摘要:

r QUARTERLY REVIEWS STEREOCHEMISTRY OF C YCLOHEXANE By 0. HASSEL DR. PHIL. H. c. (PROFESSOR OF PHYSICAL CHEMISTRY UNIVERSITY OF OSLO) OUR knowledge of the spatial distribution of atoms in aromatic compounds is to-day considerably more detailed and precise than it is for organic com- pounds in which the chemical bonds between adjacent atoms are essentially of the “ two-electron ” type. Even in the case of cyclohexane and other substances containing six-membered rings which are believed to be com- paratively “ rigid ” much more work will have to be carried out before their stereochemical properties can be treated with a similar degree of confidence. Both from a purely theoretical point of view and because of the biological importance of many compounds belonging to this class it therefore appears worth while to gather new experimental evidence about their atomic arrangement.During the last two decades efforts have in fact been made to establish fundamental facts especially about the behaviour of the cyclohexane ring by using direct methods based on the diffraction of X-rays or electrons. The present Review gives a short survey of the principal experimental results so far obtained but the complicated physical aspects are not discussed. General Properties of the cycZoHexane Ring.-The idea that the valency angles of the six carbon atoms of the cyclohexane ring are tetrahedral or nearly so has been generally accepted by chemists and the same may be said of the conclusions drawn from it by Sachse and by Mohr regarding the a possible forms of the ring. Pig. l a shows the rigid form (“ staircase ” or ‘‘ chair ” model) of the cyclohexane molecule Fig.l b a special case of the alternative ‘‘ movable ” model described as the ‘‘ boat ” or “ tub ” form. No objection can be raised against acceptance of the chair model of the cyclohexane molecule from considerations regarding the nearest approach of hydrogen atoms (we can calculate the nearest approach with considerable confidence by assuming a C-C bond distance of 1.54 and a C-H bond FIG. 1 Q 22 1 222 QUARTERLY REVIEWS distanca of 1.10 A). This model leads to a smallest hydrogen-hydrogen separation of about 2-5 A a distance which is not far from twice the van der Wads radius of hydrogen and should therefore be expected to be energetically favourable. This distance occurs 24 times within the molecule a fact which no doubt contributes appreciably to stabilisation of the chair model.In the boat form however pairs of hydrogen atoms would be con- siderably closer if the carbon atoms retain their normal valency ang1es.I All the available experimental evidence actually indicates that practically all cyclohexane molecules correspond to the chair form both in the vapour and in the liquid phase. One fact seems to have been given very little attention until 1943,l namely that the hydrogen atoms of cyclohexane in the symmetrical chair form may be divided into two geometrically different groups each com- prising six hydrogen atoms. The C-H bonds of the first group are parallel to the symmetry axis of the molecule which is usually placed vertically. Originally,l these bonds were designated as E bonds ( i ~ r ~ ~ d g = upright).The bonds of the second group are not parallel but all form an angle of & 19.5" (109.5" - 90') with the horizontal plane ; they were called K bonds ( I C ~ L ~ E Y O ~ = prostrate). Some years later the designations '' polar " ( E ) and '' equatorial " ( K ) bonds were suggested and have since been adopted by some authors. It may perhaps be objected to the original designations that they are less easily niemorised and to the latter that the expression " polar bond " may be misunderstood.* Throughout this Review the original designations will be used. If we consider for a moment a cyclohexane molecule of the synimetrical chair form to be " rigid " and replace one of its hydrogen atoms by another atom (e.g. chlorine) it will easily be seen that the new molecule will be a b FIG.2 represented by different models depending on whether we replace a K or an E hydrogen atom of the original cyclohexane molecule (Fig. 2a and b). Neither chlorocyclohexane nor any other mono-derivative shows any sign of steric isomerism so if the two forms (E and K) of a mono-derivative both exist one may easily change into the other. This is only possible if the activation energy of the process is much smaller than the energy known to be necessary for fission of a carbon-carbon bond. Obviously 1 0. Hassel Tidsskr. Kjemi 1943 3 32. 2 C. W. Beckett K. S. Pitzer and R. Spitzer J . Amer. Chem. SOC. 1947 69 2488. * These terms polar and equatorial are used in most British papers and are often abbreviated to p and e. It is to be noted that p (and not e) corresponds to E that this use of polar should not be confused with the electrochemical use (cf.above) and that Roman font is used (to avoid confusion with p = para).-F,:r>. HASSEL STEREOCHEMISTRY OF CYCLOHEXANE 22 3 therefore the process does not involve the breaking of such bonds it will however probably be closely associated with normal thermal vibrations carried out by the carbon atoms of the six-membered ring. Ring Conversion.-Now these six carbon atoms are arranged in two planes three in each with a mutual distance between the planes of approx- imately 0.5 8. The oscillation of the two triangular groups may some- times cause them to interchange positions without necessarily causing fission of chemical bonds. Such a process although not causing any change of the total configuration of cycbhexane molecule itself necessitates that each hydrogen atom of the K type shall be transformed into one of the E type and vice versa.In the case of a mono-derivative the conversion process transforms one of the two possible forms (Fig. 2a and 6 ) into the other. If on the other hand we consider a derivative obtained by substituting equal numbers of K and E hydrogen atoms by atoms or groups of the same kind it may occur that the molecule is identical in form before and after a con- version of the type just described. This is so for example for the 1&,2~,4~,5~-tetra-~ubstituted cyclo- hexanes (Fig. 3). This molecule does not possess elements of symmetry other than a two-fold axis and in this special case a chemical separation of (+)- and (-)-forms will therefore be possible at least in principle.A second example is afforded by any trans-1 3-disub- stituted cyclohexane if the two substituents are of the same kind. If the substituents are different separation of (+)- and (-)-forms is still possible but the picture of the molecule changes when a conversion of the ring takes place. Each of the two optical antipodes may therefore exist in two '' conformations " ( 1 ~ 3.5 and 1~ 3K) which will not necessarily have the k-= FIG. 3 9 FIG. 4 3 The word " conformation " is used by organic chemists to denote a particular arrangement of the atoms of a molecular species when more than one arrangement is possible. It should not of course be employed when molecular structures such as those of cis- and tram-decalin are being compared. I n the German language " constellation " is generally used in preference t o '' conformation " (cf.J. 1962 6061). 224 QUARTERLY REVIEWS same energy. The situation is the same for cis-1 2-&substituted com- pounds with two different substituents (see Fig. 4). Measurement of the relative amounts of the two configurations in the equilibrium mixture would be'very interesting because it would probably throw some light on the forces between the two kinds of substituents and their nearest (hydrogen) neighbours (compare the case of l e 2~ 4~ 5~-tetrahalogenocyclohexanes discussed below). In other cases one of the two possible conformations of a cyclohexane derivative simply represents the mirror image of the other. The equilibrium mixture then necessarily contains equal amounts of the two configurations k=& FIG.5 and a separation is not feasible except possibly at extremely low tempera- tures or in the solid state if the substance happens to crystallise in enantio- morphous crystals. Usually however crystals containing equal amounts of the two configurations (and thus corresponding to normal racemic com- pounds) will represent the stable state and not a mixture of enantiomorphous crystals. This normal behaviour is shown for instance by cis-cyclohexane- 1 2-diol (Fig. 5). It should be borne in mind that the usual representation of molecular configurations of cyclohexane derivatives and similar compounds showing a planar six-membere'd ring and carbon valencies perpendicular to the plane of the ring may be said to comprise the whole series of special shapes in which a defined chemical individual can possibly appear.Although such a representation is not useful as a basis for detailed discussions regarding the energy of the free molecule or its chemical reactions it is well suited for classification. Thus the symmetry of the picture directly indicates whether a special representation comprises a pair of mirror image molecules or not. It is therefore easy to find the number of steric isomers which correspond FIG. 6 to a given structural formula by employing models of this kind. The ex- pressions cis and trans indicating the mutual positions of two substituents linked to different carbon atom will probably continue to be used although they do not generally (compare the case of trans-1 2-derivatives Fig. 6) give direct information concerning the real spatial arrangement of the substituents.HASSEL STEREOCHEMISTRY OF CYCLOHEXANE 225 Returning to the steric configurations based on the symmetrical chair form of the cychhexane ring we may draw some general conclusions. Two substituents each having a van der Waals radius larger than that of hydrogen (1.25 A) will be expected to repel each other if they both occupy E positions on the same side of the carbon ring. This may be of significance even to a single chlorine atom in an e position because its distance from the two nearest hydrogen atoms (also in E positions) would be smaller than the sum of the radii. A certain deformation of valency angles and a corresponding rise of the internal energy of the molecule might be the consequence of this repulsion. to decide experimentally if the “ free ” chlorocycZohexane molecules exist preferentially in the K rather than the E form.By means of the then new ‘‘ sector method ” of electcon-diffraction experimental distance- distribution curves were worked out both for chlorocyclohexane and for cycibhexane itself. The curve obtained by subtracting the latter curve from On the lines of this argument an attempt was made ten years ago FIG. 7 the former should exhibit maxima for values of the abcissa (r values) corresponding to C-Cl distances occurring in the chloro-compound and not in the parent hydrocarbon. The result is shown in Fig. 7 where the arrows are proportional in length to the weight of the distances in question. The two unbroken arrows on the left correspond to C-C1 distances expected in both forms of chlorocyclohexane and are therefore without interest in the present connection.The broken arrows represent C-C1 distances present only in the model in which the chlorine atom is in the E position (Fig. 26) ; the two unbroken arrows on the right indicate the corresponding C-Cl distances in the model having the K chlorine atom (Fig. 2a). The conclu- sion cannot of course be drawn from this experiment that molecules with chlorine in the E position are not present in the vapour phase but their amount must obviously be considerably smaller than that of the K form. Similar results were obtained a few years later for cyclohexanethiol. The distance between two adjacent halogen atoms both in K positions 4 0. Hassel and H. Viervoll Tidsskr. Kjemi 1943 3 35. 0. Bastiansen and 0.Hassel ibid. 1946 6 96. 226 QUARTERLY REVIEWS would in an ideal model probably be small enough to produce repulsive forces between them. Although sufficiently accurate measurements are still lacking the experimental evidence seems to indicate that the valency angles are in fact not strictly tetrahedral in such cases a certain flattening of the carbon ring being the chief result. On the other hand it has been established beyond doubt that deformation is caused by repulsion between E chlorine atoms and the nearest hydrogen atoms in the E position ; it causes tilting of the C-Cl bond away from the chief axis of the carbon ring (by about 7°).6 If the interaction energy due t o the two carbon-halogen dipoles is also taken into account it becomes probable that the energy difference between the two forms of trans-1 2-dihalogenocyclohexanes (KK-EE ; cf.Fig. 6) may be relatively small and that their concentrations in the equilib- rium mixture therefore roughly equal.5 The number of simple derivatives of cycbhexane available for molecular- structure determinations and likely to give more precise information about the forces acting between atoms or groups linked to the carbon atoms of the ring is limited although new substances have been specially prepared for this purpose and investigated. Further work in this field seems to promise results of interest. If 1 2 3-trihalogenocyclohexanes for example capable of existing in the conformations KKE and EEK should prove to be present as “free” molecules predominantly in the KKE form this would mean that the energy rise due to the ~K-ZK interaction is smaller than that caused by the repulsion between an E halogen atom and its nearest hydrogen neighbours.The 1 2-dibromo-4 5-dichlorocyclohexane in which both the chlorine atoms and the bromine atoms are trans one pair occupying E positions and the other pair K positions (Fig. S) has been prepared in order to decide ? FIG. 8 which of the two pairs are in the E position when the molecule takes the form having the smallest possible energy. The substance proved to be isomorphous with the corresponding tetrachloro- and tetrabromo-cyclo- hexanes and the crystal structure showed that in the solid state the chlorine atoms occupy E positions the bromine atoms K positions.‘ Electron- diffraction measurements of the ‘‘ free ” molecule of the vapour confirmed this result.No indications of the presence of molecules having the “ con- verted ” structure could indeed be detected.8 This finding does not of course answer questions about the relative energy contributions of a pair of 1~ ZK halogen neighbours and of an E halogen atom. There are two E 0. Hassel and E. Wang Lund Acta Cryst. 1949 2 309. 7 Idem Acta Chem. Scand. 1952 6 238. * 0. Bastiansen and 0. Hassel ibid. 1951 5 1404. HASSEL STEREOCHEMISTRY OF CYCLOHEXANE 227 halogen atoms in the molecule but only one pair of KK neighbours and the halogen atoms are chlorine and bromine respectively. The conclusion seems justified however that the le-38 repulsion between an E halogen atom and its two nearest ( E ) hydrogen neighbours raises the energy of the molecule more in the case of bromine than in that of chlorine.The angle of deflexion of the e C-C1 bond away from the ring axis appears to be nearly the same in the dibromodichloro- as in the tetrachloro-compound. The finding that mono-derivatives have their lowest energy in the form in which the substituent occupies the K position is substantiated by many other experimental results. All the trans- 1 4-dihalogeno-compounds so far investigated have been observed only in the KK-configuration (dichloro- dibromo- di-iodo- and bromochloro-cyclohexane). Although the piperazine ring is certainly not identical in shape with that of cyclohexane their similarity should probably be great enough t o make it probable that the 1 4-dihalogenopiperazines will also be most stable in the configuration which corresponds to the KK form of the trans- 1 4-dihalogenocycZohexanes.This has been found to be the case for the dichloro-compound both in the crystalline and in the gaseous state.g It should be remembered that here the KE conformation (corresponding to the &-forms of 1 4-dihalogenocycZohexanes) might be present in the equilibrium mixture. One is of course not justified in drawing direct conclusions from the results obtained for other halogen derivatives when considering the stability of the configurations of fluorinated cyclohexanes. It has been observed that dodecafluorocyclohexane has it structure which does not deviate very much from an ideal structure with a carbon ring of the chair form. This seems to indicate that the repulsion between two E fluorine atoms in the 1 3-position is much smaller than between two other halogen atoms in similar positions (see below).Taking this and the strong electronegativity of the fluorine atom into consideration it may perhaps be justified to expect that mono- fluorocycbhexane will be more stable in the r than in the K conformation contrary to the finding in the case of other halogenocyclohexanes. An electron-diffraction investigation of this compound and also of trans-1 4- difluorocyclohexane in the vapour state would thus be of considerabIo interest. The '' Benzene Hexachlorides '' and the Cyclitols.-The molecular structure of a number of derivatives of cyczohexane has been investigated during recent years mostly for halogen derivatives. In some cases the positions of the substituents in the molecule were unknown or a t least uncertain when the structure analysis was started.So far i t appears the molecular structure actually observed for a given chemical individual always represents the configuration which we would expect to possess the lowest internal energy on simple steric considerations and accepted values of van der Waals radii. It seems appropriate to pay special attention to the series of benzene hexa- chlorides (1 2 3 4 5 6-hexachlorocycEohexanes). Theoretically eight P. Andersen and 0. Hassel Acta Chem. Scand. 1949 3 1180. 228 QUARTERLY REVIEWS steric isomers are possible of which one should be separable into mirror- image (+ and -) forms. The idea expressed several years before deter- mination of the structure was finished namely that this compound is the a-isomer the chief product of the reaction between benzene and chlorine has been found to be correct and separation of one active form in a pure state has been accomplished.1° As long as we are not interested in absolute configurations of optically active molecules a satisfactory description of the molecular structure of a cyclohexane derivative may be given simply by indicating for every sub- stituent whether it is of the E or the K type.The five structurally known benzene hexachlorides may for example be listed as follows (cf. Fig. 9) &&KKKK y &&&KKK & &KK&KK @ KKKKKK d EKKKKK By comparing the observed configurations with those of the conversion forms obtained by simply changing each & into K and vice versa it is seen that the y-isomer is the only one which would retain its configuration.It FIG. 9 is significant that the conformations listed above and actually observed for the other four isomers do not contain a single pair of &-chlorine atoms in the 1 3-position whereas the configuration obtained by " conversion " of the carbon ring would contain at least two such pairs. It strengthens the argument mentioned above that the presence of a pair of &-halogen atomsin 1 3-position will raise the energy of a cyclohexane derivative markedly. The expected deviation from an ideal structure caused by the presence in the y-isomer of the l e 3~ chlorine pair is clearly brought out by the X-ray analysis.ll? l2 l o 8. J. Cristol J . Amer. Chern. Snc. 1949 71 1894. l1 W. van Vloten C . A. Kruissink 33. Strijk and J. M. Rijvoet A c f o C ' y s t .. 1950 l2 Many more examples of halogenated cyclohexanes might be quoted ; cf. 0. 3 139. Haxsel Research 1950 3 504. IIASSEL STEREOCHEMISTRY OF CYCLOHEXANE 229 Molecular structures of the corresponding hydroxy-compounds (cycb- hexanehexols) have so far not been extensively studied. The elucidation of these structures would however be of special interest both because these substances are biologically important and because very little is known about the forces acting between hydroxy-groups in compounds of this kind. It would for example be most interesting to know whether intramolecular hydrogen bonds are formed and if so to discuss their significance for the molecular conformation. Recently a paper dealing with the reaction between cyclitols and acetone has been published which is of considerable interest in connection with problems discussed in the present Review.Experimental evidence seems to indicate that the replacement of methyl- ene groups in cyclohexane by oxygen atoms does not cause very marked changes in the shape of the six-membered ring. X-Ray work on carbo- hydrates containing pyranose rings substantiate this view. T4e stereo- chemistry of sugars therefore has to follow within certain limits the same lines as that of cyclohexane derivatives.l* Condensed Ring Systems.-Molecules in which bridges are formed between two carbon atoms of a cyclohexane ring may of course be treated in the same way as ordinary disubstituted cyclohexanes at least if the bridge contains a sufficient number of atoms to allow the valencies of the cyclohexane carbon atoms to retain their normal angles.The special case in which the bridge is formed between adjacent carbon atoms of the cyclohexane ring and contains four methylene groups is of special interest. It was early sug- gested that the so-called trans-decalin consists of two ordinary cyclohexane rings having two carbon atoms in common as indicated in Fig. 10b ; the two - I a b FIG. 10 FIG. 11 rings are joined by the use of K bonds only It would have seemed natural to assume that the cis-form of decalin is also based on normal cyclohexane rings of the chair form joined by one K and one E bond of each ring (Fig. 10a). Mohr,15 however suggested a structure (Fig. 11) built up of two cyclohexane rings of the boat form a view still to be found in textbooks although it is certainly not in agreement with experimental results and its very l 3 S.J . Angyal and C. G. Macclonalcl J. 1952 686. I4 0. Hassel aid B. Ottar Acta Chenz. Scand. 1947 1 929 ; R. E. Reeves J . Amer. l 5 E. Mohr J . p. C'hem. 1918 98 315. I6 0. Bastiansen and 0. Hassel Nature 1946 157 765; Tidsakr. Kjemi 1946 Chem. Soc. 1950 72 11-99. 6 70. 230 QUARTERLY REVIEWS unfavourable H-H distances were pointed out ten years ag0.l' The designa- tions cis and trans for the two stereoisomeric decalins orginally indicated the mutual positions of the two hydrogen atoms linked to the carbon atoms which the two cyclohexane rings have in common. As the molecular struc- tures of the two compounds are now known it seems appropriate to associate the two designations with the positions of the bonds starting and terminating the bridges rather than with the hydrogen atoms of the two CH-groups.The extension of the work carried out on the decalins to fully hydro- genated aromatic hydrocarbons originally containing more than two benzene rings (now in progress) will no doubt be of interest especially if small devia- tions from " ideal '' structures can be detected. Such deviations indications of which were observed in cis-decalin are to be expected also in other com- pounds containing two cyclohexane rings in the cis-position because some hydrogen-hydrogen distances would be a little smaller than 1.9 A if all valency angles remain strictly " tetrahedral ". The establishment of the molecular structure of cis-decalin has already had interesting consequences in the chemistry of compounds containing a greater number of condensed six-membered rings of the cyclohexane type.Before closing this short survey we think it necessary to draw atten- tion to one special point So far not a single case has been found in which the cyclohexane ring appears in the boat form (except of course when a bridge is formed between 1 4-carbon atoms). It seems likely however that rings may be stabilised in this form in certain special cases. If one or more methylene groups are replaced by carbonyl groups for example or by single bivalent atoms some of the arguments against the boat form are weakened. The findinglg that cyclohexane-1 4-dione shows a considerable dipole moment in solution is probably not a result of partial enolisation but indicates rather the presence of a considerable proportion of molecules with a carbon ring of lower symmetry than that of the chair form. Here X-ray analysis of the crystal would be of great interest but it has not so far been carried out. l* l7 0. Hassel Tidsskr. Kjemi 1943 3 91. l* The crystals probably belong to tho space group P2 with 2 molecules in the l9 C. G. LeFhvro and R. J. W. LoFGvre 3. 1935 1696. unit cell.

ISSN:0009-2681    

DOI:10.1039/QR9530700221

出版商:RSC    

年代:1953

数据来源: RSC

摘要:

STEROIDAL ALKALOIDS By JAMES MCKENNA M.Sc. PH.D. A.R.I.C. (LECTURER IN CHEMISTRY THE UNIVERSITY SHEFFIELD) SEVERAL alkaloids now recognised to possess the characteristic steroidal carbon framework have been known for nearly a century. Little progress however could be made in earlier investigations with these bases owing to lack of fundamental knowledge of the better known steroids. The clucida- tion of the molecular structures of cholesterol and cholic acid in 1932 made the way clear for rapid and successful researches with other classes of steroids of great interest and in 1936 Soltys and Wallenfels,l observing that the potato alkaloid solanidine in alcohol formed an insoluble adduct with digitonin dehydrogenated the related base solanthrine with selenium and obtained the well-known Diels hydrocarbon 3‘-methyl-1 2-cyclo- pentenophenanthrene (I).Other SoZanum and Veratrum alkaloids were subsequently also demonstrated to be steroids and in 1948 conessine and related Hohrrhena bases were shown to belong to the same class.2 Interest in these alkaloids has increased considerably within the past few years; until recently the structural formula could not be written with certainty for any of them but the position has now been reached where the structure of most of the important bases are either well established or likely to require modification only in detail. The work up to 1948 has been comprehensively reviewed by Henry 3 and by Fieser and Fieser.4 The present Review is not intended to be compre- hensive and emphasis is focused on the more important recent develop- ments.Classification.-Steroidal alkaloids may be divided into three groups on the basis of botanical source (a) solanum alkaloids obtained from 8. tuberosum (potato) 8. Zycopersicum (tomato) and other species ; ( b ) veratrum alkaloids usually from V . aZburn and V . viride (European and American “ hellebore ”) or from sabadilla ( V . sabadiZZa) seeds ; ( c ) kurchi or holarrhena alkaloids from ‘‘ kurchi ” ( H . antidysentericu) a small Indian shrub and from other Indian and African species of Holarrhena. All the kurchi alkaloids of well-authenticated structure contain a carbon framework of twenty-one atoms similar to that of the steroid hydrocarbon pregnane (11); owing however to the presence of N-methyl groups the number of carbon atoms in the molecules of these alkaloids may rise to twenty-four (the alkaloid conkurchinine may contain twenty-five carbon atoms in the molecule; see p.253). Ber. 1936 69 811. Haworth McKenna and Singh Nature 1948 162 22 ; J . 1949 831. “ T h e Plant Alkaloids ” J. and A. Churchill Ltd. London 4th Edn. 1949 pp. 661-672 700-715 742-749 where full references t o the earlier work are given “ Natural Products related t o Phenanthrene ” Reinhold Publ. Corp. New York 3rd Edn. 1949 pp. 597 et seq. 231 232 QUARTERLY REVIEWS The solanum and veratrum alkaloids include glycosidic and ester types as well as the simple bases (alkamines) ; the parent bases also readily obtainable by hydrolysis of the corresponding esters or glycosides all contain twenty-seven carbon atoms in the molecule.* The carbon framework is usually that of cholestane (111) but there is strong evidence that at least two important veratrum alkaloids contain the modified cholestane frame- work illustrated by formula (IV).The aconite delphinium and erythro- phloeum alkaloids in which the existence of a steroid nucleus (even in a modified form) has not been demonstrated are excluded from this Review ; Me Me these bases are probably more closely related to the diterpenoids. How- ever the Fritillaria alkaloids (ref. 3 pp. 732-734) of the Chinese drug " pei-mu " may ultimately prove to be steroids. General Methods of Structural Investigation.-Hofmann Emde and cyanogen bromide degradations are among the most typical procedures for investigating alkaloid structures but only in the case of conessine have these methods been applied with success. As with other polycyclic hydro- aromatic products dehydrogenation by selenium has been of outstanding value not only for identification of the carbon framework but also with the veratrum and solanum alkaloids for the provision of information about the heterocyclic portion of the molecule.Diels's hydrocarbon (I) is not obtained by dehydrogenation of all the steroidal alkaloids in some cases presence of a substituent a t or near C(ls) results in the production of a closely related 1 2qcZopentenophenanthrene ; and the alkaloids possessing the modified cholestane framework (IV) naturally cannot yield the normal product (I). The more important kurchi bases are notable both as natural steroids and as alkaloids not containing oxygen; many veratrum and solanum alkaloids however contain the 3,!l-hydroxy-group and 5 6-double bond so characteristic of other steroids and the presence of this composite * Two alkaloids veratrobasine and geralbine recently isolat,ed by Stoll and Seebeck ( J .Arner. C'?Lem. Xoc. 1952 '74 4728) from V . album have the molecular formulz C,4H,,0,N and C,,H,,O,N respectively and thus constitute an exception to this rule. Each alkaloid contains one N-methyl group likewise an exception among the veratrum and solanum alkaloids ; geralbine may thus be related to the kurchi alkaloids. McKENNA STEROIDAL ALKALOIDS 233 group is readily demonstrable in most cases by precipitation of the alkaloid with digitonin and by dehydration oxidation and reduction. In some cases valuable information has been obtained by removal of the hetero- cyclic portion of the molecule by oxidation or by otherwise establishing a close connection with a different type of steroid.Solanum Alkaloids.-Solmidine. This base of molecular formula C,7H4,0N is obtained by acid hydrolysis of the glycoalkaloid solanine C45H73015N the usual source of which is potato shoots. The sugar portion of the solanine molecule is a glucosylgalactosylrhamnose as shown by identification of the component monosaccharides and by partial hydroly- sis ; 6 6 the aglycone solanidine is glycosidically linked to the glucose unit of this unusual oligosaccharide. The correct molecular formula for solani- dine was first given by Schopf and Hermanna7 Solanidine may readily be shown to contain a secondary alcoholic function a double bond and a tertiary amino-group. The alkaloid does not contain a N-methyl or N - ethyl group an indication that the nitrogen atom is common to two rings and the resistance of the base to Hofmann degradation * 1~ 7 9 has been regarded as additional evidence for this supposition.The secondary hydroxyl group is eliminated as water when solanidine hydrochloride is heated 7 and a doubly unsaturated base solanthrene or solanidiene C,,H,,N is produced ; some of this base (in which the double bonds are in conjuga- tion) is also formed during acid hydrolysis of the glycoalkaloid. Solanidine gives a precipitate with digitonin and (like solanthrene) yields the Diels hydrocarbon (I) on selenium dehydrogenation. A steroidal structure for the alkaloid is thus indicated. Additional valuable information is afforded by the reactions of the double bond and secondary hydroxyl group l o in solanidine which are closely analogous to well-known characteristic reactions in rings A and B of cholesterol.Some of these transformations are shown in the annexed Scheme and similarities in specific rotation among compounds of each series are shown in the Table. All values of [a], in the Table are for solution in chloroform and most of them are quoted by Prelog and Szpilfogel.lo The relation between molecular rotation (M[a],/100) and structure in the steroid field has recently been studied in detail particularly by Barton and his co-workers (for a general account see Barton and Klyne) ; l1 in the present instance a simple comparison of specific rotations illustrates the point. Gerecs and Zemplen Ber. 1928 61 2294. Caronna and Oddo Ber. 1934 67 446.Rochelmeyer Arch. Pharm. 1936 274 543 ; Craig and Jacobs J . B i d Chem. lo Geyer Rochelmeyer and Shah Ber. 1938 '71 226 ; Rochelmeyer Arch. Pharm. l1 Chem. and Ind. 1948 755. * Solanidine has generally been recovered unchanged on attempted Hofmann decomposition of its metho-salts but in one case an isosolanidine is reported to have been formed ; this isomer may possibly result from recyclisation of an intermediate unsaturated base as in the case of heteroconessine (see p. 252). Ber. 1933 66 298. 8 Soltys ibid. p. 762. 1943 149 451. 1939 277 340 ; Prelog and Szpilfogel Helv. Chim. Ach 1944 27 390. 234 QUARTERLY REVIEWS (VIII) I HO CrO 3 4 13 -+ 4 F-I + Me I and Compound solanidine series I I Partial formula (XIII) . . . I Solanidine Solanidan-3/?-01* Solanidan- 3-0119 Solanidan- 3 a - 01 * Solanid-4-en-3-one 58-Solanidan-38-olt 58 -Solanidan- 3 a -01 Solani-3 5-diene (solanthrene) Solanidane [a] - 27” + 28 + 46 + 32 + 89 + 28 + 35 - 92 + 33 Compound cholesteral series Cholesterol Cholestan-3j3-01 Cholestan- 3 -one Cholestan- 3a-01 Choles t -4-en-3 -one Coprostan- 38-01 Coprostan-3 a-01 Cholesta-3 6-dime Cholestane - 37” + 28 + 40 + 32 + 89 + 28 + 31 - 123 + 24 * An a-configuration for substituents or angular hydrogen atoms in rings A to D in steroids indicates a configuration trans t o the C(18) methyl group ; a ,%configuration is cis.These configurations are written with broken and full lines respectively in structural formula t The prefix ‘’ all0 ” has been widely used in the solanidine series to indicate &-A/B ring fusion (5p-hydrogen) but is contrary to modern practice (see J .1951 3528 rule 3.7). The chemical and physical properties of all the solanidine derivatives listed in the Table are in full accord with the partial structural formule McKENNA STEROIDAL ALKALOIDS 235 shown. All six 3~-hydroxy-compounds (but not the 3a-epimers) form precipitates with digitonin and this is known to be the general rule with isomeric 3-hydroxy-steroids. The similarity between the two series of KMnO - derivatives strongly suggests that solanidine and cholesterol are structurally identical around rings A and B and indeed among the various formuh proposed for the alkaloid from time to time there has been no disparity on this point. l2 for solanidine showed considerable variations in the representation of the heterocyclic portion of the molecule about which a t first little information was available ; the demonstration by Rochelmeyer l2 that acetylsolanidine did not yield a C(,,,-ketone on oxidation with chromic acid suggested however that this part of the molecule is linked to ring D by more than one bond.In the absence of positive evidence from Hofmann degradations the first definite indication of the structure of the heterocyclic ring was the isolation of 2-ethyl-5-methylpyridine (XIV) as a product of Earlier formulz CO,H /I- 1.1.) /\ /I\ HO " H" (XVIII) l2 Clemo Morgan and Raper J . 1936 1299 ; Rochelmeyer Arch. Phurm. 1942 280 453. 236 QUARTERLY REVIEWS selenium dehydrogenation. l3 This base had previously beeh encountered as a degradation product of the veratrum alkaloids in the first instance by the distillation of cevine with zinc dust; l4 its structure was proved by oxidation to pyridine-2 5-dicarboxylic acid by the non-identity with the previously known 2-methyl-5-ethylpyridine7 and by synthesis.On the strength of the additional evidence and because of a possible structural analogy with the steroidal sapogenins Prelog and Szpilfogel proposed formula (XV) for solanidine and suggested that a correlation with sarsasa- pogenin (XVI) might prove feasible. This possibility had independently become apparent to Jacobs and Uhle who converted sarsasapogenoic acid (XVII) an oxidation product of sarsasapogenin into 5P-solanidan- 3p-01 (XVIII) by the reactions indicated in the scheme on p. 235. This synthesis the nearest approach so far * to the partial synthesis of any steroidal alkaloid from a steroid of different type provides rigid proof for formula (XV) for solanidine.The formula is stereochemically certain so far as rings A-D are concerned ; the nitrogen linkage to ring D may be assigned the /3-configuration because of the known P-configuration of the C(l,) side- chain. Neither the orientation of substituents attached to the heterocyclic rings E and F nor the'configuration at C(22) can be discussed at present for solanidine or any other steroidal alkaloid. The glycoalkaloid demissine C,oH8,020N was isolated by Kuhn and Low l6 from the leaves of the Mexican potato 8. demissum. On acid hydrolysis it yields glucose' (2 mols.) xylose (1 mol.) galactose (1 mol.) and demissidine C2,H4,0N. Demissidine is identical with solani- dan-3p-01 (XIX) also obtainable from solanidine (XV) by catalytic hydro- genation and is epimeric a t C(5) with Craig and Jacobs's synthetic product Demissidine.(XVIII). Tomatidine. The leaves of the red currant tomato (Lycopersicon pim- pinellifolium) and of various species of wild tomato plants contain tlk glycoalkaloid tomatine C,,H,,O,,N which on acid hydrolysis yields tomati- dine C2,H4,0,N and a series of sugars identical with that obtained from demissine (above). l7 Tomatidine contains a hydroxyl group an unreactive 13 Prelog and Szpilfogel Helv. Chim. Acta 1942 25 1306 ; Craig and Jacobs Science 1943 9'7 122. 1 4 Craig and Jacobs J. Biol. Chem. 1937 119 141 ; 1937 120 447 ; 1938,124,659. l5 Ibid. 1945 160 243. l6 Chem. Bet-. 1947 80 406. 17 Doolittle Fontaine Irving Ma and Yoole Arch. Biochem.1948 18 467 ; Doo- little and Fontaine Ind. Eng. Chem. 1948 26 2440 ; Pontaine and Ma Arch. Biochem. 1950 27 461 ; Kuhn and Low Chem. Ber. 1948 81 552 ; Gauthe Kuhn and Low ibid. 1950 83 448. *Added in Proof.-This is no longer correct. See footnote on p. 239. McKENNA STEROIDAL ALKALOIDS 237 oxygen atom and a secondary amino-group but no double bond. Catalytic hydrogenation or reduction with lithium aluminium hydride gives dihydro- tomatidine C,,H,,.O,N a base containing two hydroxyl groups. Tomati- dine gives a precipitate with digitonin. The molecular formula and some of the reactions suggested a steroidal structure,l* and this was confirmed in the following manner. The base (XX) yields an ON-diacetyl derivative (XXI) containing the groups OAc and >NAc (infra-red absorption; no active hydrogen) which on exposure to ultra-violet light or long boiling in acetic acid gives an isomeric diacetyl derivative (XXII) containing the groups OAc and NHAc (infra-red absorption ; one active hydrogen) ; during the isomerisation a nitrogenous ring has evidently been opened.Chromic acid oxidises the isomers (XXI) and (XXII) to respectively the acetoxy- lactone (XXIII) and a mixture of this with 3~-acetoxyctZlopregn-lG-en-20- one (XXIV). The same ketone is obtained by oxidation of a triacetyl- tomatidine (XXV) formed by acetylation under more drastic conditions ; Ac,O reflux J. AcO- C27H420 -NAc 1 (XXI) (XXII) (XXV) HO- >NH (XX) (1) CrO (2) Hydrol. (3) Acetyln. the triacetyl derivative contains the groups OAc and NAc (infra-red absorption ; no active hydrogen).The lactone (XXIII) and the acetoxy- pregnenone (XXIV) had previously been encountered as oxidation products of acetyltitogenin (XXVI) and diacetyl-y-titogenin (XXVII) respectively. From these and other reactions a close structural analogy between tomati- dine and the steroidal sapogenins was apparent and formula (XXVIII) Ard Fontaine and Ma J. Amer. Chem. SOG. 1951 73 879. 19 Katz Mossettig and Sato ibid. p. 880 ; Kuhn and Low Chem. Ber. 1952 85 416. R 238 QUARTERLY REVIEWS was advanced for the alkaloid. This was confirmed by Kuhn Low and Trischmann O by dehydrogenation of dihydrotomatidine to 2-ethyl-5- methylpyridine (XIV) and conversion of tomatidine into demissidjne (XIX) I't-E tC)H by the series of reactions illustrated. The reduction of tomatidine to dihy- drotoinatidine (XXIX) is paralleled by similar reactions of steroidal sapo- genins e.g.(XXVI) in which one of the spiroketal oxide linkages may likewise be opened by hydrogenolysis ; these reductions are possible because of the attachment of the second oxygen atom (or the nitrogen in tomatidine) to the spiro-carbon atom Xolasodine. Hydrolysis of the glycoalkaloid solasonine C,5H730,6N from the green berries of the shrubs X. sodomeum (Dead Sea apple) 8. avicuZare (Maori " poro-poro ") and other species yields solasodine C,,H4,03 together with one mol. each of glucose galactose and rhamnose. The order of the sugar units is probably the same as in solanine.21 The correct molecular formula? for solasonine and its aglycone were established by Rochelmeyer 22 and by Briggs and his co-w0rkers.2~ Solasodine has more recently been obtained by hydrolysis of solmargine,24 C,,H,,O,,N (from X.marginatum) in which the order of the sugar units (two of rhamnose and one of glucose) is indicated by partial hydrolysis to solasodine ghcoside. Solasodine forms an insoluble adduct with digitonin and yields the Diels hydrocarbon (I) on dehydrogenation with selenium ; 25 one of the pyridine bases also formed in this reaction is probably 2-ethyl-5-methylpyridine (XIV) which has also been identified as a dehydrogenation product of solasodine g l u ~ o s i d e . ~ ~ These results indicat'e that the alkaloid has a steroidal structure with a heterocyclic arrangement probably closely related to that in solanidine. Two active hydrogen atoms are present one of which is accounted for as a secondary hydroxyl group.The alkaloid also contains 20Kuhn and Low 1953 86 372; Angew. Chem. 1952 64 397. 21 Briggs and Carroll J . 1942 17. 22 Arch. Pharm. 1939 277 329. 23 Briggs Newbold and Stace J . 1942 3. 2 * Briggs Brooker Harvey and Odell J . 1952 3587. 25 Rochelmeyer Arch. Phnrm. 1937 275 336. McKENNA STEROIDAL ALKALOIDS 239 a double bond and the reactions of the hydroxyl group and the double bond together with the spectral and rotational 26 characteristics of the products give a strong indication of the presence of the familiar 5 6-unsaturated 3p-hydroxy- arrangement (as in solanidine and cholesterol). A series of derivatives corresponding to those listed in the Table on p. 234 may also be obtained from solasodine e.g. solasod-4-en-3-one and solasoda-3 5- diene with partial formulz (IX) and (XII) respectively.Erroneous conclusions were at first reached by Briggs and his CO- workers 23 regarding the function of the second oxygen atom and the basic centre in solasodine and the alkaloid was formulated as a carbinol-amine (XXX) related to an earlier structure for solanidine. The base however was later shown to be a secondary amine and formula (XXXI) was pro- posed 2 7 by analogy with the structure of sapogeniiis such as diosgenin (XXXII). The properties and reactions of solasodine are readily explicable on the basis of formula (XXXI). Thus the alkaloid is a weak base,28 difficult to acetylate owing to the inductive effect of the cyclic oxygen atom. Catalytic hydrogenation in the presence of palladium-charcoal gives solaso- danol (dihydrosolasodine ; XXXIII) but when a platinum catalyst is employed a second mol.of hydrogen is taken up ; the product dihydro- solasodanol (XXXIV) contains an additional hydroxyl group and yields a triacetate (cf. tomatidine and the steroidal sapogenins). The oxygen ring may also be opened by treatment with lithium aluminium hydride. Treat- ment of solasodine with acetic anhydride followed by oxidation with chromic acid hydrolysis with methanolic potassium hydroxide and re- acetylation gives a mixture g of 3p-acetoxypregna-5 16-dien-20-one (XXXV) and 3~-acetoxy-16a-methoxypregn-5-en-20-one (XXXVI) ; t,he methoxy-compound is evidently an artefact formed during the hydrolysis. Most of these reactions fail to distinguish unambiguously between the suggested structure (XXXI) for solasodine and the alternative (XXXVII).A preliminary not yet confirmed of conversion of N-nitrososolaso- dine into diosgenin (XXXII) in low yield by treatment with aqueous acetic acid is equally inconclusive. However dehydrogenation of the alkaloid or its glucoside to 2-ethyl-5-methylpyridine (XIV) seems decisive. Never- theless while the general nature of the heterocyclic arrangement in solasodine is clear further evidence on the structure 6f this part of the molecule is desirable. For example an established relation between solasodine and tomatidine (XXVIII) would be of great interest ; the tentative suggestion 27 that tomatidine and solasodanol (XXXIII) may be identical has not been confirmed and indeed there are appreciable discrepancies in the physical properties of the two bases and their derivatives.* A stereoisomeric relation between solasodine and the aglycone 2G Briggs and O’Shea J.1952 1654. Briggs Harvey Locker McGillivray and Seeyls J. 1950 3013 ; Briggs and 29 Miller Mosettig and Sato J . Amer. Chem. SOC. 1951 73 5009. * Added in Proof.-The conversion of kry-ptogenin into solasodine recently reported by Uhle ( J . Amer. Chern. Soc. 1953 75 2280) indicates that (XXXI) for solasodine is structurally oorrect ; tomatidine and solasodanol may thus he epimeric e.g. a t C(22). Locker ibid. p. 3020. Cf. Bloom and Briggs J. 1952 3591. 240 QUARTERLY REVIEWS -OH solauricidine (from S. auriculatum) has been suggested by Briggs and his ~o-workers,~~ but the data do not furnish clear evidence that these two bases are distinct individuals. Solanocapsine. The leaves of 8. pseudocapsicum contain the base solanocapsine C,,H,,O,N, apparently not in glycosidic combination.This alkaloid has been investigated by Barger and Fraenkel-C~nrat,~~ and by Schlittler and Uehlinger.32 Solanocapsine yields Diels's hydrocarbon (I) 30 Bsll Briggs and Carroll J . 1942 12. 31 J . 1936 1537. 32 Helv. Chim. Acta 1952 35 2034. McKENNA STEROIDAL ALKALOIDS 241 and 2-ethyl-5-methylpyridine (XIV) on selenium dehydrogenation but differs from the other solanum alkaloids in giving no precipitate with digitonin and in containing two basic centres. There are also present a hydroxyl group and an inert (ethereal) oxygen atom but no double bond. Treatment with nitrous acid eliminates a primary amino-group (with the introduction of a double bond) and nitrosates the other basic centre which is therefore secondary.The hydroxyl group cannot be acetylated elimin- ated with acid reagents or oxidised to a carbonyl group ; it is regarded as being tertiary in character. Although hydrogenolysis of the oxide ring has not been demonstrated satisfactorily Schlittler and Uehlinger suggest structure (XXXVIII) analogous to that of the other secondary solanum bases. The behaviour of the hydroxyl group indicates that it is not in position 3 and it is thus possible that the amino-group may be situated there as in some kurchi alkaloids. Veratrum Alkaloids.-Three principal sources ( Veratrum spp.) of this group of natural steroidal bases are named on p. 231 ; of these each of the two " hellebores " contains almost all the more important alkaloids free or in combination whereas sabadilla seeds yield mainly esters of cevine.Rubijervine and isorubijervine. These two isomeric tertiary bases C ,H,,O,N are neither epimers nor otherwise readily interconvert,ible as their names might suggest. Each base contains a double bond and two hydroxyl groups. The presence of the 5 ; 6-unsaturated 3/l-hydroxy-system is indicated 33 by positive digitonin reactions and by conversion of the bases via the 4 5-unsaturated 3-ketones into the corresponding unsaturated 3-alcohols which give charactmeristic purple colours with trichloroacetic acid (Rosenheim reaction). Cholesterol behaves analogously. Dehydrogenation of isorubijervine 33 yields 2-ethyl-5-methylpyridine (XIV) and 1 2-cyclo- pentenophenanthrene (XLVIII). The alkylpyridine is also obtained from rubijervine but in this case the neutral dehydrogenation product is a hydrocarbon C18H16 which has not been conclusively identified but closely resembles 5'-methyl-1 2-cyclopentenophenanthrene (for numeration see XLVIII) ; a phenol C18H160 possibly related is also formed.34 These results suggested that rubijervine and isorubi jervine were hydroxysolani- dines and this was confirmed by conversion of each base into solanidine (XV) and solanidan-3,801 (XIX) by the reaction scheme 359 3 6 3' shown.33 Craig and Jacobs J. Biol. Chem. 1945 159 617. 34 Idem ibid. 1943 148 41. 35 Jacobs and Sato ibid. 1949 179 623. 36 Jacobs and Pelletier J . Amer. Chem. SOC. 1952 74 4218 ; Burn and Rigby 37 Burn and Weisenborn Abs. Amer. Chem. SOC. Meeting Sept. 1952 No. 2 3 ~ ; Chem. and Ind. 1952 668; J. 1953 963. J . Amer.Chem. SOC. 1953 75 259. 242 QUARTERLY REVIEWS The additional hydroxyl group in rubijervine (XXXIX) is regarded as being attached at the C(lz) position (a-configuration) for the following reasons 35 ( a ) all positions in rings A and B are ruled out since rubijervine is not an a-glycol the corresponding dihydro-diketone (XL) is not a 1 3- diketone and the 3-hydroxy-ketone (XLI) which yields solanidine (XV) on Wolff-Kishner reduction does not contain an ap-unsaturated carbonyl group ; (b) the phenol C,,H,,O [probably a derivative of cyclopentenophen- anthrene (XLVIII)] rather than a hydroxypyridine (cf. jervine and vera- tramine below) is formed on dehydrogenation of rubijervine ; this together with the full carbonyl reactivity of the basic diketone (XL) exclude positions C(ll) (authentic C(,,)-ketones are very inert) C(15) (also unlikely biogenetic- ally) C(2s) C(2a) and C(26) (c) optical rotational data favour the assigned (12a-)positioii for the hydroxyl group.As it is not at all evident however how 5’-methylcyclopentenophenanthrene could be derived from (XXXIX) further investigation of the hydrocarbon C,,Hl is desirable. The additional hydroxyl group in isorubijervine (XLII) is primary dihydroisorubi jervine (XLIV) gives the aldehyde (XLIII) or a correspond- ing keto-acid on oxidation with chromic acid. The methyl ester (XLV) of the keto-acid is very resistant to hyclr~lysis,~~ indicating that it may be the ester of a tertiary carboxylic acid. [Strophanthidin derivatives carrying an angular carbometlhoxy-group between rings a and B in place of the usual angular methyl group are more readily saponified than the ester (XLV) .] The reaction product (XLVI) of isorubijervine and toluene-p- sulphonyl chloride and the corresponding iodide (XLVII) were formulated as the normal 18-toluene-p-sulphonate and primary halide respectively by Pelletier and but the infra-red spectra and neutrul character of these coinpounds show them to be quaternary ammonium salts ; 37 the alkaline reduction of the quaternary iodide to solanidine is a kind of Kmde degradation.The formation of these quaternary salts indicates that the site of the primary hydroxyl group in isorubijervine is within bonding dist’ance of the nitrogen atom and of the three possible positions [C(ls) C(zl) C(27) ; see (XLII)] C(18) is the most favoured stereochemically. As the same position (adjoining the tertiary C(13)) also appears very likely on other grounds (see above) formula (XLII) for isorubijervine may be regarded as well established.Jervine This interesting base occurs as the glucoside y - jervine,* C33H4908N and as the free alkamine C,,H,,O,N the correct molecular formula for which was established by Craig and Jacobs.39 Jervine is a heterocyclic secondary amine and contains a hydroxyl group and two double bonds one of which is a/3 to a carbonyl group. The third oxygen atom is present as a cyclic ether linkage. The presence in the alkaloid of C:C*CO is reflected in the characteristic ultra-violet absorption curve which has Amax. 250 mp(1og E 4.2). Partial hydrogenation of jervine gives 38 Jacobs and Sato J . Riol. Chem. 1951 191 63. 39 Ibid.1943 148 51. For further details of jervine chemistry see Craig Jacobs and Lavin ibid. 1941 141 51 ; Huehner and Jacobs ibid. 1947 170 635 ; Jacobs and Sato ibid. 1948 175 57. * Acid hydrolysis of y-jervine gives isojervine owing to isomerisation during the reaction. Hydrolysis of dihydropseudojervine however gives dihydrojervine. MtKENNA STEROIDAL ALKALOIDS 243 (XXXIX) 4W-K' JW-K A/ HO $/' (XLVII) NaI or KI r * The reduction of the 3-keto-group to a secondary alcoholic rather than a methylene group is abnormal. W-K = Wolff-Kishner. 244 QUARTERLY REVIEWS dihydrojervine the absorption of which (Amax. 295 mp ; log E 1.6) is typical of compounds containing an isolated carbonyl function. It follows that in the formation of dihydrojervine the double bond originally in conjugation with the carbonyl group is reduced preferentially ; further catalytic hydrogena- tion of dihydrojervine gives a tetrahydro-compound C,,H,,O,N.The ketonic group in jervine is unreactive towards most of the usual reagents but reduction of dihydrojervine with sodium and butanol yields dihydro- jervinol (isomeric with tetrahydrojervine) a compound containing an additional acylatable hydroxyl group. While t'his may be con- nected with the modified steroid skeleton described below it is of interest that the closely related alkamine veratramine gives a positive digitonin reaction (see p. 246). Nevertheless evidence for the presence of the familiar 5 6-unsaturated 30-hydroxy-steroid structure is provided by Barton's method of molecular rotation differences. For example the changes in molecular rotation on 0-acetylation and 0-benzoylation of N-acetyljervine and on hydrogenation of the double bond in dihydrojervinol are - 27" + 91" and + 228" respectively compared with average values of - 35" + 81" and + 243" for authentic 5 6-unsaturated 3P-hydroxy-steroids such as cholesterol.Oxidation of jervine dihydrojervine and dihydrojervinol gives 4 5-unsaturated 3-ketones recognisable as such by their ultra-violet spectra,* and by reduction with aluminium isopropoxide to 4 5-unsaturated 3 - hydroxy -amines which give positive Rosenheim reactions. Jervine does not form an insoluble digitonick. (XLIX) H Ac A HO Dehydrogenation of jervine with selenium yields 2-et hyl- 5 -methyl- pyridine (XIV) and a hydroxy- base C,H,,ON the ultra-violet absorption and colour reactions of which indicate that it is a /3-hydroxypyridine possibly Z-ethyl-3-hydroxy-5-methylpyridine (XLIX).By the same reaction there is also formed a mixture of unidentified hydrocarbons many of which appear on the basis of their ultra-violet absorption to be fluorene derivatives i.~. to contain non-terminab five-membered rings. The formation of fluorene * The newly introduced chromophore is an ap-unsaturated carbonyl system similar to that originally present in jervine so that the change in ultra-violet absorption on oxidation is particularly evident with the reduced bases. McKENNA STEROIDAL ALKALOIDS 245 rather than phenanthrene hydrocarbons on dehydrogenation a t first led Jacobs and his collaborators to consider the possibility that ring B in jervine was five-membered as in the partial formula (L) but subsequent proof for the presence of the 5 &unsaturated 3p-hydroxy-grouping in rings A and B ruled out structures of this sort and led instead t o the development of formula (LI).The lack of carbonyl reactivity in jervine and its derivatives suggested that the ketonic group was at C(ll); on this basis and on the assumption of a normal carbon skeleton the 8 9-position is the only one possible for the conjugated double bond. Jacobs and Sato's formula 39 (LI) however was not in satisfactory accord with the neutral dehydrogena- tion products of jervine and it was later shown to be untenable by Winter- steiner and his collaborators. 40 Oxidation of 5 6-unsaturated S-acetoxy- steroids with chromic acid yields the corresponding 7-ketones and ON- diacetyl-7-ketojervine prepared by this method had not the spectral characteristics expected of the unsaturated diketone (LII) with its extended conjugated system.Nevertheless the properties of this and related oxida- tion products leave no doubt whatever about the position of the newly introduced keto-group. Further studies 41 with jervine led to formula (LIII) which accodmodates adequately all the above reactions as well as those now to be described. Treatment of jervine with acetic anhydride and zinc chloride gives a product H (1) Acetyln. (2) ACOH-AC~O -H,SO )"d (LVII) AcO I H Me Me Alkali Me 40 Fried Iselin Moore and Winterstsiner Proc. Nat. Acad. Sci. 1951 3'7 333. d1 Fried Iselin Klingsberg and Moore J . Amer. Ghem. Soc. 1951 73 2970. 246 QUARTERLY REVIEWS H HG M- Unfurtunatoly st 6-keto-group has been omitted from formula (LVIII).(LIV) which on oxidation with chromic acid yields acetaldehyde and the yellow 1 4-diketone (LV) with the expected absorption characteristics. This diketone on treatment with alkali gives the phenol (LVI) the ultra- violet absorption of which is closely similar to that of the cestrane derivative (LVIII). Milder acetolysis of ON-diacetyljcrvine gives a product formulated as (LVII) the indanone structure in which may also be demonstrated spectroscopically. These acetolysis reactions indicate that the carbonyl group in jervine is directly attached to a six-membered ring n but a poly- hydrochrysene structure (LIX) may be excluded as follows (a) the keto- group in (LIX) would be expected to behave like an ordinary (reactive) steroidal C(12)-ketone ; ( b ) the infra-red spectrum 42 of diliydrojervine (LX) has a band at 1730 cm.-l typical of a keto-group in a five-membered ring ; (c) biogenetically (LIII) is more probable (see 11.254) ; (d) the established relation between jervine and veratramine (see p. 247) indicates that jervine has a methyl group at’tached a t C,,, in ring D (Wintersteiner’s numbering). The oxide linkage to ring D is considered to be attached to rather than to C(ls) (as e.g. in tomatidine) because of the ease of double-bond formation at this centre in the acetolysis reactions. Some further work on this point is desirable as well as confirmation of the structure of the nitrogen-free acetolysis products but formula (LIII) for jervine seems reasonably well established. The isomeric base contains a second acylatable hydroxyl group and its ultra- violet absorption is considerably modified but a profitable discussion of structure is hardly possible at present.Veratramine. Structurally related to jervine is the secondary base ~ e r a t r a m i n e ~ ~ C,,H,,O,N which contains two alcoholic hydroxyl groups a double bond and a benzene ring. .Veratramine occurs free or as the glucoside veratrosine. The presence of the aromatic ring is shown by the ultra-violet absorption curve (Amax. 268 mp ; log E 2.8) and by the prepara- tion of an aromatic nitro-compound from OON-triacetyldihydroveratramine. The ultra-violet absorption maximum of veratramine is not altered on hydrogenation to the dih-ydro-derivative and the double bond is therefore not in conjugation with the benzene ring.Veratramine forms a precipitate with digitonin and the presence of a 5 6-unsaturated 3P-hydroxy-grouping Jervine is isomerised to isojervine by methanolic hydrochloric acid. 4 2 Anliker Heusser and Jeger HeZv. Chim. Actu 1952 35 838. 43 Craig and Jacobs J. Biol. Chem. 1945 160 555 ; Jacobs and Sato ibid. 1949 181 55; 1951 191 71. McKENNA STEROIDAL ALKALOIDS 247 may be demonstrated by conventional methods. Dehydrogenation yields a hydrocarbon C22H20 probably a chrysene or fluorene derivative and 3-hydroxy-5-methylpyridine (LXI) identified by its reactions [which resem- ble those of the base (XLIX) from jervine] and by its synthesis from the corresponding picolinesulphonic acid (LXII) . The presence of the aromatic ring in veratramine rules out a conventional steroidal structure and led Tamm and Wintersteiner to propose 44 the formula (LXIII) by analogy with jervine (LIII).Strong support for this striicture was obtained by oxidation of OON-triacetyldihydroveratramine to the indanone (LXIV) containing an unreactive carbonyl function; the same ketone may be obtained 45 by catalytic hydrogenation of the acetolysis product (LVII) obtained from OX-diacetyljervine as described above. The permanganate oxidation of CO,H HO 2 ~ @ \ CO,H HO,d II \/ NaOH 1 (LVI I) I KMnO /'v'\/' (LXI 11) I HO ( I ) HS-Pt ( 2 ) AcetyIn. (3) CrO veratramine to benzene-1 2 3 4-tetracarboxylic acid (LXV) recently reported 46 is further proof of the assigned structure. Incidentally the relation demonstrated between jervine and veratramine (which contains a preformed aromatic ring D) rules out the possibility that the six-membered ring in the acetolysis product's of jervine described above arises from molecular rearrangement during acetolysis.Cevine germine and protoverine. These three highly hydroxylated bases may be obtained by alkaline hydrolysis of a series of veratrum ester- alkaloids together with various organic acids e.g. tiglic angelic veratric a.cefic or substituted acetic acids. In spite of extensive investigations 4 4 Tamm and Wintersteiner J . Amer. C'hem. Soc. 1952 74 3842. 4 5 Hosansky and Wintersteiner ibid. p. 4474. 46 Hosansky Moore and Wintersteiner ibid. 1953 in the press. 47 Ref. 3 pp. 702-705 709-711. 248 QUARTERLY REVIEWS little is known about their structure and only brief mention is possible in this Review. Cevine and germine are isomeric and have the formula C2,H430sN ; protoverine contains an extra oxygen atom.All three alkaloids give similar products on dehydrogenation including 2-ethyl-5-methyl- pyridine (XIV) a base cevanthridine C,,H,,N and a phenol cevanthrol C,,H,,O. A series of hydrocarbons has also been obtained by dehydro- genation of cevine and of these one has been identified as 4 5-benzindane (LXVI). Among the chromic acid oxidation products of cevine and germine is a hexanetetracarboxylic acid which in spite of its simple molecular formula and obvious interest has not yet been conclusively identified.48 None of the three bases contains a double bond or gives a positive digitonin reaction. Cevine is not the parent base of the " veratrine " ester-alkaloids but is formed by alkaline isomerisation of a precusor cevagenin.49 Cevagenin unlike cevine contains a carbonyl group ; cevine however has st'rong reducing properties gives a dihydro-derivative on treatment with sodium and alcohol and is represented by Barton and Eastham 50 as containing the masked cc-ketol system (LXVII) possibly derived by isomerisation and lactolisation from the corresponding hydroxy-ketone arrangement (LXVITI) -$H- -CH (OH) - -Y-CH(OH)- - -CH(OH).CO- v A (LXVII) (LXVII I) ?a /\/ I1 I OH in cevagenin. A similar relation may exist between germine and protoverine and their respective ketonic alkaline-isomerisation products isogermine and isoprotoverine. The most abundant and important of the kurchi alkaloids is conessine C2aH40N2 the correct molecular formula for which was first given by Warnecke.,l Many workers have obtained conessine from the seeds or bark of Holarrhena antidysenterica ; it may also be obtained from H .africana H . congolensis H . Wulfsbergii or H . febrifuga. Conessine is an unsaturated diacid base and on oxidation with potassium iodate in dilute sulphuric acid or by other methods yields the characteristic derivative dioxyconessine (more correctly dihydrodihydroxyconessine) C,H,,O,N,. The presence of one double bond may be demonstrated by catalytic hydrogenation to dihydroconessine C24H42N2 and since the alkaloid is a ditertiary base containing three N-methyl groups it follows that one of the nitrogen atoms is present as a dimethylamino- and the other as a cyclic methylimino-group. 5 2 9 53 These results may be expressed as (LXIX) where the nucleus contains one double bond.Kurchi Alkaloids.-Conessine. 48 For a recent discussion see Elming Jeger Prslog and Vogel Helw. Chim. Acta 40 Seebeck and Stoll ibid. p 1270 ; cf. however Jacobs and Pelletier J . Org. Ghem. 5 2 Spgth and Hromotka Ber. 1930 63 126. s3 Kanga Ayyar and Simonsen J. 1926 2123. 1952 35 2541. 1953 18 765. 50 J. 1953 424. 51 Ber. 1886 19 60. McKENNA STEROIDAL ALKALOIDS 249 Proof that conessine is a steroidal alkaloid was derived in the first in- stance by selenium dehydrogenation of an unsaturated hydrocarbon or (more probably) a mixture of hydrocarbons C,,H,, obtained by dry dis- tillation of conessine dihydriodide in a reducing atmosphere. From the dehydrogenation mixture was isolated 3'-ethyl- 1 2-cyclopentenophenan- threne (LXX) possibly in admixture with a little of Diels's hydrocarbon (I).The structure of the hydrocarbon (LXX) was demonstrated by its synthesis by methods analogous to those used for (I). The notable preponderance of the ethyl homologue in the dehydrogenation mixture is due in this case to special structural conditions (see below). The detailed chemistry of conessine has been elucidated mainly by examination of products obtained on Hofmann decomposition of the alkaloid and its derivatives. Degradation of conessine by this technique was first investigated in detail by Spath and H r ~ m a t k a ~ ~ following valuable earlier work by Kanga Ayyar and Simonsen. 53 When conessine dimethohydrox- ide (LXXI) is heated water and trimethylamine are eliminated and apoconessine (LXXII) is formed. This contains a dimethylamino-group (which is evidently derived from the methylimino-group in conessine) and +lhree double bonds of which.two are introduced during the Hofmann decomposition one is formed by fission of the conessine heterocyclic ring and the other by elimination of the dimethylamino-group of the alkaloid as trimethylamine. Two of the three double bonds present in apoconessine are in conjugation. Partial Hofmann decomposition of conessine results in + t (1) hfeI Ilfc,SC,,H,,>NMe -ZH,O (2) .AgP 140- HO- -NhIe Me,N.C,,I-I,,>NMe ~- > -> C,,IT,,.NMe (LXlX) I (LXXI) (LXXII) - NMe L-H*o ( 1 ) Me1 (2) AgCl 250 QUARTERLY REVIEWS elimination of the dimethylamino-group as trimethylamine and yields metho-salts (e.g. LXXIII) containing a heterocyclic ring and hence only two double bonds ; since these two double bonds ara in conjugation (ultra- violet absorption as for apoconessine) the conjugated system in the metho- salts and in apoconessine is derived from the original conessine double bond and that formed by elimination of the dimethylamino-group as trimethyl- amine.The isolated double bond present in apoconessine is therefore formed by fission of the conessine heterocyclic ring.54 Reductive scission of apo- conessine methochloride (Emde degradation) gives pregna-3 5 20-triene (LXXIV) identified by its reactions including hydrogenation to pregnane (LXXV) and allopregnane (LXXVI),55 and by synthesis from pregna- 5 20-dien-3p-01 (LXXVII).56p 5 7 apoConessine is therefore a dimethyl- aminopregna-3 5 20-triene and having regard to the origin of the conju- gated diene system and isolated double bond in this derived base (see above) ikis evident that conessine may be represented by the partial structural formula (LXXVIII) in which a double bond and dimethylamino-group are situated near C(s) and a methylimino-group is attached to the C(,,)-side chain and to some other carbon atom in the neighbourhood.CH:CH Me I (Ts = p-C,H,Me.SO,) I -NMe ; double bond (1) N-Demethyln. and acetyln. (at NMe,) c2) Hofmann-Emde elimin. of .NMe (2) Ac,O ' i 'i' CH:CH (1) I-Iydrogn. (2) Hydrol. (.;) X-Methyln. Et H 5 4 Haworth McKenna and Whitfield J. 1949 3127. 5 5 Haworth McKenna Powell and Woodward J . 1951 1736. 6 6 Haworth and McKenna Chem. and Iizd. 1951 312. 57 Haworth McKenna and Whitfield J. 1953 1102. McKENNA STEROIDAL ALKALOIDS 251 For the investigation of the position of the double bond and dimethyl- amino-group in conessine a series of reactions was devised 5 7 to eliminate the heterocyclic methylimino-group leaving the other two structural features intact.N-Demethylation of conessine gives the known subsidiary kurchi base isoconessimine (see p. 253) in which the dimethylamino-group of conessine is replaced by a methylamino-group. N-Acety1isoconessimin.e is a monoacid base and Hofmann degradation of the monomethohydroxide results only in fission of the heterocyclic ring ; Emde elimination of the result ant dimet hylamino -group then gives a N - ace t yl -N- me t hylaminopreg- nadiene (LXXIX) in which one of the double bonds (formed by ring fission) is a t the 20 21-position (side chain). The second double bond and the nitrogen atom in (LXXIX) obviously correspond to the double bond and dimethylamino-group in conessine.Since the physical properties (ultra- violet and infra-red spectra and molecular rotations) of conessine and a (LXXXIII) Me,N 1 ! I H,C CH:CH R'' H,C CHMe t i number of its derivatives suggest that the double bond of the alkaloid is a t the 5 6-position with the dimethylamino-group at the adjacent 3B-position 3~-N-acetyl-N-methylaminopregna-5 20-diene was synthesised from the toluene-p-sulphonate (LXXX) 58 and found to be identical with the cones- sine degradation product (LXXIX). The structure of (LXXIX) was con- firmed by reduction deacetylation and N-methylation to 3/3-dimethylamino- allopregnane (LXXXI) which was synthesised from allopregnan-3-one (LXXXII). These experiments therefore indicate that conessine is a derivative of 3B-dimethylaminopregn-5-ene (see formula LXXXIV for conessine) .The structure of the heterocyclic part of conessine i.e. the positions of attachment of the heterocyclic methylimino-group to the pregnane frame- work can be deduced as follows. 59 The methylimino-group involves C(no) or C(zl) (see partial formula LXXXIII) and some other carbon atom in the neighbourhood to which is also attached the dimethylamino-group in apo- conessine. Models show that the only possible positions are C(ls) @-orient- ated nitrogen linkage) C(15)(p) C(ls) (p) and C(ls)- Of these possibilities all Haworth McKenna and Powell ibid. p. 1110. 59 Favre Haworth McKenna Powell and Whitfield ibid. p. 1115. 252 QUARTERLY REVIEWS but C(ls) may confidently be excluded C(15) on inter alia biogenetic grounds (this is not a proved position of substitution in steroids) ; C(16) because of the relation between conessine and heteroconessine (see below) ; C(15) C(lGl and C(lz) because neither apoconessine nor its hexahydro-derivative gives a hydrocarbon on attempted Hofmann decomposition indicating the probable absence of a /3-hydrogen atom (to the nitrogen).(It is true that trimethyl- ammonium groups attached by equatorial linkages [i.e. linkages such as 3/3 (A/B trans) which are roughly in the same plane as the ring system as a whole] would not be readily eliminated by Hofmann degradation but at least a small yield of hydrocarbon might be expected 5 8 . ) On the other hand strong evidence in favour of C(ls) as the position of attachment of the methylimino-group in conessine and the dimethylamino-group in apocones- sine is obtained from Kuhn-Roth C-methyl analysis,* which also indicates that the other position of attachment of the methylimino-group in conemine is C(20) rather than C(zl) (conessine like apoconessine has two C-methyl groups).Conessine and apoconessine may therefore be written with the structural formulx! (LXXXIV ; R = W' = R" = Me) and (LXXXV) respec- tively. There is no evidence in favour of the alternative structure (LXXXVI) for conessine previously proposed by Bertho. O By heating conessine dimethiodide under carefully defined conditions in alkaline ethylene glyc01,~~~ 5 9 the isomeric base heteroconessine is obtained NMe !\ , H H,C *C I I Me KOH on /\;/\ Me1 J /\I/\/ - - /-- Sdlt (LXXXVI I I) \/ (LXXXVII) Me,N - nfeoH +- +Nilre OH I,\.h e H2C C; ?Ye Me H2C *C \lAH \I>$ /- /- 11 (LXX XI X) The stereochemical representations at C(20,* merely illustrate the isomerism ; the configurations are not shown. instead of a Hofmann degradation product. heteroConessine and conessine give the same product conessimethine (LXXXVII) on partial Hofmann decomposition and (LXXXVII) reverts to heteroconessine in hot aqueous 6o Annalen 1950 569 1 ; 1951 573 210. * C-Methyl analyses are notoriously inaccurate but precautions were taken here to avoid erroneous conclusions. E.g. 3a-dimethylaminoaZZopregnane (isomeric with hexahydroapoconessine) was shown to contain three C-methyl groups (mols. of acetic acid found 2.17) whereas hexahydroapoconessine contains two (mols. of acetic acid found 1.29).McKENNA STEROIDAL ALKALOIDS 253 glycol or ethanol. By these and other reactions it has been shown that conessine and heteroconessine are C(,,)-epimers (partial formulae LXXXVIII and LXXXIX) and that the reactions described may be represented as illustrated on p. 252. On stereochemical grounds it is not possible to formulate conessine and heteroconessine with the methylimino-group engaging C!,,, since apart from the evidence of C-methyl analysis (which shows unambiguously that hetero- conessine like conessine contains two C-methyl groups) the existence of the isomerism shows that the heterocyclic ring in a t least one base must be joined to C(20) as there cannot be C(,,)-epimers. A second isomer neoconessine was first prepared by Siddiqui and Vashistha.61 The struc- ture of neoconessine is not yet known; the base contains a double bond which cannot be hydrogenated catalytically 5 5 and is probably fully substituted (infra-red spectrum).59 This base C23H38N2 which may also be obtained by demethylation of conessine with cyanogen bromide 62 (see p.251) contains two N-methyl groups and reverts to conessine on N-methylation with formaldehyde and formic a ~ i d . 6 ~ The demethylation reaction would be expected 6 2 to take place at the dimethylamino- rather than a t the methyl- imino-group in conessine and structure (LXXXIV ; R = H R' = R" = Me) is confirmed by the reactions described on p. 251 (degradation to 38- dimethylaminoaZZopregnane) . This alkaloid isomeric with isoconessimine also contains two N-methyl groups and yields conessine on N-methylation ; 63 its struc- tural formula is therefore (LXXXIV ; R = R' = Me R" = H).Demethylation of conessine or isoconessimine gives the di- secondary base conimine,6 which contains one N-methyl group. Conimine is reconverted into conessine on meth~lation,~~ and evidently has the structure (LXXXIV ; R = R" = H R' = Me). This interesting base C2,H3,0N2 has been isolated by several groups of 64 A hydroxyl and two primary amino- groups are present in the molecule together with one double bond and Siddiqui has suggested 6 5 that the nitrogen ring in the hkterocyclic kurchi alkaloids may be formed by the condensation of the hydroxyl group in holarrhimine with one of the amino-groups. In agreement with this Hofmann degradation of tetra-N-methylholarrhimine gives a cyclic ether C2,H3,0 which appears to be a steroidal 3 5-diene (ultra-violet absorp- tion).According to Bertho,66 acid decomposition of the base conkurchinine C2,H36N2 yields conkurchine C21H32N2 and an unidentified (Their " isoconessine ') was probably a mixture.) isoConessimine. Conessimine. Conimine. HoZarrhimine. Conkurchine. 61 J . S c i . I n d . Res. ( I n d i a ) 1945 3 559. 6 2 Siddiqui and Siddiqui J . I n d i a n Chem. SOC. 1934 11 787. 63 Siddiqui ibid. p. 283. 64 Siddiqui and Pillay ibid. 1932 9 553 ; Bertho Chem. Ber. 1947 80 316. 6 5 Proc. I n d i a n Acad. Sci. 1936 3 A 249. 6 6 Arch. Pharm. 1939 277 237; Annalen 1943 555 214; Chem. Ber. 1947 80 316. S 254 QUARTERLY REVIEWS four-carbon compound. Conkurchine is stated to contain two double bonds but no N-methyl groups. Hydrogenation gives successively di- and tetra- hydroconkurchine which on N-methylation yield respectively conessine and dihydroconessine.On the basis of this evidence conkurchine could be (LXXXIV ; R’=R’=R,”=H) with an additional double bond which Bertho regards as being situated in the heterocyclic ring. This hydroxy- base C,,H,,ON, has been isolated only from the bark of the African species H . ~ongolensis.~~ The alkaloid like conessine contains three N-methyl groups. The isolation of many other subsidiary bases from extracts of various species of Holarrhena has been reported but of these some have been inadequately characterised and the individualiky of others is open to doubt. In view of the well-established occurrence of solanum and veratrum alkaloids in glycosidic combination confirmation of Irani’s claim to have obtained a crystalline glycoalkaloid from an extract of “ kurchi ” seeds would be of particular interest.Biogenetic Considerations.-The structural analogy between solanum and veratrum alkaloids of known constitution and the steroidal sapogenins (which has been obvious to many investigators) is underlined by the con- version of sarsasapogenin into 5P-solanidan-3P-01 (p. 235) and is doubtless of biogenetic significance. An artificially prepared rearrangement product of rockogenin has the modified skeletal structure (IV) of jervine and vera- tramine and the relation between these alkaloids and others of more con- ventional structure is also apparent among other groups of natural products e.g, the triterpenes. Pharmacology.3-The most important uses have been (a) of conessine or kurchi extracts for the treatment (in India) of amoebic dysentery; and ( b ) af veratrum alkaloids as insecticides or (more recently) as hypotensive agents.It is interesting that wild tomato plants are protected by their alkaloid content (tomatine) against tomato wilt ; l7 and similarly the Mexican potato (8. demissum containing demissine) is not prone to attack by the potato beetle.16 Hohrrhenine. Other kurchi alkaloids. The Reviewer thanks Professor R. D. Haworth F.R.S. for his interest and Professor L. H. Briggs Dr. D. H. R. Barton and Dr. 0. Wintersteiner for prior information about recent work. 67 Pyman J . 1919 115 163. 68 Current Sci. (India) 1946 15 229. 69 Hirschmann Snoddy and Wendler J . Amer. Chem. SOC. 1952 74 2693.

ISSN:0009-2681    

DOI:10.1039/QR9530700231

出版商:RSC    

年代:1953

数据来源: RSC

摘要:

THE ASSOCIATION OF CARBOXYLIC ACIDS By G. ALLEN B.Sc. PH.D. and E. F. CALDIN M.A. D.PHIL. (THE UNIVERSITY LEEDS) I. Introduction THAT carboxylic acids are associated in some organic solvents and in the vapour has been known for more than seventy years and this type of molecular association has recently been much investigated. The equilibrium between single and associated molecules has been experimentally studied and a considerable body of quantitative data has accumulated. It has been found that in dilute solution (and probably also in the vapour state at low pressures) the equilibrium involves single and double molecules only. The structure of the dimeric molecule has been investigated by electron dif- fraction in the case of formic acid vap0ur.l It is generally recognised that in a dimer the two carboxyl groups are linked into a ring by hydrogen bonds /O --HO 2R*CO,H T R-C ‘c.n \OH --o/ The hydrogen bond in carboxylic acids appears from spectroscopic evidence to be stronger than that in alcohols water or phenol as might be expected from the tendency of the carbonyl group to withdraw electrons from the hydroxyl group and so to weaken the 0-H bond and increase its polarity.The dimerisation of carboxylic acids is thus a promising field for the study of hydrogen bonds in that the equilibrium can be accurately measured for a series of related compounds and the effect of substituents as well as of solvent and temperature systematically investigated. In this Review we shall try to summarise the experimental work on the association of carboxylic acids in the vapour phase and in solution and to show how far the data can be interpreted in terms of intermolecular forces.Hydrogen bonds cannot be regarded as covalent bonds involving bivalent hydrogen. From the theoretical point of view this would need orbitals of much too high energy ; and from the empirical point of view the heats of formation of hydrogen bonds are much less than those of covalent bonds. There is considerable evidence that hydrogen bonds are predominantly electrostatic and can be regarded as due to the interaction of dipoles ; we shall see that there is evidence for this in the behaviour of carboxylic acids. It will be worth while to recall briefly the various accounts of the hydrogen bond that can be attempted (a) An empirical description can be given 1 (a) Karle and Brockway J .Arner. Chem. SOC. 1945 67 898. and O’Gorman ibid. 1047 69 2638; cf. also ref. 28b. 2 Coulson “ Valence ” Oxford Univ. Press 1952 p. 301 et seq. 3 M. M. Davies Ann. Reports 1946 43 6. ( b ) Schomaker 266 256 QUARTERLY REVIEWS in terms of dipole interaction. The interaction may be further interpreted in terms of valence theory by either of the two usual methods giving accounts in terms of ( b ) molecular orbitals or ( c ) resonance. If we regard hydrogen bonds as essentially electrostatic each hydrogen bond in a carboxylic acid dimer may be represented as due to the interaction of two dipoles (a) Hydrogen bonds in terms of dipole interaction. a- a+ 6- 6+ -0-H o=c < The bond dipole moments of H-O- and 0-C< are of the order of 1-6 and 2.5 D re~pectively,~ so the electrostatic interaction at the observed 0-0 distance (about 2.7 A) should be considerable.The hydrogen atom is unique in its small size which allows the negatively charged end of a dipole to approach unusually close to the centre of positive charge so that the electrostatic attraction is abnormally strong. This explains why other atoms do not allow the formation of similar bonds and why esters ethers and alkyl halides are only weakly associated by comparison with carboxylic acids. Although electrostatic forces are in general undirected it is usual for only two atoms to be linked by hydrogen bonds. This is due no doubt to the small size of the hydrogen atom which leaves no room for the approach of another atom ; moreover the electrostatic attraction is strongest along the line of the dipole.The account of the bond in terms of electrostatic attraction is thus able to explain the pronounced tendency to form a dimer with a definite structure. This empirical description of the properties of the bond is to be interpreted by valency theory. As usual we can adopt either the molecular-orbital or the valence-bond (resonance) mode of approach. The hydrogen bond has been described in the language of molecular-orbital theory by M e ~ k e ~ ~ according to whom two steps are necessary in the formation of a hydrogen bond. Van der Waals forces loosen the O-H bond and as a result the proton “acquires quantum- mechanical bonding tendencies ” to both partners (without taking up a symmetrical position between them). Mecke concludes that a hydrogen bond is formed if a hydroxyl group with an active hydrogen atom approaches a group with easily polarisable n-electrons.In molecular-orbital terminology the bonding due to electrostatic attraction is not distinct in principle from covalent binding ; both involve a distortion of atomic or molecular orbitals and a resulting increase in molecular stability. No detailed calculations from molecular-orbital theory have been reported. ( c ) Resonance or valence-bond * interpretation. The alternative interpre- tation in terms of resonance attributes the presence of dipoles to resonance between ionic and non-ionic structures or canonical forms and any degree of polarity can be represented by suitably adjusting the relative contri- butions of these canonical forms to the resonance hybrid. Thus the polar 4 Dewar “ Electronic Theory of Organic Chemistcry ” Oxford Univ.Press 1949 p. 37. 5 (a) Mecke 2. physikal. Ghern. 1950 196 A 56. ( b ) Kellner Rep. Progr. Physics 1962 15 1. (b) Molleculur-orbital interpretation. ALLEN AND CALDIN THE ASSOCIATION OF CARBOXYLTC ACIDS 257 bond -0-H is represented as a resonance hybrid of the homopolar bond -OH and the ionic bond -0-H+ and the polar CO bond similarly as a hybrid of >C=O and >C-0. In the hydrogen bonds involved in the dimerisation of carboxylic acids the canonical forms to be taken into account include also a form involving a new covalent bond and a shift of charge from one oxygen atom to the other and in the simplest model the most important canonical forms are given by Coulson + - as including (I) -0-H o=c< (111) -0- H p o=c< Estimates of the relative weights or contributions of these structures to the resonance hybrid can be made from a knowledge of the distances and dipole moments involved ; they have been calculated in one case as 65 31 and 4 respectively.Thus the contribution from the structure (111) involving the new covalent bond is small and the first two forms which give rise to the normal 0-H dipole predominate. The relative contribu- tions of these two forms interpret the dipolar character of the bond in the language of resonance theory. An earlier view in terms of resonance theory omitted all reference to ionic canonical forms and consequently led to results a t variance with experi- ment. The hydrogen bond was attributed to resonance between structures R *C >C*E ROC >C*R ROC (11) -O-H+ 0=C< + /o--H--o ‘OR 80 H-0 /O-H 0 \O-H 0 \O H-0 \O __-_ H ____ O/ (IV) (V) (VI) (IV) and (V) .The resulting resonance hybrid would be represented by (VI) in which the hydrogen atoms are symmetrically placed between the oxygen atoms and all the C-0 bonds are alike. shows however that in dimers the hydrogen is not symmetrically placed the O-H bond being in fact only slightly extended and that the C-0 bond frequencies are not all equal ; electron-diffraction studies also show that the C-0 distances are unequal. So long as the ionic canonical forms are taken into account there is thus no sharp antithesis between the electrostatic and the resonance account of the hydrogen bond. . We may however distinguish between electrostatic and ‘‘ resonance ” contributions to the energy of the bond if by the latter we mean the contribution from the form (111) as distinct from (I) and (11) which involve no new covalent bond.It would be better to speak of electro- static and covalence contributions or to specify the canonical forms con- cerned since resonance [between (I) and (II)] is involved in interpreting even purely electrostatic descriptions of the bond. The covalence contri- bution as we have mentioned appears from valency theory to be small; and calculations of the energy of the bond in terms of the electrostatic Infra-red spectroscopy Davies and Sutherland J . Chem. Phys. 1938 6 766. 258 QUARTERLY REVIEWS energy due to the dipoles of OH and CO (and induced dipoles) give values of the right order. The electrostatic account thus appears p r i m facie to be an adequate approximation for discussions of the results obtained when the substituents solvent and other factors are vaned.11. Association in the Vapour Phase The association of a carboxylic acid under given conditions is most simply given quantitative expression in terms of the equilibrium constant. For the equilibrium between monomer (A,) and dimer (A,) which may be represented as A,JL2A1 the equilibrium constant K in the gas phase may be expressed in terms of the partial pressures of monomer ( p l ) and of dimer ( p z ) as K2= pI2/p,. This equilibrium constant measures the tendency of the dimer of a given acid to dissociate. Similar equations may be written for the higher polymers sometimes postulated. The use of partial pressures assumes that the vapour obeys the laws of a perfect gas mixture; for a non-ideal mixture pressures should be replaced by fugacities.It is there- fore desirable to use fairly low pressures so that simple gas and equilibrium laws can be applied. From the variation of the equilibrium constant with temperature the changes of free energy (G) heat content and entropy on dissociation can a t once be found. The standard free energy a t a given temperature for the dissociation of the dimer is given by AG,' = - RT lnK ; the change of heat content ( H ) may be found from the temperature-variation of K, since (3 In K2/3T)p = AH,'/RT2 ; and the standard entropy change is given by AX,' = (AH,' - AG,')/T. The standard state is that at which all the partial pressures have the value unity (commonly 1 atmosphere) Of these t'hermodynamic functions the change in heat content AH,' measures chiefly the energy of the bonds broken ; and the change in standard entropy A 8 2 will depend on the change in the number of molecules due to dissociation whether this is simply a doubling as in the gas phase or whether as in solutions it also depends on a partial " freezing " of solvent molecules round the new monomer molecule.Measurements of the equilibrium con- stant are thus the normal experimental attack on the problem. Experimental methods for investigation of the gaseous equilibrium. Investi- gation of the vapour-phase association of carboxylic acids dates back more than a hundred years to work published by Bineau in 1846.' Since then only three different techniques have been employed. Of these investigation of pressure-temperature-volume relations by the Boyle's law type of apparatus has been the most popular ; more recently vapour densities have been measured by means of the quartz microbalance.Infra-red spectro- scopy has provided the third method of investigation. Many of the early investigators measured the pressure-volume-tempera- ture relations for a given mass of vapour in a bulb attached to a manometer and so determined the apparent molecular weight of the carboxylic acid; but except in the work of Nernst and von Wartenberg the results were 7 Bineau Ann. Chim. Phys. 1846 18 228. 8 Nernst and von Wartenburg 2. Elektrochem. 1916 22 37. ALLEN AND CALDIN THE ASSOCIATION OF CARBOXYLIC ACIDS 259 rather discordant. Later work has been more precise. Corrections have been made for adsorption by parallel experiments in vessels of differing surface volume ratio,g 10 though high precision has also been claimed when this precaution has not been taken.ll Johnson and Nash l2 determiped the vapour densities of aliphatic acids by means of a quartz microbalance.This balance was essentially a buoy- ancy globe counterpoised with a perforated bulb of equal surface area; a small permanent electromagnet was sealed into the beam and alteration of the current in an external electromagnet restored the beam to its reference position. This method has important advantages over the earlier technique in that it eliminates adsorption effects and does not require an estimate of the amount of material in the vapour phase. The results do however depend on a calibration,. which requires a standard substance with a mole- cular weight in the higher part of the range t o be studied.It will be interesting to see whether different investigators will obtain more concordant results by this technique than by the classical method (cf. Table 2 p. 261). Carboxy- lic acids show the characteristic O-H vibration absorption at about 2.S ,u only at higher temperatures at which the acids are known to be monomeric. At lower temperatures where dimerisation occurs this band is replaced by one at about 3.2 p which is ascribed to the vibration of the hydroxyl group in the dimer. This method of measuring the relative concentrations has the defect that the band a t 3.2 p lies in the same region as that due to the C-H vibration so that it is impossible to measure the degree of association by means of the fundamental frequency of the hydroxyl group participating in hydrogen- bond formation.If deuterated carboxylic adids R*CO,D are used however the frequency of the O-D*.-O vibration is lower by a factor of approximately 42 and the absorption band appears at 4.35 p and so can be measured. The O-H..-O and O-D-..O bands show a depend- ence on temperature ; as the temperature is increased the fraction of the dimer increases and consequently the intensities of the O-H...D or O-D...O bands diminish and those corresponding to free O-H or 0-D increase. Badger and Bauer l3 made the first attempt a t quantitative measurements by investigating the temperature-dependence of the intensity of the second harmonic of the O-H band from a,cetic acid vapour. The results agreed roughly with MacDougall’s vapour-density work l4 but the accuracy was not very high because of the difficulty of making photographic determinations of the intensities.Herman and Hofstadter l5 used this method with considerable success in studying the dimerisation of deutero- acetic acid ; they observed the intensity of the O-D-..O band over a range of temperature. I n calculating the heat of association it was necessary to The use of infra-red spectroscopy presents different problems. Coolidge J. Amer. Chern. SOC. 1925 50 2166. lo Ritter and Simons ibid. 1945 67 757. l1 Taylor ibid. 1951 73 315. l2 Johnson and Nash ibid. 1950 72 547. l3 Badger and Bauer J . Chem. Phys. 1937 5 839. l4 MacDougall J. Amer. Chern. SOC. 1936 58 2585. l5 Herman and Hofstadter J. Chem. Phys. 1939 7 460. 260 QUARTERLY REVIEWS know the degree of association a t one temperature ; this was obtained from the appropriate vapour-density data.To prove that the method was satis- factory Herman l6 investigated the dimerisation of formic acid using the temperature-dependence of an association band at 7.35 p and obtained results which were consistent with those for the deuteroacid. Several assumptions are made in applying this technique to the association of car- boxylic acids,2 and a previous knowledge of the system is required before the equilibrium constant can be evaluated but on the whole the agreement with the two vapour-density methods is satisfactory if the assumptions are accepted. The ranges of temperature and pressure used by various investigators differ considerably as may be seen from Table 1. The pressure is however always low enough for dimerisation to be the main equilibrium concerned.At higher pressures several workers have found it necessary to postulate also either higher polymers or non-ideal behaviour to account fully for their results as indicated in Table 2. These higher polymers are discussed below (p. 265). TABLE 1. Experimental conditions Acid H*CO,H . . . 7 . . . . . . Me*EO,H . . , * - ,? * * 7 - * * . . . Et*C'b,H . . . . . CMij*CO,H . . PrW0,H. . . n-C,H,,.CO,H . CF,*CO,H. . . ~ Dimerimtion.- Investigator Coolidge Ramsperger and Porter Taylor and Bruton MacDougall Nernst and von Wartenburg Ritter and Simons Johnson and Nash Taylor MacDougall Taylor and Bruton Johnson and Nash Lundin Harris and Nash ' Pressure range Ref. 1 (mm.) 9 1 3-1127 19 3-1 3 21 j < 76 14 5-25 8 1 400-1500 10 45-813 12 1 160-1150 11 13-34 17 ' 3-20 21 < 76 12 98-780 18 1 250-720 Temp.range 10- 15 6" 25-84 50-150 25-40 40-200 50-2 10 80-200 50-150 50-65 50-150 115-200 14 1-200 189-227 81-131 Structure of the dimer. Before dealing with studies on the equilibrium it is useful to recall the results of electron-diffraction work on the dimeric forms of carboxylic acids. The earliest investigation was that of Pauling and Brockway in 1934 on formic acid ; 21a they concluded that the dimeric species had a ring structure in which the C-0 bonds were all of equal length as in (VI). confirmed the ring structure but concluded that the lengths of the two C-0 bonds in the Karle and Brockway 16 Herman J . Chem. Phys. 1940 8 252. l7 MacDougall J . Amer. Chew. Soc. 1941 63 3420. I8 Lundin Harris and Nash ibid.1952 74 743. l9 Ramsperger and Porter ibid. 1926 48 1267. 2O Lundjn Harris and Nash ibid. 1952 74 4654. 21 Taylor and Bruton ibid. p. 4151. 21a Pauling and Brockway Proc. Nut. Acad. Sci. 1934 20 336. ALLEN AND CALDIN THE ASSOCIATION OF CARBOXYLIC ACIDS 261 carboxyl group were different both in the monomer and in the dimer. Spectroscopic results in solution bear this out In acetic acid the lengths were found t o be the same as in formic. The 0-0 distance across the hydrogen bond was reported to be the same within a few hundredths of an Angstrom unit for formic acetic and trifluoroacetic acid and there were no appreciable differences in the diffraction photographs of acetic and deuteroacetic acid. Although the interatomic distances in the formic acid monomer have been revised,lb the general picture of a ring structure for the dimers does not appear to be in doubt.Electron-diffraction studies tell us nothing directly however about the positions of the hydrogen atoms. Heats and entropies of dissociation of dimeric carboxylic acid molecules. These quantities are of interest because as has been mentioned they reflect the energy of the bonds broken in dissociation and the changes in trans- lational rotational and vibrational energy associated with the formation of the two new molecules. In the more recent investigations the tempera- ture has usually been varied enough for evaluation of these factors. The temperature-dependence of the equilibrium constant K for dimerisation has commonly been expressed within experimental error by the equation In K = A - B/T where A and B are constants.The values of A and B give the heat of dissociation of the dimer AH," (= B / R ) and the standard entropy change on dissociation ASz' (= A / R ) . The results of various investigators are summarised in Table 2. TABLE 2. Standard heats and entropies of dissociation of dimers of aliphatic acids in the vapour phase From K in atm. ; in the standard state partial pressures are unity. I Acid 1 Mcthod * I I H.CO,H . . 1 V.D. 7 ) . . 1 V.D. Y ? . . . . H.&,D . . Me.CO,H . . Y 9 Y7 9 ) . . . . . . . . Me.cO,D . . Et*CO,H . . . . Et*d)O,D . . Pr*CO,H . . Pr.CO,D . . C,H,,-CO,H . CMe,*CO,H . CP,-CO,H . . V.D. Spect. V.D. V.D. V.D. V.D. Spect. V.D. V.D. Spect V.D. V.D. V.D. V.D. VID. Spect. ~ _ _ _ __- Ref. 9 19 21 16 16 8 14 10 12 11 15 17 21 15 18 15 18 12 20 Diiners AH=' (kcal.mole-') 14.1 14.1 14-1 f 0.2 12.4 f 1 12.8 & 1 15.0 16.4 f 0.8 14.5 0.4 13.8 $- 0.1 15.3 &- 0.1 15.9 4 1 18.5 f 2 15.2 0.2 14.1 f 0.5 13.9 f 0.2 13.8 f 1 13.4 f 0.7 14.0 + 0.2 14.0 f 0-2 AS,' (cal deg-l mole- ) 36 36 36* - - - 41 35 33 36* 46 36* 33 33 - - - 33 36 Higher polymers suggested f Tetramer ; AH,' 2 Trimer ; AH,O 22.7 - Trimer ; AH,O 24 - Trimer ; AH,O 23 Trimer ; AH,O 23 * V.D. = vapour density. AH0 in kcsl. mole-l. 262 QUARTERLY REVIEWS Before drawing conclusions from these figures we must consider the possible errors. Values of AH,' obtained by the density-balance method 1 2 have a precision it is claimed of r;t 0.1 kcal. mole-l and those from the most recent version of the " classical " pressure-volume-temperature method a precision within -j= 0-2 kcal.mole-l. Unfortunately the two methods do not agree within these limits; the respective values of AH,' for acetic acid are 15.27 & 0.1 (Johnson and Nash) and 13.8 & 0.2 kcal. mole-l (Taylor). The discrepancy may be due to effects of adsorption which do not appear to have been taken into account in Taylor's experiments; or it may be due to the uncertainty in the calibration of the density balance. Errors in the entropy AX,' are in general difficult to estimate ; only for the most precise work such as that just mentioned are they are low as 0.2 cal. deg.-l mole-l and most of the values cited are probably uncertain to the extent of a t least 1 unit. The accuracy attained by any method depends on the temperature range covered and the high values obtained by Mac- Dougall for acet'ic l4 and propionic acid l7 may be due to the restricted range used by him (Table 1).The spectroscopic values of AH,' are in reasonable agreement with the values from vapour-density methods when it is remembered that the experimental error is of the order of 1 kcal. mole-1 for the former. Some discrepancy might be expected since the two methods will be differently affected by the presence of higher polymers and by devia- tions from the ideal gas laws. It is evident that if the factors affecting dimerisation in the gas phase are to be elucidated more extended series of acids should be investigated by precise methods under comparable conditions and the discrepancies between the two vapour-density methods should be resolved. When the uncertainties in the results are borne in mind the following comments on the values of AH,' may be made.The heats of dissociation of the carboxylic acid dimers are all in the region of 14 & 1 kcal. mole-l in the vapour phase. This is much less than would be expected for the breaking of a normal covalent bond but is of the order of magnitude to be expected for a hydrogen bond if this is due mainly to electrostatic forces (cf. below). Neither the length of the hydrocarbon chain nor chain- branching has any great effect on the heat of dissociation as was shown by Nash and his co-workers 12 l8 in the series acetic propionic heptanoic and trimethylacetic acid. Electronegative substituents do not greatly alter AH,O if trifluoroacetic and acetic acid are typical ; the range of substitut'ed acids so far studied is not very extensive.The standard entropy changes AX,' are remarkably constant a t about 35 units (apart from MacDougall's results) and there is again no marked dependence on the dimensions of the hydrocarbon chain. When a dimer dissociates into two monomeric molecules the increase in entropy will be due mainly to the increase in the number of free particles which gives rise to three extra translational degrees of freedom per molecule with a smaller contribution from the extra degrees of rotational freedom. These factors would according to ordinary statistical theory lead to an increase of entropy in the region of 35 units plus a contribution of a few unit's from ALLEN AND CALDIN THE ASSOCIATION OF CARBOXYLIC ACIDS 263 the freeing of rotation about the C-0 bond (which will be restricted in the dimer).The order of magnitude of AS,' and its approximate constancy can thus be understood in general terms. The detailed consideration of a particular acid is more complex because in the dimer molecule the monomer units will not be rigidly fixed and the bonds between them will allow several kinds of bending and stretching motions. Given really accurate values of the entropies we should be able to draw conclusions about the variation of the force constants of these bonds with the structure of the acid. Cal- culations on formic and acetic acid 2 suggest that values for the entropies are not in general reliable enough to permit this. However it is possible to deduce a rough value for the force constant for symmetrical stretching of the hydrogen bonds in formic acid namely 3 (& 1.3) x lo4 dynes per cm.and a value in this region is supported by some spectroscopic evidence.5b This would contribute about 2 units to the entropy. It is much smaller than the values for ordinary covalent X-H bonds which are in the region of 6 x lo5 dynes per cm. The equilibrium constant at a given temperature can be determined more accurately than either AH,' or A X O which depend on its variation with temperature. Thus a comparison of the values of K at one temperature is a more precise indica- tion of the effects of the group R on the dimerisation of R-CO,H though to interpret these effects we usually have to consider the heat and entropy of dimerisation which both affect K,. Direct comparison of experimental values of K a t one temperature is often not possible because of the variety of conditions used (Table 1).However for several acids values of K can be obtained a t 160° directly or by short extrapolations ; the agreement between the results of different workers is better than for AH,'. The results are collected in Table 3. This result is in fair agreement with the experimental value. Effect of substituents on equilibrium constant TABLE 3. Equilibrium constants K for dissociation of dimers aliphatic acids in the vapour phase at 160" Acid CF,*CO,H . . . . H*CO,H . . . . Me.CO,H . . . . Et*CO,H . . . n-C,H,,*CO,H . . . CMe,*CO,H. . . . Pr C O 2H . . . Ref. 20 9 20 21 11 10 20 21 20 20 20 liz (atrn ) at 160" in gas 7.4 5*5* 1-3 1.6 1-65 1.1 1.8.f 10511 a t 25" in HaO (mole 1 - I ) Strong 17.7 1-76 1.35 1-48 1.4 0.94 * Individual values are 5.5 (Coolidge 9 5.5 (Nash 2 0 ) and 5.6 (Taylor 21).t Individual values are 2.8 (Nash 2 0 ) 2.1 (Taylor 11) and 1-6 (Ritter and Simons 10). The value of K for the dissociation of the dimer (R-C02H) is seen to be little affected by the group R so long as this is a hydrocarbon group 22 Halford J. Chom. Physics 1946 14 395. 264 QUARTERLY REVIEWS but replacement by H or by CF increases K considerably. The figures for K are in the same order as the values of the acid dissociation constant i n water at 25" which for these acids reflect the inductive effect of the group R. (In solution the results are roughly parallel; see below.) The much larger dissociation of the formic and trifluoroacetic acid dimers might be due either to a lower heat of dissociation or to a larger entropy change on dissociation corresponding to a stiffer bond in the dimer.Hence we need to consider the values of AH,' and AX2*. Unfortunately the results of different workers are not in agreement. Nash and his co-workers report that trifluoroacetic formic acetic butyric and trimethylacetic acid all have AH,' = 14.0 (& 0.3) kcal. mole-l (Table 2) whereas a decrease of about 1-4 kcal. mole-1 is required to account for the considerably larger dissociation constants of the first two acids in terms of AH,'. They conclude that an explanation in terms of the heat of dissociation of the hydrogen bond is not admissible. However Taylor's results indicate just such a decrease for formic acid (14.1 kcal. moleF1 compared with 15.3 and 15-2 for acetic and propionic acid respectively).The question thus remains open from the experimental standpoint. It is indeed difficult to account for the greater dissociation of the dimers of the two strongest acids in terms of entropy changes; Lundin Harris and Nash 2o have been unable to do so from a consideration of the effects of the masses and dimensions of the substituents on the translational and rotational contributions to the entropy. If it is true that the hydrogen bonds in the dimer are predominantly electrostatic i t should be possible to show that calculations of the heat of dimerisation AH,'based on electrostatic interaction agree with the experimental values. This is cer- tainly verified as regards order of magnitude as is shown by a variety of calculations on the formic acid dimer. (Strictly such calculations yield values of the contribution of molecular interactions to the standard free energy change AG,' without regard to thermal motion ie.at absolute zero where AG,' can be identified with the corresponding internal energy of dimerisation A U,'. At finite temperatures the identification is an approxi- mation AU,' is related to AH,' by the equation AH,' = AU,' + PAV ( P = pressure ; V = volume) ; since each mole of dimer gives rise to two moles of monomer PAV = RT and so with T in the region of 400" K about 800 cals. per mole must be added to AV,' in order to obtain AH,')). A simple calculation taking account only of the interaction of the per- manent dipoles of the hydroxyl and the carboxyl group (assumed to be the same as in other compounds) gives a value of AH,' in the region of 12 kcal.m0le-1.~3 I n a later treatment 24 the induced dipoles were also taken into consideration and the permanent dipoles were treated by resolution into point charges at appropriate distances (taken from Pauling and Brock- way's results) so that allowance could be made for all the interaction^.^^ This gave AH,' = 14.2 kcal. mole-l made up of 11.2 kcal from simple dipole intereaction 2.2 kcal. from induced dipole interaction and 0.8 kcal. Theoretical calculations of the heat of dirnerisation. 23 Moelwyn-Hughes J . 1938 1243. 24 M. M. Davies Trans. Faraday SOC. 1940 36 341. ALLEN AND CALDIN THE ASSOCIATION OF CARBOXYLIC ACIDS 265 from PAV. The agreement with the experimental value of 14.1 & 0.1 kcal. mole-1 might appear very satisfactory ; but dispersion forces and repulsion forces have been neglected and contributions from these though probably minor and of opposite sign cannot be reliably estimated.A calculation taking into account the attractive forces of dispersion gives AH,' = 16.7 kcal. mole-1 while a calculation in which repulsive forces also are con- sidered 25 gives AH,' = 11.0 kcal. mole-l (on Karle and Brockway's model l). The two values lie on either side of that calculated from electrostatic forces alone and differ appreciably from it. The difficulties associated with intermolecular forces at short distances thus prevent a precise confirmation of the view that the bonds are mainly electrostatic though there is no doubt that this view gives a good approximation to the true value of AH,'. Such calculations start from the empirical values of the distances and dipole moments concerned.Quantum-mechanical calculations which assume only the ring structure of the dimer and lead to estimates of both energies and bond length have been attempted by Gillette and Sherman,26 using the resonance method. They obtained a value for AH,' on the assumption that the structures (I) and (11) participate and another on the assumption that twelve other possible non-ionic structures are also involved. Neither value was in reasonable agreement with experiment. Gillette and Sherman concluded that ionic canonical forces must be included in the calculation and as we have noted these appear from later calculations to contribute most of the energy ; but explicit calculations o f AH,' based on them seem not to have been reported.Higher Polymers.-Several workers have found that dimerisation alone does not account for their observations and the deviations have usually been ascribed either to the formation of higher polymers of the acids or to non-ideal behaviour of the mixture of gases. It is ho-wever very difficult to distinguish between deviations of a vapour from ideality and the associa- tion of its molecules since both depend on intermolecular forces. Johnson and Nash l2 used a quantitative method of differentiating between them. The experimental results were subjected to two forms of analysis. In the first the gaseous imperfections were estimated on the assumption that no higher polymers were present. In the second the possibility of polymers higher than the dimer was considered on the assumption that the gaseous mixture was ideal.Only the second analysis gave satisfactory results ; the first gave highly improbable values for the second virial coefficients of the monomer and dimer. Most workers have considered only one or other of the two possibilities. assumed that the deviations shown by formic acid were due to non-ideal behaviour of the monomeric and the dimeric species and obtained a satisfactory correlation of his experimental data without considering higher polymers. MacDougall l7 found that even a t low pressures the equilibrium constants calculated for the dimerisation of propionic acid increased with pressure. This was attributed to the existence 25 Jones Gilkerson and Gallup J. Chem. Phys. 1952 20 1048 ; MaladiBre Compt. rend. 1948 226 1600. Gillette and Sherman J. Amer.Chem. SOC. 1936 58 1135. Thus Coolidge 266 QUARTERLY REVIEWS of a trimer and the heat of trimerisation was calculated and found to be about 1.5 times the heat of dimerisation which suggested that the hydrogen bonds in the trimer were comparable in strength with those in the dimer. Ritter and Simons lo found that non-ideal behaviour could not account for their observations on the association of acetic acid vapour. They con- sidered a monomer-trimer equilibrium but rejected it in favour of an equilibrium between monomeric dimeric and tetrameric species which fitted their experimental data better. They proposed a layer structure (VII) for the tetramer in preference to a ring structure (VIII) on the ground that the ring would be strained and could only be formed by breaking exothermic bonds; the same arguments suggested that the trimer (IX) should be less stable than the tetramer.The heat of dissociation calculated for the tetramer was about twice that of the dimer suggesting strong hydro- gen bonds between the layers. Ritter and Simons assumed that in the hydrogen bonds the hydrogen atoms were symnietrically placed between the oxygen atoms ; the layer structure would be less regular when rewritten to conform with later information on the 0-H distances.l 10 H 0 CH3.C-0 H-O-CCH3 II 0 H I I ,O H 0 H 0 I I1 I (VIII) 0 0 CH3.C-0-H O-C.CH3 The most thorough investigation of higher polymers is due to Johnson and Nash.12 The equilibrium constants for the dimerisation of acetic trimethylacetic and butyric acid when calculated from the vapour density on the assumption that no higher polymers were present showed a pro- nounced variation with density.The experimental results when analysed as mentioned above indicated the presence of higher polymers notably a trimer ; the presence of a tetramer was considered to be doubtful. The strengths of the hydrogen bonds in the trimer (calculated from its heat of dissociation) were again of the same order as the bonds in the dimer. There is some uncertainty in the evaluation of the standard entropy change for ALLEN AND CALDIN THE ASSOCIATION OF CARBOXYLIC ACIDS 267 trimerisation which depends on the choice of the equation of state of the standard substance (carbon tetrachloride). Johnson and Nash have also subjected the results obtained by Coolidge for formic acid to the same analysis and found rather tenuous evidence for the presence of a trimer with a heat of dissociation AH,' = 16 kcal.mole-l. The fact that AH,' was here little more than the heat of dissociation of the dimer (14 kcal. molep1) was interpreted to mean that the aggregate was not like the dimer a stable species and might only be a transient collision complex. Johnson and Nash also discussed the structure of the trimer ; and concluded that the most likely was a non-planar 12-membered ring. It should be noted that in three of these investigations the experiments were conducted a t relatively high pressures and that in two investigations of acetic acid at lower pressures no higher polymerisation was observed. The higher polymers would be formed in greater proportion as the pressure increased ; but the deviations from ideality must also increase and this makes the evidence for the presence of higher polymers less convincing.There appears to be a need for an investigation extending over a range of pressures from say 2 to 400 mm. to elucidate the conditions under which higher polymers are formed. III. Hydrogen Bonding in Solid Carboxylic Acids Evidence on hydrogen bonding in the solid state is much less extensive than for the vapour state. Early X-ray investigations 2 7 on solid fatty acids indicated that the hydrocarbon chains lie parallel to one another and are associated in collinear pairs presumably with the carboxyl groups together. Later work has been mainly concerned with dicarboxylic acids. Oxalic acid crystallises in two forms 01 and p in both of which certain of the 0-H-0 distances are so short tha.t they indicate hydrogen bonds.2s p-Oxalic acid exhibits the cyclic structure found in formic acid vapour ; the molecules are linked end to end in infinite chains represented by (X).(The forces between the chains are comparatively weak as indicated by the 0-0 distances,) This arrangement persists in the melt. A similar structure is found in the 8-form of succinic acid and in the higher members of the dicarboxylic acid series. In cc-oxalic acid the hydro- gen bonds link the molecules into puckered sheets (XI) ; the 0-0 distance in the grouping C=O-*.HO-C is 2.71 A short enough to indicate clearly a hydrogen bond. 27 Muller J. 1923 2043 3166 ; Morrow Phys. Review 1928 31 10. z* ( a ) Hendricks 2. Kryst. 1935 91 48 ; ( b ) Cox Dougill and Jeffrey J . 1952 4854 ; (c) Ahmed and Cruickshank Acta Cryst.1953 6 385. 268 QVARTERLY REVIEWS It would be expected that hydrogen bonding would be responsible for most of the lattice energy of each of these structures. There are two hydro- gen bonds per molecule of oxalic acid and if the heat of dissociation of each bond has about the value found for monocarboxylic acids in the gas phase the lattice energy will be in the region of 14 kcal. mole-l. The (XI) II values estimated by Bradley and Cotson 29 from vapour-pressure measure- ments are 22-3 kcal. for the 16- and 23.4 kcal. for the ct-form. These are of the right order of magnitude for hydrogen bonds and it must be remem- bered that the uncertainty in estimating the dispersion and the repulsion energy as well as the electrostatic energy is even greater in the solid than in the gaseous state because of the effects of the non-nearest neighbours.Solid formic acid according to recent preliminary results has a crystal structure that precludes the possibility of dimers in the solid and suggests that the molecules are arranged in infinite chains in which each molecule is linked to two neighbours by hydrogen bonds.3* IV. Association in Solution Structure of the Dimer.-Results obtained by a variety of methods dis- cussed below show that carboxylic acids are associated in hydrocarbons and similar solvents and that in dilute solutions the associated molecule is dimeric. There is no such direct evidence that the dimer is cyclic in solu- tion as there is for the gas phase but from the parallelism of the results there is little reason to doubt it.There is direct evidence from Davies and Sutherland’s work on the infra-red spectra of formic and acetic acids in carbon tetrachloride,6 that the O-H bond of the carhoxyl group is slightly longer in the dimer than in the monomer that the hydrogen is not sym- metrically placed and that the C-0 bonds are not all equal in length. The bond lengths (8) found for the formic acid monomer and dimer are O-H C-0 C-0 Monomer 0.9 1.1 9 1.3 Dimer 1.0 1.2 1.2 The fact that the O-H distance is only slightly greater in the dimer accords with an interpretation of the association in terms of a looser inter- action than covalency. The bearing of the results on the question of the canonical forms involved in resonance has been mentioned on p. 256. Experimental Work on the Equilibrium in Solution.-The tendency to associate is measured as in the gas phase by the equilibrium constant.29 Bradley and Cleasby J. 1953 1681 ; cf. Bradley and Cotson J. 1953 1684. 30 Holtzberg Post and Fankuchen J . Chem. Phys. 1952 20 198. ALLEN AND CBLDIN THE ASSOCIATION OF CARBOXYLIC ACIDS 269 Most workers have used dilute solutions so that thermodynamic activities may be replaced by concentrations which are conveniently expressed as mole fractions ; thus if the mole fraction of monomer is x1 and of dimer x, K = x12/x2. Values of AG2' AH,' and AX,' can then be derived from K as above. The value of K in solution is sometimes expressed in moles per 1. Concentrations expressed in these units show a temperature-variation due t,o thermal expansion of the solution ; thus K will vary with the density of the solution as well as with the displacement of equilibrium; and the temperature-dependence of K will lead to values of AH,' slightly different from those derived from K in terms of mole fractions.In practice the error is only 2-3y0 for benzene solutions for example if the temperature range is about 30". A difficulty arises about units when comparisons have to be made with the gas phase Only when the values of K in the gas phase and in solution are expressed in the same units will the standard entropies derived there- from relate to the same standard state. For the gas phase we have expressed K in atm. ; since one 1. of gas a t 26" contains about 0.041 mole the unit of concentration at 25" is 0.041 mole per 1. For solutions we must adopt the same concentration scale whenever we make comparisons with the gas phase.For benzene 1 1. of liquid contains about 11-3 moles so a solution of mole fraction x contains 1 1 . 3 ~ moles and its concentration ex- pressed in the units used for the gas phase is (11*3/0*041)x. Since we cal- culate AX,' from the equation AS,' = AH,'/17 + R In K the adoption of this unit requires us to add R In (11-3/0-041) or about 11 cal. deg.-l molep1 to the value of AS,' calculated from K in mole fraction units. This is important (see p. 272). The range of acids studied in solution is greater than that in the gas phase-aromatic and unsaturated aliphatic acids have been studied as well as fatty acids-but the precision is somewhat lower. Early investi- gations produced good evidence that carboxylic acids are partly or com- pletely dimerised in solvents such as benzene or carbon tetrachloride.The existence of double molecules was inferred by Beckmann 31 in 1890 from his determinations of the molecular weights of carboxylic acids in organic solvents by the depression of freezing-point. Nernst 32 in 1891 explained the distribution of benzoic acid between water and an inorganic solvent in terms of dimerisation in the organic solvent. By studying the distribution of acid between water and benzene or chloroform Hendrixson 33 in 1897 determined the equilibrium constant for dimerisation a t two tem- peratures and thence calculated the heat of dissociation of the dimer. Some cryoscopic work by Trautz and Moschel 34 in 1926 suggested that polymers higher than the dimer were present in solutions of acetic acid in Earlier work indicating dirnerisation in aprotic solvents.31 Beckmann 2. physilcal. Chern. 1890 6 444. 32 Nernst ibid. 1891 8 110. 33 Hendrixson 2. anorg. Chem. 1897 13 73. 34 Trautz and Moschel ibid. 1926 155 13. T 270 QUARTERLY REVIEWS benzene or nitrobenzene; this has not been observed in later work with dilute solutions. In the same year Brown and Bury 35 reported cryoscopic measurements by the Beckmann method for several carboxylic acids in nitrobenzene ; their results were complicated by deviations from ideal behaviour but when later reinterpreted by Wynne- Jones and Rushbrooke 36 gave evidence for a dimerisation equilibrium ; it was found that addition of water decreased the apparent degree of association. Cryoscopic measure- ments were also carried out by Peterson and Rodeb~sh,~' who used the much superior method due originally to Adamsj3* in which equilibrium is assured by using an intimate mixture of solution and crushed solid solvent the concentration being estimated after withdrawal of a sample of the equili- briated solution ; further the freezing-point depression is directly and accurately measured by means of a multi-junction thermocouple with one set of junctions in the mixture of solution and solid solvent and the other in a mixture of solid and liquid solvent.With benzene as solvent it was possible to cover a range of concentrations in which the apparent molecular weight of acetic acid varied from nearly the single value to nearly the double value ; that of benzoic acid varied much less over the same range. Evidence that a t higher temperatures dimerisation is incomplete in benzene for a variety of acids and depends upon structure appears to have been first provided by Brocklesby who determined the boiling-point elevations of solutions in benzene and other solvents by a method accurate to about 0*01" ; he found that at about 80" the degree of association in benzene varies with concentration and that in a series of fatty acids at a fixed concentration it decreases up to dodecanoic acid after which it remains almost constant.Trichloroacetic acid and its hydrate were studied cryoscopically in benzene by Bell and who found that the anhydrous acid formed double molecules while the hydrate was present as single molecules in dilute solution. It was thus established that dimerisation equilibria were set up that the equilibrium could be measured a t reasonable temperatures and that it varied from acid to acid.It was also shown that the equilibrium was affected by the presence of water. Little was known however about the values of the equilibrium constants or their variation with temperature. Since about 1945 these have been investigated increasingly by a variety of methods. Aprotic solvents have generally been used ; work in other solvents is briefly noticed at the end of the section. The methods that have been used include use of cryoscopy vapour-pressure measurements ebullioscopy dielectric constants absorption spectroscopy and distribution measurements. Of these freezing-point measurements offer high precision and have been used by Barton and K r a ~ s ~ l but are applicable Determination of equilibrium constants for dimerisation in solution.36 Brown and Bury J. Phys. Chem. 1926 30 694. 36 Wynne-Jones and Rushbrooke Trans. Faraday SOC. 1944 40 345. 37 Peterson and Rodebush J. Phys. Chem. 1928 32 709. 3* Adams J . Amer. Chem. Xoc. 1915 37 481. 39 Brocklesby Canad. J. Res. 1936 14 By 222. 40 Bell and Arnold J. 1935 1432. 41 Barton and Kraus J. Amer. Chem. SOC. 1951 73 4561. ALLEN AND CALDTN THE ASSOCIATION OF CARBOXYLIC A4CIDS 271 only at one temperature The ebullioscopic method give results of lower precision and cannot be applied if the acid is too volatile ; but as the boiling temperature can be varied by altering the pressure the equilibrium constant can be measured over a range of temperature and the heat and entropy of dimerisation determined,42 43 though less accurately than for the best gas- phase determinations.This holds also for the vapour-pressure method which has been applied by Wall and Banes ; 44 they used an isopiestic method and obtained an accuracy of the same order as that of ebullioscopic measure- ments in determining heats of dimerisation. Dielectric constants of solu- tions of many carboxylic acids in benzene have been measured a t one temperature by Hobbs and who fitted the results in dilute solution to an equation for a dimerisation equilibrium and evaluated the equilibrium constant ; the precision of this method is difficult to assess but is certainly not high. A spectroscopic investigation of solutions of acetic acid in carbon tetrachloride was carried out by Davies and Sutherland,6 who measured the intensity of the absorption band of the hydroxyl group over a range of concentrations a t three temperatures and obtained a reasonable value for AH," ; the results for benzoic acid were however rather erratic.The distribution of acetic and propionic acid between water and various aprotic L I I I I I 3 0 32 3 4 l o p FIG. 1 log, K against 1/11. Dissociation of bentoic acid dimers in benzene as a function of temperature plot of A Freezing-point depression method.41 0 Vapour-pressure depression method.44 0 Boiling-point elevation method.43 + Dielectric-constant method.45 42 Wolf and Metzger Annalen 1949 563 157. 43 Allen and Caldin Trans. Paraday Xoc. in the press. 4 4 Wall and Banes J . Amer. Chem. SOC. 1945 61 890. 45 Hobbs and Gross J. Chem. Phys. 1941 9 408 415 ; J. Amer. Chem. Xoc. 1949 71 1671.272 QUARTERLY REVIEWS solvents has been studied a t a series of temperatures and values of the heats and entropies of dimerisation deduced; 46 this method requires a com- paratively simple technique the temperature can be varied and solutions in various solvents can be compared a t a fixed temperature ; but there is some uncertainty about the effects of the water dissolved in the organic solvent. The mutual agreement of results obtained by the first three methods in their different temperature-ranges can be seen from Fig. 1 in which values of log, K are plotted against 1/T. The agreement is evidently satisfactory. The mean value of AHz' the heat of dissociation of the dimer over the range 5"-8O0c can be determined as 8.4 & 0.5 kcal. mole-,. Comparisons between these three results and the others mentioned may be made with the aid of Table 6.The dielectric-constant method gives results that agree fairly well with those from vapour-pressure measurements for m- and o-toluic acid but not for benzoic acid. For acetic and propionic acid the dielectric- constant results do not agree with those from distribution experiments and here it is more difficult to decide which are the more reliable ; for the series of substituted fatty acids the dielectric-constant results will be used as giving a series of comparative values. Heats and entropies of dissociation of dimeric carboxylic acid molecules in benzene solution. The heats and standard entropies of dissociation of car- boxylic acid dirners in benzene (the solvent most commonly used) have of dimers of carboxylic acids in benzene at 25".Calc. from K in mole fraction units. TABLE 4. Standard free energies heats and entropies of dissociation Acid B z O H . . . . . . . . . . . . . . . . o-C,H,Me*CO,H . m-C,H,Me*CO,H . p-C,H,Me*CO,H . p-MeO*C,H,*CO,H m-C,H,I*CO,H . o-C,H,Cl*CO,H . Ph*CiC*CO,H . . Me*CO,H . . . Et-C0,H . . . Kef. __ 33 43 44 42 44 44 43 43 43 43 43 46 46 Method D i s h B.p. B.p. 8r v.p. f.p. V.p. v.p. B.p. B.p. B.p. B.p. Distn. Distn. B-P. AaZo at 25' (kcal mole- I ) - 5-18 f 0.06 5.26 f 0.04 - 5.03 f 0.04 5.44 f 0.05 6.05 f 0.06 5.17 f 0.06 4.99 f 0.08 4.74 f 0.05 4-10 f 0.05 4.3 f 0.1* 4.3 f 0*1* * Recalc. after conversion of K AH,' (kcal. mole-') 8-7 8.0 f 0.4 8.4 f 0.2 10.5 8-4 f 0.2 8.7 f 0.4 8.7 f 0.4 7.8 f 0.5 8.6 f 0.3 9.1 f 0.4 8.2 f 0.7 7.8 f 0.6 9.4 f 0.2 AS,' cal.deg.-' mole- I ) - 9.6 f 1.5 10-6 f 1.5 - 11.3 f 1.5 13.4 f 1.5 10.1 1.6 12.0 f 1.5 9.7 f 1-5 13.0 f 1.4 16.8 f 1.4 13.0 f 2* 11.5 f 2* into mole fraction units. TAS,' at 25' (kcal. mole-') - 2.8 & 0.4 3.1 f 0.2 3.4 f 0.2 4.0 f 0.2 3.0 f 0.4 3.6 f 0.4 2.9 f 0.6 3.8 f 0.4 5.0 f 0.4 3.9 -f 0.7 3.4 f 0.6 - been collected into Table 4 from all the investigations of equilibrium con- stants in which the temperature has been varied. The heats of dissociation ( b ) The values for AS,O in carbon disulphide and in nitrobenzene given in this paper have been replaced by values corrected in acoordance with a personal communication from Dr. E. A. Moelwyn-Hughes. 46 ( a ) M. M. Davies Jones Patnaik and Moelwyn-Hughes J . 1951 1249. ALLEN AND CALDIN THE ASSOCIATION OF CARBOXYLLC ACIDS 273 AH,* are mostly accurate to about 5 0.4 kcal.mole-1 and the standard entropies AX,' to about 1.5 cal. deg.-l mole-l so that TAX,' is known to about & 0.4 kcal. molep1. Values of AG,'(= - RT In K,) are given for reference ; they are known much more accurately to about & 0.06 kcal. mole- l. From the Table it appears that the heat-content change does not vary greatly from acid to acid or between aliphatic and aromatic acids ; it is in the region of 9 kcal. mole-l compared with about 14 kcal. molep1 for aliphatic acids in the gas phase. The standard entropy change AXzo is in the region of 11 units for nearly all the acids. To obtain AX,' for the standard state corresponding to that adopted for the gas phase we must add about 11 units (p. 269). The standard entropy change is then about 22 units compared with about 35 units for aliphatic acids in the gas phase.Thus the hydrogen bonds require less energy for dissociation in benzene solution than in the vapour but the effect of this in favouring monomer at the expense of dimer is offset by the smaller entropy change on dissociation the changes in AH,' and in TAX,' being of comparable magnitude and oppo- site effect on K [a(AG,") = - a(RT In K,) = a(AH,O) - a(TAX,")]. The smaller entropy and heat-content changes can be attributed to solvation and confirm the view for which there is other evidence that benzene is not an " inert " but a donor solvent. It is very probable that the mono- mers will be more solvated than the diniers on account of the local dipoles of the hydroxyl and carboxyl groups which are masked in the dimers ; even the resultant dipole moment of the molecule as a whole is known to be higher for the monomers (m.1.8 D compared with ca. 1-0 D for the dimer molecules 45). The monomer thus has a greater powcr of polarising tho solvent molecules and orientating t'hem by dipole interaction. There is therefore a partial " freezing " of molecules round the monomer which implies a gain of molecular order and a consequent decrease of entropy which partly compensates for the increase due to fhc greater number of molecules Solvation will also lower thc encrgy of the monomer more than that of the dimer and so make AH," smaller than that in the gas phase. We might thus expect a correlation between AH," and AS,' and as we shall see some correlation is found on varying the solventl whose polarity affects both quantities.There are also some signs of a correlation between AH,' and A8,O for the aromatic acids in Table 4 but this may be illusory since the changes are little greater than the experimental uncertainties. Solvent effects in aprotic solvents. For a series of solvents without specific hydrogen-bonding properties this explanation in terms of solvation would lead us to expect that the effects on AH,' and AS,' (compared with the vapour) would be greater the more polar m d polarisable the molecules of the solvent. The only systematic results on a series of aprotic solvents a t several temperatures are those obtained for acetic and propionic acid by Davies Moelwyn-Hughes et al. 46 using the distribution method ( i . e . sol- vents saturated with water).Some of their results for acetic acid are given in Table 5 . It will be seen that as the dielectric constant of the solvent (representing its polarity) increases AH,O and AX,' decrease with a roughly 274 QUARTERLY REVIEWS linear relation between them. the above interpretation in terms of solvation of dipolar molecules. Dissociation of dimeric acetic acid in various solvents These results are evidently in accord with TABLE 5. K in mole 1.-l units Solvent Vapour . C,H . . C,Hu - * CCl . . cs . . C,H5C1 . C,H,*NO, Dielectric constant at 25' 1.00 1.87 2.25 2.22 2.60 6.37 33.1 14.5 9.0 8.2 7.6 7.05 5.0 5-25 AS2* (cal deg - l mole-') 28 17 18 13 10 6 7 13.6 lWK at 35" 0.07 1.9 12-4 3.1 2 8 9.4 180 A different but ultimately equivalent way of approaching the matter is the following.The energy when two dipoles approach each other can be calculated on certain assumptions and the effect of variation of the dielectric constant of the solvent on this free energy (and so on K,) can be derived. Now the dielectric constant depends on the ability of the solvent molecules to orient themselves in an electric field. But this property also determines their solvating properties in these solutions. Thus an explana- tion in terms of dielectric constant ( e ) is equivalent to an explanation in terms of solvation. On the simplest assumption^,^^^ 46 one finds that log, K should vary linearly with l / e and that AH,' should be given by AH,' = [(l - LT)/e]AH2G where AHZG is the value in the gas phase and L = (de/dT). The first of these relations is roughly obeyed ; K increases with increase of dielectric constant (the effect on AG,' of the decrease in AH,O being greater than that of the decrease in AS,") though the plot of log, K against l / e is somewhat scattered.The second relation is obeyed more a ~ c u r a t e l y . ~ ~ Although it is not a t all clear that the assumptions involved are valid-particularly the assumption that the bulk dielectric constant is relevant to conditions near a solute molecule-it seems that the results so far available on variation of the solvent though in need of support are at least consistent with an electrostatic interpretation of the hydrogen bond. Structure of Acids and Dimerisation in Benzene.-The view that hydrogen bonding in carboxylic acid dimers is largely due to dipole interaction would lead us to expect that it should depend on the inductive and mesoineric effects of substituents.These effects are also important in determining the strengths of the acids in water so that a t a fixed temperature we might expect some relation between the dissociation constants of dimeric acids in some aprotic solvent and the acid dissociation constant in water. Such relations have been rcported for various limited series of acids all in benzene solution by several workers. Hobbs and from their nieasureinentq of dielectric constants conclude that there is a rough para811elism between the dissociation equilibrium constant K, for the dimers in benzene ALLEN AND CALDIN THE ASSOCIATION OF CARBOXYLIC ACIDS 275 (measured at 30") and the acid dissociation Ka in water (at 25") ; in the lower fatty acids an electronegative substituent increases the degree of dis- sociation of the dimer while an electropositive group decreases it.Barton and Kraus 41 from their more precise freezing-point work found that four acids-benzoic o-bromobenzoic phenylpropiolic and cinnamic-gave a roughly linear relation between log, K (at 5.5") and log, K (at 25") from which however #?-phenylpropionic acid deviated considerably. Wall and Banes's vapour-pressure data 44 show a nearly linear logarithmic relation a t 25" for benzoic m-toluic and o-toluic acid and this has been found to cover also o-chlorobenzoic m-iodobenzoic and phenylpropiolic acid the points for p-toluic and p-anisic acid deviating somewhat from the line.43 Table 6 shows the data on twenty-one acids for which the constant K is known a t 30° either directly or by extrapolation from measurements over a range of temperatures.Not all of the acid dissociation constants Ka are known a t 30" so we have used the values a t 25" (where the values of log, Ka at 30" are known they do not differ by more than 0.05 from those at 25"). TABLE 6. Dissociation of dimers in benzene at 30" and acid dissociation in water at 25" K in mole fraction units; K in mole 1 . - l units. No. in Fig. 2 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Acid ~T-C,H,MWCO,H. . . . . m-C,H,Me-CO,H . . . . BzOH . . . . . . 5 9 I . . . 9 9 . . . . . . . . . . . . f . p-MeO.C,H,X!O,H . . . o-C,H,Me.CO,H. . . . . 9 9 . . . . . m-C6H,I*C0,H . . . . . o-C6H,C1.C02H . . . . . o-C,H,F*CO,H . . . . . WJ-C,H,F.CO,H . . . . . P-C,H~F*CO,H . . . . . CMe,*CO,H .. . . . . Prn.CO,H . . . . . . E t *CO ,H . . . . . . . . . . . . . . M~.~?o,H . . . . . . CH,&h-CO,H . . . . . . . . . . CH,Ph*CH,*CO,H . . . . H*CO,H . . . . . . . CH2C1*C02H. . . . . . CH,CH.CH*CO,H . . . . PhCHCH*CO,H . . . . Ph*CiC*CO,H . . . . . Ref. 43 44 45 43 44 45 43 44 45 43 43 45 45 45 45 45 46 45 46 45 45 45 45 45 45 45 43 0.68 f 0.04 1-25 1-29 f 0-04 1.27 f 0.04 1-68 1.31 0.04 1.43 & 0-04 1.34 1.43 f 0.05 1.63 f 0.04 1.67 0.70 0-85 1.20 f 0.04 1.12 1-32 1.92 1.36 1-89 1.38 1.42 1-43 1-85 1.94 1.11 1.18 2-11 -J= 0.04 6 + lOY*Ll K a a t 85' 1.637 1.726 1.820 1.529 2-086 2.150 3.079 2.73 2.13 1.86 0.973 1.176 1-130 1.246 1.69 1.34 2.248 3-179 1.36 1.580 3-77 276 QUARTERLY REVIEWS Fig. 2 shows some approximate correlations between the dimerisation and acid-dissociation equilibria (i) The points for seven of the ten sub- stituted benzoic acids are grouped fairly closely about the upper straight line that for p-toluic acid is the only accurately known point to deviate by more than 0.2 in log, K,.The greater the dissociation as an acid in water the greater the dissociation of the dimer in benzene. (ii) The points repre- senting crotonic cinnamic and phenylpropiolic acid which form the third 5+ toy/* K2 FIG. 2 Comparison of equilibrzurn constants for dimerisation ( K2) and mid dissociation (K,) of carboxylic acids. log, ,Ka at 25" in water (in mole l.-l) ; log, K a t 30" in benzene (in mole-fraction Substituted benzoic acids. Other acids in which conjugation is possible crotonic cinnamic and phenyl- Acids without conjugation between carboxyl group and re& of molecule.UnltS). propiolic acid. 0 group in Table 6 lie near the same straight line. Thus there appears to be a general relation covering acids in which the carboxyl group is con- jugated with the rest of the molecule. (iii) The points for the second group in Table 6-those in which there is no conjugation between the carboxyl group and the rest of the molecule including fatty acids and phenylacetic and p-phenylpropionic acid-appear to be grouped about the lower line in Pig. 2. (P-Phenylpropionic acid is thus no longer anomalous on this classification.) For a given acid strength Ka this line corresponds to less dimerisation K being larger by a factor of about 2. The data used for all ALLEN AND CALDIN THE ASSOCIATION OF CARBOXYLIC ACIDS 277 these acids were obtained from dielectric-constant measurements only.however and it would be desirable to have confirmatory evidence and values of AH,O and At?,' from other methods. There are not enough data on acid dissociation constants in benzene to test whether they would give more exact relations between K and K,. (iv) The slope of each line is about 2 ; that is Ka is proportional to K2, or the effects of substituents on the free energy of acid dissociation in water are about twice as great as those on the free energy of dissociation of the dimer in benzene. Generalising the conclusions it appears that to some extent at least dimerisation and acid strength depend on the same structural factors. The effects of variation in structure are smallcr for dimerisation than for acid dissociation and acids appear to be distinguished according to whether or not mesomeric effects can influence the carboxyl group.Attempts to interpret these regularities further lead however to difficul- ties. Unfortunately the experimental results on AH,' and AX,' are not accurate enough compared with the rather small variations in K (which is much more accurately known) to allow us to decide how far each of these two quantities contributes to the variations in AG,' and K ; the variations of AH,' and !PAS2' relative to the values for say benzoic acid are comparable with the experimental uncertainties as may be seen from Table 4. The free energy of acid dissociation will depend in a series of acids R-CO,H on the energy of dissociation of the acid molecule to give a hydrogen atom (R*CO,H -+ RCO -+ H) on the electron-affinity of the remaining radical (RCO + c -+ RCO,-) and on the energy and entropy of solvation of the anion R.CO,-.It woulci be difficult to predict a priori how these quantities are related to AH; and ALY,'. (For that matter we do not know enough about them to predict values of Ka.) If solvation changos could be ignored we should conclude a posteriori from the experimentally observed relations that the energies of dissociation of dirner to monomer and of monomer to R*CO,- + H+ were linearly related and that for a given change of substituent the changes in the latter were about double those in the former. However these simple conclusioiis are not admissible because solvation plays an important part in the dimerisation equilibrium as we have seen and changes in As,' cannot be neglected ; while in determining acid strengths the entropy change is dominant in the lower fatty acids and important also in the substituted benzoic acids as may be seen from Table 7.47 More accurate results on a wider range of acids in solution and extension to aromatic acids of measurements in the vapour phase must be awaited.The difference in degree of dimerisation observed between the group of acids whose strengths are thought to be controlled by inductive effects only and t*hose for which rnesomeric effects also are important is another puzzle because both I and N effects should affect both the dis- sociation of hydrogen bonds (through the local dipole moments) and t'he dissociation of R*CO,H as an acid ; and there is no reason to expect a great difference in the solvation of the anions in which mcsomeric effects will 47 Everett and Wynne-Jones Trans.Paraday Soc. 1939 35 1380; R. P. Bell *'Acids and Bases " Methuen 1952 p. 57. 278 QUARTERLY REVIEWS not be prominent. If Confirmed the different relations for the two types of carboxylic acid might throw light on the several factors influencing the strengths of these acids. It might be thought that on the electrostatic view of hydrogen bonds there should be a correlation between the dimerisation constant K, and the dipole moment p of the monomer (which is known for many of the acids) since to a first approximation log K will vary with p2. This kind of explanation is only valid however when the entropy change in the reaction is constant ; 48 moreover one would need to know the local dipole moments of the hydroxyl and the carboxyl group not merely the resultant dipole moment of the whole molecule.No correlation is in fact found between the latter and the dimerisation constant. TABLE 7. For the first group of acids below the values given for AG,O AHa and AS,O are These quantities They are independent of concentration units since they Thermodynamic data for acid dissociutiun in water at 25". relative t o benzoic acid; for the second relative t o formic acid. are given in kcal. mole-l. refer t o the reaction R*CO,H + R'C0,- R*CO,- + R'*CO,H. Acid p-MeO*C,H,*CO,H . . . CHPh:CH*CO,H . . . m-C,H,Me*CO,H . . . BzOH . . . . . . o-C,H,Me*CO,H . . . o-C,H,Cl*CO,H . . . . Et*CO,H. . . . . . Pr*CO,H. . . . . . Me.CO,H . . . . . H*CO,H. . . . . . p-C,H,MeCO,H . . . m-C,H,I.CO,H .. . . CH,CI*CO,H . . . . 4.47 4.41 4.36 4.27 4.18 3.91 3.85 2.92 4-87 4-82 4-75 3.75 2.82 AGOO + 0.43 + 0.33 + 0-24 + 0.11 0 - 0.40 - 0.42 - 1.75 + 1.53 + 1.46 + 1.37 0 - 1.21 + 0.76 + 0.56 + 0.26 + 0.03 0 - 1.44 + 0.15 - 2.51 - 0.13 - 0.66 - 0.06 0 - 1.11 + 0.33 + 0.23 + 0.02 - 0.08 0 - 1.04 + 0.57 - 0.76 - 1.66 - 2.12 - 1.43 0 + 0.10 Association in. other solvents. In solvents which are capable of forming hydrogen-bonded complexes carboxylic acids associate with the solvent molecules rather than with their own species. Thus they give normal molecular weights in ethers esters and ketones 4 9 ; and in freezing dioxan trichloroacetic acid appears from cryoscopic measurements to be entirely monomeric although it is entirely dimerised in benzene at about the same temperature. A range of solvents was used by Brockle~by,~~ who found from ebullioscopic measurements that oleic acid exists entirely as monomer in boiling ether dioxan acetone or acetic acid. Equilibrium constants for the associat'ion of these compounds with carboxylic acids in such solvents as benzene have not yet been measured but it seems likely that we have here a way of studying quantitatively the association of carboxylic acids with ketones and other donor substances. 48 Hammett " Physical Organic Chemistry " McGraw-Hill 1940 Chap. 3. 49 Bell Lidwell and Vaughan-Jackson J. 1936 1708.

ISSN:0009-2681    

DOI:10.1039/QR9530700255

出版商:RSC    

年代:1953

数据来源: RSC

摘要:

By J. A. S. SNITH M.A. D.PHIL. (DEPT. OF INORGANIC AND PHYSICAL CHEMISTRY THE UNIVERSITY LEEDS) I. Introduction THE first prediction of nuclear magnetic resonance absorption was made by Gorter ; 1 several early attempts to detect the phenomenon in the solid state failed,l chiefly because of an unfortunate choice of compounds. However the theoretical prediction provided a stimulus for work in other fields and in 1938 the magnetic resonance method was applied to molecular beams,3 and gave results of considerable importance. * This Review is concerned with the effect in solids and liquids ; little further progress was made in this work until the close of the war. Great advances had then been made in the design of radiofrequency bridges and low noise-level amplifiers and with this assistance Bloch and Bloembergen Pound and Purcell,6 in 1946 were able to detect the phenomenon in the solid and in the liquid state.In the past six years these new methods have been used to study many problems of importance to chemistry. For example they can provide information on inter-hydrogen atom distances in crystals hindered rotation in the solid state and the internal structure of liquids. In this Review some of the more outstanding contributions of the subject to chemistry will be discussed with special reference to the internal properties of crystals. Much promising work still at an early stage of development has been omitted ; in particular we may mention recent studies of quadrupole effects the splitting of the proton resonance line at low temperatures and the fine structure of the liquid resonance line.More comprehensive reviews of these and other aspects of the subject are available.5-11 The novelty of the method in chemistry is that it is concerned essentially with the measurement of the magnetism of the atomic nucleus rather than that of the electrons from which the well-known bulk properties of diamag- netism paramagnetism and ferromagnetism originate. It has been known for some time notably from spectroscopic evidence that many atomic nuclei have a magnetic moment and that this is a consequence of the spin of the nucleus. As a rough analogy it can be imagined that just as the spin of the negatively-charged electron gives rise to a magnetic moment so the 1 Gorter Physica 1936 3 995. 4 Kusch Physica 1951 17 339. Gorter and Broer {bid. 1942 9 591. Bloch Phys.Review 1946 70 460. Rabi Zacharias Millman and Kusch Phys. Review 1938 53 318. Bloembergen Pound and Purcell &id. 1948 73 679 (referred to as BPP in Procertiings of the Internat ional Conference on Spectroscopy a t Radiofrequencies Purcell Science 1948 107 433. the text). P?ZYS~CU 1951 17 169-454. Pake Amer. J. Physics 1950 18 438 473. lo Idem Physica 1951 17 282. l1 Rollin Reports Progr. Physics 1948-1949 12 22. 279 280 QUARTERLY REVIEWS spin of a positively charged nucleus gives rise to a nuclear magnetic moment This is a highly simplified model as may be realised from the fact that the neutron which is uncharged also has a magnetic moment. This analogy to the electronic magnetic moment extends to the equations defining the nuclear magnetic moment; a nucleus with a spin of I has a magnetic moment of magnitude p = g p * v T T - G * ' (1) pn= eh/4nMc .(2) The quantity pn is the nuclear magneton,* and is defined by the equation g is the g-factor of the nucleus and may be compared with the Land6 g-factor for electronic spin ; its value is accurately known for a number of nuclei.12 The quantity equivalent to for the electron is the familiar Rohr magneton pe = eh/4nrnc and it will be seen that because of the difference in mass of the two particles nuclear moments will be a t least one thousand times smaller than electronic moments. The presence of nuclear magnets in any molecule gives rise to a nuclear paramagnetic bulk sus- ceptibility defined through the equation ~0 = Ng2/tn21(1 + 1)/3kT . * (3) which is equivalent to the Curie equation for substances displaying normal paramagnetism.Because of the factor pn2 in equation (3) nuclear para- magnetic susceptibilities will be generally lo6 times smaller than electronic paramagnetic susceptibilities and lo3 to lo4 times smaller than molecular diamagnetism ; consequently they would not be detected by the ordinary methods of measuring magnetic susceptibilities. Ideally the magnitude could be found by measuring the magnetic susceptibility of a diamag- netic compound at very low temperatures. Diamagnetism is largely inde- pendent of temperature but the relative contribution of the nuclear para- magnetism to the susceptibility is considerably increased when the tempera- ture is very low. Such an effect has been detected in one case by Lasarew and Shubnikow l3 in solid hydrogen in the range 1-76-4.22" K.When a nuclear magnet of spin I is placed in a magnetic field Ho the energy of the magnet in t,he field which is given by the equation where 8 is the angle between p and H, can assume 21 + 1 values ; i . e . there are 21 + 1 orientations of the nuclear spin with respect to t'he applied field. We shall be concerned mainly with hydrogen and fluorine nuclei for which I = + (and the quadrupole moment is zero). In this case the situation can be illustrated as in Pig. 1. Each nuclear moment has two components one of magnitude * gpn/2 parallel to the field and the other of magnitude g/c,/2/z perpendicular to the field. sign in front of the parallel components denotes the two possible nuclear magnetic energy The 12 Mack Rev. Mod. Physics 1950 22 64. 13 Lasarew and Shuknikow Physikal.2. Sowietunion 1937 11 445. * A complete list of the symbols used hero is given in the Appendix (p. 305), SMITH NUCLEAR MAGNETIC RESONANCE ABSORPTIOPU’ 281 levels (21 + 1 = 2) one in which the component is parallel and the other in which it is antiparallel to.the field the latter having the higher energy. The actual moment of a magnitude given by cquatiori (l) is inclined a t an angle to H and precesses round the field vector. As a result of this magnetic quantisation when a molecule cont,aining hydrogen or fluorine FIG. 1 atoms is placed in a magnetic field apart from the rotational and vibrational transitions which occur in the ground electronic state we also have the possibility of magnetic transitions between the two nuclear magnetic energy levels defined by their antiparallel components g/.tn/2 and - qp,/2.The Bohr frequency condition states that this difference in energy between the two levels is equivalent to a frequency Y given by For hydrogen nuclei in a field of strength 2350 gauss this frequency is 10 megacycles per second [which is in the RE (radiofrequency) region] and should be the same for all hydrogen nuclei in whatever type of molecule they happen to reside provided that the material is diamagnetic. For the 19F nucleus g is 5-257 and for the deuterium nucleus 0.8574 so that at the same frequency fluorine resonance occurs a t about 2500 gauss and deuterium resonance a t about 15,300 gauss. The precession of ,u around Ho will itself give rise to a magnetic field which we may resolve into two components a static coniponeilt H along H and a rotating component perpendicular to this.The precessional move- ment of the nuclear magnet about the applied field is familiar from Larmor’s classical theorem ; 14 any magnet of moment ,u in a field H will precess around the direction of the field with an angular velocity given by The frequency of the rotating field produced by this precessing magnet a t l4 The theorem is thoroughly discussed in Van Vleck “Electric and Magnetic Susceptibilities ” p. 22 (Oxford Univ. Press 1932). hv=gpnHo . * ( 5 ) ~0h/2n = gpnHo . - (6) 282 QUARTERLY REVIEWS right angles to H is cu/2n. If we superimpose on this system an osciEEatory magnetic field which we will call H, of the same periodicity and perpendicu- lar to H, the precessing nucleus and the oscillatory field can exchange energy and the nuclear magnet can be made to '' flop " between the parallel and the antiparallel states.Equation (6) which was derived classically is entirely equivalent to equation (5) which was derived from the Bohr frequency condition. This system of a static field H and an oscillatory field H, applied perpendicular to H, reproduces the essential coiiditions of the nuclear resonance experiment. The method is to measure the small change in the susceptibility of the material which occurs when an oscillatory field of the appropriate frequency is applied at right angles to the main field ; when equation (5) is satisfied nuclear resonance occurs. It will be shown later that there are several processes which maintain a greater number of nuclei in the ground state than in the excited state.This was in fact assumed in writing equation (3) because this is an expression for the bulk susceptibility and so depends on the number of spins in each of the two nuclear magnetic energy levels. At resonance the equilibrium distribution of the spins is disturbed thereby altering the susceptibility. Transitions between proton Zeeman levels (so-called because of the analogy to the Zeeman splitting of electronic energy levels in magnetic fields) occur in the 4-40 Me. region in fields of 1000-10,000 gauss the usual range of pract'ical values. Simultaneously with the magnetic quantisation of the nuclear magnetic levels all other types of magnetic moment possessed by the molecule will be similarly affected ; for example if the molecule possesses an unpaired spin the electronic magnetic moment will be quantised in the field with a fre- quency given by equation (5) with ,un replaced by pe.Since the latter is 1000 times larger than the former the frequency range over which resonance would be expected to occur is 4000-40,OOO Me. in the microwave region ; consequently electronic magnetic resonance or paramagnetic resonance does not interfere in nuclear resonance experiments. It will be seen that there is an important distinction between the resonance methods of measuring the nuclear or electronic susceptibility and the more familiar methods of measuring the bulk susceptibility such as the Gouy magnetic balance. In the latter the so-called static susceptibility xo is obtained denoting the magnetic properties of the molecules in a field of zero frequency.In the former the situation is very similar to the measurement of the dielectric constant of materials in high-frequency electric fields. In both cases we have the real components x' which rotate in phase with the applied frequency and the imaginary components XI' which are 90" out of phase. Because of the phase difference between x' and XI' the susceptibility x is a complex number and we must put where 1 x 1 = 4 ~ ' ~ + %'I2. by the system from the RF field Chis can be shown the other component - (7) x ==- - ix" ~ -_ If we are interested only in the energy absorbed to be governed by x" ; governs the dispersion of the real susceptibility. SMITH NUCLEAR MAGNETIC RESONAKCE ABSORPTION 283 II. Methods of Measurement One method of measuring x” the component giving rise to nuclear magnetic resonance absorption follows directly from this discussion.The RF field is produced by passing for example a 10-Mc. signal through a small induction coil containing about one C.C. of the proton- or fluorine-rich liquid and the coil is mounted with its cylindrical axis perpendicular to a homogeneous magnetic field of about 2350 gauss. The small inductance is made part of a twin-T RF bridge l5 which is so balanced that a very small change in the RF energy absorbed by the material inside the coil with consequent change of susceptibility suffices to throw the bridge off balance and allows a small signal to get through to a RF amplifier. The main FIG. 2 Block diagram of the RP bridge method (reprinted with permission from Trans. Faraday SOC. 1951 47 1264). magnetic field is not kept constant? but is made to oscillate or “ sweep ” over a range of field strength which is much larger than the region over which the nuclei of the liquid absorb.This is done by sending a low- frequency current through subsidiary coils attached to the magnet ; the frequency is known as the modulatory or sweep frequency and quite often falls in the 30-60 cps range. If the mean field is now brought to within a very small distance of the resonance field strength for the particular nucleus and frequency concerned the modulation causes the field to sweep through resonance twice for each cycle of the sinusoidal modulatory sweep. If the amplified signal from the bridge is then applied across the Y-plates of a cathode-ray oscillograph and a tapping off the sweep to the X-plates a stationary picture of the absorption line will be thrown on to the screen provided that the RF bridge has been properly adjusted.15 Tuttls Proc. Inst. Radio Eng. 1940 28 23. 284 QUARTERLY REVIEWS This is a general outline of the RP bridge method of Bloembergen Purcell and Pound (generally abbreviated to BPP) and a block diagram of the apparatus is given in Fig. 2. Fig. 3 is a reproduction of the proton resonance line of water a field of 3900 gauss being used with an inhomo- geneity of about 0.1 gauss. (The term inhomogeneity refers to the variation of the main field over the volume of the sample.) It will be seen that the line has a definite width denoting that the nuclei do not all absorb a t the single frequency that equation ( 5 ) predicts. All gases and liquids have com- paratively narrow absorption lines of this type and in fact in most liquids the lines are so narrow as to be governed entirely by the field inhomogeneity unless highly uniform fields FIG.3 The proton resonance absorption line of water. positions in the pole gap of are used. This is the case for the absorption line of water. The inhomogeneity of the field may cause the line to have a width of the order of 0.1 gauss but the expected natural line width (AHm.s,J is about 3.4 x lop5 gauss.16 The line width AHm.sl. is the distance in gauss between the two points on the absorption curve where the slope has its maximum value. As a consequence a small nuclear resonance coil containing for example water can be used in the form of a probe for measuring field strength and field inhomogeneity at various a magnet.l' Because of the sharpness of the resonance line in liquids and gases the method has been widely employed in the measurement of nuclear magnetic moments.Two recent applications include the first accurate measurement l8 of the nuclear magnetic moment of 3He and a direct determination of the magnetic moment of the proton in nuclear magnetons.19 Many solids however especially at low temperatures give much broader lines some 10-100 times wider than in liquids and showing marked fine structure. The signal in this case is correspondingly weaker and for very broad lines would be indistinguishable from the background of the line on the cathode-ray screen. This background or '' noise " arises from thermal agitation of the atoms in the coil and the bridge and the fluctuations gener- ated in the first amplifying stage.The ratio of the signal strength to the noise is the factor which determines the sensitivity of the nuclear resonance apparatus and an equation for the ratio in terms of the various constants of the circuits used and the materials studied has been derived by BPP.6 A more sensitive method of recording broad lines is to use a modulatory sweep equivalent t,o a much smaller range of field strength than the line width. The signal then passes from the RF amplifier to a low-frequency amplifier of narrow band-width which amplifies the sweep-frequency com- ponents of the signal and thence to a homodyne rectifier which converts the signal to direct current and a " lock-in ') meter which is a device for 16 Torrey Phys. Rewiew 1952 85 365. 18Anderson Phys.Rewiew 1949 76 1460. l9 Bloch and Jeffries ibid. 1950 80 305. l7 Gooden hTature 1950 165 1014. SMITH NUCLEAR MAGNETIC RESONANCE ABSORPTION 285 reducing the relative contribution of noise to the experimental signal. It can then be shown that as we move the main field through the range in which resonance occurs the derivative of the line-shape is recorded on the meter,s and this is integrated to obtain the absorption line. This part of the apparatus is also drawn in Fig. 2 . Other methods of detecting nuclear resonance may be mentioned briefly. They fall into four groups. Instead of altering the field the line can be traversed by varying the frequency and keeping the main field constant. This is a feature of the RF spectrometer designed by Pound and Knight.2o The instrument is capable of the same degree of accuracy as the bridge method and has the additional advantage that only the pure absorption line is recorded so that the centre-points of the resonance line are accurately located.In the bridge method both absorption and dispersion curves can be plotted according to the type of balance used and a mixture of the two is obtained if the bridge is not properly adjusted. The second group con- tains methods using super-regenerative techniques.21 22 They provide little information on line-shapes and have been used mainly in measuring nuclear magnetic moments. Simultaneously with t'he development of the RF bridge method a t Harvard Bloch at Stanford was successfully applying a different technique to which he gave the name of nuclear ind~ction.~ The characterist,ic feature of the method is the use of a second coil round the substance to be measured which has its axis perpendicular both to the Zeeman field and to the RF field of the first coil.At the resonance fre- quency there is a sudden change in the orientation of the nuclear moments which induces an electromotive force in the second coil. The field is modu- lated as in the RF bridge method and both the RF and the sweep-frequency components of the signal are amplified successively. The method has practically the same sensitivity as the RF bridge method and much the same range of application ; moreover it can furnish the sign of a magnetic moment which is governed by the orientation of the nuclear magnetic moment with respect to the electronic moment and so has been much used in measuring nuclear moments.23 The fourth group includes the pulse methods which have been developed recently by Hahn 24 and T0rrey,~5 and have proved to be particularly useful in measuring the fine structure of the resonance line of some liquids.III. Chemical Applications The Relaxation Time.-The line width and the shape of the derivative and absorption lines are quantities of considerable interest to chemistry. I n order to discuss their application we must consider in very general terms Ihe conclusions that can be drawn about the process of absorption from quantum mechanics. Previously transitions between nuclear energy levels 2o Pound and Knight Rev. Sci. Instr. 1950 21 219. 21 Roberts ibid. 1947 18 845. 22 Zimmerman and Williams Phys. Review 1949 76 354. 23 Proctor and Yu ibid.1950 77 716. 24 Hahn ibid. 1950 80 580. 25 Torrey ibid. 1949 76 461. U 286 QUARTERLY REVIEWS have been discussed as though they were caused only by interaction with the radiation field However at nuclear resonance frequencies the processes of absorption and induced emission are effectively equal which means that no set signal would be obtained in the nuclear resonance experiment. Clearly since signals are obtained there must be other processes which cause the nuclear spins to relax to a lower energy level; the magnetic energy of the spin in the applied field must be convertible into some other form of energy. There are several processes such as spontaneous emission and interaction with the thermal radiation field which will provide a source of relaxation but the lowest life-time which they predict for an excited state is about lo3 years whereas experiment shows that nuclear resonance life-times are of the order of seconds.We must therefore seek alternative relaxation mechanisms which will convert magnetic energy into thermal energy. They are affected only by other magnetic fields ; and in most molecules the electrons which are capable of producing large magnetic fields as a result of their orbital and spin motion are paired off in the orbitals of the molecule? their magnetic effect in this case being very small. The situation may be compared (Purcell 9 to the analogous case of the line width of spectral lines in ordinary spectroscopy. Here,.in contrast an atom or molecule can be deactivated by collision with another ; this reduces the life-time of the excited state and as a result of the Heisenberg equation Nuclear spins are well shielded from deactivation mechanisms.AE.At 2 h/2n . ' (8) a reduction in At which represents the life-time of an excited state increases the uncertainty in E the energy of the level and so broadens the spectral line. The analogy shows that in the case of nuclear resonance? the two problems of finding how an upper spin state loses its energy and how the resonance line is broadened are complementary. Any mechanism which supplies an effective deactivation of an excited spin level produces by equation (8) a contribution to the line- broadening. We shall frequently discuss line- broadening in terms of the relaxation times rather than the broadening of the energy levels. Calculations have shown 5 9 that two in.teractions are of predominant importance in determining the width of the proton resonance line.To anticipate the later discussion the first arises from interaction of the pre- cessing nucleus with oscillatory magnetic fields produced by molecular motion in the lattice and the second from interaction with the oscillatory and static magnetic fields produced around it by neighbouring nuclei with a magnetic moment. The essential distinction between the two is that the first is the mechanism whereby magnetic energy of the nuclear spins of the material (or the spin system) can be converted into thermal energy of agitation of the molecules whereas the second mechanism involves merely an exchange of energy between neighbouring spins no energy entering or leaving the spin system.The former effect clearly provides the alternative relaxation mechanism SMITH NUCLEAR MAGNETIC RESONANCE ABSORPTION 287 which we have previously shown to be necessary to obtain a net nuclear resonance signal. Its origin can be pictured in the following manner random movement of the molecules causes the nuclear magnets embedded in them to undergo oscillations with a wide range of frequencies. This produces a corresponding range of oscillatory magnetic fields and if this range overlaps the Larmor resonance frequency of the nucleus concerned energy can pass from the spin system to the thermal energy of molecular agitation. In the solid state this random motion of the molecules which is taken generally to involve complete movement of the molecule or part of the molecule from one equilibrium position to another must be dis- tinguished in its effect on the second moment from the lattice vibrations which govern the specific heat of the solid (see p.292). These are present in all crystals but their contribution to the relaxation time is considered to be negligibly small in nearly all compounds so far examined.1° The spins of the compound can now be assumed to be in thermal equilibrium which means that the spin state of the lowest energy is favoured. In other words the theoretical transition probabilities must be weighted with Boltzmann factors and this is found to give an excess number no in the ground state of no = N I g IHoh/kT * ' (9) where N is the total number of resonating nuclei. For hydrogen nuclei at 90" K in a field of 2500 gauss the difference in distribution between the two levels is only about six per million nuclei and yet it is the signal from this small number that the apparatus will detect.When the RF signal is applied this equilibrium will be disturbed and the actual excess number n between the lower and the upper state will depend on the rate at which the spins can relax. The rate a t which n approaches no when the RF field is removed is given by where t is the time. The constant T is known as the spin-lattice relaxation time and from equation (10) it can be defined as the time taken for a fraction 0.63 of the original excess number to relax to the ground state when the R F field is switched off. One method of measuring T for liquids is to increase the RF signal to the coil and simultaneously observe the maximum output on the " lock-in " meter; because of the finite relaxation time saturation will set in at a certain point defined under certain conditions by the equation n/n = 1 - e-ljT1 .(10) y2HI2T,T2 = 1 . - (11) in which T is the relaxation time corresponding to the second mechanism Hl is the R F field and y the gyromagnetic ratio of the resonating nucleus (see Appendix.). H can be found by observing the saturation point in a substance of known T and T, and equation (11) can then be used for finding T in other materials provided T is known ; in many liquids T is governed by the field inhomogeneity. The first measurement of Tl for water was made by observing the nuclear resonance line under conditions in which saturation was occurring and then photographing the rate at which the 288 QUARTERLY REVIEWS signal grew when the saturation condition was suddenly removed.g It must be noted that all measurements of the line-shape and line-width must be carried out at RF levels below the saturation limit and so the value of T is an importlant factor in determining whether or not a suitable nuclear resonance signal can be obtained from a compound.(Other methods can be used for measuring The value of TI for pure water a t 293" K is 2-33 5 0-07 seconds 26 and for ammonium chloride a t 90" K is about 100 seconds. 29 If we put the last value in equation (S) the predicted line width is about The reason for the differ- ence is that in many solids the predominant relaxation mechanism is the second of the two which we have previously mentioned that arising from neighbouring nuclear spins in the same or surrounding molecules.Any two nuclear magnetic dipoles whose moments are parallel and separated by a distance of r cm. will produce a t each other a maximum magnetic field of approximately p/rS gauss which is 4.1 gauss for the two protons in a gaseous water molecule. The effective magnetic field a t any point inside a diamagnetic crystal will be therefore that of the applied Zeeman field plus a second term which will depend on the distance of all neighbouring nuclei with magnetic moments from that point and the direction of their magnetic moments with respect to the applied field. It is this second term which gives rise to the line-broadening observed in ammonium chloride and many other proton- and fluorine-containing compounds. Alternatively we can consider the effect in terms of relaxation times rather than line-broadening.From the Heisenberg uncertainty principle the broadening will correspond to a relaxation time denoted by T and known as the spin-spin relaxation time. There are two separate contributions to T, which we can picture in physical terms as follows. Any nucleus of moment p precessing about the applied field H will produce at a neighbouring nucleus a magnetic field which can be resolved into two components. The first is a static field along H and if this has the value aHst. it will spread the precessional frequency of the second nucleus by an amount This broadening of the magnetic energy levels corresponds to a relaxation time of l/acost. seconds. The second component will be an oscillatory com- ponent perpendicular to H ; if the second nucleus has a component in the same direction and of the same frequency energy can pass from one spin to the other (a process known as spin exchange) and the effective life-time of any spin state is thereby reduced.The sum of these two effects gives the spin-spin relaxation time T,. T can be defined with respect to a quantity known as the mean-square frequency of the line and given the symbol ( A W ~ B ) ~ ~ . . Pig. 4 is a plot of an absorption line the example chosen is in fact a Gaussian curve such as 26 2 7 3 28) gauss. The experimental value is 23 gauss. (h/Wa%t. = SpnaH,,. * - (12) 28 Hahn Phys. Review 1949 76 145. 27 Pake and Gutowsky ibid. 1948 74 979. 28 Drain Proc. Phys. Xoc. 1949 62 301. 20 Sachs Turner and Purcell quoted by Purcell Physics 1951 17 282.SMITH NUCLEAR MAGNETIC RESONANCE ABSORPTION 289 is given for example by ammonium chloride a t room t e m p e r a t ~ r e . ~ ~ The quantity F ( Y ) is known as the line-shape function a t the frequency Y and is V FIG. 4 A Gaussian absorption curve. a measure of the intensity of absorption a t that particular frequency; Y* is the frequency at the centre of the line. (Aw2)Av. is then given by the equation 00 (Aw2)Av. = J P(v). (Y - v*)2dv . ' (13) - m The simplest method of treating T is to define it with respect to ( A C O ~ ) ~ ~ . by the equation Equation (14) strictly has meaning only if the line has the shape of a Gaussian curve; otherwise the equation defines a sort of mean value of T which is still useful in drawing roughly quantitative conclusions. By using equation (6) equation (14) can be rewritten as 1 P 2 ) = d(Aw2)*v.* (14) The quantity (AH2),. in units of the gauss2 is known as the second moment of the absorption line and is a very useful way of expressing the magnitude of the line-broadening. In the nuclear resonance experiment the final recorder plots the derivative curve of the absorption line and the latter is derived from the former by integration. The second moment of the absorp- tion line is defined in the same way as the mean-square frequency through the equation 00 (AH2)Av. = 5 F(H).(H - H*)2dH - (16) -co where P(H) is the line-shape function at any value of the field H and H* is the resonance field strength for an isolated nucleus Le. the value of 12 at the centre of the absorption line. The second moment can therefore be derived directly from the experimental data by finding the total area under 30 Pake and Purcell Phys.Review 1948 '74 1184 ; erratum &bid. 1949 75 634. 290 QUARTERLY REVIEWS a plot of the function F(H) . (H - H*)2 against H although a method due to Pake and Purcell O using the derivative curve is more convenient in practice. Ammonium ChZoride.4ne important application of the nuclear reson- ance method lies in the fact that under certain conditions the second moment of the absorption line bears a direct relationship to the distance between the resonating atoms. The reason is that T depends on the size of the oscillatory magnetic fields perpendicular to Ho produced by one nucleus a t another and these are very sensitive to the internuclear distance. If the resonating nuclei are hydrogen nuclei we can therefore measure interproton distances in crystals a fact of considerable importance to structure analysis.Now in ammonium chloride at 90" K T is about 5 x 10-8 sec.31 and T is nearly 100 sec. Consequently the broadening is due solely to T, and so depends on the interproton distances in the salt. The equation relating the second moment of a powdered crystal to the distance rjk between resonating atoms has been calculated theoretically by Van V l e ~ k ~ ~ and can be written as 31 <AH2)A,. = where the f subscripts refer to nuclei with magnetic moments other than the resonating nucleus which nevertheless contribute to the broadening. NI represents the total number of resonating nuclei in the particular sub- group (e.g. unit cell molecular or molecular complex) to which the broaden- ing is attributed.It will be noticed that the second moment depends on rjk-6 illustrating the high sensitivity of the former to slight changes in the interproton distance. The two important conditions for equation (17) to hold are (1) that the lattice must be rigid a t the temperatures a t which the experiment is carried out or in more precise terms the resonating nuclei must not have moved appreciably in a time comparable to T ; and (2) that there must not be paramagnetic atoms present in greater atomic ratio to the resonating atoms than If this ratio is exceeded the oscillatory magnetic fields which are produced by thermal agitation of the para- magnetic atoms or their low relaxation times will compete with the spin-spin interactions and so alter the second moment.32 A further condition which we have assumed so far is that the resonating nuclei have no quadrupole moment ; this is necessarily true for nuclei of spin i.These conditions are fulfilled for proton resonance in pure ammonium chloride at 90" K and this compound will now be discussed in more detail. The derivative and absorption lines are shown in Pig. 5.31 The distance between the two peaks on the derivative curve is 22 gauss and the second- moment of the absorption line is 57.0 & 3.0 gauss2. For our sub-group we take a single molecule of NH,C1 for which N = 4. From equation (17) 31 Gutowsky Kistiakowsky Pake and Purcell J . Chem. Phys. 1949 17 972. 33 Van Vleck Phys. Review 1948 74 1168. SMITH NUCLEAR MAGNETIC RESONANCE ABSORPTION 29 1 there will be three separate contributions to the second moment ; the first two arise from neighbouring hydrogen atoms in the same NH,+ ion and in neighbouring NH4+ ions and the third from neighbouring C1- ions both isotopes of which have a magnetic moment.The last two contributions can be calculated from the known crystal structure of NH4C1 at 88" K and prove to be very small. Subtraction of these values from the experimental I I I I I - 2 0 - 1 0 0 10 20 Gauss FIG. 5 Derivative and absorption lines of ammonium chloride at 90" K (reprinted with permzssaon from J. Chem. Phys. 1949 17 972). second moment leaves 50.5 gauss2 for the contribution of the NH,+ ion alone which is consistent with an N-H bond distance of 1.025 5 0.005 A in excellent agreement with the latest neutron-diffraction value 33 of 1-03 & 0-02 A.Ammonium chloride is one of the most favourable cases studied in this way. The interproton distance is small giving rise to a high second moment and a broad line ; T has a sufficiently low value at 90" IC to give a reasonable signal-to-noise ratio. Although there is an error of 10% in the second moment this quantity depends on the sixth power of thc interproton distance and the corresponding error in the bond distance will therefore be a sixth of this. Consequently the method is able to furnish in favourable 33 Levy and Peterson Phys. Review 1952 86 766. 292 QUARTERLY REVIEWS cases values of interproton distances as accurate as X-ray values for bonds involving heavier atoms. The most favourable cases occur in compounds preferably of known crystal structure in which most of the hydrogen atoms are in the form of XH groups as for example the CH group in CH,*CCl, the H20 molecule in hydrates and the hydroxonium ion in perchloric acid monohydrate provided that there are one or more convenient temperatures at which the lattice can be considered to be rigid and at which T has a reasonable value.The method can then furnish considerable information on the disposition of the hydrogen atoms in a crystal and can be of valuable assistance in an X-ray analysis. The Eflect of Molecular Motion.-One drawback of the method is that a " rigid lattice " (see p. 290) is required. This condition is unlikely to hold for all temperatures up to the melting point of a solid and it will certainly not hold in the liquid or the gaseous state. I n particular a number of solids including ammonium chloride are known to possess one or more transition points atl which there may be abrupt changes in the crystal symmetry the density or the specific heat.These have in certain cases been associated with the onset of rotation or vibration of molecules or ions in the crystal.34 If a molecule containing protons is free to rotate in all directions the internuclear fields will be largely cancelled out and conse- quently the line-width and second moment will decrease. The problem has been treated quantitatively by BPP whose paper ti gives a full account of the theory. A simple way of discussing the theory is to consider a system of two identical spins in a magnetic field H,. The two spins can be located in either the same or adjacent molecules. The energy of this system depends on the angles made by the nuclear magnets pl and p2 and by the vector joining the two moments r12 with the field H,.The equation as derived classically is The change in the orientation of pl and p2 with time due to the influence of (rl and T2 gives rise to the line-broadening which has already been discussed The random distribution of Y, in space but not with time is the effect we encounter in the powdered crystal. Suppose now that rI2 varies with time that is the interproton vector oscilla'tes or rotates with respect to the direction of the applied field. We can distinguish two types of motion. The first is the lattice vibrations of the crystal which for the moment we omit from discussion The second occurs in molecular crystals e.g. hydrates in which the H,O molecules can rotate over energy barriers ; because of the Boltzmann distribution of thermal energy there will always be a small fraction of molecules at any temperature which will have sufficient energy to surmount such barriers.We also include in this group other types of motion which cam6 complete reorienta- tion of a proton-containing group such as quantunz-mechanical tunnelling. Consider now what we can expect to happen as the temperature of the crystal rises. The frequency with which this Orientation of r12 occurs will rise. W12 = p 2 / ~ 1 2 ~ - 3p1 - ~ 1 2 ~ 2 - ~ 1 2 / ~ 1 2 ~ * (18) This type of motion is usually random. 3 4 Staveley Quart. Reviews 1949 3 66. SMITH NUCLEAR MAGNETIC RESONANCE ABSORPTION 293 At a certain temperature it will become comparable to the frequency of spin exchange which is the reciprocal of T,.Above this temperature the two processes will compete in reorienting pl and p but the former will rapidly overtake the latter as the temperature continues to rise. The spin-exchange process therefore becomes much less effective in inducing relaxation causing T to rise and so reducing the line width. It must be borne in mind that T makes a negligible contribution to the line width a t these temperatures and often we have to wait until the melting point before T and T approach each other in value. One important fact which has emerged from experi- ment is that the first type of motion involving lattice vibrations is generally much less effective than the second involving reorientation of r12 and there are few if any crystals in which the former can be said to be the predominant relaxation mechanism.l* However the distinction we have drawn between lattice vibrations and group rotations is rather arbitrary.In particular if the potential-energy curve of the rotation has a broad minimum the group may undergo rotational oscillation with large amplitudes e . g . of the order of 3045" even at low temperatures. The second moment of the nuclear resonance line may then be reduced and the effect has been calculated for the two-spin group by A n d r e ~ . ~ 5 In general the reduction is smaller than that caused by rotation of the group ; a sudden change in the line width is more likely to be caused by group rotation than the lattice vibrations. The calculations of BPP show that it is convenient to discuss the expected variation of T and T due to molecular motion in terms of a quantity zc known as the correlation time which can be considered to be the time in which a molecule is rotated to such an extent that the value of the second moment has altered by a detectable amount.Since the latter is governed by T, it follows directly from this definition that the line will begin to narrow when zc approaches T, and it will be recalled that this will break one of the conditions which determine the application of equation (17). Using statistical theory BPP show that T and T can be written in terms of zc and that as zc varies both quantities show the characteristic changes given in Fig. 6 which shows a plot of log T and log T against log zc. These variations are derived for a simplified model in which all interactions pro- ducing relaxamtion are assumed to have the same value of the correlation time.But zc being a quantity referring directly to the freedom of move- ment of the molecule will change with temperature. In general it will decrease as the temperature rises although its precise dependence may not be known a priori. In the case of liquids BPP show that it is often directly proportional t o the Debye relaxation time z which is the time taken for an assembly of polar molecules originally oriented by an electric field to assume a random distribution through Brownian movement when the field is removed. Since t is inversely proportional to the temperature zc in this case shows the same dependence. In solids the precise form of the equation coimecting zc and the temperature is not generally known but the graphs in Fig.G can frequently be related to the variation of T and T with t'empera- ture. Thc forins of the curves for 2' and T show some of the characteristics 35 Andrew J . Chem. Phys. 1960 18 607. 294 QUARTERLY REVIEWS of these quantities which we have already mentioned. In particular a t high values of zc when 2ntyzc > 1 i.e. in an effectively rigid lattice (Y is the resonance frequency) T tends to a constant value Tz" which proves to be the value defined through equations (17) and (15). In this region T is Zc,sec. FIG. 6 Variation of TI and T with rc (reproduced with permission from Phys. Review 1948 73 679). directly proportional to zc and so rises as the temperature falls. At low values of zc when 2zvzc << 1 i.e. in liquids T is inversely proportional to zc and so rises as the temperature increases.The minimum which TI approaches when 2z1/zc tends to unity occurs theoretically a t the point when zc equals (d?j.~B~)-~. T assumes the rigid lattice value when the correlation time is given by xc = ( ~ S ~ . I A H ~ . ~ I . ) - ' . * (19) Next we must discuss how closely these relationships are obeyed in practice. We begin by resuming the discussion of ammonium chloride. There are measurements for this compound of TlZ9 and the line width AHm.s1.36 covering the range 90-290" K. The plot of the line width is given in Fig. 7. At about 130" K the line width of ammonium chloride decreases from 23 to 5 gauss over a temperature range of 16" ; above and below this temperature it is nearly constant at 5 and 23 gauss respectively.This rather sudden change in T might not be expected from Fig. 6 a,nd we have no precise way of predicting the varicttion of zc with temperature except 36 Gutowsky and Pake J . Chem Phys. 1948 16 1164. SMITH NUCLEAR MAGNETIC RESONANCE ABSORPTION 295 for some evidence from the variation of T, which is discussed later. Up to 130" the line-width curve is horizontal representing the rigid lattice part of Fig. 3 where 23tvtc>l. The frequency of rotation of the NH4+ ion from one equilibrium position to another has meanwhile been slowly increasing and at 130" K reaches the critical value above which it begins to reduce the line width. This value is the reciprocal of T, which is 3 x lo4 cps. It will be noticed that the limiting frequency is very smdl on the thermodynamic scale where one is dealing with rotational and vibrational motions of frequen- cies greater than 1O1O cps.Con- sequently the changes in the line T O K FIG. 7 AHm8,,-T curve for ammonium chloride (re- produced with permission from J. Chem. Phys. 1948 16 1164). width need not necessarily correspond to or be related to any transition points or changes in the lattice constants although they may be related to variations in the dielectric constant when we have polar molecules in the lattice. Above the transition temperature of 130" K it has been shown 2 9 that zc varies exponentially with temperature and the results for ammonium chloride can be expressed to a very good approximation by the equation It has been suggested that the form of equation (20) shows that we are dealing with simple rotation over an energy barrier of a height of 2-37 kcal./mole,'and similar equations have been proposed for zc in a number of other compounds notably NH4BrZ9 and CH3*CC1,.37 If this picture is true the fraction of NH,+ ions having sufficient energy to surmount an energy barrier of 2370 cal./moIe is at room temperature.Free rotation of a fraction lop8 of the NH4+ ions present is sufficient to reduce the line width by a factor of 5 which shows the extraordinary sensitivity of the nuclear resonance line to this form of motion. This interpretation of the line-width transition may appear to conflict with the latest neutron-diffraction results for ammonium chloride,33 which suggest that the second-order transition a t - 30" c is of the order-disorder type. Below this temperature the four hydrogen atoms of any one NH4+ ion occupy only four of the eight positions available for them in the lattice ; above these eight positions are occupied at random.Transi- tions from one set of positions to another will be proceeding at all tempera- tures but the rate at which the ions reorient over the energy barrier hinder- ing the process will rise with temperature. At 130" K it reaches the critical zc = (26 x lO-l*) exp 2370/RT . * (20) a t 138" K and 3 X A general picture of the situation has been given by Purcell.l0 37 Gutowsky and Pake J . Chem. Phys. 1950 IS 162. 296 QUARTERLY REVIEWS value above which it begins to reduce the line width. This value is the reciprocal of T which is 3 x lo4 cps. Above 130" K the frequency rises still further until the molecule begins to spend an appreciable amount of its time in intermediate positions The intermolecular forces will then begin to change and eventually this will bring about a transition to a disordered lattice with a space-group of a higher symmetry but with little further change in the line-width.The neutron-diffraction results also indicate that at room temperature the NH,+ ion as a whole is undergoing rotatory oscilla- tion with a half-angle of about ll" but this type of motion has not so far been shown to have a detectable effect on the line width. The simple exponential form of equation (20) would not be expected if we were dealing with quantum-mechanical tunnelling through the energy barrier. Newman 38 has shown that this process might be occurring in potassium dihydrogen phosphate in which zc is found to be almost inde- pendent of temperature over a certain range.Unfortunately it is not known to what extent paramagnetic impurities contribute to the result and no precise calculation of the effect of quantum-mechanical tunnelling was made. In liquids the random thermal oscillations undergone by the molecules are called Brownian movement. This type of motion is very effective in inducing relaxation and gives rise to proton spin-1atDice relaxation times of the order of seconds and values of T ranging from 1 to sec. Measure- ments on liquids have proved very useful in testing the BPP theory. For example a plot of log T against log q/T where 7 is the viscosity should be linear and the agreement for glycerol and ethyl alcohol between - 35" and 60" c has been found to be quite satisfactory.6 Among other interesting properties of the liquid resonance line it has been found that the presence of paramagnetic ions slightly alters the value of the frequency a t which resonance occurs,39 and has a marked effect on T and T ; for example in lO-S~-manganese sulphate solution T I has dropped from 2.33 to 0.09 sec.This sensitivity has been used by Selwood and his collaborators 40 to deter- mine the effective paramagnetic activity on supported oxide catalysts. As already noted T in a number of liquids is so small that it is completely masked by the field inhomogeneity ; however it has been directly measured in a number of branched-chain hydrocarbon^,^^ in which it shows an inter- esting dependence on the symmetry of the molecule. In highly homogeneous magnetic fields the proton and fluorine resonance lines of some compounds are split into two or more peaks and even in liquids showing single lines the exact frequency of the peak centre varies from compound to compound.This additional contribution to the line broadening is small in proton resonance lines but rather more important in fluorine lines ; for example in fields of about 6365 gauss a t n fixed frequency the resonance field strength of 1H and lBF nuclei in a number of coinponncis has been found to vary Newinail J . Chem. P~L,IJs. 1950 18 670. 39 Gabillard and Soutif Compt. rend. 195-1 233 450. 40 Selwood and Schroyer Discuss. Faraduy Soc. 1950 8 337. 41 Clay Bradford and Strick J. Chm. Phys. 1951 19 1429. SMITH NUCLEAR MAGNETIC RESONANUE ABSORPTION 297 over a range of 0.12 and 3.98 gauss.42 The phenomenon promises to give results of considerable interest but the subject is developing rapidly and the reader is referred to refs.24,42 and 43 and recent issues of the Physical Review. The Line 8hpe.-There is yet another aspect of the line which we have not discussed and that is its shape. The simplest assumptions that can be made about the form of the absorption curve are 30 that it corresponds either to the familiar Gaussian distribution or to a Lorentzian distribution. The latter is characteristic of a damped oscillator and is found in the pressure broadening of spectroscopic lines. In practice few compounds show either of these simple line shapes; the Gaussian form holds for ammonium chloride at 290" K and the Lorentzian for polytetrafluoroethylene a t the same temperature.The majority of proton resonance lines have shapes somewhere in between the two forms and it is generally impossible to calculate what line shape to expect.30 The exceptions to this statement occur when the disposition of protons in the crystal is such that there are groups of resonating nuclei particularly close to each other and a considerable distance from any other sources of broadening as for example the H20 molecule in certain hydrates the CH group in CH,.CCl, and the NH4+ ion in NH4C1. Since the second moment varies as the inverse sixth power of the interproton distance these groups make by far the largest contribution to the broadening. In these cases a rigorous quantum-mechanical calcula- tion can be made of the line shape which can then be compared with the experimental curve and second moment.The calculation of the line shape for the three- spin group of a CH radical (which is assumed to be symmetrical) is carried out in three stages.44 (1) First the number and frequency of permitted transitions in the three-spin group are found quantum-mechanically ; these are nine in number. (2) Secondly the effect of a random orientation of three-spin groups over 360" is calculated in order to be able to apply the results to a powdered crystal; this gives us a line shape with nine sharp maxima. (3) The broadening effect of neighbouring three-spin groups and foreign nuclei with magnetic moments is allowed for by applying a Gaussian broadening function to every fine-structure component of the curve in (2). That is the final line-shape function F(H) at any value of the field H is found from the equation The three-spin group.F(H) 1 F(W,- - H*).S(H - H,).dH . ' (21) Sm_ where F(H0 - H*) is the line-shape function found in (2) and S(H - H,) is a Gaussian broadening function proportional to exp - (W - The area under a plot of F(H - H*) . S(H - H,) against H gives the line-shape function a t the field H. This process removes much of the fine structure of the curve and in CH3*CC1 leaves us with a triply-peaked absorption a2 Gutowsky and Hoffman J Chem. Phys. 1961 19 1259. 43 Ramsey and Purcell Phys. Review 1952 85 143. 4 4 Andrew and Bersohn J . Chern. Phys. 1960,18 169 ; erratum ibid. 1962 20 924. 298 QU9RTERLY REVIEWS curve Le. a subsidiary maximum disposed symmetrically on either side of the central main maximum.In the case of the two-spin the final curve has two maxima symmetrically placed on either side of a central minimum. The calculations for the four-spin system are quoted by Purcell,lo and the experimental curve for ammonium chloride a t 90" K is flat-topped. The rigorous calculation of these line shapes requires considerable information about the structure of the crystal in order that a reasonable form for the Gaussian broadening function may be evaluated. Otherwise the broadening effect of neighbour- ing nuclei has to be calculated by analogy with structures of similar com- pounds or by using van der Waals radii to determine the nearest distance of approach of the relevant groups. The first three-spin system studied in this way was the CH group in CH,*CCl,. The experimental line-shape and second moment were known 31 and Andrew and Bersohn then showed 44 that by assuming a symmetrical FIQ.8 The absorption line of CH,.CCI (reproduced with permission from J . Chem. Phys. 1950 18 159). A CH group with a tetrahedral value for the HCH angle and a C-H bond length of 1-10 A the best fit to the experimental curve occurred when a Gaussian broadening function of second moment 2-1 gauss2 was used. The two curves are superimposed in Fig. 8 ; the full line gives the experimental results and the broken line the calculated line-shape. There is a bad misfit at the sides of the curve which was attributed to the residual effects of the rotation of the CH3 group which reduces the line width from 28 to 3 gauss in the vicinity of 145" K . ~ ~ . It has been shown 46 that if the experimental data on the variation of the line width with temperature are combined with the 4 5 Pake J .Chem,. Phys. 1948 16 327. 4 6 Gutowsky and Pake ibid. 1960 18 162. SMITH NUCLEAR MAGNETIC RESONANCE ABSORPTION 299 BPP equations predicting the variation of z with the former quantity equation (22) can be derived. This suggests that the CH group undergoes hindered rotation over an energy barrier of height 7000 cal. ; when a fraction 10-11 of the molecules have sufficient! energy to surmount the barrier the motion begins to reduce the line width. The presence of residual low- frequency motion a t low temperatures is supported by the gradual decrease of the dielectric constant down to 150" K.47 The CH group seems to be particu- larly prone to rotation of this sort and a t 90" K the critical fraction of rotating methyl groups required to narrow the line is exceeded in a number of compounds notably C2H6 CH,*CN CH,I CH,*NO, HgMe2,46 and Van Vleck's equation for the broadening in a rigid lattice can be modified to allow for the effect of hindered rotation or quantum-mechanical tunnel- ling; 46 the equation then includes a function depending on the angle between the axis of rotation of any one interproton distance rjX and r,k itself.I n the three-spin system it is found that when the axis of the motion is normal to the plane of the three spins the intramolecular second moment (i.e. the broadening contribution of any one three-spin group in isolation) should be one-fourth as great as that in the rigid lattice. The intermolecular broadening ( i .e . between neighbouring three-spin groups) will also be reduced but not generally in the same ratio. However the latter is usually small and the effect of this type of motion is generally to reduce the total second moment by a factor of 4 as has been verified in CH,*CN and NH, and more approximately in CH,I CH,*NO, and HgMe,.46 The low second moments of o- m- and p-xylene mesitylene and hexamethylbenzene a t 95" K have also been explained by rotation or tunnelling of the CH groups about the C-C side bonds.35 Andrew and Bersohn 44 have also calculated the line shape to be expected for a symmetrical three-spin group which is rotating about the normal to its plane and by comparing this with the experimental line shape further confirmation of the proposed motion can be obtained. This analysis has been successfully carried out for CH,*CN a t 93" K,44 in which the CH group is found to be rotating round the C-C bond.One three-spin group of great interest to chemistry is the hydroxonium ion H,O+. It has been inferred for some time that this ion exists in a number of compounds in the solid state and evidence from nuclear magnetic resonance absorption has confirmed many of these conclusions. The char- acteristic triply-peaked curve has been found in three hydrates of strong acids a t 90" K namely HN03,H20,48 HC104,H20,48~ 4 9 and H2PtC16,2H20,50 all of which give very similar absorption curves. For example the outer maximum and the minimum of the absorption curves of HNO,,H,O fall a t 13.4 and 9.3 gauss those of HClO,,H.,O at 13.4 and 8.5 gauss and 47 Turkevich and Smyth J .Amer. Chem. Xoc. 1940 62 2468. 48 Richards and Smith Trans. Faraday SOC. 1951 47 1261 ; erratum 1952 48 676. 49 Kakiuchi Shono Komatsu and Kigoshi J . Phys. SOC. Japan 1952 '7 108. 6o Smith and Richards Trans. Faraday SOC. 1952 48 307. zc = (12 x 10-13) exp 7000/RT . (22) c&fe6.35 300 QDAR!K'ERLP BEVIEWS H,PtC1,,2H20 has a nearly horizontal inflection stretching from 8 to 13 gauss. The three second moments a t 90" K are 30.4 31.9 and 31.2 gauss2 (each with an error of & 5%). These values are reasonable for a rigid lattice of hydroxonium ions and NO,- ClO,- and [PtCI,- -7 ions respectively. Furthermore from the known crystal structure of HN03,H20,51 the distance apart of neighbouring H,O+ ions can be calculated and so the line shape can be predicted. A value of 2.3 gauss2 being used for the second moment of the Gaussian broadening function the best theoretical fit to the experimental line shape is obtained for a symmetrical H,O+ ion with an interproton distance of 1-72 A.This is illustrated in Fig. 9 where the full line is the calculated FIG. 9 The absorption curwe of the H,O+ ion in HNO,,H,O (reproduced with permission from Trans. Faraday SOC. 1952 48 675). curve and the white and filled circles are two sets of experimental data from different specimens of HNO,,H,O. In HClO,,H,O the temperature variation of the second moment 4 9 shows that the lattice has become rigid from the nuclear resonance point of view at 110" K and a calculation of the line shape a t this temperature reasonable van der Waals radii being assumed for the H,O+ and C10,- ions predicts an interproton distance of 1-70 A.We can conclude that because of the closely-agreeing second moments and the similarity in line shapes (especially at the '' tails " of the curves) the inter- proton distance is probably 1-72 -+ 0.02 ,& in all three compounds. How- ever this analysis requires some qualification. First the calculations have been carried out hitherto for a symmetrical H,O+ ion only. Secondly no reliable O-H bond length can be deduced because it is not possible to predict the HOH angle precisely. The X-ray data have been interpreted in terms of a hydrate lattice,51* but if it is now considered to be (H,O+)NO,- and if 51 Luzzati Acta Cryst. 1951 4 239. * Later calculations of the X-ray data for HNO,,SH,O have indicated the presence A of an H,O+ ion; see Luzzati Acta Cryst.1953 6 167. SMITH NUCLEAR MAGNETIC RESONANCE ABSORP.TION 301 the hydrogen atoms of the H,O+ ion are assumed to lie along the O-..O directions the results give a mean HOH angle of 117" which will agree with the nuclear resonance evidence if the 0-H bond length is 1.01 8 a8 n compared with ro = 0.957 8 and HOH = 105" 3' in water ~ a p o u r . ~ ~ The Line shape for the two-spin system of a water molecule has been calculated and compared with the experimental results for a single crystal of gypsum CaS0,,2H20 at room temperaf~re.~~ As already mentioned the theoretical line shape for a water molecule consists of two peaks symmetrically disposed about a central minimum. In a single crystal the distance apart of the two.peaks will depend on the angle between the line joining the two hydrogen atoms and the Zeeman field H,.The crystal structure of gypsum 53 shows that in any single crystal there are in general two possible values for this angle and therefore for the peak separation. Consequently when a single crystal is rotated in the field so that H can take any direction in the (001) plane the absorption curve consists of two superimposed double-peaked curves with different peak separations and so a t certain orientations shows four peaks. The peaks merge and then resolve again as the direction of H in the (001) plane changes and these variations follow very satisfactorily the theoretical predictions for water molecules with an interproton distance of 1.58 8. This value in conjunction with a Gaussian broadening function of 2-37 gauss2 deduced from the broadening of component lines in the resonance spectrum of the single crystal was used to calculate the line-shape for the powdered crystal and also provided a reasonable fit to the experimental results.As in the case of the H,O+ ion no precise value of the 0-H bond length is obtained because the HOH angle is not known. If the hydrogen atoms of the water groups are assumed to lie along the O...O directions which are hydrogen bonds of length 2-70 A the HOH angle will then be 108" consistent with an 0-H bond length of 0.98 8. A similar analysis of a single crystal of Li2S04,H2054 has given an interproton distance of 1-57 & 0.02 8. Similar double-peaked absorption curves have been obtained 45 from powdered specimens of KF,2H20 MgS0,,7H20 ZnS0,,7H2O and KA1(SO4),,12H2O at room temperature ; the separations of the maxima of the derivative curves all fell between 12.5 and 14.5 gauss.In borax Na2B,0,,10H20 however the absorption curve was narrow with a single peak and a width (AHmsl.) of 4.4 gauss. Ice just below 0" also shows a narrow line about 3.7 gauss wide ; this broadens as the temperature is lowered but the rigid lattice value is not reached until below 230" K. In ice this behaviour would result from the type of motion that has been adduced to explain the residual entropy of ice,55 that is transition of a water molecule from one to another of the six positions it can assume in A The two-spin group. A n 0 52 Herzberg " Infrared and Raman Spectra " p. 489 (Van Nostrand 1945). 53 Wooster 2. Krist. 1936 94 375. 54 Soutif Dreyfus and Ayant Compt.rend. 1951 233 395. 5 5 Pauling J. Amer. Chem. SOC. 1935 57 2680. X 302 QUARTERLY REVIEWS the lattice provided that the frequency of the motion is greater than 3 x lo4 cps. When the frequency drops below this value the line width rises to the rigid-lattice value. A similar explanation presumably holds for borax. Nuclear Magnetic Resonance in More Complicated Molecules.-There are many compounds in which the hydrogen atoms are not all located in one of these two- three- or four-spin groups as for example ammonium di- hydrogen phosphate. In these cases only a roughly quantitative prediction -20 -I5 -10 -5 0 5 10 15 20 G a u ~ FIG. 10 The absorption curve of NH,H,PO (reproduced with permission from J . Chem. Phys. 1950 10,670). of the line shape can be made. In the example quoted above the absorption curve obtained at 87" K ~ * given in Fig.10 is ex- plained most satisfactorily as resulting from the superposition of the broad flat-Dopped curve of a typical NH,+ ion known from NH,Cl and the much sharper-peaked curve of an H2P04- ion known from KH,PO,. If the curve is divided in two as shown by the dotted line the areas of the upper and the lower portion are in the ratio of 0.57 ; since there are half as many hydrogen atoms in the acidic phosphate groups as t'here are in the NH,+ ion the theoretical value should be 0.50. Because it is impossible to predict a very precise value of the intermolecular broadening without knowing something be- forehand about the position of the hydrogen atoms i.e. the length of the O-H bond whether or not the H atom lies on the O-..O bond directions and so on no quantitative coniparison of the theoretical and experi- mental absorption curves can be made and this general drawback of the method becomes more serious as the number of chemically distinct hydrogen atoms in the molecule increases.A number of hydrates also fall in this category. Two in particular are oxalic acid dihydrate for which there are the two possible structures H,C204,2H,0 and (H,O +),C,O,- - and potassium pentaborate tetrahydrate for which a hydroxonium structure K+(H30+),(H,B,0,,- - -) has been pr0posed.5~ The nuclear resonance absorption spectrum of oxalic acid dihydrate a t 90" K 48 is quite different in type from that of (H,O+),(PtCl,- -) and has half the second moment of the latter. The absorption curve illustrated in Fig.11 could be obtained by superimposing two separate absorption curves the first that of a water molecule of about the same dimensions as those in gypsum and the second that of a more isolated single hydrogen atom. The single peak of the latter will obliterate the cent,ral minimum of the two-spin curve giving a practically continuous absorption 6 6 ( a ) A b e d and Cruickshank ( b ) Cox Dougill and Jeffrey ( c ) Jeffrey and Parry J . in the press. 5 7 Zachariasen 2. Krist. 1937 98 266. SMITH NUCLEAR MAGNETIC RESONANCE ABSORPTION 303 curve. The results therefore support the hydrate structure H,C,04,2H,0. This is in agreement with the most recent X-ray investigation^,^^ which show that the dimensions and configuration of the oxalic acid molecule in the dihydrate are the same as in the ct-anhydrous form (which must be I I /O 15 15 1 /o I N gauss FIG.11 The derivative and absorption curves of oxalic acid dihydrate (reproduced with permission from Trans. Faraday SOC. 1951 47 1261). H,C,O,) in contrast to the non-planar oxalate ion in ammonium oxalate. A closely similar absorption curve is given by potassium pentaborate t e t r a h ~ d r a t e ~ ~ which supports the structure K+(H4B50,0-),2H,0 for the compound. In contrast a similar analysis of the monohydrates of sulphuric and selenic acid 48 50 shows that the absorption curves and second moments are most satisfactorily explained in terms of hydroxonium compounds so that in the solid state these substances should be formulated as (H,O+)(HSO,-) and (H,O+)(HSeO,-) although in the liquid state they appear to be incompletely ionised.Many compounds in which the hydrogen atoms occur in different groups and in which there is a large intermolecular broadening give absorption lines which show little or no fine structure and cannot be calculated theoretically. In such cases no precise interproton distances can be derived; but the second moment the line width and the spin-lattice relaxation time can furnish useful information on molecular motions in the 304 QUARTERLY REVIEWS crystal. Ideally to make the most complete analysis of the solid which the method offers all three quantities should be measured over a consider- able range of temperature but this has been carried out for comparatively few compounds. The measurement of the second moment involves a com- plete plot of the derivative line and during the experiment the temperature must remain reasonably constant or the RP bridge will drift off balance and the second moment itself may change.Each line-scan has to be repeated a t a number of fixed temperatures in a range from liquid-oxygen or -hydrogen temperatures to room temperature and perhaps higher. At each point the compound must be left for a considerable time to reach equilibrium and in view of the well-known tendency of many solids to show hysteresis in the vicinity of their transition temperatures measurements must usually be taken in the direction of dccrcasing as well as increasing temperature. *The materials must be carefully purified especially if the impurity is liable to give a sharp resonance line at temperatures a t which the compound under investigation gives a broad line.The sharply-peaked curve may then mask the slowly-varying background and will also cause serious errors in T,. It will also be recalled that T is very susceptible to traces of paramagnetic impurity especially when its value approaches the order of an hour or so.l0 Some inconsistencies in the published data may be attributed to neglect in making due allowance for these various factors. A considerable amount of work has been done on the variation of some of the nuclear magnetic properties of various materials with temperature. A fairly complete study of benzene over a wide temperature range has been made ; line-width data 35 show that molecular rotation presumably in the molecular plane begins to affect the line width at 100" K and from the variation of T from 85" K to 230" B an energy-barrier for the rotation of 3800 cal.can be derived.58 The results show that only a fraction lo-' of the benzene molecules need be rotating in order to affect the line width. Many compounds have been investigated in order to discover any relation- ship between line-width transitions and second-order transitions and what light the two processes may throw on each other. It has been previously emphasised that the two may be quite independent of each other. For example this is so in methane for which recent correcting some earlier results,60 has shown that there is no change in the line width at the specific-heat anomaly at 20-4" K although there is an abrupt change in TI. On the other hand in Polythene 61 the line width changes over the range - 60" to - 20" c and the density-temperature relationship also shows signs of a transition point a t - 40" c.62 Below - 60" c the second moment is consistent with a zig-zag Polythene molecule in a rigid lattice and both transition points then locate the region in which random motion of the polymer segments (involving either rotation or rotational oscillation) begins 58Dr.E. R. Andrew and R. G. Eades personal communication. 59 Thomas Alpcrt and Torrey J. Chern. Phys. 1950 18 1511. 6o Alpert Phys. Review 1949 75 398. 61Newman J. Chem. Phys. 1960 18 1303. 02 Hunter and Oakes Trans. Paraday Xoc. 1946 41 49. SMITH NUCLEAR MAGNETIC RESONANCE ABSORPTION 305 to affect the macromolecular properties and the line width. Other polymers including polytetraflu~roethylene,~~ natural rubber,so synthetic rubbers and p0lystyrene,~3 have also been examined and the results have been of some assistance in interpreting the mechanisms involved in transitions.Most of the work discussed in this Review is of recent development and it would be appropriate to conclude by summarising what has been achieved so far. Present experimental methods are sufficiently sensitive to record the nuclear resonance spectra of most nuclei with a magnetic moment in the solid liquid and in some cases the gaseous state. In the solid state the crucial factor determining the sensitivity the signal-to-noise ratio depends very much on the values of T and T, which in turn depend on the freedom of motion allowed to the nuclei in the lattice their distance apart and the concentration of paramagnetic atoms and lattice defects.These are there- fore three subjects which can be studied by the method. From the first we can derive considerable information on the low-frequency motions in certain crystals. From the second we can in the most favourable cases determine bond lengths involving hydrogen atoms as for example in ammonium chloride or at least distinguish between two or more possible structures for a compound as in the hydrates of some inorganic acids. In the liquid state paramagnetic impurities also have a large effect on TI so the method has been used to measure the small surface concentration of paramagnetic atoms on supported oxide catalysts. Probably the most accurate method at present of measuring the strength and inhomogeneity of magnetic fields is to use the resonance line of a suitable liquid in conjunction with a frequency meter and oscilloscope.Recently it has been used to study electric quadrupole effects in the solid state and the fine structure of the liquid resonance line has been shown to depend on the structure of the molecule so that for example the splitting of the fluorine resonance is different in IF5 from that in IF,. Recent experimental work has produced simple and robust apparatus for the detection of the effect and new methods have been developed for in.creasing the sensitivity. The subject is now one of the most important new fields of physics to be developed since the war and for the part they played in its development Professors Bloch and Purcell mere jointly awarded the Nobel prize for physics in 1952. In its chemical applications the future promises to be at least as interesting as the past seven years.The R'eviewer thanks the Department of Scientific and Industrial Research for a research grant during the tenure of which this Review was written. He is also indebted to Professor E. G. Cox and Professor F. S. Dainton for helpful criticism of the manuscript and to the many authors and editors of journals who have given permission to reproduce diagrams from their articles. APPENDIX The Reviewer has conformed to the generally accepted usage of symbols in this field despite the unfortunate clash with other important quantities. 63 Holroyd Codrington Mrowca and Guth J . Appl. Phys. 1951 22 696. The scope and usefulness of the method are still increasing. 306 QUARTERLY REVIEWS I n particular the capital letter T has been used with the subscripts 1 and 2 to represent the spin-lattice and spin-spin relaxation times.Magneiic field strength has been defined in units of the gauss. The other common symbols are given below in the order in which they occur in the text. The magnetic moment in absolute units. The so-called “ magnetic moment ” in the literature is generally the dimensionless number p = gI. The nuclear spin quantum number (dimensionless). The nuclear g-factor or the ratio of the nuclear magnetic moment to the nuclear angular momentum (dimensionless). The nuclear magneton eh/4nMc where e is the electronic charge h Planck’s constant M the mass of the proton and c the velocity of light. The Bohr magneton eh/4nmc where m is the mass of the electron. The applied magnetic field sometimes called the Zeeman field in Radiofrequency .Cycles per second 1 Me. = lo6 cycles per second. Frequency in cps. Angular velocity. The applied R’F field. The complex magnetic mass susceptibility equal to x’ - i ~ ’ ‘ (dimen- The real (in-phase dispersive or high-frequency) component of x. The imaginary (out-of-phase or absorptive) component of x. The static mass susceptibility. The separation (in gauss) of the two maxima of the derivative curve i.e. the two points of maximum slope positive or negative of the absorption curve. the text. sionless). The spin-lattice relaxation time. The spin-spin relaxation time. The gyromagnetic ratio ge/2Mc or the ratio of the angular Larmor frequency to the applied magnetic field. (Acu’)A,. The mean-square frequency displacement of the absorption linc. (AH2)Av. The second moment of the absorption line. F(H) The line-shape function a t any value of the applied field 11. H* The resonance field strength a t a fixed frequency of an isolated S(H-Ho) The Gaussian broadening function of form exp -(H-Ho)2,/2@2 N The total number of resonating nuclei in the sub-group to which the rjk The internuclear distance between the j t h and the lcth nuclens. Z C The correlation time. z The Debye relaxation time for dielectric relaxation. nucleus. where ,b’ is the second moment of the distribution. line broadening is attributed.

ISSN:0009-2681    

DOI:10.1039/QR9530700279

出版商:RSC    

年代:1953

数据来源: RSC

摘要:

INORGANIC CHROMATOGRAPHY By R. A. WELLS B.Sc. A.R.I.C. (PRINCIPAL SCIENTIFIC OFFICER CHEMICAL RESEARCH LABORATORY TEDDINGTON) CHROMATOGRAPHY has been simply defined as '' those processes which allow the resolution of solute mixtures by selective fixation and liberation on a solid surface or support with the aid of a fluid streaming in a definite direction ".I A great variety of chromatographic systems is possible ; thus many different materials have been used as the solid surface or sup- port usually called the adsorbent while the fluids or solvents employed include aqueous solutions organic solvents and even gases. An exact classification of all the variations is not possible but an arbitrary division into four groups may be made according to the mechanism by which the solute is fixed or adsorbed on the adsorbent.The classification has to be arbitrary since many solutes are retained on an adsorbent by more than one process. The four types of adsorbent which will now be considered are ( a ) Surface-active adsorbents which include such materials as activated alumina. ( b ) Bound or supported liquids; these give rise to partition chromatography the most popular form being paper chromatography where the supported liquid is water in the cellulose and the solute undergoes partition between this phase and the solvent. ( c ) Chemically reactive solids ; these give rise to adsorption by means of direct chemical combination with the solute to form recognised compounds. ( d ) Ion-exchange materials ; these may be considered as particular examples of chemically active solids.To whom should go the honour of having performed the first chroma- tographic separation is still a matter of discussion; there is however no doubt that the first major publication on the subject was by F. Goppels- roeder a student of C. P. Schonbein,2 who in 1861 noticed on dipping filter- paper into an aqueous solution of an electrolyte that the water rose to a greater height than did the dissolved salt. Moreover he found that dif- ferent solutes rose to different heights up the filter-paper. These observa- tions formed the basis of many years of work by Goppelsroeder who studied the behaviour of organic compounds under similar conditions and ultimately published his results in Dresden in 1910 in a book entitled "Kapillar- analyse ". As may be gathered from this title Goppelsroeder attributed his separations to capillary forces acting in the paper.That this is not the full story has been explained by I?. Feigl who has elaborated and developed the ideas of Schonbein and Goppelsroeder into the well-known technique of " Spot Tests ".3 Adsorption of the solute by the filter-paper does play an importlant p r t in inany separations of this type. L. Zechmeister " Progress in Chromatography 1938-1947 " Chapman & Hall London 1950. Pogg. Ann. 1861 114 276. " Chemistry of Spec& Selective and Sensitive Reactions," English transl. R. E. Oesper Academic Press Inc. New York 1949. 307 308 QUARTERLY REVIEWS Similarly to whom to attribute the first chromatographic separation by a truly adsorptive process is still a matter of argument but here again there is little doubt that the first major contribution came from the botanist M.Tswett.4 The technique by which he was able to separate the compo- nents of chlorophyll and the isomers of carotene has been described fully many times and will not be further dealt with here. A jump of some 37 years is necessary Do arrive a t the first application of Tswett’s technique t o inorganic separations. that when a solution of metal salts of suitable pH was filtered through a column of alumina the metals were precipitated in a series of sharply- defined zones and always in a definite order. The reader is referred to any one of a number of text books I 6 ‘ 9 * for a full description of these early experiments together with details of the first chromatographic separation of inorganic anions. The growth of inorganic chromatography has been characterised by a tendency to lag behind developments in organic chromatography.Partition chromatography was devised initially for the separation of amino-acids 0 and it was used in the separation of a number of other organic materials before it was applied to inorganic separations. Ion-exchange materials were initially prepared for the exchange of inorganic ions but the number of applications of ion-exchange chromatography to organic separations probably equals if not exceeds the applications to inorganic separations. One field is however peculiar to inorganic separations. It was introduced by H. Erlenmeyer and H. Dahn,ll who discovered the possibility of separating inorganic mixtures on columns prepared from organic precipitation reagents such as hydroxyquiiioline.This appears to have no application so far in organic separations. There is now evidence of a growing interest in inorganic chromatography and a realisation of its possibilities so it is to be hoped that the lead long held by organic chromatography may be somewhat reduced It was shown by G. M. Schwab and K. Jockers Surface Active Adsorbents The study of the separation of inorganic ions on alumina initiated by G. M. Schwab and his co-workers was continued over a number of years.12 It was shown that the length of an adsorption zone was proportional to the “ Chromophylls in the Plant and Animal World ” (Warsaw 1906) Academic Press New York 1949. Angew. Clzem. 1937 50 546. L. Zechmeister “ Principles and Practice of Chromatography ” English hransl. A. L. Bacharach and F.A. Robinson Chapman & Hall London 1941. T. I. Williams “ An Introduction to Chromatography ” Blackie PL Son London 1946. 8 H. Cassidy “ Adsorption and Chromatography ” Interscience Publishers Now York and London 1951. H. H. Strain “ Chromatographic Adsorption Analysis ” Interscience Publishers New York and London 1941. lo A. J. P. Martin and R. L. M. Synge Biochem. J . 1941 35 1358. l1 Helv. Chim. Acta 1939 22 1369. l2 G. M. Schwab and G. Dattler Angew. Chem. 1937 50 691 ; 1938 51 709 ; G. M. Schwab and A. N. Ghosh ibid. 1939 52 666; 1940 53 39. WELLS INORGANIC CHROMATOGRAPHY 309 concentration of the material present and thus it is possible to carry out an approximately quantitative estimation. By using columns of 1-2 mm. diameter and specific spot-test reagents micro-amounts of an ion present in a large volume of solution can be detected.The ions present are con- centrated by the adsorption process into a narrow zone and with the aid of a suitable reagent a sensitivity of detection equal to that of normal spot tests is obtained. The application of alumina chromatography to the general analysis of an unknown mixture is difficult because of the lack of separation between several pairs of metals. It is however possible to separate the individual metals of an analytical group after Separation into groups by the conventional inet1i0ds.l~ More recently the separation of the platinum metals on aluniina has been investigated.14 It is claimed that extremely sharp separations can be obtained thc order of increasing adsorp- tion being Rh Pd Pt Ru and Ir but certain conditions with regard to the age of the test solution and absence of interfering metals must be ful- filled.Thus the platinum group can be separated from Pb Cu Zn and Ni but in the presence of iron mixed complex formation gives rise to complicated chromatograms. The separation of organometallic complexes on columns of alumina has been demonstrated. It has been claimed l5 that one part of cobalt in 2 x lo9 parts can bz detected by the addition of a chloroform solution of the complex of the nietal with /3-nit'roso-cc-naphthol to an alumina column. After being washed with a small amount of alcohol the cobalt complex is visible as a red band a t the top of the column. The cobalt complex with Nitroso-R-salt has been utilised in the determination of cobalt in carbon steels.16 A solution of the salt was added to a suitably buffered solution of the steel.After addition of perchloric acid the solution was passed through a column of alumina which had been pretreated by washing with M-per- chloric acid. The excess of reagent and iron salts was washed through the column with hot nitric acid. The cobalt complex was finally eluted with M-sulphuric acid. A number of mechanisms have been proposed to explain the separation of inorganic ions on alumina. Technical alumina invariably contains sodium aluminate as an impurity and adsorption has been supposed to be due to cation exchange between the sodium and adsorbed ion.12 Again technical alumina has an alkaline reaction and it has been suggested that precipitation of basic salts occurs.12 Both carbonates and bicarbonates are normally present in technical alumina and precipitation of basic carbonates may take p1ace.l' Although all three of these reactions probably occur when technical alumina is used they are not essential since separations are equally possible on sodium-free alumina.The possibility of exchange betwecn cations and aluminium ions has been proposed,12 but since the l3 H. H. Fillinger and L. A. Trafton J . Chein. Educ. 1952 29 285. l4 G. M. Schwab Discuss. Fnruday Soc. 1949 7 170. l5 R. 0. Bach and A. A. Garmcndia Anal. Asoc. Quim. Argentina 1951 39 11. l6 J. A. Dean Anal. Chem. 1951 23 1096. l7 G. Siewert and A. Jungnickel Ber. 1943 76 210. 310 QUARTERLY REVIEWS position of aluminium adsorption in the chromatographic series has been located between lead and cadmium this does not explain adsorption of ions lower in the list.Other mechanisms suggested include aniphoteric adsorp- tion by alumina giving rise to both anionic and cationic exchange; pre- ferential adsorption of ions which then bind the ion of opposite charge by equivalent secondary adsorption and molecular adsorption. l8 Considerable practical evidence has been given in support of a theory of hydrolytic adsorption.lg This suggests that the hydrolysis of a metallic salt according t o the equation M" + H,O + M(OH)' + H is encouraged by the removal of hydrogen ions by aluminium-ion- hydrogen-ion exchange thus Al,O + GH' f 2A1"' + 3H,O The liberated aluminium salt is then adsorbed as a basic salt 2A1"' + 2H,O f 2Al(OH)" + 2H' This theory explains the observed separations and with a few exceptions the order of adsorption of metals is the same as the order for the degree of hydro- lysis of the metallic salts.The order of adsorption is also the same as the order of ionic potential values the reciprocal of dialysis coefficient values and the reciprocal of diffusion coefficients; moreover the most highly hydrated cations are the most strongly adsorbed. This agreement is explained by the fact that co-ordinated water molecules in an aquo-complex are orientated with the hydrogen atoms directed outwards. The opposite type of polarisation occurs with water adsorbed on alumina and the hydrogen atoms are directed towards the oxygen molecules of the adsorbent. Thus the suitably polarised water molecules of the aquo-complexes will be readily adsorbed by the alumina. Further the degree of polarisation of the mole- cules will increase with the ionic potential which is in agreement with the correlation between ionic potential values and degree of adsorption.I n addition to work on columns of alumina separations have been carried out on paper impregnated with alumina.20$ 21 The paper is prepared by soaking it in sodium aluminate solution allowing it to dry and then soaking it in sodium hydrogen carbonate solution. The paper is finally washed for several days in distilled water and then allowed to dry. About 0-1 ml. of test solution is sucked up to a height of 1-3 em. on a strip of paper which is then treated with water until the water rises to a height of about 10 em. The same type of separation as is obtained on columns of alumina occurs and the order of adsorption is identical.If the paper is pretreated with dilute nitric acid separation of anions may be obtained as with columns of alumina similarly treated. The zone length is directly proportional t o the concentration but diffcrent ions give different zone lengths. In general for ions of the same vdency zone length increases zOH. Flood ibid. p. 190. Since the alumina is in excess the process goes to completion. 1* P. W. M. Jacobs and F. C. Tompkins Trans. Faraday SOC. 1945,41,388 395,400. l9 L. Sacconi Discuss. Faraday Soc. 1949 7 173. 21 Y. Oka and A. Murata Sci. Rep. Res. Inst. Tdhoku Univ. 1961 A 3 82. WELLS INORGANIC CHROMATOGRAPHY 311 with decreasing adsorption properties. Increase in valency results in increasing zone lengths and the zone length is to some extent affected by the anion present.A disadvantage of the separations on alumina is that sepa- ration is only into zones. In most cases there is some overlapping of zones and even in the best separations the zones remain in contact. It is possible to obtain clear cut and absolute separations by addition of complexing agents to a mixture of cations. The undissociated molecules are not ad- sorbed on alumina with the result that addition of a complexing agent to a mixture of cations will permit separations depending upon the relative stability of the complexes formed. Thus a solution of glycine added to a copper solution will result in a mixture of unadsorbed copper glycinate molccules and adsorbable copper ions giving two bands in the chromato- gram. If the amount of glycine is sufficient to form a complex [CuGl,] with the whole of the copper none of the copper is adsorbed.This may be used for the determination of copper by a method known as a chromato- graphic titration. To aliquots of the unknown copper solution are added known amounts of glycine and the solutions are chromatographed. By extrapolation of the adsorbed copper zones to zero length the equivalent amount of glycine and hence the unknown copper concentration may be calculated. If two metals are present in solution and the stability constants of their complexes with glycine differ sufficiently complete separation of the two metals may be achieved. Thus on chromatographing a mixed solution of nickel and cobalt in the presence of a suitable concentration of glycine two completely separated zones are obtained one of cobalt which has moved little from its original position and another of nickel which has migrated some way along the paper.Another modification of Schwab's technique due to J. E. Meinhard and N. P. uses a thin film of alumina mixed with an adhesive on a micro- scope slide as adsorbent. Spots of test solution are applied to the film and the chromatogram of concentric rings is examined under the microscope. Other metallic oxidcs have been used as adsorbents including chromic oxide,23 which was claimed to give better results than alumina but has not proved a popular adsorbent. Zinc sulphide has been used as an adsorbent 2 4 for the separation of copper and cadmium. Since this separation is almost certainly dependent upon the formatiox and relat'ive stability of the sulphides of the two metals it ought perhaps to be included with the separations described in the next section carried out on chemically active solids.Partial separation of the isotopes of neon has been achieved by chroma- tographic separation on charcoal at the temperature of liquid nitrogen.25 The work was undertaken in order to test equations derived for the behaviour of two solutes (which gave non-linear isotherms) in the region of the frontal interboundary . Best results were obtained by immersing the charcoal 22 Anal. C%em. 1949 21 186. 2 3 H. Goto and Y. Kakita J . Chem. Xoc. Japun 1942 63 120. 2 4 J. M. Bach Anal. Asoc. Quim. Argentina 1937 183 55. 25 E. Glueckauf K. H. Barker and G. P. Kitt Dzscuss. Paraday SOC. 1949 7 199. 312 QUARTERLY REVIEWS column in liquid nitrogen and then filling it with helium.Neon was then passed through t'he column and replaced the less-adsorbed helium and advanced with a self-sharpening front boundary. Enrichment factors of up to 15 were obtained on quantities of the order of a milligram but i t is not suggested that the method could be readily used on a larger scale. The result of varying experimental conditions could be forecast by consideration of the theoretical equations and theoretical and practical values for " theoretical plate " height were in close agreement. Chemically Reactive Solids A type of chromatography which so far appears to have been confined to the inorganic field was introduced by Erlenmeyer and Dahn in 1939.11 They employed as adsorbent " oxine " (8-hydroxyquinoline) diluted with a t least its own weight of an inert material such as starch or Kieselguhr.On adding a solution of cations of preadjusted pH to a column of the adsorbent and washing wit'h a buffer solution separation of the mixture into a number of zones occurred. Each zone had the distinctive colour of the hydroxy- quinoline complex of the metal concerned. The order of adsorption pro- ceeding down the column has been given by G . Robinson 26 as V0,- Mn++ WO,- Ag+ Cut-+ Bi+++ Ni++ MOO,- Alfff Co++ Zn++ Mg++ Fe+++ UO,++. The same worker has shown that the technique may be applied to quantitative analysis. If carefully controlled conditions particularly with'regard to pH of the test and wash solutions are used the length of zone produced is reproducible and proportional to the concen- tration of metal present.The method has been used for the analysis of both zinc and silver brazing alloys. When dealing with the first an upper green zone due to copper was obtained below which were two yellow zones the top one due to nickel and the lower one to zinc. The colours of the lower zones were sufficiently alike to make measurement of the length of the zinc zone impossible but advantage was taken of the fact that the zinc complex gives a bright green fluorescence in ultra-violet light. With the silver alloy tested an upper yellow zone due to silver was obtained separated from a yellow zinc zone by a green zone due to copper. Far less satisfactory results were obtained when an attempt was made to apply the method to the analysis of ferrous alloys. The large concentration of iron led to the formation of the intensely black ferric complex.Efforts to wash the complex through the column led to a spreading of the colour throughout the column thus masking the presence of other cations. On using a slightly acid (pH 3) wash solution a much speedier movement of the ferric complex was obtained but the wash solution dissolved sufficient of the oxine to ruin the zones due to other metals. This solubility of the adsorbent in slightly acid solutions is a limiting factor which prevents a much wider application of the technique. A clear picture has not been given of the mechanism by which zoning of cations in these columns occurs. It may be due to a combination of two fact'ors difference in affinity of oxine for oxine salts and difference of solubility of the oxine salts in water. It is significant that the order of z 6 E.Glueckauf I<. H. Barker and G. P. Kitt Discuss. Paraday Soc. 1949 7 195. WELLS INORGANIC CHROMATOGRAPHY 313 adsorption is very similar to the order of stability of the oxine complexes in acid solution. Separations on columns of oxine are confined to those met'als which form insoluble oxine complexes (oxinates) but other organic reagents have been employed as adsorbents. Sodium potassium magnesium and ammonium ions have been separated on columns of violuric acid.27 All give violet to brick-red colours with the reagent. A compound column containing 4-hydroxyimino-5-oxo-3-phenylisooxazoline in the upper half and violuric acid in the lower half has been employed for the quantitative separation of sodium and potassium.2* With the aid of a column of dimethylglyoxime it is possible to detect traces of nickel in the presence of large amounts of cobalt iron and even palladium.The nickel is retained as a red complex at the top of the column while the other metals are washed through. These separations on columns of reactive chemicals suffer from the same major disadvantage as a separation on alumina namely that they are not complete. Even in the best cases where clearly defined zones are obtained the zones are in contact with each other. This disadvantage was overcome by P. P. H ~ p f ~ ~ who used as adsorbent filter paper which was impregnated first with starch or peptised alumina and then with an organic reagent. Thus he was able to show that on placing a drop of an aqueous solution containing manganese and iron on a piece of paper impregnated with alumina and formaldoxime hydrochloride complete separation took piace.On making the paper alkaline with ammonia vapour a central spot due to manganese was observed separated by a blank zone from a coloured ring due to iron. Using a formula suggested by Flood Hopf was able to apply the method to quantitative analysis. Flood had shown that the radius of a spot formed by a solute was a function of its concentration plus a constant for the paper from which Hopf deduced the expression a - r I 2 - r 2 b r32 - r42 _ _ _ _ _ _ Kab for a binary mixture. This was derived from the equations a = n(rI2 - rZ2)Ka and b = n(r,2 - r,2)Kb Kab being put for K,/Kb where a = concentration of first substance and rl and r2 are bounding radii of the annular zone which it forms b = concen- tration of second substance and r3 and r4 are bounding radii of the annular zone which it forms and K and Kb are constants which may be determined by an experiment on a solution of known concentration.The method was also utilised in the analysis of ferromolybdenum alloys by using paper impregnated with alumina and 8-hydroxyquinoline and for the separation of ammonium and potassium salts by using paper impregnated with starch and violuric acid. 27 H. Erlenmeyer and W. Schroenauer Helv. Chim. Acta 1941 24 878. 28 H. Erlenmeyer and J. Schmidlin ibfid. p. 1213. 29 F. Burriel Marti and F. Pino-Perez Anal. Chim. Acta 1949 3 468. 30 J. 1946 786. . 314 QUARTERLY REVIEWS Partition Chromatography The number of separations of inorganic materials achieved by chroma- tographic means has been greatly increased by the application of partition chromatographic techniques to mixtures of inorganic salts.In the organic field partition separations have been carried out with a number of adsorbents such as cellulose powdered rubber silica gel and Celite but for inorganic substances attention has been almost totally confined to cellulose The reader is referred t o a recent booklet for an excellent review of the use of cellulose in both organic and inorganic chromatography. 31 The cellulose has been used both in sheet form and as a powder and the solvents employed have been orga'nic liquids usually containing water and oft'en a mineral acid. To improve separations various compounds capable of coniplexing inorganic salts have been added to the solvents. In their preliminary experiments Martin and his co-workers attempted to separate the acetyl derivatives of the amino-acids by partition chroma- tography on silica gel.lo7 32 This was only partly successful and they later introduced a technique employing cellulose as the adsorbent.33 In this method the aqueous test solution is applied to a small area near one end of a ship of filter-paper and allowed to dry.The strip is then suspended vertically with the end nearest the test spot dipping into a solvent container. Both solvent and strip are enclosed by a sealed jar in order to maintain constant atmospheric conditions. The solvent diffuses down the strip passes through the spot of test solution and carries with it at differential rates the inorganic materials present. After the solvent has moved suf- ficiently far to achieve a separation of the mixture the position of the cations or anions on the strip may be found by spraying the strip with a suitable spot-test reagent A modification of this system originated by Williams and K i r b ~ 3 ~ has also proved popular with a number of investi- gators.For this technique a sheet of filter-paper is rolled to form a cylinder which can be stood on end in a solvent container. The t'est solution is applied as spots a few em. up from the lower end of the cylinder. Again the whole apparatus is enclosed by some suitable vessel to maintain constant atmospheric conditions. In this case diffusion of the solvent is upward and separation into a series of spots takes place. A third technique due to L. Rutter,36 has been used in which a circular disc of filter-paper is employed.Two parallel cuts about 2 mm. apart are made from the circumference to the centre of a disc of filter-paper. The strip of paper thus formed is bent down at right angles to the disc of paper thus forming a " tail ". The test solution is applied as a spot to the centre of the filter-paper which is then placed in a horizontal position supported by the edge of t>he bottom half of a Petri dish. The " tail " is allowed to dip into a solvent receptacle contained 31 J. N. Balston and B. E. Talbot " A Guide to Filter Paper and Collulosa Powder 32 A. H. Gordon A. J. P. Martin and R. L. M. Synge Biochem. J. 1943 37 79. 33 R. Consden A. H. Gordon A. J. P. Martin and R. L. hl. Syngc ibid. 1944,38,221. 3 4 R. J. Williams and H. Kirby Science 1948 107 481.35 Nature 1948 161 435; Analyst 1950 75 37. Chromatography" H. Reeve Angel & W. & R. Balston Lt8d. 1952. IT7ELLS INORGANIC CHROMATOGRAPHY 31 5 by the Petri dish and paper and dish are covered by the second half of the dish. The solvent rises up to the tail of filter-paper and spreads radially through the test spot. A chromatogram is formed in which the inorganic ions are found as a series of concentrfc annuli. Use of the upward or down- ward diffusion method seems largely a matter of personal choice for the operator. Rutter's technique is employed when a quick qualitative result is required. These methods of separation using sheets of filter-paper are limited in the amount of material which can be dealt with in any one experiment. The maximum amount of material normally dealt with is of the order of 1 mg.although a few cases have been described where by virtue of the special properties of the material being dealt with or by using wide strips up to 0.15 g. of metal has been employed. By using a column of cellulose pulp there is theoretically no reason why the amounts of material separated should have any limitations since it is quite practicable to use very large columns. Workers a t the Chemical Research Laboratory have in fact carried out separations on up to $ kg. amounts of metal. In using a cellulose column 369 37 the cellulose is packed in pulp form into a glass tube cont'aining the organic solvent mixture to be used during the separation. The aqueous solution containing the material solution is transferred to the top of' the column and the solvent passed through.In order $0 avoid clogging of the column and undesirable wall effects the test solution is often transferred to the top of the column by first absorbing it on a wad of cellulose and then transferring the mixture. The separated metals are collected in suitable fractions in the eluate solvent solution. I n such experiments concentration of excess of water at the top of the column is likely to be high a t the begin- ning of the run. To prevent mechanical carry down of the water between the cellulose and glass walls of the tube it has bceii foiind advantagcous to treat the inner surface of the glass tube with a water repellant such as dimethyldichlorosilane or other suitable silane. Use of cellulose columns has a number of advantages over the use of more classical adsorbents.The metal is obtained in eluate solvent solution ; it is not necessary t o extrude the adsorbent and wash the metal from a particular zone. Any residual metallic salts in the column can normally be removed by washing through with water or dilute acid. In the few cases of particularly strong adsorption complete recovery can be effected by ashing the cellulose. Apart from the technique and apparatus used the approach t o inorganic partition chromatography has varied with different workers. One group have concentrated on separating the metals of previously isolated groups of metals another has concerned itself with the separation of a large number of metals in an unknown mixture. A further type of separation is that of one metal from a very large number of others. A few inorganic salts have exceptionally high solubilities in organic solvents and advantage has been taken of these peculiarities to produce highly specific separations.Thus uranium has been tested for in solutions of minerals in nitric acid by using 3 6 F. H. Burstall G. It. Davies and R. A. Wells Discuss. Faruduy Xoc. 1949,7 178. 37 F. H. Burstall and R. A. Wells Analyst 1951 76 396. 316 QUARTERLY REVIEWS a paper-strip technique. 38 The most satisfactory results were obtained with 2-methyltetrahydrofuran as solvent but good separations were obtained with the simple ester ethyl acetate. In both cases water and nitric acid were added to the solvent and uranium was extracted in a narrow band in the solvent front. Most other metals remained in the original test-spot or moved only slightly.By cutting out the portion of the strip containing the uranium ashing and determining the uranium in the ash the separation was shown to be quantitative 39 and the method has been used for the deter- mination of uranium in minerals and ores. Uranium has also been extracted quantitatively on a larger scale from a variety of minerals by using a cellu- lose column with ethyl ether containing nitric acid as solvent.37 Mercuric chloride is extremely soluble in a number of organic solvents but unlike other inorganic salts its solubility decreases as the concentration of acid added to the solvent increases. By using as solvent a mixture of methyl acetate and water mercuric chloride has been separated from a wide range of metals many of which are themselves extracted by the same solvent in the presence of hydrochloric acid both in strip 4 0 and column e~periments.3~ By using butyl alcohol saturated with N-hydrochloric acid thallic chloride has been separated quantitatively 41 from a number of metals.Here advan- tage was taken of the fact that the thallium salt moved in the solvent front by tapering the paper strip and collecting the first few drops of solvent to drip from the end. The metallic salt was thus isolated in a small volume of solvent. A similar technique has been adopted for the quantitative separa- tion of gold by a paper-strip method.42 Auric chloride is very soluble in a number of organic solvents but there is a strong tendency for it to be reduced to the metal particularly in the presence of an organic solvent and cellulose. This has been prevented by the use of ethyl acetate containing nitric acid as solvent.The test solution was prepared in dilute hydrochloric acid and applied to a paper strip. The gold chloride was extracted in a narrow band in the solvent front and providing the solvent was saturated with water separation was achieved from all metals with the exception of uranium antimony and tin. The solubility of ferric chloride in organic solvents is well known and various solvents may be used for the paper-strip separation of iron from other metals. Mixtures of ketones with hydrochloric acid extract ferric chloride particularly efficiently. With simple mixtures of acetone containing hydrochloric acid very sharp separations from titanium nickel lead aluminium manganese cobalt etc. have been obtained. With higher ketones the list of metals from which separation is possible grows rapidly.The extreme insolubility of some inorganic salts in organic solvents malies it possible to carry out the opposite type of separation from that just des- cribed. By choice of a suitably polar solvent a large number of metals 38 T. V. Arden F. H. Burstall and R. P. Linstead J. 1949 S 311. 39 J. A. Lewis and J. M. Griffiths Analyst 1951 76 388. 40 F. H. Burstall G. R. Davies R. P. Linstead and R. A. Wells J. 1950 516. 41 J. R. A. Anderson and M. Lederer Anal. Chim. Acta 1950 4 513. 4 2 N. F. Kember and R. A. Wells Analyst 1951 76 579. WELLS ITSORGANIC CHROMATOGRAPHY 31 7 may be extracted leaving the required insoluble salt still in the original position of the test-spot or at least near the top of the strip.Thus many metals may be separated from lead as their chlorides with methyl alcohol containing dilute hydrochloric acid. Similarly the extreme insolubility of nickel aluminium and titanium salts in a number of organic solvent mix- tures has permitted the ready extraction of other metals from them. Advan- tage of the insolubility of nickel salts has been taken in the determination of impurities in nickel-plating baths.43 An aliquot of the plating solution was extracted in a column of cellulose with acetone containing hydrochloric acid. Six metals iron manganese cobalt copper zinc and cadmium were quantitatively recovered and estimated in the eluate solution. This separa- tion is particularly interesting in that the test solution was in sulphuric acid but the extracted metals were eluted as their chlorides.The mass of nickel sulphate was slowly converted into the chloride by the hydrochloric acid in the solvent and free sulphuric acid appeared in the eluate. Considerable success has been achieved in the separation of groups of metals and the metals of each of the groups of the routine qualitative analysis tables have been separated on paper strip~.~O In some cases the study of the separation of one particular group has been extended to include metals not present in the group. The metals of Group IIA lead copper bismuth cadmium and mercury have been separated as their chlorides with butyl alcohol saturated with 3~-hydrochloric acid as solvent and the behaviour of the additional metals vanadium uranium iron molybdenum and antimony with the same solvent has been studied.40 It was found that vanadium moved with copper uranium with lead antimony with bismuth and molybdenum moved very close to bismuth but a clear separation of iron between copper and bismuth was obtained.A number of naturally occurring groups or pairs of metals are separated only with difficulty by normal methods. Many of these have been separated by paper-strip technique. Thus aluminium indium gallium and zinc 44 have been readily separated as their chlorides with butyl alcohol containing 20% of concen- trated hydrochloric acid. The platinum metals Pt and Pd together with Au Ag and Cu have been separated with butyl alcohol saturated with N-hydrochloric acid 45 and both methyl propyl ketone and ethyl methyl ketone containing concentrated hydrochloric acid have been used in the separation of Au Pt Pd Ir and Rh mixtures.40 By suitable choice of solvent and acid radical the members of the following pairs have been separated aluminium and beryllium,46 zirconium and hafnium,49 and selenium and tellurium.40 Scandium and thorium have been separated from each other and from the rare earths.40 Separation of the alkali metals and alkaline earths on paper strips has been studied by a number of Although separations have been achieved location of the 4 7 9 4 8 4 3 F.H. Burstall N. F. Kember and R. A. Wells J . Electrodepositors Teclz. Soc. 4 4 2'. V. Arden F. H. Burstall G. R. Davies J. -4. Lewis and R. P. Linstead 4 6 G. H. Osborn and A. Jewsbury ibicl. 1949 164 443. 1951 27 261. Sature 1948 162 691. 4 5 M. Lederer ibid. p. TiG. Y 31 8 QUARTERLY REVIEWS metals on the strips has given some difficulty.Violuric acid has been used to test for Li Na K Be Mg Ca Sr and Ba after separation of the acetates with ethyl alcohol containing acetic acid. 47 More recently thiovioluric acid has been introduced 48 as a more sensitive reagent than violuric acid. The metallic salts fluoresce strongly and since in this particular separation the areas of the spots vary linearly with the quantity of ion present quan- titative estimations of the cations has been possible. There are a number of disadvantages in using a fluoride system for paper-strip separations but with metals such as niobium and tantalum the advantages of a fluoride solution are obvious. By using a Polythene appara- tus it has proved possible to separate niobium and tantalum from a number of metals and from each other on paper strips ethyl methyl ketone containing hydrofluoric acid being used as solvent.The R value for an element defined as the ratio of the distance moved by the element to that moved by the solvent front conveniently describes its behaviour. Reasonably constant R values are obtained if sufficient precautions are taken to standardise conditions. The R values of some 40 elements have been determined by Lede~-er,~O using the solvents ethyl alcohol plus 10% v/v ~ N - H C ~ isopropyl alcohol plus 10% v/v ~N-HC~ butyl alcohol saturated with N-HC1 and amyl alcohol saturated with ~N-HC~. These form a useful list which can be consulted to ascertain whether the metals present in a given mixture can be separated with a particular solvent. With few exceptions the R values for a metal decrease as the molecular weight of the alcohol used increases.With mixed alcohol solvents the R values obtained were less than the mean value for the individual alcohols. The R values for a large number of metals in butyl alcohol saturated with N-hydrochloric acid have been compared 51 with values obtained with butyl alcohol containing hydrogen bromide and with butyl alcohol-nitric acid mixtures. These results emphasise the importance of the nature of the acid radical in most of these separations. On using the nitrate-containing solvent uranium was the- only metal of those tested which showed appreciable movement. With a solvent containing either hydrochloric or hydrobroniic acid appreciable movement was observed with all those metals which readily form chloride or bromide complexes.E. C. Martin 52 has measured R values for a number of metals with butanol containing thio- cyanic acid using Rutter's technique. Others from the same laboratory 53 have measured R values for 22 metals using diethyl ether saturated with N-HC~ N-HBr N-H,SO, or N-HNO as solvents. The only metals to give significant movement with all four solvents were arsenic antimony and tin. Workers a t Bristol University have made considerable progress in working out a separation scheme which can be applied directly to any unknown 4' H. Erlenmeyer H. von Hahn and E. Sorkin Helc. China. Acta 1951 34 1419. 48 H. Seiler E. Sorkin and H. Erlenmeyer ibid. 1952 35 120. 49 Chemical Research Laboratory unpublished work. 50 Anal. Chim. Acta 1951 5 185. 51 M. Lederer ibid.1950 4 629. 52 E. C. Martin ibid. 1951 5 511. 53 J. R. A. Anderson and A Whitley ibid. 1952 6 517. WELLS INORGANIC CHROMATOGRAPHY 319 mixtures of metals.543 55 They have preferred to use nitrates which are relatively insoluble in organic solvents and induce solubility by the addition of strong complexing agents such as benzoylacetone. Two different proce- dures have been described for the identification of an unknown mixture. In the first three separate chromatograms are run using the following mixed solvents butanol-water-benzoylacetone collidine-water and dioxan- antipyrine. The metals are identified by their R values and by their characteristic reaction to spot tests. I n some cases two-dimensional chromatography is employed. In this process one solvent is run normally down a sheet of paper on which the test solution has been placed in an upper corner.The paper is then turned through go" and after the solvent has been allowed to dry off a second solvent is allowed to diffuse down the paper. In a few cases the process is repeated with a third solvent. In the second separation procedure a series of chromatograms are run with the butanol- water-benzoylacetone mixture. A systematic testing of the chromatogram with a series of spot test reagents then enables the unknown mixture to be identified. The behaviour of the oxine complexes of a number of cations has been studied by carrying out separations on strips of paper impregnated with 8-hydro~yquinoline.~~ The paper was prepared by treating it with an alcoholic solution of the oxine reagent and then allowing it to dry.The solvents tested included pyridine chloroform acetone and methyl ethyl n-propyl isopropyl and n-butyl alcohols and a number of separations were shown to be possible. Together with R values the workers in this field have listed a number of corresponding 8 or spreading factor values. If A is the initial band width and Bi the final band width for a cation i then Xi = &/A. By a consideration of both R and X values it is possible to predict whether a complete separation of two metal zones is possible. The interesting observation has also been made that cations may be divided into two groups those whose R values increase as the chain length of the alcohol used as solvent is increased and those whose R values decrease which supports the supposition that two types of 8-hydroxyquinoline complex may be formed.Separation of those metals which form chelate complexes with oxine has also been attempted 5 7 by using butanol containing 20% (v/v) of l2~-hydrochloric acid. A solution of the oxinates of the metals in dilute hydrochloric acid was applied to a paper strip which was then run with the butyl alcohol solvent. Although good separations were obtained it appears from the R values that the metals were in fact separated as their chloride complexes rather than as their oxinates. The chelated compounds undergo decomposition in the presence of the acid in the solvent and the oxine is found in a separate band (RF value 0.5). Use of a specific complexing agent has been applied to the separation of one metal from a group of others. Thus bismuth has been extracted with 6 4 F.H. Pollard J. F. W. McOmie and I. I. M. Elbeih J . 1951 466 470. G5 F. H. Pollard J. F. W. McOmie and H. M. Stevens J . 1951 771. asD. E. Laskowski and W. C. McCrone Anal. Chern. 1951 23 1579. 5'W. A. Reeves and T. B. Crumpler ibid. p. 1576. Y* 320 QUARTERLY REVIEWS acetone containing nitric acid from admixture with lead and copper after complexing the cations with diallyldithiocarbamidohydrazine.58 Copper has been extracted quantitatively on a paper strip from a number of other metals butanol saturated with 2~-ammonia and dimethylglyoxime being used as solvent. In addition to these separations on a column scale to which reference has already been made a variety of other metals have been handled on the larger scale made possible by the use of a cellulose column.The separation of zirconium from hafnium has been carried out 6 o by using ethyl ether containing nitric acid as solvent. The separation was not quantitative but by the use of large columns capable of dealing with 100 g. of zirconia it was possible to prepare substantial amounts of hafnia-free zirconia. Other separations which have been performed on an analytical scale include the extraction of thorium and scandium from minerals and ores with ethyl ether-nitric acid mixtures,61 62 the removal of zinc from solders by extrac- tion with butanol containing 2% v/v HC1 (d 1 ~ 1 8 ) ~ ~ ~ and the separate extrac- tion of both niobium and tantalum from a variety of minerals.64 6 5 9 66 For this last separation which was carried out in the presence of free fluoride a Polythene tube replaced the normal glass column.The test material was prepared in dilute hydrofluoric acid solution and the tantalum was extracted with ethyl methyl ketone saturated with water ; the niobium was subse- quently extracted with ethyl methyl ketone containing 124% of hydrofluoric acid (40% aqueous). The only interference came from tungsten which is partly extracted with the niobium. A simple separation which might readily be applied to quantitative analysis is the extraction of vanadium and molybdenum from t i t a n i ~ m . ~ ' Both metals have been readily separated by using acetone containing 2% v/v HC1 (d 1.18) as solvent. Ketonic solvents have also been used for the separation of the platinum metals rhodium iridium palladium and platinum. Mixtures of anions have been separated by procedures similar to those employed for cations.Solutions of the sodium salts have normally been employed in order to avoid interference but in most cases only the anionic part of the salt moves. The cations remain in the original spot or move only little. One group of workers have used pyridine containing water for the separation of fluoride chloride bromide and iodide. 4 O Another has separated chloride bromide iodide and thiocyanate with butyl alcohol saturated with 1.5~-ammonia.~~ A third group has measured the R values 58 M. M. Singh and J. Gupta J . Sci. I n d . Res. I n d i a 1951 10 B 289. 59 J. R. A. Anderson and M. Lederer A n a l . Chim. Acta 1951 5 396. 60 Chemical Research Laboratory unpublished work. 61 N. F. Kember Analyst 1952 77 78. O 2 A. F. Williams ibid.p. 297. 63 J. R. Bishop and H. Liebmann Nature 1951 167 524. 64F. H. Burstall P. Swain A. F. Williams and G. A. Wood J. 1952 1497. 65 A. F. Williams {bid. p. 3155. 66 R. A. Mercer and A. F. Williams ibid. p. 3399. 67 Chemical Research Laboratory unpublished work. 68 D. B. Rees-Evans and R. A. Wells in the press. M. Lederer Science 1949 110 116. WELLS INORGANIC CHROMATOURAPHY 321 of chloride bromide iodide chlorate bromate iodate nitrite nitrate arsenit e arsenat e carbonate phosphate ,'chromate t hioc y anate and sulphat e employing a solvent mixture of butyl alcohol pyridine and 1*5~-arnmonia in the proportions 2 1 2.54 The separation and estimation of borate from silicate and molybdate by means of acetone plus concentrated hydrochloric acid has been described.70 Following the work by C.S. Hanes and F. A. Therwood 71 on the separation of the phosphoric esters other workers have described the separation of ortho- pyro- meta- and poly-phosphates. Acid solvents have been used; thus one group of investigators have used a 70 30 5 mixture of isopropyl alcohol water and trichloroacetic a ~ i d 7 ~ another has favoured an aqueous butanol-acetic acid mixture. 73 The separ- ations have also been carried out with alkaline solvents e.g. a 40 20 39 1 mixture of isopropanol isobutanol water and ammonia (d 0.880). 7 2 By using as solvent a mixture consisting of equal volumes of n-butanol dioxan and N-ammonia the separation of phosphates phosphites and hypophos- phites has been demonstrated. 74 Cellulose has proved to be a very convenient adsorbent on which to carry out partition separations but other materials have been used.The complexes of a number of metals with 8-hydroxyquinoline have been separated on a column of silica gel.75 The oxinates were adsorbed 011 the column from a solution in chloroform and eluted with a mixture of chloroform and ethyl alcohol The order of adsorption was found to be Cu > Co > Fe > A1 > Bi > Ni > Pb. By elution with chloroform con- taining 1% of ethyl alcohol it was possible to obtain a complete separation of copper nickel and cobalt. Silica gel columns have also been used for the separation of zirconium and hafnium.76 On passage of a 1.5% solution of mixed zirconium and hafnium tetrachlorides in methyl alcohol through a silica gel column the hafnia is preferentially adsorbed and it has been shown possible to recover 60 g.of zirconia containing 0.1% of hafnia from 4.55 g. of the mixed materials with the aid of a column consisting of 290 g. of the silica adsorbent. The principles of electromigration have been combined with those of chromatography to produce separations of inorganic ions. The technique was introduced by G. Haugaard and T. D. Kroner 77 for the separation of amino-acids and has been named electro-chromatography. While the solvent was allowed to diffuse down a sheet of paper bearing the amino-acid mixture a potential gradient was applied across the paper a t right angles to the flow of solvent. The negatively charged acids moved towards the anode and the basic acids towards the cathode thus producing a two-dimensional 70 A. Lacourt G. Sommereyns and M.Claret Mikrochem. Mikrochim. Acta 1951 38 444. 71 Nature 1949 164 1107. 7 2 J. P. Ebel and Y . Volmar Compt. relad. 1951 233 415. 73 T. Ando J. Ito S. Ishi and T. Soda Bull. Chem. SOC. Japan 1952 25 78. 7 4 A. Bonnin and P. Sue Compt. rend. 1952 234 960. 76 L. B. Hilliard and H. Freiser Anal. Chem. 1952 24 752. 713 R. S. Hanscn and K. J. Gunnar J . Amer. Chem. SOC. 1949 71 4158. 77 Ibid. 1948 70 2135. 322 QUARTERLY REVIEWS separation. Using a similar technique H. A. Strain 78 has demonstrated the separation of inorganic ions. Separation of inorganic ions by electro- migration on filter-paper has been achieved by a number of worker^,^^-^^ but the combination of this technique with chromatography has introduced a number of new possibilities. Not only is it possible to carry out more difficult separations but the process may also be made continuous.An electrochromatographic cell has been designed by Strain for this purpose. 86 A sheet of filter paper 0.1’’ thick was clamped between two glass plates held vertically. Two electrodes consisting of fine platinum wire were inserted along the vertical edges of the paper. At the top of the glass plates was arranged an electrolyte holder so that electrolyte flowed continuously down the whole width of the paper. With the voltage applied a solution of the mixture to be separated was continuously applied to the centre of the top edge of the sheet of paper. As the test solution was carried down the sheet by the flow of electrolyte varying transverse movement of the ions present took place and diverging zones of the separated ions were formed.By a suitable arrangement of “ collecting wicks ’’ at the bottom of the strip it was possible to collect the electrolyte solution in portions containing the separated ions. A variety of electrolytes has been suggested for different separations. Thus conditions for the separation of large amounts of phos- phate from calcium have been given. The electrolyte 0-lM-lactic acid was allowed to flow through a 12-inch wide cell at the rate of 150 nil. per hour. A mixture of calcium chloride and phosphoric acid in IN-hydro- chloric acid constituted the original solution. A potential of 300 v giving a current of 150 milliamp. was applied. For a full discussion of the technique the reader is referred to a report by the U.S. Atomic Energy Commission.87 Ion Exchange Since the discovery by J.T. Way 8 8 in 1850 of the ion-exchange pro- perties exhibited by materials of the aluminosilicate type substances cap- able of undergoing ion exchange have been used for a wide variety of purposes. These include the concentration of dilute solutions removal of interfering ions chromatographic separations the study of complexes deter- mination of activity coefficients and use as catalysts for organic reactions. A thorough review of the whole field of ion exchange appeared in an earlier Review,89 and the reader is referred to this and to a number of excellent 78H. H. Strain and J. C. Sullivan Anal. Chem. 1951 23 816. 79 H. J. McDonald M. C. Urbin and M. B. Williamson Science 1950 112 227. 80K. A. Kraus and G. W. Smith J . Amer. Chem. SOC. 1950 72 4329. 8lM.Lederer and F. L. Ward Australian J . Sci. 1951 13 114. 82 M. Lederer Nature 1951 167 864. 83 M. Lederer and F. L. Ward Anal. Chirn. Acta 1952 6 355. 8 4 J. R. A. Anderson and M. Lederer ibid. p. 472. 8 5 H. H. Strain Anal. Chem. 1952 24 356. 86 T. R. Sato W. P. Norris and H. H. Strain Anal. Chem. 1952 24 776. S7 Idem Office of Tech. Services Dept. of Commerce Washington D.C. ; U.S.A.E.C. 88 J . Roy. Agric. SOC. 1850 11 313. 8 9 J. F. Duncan and B. A. J. Lister Quart. Reviews 1948 2 307. Document ANL 4724 (1951). WELLS INORGANIC CHROMATOGRBPHY 323 reviews and text books 8s 94 95 for a detailed treatment of the theory and practice of ion exchange. The present discussion is confined to the use of ion exchange in inorganic chromatography and deals with the principles and observations which have been utilised in obtaining separation of various substances with the aid of ion-exchange materials.Although a few chromatographic separations have been studied on naturally occurring and synthetically prepared siliceous materials e.g. zeolites and permutites and others have been attempted on ion-exchange materials prepared by the sulphonation of coal most interest has been concentrated on synthetic resin ion exchangers. The first cation exchanger of this type was prepared by B. A. Adams and E. L. Holmes 96 by the condensation of phenols with formaldehyde. The product yielded insoluble resin containing ionisable hydroxyl groups capable of ion exchange. The same workers produced anion-exchange resins by condensation of polyamines with formaldehyde. Since this early work the search for stable resins having a high capacity has led to the production of a wide variety of exchangers.Cation exchangers combining the properties of physical and chemical stability with high capacity have been produced by nuclear sulphonation of polymeric aromatic hydrocarbons and most of the recent work on the separation of cations by ion exchange has been carried out on materials of this type. The study of anionic separations has received great impetus from the introduction of strongly basic anion exchangers of high physical stability. The stability of the earlier weakly basic anion exchangers which relied upon primary secondary or tertiary groups for their ion-exchange properties left much to be desired. The new materials containing quaternary ammonium functional groups are strong bases and have a relatively high chemical and physical stability.The general equation for the equilibrium between univalent ions on an excha,nger and in solution may be written where R is the insoluble resin ion. If the solutions are dilute the ratio of activities in solution may be replaced without much error by the ratio of concentration and (1) becomes where yltB and Y E A are the activity coefficients of the ions A and B in the resin phase. Since the ionic strength in the resin is constant yaB/ynA may be assumed nearly constant over a wide variation in the ratio [RB]/[RA] and (2) becomes R A + B * + R B + A h By the mass law anBaAf/aRAaBt = K . (1) [RBly,,I[RAly,* = K,[B*l/[A*l - * (2) = KA+BIB*l/[A*l soR. Kunm Anal. Chem. 1949 21 87. Dl Ibid. 1950 22 64. 94 F.C. Nachod “ Ion Exchange ” Academic Press Inc. New York 1949. 95 R. Kunm and R. J. Myers “Ion Exchange Resms ” John Wiley & Go. Inc. 96 J . SOC. Chem. Ind. 1935 54 1 ~ ; B.P. 450,308 450,309 (1934). 92 Ibid. 1951 23 45. s3 Ibid. 1952 24 64. New York 1950. 324 QUARTERLY REVIEWS KA-tB may be termed the affiity coefficient and gives a measure of the relative af'finity of the resin for the ions A and B. By a similar argument an expression for the more general case of multivalent exchange may be derived. The affinity coefficients of a number of cations relative to hydrogen and of a series of anions relative to chloride have been measured.97-100 The numerical values obtained vary with the type of functional group and the degree of cross-linking of the resin but the order of affinities remains the same.In general the higher the charge on a cation the more strongly itt is adsorbed and for ions of the same charge resin affinity increases with decrease of hydrated ionic radius. At the present moment it is not possible to make a similar generalisation correlating anion size with order of adsorption. Consideration of the difference in affinities leads to the simplest type of chromatographic analysis named by A. Tiselius lol frontal analysis. If a solution containing a mixture of ions having different affinity coefficients is passed through a column of ion-exchange resin in sufficient quantity to exceed the exchange capacity of the resin the least adsorbed ion " breaks through " the column first. This has been made the basis of a number of inorganic separations although generally the method has proved more useful in the organic field.A solution containing O-h-KCl and O-lN-NaCl was passed through a column of cation exchanger in the hydrogen form,lo2 and the effluent solution was collected in fractions and analysed for hydrogen sodium and potassium. At first hydrogen ion was displaced a t a concentration of 0 . 2 ~ ~ Le. equal to the total influent. Sodium having a lower affinity than potassium then broke through and its concentration in the effluent rose rapidly to 0 . 2 ~ ~ i.e. it was displaced by the potassium. Later potassium appeared in the effluent and its concentration increased to 0 . 1 ~ ~ a t which point analysis of input and effluent solutions was the same. In a similar manner copper and zinc have been separated l o 3 in sulphate solution by passage through a column of cation exchanger having carboxylic functional groups.The less adsorbed zinc appeared in the effluent first and continued to be displaced by the copper. B. A. J. Lister attempted t o use this method for the separation of zirconium and hafnium.104 He passed a solution of mixed zirconyl and hafnyl nitrates in N-sulphuric acid t'hrough a column of a high-capacity cation exchanger. The experiment was not a complete success since small amounts of hafnium and zirconium were detected in the effluent almost from the beginning. Hafnium continued to be extracted a t the same concentration for some time after the break-through of zirconium until finally there was a rapid increase in concentration. The net effect was that a considerable amount of zirconium was recovered with a much reduced ha'fnia content but a complete separation was not obtained.97 G. E. Boyd J. Schubert and A. W. Adamson J . Amer. C'hem. Soc. 1047,69 2818. 98 B. H. Ketelle and G. E. Boyd ibid. p. 2800. 99 R. Kunin and R. J. Myers ibid. p. 2874. loo R. M. Wheaton and W. C. Bauman I n d . Eng. Chem. 1951 43 1088. lol Arkiv Kemi Min. Geol. 1940 B 14 No. 22. lo2 Anon. Chemistry Research 1951 D.S.I.R. p. 92. lo3 E. J. Breton and A. W. Schlecten J. Metals 1951 3 517. lo4 J. 1951 3123. WELLS INORGANIC CHROMATOGRAPHY 325 This may have been due to the presence of colloidal and anionic complexes which are formed in nitrate solutions of both metals. One of the earlier separations carried out with the aid of ion exchange was the partial separation of the isotopes of lithium by T.I. Taylor and H. C. Urey.105 This was again a frontal-analysis experiment carried out by passing a solution of lithium chloride down a column of natural zeolite. A more efficient separation was obtained by Glueckauf et who used a column of Zeo-Carb H-I ground and sieved to a grain size of 1.5 x 10-3 cm. and passed through it a solution of lithium acetate. In the first frac- tions of effluent collected the 6Li isotope content had been reduced from 7.5 to < 1.0%. Other examples of separation by frontal analysis have been reviewed previously. A further type of chromatographic separation has been termed dis- placement development. The mixture is adsorbed from solution on to a resin and then displaced with a reagent more strongly retained by the resin than any component of the mixture.The most strongly adsorbed material displaces less strongly adsorbed materials each component of the mixture in turn displacing less readily adsorbed components. If the volume of effluent is plotted against concentration a stepwise diagram is obtained in which each step as it emerges consists of one pure component all compo- nents moving through the column a t the same rate. Little use has been made of the technique as yet for inorganic separations By far the most used method of separation is that of elution development. The mixture to be separated is adsorbed in a narrow band at the top of a resin column and is then desorbed by passing down the column a solution of another ion which has a lower affinity coefficient than the components of the mixture and which is already adsorbed on the major portion of the column.The components of the mixture separate to an extent depending upon their distribution coefficients and on the nature of the adsorption iso- therm for each material. The distribution coefficient a t equilibrium of an ion has been defined lo6 as K d = (Ms/mass of resin)/(2ML/volume of liquid) where M and M are the fractions of the ion on the solid resin and in the liquid phase. If the volume of effluent solution is plotted against concen- tration' of solute an elution curve is obtained. In cases of low concen- tration where a linear isotherm is obeyed the elution curve is of the Gaussian type provided that the column is run sufficiently slowly for equilibrium conditions to have been approached. In more concentrated solutions the Langmuir and less frequently the Freundlich isotherm applies and the shape of the elution curve alters accordingly.Thus a common result is a sharpen- ing of the front edge and trailing of the back edge of the elution curve. This simple type of elution analysis can be illustrated by reference to the separation of the alkali metals. The first success was a partial separation lo' of Na K Rb and Cs in that order obtained by adsorbing the mixture on a column of Dowex 50 (sulphonic acid-type cation exchanger) and elution lo5 J. C'hem. Phys. 1938 6 429. lo6 E. R. Tomkins and S. W. Mayer J. Amer. Chent. SOC. 1947 69 2869. lo' W. E. Cohn and H. W. Kohn ibid. 1948 70 1986. 326 QUARTERLY REVIEWS with 0-15~-hydrochloric acid. Later in experiments with two-component mixtures sodium was completely separated from potassium by elution with 0-1N-perchloric acid,l O8 and sodium was separated from magnesium by elution with either this acid or 0-1N-hydrochloric acid.log By use of radio- active tracers a complete separation 110 of the four metals Na K Rb and Cs was achieved by first eluting Na and K with 0-1N-hydrochloric acid and then Rb and Cs with N-acid.A fine piece of work by W. Rieman et aZ.lll has placed the separation of Li Nay and K on an analytical basis and the method has been applied to the routine estimation of alkali metals in sili- cates.l12 The elution was carried out with 0-7~-hydrochloric acid on a column of colloidal Dowex 50 and the behaviour of a number of other metals was studied. Cadmium was eluted before lithium and mercury before cadmium; beryllium was eluted with the potassium and lead zinc and manganese were eluted between potassium and magnesium.Other workers '13J 114 investigated the separation of the alkali metals and a mix- ture of Li Be and A1 has been separated,l15 O-h-HCl being used to elute the lithium O-O~N-C~CI to elute the beryllium and ~ N - H C ~ to remove the aluminium. This type of separation depending upon the selectivity of the exchanger for a number of ions has also been utilised in the separation of anions.l16 The halides F C1 Br and I have been separated on a column of Dowex 2 (strong base exchanger) by elution with M-sodium nitrate the pH of which had been adjusted with sodium hydroxide to 10.4. The scope of elution analysis was greatly increased by the introduction in 1947,117 by workers for the Manhattan Project, of the use of complexing agents as elutriants.If two cations in solution compete for a rcsin an equilibrium is set up in which the amount of each ion on the resin phase depends upon its activity and upon its relative affinity for the resin. If now a complexing agent which will strongly complex one of the ions is added to the system the concentration of that ion in the solution will be greatly reduced and the equilibrium will move in favour of the adsorption of the second ion. This reasoning was applied to the separation of the rare earths alkaline earths fission products and transuranic elements on cation- exchange resins. The eluting agent was a suitably buffered solution of citric acid. For a tervalent rare-earth ion adsorbed on the ammonium form of a cation exchanger two reactions are involved 3NH,R + M+++ + MR + 3NH,+ .( 1 ) M+++ + 3H,Cit- + M(H,Cit) . (2) lo8 G. Kayas Compt. rend. 1949 228 1002. R. Bouchez and G. Kayas ibid. p. 1222. G Kayas J. Chim. Phys. 1950 47 408. 111 J. Beukenkamp and W. Rieman 111 Anal. Chem. 1950 22 682. 112 R. C. Sweet W. Rieman 111 and J. Beukenkamp ibid. 1952 24 953. l13H. Kakihana J . Chem. SOC. Japan 1951 72 255. 114 R. Wickbold 2. unal. Chem. 1951 132 401. l15M. Hondo J. Chem. SOC. Jupan 1950 71 118; 1951 72 361. 11' Ibid. 1947 69 2769-2881. . 116 R. W. Attenberry and G. E. Boyd J . Amer. Chem. SOC. 1950 72 4805. WELLS INORGANIC CHROMATOGRAPHY 327 The distribution of the rare earth between resin and solution will depend upon the equilibrium constants for these reactions. Citric acid gives three ions in solution H,Cit- HCit- and Cit-- according to the pH and thus the concentration of the rare-earth ion will depend on the pH of the solution.At any one pH value the rare-earth ion will exist as a mixture of several citric acid complexes so equation (2) is an approximation. How- ever if 5% citric acid solution is used as eluting agent optimum separation of the rare earths is obtained with the pH between 2 and 3. In this range citric acid is almost completely in the H,Cit and H,Cit- forms so the most probable rare-earth complex is M( H,Cit),. Those cations which form the strongest citrate complexes are also those for which the resin has the lowest affinity so that complexing elution accentuates minor differences in adsorp- tion and improves the separation. The use of citrate eluting agents for the sepa,ration of cations has been described and reviewed on numerous occa- s i o n ~ .~ * ~ l183 119 Papers continue to appear describing attempts to simplify increase the scale extend the scope or improve the degree of separation for rare earths and other metals by use of citrate solutions. A mixture of titanium zirconium and thorium has been separated l 2 0 by elution from a column of colloidal Dowex 50 with a 1% solution of citric acid buffered to pH 1.75. The eluting agent was then changed to O.O5~-diammonium hydrogen citrate in order to elute thorium in a reasonable volume. From measurement of distribution coefficients the predominant complexes over the pH range 1.59 -+ 5.39 were calculated to be TiCit2-2 ThCit,-2 ZrO(HCit),-2 (at lower pH values) and ZrOCit- (at higher pH values).Other factors being equal the sharpness of ion-exchange separation increases with decrease of resin particle size. To obtain optimum and repeatable results for the separation of titanium zirconium and thorium it was necessary to use a resin sieved to a mesh size of 100-120. Strict control of temperature (25" In general the effect of increased tem- perature is to increase the rate of exchange with resultant increase in the sharpness of elution peaks. The high affinity of cationic resins for thorium has suggested 121 a method of separation of the rare earths from this element. The rare earths can be eluted from a cation exchanger with a 10% .citric acid solution buffered to pH 3. Thorium is retained and can be subse- quently eluted with 6~-sulphuric acid. The adsorption of the rare earths from citrate solution on an anion- exchange resin indicates the presence of anionic complexes of the rare earths.Tracer quantities of promethium and curopium were adsorbed 122 from solution in 0.0125M-CitriC acid buffered to pH 2.1 with hydrochloric acid on a column of Dowex 1. They were then eluted differentially with a further portion of the same solution. A similar separation of the oxalate Zirconium was the first to be eluted followed by titanium. 0.1") was also necessary. 118 L. L. Quill Record Chem. Progress 1950 11 151. 119 R. Bock Angew. Chem. 1950 62 A 375. lZo W. E. Brown and W. Rieman 1x1 J. Amer. Chem. Soc. 1952 74 1278. lZ1 P. Radhakrishna Anal. Chim. Acta 1952 6 351. 122 E. F. Huffman and R. L. Oswalt J. Amer. Chem. SOC. 1950 72 3323. 328 QUARTERLY REVIEWS complexes of zirconium and niobium has been achieved.123 The resin was again Dowex 1 and after adsorption from oxalic acid solution ( 0 .1 4 . 4 ~ ) the met'als were eluted zirconium first with a 1M-hydrochloric acid solution 0 . 0 1 ~ with respect to oxalic acid. At first sight the separation of zirconium and hafnium on a cation- exchange resin by elution with dilute acid might appear to be completely due to differences in adsorption affinity but further study has shown it to be due to complex formation. If the two metals are eluted from a cation exchange with 34~-hydrochloric acid 124 separation occurs and the hafnium is extracted first. In an experiment 125 using 2.8 g. of the mixed oxychloride 42% of the HfO was recovered at better than 99.9% purity. Elution with N-sulphuric acid 12G 12' also produced a separation but in this case zir- conium was extracted first well in advance of the hafnium.The ease with which zirconium forms anionic complexes with sulphuric acid is well known. As will be shown later it also forms negatively charged complexes in concentrated hydrochloric acid solution. Elution with nitric and per- chloric acids produced no separation but wherea's the rate of elution of both metals with nitric acid was similar to that obtained with hydrochloric and sulphuric acids both metals were extracted much more slowly with perchloric acid indicating little complexing in perchloric acid solution. B. A. J. Lister has described experiments 12G in which 95-9Sy0 of the zirconia containing 0.01 yo of hafnia has been recovered from 20 g.of zirconyl nitrate having an original hafnia content of 1.5%. K. A. Kraus and G. E. Moore 128 have also separated zirconium and hafnium on an anion-exchange column making use of their negatively charged fluoride complexes. Elution with a mixture of O-~M-HF and ~~OM-HCI extracted zirconium first but a complete separation was not obtained E. A. Huffman and R. C. Lilly 129 also studied the same separa- tion and investigated the effect of varying the hydrochloric acid concen- tration over the range 0-15-0-6~ and the hydrofluoric acid concentration over the range 0-01-0.6~ but in no case obtained a complete separation. This work on the Auorozirconates and fluorohafnates was extended to include a study of the separation of niobium tantalum and protactinium and the separation of all three from zirconium.130 The elution constants for the four metals were measured with elutriants containing varying quantities of HCl and HF.The elution constant ED was defined as dA/V where d is the distance (crn.) a band maximum moves after the passage of 'v ml. of eluting solution through a column of A cm. cross-section. It was found that for HC1 concentrations of 0 . 2 - 9 ~ and HF concentration of 0-1-5.0~ the order of elution constants was Pa > Nb > Ta. Variation in the acid concentration outside these liniits could bring about changes in the order 123 R. E. Wacker and W. H. Baldwin U.S.A.E.C. ORNL-637 N.S.A. 1950,4 460. lZ4 K. Street and G. T. Seaborg J . Amer. Chem. SOC. 1948 70 4268. lZ5 I. E. Newnham ibid. 1951 73 5899. lZ6 B. A. J. Lister ibid. p. 3123. 12' 5.C. Hutcheon and B. A. J. Lister Reseccrch 1952 5 291. 128 J . Amer. Chem. SOC. 1949 71 3263. l30 K. A. Kraus and G. E. Moore ibid. 1949 71 3855; 1951 73 9713 2900. 12B Ibid. p. 4147 ; 1951 '73 2902. WELLS INORGANIC CHROMATOGRAPHY 329 of elution of the niobium and tantalum. Zirconium normally preceded protactinium in the series but here also by adjustment of acid concentra- tions zirconium could be more strongly retained than niobium. The behaviour of the metals in this fashion on elution depends upon the anionic species of each element existing in solution. It appears likely that bot'h protactinium and niobium in low HF concentrations exist as their univalent oxygenated complexes PaOC1,- and NbOCI,- (or more generally H,MOCl,,,,). Increase in HF concentration leads to complex formation by which the negative charge is increased.Thus the equilibrium for niobium becomes For tantalum two complexes TaX,F- and HTaX,F3-2 have been identi- figd. The practical separation of the metals depends q o n the concentration of HF necessary to bring about formation of each more highly charged and hence less adsorbed complex. All four metals were separated from each other in one experiment. After adsorption from a solution of 9&l-HC1 + 0.05~-HF on to a colurnn of Dowex 1 zirconium was eluted followed by protactliniuni with a mixture of ~ M - H C ~ and O-OO~M-HF. The HF concentration was increased to 0 - 1 8 ~ to extract niobium and finally the elutriant was changed to ~M-NH,C~ + ~M-HF to remove tantalum. Observations 31 on the behaviour of protactinium in concentrated hydrochloric acid solution have led to the development of a very useful complexing technique.Protactinium is strongly adsorbed from solutions of hydrochloric more concentrated than 4M and its distribution coefficient rises rapidly to a maximum at 8M. It may be eluted in a sharp band with acid concentrations below 4 ~ suggesting that protactinium does not form anionic complexes in low acid concentrations. Similar results were obtained with iron.132 While chromium aluminium and the rare earths were almost unadsorbed from strong hydrochloric acid solution the distribution co- efficient for iron increased rapidly in solutions from ca. 1~ to a maximum at 9 ~ . Zirconium hafnium niobium and tantalum also form anionic com- plexes in fairly concentrated (> 6 ~ ) hydrochloric acid s01ution.l~~ Advan- tage has already been taken of this in the extraction of zirconium and hafnium adsorbed on a cation exchanger.The distribution coefficients of zirconium and hafnium increase sharply above ~ N - H C ~ while those for niobium and tantalum are high in concentrated and dilute solutions but have a minimum at ~ N - H C ~ . A simple separation of cobalt and nickel has been dem0n~trated.l~~ The elution coefficient for both metals adsorbed on Dowex 1 on elution with ~ N - H C ~ is 2.5. For coba'lt this falls to 0.02 with W-HC1 and then rises to 0.055 with 12~-HC1 while ENi remains at 2.5 throughout the whole range of acid concentrations. A serious disadvantage to the use of citrate buffer solutions for t'he elution of the rare earths is the large volume of solution required.A NbX,- + nHF + jH,O + Hn-jNbX8-mFn-l-j+" + mX-l + jH,O+ 131 K. A. Kraus and G. E. Moore J . Amer. Chern. SOC. 1950 72 4293. 132 Idem ibid. p. 5792. 133 E. A. Huffman G. M. Iddings and R. C. Lilly ibid. 1951 '73 4474. 134 K. A. Kraus and G. E. Moore ibid. 1952 74 843. 330 QUARTERLY REVIEWS number of other complexes have been studied with the hope of reducing this volume. The ammonium salts of the iminodiacetic acids form com- plexes with the rare earths which show increasing stability with increase of atomic number. The possible ionic forms of these acids are represented by the equilibrium CH ,420 -0- CH ,*CO*O- CH ,420 00- I H+ I H+ I I I NRH+ I - NR - - NRH+ 7 CH,*CO ,H CH *C 0 a 0 - CH,.CO*O- (1) (2) (3) Between pH values of 5 and 7 the most suitable range for rare-earth separations forms (1) and (2) are in major concentration and rare-earth complexes of the type MA+ and MA,- are formed where M is the tervalent cation.Use of this type of complex has been demonstrated 135 by the use of nitrilotriacetic acid solution buffered to pH5 with ammonia for the separation of the cerium group on Amberlite 120 (sulphonic acid cation exchanger). C. E. Higgins and W. A. Baldwin have shown that tracer quantities of yttrium 91 and europium 136 may be separated on Dowex 1 by elution with a solution of 0*16~-versene. An interesting analytical application of complexing elution is that for the estimation of iron and titanium in ilmenite.137 A solution of the mineral was adsorbed on a cation exchanger and eluted with %-potassium cyanide to remove the iron. Titanium was subsequently removed with 10% Sulphuric acid.Uramildiacetic acid has been studied l38 as a possible eluting agent for Li K and Na. Here again the object was to cut down the volume of solution required for elution. The alkali metals form complexes with uramildiacetic acid (I) which are strongly pH dependent and except for lithium are NH-C/ CH2.C0,H stable only in alkaline solution. The resin has therefore to be used in a neutral form preferably containing an adsorbed cation whose affinity for the resin is less than that (1) of the alkali metals. In a succ.xsfu1 experi- ment the resin was used in the tetramethyl- ammonium form and the metals were eluted with a solution of the tetramethylammonium salt of uramildiacctic acid. Variation of the pH of the eluting solution between 7 and 11 enabled Li Na and K to be separated.Organic solvents containing mineral acids have recently been used to elute complex metal ions from resin exchangers. Complex cyanides such as auro- argento- ferro- cupro- and nickelo-cyanides arc readily adsorbed 0- \ +/ C-NH / \ oc / \ CH,.CO,H NH-CO 135 F. T. Fitch and D. S. Russell Canadian J. Chem. 1951 29 363. 136U.S.A.E.C. Rep. ORNL 894 N.S.A. 1951 5 542. 13' Y. Yoshino and M. Kajima Bull. Chem. SOC. Japan 1950 23 46. 138 W. Buser Helv. Chim. Acta 1951 34 1635. WELLS INORGANIC CHROMATOGRAPHY 331 from dilute sodium cyanide solutions on to an anion-exchange resin. While nickel can be removed with dilute hydrochloric acid and cupro- and ferro- cyanides with 2~-sodium cyanide it is extremely difficult to elute auro- and argento-cyanides from strong base resins of the Amberlite IRA-400 type with normal aqueous eluting agents.Both are readily removed with solvents such as acetone alcohol or ether containing hydrochloric or nitric acid. Although this is perhaps not true chromatographic elution some selectivity by the solvents was noted. In view of the high water content of ion-exchange resins it might be possible to combine an ion-exchange adsorption from aqueous solution with a " partition elution " using an organic solvent. Theoretical Considerations A general theory of chromatographic separation was first developed by J. N. Wilson and has since been revised and expanded by D. de Vault E. Glueckhauf J. Weiss L. G. Sillen and others. The mathematics of the theory is complicated and the reader is referred to the original papers and to a recent review.lgO I n general the theory permits the prediction of the chromatographic behaviour of a number of solutes in a mixture if the adsorp- tion isotherm for each is known.The equations necessary to cover linear Freundlich and Langmuir isotherms have been derived. The theory was derived on the assumption that the separations were achieved under equi- librium conditions but it has been possible to derive modified equations for use in non-ideal ~onditions.1~1 This has permitted the calculation of adsorption equilibria data from the diffuse rear boundary of an elution curve for linear or non-linear isotherms providing the conditions do not depart radically from ideality. Such variables as grain size flow rate concentra- tion pore space and diffusion constants in the liquid and in the solid adsorb- ent are catered for by the theory.Glueckauf 25 has determined the effect of some of these variables on the separation of the isotopes of neon on charcoal a t -198". He found close agreement between the practical results and theoretical predictions. A simpler but more empirical approach has been favoured by a number of workers who have employed the " plate " theory. Initially derived by A. J. P. Martin and R. L. M. Synge lo to give a quantitative explanation of partition chromatography the theory is based upon the similarity between the chromatographic process and that of distillation. A column is con- sidered to consist of a number of theoretical plates. Each plate is of such a thickness that the solution issuing from it is in equilibrium with the mean concentration of solute in the non-mobile phase throughout the layer.An expression was derived which enabled the position of a band in a column after the passage of a known volume of solvent to be predicted. Little use has so far been made of this equation in inorganic separations but a second 139 F. H. Burstall P. J. Forrest N. F. Kember and R. A. Wells in the press. 140 E. Glueckauf Nature 1951 167 715 ; J. F. Duncan and E. Glueckauf ibid. 141 E. Glueckauf J. 1949 3280. p. 714 ; E. Ekedahl E. Hogfeldt and L. G. Sillen ibid. p. 714. 332 QUARTERLY REVIEWS expression derived 33 from it for the special case of separations on a paper strip has received more attention. For separations on sheets of paper movement of band movement of solvent front R p = where R = - ____ A + aA from which a = AL(l/BF - l)/& where cc = partition coefficient of solute between water and solvent phases and AL/A = ratio of volume of solvent and water phase in the chromatogram.The theory holds only for linear isotherms but it appears likely that these apply in a large number of cases for separations on paper strips. If the partition coefficient is constant the RE values will also be constant and this has been found to be so for a large number of inorganic separations. S. W Mayer and E. II. Tompkins lo6 have also derived a '' plate " theory for ion exchange based upon the analogy of ion-exchange column procedures with fractional distillation. They derived the expression where .Lmax. = the maximum fraction of a solute in any volume v of solution ( u = volume of solution in one theoretical plate) p = number of theoretical plates in column and C = distribution ratio of solute in any plate.As with the equation derived by Martin and Synge this relation holds only for those cases where the amount of a solute adsorbed is proportional t o its concentration in solution. For most ion-exchange work this is true or nearly so for dilute solutions. The theory has to some extent been used to predict the degree of separation attainable in several cases but there is still a great tendency to produce the optimum separations by trial and error methods and then to use the practical data to test the theory. I n addition to attempts to produce a general theory of chromatography much work has been directed towards the determination of the physical and chemical processes and their kinetics which operate with various adsorbents.Recent suggestions for the mechanism of alumina columirs have already been mentioned. The mechanism of adsorption in paper- partition separations appears to be fairly straightforward but several factors have received little consideration. A number of workers have noticed that on allowing an organic solvent containing water to diffuse down a sheet of paper two solvent fronts are formed. The paper behind the first of these fronts appea'rs to contain both water and solvent whereas the area between the front appears to contain only dry solvent. It has been shown 142 that the " wet " solvent front is in fact an indication of a water concen- tration gradient which extends from the solvent container to the solvent front.Moreover if acid is present in the solvent an acidity gradient extends down the strip analogous to the distribution curve for water. What part these extreme non-equilibrium conditions play in producing separations is not known. For a detailed study of the kinetics and theory of ion exchange the reader is referred to a number of text books 8 p 04 95 and a review,143 in which 142 R. A. Wells unpublished. 143 E. R. Tompkins Analyst 1952 '7'7 970. WELLS INORGANIC CHROMATOGRAPHY 333 these matters have been fully discussed. Recent publications 144-146 have described methods for carrying out continuous ion-exchahge separations in conditions of low concentration. I n these cases separations do not neces- sarily occur under equilibrium conditions and the authors have made substantial headway in correlating the kinetic theories of ion exchange with the plate theory in order to predict the course of separations carried out under non-ideal conditions.The Reviewer would like to express his thanks to the Director of the Chemical Research Laboratory Teddington €or permission t o publish this article. 144 N. K. Hiester Tech. Rep. No. 6 U.S.A.E.C. COO-41 1951. 14& T. Vermeulen and N. K. Hiester Ind. Eng. Chem. 1952 44 637. 146 N. K. Hiester R. C. Phillips and E. T. Fields Tech. Rep. 11 U.S.A.E.C. COO-59 1952.

ISSN:0009-2681    

DOI:10.1039/QR9530700307

出版商:RSC    

年代:1953

数据来源: RSC