“ NON-RACEMIC ” AND “ RACEMIC )) LIQUIDS. 111 1 ALTHOUGH during the last few years Ladenburg lias persistently endeavoured to demonstrate the existence of raceinic liquids, most of the methods which he has used &ord results which have no bearing upon the question at issue. Thus he found (Anizcden, 18SS, 247, S7) that on converting a mixture of dextro- and lLevo-coniine in unequal quantities fractionally into its compound with cadmium iodide, an alteration in the specific rotatory power of the mixture of bases could be effectedj and conchded that this proved the rncemic nature of inactive coniine, but Fischer pointed out ( B e y . , 1894, 27, 3224), and it follows from recent work of Kipping and Pope (this vol., p. 1119), that this fact has nothing t o do with the racemic or non-racemic character of the liquid.Again, Ladenburg observed that on mixing dextro- and lrevo-coniine a fall in temperature occurs, and attributed this t o chemical change (Be).., 1S95, 28, 164) ; he noted fwther that, although the density is not altered thereby, the refractive index increases notably. This result involves the remarkable conclusion, which Ladenburg does not draw, that the molecular refraction of (6 racemic” coniine is about 1 per cent.. greater than that of its optically active components. Since so much depends upon the densities (compare Traube, Bey., 1S96, 29, 1396), which are not given in the paper referred to, judgment must be deferred ; Laden- burg, however, years ago ( A m d e n , 1888, 24’7, S l ) gave experimental numbers showing the density of active and inactive coniine to be the same.A criticism of his latest method (Bey., 1899, 32, 1822) is to be found in the following payer (p. 1119). Although the methods hitherto adopted in the study of externally compensated liquids are so unconvincing, yet it seems possible t o devise a trustworthy method. We owe to the labours of Ramsay and Shields (Zeit. plysikal. Chem., 1893, 12, 433) and others the know-11.12 POPE AND PEACHEY: A METHOD FOR DISCRIMINATING ledge that many liquid substances are associated, the physical mole- cules being aggregates of several chemical molecules ; some classes of compounds, such as bases, are highly associated, whilst others, like the hydrocarbons, have, in general, association factors approximating to unity.I n this question of association is to be found the cause of many facts relating to optical activity which are otherwise difficult to explain (compare also P. Frankland, this vol., p. 347). Thus, if n highly associated substance be dissolved in a solvent, it is to be expected that it will break down into simpler molecules moiw or less completely, the extent to which this occurs depending upon the solvent and the concenti-ation ; but if a liquid already consisting of chemical molecules be dissolved in a solvent, it cannot undergo further dissociation." Further, if in any case it is possible for a solvent to cause an increase in the association factor of an already associated solute as suggested by Frankland (this vol., p. 363), this is an unlikely contingency when the pure solute has an association factor approximating to unity.Since an optically active substance neces- sarily has different rotation constants according as it is associated to different degrees, we must expect to find that the specific rotatory power of substances having high association factors in the pure liquid state varies considerably with change of solvent and of concentration, whilst those substances having in the pure liquid state association factors approximating to unity would in solntion have specific rotatory powers but slightly dependent on the solvent and the concentration. An examination of tho rather meagre data available, shows this con- clusion to be appreciably i n accord Kith the facts, substances which from their nature should have high association factors exhibiting very different specific rotatory powers as the solvent is changed, whilst those which should be nearly monomolecular vary but slightly in specific rotatory power in like circutustances.Kesults are given in the present paper showing that, in the case of Isvotetrahydro- quinaldine, [ .]D varies from - 45.9" to - 11 7.9" in various solvents, whilst the specific rotatory power of lsvoyinene varies in different sol- vents between [a], - 33" atid - 42". Although we have no direct infor- mation respecting the association factors of these two substances in the pure state, yet from the fact that Traube (Bey., 1897, 30, 265) gives the association factors of aniline, pyridine, quinoline, and piperidine as 1.35, 1.75, 1.40, and 1.62 respectively at the ordinary tempera- ture, and the association factors of the aliphatic hydrocarbons and the higher homologues of benzene as unity, it may be judged that lsvopinene is monomolecular, whilst the association factor of laevo- tetrahydroquinaldine is about 1.5, * The question of electrolytic or of hydrolytic dissociation is naturally not referred to here.BETWEEN '' NON-RACEMIC " AND '' RACEMIC " LIQUIDS. 1113 I n a preceding paper (p.1066), we have elaborated a method by which at least one optically active component of an externally compensated base can be prepared in quantity, and in a state closely approaching purity. We purpose -continuing this work, supplementing it with a careful comparison of the active and inactive liquid bases; in the present paper me give the results obtained with the tetrnhydroquinaldines : results which justify us in definitely concluding that, at the ordinary temperature, dextro- and Isvo-tetrahydroquinaldine exist in the same state of molecular aggregation, whether separate or mixed, and there- fore do not combine t o form a liquid racemic compound.A considerable number of data respecting the densities of optically active and externally compensated liquids has been collected, but the only available accurate values we have been able to find are those given by Ladenburg (dmulen, 1888, 247, 81) in his classical work on coniine ; he finds that the densities of lcevo- and inactive coniine at Oo/Oo, are 0.5635 and 0.8626 respectively, numbers identical within the limits of experimental error.Density determinations of the active and inactive nlkylic glycerates have been given by Frankland and MacGregor (Trans,, 1893, 63, 511) ; these data, and also those given by Schiitz and Marckmald (Ber., 1896, 29, 52) for the valeric acids, allow of no conclusions being drawn as the experimental error is apparently large. Traube (Beis., 1896, 29, 1394) concludes from the densities of the active and inactive limonenes and carvones, that these substances are monomolecular, and that the inactive substances are merely mixtures ; the density numbers used are, however, only very approximate. Densities of , h v o - and Extewudly Coqvensated 27et~a~?l~~oqz~ii~akEine. We obtained laevotetrahydroquinaldine in B highly purified condi- tion by distilling its dextro-a- bromocamphorsulphonate (this vol., p.1067) with soda in a current of steam, extracting the distillate with purified ether, and distilling off the ether on the water-bath; the residual oil is then repeatedly distilled under about 50 mm. pressure. Highly purified samples of externally cornpensated tetrahydroquin- aldine have been prepared in a similar manner from the pure racemic hydrochloride (this vol., p. 1086). The densities were determined in a Sprengel tube holding about 5.4 c.c., with the following results : L.cevotet~.nl~?/droqui~~~ Zdilie. (1) Preparation B : d at 14*5'/4' = 1.03365. 9 9 A : ,, 18*5"/d0 = 1.03046. B : ,, 20.2'/4'= 1.01914, (2) (3) 931114 POPE AND PBACBEP : A IdETHOD POR bISCRIMINA'l'INu Extent ally Co nyiensa t ed I'e t rahy dr oquimld ine .(4) Preparation C : d at 14*5'/4'= 1,02362. ( 5 ) ,, D : ,, 16°50/40=1*02219. (6) ,, D : ,, 18*0°/4'=1~02083. (7) The density of a mixture of 3.8503 grams of lwvotetrahydro- quinaldine (preparation S) with 7.3455 grams of externally compen- sated tetrnhydroquinaldine (preparation D) was found to be 1-01 915 at 20*2'/Oo. An exa- mination of the above seven determinations shows that within the limits of experimental error the density of all the samples is given by the expression, d at t/4'= 1.01930 i- 0'00079 (20 - t ) , so that the lavo- base, the inactive base, and the mixture have the same density at the game tempernt ure, A dextrorotatory base and its Itxvorotatory antipodes necessarily have the same density under the same conditions; if both are asso- ciated and no chemical action occurs on admixture, it would be expected that no change in the association factor of either isomeride follows admixture.This latter condition can only be fulfilled if no alteration of density occurs; from the density determinations alone we must therefore conclude RS highly probable that no alteration in the state of molecular aggregation occurs on mixing dextro- and Isvo-tetrahydro- quinaldine. Four different preparations, A, B, C, and D, were used. Befraction C ' o n s t ~ ~ t s of h m o - m d h'xternally Covipenscded I'etral&*o- quinaldine. Tho refractive indices of ltevo- and externally compensated tetl.ahyclroquinalcline (preparations B and D) were determined for sodium and thallium light in a hollow glnas prism of about 60° angle j the results are given in the following table, the uioleculitr refractions, ?a' - I i V m, being calculated from the expression, ,@= - 122 f 2'2- ~ Sample.t . LXVO- B ............ ,, ,, ............ Iuactive D ............ ,, ,, ............ Light. Na TI Nll T1 71. 1.5iOSO 1'57724 1.57273 1.55184 47.50 48.07 47.57 48.19 The values for the I)-liue are probably more nearly accurate than those for the less luminous thallium flume j the molecular refractions, DlBETWEEN t r NON-RACEMIC ". AND RACEMIC " LIQUIDS. 1115 N-niethylpiperidiiie ............ N-iiieth yite traliydroquinoliiie. U- 9 , ( I ~ ~ v o - ) a- 9 7 (inactive). U- ............ 9 , ... for the D-line are identical within the limits of experimental error. With the identity in density between the active and inactive bases there goes therefore an identity in refractive index at the same temperature and an identity in molecular refraction.By applying Traube's arguments (Ber., 1897, 30, 43) on the relation between the refraction constants and the association factor to the case of the tetra- hydroquinaldines, it is seen that the association factors of the active and inactive bases are rigorously identical. A perusal of Briihl's magnificent work on the refraction constants of nitrogen compounds furnishes rnaterial for at1 interesting comparison, I n the appended table, we quote the vslues obtained by Briihl for the molecular refractions of N- and a-inethylpiperidine (Zeit. plqsiknl. Chem., 1 S95, 16, 222) and for ~~~-metliyltetrr~hydroquinoline (Zeit.physikal. Chem., 1897, 22, 395) together with the values which we now give for 31 -74 31-32 48.02 47'60 47.57 &LJ. I Snbstaiice. Preparation. A .............................. €3 .............................. t. a u. [alD. [ 5 1 3 D . 20" - 50 24" 59.12" - 85.44" 9 , - 59'32 - 5s 2 0 - 85 -56 A. 0'42 0 *52 0 *P5 Observer. ISriilil. W. J. P. :nd S. J. P. ,¶ 9 9 9 ) the active and inactive u-inethyltetrshydroquinoline and the differ- ences, A, between the molecular refrnctions for the D-line of the N- and a-isomerides of both series. Attcution should be drawn to the appre- ciable identity of the diiferences, A, between the molecular refractions of the secondary and tertiiiry polymethylenic bases. notation Coizstmtts of ~~votetrnl~;ydroqzii~LaZcli~te. The rotatory powers of the prepsmtioiis A.and B of lxvotetrahydro- qtiinsldine described above were determined in 100 mm. tubes for the D-line, and are very nearly the mine. The difference of O*OSo observed in a rotatory power of about 60° is probably mainly due t o the impossibility of obtaining a pure D-line even by the use of a long dichromate tube as a screen, It is to be noted that our value for aD a t 203, namely, 59*28O, is rather higher1116 than that given by Ladenburg (Bey,, 1894,27, 77) for the dextro-base, namely, aD + 58035~. Since it is certain that laevotetrahydroquinaldine is highly associated, we were not surprised to find that its specific rotatory power varies very considerably with the solvent. The rotations given in the following table refer to the preparation A, and the values of uD were obtained in 200 mm.tubes, except in the case of the pure base, when a 100 mm. tube was used. POPE AND PEACBEY: A METHOD FOR DISCRIhflNATIN~ Solvent, Piperidine ................... Ether ........................ None ......................... Acetone.. ..................... N-Methyl tetrahydro- yuinoline., ................ Ethylic alcohol ............. Methylic alcohol .......... Chloroform ................ tknzene ...................... Carbon tetrachloride . . , . , Acetic acid .................. t. 23 -8 21 +o 20 *o 19.5 19'0 21'5 19'5 21 *5 21.0 21 *5 21 *o 20. 1.2382 0'6333 0.6415 1 ,8957 0'6454 0*6398 0.6514 0.6465 0.6410 0.6672 - - t'. 15.00 25-34 25.24 10.10 25 '20 25.27 25-20 23.24 25 *29 25'29 - - i.58" - 2.55 - 59 2 4 - 3 '22 - 11 '91 - 3.28 - 3'50 - 4-43 - 4 *54 - 4.95 - 6-22 - 45'9" - 50.8 - 58-12 - 63'3 - 63.6 - 64.0 - 75.1 - 85.3 - 88 *6 - 97 *6 - 117'9 - 6i.5" - 74.7 - 93.1 - 93.5 - 94.1 - 110.4 - 125.4 - 130.3 - 143'5 - 173.3 - 85-44 A perusal of this table shows that as the solvent is changed the specific rotatory power of lawotetrahydroquinaldine alters from [a]= - 46.9' to [.ID - 117*9O, and is therefore nearly three times as great in one solvent as in another ; even disregarding the use of acetic acid as a solvent because of the chemical action which doubtless occurs, the fact remains that the specific rotatory power of the base in piperidine solution is less than one-half of what it is in carbon tetra- chloride solution.These large variations in specific rotatory power with change of solvent can only be attributed t o differences in the degree of association of the base in the various solutions.Corroborative evidence of this can be obtained in two ways, by making use of the highly probable supposition that if an associated substance be dissolved in a solvent of identical association factor, the association factor of the solute will remain unaltered. We see from the table that on dissolving Izvotetrahyclroquinaldine in its isomeride, N-methyltetrahydroquinoline, the specific rotatory power increases by some 9 per cent. ; although N-methyltetrahydroquinoline is isomeric with lsevotetrahydroquinaldine, the latter is a secondary and the former a tertiary base, so that, structurally, the two substances differ considerably, and are not likely to have similar association factors.Tetrahydroquinoline is, however, the next lower homologue of IEVO- tetrahydroquinaldine, and both are secondary bases in every wayBETWEEN (' NON-RACEMIC " AND " RACEMIC " LIQUIDS, 11 17 Solvent. -~ Tetrahjdroquinoline ...... closely resembling each other ; it is therefore, d p i o v i , highly probable that they have almost the same association factor. Further, it is t o be noted that tetrahydroquinoline contains no asymmetric carbon atom. The rotation constants of the optically active base were determined in the lower homologue as solvent with the following results : aD in t. ZU. v* 1 ,Icm. ------ 24.2" 1.0647 15.0 -4'18O - 58.9" - 86.6' t. W W lmo-base. inac. base. 2.8503' 7.3455 20'2" I n accordance with our expectations, the specific rotatory power of the base in this solvent ( - 58.9") is almost identical with the specific rotatory power in the pure liquid state ( - 58-12'), because the as- sociation factor remains almost unchanged.A gain, the determinations of the densities and refraction constants of laevo- and externally compensated tetrahydroquinaldine indicate with great probability that the association factor is the same in both, If any combination existed between the dextro- and laevo-isomerides in the externally compensated base this mould be impossible. If the optical antipodes are quite indifferent one to the other in the mixture, we should find the specific rotatory power of the lsvo-base, dissolved in the externally compensated mixture as solvent, absolutely identical with the specific rotation of the pure liquid laevo-base.If, however, the two antipodes are not quite mutually indifferent, the association factor mould change on admixture and hvotetrahydroquinaldine could not have the same specific rotatory power when dissolved in the externally compensated base as solvent as when solvent-free. The following determination made with the mixture used in the density determina- tions, shows in the most conclusive manner possible that the former alternative is the true one : dtw. 1.01916 The specific rotatory power of lsvotetrahydroquinaldine, peg* ue, is [ a ] D - 5S.12", whilst when dissolved in the externally compensated base the specific rotatory power has an identical value, namely, [ a ] D - 58.02O.It remains to be added that externally compensated tetrahydroquin- aldine is the best example of a physically perfect liquid solution to VOL. LXXV. 4 FI l l 8 '' NON-RACEMIC " AND '' RACEMIC '' LIQUIDB. C. 2 4 8 10 12 20 80 I which attention has yet been drawn-of a solution, namely, the physical properties of which are proportional means of those of the constituents. Although it mould appear that the variations in specific rotatory power of a substance dissolved in various chemically inert solvents are due mainly to changes in the associatiou factor of the solute, this does not preclude the solvent from exerting a specific action upon the rota- tion constants quite apart from its influenco upon the association factor. It would seem, however, that such a specific action is not exerted upon an active substance by the mixture of the two antipodes; this is evident from the fact that the specific rotatory powers of laevotetra- hydroquinaldine and of lievopinene respectively are the same in the solvent free state as in solutions in which the externally compensated substance is used as solvent.{a], a t 21 '2. C. [a], a t 21.2". - 36-85" 40 - 36 '97" - 36.87 60 - 36 '98 - 36.97 60 36-96 - 37 *oo 70 L 37.00 - 36.95 80 - 36-97 - 36 *98 90 - 3694 -36.99 100 - 36 '97 Solvent, Methylic alcohol ............ Ethylic alcohol.. ............. Ether ........................... Piperidine ..................... . Benzene ........................ Ace tone ....................... Etliylic acetate ............... Carbon tetrachloride ...... Acetic acid ................... Chloroform.. ................. - 33'3" - 34.8 - 34'9 - 35'5 - 37'5 - 38.8 - 39.2 - 39.5 - 41 *7 - 41 *5 - 37 *3" - 38.6 - 37 * 5 - 37.6 - 39.1 - 3 9 5 - 39.6 - 39.3 - 40.0 - 41.7 - 38.3" - 38.7 - 38.1 - 37 *6 - 39.2 - 39.4 - 39.4 - 40.7 - 40.4 - 42'0THE CHARAC'PERISA~ION OF " RACEMIC " LIQUIDS. 1119 aiderably affected by experimental error. The table shows that the specific rotatory power of lsvopinene varies in different solvents, but to a minute extent compared with the variation in specific rotatory power of laevotetrahydroquinaldine in different solvents. A series of determinations of the specific rotatory power of hvopinene dissolved in a carefully purified sample of externally compenmted pinene was also made, and the results, given in the last table, contrast very strongly with those obtained with solvents other than inactive pinene, The results obtained with the pinenes point to the same conclusion as do those with tetrabydroquinaldine, namely, that the two optical antipodes, when mixed together, are wholly without mutual influence, and that consequently there is no question of the two substances combining to form cz racemic compound. Our thanks are due to the Government Grant Committee of the Royal Society and to the Research Fund Committee of the Chemical Society for grants defraying the cost of apparatus and materials used in the above work. GOLDSMITHS' INSTITUTE, NEW CROSS.