ANNUAL REPORTSON THEPROGRESS OF CHEMISTRY.GENERAL AND PHYSICAL CHEMISTRY.1. RECENT ADVANCES IN INFRA-RED SPECTROSCOPY.THE past ten years have seen a remarkable advance in infra-red spectroscopy,particularly in its application to chemical problems. Earlier work, summar-ised in previous reports,l was somewhat physical in character, and centredmainly on the determination of the vibration frequencies and moments ofinertia of simple molecules, together with the exploration of underlyingspectroscopic theory and the intricacies of molecular dynamics.2 It wasoften possible to derive from the results other information bearing upon themolecular structure or thermodynamic properties of the substances involved,and with relatively simple molecules containing not more than five or sixatoms highly accurate structural data could thus be obtained.The scopeof such work was obviously limited, however, since the large majority ofmolecules are too heavy t o give vibrational bands which have a resolvablerotational fine structure, and the molecular complexity and general lack ofsymmetry make it impossible to assign the vibration frequencies completelyand unambiguously.The possibility of other kinds of chemical application was foreshadowedby the correlation of certain absorption bands with particular atomic group-ings, especially those measurements on hydroxylic compounds in which thedisplacement of the characteristic vibration frequency of the O-H link wascorrelated with the existence of hydrogen bonds. It was clear, however,that an essentially different experimental approach would be necessary ifthe bigger molecules generally encountered in chemistry were to be studiedwith advantage, and that this might involve as a first step the empiricalcorrelation of the spectra of a large number of molecules in carefully chosenrelated series.Ten years ago the compilation of such a background ofreference seemed impossible, mainly because of the technical difficulties ofmeasuring the spectra with reasonable speed. Using instruments of greaterresolving power, such as grating spectrometers, progress was even slower.These problems, accentuated by the needs of the war, have led to a numberof striking technical developments, as a result of which it is now possibleto measure the spectra rapidly, using instruments which require neitherAnn.Reports, 1941, 38, 46; 1938, 35, 37; 1936, 33, 53; 1935, 32, 53.a G. Herzberg, “ Infra-red and Raman Spectra,” van Nostrand, 19456 GENERAL AND PHYSICAL CHEMISTRY.elaborate laboratory facilities nor particular operational skill.3 The maincause of this progress has been the vast improvement in the methods ofdetecting infra-red radiation? although many new appropriate accessorieshave simultaneously become available. As a result, the emphasis haspassed rapidly to a wide range of chemical application^,^ many of which canbe classified as qualitative or quantitative analysis, or as structural diagnosis.Further advances on the instrumental side are certain to be made in thenear future, so that infra-red spectroscopy now ranks among the otheruseful standard physicochemical techniques.I n order to obtain a correct perspective, as well as to appreciate theparticular uses of different kinds of instrument, the experimental develop-ments must be outlined more specifically and a t rather greater length thanis usual in a report of this kind.The main difficulty in earlier work lay inthe detection and accurate measurement of the small amount of radiationavailable a t these longer wave-lengths ; especially when radiation from aaource was first spread out by some dispersing system. The most commonmethod employed a thermocouple, working with a sensitive galvanometer.The comparative insensitivity of the couples available made it neccssary touse highly sensitive galvanometers of long period and capricious stability,which implied a slow and tedious measurement of the spectrum, readingsbeing taken at successive wave-lengths point by point.Amplification bythermal or photocell relays of different kinds 5 led $0 some improvementin the technique, but i t remained essentially one for a limited number ofresearch laboratories. By contrast, a wide variety of detectors is nowavailable, so far superior to the older ones that minute amounts of infra-redradiation can be detected from emitters a t great distance and with com-parative ease.A detector can be characterised by three main properties, namely, (1)sensitivity, (2) speed of response, and (3) “ blackness,” i.e., the ability toabsorb perfectly radiation of any wave-length.Three main kinds have nowbeen developed, namely, thermocouples,6 bolometers, and the so-calledselective receivers. As mentioned below, the warming of a gas consequentupon the absorption of radiation can also be used with marked success.As regards thermocouples, the use of new alloys and new methods of mount-ing in a vacuum have led to much greater sensitivity, greater speed ofN. Wright, Id. Eng. Chern. Anal., 1941, 13, 1; R. 13. Barnes, U. Liddel, andV. Z. Williams, ibid., 1943, 15, 659; R. B. Barnes, R. S. McDonald, V. Z. Williams,and R. F. Kinnaird, ibid., 1915,16, 77; H. W. Thompson and D. H. Whiffen, J., 1945,268; G. B. B. M. Sutherland and H. W. Thompson, Trans. Faraday Soc., 1945, 41,174; R.R. Brattain and 0. Beecli, J . Appl. Physics, 1942, 13, 699; R. R. Brattain,Petroleum World, Feb. 1943; W. H. Awry, J. Opt. Soc., 1941, 31, 633; E. D. McAlister.G. L. Matheson, and W. J. Sweeny, Rev. Sci. Instr., 1941, 12, 314; L. G. Smith, ibid.,1942, 13, 54; E. Lehrer and K. F. Luft, 2. techn. Physik, 1942, 23, 169.H. W. Thompson, Endeavour, Oct. 1945.6 E. B. Moss, J. Sci. Instr., 1935, 12, 141.L. C. Roess and E. N. Dacus, Rev. Sci. lnstr., 1945, 16, 164, 172; also the Hilger-Schwarz couple, Brit. PatTHOMPSON : RECENT ADVANCES IN MFRA-RED SPECmOSCOPY. 7response, and by proper compensation to almost complete stability asregards small local fluctuations of temperature. By amplifying the outputvoltage from such thermocouples by means of electro-optical devices it hasbeen possible to build spectrometers with which the spectrum between1 and 25 p can be scanned rapidly, and by coupling a recorder to the scanningmechanism to record the energy as a function of wave-length.I n this way,rapid automatically-recording prism spectrometers have been developedwith which an absorption spectrum between 1 and 25 p can be measuredin about half an hour. Moreover, in such instruments the galvanometerscan be sufficiently robust to be handled conveniently in ordinary laboratories,and the old difficulties largely disappear. I n " single-beam " spectrometersof this kind the record is usually the " fall-off " curve of the emitter, withtroughs in the background due to absorption of the substance under examin-ation.Wave-length calibration can be made either by absolute calculationsfrom the dimensions of the instrument, or by using standard absorptionlines of water vapour,' carbon dioxide, ammonia or other substances. Inorder to simplify the measurement of records some attempts have beenmade to alter the widths of the spectrometer slits continuously as thespectrum is being traversed, so as to give a flat base line on the record.This is not easy, however, to achieve in practice.The traces obtained with the above single-beam spectrometers areadequate for most practical purposes. They suffer, however, from twominor disadvantages. First, it is necessary to measure on the record thevalues of the galvanometer throws a t all wave-lengths with respect to thevalues of these throws for the source itself when there is no absorption.When a large number of spectra are being measured over a wide range ofwave-lengths, it would save much time and computation if the record gavethe percentage absorption directly, although some quick computationaldevices have been suggested.* Secondly, unless the spectrometer and ex-ternal path of the entrant beam of radiation are evacuated or cleared ofwater vapour and carbon dioxide, absorption bands occur in the blank tracedue to these substances.I n some regions these bands are so intense andcomplex that they obscure the fine details of the absorption bands of thesubstance being examined. The difficulties can be greatly reduced byevacuation of the spectrometer or by the use of drying agents, but in eithercase there are experimental inconveniences.In an attempt to obviate these shortcomings of the single-beam insfru-ments, double-beam spectrometers have been de~eloped.~ I n these, twoequal beams of radiation pass through the spectrometer simultaneously andafter emerging fall upon two separate thermocouples. One beam is used asa blank, and the other contaiiis the absorbing substance. The spectralR.A. Oetjen, C. L. Kao, and H. 31. Randall, Rev. Sci. Instr., 1942, 13, 515.H. A. Willis and A. R. Philpots, Trans. paraday SOC., 1946, 41, 187; R. M.Q J. D. Hardy and A. I. Ryer, Physical Rev., 1939,55, 1112 ; G. B. B. M. SutherlandFuoss and D. J. Mead, Rev. Sci. Instr., 1945, 16, 223.and H.W. Thompson, ref. (3); E. Lehrer and K. F. Luft, ref. (3)8 GENERAL AND PHYSICAL CHEMISTRY.record can then be obtained in two ways. Hardy and Ryer, and Lehrerand Luft, used a variable-diaphragm shutter in the blank beam, whichcould be adjusted continuously as the spectrum was being traversed so asto maintain the radiation in both beams the same. The extent of closingthe shutter could be geared to a pen and a direct reading of the absorptionthus obtained. In the method of Hardy and Ryer this procedure involvedanticipation of the motion of a galvanometer spot, and in regions wherethere are sharp absorption bands the spectrum had to be traversed veryslowly if accurate records were to be obtained. Lehrer and Luft avoidedthis difficulty to some extent by means of a photoelectric control.Analternative method is to amplify the voltages from the two thermocouplesseparately and measure their ratio on a quick-acting potentiometric recorder,the balance of which is found by a photoelectric mechanism. The potentio-metric recorder must have a rapid response, consistent with the rate ofscanning of the spectrum, and it must give a conveniently large displacementof the pen in about one second; it must also be capable of being actuatedby a current of about one milliamphe, since this may be the greatest currentwhich can be obtained by direct-current amplification of the very smallvoltages obtained from the thermocouple.These difficulties have been successfully overcome, and the double-beamspectrometers have already proved valuable in studying the absorptionspectra of molecules containing those atomic groupings which give rise toabsorption bands in the region of 3 p and 6-7 p, such as amides and amino-compounds, and hydroxy-compounds.l* There is no doubt, however, thatthe use of direct-current amplifiers of high gain with low input voltage isstill a troublesome undertaking, and complete elimination of moving-coilgalvanometers would be a further advance in building spectrometers forgeneral industrial use in locations where vibrations may be unavoidable.This could be achieved by using alternating-current amplifiers workinga t reasonable audio-frequencies.A pulsating voltage from a thermocouplecan be obtained in two ways, either by means of a periodic shutter placedin front of the entrance slit of the spectrometer, or by feeding the outputto a commutator.I n either case a large degree of amplification is thenrequired. Both methods have now been used successfully,ll but eachinvolves some difficulty. If the commutator is used, erratic contactpotentials are not easily avoided and may mask the very low voltages beingdetermined. If an interrupted beam is used, the chopping frequency mustnot exceed a value consistent with the speed of response of the thermo-couple. The latter condition at present implies the use of very lowfrequencies, say five cycles per second, since the most sensitive thermo-couples have time constants little less than one-quarter of a second.Amplification a t this frequency, for a high gain, is difficult, and troublesarise owing to various kinds of electrical noise in the circuit.If faster thermocouples could be constructed without serious loss oflo R.E. Richards, t o be published soon.l1 E.g., L. C. Roess, Rev. Sci. Instr., 1945, 16, 172THOMPSON : RECENT ADVANCES IN INFRA-RED SPECTROSCOPY. 9sensitivity, some of the difficulties might be avoided. A very promisingalternative a t present, however, is the use of bolometers which, althoughintrinsically less sensitive, have an extremely high rate of response. Theuse of a bolometer for detecting radiation depends on the fact that itsresistance alters with temperature, so that if a steady voltage is appliedacross it, periodic irradiation will produce a small fluctuating output voltage,which can be amplified.Owing to the rapid response (low time constant),higher frequency can be used for the interrupter, and the high resistanceof the bolometer makes i t possible to use a fairly conventional tunedelectronic amplifier. Earlier bolometers were of questionable value com-pared with the best thermocouples since their temperature coefficients ofresistance were small, but other kinds are now becoming available whichhave both greater sensitivity and speed of response (some milliseconds).Some of these are made from films of mixed metallic oxides, and are calledthermistors.12 Apart from their other advantages, they are robust and inuse are not affected by stray disturbances of temperature since the amplifiercan be tuned to the frequency of the interrupter, and erratic readings dueto thermal drift or vibration of a galvanometer disappear.It is certainthat the bolometer will replace the thermocouple in many infra-red instru-ments, and there are some signs that the practical limit may soon be reachedfor the detection of infra-red radiation, judged on the basis of the electricalnoise inherent in the instruments involved. It is still possible, on the otherhand, that by using the detector at very low temperatures a further im-provement may be obtained. A minor defect of the existing bolometers isthat they may not be perfectly " black " for all spectral ranges, but thereseems no reason why this cannot be achieved if suitable surface coatingsare applied.I n spite of these marked improvements leading to the rapid measurementof spectra, there are some circumstances in which an even quicker 13 scan-ning would be advantageous.Several years ago Baker and Robb attemptedto use a cathode-ray oscillograph for this purpose. This possibility has nowbeen developed further. As indicated already, an intermittent beam ofradiation passing through a spectrometer and falling upon a bolometergives rise to an alternating voltage output. The latter can be amplified,rectified, and fed to one pair of plates of a cathode-ray tube so as to causea displacement of the spot-say, vertically-and the amount of the dis-placement will be a measure of the intensity of the incident radiation, Ifthe prism is rotated and its motion is geared to a potentiometer, a timebase can be constructed for the oscillograph, so as to move the spot horizon-tally during the scanning operation.If now the screen of the oscilloscopehas a long persistence of glow, the beginning of the trace will still be visiblewhen the end of the traverse is reached, and by means of a cam a quickfly-back can be arranged, followed by a repeat, so that effectively thespectrum remains continuously on the screen. The success of this applic-l2 Bell Telephone Laboratories, e.g., B e l l Telephone Record, 1940, 19, 106.lS E. B. Baker and C. D. Robb, Reu. Sci. Instr., 1943, 14, 356, 359, 362.A 10 GENERAL AND PHYSICAL CHEMISTRY.ation hinges primarily on the use of a bolometer of rapid response, althoughthe electronic amplifying circuits are rather elaborate.In this way,however, E. F. Daly and G. B. B. M. Sutherland l4 have been able to projecta useful spectral range on to the screen, and improvements in this kind ofspectrometer may be expected soon. The instrument may revolutionisecertain types of analysis and structural diagnosis where it is desirable toobtain a quick survey of the spectrum or to estimate qualitatively or semi-quantitatively a particular component or impurity. It should also be mostuseful in observing continuously any alterations in a spectrum brought aboutby physical and chemical changes.It is well knownthat some czsium photoelectric cells are feebly sensitive to infra-red radiationjust beyond the visible.Other substances are now known to becomephoto-conductive when exposed to selected infra-red wave-lengths. Thusa layer of thallium oxysulphide l5 (" thalofide ") has a peak sensitivitynear 1 t ~ . , and lead sulphide l6 is sensitive between 2 and 3 p. If it is requiredto study the absorption of a substance having some key band within thenarrow spectral range to which the receiver is sensitive, no further spectraldispersion is needed, and by virtue of the high resistance of the cells theeffect can be amplified electronically after interrupting the beam periodicallyin the usual way. It seems likely that other similar receivers may soonbecome available for other spectral regions, and these should be particularlyvaluable for industrial applications.This progress with the various methods of detection has stimulatedimprovements in the accessories used with spectrometers, and in theexperimental methods as a whole.Synthetic crystals of the alkali halidesare now available l7 for use as prisms or windows of absorption cells, andthese are superior to natural samples as regards both transmission and size.Synthetic fluorides of calcium and lithium l8 are also available, and providea better dispersing material than rock salt for the range 2-9 p. Silverchloride sheet l9 is also available for windows of cells, since it transmitsuniformly to about 30 p and if protected with suitable reagents does notdeteriorate markedly in light. Another new material which promises tohave wide application for absorption cells and windows, and perhaps alsofor prisms in the region of very long wave-lengths, is a clear orange-colouredglassy solid obtained from a melt of mixed thallium halides.20 Althoughthis substance has a high refractive index and reflects some 20% of theradiation incident upon it, the transmission is smooth to about 50 p, andthe material is not attacked by water or atmospheric contaminants.SeveralFor some purposes, the selective receivers can be used.l4 Nature, 1946, 157, 547.1 5 T. W. Case, Physical Rev., 1920, 15, 289; J . Opt. SOC. Amer., 1922, 6, 398.l6 German development.17 The Harshaw Chemical Co., Cleveland, Ohio; H. C. Kremers, Ind. Eng. Chem.,18 N. Wright, Rev. Sci. Instr., 1944, 15, 22.l9 R. M. FUOSB, ibicE., 1945, 16, 154.1940, 32, 1478.2o German developmentTHOMPSON : RECENT ADVANCES IN INFRA-RED SPEWROSCOPY.11of the new substances just mentioned make it possible to examine very thinaqueous layers over wide ranges in the infra-red, and this should stimulatefurther work on substances such as proteins and amides. Several labora-tories have described new designs for absorption cells,21 and in view of thegrowth of quantitative analysis much attention has been paid to theaccurate control and measurement of cell thickness.22 Cells for use a thigher temperatures have also been described,23 and a range of solventssuitable for use with different kinds of solute in different spectral regionshas been examined.24 Some alternative laboratory sources of infra-redradiation have been explored,25 but the Nernst filament and Globar rodremain the most convenient.A number of reasons have made it desirable to measure the spectra ofsubstances in the solid state.With clear or semi-transparent films thisoffers no problem, but with amorphous powders much difficulty may befound owing to irregular scattering of the incident radiation. If the powdersare ground to a fine paste with paraffin or other liquids having a goodinfra-red transmission, and the particle size is controlled with respect tothe wave-lengths being studied, good spectra can be obtained.26The use of infra-red spectroscopy in analysis is based upon very simpleprin~iples.~~ A molecule of n atoms has in general (372 - 6) normalmodes of vibration. Some, or all, of these will involve a changing molecularelectric moment and be permitted to appear as fundamentals in the infra-red spectrum; many will occur as combinations or overtone bands.Thefundamental frequencies of the vibrations will depend in magnitude uponthe nuclear masses and force constants of the bonds, i.e., upon the potentialenergy function. I n consequence, no two molecules other than a pair ofoptical enantiomorphs will have the same set of frequencies, and the infra-red spectrum will be a characteristic property of the molecule-a finger-print-and can be used for its identification. I n the case of two closely alliedmolecules containing similar atomic groupings, some of the frequencies willbe the same in the two molecules, but in principle there should be some regionof the spectrum where differences occur.This is nearly always borne outby the facts, for while a few very similar molecules such as a group ofhigher homologues cannot easily be differentiated, many others differingonly slightly in structure show marked spectral dissimilarities.21 L. Gildart and N. Wright, Rev. Sci. Instr., 1941, 12, 204; E. S. Barr, ibid., p. 396.22 D. C. Smith and 33. C. Miller, J. Opt. SOC., 1944, 34, 130; G. B. B. M. Sutherlandand H. A. Willis, Trans. Paraday SOC., 1945, 41, 181; R. R. Gordon and H. Powell,J . Sci. Instr., 1945, 22, 12.z3 R. E. Richards and H. V?. Thompson, Trans. Faraday SOC., 1945, 41, 183; L. G.Smith, Rev. Sci. Instr., 1942, 13, 66.24 P. Torkington and H. W. Thompson, Trans.Paraday SOC., 1945, 41, 184.25 L. G. Smith, Rev. Sci. Instr., 1942, 13, 63.z6 R. E. Richards and H. W. Thompson, unpublished; J. Lecomte, Cahiers dePhysique, 1943, hTo. 17.27 H. W. Thompson, Analyst, 1945, 'SO, 443; G. B. B. M. Sutherland and H. W.Thompson, Trans. Faraduy SOC., 1945, 41, 19712 GENERAL AND PHYSICAL CHEMISTRY.The complete identity of the spectrum of a synthetic product with thatof a natural extract can be regarded as almost certain proof of the identityof the two samples, and this may become very important in testing synthesesof new organic and biologically interesting substances such as penicillin. Weare often more concerned, however, with the analysis of mixtures of fairlysimple substances. It is firstnecessary to know qualitatively what substances are present, or a t anyrate to be sure that no unsuspected component has absorption bands whichinterfere with or mask the key absorption bands of the component underconsideration.If independent methods give this information, the infra-redspectrum can be used to show that no other unsuspected substances arepresent. The spectrum of each of the pure substances must be known, andall the observed absorption bands must be accounted for by reference tothese. Any bands which cannot be correlated in this way will imply thepresence of an unsuspected component. It should be noted that failure todetect a small amount of a particular component in the infra-red spectrumis not necessarily proof of its absence, since in complex mixtures it mayhappen that the only strong bands of a particular component are maskedby overlapping bands of others.The sensitivity of the method dependsupon several factors and can only be assessed by reference to each individualcase. I n some cases 0.01% of a component can be detected, but in others5% might be missed.When the components of a mixture are known, quantitative analysiscan be attempted by methods based upon the theoretical absorption laws,or by empirical calibration. The key wave-lengths to be used for eachcomponent will be selected on the basis of (a) intrinsic intensity of thebands, (b) freedom from overlap with bands of other components. Incomplex cases it- is often essential to compromise between these two factors.It may sometimes be preferable to use a relatively feeble absorption bandfor determining one particular component rather than a more intense onewhich is partly overlaid by a band of some other substance, and by a correctchoice of cell thickness or concentration the percentage absorption can oftenbe brought into a range of convenient measurement.Unless particular intermolecular interactions occur, the spectrum of amixture is obtained by simple superposition of those of the components.I n terms of the standard absorption law, the extinction coefficient of asubstance can be defined by the equation,1 1 Eih = - .log T O = - . d )CiL I CiLSuch analyses usually involve two stages.where d: = log ( I , I ) .If in a mixture the optical densities are additive,dX = d: + d: + ...+ dnA= L(C1E2 + CZEB(\ + ... + C,E,h)The values of E can be determined from the spectra of the pure components.I n order to carry out an analysis for n components, it will be necessary tTHOMJ?SON : RECENT ADVANCES IN INFRA-RED SPECTROSCOPY. 13determine the optical densities at each of n wave-lengths, at each of whichthe values of E have been determined for the pure components. Thesystem of linear homogeneous equations can then be solved for the con-centrations cl, c2 ... c,. The accuracy which can be obtained will dependupon several factors, such as overlapping of bands, and care in solving thesystem of linear equations, since these may be not quite self-consistent.28A more fundamental problem, however, is to decide upon a correct measureof the optical densities, whether in fact we should take peak heights as ameasure of the intensities of some integrated band area.While there isstill some uncertainty on this point,29 experience has shown that in manycases at least the percentage absorption at the band peaks can be safelyused, unless there is some effect such as intermolecular association leadingto excessive and unsymmetrical broadening, or if the absorption band showsa rotational structure which varies abnormally with pressure or con-centration. In this work, too, allowance must be made for the absorptionby the cell itself, and by the solvent, if this is not zero. J. R. Nielsen andD. C. Smith 30 have set out the relevant mathematical formulation in somedetail.When the standard absorption laws are not applicable, empirical calibra-tion can be set up for the analysis by reference to mixtures of knowncomposition, examined under fixed experimental conditions.Many examplesof this kind have been described.31The number of examples of infra-red analysis is now large, althoughmuch of the work has remained unpublished during the war, and muchremains shrouded by industrial secrecy. The first published work on thismethod was that of W. S. Benedict, K. Morikawa, R. B. Barnes, and H. S.who analysed isotopic mixtures of the deutero-methanes and-ethanes, and perhaps the biggest general application hitherto has occurredin studying mixtures of hydrocarbons with special reference to fuels.33This field is well suited to the method, since the spectra of isomers boilingat almost the same temperature often show pronounced differences, and itis at once possible to examine fractions from distillation columns.Thedifferent isomeric octanes, for instance, have characteristic spectral features.It is worth emphasising the point that, although other methods based onphysical properties such as refractive index or density can be used for such28 R. R. Brattain, R. S. Rasmussen, and A. M. Cravath, J. Appl. Physics, 1943,2Q J. R. Nielsen, V. Thornton, and E. B. Dale, Rev. Mod. Physics, 1944, 16, 307.30 I n d . Eng. Chem. Anal., 1943, 15, 609.31 E.g., R. B. Barnes, U. Liddel, and V. 2. Williams, ref. (3).32 J. Chem. Physics, 1937, 5, 1.33 J.R. Nielsen, Oil and Gas J., Jan. 1942; R. A. Oetjen, H. M. Randall, andW. E. Anderson, Rev. Mod. Physics, 1944, 16, 260; R. A. Oetjen and H. M. Randall,ibid., p. 265; R. R. Brattain, Petroleum World, Feb. 1943; R. C. Gore and J. B.Patberg, Ind. Eng. Chern. Anal., 1941, 13, 768; L. J. Brady, ibid., 1944, 16, 422;J. Lecomte and P. Lambert, Publ. Sci. Min. de Z’Air, 1939,42; G. B. B. M. Sutherlandand H. W. Thompson, to be published shortly.14, 418; D. L. Fry, R. E. Nusbaum, and H. M. Randall, ibid., 1946, 17, 15014 GENERAL AND PHYSICAL CHEMISTRY,analyses, they may lead to serious error unless it is known with certaintywhat the components of the fractions really are. Mere boiling point is noreliable guide in such fractionations where azeotropes can arise.Theinfra-red spectrum usually characterises the components beyond doubt.The maximum number of components which can be tolerated in a mixturevaries in different cases and also according to the sensitivity and accuracyrequired. A simple routine spectrometer was originally designed 34 for theanalysis of mixtures of butane and isobutane, needed in the production ofsynthetic octanes, but it is now used for the analysis of many mixtures ofthe lower hydrocarbon gases such as arise in cracking operations, and it wasapplied to the control of butadiene production in the manufacture ofsynthetic rubber.I n the chemical industry as a whole the applications cover a wide range.The individual isomers in cresylic acid can be determined quickly,35 andthe composition of coal-tar acids in general can be examined more satis-factorily than by other methods. The analysis of the complex mixture ofstereoisomeric benzene hexachlorides formed in the manufacture of the newinsecticide ' ' Gammexane " (666) proved comparatively easy,36 particularlyfor the determination of the important y-isomer.Mixtures of nitroparaffinshave been analysed by J. R. Nielsen and D. C. Smith.37 The method hasbeen much applied to the detection of impurities in chemical products suchas the halogenated paraffins,38 or other intermediates. Other mixtureswhich have been examined include acetic anhydride-acetic acidF9 sub-stituted anilines and tol~idines,~~ t e r p e n e ~ , ~ ~ cyclohexanone and cyclo-h e ~ a n e , ~ ~ and various intermediates for the plastics industry.Mixtures ofsynthetic and natural rubber have been examined in this way.43 It isprobable that the infra-red spectrum may come to be regarded as a standardof purity for pharmaceuticals, and for organic solvents and other reagents.It is natural that infra-red analysis should also have been applied infollowing the rate of chemical reactions when other methods are less con-venient. Most applications of this kind so far concern the polymerisationof unsaturated compounds to form polymers or plastics. In such processeswe may follow either the disappearance of a band due to the unsaturatedlinkage, or the appearance of a new band due to the polymer. This method34 R. R. Brattain and 0. Beeck, J. Appl. Physics, 1942, 13, 699.36 D.H. Whiffen and H. W. Thompson, J., 1945, 268; Trans. Paraday SOC., 1945,s6 D. H. Whiffen, P. Torkington, and H. W. Thompson, ibid., p. 206; H. W.37 Ind. Eng. Chem. Anal., 1944, 16, 609.38 N. Wright, ref. (3).30 I. F. Trotter and IS. W. Thompson, Analyst, 1945, 70, 443.40 D. H. Whiffen, P. Torkington, and H. W. Thompson, Trans. Paraday SOC., 1945,*l G. B. B. M. Sutherland, ibid., p. 207.42 R. B. Barnes, U. Liddel, and V. Z. Williams, ref. (3).43 R. B. Barnes, V. Z. Williams, A. R. Davis, and P. Giesecke, I n d . Eng. Chem.41, 200.Thompson, ref. (27).41, 203.Anal., 1944, 16, 9THOMPSON : RECENT ADVANCES IN INFRA-RED SPECTROSCOPY. 15has been used in studying the polymerisation of styrene44 and in othersimilar reactions.Obviously, too, when some reaction is being followed bymeans of a single overall change of pressure, whilst in reality several processesare occurring simultaneously, the infra-red spectrum a t different stages ofthe reaction may be very informative, and its value in this connection maybe greatly enhanced if the cathode-ray tube recorder can be used to givea continuous picture of the reaction as it proceeds. It has hitherto beenusual to measure the spectra of samples taken from the reaction mixturea t successive stages of the reaction. For industrial production and for thecontrol of flow lines it would be more advantageous if the analysis couldbe made continuously and without the use of elaborate prism spectrometers,I ' I r - -I 0 I I I9 I IFIG.1. FIG. 2.or even without a dispersing system a t all. This can be done by filteringout from a beam of infra-red radiation some band which will include thekey wave-length for the substance being estimatedt5 and instruments havenow been devised for this purpose, applicable to the analysis of both gasesand liquids.Several alternative designs are possible, two of which are illustrated inFigs. 1 and 2. I n Pig. 1 N , and N , are two nichrome filaments which emitradiation through the test cell Q and through two separate tubes P and R,each of these chambers being provided with windows which transmit thewave-lengths desired. Effectively, two separate beams of radiation fall uponthe thermocouples T,,T, which can be arranged in some form of balanced44 R.B. Barnes, U. Liddel, and V. Z . Williams, ref. (3).4s A. H. Pfund, Science, 1939, 90, 326; A. H. Pfund and G. L. Gemmill, Bull.Johns Hopkins Hospital, 1940, 67, No. 1, 6116 GENERAL AND PHYSICAL CHEMZSTRY.circuit. If, for example, it is required to determine carbon dioxide in air,tube P is filled with pure carbon dioxide, R being empty. When Q is empty,the gas in P will absorb the band of carbon dioxide at 4.25 p completely,and the thermocouple circuit will now need to be rebalanced. When thetest sample containing carbon dioxide is now introduced into Q, energy willbe removed from the beam R, but owing to the already complete extinctionin P this beam will suffer no change. Hence the balance will again bedisturbed and the extent of the disturbance will give a measure of thecontent of carbon dioxide in Q.This arrangement can be improved andmade more selective by the use of filters, and bolometers can replace thethermocouples if desired. An instrument of this kind has been used byN. Wright and L. W. Herscher 4s for the analysis of streams of styrenewith ethylbenzene, or butadiene with butene.Two beams of radiationfrom the heaters N,, N2 pass through the cells A and B and fall uponchambers C , and C,, which are separated by a membrane condenser acrosswhich a direct-current voltage is applied. Cell A is maintained empty andB is for the test sample. The gas to be determined is first introduced atequal concentrations into C , and C2. Radiation will therefore be absorbedby these cells equally.If the gas to be determined now enters B, chamberC, will become less heated, and if the beams are chopped alternately by arotating shutter X , a fluctuating voltage will be supplied from the condenser,and this voltage can be increased by a tuned amplifier sufficiently to operatea recorder. The presence of other contaminants in B will not affect thedetermination of the gas in C , and C2 unless their absorption bands overlapthose of the gas being determined, since the bands of the contaminants wouldnot be absorbed by the gas in C, in any case. When overlapping of bandsdue to many components arises, it can be mitigated and the selectivity ofthe whole arrangement increased by inserting additional filters. Theoptimum conditions of working for this instrument depend upon a numberof factors which have to be carefully considered, but it promises to be verywidely used in the continuous control of gas streams.A somewhat simplerform of infra-red analyser has been described by D. J. Mead and R. M.FUOSS,~~ and used for the analysis of mixtures of liquids.The use of infra-red spectra for structural diagnosis49 is based uponslightly different principles from those used in analysis. As alreadyexplained, the magnitudes of the vibration frequencies of a molecule aredetermined by the nuclear masses and force constants, and are in realitya characteristic set for the particular molecule considered. Some of thevibrations, however, are effectively controlled by motions of atoms formingone particular link or group, and in such cases the frequencies may persistalmost unchanged in different molecules in which the group is present.A second form of detector 47 is used in Fig.2.46 J. Opt. SOC., 1946, 36, 195.48 Rev. Sci. Instr., 1945, 16, 53.49 H. W. Thompson, J., 1944, 183; R. B. Barnes, R. C. Gore, U. Liddel, and V. Z.4 7 K. F. Luft, 2. techn. Physik, 1943, 24, 97.Willianu, “ Infra-red Spectroscopes,” Rheinhold, 1944THOMPSON : RECENT ADVANCES IN INFRA-RED SPECTROSCOPY. 17With linkages involving hydrogen where the light atom oscillates against amuch heavier residue, it is not surprising that the mass of this residuehardly affects the vibration frequency, and bands characteristic of thestretching of C-H, 0-H, N-H, S-H links are well defined.Other linkagessuch as CIC, (3x0, or CZN also show fairly characteristic absorption bands,and it now seems that some larger groups of atoms forming a structuralunit may give rise to a set of frequencies which persist through series ofrelated molecules. I n order that such rules for structural correlation canbe built up, it will be necessary to measure the spectra of a large numberof compounds chosen in the first instance in related series, and much surveywork of this kind has been begun. Thus, the spectra of esters and ketones 50reveal a number of bands which can be correlated with the CH,.CO or othergroupings, and olefins of the type CR1R2:CHR, show bands in the regionof 10 which vary in position according to whether the radicals are hydrogenatoms or alkyl groups.51 The latter result can be applied to study the typesof unsaturated compound formed in the cracking of hydrocarbons. Insome cases the small variations in frequency of a particular link or groupin different molecules are determined by the other groups to which the maingroup is attached.For example, the stretching vibration band of thecarbonyl group gives rise to a band near 6 p, but its exact position dependson whether the group is present as a ketone, aldehyde, carboxylate ion, oramide, and upon the nature of the neighbouring radicals, saturated orunsaturated alkyl groups or aryl radicals.52 J. J. Fox and A. E. Martin 53have also shown how the frequencies of C-H bonds depend upon theparticular grouping in which they occur.If more examples can be foundof such electronic influences, the correlations should be extremely valuablenot only in the determination of the molecular structure of organic molecules,but also in leading to a fuller understanding of the electronic nature of thelinkages concerned.Many of the diagnostic rules have been applied to current problems inindustrial research laboratories and remain unpublished. A typical case ofan unknown impurity in a sample of adiponitrile has been quoted,S4 wherea small amount of the contaminant was diagnosed and then extracted.The method has been much used for studying macro-molecules of all kinds.556o H. W. Thompson and P. Torkington, J., 1945, 640; J. Lecomte, J . Physique,1945, 5, 1; J .Phys. Radium, 1942, 8, 196.51 M. Tuot, J. Lecomte, and S. Lorillard, Compt. rend., 1940, 211, 586; H. W.Thompson, J., 1944, 183; P. Torkington and H. W. Thompson, Proc. Roy. Xoc., 1945,A , 184, 3.52 Unpublished work of several laboratories. See also a series of papers byJ. Lecomte, Compt. rend., 1941-1945, and Bull. SOC. chirn., 1942-1944.63 Proc. Roy. SOC., 1938, A , 167, 257; 1940, A , 175, 208.64 H. W. Thompson, ref. (27).6 6 R. B. Barnes, U. Liddel, and V. Z. Williams, Ind. Eng. Chem. Anal., 1943, 15,83; W. C . Sears, J. Appl. Physics, 1941,12, 35; A. J. Wells, ibid., 1940,11, 137; H. W.Thompson and P. Torkington, Proc. Roy. SOC., 1945, A , 184, 3, 21; Trans. FaradaySoc., 1945, 41, 246; H. W. Thompson, “ Discussion on Macromolecules,’’ Chem.SOC.,to be published shortly18 GENERAL AND PHYSICAL CHEMISTRY.The spectra of hydrocarbons which were measured as reference data foranalytical work provided many correlation rules which can be used indealing with hydrocarbon-like structures such as polythene and rubber.The occurrence of methyl groups in polythene, first suggested by Fox andMartin,56 has now been confirmed by independent absorption bands.57The possibility of distinguishing spectroscopically between olefinic structuresof the types =CH:CH, and *CH:CH- makes it possible to estimate the extentto which a diene condenses with another olefin by either 1 : 2- or 1 : 4-addition. Thus when butadiene condenses to form " Buna," 1 : 4-additiongives a straight chainwhereas 1 : 2-addition will lead to pendent vinyl groups.The resultsconfirm that under different experimental conditions of polymerisationthe proportions vary. This criterion may also become useful in settlingthe old problem of the presence of isopropenyl and isopropylidene groupsin some terpene derivatives. Other hydrocarbon-type polymers which havebeen examined include polyisobutene, polystyrene, methyl rubber, hydro-rubber, and the interpolymers of butadiene with styrene and isoprene. Theeffect of stretching on the spectrum of rubber has been e~amined,~8 andpolarised infra-red radiation has been used to investigate oriented polymers.59Many other polymers and interpolymers of vinyl derivatives have beenstudied in the same way, including the acetate, acrylates, chloride, andcyanide, and differences in the spectra of many of the products can some-times be interpreted in terms of structural differences.Work with thephenolic resins has given some indications of the types of cross linkageprevalent in such products, and the preliminary results suggest that a moredetailed examination of the hydrogen-bonding relationships by means of thebands near 3 p would be profitable.60 Cellulose ethers and esters have alsobeen examined. It is possible to determine the content of the individualacyl groups in cellulose esters, but most of the results on cellulose deriva.tivesstill await the compilation of more reference data before they can be fullyinterpreted. Other substances which have been measured include poly-esters, poly-amides, and silicon polymers. The spectra of some amino-acidshave been measured by N.Wright,61 and of some proteins by A. M. Ruswelland R. C. Gore.62 It seems likely that processes such as the denaturationof proteins may be followed by spectral changes, and the structural alter-ations may thereby in some measure be inferred.Attempts have also been made to follow the changes brought about bythe special treatment of macro-molecules , such as vulcanisation or plastic-66 Proc. Roy. Soc.. 1940, A , 175, 208.67 H. W. Thompson and P. Torkington, ref. (55).68 D. Williams and B. Pale, J. Appl. Physics, 1944, 15, 585.69 H. W. Thompson and P. Torkington, ref. (55).60 R. E. Richards, unpublished work.J. Biol. Chem., 1939, 127, 137.62 J . Physical Chem., 1942, 46, 575THOMPSON : RECENT ADVANCES IN INFRA-RED SPECTROSCOPY. 19isation. Vulcanisation by sulphur and other reagents,G3 and by sulphurchloride,64 brings about spectral changes, and efforts have been made tocorrelate them with the formation of C-S or S-S bonds, or with otherstructural changes. Similarly, the oxidation of rubber and of polythenehas been studied; polar groups such as C O can be detected and estimatedin hydrocarbon residua, even though present in very small amount.65 I nall these cases, however, the background of reference data is insufficient.The spectra of samples of coal 66 and of coal extracts show special featuresof interest, and some promising deductions have been made about thepresence of key groups.The infra-red spectra of dyes and paints G7 havebeen discussed. General accounts of other applications have been publishedby J. Lecomte.68The other recent work in this field can only be reviewed briefly. Anumber of papers have extended spectroscopic theory and our generalunderstanding of the vibrational and rotational levels of molecules. D. M.Dennison's earlier review of this subject 69 has been followed by a secondpart.70 The interaction of vibrational and rotational energy has beenconsidered by several authors,71 and a striking experimental verification ofsome earlier' predictions by Nielsen has been found in two vibration bandsof allene.72 The normal vibrations of molecules with internal torsion havebeen examined by B. L.Crawford and E. B. Wilson,73 and in a generalarticle J. Duchesne 74 has reviewed critically the potential energy functionsof molecules. A. G. Meister, F. F. Cleveland, and M. J. Murray 75 havetabulated the selection rules in the infra-red and Raman spectra for vibrationsof molecules belonging t o different symmetry groups. J. W. Linnett 76 hasdiscussed the force constants of various kinds of bond.Attempts to analyse the spectra of molecules with a view to assign thevibrational frequencies or to obtain moments of inertia have been carriedout in more cases, including nitr~rnethane,~~ nitrodeuteromethane,78N. Shepperd and G. B. B. M. Sutherland, Trans. Faraday SOC., 1945, 41, 261.64 P. Torkington and H. W. Thompson, ibid., p. 276.6 5 H.W. Thompson, ref. (55).G6 C. G. Cannon and G. B. B. M. Sutherland, ibid., p. 279.67 J. Stearus, Amer. Dyestufls Rep., 1944, 33, 1, 16.6 8 Cours. Confer. Paris, Centre Perf. Tech., Non. 1943; Bull. SOC. Franp. Electr.,6B Rev. Mod. Physics, 1931, 3, 280.70 Ibid., 1940, 12, 175.7l H. H. Nielsen, Physical Rev., 1940, 62, 151, 161; J . Chem. Physics, 1941, 9,847 ; 1943,11, 160; TV. H. Shaffer, Rev. Mod. Physics, 1944,16,245; J. Chem. Physics,1848, 10, 1 ; 1944, 12, 504; W. H. Shaffer and R. C. Herman, ibid., 1945, 12, 83;W. H. Shaffer and W. Silver, ibid., 1941, 9, 599, 607; 1940, 8, 919; W. Silver, ibid.,10, 559, 565.1942, 2, 1 ; Bull. SOC. Philomath., Paris, 1942, 124, 68.7 2 H. W. Thompson and G. P. Harris, Trans. Faraday SOC., 1944, 40, 295.73 J.Chem. Physics, 1941, 9, 323.7 6 Amer. J. Physics, 1943, 11, 239.7 7 A. J. Wells and E. B. Wilson, J . Chem. Physics, 1941, 9, 314.7 8 T. P. Wilson, ibid., 1943, 11, 361.74 Mem. SOC. Roy. LiZge, 1943, 1, 429.7g Trans. Faraduy Xoc., 1945, 41, 22320 GENERBL AND PHYSICAL CHEMISTRY.p r ~ p y l e n e , ~ ~ propaneY8O methylamine,81 acetaldehyde, and deuteroacet-aldehyde, 82 cyczohexane, 83 pyridine, 84 cyclobutane, butadiene, 86 furan,87and thiophen.88 A very important analysis has been made by W. S.Gallaway and E. F. Barker 89 of the spectra of ethylene and tetradeutero-ethylene, which claims to fix the carbon-carbon bond length in thesemolecules with high precision. There is still some doubt, however, aboutthe correctness of their vibrational assignments. A comparison of thespectra of the vinyl halides and vinyl cyanide has led to a very satisfactoryvibrational assignment for these molecules.g0 The rotational contour ofsome bands of the fluoroethylenes has also been measured,gl and the changesin some of the vibration frequencies of these molecules resulting from theintroduction of fluorine atoms is particularly noteworthy.More papers havedealt with measurements on the hydrogen bond, but these must be reservedfor a future report. The effect of temperature upon the shapes of someabsorption bands of hydrocarbons has been examined by W. H. Avery andC. F. Ellis.92Although it is not intended to discuss the Raman spectra in this report,the account would not be complete without reference to the new rapidmethod of measuring Raman spectra using large spectrometers and photo-electric cells as detect0rs.~3 If this method of recording the spectra fulfilsits present promise, it may well open up a new chapter in the applicationof such measurements.H.W. T.2. FRICTION AND LUBRICATION.When there is relative motion between surfaces which are separated bya liquid layer of appreciable thickness, the resistance to motion is due tothe viscosity of the interposed layer. This type of lubrication, which occursin well-designed journal bearings, is essentially a problem in hydro-dynamics; the friction is very small and since the surfaces are completelyseparated by the lubricant film there is no wear of the moving parts. It79 E.B. Wilson and A. J. Wells, J . Chem. Physics, 1941, 9, 319.80 V. L. Wu and E. F. Barker, ibid., p. 487.81 R. G. Owens and E. F. Barker, ibid., 1940, 8, 229.82 J. C. Morris, ibid., 1943, 11, 230. 83 R. S. Rasmussen, ibid., p. 249.84 J. Turkevich and P. C. Stevenson, ibid., p. 328; 1944, 12, 300.S 5 T. P. Wilson, ibid., 1943, 11, 369.8 6 R. S. Rasmussen, D. D. Tunnicliff, and R. R. Brattain, ibid., p. 432.8 7 L. W. Pickett, ibid., 1942, 10, 660; H. W. Thompson and R. B. Temple, Trans.8 8 Idem, ibid. J . Chem. Physics, 1942, 10, 88.Qo P. Torkington and H. W. Thompson, J . , 1944, 597, 303; Trans. Paraday SOC.,1945, 41, 240; Proc. Roy. SOC., 1945, A , 184, 21.Q1 P. Torkington and H. W. Thompson, Trans. Paraday SOC., 1945, 41, 236.O2 J . Chem.Physics, 1942, 10, 10.O3 R. H. Rank, R. J. Pfister, and P. D. Coleman, J. Opt. SOC., 1942, 32, 390; R. H.Rank, R. J. Pfister, and J. Grimm, ibid., 1943, 33, 31; R. H. Rank, R. W. Scott, andM. R. Fenske, Ind. Eng. Chem. Anal., 1942,14, 816; R. F. Straum, ibid., 1945,17, 318.Faraday SOC., 1945, 41, 27BOWDEN AND TABOR : FRICTION AND LUBRICATION. 21is clear, however, that in many practical cases fluid lubrication is impossible.At the beginning and end of a reciprocating stroke and in many slidingmechanisms it is difficult to maintain a thick continuous film of lubricantand even in rotating parts the thick film may break down and only a surfacefilm of lubricant may remain. The friction in such cases is influenced bythe nature of the underlying surfaces as well as the chemical constitution ofthe lubricant, and (Sir) W.B. Hardy 1 referred to such a state as “ boundarylubrication.’’ I n practice, boundary lubrication is of great importance, andthe nature of the surface film will determine whether serious wear or seizurewill take place.Theory of Friction.Before we can form any picture of the nature of this thin-film lubric-ation it is necessary to know something of the origin of the frictional forcebetween clean unlubricated solids. It is a matter of considerable ex-perimental difficulty to prepare and to maintain metal and other surfaceswhich are quite free from oxide and other adsorbed films (see later), and,since most friction measurements have been carried out in air, the surfacesare necessarily contaminated in this way.The early view of Coulomb 2 thatthe frictional force between solids is due to the interlocking of surfaceasperities, so that the frictional work represents the work required to liftone surface irregularity or high spot over another, is still held by someworker^.^have postulated that friction is due to an interaction between the surfacefields of force of the two solids, so that the friction may be regarded as apurely surface effect due to the molecular attraction between the two solids.In recent years, our knowledge of the surface structure and surface con-tour of solids has advanced considerably. Instruments have been de-veloped 6 ~ 7 ~ 8 ~ 9 which amplify the movement of a tracer needle as it passesslowly over the surface.These instruments will record surface irregularitiesas small as 0.05-0-1 micron. Although they can be very useful in moretechnical applications, they do not reveal much of the fine cracks, pits, orsharp steps in the surface, since their sensitivity is limited by the diameterof the tracer point, which in general cannot be much less than 5p. How-ever, more sensitive physical methods have been introduced recently.R. D. Heidenreich and L. A. Mattheson,lo using the electron microscopestereoscopically, have shown that it can reveal surface irregularities whichare of the order of 0.015-3p. R. C . Williams and R. W. G. Wyckhoff l1On the other hand, Hardy,l G. A. Tomlinson,4 and B. Derjaguin“ Collected Works,” Cambridge University Press, 1936.Mem.Math. Phys. Acad. Roy. Sci., 1785, 161.E.g., J. J. Bikerman, Rev. Mod. Physics, 1944, 16, 53.Phil. Mag., 1929, 905.E. J. Abbot, S. Bousky, and D. E. Williamson, Mech. Eng., 1938, 60, 205.Anon., Engineering, 1941, 151, 356.See, e.g., the Talysurf Surface Finish Measuring Instrument, Messra. Taylor &Anon., Engineering, 1945, 159, 427.2. Physik, 1934, 8, 66.Hobson.lo J. Appl. Physics, 1944, 15, 423. l1 Ibid., p. 71222 GENERAL AND PHYSICAL CHEMISTRY.have sputtered thin metallic films on to surfaces from an oblique angle.Small isolated peaks 30 A. high cast a “shadow” which may then bedetected by the electron microscope, although the irregularity itself is notvisible. Improved optical interference methods have also been used.12) 13,143 15S.Tolansky 14 has introduced an important modification by silvering theoptical flat and half silvering the surface in order to get multiple reflectionof the incident light. The fringes are then extremely sharp and a t highmagnification can reveal differences in height of less than 40 A. and changesof face angle as small as 0.016 minute of arc. He used this method tostudy the cleavage surface of mica and showed that small steps were present,the heights of which were integral values of the molecular length (20a.).The (100) face of a natural quartz crystal also showed minute steps of afew molecules in height (100 A. high). Similar measurements have beenmade of the detailed topography of a diamond crystal surface.16The effectiveness of the ordinary optical microscope for the study ofsurface contour can also be considerably increased by protecting the surfaceirregularities with an electrodeposit and cutting a taper section at a veryoblique angle on the surface.17 Irregularities as small as 0 .1 ~ can bedetected in this technique, which also has the advantage of revealing thestructure of the solid immediately below the surface. Electron-diffractionmethods 18,19 also give valuable information about the structure of thesurface films and surface layers of solids. It is evident that metal surfaceswhich have been prepared by the usual methods will be covered with avisible oxide film up to 100 A. thick 2 0 , 2 1 9 2 2 and a layer of altered metal.H. C. V a ~ h e r , ~ ~ using X-ray back reflection methods, has shown that thedeformed layer on steel, polished in the usual metallographic way, is 2pthick, while copper finished on a fine emery paper showed a deformed layer20p thick.It is clear that the metal surfaces normally employed both in practiceand in laboratory experiments on friction will normally be very complexand will consist of (a) surface irregularities which are very large comparedwith molecular dimensions, ( b ) an oxide film, and (c) an altered layer inthe metal itself.When two such surfaces are placed together, they will besupported on the summits of the surface irregularities, and the real areaof contact, i.e., the area over which the surfaces are within molecular range,12 T. P. Hoare and B.Chalmers, Tin Res. and Dev. Council, Pubn. A21, 1935.13 J. F. Kayser, Met. Treatment, 1943, 10, 153.1 4 Proc. Roy. SOC., 1945, A , 184, 41.l8 S. Tolansky, Nature, 1946, 157, 5831 7 H. R. Nelson, Conf. Friction and Surface Finish, MIT., 1940, 217.l5 C . Timms, J . Sci. Instr., 1945, 22, 245.G. P. Thomson and W. Cochrane, “Theory and Practice of Electron Diffraction,”l9 B. Chalmers and A. G. Quarrel, “ Physical Examination of Metals,” Arnold, 1941*O W. H. J. Vernon, F. Wormwell, and T. J. Nurse, J., 1939, 621.21 W. E. Campbell and U. B. Thomas, Trans. Amer. Electrochem. SOC., 1939, 76, 303.22 U. R. Evans, “ Metallic Corrosion, Passivity and Protection,” Arnold, 1937.23 J. Res. Nut. Bur. Stand., 1942, 29, 1’77.London, 1939.[see also refs.(94)-(102)]BOWDEN AND TABOR : FRICTION AND LUBRICATION. 23will be small. This means that, even with comparatively lightly loadedsurfaces, the local pressure a t the region of contact will be high and mayeasily exceed the yield point of the metal or other solid so that plastic flowand deformation occurs a t the point of contact. Measurements of theelectrical conductivity z4, 25 between metals in contact support this viewand show that the area of intimate contact is indeed very small, is com-paratively little influenced by the size of the surfaces, and is determinedmainly by the applied load and the yield point of the metal. Apparently,the summits of the irregularities on which the solids are supported flowplastically, and are crushed down until their cross section is sufficient t oenable them to support the applied load.Beyond this region of intensepressure there will be local elastic deformation of the solids. Except underthe lightest loads, there will be partial breakdown of the surface films andtrue metal to metal contact occurs; this is particularly true if there issome slight movement as if the surfaces are sliding. According to the viewsof Bowden and his co-workers, this local adhesion and pressure welding ofthe surfaces a t the points of contact plays a major part in the friction ofmetals and many other solids. The frictional resistance is due primarilyto the shearing of these metallic junctions and to the work of dragging orploughing the surface irregularities of the harder metal through the softerone.26 Examination of the surfaces after sliding 26$ 273 28 shows that they aretorn and distorted to a depth which is very great compared with moleculardimensions, so that friction cannot be regarded as a purely surface effect.The bulk properties of the solids, such as their relative hardness and, athigh sliding speeds, their relative melting point or softening point, play animportant part.The physical processes occurring during sliding areobviously complex, but as an approximation the frictional force F maybe written F = 8 + P , where S is the shearing term and P the ploughingterm. I n general, it follows that S = As, where A is the real area ofcontact and s the shear strength of the metal; and P = A’p where A‘ isthe cross sectional area of the torn track and p the pressure necessary tocause plastic flow of the metal.Experiments with sliders of different shapesand hardness 26p28,29 show that P is usually small compared with S, andunder the conditions where it can be neglected P = As. The real area ofcontact A is determined primarily by the load W , and W = p A . HenceP = Ws/p, and the coeficient of friction p = P/W = s / p = shear strength/flow pressure.A somewhat similar expression which includes a term for surface rough-ness has also been derived by H. Ernst and M. E. Merchant 30 and applied24 R. Holm, Wiss. Verofl. Siemens-Konz., 1929, 7 (2), 217; 1931, 10, (4), 1; “Dietechnische Physik der elektrischen Kontakte,” Springer, 1941.2s F. P.Bowden and D. Tabor, Proc. Roy. SOC., 1939, A, 169, 391.26 F. P. Bowden, A. J. W. Moore, and D. Tabor, J. Appl. Physics, 1943, 14, 80.27 F. P. Bowden and A. J. W. Moore, Nature, 1945, 155, 451.28 F. P. Bowden, Proc. Roy. SOC. N.S.W., 1945, 78, 187 (Liversidge Lectures).29 F. P. Bowden and D. Tabor, Nature, 1942, 150, 197.30 Conf. Friction and Surface Finish, MIT., 1940, 7624 GENERAL AND PHYSICAL CHEMISTRY.to the cutting of metals.31 The nature and distribution of the small weldedjunctions which are formed and broken during sliding have been deter-mined by Bowden, Moore, and Tabor,Z6 using the taper section technique.When copper is slid on clean steel, small particles are detached and adherestrongly to the steel surface. The shearing usually occurs in the copper,but the strength of the bond is such that occasionally the steel itself isdragged up above the general surface level, or plucked out altogether.Thisis a clear example of a hard metal being worn away by a softer one. Thereis also a marked deformation and hardworking of the metal to a considerabledepth beneath the torn surface. More recent work27 using the methodof electrographic surface analysis 32933 has shown that this localised metallicadhesion occurs even with lubricated surfaces. Experiments with a copperslider passing once over lubricated platinum, for example, have shown thatthe amount of copper adhering to the platinum surface may be cu. g.per mm.2 of track. The copper is not spread uniformly, but is distributedas a number of small discrete particles where local adhesion has occurred ;this is most pronounced on the high spots of the surfaces.If the surfacesare unlubricated, the amount of metallic pick up under similar conditionsmay be greater by a factor of 100 or more. If one of the surfaces is naturallyor artificially radioactive, it provides a very sensitive method for detectingpick up. Sackman, Burwell, and Irvine3* have used curved sliders ofcopper-beryllium alloy and of steel on a radioactive copper-beryllium alloyand have shown by Geiger counter methods that adhesion occurred on theslider under clean and lubricated conditions. This method can be used todetect quantities as small as 10-5 microg. J. N. Gregory,35 using radioactivelead and a photographic technique, has also shown that localised metallicadhesion and welding occurs through the lubricant film.This localisedwelding occurs a t very low sliding speeds where the temperature rise isnegligibly small. It is a " cold " welding brought about by the high localpressure a t the point of contact; at greater speeds and loads, however, thefrictional heat may raise the local surface temperature to a high value, sothat a softening or even a melting may occur a t the points of contact.The occurrence of these high temperatures may be demonstrated bymeasuring the thermal e.m.f. developed between rubbing surfaces of dis-similar metals.36 Earlier work has shown 37,38 that extremely high localtemperatures may readily be reached between metal surfaces undermoderate conditions of load and speed,39 even in the presence of lubricantfilms. When the surfaces are non-conducting, this method is no longer31 M.E. Merchant, J. Appl. Physics, 1945, 16, 267, 318.32 D. L. Masters, Metallurgia, 1943, 29, 101.33 M. S. Hunter, J. R. Churchill, and R. B. Mears, Met. Prog., 1942, 42, 1070.34 B. W. Sackman, J. T. Burwell, and J. W. Irvine, J . Appl. Physics, 1944,15,459.35 Nature, 1946, 157, 443.36 E. G. Herbert, Proc. Inst. Mech. Eng., 1926, 2, 289.37 F. P. Bowden and K. E. W. Ridler, Proc. Roy. SOC., 1936, A , 154, 640.38 T. U. Matthew, J . Roy. Tech. Coll. Qlasgow, 1940, 4, 360.30 H. Blok, Inst. Mech. Eng., Discussion on Lubrication, 1937, 2, 14BOWDEN AND TABOR : FRICTION AND LUBRICATION. 25applicable, but it is clear that for such surfaces, under any given loadand speed, the temperature rise will be higher than for conducting surfaces.If one of the surfaces is transparent, the incidence of high surface temper-atures may be directly observed by visual or photographic Ifpolished surfaces of glass or quartz are used, and the apparatus is so arrangedthat a clear image of the rubbing surfaces can be seen, it is found that anumber of tiny stars of light appear a t the interface between the rubbingsurfaces.These correspond to local hot spots, and experiments suggestthat the temperature at which they first become visible to the eye, asreddish stars, is about 500". They may be observed with metals sliding onglass, when the speed is as low as one or two feet per second.At higherloads or speeds, the points of light become whiter and brighter, correspondingto an increase in the temperature of the hot spots. It is clear that thesehigh local temperatures may greatly facilitate the formation of weldedjunctions between sliding surfaces, and will play an important part inthe adhesion and frictional behaviour of the rubbing solids. The r81eof these high surface temperatures in the process of polishing has beende~cribed.~~Intimacy of Contact.-It is clear that if the frictional force is essentiallydue to the formation and shearing of metallic junctions it will depend onthe intimacy of contact between the metal surfaces. Under most experi-mental conditions, metal surfaces are covered with a thin oxide layer andother contaminating films.During sliding, these films are torn, and manyof the surface irregularities penetrate through them so that a certain amountof intimate metallic contact does occur. We should, however, expect thatif the surface contaminations are removed the intimacy of contact betweenthe surfaces will be increased with a corresponding increase in the friction.This effect will be more marked the more effectively the surface films areremoved. Mechanical and chemical methods are relatively ineffective,* andsurfaces cleaned by even the most stringent normal laboratory methodsrarely give a coefficient of friction above about p = 1. If the surfaces areheated in a vacuum, the surface films are partially destroyed and there isan increase in f r i ~ t i o n .~ ~ , ~ ~ Unless the contaminant film is completelydestroyed, however, the effect is not very marked and the results are variable.By using high-melting metals, F. P. Bowden and T. P. Hughes44 wereable to outgas the surfaces very thoroughly in a high vacuum a t temperatureswell over 1000". They found that the results so obtained were reproducible,and the coefficient of friction extremely high. For example, for outgassed40 F. P. Bowden, M. A. Stone, and G. K. Tudor, Proc. Roy. SOC. (in the press);F. P. Bowden and M. A. Stone, Experientia, 1946, 2, 186.*l F. P. Bowden and T. P. Hughes, ibid., 1937, A, 160, 575.42 W. Claypoole and D. B. Cook, J . Franklin Inst., 1942, 233, 453.43 J. M. Macaulay, J . Roy.Tech. Coll. Glasgow, 1935, 353.4 4 Proc. Roy. SOC., 1939, A, 1'42, 263.4 5 R. schnurmann, Proc. Physical Xoc., 1941, 53, 638.* See, however, the marked adhesion between platinum surfaces cleaned by chemicalmethods,42 and the adhesion of freshly cleaved mica.426 GENERAL AND PHYSICAL CHEMISTRY.nickel or tungsten p increased from about 0.3 to between 5 and 6. In asimilar way, the presence of contaminants or lubricant films reduces theintimacy of contact between the sliding surfaces, and so results in a decreasein the frictional force. A quantitative formulation of these factors hasrecently been given by Bowden and Tabor.29Intermittent Motion.-The friction between metal surfaces depends onthree main factors. The first is the intimacy of contact, which determinesthe extent to which the adhesion between the surfaces is truly metallic,as discussed above.The second is the flow-pressure p , which determinesthe area of real contact between the surfaces. The third is the shear strengthof the metals, which determines the strength of the metallic junctions formed.If any experimental variations arise which alter one or more of these factors,there will be a corresponding alteration in the frictional force. To a certainextent, therefore, the friction will depend on the experimental conditionsunder which it is measured. For example, a t extremely slow speeds ofsliding, these factors vary in such a way that the resultant strength of themetallic junctions formed is often higher than that occurring a t higherspeeds of sliding.This means that the static friction is often higher thanthe kinetic f r i ~ t i o n . ~ 6 ~ * ~ ~ ~ ~ ~ 49 If, therefore, one of the sliding surfaces has acertain degree of elastic freedom, the motion may not be continuous, butmay be intermittent and proceed by a process of " stick-slip ". The " stick "corresponds to the static friction between the surfaces, and the " slip " tothe lower kinetic friction during the slip itself. This type of motion clearlydepends on the mechanical properties of the system, such as the naturalfrequency, the moment of inertia, and the damping of the moving parts.50It will also be influenced by the 'velocity of the main forward motion andby the friction-speed characteristics of the surfaces under consideration.51Since many moving systems possess an appreciable degree of elastic freedom,and since the static friction is often higher than the kinetic friction formetal surfaces, this type of intermittent motion is of frequent occurrencein practice.Even when the moving parts are extremely rigid, the surfaceirregularities may be capable of microscopic elastic deformation of theorder of 10-5 cm, as S. Khaikin, L. Lissovsky, and A. Solomonovitch 52 haverecently shown, using quartz crystals to measure the minute displacementsinvolved. I n such cases, the elasticity of the surface irregularities them-selves may, in the limit, be sufficient to set up vibrations or intermittentmotion in the moving parts.A fourth factor which appears to have some influence on the friction46 F.Morgan, M. Muskat, and D. W. Reed, J. Appl. Physics, 1941, 12, 743; J. B.4 7 S. Khaikin, L. Lissovsky, and A. Solomonovich, J. Physics (U.X.S.R.), 1940, 2,4 8 J. R. Bristow, Nature, 1942, 149, 169.49 B. Chalmers, P. G . Forrester, and E. F. Phelps, ?roc. Roy. SOC. (in the press):60 F. P. Bowden, L. Leben, and D. Tabor, Engineer (London), 1939, 168, 214.51 H. Blok, J . SOC. Aut. Eng., 1940, 46, 54.62 J. Physics (U.S.S.R.), 1939, 1, 455.Sampson, F. Morgan, D. W. Reed, and M. Muskat, ibid., 1943, 14, 689.253BOWDEN AND TABOR : FRICTION AND LUBRICA!L’ION. 27of metal surfaces is the degree of surface fini~h.~395~9~5 Although this effectmay be explained in terms of Coulomb’s theory,3 it may also be explainedin terms of the variation of the ploughing and shearing terms produced bya change in the geometry of the rubbing surfaces.With unlubricated metal surfaces, the friction obeys Amontons’s lawsince a change in load W produces a corresponding change in the area ofreal contact A , where A = W / p .If we neglect the ploughing term andassume a mean value S for the shear strength of the junctions, the frictionalforce is given by F = AB = Wi/p, which is directly proportional to theload. Similarly, the friction of unlubricated metal surfaces is not markedlyaffected by temperat~re.3~3 447 45 The shear strength s decreases with temper-ature, but there is a corresponding increase in the area of contact as themetal softens. Since both B and p are strength properties of the metals,they vary in a similar way with temperature, so that the term B/p is notsensibly dependent on the temperature.If, however, the heating is carriedout in a vacuum and leads to a partial removal of adsorbed layers, the effectof temperature is more complicated and less reprodu~ible.~~9~~Metallic Film Lubrication and the Theory of Bearing Alloys.-A com-pletely different situation arises when two hard rubbing metal surfaces areseparated by a thin film of a softer metal such as indium, lead, or copper.56If the metallic film is plated on to a hard flat surface and the second surfaceconsists of a hard spherical slider, the area of contact is determined largelyby the thickness of the film and the geometry of the surfaces.The under-lying hard surfaces, which support the load, are deformed comparativelylittle, so that further increases in the load have relatively little influence onthe area of contact between the slider and the metallic film. This givesa small area of contact A which is almost independent of load. Providedthe metallic film remains intact, so that little or no metallic contact occursthrough it, the shearing occurs within the soft metallic film. This gives asmall value of i5, so that the frictional force which is given by F = As issmall and almost independent of the load.29 Experiments show, in fact, thatAmontons’s law does not hold, and that extremely low coefficients may beobtained a t the higher loads; with indium films on tool steel, for example,p may be as low as 0.02.If, however, very heavy loads are used, the filmbreaks down, contact occurs between the slider and the underlying metal,and the friction rises. With metallic films, the friction decreases steadilyas the temperature is raised, and reaches a minimum when incipient meltingbegins. This is because the area of contact remains essentially constant,whilst the shear strength of the metallic film steadily decreases. Oncemelting is complete, there is a rise in friction, which is greater if the moltenmetal fails to wet the underlying surface. Thin metallic film lubrication hasbeen used by Z. J. Attlee, J. T. Wilson, and J. C. Filmer 57 to lubricate the53 J. Pr6vost, Mdcanique, 1939, 23, 139.ti4 C. A. Congwer, Conf. Friction and Surface Finish, MIT., 1940, 239.56 J.T. Burwell, J. SOC. Aut. Eng., 1842, 50, 450.6 6 F. P. Bowden andD.Tabor, J . AppZ. Physics, 1943,14,141. 6 7 Ibid., 1940,11,61128 GENERAL AND PHYSICAL CHEMISTRY.steel ball bearings of a rotating-anode X-ray tube. Thin metallic films havealso been used successfully in deep-drawing operations 58 and it is possiblethat they will find increasing use as lubricants under extreme conditions.Earlier accounts of bearing alloys have suggested that an essentialcharacteristic of a bearing alloy is that it should possess a duplex structureconsisting of hard crystals embedded in a softer matrix.59 The function ofthe hard crystals is to resist wear, and that of the softer constituents topermit a more uniform distribution of the load, by allowing any of the hardcrystals that are heavily loaded to sink into the matrix. It is also suggestedthat the hollows worn in the softer material serve as reservoirs for thelubricating oi1.60 Although many successful bearing alloys do possess astructure of this type, recent investigations show that this picture of themechanism of bearing alloys is inadequate.First, even with bearings ofthis type (white-metal bearing alloys) experiments show 61 that the hardparticles are impressed into the surface of the alloy by the sliding processand the frictional and wear properties are determined essentially by thoseof the matrix material itself. Secondly, with many modern bearing alloysthe surface layer does not possess a duplex structure at all, but consists ofa single pure Thirdly, another very wide class of bearing alloys(copper-lead type) consists of a hard matrix (copper) in which numeroussmall particles of a soft phase (lead) are dispersed.In this case the hardcopper is the continuous phase and cannot “ sink ” into the lead, Invest-igations show 56 that in bearing alloys of this type a thin film of the softermetal is extruded by the sliding process and acts as a thin metallic-filmlubricant on the surface of the harder matrix.Non-metals.-Very little work has been carried out on the friction ofnon-metallic substances.(i) Crystalline solids. R. Hutchison 63 has investigated the friction ofcrystalline substances, such as sodium and potassium halides, sulphur,paraffin-wax, and quartz on like crystals and on metals.Similar work hasbeen done by E. Hut~hinson.~~ In general, the friction coefficient is in-dependent of load and speed, but with extremely soft materials such asparaffin-wax the behaviour is essentially like that of a liquid under shear,the friction increasing rapidly with speed of sliding. The frictional behavioiiris associated more with the physical characteristics of the solids than withtheir chemical structure. In nearly all cases when non-metallic crystalsslide on a clean metal surface, there is a small but definite deposit of crystalon the surface. However, with diamond on a hard metal, there is nopick-up and the friction is extremely small, about p = 0.05. Other measure-68 A. J.W. Moore and D. Tabor, C.S.I.R. (Australia), Lubricants and BearingsSect., 1943, Report A96.LD H. N. Bassett, “ Bearing MetaIs and Alloys,” Arnold, 1937.6o C. H. Desch, “ Metallography,” Longinans Green, 1937.61 D. Tabor, J . Appl. Physics, 1945, 16, 325.62 A. Bregman, Iron Age, 1942, 150 (7), 65; (€9, 41.63 Private communication, 1938.64 Thesis (Cambridge) 1946, “ Adsorption and Lubrication a t Crystal Surfaces.BOWDEN AND TABOR : FRICTION AND LUBRICATION. 29ments on the friction of diamonds, sapphires, and jewel pivots have beencarried out by W. Claypoole,65 G. F. Shotter,66 and V. S t ~ t t . ~ ~Hutchison has also extended the earlier work of F. P. Bowden andT. P. Hughes 68 on the frictional properties of ice surfaces. The latterworkers found that a t a few degrees below the melting point of ice thestatic friction is of the same order as for other solid bodies, whilst the kineticfriction is an order of magnitude lower.This effect was explained as beingdue to a surface melting of the ice by frictional heating. Hutchison hasobtained similar results for benzophenone, dinitrobenzene, and sodiumhyposulphite (dithionite). With benzophenone, p fell from 0.2 a t slowspeeds of sliding to 0.03 at high speeds. In confirmation of the earliertheory of surface melting, it was again found that the friction was increasedby (a) increasing the thermal conductivity of the sliding surface, ( b ) decreas-ing the bulk temperature of the rubbing bodies.Some interesting experiments on the frictional properties of freshlycleaved mica surfaces, and of the effect of surface-active materials on thefriction, have been described by B.Derjaguin and 17. Lazarev 69 and otherRussian workers.(ii) Non-crystalline solids. The main recent investigations on non-crystalline materials have been on glass and on rubber. G . W. Hammerand G. Martin 70 showed that in a vacuum the friction of glass is increased,and their observations are consistent with Hardy’s and P. E. Shaw andE. W. L. Leavey’s 71 earlier work, which showed that there .is markedadhesion and tearing of the glass surfaces. F. L. Roth, R. L. Driscoll, andW. L. Holt 72 have investigated the frictional properties of rubber onground steel and on plate-glass surfaces, They find that the static frictionis lower than the kinetic friction and that, in general, rougher surfaces givesmaller coefficients of friction than smooth surfaces.In general, the frictionof a rubber compound depends more on the nature of the rubber matrixthan on the compounding ingredients and fillers.The frictional properties of natural fibres are of interest tothe manufacturers of textiles. Recent measurements by a number ofworkers on the friction of wool fibres have shown that for dry wool thefriction against the scales is higher than in the direction of the scale~.~~9 749 75This “ directional friction ” effect has been used to explain the feltingproperties of woollen fabrics.76 E. H. Mercer and M. Lipson 77 have shown(iii) Fibres.6s Trans. Amer. SOC. Mech. Eng., 1936, 61, 323.6 6 Inst.Mech. Eng., Discussion on Lubrication, 1937, 2, 140.67 Ibid., p. 145.70 Science, 1939, 90, 179.72 J . Res. N a t . Bur. Stand., 1942, 28, 439.73 M. Lipson, Nature, 1945, 158, 268.76 C. S. Whewell, L. Rigelhaupt, and A. Selim, ibid., 1944, 154, 772.76 E. H. Mercer, ibid., 1945, 155, 573; see also J. B. Speakman and E. Stott,J . Textile Inst., 1931, 22, ~ 3 3 9 .77 Ibid., 1946, 157, 134.Proc. Roy. SOC., 1939, A , 172, 280.J . Physical Chem. (U.S.S.R.), 1934, 5, 416; Kolloid-Z., 1934, 69, 11.71 Phil. Mag., 1930, 10, 809.74 L. Bohm, ibid., 155, 54730 GENERAL AND PHYSICAL CHEMISTRY.that certain agents reduce the felting by reducing the difference betweenthe pro-scalar and anti-scalar coefficients of friction, and this effect has beenexplained as being due in some cases to a destruction of the overlappingratchet-like edges of the scales by the chemical agent usecl.78 The effectof pH on the frictional properties of fibres has been used by R.E. D. Clark 79to determine the end point in acid-alkali titrations.Tribo-electricity and its R61e in Friction.-It has long been known thatwhen solids are rubbed on one another electrical charges may be left onthe rubbing surfaces. The earlier work in this field was mainly concernedwith the practical application of this phenomenon to the production of highvoltages, and with the development of a theory that would explain theorigin of these tribo-electric charges. A good review of the work carriedout until 1936 on frictional electricity is given by W.HI. Ward.80pointed out many years ago, that the electricalcharges appearing at the surfaces of rubbing solids will make some contribu-tion to the frictional force observed. Recently, R. Schnurmann andE. Warlow-navies 81 have emphasised the importance of the electrostaticcomponent of sliding friction, particularly when the boundary layer hasdielectric properties, and they have explained the intermittent motion whichoccurs between sliding surfaces in terms of charging and discharging. Theseconclusions agree in part with those of P. E. Shaw and his co-workers, whohave made extensive measurements of electrostatic phenomena. In a paperwith Leavey in 1932, Shaw suggested 82 that tribo-electricity and frictionare two aspects of the same phenomenon; when two surfaces are separated,both take up an electrostatic charge, and when two surfaces slide over oneanother the frictional work is expended in overcoming the electrostaticattraction and in deforming the surface structure, Few workers in thefield, however, believe that the tribo-electric charges can play an appreciablepart in the mechanism of metallic friction, though they may play a largepart in the friction of non-metallic materials.It is clear, as HardyLubricated Surfaces.A systematic investigation of boundary lubrication was first undertakenby Hardy,l who measured the static friction between surfaces, using homo-logous series of paraffins, fatty acids, and alcohols as lubricants.He foundthat the coefficient of friction depended on the nature of the underlyingsurface, but that in all cases it decreased linearly with the chain length ofeach family of compounds.This led to the theory that friction is due tothe surface fields of force and that the effectiveness of a lubricant in reducingthe friction is determined by the extent to which the lubricant film canmask the fields of force of the underlying surfaces.7 8 E. H. Mercer and A. L. G. Rees, Nature, 1946, 157, 589.7e J . SOC. Chem. Ind., 1940, 59, 216.81 Proc. Physical SOC., 1942, 54, 14.82 P. E. Shaw, Phil. Mag., 1930, 9, 628; with C. S. Jex, Proc. Roy. SOC., 1926, A,111, 339; 1928, A , 118, 97, 108; with E. W. L. Leavey, ibid., 1932, A , 138, 502.Rep. Prog. Physics, 1937, 247BOWDEN AND TABOR : FRICTION AND LUBRICATION.31Later workers on static friction ha,ve not fully confirmed these r e s ~ l t s . ~ ~ Measurements of the kinetic friction also show that, although the frictiondecreases with chain length, the reduction is not linear 85786 and the frictioncoefficient reaches a steady low value of about 0.1. In particular, carefulmeasurements by electrographic 27 and radioactive methods 34935 show thateven with the best boundary lubricants the surfaces are torn to a depthwhich is large compared to the dimensions of a molecule, and there is acertain amount of metallic transfer through the lubricant film. It wouldseem that, as for unlubricnted surfaces, the friction cannot be regardedentirely as a surface effect, but must also be dependent upon the bulkproperties of the solids.Film Thickness and Structure.-I.Langmuir 87 was the first to show thata monolayer of a fatty acid deposited from the Langrnuir trough is sufficientto redlzce the friction of glass surfaces from about p = 1.0 for clean glassto about p = 0.1. Numerous workers have confirmed the importance ofthe first monolayer in reducing the friction. In particular, mention may bemade of the recent work of B. V. Derjaguin,88 E. N. Dacus, E. F. Coleman,and L. C. R o ~ s s , ~ ~ T. P. Hughes and G. WhittinghamYgo J. J. F r e ~ i n g , ~ ~T. I s e m ~ r a , ~ ~ and F. P. Bowden, J. N. Gregory, and D. Tabor.93 Isemurafound a slight decrease in friction with increasing film thickness. On theother hand, Bowden and Leben,86 working on built-up layers of stearic acidvarying from 1 to 53 molecular layers, found that a monolayer of stearicacid on steel produced the same low coefficient of friction as a multilayer53 molecules thick.However, the single film was soon worn away, and,for effective lubrication capable of withstanding considerable wear, it wasfound necessary to have present a layer of lubricant several molecules thickin order to replenish the surface with fresh lubricant. Recently Derjaguin 88and Dacus, Coleman, and Roess B9 have described apparatus for investigatingthe " life " of thin lubricant films. With some metals, particularly the noblemetals, a monolayer of fatty acid is insufficient to provide adequate lubric-ation, and for platinum, for example, a stearic acid film must be at least7 molecules thick to produce a low coefficient of friction.These frictional measurements have been co-ordinated with a study ofthe structure of thin lubricant films on solid surfaces.Using electron-diffraction techniques, C. A. M u r i ~ o n , ~ ~ L. T. A n d r e ~ s , ~ ~ and 0. Beeck, J. W.Givens, and A. E. Smith 96 have found that those lubricants which are most83 J. Sameshima, H. Akamatu, T. Isemura et al., '' Studies in the Oiliness of Liquids,"Bull. Chem. SOC. Japan, 1936-39 (10 papers) ; J. Sameshima et al., Rev. Physical Chem.Japan, 1940, 14, 55.A. Fogg, Proc. Physical SOC., 1940, 52, 239.8 5 W. G. Beam and F. P. Bowden, Phil. Trans., 1935, A , 234, 329.8 6 F. P. Bowden and L.Leben, ibid., 1940, A , 239, 1.8 7 Trans. Faraday Soc., 1920, 15, 62.89 J. Appl. Physics, 1944, 15, 813.91 Proc. Rog. SOC., 1942, A , 181, 23.93 Nature, 1945, 156, 97.96 Trans. Faraduy SOC., 1936, 32, 607.8 8 See Chem. Abs., 1942, 7280'.92 Bull. Chem. SOC. Japan, 1940, 15, 467.g4 Phil. Mag., 1934, 17, 201.s6 Proc. Roy. SOC., 1940, A, 177, 90.Trans. Faraday SOC., 1942, 38, 932 GENERAL AND PHYSICAL CHEMISTRY.completely oriented on the solid surface are also those which have the bestlubricating properties. Similar investigations on the structure of surfacefilms of fatty acids, alcohols, soaps, graphitic deposits, etc., by L. H. Germer,97J. J. Trillat, and H. M o ~ z , ~ ~ E. Havings and J. de Wae1,99 R. 0. Jenkins,loOK. Tanaka,lol and G .I. Finchlo2 have tended to agree with these con-clusions. Interesting studies of a similar nature using X-rays have also beendescribed by G. L. Clark, R. R. Sterrett, and B. H. Lincoln.103 However,it would seem that in some cases the degree of orientation of the lubricantfilm bears little relation to its intrinsic lubricating properties.104The friction between metal surfaces may be profoundly modified andgreatly reduced by the presence of thin films of adsorbed vapours 443105 aswell as by liquid and solid lubricant films. In some cases, a similar effectmay be produced by adsorbed layers of gases, and this observation hasformed the basis of a number of interesting experiments on the effect ofinterfacial potential on the friction of metal surfaces in solutions of electro-1ytes.lo63 log Later work has shown that these effects correspond tothe stages a t which adsorbed gaseous layers of hydrogen and oxygen areformed a t the surfaces.For example, G. C. Barker 1 1 O has found anapproximately linear relation between the amount of hydrogen and oxygenadsorbed a t a platinum surface and the coefficient of friction, oxygen beingmore effective than hydrogen in reducing the friction. Similar, though morecomplicated, effects have been observed between metals and non-metallicThe effect of surface films on the friction of metal surfaces has an interest-ing parallel in the work of Fkhbinder and his associates on the reductionof the strength properties of solids by surface-active materials.I n earliers~~fa~s~111,112,113,114,115Q7 J . Appl. Physics, 1938, 9, 143 ; L. H. Germer and K. U. Storks, Proc. Nat. Acad.,Q8 Cornpt. rend., 1935, 200, 1299.QQ Rec. Trav. chim., 1937, 56, 375; Chem. Weekblad, 1937, 34, 694.loo Phil. Mag., 1934, 17, 457.lol Mem. Coll. Sci. Kyoto, 1938, A , 21, 85; 1939, A 22, 377.lo2 Trans. Faraduy SOC., 1935, 31, 1051.lo3 Ind. Eng. Chem., 1936, 28, 1318.lo4 D. Tabor, Dissertation, Cambridge 1939, “ The Area of Contact between Station-lo5 H. Donandt, Reib. u. Verschleiss, 1939 (see Chem. Abs., 1942, 34664).loG T. A. Edison, “Handbook of Electrical Telegraphy,” 1874, I, 474.lo7 K. R. Koch, Ann. Physik, 1879, 7 , 92.lo* M. Krouchkoll, Cornpt. rend., 1882, 95, 177; Ann. Chim. Phys., 1889, 17,lo9 K. Waitz, Ann.Physik, 1883, 20, 285.110 Private communication, 1939.ll1 W. Barrett, Nature, 1880, 21, 483.112 A. Johnsen and K. Rahbek, 2. techn. Physik, 1921, 2, 11; J . Inet. Elect. Eng.,113 K. Rottgardt, 2. techn. Physik, 1923, 4, 1.114 J. Waszik, ibid., 1924, 5, 29.1937, 23, ,390; J . Chem. Physics, 1938, 6, 280; Physical Rev., 1939, 55, 648.ary and Moving Surfaces.’’182.1923, 61, 713.H. M. Barlow, J . Inst. Elect. Eng., 1924, 62, 133BOWDEN AND TABOR : FRICTION AND LUBRICATION. 33papers 116,117 he found that the yield value of metal wires is markedlyreduced by solutions of surface-active materials such as alcohols or fattyacids, for both polycrystalline and single crystal specimens. A similarreduction of Young’s modulus was observed in the elastic deformation ofmica sheets.118 Recently Rehbinder has determined the variation of surfacehardness of pyrites, immersed in sodium chloride solutions, when a varyingpotential is applied between the specimen and the s o l ~ t i o n .~ l ~ In theseexperiments a relation very similar.to the variation of the surface tensionof a mercury-electrode interface with potential was found. These effects areattributed to a penetration by the surface-active materials into the micro-cracks produced on the surface of the solid specimen by deformation, givingrise to a “ wedging pressure ” analogous to that observed earlier l2O9 121 forthin liquid layers between solid surfaces. This explanation is confirmed bythe fact that with metals the decrease in strength-properties is accompaniedby a marked decrease in electrical condu~tivity.1~~ Although the explanationis not quantitative, it is clear that these effects have a significant bearingon the friction of lubricated surfaces.Effect of Speed.-The earlier work of Beare and Bowden 85 on the kinetiofriction of lubricated surfaces showed that pk was sensibly constant over arange of speeds from 60 to 600 cm./sec., and T.Sasaki 122 finds pk independentof sliding speeds up to 100 cm./sec. More recently, Beeck, Givens, andSmiths6 have found that for low speeds the friction is independent of thesliding speed, but that with lubricants containing polar compounds thereis a marked decrease in friction a t a certain critical velocity. This wasattributed to the action of the polar molecules in drawing in a “ wedge ”of oil between the surfaces, and so producing quasi-hydrodynamic lubrication.The effect may also be explained in terms of the formation of a viscous filmof metallic soap formed by chemical reaction between the polar bodies andthe metal surface.93In some cases, the static friction is higher than the kinetic, underconditions of boundary lubrication (see Muskat et ~ 1 .~ ~ ) . Here, if therecording system has an appreciable degree of elastic freedom, the motionwill be intermittent. With “ good ” boundary lubrication this is notgeneral, and usually the friction is low and the motion is smooth.Effect of Temperature.-The effect of temperature on the lubricatingproperties of boundary lubricants is of general interest and importance. Inmany parts of an engine high temperatures may be reached in the runningparts, and it is necessary to know the way in which this will affect the116 P.Rehbinder and E. Wenstrom, Bull. Acad. Sci. U.R.S.S., Sdr. phys., 1937,4, 531.117 P. Rehbinder, V. I. Lichtman and V. M. Maslennikov, Compt. rend. Acad. Sci.U.R.S.S., 1941, 32, 125.11* P. Rehbinder and G. Logghinov, ibid., 30, 491.ll9 P. Rehbinder and E. Wenstrom, Acta Physicochim. U.R.S.S., 1944, 19, 36.120 B. Derjaguin and E. Obuchow, ibid., 1936, 5, 1.lZ1 B. Derjaguin and M. Kussakov, Bull. Acad. Sci. U.R.S.S., S h . chim., 1936,5, 741.122 Bull. Chem. SOC. Japan, 1938, 13, 134.REP.-VOL. XLII. 34 GENERAL AND PHYSICAL CHEMISTRY.lubricant.For small temperature variations a t room temperature, there ispractically no change in the f r i ~ t i 0 n . q ~ ~ ~ At temperatures up to loo", &I.Briault 124 and F. Charron 125 found an increase in friction with temperaturein some cases, but not in others. D. Tabor 126 pointed out in 1940 thatin general there is a well-defined transition temperature (specific to eachlubricant) at which the lubricant breaks down with a corresponding increasein friction and wear. Provided the heating has not been sufficient to causeappreciable oxidation of the lubricant, these changes are reversible oncooling, and the transition temperature T was considered to correspond toa disorientation or desorption of the adsorbed film of lubricant on thesurfaces.For pure paraffins and alcohols, the transition is sharp and well defined,and occurs at the bulk melting point of the compound.86 For fatty acids,the transition temperature depends on the load, speed, and experimentalconditions, but it is usually considerably higher than the bulk melting pointof the fatty Treating the phenomenon as an equilibriumadsorption process, Frewing 128 has used the frictional measurements todetermine the heat of adsorption of fatty acids and esters on steel surfaces.A different interpretation has, however, been given by Bowden, Gregory,and Tabor g3 (see below).For temperatures above 200" in air, the maineffect of temperature is that of oxidation. At an early stage, the oxidationproducts so produced may provide improved lubri~ation.1~~ At highertemperatures, however, or after prolonged heating, gumming, corrosion, andthe production of other deleterious products will cause a deterioration inthe lubricating properties.These effects are not reversible on cooling, andare due to chemical changes in the oils and sometimes the surfaces t,hemselves.127, 128Nature of Underlying Surface and the Importance of Soap Formation.Apart from a few earlier measurements by Hardy 1 and Same~hima,*~little work of a systematic nature has hitherto been carried out on the effectof the underlying metal on the lubricating properties of given boundarylubricants. G . M. Panchenkov and K. V. Konstantinova130 in 1939 de-scribed the effect of various metal substrates on the lubricating propertiesof certain organic compounds, and the investigation was extended over awider field by Hughes and Whittingham in 1942.More recently Bowden,Gregory, and Tabor g3 have investigated the frictional properties of fattyacids on an extensive series of metals, using similar metals for both of thesliding surfaces. One of the most striking results is that, for unreactivemetals such as nickel, platinum, silver, and glass, fatty acids are scarcely123 W. E. Campbell, Trans. Amer. SOC. Mech. Eng., 1939, 61, 633.124 Pub. Sci. Tech. Ministhe Air, France, 1934, 46, 29.125 Ibid., 1935, 131, 18; 1940, 169, 26.126 Nature, 1940, 145, 308.128 J. J. Frewing, Proc. Roy. Xoc., 1944, A, 182, 270.129 F. P. Bowden, L. Leben, and D. Tabor, Tmns. Furuday SOC., 1939, 35.900.130 J . Tech. Physik, U.S.S.R., 1939, 9, 537.lZ7 D. Tabor, ibid., 1941, 147, 609BOWDEN AND TABOR : FRICTION AND LUBRICATION. 35more effective as lubricants than saturated hydrocarbons. On the otherhand, those metals that are most readily attacked chemically by the fattyacid are most effectively lubricated ; whilst the less reactive metals, suchas iron and aluminium, require a higher concentration of fatty acid to givelubrication. (The chemical reactivity of the metals in air is determinedlargely by their oxides, as was pointed out by R. Dubrisay 131 and later byC. F. Prutton et al.132) This a t once leads to a modification of the oldertheory that lubrication is due to an adsorbed monolayer, and suggests thatfatty acids are most effective as boundary lubricants only when they canreact with the surfaces to form a metallic soap ; i.e., the lubrication is ejj’ected,not by the adsorbed fatty acid itself, but by the metallic soap formed on the metalsurface. This view has been confirmed by several experiments on thelubricating properties of metal soaps.The results show that in many casesthe frictional properties of a fatty acid R-CO,H on a metal surface M arethe same as those obtained with a soap (R-CO,),M on any type of surface.93Further, the transition temperature T a t which lubrication breaks downcorresponds approximately to the softening point of the metallic soap. Thishas been confirmed by electron-diffraction experiments which show that thetransition temperature corresponds approximately to the temperature a twhich the soap film loses its high degree of lateral o r i e n t a t i ~ n .~ ~ ~ ~ ~ ~ Thereis therefore a marked similarity between the frictional behaviour of metallicsoaps and of thin metallic films deposited on hard substrates. Lubricationis effective until the lubricant film softens and melts. The behaviour is alsosimilar to that of long-chain fatty acids on unreactive metal surfaces, andof long-chain hydrocarbons on any surfaces, since these lubricate until themelting point of the film is reached. With soap films, however, the soften-ing point is often appreciably higher than thl: mAting point of the corres-ponding fatty acid or hydrocarbon, so that they will lubricate satisfactorilyto a much higher temperature. Further, a single monolayer of soap is fre-quently sufficient to lubricate the surfaces.For these reasons fatty acids,when used on reactive metal surfaces, generally provide good boundarylubrication up to relatively high temperatures. If, however, the adsorptionof the lubricant film to the solid surface is weak, and there is prosent asuperincumbent layer of oil, the lubricant film may dissolve in the excessof oil a t a temperature lower than its softening or melting p0int.~3 Suchan effect may lead to a reduction of the transition temperature.Frewing’s 128 observations are not consistent with this view, since he hasshown that esters lubricate to temperatures well above their bulk meltingpoints on steel surfaces, and has estimated their heats of adsorption fromtheir lubricating properties.Esters, however, do not lubricate above theirbulk melting points on copper and cadmium surfaces, and Frewing’s resultshave been attributed to the anomalous behaviour of steeL93 Furtherinvestigations are needed to clarify this point.The Mechanism of Boundary Lubrication.-Boundary lubrication has beenexplained in terms of surface fields of force,l dipole moments of the lubric-131 Compt. rend., 1940, 210, 533. 132 I n d . Eng. Chem., 1945, 37, 9036 GENERAL AND PHYSICAL CHEMISTRY.ant,l33~1~~ surface. tension effects,l353 1369 1379 1389 139 and a modification ofCoulomb’s theory of surface a~perities.~ Since, however, some surfacedamage always occurs to a depth that is large compared with the dimensionsof a molecule, it would seem that boundary lubrication cannot be consideredas a purely surface effect.There is a continuous formation and shearingof metallic junctions through the lubricant film. When the lubricatedsurfaces are placed in contact, plastic flow of the metals occurs until thearea is large enough to support the applied load. The pressure, however,will not be uniform over the whole region of contact ; a t some points it willbe very much higher, and a t these points a local breakdown of the lubricantfilm may occur. The extent of the breakdown will naturally depend on thenature of the lubricant film. Further, if the sliding speeds are appreciable,it will be aided by local high temperatures developed during sliding.As aresult of the partial breakdown of the lubricant film, metallic junctions,large compared with the size of a molecule, are formed between the surfaces.The resistance to motion is then due in part to the force necessary to shearthese junctions. There will also be some resistance to sliding by thelubricant itself, and we may writewhere A is the area which supports the applied load, c( is the fraction of thisarea over which breakdown of the film has occurred, s, is the shear strengthof the junctions a t the metal-metal contact, and s is the shear strength ofthe lubricating film.93 With a good lubricant, the area over which metalliccontact occurs may be very small indeed. Nevertheless, the shear strengthof these junctions may be so high compared with that of the lubricant thatthey may be responsible for an appreciable part of the resistance to motion.The main purpose of the lubricant film is, therefore, to reduce the amountof metallic contact between the surfaces by interposing a layer that is noteasily penetrated and that possesses a relatively low shear strength.Thispurpose is served effectively by thin metallic films of soft metals 56 and bythin films of certain metallic soaps. There are, however, two main differ-ences in the frictional behaviour of thin soap films and thin metallic films.First, even on rough surfaces, a single molecular layer of soap may provideeffective boundary lubrication, whereas metal films must be appreciablythicker (ca. 10-6 cm.).56 It is for this reason that Amontons’s law holds forlubricated surfaces, but not for metallic film lubrication.Secondly, thegreater portion of the resistance to motion with thin metallic films is duet o shearing within the film itself; with lubricated surfaces, an appreciable133 R. Heinze, M. Marder et al., Oel u. KohZe, 1941, 37, 8.134 E. H. Kadmer, Petroleum Refiner, 1945, 24, 321.136 J. H. Wells and J. E. Southcombe, J . Xoc. Chem. Ind., 1920, 34, 5 1 ~ .136 D. P. Barnard and R. E. Wilson, I d . Eng. Chem., 1922, 14, 682.13’ J. J. Trillat and R. VaillB, J. Chem. Physics, 1936, 33, 742.13* P. Lecomte du Noiiy, Compt. rend., 1940, 210, 101.J. L. Culbertson and F. A. Hedman, J . Phy8iUd Chem., 1937, 41 485BOWDEN AND TABOR : FRICTZON AND LUBRICATION.37part of the friction is generally due to the shearing of metallic junctionsformed through the lubricant film.Extreme Pressure hbrication.The recent development of extreme pressure lubricants has been broughtabout by the failure of conventional mineral oil lubricants t o functioneffectively a t the high pressures and temperatures developed in certainmechanisms, such as hypoid gears, heavy machining operations, etc. Duringthe last ten years, a large number of organic compounds containing " active "elements such as sulphur, phosphorus, and chlorine have been used for thispurpose. Full reviews of the numerous chemicals used are given byJ. Byers,140 M. G. Van Voorhis,l41 and W. A. Wright,142 and a review of themore common extreme pressure additives is given by E.A. Evans andJ. S. Elliot.143 I n a later paper,144 Evans discusses the chemical processesinvolved in the preparation of these compounds.The earlier work was concerned mainly with the empirical preparationof additives which functioned effectively in. practical operations. Morerecent work has been directed to elucidating the mechanism by which theseextreme pressure lubricants function. Hughes and Whittingham,go andmore recently J. N. Gregory145 and W. Davey,lP6 have found that thinsulphide and chloride films are effective in reducing the friction and wearbetween steel and copper surfaces. Campbell 123 showed that relativelythick films of sulphide reduce the friction between steel and copper surfaces,particularly in the presence of paraffin oil, and E.B. Greenhill 14' showedthat the presence of a small quantity of fatty acid in the paraffin oil produceda very much larger reduction in friction. These results explain the wideuse of sulphurised fatty oils as extreme pressure lubricants.The film-forming properties of organic sulphur compounds have beeninvestigated by G. L. Simard, H. W. Russell, and H. R. Nelson,14* usingelectron-diffraction methods. I n particular, they find that a lubricantcontaining free sulphur forms an oxide layer on iron surfaces. They havealso examined the film-forming action of a lubricant containing lead naphth-enate and free sulphur, and find that in many cases the sulphide film isformed. Using a specimen of sulphurised oil prepared with radioactivesulphur, G.L. Clark, S. G. Gallo, and B. H. Lincoln 149 have demonstratedthe formation of a film on various metals and on glass, though the actualcomposition of the film was not investigated.Similar experiments have been carried out on compounds containingchlorine. In particular, Gregory 145 has shown that chloride films are very140 Nat. Pet. News, 1936, 33, 79.142 Ibid., 1945, 37, R34.144 Ibid., 1943, 29, 333.146 C.S.I.R. (Australia), 1945, Ser. no. A , 134, No. 49.146 J . Inst. Petroleum, 1945, 31, 73, 154.ld7 C.S.I.R. (Australia), 1944, Ser. no. A , 97, No. 36.148 I n d . Eng. Chem., 1941, 33, 1352.141 Ibid., 1940, 32, R66.143 J . Inst. Petroleum, 1941, 27, 165.149 J . Appl. Physics, 1943, 14, 42838 GENERAL AND PHYSICAL CHEMISTRY.effective in reducing the friction between steel and copper surfaces in theabsence of moisture.Furthermore, chlorine compounds are effective onlywhen the formation of the metal chloride, by chemical reaction with themetal surface, is possible.It is clear from the results of these workers that the sulphur and thechlorine type of extreme pressure lubricants function by forming a sulphideor metallic chloride on the surface of the rubbing metals. Under the highpressures and temperatures developed between the rubbing surfaces, thecompounds break down, and the " active " portion of the molecule com-bines with the metal surface. These surface films are capable of preventingintimate metallic contact between the surfaces and so reduce the amount ofseizure and wear.That is to say, in terms of equation (1) they produce asmall value of u. With sulphide films, the shear strength s is not, appar-ently, very low, so that although the ability to withstand seizure is greatlyincreased the friction is not appreciably lowered unless fatty acids are alsopresent. With chloride films, however, both s and a are low, so that boththe friction and the probability of seizure are greatly reduced.Interesting confirmation of this general mechanism of film formation hasbeen furnished by 0. Beeck, J. W. Givens, and E. C. Williams I5O in theirwork on the wear-reducing properties of lubricants containing phosphorus.They showed that these compounds, under the action of high runningpressures and temperatures, form phosphide films on the metal surfaces,which then alloy with the metals themselves, producing a low-meltingeutectic.The surface asperities are thereby removed by " chemical "polishing and subsequent wear is very greatly reduced. These workers havesuggested a similar mechanism for the action of lubricants containing sulphur.The Lubrication of Internal Combustion Engines.It has been customary in the past to examine the conditions oflubrication and wear in a running engine by long- or short-range benchtests. In these tests, the engine is run, under close control of the run-ning conditions, and the total energy lost in friction estimated, whilstmeasurements may be made of the oil consumption, compression ratio,piston and cylinder-head temperatures, etc.151, 1 5 2 9 1537 1549 155 After therun is completed, the wear of the piston ring and cylinder liner, and theamount of scuffing and corrosion, are measured. In addition, an ex-amination is made of any physical or chemical changes that may have160 Proc.Roy. SOC., 1940-41, A, 177, 103.1 5 1 H. Wright Baker, Inst. Mech. Eng., 1934, 1i7, 217; 1937, 135, 35-67; Proc.Inst. Automobile Eng. 1932-3, 27, 109; 1934-5, 29, 312.G. F. Mucklow, Inst. Mech. Eng., 1932, 123, 349.153 A. H. Gibson, ibid., 1926, 221; Phil. Mag., 1924, 47, 883.154 Saharo and Sato, Tokyo Imp. Univ. Aeronautical Res. Inst. Report No. 5;L. C. Tyte, Inst. Mech. Eng. Symposium on Modern Aids t o the Investigation ofMaterials, 1944, 16.166 W. L. Bride, J. Inst.Mech. Eng., 1943, 150, 134; 1944, 151, 338BOWDEN AND TABOR : FRICTION AND LUBRICATION. 39occurred in the lubricant. For refined chemical analysis, colorimetric lS6 orradioactive tracer methods 1493 157 may be used.It is, of course, evident that the friction and wear behaviour of a runningengine will be profoundly influenced by the nature of the lubricant, andthe rubbing surfaces, as well as by the running conditions themselves.Experiments also show that it depends markedly on the surface finish of thecylinder and piston ring, since this determines the ease with which thesurfaces are run in a t an early stage of 0peration.15~ A tapered ring hasbeen found advantageous for the same reason.159 More recently, it has beensuggested that the surface finish of the cylinder liner may determine theextent to which the surface will retain a thin film of oil whilst the engine isrunning.l6O> 161The wear which occurs in a running engine may be conveniently dividedThe most outstand-ing research on corrosive wear has been carried out by C.G . Williams,l62who has shown that this type of wear is the predominant factor in engineswhich are frequently started and stopped, and is due to the depositionon the cylinder walls of acids and moisture resulting from the productsof combustion. This work has been generally confirmed by otherw0rkers,16~, 1G43 165, 166 and there is general agreement that a significantreduction of corrosive wear may be obtained by using corrosion-resistantcoatings on the cylinder walls, such as chr~rne-plating,~~~ surface harden-ing,16B etc.Some anti-corrosive coatings may also reduce the amount ofabrasive wear.A more direct investigation of the abrasive wear between the piston ringand cylinder-wall of an internal combustion engine has recently been. into two categories, corrosive wear and abrasive wear.166 H. A. Everett and G. H. Keller, Inst. Mech. Eng., General Discussion on Lubric-ation, 1937, Vol. I, 451-6, 627-8; H. A. Everett and F. C. Stewart, Penna. StateCollege Eng. Expt. Sta., 1935, Ser. Bull. No. 44, 52 pp. ; G. H. Keller, Automotive Ind.,1935, 72, 484.16' S. W. Ferris, U.S. Patent No. 2,315,845.168 W. H. Spencer, Steel, 1938, 103, No. 23, 60; M. M. Roensch, J. SOC. Aut. Eng.,1940, 46, 2 2 1 ~ ; F.Bremer, Korrosion u. Metallschutx, 1941, 17, 208.169 A. Taub, Inst. Mech. Eng., General Discussion on Lubrication, 1937, Vol. I,572-6; J . Inst. Mech. Eng., 1939, 141, 87; 429; &l. Andreev, Teoriga i Prakt. Met.,160 A. Cyril Yeates, Inst. Mech. Eng., General Discussion on Lubrication, 1937,161 Anon., Iron A g e , 1941, 148, 57; E. L. Hemingway, ibid., 1942, 149, 40; AlC2 Collected Researches on Cylinder Wear, Inst. Auto. Engrs., Auto Research163 H. Kjolsen, Ingenwyen, 1936, 45, iv, 52, 71; Chisn. et Ind., 1937, 38, 255.16* Inst. Mech. Eng., General Discussion on Lubrication, 1937, Vol. I, Group 11.165 V. S. Prever, Ind. Meccanica, 1935, 17, 489.166 R. A. Collacott, Power and Works Engineer, 1942, 51.16' H. Van Der Horst, Metal Ind.(N.Y.), 1940, 38, 76; Automobile Engineer, 1941,16* F. P. Peters and E. F. Cone, Metals and Alloys, 1941, 13, 713.1937, NO. 5, 59-71.VOl. I, 595.Composite, " Metals and Alloys," 1942, 15, 322, 326.Committee, ~ 9 4 0 .31, 405; Metal Finishing, 1942, 40, 6940 QENERAL AND PHYSICAL CHEMISTRY.undertaken by workers who have investigated the effect of the metallurgicalstructure of the rings and cylinder on the wear.169, 170, 171, 172, 1739 174 Mostof the workers in this field have emphasised the difficulty of applying ideal-ised experimental wear data to the practical case of a running engine, sincethe laboratory tests are usually carried out under conditions which are veryfar indeed from those which apply to a running engine. This difficultyhas long been felt in engine research, and for this reason special interestattaches to the work of R.Poppingal75 and of J. S. Courtney-Pratt andG . K. These workers have nieasured, by a cathode ray technique,the electrical resistance between the piston ring and cylinder wall of arunning engine. At the top and bottom dead-centre, the resistance is verylow, indicating that there is appreciable metallic contact, or at best aregime of boundary lubrication. At the centre of the stroke, the resistanceis high, implying that the conditions are largely those of fluid lubrication.These results are in agreement with those of C. A. Bouman177 and~ t h e r s , l ~ ~ , 179 though different views have been expressed by otherworkers.f80~ls1, 182s 1839 184 However, the electrical measurements show thatthere is intermittent contact between the surfaces a t all stages of the cycleso that a t no part of the stroke are the surfaces separated by an unbrokenfilm of lubricant.This technique provides an analytical method of in-vestigating the lubricating conditions which operate while the engine isrunning and the way in which these conditions are affected by the temper-ature, viscosity, compression ratio, and other variables.The effect of temperature on the lubrication of a running engine isextremely important. High temperatures, which are readily reached, notonly reduce the viscosity of the oil, but lead to a deterioration of itslubricating properties.126, 127 In addition, persistent high temperatures lead168 H.J. Young, Inst. Mech. Eng., General Discussion on Lubrication, 1937, Vol.170 J. G. Pearce, Inst. Mech. Eng., General Discussion on Lubrication, 1937, Vol.171 A. Wallichs and J. Gregor, Biesserei, 1933, 20, 517, 548.17a E. Soehnchen and E. Piwowarsky, Arch. Eisenhiittenw., 1933, 7 , 371.173 Paul S. Lane, Metal Progress, 1941, 39, 315.174 August Gimmy, Automobiltech. Z., 1939, 42, 334.176 Automobiltech. Z . , 1941, 44, No. 10, 247; 1941, 44, No. 11, 272 (R.T.P. Trans-lation Nos. 1505 and 1603, Ministry of Aircraft Production).176 C.S.I.R. (Australia), 1944, Bulletin No. 179; Engineering, 1946, 161, 69 ; Inst.Mech. Eng., Paper and Discussion, Nov. 30, 1945 (to be published shortly).17? Inst. Mech. Eng., General Discussion on Lubrication, 1937, Vol. I, 426-431.lI8 H.A. Everett, J. Soc. Aut. Eng., 1943, 51, 1 6 5 ~ .17$ R. A. Castleman, Physical Rev., 1936, 49, 410, 886.180 H. R. Ricardo, Automobile Engineer, 1922, 12, 304.lS1 T. E. Stanton, Aeronautical Research Committee, 1924, Report No. 931.lSa S. W. Sparrow and M. A. Thorne, National Advisory Cttee. for Aeronautics,lE3 C. J. Hawkes and G. F. Hardy, Trans. North East Coast Inst. Eng. and Ship-la4 M . P. Taylor, J. SOC. Aut. Eng., 1936, 38, 200.I, 599-606; Proc. Inst. Auto. Engrs., 1935-6, 30, 69.I, 546-549.1927, Report No. 262, Pt. 2.builders, 1936, 52, 143BOWDEN AND TABOR : FRICTION AND LUBRICATION. 41to oxidation and polymerisation of the lubricant. Oxidation gives rise tothe formation of acidic products which may, a t a very early stage, impartimproved boundary lubricating properties to the 0il.129,185 At a later stage,however, these acidic products lead to corrosive wear of the cylinder wallsand piston rings.Oxidation combined with polymerisation causes anincrease in the viscosity of the lubricating oil and the formation of productswhich are insoluble in the bulk of the oil. These products appear in theform of sludge or as gum-like deposits which give rise to sticking of thepiston rings and valve stems.186, 18'9 l88, 1B9 Thus oxidation and polymeris-ation lead to increased ring and cylinder wear, ring sticking, and theaccumulation of sludge.The Use of Additives and Polymers.-It is now common practice toadd substances in small proportions to the lubricating oil, in order tocounteract the effect of oxidation.Such additive agents include a widerange of substances.lW Earlier examples are hydroxy-aromatic compoundssuch as phenols and naphthols and their derivatives; nitrogen com-pounds such as amines ; organic compounds containing sulphur, chlorine,or phosphorus ; and organometallic compounds.The additive may interfere with the oxidation reaction thus retardingthe effect of oxidation, while, in addition, the inhibitor may render themetallic surface passive to corrosion. Additives containing hydroxyl oramino-radicals, such as phenylnaphthylamine, usually react with oxygen,causing a delay period in the oxidation of the lubricating oil, and compoundssuch as tributyl phosphite and tritolyl phosphate react with the metalpresent to form a protective coating.Various organic sulphur compoundsare thought to act in both ways, that is, as anti-oxidants and as metalpassifiers. In addition, some of the additives may possess detergent pro-perties; that is to say, they may tend to take up sludge and carboniferousdeposits and so reduce the amount of ring sticking.Some of the more recent anti-oxidants which have been used as additivesfor lubricating oil are : (1) Zinc salt of diisopropylsalicylic acid (" ZincDips ").191 (2) Polycarboxylic acids of high molecular weight formed bycondensing alkylenes with maleic acid or anhydride followed by saponific-a t i ~ n . l ~ ~ (3) Calcium pheny1~tearate.l~~ (4) Sulphonate salt, as ofpetroleum sulphonic acid and a multivalent metal (e.g., calcium) andarylamine, e.g., phenylnaphthylamine.lS4 ( 5 ) Chelated barium salt of anle5 C.H. Barton, Inst. Mech. Eng., General Discussion on Lubrication, 1937, Vol. I,lS6 C. F. Prutton, Inst. Spokesman, 1941, 5, No. 9, 5-8.407-413.H. W. Brownsdon, Inst. Mech. Eng., General Discussion on Lubrication, 1937,lee H. R. Luck, T. A. Rogers, and A. G. Cattaneo, J. Soo. Aut. Eng., 1943, 51, 38.lE9 33. W. J. Mardles and J. E. Ramsbottom, Inst. Mech. Eng., General DiscussionVol. 11, 2 5 6 2 6 0 .on Lubrication, 1937, Vol. 11, 354-366.M. W. Webber, Petroleum, 1945, 8, No. 4, 76.lgl U.S.P. 2,258,591. lg2 U.S.P. 2,124,628; B.P. 488,597.lg3 B.P. 509,097. 19* U.S.P. 2,270,577.B 42 GENERAL AND PHYSICAL CHEMISTRY.alkylated phenol disulphide (I).lS5 (6) Alkylated p-cres01.1~6 (7) Trichloro-benzene, hexachlorodiphenyl oxide, and inhibitor (trichlorotolyl phosphiteC,H,l R R/ \ / /7\-/-cH2-<7)/-\ T3&s- L/\O-Ba-O/ \O-CH2-O/(1.1 (11.)or phosphate).lS7 (8) Pormols (II).lS8 (9) Sulphurised cracked wax.199(10) pp'-Dichlorodiphenyl disulphide.200 (1 1) Reaction product of tritolylphosphite and octylphenoxyethanol.201 (12) A blend of basic calciumphenylstearate, a solubiliser such as lauryl alcohol or thiol, and athioamide, e.g., thiobenzanilide.202 (13) A combination of additives such as" Zinc Dips " combined with calcium hydro~ypetronate.~~3I n the operation of aero-engines, as distinct from automobile engines,another source of trouble is the formation of a stable foam in the lubricatingoil, particularly a t high altitudes.The presence of foam may cause apartial stoppage in the flow of oil through the supply channels and mayalso cause considerable loss of oil by leakage through the breathing outlets.The use of addition agents to keep foaming down to a minimum hasbeen known for some time. The higher alcohols, such as octyl alcohol, havebeen used in small proportions in industry to prevent heavy foaming duringchemical reactions.204 Anti-foaming additives have been used successfullyin lubricating but there is not much information available on theactual additives in use. The use of silicones has been claimed,206 andthey are likely to be used extensively for this purpose. Other anti-foaming additives which have been proposed are potassium oleate in sul-phurised sperm oi1,206a and compounds such as barium diethylhexyl orditetradecyl dithiophosphates.206b The latter compounds are made byreaction between phosphorus pentasulphide and branched-chain alcohols ormixtures of alcohols and ketones a t 90--100".In many types of moving mecha,nisms lubricated with petroleum oils,the dependence of the viscosity of the oil on the temperature is often aserious disadvantage.For this reason, considerable attention has recentlybeen aroused by the development of polymers containing silicon, which showextremely small variations of viscosity with temperature. These compoundsare, in general, long-chain or ring polymers derived mainly from diorgano-silanediols, with a wide range of physical properties ranging from liquids of195 U.S.P.2,139,766. lo6 U.S.P. 2,202,825. 197 U.S.P. 2,204,620.198 U.S.P. 2,250,188. lg9 U.S.P. 2,215,132. U.S.P. 2,153,432.201 U.S.P. 2,280,450. 202 U.S.P. 2,252,793. 203 U.S.P. 2,373,411.204 Ind. Eng. Chem. (News Edn.), 1935, 16, 389 ; Ann. Reports on Applied Chemistry,205 H. A. Ambrose and C. E. Trautman, J. SOC. Aut. Eng., 1945, 53, 373.206 Nat. Pet. News, 1945, 37, No. 49, 9 4 5 ~ .2060 U.S.P. 2,377,654.1938, 23, 122.206b U.S.P. 2,368,000BOWDEN AND TABOR : FRICTION AND LUBRICATIOX. 43viscosity lower than that of water to those possessing the consistency of thickgreases. These polymers are relatively non-volatile and are extremelyresistant to decomposition and oxidation.They are serviceable up to tem-peratures of over 250" and may possess pour-points lower than - 70".The boundary lubricating properties and wear-resisting characteristics ofsilicones are in general poor, and for this reason they have been used ashydraulic fluids and for mechanisms working mainly under conditions offluid lubrication. However, by suitable chemical modifications, siliconeproducts have recently been prepared which have slightly better boundarylubricating properties than straight mineral oils. Silicones and other syn-thetic polymers such as polyethylene oxides should find increasing use in anumber of practical applications. With most polymers the viscosity de-creases with increasing rate of shear, and in some cases there may be apermanent increase in viscosity due to a breakdown of the polymer.Ifthis occurs to a marked degree, it may impose a limitation on their practicaluse. A recent review of the use of silicones as lubricants is given by T. A.Kauppi and W. W. Pedersen.206Chemical Decomposition by Friction.Although a considerable amount of qualitative work has been done onchemical decomposition produced by friction, little is known concerning themechanism by which these reactions take place. The earlier work wasconcerned with the decomposition of solids (mainly endothermic) by highpressure 2077208 and by grinding in a mortar with a 209 where the solidis subjected to both pressure and shear. Carey Lea 208 investigated thedecomposition of solids such as AgCl+ Ag, HgO ---+ Hg, KMnO, +MnO,, and was able to show that decomposition by pressure was facilitatedby a shearing motion.During the shearing, frictional heat is developed,but he considered that this heat played little or no part in the decomposition.L. H. Parker 210 suggested that, when solids are ground together, reactionsof the type HgC1, + 2KI -+ HgI, + 2KC1 take place owing to the localor surface melting of the solid which results from the stress. In theseexperiments dry salts must be used since the presence of minute traces ofwater vapour affects the results ~onsiderably.20~~210I n a series of papers, P. W. Bridgman 211 has described experiments inwhich he subjected many compounds to hydrostatic pressures up to50,000 kg./cm.(or 50,000 atm.) combined with a shearing stress up to theplastic flow pressure of the material. Under these extreme conditionsmany compounds decomposed explosively, e.g., iodoform, silver nitrate,207 W. Spring, Bull. Soc. chim., 1885, 44, 166; 1886, 46, 299; Z.physika1. Chem.,208 M . Carey Lea, Phil Mug., 1891, 34, 46; 1893, 36, 351; 1894, 37, 31, 470.209 E. P. Perman, Chem. News, 1903, 88, 197; 1907, 96, 3.*lo J., 1914, 105, 1504; 1918, 113, 396.211 Physical Rev., 1935, 48, 825; J . Geol., 1936, 44, 653; Proc. Amer. Acad., 1937,1888, 2, 532, 536.71, 387; Amer. J . Sci., 1938, 36, 8144 GENERAL AND PHYSICAL CHEMISTRY.lead dioxide. Reactions between copper and sulphur, and silicon andmagnesium oxide, also took place with explosive violence.Two phenomena which are of importance in connection with frictionaldecomposition are (i) the effect of pressure on the melting point of solids,212$ 213and (ii) the production of local high-temperature flashes at surface boundarieswhen two solids are rubbed together.26~37~40~214~215 In discussing the effectof pressure on the melting point of solids, Johnston and Adams 212 havepointed out that the application of a uniform pressure to a solid-liquidphase has a comparatively slight effect on the melting point and, in fact,usually raises it by ca.10-30" per 1000 atm. On the other hand, in asystem where there is a non-uniform pressure on the solid-liquid phase, i.e.,where there is an excess pressure on the solid phase, a lowering of the meltingpoint is always obtained.213 Non-uniform compression may be visualised,for example, in the grinding of solids in a mortar with a pestle.Thisgrinding results in a melting of the surface layers a t the crystal boundarieswhere the reaction occurs. The liquid formed by melting flows into inter-stitial spaces and in this way becomes subjected to a smaller pressure thanthe adjacent solid particles. The reaction products are then removedduring the grinding, and fresh surfaces of the reactants are continuallyexposed.It has also been suggested that the decomposition is due to the frictionalheat developed when two surfaces rub together. Many experimental findingssupport this concl~sion.~~~ 40$ 41 The reduction of polishing powders suchas red lead and lead dioxide and the decomposition of calcium carbonatewhen these powders are used to polish metal surfaces has been ascribed tohot spots produced during the polishing.41~214When metals such as iron, copper, and nickel are polished or rolled,oxidation of the metal surface occurs.216% 2179 2189 2191 220 The extent of thewear of the rubbing surfaces is closely bound up with this oxidation.Ina nitrogen 216 or carbon dioxide 217 atmosphere, the wear is considerablydecreased. The oxidation in air may be accelerated by the localised surfaceh e a t h ~ g . ~ ~ , ~ ~ K. Dies 217 has also associated wear with high temperaturesproduced at the points of metallic contact. It is possible that the mechan-ical deformation and breakdown of the protective oxide film also acceleratesthe surface attack.The theory proposed by Fink and Hoffmann 216 is that212 J. Johnston, J . Amer. Chem. SOC., 1912, 34, 788; J. Johnston and L. H. Adams,21s See also H. Jeffreys, Phil. Mag., 1935, 19, 840.216 J. D. Bernal, Trans. Faraday Soc., 1938, 34, 834, 1008.216 M. Fink and U. Hoffiann, Chem. Abs., 1933,27, 1598; 1935, 29,432; 2. anorg.217 Chern. Abs., 1939, 33, 4180; 1943, 37, 5680;218 See also F. Wunderlich, ibid., 1942, 36, 3465.21s K. Lippacher, ibid., 1943, 37, 6226.220 S. Dobinski, Phil. Mag., 1937, 23, 397; see, however, E. Plessing, Physikal. Z.,Amer. J . Sci., 1913, 55, 205.J. M. Macaulay, J . Roy. Tech. Coll. Glasgow, 1931, 2, 378.Chem., 1934, 210, 100; see also F. Roll and W. Palewka, ibid., 221, 177.1944, 38, 708.1939, 40, 233BOWDEN AND TABOR : FRICTION AND LUBRIOATION.45during polishing or rolling chemically active centres are produced on themetal which readily oxidize in air, and this oxidation may penetrate to aconsiderable depth causing scaling of the outer layer.218The initiation of some solid and liquid explosives by friction is thoughtto be due to the development of localised hot spots. The temperature ofthese hot spots is sufficient to bring about a thermal decomposition of theexplosive in the neighbourhood of the hot spots, and this decomposition willdevelop by a process of self-heating into a detonation. Direct experimentalevidence for this view has been obtained by Bowden, Stone, and Tudor 40with liquid explosives such as nitroglycerin.They found that, when thiswas rubbed between solids, the incidence of explosion was determinedmainly by the thermal conductivity of the solids and by their melting points.Explosion resulted only if the melting point was above 480". Below 480"explosion of nitroglycerin could not be obtained, even under severe conditionsof load and speed.Impact experiments on solid explosives (mainly initiating explosives) byW. Taylor and A. Weale 221 have led them .to postulate a tribochemicalmechanism as the cause of initiation in solid explosives. Under the suddenlyapplied pressure of the impact (1 100 atm. in the case of mercury fulminate)the explosive crystals are subjected to normal stress forcing them into moreintimate contact and to tangential stresses tending to shear the crystalsapart. As a result of normal stress, the molecular fields of the surfacemolecules are thrust together and linkages formed.These are almostimmediately ruptured under the tangential stresses, with the result that thesurface molecules are left in highly activated states.222i 223 Experiments onliquid and plastic explosives have shown that in certain cases the initiationby gentle impact may in fact be a thermal one, due to the adiabatic heatingof small entrapped bubbles 224 and, under heavy impact, by a viscous heatingof the explosive.225, 226The effect of grinding on the transformation of the yellow modificationof lead oxide into the red modification has been studied in some228, 2293 230, 231 Clark and his co-workers 230 found changes in theX-ray diffraction patterns and in the catalytic activity of yellow ortho-221 Proc.Roy. SOC., 1932, A, 138, 92; Trans. Paraday Soc., 1938, 34, 995.222 See review by M. F. R. Mulcahy and A. Yoffe, Aust. Chem. Inst. J . Proc., 1945,223 L. R. Carl, J. Franklin Inst., 1943, 235, 553; 1940, 230, 75, 355.224 F. P. Bowden, M. F. R. Mulcahy, R. G. Vines, and A. Yoffe, Nature, 1946,225 T. M. Cherry, C.S.I.R. (Australia), Lubricants and Bearings Sect., 1945, Report2 z 6 F. Eirich and D. Tabor, ibid., Report A121.227 0. W. Brown, S. V. Cook, and J. C. Warner, J. Physical Chem., 1922, 26, 477.228 M. Leblanc and E. Eberius, 2. physikaE. Chem., 1932, A, 160, 69.229 G. Tammann and E. Jenckel, 2. anorg. Chem., 1930, 192, 245.230 G.L. Clark and R. Rowan, J . Amer. Chem. SOC., 1941, 63, 1302; G. L. Clark231 M. Peterson, ibid., 1941, 63, 2617.62, 198.157, 105.A116.and S. F. Kern, ibid., 1942, 04, 163746 GENERAL AND PHYSICAL CHEMISTRY.rhombic lead oxide when subjected to grinding, and conclude that it isconverted into a distorted red tetragonal modification of the oxide. Thepresence of small traces of water vapour appears to have a profound effectin facilitating this tran~formation.~3l It is evident that a considerableamount of work remains to be done before we have a clear understandingof the mechanism of tribochemical or frictional decomposition. At presentthe evidence shows that in certain cases the decomposition is in reality athermal one, due to the high localised temperature flashes produced duringthe rubbing of solid surfaces.We wish to thank Messrs.G. C. Barker, J. A. Burns, J. S. Courtney-Pratt,E. B. Greenhill, A. J. W. Moore, E. D. Tingle, A. Yoffe and other membersof the Research Laboratory on the Physics and Chemistry of Rubbing Solids,Department of Physical Chemistry, Cambridge, for assistance in preparingthis review, and Mr. R. I. Lewis for help with the section on additives.F. P. B.D. T.3. CRYSTALLOGRAPHY.(i) Introduction and General.From a statement now issuedl it is clear that during the recent waryears X-ray crystallography has achieved one of its greatest triumphs in thecomplete elucidation of the structure of penicillin. Details are not yetavailable, apart from the statement that a full electron distribution of therubidium salt of penicillin I1 has been obtained.This implies a full structuredetermination. Although the molecule is not unduly large, containing about25 atoms other than hydrogen, this result is significant of the rapidly growingpower of the X-ray method. The fact that it has been obtained in such ashort time from a recently isolated natural product emphasises the importanceof the X-ray method as a research tool in structural organic chemistry,especially where instability of the molecule or unusual groupings make theusual degradative processes difficult to interpret and entirely independentconfirmation of a structure is badly wanted. Further details of this workare awaited with great interest.During the .year a number of very detailed and complete structuredeterminations by the X-ray method have been published, mainly in thefield of organic structures, and the space available in this Report has beendevoted mainly to outlining the results obtained.Of these determinationsthe work on cholesteryl iodide, geranylamine hydrochloride, and dibenzylis particularly interesting because three-dimensional Fourier series methodshave been extensively employed. It has, of course, long been apparent thatsuch methods must yield far more detailed and accurate information aboutatomic distributions than the ordinary two-dimensional projection method.The latter method is very suitable for planar molecules, such as found in thearomatic hydrocarbons, but when the molecule has a more complicated shape1 “ Chemistry of Penicillin,” Nature, 1946, 156, 766ROBERTSON : CRYSTALLOGRAPHY.47no projection of it is likely to yield a clear picture of all the atomic positions.One difficulty in applying the three-dimensional method lies in the veryformidable amount of numerical calculation required. It is noteworthythat in one of the examples mentioned (dibenzyl) the help of a professionalcomputing service was required. With the provision of more facilities inthis direction and the development of mechanical aids we may expect thethree-dimensional method to be more frequently employed in the future.A far more serious obstacle to such work lies in the initial determinationof the structure; because it must be remembered that in the general caseany Fourier series method is only a means of refining a structure whoseco-ordinates are already approximately known? There is, however, onedirect method of approach which can have a very widespread applicationto complex organic structures.This method depends on the presence a t oneor more points in the structure of atoms whose atomic number is muchhigher than the atomic numbers of the remaining atoms. I n effect thisconverts the unknown phase differences of the X-ray reflections into differ-ences of amplitude, which can be measured; or, in terms of the Pattersonanalysis, it produces a set of vectors, between the heavy atom and each ofthe other atoms, so prominent that they outweigh the confusing mass ofother secondary vectors between the light atoms themselves.The result issome approach to a direct picture of the molecular structure, without anyassumptions based on chemical theory. This powerful method is beingsteadily developed and has been applied to the first two structures mentionedabove, vix., cholesteryl iodide and geranylamine hydrochloride. Fororganic structures in general the halogens, especially iodine, are likely toprovide the most useful heavy atoms; for acids, salts with heavy metalsare possible.The usual indirect method of approach to a crystal structure starts witha model based on what is known of the chemical structure. The orientationof this model in the crystal unit cell must be found by trial, and, afterapproximate agreements have been obtained for the intensities, the Fourierseries method can be applied.The dibenzyl structure was obtained in thisway, and was refined as far as possible by means of two-dimensional pro-j cctions before the three-dimensional analysis was applied.When the Fourier series method is used at its full power, as in theseexamples, the accuracy obtainable is undoubtedly high, provided that acomplete series based on carefully determined intensities has been used. Itappears that the atomic co-ordinates may be obtained to within &O.Ol or*0.02 A.* With this order of accuracy the results should ultimately be ofgreat importance in the development of valency theory. On the other hand,when it is only desired to establish the structure of a molecule in the chemicalsense of finding the relative spatial positions for all the atoms, there isno need to push the work so far, and somewhat less complete series andSee, e.g., D. Macewan and C.A. Beevers, J . Sci. Instr., 1942, 19, 150.For a discussion of principles see J. M. Robertson, J., 1945, 249.A. D. Booth, Nature, 1945, 156, 5148 GENERAL AND PHYSICAL CHEMISTRY.less accurate intensities will suffice. Cholesteryl iodide is a case in point,where the accuracy is sufficient to provide an unambiguous view of thestructure, but not sufficient for a detailed discussion of bond lengths.Two interesting new modifications of the Fourier series method haverecently been devised by A. D. Booth.5 These are called the method ofsection-projections and the method of projected sections and they possessadvantages which are intermediate between the standard two-dimensionaland three-dimensional methods, without requiring such formidable com-putations as are needed in the latter.By the use of section-projections amolecule may sometimes be separated from its companions in the unit celland so lead to a projection which is free from overlapping effects. Theprojected sections, on the other hand, have been devised mainly to reducethe amount of computation involved in making a complete set of ordinarysectional syntheses.Other work published during the year includes a detailed analysis of thecoronene structure, which shows interesting bond length variations withinthe molecule ; the complete analysis of diphenylene, which confirms that thecompound is actually dibenzcyclobutadiene ; and detailed work on certainamino-acids and on adipic acid.Amongst inorganic structures the mostinteresting work is that of Powell and Bartindale on hexamethylisocyanido-ferrous chloride trihydrate, which yields a beautiful projection and givesdetailed information about the structure of the ferrocyanides.The year has also been notable for the publication of a number ofimportant books and monographs on structural crystallography and crystalchemistry. C. W. Bunn has covered the subject of chemical crystallographyin an eminently practical manner, dealing Grst with optical and X-raypowder methods for the identification of substances, and then, in the longersection of the book, with the determination of atomic positions.Thevarious methods of analysis are fully discussed and illustrated with manyexamples of structure determinations. Much of this is new ground whichhas not previously been covered in any book. The widespread applicationsof the subject to chemical problems in general are very well brought out.The more specialised application of X-ray methods to metallurgicalproblems has been fully treated in a book by A. Taylor.’ Brief introductorychapters on the standard methods and principles and the crystal structuresof metals are followed by a full account of applications to metallographicproblems. The subjects covered include the determination of phaseboundaries, defect lattices, electron compounds, superlattices, order-disordertheory, the examination of binary and ternary systems, and an account ofrecent X-ray work on the iron-carbon system.The determination of grainsize by optical and X-ray methods and the study of grain orientation arenext described, and there is a brief account of the X-ray study of refractoryTrans. B’amday SOC., 1945, 41, 434.“ Chemical Crystallography,” Oxford University Press, 1946.“ An Introduction to X-Ray Metallography,” Chapman and Hall, Ltd., London,1945ROBERTSON : CRYSTALLOGRAPHY. 49materials. Much of the ground covered can only be found otherwise inmany scattered original papers, so the book ought to prove extremelyuseful.The much wider field of structural inorganic chemistry is covered in thesurvey given by A.F. Wells.8 The modern basis of this subject dependsvery largely on X-ray crystallographic studies. It is now realised that thefinite molecules studied by the classical methods of chemistry are only asmall part of the subject, which must be extended to include the infinitethree-dimensional arrays of atoms which constitute solids. This extensionof the field of inorganic chemistry forms the main thesis of the book. Aftera general introduction dealing with atomic structure, interatomic forces,the spatial arrangement of atoms in relation to bond type, and other relevantmatters, it goes on to treat the subject in a systematic manner in whichthe results of crystal chemistry figure very largely. The arrangement isaccording to the usual groups; hydrogen (including the acids and certainstructures involving the hydrogen bond), the halogens, oxygen and sulphur(three chapters), nitrogen and phosphorus, carbon, silicon, and boron.Finally, two short chapters deal with the stereochemistry of certain metalsand the crystal structure of metals and alloys.Again, this book bringstogether and systematises a great deal of otherwise scattered material andi t thus represents an important contribution to the literature of the subject.Many papers have also appeared recently on the more physical aspectsof crystallography, and on the structure of metals and alloys, mineralstructures, fibre structures and other less completely crystalline materials,but a review of the work in these fields is not included in the present Report.(ii) Inorganic Structures.The Iron-Carbon Bond in the Ferrocyanides.-A very interesting crystalstructure determination has been carried out by H.M. Powell andG. W. R. Bartindale 1 on hexamethylisocyanidoferrous chloride trihydrate,Fe(CNMe),C1,,3H20. The hexagonal crystal has only one molecule of thiscomposition per unit cell, and so the iron atom must occupy a special position,taken as the origin of the xy co-ordinates. Its contribution to the structureamplitudes is sufficient t o determine the phase constants and enable a directand very striking Fourier projection of the structure to be made on the basalplane of the hexagonal cell. The atoms are all beautifully resolved, but acurious ambiguity is discovered.The structure contains certain two-foldand three-fold positions which one would expect to find occupied by thetwo chlorine ions and the three water molecules respectively. Actually, theelectron-density peaks show that the reverse is true, Le., the chlorine appearsto be where the water was expected and vice versa. I n the isomorphousbromide compound similar and even more striking discrepancies occur.The explanation must be that there is a random distribution of the twohalogen ions among the three-fold positions, and, of the three watera “ Structural Inorganic Chemistry,” Oxford University Press, 1945.J . . 1945, 79950 GENERAL AND PHYSICAL CHEMISTRY.molecules, two must occupy the two-fold positions, and the remaining onebe distributed a t random among the three-fold positions.Each of the three-fold positions therefore holds statistically $Cl and QH,O. This curiousdisorder effect is due to the general architecture of the structure, whichpermits a more compact grouping of the chlorine ions with respect to thepositive ion, and consequently a lower potential energy, if the disorderedarrangement is adopted.The complex ion Fe(CNMe),++ (I) is found to be an octahedral arrange-ment with Fe-C distances of 1-85 A., indicating about 50% double-bondcharacter. The FeCNMe arms have a bend of about 7" at the nitrogenatom, confirming resonance of the two types Fe::C::N:Me and Fe:C:::N:Me.The C-N distance is 1.18 A. and the N-Me, 1-47 A.Me -/Me N+++The octahedral complexes are packed as closely together in the structureas normal separation between the methyl groups (3.70 and 3-91 A.) willallow.With this arrangement, holes are left in the structure of sufficientsize to accommodate either chlorine ions or water molecules, and these areoccupied in the random manner indicated above.8uZphides.-New determinations of the structures of a number of sulphidecompounds have recently been made. Potassium thioferrite,2 KFeS,, ismonoclinic (pseudohexagonal) and consists of chains of (FeS,),, iron beingat the centre of almost regular tetrahedra of sulphur atoms. The iron-sulphur distances are 2.20 and 2-28 A., and the tetrahedron edges 3.62-3.74 A. Potassium occupies interstices in the structure, and is surrounded inan irregular way by eight sulphur atoms, a t distances of from 3.33 to 3-48 A.Sodium thiochromitle,2 NaCrS,, has a rhombohedra1 layer lattice, andconsists of hexagonal layers of Cr, S, Na, S, Cr, etc., piled on each other inclose packing.The Na-S distance is 2.78 A., and Cr-S, 2.44 A. Thestructure resembles the NaCl type, if we disregard the difference betweenNa and Cr.The structure of chalcopyrite,3 CuFeS,, given by L. Pauling and L. 0.J. W. Boon and C . H. MacGillavry, Rec. Trav. chim., 1942, 61, 910; see alsoJ. W. Boon, Rec. Trav. chim., 1944, 63, 69.W. Rudorf€ and K. Stegemann, 2. anorg. Chem., 1943, 251, 376ROBERTSON : CRYSTALLOGRAPHY. 51Brockway4 has been confirmed, and AgFeS, is shown.to have the samestructure. I n these, as in potassium thioferrite, nearly regular tetrahedraof sulphur atoms exist.These chains of sulphur tetrahedra must remainintact during the reactions MFeS, + CuFeS, and KFeS, + AgFeS,,which can occur in the crystalline state, but certain other problems con-cerning the mechanism of these reactions remain obscure.The compounds NaBiS, and KBiS, are cubic and are reported to havethe NaCl type of s t r u ~ t u r e . ~ Apparently two sodium (or potassium) andtwo bismuth ions occupy in a random manner the four sodium positions inthe NaCl structure, while the other four chlorine positions are occupied bythe sulphur ions. The compounds appear to be stable in this structuretype even after prolonged heating, unlike LiFeO, and Li,Ti0,.6 Accordingto this investigation the Bi*++ ion has a radius of between 1.10 and 1.20 A,Sulphur Trioxide.-An interesting account of the crystal structure of they-modification, or ice-like form, of sulphur trioxide has now become avail-able.' From single-crystal measurements it is found that the orthorhombiccell contains twelve units of SO,, which are combined to give four puckeredring molecules of S30, as in (11).Here the sulphuratoms are situated a t the centres of slightly distortedtetrahedra of oxygen atoms, with S-0, 1-60 A. ; S O ,1-40 A. ; 0-0 (tetrahedron edge), 2.45 A. ; although nogreat precision is claimed for the atomic positions.The smallest intermolecular distances are about 2.9-3-0 A. This form of molecule is rather closely relatedto that found for P,01,,8 from which it may be derivedby the removal of one central atom and one oxygen atom and a slightmodification of the remaining positions. This structure for the y-modifica-tion of SO, is in agreement with recent Raman and infra-red investigations.Other Structures.--Other work which may be briefly mentioned includesa confirmation of the phosphorus pentabromide structure 9 as Dz-Pbcmwith four units of PBr, per unit cell, consisting of almost regular tetrahedralPBr,+ ions (P-Br distance, 2.2 A.) and the fifth bromine as a separate Br-ion a t 4.3~., as previously found by H.M. Powell and D. Clark.10I n tungsten hexachloride l1 the space-group is Cii and the structure is ahexagonally close-packed chlorine lattice slightly deformed so as to group thesix chlorine atoms octahedrally around a tungsten atom with W-Cl, 2.24 A.Potassium tetrachlorozincate, K2ZnCl4,l2 proves to be a very complicatedstructure with twelve molecules per unit cell in the space-group C&Pma.0Oys/ 8 ~ 0/ b dNo\s/ (f > ('I.)4 2.Krist., 1932, 82, 188.7 R. Westrik and C. H. MacGillavry, Rec. Trav. chim., 1941, 60, 794.8 G. C. Hampsonand A. J. Stosick,J. Amer. Chem. SOC., 1938, 60, 1814; H. C. J.de Decker and C. H. MacGillavry, Rev. Trav. chim., 1941, 60, 153.9 M. van Driel and C. H. MacGillavry, ibid., 1943, 62, 167.J. W. Boon, Rec. Trav. chim., 1944, 63, 32.E. Kordes, 2. Krist., 1935, 92, 139.Nature, 1940, 145, 071.J. A. A. Ketelaar and G. W. van Oosterhout, Rec. Trav. chim., 1943, 62, 197.l2 H. P.Klug and G. W. Sears, J. Arner. Chem. SOC., 1945, 67, 87852 GENERAL AND PHYSICAL CHEMISTRY.The crystal structures of cadmium cyanide and gold cyanide are discussedby G. Shdanov and E. Schugam.13 In gold cyanide the lattice is built upof chains of Au-C-N-Au- in which covalent bonds predominate.X-Ray measurements have been made on czsium fluosulphonate,l4CsSO,F, which has the scheelite type of structure; and on potassiumberyllium fluoride, K2BeF4, strontium orthosilicate, Sr,SiO,, and bariumorthosilicate, Ba,SiO,, l5 which are all isomorphous with potassium sulphate.The structural crystallography and optical properties of the variousforms of silicon carbide have been discussed,l6 and a preliminary noteindicating a full structure determination of the interesting ferromagneticmineral cubanite, %uFe,S,, has appeared.17 In this structure each metalatom appears to be surrounded by four sulphur atoms in almost undistortedtetrahedral co-ordination, and the sulphur atoms are similarly eachsurrounded by four metal atoms.There is an indication that the iron atomsare bonded to one another in pairs, in addition to their links with the sulphuratoms.(iii) Organic Structures.The Structure of Cholesteryl Iodide.-The very great part played bycrystallography in the elucidation of the chemistry of the sterols is wellknown, The early measurements and ideas put forward by J. D. Bernallwere rapidly followed by an immense amount of purely chemical work,2 andthe structure of the sterol skeleton (I) is now well established.The X-ray26!2627analysis of the crystal structure of cholesteryl iodide (I) now given by C. H.Carlisle and D. Crowfoot (Mrs. Hodgkin) is of outstanding importanceboth to crystallography and to chemistry. In crystallography cholesteryliodide is probably the most complicated organic structure which has yet13 Acta Physicochim. U.R.X.S., 1945, 20, 247, 253.l4 El. Siefert, 2. Krist., 1942, 104, 385.16 H. O'Daniel and L. Tscheischwili, ibid., p. 348.16 N. W. Thibault, Amer. Min., 1944, 29, 327; L. S. Ramsdell, ibid., p. 431.l7 M. J. Buerger, J . Amer. Chem. SOL, 1945, 67, 2056.1 Nature, 1932, 129, 277; Chem. and Ind., 1932, 51, 466; see also J. D. Bernal,2 See Ann. Reports, 1933, 30, 198; 1934, 31, 206; 1936, 33, 341; 1938, 35, 281;3 Proc. Roy.SOC., 1945, A, 184, 64.D. Crowfoot, and I. Fankuchen, Phil. Trans., 1940, A, 239, 135.1939, 36, 286; 1940, 3'7, 332; 1943, 40, 122ROBERTSON : CRYSTALLOGRAPHY. 53been fully analysed, and the analysis has been accomplished very largely bythe use of recently developed direct methods which do not involve anychemical theory. From the chemical point of view, the determination ofall the atomic positions with reasonable accuracy is of great importance asa verification of theory and in settling various outstanding points of stereo-chemical detail.The iodide was found to exist in two polymorphic modifications A andB, of closely related crystal structure, both monoclinic, PZ,, with twomolecules per unit cell. The analysis, carried out on both these crystals,depends on the fact that the phase constants of the X-ray reflections arelargely controlled by the contributions of the iodine atoms.As the twoiodine atoms (one on each molecule) occupy general positions in the unitcell, the analysis is not so straightforward as that of platinum phthalo-cyanine,4 but the general principles are the same. The position of theiodine atoms was first determined with considerable accuracy by a two-dimensional Patterson synthesis on the (010) plane. The phase anglesderived from these positions were then attached to the measured structurefactors and Fourier projections of the electron density were made on the(010) plane. In the projections so obtained from both crystals the outlinesof the sterol ring system and attached side chain were clearly visible.Itwas now apparent, however, that in the B crystal form the orientation ofthe molecule is more favourable for projection as there is less overlappingof the atom centres. The position of nearly every atom could be seen.More detailed work was therefore confined to the B form.To proceed further requires the use of three-dimensional methods, as thesterol molecule is far from being a planar structure. Line syntheses weretherefore constructed parallel to the b axis and passing as nearly as possiblethrough the projected centres of the atoms so far resolved. Here a complic-ation arises, because the phase angles derived from the two iodine positionsnecessarily introduce a centre of symmetry which does not exist in the realstructure.(This complication does not arise in the projections mentionedabove, because they actually possess centres of symmetry.) In threedimensions, the result of the calculation is to show the molecule as a wholesuperimposed upon a spurious mirror image of itself. To select the trueatomic positions it is now necessary to make use of our knowledge of normalcarbon-carbon bond distances and valency angles. It will be noted thatonly at this late stage of the analysis is any previous knowledge of chemicalstructure used, and this knowledge is of the kind that might be derivedmerely from a study of the structure of diamond.Once the atomic positions are sorted out in this manner it is possible toconstruct a very satisfactory model of the molecule, and further refinementcan proceed by well-tried methods of successive approximation. The finalresults are expressed by three very striking sections showing the electron-density distribution in planes parallel to the (010) and passing through themolecule at different levels.From these sections, combined with the lineJ. M. Robertson and (Miss) I. Woodward, J., 1940, 3654 GENERAL AND PHYSICAL CHEMISTRY.syntheses mentioned above, it is possible to estimate all the 84 co-ordinateswhich govern the structure.The accuracy claimed for the final atomic positions is not high, owingto a number of causes, e.g., intensity inaccuracies, spurious diffraction effectsresulting from the heavy atom, and incompleteness of the excessively lengthycalculations involved. A more serious and fundamental limitation isimposed by the general weakness of the X-ray reflections from planes ofsmall spacing, which may indicate some degree of disorder in the structure.Nevertheless, the results are quite sufficient to determine completely thestructure in the chemical sense, and to decide which atom is attached towhich, and the relative orientations of the atoms in space.The interatomicdistances and valency angles found are C-I, 2.08 A. ; C-C, 1.47-1.60 A.,mean 1 . 5 5 ~ . ; ' c=c, 1.30A.; <c-c-c, 91-129" 30', mean 108" 36';(C-C-C, 124" 45', 125" 33'.The molecular structure is in excellent agreement with present chemicalviews.5 The ring system is non-planar, with methyl groups attached a tC,, and C13.C3 is rather ill-defined in the electron-density maps owing tothe proximity of the heavy iodine atom, but the data definitely favour thecarbon-iodine bond being cis to methyl at C,, (the " trans " form ofL. Ruzicka, M. Furter, and M. W. Goldberg 6). Apart from the distortioncaused by the double bond (which is good evidence for its position), therings have the Sachse trans-configuration.Chemical evidence on the stereochemical relations involved in theattachment of the side chain a t C17 appears to be conflicting.' The X-rayevidence, however, is very definitely in favour of the side chain being cis tomethyl at C13. The arrangements of the carbon atoms about the bondCl3-C1, and also about C17-C,o are trans in form.In general, the staggeredtrans-configuration is followed throughout the chain and ring systems ofthe molecules.*Abnormal Bond Distances in Geranylamine Hydrochloride and Dibenzy1.-Perhaps the most complete structural analysis of a complex organic crystalso far achieved has now been described by G. A. Jeffrey9 for the hydro-chloride of geranylamine, CloH1,*NH,,HC1. A preliminary note about thisstructure has already been reported upon.1° The work is remarkable notso much for the complexity of the molecule as for the thoroughness of theX-ray analysis. The atomic positions were derived from preliminaryPatterson syntheses followed by full-scale three-dimensional Fourierayntheses in the form of sections through the two principal monocliniccrystal planes, (010) and (001).Some of the final results are illustrated in5 0. Rosenheim and H. King, Chem. and Id., 1932, 51, 464; H. Wieland andE. Duane, 2. physiol. Chem., 1932, 210, 268.f~ Helv. Chim. Acta, 1938, 21, 498.L. Ruzicka, M. Goldberg, and H. Wirz, ibid., 1935, 18, 998; H. Wieland andE. Duane, 2. physiol. Chem., 1933, 216, 98; V. Caglioti and G. Giacomello, Gazzetta,1939, 69, 245.C. W. Bum, Proc. Roy. SOG., 1942, A , 180, 67.Ibicl., 1945, A, 183, 388. lo Ann. Reports, 1943, 40, 96ROBERTSOX : CRYSTALLOGRAPHY. 55Fig. 1 in the form of superimposed sectional summations on the (010) plane,the contour lines denoting density increments of one electron per A.3, excepton the chlorine atom where the scale is two electrons per A .~ . (The zeroand first contours have been omitted in the diagram.) From this and othersimilar maps all the 36 atomic parameters for the carbon, nitrogen, andchlorine atoms have been measured directly from the maximum of each peak.The experimental data in this work consisted of structure factors derivedfrom visual estimates of intensities from 1060 planes, including all thosewithin range of Cu-Kcc radiation. This represents about 88 structure factorsFIG. 1 .-Projection of the gernnylanaine hydrochloride molecule and Fourier sections on(010). (G. A. Jeffrey, from Proc. Roy. SOC., 1945, A , 183, 391.)per atom of the asymmetric unit and appears to constitute a record forcompleteness of data. Various tests have been applied to ascertain thelimits of experimental error, and these are finally estimated at 3 0.04~.and & 4" for the bond lengths anti valency angles.Within these limits the two isoprene units in the molecule are identical,most of the bond lengths are normal (carbon-carbon single bonds, 1.52-1.55 A., carbon-carbon double bonds, 1.32 A., carbon-nitrogen, 1-48 A.), andthe valency angles have about the expected values. There is, however, oneoutstanding exception.The C,-C, bond (Fig. 1) linking the two isopreneunits has a length of 1.44 or 1.45 A., and is thus shorter than a normal singlebond by an amount two or three times the probable experimental error. A56 GENERAL AND PHYSICAL CHEMISTRY.the system is not conjugated in the usual sense, this contraction is verydifficult to explain in terms of current theory.It is suggested that thehybrid character of the bond may be the result of a hyperconjugation processin which the a-methylenic C-H electrons become partly localised in thecentral bond, as indicated in (11).H H-CH=C+C-CH=C- I ! ! I(11.1 MeH H MeOne important stereochemical feature of the structure 'concerns theorientation of this central bond relative to the planes of the isoprene unitswhich it joins. This appears to be governed by a balance of the stericrepulsions between c6.. . . . C, (3-28 A.) and between c6.. . . . C, (methyl)(3.24 A.). As only intramolecular forces are involved here, the resultshould apply generally and may have an important bearing on the stereo-chemistry of long-chain polymers.I n the structure as a whole, the molecules are so arranged that eachnitrogen atom is a t nearly equal distances (3-17-3-24 A.) from each of fourchlorine neighbours.Thepeak value of the electron density becomes progressively less for every atomas we pass along the chain away from the nitrogen atom.Similar thoughrather less pronounced effects have been noticed in other structures, e.g.,anthracene,ll as we pass along the molecule away from the meso carbonatoms. They may bedue to a variation in the thermal motion of different parts of the molecule;but the possibility of their being spurious and due to the incomplete con-vergence of the Fourier series used cannot be excluded.The interesting result mentioned above regarding the contraction of thecentral bond in the geranylamine molecule immediately raises the questionas to whether this is a general property of all similarly linked systems.Abrief note has now appeared giving the bare results of an equally comprehen-sive three-dimensional investigation of the structure of dibenzyl.12 Anearlier two-dimensional investigation of this structure l3 succeeded indefining the orientation and general shape of this molecule in the crystalwith considerable precision, but the resolution was not sufficient to yieldany very precise measurements of the length of the central -CH,-CH,- bondsituated between the benzene rings. The new determination makes thislength 1.48 0-01 A., thus indicating a contraction similar to thoughsmaller than that found in geranylamine.The explanation, in terms of ahyperconjugation process, is presumably the same.Other interesting results obtained from this new analysis of dibenzyll1 J. M. Robertson, Proc. Roy. SOC., 1933, A , 140, 79.l2 G. A. Jeffrey, Nature, 1945, 156, 82.l3 J. M. Robertson, Proc. Roy. SOC., 1934, A , 146, 473; 1935, A, 150, 348.One peculiar feature of the structure is clearly evident in Fig. 1.The explanation of these effects is rather obscureROBERTSON : CRYSTALLOGRAPHY. 57show that the sides of the benzene rings vary in length from 1-36 to 1-39 A.As the experimental error is estimated to be only 0.01 A., it seems possiblethat some of these variations may be real. All the angles associated withthe benzene rings are found to be 120" 1".Finally, the C-CH, bondlengths immediately adjacent to the rings are given as 1-50 & 0.01 A,, thusindicating a fairly large contraction from the standard single bond value of1.54 A.It is clearthat, if measurements of this order of accuracy can be established generally,much interesting work remains to be done in the detailed examination ofthe structures of simple compounds.Meanwhile, the anomalous reactivity of certain 1 : 5-dienes has beenfurther examined in a combined chemical and crystallographic study ofmethyl and ethyl A1 : 5-hexadiene-l : 1 : 3 : 3 : 4 : 4 : 6 : 6-octacarboxylates byL. Bateman and G. A. Jeffrey.14 The properties of these compounds hadpreviously been studied very fully by C. K. Ingold, M. M. Parekh, andC.W. Shoppee.15 It is now suggested that their anomalous behaviour maybe due to a similar but more extensive chain hyperconjugation than thatin geranylamine hydrochloride. X-Ray data are given for the octamethyland octaethyl esters mentioned above, as well as for the hexamethyl andhexaethyl ester diacids, the hexaethyl ester diacid chloride, the dihydro-octamethyl ester, methyl and ethyl cyczopentane hydroxy-acid ester, thedimer of methyl ay-dicarbomethoxyglutaconate, and the dimers of ethyla-dicarbethoxyglutaconate, but these data only go as far as cell dimensionsand space-group determinations, with. the object of deciding between cis,trans, and cyclic structures.Coronene.-In elucidating the structures so far described full use hasbeen made of the three-dimensional Fourier series methods.For moleculeswith complicated shapes this represents the only feasible way of refining thestructure. The main drawback of the method lies in the excessive amountof numerical computation involved, which makes its use prohibitive in somecases. There are certain special structures, however, for which an almostcomparable accuracy may be achieved by the careful use of two-dimensionalmethods. The aromatic hydrocarbon coronene, for which the full structuredetermination is now available,16 is a good example.I n the two-dimensional projection on the (010) plane a very high degreeof resolution is obtained, the great plane of the molecule being inclined atabout 44" to this projection plane. The accuracy of such work is difficultto assess in general, but in this case a rather careful examination has beenmade by conducting parallel investigations on hypothetical structures withknown atomic positions.The general conclusion is that for the coronenestructure the bond length measurements, after averaging in groups as shownA full discussion of these results is awaited with interest.l4 J . , 1945, 21 1 ; L. Bateman and H. P. Koch, ibid., p. 216.l5 J . , 1930, 142.l6 J. M. Robertson and J. G. White, J., 1945, 607. See also Ann. Reports, 1944,41, 6958 GENERAL AND PHYSIOAL CHEMISTRY.in Fig. 2, are probably accurate to about 5 0.01 A., and that the positionof individual atoms may have maximum errors of between 0.02 and 0.03 A.The results, given in Fig.2, show an interesting variation in bond lengthin different parts of the molecule. For the central ring and spokes it is1-43a., whereas the outer bonds appear to alternate between 1.385 and1.415 A., depending on their situation. These variations are large enoughto be significant, and they represent the first definite measurements ofvariable carbon-carbon bond lengths for any condensed ring aromatic hydro-carbon. They can be given a rough qualitative explanation in terms of thet tIE 7.385 FFIG. 2.-Dimensions of the coronene molecule. (From J . , 1945, p. 612, Fig. 5.)twenty stable valency bond structures which can be written for coroneneby assessing the double-bond character for each set of links and making useof L. Pauling and L. 0.Brockway’s empirical curve relating double-bondcharacter and distance.17 However, both in this treatment and in the moredetailed calculations carried out by C. A. Coulson l8 there is difficulty indistinguishing between the lengths of one set of outer bonds (1.415 A.) andthe bonds of the inner ring (1.43 A.). Much further work clearly remainsto be done in examining other aromatic systems, both experimentally andtheoretically, before final conclusions can be reached.DiphenyEene.-The results of an electron-diffraction study of this moleculehave already been discussed in these Rep0~ts.l~ A full account of thecrystal structure now available 20 confirms the previous conclusion that thel7 J. Amer. Chem. SOC., 1937, 59, 1223.2o J. Waser and C. S. Lu, J.Amer. Chem. SOC., 1944, 88, 2035.18 Nature, 1944, 154, 797.Ann. Reports, 1943, 40, 92ROBERTSON : CRYSTALLOGRAPHY. 59compound is indeed dibenzcyclobutadiene, as originally suggested by W. C.Lothrop's synthesis.21 The crystal structure is a peculiar one, with sixmolecules of diphenylene in the monoclinic space group P2,/a. The symmetryrequirements in this space-group demand that a t least two of these moleculesmust have exact centres of symmetry in special positions, which is strongpreliminary evidence for the coplanar molecule (111) already indicated bythe electron-diffraction results. Exact conclusions regarding the remainingfour niolecules are more difficult to reach, but the analysis is somewhatsimplified by the observation that the (hkO) reflections are nearly all absentexcept when h = 3n, which indicates some form of triplet grouping alongthe a axis (19.60 A.).It was finally possible to achieve a fairly detailedanalysis, with Fourier projections showing reasonably good resolution ofmany of the atoms. Owing to the complexity of the crystal structure ahigh order of accuracy is not claimed, but most of the atomic positions areknown to within 0.1 A. The structure (111) is definitely con-firmed and the dimensions found by electron diffraction(hexagon sides, 1-41 A., and lateral connecting links of the fourring, 1.46 A.) appear to be consistent, within the experimentallimits, with the X-ray results.The full implications of the interesting fact that there are two crystallo-graphically independent types of molecule in the unit cell have not quitebeen worked out.Although it is perhaps improbable, the possibility existsthat the dimensions or the shapes of these two kinds of molecule may beslightly different. In fact, some of the X-ray results seem slightly to favoursuch a conclusion. In this connection a careful study of the environmentof the two crystallographically different kinds of molecule in the unit cellis required, and the appropriate diagrams are given in the original paper.21Other crystals displaying two crystallographically distinct kinds of moleculeare to be found in the stilbene, tolan, and trans-azobenzene series.22 Thesestructures are not strictly analogous to diphenylene, as they are four-moleculecrystals with apparently exact symmetry centres in each molecule, but theyhave the same interesting possibilities of variable dimensions and differentenvironments for the two kinds of molecule.Amino-acids.-Even the simple amino-acids usually present crystalstructure problems of considerable complexity.Several such structureshave in the past been studied carefully as part of a general programme ofinvestigation of the amino-acids and proteins.23 Recently, crystalline copperand nickel salts of several of the acids have been prepared for X-ray in-vestigation, with the hope that the heavy-metal atoms would facilitate thework and enable more or less direct Fourier series methods to be employeda t an early stage. In two structures recently reported this technique has()<)21 J.Waser and C. S . Lu, J , Amer. Chem. SOC., 1941, 63, 1187 ; 1942, 64, 1698.22 J. 31. Robertson and (Miss) I. Woodward, Proc. Roy. SOC., 1937, A , 162, 568;1938, A , 164, 436; J. J. de Lange, J. M. Robertson, and (Miss) I. Woodward, ibid.,1939, A , 171, 398.23 J . Amer. Chem. SOC., 1939, 61, 1087; Ann. Reports, 1939, 36, 17960 GENERAL AND PHYSICAL CHEMISTRY.proved very successful. Copper dl-a-aminobutyrate and nickel aminoacetatedihydrate 24 are both found to belong to the monoclinic space group P2Jcwith two molecules per unit cell. The metal atoms must therefore be onthe special two-fold centro-symmetrical positions, and their co-ordinates arefixed. I n the Patterson projection of these structures the only significantpeaks will correspond to vectors with a metal atom a t one end; in otherwords, the Patterson projections will have very much the same appearanceas the corresponding Fourier projections. The situation is closely similarto that found in the metal phthalocyanine structure^.^^In the case of copper dl-a-aminobutyrate a clear projection on the acplane determines all the x and x co-ordinates of the atoms.The presumablysquare and necessarily coplanar co-ordination of the amino-nitrogen andcarboxyl oxygen atoms around the copper atom is indicated diagramaticallyin (IV). In addition, hydrogen bridges appear to connect the amino-nitrogen atoms to the carboxyl oxygen atoms of other neighbouring molecules,as shown by the dotted lines. As only one projection of the structure isavailable the y co-ordinates of the atoms cannot be determined and so noinformation on real interatomic distances is yet available.The analysis has proceeded much further for nickel aminoacetatedihydrate, Ni(NH2.CH2.C02)2,2H,0.24 The monoclinic cell dimensions aremore nearly equal, and three independent projections of the structure havebeen made along the three crystallographic axes.Unfortunately, theorientation of the molecule in the unit cell is such that some of the parametersof the light atom’s cannot be determined directly. For example, the nitrogenatom is unresolved in all the projections. By assuming values for a fewof the better known interatomic distances it is possible, however, to deducethe whole structure with a good deal of certainty.It is found to consistof distorted octahedral complexes of nickel atoms bound to two glycineresidues and two water molecules. As in the previous compound, thenickel atom makes a coplanar and almost square co-ordination to twooxygen atoms of different carboxyl groups and to two amino-nitrogenatoms, the distances being 2-08 and 2-09 A. In addition the nickel atomforms covalent bonds (2.12 A.) to the two water molecules which are situateda t the remaining vertices of the octahedron. The glycine residues arenearly flat, with dimensions closely similar to those previously found by24 A. J. Stosick, J. Amer. Chem. Soc., 1945, 67, 362, 365.*5 J. M. Robertson and (Miss) I. Woodward, J . , 1937, 219; 1940, 36ROBERTSON : CRYSTALLOGRAPHY. 61G. Albrecht and R. B. Corey.23 Each of the water molecules in the octa-hedral complexes is further connected by two moderately strong hydrogenbridges (2.72 A.) to carboxyl oxygen atoms of neighbouring complexes.The outer oxygen atoms of the carboxyl groups also form one fairly strong(2.96 4.) hydrogen bridge and one very weak (3.13 A.) bridge to the nitrogenatoms of the amino-groups of neighbouring complexes.Although it has been possible to deduce these structures very fully froma study of the Fourier projections, it is quite clear that with such structuresthe maximum amount of information can be obtained only by the use ofthree-dimensional Fourier methods. For the more complex amino-acidsand proteins, and indeed for all structures where the molecular codgurationdeparts markedly from the simple planar form, the use of three-dimensionalmethods is necessary.Adipic Acid.-A fairly complete account of the crystal structure ofadipic acid has now become available.26 The two centro-symmetricalmolecules in the monoclinic unit cell (space group P2,/a) lie along the caxis. Adjacent carboxyl groups are centro-symmetrically related andconnected by hydrogen bridges of length 2-65 A. The structure has beenrefined by means of two Fourier projections [along the (101) and the b axis]and all the parameters are given. It is, however, rather difficult to estimatethe accuracy of the bond-length determinations from this paper. Thevalues reported are C-0, 1-38 A. ; C=O, 1.28 A. ; G-C, 1.49, 1.52, 1.51 A.(reading outwards from the carboxyl group). The mean zigzag angle inthe carbon chain is 113", and the chain is reported to depart by a few degreesfrom the planar zigzag form. It is difficult to say whether the departuresfrom the standard carbon-carbon bond length of 1-54 A. are significantThey are, however, almost identical with the values reported earlier forsuccinic and it is pointed out that a fairly large number of independentstructure determinations lead to values for normal single bonds that aresomewhat less than 1-54 A . ~ * It is quite clear, however, that none of thiswork has an accuracy as high as that now obtained for dibenzyl or coronene(pp. 58, 59), and further investigation is needed.Other Organic Structures.-Certain geometrically isomeric piperazinederivatives containing two quaternary nitrogen atoms have recently beenprepared 29 and an X-ray examination by H. M. Powell has established theconfigurations in this series. The usual correlation between low meltingpoint, high solubility, and &-configuration has been shown to hold. Bymaking use of the high scattering power of the iodine atom, an electrondensity projection was obtained for trans-NN'-di-(P-chloroethyl)-NN'-di-2 6 C. H. MacGillavry, Rec. Traw. chim., 1941, 60, 605.2 7 H. 5. Verweel and C. H. MacGilIavry, Nature, 1938, 142, 161; 2. Krist., 1939,102, 60.28 Ann. Reports, 1938, 35, 196; also C. W. Bunn, Trans. Paraday SOC., 1939, 35,482; C. H. MacGillavry, 2. Krist., 1938, 98, 407; F. J. Llewellyn, E. G. Cox, andT. H. Goodwin, J . , 1937, 883; T. H. Goodwin and R. Hardy, Proc. Roy. SOC., 1938,A , 164, 369.29 W. E. Hanby and H . N. Rydon, J., 1945, 83362 GENERAL AND PHYSICAL CHEMISTRY.methylpiperazinium di-iodide (triclinic) in which the resolution is sufficientto establish the trans-configuration quite clearly.The triphenyls of bismuth, arsenic, and antimony have been examined byJ. Wet~el,~O and he has determined the structure of the bismuth compound inconsiderable detail. 0-03 A.The crystal structure of adamantane (s-tricyclodecane, CloH16) has beendetermined by W. Nowacki.31 This very interesting compound crystallisesin the cubic system, space group T;-F33m, with four molecules per unitcell, and the structure establishes the tetrahedral distribution of valencybonds for the carbon atoms. The carbon-carbon bond length is 1-54~.,and the distance between the carbon atoms on adjoining molecules is 4.15 A.The whole structure is closely similar to that of he~amethylenetetramine.~~Amongst complex compounds of biological interest, preliminary X-ray datahave been given for gliotoxin,33 crepin,3* deoxycorticosterone acetate,35 and17-isodeoxycorticosterone acetate.36 Some very striking single crystalmeasurements have been made on a tobacco necrosis virus derivative byD. Crowfoot and G . M. J. Schmidt.3' The molecular weight of this sub-stance is about 1,850,000, yet it gives single crystals measuring severalmillimetres across. These are triclinic, a = 179 A., b = 219 A., c = 243 A.,and the angles do not differ much from 90". The crystal structure suggeststhat the molecules are spherical, of radius 80-100 A. The extreme size ofthe unit cell makes it possible to apply a new type of examination withmonochromatic X-rays, known as the " still " photograph method. Whenthe crystal is stationary the cell dimensions are such that large numbers ofcrystal planes are correctly oriented for reflection and these are character-istically arranged on the photographic plate in series of concentric circles orellipses. Much information can be obtained from a detailed study of thesephotographs, which require a far shorter exposure time than the otherusual methods. Reflections from planes with spacings as small as 2 . 8 ~ .have been observed.The crystalline phases of soap have received detailed X-ray examinationin a series of papers by M. J. Buerger and and diffraction datahave also been given for a series of mono glyceride^.^^The bismuth-carbon distance is given as 2.30J. M. R.F. P. BOWDEN. J. M. ROBERTSON.D. TABOR. H. W. THOMPSON.so 2. Krist., 1942, 104, 305.32 R. G. Rickinson and A. L. Raymond, J . Amer. Chem. SOC., 1923, 45, 22;s1 Helv. Chim. Acta, 1945, 28, 1233.R. Brill, H. G. Grimm, C. Hermann, and C . Peters, Ann. Physilc, 1939, 34, 393.R. Crowfoot and 13. W. Rogers-Low, Nature, 1944, 153, 651.34 B. W. Rogers, Brit. J. Exp. Path., 1944, 25, 212.35 W. Nowacki, Helv. Chim. Acta, 1944, 27, 1622.s6 Idem, ibid., 1945, 28, 1373.38 M. J. Buerger, L. B. Smith, F. V. Ryer, and J. E. Spike, Proc. Nut. Acud. Sci.,1945, 31, 226; K. W. Gardiner, M. J. Buerger, and L. B. Smith, J. Physical Chem.,1945, 49, 417; M. J. Buerger, Amer. Min., 1945, 30, 551.s9 L. J. Filer, S. S. Sidhu, €3. F. Daubert, and H. E. Longenecker, J . Amer. Chem.SOC., 1944, 66, 1333.a 7 Nuadre, 1945, 155, 504