A RiAGNETIC STUDY OF COMPOUNDS OF WATER, ETC. 2’70’7CCLTI1.-A Magnetic Study of Compounds of Waterand of Aqueous Solutions.By FRANCIS WILLEAM GRAY and WILLIAM MILNE BIRSE.THE object of the work described in the present paper was toascertain whet her magnetic measurements can throw any light onthe state of ccrnbination of water in compounds of different types,and especially t o measure the magnetic properties of water in(1) aqueous salt solfitions, (2) hydrated crystals such as those ofcopper sulphate, and (3) organic acids, such as benzoic andphthalic acids, which may be regarded as compounds of theiranhydrides with water2708 GRAY AND BIRSE: A MAGNETIC STUDY OFI n all these classes instances were found in which tho law ofadditivity, in the molecular sense, is obeyed.Thus aqueous solu-tions of potassium ferricyanide obey the law of additivity through-out the whole range of concentration, and yield a more trustworthyvalue f o r the susceptibility of potassium ferricyanide than thatobtained from the solid.I n the case of copper sulphate, if it is assumed that the suscepti-bility of the water is not affected appreciably by the union, theni t is found that the paramagnetic susceptibility of the anhydrouscopper sulphate molecule is increased by about 11.5 per cent. whenit unites with one molecule of water. Further addition of watermolecules to form the higher hydrate has no marked influence onthe susceptibility.-%I n the case of organic acids it is found that additivity, in themolecular sense, holds for benzoic, phthalic, maleic, and fumaricacids, but not for succinic and camphoric acids.Aqueous solutions of potassium ferricyanide obey (see table I)X very nearly the equation -- - .2/ = 1, where z=the weight of10.0'7 0.72potassium f erricyanide in 100 grams of aqueous solution, andy x denotes the susceptibility of a solution of percentage x.It is not usual for an aque'ous solution to follow so closely the addi-tive law.Indeed, many of the older determinatJons of the sus-ceptibility of salts are quite valueless, since they were calculatedfrom determinations of solutions of single concentrations on thebasis of additivity, and the value obtained in this way variesusually according to the concentration of the solution. Cabreraand Moles (Arch, Sci.phys. m t . , 1913, [iv], 35, 425) have shownhow the atomic susceptibility of iron varies with the concentrationin solut'ions of ferric chloride, ferric nitrate, and sodium ferricp yrophospha t c.l n the magnetic study of solutions the following effects are to belooked f o r : (1) ionisation, (2) union of two or more molecularmagnets t o yield an astatic system not oriented in a magnetic field,(3) formation of hydrates and the stability of the same, (4) hydro-lysis. I n addition, care must be taken to ascertain whether or notthe susceptibility of a solution changes with time. Instances havebeen found of solutions originally additive which showed a gradualdeparture from additivity. Heydweiller (Ber. Deut. physikal.Ges.,1913, 15, 112) gives results for solutions of ferric chloride, man-ganese sulphate, and nitrate, nickel nitrabe, chromic sulphate,chromic nitrate, and cobalt nitrate. H e observed a maximum inthe curves for the relation between concentration and molecular* It is well known that similar differeuces have been observed in the specific heatsand heats of hydration for the different water molecules in polyhydratesCOMPOUNDS OF WATER AND OF AQUEOUS SOLUTIONS. 2'709susceptibility. This maximum may be produced by the joint actionof effect ( l ) , which causes increase of susceptibility, and effeet (a),which causes diminution of susceptibility. Oxley discusses effect (3)(Proc. Camb. Phil. SOC., 1912, 16, 421), and points out that thehydrate may be so unstable as not to affect the magnetic propertiesa t all.Wiedemann claims from magnetic measure'ments to be ablet o calculata the degree of hydrolysis of ferric chloride in aqueoussolutions. I n aqueous solutions of pchassiurn ferricyanide whichseem to obey the additive law none of the above effects can bedetected by the present method.X Using -- ' - 1 and putting x=lOO, we obtain for the sus-10.07 O--ceptibility of potassium ferricyanide t'he value + 6.43 x 10-6. Thisgives 10-97, or very nearly 11 magnetons per molecule. Weiss(Compt. rend., 1911, 152, 367) gives 10.41 magnetons.The susceptibility of solid potassium ferricyanide (powder) wefound to be + 6-77 x 10-6, the error-range being 50.16, or about-t2-2 per cent. (when calculated by the average deviation method).This figure, 6.77, gives 11-26 magnetons per molecule.Ihde (Ann.Physik, 1913, [iv], 41, 829) points out, in the caseof paramagnetic powders, that the molecules a t the surface of thegrains are most easily oriented in the magnetic field. An increasein the size of the particla causes a diminution in the total surface,and therefore in the number of surface molecules, and thus a fallin the susceptibility. This was found to hold f o r powdered potass-ium ferricyanide.On the other hand, the increased density which accompaniesincreased size of particles tends to an increased value for thesusceptibility.On the whole, therefore, with potassium ferricyanide, the sus-ceptIbility obtained from solutions is more trustworthy thal; thatfrom powders, because, with solutions the precision is much better,the law of additivity is obeyed throughout the whole range ofconcentrations, and, further, the result implies an integral numberof magnetons per molecule.I n order to account f o r the magnetic difference between coppersulphate monohydrate and copper sulphate pentahydrate, wesuggest the following hypothesis : that the water-molecules arearranged in space round the outside of the copper sulphate molecule,one in the vicinity of each oxygen atom and one in the vicinityof the copper atom.The last-mentioned water-molecule is the onlyone that causes deviation from the additivity of the magneticproperties. When two copper atoms of two anhydrous coppersulphate molecules are near one another they hamper one another'smovement in the magnetic field.Thus the orientation of the copperFOL. cv. 8 2710 GRAY AND BIRSE: A MAGNETIC STUDY OFatom is not so free in the a-nhydride as in the monohydrate o r inthe pentahydrate, in both of which water-molecules intervenebetween the copper atoms, keeping them apart and thus preventingthe mutual action above referred to. In other words, the para-magnetic susceptibility of the copper atom is less in the, anhydridethan in the monohydrate or in the pentahydrate.This theory receives support from the fact that similar hypotheseshave served to explain two observations recently made in the cryo-genic laboratory a t Leiden. Perrier and Onnes (Compt. rend.,1914, 158, 941) studied mixtures of liquid oxygen and liquidnitrogen, and found that the coefficient of magnetic susceptibilityof liquid oxygen increases as the concentration diminishes.Again,Onnes and Oosterhuis (Proc. IT. A Lad. Wetensch., Amsterdam,1913, 15, 969) in studying paramagnetism a t low temperaturesfound for hydrates of salts and anhydrous salts, in the case offerrous sulphate and manganese sulphate, that whilst the hydrateobeyed Curie’s law, xT=constant, down to the temperatares ofliquid nitrogen, the anhydrous salt followed the law, x(T+A)=constant, where x = specific susceptibility, T = absolute temperature,and A=a constant. Thus, a t any given temperature withiu acertain range the paramagnetism is increased by the union ofwater with the salt.Mlle, Feytis (Compt.rend., 1911, 153, 668) about the same timeas we made our observations obtained similar results f o r coppersulphate (see experimental part).Mlle. Feytis (Zoc. cit. and Compt. rend., 1913, 156, 886) foundfor the salts CuC12,2H20, CuC1,,2NH4C1,2B,0, CuC12,2KC1,2H20,and NiS04,6H20 departure from additivity in the same sense as incopper sulphate. On the other hand, she observed that additivityheld for the salta CoS0,,7H20, Cr,(S04),,16.74H20, andbut not for CrCl,,6H20, for which the departure was in a directionopposite to that for copper sulphate. This last case may be broughtinto line with our hypothesis by supposing that in anhydrouschromic chloride the chlorine atoms are arranged symmetricallyround the chromium atom, chlorine atoms keeping apart thechromium atoms of different salt molecules, and thus enhancing theparamagnetism.When water unites with the anhydrous moleculewe suppose that there is no longer the symmetry referred to, andthat chromium atoms can come nearer one another than before,and thus the atomic paramagnetism of the chromium is diminished.The hydrates of chromic chloride are represented thus :K2so,,cT2 (S04)&H@,[Cr(H20),]Cls (violet) and [CrC12(H20),]CI + 2H20 (green).Tha difference between these is not shown in m’agnetic measure-mentsCOMPOUNDS OF WATER AND OF AQUEOUS SOLUTIONS. 2'711We have observed departure from additivity in tlhe hydrates ofdiamagnetic salts also, and from our own results and those of otherswe have been led t o the general rule, that when there is departurefrom additivity a paramagnetic anhydride has its paramagnetismincreased and a diamagnetic anhydride has its diamagnetismdiminished by the union with water, on the assumption that thesusceptibility of the water is not affected appreciably by the union.I n a paramagnetic substance the always present diamagnetism ismasked by the larger paramagnetism.According to our generalrule, the diamagnetism in a paramagnetic substance might bediminished by union with water (when there is departure fromadditivity), and thus the apparent paramagnetism would beenhanced. We believe, however, that the departure from additivityin paramagnetic salts can be only partly explained by this cause,and that the hypothesis suggested under copper sulphate, or asubstitutle, is still required.For copper sulphate pentahydrate the theory might be broughtforward that the water of crystallisation is made up of two dihydrolmolecules and one monohydrol molecule, or one trihydrol moleculeand two monohydrol molecules, or on0 trihydrol molecule and onedihydrol molecule, or one dihydrol molecule and three monohydrolmolecules, or five monohydrol molecules.The first of these fivepossibilities is best suited for explaining how one water-moleculediffers from the other four. We think, however, that this wouldaccount for only a very small magnetic difference, judging fromthe results of Piccard (Compt. rend., 1912, 155, 1497), who studiedthe susceptibility o i water a t various temperatures from Oo t o looo.The susceptibility of water is only 0.75 per cent.greater a t looothan a t Oo; also the decreasa on solidification is 2.4 per cent., sothat variations in the proportions of trihydrol, dihydrol, and mono-hydro1 has very little eff ect-on the susceptibility.Similarly the results of Piccard do not encourage us to suppoaethat departure from additivity in a hydrate is due to any appre-ciable extent t o any change in the magnetic susceptibility of thewater, that is, to change caused by the union with the anhydride.The water which is present in a paramagnetic metallic hydroxideis usually regarded as a clear case of water of constitution, andwhen water unites with the oxide to form the hydroxide there isoften a considerable enhancing of the paramagnetic susceptibilityof the oxide molecule. Thus, there are hydrates and hydroxidesin which the magnetic r6le of water is identical, and it becomesan interesting question whether we can extend the hypothesis wegave for hydrates t o the case of metallic hydroxides.With regard to the organic acids (see tables I11 and IV) theparallelism between constitutional and magnetic similarities an2712 GRAY AND BIRSE: A MAGNETIC STUDY OFdifferences is interesting.In every case except the two acids whichshow decided depart.ure from additivity (succinic and camphoricacids), the anhydride is obtained from the interaction of twocarboxyl groups which are either near one another in the samemolecule or are in different molecules, so that interaction can takeplace without any great change in the configuration of the atomsand their electrons.On the other hand, with succinic acid there isa marked change in the relative position of the atoms when the twocarboxyl groups a t the ends of the open chain interact t’o form acyclic compound, and similarly with camphoric acid when twocarboxyl groups attached to two non-adjacent carbon atoms in thecamphoceanic ring interact to give the anhydride.In comparing our rwult8s with Pascal’s it should be noted thatwe use for the molecular susceptibility of water the experimentalvalue - 12.96 x 10-6, whilst the sum of Pascal’s atomic values is- 10.46 x In calculating the molecular susceptibility of anorganic compound, however, Pascal introduces corrections for con-stitution, so that the two methods are not necessarily inconsistent.A t any rate, our experimental figures and Pascal’s calculated valuesagree for furnaric acid and maleic acid and benzene; also, accord-ing to our method, additivity holds for these three substances.EXPERIMENTAL.The rezent concordant result;s of de Haas and Drapier (1913),of Wciss and Piccard (1913), and of S&ve (1912) yield for thespecific susceptibility of water reduced to a vacuum the value-0.72 x 10-6, which we u3e here in preference to the value-0.75 x 10-6.formerly used by Pascal, and the still older valueof Curie, namely, -0.79 x 10-0.All our results were obtained with a CurieChBneveau magneticbalance except number 8 of table 111, for which a Pascal balancewas used.The permanent magnet of the Curie-ChBneveau balancehad an average field of 232 gauss per sq. cm. over an area of5.6 sq. cm. round and a t right angles to the axis. Platinumtorsion wirea were used about 33 cm. long and of diameter 0.15mm., 0.10 mm., or 0.07 mm., according to the requirements. Thescale was placed a t a distance of more than 2 metres from themirror. The greatest precisian was obtained with a pure liquid, as,for example, with benzene, as shown in table 111. All the solidsin table I11 were in the form of powder, and the precision isusually not so great as with benzene. In some cases we improvedthe precision by heating the anhydride both before and after it wasin the tube, or by leaving the filled tube in a vacuum desiccator forsome time before the determination.With a glass tube the heatingmight cause volcme changes resulting in error, and it occurred t COMPOUNDS OF WATER AND OF AQUEOUS SOLUTIONS. 2'713us to t4ry a quartz tube, ?Ve found, however, that little, if any-thing, was gained by its use, as the susceptibility of the quartz wasmuch greater than that of the glass we used, so that a degree ofuncertainty was introduced which perhaps more than balancedany advantage obtained froni the constancy of volume of the quartz.I n the case of the substances in table IV we sometimes foundthat purification improved the precision, even when the method ofpurification produced no change in the melting point.Result number 8 in table I11 was obtained with a Pascal balance,the field of the electromagnet having an average of about 9000gauss per sq.cm. over an area of 2-21 sq. cm. a t right angles toand round the axis.The determinations were carried out at 1 5 O .I n the following numerical results we have given as many digitsas we obtained in our calculation. The precision of the estimationindicahs how many digits should be retained in each case.Magnetic Susceptibility of Aqueous Solutions of PotassiumFerricyanide.TABLE I .Numberletter or measure-Reference ofnumber. ments. z. Y- Y1. 4. yy d?.A 3 29.13 +1.394 3.1.362 +0-032 +1.354 +0*040B 2 26.22 +1*188 +1*155 +0.033 +1.148 +0*040C 3 23.53 +0.993 +0*962 +0*031 $0.956 +0-037S 2 20.35 +0.731 +0.735 -0.004 f0.731 0.0001 2 17-32 $0.513 SO.518 -0.005 +Om515 -0.0022 2 15.55 +0.391 +0.392 -0.001 +0.390 +0.0013 3 12.37 +0.162 +0.164 -0.002 +0*163 -0.0014 1 11.08 3.0'0675 +0.0722 -0.0047 +0*0718 -0*00435 2 8.27 -0.123 -0.129 +0.006 -0.128 +0*0056 1 7.416 -0.178 -0.189 $0.011 -0.188 +O.OlO7 4 6.005 -0.286 -0.291 +0.005 -0.289 +0.0038 4 4.594 -0.383 -0.392 +0.009 -0.390 +0*0079 6 2.976 -0.506 -0.507 $0-001 -0.504 -0*002In table I, x denotes the weight of potassium ferricyanide in100 grams of aqueous solution, y x 1 0 - 6 denotes the susceptibilityfound by experiment for a solution of percentage x.y1 is thez Y value of y as obtainsd from 1m - o% = 1 . d, denotes the distancein the direction of y that the: experimental point is above or belowthis straight line; + means above, - means below.Under ys andX d2 are given corresponding values obtained from ~- -!!--- -lO.07-O*716 - ''Neglecting the three solutions A , B, and C , we note that the firststraight line fits the experimental points very closely, and that thesecond fine fits still closer, but does not pass through the water-point. However, taking into account the degree of uncertainty o2714 OKAY AND BIRSE: A MAGNETIC STUDY OFthe measurements and seeing how near the lines lie tot one another,we need not push the refinement of the calculation so far, and maybe content nrith the first and simpler equation.Of the solutions, A was the most concentrated we could con-veniently use, and its composition was found by chemical analysis.By adding a weighed amount of ,4 t o a weighed amount of water,B and C were obtained.S was.prepared by adding a weighedamount of water to a weighed amount of pure potassium ferri-cyanide. A stock of S was prepared, froin which, by the weighingmethod, the series of solutions 1 . . . . 9 were obtained. Thusany error in the determination of the composition of A will affectB and G, and any error in the preparation of S will affect the set1 t o . . . . 9. The two sets are, of course, independent of oneanother.This explains why the three points obtained with A , B, and Cdeviate further from the straight line than any of the points of theset 8, 1, . . . . 9. It is probable that the analysis of A bychemical means is not so accurate' as the preparation of 8 by theweighing method.The degree of the uncertainty of the magnetic measurementsof the solutions will be seen from solution 9, for which the averagedeviation from the mean was 0.0038 or 0.76 per cent.TABLE 11.AverageSpecific Susceptibility deviation x 106.from mean, Molecular NumberSubstance. values. values. lute. cent. x 10'). ments.,-- suscepti- ofCUSO, ............... 3-8-6 8.39 0.02 0.2 -+ 1339 3CUSO,,H,O ......... +8.6 8.32 0.04 0.5 +1479 3CuS0,,5HL0 ...... +5*9 5-81 0.03 0.5 3.1450 4Mlle. Feytis' Our Abso- Per bility measure-Molecular susceptibility of CuSO,,&O= + 1479 XCorrection for water= - 13 x lo-"Susceptibility of the molecule CuSO, in CuSO,,H,O = + 1492 X lo-"Molecular susceptibility of anhydrous CuSO,= + 1339 X lo-';Difference= 153 x lo-"Thus, the union with one molecule of water has increased theparamagnetic susceptibility of the anhydrous copper sulphate rnole-cule by + 153 x 10-6 or by about 11.5 per cent.Further additionof water molecules to form the higher hydrate has no marked influ-ence on the susceptibility.Otherwise :Deviation fromMolecular susceptibility x 10';. additivity. -- -- Substance. 4 x emmental. Calculated. Absolute. Per cent.CuSO,,H,O ... 4- 1479 + 1326 153 11(from CuSO, and -0)CuSO,, 5X20 ... + 1450 + 1427 23 1.5(from CuSO,,H,O and S O COMPOUNDS OF WATER AND OF AQUEOUS SOLUTIONS. 2715Oopper Nitropusside.Average deviationSpecific from mean Number Molecular Deviatiorisuscep- & of suscep- fromtibility Abso- Per measure- tibility additivityx lW.lute. cent. ments. x106. per cent.Cu(NO)Fe(CN),,2H20 +4.54 0.05 1.0 5 +1432 8.4Cu(NO)Fe(CN), ...... +5-73 0.03 0.5 5 +1593 -This would indicate that the deviation is opposite in direction t othat for copper sulphate. However, any slight decomposition inthe preparation would make the anhydride too highly paramag-netic. A general study of nitroprusiides, on which we are a tpresent engaged, may throw more light on this question.Some Diamagnetic Salts.Potassium ferrocyanide :Molecular susceptibility of K,Fe(CN),,3H90 = - 172.0 x 10 riCorrection for 3H2O = - 38.9Susceptibility of the molecule K,Fe(CN), in K4Fe(CN),,3&0= - 133.1Molecular susceptibility of anhydrous K,Fe(CN),= - 145.1Difference= 12.By union with water thc diamagnetism of the moleculeK,Fe(CN), is diminished by about: 9 per cent.The following figures were calculated from the results of St.Meyer.Loss per cent. means the percentage by which the diamag-netic susceptibility of the anhydrous salt molecule is reduced byunion with water to form the given hydrate.Hydrate .................. MgCL,,6&0. CaCL,, 6H,O. BaCh, 2H20.Loss per cent ............ over 100 over 100 25Loss per cent ............. over 100 24 over 100masked by the diamagnetisni of the water.Hydrate .................. MgS0,,7H20. Li2S0,,H,0. Na,CO,,lOH,OThis would imply that in these salts we have paramagnetismTABLE 111.Organic Acids.SpecificReference susceptibilitynumber.Substance. x 106.1. [Benzene] ...... -0.70862. Benzoic acid ... -0.55603. Benzoicanhydride ... -0.55224. Phthalic acid ... -0.48786. Phthalicanhydride ... -0.44606. Maleic acid ...... -0.42697. Fumario wid ... -0.41588. Fummic acid ... -0.42699. Maleiotdlydride ... -0.3656Average deviationfrom mean.\Absolute. Per cent.0.0049 0.70.0123 2.20.0060 1.10.0054 1.10.0048 1 e.10.0182 4.00.0043 1.00.0096 2.20.0056 1.5Numberofmeasure-ments.566768452716 A MAGNETIC STUDY OF COMPOUNDS OF WATER, ETC.T.~BLE TJT. (continued).SpecificReference susceptibility10. Succinic acid ... -0-46151 1. Succinicanhydride ... -0.475312. Camphoricacid ............ -0.748113. Camphoricanhydride ... - 0.6204number. Substance. x loc.Average deviationfrom mean.Absolute. Per cent.0.0016 0.30.01 11 5.30.0054 0-s0.0172 2.7,TABLE IV.Molecular susceptibility x loq. Deviation *\ fromReferencenumber. Substance.2. Benzoic acid ...:I. Benzoicanhydride ...4. Phthalic acid ...5. Phthalicanhydride ...G. Maleic acid ...7. Fumaric acid ...8. Fumaric acid . .9. Maleicanhydride ...10. Succinic acid ...11. Succinicanhydride ...12. Camphoric13. Carnphoricanhydride ...acid ............Experi-mental.- 67.78- 62.37 x 2- 81.36- 66.01-48.23-49.62- 36.81- 64.46- 47.63- 149.3- 112.3- 49-62Forndditivity.- 68-85-- 78.97--48.77 - 48.77-#8*77-- 60.49-- 126.3-additivity.Per cent,.1.5-2.6-1.61.01.5-9.9-19.0-Nunilwrofmeasure -in~nts.554c,Pascal.70.63-86-0349-8749.6749.67--57.48--IMolecularsusceptibility Pas4 givesBenzene ......................... - 56.27 - 55.1 (Expt.) -- 55-03 (Calc.)Phthalic acid .................. - 81.36Mean ................... - 68.31ncnzoic acid .................. - 67.78Thus when t h molecular susceptibility of benzoic acid is calcu-lated from benzene and phthalic acid on the basis of additivity,then the experimental result differs from this calculated value byless than 1 per cent.Under the heading ‘‘ Pascal ” we give the molecular suscepti5li-ties as calculated from Pascal’s atomic values and his correctiohs forconstitution.~’HYSICAL CZIICMISTRT DEPARTMENT,ABERDEEN UNIVERSITY