COLOUR CENTRE§ IN ALKALI W I D E CRYSTALS By M. C. R. SYMONS and W. T. DOYLE* (THE UNIVERSITY SOUTHAMPTON) 1. Introduction A REMARKABLE wealth of information of interest and significance extending well beyond the field of solid-state physics has recently been forthcoming from studies of the spectrophotometric and magnetic properties of colour centres in ionic crystals. It is our aim here to describe and discuss the simplest of these systems namely additively coloured alkali halide crystals and to give some indica- tion of the situation in other systems. Our treatment will be brief and selective. The reviews by Seitz1.2 are outstanding and exhaustive other major works dealing with specific aspects will be referred to when appro- priate. In general colour centres are found whenever a new species is in- corporated into the alkali halide crystal to give what may be described crudely as a solid solution.However the geometry of the crystal imposes severe limitations on the size shape and charge of the solutes. We can distinguish two classes of “solutes” (a) Those which are derived from the cations or anions of the host crystal (sections 2 and 3) and (b) those which are unrelated to the host crystal (section 4). Alkali halide crystals have two great advantages as host crystals they are transparent from the far infrared to the far ultraviolet region (about 2000 A) where the fundamental absorption of the crystal begins. In addition the simplicity of the lattice geometry and the nearly complete ionic character of the bonding greatly simplify the interpretation. When considering colour centres of class (a) it is convenient to describe those which are derived by addition of metal as electron-excess centres and those derived from halogen as electron-deficit centres.Experimentally these can be formed separately by treatilig the crystal with metal or halogen vapour or together by exposure to high-energy radiation. Since such exposure can be carried out at low temperature several spectral features occur which are lost on warming and are not formed by the addition procedures. Additions give rise to relatively simple spectra and will be considered in the greater detail. The situation can be pictured in a simple way by considering what chemical reactions are to be expected for alkali-metal and halogen atoms in a fluid medium and then how they are likely to be modified by the crystal lattice.* National Science Foundation Postdoctoral Fellow on leave of absence from the Department of Physics Dartmouth College Hanover New Hampshire U.S.A. Seitz Rev. Mod. Phys. 1946 18 384. Seitz Rev. Mod. Phys. 1954 26 7. 62 SYMONS AND DOYLE COLOUR CENTRES IN ALKALl HALlDE CRYSTALS 63 Alkali-metal atoms have one overriding tendency namely to donate electrons to any available acceptor. In the absence of an acceptor they may dimerise but this bonding is very weak and at room temperature condensation to metal will normally occur. In contrast halogen atoms besides being powerful electron-acceptors can dimerise to give relatively stable molecules. In the presence of halide ions an equilibrium involving trihalide ions would be established (1 1 Hal + Hal- + Hal,- .. . . If the concentration of halide ions is large compared with that of halide atoms dihalide radical-ions can be formed (2) Hal. -t Hal- + Hal,- . . . . The bonding must be weak but there is evidence that these radical-ions are important intermediates in the flash photolysis of halide ions in water.3 Within the lattice in the absence of impurities the simplest electron- acceptors are anion vacancies which have an effective positive charge centred on the vacancy. It can be shown theoretically that alkali-metal atoms cannot exist as such in an undistorted lattice site and therefore they must be trapped interstitially combine to form metallic units or ionise in such a way that the electrons are trapped in anion vacancies and the result- ing cations occupy cation vacancies.An electron in an anion vacancy termed an F-centre is the simplest electron-excess centre [Fig. 1 (b)]. FIG. 1. Models for electron-excess centres. (4 (6) (c) (4 c4 V) w and represent anion and cation vacancies. represents an electron in an anion vacancy. (f) and ( g ) are alternative models for the M-centre. Other varieties of electron-excess centres formed by combination of elec- trons and vacancies are discussed in section 2. Similarly a halogen atom in an alkali halide crystal cannot exist at a regular lattice point. An interstitial atom would probably combine with a neighbouring ion to form a dihalide anion or migrate to a more stable site. If spatial requirements are satisfied formation of molecular halogen and trihalide negative ions would be expected. However it is commonly supposed that the most fundamental and simplest of electron-deficit centres would be the antimorph of an F-centre that is a system of similar configuration but with a reversal of sign.Such a unit termed a V centre by Seitz1*2 and symbolised in Fig. 2 (a) consists of a cation vacancy and a “hole” symmetrically distributed around the vacancy. That is to say that Grossweiner and Matheson J. Phys. Chem. 1957 61 1089. 64 QUARTERLY REVIEWS at any given instant one could picture one halogen atom and five halide ions around the cation vacancy but that on average all the six surround- ing halide ions share the “hole” created by the absence of an electron and hence the unit has spherical symmetry. FIG. 2. Models for electron-deficit centres (after Seitz2). (4 6) (el (4 o represents the absence of an electron from one of the halide ions surround- ing the vacancy.Since the spectra of several of these electron-deficit units such as the trihalide ions are known they can be compared with the bands observed in additively coloured crystals. These bands found in the near-ultraviolet region and generally described as V-bands are discussed in section 3. 2. Electron-excess centres Alkali halide crystals prepared with a stoicheiometric excess of alkali- metal either by heating them in the presence of the metal vapour4 or by electrolysis just below the melting point5 may exhibit a variety of optical absorption bands. In general these electron-excess bands occur in the lower- energy half of the region of transparency of the host crystal. The particular type of coloration observed depends in detail upon the optical and thermal treatment to which the crystal is subjected after introduction of the metal.Although a large variety of such colour centres exists they may be con- veniently grouped according to their mode of formation in additively coloured crystals. The F-Centre.-When a crystal is heated in the presence of alkali-metal vapour and then quenched rapidly a single narrow absorption band the F-band is found. The excess of alkali-metal ions is incorporated in the crystal and causes the formation of cation vacancies in which according to de Boer’s model of the F-centre the excess of electrons is trapped.2 This vacancy model which has been generally accepted for some time has now been established beyond reasonable doubt by the results obtained from magnetic resonance studies.In view of these results which are discussed briefly below there seems little point in considering alternative models for the F-centre such as those discussed by Seitz.2 Detailed wave functions all based on the vacancy model have been presented and recent experi- mental studies have been designed to give results which probe the suitability of these treatments. We will mention some of the most significant proper- ties of I;-centres and then give a brief resume of the basic features of the Rogener Ann. Phys. 1937 29 386. Heiland 2. Physik. 1950 127 530. SYMONS AND DOYLE COLOUR CENTRES IN ALKALI HALIDE CRYSTALS 65 mathematical treatments. For details the reader is referred to a com- prehensive and critical treatise by Gourary and Adrian,g and to a review by Dexter' in which the optical properties are discussed in detail.One of the most interesting features of the F-bands in alkali halides is that the wave- lengths of the band maxima are a simple function of the nearest-neighbour distance a. Ivey* has given an empirical equation for the wavelength A,,,. (A) = 703~zl*~~ which reproduces the experimental data within 7 %. It is noteworthy that the electron-affinity of the cations does not seem to be an important pal ameter and any theoretical treatment should reproduce this dependence upon the nearest-neighbour distance. Another important yardstick against which wave functions for the F-centre must be judged is the wealth of detailed information obtained from the application of electron-spin r e s o n a n ~ e ~ J ~ ~ ~ ~ J ~ and electron- nuclear double resonance techniques.13 When crystals containing F-centres are placed in an intense magnetic field the two-fold degeneracy of the electron spin is removed and transitions between these levels can be induced by suitable irradiation with microwave energy.The resulting absorption curve is generally broadened or split into components (described as hyperfine structure) by interactions with the magnetic moments of the surrounding nuclei. It is this hyperfine structure which provides the most useful information since one can deduce from the number of lines and their relative intensities the number of equivalent nuclei interacting and it is generally a simple matter to identify the nuclei involved. At the same time the separation of individual lines derived from a given nucleus is a function of the time the unpaired electron spends near that nucleus and hence a very detailed picture of the electron distribution can be built up.The results show that the unpaired-electron density is quite high on the six cations which define the vacancy. Somewhat surprisingly it is also appreciable on the first shell of anions although it is only just detectable on the second shell of cations. Early wave-functions for the F-centre were calculated on the assumption that the crystal could be treated as a polarisable continuum the ionic nature of the lattice and the characteristic periodic potential being largely ignored. This approximation can be correct only if both the ground and the first excited state are diffusely spread over many ions in the vicinity of the vacancy.In fact the simple models based on this approximation prove to be inconsistent since the resulting wave functions are far too compact for the continuum model to be a valid approximation and hence this simplification must be rejected.6 Indeed the results of electron-spin Gomary and Adrian Solid State Physics in the press. ' Dexter Solid State Physics 1958 6 353. * Ivey Phys. Rev. 1947 72 341. * Hutchinson ibid. 1949 75 1769. lo Schneider and England Physica 1951 17 221. l1 Kahn Kittel Levy and Portis Phys. Rev. 1953 91 1066. l2 Lord Phys. Rev. Letters 1958 I 170. l3 Feher Phys. Rev. 1957 105 1122. 3 66 QUARTERLY REVIEWS resonance show conclusively that the electron is closely confined to the vacancy and the first cation and anion shells; this volume is far too small to justify a continuum model.The other extreme would be to use an infinitely deep square-well model with a radius equal or close to the interionic distance. However a better approximation is to use a well of finite depth with a constant potential inside it and a coulombic potential outside it. This semi-continuum theory has been greatly refined by various workers and is discussed in detail by Courary and Adrian.6 These more refined calculations give good agreement with experiment when applied to the calculation of the energies and intensities of the F-bands and reasonable values for the hyperfine inter- actions can also be obtained provided one includes the reasonable exten- sion that when close to an ion the electron behaves like an s-electron in the outermost shell of that ion.The linear combination of atomic orbitals (LCAO) molecular-orbital approximation has also been used to describe the cavity model of F-centres and was particularly important in the development of the underlying theory of electron-spin resonance in F-centres.l1 However by far the most satisfactory and versatile approximation for the F-centre is the “point-ion- lattice” approximation of Gourary and Ad~-ian.~~l~ These authors have compared their treatment with the continuum and the molecular-orbital approximation in their review,6 and have extended it to cover other electron-excess centres. Agreement with experiment is good both for the energies of the F-bands and for the hyperfine splitting constants and other data derived from electron-spin resonance spectroscopy. The model uses the Hartree-Fock approximation in which the dipoles induced by the F- electron on the surrounding ions point to the centre of the vacancy rather than following the detailed motion of the electron.In view of the success of these calculations the following conclusions can be drawn with some confidence (a) the F-electrons are to be found largely within the vacancy and hence (b) polarisation of the surrounding ions is small since the trapped electron will largely cancel the effect of the missing anion; (c) displacement of the ions surrounding an F-centre from the normal lattice position is small; (4 the electronic transition which gives rise to the F-band may be described as p t s to a first approximation and even in the excited p-state the electron is closely confined to the vacancy.Although the vacancy model satisfactorily accommodated most of the properties of the F-centre until recently there was one notable exception. It can readily be shown that F-centres should luminesce on excitation with high quantum efficiency at low temperature.15JG However all evidence pointed to the contrary until in 1954 Botden van Doorn and Haven1’ l4 Gourary and Adrian Phys. Rev. 1957 105 11 80. l5 Huang and Rhys Proc. Roy. Soc. 1950,204 A 406. l6 Pekar J . Exp. Tlzeor. Phys. (U.S.S.R.) 1952 22 641. l7 Botden van Doorn and Haven Philips Res. Reports 1954 9 469. SYMONS AND DOYLE COLOUR CENTRES IN ALKALI HALIDE CRYSTALS 67 published results which showed that the predicted luminescence was readily detected if crystals were used which contained only I;-centres and that the quantum efficiency was indeed high under these circumstances.18 When there is appreciable aggregation the excitation energy is apparently dispersed by other processes involving these aggregate centres and it was for this reason that luminescence from F-centres proved so hard to detect.This aspect is discussed in more detail in the section on aggregate centres. F‘- a- and /I-Bands.-The F-band is always accompanied by a narrow band /I in the ultraviolet region close to the long-wavelength edge of the first fundamental band of the cry~ta1.l~ The band is therefore assigned to a transition involving the anions surrounding the F-centre. If the F-band is bleached by radiation of somewhat higher energy than the F-band maximum the /I-band is also bleached and a new band a is found on the long-wavelength side of the P-band.Simultaneously a broad band labelled F’ grows on the low-energy side of the narrow F-band. Since there is a decrease in paramagnetism as the F-band is bleached and since quantum-efficiency measurements show that in favourable circum- stances two F-centres are destroyed for every quantum absorbed,,* the effect of irradiation seems to involve the ejection of one F-electron from its anion vacancy trap and its capture by another F-centre. The I;’-centre is thus two electrons with paired spins trapped at a single anion vacancy. The a-band is then due to the excitation of electrons on those anions which surround an anion vacancy whilst the P-band is a similar transition when the vacancies contain F-electrons. These transitions may be similar in nature to that for anions in normal lattice sites or they may be caused by electron-transfer from an anion into the vacancy to give momentarily an F-centre (a-band) or an F‘-centre (/I-band).Aggregate Centres R M N.-When light is absorbed within theF-band at room temperature numerous new absorption bands may be formed on the long-wavelength side of the F-band;21 Seitz has suggested that the new bands are due to electrons trapped at aggregates of vacancies.1t2 The criterion of simplicity together with a consideration of possible formation mechanisms led Seitz to suggest specific models for these centres (Fig. 1). The R model is an electron trapped at a pair of adjacent anion vacancies the R2 model is two electrons trapped at a similar vacancy pair. Thus the models correspond to F2+- and F2-centres.That is if one compares the F-centre with a hydrogen atom then the R and the R model are analogues of the hydrogen molecule ion and the hydrogen molecule respectively. Seitz’s model for an M-centre is an electron trapped at a cluster of two anion vacancies and one cation vacancy. All of these centres have lower than cubic symmetry [Fig. 1 (d e andf)]. It should be emphasised that van Doorn Philbs Res. Reports 1958 13 296. l9 Delbecq Pringsheim and Yuster J. Chern. Phys. 1951 19 574; 1952,20 746. 2o Pick Ann. Physik. 1938 31 365. 21 Petroff 2. Physik. 1950 127 443. 68 QUARTERLY REVIEWS there is still no convincing evidence that they are correct. However most of the available evidence is consistent with these models. Much of the data relating to aggregate centres is still based upon plausible formation mechanisms.An example of such arguments may be mentioned. The aggregate centres are not produced simultaneously.21 The M-band appears first followed by R,. The rate of production of R,- centres is proportional to the concentration of M-centres. Cole and Friauf22 explain this genetic relation using Seitz’s models as due to capture of an electron by an M-centre followed by the ejection of a positive ion vacancy. This sort of reasoning can hardly be taken as establishing the models. More direct evidence on aggregate centres is obtained from symmetry studies. Ueta23 has shown that polarised light absorbed in the M-band leads to dichroism thus establishing that the M-centre has lower than cubic symmetry in qualitative agreement with Seitz’s model. Corroboration of this symmetry is obtained from measurements of the polarisation of luminescence excited by polarised 25 The problem is complicated by evidence presented below that centres which are close together but not adjacent can interact strongly.26 Some at least of the asymmetry detected by polarisation studies could be caused by such interaction between centres which in isolation might have cubic symmetry.Another experimental result which gives information about the symmetry of M-centres is the observation by Overhauser and RUchardt2’ that no linear Stark effect could be detected in the M-band. One can conclude from this negative result that M-centres should have inversion symmetry a conclusion which conflicts with Seitz’s model. Knox28 has suggested a modified model which whilst being in accord with all other information relating to M-centres also possesses inversion symmetry.This model can also be used to explain the otherwise puzzling absence of a dielectric loss peak.29 Knox assumes that this configuration shown in Fig. 1 (g) may actually be stabilised by lattice relaxation. Such a configuration may be thought of as a neutral alkali atom strongly dis- torted by the surrounding lattice. Since the composition of the M-centre in both Knox’s and Seitz’s models is the same no reappraisal of formation and bleaching experiments is necessary. It should be possible to distinguish easily between the two configurations by using the new electron- spin double resonance methods13 because of the difference in symmetry. Moreover one would expect a considerable hyperfine splitting due to the central nucleus for Knox’s model.Conventional electron-spin resonance 22 Cole and Friauf Phys. Rev. 1957 105 1464. 23 Ueta J. Phys. SOC. Japan 1952 7 107. 24 van Doom Philips Res. Reports 1957 12 309. 25 Lambe and Compton Php. Rev. 1957 106 684. Compton and Klick Phys. Rev. 1958 112 1620. 27 Overhauser and Ruchardt Phys. Rev. 1958 112 722. 28 Knox Phys. Rev. Letters 1959 2 87. 29 Jacobs J. Chem. Phys. 1957 27 218. SYMONS AND DOYLE COLOUR CENTRES IN ALKALI HALIDE CRYSTALS 69 has already been observed in M-~entres,~~ but the large inhomogeneous broadening of the lines has prevented any detailed conclusions regarding the nature of the centre. Seitz’s models assign R,- and R,-bands to two different centres consist- ing of one or two electrons trapped at two neighbouring anion vacancies.It has been observed however that the ratio of the intensity of the R,-band to that of the R2-band is constant over a range of concentrations. This has led Herman Wallis and Wallis31 to suggest that the two bands represent different transitions of the same centre. Again it should be possible to distinguish between these alternatives by using electron-spin resonance since the R1 model would give rise to paramagnetism whereas the R model would not. Crystals containing aggregate centres exhibit a number of interesting phenomena which may be termed interaction eflects. Of these we shall list only a few as being typical and particularly interesting. (a) The quantum efficiency of the conversion of F- into F‘-centres on irradiation decreases with repeated bleaching and thermal reconstitution of the F-band.32 (b) In a crystal containing both F- and M-centres irradiation with polarised “3’-band” light gives rise to dichroism of opposite sign in the M - b a n d ~ .~ ~ From the conditions of this experiment it appears that reorientation of M-centres is involved. (c) Irradiation in the F-band produces luminescence characteristic of R- and M-bands in crystals containing aggregate centres.28 (d) Irradiation at WOK in any of the bands produces a temporary change in the bands some being increased in intensity and others d e c r e a ~ e d . ~ ~ > ~ ~ All these effects may be explained by the existence of an efficient inter- action mechanism. Thus (a) would simply require the energy absorbed by the F-centre to be transmitted to an A-centre say which could undergo a radiationless transition to its ground state.Alternatively the energy could be utilised by an R-centre for dissociation. Similarly in case (b) the energy absorbed by the F-centre could be utilised by a neighbouring M-centre to reorient itself. The mechanism for preferential reorientation is obscure but may simply be that interaction with light is a minimum for one particular orientation which is thus favoured on irradiation. Case (c) involves energy exchange between centres followed by re-emission as light a phenomenon well known in the general field of luminescence. The high efficiency of the process however suggests a greater proximity than would be deduced from the concentration of the various centres. This may indicate a segregation of all of the centres in specific regions of the crystals pos- sibly along dislocation lines.Such a segregation is known to occur in the later stages of c~agulation.~~ Case (4 provides striking evidence for interaction for it was found that 30 Lord Phys. Rev. 1957 106 1100. 31 Herman Wallis and Wallis Phys. Rev. 1956 103 87. 32 Geiger Phys. Rev. 1955 99 1075. 33 van Doorn and Haven Phys. Rev. 1955 100 753. 34 Amelinckx Phil. Mag. 1956 47 269. 70 QUARTERLY REVIEWS irradiation in the infrared region (M-band) in sodium chloride produced changes in the absorption spectrum in the ultraviolet region. Since it is hardly likely that bands so widely spaced in wavelength could overlap appreciably this observation is interpreted as implying a particularly close proximity of centres. As Compton and Klick have pointed out,26 such strong interaction between centres may invalidate earlier conclusions regarding the symmetry of aggregate centres drawn from polarisation measurements.Interaction effects certainly complicate the analysis of the properties of aggregate centres. However they also offer a new intriguing field of study. The Colloid Band-The F-band obtained by rapidly quenching hot additively coloured crystals does not represent an equilibrium state. When the concentration of F-centres exceeds the solubility of the metal there is a tendency for the excess of metal to coagulate in the form of colloidal particles. The F-band is then replaced by a new band at longer wavelengths. This band which we shall describe as the “colloid band” is thought to be a property of the colloidal metal.All the known features of this band- position width shape integrated intensity and the fact that its properties are independent of temperature-are determined by the optical constants of the metal and the index of refraction of the host lattice. It can be shown that this result is in accord with the concept that the colloid band is caused by transitions of electrons in the colloidal metal the position of the band can be computed by using Mie’s theory35 for the absorption of light by metal ~ p h e r e s ~ ~ ~ ~ ’ and by using the free-electron theory for alkali-metals it can be shown3* that there is an intimate connection between the colloid band and the theory of collective modes of os~illation.~~ Thus the optical properties of colloid centres seem to be well understood. A considerable body of independent evidence is available to show that colloidal particles of reduced metal are present in crystals when the colloid band appears e.g.electron-diffraction rnea~urements,~~ X-ray ~cattering,~~ and nuclear magnetic resonance.42 This evidence will not be described here but a glance at the references enumerated would convince the reader that the identification of this band as a property of colloidal metal rests on firm ground. The mechanism of colloid formation is not yet well understood. In particular the action of light and the r61e of impurities deserve further study. It is known that concentrated arc light can cause coagulation of 35 Mie Ann. Phys. 1908 25 377. s6 Savostianova 2. Phys. 1930 64 262. 37 Scott Smith and Thompson J. Phys. Chem. 1953,57,757. 38 Doyle Phys.Rev. 1958 111 1067. 39 Pines Rev. Mod. Phys. 1956 28 184. 40 McLennan Cunad. J. Phys. 1951,29 122. 41 Smallman and Willis Phil. Mag. 1957 48 1018. 42 Ring O’Keefe and Bray Phys. Rev. Letters 1958 1 453. SYMONS AND DOYLE COLOUR CENTRES IN ALKALI HALIDE CRYSTALS 71 electron-excess but only if the specimen is allowed to become warm under the intense ill~mination.~~ If in high concentration F-centres will coagulate in the dark.45 A heterogeneous equilibrium analogous to a vapour-solution reaction exists between F-centres and colloid particles if light is excluded. The heat of the dark reaction was and found to be in agreement with the cohesive-energy data for bulk Although light is unnecessary for coagulation in densely coloured crystals Theisen and Scott4' found that the rate of coagulation was enormously enhanced by illumination.S h a t a l ~ v ~ ~ ? ~ ~ has examined the effect of light alone using small concentrations of F-centres so that thermal coagulation did not occur. Light within the F-band produced coagulation at 250°c even in the absence of thermal coagulation. The 3'-band was replaced by a band labelled X which was displaced by about 0.5 ev to the low-energy side of the 3'-band. This is probably the same band which Scott and B ~ p p ~ ~ called R'. Shatalov attributes it to the F2 molecular centre but its position and temperature- dependence seem to suggest that it is due to very small colloidal particles of reduced metal. The kinetics of formation and destruction of X-centres will bear closer examination since they are very early products in the coagulation process.They offer an interesting meeting ground between aggregate centres (R M etc.) which are produced optically at room temperature and the larger colloidal particles which are formed in heavily coloured crystals by thermal coagulation. The r61e of impurities in stimulating and suppressing colloid formation can throw light on the initial steps in coagulation. It is known thatsmall quantities of bivalent cations can greatly inhibit colloid formation,50 possibly by altering the equilibrium vacancy concentration^.^^ Conversely hydroxyl ions may well be necessary for the formation of colloids in irradiated crystals,52 but are not needed to promote coagulation in addi- tively coloured crystals. However hygroscopic salts such as sodium bromide are particularly hard to colour additively suggesting that hydroxyl ions may assist coagulation even in additively coloured crystals.Hydroxyl ions are present in synthetic crystals unless special pains are taken to exclude water vapour. There is every indication that this im- purity affects other properties besides the coagulation of colloids (cf. section 4).53954 43 Glaser Nachr. Akad. Wiss. Gottingen 1936 3 31 44 Scott and Bupp Phys. Rev. 1950 79 341. 45 Scott and Smith Phys. Rev. 1951 83 982. 46 Scott Phil. Mag. 1954 45 610. 47 Theisen and Scott J. Chem. Phys. 1952 20 529. 48 Shatalov Soviet Physics J.E.T.P. 1956 2 725. 49 Shatalov Optika i Spektrosk. 1957 3 610. Heiland and Kelting 2. Phys. 1949 126 689. 51 Watson and Scott J. Chem. Phys. 1958 30 342. aa Compton Phys. Rev. 1957 107 1271. 53 Rolfe Phys.Rev. Letters 1958 1 56. 54 Etzel and Patterson Phys. Rev. 1958 112 11 12. 72 QUARTERLY REVIEWS 3. Electron-deficit centres In 1954 Seitz2 made a strong plea for a systematic study of electron- deficit ( V ) centres which at that time were not well understood. Since then some light has been thrown upon the problem but much still awaits clarification. Seitzl9 offered an interpretation of the various V-bands in terms of centres which are direct antimorphs of the electron-excess centres dis- cussed in section (2). His V,-model consists of a hole combined with a cation vacancy symbolised as in Fig. 2 (a). By “hole” is meant the absence of a single electron from the outer shells of those halide ions directly surrounding this vacancy the unit is therefore centrosymmetric.Seitz’s Yz- V3- and V,-centres are also shown in Fig. 2. This approach failed to take note of the strong tendency of halogen atoms to form covalent bonds and it now seems clear that in general the V-bands are not so much properties of the imperfect solid as of isolated molecules or ions trapped within the solid. Certain bands in the ultraviolet region are always designated as Vl V, V, etc. for specific crystals but there is no certainty that the same centre is being described in different crystals. We therefore deal with each halide in turn. Iodides.-Since no magnetic studies of V-centres in alkali iodides have yet been reported only spectrophotometric results can be discussed. Additive coloration with iodine invariably gives rise to two intense bands at about 3650 and 2950 1$.55-57 These bands are found with about the same relative heights under different conditions and the separation between them is approximately equal to the energy difference between the 2P3/2 and the 2P112 state of the iodine atom.It therefore seems likely that these bands are caused by a single centre. Hershs7 has pointed out that tri-iodide ions in various solvents have two intense ultraviolet absorption bands with maxima close to the 3650 and 2950 A bands of additively coloured crystals and has suggested therefore that these bands are due to I,- ions incorporated in the crystals. This is in accord with chemical expectation. It is not clear however how the lattice changes to accom- modate these ions. It is curious that the low-energy band is more intense than that of higher energy in the crystal doublet whereas the reverse is true for I,- in solution.It is possible that other bands are present in the crystal spectrum which are not resolved under the conditions used or that distortion in the crystal considerably alters the relative intensities of the two transitions. Uchida and Nakai56 found a third band at about 2600 1$ in potassium iodide crystals. No chemical analogue is known for this band and further study is required before an assignment can be made. 55 Mollwo Ann. Phys. 1937 29 394. 54 Uchida and Nakai J. Phys. SOC. Japan 1954 9 928. 57 Hersh Phys. Rev. 1957 105 1410; J. Chern. Phys. 1957,27 1330. SYMONS AND DOYLE COLOUR CENTRES IN ALKALI HALIDE CRYSTALS 73 Molecular iodine has a fairly strong absorption band in the 5000 A region. Since the formation of tri-iodide ion is energetically favoured free iodine is not likely to be a stable species in the crystal but Hersh has reported two bands in this region in potassium iodide under unspecified condition^.^^ Bromides.-Potassium bromide coloured with bromine has been studied more extensively.An intense band in the 2700 A region generally pre- dominates. This is very close to the peak found for Br3- ions in various solvents and is probably due to these ions. M o l l ~ o ~ ~ found a weak band at about 4200 A this is close to the position expected for molecular bromine and the band shape is also similar to that found for bromine in solution. It therefore seems likely that both Br and Br3- are formed. Another band at 2300 A was found by Mollwo and also by Her~h:~' it is unlikely that this band is due to Br3- since Teegarden58 found an intense band at 2740 A but no band at 2300 A.Also the spectrum of Br3- ions in solution is a single broad line rather than a doublet. Teegarden however did find a shoulder at about 3200 A which was also found by D ~ r e n d o r f ~ ~ in crystals exposed to X-rays at -140". Since no simple molecular or ionic species is known with bands either at 3200 or at 2300A no clear identification can be made. A combined magnetic and optical study might prove helpful in this connection. Teega~den~~ found that potassium iodide crystals treated with bromine gave spectra which were indistinguishable from those treated with iodine under similar conditions. This can be understood in terms of the concepts outlined in this section since bromine readily oxidises iodide to iodine (3) Br + 21- + I + 2Br- However a similar result was found for potassium bromide treated with either bromine or iodine and the reverse reaction (3) is not energetically favourable.It is more probable that under these conditions Br,- and BrI respectively are formed. Since these ions have bands at 2700 and 2600A respectively it would be difficult to distinguish between fhem. It is possible however that the large excess of bromide ions is sufficient to upset equilibrium (3) and that iodide ions are formed from iodine. Teegarden rejected this possibility on the grounds that no band for I- was detected in the 2150 A region. However when iodide is incorporated in a potassium bromide lattice60 the first excited band is strongly shifted towards the higher-energy region being at 1930 A at 77°K.This is outside the range studied by Teegarden. Chlorides.-Unfortunately it has not been possible to form V-centres in alkali chlorides or fluorides by additive coloration. Therefore we have to 58 Teegarden J. Chem. Phys. 1956 24 161. 69 Dorendorf 2. Phys. 1951 129 317. 6o Delbecq Robinson and Yuster Phys. Rev. 1954 96 262. . . . . 4 74 QUARTERLY REVIEWS turn to the results obtained with high-energy radiation. At room tempera- ture two bands are found both for sodium chloride and for potassium chloride in the 2300 A region (Va and the 21 50 A region ( V3). By analogy with solution spectra it is probable that the V2-band is due to C&- but the assignment is not clear cut since the V2- and the V,-band lie close together. One might expect to find spectrophotometric evidence for the presence of “free” chlorine whose 117u+1C transition lies at about 3330 A.This is normally a weak transition and might well be too small for measurement but a weak band at 3400 A found by Hersh5’ for potassium chloride pellets could be due to free chlorine. When alkali chlorides or bromides are exposed to high-energy radiation at low temperatures and the spectra measured at the same temperature two new bands in the near ultraviolet region appear which are irreversibly lost on warming. Seitz2 has speculated on the nature of the centres respons- ible for these bands (which are symbolised as V and H) but there is still insufficient information concerning them to enable definite conclusions to be drawn. However electron-spin resonance studies have conclusively proved the presence of C1,- ions in potassium chloride crystals X-irradiated at low temperature.61 It was thought at first that the Y,-band was due to this radical ions1 and it has also been suggested that the resonance was caused by V,-~entres~~ [Fig.2 (c)]. In fact neither postulate is correct and none of Seitz’s models is suitable. The ions have an absorption band at 3650 A in potassium chloride which had not been detected previously because of its proximity to the V,-band (- 3500A).63 The electron-resonance results are of great importance since the wealth of detail is such that identification is unambiguous. These ions represent the simplest possible electron-deficit centre in which a chlorine atom formed by loss of an electron has combined with any one of its neighbour- ing chloride ions no lattice vacancies appear to be involved and the absorption band is close to that attributed to C1,- ions in aqueous s01ution.~ Thus the Vl-bands still remain enigmatic.There is no evidence against Seitz’s model but the inability to detect any electron-spin resonance which can be linked to this band is hard to understand in terms of this model. The band for potassium chloride is close to that for molecular chlorine but the intensity is too great and studies with polarised light suggest that the centre has cubic Cohen Kanzig and Woodruffs5 recently identified the radical ion F2- in irradiated lithium fluoride. They suggested that this unit was closely similar to Seitz’s Y,-model and it would be of great interest to make a combined magnetic and optical study similar to that of Delbecq et aL6 61 Kanzig Phys.Rev. 1955 99 1890. Bagguley and Owen Reports Progr. Phys. 1957 20 304. 63 Delbecq Smaller and Yuster Phys. Rev. 1958 111 1235. 64 Lambe and West Phys. Rev. 1957 108 634. 65 Cohen Kanzig and Woodruff Phys. Rev. 1958,107 1096. SYMONS AND DOYLE COLOUR CENTRES IN ALKALI HALIDE CRYSTALS 75 The particular items discussed in this section establish conclusively that the formation of specific molecules plays a fundamental r81e in the field of electron-deficit centres. Many outstanding problems remain and it is felt that these should be viewed in terms of interactions between vacancies and the molecular species whose formation now seem established. The results of Miessner'j6 and are of great significance in this respect their studies of mixed alkali halide crystals whilst still lacking a complete interpretation clearly show the difference between &centres and V- centres.4. Foreign ions A remarkably large number of cations and anions can be incorporated in alkali halides to give crystals which are ideal for spectrophotometric and magnetic studies. Of these we choose to discuss briefly hydride ions hydroxide ions alkaline-earth metal ions and first-row transition-metal ions of low valency. Hydride Ions.-Because it consists of only a proton and two electrons the hydride ion can adapt itself to fit any alkali halide lattice giving what is often described as a U-centre. Because of this flexibility the U-band which may be described as the first fundamental band of the hydride ion depends strongly upon the lattice constant just as does the F-band.Other bands associated with hydrogen are found after high-energy irradiation of crystals containing hydride. A combined magnetic and spectrophotometric study6* has greatly helped to elucidate the reactions involved and has established that hydride ions are photolysed with ultra- violet light to give F-centres and interstitial hydrogen atoms. A band (V,) found on the long-wavelength side of the U-band after irradiation is thought to be due to these interstitial hydrogen atoms.68 Hydroxide Ions.-These ions can replace halide ions in certain crystals and are of great importance because this may well happen unintentionally and hence phenomena caused by hydroxide ions may be ascribed to the pure c r y ~ t a l . ~ ~ ~ ~ On photolysis with ultraviolet light in the fundamental hydroxide band hydrogen atoms ( Uz) and F-centres are detected for potassium chloride but for potassium bromide a prominent V,-band is also found.Since the primary step probably gives H- and 0.- this difference can be understood in terms of the reaction (4) 0.- + Br- -+ 02- + Br* (V,) . . . Whilst this is favoured in a bromide lattice the greater electron affinity of chlorine could well prevent its occurrence with potassium chloride. Alkaline-earth Cations.-Any cation with an electron-affinity greater O6 Miessner 2. Phys. 1953 134 567. 67 Pick Z. Phys. 1953 134 604. 88 Delbecq Smaller and Yuster Phys. Rev. 1956 104 599. 76 QUARTERLY REVIEWS than that of the host cation is a potential electron-trap. This is certainly true for alkaline-earth cations and the resulting centres which are to a first approximation the corresponding singly charged cations (e.g.Ca+) give rise to characteristic optical spectra69 and electron-spin resonance ab~orption.~~ These centres are generally designated Z, irrespective of the particular cation involved. Transition-metal Cations.-Lithium fluoride has been used as host crystal for magnetic studies of ions such as Cr+ Mn2+ Fe+ Co2+ Co+ and Ni+ by Bleaney and ha ye^,^^^'^ the monopositive ions being formed by exposing crystals containing the corresponding bivalent ion to high-energy radiation. In many cases a measure of the covalent character of the bonds between impurity ions and the six neighbouring fluoride ions can be deduced because such interaction gives rise to a hyperfine structure in the electron-spin resonance spectra.Univalent ions (Cr+ Fe+ and Co+) are found to have cubic symmetry but bivalent ions (Co2+ and Mn2+) are in surroundings of less than cubic symmetry. This suggests that there is a cation vacancy close to the bivalent ions which migrates away during the irradiation. Extensive studies of this sort have also been made by Wertz and his co-workers using magnesium oxide as host crystal.73 5. Conclusions The most outstanding development in this field has been the application of electron-spin resonance methods including double resonance. When this is combined with optical studies a great deal of information can be obtained which besides establishing the basic characteristics of the centre being studied provides data which the theoretician can use as a yardstick for detailed mathematical descriptions.It is perhaps not always realised how many unusual compounds are studied in this field we have already mentioned many such as FZ2- C12- Ca+ and Fe+ and doubtless many more will soon be discovered. These highly reactive species have not only been detected but have often been subjected to more detailed structural analysis than is afforded to stable compounds. Those interested in solution chemistry can also draw from the results of these studies. This has been done for example in the field of solutions of alkali metals in ammonia and other solvents,74 and comparison with crystals may also be relevant when considering ionic s ~ l v a t i o n . ~ ~ G 9 Pick Ann. Phys. 1939 35 73. 70 Kawamura and Ishiwatari J. Phys. SOC. Japan. 1958 13 574. 71 Bleaney and Hayes Proc. Phys. SOC. 1957 70 B 626. 72 Hayes Discuss. Faraday SOC. 1958 26 58. i s Wertz Auzins Griffiths and Orton Discuss. Faraday SOC. 1958 26 66. 7 4 Symons Quart. Rev. 1959 13 99. Smith and Symons Trans. Faraday SOC. 1958 54 338 346.