly perplexing and any information derived from a spectroscopic study of the chemisorption would be pertinent in unravelling this particular heterogeneous r e a ~ t i o n . ~ l * ~ ~ Chernisorption and hydrogenation of ethylene. When ethylene is catalytically hydrogenated there also occurs an exchange reaction between a hydrogen atom of ethylene and gaseous hydrogen43 which is readily demonstrated by the use of deuterium instead of light hydrogen. Farkas and consider these two reactions as occurring independently of each other and picture the hydrogenation as consisting of simultaneous addition of two atoms of the same hydrogen molecule adsorbed on the surface of the catalyst to a presumably physically adsorbed ethylene molecule C2H4 + H* + H* + C2HG On the other hand46 they regard the catalytic exchange as involving a dissociative mechanism according to which the hydrocarbon is split on the surface of the catalyst into a hydrogen atom and a hydrocarbon radical 40 P.Tarte J. Chem. Phys. 1952 10 1570. I1 D. D. Eley Quart. Rev. 1949 3 209. J. K. Laidler Catalysis 1954 1 168. 43 A. Farkas L. Farkas and Sir Eric Rideal Proc. Roy. SOC. 1934 A 146 630. 44 A. Farkas and L. Farkas Trans. Furuduy Soc. 1937 33 827. 45 A. Farkas and L. Farkas Trans. Furuduy SOC. 1939,35 906,941.46 A. Farkas and L. Farkas J. Amer. Chem. SOC. 1938,60 22. CRAWFORD INFRARED SPECTRA OF ADSORBED GASES 395 and subsequent reunion of the radical with a deuterium atom. The dis- sociative mechanism for exchange may be formulated as H CH2 II D CH2 II C2H4 4 CH + I CH + I -+ CHa=CHD Ni Li I Ni Ni In contrast to this view Horiuti and Polanyi*’ consider ethylene to be adsorbed “associatively” i.e. by opening of the double bond. Addition of a hydrogen atom then leads to formation of the “half-hydrogenated state”. This may either lose one hydrogen atom or take up a second thus leading to either exchange or hydrogenation. According to this picture exchange and hydrogenation are alternative processes of the same primary reaction viz. that resulting in the half-hydrogenated state.The views of Horiuti and Polanyi may be written CHZ - CH2 D CH2D CHZ-CHD H I I ’ ’ + ki 4 CH2 -+ Ni ?!Ji + l!Ji Ni NI C,H* -+ I Ji Here the deuterium atom adds to give an ethylnickel radical which breaks up to form an adsorbed ethylene molecule and liberate a hydrogen atom. Since these views were propounded much effort has been devoted to the hydrogenation and exchange reactions of ethylene with the object of discriminating between the dissociative and the associative mechanisms or some modification of them. Thus Twigg and Sir Eric Ridea148 claimed that their results on ethylene hydrogenation and deuterium exchange at nickel surfaces could best be explained if the double bond opened on chemi- sorption giving a complex bound by two-point attachment. Subsequent work49 seemed to confirm this result.During the course of time it has become clear that the reaction is far from simple but the efforts of Beeck and his colleague^^^^^^^^^ have done much to elucidate the nature of the chemisorption of ethylene. By using nickel films it was shown that if ethylene is carefully admitted so as never to build up an excess pressure one ethylene molecule is adsorbed per four nickel sites. Admission of excess of ethylene leads to appearance of ethane in the gas phase. Since the hydrogen for this hydrogenation could only have come from the ethylene the data were taken to mean that the primary adsorption is a dissociation of ethylene into acetylenic residues (occupying two sites) and two hydrogen atoms (each occupying one site). Excess of ethylene immediately removes pairs of *H atoms leaving pairs of empty 47 I.Horiuti and M. Polanyi Truns. Furuduy Suc. 1934 30 1164. 48 G. H. Twigg and Sir Eric Rideal Proc. Roy. Suc. 1939 A 171 55. 49 G. K. T. Corn and G. H. Twigg Pruc. Roy. Suc. 1939 A 171,70. 50 0. Beeck Rev. Mud. Phys. 1945 17 61. 61 0. Beak Discuss. Furday Suc. 1950 8 118. 396 QUARTERLY REVIEWS sites and the formation of ethane. The recent work of Jenkins and Sir Eric Ridea152 has confirmed the results of Beeck et al. that most of the surface is covered with acetylenic and polymerised acetylenic radicals i.e. that ethylene undergoes dissociative adsorption. At the present time there is no general agreement as to the mechanism of the hydrogenation or even of the way in which ethylene is chemisorbed. The reason for this is due partly to the complicated nature of the reaction and partly to the difficulty of devising a "one-result" experiment which will enable an unambiguous decision to be made in favour of one of the competing points of view.Consequently the Reviewer looks forward to the contribution which infrared spectroscopy of the surface phase can make towards the solution of this stubborn problem. The work of Eischens and P l i ~ k i n ~ ~ . ~ ~ shows that in the case of ethylene chemisorbed on Cabosil-supported nickel either dissociative or associative chemisorption may occur depending on the prevailing experimental conditions. Fr e 4 nc y (c m:' ) 3000 2 5 0 0 1500 1400 ' I 1 I 100- 9 0 B - n .- v) v) I- 9 0 - A d 6o - 3-5 4.0 6 . 5 7 . 0 Wave I ongt h ( p ) FIG. 4. Infrared spectra of (A) ethylene chemisorbed on hydrogen-covered nickel and (Figs.4 5 and 6 are reproduced by permission from R. P. Eischens and W. A. Pliskin (a) The adsorbent was a Cabosil-supported nickel (9-2 weight %) which had been reduced at 350" and cooled and the hydrogen had been pumped out for five minutes at room temperature. It is assumed that under these conditions the nickel remains covered with adsorbed hydrogen. The spectrum obtained when ethylene is chemisorbed on such a surface is shown in Fig. 4A. The C-H stretching bands occur in the region 2967- (B) the same after treatment with hydrogen. Adv. Catalysis 1958 10 1. 52 G. T. Jenkins and Sir Eric Rideal J. 1955 2490. 53 R. P. Eischens and W. A. Pliskin A h . Catalysis 1958 10 1. 54 R. P. Eischens and W. A. Pliskin J . Chem. Phys. 1956 24 482. CRAWFORD INFRARED SPECTRA OF ADSORBED GASES 39' 2857 cm.-l in saturated hydrocarbons and above 3003 cm.-l in olefins.Consequently the bands in the 2890 cm.-l region of spectrum A are assigned to C-H stretch vibrations for groups in which there are no double bonds on the carbon. The band at 1447 cm.-l is assigned to C-H deforma- tion which changes the H-C-H angle. Its occurrence shows that there are at least two hydrogen atoms on the carbon. Hence the combined presence of bands characteristic of C-H stretch and HCH deformation implies that in this case the chemisorption is associative. When the chemisorbed ethylene H2$-5H2 (asterisks representing a bond to the adsorbent) shown in spectrum A of Fig. 4 is treated with 2 mm. of hydrogen at 25" spectrum B results. Treatment with hydrogen therefore produces a shift of the 2890 cm.-l band whereas the 1447 cm.-l band is increased in intensity and displaced to 1458 cm.-l.The most significant change however is in the appearance of the small but sharp band at 1379 cm.-l. This is assigned to the symmetrical CH deformation and the increase in 1447 cm.-l intensity is explained as being due to a super-position of the CH asymmetrical deformation on the CH deforma- tion. These effects imply the presence of the half-hydrogenated state CH,CH, i.e. ethyl radicals. Thus the information contained in Fig. 4 provides spectroscopic evidence for the theory of Horiuti and Polanyi.*' (b) When ethylene is adsorbed on "bare" nickel obtained by evacuation of the hydrogen at 350" for hour or on hydrogen-covered nickel at 150" spectrum A of Fig.5 is obtained. By comparing Figs. 4 and 5 it will be seen that the intensity of the C-H bands in A of Fig. 5 is small compared with those of the associatively chemisorbed ethylene shown in A of Fig. 4. When ethylene chemisorbed on bare nickel is treated with hydrogen at 35" the band intensities increase as shown in Fig. 5B and this would seem to indicate that the species A is the spectrum of dissociatively chemisorbed ethylene. Pickering and E c k ~ t r o m ~ ~ recently obtained infrared spectra in reflec- tion of ethylene adsorbed on rhodium and nickel films. The metals were evaporated on to the mirrors of a multiple reflection cell,56 and scans taken before and after adsorption. It was found that when ethylene is added in excess to new rhodium mirrors it is adsorbed with dissociation to produce ethane.The spectrum of acetylene chemisorbed at 35" on either a hydrogen- covered or a bare nickel surface is shown in Fig. 6. This spectrum is similar to that of Fig. 4B which was assigned to the half-hydrogenated ethylene. The same spectrum was also obtained when acetylene was chemisorbed on a deuterium-covered surface. It seems then as if chemisorption of acetylene could involve a self-hydrogenation process and Pliskin and Eischens point out that the infrared evidence is consistent with the work 55 H. L. Pickering and H. C . Eckstrom J. Phys. Chem. 1959 63. 512. 56 J. U. White J . Opt. SOC. Amer. 1942 32 285. 398 QUARTERLY REVIEWS Frcrquency (cm-') 2800 6o t 3.2 3.3 3.4 3.5 3-6 Wavelength ( p ) FIG. 5. Spectrum of (A) ethylene chemisorbed on bare nickei and (B) the same afrer treatment with hydrogen.F r equ en c y (c m? 3000 2 5 0 0 1500 I 4 0 0 I I I J 3.0 4.0 6 - 5 7 - 0 Wove I engt h ( p Fro. 6. Spectrum of (A) acetylene chemisorbed on nickel and (B) the same after treat- ment with hydrogen. CRAWFORD INFRARED SPECTRA OF ADSORBED GASES 399 of Douglas and Rabin~vitch.~’ These authors found that deuteroacetylene was self-hydrogenated on nickel supported on kieselguhr to a mixture of deuterated ethylenes the greater proportion of which is C2D,. This seems to indicate that the effect is independent of any hydrogen which might have remained on the surface as a result of preparation of the catalyst. However since residual surface hydrogen after reduction is always a problem the present conclusion must be that evidence for self-hydrogena- tion is indicative rather than rigorously convincing.(C) Effect of the Support.-Almost all the work to date has been on adsorbents which have been supported on silica or alumina. That the support is not inert has been known for some time. For example S e l ~ o o d ~ ~ noticed that the concentration of electron holes in nickel oxide films can be increased if the films are supported on y-alumina. Similarly HiittigS9 showed that zinc oxide and ferric oxide as single oxides are poor catalysts for some hydrogenation-dehydrogenations yet when they are combined their catalytic activity is markedly increased. Independent spectroscopic evidence of this activity of the support is provided by Terenin and roe^^^ for the case of oxide adsorbents. Thus nitric oxide adsorbed on nickel oxide gives a band at 1805 cm.-l but when nitric oxide is chemisorbed on nickel oxide dispersed in alumina gel this band is shifted to 1850 cm?.More pronounced changes are found in the spectrum of nitric oxide chemisorbed on mixed Fe,O,-Al,O gel. In addition to the bands at 1927 1938 and 1806 cm.-l found with ferric oxide gel alone new bands appear at 1980 and 2125 cm.-l indicating the presence of new adsorption centres. This is not very surprising for since the oxide adsorbents and supports are semiconductors the support can be expected to affect the nature and concentration of defects in the oxide adsorbent. The “inert” support is also known to affect the catalytic activity of a metal adsorbent dispersed in it. Confirmatory spectroscopic evidence for this is provided by Terenin and R o ~ v ~ ~ who deposited iron from iron pentacarbonyl on aluminium oxide and also on zinc oxide and nickel oxide.The nitric oxide band at 2008 cm.-l for the Fe203-A1,03 system was found at 1985 cm.-l for the Fe-NiO system indicating that there is indeed an interaction between the support and chemisorbed nitric oxide via the iron atom. When nitric oxide is adsorbed on iron dispersed on zinc oxide instead of the one band there are now two one being at 2040 and the other at 1915 cm.-l. It is evident that the zinc oxide is far frominert and the support can markedly influence the metal sites. Eischens and PliskinM show that for chemisorption of carbon monoxide on supported metal adsorbents changing the carrier can bring about 67 J. E. Douglas and R. S. Rabinovitch J. Amer. Chem.Soc. 1952 74 2486. 68 P. W. Selwood Bull. SOC. chirn. France 1949 489. G. F. Hiittig Discuss. Faraday Suc. 1950 8 215. O0 J. Sheridan J. 1945 470. 400 QUARTERLY REVIEWS very marked changes in the spectrum of chemisorbed carbon monoxide. The effects observed by them are shown in Fig. 7. Changing the carrier from silica to y-alumina results in (i) a displacement to lower frequency of the band due to linear carbm monoxide and (ii) a large increase in the intensity of the band due to bridging carbon monoxide. Fig. 8 shows that a similar result is obtained when hydrogen is added to carbon monoxide chemisorbed on platinum supported on silica.21 In this J) .,_,---- 6 0 - 0 b I I 20 00 8 0 - 6 0 - 4 0 - I I 4.5 5 . 0 5 . 5 Wavelenqth ( A ) FIG. 7 . Spectrum of carbon monoxide chemisorbed on (A) silica-supported and (B) (Figs.7 and 8 are reproduced by permission from R. P. Eischens and W. A. Pliskin alumina-supported platinum. Adv. Catalysis 1958 10 1. case however the effects are not nearly as marked as those demonstrated in Fig. 7. Since no bands are produced in the 3700 cm.-l region (OH) or near 2800 cm.-l (C-H region) this negative evidence might be taken as indicat- ing that no significant amounts of HO)C'Pt H or o\ H/C-Pt are formed. In addition fdrmation of these structures would diminish the intensity of the band in the 2040 cm.-l region due to linear carbon monoxide. This occurs to a slight extent but is accompanied by a simultaneous increase in CRAWFORD INFRARED SPECTRA OF ADSORBED GASES 40 1 the band at approximately 1840 cm.-l due to bridging carbon monoxide.It seems then as if addition of hydrogen has not decreased the amount of carbon monoxide chemisorbed but has converted some linearly bonded carbon monoxide into the bridging form. The material reviewed here indicates the potential value of studying the infrared spectra of adsorbed gases. So far results have been obtained almost exclusively for gases either physically adsorbed on high-area silica or chemisorbed on silica-supported metals and metal oxides. Since spectro- scopic evidence has been presented which indicates that the support is not inert it becomes imperative to determine to what extent the observed effects can be attributed to interaction between the adsorbate and the adsorbent. This will have to be done before the spectroscopic results obtained for gases adsorbed on supported metals can usefully be compared with the results obtained on evaporated metal films. Despite this it can safely be predicted that infrared spectroscopic studies will increase in the future and advances will probably come in the follow- ing three directions viz. (i) increasing the frequency range available for study (ii) the obtaining of high-resolution spectra and (iii) the obtaining of fairly reliable intensity measurements. Attainment of the last objective will not be easy but a start has been made in this direction.61 The writing of this Review was supported in part by the U.S. Air Force. 61 L. H. Little J. Phys. Chem. 1959 63 1616.

ISSN:0009-2681    

DOI:10.1039/QR9601400378

出版商:RSC    

年代:1960

数据来源: RSC

摘要:

THERMODYNAMICS QF ION ASSOCIATION IN AQUEOUS SOLUTION By G. H. NANCOLLAS (THE UNIVERSITY GLASGOW) THE acceptance of the Debye-Hiickel theory of interionic attraction some thirty years ago permitted satisfactory interpretation of the thermo- dynamic behaviour of very dilute electrolyte solutions in which long-range interionic effects are important. When the ions are close together how- ever the approximations on which the Debye-Hiickel theory depends are no longer valid. The energy of their mutual electrical attraction may be considerably greater than the thermal energy which strives to maintain a random distribution. Although there is a continual interchange of ions the result is the formation of what amounts to a new entity the ion pair in the solution and this persists through a number of collisions with solvent molecules.Very early work in the field was concerned mainly with a qualitative study of the composition of more stable association products. Thus we have the well known and very stable complexes such as CdI,2- HgC1,2-. In more recent years there has been a change in emphasis in that it is now the physicochemical principles involved in the association process that are of greater interest. This Review will deal only with equilibria of the type . . . . . . . . . M: + Xrq- + MX(”-*)+ aq (1 ) involving ion pairs and their two constituent ions. The question arises as to how close the ions have to be before they can be considered as having lost their thermodynamic independence. This problem was tackled by Bjerrum,l who from purely electrostatic considerations defined a distance between oppositely charged ions within which they are to be considered as being associated into ion pairs.This distance y the ionic separation at which the mutual potential energy is equal to 2kT represents the position of minimum probability of finding an ion of opposite charge anywhere on a sphere of radius q surrounding the central ion . . . . . . . . q = z,z-e2/2ckT. (2) in which z+ and z- are the valencies of cation and anion e is the electronic charge k Boltzmann’s constant and E the dielectric constant. Analysis of the experimental data shows that the deviations of electro- lytes from the simple Debye-Huckel law depend upon short-range inter- actions between ions of opposite charge. Part of these deviations can be ascribed to ion-pair formation by assuming that the free ions will obey the Bjerrum Kgl.danske Videnskab. Seiskab Mat.-fys. Medd. 1926,7 No. 9. 402 NANCOLLAS ION ASSOCIATION IN AQUEOUS SOLUTION 403 law. Bjermm's equation (2) predicts greater deviations or ion-pair forma- tion the higher the valencies z+ and z- and the smaller the dielectric con- stant of the medium; this is in agreement with experimental results. It should be remembered however that the model used in Bjerrum's theory contains a number of simplifying assumptions the ions are regarded as spherically symmetrical and unpolarisable and the solvent as a con- tinuum of uniform dielectric constant. These assumptions are open to question and will be discussed below. We may write for equation (l) a mass-action constant or thermo- dynamic association constant (3) in which braces enclose activities square brackets concentrations and f ' s are the corresponding activity coefficients.Various methods can be used to determine these constants and these will be discussed briefly in the following section. In a symmetrical electrolyte any ion pairs formed will be effectively uncharged and will not contribute to the conductivity of the solution. The measurement of conductivity therefore provides a valuable method of studying such solutions. The free ions being assumed to obey the limiting equation of Onsager if a is the fraction of free ions at a concentration rn and b is the Onsager slope the equivalent conductance in dilute solution is given by A = a[A - b(am)t] Experimental Methods. -Conductivity. The association constant (1 - a) - [A - b(aun)i - A ] [ A - b(arn)b] a2m fk2 (1% f&2 (4) K = ~ - and is obtained by successive approximations for a.Davies and his collaborators have used this method extensively and have shown that the deviations exhibited by many 2-2 and some 3-3 electrolytes are capable of interpretation in dilute solution. Their data for manganous oxalate provide a good example.2 In the concentration range 1-37 x 104~ a varied between 0.90 and 0.21 but the irregular variations in K were no greater than the experimental uncertainty. Very precise measurements with 3-3-valent lanthanum ferricyanide3 in the concentra- tion range 3-30 x 1 0 " ~ yielded Kvalues showing no trend and the mean at each of three temperatures showed an average deviation of 0.5 %. A notable recent advance has been the work of Fuoss and Onsagefl dealing with the time of relaxation and introducing higher Coulomb Money and Davies Trans.Faraday SOC. 1932 28 609. Fuoss and Onsager Proc. Acud. Nut. Sci. 1955,41,274,1010; J. Phys. Chem. 1957 a Davies and James Proc. Roy. SOC. 1948 A 195 116. 61 668. 404 QUARTERLY REVIEWS interaction terms in the conductivity equation. Fuoss5 has applied the theory to ion-pairing electrolytes facilitating the calculation of both the association constants and the distances of closest approach of free ions. Calculations have been carried out for several salts in dioxan-water mixtures;6 and Nash and Monk,' comparing the new method with that using the older limiting equations have shown that wherever Davies's method indicates ion association for sodium chloride potassium chloride and hydrochloric acid in aqueous dioxan the Fuoss method does not contradict it.The Fuoss K values however are uniformly lower up to K = 100 1.mole-l. With unsymmetrical electrolytes the position is more complicated since the ion pair is now a new charged entity contributing to the measured conductivity. Making reasonable assumptions for A of such ion pairs however Righellato and Daviess have shown for a number of uni-bivalent electrolytes that relatively large errors in the assumed A lead only to small errors in the calculated association constants. The conductivities of mixtures of electrolytes may also be used since the deviations from additivity may be almost entirely accounted for if the dissociation constants of the possible ion pairs are taken into considera- t i ~ n . ~ The method has been applied to a considerable number of electro- lyte systems and in particular to the association of metal ions with a number of condensed phosphate ions.l0 Solubility.The measurement of solubility of a sparingly soluble salt in the presence of another electrolyte was first used by Davies for the determination of ion-pair formation.ll It was found that in order that the activity coefficients should be the same in any solution of the same ionic strength independently of the nature of the added salt allowance was necessary for incomplete dissociation. Calcium and thallous iodates have conveniently low solubilities and have been widely s t ~ d i e d . ~ ~ ~ ~ As illustration the concentrations of ionic species in saturated solutions of thallium iodate in the presence of added electrolyte MX may be obtained from (i) the activity product [T1+] [IO,-]fl2 = K s .~ . (ii) the total con- centration of MX = EX-] + [TlX] = [M+] + [MIO,] (iii) the associa- tion constant of MIO, viz. [MIO,]/[M+] [103-] f12 and (iv) electro- neutrality [M+] + [TI+] = [X-] + [IO,-I. Successive approximations for I the ionic strength andf permit calculation of the required associa- tion constant [TIXI/ [TI+] [X- ] fi2. Fuoss J. Amer. Gem. SOC. 1957 79 3301. Fuoss and Kraus J. Amer. Chem. SOC. 1957,79 3304. Nash and Monk Trans. Faraday SOC. 1958 54 1650. Righellato and Davies Trans. Faraday SOC. 1930 26 592. Davies J. 1938 448. lo Monk et al. J. 1949 413 423 427 2693. l1 Blayden and Davies J. 1930 949. l2 Davies J. 1930 2410. l3 Bell and George Trans. Farads-v Soc. 1953 49 619.l4 Nair and Nancollas J. 1957 318. NANCOLLAS ION ASSOCIATION IN AQUEOUS SOLUTION 405 Potentiometry. Numerous potentiometric methods have been applied to the problem of ion-pair formation the commonest being the measure- ment of pH. Cells both with and without liquid junction have been used the former usually incorporating the glass electrode. Incomplete dissocia- tion in salts of weak acids and bases is readily studied in this way and workers such as Bjerrum and Schwarzenbach have made noteworthy contributions to the development of the method. Recently glass electrodes have become available having resistances as low as 0-5 megohm. E.m.f.'s can be measured with conventional Poggendorff potentiometers and no amplifiers are necessary.15 Very precise e.m.f.'s can be obtained by Harned-type cells H2,Pt IHCl(ml) MSO,(wt2)l AgC1-Ag for which -1Og mH+ = (E-Eo)/k + log ml + log YH+YcI- The concentrations of ionic species may be calculated and hence the association constants of bivalent metal ~ u l p h a t e s .~ ~ ' ~ ~ Spectrophotometry. The spectrophotometric method takes advantage of the non-associated nature of dilute solutions of metal perchlorates. The changes in absorption on the addition of other ions are analysed in order to derive extinction coefficients of an assumed new species the ion pair. The optical density in a cell of unit length for a system such as that depicted in equation (1) would be written D = EM~+[M~+] + E X ~ - [ X ~ - - ] + E M X ( ~ - ~ ) + [MX(m-")+] * ( 5 ) By using wavelengths at which either Mn+ or Xm- ions do not absorb the right-hand side of eqn.( 5 ) is reduced to two terms. Optical-density measurements in the presence of various values of [Mn+] and [Xm-] enable the determination of K and GMX(~-W+. Bell and Panckhurst18 report for thallous hydroxide a value of K which agrees with that obtained by other methods. The method which is still in course of development has also been extensively used by Monk and his co-worker~.~~ Davies and Prue20 have discussed the sources of error in precision spectrophotometry and suggested a technique to minimise their effects. Reaction kinetics. The measurement of reaction rates provides a useful method for estimating the degree of dissociation of salts in solution. When an ion of a strong electrolyte takes part in a reaction with a neutral molecule the reaction velocity is usually more closely proportional to the concentration of the ion than to its activity.21a Thus when the ion is l5 Covington and Prue J.1955 3696. l6 Jones and Monk Trans. Faraday Soc. 1952 48 929. l7 Nair and Nancollas J. 1958 3706; 1959 3934. l8 Bell and Panckhurst Rec. Trav. chim. 1956 75 725. l9 Bale Davies Morgans and Monk Discuss. Faraday Suc. 1957 24 94. 2o Davies and Prue Trans. Furaday Soc. 1955 51 1045. 21 (a) Bell and Prue J. 1949 362; (b) Bell and Panckhurst J. 1956 2836. 406 QUARTERLY REVIEWS present as a salt which is incompletely dissociated the reaction velocity may be used as a measure of its true concentration. The K values are in reasonable agreement with those derived by other methods. In certain cases the half-wave potential of the metal ion is shifted to more negative values as the concentration of associating anion is increased.Provided the reduction process is reversible the shift in half-wave potential Polarography. d E4 = -(RT/nF) In K - (pRT/nF) In [xm-] . . . . (6) in which n is the number of electrons involved and p the number of Xm- ions associated with each cation. The method therefore gives direct information concerning both the composition and stability of the ion pair but suffers from the disadvantage that in order to ensure that the current is purely diffusion-controlled the ionic strength must be maintained at a relatively high value with neutral electrolyte. Much of the work has been done at high values of [Xm-] leading to p > 1,22 since it is then feasible to assume that [Xm-] at the surface of the mercury drops is equal to the value in bulk solution.For copper malonate however concentrations at the drop surface have been calculated and experiments at low values of Ern-] have yielded the association constant for the formation of CuC3H,0 in good agreement with that obtained by other A number of other properties are affected by ion asso- ciation in solution and may be used to determine K such as cryoscopy and vapour pressure,24 the Wien effect,25 nuclear magnetic resonance,26 and sound ab~orption.~’ Other methods. The Association Constant.-Equation (3) may be written in which K is the concentration quotient. The determination of K requires the evaluation or elimination of the activity coefficients term. This can be done in two ways (a) the determination of K at various salt concentra- tions and extrapolation to infinite dilution at which the activity coefficients become unity or ( b ) the calculation of activity coefficients from theoretical relationships.Another way of dealing with the situation has been to use constant ionic media. The equilibria are studied in solutions containing relatively large concentrations of “neutral” or inert electrolyte such as alkali-metal perchlorate or nitrate. It is assumed that these ions do not form complexes aa Lingane Chem. Rev. 1941 29 1. 23 Gelles and Nancollas J. 1956 4847. 24 Brown and Rue Proc. Roy. SOC. 1955 A 232 320. a5 Patterson J. Phys. Chem. 1952 56 999. 26 Richards Chem. SOC. Special Publ. No. 12 1958 p. 173. Eigen and Wicke J. Phys. Chem. 1954 58 702. NANCOLLAS ION ASSOCIATION IN AQUEOUS SOLUTION 407 with the reacting species and also that the activity coefficients remain constant.Concentrations are then calculated from the experimental measurements leading to molarity quotients Q = [MX(n-m)+]/ [Mn+ J [Xm-j valid only at the ionic strength used. The disadvantages of this method are as follows (i) The results are not comparable unless deter- mined at the same ionic strength and unfortuntely different workers use different ionic strengths. (ii) Too little is known about the variation of activity coefficients at constant ionic strength to permit reliable estimates of uncertainties from this source. (iii) The assumption that perchlorate ions do not form complexes is open to question especially when their concentration is high. Thus Sykes28 has shown that the spectrophoto- metry of ferric perchlorate solutions indicates formation of perchlorate complexes and the same must surely be true for ions such as Hg,%.Schwarzenbach one of the first to use 0-lwneutral salt solutions has attempted29 to be consistent by using the same activity data to convert pH into concentration. Sillh and his favour the use of very much higher neutral salt concentrations and a large number of equilibria have been studied. In these cases involving several simultaneous equilibria the incorporation of activity factors in the measured constants is a necessity for a solution of the problem. Estimation of variable activity coefficients as well as concentrations of several ionic products would lead to insuper- able difficulties. Although the constant-ionic-strength procedure is useful in comparing different metals with the same ligand too many different values of I have been used.It is very desirable to have the thermodynamic constants so that different systems are theoretically comparable and other thermodynamic properties may be obtained. Measurement of K at various ionic strengths and extrapolation to zero would seem to be a most promising method of determining thermodynamic constants and it has been used in a few instances. There is still some difficulty however in the choice of the ionic-strength function to be used for extrapolation. The estimation of activity coefficients in dilute solutions has been the most widely adopted method. For this there are a number of expressions from which to choose each of which contains different assumptions. The simplest equation is the limiting law of Debye where A is the universal constant of the Debye-Hiickel theory and this provides a useful means of extrapolating experimental data to infinite dilution.The limiting law is however accurate only at very high dilutions of the order I<10-3. Sykes Chem. SOC. Special Publ. No. 1 1954 p. 64. as Schwarzenbach Kampitsch and Seiner Helv. Chim. Ada 1945 28 828. 9o SillCn Quart. Rev. 1959 13 146. 408 QUARTERLY REVIEWS The Debye-Hiickel equation may be written logf = -Az21*/(1 + Bilk) . . . . (9) where B is another fundamental constant and a“ is the closest distance of approach of the ions. It has been shown31 that when we are dealing with systems containing mixed electrolytes as is often the case it is thermo- dynamically inconsistent to use this formula for estimating activity coeffi- cients.Moreover it cannot be readily extended to include such cases. In so far as the ion-size parameter a“ is unknown the activity coefficient cannot be calculated unambiguously. There is the further effect of dielectric saturation in the vicinity of an ion since as Gurney has pointed the intensity of the field at a distance of 2 A from the centre of a singly charged ion in a vacuum amounts to 3.6 x lo8 v/cm. In such a field the free energy lost by the polar water molecules will be considerable and there is a large uncertainty in the value E to be used for calculating A and B. The universal adoption of the macroscopic E implies that we are considering the effect of a weak field on a large number of solvent molecules instead of a strong field such as is found in the ion’s vicinity on the small number of surrounding molecules.Gurney has shown however that the two methods of treatment result in approximately the same energy contribution so the predicted effect is at least of the right order of magnitude. To eliminate as many arbitrary parameters as possible equations of the type 10gfz= -Az2I* +PI . . . . . . (10) have been used by Harned and his co-w~rkers,~~ in which /3 is an empirical constant whose value is determined by the best straight line to fit the experimental points. Other workers have used the full Debye-Hiickel expression with the addition of a linear term Az21* logf = - ____ +PI . . . . . (11) 1 + Ba“l* but this suffers from the disadvantage of having two arbitrary parameters. Guntelberg3* suggested the use of the formula containing no adjustable parameter.Obviously such a formula cannot compete in accuracy with those containing parameters but it has the 31 Guggenheim “Thermodynamics,” NoAth Holland Publ. Amsterdam 1949. 32 Gurney “Ionic Processes in Solution McGraw-Hill New York 1953. 33 Harned and Owen “Physical Chemistry of Electrolytic Solutions,” 2nd edn. 34 Guntelberg Z . phys. Chem. 1926 123 243. Rheinhold New York 1950. NANCOLLAS ION ASSOCIATION IN AQUEOUS SOLUTION 409 advantage that it can be extended to solutions of several electrolytes. Davies has proposed the formula I * ) . . . ( 1 + I - logf = Az2 __- - 0.21 . . (13) which has the form of the modification of the Debye-Hiickel equation suggested by G~ggenheirn.~~ It corresponds to an a" value in eqn. (9) of 4.2-4.4 A for 2-2 electrolytes up to I = 0 0 1 .~ ~ DavieP has shown that the equation is in good agreement with the actual values of the mean ionic activity coefficients of dilute solutions of 1-1 1-2 and 2-1 electro- lytes the average deviation being about 2% in 0-1M-solution and pro- portionately less at lower concentrations. The equation has been applied to the activity data of many electrolytes in which incomplete dissociation was assumed and the derived K values agreed with those obtained from conductivity measurements. Recently Guggenheim has criticised the use of the constant 0-2 in the Davies equation for other than 1-1 electrolyte^.^^ He proposed a value of 2.0 for 2-2 electrolytes corresponding to a more realistic distance of approach to define association of about 9A. In the case of lanthanum ferrocyanide in which on the assumption that the forces involved are purely electrostatic the corresponding distance would be 40A Panckhurst and Wo~lmington~~ have shown in a spectrophotometric study that the effect on K of using the required constant 6.0 in eqn.(13) is negligible. The activity-coefficient corrections were rather small in this work since I < 6 x 10-3~. For the stable copper oxalate ion pair,39 the use of the suggested 2.0 in place of 0-2 in eqn. (1 3) leads to a reduction in K of about 15-20 %. The bivalent metal sulphates which are considerably less stable have been studied by a freezing-point method,24 and it has been shown that the results can be fitted equally well by various choices o f a". This does not seem to be the case in the same systems studied by an e.m.f.method involving mixed electr01ytes.l~ Prue40 has pointed out that the arbitrariness arises because properties that are really complicated functions of the separation of pairs of ions are treated as if they changed sharply at a certain separation. Even though the precision of the experimental results is improved this arbitrariness will persist and can only be removed by refinement of the model. Some information about the interatomic distances in complexes in solution can be obtained from X-ray studies41 and in certain cases the structure of the crystalline solid may be used to justify a particular choice of distance 35 Guggenheim Phil. Mag. 1935 19 58. 36 Davies J. 1938 2093. 37 Guggenheim Discuss. Faraday SOC. 1957 24 53. 38 Panckhurst and Woolmington Proc. Roy.SOC. 1958 A 244 124. 3Q McAuley and Nancollas Trans. Furuday SOC. 1960 56 1165. 40 Prue Ann. Reports 1958 55 14. 41 Nilsson Arkiv Kemi 1958 12 513. 410 QUARTERLY REVIEWS parameter. Beevers and Lipson's X-ray data for copper sulphate4z were used to show that at least for this salt a distance of approach of the free ions of about 4 was reasonable. In view of the difficulties involved in the calculation of activity coefficients some workers regard the derived K values more in the nature of adjustable parameters. If the concept of ion association were successful only in accounting for thermodynamic properties this view might be justified. Davies has shown however that the same K will also account for the conductivity data and in some cases yet other proper tie^.^^ Additional evidence that different experimental methods of determining Krefer to the same species has been put forward by other workers.44 Although it has been suggested that e.g.in the case of magnesium sulphate the activity-coefficient data can be interpreted in terms of the Debye-Huckel theory by using an essentially smaller a" parameter without invoking ion association the idea of ion-pair formation has been confirmed independently by sound-absorption measurement~.~~ The terms "ion pair" and "complex" have been used by different workers to distinguish between different types of association and some- times with different meanings. A distinction proposed by Smithson and Williams45 is that the former involves solvent interposed between the ions Mn+(HzO)Xm- as distinct from MX("-")+. For transition metals they suggest that ion-pair formation unlike complex formation will be expected to have very little effect on the low-intensity absorption bands in the visible spectrum.These are due to forbidden d-d transitions the wavelength depending upon the separation of the d-states which in turn is a function of the polarisation of the cation by the associating anion. Ion-pair forma- tion will be expected to have very little effect on the d-d splitting because of the large separation of Mn+ and Xm-. On this basis spectrophotometric measurements on cobaltous sulphate and thiosulphate indicate ion pairs in the former and complexes in the latter even though their thermodynamic association constants are about equal K(CoS0,) = 2.3 x lo2 and K(C0S203) = 1.2 x lo2. The free-energy change accompanying ion association may be written where Go's are standard molar free energies of the species.Using the above definition of terms Duncan and K e ~ e r t * ~ recognise two types of behaviour metal-ion complexes in which - d Go values fall with increasing G"[M"+]aq and ion pairs in which - dC" is almost independent of Go [M"+]aq. When two ions interact without disturbing their hydration sheaths the energy will depend on the ionic charges and the distance 42 Beevers and Lipson Proc. Roy. SOC. 1934 A 146 570. 4a Davies "The Structure of Electrolytic Solutions,"ed. Hamer Wiley 1959 Chapter 3. 44 Austin Matheson and Parton op. cit. ref. 43 Chap. 24. *6 Smithson and Williams J. 1958 457. 48 Duncan and Kepert op. cit. ref. 43 Chap. 25. NANCoLLAS - ION ASSOCIATION IN AQUEOUS SOLUTION 41 1 between them and this distance being a function of the hydration radius does not vary appreciably.Difficulties arise if we wish to distinguish between ion pairs resulting from electrostatic forces between the ions and complexes in which covalent forces are also involved. Bjerrum's treatment may be applied to see whether it is adequate to explain the experimental results in terms of crystallographic radii. The experimental Kvalue for lanthanum ferricyan- ide being used in the Bjerrum formula' K = 4nJ; exp(z+z-e2/ckTr)r3.dr the calculated closest distance of approach is 7.2 A a value which suggests that the cation probably retains its primary solvation sheath in the as- sociated form. In contrast to this impossibly small distances are calculated for PbC1+ and CdC1+ (which are simply first stages in the formation of higher halide complexes) and for thallous-ion pairs and this has been used as evidence for covalent-bond formation.At the other extreme the Bjerrum theory fails for a large number of 2-1-valent salts in which ion pairs cannot be detected and Davies has suggested that definite classifica- tion into ion pairs and complexes on this basis should not be made.43 W the work done in the formation of an ion pair from the separate ions in solution may be written W = dG"/N - kTln 55.5 . . . . . . . (1 5) where N is Avogadro's number. Gurney32 has emphasised the importance of the term - k T h 55.5 which makes the K values dimensionless and is characteristic of a reaction in aqueous solution in which the number of solute particles is decreased by unity 55.5 being the number of moles contained in 1000 g.of water. Wmay conveniently be regarded as consist- ing of two parts,32 Wenv representing long-range electrostatic forces dependent upon environment and increasing with temperature and Wnon representing short-range or quantum-mechanical forces insensitive to environment and independent of temperature. W = Wenv + Wnon = -kT(ln K + In 55.5) = -kT In Kx (16) Davies4' has suggested that the more associated nature of the hydroxides of most metals than of their other compounds may be due to the absence of the hydration shell which protects the cation. Whereas for most anions the configuration Mn+(H20)Xm- may represent their distance of closest approach to the cation Mn+(H,O)OH- would tend to pass over into Mn+OH-(H,O) thus displacing a water molecule of hydration.The hydroxides are thus a particularly useful series for comparison. Between OH- and Caw Srw Baw Mgw and La* ions the only short-range forces will be forces of repulsion since the electronic shells of the cations 47 Davies J 1951 1256. 412 QUARTERLY REVIEWS contain the same number of electrons as neutral rare-gas atoms. The association constants for these ion pairs will therefore be only moderately large (Table l) governed mainly by Wenv and it is not surprising that log K varies linearly with z2/r+ where r+ is the radius of the cation a relationship shown by Davie~.~' On the other hand Few VW and V e do not have rare-gas structures and quantum-mechanical forces of attraction together with the electrostatic attraction will lead to the large K values in Table 1.As would be expected this second group of ion pairs does not fit in with Davies's pK relationship. TABLE 1. Ion association reactions Reaction Pb2+ + NO3- Pb2+ + Cl- Pb2+ + Br- Pb2+ +I- Cd2+ + C1- Cd2+ + Br- Cd2+ + I- Tl+ + OH- T1+ + C1- T1+ + Br TI+ + CNS- T1+ + N3- Ag+ + C1- Ag+ + 103- Ca2+ + OH- Ca2+ + OH- Sr2+ + OH- Ba2+ + OH- Mg2+ + OH- La3+ + OH- Fe3+ + OH- V3+ + OH-" U4+ +OH- Ca2+ + S042- Mg2+ + Mn2+ + S042- Co2+ + S042- Ni2+ + S042- Zn2+ +SO4+ T1+ + NOS- K298 (1.mole-l) 15.1 41 72 83 91 141 282 6-67 4.76 7.67 2.15 6-25 2.44 6.8 2.0 x 103 25 23 6.7 4.4 380 2 x 103 2.5 x 1013 6.4 x loll 5 x 10l2 200 234 181 230 21 1 240 Mg2+ + CH3C02- 17.6 A Go (kcal. m o le-l) -1.62 -2.20 -2.53 -2.62 -2.67 -2.93 -3.34 -1.12 -0.93 -1.2 -0.45 -1.09 -0.53 -4.5 -1.13 -1.91 - 1.86 -1.12 -0.87 -3.5 -4.1 -16.1 - 17.2 -18.2 -3.15 -3.22 -3.07 -3.21 -3.16 -3.25 -1.69 A H" (kcal.mole-l) 1.25 0.3 0.98 -0.57 -0.3 -0.32 -2.05 0.37 -1.43 -2.45 -0.65 -2.96 - 1-33 -2.7 5-14 1.19 1-25 1.15 1.75 - - -3.0 -3.7 -2.5 1.65 4.55 3.37 1-74 3.3 1 4.0 1 -1.52 A So (cal .deg.-l mole-l) 3.5 11-6 9.4 7.8 12.2 8.8 4.3 5-1 -1.7 -4.2 -1.0 -6.4 -2.7 6 20.3 10.4 11.3 7.6 8.8 - - 44 45.5 52 16-1 26.1 22-6 16.6 21.7 24.4 0-6 Ref. a 44 44 44 44 44 44 13 13 14 14 13 14 b 13 d d d e 47 g h 13 17 17 17 17 17 C f n NANCOLLAS ION ASSOCIATION IN AQUEOUS SOLUTION 413 K298 (1.mole-l) Reaction La3+ + CH2(C02)22- 1.0 x lo5 Gd3+ + CH2(C02)22- 2.5 x lo5 Lu3+ + CH2(C02)22- 5.3 x lo5 Law + Fe(CN),3- 5.5 x lo3 Mgw + edtaGb 3.5 x los Ca2+ + edta4-b 3.3 x 1O1O Co2+ + edta4-b 1-15 x 10l6 Ni2+ + edta4-b 2.0 x l0lS Zn2+ + edta4-b 1 .7 ~ 10l6 AP+ + edta4-b 6.9 x 1015 Y3+ + edta4-b 6-2 x 1017 Ca2+ + CH3.C02- 17.5 La3+ + CO(CN)G~- 5.8 x lo3 A Go (kcal. mole-l) - 1.69 -6.79 -7.32 -7-77 - 5.09 -5.13 -11.65 - 14.34 -21.9 -24.96 -22.12 -21.60 - 24.26 A H" (kcal. mole-l) 0-9 1 4.8 5.1 5.2 2.0 1.33 3.14 -6.45 -4.4 -8.35 -5.61 12.58 0.32 A So (cal. deg. -1 mole-l) 8.7 39 42 44 23.9 21.7 50-5 26.9 59.7 56.7 56.3 116-6 83.8 Ref. a i i i 3 j k k k k k k k a Ionic strength = 1.0. b Ionic strength = 0.1. References a Nancollas J. 1955 1458; 1956 744; b Jonte and Martin J. Amer. Chem. Soc. 1952,74,2053. c Renier and Martin ibid. 1956,78 1833. d Gimblett and Monk Trans. Faraday SOC. 1954 50 965. e Stock and Davies ibid. 1948 44 856. f Milburn J. Amer. Chem. SOC. 1957 79 537. g. Furman and Garner ibid.1950,72 1785. h Betts Canad. J. Chem. 1955,33 1775. i Gelles and Nancollas Trans. Faraday SOC. 1956 52 680.1 James and Monk ibid. 1950 46 1041. k Staveley and Randall Discuss. Faraday SOC. 1958 26 157. Returning to the case of thallous ion pairs D a v i e ~ ~ ~ found that the hydroxide behaved similarly to the second group above and gave this as additional evidence that covalent bonding is involved in this ion pair. Although this is supported by the impossibly low Bjerrum distance Bell and George13 have pointed out that such calculations involve the macro- scopic E of water which it has been shown is inappropriate in the vicinity of an ion. The relative positions of the maxima in the spectra of thallous hydroxide chloride and bromide are very similar to those of the free ions OH- C1- Br- and this again points to the absence of specific covalent effects.4s The appearance of new lines in the Raman spectrum is strong evidence for covalent-bond formation but no such effect could be found for a nearly saturated (1.5~) solution of thallous hydroxide.40 The new technique of nuclear magnetic resonance may also provide information about solution structure shifts in the nuclear magnetic resonance frequency of various nuclei being ascribed to the changes in their immediate environment.Richards26 has found that the thallium resonance line shifts with increasing concentration of thallous hydroxide owing to the varying concentrations of the species in equilibrium TI+ OH- and TlOH. A detailed study gave K(T1OH) = 5 in agreement with solubility (6.7),13 spectrophotometric (5),48 and kinetic (7)21b estimates.48 Bell and Panckhurst Rec. Trav. chim. 1956 75 725. 4 414 QUARTERLY REVIEWS The type of bonding must still remain an open question although there is strong evidence that the association properties of the thallous ion are determined by other than purely electrostatic forces. There have been several attempts to correlate experimental association constants with properties of the ions concerned. Williams49 has emphasised however that even with a closely related series of cations the association constants may refer to the formation of ion pairs having different degrees of hydration. The alkaline-earth ions constitute an interesting series and if we assume that the ion pairs are formed as a result of purely electro- static interactions the K's should be in the inverse order of the cationic radii i.e.Mg > Ca > Sr > Ba. We have already noted that Davies observed this to be the case with the hydroxides and the organic salts also follow this general order. The K for magnesium oxalate is 20 times that for the barium oxalate or magnesium sulphate and such a large difference may denote actual contact between the magnesium and oxalate ions.43 With acetate and formate ions the entropies of formation of magnesium ion pairs are anomalous (Table 1) and there is some evidence that the anion can displace a water molecule from the hydrated Ca2+ Sr2+ and Ba2+ ions but not from the smaller Mg2+. In the nitrates iodates sulphates and thiosulphates the order of K is completely reversed and the measured values are undoubtedly related to the radii of the hydrated and not the bare cations.The K values are of a magnitude to be expected from Bjerrum's theory the anion competing rather unsuccessfully with the water molecule for positions adjacent to the cations.43 Comparisons of association constants with such properties as ionisation potentials of the metal atoms and lattice energies of the salts have also been made with varying success in closely related systems. The extension of these correlations to bond type should be made with care. There is no doubt that covalent character is an important factor contributing to stability but it is not safe to assume perfect correlation. A number of authors have found50 that log KMX is linearly related to log K m for a large number of systems and such a relation is of use for predicting hitherto unknown K values.Heat and Entropy Changes.-Unfortunately one cannot use the free- energy changes as diagnostic of the type of association without considering the heat AH" and entropy AS" contributions in detail. AGO may be written AGO= -RTlnK= AH"-TAS" . . . . . (17) and in order to gain an insight into the factors which affect the equilibrium it is more useful to regard the free-energy change as being a consequence of the changes in heat and entropy. There are for instance endothermic *@ Williams J. 1952 3770. 6o Eg. Irving and Rossotti Acta Chem. Scand. 1956,10,72. NANCOLLAS ION ASSOCIATION IN AQUEOUS SOLUTION 415 reactions which are made possible by favourable entropy changes also there are exothermic reactions which do not take place because of un- favourable entropy changes.Any comparison which is made of K values at a single temperature involves the assumption that AGO is dominated by the heat term and that changes in ASo may be neglected. It is clearly desirable therefore to obtain AH" and this can be done in two ways (i) studies of the association constant over a range of temperature leading to the calculation of AH" from the van't Hoff isochore d log K/dt = AH"/RT2 and (ii) direct calorimetric methods. The second method is more reliable since there is rather a large uncertainty in heat changes derived from temperature coefficients of equilibrium constants. However there are comparatively few equilibria for which heat and entropy data are available and most of these have been obtained by method (i). Therefore although we must not put too much weight on their exact numerical values they are useful for making com- parisons Measurements over too small a temperature range may lead to large errors in AH".Thus conductivity data51 at 25" and 35" give A S = 22 cal.deg.-l mole-l for PbCl+ which is considerably higher than the more precise calorimetric value in Table 1. The use of a larger tempera- ture range leads to a more reliable A H ". Thus an e.m.f. method52 over a 45' range yields AH" = - 5.6 kcal./mole for the formation of the HS0,- ion in good agreement with Pitzer's calorimetric value,53 - 5-2 kcal./mole. It shouId be emphasised however that the use of a large temperature range does not completely eliminate the errors in AH" and there is a need for more calorimetric measurements. The proton acids form a very large group of compounds for which accurate data are available over a wide range of temperature.Some selected data are given in Table 2. Assuming in eqn. (16) that Wenv is almost entirely electrostatic in origin we may replace it by We1 and assume that as a function of temperature it is inversely proportional to E . ~ ~ It will not be strictly proportional to l owing to dielectric saturation in the neighbourhood of the ion. Thus we expect W to be composed of Wel increasing with temperature and Wnon independent of temperature. For the systems in Table 2 we find that the Kvalues may decrease or increase or may pass through a maximum. However in no case does -kT In Kx (= W ) pass through a maximum or minimum but it rises steadily providing good evidence for this picture of ionic forces.It is possible to interpret qualitatively the temperatures at which the K values pass through a maximum or minimum Extending the arguments to ion-association reac- tions of type (l) we may express E by means of the empirical formula 51 Nancollas J. 1955 1458; 1956 744. 53 Nair and Nancollas J. 1958 4144. 53 Pitzer J. Amer. Chem. Soc. 1937 59 2365. 416 QUARTERLY REVIEWS TABLE 2. Proton transfer reactions Reaction H+ + CH3*CO2- H+ + C2H,C02- H+ + C3H,C02- H+ + H2NCH2*C02- H+ + S042- H+ + S042- H+ + H2P04- H+ + HP04" H+ + PO4% H+ + H,Cit- H+ + H Cit2- H+ + C P d H" AS" K (kcal. (cal.deg.-l (1.mole-l) mole-l) mole-l) Ref. 5.7 x 104 6.0 x 109 7.4 x 104 6-6 x lo4 83 110 130 1.6 x 107 1-34 x 103 8-9 x loll 5.77 x 10* 2-49 x log 0.10 0.17 0.70 5.2 5.6 1.88 -0*80 -3.50 - 1-01 -0.58 0.80 - 10.60 22.1 a 22.9 a 24-4 a 9.2 a 26.3 53 26.7 52 16.0 53 30-3 53 43 53 11.0 b 19.8 b 32 b H,Cit = Citric acid.References a Everett and Wynne Jones Trans. Faraday Soc. 1939 35 1384. b Bates and Pinching J. Amer. Chem. Soc. 1949 71 1274. where v is a temperature characteristic of the solvent and e0 is a constant.32 If Wel were proportional to E - ~ eqn. (16) could then be written In KX = -C(a + eT/">IT . . . . . (19) where Ca = Wnodk and CeTlV = Wel/k. Differentiation of (19) with respect to Tand equation to zero gives the temperature (T") at which the associa- tion constant passes through a minimum or (20) . . . . . } T* = v(l + a/eT*/Y) T* = ~ ( 1 + Wnon/Wel*) where WeI* is the value of We1 at T = T*. The value of K will rise over the whole range of temperature if T* lies below freezing point of the solvent and it is noteworthy that the position of the minimum depends not on the value of (Wnon + Wel) but on their ratio.When Wnon/Wel is small as would be expected in the alkaline-earth hydroxide ion pairs the predominant effect with rise of temperature will be to decrease the mutual potential energy and K will increase. For other ion pairs involving the OH- ion Wnon becomes important Wnon/Wel may be large and the predominant effect will be the tendency for thermal agitation to cause dissociation of the ion pairs.This predicts the right kind of behaviour as is shown in Table 1 ; AH" is positive for the alkaline-earth hydroxides and negative for the others. If the reaction (1) involves the displacement of a water molecule by the associating ion the heat changes may be made up in the following NANCOLLAS ION ASSOCIATION IN AQUEOUS SOLUTION 417 energetically equivalent manner:46 (1) Removal of the hydrated ions Mn+ and Xm- from solution and placing them in the gas phase with loss of their co-ordinated water molecules.(2) Movement of the two ions into adjacent positions. (3) Return of the ion pair into solution with reorientation of water molecules each ion being assumed to occupy a single co-ordination position in the octahedral arrangement of the other ion. By using this thermodynamic cycle calculations from which can be regarded as being accurate only to about 5 5 kcal. mole-1 it is possible to explain a number of observed facts. For instance calculations show that in sodium chloride the formation of ion pairs in which the Na+ ion is adjacent to the C1- ion is energetically most unlikely.46 The interposition of a water molecule between them however makes the ion association exothermic and the degree of association a in a 0.1 iM-solution would be about 0.33 "/o for E = 1 (this value of E being assumed in the vicinity of the ions for E = 79 a = 0.0041 %).Eigen and Wicke2' have obtained 0.2 % from activity data so it would appear that the small amount of ion pair formed in sodium chloride solution consists of Na+(H,O)Cl-. When this approach is applied to nickel sulphate there is a wide discrepancy between the calculated and the experimental degree of association. Since the calculations were made on a model having adjacent nickel and sulphate ions it is concluded that no such ion pairs exist in these solutions.46 The absence of any spectral shift on adding sodium sulphate to nickel sulphate solutions also indicates that any association must take place through water molecules.If the heat changes upon ion-pair formation depended only upon the hydration steps in the above picture then we should expect to find a simple relation between heats of hydration of gaseous ions and z / r where r is the ionic radius. Williams54 has shown that no such simple relation exists either in this case or when the heats are correlated with ionisation potentials. Staveley and his c ~ - w o r k e r s ~ ~ * ~ ~ have measured by direct calorimetry the heats of formation of some bivalent metal ethylenedia- minetetra-acetates and have shown that there are no obvious regularities in the values.AH" values must be associated with electrostatic and covalent interactions between the ions together with additional stabilisation due to the different available bonding orbitals of the cations. For the transition series of metals OrgeI5' has proposed that this additional stabilisation is due to the splitting of the d-level in the electrostatic field of the ligand groups and has shown that this accounts for the main features of the heats of solvation throughout the series. George58 has suggested that this phenomenon is of primary importance in the case of complex ions. Examination of the data in Tables 1 and 2 shows that in most cases for 54 Williams J. Phys. Chem. 1954 58 121. 55 Care and Staveley J. 1956 4571. 66 Staveley and Randall Discuss. Faraduy SOC. 1958 26 157.67 Orgel J. 1952 4756. 58 George Rec. Trav. chim. 1956 75 671. 418 QUARTERLY REVIEWS reactions of type (1) between cations and anions with a resultant reduction in the number of ions in solution and a concomitant neutralisation of charge there is an entropy change favouring ion-pair formation. Frank and EvansK9 have suggested that the ions in solution order water molecules around them to form an “iceberg,” the process being similar to the localised freezing of a liquid. Ion-pair formation tends to break down this “ice- berg” structure and the decreased orientation of solvent molecules leads to a positive entropy change. At first sight this is not always the case since one must take into account with polyatomic ions the loss in con- figurational and rotational entropy which accompanies the association.Also it is important to compare only those reactions which involve the same change in number of solute species and this has been emphasised by a number of woi-kers.60-61b. In eqn. (17) the entropy value depends upon the concentration units used in the calculation of K AS” referring to the molality scale. Gurney32 was the first to recognise this problem of standard states and using the arguments already outlined he defined a quantity called the unitary entropy So’ as the difference between the partial molar entropy So and the absolute value of the partial molal configurational entropy So’ = S o - Rln55-5 =So - 8 An . . . . . . (21) An being the change in number of solute species when the reaction takes place. Williams6* compared AS of a large number of reactions in which com- plexes of the type- MXp are formed where p = 1-4 without correction for An.He found four systems in which AS was negative instead of positive. Bentso applied the above corrections to Williams’s data and showed that in all cases excepting silver thiosulphate the anomalies were removed. In Ag(S203)23- it is possible to explain the negative entropy as a loss of configurational and librational entropy of the S203” ion upon ion-pair formation. There is more regularity in the entropy changes on ion-pair formation than in the heat changes and reactions which have been rather thoroughly studied are those between the anions of weak acids and the hydrogen ion. Some data are included in Table 2 and it can be seen that a AS of about 22 cal. deg.-l mole-l is fairly common for this type.61a The value of AS” for the association of hydrogen and glycine anions is not as positive as +22 cal.deg.-l mole-l.In this case the resulting molecule whilst neutral has a zwitterion structure which will exert strong orientating effects on surrounding water molecules just as did the separated ions. If we can assume that the interaction of the carboxylate group with water molecules is not appreciably altered by the association with the hydrogen ion at the 6 8 Frank and Evans J. Chem. Phys. 1945,13,507. 6o Bent J. Phys. Chem. 1956,60 123. 61 King (u)J. Chem. Educ. 1953,30,71; (6)J. Phys. Chem. 1959,63 1070. NANCOLLAS ION ASSOCIATION IN AQUEOUS SOLUTION 41 9 other end of the molecule then AS" for the association of the glycine and hydrogen ions might be expected to be similar to AS" for reactions such as Hf + Me-NH $ Me*NH,+ and King has shown this to be the case.s10 The other acids in Table 2 for which AS" < + 22 cal.deg.-l mole-1 for the reaction between univalent ions are those in which the functional acidic group is not the only one possessing polarity.The entropy of association becomes appreciably more positive as the negative charge on the anion increases (Table l) and this is due to d>l in the dependence of Soi of polyatomic ions upon zd. In the case of the strong acids which are completely dissociated in solution while associa- tion between the hydrogen ion and the acid anion will be accompanied by a favourable entropy change the heat change must be unfavourable.sla After making corrections for the difference of the symmetry numbers of the acid and conjugated base King has shown61b that the AS" values for the closely related phosphoric and citric acid dissociation reactions are linearly related to Az2 Az2 being the difference between the sum of the squares of the charges on the products and on the reactants.We may write the entropy change in reactions of type (1) in terms of an entropy cycle62 AS3 Mn+(g) +- Xm-(g) --+ MX("-")+(g) I ? T I i 3. I 1 AS1 1 AS I As4 Mn+(aq) XTn-(aq) t- MX,q(n-m)+ - AS" where - AS is the entropy change accompanying the hydration of the gas-phase cation - AS and AS are the corresponding values for the anion and ion pair respectively and AS is the association entropy change in the gas phase. The observed entropy change in solution may thus be written AS" = A S + A S + AS,+ AS4 .. . . . . (22) If AS does not change appreciably for a series of related reactions AS" should be dependent upon the differences in hydration effects AS + AS + AS,. For a given cation and a series of simple anions in which AS4 remains approximately the same we should expect a linear relation between AS" and the hydration entropy of the anion. This has been found to hold approximately in a number of cases.44*62*83 Staveley and have observed for a number of bivalent metal ethylene- diaminetetra-acetates direct linear relation between AS (association) and l/r+. These fall into two groups as is shown in Fig. 1 (curve iv) those metal Evans and Nancollas Trans. Farday SOC. 1953,49 363. Evans and Uri S.E.B. Symposium No. 5 Cambridge 1951. 420 QUARTERLY REVIEWS -AS ( M n + ) hYd PO 40 6 0 8 0 100 120 I 2 0 E 8 0 c 0 n 0-6 I *O I -4 1.8 I/r + FIG.1. Metal-ethylenediaminetetra-acetate complexes. (ii) (iii) and (iv) plots of AS (association) against 1 /r+ (reproduced by permission from Discuss. Faraday SOC. 1958,26 159); (i)plot of AS (association) against - A s h y d ( ~ n + ) . ions with inert-gas structures and (iii) the transition-metal ions. The relationship is no longer linear for the tervalent cations (curve ii). In Fig. 1 AS is also plotted against - ASh,d(Mn+) and it is noteworthy that a single quite reasonable linear relation (line i) is sufficient to include all charge types. George64 has shown more generally that the equation AS = AS + Constant . . . . - (23) satisfactorily interprets the results of a number of cation-anion associa- tion reactions.The value of the constant to be used for different systems 84 George J. Amer. Chem. SOC. 1959 81 55. NANCOLLAS ION ASSOCIATION IN AQUEOUS SOLUTION 42 1 varies from +10 for Few with OH- F- C1- Br- to -40 for CH,-C02- with Ca2+ Srzf Ba2+. The relation is illustrated in Fig. 2 which differs from that of George in that the more precise calorimetric data have been used for lead and cadmium halides.44 It is seen that there is good evidence for the validity of this approximate equation and the implication is that the change of hydration of the ions on ion-pair formation is the important factor in these cases. 60 4 0 0 2 0 0 -10 I I I I I -10 0 20 40 60 d S,+Constant FIG. 2. Plot of AS" against (ASz+ Constant). For the hydroxide ion pairs 8 cal. deg.-l mole-l is added to AS" to allow for loss of rotational entropy on association (reproduced by permission from J .Amer. Chem. Sac. 1959,81,5530 with substitu- tion of calorimetric PbX+ and CdX+ data44). Reaction Constant Reaction Constant -!- Co(NH,) with C1- Br- I- + 2 A HCO with Ca2+ Sr2+ Ba2+ -39 3 Cdw with Cl- Br- 1- - 10 V CH,.CO; with Ca2+ Srw Ba2+ -40 Sn2+ with OH- C1- Br- - l l 0 Few with OH- F C1- Br- +10 0 T1+ with OH- Cl- Br- - 26 3 Pbw with C1- Br- I- - 9 GD C1- with T1+ A@ - 16 The calculation of the terms on the right-hand side of eqn. (22) makes possible the derivation of AS". The entropy change in the gas phase contains both translational and rotational terms : 422 QUARTERLY REVIEWS IA IB IC = + I, - I, - Izz -L + IW - I V Z -I,z - Iy 4- 12 I, = Emi (yi2 -+ zr2) . . . Izu = cm,x,yi .. . where M = Crni In some cases the entropy change due to the loss of free internal S(f.i.r.) = 2.287 (log T + log I - 2 log n) + 89.93 . . . (27) rotation accompanying association must be taken into account :65 where n is the number of indistinguishable positions of the groups with respect to each other and I,. the reduced moment of inertia is given by = 1," [I - f r o (&A2/IA + &B2/IB)] Tn this formula I," is the moment of inertia of the "rotating top" and X,A and &B are direction cosines. Such corrections have been applied in the case of iodate and nitrate ion pairs.67 The entropies of the free ions required for the calculation of AS and AS2 are in many cases known with some certainty68 and there are a number of empirical relationships by means of which estimates may be made when individual data are not available.Since the binding of a water molecule to an ion will occur with a decrease in entropy the partial molar entropies of the various ions will be related to their respective degrees of 05 Herzberg "Infra-red and Rarnan Spectra," Van Nostrand New York 1945 66 Wilson Chern. Rev. 1940 27 17. 87 Nancollas Discuss. Faraduy Sac. 1957 24 108. 88 Latimer "Oxidation Potentials," Prentice-Hall New York. 1952. NANCOLLAS ION ASSOCIATION IN AQUEOUS SOLUTION 423 hydration. A number of workers have noted correlations with functions of the charges and radii of the ions. Powell and LatimersQ have shown that for all the simple monatomic ions the equation 3 2 so = - R l n M + 37 - 270z/r,2. . . . . . (28) holds very well. re the effective radius includes an approximate correction for hydration and is taken as r+ + 2-00 A for cations and r- + 1.00 A for anions.In terms of an electrostatic model Born has proposed a value z2/2er for the potential energy of a conducting sphere in a continuum of dielectric constant and this leads to an entropy contribution related to z2/r and not to z/r2. This has been emphasised by Laidler,’O who pointed out that relations such as eqn. (28) using a corrected radius and referring to SH+’ = 0 have no fundamental validity. Although ionic entropies are usually based on this standard there i! good evidence32 that SH+’ (absolute) = -5.5 cal.deg.-l mole-l and when the entropies were referred to this Laidler found for all monatomic cations the relation following z2/r+ 3 S:abs) = 3 R In M + 10.2 - 1 l*6z2/r+ r+ being the unadjusted ionic radius.This represents a true dependence according to the Born model. The entropies of a large number of complex ions have been treated by Cobble,’l who proposed for inorganic complexes the equation So = 49 - 99zIf(r+ + r-)] + nSH20 . . . . . . (30) f i s a structural factor and n is the number of water molecules displaced from the hydration shell of the cation on ion association. The need for the structural factor is thought to be due to the “open” structure of the complexes. Owing to the non-uniform charge distribution their centre of gravity does not approach the central ion as closely as with a monatomic ion so that the sum of the crystal radii is smaller than the effective radius of the ion. Cobble found that anfvalue of 0.65 made possible the inter- pretation of the entropies of a large number of complex ions.The agree- ment was not very good in all cases but on the other hand the accuracy of the entropy data is not high. The equation is an empirical one but it is also related to a model in which the anions replaae n molecules of water from the hydration shells of the fully co-ordinated aquated cations. Powell and Latimer J. Chem. Phys. 1951 19 1139. 7 0 Laidler Canad. J. Chem. 1956 34 1107. 71 Cobble J. Chem. Phys. 1953 21 1443 1446 1451. 424 QUARTERLY REVIEWS In comparing the entropy changes upon ion association it is important to allow for the various translational and rotational contributions. It is therefore useful to use eqn. (22) to calculate dShyd(MX(n-m)+) from the estimated ASl-,. Data are given in Table 3 ; and in Fig 3 dShyd(MX(n-m)+) has been plotted against z/(r+ + r-).It is seen that for the hydroxide ion pairs the values fall fairly well on a straight line-although there is little ii’c. + r) 0.30 0.32 0 . 3 4 0 . 3 6 n x I W 0 *. m= 7 I I I I 2 6 - 2 4 2 2 - - AqCl 2 0 - 18 - 16 - BqOHr I I I 1 I I b (0 9 0 < 70 5 0 + a r7 =hJ FIG. 3. Plots of dShyd(MX) against l/(r+ + Y-) and of - d&d(MOHz+) against z/(r+ + Y-1. to choose between a z/(r+ + r-) and a .z2/(r+ + r-) relationship. The system UOH3+ refers to experiments at Z = 1 and is therefore not strictly com- parable with the others; the value of d&yd(UOH3+) will probably be in error by a few entropy units. The experimental and interpretative un- certainties for the alkaline-earth hydroxides are likely to be greatest owing to their lower association constants.For ion pairs carrying no residual charge Fig. 3 shows the good linear relation between DShyd(MX) and (r+ + r.-.)-’. The values of dShyd(TlN3) and dShyd(TICNS) in Table 2 are considerably lower than would be expected from this relation. There is some doubt about the So values for the azide and thiocyanate ions but it is unlikely that they will be in error by more than 2-3 cal. deg.-l mole-l. Also entropy contributions due to bending vibrations will account for some of the discrepancy. The latter will undoubtedly be present in the nitrate and iodate ion pairs but it is possible that in these cases the entropy is already over-corrected in the term for free internal rota tion. NANCOLLAS ION ASSOCIATION IN AQUEOUS SOLUTION 425 It is interesting to consider the way in which the entropy of hydration of the ion pair is influenced by the values of Wenv and Wnon.If we compare BaOH+ and PbCl+ the values of (r+ + r-) are not very different but whereas in BaOH+ the only short-range forces will be those of repulsion in PbClf Wnon will contribute leading to a greater neutralisation of charge. This is reflected by the considerably smaller value of - d S h r d (PbCl+) in Table 3. TABLE 3. Thermodynamic properties." Ion pair CaOH+ SrOH+ BaOH+ VOH2+ FeOH2+ CrOH2+ TlOH TlCl TlBr TlNO TlCNS TlN AgCl AgIO3 MgS04 CaSO MnSO CoSO NiS0 ZnSO CH,-C02Mg+ CH,CO $a+ UOH~+ 53.8 56.0 58.4 53.7 53.9 53.8 58.5 58.8 60.9 62-8 80-4 69.8 66.7 58.9 82.2 68.2 69-5 70.1 70.2 70.2 70-3 65.0 66.5 10.8 7.6 8.8 45.5 44 52 5.1 - 1.7 - 4.2 - 1.0 - 6.4 - 2.7 - 6 20.3 26.1 16.1 22-6 16.6 21-7 24.4 0.6 8.7 So - AShyd (r++r-)-' (MX(n-">+) (MX(n-">+) - 4.8 58.6 0.397 - 4.3 60.3 0.376 9.3 49.1 0.347 -22 75.7 0.457 -23.2 77.1 0-469 - 16.4 70.2 0.462 -28.5 8 6.0 0.4 1 3 33.0 25.8 0.352 41-9 19.0 0.308 45.5 17.3 0-295 64.4 16-0 0.289 60 9.8 ca.0.30 59.7 7-0 ca. 0.32 36.5 22-4 0.326 66-0 16.2 ca. 0.29 2.0 66.2 0.26 7-0 62.5 0.24 6-7 63-4 0.25 - 1 71 0.25 3.0 67.3 0.25 3 67 0.256 - 1.0 66.1 ca. 0.35 22.2 44.3 ca. 0.31 * All S values are in terms of cal. deg.-l mole-'. The entropy of association can yield qualitative information about the degree of hydration of the ions in the ion pair. If the ion-pair formation takes place without the elimination of water molecules of hydration then the entropy change will be expected to be less positive.Considering only the charges on the ions it would be expected that AS" for the formation of LaFe(CN) and LaCo(CN) would be considerably more positive than the values in Table 1. We have already seen that there is good evidence that these ion pairs involve the hydrated lanthanum ion and the smaller 426 QUARTERLY REVIEWS positive value of AS" substantiates this view. The water molecules in the first sphere of co-ordination would not gain freedom when the ion pair is formed. The measurement of the entropy changes accompanying ion-association reactions has important implications in reaction kinetics of ionic processes in aqueous solution. We can write the velocity constant of ionic reactions of the type. An+ + Bm- = A(n-114- + B(M-1)- as and enquire how AS* the difference in entropy between the activated complex and the reactants varies with the ionic charges m and n.Since the transition state is an association complex this entropy change (AS*) will vary in the same way as AS (association) for the same ions. Confirma- tion of this is obtained by comparing AS* for the reaction Few + 02H- = Fe2+ + 02H k = v exp ( AS*/R) exp ( AH*/RT) with AS (association) for the equilibrium Few + OzH- + Fe02H2+ These values are 52 and 49 cal. deg? mole-1 respectively.62 NH$ + OCl- = [NH,OCl]* = NH2C1 + H,O Similar observations have been made by for reactions such as An empirical relation gives the translational entropy change for the forma- tion of a single molecule by the combination of two non-electrolyte molecules as -14 cal.deg.-l mole-l. Thus if we imagine the activated complex as being formed from NH (26 cal.deg.-l mole-l) and HOCl (31 cal.deg.-l mole-l) its estimated entropy is 26 + 31 - 14 = 43.This compares very well with the experimental entropy value 44 cal.deg.-l mole-l. An attempt has been made throughout this Review to present the physi- cochemical principles involved in studies of ion-pair formation. Clearly the determination of reliable thermodynamic data is very desirable and so far as the heats of formation are concerned a calorimetric approach seems to be promising. Entropy values are important in that they give information about the structures of ion-solvent and ion pair-solvent complexes. However accurate the experimental data these entropies will always involve the assumptions implicit in the calculation of AGO. The Reviewer thanks Dr. J. C. Speakman for reading this manuscript. 7 2 Powell J. Phys. Chem. 1954 58 528.

ISSN:0009-2681    

DOI:10.1039/QR9601400402

出版商:RSC    

年代:1960

数据来源: RSC

摘要:

ELECTRON RESONANCE IN CRYSTALLINE TRANSITION-METAL COMPOUNDS By A. CARRINGTON and H. C. LONGUET-HIGGINS (DEPARTMENT OF THEORETICAL CHEMISTRY UNIVERSITY OF CAMBRIDGE) 1. Introduction WHEN a solid or a liquid specimen is placed between the poles of a permanent magnet and subjected to microwave radiation power may be absorbed by the specimen at a particular frequency or set of frequencies which depend on the strength of the permanent field. If the magnetic field is about 10,OOO gauss and resonant absorption occurs at frequencies of the order of 1 cm.-l then it is safe to attribute the resonance to unpaired electrons. This phenomenon has been variously termed “paramagnetic resonance” “electron magnetic resonance” and “electron spin resonance”. The first name emphasises that pure diamagnetic substances exhibit no such resonance; the second that it is the magnetic component of the micro- wave field which induces the observed transitions; and the third that it is the spins of the unpaired electrons which make them susceptible to the applied magnetic field.As a rule electron-resonance measurements are made in dilute solution solid or liquid to minimise interference between neighbouring para- magnetic units. Though organic free radicals often exhibit electron resonance in liquid solution it is relatively unusual for transition-metal compounds to do so. To obtain resonance spectra from such compounds it is usually necessary to grow and study dilute single crystals. It is also convenient to be able to work at very low temperatures where line- broadening effects are reduced.Crystal studies have the compensating advantage that one can observe how the resonance spectrum depends on the orientation of the crystal in the external magnetic field; if the crystal structure is known the orbitals of the unpaired electrons may thus be identified; if not electron resonance may provide a useful supplement to crystallographic studies. Finally if magnetic nuclei are present it may be possible to observe “hyperfine” structure in the electron-resonance spectrum and from this to determine how far the magnetic electrons are delocalised into the orbitals of neighbouring atoms. For these reasons the study of inorganic crystals by electron resonance is a tool of considerable utility for the student of molecular structure. It is therefore of advantage for the inorganic chemist to know something of the theory and practical applications of the technique.There are many good reviews of the subject some of which go deeply into the basic theory,l.2 while others give comprehensive accounts of the applications of electron- B. Bleaney and K. W. H. Stevens Reports Progr. Phys. 1953,16,108. K. D. Bowers and J. Owen Reports Progr. Phys. 1955,18,304. 427 428 QUARTERLY REVIEWS resonance spectros~opy.~~~ In this article we have chosen to expound the theory by the detailed consideration of one or two representative examples in the hope that the reader may then feel able to tackle more confidently the more advanced papers. We shall find it necessary to use the methods of quantum mechanics and to quote general results without proof; but we shall not assume any previous knowledge of the theory of electron reso- nance.A few words on notation may be helpful at this point as it is not in all respects identical with that used in general chemical work. First as is customary we shall distinguish vectors from scalars by printing the former in bold type and the latter in italic type. Thus H the magnetic field is a vector whereas H, its z-component is a scalar quantity. Secondly we shall place a circumflex over every symbol (whether a vector or a scalar) which stands for an operator. Thus and S^(which are both vectors) denote the angular momentum operators associated with the orbital motion and the spin of an electron or group of electrons measured in units of h/272 while xz and s” are the (scalar) z-components of these operators. By contrast L S ML and Ms are quantum numbers used to label the various electronic states.Thus the symbol IL,ML) denotes a state for which 6 has the eigenvalue L(L + 1) and has the eigenvalue ML (which can take any one of the 2L + 1 values -L . . . +L). Likewise IML,M~) is an eigenstate of both and $ with eigenvalues ML and Ms respectively (the numerical values of L and S having been previously specified). Finally the Bohr magneton /3 has the value eh/4nrnc where -e and m are the charge and mass of the electron c being the velocity of light. 2. General theory An electron interacts with an external magnetic field in two ways by virtue of its charge and by virtue of its spin. A charged particle revolving round a fixed point is equivalent to a current flowing in a loop; likewise an electron of charge -e with angular momentum L interacts with a magnetic field H by virtue of its orbital motion.This interaction is -H.& where is called the orbital magnetic moment and is equal to -@ where /I is the Bohr magneton. In addition the spin Kconfers on the electron a magnetic moment and this also interacts with the external magnetic field H. The total inter- action is therefore - H.(FL + c) = pH.(% + 2s) A A A ps = -2ps D. J. E. Ingram “Spectroscopy at Radio and Microwave Frequencies” Butter- J. W. Orton Reports Progr. Phys. 1959,22,204. woI;ths Scientific Publications London 1955. CARRINGTON AND LONGUET-HIGGINS ELECTRON RESONANCE IN CRYSTALS 429 The simplest possible situation is that of a single electron in an s orbital as in the hydrogen atom. Here there is no orbital angular momentum so the interaction is simply 2PH.Kwhich becomes 2/3Hzz if the z direction is chosen as that of the permanent field.The field H thus quantises the z component of the spin whose eigenvalues Ms are well known to be st; and - for a single electron. The magnetic energies of the two correspond- ing levels are therefore PHZ -pH and the separation between them may be written as gPH where the quantity g is called the spectroscopic splitting factor and has the value 2 (more precisely 2.0023). We have here dropped the suffix from H to stress the obvious fact that the Zeeman splitting is independent of the direction of the applied field; we may say that g is isotropic. In crystals however isotropic g values are the exception rather than the rule as we shall see in a moment. Having split the spin levels apart by the field H, we may now induce transitions between them by applying an oscillating magnetic field H,' cos 2nvt in a perpendicular direction.This is because the resulting perturbation 2/IH,'$ cos 2nvt involves S, which connects i.e. has matrix elements between eigenstates of s^ differing in eigenvalue Ms by 1. The selection rule is therefore in this example AMS = & 1 and the resonance condition is of course hv = gPH The frequency of resonance is thus proportional to the magnetic field and vice versa. In practice it is more convenient to vary the magnetic field than to vary the frequency of the microwave source so the spectrum is usually presented as a relation between absorption and magnetic field for a given microwave frequency; for instance the width of a resonance line is normally quoted in gauss.Before passing on to consider the effects of orbital motion which are of crucial importance in crystal studies we may interpose a word about hyperfine structure. Atomic hydrogen actually shows two electron- resonance lines and the explanation of this is that the proton like the electron has a spin of magnitude $ which confers on it a magnetic moment of about Bohr magneton. This interacts both with the external field and with the magnetic field of the electron the latter interaction being a good deal larger than the former. If the proton spin is denoted b y x the total magnetic interaction takes the form g/3H.Z + &.r+ BH.2 A 430 QUARTERLY REVIEWS where B depends on the magnetic moment of the proton and A depends also on the probability of the electron's being at the position of the proton.Again taking H to be in the z direction we find that if H>>A/gP the eigenstates of this operator are states for which Ms and MI are both good quantum numbers having the alternative values + and -6. Fig. 1 shows the resulting set of four energy levels (not to scale). Neglecting the term 9pffz+ !g A+ 6 H z FIG. 1. Nuclear hyperfine energy levels of a 1s hydrogen atom in a magnetic field. BH in comparison with A we see that the selection rules ~ M s = kl AM = 0 allow the two transitions indicated with energies gpH -+ $A. The two members of the hyperfine doublet are therefore separated by an amomt A on the energy scale. Although such a simple discussion is seldom possible it may be said that in general the appearance of hyperfine structure is clear evidence that the unpaired electron (or electrons) is interacting with one or more magnetic nuclei either by direct contact as here or through the magnetic field associated with its orbital motion.3. Spin-orbit coupling When an atom is placed in a chemical environment the orbital motion of its electrons is strongly perturbed. To take an example the isolated Ti3+ ion has a 2D ground term ( L = 2 S = 3) which has a five-fold orbital degeneracy (-L<ML< +L); but when the ion is placed in a chemical environment this degeneracy is partly lifted. Were the degeneracy com- pletely lifted so that finally all of the five 3d orbitals had different energies it would be found that the expectation value of L^ would vanish for each of them and we should then say that the orbital angular momen- tum was completely quenched.Now there is a theorem due to Jahn and Teller5 which states that if a symmetrical non-linear molecule is in an orbitally degenerate state then it will always distort in such a way as to lift the degeneracy. Why then is the orbital angular momentum of an atom or ion not always completely quenched in a chemical environment? The answer is to be found in the phenomenon of spin-orbit coupling. Though this effect is usually insignificant in chemistry it is vital to an H. A. Jahn and E. Teller Proc. Roy. Soc. 1937 A 161,220. CARRINGTON AND LONGUET-HIGGINS ELECTRON RESONANCE IN CRYSTALS 43 1 understanding of electron resonance. In simple terms an electron tends to align its spin anti-parallel to its orbital angular momentum; hence if it has orbital angular momentum this tends to be kept going by being weakly coupled to the spin; and if it has spin this tends to generate orbital angular momentum.There is therefore a competition between the quenching effect of the ligands-the “crystal field’’-and the sustaining effect of the spin-orbit coupling. Were it not for spin-orbit coupling the orbital magnetic moment would be of no consequence in electron reso- nance and we should always observe without much interest an isotropic g value of 2. In the free atom or ion the most important effect of spin-orbit coupling is to split apart the different members of a particular “Russell-Saunders term’’ such as 2D. Thus in Ti3+ the level 2D3/2 with 2 and antiparallel is slightly more stable than 2D5/2 with z and3 parallel.In considering only the levels which arise from a particular Russell-Saunders term it is sufficient to take the spin-orbit interaction to be of the form &.g where the parameter his an energy which depends on the number and nature of the outer electrons (for example n 3d electrons) and on the values of L and S for the term in question. h also depends very strongly on the atomic number of the ion increasing roughly as the cube of the atomic number. Values of h for most of the ions of chemical interest are to be found in Condon and Shortley’s “Theory of Atomic Spectra”.6 They are mostly in the range 100-1,OOO cm.-l. 4. The caleulation of g values When the ion is placed in a chemical environment the orbital degeneracy characteristic of the free ion will as already remarked be partially lifted; but nevertheless the valency state of the bound ion may approximate quite closely to a mixture of states of the free ion all of which arise from the same Russell-Saunders term.Thus in an octahedral environment the ground state of the Ti3+ ion (2T2g) and its first excited state (2Eg) both arise from the 2D term of the free ion other such terms probably making insignificant contributions. If the crystal field is very strong this is no longer quite true but for the present exposition we shall assume it to be the case and will therefore take the spin-orbit interaction to be of the same form as it is in the related Russell-Saunders term of the free ion. In order to make the situation clearer we shall now work through two examples to show how the observed resonance spectra are related to the spin-orbit coupling and crystal-field splitting parameters.(i) The Ti% ion in a tetrahedral environment. It is well known that if a transition-metal ion is placed in an environment with octahedral or tetra- * E. U. Condon and G. H. Shortley “Theory of Atomic Spectra ” Cambridge University Press 1951 p. 197 (see also D. S. McClure Solid State Phys. 1959 9 428). 432 QUARTERLY REVIEWS hedral symmetry the d orbitals are split into a group of three and a group of two by the electrostatic crystal field.' If the ion is tetrahedrally co- ordinated the orbital doublet lies below the orbital triplet as shown in Fig. 2 where we have indicated how a small tetragonal distortion- / I I JA \ i Free Tetrahedral Tetragonal ion co-ordn. distortion FIG. 2.Splitting of the orbitals of the Ti3+ ion in a tetrahedralfield with tetragonal distortion. actually a slight flattening of the tetrahedron-further splits the orbital doublet. (We shall not need to discuss how the upper triplet is split by this additional perturbation.) We have also indicated the explicit forms of the orbitals in terms of the complex 3d orbitals; IML) represents a 3d orbital with ML units of angular momentum around the z axis which is taken to be the axis along which the tetrahedron is flattened. Some such distortion is to be expected from the Jahn-Teller theorem and we are supposing that it is such as to make the orbital (0) the more stable.* We now fix attention on the lowest orbital and note that it contains a single electron in the ground state of the complex. If spin-orbit coupling is ignored the ground state may be represented by the pair of symbols lo,&) lo,-&) where the pair of numbers in each "ket" represent the values of ML and Ms respectively i.e.the eigenvalues of L and S,. When spin-orbit coupling is taken into account however these expressions are modified slightly by the admixture of small amounts of states with different values of ML and MS. n n J. S. Griffith and L. E. Orgel Quart. Rev. 1957 11 381. * The orbitals 10) and \It( 12) + I -2)) are actually the orbitals more commonly known as dz2 and dzs-yi. Of the other three orbitals d4(\2) - I -2>) is the orbital dzY and I 1 ) and I - 1 > are the complex combinations z/+(d, f idyz). But as we are dealing with angular momentum it is not convenient to use the more familiar real forms for all five d orbitals.CARRINGTON AND LONGUET-HIGGINS ELECTRON RESONANCE IN CRYSTALS 433 To calculate the admixing coefficients we use first-order perturbation theory. According to this theory the perturbation hz% changes the state lo,$) into where the sum is over all other states IML,Ms) and the denominator is the difference in energy between the unperturbed state and the state which is being mixed with it. We must therefore have a look at the values of the matrix elements (ML,Ms 1AL.S @,T) and (ML,Ms 1hL.S lo,-+). First we note that the perturbation hc.g may be expanded in the form A n A- nn + cg + C$) = +X(L+S- + LS+) +AL;g n n n A A n where ,!, = L & iL and S = S & is Secondly we make use of standard formula for the matrix elements in the last expression.It is a standard result that the only matrix elements of these types which do not vanish are the following (ML + 1 IG IML> (MY + 1 I?+ IMS) (ML - 1 IML> = ~ / [ L ( L + 1) - ML(ML - 111 (MS - 1 IS- IMS) = 1/[L(L + 1) - ML(ML + I)] 2/[S(S + 1) - M s W s + I)] .\/[S(S + 1) - MS(MS - l)] = n = <ML[L?IML) = ML < M S / X I M S ) = MS ~n From these formula we can see that the term &AL+S- mixes lo,&> with a little bit of [ 1,-+) but that &AL^_g cannot mix lo,&) with anything because MS cannot have the value 3/2 and AL,S is also ineffective because when operates on lo,&) it multiplies it by the eigenvalue ML = 0. The perturbed form of 10,;) is therefore nn lo,+,> + c [I,-&) = I"&") say; 434 QUARTERLY REVIEWS the admixing coefficient c is given by n A ( I + IK.qo,a> - - g m - * IL+S- lo,+> c = E(O,+) - E( l + ) - A = - 0(/2A)<l I& 10) <-& Is? 19) = - ( h / 2 4 1 / ( 2 x 3 - 1 x O ) l / ( & x ; - (-4) x +) = - 1/(3/2)h/A In a precisely similar way we can demonstrate that the unperturbed state I O,-+) becomes lo,-*> + c I-LO = I“-$”> say when the effect of spin-orbit coupling is taken into account.We now observe that the states called I”S”> and I”-+”> are strictly degenerate in the absence of a magnetic field and constitute what is known as a Kramers doublet.s A magnetic field will however resolve this degeneracy and we now determine the splitting produced by fields in the z or x direction. A field H produces a perturbation /3Hz(C + 25;‘,) and we have to find the matrix elements of this between the states I”&&”) in order to determine how they are mixed and to find the energy separation between the resulting pair of levels.First (c + 2 z 2 ) has no matrix element between I”4”} and l”-$“} because the states IML,MS) are eigenstates of both 6 and g and there are no such states common to I”&”) and l”-~”). Hence only the diagonal elements are non-zero and these are Hence if the field is applied in the z direction we obtain just the same split- ting namely 2PH, as we should if only the electron spin interacted with the field. We describe this situation by saying that 811 = 2 where the subscript II means parallel to the z axis. The situation is different however if the field is applied in the x direc- tion. Now the perturbation is PH,(L^ + 2$) which may be written PH,(+L+ + +c + ,!$ + K).This operator has no diagonal elements within the pair of states I”-+-”) since each member of the Kramers doublet comprises two terms which differ in both ML and MS. The off- H. A. Kramers “Quantum Mechanics” North-Holland Publ. Co. Amsterdam 1957 p. 384. n CARRINGTON AND LONGUET-HIGGINS ELECTRON RESONANCE IN CRYSTALS 435 diagonal elements on the other hand do not vanish since each term in ("ij") is connected with one (or both) in I"-+") through one of the operators L+ L- S+ or S-. The two off-diagonal terms ("&" I#3Hz(Tz + 2 g ) I"-&") and ("-4" I/3Hz(c + 2 s ) I"+") A h A A are both equal to P f w O & IKlO,-&> + &C(O,4 IG l-1,8> + W-1,-& IL lo,-&>] = #3HX[I 3- &c(d6 -1- d6)] = /3Hz(1 - 3WA) The two states therefore split into their sum and their difference with a separation 2#3Hx(1 - 3h/A) and we describe this by saying that gl = 2(1 - 3h/A).Abragam and Prycee showed that the observable behaviour of this type of system could be described by what they called a "spin hamiltonian". The two splittings we have calculated and the splitting which occurs if the field is applied in an arbitrary direction are precisely what one would obtain if one ignored the orbital angular momentum and replaced its effect by an anisotropic coupling between the electron spin and the external magnetic field of the form l3H.g.g = Pg,,HzS^ + jSg,(H& 3- Physically the orbital angular momentum induced by the spin makes the latter easier to orient in some directions than in others and this effect is what is represented by the spin hamiltonian.It may be noted that the axes of the g tensor must coincide with the symmetry axes of the system; thus we can use the observed anisotropy of the electron-resonance spectrum to determine the orientation of a paramagnetic unit in a crystalline lattice. An outstanding example of this was the determination of the orientations of the four haem groups in the haemoglobin crystal,1° discussed below. We work out this example as well as the previous one because the directional effects are much more pronounced. The situation is as illustrated in Fig. 3. This time we consider the Jahn-Teller distortion to be a slight compression of the octahedron in the direction of one of its three-fold axes which we take as the z axis. Such a distortion occurs for instance in the compound CsTi(SO,), in which the site symmetry of the octahedrally co-ordinated Ti3+ ion is trigonal.ll In terms of the eigenstates of L, namely /AIL) the orbitals then take the forms indicated in Fig.3 the lowest being lo} which is axially symmetrical about the z axis. (ii) The Ti- ion in an octahedral environment. A A. Abragam and M. H. L. Pryce Proc. Roj. SOC. 1951 A 205 135. lo J. Bennett J. F. Gibson and D. J. E. Ingram Proc. Roy. SOC. 1957 A 240 67. l1 B. Bleaney Proc. Phys. Soc. 1950 A 63 407. 436 QUARTERLY REVIEWS As in the previous example the spin-orbit interaction mixes lo,+> with states involving I 1,-$) and mixes lo,-&) with states containing I-l,&). Now since 8 is much smaller than A the mixing of 10) with the two states immediately above it will be much larger than with the two Free Octahedral Trigonal ion f i e l d distort ion FIG.3. Splitting of the orbitals of the Ti3+ ion in an octahedralfield with a trigonal distortion. states at the top of the diagram. The result is that the perturbed levels take the approximate forms 1 / ( 1 - c2) lo,+> + c [ d 5 1L-i) + 48 l-2,-$>1 1 / ( 1 - c2) lo,-+> + C [ d & 1 - 1 9 8 ) - 1 / $ 1 2 9 &)I and where c is much larger than in the previous example. These states are split to different extents by magnetic fields in the z and the x direction; the corresponding g-values are found to be (8 - 3x12) gL = 2/[(8 + X/2)2 + 2 P ] + In general if the t shell in a distorted octahedral or tetrahedral complex is incompletely filled (but not if it is exactly half-filled) the observed split- ting is much more sensitive to direction than if the odd electron (or hole) is in an e orbital.CARRINGTON AND LONGUET-HIGGINS ELECTRON RESONANCE IN CRYSTALS 437 5. Zero-field splittings In the previous section we saw that for an ion with one d electron the degeneracy of the 2D term is largely lifted by the combined effects of the crystal field and spin-orbit coupling but that there remains a two-fold degeneracy in each level which is only removed by an external magnetic field. Kramers* showed that this sort of degeneracy always occurs if the number of electrons is odd because in the absence of a magnetic field a simultaneous reversal of the spin and orbital angular momenta leaves the energy unchanged. When however there is an even number of unpaired electrons the spin degeneracy of a triplet or quintet level may be removed entirely by the crystal field alone.We will now illustrate this effect known as a “zero-field splitting” by the examples of the ferrate ion FeO!- the octahedrally co-ordinated V3+ ion and the Mn2+ ion in turn. It is possible to grow single crystals of K,CrO containing measurable amounts of the Fe0;- ion and the spectrum of such a crystal shows not a single resonance but two resonance lines when the external magnetic field is applied in the crystallographic ab or ac planes.12 Furthermore the position of the resonance lines depends on the direction of the field relative to the crystal axes. We now discuss the interpretation of these results. Regarding the FeOt- ion as a very strongly perturbed Fe6+ ion we anticipate that its ground state has the electronic configuration 3d2.The ion is approximately tetrahedral and the two 3d electrons will occupy the two orbitals of symmetry E giving a 3A2 state. If the tetrahedral symmetry were perfect the spin degeneracy of this triplet would be maintained but an orthorhombic distortion is observed which resolves the triplet into three components with electronic spin parallel to the x y and z axes of the distortion. (These symmetry arguments make it plausible that the splitting can occur; why it does occur we shall not discuss here.) Two of the axes of the distortion do not coincide with those of the crystal but we shall indicate how the electron-resonance spectrum is determined by the directions in which the individual ions distort. When discussing the Ti3+ ion we saw that it was possible to describe the behaviour of the levels by a spin hamiltonian of the form We now seek a corresponding expression for the ferrate ion.In this we are guided by two considerations. First since the effect of an external magnetic field would necessarily be isotropic if the ion were perfectly tetrahedral we represent this interaction by the isotropic expression Secondly we must have some anisotropic terms to represent the zero-field l2 A. Carrington D. J. E. Ingram K. A. K. Lott D. S. Schonland and M. C. R. Symons Proc. Roy. SOC. 1960 A 254 101. 438 QUARTERLY REVIEWS splitting and these cannot involve H. The simplest second-order expression which has the same symmetry as the orthorhombic field is D 2 + D g + D,? We therefore adopt as our spin hamiltonian the form where D, D, and D are constants with the dimensions of energy; their sum is taken to be zero.The problem is now to see what sort of electron- resonance spectrum emerges from such a spin hamiltonian. What we must first do is to determine the possible spin states of the system and their energies. These are the eigenstates and eigenvalues of the spin hamiltonian. We therefore begin by determining the matrix elements of &' between the three levels Il) lo) and 1-1) which are defined as the eigenstates of S with the eigenvalues Ms = 1 0 -1 respectively. To do this we require the matrix elements of the operators S, S, S, Sz Si and 2;. We have already given expressions for determining the matrix elements of the operators S+ and S- and the complete matrices are n A A ~ A A A n n 11) 10) 1-1) d2 0 0 0 11) 10) 1-1) 0 A 0 0 A n n n 6 A - A From the relations S+ = S + is, S- = S - is, we see that S = +(g + c) and = - $i(?+ - s?>.Hence we can write down the matrices for S, S, and g, remembering that Il> lo) and I-l} are eigen- functions of Zz. These matrices are A - 1 / 4 2 0 1 / 4 2 0 1 / 4 2 ] [ 1 / 4 2 0 $ = $ = s ^ = 0 1 0 - i / d 2 0 i / 4 2 0 - i / 4 2 i I 4 2 0 1 0 CARRINGTON AND LONGUET-HIGGINS ELECTRON RESONANCE IN CRYSTALS 439 The matrix elements of the operators @ $ and @ are obtained by simply squaring these matrices; the rules for matrix multiplication give We are now tonian. For 1-4 0 -k 1 0 0 0 0 [ ! 0 1 g j = in a position to find the matrix elements of the spin hamil- instance consider the matrix element (1 $? 11).This is . . . expanded and evaluated as follows (1 l a l > = gPWx(1 ICIl> + H,(1 Ell> + Hz(1 Is^. l1>1 + DX(113 11) + D,(11@ 11) + D,<l I$p> = gPfH,.O + H,.O + H z . l ] + Dx.i + D,.& + D,.1 = gSHz + W X 4- D,) + Dz The other matrix elements are determined in precisely the same manner. The equation for the eigenvalues E and the eigenstates c1 11) + c, 10) + c - ~ I - 1 ) therefore is sBHz +MDx+D,) +a -E d*gP(Hx -ifGI Q(Dz -4ll c1 o= d4gP(Hx+iH,) D,+D,-E d4gS(Hx-iHJ] [co ] [ 8(DT,-DA -\/*gPWx+iH,) 9 -gPHz+1E-(Dx+D,)+Dz-E c-1 In principle we can now solve for the energies of the three levels but in practice it is better to examine the energies under different limiting condi- tions. First we consider the energy levels when no external field is applied.Putting H = H = H = 0 and expanding the secular determinant as a cubic equation in E we find that the three energies (and the correspond- ing states) are D + D 0 Dz + Dx dHI0 + 1- 1 ) ) D + Dl/ Z/S(Il> - I- 1)) Hence we see that the three spin levels have different energies even in zero magnetic field. We have determined these energies in terms of constants which as we shall see shortly are obtained from the behaviour of the spectrum in the presence of an external magnetic field. Let us now examine the effect of a magnetic field H applied parallel to . 440 QUARTERLY REVIEWS the z axis ( H = Hv = 0). Returning to the secular equation we see that the energy of 10) is unaffected by H whereas the other two energies are modified and are obtained by solving a quadratic equation in E.The roots of this equation are Provided the value of the term 4(gj3H,)2 is large compared with that of ( D - DJ2 the energies are to a very good approximation + gBHz + D + ;(Ox + 0,) + ( D x - D,>"~SPHZ -gBHz + D + B P x + DY) - (Dz - D,)"8gPHz (This approximation is equivalent to that of second-order perturb c? t' ion theory.) Making use of the fact that D + D + D = 0 and neglecting the term (D - DJ2/8gPH we see that the energies and corresponding eigenstates are for large values of H, Now as indicated in section 2 the selection rule for transitions induced by an oscillating field perpendicular to the z axis is AMS = & 1. Hence the allowed transitions are for large H, 10) +-+ 11) and 10) t+ 1-1) and their energies are respectively gPH + :Dz and gPH - ill,.We now see why two absorption lines are observed for certain orientations of the crystal. (Actually the z axis turns out to be parallel to the crystallo- graphic a axis for every FeOi- ion in the crystal.) The general way in which the level separation varies with H is shown in Fig. 4. We note that when the field H is strong compared with the separation D - D (measured in gauss) the two resonances are separated by an amount 30,. It is interesting that transitions between the levels whose asymptotic form is 11) and I - 1) are induced by an oscillating field HI cos 2nvt parallel to the permanent field H if the latter is not too strong. This is because the matrix element of s^ between the states dQ( 11) rt 1-1)) is A A *(l IS 11) - i{-l Is 1-1) = &(l-(-l)) = 1 # 0 If however the permanent field becomes very strong these transitions disappear since S has no matrix element between the limiting forms of these states.Our second example is the octahedrally co-ordinated V3+ ion whose ground state would be one in which two electrons occupied tZg orbitals A CARRINGTON AND LONGUET-HIGGINS ELECTRON RESONANCE IN CRYSTALS 44 1 with parallel spins. We say “would be” because in fact the resulting level 3Tlg is orbitally degenerate and distorts in the manner indicated in Fig. 3 giving a complex of symmetry Dsd. The 3T,g state therefore splits into a 3A2 and a 3E component of which the former is the ground state (see Fig. 5).13 E FIG. 4. Magnetic field dependence of levels of a spin triplet exhibiting a zero-field splitting. Ms= t l Ms= 0 \ q- Octahedral Trigonal Spln orbit field distortion coupling FIG.5. Splitting of the ground state of an octahedrally co-ordinated V3+ ion. Spin-orbit coupling further splits the 3A2 level into a non-degenerate level with Ms = 0 and a doubly degenerate one with Ms = 1 where the axis of quantisation of zz is the axis of trigonal distortion. This zero-field splitting is a good deal larger (about 8 cm.-l)14 than that in FeOf (D = 0.0504 cm.-l D = 0.0180 cm.-l). The reason is that the spin-orbit l3 M. H. L. Pryce and W. A. Runciman Discuss. Faraday SOC. 1958,26,34. l4 G . M. Zverev and A. M. Prokorov J . Exp. Theor. Phys. 1958 7 707 (transl. of Soviet Phys. English pagination). 442 QUARTERLY REVIEWS coupling operates by contaminating the 3A2 sublevels with the 3E sublevels and this effect is large because of the small separation between the 3A2 and the 3E states; in FeOq2- by contrast there are no such low-lying excited states.The large zero-field splitting in V3+ makes impracticable the observation of electron resonance involving transitions with AMS = 5 I ; it is possible however to observe weak transitions between the levels with Ms = &l under certain rather special conditions.14 Our third example of a zero-field splitting is that of the 6S (d5) ground state of the Mn2+ ion in an octahedral field. We shall not go into details but the splitting arises essentially in the following way. In the free ion spin-orbit coupling can mix together states with the same value of J ; hence the 6S5,2 state is slightly mixed with 4P5/2 which in turn is con- taminated with 2D5/2. But a 2D state is split by an octahedral field for the same reason that the degeneracy of the d orbitals is lifted in such a field.This splitting reacts back on the 6S5/2 state and this state consequently splits into a doublet and a quartet. The quartet is often found to split further into two doublets as the result of distortion. The effect of an external magnetic field on these levels is shown in Fig. 6. The selection Free ion Crystal field Magnetic f icld FIG. 6. Zero-field splitting in the Mn2+ ion leading to fine structure in the electron- ?esonance spectrum. rules permit the transit.ions indicated and the spectrum consists of five lines. This "fine structure" is further complicated by hyperfine structure due to interaction with the magnetic moment of the Mn n~c1eus.l~ 6. Rare-earth ions The tervalent rare-earth ions have the electron configuration 1 s22s22p6- 3s23pe3d104s24p64d104fn5s25p6 where It runs from 0 (Law) to 14 (Lu"+) and the paramagnetism thus arises from unpaired 4f electrons.Although B. Bleaney and D. J. E. Ingram Proc. Roy. SOC. 1951 A 205 336. CARRINGTON AND LONGUET-HIGGINS ELECTRON RESONANCE IN CRYSTALS 443 the same perturbations as before must be introduced in order to explain the electron-resonance properties of these ions the relative importance of these perturbations is now quite different. The 4f electrons are shielded by the outer electrons ; consequently the crystal-field splittings are rather small and spin-orbit coupling is the dominant effect reducing the degener- acy of the ground state of the free ion. It will be remembered that to obtain the energy levels of the Ti3+ ion (which has one 3d electron) we first con- sidered the effect of the crystal field in splitting the orbital levels apart and only afterwards examined the effect of spin-orbit coupling in mixing levels with different spin and orbital quantum numbers.The situation with the rare-earth ions is the reverse and as an example we consider the theory of the Ce3+ ion (with one 4f electron) in some detail. The lowest electron configuration of the free Ce3+ ion is 2F ( L = 3 S = 4) which gives rise to a 2F5/2 (J = 3 - 8 ) and a 2F,/2 (J = 3 + &) state the former being the ground state. The energy separation between these two states arises from the spin-orbit coupling &S^which may be written in the form n A A n ,-. A(E@ = - L2 - S2] = $A[J2 - L2 -S2].For the states 2F5/2 and 2F,/2 J L and S are good quantum numbers so that the expectation values of J2 L2 and s^2 are J(J + l) L(L + l) and S(S + 1) respectively. Hence the energy difference is just E(2F,/2) - E(2F)512 = *A[7/2(7/2 + 1) - 5/2(5/2 + l)] = (7/2)A and has been found experimentally to have the value + 2,250 cm.-l. The simplest approximation that can usefully be made is to suppose that the ground state of the Ce3+ ion in the crystal is derived entirely from the ‘F5/2 level of the free ion.lS If the crystal field has the trigonal symmetry Csh or D3h its effective potential splits the six-fold degenerate 2F5/2 ground state of the free ion into three Kramers doublets which turn out to be eigenfunctions of the operator? with eigenvalues j-4 j- and -J-i respectively.We shall not explain in detail why this happens but merely write down the wave functions of the MJ = &h states which are 14) = dqqO,*) - .\/4/7]I,-i) and 1-4) = d 4 ~ ~ - l + ) - d ~ ~ O - ~ ) where the two numbers in each “ket” on the right-hand side are the eigen- values ML Ms of the operators L, S,. To calculate gll and g for transi- tions between these levels we need the matrix elements of PH,& $- 2gz) and + 2s). The diagonal elements of the former are PHd-4 I(L2 + 2 6 1-8) = (-3/7)PHz n n PK<B @z + 2% 13) = (3/7)PHz and la R. J. Elliott and K. W. H. Stevens Proc. Roy. SOC. 1952 A 215,437. 444 QUARTERLY REVIEWS and there are no off-diagonal elements. Hence gll = 6/7 = 0.857. For a field applied in the perpendicular direction we have (8 I P K L + 2% 1-4) = (9/7)Pz = <-B IPKdLh + 2% 14) and this time there are no diagonal elements.Hence gl = 18/7 = 2.571. Experimentally it is found1’ that a diluted single crystal of Ce3+ in lanthanum ethyl sulphate gives two resonances; one has gll = 0-955 g = 2-185 and the other arises from a doublet about 3 cm.-l higher having 811 = 3.72 gl = 0.2. The g values calculated for the MJ = i-; doublet are 4.286 and 0 respectively. The situation is illustrated in Fig. 7. Free ion Spln orbit Crystal Magnetic coupling field field FIG. 7. Splitting of the ground state of the Ce3+ ion in a field of symmetry D The MJ = & doublet is apparently too high in energy to be appreciably populated at the temperatures of measurement. In the foregoing paragraphs we have discussed only the very simplest rare-earth ion with one 4f electron and have applied only the most elementary theory to it.A more accurate theory would allow for mixing of levels with different J under the influence of the crystal field.16 Never- theless to ignore this effect in the rare earths leads only to a slightly inaccurate description whereas in the transition-metal series it would be completely misleading to think of the ions in the crystal field as being in eigenstates of 3. 7. Nuclear hyperfine structure As mentioned in section 2 the presence of one or more magnetic nuclei in a paramagnetic substance can impart to the electron-resonance spectrum a “hyperfine” structure in which each individual line arises from an elec- l7 B. Bleaney Phil. Mug. 1951 42 441. CARRINGTON AND LONGUET-HIGGINS ELECTRON RESONANCE IN CRYSTALS 445 tronic transition in which the nuclear spin (or spins) remains oriented in a particular direction.If there is one magnetic nucleus with spin magnitude f one may observe (21 + 1) hyperfine components since MI the eigen- value of 6 can take any one of the (21 + 1) values - I - I -+ 1 . . . + I. These components are usually of equal intensity and equally spaced. The magnitude of the spacing depends on two factors (i) the resultant magnetic field and (ii) the electron spin density at the given nucleus. These effects need separate discussion. The magnetic field experienced by a nucleus arises from the spin and orbital angular momentum of the electrons revolving in its neighbourhood the direct contribution of the external magnetic field being usually negligible. The contribution of a particular electron falls off inversely as the cube of its distance from the nucleus so that electrons which do not get into the valency shell of an atom have a negligible effect on its nucleus.Another feature of this interaction is its directional character; for a given electronic level its average over all orientations of the molecule is zero. (This is why this particular interaction need not be considered in inter- preting the electron-resonance spectra of paramagnetic species in liquid solution.) The other factor determining the hyperfine spacing is the unbalance of electron spin in the immediate neighbourhood of the nucleus. This is the so-called Fermi contact interaction and can arise only if some unpaired spin finds its way into s orbitals of the atom because orbitals of pure p d and f type have vanishing amplitudes at the nucleus.When an external field is applied in a given direction the spin of the electrons tends to become aligned in one of (2s + 1) directions (though spin- orbit coupling will partly upset this alignment) and the average direction of the electron spin near the nucleus will determine the axis of quantisation of the nuclear spin I. It may be noted that as far as any nucleus is concerned the effect of the external field is only to determine the manner in which the electrons move in its neighbourhood; the magnitude of the hyperfine spacing does not therefore? depend on the strength of the external field but only on its direction (though in certain special circumstances this statement needs modification). We have already mentioned briefly one situation in which the nuclear spin enters the hamiltonian for a paramagnetic substance; in section 2 we showed that the spin hamiltoniait for the hydrogen atom could be written as n and that the last term was negligible.Later we showed that when allow- ance was made for the coupling between S and the orbital angular momen- A 9 446 QUARTERLY REVIEWS n tum L the spin hamiltonian for an axially symmetrical system (excluding I’ could be written as A gls(H3cK + H,$ + g1,BHzK where gl and gll were not necessarily equal to one another or to 2.0023 the value for a free electron. These observations lead us to expect that in a situation in which both EandTenter the full hamiltonian it may still be possible to describe the spin-resonance properties of an axially sym- metrical system by an effective hamiltonian of the form &B(H,E + HYK) + 811 PHZK + A,($2 3- KG + A ll~;z This expression suggested by analogy can be justified by detailed analysis provided the direct effect of H on the nucleus can be neglected and pro- vided also that the nucleus has no quadrupole moment.(Terms in S& If etc. are also assumed to be negligible.) The eigenvalues of this effec- tive hamiltonian can be found by methods similar to those already out- lined.2 For example if the field H is in the z direction the spin hamiltonian becomes gll/3Hzgz + A l z z E + +Al(KZ + cc) A A A A A A A where I + = I + iIy and I- = I - iIy. These operators operate on the nuclear spin states in precisely the same manner as the operators S+ and A A S- operate on the electronic spin states.For an electronic doublet (Ms = 5 4 ) and a nucleus of spin i (MI = &&) the matrix of the above hamil- tonian within the set of states IMS,Mj) is 1898) 14 -9 I -Q 8) I-& -3) 1 4gllmz+MI 0 0 0 I Z s l l P H Z - *All &A1 0 -4grlBHZ-aAIl 0 0 - 4gllrBH~faAll Ml 0 of which the eigenvalues are &4 It higll PITz and -&f 11 -J-d[(&gll +(8AJ2]1; when A,<g,l the eigenvalues are therefore close to *&A 11 &$tg11 PHz as asserted in section 2. 8. Electron delocalisation So far our discussion of transition-metal compounds has been based on the implicit assumption that the unpaired electrons are localised entirely in the valency shell of the metal ion. In reality the unpaired electrons may be partly delocalised on to the ligands or the ligand electrons may be partially delocalised into the metal orbitals.Electron resonance provides a sensitive method for the study of this effect,l8*l9 as we shall now explain. l8 K. W. H. Stevens Proc. Roy. SOC. 1953 A 219 542. l9 J. Owen Discuss. Faraday SOC. 1955 19 127. CARRINGTON AND LONGUET-HIGGINS ELECTRON RESONANCE IN CRYSTALS 447 A quantitative definition of electron delocalisation may be obtained from the molecular-orbital theory as follows. If 4 is one of the d orbitals of the metal and x is a combination of ligand orbitals having the same symmetry as 4 then a molecular orbital of the combined ion will have the form 45 3- ax * = (1 + 2aa + a2) where a is the overlap integral between4 and x; a then measures the extent of delocalisation of an unpaired electron in the orbital +.Estimates of 01 can be obtained in several ways namely (a) From the g values. A convenient example is the IrCli- ion (t2g5) whose electron-resonance spectrum can be observed in dilute solution in the diamagnetic crystal K2PtC1,.20 The orbital of the “unpaired hole’’ is a d orbital of type t2g which can overlap r-wise with a suitable combina- tion of C13pr orbitals. It has been shown that for a = 0 when distortions of the ion are neglected one should obtain an isotropic g value of 2 where- as for a very large the g value would be isotropic with the value 4/3. The actual value is 1.8 indicating1* that a2/(1 + 2aa + a2) lies between 0 and 0.3. Roughly then the extent of delocalisation az must be about 15 %. It has been shownz1 that o-bonding also results in a reduction of the g values.In the hydrated Ni2+ ion for example the (isotropic) g value should be where h is the cg coupling constant of the bound ion d is the crystal field splitting of the 3d orbitals and overlap has been neglected. Un- fortunately this formula cannot be used to obtain a very accurate estimate of a2 since h is known less accurately for the bound ion than for the free ion. (b) From ligand hyperfine structure. An example of this is the electron- resonance spectrum of IrCli- where electron transfer is revealed very directly by the appearance of chlorine hyperfine structure superimposed upon the quartet hyperfine structure due to the iridium nucleus.20 This phenomenon like the reduced g value is clearly due to partial transfer of C13pn electrons into the tZg “hole” on the metal ion resulting in a transfer of unbalanced spin in the opposite direction.For an arbitrary orientation of the IrC1;- octahedron in the external magnetic field the C1 hyperfine structure is rather complicated but if the field is applied along a C1-Ir-C1 axis the two C1 nuclei on the axis split each Ir hyperfine line into a septet whose spacing depends on a2 and on the mean value of r3 for an electron 2o J. H. E. Grfiths J. Owen and I. M. Ward Proc. Roy. SOC. 1953 A 219 526. 21 M. Tinkham Proc. Roy. SOC. 1956 A 236 535 549. 448 QUARTERLY REVIEWS in a 3pn C1 orbital. The value of a2 obtained from the septet spacing is approximately 0.26 overlap being neglected. (c) From reduced metal hyperfine structure. It is natural to suppose that delocalisation of the magnetic electrons on to the ligands will result in decreased interaction with the metal nucleus.22 An example is the Cu(H20)i+ ion where transfer of electrons from the H,O molecules into the eg “hole” in the valency shell of the metal ion reduces the unpaired spin density in the neighbourhood of the Cu nucleus.Thus the Cu hyper- fine spacing is appreciably smaller than would be expected from a calcula- tion in which a is set equal to zero. Many other examples of these three effects have been found experi- mentally. 9. Exchange interaction An important phenomenon which can be studied by electron-resonance methods is electron exchange between two paramagnetic ions in a crystal. This type of interaction has been in salts of the IrCli- ion the technique being to grow crystals in which 95 % of the Ir atoms are replaced by diamagnetic Pt atoms.In these crystals there is an appreciable. number of nearest-neighbour pairs of Ir ions which have no other Ir neighbours. The structure of such a nearest-neighbour pair can be represented dia- grammatically thus Tr - \ . . . Cl/ Ir We have already discussed the electron transfer which occurs in the 1rCl:- ion. The 3pr electrons of the Cl- ions spend 5 % of their time on the Ir4+ ions and there is thus a certain definite probability of finding two chlorine atoms in neighbouring positions with spins either parallel or antiparallel; in the latter case we may think of an elongated Cl “molecule”. The resulting energy levels are a triplet (S = 1) at i K - gPH BK &K + gPH and a singlet at - iK where Kis the C1-Cl exchange energy and one observes an electron-resonance spectrum arising from transitions between the levels of the triplet.The intensity of these transitions decreases as the temperature is lowered indicating thermal depopulation of the triplet level. Thus the triplet lies above the singlet and the exchange is said to be antiferromagnetic. Equally interesting examples of exchange coupling are found in some copper salts. In copper acetate2* the fine structure of the spectrum is 22 B. Bleaney K. D. Bowers and M. H. L. Pryce Pruc. Roy. Suc. 1955 A 228 166. ** B. Bleaney and K. D. Bowers Pruc. Roy. SOC. 1952 A 214,451. J. Owen Discuss. Furuday Suc. 1958 26 53. CARRINGTON AND LONGUET-HIGGINS ELECTRON RESONANCE IN CRYSTALS 449 characteristic of a triplet spin state and in addition the hyperfine structure shows that the odd electrons interact with two equivalent copper nuclei.This indicates that the copper ions interact strongly in pairs through exchange forces each pair forming a paramagnetic triplet spin state and a lower diamagnetic singlet spin state. The nuclear spins of the two stable copper isotopes are both 3/2 and their magnetic moments are so close that except under conditions of very high resolution only one set of hyperfine lines is expected. This will normally consist of four lines with equal spacing and intensity. It is easy to show that two equivalent copper nuclei will give rise to seven lines with relative intensities 1 2 3 4 3 2 1 and this is just what is observed for copper acetate. The intensity of the electron-resonance spectrum again decreases as the temperature is lowered.It is interesting to note that the deduction that the copper ions form isolated pairs was subsequently confirmed by X-ray crystallography; the Cu-Cu distance was found to be 2-6A. 10. Univalent transition-metal ions An interesting and recent development is the detection and study of some unipositive transition-metal ions formed by irradiation of the bivalent ions present as impurities in single crystals such as NaF. We can best describe this by referring to one or two specific examples.25 Crystals of NaF to which Cr3+ has been added do not show resonance down to 20°K. This can be understood if the chromium is present as Cr2+; this ion with an even number of unpaired electrons might well not show resonance because of a large zero-field splitting of the spin quintet. However on irradiation with y-rays X-rays or high-energy electrons an isotropic spectrum appears whose fine structure is characteristic of a 6S ground state.As in the spectrum of Mn2+(3d5 %) metal hyperfine structure is observed but is found to arise from a nucleus with spin 3/2 which is the spin of 53Cr. There is thus very good evidence that the spectrum observed is due to Cr+ (3d5). Similarly when crystals of NaF containing Fe2+ ions are irradiated a resonance line appears at 20°K with a g value of 4.344. This g value is very close to that observed for Co2+ in MgO crystals (4.278). However the line observed is not due to the presence of cobalt as an impurity since the hyperfine structure characteristic of that metal is absent. The obvious conclusion is that it is due to Fe+ (3d7) which is isoelectronic with Co2+ (3d7 *F) but has no nuclear spin.The Table summarises some of the available data for transition-metal- ion impurities in NaF and MgO both of which have the NaCl crystal structure. While the g values are in general extremely sensitive to the local structure of the crystal the results quoted have been chosen because 2s W. Hayes Discuss. Faraday Soc. 1958,26 58. 5* 450 QUARTERLY REVlEWS d6 d7 d8 d9 No. of d Ion Temp. (K) g Value electrons d5 Mn2+( Na F) Room temp. 2-00 & 0.01 Mnw(Mg0) 9 2-00 14 Few(Mg0) 9 9 2.0037 Crt(NaF) 9 2.000 Fez+( MgO) 4” 3.428 Mn+(NaF) Resonance not observed down to 20”~ but intensity of Mn2+ spectrum decreases by 90% on irradiation Coz+(MgO) 20” 4.278 Co2+(NaF) 20 g = 4.3 g = 3.3 g = 5.7 Fe+(MgO) 20 4.15 Niw(Mg0) 77 2.227 Co+( N a F) 90 2.3 1 Cu2+(Mg0 powder) 2.190 Ni+(NaF) 20 811 = 2.766 Fe+( NaF) 20 4.344 gL = 2.114 they refer to ions which probably have similar crystal environments.In these circumstances the g values provide strong evidence for the existence of the Crt Fe+ Co+ and Ni+ ions in the irradiated crystals. 11. The investigation of crystal structures It will by now be apparent that the electron-resonance spectrum of a compound is largely determined by its crystal structure. In some cases important features of an unknown crystal structure can be inferred from the electron-resonance spectrum. We have already quoted cupric acetate as a case in point and another good example is provided by recent work on hzemoglobin and myoglobin and some of their derivatives.1° In these substances the electron-resonance spectrum arises entirely from the co- ordinated iron atoms so that the hzem groups can be studied without inter- ference from the rest of the molecule.Fig. 8 represents the structure of one of the haem units. The iron atom is centrally placed between the four nitrogen atoms; the hzm group is linked to the rest of the molecule through a fifth co-ordination point and various other groups can be attached at the sixth position (R). In the case of the “acid-met” derivative this position is occupied by a water molecule and Fig. 8 shows that the line joining the water molecule to the iron atom is an axis of local symmetry. What one actually observes in studying “type A” myoglobin crystals is a pair of resonance lines whose g values depend differently upon the crystal CARRINGTON AND LONGUET-HIGGINS ELECTRON RESONANCE IN CRYSTALS 451 orientation.Each line has a minimum g value of 2.0 for one particular crystal orientation and a maximum g value of 6.0 for any perpendicular orientation. This clearly indicates the presence of two hzem groups in the unit cell and the directions of minimum g value for the two groups may R Globin FIG. 8. Immediate environment of an iron atom in hamoglobin. be identified with the Fe-R directions. The orientations of the ha=m groups in myoglobin and in hzmoglobin were thus determined relatively to the crystallographic axes within an accuracy of 2" by electron resonance before the recent comprehensive X-ray studies of these molecules had been completed. 12. Line widths For a crystal in a perfectly uniform magnetic field subjected to weak monochromatic microwave radiation there are two main sources of line broadening namely spin-lattice and spin-spin interaction.Spin-lattice interaction is the means whereby the magnetic energy of the paramagnetic ions is transferred to the vibrational degrees of freedom of the lattice. We have already seen that it is the spins of the electrons that mainly give rise to the magnetic energy; hence spin-lattice relaxation requires some kind of coupling between the electron spins and the varying electrostatic field due to the lattice motions. This coupling is none other than the spin-orbit interaction already discussed at some length. If spin-lattice relaxation were the only source of line broadening one would obtain a line of half-width -(27rT)-l cycles where T is the relaxation time for transfer of energy from the spin system to the lattice.The magni- tude of T is sensitive to two factors. First for a given substance T depends on the temperature since as the temperature is raised the violence of the interatomic motion increases and the relaxation time is consequently shortened. Secondly if we are comparing different substances r will depend on the sensitivity of the g value to the geometry of the ionic environment. Thus Mn2+ has a long spin-lattice relaxation time whereas the octahedrally co-ordinated Ti* ion whose g values are sensitive to 452 QUARTERLY REVIEWS small distortions has a very much smaller value of r and only gives sharp lines at low temperatures. The sensitivity of the magnetic levels to the environment will in general be greater the smaller the excitation energy to higher magnetic states; hence line widths will in general be greater if there are low-lying excited states.The other source of line broadening is direct interaction between different paramagnetic units. This is of course very small in paramagnetic- ally dilute crystals but becomes important in undiluted paramagnetic materials. There are two main mechanisms of spin-spin interaction; one is the direct dipole-dipole interaction between ions regarded as fixed bar magnets and the other is exchange coupling of the electron spins possibly mediated by intervening atoms (see section 9). Both mechanisms have the effect of replacing each N-fold degenerate state of a group of N independent ions by a set of N states differing slightly in energy. The result is that the various excitation energies are no longer exactly equal and resonance is obtained over a range of magnetic fields. If the ions are non- equivalent each resonance line is visibly broadened ; if they are equivalent the line appears quite sharp (“exchange narrowing”) but its fourth moment is actually greater than it would be if there were no exchange coupling. It should be noted that varying the temperature has no effect upon spin-spin broadening since the motion of the lattice does not alter significantly the distances between neighbouring ions. One can of course study line widths with a view to evaluating spin- lattice relaxation times or spin-spin coupling constants. On the whole however the broadening of resonance lines is something that one tries to avoid experimentally and this is best done by working with paramagnetic- ally dilute crystals at low temperatures.

ISSN:0009-2681    

DOI:10.1039/QR9601400427

出版商:RSC    

年代:1960

数据来源: RSC