THE SPECTRA OF METAL MOLECULES .sx BY S. BARRATT ( Chemical Department , University College , London) .sx Received 30th August , 1929 .sx It has long been known , as the result of physico-chemical investigations , that metals and alloys in the liquid state have in general a high molecular complexity .sx Nothing , however , could be deduced as to the behaviour of the same symptoms in the vapour state , and so far as vapour density determinations , etc. , were able to show , the vapours of the more electro-positive elements were uniformly monatomic .sx A study of the absorption spectra of the same vapours has revealed that this conclusion is not strictly true , and that in many metallic vapours there must be a small proportion of di- or polyatomic molecules , for they possess well-defined , and often intensely-developed , band spectra .sx Perhaps the best-known examples of such spectra are those of the alkali metals and of mercury .sx The important suggestions made by Professor Franck at a previous meeting of the Society gave a further impetus to the study of these metallic molecules .sx They would seem to have a very simple homopolar structure , and the original electronic configurations of the component atoms may only suffer comparatively small disturbances in their formation .sx In accordance with this view , many of the band spectra are closely associated with the atomic line absorption spectra ; and the study of these relationships , and their explanation in terms of molecular structure , is a field of increasing interest .sx It has also become evident from an examination of the absorption spectra of mixtures of metallic vapours that many such mixtures contain compound molecules ( say of the type MM' when M and M' are two different metals ) , since characteristic band spectra are developed in the vapours .sx The total number of metal molecular spectra that becomes available for study is thus enormously increased , and it should be possible soon to accumulate sufficient experimental data for a general study of the structure and behaviour of this class of molecule .sx Among the spectra of these compounds , so far established , are those of the " mixed molecules " of the alkali metals ( e.g. , NaK , LiCs , etc. ) , and there would seem to be a characteristic spectrum for each of the possible pairs of these metals .sx Magnesium , in conjunction with the alkalis , yields another group of spectra .sx The spectra of the ZnHg , and of the Cd Hg molecules have also been recognised .sx It is proposed to describe in the present communication experiments recently undertaken from this point of view upon the vapours of mixtures of cadmium , zinc , and mercury , with those of the alkali metals .sx Since a series of new spectra , which is virtually a complete one , has been identified , it is evident that all these alloys must yield a certain proportion of compound molecules on vaporisation-though the absolute amount of the association is probably small .sx It is not necessary to enter here into details of apparatus , especially as these have been given in previous papers on the alkali metal spectra , but the following outline may be useful .sx For preliminary observations , the metals were vaporised in a steel tube electrically heated , with water-cooled end windows .sx This simple arrangement is easily and rapidly worked , but it has an important disadvantage when experimenting with mixtures of metals with widely-separated boiling-points .sx The more volatile metal is apt to distil away completely from the heated portion of the tube before the other component has given an appreciable density of vapour , so that the compound spectra do not develop , or do so only for a short time .sx The apparatus used in the final experiments consisted of a sealed hard-glass tube , about 9 ins. long , with the ends blown as thin and as uniform as possible , to minimise the distortion of the light beam when passing through them .sx This container was enclosed in an electric furnace , and had attached to it a long side-tube with an in-dependent heating system .sx By placing the more volatile metal in the side-tube and the other in the main tube , it was possible to obtain a mixture of the vapours in any desired proportion by regulating the temperatures of the main and the side-tube .sx Such an apparatus , it is true , is not practicable for observations in the ultra-violet , but the preliminary work gave indications only of spectra in the visible region , so that the limitation was not of importance .sx The alkali metals themselves , and their alloys with each other , possess extensive band spectra in the same spectral region , so that certain of these bands were necessarily present in each absorption spectrogram obtained , as well as the new spectra under discussion .sx Fortunately , long experience with these metals , and the wide range of comparison plates that was available , made it easy to allow for all such complications .sx The alkali bands could , in addition , be repressed to some extent by using large excess of the other component ( Cd , Zn , or Hg) .sx None of the metals of Group IIB .sx , when examined alone , show absorption in this region .sx The new spectra recognised in this way resemble in their general characteristics the metallic molecule spectra which have been previously examined , that is to say , they contain two distinct types of absorption :sx ( I ) Electronic band systems , showing fine structure ; and ( 2 ) Isolated diffuse bands with no fine structure , which terminate abruptly on the long wave-length side and broaden , with increased density of vapour , towards the ultra-violet .sx In some cases , such bands are definitely the starting point of a weak continuous absorption stretching towards shorter wave-lengths .sx The spectra under discussion differ from those of the alkali metals in that these diffuse bands are their more prominent feature .sx With many pairs of metals , in fact , the diffuse bands were the only new absorptions registered on the plates , and the remaining pairs for which the ordinary type of band system was also found , gave it only weakly .sx The positions of the heads of the diffuse bands corresponding to the different molecules are given in the table on p. 760 .sx The width of these bands , when well developed , is about 20 or 30 A. It is dependent to some extent on the density of the vapour in the absorbing column .sx Their appearance is quite unaffected by the presence or absence of an atmosphere of an inert gas argon and hydrogen were both introduced on occasion .sx On plates secured with dense vapours it could be seen quite distinctly that the bands were the beginning of a weak continuous absorption extending indefinitely on the short wave-length side .sx Only a few bands belonging to this group of molecules have been found which show fine structure .sx The component heads of these are much more closely packed than are those in the alkali metal band spectra , and they are unsatisfactory subjects for work under low dispersion .sx Observations on them are still in hand , but the details are not ready for publication .sx Fine structure bands have been identified only in the absorptions of " Cs Hg " ( three band heads at 4991 , 4984 , and 4975 ) , and of mixtures of rubidium with zinc , cadmium , and mercury .sx These three spectra all show characteristic bands in the region 4400-4200 .sx The three systems are quite distinct , but they will require careful examination before details can be published .sx To summarise , the most important characteristic of this group of spectra is the uniform presence of the diffuse type of band .sx The same type is found in the alkali metal spectra ( both of the single metals and of binary mixtures ) and in the magnesium-alkali spectra .sx The bands are such a prominent feature of these spectra that a satisfactory explanation of them will undoubtedly afford much information about the molecules responsible for them .sx The hypothesis has been advanced in a previous paper that these diffuse bands represent the dissociation spectra of the corresponding molecules ; or in other words that the bands are analogous in origin to the continuous absorption spectra of the halogens .sx This view accounts satisfactorily for the sharp long wave-length limit of the bands , for their " degradation " towards the ultra-violet , and for the weak continuous spectra known to be associated with some of them .sx It is possible to calculate , on this basis , the heat of formation of the molecules from the position of the bands assuming , for example , that the molecule is a diatomic one , and dissociates into one normal and one excited atom .sx A comparison of these values with those obtained by independent methods would give a quantitative test of the hypothesis .sx The data required for such a comparison are available at present for the K2 molecule alone , but there the agreement is good .sx Carelli and Pringsheim estimate a heat of formation for this molecule of 0.51 to 0.63 volts , from observations on the intensity of fluorescence in the vapour .sx The value obtained from the band position on the foregoing assumptions is 0.53 volts .sx It is unfortunate that non-spectroscopic estimates for the heats of formation of other molecules of this class are not available , to test whether this agreement is only fortuitous .sx If the same hypothesis is to be applied to the new spectra which have been described above , it will clearly need amplification , for each molecule gives rise not to one such band , but to several , and there are instances of two diffuse bands occurring at quite a short wave-length interval .sx A greater complexity in .sx the spectrum would naturally be expected if the molecules responsible for them are more than diatomic .sx This is itself quite an open question , which will probably only be settled by an analysis of the fine structure bands , when these have been fully recorded .sx There is , at least , no doubt of the experimental fact that diffuse bands with the same general characteristics are to be found in all the known spectra of metal molecules containing an alkali metal atom .sx It so happens that very few metal molecules have as yet been recognised which have not an alkali as one component , and it is not possible to generalise further .sx There are certainly continuous absorptions of molecular origin in the cadmium and zinc spectra , but these , unlike the bands in question , broaden towards longer wave-lengths , and so cannot be regarded as analogous .sx Winans has suggested an ingenious explanation of the phenomena in these two spectra .sx There is a further remark which it is desirable to make in connection with future work on this class of spectra the necessity of great experimental caution in assigning observed absorption band spectra to a purely metallic origin .sx The great difficulty of determining the true origin of many emission band spectra is well known , but for obvious reasons the uncertainty is less when the spectra are observed in absorption , and with care a definite conclusion can usually be drawn .sx The sensitivity of the absorption method for the detection of certain classes of molecule is , however , extremely great and , if any of these highly absorbent molecules are present as traces in the vapour under examination , their spectra will certainly be registered and may ( and have been ) wrongly put to the account of the metal present in bulk .sx To take an example , it has been established in this laboratory that many metals ( e.g. , those of Groups II .sx and III .sx , lead , etc. ) are capable of forming , under suitable conditions , volatile subchlorides which possess strongly developed absorption spectra .sx Most laboratory samples of these metals contain a small amount of chloride as impurity , and if they are vaporised in the ordinary way , the subchloride bands form a prominent feature of the resulting absorption .sx Such bands have several times been catalogued as part of the true molecular spectra of the metals themselves .sx Thus , band spectra due to cadmium subchloride ( Cd Cl ) , and to zinc subchloride ( Zn Cl ) , have been considered part of the cadmium and zinc spectra .sx Another rather interesting example of the same difficulty is afforded by a spectrum in the region 3200-3300 , which has recently been attributed , on two separate occasions :sx To a cadmium molecule , Cd2 ; .sx To the molecule of a thallium mercury compound .sx It is , however , developed much more completely in thallium chloride vapour , under conditions precluding the presence of any free metal , and is undoubtedly the absorption spectrum of this salt itself .sx